Pharmacogenetics of Breast Cancer - JOHN J. HADDAD, Ph.D.

Translational Medicine Series 7
Pharmacogenetics
of Breast Cancer
Towards the
Individualization of Therapy
Edited by
Brian Leyland-Jones

TRANSLATIONAL MEDICINE SERIES
1. Prostate Cancer: Translational and Emerging Therapies, edited by
Nancy A. Dawson and W. Kevin Kelly
2. Breast Cancer: Translational Therapeutic Strategies, edited by Gary
Lyman and Harold Burstein
3. Lung Cancer: Translational and Emerging Therapies, edited by
Kishan J. Pandya, Julie R. Brahmer, and Manuel Hidalgo
4. Multiple Myeloma: Translational and Emerging Therapies, edited by
Kenneth C. Anderson and Irene Ghobrial
5. Cancer Supportive Care: Advances in Therapeutic Strategies,
edited by Gary H. Lyman and Jeffrey Crawford
6. Cancer Vaccines: Challenges and Opportunities in Translation,
edited by Adrian Bot and Mihail Obrocea
7. Pharmacogenetics of Breast Cancer: Towards the Individualization
of Therapy, edited by Brian Leyland-Jones

Informa Healthcare USA, Inc.
52 Vanderbilt Avenue
New York, NY 10017
# 2008 by Informa Healthcare USA, Inc.
Informa Healthcare is an Informa business
No claim to original U.S. Government works
Printed in the United States of America on acid-free paper
10987654321
International Standard Book Number-10: 1-4200-5293-4 (Hardcover)
International Standard Book Number-13: 978-1-4200-5293-0 (Hardcover)
This book contains information obtained from authentic and highly regarded sources. Reprinted material is
quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable
efforts have been made to publish reliable data and information, but the author and the publisher cannot
assume responsibility for the validity of all materials or for the consequence of their use.
No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic,
mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and
recording, or in any information storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright
.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC) 222 Rosewood
Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and
registration for a variety of users. For organizations that have been granted a photocopy license by the CCC,
a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used
only for identification and explanation without intent to infringe.
Library of Congress Cataloging-in-Publication Data
Pharmacogenetics of breast cancer : towards the individualization of therapy / edited by
Brian Leyland-Jones.
p. ; cm. — (Translational medicine series ; 7)
Includes bibliographical references and index.
ISBN-13: 978-1-4200-5293-0 (hardcover : alk. paper)
ISBN-10: 1-4200-5293-4 (hardcover : alk. paper)
1. Breast—Cancer—Genetic aspects. 2. Breast—Cancer—Chemotherapy.
3. Pharmacogenetics. I. Leyland-Jones, Brian. II. Series.
[DNLM: 1. Breast Neoplasms—genetics. 2. Breast Neoplasms—drug therapy.
WP 870 P536 2008]
RC280.B8P43 2008
616.99 0 449—dc22
2008001013
For Corporate Sales and Reprint Permissions call 212-520-2700 or write to: Sales Department,
52 Vanderbilt Avenue, 16th floor, New York, NY 10017.
Visit the Informa Web site at
www.informa.com
and the Informa Healthcare Web site at
www.informahealthcare.com

Preface
I write this preface on Christmas Day, 2007, in Costa Rica. Christmas is always a
special time for me, since my parents were married on Christmas Day sixty
nine years ago. It is almost unimaginable to think how far we have progressed
since 1938. My parents were married in the year before the Second World
War. The nitrogen mustards, from which arose the first chemotherapies, were
developed during the war. I was fortunate during my training at Sloan-Kettering
to work in the laboratories of Fred Phillips, Joe Burchenal, and Jack Fox, which
conducted some of the pivotal work on these compounds. Indeed, the first task
force for childhood leukemia was set up in Washington in the year of my birth.
However, we all have to remember how slowly treatments are disseminated to
the community; I can remember, as a child, my next-door neighbor having a leg
amputated for metastatic lung cancer since no other treatments were accessible
locally!
The rest is history, but again it is difficult to believe that the Nobel Prize
for the discovery of the cellular origin of retroviral oncogenes was awarded to
Bishop and Varmus only in 1989.
Now we face the challenge of too few patients eligible for clinical trials to
adequately test the myriad of targeted therapies available as well as the availability
of high-quality starting material with associated clinical and phenotypic
data. Various platforms are available for the identification of biomarkers: early
technologies have focused on reverse transcriptase polymerase chain reaction
and expression arrays. Now single nucleotide polymorphisms, array comparative
genomic hybridization, methylation signatures, microRNAs, and proteomics are
increasingly included in biomarker profiling; nanotechnology, molecular imaging,
and the use of circulating tumor cells are advancing at a tremendous pace.
By the time this book is released, the specimen collection guidelines for breast
cancer, representing a joint effort between the North American cooperative
iii

iv
Preface
groups and those comprising the Breast International Group (BIG), will have
been published and be available for open Web access. We have attempted to
cover the breadth of these diverse areas in this book.
I am especially proud of three aspects of the book. The first is the timeliness,
as so many new approaches to individualization of therapy have emerged, and
when two very large randomized trials are aggressively accruing on the basis of
Oncotype and mammoprint. Second, the diversity of the topics covered, ranging
from pathology, tissue handling, DNA-, expression-, and epigenetic-based
approaches, nanotechnology, stem cells, circulating tumor cells, endocrine, and
chemotherapeutic through to pharmacologic aspects and the key pivotal trials.
Third, I am proud that so many of my most distinguished friends and colleagues
from around the world have been so generous with their time and devotion to
make this book outstanding.
Finally, in writing this preface from one of the most beautiful places on
earth, I am deeply reminded of the disparities in access to healthcare. We are all
immensely grateful that my friends and colleagues John Seffrin and Otis
Brawley at the American Cancer Society have been leaders in bringing this
critical issue to the forefront of public consciousness, but we must never fail to
remember that severe disparities in access to care remain, not only in the United
States but globally. Individualized therapy is not only meant to direct therapy to
those who will benefit most but also intended to reduce waste and improve cure
rates dramatically, so that this therapy is available to all who suffer. Let us, the
privileged, never move our focus from the most disadvantaged in our world and
aim at using individualized therapy as a means to make our most sophisticated
treatment available and affordable to all.
Brian Leyland-Jones, M.D., Ph.D.

Acknowledgments
Many thanks are owed on a personal level, especially to all my friends and
coauthors who have been immensely generous in dedicating their nonexistent
spare moments to contributing the various chapters. I am deeply grateful to the
publisher, Informa Healthcare, who not only had the insight to partner with
me on this book but also had the patience to wait for the “nonexistent spare
moments” of the contributing authors to materialize. I must also profoundly
apologize to several colleagues who were approached early on to contribute
chapters but who were unable to because of the original time lines and now must
be frustrated because the time lines were forced to be extended considerably.
Most especially, I must thank my dear, dear colleague, Dr. Brian R. Smith, who
was relentless in organization, follow-up, and advice, a man whose wisdom I
rely upon at so many critical moments and to whose structured approach I aspire.
I am deeply, deeply grateful to Charles Bronfman, whose chair in the memory of
his dear sister, Minda, sustained me over the past 17 years and who now has
inspirationally established his own Institute of Personalized Cancer Therapy in
New York. I am grateful also to the Flanders family who have shown me a depth
of kindness, support, and friendship that is without equal. Finally, I am humbled
by my research and clinical colleagues who instruct and guide me constantly,
and especially by each and every patient who inspires us and teaches, in each and
every encounter, just how personalized breast cancer therapy really is.
v

Contents
Preface . . . . . . . . . . iii
Acknowledgments . . . . v
Contributors . . . . . . . . xi
1. Pharmacogenetics of Breast Cancer: Toward the Individualization
of Therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Jenny C. Chang, Susan G. Hilsenbeck, and Suzanne A. W. Fuqua
2. Pharmacology, Pharmacogenetics, and Pharmacoepidemiology:
Three P’s of Individualized Therapy . . . . . . . . . . . . . . . . . . . . 11
Shaheenah Dawood
3. Role of Genetic Variability in Breast Cancer Treatment
Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Kandace L. Amend, Ji-Yeob Choi, and Christine B. Ambrosone
4. The Role of Epigenetics in Breast Cancer: Implications for
Diagnosis, Prognosis, and Treatment . . . . . . . . . . . . . . . . . . . . 45
Amy M. Dworkin, Tim H.-M. Huang, and Amanda E. Toland
5. Comparative Genomic Hybridization and Copy Number
Abnormalities in Breast Cancer . . . . . . . . . . . . . . . . . . . . . . . . 61
Nicholas Wang and Joe Gray
vii

viii
Contents
6. Gene Expression Profiling as an Emerging Diagnostic Tool
to Personalize Chemotherapy Selection for Early Stage
Breast Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Cornelia Liedtke and Lajos Pusztai
7. Gene Expression–Based Predictors of Prognosis and Response
to Chemotherapy in Breast Cancer . . . . . . . . . . . . . . . . . . . . . 97
Soonmyung Paik
8. Utilization of Genomic Signatures for Personalized
Treatment of Breast Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . 107
P. Kelly Marcom, Carey K. Anders, Geoffrey S. Ginsburg,
Joseph R. Nevins, and Anil Potti
9. Personalized Medicine by the Use of Microarray Gene
Expression Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Stella Mook, Laura J. van ’t Veer, and Fatima Cardoso
10. Predictive Markers for Targeted Breast Cancer Treatment . . . 135
Hans Christian B. Pedersen and John M. S. Bartlett
11. Circulating Tumor Cells in Individualizing Breast Cancer
Therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
James M. Reuben and Massimo Cristofanilli
12. Individualization of Endocrine Therapy in Breast Cancer . . . . 167
Amelia B. Zelnak, Ruth M. O’Regan, and Clodia Osipo
13. TAILORx: Rationale for the Study Design . . . . . . . . . . . . . . . 185
Joseph A. Sparano
14. Taking Prognostic Signatures to the Clinic . . . . . . . . . . . . . . . 197
Michail Ignatiadis and Christos Sotiriou
15. Tissue is the Issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
Adekunle Raji
16. Acquisition and Preservation of Tissue for Microarray
Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Brian R. Untch and John A. Olson

Contents
ix
17. Tapping into the Formalin-Fixed Paraffin-Embedded (FFPE)
Tissue Gold Mine for Individualization of Breast
Cancer Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
Mark Abramovitz and Brian Leyland-Jones
18. Breast Cancer Stem Cells and Their Niche: Lethal Seeds in
Lethal Soil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
Danuta Balicki, Brian Leyland-Jones, and Max S. Wicha
19. Molecular Imaging in Individualized Cancer Management . . . 291
David M. Schuster and Diego R. Martin
20. Cancer Nanotechnology for Molecular Profiling and
Individualized Therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
May Dongmei Wang, Jonathan W. Simons, and Shuming Nie
Index . . . . . . . 323

Contributors
Mark Abramovitz
VM Institute of Research, Montreal, Québec, Canada
Christine B. Ambrosone Department of Cancer Prevention and Control,
Roswell Park Cancer Institute, Buffalo, New York, U.S.A.
Kandace L. Amend Department of Oncological Sciences, Mount Sinai School
of Medicine, New York, New York, U.S.A.
Carey K. Anders Duke Institute for Genome Sciences & Policy, Department
of Medicine, Duke University Medical Center, Durham, North Carolina, U.S.A.
Danuta Balicki Department of Medicine and Research Centre, Centre
Hospitalier de l’Université de Montréal; Department of Medicine, Université de
Montréal; and Breast Cancer Prevention and Genetics and Breast Cancer Risk
Assessment Clinic, Ville Marie Multidisciplinary Breast Cancer Centre,
Montréal, Québec, Canada
John M. S. Bartlett Endocrine Cancer Group, Cancer Research Centre,
Western General Hospital, Edinburgh, United Kingdom
Fatima Cardoso
Brussels, Belgium
Department of Medical Oncology, Institute Jules Bordet,
Jenny C. Chang Breast Center, Dan L. Duncan Cancer Center, Baylor College
of Medicine, Houston, Texas, U.S.A.
Ji-Yeob Choi Department of Cancer Prevention and Control, Roswell Park
Cancer Institute, Buffalo, New York, U.S.A.
xi

xii
Contributors
Massimo Cristofanilli Department of Breast Medical Oncology, The
University of Texas M. D. Anderson Cancer Center, Houston, Texas, U.S.A.
Shaheenah Dawood Department of Medical Oncology, Dubai Hospital,
Dubai, United Arab Emirates
Amy M. Dworkin Human Cancer Genetics Program, Department of
Molecular Virology, Immunology, and Medical Genetics, The Ohio State
University, Columbus, Ohio, U.S.A.
Suzanne A. W. Fuqua Breast Center, Dan L. Duncan Cancer Center, Baylor
College of Medicine, Houston, Texas, U.S.A.
Geoffrey S. Ginsburg Duke Institute for Genome Sciences & Policy,
Department of Medicine, Duke University Medical Center, Durham,
North Carolina, U.S.A.
Joe Gray Life Sciences Division, Lawrence Berkeley National Laboratory,
Berkeley, California, U.S.A.
Susan G. Hilsenbeck Breast Center, Dan L. Duncan Cancer Center, Baylor
College of Medicine, Houston, Texas, U.S.A.
Tim H.-M. Huang Human Cancer Genetics Program, Department of
Molecular Virology, Immunology, and Medical Genetics, The Ohio State
University, Columbus, Ohio, U.S.A.
Michail Ignatiadis Department of Medical Oncology, Translational Research
Unit, Institut Jules Bordet, Université Libre de Bruxelles, Brussels, Belgium
Brian Leyland-Jones Winship Cancer Institute, Emory University School of
Medicine, Atlanta, Georgia, U.S.A.
Cornelia Liedtke Department of Breast Medical Oncology, The University of
Texas, Anderson Cancer Center, Houston, Texas, U.S.A.
P. Kelly Marcom Duke Institute for Genome Sciences & Policy, Department
of Medicine, Duke University Medical Center, Durham, North Carolina, U.S.A.
Diego R. Martin
Georgia, U.S.A.
Department of Radiology, Emory University, Atlanta,

Contributors
xiii
Stella Mook Department of Pathology, The Netherlands Cancer Institute,
Amsterdam, The Netherlands
Joseph R. Nevins Duke Institute for Genome Sciences & Policy, Department
of Molecular Genetics and Microbiology, Duke University Medical Center,
Durham, North Carolina, U.S.A.
Shuming Nie Departments of Biomedical Engineering and Chemistry and
the Winship Cancer Institute, Emory University and Georgia Institute of
Technology, Atlanta, Georgia, U.S.A.
Ruth M. O’Regan
Emory Winship Cancer Institute, Atlanta, Georgia, U.S.A.
John A. Olson Institute for Genome Sciences and Policy and the Department
of Surgery, Duke University Medical Center, Durham, North Carolina, U.S.A.
Clodia Osipo
Loyola University, Maywood, Illinois, U.S.A.
Soonmyung Paik Division of Pathology, NSABP Foundation, Pittsburgh,
Pennsylvania, U.S.A.
Hans Christian B. Pedersen Endocrine Cancer Group, Cancer Research
Centre, Western General Hospital, Edinburgh, United Kingdom
Anil Potti Duke Institute for Genome Sciences & Policy, Department of
Medicine, Duke University Medical Center, Durham, North Carolina, U.S.A.
Lajos Pusztai Department of Breast Medical Oncology, The University of
Texas, Anderson Cancer Center, Houston, Texas, U.S.A.
Adekunle Raji Eastern Cooperative Oncology Group Pathology Coordinating
Office, Northwestern University, Evanston, Illinois, U.S.A.
James M. Reuben Department of Hematopathology, The University of Texas
M. D. Anderson Cancer Center, Houston, Texas, U.S.A.
David M. Schuster Division of Nuclear Medicine and Molecular Imaging,
Department of Radiology, Emory University, Atlanta, Georgia, U.S.A.
Jonathan W. Simons
California, U.S.A.
The Prostate Cancer Foundation, Santa Monica,

xiv
Contributors
Christos Sotiriou Department of Medical Oncology, Translational Research
Unit, Institut Jules Bordet, Université Libre de Bruxelles, Brussels, Belgium
Joseph A. Sparano Albert Einstein College of Medicine, Montefiore Medical
Center, Bronx, New York, U.S.A.
Amanda E. Toland Human Cancer Genetics Program, Department of
Molecular Virology, Immunology, and Medical Genetics, The Ohio State
University, Columbus, Ohio, U.S.A.
Brian R. Untch Institute for Genome Sciences and Policy and the Department
of Surgery, Duke University Medical Center, Durham, North Carolina, U.S.A.
Laura J. van ’t Veer Department of Pathology, The Netherlands Cancer
Institute, Amsterdam, The Netherlands
May Dongmei Wang Departments of Biomedical Engineering and Electrical
and Computer Engineering, Georgia Institute of Technology and Emory
University, Atlanta, Georgia, U.S.A.
Nicholas Wang Life Sciences Division, Lawrence Berkeley National
Laboratory, Berkeley, California, U.S.A.
Max S. Wicha Comprehensive Cancer Centre, Department of Internal
Medicine, Division of Hematology-Oncology, University of Michigan, Ann
Arbor, Michigan, U.S.A.
Amelia B. Zelnak
Emory Winship Cancer Institute, Atlanta, Georgia, U.S.A.

1
Pharmacogenetics of Breast Cancer:
Toward the Individualization of Therapy
Jenny C. Chang, Susan G. Hilsenbeck, and Suzanne A. W. Fuqua
Breast Center, Dan L. Duncan Cancer Center,
Baylor College of Medicine, Houston, Texas, U.S.A.
INTRODUCTION
Mortality from breast cancer results from the ability of some tumors to metastasize
to distant sites. Selecting patients with micrometastases at diagnosis is
crucial for clinicians in deciding who should, and who should not, receive toxic
and expensive adjuvant chemotherapy to eradicate these metastatic cells.
Although many individual biomarkers were originally attractive, over the years
most have failed to become clinically useful. In addition, the management of
breast cancer has changed, with the majority of node-negative patients now
undergoing systemic adjuvant therapy because we cannot precisely determine an
individual’s risk of recurrence. A majority of node-negative patients are being
unnecessarily overtreated because if left systemically untreated, only about 25%
of node-negative patients would ever develop recurrence. There is therefore a
critical need to identify patients with sufficiently low risk of breast cancer
recurrence so as to avoid further treatment. In addition, in patients at risk of
recurrence and in need of therapy, optimal therapeutic selection is an increasingly
important objective. Recent developments in applying microarray technologies
to breast tumor samples suggest that these new techniques may provide
for the transition of molecular biological discoveries to clinical application and
1

2 Chang et al.
will generate clinically useful genomic profiles that more accurately predict the
long-term outcome of individual breast cancer patients.
BACKGROUND
Until recently, evaluations of prognostic and predictive factors have considered
one factor at a time or have used small panels of markers. However, with the
advent of new genomic technologies such as microarrays, capable of simultaneously
measuring thousands of genes or gene products, we are beginning to
construct molecular fingerprints of individual tumors so that accurate prognostic
and predictive assessments of each cancer may be made. Clinicians may one day
base clinical management on each woman’s personal prognosis and predict the
best individual therapies according to the genetic fingerprint of each individual
cancer.
Breast cancer is characterized by a very heterogeneous clinical course. A
major goal of recent studies is to determine whether RNA microarray expression
profiling or DNA array gene amplification or gene loss patterns can accurately
predict an individual’s long-term potential for recurrence of breast cancer, so that
appropriate treatment decisions can be made. Microarrays can be used to measure
the mRNA expression of thousands of genes at one time or survey genomic
alterations that may distinguish molecular phenotypes associated with long-term,
recurrence-free survival or clinical response to treatment. These new technologies
have been successfully applied to primary breast cancers and may
eventually outperform currently used clinical parameters in predicting disease
outcome.
Since RNA expression microarray technology provided a method for
monitoring the RNA expression of many thousands of human genes at a time,
there was considerable anticipation that it would quickly and easily revolutionize
our approaches to cancer diagnosis, prognosis, and treatment. The reality
remains extremely promising but is also complex. A potential complication in
the application of microarray technology to primary human breast tumor samples
is the presence of variable numbers of normal cells, such as stroma, blood
vessels, and lymphocytes, in the tumor. Indeed, it has been demonstrated using
gross analysis of human breast cancer specimens compared with breast cancer
cell lines that the tumors expressed sets of genes in common not only with these
cell lines but also with cells of hematopoietic lineage and stromal origin (1,2).
Laser capture microdissection has also been successfully used to isolate pure-cell
populations from primary breast cancers for array profiling (3). Sgroi et al. (3)
utilized laser capture microdissection to isolate morphologically “normal” breast
epithelial cells, invasive breast cancer cells, and metastatic lymph node cancer
cells from one patient and were able to demonstrate the feasibility of using
microdissected samples for array profiling as well as following potential progression
of cancer in this patient. However, with the emerging data supporting
important roles for the surrounding stroma in breast cancer progression and the

Pharmacogenetics of Breast Cancer 3
labor-intensive and technically challenging nature of laser capture technology
with subsequent amplification of RNA for quantitation, most published investigations
to date have evaluated total gene expression to identify prognostic
profiles, as will be described in the next section.
MOLECULAR CLASSIFICATION OF BREAST CANCER
A study of sporadic breast tumor samples by Perou et al. (2) was the first to show
that breast tumors could be classified into subtypes distinguished by differences
in their expression profiles. Using 40 breast tumors and 20 matched pairs
of samples before and after doxorubicin treatment, an “intrinsic gene set” of
476 genes was selected that was more variably expressed between the 40 sporadic
tumors than between the paired samples. This intrinsic gene set was then used to
cluster and segregate the tumors into four major subgroups: a “luminal cell-like”
group expressing the estrogen receptor (ER); a “basal cell-like” group expressing
keratins 5 and 17, integrinb4, and laminin, but lacking ER expression; an
“Erb-B2-positive” group; and a “normal” epithelial group (Fig. 1).
In a subsequent study with 38 additional cancers, the investigators found
the same subgroups as before (4), except that the luminal, ER-positive group was
further subdivided into subsets with distinctive gene expression profiles. In
univariate survival analysis, performed on the 49 patients diagnosed with locally
advanced disease but without evidence of distant metastasis, ER positivity was
not a significant prognostic factor on its own, but the luminal-type group enjoyed
a more favorable survival compared with the other groups. Conversely, the
basal-like group had a significantly poorer prognosis. Although small and
exploratory, this study suggests that important differences in outcome can be
ascertained from microarray expression profiling.
An interesting study was reported by Gruvberger et al. (5), who profiled 58
grossly dissected primary invasive breast tumors and used artificial neural network
analysis to predict the ER status of the tumors on the basis of their gene expression
patterns. They then determined which specific genes were the most important for ER
classification. By comparing to Serial Analysis of Gene Expression (SAGE) data
from estradiol-stimulated breast cancer cells, they determined that only a few genes
of the many genes that were associated with ER expression in tumors were indeed
estrogen responsive in cell culture. This observation lent further support to the
hypothesis developed by Perou et al. (1) that basic cell lineages, such as the luminal
ER-positive cell type, can be partly explained by observed genomic gene expression
patterns rather than downstream effectors of only one pathway, such as the ER.
PROGNOSTIC IMPLICATIONS
Microarrays have also been used to predict lymph node status and very shortterm
relapse-free survival in two groups (n ¼ 37 and 52, respectively) of heterogeneously
treated patients (6). Although prediction of nodal status is of

4 Chang et al.
Figure 1 Supervised classification of prognosis signatures (van ’t Veer). The 78 tumors
are listed vertically, and the 70 “prognostic” genes horizontally. (A) Study design. (B)Heatmap
of prognostic signature.
limited interest clinically, the study uses innovative statistical methods, rigorously
generates estimates of future classifier performance, and further demonstrates
the feasibility of accurate prediction of tumor biology using expression
arrays. In a more focused and somewhat more clinically relevant study, van ’t Veer
et al. (7) used RNA expression microarray analyses to identify a 70-gene
prognostic gene signature (“classifier”) in young, untreated, axillary lymph
node–negative patients using a training set of 44 good (disease free for more than
5 years) and 34 poor (distant relapse in less than 5 years) outcome tumors and
then tested the classifier in a validation set of 19 tumors. The same group (8) has
now extended the study to a total of 295 young (

Pharmacogenetics of Breast Cancer 5
70-gene classifier (7). The microarray-based predictions are consistent with, and
perhaps better than, estimates that can be obtained with current prognostic
indices.
GENETIC SUSCEPTIBILITY
A few studies have utilized new genomic approaches for the study of inherited
breast cancer (8). There is accumulating evidence, both epidemiological and
histological, that tumors arising as a result of mutations in the two breast cancer
susceptibility gene families (BRCA1 and BRCA2) are biologically distinct. For
instance, BRCA1 breast cancers are most often ER and PR negative, but BRCA2
cancers more often tend to be positive for these receptors (9). In a seminal paper
published by Hedenfalk et al. (10), seven tumors each from BRCA1 and BRCA2
gene mutation carriers or sporadic breast cancers were compared by expression
microarray analysis. They found that the gene expression profiles of the three
tumor groups differed significantly from each other, underscoring the fundamental
differences between BRCA1 and BRCA2 mutation-associated tumors. Of
course a potential confounding issue was the differential distribution of ER
between the BRCA1 and BRCA2 tumors. However, even after removal of ERor
PR-associated genes from the analysis, the two inherited tumor groups were
still discernable. Thus, ER status alone does not fully explain the observed
differences in gene expression profiles. Although this study is obviously very
small, and other confounding issues such as tumor stage, grade, and treatment
cannot be considered, it does set a foundation for larger validation studies to
confirm differential genes, which could then provide important clues to the
etiology of inheritable breast cancer.
Microarrays are also being studied as a way to predict response to systemic
therapy. The neoadjuvant setting is especially attractive for these studies for
several reasons, including early assessment of response to therapy, biopsiable
access to the primary tumor, and considerable reduced sample sizes compared to
those required in the adjuvant setting.
PREDICTIVE IMPLICATIONS
Methods for assessing response in neoadjuvant trials remain problematic. Clinical
response to neoadjuvant chemotherapy is a validated surrogate marker for
improved survival (11,12). Women who achieve pathologic complete response
are most likely to have the best clinical outcome, although survival is still
improved in those who clinically respond but do not achieve pathologic complete
response.
In an early study, Buchholz et al. (13) obtained sufficient RNA from core
biopsies of five patients to perform serial microarray expression profiles and
showed that despite differences in therapy, patients with good pathological

6 Chang et al.
responses to neoadjuvant treatment appeared to have gene profiles that clustered
distinctly than those of patients who were poor responders to treatment. More
recently, Chang et al. (14) have shown that gene profiling can be used to
accurately predict response to neoadjuvant docetaxel. The study enrolled
24 subjects, extracted sufficient RNA from all core needle biopsies, and constructed
a 92-gene predictor of response. In a complete cross-validation analysis, which
gives an unbiased estimate of performance on future samples, the classifier
correctly identified 10 of 11 responders and 11 of 13 nonresponders with an
overall accuracy of 88%. In a small validation set, this 92-gene classifier successfully
predicted response in six patients (Fig. 2). This compares very favorably
with the best existing predictive factors for response to specific therapy and
strongly suggests that after appropriately extensive validation, microarray
profiling will be useful for treatment selection. A second neoadjuvant study was
recently published that used cDNA arrays to develop predictors for paclitaxel,
fluorouracil, doxorubicin, and cyclophosphamide, involving 24 samples. A
classifier with 74 markers was developed, with 78% accuracy, suggesting that
transcriptional profiling has the potential to identify a gene expression pattern in
breast cancer that may lead to clinically useful predictors of chemotherapy
response (15).
In vitro drug sensitivity data in NCI60 cell lines sensitive and resistant to
specific chemotherapeutic agents (adriamycin, cyclophosphamide, docetaxel,
etoposide, 5-fluoruracil, paclitaxel, topotecan) were recently published (16).
Gene set enrichment analysis was performed, and gene expression signatures
predictive of sensitivity to individual chemotherapeutic drugs were reported.
Each signature was validated with response data from an independent set of cell
line studies. These signatures were then used to predict clinical response in
individuals treated with these drugs. Notably, signatures developed to predict
response to individual agents, when combined, were found to also predict
response to multidrug regimens (16). These profiles are being validated in
prospective clinical trials.
However, we acknowledge that studies to construct and validate arraybased
prognostic and predictive “markers” are complex. These studies must
address all of the concerns associated with ordinary, single-gene markers as well
as a number of considerations unique to array studies. Recommendations for
development of array-based prognostic classifiers have recently been enunciated
by Simon et al. (17). Among the most important points, they recommend that
studies should include (1) adequately large sample sizes in both training and
validation sets, (2) complete iteration of the entire classifier construction process
in estimating cross-validated prediction rates, (3) head-to-head comparison of
alternative classifiers on the same dataset, and (4) the full diversity of cases in
any validation set. In addition, gene expression patterns may be confounded by
several other factors, including ovarian ablation in premenopausal, ER-positive
patients and a different mechanism of action of combination chemotherapies.

Pharmacogenetics of Breast Cancer 7
Figure 2 Hierarchical clustering of genes correlated with docetaxel response. Sensitive
tumors (S) are defined as 25% residual disease, and resistant tumors (R) are defined as
greater than 25% residual disease.
Groups of patients with different characteristics, such as menopausal and ER
status or human epidermal growth factor receptor 2 (HER2) overexpression, may
be necessary to definitively determine classifying patterns in these subsets of
patients.

8 Chang et al.
CONCLUSIONS
The goal of comprehensive, genomewide approaches is to identify clinically
useful genetic profiles that will accurately predict the outcome to therapy and
prognosis of patients with breast cancer. Despite improvements in technology,
complex mechanisms driving breast cancer evolution continue to present challenges
for the use of genomic approaches to better understand breast and other
cancers. This, combined with our use of different markers, methods, tumors (e.g.,
differing ER and HER2), and measurements of clinical outcomes, impedes the
development of a consensus regarding predictive and prognostic markers for
breast cancer.
As this field matures, genomic studies examining identical breast tumor
sets with multiple complementary technologies (e.g., loss of heterozygosity,
comparative genomic hybridization, gene expression array analysis) will prove
essential to unraveling the genetic heterogeneity characteristic of this disease. A
combined genomic approach is necessary to define the underlying heterogeneous
complexity characteristic of breast cancer. These data should lead to the identification
and characterization of breast cancer subtypes, define the malignant
potential of a given lesion, and predict its sensitivity to specific therapies. These
multidisciplinary approaches should contribute to a better biological understanding,
and therefore improved clinical management, of breast cancer.
ACKNOWLEDGMENTS
This work was supported by funding provided by National Institutes of Health
(NIH) grants P50 CA58183 and P01 CA30195, the Breast Cancer Research
Foundation, the Emma Jacobs Clinical Breast Cancer Fund, and the U.S. Army
Medical Research and Materiel Command DAMD17-01-0132 and W81XWH-
04-1-0468.
REFERENCES
1. Perou CM, Jeffrey SS, van de Run M, et al. Distinctive gene expression patterns in
human mammary epithelial cells and breast cancers. Proc Natl Acad Sci USA 1999;
96:9212–9217.
2. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours.
Nature 2000; 406:747–752.
3. Sgroi DC, Teng S, Robinson G, et al. In vivo gene expression profile analysis of
human breast cancer progression. Cancer Res 1999; 59:5656–5661.
4. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas
distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci USA 2001;
98:10869–10874.
5. Gruvberger S, Ringner M, Chen Y, et al. Estrogen receptor status in breast cancer is
associated with remarkably distinct gene expression patterns. Cancer Res 2001; 61:
5979–5984.

Pharmacogenetics of Breast Cancer 9
6. Huang E, Cheng SH, Dressman H, et al. Gene expression predictors of breast cancer
outcomes. Lancet 2003; 361:1590–1596.
7. van ’t Veer LJ, Dai H, van De Vijver MJ, et al. Gene expression profiling predicts
clinical outcome of breast cancer. Nature 2002; 415:530–536.
8. van de Vijver MJ, He YD, van ’t Veer LJ, et al. A gene-expression signature as a
predictor of survival in breast cancer. N Engl J Med 2002; 347:1999–2009.
9. Verhoog LG, Brekelmans CTM, Seynaeve C, et al. Survival in hereditary breast cancer
associated with germline mutations of BRCA2. J Clin Oncol 1999; 17:3396–3402.
10. Hedenfalk I, Duggan D, Chen Y, et al. Gene-expression profiles in hereditary breast
cancer. N Engl J Med 2001; 344:539–548.
11. Fisher B, Bryant J, Wolmark N, et al. Effect of preoperative chemotherapy on the
outcome of women with operable breast cancer. J Clin Oncol 1998; 16:2672–2685.
12. Chang J, Powles TJ, Allred DC, et al. Biologic markers as predictors of clinical
outcome from systemic therapy for primary operable breast cancer. J Clin Oncol
1999; 17:3058–3063.
13. Buchholz TA, Stivers DN, Stec J, et al. Global gene expression changes during
neoadjuvant chemotherapy for human breast cancer. Cancer J 2002; 8:461–468.
14. Chang JC, Wooten EC, Tsimelzon A, et al. Gene expression profiling predicts
therapeutic response to docetaxel (Taxotere TM ) in breast cancer patients. Lancet
2003; 362:280–287.
15. Ayers M, Symmans WF, Stec J, et al. Gene expression profiles predict complete
pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and
cyclophosphamide chemotherapy in breast cancer. J Clin Oncol 2004; 22:2284–2293.
16. Potti A, Dressman HK, Bild A, et al. Genomic signatures to guide the use of chemotherapeutics.
Nat Med 2006; 12:1294–1300.
17. Simon R, Radmacher MD, Dobbin K, et al. J Natl Cancer Inst 2003; 95:14–18.

2
Pharmacology, Pharmacogenetics,
and Pharmacoepidemiology: Three
P’s of Individualized Therapy
Shaheenah Dawood
Department of Medical Oncology, Dubai Hospital,
Dubai, United Arab Emirates
If it were not for the great variability among individuals medicine might as
well be a science and not an art.
Sir William Osler, 1892
INTRODUCTION
Over the last decade significant advances have been made in the management of
breast cancer. This is exemplified by the development of several new agents and
significantly improved survival rates. However, despite our best efforts breast
cancer remains the second most common cause of cancer-related death in the
United States with approximately 40,910 women expected to die of this disease
in 2007 (1). Unfortunately, such trends are not restricted to only breast cancer but
have been observed across a range of solid tumors. A large portion of the
underlying problem is attributed to the fact that the choice of a chemotherapeutic
agent is often empirical based on guidelines that assumes all patients with a
particular malignancy to be homogeneous. The unfortunate outcome of this
11

12 Dawood
erroneous belief is that approximately 40% of patients may be receiving the
wrong drug (2).
Numerous studies have consistently observed that the efficacy and toxicity
of chemotherapeutic agents are heterogeneous rather than homogeneous across
populations. Despite a multitiered approach to determining the efficacy and safety
profiles of chemotherapeutic agents, spanning from in vitro assays, animal studies
through to large human population studies, the current practice of administering
the same dose of a given anticancer drug to a population of patients results in
considerable variation in clinical outcome. These effects cover a spectrum of
possibilities ranging from success to failure when considering treatment outcome
as an endpoint, and from no effect to a lethal event when treatment-associated
toxicity is the variable under study (3). Complicating the problem further is the
fact that currently available anticancer drugs show limited efficacy in up to 70% of
patients with adverse drug reactions accounting for as many as 100,000 patient
deaths and two million hospitalizations in the United States every year (4).
The key is to be able to, at the individual level, prospectively identify
patients at risk for severe toxicity and those likely to attain maximum benefit
from a given chemotherapeutic agent. Understanding that an individual does not
live in isolation but is in fact a product of the gene-environment interaction is
also vital to our understanding of the concept of individualized therapy (Fig. 1).
Figure 1
Sources of pharmacologic and pharmacogenetic variability.

Pharmacology, Pharmacogenetics, and Pharmacoepidemiology 13
Designing treatment strategies based on individual efficacy and toxicity predictions
will shift clinical practice from simply an “art” to a “science” by
incorporating a combination of physiological variables, genetic characteristics,
and environmental factors (all known to alter the dose-concentration time profile)
as well as population epidemiological elements resulting in a practice
paradigm of “personalized medicine.”
THE COMPONENTS OF INDIVIDUALIZED THERAPY: THE THREE P’s
When we administer a drug to a patient, our goal is to deliver a dose that will
achieve maximum benefit with little, if any, associated toxicity. Seldom is this
goal achieved. Individual differences in tumor response and normal tissue toxicity
have been consistently observed with most anticancer agents. These differences
are due to a variety of clinical variables (e.g., gender, age, diet, and
drug-drug interactions) as well as variations in drug metabolism and transport.
It is believed that 30% of patients treated with a particular drug will exhibit
beneficial effects from the drug, 30% will not attain any form of beneficial
effect, 10% will experience only side effects from the drug, and 30% are noncompliant
with the drug often as a result of drug-related side effects or perceived
insufficient effect (5). This results in a waste of allocated resources with large
proportions of patients being unnecessarily exposed to adverse side effects. Here
we will describe three components that aim to explain the observed differences
in responses to individual drugs: pharmacology, pharmacoepidemiology, and
pharmacogenetics.
Current dosing of anticancer drugs is based on either a fixed quantity of
the drug or on a dose that is normalized to the individual body surface area.
This method makes the assumption that within a group of individuals there
will be a uniform degree of systemic exposure to the drug. With studies
reporting 2- to 10-fold variations in drug clearance, this assumption is clearly
not valid (6). With drugs used in the field of oncology having narrow therapeutic
indices, it thus becomes imperative that we understand the mechanisms
behind the observed variability in drug response and toxicity when treating a
patient with cancer. At the clinical level, variability in drug response can be
explained by its pharmacology that describes the pharmacokinetic and pharmacodynamic
profiles of the drug. Pharmacokinetics explain “what the body
does to the drug” by describing the relationship between time and plasma
concentrations of the drug metabolites. Thisrelationshipisaffectedbyvariables
such as absorption, distribution, metabolism, and excretion of the drug.
The underlying objective of pharmacodynamics is to describe “what the drug
does to the body” by correlating drug concentration to drug effect (both beneficial
and adverse effects). Both the pharmacokinetic and pharmacodynamic
components of a drug are not independent, but represent a spectrum of continuous
events starting with the ingestion of the drug and culminating in the
observed clinical effect.

14 Dawood
Variability in any of the pharmacokinetic components will affect the
observed drug effect. This is typically observed in patients with organ dysfunction
where the inability to either metabolize or excrete the drug will lead to
exaggerated drug effects. For example, the use of carboplatin in patients with
renal dysfunction exaggerates the thrombocytopenic effect of this drug. Dosing
formulas using creatinine clearance are used to avoid this problem while maximizing
dose intensity (7). Another example is the use of taxanes, which are
primarily metabolized and excreted by the liver. When administered to patients
with hepatic dysfunction, prolonged bouts of severe neutropenia are observed.
These examples illustrate how clinical variables can affect the pharmacokinetic
profile of a drug leading to disturbances in effect. Good clinical assessment of
the patients is therefore the first and most important step in designing an individual
treatment regimen that will maximize drug benefit.
Pharmacoepidemiology is a sub discipline of epidemiology that focuses on
understanding why individuals respond differently to drug therapy. It is defined
“as the study of the distribution and determinants of drug-related events in
populations and the application of this study to efficacious drug treatments” (8).
These drug-related events pertain to both beneficial and adverse events. Pharmacoepidemology,
through the use of observational methods, aims to quantify
drug exposure and to describe, as well as predict, the effectiveness and safety of
the drug within defined populations at specific places and time. Factors
important in interpreting the variability in drug outcome include a patient’s
health profile, sex, age, diet, associated comorbidities, disease severity and
prognosis, drug compliance, drug prescribing and dispensing quality, and genetic
profile of the patient and tumor (9,10). Depending on drug effect there are four
main groups of patients: (i) responders, (ii) non-responders, (iii) toxic responders,
and (iv) toxic non-responders (Fig. 2).
Pharmacogenetics, a third and important component of individualized
therapy, focuses on describing the extent to which an individual’s genetic
makeup is responsible for the observed differences in drug efficacy and toxicity
profiles (11,12). This information is then used to make predictions about the
safety, toxicity, and efficacy of drugs in individual patients. Inherited variability
of drug targets, drug metabolizing enzymes, and drug transporters may all have a
major impact on overall drug response, disposition, and associated adverse side
effects by altering the pharmacokinetic and pharmacodynamic properties of
drugs (Table 1).
Genetic variations include nucleotide repeats, insertions, deletions, and
single nucleotide polymorphisms (SNPs). SNPs are the simplest and most
commonly studied DNA polymorphism that occur once every 1900 base pairs in
the 3 billion bases in the human genome (13) and account for more than 90% of
the genetic variation observed. More than 1.4 million SNPs have been identified
in the human genome. Of these more than 60,000 SNPs occur in the coding
region of genes with some SNPs being associated with variations in drug
metabolism and effects (13). When polymorphisms occur in the promoter region

Pharmacology, Pharmacogenetics, and Pharmacoepidemiology 15
Figure 2 A group of individuals with the same diagnosis do not respond homogeneously
to a single drug. Through the elements of phamacology, pharmacoepidemiology, and
pharmacogenetics, the aim is to identify patients at risk for toxicity or reduced response to
therapy prior to medication initiation.
Table 1
Examples of Genetic Variations and Their Associated Phenotypic Changes
Enzymes Substrate Phenotype
CYP3A4/3A5 D, P, E, T Interindividual variability in
pharmacokinetics
DPD 5-FU Increased toxicity
P-glycoprotein Various anticancer agents Drug resistance
TS 5-FU and antifolates Drug resistance
TPMT 6-MP Increased hematological toxicity
UGT1A1 Irinotecan Increased diarrhea and neutropenia
Abbreviations: D, docetaxel, P, paclitaxel; E, etoposide, T, tamoxifen; 6-MP, mercaptopurine, TPMT,
thiopurine methyltransferases, UGT1A1, UDP-glucuronosyltrasferase 1A1, DPD, dihydropyrimidine
dehydrogenase.
it can effect the transcription of the gene. When the polymorphism occurs in the
exon (and intron) region or 3 0 UTR of the gene, it can affect translation and RNA
stability, respectively, resulting in either reduced or enhanced activity of the
encoded protein (3).
Two approaches have been used to identify clinically relevant polymorphisms.
In the reverse genetics approach, a drug is first given to a group
individuals and the variability in response to and disposition of the drug is
observed. Phenotypes of clinical importance are prioritized and then experiments

16 Dawood
are conducted to determine if a genetic component is present to explain the
observed variability. Finally, the biochemical and molecular nature of the variability
is determined through extensive resequencing techniques that will link the
polymorphisms with the observed phenotypes. The second and more cost effective
method is the forward genetics approach. This method involves first determining
the presence of genetic variability at specific loci in a group of individuals. A
hypothesis is then formulated regarding these polymorphisms causing phenotypic
variability. Finally, an experiment is performed to investigate the clinical consequences
of the genetic variability. Because of the multitude of polymorphisms,
the challenge today is to be able to identify polymorphisms that occur in gene
regulatory or coding regions that have clinical relevance. The field of oncology
adds another level of complexity as issues such as clinically relevant tumor
genome variations not found in the germline and somatic mutations acquired
before or after chemotherapy can affect drug efficacy and toxicity. Moreover,
different polymorphisms of certain genes may exhibit different phenotypes among
different malignancies. Polymorphisms of glutathione S-transferases (GSTs) are a
common example. GSTs facilitate the conjugation of glutathione to xenobiotics,
including anticancer agents (e.g., cyclophosphamide and anthracyclines), environmental
carcinogens, and reactive oxygen products (14). A variety of functionally
important polymorphisms of GSTs have been described. The presence of
the GSTT1 polymorphism has been shown to predispose children treated intensively
for acute myelogenous leukemia to greater toxicity and death during
remission (15), while women with breast cancer treated with cyclophosphamide
carrying the GSTP1 polymorphism are more likely to survive (16). This example
illustrates how the phenotypic effect of polymorphisms depends on the underlying
tumor as well as the intensity of regimen used to treat the malignancy. However,
as we begin to unravel and accurately identify the mechanism of action of many
anticancer drugs, polymorphisms in candidate genes likely to influence drug
efficacy and associated adverse events can be identified. Indeed the field of
pharmacogenetics has transitioned from one that determined the genetic basis of
observed inter patient phenotypic variability to one that determines the phenotypic
consequence of an observed genotypic variability.
Linking the information obtained from the pharmacology and pharmacoepidemiology
of a drug and the pharmacogenetics that describes the genetic
makeup of the individual being treated with the drug is thus vitally important in
developing an optimal approach to individualized therapy. The use of isoniazid
in the treatment of tuberculosis is a good example. Epidemiological evidence
shows that polymorphisms differ in frequency among ethnic and racial groups.
This phenomenon was shown early on when interindividual and interethnic
variations of efficacy and toxicity associated with isoniazid treatment was
observed in a cohort of individuals with tuberculosis. The enzyme responsible
for the metabolism of isoniazid is N-acetyltransferase, a phase II conjugating
liver enzyme. Early studies of isoniazid revealed a bimodal distribution in its
plasma concentration among individuals. High plasma concentrations were

Pharmacology, Pharmacogenetics, and Pharmacoepidemiology 17
observed among individuals with the slow acetylator phenotype often resulting
in peripheral nerve damage, while low plasma concentrations of isoniazid were
observed among individuals with fast acetylator phenotype and were not affected
(17). Studies have shown large variations in the incidence of the slow acetylator
phenotype (inherited as an autosomal recessive trait) among different ethnic
groups (40–70% of Caucasians and African Americans, 10–20% of Japanese, and
more than 80% of Egyptians) (18,19). Amonafide (20), a topoisomerase II
inhibitor, is another substrate of N-acetyltransferase whose clinical development
has been hampered by highly variable and unpredictable toxicity caused, at least,
in part, by interindividual differences in N-acetylation with fast acetylators
experiencing greater myelosuppression than slow acetylators. This example
illustrates both the importance of pharmacoepidemiological evidence in identifying
and characterizing phenotypic variance to individual drugs and how the
pharmacological effects of these drugs depend on pharmacogenetic determinants
that define sensitivity to the drug, and ultimately describes the drug’s behavior.
THE LEUKEMIA STORY: WHAT DID WE LEARN?
Pharmacological and pharmacogenetic studies are increasingly being used to
optimize existing drug therapy with the goal of reducing toxicity while maximizing
efficacy, leading to improved survival. By utilizing this method
remarkable progress has been observed in the treatment of acute lymphoblastic
leukemia (ALL) with cure rates increasing from a mere 10% (four decades ago)
to nearly 80% in children and 40% in adults (21). This has been achieved
through optimization of use of a variety of commonly used antileukemic drugs
such as mercaptopurine and methotrexate.
Thiopurine methyltransferase (TPMT) is an enzyme that converts thiopurine
prodrugs, such as mercaptopurine, into inactive methylated metabolites.
Prior to inactivation, mercaptopurine is converted into a variety of active
nucleotide metabolites that are incorporated into DNA and RNA, thereby
inducing an antileukemic effect. Patients with an inherited TPMT deficiency
suffer severe, potentially fatal hematopoietic toxicity when exposed to standard
doses of mercaptopurine (and other thiopurine prodrugs such as azathioprine and
thioguanine). More than 95% of the clinically relevant TPMT mutations are
accounted for by three nonsynonymous SNPs (TPMT*2, TPMT*3A, and
TPMT*3C) (22). TPMT activity is inherited as an autosomal codominant trait
with approximately 95% of the population homozygous for the wild-type allele
TPMT*1 (that have full enzyme activity), 10% heterozygous for the polymorphism
(that have intermediate levels of enzyme activity), and 1 in 300
individuals will carry two mutant TPMT alleles (that do not express functional
TPMT) (23). Using a pharmacogenetic test, developed at St. Jude Hospital,
patients are often classified according to their levels of TPMT activity into
normal, intermediate, and deficient. This test has a concordance rate of 100%
between genotype and phenotype with studies revealing no long- term outcome

18 Dawood
Figure 3 Individualization of dosing based on TPMT status does not affect cumulative
incidence of relapse. Source: From Ref. 24.
changes related to dosage individualization based on the TPMT status of a
patient (24) (Fig. 3). This illustrates how the clinical application of pharmacogenetics
can be used to minimize toxicity without compromising efficacy.
Methyltetrahydrofolate reductase (MTHFR), a crucial enzyme in the
metabolism of folic acid, catalyses the reduction of 5,10-methylenetetrahydrofolate
to 5-methyltetrahydrofolate (the predominant circulatory form of folate
and the carbon donor for the demethylation of homocysteine to methionine). As
methotrexate increases serum levels of homocysteine and the active polyglutamate
metabolites of methotrexate inhibit MTHFR function, patients with a
low level of MTHFR function are at an increased risk of methotrexate-induced
toxicities such as oral mucosities. Two polymorphisms associated with reduced
MTHFR activity have been described (25): (i) 677C > T polymorphism that is
homozygous in 10% of the population where it encodes the enzyme with 30% of
the wild-type enzyme activity and heterozygous in 40% of the population where
it encodes an enzyme that has 60% of the wild type-enzyme activity and (ii)
1298A > C (less common than 677C > T). Recent findings have also suggested
that MTHFR polymorphisms may protect against the development of both adult
and pediatric ALL (26).
5-FLUOROURACIL
5-Fuorouracil (5-FU) is a uracil analog widely used to treat solid tumors such as
breast and colorectal cancer (27). 5-FU is inactivated by dihydropyrimidine
dehydrogenase (DPD), an enzyme that exhibits up to 20-fold variation in activity
among individuals (28). Mutations that result in low DPD activity predispose to
potentially life-threatening gastrointestinal, hematopoietic, and neurological
toxicities (29). Up to 20 mutations (30) have been identified that cause inactivation
of DPD. In the general population it has been estimated that 3% to 5% of
individuals are heterozygous and 0.1% are homozygous carriers for the mutation
that inactivates DPD, with DPYD*2A being the most inactivating allele (31).
Although polymorphisms associated with DPD inactivation have been identified,
a number of patients still exhibit severe 5-FU toxicity with no detectable

Pharmacology, Pharmacogenetics, and Pharmacoepidemiology 19
mutations in the coding region of the DPD gene explaining the difficulty in
implementation of DPD pharmacogenetics in the prospective identification of
high-risk patients for severe 5-FU toxicity.
One of the primary mechanisms of action of 5-FU is inhibition, via fluorodeoxyuridine
monophosphate (an active metabolite of 5-FU), of thymidylate
synthase (TS), a critical enzyme required for de novo thymidylate synthesis,
which is required for DNA synthesis and repair (32). Polymorphisms in the gene
encoding TS are one of the few examples of a polymorphism that is associated
with the genetic makeup of the tumor with clinical studies showing an inverse
relationship between TS levels and antitumor response (33). The TSER polymorphism
of TS is characterized by tandem repeats in the TS promoter enhancer
region of the gene. Higher levels of TS expression and enzyme activity have been
observed with increasing copies of tandem repeats, which result in lower tumor
response rates to 5-FU (34). In a study by Marsh et al (35), it was observed that in
a cohort of patients with metastatic colorectal tumors receiving 5-FU-based chemotherapy
2 tandem repeats was nearly twice as common in responders than in
non-responders with longer median survival observed in patients with 2 tandem
repeats compared to those with 3 tandem repeats (16 months vs. 12 months) (35).
Polymorphisms of TS that result in 5-FU variability of its pharmacodynamic
and pharmacokinetic properties illustrate the important fact that cancer
being a polygenic disease will need a polygenic solution to guide therapy. To
accurately select patients that are most likely to tolerate and respond to 5-FU
therapy, one needs to use combined genotyping of DPYD and TSER functional
variants as well as taking into account nongenetic factors.
TAXANES
Paclitaxel and Docetaxel are the two most commonly used taxanes in the
treatment of both early and advanced stage breast cancer (36). They exert their
cytotoxic effects by binding to B-tubulin that subsequently results in the stabilization
of microtubules and disruption of mitotic spindle formation during cell
division ultimately resulting in cell death. Genes involved in drug transport,
metabolism, and drug target all play an important role in overall taxane efficacy,
with their polymorphisms partly explaining the variability observed in toxicity
and response to these drugs.
The multidrug-resistant transporter, P-glycoprotein, encoded by ABCB1
(MDR1) and expressed in multiple cell types, has been shown to mediate both
paclitaxel and docetaxel influx (37,38). Three polymorphisms in ABCB1 have
been commonly studied (3435C>T, 1236C>Y, and 2677G>T/A) (39), with
most studies showing inconsistent results when trying to establish an association
between these polymorphisms and drug efficacy/toxicity. Studies focusing on
looking at ABCB1 polymorphism in combination may yield more useful clinical
information. Some of the recent studies have reported an association of the
various ABCB1 polymorphisms with tumor response to paclitaxel (ABCB1

20 Dawood
2677G>T/A) (40), risk of neutropenia with docetaxel (ABCB1 3435C>T) (41),
and risk of neuropathy with paclitaxel (ABCB1 3435C>T) (42). Oxidative
taxane metabolism occurs via the CYP450 pathway, with CYP2C8 and CYP3A4
primarily responsible for paclitaxel metabolism while docetaxel is mainly
metabolized by CYP3A4 (43,44). The functional significance of the polymorphisms
of these enzymes has not been established with associations seen in
in vivo studies (45) not observed in studies of cancer patients (46).
TAMOXIFEN
Tamoxifen is a selective estrogen receptor modulator (SERM) that is widely
used in the treatment and prevention of breast cancer. When given in the
adjuvant setting, it reduces the annual risk of recurrence of breast cancer by half
and the breast cancer mortality rate by one-third (47). The pharmacology and
metabolism of tamoxifen is complex and undergoes extensive phase I and II
metabolism with the production of endoxifen and to a lesser extent 4-hydroxytamoxifen
(48). These metabolites play an important role in the anticancer
effect of tamoxifen through their high-affinity binding to estrogen receptors and
suppression of estradiol-stimulated cell proliferation (49).
The clinical effects of tamoxifen are highly variable (50) due to the variability
observed in the plasma levels of the more potent metabolites of tamoxifen.
Tamoxifen is converted to endoxifen (4-hydroxy-N-desmethyltamoxifen) primarily
by CYP2D6-mediated oxidation of N-desmethyl tamoxifen, while the
conversion of tamoxifen to 4-hydroxy tamoxifen is catalyzed by multiple enzymes
(51). Plasma concentrations of endoxifen is thus sensitive to variants of the
CYP2D6 genotype with the most common variant making the enzyme inactive
(common allele associated with CYP2D6 poor metabolizer phenotype is
CYP2D6*4). One prospective trial revealed that women treated with tamoxifen
who either carried the genetic variants associated with low or absent CYP2D6
activity or were treated concomitantly with drugs that inhibited the activity of
CYP2D6 had lower levels of endoxifen (52). Subsequently, using tissue sample
from a cohort of early stage breast cancer patients treated with tamoxifen, it was
shown that women with a homozygous CYP2D6 phenotype tended to have a
higher risk of disease relapse and a lower incidence of hot flashes (53).
Polymorphisms associated with the activity of the metabolites of tamoxifen
teach us two important lessons. First, the interindividual variability in the
response to tamoxifen is explained both by genetic variation in CYP2D6 and
the coadministration of drugs that inhibit the activity of CYP2D6. Second, the
toxicity associated with drug can help identify responders.
CONCLUSION
At present time, dosages of anticancer drugs are administered based on the
assumption of homogeneity among individuals and are rarely based on individual
characteristics. As a result we encounter a situation where by a large

Pharmacology, Pharmacogenetics, and Pharmacoepidemiology 21
number of individuals experience unwanted adverse side effects after treatment
without attaining maximum drug efficacy. Such adverse drug reactions impose a
huge burden not only on the patient but also on the health care system. A better
understanding of the pharmacoepidemiological, pharmacological, and pharmacogenetic
determinants of a chemotherapeutic agent is needed to optimize drug
dosing regimens to obtain maximal efficacy with minimal toxicity. Advancements
in the field of pharmacogenetics hold promise in helping to identify which
polymorphisms are likely to be important in clinical oncology thereby enabling
the development of effective agents that will make their way to the clinic in a
timely manner. Realizing the importance of each of the components of individualized
therapy and incorporating each one into the infrastructure of clinical
trials will allow us to further explore and define the relationships between different
polymorphisms and observed phenotypes. Considering the fact that anticancer
drugs have narrow a therapeutic range oncology is a field that will
certainly benefit from the application of pharmacogenetic and pharmacological
principles to dosage individualization. The idea of the “one model fits all” should
be accepted as grossly flawed and, as we progress in the field of cancer therapeutics,
will be replaced in the future with the application of “personalized
cancer chemotherapy” into routine clinical practice.
REFERENCES
1. Jemal A, Siegal R, Ward E, et al. Cancer statistics 2007. CA Cancer J Clin 2007;
57:43–66.
2. Bordet R, Gautier S, Le Louet H, et al. Analysis of the direct cost of adverse drug
reactions in hospitalized patients. Eur J Clin Pharmacol 2001; 56:935–941.
3. Evans WE, Reilling MV. Pharmacogenomics: Translating functional genomics into
rational therapeutics. Science 1999; 286:487–491.
4. Jeffrey S, Ross MD, David P, et al. Pharmacogenomics. Adv Anat Pathol 2004;
11:211–220.
5. Maitland-van der Zee AH, de Boer A, Leufkens HG. The interface between pharmacoepidemiology
and pharmacogenetics. Eur J Pharmacol 2000; 410:121–130.
6. Gaplin AJ, Evans WE. Therapeutic drug monitoring in cancer management. Clin
Chem 1993; 39:2419–2430.
7. Calvert A, Newell D, Gumbrell L, et al. Carboplatin dosage: prospective evaluation
of a simple formula based on renal function. J Clin Oncol 1989; 7:1748.
8. Last JM. A Dictionary of Epidemiology. Oxford: Oxford University Press, 2000.
9. Sander C. Genomic medicine and future of health care. Science 2000; 287:1977–
1978.
10. Vesell E.S. Therapeutic lessons from pharmacogenetics. Ann Intern Med 1997;
126:653–655.
11. Evans WE, Relling MV. Pharmacogenomics: Translating functional genomics into
rational therapeutics. Science 1999; 286:487–491.
12. Relling MV, Dervieux T. Pharmacogenetics and cancer therapy. Nat Rev Cancer
2001; 1:99–108.

22 Dawood
13. Sachidanadam R, Weissman D, Schmidt SC, et al. A map of human genome
sequence variation containing 1.42 million single nucleotide polymorphisms. Nature
2001; 409:928–933.
14. Ketterer B. Protective role of glutathione and glutathione transferases in mutagenesis
and carcinogenesis. Mutat Res 1988; 202:343–361.
15. Davies SM, Robison LL, Buckley JD, et al. Glutathione S-transferase polymorphisms
and outcome of chemotherapy in childhood acute myeloid leukemia.
J Clin Oncol 2001; 19:1279–1287.
16. Sweeney C, McClure GY, Fares MY, et al. Association between survival after
treatment for breast cancer and glutathione S-transferase P1 Ile105Val polymorphism.
Cancer Res 2000; 60:5621–5624.
17. Evans DA, Manley KA, Mckusch VA, et al. Genetic control of isoniazid metabolism
in man. Br Med J 1960; 2:485–491.
18. Knight RA, Selin MJ, Harris HW, et al. Genetic factors influencing isoniazid blood
levels in humans. Trans Conf Chemother Tuberc 1959; 8:52–56.
19. Evans Da. Survey of human acetylator polymorphism in spontaneous disorders.
J Med Genet 1984; 21:243–253.
20. Ratain MJ, Mick R, Berezin F, et al. Paradoxical relationship between acetylator
phenotype and amonafide toxicity. Clin Pharmacol Ther 1991; 50:573–579.
21. Pui C-H, Evans WE. Acute lymphoblastic leukemia. N Engl J Med 1998; 339: 605–615.
22. McLeod HL, Coulthard S, Thomas, AE et al. Analysis of thiopurine methyltransferase
variant alleles in childhood acute lymphoblastic leukemia. Br J Hematol
1999; 105:696–700.
23. Krynetski EY, Tai HL, Yates CR, et al. Genetic polymorphisms of thiopurine
s-methyltransferase: clinical importance and molecular mechanisms. Pharmacogenetics
1996; 6:279–290.
24. Relling MV, Pui C-H, Cheng C, et al. Thiopurine methyltransferase in acute lymphoblastic
leukemia. Blood 2006; 107:843–844.
25. Molloy AM, Daly S, Mills JL, et al. Thermolabile variant of 5,10-methyenetetrahydrofolate
reductase associated with low red-cell folates: implications for folate
intake recommendations. Lancet 1997; 349:1591–1593.
26. Franco RF, Simoes BP, Tone LG, et al. The methylenetetrahydrofolate reductase
C677T gene polymorphism decreases the risk of childhood acute lymphocytic leukemia.
Br J Haematol 2001; 115:616–618.
27. Heidelberger C, Chaudhuri NK, Danneberg P, et al. Fluorinated pyrimidines, a new
class of tumor – inhibitory compounds. Nature 1957; 179:663–666.
28. Etienne MC, Lagrange JL, Dassonville O, et al. Population study of dihydropyrimidine
dehydrogenase in cancer patients. J Clin Oncol 1994; 12:2248–2253.
29. Raida M, Schwabe W, Hausler P, et al. Prevalence of a common point mutation in
the dihydropyrimidine dehydrogenase (DPD) gene within the 5’-splice donor site of
intron 14 in patients with severe 5-fluorouracil (5-FU)-related toxicity compared
with controls. Clin Cancer Res 2001; 7:2832–2839.
30. McLeod HL, Collie-Duguid ES, Vreken P, et al. Nomenclature for human DPYD
alleles. Pharmacogenetics 1998; 8:455–459.
31. Lu A, Zhang R, Diasio RB. Dihydropyrimadine dehydrogenase activity in human
peripheral blood mononuclear cells and liver: population characteristics, newly
identified deficient patients, and clinical implication in 5-fluorouracil chemotherapy.
Cancer Res 1993; 53:5433–5438.

Pharmacology, Pharmacogenetics, and Pharmacoepidemiology 23
32. Grem JL. 5-Fluorouracil: forty-plus and still ticking. A review of its preclinical and
clinical development. Invest New Drugs 2000; 18:299–313.
33. Leichman CG, Lenz HJ, Leichman L, et al. Quantitation of intratumoral thymidylate
synthase expression predicts for disseminated colorectal cancer response and resistance
to protracted-infusion fluorouracil and weekly leucovorin. J Clin Oncol 1997;
15:3223–3229.
34. Horie N, Aiba H, Oguro K, et al. Functional analysis and DNA polymorphism of the
tandemly repeated sequences in the 5’-terminal regulatory region of the human gene
for thymidylate synthase. Cell Struct Funct 1995; 20:191–197.
35. Marsh S, Mckay JA, Cassidy J, et al. Polymorphism in the thymidylate synthase
promoter enhancer region in colorectal cancer. Int J Oncol 2001; 19:383–386.
36. Nabholtz JM, Gligorov J. The role of taxanes in the treatment of breast cancer.
Expert Opin Pharmacother 2005; 6:1073–1094.
37. Shirakawa K, Takara K, Tanigawara Y, et al. Interaction of docetaxel (‘Taxotere’)
with human P-glycoprotein. Jpn J Cancer Res 1999; 90:1380–1386.
38. Sparreboom A, Van Asperen J, Mayer U, et al. Limited oral bioavailability and
active epithelial excretion of paclitaxel (Taxol) caused by P-glycoprotein in the
intestine. Proc Natl Acad Sci U S A 1997; 94:2031–2035.
39. Lepper ER, Nooter K, Verweij J, et al. Mechanisms of resistance to anticancer drugs:
the role of the polymorphic ABC transporters ABCB1 and ABCG2. Pharmacogenomics
2005; 6:115–138.
40. Green H, Soderkvist P, Rosenberg P, et al. mdr-1 Single nucleotide polymorphisms
in ovarian cancer tissue: G2677T/A correlates with response to paclitaxel chemotherapy.
Clin Cancer Res 2006; 12:854–859.
41. Tran A, Jullien V, Alexandre J, et al. Pharmacokinetics and toxicity of docetaxel:
role of CYP3A, MDR1, and GST polymorphisms. Clin Pharmacol Ther 2006;
79:570–580.
42. Sissung TM, Mross K, Steinberg SM, et al. Association of ABCB1 genotypes with
paclitaxel-mediated peripheral neuropathy and neutropenia. Eur J Cancer 2006;
42:2893–2896.
43. Taniguchi R, Kumai T, Matsumoto N, et al. Utilization of human liver microsomes
to explain individual differences in paclitaxel metabolism by CYP2C8 and CYP3A4.
J Pharmacol Sci 2005; 97:83–90.
44. Leyseng-Williamson KA, Fenton C. Docetaxel a review of its use in metastatic
breast cancer. Drugs 2005; 65:2513–2531.
45. Bahadur N, Leathart JB, Mutch E, et al. CYP2C8 polymorphisms in Caucasions and
their relationship with paclitaxel 6 á- hydroxylase activity in human liver micorsomes.
Biochem Pharmacol 2002; 64:1579–1589.
46. Henningsson A, Marsh S, Loos WJ, et al. Association of CYP2C8, CYP3A4,
CYP3A5, and ABCB1 polymorphisms with pharmacokinetics of paclitaxel. Clin
Cancer Res 2005; 11:8097–8104.
47. Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and
15-year survival: An overview of the randomized trials. Lancet. 2005; 365:1687–1717.
48. Lonning PE, Lien EA, Lundgren S, et al. Clinical pharmacokinetics of endocrine
agents used in advanced breast cancer. Clin Pharmacokinet 1992; 22:327–358.
49. Stearns V, Johnson MD, Rae JM, et al. Active tamoxifen metabolite plasma concentrations
after co administration of tamoxifen and the selective serotonin reuptake
inhibitor paroxetine. J Natl Cancer Inst 2003; 95:1758–1764.

24 Dawood
50. Osborne CK. Tamoxifen in the treatment of breast cancer. N Engl J Med 1998;
339:1609–18.
51. Desta Z, Ward BA, Soukhova NV, et al. Comprehensive evaluation of tamoxifen
sequential biotransformation by the human cytochrome P450 system in vitro: Prominent
roles for CYP3A and CYP2D6. J Pharmacol Exp Ther 2004; 310:1062–1075.
52. Jin Y, Desta Z, Stearns V, et al. CYP2D6 genotype, antidepressant use, and tamoxifen
metabolism during adjuvant breast cancer treatment. J Natl Cancer Inst 2005; 97:30–39.
53. Goetz MP, Rae JM, Suman VJ, et al. Pharmacogenetics of tamoxifen biotransformation
is associated with clinical outcomes of efficacy and hot flashes. J Clin
Oncol 2005; 23:9312–9318.

3
Role of Genetic Variability in Breast
Cancer Treatment Outcomes
Kandace L. Amend
Department of Oncological Sciences, Mount Sinai School of Medicine,
New York, New York, U.S.A.
Ji-Yeob Choi and Christine B. Ambrosone
Department of Cancer Prevention and Control, Roswell Park
Cancer Institute, Buffalo, New York, U.S.A.
INTRODUCTION
Variability in toxicity and response to the same dose of medication may occur
among cancer patients and contribute to the unpredictability of clinical outcomes.
In addition to the potential importance of clinical factors in determining
drug effects (1), inherited differences in drug metabolism are likely to have a
profound effect on treatment outcomes (2). Pharmacogenetics, the study of the
contribution of genetic variation to these interindividual differences, may allow
for the individualization of therapy and improved prediction of pharmacodynamic
events, including host toxicity and treatment efficacy. To date, the role of
pharmacogenetics in breast cancer therapy has not been clearly defined. Herein
we discuss the drugs most commonly used to treat breast cancer and describe
their metabolic pathways, as well as provide epidemiologic support for the role
of pharmacogenetics in breast cancer outcomes.
25

26 Amend et al.
BREAST CANCER THERAPEUTICS
The most frequently used agents to treat breast cancer include cyclophosphamide
(CP) in combination with doxorubicin [Adriamycin 1 (A)] or with methotrexate
(MTX) and 5-flourouracil (5FU) (3). The taxanes paclitaxel (Taxol 1 ) and
docetaxel (Taxotere 1 ) have an important role in the treatment of early-stage
breast cancer as well as metastatic disease. In patients with estrogen receptor
(ER)-positive tumors, tamoxifen (TAM) reduces the risk of recurrence by 31%
and breast cancer morality by one-third (4), and randomized trials have indicated
that treatment with third-generation aromatase inhibitors (AIs) is now an
important option for the adjuvant treatment of hormone receptor–positive breast
cancer (5,6).
Multiple enzymatic pathways are involved in the metabolism and detoxification
of chemotherapeutic agents, and common genetic polymorphisms in
drug metabolizing enzymes are likely to contribute to interindividual variability
in toxicity and cancer recurrence (7).
Cyclophosphamide
Cyclophosphamide requires multistep activation before it can function as an
antitumor alkylating agent. The drug is first oxidized in the liver to the active
4-hydroxy metabolite mainly by CYP2B6, although other enzymes, including
CYP2A6, CYP3A4, CYP2C8, CYPC29, and CYP2C19, are involved. Drug
intermediates are transported in blood and decompose spontaneously to phosphoramide
mustard and acrolein, both of which can react with nucleophilic
groups, forming DNA cross-links and inhibition of protein synthesis (8). Inactivating
glutathionyl conjugation reactions of C intermediates are catalyzed by
glutathione S-transferases (GSTs) GSTA1 and GSTP1 (9). Variability in the
activity of these activating and detoxifying enzymes is likely to affect ultimate
tumor tissue dose and efficacy of treatment.
The human cytochrome P450 CYP2B6 gene is involved in the metabolic
activation of a number of clinically important chemotherapeutic drugs for breast
cancer, including CP (10,11) as well as the antioestrogen TAM (12). Several
single nucleotide polymorphisms (SNPs) in CYP2B6 have been described, some
having functional significance (13) and affecting the pharmacokinetic parameters
of CP. In cell culture, enzymatic activities in microsomes from COS-1 cells
expressing a K262R amino acid substitution (CYP2B6*4, CYP2B6*6, and
CYP2B6*7) showed increased values for V max and V max /K m compared with that
of the wild type (CYP2B6*1) (14), while a second study reported that the amino
acid substitution Q172H resulting from the CYP2B6*6 allele enhanced
7-ethoxycoumarin O-deethylase activity (15). A study by Lang et al. (2001)
identified a total of nine SNPs by sequencing DNA from 35 subjects and
reported extensive variability in the expression and activity of CYP2B6, where
CYP2B6 expression was significantly reduced in carriers of the R487T, Q172H,

Role of Genetic Variability in Breast Cancer Treatment Outcomes 27
and K262R polymorphisms (16). Ethnic variation in allelic variants and CYP2B6
expression has been also been observed (17,18).
CYP3A4 also plays a large role in the metabolism of CP, and the CYP3A4
gene has a number of known variant alleles, including some having functional
significance (19). A single base substitution in the 5 0 promoter region of the gene
(20), present in 9% of European-Americans and 53% of African-Americans (21),
with the CYP3A4*1B allele associated with expression levels is 1.4-fold higher
than that in common alleles (22). Using enzyme kinetic analyses to evaluate
intersample variation in activation of CP by CYP3A and 2B in human liver, both
Chang et al. (1993) and Roy et al. (1999) found considerable interindividual
differences in CYP3A and CYP2B levels and variation over a wide range in CP
hydroxylation (10,23). Variation in CYP3A enzyme levels and expression may
be related, in part, to genetic differences or to other chemotherapeutic and
natural agents that have been shown to increase CYP3A4 expression (24).
CYP3A5 accounts for at least 50% of the total CYP3A content in most
Caucasians and African-Americans and is thus considered an important contributor
to interindividual differences in CYP3A-dependent drug clearance (25).
Although CYP3A5 is polymorphically expressed at high levels in a minority of
Caucasians, the expression in human liver is more frequent in African-
Americans (60%) than in Caucasians (33%) (25). Reduced expression in some
individuals may be due to a common polymorphism in intron 3 of CYP3A5
(A6986G), which generates a cryptic splice site (CYP3A*6) resulting in a
truncated amino acid (25,26).
Genetic polymorphisms have been described in several of the CYP2C
family members, including CYP2C8, CYP2C29, CYP2C18, and CYP2C19,
although the catalytic contribution of CYP2C8 and CYP2C18 to activation of CP
is likely minor compared with the contributions of CYP2C9 and CYP2C19. In
vitro, CYP2C19 was found to bioactivate CP in human liver microsomes (27),
and while both CYP2C9 and CYP2C19 were found to possess CP 4-hydroxylase
activity in vitro, CYP2C19 exhibited higher CP 4-hydroxylase activity following
only CYP2B6 (23). Although known variants in CYP2C9 and CYP2C19 exist,
little is known regarding their contribution to pharmacokinetic variability in
relation to breast cancer prognosis. In a study by Xie et al., CYP2C9*2,
CYP2C9*3, CYP2C19*2, and CYP2C19*3 were not significantly correlated with
the metabolism of CP in 19 human liver specimens (11).
Phase II drug metabolism is catalyzed by GSTs, which consist of several
isoenzymes, including mu (GSTM1), pi (GSTP1), alpha (GSTA1), theta
(GSTT1), and zeta. The GSTs detoxify many chemotherapeutic drugs, including
CP, are induced by oxidative stress, and detoxify the reactive oxygen species
(ROS) that are produced by chemotherapeutic drugs and radiation therapy.
GSTA1, the enzyme most active in glutathione conjugation reactions with CP
intermediates, has a polymorphism in the 5 0 promoter region of the gene (28).
Coles et al. (2001) identified substitutions designated as GSTA1*A ( 567, 69,
and 52) and GSTA1*B ( 631, 567, and 69) (29), with the GSTA1*B variant

28 Amend et al.
showing decreased hepatic expression and increased GSTA2 expression compared
with GSTA1*A homozygotes. Given that GSTA1 increases the conjugation
of CP intermediates (9), these differences in hepatic expression may affect the
detoxification of CP metabolites and, ultimately, the outcomes of treatment and
survival.
For the GSTP1 gene, two variant alleles, GSTP1*B and GSTP1*C, have
been detected (30). A single amino acid change from isoleucine to valine occurs
at codon 105 (Ile105Val) of the GSTP1 gene, with genotype frequencies in
Caucasians 51% homozygous for Ile/Ile, 43% heterozygous for Ile/Val, and 6%
homozygous for the variant allele Val/Val (31). The allelic variant 105Val
exhibited lower glutathione conjugation of the alkylating agent thiopeta compared
with the isoleucine allele (32). GSTP1 variants may also differ in activity
toward CP, structurally similar to thiotepa, and patients with GSTP1 allelic
variants may have reduced activity in removal of active CP.
Anthracyclines
The commonly used anthracyclines include doxorubicin, daunorubicin, and
epirubicin, although doxorubicin and epirubicin are used more frequently for the
treatment of breast cancer. Several different mechanisms have been proposed for
anthracycline tumor activity, including free radical generation, DNA binding and
alkylation, DNA cross-linking, interference with DNA unwinding or strand
separation, direct membrane effects, inhibition of topoisomerase II, and induction
of apoptosis (33). Doxorubicin, like other anthracyclines, results in the
formation of quinone-mediated free radicals, which have the capacity to cause
oxidative damage and cytotoxicity, resulting in significant cardiotoxicity as a
result of the production of ROS (34). Cellular oxidoreductases reduce Adriamycin
to a semiquinone radical that is subsequently reoxidized by oxygen to a
superoxide anion and the parent quinine (35–37). Superoxide anions can dismutate
to form hydrogen peroxide and/or react with nitric oxide to form peroxynitrite.
Lipid peroxides resulting from doxorubicin can further break down to
yield hydroxyalkenals, which are substrates for glutathione-conjugating isozymes
(38), indicating the potential importance of GSTs in mediating therapeutic
outcomes.
Carbonyl reductase (CBR) is a monomeric, NADPH-dependent oxidoreductase
that catalyzes the reduction of a variety of carbonyl compounds,
including doxorubicin and daunorubicin. In humans, two CBRs have been
identified, CBR1 and CBR3. An SNP has been identified in CBR3 (CBR3
V244M), where the M244 isoform results in higher V max and is 100% more
efficient in catalyzing the reduction of substrate (39). To our knowledge, CBRs
have not been investigated in relation to breast cancer outcomes following
treatment with anthracyclines.
In addition to its mechanistic role in cytotoxicity of anthracyclines, it is well
documented that oxidative stress plays a role in the cell-killing abilities of many

Role of Genetic Variability in Breast Cancer Treatment Outcomes 29
other chemotherapy agents, as well as in radiation therapy. Specifically for breast
cancer, CP, doxorubicin (Adriamycin), and paclitaxel and radiation therapy generate
ROS and induce apoptosis, which may contribute to treatment efficacy and
increased survival. ROS can cause mitochondrial permeability, and once the
mitochondrial membrane barrier function is lost, a number of other factors contribute
to cell death. While ROS, among other factors, induce or facilitate mitochondrial
permeability, glutathione and antioxidant enzymes such as manganese
superoxide dismutase (MnSOD), catalase (CAT), and glutathione peroxidase
(GPX1) inhibit it (40). In the mitochondrion, MnSOD catalyzes the dismutation of
two superoxide radicals, producing H 2 O 2 and oxygen. If H 2 O 2 is not neutralized, it
may contribute to further generation of ROS by a reaction catalyzed by myeloperoxidase
(MPO). Thus, levels of MnSOD, CAT, GPX1, and MPO are important
in determining the amount of ROS or detoxification of H 2 O 2 . It is plausible that
interindividual variability resulting from polymorphisms in these genes may
influence the efficacy of cancer treatment and survival.
Two genetic polymorphisms appear to affect mitochondrial levels of
MnSOD. A single-base substitution variant in the MnSOD gene results in an
amino acid substitution from alanine to valine at position 9 (ala-9) in the
mitochondrial targeting sequence (41). The substitution alters the secondary
structure of the protein, affecting the cellular distribution of the enzyme and
transport of MnSOD in the mitochondrion, where it would be biologically
available (41). In the mitochondria, the MnSOD Ala variant has been shown to
generate more active MnSOD protein compared with the Val allele (42). A
second polymorphic variant has been identified in MnSOD, and Ile or Thr at
amino acid position 58 (43) and the Thr 58 MnSOD variant was shown to exhibit
only half the enzymatic activity of Ile 58 MnSOD (43,44). It can be hypothesized
that the variant (ala-9 and Thr 58 MnSOD) would confer reduced protection of
breast tumor cells from ROS produced during therapy. Thus, lesser MnSOD
activity due to polymorphisms could increase treatment efficacy but also result
in greater toxicity due to ROS.
CAT is most abundant in the liver, kidneys, and erythrocytes, where it
decomposes H 2 O 2 to H 2 O and O 2 . A common functional polymorphism
(C-262T) alters transcription factor binding and the expression of CAT in
erythrocytes. We previously found that individuals with C alleles had significantly
higher CAT levels compared with those with T/T genotypes (45).
H 2 O 2 , if not neutralized, may contribute to further generation of ROS by a
reaction catalyzed by MPO. MPO generates ROS endogenously by functioning
as an antimicrobial enzyme, catalyzing a reaction between H 2 O 2 and chloride to
generate hypochlorous acid (HOCl), a potent oxidizing agent. HOCl further
reacts with other biological molecules to generate secondary radicals (46). The
promoter region of the MPO gene has a G-to-A base substitution at position 463,
which creates high- (G) and low-expression (A) alleles (47). Variability in genes
in this oxidative stress-related pathway may influence ultimate levels of ROS
and, thereby, tumor cell kill.

30 Amend et al.
5-Fluorouracil and Methotrexate
The agents 5FU and MTX are involved in folate metabolism, in the conversion
of homocysteine to methionine, and in purine and pyrimidine synthesis. The
activated metabolites of 5FU, similar to pyrimidine nucleotides, competitively
and reversibly bind fluoropyrimidine thymidylase synthase (TS) inhibitors,
forming covalently bound complexes with the TS enzyme, thus inhibiting the
enzyme’s activity (48). Folate antagonists, such as MTX, are structurally similar
to folate and inhibit the action of various enzymes such as dihydrofolate
reductase (DHFR) (49). Their action is accomplished through incorporation as
false precursors in DNA or RNA or through inhibition of proteins involved in
nucleotide metabolism.
Common polymorphisms have been reported in many enzymes and transport
proteins involved in intracellular folate metabolism. Using a candidate gene
approach, pharmacogenetic research, to date, has focused on TS because of its role
as a drug target, dihydropyrimidine dehydrogenase (DPD) because of its
involvement in pyrimidine degradation, and methylenetetrahydrofolate reductase
(MTHFR) because of its key involvement in the control of the flux of intracellular
folate metabolites (50). A genetic polymorphism in the 5 0 regulatory region of the
TS gene promoter, consisting of either double (2R) or triple (3R) repeats of a 28-bp
sequence (51), has been shown to influence TS expression, with higher expression
in 3R/3R tumors (52). A G>C polymorphism has been identified within the
second repeat of 3R alleles that alters the transcriptional activation of the TS gene
constructs bearing this genotype (53). A 6-bp deletion (1494del6) within the
3 0 UTR (54) has been also associated with decreased mRNA stability (55). Several
clinical studies have shown the potential contribution of the above TS germinal
polymorphism in the prediction of tumor responsiveness and/or toxicity following
treatment by fluoropyrimidines (56). However, the role of TS polymorphism as a
predictor of treatment outcome is still unclear.
MTHFR catalyzes the conversion of 5,10-methyltetrahydrofolate to 5-methyltetrahydrofolate,
which is an essential cofactor in the biosynthesis of deoxythymidine
monophosphate. A common C677T transition in exon 4 and an A1298C
transition in exon 7 of the MTHFR gene result in lower specific activity (57,58).
Most studies, to date, have examined the association between 5FU and MTHFR
polymorphisms in the context of colorectal cancer, but Toffoli et al. (59) reported
that of six patients who developed severe acute toxicity in the first cycle of adjuvant
CMF, five had the MTHFR 677 T/T genotype and one had the C/C genotype.
Because 5FU has a relatively narrow therapeutic index, toxicity increases
as the dose is increased, and more than 80% of administered 5FU is catabolized
by DPD (1). Patients with complete or partial deficiency of DPD often experience
severe or even life-threatening toxicity after the administration of 5FU (60).
At least 11 of 33 mutations identified in the dihydropyrimidine dehydrogenase
(DPYD) gene have been detected in patients suffering from severe 5FU-associated
toxicity (61–63).

Role of Genetic Variability in Breast Cancer Treatment Outcomes 31
There have been few studies of MTHFR and TS polymorphisms in relation
to breast cancer treatment outcomes. In a study with 19 human cancer cell lines,
including breast cancer, Etienne et al. (64) found that there was a marked trend
for greater 5FU efficacy in mutated MTHFR 1298C cell lines compared with
those with A/A genotypes (P ¼ 0.055), but MTHFR C677T and TS 2R/3R
polymorphisms were not associated with 5FU sensitivity. Largillier et al. (65)
noted that genetic polymorphisms of MTHFR (677C>T and 1298A>C), DPD
(IVS14þ1G>A), and TS were associated with efficacy and toxicity associated
with capecitabine, an oral 5FU prodrug, in 105 breast cancer patients receiving
capecitabine as monotherapy.
Although MTX has been widely used to treat many cancers, including
leukemia, lymphomas, and breast cancer, as well as some autoimmune diseases,
there has been limited research to evaluate the effects of genetic variability on
treatment outcomes. Associations have been noted for polymorphisms in folaterelated
genes in leukemia (66–69), and Sohn et al. (70) found that in vitro
transfection of mutant 677T MTHFR cDNA in breast and colon cancer cells
increased the chemosensitivity of these cells to 5FU but decreased the chemosensitivity
of breast cancer cells to MTX, suggesting at least a modulating role of
this genotype on cellular fluouropyrimidine sensitivity.
Taxanes
The taxanes paclitaxel and docetaxel interfere with microtubular disassembly,
ultimately resulting in DNA fragmentation and features of apoptosis (71). The
contributions of genetic variability to paclitaxel metabolism and breast cancer
outcomes is still unclear. CYP2C8 is thought to be the main CYP-metabolizing
enzyme of paclitaxel, while CYP3A4 and CYP3A5 play more minor roles in drug
metabolism (19). At least three known variants in CYP2C8 exist, and the CYP2C8*3
variant allele has been associated with decreased paclitaxel 6-ahydroxylase activity
in human cell lines and human liver microsomes (72). In a recent study of
paclitaxel metabolism in breast cancer, however, CYP2C8 genotypes (CYP2C8*3
and CYP2C8*4) were not significantly associated with paclitaxel clearance (73).
Ethnic variation exists in the distribution of CYC2C8 alleles, with the CYP2C8*2
allele more common in African-Americans than Caucasians (18% vs. 0%,
respectively) and the CYP2C*3 allele more common in whites (13%) than
African-Americans (2%). The CYPC2*5 allele is found in 9% of Asians (74).
Antiestrogens
The pharmacology and metabolism of TAM is complex, and differences in
patient outcomes could be due to individual variation in the metabolism of the
drug. TAM is metabolized primarily by CYP3A4, CYP2D6, CYP2C9,
SULT1A1, and UGT2B15 (75–78). The main metabolites of TAM are
N-desmethyltamoxifen (formed by CYP3A4) and endoxifen and 4-OH-TAM

32 Amend et al.
(formed by CYP2D6) (8,76). CYP2D6 is polymorphic with 46 known allelic
variants, which are categorized as resulting in abolished, decreased, normal,
increased, or qualitatively altered catalytic activity (79). The major polymorphic
variant CYP2D6 alleles include CYP2D6*2, CYP2D6*4, CYP2D6*5,
CYP2D6*10, CYP2D6*17, and CYP2D6*41 (79).
Human sulfotransferase 1A1 (SULT1A1) catalyzes the sulfation of a
variety of phenolic and estrogenic compounds, including endogenous and
environmental estrogens and the 4-hydroxy metabolite of TAM, 4-OH TAM
(80). An exon 7 polymorphism in SULT1A1 (SULT1A1*2) results in a translated
protein with approximately twofold lower catalytic activity and decreased
thermostability than the common allele (SULT1A1*1) (81). We found, as
hypothesized, that women treated with TAM who were homozygous for the
variant SULT1A1*2/*2 allele had approximately three times the risk of death
compared with those homozygous for the common allele or heterozygotes (82).
A wide range of drugs are substrates for glucuronosyltransferases (UGTs)
and this pathway has been estimated to account for approximately one-third of
all drugs metabolized by phase II drug metabolizing enzymes (2). SULT1A1 and
UGT2B15 have overlapping substrate specificity for 4-OH TAM. A functional
polymorphism in UGT2B15, a G?T substitution at codon 85 within the putative
substrate recognition site of the gene, results in higher V max compared with the
common allele (83). In breast cancer patients treated with TAM, the combined
effects of SULT1A1 and UGT2B15 were evaluated, and an increasing number of
variant SULT1A1 and UGT2B15 alleles was associated with poorer survival (84).
Aromatase Inhibitors
AIs, including exemestane, anastrozole, and letrozole, have come into clinical
use more recently, developed primarily for hormone-dependent breast cancer in
postmenopausal women. AIs act by inhibiting the cytochrome P450 enzyme,
aromatase (CYP19), which catalyzes the conversion of androgens to estrogens.
AIs block this enzyme, decreasing circulating levels of estrogens. It has been
demonstrated that exemestane is metabolized by CYP3A4 but is also metabolized
by aldoketoreductases (85). Letrazole is metabolized by CYP2A6 and
CYP3A4 to a pharmacologically inactive carbinol metabolite (86), and anastrazole
is metabolized by N-deaklylation and hydroxylation reactions followed
by glucuronidation. However, specific isoforms involved and the effects on drug
levels have not yet been identified.
MOLECULAR EPIDEMIOLOGIC STUDIES OF PHARMACOGENETICS
AND BREAST CANCER OUTCOMES
To date, most analytical studies of pharmacogenetics and breast cancer outcomes
have been conducted among patients receiving heterogeneous chemotherapy
treatments, although most chemotherapy regimens included CP. Several studies

Role of Genetic Variability in Breast Cancer Treatment Outcomes 33
have evaluated polymorphisms in phase II or GST enzymes in relation to
prognosis after treatment for breast cancer. In a study of 92 women with
advanced breast cancer, the GSTM1-null genotype was not significantly associated
with tumor characteristics, disease-free survival (DFS), or overall survival
(OS) (87), but in a retrospective study with a larger sample size, we found that
women with low-activity homozygous genotypes for the GSTA1*B allele had
better survival, with the hazard of death reduced by 30% (88,89), and women
with low-activity GSTP1 Val/Val genotype had a similar risk reduction (89).
Deletion polymorphisms in GSTM1 and GSTT1, genes responsive to ROS and
resulting from radiation or chemotherapy and subsequent lipid peroxidation,
were associated with a reduced hazard of death compared with women with both
alleles present [HR ¼ 0.59; 95% confidence interval (CI), 0.36–0.97; and HR ¼
0.51; 95% confidence interval (CI), 0.29–0.90] (90). In a study in China (91),
women with GSTP1 Val/Val genotypes experienced a 60% reduction in
overallmortality (HR ¼ 0.4; 95% confidence interval (CI), 0.2–0.8), although
no associations were found with GSTM1 or GSTT1 genotypes. A smaller study
(n ¼ 85) evaluated polymorphisms in both phase I and phase II metabolism of
CP among women with metastatic or inflammatory breast cancer treated with
high-dose CP, cisplatin, and carmustine. Patients with the GSTM1-null genotype
had prolonged OS, while those with variants in CYP3A4*1B and CYP3A5*1 had
shortened survival. To our knowledge, only one genotype association study has
focused on paclitaxel metabolism. In that study of 93 patients (73), there were no
significant associations between ABCB1, ABCG2, CYP1B1, CYP3A4, CYP3A5,
and CYP2C8 and paclitaxel clearance. However, patients homozygous for
CYP1B1*3 Leu/Leu genotype had significantly longer progression-free survival
(PFS) compared with patients with at least one Val allele (Table 1).
Few studies have examined the effects of MTHFR genotypes on breast
cancer survival. In the Shanghai Breast Cancer Study (92), no significant
associations were found between MTHFR genotypes and the risk of death. A
second study evaluated these same genotypes in African-American and Caucasian
women from the Baltimore area (93) and found that carriers of the lowactivity
allele at codon 1298 (A/C or C/C) had significantly poorer survival,
while those with the variant allele at codon 677 (C/T or T/T) had improved
survival. Associations were stronger in patients with ER-negative tumors, and
African-American women had improved survival if they were carriers of either
variant allele (A1298C, P interaction ¼ 0.088, C677T, P interaction ¼ 0.026).
Molecular epidemiologic studies have indicated that genetic variation in
phases I and II conjugating enzymes can influence the efficacy of TAM therapy
for breast cancer. In a recent study in Sweden (94), patients homozygous for
CYP2D6*4 alleles experienced significantly better DFS compared with patients
with at least *1 allele (P ¼ 0.05). In a group of women randomized to 5, rather
than two years of TAM, patients homozygous for CYP3A5*3 experienced
improved relapse-free survival (RFS) (HR ¼ 0.02; CI, 0.07–0.55). However, this
finding should be interpreted with caution since the number of patients within

34 Amend et al.
Table 1 Epidemiologic Studies of Genetic Variation in Chemotherapeutic Metabolizing Genes and Breast Cancer Outcomes
Genes SNP Participants Treatment Outcome Risk (95% CI) Reference
Cyclophosphamide
GSTM1 GSTM1-null
genotype
92 patients CAF OS or DFS GSTM1 not significant 87
GSTA1 GSTA1*B/*B 245 patients CP based 5-yr survival GSTA1*B/*B HR ¼ 0.3 (0.1–0.8) 88
GSTP1 GSTP Val/Val 240 patients CP based OS GSTP Val/Val HR ¼ 0.3 (0.1–1.0) 89
GSTP1 GSTP Val/Val 1034 patients CP based OS GSTP Val/Val HR ¼ 0.4 (0.2–0.8) 91
6 genes, CYPs,
ABCB1,
ABCG2
17 genes, CYPs,
METs, GSTs
8 SNPs 93 patients High-dose
paclitaxel and
CP and A
29 SNPs 85 patients CP and cisplatin
and carmustine
PFS CYP1B1*3 Leu/Leu Longer PFS (P ¼ 0.037) 73
PK of CP
and OS
GSTM1-null genotype; better OS; at least one
variant allele in CYP3A4*1B or CYP3A5*1
(98)
Methotrexate and 5-fluorouracil
MTHFR MTHFR C677T;
MTHFR
A1298C
MTHFR MTHFR C677T;
Tamoxifen
CYP3A5;
CYP2D6;
SULT1A1;
UGT2B15
CYP3A5;
CYP2D6
MTHFR
A1298C
CYP3A5*3;
CYP2D6*4;
SULT1A1*2;
UGT2B15*2
CYP3A5*3;
CYP2D6*4;
1067 patients Combination
(most received
5FU)
OS MTHFR genotypes not significant with OS 92
248 patients Combination OS MTHFR 1298 A/C or C/C HR ¼ 2.05
(1.05–4.0);
MTHFR 677 C/T or T/T HR ¼ 0.65 (0.31–1.35)
677 patients
(238
randomized
to TAM)
TAM RFS CYP3A5 *3/*3 in 5-year
TAM group HR ¼ 0.20 (0.07–0.55);
CYP2D6 *4/*4 better
RFS (n ¼ 677) (P ¼ 0.05)
223 patients TAM RFS or DFS
or OS
CYP2D6 *4/*4
HR ¼ 2.71 (1.15–6.41) for RFS;
HR ¼ 2.44 (1.22–4.90) for DFS (both ratios
unadjusted)
(93)
94
(95)

Role of Genetic Variability in Breast Cancer Treatment Outcomes 35
Table 1 Epidemiologic Studies of Genetic Variation in Chemotherapeutic Metabolizing Genes and Breast Cancer Outcomes (Continued )
Genes SNP Participants Treatment Outcome Risk (95% CI) Reference
CYP2D6;
UGT2B15;
SUL1A1
CYP2D6*4;
UGT2B15*2;
SULT1A1*2
337 patients
(162 patients
treated with
TAM)
SULT1A1 SULT1A1*2 377 patients
(160 treated
with TAM)
TAM OS or PFS CYP2D6 NS;
UGT2B15*2 and SULT1A1*2
HR ¼ 4.4 (1.17–16.55) in TAM group
TAM OS SULT1A1 *2/*2
HR ¼ 2.9 (1.1–7.6) in TAM group
84
(82)
Oxidative stress
GSTM1/GSTT1 GSTM1 and
GSTT1-null
genotypes
MPO, MnSOD,
CAT
MnSOD Ala-9Val,
CAT-262C?T,
MPO-463G?A
MnSOD MnSOD-102 C?T 291 patients
(230 treated
with chemotherapy)
251 patients Combination OS and DFS GSTM1-null genotype
HR ¼ 0.59 (0.36–0.97;
GSTT1-null genotype
HR ¼ 0.51 (0.29–0.90)
279 patients Combination OS MPO GG genotype
HR ¼ 0.60 (0.38–0.95),
MnSOD CC genotype HR ¼ 0.70 (0.40–1.24),
genotypes combined HR ¼ 0.33 (0.13–0.80)
Combination RFS MnSOD CT þ TT HR ¼ 0.65 (0.40–1.05)
for chemotherapy group
90
96
97
Abbreviations: CAF, Cyclophosphamide/doxorubicin/fluorouracil; A, doxorubicin; CP, cyclophosphamide; T, DFS, disease-free survival; NS, not significant; OS,
overall survival; PFS, progression-free survival; RFS, relapse-free survival; TAM, tamoxifen; 5FU, 5-fluorouracil; GST, glutathione S-transferase; SNP, single
nucleotide polymorphism; METs, Metallothioneins; MTHFR, methylenetetrahydrofolate reductase; MPO, myeloperoxidase; MnSOD, manganese superoxide dismutase;
CAT, catalase; PT, Pharmacokinetics.

36 Amend et al.
this group was small. No statistically significant differences were found for
CYP2D6, SULT1A1, or UGT2B15 and the length of TAM treatment. In contrast,
Goetz et al. (95) found no significant associations between CYP3A5*3 and breast
cancer outcomes; however, women with the CYP2D6 *4/*4 genotype (n ¼ 13)
had statistically significant worse RFS (P ¼ 0.023) and DFS (P ¼ 0.12) but not
OS (P ¼ 0.169). However, after adjustment for nodal status and tumor size,
associations were no longer statistically significant. These results are in agreement
with our previous report, wherein we noted no significant associations
between the CYP2D6*4 genotype and OS or PFS among breast cancer patients
treated with TAM (84).
In addition to CYP isoforms, phase II enzymes, including sulfotransferase
1A1 (SULT1A1) and UDP-glucuronosyltransferase isoform 2B15 (UGT2B15),
have been implicated in the metabolism of TAM, and genetic variability in these
genes together may influence breast cancer survival. In a cohort of women with
breast cancer who were treated with TAM, those who were homozygous for the
low-activity SULT1A1*2 allele had a threefold higher risk of death compared
with those with at least *1 common allele (HR ¼ 2.9; 1.1–7.6), while no significant
associations were found among women who did not receive TAM (82).
When considering the independent and combined effects of SULT1A1 and
UGT2B15 genotypes on OS and recurrence among women treated with TAM,
there was a nonsignificant trend toward increased risk of disease recurrence and
poorer survival with increasing number of UGT2B15*2 alleles (P ¼ 0.11) (84).
However, there was a significant trend toward decreased OS among women
treated with TAM, with increasing number of variant alleles when UGT2B15 and
SULT1A1 genotypes were combined (P ¼ 0.03), although these risk estimates
may be unstable because only 16 cases had two variant alleles.
We previously found that genotypes for MnSOD and MPO resulting in
higher levels of ROS were associated with better survival in a study of women
from Arkansas, and the combined effects of MnSOD and MPO were greater than
those for either polymorphism alone (HR ¼ 0.33; CI, 0.13–0.80) (96). An
additional MnSOD variant, located in the promoter region of the gene, was also
associated with a significant decrease in RFS (CTþTT genotype HR ¼ 0.65, CI,
0.40–1.05), although no significant associations were found with OS (97).
In the context of a clinical trial for breast cancer with CAF versus CMF
(SWOG 8897) (unpublished data, presented at the San Antonio Breast Cancer
Symposium, 2006), as in the Arkansas study, MPO genotypes encoding for
higher activity were associated with better survival.
FUTURE DIRECTIONS
Studies, to date, indicate that interindividual variability in drug metabolism may
affect treatment-related toxicities and DFS. However, much of the data are
inconsistent, and it is likely that the evaluation of single genes in complex
pathways presents only part of a complicated picture. Thus, future studies of

Role of Genetic Variability in Breast Cancer Treatment Outcomes 37
pharmacogenetics and treatment outcomes should focus on entire drug metabolism
pathways in large sample sets receiving homogeneous treatment regimens.
Furthermore, assessment of variability across an entire gene, through construction
of haplotype blocks and tag SNPs, may give more complete information
than a study of single SNPs. It is likely that, in the future, other, more systemic
markers of inherited drug metabolism capabilities may better predict treatment
outcomes, and scans of variability across the entire genome may reveal factors
that are likely to determine treatment outcomes. The consideration of individual
genotypes, as well as tumor characteristics, in therapeutic decision making holds
promise for individualized treatment for breast cancer in the future.
REFERENCES
1. Heggie GD, Sommadossi JP, Cross DS, et al. Clinical pharmacokinetics of 5-
fluorouracil and its metabolites in plasma, urine, and bile. Cancer Res 1987; 47(8):
2203–2206.
2. Evans WE, Relling MV. Pharmacogenomics: translating functional genomics into
rational therapeutics. Science 1999; 286(5439):487–491.
3. DeVita VT Jr., Hellman S, Rosenberg SA (Eds.). Cancer: Principles and Practice of
Oncology. 5th ed. New York, NY: Lippincott-Raven, 1997.
4. Early Breast Cancer Trialists’ Collaborative Group (EBCTG). Effects of chemotherapy
and hormonal therapy for early breast cancer on recurrence and 15-year
survival: an overview of the randomised trials. Lancet 2005; 365:1687–1717.
5. Cuzick J. Aromatase inhibitors for breast cancer prevention. J Clin Oncol 2005;
23(8):1636–1643.
6. Conte P, Frassoldati A. Aromatase inhibitors in the adjuvant treatment of postmenopausal
women with early breast cancer: putting safety issues into perspective.
Breast J 2007; 13(1):28–35.
7. Choi JY, Nowell SA, Blanco JG, et al. The role of genetic variability in drug
metabolism pathways in breast cancer prognosis. Pharmacogenomics 2006; 7(4):
613–624.
8. Scripture CD, Sparreboom A, Figg WD. Modulation of cytochrome P450 activity:
implications for cancer therapy. Lancet Oncol 2005; 6(10):780–789.
9. Dirven HA, van Ommen B, van Bladeren PJ. Involvement of human glutathione
S-transferase isoenzymes in the conjugation of cyclophosphamide metabolites with
glutathione. Cancer Res 1994; 54(23):6215–6220.
10. Chang TK, Weber GF, Crespi CL, et al. Differential activation of cyclophosphamide
and ifosphamide by cytochromes P-450 2B and 3A in human liver microsomes.
Cancer Res 1993; 53(23):5629–5637.
11. Xie HJ, Yasar U, Lundgren S, et al. Role of polymorphic human CYP2B6 in
cyclophosphamide bioactivation. Pharmacogenomics J 2003; 3(1):53–61.
12. White IN, De MF, Gibbs AH, et al. Species differences in the covalent binding of
[14C]tamoxifen to liver microsomes and the forms of cytochrome P450 involved.
Biochem Pharmacol 1995; 49(8):1035–1042.
13. http://www.imm.ki.se/CYPalleles/cyp2b6.htm (accessed January 2007). 2007. 7. Ref
Type: Electronic Citation

38 Amend et al.
14. Jinno H, Tanaka-Kagawa T, Ohno A, et al. Functional characterization of cytochrome
P450 2B6 allelic variants. Drug Metab Dispos 2003; 31(4):398–403.
15. Ariyoshi N, Miyazaki M, Toide K, et al. A single nucleotide polymorphism of
CYP2b6 found in Japanese enhances catalytic activity by autoactivation. Biochem
Biophys Res Commun 2001; 281(5):1256–1260.
16. Lang T, Klein K, Fischer J, et al. Extensive genetic polymorphism in the human
CYP2B6 gene with impact on expression and function in human liver. Pharmacogenetics
2001; 11(5):399–415.
17. Hiratsuka M, Takekuma Y, Endo N, et al. Allele and genotype frequencies of
CYP2B6 and CYP3A5 in the Japanese population. Eur J Clin Pharmacol 2002; 58(6):
417–421.
18. Lamba V, Lamba J, Yasuda K, et al. Hepatic CYP2B6 expression: gender and ethnic
differences and relationship to CYP2B6 genotype and CAR (constitutive androstane
receptor) expression. J Pharmacol Exp Ther 2003; 307(3):906–922.
19. Daly AK. Pharmacogenetics of the major polymorphic metabolizing enzymes.
Fundam Clin Pharmacol 2003; 17(1):27–41.
20. Rebbeck TR, Jaffe JM, Walker AH, et al. Modification of clinical presentation of
prostate tumors by a novel genetic variant in CYP3A4. J Natl Cancer Inst 1998;
90(16):1225–1229.
21. Walker AH, Jaffe JM, Gunasegaram S, et al. Characterization of an allelic variant
in the nifedipine-specific element of CYP3A4: ethnic distribution and implications
for prostate cancer risk. Mutations in brief no. 191. Hum Mutat 1998; 12(4):289
(online).
22. Amirimani B, Walker AH, Weber BL, et al. RESPONSE: re: modification of clinical
presentation of prostate tumors by a novel genetic variant in CYP3A4. J Natl Cancer
Inst 1999; 91(18):1588–1590.
23. Roy P, Yu LJ, Crespi CL, et al. Development of a substrate-activity based approach
to identify the major human liver P-450 catalysts of cyclophosphamide and ifosfamide
activation based on cDNA-expressed activities and liver microsomal P-450
profiles. Drug Metab Dispos 1999; 27(6):655–666.
24. Raucy JL. Regulation of CYP3A4 expression in human hepatocytes by pharmaceuticals
and natural products. Drug Metab Dispos 2003; 31(5):533–539.
25. Kuehl P, Zhang J, Lin Y, et al. Sequence diversity in CYP3A promoters and characterization
of the genetic basis of polymorphic CYP3A5 expression. Nat Genet
2001; 27(4):383–391.
26. Hustert E, Haberl M, Burk O, et al. The genetic determinants of the CYP3A5
polymorphism. Pharmacogenetics 2001; 11(9):773–779.
27. Griskevicius L, Yasar U, Sandberg M, et al. Bioactivation of cyclophosphamide: the
role of polymorphic CYP2C enzymes. Eur J Clin Pharmacol 2003; 59(2):103–109.
28. Morel F, Rauch C, Coles B, et al. The human glutathione transferase alpha locus:
genomic organization of the gene cluster and functional characterization of the
genetic polymorphism in the hGSTA1 promoter. Pharmacogenetics 2002; 12(4):
277–286.
29. Coles BF, Morel F, Rauch C, et al. Effect of polymorphism in the human glutathione
S-transferase A1 promoter on hepatic GSTA1 and GSTA2 expression. Pharmacogenetics
2001; 11(8):663–669.
30. Ali-Osman F, Akande O, Antoun G, et al. Molecular cloning, characterization, and
expression in Escherichia coli of full-length cDNAs of three human glutathione

Role of Genetic Variability in Breast Cancer Treatment Outcomes 39
S-transferase Pi gene variants. Evidence for differential catalytic activity of the
encoded proteins. J Biol Chem 1997; 272(15):10004–10012.
31. Harries LW, Stubbins MJ, Forman D, et al. Identification of genetic polymorphisms
at the glutathione S-transferase Pi locus and association with susceptibility to
bladder, testicular and prostate cancer. Carcinogenesis 1997; 18(4):641–644.
32. Srivastava SK, Singhal SS, Hu X, et al. Differential catalytic efficiency of allelic
variants of human glutathione S-transferase Pi in catalyzing the glutathione conjugation
of thiotepa. Arch Biochem Biophys 1999; 366(1):89–94.
33. Minotti G, Menna P, Salvatorelli E, et al. Anthracyclines: molecular advances and
pharmacologic developments in antitumor activity and cardiotoxicity. Pharmacol
Rev 2004; 56(2):185–229.
34. Ravid A, Rocker D, Machlenkin A, et al. 1,25-Dihydroxyvitamin D3 enhances the
susceptibility of breast cancer cells to doxorubicin-induced oxidative damage.
Cancer Res 1999; 59(4):862–867.
35. Kong Q, Lillehei KO. Antioxidant inhibitors for cancer therapy. Med Hypotheses
1998; 51(5):405–409.
36. Kalyanaraman B, Perez-Reyes E, Mason RP. Spin-trapping and direct electron spin
resonance investigations of the redox metabolism of quinone anticancer drugs.
Biochim Biophys Acta 1980; 630(1):119–130.
37. Conklin KA. Dietary antioxidants during cancer chemotherapy: impact on chemotherapeutic
effectiveness and development of side effects. Nutr Cancer 2000; 37(1):
1–18.
38. Sellin S, Holmquist B, Mannervik B, et al. Oxidation and reduction of 4-hydroxyalkenals
catalyzed by isozymes of human alcohol dehydrogenase. Biochemistry
1991; 30(9):2514–2518.
39. Lakhman SS, Ghosh D, Blanco JG. Functional significance of a natural allelic
variant of human carbonyl reductase 3 (CBR3). Drug Metab Dispos 2005; 33(2):
254–257.
40. Costantini P, Jacotot E, Decaudin D, et al. Mitochondrion as a novel target of
anticancer chemotherapy. J Natl Cancer Inst 2000; 92(13):1042–1053.
41. Shimoda-Matsubayashi S, Matsumine H, Kobayashi T, et al. Structural dimorphism
in the mitochondrial targeting sequence in the human manganese superoxide dismutase
gene. A predictive evidence for conformational change to influence mitochondrial
transport and a study of allelic association in Parkinson’s disease. Biochem
Biophys Res Commun 1996; 226(2):561–565.
42. Sutton A, Khoury H, Prip-Buus C, et al. The Ala16Val genetic dimorphism modulates
the import of human manganese superoxide dismutase into rat liver mitochondria.
Pharmacogenetics 2003; 13(3):145–157.
43. Ho YS, Crapo JD. Isolation and characterization of complementary DNAs encoding
human manganese-containing superoxide dismutase. FEBS Lett 1988; 229(2):
256–260.
44. Borgstahl GE, Parge HE, Hickey MJ, et al. Human mitochondrial manganese
superoxide dismutase polymorphic variant Ile58Thr reduces activity by destabilizing
the tetrameric interface. Biochemistry 1996; 35(14):4287–4297.
45. Forsberg L, Lyrenas L, de FU, et al. A common functional C-T substitution polymorphism
in the promoter region of the human catalase gene influences transcription
factor binding, reporter gene transcription and is correlated to blood catalase levels.
Free Radic Biol Med 2001; 30(5):500–505.

40 Amend et al.
46. Klebanoff SJ. Myeloperoxidase: occurrence and biological function. In: Everse J,
Everse KE, Grisham MB, eds. Peroxides in Chemistry. Boca Raton: CRC
Press, 1991.
47. Piedrafita FJ, Molander RB, Vansant G, et al. An Alu element in the myeloperoxidase
promoter contains a composite SP1-thyroid hormone-retinoic acid response
element. J Biol Chem 1996; 271(24):14412–14420.
48. Chabner BA, Ryan DP, Paz-Ares LP, et al. Antineoplastic agents. In: Hardman JG,
Limbird LE, Gilman AG, eds. Goodman and Gilman’s the Pharmacological Basis of
Therapeutics. New York, NY: McGraw-Hill, 2001:1389–1460.
49. Calvert H. An overview of folate metabolism: features relevant to the action and
toxicities of antifolate anticancer agents. Semin Oncol 1999; 26(2 suppl 6):3–10.
50. Robien K, Boynton A, Ulrich CM. Pharmacogenetics of folate-related drug targets in
cancer treatment. Pharmacogenomics 2005; 6(7):673–689.
51. Horie N, Aiba H, Oguro K, et al. Functional analysis and DNA polymorphism of the
tandemly repeated sequences in the 5 0 -terminal regulatory region of the human gene
for thymidylate synthase. Cell Struct Funct 1995; 20(3):191–197.
52. Pullarkat ST, Stoehlmacher J, Ghaderi V, et al. Thymidylate synthase gene polymorphism
determines response and toxicity of 5-FU chemotherapy. Pharmacogenomics
J 2001; 1(1):65–70.
53. Mandola MV, Stoehlmacher J, Muller-Weeks S, et al. A novel single nucleotide
polymorphism within the 5 0 tandem repeat polymorphism of the thymidylate synthase
gene abolishes USF-1 binding and alters transcriptional activity. Cancer Res
2003; 63(11):2898–2904.
54. Ulrich CM, Bigler J, Velicer CM, et al. Searching expressed sequence tag databases:
discovery and confirmation of a common polymorphism in the thymidylate synthase
gene. Cancer Epidemiol Biomarkers Prev 2000; 9(12):1381–1385.
55. Mandola MV, Stoehlmacher J, Zhang W, et al. A 6 bp polymorphism in the thymidylate
synthase gene causes message instability and is associated with decreased
intratumoral TS mRNA levels. Pharmacogenetics 2004; 14(5):319–327.
56. Danenberg PV. Pharmacogenomics of thymidylate synthase in cancer treatment.
Front Biosci 2004; 9:2484–2494.
57. Stern LL, Bagley PJ, Rosenberg IH, et al. Conversion of 5-formyltetrahydrofolic acid
to 5-methyltetrahydrofolic acid is unimpaired in folate-adequate persons homozygous
for the C677T mutation in the methylenetetrahydrofolate reductase gene.
J Nutr 2000; 130(9):2238–2242.
58. Weisberg I, Tran P, Christensen B, et al. A second genetic polymorphism in
methylenetetrahydrofolate reductase (MTHFR) associated with decreased enzyme
activity. Mol Genet Metab 1998; 64(3):169–172.
59. Toffoli G, Veronesi A, Boiocchi M, et al. MTHFR gene polymorphism and severe
toxicity during adjuvant treatment of early breast cancer with cyclophosphamide,
methotrexate, and fluorouracil (CMF). Ann Oncol 2000; 11(3):373–374.
60. Etienne MC, Lagrange JL, Dassonville O, et al. Population study of
dihydropyrimidine dehydrogenase in cancer patients. J Clin Oncol 1994; 12(11):
2248–2253.
61. van Kuilenburg AB, Haasjes J, Richel DJ, et al. Clinical implications of dihydropyrimidine
dehydrogenase (DPD) deficiency in patients with severe 5-
fluorouracil-associated toxicity: identification of new mutations in the DPD gene.
Clin Cancer Res 2000; 6(12):4705–4712.

Role of Genetic Variability in Breast Cancer Treatment Outcomes 41
62. van Kuilenburg AB, De Abreu RA, van Gennip AH. Pharmacogenetic and clinical
aspects of dihydropyrimidine dehydrogenase deficiency. Ann Clin Biochem 2003;
40(pt 1):41–45.
63. Collie-Duguid ES, Etienne MC, Milano G, et al. Known variant DPYD alleles do not
explain DPD deficiency in cancer patients. Pharmacogenetics 2000; 10(3):217–223.
64. Etienne MC, Ilc K, Formento JL, et al. Thymidylate synthase and methylenetetrahydrofolate
reductase gene polymorphisms: relationships with 5-fluorouracil
sensitivity. Br J Cancer 2004; 90(2):526–534.
65. Largillier R, Etienne-Grimaldi MC, Formento JL, et al. Pharmacogenetics of
capecitabine in advanced breast cancer patients. Clin Cancer Res 2006; 12(18):
5496–5502.
66. Laverdiere C, Chiasson S, Costea I, et al. Polymorphism G80A in the reduced folate
carrier gene and its relationship to methotrexate plasma levels and outcome of
childhood acute lymphoblastic leukemia. Blood 2002; 100(10):3832–3834.
67. Costea I, Moghrabi A, Laverdiere C, et al. Folate cycle gene variants and chemotherapy
toxicity in pediatric patients with acute lymphoblastic leukemia. Haematologica
2006; 91(8):1113–1116.
68. Seidemann K, Book M, Zimmermann M, et al. MTHFR 677 (C–>T) polymorphism
is not relevant for prognosis or therapy-associated toxicity in pediatric NHL:
results from 484 patients of multicenter trial NHL-BFM 95. Ann Hematol 2006;
85(5):291–300.
69. Jazbec J, Kitanovski L, Aplenc R, et al. No evidence of association of methylenetetrahydrofolate
reductase polymorphism with occurrence of second neoplasms
after treatment of childhood leukemia. Leuk Lymphoma 2005; 46(6):893–897.
70. Sohn KJ, Croxford R, Yates Z, et al. Effect of the methylenetetrahydrofolate
reductase C677T polymorphism on chemosensitivity of colon and breast cancer cells
to 5-fluorouracil and methotrexate. J Natl Cancer Inst 2004; 96(2):134–144.
71. Schiff PB, Fant J, Horwitz SB. Promotion of microtubule assembly in vitro by taxol.
Nature 1979; 277(5698):665–667.
72. Soyama A, Saito Y, Hanioka N, et al. Non-synonymous single nucleotide alterations
found in the CYP2C8 gene result in reduced in vitro paclitaxel metabolism. Biol
Pharm Bull 2001; 24(12):1427–1430.
73. Marsh S, Somlo G, Li X, et al. Pharmacogenetic analysis of paclitaxel transport and
metabolism genes in breast cancer. Pharmacogenomics J 2007.
74. van Schaik RH. Implications of cytochrome P450 genetic polymorphisms on the
toxicity of antitumor agents. Ther Drug Monit 2004; 26(2):236–240.
75. Crewe HK, Ellis SW, Lennard MS, et al. Variable contribution of cytochromes P450
2D6, 2C9 and 3A4 to the 4-hydroxylation of tamoxifen by human liver microsomes.
Biochem Pharmacol 1997; 53(2):171–178.
76. Dehal SS, Kupfer D. CYP2D6 catalyzes tamoxifen 4-hydroxylation in human liver.
Cancer Res 1997; 57(16):3402–3406.
77. Stearns V, Johnson MD, Rae JM, et al. Active tamoxifen metabolite plasma concentrations
after coadministration of tamoxifen and the selective serotonin reuptake
inhibitor paroxetine. J Natl Cancer Inst 2003; 95(23):1758–1764.
78. Nishiyama T, Ogura K, Nakano H, et al. Reverse geometrical selectivity in glucuronidation
and sulfation of cis- and trans-4-hydroxytamoxifens by human liver
UDP-glucuronosyltransferases and sulfotransferases. Biochem Pharmacol 2002;
63(10):1817–1830.

42 Amend et al.
79. Ingelman-Sundberg M. Genetic polymorphisms of cytochrome P450 2D6
(CYP2D6): clinical consequences, evolutionary aspects and functional diversity.
Pharmacogenomics J 2005; 5(1):6–13.
80. Falany CN, Wheeler J, Oh TS, et al. Steroid sulfation by expressed human cytosolic
sulfotransferases. J Steroid Biochem Mol Biol 1994; 48(4):369–375.
81. Raftogianis RB, Wood TC, Otterness DM, et al. Phenol sulfotransferase pharmacogenetics
in humans: association of common SULT1A1 alleles with TS PST phenotype.
Biochem Biophys Res Commun 1997; 239(1):298–304.
82. Nowell S, Sweeney C, Winters M, et al. Association between sulfotransferase 1A1
genotype and survival of breast cancer patients receiving tamoxifen therapy. J Natl
Cancer Inst 2002; 94(21):1635–1640.
83. Levesque E, Beaulieu M, Green MD, et al. Isolation and characterization of
UGT2B15(Y85): a UDP-glucuronosyltransferase encoded by a polymorphic gene.
Pharmacogenetics 1997; 7(4):317–325.
84. Nowell SA, Ahn J, Rae JM, et al. Association of genetic variation in tamoxifenmetabolizing
enzymes with overall survival and recurrence of disease in breast
cancer patients. Breast Cancer Res Treat 2005; 91(3):249–258.
85. Buzdar AU, Robertson JF, Eiermann W, et al. An overview of the pharmacology and
pharmacokinetics of the newer generation aromatase inhibitors anastrozole, letrozole,
and exemestane. Cancer 2002; 95(9):2006–2016.
86. Sioufi A, Gauducheau N, Pineau V, et al. Absolute bioavailability of letrozole in
healthy postmenopausal women. Biopharm Drug Dispos 1997; 18(9):779–789.
87. Lizard-Nacol S, Coudert B, Colosetti P, et al. Glutathione S-transferase M1 null
genotype: lack of association with tumour characteristics and survival in advanced
breast cancer. Breast Cancer Res 1999; 1(1):81–87.
88. Sweeney C, Ambrosone CB, Joseph L, et al. Association between a glutathione
S-transferase A1 promoter polymorphism and survival after breast cancer treatment.
Int J Cancer 2003; 103(6):810–814.
89. Sweeney C, McClure GY, Fares MY, et al. Association between survival after
treatment for breast cancer and glutathione S-transferase P1 Ile105Val polymorphism.
Cancer Res 2000; 60(20):5621–5624.
90. Ambrosone CB, Sweeney C, Coles BF, et al. Polymorphisms in glutathione
S-transferases (GSTM1 and GSTT1) and survival after treatment for breast cancer.
Cancer Res 2001; 61(19):7130–7135.
91. Yang G, Shu XO, Ruan ZX, et al. Genetic polymorphisms in glutathione-S-transferase
genes (GSTM1, GSTT1, GSTP1) and survival after chemotherapy for invasive
breast carcinoma. Cancer 2005; 103(1):52–58.
92. Shrubsole MJ, Shu XO, Ruan ZX, et al. MTHFR genotypes and breast cancer survival
after surgery and chemotherapy: a report from the Shanghai Breast Cancer
Study. Breast Cancer Res Treat 2005; 91(1):73–79.
93. Martin DN, Boersma BJ, Howe TM, et al. Association of MTHFR gene polymorphisms
with breast cancer survival. BMC Cancer 2006; 6:257.
94. Wegman P, Elingarami S, Carstensen J, et al. Genetic variants of CYP3A5,
CYP2D6, SULT1A1, UGT2B15 and tamoxifen response in postmenopausal patients
with breast cancer. Breast Cancer Res 2007; 9(1):R7.
95. Goetz MP, Rae JM, Suman VJ, et al. Pharmacogenetics of tamoxifen biotransformation
is associated with clinical outcomes of efficacy and hot flashes. J Clin
Oncol 2005; 23(36):9312–9318.

Role of Genetic Variability in Breast Cancer Treatment Outcomes 43
96. Ambrosone CB, Ahn J, Singh KK, et al. Polymorphisms in genes related to oxidative
stress (MPO, MnSOD, CAT) and survival after treatment for breast cancer. Cancer
Res 2005; 65(3):1105–1111.
97. Martin RC, Ahn J, Nowell SA, et al. Association between manganese superoxide
dismutase promoter gene polymorphism and breast cancer survival. Breast Cancer
Res 2006; 8(4):R45.
98. Petros WP, Hopkins PJ, Spruill S, et al. Associations between drug metabolism
genotype, chemotherapy pharmacokinetics, and overall survival in patients with
breast cancer. J Clin Oncol 2005; 23(25):6117–6125.

4
The Role of Epigenetics in Breast Cancer:
Implications for Diagnosis, Prognosis,
and Treatment
Amy M. Dworkin, Tim H.-M. Huang, and Amanda E. Toland
Human Cancer Genetics Program, Department of Molecular Virology,
Immunology, and Medical Genetics, The Ohio State University,
Columbus, Ohio, U.S.A.
INTRODUCTION
In the past 10 years, the field of epigenetics has expanded into an essential
component of clinical cancer research and therapeutics. Epigenetic modifications
in DNA or chromatin are stably transmitted over rounds of cell division but do
not result in any permanent alterations to the underlying DNA sequence. These
changes, such as histone modifications and DNA methylation of tumor suppressor
genes, are considered a hallmark of certain cancers, including breast
cancer. Unlike germline mutations, epigenetic modifications can potentially be
reversed, making them very appealing for preventative care and therapeutics
against cancers (1). Because of this, demethylating agents are currently being
evaluated for efficacy in the treatment of breast cancer.
There are distinct mechanisms that initiate and sustain epigenetic modifications
(2–8). Of these, DNA methylation and posttranslational modifications
of histone proteins are the best understood. DNA methylation is a covalent
addition of a methyl group to DNA, usually to a cytosine located 5 0 to guanasine
(CpG dinucleotide). CpG dinucleotides are underrepresented in the genome,
45

46 Dworkin et al.
Figure 1 Methylation of CpG islands. Unmethylated CpG islands (in light gray) allow
for transcription, while methylated CpG islands (in dark gray) repress transcription.
except for small clusters, referred to as CpG islands, located in or near the
promoter of approximately half of all genes (1,9–11). In addition, other epigenetic
modifications have recently been explored, including the Polycomb
group (PcG) proteins, which repress gene function and can only be overcome by
germline differentiation processes, and small noncoding RNA molecules, which
regulate gene expression by targeting RNA degradation (3). Small noncoding
RNA molecules have recently been found to also target gene promoters and
result in transcriptional gene silencing (11,12).
DNA methylation plays a variety of roles in normal cells: silencing of
transposable elements and viral sequences, maintenance of chromosomal
integrity, X chromosome inactivation, and transcriptional regulation of genes
(1). Methylation is established and maintained by at least four DNA methyltransferases
(DNMTs), DNMT1, DNMT3A, DNMT3B, and DNMT3L, which
catalyze the covalent addition of methyl groups to cytosine in the CpG dinucleotide.
Methylation of CpG islands causes transcriptional inactivation (Fig. 1).
These methyltransferases and their isoforms have been demonstrated to be
essential for development; a knockout of any of these enzymes in mouse
embryos results in embryonic or perinatal lethality (1). Functionally, there are
two types of methyltransferases, maintenance and de novo. Maintenance methyltransferases
are essential for copying DNA methylation patterns during cell
division, and de novo methyltransferases introduce new patterns of methylation
during early development. There are duplications of functions between these
methylases so that perturbation of both types may be necessary to change overall
methylation patterns (6). DNMT1 functions as a maintenance methyltransferase
by copying DNA methylation patterns to daughter strands, following replication.
It is the most abundant and catalytically active form. DNMT3A and DNMT3B
act on both unmethylated and hemimethylated DNA. However, they are thought
to have different methylation targets depending on the cell type and stage of
development (13). In vitro knockouts of DNMT3a or DNMT3b each exhibit
distinct phenotypes. DNMT3a-null cells lack methylation at imprinting loci,
while DNMT3b-null cells result in loss of DNA methylation on endogenous

The Role of Epigenetics in Breast Cancer 47
Figure 2 Histone modification and chromatin remodeling. Histone modifications, like
acetylation, change the affinity of the histones to DNA. The acetylated histones on the left
are in an open euchromatin structure and can be transcribed. The unacetylated histones on
the right are in a heterochromatin structure and repress transcription. Source: Cartoon
compliments of Kristi Bennett.
C-type retroviral DNA. Deletion of both DNMT3a and DNMT3b abolishes de
novo methylation, while DNMT1 deletion produces bulk DNA demethylation
(7). Of the methyltransferases, DNMT3L is unique. It stimulates de novo
methylation by DNMT3a, does not enhance its binding to DNA, and alone does
not bind to DNA (14).
Along with methylation, posttranslational modification of histones plays
an important role in epigenetics. Histone proteins associate with DNA to form
nucleosomes, permitting large amounts of DNA to be neatly packaged into the
nucleus. There are two configurations of chromatin, heterochromatin and
euchromatin. Heterochromatin is condensed and transcriptionally inactive, while
euchromatin has an “open” configuration and is favorable for gene transcription
(Fig. 2). The N-terminal tails of histones “stick out” of the nucleosome and are
subject to posttranslational modification such as acetylation, phosphorylation,
methylation, ubiquitination, and sumoylation (2,4,7,15). The pattern of histone
modification creates a “code,” which is read by proteins involved in chromatin
remodeling and the dynamics of gene transcription (6,11). Acetylation and
deacteylation are the most common types of histone modification. Histone
deacetylases (HDACs) are a class of enzymes that remove acetyl groups from a
e-N-acetyl lysine amino acid on a histone. Its action is opposite to that of histone
acetyltransferases (HATs). Deacetylation removes acetyl groups from histone
tails, causing the histones to wrap more tightly around the DNA and interfering
with transcription by blocking access to transcription factors. The overall result
of histone deacetylation is a global (nonspecific) reduction in gene expression
(Table 1).
METHYLATION IN CANCER
Cancer development is complex due to progressive changes of the genome that
result in altering normal cellular processes such as cell growth, apoptosis, and
angiogenesis (12). Both genetic and epigenetic modifications play a strong role
in tumorigenesis (7). In classic tumor biology, cells lose both copies of tumor

48 Dworkin et al.
Table 1
Histone Modifiers
Histone modifier Mechanism Catalysis Role in transcription
Acetylation
Deacetylation
Phosphorylation
Dephosphorylation
Methylation
Demethylation
Ubiquitination
Deubiquitination
Sumoylation
Introduction of the
acetyl group to
the lysine residue
at the N terminus
of the histone
protein
Removal of the
acetyl group
from the lysine
residue at the N
terminus of the
histone protein
Addition of
phosphate group
Removal of
phosphate group
Addition of 1, 2, or
3 methyl groups
on either lysine
or arginine
residues
Removal of methyl
groups
Covalent
attachment
of ubiquitin
Removal of
ubiquitin
Covalent
attachment
of SUMO to
lysine residues
HATs
HDACs
Histone kinases
Histone
phosphatases
HMTs
Cytochrome P450
family of liver
enzymes
Enzymatic cascade,
including
ubiquitin ligase,
E1 ubiquitin
activating
enzyme, and
ATP
Isopeptidases
Enzymatic cascade,
including SUMOactivating
enzyme
and ATP
Enhances transcription
(removes positive
charge and reduces
affinity between
histones and DNA)
Represses transcription
Enhances transcription
(neutralizes histones’
basic charge and
reduces affinity to
DNA)
Represses transcription
Silences genes
Activates genes
Labels protein for
degradation
Reverses
ubiquitination
Can mediate gene
silencing through
recruitment of HDAC
and heterochromatin
protein 1
Abbreviations: HATs, histone acetyltransferases; HDACs, histone deacetylases; HMT, histone
methyltranferases; SUMO, small ubiquitin-like modifier; ATP, adenosine triphosphate.

The Role of Epigenetics in Breast Cancer 49
suppressor genes through mutation or deletion. Epigenetic modifications, such
as DNA methylation, serve a similar role to somatic mutations, leading to
gene silencing. Overall, tumors tend to exhibit hypomethylation of repeat
elements and pericentromeric regions and hypermethylation of normally
unmethylated CpG islands. Gene-specific hypermethylation of CpG islands is
associated with chromosomal instability, resulting in increased mutation rates
and abnormal gene expression. For example, hypermethylation of critical
genes, such as tumor suppressors, adhesion molecules, DNA repair genes,
inhibitors of angiogenesis, and inhibitors of metastasis, is observed in
tumorigenesis (7). Hypermethylation of DNA repair genes (e.g., hMLH1) is
associated with genomic instability through disruption of chromosome replication
control (7). Unlike gene mutations, DNA methylation can be reversed
by DNMT inhibitors.
An important component of the cancer epigenome is an altered DNA
methylation pattern of global demethylation and localized promoter hypermethylation.
Aberrant methylation patterns are seen in almost every cancer type
(7,16). In addition to aberrant methylation, tumors also show altered expression
of HATs, making them another potential target for therapy (11). It is believed
that histone modifications precede promoter hypermethylation to initiate transcriptional
silencing of tumor suppressor genes. Repressive histone modifications
are first established. Over time the density of DNA methylation
increases in the gene promoter regions, which creates a “permanent” mode of
silencing (17).
THE ROLE OF EPIGENETICS IN BREAST CANCER
In the United States, breast cancer is one of the most common cancers, with
approximately 200,000 women diagnosed with breast cancer each year (http://
www.cancer.org). In addition to well-understood DNA mutations in breast
tumors, epigenetic alterations of gene expression are key contributors (16,18–24).
A large number of genes are found to be hypermethylated in breast tumors but
not in normal breast tissue. Many of these genes have well-characterized functions
in tumor development. Genes commonly methylated in breast cancer and
their potential roles in tumor development are listed in Table 2. Four key genes,
RASSF1A, BRCA1, ESR1, and ESR2, are further discussed in detail.
RASSF1A
The CpG island of the Ras-associated domain family 1A gene (RASSF1) is
hypermethylated in 60% to 77% of breast cancers, and its transcripts are frequently
inactivated in cancer cell lines and primary tissues (13,22). When cells
lacking RASSF1A expression are treated with a DNMT inhibitor, such as 5-aza-
2 0 -deoxycytidine, expression can be reactivated (25). RASSF1A is involved in

50 Dworkin et al.
Table 2
Genes Methylated in Breast Cancer
Gene Cancer hallmark contribution Reference
RASSF1A Apoptosis 15
CYP26A1 Microenvironment, encodes for cytochrome p450 enzymes 17
KCNAB1 Microenvironment 17
SNCA Microenvironment 17
RARb Growth regulation 15
TWIST Cell lineage and differentiation 15
Cyclin D2 Cell cycle regulation 15
APC Cell cycle regulation and apoptosis 8
BRCA1 Cell cycle regulation 8,15
BRCA2 Tumor suppressor gene 15
p16 Cell cycle regulation 8,15
DAPK Apoptosis 8
GSTP1 Cellular metabolism and carcinogen detoxification 8
TMS1 Apoptosis 8
PR Growth regulation 8
CDH1 Angiogenesis 8
WWOX Apoptosis 38
growth regulation and apoptotic pathways. Specifically, RASSF1A is a Ras
effector and induces apoptosis through its interactions with pro-apoptotic kinase
MST1. Mouse knockout studies show that RASSF1a –/– mice are prone to
spontaneous development of lung adenomas, lymphomas, and breast adenocarcinomas.
Knockout RASSF1A mice are prone to early spontaneous tumorigenesis
and show a severe tumor susceptibility phenotype compared to wild-type
mice (25). In addition to breast tumors, hypermethylation of RASSF1A can be
detected in nonmalignant breast cells and patient sera. In one study, RASSF1A
hypermethylation in sera of breast cancer patients was detected in six out of 26
cases. These data suggest that RASSF1A hypermethylation may be a promising
biomarker to screen putative cancer patients (25).
BRCA1
Germline mutations of BRCA1 and BRCA2 are responsible for familial breast
cancers. Somatic mutations in sporadic cases are rare, but chromosomal losses
occur in 30% to 50% of sporadic tumors, respectively (26). BRCA1 acts as a
tumor suppressor gene for both breast and ovarian cancer (16). It encodes a
multifunctional protein involved in DNA repair, cell cycle checkpoint control,
protein ubiquitinylation, and chromatin remodeling (19). In vitro studies showed
that decreased BRCA1 expression in cells led to increased levels of tumor
growth, while increased expression of BRCA1 led to growth arrest and apoptosis.
Inactivation of BRCA1 by promoter methylation is seen in 7% to 31% of

The Role of Epigenetics in Breast Cancer 51
sporadic breast cancer cases (27). It is believed that aberrant methylation pairs
with loss of heterozygosity to reduce BRCA1 expression in invasive sporadic
breast tumors (16,19,22,26,28). The magnitude of the decrease of functional
BRCA1 protein correlates with disease prognosis (19,22). Phenotypically,
BRCA1-methylated tumors are similar to tumors with germline mutations.
Higher-grade tumors tend to show complete loss of BRCA1 protein. In contrast to
BRCA1, BRCA2 does not show as high a degree of promoter methylation (16).
ESR1 and ESR2
Along with other genes, the estrogen receptors (ERs) ERa and ERb have been
implicated in breast cancer development. ERa is encoded by ESR1, and when
estrogen is activated, it stimulates cell proliferation. ERb is encoded by ESR2
and is known to inhibit the proliferation and invasion of breast cancer cells. ERa
and ERb both have promoter-associated CpG islands that can be abnormally
methylated in breast cancer, but ESR2 methylation has been less well studied
(16,18). Almost all breast cancers show some degree of DNA methylation of the
ESR1 gene promoter, but this methylation is only associated with gene repression
in ~30% of tumors (16,18). Approximately 66% of breast cancers express
ERa. A fraction of breast cancers that are initially ERa positive lose ER
expression during tumor progression, but it is unclear if this is due to methylation
or other causes (18).
The presence or absence of ERs plays an important role in both therapy
and survival (16,18). ER-positive breast cancer can be treated with antiestrogenic
drugs, such as tamoxifen, which binds to the ER, preventing estrogen from
binding. This results in a decrease in growth of ER-positive breast cancer cells.
ER-negative cells are no longer responsive to estrogen; therefore, antiestrogenic
drugs have no effect (18). Reversal of ER-negative status is an attractive target
for breast cancer therapy.
BIOMARKERS AND DETECTION
Gene-specific epigenetic changes for breast cancer are likely to occur early in
tumorigenesis and have potential to be used as early detection and prevention
methods (2). As epigenetic modifications become clear biomarkers for breast
cancer treatment and therapeutic intervention, it is important to understand the
many different techniques available for studying and detecting the presence of
methylation, histone modifications, and microRNAs (11).
New, promising high-throughput methylation detection methods are
available, which enable researchers and clinicians to identify an “epigenetic
signature” specific to breast cancer. Biomarkers based on unique DNA methylation
and histone modification patterns of breast tumors may be used in the
future for diagnosis and early detection. Currently, breast cancer detection relies
on various screening methods. Cytology is the main standard for identification

52 Dworkin et al.
of abnormalities typical of cellular transformation. Quantitative multiplex
methylation-specific polymerase chain reaction (PCR) (QM-MSP) is a highly
sensitive method to quantitate cumulative gene promoter hypermethylation in
samples where DNA is limiting. First, gene-specific primer pairs are used to
coamplify genes, independent of their methylation status. Next, gene-specific
primers (methylated and unmethylated) conjugated with fluorescent labels are
combined with the first PCR product. The signal generated is proportional to the
extent of DNA amplification (15). Since it is known that multigene methylation
of CpG islands is common in early breast cancer, a panel of nine genes were
tested for their sensitivity and specificity of detecting premalignant changes
using QM-MSP. These genes included RASSF1A, RARb, TWIST, HIN1, Cyclin
D2, APC, BRCA1, BRCA2, and p16. Tests were done using ductal lavage, nipple
aspiration fluids, and fine needle aspirates.
Using this detection panel, the rate of detection of breast cancer cells rises
from 43% sensitivity with cytologic examination alone to 71.4% sensitivity (15).
Hypermethylation of genes that are commonly methylated in breast cancer in
sera in breast cancer patients has also been used to detect early malignant
changes. Sera methylation can be detected using methylation-specific PCR (MS-
PCR), a noninvasive PCR-based technique. Early studies indicate that MS-PCR
is a promising approach to screen putative cancer patients (25). This technique
utilizes primers designed for methylated or unmethylated bisulfite-modified
DNA (29). RASSF1A gene hypermethylation has been detected using MS-PCR
in sera of ovarian cancer patients, with 100% specificity (25).
Recurrence rates for breast cancer in the same breast have been recorded as
high as 40% despite negative pathological margins. Researchers believe that a
primary tumor may serve as a locus from which methylation density progressively
diffuses outward to surrounding tissues. CYP26A1, KCNAB1, RASSF1A,
and SNCA have found to be frequently methylated in adjacent normal tissue. A
methylation profile containing genes such as these, or those identified in other
studies (Table 2), could be used by clinicians to determine whether patients
should undergo radiological therapy or other treatments to prevent future
recurrence (17).
MassARRAY and pyrosequencing, two additional methods of detecting
methylation, are currently being developed for use in clinical settings.
MassARRAY is a highly sensitive technique that uses base-specific cleavage and
matrix-assisted laser desorption/ionization time-to-flight mass spectrometry
(MALDI-TOF MS). It is capable of detecting DNA methylation levels as low as
5%. It is suitable for testing methylation patterns on various sources, including
archival tissues and laser capture microdissected specimens (1). Pyrosequencing
is a locus-specific quantitative method that utilizes the detection of pyrophosphate,
which is liberated from incorporated nucleotides by DNA polymerase
during strand elongation. Free pyrophosphates are converted to adenosine triphosphate
(ATP), which provides energy for the oxidation of luciferin to then
generate light. Nucleotides are added sequentially to enable base calling.

The Role of Epigenetics in Breast Cancer 53
Pyrosequencing has two major advantages over MS-PCR. First, the data are
actual nucleotide sequences rather than fluorescence data. Second, pyrosequencing
can detect partially methylated sequences that are outside of the priming
sites (30).
Commonly used methods for detecting histone modifications usually rely
on chromatin immunoprecipitation (ChIP). In ChiP, DNA is cross-linked to
chromatin, followed by co-immunoprecipitation using antibodies against specific
histone residues or chromatin proteins. “ChIP” DNA can be used in microarrays,
cloning, and sequencing (11). Techniques for detecting histone modifications
have not yet been adapted for clinical application.
CLINICAL APPLICATION
Breast cancer prevention, treatments, and diagnostics are being developed to
target epigenetic changes leading to breast cancer. Treatments for breast cancer
currently being studied extensively focus on reversing aberrant DNA methylation
and histone acetylation of tumor suppressor genes and genes involved in
therapeutic response. Combinations of epigenetic-targeted therapies with conventional
chemotherapeutic agents may provide a way to resensitize drugresistant
tumors (Table 3).
Methylation Therapy
DNA methylation in proliferating cells is dependent on continued expression of
DNMTs. Treatments that focus on inhibiting the expression of these proteins
would therefore result in progressive reduction in DNA methylation in newly
divided cells. There are three main classes of DNA methylation inhibitors:
nucleoside inhibitors, nonnucleoside weak inhibitors, and rationally designed
inhibitors. Nucleoside inhibitors trap DNMTs for degradation. However, they
require DNA incorporation and active DNA synthesis, which limits the drug in
hypoproliferating cells. 5-Azacytidine (5-aza-dC) and its deoxy analog are two
of the first hypomethylating agents approved by the U.S. Food and Drug
Administration for cancer treatment (31). They work by two mechanisms: (i) as
demethylators by incorporating into CpG sites opposite a methylated CpG site on
the template strand and (ii) by binding to the cytosine in the catalytic domain of
the DNMT1 enzyme. Within a few hours of treatment with 5-aza-dC, there is
rapid passive loss of methylation (32). Nonnucleoside inhibitors are not as well
understood and are limited by their low level of hypomethylation induction.
Nonnucleoside inhibitors were first discovered for their use in reducing autoreactivity
for autoimmune diseases. During this study researchers discovered their
potential as demethylating agents. Researchers are currently using clinically
approved nonnucleoside inhibitors as a starting point to study their use as
demethylating agents. Rationally designed inhibitors are also in development.
These drugs are designed on the basis of knowledge of specific chemical

54 Dworkin et al.
Table 3
Drugs in Use for Breast Cancer Treatment
Nucleoside inhibitors
Drug Use Research level Reference
5-Azacytidine Trap DNMTs for degradation Clinical use 31,40
5-aza-
Trap DNMTs for degradation Clinical use 31
2 0 deoxycytidine
Zebularine Unknown mechanism Shown promise in vitro 31
5 0 -fluoro-
2 0 deoxycytidine
Unknown mechanism In clinical trials 31
Nonnucleoside inhibitors
Drug Use Research level Reference
Procainamide
Hydralazine
Mechanism not well
understood but proven to
decrease expression of
DNMT1 and DNMT3a
Mechanism not well
understood but proven to
decrease expression of
DNMT1 and DNMT3a
Rationally designed inhibitors
Clinically used as
an antihypertensive
drug
Clinically used as a
vasodilator
31,41,42
31,41,42
Drug Use Research level Reference
NSC303530 Blocks the active site of DNMTs Basic research 31,43
NSC401077 Blocks the active site of DNMTs Basic research 31,43
HDAC inhibitors
Drug Use Advantages Disadvantages
Research
level
Reference
LAQ824
Magnesium
Valporate
and
Hydralazine
HDAC
inhibitor
N-hydroxy-2- HDAC
propenamide inhibitor
Unknown
HDAC and Increases
methyltransferase
conventional
efficacy of
inhibitor cytotoxic
agents
Suppresses
mammary
tumor growth
in cells
Important to
pay attention
to schedule of
administration
Unknown
Unknown
Preclinical
level; studies
on hematological
malignancies
Phase III
clinical
trials
Research
level
40
44
l45
Abbreviations: DNMTs, DNA methyltransferases; HDAC, histone deacetylase.

The Role of Epigenetics in Breast Cancer 55
Figure 3 Drug impact on histone modifications. HDAC modifies histones and leads to
chromatin remodeling so that genes are transcriptionally inactive. HAT can reverse the
effects of HDAC to acetylate histones and open the confirmation into an active state.
Drugs such as Scriptaid, TSA, LAQ824, and N-hydroxy-2-propenamide can act on this
pathway to block HDACs and keep DNA transcriptionally active. Abbreviations: HDAC,
histone deactylase; HAT, histone acetyltransferase; TSA, trichostatin A. Source: Cartoon
compliments of Kristi Bennett.
responses in the body, tailoring combinations of these to fit a treatment profile.
An example of a rational design involves the use of three-dimensional information
to target the drug. Limitations of DNA methylation inhibitors include the
cooperation between different DNMTs such that the drugs must inhibit several of
them simultaneously. All demethylating agents suffer from nonspecificity and
toxicity. The next phase of drug development is how to target gene-specific
hypomethylation (e.g., unmethylated oligonucleotides) (31).
Histone Modification Therapy
Histone deacetylases (HDACs) are also targets for therapy (Fig. 3). HDACs are
important because of their role in cell cycle arrest and inhibition of apoptosis.
These drugs have been studied in isolation and synergistically with DNMTs.
Therapies that target both DNMTs and HDACs act globally, which can cause
reactivation of other genes that are typically silenced, such as oncogenes (18,33–37).
ER Therapy
Drugs are in development for different types of breast cancer. A therapy that could
restore gene expression of ERS1 (for ER-negative cancer patients) could reestablish
cancer cell growth regulation through estrogen. After reexpression of
ERS1, antiestrogenic drugs could subsequently be used. A number of drugs,
including DNMT inhibitors and HDAC inhibitors, are being used to reactivate ER
expression (Table 4). DNMTs and HDACs work synergistically to silence gene
expression of ER. One problem found in treatments that target DNMTs and HDACs
is that alone they may not be enough to reverse ER-promoter hypermethylation.

56 Dworkin et al.
Table 4
Drugs Involved in Reactivation of ER Expression
Treatment Result Reference
5-aza-Dc
(methyltransferase
inhibitor)
Antisense oligonucleotides
(for DNMT1)
Partial demethylation of ER CpG islands,
reexpression of ER mRNA, and synthesis
of functional ER protein
ER gene reexpression and estrogen
responsiveness reestablished
18,36,37,46
18,35
TSA (HDAC inhibitor) Reexpression of ERa protein 18,33
DNMT and HDAC Reactivation ERa protein and expression 18,37
inhibitor
Scriptaid (HDAC inhibitor) Reexpression of functional ER in vivo 18,34
Scriptaid and 5-aza-Dc More effective in inducing ER expression
than just Scriptaid or 5-aza-Dc alone
18,34
Abbreviations: ER, estrogen receptor; DNMT, DNA methyltransferase; HDAC, histone deacetylase;
TSA, trichostatin A.
Tamoxifen is a commonly used drug for ERa-positive breast cancers.
Unfortunately, acquired resistance of tamoxifen occurs in nearly 40% of patients.
Therefore, it is important to identify patients most likely to respond to
Tamoxifen and to identify individuals likely to acquire resistance. One promising
biomarker for Tamoxifen response is WWOX. Preliminary studies show
that WWOX expression levels predict tamoxifen resistance better than the two
previously known biomarkers, PR and HER2 (38). WWOX appears to mediate
tamoxifen sensitivity, and its expression is reduced in a large fraction (63.2%) of
breast cancers (39). The primary mechanism of downregulation of WWOX is
through DNA hypermethylation of its regulatory region.
FUTURE DIRECTIONS
The next 10 years are likely to bring significant advances for breast cancer
treatment. Therapeutics that target methylation and histone modifications may
play a leading role in treating breast cancer. Since epigenetic modifications can
also be used as biomarkers, targeted therapies may some day be used as preventative
measures.
REFERENCES
1. Brena RM, Huang TH, Plass C. Quantitative assessment of DNA methylation:
potential applications for disease diagnosis, classification, and prognosis in clinical
settings. J Mol Med 2006; 84(5):365–377.
2. Balch C, Montgomery JS, Paik HI, et al. New anti-cancer strategies: epigenetic
therapies and biomarkers. Front Biosci 2005; 10:1897–1931.
3. Luczak MW, Jagodzinski PP. The role of DNA methylation in cancer development.
Folia Histochem Cytobiol 2006; 44(3):143–154.

The Role of Epigenetics in Breast Cancer 57
4. Miyamoto K, Ushijima T. Diagnostic and therapeutic applications of epigenetics. Jpn
J Clin Oncol 2005; 35(6):293–301.
5. Reamon-Buettner SM, Borlak J. A new paradigm in toxicology and teratology:
altering gene activity in the absence of DNA sequence variation. Reprod Toxicol
2007; 24(1):20–30.
6. Szyf M. Therapeutic implications of DNA methylation. Future Oncol 2005; 1(1):
125–135.
7. Ting AH, McGarvey KM, Baylin SB. The cancer epigenome—components and
functional correlates. Genes Dev 2006; 20(23):3215–3231.
8. Widschwendter M, Jones PA. DNA methylation and breast carcinogenesis. Oncogene
2002; 21(35):5462–5482.
9. Hellebrekers DM, Griffioen AW, van Engeland M. Dual targeting of epigenetic
therapy in cancer. Biochim Biophys Acta 2007; 1775(1):76–91.
10. Hellebrekers DM, Jair KW, Viré E, et al. Angiostatic activity of DNA methyltransferase
inhibitors. Mol Cancer Ther 2006; 5(2):467–475.
11. Curt Balch TH-MH, Nephew KP. High-throughput assessments of epigenetics in
human disease. In: Kim S, Tang H, Mardis ER, eds. Advances in Genome Sequencing
Technology and Algorithms, Norwood, MA: Artech House Publishers, 2007.
12. Han J, Kim D, Morris KV. Promoter-associated RNA is required for RNA-directed
transcriptional gene silencing in human cells. Proc Natl Acad Sci U.S.A. 2007; 104(30):
12422–12427.
13. Oka M, Rodic N, Graddy J, et al. CpG sites preferentially methylated by Dnmt3a in
vivo. J Biol Chem 2006; 281(15):9901–9908.
14. Ooi SK, Qiu C, Bernstein E, et al. DNMT3L connects unmethylated lysine 4 of
histone H3 to de novo methylation of DNA. Nature 2007; 448(7154):714–717.
15. Fackler MJ, Malone K, Zhang Z, et al. Quantitative multiplex methylation-specific
PCR analysis doubles detection of tumor cells in breast ductal fluid. Clin Cancer Res
2006; 12(11 Pt 1):3306–3310.
16. Campan M, Weisenberger DJ, Laird PW. DNA methylation profiles of female steroid
hormone-driven human malignancies. Curr Top Microbiol Immunol 2006;
310:141–178.
17. Yan PS, Venkataramu C, Ibrahim A, et al. Mapping geographic zones of cancer risk
with epigenetic biomarkers in normal breast tissue. Clin Cancer Res 2006; 12(22):
6626–6636.
18. Giacinti L, Claudio PP, Lopez M, et al. Epigenetic information and estrogen receptor
alpha expression in breast cancer. Oncologist 2006; 11(1):1–8.
19. Mirza S, Sharma G, Prasad CP, et al. Promoter hypermethylation of TMS1, BRCA1,
ERalpha and PRB in serum and tumor DNA of invasive ductal breast carcinoma
patients. Life Sci 2007; 81(4):280–287.
20. Sharma D, Blum J, Yang X, et al. Release of methyl CpG binding proteins and histone
deacetylase 1 from the Estrogen receptor alpha (ER) promoter upon reactivation in
ER-negative human breast cancer cells. Mol Endocrinol 2005; 19(7):1740–1751.
21. Sui M, Huang Y, Park BH, et al. Estrogen receptor alpha mediates breast cancer cell
resistance to paclitaxel through inhibition of apoptotic cell death. Cancer Res 2007;
67(11):5337–5344.
22. Vincent-Salomon A, Ganem-Elbaz C, Manié E, et al. X inactive-specific transcript
RNA coating and genetic instability of the X chromosome in BRCA1 breast tumors.
Cancer Res 2007; 67(11):5134–5140.

58 Dworkin et al.
23. Visvanathan K, Sukumar S, Davidson NE. Epigenetic biomarkers and breast cancer:
cause for optimism. Clin Cancer Res 2006; 12(22):6591–6593.
24. Zhou Q, Davidson NE. Silencing estrogen receptor alpha in breast cancer cells.
Cancer Biol Ther 2006; 5(7):848–849.
25. Pfeifer GP, Dammann R. Methylation of the tumor suppressor gene RASSF1A in
human tumors. Biochemistry (Mosc) 2005; 70(5):576–583.
26. Esteller M, Silva JM, Dominguez G, et al. Promoter hypermethylation and BRCA1
inactivation in sporadic breast and ovarian tumors. J Natl Cancer Inst 2000; 92(7):
564–569.
27. Wei M, Grushko TA, Dignam J, et al. BRCA1 promoter methylation in sporadic
breast cancer is associated with reduced BRCA1 copy number and chromosome 17
aneusomy. Cancer Res 2005; 65(23):10692–10699.
28. Catteau A, Morris JR. BRCA1 methylation: a significant role in tumour development?
Semin Cancer Biol 2002; 12(5):359–371.
29. Herman JG, Graff JR, Myöhänen S, et al. Methylation-specific PCR: a novel PCR
assay for methylation status of CpG islands. Proc Natl Acad Sci U.S.A. 1996; 93(18):
9821–9826.
30. Shames DS, Minna JD, Gazdar AF. Methods for detecting DNA methylation in
tumors: from bench to bedside. Cancer Lett 2007; 251(2):187–198.
31. Issa JP. DNA methylation as a therapeutic target in cancer. Clin Cancer Res 2007; 13(6):
1634–1637.
32. Christman JK. 5-Azacytidine and 5-aza-2 0 -deoxycytidine as inhibitors of DNA
methylation: mechanistic studies and their implications for cancer therapy. Oncogene
2002; 21(35):5483–5495.
33. Kawai H, Li H, Avraham S, et al. Overexpression of histone deacetylase HDAC1
modulates breast cancer progression by negative regulation of estrogen receptor
alpha. Int J Cancer 2003; 107(3):353–358.
34. Keen JC, Yan L, Mack KM, et al. A novel histone deacetylase inhibitor, scriptaid,
enhances expression of functional estrogen receptor alpha (ER) in ER negative
human breast cancer cells in combination with 5-aza 2 0 -deoxycytidine. Breast Cancer
Res Treat 2003; 81(3):177–186.
35. Yan L, Nass SJ, Smith D, et al. Specific inhibition of DNMT1 by antisense oligonucleotides
induces re-expression of estrogen receptor-alpha (ER) in ER-negative
human breast cancer cell lines. Cancer Biol Ther 2003; 2(5):552–556.
36. Yang X, Ferguson AT, Nass SJ, et al. Transcriptional activation of estrogen receptor
alpha in human breast cancer cells by histone deacetylase inhibition. Cancer Res
2000; 60(24):6890–6894.
37. Yang X, Phillips DL, Ferguson AT, et al. Synergistic activation of functional
estrogen receptor (ER)-alpha by DNA methyltransferase and histone deacetylase
inhibition in human ER-alpha-negative breast cancer cells. Cancer Res 2001; 61(19):
7025–7029.
38. Guler G, Iliopoulos D, Guler N, et al. Wwox and Ap2g expression levels predict
tamoxifen response. Clin Cancer Res 2007; 13(20).
39. Guler G, Uner A, Guler N, et al. The fragile genes FHIT and WWOX are inactivated
coordinately in invasive breast carcinoma. Cancer 2004; 100(8):1605–1614.
40. Hurtubise A, Momparler RL. Effect of histone deacetylase inhibitor LAQ824 on
antineoplastic action of 5-Aza-2 0 -deoxycytidine (decitabine) on human breast carcinoma
cells. Cancer Chemother Pharmacol 2006; 58(5):618–625.

The Role of Epigenetics in Breast Cancer 59
41. Chuang JC, Yoo CB, Kwan JM, et al. Comparison of biological effects of nonnucleoside
DNA methylation inhibitors versus 5-aza-2 0 -deoxycytidine. Mol Cancer
Ther 2005; 4(10):1515–1520.
42. Cornacchia E, Golbus J, Maybaum J, et al. Hydralazine and procainamide inhibit
T cell DNA methylation and induce autoreactivity. J Immunol 1988; 140(7):
2197–2200.
43. Siedlecki P, Boy RG, Musch T, et al. Discovery of two novel, small-molecule
inhibitors of DNA methylation. J Med Chem 2006; 49(2):678–683.
44. Arce C, Pérez-Plasencia C, Gonzälez-Fierro A, et al. A proof-of-principle study of
epigenetic therapy added to neoadjuvant doxorubicin cyclophosphamide for locally
advanced breast cancer. PLoS ONE 2006; 1:e98.
45. Joung KE, Min KN, An JY, et al. Potent in vivo anti-breast cancer activity of IN-
2001, a novel inhibitor of histone deacetylase, in MMTV/c-Neu mice. Cancer Res
2006; 66(10):5394–5402.
46. Ferguson AT, Vertino PM, Spitzner JR, et al. Role of estrogen receptor gene
demethylation and DNA methyltransferase. DNA adduct formation in 5-aza-
2 0 deoxycytidine-induced cytotoxicity in human breast cancer cells. J Biol Chem
1997; 272(51):32260–32206.

5
Comparative Genomic Hybridization
and Copy Number Abnormalities in
Breast Cancer
Nicholas Wang and Joe Gray
Life Sciences Division, Lawrence Berkeley National Laboratory,
Berkeley, California, U.S.A.
INTRODUCTION
The progression of normal breast epithelial cells from a normal state toward one
characterized by uncontrolled growth and metastatic behavior is caused by the
deregulation of key cellular processes and signaling pathways. These alterations
in normal cellular behavior are rooted in the accumulation of genomic and
epigenomic lesions that impact hallmarks of cancer, such as the ability of the cell
to control proliferation, undergo apoptosis, increase motility leading to invasion,
and alter angiogenesis. A suite of technologies has now been developed to assess
genomic and epigenomic aberrations that contribute to cancer progression. The
application of these has shown that genome copy number abnormalities (CNAs)
are among the most frequent genomic aberrations. Remarkably, these studies
have revealed that 10% to 15% of the genes in a typical carcinoma tumor may be
deregulated by recurrent genome CNAs and regions. Some of these genes
influence disease progress and so may be assessed to facilitate prognosis. Others
influence response to therapy and so may be assessed as predictive markers.
Some of these enable oncogenic processes, on which tumors depend for survival,
and so are candidate therapeutic targets. In most cases, array comparative
61

62 Wang and Gray
genomic hybridization (CGH) is the method of choice for their discovery and
may be used in some setting for clinical assessments of these abnormalities.
Accordingly, we review here several of the current array CGH technologies
available and the considerations needed when determining the technology most
applicable to a given study.
CGH analyses of thousands of primary breast tumors and cell lines have
now defined recurrent CNAs throughout the genome and identified genes
encoded in these aberrant regions that are deregulated by the abnormalities (1,2–4).
In breast cancer, for example, 10% to 15% of known genes are deregulated by
these CNAs (2–4). The aberrations tend to fall into two general classes that may
contribute in different ways to cancer pathophysiology: low-level CNAs that
occur through the gain or loss of extended regions of single chromosomes or
extended regions thereof and high-level amplification or homozygous deletions
that tend to be narrower in genomic extent.
Most genes are deregulated by low-level aberrations. The contributions of
these genes to cancer pathophysiology are not well understood; however, evidence
is now emerging that these may provide, en ensemble, a general metabolic
advantage (4–6). These aberrations typically occur relatively early in the carcinogenesis
process. It has been proposed that these aberrations are selected during
successful transition through a telomere crisis (7).
High-level amplification and narrow homozygous deletions provide strong
evidence of active selection during cancer progression and seem to occur later
during cancer progression. In breast cancer, regions of high-level amplification
encode well-studied oncogenes, such as ERBB2 (17q12) (8), FGFR1 (8p11) (9),
PVT1 and MYC (8q24) (10,11), CCND1 and EMS1 (11q13) (12), and ABI1 and
ZNF217 (20q13) (13,14), while deletions contribute to the inactivation of tumor
suppressor genes such as TP53 (17p13) (15), CDKN2A (9p21) (16), and BRCA1
(13q14) (17). While it is often assumed that these oncogenes and tumor suppressor
genes are the sole targets for these aberrations, a growing body of
functional data suggests that several genes in each region of recurrent abnormality
may be coselected by the event. In breast cancer, for example, regions of
amplification in which more than one gene appears to be functionally important
include 8p11, 8q24, 11q13, 17q21, and 20q13 (4). Table 1 summarizes published
studies for genes in these regions of amplification. The concept of multigene
selection is also supported by studies in mouse models, where the extents of
amplicons in mice and humans are similar. Similarities in the extent of amplification
around ERBB2 in humans and in polyomavirus middle T-induced
mammary tumors in mice are particularly striking (18). Similar studies are
emerging at other organ sites (19). As a result, it is critically important that the
extent of each region be defined as precisely as possible. High-resolution CGH
analyses are now contributing to this process.
Several recurrent high-level CNAs are associated with clinical endpoints,
including reduced survival duration and response to aberration-targeted therapies.
Assays for recurrent aberrations or genes deregulated by CNA abnormalities have

Comparative Genomic Hybridization in Breast Cancer 63
Table 1
Array CGH Analysis Platform Summary
Platform
composition Reference Density
Allele
specific
Source
cDNA 5 40K No Academic
BAC 68 25K No Academic/Spectral
Genomics
Oligonucleotide http://affymetrix.com 950K Yes Affymetrix
Oligonucleotide http://www.
400K No Nimblegen
nimblegen.com
Oligonucleotide http://www.chem. 240K No Agilent
agilent.com
Oligonucleotide http://www.illumina. 1M Yes Illumina
com
MIP 41 20K Yes Academic
Abbreviations: CGH, comparative genomic hybridization; BAC, bacterial artificial chromosome;
MIP, molecular inversion probe.
already been incorporated into molecular indicators of outcome. In addition,
several therapeutic agents are now being developed to attack genes deregulated
fully or in part by recurrent genome abnormalities so that assays for these targets
may be useful as predictors of response. In breast cancer, the most prominent
examples are trastuzumab (20,21) and lapatinib (22) (NEJM PMID 17192538),
developed to attack tumors driven by amplification of ERBB2. In addition,
therapies have been developed to treat tumors arising from deletion or inactivation
of TP53 (23) and activation of elements of the PI3-kinase pathway, including
PIK3CA, PTEN, AKT, and ZNF217 (24–27).
The ability to detect and define the extents of CNAs has steadily improved
with the evolution of CGH analysis technologies (Table 2). The first CGH
studies mapped CNAs onto metaphase chromosomes, but the resolution of such
analyses was limited to about 10 Mbp by the complex packaging of the DNA
into chromosomes (28). The introduction of array CGH in which CNAs were
mapped onto arrays of cloned probes substantially increased resolution. Early
arrays comprised hundreds to thousands of probes and improved resolution to
about 1 Mbp (29,30). Current cloned probe arrays interrogate the genome at
approximately single-gene resolution (5). Today, several commercial platforms
map abnormalities onto arrays of oligonucleotide probes with subgene resolution
(either on flat substrates or attached to microbeads) (31), and some allow
analyses in an allele-specific manner (32). Applications of these platforms are
further refining the extents of recurrent genomic aberrations but also reveal the
somewhat surprising existence of germline copy number polymorphisms that
must be taken into account when interpreting CGH measurements. Several
aspects of these important array CGH analysis technologies are reviewed in the
following sections.

64 Wang and Gray
Table 2
Data Analysis Algorithms
Platform Program Reference Type
Required
software
Two-color arrays aCGH 54 HMM R
GLAD 56 Adaptive weight R
smoothing
CLAC 57 Clustering along R
chromosomes
DNAcopy 55 Segmentation R
Single-color arrays dChipSNP 37 HMM none
(Affymetrix)
CARAT 69 none
CNAG 70 Regression none
GEMCA 71 Regression none
Abbreviations: aCGH, array comparative genomic hybridization; HMM, hidden Markov model;
GLAD, Gaussian model–based approach; CART, copy number analysis with regression and tree; CLAC,
clustering along each chromosome; SNP, single nucleotide polymorphism; CNAG, Copy Number
Analyzer for GeneChip; GEMCA, Genotyping microarray based copy number variation analysis.
COMPARATIVE GENOMIC HYBRIDIZATION
CGH is the primary technology used for assessment of genome copy number.
Typically, analysis proceeds by hybridizing differentially labeled tumor and normal
DNA to a representation of the normal genome, although some platforms
perform tumor and normal reference hybridization in separate representations. The
ratio of tumor label to normal label, hybridized to each locus on the representation
is then measured as an estimate of relative copy number at that locus.
The first CGH experiments employed metaphase chromosomes as the
representation onto which information was mapped (28,33). While useful, the
genomic resolution was limited by the nonlinear organization of DNA along
metaphase chromosomes. Thus, chromosome CGH has now been largely supplanted
by array CGH, in which information is mapped onto arrays of DNA
probes (Fig. 1). The initial array CGH platforms comprised large genomic cloned
probes, such as yeast artificial chromosomes (YACs) (34), bacterial artificial
chromosomes (BACs), and P1-derived artificial chromosomes (PACs) (29), or
shorter cDNA sequences spotted on glass slides (35). More recently, synthetic
oligonucleotide probes have been used successfully (36–39). Several platforms
are summarized in Table 1 and in the following sections.
Cloned probe arrays are typically available from large academic cores with
the experience and resources to maintain and print the arrays. These arrays are
developed using the large collections of cDNAs (35) or BACs (29,30,40)
developed during the course of the human genome program. The advantages of

Comparative Genomic Hybridization in Breast Cancer 65
Figure 1 The use of a dual-color hybridization protocol for the detection of genome copy
number aberrations. Both a normal sample (representing two copies of the genome) and a
tumor sample are randomly labeled with differential fluorescent dyes. They are then
cohybridized onto an array platform (BACs or oligonucleotides) with the subsequent log 2
dye ratios of the tumor sample compared with the normal sample used to detect copy
number changes. Abbreviation: BACs, bacterial artificial chromosomes.
this approach are that the clones are readily available in the public domain and
the signals produced during hybridization are large so that measurement precision
is relatively high. The disadvantages are that the clones are not perfectly
curated and mapped, so some may be mismapped. In addition, managing the
large number is daunting, and the ultimate resolution of the analysis platforms is
limited by the relatively genomic extents of the cloned probes.
Commercial vendors, such as Affymetrix, Agilent, Illumina, and NimbleGen,
have begun offering oligonucleotide-based arrays with which to study
genomewide CNAs (41–43). The Illumina arrays are unique in that the oligonucleotide
probes are attached to microbeads (39), while the NimbleGen arrays
are synthesized on demand and so can be easily customized (44). The smaller
target sequences, 25 to 85 bp, also offer the possibility of higher resolution than
is possible using cloned probe arrays. Some protocols generate relative copy
number information by comparing tumor and normal hybridization signals using
two-color cohybridization approaches, while others rely on hybridizations of
tumor and normal reference DNA to separate arrays. The latter approach is
technically simpler but requires additional normalization steps to account for
array-to-array differences, introducing additional experimental variables when
analyzing data sets.

66 Wang and Gray
Since each oligonucleotide probe is synthesized to be complementary to a
known sequence, oligonucleotide array platforms do not suffer as much from
incorrect mapping, and probes can be selected to be uniquely represented in the
genome, thereby minimizing cross-hybridization artifacts. The smaller sequences
also are useful in detecting single nucleotide polymorphisms (SNPs), since
hybridization rates to 25 mer are significantly reduced by a single nucleotide
mismatch (37,45). The additional information provides researchers the ability to
detect loss of heterozygosity (LOH) in regions of neutral copy change and also
provide data for use in association studies (42). Regions of LOH that do not carry
imbalances in copy number may provide evidence of selective pressures duplicating
one allele coupled with the loss of the other. This is an important
mechanism enabling removal of wild-type alleles while selecting mutations.
Allele-specific selection can be either positive or negative. For example, loss of
the wild-type allele and duplication of an inactivating mutation might lead to
inactivation of a tumor suppressor gene. The high frequency of LOH on 17p
where p53 is located is an example of this type of selection (Fig. 2). On the other
hand, amplification of a polymorphism that increases gene activity may contribute
an advantage to the tumor. Amplification of a specific polymorphism in
the aurora kinase STK15 is one example (46). The utility of allele-specific copy
number analysis may also contribute to the identification of polymorphisms that
confer increased susceptibility to cancer, since evidence is now emerging that
tumors may increase the number of copies of polymorphisms that confer
increased risk and may decrease the number of copies of polymorphisms that
confer decreased risk of developing cancer (47). The frequency of allele-specific
copy increases and decreases in individual tumors is substantial, as illustrated in
Figure 3.
While commercial oligonucleotide arrays offer advantages, the smaller
target sequences also present some disadvantages. Foremost among these is the
relatively low signal-to-noise ratio generated during hybridization. Because the
hybridization sequences are smaller, there is less nucleic acid sequence available
to label, leading to less intense labeling. This typically leads to higher variability
between probe sets in oligonucleotide arrays, and the need to combine multiple
data points to estimate copy number, which in turn reduces the actual resolution
of these arrays (38,48). The increased variability compared to BAC arrays also
makes the detection of low-level changes, such as single-copy gains and losses,
more difficult to detect, although performance with oligonucleotide arrays is
improving steadily. Typically, microgram quantities of starting DNA are needed
for an analysis.
The most recent array CGH analysis platform does not operate through
hybridization of tumor and normal DNA to a normal representation. Instead, it
utilizes a molecular inversion probe (MIP) interrogation strategy to perform
allele-specific copy number analysis (45). In this approach, each locus is interrogated
by a linear MIP probe that has sequence homology to approximately
20-bp regions flanking a single base to be tested. Each MIP also contains a unique

Comparative Genomic Hybridization in Breast Cancer 67
Figure 2 LOH detection using Affymetrix GeneChip 1 Human Mapping 50K Array Xba
240. A panel of breast cancer cell lines is individually hybridized to oligonucleotide
arrays, and copy number changes and LOH are determined using dChipSNP. (A) The
expected result of a genomic deletion is the presence of LOH, as seen on chromosome 8.
(B) However, chromosome 17 shows much higher levels of LOH compared with deletions,
suggesting that these regions have undergone genomic changes such as a deletion or
duplication event or genomic conversion of one homolog so that you maintain normal
copy number but not allelic information. This may result in the loss of a functional allele
of a tumor suppressor gene or an active gain-of-function allele in the case of an oncogene.
These genomic changes are only evident using a platform capable of detecting allelic
changes. Abbreviation: LOH, loss of heterozygositiy.
tag sequence and universal polymerase chain reaction (PCR) primer-binding
sites facing away from each other so that PCR with primers to the universal
sequences will not amplify. The first step in MIP CGH is to hybridize the linear
MIPs to the DNA sample to be interrogated. After hybridization, the reaction
mix is separated into multiple tubes, each containing polymerase, ligase, and one
of the four nucleotides. The polymerase and ligase then fill the gaps in each
reaction mix that contains a nucleotide complementary to the base at the test
locus. The linear DNA and probes are then eliminated, reactions are mixed, and
the ligated probes are amplified and labeled using PCR with the universal primers.
The resulting PCR products are hybridized to an array comprising oligonucleotides
homologous to the tag sequences in the MIPs. To date, MIP analyses
interrogating over 20,000 loci have been reported (45). This approach reports

68 Wang and Gray
Figure 3 Allelic copy number changes in breast cancer cell lines from a 50K MIP
array platform. Allele copy numbers are represented as the higher copy allele and
lower copy allele. (A) Allele copy number data from HCC38 from chromosome 5.
The lack of copies of the lower allele is indicative of LOH. The array detects a region
of LOH without copy number change. (B) Allele copy number data from ZR75.30
from chromosome 8. 8q displays a region of monoallelic amplification in the region
that contains MYC. Abbreviations: MIP, molecular inversion probe; LOH, loss of
heterozygositiy.

Comparative Genomic Hybridization in Breast Cancer 69
copy number in an allele-specific manner and works well with FFPE material
since only ~40 bp of intact DNA is needed at each locus for proper analysis.
Typically, only about 75 ng of DNA is needed for an analysis since the reaction
itself contains a PCR amplification step. The approach has the advantage that the
tag sequences and arrays can be designed to have uniform hybridization conditions
to copy number and the hybridization does not depend on the quality of
the hybridized DNA.
DNA REQUIREMENTS
One of the most important factors when considering a clinical CGH tool is the
quality and amount of DNA available for analysis. Depending on which array
platform is chosen, the amounts needed for successful hybridization can range
from 50 ng to 3 mg. In addition, some protocols have been developed that allow
analysis of material from formalin-fixed, paraffin-embedded (FFPE) samples
(39,49). The largest amounts of starting material generally are used in protocols
in which entire genomes are labeled and hybridized to the array, as is usual for
array CGH using BAC or Agilent oligonucleotide arrays. In these experiments,
the starting DNA is typically labeled with fluorescent nucleotides using random
priming. For BAC arrays, unlabeled repeat-rich DNA is included to suppress
hybridization of repeated sequences to repetitive sequences that are present in
the BAC probes. Larger amounts of DNA are needed to “drive” these wholegenome
hybridization reactions. The amount of starting material needed for
analysis is reduced if the DNA is amplified or reduced in complexity—for
example, through a PCR amplification step preceded by restriction digestion that
reduces the complexity of the hybridized DNA (50). This complexity reduction
results in an increase of template sequences, which can utilized in downstream
hybridization steps. Both academically and commercially developed platforms
result in data sets that are able to detect gains and losses with the caveat that
variations and noise may differ between platforms (Fig. 4).
While there is a range of starting DNA amounts between the various
platforms, recent protocols have been developed that can amplify a large
quantity of DNA from small starting amounts. These applications can be critical
when dealing with samples that are limiting. These whole-genome amplifications
are useful for both intact, high-quality DNA and degraded samples and
involve the use of j29 enzyme or PCR (51,52). Alternately, the MIP platform has
a targeted PCR amplification inherent to the protocol, so it is particularly tolerant
of small amounts of starting DNA (45). To date, successful data sets have been
generated using as little material as 1 ng or a single cell (53). However, as with
any random amplification procedure, the protocols may artificially create misrepresentations
of the genome, resulting in false positives for amplifications and
deletions. Determining the amount and quality of starting material along with the

70 Wang and Gray
Figure 4 Chromosomal copy number changes in the breast cancer cell line MDAMB453. DNA from MDAMB453 is applied to three separate
platforms: (A) a 3K BAC array, (B) a 20K MIP array, and (C) a 50K XbaI Affymetrix SNP array. Data from each platform are scaled using adjacent data
points so that the resulting resolutions are comparable. The data are then smoothed using the GLAD software package. Abbreviations: BAC, bacterial
artificial chromosome; MIP, molecular inversion probe; SNP, single nucleotide polymorphism; GLAD, Gaussian model–based approach.

Comparative Genomic Hybridization in Breast Cancer 71
type of information required will help determine the appropriate array CGH
platform and DNA preparation technique to utilize.
ABNORMALITY DETECTION AND INTERPRETATION
Computational strategies to identify copy number gains and losses are a critical
part of array CGH. The first step in the analysis of data is the identification of
any features with anomalous intensity readings. These could result from a high
background, errors in printing, contamination, or air bubbles prohibiting
hybridization of the sample to the array. After removal of these spurious data
points, a normalization step is applied to calculate the tumor hybridization
intensity along the genome, relative to the normal hybridization intensity. A
statistical analysis is then applied to identify genome segments with different
copy number levels. Several strategies have been developed for this purpose and
are summarized in Table 2. These include the hidden Markov model (HMM)
(54), circular binary segmentation (55), a Gaussian model–based approach
(GLAD) (56), hierarchical clustering along each chromosome (CLAC) (57), an
expectation maximization–based method (58), a Bayesian model (59), and a
wavelet approach (60). Several of these algorithms have been made available as
R packages through Bioconductor (www.bioconductor.org). Comparsions of
several of these methods have now been published (61,62).
Detection of the differences in copy number between tumor and normal
samples is increasingly straightforward and precise as CGH procedures improve.
However, the interpretation of CGH analyses requires additional interpretation.
First, high-resolution array CGH analyses are important now that focal, germline
genome copy number polymorphisms are relatively common (36). These may be
scored as CNAs if the samples for tumor versus normal CGH analyses are not
from the same individual. In addition, it is tempting to think that the regions of
the genome not detected as aberrant during CGH analyses are present in cancer
genomes in two copies. This is often not the case since tumor genomes may be
highly aneuploid. Thus, a CGH comparison of a perfectly tetraploid tumor with a
diploid normal sample will not show any evidence of abnormality. Thus, if
absolute copy number assessments are important, it will be necessary to normalize
the CGH analyses using information about absolute copy number at one or a few
loci measured using fluorescent in situ hybridization (FISH) or quantitative PCR.
CONCLUSIONS AND FUTURE DIRECTIONS
Array CGH technologies are becoming increasingly powerful tools for the
identification of recurrent genome CNAs that are associated with outcome or
response to therapy in breast and other cancers (63). The ability to detect genetic
copy number changes of both large and small scales will be critical in evaluating
the most effective therapies on an individual patient basis. In the short term,
array CGH will likely be accomplished using commercial oligonucleotide
analysis platforms that allow assessment of allele-specific copy number using

72 Wang and Gray
material from FFPE samples. In the long term, however, array CGH may be
supplanted by high-throughput sequencing technologies that allow assessment of
both copy number and DNA sequence through deep sequencing. It is already clear
that shotgun sequencing, if done at sufficient redundancy, offers the same or better
resolution than CGH platforms (64,65). Next-generation sequencing platforms
based on massively parallel single-molecule sequencing promise sufficiently lowcost
sequencing, so that these platforms may become a viable alternative to array
CGH (66,67). Today, this is not cost effective, but costs of high-throughput
sequencing are falling rapidly, so these approaches should be followed closely.
REFERENCES
1. Knuutila S, Autio K, Aalto Y. Online access to CGH data of DNA sequence copy
number changes. Am J Pathol 2000; 157:689.
2. Neve RM, Chin K, Fridlyand J, et al. A collection of breast cancer cell lines for the
study of functionally distinct cancer subtypes. Cancer Cell 2006; 10:515–527.
3. Pollack JR, Sorlie T, Perou CM, et al. Microarray analysis reveals a major direct role
of DNA copy number alteration in the transcriptional program of human breast
tumors. Proc Natl Acad Sci U.S.A. 2002; 99:12963–12968.
4. Chin K, DeVries S, Fridlyand J, et al. Genomic and transcriptional aberrations linked
to breast cancer pathophysiologies. Cancer Cell 2006; 10:529–541.
5. Bergamaschi A, Kim YH, Wang P, et al. Distinct patterns of DNA copy number
alteration are associated with different clinicopathological features and gene-expression
subtypes of breast cancer. Genes Chromosomes Cancer 2006; 45:1033–1040.
6. Callagy G, Pharoah P, Chin SF, et al. Identification and validation of prognostic
markers in breast cancer with the complementary use of array-CGH and tissue
microarrays. J Pathol 2005; 205:388–396.
7. Chin K, de Solorzano CO, Knowles D, et al. In situ analyses of genome instability in
breast cancer. Nat Genet 2004; 36:984–988.
8. Slamon DJ, Godolphin W, Jones LA, et al. Studies of the HER-2/neu proto-oncogene
in human breast and ovarian cancer. Science 1989; 244:707–712.
9. Theillet C, Adelaide J, Louason G, et al. FGFRI and PLAT genes and DNA
amplification at 8p12 in breast and ovarian cancers. Genes Chromosomes Cancer
1993; 7:219–226.
10. Escot C, Theillet C, Lidereau R, et al. Genetic alteration of the c-myc protooncogene
(MYC) in human primary breast carcinomas. Proc Natl Acad Sci U.S.A. 1986;
83:4834–4838.
11. Guan Y, Kuo WL, Stilwell JL, et al. Amplification of PVT1 contributes to the
pathophysiology of ovarian and breast cancer. Clin Cancer Res 2007; 13:5745–5755.
12. Schuuring E, Verhoeven E, Mooi WJ, et al. Identification and cloning of two
overexpressed genes, U21B31/PRAD1 and EMS1, within the amplified chromosome
11q13 region in human carcinomas. Oncogene 1992; 7:355–361.
13. Anzick SL, Kononen J, Walker RL, et al. AIB1, a steroid receptor coactivator
amplified in breast and ovarian cancer. Science 1997; 277:965–968.
14. Tanner MM, Tirkkonen M, Kallioniemi A, et al. Amplification of chromosomal
region 20q13 in invasive breast cancer: prognostic implications. Clin Cancer Res
1995; 1:1455–1461.

Comparative Genomic Hybridization in Breast Cancer 73
15. Varley JM, Brammar WJ, Lane DP, et al. Loss of chromosome 17p13 sequences and
mutation of p53 in human breast carcinomas. Oncogene 1991; 6:413–421.
16. Kamb A, Gruis NA, Weaver-Feldhaus J, et al. A cell cycle regulator potentially
involved in genesis of many tumor types. Science 1994; 264:436–440.
17. Friedman LS, Ostermeyer EA, Lynch ED, et al. The search for BRCA1. Cancer Res
1994; 54:6374–6382.
18. Hodgson JG, Malek T, Bornstein S, et al. Copy number aberrations in mouse breast
tumors reveal loci and genes important in tumorigenic receptor tyrosine kinase
signaling. Cancer Res 2005; 65:9695–9704.
19. Maser RS, Choudhury B, Campbell PJ, et al. Chromosomally unstable mouse tumors
have genomic alterations similar to diverse human cancers. Nature 2007; 447:966–971.
20. Yeon CH, Pegram MD. Anti-erbB-2 antibody trastuzumab in the treatment of HER2-
amplified breast cancer. Invest New Drugs 2005; 23:391–409.
21. Vogel CL, Cobleigh MA, Tripathy D, et al. First-line Herceptin monotherapy in
metastatic breast cancer. Oncology 2001; 61(suppl 2):37–42.
22. Geyer CE, Forster J, Lindquist D, et al. Lapatinib plus capecitabine for HER2-
positive advanced breast cancer. N Engl J Med 2006; 355(26):2733–2743.
23. Kumar S, Walia V, Ray M, et al. p53 in breast cancer: mutation and countermeasures.
Front Biosci 2007; 12:4168–4178.
24. Jonsson G, Staaf J, Olsson E, et al. High-resolution genomic profiles of breast cancer
cell lines assessed by tiling BAC array comparative genomic hybridization. Genes
Chromosomes Cancer 2007; 46:543–558.
25. Wu G, Xing M, Mambo E, et al. Somatic mutation and gain of copy number of
PIK3CA in human breast cancer. Breast Cancer Res 2005; 7:R609–R616.
26. Huang G, Krig S, Kowbel D, et al. ZNF217 suppresses cell death associated with
chemotherapy and telomere dysfunction. Hum Mol Genet 2005; 14:3219–3125.
27. Bellacosa A, de Feo D, Godwin AK, et al. Molecular alterations of the AKT2
oncogene in ovarian and breast carcinomas. Int J Cancer 1995; 64:280–285.
28. Kallioniemi A, Kallioniemi OP, Sudar D, et al. Comparative genomic hybridization
for molecular cytogenetic analysis of solid tumors. Science 1992; 258:818–821.
29. Pinkel D, Segraves R, Sudar D, et al. High-resolution analysis of DNA copy number
variation using comparative genomic hybridization to microarrays. Nat Genet 1998;
20:207–211.
30. Snijders AM, Nowak N, Segraves R, et al. Assembly of microarrays for genomewide
measurement of DNA copy number. Nat Genet 2001; 29:263–264.
31. Toruner GA, Streck DL, Schwalb MN, et al. An oligonucleotide based array-CGH
system for detection of genome wide copy number changes including subtelomeric
regions for genetic evaluation of mental retardation. Am J Med Genet A 2007;
143:824–829.
32. George RE, Attiyeh EF, Li S, et al. Genome-wide analysis of neuroblastomas using
high-density single nucleotide polymorphism arrays. PLoS ONE 2007; 2:e255.
33. Kallioniemi A, Kallioniemi OP, Piper J, et al. Detection and mapping of amplified
DNA sequences in breast cancer by comparative genomic hybridization. Proc Natl
Acad Sci U.S.A. 1994; 91:2156–2160.
34. Solinas-Toldo S, Lampel S, Stilgenbauer S, et al. Matrix-based comparative genomic
hybridization: biochips to screen for genomic imbalances. Genes Chromosomes
Cancer 1997; 20:399–407.

74 Wang and Gray
35. Pollack JR, Perou CM, Alizadeh AA, et al. Genome-wide analysis of DNA copynumber
changes using cDNA microarrays. Nat Genet 1999; 23:41–46.
36. Lucito R, Healy J, Alexander J, et al. Representational oligonucleotide microarray
analysis: a high-resolution method to detect genome copy number variation. Genome
Res 2003; 13:2291–2305.
37. Zhao X, Li C, Paez JG, et al. An integrated view of copy number and allelic
alterations in the cancer genome using single nucleotide polymorphism arrays.
Cancer Res 2004; 64:3060–3071.
38. Barrett MT, Scheffer A, Ben-Dor A, et al. Comparative genomic hybridization using
oligonucleotide microarrays and total genomic DNA. Proc Natl Acad Sci U.S.A.
2004; 101:17765–17770.
39. Oosting J, Lips EH, van Eijk R, et al. High-resolution copy number analysis of
paraffin-embedded archival tissue using SNP BeadArrays. Genome Res 2007; 17:
368–376.
40. Hodgson G, Hager JH, Volik S, et al. Genome scanning with array CGH delineates
regional alterations in mouse islet carcinomas. Nat Genet 2001; 29:459–464.
41. Iljin K, Wolf M, Edgren H, et al. TMPRSS2 fusions with oncogenic ETS factors in
prostate cancer involve unbalanced genomic rearrangements and are associated with
HDAC1 and epigenetic reprogramming. Cancer Res 2006; 66:10242–10246.
42. Zhou X, Mok SC, Chen Z, et al. Concurrent analysis of loss of heterozygosity (LOH)
and copy number abnormality (CNA) for oral premalignancy progression using the
Affymetrix 10K SNP mapping array. Hum Genet 2004; 115:327–330.
43. Sharp AJ, Hansen S, Selzer RR, et al. Discovery of previously unidentified genomic
disorders from the duplication architecture of the human genome. Nat Genet 2006;
38:1038–1042.
44. Selzer RR, Richmond TA, Pofahl NJ, et al. Analysis of chromosome breakpoints in
neuroblastoma at sub-kilobase resolution using fine-tiling oligonucleotide array
CGH. Genes Chromosomes Cancer 2005; 44:305–319.
45. Wang Y, Moorhead M, Karlin-Neumann G, et al. Allele quantification using
molecular inversion probes (MIP). Nucleic Acids Res 2005; 33:e183.
46. Ewart-Toland A, Briassouli P, de Koning JP, et al. Identification of Stk6/STK15 as a
candidate low-penetrance tumor-susceptibility gene in mouse and human. Nat Genet
2003; 34:403–412.
47. Balmain A, Gray J, Ponder B. The genetics and genomics of cancer. Nat Genet 2003;
33:238–244.
48. Carvalho B, Ouwerkerk E, Meijer GA, et al. High-resolution microarray comparative
genomic hybridisation analysis using spotted oligonucleotides. J Clin Pathol 2004;
57:644–646.
49. Nassiri M, Gugic D, Olczyk J, et al. Preservation of skin DNA for oligonucleotide
array CGH studies: a feasibility study. Arch Dermatol Res 2007; 299:353–357.
50. Hicks J, Muthuswamy L, Krasnitz A, et al. High-resolution ROMA CGH and FISH
analysis of aneuploid and diploid breast tumors. Cold Spring Harb Symp Quant Biol
2005; 70:51–63.
51. Guillaud-Bataille M, Valent A, Soularue P, et al. Detecting single DNA copy number
variations in complex genomes using one nanogram of starting DNA and BAC-array
CGH. Nucleic Acids Res 2004; 32:e112.

Comparative Genomic Hybridization in Breast Cancer 75
52. Lage JM, Leamon JH, Pejovic T, et al. Whole genome analysis of genetic alterations
in small DNA samples using hyperbranched strand displacement amplification and
array-CGH. Genome Res 2003; 13:294–307.
53. Le Caignec C, Spits C, Sermon K, et al. Single-cell chromosomal imbalances
detection by array CGH. Nucleic Acids Res 2006; 34:e68.
54. Fridlyand J, Snijders A, Pinkel D, et al. Hidden Markov models approach to the
analysis of array CGH data. J Multivar Anal 2004; 90:132–153.
55. Olshen AB, Venkatraman ES, Lucito R, et al. Circular binary segmentation for the
analysis of array-based DNA copy number data. Biostatistics 2004; 5:557–572.
56. Hupe P, Stransky N, Thiery JP, et al. Analysis of array CGH data: from signal ratio
to gain and loss of DNA regions. Bioinformatics 2004; 20:3413–3122.
57. Wang P, Kim Y, Pollack J, et al. A method for calling gains and losses in array CGH
data. Biostatistics 2005; 6:45–58.
58. Myers CL, Dunham MJ, Kung SY, et al. Accurate detection of aneuploidies in array
CGH and gene expression microarray data. Bioinformatics 2004; 20:3533–3543.
59. Daruwala RS, Rudra A, Ostrer H, et al. A versatile statistical analysis algorithm to detect
genome copy number variation. Proc Natl Acad Sci U.S.A. 2004; 101:16292–16297.
60. Hsu L, Self SG, Grove D, et al. Denoising array-based comparative genomic
hybridization data using wavelets. Biostatistics 2005; 6:211–226.
61. Lai WR, Johnson MD, Kucherlapati R, et al. Comparative analysis of algorithms for
identifying amplifications and deletions in array CGH data. Bioinformatics 2005;
21:3763–3770.
62. Willenbrock H, Fridlyand J. A comparison study: applying segmentation to array
CGH data for downstream analyses. Bioinformatics 2005; 21:4084–4091.
63. Gunn SR, Robetorye RS, Mohammed MS. Comparative genomic hybridization arrays
in clinical pathology: progress and challenges. Mol Diagn Ther 2007; 11:73–77.
64. Volik S, Raphael BJ, Huang G, et al. Decoding the fine-scale structure of a breast
cancer genome and transcriptome. Genome Res 2006; 16:394–404.
65. Volik S, Zhao S, Chin K, et al. End-sequence profiling: sequence-based analysis of
aberrant genomes. Proc Natl Acad Sci U.S.A. 2003; 100:7696–7701.
66. Albert TJ, Molla MN, Muzny DM, et al. Direct selection of human genomic loci by
microarray hybridization. Nat Methods 2007; 4:903–905.
67. Bennett ST, Barnes C, Cox A, et al. Toward the 1,000 dollars human genome.
Pharmacogenomics 2005; 6:373–382.
68. Ishkanian AS, Malloff CA, Watson SK, et al. A tiling resolution DNA microarray
with complete coverage of the human genome. Nat Genet 2004; 36:299–303.
69. Huang J, Wei W, Chen J, et al. CARAT: a novel method for allelic detection of DNA
copy number changes using high density oligonucleotide arrays. BMC Bioinformatics
2006; 7(83):83.
70. Nannya Y, Sanada M, Nakazaki K, et al. A robust algorithm for copy number
detection using high-density oligonucleotide single nucleotide polymorphism genotyping
arrays. Cancer Res 2005; 65(14):6071–6079.
71. Komura D, Shen F, Ishikawa S, et al. Genome-wide detection of human copy number
variations using high-density DNA oligonucleotide arrays. Genome Res 2006; 16(12):
1575–1584.

6
Gene Expression Profiling as an Emerging
Diagnostic Tool to Personalize
Chemotherapy Selection for
Early Stage Breast Cancer
Cornelia Liedtke and Lajos Pusztai
Department of Breast Medical Oncology, The University of Texas,
Anderson Cancer Center, Houston, Texas, U.S.A.
INTRODUCTION
The selection of adjuvant chemotherapy for breast cancer is currently more of an
art than a science. Several combination chemotherapy regimens can improve
disease-free and overall survival of stage I to stage III breast cancer when used as
adjuvant (postoperative) or neoadjuvant (preoperative) therapy (1,2). However,
no chemotherapy regimen is universally effective for all patients, and it is currently
impossible to predict which of these regimens will actually work for a
particular individual. Choosing the best possible adjuvant or neoadjuvant chemotherapy
regimen for an individual is important because the right choice may
make the difference between cure and relapse. Results from traditionally
designed randomized clinical trials offer little guidance for individualized
treatment selection. These studies compare the activity of different treatments to
establish which is the most effective for the entire study population. This
approach ignores the fact that many patients are cured with earlier generation
adjuvant regimens and therefore are not well served by the administration of
77

78 Liedtke and Pusztai
longer, potentially more toxic, and more expensive second- or third-generation
combination therapies. These studies also cannot reveal whether some cancers
are particularly sensitive to a specific drug (or a combination of drugs) and
therefore can only be cured by that treatment. The most effective treatment for
an individual may or may not be the same regimen that is the most effective for
the whole population. These theoretical concerns and the increasing number of
adjuvant chemotherapy options, as well as the advent of new molecular analytical
techniques, fuel a renewed interest in developing chemotherapy response
predictors that may lead to more personalized treatment recommendations.
THE ROLE OF NEOADJUVANT CHEMOTHERAPY
IN THE TREATMENT OF BREAST CANCER
Administration of chemotherapy before surgical resection of the cancer (preoperative
or neoadjuvant chemotherapy) provides a unique opportunity to
identify molecular markers of response. Preoperative chemotherapy is the current
standard of care for locally advanced breast cancer because this treatment
often renders previously inoperable cancers amenable to surgery due to frequent
and substantial tumor responses (3). Preoperative chemotherapy is also
increasingly used in the management of operable breast cancer because it can
reduce tumor size and allow for more limited surgery and possibly improved
cosmetic outcome (4). Importantly, there is no statistically significant difference
in survival between those who receive chemotherapy before surgery (i.e., neoadjuvant)
and those who are treated with the same regimen postoperatively (i.e.,
adjuvant) (5–7). Therefore, preoperative chemotherapy is an appropriate option
for most patients who present with a breast cancer that requires systemic therapy.
Furthermore, eradication of all invasive cancer from the breast and regional
lymph nodes by the preoperative therapy [i.e., pathologic complete response,
(pCR)], is associated with excellent long-term, disease-free, and overall survival
(7,8). Therefore, one could propose that the regimen most likely to induce pCR is
the most effective treatment for a particular individual, and that such treatment
could be ideally suited to that patient either as adjuvant or neoadjuvant therapy.
This is the rationale why pCR is a commonly used clinical endpoint for breast
cancer response predictors. However, it is important to realize that most patients
with less than pCR had at least some clinical response to chemotherapy, and
therefore cannot be considered truly resistant to treatment, but rather, cases with
pCR represent extreme chemotherapy sensitivity.
CURRENT CHEMOTHERAPY RESPONSE PREDICTORS
There are several clinical and pathologic variables of breast cancer that are consistently
associated with a greater likelihood of response to neoadjuvant chemotherapy.
These include estrogen receptor (ER)-negative status, smaller tumor
size, increased proliferative activity, and high nuclear or histological grade (9–12).

Gene Expression Profiling to Personalize Chemotherapy Selection 79
Tumor size, grade, and ER-status are routinely available parameters that can be
combined into a multivariate response prediction model (13). One such model
is freely available at our institutional Web site, http://www.mdanderson.org/
Care_Centers/BreastCenter/dIndex.cfm?pn¼448442B2-3EA5-4BAC-
98310076A9553E63. However, clinical variable–based predictions are limited to
predict response to chemotherapy in general and lack regimen specificity.
A large number of molecular markers were proposed as potential predictors
of response to various chemotherapy agents (11,12,14–16). However, currently
there are no molecular markers of response that could be used in the clinic to select
patients for a particular therapy. Biomarker studies were traditionally conducted as
correlative science projects attached to therapeutic trials as optional components.
This has imposed serious limitations on these studies. Sample sizes for marker
discovery were limited by the availability of specimens that has lead to underpowered
observations and introduced potential bias due to subset analysis. Most
commonly, only associations between marker and response were reported, but no
formal prediction models were developed. The methodologies to detect a marker
were rarely standardized, and definitions for test-positive and test-negative cases
often differed from study to study, even for the same marker.
In addition to these technical limitations, there are important theoretical
considerations, which also suggest that single gene molecular markers expressed
by the cancer may have limited predictive value. Factors that reside outside the
neoplastic cells could influence drug response. The importance of interactions
between stroma and neoplastic cells, including neovascularization, changes in
interstitial fluid pressure, and immune/inflammatory response are increasingly
recognized as potential determinants of prognosis and possibly response to
therapy (17,18). Rates of drug metabolism are also variable and can have an
effect on drug activity. Cellular response to toxic insults is a complex molecular
process that involves simultaneous induction of several proapoptotic and survival
pathways, including activation of repair and drug elimination mechanisms
(19,20). Tumor response may depend on the balance of these pathways, and
therefore any single gene may contain only limited information about the
complexity of the mechanisms that determine response to chemotherapy.
GENE EXPRESSION PROFILING AS A NOVEL TOOL
FOR RESPONSE PREDICTION
Gene expression profiling with DNA microarrays represents a relatively recent
tissue analysis tool that was developed in the early 1990s (21). This technology
enables investigators to measure, in a semiquantitative manner in a single
experiment, the expression of almost all mRNAs expressed in a small piece of
cancer tissue. The rationale behind pharmacogenomic a research is that by using
a Pharmacogenomics is defined in this review as the use of mRNA expression profiles of the tumor to
predict drug efficacy. Pharmacogenetics is defined as the study of DNA variations in the germline to
predict toxicity or treatment response.

80 Liedtke and Pusztai
this technology, one can identify individually predictive genes and combine
these into a multivariate prediction model that will be more accurate than any
single gene alone. Multivariate outcome prediction models are not new to
medicine. Independent and individual weakly predictive clinical variables have
been combined into useful and validated prognostic or predictive models. There
are several established statistical methods for combining variables into prediction
models, including the commonly used logistic regression analysis and
the Cox proportional hazards regression model (22). Predictors could be
binary—that is, predicting response versus no response—or could predict the
probability of response on a continuous scale. The Adjuvant Online (http://www
.adjuvantonline.com) software represents a freely available validated multivariate
prognostic prediction tool for breast cancer.
The general strategy for multigene marker development in neoadjuvant
studies is to compare the transcriptional profiles of cancers that responded to
therapy with those that did not and to identify differentially expressed genes that
can be combined into a prediction model. There are several caveats that make
pharmacogenomic predictor discovery challenging.
1. The multiple comparison problem inherent to microarray analysis is well
known. It refers to the large number of variables (i.e., genes) that are
compared between two small data sets. As a consequence, observation of
many small p values may be due to chance. This could result in spuriously
identifying genes as associated with response when in fact they are not.
Several methods used to adjust for this phenomenon, at least to some extent,
by calculating false discovery rates for particular p values were described (23).
2. Another confounder relates to the coordinated expression of thousands of genes.
Individual transcripts do not represent independent variables, but rather their
expression is highly correlated with one another. Many phenotypic characteristics
of breast cancer are also associated with, and likely caused by, the coordinated
expression of thousands of genes. For example, ER-positive cancers
differ from ER-negative tumors in the expression of thousands of genes (24). The
gene expression pattern of high-grade tumors is also different from low-grade
tumors (25). These large-scale gene expression patterns associated with clinical
phenotypes have a profound influence on the pharmacogenomic discovery
process, because ER-negative, high-grade breast cancers are more sensitive
to many different types of chemotherapies compared with ER-positive and
low-grade tumors. Unadjusted pharmacogenomic comparison of responders
with nonresponders will yield gene lists that are dominated by ER- and gradeassociated
genes, and the resulting pharmacogenomic predictor may represent
a predictor of clinical phenotype. When the discovery sample size is large, it
is possible to adjust for these phenotypic confounders; however, in small
studies, these adjustments may be difficult to do.
3. The probability that a supervised pharmacogenomic discovery approach, when
responders are compared with nonresponders, will lead to regimen-specific

Gene Expression Profiling to Personalize Chemotherapy Selection 81
predictors depends on to what extent the response groups are balanced for
strong phenotypic markers and on the extent of molecular differences that
determine drug-specific response. If drug sensitivity is influenced by the
two- to threefold higher or lower expression of a few dozen genes, these
differences may not be readily discovered through supervised pharmacogenomic
analysis of small data sets. The modest gene expression differences
between responders and nonresponders can be masked by the larger scale
molecular differences due to any phenotypic imbalance between the
response groups and the technical noise of microarray experiments. For
example, when 24,000 measurements are performed (e.g., Affymetrix
U133A gene chip) and the overall concordance is 97.98% in a technical
replicate (i.e., same RNA profiled twice), 1.31% of all measurements can
have twofold or greater variation. This means that 314 genes decreased or
increased twofold from one experiment to another due to technical noise
alone (26). If the sample size is large, this technical noise can be separated
from the true signal, but when sample sizes are small it is more difficult to
separate noise from the true signal.
PHARMACOGENOMIC MARKER DEVELOPMENT PROCESS
Ideally, the development process of molecular markers should resemble the
development stages of therapeutic agents (Fig. 1) (27). In this context, a phase I
marker discovery study would demonstrate the technical feasibility of developing a
predictor and assess the reproducibility of marker results. The second phase
Figure 1
Schema of the developmental stages of molecular markers.

82 Liedtke and Pusztai
includes phase II validation studies of statistically adequate power to estimate
more precisely the predictive performance of the test in independent validation
cases. The next level of marker evaluation would include the demonstration of
clinical value by showing that a particular clinical endpoint improves when the
new test is used compared with the current standard, which could be no testing
or another type of testing. This may require a prospective phase III marker
validation study. Currently, no pharmacogenomic response predictors have
completed all three levels of clinical evaluation. Most published studies represent
phase I marker discovery and phase II marker validation studies.
FEASIBILITY OF MICROARRAY-BASED DIAGNOSTIC TESTS
Technical Reproducibility
The reproducibility of microarray measurements has been examined systematically
by the Microarray Quality Control Project (MAQC), which included the US
Food and Drug Administration and 51 academic and industry collaborators. The
goal of this project was to assess the within- and between-laboratory technical
reproducibility of microarray measurements and compare the results across
different microarray platforms. Four different RNA samples were profiled in five
replicates on seven different array platforms, three different laboratories representing
each platform. The same RNA samples were also profiled by three
different reverse-transcription polymerase chain reaction (RT-PCR) methods.
The most important finding of this collaborative effort was that the microarray
measurements were highly reproducible within and across the different commercially
available microarray platforms (28,29). The median coefficient of
variation (CV) for within-laboratory replicates ranged from 5% to 15% for the
various platforms and it was 10% to 20% for between-laboratory replicates. Most
importantly, the concordance between the microarray based mRNA measurements
and RT-PCR results were also high. The correlation coefficients for gene
expression ratios measured by arrays and by RT-PCR in two different samples
ranged from 0.79 to 0.92 for several hundreds of genes.
Between-platform variation in gene expression measurements is due to
sequence variations in the probe sets that target the same gene at different
locations. Factors that determine signal intensities generated by distinct probes
for the same gene include GC content of the sequence, sequence length, crossmatch
opportunities, and the location of the probe sequence in relation to the
3 0 end of the target gene. The concordance between signal intensity for probes
that target the same gene on different platforms is directly related to sequence
homology (30,31). Probes with complete sequence match yield highly concordant
results across platforms.
Few studies examined the within- and cross-platform reproducibility of
multigene prediction scores, as opposed to the reproducibility of single gene
measurements. Two reports indicated high reproducibility of pharmacogenomic

Gene Expression Profiling to Personalize Chemotherapy Selection 83
prediction results in replicate experiments from the same RNA, when the same
experimental procedures were used (26,32). Furthermore, when random noise
was added to corrupt the gene expression data, a remarkable degree of robustness
was observed in the face of noise.
However, caution must be exercised when trying to reproduce prediction
results across different platforms. There are multiple sources of variation that
will decrease reproducibility including differences in probe set and different
normalization methods. Essentially, all published results that attempted crossplatform
testing of gene signatures reported diminished, but not completely
lost, classification accuracy on data generated by other than the original
platform (33).
Preanalytical Sources of Variation
Another potentially important source of variation in microarray data is preanalytical
variations, including differences in tissue sampling and RNA degradation.
There are several different techniques to obtain tissue from a patient for
pharmacogenomic studies, including excisional biopsy (with or without microdissection),
fine needle aspiration (FNA), and core needle biopsy (CNB). Both
needle biopsy techniques yield sufficient RNA for gene expression profiling with
DNA microarrays (1–3 mg) (34–36). However, different sampling methods may
result in different cellular composition and distinct gene expression profiles (37).
FNA specimens contain mostly neoplastic cells with few infiltrating leukocytes,
whereas the tumor cell content of CNB can vary from 10% to 90%. It is
important to consider the different amount of stromal cell that might be present
in a specimen, because the presence or absence of stromal genes might influence
prognostic or predictive gene signatures. Also, in studies comparing pre- and
postchemotherapy samples, an important source of variation may be the different
cellular composition of the specimens.
RNA is an easily degradable molecule and therefore was considered
suboptimal target for marker measurements compared with DNA or proteins.
The development of single step universal RNA stabilizing and preserving
reagents (RNA later, Ambion) has completely changed the field of RNA-based
diagnostics (38). Today, RNA can be stabilized within seconds after a fresh
tissue biopsy has been obtained and stored at room temperature for several days
(if needed) due to readily available commercial RNA stabilizing solutions.
Under these circumstances, RNA degradation from properly collected fresh
tissues is now rare. Interestingly, it is now more cumbersome to preserve proteins
for future analysis than it is to preserve RNA. A single step, universal
protein preserving solution is yet to be discovered. However, it is important to
realize that surgical resection methods (i.e., length of warm ischemia) and the
length of time between removing a piece of tissue from the body and placing it
into RNA preservative have a major effect on RNA quality (39).

84 Liedtke and Pusztai
STATISTICAL DESIGN OF PHARMACOGENOMIC
DISCOVERY AND VALIDATION STUDIES
Clinical biomarker discovery and validation is governed by similar statistical
rules as therapeutic clinical trials (40). Estimation of discovery sample size is
essential to develop a reliable pharmacogenomic predictor, and appropriately
powered independent validation studies are needed to convincingly demonstrate
its predictive values. There are several statistical methods that could be used to
estimate the number of patients needed to develop a clinical outcome predictor
(41). One simple approach is to power the study to discover individual genes that
show a predefined level of differential expression (often expressed in statistical
terms as standard effect size) between outcome groups. However, the number
and extent of differentially expressed genes are often unknown in advance, and
the multiple comparisons (i.e., thousands of different measurements on each
case) represent a major bias in this approach. Another approach takes into
account accumulating preliminary results to estimate the sample size needed for
discovery (42). With multivariate prediction algorithms, it is expected that the
predictive accuracy increases as the predictor is developed from and trained on
larger and larger data sets until it reaches a plateau (i.e., best possible predictor
within the constrains of the technology). The general principle is to fit a learning
curve into existing data by incrementally increasing the training set size and
extrapolate results to larger sample sizes. This is an intuitively appealing
approach, but it requires a substantial amount of preliminary results from many
dozens of cases. It also implies that discovery sample size will vary depending
on the difficulty of separating the two response groups on the basis of gene
expression results. When we applied this method to human breast cancer gene
expression data to estimate the number of cases needed for discovery of a
chemotherapy response predictor, the results indicated that 80 to 100 cases
would yield a predictor as good as extending the sample size to 200 (32).
To assess the accuracy of the predictor, it is necessary to test it on independent
patient cohorts. Often, there are not enough cases available for validation,
and investigators report cross-validation results. In cross-validation, some
samples are left out from the marker development process and later on used to
test the accuracy of the predictor, and this process is repeated many times. This
leave-one-out cross-validation method gives a reliable estimate of the overall
accuracy of the marker development process, but it is not the validation of a
particular predictor. The genes that contribute to the predictors differ from
iteration to iteration. The final single best predictor that is developed from the
entire training data must be tested on completely separate cases to assess its
predictive value. The goal of an independent validation study is to (i) define the
true sensitivity, specificity, and the positive and negative predictive values (PPV
and NPV) with narrow confidence intervals (CI) and (ii) to prove clinical utility
of the test. Since the estimated performance characteristics of the test are known
from the cross-validation results, the design of validation studies is more

Gene Expression Profiling to Personalize Chemotherapy Selection 85
straightforward. Sample size for a single arm validation studies can be calculated
on the basis of the known historical response rate in unselected patients and the
estimated PPV of the genomic predictor. For example, if we assume that the
response rate in unselected patients is 30% and the true PPV of the test is 60%
(two-fold greater chance of response in “test-positive cases”), the sample
size could be calculated as follows: The lower bound of a two-sided 95% CI
should not be less than 0.3 (since the expected maximum pCR rate is 30% in
unselected cases). This implies that the standard error for the PPV is less than 0.1
and therefore the study must include at least 24 “test-positive” patients. The
proportion of patients with response is expected to be around 30%, and an
accurate prediction algorithm should generate a proportion that test positive also
about 0.20 to 0.40, this equates to a total sample size of 80 to 120 patients for
validation. If the predictor performs better and the true PPV is higher than 60%,
fewer patients may be sufficient. A randomized trial design could simultaneously
address response rates in selected and unselected patients and could also assess
treatment specificity of the predictor.
BREAST CANCER PHARMACOGENOMIC RESPONSE PREDICTORS
Supervised Pharmacogenomic Predictors Developed from Human Data
Results from several phase I genomic response marker discovery studies were
reported that compared gene expression profiles of cases with response and noresponse,
and used the genes that were differentially expressed to construct
multigene predictors. Some of these markers were also assessed in small independent
validation cohorts. Table 1 summarizes the currently available literature
in this field. Chang et al. presented the first study that suggested the feasibility of
developing multigene predictors of response to neoadjuvant chemotherapy. They
developed a 92-gene predictor for clinical response to docetaxel using gene
expression data from 20 patients with locally advanced breast cancer (36).
Leave-one-out partial cross-validation indicated a predictive accuracy of 88%
(95% CI, 68–97%), sensitivity of 85% (95% CI, 55–98%), and specificity of 91%
(95% CI, 59–100%). Larger scale independent validation of this marker set has
not yet been reported. Another small study examined 24 breast cancer patients
treated with neoadjuvant paclitaxel, 5-fluoruracil, doxorubicin, and cyclophosphamide
(T/FAC) chemotherapy and developed a 74-gene predictor for
pCR (43). This multigene predictor model, tested on an independent cohort of
12 patients, showed a predictive accuracy of 78% (95% CI, 52–94%), PPV of 100%
(95% CI, 29–100%), NPV of 73% (95% CI, 45–92%), sensitivity of 43% (95%
CI, 10–82%), and specificity of 100% (95% CI, 72–100%). A larger follow-up
study included 82 cases for marker discovery and examined 780 distinct pharmacogenomic
predictors in cross-validation (32). Many of the distinct predictors
demonstrated statistically equal performance in complete cross-validation. This
(text continues on page 91)

86 Liedtke and Pusztai
Table 1 Overview of Published Studies on Pharmacogenomic Predictors of Response to Neoadjuvant Chemotherapy in Breast Cancer
Patient characteristics and
treatment Method/hypothesis
Gene expression
platform Results Comment
Chang et al.,
Lancet 2003
(36)
24 (¼dev.) and 6 (¼val.) pts.
with locally advanced breast
cancer
? 4 cycles 3-weekly
docetaxel
Development,
LOOCV and
independent
validation of a
gene expression
profile
predictive for
clinical response
(25% RD)
HgU95-Av2
(12,625 genes)
92 differentially
expressed genes (p <
0.001) from different
functional classes;
Combination into
prediction model
LOOCV: 88% accuracy,
92% PPV, 83%
NPV, 85% sensitivity,
91% specificity
Independent validation:
correct classification
of six pts. with sensitive
tumors
First study developing
a multigene
predictor for
response to
chemotherapy in
breast cancer
patients
Ayers et al., JCO
2004 (43)
24 (¼dev.) and 12 (¼val.) pts.
with primary breast cancer
? 3 cycles 3-weekly or 12
cycles weekly paclitaxel
followed by 4 cycles of FAC
(T/FAC)
Development and
independent
validation of
gene expression
predictor for
pCR
cDNA arrays
(30,721 genes)
Development of
74-gene-multigene
predictor
Independent validation:
78% accuracy, 100%
PPV, 73% NPV,
43% sensitivity,
100% specificity
No single gene sufficiently
associated
with pCR to serve as
single valid marker
Proofofprinciple
for building multigene
predictors for
chemotherapy
response

Gene Expression Profiling to Personalize Chemotherapy Selection 87
Table 1 Overview of Published Studies on Pharmacogenomic Predictors of Response to Neoadjuvant Chemotherapy in Breast Cancer (Continued )
Patient characteristics and
treatment Method/hypothesis
Gene expression
platform Results Comment
Iwao-Koizumi,
JCO 2005 (44)
44 (¼dev.) and 26 (¼val.) pts.
with stage II/IV breast cancer
? 4 cycles of docetaxel
Development and
independent
validation of
gene expression
predictor for
clinical response
(WHO criteria)
ATAC-PCR (2,453
selected genes)
Development of
85-gene signature
Independent validation:
81% accuracy, 73%
PPV, 91% NPV,
92% sensitivity, 71%
specificity
Genes controlling the
cellular redox
environment associated
with docetaxel
resistance
? supported by
in vitro results
Hannemann et al.,
JCO 2005 (45)
48 pts. with LABC
? cycles of doxorubicin/
docetaxel (AD, n ¼ 24)
or doxorubicin/cyclophosphamide
(AC, n ¼ 24)
Identification of
gene expression
predictor for
pCR or nPCR
for response to
AD and/or AC
Proprietary array
(18,432 genes)
Failure to identify differentially
expressed
genes or define multigene
predictor
Hierarchical clustering:
pCR/npCR samples
did not cluster distinctly
compared
to RD
Comparison of gene
expression between
pre- and
postchemotherapy
samples: more
transcriptional
changes among
sensitive compared
to resistant tumors
Gianni et al., JCO
2005 (46)
89 (¼dev.) and 82 (¼val.) pts.
with LABC
? 3 cycles of 3-weekly
doxorubicin/paclitaxel
followed by 12 cycles
weekly paclitaxel
? 3 cycles 3-weekly or
12 cycles weekly paclitaxel
followed by 4 cycles of
FAC (T/FAC)
Identification of
gene expression
markers that
predict response
Correlation
between RS and
response
RT-PCR (384
genes)
Affymetrix HG-
U133 Plus 2.0
86 genes correlating
with pCR (p < 0.05)
forming three clusters
(ER, proliferation
and immunerelated
gene cluster)
RS correlated with pCR
86-gene model developed
and successfully
validated
(Continued)

88 Liedtke and Pusztai
Table 1 Overview of Published Studies on Pharmacogenomic Predictors of Response to Neoadjuvant Chemotherapy in Breast Cancer (Continued )
Patient characteristics and
treatment Method/hypothesis
Gene expression
platform Results Comment
Folgueira et al.,
Clin Cancer Res
2005 (47)
38 (¼dev.) and 13 (¼val.)
31 (¼dev.) and 13 (¼val.)
? 4 cycles 3-weekly
doxorubicin/cyclophosphamide
(AC)
Development and
independent
validation of
gene expression
predictor for
pCR
692 cDNA
microarray
4,608 gene
platform
25 differentially
expressed transcripts
between responders
and nonresponders,
3 gene classifier could
not be validated;
successful development
and validation
(LOOCV and independent
dataset) of
distinct three-gene
classifier
Dressman et al.,
Clin Cancer Res
2006 (48)
37 pts. with stage IIB/III
breast cancer
? 4 cycles of liposomal
doxorubicin/paclitaxel þ
local whole breast
hyperthermia
Definition of gene
signature for
(a) inflammatory
breast cancer
(IBC), (b) tumor
hypoxia, and (c)
clinical response
and validation in
two independent
datasets
Affymetrix
HG-U133
Plus 2.0
Identification and validation
(LOOCV) of
(a) 22-gene signature
for IBC
(b) 18-gene signature
for tumor hypoxi
Failure to identify gene
signature predicting
clinical response
Development and
successful validation
(2 independent
datasets)ofa
signature predictive
for lymph node
persistence

Gene Expression Profiling to Personalize Chemotherapy Selection 89
Table 1 Overview of Published Studies on Pharmacogenomic Predictors of Response to Neoadjuvant Chemotherapy in Breast Cancer (Continued )
Patient characteristics and
treatment Method/hypothesis
Gene expression
platform Results Comment
Thuerigen et al.,
JCO 2006 (49)
52 (¼dev/GEsDoc) and 48
(¼val./GEDoc) pts. with
breast cancer
? 5 cycles biweekly gemcitabine/epirubicin
followed by
4 cycles biweekly docetaxel
(GEsDoc)
? 6 cycles of 3-weekly
gemcitabine/epirubicin/
docetaxel (GEDoc)
Development and
independent
validation of
gene expression
predictor for
pCR
Oligonucleotide
array (21329
genes)
512-gene-signature
predictive of pCR
with: independent
validation: 88%
accuracy, 64% PPV,
95% NPV, 78%
sensitivity, 90%
specificity
HER2/neu expression
independently
predictive of
response
Park et al., Breast
Cancer Res
Treat 2006 (50)
21 pts. with stage II/III breast
cancer
? 4 cycles of 3-weekly
FEC followed by 12 cycles
of weekly paclitaxel
Development and
LOOCV of
ABC transporter
gene expression
profiles
predictive for
pCR
Affymetrix HG-
U133 Plus 2.0
(49 ABC transporters)
11 differentially expressed
ABC transporters multigene
predictor model of
differentially expressed
ABC transporters predictive
of pCR
LOOCV: 93% accuracy,
93% PPV, 94%
NPV, 88% sensitivity,
96% specificity
Cleator et al.,
Breast Cancer
Res Treat 2006
(51)
40 pts. with primary breast
cancer
? 6 cycles doxorubicin/
cyclophosphamide (AC)
Development and
LOOCV of
predictor for
clinical response
Affymetrix
HG-U133A
253 differentially
expressed genes;
75 and 178 genes overexpressed
in resistant
and sensitive tumors,
respectively
Development and
LOOCV of gene predictor
for response
Previously identified
92-gene docetaxel
classifier (36) predicted
only 39–61%
tumors correctly
? gene expression
pattern for docetaxel
is not predictive for
AC response
(Continued)

90 Liedtke and Pusztai
Table 1 Overview of Published Studies on Pharmacogenomic Predictors of Response to Neoadjuvant Chemotherapy in Breast Cancer (Continued )
Patient characteristics and
treatment Method/hypothesis
Gene expression
platform Results Comment
Hess et al., JCO
2006 (32)
82 (¼dev.) and 51 (¼val.) pts.
with primary breast cancer
? 3 cycles 3-weekly or
12 cycles weekly paclitaxel
followed by 4 cycles of FAC
(T/FAC)
Development and
independent
validation of
gene expression
predictor for
pCR
Affymetrix
HG-U133A
Identification of 30-
gene predictor
Validation fivefold
cross-validation and
permutation testing
Independent validation:
76% accuracy, 52%
PPV, 96% NPV,
92% sensitivity, 71%
specificity
Follow-up study
of (43)
Predictor preserved its
predictive accuracy
when applied to
replicate samples
Largest study to date
Mina et al., Breast
Cancer Res
Treat 2006 (52)
45 pts. with stage II/III breast
cancer (including 6 pts. with
inflammatory BC)
? 3 cycles biweekly doxorubicin
and 6 cycles weekly
docetaxel
Identification of
genes correlating
with pCR
Correlation of pCR
with RS or ERrelated
genes
Identification of
genes correlating
with inflammatory
phenotype
RT-PCR assay
(274 genes)
22 of 274 candidate genes
correlated with pCR (p <
0.05) forming three large
clusters (angiogenesis,
proliferation, and invasion-related
genes)
Lack of correlation
between pCR and
RS/ER-related genes
24/274 genes correlated
with inflammatory BC
Abbreviations: ABC transporter, ATP-binding cassette transporter; dev., development/training set; IBC, inflammatory breast cancer; LABC, locally advanced breast
cancer; LOOCV, leave-one-out cross-validation; npCR, near pCR; NPV, negative predictive value; pCR, pathological complete response; PPV, positive predictive
value; pts., patients; RD, residual disease; RS, recurrence score (Oncotype DX # ) (53); RT-PCR, reverse transcription polymerase chain reaction; val., validation set.

Gene Expression Profiling to Personalize Chemotherapy Selection 91
observation is confirmed by similar reports in the genomic literature, which
indicates that several different but equally good predictors can be developed
from the same data. This is due to the coordinated expression of many thousands
of genes. Any combination from a particular group of coexpressed genes can
carry similar information. However, a nominally best 30-probe set genomic
predictor of pCR (DLDA-30) was selected for independent validation on 51
cases, and its performance was compared with a multivariate clinical variable–
based predictor including age, estrogen receptor status, and grade. The genomic
predictor showed a sensitivity of 92% that was significantly higher than that of
the clinical predictor (61%). The NPV (96% vs. 86%) and area under the
receiving operator curve (0.877 vs. 0.811) were also slightly better. Several other
smaller studies demonstrated similar proof of concept that this supervised
approach to marker development is feasible (Table 1).
Genomic Predictors from Cell Line Models
An interesting alternative approach is to use cancer models to develop chemotherapy
response predictors and test the performance of these in human data. The
gene expression profiles of breast cancer cell lines are different from those of
human breast cancers, but nevertheless, they recapitulate many biological and
genomic properties of primary breast cancers (54). Several groups reported drugspecific
transcriptional changes in response to chemotherapy exposure in vitro
that might be exploited as predictors of therapy response in patients (55).
However, the predictive value of these in vitro generated signatures in patients
has not yet been studied extensively. Nevertheless, this is an important research
direction that if working can have important implications for individualizing
treatment regimens for patients. In fact one could envisage custom-assembled
chemotherapy regimens based on sensitivity profiles for individual drugs. It is
rapidly becoming feasible to test such concepts in silico. Several groups have
made their complete gene expression data from neoadjuvant biopsies publicly
available. Cell line-based predictive signatures can be tested on these already
existing and clinically annotated human data sets (56).
Gene Expression Changes in Serial Biopsies
Similar to cell lines, transcriptional response to chemotherapy was also observed
in serial biopsies before and during chemotherapy. The rationale is that these
changes may be more predictive of response than baseline gene expression
profiles. However, the few small studies that examined this approach reported
very heterogeneous transcriptional response to chemotherapy (57). In essence,
different genes change in different individuals. Sotiriou et al. examined gene
expression changes after one complete cycle of neoadjuvant chemotherapy and
demonstrated that tissue samples, obtained from patients with good response,
exhibited much greater changes in gene expression than those from poor

92 Liedtke and Pusztai
responders (34). These observations were corroborated by results obtained by
Hannemann et al., who compared gene expression profiling on pretreatment
biopsies (n ¼ 46) and posttreatment tumors (n ¼ 17) obtained at time of surgery.
They also demonstrated that neoadjuvant chemotherapy causes major changes in
gene expression in those cases who responded to treatment compared with those
who did not (58,59). Several technical and theoretical concerns limit our
enthusiasm for this approach to predictor development. It is difficult to obtain
serial biopsies at regular intervals from patients who undergo chemotherapy.
There is substantial variation in the actual timing of the biopsy compared with
the planned time of obtaining the follow-up biopsy. For example, biopsies
planned at 24 hours after chemotherapy are in fact usually collected during an
18- to 36-hour window. Since the kinetics of important gene expression changes
are unknown, this introduces unpredictable variation into already highly variable
data. Also, there are altogether missing time points, which further reduce the
already small sample size for discovery. These factors increase the risk for false
discovery from serially collected small pharmacogenomic data sets. More
importantly, the practical value of a response predictor that is applied to a biopsy
after chemotherapy was given is not particularly appealing.
FUTURE POTENTIAL OF PHARMACOGENOMICS
Pharmacogenomic predictive marker discovery is a new science. It is relatively
less mature than the field of prognostic gene signatures because of the lack of
preexisting tissue banks from patients who received preoperative chemotherapy.
All human biopsies had to be collected in prospective clinical trials in the past
few years, unlike the readily available frozen and archived tumor banks that were
suitable for prognostic signature discovery and validation. However, several
such data sets now exist and the field has likely entered an accelerated phase of
development.
We have learned several important lessons. (i) It is clear that the gene
expression profiles of breast cancers with extreme chemotherapy sensitivity are
different from those with lesser sensitivity. Unfortunately, these gene expression
differences correspond to major clinical pathological phenotypes such as ERnegative
status and high grade that somewhat limits the clinical value of the
current predictors. However, as larger data sets become available, it will be
possible to attempt to develop predictors after stratification by known clinical
pathological variables. (ii) It is also clear from many different reports in the
genomic literature that there is no unique, single best signature for any particular
classification problem. There are several distinct predictors that can perform
equally well (60). (iii) The availability of large, public, clinically annotated,
comprehensive gene expression databases started to revolutionize the biomarker
discovery process. It is now possible to test in silico any mRNA-based marker
for prognostic or chemotherapy or endocrine therapy predictive values, respectively,
on data sets that were prospectively assembled for this purpose (61). As

Gene Expression Profiling to Personalize Chemotherapy Selection 93
more neoadjuvant data sets become available from patient cohorts treated with
different treatment regimens, it will be possible to examine the regimen specificity
of the various predictors. (iv) The most important long-term benefit from
pharmacogenomic research may not be the current predictive signatures or the
ones that will be developed in the near future, which still represent relatively
crude attempts to match patients with existing drugs, but rather the large
molecular data sets that this research generates. These databases could allow the
identification of novel drug targets in human data and the development of drugs
that target subsets of patients with particular genomic abnormalities.
REFERENCES
1. National Comprehensive Cancer Network. Available at: http://www.nccn.org. Accessed
June 2007.
2. Goldhirsch A, Wood WC, Gelber RD, et al. Meeting highlights: updated international
expert consensus on the primary therapy of early breast cancer. J Clin Oncol
2003; 21(17):3357–3365.
3. Scholl SM, Fourquet A, Asselain B, et al. Neoadjuvant versus adjuvant chemotherapy
in premenopausal patients with tumours considered too large for breast
conserving surgery: preliminary results of a randomised trial: S6. Eur J Cancer 1994;
30A(5):645–652.
4. Boughey JC, Peintinger F, Meric-Bernstam F, et al. Impact of preoperative versus
postoperative chemotherapy on the extent and number of surgical procedures in
patients treated in randomized clinical trials for breast cancer. Ann Surg 2006; 244(3):
464–470.
5. Makris A, Powles TJ, Ashley SE, et al. A reduction in the requirements for mastectomy
in a randomized trial of neoadjuvant chemoendocrine therapy in primary
breast cancer. Ann Oncol 1998; 9(11):1179–1184.
6. Mauriac L, Durand M, Avril A, et al. Effects of primary chemotherapy in conservative
treatment of breast cancer patients with operable tumors larger than 3 cm.
Results of a randomized trial in a single centre. Ann Oncol 1991; 2(5):347–354.
7. Fisher B, Bryant J, Wolmark N, et al. Effect of preoperative chemotherapy on the
outcome of women with operable breast cancer. J Clin Oncol 1998; 16(8):2672–2685.
8. Kuerer HM, Newman LA, Smith TL, et al. Clinical course of breast cancer patients
with complete pathologic primary tumor and axillary lymph node response to
doxorubicin-based neoadjuvant chemotherapy. J Clin Oncol 1999; 17(2):460–469.
9. Abu-Farsakh H, Sneige N, Atkinson EN, et al. Pathologic predictors of tumor
response to preoperative chemotherapy in locally advanced breast carcinoma. Breast
J 1995; 1(2):96–101.
10. Colleoni M, Minchella I, Mazzarol G, et al. Response to primary chemotherapy in
breast cancer patients with tumors not expressing estrogen and progesterone receptors.
Ann Oncol 2000; 11(8):1057–1059.
11. Chang J, Powles TJ, Allred DC, et al. Biologic markers as predictors of clinical
outcome from systemic therapy for primary operable breast cancer. J Clin Oncol
1999; 17(10):3058–3063.
12. Ellis P, Smith I, Ashley S, et al. Clinical prognostic and predictive factors for
primary chemotherapy in operable breast cancer. J Clin Oncol 1998; 16(1):107–114.

94 Liedtke and Pusztai
13. Rouzier R, Pusztai L, Delaloge S, et al. Nomograms to predict pathologic complete
response and metastasis-free survival after preoperative chemotherapy for breast
cancer. J Clin Oncol 2005; 23(33):8331–8339.
14. Rozan S, Vincent-Salomon A, Zafrani B, et al. No significant predictive value of
c-erbB-2 or p53 expression regarding sensitivity to primary chemotherapy or
radiotherapy in breast cancer. Int J Cancer 1998; 79(1):27–33.
15. Petit T, Borel C, Ghnassia JP, et al. Chemotherapy response of breast cancer depends
on HER-2 status and anthracycline dose intensity in the neoadjuvant setting. Clin
Cancer Res 2001; 7(6):1577–1581.
16. Zhang F, Yang Y, Smith T, et al. Correlation between HER2 expression and response
to neoadjuvant FAC chemotherapy in breast cancer. Cancer 2003; 97(7):1758–1765.
17. Silberstein GB. Tumour-stromal interactions. Role of the stroma in mammary
development. Breast Cancer Res 2001; 3(4):218–223.
18. Coussens LM, Werb Z. Inflammation and cancer. Nature 2002; 420(6917):860–867.
19. Burns TF, El-Deiry WS. Microarray analysis of p53 target gene expression patterns
in the spleen and thymus in response to ionizing radiation. Cancer Biol Ther 2003; 2(4):
431–443.
20. Reed JC. Dysregulation of apoptosis in cancer. J Clin Oncol 1999; 17(9):2941–2953.
21. Fodor SP, Rava RP, Huang XC, et al. Multiplexed biochemical assays with biological
chips. Nature 1993; 364(6437):555–556.
22. Kattan MW, Scardino PT. Prediction of progression: nomograms of clinical utility.
Clin Prostate Cancer 2002; 1(2):90–96.
23. Pounds S, Morris SW. Estimating the occurrence of false positive and false negatives
in microarray studies by approximating and partitioning the empirical distribution of
p-values. Bioinformatics 2003; 19(10):1236–1242.
24. Pusztai L, Ayers M, Stec J, et al. Gene expression profiles obtained from fine-needle
aspirations of breast cancer reliably identify routine prognostic markers and reveal
large-scale molecular differences between estrogen-negative and estrogen-positive
tumors. Clin Cancer Res 2003; 9(7):2406–2415.
25. Sotiriou C, Wirapati P, Loi S, et al. Gene expression profiling in breast cancer:
understanding the molecular basis of histologic grade to improve prognosis. J Natl
Cancer Inst 2006; 98(4):262–272.
26. Anderson K, Hess KR, Kapoor M, et al. Reproducibility of gene expression signature
based predictions in replicate experiments. Clin Cancer Res 2006; 12(6):1721–1727.
27. Pusztai L. Perspectives and challenges of clinical pharmacogenomics in cancer.
Pharmacogenomics 2004; 5(5):451–454.
28. MAQC Consortium, . The MicroArray Quality Control (MAQC) project shows interand
intraplatform reproducibility of gene expression measurements. Nat Biotechnol
2006; 24(9):1151–1161.
29. Klebanov L, Yakovlev A. How high is the level of technical noise in microarray
data? Biol Direct 2007; 2(1):9.
30. Mecham BH, Klus GT, Strovel J, et al. Sequence-matched probes produce increased
cross-platform consistency and more reproducible biological results in microarraybased
gene expression measurements. Nucleic Acids Res 2004; 32(9):e74.
31. Ji Y, Coombes K, Zhang J, et al. RefSeq refinements of UniGene-based gene
matching improve the correlation of expression measurements between two microarray
platforms. Appl Bioinformatics 2006; 5(2):89–98.

Gene Expression Profiling to Personalize Chemotherapy Selection 95
32. Hess KR, Anderson K, Symmans WF, et al. Pharmacogenomic predictor of sensitivity
to preoperative chemotherapy with paclitaxel and fluorouracil, doxorubicin,
and cyclophosphamide in breast cancer. J Clin Oncol 2006; 24(26):4236–4244.
33. Stec J, Wang J, Coombes K, et al. Comparison of the predictive accuracy of DNA
array-based multigene classifiers across cDNA arrays and Affymetrix GeneChips.
J Mol Diagn 2005; 7(3):357–367.
34. Sotiriou C, Powles TJ, Dowsett M, et al. Gene expression profiles derived from fine
needle aspiration correlate with response to systemic chemotherapy in breast cancer.
Breast Cancer Res 2002; 4(3):R3.
35. Assersohn L, Gangi L, Zhao Y, et al. The feasibility of using fine needle aspiration
from primary breast cancers for cDNA microarray analyses. Clin Cancer Res 2002;
8(3):794–801.
36. Chang JC, Wooten EC, Tsimelzon A, et al. Gene expression profiling for the prediction
of therapeutic response to docetaxel in patients with breast cancer. Lancet
2003; 362(9381):362–369.
37. Symmans WF, Ayers M, Clark EA, et al. Total RNA yield and microarray gene
expression profiles from fine-needle aspiration biopsy and core-needle biopsy
samples of breast carcinoma. Cancer 2003; 97(12):2960–2671.
38. Mutter GL, Zahrieh D, Liu C, et al. Comparison of frozen and RNALater solid tissue
storage methods for use in RNA expression microarrays. BMC Genomics 2004; 5(1):88.
39. Dash A, Maine IP, Varambally S, et al. Changes in differential gene expression
because of warm ischemia time of radical prostatectomy specimens. Am J Pathol
2002; 161(5):1743–1748.
40. Simon R. Roadmap for developing and validating therapeutically relevant genomic
classifiers. J Clin Oncol 2005; 23(29):7332–7341.
41. Pusztai L, Hess KR. Clinical trial design for microarray predictive marker discovery
and assessment. Ann Oncol 2004; 15(12):1731–1737.
42. Mukherjee S, Tamayo P, Rogers S, et al. Estimating dataset size requirements for
classifying DNA microarray data. J Comput Biol 2003; 10(2):119–142.
43. Ayers M, Symmans WF, Stec J, et al. Gene expression profiles predict complete
pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and
cyclophosphamide chemotherapy in breast cancer. J Clin Oncol 2004; 22(12):
2284–2293.
44. Iwao-Koizumi K, Matoba R, Ueno N, et al. Prediction of docetaxel response in human
breast cancer by gene expression profiling. J Clin Oncol 2005; 23(3):422–431.
45. Hannemann J, Oosterkamp HM, Bosch CA, et al. Changes in gene expression
associated with response to neoadjuvant chemotherapy in breast cancer. J Clin Oncol
2005; 23(15):3331–3342.
46. Gianni L, Zambetti M, Clark K, et al. Gene expression profiles in paraffin-embedded
core biopsy tissue predict response to chemotherapy in women with locally advanced
breast cancer. J Clin Oncol 2005; 23(29):7265–7277.
47. Folgueira MA, Carraro DM, Brentani H, et al. Gene expression profile associated
with response to doxorubicin-based therapy in breast cancer. Clin Cancer Res 2005;
11(20):7434–7443.
48. Dressman HK, Hans C, Bild A, et al. Gene expression profiles of multiple breast
cancer phenotypes and response to neoadjuvant chemotherapy. Clin Cancer Res
2006; 12(3 pt 1):819–826.

96 Liedtke and Pusztai
49. Thuerigen O, Schneeweiss A, Toedt G, et al. Gene expression signature predicting
pathologic complete response with gemcitabine, epirubicin, and docetaxel in primary
breast cancer. J Clin Oncol 2006; 24(12):1839–1845.
50. Park S, Shimizu C, Shimoyama T, et al. Gene expression profiling of ATP-binding
cassette (ABC) transporters as a predictor of the pathologic response to neoadjuvant
chemotherapy in breast cancer patients. Breast Cancer Res Treat 2006; 99(1):9–17.
51. Cleator S, Tsimelzon A, Ashworth A, et al. Gene expression patterns for doxorubicin
(Adriamycin) and cyclophosphamide (cytoxan) (AC) response and resistance. Breast
Cancer Res Treat 2006; 95(3):229–233.
52. Mina L, Soule SE, Badve S, et al. Predicting response to primary chemotherapy:
gene expression profiling of paraffin-embedded core biopsy tissue. Breast Cancer
Res Treat 2007; 103(2):197–208.
53. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifentreated,
node-negative breast cancer. N Engl J Med 2004; 351(27):2817–2826.
54. Neve RM, Chin K, Fridlyand J, et al. A collection of breast cancer cell lines for the
study of functionally distinct cancer subtypes. Cancer Cell 2006; 10(6):515–527.
55. Troester MA, Hoadley KA, Sorlie T, et al. Cell-type-specific responses to chemotherapeutics
in breast cancer. Cancer Res 2004; 64(12):4218–4226.
56. Potti A, Dressman HK, Bild A, et al. Genomic signatures to guide the use of chemotherapeutics.
Nat Med 2006; 12(11):1294–1300.
57. Buchholz TA, Stivers DN, Stec J, et al. Global gene expression changes during
neoadjuvant chemotherapy for human breast cancer. Cancer J 2002; 8(6):461–468.
58. Hannemann J, Oosterkamp HM, Bosch CAJ, et al. Changes in gene expression
profiling due to primary chemotherapy in patients with locally advanced breast
cancer. Proc Am Soc Clin Oncol 2004; 22(14S):502 (abstract).
59. Modlich O, Prisack HB, Munnes M, et al. Immediate gene expression changes after
the first course of neoadjuvant chemotherapy in patients with primary breast cancer
disease. Clin Cancer Res 2004; 10(19):6418–6431.
60. Fan C, Oh DS, Wessels L, et al. Concordance among gene-expression-based predictors
for breast cancer. N Engl J Med 2006; 355(6):560–569.
61. Andre F, Hatzis C, Anderson K, et al. Microtubule-associated protein-tau is a
bifunctional predictor of endocrine sensitivity and chemotherapy resistance in
estrogen receptor-positive breast cancer. Clin Cancer Res 2007; 13(7):2061–2067.

7
Gene Expression–Based Predictors
of Prognosis and Response to
Chemotherapy in Breast Cancer
Soonmyung Paik
Division of Pathology, NSABP Foundation, Pittsburgh, Pennsylvania, U.S.A.
INTRODUCTION
With the explosion of genomic technology, numerous studies have demonstrated
the potential clinical utility of gene expression profiles as a predictor of prognosis
and response to therapy. Of particular interest is the demonstration that
there is a general link between poor prognostic features and response to chemotherapy.
In addition, linearity of RNA-based assays shows a strong promise
that they can be very useful as predictive assays for targeted therapies.
Since many excellent reviews have been written on this subject, this
chapter will concentrate mainly on conceptual background to aid the clinical
readers understand what is going on in the field rather than going into details.
GENE EXPRESSION PROFILING PROVIDES
PROGNOSTIC INFORMATION
Gene expression profiling using microarray platforms allows interrogation of
expression levels of essentially all known genes in the tumor cells (1). In most
studies, gross tumor rich area is homogenized to extract RNA. Therefore, the
profile that is achieved is mean expression levels of each gene among different
97

98 Paik
cell population in the tumor, including host stromal cells or inflammatory infiltrates.
It is also important to realize that current methods do not have sensitivity to
allow absolute measurement of gene expression levels of all genes printed on the
microarray. Instead only about 50% of genes will have enough signal above
background level (so-called present call). Different microarray platforms use DNA
probes of various lengths (ranging from 22-mer oligonucleotide in Affymetrix
GeneChip, 60-mer oligonucleotides for Agilent arrays, and average 1 kilobase for
cDNA arrrays) representing different positions in each gene between the arrays.
Sensitivity and dynamic range for measuring a specific gene could be widely
different among different array platforms. Thus different studies using different
platforms may result in identifying different set of genes representing same clinical
phenotype (2). Obviously most studies also differ in the study population.
Until recently microarray gene expression profiling required fresh or snap frozen
tissue, seriously limiting the scale of studies conducted so far.
It is not entirely surprising to find gene expression profiling to provide
prognostic information, since after all morphology-based prognostic factors such
as tumor histological grade is essentially end result of gene expression changes in
the tumor cells. Two seminal papers using microarray gene expression profiling,
one by Sorlie et al. (3) and the other by van ’t Veer et al. (4), clearly demonstrated
the association of gene expression profiles and prognosis of patients diagnosed
with early breast cancer. However, these two studies used very different
approaches. In the seminal paper published in 2000, Perou et al. used unsupervised
analysis of gene expression data and identified distinct biological subtypes
(normal-like, luminal A and B, basal-like, and HER2) (5). Subsequently Sorlie
et al. demonstrated that these subtypes confer differing prognosis with basal-like
and HER2 associated with poor clinical outcome (3). On the other hand, van ’t
Veer et al. conducted a supervised analysis to develop prognostic index based on
expression levels of 70 genes identified from microarray analysis (4). This was
then validated in a larger cohort (6). Recently Dr. Perou’s group has demonstrated
that there is a significant overlap of information between the two approaches
using same dataset—meaning those classified as good prognosis using 70-gene
van ’t Veer set was found to be largely populated by luminal A subtype (2).
While studies described above had enormous impact on how we view
breast cancer, despite recent clearance of 70-gene Mammaprint 1 assay by the
U.S. Food and Drug Administration (FDA), actual clinical implementation has
been impeded by the need for fresh or snap frozen tumor tissue because of the
requirement for highest quality RNA as a starting material for the microarray
gene expression profiling.
DEVELOPMENT AND VALIDATION OF OncotypeDx TM
RECURRENCE SCORE ASSAY
RNA isolated from formalin-fixed, paraffin-embedded tissue (FFPET) samples
is a poor material for gene expression profiling. Masuda et al. analyzed the RNA
extracted from FFPET for chemical modifications (2,7). In freshly made FFPET

Predictors of Prognosis and Response to Chemotherapy 99
that were fixed and processed in ideal conditions (fixed in 10% buffered formalin
at 48C), RNAs are fairly well preserved. Although the extracted RNAs did not
show any sign of degradation compared with fresh samples, they were poor
substrates for cDNA synthesis and subsequent polymerase chain reaction (PCR)
amplification, so that only PCR amplification of short targets was possible. The
investigators found addition of monomethylol (CH 2 OH) groups to all four bases
to varying degrees and some adenines dimerized through metheylene bridging.
In addition to the chemical modification, however, RNAs in FFPET continue to
be degraded or fragmented over time during storage for reasons unclear. Cronin
et al. systematically examined the quality of RNA extracted from breast cancer
FFPET specimens taken at different times. RNA from FFPET archived for
approximately 1 year were less fragmented than RNA archived for nearly 6 or
17 years (8).
There have been reports on the use of reverse transcription polymerase
chain reaction (RT-PCR) to measure expression levels of few genes using RNA
isolated from FFPET, but large-scale RT-PCR was not reported until recently.
Cronin et al. from Genomic Health Inc. has invested on developing high quality
controlled robotic assays for RT-PCR and published proof of concept study
(8). In the latter study, while RNAs isolated from old formalin fixed tissue were
heavily fragmented with average size under 200 base pairs and absolute signal
was much less for older samples, normalization of signal using a set of reference
genes could largely correct these problems.
Encouraged by these results, the National Surgical Adjuvant Breast and
Bowel Project (NSABP) has engaged in a collaborative study with Genomic
Health Inc. to develop prognostic gene expression assay for node negative (N )
estrogen receptor–positive (ERþ) breast cancer treated with tamoxifen (9).
While there are many other outstanding clinical questions that need to be
addressed, we were compelled to first address the question of who among N
ERþ tamoxifen treated patients need chemotherapy. In addition, there were two
independent tamoxifen treated cohorts (tamoxifen arms of B-14 and B-20 trials)
in the NSABP tissue bank so that discovery and validation could be performed
without difficulty. We wanted to make sure that by the time of reporting, the
assay was validated in a completely independent cohort. At that time, technology
and cost limited Genomic Health investigators to examine only 250 genes
maximum in RNA extracted from 60 mm thick section from a paraffin block.
Therefore, existing cancer literature and microarray data were surveyed and
250 candidate genes were selected. Before tumor blocks from B-20 tamoxifen
treated cohort were examined, two other cohorts, one from mixed patient
population and the other from those with more than 10 positive nodes were
examined with the same assay (10). We did not believe that there will be different
prognostic genes based on nodal status and therefore selected the prognostic
genes that were equally prognostic across all three cohorts as a basis for
prognostic model building. From the beginning, because of the continuous nature
of RNA expression data, we believed strongly that a linear prediction model
needed to be developed.

100 Paik
Correlating clinical outcome with expression levels yielded many genes
from these three studies, and 16 top-performing genes were identified for final
model building and validation. Relative expression levels of the 16 genes are
measured in relationship to average expression levels of 5 reference genes.
While a majority of genes comprising 16 genes are ER (ER, PGR, BCL2,
SCUBE2) and proliferation (Ki67, STK15, Survivin, CCNB1, MYBL2) related,
there are other genes (HER2, GRB7, MMP11, CTSL2, GSTM1, CD68, BACG1)
(9). The unscaled recurrence score (RSu) was calculated with the use of coefficient
that are defined on the basis of regression analysis of gene expression.
Recurrence score (RS) was rescaled from the unscaled recurrence score (Rsu) as
follows: RS ¼ 0 if Rsu < 0; RS ¼ 20 (Rsu–6.7) if 0 Rsu 100; and RS ¼
100 if Rsu > 100 (9). This resulted in RS ranging from 0 to 100.
Final validation of the RS was achieved by examination of its performance
in a completely independent cohort from NSABP trial B-14, which was not used
in the model building process (9). The validation study was conducted with
rigorous predefined statistical analysis plan with prespecified outcome end
points and cutoffs for RS. The predefined low-risk group (RS below 18) had
significant better prognosis compared with higher risk groups. More importantly,
as shown in the published report, there is a linear relationship between
RS and risk of distant failure at 10 years among N ERþ tamoxifen-treated
patients (9).
To test the applicability of RS in community oncology setting, Habel et al.
conducted a nested case-control study in breast cancer patients diagnosed from
1985 to 1994 at 14 Kaiser hospitals (11). Eligibility included negative nodes, age
less than 75 years, and no chemotherapy. Cases died of breast cancer prior to
2002. Up to three controls were matched to each case on age, race, tamoxifen
treatment, facility, diagnosis year, and follow-up time. Among 4964 potentially
eligible patients, 220 cases and 570 matched controls were identified. Median
age was 59 years (range, 28–74 years); 30.9% were tamoxifen treated (mostly
after 1988). Median follow-up time was 4.9 years for cases (time to death) and
12.9 years for controls. RS was significantly associated with breast death in ERþ
patients treated with or without tamoxifen (p ¼ 0.002). This study is important
since it provides validation of RS assay used in a community setting.
PREDICTION OF RESPONSE TO CHEMOTHERAPY
Among gene expression–based markers, only RS has been tested whether it
predicts response to chemotherapy in adjuvant setting (12). Since many of the
genes comprising 21 genes in the RS assay are ER or proliferation related, we
hypothesized early on that RS would be predictive of response to chemotherapy.
Existence of chemotherapy arm in trial B-20 allowed us to ask this
question (in fact this exactly was the reason why we first decided to work with
N ERþ tamoxifen-treated cohort since we knew that we could not only discover
and validate the prognostic index, but also test whether it predicts response

Predictors of Prognosis and Response to Chemotherapy 101
to chemotherapy using banked materials readily available to us) (12). There were
651 evaluable patients (227 randomized to tamoxifen and 434 randomized to
tamoxifen plus chemotherapy). Patients with tumors who had high RS greater
than 30 had a large absolute benefit of chemotherapy [with an absolute increase
in distant recurrence free survival (DRFS) at 10 years of 27.6 8.0%, mean
SE]. Patients with tumors who had low RS less than 18 derived minimal, if any,
benefit from chemotherapy (with an absolute increase in DRFS at 10 years of
–1.1 2.2%, mean SE). The test for interaction between chemotherapy
treatment and RS was statistically significant (p ¼ 0.038). Of particular
importance is the finding that there is a linear relationship between RS and
degree of benefit from chemotherapy when RS was examined as a continuous
variable in Cox Model.
While the relationship between RS and chemotherapy response has not
been validated in an independent cohort as was done for prognostic aspect of the
assay, supportive evidences exist to allow its use in clinical decision making.
The most important data are from a cohort of patients treated with neoadjuvant
chemotherapy at Milan reported by Gianni et al. (13). In this cohort, higher RS
correlated with higher incidence of complete pathological response (pCR). In
addition, due to the fact that many genes in RS are proliferation and ER related, we
can make reasonable assumption that B-20 data are true. Since the decision to get
chemotherapy has to be based on evaluation of baseline risk and assessment of
the degree of potential benefit from adding chemotherapy, it is reasonable to
conclude that those with lower RS may not need chemotherapy since their
baseline risk is low on the basis of B-14 data, and expected benefit is low on the
basis of B-20 data. One headache is the fact that because of the continuous nature
of the RS, it is difficult to pinpoint from which RS score patients start to gain
benefit from chemotherapy. Clinical oncologists are not used to this kind of data.
Uncertainties regarding the degree of benefit from chemotherapy in patients with
intermediate range of RS resulted in launching a large randomized clinical trial
named Trial Assigning IndividuaLized Options for Treatment (TAILORx) by
North American Breast Cancer Intergroup (14). Many questioned about the
reasons for the definition of the intermediate risk group in TAILORx trial being
different from the definition used in the original publication in B-14 cohort. This
was due to the fact that primary clinical end point in TAILORx is disease-free
survival instead of distant disease-free survival and consideration of 95% confidence
intervals in the baseline risk assessment. Again one has to realize that
these cutoffs are totally arbitrary.
PREDICTION OF RESPONSE TO NEOADJUVANT CHEMOTHERAPY
Preoperative systemic therapy trials provide unique opportunities to assess
sensitivity of tumor cells to systemic therapy in vivo while treatment is ongoing.
Numerous studies have demonstrated the correlation of pCR and long-term
outcome in breast cancer patients, potentially eliminating the need for long-term

102 Paik
followup. Together with ease of obtaining fresh pretreatment core biopsies
allowing high throughput assessment of molecular markers, preoperative trials
may allow speedy discovery and validation of predictive markers for response to
systemic therapy. So it is not surprising that many investigators attempted to
utilize neoadjuvant setting to develop predictive markers of response to chemotherapy
using microarray-based gene expression profiling. Regrettably the
cumulative experience has so far been somewhat disappointing. No profile has
been found to have enough accuracy to be directly clinically applicable, although
Hess et al. demonstrated that gene expression–based predictor still provided
better prediction than clinical predictors (15). But more importantly, there has
been no study with large enough sample size to provide conclusive answers as to
whether one could reliably predict pCR using gene expression levels.
It probably is somewhat naïve to expect gene expression profile alone to
predict pCR. For example, one would expect a smaller tumor would have higher
chance of getting into pCR than a bigger tumor with same gene expression
phenotype. Host factors such as drug metabolism and distribution may influence
the chance for pCR. Prediction of combination regimens also may be more
difficult than predicting response to single regimen. However, in the latter situation,
because of low pCR rate, it is very difficult to develop predictor unless
sample size is very high. In this regard, perhaps what investigators have achieved
so far is as far as they can get, unless these other factors are integrated into
algorithm development. In addition, pCR itself is not a perfect measure of
response either. As many as 80% of tumors show clinical response on taxanebased
neoadjuvant chemotherapy, and therefore among those not achieving pCR,
there is a continuous spectrum of residual burden in the breast. So it may make
more sense to correlate gene expression with residual tumor burden than pCR,
which has a somewhat arbitrary cutoff.
Despite these limitations, certain consensus has emerged that are in
agreement with what we found in B-20 study—that the tumors with poor
prognostic signatures have higher chance of responding to chemotherapy. In
addition to above described study by Gianni et al. (13), Rouzier et al. (using gene
expression array) (16) and Carey et al. (using IHC surrogates of intrinsic subtypes)
(17) has examined whether Perou’s intrinsic subtypes correlate with
chance of pCR. As expected from B-20 data, basal and HER2 subtypes have
higher chance of pCR compared to luminal subtypes. So there is a general
agreement in the field that genetic program that determines the prognosis is
tightly linked with general sensitivity to chemotherapy.
ASSESSMENT OF RESIDUAL RISK AFTER CHEMOTHERAPY
Despite the fact that gene expression data alone have not been shown to be able
to predict pCR with high accuracy, gene expression profiling can be still
important in patients treated with neoadjuvant chemotherapy. In an unpublished
study, using pretreatment core biopsy samples from NSABP B-27, we found that

Predictors of Prognosis and Response to Chemotherapy 103
the combination of prognostic profile with pCR may be used to define residual
clinical risk after chemotherapy, since those with pCR had a very good prognosis
despite having a poor prognostic gene signature. This means that patients with
poor prognostic gene signature who did not achieve pCR are at high risk of
recurrence after chemotherapy. Such data will allow identification of patients
who will require more than chemotherapy without the need for waiting until they
develop recurrences and opens door for a concept for post-neoadjuvant targeted
therapy trials.
CLINICAL UTILITY OF RS ASSAY
How Is RS Compared to NCCN or St. Gallen Clinical Practice Guidelines?
In B-14 study, less than 10% of the tamoxifen treated patients are classified as
low risk by either NCCN or St. Gallen clincial guidelines. These patients did
have excellent prognosis with less than 10% distant failure within 10 years. In
contrast, RS classified 51% of the same cohort into low-risk category with very
similar 10-year failure rate. Therefore 40% more ERþ tamoxifen-treated
patients would be spared from chemotherapy if RS is used in comparison to
NCCN or St. Gallen clinical guidelines. In addition, unlike NCCN and St. Gallen
guidelines, which assumed that treatment benefit is equal in all subgroups,
NSABP B-20 data showed that the benefit from chemotherapy is nearly zero in
low-risk group defined by RS.
How Is RS Compared to Adjuvant on Line Algorithm?
The high cost of the RS assay (estimated at $3400) can only be justified by the
additional information that it provides compared with other assays. Bryant and
colleagues, in collaboration with Peter Ravdin, compared RS with Adjuvant on
Line (AOL) in the NSABP B-14 cohort in a presentation made at the 2005
St. Gallen Breast Cancer Symposium (personal communications with the
late Dr. John Bryant). The AOL program provides individualized estimates of
recurrence risk on the basis of information such as patient age, tumor size, nodal
involvement, and histological grade. It is actually not that straightforward to
compare RS and AOL, since AOL does not provide risk scores. Therefore, it is
difficult to utilize the numerical output from AOL when comparisons are made
to RS. To compare the two, Dr. Bryant rank ordered output from AOL so that
similar proportion of patients belong to low- (50%), intermediate- (25%), and
high-risk (25%) categories as defined by RS. Rank-ordered AOL (R-AOL)
performed remarkably well as a prognostic factor (in B-14) as well as predictive
factor for benefit from chemotherapy (in B-20). RS and R-AOL were both
independently prognostic in B-14 in multivariate analysis, suggesting that both
gene expression signature and clinical features influence patient prognosis.
However, agreement between the two assays was only 48%. RS was able to tease

104 Paik
out patients with poor prognosis from the R-AOL low-risk category, so that the
10-year distant relapse rate was 5.6% for patients categorized as low risk by both
R-AOL and RS. R-AOL low-risk but RS intermediate- and high-risk patients had
a 12.9% distant relapse rate at 10 years in the tamoxifen-treated cohort (p ¼
0.04). For R-AOL intermediate- or high-risk patients, the distant relapse rate was
8.9% when the RS risk was low, whereas the distant relapse rate was 30.7%
when the RS risk was intermediate or high (p < 0.0001). These data suggest that
the RS adds to existing prognostic markers and it would be ideal to incorporate
RS and AOL into a single algorithm. However, at this point we don’t have a
clear way to actually implement this finding into AOL in clinical use.
Applicability in Node-Positive Patients
One question about RS is whether it will work for Nþ ERþ patients. There is no
reason why it should not work when it actually works even in 10þ node patients.
However, risk estimates for RS will be different between N and Nþ patients
although overall direction of the data would be the same, so one cannot simply
apply the current RS algorithm to Nþ patients without a study.
Categorical Vs. Continuous Variable
One of the biggest problems in the field currently in my view is the use of RS as
a categorical variable (low-, intermediate-, and high-risk categories), which was
arbitrarily defined for validation study to ease the understanding of the data,
because clinicians are not used to seeing a continuous data (9). It is very
important to recognize the continuous nature of the RS. For example, a patient
with score 17 (categorized as low risk) has nearly identical risk as a patient with
score 19 (categorized as intermediate risk). On the other hand, a patient with RS
of 3 would have much different prognosis than the one with RS of 17, although
they are both categorized as low risk in the published report.
FUTURE PERSPECTIVE
The problem of current gene expression profiling methods is the need for fresh or
snap frozen tissue and/or high cost. There are alternative methods that may be
able to overcome these barriers. For example, Quantiplex assay (Panomics)
based on combination of branched DNA amplification and fluorescence-labeled
beads may allow multiplexing as many as 30 genes at a time in a 96 well plate
with reasonable amount of input RNA extracted from FFPET. cDNA-mediated
annealing, selection, extension, and ligation (DASL) assays from Illumina
is another promising example (18). These assays cost fraction of the cost for
RT-PCR or microarray-based methods and can utilize FFPET as starting materials.
Replacement of standard markers such as ER and HER2 by these assays or
QRT-PCR is desirable since they will provide more linear prediction of response.

Predictors of Prognosis and Response to Chemotherapy 105
Availability of gene expression profiling methods for FFPET opens the
possibility that archived materials from adjuvant chemotherapy trials can be
analyzed, and therefore current approach of using gene expression analysis in
neoadjuvant setting as a mean to identify predictive markers will become less
important. The latter approach always will require validation in adjuvant setting
since the FDA will not approve marker based on pCR as a surrogate—therefore
it makes more sense to perform discovery and validation of predictive markers
purely in the adjuvant setting.
REFERENCES
1. Schena M, Shalon D, Davis RW, et al. Quantitative monitoring of gene expression
patterns with a complementary DNA microarray. Science 1995; 270:467–470.
2. Fan C, Oh DS, Wessels L, et al. Concordance among gene-expression-based predictors
for breast cancer. N Engl J Med 2006; 355:560–569.
3. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas
distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A
2001; 98:10869–10874.
4. van ’t Veer LJ, Dai H, van de Vijver MJ, et al. Gene expression profiling predicts
clinical outcome of breast cancer. Nature 2002; 415:530–536.
5. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours.
Nature 2000; 406:747–752.
6. van de Vijver MJ, He YD, van ’t Veer LJ, et al. A gene-expression signature as a
predictor of survival in breast cancer. N Engl J Med 2002; 347:1999–2009.
7. Masuda N, Ohnishi T, Kawamoto S, et al. Analysis of chemical modification of RNA
from formalin-fixed samples and optimization of molecular biology applications for
such samples. Nucleic Acids Res 1999; 27:4436–4443.
8. Cronin M, Pho M, Dutta D, et al. Measurement of gene expression in archival
paraffin-embedded tissues: development and performance of a 92-gene reverse
transcriptase-polymerase chain reaction assay. Am J Pathol 2004; 164:35–42.
9. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifentreated,
node-negative breast cancer. N Engl J Med 2004; 351:2817–2826.
10. Cobleigh MA, Tabesh B, Bitterman P, et al. Tumor gene expression and prognosis in
breast cancer patients with 10 or more positive lymph nodes. Clin Cancer Res 2005;
11:8623–8631.
11. Habel LA, Shak S, Jacobs MK, et al. A population-based study of tumor gene
expression and risk of breast cancer death among lymph node-negative patients.
Breast Cancer Res 2006; 8:R25.
12. Paik S, Tang G, Shak S, et al. Gene expression and benefit of chemotherapy in women
with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol 2006;
24:3726–3734.
13. Gianni L, Zambetti M, Clark K, et al. Gene expression profiles in paraffin-embedded
core biopsy tissue predict response to chemotherapy in women with locally advanced
breast cancer. J Clin Oncol 2005; 23:7265–7277.
14. Sparano JA. TAILORx: trial assigning individualized options for treatment (Rx).
Clin Breast Cancer 2006; 7:347–350.

106 Paik
15. Hess KR, Anderson K, Symmans WF, et al. Pharmacogenomic predictor of sensitivity
to preoperative chemotherapy with paclitaxel and fluorouracil, doxorubicin, and cyclophosphamide
in breast cancer. J Clin Oncol 2006; 24:4236–4244.
16. Rouzier R, Perou CM, Symmans WF, et al. Breast cancer molecular subtypes respond
differently to preoperative chemotherapy. Clin Cancer Res 2005; 11:5678–5685.
17. Carey LA, Dees EC, Sawyer L, et al. The triple negative paradox: primary tumor
chemosensitivity of breast cancer subtypes. Clin Cancer Res 2007; 13:2329–2334.
18. Bibikova M, Talantov D, Chudin E, et al. Quantitative gene expression profiling in
formalin-fixed, paraffin-embedded tissues using universal bead arrays. Am J Pathol
2004; 165:1799–1807.

8
Utilization of Genomic Signatures for
Personalized Treatment of Breast Cancer
P. Kelly Marcom, Carey K. Anders,
Geoffrey S. Ginsburg, and Anil Potti
Duke Institute for Genome Sciences & Policy, Department of Medicine,
Duke University Medical Center, Durham, North Carolina, U.S.A.
Joseph R. Nevins
Duke Institute for Genome Sciences & Policy, Department of Molecular
Genetics and Microbiology, Duke University Medical Center, Durham,
North Carolina, U.S.A.
INTRODUCTION
The challenge of personalized cancer treatment lies in matching the most effective
therapy to the characteristics of an individual patient’s tumor. The selection
of therapy for both early- and late-stage breast cancer, as is the case for most
cancer therapy, is still largely empiric and guided by large, population-based,
randomized clinical trials. Although widely adopted, this approach is inadequate for
the selection of individualized chemotherapy regimens. Estimates of benefit are
extrapolations from the effects observed in these large trials and are not necessarily
applicable to individual patients.
The use of genomic data, in particular gene expression profiles derived from
DNA microarray analysis, has the promise of providing the basis for truly individualized
therapy. This approach has the potential to transform the treatment of
107

108 Marcom et al.
cancer from a population-based approach to an individualized one directed toward
unique patient characteristics. This includes the use of gene expression data to
identify subtypes of cancer not previously recognized by traditional methods of
analysis. Profiling genomic patterns to identify new subclasses of tumors has
previously been described in distinguishing differences between acute myeloid
leukemia and acute lymphoblastic leukemia (1,2). Similar studies have employed
gene expression data to identify distinct subtypes of breast cancer with prognostic
significance (3).
A landmark study in breast cancer published in 2002 described the use of a
DNA microarray-based, 70-gene expression profile as a prognostic factor in
breast cancer (4). This genomic marker has demonstrated predictive value
stratifying stage I and II breast cancer patients into two broad subgroups, relatively
high-risk versus low-risk groups as defined by a 10-year risk of breast
cancer recurrence. Various other studies have described the development of
expression profiles that have prognostic value in breast cancer (5–9).
The Netherlands 70-gene predictor has provided the basis for an expanded
European study that aims to measure its performance in a more diverse set of
patients (10). The parallel, prospective clinical trial MINDACT (Microarray In
Node-negative Disease may Avoid ChemoTherapy) aims to measure the effectiveness
of this genomic predictor in guiding adjuvant chemotherapy when
compared to predictions based solely on the traditional clinicopathologic features.
Likewise, a large prospective trial in the United States, TAILORx (Trial
Assigning IndividuaLized Options for Treatment), is currently evaluating the
Oncotype DX test (Genomic Health, Inc., Redwood City, California, U.S.),
a polymerase chain reaction (PCR)-based assay of 21 genes correlated with
10-year breast cancer recurrence risk (11), to guide the administration of hormonal
therapy with or without chemotherapy among patients with estrogen
receptor (ER)-positive, early-stage breast cancer.
Of equal, if not greater, importance in achieving the goal of personalized
treatment of breast cancer is the ability to predict response to specific chemotherapeutics,
particularly the standard-of-care cytotoxic regimens incorporated
into routine clinical practice. Various studies have described approaches to
developing genomic predictors of response to chemotherapeutic agents promising
a more effective strategy of assigning chemotherapeutic regimens to
patients, as guided by unique characteristics of their breast tumors (12–14).
The overall aim of this chapter is to highlight both opportunities presented
by recent advances in the development of genomic-based approaches within the
field of breast cancer and corresponding strategies for implementation of these
techniques into clinical practice. Rather than relying on historic paradigms of
chemotherapeutic selection as guided by broad, population-based studies, the
goal of this unique methodology is to improve patient outcome, and ultimately
patient survival, through the optimal selection of patient-specific chemotherapeutic
agents as guided by genomic-based strategies.

Genomic Signatures for Personalized Treatment of Breast Cancer 109
CYTOTOXIC THERAPIES FOR EARLY-STAGE BREAST CANCER
Currently, the standard regimens for the adjuvant treatment of early-stage breast
cancer combine anthracyclines, taxanes, and cyclophosphamide in limited permutations
of combination, dose, and administration schedule. The addition of a
taxane, either sequentially or in combination, to a core anthracycline/cyclophosphamide
regimen in the setting of node-positive breast cancer has been
shown to improve disease-free survival (DFS) by 4% to 7% in four separate
randomized trials (15–18). However, it is unclear whether all breast cancer
subtypes are equally affected by this therapeutic approach. Recent work examining
a series of breast cancer adjuvant trials conducted over the past two decades
further confirms the importance of defining breast cancer subtypes and their
relationship to treatment benefit. This work demonstrated a lack of benefit for
this broad treatment approach across a subset of ER-positive patients (19).
Optimizing chemotherapy outcomes through the manipulation of dosing
schedules remains controversial. In CALGB 9741 (Cancer and Leukemia Group
B Intergroup 9741), decreasing the dosing interval from every three weeks to
every two weeks [i.e., dose-dense (DD) scheduling] was shown to improve DFS
and overall survival (OS) at three-year median follow-up [risk ratio (RR) ¼ 0.74,
p ¼ 0.010] (20). However, an update at five-year median follow-up showed a
less impressive overall improvement in DFS (RR ¼ 0.80, p ¼ 0.012) and
perhaps no improvement in ER-positive disease (21). Moreover, a randomized
Phase III trial from the Gruppo Oncologico Nord Ovest–Mammella Inter
Gruppo (GONO–MIG) examining DD scheduling of a non-taxane-containing
regimen did not show a benefit in DFS or OS (22). In an accompanying editorial
to this publication, the ambiguity regarding the benefit of DD therapy
was acknowledged and emphasis was placed on the need to define biologically
relevant breast cancer subtypes so that individualized treatment strategies could
be developed (23).
A ROLE FOR PREOPERATIVE SYSTEMIC CHEMOTHERAPY
TO EVALUATE GENOMIC STRATEGIES
Preoperative, systemic chemotherapy (PST), also known as neoadjuvant therapy,
is an effective means for assessing the sensitivity of an individual breast tumor
to chemotherapy. Multiple trials of preoperative systemic chemotherapy have
been performed. Initially, this approach was primarily used for locally advanced,
inoperable, or inflammatory breast cancer. However, subsequent trials have
evaluated PST in the setting of smaller, operable breast tumors both to increase
breast conservation rates and to assess sensitivity to chemotherapy.
Over the past two decades, PST for early-stage breast cancer has been
shown to be a safe and effective method for assessing the chemosensitivity of
primary breast cancers. Sequencing systemic therapy first, with the consequent

110 Marcom et al.
three- to six-month delay in surgery, has not been shown to adversely affect
outcomes (24). Assessment of pathologic response to PST, both in the breast and
regional lymph nodes, provides the most powerful prognostic information
available. Patients who obtain a complete pathologic response in the breast and
lymph nodes have a greater than 95% survival at seven years (24,25). Although
PST provides a powerful tool for in vivo assessment of chemosensitivity,
exploiting the information gained in the aforementioned trials to improve
management of individual patients has been more difficult.
Studies employing a PST approach have yielded conflicting results illustrating
the heterogeneity of primary breast tumors and responsiveness to different
chemotherapeutic agents. This observation begs the question on whether
or not breast tumors commonly possess intrinsic ‘pan-chemotherapy’ responsiveness
or resistance. For example, patients responsive to preoperative treatment
with CVAP (Cytoxan, vincristine, Adriamycin, and prednisone) were
either continued on the same regimen or crossed over to docetaxel (26).
Responding patients who received docetaxel had a significantly higher pathologic
complete response (pCR) rate than those who continued on the CVAP
regimen, supporting differential breast tumor sensitivity to anthracyclines and
taxanes. Additionally, the pCR rate in the NSABP (National Surgical Adjuvant
Breast and Bowel Project) B-27 clinical trial was doubled by the addition of
preoperative docetaxel to preoperative doxorubicin/cyclophosphamide (AC)
chemotherapy (26.1% vs. 12.9%, p < 0.0001) supporting this conclusion.
However, the addition of preoperative docetaxel did not yield a statistically
significant difference in overall DFS when added to preoperative AC alone
[hazard ratio (HR) ¼ 0.9, p ¼ 0.22]. The question remains open on whether
patients only partially responsive to AC, who achieved a pCR following
administration of docetaxel, might have achieved a pCR with docetaxel alone,
thus avoiding the risk of cardiotoxicity inherent to anthracyclines. Unfortunately,
this study is somewhat flawed by the concurrent administration of tamoxifen
with chemotherapy in patients with known ER-positive breast tumors (25). An
additional study by the GEPARDUO (German Preoperative Adriamycin Docetaxel)
study compared DD doxorubicin/docetaxel (AT) with AC followed by
docetaxel, illustrating improved response among patients receiving cyclophosphamide.
This study, however, was more widely interpreted as a comparison
of sequential versus DD therapy (27).
The largest trials investigating response to PST with AC come from the
NSABP B-18 and the aforementioned B-27 clinical trials, where the pCR rates,
allowing for residual in situ disease, were consistently approximately 13%
(25,28). Data for single-agent paclitaxel are more limited; however, a study
conducted at the M.D. Anderson Cancer Center reported pCR rates of 6% when
excluding in situ diseases (29). A comprehensive review of these and other PST
trials supports that the pCR rate in the breast with most doublet chemotherapies
is approximately 10% to 15% (24). The National Comprehensive Cancer

Genomic Signatures for Personalized Treatment of Breast Cancer 111
Network (NCCN) guidelines support that standard adjuvant regimens can also be
used as preoperative therapy (http://www.nccn.org).
Although response to PST is a powerful tool for assessing tumor chemosensitivity,
it is a post hoc evaluation and requires the exposure of individual
patients to toxic and potentially ineffective therapies. Consequently, it has been
difficult to exploit the knowledge gained about both tumor sensitivity and
resistance to the patient’s benefit. In addition, to our knowledge, no studies have
examined the influence of PST on the postoperative management of patients who
have minimal or no response to PST.
The relative importance of including either an anthracycline or taxane or
both as part of the adjuvant treatment regimen of early stage breast cancer is an
area of active research. These two classes of agents, in combination with
cyclophosphamide, provide the foundation for the treatment of early-stage breast
cancer, and several clinical trials provide insight into the current use of these
agents. A recently published randomized phase III trial compared AC with
docetaxel/cyclophosphamide (TC) (30). While this trial demonstrated improvement
in DFS favoring TC (86% vs. 80%, p ¼ 0.015), the confidence interval (CI)
approached 1.0 (CI ¼ 0.50–0.94) and there was no difference in OS despite
median follow-up of 5.5 years. Interestingly, a previous trial comparing AT with
AC showed no difference in DFS at four years (31). Although both of these trials
enrolled a lower-risk population of patients, reconciling differences in outcome
across these two studies is difficult. An ongoing trial in patients with fewer than
four positive axillary nodes is comparing AC with single-agent paclitaxel as
adjuvant therapy (CALGB 40101), in part on the basis of responses observed in
the aforementioned PST trials. In summary, although many active chemotherapeutic
agents suitable for the treatment of early-stage breast cancer exist,
data available in 2007 do not provide the necessary guidance to select the most
effective chemotherapeutic agent for each unique patient. This observation
speaks to the importance of developing novel methods employing genomic
analyses of tumor biology to individualize and optimize patient care.
GENE EXPRESSION PROFILES TO PREDICT THERAPEUTIC RESPONSE
Given the current status of neoadjuvant and adjuvant chemotherapy for breast
cancer, an opportunity exists to use genomic information to improve the selection
of patients who would derive the greatest benefit from treatment with cytotoxic
chemotherapy. That is, prognostic tools that can identify patients clearly in need of
further treatment can help to refine the decisions based now on clinical information.
Additionally, the development of predictive signatures specific for
response to individual cytotoxic agents would provide an invaluable clinical tool
to enable the better use of available standard-of-care agents. Historically, predictive
biomarkers in the setting of breast cancer have played an important role in
selecting patients for targeted therapies (i.e., trastuzumab and endocrine therapies

112 Marcom et al.
including tamoxifen and aromatase inhibitors). However, validated biomarkers for
commonly used cytotoxic agents are not yet clinically available. The importance
of cytotoxic agents, even with the advent of the targeted molecular therapy era,
cannot be discounted. To date, many targeted therapies demonstrate greatest
efficacy when combined with chemotherapeutic agents. To address this relevant
clinical need, a variety of studies have been performed with the aim of developing
predictors of sensitivity for a spectrum of cytotoxic chemotherapies.
Gene expression profiling of breast cancers in the context of preoperative
chemotherapy has shown promise in providing predictive signatures of response
to specific agents (12–14). The availability of clinical datasets provides an
opportunity to move ahead with the prospective assessment of gene expression
profiling as a means for guiding PST. This methodology can also simultaneously
provide insight into breast cancer biology, as the investigation of profiles will
shed light on the deregulation of molecular pathways.
Several studies have correlated breast cancer gene expression profiles with
response to preoperative chemotherapy. Chang et al. performed gene expression
profiling on locally advanced breast cancers treated with docetaxel to identify
genes predictive of response to this therapeutic modality (32). Similar work has
been done with preoperative doxorubicin therapy and sequential taxane/anthracycline
therapy (33,34). While some studies have suggested that single markers
such as NF-kB may predict resistance to anthracycline therapy, most have found
that a multigene classifier is needed. In comparing a signature for doxorubicin/
cyclophosphamide sensitivity with a signature for docetaxel sensitivity, Cleator
et al. concluded that unique signatures for sensitivity to different chemotherapy
regimens are likely to exist (33).
More recently, studies of PST have examined the pCR rates for clinically
and molecularly defined breast cancer subtypes. These retrospective analyses of
PST trials demonstrate marked differences in pCR rates depending on breast
cancer subtype (35,36). A recent publication by the M.D. Anderson Breast Group
has retrospectively defined gene expression profiles that predict response to a
sequential regimen of paclitaxel (Taxol) followed by 5-fluorouracil, doxorubicin
(Adriamycin), and cyclophosphamide (T-FAC) (13).
As an alternate strategy, drug sensitivity data from the NCI-60 panel of
cancer cell lines, coupled with baseline Affymetrix gene expression data unique
to these cells, has been utilized to generate a series of gene expression signatures
with the potential to predict sensitivity to various chemotherapeutic agents (37)
(Fig. 1). To test the capacity of the in vitro docetaxel sensitivity predictor to
accurately identify the patients who responded to docetaxel, we evaluated data
from a previously published neoadjuvant study that linked gene expression data
with clinical response to docetaxel (32). The in vitro–generated profile correctly
predicted docetaxel response in 22 of 24 patient samples, achieving an overall
accuracy of 91.6%. Findings were further validated in an ovarian cancer dataset,
where accuracy for predicting the response to salvage docetaxel therapy
exceeded 85% (37).

Genomic Signatures for Personalized Treatment of Breast Cancer 113
Figure 1 Signatures that predict response to cytotoxic chemotherapies. Heatmaps of
expression profiles selected to predict sensitivity to the indicated chemotherapies. Each
column in the figure represents individual samples. Each row represents an individual
gene, ordered from top to bottom according to regression coefficients. Shown below is
prediction of single agent therapy response in patient samples using cell line based
expression signatures of chemosensitivity. Source: Adapted from Potti et al., Nature
Medicine 12:1294 –1300, 2006.
Given that the in vitro developed gene expression profiles predicted
clinical response to docetaxel, the approach has been extended to test the ability
of additional signatures to predict response to commonly used salvage therapies
in an independent dataset of ovarian cancer patients previously treated with
Adriamycin. As shown in Figure 1B, each of these predictors was capable of
accurately predicting the response to the drugs in patient samples, achieving an
accuracy in excess of 81% overall. Importantly, in these examples, the overall
clinical response rate varied from 25% to 45%. Using the expression signatures
to predict patients likely to respond would essentially increase the “effective”
response rate (the positive predictive value for chemosensitivity for any given
drug) to greater than 85%, by selecting those patients likely to respond, based on
the drug sensitivity predictor. These data suggest that the genomic-based predictors
of chemotherapeutic response provide a unique opportunity to guide the
selection of a cytotoxic agent most effective for an individual patient—the
essence of personalized medicine.
Importantly, it is also evident from further analyses that while there are
overlaps in the predicted sensitivities to the chemotherapeutic agents among
ovarian patients, there are also distinct groups of patients that are predicted to be
sensitive to various single-agent salvage therapies (Fig. 2). Many therapeutic
regimens make use of combinations of chemotherapeutic cytotoxic agents,
raising a question on the extent that signatures of individual therapeutic response
will also predict response to a combination of agents. To address this question,

114 Marcom et al.
Figure 2 Patterns of predicted sensitivity to commonly used chemotherapies. A
collection of breast cancer samples represented by microarray data was used to predict
the likely sensitivity to the indicated agents based on the signatures at the top of the
figure. Predictions are plotted as a heatmap representing probability of sensitivity.
Adapted from Potti et al., Nature Medicine 12:1294 –1300, 2006.
data from a breast neoadjuvant trial were utilized to evaluate the ability of singleagent
signatures for T-FAC to predict patient response measured by pCR (34).
Utilization of individual chemotherapy sensitivity signatures to assess patient
response illustrated a significant distinction between the responders (n ¼ 13) and
nonresponders (n ¼ 38), with the exception of 5-flourouracil. Importantly, the
combined probability of sensitivity to all four agents in the T-FAC neoadjuvant
regimen was calculated using the probability theorem. It was clear from this
analysis that the prediction of response based on a combined probability of
sensitivity, built from individual chemosensitivity predictions, yielded a statistically
significant (p < 0.0001, Mann–Whitney U test) distinction between the
responders and nonresponders.
A STRATEGY FOR USING CHEMOTHERAPY
RESPONSE PREDICTORS IN PST
Although promising, studies to date have included post hoc gene expression
profiling of tumor samples. The ability to effectively select the most appropriate
preoperative systemic cytotoxic agents for the treatment of early-stage breast
cancer must ultimately be tested and validated in prospective clinical trials. This
rationale approach will not only provide an opportunity to confirm biologically
relevant subtypes, but also establish their predictive capacities regarding
response to specific, conventional cytotoxic agents.

Genomic Signatures for Personalized Treatment of Breast Cancer 115
The development of genomic signatures predictive of response to commonly
employed chemotherapeutics offers an opportunity to significantly
expand the identification of combinations of cytotoxic agents best suited for an
individual patient. With this in mind, several prospective clinical studies have
been proposed to test the capability of genomic predictors to most effectively
match cytotoxic agents with individual patients. This concept is best evaluated in
the context of neoadjuvant trials where hypotheses regarding the effectiveness of
genomic predictions can be accurately assessed through pathologic measurements
of tumor response.
We note that the availability of chemotherapy response predictors provides
an opportunity to employ genomic predictors in present-day practice, where
patients are commonly treated with one or more therapeutic regimens of equal
efficacy (Fig. 3). The ability to select the most effective therapy, among a panel
of standard-of-care agents, is a rational strategy because standard treatment
procedures would not be greatly altered. The goal of these trials would be to
more effectively utilize currently available standard agents. For example, a
breast neoadjuvant study has been designed and is currently being implemented
by employing this approach. Patients with early-stage breast cancer will be
treated with one of two current standard-of-care therapies for which genomic
predictors are available. This study design would compare the response rates
for the randomly or doctor-selected A or B regimens versus assignment to
therapy A or B via genomic predictor of response. The success of this study
could then set the stage for larger studies utilizing genomic predictors of
chemotherapy sensitivity to more accurately predict response to cytotoxic
therapies. The next phase of clinical studies would then focus on novel therapeutics
(i.e., targeted therapies) for the individual patients unlikely to respond
to either regimen.
Figure 3 Schematic diagram of prospective clinical trial that would evaluate the
capacity of predictors of chemotherapy response to improve response rates.

116 Marcom et al.
OTHER GENOMIC OPPORTUNITIES TO GUIDE
BREAST CANCER THERAPEUTICS
The successful development and validation of predictors of response to standard
chemotherapeutics will foster the more efficient use of these agents through the
identification of patients with the greatest likelihood of response. Concurrently,
these strategies will also identify a subgroup of patients resistant to all available
standard-of-care chemotherapeutics, necessitating the implementation of targeted
strategies. This is a critical point as it is of little benefit to patients to
identify resistance if a plan for alternative therapeutic strategies does not exist.
Currently, most patients resistant to standard chemotherapeutic agents are traditionally
treated with second- or third-line chemotherapeutic agents or are
enrolled on a clinical trial. The more effective alternative is to identify patients
likely to be resistant prior to initiation of therapy so that opportunities for
alternative treatments might be employed in the first-line setting.
A viable option is to exploit the wealth of pathway-specific therapeutics
that have been developed over the past two decades. To this goal, gene
expression patterns reflecting deregulation of oncogenic pathways have been
developed, as depicted in Figure 4 (38). These oncogenic pathway signatures
have been shown to accurately predict the status of pathway deregulation in a
series of both tumors derived from murine models and human tumors.
These results demonstrate an ability to predict pathway profiles unique to
individual human tumors, thus characterizing the genetic events associated with
tumor development. Importantly, the assay of gene expression profiles provides
a measure of the consequence of the oncogenic process, irrespective of how the
pathway might have been altered. Thus, even if the known oncogene is not
mutated, but rather another component of the pathway is altered, the gene
expression profile will detect this alteration.
The analysis of oncogenic pathways through gene expression profiling
offers an opportunity to identify new therapeutic options for patients by providing
a potential basis for guiding the use of pathway-specific drugs. To
demonstrate the value of pathway prediction to guide drug selection, pathway
deregulation was mapped among a series of breast cancer cell lines to screen
potential therapeutic drugs. In parallel with pathway status mapping, the cell
lines were treated with a variety of drugs known to target specific activities
within given oncogenic pathways. A variety of available reagents known to
target oncogenic pathways for which signatures have been developed [i.e., a farnesyl
transferase inhibitor (FTS) and an Src inhibitor (SU6656)] were employed.
In each case, a clear relationship was demonstrated between prediction of pathway
deregulation and sensitivity to the respective targeted therapeutic. These preliminary,
yet promising, results have been further validated in a larger set of 50 cancer
cell lines, including lung, ovarian, and melanoma tumor types.
Taken together, these data demonstrate a potential approach to identify
therapeutic options for chemotherapy-resistant patients, as well as identify novel

Genomic Signatures for Personalized Treatment of Breast Cancer 117
Figure 4 Patterns of predicted cell signaling pathway activation. A collection of
breast cancer samples represented by microarray data was used to predict the likely state
of activation of the indicated cell signaling pathways based on the signatures at the top of
the figure developed to represent pathway activation. Predictions are plotted as a heatmap
representing probability of pathway activation. A Kaplan-Meier survival analysis is
shown at the bottom reflecting the overall survival for patients in each cluster defined by
pathway patterns. Source: Adapted from Bild et al., Nature 439:353–357, 2006.
combinations for chemotherapy-sensitive patients. This unique approach represents
a potential strategy to optimize effective treatments for cancer patients and
warrants future, prospective, validations trials in both the adjuvant and metastatic
settings.
Pathway signatures have the potential to identify patients most sensitive
to standard-of-care cytotoxic drugs, as well as those resistant to standard
regimens. This approach creates an opportunity to combine chemotherapy
sensitivity profiles with concurrent oncogenic pathway deregulation within an
individual patient’s tumor. As depicted in Figure 5, predictions of sensitivity to

118 Marcom et al.
Figure 5 Combining information regarding status of individual patient’s tumors.
The figure depicts an analysis of a set of breast tumors for predicted sensitivity to various
chemotherapies (top figure) and then the pattern of pathway activation in relation to one
of these (docetaxel sensitivity).
various cytotoxic agents can be combined with predictions of oncogenic
pathway deregulation with the goal of defining the most optimal regimen. For
example, a patient whose tumor is resistant to docetaxel may concurrently
illustrate a high probability of Src pathway deregulation. Src inhibitors are
currently being tested in clinical trials in a variety of solid tumors, including
breast tumors. The selection of a therapy targeted against the Src pathway may
prove more efficacious for this individual patient compared with the empiric
selection of a cytotoxic agent. In addition, this strategy could be employed to
identify novel combinations of therapies—either multiple cytotoxic agents or
cytotoxic agents combined with targeted agents—as guided by a more rational
and informed approach.
IMPACT AND FUTURE
The practice of oncology continually faces the challenge of matching the optimal
therapeutic regimen with the most appropriate patient, balancing relative benefit
with risk, to achieve the most favorable outcome. This challenge is often
daunting with marginal success rates in many advanced disease contexts, likely
reflecting the enormous complexity of the disease process coupled with an
inability to properly guide the use of available therapeutics.
The importance of selecting patients who will respond to a given therapeutic
agent is perhaps best illustrated by the example of trastuzumab. In the
absence of selection as guided by Her2 protein overexpression or gene amplification,
the overall response rate among breast cancer patients is on the order of

Genomic Signatures for Personalized Treatment of Breast Cancer 119
10%. In contrast, for patients selected on the basis of Her2 overexpression or
amplification, the overall response rate when combined with chemotherapy
approached 50% to 60% (39–41). The use of trastuzumab in the treatment of
both early-stage and advanced breast cancer is guided by Her2 protein overexpression
of gene amplification. Gene expression signatures predicting
response to various cytotoxic chemotherapeutic agents provide an equally
important opportunity to optimize response to various cytotoxic agents.
Importantly, clinically-available, validated predictors of chemotherapy
response could provide an invaluable tool in present day practice. Commonly,
patients are treated with one or more therapeutic regimens that, on a population
basis, have individually demonstrated equal efficacy. Validated chemotherapy
predictive of chemotherapy sensitivity provides an opportunity to select the most
effective therapy among a panel of standard-of-care agents individualized to a
particular patient’s tumor biology. It is essential that this approach be incorporated
into modern day clinical trials. Ultimately, this novel concept has the
potential to augment future treatment strategies incorporating likelihood of
sensitivity to common cytotoxic agents, as well as various targeted therapeutic
agents. Incorporating this approach into the routine management of patients with
various malignancies has the potential to more rationally assign combination
therapies, both cytotoxic and targeted agents, as rationally guided by each
individual’s unique tumor biology.
SUMMARY
Treatment within the field of oncology, in particular that of breast cancer, is
shifting away from a broad, population-based approach toward individualized
care as driven by unique tumor biology. Validated signatures of chemotherapy
sensitivity, coupled with an understanding of oncogenic pathway signaling
within a given breast tumor, provides an opportunity to improve the utilization of
standard-of-care therapies. This approach also provides an opportunity to take
the next step toward personalized cancer care—incorporation of targeted
agents—as guided by advances in genomic technology.
REFERENCES
1. Alizadeh AA, Eisen MB, Davis RE, et al. Distinct types of diffuse large B-cell
lymphoma identified by gene expression profiling. Nature 2000; 403:503–511.
2. Golub TR, Slonim DK, Tamayo P, et al. Molecular classification of cancer: class discovery
and class prediction by gene expression monitoring. Science 1999; 286:531–537.
3. Perou CM, Jeffrey SS, van de Rijn M, et al. Distinctive gene expression patterns in
human mammary epithelial cells and breast cancers 1999; 96:9212–9217.
4. van ’t Veer L, Dai H, van de Vijver MJ, et al. Gene expression profiling predicts
clinical outcome of breast cancer. Nature 2002; 415:530–536.

120 Marcom et al.
5. Huang E, Cheng SH, Dressman H, et al. Gene expression predictors of breast cancer
outcomes. Lancet 2003; 361:1590–1596.
6. Pittman J, Huang E, Dressman H, et al. Integrated modeling of clinical and gene
expression information for personalized prediction of disease outcomes. Proc Natl
Acad Sci U S A 2004; 101:8431–8436.
7. Ma XJ, Wang Z, Ryan PD, et al. A two-gene expression ratio predicts clinical outcome
in breast cancer patients treated with tamoxifen. Cancer Cell 2004; 5:607–616.
8. Chang HY, Sneddon JB, Alizadeh AA, et al. Gene expression signature of fibroblast
serum response predicts human cancer progression: similarities between tumors and
wounds. PLOS Biol 2004; 2:206–214.
9. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumors.
Nature 2000; 406:747–752.
10. van de Vijver MJ, He YD, van’T Veer LJ, et al. A gene-expression signature as a
predictor of survival in breast cancer. N Engl J Med 2002; 347:1999–2009.
11. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifentreated,
node-negative breast cancer. N Engl J Med 2004; 351:2817–2826.
12. Chang JC, Wooten EC, Tsimelzon A, et al. Gene expression profiling for the prediction
of therapeutic response to docetaxel in patients with breast cancer. Lancet
2003; 362:362–369.
13. Hess KR, Anderson K, Symmans WF, et al. Pharmacogenomic predictor of sensitivity
to preoperative chemotherapy with paclitaxel and fluorouracil, doxorubicin,
and cyclophosphamide in breast cancer. J Clin Oncol 2006; 24:4236–4244.
14. Modlich O, Prisack HB, Munnes M, et al. Predictors of primary breast cancers
responsiveness to preoperative epirubicin/cyclophosphamide-based chemotherapy:
translation of microarray data into clinically useful predictive signatures. J Transl
Med 2005; 3:1–18.
15. Mamounas EP, Bryant J, Lembersky B, et al. Paclitaxel after doxorubicin plus
cyclophosphamide as adjuvant chemotherapy for node-positive breast cancer: results
from NSABP B-28. J Clin Oncol 2005; 23:3686–3696.
16. Henderson IC, Berry DA, Demetri GD, et al. Improved outcomes from adding
sequential Paclitaxel but not from escalating Doxorubicin dose in an adjuvant chemotherapy
regimen for patients with node-positive primary breast cancer. J Clin
Oncol 2003; 21:976–983.
17. Martin M, Pienkowski T, Mackey J, et al. Adjuvant Docetaxel for node-positive
breast cancer 2005; 52:2302–2313.
18. Roche H, Fumoleau P, Spielmann M, et al. Five years analysis of the PACS 01 trial:
6 cycles of FEC100 vs 3 cycles of FEC100 followed by 3 cycles of docetaxel (D) for
the adjuvant treatment of node positive breast cancer. Breast Cancer Res Treatment
2004; 88:27 (abstr).
19. Berry DA, Cirrincione C, Henderson IC, et al. Estrogen-receptor status and outcomes
of modern chemotherapy for patients with node-positive breast cancer. JAMA 2006;
295:1658–1667.
20. Citron ML, Berry DA, Cirrincione C, et al. Randomized trial of dose-dense versus
conventionally scheduled and sequential versus concurrent combination chemotherapy
as postoperative adjuvant treatment of node-positive primary breast cancer:
first report of Intergroup Trial C9741/Cancer and Leukemia Group B Trial 9741.
J Clin Oncol 2003; 21:1431–1439.

Genomic Signatures for Personalized Treatment of Breast Cancer 121
21. Hudis C, Citron M, Berry D, et al. Five year follow-up of INT C9741: dose-dense
chemotherapy is safe and effective. Presented at: The 28th Annual San Antonio
Breast Cancer Symposium, December 8–11, 2005.
22. Venturini M, Del Mastro L, Aitini E, et al. Dose-dense adjuvant chemotherapy in
early breast cancer patients: results from a randomized trial. J Natl Cancer Inst 2005;
97:1724–1733.
23. Lin NU, Gelman R, Winer EP. Dose density in breast cancer: a simple message?
J Natl Cancer Inst 2005; 97:1712–1714.
24. Kaufmann M, Hortobagyi GN, Goldhirsch A, et al. Recommendations from an
international expert panel on the use of neoadjuvant (primary) systemic treatment of
operable breast cancer: an update. J Clin Oncol 2006; 24:1940–1949.
25. Bear HD, Anderson S, Smith RE, et al. Sequential preoperative or postoperative
docetaxel added to preoperative doxorubicin plus cyclophosphamide for operable
breast cancer: national Surgical Adjuvant Breast and Bowel Project Protocol B-27.
J Clin Oncol 2006; JCO.2005.04:1665.
26. Smith IC, Heys SD, Hutcheon AW, et al. Neoadjuvant chemotherapy in breast cancer:
significantly enhanced response with docetaxel. J Clin Oncol 2002; 20:1456–1466.
27. von Minckwitz G, Raab G, Caputo A, et al. Doxorubicin with cyclophosphamide
followed by docetaxel every 21 days compared with doxorubicin and docetaxel every
14 days as preoperative treatment in operable breast cancer: the GEPARDUO study
of the German Breast Group. J Clin Oncol 2005; 23:2676–2685.
28. Fisher B, Bryant J, Wolmark N, et al. Effect of preoperative chemotherapy on the
outcome of women with operable breast cancer. J Clin Oncol 1998; 16:2672–2685.
29. Buzdar AU, Singletary SE, Theriault RL, et al. Prospective evaluation of paclitaxel
versus combination chemotherapy with fluorouracil, doxorubicin, and cyclophosphamide
as neoadjuvant therapy in patients with operable breast cancer. J Clin
Oncol 1999; 17:3412–3417.
30. Jones SE, Savin MA, Holmes FA, et al. Phase III trial comparing doxorubicin plus
cyclophosphamide with docetaxel plus cyclophosphamide as adjuvant therapy for
operable breast cancer. J Clin Oncol 2006; 24:5381–5387.
31. Goldstein L, O’Neill A, Sparano J, et al. E2197: Phase III AT (doxorubicin/docetaxel)
vs. AC (doxorubicin/cyclophosphamide) in the adjuvant treatment of node positive and
high risk node negative breast cancer. American Society of Clinical Oncology Annual
Meeting, May 13–17, 2005.
32. Chang JC, Wooten EC, Tsimelzon A, et al. Patterns of resistance and incomplete
response to docetaxel by gene expression profiling in breast cancer patients. J Clin
Oncol 2005; 23:1169–1177.
33. Cleator S, Tsimelzon A, Ashworth A, et al. Gene expression patterns for doxorubicin
(Adriamycin) and cyclophosphamide (Cytoxan) (AC) response and resistance. Breast
Cancer Res Treat 2006 95:229–233.
34. Ayers M, Symmans WF, Stec J, et al. Gene expression profiles predict complete
pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and
cyclophosphamide chemotherapy in breast cancer. J Clin Oncol 2004; 22:2284–2293.
35. Guarneri V, Frassoldati A, Giovannelli S, et al. Primary systemic therapy for
operable breast cancer: a review of clinical trials and perspectives. Cancer Lett 2007;
248:175–185.
36. Rouzier R, Perou CM, Symmans WF, et al. Breast cancer molecular subtypes respond
differently to preoperative chemotherapy. Clin Cancer Res 2005; 11(16):5678–5685.

122 Marcom et al.
37. Potti A, Lancaster JM, Dressman HK, et al. A genomic strategy to guide the use of
chemotherapeutic drugs in solid tumors. Nat Med 2006; 12:1294–1300.
38. Bild A, Yao G, Chang JT, et al. Oncogenic pathway signatures in human cancers as a
guide to targeted therapies. Nature 2006; 439:353–357.
39. Chan A, Martin M, Untch M, et al. Vinorelbine plus trastuzumab combination as
first-line therapy for HER 2-positive metastatic breast cancer patients: an international
phase II trial. Br J Cancer 2006; 95:788–793.
40. Marty M, Cognetti F, Maraninchi D, et al. Randomized phase II trial of the efficacy
and safety of trastuzumab combined with docetaxel in patients with human epidermal
growth factor receptor 2-positive metastatic breast cancer administered as firstline
treatment: the M77001 study group. J Clin Oncol 2005; 23:4265–4274.
41. Robert N, Leyland-Jones B, Asmar L, et al. Randomized phase III study of trastuzumab,
paclitaxel, and carboplatin compared with trastuzumab and paclitaxel in women with
HER-2-overexpressing metastatic breast cancer. J Clin Oncol 2006; 24:2786–2792.

9
Personalized Medicine by the Use of
Microarray Gene Expression Profiling
Stella Mook and Laura J. van ’t Veer
Department of Pathology, The Netherlands Cancer Institute,
Amsterdam, The Netherlands
Fatima Cardoso
Department of Medical Oncology, Institute Jules Bordet, Brussels, Belgium
INTRODUCTION
For accurate management of early breast cancer, several important decisions
need to be made by patients and clinicians regarding the type of surgery, the
need for adjuvant radiotherapy, and particularly the type of adjuvant systemic
therapy (chemotherapy and/or endocrine therapy). The decision about adjuvant
systemic therapy is of paramount importance since breast cancer is mainly a
systemic disease, with a majority of deaths occurring as a consequence of distant
metastases. This decision depends on the likelihood of recurrence; patientrelated
factors such as biological age, menopausal status, performance status,
and co-morbidities; and tumor-related factors such as size, grade, lymph node
status, and hormonal and human epidermal growth factor receptor 2 (HER2)
status. Most of these variables are combined into prognostic models or guidelines
such as the St. Gallen consensus or the National Comprehensive Cancer Network
(NCCN) guidelines, in order to assist patients and clinicians in the decisionmaking
process (1–3). Although these guidelines provide valuable information
about prognosis in certain breast cancer populations, the prediction of disease
123

124 Mook et al.
outcome for the individual patient is far from accurate. In reality, patients with
similar clinicopathological characteristics can have strikingly different outcomes.
Because of these limitations, and because of the recognition of the incurable
nature of metastatic breast cancer, clinicians consider prescribing adjuvant
chemotherapy in the majority of early-stage breast cancer patients to reduce the
risk of relapse. However, a large proportion of these patients is probably cured
with locoregional treatment alone or in combination with adjuvant endocrine
therapy, thus unnecessarily exposing many patients to chemotherapy side effects.
The identification of robust and reliable prognostic markers that accurately select
patients not requiring adjuvant chemotherapy is essential to decrease the problem
of overtreatment. Additionally, a more accurate patient selection will also contribute
to the decrease in undertreatment, a less common but dangerous problem.
The introduction of high-throughput techniques, such as microarray-based
gene expression profiling and sequencing of the human genome, has led to a novel
approach in breast cancer research. Measuring the expression level of thousands of
genes in one experiment using microarray technology enables us to better
understand the biology and heterogeneity of breast cancer. The first study to
examine gene expression patterns in breast cancer showed the existence of at least
four molecular subtypes of breast cancer—luminal A, luminal B, basal, and HER2
positive—distinguished by extensive differences in gene expression (4). Several
subsequent studies confirmed these findings (5–8). In addition, microarray techniques
have been used to identify profiles that correlate with disease outcome
(prognostic profile) or with response to treatment (predictive profile) (9–16).
One of the first prognostic profiles was the 70-gene profile (Mamma-
Print TM ) identified by van ’t Veer et al. (13). This dichotomous microarray
classifier can accurately distinguish breast cancer patients who have a high
likelihood of remaining free of distant metastases (good profile) from those who
are at high risk of developing distant metastases (poor profile) (Fig. 1).
DEVELOPMENT AND FIRST RETROSPECTIVE VALIDATION
OF THE 70-GENE PROFILE (MAMMAPRINT TM )
The 70-gene prognostic profile was developed in a training set of 78 lymph
node–negative, invasive breast tumors less than 5 cm in diameter. Patients were
younger than 55 years at diagnosis, had no prior malignancies, and were treated
in the Netherlands Cancer Institute (NKI) with breast conserving therapy or
mastectomy. Five out of 78 patients received adjuvant systemic treatment
comprising chemotherapy (n ¼ 3) or endocrine therapy (n ¼ 2). Thirty-four
patients developed distant metastases within five years of diagnosis, while the
remaining 44 patients remained free of distant metastases. The mean follow-up
of patients who remained free of distant metastases was 8.7 years, and the mean
time to metastases was 2.5 years.
For all 78 samples, the percentage of tumor cells in a hematoxylin- and
eosin-stained section, before and after cutting sections for RNA isolation, was at

Microarray Gene Expression Profiling 125
Figure 1 Steps in 70-gene microarray analysis. Source: From Ref. 32. copyright # 2002
Massachusetts Medical Society. All rights reserved. Adopted with permission.
least 50%. RNA was isolated and labeled with a fluorescent dye. Reference RNA
consisted of an equal amount of RNA from all tumors. Both sample RNA and
reference RNA were hybridized on an oligonucleotide microarray platform
containing approximately 25,000 genes (produced by Agilent Technologies, Palo
Alto, California, U.S.) (17). After hybridization, microarray slides were washed
and scanned using a confocal laser scanner (Agilent). Fluorescence intensities
were quantified, corrected for background noise, and normalized (13).
Using a statistical method called “supervised classification,” 231 genes
that were differentially expressed in tumors that metastasized compared with
tumors that did not were selected. These 231 genes were then ranked on the basis
of their correlation coefficient with disease outcome. Using a leave-one-out
cross-validation procedure, the top 70 of these genes appeared to be the optimal
set of genes to accurately discriminate between tumors with a good disease
outcome (free from distant metastases for at least five years) and tumors with a
poor disease outcome (distant metastases within five years of diagnosis).
All 78 tumors were ranked according to their correlation with the average
expression of the 70 genes of the patients who did not develop distant metastases.
The sensitivity was optimized by setting a threshold, which resulted in the

126 Mook et al.
misclassification of 3 out of 34 patients with a poor disease outcome being
classified as good prognosis and therefore would erroneously have had chemotherapy
withheld (9% misclassification). Subsequently, the 70-gene profile was
validated in 19 breast cancer tumors—7 tumors from patients with a good
clinical outcome (no distant metastases within five years of diagnosis) and
12 tumors from patients who did develop distant metastases within five years of
diagnosis. Using the 70-gene profile, only 2 out of 19 tumors were classified
incorrectly, thereby confirming the initial results (13,18).
The same group of investigators performed a first retrospective validation
study using a consecutive series of 295 breast tumors from NKI (144 lymph
node–positive and 151 lymph node–negative tumors). All patients were younger
than 53 years at diagnosis and were treated with locoregional therapy alone
(56%) or in combination with adjuvant systemic therapy comprising chemotherapy
alone (31%), hormonal therapy alone (7%), or a combination (7%). In
this consecutive series of 295 patients, 61 were also part of the previous series
used to develop the 70-gene profile (14).
All 295 tumors were classified as good or poor profile by calculating the
correlation coefficient with the established average expression level of the
70 genes in the previously studied good-outcome group. For the 234 samples that
were not included in the previous study, tumors with a correlation coefficient
above the threshold of 0.4 were classified as good profile; for the 61 tumors
that were also part of the previous study, the threshold was adapted to 0.55 to
correct for overfitting (18).
The 70-gene profile accurately classified 115 tumors (39%) as a goodprognosis
group (good profile) with a 10-year overall survival of 95% (2.6%)
and 180 tumors (61%) as a poor-prognosis group (poor profile) with a 10-year
overall survival of 55% (4.4%). Interestingly, the 70-gene profile was not only
associated with disease outcome in the lymph node–negative patients [hazard
ratio (HR) for distant metastases 5.5; 95% confidence interval (CI) 2.5–12.2; p <
0.001], but also strongly associated with disease outcome in the group of 144
lymph node–positive patients (HR for distant metastases 4.5; 95% CI 2.0–10.2;
p < 0.001). Furthermore, the only significant independent factors for prediction
of distant metastases were a poor-prognosis profile, a larger tumor, the presence
of vascular invasion, and no chemotherapy treatment. The 70-gene profile was
by far the most powerful predictor of distant metastases, with an adjusted HR of
4.6 (95% CI 2.3–9.2; p < 0.001) (14,18).
THE PROGNOSTIC VALUE OF THE 70-GENE
PROFILE IN A CLINICAL CONTEXT
To assess the potential added value of this prognostic tool in a clinical context,
the 70-gene profile was compared with the St. Gallen consensus guidelines or
the NIH criteria used at that time (3,19). In comparison with these guidelines, the
70-gene profile assigned more patients to the good-prognosis group than the
St. Gallen consensus guidelines or the NIH criteria (40% vs. 15% and 7%,

Microarray Gene Expression Profiling 127
respectively). Moreover, patients identified by the 70-gene profile as having
good prognosis were more likely to remain free of distant metastases than
patients identified as having good prognosis according to the St. Gallen consensus
guidelines or the NIH criteria. On the other hand, patients who were
assigned to the poor-prognosis group by the 70-gene profile had a higher risk of
developing distant metastases as compared with the poor-prognosis group
classified by the St. Gallen consensus guidelines or the NIH criteria.
The 70-gene profile was built to have a minimum of misclassified patients
with a poor disease outcome; in other words, the low-risk group should be
maximized without causing considerable undertreatment. Consequently, in this
retrospective validation series, the 70-gene profile had a high negative predictive
value of 96% for the 234 new patients and a positive predictive value of only
38%. That is to say, of all patients with a good prognosis profile, 4% will still
develop distant metastases and would erroneously have had chemotherapy
withheld, whereas of all patients classified as having a poor prognosis, 38% will
indeed develop distant metastases. Consequently, treatment decisions based on
the 70-gene profile would still lead to overtreatment; but since the total proportion
of patients identified as having a poor profile is much smaller than the
proportion of high-risk patients according to the St. Gallen consensus guidelines
or the NIH criteria, overtreatment would be reduced by 25% to 30%. Furthermore,
the overall prediction of disease outcome and consequent treatment
decision according to the 70-gene profile seems to be more accurate.
INDEPENDENT RETROSPECTIVE VALIDATION
BY THE TRANSBIG CONSORTIUM
To further substantiate the prognostic value of the 70-gene profile, the European
Union’s sixth Framework Project TRANSBIG (http://www.breastinternationalgroup.org/TransBIG.aspx)
conducted an independent retrospective validation
study (20). Frozen tumor material and clinical data were collected from
302 patients from five different cancer centers in the United Kingdom, Sweden,
and France. Patients were younger than 61 years at diagnosis, had a lymph node–
negative T1 or T2 tumor, and had not received adjuvant systemic therapy. The
patients were treated before 1999, and the median follow-up in this series was
13.6 years. The frozen tumor samples were sent to Agendia, the microarray facility
in Amsterdam, where RNA was isolated and hybridized on a custom-designed
microarray, measuring the gene expression level of the 70 genes in triplicate.
Tumors were classified as poor profile if the correlation coefficient of the average
expression was under 0.4. The microarray facility had no access to the clinical
data, and the researchers collecting the clinical data were blinded for the 70-gene
profile results. Only an independent statistical office had access to both the clinical
and the microarray data simultaneously. Eighty percent of the samples were
centrally reviewed for estrogen receptor (ER) status and tumor grade, and the
centers were independently audited for quality of data control.

128 Mook et al.
This independent study confirmed that the 70-gene profile can accurately
select patients who are at high risk of developing distant metastases (poor prognosis)
from patients who are at low risk of developing distant metastases and
hence can be safely spared chemotherapy (good profile), with HRs of 2.79 (95%
CI 1.60–4.87) and 2.32 (95% CI 1.35–4.0) for distant metastases–free survival and
overall survival, respectively. Moreover, patients with a good profile had a 10-year
overall survival of 89% compared with 69% in patients with a poor profile. The
70-gene profile outperformed several prognostic methods such as the Nottingham
Prognostic Index and the St. Gallen consensus guidelines. The performance of the
70-gene profile was also compared with Adjuvant! Online. The Adjuvant! Software—available
at www.adjuvantonline.com—calculated a 10-year overall survival
probability based on clinicopathological criteria (the patient’s age, tumor
size, tumor grade, ER status, and comorbidity) (21,22). Patients were considered
as having low clinical risk when the 10-year breast cancer–specific survival
probability calculated by Adjuvant! was more than 88% if their tumors were ER
positive and more than 92% if they were ER negative. The rationale for these two
cutoffs was the assumption that patients with ER-positive tumors would now
receive hormonal therapy, with an average absolute benefit of 4%. In the discordant
cases, the 70-gene profile provided much stronger prognostic information,
with almost identical survival rates for the low and high clinical risk groups within
each 70-gene profile risk, suggesting that the 70-gene profile predicts disease
outcome independently of clinical and pathological characteristics (Fig. 2) (20).
Figure 2 Kaplan–Meier curve by clinical and 70-gene expression signature risk groups.
Overall survival. Source: From Ref. 20. By permission of Oxford University Press.

Microarray Gene Expression Profiling 129
Another important finding of this validation study is the time dependency
of the 70-gene profile. The median follow-up in this series was twice as long as
the follow-up of the initial series. To determine the influence of the time of
distant metastases on the prognostic value of the profile, HRs were calculated
with arbitrary censoring of all observations at different time points. The adjusted
HR for distant metastases at 5 years was 4.7 in comparison with an adjusted HR
of 3.5 at 10 years, suggesting a better prediction of early (within five years of
diagnosis) metastases by the 70-gene profile (18,23,24).
OTHER PROGNOSTIC PROFILES IN BREAST CANCER
Several other prognostic gene expression signatures were developed, such as the
76-gene profile, the wound signature, and the genomic grade index (15,25,26).
The 76-gene prognostic profile was developed on a different microarray platform
(Affymetrix, Santa Clara, California, U.S.) and using a slightly different methodology;
the genes associated with relapse of the disease were selected separately
for ER-positive and ER-negative patients. The genes selected in each
group were then combined in a 76-gene profile (15). A retrospective validation
study confirmed the prognostic value of this profile in early-stage breast cancer
patients (27). Recently, the 76-gene profile has also been validated independently
by the TRANSBIG consortium in the same patient series, using the same
methodology as the validation of the 70-gene profile (24). Although only
three genes were common to both profiles, the biological pathways represented
are the same, and the prognostic value of both the 70-gene profile and the
76-gene profile was validated in this patient series (20,24). Since the profiles
perform equally, and the 70-gene profile has proven to be reproducible and
robust and is available as a Food and Drug Administration (FDA)-cleared
diagnostic test (28), the TRANSBIG consortium decided to use the 70-gene
profile in a large prospective clinical trial, the Microarray In Node-negative
Disease may Avoid ChemoTherapy (MINDACT) trial.
IMPLEMENTATION OF THE 70-GENE PROFILE:
THE RASTER TRIAL AND THE LOGISTICS PILOT STUDY
In addition to scientific evidence provided by validation studies, implementation
of these microarray profiles in both prospective clinical trials and in daily practice
requires logistical feasibility. The requirement of fresh tissue as the source of highquality
RNA has several consequences for tumor sample handling. Traditional
fixation of fresh tissue in formaldehyde, for example, causes degradation of RNA,
thereby making tumor samples unsuitable for microarray experiments. In addition,
to avoid interlaboratory variability, fresh tumor samples have to be shipped to one
microarray facility. Consequently, for the implementation of microarray profiles,
local procedures have to be adapted, and close collaboration between pathologists,
surgeons, and medical oncologists is of outmost importance.

130 Mook et al.
Recently, two multicenter feasibility studies were conducted, the RASTER
study and a logistics pilot study (18,29). The first study was coordinated by NKI
with financial support from the Dutch Health Care Insurance Board. The aim of
the study was to assess the feasibility of collecting good-quality tissue from
several community hospitals in the Netherlands to perform the 70-gene profile
test. For all patients included in this study, tumor samples from excised specimens
were obtained and placed in a commercially available preservation fluid
(RNAlater 1 , Qiagen GmBH, Hilden, Germany) at room temperature. Subsequently,
the samples were sent by conventional mail to NKI, where they
were snap-frozen and stored at –808C. The 70-gene profile was performed at the
microarray facility in Amsterdam. Preliminary results showed that it is feasible
to adapt local procedures and to collect good-quality tissue for microarray testing
in a multicentric setting (18,29,30).
In the RASTER study, tissue was preserved temporarily in RNAlater and
sent by conventional mail at room temperature. Preservation of tumor tissue in
RNAlater may influence several processes in the tissue, such as protein levels.
Since one of the additional aims of the MINDACT trial is the establishment of a
biological materials bank for future research, including proteomics, storage of
tissue in RNAlater is not suitable for the MINDACT trial and fresh, frozen tissue
will be mandatory (31). Therefore, a second feasibility study was conducted to
test the collection, storage, and shipment of fresh, frozen tissue in six European
hospitals. Women younger than 71 years, with early-stage breast cancer, were
included. After surgery, the pathologist obtained representative tumor samples
using a 6-mm biopsy puncher. Tumor samples were snap-frozen in liquid
nitrogen and stored at –808C within one hour of surgery. The samples were
shipped to Amsterdam on dry ice by a contracted courier specializing in the
transportation of frozen tissue. At Agendia, RNA was isolated when the samples
were representative of the tumor (i.e. tumor cells 50%). After measurement of
the quality and quantity of the RNA, the samples were hybridized on the 70-gene
profile platform. Results showed that in general it is feasible to collect goodquality
fresh, frozen tissue for microarray tests in a multicentric and multinational
setting. Additionally, in the course of this feasibility study, procedures
were adapted and optimized for use in the MINDACT trial (18).
PROSPECTIVE VALIDATION OF THE 70-GENE
PROFILE IN THE MINDACT TRIAL
Despite the completion of all necessary technical validation procedures and the
FDA clearance as the in vitro diagnostic multivariate index assay (IVDMIA),
including approval of clinical utility, the clinical validation of the 70-gene
profile can still be further optimized in currently diagnosed, adjuvantly treated
populations to achieve level-1 evidence. The TRANSBIG consortium has
decided to further validate the 70-gene prognostic alongside the classical clinicopathological
factors (31). This trial, the MINDACT trial [EORTC (European

Microarray Gene Expression Profiling 131
Organization for Research and Treatment of Cancer) 10041 BIG (Breast International
Group) 3-04], will enroll 6000 women between 18 and 71 years of age,
with primary invasive breast cancer at diagnosis. Tumors should be unilateral,
stage T1, T2, or operable T3; carcinomas in situ are eligible provided invasive
cancer is present. Patients should be treated with breast conserving treatment or
mastectomy and should have a negative sentinel node or axillary clearance.
Patients with primary malignancies or previous chemotherapy or radiotherapy
will not be included. The main goal of the trial is to prove that the 70-gene
profile will safely assign chemotherapy to a smaller proportion of node-negative
breast cancer patients. Therefore, all patients included in MINDACT will have a
risk assessment based on the 70-gene profile and according to the currently used
clinicopathological criteria. This last risk assessment is provided by an updated
version of Adjuvant! Online. If, according to both risk assessments, the patients’
risk of developing distant metastases is low (an estimated 13% of all patients), no
chemotherapy will be given. If both methods classify the patients’ risk of developing
distant metastases as high (an estimated 55% of the patients), chemotherapy
will be proposed. If the risk assessments are discordant (an estimated 32% of
all patients), patients will be randomized to the clinicopathological criteria or
the 70-gene profile to be used for adjuvant chemotherapy decision making
(Fig. 3). The group of most interest to answer the primary research question
comprises patients with a low-risk 70-gene profile (good profile) and high-risk
Figure 3 MINDACT study design. Source: From Ref. 18. With permission.
Abbreviations: R-T, treatment decision randomization: CT, chemotherapy.

132 Mook et al.
clinicopathological criteria who will be randomized to use the 70-gene profile
for the treatment decision and thus receive no chemotherapy. In this group a
null hypothesis of a five-year distant metastases–free survival of 92% will be
tested (31).
Additional study objectives are related to the type of chemotherapy and
endocrine therapy. Patients who are to receive chemotherapy may be randomized
between an anthracycline-based regimen and a regimen of docetaxel and capecitabine
in order to compare the efficacy and safety of both regimens. The latter
regimen may have increased efficacy with less long-term side effects such as
cardiotoxicity and leukemia. Post-menopausal (spontaneous or induced) patients
with hormone receptor–positive tumors will be randomized for endocrine therapy
with two years of tamoxifen, followed by five years of letrozole or seven
years of letrozole upfront.
In addition to these three research questions, whole genome arrays will be
performed for all 6000 patients, and frozen and paraffin-embedded tumor samples
together with blood samples will be collected from all patients. These
microarray data and the biological materials will be stored in an independent
biobank and will provide a valuable source of information for future research
that may result in the identification of new prognostic or predictive biomarkers
and new drug targets.
FUTURE PROSPECTS
The MINDACT trial is the first randomized clinical trial that will prospectively
evaluate the prognostic value of a microarray-based prognostic tool, the 70-gene
profile or Mammaprint, in an adjuvantly treated population. The implementation
of molecular diagnostics based on the true biology of a patient’s tumor will
greatly expedite the introduction of personalized medicine.
The details of the rationale, design, and logistics of the MINDACT trial are
described in references 18,23, and 31, and the authors strongly recommend these
articles to the reader.
ACKNOWLEDGMENTS
The studies mentioned in this chapter were supported by the European Commission
Framework Programme VI, the Center of Biomedical Genetics, the
Dutch Health Care Insurance Board, the Dutch National Genomic Initiative:
Cancer Genomics Program, the Breast Cancer Research Foundation, the EBCC-
Breast Cancer Working Group–association sans but lucratif (asbl), the Fondation
Belge Contre Le Cancer, the S. G. Komen for the Cure Foundation, Jacqueline
Seroussi Memorial Foundation for Cancer Research, and the European Organisation
for Research and Treatment of Cancer (EORTC)–Breast Cancer Group.
S. Mook was partially supported by the traineeship program of the TRANSBIG
network. The authors thank the numerous individuals who have contributed

Microarray Gene Expression Profiling 133
to the studies mentioned in this review, especially those from the TRANSBIG
consortium and the EORTC, and all the patients who have and still are participating
in these studies.
CONFLICTS OF INTEREST
Dr. L. J. van ’t Veer is a named inventor on a patent application for Mammaprint
and reports holding equity in Agendia BV.
REFERENCES
1. National Comprehensive Cancer Network. NCCN Clinical Practice Guidelines in
Oncology. Breast Cancer V.I.2008. Available at: http://www.nccn.org/professionals/
physician_gls/PDF/breast.pdf.
2. Carlson RW, McCormick B. Update: NCCN breast cancer clinical practice guidelines.
J Natl Compr Canc Netw 2005; 3(suppl 1):S7–S11.
3. Goldhirsch A, Glick JH, Gelber RD, et al. Meeting highlights: international expert
consensus on the primary therapy of early breast cancer 2005. Ann Oncol 2005;
16(10):1569–1583.
4. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours.
Nature 2000; 406(6797):747–752.
5. Hu Z, Fan C, Oh DS, et al. The molecular portraits of breast tumors are conserved
across microarray platforms. BMC Genomics 2006; 796.
6. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas
distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci
U S A. 2001; 98(19):10869–10874.
7. Sorlie T, Tibshirani R, Parker J, et al. Repeated observation of breast tumor subtypes
in independent gene expression data sets. Proc Natl Acad Sci U S A. 2003; 100(14):
8418–8423.
8. Sotiriou C, Neo SY, McShane LM, et al. Breast cancer classification and prognosis
based on gene expression profiles from a population-based study. Proc Natl Acad Sci
U.S.A. 2003; 100(18):10393–10398.
9. Chang JC, Wooten EC, Tsimelzon A, et al. Gene expression profiling for the prediction
of therapeutic response to docetaxel in patients with breast cancer. Lancet
2003; 362(9381):362–369.
10. Huang E, Cheng SH, Dressman H, et al. Gene expression predictors of breast cancer
outcomes. Lancet 2003; 361(9369):1590–1596.
11. Jansen MP, Foekens JA, van SI, et al. Molecular classification of tamoxifen-resistant
breast carcinomas by gene expression profiling. J Clin Oncol 2005; 23(4):732–740.
12. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifentreated,
node-negative breast cancer. New England J Med 2004; 351(27):2817–2826.
13. van ’t Veer LJ, Dai H, van de Vijver MJ, et al. Gene expression profiling predicts
clinical outcome of breast cancer. Nature 2002; 415(6871):530–536.
14. van de Vijver MJ, He YD, van ’t Veer LJ, et al. A Gene-expression signature as a
predictor of survival in breast cancer. New England J Med 2002; 347(25):1999–2009.
15. Wang Y, Klijn JG, Zhang Y, et al. Gene-expression profiles to predict distant metastasis
of lymph-node-negative primary breast cancer. Lancet 2005; 365(9460):671–679.

134 Mook et al.
16. West M, Blanchette C, Dressman H, et al. Predicting the clinical status of human
breast cancer by using gene expression profiles. Proc Natl Acad Sci U.S.A. 2001;
98(20):11462–11467.
17. Hughes TR, Mao M, Jones AR, et al. Expression profiling using microarrays fabricated
by an ink-jet oligonucleotide synthesizer. Nat Biotech 2001; 19(4):342–347.
18. Mook S, ’van ’t Veer LJ, Rutgers EJ, et al. Individualization of therapy using
Mammaprint: from development to the MINDACT trial. Cancer Genomics Proteomics
2007; 4(3):147–155.
19. Eifel P, Axelson JA, Costa J, et al. National Institutes of Health Consensus Development
Conference Statement: adjuvant therapy for breast cancer, November 1–3,
2000. JNCI Cancer Spectrum 2001; 93(13):979–989.
20. Buyse M, Loi S, van ’t Veer L, et al. Validation and clinical utility of a 70-gene
prognostic signature for women with node-negative breast cancer. JNCI Cancer
Spectrum 2006; 98(17):1183–1192.
21. Olivotto IA, Bajdik CD, Ravdin PM, et al. Population-based validation of the
prognostic model ADJUVANT! for early breast cancer. J Clin Oncol 2005; 23(12):
2716–2725.
22. Ravdin PM, Siminoff LA, Davis GJ, et al. Computer program to assist in making
decisions about adjuvant therapy for women with early breast cancer. J Clin Oncol
2001; 19(4):980–991.
23. Cardoso F, van ’t Veer L, Rutgers E, et al. Clinical application of the 70-gene profile
(Mammaprint TM ): the MINDACT trial. J Clin Oncol 2008; 26(5):729–735.
24. Desmedt C, Piette F, Loi S, et al. Strong time dependence of the 76-gene prognostic
signature for node-negative breast cancer patients in the TRANSBIG multicenter
independent validation series. Clin Cancer Res 2007; 13(11):3207–3214.
25. Chang HY, Nuyten DSA, Sneddon JB, et al. Robustness, scalability, and integration
of a wound-response gene expression signature in predicting breast cancer survival.
Proc Natl Acad Sci U S A. 2005; 102(10):3738–3743.
26. Sotiriou C, Wirapati P, Loi S, et al. Gene expression profiling in breast cancer:
understanding the molecular basis of histologic grade to improve prognosis. JNCI
Cancer Spectrum 2006; 98(4):262–272.
27. Foekens JA, Atkins D, Zhang Y, et al. Multicenter validation of a gene expressionbased
prognostic signature in lymph node-negative primary breast cancer. J Clin
Oncol 2006; 24(11):1665–1671.
28. Couzin J. DIAGNOSTICS: amid debate, gene-based cancer test approved. Science
2007; 315(5814):924.
29. Bueno-de-Mesquita JM, van Hakten WH, Retel VP, et al. use of 70-gene signature to
predict prognosis of patients with node-negative breast cancer: a prospective community-based
feasibility study (RASTER). Lancet Oncol 2008; 8: 1079–1087.
30. van de Vijver M. Gene-expression profiling and the future of adjuvant therapy.
Oncologist 2005; 10(suppl 2):30–34.
31. Bogaerts J, Cardoso F, Buyse M, et al. Gene signature evaluation as a prognostic
tool: challenges in the design of the MINDACT trial. Nat Clin Pract Oncol 2006; 3(10):
540–551.
32. Sauter G, Simon R. Predictive molecular pathology. N Engl J Med 347 2002; (25):
1995–1996.

10
Predictive Markers for Targeted Breast
Cancer Treatment
Hans Christian B. Pedersen a and John M. S. Bartlett
Endocrine Cancer Group, Cancer Research Centre,
Western General Hospital, Edinburgh, United Kingdom
INTRODUCTION
Currently selection of breast cancer treatment is guided predominantly by patient
prognosis using classical pathological assessment of tumors, with higher risk
patient groups being offered more aggressive therapy. The choice of treatment
regimen is guided by a very small number of predictive biomarkers (Table 1) (1).
However, it is now clearly recognized that not only may the risks of treatment
outweigh the benefits in some patient groups but that not all patients are equal
with respect to their response to and benefit gained from exposure to novel and
established therapies. There is increasing interest in the identification of biological
markers to improve the targeting of established therapies to those cancers
likely to respond. In addition, development of new molecular targeted drugs has
led to an urgent need for research into predictive markers that can distinguish
patients who benefit from treatment. New proteomic technologies are leading to
markers being discovered at an increasing rate, with protein expression, cellular
localization, and posttranslational modification being recognized as those of
potential importance. Many new immunohistochemical (IHC) markers are
a Hans Christian B. Pedersen is employed on a Marie Curie Transfer of Knowledge project between
Dako A/S, Glostrup, Denmark, and University of Edinburgh.
135

136 Pedersen and Bartlett
Table 1
Current FDA-Approved Predictive Markers in Breast
Predictive Marker Method Drug
ERa IHC Tamoxifen, AIs
PR IHC Tamoxifen, AIs
HER2/neu IHC Trastuzumab (Herceptin)
HER2/neu FISH Trastuzumab (Herceptin)
Abbreviations: ERa, estrogen receptor a; PR, progesterone receptor; AIs,
aromatase inhibitors; FISH, fluorescence in situ hybridization.
currently being tested for predictive potential for various therapies in breast
cancer. Through large-scale gene expression studies using microarrays, it has
been realized that tumors have molecular signatures that may determine response
to treatment and patient prognosis. Many candidate-predictive and prognostic
signatures in breast cancer are being explored, and some are already in use to aid
patient management (e.g., OncoType Dx 1 , Genomic Health, Redwood City,
California, U.S.). The challenge for the future is to identify both the biological
and diagnostic determinates of treatment response in order to allow matching of
appropriate treatments to specific patients.
Future breast cancer treatment is likely to be guided by predictive markers
along with genomic signatures, treating patients with mixtures of molecular
targeted therapeutics. This chapter will review current IHC and fluorescent in
situ hybridization (FISH)-based predictive markers in breast and the challenges
associated with the development and implementation of predictive markers in
future.
CURRENT PREDICTIVE MARKERS IN BREAST CANCER: LESSONS LEARNT
A predictive marker is defined as a factor that indicates sensitivity or resistance
to a specific treatment (2). In contrast, a prognostic marker predicts a tumor’s
future behavior independent of systemic adjuvant treatment. Some markers can
be both predictive and prognostic, as exemplified by the estrogen receptor a
(ERa) expression, which both predicts response to endocrine treatment and
correlates with prognosis.
Hormonal Therapy and Hormone Receptors
ERa and progesterone receptor (PR) status were the first predictive markers used
in the clinic to select patients for breast cancer treatment. Today, IHC ERa or PR
staining is done routinely on patient tumor samples to predict response to
endocrine therapy. More correctly, the assay is used to exclude ERa- or PRnegative
patients from ERa- or PR-targeted therapy, as they will have a low
probability of responding to treatment (3). Historically however, ERa was first
discovered in the late 1960s on its ability to bind radiolabeled ligand (4,5).

Predictive Markers for Targeted Breast Cancer Treatment 137
Within 10 years of its initial discovery, assays had been developed that allowed
ER to be used a marker for hormonal responsiveness and clinical aggressiveness
of the tumor (6). Many years passed before it was realized that only ERapositive
patients could benefit from treatment and for these finding to be applied
as clinical practice (3). It is now almost universally accepted that ERa testing has
a central role in breast cancer management, but it is worth noting that no prospective
trial has yet tested the predictive value of ERa or PR as predictive
biomarkers. Later evidence has emerged supporting testing for PR as well as
ERa (7). PR is a transcriptional target of ligand-bound ERa, and it is has been
proved that including the PR score in the ERa score provides stronger evidence
for true ERa positivity and therefore makes patients more likely to respond to
therapy (8–10). However, controversy remains over the value of supplementary
markers to ERa, and in several countries, PR testing in not routinely performed.
The drugs administered to ERa- or PR-positive patients have been through
many generations through the last three decades, from pure ERa-antagonist
(fulvestrant) to ERa–partial antagonist tamoxifen and drugs inhibiting the enzyme
aromatase that synthesise oestrogen (11). Current clinical practice in breast cancer
management utilizes a combination adjuvant tamoxifen followed up by aromatase
inhibitors (AIs) or AIs alone. However, ER or PR do not select between these
therapies, and despite evidence that PR is predictive of poor response on tamoxifen,
there is no evidence, to date, that AIs circumvent this resistance.
Despite the long history of the ERa or PR status as a predictive and
prognostic marker, no standardized protocols for fixation and IHC staining are
yet widely applied. Conflicting evidence exists for when patients are considered
ER negative and therefore will not respond to treatment, but above 10% positivestaining
tumor cells are generally considered ER positive (12). Perhaps because
ERa has been available for testing as a predictive marker for such a long
time, before emphasis was put on standardizing IHC testing, this has lead to the
development of a multitude of different protocols using a variety of antibodies.
ER IHC results have been shown to be highly influenced by the efficiency of the
antigen-retrieval step (13). External quality assurance initiatives such as the
United Kingdom National External Quality Assessment Scheme for Immunocytochemistry
(UK NEQAS-ICC) exist now to improve quality across clinical
laboratories through external assessment of analytical and interpretative performance;
such schemes are however not uniformly successful.
In addition to standardizing ERa IHC for scoring tumors as positive or
negative, it has been suggested that quantification of ERa levels may be necessary
in some clinical settings because tumors expressing low or intermediate
ERa levels have been shown to benefit from chemotherapy in addition to
endocrine therapy (14). Current ERa IHC protocols are semiquantitative, so
alternatives such as radiolabeled ligand-binding assays or quantitative polymerase
chain reaction (qPCR) have been suggested. Gene expression of ER and
PR correlate well with protein expression, and it has been suggested to be an
alternative to IHC as it is sensitive and quantitative over a large dynamic

138 Pedersen and Bartlett
range (15). It has been suggested that more quantitative methods of measuring
ERa and PR may provide additional information to guide choice of therapy. We
will discuss this in detail later (12).
HER2/neu and Trastuzumab (Herceptin 1 )
HER2/neu or human epidermal growth factor receptor 2, a product of the protooncogene
HER2 first discovered in 1985, belongs to the HER family of tyrosine
kinase receptors (16–18). Numerous studies have documented HER2/neupositive
breast cancers display more aggressive disease and shortened diseasefree
survival (19). Using recombinant technologies, trastuzumab (Herceptin 1 ,
Genentech, South San Francisco, California, U.S.), a monoclonal immunoglobulin
G1 class, humanized murine antibody, was developed to specifically target
patients with breast cancer that overexpressed the HER2/neu protein (20,21).
Clinical trials demonstrated the efficiency of trastuzumab for treating metastatic
breast cancers and subsequent trials in combination therapy (21–25) for HER2
positive disease. Again, HER2/neu, like ERa, is more predictive of which
patients will not respond to treatment.
As in the case for endocrine therapy, most predictive marker development
followed drug development, leading to some confusion. In initial clinical trials,
patient selection was based on an IHC assay using the mouse monoclonal
antibody that had been humanized to create trastuzumab (26,27). Today most
patients are selected for trastuzumab therapy on the basis of HER2/neu
expression levels in tumor cells (Fig. 1). In 2001, HER2/neu testing achieved
standard care of practice in the American Society of Clinical Oncology (ASCO)
breast cancer clinical practice guidelines. In most laboratories, tumors are
screened with FDA-approved IHC tests, HercepTest 1 (Dako, Glostrup, Denmark)
or HER2/neu (CB11) PATHWAY 1 (Ventana Medical Systems, Inc., Tucson,
Arizona, U.S.). Ambiguous cases are then analyzed with FDA-approved HER2/
neu FISH from Vysis (Downers Grove, Illinois, U.S.), Dako, or Ventana. Alternatively,
in some laboratories, FISH is the only method used for HER2/neu testing.
HER2/neu testing for trastuzumab therapy has undergone considerable scrutiny
in numerous studies, because evidence emerged indicated that the IHC tests
generated considerable amount of false positives and negatives (28–31). This
discrepancy is due in part to lack of adherence to protocols, variable fixation
methods, and subjectivity related to tumor scoring.
Only 20% to 35% of HER2/neu positive tumors respond to treatment,
depending on the context the drug is given in, and studies have demonstrated that
this may be attributed to a combination of factors, including lack of pathway
dependence, receptor being inactive despite overexpressed, or the induction of
resistance through various mechanisms (32). Controversy persists with regard to
the optimal method for HER2/neu testing and interpretation of results (33). It has
been reported that it may be more cost effective to test all tumors with HER2/neu
FISH, avoiding false positives and false negatives (28). This is a possibility, as

Predictive Markers for Targeted Breast Cancer Treatment 139
Figure 1 Current HER2/neu testing in breast cancer. IHC staining performed on patient
tumor samples and scored using the HercepTest. Patient testing 0 to 1þ go on to receive
other treatments. Patients testing 2þ go on to HER2/neu FISH testing and if amplified for
HER2/neu receive trastuzumab. Patients scoring 3þ receive trastuzumab. Alternatively all
patients undergo HER2/neu FISH and patients scored as amplified receive trastuzumab.
Abbreviations: IHC, immunohistochemical; FISH, fluorescence in situ hybridization.
FISH and, more critically, chromogenic in situ hybridization (CISH) technology
become more widely distributed, and an agreement can be made on the best
choice of HER2/neu testing.
DEVELOPMENT OF PREDICTIVE MARKERS
The successful targeting of HER2/neu with trastuzumab inspired the development
of a wave of new therapeutic antibodies and kinase inhibitors that were
aimed at inhibiting growth factor–regulated signaling. However, repeating the
success of HER2/neu testing for trastuzumab therapy has proven difficult.
Inhibitors of mammalian Target Of Rapamycin (mTOR, e.g., temsirolimus,
Toricel 1 , Wyeth Pharmaceuticals, Inc., Madison, New Jersey, U.S.), epidermal
growth factor receptor (EGFR, e.g., cetuximab, Erbitux 1 , ImClone Systems, Inc.,
Branchburg, New Jersey, U.S.) and vascular endothelial growth factor receptor
(VEGFR, e.g., bevacizumab, Avastin 1 , Genentech) have shown modest or no
results in clinical trials as single agent therapies in breast cancer (34–36). This may
be partly attributed to lack of predictive markers of response in combination with
drugs simply not being effective as single agents. However, this failure may also

140 Pedersen and Bartlett
reflect poorly designed approaches toward the discovery of predictive biomarkers.
Predictive markers for EGFR-targeted therapy have been in focus for the past five
years, and we will therefore briefly go through the history of EGFR-targeted
therapy and what has been learnt for future predictive marker development. In
addition, a model for codevelopment of predictive markers in relation to traditional
drug development is presented.
EGFR—A Cautionary Tale
EGFR or HER1 belongs to the same family of tyrosine kinase receptors as
HER2/neu and EGFR is considered an attractive target for therapeutic intervention
in many cancer forms due to the receptor being frequently mutated and
overexpressed (34). EGFR overexpression is associated with reduced diseasefree
survival and resistance to chemotherapy (37). EGFR inhibitors (e.g., erlotinib
and gefitinib) and therapeutic antibodies against the extracellular domain of
the receptor (e.g., cetuximab, matuzumab, and panitumumab) are currently in
clinical trials or late-stage development (38). Most clinical trials have investigated
EGFR inhibitors in cancers of the lung and colon, and cetuximab has
been approved for treatment of colorectal cancer (34). Clinical trials have also
been conducted in breast cancer, with limited success (39–47). In several of these
trials, patients have been selected for trial on the basis of EGFR overexpression.
However, this was later shown to be a mistake because EGFR expression, as
currently evaluated, did not correlate with clinical response (48,49) or with
breast cancer cell line inhibition (50). Clinical trials in breast cancer have clearly
demonstrated that EGFR inhibitors and therapeutic antibodies are targeting
EGFR receptors in both tumor and skin from patients, so the lack of effect is not
due to the drug’s failure to reach the target or the target’s lack of inhibition (41).
So EGFR inhibitors hit their target, and certain cancer patients clearly benefit
from these drugs. Therefore, it is surprising that although EGFR-targeted therapy
has been in focus for the past decade and drugs are now being administered to
patients, no FDA-approved test or profile exists at present to guide clinicians on
the use of EGFR inhibitors. Studies have shown EGFR gene copy number,
mutations, and downstream signalling to be associated with clinical response in a
series of lung cancer trials (51–55). However, these were retrospective, poorly
powered studies, which are of insufficient impact to influence molecular diagnostics.
Taken together, what has emerged from clinical trials in breast and other
cancers along with findings from in vitro studies is that EGFR inhibition is not
similar to HER2/neu-targeted therapy in breast cancer, where a single, wellcharacterized
molecular event can be used to target treatment. Therefore, what
constitutes a good predictive marker for EGFR-targeted therapy in breast cancer
is currently not clear. For future clinical trials with EGFR inhibitors to be successful
and current, approved EGFR inhibitors to be administered to the right
patients, it will be necessary to understand better the biology of EGFR-directed
signaling and the relative importance of mutations, gene copy number, and

Predictive Markers for Targeted Breast Cancer Treatment 141
downstream signaling pathways have on therapeutic response. These findings
can then be used to setup adequately designed molecular pathology and translational
studies within clinical trials that will address predictive markers for
EGFR-targeted therapy.
Predictive Markers in Drug Development
Future clinical trials should utilize findings from HER2/neu and EGFR-targeted
therapy. We have learnt that overexpression or amplification alone does not
necessarily mean that protein can be targeted with drugs or that it is indicative of
aberrant activation. Posttranslational modifications such as phosphorylation,
ubiquitination and acetylation, and localization are critical for protein function.
Furthermore, what has become clear is that preclinical models do not reflect the
complexity of human tumors well, and, therefore, predictive marker selection
must be validated in clinical settings. Despite all these challenges, drug development
must try and entail predictive marker development so that stratification
of patients can be made possible, thus avoiding unnecessary side effects and
improving the chance of a trial being successful. Without an increased focus on
predictive marker–directed patient selection, many of the drugs currently in
clinical trials run the risk of failing to reach the clinic despite being efficacious
for a subpopulation of patients. Therefore, it should be cost effective for pharmaceutical
companies to build predictive marker strategies in drug development,
as this may improve the success rate of drug development in clinicals. In
addition, it would be useful to develop predictive markers of response to many of
the drugs already in the market, thus improving patient care and reducing
treatment costs by only treating patients who benefit from the treatment.
The major risk we face at present is that we will continue to use poorly
designed retrospective analyses of existing trial populations to seek to identify
markers on the basis of incomplete understanding of the biology of drug
response. This is, in part, due to the fact that we have no reference criteria
against which to judge biomarker development. For ERa, despite no level-1
evidence being available, it is unthinkable that this biomarker be regarded as
anything less than essential to the management of breast cancer in the twentyfirst
century. For HER2, only after the central importance of this biomarker was
established were steps put in place to produce robust, accurate, and clinically
viable test methods. However, clearly for future biomarkers, a structured
approach will improve the likelihood of selecting appropriate markers. For this
to succeed, we need to progress from the current model to one where, as with
drug development, critical attention is paid to each step of the process by which
novel biomarkers are developed (56).
To develop and validate predictive markers, a number of hurdles have to be
passed before they can be considered for clinical practice. These processes should
be addressed in a structured manner to maximise the possibility of success. For
example, the development of robust and accurate diagnostic assays should

142 Pedersen and Bartlett
Figure 2 Development of IHC- and FISH-based predictive markers. Abbreviations: IHC,
immunohistochemical; FISH, fluorescence in situ hybridization.
precede, not follow, the identification of a potential clinical role for biomarkers
(56). To illustrate this, we have adapted a biomarker development scheme from
Pepe et al. and Hayes et al. (57,58) into four phases (Figure 2).
l
l
Phase 1: Discovery stage. In order to discover predictive markers of
response, high-throughput gene expression microarrays or proteomic
technologies are often used. The primary output of this phase is to develop
a set of potential predictive markers that allow selecting patients who
respond to treatment. Ideally, samples from the neoadjuvant setting should
be used so that information about the influence that treatment has on the
markers selected can be obtained.
Phase 2: Development stage. In this stage, the predictive marker assay now
needs make the transition from the research setting to the clinical diagnostic
laboratory. At this stage, it is required to develop the assay for
correct determination of predictive marker of response to a level where it

Predictive Markers for Targeted Breast Cancer Treatment 143
can be used by clinical institutions, according to guidelines put forward by
ASCO and EORTC (The European Organization for Research and
Treatment of Cancer) (58,59). This means that proper instructions need to
be in place for all aspects of predictive marker testing, including sample
handling and pretreatment, protocols and standards for ensuring uniform
testing across institutions, and external quality assurance. It is desirable
that a set of positive controls are developed for interlaboratory variation to
be accounted for in the same fashion as the HercepTest comes with positive
control cell lines.
l Phase 3: Validation phase. In order to validate markers found in phase 1,
retrospective validation studies need to be set up. When setting up clinical
trial at this stage, it is important that one of the endpoints of the trial is to
assess predictive markers. One of the primary outcomes of this phase is to
evaluate how well the predictive markers perform in independent
validation studies and what predictive markers perform compared better
with traditional diagnostic markers such as nodal status, grade, etc.
l Phase 4: Clinical utility phase. If success is obtained in phase 3, the marker
can be move to phase 4, with clinical studies utilizing randomized
prospective screening with predictive markers. The primary outcome of
phase 4 is to evaluate if patients selected with predictive markers were the
ones that benefited from treatment and if it is ethically defendable to select
patients on the basis of the results of the clinical studies. The predictive
marker then needs to go through regulatory approval for the defined
indication before being applied in the clinic.
Predictive marker development requires the joint effort of multiple academic
institutions and industry to cover all the aspects of development. It takes
many years and large investments, and the risk for failure is high because of the
nature of drug development. That may be the reason why so few new markers
have been introduced in recent years despite major research and development
efforts and new techniques.
Future Perspectives
Cancer is a disease that involves the complex interplay of multiple genes in a
framework of mutations in the cancer cell. In addition, tumor cells have interactions
with other cells, thereby further promoting growth and invasiveness.
Patients respond differently to treatment based on their tumor characteristics.
Despite recent challenges, we foresee the strong likelihood that current research
and drug development will introduce more predictive markers into IHC laboratories
in the same fashion as trastuzumab has resulted in HER2/neu being
routinely tested for (see above). This foresight is based on the premise that more
care and more robust approaches are applied to the development of predictive
biomarkers. It is likely that within two to five years, a panel of markers will

144 Pedersen and Bartlett
be tested on every tumor sample in order to decide optimal therapy regimen.
Currently only three predictive IHC markers (ERa, PR, and HER2/neu) are tested
for routinely in breast, but already some pathologists use additional markers to
inform decisions, e.g., p53, Ki67, and TOP2A. As more and more predictive IHC
markers are introduced in clinical laboratories, this will lead to an increased
pressure on pathology laboratories to deliver fast and accurate decisions. How can
this be achieved and what technological advances can improve IHC testing?
Current IHC testing in pathology laboratories is subjective and semiquantitative,
as slides are scored manually and using chromogenic dyes (56–61).
The use of computer-assisted analysis of predictive markers has been proven to
reduce slide-scoring variability among pathologists (31). While FISH/CISH
analysis is more robust, there remains room for improvement in this area also.
We foresee the introduction of image analysis as a standard tool for IHC scoring,
which will begin the process of moving the analytical phase from the microscope
to the computer screen.
As analytical challenges increase, accurate quantitation will become
increasingly important. Already evidence suggests that additional information
could be derived if it were possible to robustly quantify ERa and Ki67. There is,
therefore, a strong opportunity for systems such as the Automated QUantitative
Analysis (AQUA 1 ) from HistoRX, Inc., New Haven, Connecticut, U.S., to pave
the way for a switch from traditional chromogenic staining techniques to fluorescent
staining techniques which are potentially quantitative (wide dynamic
range) (62). A further advantage is that fluorescent labels can be multiplexed
together, yielding spatial information and quantization of proteins in different
compartments of the cell. New techniques being developed, such as the proximity
ligation in situ assay, show promising results for quantitative analysis of
specific protein interactions in FFPE, i.e., formalin-fixed, paraffin-embedded,
tissue (63). In some cases, this may be superior to the measurement of protein
concentrations or posttranslational modifications, as most proteins take part in
multiprotein complexes and their inclusion in the complex indicates relevant
activity. In addition, several protein-interaction inhibitors are currently in clinical
trials, and, therefore, measurement of the amount of target (interacting proteins) in
the tissue may be a predictive marker for responsiveness to the drug.
There remains a key question: Will we be testing protein and DNA
markers on slides at all in 5 to 10 years? There are a number of developments in
process now miniaturizing assays onto small chips that will measure hundreds of
protein or DNA markers quantitatively and in a reproducible fashion. Such
technological advances will change the pathology workflow in the future so that
every cancer sample coming into the clinic will be analyzed by the pathologist,
tumor cells microdissected into solution, and DNA, protein, and RNA extracted
and applied onto analytical chips. The readout from these analyses could then be
used to determine optimal therapy regimen for the patient and hopefully cure the
disease. However, until these techniques have been properly validated and
developed for clinical use and trials have proven their value, we are likely to rely

Predictive Markers for Targeted Breast Cancer Treatment 145
on traditional in situ techniques for protein analysis (IHC) and molecular
determination of gene amplifications (FISH/CISH).
In summary, introduction of quality control, automated platforms, and
computer-assisted scoring will lead to higher throughput and more accurate
decisions with regard to treatment and facilitate the development of future
predictive markers.
CONCLUSION
As our understanding of cancer as a disease progresses and the tools to investigate
cancer biology develops, the complexity of the disease is unraveling.
Breast cancers are now being classified not only on the basis of traditional
pathology traits but also according to molecular phenotype and, more recently,
gene expression profiles. New molecular targeted therapeutics will lead to more
predictive markers being introduced, and these markers must be prospectively
validated, internally and externally quality assured, and rigorously quantitative to
allow correct decisions for treatment. Therefore, an evolution of the pathology
laboratory is likely to take place in the near future, leading to a switch to
fluorescent-based techniques, digital imaging of slides, and computerized scoring
of predictive markers. On a longer time horizon, new technologies being
developed and applied are likely to change cancer diagnostics so that molecular
phenotyping of patients will take center stage in patient management. In the not
too distant future, there is the possibility that pathologists will type and stage
tumors and simultaneously identify key areas of tissue for automated microdissection
and processing. The subsequent process of molecular phenotyping,
which may include kinomic, proteomic, and transcriptomic assays performed
simultaneously could be carried out in specialised laboratories using robotics and
microfluidics on high-density protein and gene arrays to determine optimal
treatment regimens for tumors with specific molecular profiles linked to drug
response. Such a future vision may represent the more radical view of what could
be accomplished by linking of current concepts and technologies to maximize
patient benefit from therapeutic options. However, we must dream of such
futures if we are to maximize progress in the area of molecular pathology for the
benefit of the most important people, our patients.
REFERENCES
1. Edwards BK, Brown ML, Wingo PA, et al. Annual report to the nation on the status
of cancer, 1975-2002, featuring population-based trends in cancer treatment. J Natl
Cancer Inst 2005; 97(19):1407–1427.
2. Duffy MJ. Predictive markers in breast and other cancers: A review. Clin Chem
2005; 51(3):494–503.
3. Osborne CK. Steroid hormone receptors in breast cancer management. Breast Cancer
Res Treat 1998; 51(3):227–238.

146 Pedersen and Bartlett
4. Jensen EV, DeSombre ER. Mechanism of action of the female sex hormones. Annu
Rev Biochem 1972; 41:203–230.
5. Gorski J, Toft D, Shyamala G, et al. Hormone receptors: studies on the interaction of
estrogen with the uterus. Recent Prog Horm Res 1968; 24:45–80.
6. Osborne CK, Yochmowitz MG, Knight WA III, et al. The value of estrogen and
progesterone receptors in the treatment of breast cancer. Cancer 1980; 46(12 suppl):
2884–2888.
7. Horwitz KB, McGuire WL. Predicting response to endocrine therapy in human
breast cancer: a hypothesis. Science 1975; 189(4204):726–727.
8. Elledge RM, Green S, Pugh R, et al. Estrogen receptor (ER) and progesterone
receptor (PgR), by ligand-binding assay compared with ER, PgR and pS2, by
immuno-histochemistry in predicting response to tamoxifen in metastatic breast
cancer: a Southwest Oncology Group Study. Int J Cancer 2000; 89(2):111–117.
9. Ravdin PM, Green S, Dorr TM, et al. Prognostic significance of progesterone
receptor levels in estrogen receptor-positive patients with metastatic breast cancer
treated with tamoxifen: results of a prospective Southwest Oncology Group study.
J Clin Oncol 1992; 10(8):1284–1291.
10. Bardou VJ, Arpino G, Elledge RM, et al. Progesterone receptor status significantly
improves outcome prediction over estrogen receptor status alone for adjuvant endocrine
therapy in two large breast cancer databases. J Clin Oncol 2003; 21(10): 1973–1979.
11. Jordan VC, Brodie AM. Development and evolution of therapies targeted to the
estrogen receptor for the treatment and prevention of breast cancer. Steroids 2007;
72(1):7–25.
12. Allred DC, Harvey JM, Berardo M, et al. Prognostic and predictive factors in breast
cancer by immunohistochemical analysis. Mod Pathol 1998; 11(2):155–168.
13. Rhodes A, Jasani B, Barnes DM, et al. Reliability of immunohistochemical demonstration
of oestrogen receptors in routine practice: interlaboratory variance in the
sensitivity of detection and evaluation of scoring systems. J Clin Pathol 2000; 53(2):
125–130.
14. Schnitt SJ. Estrogen receptor testing of breast cancer in current clinical practice:
what’s the question? J Clin Oncol 2006; 24(12):1797–1799.
15. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifentreated,
node-negative breast cancer. N Engl J Med 2004; 351(27):2817–2826.
16. Coussens L, Yang-Feng TL, Liao YC, et al. Tyrosine kinase receptor with extensive
homology to EGF receptor shares chromosomal location with neu oncogene. Science
1985; 230(4730):1132–1139.
17. King CR, Kraus MH, Aaronson SA. Amplification of a novel v-erbB-related gene in
a human mammary carcinoma. Science 1985; 229(4717):974–976.
18. Yamamoto T, Ikawa S, Akiyama T, et al. Similarity of protein encoded by the human
c-erb-B-2 gene to epidermal growth factor receptor. Nature 1986; 319(6050):230–234.
19. Ross JS, Fletcher JA, Bloom KJ, et al. Targeted therapy in breast cancer: the HER-2/
neu gene and protein. Mol Cell Proteomics 2004; 3(4):379–398.
20. Carter P, Presta L, Gorman CM, et al. Humanization of an anti-p185HER2 antibody
for human cancer therapy. Proc Natl Acad Sci U S A 1992; 89(10):4285–4289.
21. Hortobagyi GN. Overview of treatment results with trastuzumab (Herceptin) in
metastatic breast cancer. Semin Oncol 2001; 28(6 suppl 18):43–47.
22. Piccart-Gebhart MJ, Procter M, Leyland-Jones B, et al. Trastuzumab after adjuvant
chemotherapy in HER2-positive breast cancer. N Engl J Med 2005; 353(16):1659–1672.

Predictive Markers for Targeted Breast Cancer Treatment 147
23. Vogel CL, Cobleigh MA, Tripathy D, et al. Efficacy and safety of trastuzumab as a
single agent in first-line treatment of HER2-overexpressing metastatic breast cancer.
J Clin Oncol 2002; 20(3):719–726.
24. Cobleigh MA, Vogel CL, Tripathy D, et al. Multinational study of the efficacy and
safety of humanized anti-HER2 monoclonal antibody in women who have HER2-
overexpressing metastatic breast cancer that has progressed after chemotherapy for
metastatic disease. J Clin Oncol 1999; 17(9):2639–2648.
25. Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a monoclonal
antibody against HER2 for metastatic breast cancer that overexpresses HER2.
N Engl J Med 2001; 344(11):783–792.
26. Pegram MD, Lipton A, Hayes DF, et al. Phase II study of receptor-enhanced
chemosensitivity using recombinant humanized anti-p185HER2/neu monoclonal
antibody plus cisplatin in patients with HER2/neu-overexpressing metastatic breast
cancer refractory to chemotherapy treatment. J Clin Oncol 1998; 16(8):2659–2671.
27. Baselga J, Tripathy D, Mendelsohn J, et al. Phase II study of weekly intravenous
recombinant humanized anti-p185HER2 monoclonal antibody in patients with HER2/
neu-overexpressing metastatic breast cancer. J Clin Oncol 1996; 14(3):737–744.
28. Fornier M, Risio M, Van Poznak C, et al. HER2 testing and correlation with efficacy
of trastuzumab therapy. Oncology (Williston Park) 2002; 16(10):1340–1342.
29. Press MF, Slamon DJ, Flom KJ, et al. Evaluation of HER-2/neu gene amplification and
overexpression: comparison of frequently used assay methods in a molecularly characterized
cohort of breast cancer specimens. J Clin Oncol 2002; 20(14): 3095–3105.
30. Lebeau A, Deimling D, Kaltz C, et al. Her-2/neu analysis in archival tissue samples
of human breast cancer: comparison of immunohistochemistry and fluorescence in
situ hybridization. J Clin Oncol 2001; 19(2):354–363.
31. Wang S, Saboorian MH, Frenkel E, et al. Laboratory assessment of the status of
Her-2/neu protein and oncogene in breast cancer specimens: comparison of immunohistochemistry
assay with fluorescence in situ hybridisation assays. J Clin Pathol
2000; 53(5):374–381.
32. Moasser MM. Targeting the function of the HER2 oncogene in human cancer
therapeutics. Oncogene 2007; 26(46):6577–6592.
33. Bartlett JM. Pharmacodiagnostic testing in breast cancer: focus on HER2 and trastuzumab
therapy. Am J Pharmacogenomics 2005; 5(5):303–315.
34. Scaltriti M, Baselga J. The epidermal growth factor receptor pathway: a model for
targeted therapy. Clin Cancer Res 2006; 12(18):5268–5272.
35. Chan S, Scheulen ME, Johnston S, et al. Phase II study of temsirolimus (CCI-779), a
novel inhibitor of mTOR, in heavily pretreated patients with locally advanced or
metastatic breast cancer. J Clin Oncol 2005; 23(23):5314–5322.
36. Schneider BP, Sledge GW, Jr. Drug insight: VEGF as a therapeutic target for breast
cancer. Nat Clin Pract Oncol 2007; 4(3):181–189.
37. Tsutsui S, Kataoka A, Ohno S, et al. Prognostic and predictive value of epidermal
growth factor receptor in recurrent breast cancer. Clin Cancer Res 2002; 8(11):
3454–3460.
38. Johnston JB, Navaratnam S, Pitz MW, et al. Targeting the EGFR pathway for cancer
therapy. Curr Med Chem 2006; 13(29):3483–3492.
39. Moulder SL, Arteaga CL. A Phase I/II Trial of trastuzumab and gefitinib in patients
with Metastatic Breast Cancer that overexpresses HER2/neu (ErbB-2). Clin Breast
Cancer 2003; 4(2):142–145.

148 Pedersen and Bartlett
40. Tan AR, Yang X, Hewitt SM, et al. Evaluation of biologic end points and pharmacokinetics
in patients with metastatic breast cancer after treatment with erlotinib, an
epidermal growth factor receptor tyrosine kinase inhibitor. J Clin Oncol 2004; 22(15):
3080–3090.
41. Baselga J, Albanell J, Ruiz A, et al. Phase II and tumor pharmacodynamic study of
gefitinib in patients with advanced breast cancer. J Clin Oncol 2005; 23(23):5323–5333.
42. Fountzilas G, Pectasides D, Kalogera-Fountzila A, et al. Paclitaxel and carboplatin as
first-line chemotherapy combined with gefitinib (IRESSA) in patients with advanced
breast cancer: a phase I/II study conducted by the Hellenic Cooperative Oncology
Group. Breast Cancer Res Treat 2005; 92(1):1–9.
43. Gasparini G, Sarmiento R, Amici S, et al. Gefitinib (ZD1839) combined with weekly
epirubicin in patients with metastatic breast cancer: a phase I study with biological
correlate. Ann Oncol 2005; 16(12):1867–1873.
44. Polychronis A, Sinnett HD, Hadjiminas D, et al. Preoperative gefitinib versus
gefitinib and anastrozole in postmenopausal patients with oestrogen-receptor positive
and epidermal-growth-factor-receptor-positive primary breast cancer: a double-blind
placebo-controlled phase II randomised trial. Lancet Oncol 2005; 6(6):383–391.
45. von Minckwitz G, Jonat W, Fasching P, et al. A multicentre phase II study on
gefitinib in taxane- and anthracycline-pretreated metastatic breast cancer. Breast
Cancer Res Treat 2005; 89(2):165–172.
46. Ciardiello F, Troiani T, Caputo F, et al. Phase II study of gefitinib in combination
with docetaxel as first-line therapy in metastatic breast cancer. Br J Cancer 2006;
94(11):1604–1609.
47. Modi S, D’Andrea G, Norton L, et al. A phase I study of cetuximab/paclitaxel in
patients with advanced-stage breast cancer. Clin Breast Cancer 2006; 7(3):270–277.
48. Chung KY, Shia J, Kemeny NE, et al. Cetuximab shows activity in colorectal cancer
patients with tumors that do not express the epidermal growth factor receptor by
immunohistochemistry. J Clin Oncol 2005; 23(9):1803–1810.
49. Vincenzi B, Santini D, Rabitti C, et al. Cetuximab and irinotecan as third-line
therapy in advanced colorectal cancer patients: a single centre phase II trial. Br
J Cancer 2006; 94(6):792–797.
50. Campiglio M, Locatelli A, Olgiati C, et al. Inhibition of proliferation and induction
of apoptosis in breast cancer cells by the epidermal growth factor receptor (EGFR)
tyrosine kinase inhibitor ZD1839 (‘Iressa’) is independent of EGFR expression level.
J Cell Physiol 2004; 198(2):259–268.
51. Uramoto H, Mitsudomi T. Which biomarker predicts benefit from EGFR-TKI
treatment for patients with lung cancer? Br J Cancer 2007; 96(6):857–863.
52. Lynch TJ, Bell DW, Sordella R, et al. Activating mutations in the epidermal growth
factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib.
N Engl J Med 2004; 350(21):2129–2139.
53. Yun CH, Boggon TJ, Li Y, et al. Structures of lung cancer-derived EGFR mutants
and inhibitor complexes: mechanism of activation and insights into differential
inhibitor sensitivity. Cancer Cell 2007; 11(3):217–227.
54. Benvenuti S, Sartore-Bianchi A, Di Nicolantonio F, et al. Oncogenic activation of
the RAS/RAF signaling pathway impairs the response of metastatic colorectal cancers
to anti-epidermal growth factor receptor antibody therapies. Cancer Res 2007;
67(6):2643–2648.

Predictive Markers for Targeted Breast Cancer Treatment 149
55. Pao W, Wang TY, Riely GJ, et al. KRAS mutations and primary resistance of lung
adenocarcinomas to gefitinib or erlotinib. PLoS Med 2005; 2(1):E17.
56. Bartlett JM, Forsyth A. Detection of HER2 gene amplification by fluorescence in
situ hybridization in breast cancer. Methods Mol Med 2006; 120:309–322.
57. Pepe MS, Etzioni R, Feng Z, et al. Phases of biomarker development for early
detection of cancer. J Natl Cancer Inst 2001; 93(14):1054–1061.
58. Hayes DF, Bast RC, Desch CE, et al. Tumor marker utility grading system:
a framework to evaluate clinical utility of tumor markers. J Natl Cancer Inst 1996;
88(20):1456–1466.
59. Sweep FC, Fritsche HA, Gion M, et al. Considerations on development, validation,
application, and quality control of immuno(metric) biomarker assays in clinical
cancer research: an EORTC-NCI working group report. Int J Oncol 2003; 23(6):
1715–1726.
60. Bartlett J, Mallon E, Cooke T. The clinical evaluation of HER-2 status: which test to
use? J Pathol 2003; 199(4):411–417.
61. Bartlett JM, Going JJ, Mallon EA, et al. Evaluating HER2 amplification and overexpression
in breast cancer. J Pathol 2001; 195(4):422–428.
62. Rimm DL. What brown cannot do for you. Nat Biotechnol 2006; 24(8):914–916.
63. Soderberg O, Gullberg M, Jarvius M, et al. Direct observation of individual endogenous
protein complexes in situ by proximity ligation. Nat Methods 2006; 3(12):995–1000.

11
Circulating Tumor Cells in Individualizing
Breast Cancer Therapy
James M. Reuben
Department of Hematopathology, The University of Texas
M. D. Anderson Cancer Center, Houston, Texas, U.S.A.
Massimo Cristofanilli
Department of Breast Medical Oncology, The University of Texas
M. D. Anderson Cancer Center, Houston, Texas, U.S.A.
INTRODUCTION
Breast cancer is the most frequently diagnosed nonskin cancer among women
in the United States and is second only to lung cancer in causing cancer-related
deaths among them (1). The vast majority of deaths are due to recurrent
metastatic disease. Occult dissemination of tumor cells is the main cause of
recurrent metastatic breast cancer (MBC) in patients who have undergone
resection of their primary tumor (2). Approximately 5% of patients with breast
cancer have clinically detectable metastases at the time of initial diagnosis, and
a further 30% to 40% of patients who appear clinically free of metastases
harbor occult metastases (3,4). The formation of metastatic colonies is a
continuous process, commencing early during the growth of the primary tumor.
Metastasis occurs through a cascade of linked sequential steps involving
multiple host-tumor interactions. This complex process requires the cells to
enter the circulation, arrest at the distant vascular bed, extravasate into the
organ interstitium and parenchyma, and proliferate as a secondary colony.
151

152 Reuben and Cristofanilli
Several experimental studies suggest that during each stage of the process, only
the fittest tumor cells survive (2).
The first report on tumor cells in the peripheral circulation was attributed
to Ashworth in 1869 (5). Since then, the existence, origin, and clinical significance
of circulating tumor cells (CTCs) have been debated. The introduction of
sensitive and specific immunohistochemical techniques in the late 1970s led to
renewed interest in the detection of CTCs and their possible association with
minimal residual disease in solid malignancies (6,7). However, there is little
understanding of the genetic and phenotypic characteristics of CTCs in peripheral
blood. It has been speculated that mammary stem cells represent the cellular
origins of cancer because they exist quiescently over long periods and can
accumulate multiple mutations over the lifetime, ultimately giving rise to tumors
when stimulated to proliferate (8–10). It was recently reported that highly
tumorigenic cells possessing properties consistent with those of stem or progenitor
cells could be isolated from human breast cancers (9). Furthermore,
transplantation of as many as several hundred of these cells into the cleared
mammary fat pads of etoposide-treated mice resulted in the growth of breast
tumors (10). Cancer stem cells not only exist in human breast cancer but also
have different biologic features than the more differentiated cells that constitute
the majority of cells in human breast cancers. These cancer stem cells may be
detectable in the peripheral blood and bone marrow of patients with MBC and
may account for their worse prognosis. Hence the development of therapies
specifically targeting cancer stem cells may have great benefits for patients with
CTCs who possess the properties of cancer stem cells.
When CTCs migrate to distal sites, they can potentially give rise to
microscopic lesions that result in metastasis.Microscopic disease may be present in
lymph nodes, bone marrow (primary breast cancer), and peripheral blood (metastatic
disease) at the time of breast cancer diagnosis (3,4,11,12). Most studies have
demonstrated that the detection of microscopic disease in patients with breast
cancer contributes prognostic information and, in selected cases, can predict the
efficacy of treatments (12). In patients with primary breast cancer, the detection of
microscopic disease in lymph nodes and bone marrow has led to a better understanding
of the role of minimal residual disease in establishing metastasis.
Over the past few years, immunomagnetic separation technology, with its
higher level of sensitivity and specificity, has been used to improve the detection
of CTCs compared with the detection of occult CTCs by reverse transcriptasepolymerase
chain reaction (RT-PCR) (13–21). In this chapter, we review the
methods of detecting CTCs, the prognostic implications of CTCs, and the
potential means of exploiting CTCs in developing individualized therapy.
DETECTION OF CIRCULATING TUMOR CELLS
Advances in technology have facilitated the detection of even very small numbers
of CTCs in the peripheral blood of cancer patients. The principle of these
methods is based on the fact that breast cancer cells express epithelial cell

Circulating Tumor Cells in Individualizing Breast Cancer Therapy 153
adhesion molecule-1 (EpCAM). EpCAM, a 40-kDa glycoprotein, is present on
most epithelial carcinomas, and recent studies indicate that this molecule has a
major morphoregulatory function, relevant not only to epithelial tissue development
but also to carcinogenesis and tumor progression (22,23).
Detection of CTCs by the CellSearch TM System
The only assay approved by the U.S. Food and Drug Administration is the
CellSearch TM system (Veridex, LLC, Warren, New Jersey, U.S.). With this assay,
a small sample of peripheral blood is allowed to react with ferrofluids coated with
antibodies to EpCAM. Next, the sample is processed through a magnetic field to
retain the EpCAM-positive cells, while the EpCAM-negative cells, primarily of
hematopoietic origin, are eluted from the column and discarded. Thereafter, the
enriched EpCAM-positive cells are allowed to react with the nucleic acid dye
4 0 ,6-doamidino-2-phenylindole (DAPI), monoclonal antibodies specific for leukocytes
(anti-CD45 antibodies conjugated with allophycocyanin), and epithelial
cells (phycoerythrin-conjugated antibodies to cytokeratin 8, 18, and 19) and
analyzed by the CellSpotter TM Analyzer (Veridex). The CellSpotter Analyzer is a
semiautomated fluorescence-based microscopy system that enables computergenerated
reconstruction of cellular images. The analyzer interrogates each cell
image to determine if it meets the very stringent requirements of an algorithm to
be classified as a CTC. The algorithm insures that a CTC expresses EpCAM and
not leukocyte lineage-specific antigens (represented by CD45), exhibits cytoplasmic
expression of cytokeratin, and contains a nucleus that binds DAPI.
Absence of any of these characteristics disqualifies a cell image as a CTC. In
identifying CTCs, the procedure identifies cell objects that possess some, but not
all, of the required characteristics to be a CTC and are labeled as “unclassified”
objects (24). The stringent criteria of CellSearch for identifying a CTC are
laudable; however, the merits of the assay continue to be debated as it minimizes
the scientific or clinical significance of unclassified objects.
Although CellSearch is a very reproducible and reliable method for the
identification and enumeration of CTCs in the peripheral blood of patients with
breast cancer, its clinical utility has been limited by its high cost. The semiautomated
assay is also somewhat labor intensive. In addition, the need to permeabilize
the cell membrane of a viable cell to introduce DAPI and anticytokeratin
antibodies to label intracellular structures further deters the possibility of interrogating
viable cells for their growth potential, colony formation, and genomics,
among others. To address these issues, the CellProfile TM kit was introduced that
permits the retrieval of EpCAM-positive cells prior to cell permeabilization and
interrogation for genomic or proteomic profiles of the CTCs.
Isolation of CTCs by the AutoMACS TM System
EpCAM, also known as human epithelial cell antigen (HEA) or CD326, is
expressed on the majority of tumor cells of epithelial cell origin but not on

154 Reuben and Cristofanilli
circulating B cells, T cells, or monocytes. In another assay that exploits this
characteristic of CTCs, peripheral blood mononuclear cells are incubated with
magnetic beads coated with anti-CD326 (Miltenyi Biotec, Inc., Auburn, California,
U.S.) prior to passage through a magnetic column to enrich for CTCs. Typically,
mononuclear cells are isolated from peripheral blood, bone marrow, pleural
fluid, paracentesis fluid, or other tissues, mixed with anti-CD326 microbeads, and
incubated at 48C to88C for 15 minutes. After washing, the cell pellet is resuspended
and loaded onto the magnetic column of an automatic magnetic cell
sorting (AutoMACS TM ) system (Miltenyi Biotec) (25). CD326-positive CTCs are
isolated using the positive selection protocol and can then be centrifuged onto
glass slides to determine the morphology, viability, and purity of the preparation.
In one study, Papanicolaou (Pap)-stained cytospin smears were analyzed in
detail to identify tumor cells on the basis of conventional cytologic criteria. In
cytospin slides of enriched CTCs, large cells with cytomorphological features
suspicious for tumor cells were detected, and subsequently the cytospin slides
were immunocytochemically stained with the avidin-biotin complex peroxidase
technique using pancytokeratin antibody for the identification of tumor cells
(Fig. 1A). (The immunostains of CTCs were performed and interpreted by
Dr. Savitri Krishnamurthy, Department of Pathology, The University of Texas
M. D. Anderson Cancer Center).
A second cytospin slide of CD326-positive bone marrow tumor cells from
the same patient was stained with a cocktail of anticytokeratin antibodies using
the avidin-biotin complex peroxidase technique for the identification of tumor
cells (Fig. 1B). In the majority of cases, the suspicious cells were positive for
cytokeratin, thereby confirming the presence of micrometastatic carcinoma.
In addition to bone marrow, it is possible to isolate CTCs from the
paracentesis fluid of a patient with MBC. Figure 2 shows a cluster of CTCs
Figure 1A A cytospin slide of CTCs from the bone marrow of a patient with primary
breast cancer using anti-CD326 antibody–coated microbeads and reacted with Pap stain.
The figure shows a cluster of CTCs with cellular morphology consistent with that of
epithelial cells. Abbreviations: CTCs, circulating tumor cells; Pap, Papanicolaou.

Circulating Tumor Cells in Individualizing Breast Cancer Therapy 155
Figure 1B CD326-positive CTCs are cytokeratin-positive epithelial cells. Abbreviation:
CTCs, circulating tumor cells.
Figure 2 CD326-positive CTCs from the paracentesis fluid of a patient with MBC.
Abbreviations: CTCs, circulating tumor cells; MBC, metastatic breast cancer.
isolated from the paracentesis fluid of a patient with MBC that were reacted with
Pap stain.
Isolation of CTCs by the EasySep TM System
Another immunomagnetic cell selection system is EasySep TM (StemCell Technologies,
Inc., Vancouver, Canada), which combines the specificity of monoclonal
antibodies with the simplicity of a column-free magnetic system. In this
procedure, highly enriched cells are obtained by positively selecting the cells of
interest with antibodies to a specific cell-surface antigen. The cells targeted for

156 Reuben and Cristofanilli
selection are cross-linked to EasySep nanoparticles via the formation of tetrameric
antibody complexes (StemCell Technologies) in a standard centrifuge
tube, which is then placed in the EasySep magnetic chamber. The handheld
magnetic chamber is gently inverted to discard the cells that are not bound to
the specific nanoparticle complexes that adhere to the walls of the tube in the
magnetic chamber. To obtain a highly pure population, the residual cells in the
tube are rinsed twice and then harvested by removing the tube from the chamber.
Enriched CTCs can then be used for further interrogation for the differential
expression of EpCAM or other markers of interest.
Positive selection is most effective when a specimen has a generous
complement of the desired cell population. When a specimen, such as peripheral
blood, contains few or inadequate numbers of the cells of interest, a negativeselection
approach can be taken to enrich for these cells. One negative-selection
technique is the RosetteSep TM (StemCell Technologies) method that uses a
highly purified combination of mouse and rat monoclonal antibodies. When
reacted with peripheral whole blood, these antibodies bind in bispecific tetrameric
antibody complexes directed against cell surface antigens on human
hematopoietic cells (CD2, CD16, CD19, CD36, CD38, CD45, and CD66b) and
glycophorin A on red blood cells, or erythrocytes. The unwanted cells of hematopoietic
origin form rosettes by cross-linking to multiple erythrocytes that can
be easily removed by performing a typical Ficoll-Paque TM (GE Amersham-
Pharmacia Diagnostics, Uppsala, Sweden) density gradient procedure for the
separation of mononuclear cells from peripheral blood. The rosettes, free
erythrocytes, and granulocytes form a pellet, while the unlabeled, desired CTCs
are collected from the interface between the plasma and the buoyant density
medium. Figures 3A and 3B represent cytospin slides of CTCs isolated from the
Figure 3A Pap stain of a cluster of CTCs isolated from the peripheral blood of a patient
with MBC using RosetteSep TM . Abbreviations: Pap, Papanicolaou; CTCs, circulating
tumor cells; MBC, metastatic breast cancer.

Circulating Tumor Cells in Individualizing Breast Cancer Therapy 157
Figure 3B Cytokeratin-positive CTCs from the peripheral blood of a patient with MBC
using RosetteSep TM . Abbreviations: CTCs, circulating tumor cells; MBC, metastatic
breast cancer.
peripheral blood of a patient with MBC that were reacted with Pap stain
(Fig. 3A) and anticytokeratin antibodies (Fig. 3A).
Detection of CTCs by AdnaGen’s Technology
In addition to magnetic separation methodologies, PCR greatly facilitates the
detection of occult tumor cells through the use of nucleic acid analysis. The
PCR-based assay, Adna Test Breast Cancer Select, developed by AdnaGen AG,
Langenhagen, Germany, makes use of RT-PCR to identify putative transcripts of
genes in EpCAM-positive cells that are isolated by a magnetic separation
method. This approach takes advantage of the prevailing state of the art of CTC
isolation and enhances the chance of detecting very low numbers of CTCs
or occult CTCs on the basis of their expression of tumor-associated genes.
AdnaGen’s two-step “combination-of-combinations principle” involves initially
the enrichment of CTCs using an antibody mix linked to magnetic particles (26).
Thereafter, mRNA is extracted from the CTCs and transcribed into cDNA before
being subjected to multiplex RT-PCR for the detection of tumor-associated
genes. Owing to the combination of different selection and tumor markers, both
the heterogeneity of the tumor cells and the possible therapy-induced deviations
in the expression patterns are taken into account. The AdnaGen system has high
analytical sensitivity (two cells in 5 mL of blood) and high specificity (>90%)
achieved through the combination of multimarker tumor cell enrichment and
multiplex gene expression profiling.

158 Reuben and Cristofanilli
CLINICAL SIGNIFICANCE AND PROGNOSTIC VALUE OF CTCs in MBC
Tumor cells can be detected in the locoregional lymph nodes of women with
early breast cancer, and the presence of these cells has been shown to have a
negative effect on long-term prognosis (3,12). Despite evidence of the prognostic
value of CTCs in some studies, the detection of micrometastases was never
incorporated into cancer-staging protocols or considered a valuable clinical tool.
This may be the result of a combination of factors, such as variable antigen
expression in poorly differentiated tumors and reports of cytokeratin and epithelial
membrane antigen positivity in cells that are not of epithelial origin,
which demonstrated a need for more sensitive and specific methods of detection
than were available at the time. This need was filled to some degree by sensitive
PCR techniques (14,16,17). In the past decade, a few studies have shown that the
detection of occult disease by PCR has prognostic significance in some solid
tumors (13). However, PCR-based assays for the detection of occult tumor cells
have limitations, particularly contamination of samples, specificity of the assays,
and inability to quantify tumor cells. Moreover, PCR-based methods cannot be
used to perform functional assays. These factors have precluded the widespread
use of PCR in in vitro diagnostic applications.
Using an immunomagnetic detection approach, Austrup et al. reported the
prognostic significance of genomic alterations (e.g., c-erbB-2 overexpression)
present in CTCs purified from the blood of patients with breast cancer (27). The
authors investigated genomic imbalances, such as mutation, amplification, and
loss of heterozygosity, of 13 tumor suppressor genes and 2 proto-oncogenes
using DNA isolated from minimal residual cancer cells. The presence and the
number of genomic imbalances measured in disseminated tumor cells were
significantly associated with a worse prognosis (27).
Subsequently, Meng et al. demonstrated that CTCs recapitulate the human
epidermal growth factor receptor 2 (HER2) status of the primary tumor (28).
Furthermore, the authors demonstrated that a fraction of patients with HER2-
negative primary tumors had detectable HER2 gene amplification in their CTCs,
suggesting acquisition or selection of this phenotype during cancer progression
(37.5%; 95% confidence interval, 18.8–59.4%). Intriguingly, the initiation of
trastuzumab-based therapy in a few cases with altered HER2 status in CTCs was
associated with a clinical response (29). These important observations support
the argument that detection of CTCs and determination of their gene amplification
or expression can be used for better tailoring of therapies.
The predictive and prognostic roles of CTCs were investigated in a prospective
multicenter clinical trial led by researchers at the University of Texas
M. D. Anderson Cancer Center (11). In this study, the CellSearch system was
used to prospectively determine the prognostic and predictive value of CTCs in
patients with MBC who were about to start a new systemic treatment. The 177
enrolled patients underwent peripheral blood collection at monthly intervals for
up to six months after enrollment. A cutoff of 5 CTCs/7.5 mL was used to

Circulating Tumor Cells in Individualizing Breast Cancer Therapy 159
stratify patients into positive and negative groups (positive, 5 CTCs/7.5 mL;
negative, 18 months; P <
0.0001). On multivariate Cox hazards regression analysis, CTC levels, both at
baseline and at first follow-up, were the most significant predictors of progression-free
and overall survival (11).
An analysis restricted to patients who were about to start first-line systemic
therapy after the diagnosis of MBC had similar findings (30). A subsequent
investigation including patients with either measurable or evaluable MBC confirmed
that CTCs are an independent predictor of survival that is more powerful
than standard prognostic measures, such as hormone receptor status, and measures
of tumor burden (e.g., the Swenerton score or level of CA27-29) (31).
The detection of changes in CTC status may also have predictive utility.
The CTC detection rate at first follow-up was lower than at baseline, particularly
in patients undergoing first-line treatment (25% versus 52%) and those
with visceral disease (28% versus 50%) (11). In patients with newly diagnosed
disease treated with chemotherapy alone (n ¼ 37), the percentage of patients
who were CTC positive decreased significantly from baseline to first follow-up
(57% to 37%, P ¼ 0.001) and first imaging-visit (at eight to nine weeks) blood
draw (57% to 31%, P ¼ 0.004) (31). More importantly, in patients receiving
trastuzumab-containing regimens, CTC-positive patients decreased from
baseline to first follow-up (59% to 7%, P ¼ 0.600) and first imaging-visit
blood draw (59% to 0%, P ¼ not applicable). This data indicates that patients
with newly diagnosed disease who are about to start first-line therapy can have
detectable CTCs and that the changes in CTC status at three to four weeks may
indicate a benefit from systemic treatment (particularly trastuzumab-based
regimens). Moreover, the determination of CTCs at baseline and follow-up
appears to be a superior prognostic marker in patients with measurable MBC
compared with standard imaging assessments (32). CTCs could be used for
assessment of the clinical benefit of systemic treatments, for prognostic
stratification, and for evaluation of patients with nonmeasurable disease in
future prospective validation trials.
STAGING MBC USING CTCs
The American Joint Committee on Cancer (AJCC) classification system, based
on a schema developed by the International Union Against Cancer (UICC)
(33–35), includes much of the traditional prognostic information used by clinicians
when developing a comprehensive treatment plan. In brief, the AJCC

160 Reuben and Cristofanilli
system attempts to define the disease by incorporating all aspects of cancer
distribution in terms of the primary tumor (T), lymph nodes (N), and distant
metastasis (M). A “stage” group (0, I, II, III, or IV) is then assigned on the basis
of the possible TNM permutations, with 0 reflecting minimal involvement and
IV either the maximum tumor involvement or distant metastasis. This basic
TNM staging is then further subdivided according to the time the evaluation is
performed: c ¼ clinical (before surgical treatment) and p ¼ pathologic (after
surgical specimen analysis) (10,35).
Over the past 45 years, the AJCC has regularly updated its staging standards
to incorporate advances in prognostic technology. The committee’s current
work concerns the development of prognostic indices based on molecular
markers. However, until these changes are incorporated, TNM staging continues
to quantify only the physical extent of the disease. Although it covers the
approximately 2% of women who present with metastatic disease (stage IV), the
prognostic information is applicable only to in situ, local, and regional primary
breast cancers. Given the heterogeneity of the disease in primary and metastatic
settings, the potential for continued mutation, and the variety of treatment
options available, the information acquired at the time of the initial diagnosis of
breast cancer may not be as relevant to planning the treatment for recurrence in
the approximately 30% of women who have recurrent MBC years after their
initial diagnosis.
The recent demonstration that the presence of CTCs predicts the prognosis
of two subgroups of patients with MBC raises the possibility that this method
will allow for a true “biologic staging” of breast cancer (e.g., stages IVA and
IVB) (11). The main limitations of the previous study were the sample size (only
177 patients); the inclusion only of patients with measurable disease, which does
not reflect the heterogeneity of MBC; and the inclusion of patients at different
points in treatment—those with newly diagnosed disease and those undergoing
second- or third-line treatment. To overcome these limitations, a larger international
validation trial, the International Stage IV Stratification Study, was
begun. The objective of this study was to confirm the prognostic value of CTC
detection in a larger and more homogeneous cohort of patients with newly
diagnosed MBC. The study aimed at enrolling 660 patients and providing a more
detailed analysis of the association between CTC detection and other factors
such as ethnicity (black compared with white and Hispanic groups) and specific
disease sites (visceral versus nonvisceral disease).
CLINICAL APPLICATION OF CTCs
Despite years of clinical research, the odds of patients with MBC achieving a
complete response remain extremely low. Thus, the main goal in the management
of MBC is palliation (36). Only a few patients who experience a complete
response after chemotherapy remain in this state for prolonged periods, although

Circulating Tumor Cells in Individualizing Breast Cancer Therapy 161
some have remained in remission for more than 20 years (37). These long-term
survivors are usually young, have an excellent performance status, and, more
importantly, have limited metastatic disease (37,38). Most patients with MBC
respond only transiently to conventional therapies and develop evidence of
progressive disease within 12 to 24 months of starting treatment. For these
patients, systemic treatment does not result in a significant improvement in
survival time but may improve their quality of life.
At present, clinicians use three different systemic treatment modalities for
advanced breast cancer: endocrine therapy, chemotherapy, and biologic targeted
therapy (39–42). Appropriate selection of patients for these modalities is based
mostly on tissue assessment of hormone receptor status (estrogen and progesterone
receptors) and c-erbB-2 status. In patients who lack expression of hormone
receptors or who demonstrate no amplification of c-erbB-2, cytotoxic
chemotherapy is used. However, no standard therapeutic regimen has been
defined. For example, the optimal schedule of chemotherapy administration in
MBC (i.e., concurrent vs. sequential) remains controversial, and the decision
must be individualized. Sledge et al. addressed this issue in a prospective study
that included 739 chemotherapy-naïve patients who were randomly assigned to
receive, at progression, doxorubicin, paclitaxel, or both (39). Although the
response rates and times to treatment failure were improved with the combination
regimen, the overall survival rate was comparable in all groups. Other trials
have demonstrated a survival advantage with combination regimens over singleagent
chemotherapy, but all of these studies have shown differences in the actual
median overall survival time between 10 and 13 months across studies, suggesting
that even with comparable inclusion criteria, the heterogeneity that is
typical of MBC cannot be eliminated (41,42). This heterogeneity is indeed one
of the major limitations to the development of more personalized treatments for
patients. Therefore, although the data demonstrate that patients can benefit from
combination therapy, they do not clearly identify the subsets of patients who will
most benefit or who will experience only additional toxicity.
In this context, the use of CTCs to stratify patients at the time of disease
recurrence may be an appropriate way to design personalized therapeutic
approaches. CTC detection may enable a more rational selection of treatments
for patients with newly recurrent disease, and this approach could maximize
the chance of a particular combination or single new drug showing clinical
benefit and, eventually, prolonging survival. It is possible that CTC detection
could be used in the design of efficacy trials of different therapeutic
approaches. The efficacy of these treatments could be more easily assessed if
patients were stratified by their prognosis, leading to more tailored treatment
strategies. We believe that the challenge for the next generation of clinical
trials, and the responsibility for both clinical investigators and the pharmaceutical
industry will be to incorporate these concepts into the process of drug
development.

162 Reuben and Cristofanilli
FUTURE USES OF CTCs
The process of sorting cancer cells from other cellular components (e.g., blood
and stromal cells) in clinical samples is fundamentally important for the future of
genomic and proteomic analysis (43–45). Collecting representative tissue from
solid tumor metastases usually requires more invasive procedures that increase
the risk of complications and discomfort. Furthermore, these procedures may not
provide an adequate specimen for detailed analysis and typically cannot be
repeated for dynamic evaluation of the biologic changes during treatments.
Theoretically, CTC detection would allow specific genes [e.g., c-erbB-2, epidermal
growth factor receptor (EGFR), and mammaglobin B (MGB)] or more global
gene expression to be analyzed while using specific targeted treatments for MBC
based on the expression of the CTCs (29,46–48). This information could then be
used to design specific treatments that more appropriately reflect the dynamics and
heterogeneity of MBC.
CONCLUSIONS
The detection of microscopic disease in the peripheral blood of patients with
MBC provides prognostic information. This information will allow appropriate
risk stratification and modification of the current staging system for advanced
disease. Furthermore, the intriguing hypothesis that CTCs are, at least in part,
representative of tumorigenic cancer stem cells may present an additional
challenge for translational research. In fact, the phenotypic evaluation of CTCs is
indicating the possibility of more critical design of targeted therapies in the
setting of MBC. Moreover, with the introduction of this novel concept, it is
expected that investigators and the pharmaceutical industry will derive benefits
from these sorting technologies that will contribute to more tailored treatments
and sophisticated trial designs.
REFERENCES
1. Surveillance Epidemiology and End Results (SEER). National Cancer Institute
Website. Available at: http://seer.cancer.gov/statfacts/html/breast.htm.
2. Folkman J. Tumor angiogenesis: therapeutic implications. N Engl J Med 1971; 285
(21):1182–1186. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=
Retrieve&db=PubMed&dopt=Citation&list_uids=4938153.
3. Clare SE, Sener SF, Wilkens W, et al. Prognostic significance of occult lymph node
metastases in node-negative breast cancer. Ann Surg Oncol 1997; 4(6):447–451.
Available at http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=
PubMed&dopt=Citation&list_uids=9309332.
4. Braun S, Pantel K, Muller P, et al. Cytokeratin-positive cells in the bone marrow
and survival of patients with stage I, II, or III breast cancer. N Engl J Med 2000; 342(8):
525–533. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve
&db=PubMed&dopt=Citation&list_uids=10684910.

Circulating Tumor Cells in Individualizing Breast Cancer Therapy 163
5. Ashworth T. A case of cancer in which cells similar to those in the tumours were
seen in the blood after death. Aust Med J 1869; 14:146.
6. Roberts S, Watne A, McGrath R. Technique and results of isolation of cancer cells
from the circulating blood. Am Med Assoc Arch Surg 1958; 76:334–346.
7. Long L, Jonasson O, Roberts S, et al. Cancer cells in blood: results of a simplified
isolation technique. Arch Surg 1960; 80:910–919. Available at http://www.ncbi.nlm.
nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=
14418112.
8. Dontu G, Al-Hajj M, Abdallah WM, et al. Stem cells in normal breast development
and breast cancer. Cell Prolif 2003; 36(suppl 1):59–72. Available at: http://www
.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_
uids=14521516.
9. Al-Hajj M, Wicha MS, Benito-Hernandez A, et al. Prospective identification of
tumorigenic breast cancer cells. Proc Natl Acad Sci U S A 2003; 100(7):3983–3988.
Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=Pub
Med&dopt=Citation&list_uids=12629218.
10. Al-Hajj M, Becker MW, Wicha M, et al. Therapeutic implications of cancer stem
cells. Curr Opin Genet Dev 2004; 14(1):43–47. Available at: http://www.ncbi.nlm.
nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=
15108804.
11. Cristofanilli M, Budd GT, Ellis MJ, et al. Circulating tumor cells, disease progression,
and survival in metastatic breast cancer. N Engl J Med 2004; 351(8):781–791.
Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=Pub
Med&dopt=Citation&list_uids=15317891.
12. Reed W, Bohler PJ, Sandstad B, et al. Occult metastases in axillary lymph nodes as a
predictor of survival in node-negative breast carcinoma with long-term follow-up.
Breast J 2004; 10(3):174–180. Available at: http://www.ncbi.nlm.nih.gov/entrez/
query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=15125740.
13. An XY, Egami H, Hayashi N, et al. Clinical significance of circulating cancer cells in
peripheral blood detected by reverse transcriptase-polymerase chain reaction in
patients with breast cancer. Tohoku J Exp Med 2001; 193(2):153–162. Available at:
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=
Citation&list_uids=11318030.
14. Bossolasco P, Ricci C, Farina G, et al. Detection of micrometastatic cells in breast cancer
by RT-pCR for the mammaglobin gene. Cancer Detect Prev 2002; 26(1):60–63.
Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=Pub
Med&dopt=Citation&list_uids=12088204.
15. Gross HJ, Verwer B, Houck D, et al. Model study detecting breast cancer cells in
peripheral blood mononuclear cells at frequencies as low as 10(-7). Proc Natl Acad
Sci U S A 1995; 92(2):537–541. Available at: http://www.ncbi.nlm.nih.gov/entrez/
query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=7831325.
16. Hawes D, Neville AM, Cote RJ. Methods used for the detection of occult metastasis.
Biomed Pharmacother 2001; 55(4):229–242. Available at: http://www.ncbi.nlm.nih.
gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=
11393810.
17. Hu XC, Chow LW. Detection of circulating breast cancer cells with multiple-marker
RT-PCR assay. Anticancer Res 2001; 21(1A):421–424. Available at: http://www

164 Reuben and Cristofanilli
.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation
&list_uids=11299772.
18. Martin KJ, Graner E, Li Y, et al. High-sensitivity array analysis of gene expression
for the early detection of disseminated breast tumor cells in peripheral blood. Proc
Natl Acad Sci U S A 2001; 98(5):2646–2651. Available at http://www.ncbi.nlm.nih.
gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=
11226293.
19. Pachmann K, Heiss P, Demel U, et al. Detection and quantification of small numbers
of circulating tumour cells in peripheral blood using laser scanning cytometer (LSC).
Clin Chem Lab Med 2001; 39(9):811–817. Available at http://www.ncbi.nlm.nih.
gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=
11601678.
20. Racila E, Euhus D, Weiss AJ, et al. Detection and characterization of carcinoma
cells in the blood. Proc Natl Acad Sci U S A 1998; 95(8):4589–4594. Available at
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=
Citation&list_uids=9539782.
21. Witzig TE, Bossy B, Kimlinger T, et al. Detection of circulating cytokeratin-positive
cells in the blood of breast cancer patients using immunomagnetic enrichment and digital
microscopy. Clin Cancer Res 2002; 8(5):1085–1091. Available at http://www.ncbi.nlm.
nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=
12006523.
22. Kagan M, Howard D, Bendele T, et al. A sample preparation and analysis system for
identification of circulating tumor cells J Clin Ligand Assay 2002; 25(1):104–110.
23. Terstappen LW, Rao C, Gross S, et al. Peripheral blood tumor cell load reflects the
clinical activity of the disease in patients with carcinoma of the breast. Int J Oncol
2000; 17(3):573–578. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?
cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=10938400.
24. Riethdorf S, Fritsche H, Muller V, et al. Detection of circulating tumor cells in
peripheral blood of patients with metastatic breast cancer: a validation study of the
CellSearch system. Clin Cancer Res 2007; 13(3):920–928. Available at: http://
clincancerres.aacrjournals.org/cgi/content/abstract/13/3/920.
25. Taubert H, Blumke K, Bilkenroth U, et al. Detection of disseminated tumor cells in
peripheral blood of patients with breast cancer: correlation to nodal status and
occurrence of metastases. Gynecol Oncol 2004; 92:256–261.
26. Lankiewicz S, Gutierrez B, Böcher O. Quantitative real-time RT-PCR of disseminated
tumor cells in combination with immunomagnetic cell enrichment. Mol
Biotechnol 2006; 34:15–27.
27. Austrup F, Uciechowski P, Eder C. Prognostic value of genomic alterations in
minimal residual cancer cells purified from the blood of breast cancer patients. Br J
Cancer 2001; 83:1664–1673.
28. Meng S, Tripathy D, Shete S, et al. HER-2 gene amplification can be acquired as
breast cancer progresses. Proc Natl Acad Sci U S A 2004; 101(25):9393–9398.
Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=Pub
Med&dopt=Citation&list_uids=15194824.
29. Meng S, Tripathy D, Shete S, et al. uPAR and HER-2 gene status in individual
breast cancer cells from blood and tissues. Proc Natl Acad Sci U S A 2006; 103
(46):17361–17365. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=
Retrieve&db= PubMed&dopt=Citation&list_uids=17079488.

Circulating Tumor Cells in Individualizing Breast Cancer Therapy 165
30. Cristofanilli M, Hayes DF, Budd GT, et al. Circulating tumor cells: a novel prognostic
factor for newly diagnosed metastatic breast cancer. J Clin Oncol 2005; 23(7):
1420–1430. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=
Retrieve&db=PubMed&dopt=Citation&list_uids=15735118.
31. Cristofanilli M, Broglio KR, Guarneri V, et al. Circulating tumor cells in metastatic
breast cancer: biologic staging beyond tumor burden. Clin Breast Cancer 2007; 7(6):
471–479. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve
&db=PubMed&dopt=Citation&list_uids=17386124.
32. Budd GT, Cristofanilli M, Ellis MJ. Circulating tumor cells versus imaging:
Predicting overall survival in metastatic breast cancer. Clin Cancer Res 2006; 12(21):
6403–6409.
33. Internation Union Against Cancer (UICC). TNM Classification of Malignant Tumours.
1st ed. Geneva, Switzerland: UICC, 1968.
34. Denoix P. Tumor, node, and metastasis (TNM). Bull Inst Nat Hyg (Paris) 1944; 1:1–69.
35. Berndt H, Titze U. TNM clinical stage classification of breast cancer. Int J Cancer
1969; 4(6):837–844. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?
cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=5378609.
36. Greene FL, Page DI, Fleming ID, eds. AJCC Cancer Staging Manual. 6th ed. New
York, NY: Springer-Verlag, 2002.
37. Ellis MJ, Hayes DF, Lippman ME. Treatment of Metastatic Disease. Philadelphia,
PA: Lippincott-Raven, 2000.
38. Swenerton KD, Legha SS, Smith T, et al. Prognostic factors in metastatic breast
cancer treated with combination chemotherapy. Cancer Res 1979; 39(5):1552–1562.
Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=Pub
Med&dopt=Citation&list_uids=427797.
39. Bernard-Marty C, Cardoso F, Piccart MJ. Facts and controversies in systemic
treatment of metastatic breast cancer. Oncologist 2004; 9(6):617–632. Available at:
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=
Citation&list_uids=15561806.
40. Sledge GW, Neuberg D, Bernardo P, et al. Phase III trial of doxorubicin, paclitaxel,
and the combination of doxorubicin and paclitaxel as front-line chemotherapy for
metastatic breast cancer: an intergroup trial (E1193). J Clin Oncol 2003; 21(4):
588–592. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve
&db=PubMed&dopt=Citation&list_uids=12586793.
41. O’Shaughnessy J, Miles D, Vukelja S, et al. Superior survival with capecitabine plus
docetaxel combination therapy in anthracycline-pretreated patients with advanced
breast cancer: phase III trial results. J Clin Oncol 2002; 20(12):2812–2823. Available
at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=
Citation&list_uids=12065558.
42. O’Shaughnessy J, Nag S, Calderillo-Ruiz G. Gemcitabine plus paclitaxel versus
paclitaxel as first-line treatment for anthracycline pre-treated metastatic breast cancer:
interim results of a global phase III study. Proc Am Soc Clin Oncol 2003; 22:7.
43. Hancock W, Apffel A, Chakel J, et al. Integrated genomic/proteomic analysis. Anal
Chem 1999; 71(21):742A–748A. Available at: http://www.ncbi.nlm.nih.gov/entrez/
query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=10565269.
44. Marshall A, Hodgson J. DNA chips: an array of possibilities. Nat Biotechnol 1998; 16(1):
27–31. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&
db=PubMed&dopt=Citation&list_uids=9447589.

166 Reuben and Cristofanilli
45. Page MJ, Amess B, Townsend RR, et al. Proteomic definition of normal human
luminal and myoepithelial breast cells purified from reduction mammoplasties. Proc
Natl Acad Sci U S A 1999; 96(22):12589–12594. Available at: http://www.ncbi.nlm.
nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=
10535966.
46. Smirnov DA, Zweitzig DR, Foulk BW, et al. Global gene expression profiling of
circulating tumor cells. Cancer Res 2005; 65(12):4993–4997. Available at: http://www
.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&
list_ uids=15958538.
47. Speicher MR, Carter NP. The new cytogenetics: blurring the boundaries with
molecular biology. Nat Rev Genet 2005; 6(10):782–792. Available at: http://www
.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&
list_uids=16145555.
48. Reuben JM, Lee BN, Li C. Genomic of circulating tumor cells in metastatic breast
cancer. Proc ASOC 25:32s (abstr 1002).

12
Individualization of Endocrine Therapy
in Breast Cancer
Amelia B. Zelnak and Ruth M. O’Regan
Emory Winship Cancer Institute, Atlanta, Georgia, U.S.A.
Clodia Osipo
Loyola University, Maywood, Illinois, U.S.A.
INTRODUCTION
Discovery of estrogen receptors (ER) was critical for the development of
endocrine therapy in breast cancer. Expression of ERa, thepredominantisoform,
in breast tumors of both premenopausal and postmenopausal women is a
highly predictive marker for response to antiestrogen treatment in women with
ERa-positive breast cancer. Today, endocrine therapies include the use of
antiestrogens such as tamoxifen and aromatase inhibitors such as anastrozole,
exemestane, and letrozole. Tamoxifen, the first antiestrogen to be used for the
treatment of ERa-positive breast cancer, competitively blocks the actions of
17b-estradiol (E 2 ), the female hormone that binds and activates ERa in tumors.
In postmenopausal women, peripheral aromatization of androgens to estrogens
is the major source of plasma estrogen. Aromatase inhibitors inhibit this
reaction and consequently suppress the production of circulating estrogen in
postmenopausal women. Endocrine therapies are effective at reducing recurrence,
increasing overall survival, and reducing contralateral breast cancer up
to 50%. However, about 50% of patients with ERa-positive breast cancer have
167

168 Zelnak et al.
intrinsic resistance to antiestrogen therapy and therefore do not benefit. In
contrast to patients with intrinsically resistant tumors, there are patients who do
initially respond to antiestrogen therapy; however, most of these patients
develop acquired resistance during the treatment regimen. Therefore, the
current goal in breast cancer research is to elucidate the mechanisms of both
intrinsic and acquired resistance to tamoxifen and the aromatase inhibitors in
order to develop new therapeutic strategies to prevent and/or treat resistant
breast cancer.
ANTIESTROGENS
Nonsteroidal antiestrogens were initially developed as contraceptives in the
1960s. Walpole and colleagues synthesized tamoxifen (termed ICI 46, 474), a
potent antiestrogen with antifertility properties in rats. However, in humans,
tamoxifen induced ovulation in subfertile women. Therefore, the development
of tamoxifen as a contraceptive was terminated. However, Walpole also
patented the application of tamoxifen as a drug treatment for hormonedependent
cancers. Thus, clinical trials were started to evaluate tamoxifen
against the standard endocrine treatment at the time, diethylstilbestrol (DES),
for the treatment of advanced breast cancer in postmenopausal women (1).
Tamoxifen not only was as effective as DES for the treatment of advanced
breast cancer but also had fewer side effects. Therefore, the advantage of
tamoxifen over DES was crucial for its subsequent evaluation as a treatment for
all stages of breast cancer.
In 1962, Jensen and colleagues discovered ERa. Jensen demonstrated
that E 2 , the circulating female hormone that promoted breast cancer growth,
binds to diverse tissue sites around a woman’s body but is retained in estrogentarget
tissues, for example, the uterus and vagina (2). The identification of
ERa as the target of E 2 action in the breast and antiestrogens blocking the
binding of E 2 to ERa provided a therapeutic target and an approach for the
treatment of breast cancer. Although many antiestrogens were discovered and
tested during the 1960s and 1970s, only tamoxifen was considered safe enough
for extensive clinical evaluation (3). Clinical trials ultimately demonstrated
that patients with ERa-positive breast cancer benefited the most from
tamoxifen therapy, whereas women with ERa-negative breast cancer were
found to be unaffected. During subsequent clinical trials, five years of adjuvant
tamoxifen treatment was found to be more effective than less than five
years of treatment in improving time to tumor recurrence and overall survival
(4). In contrast to the beneficial effects of tamoxifen as a treatment for breast
cancer, both laboratory and clinical results showed that tamoxifen increased
the risk of endometrial cancer in the uterus fourfold in postmenopausal women
compared with untreated women (5,6). These results strongly indicated that
tamoxifen was not a pure antiestrogen but had selective functions depending
on the target tissue.

Individualization of Endocrine Therapy in Breast Cancer 169
Resistance to Tamoxifen
Overview
Five years of adjuvant tamoxifen therapy is a standard of care for early-stage ERapositive
breast cancer. Although there is increasing use of aromatase inhibitors in
postmenopausal women, tamoxifen is the only approved effective therapy in premenopausal
women. Approximately 50% of patients with ERa-positive breast
cancer benefit from tamoxifen treatment in the advanced- and early-stage settings
(4,7). Unfortunately, the other 50% of patients do not respond, and the ones that do
respond initially eventually progress. Thus, the problem with tamoxifen therapy is
that breast tumors can either be resistant (intrinsic or de novo) prior to endocrine
treatment or become resistant (acquired) during therapy. Numerous studies
have looked at the mechanisms responsible for this resistant phenotype. They
include the loss of the ERa gene and/or protein in tumors during short-term and
long-term treatment, identification of mutations within the ERa gene, changes in
the pharmacologic response of resistant breast cancer cells to tamoxifen, activation
of ERa-mediated gene transcription in the absence of ligand, modifications in gene
transcription by changes in interactions between ERa and coregulators, and the
development of enhanced surface-to-intracellular signaling cross-talk between
growth factor receptors and ERa (Fig. 1).
Figure 1
Mechanisms of hormone resistance.

170 Zelnak et al.
ER-Negative Breast Cancer and Resistance
Breast cancers initially lacking expression of ERa have intrinsic resistance to
endocrine therapy. In addition, loss of ERa expression upon recurrence of breast
cancer during tamoxifen therapy has been reported in as many as 25% of tumors
(8–10). However, most of the acquired resistance during adjuvant tamoxifen
treatment in ERa-positive tumors is not due to changes in ERa expression.
Interestingly, most tamoxifen-resistant breast tumors retain ERa and respond to
secondary endocrine treatments. Overall, a loss of ERa status does not seem to
be the major mechanism for the development of acquired antiestrogen resistance.
ER Mutations
Several mutant forms of both the ERa and ERb, another ER isoform, have been
identified and previously reviewed (11–13). However, the functional significance
of these mutants as a mechanism of antiestrogen resistance is yet to be
elucidated. Most of the data demonstrate the existence of splice variants for ERa
and ERb mRNAs without evidence of translation into functional proteins. Most
of the tumors express both mutant and wild-type ER with the wild type being the
predominant species. A single-point mutation in the ligand-binding domain
(LBD) of ERa has been reported (D351Y) that converts tamoxifen and ralox
ifene from antiestrogen to estrogens in in vivo and in vitro models of antiestrogenstimulated
MCF-7 tumor cells (14,15). However, the D351Y ERa mutant has yet
to be detected in either intrinsic or acquired resistant breast tumors from patients.
In addition, mutations in the F-region of the ER have been shown to affect
the activities of both E 2 and 4-hydroxytamoxifen (4OHT), the active metabolite
of tamoxifen (16). Therefore, the clinical relevance of ER mutants is unclear to
date as a mechanism of resistance to antiestrogens. However, our understanding
of ER mutations in response to antiestrogens might lead to the development of
better antiestrogens or other drugs for breast cancer treatment.
Role of Pharmacogenetics
Tamoxifen undergoes extensive primary and secondary metabolism. The secondary
metabolite 4-hydroxy-N-desmethyl tamoxifen (endoxifen) is formed by
the CYP2D6-mediated oxidation of N-desmethyl tamoxifen and has been shown
to play an important role in the anticancer effect of tamoxifen. CYP2D6 genotyping
has revealed that women can be classified as either poor, intermediate, or
extensive metabolizers of tamoxifen. Women who are poor metabolizers of
tamoxifen, and, therefore, have lower levels of endoxifen, have been shown to
have worse disease-free survival. No difference in overall survival has been
shown (17). Women who are extensive metabolizers of tamoxifen have also been
shown to experience more hot flashes. A recent study showed that women who
experience hot flashes while on tamoxifen had a hazard ratio of recurrence of

Individualization of Endocrine Therapy in Breast Cancer 171
0.51 compared with women without hot flashes (18). This study suggests that
vasomotor symptoms and efficacy of tamoxifen may be related to a women’s
pharmacogenetics.
Role of Coregulators in Resistance
The recruitment of coactivators or corepressors to the ER determines the switch
between ER activation and repression. In addition, the coordinated action of
ligand, ER, and coregulators determines which genes are transcribed or repressed
depending on the cellular context and thus which cells will or will not proliferate.
The overall data indicate that intricate modulation of the ER-to-coregulator ratio
in breast cancer cells could determine resistance to antiestrogens. For example,
the coactivator SRC-1 activates ER in a ligand-independent manner while
increasing 4OHT’s agonist activity (19). On the other hand, the corepressor
SMRT blocks the agonist activity of 4OHT-induced by SRC-1 (20). The other
member of the p160 coactivator family is amplified in breast cancer (AIB1,
SRC-3, and RAC3). AIB1 mRNA was found to be amplified in 60% of breast
cancers, and the protein product was found to be overexpressed in about 10% of
tumors. Recently, an association was discovered between high AIB1 and human
epidermal growth factor receptor 2 (HER2/neu), a member of the epidermal
growth factor receptor (EGFR) family of receptor tyrosine kinases and tamoxifenresistant
breast cancers (20). The potential mechanism for this resistance requires
activation of the mitogen-activated protein kinase (MAPK) signaling cascade by
HER2/neu, which in turn potentially phosphorylates and activates AIB1. Activated
AIB1 in turn can convert 4OHT from an antiestrogen to an estrogen in
regards to ERa activity. In contrast, low levels of N-CoR mRNA, a corepressor,
have been implicated with resistance to tamoxifen (21). Therefore, the identification
of abnormally expressed coregulators will probably assist in the design
of future therapies that improve the efficacy of endocrine treatment without the
development of resistance.
EGFR and HER2/Neu and Antiestrogen Resistance
Numerous laboratory and clinical studies indicate that overexpression and/or
aberrant activity of the HER2/neu (erbB2) signaling pathway is associated with
antiestrogen resistance in breast cancer (22). The HER2/neu receptor is a
member of the EGFR family of receptor tyrosine kinases, which include HER3
(erbB3) and HER4 (erbB4) (23–26). Upon dimerization, the tyrosine kinase
domains located within the COOH-terminal regions of receptors are activated by
an autophosphorylation cascade on specific tyrosine residues, which activate
downstream effectors, such as MAPK and Akt, which promote cellular proliferation,
survival, anti-apoptosis, and transformation. HER2/neu is overexpressed
and/or amplified in 25% to 30% of breast tumors and is associated with a more
aggressive phenotype and poor prognosis (27). Patients with breast tumors
overexpressing HER2/neu exhibit much lower response rates to antiestrogen

172 Zelnak et al.
therapy (27). Thus, it is suggested that one possible mechanism of resistance to
tamoxifen is overexpression of HER2/neu in ERa-positive breast cancers. In
addition, overexpression of EGFR and its ligands are observed in several human
cancers including breast cancer (28–30). The increased expression of EGFR is
frequently associated with tumor progression and resistance to antiestrogens.
These data suggest that EGFR and/or HER2/neu are possible targets for preventing
or treating antiestrogen resistance in breast cancer. Strategies to target
EGFR and HER2/neu include the use of humanized monoclonal antibodies to
the receptors (31), tyrosine kinase inhibitors that block reduction of adenosine
triphosphate (ATP) to adenosine diphospohate (ADP) þ Pi, and receptor antisense
molecules (32). Gefitinib, (ZD 1839, Iressa 1 ), an EGFR-specific tyrosine kinase
inhibitor, has been shown to inhibit growth of breast cancer cell lines in vitro that
are resistant to tamoxifen (33). More importantly, the combination of gefitinib
and tamoxifen was shown to prevent resistance (34), demonstrating that (1)
EGFR might be a key player in the development of antiestrogen resistance and
(2) inhibiting the activity of this receptor might be therapeutically beneficial in
preventing resistance to antiestrogens. In addition to gefitinib, trastuzumab
(Herceptin 1 ) is a humanized monoclonal antibody directed against the ectodomain
of the HER2/neu receptor. It has been shown to restore breast cancer cell
sensitivity to tamoxifen in HER2/neu-overexpressing cells (35).
Role of MAPK
MAPK is a serine/threonine kinase that is activated by phosphorylation cascades
originating from GTP-bound Ras, a downstream effector of EGFR and HER2/
neu. MAPK has been shown to be hyperactivated in MCF-7 breast cancer cells
overexpressing either EGFR or HER2/neu (36–38). This increased activity of
MAPK promoted an enhanced association of ERa with coactivators and
decreased interaction with corepressors, thereby enhancing hormone-dependent
gene transcription and possibly leading to tamoxifen-resistant breast cancer. The
role of the MAPK signaling pathway in tamoxifen resistance has been demonstrated
primarily in vitro using specific inhibitors such as U0126 (MAPKK,
MEK1/2 inhibitor), AG1478 (EGFR and HER2/neu inhibitors), and PD98059
(MAPK, ERK1/2 inhibitor) (36–38). The exact mechanism by which hyperactivated
MAPK converts the antiestrogenic activity of the tamoxifen-ERa
complex to more of an estrogenic one in resistant breast cancer cell is yet
unclear. However, it has been demonstrated that activation of the Ras/MAPK
pathway by EGFR/HER2/neu receptor activation could phosphorylate Ser-118 in
the AF-1 domain of ERa, thus promoting ligand-independent transcription of
ERa and possibly loss of tamoxifen-induced inhibition of ERa-mediated gene
transcription. On the basis of these preclinical models, several clinical trials are
under way to determine whether treating patients with EGFR inhibitors (i.e.,
Iressa) and/or MAPK-specific inhibitors (U0126) can effectively treat and/or
prevent antiestrogen resistance in breast tumors.

Individualization of Endocrine Therapy in Breast Cancer 173
Role of PI3-K/AKT Signaling
Phosphatidylinositol 3-kinase (PI3-K) is a heterodimer complex consisting of
a regulatory subunit, p85, and a catalytic subunit, p110. The regulatory subunit
p85 is phosphorylated by either receptor tyrosine kinases such as EGFR/
HER2/neu or intracellular adapter kinases such as Shc or Src and thereafter
interacts and activates the catalytic subunit p110 (39). The activated PI3-K
subsequently activates downstream effector kinases such as Akt (PKB), a
serine/threonine kinase. Akt has been shown to activate either directly or
indirectly proteins responsible for prosurvival and anti-apoptosis in cancer
cells. Several reports indicate that the PI3-K/Akt pathway interacts with the
ERa pathway, resulting in bidirectional cross talk. It has been shown that PI3-
K/Akt can mediate E 2 -induced transcription of cyclin D1 and entry of MCF-7
breast cancer cells into S-phase (40). In addition, the PI3-K/Akt pathway can
induce ERa phosphorylation on Ser-167 to promote ERa-mediated transcription
and thus protect cells against tamoxifen-induced apoptosis (41). The
accumulating data suggest that ERa-positive breast tumors with alterations of
PI3-K and Akt signaling, whether dependent or independent of EGFR/HER2/
neu, might be insensitive to antiestrogens and thus lead to resistance. Future
studies are needed to confirm whether blocking PI3-K and/or Akt signaling in
ERa-positive breast cancer will be beneficial in subverting resistance to
antiestrogens.
New SERMS
New drug discovery for selective ER modulators (SERMs) is currently driven
by the known side effects of tamoxifen, which include an increase in the
incidence of endometrial cancer, development of resistance, and the recent
report of the negative effects of hormone replacement therapy (HRT) on coronary
heart disease (42). Several approaches are being pursued by altering the
antiestrogenic side chain or improving the pharmacokinetics of existing molecules.
Several novel antiestrogen compounds have been developed to replace
tamoxifen for ERa-positive breast cancer without the agonist actions. These
compounds have the potential to be more effective than tamoxifen, having
the advantage of reducing the incidence of endometrial cancer and possibly
acquired resistance during therapy. Two groups of antiestrogenic drugs exist
today with the potential to replace tamoxifen: (1) The SERMs that include
tamoxifen-like compounds (triphenylethylenes) such as toremifene, idoxifene,
and GW 5638 and fixed-ring compounds (benzothiophenes) such as raloxifene,
arzoxifene, EM 652, and CP 336,156, and (2) selective ER downregulators
(SERDs and pure antiestrogen) such as ICI 182,780 (Fulvestrant 1 ). In phase II
trials in patients with tamoxifen-resistant metastatic breast cancer, the new
SERMs showed low response rates (0–15%) (43,44), suggesting crossresistance
to tamoxifen. No difference in outcome was noted between
tamoxifen and toremifene in patients with metastatic or early-stage breast

174 Zelnak et al.
cancer (45,46). In contrast, few clinical trials exist today evaluating the
effectiveness of the fixed-ring compounds to tamoxifen. The Study of
Tamoxifen and Raloxifene (STAR) trial compared the efficacy of raloxifene
over tamoxifen as a chemopreventive for women at high risk for breast cancer.
Both drugs reduced the risk of invasive breast cancer by approximately 50%.
Tamoxifen also lowered the incidence of lobular carcinoma in situ (LCIS) and
ductal carcinoma in situ (DCIS) by about 50 percent; however, raloxifene did
not have an effect on the incidence of LCIS and DCIS (47). Of note, the
incidence of uterine cancer was 36% lower, and the incidence of thromboembolic
events was 29% lower in the raloxifene arm (47). Fulvestrant and
other SERDs have been shown to have high affinity for ERs compared with
tamoxifen with none of the agonist activities. In clinical trials, fulvestrant
initially showed significant promise for the treatment of advanced breast
cancer (48,49). However, in a phase III clinical trial versus tamoxifen for firstline
therapy of advanced breast cancer, fulvestrant was not superior to
tamoxifen in terms of time-to-treatment failure (50). Therefore, more work is
needed to evaluate the efficacy of SERDs in comparison with SERMs for ERapositive
breast cancer.
AROMATASE INHIBITORS
An alternate strategy to endocrine therapy, which specifically inhibits binding of
E 2 to the ER, is to inhibit the production of E 2 by blocking the cytochrome p450
aromatase enzyme, the rate-limiting enzyme that converts androgens (i.e., testosterone
and androstenedione) to estrogens (i.e., E 2 and estrone) in the adrenal
gland, surrounding stroma, and adipose tissue of the breast tumor. The main
drugs of this type are aromatase inhibitors, which include Type I (steroidal) or
Type II (nonsteroidal) (Fig. 2). Steroidal inhibitors are competitive-substrate
mimics of androstenedione. These include formestane and exemestane, which
are irreversible inhibitors that bind with high affinity to the binding site of
aromatase and are converted to a covalently bound intermediate. Nonsteroidal
inhibitors include the first-generation aromatase inhibitor aminoglutethimide and
the second-generation compounds anastrozole and letrozole. All nonsteroidal
aromatase inhibitors act by binding reversibly to the enzyme and competitively
inhibiting binding of the substrate androstenedione. The benefits of using aromatase
inhibitors over tamoxifen are believed to be the complete deprivation of
E 2 and thus better efficacy for ERa-positive breast cancer (51). Recent clinical
data have clearly demonstrated that anastrozole (52), letrozole (53), and
exemestane (54) are more effective than tamoxifen as first-line treatments in
patients with metastatic breast cancer. On the basis of clinical results (52,53),
currently both anastrozole and letrozole are approved by the Food and Drug
Administration for first-line treatment of postmenopausal, ERa-positive
advanced breast cancer. The data from advanced breast cancer trials provided the
rationale to perform large-scale clinical trials to determine whether there is an

Individualization of Endocrine Therapy in Breast Cancer 175
Figure 2
Structures of selective aromatase inhibitors.
advantage of using aromatase inhibitors over tamoxifen in the adjuvant setting.
The anastrozole, tamoxifen, and the combination of anastrozole and tamoxifen
(ATAC) trial has sufficient follow-up data to confirm that anastrozole is superior
to tamoxifen as a first-line adjuvant therapy in ERa-positive breast cancer with
regard to disease-free survival and incidence of contralateral breast cancer (55).
The Breast International Group (BIG) 1-98 study demonstrated a similar
improvement in event-free survival to confirm that letrozole is superior to
tamoxifen as first-line adjuvant hormonal therapy (56). Other trials have demonstrated
an improved outcome for postmenopausal patients with early-stage
breast cancer treated with two to three years of tamoxifen, followed by an
aromatase inhibitor, compared with five years of tamoxifen (57,58). The BIG 1-98
trial will help to determine whether these sequencing/switching approaches are
equivalent to upfront aromatase inhibitors.
Mechanisms of Resistance
Clinically, patients that relapse after a previous response to tamoxifen usually
have a clinical response to aromatase inhibitors (59,60). These results strongly
indicate that the ERa continues to be expressed and is functional in breast tumors
that are resistant to antiestrogens. However, although estrogen deprivation
treatment might be more effective than tamoxifen in delaying resistance, eventually
resistance to aromatase inhibitors will also develop. To date, it is unclear

176 Zelnak et al.
whether similar mechanisms of actions that have been identified for tamoxifen
resistance are also involved in resistance to aromatase inhibitors. The exact
mechanisms contributing to aromatase inhibitor resistance has yet to be fully
elucidiated. However, in vitro studies have identified mutations within the aromatase
gene that confers resistance to aromatase inhibitors (61). These mutations
have not yet been identified in human breast carcinomas (62). Other studies have
demonstrated that estrogen deprivation supersensitizes the breast cancer cell to
low levels of estrogen, thus creating a hypersensitive environment to overcome
estrogen deprivation resulting in resistance (63–65). In addition, results suggest
that there is increased cross talk between growth factor receptor signaling
pathways and ERa. ERa has been shown to become activated and supersensitized
by several different intracellular kinases, including MAPKs, insulin-like growth
factors, and the PI3-K/Akt pathway (41,66–69). Therefore, the data suggest that
ERa continues to be an integral part of the breast cancer cell signaling pathway
even after resistance to aromatase inhibitors has developed.
Role of Progesterone Receptor and HER2/Neu
There is emerging evidence to suggest that ER-positive cancers that do not
express the progesterone receptor (PR) and/or HER2/neu are somewhat intrinsically
resistant to tamoxifen and perhaps hormonal therapy in general. Arpino et
al. have demonstrated an increased relapse rate in patients with ER-positive, PRnegative
cancers compared with ER-positive, PR-positive cancers, treated with
tamoxifen (70). Patients treated with tamoxifen with ER-positive, HER2-positive
metastatic breast cancers have a shorter time to treatment failure compared with
ER-positive, HER2-negative cancers (71). In fact, Arpino et al. have demonstrated
an increase in both HER1 and HER2 in ER-positive, PR-negative cancers
compared to ER-positive, PR-positive cancers, suggesting an interplay between
the ER and epidermal growth factor pathways (70).
Two very small trials demonstrated a significantly increased clinical
response rate in patients with ER-positive, HER2-positive cancers treated with
preoperative aromatase inhibitors compared to preoperative tamoxifen (72,73).
This led to a widely accepted hypothesis that aromatase inhibitors were a better
choice than tamoxifen in patients with ER-positive, HER2-positive cancers.
However, an analysis of the BIG-1-98 trial demonstrates that letrozole improves
outcome compared to tamoxifen in both ER-positive, HER2-positive cancers
(HR 0.68) and in ER-positive, HER2-negative cancers (HR 0.72) (74). A recent
subanalysis of the ATAC trial demonstrated a significantly improved outcome in
ER-positive, HER2-negative cancers (HR 0.66) but not in ER-positive, HER2-
positive cancers (HR 0.92), but this may have been due to the small number of
patients in the HER-positive group (75). Are HER2-positive cancers somewhat
resistant to not just tamoxifen but also to aromatase inhibitors? As outlined
above, a recent trial randomized patients with HR-positive, HER2-positive
metastatic breast cancers to anastrozole alone or to anastrozole plus trastuzumab

Individualization of Endocrine Therapy in Breast Cancer 177
(76). Although there was no significant difference in overall survival, possibly
because patients randomized to anastrozole alone could receive trastuzumab at
disease progression, the time to progression was doubled from 2.4 months in the
anastrozole-alone arm to 4.8 months in the combined arm (p ¼ 0.0016) (76). The
clinical benefit rate in the combination arm was 42%—significantly higher than
in the anastrozole alone arm. A trial that evaluated single-agent trastuzumab as
first-line therapy for patients with HER2-positive cancers demonstrated a clinical
benefit rate of 48% (77). This suggests the intriguing possibility that HRpositive,
HER2-positive cancers are driven by the HER2 pathway, which renders
the cancers partly resistant to hormonal therapies.
An initial evaluation of the ATAC trial using case report forms revealed
that TTR was longer for anastrozole in both ER-positive/PR-positive and ERpositive/PR-negative
subgroups, but the benefit was more pronounced in the ERpositive/PR-negative
subgroup [HR 0.84, 95% confidence interval (CI) 0.69–
1.02 vs. 0.43, 95% CI 0.31–0.61] (78). Importantly, the ER and PR analyses were
not performed centrally. More recently, a central analysis of about 2000 patients
on the ATAC trial demonstrated similar improvements with the use of anastrozole
compared to tamoxifen, regardless of PR status (HR anastrozole vs.
tamoxifen 0.72 for ER-positive, PR-positive subgroups and 0.66 for ER-positive,
PR-negative subgroups) (75). In the BIG-1-98 trial, similar benefits for letrozole
compared to tamoxifen were seen in the ER-positive/PR-positive and ERpositive/PR-negative
subgroups (74).
On the basis of this data, decisions regarding whether to start a patient on
tamoxifen or an aromatase inhibtor should not be based on PR or HER2 status.
Further molecular profiling may help in the future in making decisions regarding
optimal hormonal therapies.
Role of Gene Expression Analysis
Recent studies have shown that that gene expression profiling of breast tumors
can be used to predict outcome. The initial gene expression signatures were
developed using snap-frozen tissue, limiting their clinical application (79). An
assay (Oncotype DX, Genomic Health, Inc., Redwood City, California, U.S.) has
been developed and validated in women with early-stage, node-negative breast
cancer using fixed, paraffin-embedded tissue. The 21-gene panel includes genes
involved with proliferation, invasion, and hormone response and classifies a
tumor as being low, intermediate, or high risk of recurrence based on the level of
expression (80). Additional studies have shown that a woman’s recurrence score
also predicts response to chemotherapy: women with low recurrence scores
derive little benefit from adjuvant chemotherapy, whereas women with high
recurrence scores benefit significantly from chemotherapy in addition to hormonal
therapy. Most strikingly are the findings that breast cancers with a high
recurrence score obtain minimal, if any, benefit from five years of tamoxifen
(81). Given the fact that these cancers likely have higher expression of HER2

178 Zelnak et al.
and lower quantitative hormone receptors, this is in keeping with the data outlined
above, suggesting that HER2-positive cancers have intrinsic resistance to
tamoxifen. It is not known if women with intermediate recurrence scores benefit
from chemotherapy in addition to hormonal therapy (82). The TAILORx (Trial
Assigning Individualized Options for Treatment) trial is under way, randomizing
early-stage breast cancer patients with intermediate recurrence scores to chemotherapy
plus hormonal therapy versus hormonal therapy alone.
FUTURE RESEARCH AND THERAPEUTICS
It is clear that the central regulator of growth for ERa-positive breast tumors is
ERa. ER signaling is evidently not isolated from other cellular signaling pathways.
Research studies demonstrate that there is bidirectional cross talk between
EGFR/HER2/neu, MAPK, PI3-K/Akt, and ERa, which ultimately affects the
cellular response of a cell to estrogens and mitogens. ERa integrates numerous
signals from hormones, growth factors, and intracellular kinases along with the
array of coregulators to modulate cellular physiology and tumor pathology. The
first generation antiestrogen/SERM, tamoxifen, has been effective at increasing
overall survival of patients with ERa-positive breast cancer. However, the
eventual development of resistance to tamoxifen is common because of the
multiple signaling networks affecting the function of ERa. As a result of this
limitation, intense research over the past 25 years has revealed the need for
alternate treatment strategies to tamoxifen such as newer and better SERMs with
less agonist activity in the uterus while being full antagonists in the breast.
In addition to SERMs, other therapeutic approaches have emerged such as
the use of aromatase inhibitors to prevent the synthesis of E 2 and thus deprive
breast cancer cells and other cells of E 2 . The strategy of estrogen deprivation
may also have consequences such as resistance, bone loss, and/or dementia. In
the ATAC and BIG 1-98 adjuvant trials, changes in bone mineral density (BMD)
and increased incidence of bone fractures were reported. Longer follow-up in
these studies will help determine if the decrease in BMD stabilizes over time or
if it continues throughout the course of treatment.
Therapeutic opportunities also exist from research studies investigating the
role of growth factor receptors EGFR and HER2/neu and their intracellular
signaling communicators MAPK and PI3-K/Akt in the development of tamoxifen
resistance. Trastuzumab, the HER2/neu-specific antibody, has been shown
to be highly effective in treating breast tumors overexpressing HER2/neu.
Moreover, research data indicate a role for HER2/neu in antiestrogen resistance,
demonstrating a rationale for using trastuzumab to treat or prevent antiestrogenresistant
breast cancer. EGFR inhibitors and other tyrosine kinase inhibitors may
also prove to be beneficial in preventing resistance when given in combination
with tamoxifen or other SERMs or SERDs such as fulvestrant.
In addition to EGFR and HER2/neu as important modulators of tamoxifen
action, dissecting the specific interrelationship between ERa and its coregulators

Individualization of Endocrine Therapy in Breast Cancer 179
may open the door for novel therapeutics. An understanding of how ERamediated
gene transcription is dysregulated in breast cancer and how this
changes in resistance during endocrine therapy will hopefully lead to improved
strategies for the use of current and future antiestrogen therapies.
REFERENCES
1. Cole MP, Jones CT, Todd ID. A new anti-oestrogenic agent in late breast cancer. An
early clinical appraisal of ICI46474. Br J Cancer 1971; 25:270–275.
2. Jensen EV, Jacobson HI. Basic guides to the mechanism of estrogen action. Recent
Prog Horm Res 1962; 18:387–414.
3. Jordan VC, Dowse LJ. Tamoxifen as an anti-tumour agent: effect on oestrogen
binding. J Endocrinol 1976; 68:297–303.
4. Fisher B, Redmond C, Brown A, et al. Adjuvant chemotherapy with and without
tamoxifen in the treatment of primary breast cancer: 5-year results from the National
Surgical Adjuvant Breast and Bowel Project Trial. J Clin Oncol 1986; 4:459–471.
5. Fisher B, Costantino JP, Redmond CK, et al. Endometrial cancer in tamoxifentreated
breast cancer patients: findings from the National Surgical Adjuvant Breast
and Bowel Project (NSABP) B-14. J Natl Cancer Inst 1994; 86:527–537.
6. Gottardis MM, Robinson SP, Satyaswaroop PG, et al. Contrasting actions of
tamoxifen on endometrial and breast tumor growth in the athymic mouse. Cancer
Res 1988; 48:812–815.
7. Muss HB, Case LD, Atkins JN, et al. Tamoxifen versus high-dose oral medroxyprogesterone
acetate as initial endocrine therapy for patients with metastatic breast
cancer: a Piedmont Oncology Association study. J Clin Oncol 1994; 12:1630–1638.
8. Kuukasjarvi T, Kononen J, Helin H, et al. Loss of estrogen receptor in recurrent
breast cancer is associated with poor response to endocrine therapy. J Clin Oncol
1996; 14:2584–2589.
9. Bachleitner-Hofmann T, Pichler-Gebhard B, Rudas M, et al. Pattern of hormone
receptor status of secondary contralateral breast cancers in patients receiving adjuvant
tamoxifen. Clin Cancer Res 2002; 8:3427–3432.
10. Johnston SR, Saccani-Jotti G, Smith IE, et al. Changes in estrogen receptor, progesterone
receptor, and pS2 expression in tamoxifen-resistant human breast cancer.
Cancer Res 1995; 55:3331–3338.
11. Hopp TA, Fuqua SA. Estrogen receptor variants. J Mammary Gland Biol Neoplasia
1998; 3:73–83.
12. Murphy LC, Dotzlaw H, Leygue E, et al. The pathophysiological role of estrogen
receptor variants in human breast cancer. J Steroid Biochem Mol Biol 1998; 65:175–180.
13. Fuqua SA. The role of estrogen receptors in breast cancer metastasis. J Mammary
Gland Biol Neoplasia 2001; 6:407–417.
14. Jiang SY, Langan-Fahey SM, Stella AL, et al. Point mutation of estrogen receptor
(ER) in the ligand-binding domain changes the pharmacology of antiestrogens in ERnegative
breast cancer cells stably expressing complementary DNAs for ER. Mol
Endocrinol 1992; 6:2167–2174.
15. Liu H, Park WC, Bentrem DJ, et al. Structure-function relationships of the raloxifeneestrogen
receptor-alpha complex for regulating transforming growth factor-alpha
expression in breast cancer cells. J Biol Chem 2002; 277:9189–9198.

180 Zelnak et al.
16. Schwartz JA, Zhong L, Deighton-Collins S, et al. Mutations targeted to a predicted helix
in the extreme carboxyl-terminal region of the human estrogen receptor-alpha alter its
response to estradiol and 4-hydroxytamoxifen. J Biol Chem 2002; 277:13202–13209.
17. Goetz MP, Rae JM, Suman VJ, et al. Pharmacogenetics of tamoxifen biotransformation
is associated with clinical outcomes of efficacy and hot flashes.
J. Clin Onc 2005; 23:9312–8.
18. Mortimer J, Flatt S, Parker B, et al. Tamoxifen, hot flashes, and breast cancer
recurrence: support for pharmacogenetics. Proc ASCO 2007 (abstr 500).
19. Smith CL, Nawaz Z, and O’Malley BW. Coactivator and corepressor regulation of
the agonist/antagonist activity of the mixed antiestrogen, 4-hydroxytamoxifen. Mol
Endocrinol 1997; 11:657–666.
20. Osborne CK, Bardou V, Hopp TA, et al. Role of the estrogen receptor coactivator
AIB1 (SRC-3) and HER-2/neu in tamoxifen resistance in breast cancer. J Natl
Cancer Inst 2003; 95:353–361.
21. Girault I, Tozlu S, Lidereau R, et al. Expression analysis of DNA methyltransferases
1, 3A, and 3B in sporadic breast carcinomas. Clin Cancer Res 2003; 9:4415–4422.
22. Kurokawa H, Arteaga CL. ErbB (HER) receptors can abrogate antiestrogen action in
human breast cancer by multiple signaling mechanisms. Clin Cancer Res 2003;
9:511S–515S.
23. Yarden Y, Sliwkowski MX. Untangling the ErbB signalling network. Nat Rev Mol
Cell Biol 2001; 2:127–137.
24. Yarden Y. The EGFR family and its ligands in human cancer. signalling mechanisms
and therapeutic opportunities. Eur J Cancer 2001; 37(suppl 4):S3–S8.
25. Yarden Y. Biology of HER2 and its importance in breast cancer. Oncology 2001;
61(suppl 2):1–13.
26. Rubin I, Yarden Y. The basic biology of HER2. Ann Oncol 2001; 12(suppl 1):S3–S8.
27. Spigel DR, Burstein HJ. HER2 overexpressing metastatic breast cancer. Curr Treat
Options Oncol 2002; 3:163–174.
28. Klijn JG, Berns PM, Schmitz PI, et al. The clinical significance of epidermal growth
factor receptor (EGF-R) in human breast cancer: a review on 5232 patients. Endocr
Rev 1992; 13:3–17.
29. Klijn JG, Look MP, Portengen H, et al. The prognostic value of epidermal growth
factor receptor (EGF-R) in primary breast cancer: results of a 10 year follow-up
study. Breast Cancer Res Treat 1994; 29:73–83.
30. Tang CK, Gong XQ, Moscatello DK, et al. Epidermal growth factor receptor vIII
enhances tumorigenicity in human breast cancer. Cancer Res 2000; 60:3081–3087.
31. Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a monoclonal
antibody against HER2 for metastatic breast cancer that overexpresses HER2.
N Engl J Med 2001; 344:783–792.
32. Nahta R, Hortobagyi GN, Esteva FJ. Growth factor receptors in breast cancer:
potential for therapeutic intervention. Oncologist 2003; 8:5–17.
33. Morris C. The role of EGFR-directed therapy in the treatment of breast cancer.
Breast Cancer Res Treat 2002; 75(suppl 1):S51–S55, S57–S59 (discussion).
34. Johnston SR, Head J, Pancholi S, et al. Integration of signal transduction inhibitors
with endocrine therapy: an approach to overcoming hormone resistance in breast
cancer. Clin Cancer Res 2003; 9:524S–532S.
35. Kurokawa H, Lenferink AE, Simpson JF, et al. Inhibition of HER2/neu (erbB-2)
and mitogen-activated protein kinases enhances tamoxifen action against

Individualization of Endocrine Therapy in Breast Cancer 181
HER2-overexpressing, tamoxifen-resistant breast cancer cells. Cancer Res 2000; 60:
5887–5894.
36. Oh AS, Lorant LA, Holloway JN, et al. Hyperactivation of MAPK induces loss of
ERalpha expression in breast cancer cells. Mol Endocrinol 2001; 15:1344–1359.
37. Kato S, Endoh H, Masuhiro Y, et al. Activation of the estrogen receptor through
phosphorylation by mitogen-activated protein kinase. Science 1995; 270:1491–1494.
38. Bunone G, Briand PA, Miksicek RJ, et al. Activation of the unliganded estrogen
receptor by EGF involves the MAP kinase pathway and direct phosphorylation.
Embo J 1996; 15:2174–2183.
39. Fry MJ. Phosphoinositide 3-kinase signalling in breast cancer: how big a role might
it play? Breast Cancer Res 2001; 3:304–312.
40. Castoria G, Migliaccio A, Bilancio A, et al. PI3-kinase in concert with Src promotes the
S-phase entry of oestradiol-stimulated MCF-7 cells. Embo J 2001; 20: 6050–6059.
41. Campbell RA, Bhat-Nakshatri P, Patel NM, et al. Phosphatidylinositol 3-kinase/
AKT-mediated activation of estrogen receptor alpha: a new model for anti-estrogen
resistance. J Biol Chem 2001; 276:9817–9824.
42. Rossouw JE, Anderson GL, Prentice RL, et al. Risks and benefits of estrogen plus
progestin in healthy postmenopausal women: principal results From the Women’s
Health Initiative randomized controlled trial. JAMA 2002; 288:321–333.
43. Vogel CL, Shemano I, Schoenfelder J, et al. Multicenter phase II efficacy trial of
toremifene in tamoxifen-refractory patients with advanced breast cancer. J Clin
Oncol 1993; 11:345–350.
44. Johnston SR, Gumbrell LA, Evans TR, et al. A cancer research (UK) randomized
phase II study of idoxifene in patients with locally advanced/metastatic breast cancer
resistant to tamoxifen. Cancer Chemother Pharmacol 2004; 53:341–348.
45. Pyrhonen S, Ellmen J, Vuorinen J, et al. Meta-analysis of trials comparing toremifene
with tamoxifen and factors predicting outcome of antiestrogen therapy in postmenopausal
women with breast cancer. Breast Cancer Res Treat 1999; 56: 133–143.
46. Pagani O, Gelber S, Price K, et al. Toremifene and tamoxifen are equally effective
for early stage breast cancer: first results of International Breast Cancer Study Group.
Ann Oncol 2004; 15:1749–1759.
47. Wickerham DL, Costantino JP, Vogel V, et al. The study of tamoxifen and raloxifene
(STAR): initial findings from the NSABP P-2 breast cancer prevention study. ASCO
Annual Meeting Proceedings Part I. J Clin Oncol 2006; 24(suppl 18):LBA5.
48. Howell SJ, Johnston SR, Howell A. The use of selective estrogen receptor modulators
and selective estrogen receptor down-regulators in breast cancer. Best Pract
Res Clin Endocrinol Metab 2004; 18:47–66.
49. Osborne CK, Wakeling A, Nicholson RI. Fulvestrant: an oestrogen receptor antagonist
with a novel mechanism of action. Br J Cancer 2004; 90(suppl 1):S2–S6.
50. Howell A, Robertson JF, Abram P, et al. Comparison of fulvestrant versus tamoxifen
for the treatment of advanced breast cancer in postmenopausal women previously
untreated with endocrine therapy: a multinational, double-blind, randomized trial.
J Clin Oncol 2004; 22:1605–1613.
51. Johnston SR, Dowsett M. Aromatase inhibitors for breast cancer: lessons from the
laboratory. Nat Rev Cancer 2003; 3:821–831.
52. Bonneterre J, Buzdar A, Nabholtz JM, et al. Anastrozole is superior to tamoxifen as
first-line therapy in hormone receptor positive advanced breast carcinoma. Cancer
2001; 92:2247–2258.

182 Zelnak et al.
53. Mouridsen H, Gershanovich M, Sun Y, et al. Superior efficacy of letrozole versus
tamoxifen as first-line therapy for postmenopausal women with advanced breast
cancer: results of a phase III study of the International Letrozole Breast Cancer
Group. J Clin Oncol 2001; 19:2596–2606.
54. Paridaens R, Dirix L, Lohrisch C, et al. Mature results of a randomized phase II
multicenter study of exemestane versus tamoxifen as first-line hormone therapy for
postmenopausal women with metastatic breast cancer. Ann Oncol 2003; 14:1391–1398.
55. Baum M, Budzar AU, Cuzick J, et al. Anastrozole alone or in combination with
tamoxifen versus tamoxifen alone for adjuvant treatment of postmenopausal women
with early breast cancer: first results of the ATAC randomised trial. Lancet 2002;
359:2131–2139.
56. Thurlimann B, Keshaviah A, Coates AS, et al. A comparison of letrozole and
tamoxifen in postmenopausal women with early breast cancer. N Engl J Med 2005;
353:2747–2757.
57. Coombes RC, Kilburn LS, Snowdon CF, et al. Survival and safety of exemestane
versus tamoxifen after 2 to 3 years tamoxifen treatment (Intergroup Exenestane
Study): a randomized controlled trial. Lancet 2007; 369:559–570.
58. Kaufmann M, Jonat W, Hilfrich J, et al. Improved overall survival in postmenopausal
women with early breast cancer after anastrozole initiated after treatment with
tamoxifen compared with continued tamoxifen: the ARNO 95 Study. J Clin Oncol
2007; 25:2664–2670.
59. Buzdar A, Jonat W, Howell A, et al. Anastrozole, a potent and selective aromatase
inhibitor, versus megestrol acetate in postmenopausal women with advanced breast
cancer: results of overview analysis of two phase III trials. Arimidex Study Group.
J Clin Oncol 1996; 14:2000–2011.
60. Dombernowsky P, Smith I, Falkson G, et al. Letrozole, a new oral aromatase
inhibitor for advanced breast cancer: double-blind randomized trial showing a dose
effect and improved efficacy and tolerability compared with megestrol acetate. J Clin
Oncol 1998; 16:453–461.
61. Kao YC, Cam LL, Laughton CA, et al. Binding characteristics of seven inhibitors
of human aromatase: a site-directed mutagenesis study. Cancer Res 1991; 56:
3451–3460.
62. Sourdaine P, Parker MG, Telford J, et al. Analysis of the aromatase cytochrome P450
gene in human breast cancers. J Mol Endocrinol 1994; 13:331–337.
63. Masamura S, Santner SJ, Heitjan DF, et al. Estrogen deprivation causes estradiol
hypersensitivity in human breast cancer cells. J Clin Endocrinol Metab 1995;
80:2918–2925.
64. Santen R, Jeng MH, Wang JP, et al. Adaptive hypersensitivity to estradiol: potential
mechanism for secondary hormonal responses in breast cancer patients. J Steroid
Biochem Mol Biol 2001; 79:115–125.
65. Chan CM, Martin LA, Johnston SR, et al. Molecular changes associated with the
acquisition of oestrogen hypersensitivity in MCF-7 breast cancer cells on long-term
oestrogen deprivation. J Steroid Biochem Mol Biol 2002; 81:333–341.
66. Martin LA, Farmer I, Johnston SR, et al. Enhanced estrogen receptor (ER) alpha,
ERBB2, and MAPK signal transduction pathways operate during the adaptation
of MCF-7 cells to long term estrogen deprivation. J Biol Chem 2003; 278:
30458–30468.

Individualization of Endocrine Therapy in Breast Cancer 183
67. Jeng MH, Yue W, Eischeid A, et al. Role of MAP kinase in the enhanced cell
proliferation of long term estrogen deprived human breast cancer cells. Breast
Cancer Res Treat 2000; 62:167–175.
68. Shim WS, Conaway M, Masamura S, et al. Estradiol hypersensitivity and mitogenactivated
protein kinase expression in long-term estrogen deprived human breast
cancer cells in vivo. Endocrinology 2000; 141:396–405.
69. Stephen RL, Shaw LE, Larsen C, et al. Insulin-like growth factor receptor levels are
regulated by cell density and by long term estrogen deprivation in MCF7 human
breast cancer cells. J Biol Chem 2001; 276:40080–40086.
70. Arpino G, Weiss H, Lee AV, et al. Estrogen receptor-positive, progesterone receptornegative
breast cancer: association with growth factor receptor expression and
tamoxifen resistance. J Natl Cancer Inst 2005; 97:1254–1261.
71. De Laurentiis M, Arpino G, Massarelli E, et al. A meta-analysis on the interaction
between HER-2 expression and response to endocrine treatment in advanced breast
cancer. Clin Cancer Res 2005; 11:4741–4748.
72. Ellis MJ, Tao Y, Young O, et al. Estrogen-independent proliferation is present in
estrogen-receptor HER2-positive primary breast cancer after neoadjuvant letrozole.
J Clin Oncol 2006; 24:3019–3025.
73. Smith IE, Dowsett M, Ebbs SR, et al. Neoadjuvant treatment of postmenopausal
breast cancer with anastrozole, tamoxifen, or both in combination: the Immediate
Preoperative Anastrozole, Tamoxifen, or Combined with Tamoxifen (IMPACT)
multicenter double-blind randomized trial. J Clin Oncol 2005; 23:5108–5116.
74. Viale G, Regan MM, Maiorano E, et al. Prognostic and predictive value of centrally
reviewed expression of estrogen and progesterone receptors in a randomized trial
comparing letrozole and tamoxifen adjuvant therapy for postmenopausal early stage
breast cancer: BIG 1-98. J Clin Oncol 2007; 25:3846–3852.
75. Dowsett M, Allred DC, for the TransATAC investigators. Relationship between
quantitative ER and PgR expression and HER2 status with recurrence in the ATAC
trial. SABCS 2006 (abstr 48).
76. Mackey JR, Kaufman B, Clemens M, et al. Trastuzumab prolongs progression-free
survival in hormone-dependent and HER2-positive metastatic breast cancer. SABCS
2006 (abstr 3).
77. Vogel CL, Cobleigh MA, Tripathy D, et al. Efficacy and safety of trastuzumab as a
single agent in first-line treatment of HER2-overexpressing metastatic breast cancer.
J Clin Oncol 2002; 20:719–726.
78. Dowsett M, Cuzick J, Wale C, et al. Retrospective analysis of time to recurrence in
the ATAC trial according to hormone receptor status: an hypothesis-generating
study. J Clin Oncol 2005; 23:7512–7517.
79. Van de Vijver MJ, He YD, van ’t Veer LJ, et al. A gene-expression signature as a
predictor of survival in breast cancer. N Eng J Med 2002; 347:1999–2009.
80. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifentreated,
node-negative breast cancer. N Engl J Med 2004; 351:2817–2826.
81. Paik S, Shak S, Tang G, et al. Expression of the 21 genes in the recurrence score
assay and prediction of clinical benefit from tamoxifen in NSABP study B-14 and
chemotherapy in NSABP study B-20. Breast Cancer Res Treat 2004; 88 (abstr 24).
82. Paik S, Tang G, Shak S, et al. Gene expression and benefit of chemotherapy in
women with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol
2006; 24:3796–3834.

13
TAILORx: Rationale for the Study Design
Joseph A. Sparano
Albert Einstein College of Medicine, Montefiore Medical Center,
Bronx, New York, U.S.A.
INTRODUCTION
There have been substantial technical and analytical advances within the past
decade that have facilitated high-throughput analysis of clinical specimens, a
process that has been referred to as “molecular profiling.” This term, as applied
here, may refer to evaluating various “markers,” including genomic, proteomic,
and epigenomic expression patterns, or a combination of these patterns, in
clinical specimens. Technological advances have led to the ability to measure
thousands of markers with a variety of validated methods (1), and analytical
models have been developed that facilitate analysis of the voluminous amount of
data that are generated (2). The technology has also led to an ability to perform
“discovery-based research,” in which large volumes of data are generated and
analyzed without a specific hypothesis, in contrast to the traditional scientific
paradigm of “hypothesis-based research,” in which a limited number of genes or
proteins are based upon a specific hypothesis and rationale (3). Discovery-based
research and hypothesis-based research are not mutually exclusive; however,
profiling may also be used to test specific hypotheses that are based on sound,
scientific rationale. A series of studies have been reported over the past few
years that have utilized gene expression profiling to discover “molecular
markers,” which may identify (i) distinct molecular subtypes of breast cancer
185

186 Sparano
(i.e., genotypic-phenotypic correlation), (ii) molecular signatures–associated
prognosis (i.e., prognostic factors), and (iii) molecular signatures that predict
the benefit from specific therapies (i.e., predictive factors) (4). When a
“molecular marker” is shown to perform more reliably than clinical features in
predicting prognosis or the benefit from a specific intervention, then there is
interest in further evaluating the utility of the marker in clinical practice (5).
PROGRAM FOR THE ASSESSMENT OF CLINICAL CANCER TESTS
In order to address the issue of integrating molecular diagnostic testing into
clinical practice, the U.S. National Cancer Institute initiated the program for the
assessment of clinical cancer tests (PACCT) (http://www.cancerdiagnosis.nci
.nih.gov/assessment/index.html). The goals of PACCT are to ensure translation
of new knowledge about cancer and new technologies to clinical practice and
development of more informative laboratory tools to help maximize the impact
of cancer treatments. This effort has led to the publication of the REMARK
(REporting recommendations for tumour MARKer prognostic studies)
guidelines (6) and to the TAILORx (Trial Assigning IndividuaLized Options
for Treatment) trial developed by the North American Breast Cancer Intergroup
(http://www.cancer.gov/clinicaltrials/digestpage/TAILORx). In reviewing
the rationale for design of the TAILORx trial in breast cancer, the following
questions were considered by the Breast Cancer Intergroup and are reviewed
herein:
l
l
l
l
What therapeutic intervention should the marker be used to select?
What patient population should be evaluated?
Which marker should be evaluated?
What trial design would integrate current knowledge about the marker and
address gaps in our knowledge about the current marker and future potential
markers?
WHICH TREATMENT INTERVENTION?
Therapeutic options are used in an adjunctive manner for patients with earlystage
breast cancer after surgery, irradiation, endocrine therapy, chemotherapy,
and trastuzumab (Table 1). In contrast to irradiation, endocrine therapy, and
trastuzumab, there are no clinical factors that may be used to predict the benefit
from adjuvant chemotherapy. Chemotherapy is usually recommended on the
basis of the risk of recurrence (i.e., prognostic factors) and age (7). In contrast
to other therapies, the treatment effect of chemotherapy is modest, acute toxicities
are frequent, and the cost is high (8). Therefore, evaluation of a marker
that has the potential for predicting benefit from chemotherapy would be
desirable.

TAILORx 187
Table 1
Standard Adjuvant Therapies for Operable Breast Cancer
Treatment
Selection
Percent
eligible
Treatment
effect (%)
Acute
toxicity
Relative
cost
Chemotherapy Recurrence risk Up to 90 25–35 High þþþ
> 5–10%
Endocrine ER- and/or PRpositive
70 50 Low þþ
therapy
tumor
Trastuzumab HER2/neu-positive 15 50 Low þþþþ
tumor
Radiation Breast-sparing surgery 40–60 90 Low þ
Abbreviations: ER, estrogen receptor; PR, progesterone receptor; HER2, human epidermal growth
factor receptor 2.
WHICH PATIENT POPULATION?
In 2006, approximately 124,000 women in the United States were diagnosed
with estrogen receptor (ER)–positive breast cancer associated with negative
axillary lymph nodes, accounting for about 50% of all breast cancer diagnosed
each year and nearly 9% of all new cases of cancer (9). Treatment recommendations
have generally been based on prognostic factors, with chemotherapy
recommended if the residual risk of recurrence exceeds approximately 5% to
10% despite adjuvant hormonal therapy (i.e., the tumor size exceeds 1 cm or
smaller tumors are associated with unfavorable histologic features) (7). By these
criteria, adjuvant chemotherapy could be recommended annually for up to
75,000 women younger than 70 years in the United States. Currently, it is estimated
that approximately 25% of all women, or about 30,000 annually, with earlystage
ER-positive breast cancer receive adjuvant chemotherapy. Its use has
increased substantially over the past 15 years (10) is highly dependent on age, and
is estimated to be given to 75% of those younger than 50 years, 30% between the
ages of 50 and 69 years, and about 5% of those 70 years or older.
Approximately 85% of women with early-stage ER-positive breast cancer
are therefore adequately treated with adjuvant hormonal therapy. Although
adding chemotherapy reduces the risk of recurrence on average by about 30%,
the absolute benefit for an individual patient is small, ranging from 1% to 5%
(11). The vast majority of patients with ER-positive breast cancer are therefore
overtreated with chemotherapy, since most would have been cured with hormonal
therapy alone. Decision aids such as Adjuvant! Online (12), decision
boards (13), and other tools, are often useful to assist patients and caregivers with
information regarding absolute benefits that might be expected from chemotherapy
(14). Although such decision aids may assist some patients in making a
more informed decision regarding whether to accept adjuvant chemotherapy,
when faced with a choice, many patients and their clinicians err on the side of
overtreatment because of the imprecise nature of predicting the treatment benefit

188 Sparano
(15). Therefore, developing a clinical trial that employs a molecular marker to
define the chemotherapy benefit was judged by the Breast Cancer Intergroup to
be the preferred setting. Recent evidence demonstrating that the incremental
benefits associated with contemporary chemotherapy regimens had minimal
impact in high-risk ER-positive disease reinforces this approach (16).
WHICH MARKER?
A number of molecular diagnostic tests that provide more accurate prognostic
information that standard clinical features have been developed and externally
validated and are summarized in Table 2, including the Amsterdam 70-gene
profile (17,18), Rotterdam 76-gene profile (19,20), and a 21-gene profile (21,22).
Recently, these assays have been shown to have comparable discriminatory
value in early-stage breast cancer (23).
The 21-gene Oncotype DX TM (Genomic Health, Inc., Redwood City,
California U.S.) includes 16 tumor genes and 5 reference genes, with the result
expressed as a computed “recurrence score” (RS) (Fig. 1) (21). The genes
utilized in the assay were identified in several training set trials (24), followed by
a validation study that was performed in patients with ER-positive, node-negative
breast cancer who received a five-year course of tamoxifen (TAM) in
National Surgical Adjuvant Breast and Bowel Project (NSABP) trial B-14 (21).
A higher RS was associated with an elevated risk of distant recurrence, whether
evaluated as a categorical or a continuous variable (Fig. 2). This assay was
selected for evaluation in the TAILORx trial for several scientific and practical
reasons that are summarized in Table 3 and discussed herein. In addition to the
prospective validation study illustrated by Figure 2, a population-based external
validation study performed in patients treated in the Kaiser Permanente health
care system provided further evidence for the robustness of the assay (22).
Moreover, subsequent studies in patients treated with TAM with or without
Table 2
Externally Validated Multigene Breast Cancer Markers
Number
Population
Commercially
Refs. of genes Tissue ER/PR Nodes Adjuvant therapy available
27,28 534 Frozen þ/– þ/– Preoperative chemo Not available
17,18 70 Frozen þ/– þ/– þ/– Chemo; MammaPrint 1
þ/– hormones
21,22 21 Paraffin þ TAM Oncotype DX
embedded
19,20 76 Frozen þ/– No therapy Not available
Abbreviations: ER, estrogen receptor; PR, progesterone receptor; Chemo, chemotherapy; TAM,
tamoxifen.

TAILORx 189
Figure 1
Oncotype DX RS: genes and algorithm. Abbreviation: RS, recurrence score.
Figure 2 Oncotype DX RS predicts distant recurrence as either a categorical variable
(left) or continuous variable (right). Abbreviation: RS, recurrence score.
adjuvant chemotherapy in NSABP trial B20 indicated that only patients with an
elevated RS (RS > 30) derived benefit from adjuvant chemotherapy (25). Other
analyses also indicated that an RS also predicts local recurrence, which may
occur in up to 20% of patients with operable breast cancer and which may be
reduced by adjuvant chemotherapy (26). Furthermore, an RS more accurately
predicted recurrence than the Adjuvant! Online model, providing evidence that it
offers information that is complementary to Adjuvant! (12). Finally, although
Oncotoype DX includes a reverse transcriptase polymerase chain reaction
(RT-PCR) assay-based evaluation of ER expression, a tumor with high ER
content by RT-PCR may have either a low or a high RS, indicating that it reflects
a measure other than simply ER expression (25).

190 Sparano
Table 3
Scientific Rationale for Selecting Oncotype DX Assay for TAILORx
Rationale
Ref.
Validation study demonstrating that an RS is a prognostic factor for risk of distant 21
recurrence in tamoxifen-treated patients, which outperforms clinical criteria in
the multivariate model
External population-based validation study demonstrating an RS predictive of 22
breast cancer death
A high RS predictive of chemotherapy benefit in tamoxifen-treated patients 25
An RS correlating with risk of local recurrence 26
An RS predicting the outcome more accurately than Adjuvant! Online 29
Concordance in the predictive value of Oncotype DX compared with other 23
multigene markers
Abbreviation: RS, recurrence score.
There were also several practical reasons for selecting Oncotype DX. First
and foremost is the fact that it involves an RT-PCR-based technique that may be
employed on as little as three 10-mm sections of routinely processed and stored
formalin-fixed, paraffin-embedded tissue. In addition, the assay was approved by
Clinical Laboratory Improvement Amendments (CLIA) in the United States and
has received favorable local coverage decisions regarding reimbursement by
Medicare, indicating acceptance of the value of the assay by third-party payers.
Finally, selecting Oncotype DX for further evaluation by the Breast Cancer
Intergroup represents a logical extension of the successful public-private collaboration
between NSABP and Genomic Health.
WHAT TRIAL DESIGN?
In designing a trial integrating a molecular diagnostic test into clinical decision
making, three factors must be considered. First, it is essential to incorporate
features already known about the prognostic and predictive value of the test into
the trial design. In this regard, it is clear that patients with a very low RS do very
well with hormonal therapy alone, and patients with a very high RS derive great
benefit from chemotherapy, thereby providing sufficient information to make
definitive treatment recommendations for patients who fall into these groups.
Second, the trial must be designed to address the clinical utility of the test in
circumstances where its utility is uncertain or result uninformative. This scenario
applies to patients with a tumor that has a mid-range RS, in which the risk of
relapse may be sufficiently high to recommend chemotherapy, but in which
chemotherapy may be not be beneficial. Third, it is important to design the trial
in a manner that offers an opportunity to evaluate other clinical cancer tests as
they evolve, obviating the need to repeat a large-scale clinical trial and providing
an opportunity to evaluate the test within a very short time frame.

TAILORx 191
Figure 3
TAILORx schema. Abbreviation: RS, recurrence score.
The factors outlined in the previous paragraph were taken into account
while designing TAILORx. Only patients who meet established clinical criteria
for recommending adjuvant chemotherapy, who are medically suitable candidates
for chemotherapy, and who agree to have their treatment assigned or
randomized on the basis of the molecular test are eligible. After informed consent
and “preregistration,” a tumor specimen is forwarded to Genomic Health for the
Oncotype DX assay, with the result reported to the site within 10 to 14 days. After
the Oncotype DX RS is known, patients are then registered and have their treatment
assigned or randomized on the basis of the RS (Fig. 3), as described below:
l
l
l
RS is 10 or less: Hormonal therapy alone
RS is 26 or higher: Chemotherapy plus hormonal therapy
RS is 11 to 25: Patients randomized to receive chemotherapy plus hormonal
therapy (the standard treatment arm) versus hormonal therapy alone
(the experimental treatment arm)
Because the study includes only women who meet standard clinical criteria
for adjuvant chemotherapy and are medically suitable candidates for therapy,
chemotherapy plus hormonal therapy is considered the standard treatment arm
for patients with a mid-range RS (i.e., 11–25). Hormonal therapy is considered
the experimental treatment arm for the mid-range group. The primary objective

192 Sparano
of the trial is to determine whether hormonal therapy alone is not inferior to
chemotherapy plus hormonal therapy. The trial utilizes a “noninferiority” design
for the mid-range group and is adequately powered to detect a 3% or greater
difference between the two treatment arms.
The chemotherapy regimen and hormonal agents used are prescribed at the
discretion of the treating physician but must be consistent with one of several
standard options described in the protocol document. Patients may also enroll in
other Intergroup trials as long as the protocol treatment is consistent with the
TAILORx-assigned treatment arm (i.e., hormonal therapy or chemotherapy plus
hormonal therapy). In addition, patients who have already had the Oncotype DX
assay performed are eligible to enroll if the RS is 11 to 25.
At the time of registration and treatment assignment, patients are asked to
provide blood samples for banking of plasma and peripheral blood mononuclear
cells. Tissue specimens are collected by the Eastern Cooperative Oncology
Group (ECOG) Pathology Coordinating Office, with central testing and confirmation
of ER and progesterone receptor (PR) expression, creation of tissue
microarrays, and RNA extraction for future studies.
The trial is available through the Cancer Trials Support Unit (www.ctsu.org)
and is being coordinated by ECOG. Other participants include NSABP and all
members of the North American Breast Intergroup, including the Southwest
Oncology Group (SWOG), Cancer and Acute Leukemia Group B (CALGB), North
Central Cancer Treatment Group (NCCTG), National Cancer Institute of Canada
(NCIC), and the American College of Surgeons Oncology Group (ACOSOG).
RATIONALE FOR THE RS RANGES IN TAILORx
The RS ranges used in TAILORx are different from those originally described
for the low- (30) risk groups as
originally reported for the assay. The range was adjusted in order to minimize
the potential for undertreatment in both the high-risk group (also called “secondary
study group 2”) and the randomized group (also called the “primary study
group”). When the NSABP B20 data were analyzed using the RS ranges used in
TAILORx, the treatment effect of chemotherapy was similar to the original
analysis (Table 4). For the analyses described in Table 4, disease-free survival
(DFS) is defined as time to first local, regional, or distant recurrence, second
primary cancer, or death due to any cause, and distant recurrence-free survival
(DRFS) is defined as time to first distant recurrence or death from breast cancer
(contralateral breast cancers, other second primary cancers, and deaths due to
other cancers are censored, and local and regional recurrences are ignored). DFS
has been the primary endpoint evaluated in most breast cancer trials and is the
primary endpoint for TAILORx, whereas DRFS was the primary endpoint in
the NSABP trials evaluating RS. Although a trend favoring the addition of
chemotherapy becomes evident at an RS of approximately 11 when the risk of
relapse is analyzed in a linear fashion, the 95% confidence intervals completely

TAILORx 193
Table 4 RS and Response to Chemotherapy in B20 Trial (n ¼ 651) by RS
RS
Number of
patients
TAM 10-yr
DRFS (%)
TAM þ
chemo
10-yr
DRFS (%)
HR (95% CI) for
recurrence by
addition of
chemo P-value
TAM
10-yr
DFS (%)
TAM þ
chemo
10-yr
DFS (%)
HR (95% CI) for
recurrence by
addition of chemo P-value
25 195 (30%) 63 88 0.285 (0.148, 0.551)

194 Sparano
Figure 4 Treatment effect of chemotherapy by Oncotype DX RS. Abbreviation: RS,
recurrence score.
overlap in the 11 to 25 RS range (Fig. 4). Finally, a high RS has also been
associated with an elevated risk of local relapse, which may also be reduced by
giving adjuvant chemotherapy (26). An RS of 11 is associated with a risk of both
local and distant relapse of about 10%, a threshold that has been typically used
for recommending adjuvant chemotherapy.
CONCLUSION
TAILORx represents a major step forward into the era of personalized medicine
for breast cancer. By integrating a molecular diagnostic test into clinical decision
making, patients and clinicians will be able to make more informed decisions
regarding the most appropriate treatment options for a subset of patients for
whom the test results in a clear treatment path and refine the utility of the assay
for those for whom the result may be uninformative. This trial will also serve as
an important resource for evaluating new molecular signatures and other technologies,
such as proteomics, epigenomics, and pharmacogenomics, as these
technologies evolve.
REFERENCES
1. Hanash S. Integrated global profiling of cancer. Nat Rev Cancer 2004; 4:638–644.
2. Simon R. Diagnostic and prognostic prediction using gene expression profiles in
high-dimensional microarray data. Br J Cancer 2003; 89:1599–1604.

TAILORx 195
3. Ransohoff DF. Evaluating discovery-based research: when biologic reasoning cannot
work. Gastroenterology 2004; 127:1028.
4. Sparano JA, Fazzari MJ, Childs G. Clinical application of molecular profiling in
breast cancer. Future Oncol 2005; 1:485–496.
5. Hayes DF, Bast RC, Desch CE, et al. Tumor marker utility grading system: a framework
to evaluate clinical utility of tumor markers. J Natl Cancer Inst 1996; 88: 1456–1466.
6. McShane LM, Altman DG, Sauerbrei W, et al. REporting recommendations for
tumour MARKer prognostic studies (REMARK). Eur J Cancer 2005; 41:1690–1696.
7. Carlson RW, Brown E, Burstein HJ, et al. NCCN Task Force Report: Adjuvant
Therapy for Breast Cancer. J Natl Compr Canc Netw 2006; 4(suppl 1):S1–S26.
8. Hassett MJ, O’Malley AJ, Pakes JR, et al. Frequency and cost of chemotherapyrelated
serious adverse effects in a population sample of women with breast cancer.
J Natl Cancer Inst 2006; 98:1108–1117.
9. Jemal A, Siegel R, Ward E, et al. Cancer statistics, 2006. CA Cancer J Clin 2006;
56:106–130.
10. Harlan LC, Clegg LX, Abrams J, et al. Community-based use of chemotherapy and
hormonal therapy for early-stage breast cancer: 1987-2000. J Clin Oncol 2006; 24:
872–877.
11. Fisher B, Jeong JH, Dignam J, et al. Findings from recent National Surgical Adjuvant
Breast and Bowel Project adjuvant studies in stage I breast cancer. J Natl Cancer Inst
Monogr 2001; 30:62–66.
12. Olivotto IA, Bajdik CD, Ravdin PM, et al. Population-based validation of the prognostic
model ADJUVANT! for early breast cancer. J Clin Oncol 2005; 23:2716–2725.
13. Whelan T, Levine M, Willan A, et al. Effect of a decision aid on knowledge and
treatment decision making for breast cancer surgery: a randomized trial. JAMA
2004; 292:435–441.
14. Whelan TJ, Loprinzi C. Physician/patient decision aids for adjuvant therapy. J Clin
Oncol 2005; 23:1627–1630.
15. Simes RJ, Coates AS. Patient preferences for adjuvant chemotherapy of early breast
cancer: how much benefit is needed? J Natl Cancer Inst Monogr 2001; 30:146–152.
16. Berry DA, Cirrincione C, Henderson IC, et al. Estrogen-receptor status and outcomes
of modern chemotherapy for patients with node-positive breast cancer. JAMA 2006;
295:1658–1667.
17. van de Vijver MJ, He YD, van ’t Veer LJ, et al. A gene-expression signature as a
predictor of survival in breast cancer. N Engl J Med 2002; 347:1999–2009.
18. Buyse M, Loi S, van ’t Veer LJ, et al. Validation and clinical utility of a 70-gene
prognostic signature for women with node-negative breast cancer. J Natl Cancer Inst
2006; 98:1183–1192.
19. Wang Y, Klijn JG, Zhang Y, et al. Gene-expression profiles to predict distant
metastasis of lymph-node-negative primary breast cancer. Lancet 2005; 365:671–679.
20. Foekens JA, Atkins D, Zhang Y, et al. Multicenter validation of a gene expressionbased
prognostic signature in lymph node-negative primary breast cancer. J Clin
Oncol 2006; 24:1665–1671.
21. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifentreated,
node-negative breast cancer. N Engl J Med 2004; 351:2817–2826.
22. Habel LA, Shak S, Jacobs MK, et al. A population-based study of tumor gene
expression and risk of breast cancer death among lymph node-negative patients.
Breast Cancer Res 2006; 8:R25.

196 Sparano
23. Fan C, Oh DS, Wessels L, et al. Concordance among gene-expression-based
predictors for breast cancer. N Engl J Med 2006; 355:560–569.
24. Cobleigh MA, Tabesh B, Bitterman P, et al. Tumor gene expression and prognosis in
breast cancer patients with 10 or more positive lymph nodes. Clin Cancer Res 2005;
11:8623–8631.
25. Paik S, Tang G, Shak S, et al. Gene expression and benefit of chemotherapy in
women with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol
2006; 24(23):3726–3734.
26. Mamounas ETG, Bryant J, Paik S, et al. Association between the 21-gene recurrence
score assay (RS) and risk of loco-regional failure in node-negative, ER-positive
breast cancer: results from NSABP B-14 and NSABP B-20. Breast Cancer Res 2005;
94(suppl 1):S16 (abstr 29).
27. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas
distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci
U S A 2001; 98:10869–10874.
28. Sorlie T, Tibshirani R, Parker J, et al. Repeated observation of breast tumor subtypes
in independent gene expression data sets. Proc Natl Acad Sci U S A 2003;
100:8418–8423.
29. Bryant J. Oncotype DX correlates more closely with prognosis than Adjuvant
Online. Paper presented at: 9th International Conference on Primary Therapy of
Early Breast Cancer January 26-29, 2005; St. Gallen, Switzerland. Available at:
http://www.oncoconferences.ch/2005/program/overview.htm.

14
Taking Prognostic Signatures to the Clinic
Michail Ignatiadis and Christos Sotiriou
Department of Medical Oncology, Translational Research Unit,
Institut Jules Bordet, Universite´ Libre de Bruxelles, Brussels, Belgium
INTRODUCTION
The advent of microarray technology and the sequencing of the human genome
have enabled scientists to simultaneously interrogate the expression of thousands
of genes. These gene expression profiling studies have provided (i) a molecular
classification of breast cancer into clinically relevant subtypes, (ii) new tools to
predict disease recurrence and response to different treatments, and (iii) novel
insight into various oncogenic pathways and the process of metastatic progression.
Breast cancer prognosis has been extensively studied by gene expression
profiling, resulting in different signatures with little overlap in their constituent
genes. In a large meta-analysis of all the above studies, despite the disparity in
their gene lists, molecular subtyping and different prognostic signatures were
shown to carry similar prognostic information. In all these signatures, proliferation
genes were the common driving force with regard to prognostication (1).
Gene expression signatures, despite their current limitations may further
refine breast cancer prognosis and prediction of benefit from systemic therapy,
beyond conventional clinicopathological prognostic and predictive factors (2). In
this chapter, we will not refer to gene expression signatures that predict response
to systemic therapy. We will focus only on genomic predictors related to prognosis
and the way these predictors are expected to impact patient management in the
future and accelerate the transition from the “empirical” to “tailored” oncology.
197

198 Ignatiadis and Sotiriou
MOLECULAR CLASSIFICATION OF BREAST CANCER
BY GENE EXPRESSION PROFILING
Clinicians have long recognized that the diagnosis of breast cancer includes tumor
types with different natural histories and responses to various treatments. Nevertheless,
traditional histopathological characteristics have been unable to capture the
biological heterogeneity of breast tumors. The first studies employing microarray
technology to address this issue were performed by Perou and Sorlie (3–5). They
proposed a molecular classification of breast cancer with at least five subgroups on
the basis the expression patterns of 500 “intrinsic genes.” Three subgroups were
characterized by low to absent expression of the estrogen receptor (ER) and ERrelated
genes, the basal-like, the HER2þ, and the normal breast-like subgroup. The
first subtype was characterized by high expression of genes found in myoepithelial
cells in the outer basal layer of a normal breast duct, like keratins 5/6 and 17, and
was therefore named basal-like. The HER2þ subgroup was characterized by high
expression of several genes of the ERBB2 amplicon. The normal breast-like subgroup
showed genes expressed by adipose and nonepithelial tissues and also
revealed a strong expression of basal epithelial genes. The other two subtypes
consisted of ER-positive tumors and expressed genes of the luminal epithelial cells
that are found in the inner layer of a normal breast duct, and, as a result, they were
termed luminal. The luminal A subgroup had the highest expression of the ERa and
the GATA-binding protein 3 gene, and the second subgroup (luminal B and C)
showed low to moderate expression of these genes. To assign each sample to one of
the five molecular subtypes (basal-like, HER2þ/ER , luminal A, luminal B, normal
breast-like), Hu et al. have created a single sample predictor by calculating the
mean expression profiles (centroids) of a new “intrinsic set of 300 genes” for each
subgroup (6). In a multivariate analysis, the basal-like, HER2þ/ER , luminal A,
and luminal B subgroups provided important prognostic information beyond those
provided by traditional predictors like tumor size and grade.
Despite using a different microarray platform, Sotiriou et al. confirmed the
consistency and the prognostic relevance of the above molecular classification of
breast cancer in a different patient population (7). Again, ER status was the most
important discriminator of gene expression patterns. Interestingly, although
tumor grade was also strongly reflected in the transcriptome profile of the primary
tumor, this was not the case for tumor size and axillary nodal invasion.
These results confirmed the important role of ER and tumor grade biology in
defining breast cancer molecular subtypes. In conclusion, microarray studies
with different technology platforms and in different patient populations treated
with heterogeneous treatments have been remarkably consistent in reproducing a
similar molecular classification of breast cancer.
However, it should be pointed out that the number of robust breast cancer
molecular subtypes is not certain, and there is no standardized method for assigning
a molecular subtype to a new case (8). Nevertheless, the breast cancer molecular
classification has begun to alter the way clinical trials are designed. As an example,

Taking Prognostic Signatures to the Clinic 199
new trials have been designed to test the efficacy of targeted agents only in the
basal-like breast tumors.
GENE EXPRESSION SIGNATURES AS TOOLS FOR IMPROVING
PROGNOSIS IN BREAST CANCER
Besides the above studies aiming at providing a molecular taxonomy of breast
tumors, several independent groups, including our own, have conducted gene
expression profiling studies with the objective of improving upon traditional
prognostic markers used in the clinic for prediction of outcome. Two conceptually
different, supervised approaches for prognostic marker discovery on a
genome-wide scale have been applied so far, the “top-down” approach and the
“bottom-up” or “hypothesis-driven” approach. The former derives from a
prognostic model simply by looking for gene expression patterns associated with
the clinical outcome without any a priori biological scenario, whereas the latter
approach first identifies gene expression profiles linked with a specific biological
phenotype and subsequently correlates these findings to survival.
Using the top-down approach, researchers from the Netherlands Cancer
Institute (NKI) first identified a 70-gene prognostic signature in a series of 78
systemically untreated node-negative breast cancer patients. This signature
(MammaPrint 1 Agendia, Amsterdam, the Netherlands) was then validated on a
larger set of 295 young treated and untreated patients, with either node-negative
or -positive breast tumors, from the same institution (9,10). The Rotterdam group
later identified, using a training set of 115 breast cancer patients, a 76-gene
signature to predict the development of distant metastases in untreated nodenegative
breast cancer patients (11). Interestingly, this group used the ER status
to divide the training set into two subgroups, and the genes selected from each
subgroup (60 and 16 genes for the ER-positive and ER-negative subgroups,
respectively) were combined to form a single signature. The 76-gene signature
(VDX2 array, Veridex LLC, Warren, New Jersey, U.S.) was recently validated in
180 lymph node–negative, untreated patients from multiple institutions (12). The
two groups have used different technology platforms (the NKI group used the
Agilent platform, Santa Clara, California, U.S. and the Rotterdam group used
the Affymetrix platform, Santa Clara, California, U.S.), and although there was
only a three-gene overlap between them, both signatures outperformed the
National Institutes of Health (NIH) (13) and St. Gallen criteria (14) in stratifying
patients according to risk of relapse. Indeed, while the prognostic signatures
correctly identified those patients who will relapse, they could also identify
many low-risk patients who could be spared from unnecessary chemotherapy.
However, the identification of high-risk patients could still be improved, since
half of the patients assigned to the poor-signature group, in fact, did not relapse.
A major challenge for gene expression profiling studies, especially those
with clear clinical implications, is independent validation. Consequently,
TRANSBIG, the translational research network of the Breast International Group

200 Ignatiadis and Sotiriou
(BIG), conducted an independent validation study of these two prognostic signatures
in a series of 302 untreated patients from five different centers and across
different statistical facilities (15,16). The two signatures were robust, and their
performance was reproducible and similar in this retrospective series. Interestingly,
both signatures were strong predictors of the development of metastasis
within five years of diagnosis (early metastasis), but their performance decreased
for the detection of late metastasis. This could be easily explained, since the
training set in both studies consisted of patients who have all relapsed within five
years. However, an intriguing question arises about the potential differences in
the biology and the oncogenic pathways that drive early and late metastasis.
Another group explored the correlation between the expression of 250
genes (selected from publicly available microarray data linked with breast cancer
biology) and the clinical outcome in 447 patients from three studies including the
tamoxifen-only group of the NASBP trial B-20 to build the Oncotype DX 1
(Genomic Health Inc, Redwood City, California, U.S.) Recurrence Score 1 (RS)
(17). This score (low, intermediate, and high) was based on the expression of a
final selected panel of 16 cancer-related genes and five reference genes, as
assessed by reverse transcriptase-polymerase chain reaction (RT-PCR) in paraffin-embedded
tumor tissue. Then, the Oncotype DX RS was validated for its
ability to quantify the likelihood of distance recurrence in the tamoxifen treated,
ER-positive, node-negative group of the NSABP trial B-14 (17). Using the RS,
around 50% of the patients in the above study were characterized as low risk
(low RS), with a 6.8% probability of distant recurrence at 10 years.
Using the second approach, the hypothesis-driven approach, Chang et al.
tried to identify the molecular similarities between cancer invasion/metastasis
and wound healing (18). They developed a gene signature to characterize the
response of fibroblasts to serum. When they applied these genes to primary
gene expression profiling data from multiple data sets, they observed that the
presence of a wound-healing phenotype in the primary tumor was related with
worse clinical outcome in various cancers, including breast cancer. Similar to the
70- and 76-gene signatures, the wound-response signature provided better risk
stratification than the NIH and St. Gallen criteria. Furthermore, it was able to
refine prognosis in the 295 patient data sets of van de Vijver et al. (10) beyond
existing clinicopathological risk factors and the 70-gene signature (19). In
conclusion, Chang et al., using a hypothesis-driven approach, apart from identifying
a gene expression prognostic signature, have offered novel insight into
the molecular pathogenesis of metastasis by linking this process with wound
healing.
Another example of deriving a prognostic gene expression signature using
a hypothesis-driven approach is the study by our group (20) that focused on
histological grade, a well-established pathological parameter rooted in the cell
biology of breast cancer. Applying a supervised analysis, we developed a Gene
expression Grade Index (GGI) score based on 97 genes. These genes were
mainly involved in cell cycle regulation and proliferation and were consistently

Taking Prognostic Signatures to the Clinic 201
differentially expressed between low- and high-grade breast carcinomas. Interestingly,
the GGI, which essentially quantifies the degree of similarity between
the tumor expression pattern of these 97 genes and tumor grade, was able to
reclassify patients with histological grade 2 tumors into two groups with distinct
clinical outcomes similar to those of histological grades 1 and 3, respectively
(20). This observation challenges the existence and clinical relevance of an
intermediate grade classification. When we explored the implications of the joint
distribution of ER status and GGI in predicting clinical outcome, we found that
almost all ER-negative tumors with poor clinical outcome were associated with a
high GGI score (high grade), whereas ER-positive tumors were associated with a
heterogeneous mixture of gene expression grade values. Recently, our group has
reported that GGI can be used to define two clinically relevant molecular subtypes
in ER-positive breast cancer (21). These two ER-positive molecular subgroups
(high and low genomic grade) were highly comparable to the luminal A
and B classifications and were associated with distinct clinical outcome in
systemically untreated or tamoxifen-only-treated ER-positive breast cancer
populations involving more than 650 patients (21). GGI appeared to be the
strongest predictor of clinical outcome, highlighting the prognostic importance
of proliferation genes in ER-positive subgroups, as have been already reported
(17). Indeed, proliferation was one of the five components associated with the
highest weight in the Oncotype DX RS.
Four additional studies have used the hypothesis-driven approach to build
gene predictors of clinical outcome. Oh et al. used estrogen-induced genes,
identified by treating the MCF-7 cell line with 17b-estradiol to build a predictor
of clinical outcome on a training set of 65 patients with ER-positive breast
tumors (22). A total of 822 genes, mainly proliferation and antiapoptosis genes
could separate these tumors into two subgroups with different clinical outcomes.
This gene list was then mapped in three published datasets (4,5,19,23) and an
average expression index or “centroid” was created that could classify the ERpositive
tumors into two clinically relevant subgroups. In another study, Miller et
al. generated a 32-gene expression signature that distinguishes p53 mutant and
wild type tumors. The aim of this research was to provide a more accurate
assessment of the functional status of the p53 pathway than that provided by
mutation analysis using sequencing (24). It should be noted that none of the
signature genes were known transcriptional targets of p53, nor have they been
related to the p53 pathway. Instead, they were mainly proliferation-related genes,
various transcription factors, and ion transport genes. Once more, the proliferation
genes were a prominent part of this signature. The signature outperformed
p53 mutation status, as identified by sequencing in prognosis and prediction of
response to hormonal therapy and chemotherapy. By combining the hypothesisdriven
approach with comparative cross-species functional genomics, Glinsky
et al. identified an 11-gene “death-from-cancer” signature (25). This signature
can identify a lethal subset of human cancers of various origins (including breast
cancer) with a high propensity to metastasize. These tumors display the

202 Ignatiadis and Sotiriou
“phenotype” of an activated BMI-1 oncogenic pathway that is essential for the
self-renewal of normal stem cells. Again, proliferation- and cell cycle–regulating
genes comprise the most essential part of this signature. Along the same lines,
Liu et al. isolated and compared at the transcriptional-level tumorigenic CD44þ/
CD24 /low breast cancer cells derived from a small set of six breast cancer
patients with normal breast epithilium (26). They developed an “invasiveness”
gene signature (IGS) that was associated with the risk of death and metastasis not
only in breast cancer but also in other tumor types. Again, there was a small gene
overlap between IGS and other prognostic signatures, and, not surprisingly, its
prognostic power was significant only in ER-positive and intermediate-grade
tumors, a feature that had already been observed for other prognostic profiles.
CAN WE INTEGRATE THE DIFFERENT GENE EXPRESSION SIGNATURES?
The above gene expression signatures developed to improve breast cancer
prognostication have little overlap in their constituent genes. A representative
list of these signatures is depicted in Table 1. The inevitable question is whether
the “intrinsic subtypes” developed by unsupervised cluster analysis and the
other prognostic signatures have also little overlap in the prognostic information
they convey. Fan et al. recently compared the prognostic ability of five gene
expression–based models—intrinsic subtypes (6), 70-gene profile (9), wound
response (19), RS (17), and two-gene ratio (23)—in a single dataset of
295 patients and found that four of five predictors had significant agreement in
outcome predictions for individual patients (27). However, this study was limited
to only one dataset of 295 patients, examined only five gene expression signatures,
and did not investigate the biology that may drive prognosis among the five
signatures. To address this issue, a large meta-analysis of publicly available gene
expression and clinical data from 2833 patients (4–7,9,12,15,20,23,24,28–39) with
breast cancer was performed by the Swiss Institute of Bioinformatics in collaboration
with our group (Table 1). In this meta-analysis, the concept of “coexpression
modules” (comprehensive list of genes with highly correlated
expression) was used to describe important biological processes in breast cancer,
like proliferation, ER and HER2 signaling, tumor invasion, and immune response.
In this study, we sought to depict the connection between these modules, the
previously reported molecular classification, several gene prognostic classifiers,
and the most established clinicopathological variables. A number of interesting
conclusions were drawn from this collaborative effort. First, the disparity of gene
lists of the various gene signatures may be attributed to (i) heterogeneity in patient
populations studied, (ii) the use of different microarray platforms with different
probe sets and different methods for data normalization, and (iii) sampling variation
due to small sample size relative to the number of genes examined. Second,
using these co-expression modules, instead of the five molecular portraits reported
with the intrinsic genes, breast tumors were now grouped into three main subgroups
namely ER /HER , HER2þ, and ERþ/HER2 . Third, the ER /HER ,

Taking Prognostic Signatures to the Clinic 203
Table 1 Gene Expression Signatures Related to Breast Cancer Prognosis
Gene expression
signatures Biological scenario Microarray platform
Number of
genes in the
signature
Independent
validation
Prospective
clinical validation Ref.
70-gene signature Clinical outcome Agilent (oligonucleotides) 70 yes Yes (MINDACT
Trial)
76-gene signature Clinical outcome Affymetrix
(oligonucleotides)
(9,10,15)
76 yes No (11,12,16)
Recurrence Score Clinical outcome RT-PCR 21 yes Yes (TAILORX
trial)
Wound-response
signature
Wound healing and tumor
progression
Genomic grade Histological grade and tumor
progression
(17,37)
CDNA (custom made) 512 yes No (18,19)
Affymetrix
(oligonucleotides)
P-53 signature Functional status of p-53 Affymetrix
(oligonucleotides)
Death-from-cancer
signature
Invasiveness gene
signature
Bmi-1 oncogenic pathway
self-renewal
Tumorigenic cancer cells
CD44þ/CD24 , low
Affymetrix
(oligonucleotides)
Affymetrix
(oligonucleotides)
97 yes No (20)
32 yes No (24)
11 Yes No (25)
186 Yes No (26)

204 Ignatiadis and Sotiriou
Figure 1 Despite the disparity in their gene lists, different prognostic signatures showed
similar prognostic performance. Proliferation is the common driving force responsible for
their prognostic performance.
HER2þ tumors were characterized by high proliferation, whereas the
ERþ/HER2 was divided in two subgroups, the ERþ/low proliferation and the
ERþ/high proliferation tumors resembling to luminal A and B subtypes,
respectively. Fourth, 10 prognostic signatures [RS (17), 70-gene signature (9),
76-gene signature (11), wound response (18), p53 signature (24), GGI (20),
NCH70 (33), CON52 (38), CCYC (39), ZCOX], despite the disparity in their
gene lists, showed similar prognostic performance (Fig. 1). It is interesting to
note that they all identified as “low risk” the ERþ subtype associated with low
proliferation. Fifth, proliferation was the common driving force responsible for
the performance of the various prognostic signatures (Fig. 1). Indeed, to further
investigate the role of proliferation genes in relation to the different prognostic
signatures, we divided each signature into two “partial signatures”—one with only
proliferation genes and the other with the complementary nonproliferation genes.
Interestingly, when only proliferation genes were used, the total performance was
not degraded. In contrast, the nonproliferation partial signatures typically show
degraded performance. Sixth, combining the signatures did not improve the performance
as expected from the high concordance in their classifications. Finally,
in the multivariate analysis, including the genomic predictor derived from coexpression
modules, the standard clinicopathological variables, namely tumor size
and nodal status, retained their prognostic value.
PROSPECTIVE VALIDATION OF TWO GENOMIC PREDICTORS
The initial promising results that breast cancer prognostication can be improved
using gene expression signatures need to be validated by well-designed prospective
clinical trials. Two gene expression predictors, the 70-gene expression

Taking Prognostic Signatures to the Clinic 205
signature, (MammaPrint) discussed earlier and the Oncotype DX RS have
reached the final step of prospective clinical testing. The BIG and the EORTC
have jointly launched the MINDACT (Microarray for Node-Negative Disease
may Avoid Chemotherapy) trial that directly compares two approaches for
deciding whether to offer adjuvant chemotherapy in node-negative breast cancer
patients, i.e., either using the prognostic information provided by standard
clinicopathological criteria (control arm) or by the 70-gene expression signature
(MammaPrint) (experimental arm). Six thousand women will participate, and
the primary endpoint is to prove that five-year disease-free survival in the
experimental arm will not be inferior to that of the control arm; at the same time,
it is estimated that 10% to 15% fewer women will be treated with chemotherapy
in the experimental arm.
A similarly ambitious initiative is taking place in the United States. The
Oncotype DX RS is being evaluated in the TAILORx [Trial Assigning
IndividuaLized Options for Treatment (Rx)] trial, which is part of the
National Cancer Institute (NCI) Program for the Assessment of Clinical
Cancer Tests (PACCT). This trial randomizes patients with intermediate RS
to receive either hormonal therapy alone or hormonal therapy plus chemotherapy.
In it, patients with low RS will be treated with hormonal therapy
alone and those with high RS will receive chemotherapy plus hormonal
therapy.
These two trials will provide level I evidence that these gene expression
predictors provide prognostic information beyond standard clinicopathological
factors and are ready for use in clinical practice. Both trials will provide an
excellent opportunity to address issues related to tissue handling and shipping,
reproducibility, quality controls, and standardization of the microarray experiments.
Furthermore, they will provide a valuable database for future translational
research.
FUTURE DIRECTION–CONCLUSIONS
The real challenge, once the above trials prospectively validate the two genomic
predictors, will be to integrate these predictors into a comprehensive clinical
decision-making algorithm that will lead toward personalized medicine.
Parameters that need to be included in this algorithm include the following:
Traditional Clinicopathological Parameters
These include patient comorbidities, age, and primary tumor characteristics like
tumor size, tumor grade, nodal status, ER status, HER2/neu status, and lymphovascular
invasion. Many of these parameters are incorporated in various
prognostic tools like the Nottingham Prognostic Index (40), Adjuvant! Online
(41), and the NIH (13) and St. Gallen (42) Guidelines.

206 Ignatiadis and Sotiriou
Primary Tumor Gene Expression Signatures
From the above discussion, it is apparent that two genomic predictors are in the
late phase of clinical validation. However, apart from the question of who can
be spared unnecessary adjuvant chemotherapy (prognosis), a more important
question is which therapy works better in an individual patient (prediction).
Genomic predictors of response to multidrug regimens and newer targeted agents
have also been reported (43–49), but their description is beyond the scope of this
chapter. Other important technology platforms are being developed to analyze
epigenetic changes in DNA (50), the role of microRNAs (51) that will provide us
with a more comprehensive picture of the biology of the tumor and further refine
prognosis and prediction.
Micrometastatic Cells
The presence of cytokeratin positive tumor cells in the bone marrow of early breast
cancer patients has been correlated with worse outcome in a meta-analysis
involving more than 4500 patients (52). We have reported that the detection with a
real-time RT-PCR assay of Cytokeratin-19 (CK19) mRNA-positive circulating
tumor cells (CTCs) in patients with early breast cancer before (53,54) and after
(55,56) the initiation of adjuvant systemic treatment is associated with poor outcome.
However, the additional prognostic information of micrometastatic cells to
primary tumor gene expression profiling data has not yet been examined. Moreover,
there are pilot studies where HER2-positive CTCs and/or disseminated tumor cells
(DTCs) were successfully eliminated with trastuzumab (57), suggesting that CTCs
and/or DTCs may be tested also as tools to predict response to a particular drug.
Molecular Imaging
PET (positron emission tomography) scan using novel PET tracers that target
critical cellular processes like angiogenesis may provide important staging and
functional information and is under investigation in breast cancer (58).
Pharmacogenetics (Host)
The importance of pharmacogenetics in predicting clinical outcome in breast
cancer has recently been illustrated in studies focused on the polymorphism of
enzymes that metabolize tamoxifen (59,60). The study of the host is an important
piece of the clinical decision-making algorithm.
Proteomics
Proteomics provides the hope of discovering novel biological markers that can
be used for early detection, disease diagnosis, prognostication, and prediction of
response to therapy in early breast cancer (61).

Taking Prognostic Signatures to the Clinic 207
Adjuvant Treatment
It is well documented that adjuvant treatment affects clinical outcome of breast
cancer patients (62). From that perspective, the positive adjuvant trials with the
aromatase inhibitors (63) and trastuzumab (64,65), the concept of dose-dense
adjuvant chemotherapy (66), and the emerging data about the use of taxanes
(67,68) and anthracyclines (69) in the adjuvant setting are rendering both the
treatment decision-making process and prediction of outcome increasingly more
complex.
Several efforts have been made to combine at least some of the above
parameters to predict the clinical outcome of breast cancer patients. For instance,
Pittman et al. have used a modeling approach based on classification trees to
combine clinical and genomic data (70). They nicely showed that by combining
multiple metagenes (defined as summary measures of the expression of subsets
of potentially related genes) with nodal status, they could create integrative
clinicogenomic models. These models provided improved prediction accuracy
compared with either clinical or genomic data alone. Another example of such an
effort is the Adjuvant! Online for breast cancer, Genomic Version 7.0, which
combines traditional clinical data, the RS, and information about adjuvant
treatment to provide an estimate of clinical outcome in patients with early breast
cancer (71).
However, the accumulation of highly complex and multilevel information
will ultimately impose specific data integration requirements. So far these different
types of data are not only maintained in a variety of different data sources
but are also managed by different complex data management and analytical
systems. Additionally, they involve many institutions and numerous complex
workflows. To address all these critical issues, two initiatives have been
launched that intend to integrate all this information into one comprehensive
biomedical platform—the cancer Biomedical Informatics Grid, “caBIG” (https://
cabig.nci.nih.gov/) and the Advancing Clinico Genomic Trials, “ACGT” (http://
www.eu-acgt.org/) funded by the NCI and European Union, respectively.
In the future, medical oncologists should use a clinically validated treatment
decision-making algorithm that will take into account all this highly
complex and rapidly evolving, multilevel information. In this way, we will move
from the “empirical” to the new exciting era of “individual-tailored” oncology.
REFERENCES
1. Sotiriou C, Wirapati P, Kunkel S, et al. Meta-analysis of gene-expression profiles in
breast cancer: towards a unified understanding of breast cancer sub-typing and
prognosis signatures. Breast Cancer Res Treat. 2007; 106(suppl 1):1047(abstr).
2. Ignatiadis M, Desmedt C. Predicting risk of breast cancer recurrence using geneexpression
profiling. Pharmacogenomics 2007; 8(1):101–111.
3. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours.
Nature 2000; 406(6797):747–752.

208 Ignatiadis and Sotiriou
4. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas
distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci
U S A 2001; 98(19):10869–10874.
5. Sorlie T, Tibshirani R, Parker J, et al. Repeated observation of breast tumor subtypes
in independent gene expression data sets. Proc Natl Acad Sci U S A 2003; 100(14):
8418–8423.
6. Hu Z, Fan C, Oh DS, et al. The molecular portraits of breast tumors are conserved
across microarray platforms. BMC Genomics 2006; 7:96.
7. Sotiriou C, Neo SY, McShane LM, et al. Breast cancer classification and prognosis
based on gene expression profiles from a population-based study. Proc Natl Acad Sci
U S A 2003; 100(18):10393–10398.
8. Andre F, Pusztai L. Molecular classification of breast cancer: implications for
selection of adjuvant chemotherapy. Nat Clin Pract Oncol 2006; 3(11):621–632.
9. van ’t Veer LJ, Dai H, van de Vijver MJ, et al. Gene expression profiling predicts
clinical outcome of breast cancer. Nature 2002; 415(6871):530–536.
10. van de Vijver MJ, He YD, van ’t Veer LJ, et al. A gene-expression signature as a
predictor of survival in breast cancer. N Engl J Med 2002; 347(25):1999–2009.
11. Wang Y, Klijn JG, Zhang Y, et al. Gene-expression profiles to predict distant
metastasis of lymph-node-negative primary breast cancer. Lancet 2005; 365(9460):
671–679.
12. Foekens JA, Atkins D, Zhang Y, et al. Multicenter validation of a gene expressionbased
prognostic signature in lymph node-negative primary breast cancer. J Clin
Oncol 2006; 24(11):1665–1671.
13. Eifel P, Axelson JA, Costa J, et al. National Institutes of Health Consensus Development
Conference Statement: adjuvant therapy for breast cancer, November 1–3,
2000. J Natl Cancer Inst 2001; 93(13):979–989.
14. Goldhirsch A, Wood WC, Gelber RD, et al. Meeting highlights: updated international
expert consensus on the primary therapy of early breast cancer. J Clin Oncol
2003; 21(17):3357–3365.
15. Buyse M, Loi S, van ’t Veer LJ, et al. Validation and clinical utility of a 70-gene
prognostic signature for women with node-negative breast cancer. J Natl Cancer Inst
2006; 98(17):1183–1192.
16. Desmedt C, Piette F, Loi S, et al. Strong time-dependency of the 76-gene prognostic
signature for node-negative breast cancer patients in the TRANSBIG multi-centre
independent validation series. Clin Cancer Res 2007; 13:3207–3214.
17. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifentreated,
node-negative breast cancer. N Engl J Med 2004; 351(27):2817–2826.
18. Chang HY, Sneddon JB, Alizadeh AA, et al. Gene expression signature of fibroblast
serum response predicts human cancer progression: similarities between tumors and
wounds. PLoS Biol 2004; 2(2):E7.
19. Chang HY, Nuyten DS, Sneddon JB, et al. Robustness, scalability, and integration of
a wound-response gene expression signature in predicting breast cancer survival.
Proc Natl Acad Sci U S A 2005; 102(10):3738–3743.
20. Sotiriou C, Wirapati P, Loi S, et al. Gene expression profiling in breast cancer:
understanding the molecular basis of histologic grade to improve prognosis. J Natl
Cancer Inst 2006; 98(4):262–272.

Taking Prognostic Signatures to the Clinic 209
21. Loi S, Haibe-Kains B, Desmedt C, et al. Definition of clinically distinct molecular
subtypes in estrogen receptor-positive breast carcinomas through genomic grade.
J Clin Oncol 2007; 25(10):1239–1246.
22. Oh DS, Troester MA, Usary J, et al. Estrogen-regulated genes predict survival in
hormone receptor-positive breast cancers. J Clin Oncol 2006; 24(11):1656–1664.
23. Ma XJ, Wang Z, Ryan PD, et al. A two-gene expression ratio predicts clinical
outcome in breast cancer patients treated with tamoxifen. Cancer Cell 2004; 5(6):
607–616.
24. Miller LD, Smeds J, George J, et al. An expression signature for p53 status in human
breast cancer predicts mutation status, transcriptional effects, and patient survival.
Proc Natl Acad Sci U S A 2005; 102(38):13550–13555.
25. Glinsky GV, Berezovska O, Glinskii AB. Microarray analysis identifies a deathfrom-cancer
signature predicting therapy failure in patients with multiple types of
cancer. J Clin Invest 2005; 115(6):1503–1521.
26. Liu R, Wang X, Chen GY, et al. The prognostic role of a gene signature from
tumorigenic breast-cancer cells. N Engl J Med 2007; 356(3):217–226.
27. Fan C, Oh DS, Wessels L, et al. Concordance among gene-expression-based predictors
for breast cancer. N Engl J Med 2006; 355(6):560–569.
28. Calza S, Hall P, Auer G, et al. Intrinsic molecular signature of breast cancer in a
population-based cohort of 412 patients. Breast Cancer Res 2006; 8(4):R34.
29. Pawitan Y, Bjohle J, Amler L, et al. Gene expression profiling spares early breast
cancer patients from adjuvant therapy: derived and validated in two population-based
cohorts. Breast Cancer Res 2005; 7(6):R953–R964.
30. Huang E, Cheng SH, Dressman H, et al. Gene expression predictors of breast cancer
outcomes. Lancet 2003; 361(9369):1590–1596.
31. Huang E, Ishida S, Pittman J, et al. Gene expression phenotypic models that predict
the activity of oncogenic pathways. Nat Genet 2003; 34(2):226–230.
32. Korkola JE, DeVries S, Fridlyand J, et al. Differentiation of lobular versus ductal breast
carcinomas by expression microarray analysis. Cancer Res 2003; 63(21): 7167–7175.
33. Naderi A, Teschendorff AE, Barbosa-Morais NL, et al. A gene-expression signature
to predict survival in breast cancer across independent data sets. Oncogene 2007;
26(10):1507–1516.
34. Farmer P, Bonnefoi H, Becette V, et al. Identification of molecular apocrine breast
tumours by microarray analysis. Oncogene 2005; 24(29):4660–4671.
35. Richardson AL, Wang ZC, De NA, et al. X chromosomal abnormalities in basal-like
human breast cancer. Cancer Cell 2006; 9(2):121–132.
36. Espinosa E, Vara JA, Redondo A, et al. Breast cancer prognosis determined by gene
expression profiling: a quantitative reverse transcriptase polymerase chain reaction
study. J Clin Oncol 2005; 23(29):7278–7285.
37. Paik S, Tang G, Shak S, et al. Gene expression and benefit of chemotherapy in
women with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol
2006; 24(23):3726–3734.
38. Teschendorff AE, Naderi A, Barbosa-Morais NL, et al. A consensus prognostic gene
expression classifier for ER positive breast cancer. Genome Biol 2006; 7(10):R101.
39. Whitfield ML, George LK, Grant GD, et al. Common markers of proliferation. Nat
Rev Cancer 2006; 6(2):99–106.

210 Ignatiadis and Sotiriou
40. Galea MH, Blamey RW, Elston CE, et al. The Nottingham Prognostic Index in
primary breast cancer. Breast Cancer Res Treat 1992; 22(3):207–219.
41. Olivotto IA, Bajdik CD, Ravdin PM, et al. Population-based validation of the
prognostic model ADJUVANT! for early breast cancer. J Clin Oncol 2005; 23(12):
2716–2725.
42. Goldhirsch A, Glick JH, Gelber RD, et al. Meeting highlights: international expert
consensus on the primary therapy of early breast cancer 2005. Ann Oncol 2005;
16(10):1569–1583.
43. Gianni L, Zambetti M, Clark K, et al. Gene expression profiles in paraffin-embedded
core biopsy tissue predict response to chemotherapy in women with locally advanced
breast cancer. J Clin Oncol 2005; 23(29):7265–7277.
44. Chang JC, Wooten EC, Tsimelzon A, et al. Patterns of resistance and incomplete
response to docetaxel by gene expression profiling in breast cancer patients. J Clin
Oncol 2005; 23(6):1169–1177.
45. Jansen MP, Foekens JA, van Staveren IL, et al. Molecular classification of tamoxifen-resistant
breast carcinomas by gene expression profiling. J Clin Oncol 2005;
23(4):732–740.
46. Hess KR, Anderson K, Symmans WF, et al. Pharmacogenomic predictor of sensitivity
to preoperative chemotherapy with paclitaxel and fluorouracil, doxorubicin,
and cyclophosphamide in breast cancer. J Clin Oncol 2006; 24(26):4236–4244.
47. Bild AH, Yao G, Chang JT, et al. Oncogenic pathway signatures in human cancers as
a guide to targeted therapies. Nature 2006; 439(7074):353–357.
48. Potti A, Dressman HK, Bild A, et al. Genomic signatures to guide the use of chemotherapeutics.
Nat Med 2006; 12(11):1294–1300.
49. Solit DB, Garraway LA, Pratilas CA, et al. BRAF mutation predicts sensitivity to
MEK inhibition. Nature 2006; 439(7074):358–362.
50. Herman JG, Baylin SB. Gene silencing in cancer in association with promoter
hypermethylation. N Engl J Med 2003; 349(21):2042–2054.
51. Iorio MV, Ferracin M, Liu CG, et al. MicroRNA gene expression deregulation in
human breast cancer. Cancer Res 2005; 65(16):7065–7070.
52. Braun S, Vogl FD, Naume B, et al. A pooled analysis of bone marrow micrometastasis
in breast cancer. N Engl J Med 2005; 353(8):793–802.
53. Xenidis N, Perraki M, Kafousi M, et al. Predictive and prognostic value of peripheral
blood cytokeratin-19 mRNA-positive cells detected by real-time polymerase chain
reaction in node-negative breast cancer patients. J Clin Oncol 2006; 24(23):3756–3762.
54. Ignatiadis M, Xenidis N, Perraki M, et al. Different prognostic value of Cytokeratin-
19 mRNA-positive Circulating Tumor Cells according to estrogen receptor status in
early breast cancer. J Clin Oncol 2007; 25(33):5194–5202.
55. Xenidis N, Vlachonikolis I, Mavroudis D, et al. Peripheral blood circulating cytokeratin-19
mRNA-positive cells after the completion of adjuvant chemotherapy in
patients with operable breast cancer. Ann Oncol 2003; 14(6):849–855.
56. Xenidis N, Mavroudis D, Apostolaki S, et al. Effect of adjuvant treatment on the
circulating CK19mRNA positive tumor cells in patients with early stage breast
cancer. Proc Am Soc Clin Oncol 2006; 24(No18S):20012 (abstr).
57. Bozionellou V, Mavroudis D, Perraki M, et al. Trastuzumab administration can
effectively target chemotherapy-resistant cytokeratin-19 messenger RNA-positive
tumor cells in the peripheral blood and bone marrow of patients with breast cancer.
Clin Cancer Res 2004; 10(24):8185–8194.

Taking Prognostic Signatures to the Clinic 211
58. Quon A, Gambhir SS. FDG-PET and beyond: molecular breast cancer imaging.
J Clin Oncol 2005; 23(8):1664–1673.
59. Goetz MP, Rae JM, Suman VJ, et al. Pharmacogenetics of tamoxifen biotransformation
is associated with clinical outcomes of efficacy and hot flashes. J Clin
Oncol 2005; 23(36):9312–9318.
60. Nowell S, Sweeney C, Winters M, et al. Association between sulfotransferase 1A1
genotype and survival of breast cancer patients receiving tamoxifen therapy. J Natl
Cancer Inst 2002; 94(21):1635–1640.
61. Somiari RI, Somiari S, Russell S, Shriver CD. Proteomics of breast carcinoma.
J Chromatogr B Analyt Technol Biomed Life Sci 2005; 815(1–2):215–225.
62. Early Breast Cancer Trialists’ Collaborative Group (EBCTCG). Effects of chemotherapy
and hormonal therapy for early breast cancer on recurrence and 15-year
survival: an overview of the randomised trials. Lancet 2005; 365(9472):1687–1717.
63. Winer EP, Hudis C, Burstein HJ, et al. American Society of Clinical Oncology
technology assessment on the use of aromatase inhibitors as adjuvant therapy for
postmenopausal women with hormone receptor-positive breast cancer: status report
2004. J Clin Oncol 2005; 23(3):619–629.
64. Piccart-Gebhart MJ, Procter M, Leyland-Jones B, et al. Trastuzumab after adjuvant
chemotherapy in HER2-positive breast cancer. N Engl J Med 2005; 353(16):1659–1672.
65. Romond EH, Perez EA, Bryant J, et al. Trastuzumab plus adjuvant chemotherapy for
operable HER2-positive breast cancer. N Engl J Med 2005; 353(16):1673–1684.
66. Citron ML, Berry DA, Cirrincione C, et al. Randomized trial of dose-dense versus
conventionally scheduled and sequential versus concurrent combination chemotherapy
as postoperative adjuvant treatment of node-positive primary breast cancer:
first report of Intergroup Trial C9741/Cancer and Leukemia Group B Trial 9741.
J Clin Oncol 2003; 21(8):1431–1439.
67. Mamounas EP, Bryant J, Lembersky B, et al. Paclitaxel after doxorubicin plus
cyclophosphamide as adjuvant chemotherapy for node-positive breast cancer: results
from NSABP B-28. J Clin Oncol 2005; 23(16):3686–3696.
68. Martin M, Pienkowski T, Mackey J, et al. Adjuvant docetaxel for node-positive
breast cancer. N Engl J Med 2005; 352(22):2302–2313.
69. Piccart-Gebhart MJ. Anthracyclines and the tailoring of treatment for early breast
cancer. N Engl J Med 2006; 354(20):2177–2179.
70. Pittman J, Huang E, Dressman H, et al. Integrated modeling of clinical and gene
expression information for personalized prediction of disease outcomes. Proc Natl
Acad Sci U S A 2004; 101(22):8431–8436.
71. Adjuvant! For Breast Cancer, Genomic Version 7.0. http://www.adjuvantonline.com/
breastgenomic.jsp 2006.

15
Tissue is the Issue
Adekunle Raji
Eastern Cooperative Oncology Group Pathology Coordinating Office,
Northwestern University, Evanston, Illinois, U.S.A.
INTRODUCTION
All eyes are on tissue. The biggest challenge facing biomedical discoveries today
is the issue with tissue. The era of personalized, individually tailored medicine has
begun; therefore, cells, tissue and its intra- and interenvironment, are the focus of
pharmacogenetics. The intracellular kinetic and intercellular dynamics are key
factors in the understanding of cell behavior before, during, and after therapy.
Undoubtedly the solid type of tissue carries the bulk of these issues. Other
fluid forms of tissue, however, do have their own share of these issues, of course.
Tissue will be defined here as biological material comprising a network of
characteristically diverse cells. In essence, blood, urine, surgical resections,
biopsies, marrow, lavage, hair, nail, and other specimen taken from a person will
be considered as tissue. Turn to any biomedical community member and most, if
not all, will have some sort of opinion about tissue. Between the ultimate giver
(patients) and ultimate user (laboratorians) lie the challenges. To the patient, care
team, and several entities, the tissue is the issue, and to the laboratories, technologists,
and the interpreters of results, the issue is really the tissue.
In the past two years, several working groups were established not only to
understand current issues as well as future anticipated issues but also to address
them. These working groups are usually comprise various clinical and financial
institutions and experts from multiple disciplines, such as medical scientists,
basic scientists, surgeons, laboratories, pathologists, funding agencies, agent
213

214 Raji
manufacturers, patient advocates, statisticians, bioinformatics, research associates,
research administrators, medical staff, and clinical administrators. These
groups guide the interdisciplinary patient care team by addressing the procurement,
collection, preservation, acquisition, processing, transportation, and
storage of biological materials of high quality and usability that are adequate
for analysis and interpretation. Good news is that specimens with clinical data
are archived as far back as 40 years ago, the bad news is that they were
collected non-uniformaly; today’s good news is that the collection processing
andassociateddataareundergoingharmonization (uniformity). Contingency
plans are also being developed for future anticipated issues. The key areas of
the tissue issue are obviously procurement, collection, acquisition, processing,
preservation, storage, distribution, and bioinformatics that are successfully
conducted within stringent policies, procedures, rules, and regulations.
PROCUREMENT
A strategy for collecting any biological specimens from patients is needed to
ensure that the behavior of cells or tissue after it leaves the body of the patient is
consistent with what it was while in the patient’s body. Regardless of the
strategy, clear and simple instructions should be written and made available to
the procurement team comprising a preoperation nurse, phlebotomist, surgery
assistant, physician/surgeon, physician assistant, pathology assistant, research
coordinator, patient care nurse, internal specimen transporters, and medical/
research technologist. These instructions are typically included in the local
manual of operational procedures and/or clinical protocols and should provide
details of the type of collection devices such as vacutainers, syringes, punch size,
preservatives, buffers, reagents and the containers needed. The manual must also
include procedures for preparing reagents or buffers such as formalin, RNAlater,
and the optimal cutting temperature (OCT) compound. It must contain information
about the precollection requirements with regard to time, temperature,
and pressure for certain types of specimens (Table 1)
COLLECTION
There are several choices of collection devices, to the extent that they have
created a great deal of confusion in the medical field. It is not surprising to see
several new clinical trial protocols now restricting collection to the use of
preassembled collection devices, with instructions to assist the procurement
plan and ensure uniform collection, particularly in a multi-institutional study.
While it is possible to obtain serum from redtop, royal blue top, serum separator
top, and plain syringe, it is true that tissue can be fixed in paraformaldhyde,
formalin, acid alcohol, zinc formalin, and many other reagents. Each of these
fixatives presents its own advantages and disadvantages toward downstream
applications such as immunohistochemistry (IHC), in situ hybridization (ISH),

Tissue is the Issue 215
Table 1
Roles
Roles and Responsibility
Responsibility
Physician/surgeon including
but not limited to pathologist
and intervention radiologist
Nurse
Phlebotomist, nurse, medical
technologist
Physician/surgery assistant
Transporter
Research coordinators, project
manager, protocol specialist,
data manager, nurse
coordinator, clinical
research associates
Medical/research technologist
Inform patient, order test, perform surgery/biopsy/
imaging/morphology/assay procedures.
Inform patient, determine timing of procedure/draw.
Draw blood via port.
Draw blood into devices. Perform aspiration.
Guide physician and expedite preservation and delivery.
Ensure timely delivery from point of collection to next
critical step/location while preserving samples.
Ensure all team members are familiar with their roles
and have adequate information to assist in
performing task. Liaison among team members as
well as facilitate collaboration.
Ensure availability of appropriate supplies for
collection and preservation. Effectively and
efficiently process specimens and analyze samples.
comparative genomic hybridization (CGH), polymerase chain reaction (PCR),
blotting (Western, Northern, Southern), enzyme-linked immunosorbent assay
(ELISA), radio immunoassay (RIA), enzyme immunoassay (EIA), and chemiluminescence
assay (CLA). It is also true that plasma can be obtained from all
anticoagulated vacutainers as well as heparin-coated syringes. Protein and
nucleic acids can be obtained from biological samples regardless of collection
or preservation buffer. However, experience has shown a great deal of variation
in the yield and quality of sample products, as well as the analytical results
obtained from the samples. It is therefore important to select one device and one
alternate backup for the collection, preservation, and fixation that will provide
the best representation of the cellular content and its content behavior as it was
just before leaving the patient’s body. It is important to select a device and
preservative that are best for yield and quality, as well as most appropriate for
the next step in processing or analysis. Table 2 lists the collection devices,
preservatives, and fixatives that have been selected by large research groups
because of their overall performance success.
Fresh tissue. Nonfrozen and nonfixed tissue should be collected directly
into a labeled sterile screw-cap plastic container, with one of the
following buffers:
1. Phosphate buffered saline (PBS) for tissue culture and cell viability
studies.
2. Neutral buffered formalin (10%) for gradual fixation.

216 Raji
Table 2 Common Blood Collection Tubes
Closure/stopper color Anticoagulant or additive
Number of
inversions to mix Examples for use
Red/red in glass tube None None Serum for chemistry, proteomics, immunology, and
banking
Red/red in plastic tube Clot activator 5–10 Serum for chemistry, proteomics, immunology, and
banking
Gold/red marbled SST Clot activator and separation gel 5–10 Serum for chemistry, proteomics, immunology, and
banking. Remote collection.
Lavender/lavender
(purple)
Sodium EDTA (dried), or
potassium EDTA (liquid)
Light blue/light blue Liquid 3.2% sodium citrate, a or
3.8% sodium citrate
Cell preparation
tube CPT
Liquid 3.2% sodium citrate, a or
3.8% sodium citrate. Ficol and
gel separator
Paxgene RNA/DNA Preanalytic/Qiagen nucleic acid
stabilizer
5–10 Whole blood for hematology, certain whole blood
chemistries, e.g., hemoglobin A1c or blood typing.
Proteomics, floating DNA, genomic DNA, cells.
5–10 Plasma for coagulation, Platelet poor plasma. Growth
factors, fragile proteins, viable cells.
5–10 Plasma for coagulation, Platelet poor plasma. Growth
5–10 Genomic expressions.
factors, fragile proteins, viable cells. Remote
collection of viable mononuclear cells.
P100 Protein stabilizer 5–10 Small size protein measurement. Proteomic studies.
Tan Sodium heparin (glass), or 5–10 Whole blood for lead, heavy metals, toxicology studies.
potassium EDTA (plastic)

Tissue is the Issue 217
Table 2 Common Blood Collection Tubes (Continued )
Closure/stopper color Anticoagulant or additive
Number of
inversions to mix Examples for use
Green/green Lithium heparin, or 5–10 Plasma or whole blood for chemistry, PK, and PD.
Sodium heparin 5–10 Plasma for chemistry
Light green/green
marbled
Lithium heparin and separation gel 5–10 Plasma for chemistry
Gray/gray Potassium oxalate/sodium fluoride,
or sodium fluoride, or lithium
iodoacetate, or lithium
iodoacetate/lithium heparin
5–10 Serum or plasma for glucose or alcohol requiring
inhibition of glycolysis, C-peptides.
Dark royal blue Sodium heparin, EDTA, or plain 5–10 Plasma or serum for trace elements, toxicology, and
nutrients.
Pale yellow/bright yellow Sodium polyanetholesulfonate 5–10 Blood cultures, microbiology, and infectious diseases
markers.
Yellow Acid citrate dextrose (ACD) 5–10 Blood bank, viable cells alternate to CPT and CellSave
Orange/yellow and
marbled
Clot activator (thrombin) 5–10 Serum for chemistry, ELISA
Pink/pink Potassium EDTA 5–10 Molecular/viral load testing, blood bank, hematology.
a www.bd.com
Abbreviations: SST, serum separator top; PK, pharmacokinetic; PD, pharmacodynamic.

218 Raji
3. RPMI (Roswell Park Memorial Institute) media for tissue culture, cell
viability studies, protein extraction, and multiphoton studies.
4. RNAlater for RNA extraction.
5. OCT bottom and top prior to freezing.
PROCESSING
Generally, processing methods that will ensure optimal quality and yield should be
selected for the processing of specimens and samples into next product aliquot.
Blood. It is important to perform all centrifugation in refrigerated and
temperature-controlled chambers. This is to avoid unnecessary heat
incubation of samples while spinning. In some instances, it will be
necessarytospinthesampleataslowspeedfollowedbyhighspeedto
eliminate platelets. Selection of complex processes to obtain delicate
and fragile by-products is important. Aliquoting into the smallest
useable quantity to avoid freeze-thaw-freeze cycle is also necessary.
For example, serum and plasma can be apportioned into 200 mL, DNA
andRNAcanbeapportionedinto75mg and1mg, respectively.
Fresh frozen tissue. OCT-protected frozen tissue should be cryosectioned
for hematoxylin and eosin staining to assess tumor area and percentage
as part of the annotation. The information will guide case selection as
well as steps in macro and micro dissection. Integrity of the sample must
be maintained through sterile processes to avoid degradation and/or
contamination. The tissue must be kept frozen until submersed into
extraction buffer. There are many choices of methods for the extraction
of cellular components such as RNA, DNA, and proteins.
Fixed tissue. Tissue biopsies and/or resection must undergo complete and
adequate fixation in a buffered environment. This is important for the
morphologic review as well as the evaluation of reference and novel
markers. Neutral buffered formalin has been widely used and is most
appropriate for current and future translational processes. As previously
mentioned, there are many choices of methods for the extraction and
isolation of DNA; however, there are only few methods for the
extraction of RNA and protein from fixed tissue. Care must be taken
in making selections to avoid loss of valuable tissue resource.
Sectioning should be performed with new blades and clean water.
Nonprecipitating formalin should be avoided.
Xylene rather than its substitutes should be used.
Manual rather than automated processes should be used for small,
starting samples.
Washes between steps in any and all processes must be complete to
avoid reagent carry over.

Tissue is the Issue 219
CURRENT NOVEL PROCESSING OF BLOOD AND FIXED TISSUE
Isolation of Mononuclear Cells
Monocytes, a type of white blood cells, are found both in bone marrow and blood.
The following procedure describes the isolation of monocytes in peripheral blood
(peripheral blood mononuclear cells, PBMC).
Steps
1. Centrifuge CPT tubes containing whole blood for 15 minutes at 258C at
2700 rpm.
2. Pipet out the top layer (plasma) and transfer into two labeled tubes and
freeze.
3. Pipet out the lymphocyte layer (the cloudy layer) with a transfer pipette,
placing the layer into a fresh 15-mL tube. Fill tube with room-temperature
PBS.
4. Centrifuge for 10 minutes at 1800 rpm.
5. Discard supernatant and resuspend pellet in 1 mL PBS.
6. Fill tube with PBS, inverting eight times to mix and wash.
7. Centrifuge for 15 minutes at 2000 rpm.
8. Discard supernatant.
9. Proceed to cell sorting or resuspend and freeze pellet in 1 mL of RPMI or
DMSO (dimethyl sulfoxide) for later sort.
Note:
For blood collected in any other anticoagulated tube, dilute 1:1 with
RPMI, layer on Histopaque 1077, then proceed with step 1.
Specialized Sorting of Circulating Tumor/Endothelial Cells
1. Magnetic labeling (*45 min)
a. Centrifuge sample for 10 minutes at 300g at room temperature.
b. Carefully remove the supernatant completely.
c. Resuspend cell pellet in appropriate amount of buffer as specified by
the microbeads package insert. For fewer cells, use the same volume.
d. Add microbeads per package insert.
e. Incubate for 15 minutes at 68Cto128C. For less than 5 10 6 total cells,
use the same volume.
f. Wash cells by adding 10 to 20 times the labeling volume of buffer and
centrifuge at 300g for 10 minutes.
g. Remove supernatant completely (save the supernatant in appropriately
labeled 1 mL cryovial) and resuspend cell pellet in appropriate amount
of buffer (1 mL of buffer/10 8 total cells).
h. Proceed to magnetic separation.

220 Raji
2. Magnetic separation (*60 min)
a. Turn on the manual and/or automated cell sorter.
b. When the machine has finished initializing, perform a column exchange
to remove blank columns and insert magnetic columns. (To save
time, try to perform these functions while the sample and beads
are incubating.)
c. Wipe down probes with alcohol to disinfect.
d. Place the sample in the uptake position, and place empty 15-mL conical
tubes in the rack under the Neg1, Pos1, and Pos2 probes.
e. Choose “Separation” on the main menu.
f. Use the down arrows to select “possel_s” and press “start.”
g. For both the positive and negative cells, separate elution container
must be used (*15 min).
h. Prepare and/or submit samples for next step in processing (DNA extraction,
thin prep slides, etc.).
Extraction of RNA from Peripheral Blood in PAXgene
1. Centrifuge the PAXgene Blood RNA Tube for 10 minutes at 3000g to
5000g using a swing-out rotor.
2. Remove the supernatant by decanting or pipetting. Add 4 mL RNase-free
water to the pellet, and close the tube using a fresh secondary Hemogard
closure.
3. Vortex until the pellet is visibly dissolved, and centrifuge for 10 minutes at
3000g to 5000g using a swing-out rotor. Remove and discard the entire
supernatant.
4. Add 350 mL Buffer BR1, and vortex until the pellet is visibly dissolved.
5. Pipet the sample into a 1.5-mL microcentrifuge tube. Add 300 mL Buffer
BR2 and 40 mL proteinase K. Mix by vortexing for 5 seconds, and incubate
for 10 minutes at 558C using a shaker/incubator at 400 to 1400 rpm.
After incubation, set the temperature of the shaker/incubator to 658C (for
step 20).
6. Pipet the lysate directly into a PAXgene Shredder spin column (lilac)
placed in a 2-mL processing tube, and centrifuge for 3 minutes at maximum
speed (but not to exceed 20,000g).
7. Carefully transfer the entire supernatant of the flow-through fraction to a
fresh 1.5-mL microcentrifuge tube without disturbing the pellet in the
processing tube.
8. Add 350 mL ethanol (96–100%, purity grade p.a.). Mix by vortexing, and
centrifuge briefly (1–2 seconds at 500–1000g) to remove drops from the
inside of the tube lid.
9. Pipet 700 mL sample into the PAXgene RNA spin column (red) placed in a
2-mL processing tube and centrifuge for 1 minute at 8000 to 20,000g.

Tissue is the Issue 221
Place the spin column in a new 2-mL processing tube, and discard the old
processing tube containing flow-through.
10. Pipet the remaining sample into the PAXgene RNA spin column, and
centrifuge for 1 minute at 8000 to 20,000g. Place the spin column in a
new 2-mL processing tube, and discard the old processing tube containing
flow-through.
11. Pipet 350 mL Buffer BR3 into the PAXgene RNA spin column. Centrifuge
for 1 minute at 8000 to 20,000g. Place the spin column in a new 2-mL
processing tube and discard the old processing tube containing flow-through.
12. Add 10 mL DNase I stock solution to 70 mL Buffer RDD in a 1.5-mL
microcentrifuge tube. Mix by gently “flicking” the tube and centrifuge
briefly to collect residual liquid from the sides of the tube.
13. Pipet the DNase I incubation mix (80 mL) directly onto the PAXgene RNA
spin column membrane, and place on the bench top (20–308C.) for
15 minutes.
14. Pipet 350 mL Buffer BR3 into the PAXgene RNA spin column, and
centrifuge for 1 minute at 8000g to 20,000g. Place the spin column in a
new 2-mL processing tube and discard the old processing tube containing
flow-through.
15. Pipet 500 mL Buffer BR4 to the PAXgene RNA spin column, and
centrifuge for 1 minute at 8000g to 20,000g. Place the spin column in a
new 2-mL processing tube, and discard the old processing tube containing
flow-through.
16. Add another 500 mL Buffer BR4 to the PAXgene RNA spin column.
Centrifuge for 3 minutes at 8000g to 20,000g.
17. Discard the tube containing the flow-through, and place the PAXgene
RNA spin column in a new 2-mL processing tube. Centrifuge for 1 minute
at 8000g to 20,000g.
18. Discard the tube containing the flow-through. Place the PAXgene RNA
spin column in a 1.5-mL microcentrifuge tube, and pipet 20 mL Buffer
RNAse-free water directly onto the PAXgene RNA spin column membrane.
Centrifuge for 1 minute at 8000g to 20,000g to elute the RNA.
19. Repeat the elution step by adding the eluted RNA to the same PAXgene
RNA spin column membrane. Centrifuge for 1 minute at 8000g to 20,000g
to elute the RNA.
20. Incubate the eluate for 5 minutes at 658C. in the shaker/incubator without
shaking. After incubation, chill immediately on ice.
21. If the RNA samples will not be used immediately, store at 208Cor 708C.
Laser Capture Microdissection
Laser capture microdissection (LCM) allows for the microscopic removal of
specific cells from paraffin or frozen tissue sections mounted on glass slides.

222 Raji
This procedure describes the isolation of a pure population of cells followed by
its nucleic acid or protein extraction.
1.1. Prepare paraffin section on glass slide.
a. Collect 5 mm paraffin section onto uncoated glass slide, using microtome
and water bath.
b. Allow to air-dry overnight at room temperature or at 428C for up to
8 hours.
c. Deparaffinize and lightly stain slide(s) by immersing in the following:
i. Xylene 2: 5 minutes each
ii. 100% ethanol: 1 minute
iii. 95% ethanol: 1 minute
iv. 70% ethanol: 1 minute
v. Distilled water: 1 minute
vi. Harris hematoxylin: 1 dip (Fisher’s Protocol #245–678)
vii. Tap water: 2 minutes
viii. Distilled water: 1 minute
ix. 95% ethanol: 1 minute
x. Eosin Y: 1 dip (Fisher’s Protocol #314–631)
xi. 95% ethanol: 1 minute
xii. Allow to air-dry
or
1.2. Prepare frozen section on glass slide.
a. Collect 8 mm frozen section on an uncoated glass slide, using cryostat.
b. Store frozen until ready to postfix.
c. Postfix and lightly stain.
d. Remove a maximum of four slides from cryostat or freezer, place on
clean paper towel, i.e., KimWipe, and allow to thaw for approximately
30 seconds.
e. Place the slides in 75% ethanol for 30 seconds.
f. Transfer the slides to distilled water for 30 seconds.
g. Place slides on KimWipe, apply 100 mL of histogene staining
solution, and stain for approximately 20 seconds.
h. Place the slides in distilled water for 30 seconds.
i. Place the slides for 30 seconds in each of the following ethanols in
increasing concentration: 75%, then 95%, and then 100%.
j. Transfer the slides to xylene for 5 minutes.
k. Place the slides on a KimWipe to dry in a ventilation hood for 5 minutes.
2. Dissect area of tissue from slide (to be performed same day as steps 1.1
or 1.2).
a. Place slide on Pixcell II LCM [Arcturus (now part of Molecular
Devices), Mountain View, California, U.S.] and locate area of
interest using 4 or 10 objective.

Tissue is the Issue 223
b. Load CapSure HS caps.
c. Turn on laser key, push laser enable button, and locate laser spot on
screen.
d. Focus laser beam to bright, well-defined spot.
e. Test laser, observing melted plastic ring(s).
See following table for power and duration.
Laser spot size (mm) Power (mW) Duration
CapSure HS Small (7.5) 65–75 650–750 ms
Medium (15) 35–45 2.5–3.0 ms
Large (30) 45–55 6.0–7.0 ms
f. View captured cells on cap.
g. Move cap to capping station.
h. Transfer to ExtractSure device from CapSure HS cap.
3. Extract DNA from dissected tissue.
a. Pipet 155 mL of reconstitution buffer into one vial of proteinase K.
Gently vortex the tube to mix the reagents and immediately place
the tube on ice. Completely dissolve reagent, excessive mixing can
denature proteinase K.
b. Assemble the ExtractSure Sample Extraction Device onto the CapSure
HS LCM Cap.
c. Using clean forceps, remove the cap from the cap insertion tool and
place into the alignment tray (sample facing up).
d. Make sure that the cap snaps securely and lies flat on the bottom of the
opening of the alignment tray.
e. Position the ExtractSure Device over the cap using clean forceps. The
fill-port should be facing up.
f. Using forceps, push the ExtractSure Device down onto the cap until it
snaps securely into place. Use forceps to ensure that the cap is firmly
attached to the ExtractSure Device.
g. Pipet 10 mL of proteinase K extraction solution into the microchamber
formed by the assembled ExtractSure Device and CapSure HS LCM
Cap.
h. Using gloved hands, place a 0.5-mL thin-walled reaction tube over the
ExtractSure device.
i. Cover the assembly in the alignment tray with the incubation block and
incubate it overnight at 658C.
j. After incubation, remove the assembled CapSure HS LCM Cap and
ExtractSure Device from the incubator, place the assembly into a
microcentrifuge and centrifuge for 1 minute at 4000g.

224 Raji
k. Separate the CapSure HS LCM Cap and ExtractSure Device. Close the
microcentrifuge tube containing the extract and heat to 958C for
10 minutes in a heating block to inactivate the proteinase K. Cool
the sample to room temperature.
Isolation of DNA, RNA, Protein Using TRIZOL Method
Isolation of RNA by TRI Reagent
Homogenization Homogenize fresh tissue samples (that have been stored in
RNAlater) in TRI Reagent (1 mL/50–100 mg tissue) using a glass-Teflon or
Polytron homogenizer. Sample volume should not exceed 10% of the volume of
TRI Reagent used for homogenization.
Phase separation
1. Store the homogenate for 5 minutes at room temperature to permit the
complete dissociation of nucleoprotein complexes.
2. Add 0.2 mL chloroform to the homogenate, cover the samples tightly, and
shake vigorously for 15 seconds.
3. Store the resulting mixture at room temperature for 2 to 15 minutes and
centrifuge at 12,000g for 15 minutes at 48C.
4. Following centrifugation, the mixture separates into a lower red phenolchloroform
phase, interphase, and the colorless upper aqueous phase. RNA
remains exclusively in the aqueous phase, whereas DNA and proteins are
in the interphase and organic phase. The volume of the aqueous phase is
about 60% of the volume of TRI Reagent used for homogenization.
RNA precipitation
1. Transfer the aqueous phase to a fresh tube.
2. Precipitate RNA from the aqueous phase by mixing with 0.5 mL of
isopropanol.
3. Store samples at room temperature for 5 to 10 minutes and centrifuge at
12,000g for eight minutes at 48C to258C. RNA precipitate (often invisible
before centrifugation) forms a gel-like or white pellet on the side and
bottom of the tube.
RNA wash
1. Remove the supernatant and wash the RNA pellet (by vortexing) with 75%
ethanol (1 mL).
2. Centrifuge at 7500g for 5 minutes at 48C to258C.

Tissue is the Issue 225
RNA solubilization
1. Remove the ethanol wash and briefly air-dry the RNA pellet for 3 to 5 minutes.
2. Dissolve RNA water by passing the solution a few times through a pipette
tip and incubating for 10 to 15 minutes at 55 to 608C.
Isolation of DNA by TRI Reagent
The DNA is isolated from the interphase and phenol phase separated from the
initial homogenate as described in the RNA isolation protocol.
DNA precipitation
1. Remove the remaining aqueous phase overlying the interphase. Precipitate
DNA from the interphase and organic phase with ethanol.
2. Add 0.3 mL of 100% ethanol and mix samples by inversion.
3. Store the samples at room temperature for 2 to 3 minutes and sediment
DNA by centrifugation at 2000g for 5 minutes at 48C to258C.
DNA wash
1. Wash the DNA pellet twice in a solution containing 0.1 M trisodium citrate
in 10% ethanol (1 mL).
2. Store the DNA pellet in the washing solution for 30 minutes at room temperature
with periodic mixing and centrifuge at 2000g for 5 minutes at 48Cto258C.
3. Suspend the DNA pellet in 75% ethanol (1.5–2 mL of 75% ethanol). Store
for 10 to 20 minutes at room temperature with periodic mixing and centrifuge
at 2000g for 5 minutes at 48C to258C. This ethanol wash removes
pinkish color from the DNA pellet.
DNA solubilization
1. Remove the ethanol wash and briefly air-dry the DNA pellet by keeping
tubes open for 3 to 5 minutes at room temperature.
2. Dissolve the DNA pellet in 8 mM NaOH by slowly passing through a
pipette. Add (0.3 mL) of 8 mM NaOH.
3. Centrifuge at 12,000g for 10 minutes and transfer the resulting supernatant
containing DNA to new tubes.
Isolation of Proteins by TRI Reagent
Protein precipitation
1. Aliquot a portion of the phenol-ethanol supernatant (0.2–0.5 mL, 1 volume)
into a microfuge tube.

226 Raji
2. Precipitate proteins by adding 3 volumes of acetone.
3. Mix by inversion for 10 to 15 seconds to obtain a homogeneous solution.
4. Store samples for 10 minutes at room temperature and sediment the protein
precipitate at 12,000g for 10 minutes at 48C.
Protein wash
1. Decant the phenol-ethanol supernatant and disperse the protein pellet in
0.5 mL of 0.3 M guanidine hydrochloride in 95% ethanol þ 2.5 %
glycerol.
2. Disperse the pellet using a pipette tip. After dispersing the pellet, add
another 0.5 mL aliquot of the guanidine hydrochloride/ethanol/glycerol
wash solution to the sample and store for 10 minutes at RT.
3. Sediment the protein at 8000g for 5 minutes.
4. Decant the wash solution and perform two more washes in 1 mL each of
the guanidine/ethanol/glycerol wash solution. Disperse the pellet by vortexing
after each wash to efficiently remove residual phenol.
5. Wash in 1 mL of ethanol containing 2.5 % glycerol.
6. Sediment the protein at 8000g for 5 minutes.
7. Decant the alcohol, invert the tube, and dry the pellet for 7 to 10 minutes at
room temperature.
Protein solubilization
1. After briefly air-drying the protein pellet, add 1% SDS (0.2 mL) to the
protein pellet.
2. Gently disperse and solubilize the pellet for 15 to 20 minutes by flicking
the tube or pipetting as required.
Storage of RNA, DNA, and/or Protein
Store the solubilized proteins and DNA at 208C or colder, store RNA at 708C
or colder.
Tissue Microarray Creation
This procedure describes the use of the MTA-1 (Manual Tissue Arrayer) by
Beecher Instruments, Inc. (Sun Prairie, Wisconsin, U.S.) to create tissue microarray
(TMA) blocks. TMA is a technology which facilitates research by permitting
the collection of biological specimens representing a great many individuals
without requiring a great deal of storage space. TMA eases comparison of the
histological characteristics of different patients’ pathologies by placing many
different tissue samples together on the same microscope slide for observation.

Tissue is the Issue 227
Also, TMA helps prevent the unnecessary exhaustion of irreplaceable pathological
material, since it permits the same diagnostic and research objectives to be achieved
with far less tissue than was required for such objectives before TMA
existed.
A core of paraffin is removed from a “recipient” paraffin block (one
embedded without tissue) and the remaining empty space is filled with a core of
paraffin-embedded tissue from a “donor” block. This process can be repeated up
to 230 times for one TMA block, resulting in a block containing mapped pieces
from many different blocks, and when sectioned, produces a slide containing
mapped pieces from many different blocks. A corresponding map records each
core’s location in the grid and the information connected with it (patient ID
number, study arm, etc.). The array sections can be used for all histological
staining, including hematoxylin & eosin (H&E), IHC, and ISH.
1. Prepare the donor blocks and corresponding H&E slides.
a. Re-embed tissue, as needed, for more uniform wax consistency and
optimal tissue orientation.
b. Donor tissue should be at least 1.0 mm thick for best results.
c. Prepare a fresh H&E slide from each re-embedded potential donor block.
2. Pathologist characterizes each H&E slide for appropriateness of diagnosis
and percentage of tumor.
a. Examine each H&E slide microscopically and indicate, with marker,
the area of interest (AOI). This area corresponds to the area on the
block from which the core(s) will be taken.
3. Prepare the recipient (TMA) block.
a. Label a blank cassette, using the automated cassette labeler.
b. Determine the size of the TMA block and use:
M—the medium mold (30 mm 23 mm) or
L—the large mold (35 mm 23 mm)
c. Make a paraffin block, containing no tissue, using the labeled cassette.
d. Assure that the paraffin block surface is flat and parallel to the
underside of plastic cassette and that it contains no air bubbles.
4. Design the array.
a. Determine the diameter of the cores to be removed from the donor
blocks. Available core sizes include 0.6, 1.0, 1.5, and 2.0 mm. Core
size is often indicated by protocol, i.e., breast cancer studies require
the smallest core size—0.6 mm—to conserve tissue while removing
three cores from each donor block during the routine production
of three TMA blocks from each group of subjects.
b. Allow for enough space (3.0 mm) around the edge of the block to avoid
cracking paraffin.

228 Raji
c. Determine the number of columns and number of rows to use by noting
the following table, which describes the maximum number of cores per
TMA block, with respect to core size and mold size.
Maximum Number of Cores per TMA Block
Core size (mm)
Spacing between
cores (mm)
Mold size
Maximum no.
of cores/spaces
0.6 a yellow 1.5 LM 20 12 ¼ 240
1.0 a green 2.0 LM 15 9 ¼ 135
1.5 a black 2.5 LM 12 7 ¼ 84
2.0 a white 3.0 LM 10 6 ¼ 60
a Punches are color coded.
5. Create a map, identifying each core’s position on the grid its donor block.
a. Enter information into Excel Program (TMA Mapping Tool, Attachment
1) located on Server/Pathcore/TMA. TMA may be mapped using
alternate methods.
b. Include cores from tissue and cell line control blocks, in duplicate.
i. Tissue blocks, i.e., for breast protocols:
Fibroadenoma
Normal endometrium
Normal salivary gland
Normal testis
ii.
Cell line blocks, i.e., for breast protocols:
MCF7
SKBR3
T47D
c. Each TMA block will be asymmetric to assure map orientation, i.e.:
2nd column blank,
4th row blank, and
bottom right space blank
6. Construct the TMA.
a. Install the pair of tissue punches, for desired core size, in MTA-1. Install
the slightly larger of the pair (identified in blue) on the right (for the
donor block). Install the slightly smaller of the pair (identified in red) on
the left for the (for the recipient or TMA block).
b. Insert the recipient (TMA) paraffin block into the MTA-1 block holder.
Secure in place by tightening two screws in block holder with MTA-1
allen wrench.

Tissue is the Issue 229
c. Remove the first core from TMA block.
i. Assure that recipient punch is swung into place.
ii. Align punch over site of first core, usually top left corner, by
turning micrometers.
iii. Zero micrometers by pressing the X and Y “ZERO/ABS” buttons.
iv. Gently lower punch into TMA block, using care not to punch too
deeply. Maximum punch depth can be adjusted with large vertical
screw on the left.
v. Raise punch.
vi. Remove paraffin core from punch, using punch’s stylet.
d. Remove core from first donor block.
i. Place the donor block bridge over the TMA block holder.
ii. Swing donor punch into place.
iii. Manually align donor block so that AOI is directly under the
donor punch.
iv. Lower and raise punch.
v. Remove core of tissue from punch using punch stylet.
e. Place donor core into TMA block.
i. Remove donor block bridge from over the TMA block holder,
noting location of delicate loose core of tissue.
ii. Using forceps, align donor core over hole created for it in the
TMA block.
iii. Lower donor core into hole and assure flat level block surface by
pressing evenly with clean glass slide.
f. Adjust the micrometer(s) to move the precision guide to the next X/Y
position. Follow the Spacing Between Cores (mm) Column in the
maximum number of cores per TMA block table above to determine
the appropriate increments.
g. Repeat this cycle (steps 6c–6f) to construct the whole array.
h. To punch holes on right half of TMA block, rotate it 1808 and replace
in recipient block holder, because the x axis micrometer should not be
extended beyond 9 mm.
i. Smooth and level surface of TMA block.
i. Remove the TMA block from the recipient block holder.
ii. Place the TMA block in oven at 378C, maximum temperature, for
30 minutes.
iii. Gently press a glass slide on the top of the TMA block, applying
even pressure to push all tissue cores on the block to same level.

230 Raji
7. Section TMA block.
i. Cut a section from the TMA block and stain the slide with H&E for
QA/QC review by the pathologist.
ii. Store TMA block and save TMA map for any future sectioning requests.
Label side of block
2nd column left blank.
4th row left blank
Bottom right corner left blank

16
Acquisition and Preservation of Tissue
for Microarray Analysis
Brian R. Untch and John A. Olson
Institute for Genome Sciences and Policy and the Department of Surgery,
Duke University Medical Center, Durham, North Carolina, U.S.A.
INTRODUCTION
Increasingly, collection of normal and neoplastic human tissue is being incorporated
in the design of clinical trials and in routine clinical settings under tissue
banking programs. The goal behind these activities is either to facilitate identification
and development of robust predictive or prognostic biomarkers or to
enable basic investigation of tumor or host biology.
As experience in tissue banking has broadened, the role of rigorous collection
protocols, tissue handling, and annotation has become increasingly appreciated.
To this end, several groups have published guiding principles and standard
operating procedures (SOPs) designed to advance the fields of tissue banking,
biomarker development, and tumor biology (1,2).
This chapter is designed to summarize key points in the process of informed
consent, tissue collection, tissue preservation and storage, as well as extraction of
macromolecules of interest.
INFORMED CONSENT
Tissue banking requires informed consent by patients donating their tissues. The
informed consent document should include standard topics such as study design,
study purpose, and patient risks/benefits. Privacy rights are paramount for
231

232 Untch and Olson
patients when their health information is stored in databases, perhaps indefinitely.
The Health Insurance Portability and Accountability Act and the Common
Rule are federal regulations that require protection of patient identification.
Database information and patient identifiers (name, social security number,
institutional medical record ID) should remain separate. An arbitrary code system
should be created to uniquely label specimens and data. The informed
consent document should specify which investigator will keep the anonymous
identifiers and patient identifiers together. In the event that third-party sharing of
patient identifier data becomes necessary, written authorization must be obtained
from the patient. Also specified in the informed consent document is the
patients’ right to withdraw from the study or have their specimen removed from
the tissue databank. Contact information for these requests is provided to the
patient at the time of document signing.
A recent topic of debate has been whether giving consent for tissue
banking also allows for indiscriminate investigations related to and unrelated to
the initial study design. Implicit to the potential of tissue banking is the use of
specimens for areas of research that are unanticipated or currently unknown.
However, the use of specimens for research not specifically noted in the
informed consent document may violate religious and personal beliefs. The
inclusion of broad consent for any possible project has been proposed by some as
a solution (3). Others argue that this does not constitute informed consent, as the
possibilities for potential research are unpredictable (4,5). Our approach has been
to specifically describe research to be performed on collected specimens. This
prevents any misunderstanding between patient and researcher. This is not useful
for large repositories that collect tissue for future investigations.
TISSUE ACQUISITION
When tissues are devascularized and removed from the body, cellular metabolism
is altered as hypoxia and acidification occur and nutrient delivery ceases.
At a minimum, macromolecules in ischemic tissues are subject to degradation.
Further, cellular response to ischemia may lead to stress-specific changes in gene
expression. This may ultimately lead to changes in gene and protein expression
or degradation until the cellular machinery stops or tissue preservation occurs.
To accurately and reproducibly analyze gene and protein expression events
within cells, tissue must be preserved as quickly and uniformly as possible.
For all protocols where banking is superimposed upon clinical care, above
all else, banking strategies must not interfere with pathologic assessment of
either the tumor or specimen resection margins. When designing these protocols,
involvement of the pathologist who is custodian of the tissue record is critical.
This is not to say that banking cannot occur before specimen processing in
pathology, but that overall specimen integrity must be maintained during the
research specimen collection (e.g., core biopsy of lumpectomy specimens).
When a pathologist is immediately available, specimen banking after pathologic

Acquisition and Preservation of Tissue for Microarray Analysis 233
gross evaluation is preferred. However, if a pathologist is not immediately
available, then tumor specimens must be collected before pathologic assessment
but without specimen disruption. Regardless of when during the processing of
the pathologic specimen banking occurs, representative samples of all banked
tissue should be reviewed by the pathologist to both ensure that critical diagnostic
material is not lost and to document that research tissue samples are
indeed tumor tissues.
To further ensure that research tissue acquisition will not have an impact
on diagnosis, frozen specimens should be held and not analyzed until definitive
pathologic assessment of the entire remaining specimen has been completed. At
the time of definitive diagnosis of the surgical specimen, research specimens can
then be deposited in the tumor bank (Fig. 1). If analysis of the surgical specimen
does not demonstrate the abnormal pathology, then research specimens may be
Figure 1 Flow diagram demonstrating the sample handling of core biospy specimens in
patients suspected of having breast cancer. Research and clinical core biopsies should be
pathologically examined so that the patient’s diagnosis is not affected and that research
specimens contain the tissue of interest.

234 Untch and Olson
used for frozen-section analysis. Should the frozen section of research specimens
reveal abnormal histology not present in the surgical specimen (e.g., neoplasia
is present in the research specimens and not in the surgical specimen), all
samples can be turned over to pathology for routine processing. While this may
result in a short delay as surgical pathology is reevaluating frozen research
specimens, this can be explained in a preoperative consent form.
In all cases, frozen research specimens should be held frozen until two
weeks after a definitive pathologic report is issued on the diagnostic specimens.
This hold period is desirable, as once DNA and RNA are extracted, no further
diagnostic information relevant to patient care will be available.
For clinical trials and research protocols, specimen collection needs to be
standardized so that tumor samples are treated in a similar manner to minimize
technical variability in gene and protein expression data. Several cooperative
groups have developed banking SOPs for collection of formalin-fixed specimens
and blood samples; however, few have until recently focused on frozen-tissue
collection. The American College of Surgeons Oncology Group (ACOSOG) has
developed effective procedures for collecting tumor tissue from breast cancer
patients in neoadjuvant studies with good success. The European Human Frozen
Tumor Tissue Bank has also developed similar SPOs that are followed across
institutions and countries in an effort to reduce variability (2). In the former case,
specialized collection kits have been privately developed and are in use by
ACOSOG investigators for multisite tissue and blood collection at the time of
diagnostic biopsy and again at surgical resection (Fig. 2). Technicians responsible
for tumor collection prepare reagents and containers before excision.
Figure 2 A stereotactic biopsy device to retrieve breast cancer samples from excised
lumpectomy specimens. (A, arrow) Excised lumpectomy specimens are placed in the
device in an oriented fashion, and a specimen radiogram is taken to identify the location of
the tumor. (B) Coordinates are taken from the radiogram and are used to guide core biopsies
taken from the tumor within an example lumpectomy specimen held in fixed orientation in
the device.

Acquisition and Preservation of Tissue for Microarray Analysis 235
Samples are required to be snap frozen within five minutes of excision,
regardless of the method of tissue fixation (described subsequently). Rapid and
uniform handling of specimens from patient to preservative is essential.
TECHNIQUES OF CRYOPRESERVATION
RNA degradation by tissue RNases can be detrimental to RNA yield and quality,
with implications on microarray and quantitative PCR experiments. Traditional
practice has been to snap freeze samples in liquid nitrogen followed by permanent
storage at a minimum of 808C. This prevents changes to the RNA, and
tumor banks using this method have been highly successful in producing highquality
extracted RNA for microarray analysis. Given the complexity of sample
acquisition and multi-institutional collection, recent studies have tried to determine
the importance of immediate ex vivo freezing. Micke et al. (6) used ex vivo
tumors at room temperature to determine the amount of time until RNA degradation
occurred. Distinct ribosomal peaks were identified up to 16 hours,
indicating that this tissue can remain stable for extended periods of time. Other
sample-handling issues, such as freeze-thawing cycles, have also been studied.
Jochumsen et al. (7) established that gene expression analysis is not altered in up
to three freeze-thaw cycles. Despite these reports, it has been our experience that
RNA degradation can be extremely variable, even when SOPs are employed
(Figs. 3,4). To ensure quality RNA, samples should be frozen as soon as possible
after being devascularized.
Figure 3 RNA degradation is variable between two breast tumor samples. Samples are
subjected to the same conditions over 120 minutes. Sample 2 degrades faster than sample 1,
demonstrating the stability variation.

236 Untch and Olson
Figure 4 The effect of two-hour extended ischemia time on microarray data. The same
sample had RNA extracted at time 0 and after two hours and hybridized to the U95
Affymetrix platform. Raw expression values were log normalized and graphed. Data
demonstrate that at lower expression values greater variability exists between prolonged
ischemic and nonischemic tissue. There was also variability demonstrated in published
gene signatures describing prognosis. Source: From Refs. 26, 27, 31.
Preservation of tissue morphology can be difficult in samples that are snap
frozen. These samples are also subject to desiccation or freezer burn. Samples
can be immersed in optimal cutting temperature (OCT) compound to preserve
tissue morphology and prevent sample damage. This method has been shown to
generate reproducible genomic DNA extraction results (8). Our group has not
observed differences in microarray data when using OCT (unpublished observation).
However, Turbett et al. (9) reported inhibition of a polymerase chain
reaction (PCR) assay when compared with non-OCT compound–preserved
specimens. OCT compound can also make specimen thawing and dissection
more difficult. When specimens contain a mix of tissue (as in the case of
undissected tumor specimens with margins), separating whole tumor for RNA
extraction can be difficult, particularly when trying to avoid time-dependant
RNA degradation. Nonetheless, few studies have investigated OCT compound
and its potential to alter microarray results. A recommendation for or against its
use cannot be made at this time.
Because of the challenges of sample handling (pathology analysis) and the
potential for RNA degradation, RNA stabilization buffers have been developed.
Commercially available RNAlater (Ambion inc.) has been shown to maintain
RNA quality before snap freezing. Several studies have shown consistent
microarray results when storing samples in RNA stabilization buffers (10–12).
One such study found that samples could be stored up to one hour at 378C,
without affecting the RNA or subsequent microarray data (13). These buffers

Acquisition and Preservation of Tissue for Microarray Analysis 237
will be important for tissue acquisition protocols, particularly with regard to
multi-institutional trials that will require transfering tissue specimens to a central
tissue bank. This approach allows the samples to be shipped refrigerated rather
than frozen. The principal disadvantages are that proteins are degraded, and
subsequent sectioning for histologic verification of tumor identity is problematic.
PARAFFIN-EMBEDED TISSUE FOR MICROARRAY ANALYSIS
The discovery of paraffin wax is attributed to the noted hydrocarbon chemist,
Dr. Carl (Karl) Ludwig von Reichenbach (1788–1869). After being developed in
1830, pathology laboratories discovered its potential as a method for storing
and preserving specimens. Throughout the twentieth century, formalin-fixed,
paraffin-embedded tissue became a reliable way to archive tissue. Given the
large quantity of specimens available in paraffin, these tissue banks were identified
as a possible tissue source for array-based techniques. Numerous challenges have
been identified that complicate the macromolecular extraction process from these
samples.
Sample handling is an important step that can lead to variability observed
in microarray results. Ischemia time can profoundly affect protein, DNA, and
RNA extraction because of enzymatic degradation. As noted above, sample
collection is ideally performed by snap-freezing tissue in liquid nitrogen
immediately after being devascularized and removed from a patient. Sample
handling of formalin-fixed, paraffin-embedded tissue is variable, and this makes
application of the array data potentially unreliable.
Many investigators have attempted to develop methods for extraction of
these samples with mixed results. Given the cross-linking effects of formalin, it
was thought that protein extraction for proteomic analysis would be impossible.
However, attempts have been made to isolate proteins from paraffin, but results
have been mixed and more technical development is required (14).
Extraction of nucleic acids can be accomplished using commercially
available proteinase digestion buffers (15). This method also eliminates the risk
of Rnase activity in the sample as all previously fixed proteins are destroyed.
Formalin fixation does not appear to affect the yield of RNA, unless the tissue is
fixed for an extended period of time. This can result in cross-linking between
RNA and protein (16). As long as the methylol groups added to the bases during
fixation can be removed, the quality of the RNA is not perturbed (17) and results
have been comparable with frozen-tissue controls (18).
The most difficult problem encountered with these specimens has been that
the total DNA or RNA yield from these samples is inadequate and often
degraded (19). Yield of DNA or RNA is dependent on duration of protein
digestion, temperature, and pH (20). Most array platforms will require several
micrograms of RNA for hybridization. Amplification of the RNA has been
suggested as means to increase quantity, but these techniques can result in
considerable variability between samples (21). Further, amplification of partially

238 Untch and Olson
degraded samples can also result in unreliable data. Statistical adjustments for
nucleic acid amplification in microarray experiments are being developed and
may eventually prove to be useful.
As extraction, amplification, and statistical techniques improve, formalinfixed,
paraffin-embedded tissue may provide an additional resource for researchers.
Nevertheless, the variability in sample handling history makes application of the
array data from these samples less than reliable. At this point in time, use of these
samples for application to clinical investigations is discouraged.
FINE NEEDLE ASPIRATION, CORE NEEDLE BIOPSY,
AND SURGICAL SPECIMENS IN MICROARRAY ANALYSIS
Using micorarray analysis, numerous genomic signatures have been developed
that attempt to predict tumor phenotypes, including lymph node metastasis,
tumor growth patterns, postoperative margin positivity, chemosensitivity, and
oncogenic pathway activation. Most of these analyses have been based on
samples collected at the time of surgery. These large volume samples can yield
adequate RNA for array hybridization and also afford the opportunity to assess
multiple areas of the tumor and subcellular compartments via laser capture
microdissection. However, tumor size and growth is known to affect gene
expression, and successful application of the developed models to earlier stage,
preoperative biopsy specimens may not be possible.
Nonetheless, application of these predictive tests depends on obtaining
adequate tissue for microarray analysis before a definitive surgical intervention.
The preoperative diagnosis of breast cancer is typically achieved though two
methods of tumor biopsy: core needle biopsy (CNB) and fine needle aspiration
(FNA). For micoarray-based predictive testing, an optimal protocol would allow
these two modalities to collect tissue for pathologic analysis and for molecular
extraction.
FNA is performed by inserting a small gauge needle, typically under image
guidance, into a tumor or lesion. Tissue is removed through a negative-pressure
syringe. Tumor architecture is typically destroyed during this process, but the
resultant sample can be analyzed under the microscope for tumor cell characteristics.
Likewise, this small amount of tissue can be harvested for RNA and
microarray analysis. However, concern exists over adequate sample amount (22).
Several studies have shown that FNA can produce tissue in amounts appropriate
for pathology review and array applications (23,24).
CNB is performed under image or visual guidance in a fashion similar to
FNA. However, with CNB, a larger bore needle is used, and the resultant
specimen has maintained tissue architecture, allowing for a more accurate
pathologic analysis. This larger amount of tissue can also be used for array-based
studies. Several studies have addressed adequacy of tissue acquisition in core

Acquisition and Preservation of Tissue for Microarray Analysis 239
biopsy samples (13,24). Studies from our laboratory indicate that core biopsy
specimens can be collected before surgery and can undergo successful RNA
extraction (unpublished observation). The average yield from core biopsy
specimens was 10.4 mg of RNA, enough for most array platforms. After
extraction, RNA samples underwent hybridization to the HU95 Affymetrix
microarray platform and then chemosenstivity and oncogenic pathway signatures
were successfully applied. This is consistent with the work of Ellis et al. (13),
who was able to extract RNA from core biopsy specimens and distinguish cancer
and noncancer phenotypes.
LASER CAPTURE MICRODISSECTION
Recently, focus has shifted away from whole-tumor microarray analysis to
analysis of individual cell types within a tumor. A tumor is composed of a
heterogeneous group of cells, malignant cells, premalignant cells, normal
parenchymal cells, and supportive cells. This cellular heterogeneity must be
taken into account when applying array-based techniques (25). Numerous successful
expression-profiling studies have been performed using whole-tumor
macromolecular extraction (26–28). These analyses thus incorporate a variety of
different cell types. Laser capture microdissection allows one to isolate histologically
distinct cell types and perform subsequent analysis. The technique
employs an infrared laser in conjunction with light microscopy. Prepared histologic
slides are covered with a transfer film. When cells of interest are identified,
the laser is activated over the cells. This process causes the cells to attach
to the transfer film. The remainder of the sample remains, and the isolated cells
can undergo extraction procedures. Notably, this procedure can result in lower
yields of cells requiring amplification of nucleic acids for use in microarray
experiments (29,30). Laser capture microdissection has the potential to not only
improve our understanding of cellular deregulation but also to help determine
how other cells within the tumor contribute to the overall tumor phenotype.
Immune cells, vascular structures, and normal parenchymal cells within a tumor
may contribute significantly to the growth and metastatic potential. Expression
analysis of these individual cell types may help elucidate their roles.
SUMMARY
Acquisition and preservation of tissue for microarray experiments requires
understanding of informed consent, sample handling, macromolecular extraction,
and sampling methods. Significant advances in these areas have been made
in recent years to improve the quality of microarray data. Appreciating these
aspects is paramount if the application of microarray techniques is to contribute
meaningfully to the care of the cancer patient.

240 Untch and Olson
REFERENCES
1. McShane LM, Altman DG, Sauerbrei W, et al. Reporting recommendations for tumor
marker prognostic studies (REMARK). J Natl Cancer Inst 2005; 97(16):1180–1184.
2. Mager SR, Oomen MN, Morente MM, et al. Standard operating procedure for the
collection of fresh frozen tissue samples. Eur J Cancer 2007; 43(5):828–834.
3. Hansson MG, Dillner J, Bartram CR, et al. Should donors be allowed to give broad
consent to future biobank research? Lancet Oncol 2006; 7(3):266–269.
4. Maschke KJ. Alternative consent approaches for biobank research. Lancet Oncol
2006; 7(3):193–194.
5. Rothstein MA. Expanding the ethical analysis of biobanks. J Law Med Ethics 2005;
33(1):89–101.
6. Micke P, Ohshima M, Tahmasebpoor S, et al. Biobanking of fresh frozen tissue:
RNA is stable in nonfixed surgical specimens. Lab Invest 2006; 86(2):202–211.
7. Jochumsen KM, Tan Q, Dahlgaard J, et al. RNA quality and gene expression analysis
of ovarian tumor tissue undergoing repeated thaw-freezing. Exp Mol Pathol 2007;
82(1):95–102.
8. Sun Y, Hegamyer G, Colburn NH. A simple method using PCR for direct sequencing
of genomic DNA from frozen tumor tissue embedded in optimal cutting temperature
compound. Biotechniques 1992; 12(5):639–640.
9. Turbett GR, Sellner LN. The use of optimal cutting temperature compound can inhibit
amplification by polymerase chain reaction. Diagn Mol Pathol 1997; 6(5):298–303.
10. Gotzer MA, Patti R, Geoerger G, et al. Biological stability of RNA isolated from
RNAlater-treated brain tumor and neuroblastoma xenografts. Med Pediatr Oncol
2000; 34(6):438–442.
11. Florell SR, Coffin CM, Holden JA, et al. Preservation of RNA for functional
genomic studies: a multidisciplinary tumor bank protocol. Mod Pathol 2001; 14(2):
116–128.
12. Mutter GL, Zahrieh D, Liu C, et al. Comparison of frozen and RNALater solid tissue
storage methods for use in RNA expression microarrays. BMC Genomics 2004;
5(1):88.
13. Ellis M, Davis N, Coop A, et al. Development and validation of a method for using
breast core needle biopsies for gene expression microarray analyses. Clin Cancer Res
2002; 8(5):1155–1166.
14. Becker KF, Schott C, Hipp S, et al. Quantitative protein analysis from formalin-fixed
tissues: implications for translational clinical research and nanoscale molecular
diagnosis. J Pathol 2007; 211(3):370–378.
15. Krafft AE, Duncan BW, Bijwaard KE, et al. Optimization of the isolation and
amplification of RNA from formalin-fixed, paraffin-embedded tissue: The Armed
Forces Institute of Pathology Experience and Literature Review. Mol Diagn 1997;
2(3):217–230.
16. Moller K, Rinke J, Ross A, et al. The use of formaldehyde in RNA-protein crosslinking
studies with ribosomal subunits from Escherichia coli. Eur J Biochem 1977;
76(1):175–187.
17. Lewis F, Maughan NJ, Smith V, et al. Unlocking the archive: gene expression in
paraffin-embedded tissue. J Pathol 2001; 195(1):66–71.

Acquisition and Preservation of Tissue for Microarray Analysis 241
18. Specht K, Richter T, Muller U, et al. Quantitative gene expression analysis in
microdissected archival formalin-fixed and paraffin-embedded tumor tissue. Am J
Pathol 2001; 158(2):419–429.
19. Jackson DP, Lewis FA, Taylor GR. Tissue extraction of DNA and RNA and analysis
by the polymerase chain reaction. J Clin Pathol 1990; 43(6):499–504.
20. Gilbert MT, Haselkorn T, Bunce M, et al. The isolation of nucleic acids from fixed,
paraffin-embedded tissues-which methods are useful when? PLoS ONE 2007; 2(6):
E537.
21. Coombs NJ, Gough AC, Primrose JN. Optimisation of DNA and RNA extraction
from archival formalin-fixed tissue. Nucleic Acids Res 1999; 27(16):E12.
22. Centeno BA, Enkemann SA, Coppola D, et al. Classification of human tumors using
gene expression profiles obtained after microarray analysis of fine-needle aspiration
biopsy samples. Cancer 2005; 105(2):101–109.
23. Rapkiewicz A, Expina V, Zujewski JA. The needle in the haystack: application of
breast fine-needle aspirate samples to quantitative protein microarray technology.
Cancer 2007; 111(3):173–184.
24. Symmans WF, Ayers M, Clark EA, et al. Total RNA yield and microarray gene
expression profiles from fine-needle aspiration biopsy and core-needle biopsy
samples of breast carcinoma. Cancer 2003; 97(12):2960–2971.
25. Fuller AP, Palmer-Toy D, Erlander MG, et al. Laser capture microdissection and
advanced molecular analysis of human breast cancer. J Mammary Gland Biol
Neoplasia 2003; 8(3):335–345.
26. van de Vijver MJ, He YD, van ’t Veer LJ, et al. A gene-expression signature as a
predictor of survival in breast cancer. N Engl J Med 2002; 347(25):1999–2009.
27. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas
distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci
U S A 2001; 98(19):10869–10874.
28. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours.
Nature 2000; 406(6797):747–752.
29. Caretti E, Devarajan K, Coudry R, et al. Comparison of RNA amplification methods
and chip platforms for microarray analysis of samples processed by laser capture
microdissection. J Cell Biochem 2008; 103(2):556–563.
30. Schuetz CS, Bonin M, Clare SE, et al. Progression-specific genes identified by
expression profiling of matched ductal carcinomas in situ and invasive breast tumors,
combining laser capture microdissection and oligonucleotide microarray analysis.
Cancer Res 2006; 66(10):5278–5286.
31. Ramaswamy S, Ross KN, Lander ES, et al. A molecular signature of metastasis in
primary solid tumors. Nat Genet 2003; 33(1):49–54.

17
Tapping into the Formalin-Fixed
Paraffin-Embedded (FFPE) Tissue
Gold Mine for Individualization
of Breast Cancer Treatment
Mark Abramovitz
VM Institute of Research, Montreal, Que´bec, Canada
Brian Leyland-Jones
Winship Cancer Institute, Emory University School of Medicine,
Atlanta, Georgia, U.S.A.
INTRODUCTION
It is now clearly recognized that breast cancer is a heterogeneous disease that
requires personalized targeted therapy in order to have the greatest impact on
survival. This necessitates the development of biomarkers that can be used to
predict outcome and response to therapy. Unfortunately, very few biomarkers
have made their way from the research bench to the clinic. The main reason for
this is that in order to validate breast tumor markers, large sample sets from welldefined
patient groups are required, although this could be overcome by tapping
into large numbers of well-defined banked tumor specimens. The efforts that are
being made, though, toward the development of biomarkers in order to improve
the assessment of cancer prognosis and diagnosis are great, given the enormous
potential that they embody.
243

244 Abramovitz and Leyland-Jones
Formalin-fixed, paraffin-embedded (FFPE) tissue samples represent the
largest collection of well-annotated, clinical tumor samples that are readily
available for conducting retrospective studies and, therefore, represent a potential
gold mine from which important biomarkers are just waiting to be discovered,
given that the right tools can be brought to bear upon this invaluable resource.
However, until relatively recently, a major caveat regarding the use of these FFPE
specimens for genomewide studies was that due to the formalin fixation process,
the standard procedure for handling tumor tissue samples in pathology labs around
the world, RNA degradation occurred, greatly limiting their potential utility. In
fact, with regard to nucleic acid–based assays, FFPE specimens have been mainly
subjected to reverse transcribed-polymerase chain reaction (RT-PCR) analysis.
This chapter will describe the progress that has been made in greatly expanding the
use of FFPE tumor specimens for several genomic-based applications that will
greatly increase our ability to tap into this gold mine and extract the biomarker and
genomic data nuggets that will improve cancer prognosis, diagnosis, and treatment.
EXTRACTION OF RNA FROM FFPE TUMOR TISSUE
SAMPLES FOR GENE EXPRESSION ANALYSIS
FFPE tissue samples, a vast archive of pathologically well-characterized clinical
samples from randomized trials, are an enormous potential resource that will allow
translational scientists to more fully describe breast cancers at the molecular level
and will ultimately have important etiologic and clinical implications. Even
though the degradation of RNA that occurs because of the formalin fixation
process results in an RNA species with an average size of approximately 200 nt
(1), several groups have illustrated that it is feasible to extract and purify RNA
from such fixed tissue and perform gene expression profiling (2–9). With the
development of real-time quantitative (q) RT-PCR technology that has high
detection sensitivity and high dynamic range, it is now possible to detect even rare
messages in FFPE tissue and to examine the variation of expression over quite a
large dynamic range. Amplicons are designed specifically on small segments of
DNA [

The Formalin-Fixed Paraffin-Embedded (FFPE) Tissue Gold Mine 245
method with multianalyte capability for potential use in clinical research and
diagnostic assays. Two multianalyte assays, one with 48 genes and the other with
92 genes, were designed and tested in different experiments with preserved
human breast cancer tissue. In the first case, all of the genes profiled in FFPE
RNA yielded measurable values, and the overall profile was validated by its
similarity to that generated with well-preserved RNA from matched freshly
frozen tissue. In the second experiment using a 92-gene assay, only one of the
tested genes failed to yield a signal. Measured levels of estrogen receptor (ER),
progesterone receptor (PR), and human epidermal growth factor receptor 2
(HER2) mRNAs were concordant with the levels of the respective proteins as
measured by immunohistochemistry (IHC) at an independent clinical reference
laboratory. Approximately 90% concordance was obtained when RT-PCR
expression results for ER and PR were dichotomized into positive and negative
values and compared with ER- and PR-positive and ER- and PR-negative
assignments based on IHC. It is noteworthy that similar levels of concordance
were found in other comparative studies of IHC versus RT-PCR (15,16).
DEVELOPMENT OF AN RT-PCR-BASED BREAST CANCER
CLINICAL DIAGNOSTIC TEST: ONCOTYPE DX ASSAY
A major development in this area has been the development of a multigene qRT-
PCR assay to predict the likelihood of breast cancer recurrence in node-negative,
ER-positive, tamoxifen-treated patients (17). The authors studied the likelihood
of distant recurrence in breast cancer patients who had involved lymph nodes and
ER-positive tumors that were poorly defined by clinical and histopathologic
measures. They tested whether a prospectively-defined 21-gene RT-PCR assay
of FFPE tumor tissue could predict the likelihood of distant recurrence in
patients in the National Surgical Adjuvant Breast and Bowel Project (NSABP)
B-14 clinical trial. Measured levels of gene expression for 16 cancer-related and
5 reference genes were used, by a prospectively defined algorithm, to calculate a
recurrence score (RS) and risk group (low, intermediate, or high) for each
patient. Adequate RT-PCR profiles were obtained in 668 of 675 tumor blocks.
The proportion of patients categorized as low, intermediate, or high risk by the
RT-PCR assay was 51%, 23%, and 27%, respectively. The Kaplan–Meier estimates
and 95% confidence intervals for the rates of distant recurrence at 10 years
were 6.8% (4.0%, 9.6%), 14.3% (8.3%, 20.3%), and 30.5% (23.6%, 37.4%),
respectively, for the low-, intermediate-, and high-risk groups; the rate for
the low-risk group was significantly lower than the rate for the high-risk group
(p < 0.001). In a multivariate Cox model, RS provides significant (p < 0.001)
predictive power that goes beyond age and tumor size. The RS was also predictive
of overall survival (p < 0.001) and could be used as a continuous function
to predict distant recurrence for individual patients. Finally, the RS was prospectively
validated to predict the likelihood of distant recurrence in tamoxifentreated
breast cancer patients with node-negative, ER-positive tumors (18,19).

246 Abramovitz and Leyland-Jones
This methodology has been sufficiently successful, so that the Food and
Drug Administration (FDA) approved it in 2004 to become commercially available
[Genomic Health, Inc. (GHI), Redwood City, California, U.S.]. Oncotype DX
is a diagnostic assay that quantifies the likelihood of breast cancer recurrence in
women with newly diagnosed stage I or II, node-negative, ER-positive breast
cancer who will be treated with tamoxifen. More importantly, this test will pave
the way for individualization of cancer therapy. While the St. Gallen classification
distinguished 7.9% of patients as low risk, the GHI RS identified 50.6% of
patients as low risk. Similarly, St. Gallen overestimated the high-risk patients at
58.8%, while the GHI test classified high-risk patients at only 27.1%. It is obvious
that St. Gallen underestimated the low-risk patients and overestimated the highrisk
ones. The advantage of the newly developed multigene RS is, therefore, that
low-risk patients can be saved from potentially toxic, high-dose chemotherapy.
RNA-BASED EXPRESSION PROFILING OF
FFPE BREAST TUMOR SPECIMENS
Expression Profiling: The DASL Assay
Gene expression profiling has become an important aspect in prediction, prognosis,
and cancer modeling. Gene signatures resulting from expression profiling
studies have the potential to define cancer subtypes, predict the clinical outcome
(recurrence of disease) and response to specific therapies, and analyze oncogenic
pathways (20–23). Investigations of gene pathways and interactions indicated by
gene signatures that are truly predictive of the clinical endpoints are necessary to
understand the biology underlying this predictive value. More importantly, when
combined with clinical and demographic factors, multiple forms of molecular
(protein- and gene-based) data can provide information that has the potential to
identify unique characteristics of individuals and so lead to individualized
treatment strategies (24,25).
Although most array-based assays utilize RNA prepared from frozen
specimens, the newer cDNA-mediated annealing, selection, extension, and
ligation (DASL) assay (Illumina, Inc., San Diego, California, U.S.) platform was
specifically designed to profile small transcripts typically found in FFPE tissue
(26–31). The DASL platform is based on massively multiplexed RT-PCR
applied in a microarray format that allows for the determination of expression of
up to 512 genes (502 genes in the standard cancer panel available from Illumina)
using a minimal amount of total RNA (100 to *200 ng per assay) isolated
from 96 FFPE tumor tissue samples in a high-throughput format (29,30). In the
procedure, biotinylated random nonamers (biotin-d(N)9) and oligo d(T)18 are
used for cDNA synthesis, and therefore probes are designed so that they can
target any unique region of the gene without limiting the selection of the optimal
probe to the 3 0 ends of transcripts. In addition, because of the small size of the
targeted gene sequence (50 nucleotides), along with the use of random primers in

The Formalin-Fixed Paraffin-Embedded (FFPE) Tissue Gold Mine 247
Figure 1 The DASL platform for profiling both RNA and miRNA. (A) The DASL assay.
FFPE tumor tissue samples, 5- to 10-mm shaved sections or from microscope slides, are
first subjected to deparaffinization and then RNA extraction. The extracted RNA is
fragmented with an average size of approximately 200 nucleotides. Biotinylated cDNA is
made from the fragmented RNA using random primers (9-mers). Complex oligonucleotides
that contain a gene-specific sequence and a unique address sequence are used to
selectively bind to the cDNA. Extension and then ligation follow, resulting in cDNA
products that are subsequently amplified using fluorescently labeled common primers.
Depicted are three gene-specific oligonucleotide probe sets hybridized to cDNAs from a
single gene. The amplified products contain 1536 unique addresses (3 addresses per gene
and therefore 512 genes) and are hybridized to the 50,000-bead array (each unique
complementary address bead at *30-fold redundancy). The address sequence directs
hybridization to their complements on the universal bead arrays. (B) The miRNA array.
cDNA derived from total RNA is recognized as a specific miRNA by MSO hybridization.
Primer extension imparts further target specificity. The product of this extension is used as
a template for universal PCR. Abbreviations: DASL, cDNA-mediated annealing, selection,
extension, and ligation; FFPE, formalin-fixed, paraffin-embedded; PCR, polymerase
chain reaction.
the cDNA synthesis, RNAs that are otherwise too degraded for conventional
microarray analysis can be readily detected. Sequence-specific query oligonucleotides
encompassing primer extension, ligation, and universal PCR in highlyplexed
reactions (1536-plex), two-color labeling, and redundant (*30-fold
redundancy of each bead type) feature representations are used to probe up to
three different sites on each cDNA (Fig. 1A) and, therefore, lend the assay the
necessary sensitivity and reproducibility for quantitative detection of differential
expression using RNA from FFPE tissue (29,30). In an initial study, a 231-gene

248 Abramovitz and Leyland-Jones
cancer panel was used in the DASL assay in order to profile both breast and
colon cancer FFPE tumor samples. Cluster analysis was able to separate breast
from colon tissue types and, subsequently, divide each tissue sample set into
cancer versus normal (32). More recently, the DASL assay has been used to
identify RNA signatures in prostate cancer, including a 16-gene set that correlates
with prostate cancer relapse (31,33).
We have designed our own custom 512-gene breast cancer–related gene
panel so that it incorporates previously identified signature genes from various
breast cancer microarray studies that have been used in the classification of
breast tumors (34,35), in prognosis (MammaPrint) (36) and as predictors of
outcome to treatment (17,37–40), as well as a host of additional genes that have
been implicated as playing a role in a number of breast cancer–related processes,
including proliferation, angiogenesis, (41,42), metastasis (43,44), DNA repair,
apoptosis (45), thrombosis (46), and custom-selected genes taken from the most
recently published data on breast cancer (47–49). Additional cancer-related
genes include oncogenes, cell cycle genes, telomerase-related genes, amplified
genes, breast cancer stem cell genes, and senescence-related genes (50–52).
MiRNA Profiling
MiRNAs (miRNAs) are small noncoding RNA species that regulate the translation
and stability of mRNA messages for hundreds of downstream target genes
via partial complementarity to short sequences in the 3 0 -untranslated region
(3 0 UTR) of mRNAs. The aberrant expression or mutation of miRNAs, a distinct
class of RNAs that control gene expression via translation repression or degradation,
has been noted in cancers and suggests a key role for miRNAs in
tumorogenesis. MiRNAs can act both as tumor suppressors and as oncogenes,
depending on the sets of downstream targets that they regulate (53). Studies have
recently implicated nearly 30 of the over 450 known human miRNAs in breast
cancer, as tumor suppressor genes or as oncogenes, with significant deregulation
among only a handful (54). Others are correlated with specific tumor
types as identified by protein expression. For example, the mir-30 cluster is
downregulated in both ER-negative, PR-positive and ER-positive, PR-negative
tumors (54). Differential expression has also been noted by tumor stage or proliferative
indices. The tumor suppressor, p53, was recently shown to regulate
expression of mir-34b and mir-34c (55). One key set of tumor suppressor miRNAs
includes mir15a-16, which targets the translation of the suppressor of apoptosis,
Bcl-2 (56). These observations suggest that specific miRNAs are associated with
major breast cancer subgroups, including hormone-positive and hormone-negative
subgroups and the basal or triple negative subtype, and that the deregulated
miRNAs may differ between African-American and Caucasian women.
These low-molecular weight miRNAs can be extracted at the same time as
total RNA is being isolated from FFPE tissue samples. qRT-PCR can be used to
screen tumor samples for miRNA expression. The TaqMan TM array human

The Formalin-Fixed Paraffin-Embedded (FFPE) Tissue Gold Mine 249
miRNA panel from Applied Biosystems allows for the expression profiling of
365 miRNAs from one RNA sample. Illumina, Inc., has recently adapted its
DASL assay for high-throughput multiplexed miRNA expression profiling
(Fig. 1B). It has made a human miRNA panel that can be used to detect 470
miRNAs described in the miRBase database and an additional 265 potential
miRNAs that have been identified in an RNA-primed array-based Klenow
extension (RAKE) analysis study (57,58). With these arrays and tools now
available, our understanding of the role that miRNAs play in breast cancer will
be greatly enhanced.
Whole-Genome Expression Profiling
In order to conduct expression profiling using microarrays, it has been necessary
to obtain freshly frozen tissue as the starting material for RNA extraction. This
presents limitations with regard to obtaining enough samples, enough tissue per
sample, as well as the logistics of storing this type of material. Most recently,
new technologies have come into play that enable the use of RNA extracted from
FFPE to be used in genomewide gene expression profiling. The company
NuGEN Technologies, Inc. (San Carlos, California, U.S.) has developed an
amplification system that can take picogram amounts of total RNA to yield
micrograms of amplified cDNA that can be fragmented and labeled for same-day
hybridization to a variety of array formats, including Affymetrix, Agilent, and
Illumina microarrays (59). This type of technology will undoubtedly generate
significant microarray data from FFPE breast tumor samples in the near future
further adding to the utility of FFPE specimens.
DNA-BASED ARRAY PROFILING THAT EMPLOYS FFPE SAMPLES
Array CGH and SNP Profiling
DNA copy number alterations (CNAs), which are the result of genomic instability,
can give rise to gene amplifications and gene deletions that play important
roles in tumor progression, and it has become apparent that an increased number
of CNAs correlate with a poor prognosis for the patient. Several genes that
undergo CNAs have been identified that act as either oncogenes or tumor suppressors,
which are crucial factors in the disease process. With regard to breast
cancer, array comparative genomic hybridization (CGH) has been used principally
to identify novel breast cancer-related genes, genes that can be used to classify
different types of breast cancer, genes that yield prognostic or predictive information,
or genes that may represent important therapeutic targets. Single nucleotide
polymorphisms (SNPs) have been linked to individual variations in breast cancer
susceptibility. Genetic variations in genes involved in ER and drug metabolism and
genes involved in cell cycle control as well as variations depending on ethnic
background have been found to contribute to breast cancer susceptibility.

250 Abramovitz and Leyland-Jones
Array CGH has been used in the classification of breast tumors and has
shown that subtypes can be discerned on the basis of genomic instability patterns.
In one study, array CGH, made up of 2464 genomic clones, was applied to
the analysis of CNAs in 62 sporadic invasive ductal carcinomas, and three breast
tumor subtypes were delineated based on genomic DNA CNAs (60). In another
study, 89 breast cancer tumors of the luminal A, luminal B, HER2, and basal-like
subtypes were also found to yield distinct CNA profiles, suggesting that genomic
instability mechanisms differ between subtypes (61). Most array CGH studies
have made use of DNA prepared from fresh or frozen tumor samples; however,
the ability to tap into archival FFPE clinical samples would greatly accelerate the
rate at which disease-related genomic change, including translocation breakpoints,
could be more easily identified. Recent studies have addressed this
important issue and shown that DNA can be prepared from FFPE tissue samples
and used successfully in array CGH (62–64). Van Beers and colleagues (2006)
have also developed a multiplex PCR predictor that can be used to assess the
quality of the DNA extracted from FFPE samples (65) prior to use in array CGH.
Recently, Oosting and co-workers (2007) (66) have compared the Affymetrix
GeneChip and Bac array CGH with the Illumina BeadArray platform, which is a
high-density SNP microarray that allows for the measurement of both genomic
copy number and loss of heterozygosity (LOH). The BeadArray platform (Fig. 2)
is amenable for use with fragmented DNA from FFPE tissue because of the fact
that only short, intact genomic segments of approximately 40 bp flank each SNP
on the array. The authors found that when paired comparisons of DNA prepared
from freshly frozen and FFPE tissue were conducted, the patterns of genomic
aberrations were basically identical.
The use of array CGH and SNP profiling in conjunction with the vast
archive of FFPE clinical breast cancer samples will allow for the identification of
chromosomal aberrations as well as novel variants that contribute to breast
cancer, and this will lead to uncovering novel breast cancer-related genes that
will be key to understanding the molecular mechanisms underlying the disease.
Methylation Profiling
Epigenetic changes, which involve very subtle molecular modifications, can
cause activation or suppression of genes that can lead to abnormalities resulting
in tumorigenesis (67), and these epigenetic changes have been found to occur in
many cancers, including breast, colon, prostate, and blood cancer. The epigenotype
is affected by such factors as the environment, age, epigenotype of the
parents, and the genotype at the loci that regulate DNA methylation and chromatin,
and these factors can cause aberrant methylation of the cytosine residue of
DNA, resulting in gene regulation.
Illumina has developed an array-based platform (Fig. 3) that can be used in
conjunction with its standard Cancer Methylation Panel, which spans 1505 CpG
loci selected from 807 genes with 71% of those selected genes containing at least

The Formalin-Fixed Paraffin-Embedded (FFPE) Tissue Gold Mine 251
Figure 2 Whole-genome expression profiling on the BeadChip array. (A) Gene-specific
probes are attached to beads, which are then assembled into the arrays. For simplicity,
this diagram shows only one oligomer attached to the bead; actual beads have hundreds
of thousands of copies of the same sequence attached. (B) Sentrix Human-6 and
HumanRef-8 Expression BeadChips. More than 48,000 different bead types, each with
a unique sequence, are represented on the Human-6 Expression BeadChip, greater
than 24,000 on the HumanRef-8 Expression BeadChip. Each bead contains hundreds
of thousands of copies of covalently attached oligonucleotide probes. Beads are
assembled into greater than 1.6 million pits, each measuring 3 mm in diameter,
generating an average 30-fold redundancy for each sequence represented on the array.
2 CpG sites. The Cancer Methylation Panel includes genes selected from various
functional classes, including oncogenes and tumor suppressor genes; genes
involved in cell cycle control, DNA repair, differentiation, and apoptosis; and
X-linked and imprinted genes (68). Although DNA prepared from FFPE tumor
samples has not yet been specifically validated for use in the assay, it would be
useful to be able to scan large numbers of FFPE samples for methylation patterns,
which could provide important clues regarding tumor type.
THE WAY FORWARD
The goal is to obtain as complete a molecular picture of a breast tumor as possible,
which can now be realistically undertaken using FFPE tissue specimens, truly
untapped gold now ready to be brought to the surface and subjected to a thorough
nucleic acid characterization. Profiling platforms can now be brought to
bear on these invaluable samples in order to build a comprehensive picture of the
tumor. In the scheme shown in Figure 4, FFPE samples, either sections or cores,
are used to extract DNA, RNA, and miRNA. These nucleic acids can now be

252 Abramovitz and Leyland-Jones
Figure 3 The DNA methylation assay. (A) In order to perform a DNA methylation assay,
genomic DNA must first undergo bisulfite conversion. Cytosines are converted to uracils,
but methyl-cytosine is unreactive. (B) In order to take advantage of this differential
reactivity to bisulfite treatment, two pairs of complex oligonucleotide probe sets are
designed for each CpG site. The first probe pair is made up of an ASO and a LSO to
interrogate the methylated state of the CpG site and a corresponding ASO-LSO pair for the
unmethylated state. The ASO is used to determine whether a site is methylated or not and
also incorporates a universal PCR primer sequence, P1 or P2. P1 and P2 are fluorescently
labeled, each with a different dye, and associated with the “T” (unmethylated) allele or
the “C” (methylated) allele, respectively. The LSO is made up of three parts: a CpG
(Continued)

The Formalin-Fixed Paraffin-Embedded (FFPE) Tissue Gold Mine 253
Figure 4 Analysis of FFPE specimens on multiple array platforms. FFPE tissue samples,
as sections on either slides, sections, or cores are available for extraction of nucleic acids.
DNA, RNA, and miRNA can be isolated and applied to various array platforms in order to
maximize the molecular information that can be obtained from a single tumor sample.
DNA can be subjected to array CGH, SNP analysis, LOH profiling, and methlylation
analysis. RNA can be used in conjunction with expression profiling platforms that include
the DASL assay and whole-genome expression profiling on different microarray platforms.
MiRNA can be used to conduct miRNA profiling. This ability to utilize FFPE
tumor tissues for detailed molecular characterization will undoubtedly have a profound
impact on our understanding of tumor biology. Abbreviations: FFPE, formalin-fixed,
paraffin-embedded; CGH, comparative genomic hybridization; SNP, single nucleotide
polymorphisms; LOH, loss of heterozygosity; DASL, cDNA-mediated annealing, selection,
extension, and ligation.
<
locus-specific sequence, an address sequence in the middle corresponding to a complementary
address sequence on the bead array, and a universal PCR priming site (P3) at the
3 0 -end. The pooled assay oligonucleotides are first annealed to bisulfite-converted
genomic DNA. An allele-specific primer extension step is then carried out using ASOs
that are extended only if their 3 0 -base is complementary to their cognate CpG site in the
genomic DNA template. Allele-specific extension is followed by ligation of the extended
ASOs to their corresponding LSOs to create PCR templates. Common primers P1, P2, and
P3 are then used to amplify the ligated products, which are subsequently hybridized to a
microarray bearing the complementary address sequences. Abbreviations: ASO, allelespecific
oligonucleotide; LSO, locus-specific oligonucleotide; PCR, polymerase chain
reaction.

254 Abramovitz and Leyland-Jones
subjected to a panoply of analysis platforms that can be used to detect CNAs,
analyze SNPs, and investigate methylation aberrations, regulated gene expression,
and signaling pathway changes, as well as miRNA alterations. Data derived from
tumor tissue analyses can be used to further our understanding of the different
subtypes, which could lead to new insights into breast cancer etiology, prediction
of disease outcome, response to therapy, and ultimately to better and personalized
patient management.
REFERENCES
1. Masuda N, Ohnishi T, Kawamoto S, et al. Analysis of chemical modification of RNA
from formalin-fixed samples and optimization of molecular biology applications for
such samples. Nucleic Acids Res 1999; 27:4436–4443.
2. Abrahamsen HN, Steiniche T, Nexo E, et al. Towards quantitative mRNA analysis in
paraffin-embedded tissues using real-time reverse transcriptase-polymerase chain
reaction: a methodological study on lymph nodes from melanoma patients. J Mol
Diagn 2003; 5:34–41.
3. Finke J, Fritzen R, Ternes P, et al. An improved strategy and a useful housekeeping
gene for RNA analysis from formalin-fixed, paraffin-embedded tissues by PCR.
Biotechniques 1993; 14:448–453.
4. Godfrey TE, Kim SH, Chavira M, et al. Quantitative mRNA expression analysis
from formalin-fixed, paraffin-embedded tissues using 5 0 nuclease quantitative
reverse transcription-polymerase chain reaction. J Mol Diagn 2000; 2:84–91.
5. Reichmuth C, Markus MA, Hillemanns M, et al. The diagnostic potential of the
chromosome translocation t(2;13) in rhabdomyosarcoma: a Pcr study of fresh-frozen
and paraffin-embedded tumour samples. J Pathol 1996; 180:50–57.
6. Rupp GM, Locker J. Purification and analysis of RNA from paraffin-embedded
tissues. Biotechniques 1988; 6:56–60.
7. Sheils OM, Sweeney EC. TSH receptor status of thyroid neoplasms–TaqMan RT-
PCR analysis of archival material. J Pathol 1999; 188:87–92.
8. Specht K, Richter T, Muller U, et al. Quantitative gene expression analysis in
microdissected archival formalin-fixed and paraffin-embedded tumor tissue. Am
J Pathol 2001; 158:419–429.
9. Stanta G, Bonin S. RNA quantitative analysis from fixed and paraffin-embedded
tissues: membrane hybridization and capillary electrophoresis. Biotechniques 1998;
24:271–276.
10. Abramovitz M, Leyland-Jones B. Application of array-based genomic and epigenomic
technologies to unraveling the heterogeneous nature of breast tumors: on
the road to individualized treatment. Cancer Genomics Proteomics 2007; 4:135–145.
11. Chen J, Byrne GEJr., Lossos IS. Optimization of RNA extraction from formalinfixed,
paraffin-embedded lymphoid tissues. Diagn Mol Pathol 2007; 16:61–72.
12. Chung JY, Braunschweig T, Hewitt SM. Optimization of recovery of RNA from
formalin-fixed, paraffin-embedded tissue. Diagn Mol Pathol 2006; 15:229–236.
13. Penland SK, Keku TO, Torrice C, et al. RNA expression analysis of formalin-fixed
paraffin-embedded tumors. Lab Invest 2007; 87:383–391.

The Formalin-Fixed Paraffin-Embedded (FFPE) Tissue Gold Mine 255
14. Cronin M, Pho M, Dutta D, et al. Measurement of gene expression in archival
paraffin-embedded tissues: development and performance of a 92-gene reverse
transcriptase-polymerase chain reaction assay. Am J Pathol 2004; 164:35–42.
15. Chevillard S, Muller A, Levalois C, et al. Reverse transcription-polymerase chain
reaction (RT-PCR) assays of estrogen and progesterone receptors in breast cancer.
Breast Cancer Res Treat 1996; 41:81–89.
16. Knowlden JM, Gee JM, Bryant S, et al. Use of reverse transcription-polymerase
chain reaction methodology to detect estrogen-regulated gene expression in small
breast cancer specimens. Clin Cancer Res 1997; 3:2165–2172.
17. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifentreated,
node-negative breast cancer. N Engl J Med 2004; 351:2817–2826.
18. Paik S, Tang G, Shak S, et al. Gene expression and benefit of chemotherapy in
women with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol
2006; 24:3726–3734.
19. Paik S. Development and clinical utility of a 21-gene recurrence score prognostic
assay in patients with early breast cancer treated with tamoxifen. Oncologist 2007;
12:631–635.
20. Bild AH, Yao G, Chang JT, et al. Oncogenic pathway signatures in human cancers as
a guide to targeted therapies. Nature 2006; 439:353–357.
21. Huang E, Ishida S, Pittman J, et al. Gene expression phenotypic models that predict
the activity of oncogenic pathways. Nat Genet 2003; 34:226–230.
22. Huang ES, Black EP, Dressman H, et al. Gene expression phenotypes of oncogenic
signaling pathways. Cell Cycle 2003; 2:415–417.
23. Ramaswamy S, Tamayo P, Rifkin R, et al. Multiclass cancer diagnosis using tumor
gene expression signatures. Proc Natl Acad Sci U S A 2001; 98:15149–15154.
24. Nevins JR, Huang ES, Dressman H, et al. Towards integrated clinico-genomic models
for personalized medicine: combining gene expression signatures and clinical factors
in breast cancer outcomes prediction. Hum Mol Genet 2003; 12(2):R153–R157.
25. van de Vijver MJ, He YD, van ’t Veer LJ, et al. A gene-expression signature as a
predictor of survival in breast cancer. N Engl J Med 2002; 347:1999–2009.
26. Abramovitz M, Ordanic-Kodani M, Wang Y, et al. Optimization of RNA extraction
from FFPE tissues for expression profiling in the DASL assay. Biotechniques 2007
(in press).
27. Abramovitz M, Leyland-Jones B. A systems approach to clinical oncology: focus on
breast cancer. Proteome Sci 2006; 4:5.
28. Bibikova M, Yeakley JM, Chudin E, et al. Gene expression profiles in formalinfixed,
paraffin-embedded tissues obtained with a novel assay for microarray analysis.
Clin Chem 2004; 50:2384–2386.
29. Bibikova M, Talantov D, Chudin E, et al. Quantitative gene expression profiling in
formalin-fixed, paraffin-embedded tissues using universal bead arrays. Am J Pathol
2004; 165:1799–1807.
30. Fan JB, Yeakley JM, Bibikova M, et al. A versatile assay for high-throughput gene
expression profiling on universal array matrices. Genome Res 2004; 14:878–885.
31. Li HR, Wang-Rodriguez J, Nair TM, et al. Two-dimensional transcriptome profiling:
identification of messenger RNA isoform signatures in prostate cancer from archived
paraffin-embedded cancer specimens. Cancer Res 2006; 66:4079–4088.

256 Abramovitz and Leyland-Jones
32. Bibikova M, Talantov D, Chudin E, et al. Quantitative gene expression profiling in
formalin-fixed, paraffin-embedded tissues using universal bead arrays. Am J Pathol
2004; 165:1799–1807.
33. Bibikova M, Chudin E, Arsanjani A, et al. Expression signatures that correlated with
Gleason score and relapse in prostate cancer. Genomics 2007; 89:666–672.
34. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas
distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci
U S A 2001; 98:10869–10874.
35. Sotiriou C, Neo SY, McShane LM, et al. Breast cancer classification and prognosis
based on gene expression profiles from a population-based study. Proc Natl Acad Sci
U S A 2003; 100:10393–10398.
36. van ’t Veer LJ, Dai H, van de Vijver MJ, et al. Gene expression profiling predicts
clinical outcome of breast cancer. Nature 2002; 415:530–536.
37. Ayers M, Symmans WF, Stec J, et al. Gene expression profiles predict complete
pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and
cyclophosphamide chemotherapy in breast cancer. J Clin Oncol 2004; 22:2284–2293.
38. Chang JC, Wooten EC, Tsimelzon A, et al. Patterns of resistance and incomplete
response to docetaxel by gene expression profiling in breast cancer patients. J Clin
Oncol 2005; 23:1169–1177.
39. Jansen MP, Foekens JA, van Staveren IL, et al. Molecular classification of tamoxifenresistant
breast carcinomas by gene expression profiling. J Clin Oncol 2005; 23:
732–740.
40. Pawitan Y, Bjohle J, Amler L, et al. Gene expression profiling spares early breast
cancer patients from adjuvant therapy: derived and validated in two population-based
cohorts. Breast Cancer Res 2005; 7:R953–R964.
41. Heuze-Vourc’h N, Liu M, Dalwadi H, et al. IL-20, an anti-angiogenic cytokine that
inhibits COX-2 expression. Biochem Biophys Res Commun 2005; 333:470–475.
42. Zhong H, Chiles K, Feldser D, et al. Modulation of hypoxia-inducible factor 1alpha
expression by the epidermal growth factor/phosphatidylinositol 3-kinase/PTEN/
AKT/FRAP pathway in human prostate cancer cells: implications for tumor angiogenesis
and therapeutics. Cancer Res 2000; 60:1541–1545.
43. Kang Y, Siegel PM, Shu W, et al. A multigenic program mediating breast cancer
metastasis to bone. Cancer Cell 2003; 3:537–549.
44. Minn AJ, Gupta GP, Siegel PM, et al. Genes that mediate breast cancer metastasis to
lung. Nature 2005; 436:518–524.
45. Miller LD, Smeds J, George J, et al. An expression signature for p53 status in human
breast cancer predicts mutation status, transcriptional effects, and patient survival.
Proc Natl Acad Sci U S A 2005; 102:13550–13555.
46. Kwaan HC, Parmar S, Wang J. Pathogenesis of increased risk of thrombosis in
cancer. Semin Thromb Hemost 2003; 29:283–290.
47. Gonzalez MA, Pinder SE, Callagy G, et al. Minichromosome maintenance protein 2
is a strong independent prognostic marker in breast cancer. J Clin Oncol 2003;
21:4306–4313.
48. McLean GW, Carragher NO, Avizienyte E, et al. The role of focal-adhesion kinase
in cancer: a new therapeutic opportunity. Nat Rev Cancer 2005; 5:505–515.
49. Zhao C, Yasui K, Lee CJ, et al. Elevated expression levels of NCOA3, TOP1, and
TFAP2C in breast tumors as predictors of poor prognosis. Cancer 2003; 98:18–23.

The Formalin-Fixed Paraffin-Embedded (FFPE) Tissue Gold Mine 257
50. Brabletz T, Jung A, Spaderna S, et al. Opinion: migrating cancer stem cells - an
integrated concept of malignant tumour progression. Nat Rev Cancer 2005; 5:744–749.
51. Collado M, Gil J, Efeyan A, et al. Tumour biology: senescence in premalignant
tumours. Nature 2005; 436:642.
52. Dikmen ZG, Gellert GC, Jackson S, et al. In vivo inhibition of lung cancer by
GRN163L: a novel human telomerase inhibitor. Cancer Res 2005; 65:7866–7873.
53. Hammond SM. MicroRNAs as oncogenes. Curr Opin Genet Dev 2006; 16:4–9.
54. Iorio MV, Ferracin M, Liu CG, et al. MicroRNA gene expression deregulation in
human breast cancer. Cancer Res 2005; 65:7065–7070.
55. Corney DC, Flesken-Nikitin A, Godwin AK, et al. MicroRNA-34b and microRNA-
34c are targets of p53 and cooperate in control of cell proliferation and adhesionindependent
growth. Cancer Res 2007; 67:8433–8438.
56. Cimmino A, Calin GA, Fabbri M, et al. miR-15 and miR-16 induce apoptosis by
targeting BCL2. Proc Natl Acad Sci U S A 2005; 102:13944–13949.
57. Berezikov E, van Tetering G, Verheul M, et al. Many novel mammalian microRNA
candidates identified by extensive cloning and RAKE analysis. Genome Res 2006;
16:1289–1298.
58. Griffiths-Jones S, Grocock RJ, van Dongen S, et al. miRBase: microRNA sequences,
targets and gene nomenclature. Nucleic Acids Res 2006; 34:D140–D144.
59. Turner, L, de Feo, G, Heath, JD, et al. WT-Ovation TM FFPE system: RNA amplification
for gene expression profiling using RNA from formalin fixed, paraffin
embedded (FFPE) tissue. Los Angeles, CA: American Association of Cancer
Research (AACR) 20073 (abstr).
60. Fridlyand J, Snijders AM, Ylstra B, et al. Breast tumor copy number aberration
phenotypes and genomic instability. BMC Cancer 2006; 6:96.
61. Bergamaschi A, Kim YH, Wang P, et al. Distinct patterns of DNA copy number alteration
are associated with different clinicopathological features and gene-expression subtypes
of breast cancer. Genes Chromosomes Cancer 2006; 45:1033–1040.
62. Johnson NA, Hamoudi RA, Ichimura K, et al. Application of array CGH on archival
formalin-fixed paraffin-embedded tissues including small numbers of microdissected
cells. Lab Invest 2006; 86:968–978.
63. Joosse SA, van Beers EH, Nederlof PM. Automated array-CGH optimized for archival
formalin-fixed, paraffin-embedded tumor material. BMC Cancer 2007; 7:43.
64. Little SE, Vuononvirta R, Reis-Filho JS, et al. Array CGH using whole genome
amplification of fresh-frozen and formalin-fixed, paraffin-embedded tumor DNA.
Genomics 2006; 87:298–306.
65. van Beers EH, Joosse SA, Ligtenberg MJ, et al. A multiplex PCR predictor for aCGH
success of FFPE samples. Br J Cancer 2006; 94:333–337.
66. Oosting J, Lips EH, van Eijk R, et al. High-resolution copy number analysis of paraffinembedded
archival tissue using SNP BeadArrays. Genome Res 2007; 17:368–376.
67. Esteller M. Cancer epigenomics: DNA methylomes and histone-modification maps.
Nat Rev Genet 2007; 8:286–298.
68. Bibikova M, Lin Z, Zhou L, et al. High-throughput DNA methylation profiling using
universal bead arrays. Genome Res 2006; 16:383–393.

18
Breast Cancer Stem Cells and Their
Niche: Lethal Seeds in Lethal Soil
Danuta Balicki
Department of Medicine and Research Centre, Centre Hospitalier de
l’Universite´ de Montre´al; Department of Medicine, Universite´ de Montre´al;
and Breast Cancer Prevention and Genetics and Breast Cancer Risk
Assessment Clinic, Ville Marie Multidisciplinary Breast Cancer Centre,
Montre´al, Que´bec, Canada
Brian Leyland-Jones
Winship Cancer Institute, Emory University School of Medicine,
Atlanta, Georgia, U.S.A.
Max S. Wicha
Comprehensive Cancer Centre, Department of Internal Medicine, Division of
Hematology-Oncology, University of Michigan, Ann Arbor, Michigan, U.S.A.
INTRODUCTION
The quest for optimal breast cancer therapy has been thwarted by recurrent
treatment failures. As a result, the necessity has emerged to concomitantly rethink
the fundamental basis of breast cancer, and to ensure that it is adequately and
optimally targeted by therapeutic agents, and combinations of these agents.
Adding to the complexity of this disease is the heterogeneity of breast cancer, and
the realization that breast cancer is not a single disease, but a family of diseases.
The daunting challenge that we face with a newly diagnosed breast cancer patient
is one-to-one communication that tailors an individualized treatment strategy
259

260 Balicki et al.
based on the corresponding breast cancer pathophysiology. At the same time, we
are mandated to address the urgency of this therapy on the basis of the underlying
clinical assessment, the risks of metastases and death, as well as the benefits and
adverse effects associated with therapeutic regimens to ensure that an enlightened
decision among available therapeutic options can be mutually reached with an
individual patient. As our understanding of the pathophysiology of breast cancer is
incomplete, it is critical to channel the information that we glean from observations
that deviate from our current models into a unified paradigm shift where therapeutic
targets are validated, and subsequently attacked and destroyed. Given the
complexity of breast cancer, the highest level of therapeutic success is most
likely to be achieved when the uniqueness of an individual patient’s breast
cancer is matched with individualized, combined, and targeted therapy.
CANCER STEM CELL HYPOTHESIS
The cancer stem cell hypothesis is a major player in the current paradigm shift in
oncology (1,2). Cancer is, in essence, a genetic disease where three groups of genes
are responsible for tumor progression, namely oncogenes, tumor suppressor genes,
and stability genes (3,4). Cancer is a multistage process involving the accumulation
of genetic and epigenetic changes in genes involved in the regulation of cell
growth, DNA repair, and metastasis. In the traditional stochastic model, the vast
majority of cancer cells can proliferate extensively to form new tumors. In contrast,
the cancer stem cell model suggests that only rare cancer stem cells have the
intrinsic ability to proliferate extensively to form new tumors (Fig. 1) (1). Clinical
observations indicate that breast cancer responses to therapy do not necessarily
correlate with patient survival. Tumor-initiating cells with the concomitant stem
cell properties of self-renewal and successive maturation of progenitor cells into
highly differentiated cells are referred to as “cancer stem cells” (Fig. 2).
The cancer stem cell hypothesis provides an explanation for our clinical
observations, including those which deviate from the stochastic model, by suggesting
that therapies that kill cancer cells but not cancer stem cells may lead to tumor
regression, but that the disease will subsequently recur as the cancer stem cells act as
lethal seeds to repopulate the tumor (5). Thus, tumor regression in itself may represent
an incomplete clinical end point, as it reflects cell death limited to differentiated tumor
cells, while sparing cancer stem cells (Fig. 3). The “dandelion hypothesis” has been
used to illustrate this pattern of clinical activity as it is analogous to cutting off a
dandelion (or other weed) at ground level, which may initially appear to produce the
desired effect, but only the elimination of the root will prevent the weed from
regrowing (6). Therefore, effective therapies should target cancer stem cells, with
limited effects on normal cells. In the presence of therapies that kill cancer stem
cells, tumors are predicted to lose their ability to generate new cancer cells,
subsequently degenerate, leading to complete tumor eradication.
Stem cells are multipotent with high proliferative potential and with
concomitant properties of self-renewal and differentiation into multilineage

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 261
Figure 1 A comparison of the stochastic model where most cancer cells can proliferate
extensively to give rise to new tumors with the cancer stem cell model where only a few
cancer stem cells can give rise to new tumors. Abbreviation: CSC, cancer stem cell.
cells. Recent findings suggest that stem cell properties are integral and fundamental
to the formation and perpetuation of human cancers, including breast
cancer (7). The quiescence, self-renewal, and differentiation of stem cells are
regulated by both intrinsic and extrinsic mechanisms. Intrinsic mechanisms
include those governing the epigenetic state of stem cells, for example, hematopoietic
stem cells are controlled by chromatin remodelers such as polycomb
group proteins (8,9). The extrinsic mechanisms include changes in stem cell
fate that are dictated by the environment, i.e., the stem cell niche. As our tissues
are organized into stem cell hierarchies ranging from stem cells with extensive
proliferative and self-renewal capacity to mature cells with little or no capacity
for cell division, it is likely that cancer cell populations might also be organized
in stem cell hierarchies, ranging from a small number of cells that are responsible
for fueling the uncontrolled growth of the tumor cells to a large array of differentiated
daughter cells (10). While the vast majority of cells in a tumor are
merely the nontumorigenic daughter cells of cancer stem cells, the highly proliferative
capacities of stem cells may drive the continued expansion of malignant
cells. As cellular proliferation decreases with differentiation, the long-lived
and slowly dividing stem and progenitor cells are likely candidates for the
accumulation of mutations associated with carcinogenesis (11,12).

262 Balicki et al.
Figure 2 NSCs and CSCs have the capacity of self-renewal and differentiation into
progenitor cells. NSCs divide asymmetrically giving rise to a stem cell and a differentiated
progenitor cell. CSCs may lose asymmetric division, leading to selectively selfrenewal,
thereby increasing the number of CSCs and the subsequent tumor burden.
Abbreviations: NSCs, normal stem cells; CSCs, cancer stem cells.
The multilineage differentiation of cancer stem cells may contribute to the
heterogeneity of certain tumors, including breast cancer. The parallels between
normal stem cells and cancer stem cells suggest that some cancer stem cells
arise from mutated normal stem cells, and are responsible for the proliferative
capacity of cancers (Fig. 2) (7). The observation that stem and cancer cells share
certain signaling pathways that regulate their self-renewal suggests that normal
stem cells can, in fact, give rise to cancer cells (8). Mutated progenitor cells
which acquire self-renewal properties may also become cancer stem cells
(Fig. 2) (7,13). Epigenetic gene silencing and associated promoter CpG island
DNA hypermethylation may precede genetic changes in premalignant cells and
foster the accumulation of additional genetic and epigenetic hits that contribute

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 263
Figure 3 Therapies which kill cancer cells, but not CSCs lead to an initial tumor
regression, but as the CSCs are not eliminated, the tumor will regenerate. In contrast,
therapies that target CSCs also eliminate the genesis of new tumor cells and result in
tumor regression. Abbreviation: CSCs, cancer stem cells.
to the generation of cancer stem cells from stem or early progenitor cells (14).
The accumulation of mutations in cancer stem cells may disrupt the tight control
of normal stem cells, leading to the dysregulation of self-renewal, tumorigenesis,
loss of asymmetric division, and aberrant differentiation (Fig. 2) (12,15). The
fate of stem cells, including self-renewal and differentiation, may also be programmed
by the stem cell niche, i.e., the cellular microenvironment providing
necessary support and stimuli to sustain self-renewal and other environmental
factors (11,13,16–19). In normal breast development, mammary stem cells differentiate
into myoepithelial, alveolar epithelial, and ductal epithelial cells (20).
Cancer stem cells may acquire additional features associated with tumor progression,
metastases, and therapeutic failure, including genetic instability and
drug resistance (11). Mechanisms of therapeutic resistance include cellular
quiescence and the acquisition of protein expression responsible for the cellular
efflux of chemotherapeutic agents. The elimination of the cancer stem cell
compartment of a tumor may be necessary to achieve stable, long-lasting cancer
remissions. Therapeutic strategies designed to eradicate the cancer stem

264 Balicki et al.
compartment should selectively, or at least preferentially, target the self-renewal,
survival, and proliferative pathways specific to cancer stem cells, or induce
differentiation. Thus, advances in our knowledge of stem cell properties are
pivotal to the specific and effective targeting of these cells in therapeutic
strategies, including cancer therapies.
CURRENT STATUS OF BREAST STEM CELL RESEARCH
The human mammary gland is organized during development as a hierarchy of
stem and progenitor cells that are successively limited in their multilineage
potential and proliferative capacities. Stem cells represent approximately 1 in
250 epithelial cells of the normal human breast (17). The transplantation of
mammary cells into cleared murine fat pads is a functional assay used to identify
mammary epithelial cells with stem cell properties based on their ability to
generate tumors (21,22). The existence of a stem cell origin of murine mammary
development was established using transplantation of cells at limiting dilutions.
These experiments confirmed the clonal nature of the regenerated alveolar,
ductal, lobular, and more complex structures (22–24). The unequivocal demonstration
of the existence of a murine mammary stem cell occurred with the
reconstitution of a complete and functional mammary gland from a single lineage
negative (Lin )/cluster of differentiation (CD)24 þ /CD29 þ cell, marked
with a b-galactosidase (LacZ) transgene (25).
Three types of murine mammary epithelial cell progenitors have been
identified thus far (22–24). The first is a bipotent progenitor with features of
both luminal and myoepithelial characteristics, the second has luminal features,
and the third has myoepithelial features. Analogous human studies of stem and
progenitor cells have been limited by the challenges associated with the development
of suitable in vivo xenotransplantation assays. Alternative strategies
have been undertaken to assay human breast stem cells and to optimize conditions
that support the in vitro growth and differentiation of primitive human epithelial
cells. These include the proliferation of nonadherent mammospheres in suspension
from human breast cancer tissues (17,26) and murine mammary fat pads
(27). These mammospheres are enriched in stem and progenitor cells, as illustrated
by their ability to differentiate along all three mammary epithelial lineages
and to clonally generate complex functional structures in reconstituted threedimensional
culture systems (17). Additional studies to characterize stem and
progenitor cells include the identification of side populations using flow cytometry,
i.e., low Hoechst 33342 dye-retaining populations (28–30). Side populations
are enriched 30-fold in mammospheres derived from human tissues (17).
The use of mammosphere assays subsequently lead to the discovery that
the Notch signaling pathway plays a pivotal role in normal human mammary
development through its activity in both stem and progenitor cells, concurrently
affecting self-renewal and lineage-specific differentiation. As a result, it was
proposed that abnormal Notch signaling contributes to mammary carcinogenesis

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 265
by dysregulating the self-renewal of normal mammary stem cells (26). Additional
pathways thought to be involved in stem cell self-renewal include
hedgehog (HH), Bmi-1, wingless (Wnt/b-catenin) and phosphatase and tensin
homolog deleted on chromosome ten (PTEN) (11,20,31).
Further studies in human breast stem/progenitor cells identified a subset of
human breast cancer cells using flow cytometry with the phenotype CD44 þ /
CD24 /Lin that possess highly tumorigenic properties in a nonobese diabetic/
severe combined immunodeficiency (NOD/SCID) xenograft model (21). As few
as 200 cells with the epithelial-specific antigen (ESA) þ , CD44 þ /CD24 /Lin
phenotype consistently generated tumors in mice, whereas even 20,000 cells
with alternative phenotypes were ineffective. The isolation and serial passage of
the tumorigenic CD44 þ /CD24 /Lin cell population in NOD/SCID mice gave
rise to additional tumorigenic CD44 þ /CD24 /Lin cells, as well as phenotypically
diverse nontumorigenic cells, that composed the bulk of the tumors,
thereby confirming the self-renewal and differentiating capacities of these cells.
While the unequivocal demonstration of stem cell characteristics necessitates
model systems capable of tumor generation from a single cell, a population of
CD44 þ /CD24 /Lin cells appears to exhibit properties of human breast stem/
progenitor cells in NOD/SCID mice.
GENETIC SIGNATURES AND POLYMORPHISMS WHICH
MAY INFLUENCE BREAST CANCER DEVELOPMENT
AND TREATMENT OUTCOMES
The study of molecular breast cancer portraits identifies patterns of molecular
breast cancer markers using gene expression microarrays, which subsequently
are useful in the selection of therapy. These portraits also seek to overcome
potential differences in drug sensitivity or target expression between tumorinitiating
cells and the more frequent nontumorigenic cells (32). The rationale for
the pursuit of new drug discovery based on molecular targets results from the
significant increment of the effectiveness of these drugs as compared with
conventional chemotherapy. A recent systematic analysis of the consensus of
human breast and colorectal cancers included the analysis of 13,023 genes in
11 breast and 11 colorectal cancers, and revealed that individual tumors accumulate
an average of approximately 90 mutated genes, but that only a subset of
these contribute to the neoplastic process (33). Using stringent criteria, 189 genes
(average of 11 per tumor) were found to be mutated at significant frequencies.
The vast majority of these genes were not known to be genetically altered in
tumors and are predicted to affect a wide range of cellular functions, including
transcription, adhesion, and invasion. The genetic landscape of breast and colorectal
cancer defined in this study provides new targets for diagnostic and
therapeutic intervention.
Gene signatures of tumor invasiveness in tumorigenic CD44 þ CD24 /low
breast cancer cell populations correlate with overall and metastasis-free survival,

266 Balicki et al.
independently of established clinical and pathological variables (34).
CD44 þ CD24 /low breast cancer cells were also shown to increase from a median
of 4.8% to 14.8% after chemotherapy in paired breast cancer biopsies from
35 patients obtained before and after 12 weeks of neoadjuvant chemotherapy.
Microarray analyses of CD44 þ CD24 /low breast cancer cells using the Affymetrix
U133 GeneChip demonstrated an increased expression of self-renewal pathways.
An increase in mammosphere formation of residual tumors was also observed
postchemotherapy. Taken together, these results suggest that therapy resistant
cancer stem cells may exist in the neoadjuvant setting of breast cancer therapy, and
that breast cancer patients may benefit from the therapeutic targeting designed to
eradicate these cells (35).
In addition to chemoresistance, progenitor cells may also be resistant to
radiation. This radioresistance has been shown in an animal model to be
mediated at least in part by Wnt signaling, which has been implicated in stem
cell survival (36). Radioresistance was investigated by treating primary BALB/c
mouse mammary epithelial cells with clinically relevant doses of radiation. An
enrichment in normal progenitor cells [stem cell antigen (SCA) 1-positive and
side population progenitors] was identified.
Specifically, radiation selectively enriched for progenitors in mammary
epithelial cells isolated from transgenic mice with activated Wnt/b-catenin
signaling but not for background-matched controls. Irradiated stem cell antigen
1-positive cells had a selective increase in active b-catenin and survivin
expression compared with SCA 1-negative cells. Radiation also induced enrichment
of side population progenitors cells of the MCF-7 human breast adenocarcinoma
cell line. In comparison to differentiated cells, progenitor cells have
different cell survival properties that may facilitate the development of targeted
antiprogenitor cell therapies.
Closely intertwined with response prediction are an individual patient’s
genetic polymorphisms, which may increase breast cancer susceptibility and
influence the efficacy or safety of breast cancer treatments. The accumulation of
genetic factors that underlie breast cancer susceptibility or therapeutic escape
may trigger the transformation of a normal stem or progenitor cell into a cancer
stem cell. The variation of treatment responses between individuals also lies in
differences in DNA sequences that alter the expression or function of proteins
that are drug targets. These proteins are encoded by genes identified in early
studies that are linked to highly penetrant, single-gene traits. Germline polymorphisms
which confer susceptibility to breast cancer are associated with the
highly penetrant breast cancer genes BRCA1 and BRCA2, p53, PTEN, ataxia
telangiectasia mutated (ATM), and candidate BRCA3 genes (37). In addition,
BRCA1 has been suggested to function as a breast stem cell regulator (38). Thus,
BRCA1 may contribute to a conceptual link between hereditary and sporadic
breast cancer. The loss of BRCA1 function may produce a block in cell differentiation
and an increase in the self-renewal of mammary stem cells, resulting
in the expansion of the stem cell pool. Taken together with its known function in

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 267
DNA repair, BRCA1 may generate genetically unstable stem cells which are targets
for additional carcinogenic events and the aggressive triple negative “basal” stem
cell–like phenotype (39).
Low-penetrance genes that may influence breast cancer susceptibility
alone or in combination with environmental factors include genes in DNA repair,
metabolic pathways, and estrogen-related pathways including HRAS1 minisatellite,
CHEK2, XRCC1, XRCC2, RAD51, XRCC3, and LIG4 (37). Metabolic
enzymes such as N-acetyltransferases (NAT) and gluthathione-S-transferases
(GST) are involved in the metabolic inactivation of carcinogens. Mutations in
these enzymes can result in increased toxicity and exposure to mutagenic substances,
thereby increasing mutations, which contribute to carcinogenesis.
Superoxide dismutases (SODs) are metalloenzymes that protect cells from free
radicals, and their mutations can lead to increased breast cancer susceptibility
(37). Cytochrome p450s (CYP) metabolically activate carcinogens into chemically
reactive electrophiles. CYP1B1 is the most common form in the breast,
and its alanine and serine (ALA/SER) polymorphism significantly increases
breast cancer susceptibility (37). Additional targets of polymorphisms that
increase breast cancer susceptibility include COMT, CYP17, CYP19, PPAR-g,
TGF-b, HSP70-1, HSP70-2, and HSP70-HOM (37). Germline polymorphisms
may also influence an individual’s susceptibility to the effects of breast cancer
treatments (37). For example, dihydropyrimidine dehydrogenase (DPD) and
thymidylate synthase (TS) are pivotal to the metabolism of 5-FU, and specific
polymorphisms of these enzymes lead to 5-FU toxicity. Paclitaxel is metabolized
by cytochrome P450, and specific CYP polymorphisms lead to defective metabolism.
Additional breast cancer treatments with known polymorphisms, which
may effect drug efficacy or safety include tamoxifen (ER, CYP2D6, SULT1A1),
aromatase inhibitors (CYP19, CYP1A2, CYP2C9, CYP3A), cyclophosphamide
(GST), methotrexate (MTHFR), doxorubicin (GST), and epirubicin (UGT2B7)
(40). Future advances hinge on the more difficult challenge of elucidating multigene
determinants of breast cancer susceptibility and drug responses through an
intersection of genomics and medicine, which has the potential to yield a new set
of molecular diagnostic tools that can be used to individualize and optimize drug
therapy (41). Additional refinements can be accomplished through coupling in
vitro drug sensitivity data with microarray data, to develop gene expression signatures
that predict sensitivity to individual chemotherapeutic drugs (42). Thus, as
we continue to fine-tune our stratification, we also approach our ultimate goal of
defining individual molecular breast cancer portraits.
STEM CELL SANCTUARIES AS A MECHANISM
OF THERAPEUTIC RESISTANCE
Targeting of stem cells is particularly complex since this cellular population is
largely quiescent. Thus, therapeutic agents directed against cycling cells are
predictably ineffective in this population, which is essentially shielded from

268 Balicki et al.
these drugs. Cancers that respond to therapy initially may acquire drug resistance
during the course of treatment, while other cancers appear to be intrinsically
resistant. The cancer stem cell hypothesis suggests that in both of these cases, the
resting cancer stem cell, which is both the cancer-initiating cell and its source
of replenishment under selective pressure, has innate drug resistance by virtue of
its quiescent stem cell phenotype. Acquired drug resistance in more differentiated
cancer cells may occur through gene amplification or rearrangement,
thereby contributing to an aggressive phenotype. Through the accumulation of
mutations that distinguish cancer stem cells from their normal counterparts, these
cells may acquire additional features associated with tumor progression,
metastases, and therapeutic failure, including genetic instability, radioresistance,
and chemoresistance (11,36). In the case of gliomas, cancer stem cells contribute
to radioresistance through the preferential activation of the DNA damage
checkpoint response and an increase in DNA repair capacity (43). Radioresistance
may also result from a selective increase of cell survival pathways in
progenitor cells, including b-catenin and survivin (36).
Stem cell niches regulate stem cell proliferation and cell fate decisions, and
they also play a protective role by shielding stem cells from environmental
insults. In this manner, vascular niches might protect brain cancer stem cells
from chemo- and radiotherapies, and the opportunity for these cells to reform a
tumor mass in spite of an initial clinical response (44). With the acquisition of
properties of tumor progression and metastases, cancer stem cells may travel to
sanctuary sites, such as across the blood-brain barrier, where it may be difficult
to eradicate them, even with targeted therapies.
The drug resistance of normal tissue stem cells is constitutively mediated
by multidrug resistance (MDR) transporters and detoxifying enzymes (45). DNA
repair mechanisms, resistance to apoptosis, and telomerase activity also contribute
to the stability of normal tissue stem cells. Among the mechanisms
responsible for stem cell drug resistance, there is mounting speculation that ABC
transporters repress the maturation and differentiation of stem cells (46,47).
However, since breast cancer resistance protein 1 (BCRP1) knockout mice
display normal hematopoiesis instead of accelerated hematopoietic maturation,
it is clear that multiple factors determine the maturation and differentiation of
stem cells (48). In addition, protection from hypoxia appears to be another
function of ABC transporters. As a result, it has been postulated that Bcrp
expression protects hematopoietic stem cells from hypoxic environments by
preventing porphyrin accumulation that causes mitochondrial death (49).
Thus, leading edge-targeted breast cancer therapies must overcome all of
the obstacles associated with cancer stem cell biology, including both the
intrinsic and extrinsic properties of resistance, which define these enigmatic cells
and the cloisters that keep them safe. From the cancer stem cell hypothesis, it
follows that systemic administration of an efficacious MDR reversal agent would
render tumor and normal tissue stem cells equally susceptible to the chemotherapeutic
agents, offering no net gain in therapeutic index (45). Thus, targeting

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 269
the properties that distinguish cancer stem cells from their normal stem cell
counterparts is necessary.
CELLULAR BREAST CANCER STEM CELL TARGETS
The identification of cancer stem cell markers is necessary for the isolation
and purification of cancer stem cells as they are potential targets of novel drug
discovery and development. The functional targets that define stem cells are
their abilities to self-renew and differentiate using serial passages in xenotransplantation
models. The current breast cancer stem cell hypothesis suggests that
the most primitive human breast stem cell does not express the estrogen receptor
(ER) while more mature progenitors may or may not express ER depending on
their location in the differentiation pathway (12). A recent study using reverse
transcriptase polymerase chain reaction (RT-PCR) and immunohistochemical
analyses established that mouse mammary stem cells were negative for both ERa
and the progesterone receptor (50). An independent study further supported this
result by showing that ER-positive cells retained very little stem cell activity (51).
Throughout much of mammary development during embryogenesis, the developing
gland consists of a simple duct system, with a primitive mammary bud, the
growing terminal end of the duct, with no ERa expression until the 30th week of
gestation. These structures are thought to be made of primitive ductal and
myoepithelial progenitors. Later in gestation, ERa-positive cells are present in
the growing terminal ends of these ducts, followed by a surge of ERa-positive
cells in this region after birth (12). In the developing mammary gland, insulin
growth factor-1 (IGF-1) is required to form ducts and terminal end buds, indicating
its critical role in stimulating the proliferation and differentiation of
primitive breast epithelium (12).
ER-positive progenitors arise after 30 weeks of gestation. They respond to
estrogen stimulation by proliferating and generating ductal epithelial cells and
alveolar cells that form the adult mammary tree. Maintenance of the myoepithelial
progenitor pool and the stem cell compartment may rely on paracrine
signaling initiated by the ER-positive progenitor cells. Breast carcinogenesis
results from mutations affecting stem and progenitor cells. Thus, the phenotype
of different subtypes of breast cancer is determined both by the cell of origin and
the particular mutations driving carcinogenesis. Thus, ER-negative tumors arise
from stem cells or early ER-negative progenitor cells, while ER-positive tumors
arise from either ER-negative or ER-positive progenitors (12).
On the basis of these observations, a breast tumor model has been proposed
consisting of three types of tumors. The first is an ER-negative tumor termed
type 1, which arises from the most primitive ER-negative stem or early progenitor
cells. Using classic histopathology and molecular profiling studies, these
tumors are poorly differentiated and display a basal phenotype with expression
markers of both luminal epithelial and myoepithelial cells. Clinically, these
tumors are highly aggressive and are associated with a poor prognosis. Type 2

270 Balicki et al.
tumors also arise from early ER-negative stem or early progenitor cells. However,
this group includes stem cell mutations, which allow for the differentiation
of a specific subset of tumor cells into ER-positive cells, which may range from a
low percentage (6–10%) similar to that seen in normal tissue to a high percentage
(70–100%). Thus, these tumors are heterogeneous, with intermediate differentiation
and prognosis. Type 3 tumors arise from the transformation of ER-positive
progenitors. They are predicted to exclusively express luminal markers and have
the best prognosis as they contain differentiated cells and respond to antiestrogen
therapy (12). Additional breast stem cell markers include CD44, CD24, aldehyde
dehydrogenase (ALDH), Oct-4, Bmi-1, and Wnt/b-catenin.
As discussed above, a small population of cancer cells characterized by
CD44 expression but low or undetectable levels of CD24 (CD44 þ CD24 /low )
have a high tumorigenic capacity when injected into immunodeficient mice (52).
In contrast, the rest of the cancer cells, called nontumorigenic breast cancer cells,
have little or no tumorigenic ability. Tumors in mice that originate from purified
tumorigenic breast cancer cells contain a mixture of both tumorigenic and
nontumorigenic breast cancer cells, which confirms that the CD44 þ CD24 /low
population shares the capacity for self-renewal with normal stem cells. Gene
expression profiling of tumorigenic breast cancer cells was utilized to develop
prognostic tools to assess survival in patients with breast cancer and to predict
clinical outcomes in breast cancer (34). An invasiveness gene signature of 186
genes was developed to predict the likelihood of a tumor to metastasize. This
invasiveness gene signature is associated with the risk of death and metastasis
not only in breast cancer but also in lung cancer, prostate cancer, and medulloblastoma.
These results suggest the critical relevance of the CD44 þ CD24 /low
tumorigenic subclass of breast cancer cells as a pivotal contributor to the association
of the invasiveness signature to clinical outcome. In addition, CD44, a
cell-surface receptor that mediates cell–cell contacts, promotes the engraftment
of leukemic stem cells in the bone marrow and may thus have an additional role
as a critical component of the cancer stem cell niche (53).
As a result of the growing interest of identifying new markers for identifying
and isolating stem cells, assays based on ALDH activity have been proposed
to be a promising alternative in both murine and human mammary stem
cells. ALDH is a family of cytosolic enzyme isoforms responsible for oxidizing
intracellular aldehydes, including vitamin A, to carboxylic acids. It is highly
expressed in hematopoietic progenitors, in intestinal crypt cells, as well as in
breast tumor cells (54–56). ALDH1 or (ALDH isoenzyme 1 or ALDH1A1) has
been shown to be responsible for the resistance observed in hematopoietic stem
cells and breast cancer tumor cells to the alkylating agent, cyclophosphamide
(56–58). A novel system has been designed to detect ALDH using a visible light
excitable fluorochrome named Aldefluor 1 (BODIPY-conjugated aminoacetaldehyde,
BAAA) metabolized by both murine and human cells (59).
Fluorescence-activated cell sorter (FACS) of the normal breast cell population
revealed that 6% of this population is ALDH-positive (60). In breast cancer cell

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 271
lines, ALDH-positive cells were more frequently observed in basal-like than in
luminal cell lines. ALDH-positive cells form mammospheres at a high efficiency
and are tumorigenic in NOD/SCID mice (61).
A recent study showed that when acute myeloid leukemic (AML) patient
cells were sorted on the basis of Aldefluor staining, three distinct possibilities
were observed (62). High levels of aldehyde dehydrogenase (ALDH hi ) cells from
a first group of patients were identified as normal nonleukemic stem cells, were
present in low number, and were able to reconstitute a complete hematopoietic
system when transplanted into NOD/SCID mice. However, a second group of
patients had ALDH hi cells displaying a higher side scattering pattern than normal
stem cells and these cells were also significantly more abundant. These cells
correspond to leukemic stem cells. Finally, a third group of patients showed no
ALDH hi cells. This study demonstrates that FACS analyses based on the first
ALDH isozyme (ALDH1 or ALDHA1) activity can be used to identify and
isolate leukemic stem cells (58). In contrast to the toxicity associated with
Hoechst 33342, ALDH is relatively mild and allows for cell sorting using viable
cell populations, enriched in stem cells. The availability of antibodies derived
against ALDH1 allows for the study of their expression in fixed tissue samples
using immunohistochemistry (Fig. 4). Immunohistochemistry of 577 human
breast carcinoma specimens from two different sources detected 24% and 30%
ALDH-positive cells, respectively. In this study, ALDH expression was strongly
correlated with tumor size, grade, ER status, human epidermal growth factor
receptor 2 (HER2) overexpression, and poor prognosis (60).
OCT-4 is an additional cancer stem cell marker. The OCT-4 oncogene
encodes a nuclear protein that belongs to a family of transcription factors containing
the POU DNA–binding domain that plays an important role in maintaining
the pluripotent state of embryonic stem cells (63). It may prevent expression of
genes activated during differentiation, and its expression is downregulated during
differentiation. OCT-4 is normally found in the pluripotent stem cells of pregastrulation
embryos, including oocytes, early cleavage-stage embryos, and the inner
cell mass of the blastocyst, and knockout of OCT-4 causes early lethality in mice
because of the absence of an inner cell mass (63). Thus, these activities associated
with OCT-4 suggest that it plays a pivotal role in mammalian development and in
the self-renewal of embryonic stem cells. The expression of genes such as OCT-4
is potentially correlated with tumorigenesis and may affect some aspects of tumor
behavior, such as tumor recurrence or resistance to therapies. OCT-4 is expressed
in bladder cancer, as well as in CD44 þ /CD24 breast cancer–initiating cells
(64,65). It may serve as a multifunctional factor involved in major biological
processes such as embryonic development, control of differentiation, and stem
cell–based carcinogenesis (64).
Bmi-1 is a cancer stem cell marker that belongs to the polycomb group
family of proteins, and was first identified as a c-myc cooperating oncogene in
murine lymphomagenesis. Bmi-1 is involved in the regulation of diverse biological
processes involved in X chromosome inactivation, repression of cellular

272 Balicki et al.
Figure 4 ALDH1 immunostaining. Immunohistochemistry using anti-ALDH1 on tissue
microarray samples from two patients (A, B) with infiltrating ductal carcinoma of the breast,
both magnified 200, with a scale bar of 100 mm. Abbreviation: ALDH1, aldehyde dehydrogenase
isoenzyme 1.
senescence, carcinogenesis, and stem cell renewal (66). Bmi-1 oncoprotein
overexpression correlates with axillary lymph node metastases in invasive ductal
breast cancer (67). In addition, the Bmi-1 oncogene induces telomerase activity
and immortalizes human mammary epithelial cells (68). Bmi-1 regulates selfrenewal
of normal and malignant human mammary stem cells, stimulates
mammosphere formation and colony size in vitro, and increases ductal/alveolar
development in humanized NOD/SCID mammary fatpads (20).
The final stem cell marker in this chapter is wingless (Wnt)/b-catenin. The
canonical Wnt/b-catenin signaling pathway modulates the delicate balance
between stemness, differentiation, and tumorigenesis in several adult stem cell

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 273
niches such as the hair follicles in the skin, the mammary gland, hematopoiesis,
and the intestinal crypt (8,69). Differences in the levels of Wnt signaling activity
reflect tumor heterogeneity and are likely to account for distinct cellular activities
such as proliferation and epithelial-mesenchymal transitions, which prompt
tumor growth and malignant behavior, respectively. This pivotal pathway is
highly conserved in evolution and is known to regulate cell-fate decisions, cell
proliferation, morphology, migration, apoptosis, differentiation, and stem cell
self-renewal. Recent evidence which suggests that Wnt signaling plays a role in
human breast cancer include the detection of elevated levels of nuclear and/or
cytoplasmic b-catenin using immunohistochemistry, and the overexpression or
downregulation of specific Wnt proteins. The Wnt pathway has also been
implicated in normal stem cell self-renewal in vivo, and there is evidence that
dysregulation of this pathway in the mammary gland and other organs may play
a key role in carcinogenesis (70).
The Wnt family of secreted proteins includes the well-characterized
canonical Wnt signaling pathway, in which Wnt ligands signal through the
stabilization of b-catenin. In this pathway, Wnt proteins bind to a family of
frizzled receptors in a complex with the low-density lipoprotein receptor–related
proteins 5 and 6 (LRP5/6) coreceptors to activate Dishevelled (Dsh). Subsequently,
Dsh inhibits the activity of the b-catenin destruction complex [adenomatous
polyposis coli (APC), axin, and glycogen synthase kinase-3b (GSK-
3b)], which phosphorylates b-catenin in the absence of the ligands. As a result,
b-catenin is stabilized and translocated to the nucleus, where it recruits transactivators
to high mobility group (HMG)–box DNA-binding proteins of the
lymphoid-enhancer factor/T-cell factor (LEF/TCF) family. In the absence of
Wnt signaling, b-catenin remains in the cytoplasm, where it forms the b-catenin
destruction complex. GSK-3b phosphorylates b-catenin, which targets the protein
for ubiquitin-mediated degradation. When the Wnt pathway is activated,
GSK-3b is inhibited, blocking b-catenin phosphorylation and its subsequent
degradation (71). In addition, several b-catenin-independent Wnt signaling
pathways, known as noncanonical, have been shown to be crucial for different
aspects of vertebrate embryo development (71).
MOLECULAR BREAST CANCER STEM CELL TARGETS
Molecular signatures of stem cells are currently being identified. The mutations that
accumulate in cancer stem cells could inappropriately activate Bmi-1, Wnt/b-catenin,
Notch, sonic hedgehog (SHH), and PTEN pathways that are involved in the regulation
of self-renewal of normal stem cells and could potentially lead to cancer
(Figs. 5 and 6) (26,72–74). These molecular markers can be exploited to purify
cancer stem cells, which can then be characterized in the most intimate detail,
by the modern tools of gene expression profiling and informatics. The failure to
shut off the self-renewal pathways could confer self-renewal on cells and microenvironments
capable of evading homeostatic regulation, leading to tumorigenesis.

274 Balicki et al.
Figure 5 Stem cell self-renewal pathways include the HH, Notch, Bmi-1, Wnt, and
PTEN pathways. Dysregulation of the self-renewal pathways of NSCs and progenitor
cells may lead to CSCs and tumor formation. NSCs are ER negative and they differentiate
into ER-positive progenitor cells, which further differentiate into myoepithelial, alveolar
epithelial, and ductal epithelial cells. CSCs in their niche may lose asymmetric division,
leading to the clonal expansion of cancerous cells and tumor formation. Abbreviations:
NSCs, normal stem cells; CSCs, cancer stem cells; ER, estrogen receptor.
The intimate interrelationship of multiple pathways critical to tumorigenesis and
to normal stem cell self-renewal raises the specter that cancer therapies targeting
such pathways might ablate resident stem cells (75). Drugs that selectively target
the self-renewal pathways of tumors will be attractive candidates for cancer
therapies if they can effectively spare the normal bystander stem cells that use
self-renewal as part of their regular functional repertoire. Alternatively,

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 275
Figure 6 A model of the interrelationship of stem cell self-renewal pathways including a
conceptual link between sporadic and hereditary breast cancers. BRCA1 may inhibit stem
cell self-renewal, while HER2 overexpression may increase the stem cell pool through
regulation of self-renewal. In this model, there is bidirectional cross talk between the
Notch and HH pathways with subsequent regulation of the polycomb gene Bmi-1 and the
Wnt/b-catenin pathway. SHH, IHH, and DHH are HH ligands, while Gli1 and 2 are two
HH transcription factors that are positive mediator of HH signaling. Cylcopamine inhibits
the HH pathway. The DSL peptide activates Notch signaling, while the GSI blocks the
intramembrane cleavage of Notch required for signaling. PTEN affects Wnt signaling
through GSK-3b. Exisulind, bromoindirubin-3 0 -oxime, and imatinib inhibit the Wnt/bcatenin
pathway. Abbreviations: BRCA, breast cancer; HER, human epidermal growth
factor receptor 2; HH, hedgehog; SHH, sonic hedgehog; IHH, Indian hedgehog; DHH,
desert hedgehog; PTEN, phosphatase and tensin homolog deleted on chromosome ten;
GSI, gamma secretase inhibitor.

276 Balicki et al.
differentiating therapies may also be effective if they can efficiently shift the
balance of stem cell divisions toward differentiation rather than self-renewal and
stem cell expansion. In addition to Bmi-1, and Wnt/b-catenin discussed above,
Notch, SHH pathways, and PTEN will be reviewed below.
The Notch signaling pathway plays an important role in regulating proliferation,
survival, and differentiation of many cell types, including the regulation
of self-renewal of adult stem cells and differentiation of progenitor cells
along a particular lineage. These outcomes include increased survival or death,
proliferation or growth arrest, and commitment to or blockage of differentiation.
There are four mammalian Notch homologues, Notch 1 to 4, which interact with
the DSL ligands: Delta, Delta-like, Jagged1, and Jagged2 in vertebrates (71).
Notch 1 and Notch 4 homologues are involved in the normal development
of the mammary gland, and mouse mammary tumors are associated with their
mutated forms. Notch signaling is critical to normal human mammary development
where it acts on both stem cells and progenitor cells, affecting selfrenewal
and lineage-specific differentiation. Notch signaling may also contribute
to mammary carcinogenesis by dysregulating the self-renewal pathways of
normal mammary stem cells. There is bidirectional cross talk between the Notch
and HH pathways with subsequent regulation of the polycomb gene Bmi-1, as
well as the Wnt signaling network (71). The DSL ligands activate Notch signaling,
while the gamma secretase inhibitor (GSI) blocks the intramembrane
cleavage of Notch required for signaling (71). The abnormal expression of Notch
receptors in epithelial metaplastic lesions and neoplastic lesions suggests that
Notch may act as a protooncogene (26). In spite of the simplicity of the Notch
signaling pathway, it yields varied outcomes in different contexts (76). These
different outcomes are mediated through a novel signaling pathway in which
Notch receptors on the cell surface are processed to give rise to a nuclear
transcriptional activation complex. Activated Notch modulates stem and progenitor
cell renewal through the upregulation of the cyclin-dependent kinase
inhibitors p21 cip1 and p27 kip1 (77). Notch signaling is also required for the
survival of niche cells (78). Notch receptor activation promotes the survival of
neural stem cells by inducing the expression of the specific target genes hairy and
enhancer of split 3 (Hes3) and SHH. The underlying mechanism occurs through
the rapid activation of cytoplasmic signals, including the serine/threonine kinase
AKT, the transcription factor STAT3, and mammalian targets of rapamycin
(mTOR) molecular targets of rapamycin (79).
The HH signaling pathway has important roles in embryonic development
and tumorigenesis (80). In a number of tumors, including prostatic and pancreatic
cancers, HH signaling is abnormally upregulated, and effective signaling
inhibition can repress tumor growth and induce apoptosis (81–83). The core
components of the HH pathway are evolutionarily conserved from insects to
mammals. They include the secreted ligand HH, the membrane receptor Patched
(PTCH), the membrane signal transductor Smoothened (SMO), and the transcription
factor Ci/glioma (GLI). However, while in Drosophila, there is only

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 277
one secreted ligand (HH), one receptor (Ptc), and one transcription factor (Ci),
there are three secreted ligands, sonic HH (SHH )/desert HH (DHH)/Indian HH
(IHH), two receptors (PTCH1/2), and three transcription factors (GLI1/2/3) in
mammals (81). The HH signaling components PTCH1 and GLI1/2 are highly
expressed in normal mammary stem and progenitor cells, and activated HH signaling
is mediated by Bmi-1 (30). Addition of Hh ligands increases the expression
of the transcription factors Gli1 and Gli2, which are positive mediators of Hh
signaling (20). SHH can potentiate tumor cell proliferation. Components of the
SHH–GLI pathway are expressed in adult human prostate cancer, often with
enhanced levels in tumors versus normal prostatic epithelia. Cyclopamine and
anti-SHH antibodies inhibit the proliferation of GLI1–PSA primary prostate tumor
cultures (83). The SHH signaling pathway also regulates adult neuronal selfrenewal,
and may play a role in brain tumor biology (4).
The final pathway discussed herein is PTEN, which is a phosphatase that
inhibits cellular proliferation by negative regulation of signaling through the
phosphatidylinositol-3-OH kinase (PI3K) pathway. The activation of PI3K leads
to phosphorylation and activation of the Akt protein, which in turn can potentially
phosphorylate a multitude of proteins (84). The PTEN–Akt pathway regulates
stem cell activation by helping control nuclear localization of the Wnt
pathway effector b-catenin. PTEN inhibits, whereas Akt enhances, the nuclear
localization of b-catenin. Akt is thought to coordinate with Wnt signaling to
assist in the activation of b-catenin in intestinal stem cells, either through
phosphorylation of GSK-3b or by phosphorylation of b-catenin itself. The
phosphorylation by Akt inactivates GSK-3b, helping the Wnt signal to permit
b-catenin to escape proteasome-mediated degradation and facilitating its nuclear
translocation and activity in transcriptional regulation (85).
In addition to other downstream effectors, the highly branched PI3K
pathway activates the mammalian target of rapamycin (mTOR). Molecules such
as p21 Cip1/Waf1 are found downstream to PTEN, and they influence stem cell
proliferation, without necessarily implicating self-renewal. FOXO regulatory
proteins may have the effects of PTEN on p21 Cip1/Waf1 by influencing the
expression of a range of crucial genes for stem cell function, including cell
division, and tolerance of oxidative stress. Thus, the perturbation of PTEN may
thus result in nonrenewing cell divisions or death of stem cells (84). PTEN has
also been reported to be a guardian of genome integrity and of the maintenance
of chromosomal stability through the physical interaction with centromeres and
control of DNA repair. PTEN acts on chromatin and regulates the expression of
Rad51, which reduces the incidence of spontaneous double strand breaks (86).
PTEN is commonly deleted or otherwise inactivated in diverse cancers,
including hematopoietic malignancies, glioblastomas, endometrial carcinomas,
breast carcinomas, and prostate carcinomas (86). PTEN deletion or HER2
overexpression may modulate stem cell function leading to increased invasion
and tumorigenicity (87). PTEN deletions are implicated in the generation of
transplantable leukemia-initiating cells, and the depletion of normal hematopoietic

278 Balicki et al.
stem cells (88). PTEN normally maintains hematopoietic stem cells in a quiescent
state, and in the absence of PTEN, hematopoietic stem cells are driven into the cell
cycle, eventually leading to depletion of hematopoietic stem cell reserves. Thus,
PTEN activity identifies a mechanistic difference between the maintenance of
normal stem cells and cancer stem cells. These effects were mostly mediated by
mTOR as they are inhibited by rapamycin, which depletes leukemia-initiating
cells but also restores normal hematopoietic stem cell function (88). This mechanistic
distinction shall enable the design of therapies that target the PTEN
pathway to combat leukemia with the goal of killing leukemic stem cells, without
damaging the normal stem cell pool (75). In addition to leukemia, this pathway
may be an attractive therapeutic target for treating a broad spectrum of tumors
carrying inactivated PTEN because mTOR activity can be suppressed by a number
of compounds including rapamycin (75). Genes that promote stem cell quiescence
might be particularly amenable to such targeted strategies because although normal
stem cells thrive in a quiescent world, forcing cancer stem cells to adopt
quiescence could lead to their eradication (75).
TARGETS OF THE STEM CELL NICHE
While cancer stem cells may indeed be lethal seeds in themselves, they may
whither and die unless they are supported by environmental factors, which
perpetuate their survival and self-renewal. The stem cell niche provides this
microenvironment for cancer stem cells, and the combination of lethal seeds in
lethal soil is likely to lead to the most aggressive forms of breast cancer, with the
most fatal course. The stem cell niche determines cell fate decisions based on
surrounding cells, extracellular matrix components, and secreted local and systemic
factors (11). The 1889 proposal by Stephen Paget that metastasis depends
on cross talk between selected cancer cells (the “seeds”) and specific organ
microenvironments (the “soil”) still holds forth today, and may also apply to the
interrelationship between cancer stem cells and their niche (89–91). Cancer
metastasis requires the seeding, homing, and successful colonization of specialized
cancer stem cells at distant sites (92). In this context, it is not surprising
that the inhibition of the homing factor CXCR4 has been shown to effectively
prevent both primary tumor formation as well as metastasis in animal models
(92). Stem cells of various tissues exist within protective niches that are composed
of differentiated cells which provide direct cell contacts and secreted
factors that maintain stem cells primarily in a quiescent state by providing cues
for the inhibition of proliferation (8,44). The mutations that distinguish cancer
stem cells from normal stem cells may enable them to escape niche control.
Alternatively, the niche provides proliferation or differentiation-inducing signals
at times when high numbers of progenitors are needed and can quickly give rise
to all committed cell lineages (8). As a result, the uncontrolled proliferation of
stem cells and tumorigenesis may be a consequence of the dysregulation of
extrinsic factors within the niche. In this case, the stem cell niche might represent

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 279
targets for cancer therapy (44). Given that postnatal estrogen exposure is related to
breast cancer risk, this effect could also apply in utero, where the developing fetus
is exposed to estrogen levels up to 10-fold higher than at any other time in the
individual’s lifetime. Under these circumstances, estrogens or other hormones may
create the fertile soil that is responsible for an increase in the responsiveness of
estrogen-responsive tissue to undergo oncogenesis later in life (93).
Epigenetic modulation of gene expression is highly abnormal in cancers,
which often show aberrant promoter CpG island hypermethylation and transcriptional
silencing of tumor suppressor genes and prodifferentiation factors.
The aberrant epigenetic landscape of the cancer cell is also characterized by a
massive genomic hypomethylation, an altered histone code for critical genes, and
a global loss of monoacetylated and trimethylated histone H4 (94). In embryonic
stem cells, these genes are maintained in a state that is ready for transcription
mediated by a bivalent promoter chromatin pattern consisting of the repressor,
histone H3 methylated at Lys27 (H3K27) by polycomb group proteins, plus the
activator, methylated H3K4. However, in contrast to embryonic stem cells,
embryonic carcinoma cells add two key repressors, dimethylated H3K9 and
trimethylated H3K9, which are both associated with DNA hypermethylation in
adult cancers (14). Polycomb group proteins reversibly repress genes required
for differentiation in embryonic stem cells. Stem cell polycomb group targets are
up to 12-fold more likely to have cancer-specific promoter DNA hypermethylation
than nontargets, supporting a stem cell origin of cancer where
reversible gene repression is replaced by permanent silencing, locking the cell
into a perpetual state of self-renewal, thereby predisposing to subsequent
malignant transformation (95). Thus, cell chromatin patterns and transient
silencing of these important regulatory genes in stem or progenitor cells may
leave these genes vulnerable to aberrant DNA hypermethylation and heritable
gene silencing during tumor initiation and progression.
Vascular niches have been described in the case of neural stem cells, which
were found to be concentrated in regions that are rich in blood vessels (18).
These vessels shelter neural stem cells from apoptotic stimuli and maintain a
proper balance between self-renewal and differentiation. The endothelial cells
that line blood vessels are key components of the vascular niche as they secrete
factors that promote stem cell survival and self-renewal. As the cotransplantation
of tumor cells with endothelial cells leads to more rapid tumor formation, the
vascular niche may contribute to tumor initiation and development (44). In an
established tumor, cancer stem cells that find a vascular niche may continue to
self-renew, while the remaining cells may differentiate to form the bulk of the
tumor, but not to its long-term maintenance (18). Thus, the identification of
endothelial cell signals, which regulate the self-renewal of cancer stem cells,
may also contribute to our understanding of the interrelationship of cancer stem
cells and their niche. There is some evidence that there is a bidirectional relationship
between cancer stem cells and their niche. For example, gliomas secrete
elevated levels of vascular endothelial growth factor (VEGF), which increases

280 Balicki et al.
endothelial cell migration and tube formation (43). In addition, VEGF promotes
neural stem cell survival (96). Thus, the effectiveness of antiangiogenic agents in
cancer therapy may relate to the prevention of new blood vessels that inhibit
tumor growth. In addition, antiangiogenic therapy may disrupt a vascular niche,
which is necessary for cancer stem cell renewal (18).
In the case of hematopoietic stem cell niches, studies have focused on the
endosteum, the inner surface of the bone that interfaces with the bone marrow.
Regulation of hematopoietic stem cell function is governed by factors secreted
by osteoblasts, osteoclasts, and stromal fibroblasts within the endosteum (8).
Imaging data suggests that hematopoietic stem cells reside near the endosteum,
and that this localization and maintenance is also usually adjacent to CXCL12-
secreting reticular cells, irrespective of whether the hematopoietic stem cells
are in vascular or endosteal locations (97,98). The stimulation via parathyroid
hormone–related protein receptors (PPR) or an increase in number of osteoblastic
cells in the periosteum with high levels of the Notch ligand Jagged1,
enlarged the stem cell niche with a resultant increase in the number of hematopoietic
stem cells with evidence of Notch1 activation in vivo (99). The
activation of Notch1 has been shown to result in enhanced self-renewal of
hematopoietic stem cells possibly by inducing self-renewal genes including
Hes1 (8). In addition, the bone morphogenetic protein signal has an essential
role in inducing haematopoietic tissue during embryogenesis within the hematopoietic
stem cell niche (100). Mesenchymal stem cells have been shown to
improve engraftment of hematopoietic stem cells, and suppress graft-versushost
disease after allogeneic hematopoietic stem cell transplantation. In addition,
when tumor cells are injected into NOD/SCID mice in conjunction with
mesenchymal stem cells, their growth is much faster as compared to the group
receiving only tumor cells. To explain this observation, it was suggested that
mesenchymal stem cells have the ability to form a cancer stem cell niche in
which tumor cells can preserve their proliferative potential to sustain the
malignant process (101).
Activated, irradiated, mammary fibroblasts are critical to the microenvironment
of mammary stem cells as they are required for the successful
repopulation of cleared mammary fat pads in immunocompromised mice (102).
These epithelial-stromal interactions are also important in malignant transformation
and likely to be important with cancer stem cells (11). Fibroblasts secrete
matrix metalloproteinases and protease inhibitors and mobilize and activate
growth factors such as tumor growth factor-b (TGF-b), which has been shown to
promote tumor progression and invasion (103). Carcinoma-associated fibroblasts
promote angiogenesis, as well as tumor growth, by secreting stromal cell-derived
factor-1 (SDF-1), which interacts with the chemokine receptor CXCR4 on tumor
cells and endothelial cells (104). As radiographic breast density correlates with
increased breast cancer risk, as well as increased fibroblasts and collagen
deposition, this modified niche may promote the transformation of initiated stem
cells into cancer stem cells (11,105).

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 281
FUTURE CHALLENGES: INDIVIDUALIZED THERAPY BASED
ON BREAST CANCER STEM CELLS AND THEIR NICHE
In the last few years, there has been a significant evolution in breast cancer care, if
not a revolution (106). Whereas, we previously focused on tumor size, nodal
status, and certain prognostic features based on tumor pathology, the paradigm has
now shifted to include the underlying molecular pathways that drive cancerigenesis,
as well as the cellular targets of breast cancer, including cancer stem cells
and their niche. The intricate circuitry of pathways that define the hallmarks of
cancer and the corresponding complex molecular interactions are being translated
into therapeutic benefit (107,108). Individualized drivers are gradually being
factored into patient assessments in order to astutely and accurately determine the
clinical course of breast cancer patients on the basis of their unique and complex
cellular and molecular breast cancer portraits.
Breast cancer stem cell markers and the critical molecules in the pathways
that drive their self-renewal are necessary to complete the molecular portrait of
breast cancer. Our failure to eradicate most cancers may be as fundamental as a
misidentification of the target cancer stem cell (Fig. 3) (6). The fact that stem cells
rarely divide and have unique cellular properties, coupled with the observation that
they may have high levels of drug transporters to pump chemotherapy agents out
of the cell, has led many to believe that traditional chemo- or radiation therapies
are insufficient to clear these tumor-initiating cells from the body (92). Clinical
observations indicate that breast cancer responses to therapy do not necessarily
correlate with patient survival and are incomplete clinical endpoints. Instead, the
effectiveness of these treatments should be evaluated by decreased cancer recurrence
and metastases as measures of therapeutic efficacy (92). As the molecular
pathways that govern normal and cancer stem cells overlap, the targeting of these
pathways could have harmful effects on the homeostatic function of normal stem
cells. Targeted therapies that specifically eliminate cancer stem cells could further
revolutionize the ways by which we treat cancer.
As the proliferation of cancer stem cells is important for the growth of the
primary tumor, as well as progression to metastatic disease, inhibition of the selfrenewal
capacity of these cells is an obvious target for their elimination. Therapies
such as cyclopamine (targeting the HH signaling) and exisulind, bromoindirubin-3 0 -
oxime, and imatinib (targeting the Wnt/b-catenin pathways) were designed to
inhibit self-renewal pathways critical to cancer stem cells, and have achieved
varying levels of success (92). Another strategy to eliminate self-renewal is to force
cancer stem cells to differentiate (Fig. 6).
As cancer stem cells are currently being identified in a broad spectrum of
cancers (44,109–112), significant advances in their fundamental properties are
forthcoming, and will serve as a scaffold to expedite our understanding and identification
of cancer stem cells within specific types of cancers, including breast
cancer. The existence of overlapping cancer stem cell pathways within different
cancer types will also stimulate the discovery and development of new drugs to

282 Balicki et al.
target these pathways, as they are likely to be effective and marketable in more than
merely a small subset of cancer patients.
The addition of molecular markers in our patient assessments acknowledges
the pivotal role that the corresponding molecular pathways play in fueling
tumorigenesis. This fine-tuning of patient prognostic factors will ensure that the
oncogenic drivers of each individual patient’s tumor pathophysiology are both
identified and tailored into the definition and optimization of their specific
therapeutic strategy, which will translate into their best possible clinical outcome
and overall survival. As a result, it is critical that our therapeutic armamentarium
target the specific fingerprint of cellular and molecular pathways defined for an
individual patient. Therapeutic resistance via quiescence and the sequestration of
cancer stem cells into sanctuaries that are impermeable to most therapeutic
agents must be taken into consideration in the design of individualized breast
cancer therapies. Therapeutic resistance may also result from an inadequate
depletion of key cellular and molecular elements within the cancer stem cell
niche. The elucidation of the secreted factors that set up the premetastasis and
metastatic niche could provide both diagnostic as well as therapeutic benefits.
Drugs such as Bevacizumab (Avastin 1 ) that reduce intratumoral pressure
associated with leaky tumoral vasculature and thereby increase the introduction
of intravenous agents within the tumor bed may have the critical synergy necessary
in combinations of targeted agents to permeate cancer stem cell sanctuaries
and cancer stem cell niches (18,44,113,114). In addition, they may target
both the vascular niche and the associated cancer stem cells (44). Thus, it is not
surprising that clinical trials of Bevacizumab combined with the chemotherapeutic
drug CPT-11 suggest that this combination is one of the most
effective treatments of glioblastoma (115). Other agents that stimulate cycling of
quiescent cancer stem cells may also be beneficial in combined therapies, which
include agents that eradicate cycling cells. In addition, individualized therapy
must also have the long-term objectives of predicting and preparing to eradicate
the molecular pathways that may circumvent those pathways that drove the
initial breast cancer pathophysiology, and were targeted by first-line targeted
therapies. This strategy also underlines the critical objective of eradicating every
single cancer stem cell and its niche in newly diagnosed breast cancer patients,
thereby striving to eliminate all possibilities that these lethal seeds in lethal soil
may ignite tumor recurrence in the future.
Therefore, the genesis of individualized breast cancer therapy including
the eradication of cancer stem cells has become a necessity to achieve optimal
breast cancer patient care. It is imperative to understand the biology of cancer
stem cells and their niche to optimize our cancer treatment strategies to prevent
the longevity of cancer stem cells from interfering with our own (18). The
associated costs of its future reality precipitate a call to rethink our healthcare
strategies with a focus on cancer prevention and therapeutic schedules which
deliver targeted agents when they are both necessary and most effective in our
quest to optimally treat an individual breast cancer patient with the ultimate goal

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 283
of a stable, long-term remission. In this setting, the identification and targeting of
breast cancer stem cells and their niche hold great promise in the development of
new therapeutic translational strategies using state-of-the-art individualized
therapies for breast cancer patients.
ACKNOWLEDGMENTS
Rhyna Salinas (CHUM) is gratefully acknowledged for her expert assistance in
the preparation of this chapter. The contributions made by Drs. Francine
Jolicoeur, Isabel Garcia de la Fuente, and Louis Gaboury (CHUM) to Figure 4
during their collaboration with D.B. are greatly appreciated. Drs. Connie Eaves,
Gabriela Dontu, Suling Liu, Christophe Ginestier, Hasan Korkaya, and Julie
Dutcher are thankfully acknowledged for insightful and inspirational discussions.
Dr. Connie Eaves is also gratefully recognized for her invaluable role
as D.B.’s first stem cell research mentor and role model, who initially planted the
seed that she continues to nourish. D.B. is a scholar of Fonds de la recherche en
santé du Québec (FRSQ) and principal investigator of the National Cancer
Institute of Canada’s Clinical Trials Group (NCIC CTG). This chapter was also
made possible through a Canadian Institutes for Health Research (CIHR)
operating grant to D.B..
REFERENCES
1. Wicha MS, Liu S, Dontu G. Cancer stem cells: an old idea—a paradigm shift.
Cancer Res 2006; 66(4):1883–1890.
2. Balicki D. Cancer stem cells and resistance to treatment. Oncol Exchange 2007;
6(1): 37–40.
3. Vogelstein B, Kinzler KW. Cancer genes and the pathways they control. Nat Med
2004; 10(8):789–799.
4. Massard C, Deutsch E, Soria JC. Tumour stem cell-targeted treatment: elimination
or differentiation. Ann Oncol 2006; 17(11):1620–1624.
5. Wicha MS. Cancer stem cells and metastasis: lethal seeds. Clin Cancer Res 2006;
12(19):5606–5607.
6. Jones RJ, Matsui WH, Smith BD. Cancer stem cells: are we missing the target?
J Natl Cancer Inst 2004; 96(8):583–585.
7. Jordan CT, Guzman ML, Noble M. Cancer stem cells. N Engl J Med 2006; 355(12):
1253–1261.
8. Rizo A, Vellenga E, de Haan G, et al. Signaling pathways in self-renewing hematopoietic
and leukemic stem cells: do all stem cells need a niche? Hum Mol Genet
2006; 15(spec no 2):R210–R219.
9. Schwartz YB, Pirrotta V. Polycomb silencing mechanisms and the management of
genomic programmes. Nat Rev Genet 2007; 8(1):9–22.
10. Bernstein A. Cancer stem cells: the centrality of translational research to cancer
control. CMAJ 2007; 176(1):29–30.

284 Balicki et al.
11. Dontu G, Liu S, Wicha MS. Stem cells in mammary development and carcinogenesis:
implications for prevention and treatment. Stem Cell Rev 2005; 1(3):207–213.
12. Dontu G, El Ashry D, Wicha MS. Breast cancer, stem/progenitor cells and the
estrogen receptor. Trends Endocrinol Metab 2004; 15(5):193–197.
13. Krivtsov AV, Twomey D, Feng Z, et al. Transformation from committed progenitor
to leukaemia stem cell initiated by MLL-AF9. Nature 2006; 442(7104):818–822.
14. Ohm JE, McGarvey KM, Yu X, et al. A stem cell-like chromatin pattern may predispose
tumor suppressor genes to DNA hypermethylation and heritable silencing.
Nat Genet 2007; 39(2):237–242.
15. Clevers H. Stem cells, asymmetric division and cancer. Nat Genet 2005; 37(10):
1027–1028.
16. Scadden DT. The stem-cell niche as an entity of action. Nature 2006; 441(7097):
1075–1079.
17. Dontu G, Abdallah WM, Foley JM, et al. In vitro propagation and transcriptional
profiling of human mammary stem/progenitor cells. Genes Dev 2003; 17(10):
1253–1270.
18. Yang ZJ, Wechsler-Reya RJ. Hit ’em where they live: targeting the cancer stem cell
niche. Cancer Cell 2007; 11(1):3–5.
19. Moore KA, Lemischka IR. Stem cells and their niches. Science 2006; 311(5769):
1880–1885.
20. Liu S, Dontu G, Mantle ID, et al. Hedgehog signaling and Bmi-1 regulate selfrenewal
of normal and malignant human mammary stem cells. Cancer Res 2006;
66(12):6063–6071.
21. Al-Hajj M, Wicha MS, ito-Hernandez A, et al. Prospective identification of
tumorigenic breast cancer cells. Proc Natl Acad Sci U S A 2003; 100(7):3983–3988.
22. Stingl J, Eirew P, Ricketson I, et al. Purification and unique properties of mammary
epithelial stem cells. Nature 2006; 439(7079):993–997.
23. Smith GH. Experimental mammary epithelial morphogenesis in an in vivo model:
evidence for distinct cellular progenitors of the ductal and lobular phenotype. Breast
Cancer Res Treat 1996; 39(1):21–31.
24. Kordon EC, Smith GH. An entire functional mammary gland may comprise the
progeny from a single cell. Development 1998; 125(10):1921–1930.
25. Shackleton M, Vaillant F, Simpson KJ, et al. Generation of a functional mammary
gland from a single stem cell. Nature 2006; 439(7072):84–88.
26. Dontu G, Jackson KW, McNicholas E, et al. Role of Notch signaling in cell-fate
determination of human mammary stem/progenitor cells. Breast Cancer Res 2004;
6(6):R605–R615.
27. Youn BS, Sen A, Behie LA, et al. Scale-up of breast cancer stem cell aggregate
cultures to suspension bioreactors. Biotechnol Prog 2006; 22(3):801–810.
28. Hadnagy A, Gaboury L, Beaulieu R, et al. SP analysis may be used to identify
cancer stem cell populations. Exp Cell Res 2006; 312(19):3701–3710.
29. Goodell MA, Brose K, Paradis G, et al. Isolation and functional properties of
murine hematopoietic stem cells that are replicating in vivo. J Exp Med 1996;
183(4):1797–1806.
30. Balicki D, Beaulieu B. Cancer stem cell side populations. In: Bapat SA, ed. Cancer
Stem Cells. Hoboken: John Wiley and sons, Inc., 2008 (in press).
31. Dontu G, Al-Hajj M, Abdallah WM, et al. Stem cells in normal breast development
and breast cancer. Cell Prolif 2003; 36(suppl 1):59–72.

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 285
32. Al-Hajj M, Becker MW, Wicha M, et al. Therapeutic implications of cancer stem
cells. Curr Opin Genet Dev 2004; 14(1):43–47.
33. Sjoblom T, Jones S, Wood LD, et al. The consensus coding sequences of human
breast and colorectal cancers. Science 2006; 314(5797):268–274.
34. Liu R, Wang X, Chen GY, et al. The prognostic role of a gene signature from
tumorigenic breast-cancer cells. N Engl J Med 2007; 356(3):217–226.
35. Chang JC, Li X, Wong H, et al. Journal of Clinical Oncology, 2007 ASCO Annual
Meeting Proceedings Part I. Vol. 25, No. 18S (June 20 Supplement), 2007:9S.
36. Woodward WA, Chen MS, Behbod F, et al. WNT/beta-catenin mediates radiation
resistance of mouse mammary progenitor cells. Proc Natl Acad Sci U S A 2007;
104(2):618–623.
37. Lymberis SC, Parhar PK, Katsoulakis E, et al. Pharmacogenomics and breast
cancer. Pharmacogenomics 2004; 5(1):31–55.
38. Foulkes WD. BRCA1 functions as a breast stem cell regulator. J Med Genet 2004;
41(1):1–5.
39. Liu S, Ginestier S, Dontu G, et al. BRCA1 regulates human mammary stem cell
self-renewal and differentiation. In: Proceedings of the 98th Annual Meeting of the
American Association for Cancer Research, Los Angeles, CA, 14–18 April 2007.
Philadelphia, PA: AACR; 2007:1346–1347. Abstract nr 5700.
40. Stearns V, Davidson NE, Flockhart DA. Pharmacogenetics in the treatment of
breast cancer. Pharmacogenomics J 2004; 4(3):143–153.
41. Evans WE, Relling MV. Moving towards individualized medicine with pharmacogenomics.
Nature 2004; 429(6990):464–468.
42. Potti A, Dressman HK, Bild A, et al. Genomic signatures to guide the use of
chemotherapeutics. Nat Med 2006; 12(11):1294–1300.
43. Bao S, Wu Q, McLendon RE, et al. Glioma stem cells promote radioresistance
by preferential activation of the DNA damage response. Nature 2006; 444(7120):
756–760.
44. Calabrese C, Poppleton H, Kocak M, et al. A perivascular niche for brain tumor
stem cells. Cancer Cell 2007; 11(1):69–82.
45. Donnenberg VS, Donnenberg AD. Multiple drug resistance in cancer revisited: the
cancer stem cell hypothesis. J Clin Pharmacol 2005; 45(8):872–877.
46. Kim M, Turnquist H, Jackson J, et al. The multidrug resistance transporter ABCG2
(breast cancer resistance protein 1) effluxes Hoechst 33342 and is overexpressed in
hematopoietic stem cells. Clin Cancer Res 2002; 8(1):22–28.
47. Zhou S, Schuetz JD, Bunting KD, et al. The ABC transporter Bcrp1/ABCG2 is
expressed in a wide variety of stem cells and is a molecular determinant of the sidepopulation
phenotype. Nat Med 2001; 7(9):1028–1034.
48. Zhou S, Morris JJ, Barnes Y, et al. Bcrp1 gene expression is required for normal
numbers of side population stem cells in mice, and confers relative protection to
mitoxantrone in hematopoietic cells in vivo. Proc Natl Acad Sci U S A 2002;
99(19):12339–12344.
49. Krishnamurthy P, Ross DD, Nakanishi T, et al. The stem cell marker Bcrp/ABCG2
enhances hypoxic cell survival through interactions with heme. J Biol Chem 2004;
279(23):24218–24225.
50. Asselin-Labat ML, Shackleton M, Stingl J, et al. Steroid hormone receptor status of
mouse mammary stem cells. J Natl Cancer Inst 2006; 98(14):1011–1014.

286 Balicki et al.
51. Sleeman KE, Kendrick H, Robertson D, et al. Dissociation of estrogen receptor
expression and in vivo stem cell activity in the mammary gland. J Cell Biol 2007;
176(1):19–26.
52. Al Hajj M, Wicha MS, Benito-Hernandez A, et al. Prospective identification of
tumorigenic breast cancer cells. Proc Natl Acad Sci U S A 2003; 100(7):3983–3988.
53. Williams DA, Cancelas JA. Leukaemia: niche retreats for stem cells. Nature 2006;
444(7121):827–828.
54. Kastan MB, Schlaffer E, Russo JE, et al. Direct demonstration of elevated aldehyde
dehydrogenase in human hematopoietic progenitor cells. Blood 1990; 75(10):
1947–1950.
55. Russo JE, Hilton J, Colvin OM. The role of aldehyde dehydrogenase isozymes in
cellular resistance to the alkylating agent cyclophosphamide. Prog Clin Biol Res
1989; 290:65–79.
56. Sladek NE, Kollander R, Sreerama L, et al. Cellular levels of aldehyde dehydrogenases
(ALDH1A1 and ALDH3A1) as predictors of therapeutic responses to cyclophosphamidebased
chemotherapy of breast cancer: a retrospective study. Rational individualization
of oxazaphosphorine-based cancer chemotherapeutic regimens. Cancer Chemother
Pharmacol 2002; 49(4):309–321.
57. Gordon MY, Goldman JM, Gordon-Smith EC. 4-Hydroperoxycyclophosphamide
inhibits proliferation by human granulocyte-macrophage colony-forming cells (GM-
CFC) but spares more primitive progenitor cells. Leuk Res 1985; 9(8):1017–1021.
58. Sladek NE. Human aldehyde dehydrogenases: potential pathological, pharmacological,
and toxicological impact. J Biochem Mol Toxicol 2003; 17(1):7–23.
59. Storms RW, Trujillo AP, Springer JB, et al. Isolation of primitive human hematopoietic
progenitors on the basis of aldehyde dehydrogenase activity. Proc Natl
Acad Sci U S A 1999; 96(16):9118–9123.
60. Ginestier C, Hur M, Charafe-Jauffret E, et al. ALDH1 is a marker of normal and
breast cancer stem cells and predicts poor clinical outcome. Targeted Oncol 2007;
2(suppl 1):S9.
61. Charafe-Jauffret E, Ginestier C, Hur MH, et al. Aldefluor-positive populations in
breast cell lines: evidence for cancer stem cells properties and gene expression profile
(GEP). In: Proceedings of the 98th Annual Meeting of the American Association for
Cancer Research, Los Angeles, CA, 14–18 April 2007. Philadelphia, PA: AACR,
2007:310. Abstract 1307.
62. Pearce DJ, Taussig D, Simpson C, et al. Characterization of cells with a high
aldehyde dehydrogenase activity from cord blood and acute myeloid leukemia
samples. Stem Cells 2005; 23(6):752–760.
63. Lee J, Kim HK, Rho JY, et al. The human OCT-4 isoforms differ in their ability to
confer self-renewal. J Biol Chem 2006; 281(44):33554–33565.
64. Atlasi Y, Mowla SJ, Ziaee SA, et al. OCT-4, an embryonic stem cell marker, is
highly expressed in bladder cancer. Int J Cancer 2007; 120(7):1598–1602.
65. Ponti D, Costa A, Zaffaroni N, et al. Isolation and in vitro propagation of tumorigenic
breast cancer cells with stem/progenitor cell properties. Cancer Res 2005;
65(13):5506–5511.
66. Kang MK, Kim RH, Kim SJ, et al. Elevated Bmi-1 expression is associated with
dysplastic cell transformation during oral carcinogenesis and is required for cancer
cell replication and survival. Br J Cancer 2007; 96(1):126–133.

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 287
67. Kim JH, Yoon SY, Jeong SH, et al. Overexpression of Bmi-1 oncoprotein correlates
with axillary lymph node metastases in invasive ductal breast cancer. Breast 2004;
13(5):383–388.
68. Dimri GP, Martinez JL, Jacobs JJ, et al. The Bmi-1 oncogene induces telomerase
activity and immortalizes human mammary epithelial cells. Cancer Res 2002;
62(16):4736–4745.
69. Fodde R, Brabletz T. Wnt/beta-catenin signaling in cancer stemness and malignant
behavior. Curr Opin Cell Biol 2007; 19(2):150–8.
70. Dontu G, Wicha MS. Survival of mammary stem cells in suspension culture:
implications for stem cell biology and neoplasia. J Mammary Gland Biol Neoplasia
2005; 10(1):75–86.
71. Liu S, Dontu G, Wicha MS. Mammary stem cells, self-renewal pathways, and
carcinogenesis. Breast Cancer Res 2005; 7(3):86–95.
72. Warner JK, Wang JC, Takenaka K, et al. Direct evidence for cooperating genetic
events in the leukemic transformation of normal human hematopoietic cells. Leukemia
2005; 19(10):1794–1805.
73. He XC, Zhang J, Tong WG, et al. BMP signaling inhibits intestinal stem cell selfrenewal
through suppression of Wnt-beta-catenin signaling. Nat Genet 2004;
36(10):1117–1121.
74. Adolphe C, Narang M, Ellis T, et al. An in vivo comparative study of sonic, desert
and Indian hedgehog reveals that hedgehog pathway activity regulates epidermal
stem cell homeostasis. Development 2004; 131(20):5009–5019.
75. Rossi DJ, Weissman IL. Pten, tumorigenesis, and stem cell self-renewal. Cell 2006;
125(2):229–231.
76. Roy M, Pear WS, Aster JC. The multifaceted role of Notch in cancer. Curr Opin
Genet Dev 2007; 17(1):52–59.
77. Bullock TE, Wen B, Marley SB, et al. Potential of CD34 in the regulation of
symmetrical and asymmetrical divisions by hematopoietic progenitor cells. Stem
Cells 2006; 25(4):844–851.
78. Song X, Call GB, Kirilly D, et al. Notch signaling controls germline stem cell niche
formation in the Drosophila ovary. Development 2007; 134(6):1071–1080.
79. Androutsellis-Theotokis A, Leker RR, Soldner F, et al. Notch signalling regulates
stem cell numbers in vitro and in vivo. Nature 2006; 442(7104):823–826.
80. Kalderon D. The mechanism of hedgehog signal transduction. Biochem Soc Trans
2005; 33(pt 6):1509–1512.
81. Shimokawa T, Svard J, Heby-Henricson K, et al. Distinct roles of first exon variants
of the tumor-suppressor Patched1 in Hedgehog signaling. Oncogene 2007; 26(34):
4889–4896; [Epub 2007 February 19].
82. Sanchez P, Hernandez AM, Stecca B, et al. Inhibition of prostate cancer proliferation
by interference with Sonic Hedgehog-GLI1 signaling. Proc Natl Acad Sci U S A
2004; 101(34):12561–12566.
83. Ohuchida K, Mizumoto K, Fujita H, et al. Sonic hedgehog is an early developmental
marker of intraductal papillary mucinous neoplasms: clinical implications
of mRNA levels in pancreatic juice. J Pathol 2006; 210(1):42–48.
84. Janzen V, Scadden DT. Stem cells: good, bad and reformable. Nature 2006;
441(7092):418–419.
85. He XC, Yin T, Grindley JC, et al. PTEN-deficient intestinal stem cells initiate
intestinal polyposis. Nat Genet 2007; 39(2):189–198.

288 Balicki et al.
86. Shen WH, Balajee AS, Wang J, et al. Essential role for nuclear PTEN in maintaining
chromosomal integrity. Cell 2007; 128(1):157–170.
87. Korkaya H, Iovino F, Wicha MS. PTEN and HER2 regulate self-renewal and
invasion of human mammary stem cells. In: Proceedings of the 98th Annual
Meeting of the American Association for Cancer Research, Los Angeles, CA, 14–18
April 2007. Philadelphia, PA: AACR, 2007:305. Abstract nr 1287.
88. Yilmaz OH, Valdez R, Theisen BK, et al. Pten dependence distinguishes haematopoietic
stem cells from leukaemia-initiating cells. Nature 2006; 441(7092):475–482.
89. Fidler IJ. The pathogenesis of cancer metastasis: the ‘seed and soil’ hypothesis
revisited. Nat Rev Cancer 2003; 3(6):453–458.
90. Fidler IJ, Yano S, Zhang RD, et al. The seed and soil hypothesis: vascularisation
and brain metastases. Lancet Oncol 2002; 3(1):53–57.
91. Fidler IJ. Seed and soil revisited: contribution of the organ microenvironment to
cancer metastasis. Surg Oncol Clin N Am 2001; 10(2):257–269, vii–viiii.
92. Li F, Tiede B, Massague J, et al. Beyond tumorigenesis: cancer stem cells in
metastasis. Cell Res 2007; 17(1):3–14.
93. Savarese TM, Low HP, Baik I, et al. Normal breast stem cells, malignant breast
stem cells, and the perinatal origin of breast cancer. Stem Cell Rev 2006; 2(2):
103–110.
94. Esteller M. The necessity of a human epigenome project. Carcinogenesis 2006;
27(6):1121–1125.
95. Widschwendter M, Fiegl H, Egle D, et al. Epigenetic stem cell signature in cancer.
Nat Genet 2007; 39(2):157–158.
96. Wada T, Haigh JJ, Ema M, et al. Vascular endothelial growth factor directly inhibits
primitive neural stem cell survival but promotes definitive neural stem cell survival.
J Neurosci 2006; 26(25):6803–6812.
97. Sugiyama T, Kohara H, Noda M, et al. Maintenance of the hematopoietic stem cell
pool by CXCL12-CXCR4 chemokine signaling in bone marrow stromal cell niches.
Immunity 2006; 25(6):977–988.
98. Kiel MJ, Morrison SJ. Maintaining hematopoietic stem cells in the vascular niche.
Immunity 2006; 25(6):862–864.
99. Calvi LM, Adams GB, Weibrecht KW, et al. Osteoblastic cells regulate the
haematopoietic stem cell niche. Nature 2003; 425(6960):841–846.
100. Zhang J, Niu C, Ye L, et al. Identification of the haematopoietic stem cell niche and
control of the niche size. Nature 2003; 425(6960):836–841.
101. Ramasamy R, Lam EW, Soeiro I, et al. Mesenchymal stem cells inhibit proliferation
and apoptosis of tumor cells: impact on in vivo tumor growth. Leukemia 2007;
21(2):304–310.
102. Kuperwasser C, Chavarria T, Wu M, et al. Reconstruction of functionally normal
and malignant human breast tissues in mice. Proc Natl Acad Sci U S A 2004;
101(14):4966–4971.
103. Bhowmick NA, Chytil A, Plieth D, et al. TGF-beta signaling in fibroblasts
modulates the oncogenic potential of adjacent epithelia. Science 2004; 303(5659):
848–851.
104. Orimo A, Gupta PB, Sgroi DC, et al. Stromal fibroblasts present in invasive human
breast carcinomas promote tumor growth and angiogenesis through elevated SDF-1/
CXCL12 secretion. Cell 2005; 121(3):335–348.

Breast Cancer Stem Cells and Their Niche: Lethal Seeds in Lethal Soil 289
105. Boyd NF, Guo H, Martin LJ, et al. Mammographic density and the risk and
detection of breast cancer. N Engl J Med 2007; 356(3):227–236.
106. Hortobagyi GN. Trastuzumab in the treatment of breast cancer. N Engl J Med 2005;
353(16):1734–1736.
107. Chabner BA. In defense of cell-line screening. J Natl Cancer Inst 1990; 82(13):
1083–1085.
108. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000; 100(1):57–70.
109. Li C, Heidt DG, Dalerba P, et al. Identification of pancreatic cancer stem cells.
Cancer Res 2007; 67(3):1030–1037.
110. Dirks PB. Brain tumour stem cells: the undercurrents of human brain cancer and
their relationship to neural stem cells. Philos Trans R Soc Lond B Biol Sci 2008;
363(1489):139–52.
111. Piccirillo SG, Reynolds BA, Zanetti N, et al. Bone morphogenetic proteins inhibit
the tumorigenic potential of human brain tumour-initiating cells. Nature 2006;
444(7120):761–765.
112. Prince ME, Sivanandan R, Kaczorowski A, et al. Identification of a subpopulation
of cells with cancer stem cell properties in head and neck squamous cell carcinoma.
Proc Natl Acad Sci U S A 2007; 104(3):973–978.
113. Jain RK, Forbes NS. Can engineered bacteria help control cancer? Proc Natl Acad
Sci U S A 2001; 98(26):14748–14750.
114. Heldin CH, Rubin K, Pietras K, et al. High interstitial fluid pressure—an obstacle in
cancer therapy. Nat Rev Cancer 2004; 4(10):806–813.
115. Reardon DA, Wen PY. Therapeutic advances in the treatment of glioblastoma:
rationale and potential role of targeted agents. Oncologist 2006; 11(2):152–164.

19
Molecular Imaging in Individualized
Cancer Management
David M. Schuster
Division of Nuclear Medicine and Molecular Imaging, Department
of Radiology, Emory University, Atlanta, Georgia, U.S.A.
Diego R. Martin
Department of Radiology, Emory University, Atlanta, Georgia, U.S.A.
INTRODUCTION
X-rays were discovered by Karl Roentgen in 1895, and with that discovery came
the ability to see structures within the living human body without having to directly
sample the tissue. Imaging, particularly cross-sectional techniques, has revolutionized
the practice of medicine. These techniques employ ionizing radiation, such
as computed tomography (CT) scanning, single-photon emission computed
tomography (SPECT), and positron emission tomography (PET), as well as nonionizing
energy, such as ultrasound and magnetic resonance imaging (MRI). Some
of the most exciting and practical major advances are unfolding within the MRI and
PET technologies, specifically with regard to molecular imaging.
Molecular imaging is defined as the in vivo characterization and measurement
of biological processes at cellular and molecular levels. By definition,
molecular imaging is not restricted to any one tool; it may involve optical,
ultrasound, gamma ray, positron, or magnetic imaging (1). Since the 1960s, bone
scanning has played a major role in the management of breast cancer. However,
in the past decade, the role of radionuclide molecular imaging has expanded
291

292 Schuster and Martin
significantly in the clinical management of breast cancer because of PET,
mammoscintigraphy, and sentinel lymph node techniques. Molecular imaging is
also instrumental in drug development, gene therapy, and basic science research
of breast cancer. Mammography and ultrasound have been and continue to be
very important tools in the diagnosis and management of breast cancer, but other
modalities, such as MRI, have enhanced our ability to detect occult disease,
especially in those with high genetic risk, and to follow breast carcinoma during
and after therapy (2,3).
In this chapter, we will examine MRI and PET with regard to molecular
imaging as applied to individualized breast cancer management; this chapter is
not intended to be a comprehensive review of molecular imaging methods. We
have specifically chosen MRI and PET because we believe the integration of
both technologies will lead to further progress in our ability to interrogate disease
processes in a manner analogous to the classical pathologist, but in a
noninvasive fashion. This ability will continue to expand as technology progresses.
But in essence, the future is already here.
RADIONUCLIDE TECHNIQUES:
CURRENT CONCEPTS AND APPLICATIONS
The advantage of radionuclide techniques over other imaging approaches is their
ability to label any chemical species with an isotope of choice and to detect that
radiotracer with high sensitivity. When certain radionuclides decay, gamma rays
(photons) are given off, which are detected with special crystal detectors. The
gamma emissions of SPECT or PET radiotracers are energetic enough to travel
through tissues and thus enable imaging of the whole body. Carbon, nitrogen,
hydrogen, and oxygen have positron-emitting isotopes and are ideally suited for
designing probes to image natural biochemical and molecular process in vivo.
Probes can either home in on targets on cell surfaces, such as antigens and
receptors, or intracellular structures such as internalized receptors.
Single-photon imaging is the basis of simple techniques such as planar
bone scanning or more sophisticated cross-sectional approaches such as SPECT.
High-energy dual-photon imaging is utilized for PET, permitting better resolution
and sensitivity than single-photon techniques. Yet, each method has its
place. For example, sentinel lymph node procedures, employing single-photon
radiotracers, are rapidly becoming the standard of care in breast cancer lymph
node staging (4). For all practical purposes modern PET scanning is now
accomplished as part of a PET-CT device, which fuses the functional imaging of
PET with the anatomic detail of CT.
In 1984, Beaney and coworkers reported (5) on the use of [ 15 O]H 2 O PET
to study blood flow, the oxygen extraction ratio, and oxygen utilization in breast
cancer. The earliest studies on the use of 18 F-fluorodeoxyglucose (FDG) in the
diagnosis of breast cancer were conducted in 1989 (6,7). Figure 1 is an example
of 18 F-FDG PET uptake in a primary breast mass and axillary lymph nodes.

Molecular Imaging in Individualized Cancer Management 293
Figure 1 (A) Axial CT, (B) attenuation corrected PET, and (C) fused image in a 50-year-old
female with a 4-cm primary right breast cancer (arrow) and malignant axillary lymph
nodes (arrowhead ). Abbreviations: CT, computed tomography; PET, positron emission
tomography.
Fluorodeoxyglucose is actively transported into the cell, where it is irreversibly
phosphorylated by hexokinase, trapping it within the cell, where it does
not proceed further down the glycolytic pathway. There is a correlation between
FDG uptake and glucose transporter-1 (Glut-1) expression, the mitotic activity
index, percent necrosis (likely reflecting the proliferation rate), tumor cells/
volume, hexokinase-1 expression, the number of lymphocytes (but not macrophages),
microvessel density, and possibly the expression of the p53 gene (8–12).
PET is currently of limited value in routine screening for breast cancer (13)
and cannot take the place of fine sectioning and immunohistochemical lymph
node evaluation, but it has an important role to play in a select group of problematic
diagnostic situations and in staging patients with high-risk disease (14–17).
PET can provide important information by detecting locoregional lymph node
involvement as well as distant metastases. Because of its high positive predictive
value, FDG PET may obviate a sentinel lymph node procedure and axillary
dissection.
The use of neoadjuvant chemotherapy has increased the rate of breast
conservation surgery (18). Clinical response does not necessarily correlate with
pathologic response. FDG PET has been shown to aid in monitoring response to
chemotherapy and has prognostic benefit (19). Nonresponders and those
developing progressive disease or distant metastases can be identified earlier,
and this information may prove useful in changing therapies and avoiding side
effects of chemotherapy that is not effective. Baseline PET combined with PET
after the first course of chemotherapy is highly accurate in this regard (16). A
meta-analysis of the FDG PET literature (20) reports an overall 81% sensitivity,
96% specificity, and 92% accuracy for monitoring response to therapy. Elevated
FDG uptake was found to be an independent predictor of worse prognosis and
decreased relapse-free survival in breast cancer (10). Figure 2 illustrates a dramatic
response of metastatic breast cancer to fulvestrant (Faslodex) therapy.
FDG PET has been shown to locate unsuspected metastases with high
sensitivity and specificity (21,22). Detection of early recurrence may have

294 Schuster and Martin
Figure 2 (A) Anterior view from a maximal intensity projection of an FDG-PET scan
showing extensive breast cancer pleural implants in the left chest and (B) after one dose of
fulvestrant (Faslodex) after which the implants resolved. Note normal cardiac (arrow) and
renal (arrowhead) uptake is better seen in B than in A. The patient was status post
unrelated remote right nephrectomy. Abbreviations: FDG,
18 F-fluorodeoxyglucose;
PET, positron emission tomography.
important survival benefits, prompting the use of new therapies as well as curative
or palliative surgery. FDG PET is considered of great efficacy in the evaluation of
patients with suspected recurrent breast cancer, especially to differentiate true
recurrence from postsurgical and radiation sequelae. The ability of PET to detect
suspected clinical recurrence and in patients with elevated tumor markers has been
demonstrated in a number of studies (23–32).
MRI: CURRENT CONCEPTS AND APPLICATIONS
In 1973, Paul Lauterbur discovered that introduction of magnetic gradients
superimposed over a magnetic field makes it possible to create two-dimensional
images of structures, with the first experiments showing the ability to differentiate
between tubes filled with regular or heavy water (33). No other imaging
method can achieve this differentiation. This work was the foundation for a
revolutionary tissue-imaging technique that generates images on the basis of
molecular differences between or within tissues.
To understand the added value of MRI to clinical sciences, it is useful to
compare it with CT. CT can produce images of high spatial resolution, which are
interpreted on the basis of anatomy. A distinct feature of MRI is that anatomically
correct images are produced, but these images are based on molecular
events that produce much higher soft tissue contrast between tissues and between

Molecular Imaging in Individualized Cancer Management 295
Figure 3 Liver cross-sectional imaging showing multiple hypervascular breast metastases.
(A) On contrast-enhanced CT, the liver lesions are not visualized. (B) On contrastenhanced
MRI, the lesions are conspicuous with innumerable tumors scattered throughout
the liver (6 are shown: B, black arrows). Abbreviations: CT, computed tomography; MRI,
magnetic resonance imaging.
abnormal and normal tissues. It has been shown for most clinically important
applications, including the brain and the solid organs of the body, that soft tissue
contrast resolution is more important than spatial resolution. Even on contrastenhanced
CT, the ability to generate soft tissue contrast resolution is less than
that of MRI, and this difference may contribute to the relative advantage that
MRI provides for lesion detection (Fig. 3).
Unlike CT, which relies on the degree of X-ray absorption to differentiate
tissues, MRI contrast differences are based on molecular interactions between
protons located in water molecules within and between cells, neighboring
molecules that influence mobility, such as charged glycoproteins and membranes,
and species that can alter the magnetic field at the molecular level, such
as iron and lipid. There are essentially an infinite variety of formulations of MRI
techniques that can exploit these interactions to generate different contrast patterns.
Similar to the concept of a pathologist using different histochemical stains
to visualize molecular species, the expert MR imager will utilize an optimized
set of MRI sequences to visualize the characteristics that may be unique to
specific pathologies. Though the initial applications of MRI had been for
imaging diseases of the brain, MRI is now able to image faster and with concomitant
improvement in image quality. This has been critical for improving the
capacity of MRI for rapidly imaging the chest, abdomen, and pelvis, where
motion from respiration or cardiac contractions may deteriorate images (34,35).
Breast MRI (Fig. 4) is a recent additional imaging method for the evaluation
of primary breast carcinoma and is useful for early detection and staging
(36–40). MRI has been found to provide maximal sensitivity, as compared to
traditional mammography and ultrasound, but with comparatively lower specificity.
The fundamental components to the current state of the art of breast MRI
is to use the combination of time-resolved gadolinium uptake analysis in combination
with the evaluation of the contours of a mass using high-contrast and

296 Schuster and Martin
Figure 4 (A) Right breast carcinoma shown on sagittal plane dynamically enhanced
gadolinium MRI at the indicated time intervals. Subsequently, (B) a region of interest is
applied within the tumor, and a time-intensity curve is generated. (C) A lymph node
showing a similar enhancement pattern (arrow) is shown on a more lateral slice through
the right axilla on a representative 3-min sagittal image. (D) The enhancing primary breast
carcinoma and the metastatic lymph node in the right axilla can be seen on an axial MRI
image of both breasts (arrows), which correlates with findings on (E) PET-CT (arrows),
confirming the presence of metastasis in the node. Note that the position of the breast is
different on the MRI compared with the PET-CT due to prone MRI versus supine PET
positioning. Abbreviations: MRI, magnetic resonance imaging; PET, positron emission
tomography; CT, computed tomography.
high-resolution imaging. MRI techniques have benefited from improvements in
surface coils specifically dedicated to imaging both breasts simultaneously.
The most important sequence used for routine breast MRI is the threedimensional
gradient echo (3D GRE) technique. Uniform and complete suppression
of the breast lipid signal has been a key area of focus for improvement
of these sequences. With time-resolved imaging after an MRI contrast injection
(Fig. 4), malignant focal breast carcinomas typically show enhancement that

Molecular Imaging in Individualized Cancer Management 297
remains constant over several minutes or may diminish. Benign tumors such as
fibromas may show differential contrast uptake, with progressive increase in
contrast accumulation within the benign mass. However, the specificity for these
differentiating features is relatively moderate. A more recent approach has
incorporated lessons learned previously from conventional X-ray mammography;
the margins of the tumor may show patterns specific to malignancy. Data
are developing to support that the analysis of tumor margin morphology will
improve MRI specificity for the diagnosis of breast carcinoma and separate this
condition from benign tumors. Here, further development of newer MRI strategies,
such as perfusion studies or spectroscopy, or a possible combination with
metabolic imaging provided by PET, may help improve this apparent current
limitation.
Lymphatic spread may result in local lymph node involvement. Findings
on MRI may be relatively specific when the lymph nodes are found to be over
1 cm in diameter and show contrast enhancement that is similar in time course
and intensity to those seen in the primary breast mass. However, for smaller
lymph nodes the specificity may be significantly lower. Combined imaging with
FDG-PET and MRI may help aid the specificity of the MRI study and the
identification of the most concerning lymph nodes (Fig. 4).
Currently accepted applications for MRI include screening for high-risk
women, evaluation of dense fibroglandular breast tissue, and surveillance after
surgical alteration or breast implant placement (36–40). Generalized use of MRI
for evaluation of all women in the screening age group has not been widely
accepted.
With regard to hematogenous tumor spread, the liver is the organ most
commonly involved in metastatic diseases, usually from breast, colon, pancreas,
kidney, and carcinoid cancers, or melanoma. However, most liver tumors are
benign, and benign liver lesions are common, including cysts, bile duct hamartomas,
hemangiomas, adenomas, and focal nodular hyperplasia. The specificity of
a noninvasive test is the litmus test for utility in tumor evaluation. We have shown
that although MRI is slightly more sensitive than CT for tumor detection, the most
important distinction is that MRI is highly specific, approaching 98%, while CT is
significantly less specific, approaching 80% (41–44). The ability to image rapidly
and include the evaluation of the entire body is undergoing development. The
potential for assessing metastatic burden will be a primary potential role for this
capability.
FUTURE DEVELOPMENTS
The resolution of PET scanners is currently limited in comparison with anatomic
imaging, but continued technological improvements are expected
(45,46). Sensitivity for initial diagnosis will likely improve as device resolution
advances and as special focused field PET-mammography imaging systems are
developed (47–49). Small field-of-view cameras not only offer better sentinel

298 Schuster and Martin
lymph node and primary tumor detection because of increased spatial resolution
but can also be mounted on maneuverable arms and used in an intraoperative
setting, helping the surgeon realize a tumor-free margin (50).
MR spectroscopic analysis has the potential to elucidate pathophysiological
processes and may become useful for the assessment of primary or
metastatic neoplasms, or the assessment of tumor response to therapy. MRI
strategies are providing new diagnostic methods of evaluating dynamic pathophysiological
processes using time-resolved imaging. These methods include the
use of an injected contrast agent that can be visualized as it perfuses and then
redistributes within the tissues, much as a radiotracer is sometimes used for
nuclear imaging techniques (51,52). The advantages of MRI for this application
include the capacity for high temporal and spatial resolution and the ability to
combine this component of an examination with other elements of MRI to
provide comprehensive evaluation during a single study. This approach is being
used to detect and characterize diseases, including neoplasms, inflammation and
infection, and fibrotic processes (53–57).
Apoptosis, or programmed cell death, is an elementary process of life that
maintains homeostasis. Apoptosis is physiologic and is different from necrosis,
which is pathologic, often eliciting an inflammatory reaction. Successful chemotherapy
increases apoptosis and enhances the uptake of annexin in the tumor.
As the tumor becomes necrotic, the uptake of annexin reduces. PET and SPECT
imaging with radiolabeled annexin have been used to assess apoptosis (58–61).
Molecular process that occur within breast cancer cells can be imaged by
using a radiotracer based on amino acid metabolism, such as methionine, which
can serve as an indirect measure of protein synthesis, or an amine, such as
choline, which can reflect membrane biosynthesis.
11 C-methionine and
11 C-choline have been used in assessing breast cancer and metastases (62,63).
DNA synthesis within breast cancer has been examined with thymidine analogues
such as 3 0 -deoxy-3 0 -[ 18 F]fluorothymidine (FLT), which has been
employed to image intratumoral proliferation (64,65). [ 15 O]H 2 O PET can also be
useful for tumor perfusion (65). These radiotracers can serve to characterize
tumor as well as measure therapeutic response.
Estrogen and progestin receptors offer another avenue to image breast
cancer. Estrogen receptors (ERs) have been successfully targeted using 16a-[ 18 F]
fluoro-17b-estradiol (FES), and early clinical studies have shown them to be as
sensitive as FDG imaging and probably more specific (66). In one study of
patients with ER-positive breast cancer, the patients also underwent FES PET.
PET was able to predict global response to hormonal therapy in human epidermal
growth factor receptor 2 (HER2)-negative patients better than pathology
(24% vs. 47%) (67,68). Radiolabeled chemotherapeutic agents such as [ 18 F]
fluorotamoxifen may be useful to detect ER-positive metastases and can also
help in following response to treatment (69). 18 F-labeled paclitaxel has also been
shown in a mouse model with a breast carcinoma xenograft to predict tumor
response to paclitaxel chemotherapy (70).

Molecular Imaging in Individualized Cancer Management 299
Peptides that localize to receptors are small-sized molecules and clear
rapidly via the kidney, making these peptides attractive imaging probes. Peptides,
such as bombesin (BN), labeled with single-photon and dual-photon
radionuclides are being studied both for imaging and for therapeutic purposes
(71). Gastrin-releasing peptide (GRP) receptors, a subtype of BN receptors, are
overexpressed in breast carcinoma. Octreoscan, a single-photon radiotracer
somatostatin analogue, has also been used to study breast cancer (72).
Several gene therapy approaches have been considered for the management
of breast cancer, such as ablation of oncogenic products, restoration of ER
expression, alteration of genes that are involved in apoptosis, and activation of
tumor suppressor genes. Reporter genes have been used to study promoter or
regulatory elements involved in gene expression conventionally studied by tissue
biopsy and immunohistochemistry. Now noninvasive in vivo methods are
available in the form of radionuclide techniques or optical imaging. The ability
to image deeper structures in larger animal models and in humans is the main
advantage of radionuclide technique over optical imaging. For example, 8-[ 18 F]
Fluoropenciclovir has been studied as a reporter probe for herpes simplex virus
type 1 thymidine kinase (HSV1-tk) gene expression (73). Other possibilities still
in very early development include specially designed, small radiolabelled antisense
oligodeoxynucleotides that may help image gene expression (20). PET
imaging with an appropriate PET reporter gene and probe combination may
become a powerful tool for molecular imaging of human gene expression (74).
HER2/neu overexpression in breast cancer is associated with poor response
to chemotherapy and an unfavorable prognosis. Phase I clinical trials of E1A
gene therapy targeting HER2/neu-overexpressed breast cancer has shown
increased apoptosis (75). A novel PET radiotracer, (68)Ga-DOTA-F(ab 0 )(2)-
herceptin, has been developed, which has the potential to follow tumor
response to the heat shock protein 90 (Hsp90) inhibitor 17-allylamino-17-
demethoxygeldanamycin (17AAG) (76). In an animal model, downregulation of
HER2 occurs independently of glycolytic response as measured by 18 F-FDG-
PET. This method could be used to rapidly screen tumor response to a variety of
HER2 inhibitors and choose the correct agent within hours rather than wait for
weeks or months to determine if there is a response. Multidrug resistance as
manifested by MDR1 P-glycoprotein transport activity also has the potential to
be studied with SPECT and PET radiotracers (77).
Angiogenesis, the outgrowth of new capillaries, is an important component
in tumor formation and spread. Imaging of angiogenesis for diagnosis and following
therapy can be performed with PET and SPECT radiotracers targeting
integrin, VEGF, and other biomarkers (78). For example, (64)Cu-DOTA-VEGF
(121) PET has been successful in small animals to specifically image in vivo
expression of VEGF (79).
Tumor hypoxia may result in resistance to therapy. [ 18 F]Fluoromisonidazole-3-fluoro-1-(2
0 -nitro-1 0 -imidazolyl)-2-propanol ([ 18 F]MISO) and Cu-diacetylbis(N4-methylthiosemicarbazone)
(Cu-ATSM) radiotracers for PET and blood

300 Schuster and Martin
oxygen level–dependent (BOLD) MRI are some of the techniques being utilized
for in vivo imaging but require further study (80,81).
Finally, nanoparticles technology may be utilized to achieve diagnostic
imaging with both PET and MRI and enable combined modality radionuclide
therapy with intracellular drug delivery (82). The feasibility of magnetic
frequency–activated thermal necrosis with antibody-linked iron oxide nanoparticles
conjugated with the single-photon radiotracer Indium-111 has been
recently demonstrated in breast cancer xenografts (83).
An understanding of pathology, together with the capabilities of MRI,
facilitates the use of a particular MRI feature for the evaluation of a range of
diseases that affect different organ systems. The application of MRI timeresolved
perfusion imaging for tumor ranges from primary diagnosis to the
determination of the degree of tumor neoangiogenesis. The ability to quantify
tumor angiogenesis may provide a useful surrogate marker of tumor responsiveness
prior to initiating therapy or indicate tumor response even prior to any
change in the overall volume of tumor tissue (Fig. 4). MRI may be useful for the
specific evaluation of chemotherapeutic antiangiogenic agents such as anti-
VEGF (84,85). The possibility of using MRI molecular imaging capabilities for
directing therapy is being explored, including the detection of reporter gene
transcription, the use of MRI tissue thermometry to guide therapeutic ultrasound
for noninvasive tumor ablation, and the use of functional brain MRI to plan the
surgical resection of tumors, including metastases to the brain (51,52,86–91).
Most importantly, in a manner analogous to the fusion of PET and CT
scanning, the fusion of PET and MRI scanning holds even greater promise in
providing a one-stop shop for molecular function and exquisite anatomic detail
but with a markedly reduced radiation dose. There have been a number of studies
examining the utility of MRI and FDG PET. For example, Walter et al. studied
42 lesions in 40 patients preoperatively (92). The sensitivity of FDG PET was
63%, and the specificity was 91%; the sensitivity of MRI was 89%, and the
specificity was 74%. Thus, both modalities can play complementary roles, with
overall improved accuracy when used in conjunction. Other papers examining
the complementary strengths and weaknesses of both modalities have emerged
(93–95). A recent intriguing study in a small group of patients examines the
utility of MRI with ultrasmall superparamagnetic iron oxide particles in conjunction
with PET, achieving 100% sensitivity and specificity on a per-patient
basis for local lymph node staging when both modalities are employed (96).
Further technical development is anticipated to facilitate combining MRI
and PET (97). We have already started accumulating evidence of the advantages
in this pairing of technologies, as compared with using the combination of
CT and PET. Figure 5 is an example from our institution of how PET and MRI
data can be combined utilizing special software registration. While the above
PET-MRI studies have been performed on separate devices, necessitating
moving and scheduling the patients between two physically separate units, a
combined PET-MRI scanner in a single gantry is now a reality (45,46,97,98).

Molecular Imaging in Individualized Cancer Management 301
Figure 5 Breast cancer patient (from Fig. 4) with axial MRI, PET, and CT images through
the pelvis. (A) Shows the MRI and PET images before and after deformable registration of
the PET to the MRI. (B) The original CT (upper left) andPET-CT(lower left), compared
with the MRI (upper right) and deformable registered PET-MRI (lower right). Increased
FDG activity in the perirectal space on PET-CT (arrow) is of uncertain origin; however, the
activity from the PET scan clearly corresponds to a benign fibroid on the MRI, which is not
visible on CT. Benign fibroids may occasionally take up FDG in premenopausal women.
The fusion of PET and MRI increased diagnostic accuracy in this patient. Abbreviations:
MRI, magnetic resonance imaging; PET, positron emission tomography; CT, computed
tomography; FDG, 18 F-fluorodeoxyglucose.
OUR VISION
Molecular imaging will continue to have an evolving role in diagnosis, staging,
personalization of therapy, monitoring response to therapy, and detecting
recurrence of breast carcinoma, both local and widespread. A patient will arrive
in the imaging suite and have the entire disease process interrogated on a specially
modified PET-MRI system, which will utilize dynamic MRI techniques
and MR spectroscopy as well as novel PET radiotracers. Complex processes
such as glycolysis, tumor perfusion, cell proliferation, protein synthesis, genetic
expression, angiogenesis, and various other profiling will be examined. There is
potential for additional information and improvements over current invasive
tissue sampling and histopathology, which, for example, may not be truly representative
of the heterogeneity present within a primary tumor or metastases.
This process will not simply serve as a baseline to monitor therapy but
will help individualize therapy since all the parameters can be entered into a
matrix and therapy chosen on the basis of data derived from a cumulative
outcomes database, including surgical, chemotherapeutic, radiotherapy, and
genetic therapy options. Drug-target matching and optimal dosing can be
assessed utilizing the powerful molecular techniques described above. These
techniques will also facilitate targeted biopsies, which will improve genomic
and proteomic profiling and permit a more tailored radiotherapy. Designer
therapeutic and imaging agents may also be synthesized on the basis of a
patient’s particular profile.

302 Schuster and Martin
These same methods may then be used to quickly determine treatment
response as measured by glycolysis, amino acid transport, tumor perfusion,
proliferation, and apoptosis, among other metabolic indices, and also monitor for
resistant or mutating clones and for recurrence. Chemotherapeutic and genetic
therapies may then be altered or doses optimized on the basis of rapid molecular
imaging feedback rather than maximizing therapy based only on toxicity effects.
The possibility of therapy using nanoparticle delivery of killing doses of
radiotherapy combined with chemotherapy may also be possible. The same
nanoparticles could also carry MR and PET imaging tracers.
The concept of the “one magic bullet” for the generic enemy of cancer has
fallen by the wayside. A host of ordinary tools optimized for a particular patient
and his or her own cancer can be combined in a unique fashion with the help of
the “magic window” afforded by molecular imaging. Though defeating the
cancer is a lofty goal, these tools may also enable patients to live with cancer and
manage the entire disease process holistically in cooperation with their medical
caregivers. The true cost-benefit ratios of therapies can then be ascertained. We
believe the technology already exists, but further research, social, and financial
investment, as well as societal vision is required for this dream to flower into
everyday reality.
REFERENCES
1. Basilion JP. Current and future technologies for breast cancer imaging. Breast
Cancer Res 2001; 3(1):14–16.
2. Trecate G, Vergnaghi D, Manoukian S, et al. MRI in the early detection of breast
cancer in women with high genetic risk. Tumori 2006; 92(6):517–523.
3. Bartella L, Smith CS, Dershaw DD, et al. Imaging breast cancer. Radiol Clin North
Am 2007; 45(1):45–67.
4. Krynyckyi BR, Kim CK, Goyenechea MR, et al. Clinical breast lymphoscintigraphy:
optimal techniques for performing studies, image atlas, and analysis of images.
Radiographics 2004; 24(1):121–145 (discussion 139–145).
5. Beaney RP, Lammertsma AA, Jones T, et al. Positron emission tomography for invivo
measurement of regional blood flow, oxygen utilisation, and blood volume in
patients with breast carcinoma. Lancet 1984; 1(8369):131–134.
6. Minn H, Soini I. [18F]fluorodeoxyglucose scintigraphy in diagnosis and follow up of
treatment in advanced breast cancer. Eur J Nucl Med 1989; 15(2):61–66.
7. Kubota K, Matsuzawa T, Amemiya A, et al. Imaging of breast cancer with [18F]
fluorodeoxyglucose and positron emission tomography. J Comput Assist Tomogr
1989; 13(6):1097–1098.
8. Bos R, van Der Hoeven JJ, van Der Wall E, et al. Biologic correlates of (18)
fluorodeoxyglucose uptake in human breast cancer measured by positron emission
tomography. J Clin Oncol 2002; 20(2):379–387.
9. Brown RS, Wahl RL. Overexpression of Glut-1 glucose transporter in human breast
cancer. An immunohistochemical study. Cancer 1993; 72(10):2979–2985.

Molecular Imaging in Individualized Cancer Management 303
10. Oshida M, Uno K, Suzuki M, et al. Predicting the prognoses of breast carcinoma
patients with positron emission tomography using 2-deoxy-2-fluoro[18F]-D-glucose.
Cancer 1998; 82(11):2227–2234.
11. Crippa F, Seregni E, Agresti R, et al. Association between [18F]fluorodeoxyglucose
uptake and postoperative histopathology, hormone receptor status, thymidine labelling
index and p53 in primary breast cancer: a preliminary observation. Eur J Nucl
Med 1998; 25(10):1429–1434.
12. Smith TA, Sharma RI, Thompson AM, et al. Tumor 18F-FDG incorporation is
enhanced by attenuation of P53 function in breast cancer cells in vitro. J Nucl Med
2006; 47(9):1525–1530.
13. Samson DJ, Flamm CR, Pisano ED, et al. Should FDG PET be used to decide
whether a patient with an abnormal mammogram or breast finding at physical
examination should undergo biopsy? Acad Radiol 2002; 9(7):773–783.
14. Noh DY, Yun IJ, Kang HS, et al. Detection of cancer in augmented breasts by
positron emission tomography. Eur J Surg 1999; 165(9):847–851.
15. Wahl RL, Helvie MA, Chang AE, et al. Detection of breast cancer in women after
augmentation mammoplasty using fluorine-18-fluorodeoxyglucose-PET. J Nucl Med
1994; 35(5):872–875.
16. Zimny M, Siggelkow W. Positron emission tomography scanning in gynecologic and
breast cancers. Curr Opin Obstet Gynecol 2003; 15(1):69–75.
17. Czernin J. FDG-PET in breast cancer: a different view of its clinical usefulness. Mol
Imaging Biol 2002; 4(1):35–45.
18. Rose C, Dose J, Avril N. Positron emission tomography for the diagnosis of breast
cancer. Nucl Med Commun 2002; 23(7):613–618.
19. Kostakoglu L, Goldsmith SJ. 18F-FDG PET evaluation of the response to therapy for
lymphoma and for breast, lung, and colorectal carcinoma. J Nucl Med 2003; 44(2):
224–239.
20. Gambhir SS, Czernin J, Schwimmer J, et al. A tabulated summary of the FDG PET
literature. J Nucl Med 2001; 42(suppl 5):S1–S93.
21. Avril N, Dose J, Janicke F, et al. Assessment of axillary lymph node involvement in
breast cancer patients with positron emission tomography using radiolabeled 2-
(fluorine-18)-fluoro-2-deoxy-D-glucose. J Natl Cancer Inst 1996; 88(17):1204–1209.
22. Dose J, Bleckmann C, Bachmann S, et al. Comparison of fluorodeoxyglucose positron
emission tomography and “conventional diagnostic procedures” for the detection of
distant metastases in breast cancer patients. Nucl Med Commun 2002; 23(9):857–864.
23. Kim TS, Moon WK, Lee DS, et al. Fluorodeoxyglucose positron emission tomography
for detection of recurrent or metastatic breast cancer. World J Surg 2001; 25(7):
829–834.
24. Eubank WB, Mankoff DA, Vesselle HJ, et al. Detection of locoregional and distant
recurrences in breast cancer patients by using FDG PET. Radiographics 2002; 22(1):
5–17.
25. Lonneux M, Borbath II, Berliere M, et al. The place of whole-body PET FDG for the
diagnosis of distant recurrence of breast cancer. Clin Positron Imaging 2000; 3(2):
45–49.
26. Siggelkow W, Zimny M, Faridi A, et al. The value of positron emission tomography
in the follow-up for breast cancer. Anticancer Res 2003; 23(2C):1859–1867.

304 Schuster and Martin
27. Moon DH, Maddahi J, Silverman DH, et al. Accuracy of whole-body fluorine-
18-FDG PET for the detection of recurrent or metastatic breast carcinoma. J Nucl
Med 1998; 39(3):431–435.
28. Grahek D, Montravers F, Kerrou K, et al. [18F]FDG in recurrent breast cancer:
diagnostic performances, clinical impact and relevance of induced changes in
management. Eur J Nucl Med Mol Imaging 2004; 31(2):179–188.
29. Suarez M, Perez-Castejon MJ, Jimenez A, et al. Early diagnosis of recurrent breast
cancer with FDG-PET in patients with progressive elevation of serum tumor
markers. Q J Nucl Med 2002; 46(2):113–121.
30. Liu CS, Shen YY, Lin CC, et al. Clinical impact of [(18)F]FDG-PET in patients with
suspected recurrent breast cancer based on asymptomatically elevated tumor marker
serum levels: a preliminary report. Jpn J Clin Oncol 2002; 32(7):244–247.
31. Pecking AP, Mechelany-Corone C, Bertrand-Kermorgant F, et al. Detection of occult
disease in breast cancer using fluorodeoxyglucose camera-based positron emission
tomography. Clin Breast Cancer 2001; 2(3):229–234.
32. Maldonado A, Sancho F, Calvo FA, et al. The role of FDG-PET in diagnosing
suspected tumor recurrence based on increased levels of plasma markers. Clin
Positron Imaging 1998; 1(4):246.
33. Pearson H. Magnetic pioneers net Nobel for putting medicine in the picture. Nature
2003; 425(6958):547.
34. Martin DR, Danrad R, Hussain SM. MR imaging of the liver. Radiol Clin North Am
2005; 43(5):861–886, viii.
35. Martin DR, Friel HT, Danrad R, et al. Approach to abdominal imaging at 1.5 Tesla and
optimization at 3 Tesla. Magn Reson Imaging Clin N Am 2005; 13(2):241–254, v–vi.
36. Deurloo EE, Peterse JL, Rutgers EJ, et al. Additional breast lesions in patients eligible
for breast-conserving therapy by MRI: impact on preoperative management and
potential benefit of computerised analysis. Eur J Cancer 2005; 41(10):1393–1401.
37. Lalonde L, David J, Trop I. Magnetic resonance imaging of the breast: current
indications. Can Assoc Radiol J 2005; 56(5):301–308.
38. Lehman CD, Blume JD, Weatherall P, et al. Screening women at high risk for breast
cancer with mammography and magnetic resonance imaging. Cancer 2005; 103(9):
1898–1905.
39. Liberman L. Breast MR imaging in assessing extent of disease. Magn Reson Imaging
Clin N Am 2006; 14(3):339–349, vi.
40. Warner E, Causer PA. MRI surveillance for hereditary breast-cancer risk. Lancet
2005; 365(9473):1747–1749.
41. Eisenberg LB, Warshauer DM, Woosley JT, et al. CT and MRI of hepatic focal
nodular hyperplasia with peripheral steatosis. J Comput Assist Tomogr 1995; 19(3):
498–500.
42. Martin DR, Semelka RC. Imaging of benign and malignant focal liver lesions. Magn
Reson Imaging Clin N Am 2001; 9(4):785–802, vi–vii.
43. Semelka RC, Martin DR, Balci C, et al. Focal liver lesions: comparison of dualphase
CT and multisequence multiplanar MR imaging including dynamic gadolinium
enhancement. J Magn Reson Imaging 2001; 13(3):397–401.
44. Lee VS, Lavelle MT, Krinsky GA, et al. Volumetric MR imaging of the liver and
applications. Magn Reson Imaging ClinN Am 2001; 9(4):697–716, v–vi.
45. Lucignani G. Time-of-flight PET and PET/MRI: recurrent dreams or actual realities?
Eur J Nucl Med Mol Imaging 2006; 33(8):969–971.

Molecular Imaging in Individualized Cancer Management 305
46. Cherry SR. The 2006 Henry N. Wagner Lecture: Of mice and men (and positrons)–
advances in PET imaging technology. J Nucl Med 2006; 47(11):1735–1745.
47. Raylman RR, Majewski S, Weisenberger AG, et al. Positron emission mammography-guided
breast biopsy. J Nucl Med 2001; 42(6):960–966.
48. Murthy K, Aznar M, Thompson CJ, et al. Results of preliminary clinical trials of the
positron emission mammography system PEM-I: a dedicated breast imaging system
producing glucose metabolic images using FDG. J Nucl Med 2000; 41(11):1851–1858.
49. Adler LP, Weinberg IN, Bradbury MS, et al. Method for combined FDG-PET and
radiographic imaging of primary breast cancers. Breast J 2003; 9(3):163–166.
50. Aarsvold JN, Mintzer RA, Yasillo NJ, et al. A miniature gamma camera. Ann N Y
Acad Sci 1994; 720:192–205.
51. Curran SD, Muellner AU, Schwartz LH. Imaging response assessment in oncology.
Cancer Imaging 2006; 6:S126–S130.
52. Kiessling F, Morgenstern B, Zhang C. Contrast agents and applications to assess tumor
angiogenesis in vivo by magnetic resonance imaging. Curr Med Chem 2007; 14(1):
77–91.
53. Ansari M, Araoz PA, Gerard SK, et al. Comparison of late enhancement cardiovascular
magnetic resonance and thallium SPECT in patients with coronary disease
and left ventricular dysfunction. J Cardiovasc Magn Reson 2004; 6(2):549–556.
54. Goldfarb JW, Shinnar M. Free-breathing delayed hyperenhanced imaging of the
myocardium: a clinical application of real-time navigator echo imaging. J Magn
Reson Imaging 2006; 24(1):66–71.
55. Saeed M, Weber O, Lee R, et al. Discrimination of myocardial acute and chronic
(scar) infarctions on delayed contrast enhanced magnetic resonance imaging with
intravascular magnetic resonance contrast media. J Am Coll Cardiol 21 2006; 48(10):
1961–1968.
56. Schneider G, Ahlhelm F, Seidel R, et al. Contrast-enhanced cardiovascular magnetic
resonance imaging. Top Magn Reson Imaging 2003; 14(5):386–402.
57. Martin DR, Semelka RC. Magnetic resonance imaging of the liver: review of
techniques and approach to common diseases. Semin Ultrasound CT MR 2005; 26(3):
116–131.
58. Yagle KJ, Eary JF, Tait JF, et al. Evaluation of 18F-annexin V as a PET imaging
agent in an animal model of apoptosis. J Nucl Med 2005; 46(4):658–666.
59. Murakami Y, Takamatsu H, Taki J, et al. 18F-labelled annexin V: a PET tracer for
apoptosis imaging. Eur J Nucl Med Mol Imaging 2003.
60. Dekker B, Keen H, Shaw D, et al. Functional comparison of annexin V analogues
labeled indirectly and directly with iodine-124. Nucl Med Biol 2005; 32(4):403–413.
61. Keen HG, Dekker BA, Disley L, et al. Imaging apoptosis in vivo using 124I-annexin
V and PET. Nucl Med Biol 2005; 32(4):395–402.
62. Inoue T, Kim EE, Wong FC, et al. Comparison of fluorine-18-fluorodeoxyglucose
and carbon-11-methionine PET in detection of malignant tumors. J Nucl Med 1996;
37(9):1472–1476.
63. Leskinen-Kallio S, Nagren K, Lehikoinen P, et al. Carbon-11-methionine and PET is
an effective method to image head and neck cancer. J Nucl Med 1992; 33(5):691–695.
64. Smyczek-Gargya B, Fersis N, Dittmann H, et al. PET with [18F]fluorothymidine for
imaging of primary breast cancer: a pilot study. Eur J Nucl Med Mol Imaging 2004;
31(5):720–724.

306 Schuster and Martin
65. Buck AK, Schirrmeister H, Mattfeldt T, et al. Biological characterisation of breast
cancer by means of PET. Eur J Nucl Med Mol Imaging 2004; 31(suppl 1):S80–S87.
66. Dehdashti F, Mortimer JE, Siegel BA, et al. Positron tomographic assessment of
estrogen receptors in breast cancer: comparison with FDG-PET and in vitro receptor
assays. J Nucl Med 1995; 36(10):1766–1774.
67. Linden HM, Stekhova S, Link JM, et al. HER2 expression and uptake of 18F-fluoroestradiol
predict response of breast cancer to hormonal therapy. J Nucl Med 2004;
45:85–86.
68. Shields AF. Positron emission tomography measurement of tumor metabolism and
growth: its expanding role in oncology. Mol Imaging Biol 2006; 8(3):141–150.
69. Inoue T, Kim EE, Wallace S, et al. Preliminary study of cardiac accumulation of F-18
fluorotamoxifen in patients with breast cancer. Clin Imaging 1997; 21(5):332–336.
70. Hsueh WA, Kesner AL, Gangloff A, et al. Predicting chemotherapy response to
paclitaxel with 18F-Fluoropaclitaxel and PET. J Nucl Med 2006; 47(12):1995–1999.
71. Zhang H, Schuhmacher J, Waser B, et al. DOTA-PESIN, a DOTA-conjugated
bombesin derivative designed for the imaging and targeted radionuclide treatment of
bombesin receptor-positive tumours. Eur J Nucl Med Mol Imaging 2007; 34(8):
1198–1208.
72. van Eijck CH, Kwekkeboom DJ, Krenning EP. Somatostatin receptors and breast
cancer. Q J Nucl Med 1998; 42(1):18–25.
73. Iyer M, Barrio JR, Namavari M, et al. 8-[18F]Fluoropenciclovir: an improved
reporter probe for imaging HSV1-tk reporter gene expression in vivo using PET.
J Nucl Med 2001; 42(1):96–105.
74. Penuelas I, Boan J, Marti-Climent JM, et al. Positron emission tomography and gene
therapy: basic concepts and experimental approaches for in vivo gene expression
imaging. Mol Imaging Biol 2004; 6(4):225–238.
75. Hung MC, Hortobagyi GN, Ueno NT. Development of clinical trial of E1A gene
therapy targeting HER-2/neu-overexpressing breast and ovarian cancer. Adv Exp
Med Biol 2000; 465:171–180.
76. Smith-Jones PM, Solit D, Afroze F, et al. Early tumor response to Hsp90 therapy using
HER2 PET: comparison with 18F-FDG PET. J Nucl Med 2006; 47(5):793–796.
77. Sharma V, Prior JL, Belinsky MG, et al. Characterization of a 67Ga/68Ga radiopharmaceutical
for SPECT and PET of MDR1 P-glycoprotein transport activity in
vivo: validation in multidrug-resistant tumors and at the blood-brain barrier. J Nucl
Med 2005; 46(2):354–364.
78. Choe YS, Lee KH. Targeted in vivo imaging of angiogenesis: present status and
perspectives. Curr Pharm Des 2007; 13(1):17–31.
79. Cai W, Chen K, Mohamedali KA, et al. PET of vascular endothelial growth factor
receptor expression. J Nucl Med 2006; 47(12):2048–2056.
80. Padhani AR, Krohn KA, Lewis JS, et al. Imaging oxygenation of human tumours.
Eur Radiol 2007; 17(4):861–872.
81. O’Donoghue JA, Zanzonico P, Pugachev A, et al. Assessment of regional tumor
hypoxia using 18F-fluoromisonidazole and 64Cu(II)-diacetyl-bis(N4-methylthiosemicarbazone)
positron emission tomography: comparative study featuring microPET
imaging, Po2 probe measurement, autoradiography, and fluorescent microscopy in
the R3327-AT and FaDu rat tumor models. Int J Radiat Oncol Biol Phys 2005; 61(5):
1493–1502.

Molecular Imaging in Individualized Cancer Management 307
82. DeNardo SJ, Denardo GL. Targeted radionuclide therapy for solid tumors: an
overview. Int J Radiat Oncol Biol Phys 2006; 66(suppl 2):S89–S95.
83. Denardo SJ, Denardo GL, Natarajan A, et al. Thermal dosimetry predictive of
efficacy of 111In-ChL6 nanoparticle AMF-induced thermoablative therapy for
human breast cancer in mice. J Nucl Med 2007; 48(3):437–444.
84. Kanematsu M, Semelka RC, Leonardou P, et al. Angiogenesis in hepatocellular
nodules: correlation of MR imaging and vascular endothelial growth factor. J Magn
Reson Imaging 2004; 20(3):426–434.
85. Kanematsu M, Semelka RC, Osada S, et al. Magnetic resonance imaging and
expression of vascular endothelial growth factor in hepatocellular nodules in cirrhosis
and hepatocellular carcinomas. Top Magn Reson Imaging 2005; 16(1):67–75.
86. Andrew ER. Nuclear magnetic resonance and the brain. Brain Topogr 1992; 5(2):
129–133.
87. Cai W, Rao J, Gambhir SS, et al. How molecular imaging is speeding up antiangiogenic
drug development. Mol Cancer Ther 2006; 5(11):2624–2633.
88. Charles-Edwards EM, deSouza NM. Diffusion-weighted magnetic resonance imaging
and its application to cancer. Cancer Imaging 2006; 6:135–143.
89. Chenevert TL, Sundgren PC, Ross BD. Diffusion imaging: insight to cell status and
cytoarchitecture. Neuroimaging Clin N Am 2006; 16(4):619–632, viii–ix.
90. Lee JH, Huh YM, Jun YW, et al. Artificially engineered magnetic nanoparticles for
ultra-sensitive molecular imaging. Nat Med 2007; 13(1):95–99.
91. Lee JH, Jun YW, Yeon SI, et al. Dual-mode nanoparticle probes for high-performance
magnetic resonance and fluorescence imaging of neuroblastoma. Angew
Chem Int Ed Engl 2006; 45(48):8160–8162.
92. Walter C, Scheidhauer K, Scharl A, et al. Clinical and diagnostic value of preoperative
MR mammography and FDG-PET in suspicious breast lesions. Eur Radiol
2003; 13(7):1651–1656.
93. Antoch G, Vogt FM, Freudenberg LS, et al. Whole-body dual-modality PET/CT and
whole-body MRI for tumor staging in oncology. JAMA 2003; 290(24):3199–3206.
94. Schmidt GP, Schoenberg SO, Schmid R, et al. Screening for bone metastases: wholebody
MRI using a 32-channel system versus dual-modality PET-CT. Eur Radiol
2006; 17(4):939–949.
95. Brix G, Henze M, Knopp MV, et al. Comparison of pharmacokinetic MRI and [18F]
fluorodeoxyglucose PET in the diagnosis of breast cancer: initial experience. Eur
Radiol 2001; 11(10):2058–2070.
96. Stadnik TW, Everaert H, Makkat S, et al. Breast imaging. Preoperative breast cancer
staging: comparison of USPIO-enhanced MR imaging and 18F-fluorodeoxyglucose
(FDC) positron emission tomography (PET) imaging for axillary lymph node staging:
initial findings. Eur Radiol 2006; 16(10):2153–2160.
97. Schmidt GP, Haug AR, Schoenberg SO, et al. Whole-body MRI and PET-CT in the
management of cancer patients. Eur Radiol 2006; 16(6):1216–1225.
98. Catana C, Wu Y, Judenhofer MS, et al. Simultaneous acquisition of multislice PET
and MR images: initial results with a MR-compatible PET scanner. J Nucl Med
2006; 47(12):1968–1976.

20
Cancer Nanotechnology for Molecular
Profiling and Individualized Therapy
May Dongmei Wang
Departments of Biomedical Engineering and Electrical
and Computer Engineering, Georgia Institute of Technology
and Emory University, Atlanta, Georgia, U.S.A.
Jonathan W. Simons
The Prostate Cancer Foundation, Santa Monica, California, U.S.A.
Shuming Nie
Departments of Biomedical Engineering and Chemistry
and the Winship Cancer Institute, Emory University and
Georgia Institute of Technology, Atlanta, Georgia, U.S.A.
INTRODUCTION
Cancer nanotechnology is an interdisciplinary area of research in science, engineering,
and medicine with broad applications in molecular profiling and individualized
therapy (1–4). The basic rationale is that nanometer-sized particles such as
biodegradable micelles, semiconductor quantum dots (QDs), and iron oxide nanocrystals
have functional or structural properties that are not available from either
molecular or bulk materials (5–15). When linked with biotargeting ligands such as
monoclonal antibodies, peptides, or small molecules, these nanoparticles are used to
target malignant tumors with high affinity and specificity. In the “mesoscopic” size
range of 5 to 100 nm diameter, nanoparticles also have large surface areas and
functional groups for conjugating to multiple diagnostic (e.g., optical, radioisotopic,
309

310 Wang et al.
or magnetic) and therapeutic (e.g., anticancer) agents. The emergence of nanotechnology
has opened new opportunities for personalized oncology in which cancer
detection, diagnosis, and therapy are tailored to each individual’s molecular profile,
and also for predictive oncology in which genetic or molecular information is used to
predict tumor development, progression, and clinical outcome.
Recent advances have led to the development of functional nanoparticles
that are covalently linked to biological molecules such as peptides, proteins,
nucleic acids, or small-molecule ligands (1–4). Medical applications have also
appeared, such as the use of superparamagnetic iron oxide nanoparticles as a
contrast agent for lymph node prostate cancer detection (16), and the use of
polymeric nanoparticles for targeted gene delivery to tumor vasculatures (17). In
particular, the U.S. Food and Drug Administration (FDA) recently approved
Abraxane TM , an albumin-paclitaxel (Taxol TM ) nanoparticle for the treatment of
metastatic breast cancer (18). A phase I clinical trial determined that the maximum
tolerated dose (MTD) of single-agent albumin-bound paclitaxel every
three weeks was 300 mg/m 2 in patients with solid tumors (breast cancer and
melanoma). A second phase I trial demonstrated five patient responses among
39 pretreated patients with advanced solid tumors, including one response in a
patient with non–small cell lung cancer (NSCLC), three responses in patients
with ovarian cancer, and one in breast cancer. The dose-limiting toxicity was
myelosuppression, the MTD was 270 mg/m 2 , and premedication was not
required. Subsequent use of Abraxane in both phase II and phase III trials proved
that this new formulation was considerably superior to Taxol. The use of
nanoparticles for drug delivery and targeting is one of the most exciting and
clinically important applications of cancer nanotechnology.
CANCER BIOMARKERS AND NANOTECHNOLOGY
Cancer markers are broadly defined as altered or mutant genes, RNA, proteins,
lipids, carbohydrates, small metabolite molecules, and altered expression of
those that are correlated with a biological behavior or a clinical outcome (19–22).
Most cancer biomarkers are discovered by molecular profiling studies on the
basis of an association or correlation between a molecular signature and cancer
behavior. In the cases of both breast and prostate cancers, a major progression
step is the appearance of so-called “lethal phenotypes” (causing patient death)
such as bone metastatic, hormone-independent, and radiation- and chemotherapyresistant
phenotypes. It has been hypothesized that each of these aggressive
behaviors or phenotypes could be understood and predicted by a defining set of
biomarkers (20). By critically defining the inter-relationships between these
biomarkers, it could be possible to diagnose and determine the prognosis of a
cancer on the basis of a patient’s molecular profile, leading to personalized and
predictive medicine.
One example is the drug trastuzumab (Herceptin TM ; Genentech/Roche,
California, U.S./Basel, Switzerland), a monoclonal antibody designed to target

Cancer Nanotechnology 311
amplified and overexpressed ERBB2 (also known as HER2) tyrosine kinase
receptor found in only approximately 25% to 30% of breast cancers. FDA
approval of trastuzumab was predicated on the availability of a test to detect
ERBB2 overexpression. Both an immunohistochemistry (IHC) assay for the
expressed protein (HercepTest TM ; Dako, Glostrup, Denmark) and a nucleic acid–
based fluorescence in situ hybridization (FISH) test (PathVysion; Abbott Laboratories,
Illinois, U.S.) have been approved as in vitro diagnostics to guide
trastuzumab treatment decisions. In another example, the clinical response of
lung cancer patients to the epidermal growth factor receptor (EGFR) tyrosine
kinase inhibitor gefitinib (Iressa TM ; AstraZeneca, London, U.K.) is associated
with a small number of genetic mutations (23,24). Thus, a molecular diagnostic
test could be used to identify patients that are most likely to respond to this drug.
Despite these advances, critical studies that can clearly link biomarkers
with cancer behavior remain a significant challenge. One difficulty is that most
cancer tumors (especially prostate and breast cancer) are highly heterogeneous,
containing a mixture of benign, cancerous, and stromal cells. Current technologies
for molecular profiling including reverse-transcriptase polymerase
chain reaction (RT-PCR), gene chips, protein chips, two-dimensional gel electrophoresis,
biomolecular mass spectrometry (e.g., MALDI-MS, i.e., matrix
assisted laser desorption ionization mass spectrometry, ES-MS, i.e., electro spray
mass spectroscopy, and SELDI-MS, i.e., surface enhanced laser desorption/
ionization) are not designed to handle this type of heterogeneous samples
(25,26). Furthermore, a limitation shared by all these technologies is that they
require destructive preparation of cells or tissue specimens into a homogeneous
solution, leading to a loss of valuable three-dimensional cellular and tissue
morphological information associated with the original tumor. The development
of nanotechnology, especially bioconjugated nanoparticles, provides an essential
link by which biomarkers could be functionally correlated with cancer behavior
(Fig. 1).
NANOTYPING
Semiconductor QDs are tiny light-emitting particles on the nanometer scale, and
are emerging as a new class of fluorescent labels for biology and medicine
(8–15). In comparison with organic dyes and fluorescent proteins, QDs have
unique optical and electronic properties such as size-tunable light emission,
superior signal brightness, resistance to photobleaching, and simultaneous excitation
of multiple fluorescence colors (Fig. 2). These properties are most promising
for improving the sensitivity and multiplexing capabilities of molecular histopathology
and disease diagnosis. Recent advances have led to highly bright and
stable QD probes that are well suited for profiling genetic and protein biomarkers
in intact cells and clinical tissue specimens (2–4). In contrast to in vivo imaging
applications where the potential toxicity of cadmium-containing QDs is a major
concern, immunohistological staining is performed on in vitro or ex vivo clinical

312 Wang et al.
Figure 1 Size-tunable optical properties of semiconductor QDs for multiplexed molecular
profiling of cancer. (A) Fluorescence emission spectra of different sizes of CdSe QDs.
(B) Relative sizes of the five sizes of CdSe QDs corresponding to the above spectrum.
From left to right the diameters are: 2.1 nm, 2.5 nm, 2.9 nm, 4.7 nm, and 7.5 nm.
Abbreviations: CdSe, cadium selenide; QDs, quantum dots.
patient samples. As a result, the use of multicolor QD probes in IHC is likely one
of the most important and clinically relevant applications in the near term (27).
Indeed, significant opportunities exist at the interface between QD nanotechnology
and signature biomarkers for cancer diagnosis and individualized
therapy. In particular, QD nanoparticle probes have been used to quantify a panel
of biomarkers on intact cancer cells and tissue specimens, allowing a correlation
of traditional histopathology and molecular signatures for the same material
(2–4,27). A single nanoparticle is large enough for conjugation to multiple
ligands, leading to enhanced binding affinity and exquisite specificity through a
“multivalency” effect (28). These features are especially important for the
analysis of cancer biomarkers that are present at low concentrations or in small
numbers of cells.
Quantitative biomarker information can be obtained by using a spectrometer
attached to the fluorescence microscope. It is however very important to
use a common protein such as b actin or glyceraldehyde 3-phosphate dehydrogenase
(GAPDH) as an “internal control.” That is, one of the QD–Ab (antibody)
conjugates should be designed to measure the product of a housekeeping gene
that is expressed at relatively constant levels in all cells. The use of an internal
control holds great promise for overcoming a number of major problems in
biomarker quantification, such as differences in the probe brightness, variations
in probe binding efficiency, uneven light illumination, and detector responses.

Cancer Nanotechnology 313
Figure 2 Nanotechnology linking biomarkers with cancer behavior and clinical outcome.
The schematic diagram shows multiplexed detection and quantification of cancer
biomarkers on intact cells or tissues with multicolor nanoparticle probes. (Left) The
images show cancer cells labeled with quantum dots, and (right) the drawings suggest
how surface and intracellular biomarkers could be quantified by wavelength-resolved
spectroscopy or spectral imaging.
The majority of available tumor specimens are archived, formalin-fixed paraffinembedded
(FFPE) tissues that might be several decades old. As the clinical
outcomes of these tissues are already known, these specimens are well suited for
examining the relationship between molecular profile and clinical outcome in
retrospective studies.
It is also critically important to validate the QD staining data with other
available techniques. For this purpose, O’Regan and coworkers (2) have
obtained QD molecular profiling data from standard human breast cell specimens
and have compared the corresponding biomarker data with traditional IHC
and FISH techniques. Briefly, slides from formaldehyde-fixed paraffin cell
blocks were stained in accordance with standard pathological protocols for three
breast cancer biomarkers—estrogen receptor (ER), progesterone receptor (PR),
and HER2. This panel of protein biomarkers was selected because of its clinical
significance in human breast cancer diagnosis and treatment (29–32). The traditional
IHC results were analyzed by two independent observers and scored
with a standard scale from 0 (no visible staining in the nucleus or membrane) to

314 Wang et al.
3þ (strong and complete membrane or nuclear staining in > 10% of malignant
stained cells). For a comparative analysis of QD profiling with traditional IHC, it
is necessary to normalize the absolute fluorescence intensities of QD–Ab signals,
so that relative percentage values are calculated from the maximum signal
strength.
The results reveal that a 3þ score for ER, PR, or HER2 by traditional IHC
corresponds to 85% to 100% relative expression of the antigen by QD–Ab measurement,
and that 1þ or 2þ scores by traditional IHC correspond to 11% to 48%
expression as determined by QD quantification. It should be noted that classification
of antigens expressed at low levels (1þ or 2þ) is subjective, requiring
experience and often resulting in considerable interobserver variations. In contrast,
quantitative QD measurements allow accurate determination of tumor antigens at
low levels. For example, PR expression in MCF-7 cells and ER expression in BT-
474 cells are both classified as 1þ by traditional IHC, but quantitative QD
measurements indicate major differences in PR expression (16.8%) and ER
expression (47.7%) in these two cell lines. This indicates that the quantitative
nature of QD-based molecular profiling could simplify and standardize categorization
of antigens that are expressed at low levels. This is of fundamental
importance in the management of breast cancer, as the likely benefit of hormonal
therapies and trastuzumab depends directly on not just the presence but also the
quantity of hormone or HER2 receptors (33–35) (Fig. 3).
Wang and coworkers (27) have developed an integrated image processing
and bioinformatics software tool called Q-IHC for quantitative analysis of biomarker
expression and distribution in IHC images. In comparison with previous
image processing software for automated feature extraction and quantitative
Figure 3 Targeting the HER 2/Neu antigen on breast cancer cells by using antibodyconjugated
quantum dots. The image shows individual breast cancer cells with
nuclear counterstaining.

Cancer Nanotechnology 315
analysis (36,37), this software system is capable of handling imaging data from
both traditional and QD-based IHC. To measure the distribution of labeled
antigens, multiple slides of IHC imaging data are acquired to capture selected
tissue structures. After image acquisition, an image-processing module carries
out automatic boundary identification, semiautomatic image segmentation, and
color-based tissue classification on the basis of biomarker staining. Then, an
image analysis module quantifies the various biomarker features into numerical
values. These values become distinct features and are used for comparison with
clinical diagnosis. After validation by a physician, the quantitative data and rules
describing biomarker features are stored in a database. This semiautomatic
image processing and quantification system is designed to provide molecular
profiling data that are more objective, more consistent, and more reproducible
than completely manual or automated quantification methods. The software tools
process image files from slide scanners in Matlab, which is a collection of
various engineering processing tools. In addition, a user-friendly graphical
interface allows users to give input and feedback to improve the system quality.
NANOTHERAPEUTICS
Most current anticancer agents do not greatly differentiate between cancerous
and normal cells, leading to systemic toxicity and adverse effects. Consequently,
systemic applications of these drugs often cause severe side effects in other tissues
(such as bone marrow suppression, cardiomyopathy, and neurotoxicity), which
greatly limit the maximal allowable dose of the drug. In addition, rapid elimination
and widespread distribution into nontargeted organs and tissues require administration
of a drug in large quantities, which is not economical and often complicated
because of nonspecific toxicity. Nanotechnology offers a more targeted approach
and could thus provide significant benefits to cancer patients.
Rapid vascularization in fast-growing cancerous tissues is known to result
in leaky, defective architecture and impaired lymphatic drainage. This structure
allows an enhanced permeation and retention (EPR) effect (38–42), as a result of
which nanoparticles accumulate at the tumor site. For such passive targeting
mechanism to work, the size and surface properties of drug delivery nanoparticles
must be controlled to avoid uptake by the reticuloendothelial system
(RES) (43). To maximize circulation time and targeting ability, the optimal size
should be less than 100 nm in diameter and the surface should be hydrophilic to
circumvent clearance by macrophages. A hydrophilic surface of the nanoparticles
safeguards against plasma protein adsorption, and it can be achieved
through hydrophilic polymer coatings such as polyethylene glycol (PEG),
poloxamines, poloxamers, polysaccharides or through the use of branched or
block amphiphilic copolymers (44–47). The covalent linkage of amphiphilic
copolymers (polylactic acid, polycaprolactone, and polycyanonacrylate chemically
coupled to PEG) is generally preferred, as it avoids aggregation and ligand
desorption when in contact with blood components.

316 Wang et al.
An alternative passive targeting strategy is to utilize the unique tumor
environment in a scheme called tumor-activated prodrug therapy. The drug is
conjugated to a tumor-specific molecule and remains inactive until it reaches the
target (48). Overexpression of the matrix metalloproteinase (MMP), MMP-2, in
melanoma has been shown in a number of preclinical as well as clinical
investigations. Mansour et al. (49) reported a water-soluble maleimide derivative
of doxorubicin (DOX) incorporating an MMP-2 specific peptide sequence (Gly-
Pro-Leu-Gly-Ile-Ala-Gly-Gln) that rapidly and selectively binds to the cysteine-
34 position of circulating albumin. The albumin-DOX conjugate is efficiently
and specifically cleaved by MMP-2, releasing a DOX tetrapeptide (Ile-Ala-Gly-
Gln-DOX) and subsequently DOX. pH and redox potential have been also
explored as drug-release triggers at the tumor site (50). Another passive targeting
method is the direct local delivery of anticancer agents to tumors. This approach
has the obvious advantage of excluding the drug from the systemic circulation.
However, administration can be highly invasive, as it involves injections or
surgical procedures. For some tumors such as lung cancers that are difficult to
access, the technique is nearly impossible to use.
Active targeting is usually achieved by conjugating to the nanoparticle a
targeting component that provides preferential accumulation of nanoparticles in
the tumor-bearing organ, in the tumor itself, individual cancer cells, or intracellular
organelles inside cancer cells. This approach is based on specific
interactions such as lectin-carbohydrate, ligand-receptor, and antibody-antigen
(51). Lectin-carbohydrate is one of the classic examples of targeted drug delivery
(52). Lectins are proteins of nonimmunological origin, which are capable of
recognizing and binding to glycoproteins expressed on cell surfaces. Lectin
interactions with certain carbohydrates are very specific. Carbohydrate moieties
can be used to target drug delivery systems to lectins (direct lectin targeting), and
lectins can be used as targeting moieties to target cell surface carbohydrates
(reverse lectin targeting). However, drug delivery systems based on lectincarbohydrate
have mainly been developed to target whole organs (53), which can
pose harm to normal cells. Therefore, in most cases, the targeting moiety is
directed toward specific receptors or antigens expressed on the plasma membrane
or elsewhere at the tumor site.
The cell surface receptor for folate is inaccessible from the circulation to
healthy cells because of its location on the apical membrane of polarized epithelia,
but it is overexpressed on the surface of various cancers like ovary, brain,
kidney, breast, and lung malignancies (54,55). Surface plasmon resonance
studies revealed that folate-conjugated PEGylated cyanoacrylate nanoparticles
had a 10-fold higher affinity for the folate receptor than the free folate ones did
(56). Folate receptors are often organized in clusters and bind preferably to the
multivalent forms of the ligand. Furthermore, confocal microscopy demonstrated
selective uptake and endocytosis of folate-conjugated nanoparticles by tumor
cells bearing folate receptors. Interest in exploiting folate receptor targeting
in cancer therapy and diagnosis has rapidly increased, as attested by many

Cancer Nanotechnology 317
conjugated systems including proteins, liposomes, imaging agents, and neutron
activation compounds (54,55).
For enhanced tumor-specific targeting, the differences between cancerous
cells and normal cells may be exploited. By virtue of their small size, nanoparticles
entail a high surface area that not only paves a way for more efficient
drug release but also a better strategy for functionalization. There is a growing
body of knowledge on unique cancer markers due to recent advances in proteomics
and genomics. They form the basis of complex interactions between
bioconjugated nanoparticles and cancer cells. Carrier design and targeting
strategies may vary according to the type, developmental stage, and location of
cancer (57). There is much synergy between imaging and nanotechnology in
biomedical applications. Many of the principles used to target delivery of drugs
to cancer may also be applied to target imaging and diagnostic agents to enhance
detection sensitivity in medical imaging. With engineered multifunctional
nanoparticles, the full in vivo potential of cancer nanotechnology in targeted
drug delivery and imaging can be realized.
BIONANOINFORMATICS
For the long-term goal of personalized and predictive oncology, it is important to
integrate biotechnology and nanotechnology with information technology
(Fig. 4). A major step in this direction is the development of the Cancer Bioinformatics
Grid (caBIG), the “World Wide Web for Cancer Research” spearheaded
by the U.S. National Cancer Institute (NCI) (58). Through data sharing
and standardization, this national initiative aims at improving the infrastructures
of health care information technology, to facilitate the process from clinical data
Figure 4 Schematic diagram showing the integration of biotechnology, nanotechnology,
and information technology for personalized oncology.

318 Wang et al.
collection to computational data mining, and to accelerate the use of biomolecular
markers for personalized treatment. The caBIG organizational structure
is divided into working groups, each focused on different aspects of the
interoperability problem specific to cancer research. Two overarching work
groups, “Architecture” and “Vocabularies and Common Data Elements” have
begun to produce detailed specifications that will guide the design of interoperable
bioinformatics tools. In practice, caBIG provides a common platform
from which to launch even more ambitious integrated solutions. Bioinformatics
laboratories can adopt the basic infrastructure, databases, and functionality
provided by caBIG working groups to fill gaps in knowledge and to carry out
information management responsibilities. In addition to these shared tools and
architectures, caBIG is the perfect test bed for a suite of standard tools that can
be understood, verified, and used by clinicians around the world. Another major
effort is Cancer Bioinformatics Infrastructure Objects (caBIO), an application
programming interface (API) that was first developed as Java objects and was
later extended to support various programming platforms. These objects represent
the most fundamental concepts in bioinformatics research such as genes,
ontology, and sequences. Also under development are network programs such as
caBIONet based on Microsoft’s .NET technology for data analysis and integration
(59).
OUTLOOK
Looking into the future, there are a number of research themes or directions that
are particularly promising but require concerted effort for success. The first
direction is the design and development of nanoparticles with monofunctions,
dual functions, three functions, or multiple functions. For cancer and other
medical applications, important functions include imaging (single or dualmodality),
therapy (single drug or combination of two or more drugs), and
targeting (one or more ligands). With each added function, nanoparticles could
be designed to have novel properties and applications. For example, binary
nanoparticles with two functions could be developed for molecular imaging,
targeted therapy, or for simultaneous imaging and therapy (but without targeting).
Bioconjugated QDs with both targeting and imaging functions will be used
for targeted tumor imaging and molecular profiling applications. Conversely,
ternary nanoparticles with three functions could be designed for simultaneous
imaging and therapy with targeting, targeted dual-modality imaging, or for targeted
dual-drug therapy. Quaternary nanoparticles with four functions can be
conceptualized in future to have the abilities of tumor targeting, dual-drug
therapy, and imaging. The second direction is nanoparticle molecular profiling
(nanotyping) for clinical oncology; i.e., the use of bioconjugated nanoparticle
probes to predict cancer behavior, clinical outcome, treatment response, and
individualize therapy. This should start with retrospective studies of archived
specimens because the patient outcome is already known for these specimens.

Cancer Nanotechnology 319
The key hypotheses to be tested are that nanotyping a panel of tumor markers
will allow more accurate correlations than single tumor markers, and that the
combination of nanotyping tumor gene expression and host stroma are both
important in defining the aggressive phenotypes of cancer as well as determining
the response of early-stage disease to treatment (chemotherapy, radiation, or
surgery). The third important direction is to study nanoparticle distribution,
excretion, metabolism, and pharmacodynamics in in vivo animal models. These
investigations will be very important in the development of nanoparticles for
clinical applications in cancer imaging or therapy.
In conclusion, nanotechnology is emerging as an enabling technology for
personalized oncology in which cancer detection, diagnosis, and therapy are
tailored to each individual’s tumor molecular profile and also for predictive
oncology in which genetic or molecular markers are used to predict disease
development, progression, and clinical outcomes (60,61).
ACKNOWLEDGEMENTS
We are grateful to Dr. Leland Chung, Dr. Ruth O’Regan, and Dr. Dong Shin for
insightful discussions. Our research in cancer nanotechnology was supported by
National Institutes of Health (NIH) grants (R01 CA108468-01 and U54CA119338)
and the Georgia Cancer Coalition Distinguished Cancer Scholars Program (to May
Dongmei Wang and Shuming Nie).
REFERENCES
1. Ferrari M. Cancer nanotechnology: Opportunities and challenges. Nat Rev Cancer
2005; 5:161–171.
2. Yezhelyev M, Gao XH, Xing Y, et al. Emerging use of nanoparticles in diagnosis
and treatment of breast cancer. Lancet Oncol 2006; 7:657–767.
3. Sinha R, Kim GK, Nie SM, et al. Nanotechnology in cancer therapeutics: bioconjugated
nanoparticles for drug delivery. Mol Cancer Ther 2006; 5:1909–1917.
4. Xing Y, Kim GJ, Nie SM, et al. Nanotechnology Applications in Cancer. Annu Rev
Biomed Eng 2007; 9:257–288 (review).
5. Henglein A. Small-particle research - physicochemical properties of extremely small
colloidal metal and semiconductor particles. Chem Rev 1989; 89:1861–1873.
6. Schmid G. Large clusters and colloids - metals in the embryonic state. Chem Rev
1992; 92:1709–1727.
7. Niemeyer CM. Nanoparticles, proteins, and nucleic acids: biotechnology meets
materials science. Angew Chem Int Ed Engl 2001; 40:4128–4158.
8. Alivisatos AP. The use of nanocrystals in biological detection. Nat Biotechnol 2004;
22:47–52.
9. Alivisatos AP. Semiconductor clusters, nanocrystals, and quantum dots. Science
1996; 271:933–937.
10. Alivisatos AP, Gu WW, Larabell C. Quantum dots as cellular probes. Annu Rev
Biomed Eng 2005; 7:55–76.

320 Wang et al.
11. Pinaud F, Michalet X, Bentolila LA, et al. Advances in fluorescence imaging with
quantum dot bio-probes. Biomaterials 2006; 27:1679–1687.
12. Michalet X, Pinaud FF, Bentolila LA, et al. Quantum dots for live cells, in vivo
imaging, and diagnostics. Science 2005; 307:538–544.
13. Gao XH, Yang LL, Petros JA, et al. In vivo molecular and cellular imaging with
quantum dots. Curr Opin Biotechnol 2005; 16: 63–72.
14. Smith AM, Gao X, Nie S. Quantum dot nanocrystals for in vivo molecular and
cellular imaging. Photochem Photobiol 2004; 80:377–3785.
15. Chan WC, Maxwell DJ, Gao X, et al. Luminescent quantum dots for multiplexed
biological detection and imaging. Curr Opin Biotechnol 2002; 13:40–46.
16. Harisinghani MG, Barentsz J, Hahn PF, et al. Noninvasive detection of clinically
occult lymph-node metastases in prostate cancer. N Engl J Med 2003; 348:2491–2495.
17. Hood JD, Bednarski M, Frausto R, et al. Tumor regression by targeted gene delivery
to the neovasculature. Science 2002; 296:2404–2407.
18. Robinson DM, Keating GM. Albumin-bound paclitaxel in metastatic breast cancer.
Drugs 2006; 66:941–948.
19. Hahn WC, Weinberg RA. Modelling the molecular circuitry of cancer. Nat Rev
Cancer 2002; 2:331–341.
20. Liotta L, Petricoin E. Molecular profiling of human cancer. Nat Rev Genet 2000;
1:48–56.
21. Petricoin EF, Zoon KC, Kohn EC, et al. Clinical proteomics: translating benchside
promise into bedside reality. Nat Rev Drug Discov 2002; 1:683–695.
22. Menon U, Jacobs IJ. Recent developments in ovarian cancer screening. Curr Opin
Obstet Gynecol 2000; 12:39–42.
23. Lynch TJ, Bell DW, Sordella R, et al. Activating mutations in the epidermal growth
factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib.
N Engl J Med 2004; 350:2129–2139.
24. Paez JG, Janne PA, Lee JC, et al. EGFR mutations in lung cancer: correlation with
clinical response to gefitinib therapy. Science 2004; 304:1497–1500.
25. Michener CM, Ardekani AM, Petricoin EF, et al. Genomics and proteomics:
application of novel technology to early detection and prevention of cancer. Cancer
Detect Prev 2002; 26:249–255.
26. Srinivas PR, Verma M, Zhao Y, et al. Proteomics for cancer biomarker discovery.
Clin Chem 2002; 48:1160–1169.
27. Xing Y, Chaudry Q, Shen C, et al. Bioconjugated quantum dots for multiplexed and
quantitative immunohistochemistry. Nat Protoc 2007; 2:1152–1165.
28. Akerman ME, Chan WC, Laakkonen P, et al. Nanocrystal targeting in vivo. Proc
Natl Acad Sci U S A 2002; 99:12617–12621.
29. Umemura S, Osamura RY. Utility of immunohistochemistry in breast cancer practice.
Breast Cancer 2004; 11:334–338.
30. Thomson TA, Hayes MM, Spinelli JJ, et al. HER-2/neu in breast cancer: interobserver
variability and performance of immunohistochemistry with 4 antibodies compared
with fluorescent in situ hybridization. Mod Pathol 2001; 14:1079–1086.
31. Vogel CL, Cobleigh MA, Tripathy D, et al. Efficacy and safety of trastuzumab as a
single agent in first-line treatment of HER2-overexpressing metastatic breast cancer.
J Clin Oncol 2002; 20:719–726.
32. Elledge RM, Green S, Pugh R, et al. Estrogen receptor (ER) and progesterone
receptor (PR), by ligand-binding assay compared with ER, PR and pS2, by

Cancer Nanotechnology 321
immunohistochemistry in predicting response to tamoxifen in metastatic breast
cancer: a Southwest Oncology Group Study. Int J Cancer 2000; 89:111–117.
33. Marquez DC, Pietras RJ. Membrane-associated binding sites for estrogen contribute
to growth regulation of human breast cancer cells. Oncogene 2001; 20:5420–5430.
34. Hicks DG, Tubbs RR. Assessment of the HER2 status in breast cancer by fluorescence
in situ hybridization: a technical review with interpretive guidelines. Hum
Pathol 2005; 36:250–261.
35. Konecny G, Pauletti G, Pegram M, et al. Quantitative association between HER-2/
neu and steroid hormone receptors in hormone receptor-positive primary breast
cancer. J Natl Cancer Inst 2003; 95:142–153.
36. Matkowskyj KA, Schonfeld D, Benya RV. Quantitative immunohistochemistry by
measuring cumulative signal strength using commercially available software Photoshop
and Matlab. J Histochem Cytochem 2000; 48:303–311.
37. Matkowskyj KA, Schonfeld D, Benya RV. Quantitative immunohistochemistry by
measuring cumulative signal strength accurately measures receptor number. J Histochem
Cytochem 2003; 51:205–214.
38. Matsumura Y, Maeda H. A new concept for macromolecular therapeutics in cancer
chemotherapy: mechanism of tumoritropic accumulation of proteins and the antitumor
agent smancs. Cancer Res 1986; 46:6387–6392.
39. Duncan R. The dawning era of polymer therapeutics. Nat Rev Drug Discov 2003;
2:347–360.
40. Jain RK. Transport of molecules, particles, and cells in solid tumors. Annu Rev
Biomed Eng 1999; 1:241–263.
41. Jain RK. Delivery of molecular medicine to solid tumors: lessons from in vivo
imaging of gene expression and function. J Control Release 2001; 74:7–25.
42. Jain RK. Understanding barriers to drug delivery: High resolution in vivo imaging is
key. Clin Cancer Res 1999; 5:1605–1606.
43. Gref R, Minamitake Y, Peracchia MT, et al. Biodegradable Long-Circulating
Polymeric Nanospheres. Science 1994; 263:1600–1603.
44. Ringsdorf H. Structure and properties of pharmacologically active polymers. J Polymer
Sci Polymer Symp 1975; 51:135–153.
45. Davis FF. The origin of pegnology. Adv Drug Deliv Rev 2002; 54:457–458.
46. Moghimi SM, Hunter AC. Poloxamers and poloxamines in nanoparticle engineering
and experimental medicine. Trends Biotechnol 2000; 18:412–420.
47. Park EK, Lee SB, Lee YM. Preparation and characterization of methoxy poly(ethylene
glycol)/poly(epsilon-caprolactone) amphiphilic block copolymeric nanospheres
for tumor-specific folate-mediated targeting of anticancer drugs. Biomaterials 2005; 26:
1053–1061.
48. Chari RV. Targeted delivery of chemotherapeutics: tumor-activated prodrug therapy.
Adv Drug Deliv Rev 1998; 31:89–104.
49. Mansour AM, Drevs J, Esser N, et al. A new approach for the treatment of malignant
melanoma: Enhanced antitumor efficacy of an albumin-binding doxorubicin prodrug
that is cleaved by matrix metalloproteinase 2. Cancer Res 2003; 63:4062–4066.
50. Guo X, Szoka FC. Chemical approaches to triggerable lipid vesicles for drug and
gene delivery. Acc Chem Res 2003; 36:335–341.
51. Allen TM. Ligand-targeted therapeutics in anticancer therapy. Nat Rev Cancer 2002;
2:750–763.

322 Wang et al.
52. Kannagi R, Izawa M, Koike T, et al. Carbohydrate-mediated cell adhesion in cancer
metastasis and angiogenesis. Cancer Sci 2004; 95:377–384.
53. Yamazaki N, Kojima S, Bovin NV, et al. Endogenous lectins as targets for drug
delivery. Adv Drug Deliv Rev 2000; 43:225–244.
54. Leamon CP, Reddy JA. Folate-targeted chemotherapy. Adv Drug Deliv Rev 2004;
56:1127–1141.
55. Leamon CP, Low PS. Folate-mediated targeting: from diagnostics to drug and gene
delivery. Drug Discov Today 2001; 6: 44–51.
56. Stella B, Arpicco S, Peracchia MT, et al. Design of folic acid-conjugated nanoparticles
for drug targeting. J Pharm Sci 2000; 89:1452–1464.
57. Vicent MJ, Duncan R. Polymer conjugates: nanosized medicines for treating cancer.
Trends Biotechnol 2006; 24:39–47.
58. Covitz PA, Hartel F, Schaefer C, et al. caCORE: a common infrastructure for cancer
informatics. Bioinformatics 2003; 19:2404–2412.
59. Kraj P, McIndoe RA. caBIONet: a .NET wrapper to access and process genomic data
stored at the National Cancer Institute’s Center for Bioinformatics databases. Bioinformatics
2005; 21:3456–3458.
60. Weinshilboum R, Wang LW. Pharmacogenomics: Bench to bedside. Nat Rev Drug
Discov 2004; 3:739–748.
61. Evans WE, Relling MV. Moving towards individualized medicine with pharmacogenomics.
Nature 2004; 429:464–468.

Index
ABCB1 polymorphisms, 19–20
Abraxane TM , 310
Acute myeloid leukemic (AML) cells, 271
Adenosine triphosphate (ATP), 52
Adjuvant chemotherapy, 100–101, 207
Adjuvant! Online, 187, 189, 205, 207
Adjuvant on Line (AOL), vs. RS assay,
103–104
AdnaGen system, 157
Adriamycin. See Doxorubicin
Advancing Clinico Genomic Trials
(ACGT), 207
16a-[ 18 F] fluoro-17b-estradiol (FES), 298
Affymetrix GeneChip, 250, 266
Affymetrix microarray, 239
AIB1, 171
Akt, 173, 277
Aldefluor 1 , 270
Aldehyde dehydrogenase (ALDH), 270–271
ALDH. See Aldehyde dehydrogenase
(ALDH)
17-allylamino-17-demethoxygeldanamycin
(17AAG), 299
American College of Surgeons Oncology
Group (ACOSOG), 234
American Joint Committee on Cancer
(AJCC) classification system,
159–160
Aminoglutethimide, 174
Amonafide, 17
Amphiphilic copolymers, covalent linkage
of, 315
Amplicons, 244
Anastrozole, 32, 174
chemical structure of, 175
vs. tamoxifen, 177
Angiogenesis
PET and SPECT imaging of, 299
Animal model, 299
Anthracyclines
CBR and, 28
oxidative stress and, 28–29
Antiapoptosis genes, 201, 202
Anticytokeratin antibodies, 153, 154
Antiestrogens, 168–174
Anti-VEGF, 300
API. See Application programming
interface (API)
Apoptosis, 298
Application programming interface
(API), 318
Aromatase inhibitors (AI), 32, 137
analysis of gene expression with,
177–178
chemical structures of, 175
mechanisms of resistance, 175–176
steroidal and nonsteroidal, 174
vs. tamoxifen, 176–177
323

324 Index
Array CGH, 250
ATAC trial, 175–177
AutoMACS TM system, 153–155
Automatic magnetic cell sorting system.
See AutoMACS TM system
Avidin-biotin complex peroxidase
technique, 154
BAC. See Bacterial artificial
chromosomes (BAC)
Bacterial artificial chromosomes (BAC),
64–65
b-catenin, 273
Bcl-2 tumor supressor, 248
BCRP1 (breast cancer resistance
protein 1), 268
17b-estradiol (E2), 168, 201
Bioinformatics laboratories, 318
Biomarkers, 51. See also Cancer
biomarkers
Bionanoinformatics, 317–318
Biopsies, gene expression profiling in,
91–92
Biotinylated random nonamers, 246
Blood collection tubes, 216–217
Blood tisssue, 218
Bmi-1, 271–272
BMI-1 oncogenic pathway, 202
BOLD (blood oxygen level-dependent)
MRI, 300
Bone marrow
cancer stem cells in, 152
cytokeratin positive tumor cells in, 206
micrometastasis, 154
tumor cells, 154
Bone mineral density, 178
Bone scanning, 291, 292
BRCA1, 5, 50–51, 266, 267
BRCA2, 5
Breast cancer
genetic susceptibility, 5
luminal, 198
molecular classification of, 3, 198–199
patients information
storage and protection of, 231–232
treatment with genetic
polymorphism, 267
Breast cancer cells
circulating. See Circulating tumor cells
subgroups of, 202–204
Breast Cancer Intergroup, 186, 188, 190
Breast International Group (BIG) 1-98
study, 175
Breast MRI, 295–296
Breast tumor model, 269
caBIG. See Cancer Bioinformatics Grid
(caBIG)
caBIONet, 318
Cancer Bioinformatics Infrastructure
Objects (caBIO) API, 318
Cancer biomarkers, 310–311
Cancer Biomedical Informatics Grid
(caBIG), 207
development of, 317
organizational structure of, 318
Cancer cells vs. normal cells, 317
Cancer metastasis, 278
Cancer Methylation Panel, 250, 251
Cancerous tissues, vascularization in, 315
Cancer stem cells. See Stem cells
Cancer Trials Support Unit, 192
CapSure HS LCM Cap, 223, 224
Carbohydrate moieties, 316
Carbonyl reductase (CBR), 28
Carcinogenesis, 269
CAT. See Catalase
Catalase (CAT), 29
11 C-choline, 298
CD24 breast cancer cells, 202
CD44+ breast cancer cells, 202
CD44 + CD24 /loo breast cancer cells,
265, 266, 270
CD44 expression, 270
cDNA synthesis, 246, 247
CellProfile TM kit, 153
CellSearch TM system, 153
for prognostic and predictive value of
CTCs, 158
CellSpotter TM Analyzer, 153
c-erbB-2, 161
CGH. See Comparative genomic
hybridization (CGH)
Chemotherapeutic agents, 263

Index 325
Chemotherapy, 159
adjuvant. See Adjuvant chemotherapy
cytotoxic, 161
evaluation of marker, 186
neoadjuvant. See Neoadjuvant
chemotherapy
residual clinical risk and, 102–103
Chromatin immunoprecipitation (ChIP), 53
Chromogenic in situ hybridization
(CISH), 139
Circulating female hormone, 168
Circulating tumor cells (CTCs)
baseline and follow-up, 159
CD326-positive, 154, 155
clinical application of, 160–161
detection, 153–157
clinical benefits of, 161
future uses of, 162
immunostains of, 154
magnetic labeling and separation of,
219–220
prognostic value of, 158
for staging of MBC, 159–160
CISH. See Chromogenic in situ
hybridization (CISH)
Classification of breast tumors, 250
Clinical decision making, 190
Clinical Laboratory Improvement
Amendments (CLIA), 190
Clinical trials
for antiestrogens, 168
designing, 190–192
protocols, 214, 228
for treatment with MAPK-specific
inhibitors, 172
Clinicogenomic models, 207
Cloned probe arrays, in CGH, 65–66
Cluster analysis, 248
11 C-methionine, 298
CNA, Copy number abnormalities (CNA)
CNB. See Core needle biopsy (CNB)
Coexpression modules, 202
Combined imaging, with FDG-PET and
MRI, 297
Common Rule, 232
Comparative genomic hybridization
(CGH), 62, 249
BAC, 64–65
[Comparative genomic hybridization
(CGH)]
cloned probe arrays in, 65–66
DNA in, 69–71
metaphase chromosomes in, 65
MIP, 66–69
oligonucleotide-based arrays, 65–66
Complete pathological response (pCR),
101–102
Computed tomography (CT) scanning, 294
Contingency plans, 214
Copy number abnormalities (CNA), 61, 249
assessment by CGH. See Comparative
genomic hybridization (CGH)
gene deregulation by, 62–63
Coregulators, 171
Core needle biopsy (CNB), 83, 238–239
Cox model, 245
Cox regression, 159
CpG dinucleotides, 45–46
Cryopreservation techniques, 235–237
CTCs. See Circulating tumor cells (CTCs)
Cu-ATSM, 299
Cu-DOTA-VEGF, 299
CXCR4 chemokine receptor, 278, 280
Cycling cells, 267, 282
Cyclophosphamide
CYP3A4 in, 27
CYP2B6 in, 26–27
and GST, 27–28
CYP. See Cytochrome p450s (CYP)
CYP3A4, in cyclophosphamide, 27
CYP1B1, 267
CYP2B6, in cyclophosphamide, 26–27
CYP2C8
in paclitaxel, 31
CYP2D6
genotyping, 170
in TAM, 32
Cytochrome p450s (CYP), 267
Cytology, 51–52
Cytotoxic agents, in neoadjuvant
chemotherapy, 114–115
Cytotoxic therapies, 109
Dandelion hypothesis, 260
DASL assay, 246

326 Index
Database information, and patient
identifiers, 232
Daunorubicin, 28
DCIS. See Ductal carcinoma in situ
(DCIS)
Deacetylation, of histones, 47
Decision aids, 187
Decision boards, 187
DES. See Diethylstilbestrol (DES)
DFS. See Disease-free survival
3D GRE technique, 296–297
Diethylstilbestrol (DES), vs.
tamoxifen, 168
Dihydropyrimidine dehydrogenase
(DPD), 18–19
Discovery-based research, 185
Disease-free survival (DFS), 32–36, 109
defined, 192
GST in, 33
MnSOD, 36
MTHFR in, 33
SULT1A1, 36
UGT2B15, 36
Disseminated tumor cells (DTC), 206
Distant recurrence-free survival (DRFS)
defined, 192
DNA, in CGH, 69–71
DNA hypermethylation, 279
DNA isolation, 225–226
DNA methylation, 45
DNMT in, 46–47
inhibitors, 53–54, 55
DNA methyltransferases (DNMT), 46–47
DNA microarray analysis, 107–108
DNA synthesis, 298
4 0 ,6-doamidino-2-phenylindole
(DAPI), 153
Docetaxel, 19–20
Doxorubicin (DOX), 28, 316
DPD. See Dihydropyrimidine
dehydrogenase (DPD)
Drosophila, 276
Drugs
chemotherapy, 159
development
chemotherapy, 159
for CTC, 161
predictive markers in, 141–143
DSL ligands, 276
Ductal carcinoma in situ (DCIS), 174
E2. See 17b-estradiol (E2)
Eastern Cooperative Oncology Group
(ECOG), 192
EasySep TM system, 155–157
EGFR. See Epidermal growth factor
receptor (EGFR)
Embryogenesis, 269
Embryonic stem cells, 271, 279
Endocrine therapy, 167–168
with antiestrogens, 168–174
with aromatase inhibitors,
174–178
Endothelial cells, 279, 280
magnetic labeling and separation of,
219–220
Endoxifen. See 4-hydroxy-N-desmethyl
tamoxifen
Enzymes, metabolic, 267
EpCAM. See Epithelial cell adhesion
molecule-1 (EpCAM)
Epidermal growth factor receptor (EGFR),
140–141, 171, 172
Epigenetics
DNA methylation. See DNA
methylation
histones. See Histones
Epithelial cell adhesion molecule-1
(EpCAM), 153
Epithelial stem cells, 263
ER. See Estrogen receptors (ER)
ERa, 136–138, 168
gene transcription, 169
ERa-positive breast cancer, 168, 173
ERa-positive cells, 269
ERBB2, 62, 311
ER-negative breast cancer, 168, 170
ER-negative tumor cells, 269–270
ER-positive breast cancer, 187, 246
ER-positive tumors, 198, 245
Estrogen receptors (ER), 51, 108, 298
ERa, 136–138, 168
gene transcription, 169
gene expression of, 55–56
mutations, 170

Index 327
European Human Frozen Tumor Tissue
Bank, 234
European Union, 207
Exemestane, 174
chemical structure of, 175
ExtractSure device, 223, 224
FDG PET, 293–294
FES. See 16a-[ 18 F] fluoro-17b-estradiol
(FES)
FES PET, 298
18 F-fluorodeoxyglucose (FDG), 292
[ 18 F] fluorotamoxifen, 298
FFPE. See Formalin-fixed
paraffin-embedded (FFPE)
FFPE tissues, 313
Fibroblasts, 280
Fibromas, 297
Ficoll-Paque TM , density gradient
procedure, 156
Fine needle aspiration (FNA), 83, 238
Fixed tissue, 218
18 F-labeled paclitaxel, 298
Flow cytometry, 264
Fluorescence-activated cell sorter
(FACS), 270, 271
Fluorescence microscope, 312
Fluorescent in situ hybridization
(FISH), 311
for HER2/neu testing, 138–139
Fluorodeoxyglucose, 293
5-Fluorouracil (5FU)
and DPD, 18–19
DPD and, 30
MTHFR and, 30, 31
TS and, 31
TS inhibition by, 19
[ 18 F]MISO ([ 18 F]fluoromisonidazole-
3-fluoro-1-(2 0 -nitro-1 0 -imidazolyl)-
2-propanol), 299
FNA. See Fine needle aspiration (FNA)
Folate receptors, 316
Food and Drug Administration
(FDA), 153, 174, 246,
311, 3111
Formalin, 215, 218
cross-linking effects of, 237
Formalin-fixed paraffin-embedded
(FFPE) tissue
extraction of RNA from tissue samples
of, 98–99, 244–245
RNA-based expression profiling of, 246
Forward genetics approach, 16
FOXO regulatory proteins, 277
Freeze-thawing cycles, 235
Fulvestrant, 174
Ga-DOTA-F(ab 0 )(2)-herceptin, 299
Gamma rays, 292
Gamma secretase inhibitor (GSI), 276
Gefitinib, 172, 311
Gene expression
analysis, 177–178, 244–245
profiling. See Gene expression profiling
70-gene expression, 108, 124–126, 205.
See also Gene expression signature
in MINDACT trial, 130–132
prognostic value of, 126–127
in RASTER trial, 130
in TRANSBIG study, 127–129
Gene expression analysis, 177–178,
244–245
Gene expression Grade Index (GGI),
200–201
Gene expression profiling, 79–81, 97–98,
198–199, 246, 270
and chemotherapy, 100–101, 111–114
in neoadjuvant setting, 101–102
for classification of breast cancer,
198–199
and docetaxel, 112–113
oncogenic pathway analysis by, 116–118
pCR prediction, 101–102
RNA expression, 98–100
RS in. See Recurrence score (RS) assay
RT-PCR for, 99
Gene expression signatures, 265–267. See
also Gene expression profiling
for improving prognosis, 199–202
integration with, 202–204
Gene methylation
BRCA1, 50–51
estrogen receptors (ER), 51
RASSF1A, 49–50

328 Index
21-gene profile. See Oncotype DX TM
76-gene profile, 129
Genetic polymorphisms, 266–267
Genetic susceptibility, 5
Gene transcription, 19
Genomic alterations
prognostic value of, 158
Genomic Health, 190, 191
Genomic Health, Inc. (GHI), 246
Genomic hypomethylation, 279
Genomic predictors, 204–205
Germline polymorphisms, 267
Gestation, 269
GHI. See Genomic Health, Inc. (GHI)
Gliomas, 268
Glutathione-S-transferases (GST), 16, 267
and cyclophosphamide, 27–28
in DFS, 33
Glycophorin A, 156
Graphical interface, user-friendly, 315
GRP (gastrin-releasing peptide)
receptors, 299
GSK-3b, 273
Harmonization, 214
HAT. See Histone acetyltransferases
(HAT)
Health Insurance Portability and
Accountability Act, 232
Hedgehog (HH), 265
ligands, 276–277
signaling pathway, 276–277
transcription factors, Gli1 and
Gli2, 277
H & E (hematoxylin & eosin), 227
Hematopoietic cells, human, 156
Heparin-coated syringes, 215
HER2. See Human epidermal growth
factor receptor 2 (HER2)
HER2+, 198
HER3, 171
HER4, 171
HER2/neu, 138–139, 171–172
Heterochromatin, 47
HH. See Hedgehog (HH)
Histone acetyltransferases (HAT), 47
Histone deacetylases (HDAC), 55
Histones
deacetylation, 47
HAT, 47
HDAC. See Histone deacetylases
(HDAC)
methylation of, 278
modifiers, 48
posttranslational modification of, 47
ChIP for, 53
H3K9 repressors, 278
Homogenization, 224
Hormonal therapy, 191
adjuvant, 187
Hormone receptors, 161
Hormone replacement therapy
(HRT), 173
Human epidermal growth factor receptor
2 (HER2). See also HER2/neu
gene amplification in CTCs, 158
Hybridization. See Comparative genomic
hybridization (CGH)
4-hydroxy-N-desmethyl tamoxifen,
20, 170
Hypothesis-based research, 185
IHC. See Immunohistochemistry (IHC)
Illumina, Inc., 249
Illumina BeadArray, 250
Illumina microarrays, 249
Image file processing, 315
Immunohistochemistry (IHC), 135–136,
227, 311, 313
for ERa, 136–138
for HER2/neu, 138
for PR, 136–138
Immunohistological staining, 311–312
Indium-111, 300
In situ hybridization (ISH), 227
International Stage IV Stratification
Study, 160
International Union Against Cancer
(UICC), 159
Invasiveness gene signature (IGS), 202
Iressa. See Gefitinib
Isoniazid, 16–17

Index 329
Java objects, 318
Kaiser Permanente health care system,
188
Kaplan-Meier estimator, 245
Krishnamurthy, Savitri, 154
Laser capture microdissection (LCM),
2–3, 221–224, 239
Lauterbur, Paul, 294
LCM. See Laser capture microdissection
(LCM)
Lectin-carbohydrate, 316
Lectins, 316
Lethal phenotypes, 310
Letrozole, 32, 174
chemical structure of, 175
vs. tamoxifen, 176, 177
Ligand-binding domain (LBD), 170
Liver tumors, 297
Lobular carcinoma in situ (LCIS), 174
Loss of heterozygosity (LOH), 66, 250
Luminal epithelial cells, genes of, 198
Lymph nodes, 297
microarrays for, 3–5
recurrence of breast cancer in, 245
Macromolecular extraction, 237, 239
Magnetic labeling and separation, of
tumor tissue, 219–220
MammaPrint. See 70-gene expression
Mammary gland, 264
Mammography, 292
Manganese superoxide dismutase
(MnSOD), 29
in DFS, 36
MAPK. See Mitogen-activated protein
kinase (MAPK)
MassARRAY, 52
Matlab slide scanners, 315
Matrix metalloproteinase (MMP), 316
MBC. See Metastatic breast cancer
(MBC)
MCF-7 breast cancer cells, 170, 172
Medicare, 190
Mercaptopurine, 17
Metaphase chromosomes, in CGH, 65
Metastatic breast cancer (MBC)
development of, 151
diagnosis of, 158–159
stage IV, 160
HER2-positive, 177, 178
staging, 159–160
Methotrexate (MTX), 18, 30, 31
Methylation
detection of, 51–53
of DNA. See DNA methylation
of genes. See Gene methylation
profiling, 250–251
therapy, 53–55
Methyltetrahydrofolate reductase
(MTHFR), 18
in DFS, 33
and 5-Fluorouracil (5FU), 30, 31
Microarray for Node-Negative Disease
may Avoid Chemotherapy
(MINDACT) trial, 130–132, 205
Microarray Quality Control Project
(MAQC), 82
Microarrays
for lymph node status, 3–5
with OCT, 236
with paraffin-embeded tissue, 236–237
RNA expression, 2, 4
with RNA stabilization buffers,
237–238
technical reproducibility of, 82–83
using CNB, 238–239
using FNA, 238
using LCM, 239
Micrometastatic carcinoma, 154
Micrometastatic cells, 206
microRNAs, 206
Microsoft Excel, 228
MIP. See Molecular inversion probe (MIP)
miRNA profiling, 248–249
Mitogen-activated protein kinase
(MAPK), 172
MMP. See Matrix metalloproteinase
(MMP)
MMP-2, 316

330 Index
MnSOD. See Manganese superoxide
dismutase (MnSOD)
Molecular classification, of breast
cancer, 3
Molecular diagnostic test, 186, 188,
190, 311
Molecular epidemiologic studies, 32–36
Molecular imaging, 291–292. See also
Positron emission tomography
(PET)
Molecular inversion probe (MIP), 66–69
PCR in, 67
Molecular markers, 79, 160, 188–190
defined, 185–186
development process of, 81–82
in stem cells, 265, 273
Molecular profiling
defined, 185
developmental technologies for, 311
Molecular signatures, 186
Monoclonal antibodies, 153, 156
Monocytes, 219
MPO. See Myeloperoxidase (MPO)
MRI, 294–297
applications for, 297
mRNAs, 245
MTA-1 (Manual Tissue Arrayer), 226
MTHFR. See Methyltetrahydrofolate
reductase (MTHFR)
mTOR (mammalian targets of
rapamycin), 276
MTX. See Methotrexate
Mutations
estrogen receptors, 170
Myeloperoxidase (MPO), 29
N-acetyltransferases, 16, 267
Nanoparticles. See also Nanotechnology
bioconjugated, 311, 317
design and development of, 318
molecular profiling. See Nanotyping
superparamagnetic iron oxide, 310
Nanotechnology, 311
for cancer detection, diagnosis, and
therapy, 319
QD, 312
Nanotherapeutics, 315–317
Nanotyping, 311–315, 318–319
National Cancer Institute (NCI), 186,
207, 317
National Institutes of Health (NIH),
199, 200
cancer guidelines, 205
National Surgical Adjuvant Breast and
Bowel Project (NSABP), 99–100,
188, 190, 192, 245
N-CoR, 171
Neoadjuvant chemotherapy, 5–7, 78,
109–111
cytotoxic agents in, 114–115
predictive markers of response to,
101–102
response predictors of, 78–79
use of, 293
Netherlands Cancer Institute (NKI), 199
.NET technology, 318
Noncanonical pathways. See Wnt
signaling
Nonnucleoside inhibitors, 53, 54
Non-small cell lung cancer (NSCLC), 310
Nontumorigenic cells, 265
North American Breast Intergroup, 192
Notch receptors, 276
Notch signaling pathway, 264, 276
Nottingham Prognostic Index, 205
NSABP. See National Surgical Adjuvant
Breast and Bowel Project
(NSABP)
Nucleic acids
amplification, 238
collection, 215
extraction, 237
profilation, 251
Nucleoside inhibitors, 53, 54
NuGEN Technologies, Inc., 249
OCT. See Optimal cutting temperature
(OCT)
OCT-4, 271
Octreoscan, 299
[ 15 O]H 2 O PET, 292
4OHT (4-hydroxytamoxifen), 171
Oligonucleotide-based arrays,
in CGH, 65–66

Index 331
Oligonucleotides, 247
Oncogenes, 62, 251
Oncogenic pathways, deregulation of,
116–118
cytotoxic agents, 118
Oncologists, 207
Oncotype DX assay, 188, 190, 192, 194,
200, 205, 246
Optimal cutting temperature (OCT),
218, 236
p53
signature genes, 201
tumor suppressor, 248
PACCT. See Program for the assessment
of clinical cancer tests (PACCT)
Paclitaxel, 19–20, 267
CYP2C8 in, 31
Paget, Stephen, 278
Papanicolaou (Pap)-stained cytospin, 154
Paraffin, 222, 227
Paraffin-embedded tissue, 237–238
Paraffin wax, discovery of, 237
Partial signatures, 204
Pathologic report, 234
Pathologist, 232, 233
Patient identifiers, 232
PAXgene, 220–221
p21 Cip1/Waf1 , 277
PCR. See Polymerase chain reaction
(PCR); Polymerase chain reaction
(PCR), in MIP
pCR. See Complete pathological response
(pCR)
Peptides, 299
Peripheral blood
cancer stem cells in, 152
extraction of RNA from, 220–221
isolation of monocytes in, 219
separation from mononuclear cells, 156
Peripheral blood mononuclear cells
(PBMC), 219
PET. See Positron emission tomography
(PET)
PET-MRI scanner, 300
PET scanners, 297
Pharmacodynamics, 13
Pharmacoepidemiology, 14
Pharmacogenetics, 14, 206
molecular epidemiologic studies, 32–36
Pharmacogenomic predictors, 79–80
published studies, 86–90
validation studies of, 84–85
Pharmacokinetics, 13
Pharmacology, 13–14
Phosphatase and tensin homolog deleted
on chromosome ten (PTEN),
265, 277
Phosphate buffered saline (PBS), 215
Phosphatidylinositol-3-OH kinase (PI3K),
173, 277
Polycomb group (PcG) proteins, 46,
261, 279
Polymerase chain reaction (PCR), 236
detection of tumor cells, 158
in MIP, 67
Positron emission tomography (PET),
206, 292
Posttranslational modification,
of histones, 47
PR. See Progesterone receptor (PR)
Predictive markers
defined, 136
in drug development, 141–143
ERa, 136–138
HER2/neu, 138–139
PR, 136–138
Preoperative chemotherapy. See
Neoadjuvant chemotherapy
Preoperative systemic chemotherapy
(PST). See Neoadjuvant
chemotherapy
Preregistration, 191
Preservatives, 214, 215
Primary study group, 192
Privacy rights, 231
Progenitor cells, mutated, 262, 266
Progesterone receptor (PR), 136–138,
176–177
Prognosis, in breast cancer
improving, 199–202
Prognostic marker, 199
Program for the assessment of clinical
cancer tests (PACCT), 186
Proliferation genes, 201, 202, 204

332 Index
Prostate cancer, RNA signatures in, 248
Proteinase digestion buffers, 237
Protein biomarkers, 313. See also Cancer
biomarkers
Protein collection, 215
Protein isolation, 225–226
Proteomics, 206
PTEN. See Phosphatase and tensin
homolog deleted on chromosome
ten (PTEN)
PTEN pathway, 277–278
Pyrosequencing, 52–53
QD-Ab, 312
QDs. See Quantum dots (QDs)
Q-IHC tools, 314
qRT-PCR (realtime quantitative-PCR)
for screening tumor samples for
miRNA expression, 248
Quantification system, 315
Quantum dots (QDs)
multicolor, 312
nanotechnology, 312
semiconductor, 311
Radiation, resistance to
(Radioresistance), 266
Radionuclide techniques, 292–294
vs. optical imaging, 299
RAKE (RNA-primed array-based Klenow
extension) analysis study, 249
Randomization, 192
Rapamycin, molecular targets of. See
MTOR (mammalian targets of
rapamycin)
RASSF1A, 49–50
RASTER trial, 130
Rationally designed inhibitors, 53,
54, 55
Reactive oxygen species (ROS), 27
and mitochondrial permeability, 29
Recurrence score (RS), 100, 200, 245
and adjuvant chemotherapy, 100–101
clinical utility of, 103–104
GHI, 246
for node positive patients, 104
[Recurrence score (RS)]
and pCR, 101
to predict recurrence of breast
cancer, 245
vs. AOL, 103–104
REMARK guidelines, 186
Reporter genes, 299
Response predictors, of neoadjuvant
chemotherapy, 5–7, 78–79
molecular markers as, 79
Reticuloendothelial system (RES), 315
Reverse transcriptase-polymerase chain
reaction (RT-PCR), 99, 244
assay-based evaluation of ER
expression, 189
and gene signature, 200
RNA
amplification, 237–238
degradation, 235, 247
expression microarray, 2, 4
hybridization, 238, 239
isolation, 224–225
RNAlater, 218, 236
RNA stabilization buffers, 236–237
RosetteSep TM method, 156
Rotterdam group, 199
RPMI (Roswell Park Memorial Institute)
media, 218
RS. See Recurrence score (RS) assay
RT-PCR assay, 245
Sample handling, 237
Selective ER downregulators (SERDs),
173, 174
Selective ER modulators (SERMs), 173
Semiautomatic image processing, 315
Single nucleotide polymorphism (SNP),
14–15, 249
profiling, 250
Single-photon imaging, 292
SMRT, 171
SNP. See Single nucleotide polymorphism
(SNP)
Sonic hedgehog (SHH), 276, 277
SRC-1, 171
St. Gallen classification, 246
St. Gallen criteria, 199, 200

Index 333
St. Gallen guidelines, 205
Standard operating procedures (SOPs)
for collection of tissue samples, 234
for RNA degradation, 235
STAR (Study of Tamoxifen and
Raloxifene) trial, 173
STAT3 transcription factor, 276
Stem cell lines, derivation of, 266
Stem cell niche, 261, 263, 268
molecular targets of, 278–280
Stem cell regulation, 280
Stem cells
cellular proliferation, 261
clinical research on, 264–265
differentiation of, 261, 262
division of, 276
drug resistance of, 268
embryonic, 271, 279
functional targets, 269–273
gene expression profiling of, 270
hematopoietic, 278, 280
hypothesis, 260–264, 268
intestinal, 277
leukemic, 278
mesenchymal, 280
molecular markers, 265, 273
mutations, 263, 270, 278
nontumorigenic daughter cells of, 261
Stem cell sanctuaries, 267–269
Stem cell survival, neural, 279, 280
Stochastic model, 260, 261
SULT1A1
in DFS, 36
in TAM, 32
Superoxide dismutases (SODs), 267
Swiss Institute of Bioinformatics, 202
TAM. See Tamoxifen (TAM)
Tamoxifen (TAM), 20, 31–32, 56, 137
adjuvant, 169
CYP2D6 in, 32
metabolizers of, 170
SULT1A1 in, 32
UGT2B15 in, 32
vs. aromatase inhibitors, 176–177
vs. DES, 168
TaqMan TM array, 248–249
Taxanes, 19–20
CYP2C8 in, 31
Thiopurine methyltransferase (TPMT),
17–18
Thymidylate synthase (TS), 19
Tissue
acquisition of, 232–235
biopsies, 218
collection, 214–218
control blocks, 228
defined, 213
processing, 218
procurement, 214
Tissue banking, 231–232
Tissue microarray (TMA), 226–227
blocks, 228, 229, 230
mapping, 227, 228
TMA. See Tissue microarray (TMA)
TRANSBIG study, 127–129, 199–200
Transcriptional regulation, 277
Trastuzumab, 138, 172, 310–311
Trial Assigning Individualized Options for
Treatment (TAILORx) trial, 178, 186
designing of, 191
recurrence score ranges in, 192–194
TRI reagent
for isolation of DNA, 225–226
for isolation of RNA, 224–225
for tissue microarray creation, 226–230
TRIZOL method, 224–230
TS. See Thymidylate synthase (TS)
Tumor-activated prodrug therapy, 316
Tumor growth factor-b (TGF-b), 280
Tumor hypoxia, 299
Tumorigenesis, 250, 273, 274
Tumors
detection of, 158
invasion, 202
regression, 260
size and growth, 238, 239
Tumor suppressor genes, 251
Tumor tissue. See Tissue
UGT2B15
in DFS, 36
in TAM, 32
Ultrasound, 292

334 Index
Vacutainers, use of, 215
Vascular endothelial growth factor
(VEGF), 279, 280
Vascularization, 315
VEGF. See Vascular endothelial growth
factor (VEGF)
White blood cells. See Monocytes
Whole-genome expression profiling, 249
Wnt/b-catenin, 272
Wnt signaling, 266, 273
World Wide Web for Cancer Research,
317
Wound-healing, gene phenotype, 200
WWOX expression, 56
Xenograft model, 265
Xenotransplantation, 264, 269
X-ray mammography, 297
Xylene, 218

Oncology
about the book…
Pharmacogenetics is becoming increasingly relevant in the diagnosis, treatment,
and recovery of cancer patients. A major problem facing oncologists is the outstanding
varied efficacy of treatment. Promising advances in pharmacogenetics
have allowed the development of effective agents which will enable personalized
cancer chemotherapy to become routine for the clinical practice.
Written by experts in the field and combining information that is unable to be
found in a single source volume, Pharmocogenetics of Breast Cancer:
• combines a complete overview of pharmacogenetics and how it relates to
breast oncology for diagnosis, treatment, and the recovery of patients using
an individualized therapy model
• is the first single source reference demonstrating the application of
pharmacogenetics in the care of breast cancer patients
• enables physicians a coherent interpretation of the emerging science of
pharmacogenetics, aiding them to incorporate individual therapies in their
own practice
• gives practical guidance on various forms of specimen collection, tissue
selection, and handling procedures
about the editor...
BRIAN LEYLAND-JONES is Associate Vice President and Director, Emory Winship
Cancer Institute, Emory University, Atlanta, Georgia. Dr. Leyland-Jones’ main
research interests are pharmacodynamics, pharmacokinetics, and pharmacogenetics
in oncological clinical trials; translation of preclinical models into the clinic;
biomarker endpoints in Phase I/II clinical trials; and screening and mechanistic
studies of novel targeted and chemotherapeutic anti-cancer agents. He has
authored more than 125 peer-reviewed articles and book contributions, 150
abstracts, and 29 patents.
Printed in the United States of America
H8637