02.03.2013 Views

Target Discovery and Validation Reviews and Protocols

Target Discovery and Validation Reviews and Protocols

Target Discovery and Validation Reviews and Protocols

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

98 Sørlie<br />

are not internally heterogeneous, when millions of cells are sampled. The finding<br />

that the two primary tumors <strong>and</strong> the corresponding metastases were similar in<br />

their overall pattern of gene expression suggests that the molecular program of a<br />

primary tumor may be retained in its metastases. This idea has been explored by<br />

other investigators <strong>and</strong> suggests that metastatic capacity is an intrinsic feature of<br />

the primary tumor <strong>and</strong> that the fate of the tumor might have been programmed at<br />

an early stage of the disease (21). Hence, specific expression patterns in the primary<br />

tumors may help identifying genes, pathways, or both that are important for<br />

the metastatic process <strong>and</strong> for tumor progression in general.<br />

3.2. Subclassification<br />

To explore the possibilities for further refining these distinctions among subtypes<br />

of breast tumors, we took advantage of the paired tumor samples. The<br />

specific features of a gene expression pattern that are to be used as the basis for<br />

classifying tumors should be similar in any sample taken from the same tumor,<br />

<strong>and</strong> they should vary among different tumors. The paired samples therefore provided<br />

a unique opportunity for a deliberate <strong>and</strong> systematic search for genes<br />

whose expression levels reflected such intrinsic characteristics of the tumors.<br />

Using well measured expression data from the paired samples, a subset of genes<br />

termed “intrinsic gene set” was selected that consisted of genes with significantly<br />

greater variation in expression between different tumors than between<br />

paired samples from the same tumor (14). The rationale behind this list of genes<br />

is that it is enriched for those genes whose expression patterns were characteristic<br />

of each tumor as opposed to those that varied as a function of tissue sampling;<br />

hence, they would be ideally suited for classification.<br />

Our extended analyses have since included an intrinsic gene set of 540 genes<br />

selected from expression data of 45 tumor pairs (including two primary lymph<br />

node pairs) <strong>and</strong> approx 8000 common genes for all samples (22,23). Together,<br />

122 tissue samples were included in the clustering analysis: 115 carcinomas<br />

<strong>and</strong> 7 nonmalignant tissues. Most of the tumors were sampled as part of two<br />

independent studies evaluating response to chemotherapy of locally advanced<br />

breast cancer in an adjuvant setting. From the first cohort of patients treated<br />

with doxorubicin monotherapy (19), 55 tumor samples were analyzed, <strong>and</strong><br />

from the second cohort of patients treated with 5-fluorouracil <strong>and</strong> mitomycin C<br />

(20), 34 tumor samples were analyzed. The remaining 26 samples were primary<br />

tumor specimens collected either at Stanford or in Norway.<br />

The overall expression patterns showed that the tumors could be classified<br />

into five distinct subtypes, <strong>and</strong> the main distinction was between tumors that<br />

expressed genes characteristic of luminal epithelial cells (two groups), including<br />

the estrogen receptor (ER) <strong>and</strong> those that were negative for these genes (three<br />

groups) (Fig. 3). Among the luminal epithelial-type tumors, the largest group,

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!