Untitled
Untitled
Untitled
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
in this issue<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Sterile bollworms bolster Bt control<br />
Evolution of resistance<br />
by insect pests<br />
threatens to cut short<br />
the benefits of crops<br />
engineered to produce<br />
insecticidal proteins<br />
from the bacterium<br />
Bacillus thuringiensis<br />
(Bt). Tabashnik and<br />
colleagues use computer<br />
simulations and data<br />
from an extensive 4-year field trial to demonstrate that a multitactical<br />
program, including releases of sterile pink bollworm<br />
moths, can thwart the emergence of the pest’s resistance to<br />
Bt cotton. Although the sterile insect technique is not new,<br />
this is the first time it has been used jointly with a transgenic<br />
crop. From 2006 to 2009, 8 billion pink bollworm caterpillars<br />
were reared to maturity (pictured), and moths sterilized by<br />
irradiation were released over the cotton fields of Arizona. This<br />
substantially increased the likelihood that any rare resistant<br />
moths emerging from Bt cotton would mate with sterile moths,<br />
blocking their reproduction. The results were no increase in<br />
pink bollworm resistance to Bt cotton, a 99.9% decrease in<br />
the size of Arizona’s pink bollworm population and a sharp<br />
reduction in the use of insecticide sprays. It remains to be seen<br />
whether the sterile insect technique will be as effective for<br />
other combinations of Bt crops and the insect pests that they<br />
target. [Letters, p. 1304; News and Views, p. 1273] PH<br />
Synthesizing DNA aplenty<br />
Long synthetic DNA<br />
molecules have been<br />
assembled from inexpensive<br />
short oligonucleotides<br />
created on<br />
a microarray, but this<br />
approach has not been<br />
widely adopted owing<br />
to high error rates<br />
and limited scalability.<br />
Studies by two groups working with George Church address these<br />
problems. They describe clever approaches to selecting subsets of<br />
oligos from the complex pool released from the microarray. Without<br />
performing this selection step, the oligos from a microarray are not<br />
abundant enough and may cross-hybridize with each other, interfering<br />
with the assembly process. Kosuri et al. select subpools of oligos<br />
using PCR, and they lower error rates using high-fidelity microarrays<br />
Written by Kathy Aschheim, Michael Francisco, Peter Hare,<br />
Brady Huggett, Craig Mak & Lisa Melton<br />
and optimized assembly protocols. Matzas et al. describe a novel<br />
use of next-generation sequencing, reducing error rates and selecting<br />
oligos simultaneously by reading oligos with a high-throughput<br />
sequencer to identify the ones that do not have errors and then manually<br />
picking these off the sequencing instrument with a micropipette.<br />
The authors use their methods to synthesize a variety of functioning<br />
proteins, including fluorescent proteins and therapeutic antibody<br />
fragments. The two approaches each reduce DNA synthesis<br />
costs to
© 2010 Nature America, Inc. All rights reserved.<br />
in this issue<br />
urine. The results of this study could help in designing nanoparticles<br />
with desired biodistribution properties after inhalation.<br />
[Letters, p. 1300; News and Views, p. 1275]<br />
KA<br />
Patent roundup<br />
Identifying and isolating DNA without further manipulation does<br />
not constitute an invention and is not patentable, a US district<br />
court said in October. The court stated its position in response to a<br />
lawsuit involving Myriad Genetics’ (Salt Lake City, UT, USA) breast<br />
cancer genes BRCA1 and BRCA2. In a Commentary, Kenneth<br />
Chahine discusses a new framework for assessing the eligibility of<br />
DNA patents [News, p. 1226; Commentary, p. 1251] LM<br />
Patent owners who can claim a humanitarian use for their<br />
patents will be offered a fast-track re-examination voucher<br />
under a system proposed by the US Patent and Trademark<br />
Office. [News, p. 1226]<br />
LM<br />
‘Blocking patents’ have killed many a startup. Here is advice on<br />
how to deal with them. [Building a Business, p. 1239] BH<br />
A whole host of approaches fall under the rubric ‘therapeutic<br />
nanoparticles’. Burgess et al. outline the reasons why innovative<br />
pharmaceutical products based on therapeutic nanoparticles offer<br />
particular advantages in terms of intellectual property protection<br />
from generic competition. [Patent Article, p. 1267]<br />
MF<br />
Recent patent applications in pharmacogenomics.<br />
[New patents, p. 1271]<br />
MF<br />
Minicircles for all<br />
Transgenes are often ferried into cells on plasmids, but an efficient way<br />
of producing minicircle DNA could make minicircles the method of<br />
choice for delivering exogenous genes to mammalian cells. Minicircles<br />
are circular expression vectors that lack the plasmid backbone. Longterm<br />
transgene expression from minicircles is much higher than from<br />
minicircles, but broad adoption of this technology has been hindered<br />
by a cumbersome production process. Kay and colleagues present<br />
an optimized protocol that yields purified minicircles in about the<br />
same time frame as that required for routine plasmid preparation.<br />
[Brief Communications, p. 1287]<br />
KA<br />
Next month in<br />
• Safe harbors in human iPS cells<br />
• Genome-wide profiling of cytosine hydroxymethylation<br />
• Experimental halotype phasing<br />
• Improving in vivo shRNA screens with fluorescent<br />
reporters<br />
• Genetic modification of rodents with zinc finger<br />
nucleases<br />
viii<br />
volume 28 number 12 December 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.<br />
www.nature.com/naturebiotechnology<br />
EDITORIAL OFFICE<br />
biotech@us.nature.com<br />
75 Varick Street, Fl 9, New York, NY 10013-1917<br />
Tel: (212) 726 9200, Fax: (212) 696 9635<br />
Chief Editor: Andrew Marshall<br />
Senior Editors: Laura DeFrancesco (News & Features), Kathy Aschheim (Research),<br />
Peter Hare (Research), Michael Francisco (Resources and Special Projects)<br />
Business Editor: Brady Huggett<br />
Associate Business Editor: Victor Bethencourt<br />
News Editor: Lisa Melton<br />
Associate Editors: Markus Elsner (Research), Craig Mak (Research)<br />
Editor-at-Large: John Hodgson<br />
Contributing Editors: Mark Ratner, Chris Scott<br />
Contributing Writer: Jeffrey L. Fox<br />
Senior Copy Editor: Teresa Moogan<br />
Managing Production Editor: Ingrid McNamara<br />
Production Editor: Amanda Crawford<br />
Senior Illustrator: Katie Vicari<br />
Illustrator: Marina Corral<br />
Cover design: Erin DeWalt<br />
Senior Editorial Assistant: Ania Levinson<br />
MANAGEMENT OFFICES<br />
NPG New York<br />
75 Varick Street, Fl 9, New York, NY 10013-1917<br />
Tel: (212) 726 9200, Fax: (212) 696 9006<br />
Publisher: Melanie Brazil<br />
Exectutive Editor: Veronique Kiermer<br />
Chief Technology Officer: Howard Ratner<br />
Head of Nature Research & Reviews Marketing: Sara Girard<br />
Circulation Manager: Stacey Nelson<br />
Production Coordinator: Diane Temprano<br />
Head of Web Services: Anthony Barrera<br />
Senior Web Production Editor: Laura Goggin<br />
NPG London<br />
The Macmillan Building, 4 Crinan Street, London N1 9XW<br />
Tel: 44 207 833 4000, Fax: 44 207 843 4996<br />
Managing Director: Steven Inchcoombe<br />
Publishing Director: Peter Collins<br />
Editor-in-Chief, Nature Publications: Philip Campbell<br />
Marketing Director: Della Sar<br />
Director of Web Publishing: Timo Hannay<br />
NPG Nature Asia-Pacific<br />
Chiyoda Building, 2-37 Ichigayatamachi, Shinjuku-ku, Tokyo 162-0843<br />
Tel: 81 3 3267 8751, Fax: 81 3 3267 8746<br />
Publishing Director — Asia-Pacific: David Swinbanks<br />
Associate Director: Antoine E. Bocquet<br />
Manager: Koichi Nakamura<br />
Operations Director: Hiroshi Minemura<br />
Marketing Manager: Masahiro Yamashita<br />
Asia-Pacific Sales Director: Kate Yoneyama<br />
Asia-Pacific Sales Manager: Ken Mikami<br />
DISPLAY ADVERTISING<br />
display@us.nature.com (US/Canada)<br />
display@nature.com (Europe)<br />
nature@natureasia.com (Asia)<br />
Global Head of Advertising and Sponsorship: Dean Sanderson, Tel: (212) 726 9350,<br />
Fax: (212) 696 9482<br />
Global Head of Display Advertising and Sponsorship: Andrew Douglas, Tel: 44 207 843 4975,<br />
Fax: 44 207 843 4996<br />
Asia-Pacific Sales Director: Kate Yoneyama, Tel: 81 3 3267 8765, Fax: 81 3 3267 8746<br />
Display Account Managers:<br />
New England: Sheila Reardon, Tel: (617) 399 4098, Fax: (617) 426 3717<br />
New York/Mid-Atlantic/Southeast: Jim Breault, Tel: (212) 726 9334, Fax: (212) 696 9481<br />
Midwest: Mike Rossi, Tel: (212) 726 9255, Fax: (212) 696 9481<br />
West Coast: George Lui, Tel: (415) 781 3804, Fax: (415) 781 3805<br />
Germany/Switzerland/Austria: Sabine Hugi-Fürst, Tel: 41 52761 3386, Fax: 41 52761 3419<br />
UK/Ireland/Scandinavia/Spain/Portugal: Evelina Rubio-Hakansson, Tel: 44 207 014 4079,<br />
Fax: 44 207 843 4749<br />
UK/Germany/Switzerland/Austria: Nancy Luksch, Tel: 44 207 843 4968, Fax: 44 207 843 4749<br />
France/Belgium/The Netherlands/Luxembourg/Italy/Israel/Other Europe: Nicola Wright,<br />
Tel: 44 207 843 4959, Fax: 44 207 843 4749<br />
Asia-Pacific Sales Manager: Ken Mikami, Tel: 81 3 3267 8765, Fax: 81 3 3267 8746<br />
Greater China/Singapore: Gloria To, Tel: 852 2811 7191, Fax: 852 2811 0743<br />
NATUREJOBS<br />
naturejobs@us.nature.com (US/Canada)<br />
naturejobs@nature.com (Europe)<br />
nature@natureasia.com (Asia)<br />
US Sales Manager: Ken Finnegan, Tel: (212) 726 9248, Fax: (212) 696 9482<br />
European Sales Manager: Dan Churchward, Tel: 44 207 843 4966, Fax: 44 207 843 4596<br />
Asia-Pacific Sales & Business Development Manager: Yuki Fujiwara, Tel: 81 3 3267 8765,<br />
Fax: 81 3 3267 8752<br />
SPONSORSHIP<br />
g.preston@nature.com<br />
Global Head of Sponsorship: Gerard Preston, Tel: 44 207 843 4965, Fax: 44 207 843 4749<br />
Business Development Executive: David Bagshaw, Tel: (212) 726 9215, Fax: (212) 696 9591<br />
Business Development Executive: Graham Combe, Tel: 44 207 843 4914, Fax: 44 207 843 4749<br />
Business Development Executive: Reya Silao, Tel: 44 207 843 4977, Fax: 44 207 843 4996<br />
SITE LICENSE BUSINESS UNIT<br />
Americas: Tel: (888) 331 6288<br />
institutions@us.nature.com<br />
Asia/Pacific: Tel: 81 3 3267 8751<br />
institutions@natureasia.com<br />
Australia/New Zealand: Tel: 61 3 9825 1160<br />
nature@macmillan.com.au<br />
India: Tel: 91 124 2881054/55<br />
npgindia@nature.com<br />
ROW: Tel: 44 207 843 4759<br />
institutions@nature.com<br />
CUSTOMER SERVICE<br />
www.nature.com/help<br />
Senior Global Customer Service Manager: Gerald Coppin<br />
For all print and online assistance, please visit www.nature.com/help<br />
Purchase subscriptions:<br />
Americas: Nature Biotechnology, Subscription Dept., 342 Broadway, PMB 301, New York, NY 10013-<br />
3910, USA. Tel: (866) 363 7860, Fax: (212) 334 0879<br />
Europe/ROW: Nature Biotechnology, Subscription Dept., Macmillan Magazines Ltd., Brunel Road,<br />
Houndmills, Basingstoke RG21 6XS, United Kingdom. Tel: 44 1256 329 242, Fax: 44 1256 812 358<br />
Asia-Pacific: Nature Biotechnology, NPG Nature Asia-Pacific, Chiyoda Building,<br />
2-37 Ichigayatamachi, Shinjuku-ku, Tokyo 162-0843. Tel: 81 3 3267 8751, Fax: 81 3 3267 8746<br />
India: Nature Biotechnology, NPG India, 3A, 4th Floor, DLF Corporate Park, Gurgaon 122002, India.<br />
Tel: 91 124 2881054/55, Tel/Fax: 91 124 2881052<br />
REPRINTS<br />
reprints@us.nature.com<br />
Nature Biotechnology, Reprint Department, Nature Publishing Group, 75 Varick Street, Fl 9,<br />
New York, NY 10013-1917, USA.<br />
For commercial reprint orders of 600 or more, please contact:<br />
UK Reprints: Tel: 44 1256 302 923, Fax: 44 1256 321 531<br />
US Reprints: Tel: (617) 494 4900, Fax: (617) 494 4960
volume 28 number 12 DECEMBER 2010<br />
editorial<br />
1221 Vive la différence?<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Oligonucleotides generated on a<br />
microarray provide an abundant source<br />
of raw material for assembling long<br />
synthetic DNA molecules. Kosuri et al.<br />
and Matzas et al. reduce the error<br />
rate and improve the scalability<br />
of approaches that use microarray<br />
oligonucleotides to synthesize genes<br />
and other DNA elements (pp 1291 and<br />
1295). Credit: Marina Corral, based<br />
on idea courtesy of Sriram Kosuri and<br />
Mark Matzas.<br />
100<br />
80<br />
60<br />
40<br />
20<br />
0<br />
Weak<br />
Moderate<br />
Strong<br />
0 500<br />
1,000<br />
1,500<br />
Midpoint for antibody-based<br />
proteomics initiative, p 1248<br />
news<br />
1223 Will J&J turbocharge Crucell’s vaccine portfolio?<br />
1224 Amylin’s diabetes shock<br />
1225 Synthetic DNA firms embrace hazardous agents guidance but remain wary of<br />
automated ‘best-match’<br />
1226 DNA not patentable<br />
1226 USPTO’s humanitarian vouchers<br />
1227 BARDA funds vaccine makers aiming to phase out eggs<br />
1229 Ethanol blend hike promises jump start for cellulosic investment<br />
1229 Singapore injects $12.5 billion<br />
1230 Good ideas across borders<br />
1230 Airlines ahead on algae<br />
1232 News feature: Crossing the line<br />
1236 News feature: New twists on proteasome inhibitors<br />
Bioentrepreneur<br />
Building a business<br />
1239 Around the block<br />
Y Philip Zhang<br />
opinion and comment<br />
CORRESPONDENCE<br />
1242 Wrong fixes for gene patents<br />
1243 Stem cell clinics in the news<br />
1246 Tracking and assessing the rise of state-funded stem cell research<br />
1248 Towards a knowledge-based Human Protein Atlas<br />
Nature Biotechnology (ISSN 1087-0156) is published monthly by Nature Publishing Group, a trading name of Nature America Inc. located at 75 Varick Street,<br />
Fl 9, New York, NY 10013-1917. Periodicals postage paid at New York, NY and additional mailing post offices. Editorial Office: 75 Varick Street, Fl 9, New York,<br />
NY 10013-1917. Tel: (212) 726 9335, Fax: (212) 696 9753. Annual subscription rates: USA/Canada: US$250 (personal), US$3,520 (institution), US$4,050<br />
(corporate institution). Canada add 5% GST #104911595RT001; Euro-zone: €202 (personal), €2,795 (institution), €3,488 (corporate institution); Rest of world<br />
(excluding China, Japan, Korea): £130 (personal), £1,806 (institution), £2,250 (corporate institution); Japan: Contact NPG Nature Asia-Pacific, Chiyoda Building,<br />
2-37 Ichigayatamachi, Shinjuku-ku, Tokyo 162-0843. Tel: 81 (03) 3267 8751, Fax: 81 (03) 3267 8746. POSTMASTER: Send address changes to Nature<br />
Biotechnology, Subscriptions Department, 342 Broadway, PMB 301, New York, NY 10013-3910. Authorization to photocopy material for internal or personal<br />
use, or internal or personal use of specific clients, is granted by Nature Publishing Group to libraries and others registered with the Copyright Clearance Center<br />
(CCC) Transactional Reporting Service, provided the relevant copyright fee is paid direct to CCC, 222 Rosewood Drive, Danvers, MA 01923, USA. Identification<br />
code for Nature Biotechnology: 1087-0156/04. Back issues: US$45, Canada add 7% for GST. CPC PUB AGREEMENT #40032744. Printed by Publishers<br />
Press, Inc., Lebanon Junction, KY, USA. Copyright © 2010 Nature America, Inc. All rights reserved. Printed in USA.<br />
i
volume 28 number 12 DECEMBER 2010<br />
COMMENTARY<br />
1251 Anchoring gene patent eligibility to its constitutional mooring<br />
Kenneth G Chahine<br />
1256 The environmental impact subterfuge<br />
Gregory Conko & Henry I Miller<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Curbing pink bollworm resistance<br />
to Bt, p 1273<br />
Simulating metabolic interactions<br />
between cells, p 1279<br />
feature<br />
1259 To selectivity and beyond<br />
George S Mack<br />
patents<br />
1267 On firm ground: IP protection of therapeutic nanoparticles<br />
Paul Burgess, Peter Barton Hutt, Omid C Farokhzad, Robert Langer, Scott Minick &<br />
Stephen Zale<br />
1271 Recent patent applications in pharmacogenomics<br />
NEWS AND VIEWS<br />
1272 Megabases for kilodollars<br />
Mikkel Algire, Radha Krishnakumar & Chuck Merryman<br />
see also pp 1291 and 1295<br />
1273 No refuge for insect pests<br />
Kongming Wu see also p 1304<br />
1275 Nanoparticles in the lung<br />
Wolfgang G Kreyling, Stephanie Hirn & Carsten Schleh see also p 1300<br />
1277 Research highlights<br />
computational biology<br />
Analysis<br />
1279 Large-scale in silico modeling of metabolic interactions between cell types in the<br />
human brain<br />
Nathan E Lewis, Gunnar Schramm, Aarash Bordbar, Jan Schellenberger,<br />
Michael P Andersen, Jeffrey K Cheng, Nilam Patel, Alex Yee, Randall A Lewis,<br />
Roland Eils, Rainer König & Bernhard Ø Palsson<br />
research<br />
BRIEF COMMUNICATIONS<br />
1287 A robust system for production of minicircle DNA vectors<br />
Mark A Kay, Cheng-Yi He & Zhi-Ying Chen<br />
Combining DNA reading and writing,<br />
p 1291<br />
letters<br />
1291 High-fidelity gene synthesis by retrieval of sequence-verified DNA identified using<br />
high-throughput pyrosequencing<br />
Mark Matzas, Peer F Stähler, Nathalie Kefer, Nicole Siebelt, Valesca Boisguérin,<br />
Jack T Leonard, Andreas Keller, Cord F Stähler, Pamela Häberle,<br />
Baback Gharizadeh, Farbod Babrzadeh & George M Church see also p 1272<br />
nature biotechnology<br />
iii
volume 28 number 12 DECEMBER 2010<br />
1295 Scalable gene synthesis by selective amplification of DNA pools from high-fidelity<br />
microchips<br />
Sriram Kosuri, Nikolai Eroshenko, Emily M LeProust, Michael Super, Jeffrey Way,<br />
Jin Billy Li & George M Church see also p 1272<br />
1300 Rapid translocation of nanoparticles from the lung airspaces to the body<br />
Hak Soo Choi, Yoshitomo Ashitate, Jeong Heon Lee, Soon Hee Kim, Aya Matsui,<br />
Numpon Insin, Moungi G Bawendi, Manuela Semmler-Behnke, John V Frangioni &<br />
Akira Tsuda see also p 1275<br />
Scalable DNA synthesis, p 1295<br />
1304 Suppressing resistance to Bt cotton with sterile insect releases<br />
Bruce E Tabashnik, Mark S Sisterson, Peter C Ellsworth, Timothy J Dennehy,<br />
Larry Antilla, Leighton Liesner, Mike Whitlow, Robert T Staten, Jeffrey A Fabrick,<br />
Gopalan C Unnithan, Alex J Yelich, Christa Ellers-Kirk, Virginia S Harpold,<br />
Xianchun Li & Yves Carrière see also p 1273<br />
1308 errata and corrigenda<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Nanoparticle behavior in the lungs,<br />
p 1300<br />
careers and recruitment<br />
1309 Catching the wave in China<br />
Connie Johnson Hambley<br />
1312 people<br />
ADVERTISEMENT<br />
Modern evolution of the ideal science park<br />
Since their inception, science parks have become much more than<br />
shelter for technical industries. Science parks are now a feature of the<br />
global landscape and their numbers are still growing in the 21 st century.<br />
The report takes a look at the policies and dynamics that help to shape<br />
the modern science parks, especially those policies that facilitate growth<br />
of the biotech industry. The report follows Errata and Corrigenda after<br />
page 1308.<br />
nature biotechnology<br />
v
Editorial<br />
Vive la différence?<br />
The Obama administration’s $1 billion tax program at the very least signals a continued commitment to innovative<br />
biotech. The same cannot be said of plans afoot by the French government.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
At the beginning of November, the US Therapeutic Discovery<br />
Project Program (TDPP) handed out a billion dollars in tax credits/grants<br />
to small- to medium-sized (SME) companies in the life<br />
sciences. The program was overseen by the federal tax authorities<br />
and was a direct result of input by BIO, the Biotechnology Industry<br />
Organization. Although the amounts handed out to each company<br />
were rather small, the Obama administration should be congratulated<br />
on providing decisive support to the US biotech sector. But perhaps<br />
most important, this latest initiative continues decades of fiscal policy<br />
under different US administrations that has consistently supported the<br />
entrepreneurial life science sector.<br />
There is little doubt that the US biotech sector (and that in most<br />
other countries) is in need of a shot in the arm. BIO estimates that<br />
since 2007, more than 100 public biotech companies have closed their<br />
doors and countless more private companies have ceased operations.<br />
The percentage of companies with less than one year’s cash remains<br />
25% (down only slightly from 29% in 2007).<br />
TDPP was a one-off program, approved as part of the 2010 Patient<br />
Protection and Affordable Care Act, designed specifically for companies<br />
with fewer than 250 employees. The credits cover up to 50% of<br />
eligible R&D expenses in tax years 2009 or 2010. Applications were<br />
accepted from companies for one month from June 21 and the submissions<br />
reviewed by the US National Institutes of Health (NIH).<br />
The program was a roaring success—almost too successful. The US<br />
Internal Revenue Service had expected ~1,500 applications; it was deluged<br />
with over 5,600. To cope with this onslaught, a decision was made<br />
to turn the $1 billion cake into 4,606 crumbs, each worth a maximum<br />
of $244,479. Although some companies applied for and were granted<br />
multiple project awards (28% of total), 2,923 organizations in 47 states<br />
got a piece of the action.<br />
Project reviewers for TDPP were asked to assess, among a very few<br />
other criteria, whether a project was ‘likely’ to result in new therapies,<br />
or reduce long-term healthcare costs or contribute to the goal of curing<br />
cancer. The fact that 4,000 projects got the thumbs-up on this<br />
criterion at the very least severely stretches the meaning of ‘likely’. The<br />
best that can realistically be claimed is that it is likely that a handful<br />
of the projects will succeed, and that the chances of success in every<br />
project will be marginally enhanced by an extra couple of hundred<br />
thousand dollars.<br />
That said, by casting a wide net, the US authorities have ensured that<br />
their biotech sector retains more shots on the goal—a wise investment<br />
in the future.<br />
The same cannot be said for policy proposals under discussion in<br />
France.<br />
Back in September, the government of Nicolas Sarkozy introduced<br />
its 2011 finance bill in the French parliament. Amongst the cost-cutting<br />
measures and edicts designed to cut back on government spending<br />
was a small section, Article 78, which will result in an estimated annual<br />
saving to the French exchequer of $57 million.<br />
Article 78 alters radically the conditions attached to the Jeune<br />
Enterprise Innovante (JEI; Young Innovative Company) status in<br />
France. Currently, any research-based independent SME under 8<br />
years old is, in effect, excused from paying social taxes for any of its<br />
researcher employees. Because the social taxes represent 30–40% of<br />
salary costs, this represents a significant saving and an incentive to<br />
life science venture investment in the country (Nat. Biotechnol. 23,<br />
1187, 2005). Article 78 proposes to cap the savings at €103,680 per<br />
year per company and to tail off the benefit between years 4 and 8 of<br />
a company’s lifetime.<br />
A vote in the French Senate on the 2011 Finance Bill is expected<br />
early this month, with implementation in January. Meanwhile, the<br />
French industry association, France Biotech is furiously lobbying<br />
ministers, senators and senior figures in the civil administration in<br />
a desperate bid to axe Article 78. They fear, with justification, that it<br />
will discourage entrepreneurs and investors, and worse, that it will be<br />
a disincentive to innovate.<br />
What is most disingenuous about the French policy shift, though,<br />
is that the government is saying the benefit will actually provide<br />
SMEs with a ‘softer landing’ before returning to full taxation. In other<br />
words, they are claiming to understand the long-term effects of this<br />
volte face in policy. And when it comes to stargazing, the French government<br />
doesn’t have a good track record in biotech. Twenty years<br />
ago, the ‘BioAvenir’ initiative invested a billion francs in academic<br />
life science projects, but elected to make one (French) company—<br />
Rhone-Poulenc—the sole beneficiary of any commercialization. We<br />
don’t know what happened to the hundreds of BioAvenir projects, but<br />
Rhone Poulenc didn’t turn into the biotech driver the government<br />
envisaged—it is now known as Sanofi-aventis.<br />
As the BioAvenir experience illustrates, if there is one thing that<br />
can be guaranteed about the future, it is that it will not look like the<br />
present. That is something the US government seems to have grasped<br />
with TDPP, arguably to a fault: its $1 billion has been flung into any<br />
company that is both innovative and not established, into any company<br />
that might be the future simply by virtue of its not being a big part of<br />
the present.<br />
The French government, conversely, still clings to the idea that it can<br />
extrapolate the future from the present: each year, it gives an estimated<br />
€1 billion in R&D tax credits to Sanofi-aventis. There are worse places<br />
to spend a €1 billion, no doubt, but providing such largesse to one<br />
of the world’s biggest pharmaceutical companies while begrudging<br />
a fraction ($57 million) of it to innovative biotech doesn’t look like a<br />
government investing in the future.<br />
nature biotechnology volume 28 number 12 december 2010 1221
in this section<br />
Screening for<br />
malicious use of<br />
synthetic DNA<br />
p1225<br />
BARDA funds<br />
boost creative<br />
vaccine makers<br />
p1227<br />
news<br />
Ethanol blend<br />
for cars rises after<br />
35 years p1229<br />
Will J&J turbocharge Crucell’s vaccine portfolio?<br />
© 2010 Nature America, Inc. All rights reserved.<br />
The emergence of contamination at Crucell’s<br />
production facility in Shingal, Yongin City,<br />
Korea, came at an awkward time. The vaccine<br />
maker, based in Leiden, The Netherlands,<br />
disclosed the news in late October as Johnson<br />
& Johnson (J&J) prepared to issue an update<br />
on progress of its $2.3 billion bid for outright<br />
ownership of the company. Nevertheless, J&J,<br />
of New Brunswick, New Jersey, intends to<br />
proceed according to plan. The large pharma’s<br />
tender offer for Crucell’s shares is due to<br />
be considered at a meeting of Crucell shareholders<br />
in Amsterdam on 10 December. The<br />
implications—if any—that the manufacturing<br />
disruption may have for the transaction are<br />
not clear at this point. “I don’t expect this to be<br />
a deal breaker, but there’s no guarantee it won’t<br />
be a deal adjuster,” says Jan Van den Bossche,<br />
analyst at Petercam in Brussels. “It will all<br />
depend on what’s going on out there.”<br />
Although shipments of the two products<br />
affected by the quality failures—the pediatric<br />
liquid pentavalent combination vaccine<br />
Quinvaxem (diphtheria (Corynebacterium<br />
diphtheriae), toxoid/tetanus (Clostridium<br />
tetani), toxoid/whole cell Bordetella pertussis<br />
(DTwP); Haemophilus influenzae type B (Hib)<br />
conjugate vaccine; and recombinant hepatitis<br />
B virus surface antigen (HBsAg) vaccine;<br />
trade name, Hepavax-Gene)—were, as Nature<br />
Biotechnology went to press, expected to<br />
resume by late November, full production will<br />
not be restored until February. Quinvaxem,<br />
which protects against five pediatric infections,<br />
is the company’s most important product, having<br />
secured tenders worth $910 million since<br />
the end of 2006, including a $110 million order<br />
from the New York–based United Nations<br />
Children’s Fund in May. Crucell has made<br />
provision for a €22.8 million ($31.8 million)<br />
charge on the inventory lost through the disruption.<br />
Even before the problems emerged,<br />
the company had been in the process of transferring<br />
production of the two vaccines to a new<br />
€50 million ($69.8 million) facility in Incheon,<br />
South Korea, which is scheduled to become<br />
fully operational next year.<br />
Crucell’s problems echo those of Shantha<br />
Biotechnics, of Hyderabad, India. Not<br />
long after Lyon-based Sanofi Pasteur<br />
The Dutch vaccine manufacturer Crucell (headquarters in Leiden pictured above) will run as an<br />
autonomous unit following J&J’s expected acquisition.<br />
gained an 80% stake in Shantha, the latter<br />
firm’s Shan5 (DTwP-HBsAg-Hib) vaccine,<br />
a direct competitor to Quinvaxem,<br />
lost its accreditation on the World Health<br />
Organization’s (Geneva) list of recommended<br />
vaccines because of the appearance of white<br />
sediment on vaccine vials. “They’ve had<br />
massive manufacturing setbacks,” says Peter<br />
Welford, analyst at Jefferies International, in<br />
London. “Crucell [has] had the least problems.”<br />
Others competing in this product<br />
segment include Panacea Biotech, of New<br />
Delhi, India, and the GSK Biologicals arm<br />
of London-based GlaxoSmithKline.<br />
J&J’s interest in the Dutch company came<br />
to the fore in September last year, when it paid<br />
around €302 ($407) million, or €20.6 ($27.8)<br />
per share, to amass an 18% stake in Crucell.<br />
The healthcare giant’s current $2.3 billion bid<br />
for the rest of the company it does not already<br />
own is well in excess of a reported $1.35 billion<br />
bid that Wyeth of Madison, New Jersey,<br />
was preparing early last year. That deal was<br />
derailed by New York–based Pfizer’s $68 billion<br />
acquisition of Wyeth. Recent contamination<br />
problems are unlikely to affect J&J’s offer,<br />
although Welford says, this will eliminate the<br />
possibility that dissenting shareholders may<br />
hold out for a higher price than the €24.75<br />
($35.18) per share that J&J has tabled.<br />
Crucell’s history can be traced back to Leidenbased<br />
IntroGene, which was formed in 1993 to<br />
commercialize the PER.C6 human cell culture<br />
technology, which can produce as much as 27 g/l<br />
of IgGs in bioreactor culture. The Crucell<br />
name came into being in 2000, after a merger<br />
with Utrecht Biotechnology Systems in The<br />
Netherlands, an antibody discovery firm that<br />
had developed a ‘subtractive’ phage display<br />
technology called MAbstract. In that platform,<br />
a phage antibody library is admixed with a<br />
heterogeneous cell mixture, where the particular<br />
cell of interest is labeled with a fluorescent<br />
tag and sorted by flow cytometry to identify<br />
binding phage antibodies, dispensing with the<br />
need for repeatedly constructing new libraries.<br />
These technologies, as well as an AdVac adenoviral<br />
vector platform, underpin the company’s<br />
development pipeline. However, its commercial<br />
product portfolio is based on recent acquisitions<br />
of Stockholm-based SBL Vaccin, and, especially,<br />
of Bern, Switzerland–based Berna Biotech.<br />
Berna had previously acquired Rhein Biotech,<br />
of Maastricht, The Netherlands, but never<br />
managed to achieve its stated growth ambitions<br />
(Table 1). Crucell, in contrast, has proven adept<br />
at winning sales, particularly in emerging markets.<br />
“They’ve done a marvelous job of turning<br />
around Berna Biotech,” Van den Bossche says.<br />
J&J plans to run Crucell as an autonomous<br />
Crucell<br />
nature biotechnology volume 28 number 12 december 2010 1223
NEWS<br />
© 2010 Nature America, Inc. All rights reserved.<br />
in brief<br />
Amylin’s<br />
diabetes<br />
shock<br />
In a major<br />
setback for<br />
Amylin, the US<br />
Food and Drug<br />
Fears over long QT. Administration<br />
called for<br />
further tests on Amylin’s Bydureon, a onceweekly<br />
drug for type 2 diabetes. On October<br />
19, the San Diego-based biotech Amylin, and<br />
partner Eli Lilly of Indianapolis, announced that<br />
the agency had issued a Complete Response<br />
Letter calling for a study known as thorough<br />
QT to further investigate the potential effect of<br />
the drug’s cardiac effects. Bydureon is a longacting<br />
formulation of Amylin’s approved drug<br />
Byetta (exenatide), an analog of the insulinboosting<br />
glucagon-like peptide 1 (GLP-1)<br />
(Nat. Biotechnol. 28, 109, 2010). Both<br />
contain exenatide, but Bydureon contains<br />
a higher concentration of the active agent,<br />
delivered into the blood by controlled release.<br />
The first-in-class synthetic gut hormone drug<br />
Byetta generated $677 million in 2009 in the<br />
US alone. But Bydureon’s formulation—once<br />
weekly injections compared with twice a day—is<br />
predicted to gain competitive edge over its<br />
predecessor if approved. Simos Simeonidis,<br />
managing director and senior biotech analyst<br />
at Rodman & Renshaw in New York, says that<br />
Bydureon had “blockbuster” potential and could<br />
have brought in at least $1 billion per year<br />
for Amylin after the first few years. The FDA’s<br />
concerns over supratherapeutic concentrations<br />
of exenatide may have been triggered by<br />
observations in a Byetta study, in which a few<br />
patients with blood levels of 500 pg/ml had<br />
lengthened QT intervals. In patients with kidney<br />
problems, once-weekly Bydureon can reach<br />
five times the normal 200 pg/ml seen with<br />
Byetta. Given the delay in approving Bydureon,<br />
alternate treatments for type 2 diabetes like<br />
Victoza (liraglutide) from Novo Nordisk, of<br />
Princeton, New Jersey, and Amylin’s own Byetta,<br />
are now expected to flourish. London-based<br />
GlaxoSmithKline’s investigational Syncria(R)<br />
(albiglutide), also a GLP-1 analog, is currently<br />
undergoing phase 3 clinical trials. “They were<br />
considerably behind,” says Yaron Werber, senior<br />
biotech analyst at Citigroup in New York. “But<br />
now I think they have a chance to close a big<br />
part of the gap.” <br />
Nidhi Subbaraman<br />
iStockphoto/Andrius Gruzdaitis<br />
in their words<br />
“At this juncture, I don’t know that individuals<br />
will have all that much to gain from publishing<br />
their data, but I think that Genomes Unzipped<br />
will help to prove that there’s not all that much<br />
to lose, either.” Linda Avey, cofounder of<br />
23andMe, comments on a move by 11 Britishbased<br />
scientists and a US lawyer to make their<br />
own genetic tests publicly available in the hope<br />
of encouraging others to share their genome<br />
information. (The Times, 11 October 2010)<br />
Table 1 Vaccine mergers & acquisitions<br />
Acquirer Target (location) Value Year<br />
Berna Biotech Rhein Biotech $257 million 2002<br />
Crucell Berna Biotech $449 million 2005<br />
Crucell SBL Vaccin $52 million 2006<br />
GlaxoSmithKline ID Biomedical (Vancouver, Canada) $1.4 billion 2005<br />
GlaxoSmithKline Corixa (Seattle) $300 million 2005<br />
Intercell Iomai (Gaithersburg, Maryland) $189 million 2008<br />
Novartis Chiron (Emeryville, California) $5.4 billion a 2005<br />
Sanofi Pasteur Acambis $549 million 2008<br />
Sanofi Pasteur Shantha Biotechnics $781 million b 2009<br />
Sanofi Pasteur VaxDesign (Orlando, Florida) $55 million c 2010<br />
a The transaction involved the remaining 58% of Chiron’s stock not already held by Novartis. b Total valuation of the company implied<br />
by the terms of the deal. Sanofi Pasteur acquired an 80% stake. c The deal includes $5 million in potential milestone payments.<br />
unit, just as it has with several earlier acquisitions,<br />
such as antiviral drug developer Tibotec<br />
of Antwerp, Belgium, and antibody developer<br />
Centocor located in Horsham, Pennsylvania.<br />
J&J’s imminent acquisition of Crucell is further<br />
evidence that the vaccines area, although<br />
small in the context of overall pharma sales, is<br />
firmly back on the industry’s agenda, after a<br />
couple of decades during which it had fallen out<br />
of favor. Other acquisitions where vaccines have<br />
featured prominently (though not exclusively)<br />
are Pfizer’s takeover of Wyeth and Londonbased<br />
AstraZeneca’s $15.2 billion acquisition of<br />
Gaithersburg, Maryland–based MedImmune.<br />
The vaccines market’s steady growth equilibrium<br />
is punctuated by irregular spurts<br />
caused by significant new product introductions.<br />
For instance, Wyeth’s Prevnar (pneumococcal<br />
7-valent conjugate) or Gardasil (human<br />
papillomavirus quadrivalent (types 6, 11, 16<br />
and 18) recombinant vaccine) from Merck<br />
of Whitehouse Station, New Jersey, quickly<br />
added substantial chunks of revenue to the<br />
industry total.<br />
The complexities of manufacturing, combined<br />
with the economies of large-scale production,<br />
have conspired to make the sector<br />
strongly oligopolistic, with the market dominated<br />
by a small number of firms. The top<br />
five players, Sanofi Pasteur, GSK Biologicals,<br />
Merck, Pfizer and Novartis, control around<br />
85% of the market, with around $18 billion<br />
of sales in 2009, according to market analysts<br />
VacZine Analytics, of Bishop’s Stortford,<br />
UK. “The industry last year grew by about<br />
9, 10% over the previous year,” says VacZine<br />
Analytics director John Savopoulos. By his<br />
reckoning Crucell’s 2009 vaccine sales of €304<br />
million ($424.3 million) would rank the company<br />
in ninth place, behind CSL, of Parkville,<br />
Australia, and London-based AstraZeneca’s<br />
MedImmune unit. The H1N1 influenza pandemic<br />
had the biggest impact on sales in 2009,<br />
bringing in around $3.5 billion in additional<br />
revenue. In the first three-quarters of 2010,<br />
the top five companies posted combined sales<br />
of $16.9 billion, excluding the contribution of<br />
Sanofi Pasteur MSD, a European joint venture<br />
between Sanofi Pasteur and Merck.<br />
Nevertheless, the industry’s pipeline remains<br />
widely distributed. “Seventy companies are now<br />
targeting 40 plus pathogens,” Savopoulos says.<br />
Some 160 vaccines are in clinical development,<br />
around 100 of which are in phase 1 trials. Crucell<br />
has the industry’s fifth largest pipeline, he notes.<br />
It has clinical-stage programs in yellow fever,<br />
tuberculosis (TB), malaria, HIV, Ebola virus<br />
and Marburg virus as well as preclinical efforts<br />
in seasonal influenza, respiratory syncytial<br />
virus and in the development of a universal flu<br />
vaccine. Only GSK Biologicals, Sanofi Pasteur,<br />
Novartis and Merck have more vaccine programs<br />
in the clinic. “They look reasonably well<br />
positioned. The problem is when you probability-adjust<br />
their pipeline—malaria, TB—they’re<br />
tough areas to be in,” Savopoulos says. A preclinical<br />
antibody development program, which<br />
targets both seasonal and pandemic flu strains,<br />
and has attracted $40.7 million in US National<br />
Institutes’ of Health funding (with potentially<br />
$28.4 million more to come), was the main<br />
focus of J&J’s 2009 alliance with Crucell. “That<br />
technology, if it can be harnessed for an active<br />
vaccine, that’s a quantum leap above the existing<br />
flu market,” Savopoulos points out. Jan Van den<br />
Bossche identifies a combination monoclonal<br />
antibody development program for treating<br />
individuals exposed to rabies infection as particularly<br />
promising. “It’s a clear, understandable<br />
program,” he says. The program, in phase 2 trials,<br />
aims to replace an existing equine immunoglobulin<br />
therapy.<br />
Although J&J will seek further growth from<br />
Crucell’s existing product portfolio, J&J’s real<br />
measure of success will be in the extent to<br />
which it can convert pipeline promise into<br />
commercial reality. Sanofi Pasteur ‘turbocharged’<br />
Acambis’s ChimeriVax platform<br />
when it acquired the Cambridge, UK–based<br />
firm, says Savopoulos. “The question is, can<br />
J&J do the same thing?”<br />
Cormac Sheridan, Dublin<br />
1224 volume 28 number 12 december 2010 nature biotechnology
NEWS<br />
© 2010 Nature America, Inc. All rights reserved.<br />
in brief<br />
Amylin’s<br />
diabetes<br />
shock<br />
In a major<br />
setback for<br />
Amylin, the US<br />
Food and Drug<br />
Fears over long QT. Administration<br />
called for<br />
further tests on Amylin’s Bydureon, a onceweekly<br />
drug for type 2 diabetes. On October<br />
19, the San Diego-based biotech Amylin, and<br />
partner Eli Lilly of Indianapolis, announced that<br />
the agency had issued a Complete Response<br />
Letter calling for a study known as thorough<br />
QT to further investigate the potential effect of<br />
the drug’s cardiac effects. Bydureon is a longacting<br />
formulation of Amylin’s approved drug<br />
Byetta (exenatide), an analog of the insulinboosting<br />
glucagon-like peptide 1 (GLP-1)<br />
(Nat. Biotechnol. 28, 109, 2010). Both<br />
contain exenatide, but Bydureon contains<br />
a higher concentration of the active agent,<br />
delivered into the blood by controlled release.<br />
The first-in-class synthetic gut hormone drug<br />
Byetta generated $677 million in 2009 in the<br />
US alone. But Bydureon’s formulation—once<br />
weekly injections compared with twice a day—is<br />
predicted to gain competitive edge over its<br />
predecessor if approved. Simos Simeonidis,<br />
managing director and senior biotech analyst<br />
at Rodman & Renshaw in New York, says that<br />
Bydureon had “blockbuster” potential and could<br />
have brought in at least $1 billion per year<br />
for Amylin after the first few years. The FDA’s<br />
concerns over supratherapeutic concentrations<br />
of exenatide may have been triggered by<br />
observations in a Byetta study, in which a few<br />
patients with blood levels of 500 pg/ml had<br />
lengthened QT intervals. In patients with kidney<br />
problems, once-weekly Bydureon can reach<br />
five times the normal 200 pg/ml seen with<br />
Byetta. Given the delay in approving Bydureon,<br />
alternate treatments for type 2 diabetes like<br />
Victoza (liraglutide) from Novo Nordisk, of<br />
Princeton, New Jersey, and Amylin’s own Byetta,<br />
are now expected to flourish. London-based<br />
GlaxoSmithKline’s investigational Syncria(R)<br />
(albiglutide), also a GLP-1 analog, is currently<br />
undergoing phase 3 clinical trials. “They were<br />
considerably behind,” says Yaron Werber, senior<br />
biotech analyst at Citigroup in New York. “But<br />
now I think they have a chance to close a big<br />
part of the gap.” <br />
Nidhi Subbaraman<br />
iStockphoto/Andrius Gruzdaitis<br />
in their words<br />
“At this juncture, I don’t know that individuals<br />
will have all that much to gain from publishing<br />
their data, but I think that Genomes Unzipped<br />
will help to prove that there’s not all that much<br />
to lose, either.” Linda Avey, cofounder of<br />
23andMe, comments on a move by 11 Britishbased<br />
scientists and a US lawyer to make their<br />
own genetic tests publicly available in the hope<br />
of encouraging others to share their genome<br />
information. (The Times, 11 October 2010)<br />
Table 1 Vaccine mergers & acquisitions<br />
Acquirer Target (location) Value Year<br />
Berna Biotech Rhein Biotech $257 million 2002<br />
Crucell Berna Biotech $449 million 2005<br />
Crucell SBL Vaccin $52 million 2006<br />
GlaxoSmithKline ID Biomedical (Vancouver, Canada) $1.4 billion 2005<br />
GlaxoSmithKline Corixa (Seattle) $300 million 2005<br />
Intercell Iomai (Gaithersburg, Maryland) $189 million 2008<br />
Novartis Chiron (Emeryville, California) $5.4 billion a 2005<br />
Sanofi Pasteur Acambis $549 million 2008<br />
Sanofi Pasteur Shantha Biotechnics $781 million b 2009<br />
Sanofi Pasteur VaxDesign (Orlando, Florida) $55 million c 2010<br />
a The transaction involved the remaining 58% of Chiron’s stock not already held by Novartis. b Total valuation of the company implied<br />
by the terms of the deal. Sanofi Pasteur acquired an 80% stake. c The deal includes $5 million in potential milestone payments.<br />
unit, just as it has with several earlier acquisitions,<br />
such as antiviral drug developer Tibotec<br />
of Antwerp, Belgium, and antibody developer<br />
Centocor located in Horsham, Pennsylvania.<br />
J&J’s imminent acquisition of Crucell is further<br />
evidence that the vaccines area, although<br />
small in the context of overall pharma sales, is<br />
firmly back on the industry’s agenda, after a<br />
couple of decades during which it had fallen out<br />
of favor. Other acquisitions where vaccines have<br />
featured prominently (though not exclusively)<br />
are Pfizer’s takeover of Wyeth and Londonbased<br />
AstraZeneca’s $15.2 billion acquisition of<br />
Gaithersburg, Maryland–based MedImmune.<br />
The vaccines market’s steady growth equilibrium<br />
is punctuated by irregular spurts<br />
caused by significant new product introductions.<br />
For instance, Wyeth’s Prevnar (pneumococcal<br />
7-valent conjugate) or Gardasil (human<br />
papillomavirus quadrivalent (types 6, 11, 16<br />
and 18) recombinant vaccine) from Merck<br />
of Whitehouse Station, New Jersey, quickly<br />
added substantial chunks of revenue to the<br />
industry total.<br />
The complexities of manufacturing, combined<br />
with the economies of large-scale production,<br />
have conspired to make the sector<br />
strongly oligopolistic, with the market dominated<br />
by a small number of firms. The top<br />
five players, Sanofi Pasteur, GSK Biologicals,<br />
Merck, Pfizer and Novartis, control around<br />
85% of the market, with around $18 billion<br />
of sales in 2009, according to market analysts<br />
VacZine Analytics, of Bishop’s Stortford,<br />
UK. “The industry last year grew by about<br />
9, 10% over the previous year,” says VacZine<br />
Analytics director John Savopoulos. By his<br />
reckoning Crucell’s 2009 vaccine sales of €304<br />
million ($424.3 million) would rank the company<br />
in ninth place, behind CSL, of Parkville,<br />
Australia, and London-based AstraZeneca’s<br />
MedImmune unit. The H1N1 influenza pandemic<br />
had the biggest impact on sales in 2009,<br />
bringing in around $3.5 billion in additional<br />
revenue. In the first three-quarters of 2010,<br />
the top five companies posted combined sales<br />
of $16.9 billion, excluding the contribution of<br />
Sanofi Pasteur MSD, a European joint venture<br />
between Sanofi Pasteur and Merck.<br />
Nevertheless, the industry’s pipeline remains<br />
widely distributed. “Seventy companies are now<br />
targeting 40 plus pathogens,” Savopoulos says.<br />
Some 160 vaccines are in clinical development,<br />
around 100 of which are in phase 1 trials. Crucell<br />
has the industry’s fifth largest pipeline, he notes.<br />
It has clinical-stage programs in yellow fever,<br />
tuberculosis (TB), malaria, HIV, Ebola virus<br />
and Marburg virus as well as preclinical efforts<br />
in seasonal influenza, respiratory syncytial<br />
virus and in the development of a universal flu<br />
vaccine. Only GSK Biologicals, Sanofi Pasteur,<br />
Novartis and Merck have more vaccine programs<br />
in the clinic. “They look reasonably well<br />
positioned. The problem is when you probability-adjust<br />
their pipeline—malaria, TB—they’re<br />
tough areas to be in,” Savopoulos says. A preclinical<br />
antibody development program, which<br />
targets both seasonal and pandemic flu strains,<br />
and has attracted $40.7 million in US National<br />
Institutes’ of Health funding (with potentially<br />
$28.4 million more to come), was the main<br />
focus of J&J’s 2009 alliance with Crucell. “That<br />
technology, if it can be harnessed for an active<br />
vaccine, that’s a quantum leap above the existing<br />
flu market,” Savopoulos points out. Jan Van den<br />
Bossche identifies a combination monoclonal<br />
antibody development program for treating<br />
individuals exposed to rabies infection as particularly<br />
promising. “It’s a clear, understandable<br />
program,” he says. The program, in phase 2 trials,<br />
aims to replace an existing equine immunoglobulin<br />
therapy.<br />
Although J&J will seek further growth from<br />
Crucell’s existing product portfolio, J&J’s real<br />
measure of success will be in the extent to<br />
which it can convert pipeline promise into<br />
commercial reality. Sanofi Pasteur ‘turbocharged’<br />
Acambis’s ChimeriVax platform<br />
when it acquired the Cambridge, UK–based<br />
firm, says Savopoulos. “The question is, can<br />
J&J do the same thing?”<br />
Cormac Sheridan, Dublin<br />
1224 volume 28 number 12 december 2010 nature biotechnology
Synthetic DNA firms embrace hazardous<br />
agents guidance but remain wary of<br />
automated ‘best-match’<br />
news<br />
© 2010 Nature America, Inc. All rights reserved.<br />
After more than four years of public and<br />
private discussion and review, the US government<br />
has officially issued its Screening<br />
Framework Guidance for Providers of Synthetic<br />
Double-Stranded DNA (http://www.phe.<br />
gov/Preparedness/legal/guidance/syndna/<br />
Documents/syndna-guidance.pdf), a set of<br />
voluntary guidelines intended to help gene<br />
synthesis companies intercept unauthorized<br />
purchases of genetic components from human<br />
or agricultural pathogens. Many are relieved that<br />
standardized guidelines have finally been established,<br />
enabling the industry to harmonize practices<br />
and provide reassurance to large corporate<br />
clients that represent their bread and butter.<br />
“Overall it’s not a bad framework, and I<br />
think it’s been designed with a lot of expertise,”<br />
says Markus Fischer, director and cofounder<br />
of Entelechon in Regensburg, Germany, a gene<br />
synthesis provider, also part of the International<br />
Association Synthetic Biology (IASB), one of<br />
two major industry groups representing gene<br />
synthesis companies. In the absence of clear government<br />
guidance, the IASB and its counterpart,<br />
the International Gene Synthesis Consortium<br />
(IGSC), each developed their own ‘best practices’<br />
for screening both DNA orders and the customers<br />
that place them. To draw up its official guideline,<br />
the US Department of Health and Human<br />
Services (HHS), in Washington, DC, received<br />
comments from 22 organizations and individuals<br />
since publishing its draft Guidance in November<br />
2009 (Table 1). The final version—released in<br />
October—includes only a few notable changes,<br />
such as the elimination of a size cut-off for screening<br />
decisions on double-stranded segments.<br />
Although the Guidance structurally resembles<br />
preexisting protocols, critics such as Stephen<br />
Maurer of the University of California at Berkeley<br />
are concerned that its recommendations are<br />
weaker than what is needed and may encourage<br />
companies to cut corners in the future. “Industry<br />
had embraced a higher standard, and now the<br />
government is going to lead us to a lower standard,”<br />
he says. Chief among his concerns is the<br />
proposed mechanism for screening sequences.<br />
The government proposes a ‘best match’<br />
strategy, in which orders are compared against<br />
GenBank in 200-bp segments, based on both<br />
nucleotide and all six possible peptide sequences.<br />
If the top ‘hit’ is from a pathogen on the government’s<br />
list of select agents and toxins (http://www.<br />
selectagents.gov/Select%20Agents%20and%20<br />
Toxins%20List.html) or, for international orders,<br />
the ‘Commerce Control List’ (http://www.gpo.<br />
The select agents list omits many known human<br />
pathogens, such as the SARS coronavirus.<br />
gov/bis/ear/pdf/ccl1.pdf), it should be considered<br />
a ‘sequence of concern’ for further expert<br />
analysis, in conjunction with a careful assessment<br />
of the ordering customer’s credentials.<br />
Maurer suggests that by not expressly calling<br />
for human review of database matches—<br />
regardless of whether or not they are on the<br />
select agents list—this strategy is inherently<br />
less effective than the ‘top homology’ method<br />
already in use at several companies, including<br />
Entelechon, in which all GenBank results are<br />
manually assessed. “We mandate that one of<br />
our employees reviews the complete list of hits,<br />
and not just the ones that have been automatically<br />
flagged,” says Fischer. “A fully automated<br />
screening system leaves significant biosecurity<br />
questions unanswered.”<br />
According to Theresa Lawrence, a senior<br />
science advisor with HHS, top homology was<br />
rejected in the interest of applying a consistent<br />
standard for distinguishing potential threats based<br />
on analysis of an established data source. “There<br />
was concern with the top homology approach that<br />
we would have to designate an arbitrary threshold,”<br />
she says, “and this approach needs human<br />
screeners, which can represent an inconsistent<br />
mechanism from provider to provider.”<br />
Although GenBank represents a rich resource<br />
for genetic data and is therefore a powerful<br />
foundation for such screens, it is nevertheless<br />
a product of community curation and potential<br />
‘sequences of concern’ may be inconsistently<br />
designated. “GenBank is just a repository,” says<br />
Sean Eddy, a computational biologist at the<br />
Howard Hughes Medical Institute’s Janelia Farm<br />
Research Campus in Loudon County, Virginia.<br />
CMSP<br />
nature biotechnology volume 28 number 12 december 2010 1225
NEWS<br />
© 2010 Nature America, Inc. All rights reserved.<br />
in brief<br />
DNA not patentable<br />
In a spine-chilling announcement, the<br />
US Department of Justice in October said<br />
unmodified human DNA should not be eligible<br />
for patent. The department’s stance conflicts<br />
with the body of case law on the matter and a<br />
longstanding position held by the US Patent<br />
and Trademark Office, which has issued<br />
more than 10,000 of these patents. The<br />
Justice Department announced its position<br />
in response to a lawsuit involving patents on<br />
breast cancer genes BRCA1 and BRCA2. A US<br />
district court in March declared the patents<br />
invalid, saying that the genes are products of<br />
nature rather than human-made inventions.<br />
Patent holders University of Utah and Myriad<br />
Genetics, based in Salt Lake City, appealed<br />
in June. In an amicus brief filed with the<br />
Federal Circuit Court of Appeals the Justice<br />
Department agreed that identifying and<br />
isolating DNA without further manipulation<br />
is not an invention, or patent eligible. How<br />
the agency’s declaration will influence<br />
justices and the patent office worries biotech<br />
companies. But the patent office isn’t easily<br />
swayed, says Thomas Kowalski, an attorney<br />
with Vedder Price in New York. “The patent<br />
office is not going to change what it’s doing in<br />
view of what the Department of Justice says,”<br />
he says. Besides, international agreements<br />
between the American, European and<br />
Japanese patent offices, known as Trilateral<br />
Co-operation, have concluded that unmodified<br />
DNA is patentable.<br />
Emily Waltz<br />
USPTO’s do-good vouchers<br />
Patent owners may soon be able to cut the<br />
time needed to have a patent reexamined by<br />
up to two-thirds if they can demonstrate a<br />
humanitarian use. The US Patent & Trademark<br />
Office (USPTO) is proposing a system that would<br />
offer fast-track reexamination vouchers as an<br />
incentive to stimulate “creation or licensing<br />
that addresses humanitarian needs.” Because<br />
patents under reexamination are often the most<br />
valuable commercially, a fast-track procedure<br />
would let patent owners “more readily and less<br />
expensively affirm the validity of their patents,”<br />
according to the Federal Register notice. The<br />
USPTO currently takes 19 to 20 months for<br />
such reexaminations, whereas the expedited<br />
review promises a six-month turnaround. The<br />
system is modeled on the US Food & Drug<br />
Administration’s priority review vouchers given<br />
to entities that develop drugs to treat neglected<br />
tropical diseases. In this case, patent holders<br />
who receive the fast-track reexamination voucher<br />
could use it on any other patent they own or<br />
transfer it to the open market. Although the<br />
intent is worthy, there are too many unanswered<br />
questions, worries Thomas Kowalski of Vedder<br />
Price. “What will the USPTO do to ensure that<br />
those in the developing world as well as the<br />
poor in the developed world can gain access<br />
to the technology? Also, the voucher should<br />
be tied specifically to the technology with the<br />
humanitarian use instead of being independent<br />
and transferable.” Michael Francisco<br />
Table 1 Timeline of events leading up to the synthetic DNA guidance<br />
December 2006 The US government’s National Science Advisory Board for Biosecurity issues a report<br />
recommending the establishment of “uniform and standardized screening practices<br />
among providers of synthetic DNA.”<br />
June 2007 The US Government convenes an interagency working group on synthetic nucleic acid<br />
screening.<br />
November 3, 2009 Companies of the International Association Synthetic Biology (Entelechon,<br />
ATG:biosynthetics, Biomax, febit and Sloning Biotechnology) present their “Code of<br />
Conduct” for the screening of gene orders and customers.<br />
November 19, 2009 Companies of the International Gene Synthesis Consortium (Blue Heron Biotechnology,<br />
DNA2.0, GENEART, GenScript and Integrated DNA Technologies) release a similar set of<br />
‘best practices’ guidelines, the “Harmonized Screening Protocol.”<br />
November 29, 2009 The US Department of Health and Human Services releases a draft version of its<br />
Screening Framework Guidance for Providers of Synthetic Double-Stranded DNA for<br />
public comment.<br />
January 11, 2010 The American Association for the Advancement of Science (AAAS) hosts a meeting for policy<br />
specialists and scientists from academia and industry on “Minimizing the Risks of Synthetic<br />
DNA: Scientists’ Views on the U.S. Government’s Guidance on Synthetic Genomics.”<br />
January 26, 2010 Public comment period ends on the draft guidance.<br />
October 13, 2010 HHS publishes its final version of the Screening Framework Guidance.<br />
He adds, “The annotation is as provided by the<br />
person that deposited the sequence.” Screening<br />
effectiveness could also be constrained by<br />
biases in the database contents, according to<br />
James Diggans, a researcher at MITRE, a notfor-profit<br />
national technology resource that<br />
focuses on security issues, located in Bedford,<br />
Massachusetts and McLean, Virginia. “There are<br />
far more harmless sequences in these databases<br />
than there are sequences that could be used to<br />
harm human health.”<br />
For other scientists, the reliance on the select<br />
agents list is also problematic. Eighty-two items<br />
currently listed represent known risks to human,<br />
plant or animal health, and are unambiguously<br />
regulated by federal law. But many known<br />
human pathogens, such as severe acute respiratory<br />
syndrome virus, are omitted, and others<br />
worry about the problems that could be posed<br />
by the yet-unknown sequence variants. “If you<br />
synthesize a genome without creating the actual<br />
organism it encodes—and where now you aren’t<br />
even limited to the variability found in nature—<br />
how do you taxonomically classify that genome<br />
sequence?” says Eddy.<br />
Eddy and other scientists recently partnered<br />
with the US National Research Council in an<br />
effort to bring some clarity to the characterization<br />
of high-risk genes. The resulting report,<br />
Sequence-Based Classification of Select Agents:<br />
A Brighter Line (http://www.nap.edu/catalog/<br />
12970.html), concludes that although it is presently<br />
impossible to reliably predict gene function<br />
based on sequence, it should nevertheless<br />
be within reach to develop mechanisms that<br />
can help categorize sequences as belonging to<br />
predefined ‘hazardous’ or ‘safe’ classes of genes,<br />
an effort that could greatly improve the future<br />
efficiency of synthetic gene order screening.<br />
Several parallel efforts are also underway to<br />
develop more sophisticated and comprehensive<br />
pathogen databases. Fischer and Maurer are<br />
collaborating on Virulence Factor Information<br />
Repository (VIREP), a repository for annotated<br />
information about known virulence genes,<br />
based at UC, Berkeley. The IGSC has also stated<br />
its intention to develop an extensive regulated<br />
pathogens database, which could offer a broadly<br />
useful community resource. However, both<br />
groups are waiting on government support to<br />
help move these projects forward.<br />
For now, the member companies of the IGSC,<br />
which are predominantly based in the US, are<br />
moving to adapt their standards to comply with<br />
the HHS recommendations. However, the guidance<br />
also invites companies to apply their own<br />
“equivalent or superior” screening standards<br />
and several companies indicate that they will<br />
continue to err on the side of caution in their<br />
screening procedures. “If we get a gene in, we<br />
screen it,” says Robert Dawson, director of bioinformatics<br />
at Coralville, Iowa–based Integrated<br />
DNA Technologies. “There’s never a case where<br />
we would have a gene go right into production<br />
without a human being having looked at both<br />
the sequence and the prospective customer.”<br />
HHS has also made it clear that these are minimum<br />
screening recommendations and not the<br />
final word, and discussions are ongoing.<br />
Given the early stage of the field, when the<br />
risk from synthetic biology is still seen as relatively<br />
low—to date, no IGSC member company<br />
reports having received an order for a ‘sequence<br />
of concern’ that also came from a dubious customer—some<br />
hope that there will be sufficient<br />
opportunity for these guidelines to grow into a<br />
more effective monitoring strategy. “It’s a line<br />
in the sand drawn by the US government that<br />
now serves as something to be improved over<br />
time,” says Diggans. “All of these things make<br />
a direct contribution to maintaining near-term<br />
biosecurity, but it will need to evolve quickly—<br />
the technology is moving ever faster.”<br />
Michael Eisenstein, Philadelphia<br />
1226 volume 28 number 12 december 2010 nature biotechnology
NEWS<br />
© 2010 Nature America, Inc. All rights reserved.<br />
in brief<br />
DNA not patentable<br />
In a spine-chilling announcement, the<br />
US Department of Justice in October said<br />
unmodified human DNA should not be eligible<br />
for patent. The department’s stance conflicts<br />
with the body of case law on the matter and a<br />
longstanding position held by the US Patent<br />
and Trademark Office, which has issued<br />
more than 10,000 of these patents. The<br />
Justice Department announced its position<br />
in response to a lawsuit involving patents on<br />
breast cancer genes BRCA1 and BRCA2. A US<br />
district court in March declared the patents<br />
invalid, saying that the genes are products of<br />
nature rather than human-made inventions.<br />
Patent holders University of Utah and Myriad<br />
Genetics, based in Salt Lake City, appealed<br />
in June. In an amicus brief filed with the<br />
Federal Circuit Court of Appeals the Justice<br />
Department agreed that identifying and<br />
isolating DNA without further manipulation<br />
is not an invention, or patent eligible. How<br />
the agency’s declaration will influence<br />
justices and the patent office worries biotech<br />
companies. But the patent office isn’t easily<br />
swayed, says Thomas Kowalski, an attorney<br />
with Vedder Price in New York. “The patent<br />
office is not going to change what it’s doing in<br />
view of what the Department of Justice says,”<br />
he says. Besides, international agreements<br />
between the American, European and<br />
Japanese patent offices, known as Trilateral<br />
Co-operation, have concluded that unmodified<br />
DNA is patentable.<br />
Emily Waltz<br />
USPTO’s do-good vouchers<br />
Patent owners may soon be able to cut the<br />
time needed to have a patent reexamined by<br />
up to two-thirds if they can demonstrate a<br />
humanitarian use. The US Patent & Trademark<br />
Office (USPTO) is proposing a system that would<br />
offer fast-track reexamination vouchers as an<br />
incentive to stimulate “creation or licensing<br />
that addresses humanitarian needs.” Because<br />
patents under reexamination are often the most<br />
valuable commercially, a fast-track procedure<br />
would let patent owners “more readily and less<br />
expensively affirm the validity of their patents,”<br />
according to the Federal Register notice. The<br />
USPTO currently takes 19 to 20 months for<br />
such reexaminations, whereas the expedited<br />
review promises a six-month turnaround. The<br />
system is modeled on the US Food & Drug<br />
Administration’s priority review vouchers given<br />
to entities that develop drugs to treat neglected<br />
tropical diseases. In this case, patent holders<br />
who receive the fast-track reexamination voucher<br />
could use it on any other patent they own or<br />
transfer it to the open market. Although the<br />
intent is worthy, there are too many unanswered<br />
questions, worries Thomas Kowalski of Vedder<br />
Price. “What will the USPTO do to ensure that<br />
those in the developing world as well as the<br />
poor in the developed world can gain access<br />
to the technology? Also, the voucher should<br />
be tied specifically to the technology with the<br />
humanitarian use instead of being independent<br />
and transferable.” Michael Francisco<br />
Table 1 Timeline of events leading up to the synthetic DNA guidance<br />
December 2006 The US government’s National Science Advisory Board for Biosecurity issues a report<br />
recommending the establishment of “uniform and standardized screening practices<br />
among providers of synthetic DNA.”<br />
June 2007 The US Government convenes an interagency working group on synthetic nucleic acid<br />
screening.<br />
November 3, 2009 Companies of the International Association Synthetic Biology (Entelechon,<br />
ATG:biosynthetics, Biomax, febit and Sloning Biotechnology) present their “Code of<br />
Conduct” for the screening of gene orders and customers.<br />
November 19, 2009 Companies of the International Gene Synthesis Consortium (Blue Heron Biotechnology,<br />
DNA2.0, GENEART, GenScript and Integrated DNA Technologies) release a similar set of<br />
‘best practices’ guidelines, the “Harmonized Screening Protocol.”<br />
November 29, 2009 The US Department of Health and Human Services releases a draft version of its<br />
Screening Framework Guidance for Providers of Synthetic Double-Stranded DNA for<br />
public comment.<br />
January 11, 2010 The American Association for the Advancement of Science (AAAS) hosts a meeting for policy<br />
specialists and scientists from academia and industry on “Minimizing the Risks of Synthetic<br />
DNA: Scientists’ Views on the U.S. Government’s Guidance on Synthetic Genomics.”<br />
January 26, 2010 Public comment period ends on the draft guidance.<br />
October 13, 2010 HHS publishes its final version of the Screening Framework Guidance.<br />
He adds, “The annotation is as provided by the<br />
person that deposited the sequence.” Screening<br />
effectiveness could also be constrained by<br />
biases in the database contents, according to<br />
James Diggans, a researcher at MITRE, a notfor-profit<br />
national technology resource that<br />
focuses on security issues, located in Bedford,<br />
Massachusetts and McLean, Virginia. “There are<br />
far more harmless sequences in these databases<br />
than there are sequences that could be used to<br />
harm human health.”<br />
For other scientists, the reliance on the select<br />
agents list is also problematic. Eighty-two items<br />
currently listed represent known risks to human,<br />
plant or animal health, and are unambiguously<br />
regulated by federal law. But many known<br />
human pathogens, such as severe acute respiratory<br />
syndrome virus, are omitted, and others<br />
worry about the problems that could be posed<br />
by the yet-unknown sequence variants. “If you<br />
synthesize a genome without creating the actual<br />
organism it encodes—and where now you aren’t<br />
even limited to the variability found in nature—<br />
how do you taxonomically classify that genome<br />
sequence?” says Eddy.<br />
Eddy and other scientists recently partnered<br />
with the US National Research Council in an<br />
effort to bring some clarity to the characterization<br />
of high-risk genes. The resulting report,<br />
Sequence-Based Classification of Select Agents:<br />
A Brighter Line (http://www.nap.edu/catalog/<br />
12970.html), concludes that although it is presently<br />
impossible to reliably predict gene function<br />
based on sequence, it should nevertheless<br />
be within reach to develop mechanisms that<br />
can help categorize sequences as belonging to<br />
predefined ‘hazardous’ or ‘safe’ classes of genes,<br />
an effort that could greatly improve the future<br />
efficiency of synthetic gene order screening.<br />
Several parallel efforts are also underway to<br />
develop more sophisticated and comprehensive<br />
pathogen databases. Fischer and Maurer are<br />
collaborating on Virulence Factor Information<br />
Repository (VIREP), a repository for annotated<br />
information about known virulence genes,<br />
based at UC, Berkeley. The IGSC has also stated<br />
its intention to develop an extensive regulated<br />
pathogens database, which could offer a broadly<br />
useful community resource. However, both<br />
groups are waiting on government support to<br />
help move these projects forward.<br />
For now, the member companies of the IGSC,<br />
which are predominantly based in the US, are<br />
moving to adapt their standards to comply with<br />
the HHS recommendations. However, the guidance<br />
also invites companies to apply their own<br />
“equivalent or superior” screening standards<br />
and several companies indicate that they will<br />
continue to err on the side of caution in their<br />
screening procedures. “If we get a gene in, we<br />
screen it,” says Robert Dawson, director of bioinformatics<br />
at Coralville, Iowa–based Integrated<br />
DNA Technologies. “There’s never a case where<br />
we would have a gene go right into production<br />
without a human being having looked at both<br />
the sequence and the prospective customer.”<br />
HHS has also made it clear that these are minimum<br />
screening recommendations and not the<br />
final word, and discussions are ongoing.<br />
Given the early stage of the field, when the<br />
risk from synthetic biology is still seen as relatively<br />
low—to date, no IGSC member company<br />
reports having received an order for a ‘sequence<br />
of concern’ that also came from a dubious customer—some<br />
hope that there will be sufficient<br />
opportunity for these guidelines to grow into a<br />
more effective monitoring strategy. “It’s a line<br />
in the sand drawn by the US government that<br />
now serves as something to be improved over<br />
time,” says Diggans. “All of these things make<br />
a direct contribution to maintaining near-term<br />
biosecurity, but it will need to evolve quickly—<br />
the technology is moving ever faster.”<br />
Michael Eisenstein, Philadelphia<br />
1226 volume 28 number 12 december 2010 nature biotechnology
news<br />
BARDA funds vaccine makers aiming to phase out eggs<br />
© 2010 Nature America, Inc. All rights reserved.<br />
The US government handed out contracts worth<br />
potentially more than $100 million to boost<br />
innovative technologies that can deliver vaccines<br />
in large quantities—and fast—as part of a<br />
$1.9 billion initiative to protect Americans from<br />
biologic threats of the future. The Biomedical<br />
Advanced Research and Development Authority<br />
(BARDA) has awarded money to eight firms<br />
engaged in creative ways to make vaccines<br />
against naturally occurring diseases, such as<br />
the H1N1 influenza, which was pandemic last<br />
year, and others it views as potential bioterrorist<br />
weapons. The goal, Kathleen Sebelius,<br />
Department of Health and Human Services<br />
(HHS) secretary says, is to create “nimble, flexible<br />
capacity to produce medical countermeasures<br />
rapidly in the face of any attack or threat.”<br />
Contract winners announced in September<br />
(Table 1) include VaxDesign, Novartis partnered<br />
with Synthetic Genomics Vaccines, the<br />
nonprofits PATH (Program for Appropriate<br />
Technology in Health) and the Infectious<br />
Disease Research Institute, Pfenex, Rapid<br />
Micro Biosystems, Emergent Biosolutions, 3M<br />
and Northrop Grumman Security Systems.<br />
BARDA’s push to make the vaccine development<br />
and production process move beyond<br />
half-century-old systems that rely on fertilized<br />
eggs to culture most vaccines has been<br />
several months in gestation. In August, two<br />
reports looked at the US’s ability to produce<br />
vaccines quickly and efficiently and came<br />
away holding their noses. “The closer we<br />
looked at the countermeasure pipeline, the<br />
more leaks, choke points and dead ends we<br />
saw. In this age of new threats, we aren’t generating<br />
enough new products,” said Sebelius<br />
A half-century-old technology. Fertilized eggs were used to produce a vaccine for the H1N1 flu<br />
virus at the Sinovac plant in Beijing during the 2009 outbreak.<br />
at a press conference on August 19 during<br />
which both reports were presented.<br />
The first report, The Public Health Emergency<br />
Medical Countermeasures Enterprise Review<br />
(https://www.medicalcountermeasures.gov/<br />
documents/MCMReviewFinalcover-508.<br />
pdf), commissioned by the HHS called for<br />
modernization and speed-up of the regulatory<br />
procedures, the development of more flexible<br />
vaccine manufacturing processes to allow<br />
more than one product to be produced at the<br />
same facility, and an upgrade and modernization<br />
of everything related to rapid influenza<br />
vaccine production. Nearly $2 billion was put<br />
aside to reach the report’s goals.<br />
More quantitative was the President’s<br />
Council of Advisors on Science and<br />
Technology’s (PCAST) report entitled<br />
Reengineering the Influenza Vaccine Production<br />
Enterprise to Meet the Challenges of Pandemic<br />
Influenza (http://www.whitehouse.gov/sites/<br />
default/files/microsites/ostp/Influenza%20<br />
Vaccinology.pdf). It outlines every stumbling<br />
block to speeding up vaccine development but<br />
more significantly lists how much time could<br />
be saved by adapting improvements already<br />
in the pipeline. The Influenza Vaccinology<br />
Working group, chaired by Eric Lander, director<br />
of the Broad Institute of Harvard and MIT<br />
in Cambridge, Massachusetts, and Harold<br />
AP Photo/Greg Baker<br />
Table 1 Vaccine manufacturers awarded BARDA contracts<br />
Company/location Project aims Contract value<br />
Novartis Vaccines and Diagnostics/<br />
Cambridge, Massachusetts<br />
Pfenex/<br />
San Diego<br />
VaxDesign/<br />
Orlando, Florida<br />
Northrop Grumman Security Systems/<br />
Baltimore<br />
PATH/<br />
Seattle<br />
Infectious Disease Research<br />
Institute (IDRI)/<br />
Seattle<br />
Rapid Micro Biosystems/<br />
Bedford, Massachusetts<br />
3M/<br />
St. Paul, Minnesota<br />
Novartis Vaccines and Diagnostics will investigate techniques for the rapid development of optimized influenza<br />
seed virus.<br />
Pfenex will apply its Pfenex Expression Technology Platform to the development of optimized bioprocesses for<br />
high-yield production of a stable candidate anthrax vaccine.<br />
VaxDesign will further develop its MIMIC platform, an in vitro human immune system mimetic designed to accelerate<br />
evaluation of candidate and stockpiled vaccine safety and effectiveness by supplementing animal testing.<br />
Northup will develop integrated diagnostic capabilities for rapid, high-throughput surveillance and molecular<br />
diagnostics.<br />
~ $24 million<br />
over three years<br />
~$18.8 million<br />
over three years<br />
~$17.1 million<br />
over three years<br />
~$9.8 million<br />
over one year<br />
PATH will test multiple innovative formulation chemistries and strategies to increase the shelf life of influenza ~$9.4 million<br />
vaccines, which has implications for the national vaccine stockpile as well as cold-chain requirements domestically<br />
and in developing<br />
over three years<br />
countries.<br />
IDRI will develop and evaluate innovative adjuvant formulations to enhance influenza vaccine immunogenicity<br />
and cross-protection to make them more effective against novel viral strains that may cause the next pandemic.<br />
Rapid Micro Biosystems will develop methods for accelerated sterility testing. Together, these improvements could<br />
shave weeks off the influenza vaccine manufacturing and product release schedule.<br />
3M will develop integrated diagnostic capabilities for rapid, high-throughput surveillance and molecular<br />
diagnostics.<br />
~$8.5 million<br />
over three years<br />
~$6.8 million<br />
over three years<br />
~$6 million<br />
over two years<br />
nature biotechnology volume 28 number 12 december 2010 1227
NEWS<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Vaccine makers’ immunity questioned in court<br />
The US Supreme Court has begun considering how much liability vaccine makers have if<br />
the side effects of their products are believed to have injured or killed someone. The case<br />
was brought against Wyeth (now merged with Pfizer of New York) by parents of Hannah<br />
Bruesewitz, who in 1992 began suffering seizures and developmental problems after<br />
being given the combined Corynebacterium diphtheriae toxoid/Clostridium tetani toxoid/<br />
polio (DTP) vaccine against diphtheria, tetanus and pertussis (whooping cough). A few<br />
years later, DTP was removed from the market and replaced by a vaccine with fewer side<br />
effects. The Bruesewitzes believed their daughter’s injuries were avoidable because Wyeth<br />
should have put a product with fewer side effects on the market earlier.<br />
What is most notable about the Bruesewitz v. Wyeth case, which was argued on October<br />
12 in Washington, DC, is that many in the US drug industry had believed that the issue<br />
had been completely resolved with the adoption in 1986 of the National Childhood<br />
Vaccine Injury Act. The act set up a Vaccine Court to adjudicate claims of injury on a nofault<br />
basis and pay successful claimants with money generated from a tax on vaccines.<br />
The Vaccine Act was put into effect because of a fear at the time that lawsuits claiming<br />
‘design defects’ would force companies to stop making vaccines. Accordingly, the act<br />
says suits cannot be filed against manufacturers “if the injury or death resulted from side<br />
effects that were unavoidable, even though the vaccine was properly prepared and was<br />
accompanied by proper directions and warnings.”<br />
There is a back door to the law that allows families to go to a federal court if they lose<br />
in Vaccine Court or they don’t like the amount of its judgment. However, those suits are<br />
governed by the Vaccine Act, too.<br />
But neither the Vaccine Court nor a lower US federal court accepted the<br />
Bruesewitzes’ argument that their daughter’s injuries could have been avoided by<br />
the manufacturer. However, the justices found the wording in the Act, and especially<br />
its use of the word “unavoidable” quite confused. Justice Stephen Breyer remarked<br />
“it’s pretty hard to say the word unavoidable means avoidable.” A final judgment is<br />
expected in early spring of 2011.<br />
Stephen Strauss<br />
Varmus, then at the Memorial Sloan-Kettering<br />
Cancer Center, also estimates the time it would<br />
take before these changes could be instituted.<br />
From rapid sterility testing, to accelerated<br />
virus seed production and improved adjuvants,<br />
each advance could slice several weeks off the<br />
time for the first dose to reach the market. The<br />
PCAST committee members believed these<br />
changes individually could be put in place within<br />
1 to 3 years. To change egg-based vaccine production<br />
systems for alternative cell or recombinant<br />
DNA platforms would require longer—up<br />
to a decade—to reach market penetration.<br />
BARDA’s R&D money will help push a<br />
broad swath of potentially game-changing new<br />
technologies, but deputy assistant secretary<br />
of BARDA, Robin Robinson, admits that it<br />
doesn’t cover the gamut of vaccine innovations<br />
in development. In particular, Robinson points<br />
to efforts to grow vaccines in plants and insect<br />
cells. Some of these projects are being funded<br />
by other US government agencies, most notably<br />
the Defense Advanced Research Project<br />
Agency, which is supporting four tobaccobased<br />
vaccine production platforms.<br />
In plants, the process is quicker than in<br />
eggs. Andy Sheldon, president and CEO of<br />
Medicago of Quebec City, Canada, says “it<br />
takes five weeks to grow the tobacco, the<br />
plants start expressing the protein in five days,<br />
and then it takes two days to purify the VLPs<br />
[virus-like particles].” He compares this to<br />
the six months egg-based vaccine production<br />
takes. Medicago is entering phase 2 clinical trials<br />
with its plant-derived flu vaccine.<br />
Plant-based production is cheaper too. The<br />
manufacturing facilities Medicago plans for<br />
Raleigh, North Carolina, will cost $25 million<br />
to build, a far smaller investment than the $250<br />
million required for an egg-based production<br />
plant and the $1 billion that Novartis recently<br />
spent on a new Holly Springs, North Carolina<br />
facility. If approval is granted, it will become<br />
the first facility in the United States licensed to<br />
use mammalian cells to produce flu vaccines<br />
and is expected to be operational in 2013.<br />
A seasonal influenza vaccine, FluBlok,<br />
produced in insect cell culture, could be on<br />
the market next year. Protein Sciences of<br />
Meriden, Connecticut, received a BARDA<br />
contract in 2009 to use cells from fall armyworm<br />
(Spodoptera frugiperda) with a baculovirus<br />
system to generate influenza VLPs.<br />
Protein Science’s president and CEO, Manon<br />
Cox, says it takes about two months from virus<br />
discovery to vaccine production using insect<br />
cells for production. It is also cheaper than<br />
egg-based vaccine manufacturing. “Licensed<br />
vaccines cost approximately $1 a dose to make<br />
the active ingredients. Our estimates are that<br />
we can make three times more product for that<br />
price,” claims Cox. Protein Sciences is waiting<br />
for US Food and Drug Administration<br />
approval of FluBlok.<br />
Pfenex, too, says its technology is nearing<br />
the market. Last year, the Defense Threat<br />
Reduction Agency provided the company<br />
with a DNA sequence of an unknown antigen<br />
and challenged them to develop both a production<br />
strain and a high-speed, high-quality,<br />
low-cost, antigen-production process. In conjunction<br />
with partner organizations, Pfenex<br />
used its screening technology to do this within<br />
42 days. And there were cost savings. “If you<br />
scaled up to production levels, the antigen<br />
can be produced for ~50 cents per dose,” says<br />
Patrick Lucy, Pfenex’s vice president of business<br />
development.<br />
But the issue that looms largest in the push<br />
to modernize and speed up vaccine development<br />
relates to the business end of things. How<br />
are these innovative technologies going to fit<br />
into an existing vaccine marketplace that—<br />
influenza pandemics and potential terrorist<br />
bio-attacks aside—generally satisfies the<br />
world’s vaccine needs?<br />
For example, the PCAST report pointed<br />
out that although the Novartis cell culture<br />
facility was likely to generate annual profits<br />
of $30 million, it “would take over 30 years to<br />
recover the [$1 billion] investment in nominal<br />
dollars (leaving aside the need for a return<br />
on investment).”<br />
Protein Science’s Cox argues this naturally<br />
leads vaccine manufacturers, using egg-based<br />
technologies, to resist any change. “They are<br />
not going to easily let that [advantage] be<br />
taken away by a new technology in which<br />
their learning curve is going to be as steep<br />
as anybody’s else’s,” she says. Rafick-Pierre<br />
Sékaly, co-director and scientific director of<br />
the Vaccine and Gene Therapy Institute of Port<br />
St. Lucie, Florida, concurs. “The president and<br />
the committee can make all the recommendations<br />
they want but if the big vaccine makers<br />
say it is too costly or there is too much R&D,<br />
then changes are going to be treated not as a<br />
solution but as an added burden.”<br />
On this point, BARDA’s Robinson says, “we<br />
understand, and that is why we are pushing<br />
things that will definitely benefit all vaccines,<br />
including eggs.” Indeed, in October BARDA<br />
awarded Sanofi Pasteur of Lyon, France, a<br />
3-year, $57 million contract to make more fertilized<br />
eggs available for vaccine production on<br />
a year-round basis.<br />
Stephen Strauss, Toronto<br />
1228 volume 28 number 12 december 2010 nature biotechnology
NEWS<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Vaccine makers’ immunity questioned in court<br />
The US Supreme Court has begun considering how much liability vaccine makers have if<br />
the side effects of their products are believed to have injured or killed someone. The case<br />
was brought against Wyeth (now merged with Pfizer of New York) by parents of Hannah<br />
Bruesewitz, who in 1992 began suffering seizures and developmental problems after<br />
being given the combined Corynebacterium diphtheriae toxoid/Clostridium tetani toxoid/<br />
polio (DTP) vaccine against diphtheria, tetanus and pertussis (whooping cough). A few<br />
years later, DTP was removed from the market and replaced by a vaccine with fewer side<br />
effects. The Bruesewitzes believed their daughter’s injuries were avoidable because Wyeth<br />
should have put a product with fewer side effects on the market earlier.<br />
What is most notable about the Bruesewitz v. Wyeth case, which was argued on October<br />
12 in Washington, DC, is that many in the US drug industry had believed that the issue<br />
had been completely resolved with the adoption in 1986 of the National Childhood<br />
Vaccine Injury Act. The act set up a Vaccine Court to adjudicate claims of injury on a nofault<br />
basis and pay successful claimants with money generated from a tax on vaccines.<br />
The Vaccine Act was put into effect because of a fear at the time that lawsuits claiming<br />
‘design defects’ would force companies to stop making vaccines. Accordingly, the act<br />
says suits cannot be filed against manufacturers “if the injury or death resulted from side<br />
effects that were unavoidable, even though the vaccine was properly prepared and was<br />
accompanied by proper directions and warnings.”<br />
There is a back door to the law that allows families to go to a federal court if they lose<br />
in Vaccine Court or they don’t like the amount of its judgment. However, those suits are<br />
governed by the Vaccine Act, too.<br />
But neither the Vaccine Court nor a lower US federal court accepted the<br />
Bruesewitzes’ argument that their daughter’s injuries could have been avoided by<br />
the manufacturer. However, the justices found the wording in the Act, and especially<br />
its use of the word “unavoidable” quite confused. Justice Stephen Breyer remarked<br />
“it’s pretty hard to say the word unavoidable means avoidable.” A final judgment is<br />
expected in early spring of 2011.<br />
Stephen Strauss<br />
Varmus, then at the Memorial Sloan-Kettering<br />
Cancer Center, also estimates the time it would<br />
take before these changes could be instituted.<br />
From rapid sterility testing, to accelerated<br />
virus seed production and improved adjuvants,<br />
each advance could slice several weeks off the<br />
time for the first dose to reach the market. The<br />
PCAST committee members believed these<br />
changes individually could be put in place within<br />
1 to 3 years. To change egg-based vaccine production<br />
systems for alternative cell or recombinant<br />
DNA platforms would require longer—up<br />
to a decade—to reach market penetration.<br />
BARDA’s R&D money will help push a<br />
broad swath of potentially game-changing new<br />
technologies, but deputy assistant secretary<br />
of BARDA, Robin Robinson, admits that it<br />
doesn’t cover the gamut of vaccine innovations<br />
in development. In particular, Robinson points<br />
to efforts to grow vaccines in plants and insect<br />
cells. Some of these projects are being funded<br />
by other US government agencies, most notably<br />
the Defense Advanced Research Project<br />
Agency, which is supporting four tobaccobased<br />
vaccine production platforms.<br />
In plants, the process is quicker than in<br />
eggs. Andy Sheldon, president and CEO of<br />
Medicago of Quebec City, Canada, says “it<br />
takes five weeks to grow the tobacco, the<br />
plants start expressing the protein in five days,<br />
and then it takes two days to purify the VLPs<br />
[virus-like particles].” He compares this to<br />
the six months egg-based vaccine production<br />
takes. Medicago is entering phase 2 clinical trials<br />
with its plant-derived flu vaccine.<br />
Plant-based production is cheaper too. The<br />
manufacturing facilities Medicago plans for<br />
Raleigh, North Carolina, will cost $25 million<br />
to build, a far smaller investment than the $250<br />
million required for an egg-based production<br />
plant and the $1 billion that Novartis recently<br />
spent on a new Holly Springs, North Carolina<br />
facility. If approval is granted, it will become<br />
the first facility in the United States licensed to<br />
use mammalian cells to produce flu vaccines<br />
and is expected to be operational in 2013.<br />
A seasonal influenza vaccine, FluBlok,<br />
produced in insect cell culture, could be on<br />
the market next year. Protein Sciences of<br />
Meriden, Connecticut, received a BARDA<br />
contract in 2009 to use cells from fall armyworm<br />
(Spodoptera frugiperda) with a baculovirus<br />
system to generate influenza VLPs.<br />
Protein Science’s president and CEO, Manon<br />
Cox, says it takes about two months from virus<br />
discovery to vaccine production using insect<br />
cells for production. It is also cheaper than<br />
egg-based vaccine manufacturing. “Licensed<br />
vaccines cost approximately $1 a dose to make<br />
the active ingredients. Our estimates are that<br />
we can make three times more product for that<br />
price,” claims Cox. Protein Sciences is waiting<br />
for US Food and Drug Administration<br />
approval of FluBlok.<br />
Pfenex, too, says its technology is nearing<br />
the market. Last year, the Defense Threat<br />
Reduction Agency provided the company<br />
with a DNA sequence of an unknown antigen<br />
and challenged them to develop both a production<br />
strain and a high-speed, high-quality,<br />
low-cost, antigen-production process. In conjunction<br />
with partner organizations, Pfenex<br />
used its screening technology to do this within<br />
42 days. And there were cost savings. “If you<br />
scaled up to production levels, the antigen<br />
can be produced for ~50 cents per dose,” says<br />
Patrick Lucy, Pfenex’s vice president of business<br />
development.<br />
But the issue that looms largest in the push<br />
to modernize and speed up vaccine development<br />
relates to the business end of things. How<br />
are these innovative technologies going to fit<br />
into an existing vaccine marketplace that—<br />
influenza pandemics and potential terrorist<br />
bio-attacks aside—generally satisfies the<br />
world’s vaccine needs?<br />
For example, the PCAST report pointed<br />
out that although the Novartis cell culture<br />
facility was likely to generate annual profits<br />
of $30 million, it “would take over 30 years to<br />
recover the [$1 billion] investment in nominal<br />
dollars (leaving aside the need for a return<br />
on investment).”<br />
Protein Science’s Cox argues this naturally<br />
leads vaccine manufacturers, using egg-based<br />
technologies, to resist any change. “They are<br />
not going to easily let that [advantage] be<br />
taken away by a new technology in which<br />
their learning curve is going to be as steep<br />
as anybody’s else’s,” she says. Rafick-Pierre<br />
Sékaly, co-director and scientific director of<br />
the Vaccine and Gene Therapy Institute of Port<br />
St. Lucie, Florida, concurs. “The president and<br />
the committee can make all the recommendations<br />
they want but if the big vaccine makers<br />
say it is too costly or there is too much R&D,<br />
then changes are going to be treated not as a<br />
solution but as an added burden.”<br />
On this point, BARDA’s Robinson says, “we<br />
understand, and that is why we are pushing<br />
things that will definitely benefit all vaccines,<br />
including eggs.” Indeed, in October BARDA<br />
awarded Sanofi Pasteur of Lyon, France, a<br />
3-year, $57 million contract to make more fertilized<br />
eggs available for vaccine production on<br />
a year-round basis.<br />
Stephen Strauss, Toronto<br />
1228 volume 28 number 12 december 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.<br />
Ethanol blend hike to jump start cellulosic<br />
investment<br />
A move announced on October 13 by the US<br />
Environmental Protection Agency (EPA) to<br />
raise the permitted ethanol blend in gasoline<br />
for motor vehicles has been widely welcomed<br />
by ethanol manufacturers and biofuel<br />
advocate groups. The increase of ethanol in<br />
gasoline, from 10% to 15%, revises a mandate<br />
that has stood for 35 years and is thought to<br />
be a critical step toward bringing cellulosic<br />
ethanol to consumers.<br />
The EPA’s decision applies only to light<br />
motor vehicles made since 2007—a second<br />
decision on vehicles made from 2001 to 2006<br />
is expected by the year end—but it’s still<br />
enough to give hope to biotechs working<br />
on cellulosic ethanol (and other non-corn<br />
feedstock ‘second-generation’ biofuels). The<br />
problem for cellulosic ethanol producers<br />
was that the so-called E10 blend limit—10%<br />
ethanol—left the market saturated with<br />
corn ethanol, says Jim Sturdevant, director<br />
of Project Liberty, the cellulosic operation<br />
of Poet, of Sioux Falls, South Dakota. Raising<br />
that limit should open the door to companies<br />
like his—critical, Sturdevant says, because<br />
“without a demand, it’s not worth investing<br />
in cellulosic ethanol.”<br />
The E15 decision could indeed revive<br />
investor interest in non-grain-based ethanol<br />
companies. Investor interest—in both biofuels<br />
and bio-based materials—grew during the<br />
first few years of this decade, reaching a peak<br />
in 2007 (Fig. 1). That meant companies were<br />
able to quickly acquire seed money and take<br />
their technologies to the pilot plant level,<br />
but as financing became harder to acquire<br />
and interest petered out, progress stalled.<br />
That was all part of “a natural dropoff ” in<br />
the cycle of early technology development,<br />
says David Berry, a principal at Flagship<br />
Ventures in Cambridge, Massachusetts, but<br />
the blend wall decision could start money<br />
flowing again.<br />
Berry says the biggest impact would be<br />
on the later-stage companies in need of<br />
financing to “cross the chasm” from pilotscale<br />
facilities to large manufacturing plants.<br />
An example is Mascoma, of Lebanon, New<br />
Hampshire, which is in the process of securing<br />
funding for its first industrial-scale cellulosic<br />
plant, in Kinross, Michigan. After the<br />
EPA’s decision, it will be “considerably easier<br />
to raise equity for this sort of facility,” says<br />
Justin van Rooyen, Mascoma’s director of<br />
business development.<br />
But corn ethanol—for all the roads it<br />
has paved—can still be an impediment to<br />
Car drivers in the US can now fill their tanks with<br />
a higher blend of ethanol. But cellulosic biofuel<br />
producers want to see the limit rise further.<br />
second-generation biofuels. It’s part of the<br />
reason 54 ethanol manufacturers and Growth<br />
Energy, an ethanol supporters’ group based in<br />
Washington, DC, submitted the E15 Green Jobs<br />
Waiver to the EPA in the spring of 2009, petitioning<br />
the organization to amend the ethanol<br />
blend limit to 15%. The lobby group realized<br />
the 10% ethanol opportunity in gasoline was<br />
becoming saturated by existing corn ethanol<br />
manufacturers.<br />
Cynthia Bryant, marketing manager, global<br />
fuels at Novozymes North America, with<br />
headquarters in Bagsvaerd, Denmark calls<br />
the EPA’s decision “a validation of what we’ve<br />
been saying all along—that higher blends of<br />
ethanol can be used in our cars today.” A 15%<br />
blend puts ethanol into 43 million cars; if the<br />
EPA allows the blend into cars built in and<br />
after 2001, it would add another 83 million<br />
cars. That could, in effect, reach 54% of the<br />
cars in the US.<br />
That’s a huge step, though the government<br />
is also assisting the growth of biofuels in<br />
other ways. In October the US Department of<br />
Agriculture said its Biomass Crop Assistance<br />
Program will supply $525 million in subsidies<br />
to farmers growing second-generation biomass<br />
crops over a 15-year period. But the future of<br />
cellulosic ethanol in the US will eventually<br />
come down to the hypothetical consumer,<br />
and he or she is already being discussed. The<br />
EPA E15 decision came with a suggested<br />
label to identify pumps dispensing the new<br />
blend: bright orange, with the word ‘caution’<br />
printed in bold at the top. It is not final, but<br />
has raised eyebrows anyway, as those in the<br />
biofuels industry say the label could scare away<br />
customers. “It’s a little ominous looking,” says<br />
Dreyer, particularly because “all of our testing<br />
has shown that E15 is perfectly safe.”<br />
Growth Energy is seeking federal legislation<br />
that will require country-of-origin<br />
Poet<br />
in brief<br />
Singapore injects $12.5<br />
billion<br />
news<br />
The Singaporean government will spend<br />
S$16.1 ($12.5) billion on research<br />
innovation over the next 4 years—a 20%<br />
increase over the previous budget. A quarter<br />
of the funding—S$3.7 ($2.9) billion —is<br />
allotted to biomedical science, according to<br />
the September announcement. The funding<br />
boost “reflects our steady commitment<br />
to transforming our economy,” says Beh<br />
Kian Teik, director of biomedical sciences<br />
of Singapore’s Economic Development<br />
Board. Since 2000, when the Singaporean<br />
government outlined its long-term plan<br />
to transform this tiny nation-state into a<br />
knowledge-based economy, the government<br />
ploughed roughly S$25 ($19.3) billion to<br />
shore up research infrastructure and create a<br />
talent pool to attract private investment. The<br />
figures suggest Singapore has succeeded.<br />
The country’s biomedical manufacturing<br />
output alone more than tripled from S$6.3<br />
billion ($4.9 billion) in 2000 to S$21 billion<br />
($16.2 billion) in 2009. The biomedical<br />
sector now represents 10% of the country’s<br />
manufacturing output, up from 4% in 2000.<br />
Most major pharma companies have a<br />
presence in Singapore, although in October,<br />
Eli Lilly of Indianapolis announced the<br />
closure of its Singapore Centre for Drug<br />
Discovery. The new funding budget will<br />
make “more money available for everyone,”<br />
says Beh, with an emphasis on initiatives<br />
that favor “economic outcomes.” In fact,<br />
some are questioning whether the shift in<br />
emphasis toward translational outcomes and<br />
away from blue sky research is too myopic.<br />
Some of the funding has been earmarked<br />
to support the Biomedical Science Industry<br />
Partnership Office (IPO), a fledgling agency<br />
that supports public-private research projects<br />
and helps companies pool expertise across<br />
Singaporean research institutes and medical<br />
centers. The IPO office has succeeded in<br />
setting up collaborations for Basel-based<br />
Roche and London-based GlaxoSmithKline<br />
with Singaporean researchers. Roche<br />
plans to spend S$130 ($100.6) million to<br />
support a center for translational medicine<br />
to develop drugs for the Asian market;<br />
whereas GlaxoSmithKline will participate in<br />
four public-private research collaborations<br />
focused on early-stage research in<br />
ophthalmology, regenerative medicine and<br />
neurodegeneration. “This should be good<br />
news for the biotech sector,” says Jan-Anders<br />
Karlsson, CEO of S*BIO, a drug discovery<br />
company set up as a joint venture between<br />
the Singapore Economic Development Board<br />
Investments and Chiron Corporation in 2000.<br />
Government estimates that the nation’s<br />
economy will have grown by 13–15% by<br />
year’s end—higher than estimated growth<br />
for China and India. In October, Basel-based<br />
agribusiness Syngenta opened an R&D<br />
facility in Singapore to support technology<br />
development in the Asia Pacific region.<br />
<br />
Gunjan Sinha<br />
nature biotechnology volume 28 number 12 december 2010 1229
NEWS<br />
© 2010 Nature America, Inc. All rights reserved.<br />
in brief<br />
Good ideas across borders<br />
The European Commission (EC) has made a<br />
bold move to improve Europe’s competitiveness<br />
and create an environment conducive to<br />
innovation with the launch of Innovation Union.<br />
Part of the Europe 2020 initiative for recovery<br />
from the recession and economic growth over<br />
the next decade, the program, announced<br />
October 6, aims to “boost the whole innovation<br />
chain from ‘research to retail’,” says Mark<br />
English, spokesperson for the Commissioner for<br />
Research, Innovation and Science in Brussels.<br />
The EC is calling on member states to spend<br />
3% of the gross domestic product (GDP) on<br />
R&D by 2020, a hike from the current 1.9% of<br />
GDP. Europe is the largest market in the world,<br />
but its policies and support for R&D remain<br />
fragmented. The EC is proposing to boost<br />
cooperation across borders, among companies,<br />
and between private and public sectors. Plans<br />
to adopt a single EU patent system, and provide<br />
tax incentives and legislation that will allow<br />
venture capital funds to invest freely anywhere<br />
in the EU are also part of the Innovation Union’s<br />
commitments. Monika Wcislo, press officer for<br />
Research, Innovation and Science, notes that<br />
funding is not earmarked for particular sectors<br />
so they expect the impact to be widespread.<br />
“[Innovation Union] will generate more R&D,<br />
more startups and more major EU companies<br />
with the potential to be international players,”<br />
says English.<br />
Nidhi Subbaraman<br />
Airlines ahead on algae<br />
Plans to grow algal biomass in airport grounds<br />
for aircraft fuel are moving forward supported<br />
by Airbus, British Airways and Cranfield<br />
University. On September 16, European aircraft<br />
manufacturer Airbus of Blagnac, France,<br />
London-based British Airways, and Gatwick<br />
airport in Surrey, announced a collaboration<br />
with Cranfield University, UK, through the<br />
Sustainable Use of Renewable Fuels (SURF)<br />
consortium. For the past three years, Cranfield<br />
University scientists have been researching<br />
algal-derived biofuel for use in aviation. Now,<br />
with their partners at SURF, they are working<br />
to scale up from their pilot plant (a thousand<br />
gallons per batch) for commercial output. Unlike<br />
the automobile industry, the aviation sector<br />
lacks the option to use electric alternatives for<br />
energy, says Feargal Brennan, head of Cranfield<br />
University’s Department of Offshore, Process and<br />
Energy Engineering. “For the foreseeable future<br />
the aviation industry will depend on biofuels, so<br />
they have to take a lead in commercial biofuel<br />
replacements,” Brennan says. Paul Nash,<br />
Airbus’ head of New Energies and Environmental<br />
Affairs notes that the firm is collaborating on<br />
biofuel projects with research groups from Brazil<br />
and Qatar. Later this year Brazilian airline TAM<br />
will test a fuel mixture of bio-kerosene derived<br />
from the native jatropha plant in an Airbus<br />
aircraft. “What we’re finding today is that the<br />
industry is moving faster [through partnerships]<br />
than R&D or governments would do,” says Nash.<br />
“As an industry, we can say: ‘our aircrafts are<br />
ready for this’.”<br />
Nidhi Subbaraman<br />
Total transaction value (millions)<br />
$1,500<br />
$1,200<br />
$900<br />
$600<br />
$300<br />
$0<br />
88.0 112.8<br />
2004<br />
2005<br />
788.7<br />
2006<br />
Year<br />
1,378.3<br />
2007<br />
1,190.9<br />
2008<br />
876.7<br />
2009<br />
Figure 1 Investment into industrial biotech. After<br />
cresting in 2007, total investment into biofuels<br />
and bio-based materials has been in decline.<br />
labeling on all fuel and fuel sources to entice<br />
consumers by letting them know where their<br />
fuel is coming from.<br />
“You have labels on individual pieces of<br />
fruit to tell you where they came from,” says<br />
Glenn Nedwin of Genencor, in Rochester,<br />
New York, a division of Danish Danisco, and<br />
one of the biggest developers of industrial<br />
enzymes in the world. “But you don’t know<br />
where fuel is coming from. If [the pump] said<br />
it was coming from Venezuela, Saudi Arabia<br />
or Iowa, let the consumer make the choice.”<br />
In fact, much of the current biofuels market<br />
is being propped up by government<br />
decrees and subsidies. The EPA’s Renewable<br />
Fuel Standard (RFS) mandate of 2007<br />
requires that 36 billion gallons of biofuel<br />
be blended with gasoline tanks by the year<br />
2022. Cellulosic ethanol itself, the mandate<br />
says, will furnish 16% of that, with the rest<br />
supplied by corn ethanol and other secondgeneration<br />
biofuels such as biodiesel.<br />
That RFS mandate has helped drive<br />
demand, and experts like Nedwin believe an<br />
aggressive infrastructural overhaul is neces-<br />
in their words<br />
“There’s no scientist out<br />
there that has been wary<br />
of me, and guys in the<br />
venture-capital world I<br />
dealt with before—it’s<br />
as if I wasn’t gone a<br />
day!” Ex-convict Sam<br />
Waksal, former ImClone<br />
Systems CEO, is out<br />
of jail and back in<br />
business, raising $50<br />
million to acquire hepatitis C treatment maker<br />
Three Rivers Pharmaceuticals. (Pharmalot, 26<br />
October 2010)<br />
“The pendulum has swung too far at the FDA if<br />
they influence Advisory Committees to vote no for<br />
a drug that poses very little risk to the public at<br />
large…and where there is a demonstrable clinical<br />
sary to eventually cement biofuels’ place in<br />
everyday life. Growth Energy has a ‘Fueling<br />
Freedom’ plan designed to do just that, calling<br />
for additional tax credits for retailers to<br />
build 200,000 blender pumps (which allow<br />
consumers to regulate the percentage of<br />
ethanol they put in their car using a mixture<br />
of gasoline and E85), a loan guarantee<br />
for an ethanol pipeline and flex-fuel vehicles<br />
to be released on the roads. Those moves<br />
would require more involvement by the government,<br />
but there is still another concern:<br />
price. Without subsidies, gasoline remains<br />
the cheapest fuel drivers can put into their<br />
cars. “We’re seeing technologies that are projecting<br />
the ability to be competitive [with<br />
gasoline],” says Flagship Ventures’ Berry,<br />
“But…people aren’t ready to be cost competitive<br />
without subsidy yet.”<br />
The unknowns with consumers and price<br />
explain why the move to a 15% blend falls<br />
short of making cellulosic producers feel the<br />
golden years are just ahead. “I think this is a<br />
positive first step and investors will look at<br />
this, see it as a good sign. [But i]t doesn’t add<br />
enough gallons to where we ultimately need<br />
to get to,” says Wes Bolsen, chief marketing<br />
officer and vice president, government affairs<br />
at Coskata, of Warrenville, Illinois.<br />
As for the next steps, Sturdevant would<br />
like to see the limit moved to 20%, 30%<br />
and higher. “I believe that the scientific<br />
data supports moving more quickly now to<br />
higher blends. I believe that the EPA and<br />
[Department of Energy] will see that, and<br />
they’ll see that there’s no harm to engines at<br />
higher blends, and I can only hope that they<br />
will move faster in the future.”<br />
Nidhi Subbaraman, New York<br />
benefit in terms of additional weight loss and<br />
reduction in cardiovascular risk factors.” A petition<br />
from ‘Citizens for Lorcaserin’ to FDA Commissioner<br />
Margaret Hamburg protesting the 16 September<br />
advisory committee decision to turn down approval<br />
of Arena Pharmaceuticals diet drug.<br />
“There are things I know something about, and<br />
things I know little about. The stock market is one<br />
of the latter.” Sangamo CEO Ed Lanphier admits,<br />
following a 52-week slide in the company’s stock<br />
value. (Xconomy, 18 October 2010)<br />
“We are left relying on 20th century approaches<br />
for the cures of the 21st century.” FDA<br />
Commissioner Margaret Hamburg argues for a<br />
move away from randomized controlled trials<br />
to testing drugs in combination and using<br />
biomarkers. (Financial Times, 15 October 2010)<br />
1230 volume 28 number 12 december 2010 nature biotechnology
NEWS feature<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Crossing the line<br />
The list of drug companies forced into several hundred million<br />
dollar settlements for making fraudulent product claims continues<br />
to lengthen. And the signs are that the US government will<br />
continue to ramp up its efforts, using new theories of liability and<br />
handing out even stiffer penalties. Mark Ratner investigates.<br />
It has been more than a year since the US<br />
Department of Justice (DoJ) announced its<br />
$2.3 billion settlement with New York–based<br />
Pfizer for Pfizer’s illegal, off-label marketing<br />
of four drugs—the largest fraud suit to hit the<br />
pharmaceutical industry (Nat. Biotechnol. 27,<br />
961–962, 2009). And the beat goes on, with<br />
several other high-profile fraud settlements<br />
coming to light (Table 1).<br />
Over the past five years, there has been<br />
no letup in government investigations<br />
of the marketing and pricing practices<br />
of pharma, biotech and medical device<br />
companies. These activities are being<br />
heavily scrutinized on many fronts—by<br />
the US Food and Drug Administration,<br />
DoJ, US Congress and even investors.<br />
But whereas industry watchers consider<br />
these ‘plain vanilla’ off-label suits as ‘legacy<br />
conduct’—part of a bygone era—it is<br />
unclear whether the government’s and<br />
industry’s perspectives on this are truly<br />
aligned. Indeed, the government is now<br />
focusing on new theories of liability.<br />
“There is a very careful mining of<br />
operations at pharmaceutical companies<br />
to find cases,” says T. Reed Stephens,<br />
a former DoJ lawyer now with the<br />
Washington, DC, law firm McDermott<br />
Will & Emery. “The government now<br />
has the resources available to it to evaluate<br />
a higher number of new cases,” and<br />
to find cases that are consistent with the<br />
dual policy goals of deterring inappropriate<br />
conduct and maintaining patient<br />
safety, he says. By most accounts, there<br />
may be 150 or more cases still in the<br />
prosecutorial pipeline.<br />
Filling government coffers<br />
For every dollar the government puts into<br />
healthcare fraud enforcement, it gets back<br />
more than fifteen times that amount, notes<br />
Shelley Slade, a former DoJ attorney now<br />
representing corporate whistle-blowers<br />
through the Washington, DC, law firm of<br />
Vogel, Slade & Goldstein. Government press<br />
releases tout the amounts: according to DoJ, it<br />
secured more than $2.5 billion in settlements<br />
and judgments in False Claims Act (FCA)<br />
matters alleging healthcare fraud in the fiscal<br />
year ending September 30, 2010. This<br />
was more than ever before obtained in a<br />
single year and represents a 66% increase<br />
over fiscal year 2009, in which $1.68 billion<br />
was obtained. Just after the Pfizer settlement<br />
in September 2009, recoveries over the last<br />
15 years in the federal judicial district of<br />
Although Pfizer’s multi-billion dollar 2009 settlement<br />
remains the poster child for healthcare fraud, significant<br />
settlements continued to pile up this year including a<br />
half-billion dollars or more each by drug makers Novartis,<br />
AstraZeneca and Allergan. (© Novartis AG)<br />
Massachusetts alone, where that suit was<br />
prosecuted, had reached $6.8 billion. “It’s a<br />
no-brainer in terms of financial investment<br />
by the government,” says Slade.<br />
Whistle-blowers also have a vested interested<br />
in finding new theories they can file on,<br />
Stephens points out, as do the attorneys who<br />
specialize in whistle-blower cases. The private<br />
incentive alone—whistle-blowers may receive<br />
upwards of 15% of any amounts recovered in<br />
claims by the federal government—is enough<br />
to assure a steady flow of informants and creative<br />
new theories of potential liability.<br />
On the other hand, if whistle-blowing is effective<br />
at rooting out fraud scenarios, it is a relatively<br />
inefficient way to protect patient interests.<br />
“I think it’s specious for someone to go to the<br />
government because they think something is<br />
dangerous,” says Retta Riordan, an independent<br />
compliance consultant. “We know how long it<br />
takes for the government to develop a case—<br />
four to six years. If you really want the activity<br />
to stop you need to do that immediately and the<br />
best way to do that is in-house.”<br />
To that end, almost every fraud settlement<br />
is accompanied with a corporate integrity<br />
agreement (CIA) between the company<br />
and the government—specifically the US<br />
Department of Health & Human Services<br />
(HHS) Office of the Inspector General<br />
(OIG). CIAs mandate a set of procedures and<br />
reporting responsibilities to monitor compliance<br />
and prevent similar conduct from<br />
recurring. CIAs also help ensure that<br />
internal processes exist to bring future<br />
bad conduct to light (Box 1).<br />
Medical affairs or promotion?<br />
Recent cases and the resulting terms of the<br />
CIAs show that government officials have<br />
been focusing on fraud arising from medical<br />
affairs operations, both internally and<br />
externally. In many situations, the misconduct<br />
has centered on some form of kickback.<br />
“It’s a somewhat consistent theme<br />
in our investigations,” says Mary Riordan,<br />
senior counsel in the HHS OIG. But the<br />
form of those kickbacks has changed over<br />
time. “I think we see less blatant cash-tophysicians<br />
type[s] of kickbacks,” she says.<br />
“Now the type of activity we are looking<br />
at is more sophisticated—contractual<br />
arrangements between doctors and the<br />
drug companies who engage them.”<br />
The recent AstraZeneca settlement,<br />
for one, involved clinical trials activities,<br />
publications and grants in addition to<br />
inappropriate product promotion. The<br />
Pfizer case, which focused largely on the<br />
promotion of their pain medication Bextra<br />
(valdecoxib), also had a medical affairs–<br />
publications component. These areas had not<br />
traditionally been part of the government’s<br />
focus. Companies are also now being investigated<br />
more often for whether they may have<br />
retained clinicians to conduct ‘sham’ clinical trials—essentially<br />
another form of kickback. Such<br />
a claim was made against Boston Scientific’s<br />
Guidant subsidiary, in Indianapolis, as part of a<br />
$22 million December 2009 settlement in a case<br />
that included the implantation of pacemakers<br />
and defibrillators in clinical trials.<br />
1232 volume 28 number 12 DECEMBER 2010 nature biotechnology
news feature<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Table 1 Recent settlements of healthcare fraud cases<br />
Company (date announced)<br />
Novartis<br />
(Basel) (September 2010)<br />
Allergan<br />
(Irvine, California) (September 2010)<br />
Forest Laboratories<br />
(New York) (September 2010)<br />
Novartis<br />
(May 2010)<br />
Ortho-McNeil-Janssen<br />
(Titusville, New Jersey) (April 2010)<br />
Schwarz Pharma<br />
(UCB, Brussels) (April 2010)<br />
AstraZeneca<br />
(London) (April 2010)<br />
Alpharma<br />
(King Pharmaceuticals,<br />
Bristol, Tennessee) (March 2010)<br />
Boston Scientific<br />
(Guidant division, Indianapolis)<br />
(December 2009)<br />
Biovail<br />
(Valeant, Mississauga, Canada)<br />
(September 2009)<br />
Elan<br />
(Dublin) (settlement pending)<br />
Teva<br />
(Jerusalem) (settlement pending)<br />
There’s also been more focus on the intersection<br />
of medical affairs and commercial<br />
functions within companies—the grant evaluation<br />
process, how publications are developed<br />
and whether they are being used as disguised<br />
promotion. An abundance of ten-patient studies<br />
in off-label areas could be seen as geared<br />
merely to generate interest in providers. The<br />
government also is concerned that physicians<br />
are brought on to company boards so that these<br />
key opinion leaders can be influenced favorably<br />
toward in-house products and promote<br />
the message in the community, rather than<br />
using the physicians in their advisory capacity.<br />
Or if a speakers’ bureau brings 500 speakers<br />
to a nice hotel for training but only uses 100<br />
of them, it would in effect be a promotional<br />
activity. “I think our analysis of the situations<br />
is becoming more complex, as are the situations<br />
we are seeing,” says Mary Riordan.<br />
Within a company, even the seemingly<br />
simple act of responding to an information<br />
request can blur the line between medical and<br />
commercial activities and put a company in a<br />
difficult situation. The Frazer, Pennsylvania–<br />
based biotech Cephalon gets thousands of such<br />
requests every quarter, says chief compliance<br />
officer Valli Baldassano. “We get all kinds of<br />
requests, not just for off-label uses—about the<br />
Amount<br />
($ millions) Issues settled<br />
422.5 Criminal and civil charges relating to the off-label marketing of Trileptal (oxcarbazepine) and five other drugs in<br />
2000 and 2001. Whistle-blowers, all former employees of Novartis, receive over $25 million from the settlement.<br />
600.0 Criminal and civil charges relating to the off-label marketing of Botox (onabotulinumtoxinA) in 2003.<br />
Whistle-blowers receive $37.8 million.<br />
313.0 Criminal charges (including obstruction of justice) around the distribution of an unapproved drug, Levothroid<br />
(levothyroxine), civil liabilities relating to the off-label marketing of Celexa (citalopram) and fraud in the pricing<br />
of those drugs as well as Lexapro (escitalopram). Conduct occurred between 2001 and 2003. Whistle-blowers<br />
receive ~$14 million.<br />
72.5 Civil liabilities relating to the off-label marketing of inhaled tobramycin between 2001 and 2006. Whistleblowers,<br />
all former employees of Chiron (now part of Novartis), receive $7.8 million.<br />
81.0 Criminal and civil charges relating to the off-label marketing of Topamax (topiramate). Whistle-blowers receive<br />
more than $9 million.<br />
22.0 Civil charges relating to the billing of two unapproved drugs to the Centers for Medicare and Medicaid Services.<br />
Whistle-blowers receive more than $1.8 million.<br />
520.0 Civil charges relating to the off-label marketing of Seroquel (quetiapine) from 2001 to 2006. The whistle-blower<br />
receives more than $45 million.<br />
42.5 Civil charges relating to bribes paid to healthcare providers to induce them to promote or prescribe Kadian<br />
(morphine sulfate), and misrepresentations about the drug’s safety and efficacy, from 2000 to 2006.<br />
The whistle-blower receives $5.33 million.<br />
22.0 Civil charges that its subsidiary, Guidant, used post-market studies as vehicles to funnel kickbacks to physicians<br />
to induce them to use the company’s pacemakers and defibrillators.<br />
24.6 Guilty plea on criminal conspiracy and kickback charges relating to its drug Cardizen (diltiazem) beginning in<br />
2003. Amount includes a $2.4 million civil fine.<br />
203.5 Company has announced an agreement with the US Attorney’s Office in Massachusetts to settle claims around<br />
the marketing of Zongran (zonisamide).<br />
169.0 Texas Attorney General’s Office announced an agreement with the company to settle claims relating to improper<br />
price reporting for its drugs. Whistle-blower Ven-A-Care triggered the investigation.<br />
Source: Press releases from the US Department of Justice; US Attorney’s Office District of Massachusetts; and Attorney General of Texas.<br />
science of the drugs, their pharmacokinetics,<br />
their uses and what data are available for both<br />
on- and off-label uses. They can come in<br />
through a phone call, e-mail or because a doctor<br />
has asked a sales rep, which is how many of<br />
the sensational cases arise.”<br />
Follow the money<br />
By dint of its settlement over off-label promotion<br />
of its narcolepsy drug Provigil (modafinil),<br />
which included a CIA, Cephalon has helped<br />
set the industry standard for the processes for<br />
tracking spending on speakers, consultants<br />
and the like—a focus of DoJ’s inquiry into the<br />
company and also an element in the Physicians<br />
Payment Sunshine Act of 2009, now incorporated<br />
into the recent federal healthcare reform<br />
legislation, which takes effect for all companies<br />
in 2013. ProPublica now posts these payments<br />
by Cephalon on its website, along with those<br />
from the six large drug companies that have<br />
also made them public (http://projects.propublica.org/docdollars/).<br />
The mere disclosure of payments is an area<br />
where consumers, providers, competitors<br />
and the public at large can see if there are particular<br />
trends or payments. “With ‘Sunshine,’<br />
presumably comes a desire to reduce potential<br />
conflicts of interest,” says Howard Young, an<br />
attorney with Morgan Lewis in Washington,<br />
DC. (Academic institutions, for their part, are<br />
also cracking down on this potential conflict<br />
of interest; Harvard University, for example,<br />
is implementing a policy starting January 1,<br />
2011, prohibiting faculty from giving promotional<br />
talks for drug and device makers or<br />
receiving gifts, travel stipends and meals.)<br />
Pharmaceutical companies pay doctors to<br />
do services for them under fee-for-service<br />
arrangements. These can include speaker<br />
programs, promotional talks or participation<br />
advisory boards, including commercially oriented<br />
boards that brainstorm marketing messages.<br />
There’s also a bucket of spending for fees<br />
around research participation, including conducting<br />
clinical trials. What’s more, there are<br />
incidental expenses for travel and meals.<br />
“The popular thought is that pharma companies<br />
can push a button and miraculously<br />
every dollar you paid to Dr. X shows up in<br />
a database,” says Baldassano. “That’s just not<br />
true. It’s a huge undertaking and different for<br />
every company.” When the government wanted<br />
Cephalon to calculate aggregate expenditure as<br />
part of its CIA, the company balked. “We said<br />
we’ve been trying to build it, have some components<br />
of it for state reporting, but we don’t have<br />
it the way you are envisioning it,” she says. “So<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1233
NEWS feature<br />
© 2010 Nature America, Inc. All rights reserved.<br />
you need to give us time to build it.” As a result,<br />
the aggregate expenditure system in the CIA is<br />
staged, with various deadlines.<br />
For example, it’s easy to capture a consulting<br />
fee paid for services; there’s usually a contract<br />
and a direct payment. Cephalon is already<br />
posting those payments publicly and every<br />
quarter going forward. But other spending—<br />
for a pizza or sandwich, and for travel fees<br />
ancillary to research services—may come<br />
through an expense report system or a logistics<br />
vendor. “It doesn’t sound like it may be complicated,”<br />
says Baldassano. “But it is, to make<br />
sure you capture it all accurately.”<br />
Trend spotting<br />
A close examination of the provisions in<br />
CIAs can help other companies determine if<br />
their own processes represent the best practices<br />
within the industry, and whether they<br />
should be adopting other measures. “From<br />
an educational perspective, they’ve had that<br />
kind of broad-based effect on the industry,”<br />
says Young.<br />
Box 1 Getting the house in order<br />
Many pharma and biotech companies have responded to increased<br />
government scrutiny of their sales and marketing practices by<br />
adding substantially more training of staff with respect to off-label<br />
promotion. “We are seeing a real effect,” says Howard Young, an<br />
attorney with Morgan Lewis in Washington, DC, and former senior<br />
attorney and deputy branch chief with the Office of the Inspector<br />
General (OIG) of the Department of Health & Human Services.<br />
“The prosecutors themselves have stated that the industry appears<br />
to be moving in the right direction,” he notes.<br />
“I think the amount of money being spent and the number<br />
of FTEs [full-time employees] being devoted toward regulatory<br />
compliance over the past five to ten years reflects a real<br />
understanding of most of the industry that these are obligations<br />
that are here to stay and they are being institutionalized in a<br />
meaningful way,” adds T. Reed Stephens, a former DoJ lawyer now<br />
with the Washington, DC, law firm McDermott Will & Emery.<br />
OIG came out with a guidance for pharmaceutical companies on<br />
compliance practices in 2003, to prevent and reduce fraud and<br />
abuse in federal healthcare programs (http://oig.hhs.gov/authorities/<br />
docs/03/050503FRCPGPharmac.pdf). Since then, most large<br />
pharma companies have established formal programs, as have most<br />
mid-sized biotech companies and even many smaller companies.<br />
For companies that have run afoul of the law, OIG requires—almost<br />
without exception—that those companies sign a CIA mandating<br />
the establishment of compliance procedures aimed at preventing<br />
the prior contentious conduct from recurring. This is not unique to<br />
the healthcare industry: Any company involved with civil or even<br />
criminal claims is going to be entering into a CIA whether in the<br />
insurance business, the defense business or pharma.<br />
“At that point, DoJ has taken care of your sins of the past<br />
and OIG is there to make sure you fly right in the future,” says<br />
Thomas Gunning, general counsel of EMD Serono in Rockland,<br />
Massachusetts, which settled a case involving off-label promotion<br />
of its human growth hormone, Serostim, with DoJ in 2005.<br />
“Every CIA is an opportunity to look at new<br />
provisions and see what the government is focusing<br />
on,” adds Retta Riordan. “It’s an opportunity<br />
to see what went wrong. Was it a lack of knowledge?<br />
A lack of processes in place—policies and<br />
procedures? A lack of training? And use that as<br />
a tool to reassess an existing program.” When<br />
Indianapolis-based Eli Lilly, for example, settled<br />
a case with DoJ in 2005 over allegations it had<br />
been promoting its osteoporosis drug Evista<br />
(raloxifene) for three indications where it was<br />
only approved for one, Riordan pondered how<br />
it could have happened, “especially in a company<br />
the size of Lilly that has a very good compliance<br />
program,” she says.<br />
There was evidence that Lilly had developed<br />
an internal best practices video emphasizing<br />
the off-label message. Yet although all companies,<br />
including Lilly, review material going<br />
out to the public, as it turned out, Lilly’s promotional<br />
materials review committee was not<br />
reviewing internal training materials.<br />
Experienced whistle-blower attorneys will<br />
focus on these kinds of situations: products<br />
that have been on the market for a long time<br />
and have a track record of sales and profits<br />
that will potentially constitute the measure of<br />
damages to gauge the expected payoff from<br />
the case, says McDermott Will & Emery’s<br />
Stephens. “You have to be looking at cases<br />
where there’s been a systematic approach to<br />
a claim backed by the organization and it is a<br />
strategy they have undertaken. The cases with<br />
the greatest exposure will be those where the<br />
company took a sustained approach and where<br />
the plaintiff ’s counsel can at least theorize a<br />
substantial amount—tens of millions or hundreds<br />
of millions of revenue and on top of that<br />
a proportionate amount of profit that involves<br />
federal dollars.”<br />
Warning letters to companies from FDA,<br />
many of them directed at communications to<br />
medical professionals about their products<br />
that do not adequately discuss the risks of<br />
therapies, may also tip off a whistle-blower<br />
attorney to possible fertile ground. They<br />
could even foreshadow a future DoJ investigation—or<br />
a lawsuit by competitors—as DoJ<br />
In virtually every case over the past six years, CIAs have<br />
included provisions forcing a separation of the compliance<br />
function from the legal department, in effect instituting another<br />
set of checks and balances. CIAs also call for the retention of an<br />
outside independent review organization, which performs annual<br />
audits of compliance processes and training, and reports to the<br />
company as a financial auditor would. In cases involving offlabel<br />
promotion, a CIA would likely have provisions that require<br />
the company to get ‘verbatims’ from doctors—an independent<br />
review of what was said in face-to-face sessions with sales<br />
reps. If a case is focused on kickbacks, the CIA might focus<br />
on how consultants and speakers are hired and the exact use<br />
of those monies. CIAs and investigations that focused on best<br />
pricing issues, rebates and whether Medicare and Medicaid<br />
bill the right price would get deep into the internal policies and<br />
practices for price reporting.<br />
Compliance may be “a further opportunity to embed, from a<br />
business standpoint, the entire area of process improvement,”<br />
notes Paul Silver, managing director of The Huron Group, a<br />
consultancy in Atlanta that often acts as an independent review<br />
organization. For example, after Cephalon’s settlement of claims<br />
that the company promoted its narcolepsy drug Provigil (modafinil)<br />
off-label to treat fatigue associated with multiple sclerosis,<br />
it essentially “redesigned how we did business,” says chief<br />
compliance officer Valli Baldassano.<br />
Compliance systems also help to identify areas of abuse<br />
proactively. “We’ve made it into a much broader risk management<br />
function that is much more engrained into the business,”<br />
Baldassano says. “If you can design a good way to slice and dice<br />
the data, you may be able to find the trend you are looking for to<br />
see if you’ve got a problem out there. Then you can either do more<br />
investigation, stronger training, some kind of intervention that can<br />
affect that behavior. You have to have the system to smoke out the<br />
behavior, but the system itself cannot be the panacea.”<br />
1234 volume 28 number 12 DECEMBER 2010 nature biotechnology
news feature<br />
© 2010 Nature America, Inc. All rights reserved.<br />
closely examines marketing claims that are<br />
regulated by FDA to determine whether they<br />
also constitute off-label violations.<br />
Expanding on the claims beyond what are<br />
in FDA-approved statements is a little bit different<br />
from off-label promotion, Stephens<br />
says. It’s typically an FDA-focused issue but<br />
also raises the specter of making comparative<br />
claims about a product that run afoul of the<br />
federal civil FCA. The FCA is the principal<br />
mechanism for pursuing claims of off-label<br />
promotion and, in cases where the government<br />
is the customer, kickbacks and illicit pricing<br />
schemes, including rebates to Medicare. The<br />
so-called qui tam provisions of FCA also keep<br />
in place a structure for whistle-blowers to<br />
come forward and alert the government to<br />
some of those schemes. And once it becomes<br />
known that a company is under investigation,<br />
many states then start class actions.<br />
Because of the huge advertising potential of<br />
social media—Twitter, Facebook, Wikipedia<br />
and the like—their use is bound to be fertile<br />
ground for pharmaceutical companies to<br />
get out their message, and for investigation.<br />
“It’s going to create some real challenges for<br />
pharmaceutical companies to track where<br />
their content goes,” says Stephens. FDA is<br />
already working on guidelines for how the<br />
pharmaceutical industry can use social media<br />
(Nat. Biotechnol. 28, 641–643, 2010). But the<br />
mere fact that FDA is going to regulate those<br />
types of interactions means there are going<br />
to be opportunities for mistakes to be made<br />
and for companies to be overly aggressive,<br />
before guidelines come into place, according<br />
to Stephens.<br />
In addition, companies operating under<br />
CIAs are required to submit marketing and<br />
promotional plans to the government ahead of<br />
time. “That’s one of the most exacting aspects<br />
of these CIAs, giving the government the ability<br />
to preview what the company is planning<br />
on doing,” Stephens says. “DoJ could make<br />
the allegation, it’s not just an FDA violation<br />
but that you are promoting off-label by saying<br />
something lacking a scientific basis,” he<br />
says. “[In addition], they could solicit whistle-blowers<br />
from within that company—they<br />
would explain what the theory is and hope<br />
they could bring to them factual allegations<br />
they could attach to their theory and quickly<br />
file a case,” Stephens says.<br />
Criminal prosecutions<br />
According to Shelley Slade, government<br />
officials are coming to the conclusion that<br />
although the FCA “is somewhat of a deterrent,<br />
it’s not enough.” Increasingly, DoJ is turning<br />
to the Foreign Corrupt Practices Act (FCPA),<br />
which gives it some jurisdiction over conduct<br />
outside the US. For example, this August,<br />
Merck disclosed that it is being investigated<br />
for violations under the FCPA, which DoJ<br />
had used to penalize Novo Nordisk in 2009,<br />
for kickbacks paid to the former Iraqi government<br />
as an incentive to obtain insulin supply<br />
contracts. FCPA-related investigations are also<br />
ongoing for several other pharmas.<br />
But the real big stick will come in the form<br />
of criminal prosecutions of individuals spearheading<br />
these schemes, Slade says. Indeed, a<br />
trend over the past several years has been for<br />
DoJ and OIG to assess individual liability and<br />
responsibility for fraudulent schemes, along<br />
with requiring boards and senior management<br />
to be more active in compliance programs.<br />
As with any industry, compliance units<br />
within pharma and biotech companies come<br />
up against the fact that people are rewarded<br />
for improving a firm’s financial performance—<br />
right up the chain to the CEO, Slade says.<br />
“Their concerns are pleasing Wall Street and<br />
shareholders, and bonuses are driven by the<br />
financial performance. There’s a real conflict<br />
between what the compliance folks want them<br />
to do and what’s in their immediate financial<br />
self-interest. Sometimes the compliance people<br />
don’t get the full story from the people in<br />
operations.” Thus, despite the best of intentions,<br />
prosecutors know that when they speak<br />
at compliance conferences they are basically<br />
preaching to the choir.<br />
“Holding individuals accountable, and<br />
criminal settlements—the whole combination<br />
of steps—will lead to better compliance,”<br />
says OIG’s Mary Riordan. “We are increasingly<br />
looking at ways of holding individuals<br />
accountable,” she says, both through CIAs<br />
and also through so-called exclusion actions<br />
against individuals, which would bar them<br />
from participating in the industry. OIG currently<br />
has an initiative underway to target<br />
the responsible corporate executives through<br />
exclusion actions of individuals— individuals<br />
they believe were in positions to prevent<br />
alleged fraudulent conduct or should have<br />
been aware of it and stopped it.<br />
The most dramatic case in point has been<br />
the criminal conviction (now under appeal)<br />
of three individuals from Purdue Pharma<br />
in Stamford, Connecticut, who were found<br />
responsible for not preventing deceptive marketing<br />
practices for the narcotic Oxycontin<br />
(oxycodone). And on November 8, DoJ<br />
indicted a former attorney for London-based<br />
GlaxoSmithKline (GSK, Brentford, UK) charging<br />
her with obstruction of justice and making<br />
false statements to FDA in connection with<br />
an investigation into GSK’s promotion of the<br />
antidepressant Welbutrin (buproprion) early<br />
in the decade. That case is still pending, but in<br />
the fourth quarter of 2008, GSK set aside $400<br />
million, anticipating a settlement.<br />
To date, the Purdue case is the only situation<br />
where OIG has exercised its authority in<br />
this way. But there’s whispering among attorneys<br />
that more such cases are coming down<br />
the pike. Given an appropriate case with clear<br />
criminal violations and egregious circumstances<br />
where patient health is being compromised<br />
by the off-label use, and executives well<br />
aware of what the sales force has been doing<br />
and of the risk of harm to patients, “the only<br />
way the executives’ behavior is going to change<br />
is if the government treats these as criminal<br />
violations and goes after the executives,” says<br />
Slade. “If they see their peers going to jail, it<br />
will offset the financial calculus for them.”<br />
OIG could also exclude an entire company<br />
from the industry, but that’s a ‘nuclear bomb’<br />
scenario that would raise serious issues of<br />
patients’ rights and access to drugs, as well as<br />
antitrust.<br />
Too early to know<br />
In the meantime, the number of fines and<br />
prosecutions continues to increase. And<br />
industry is devoting more and more resources<br />
to ensure that it does not fall foul of government<br />
officials. “We know that companies are<br />
absolutely continuing to ramp up their compliance<br />
efforts and staffing up their compliance<br />
groups internally,” says New York attorney<br />
Arnold Friede. “They are hiring more people<br />
and propagating their compliance policies<br />
more forcefully internally. But whether that’s<br />
accompanied by a cultural shift, particularly in<br />
an era of organizational uncertainty, is an open<br />
question. Process and culture are not necessarily<br />
coincident, and it probably takes an equal<br />
measure of both to make sure that you are<br />
working towards effective compliance.”<br />
Overall, “it’s too early to tell” whether the<br />
recent ramp-up of enforcement activities will<br />
have the desired deterrent effect, says Jerry<br />
Avorn, professor of medicine at Harvard<br />
Medical School and the author of Powerful<br />
Medicines, one of the first books to examine<br />
pharmaceutical detailing practices—the oneon-one<br />
conversations between sales reps and<br />
physicians—in detail.<br />
“We don’t have a good objective way to measure<br />
changes in corporate behavior, but we certainly<br />
hope the CIAs are an effective tool,” says<br />
HHS’s Mary Riordan.<br />
Mark Ratner, Cambridge, Massachusetts<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1235
NEWS feature<br />
© 2010 Nature America, Inc. All rights reserved.<br />
New twists on proteasome inhibitors<br />
Only one drug targeting the proteasome protein degradation<br />
pathway has been approved, but several second-generation<br />
inhibitors are making progress in trials. Jim Kling reports.<br />
In July, Onyx Pharmaceuticals of Emeryville,<br />
California, announced positive phase 2 results<br />
for its proteasome inhibitor carfilzomib, which<br />
is currently being tested in multiple myeloma.<br />
The drug achieved a 24% response rate in individuals<br />
with advanced cancer, who had failed<br />
on all prior therapies, more than double the<br />
expected response rate (11%). This was not<br />
only good news for the company—whose stock<br />
jumped 21% and which clinched a marketing<br />
deal worth >$300 million with Japanese firm<br />
Ono Pharmaceuticals of Osada—but also a<br />
boost for the field of proteasome inhibition,<br />
which thus far has only a single success story,<br />
Velcade (bortezomib), which was first approved<br />
in multiple myeloma in 2003.<br />
Carfilzomib, however, is one of several<br />
proteasome inhibitors currently progressing<br />
through the clinic. Many of the compounds<br />
causing the most buzz are those that target<br />
another part of the cell’s proteolytic machinery,<br />
the ubiquitin ligases, which number in the<br />
hundreds (Fig. 1) and could potentially provide<br />
greater clinical specificity in the clinic.<br />
The trailblazer<br />
First approved as a third-line therapy for<br />
refractory multiple myeloma, Velcade was<br />
extended to mantle cell lymphoma in 2006<br />
and became a frontline treatment for multiple<br />
myeloma in the US in 2008. An injectable<br />
tripeptide (modified with a boronic acid at<br />
one end, Pyz-Phe-boroLeu), it binds the catalytic<br />
site of the 26S proteasome (Box 1). The<br />
mechanism of action, with respect to multiple<br />
myeloma, is straightforward: myeloma<br />
cells are essentially protein factories, having<br />
descended from B lymphocytes, which create<br />
antibodies. “They fail to turn off their<br />
protein production machinery, causing proteotoxic<br />
stress. Cells deal with it by upregulating<br />
the proteasome. So the cell is gummed<br />
up with proteins anyway, and then we shut<br />
off the primary mechanism by which they’re<br />
getting rid of the stress. That’s why Velcade<br />
is effective,” says Joe Bolen, chief scientific<br />
officer of Millennium (now part of Takeda in<br />
Osaka), the company that brought the drug<br />
to market.<br />
Sensitivity of tumor cells to proteasome<br />
inhibitors has been documented 1 , and such<br />
observations have prompted companies to<br />
consider them for treating other cancers,<br />
particularly solid tumors. But Velcade turns<br />
out to be not all that precise in its targeting. In<br />
disrupting the proteasome, it affects the breakdown<br />
of a range of proteins and interferes with<br />
some proteases, which is thought to be associated<br />
with the observed neurotoxicity.<br />
The next generation<br />
Velcade is an unqualified success in the marketplace—the<br />
drug generates about $1.5 billion<br />
in revenues worldwide annually, according<br />
to David Moskowitz, who follows Onyx in<br />
his role as managing director with Madison<br />
Williams investment bank in New York. But<br />
its limitations, which include the development<br />
of resistance to it in addition to neurotoxicity,<br />
have inspired the development of a second<br />
generation of proteasome inhibitors, including<br />
compounds like carfilzomib. Carfilzomib is an<br />
irreversible epoxomicin-related peptide that<br />
binds specifically the chymotrypsin-like protease<br />
activity of the proteasome, hitting fewer<br />
nontarget proteases than Velcade, which may<br />
account for its reduced toxicity.<br />
Nereus Pharmaceuticals in San Diego is<br />
developing an entirely different class of compounds<br />
represented by NPI-0052 (salinosporamide<br />
A), a β-lactam isolated from the<br />
marine bacterium Salinispora tropica. NPI-<br />
0052 blocks all three catalytic activities—<br />
chymotrypsin-like hydrolysis, trypsin-like<br />
hydrolysis and peptidylglutamyl peptide<br />
hydrolysis—whereas Velcade inhibits just<br />
the chymotrypsin-like activity. NPI-0052 has<br />
Box 1 A protein degrading machine<br />
shown activity against multiple myeloma cells<br />
resistant to Velcade, and in animal models it<br />
inhibits colon, pancreatic and lung cancers.<br />
The drug is currently in phase 1 clinical trials<br />
for multiple myeloma and lymphoma, as a<br />
single agent and in combinations (Table 1).<br />
In addition to carfilzomib, which entered<br />
the Onyx pipeline with its 2009 acquisition of<br />
the biotech company Proteolix, Onyx is developing<br />
two other Proteolix compounds. ONX<br />
0912, an oral proteasome inhibitor, is currently<br />
in phase 1 clinical trials for advanced refractory<br />
or recurring solid tumors. ONX 0914<br />
inhibits the related immunoproteasome, a<br />
specialized type of proteasome mainly found<br />
in monocytes and lymphocytes. In preclinical<br />
models, it was shown to be 10- to 50-fold more<br />
selective for the immunoproteasome than for<br />
the general proteasome. Possible indications<br />
include rheumatoid arthritis, inflammatory<br />
bowel disease and lupus.<br />
Michael Kauffman, Onyx’s chief medical<br />
officer, thinks carfilzomib has the potential to<br />
be a frontline treatment for multiple myeloma<br />
because it shows low levels of neuropathy, neutropenia<br />
and other toxicities associated with other<br />
multiple myeloma treatments. “We’re seeing very<br />
minimal long-term side effects in patients, and<br />
none so far that have precluded indefinite dosing,”<br />
he adds. That’s important because the goal is<br />
to make multiple myeloma a chronic condition.<br />
Kauffman also believes that the ability to keep<br />
patients on the drug should help prevent tumors<br />
from developing resistance.<br />
Millennium is also working on a secondgeneration<br />
proteasome inhibitor, an orally<br />
active peptide boronic acid analog, MLN 9708.<br />
Unlike Velcade, MLN 9708 has activity against<br />
a range of solid tumors in preclinical studies<br />
and is currently in phase 1 trials for hematologic<br />
malignancies and solid tumors. “We’re<br />
really striving to bring this next generation of<br />
proteasome inhibitors beyond the hematologic<br />
malignancies,” says Bolen.<br />
The proteasome is a multicatalytic enzyme comprising a hollow 20S core particle complexed<br />
with two 19S regulatory particles to form a 26S structure where protein degradation takes<br />
place. The two regulatory caps are bound to the core particle through ubiquitin-binding sites.<br />
The 20S particle has three catalytic actions: chymotrypsin-like activity, trypsin-like activity<br />
and peptidylglutamyl peptide hydrolyzing activity.<br />
Proteasome activity occurs through a multistep process. Proteins are marked for<br />
destruction by ubiquitin tags. This process begins with activation of ubiquitin (E1 enzyme),<br />
followed by conjugation (E2 enzyme) and finally specific ligases (E3) that mediate<br />
attachment of ubiquitin chains to target proteins. The pathway is hierarchical, with a single<br />
(or small number of) E1 enzyme(s) interacting with a small number of E2 enzymes, and<br />
several hundred E3 ubiquitin ligases serving specific subsets of proteins (Fig. 1). Specific<br />
ubiquitin ligases tag target proteins with a string of ubiquitins, which act as a signal to the<br />
proteasome to destroy it.<br />
1236 volume 28 number 12 DECEMBER 2010 nature biotechnology
news feature<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Even so, achieving success in solid tumors<br />
rather than hematological cancers will be a challenge<br />
as blood cells have high concentrations of<br />
proteasomes; indeed, Millennium researchers<br />
suspect that they act as a sink for tight-binding<br />
proteasome inhibitors, such as Velcade, thereby<br />
conferring efficacy in blood cancers. “We reasoned<br />
that a proteasome inhibitor with a shorter<br />
half-life [on the proteasome] might have better<br />
tissue distribution. This is of particular interest<br />
for the potential utility of proteasome inhibitors<br />
for treating nonhematological malignancies,”<br />
says Bolen. From a drug discovery textbook<br />
point of view, however, it was completely counterintuitive.<br />
“The mantra of drug development is<br />
to be specific and to stick to the target with high<br />
affinity or irreversibly,” he adds.<br />
Two hits are better than one<br />
Preclinical data support using proteasome<br />
inhibitors in combination with other drugs.<br />
“They should combine favorably with pretty<br />
much any type of chemotherapy,” says<br />
Kauffman. Some combination strategies have<br />
a theoretical basis as well, according to Robert<br />
Orlowski, a professor of cancer medicine at the<br />
University of Texas M.D. Anderson Cancer<br />
Center (Houston) 2 .<br />
The website ClinicalTrials.gov lists almost<br />
300 clinical trials with Velcade in combination<br />
with other drugs, including other targeting<br />
agents, such as Bristol-Myers Squibb’s<br />
(Princeton, NJ, USA) elotuzumab (an antibody<br />
targeting the myeloma cell-surface protein<br />
CS-1), Centocor Ortho Biotech’s (Horsham,<br />
PA, USA) siltuximab (an anti-interleukin-6<br />
antibody) as well as Pfizer’s (New York) Akt<br />
inhibitors (Viracept, nelfinavir mesylate) and<br />
heat shock protein 90 inhibitors (tanespimycin,<br />
Bristol-Meyers Squibb, New York). Velcade<br />
has also been combined with DNA-damaging<br />
therapies such as radiation and anthracyclines,<br />
because proteasome inhibitors suppress DNA<br />
E1<br />
E2s<br />
E3s<br />
?<br />
Substrates<br />
repair proteins, says Ken Anderson, a professor<br />
of medicine at the Dana-Farber Cancer Institute<br />
and Harvard Medical School and collaborator<br />
with Millennium and Nereus. “We’re nowhere<br />
near yet to learning the best combination, nor<br />
are we near to maximizing the benefit of this<br />
drug for patients,” says Orlowski.<br />
Palladino and others are hopeful that these<br />
combinations will allow proteasome inhibitors<br />
to expand beyond multiple myeloma, but<br />
not everyone is convinced. “It feels like they’re<br />
just throwing stuff up against the wall, looking<br />
for a signal. The risk-reward ratio (for solid<br />
tumors) is viewed favorably [by investors], but<br />
expectations are very low,” says Moskowitz.<br />
Beyond the proteasome<br />
Farther on the horizon, some companies are<br />
looking to attack the other half of the proteasome-ubiquitin<br />
pathway. Ubiquitin ligases, of<br />
which there are hundreds in mammalian cells,<br />
tag proteins with ubiquitin, marking them<br />
?<br />
?<br />
?<br />
? ?<br />
Figure 1 The pyramidal ubiquitin cascade. The ubiquitin cascade is pyramidal in design. A single<br />
E1-activating enzyme transfers ubiquitin to roughly three dozen E2s, which function together with<br />
several hundred different E3 ubiquitin ligases to ubiquitinate thousands of substrates. (Reprinted with<br />
permission from Nalepa, G. et al. Nat. Rev. Drug Discov. 5, 596–613, 2006)<br />
for destruction by the proteasome (Box 1).<br />
If researchers can find ways to modulate<br />
specific ligases, or their counterparts, the deubiquitinases,<br />
they may obtain greater specificity.<br />
“Proteasome inhibitors block the whole<br />
protein degradation machinery. I think at the<br />
end of the day, people will realize that this<br />
is not where selective therapy is going,” says<br />
Tauseef Butt, president and CEO of Progenra<br />
(Malvern, PA, USA), which is dedicated to<br />
drug discovery in the ubiquitin area.<br />
Millennium has such an inhibitor in early<br />
trials: MLN 4924, currently in phase 1 trials<br />
for hematological and solid tumors, inhibits<br />
NEDD8-activing enzyme. NEDD8 is a<br />
ubiquitin-like protein that, when activated,<br />
controls the activity of cullin-RING ligases,<br />
the largest class of E3 ligases. The company<br />
also has early-stage programs in ubiquitin and<br />
ubiquitin-like protein pathways.<br />
To date, ubiquitin ligases have shown a<br />
high degree of substrate specificity, making<br />
?<br />
Table 1 Selected agents targeting protein degradation in trials<br />
Company Drug/drug target Phase of development<br />
Takeda/Millennium<br />
Onyx Pharmaceuticals<br />
Velcade/proteasome<br />
MLN 9708/proteasome<br />
MLN 4924/Nedd8-activating enzyme<br />
MLN 519<br />
Carfilzomib/proteasome<br />
Approved for multiple myeloma (5/03)<br />
Approved for NHL (12/06)<br />
Phase 2 for breast cancer, ovarian cancer, myelodysplastic syndrome or<br />
Waldenstrom macroglobulinemia<br />
Phase 1/2 for amyloidosis or multiple myeloma (relapsed, refractory)<br />
Phase 1 for small cell lung cancer or solid tumors<br />
Phase 1/2 for multiple myeloma<br />
Phase 3 for multiple myeloma<br />
Phase 1/2 for solid tumors<br />
Phase 1 for solid tumors<br />
ONX 0912/proteasome<br />
Cephalon (Fraser, PA, USA) CEP-1877/proteasome Phase 1/2 for multiple myeloma<br />
Nereus Pharmaceuticals NPI-0052/proteasome Phase 1 for solid tumor or multiple myeloma<br />
Johnson & Johnson JNJ-26854165/MDM2 Phase 1 for solid tumors<br />
Roche RG7112/MDM2 Phase 1 for hematological cancers or solid tumors<br />
NHL, non-Hodgkin lymphoma.<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1237
NEWS feature<br />
Box 2 The other protein degrading system<br />
The proteasome and ubiquitin systems are not the sole systems for<br />
degrading protein in cells. Autophagy—whereby proteins are first<br />
packaged in vesicles that then fuse with lysozymes—is another<br />
process that has attracted attention.<br />
Tony Wyss-Coray, at the Stanford University School of<br />
Medicine, along with Beth Levine, at the University of Texas<br />
Southwestern Medical Center (Dallas), found that knocking<br />
out a key autophagy protein, Beclin-1, in mice could lead to<br />
greater neurodegeneration and beta-amyloid accumulation<br />
in the brain. Furthermore, there is evidence that Beclin-1<br />
levels are low in Alzheimer’s disease patients’ brains, whereas<br />
some other proteins involved in autophagy are unchanged.<br />
In mice, restoration of beclin-1 levels can reduce β-amyloid<br />
accumulation and reverse some of the degenerative changes.<br />
“I think there is a possibility that this may work in humans<br />
as well,” says Wyss-Coray. His team is now screening for<br />
compounds that can restore autophagy in cells that are<br />
deficient in Beclin-1.<br />
Autophagy also has a strong connection to cancer. The beclin-<br />
1–deficient mice are more prone to developing cancer at an<br />
advanced age than normal mice, and Beclin-1 is deleted in humans<br />
in about 40% of prostate cancers, 50% of breast cancers and 75%<br />
of ovarian cancers, according to Shengkan Jin, at Rutgers University<br />
in Piscataway, New Jersey. Autophagy also provides a means for cells<br />
to survive periods of stress, such as starvation, and some tumor cells<br />
may use it to survive the onslaught of chemotherapy.<br />
Like the proteasome, autophagy appears able to show some<br />
specificity in the proteins it degrades, but that process isn’t well<br />
understood. The two processes are somewhat complementary.<br />
“There are suggestions that if you inhibit one pathway, the other can<br />
take over. But I’m not aware of key proteins that play roles in both<br />
processes,” says Wyss-Coray.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
them attractive drug targets. “These [ubiquitin<br />
ligases] are the regulatory proteins of<br />
the ubiquitin system. It is the most selective<br />
intervention point,” says Avishai Levy, CEO<br />
of Proteologics (Orangeburg, NY, USA). His<br />
company is collaborating with Teva (Tel Aviv,<br />
Israel) to discover inhibitors of ring finger E3<br />
ligases, including drugs against inhibitors of<br />
the POSH polypeptide currently in preclinical<br />
testing for the potential treatment of HIV<br />
infection and cancer.<br />
Several groups are looking at key points in<br />
signal transduction pathways where ubiquitin<br />
ligases play a role. “There is [evidence] that G<br />
protein–coupled receptors are being managed<br />
by ubiquitination. Channels are being ubiquinated<br />
and recycled,” says Butt. For example,<br />
growth hormone receptors are internalized<br />
after binding growth hormone, followed by<br />
phosphorylation, ubiquitination and deposition<br />
into the lysosome. There, the ligand<br />
drops off the receptor, the terminal ubiquitin is<br />
cleaved by a de-ubiquitinase, and the receptor<br />
is returned to the cell surface, which can contribute<br />
to tumorigenesis. “If you inhibit the deubiquitinases<br />
involved, the ubiquitin remains<br />
in place and the receptor becomes a target for<br />
the proteasome. Now the cell doesn’t make a<br />
commitment to proliferation,” says Butt.<br />
The key is determining the ubiquitin status<br />
of disease-causing proteins. “It’s like the<br />
phosphorylation status of a kinase substrate,”<br />
says Donald Payan, chief scientific officer at<br />
Rigel Pharmaceuticals, which has a ubiquitin<br />
ligase program in cancer that focuses<br />
on the SCF/p27 complex. The p27 protein is<br />
a cyclin-dependent kinase inhibitor involved<br />
in the transition of a cell from G1 to S phase.<br />
SCF ubiquitinates p27. Inhibiting SCF should<br />
leave more p27 in place to act as a brake on<br />
cell cycle progression.<br />
Just as with kinases, which took 20 years to<br />
become established as drug targets, it could<br />
easily take as long for the ubiquitin ligase field<br />
to mature. One problem with designing U3<br />
ligase inhibitors is the difficulty of interfering<br />
with protein-protein interactions. “You have<br />
to look for sites or regions that are druggable,”<br />
says Payan. Even so, Orlicky et al. described<br />
a small-molecule inhibitor of the yeast F-box<br />
protein Cdc4, which is a central component<br />
of the cullin-RING ligases 3 . The inhibitor<br />
targets the WD40 propeller domain, and the<br />
structural changes induced by the inhibitor<br />
affect the substrate-binding pocket. If similar<br />
interactions occur in mammalian ligases,<br />
the strategy could open up a number of novel<br />
drug targets.<br />
Indeed, Payan claims an ongoing collaboration<br />
between Rigel and Daiichi Sankyo<br />
(Tokyo), initiated in 2002, is about to start<br />
human testing of a ubiquitin ligase inhibitor<br />
in an undisclosed cancer. Elsewhere, Apeiron<br />
Biologics (Vienna) is investigating short interfering<br />
RNAs that silence the E3 ubiquitin<br />
ligase Casitas B-cell lymphoma-b (Cbl-b) as a<br />
means of modulating T-cell activity for cancer<br />
treatment. “It’s a field where people have been<br />
wringing their hands about how the targets are<br />
not druggable, but there are four or five companies<br />
(nearing) the clinic,” he says<br />
Rigel’s own focus is not on cancer, but muscle<br />
atrophy. The ligases atrogin-1 and MuRF-1<br />
are upregulated in patients who experience<br />
muscle atrophy when being intubated 4 . “If<br />
you could block MuRF-1 and/or atrogin, you<br />
might have a big impact on muscle atrophy,<br />
which is a big opportunity,” says Payan.<br />
In many ways, the ubiquitin field is even<br />
more diverse than the kinase field, and significantly<br />
more difficult. “Key proteins all<br />
have their ligase hovering above them, ready<br />
to pounce. What has held the field back is that<br />
these are intracellular targets, so you’re stuck<br />
with small molecules. Companies must determine<br />
what’s the best way to go after these [protein<br />
interactions] chemically,” says Payan.<br />
The yin and the yang<br />
Not everyone is convinced of the potency of<br />
ligases. “There have been failures with ligases,”<br />
says Butt. He thinks the future may lie in deubiquitinases,<br />
which remove ubiquitin from<br />
proteins, about 100 of which have been identified<br />
in humans.<br />
“If the yin is the ligase, the yang is the deubiquitinase,<br />
and the yang is really taking off<br />
at the moment,” says Butt. Progenra has an<br />
inhibitor of the de-ubiquitinase USP7, which<br />
acts on the p53 tumor suppressor protein. The<br />
drug candidate has shown good anti-tumor<br />
activity in animal models and will be developed<br />
for multiple myeloma, Butt says.<br />
Ubiquitinases and de-ubiquitinases are<br />
just the latest targets in protein homeostasis,<br />
which will continue to evolve (Box 2).<br />
The ubiquitin field has significant challenges<br />
ahead. Researchers must achieve selectivity<br />
for the ligase of interest, and then build the<br />
infrastructure to study it.<br />
It’s a story Payan has heard before. “The<br />
guys who were involved in developing Velcade<br />
kept pushing and pushing, when people said it<br />
would never work. Kudos to the proteasome<br />
guys [for getting to the market]. That’s probably<br />
the least specific site to intervene.”<br />
Jim Kling, Bellingham, Washington<br />
1. Adams, J. Cancer Cell 5, 417–421 (2004).<br />
2. Rumpold, H. Biochem. Biophys. Res. Commun. 361,<br />
549–554 (2007).<br />
3. Orlicky, S. et al. Nat. Biotechnol. 28, 733–737<br />
(2010).<br />
4. Levine, S. et al. N. Engl. J. Med. 358, 1327–1335<br />
(2008).<br />
1238 volume 28 number 12 DECEMBER 2010 nature biotechnology
uilding a business<br />
Around the block<br />
Y Philip Zhang<br />
A blocking patent can kill your business before it’s off the ground.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
One of the most unpleasant surprises for<br />
any entrepreneur at a startup company is<br />
learning that a blocking patent threatens the<br />
commercial viability of their lead product. This<br />
could seriously jeopardize the company’s ability<br />
to raise financing, its attractiveness to strategic<br />
partners and ultimately its overall viability<br />
as a business.<br />
In this article, I provide a few pointers on<br />
how to avoid blocking patents in the first place<br />
and then go on to explain how to take evasive<br />
action if you are in the unfortunate position of<br />
sweating one out.<br />
First, avoid<br />
Officially, a blocking patent is one that is<br />
valid and enforceable and could prevent<br />
your potential product from coming to or<br />
perhaps staying on the market. In the United<br />
States and most industrialized countries, an<br />
actual or imminent patent infringement is<br />
grounds for government-sanctioned injunctions<br />
and monetary damages. Faced with<br />
hefty capital demands and long, perilous<br />
clinical trials, drug developers tend to avoid<br />
product candidates that present patent uncertainties<br />
and litigation risks, even if the drug<br />
satisfies unmet needs of patients. Concerns<br />
over patent infringement routinely interrupt<br />
and sometimes derail commercialization of<br />
drugs. Thus, freedom to operate (FTO)—the<br />
ability to develop and market a drug or device<br />
without infringing the valid and enforceable<br />
patent rights of others—should be high on<br />
your CEO agenda.<br />
The starting point for securing FTO is to<br />
clearly identify the business space for your<br />
company. A blocking patent is relevant only<br />
if it affects your business. A first-level inquiry<br />
asks: What are the current and future (including<br />
Y. Philip Zhang is cofounder and co–managing<br />
principal at Milstein Zhang & Wu, Newton,<br />
Massachusetts, USA.<br />
e-mail: philip.zhang@mzwiplaw.com<br />
potential) businesses that you are conducting<br />
and plan to conduct? A company in the drug<br />
discovery space faces different challenges than a<br />
company that is in the in vitro diagnostics space,<br />
for instance. A developer of new drugs typically<br />
must have a good understanding of discrete<br />
patent spaces around its drug targets and candidates<br />
in the pipeline. A company in the diagnostic<br />
or genetic testing business, on the other<br />
hand, may have to walk through the minefield<br />
of ever-increasingly patented biomarkers and<br />
immunoassays. A small-molecule drug developer<br />
generally is not too concerned with drug<br />
manufacturing. If you are developing recombinant<br />
proteins, however, manufacturing patents<br />
are plentiful and should be watched for.<br />
Next, you should identify the technologies<br />
that are vital to your company, and this may be<br />
tougher than determining your business space.<br />
It is important for you to have a mental list of the<br />
critical and enabling technologies for successful<br />
commercialization of your products. You need<br />
to know what technologies are necessary to<br />
build your products and clarify which are your<br />
own innovations and which are other people’s<br />
technology that you need access to. A developer<br />
of a single-molecule DNA-sequencing<br />
technology, for example, must evaluate FTO<br />
for its core sequencing platform but may also<br />
need to review reagents, optics, electronics and<br />
other enabling technologies. A bioethanol producer<br />
developing a new fermentation process<br />
may have to review its biomass pretreatment<br />
protocols and microorganism production steps<br />
as well as overall process integration.<br />
Once you understand your business and<br />
technology space, you’ve paved the way for<br />
the next step: understanding the patent environment<br />
in which your business operates. The<br />
patent landscape affects your patent needs for<br />
protecting your business and your FTO. A<br />
common misconception is that if you have a<br />
patent on something then you are free to practice<br />
the technology disclosed or claimed in<br />
your patent. This is not true.<br />
A patent grants its owner the right to exclude<br />
others from practicing the claimed invention<br />
without the owner’s authorization, but it does<br />
not grant the owner a right to practice the<br />
claimed invention or technology disclosed in<br />
the patent. So even if you patented a technology,<br />
it doesn’t mean that you necessarily can<br />
use it yourself because your patented technology<br />
could be dominated by another patent.<br />
Thus, it is important early on to have a<br />
solid grasp of the patent environment surrounding<br />
key product candidates through<br />
rigorous patent search and review. The patent<br />
search and review process is preferably conducted<br />
by experienced legal counsel working<br />
closely with your project team. Scientists<br />
should resist the temptation to play patent<br />
lawyer. Understanding the science in a patent<br />
is important, but it’s only the first step<br />
in understanding the impact of the patent,<br />
which is better done by, or with the guidance<br />
of, experienced counsel. A good patent lawyer<br />
is one who understands your business and<br />
the industry you operate in and can effectively<br />
communicate with your management and<br />
scientists. He or she understands the business<br />
ramifications of the patent issue at hand and<br />
helps you make the right judgment calls.<br />
You and your counsel should map out the<br />
existing patent landscape along the entire commercialization<br />
path, even if your business plan<br />
is to seek exit before market entrance. If you<br />
find a patent that presents FTO risks, do not<br />
panic. The good thing is that not everything is<br />
patented (even though it might seem like it),<br />
and more often than not you will be able to find<br />
a path forward. Ideally, patent searches should<br />
cover both granted patents and published patent<br />
applications and additionally encompass all<br />
intended international markets for the product.<br />
Knowing where a patent is granted or pending<br />
helps with assessing its overall impact and<br />
affords you an early appreciation for the scope<br />
of your need for licensing, design-around<br />
options or patent challenge.<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1239
uilding a business<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Box 1 Blocking patent? No<br />
Determining patent infringement involves a two-step analysis: first, construction of the<br />
claim to determine the coverage of a patent, and second, comparison of the alleged<br />
infringing composition, device or method with the construed claims. Thus, overcoming a<br />
patent means understanding its claim scope and designing the product outside such scope.<br />
Consider the patent fight between Novartis Pharmaceuticals and Eon Labs<br />
manufacturing. Novartis has a patent on a drug formulation that claims a hydrosol,<br />
which includes solid particles of the drug compound and a stabilizer that maintains<br />
the size distribution of the particles. The claim further specifies the water solubility<br />
of the drug compound as well as the ratios of drug compound to water and drug<br />
compound to the stabilizer in the hydrosol.<br />
Along comes Eon. It sells a drug product that includes the same drug compound<br />
but not in hydrosol form. Eon instead makes capsules that contain the drug compound<br />
dissolved in a small amount of ethanol. There is no water in these capsules, and the drug<br />
compound inside the capsule is completely dissolved in ethanol (not in particle form).<br />
Novartis sued Eon for infringement, asserting that when Eon’s capsule is ingested<br />
by a patient, an infringing hydrosol is formed when the capsule mixes with the<br />
aqueous environment of the patient’s stomach. This unusual infringement theory did<br />
not survive summary judgment at the trial court (and on appeal) as Novartis’ hydrosol<br />
claim was construed to be limited to a medicinal preparation consisting of a dispersion<br />
of solid particles in an aqueous colloidal solution prepared outside of the body, which<br />
Eon’s capsules clearly are not 1 .<br />
Another factor that may require some soul<br />
searching on your part is the risk profile of<br />
your company. Clearly articulating your risk<br />
tolerance can be difficult, but in many ways<br />
it shapes the appropriate approach to a patent<br />
problem. Your risk profile may change over<br />
time and is determined by a number of factors,<br />
including the business sector you operate<br />
in, the stage of your company, your financial<br />
situation, the liability exposure and the personalities<br />
of the management. The life science<br />
community in general and venture capital<br />
investors in particular are quite risk averse<br />
when it comes to patents and FTO. Because<br />
of the large capital need and lengthy and risky<br />
development process, investors do not like to<br />
deal with patent uncertainties and litigation<br />
threats. Technology and regulatory risks are<br />
already plentiful in biotech—you don’t need<br />
anything more. Heightened patent risks could<br />
drive your board and the investors nuts.<br />
A review of your FTO should not be a onetime<br />
occurrence, because a thorough search<br />
and review doesn’t necessarily uncover<br />
all relevant patents. For one thing, patent<br />
applications are published 18 months after<br />
filing, so there is always a black box in the<br />
patent space. For another thing, your product<br />
design could change and your need for<br />
enabling technologies may evolve over time.<br />
A good grasp of the patent landscape affords<br />
you the ability to make adjustments along<br />
the product development path. In the life<br />
science sector, the point of each significant<br />
go or no-go decision is often a good time<br />
to revisit the patent situation. Competent<br />
counsel can guide you through the process.<br />
Ignorance is rarely a viable approach—<br />
sophisticated investors and collaborators will<br />
conduct thorough due diligence on your patent<br />
position. It is better that you know your<br />
shortcomings before potential investors or<br />
partners point them out.<br />
Next, overcome<br />
If the worst happens and you do identify a<br />
potential blocking patent or patent application,<br />
it is critical to fully understand how it<br />
impacts your business. For example, does it<br />
affect the technology platform at the core<br />
level, a particular product composition, a<br />
manufacturing step or a specific use of a<br />
product? You and your patent counsel should<br />
understand the patent’s legal status and the<br />
claim scope. Information on the patent<br />
(whether the claims are allowed, issued, on<br />
appeal, in interference, under reexamination<br />
or challenged in litigation) can be obtained<br />
by your counsel. Often a review of the patent<br />
file history is needed to fully grasp the scope<br />
of a patent and its real or potential weaknesses.<br />
Claims in different countries quite<br />
often are not identical; therefore, you should<br />
consider obtaining legal advice from a lawyer<br />
in each of the countries of interest.<br />
In addition, you should gather background<br />
information about the patent owner as a prelude<br />
to a licensing effort or a preparation for<br />
challenge. Where the patent is licensed to a<br />
third party, information on the licensee could<br />
be valuable as well. They could become your<br />
competitor, collaborator, licensor, licensee<br />
or all of the above. Knowing their business<br />
situation, their financial clout, their technology<br />
and their patent needs can help you<br />
gain leverage and affect how you deal with<br />
the blocking patent and its owner. Patent<br />
ownership and licensing information may<br />
be obtained from appropriate databases,<br />
such as ones found at the US Patent and<br />
Trademark Office, the World Intellectual<br />
Property Organization and the US Securities<br />
Exchange Commission.<br />
You’ll then need to explore design-around<br />
options. Quite often, a blocking patent is relevant<br />
only because of one or two features of<br />
a candidate compound or device. When such<br />
features can be removed or modified without<br />
sacrificing the needed functionality, you<br />
might find a successful design-around solution.<br />
Working closely with your scientific<br />
team, an experienced lawyer can make significant<br />
contributions in identifying and finetuning<br />
successful design-around options.<br />
In a hypothetical drug development scenario,<br />
a drug is approved by the US Food and<br />
Drug Administration for treating disease A.<br />
The owner of the drug has obtained patents<br />
on the drug compound, formulation and use<br />
of the compound to treat disease A. The drug,<br />
however, is now suspected to have potency<br />
against disease B. The suspicion is further<br />
borne out by your preclinical studies.<br />
In this scenario, FTO issues are clearly<br />
present due to the patent estate around the<br />
approved drug. These patents, however,<br />
might be limited to the approved drug compound<br />
itself and certain close derivatives.<br />
Some chemical compound patents suffer<br />
from insufficient enablement regarding<br />
distant derivatives and analogs. Analysis of<br />
the patents might reveal that all valid patent<br />
claims are tied to a certain core chemical<br />
scaffold or more commonly to certain<br />
specific functionalities on the core chemical<br />
scaffold. Working together, your scientists<br />
and patent counsel can design exploratory<br />
compound libraries that investigate the surrounding<br />
chemical space uncovered by the<br />
patent. Knowledge of the structure-activity<br />
relationship and the patent landscape is critical<br />
in successfully navigating the hit-to-lead<br />
evolution process. A good practice from a<br />
patent counsel’s perspective is to fully grasp<br />
both technical and legal parameters while<br />
communicating effectively and working intimately<br />
with your project team. The objective<br />
should be to obtain a design-around solution<br />
that minimizes infringement risks and<br />
allows solid, fresh patent protection for the<br />
redesigned compound (Box 1).<br />
When faced with a blocking patent, you<br />
should also explore the availability and cost of a<br />
1240 volume 28 number 12 DECEMBER 2010 nature biotechnology
uilding a business<br />
© 2010 Nature America, Inc. All rights reserved.<br />
patent license. Too often the project team drops<br />
a compound after the discovery that it is covered<br />
by a third-party patent. Licensing opportunities<br />
often are not carefully investigated and<br />
followed up, resulting in premature loss of good<br />
drug candidates. It is more desirable from efficiency<br />
and productivity perspectives that biotech<br />
startups become more open to licensing<br />
both at the receiving and giving ends.<br />
You’ll need your counsel to closely guide<br />
the licensing process, as many pitfalls exist<br />
that could entrap the unwary. A misstep<br />
could expose you to various risks, ranging<br />
from prematurely exposing your licensing<br />
needs to the patent owner to a loss of your<br />
right to later challenge the patent. It is generally<br />
a good idea to enter a confidentiality<br />
agreement before exchanging sensitive information<br />
and starting negotiations. You can<br />
also sometimes enter a joint privilege agreement<br />
if sensitive opinions or legal documents<br />
are shared with the other side.<br />
Keep in mind, though, that the patent<br />
licensing process could be lengthy and<br />
exhaustive, and it can easily take months to<br />
complete. This means it’s important to set<br />
a timetable and move forward accordingly.<br />
Your knowledge of the weaknesses of the<br />
desired patent estate and its holder could help<br />
steer the process favorably for you. In this<br />
regard, due diligence on the desired patent<br />
estate and its owner is must-do homework<br />
that should be completed before approaching<br />
the owner. Understanding the value of<br />
the license to you is important because the<br />
deal inevitably involves setting licensing<br />
fees and royalty rates. Experienced counsel<br />
might know the prevailing licensing rates<br />
and industry terms. Your ultimate objective<br />
in a licensing discussion is to fully explore<br />
the potential of a license in view of all other<br />
options available to you so that you can make<br />
a cogent business decision.<br />
Though rarely enjoyable, sometimes the<br />
best approach is to simply challenge a blocking<br />
patent face on. It’s preferable to have proactive<br />
patent challenges at favorable times and<br />
venues rather than mere reactive defenses.<br />
Well-planned and executed strategies for<br />
patent challenges often help realize the full<br />
commercial potential of your product.<br />
Legal procedures vary from country to<br />
country, but some forms of patent validity<br />
challenge are available in all major jurisdictions.<br />
For example, Europe, Japan, China,<br />
Australia and Canada all have post-grant<br />
opposition or cancellation procedures. In<br />
the United States, the primary procedures<br />
for validity challenge at the US Patent and<br />
Trademark Office are ex parte and inter parte<br />
reexaminations. Currently, post-grant opposition<br />
is not available in the United States,<br />
but it is a part of most patent reform bills and<br />
could become available soon, as it’s believed<br />
that some iteration of patent reform is likely<br />
to pass into law.<br />
At any time after US patent issuance<br />
and before expiration, anyone may request<br />
reexamination on the grounds of a substantial<br />
new question of patentability based on<br />
printed prior art or patent references. A<br />
notable benefit is that the outcome of the ex<br />
parte reexamination does not have a binding<br />
effect on the requester, which means that you<br />
may raise the same basis of invalidity again in<br />
a pending or future patent litigation, essentially<br />
allowing you two bites at the apple. A<br />
potential downside, however, is that you do<br />
not have much direct participation in the proceedings,<br />
and the patent holder could steer<br />
that reexamination process to strengthen his<br />
or her patent. Inter parte reexamination, in<br />
contrast, allows more direct involvement,<br />
but the outcome has a binding effect on the<br />
requester. Heed your counsel’s advice when<br />
formulating and implementing effective<br />
patent challenge strategies.<br />
Conclusions<br />
FTO is critical to life science startups. Not having<br />
it from the beginning of a venture is like<br />
a building without a foundation. A biotech<br />
startup is wise to craft and implement an appropriate<br />
patent strategy early on that steers it clear<br />
of and minimizes the impact of hostile patents.<br />
But should one appear, there are approaches to<br />
try in order to maneuver around it. Still, nothing<br />
can substitute for the commitment of the<br />
company and the teamwork of management,<br />
scientists and counsel.<br />
1. Novartis Pharms. Corp. v. Eon Labs Mfg., Inc., 363 F.<br />
3d 1306 (Fed. Cir. 2004).<br />
To discuss the contents of this article, join the Bioentrepreneur forum on Nature Network:<br />
http://network.nature.com/groups/bioentrepreneur/forum/topics<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1241
correspondence<br />
Wrong fixes for gene patents<br />
© 2010 Nature America, Inc. All rights reserved.<br />
To the Editor:<br />
In your August issue, the Commentary<br />
entitled “DNA patents and diagnostics: not<br />
a pretty picture” by Robert Cook-Deegan<br />
and colleagues 1 draws conclusions and<br />
recommendations that are unwarranted by<br />
the full scope of available data and unfairly<br />
criticizes the activities of the Biotechnology<br />
Industry Organization (BIO; Washington,<br />
DC, USA) with respect to this issue. As<br />
general counsel and senior vice president for<br />
legal & intellectual property for BIO, I would<br />
like to respond to the criticisms in this article.<br />
The authors admit that there is “limited”<br />
evidence that patents are negatively impacting<br />
research on, or the accessibility of, genetic<br />
diagnostic tests, but then go on to assert that,<br />
despite this lack of evidence, “concerns about<br />
DNA patents persist.” They reference many<br />
studies for this proposition, yet the cumulative<br />
evidence contained in those studies does not<br />
support any systemic basis for such concerns.<br />
Anecdotal examples of access problems,<br />
such as those contained in the specific case<br />
studies heavily relied upon by the authors,<br />
may be helpful to consider but, as the authors<br />
admit, the evidence of problems from<br />
these particular examples is “not uniform,<br />
consistent or pervasive.” To draw from such<br />
anecdotal evidence the general conclusion<br />
that patents, or the exclusive licensing thereof,<br />
are the cause of such problems, or are not<br />
necessary to incentivize the development and<br />
marketing of diagnostic tests except in rare<br />
cases, is simply not justified.<br />
Yet the authors assert that these case<br />
studies—which, although well done, are<br />
truly case studies and are not purported to<br />
be representative of the entire field of genetic<br />
diagnostic testing—show that “Exclusive<br />
licensing practices consistently reduce<br />
availability” of genetic diagnostic tests. The<br />
weakness in this argument, beyond the<br />
unsupported generalization noted above, is<br />
found in the phrase immediately following<br />
this conclusion—“at least as measured by the<br />
number of available laboratories offering a<br />
test” (emphasis added). Of course, it should<br />
come as no surprise—and certainly we did<br />
not need a series of case studies to teach us—<br />
that exclusive licensing limits the number of<br />
laboratories offering a test and, as the authors<br />
further complain, “reduce[s] competition.”<br />
In fact, this is true of every product that is<br />
patented or exclusively licensed, regardless of<br />
the field. The key fact that should concern us,<br />
however, is not how many entities provide the<br />
test, but whether individuals can reasonably<br />
gain access to it. Just because several academic<br />
research laboratories may offer a particular<br />
test absent an exclusive license to another<br />
party, it does not automatically follow that<br />
people would have greater access to that test.<br />
Indeed, it is just as reasonable to project that,<br />
without a single entity having a financial<br />
incentive to develop not just the test but<br />
also the market for the test, many patients<br />
who should be screened (and their doctors)<br />
may never learn about the availability and<br />
usefulness of the test at all. Exclusive licensees<br />
also would seemingly have a much greater<br />
incentive to validate and improve the test, and<br />
to help obtain coverage for such tests from<br />
payors and insurers. As the studies do reveal,<br />
lack of insurance coverage can be a major<br />
obstacle to patient access to diagnostic tests,<br />
particularly for poor women. In addition,<br />
concern about reduced competition in and<br />
of itself is misplaced; the concern should be<br />
whether, in light of the reduced competition,<br />
there is public harm in the form of higher<br />
prices or inferior products. Again, the case<br />
studies do not reveal any significant evidence<br />
of either.<br />
The Commentary then criticizes BIO<br />
for having “chosen not to acknowledge the<br />
real problems” that exist with respect to<br />
the genetic diagnostics market, and for its<br />
“quick and vociferous” opposition to the<br />
recommendations of the US Secretary’s<br />
Advisory Committee on Genetics, Health<br />
& Society (SACGHS). To the contrary,<br />
BIO publicly testified before SACGHS on<br />
its original, more balanced draft report<br />
and praised the work of the committee.<br />
BIO cautioned the committee that some<br />
of the proposals under consideration<br />
could undermine the goal of enhancing<br />
patient access to genetic tests and urged the<br />
committee to partner with BIO in crafting<br />
sensible recommendations that would actually<br />
enhance patient access to such tests (e.g.,<br />
expanded insurance coverage). Instead, the<br />
committee rewrote its original draft report<br />
in an apparent attempt to support a series of<br />
divisive, controversial and highly impractical<br />
recommendations to alter patent law and<br />
impose one-size-fits-all licensing strategies<br />
enforced by the federal government. Not only<br />
were these recommendations not supported<br />
by the committee’s own review and case<br />
studies, but they also chased red herrings such<br />
as patenting and exclusive licensing practices,<br />
while avoiding more practical approaches that<br />
would not only have garnered more consensus<br />
among external stakeholders and the<br />
committee’s own members (three of whom<br />
dissented in a highly unusual maneuver)<br />
but also would have been more effective in<br />
achieving our shared goals.<br />
The authors also chastise BIO for not<br />
proposing constructive solutions, such as the<br />
development of a ‘gene patent supermarket’<br />
or similar licensing pools that could help<br />
facilitate broader licensing of such patents,<br />
while maintaining appropriate incentives.<br />
In actuality, BIO has been discussing such<br />
approaches with our members and with<br />
the US National Institutes of Health for<br />
several months now—not because there is<br />
evidence that DNA patents are impeding<br />
the development of more complex or ‘whole’<br />
genomic testing products, but to help avoid<br />
any chilling effect that might arise in the<br />
future. BIO also has been working with<br />
other organizations in developing ideas<br />
to address the ‘second opinion’ quandary<br />
that may arise when there is a sole provider<br />
of an important medical test that drives a<br />
particular treatment regimen. The key point,<br />
however, is that proposed solutions must be<br />
targeted to deal with specific problems, while<br />
avoiding unintended and counterproductive<br />
consequences for innovation and patient care.<br />
Finally, the authors criticize BIO for<br />
“failing to enforce the established norms”<br />
of nonexclusive licensing and for not<br />
“criticiz[ing] deviations from them” by our<br />
members. Leaving aside the obvious antitrust<br />
and other legal and practical concerns about<br />
1242 volume 28 number 12 DECEMBER 2010 nature biotechnology
correspondence<br />
© 2010 Nature America, Inc. All rights reserved.<br />
a trade association seeking to require or<br />
‘enforce’ specific licensing practices by its<br />
membership, the authors’ suggestion that<br />
nonparties to a specific licensing transaction<br />
would have the information necessary to<br />
reasonably judge (presumably, after the<br />
fact) whether an exclusive agreement is<br />
or is not appropriate under the particular<br />
circumstances involved is simply not realistic,<br />
as much of the relevant information would<br />
likely be proprietary. In addition, BIO’s<br />
membership is largely made up of therapeutic<br />
research and development companies, rather<br />
than the type of companies whose business<br />
models are the focus of the authors’ critique.<br />
In short, BIO never has denied that certain<br />
problems exist with respect to access to<br />
genetic diagnostic testing; we just disagree<br />
as to the causes of such problems and how<br />
best to fix them. BIO will continue its efforts<br />
to work with other organizations to help<br />
improve patient access to genetic testing, but<br />
will also continue to oppose—vigorously<br />
when necessary—any ill-considered and<br />
misguided proposals that would undermine<br />
the development of new diagnostics and<br />
therapies and do more harm than good for the<br />
patients of today and tomorrow.<br />
COMPETING FINANCIAL INTERESTS<br />
The author declares no competing financial interests.<br />
Tom DiLenge<br />
Biotechnology Industry Organization,<br />
Washington, DC, USA.<br />
e-mail: tdilenge@bio.org<br />
1. Carbone, J. et al. Nat. Biotechnol. 28, 784–791<br />
(2010).<br />
Robert Cook-Deegan,<br />
Subhashini Chandrasekharan,<br />
Misha Angrist, Bhaven Sampat,<br />
E Richard Gold, Julia Carbone &<br />
Lori Knowles reply:<br />
We thank Tom DiLenge of BIO for his<br />
thoughtful comments. We agree with many<br />
points, but focus here on remaining points of<br />
disagreement.<br />
First, although we agree there is no<br />
evidence of systematic and pervasive<br />
harm from patenting and licensing in<br />
DNA diagnostics, we reiterate that there<br />
is unequivocal evidence of problems in<br />
some cases. We agree there may well be a<br />
role for patent incentives in DNA testing;<br />
we do not believe, however, that this means<br />
carte blanche for patent holders. We are<br />
particularly wary of exclusive licensing<br />
to sole providers of genetic tests unless<br />
nonexclusive licensing will fail to bring a<br />
product to market. This is decidedly not the<br />
case in empirical studies to date. We say this<br />
for three main reasons. First, in instances<br />
where no test is available and yet patents are<br />
being enforced, as was the case with long-QT<br />
testing from 2002 to 2004, there are clearly<br />
access problems by any definition. These<br />
situations may be rare, and we hope they are,<br />
but denying a problem that has historically<br />
occurred is not a winning argument. The<br />
BIO letter is silent on such problems.<br />
Second, it is simply not true that exclusive<br />
licensing needs to lead to monopolies. If<br />
a particular laboratory does not offer a<br />
particular form of service (e.g., prenatal<br />
testing), does not have a payment agreement<br />
with an insurer or health plan, or has already<br />
gotten paid to do a test, and the patient (or<br />
doctor) wants verification, then prudent<br />
business practice would suggest sublicensing,<br />
a permissive testing policy or some other<br />
way to ensure testing can be done by others.<br />
Policies on sublicensing or testing by others<br />
are under control of the patent holder and<br />
could be remedied by them without breaking<br />
patents. It is thus puzzling that patentholders<br />
have not adopted such policies.<br />
Third, although we agree that reducing<br />
the number of laboratories offering a test<br />
does not necessarily reduce patient access,<br />
there is a very consistent pattern revealed<br />
in our case studies and the survey of<br />
laboratory directors that we cited by Cho<br />
et al. 1 : the holder of exclusive patent rights<br />
is consistently not first to market with a<br />
genetic test. The effect of patents has been<br />
solely to reduce competition, not to create<br />
new products that would not otherwise exist.<br />
Suppression of competitors who have beaten<br />
the holder of exclusive rights to market is<br />
not what is usually observed with patents.<br />
Pharmaceutical firms and instrument<br />
companies generally do not enforce patents<br />
against universities and research institutions,<br />
for example, and yet this is what we find in<br />
DNA diagnostics in several cases. In this<br />
respect, diagnostics are unusual compared<br />
with other domains where patent exclusivity<br />
has a role. We agree the evidence of harms<br />
from exclusive licensing is not systematic,<br />
but the evidence of benefit from patents<br />
in genetic diagnostics historically is even<br />
weaker.<br />
Finally, we appreciate there are indeed<br />
limits to BIO’s actions when questions<br />
of antitrust would arise in enforcing the<br />
existing norms on patenting and licensing<br />
genomic inventions. The licensing norms<br />
developed by the Organization for Economic<br />
Cooperation and Development 2 (Paris),<br />
the US National Institutes of Health 3 and<br />
the ‘Nine Points’ document on university<br />
technology licensing 4 are all pro-competitive<br />
however, not anti-competitive. If a company<br />
is deviating from those norms, therefore,<br />
antitrust concerns would not arise; quite<br />
the reverse. We don’t suggest BIO act<br />
when antitrust would loom as an issue, but<br />
commenting on policies—such as enforcing<br />
patents when no test is available to patients—<br />
would rarely confront antitrust policy.<br />
The main underlying point is that<br />
problems with patents and exclusive licensing<br />
distinctive to diagnostics can be identified<br />
and dealt with, but only if the problems are<br />
acknowledged and acted upon. If BIO is<br />
turning its attention to these issues, then we<br />
will all benefit.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
1. Cho, M.K., Illangasekare, S., Weaver, M.A., Leonard,<br />
D.G.B. & Merz, J.F. J. Mol. Diagn. 5, 3–8 (2003).<br />
2. Organisation for Economic Co-operation and<br />
Development. Guidelines for the Licensing of Genetic<br />
Inventions (OECD, Paris, 2006).<br />
3. National Institutes of Health. “Best Practices for the<br />
Licensing of Genomic Inventions,” Federal Register 70<br />
(No. 68): 18412–18415.<br />
4. Association of University Technology Managers. In the<br />
Public Interest: Nine Points to Consider in Licensing<br />
University Technology (AUTM, Deerfield, Illinois, USA,<br />
2007).<br />
Stem cell clinics in the news<br />
To the Editor:<br />
As highlighted in a News Feature “Trading<br />
on hope” published in this journal 1 , stem<br />
cell tourism is a growing and increasingly<br />
contentious phenomenon. By ‘stem cell<br />
tourism’, we refer to the emerging practice<br />
that sees patients travel abroad to receive<br />
(largely) unproven stem cell treatments that<br />
are generally not approved or available in their<br />
home country 2 . Although precise numbers are<br />
unknown, current information suggests that<br />
potentially thousands of patients each year<br />
from various countries are travelling around<br />
the world to receive stem cell therapies for a<br />
wide range of conditions 3–5 .<br />
The stem cell tourism phenomenon<br />
is highly controversial. The therapeutic<br />
possibilities promised by the clinics involved<br />
engage the hopes of often desperate patients<br />
and their families, including those who feel<br />
there are no other options. These individuals<br />
are understandably anxious to have a<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1243
correspondence<br />
© 2010 Nature America, Inc. All rights reserved.<br />
a trade association seeking to require or<br />
‘enforce’ specific licensing practices by its<br />
membership, the authors’ suggestion that<br />
nonparties to a specific licensing transaction<br />
would have the information necessary to<br />
reasonably judge (presumably, after the<br />
fact) whether an exclusive agreement is<br />
or is not appropriate under the particular<br />
circumstances involved is simply not realistic,<br />
as much of the relevant information would<br />
likely be proprietary. In addition, BIO’s<br />
membership is largely made up of therapeutic<br />
research and development companies, rather<br />
than the type of companies whose business<br />
models are the focus of the authors’ critique.<br />
In short, BIO never has denied that certain<br />
problems exist with respect to access to<br />
genetic diagnostic testing; we just disagree<br />
as to the causes of such problems and how<br />
best to fix them. BIO will continue its efforts<br />
to work with other organizations to help<br />
improve patient access to genetic testing, but<br />
will also continue to oppose—vigorously<br />
when necessary—any ill-considered and<br />
misguided proposals that would undermine<br />
the development of new diagnostics and<br />
therapies and do more harm than good for the<br />
patients of today and tomorrow.<br />
COMPETING FINANCIAL INTERESTS<br />
The author declares no competing financial interests.<br />
Tom DiLenge<br />
Biotechnology Industry Organization,<br />
Washington, DC, USA.<br />
e-mail: tdilenge@bio.org<br />
1. Carbone, J. et al. Nat. Biotechnol. 28, 784–791<br />
(2010).<br />
Robert Cook-Deegan,<br />
Subhashini Chandrasekharan,<br />
Misha Angrist, Bhaven Sampat,<br />
E Richard Gold, Julia Carbone &<br />
Lori Knowles reply:<br />
We thank Tom DiLenge of BIO for his<br />
thoughtful comments. We agree with many<br />
points, but focus here on remaining points of<br />
disagreement.<br />
First, although we agree there is no<br />
evidence of systematic and pervasive<br />
harm from patenting and licensing in<br />
DNA diagnostics, we reiterate that there<br />
is unequivocal evidence of problems in<br />
some cases. We agree there may well be a<br />
role for patent incentives in DNA testing;<br />
we do not believe, however, that this means<br />
carte blanche for patent holders. We are<br />
particularly wary of exclusive licensing<br />
to sole providers of genetic tests unless<br />
nonexclusive licensing will fail to bring a<br />
product to market. This is decidedly not the<br />
case in empirical studies to date. We say this<br />
for three main reasons. First, in instances<br />
where no test is available and yet patents are<br />
being enforced, as was the case with long-QT<br />
testing from 2002 to 2004, there are clearly<br />
access problems by any definition. These<br />
situations may be rare, and we hope they are,<br />
but denying a problem that has historically<br />
occurred is not a winning argument. The<br />
BIO letter is silent on such problems.<br />
Second, it is simply not true that exclusive<br />
licensing needs to lead to monopolies. If<br />
a particular laboratory does not offer a<br />
particular form of service (e.g., prenatal<br />
testing), does not have a payment agreement<br />
with an insurer or health plan, or has already<br />
gotten paid to do a test, and the patient (or<br />
doctor) wants verification, then prudent<br />
business practice would suggest sublicensing,<br />
a permissive testing policy or some other<br />
way to ensure testing can be done by others.<br />
Policies on sublicensing or testing by others<br />
are under control of the patent holder and<br />
could be remedied by them without breaking<br />
patents. It is thus puzzling that patentholders<br />
have not adopted such policies.<br />
Third, although we agree that reducing<br />
the number of laboratories offering a test<br />
does not necessarily reduce patient access,<br />
there is a very consistent pattern revealed<br />
in our case studies and the survey of<br />
laboratory directors that we cited by Cho<br />
et al. 1 : the holder of exclusive patent rights<br />
is consistently not first to market with a<br />
genetic test. The effect of patents has been<br />
solely to reduce competition, not to create<br />
new products that would not otherwise exist.<br />
Suppression of competitors who have beaten<br />
the holder of exclusive rights to market is<br />
not what is usually observed with patents.<br />
Pharmaceutical firms and instrument<br />
companies generally do not enforce patents<br />
against universities and research institutions,<br />
for example, and yet this is what we find in<br />
DNA diagnostics in several cases. In this<br />
respect, diagnostics are unusual compared<br />
with other domains where patent exclusivity<br />
has a role. We agree the evidence of harms<br />
from exclusive licensing is not systematic,<br />
but the evidence of benefit from patents<br />
in genetic diagnostics historically is even<br />
weaker.<br />
Finally, we appreciate there are indeed<br />
limits to BIO’s actions when questions<br />
of antitrust would arise in enforcing the<br />
existing norms on patenting and licensing<br />
genomic inventions. The licensing norms<br />
developed by the Organization for Economic<br />
Cooperation and Development 2 (Paris),<br />
the US National Institutes of Health 3 and<br />
the ‘Nine Points’ document on university<br />
technology licensing 4 are all pro-competitive<br />
however, not anti-competitive. If a company<br />
is deviating from those norms, therefore,<br />
antitrust concerns would not arise; quite<br />
the reverse. We don’t suggest BIO act<br />
when antitrust would loom as an issue, but<br />
commenting on policies—such as enforcing<br />
patents when no test is available to patients—<br />
would rarely confront antitrust policy.<br />
The main underlying point is that<br />
problems with patents and exclusive licensing<br />
distinctive to diagnostics can be identified<br />
and dealt with, but only if the problems are<br />
acknowledged and acted upon. If BIO is<br />
turning its attention to these issues, then we<br />
will all benefit.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
1. Cho, M.K., Illangasekare, S., Weaver, M.A., Leonard,<br />
D.G.B. & Merz, J.F. J. Mol. Diagn. 5, 3–8 (2003).<br />
2. Organisation for Economic Co-operation and<br />
Development. Guidelines for the Licensing of Genetic<br />
Inventions (OECD, Paris, 2006).<br />
3. National Institutes of Health. “Best Practices for the<br />
Licensing of Genomic Inventions,” Federal Register 70<br />
(No. 68): 18412–18415.<br />
4. Association of University Technology Managers. In the<br />
Public Interest: Nine Points to Consider in Licensing<br />
University Technology (AUTM, Deerfield, Illinois, USA,<br />
2007).<br />
Stem cell clinics in the news<br />
To the Editor:<br />
As highlighted in a News Feature “Trading<br />
on hope” published in this journal 1 , stem<br />
cell tourism is a growing and increasingly<br />
contentious phenomenon. By ‘stem cell<br />
tourism’, we refer to the emerging practice<br />
that sees patients travel abroad to receive<br />
(largely) unproven stem cell treatments that<br />
are generally not approved or available in their<br />
home country 2 . Although precise numbers are<br />
unknown, current information suggests that<br />
potentially thousands of patients each year<br />
from various countries are travelling around<br />
the world to receive stem cell therapies for a<br />
wide range of conditions 3–5 .<br />
The stem cell tourism phenomenon<br />
is highly controversial. The therapeutic<br />
possibilities promised by the clinics involved<br />
engage the hopes of often desperate patients<br />
and their families, including those who feel<br />
there are no other options. These individuals<br />
are understandably anxious to have a<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1243
correspondence<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Table 1 Demographics of patient undergoing stem cell treatment as reported in news<br />
media articles from October 2006 to September 2009<br />
Gender<br />
Male<br />
(134; 59.8%)<br />
Female<br />
(87; 38.8%)<br />
Unspecified<br />
(3; 1.3%)<br />
Age (adult<br />
versus minor)<br />
Adult<br />
(126; 56.3%)<br />
Minors<br />
(98; 43.8%)<br />
Country of origin (in<br />
descending order)<br />
US (84; 37.5%)<br />
UK (80; 35.7%)<br />
Australia (32; 14.3%)<br />
Canada (12; 5.4%)<br />
reason to feel optimistic about their medical<br />
prospects, and often frustrated with their<br />
native medical systems. Patients pay for these<br />
treatments personally, and research reveals<br />
costs ranging from $5,000 to $39,500 (ref. 4),<br />
or an average cost of $21,500 (excluding travel<br />
and accommodation) 3 . These treatments<br />
generally have not proceeded through the<br />
clinical trial process, and published research<br />
reveals little to no evidence of efficacy for the<br />
services advertised by these clinics 3 . There also<br />
appears to be an overall lack of transparency<br />
regarding the specifics of treatment protocols<br />
and an absence of comprehensive posttreatment<br />
follow-up. Accordingly, many<br />
experts are highly skeptical of the claims<br />
made by stem cell clinics and, particularly in<br />
light of emerging evidence regarding adverse<br />
effects from such treatments 6 , concerned<br />
about potential risks 7,8 . Other emerging<br />
issues include a lack of informed consent<br />
because of inadequate information provided<br />
to prospective patients 4,9 , the potential for<br />
vulnerable individuals (including children) to<br />
be taken advantage of or put at risk 10 , and the<br />
obligation to provide continuing care upon a<br />
patient’s return.<br />
Not surprisingly, treatment providers are<br />
often highly motivated to defend their work<br />
and to advocate for the use of novel treatments.<br />
This range of conflicting and competing<br />
perspectives regarding stem cell tourism serves<br />
to create a confusing picture of the field in the<br />
news media, especially for patients and their<br />
families. Previous research points to two key<br />
mechanisms by which news media portrayals<br />
are likely to influence audiences. First, through<br />
a process of agenda setting, the news media<br />
calls attention to specific issues, events or<br />
developments 11 , such as the claims of stem<br />
cell clinics or developments in the field. Given<br />
high levels of motivation, patients and their<br />
families are likely to pay closer attention to<br />
stem cell-related news coverage and to actively<br />
Five most common<br />
conditions<br />
Multiple sclerosis<br />
(22; 9.8%)<br />
Cerebral palsy<br />
(18; 8.0%)<br />
Septo-optic dysplasia<br />
(15; 6.7%)<br />
Optic nerve hypoplasia<br />
(13, 5.8%)<br />
New Zealand (12; 5.4%) Unspecified blindness<br />
(13; 5.8%)<br />
Israel (1; 0.5%)<br />
Brazil (1; 0.5%)<br />
Three most common<br />
treatment destinations<br />
China (100; 44.6%)<br />
Germany (17; 7.6%)<br />
Mexico (17; 7.6%)<br />
seek information about therapies online<br />
(Supplementary Discussion 1).<br />
Not only do the news media—and<br />
newspapers especially—call the public’s<br />
attention to medical claims and (unproven)<br />
therapies, but news coverage also selectively<br />
frames the nature of these claims. ‘Frames’<br />
is the conceptual term for interpretative<br />
storylines that emphasize specific dimensions<br />
of a complex topic over others, often reducing<br />
complexity and uncertainty, and leading<br />
audiences to consider certain considerations<br />
over others in reaching judgments and making<br />
decisions (Supplementary Discussion 1) 12 .<br />
Given the likely importance of the news<br />
media—and newspapers in particular—in<br />
setting the agenda of the public and in shaping<br />
public interpretations, we thought it essential<br />
to investigate the level of attention to stem<br />
cell tourism in the print media and how the<br />
emerging industry is characterized. Our<br />
analysis also provides information about the<br />
individuals accessing these services, where the<br />
clinics reside and what services are provided.<br />
We searched the Factiva database for<br />
newspaper articles about stem cell tourism<br />
from Canada, the United States, the United<br />
Kingdom, Australia and New Zealand between<br />
October 1, 2006, and September 30, 2009.<br />
Our search used ‘stem cell treatment’ or ‘stem<br />
cell therapy’ and one or more terms related to<br />
overseas travel (e.g., overseas, abroad, China<br />
or India; Supplementary Methods). Our<br />
data set contained a combined 445 articles<br />
across countries and outlets (Supplementary<br />
Table 1). Our coding frame was informed by<br />
previous work that identified communication<br />
issues associated with the stem cell tourism<br />
phenomenon, particularly issues relevant to<br />
representations of efficacy and risk 3,4 . The<br />
coding frame included patient demographic<br />
information, clinic location, treatment details,<br />
cost, donation information, discussions of<br />
policy, relevant risks, state of the science,<br />
presence of patient testimonials and the tone<br />
of the article. Given the subjectivity of news<br />
content analysis, we tested 10% of the articles<br />
for inter-coder reliability using Cohen’s kappa,<br />
which produced a mean score of k = 0.857<br />
(Supplementary Methods).<br />
We found coverage of stem cell tourism<br />
in each country. The majority of articles<br />
(234 or 52.6%) stemmed from the United<br />
Kingdom, followed by the United States with<br />
99 articles (22.2%), Australia with 74 articles<br />
(16.6%), New Zealand with 21 articles (4.7%)<br />
and Canada with 17 articles (3.8%). These<br />
445 articles discussed 224 different patients<br />
and, as a result, we were able to obtain some<br />
interesting patient demographic information<br />
(Table 1). Of course, there are various<br />
limitations with these data; newspapers<br />
represent only one form of media and our data<br />
only reflect individuals whose stories were<br />
covered by the press captured in our sample<br />
currently available in the Factiva database.<br />
In total, over 19 different countries<br />
were discussed as treatment destinations,<br />
supporting the findings of earlier research<br />
indicating the global nature of this<br />
phenomenon 4,5 . Destinations in addition to<br />
those outlined included India, the Dominican<br />
Republic, the Netherlands, Thailand, Russia,<br />
Costa Rica, the United States, Portugal,<br />
the Ukraine, Israel, the United Kingdom,<br />
Argentina, Ecuador, Chile and Brazil. Our<br />
results also verified the high financial costs<br />
associated with these treatments. The average<br />
cost noted was ~$47,315. The lowest was<br />
~$3,500 and the highest (paid for multiple<br />
treatments over time) was $399,687. In<br />
many cases, articles discussed the intense<br />
fundraising efforts made by patients, their<br />
families, friends and supporters. In fact, 134<br />
(30%) of the articles provided readers with<br />
information on how they could donate funds.<br />
Given these treatments’ unproven status and<br />
the inconsistent information about them<br />
that is generally available to the public, the<br />
involvement of the media in helping raise<br />
funds for these efforts adds another potentially<br />
concerning element to the dynamic of stem<br />
cell tourism.<br />
In addition, the high prevalence of minor<br />
patients in our data set seems particularly<br />
worrisome, especially when one considers<br />
the concerns outlined above, including the<br />
potential for adverse events and the lack of<br />
evidence regarding safety or efficacy.<br />
Examining the articles’ content for<br />
discussions of policy (e.g., legality of the<br />
treatment or regulatory environment),<br />
risk (e.g., tumors or rejection), evidence of<br />
treatment efficacy (e.g., cautions from stem<br />
cell scientists or anecdotal reports from<br />
1244 volume 28 number 12 DECEMBER 2010 nature biotechnology
correspondence<br />
© 2010 Nature America, Inc. All rights reserved.<br />
the clinics) and procedural information<br />
(that is, types of stem cells and method<br />
of application) revealed interesting<br />
jurisdictional differences in the relative<br />
prominence of these topics (Fig. 1).<br />
Policy issues arose in just 21% of the<br />
articles, the majority of which raised them<br />
only in passing (e.g., “it’s so experimental<br />
it can’t legally be done in the United States<br />
because it isn’t approved by the Food and<br />
Drug Administration,” which appeared<br />
in an article entitled “Mailia’s Miracle” in<br />
the 24 August 2009 issue of the Tri-City<br />
Herald, Kennewick, WA, USA) as opposed<br />
to including nuanced discussion of policy<br />
issues or debates. An even smaller proportion<br />
of articles (13%) mentioned the concept of<br />
risk. Of those that did, the majority (70.5%)<br />
did not discuss any specific risks; some of the<br />
risks that were identified included infection,<br />
tumors, inflammation, immune rejection<br />
and inadequate screening (Supplementary<br />
Discussion 2). This relative lack of focus<br />
in the print media on policy issues or risks<br />
associated with stem cell tourism as compared<br />
with emerging commentary in the academic<br />
literature on the topic 9,13 reveals an apparent<br />
disconnect between these two realms. It also<br />
raises the question of whether the general<br />
public (which includes prospective stem cell<br />
tourists) is receiving sufficient information to<br />
reach effective decisions and judgments on<br />
the issue.<br />
In contrast to the comparably lower<br />
prevalence of discussions regarding policy<br />
and risk, 225 articles (50.3%) contained<br />
information about the stem cells being used<br />
in the treatment, the application procedures<br />
or both. The following types of stem cells were<br />
referenced: umbilical cord (107), embryonic<br />
(27), bone marrow (27), adult (24), stem<br />
cells from pelvis (2), stem cells from hip<br />
(2), fetal (2), stem cells from fat tissue (1)<br />
and a combination of bone marrow and<br />
umbilical cord blood (5). In total, 20 different<br />
application procedures were discussed. The<br />
five most common included the following:<br />
injection (location not disclosed) (77),<br />
injection into spinal cord and/or bloodstream<br />
(including intravenous injection) (69),<br />
injection into brain (8), intravenous injection<br />
through hairline towards optic nerve (8) and<br />
catheter/injection into heart (8). The average<br />
length of treatment noted was 4–5 weeks, with<br />
a range of only a few days up to 3 months. It<br />
is important to note that these simply reflect<br />
media references. It remains unclear what,<br />
if any, kind of stem cells were injected in the<br />
procedures used at clinics.<br />
It is also surprising that few articles<br />
contained information regarding the efficacy<br />
Percentage of articles<br />
from each country<br />
80<br />
70<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
United<br />
States<br />
Policy<br />
Canada<br />
Risk<br />
Evidence<br />
United<br />
Kingdom<br />
Country<br />
Australia<br />
Stem cells<br />
New<br />
Zealand<br />
Figure 1 Percentage of news media articles<br />
originating for different countries, which focus<br />
on policy, risk, evidence and the procedure<br />
associated with stem cell treatments.<br />
of the stem cell treatment being addressed<br />
(118 articles, ~25%). In addition, the sources<br />
of this information varied widely (e.g., stem<br />
cell scientists and experts, clinics providing<br />
the therapy, patient groups and family<br />
members). The evidence or information<br />
regarding efficacy ranged from being<br />
supportive of the treatment to cautioning<br />
against it (Supplementary Discussion 2). It<br />
is important to note that in many cases, the<br />
discussions of efficacy did not constitute a<br />
particularly important aspect of the article as a<br />
whole, nor were they necessarily predictive of<br />
its overall tone, discussed below.<br />
Interestingly, 227 articles (51%) included<br />
patient testimonials from the patient, friends<br />
or relatives, other patients with similar<br />
conditions, the clinic’s previous patients or<br />
anecdotal reports (e.g., from media sources).<br />
Considering the prominent role patient<br />
testimonials have on many clinic websites,<br />
this confirmation of their significant presence<br />
in media reports as well further highlights<br />
the dominance this method of information<br />
dissemination and promotion may have to the<br />
growth of this market.<br />
Finally, and perhaps most important, the<br />
tone of the discussions is intriguing (Fig. 2).<br />
Our coders determined whether, overall,<br />
each article was positive, neutral or negative<br />
towards stem cell tourism. The results show<br />
that over time, the tone of articles has become<br />
increasingly positive (Supplementary<br />
Table 2). One exception is the period<br />
between October and December 2008<br />
(Fig. 2), which may be linked with the release<br />
of the International Society for Stem Cell<br />
Research (ISSCR)’s Guidelines for the Clinical<br />
Translation of Stem Cells in December, 2008<br />
(ref. 7). Articles coded as positive presented<br />
stem cell therapy as being a hopeful and/<br />
or successful alternative for patients that<br />
indicated little or no limitations or possible<br />
risks of the treatment. Given that the news<br />
media—and newspapers in particular—are<br />
the dominant source of information about<br />
stem cell tourism for the public, this apparent<br />
trend to frame stem cell tourism in terms of<br />
promise and hope with limited emphasis on<br />
uncertainty may have important implications<br />
for public evaluations of clinic claims,<br />
especially among patients and their families.<br />
On the whole, these data suggest that print<br />
media portrayals of stem cell tourism are<br />
largely and increasingly positive in nature.<br />
Furthermore, their focus is primarily on the<br />
individual patient, their hopes and specific<br />
treatment plans or approach, rather than<br />
on the potential risks associated with these<br />
unproven treatments, current scientific<br />
and clinical limitations, or the various<br />
policy issues implicated by this emerging<br />
phenomenon. In many respects, these<br />
results are unsurprising; the circumstances<br />
surrounding an individual’s pursuit of stem<br />
cell tourism often make for a powerful<br />
personal interest story. It is impossible<br />
not to empathize with the plights of those<br />
desperately seeking treatment for themselves<br />
or their loved ones, and important to be wary<br />
of overly broad generalizations of a field that<br />
appears to encompass an incredibly broad<br />
range of potential treatments and treatment<br />
providers. Nonetheless, there are serious<br />
concerns associated with the rise of stem cell<br />
tourism, including physical and financial<br />
risks, lack of transparency and appropriate<br />
review procedures, exploitation of vulnerable<br />
patients (including minors) and various<br />
policy issues, which must not be minimized.<br />
Although some of these issues are making<br />
their way into the print media, it is clear that<br />
continued efforts are necessary to improve<br />
the balance in media reporting of this topic.<br />
When medical experts and organizations<br />
do actively seek to engage journalists and<br />
the public on the uncertainty of stem cell<br />
tourism and the need for regulation, as was<br />
the case in 2008 with the ISSCR’s guidelines,<br />
our findings do point to an influence on<br />
coverage, with the effort balancing the<br />
otherwise overwhelmingly positive portrayal<br />
in the press. This balance is an important<br />
element in promoting a knowledgeable<br />
public, necessary both for facilitating<br />
Percentage of<br />
10/06–12/06all articles<br />
100<br />
90<br />
80<br />
70<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
01/07–03/07<br />
04/07–06/07<br />
Positive<br />
Neutral<br />
Negative<br />
07/07–09/07<br />
10/07–12/07<br />
01/08–03/08<br />
04/08–06/08<br />
07/08–09/08<br />
10/08–12/08<br />
01/09–04/09<br />
04/09–06/09<br />
07/09–09/09<br />
Figure 2 Overall ‘slant’ of different news media<br />
reports that covered stem cell treatments and<br />
appeared from October 2006 to September 2009.<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1245
correspondence<br />
© 2010 Nature America, Inc. All rights reserved.<br />
informed debate regarding stem cell tourism<br />
and for protecting potentially vulnerable<br />
individuals.<br />
Note: Supplementary information is available on the<br />
Nature Biotechnology website.<br />
Acknowledgments<br />
We would like to thank Canada’s Stem Cell Network<br />
for funding and the University of Alberta’s Health Law<br />
Institute for administrative support. We also gratefully<br />
acknowledge C. Scott, J. McCormick, T. Bubela and<br />
C. Murdoch for their helpful suggestions regarding data<br />
collection. Finally, we thank C. Toole for her research<br />
support, and G. Barr and T. Adido for their assistance in<br />
coding the data.<br />
AUTHOR CONTRIBUTIONS<br />
All authors contributed to this work. T.C., C.R. and<br />
A.Z. conceived the research concept. C.R. collected the<br />
data. C.R. and A.Z. coordinated the coding of the data.<br />
A.Z. and T.C. prepared the draft manuscript and M.N.<br />
helped with interpretation and with theoretical analysis.<br />
All authors discussed the results and implications and<br />
commented on the manuscript at all stages.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
Amy Zarzeczny 1 , Christen Rachul 1 ,<br />
Matthew Nisbet 2 & Timothy Caulfield 1,3<br />
1 Health Law Institute, Law Centre, University of<br />
Alberta, Edmonton, Alberta, Canada. 2 School<br />
of Communication, American University,<br />
Washington, DC, USA 3 . Faculty of Law and<br />
School of Public Health, Law Centre, University<br />
of Alberta, Edmonton, Alberta, Canada.<br />
e-mail: tcaulfld@law.ualberta.ca<br />
1. Qui, J. Nat. Biotechnol. 27, 790–792 (2009)<br />
2. Lindvall, O. & Hyun, I. Science 324, 1664–1665<br />
(2009).<br />
3. Lau, D. et al. Cell Stem Cell 3, 591–594 (2008).<br />
4. Regenberg, A., Hutchinson, L., Schanker, B. & Matthews, D.<br />
Stem Cells 27, 2312–2319 (2009).<br />
5. Ryan, K., Sanders, A., Wang, D. & Levine, A.<br />
Regenerative Med. 5, 27–33 (2010).<br />
6. Amariglio, N. et al. PLoS Med 6, e1000029 (2009).<br />
7. http://www.isscr.org/clinical_trans/pdfs/ISSCRGL<br />
ClinicalTrans.pdf<br />
8. MacReady, N. Lancet Oncol. 10, 317–318 (2009).<br />
9. Kiatpongsan, S. & Sipp, D. Nat. Rep. Stem Cells<br />
published online 3 December, 2008, doi:10.1038/<br />
stemcells.2008.151 (2008).<br />
10. Zarzeczny, A. & Caulfield, T. Am. J. Bioeth. 10, 1–13<br />
(2010).<br />
11. Iyengar, S. & Kinder, D. News that Matters: Television<br />
and American Public Opinion. (University of Chicago<br />
Press, Chicago, IL, USA, 1987).<br />
12. Nisbet, M.C. & Mooney, C. Science, 316, 56<br />
(2007).<br />
13. Creasy, G. & Scott, C. Nat. Biotechnol. 27, 21–22<br />
(2009).<br />
Tracking and assessing the rise of<br />
state-funded stem cell research<br />
To the Editor:<br />
The editorial in your October issue 1<br />
highlights the legal challenges to new<br />
guidelines issued by the US National<br />
Institutes of Health (NIH) in July 2009 for the<br />
federal funding of human embryonic stem<br />
cell (hESC) research. In the eight-year period<br />
preceding these most recent NIH guidelines,<br />
only a small number of cell lines could be<br />
studied with federal funds. During this time,<br />
six states—California, Connecticut, Illinois,<br />
Maryland, New Jersey and New York—took<br />
on a role typically played by the NIH and<br />
created funding programs specifically<br />
designed to support stem cell research,<br />
including hESC research. These are not the<br />
first state programs to fund scientific research<br />
but their commitment to basic research is<br />
atypical, as most state science and technology<br />
programs have focused on science closer to<br />
commercialization 2 . Although the state stem<br />
cell programs differ, they each share at least<br />
two goals: advancing promising science,<br />
including research not eligible for federal<br />
funding during the Bush Administration, and<br />
returning economic benefits to their state.<br />
In this article, we report an initial<br />
attempt to track and assess the impact of<br />
these state funding programs. Existing<br />
work on state stem cell policy has focused<br />
on identifying policy differences between<br />
various jurisdictions 3,4 , assessing the impact<br />
of state decisions to support or restrict hESC<br />
science 5–7 and examining the role of states<br />
in governing controversial science 8,9 . The<br />
analysis reported here extends this literature<br />
though use of a novel data set of the grants<br />
these states have awarded. These data provide<br />
insight into how states have prioritized their<br />
funding, including the extent to which they<br />
have supported hESC research generally<br />
and hESC research not eligible for federal<br />
funding during the Bush Administration<br />
more specifically, as well as the extent to<br />
which these states have drawn new scientists<br />
into the field. The underlying data have been<br />
publicly released on a new website (http://<br />
www.stemcellstates.net) designed to facilitate<br />
additional analysis of state-funded stem<br />
cell science and improve public awareness<br />
of these programs. Given ongoing legal<br />
uncertainties surrounding federal funding<br />
of hESC research and the likelihood that<br />
voters, at least in California, will be asked to<br />
approve additional state stem cell funding in<br />
the future, understanding and evaluating the<br />
effects of state-funded stem cell research is<br />
both timely and useful.<br />
The database that forms the basis for the<br />
analysis described here contains the title,<br />
principal investigator, institution, abstract<br />
and amount for each grant awarded by<br />
the agency overseeing stem cell research<br />
funding in these six states (Supplementary<br />
Methods). In all, between December 2005<br />
when New Jersey awarded the first state stem<br />
cell grants and the end of 2009, the six stem<br />
cell states awarded nearly 750 grants totaling<br />
just over $1.25 billion. The scale of these<br />
programs varies substantially, ranging from<br />
the roughly $15 million awarded by Illinois<br />
and New Jersey to the $1.02 billion awarded<br />
by California. On a per capita basis, funding<br />
awarded through the end of 2009 ranges<br />
from just over $1 in Illinois to nearly $28 in<br />
California (Table 1).<br />
States funding stem cell research<br />
can choose to support several different<br />
activities, ranging from investigatorinitiated<br />
research grants to new facilities<br />
to workforce development. To investigate<br />
how states prioritized these various types<br />
of funding, each grant was classified by<br />
its primary purpose (Supplementary<br />
Methods). Research grants and support for<br />
scientific infrastructure were the two largest<br />
categories, accounting for more than 90%<br />
of all state stem cell funding (Table 1). The<br />
infrastructure category was dominated by<br />
the $271 million California awarded for the<br />
construction of 12 major stem cell research<br />
facilities, although several other states also<br />
dedicated a substantial portion of their<br />
funding to infrastructure, such as shared<br />
equipment or core laboratories. In contrast<br />
to supporting basic investigator-initiated<br />
research, spending money on infrastructure<br />
is a classic state economic development<br />
approach, but, in these cases, spending<br />
was motivated, at least in part, by the need<br />
to create separate laboratories to facilitate<br />
research on unapproved hESC lines.<br />
The restrictions on federal funding for<br />
hESC research instituted by former President<br />
George W. Bush were an important rationale<br />
behind the adoption of most state stem cell<br />
programs, yet it is not clear to what extent<br />
state programs focused on hESC research<br />
generally or hESC research not eligible for<br />
federal funding more specifically. To address<br />
these questions, each research grant awarded<br />
through the end of 2009 was analyzed to<br />
assess if funded research used hESCs and,<br />
if applicable, appeared ineligible for federal<br />
funding under the Bush Administration<br />
rules (Supplementary Methods). The<br />
percentage of grants that supported hESC<br />
1246 volume 28 number 12 DECEMBER 2010 nature biotechnology
correspondence<br />
© 2010 Nature America, Inc. All rights reserved.<br />
informed debate regarding stem cell tourism<br />
and for protecting potentially vulnerable<br />
individuals.<br />
Note: Supplementary information is available on the<br />
Nature Biotechnology website.<br />
Acknowledgments<br />
We would like to thank Canada’s Stem Cell Network<br />
for funding and the University of Alberta’s Health Law<br />
Institute for administrative support. We also gratefully<br />
acknowledge C. Scott, J. McCormick, T. Bubela and<br />
C. Murdoch for their helpful suggestions regarding data<br />
collection. Finally, we thank C. Toole for her research<br />
support, and G. Barr and T. Adido for their assistance in<br />
coding the data.<br />
AUTHOR CONTRIBUTIONS<br />
All authors contributed to this work. T.C., C.R. and<br />
A.Z. conceived the research concept. C.R. collected the<br />
data. C.R. and A.Z. coordinated the coding of the data.<br />
A.Z. and T.C. prepared the draft manuscript and M.N.<br />
helped with interpretation and with theoretical analysis.<br />
All authors discussed the results and implications and<br />
commented on the manuscript at all stages.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
Amy Zarzeczny 1 , Christen Rachul 1 ,<br />
Matthew Nisbet 2 & Timothy Caulfield 1,3<br />
1 Health Law Institute, Law Centre, University of<br />
Alberta, Edmonton, Alberta, Canada. 2 School<br />
of Communication, American University,<br />
Washington, DC, USA 3 . Faculty of Law and<br />
School of Public Health, Law Centre, University<br />
of Alberta, Edmonton, Alberta, Canada.<br />
e-mail: tcaulfld@law.ualberta.ca<br />
1. Qui, J. Nat. Biotechnol. 27, 790–792 (2009)<br />
2. Lindvall, O. & Hyun, I. Science 324, 1664–1665<br />
(2009).<br />
3. Lau, D. et al. Cell Stem Cell 3, 591–594 (2008).<br />
4. Regenberg, A., Hutchinson, L., Schanker, B. & Matthews, D.<br />
Stem Cells 27, 2312–2319 (2009).<br />
5. Ryan, K., Sanders, A., Wang, D. & Levine, A.<br />
Regenerative Med. 5, 27–33 (2010).<br />
6. Amariglio, N. et al. PLoS Med 6, e1000029 (2009).<br />
7. http://www.isscr.org/clinical_trans/pdfs/ISSCRGL<br />
ClinicalTrans.pdf<br />
8. MacReady, N. Lancet Oncol. 10, 317–318 (2009).<br />
9. Kiatpongsan, S. & Sipp, D. Nat. Rep. Stem Cells<br />
published online 3 December, 2008, doi:10.1038/<br />
stemcells.2008.151 (2008).<br />
10. Zarzeczny, A. & Caulfield, T. Am. J. Bioeth. 10, 1–13<br />
(2010).<br />
11. Iyengar, S. & Kinder, D. News that Matters: Television<br />
and American Public Opinion. (University of Chicago<br />
Press, Chicago, IL, USA, 1987).<br />
12. Nisbet, M.C. & Mooney, C. Science, 316, 56<br />
(2007).<br />
13. Creasy, G. & Scott, C. Nat. Biotechnol. 27, 21–22<br />
(2009).<br />
Tracking and assessing the rise of<br />
state-funded stem cell research<br />
To the Editor:<br />
The editorial in your October issue 1<br />
highlights the legal challenges to new<br />
guidelines issued by the US National<br />
Institutes of Health (NIH) in July 2009 for the<br />
federal funding of human embryonic stem<br />
cell (hESC) research. In the eight-year period<br />
preceding these most recent NIH guidelines,<br />
only a small number of cell lines could be<br />
studied with federal funds. During this time,<br />
six states—California, Connecticut, Illinois,<br />
Maryland, New Jersey and New York—took<br />
on a role typically played by the NIH and<br />
created funding programs specifically<br />
designed to support stem cell research,<br />
including hESC research. These are not the<br />
first state programs to fund scientific research<br />
but their commitment to basic research is<br />
atypical, as most state science and technology<br />
programs have focused on science closer to<br />
commercialization 2 . Although the state stem<br />
cell programs differ, they each share at least<br />
two goals: advancing promising science,<br />
including research not eligible for federal<br />
funding during the Bush Administration, and<br />
returning economic benefits to their state.<br />
In this article, we report an initial<br />
attempt to track and assess the impact of<br />
these state funding programs. Existing<br />
work on state stem cell policy has focused<br />
on identifying policy differences between<br />
various jurisdictions 3,4 , assessing the impact<br />
of state decisions to support or restrict hESC<br />
science 5–7 and examining the role of states<br />
in governing controversial science 8,9 . The<br />
analysis reported here extends this literature<br />
though use of a novel data set of the grants<br />
these states have awarded. These data provide<br />
insight into how states have prioritized their<br />
funding, including the extent to which they<br />
have supported hESC research generally<br />
and hESC research not eligible for federal<br />
funding during the Bush Administration<br />
more specifically, as well as the extent to<br />
which these states have drawn new scientists<br />
into the field. The underlying data have been<br />
publicly released on a new website (http://<br />
www.stemcellstates.net) designed to facilitate<br />
additional analysis of state-funded stem<br />
cell science and improve public awareness<br />
of these programs. Given ongoing legal<br />
uncertainties surrounding federal funding<br />
of hESC research and the likelihood that<br />
voters, at least in California, will be asked to<br />
approve additional state stem cell funding in<br />
the future, understanding and evaluating the<br />
effects of state-funded stem cell research is<br />
both timely and useful.<br />
The database that forms the basis for the<br />
analysis described here contains the title,<br />
principal investigator, institution, abstract<br />
and amount for each grant awarded by<br />
the agency overseeing stem cell research<br />
funding in these six states (Supplementary<br />
Methods). In all, between December 2005<br />
when New Jersey awarded the first state stem<br />
cell grants and the end of 2009, the six stem<br />
cell states awarded nearly 750 grants totaling<br />
just over $1.25 billion. The scale of these<br />
programs varies substantially, ranging from<br />
the roughly $15 million awarded by Illinois<br />
and New Jersey to the $1.02 billion awarded<br />
by California. On a per capita basis, funding<br />
awarded through the end of 2009 ranges<br />
from just over $1 in Illinois to nearly $28 in<br />
California (Table 1).<br />
States funding stem cell research<br />
can choose to support several different<br />
activities, ranging from investigatorinitiated<br />
research grants to new facilities<br />
to workforce development. To investigate<br />
how states prioritized these various types<br />
of funding, each grant was classified by<br />
its primary purpose (Supplementary<br />
Methods). Research grants and support for<br />
scientific infrastructure were the two largest<br />
categories, accounting for more than 90%<br />
of all state stem cell funding (Table 1). The<br />
infrastructure category was dominated by<br />
the $271 million California awarded for the<br />
construction of 12 major stem cell research<br />
facilities, although several other states also<br />
dedicated a substantial portion of their<br />
funding to infrastructure, such as shared<br />
equipment or core laboratories. In contrast<br />
to supporting basic investigator-initiated<br />
research, spending money on infrastructure<br />
is a classic state economic development<br />
approach, but, in these cases, spending<br />
was motivated, at least in part, by the need<br />
to create separate laboratories to facilitate<br />
research on unapproved hESC lines.<br />
The restrictions on federal funding for<br />
hESC research instituted by former President<br />
George W. Bush were an important rationale<br />
behind the adoption of most state stem cell<br />
programs, yet it is not clear to what extent<br />
state programs focused on hESC research<br />
generally or hESC research not eligible for<br />
federal funding more specifically. To address<br />
these questions, each research grant awarded<br />
through the end of 2009 was analyzed to<br />
assess if funded research used hESCs and,<br />
if applicable, appeared ineligible for federal<br />
funding under the Bush Administration<br />
rules (Supplementary Methods). The<br />
percentage of grants that supported hESC<br />
1246 volume 28 number 12 DECEMBER 2010 nature biotechnology
correspondence<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Table 1 The scale and prioritization of state stem cell funding programs a<br />
research varied substantially among these<br />
states (Table 1). Large majorities of the<br />
research grants awarded in Connecticut<br />
and California supported studies involving<br />
hESCs, whereas only a minority of grants<br />
supported hESC research in the other states.<br />
These disparities likely reflect differences<br />
in the types of stem cell scientists present in<br />
these states as well as priorities of the various<br />
state funding bodies.<br />
Only a subset of grants for hESC<br />
research supported science that was<br />
clearly ineligible for NIH funding during<br />
the Bush Administration. California and<br />
Connecticut focused the most on this sort<br />
of research—which typically involved the<br />
derivation of new hESC lines or the use<br />
of newer unapproved cell lines—but even<br />
in these states fewer than a fifth of grants<br />
went to clearly ineligible research. Many<br />
scientists indicated plans to use existing<br />
hESC lines but did not specify which lines<br />
they planned to use. Given evidence that a<br />
handful of approved cell lines account for a<br />
large proportion of the hESC lines actually<br />
distributed to scientists and an even larger<br />
share of published literature 10 , most of these<br />
projects probably used approved hESC lines.<br />
In some cases, however, scientists may have<br />
chosen to use ineligible cell lines but not<br />
clearly indicated these plans. Thus, the share<br />
of grants reported here as clearly ineligible<br />
for NIH funding should be viewed as a lower<br />
bound on the amount of research each state<br />
funded that was ineligible for federal funding.<br />
Several factors could explain the relatively<br />
small share of grants that went toward clearly<br />
ineligible research. Some scientists who<br />
wished to pursue this research may have been<br />
unable to access the raw materials or acquire<br />
the intellectual property rights required to do<br />
so. Alternatively, these findings could simply<br />
reflect scientific interest. The discovery of<br />
induced pluripotent stem cells 11 may, for<br />
instance, have reduced scientific interest in<br />
the derivation of new hESC lines. Finally,<br />
these findings may reflect a preference on the<br />
part of scientists to use well-established and<br />
well-studied hESC lines. This last explanation<br />
may be particularly relevant for new scientists<br />
entering the field of hESC research, as using<br />
recognized cell lines may give their initial<br />
research efforts greater credibility.<br />
In addition to supporting research not<br />
eligible for federal funding, focused state<br />
programs might serve to draw new scientists<br />
into the field of stem cell research. To<br />
evaluate this potential impact, the recent NIH<br />
funding portfolio of each scientist receiving a<br />
state stem cell grant with a primary purpose<br />
of research was examined (Supplementary<br />
Methods). Although most scientists had<br />
received NIH funding, a substantial number<br />
(ranging from 42% in California to 71% in<br />
Maryland) had not received NIH funding<br />
for stem cell research (Table 1). Similar,<br />
but more pronounced, results are observed<br />
when the NIH funding portfolio of scientists<br />
receiving state funding for hESC research is<br />
examined, as only a small minority of these<br />
scientists also had NIH grants supporting<br />
hESC research. Given the importance of<br />
NIH funding for biomedical research in the<br />
United States, these results suggest that the<br />
existence of state funding programs for stem<br />
cell research has drawn many new scientists<br />
into the field of stem cell research, or at least<br />
encouraged scientists to consider how stem<br />
State program<br />
Grants and funding California Connecticut Illinois Maryland New Jersey New York<br />
Year first grants awarded 2006 2006 2006 2007 2005 2008<br />
Total funding pledged/time period<br />
$3 billlion/<br />
10 years<br />
$100 million/<br />
10 years<br />
N/A N/A N/A $600 million/<br />
11 years<br />
Number of grants awarded 329 69 17 140 35 158<br />
Funding awarded ($ millions) 1,024 40 15 54 15 121<br />
Funding per capita b ($) 28 11 1 10 2 6<br />
Research prioritization c<br />
Percentage of funding for research 58% 76% 100% 93% 64% 61%<br />
Percentage of funding for infrastructure 31% 24% 0% 0% 36% 34%<br />
Percentage of grants for hESC research 75% 97% 35% 42% 21% 21%<br />
Percentage of grants clearly not NIH eligible 18% 16% 12% 3% 6% 0%<br />
NIH funding status of state grant recipients d<br />
Percentage of state PIs without NIH stem cell funding 42% 61% 65% 71% 61% 49%<br />
Percentage of state hESC PIs without NIH hESC funding 77% 91% 67% 79% 100% 66%<br />
a Includes grants awarded through the end of 2009. b Per capita funding based on state population from US Census Bureau 2009 Population Estimates. c hESC prioritization analysis includes<br />
only grants with a primary purpose of research. d NIH funding was examined from FY2005 to present. PIs, principal investigators.<br />
cell research could complement their existing<br />
research programs.<br />
These data also permit a more nuanced<br />
comparison between state stem cell funding<br />
and NIH stem cell funding than has<br />
previously been available (Fig. 1). Total state<br />
funding for all types of stem cell research has<br />
risen rapidly since grants were first awarded<br />
in 2005, but states still spend less than half of<br />
what the NIH spends each year on stem cell<br />
research. The situation is different for hESC<br />
research, as state funding for hESC research<br />
grants first exceeded comparable NIH<br />
funding in 2007 and equaled or exceeded it in<br />
2008 and 2009.<br />
Considered together, these data and<br />
analyses indicate that state funding for stem<br />
cell research has grown into a substantial<br />
enterprise that has provided funding on a<br />
scale comparable to the NIH. Although states<br />
vary in the degree to which they have focused<br />
on hESC research, as a whole, state funding<br />
for hESC research has been substantial,<br />
exceeding, in cumulative terms, NIH funding<br />
for this research between 2005 and 2009.<br />
Most state hESC funding appears to have<br />
supported research also eligible for federal<br />
funding during the Bush Administration.<br />
This finding is surprising, given the<br />
explicit intent of several state programs to<br />
preferentially support science not eligible for<br />
federal funding, but likely reflects the nature<br />
of the grant proposals state agencies received,<br />
particularly given the number of grants states<br />
awarded to scientists relatively new to the<br />
field of hESC research.<br />
In the light of the recent change in<br />
federal stem cell policy and the ongoing<br />
economic downturn, the future of state<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1247
correspondence<br />
© 2010 Nature America, Inc. All rights reserved.<br />
a<br />
Funding awarded<br />
($ millions)<br />
1,250<br />
1,000<br />
750<br />
500<br />
250<br />
0<br />
609<br />
2005<br />
5<br />
643<br />
2006<br />
Figure 1 Comparing state and NIH stem cell funding. (a) Total amount of all NIH stem cell grants and<br />
all stem cell grants awarded by the six states. (b) Total amount of all NIH and state hESC research<br />
grants. Only grants with a primary purpose of research are included. State funding is by calendar year.<br />
NIH funding is by fiscal year.<br />
stem cell programs, as well as similar state<br />
programs supporting other areas of science,<br />
is uncertain. The analysis here suggests<br />
that state stem cell funding programs<br />
are sufficiently large and established that<br />
simply ending the programs, at least in<br />
the absence of substantial investment in<br />
the field by other funding sources, could<br />
have deleterious effects. Such action would<br />
fail to capitalize on the initial efforts of<br />
scientists who have been drawn to the field<br />
of stem cell research by state programs and<br />
leave many stem cell scientists suddenly<br />
searching for funding to continue their<br />
research.<br />
Large-scale state funding for basic<br />
research is a relatively new phenomenon,<br />
and many questions remain about<br />
the impact of these programs on the<br />
development of scientific fields and the<br />
careers of scientists. The influence of state<br />
funding programs on the distribution<br />
of research publications, the acquisition<br />
of future external funding, the creation<br />
of new companies and the translation<br />
of basic research into medical practice,<br />
for instance, are important unanswered<br />
questions. Similarly, comparing state<br />
funding programs with federal funding<br />
programs as well as foundations could offer<br />
new insight into the relative priorities of<br />
different funding bodies and the extent<br />
to which their funding portfolios overlap<br />
or are distinct. We hope the analysis<br />
presented here and the public release of the<br />
underlying database will inspire additional<br />
analysis of state science funding programs<br />
generally and state-funded stem cell science<br />
in particular.<br />
Note: Supplementary information is available on the<br />
Nature Biotechnology website.<br />
Acknowledgements<br />
The authors gratefully acknowledge financial<br />
support from the Roadmap for an Entrepreneurial<br />
Economy Program, funded by the Kauffman<br />
73<br />
657<br />
2007<br />
246<br />
938<br />
2008<br />
426<br />
1,231<br />
2009<br />
519<br />
NIH grants<br />
b<br />
Funding awarded<br />
($ millions)<br />
180<br />
150<br />
120<br />
90<br />
60<br />
30<br />
0<br />
33<br />
2005<br />
1<br />
State grants<br />
31<br />
2006<br />
21<br />
Foundation and the Georgia Research Alliance,<br />
and Georgia Tech. They thank J. Walsh at Georgia<br />
Tech for helpful comments on an earlier version of<br />
this manuscript. They also appreciate the assistance<br />
they received with data collection from officials<br />
in various state stem cell agencies. A.D.L. would<br />
also like to thank A. Jakimo, whose comment at<br />
a meeting of the Interstate Alliance on Stem Cell<br />
Research inspired collection of these data.<br />
35<br />
2007<br />
76<br />
2008<br />
144<br />
125<br />
To the Editor:<br />
We report on the launch of version 7 of<br />
the Human Protein Atlas with subcellular<br />
localization data and expression data<br />
for all major human tissues and organs.<br />
A milestone has been achieved with the<br />
inclusion of expression data for >50% of the<br />
human protein-coding genes. The main new<br />
feature of the release is an attempt towards<br />
a knowledge-based portal, including an<br />
annotated protein expression feature for<br />
protein targets analyzed with two or more<br />
antibodies, and the establishment of the main<br />
subcellular localization of protein targets.<br />
In 2005, the first version of the Human<br />
Protein Atlas (http://www.proteinatlas.<br />
org/) was released with protein profile<br />
data based on immunohistochemistry on<br />
tissue microarrays covering 48 different<br />
human tissues and organs, including kidney,<br />
liver, heart, brain and pancreas 1 . The first<br />
version included data from 718 antibodies<br />
corresponding to 650 human protein-coding<br />
genes. High-resolution images were published<br />
along with annotation of the presence or<br />
absence of a particular protein target in all<br />
represented tissues. The 2005 Human Protein<br />
Atlas also contained information regarding<br />
protein profiles from 20 different types of<br />
human cancer, including breast, colorectal,<br />
155<br />
76<br />
2009<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
Ruchir N Karmali 1 , Natalie M Jones 1 &<br />
Aaron D Levine 1,2<br />
1 School of Public Policy, Georgia Institute of<br />
Technology, Atlanta, Georgia, USA. 2 Institute of<br />
Bioengineering & Bioscience, Georgia Institute<br />
of Technology, Atlanta, Georgia, USA.<br />
e-mail: aaron.levine@pubpolicy.gatech.edu<br />
1. Anonymous. Nat. Biotechnol. 28, 987 (2010).<br />
2. Plosila, W.H. Econ. Dev. Q. 18, 113–126 (2004).<br />
3. Stayn, S. BNA Med. Law Pol. Rep. 5, 718–725<br />
(2006).<br />
4. Lomax, G. & Stayn, S. BNA Med. Law Pol. Rep. 7,<br />
695–698 (2008).<br />
5. Levine, A.D. Public Adm. Rev. 68, 681–694 (2008).<br />
6. Levine, A.D. Nat. Biotechnol. 24, 865–866 (2006).<br />
7. McCormick, J.B., Owen-Smith, J. & Scott, C.T. Cell<br />
Stem Cell 4, 107–110 (2009).<br />
8. Fossett, J.W., Ouellette, A.R., Philpott, S., Magnus, D.<br />
& Mcgee, G. Hastings Cent. Rep. 37, 24–35 (2007).<br />
9. Mintrom, M. Publius 39, 606–631 (2009).<br />
10. Scott, C.T., McCormick, J.B. & Owen-Smith, J. Nat.<br />
Biotechnol. 27, 696–697 (2009).<br />
11. Takahashi, K. & Yamanaka, S. Cell 126, 663–676<br />
(2006).<br />
Towards a knowledge-based Human<br />
Protein Atlas<br />
lung and prostate cancer. The data in the<br />
portal were made available freely both for<br />
academia and industry without restrictions or<br />
password protection. In 2007, the portal was<br />
extended to also include subcellular profiling<br />
data 2 using immunofluorescence-based<br />
confocal microscopy in three human cancer<br />
cell lines of different (glial, mesenchymal and<br />
epithelial) origin. More data have been added<br />
to the portal every year since the first release 3<br />
and version 6, launched in March 2010,<br />
contained 11,274 antibodies corresponding to<br />
8,489 protein-coding genes. This entire effort<br />
depends heavily on the availability of good<br />
quality antibodies, and recently a communitybased<br />
portal, Antibodypedia (http://www.<br />
antibodypedia.org/), has been launched to<br />
allow antibodies from different providers to<br />
be listed and compared 4,5 , although the main<br />
source of information so far comes from<br />
the providers’ own validation data, not by<br />
independent third-party users. At present,<br />
the Antibodypedia contains close to 100,000<br />
antibodies, corresponding to >70% of the<br />
protein-coding genes in humans.<br />
An important objective has now been<br />
reached with the inclusion of 10,118 proteincoding<br />
genes corresponding to >50% of the<br />
19,559 human entries as defined by UniProt,<br />
including only entries with evidence at protein<br />
1248 volume 28 number 12 DECEMBER 2010 nature biotechnology
correspondence<br />
© 2010 Nature America, Inc. All rights reserved.<br />
a<br />
Funding awarded<br />
($ millions)<br />
1,250<br />
1,000<br />
750<br />
500<br />
250<br />
0<br />
609<br />
2005<br />
5<br />
643<br />
2006<br />
Figure 1 Comparing state and NIH stem cell funding. (a) Total amount of all NIH stem cell grants and<br />
all stem cell grants awarded by the six states. (b) Total amount of all NIH and state hESC research<br />
grants. Only grants with a primary purpose of research are included. State funding is by calendar year.<br />
NIH funding is by fiscal year.<br />
stem cell programs, as well as similar state<br />
programs supporting other areas of science,<br />
is uncertain. The analysis here suggests<br />
that state stem cell funding programs<br />
are sufficiently large and established that<br />
simply ending the programs, at least in<br />
the absence of substantial investment in<br />
the field by other funding sources, could<br />
have deleterious effects. Such action would<br />
fail to capitalize on the initial efforts of<br />
scientists who have been drawn to the field<br />
of stem cell research by state programs and<br />
leave many stem cell scientists suddenly<br />
searching for funding to continue their<br />
research.<br />
Large-scale state funding for basic<br />
research is a relatively new phenomenon,<br />
and many questions remain about<br />
the impact of these programs on the<br />
development of scientific fields and the<br />
careers of scientists. The influence of state<br />
funding programs on the distribution<br />
of research publications, the acquisition<br />
of future external funding, the creation<br />
of new companies and the translation<br />
of basic research into medical practice,<br />
for instance, are important unanswered<br />
questions. Similarly, comparing state<br />
funding programs with federal funding<br />
programs as well as foundations could offer<br />
new insight into the relative priorities of<br />
different funding bodies and the extent<br />
to which their funding portfolios overlap<br />
or are distinct. We hope the analysis<br />
presented here and the public release of the<br />
underlying database will inspire additional<br />
analysis of state science funding programs<br />
generally and state-funded stem cell science<br />
in particular.<br />
Note: Supplementary information is available on the<br />
Nature Biotechnology website.<br />
Acknowledgements<br />
The authors gratefully acknowledge financial<br />
support from the Roadmap for an Entrepreneurial<br />
Economy Program, funded by the Kauffman<br />
73<br />
657<br />
2007<br />
246<br />
938<br />
2008<br />
426<br />
1,231<br />
2009<br />
519<br />
NIH grants<br />
b<br />
Funding awarded<br />
($ millions)<br />
180<br />
150<br />
120<br />
90<br />
60<br />
30<br />
0<br />
33<br />
2005<br />
1<br />
State grants<br />
31<br />
2006<br />
21<br />
Foundation and the Georgia Research Alliance,<br />
and Georgia Tech. They thank J. Walsh at Georgia<br />
Tech for helpful comments on an earlier version of<br />
this manuscript. They also appreciate the assistance<br />
they received with data collection from officials<br />
in various state stem cell agencies. A.D.L. would<br />
also like to thank A. Jakimo, whose comment at<br />
a meeting of the Interstate Alliance on Stem Cell<br />
Research inspired collection of these data.<br />
35<br />
2007<br />
76<br />
2008<br />
144<br />
125<br />
To the Editor:<br />
We report on the launch of version 7 of<br />
the Human Protein Atlas with subcellular<br />
localization data and expression data<br />
for all major human tissues and organs.<br />
A milestone has been achieved with the<br />
inclusion of expression data for >50% of the<br />
human protein-coding genes. The main new<br />
feature of the release is an attempt towards<br />
a knowledge-based portal, including an<br />
annotated protein expression feature for<br />
protein targets analyzed with two or more<br />
antibodies, and the establishment of the main<br />
subcellular localization of protein targets.<br />
In 2005, the first version of the Human<br />
Protein Atlas (http://www.proteinatlas.<br />
org/) was released with protein profile<br />
data based on immunohistochemistry on<br />
tissue microarrays covering 48 different<br />
human tissues and organs, including kidney,<br />
liver, heart, brain and pancreas 1 . The first<br />
version included data from 718 antibodies<br />
corresponding to 650 human protein-coding<br />
genes. High-resolution images were published<br />
along with annotation of the presence or<br />
absence of a particular protein target in all<br />
represented tissues. The 2005 Human Protein<br />
Atlas also contained information regarding<br />
protein profiles from 20 different types of<br />
human cancer, including breast, colorectal,<br />
155<br />
76<br />
2009<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
Ruchir N Karmali 1 , Natalie M Jones 1 &<br />
Aaron D Levine 1,2<br />
1 School of Public Policy, Georgia Institute of<br />
Technology, Atlanta, Georgia, USA. 2 Institute of<br />
Bioengineering & Bioscience, Georgia Institute<br />
of Technology, Atlanta, Georgia, USA.<br />
e-mail: aaron.levine@pubpolicy.gatech.edu<br />
1. Anonymous. Nat. Biotechnol. 28, 987 (2010).<br />
2. Plosila, W.H. Econ. Dev. Q. 18, 113–126 (2004).<br />
3. Stayn, S. BNA Med. Law Pol. Rep. 5, 718–725<br />
(2006).<br />
4. Lomax, G. & Stayn, S. BNA Med. Law Pol. Rep. 7,<br />
695–698 (2008).<br />
5. Levine, A.D. Public Adm. Rev. 68, 681–694 (2008).<br />
6. Levine, A.D. Nat. Biotechnol. 24, 865–866 (2006).<br />
7. McCormick, J.B., Owen-Smith, J. & Scott, C.T. Cell<br />
Stem Cell 4, 107–110 (2009).<br />
8. Fossett, J.W., Ouellette, A.R., Philpott, S., Magnus, D.<br />
& Mcgee, G. Hastings Cent. Rep. 37, 24–35 (2007).<br />
9. Mintrom, M. Publius 39, 606–631 (2009).<br />
10. Scott, C.T., McCormick, J.B. & Owen-Smith, J. Nat.<br />
Biotechnol. 27, 696–697 (2009).<br />
11. Takahashi, K. & Yamanaka, S. Cell 126, 663–676<br />
(2006).<br />
Towards a knowledge-based Human<br />
Protein Atlas<br />
lung and prostate cancer. The data in the<br />
portal were made available freely both for<br />
academia and industry without restrictions or<br />
password protection. In 2007, the portal was<br />
extended to also include subcellular profiling<br />
data 2 using immunofluorescence-based<br />
confocal microscopy in three human cancer<br />
cell lines of different (glial, mesenchymal and<br />
epithelial) origin. More data have been added<br />
to the portal every year since the first release 3<br />
and version 6, launched in March 2010,<br />
contained 11,274 antibodies corresponding to<br />
8,489 protein-coding genes. This entire effort<br />
depends heavily on the availability of good<br />
quality antibodies, and recently a communitybased<br />
portal, Antibodypedia (http://www.<br />
antibodypedia.org/), has been launched to<br />
allow antibodies from different providers to<br />
be listed and compared 4,5 , although the main<br />
source of information so far comes from<br />
the providers’ own validation data, not by<br />
independent third-party users. At present,<br />
the Antibodypedia contains close to 100,000<br />
antibodies, corresponding to >70% of the<br />
protein-coding genes in humans.<br />
An important objective has now been<br />
reached with the inclusion of 10,118 proteincoding<br />
genes corresponding to >50% of the<br />
19,559 human entries as defined by UniProt,<br />
including only entries with evidence at protein<br />
1248 volume 28 number 12 DECEMBER 2010 nature biotechnology
correspondence<br />
© 2010 Nature America, Inc. All rights reserved.<br />
or transcript level and proteins inferred<br />
from homology 6 (Fig. 1). The chromosomal<br />
coverage of protein-coding genes is shown<br />
in Figure 1a and the status for a selection<br />
of important protein classes is reported in<br />
Figure 1b. Almost 80% of the human kinases<br />
and Src-homology 2 domain–containing<br />
proteins and >50% of the transcription factors<br />
have protein profiling data in the atlas.<br />
We introduce the concept of annotated<br />
protein expression for paired antibodies, in<br />
which two or more independent antibodies<br />
are used to validate the staining pattern<br />
of each other. The immunohistochemical<br />
staining in each tissue or organ by the<br />
independent antibodies is compared and a<br />
new annotated protein expression is manually<br />
curated for each cell type in each tissue or<br />
organ. A reliability score is generated as an<br />
estimation of the degree of knowledge-based<br />
certainty of the reported expression profile.<br />
The reliability score is based on the similarity<br />
of the staining for the different antibodies, but<br />
also takes into account available information<br />
from literature, bioinformatics predictions<br />
and additional experimental evidence, such<br />
as western blots, transcript profiling and/<br />
or small interfering RNA knockdowns<br />
(Supplementary Notes; http://www.<br />
proteinatlas.org/about/quality+scoring).<br />
Approximately 2,000 protein-coding genes<br />
have annotated protein expression patterns<br />
in this launch and >75% of these have been<br />
scored with a high or medium reliability score<br />
(Fig. 1c). It is noteworthy that, at present,<br />
the majority of the proteins in the atlas have<br />
only been analyzed with a single antibody<br />
and thus these genes are reported simply as<br />
antibody staining. The long-term objective of<br />
the Human Protein Atlas project is to generate<br />
at least two antibodies for all human proteincoding<br />
genes to allow knowledge-based<br />
annotated protein expression data for the<br />
complete human proteome.<br />
The new portal contains a summary page<br />
for every human gene (Fig. 2) to allow for<br />
an easy entry point to navigate the detailed<br />
protein expression data. The summary page<br />
also serves as a convenient entry page from<br />
other related databases with a similar genecentric<br />
format. The results from the curation<br />
and annotation of the staining patterns from<br />
one or several antibodies to the same protein<br />
target are summarized with links to details<br />
pages, including the original images. A new<br />
feature is a section for subcellular localization<br />
and here, the annotated knowledge-base of<br />
the subcellular distribution reports the main<br />
and additional subcellular location for each<br />
protein target in the analyzed cell. Because<br />
the location of human proteins is known only<br />
a<br />
Number of genes<br />
2,000 No antibody<br />
Single antibody<br />
Multiple antibodies<br />
c<br />
1,500<br />
1,000<br />
500<br />
0<br />
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 X Y<br />
Medium<br />
36%<br />
Chromosome<br />
High<br />
42%<br />
Low<br />
17%<br />
Very low<br />
5%<br />
for a small fraction of the human genes, these<br />
subcellular localization data are a valuable<br />
resource for defining the protein content of<br />
the various compartments of the human cell<br />
and can thus constitute a starting point for<br />
further in-depth functional studies.<br />
A new organ view has been designed to<br />
allow cell types of similar origin to be easily<br />
compared. The tissues and organs have<br />
been divided into 12 functional classes,<br />
including central nervous system (CNS),<br />
hematopoietic, cardiovascular and female<br />
and male reproductive tissues. In this release,<br />
a total of 66 normal cell types from 46 tissues<br />
and organs have been scored. As an example,<br />
in Figure 2, we show part of the protein<br />
profiles for the human estrogen receptor 1<br />
(ESR1), including the results from antibody<br />
staining of three independent antibodies and<br />
b<br />
d<br />
Number of genes<br />
Fraction of cell types (%)<br />
1,400<br />
1,200<br />
1,000<br />
800<br />
600<br />
400<br />
200<br />
0<br />
100<br />
80<br />
60<br />
40<br />
20<br />
Kinases<br />
CD markers<br />
Transporters<br />
0<br />
0 500<br />
No antibody<br />
Single antibody<br />
Multiple antibodies<br />
Weak<br />
Moderate<br />
Strong<br />
the resulting annotated protein expression.<br />
The annotated protein expression in<br />
the various cell types are obtained by a<br />
knowledge-based process taking into<br />
account staining from at least two antibodies,<br />
literature and additional experimental data.<br />
For the ESR1 (Fig. 2), one of the antibodies<br />
shows weak staining in tissues such as those<br />
of the prostate and urinary bladder, most<br />
likely due to nonspecific staining, but the<br />
conflation of data from the three antibodies<br />
suggests exclusive expression in breast and<br />
female reproductive tissues. It is important<br />
to point out that this annotated protein<br />
expression should be considered as the best<br />
estimate of the true expression based on<br />
the experimental data available, but that<br />
additional data and input from the scientific<br />
community may lead to subsequent revisions<br />
Peptidases<br />
GPCRs (excl.<br />
olfactory receptors)<br />
1,000<br />
Genes<br />
Figure 1 Overview of the data in the Human Protein Atlas version 7.0. (a,b) The total coverage in terms<br />
of number of genes analyzed with single or multiple antibodies for each of the human chromosomes (a)<br />
and a selection of protein classes 3 (b). (c) Reliability scores for the ‘annotated protein expression’ are<br />
given by a four-graded scale, ranging from ‘very low’ to ‘high’. (d) The percent of cell types with the<br />
annotated expression levels weak, moderate and strong for each of the 2,000 genes with ‘annotated<br />
protein expression’. The genes are arranged according to the total abundance in the analyzed cell types,<br />
with genes detected only in a single cell type to the left and genes detected in all cell types to the far<br />
right. GPCRs, G protein–coupled receptors. SH2, Src-homology 2 domain.<br />
Transcription factors<br />
SH2–domain<br />
containing proteins<br />
SH3–domain<br />
containing proteins<br />
1,500<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1249
correspondence<br />
Normal tissues<br />
Cancer tissues<br />
Annotated protein expression<br />
this manner, the antibody-based portal can<br />
be complemented with other efforts, such<br />
as mass spectrometry–based proteomics<br />
and gene fusion technologies, to provide a<br />
knowledge-base for the human proteins,<br />
including isoforms, subcellular localization,<br />
tissue profiles and interaction networks.<br />
Because all the data on the Human Protein<br />
Atlas are publicly available, the information<br />
can be integrated into other databases. The<br />
availability of the annotated expression<br />
patterns opens up the possibility for a<br />
community-based dialog to provide input<br />
by researchers with specialized knowledge<br />
about particular protein targets.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Figure 2 The Human Protein Atlas web portal. The Human Protein Atlas summary view for the human<br />
estrogen receptor 1 (ESR1), displaying expression data for the target protein in normal and diseased<br />
tissues. To the right, a detailed organ view of the expression pattern of the ESR1 in normal human<br />
tissues with the cell types grouped into functional classes, here focused on the female and male tissues<br />
and the urinary tract. Three different antibodies have been used to generate the staining patterns, from<br />
which an annotated protein expression has been retrieved. The blue color scale represents annotated<br />
protein expression, and the red color scale represents relative antibody staining. The intensity of the<br />
colors indicates the level of expression and/or staining.<br />
of the interpretation of protein expression in<br />
individual cells, tissues or organs.<br />
A new cancer view has been designed,<br />
allowing an easy overview of the results for<br />
all antibodies towards a particular protein<br />
target in tumor tissues from 216 patients<br />
representing 20 different types of cancer<br />
(Fig. 2). This overview also contains the<br />
results for the corresponding normal tissues<br />
for each cancer form. An advanced search<br />
function has been created to allow complex<br />
queries involving both normal and cancer<br />
tissues, including subcellular localization,<br />
validation results for the antibody and/or<br />
target protein classes. An additional feature is<br />
the possibility to customize the display of the<br />
results page by adding or deleting columns<br />
based on interest. An important objective is<br />
to allow the integration of this database with<br />
other ’omics databases covering genomics 7 ,<br />
gene variation 8 , transcript profiling 9 ,<br />
proteomics 10 or metabolomics 11 . All the<br />
expression data and the antibody staining<br />
data are therefore available for downloads<br />
to facilitate comparative expression studies<br />
and systems biology approaches. In the near<br />
future, we also plan to include additional<br />
experimental data from mass spectrometry–<br />
based proteomics 12 and transcriptomics<br />
profiling using next generation RNA-Seq 13 .<br />
An issue with the new gene-centric<br />
design is the presence of protein isoforms,<br />
including proteolytic variants, posttranslational<br />
modifications and splice<br />
variants. Although the antigen sequence<br />
is reported for most antibodies on the<br />
detailed antibody validation page, and<br />
thus makes it possible to determine if a<br />
particular antibody should recognize all<br />
splice variants or not, the annotated protein<br />
expression is based on a consensus model<br />
in which a representative protein from the<br />
gene is reported. In the future, it might be<br />
relevant to move toward separate annotated<br />
protein expression profiles for each splice<br />
variant or other major alternative isoforms<br />
of a particular protein target.<br />
The annotated protein expression profiles<br />
allow a refinement of the global analysis<br />
of tissue specificity recently reported 14 ,<br />
in which the staining of antibodies<br />
corresponding to one-third of the proteincoding<br />
genes suggested that a large portion<br />
of the human proteins are present across<br />
tissues in a relatively ubiquitous manner.<br />
An analysis based on the 2,000 protein<br />
targets with annotated protein expression<br />
profiles can now be performed and the result<br />
suggests a similar profile with a large subset<br />
of proteins detected across various cell types<br />
(Fig. 1d).<br />
In conclusion, we here report on a new<br />
version of the Human Protein Atlas, where<br />
we have moved towards a knowledge-based<br />
portal with gene-centric expression profiles<br />
based on the annotation of several antibodies<br />
towards the same protein target. The Human<br />
Protein Atlas effort described here fits well<br />
into recent suggestions to initiate a genecentric<br />
Human Proteome Project to map and<br />
characterize a representative protein from<br />
every protein-coding gene of humans 15,16 . In<br />
Note: Supplementary information is available on the<br />
Nature Biotechnology website.<br />
Acknowledgments<br />
The authors would like to acknowledge the entire<br />
staff of the Human Protein Atlas project. This work<br />
was supported by grants from the Knut and Alice<br />
Wallenberg Foundation and EU 7 th framework<br />
program PROSPECTS.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
Mathias Uhlen 1,2 , Per Oksvold 1 ,<br />
Linn Fagerberg 1 , Emma Lundberg 2 ,<br />
Kalle Jonasson 1 , Mattias Forsberg 1 ,<br />
Martin Zwahlen 1 , Caroline Kampf 3 ,<br />
Kenneth Wester 3 , Sophia Hober 1 ,<br />
Henrik Wernerus 1 , Lisa Björling 1 &<br />
Fredrik Ponten 3<br />
1 School of Biotechnology, AlbaNova University<br />
Center, Royal Institute of Technology (KTH),<br />
Stockholm, Sweden. 2 Science for Life Laboratory,<br />
Royal Institute of Technology (KTH), Stockholm,<br />
Sweden. 3 Department of Genetics and Pathology,<br />
Rudbeck Laboratory, Uppsala University,<br />
Uppsala, Sweden.<br />
1. Uhlen, M. et al. Mol. Cell Proteomics 4, 1920–1932<br />
(2005).<br />
2. Barbe, L. et al. Mol. Cell Proteomics 7, 499–508<br />
(2008).<br />
3. Berglund, L. et al. Mol. Cell Proteomics 7, 2019–2027<br />
(2008).<br />
4. Bjorling, E. & Uhlen, M. Mol. Cell Proteomics 7, 2028–<br />
2037 (2008).<br />
5. Kiermer, V. Nat. Methods 5, 860 (2008).<br />
6. The UniProt Consortium. Nucleic Acids Res. 38,<br />
D142–148 (2010).<br />
7. Flicek, P. et al. Nucleic Acids Res. 38, D557–D562<br />
(2010).<br />
8. The International HapMap Project. Nature 426, 789–<br />
796 (2003).<br />
9. Lukk, M. et al. Nat. Biotechnol. 28, 322–324<br />
(2010).<br />
10. Mathivanan, S. et al. Nat. Biotechnol. 26, 164–167<br />
(2008).<br />
11. Wishart, D.S. et al. Nucleic Acids Res. 35, D521–526<br />
(2007).<br />
12. Ong, S.E. et al. Mol. Cell Proteomics 1, 376–386<br />
(2002).<br />
13. Sultan, M. et al. Science 321, 956–960 (2008).<br />
14. Ponten, F. et al. Mol. Syst. Biol. 5, 337 (2009).<br />
15. Anonymous. Nat. Methods 7, 661 (2010).<br />
16. Anonymous. Mol. Cell Proteomics 9, 427–429<br />
(2010).<br />
1250 volume 28 number 12 DECEMBER 2010 nature biotechnology
commentary<br />
Anchoring gene patent eligibility to its<br />
constitutional mooring<br />
Kenneth G Chahine<br />
© 2010 Nature America, Inc. All rights reserved.<br />
The antiquated legal standard that natural laws and products are not eligible for patent protection is ill-suited for gene<br />
and diagnostics patents. Here, I propose a new, technology-agnostic framework for determining patent eligibility that<br />
is tailored to the meet the US Constitutional objective of promoting innovation.<br />
The escalating debate over the validity of gene<br />
and diagnostic patents in the United States<br />
threatens to strip the incentives necessary to<br />
translate scientific findings from the bench to<br />
the bedside. Although a strong aversion to patents<br />
in new technological areas is nothing new,<br />
what sets the current debate apart is the extent<br />
of the misinformation, the personal nature of<br />
the subject matter, the support of powerful<br />
organizations and the potential damage an illinformed<br />
decision would have on the future<br />
of medicine and healthcare reform. The latest<br />
ruling in Association for Molecular Pathology<br />
v. US Patent & Trademark Office 1 (The Myriad<br />
Genetics Case) that invalidated patents covering<br />
the isolated breast cancer genes (BRCA1<br />
and BRCA2) and methods for their use in<br />
assessing the risk of breast and ovarian cancer<br />
provides an example of a decision based on<br />
faulty scientific reasoning, murky legal precedent<br />
and unsupported policy arguments.<br />
Any judge-made doctrine for assessing patent<br />
eligibility should be based on sound legal<br />
principles that are closely aligned with the constitutional<br />
purpose to “promote the Progress of<br />
Science and the Useful Arts” 2 . The challenge<br />
lies in finding the proper balance between<br />
encouraging innovation by awarding limited<br />
exclusivity to inventors and stymieing basic<br />
research through overprotection. Refocusing<br />
the legal reasoning for the natural products<br />
doctrine on this constitutional language offers<br />
a fresh perspective on the debate and opens the<br />
door to a new proposal for determining patent<br />
Kenneth G. Chahine is professor of law and<br />
member of the BioLaw Project, S.J. Quinney<br />
College of Law at University of Utah,<br />
Salt Lake City, Utah, USA.<br />
e-mail: chahinek@law.utah.edu<br />
eligibility, which is closely aligned to the goal<br />
of promoting innovation.<br />
In this article, I provide five criteria that<br />
would be useful for courts, legislatures and<br />
businesses when confronted with a question<br />
of subject matter eligibility. Taken together,<br />
these criteria incorporate many of the principles<br />
previously announced by the courts. They<br />
also allow courts to build a solid evidentiary<br />
foundation from which to decide the scope<br />
of an invention and its eligibility for patent<br />
protection. Importantly, the criteria may provide<br />
greater predictability to the legal, investment<br />
and research community by creating a<br />
legal framework for patent eligibility with one<br />
unequivocal and constitutionally consistent<br />
objective—promoting innovation.<br />
The natural products doctrine<br />
US courts have long held that “laws of nature,<br />
natural phenomena, and abstract ideas” are<br />
not eligible for patent protection 3 . Under<br />
this ‘natural laws’ or ‘natural products’ doctrine,<br />
the eligibility of two classes of patents<br />
are being questioned: first, patents covering<br />
isolated gene sequences; and second, methods<br />
of comparing genetic sequences, proteins and<br />
metabolites to diagnose disease or individualize<br />
treatment 4,5 .<br />
It has become increasingly clear, however,<br />
that advances in medicine and other information-based<br />
technologies are calling into<br />
question the applicability of the natural products<br />
doctrine. During the oral arguments in<br />
Bilski v. Kappos, US Supreme Court Justice<br />
Breyer voiced his own uncertainty on where<br />
to draw the line when he asked counsel—“All<br />
right, so what do I do?” 6 . Although the Bilski<br />
decision was hailed as a victory by the biotech<br />
community, it was based on the court’s<br />
reluctance to make a sweeping decision that<br />
might cause collateral damage to other industries<br />
7 . In rejecting the US Federal Circuit’s<br />
rigid test for determining patent eligibility,<br />
the Supreme Court once again affirmed the<br />
need for other criteria to establish patent eligibility<br />
that was consistent with the ‘purpose’<br />
of our patent system 8 .<br />
In examining the legal precedence for the<br />
natural products doctrine, I found there is<br />
little consistency in the manner in which the<br />
doctrine is applied. Because at a basic level all<br />
innovations comprise products of nature, even<br />
proponents of the doctrine have difficulty<br />
Judge Robert Sweet of the NY federal circuit<br />
court, who in March invalidated Myriad Genetics’<br />
BRCA1 and BRCA2 patents.<br />
©Joe Lawton<br />
nature biotechnology volume 28 number 12 december 2010 1251
COMMENTARY<br />
© 2010 Nature America, Inc. All rights reserved.<br />
defining when an invention crosses the line<br />
and becomes ineligible for patent protection.<br />
In distinguishing the patent ineligibility of<br />
isolated genes and the patent eligibility of<br />
purified adrenaline the Department of Justice<br />
argues that “patent eligibility may arise when<br />
a natural compound has been so refined and<br />
purified through human intervention as to<br />
become a substance different in kind from the<br />
natural product” 9 . The Department, however,<br />
offers no suggestion for determining when a<br />
product has been simply purified and remains<br />
patent ineligible (e.g., isolated genes) and<br />
when it has been purified to the extent that<br />
it now has changed ‘in kind’ and is patent eligible<br />
(e.g., adrenaline).<br />
Early decisions boldly declared without<br />
much support that natural products are “free<br />
to all men and reserved exclusively to none” 10 .<br />
More recently, the courts have relied on the<br />
justification that §101 of the US Patent Act<br />
excludes products of nature. As the Supreme<br />
Court has admitted, however, “[t]he plain language<br />
of §101 does not answer the question” 11<br />
and Chief Judge Rader of the Federal Circuit<br />
Court has called it a blunt tool for defining<br />
patent eligibility (http://www.patentlyo.com/<br />
patent/2010/08/gene-patents-on-appeal-aclusrecusal-motion.html).<br />
Indeed, §101 broadly<br />
embraces “any new and useful” invention or<br />
discovery, causing courts to strain the plain<br />
language of the statute to justify their exclusion<br />
of laws and products of nature 12 . Notably,<br />
both the US Constitution and the Patent Act<br />
explicitly embrace not only inventions, but also<br />
discoveries. The plain language of the statute<br />
and US Constitution, therefore, discredits the<br />
often-used argument that natural products are<br />
ineligible because they have been discovered<br />
rather than invented and cast doubt over the<br />
Department of Justice’s claim that patent law<br />
embraces only “human-made inventions” 13 .<br />
Thus, the natural products doctrine seems<br />
to be based on a legal house of cards. Decisions<br />
in the US courts are best rationalized based<br />
on how judges perceive the invention rather<br />
than any available legal theory. The discourse<br />
between Dan Ravicher representing Association<br />
for Molecular Pathology and Chief Judge Rader<br />
at a conference at the Fordham Law School in<br />
April succinctly captures nearly 150 years of<br />
legal precedence on the issue. Ravicher (pointing<br />
to a bottle of water), asked “was that [purification]<br />
sufficient intervention between what<br />
God gave us...and what man created to merit a<br />
patent?” Chief Judge Rader responded, “how<br />
many people have died of water pollution over<br />
the course of human events? Probably billions.”<br />
Ravicher is narrowly focusing on the structural<br />
differences between water and purified<br />
water without regard to its new utility (that<br />
is, potability); Judge Rader is less concerned<br />
with the magnitude of the structural differences<br />
and instead focuses on the impact of its<br />
new utility (that is, saving lives) 14 . Similarly,<br />
looking at the structural differences between a<br />
gene in one’s body and a patented isolated gene<br />
has led many to conclude that gene patents are<br />
not deserving of protection. This is the logic<br />
Judge Sweet employed in his reasoning in the<br />
Myriad Genetics (Salt Lake City, Utah, USA)<br />
case, which emphasized that isolated and complementary<br />
DNA is not “markedly different”<br />
from native DNA 15 .<br />
On the other hand, those focused on the<br />
novel medical uses of an isolated gene to diagnose<br />
disease come to a different conclusion.<br />
This is consistent with the Federal Circuit’s reasoning<br />
in Prometheus Laboratories, Inc. v. Mayo<br />
Collaborative Services in which the court stated<br />
that “methods of treatment … are always” eligible<br />
for patent protection 16 . Regardless of the<br />
ideological view held, it seems clear that we<br />
need a more objective tool for making decisions<br />
on the eligibility of inventions. That tool<br />
should be closely aligned with an objective we<br />
can all embrace—encouraging scientific innovation<br />
and bringing new technology to market<br />
that improves living conditions around the<br />
world now and for future generations.<br />
Do we even need a natural products<br />
doctrine?<br />
The distinction between patent eligibility and<br />
patentability is often blurred, yet it remains an<br />
important distinction. An invention is patentable<br />
if it is useful, novel and not obvious to<br />
one in the field provided that other sections<br />
of the Patent Act are met (e.g., it is adequately<br />
described and enabled). An invention that is<br />
declared ineligible cannot be patented, even<br />
if it would have met all of the patentability<br />
requirements of the Patent Act. As discussed in<br />
greater detail below, this distinction is critical<br />
to building a sound legal argument on which<br />
to rest a decision on the eligibility of gene and<br />
diagnostic patents.<br />
The natural products doctrine for patent<br />
ineligibility is a judicially created doctrine<br />
without any direct support in the Patent Act<br />
or US Constitution. The doctrine was created<br />
to prevent patents being given on fundamental<br />
discoveries that might stymie innovation.<br />
It is not entirely clear, however, whether this<br />
doctrine is a necessary legal tool to reign in<br />
patents whose breadth threatens innovation.<br />
The examples used by the courts to justify the<br />
natural products doctrine are not convincing,<br />
especially when one considers the less controversial<br />
patentability requirements of the<br />
Patent Act that can adequately address similar<br />
concerns. Take, for example, the Supreme<br />
Court’s often quoted statement that “Einstein<br />
could not patent his celebrated law that E = mc 2 ;<br />
nor could Newton have patented the law of<br />
gravity” 17 . Rather than arguing that gravity is<br />
not eligible for patent protection because it is<br />
a “law of nature,” one can persuasively argue<br />
instead that gravity was “known or used by<br />
others…before the invention thereof by the<br />
applicant for patent” and, therefore, is not novel<br />
under section 102 of the Patent Act 18 . Put differently,<br />
Newton could not have patented gravity<br />
because it would have restricted public use<br />
of something the public was enjoying before<br />
his discovery—an explanation for how something<br />
already in the public domain works has<br />
never been per se patentable (for a review of<br />
law of anticipation, see ref. 19). Discovering the<br />
mechanism of action of aspirin is not patentable<br />
if the public has, albeit unknowingly, been<br />
enjoying those benefits.<br />
Similarly, Samuel Morris’s attempt to broadly<br />
patent all forms of communication that used<br />
“electomagnetism” has been used to support<br />
the idea that laws of nature are not patent eligible.<br />
The court, however, also engaged in a<br />
lengthy discussion that the claim was invalid<br />
because it did not teach all methods of communicating<br />
by means of electromagnetism, only<br />
now by the telegraph 20 . The court’s reasoning<br />
is now codified as § 112 of the Patent Act and<br />
forms a sound basis for invalidity in instances<br />
when the patentee attempts to capture more<br />
intellectual estate than taught in the patent.<br />
Even in the Myriad case, Christopher<br />
Holman and Robert Cook-Deegan have<br />
pointed to vulnerabilities in the patent claims<br />
and suggest that “other doctrines of patentability<br />
can limit gene-based patent claims to an<br />
appropriate scope, rather than using the patent<br />
eligibility doctrine that would unnecessarily<br />
invalidate all DNA-based patents indiscriminately<br />
and regardless of their merit” 21 .<br />
By using other legal grounds, such as novelty<br />
and enablement, the court would obviate<br />
the need to clearly and predictably define what<br />
constitutes laws of nature, natural phenomena,<br />
and abstract ideas—a task it has struggled with<br />
unsuccessfully for over a century. It would also<br />
allow courts to make a decision on a case-bycase<br />
basis rather than invalidate an entire class<br />
of patents and risk hobbling the diagnostic<br />
industry by inadvertently eliminating the<br />
incentive to invent and invest.<br />
A new proposal for assessing patent<br />
eligibility<br />
Judicial reliance on §101 of the Patent Act to<br />
justify patent ineligibility is misplaced and<br />
serves only to confound the issue. Section 8 of<br />
Article 1 in the Constitution gives Congress<br />
plenary power to define what is and is not<br />
1252 volume 28 number 12 december 2010 nature biotechnology
COMMENTARY<br />
© 2010 Nature America, Inc. All rights reserved.<br />
patentable subject matter under one fundamental<br />
condition—it must “promote the<br />
Progress of Science and the Useful Arts” 22 .<br />
Therefore, a more sound legal justification is<br />
that an invention is not eligible for patent protection<br />
if it attempts to patent subject matter<br />
so fundamental that granting even a limited<br />
monopoly would impede scientific progress 23 .<br />
If so, courts could then logically hold that the<br />
patenting of fundamental products of nature<br />
are unconstitutional for failing to “promote the<br />
Progress of Science and useful Arts” even if the<br />
Patent Act does not specifically proscribe this<br />
class of invention or discovery 2 .<br />
Focusing on the constitutional underpinning<br />
for patent eligibility is not merely an exercise in<br />
legal gymnastics. Pushing past the arguments<br />
stemming from the legal precedent focused on<br />
§101 toward agreement on why certain inventions<br />
should not be eligible for patent protection<br />
may pave the way for a balanced and reasoned<br />
approach to determine when an invention or<br />
discovery is or is not promoting innovation.<br />
The current doctrine could then be viewed<br />
as one of several factors in considering patent<br />
eligibility. Just as a physician would never rely<br />
on one abnormal cholesterol score to determine<br />
whether to conduct open-heart surgery,<br />
courts should not rely solely on this doctrine to<br />
invalidate countless biotech patents. The court<br />
should incorporate into the analysis additional<br />
factors that can better define the breadth of an<br />
invention and its threat to our constitutional<br />
mandate to promote innovation.<br />
Using a balanced, multipronged approach<br />
to defining the scope of intellectual property<br />
rights is not new. In copyright law, a<br />
four-pronged approach is used to determine<br />
whether a reproduction is made for the purposes<br />
of “criticism, comment, news reporting,<br />
teaching … scholarship, or research,” and thus,<br />
exempt from infringement under the fair use<br />
doctrine. The statute specifically states: “In<br />
determining whether the use made of a work…<br />
is a fair use the factors to be considered shall<br />
include—(1) the purpose and character of the<br />
use, including whether such use is of a commercial<br />
nature or is for nonprofit educational purposes;<br />
(2) the nature of the copyrighted work;<br />
(3) the amount and substantiality of the portion<br />
used in relation to the copyrighted work as<br />
a whole; and (4) the effect of the use upon the<br />
potential market for or value of the copyrighted<br />
work.” What is appealing about these factors<br />
is how it balances the copyright holders’ economic<br />
interests with First Amendment rights<br />
to free speech and basic research. Over time,<br />
the court has applied these factors and brought<br />
clarity and predictability to the field.<br />
I propose that the US Congress and courts<br />
should take a similar approach to determining<br />
patent eligibility using criteria that refocus the<br />
determination on factual considerations that<br />
bear directly on the objective of promoting<br />
innovation. In developing this framework, I<br />
have reviewed three areas: first, Supreme Court<br />
precedence, which clarified the factors and<br />
rationales that guided the Court when making<br />
patent eligibility determinations; second,<br />
factual circumstances, case studies and market<br />
forces relevant to determining the risks of<br />
overprotection, including the specific contribution<br />
of the patent to the overall concerns;<br />
and finally, the roles of existing laws that may<br />
limit the abuse by patent holders. This review<br />
allowed the formulation of five criteria that I<br />
believe would be useful for courts, legislatures<br />
and business when confronted with a question<br />
of subject matter patentability.<br />
1. Does the patented invention cover a<br />
product or law of nature unaltered by<br />
humans?<br />
2. Do other limitations in the claims and<br />
prosecution history confer meaningful<br />
limitations on the scope of the patented<br />
claims?<br />
3. Are there reasonable approaches to<br />
design around or circumvent the patented<br />
invention?<br />
4. Is there evidence that basic research is<br />
being hindered, and if so, is the patent<br />
solely or substantially responsible?<br />
5. And are there other laws in place that<br />
would mitigate the risk of overprotection?<br />
Taken together, these criteria incorporate<br />
many of the principles previously adhered to<br />
by the courts. They also allow courts to build<br />
a solid evidentiary foundation from which to<br />
decide the scope of an invention and its eligibility<br />
for patent protection. Importantly, the<br />
criteria may provide greater predictability by<br />
creating a legal framework for patent eligibility<br />
with one unequivocal and constitutionally consistent<br />
objective—promoting innovation. I discuss<br />
each of the criteria in more detail below.<br />
Does the patented invention cover a product<br />
or law of nature unaltered by humans? This<br />
first criterion largely preserves the US courts’<br />
previously announced laws and products of<br />
nature doctrine. The US Supreme Court has<br />
stated that “anything under the sun made by<br />
man was patentable” 23 . Therefore, as the US<br />
Department of Justice and others agree, all<br />
human-made inventions are per se patent eligible.<br />
In cases where there is a question whether<br />
the invention is human-made as with isolated<br />
DNA, or the discovery employs a correlation<br />
or algorithm as with the diagnostic claims,<br />
additional factors should be considered when<br />
determining patent eligibility of the invention<br />
as a whole.<br />
For more information on the natural products<br />
doctrine, readers are referred to relevant<br />
reviews in the literature 17 .<br />
Do other limitations in the claims and prosecution<br />
history confer meaningful limitations<br />
on the scope of the patented claims?<br />
This second criterion considers the impact of<br />
claim limitations on the scope of the exclusivity<br />
sought and the risk of overprotection. After<br />
all, an invention is nothing more than the use<br />
of natural laws and products in an application<br />
not found in nature. A mixture of naturally<br />
occuring elements that make a new alloy with<br />
unique properties is patent eligible. Likewise,<br />
inventions do not become ineligible for patent<br />
protection because they employ a natural law,<br />
even if the law in central to its utility. A novel<br />
airplane wing design that improves fuel efficiency<br />
is patent eligible, even if central to its<br />
operation is the Bernoulli principle that creates<br />
lift and permits flight. In both instances, others<br />
are free to use the individual elements and<br />
the Bernoulli principle in applications outside<br />
those claimed. What is not eligible, therefore,<br />
is the individual elements and the Bernoulli<br />
principle per se because exclusivity to those<br />
products and laws without limits would grant<br />
an unreasonable broad monopoly.<br />
Indeed, the Supreme Court has recognized<br />
that “an application of a law of nature…may<br />
well be deserving of patent protection” 24 ,<br />
and at times has even used that reasoning to<br />
uphold patent eligibility. In Diamond v. Diehr,<br />
the Supreme Court held that although the<br />
Arrhenius equation alone was not patent eligible,<br />
the scope of a claim for curing rubber<br />
that used the equation was eligible because it<br />
didn’t foreclose the use of the equation in other<br />
processes 25 .<br />
Likewise, in gene and diagnostic patents<br />
the court should look at limitations, including<br />
the following: the gene covered, the mutations<br />
recited, the diseases it claims to diagnose and<br />
the tissue source. The Myriad Genetics patents<br />
do not seek to exclude all genetic correlations,<br />
of all genes, to diagnose all diseases—they<br />
claim two genes for diagnosing breast and ovarian<br />
cancer. As such, these limitations should<br />
be considered in a patentability determination,<br />
especially if the justification for the doctrine is<br />
to reign in monopolies that unreasonably preempt<br />
future progress.<br />
Are there reasonable approaches to design<br />
around or circumvent the patented invention?<br />
This third criterion considers the ability<br />
to design around the invention. As Justice<br />
nature biotechnology volume 28 number 12 december 2010 1253
COMMENTARY<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Breyer notes in his decent to the revocation of a<br />
grant of certiori in Laboratory Corp. of America<br />
Holdings v. Metabolite Labs., Inc., “[p]atent law<br />
[also]…seeks to avoid the diminished incentive<br />
to invent that underprotection can threaten” 26 .<br />
If patents do not place meaningful barriers to<br />
ward off imitators the entire constitutional<br />
purpose of promoting the progress of science<br />
would be subverted.<br />
It is important, therefore, for the court to<br />
consider alternatives that exist, or could be<br />
developed, which would reasonably circumvent<br />
the patent, avoid underprotection and<br />
promote the progress of science. This factor<br />
is particularly useful when applied to other<br />
inventions the courts have previously held<br />
ineligible. It is not possible to design around a<br />
truly fundamental product or principle such as<br />
a new element, Einstein’s law of relativity and<br />
Newton’s law of universal gravitation 27 .<br />
History instructs us that this inquiry should<br />
not be narrowly limited to the precise technology<br />
claimed. In the music industry, for example,<br />
the biggest threats have not come from improving<br />
the vinyl records but from the evolution<br />
of disruptive technologies, such as the cassette<br />
player, CDs, DVDs and digital recordings that<br />
were developed over a 20-year span.<br />
In gene and diagnostic patents the inquiry<br />
is too narrowly focused on circumventing<br />
the precise invention. In a genetic test where,<br />
for example, mutations or polymorphisms<br />
are a proxy for deficient protein activity, it is<br />
entirely possible that another assay could be<br />
developed—a protein, metabolite or functional<br />
assay—that would better predict a predisposition<br />
to the disease. In the Myriad case,<br />
a functional assay for BRCA1 and BRCA2 proteins<br />
would obviate the need to examine the<br />
gene sequences and circumvent the patents.<br />
Evidence of a potential design-around, therefore,<br />
would weigh heavily against finding the<br />
claim overly broad and unconstitutional.<br />
Is there evidence that basic research is being<br />
hindered, and if so, is the patent solely or<br />
substantially responsible? This fourth criterion<br />
considers whether the patent in question<br />
is solely or substantially responsible for the<br />
alleged industry concerns. As a recent Nature<br />
Biotechnology article by Carbone et al. 28 suggests,<br />
“the potential obstacles to innovation<br />
that patents cause in diagnostics may not be as<br />
high…as some had feared.” The authors’ policy<br />
changes in licensing practices would solve many<br />
of the problems that patent ineligibility is meant<br />
to address. This approach is supported by a case<br />
study by Fore et al. 29 on the polymerase chain<br />
reaction (PCR) technology. At the time PCR<br />
was patented, a strong argument could have<br />
been made that it was a product of nature and<br />
patenting its use would foreclose basic research<br />
and stymie innovation—after all, the enzyme<br />
in a test tube merely does what it naturally<br />
evolved to do. However, the authors found<br />
that despite “heavy patent protection” PCR<br />
was “disseminate[d] broadly in the molecular<br />
biology world, becoming an indispensable<br />
research tool employed in nearly every biological<br />
field” 29 . This was due to deliberate licensing<br />
and business practices on the part of the patent<br />
holders, including ‘rational forbearance’ where<br />
companies refrain from suing researchers for<br />
patent infringement 30 .<br />
The Carbone et al. 28 thesis, that licensing<br />
practices will resolve most of the issues, should<br />
be tempered for several reasons. First, licensing<br />
practices may increase the number of labs that<br />
offer the diagnostic test but it does not guarantee<br />
lower pricing, reimbursement and broader<br />
accessibility. As another paper 31 by several of<br />
the authors in Carbone et al. reported, the per<br />
unit price of the exclusively licensed BRCA1/<br />
BRCA2 diagnostic is lower than a non-exclusively<br />
licensed colorectal cancer test. There are,<br />
therefore, other market forces outside patents<br />
that conspire to sustain higher prices and limit<br />
accessibility.<br />
Second, an issue not often discussed is the<br />
impact of regulatory hurdles. Carbone et al. 28<br />
suggest that unlike the diagnostic industry,<br />
exclusive licenses to therapeutic proteins,<br />
such as erythropoietin, growth hormone and<br />
interferon, are “commercially significant” and<br />
justified. Although true, the reason exclusive<br />
licensing of therapeutics is commercially significant<br />
(and why erythropoietin at the time<br />
was likely not challenged as a natural product)<br />
is the requirement for US Food and Drug<br />
Administration (FDA; Silver Spring, MD)<br />
approval. The lack of FDA oversight over many<br />
diagnostics not only lowers the barriers to entry<br />
but also substantially reduces the development<br />
costs. If the FDA began regulating all diagnostics<br />
as it currently does for therapeutics, this<br />
debate would likely become mute as most of<br />
the laboratories and associations opposing<br />
gene and diagnostic patents would be unable<br />
(or unwilling) to develop FDA-approved diagnostic<br />
tests, even in the absence of patents and<br />
exclusive license arrangements. Most hospital<br />
and university-based labs draw a bright line<br />
between developing tests that can be offered<br />
by the Commercial Laboratory Improvement<br />
Act (CLIA) and FDA 510k or premarketing<br />
approval (PMA). Given the FDA’s increased<br />
scrutiny over these so-called laboratory developed<br />
tests, we should not ignore that a change<br />
in the regulatory winds may fundamentally<br />
alter market forces and take the wind out of<br />
the sails of the most vocal opponents to gene<br />
and diagnostic patents.<br />
And finally, most analysis starts with the<br />
assumption that there is a robust reimbursement<br />
policy in place for diagnostics. This gives the<br />
false impression of lower barriers to entry, especially<br />
for academic and hospital labs. A role of<br />
the exclusive licensee in a disruptive technology<br />
is to create a market and pioneer reimbursement.<br />
Establishing a commercially viable<br />
market for the product is a costly undertaking<br />
that all subsequent providers benefit from at a<br />
fraction of the investment. It is worth considering<br />
what the availability of the BRCA1/BRCA2<br />
diagnostic would have been in the absence of<br />
Myriad Genetics having pioneered the market<br />
and established a reimbursement policy.<br />
To ensure that patents are not unjustly<br />
blamed for the ills of the diagnostic industry,<br />
this criterion considers other factors that may<br />
impact dissemination and access. Overreacting<br />
to the public outcry concerning gene patents<br />
without considering additional factors may<br />
fundamentally cause a change in patent law<br />
without resolving the present concerns of the<br />
industry. Moreover, even assuming patents currently<br />
do not play a significant role in the diagnostic<br />
industry, market and regulatory forces<br />
may shift in the future necessitating once again<br />
strong incentives to invent and invest to bring<br />
diagnostics to market.<br />
Are there other laws in place that would mitigate<br />
the risk of over-protection? This final<br />
criterion takes into consideration other mechanisms<br />
that substantially mitigate the risk of<br />
over-protection and abuse by patent holders. For<br />
example, under the Bahy-Dole Act the government<br />
retains a “nontransferrable, irrevocable,<br />
paid-up license…to practice the invention…<br />
throughout the world” 32 . The Act further<br />
specifies that the government may, under certain<br />
conditions, exercise “March-in-Rights” to<br />
make the invention available to the public. In<br />
federally funded research, therefore, the courts<br />
should consider the government’s rights to reign<br />
in any potential abuse by a patent holder.<br />
Other laws could be revisited or created to<br />
address other concerns. There is, for example, a<br />
reasonable public policy argument that an individual<br />
should have a right to obtain a second<br />
opinion before undergoing a substantial medical<br />
procedure. If so, the courts could examine<br />
an extension of the doctrine of patent exhaustion.<br />
Under the doctrine (also referred to as<br />
the first sale doctrine), the first unrestricted<br />
sale of a patented item or method exhausts<br />
the patentee’s control over that particular item<br />
or method. This compromise would uphold<br />
the patentee’s rights and economic incentives<br />
by retaining exclusivity over the first sale, yet<br />
allow second opinions to be rendered without<br />
the threat of infringement. Similarly, concerns<br />
1254 volume 28 number 12 december 2010 nature biotechnology
COMMENTARY<br />
© 2010 Nature America, Inc. All rights reserved.<br />
that a physician’s fear of infringement may chill<br />
the practice of medicine could be addressed<br />
by US Congress as it has by excluding medical<br />
practitioners from liability for infringing a<br />
patent by performing of a medical or surgical<br />
procedure on a body 33 .<br />
Conclusions<br />
In the ongoing debate, both sides are talking<br />
over each other in part because there is a lack<br />
of well-articulated criteria for what constitutes<br />
a law of nature, a natural phenomena and an<br />
abstract idea, as well as the nexus between the<br />
doctrine and its putative objective of promoting<br />
innovation. Focusing the analysis on the<br />
constitutional objective that inventions and discoveries<br />
must promote the progress of science<br />
provides the clearest framework for determining<br />
what inventions should be eligible for patent<br />
protection. The proposal made for using a<br />
balanced, multipronged approach that emphasizes<br />
a factual determination directly bearing<br />
on this constitutional mandate is intended as<br />
the first step on the road to a workable compromise<br />
and a predictable legal framework.<br />
By mirroring the approach to determining<br />
copyright fair use, this proposal allows courts<br />
to thoughtfully assess patent eligibility on a<br />
case-by-case basis and build, over time, a predictable<br />
legal framework for different classes of<br />
inventions and discoveries.<br />
Without clear evidence that DNA patents<br />
are the sole or significant factor in disrupting<br />
basic research and dissemination, US courts<br />
should be loath to invalidate an entire class<br />
of patents. One misstep may have unforeseeable<br />
negative consequences on the future of<br />
personalized medicine and cripple the biotech<br />
industry. Moreover, a flexible approach<br />
to patent eligibility is more likely to withstand<br />
Supreme Court scrutiny, which has repeatedly<br />
emphasized its dislike for rigid tests previously<br />
proposed by the Federal Circuit. Although<br />
there will no doubt be a healthy debate as to<br />
where to draw the line between overprotection<br />
and underprotection, both the scientific<br />
and commercial communities should recognize<br />
that neither can advance if a single<br />
party owns exclusive and absolute rights to<br />
an invention involving a truly fundamental<br />
law or product of nature.<br />
ACKNOWLEDGMENTS<br />
The author is grateful to J. Mixco for his contribution<br />
and assistance. The views expressed are solely those of<br />
the author.<br />
COMPETING FINANCIAL INTERESTS<br />
The author declares no competing financial interests.<br />
1. No. 09 Civ. 4515 (RWS), 2010 U.S. Dist. LEXIS 35418,<br />
at *2 (S.D.N.Y. Mar. 29, 2010, revised Apr. 2, 2010).<br />
2. U.S. Const. art. I, § 8, cl. 8.<br />
3. Diamond v. Diehr, 450 U.S. 175 (1981).<br />
4. See Association for Molecular Pathology, supra note 2.<br />
5. See Prometheus Laboratories, Inc. v. Mayo Collaborative<br />
Services, 581 F.3d 1336 (Fed. Cir. 2009).<br />
6. Transcript of Oral Argument at 20, Bilski v Kappos, 130<br />
S. Ct. 3218 (2010) (No. 08–964).<br />
7. Bilski v. Kappos, 130 S. Ct. 3218 (2010).<br />
8. Ibid.<br />
9. Brief For The United States as Amicus Curiae in<br />
AMP v. USPTO (No. 2010–1406) at 20. (emphasis<br />
in original)<br />
10. Funk Brothers Seed Co. v. Kalo Inoculant Co., 333 U.S.<br />
127 (1948).<br />
11. Parker v. Flook, 437 U.S. 584 (1978).<br />
12. “Whoever invents or discovers any new and useful process,<br />
machine, manufacture, or composition of matter,<br />
or any new and useful improvement thereof, may<br />
obtain a patent therefore, subject to the conditions and<br />
requirements of this title.” 35 U.S.C. §101.<br />
13. Supra note 15 at 12.<br />
14. Motion by Plaintiffs-Appellees for Recusal of Chief<br />
Judge Randall R. Rader, AMP v. USPTO, No. 09-CV-<br />
4515 (Fed. Cir. June 29, 2010).<br />
15. See Association for Molecular Pathology, supra note<br />
2 at 114.<br />
16. Prometheus Labs, Inc v. Mayo Collaborative Services,<br />
581 F.3d 1336, 1346 (Fed. Cir. 2009)<br />
17. Diamond v. Chakrabarty, 447 U.S. 303 (1980).<br />
18. 35 U.S.C. §102(a) (2010).<br />
19. Dan, L. Burk & Mark A. Lemley, Inherency, 47. William<br />
Mary Law Rev., 371 (2005).<br />
20. O’Reilly v. Morse, 56 U.S. 62, 117–18 (1853).<br />
21. Brief of Amici Curiae Christopher M. Holman and<br />
Robert Cook-Deegan in Support of Neither Party in<br />
AMP v. USPTO (No. 2010-1406) at 24.<br />
22. “...the reason for exclusion [of laws and products of<br />
nature] is that sometimes too much patent protection<br />
can impede rather than ‘promote the Progress of<br />
Science and useful Arts,’ the constitutional objective<br />
of patent... protection.” Laboratory Corp. of America<br />
Holdings v. Metabolite Labs., Inc., cert. dismissed as<br />
improvidently granted, 548 U.S. 124, 126–27 (2006)<br />
(Breyer, J., dissenting).<br />
23. Chakrabarty, supra note 17.<br />
24. Diamond v. Diehr, supra note 3.<br />
25. Ibid.<br />
26. Laboratory Corp. supra note 22.<br />
27. Chakrabarty, supra note 17 at 309.<br />
28. Carbone, J. et al. Nat. Biotechnol. 28, 784–791<br />
(2010).<br />
29. Fore, J. Jr. et al. J. Biomed. Discov. Collab. 1, 7<br />
(2006).<br />
30. Ibid.<br />
31. Cook-Deegan, R. et al. Genet. Med. 12 Suppl, S15–S38<br />
(2010).<br />
32. 35 U.S.C. §§200–12 (2010).<br />
33. 35 U.S.C. §287(c) (2010).<br />
nature biotechnology volume 28 number 12 december 2010 1255
commentary<br />
The environmental impact subterfuge<br />
Gregory Conko & Henry I Miller<br />
Action needs to be taken to prevent anti-biotech activists from co-opting environmental law to derail the planting of<br />
transgenic crops that have already received regulatory approval.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
The latest weapon used in the misinformation<br />
war against recombinant DNA technology<br />
and its agricultural applications is an obscure<br />
environmental law from the seventies. Green<br />
activists and organic farmers are exploiting<br />
the National Environmental Policy Act of 1970<br />
(NEPA) to convince courts that inconsequential<br />
paperwork oversights by regulators at the<br />
US Department of Agriculture warrant the revocation<br />
of two final approvals for recombinant<br />
DNA–modified crop varieties and of the issuance<br />
of permits to test several others. At least<br />
one more case is pending.<br />
Under NEPA, all US federal government<br />
agencies are required to consider the effects<br />
that any “major actions” they take may have on<br />
the “human environment.” Agencies can exempt<br />
whole categories of routine or repetitive activities<br />
but most other decisions—such as the issuance<br />
of a new regulation, the location of a new bridge<br />
or the approval of a new agricultural technology—trigger<br />
the NEPA obligation to evaluate<br />
environmental impacts. If the agency concludes<br />
that the action will have “no significant impact”<br />
(a legal term of art), it issues a relatively brief<br />
Environmental Assessment explaining the basis<br />
for that decision. If significant effects are likely,<br />
though, the agency must prepare a comprehensive<br />
Environmental Impact Statement (EIS),<br />
which typically requires thousands of hours of<br />
work, details every imaginable effect and runs to<br />
hundreds (or even thousands) of pages.<br />
The obligation under NEPA is wholly procedural,<br />
which means that even significant environmental<br />
effects do not prohibit the agency<br />
Gregory Conko is a senior fellow at the<br />
Competitive Enterprise Institute in Washington,<br />
DC, and Henry I. Miller, a physician and fellow<br />
at Stanford University’s Hoover Institution,<br />
was the founding director of the FDA’s Office of<br />
Biotechnology.<br />
e-mail: henry.miller@stanford.edu<br />
In August, a federal judge revoked the USDA’s approval of Roundup Ready sugar beets, which represent<br />
95% of the crop now grown in the United States.<br />
from ultimately taking the proposed action.<br />
Its purpose is solely to force government agencies<br />
to in fact consider possible environmental<br />
effects. However, courts have interpreted the law<br />
broadly by requiring a comprehensive review of<br />
every imaginable effect on the “human environment.”<br />
This category now encompasses not only<br />
harm to the natural ecology but also economic,<br />
social and even aesthetic impacts.<br />
Thus, if agencies miss some tangential or<br />
speculative issue, they can be tripped up by an<br />
irresponsible litigant who alleges that the environmental<br />
review was incomplete. Even when<br />
regulators actually do consider a potential<br />
impact but reject the concern owing to its unimportance<br />
or improbability, they can run afoul of<br />
NEPA by failing to extensively and comprehensively<br />
document their reasoning. This latter phenomenon<br />
has lately plagued USDA approvals of<br />
recombinant DNA–modified crops.<br />
NEPA and biotech<br />
The USDA has approved 74 different recombinant<br />
DNA–modified crop varieties, or<br />
transformation events, for commercial-scale<br />
cultivation, including varieties of corn, canola,<br />
squash, soybean, potato, tomato and other<br />
species. In each case, the department’s Animal<br />
and Plant Health Inspection Service (APHIS)<br />
reviewed copious amounts of data from several<br />
years’ worth of controlled field experiments<br />
to evaluate the variety’s agronomic and<br />
environmental effects. And because each one<br />
of these plants is highly similar to conventionally<br />
bred varieties already grown throughout<br />
the United States—differing only in the addition<br />
of a small number of well characterized<br />
genes that introduce useful traits—APHIS<br />
concluded that they would have no significant<br />
environmental impact.<br />
But some anti-biotech activists are unwilling<br />
to permit facts or the judgments of experts to<br />
get in the way of their antagonism, so a group<br />
of environmental organizations led by the antibiotech<br />
Center for Food Safety joined with a<br />
handful of organic farmers to sue in federal<br />
court to halt the planting of alfalfa, sugar beets<br />
and turf grass modified for resistance to the<br />
David R. Frazier/DanitaDelimont.com/Newscom<br />
1256 volume 28 number 12 december 2010 nature biotechnology
COMMENTARY<br />
© 2010 Nature America, Inc. All rights reserved.<br />
herbicide glyphosate (better known by its<br />
trade name Roundup (Monsanto Company;<br />
St. Louis, MO)). The plaintiffs made no substantive<br />
arguments that these crops were<br />
unsafe, but instead claimed that APHIS’s<br />
Environmental Assessment was procedurally<br />
insufficient and that a comprehensive EIS<br />
should have been prepared.<br />
What were the supposedly significant environmental<br />
impacts that APHIS allegedly failed<br />
to account for in its environmental analysis?<br />
The plaintiffs made two claims: first, that crosspollination<br />
by recombinant DNA–modified<br />
crops would ‘contaminate’ non-genetically engineered<br />
alfalfa and produce negative economic<br />
and socioeconomic effects, such as the loss of<br />
certification for organic crops and an inability<br />
to produce non-genetically engineered seed;<br />
and second, that the widespread adoption of<br />
Roundup Ready crops and increased use of<br />
glyphosate could accelerate the development of<br />
glyphosate-resistant weeds and force farmers to<br />
switch to other, less convenient herbicides. Both<br />
claims are dubious.<br />
Cross-pollination by recombinant DNA–<br />
modified plants does not, in fact, jeopardize the<br />
organic certification of those crops because the<br />
USDA’s rules for organic production are based on<br />
process, not outcome. As long as organic growers<br />
adhere to permissible practices, take reasonable<br />
steps to prevent unintentional contact with transgenes<br />
and do not intentionally plant recombinant<br />
DNA–modified seeds, unintentional crosspollination<br />
does not cause those crops to lose<br />
their organic status. Furthermore, organic rules<br />
require growers to create distinct buffer zones<br />
to prevent unintended contact with prohibited<br />
substances, such as pollen from recombinant<br />
DNA–modified plants and synthetic pesticides.<br />
Seed breeders are also accustomed to using isolation<br />
distances as well as physical and biological<br />
buffer zones to maintain genetic purity, and the<br />
net effect of approving these recombinant plants<br />
would be, at worst, minimal.<br />
Similarly, the development of herbicideresistant<br />
weeds is not unique to recombinant<br />
DNA–modified varieties, but occurs commonly<br />
in all crops exposed to herbicides. APHIS is well<br />
aware of the phenomenon and considers it routinely<br />
when making approval decisions. The<br />
agency made clear to the court that good stewardship<br />
is the only reasonable defense against<br />
the development of herbicide-resistant weeds,<br />
but the court criticized APHIS for not evaluating<br />
the cumulative extra impact of approving new<br />
Roundup Ready varieties.<br />
From a NEPA perspective, what makes recombinant<br />
DNA–modified herbicide-resistant plants<br />
fair game is that they are regulated by a federal<br />
government agency and that approval to commercialize<br />
them constitutes, in NEPA-speak, a<br />
“major action.” Even though plant scientists are<br />
virtually unanimous that recombinant DNA is<br />
an extension, or refinement, of earlier techniques<br />
and that it is safer and more predictable than<br />
many other conventional breeding methods,<br />
such as wide-cross hybridization and mutation<br />
breeding, only recombinant DNA–modified<br />
varieties are subject to the kind of government<br />
approval that triggers the EIS obligation.<br />
It is ironic that recombinant DNA–modified<br />
crop varieties, whose development and environmental<br />
impacts are carefully scrutinized by<br />
the USDA, are subject to the extra burden of a<br />
compelled EIS, whereas varieties developed with<br />
cruder conventional methods are subject to neither<br />
premarket assessment nor the NEPA rules.<br />
Indeed, the EIS requirement for recombinant<br />
DNA–modified crop approvals could have been<br />
avoided if the USDA and other federal regulatory<br />
agencies had heeded the advice of the scientific<br />
community some 25 years ago and chosen not to<br />
subject the products of recombinant DNA technology<br />
to special, discriminatory government<br />
regulation. Had the regulators instead followed<br />
the scientific consensus, they would not have<br />
been tripped up by NEPA’s paperwork requirement<br />
because, in the absence of an approval<br />
process, there would be no “major actions” to<br />
trigger the EIS obligation.<br />
Environmental creep<br />
Since its enactment, the scope of NEPA’s coverage<br />
has experienced continual creep. Through<br />
a combination of prodding by environmental<br />
activists and judicial overreach, the meaning<br />
of the term “human environment” has been<br />
expanded to include not just tangible ecological<br />
harms that affect people and human communities<br />
but also impacts that are economic, social,<br />
cultural, historic and aesthetic. A decision by a<br />
federal district court in Minnesota illustrates<br />
how liberally the statute has been interpreted:<br />
“Relevant as well is whether the project<br />
will affect the local crime rate, present fire<br />
dangers, or otherwise unduly tap police<br />
and fire forces in the community…the project’s<br />
impact on social services, such as the<br />
availability of schools, hospitals, businesses,<br />
commuter facilities, and parking…harmonization<br />
with proximate land uses, and a blending<br />
with the aesthetics of the area…[and a]<br />
consideration of the project’s impact on the<br />
community’s development policy…”<br />
In other words, almost any possible effect<br />
that a clever plaintiff can imagine may constitute<br />
a significant impact under NEPA if a<br />
sympathetic judge agrees.<br />
In the first ever NEPA challenge to a product<br />
made with recombinant DNA technology,<br />
environmental activist Jeremy Rifkin successfully<br />
overturned a 1984 decision by the US<br />
National Institutes of Health (NIH) to permit<br />
the field testing of recombinant DNA–modified<br />
Pseudomonas syringae bacteria, engineered to<br />
help protect crop plants from frost damage.<br />
Researchers at the University of California,<br />
Berkeley, discovered that P. syringae contains<br />
an ‘ice nucleation’ protein that helps initiate<br />
the growth of ice crystals that damage growing<br />
crops. These scientists deleted the gene sequence<br />
coding for the ice-nucleation protein in the hope<br />
that spraying the resulting ‘ice minus’ variants<br />
on plants could inhibit ice crystal formation and<br />
thereby reduce frost damage.<br />
Rifkin’s Foundation on Economic Trends,<br />
Bethesda, sued to stop the tests and argued that<br />
the NIH had not sufficiently considered, among<br />
other things, the possibility that spraying the<br />
modified bacteria might disrupt wind circulation<br />
patterns and affect aircraft flying overhead.<br />
Not surprisingly, the NIH had dismissed such a<br />
ridiculous theory out of hand because P. syringae<br />
bacteria with a missing or non-functioning<br />
ice-nucleation gene were known to arise spontaneously<br />
in nature owing to natural mutations,<br />
and the permit was for a single small-scale field<br />
trial. Astonishingly, however, both a federal district<br />
court and federal appeals court allowed the<br />
challenge and overturned the NIH decision. It<br />
is little wonder, then, that federal courts would<br />
recognize the possibility of agronomic impacts<br />
as the basis to challenge the approval of recombinant<br />
DNA–modified crop plants.<br />
NEPA’s effect on farmers<br />
Although unsurprising, the resolution of these<br />
lawsuits has been a nightmare for US plant<br />
breeders and farmers. In one case, the plaintiffs<br />
challenged APHIS’s decision to permit<br />
the Scotts, Marysville, Ohio, garden products<br />
company to conduct field trials of glyphosateresistant<br />
turf grass. Another case involved APHIS<br />
field trial permits to four different seed companies<br />
to cultivate small, geographically isolated<br />
plots of corn and sugarcane genetically modified<br />
to produce proteins that would be used in medical<br />
products. The cases of glyphosate-resistant<br />
alfalfa and sugar beet were particularly vexing<br />
to farmers, however, because the lawsuits were<br />
filed after the USDA had approved the varieties<br />
for commercial release and farmers had already<br />
begun to cultivate the seeds.<br />
The alfalfa case involves the 2004 approval<br />
of a glyphosate-resistant variety sold under<br />
the trade name Roundup Ready by its codevelopers<br />
Monsanto and Forage Genetics<br />
(Nampa, ID). To secure approval, the firms<br />
conducted almost 300 government-monitored<br />
field trials over a period of eight years.<br />
APHIS scientists evaluated the data generated<br />
nature biotechnology volume 28 number 12 december 2010 1257
COMMENTARY<br />
© 2010 Nature America, Inc. All rights reserved.<br />
from these trials to determine the likely environmental<br />
impacts of an approval—called ‘deregulation’<br />
in agency parlance—and of widespread<br />
use of Roundup Ready alfalfa by US farmers.<br />
Because the same genetic trait had already<br />
been incorporated into dozens of approved varieties<br />
of corn, soybeans, canola and other crop<br />
plants grown in the United States since 1996,<br />
APHIS was quite familiar with the glyphosateresistance<br />
trait’s likely effects—not just from the<br />
field tests needed to secure approval, but also<br />
from nearly a decade of real-world experience.<br />
Regulators naturally concluded that deregulating<br />
the crop would not have any significant environmental<br />
impact because there is no evidence that<br />
the glyphosate-resistance gene harms humans<br />
or other animals and it was already in common<br />
use in US agriculture. In turn, the USDA<br />
prepared a shorter Environmental Assessment<br />
explaining its “finding of no significant impact”<br />
and deregulated Roundup Ready alfalfa.<br />
Roughly 5,500 farmers across the United<br />
States had planted more than a quarter million<br />
acres of Roundup Ready alfalfa by 2007 when<br />
a federal district judge in San Francisco ruled<br />
that the USDA’s Environmental Assessment<br />
was legally insufficient. The court issued an<br />
injunction revoking the approval and prohibiting<br />
new seeds from being sold until<br />
the USDA completed a full EIS to evaluate<br />
concerns about glyphosate-resistant weeds<br />
and the possible impacts on organic farmers.<br />
Biotech advocates hailed the decision in<br />
June of the US Supreme Court reversing the<br />
blanket injunction on sales of new Roundup<br />
Ready alfalfa seed, but the actually decision<br />
was a purely procedural one that reinforces the<br />
USDA’s obligation to complete an EIS before<br />
permitting more seed sales.<br />
It took two years, but by November of 2009<br />
APHIS had published a draft EIS and solicited<br />
public comments. Unsurprisingly, the EIS concluded<br />
what agency scientists and the scientific<br />
community already knew: the potentially<br />
negative impacts of approving Roundup Ready<br />
alfalfa are minimal and manageable, and they<br />
are in any event far outweighed by the crop’s<br />
many substantial benefits. The USDA must<br />
now publish a final version of the EIS and<br />
reapprove the crop; however, although the substantive<br />
work is complete, the administrative<br />
process of moving to final reapproval could<br />
take many more months and probably will not<br />
be finalized in time for new Roundup Ready<br />
alfalfa seed to be planted in 2011. Fortunately,<br />
the farmers who planted Roundup Ready<br />
alfalfa when it was first approved were permitted<br />
to continue growing and then harvest that<br />
initial crop, so the overall impact will be limited.<br />
A far worse fate is in store for America’s<br />
sugar beet growers.<br />
In August, another federal district judge<br />
revoked the USDA’s approval of Roundup Ready<br />
sugar beets. That decision will sow monumental<br />
confusion because an estimated 95% of the<br />
sugar beets currently grown in the United States<br />
are of the Roundup Ready variety. As was the<br />
case with Roundup Ready alfalfa, the court’s<br />
injunction will permit the already planted crops<br />
to be harvested and processed, but before the<br />
USDA may re-approve the variety, a full EIS will<br />
be necessary. The process will take years, leaving<br />
farmers in regulatory limbo in the meantime.<br />
Perhaps the biggest short-term problem arising<br />
from the decision, however, is a looming shortage<br />
of conventional sugar beet seeds for next<br />
spring’s planting.<br />
Because the Roundup Ready variety so quickly<br />
became popular with American beet growers,<br />
seed companies cut back on their production of<br />
non-genetically engineered seed. According to<br />
Duane Grant, chairman of the Snake River Sugar<br />
(Boise, ID, USA), which produces about 20% of<br />
the nation’s beet sugar, many growers fear they<br />
will have nothing to plant in 2011. Consumers<br />
will surely feel the pinch of sharply rising prices,<br />
and farmers and others who care deeply about<br />
protecting the environment will be denied the<br />
proven beneficial effects of these crops until they<br />
are restored to the marketplace.<br />
Time for NEPA reform<br />
Remarkably, because the NEPA obligation is<br />
purely procedural, US courts are not permitted<br />
to consider the fact that recombinant DNA–<br />
modified herbicide-resistant crop varieties have<br />
offsetting benefits to farmers, consumers and<br />
the natural environment. The mere fact that the<br />
USDA did not properly document its evaluation<br />
of potential negative effects is sufficient grounds<br />
for revoking the approval.<br />
The only function of NEPA is to ensure<br />
that agencies do in fact consider whether their<br />
actions may harm the environment. This is a<br />
laudable goal, but it is manifestly not the intention<br />
of most NEPA litigation. Instead, the statute<br />
has been hijacked by environmental activists to<br />
slow down or prevent government agencies from<br />
taking actions the activists do not like. Because<br />
the law requires agencies to consider almost<br />
any conceivable impact, the statute offers fertile<br />
ground for bad-faith, obstructionist litigation no<br />
matter how meticulous the agency is in preparing<br />
an Environmental Assessment or EIS.<br />
NEPA is therefore a recipe for stagnation, a<br />
particular problem for ‘gatekeeper’ regulatory<br />
agencies that must grant approvals before a<br />
product can be tested or commercialized.<br />
Something must be done to change the system.<br />
But what? Short of substantive reform<br />
of the underlying statute by Congress—the<br />
preferable and definitive solution—agencies<br />
themselves can take some minor steps to<br />
mitigate the Act’s worst effects.<br />
Under the NEPA statute, every agency may<br />
establish a set of ‘categorical exclusions’ that<br />
exempt whole classes or types of activities from<br />
the EIS obligation. These may include routine or<br />
repetitive actions that, on the basis of past experience,<br />
do not involve significant impacts on<br />
natural, cultural, recreational, historic or other<br />
resources; and also those that do not otherwise,<br />
either individually or cumulatively, have any significant<br />
environmental impacts. Because they<br />
fall into those categories, APHIS has already categorically<br />
excluded most small-scale field trials<br />
of recombinant DNA–modified plants.<br />
The exclusion stipulates that all large-scale<br />
field tests, as well as any field release of recombinant<br />
DNA–modified organisms involving<br />
unusual species or novel modifications, still generally<br />
require an Environmental Assessment or<br />
EIS. But the list of excluded or included activities<br />
can be modified through notice-and-comment<br />
rulemaking, in which the agency sets forth the<br />
complete analysis and rationale for excluding<br />
the activity. At the very least, APHIS should<br />
consider categorically excluding some of the<br />
classes of recombinant DNA–modified crops<br />
with which it now has more than two decades’<br />
worth of precommercial and commercial experience,<br />
including herbicide-tolerant varieties of<br />
common crop species.<br />
The most rational and definitive approach,<br />
however, would be to eliminate the agency<br />
action that triggers the NEPA obligation initially—namely,<br />
case-by-case reviews of virtually<br />
all field trials and the commercialization<br />
of recombinant DNA–modified plant varieties.<br />
That would offer the dual advantages of relieving<br />
the USDA’s NEPA difficulties and also making<br />
regulators’ approach to recombinant DNA<br />
technology more scientifically defensible and<br />
risk based. As the scientific community has<br />
explained for over two decades, the decision to<br />
subject recombinant and conventional organisms<br />
to different regulatory standards cannot be<br />
justified scientifically. The increasing prevalence<br />
of obstructionist litigation now shows that doing<br />
so is also wholly impractical.<br />
Recent NEPA lawsuits have prevented the<br />
marketing of products that offer palpable, demonstrated<br />
benefits to farmers, consumers and<br />
the environment. Nuisance litigation intended<br />
to slow the advance of socially responsible technologies<br />
are abusive, irresponsible and antisocial.<br />
And so are those who file them. It is long<br />
past time for NEPA’s burdensome paperwork<br />
requirements to be lifted from such an important<br />
and beneficial technology.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
1258 volume 28 number 12 december 2010 nature biotechnology
FEATURE<br />
To selectivity and beyond<br />
George S Mack<br />
First-generation epigenetic drugs have proven clinically useful in several hematological cancers. But newer enzyme<br />
inhibitors in the pipeline aim to be more selective and promise to broaden the portfolio of therapeutic uses.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Fifty years after covalent chromatin modifications<br />
were first discovered, four drugs that<br />
modulate the cellular epigenetic machinery<br />
are now registered by the US Food and Drug<br />
Administration (FDA) for use in prolonging<br />
the lives of people afflicted with blood and lymphoid<br />
cancers. Although clinical responses to<br />
these first-generation products have been reasonable—given<br />
that they act on a pan-genomic<br />
scale—drug developers are now aiming to produce<br />
a second generation of agents that target<br />
newly recognized enzymes involved in DNA and<br />
chromatin modification. Because these enzymes<br />
alter expression of a narrower, more specific set<br />
of modifications, it is hoped the new experimental<br />
agents will not only have an improved<br />
therapeutic index but also open up epigenetic<br />
therapy to indications outside of cancer.<br />
Early validation<br />
Among the current approved chromatin modulating<br />
agents are drugs that were first investigated<br />
in the 1970s and early 1980s (Table 1).<br />
In a few cases, their clinical value in treating<br />
malignancies only became apparent decades<br />
later. In the intervening years, advances in our<br />
understanding of the mechanisms of chromatin<br />
remodeling and DNA modification have<br />
lent support to the notion that in certain diseases,<br />
the dysregulation of gene expression<br />
arising from alterations in the epigenetic ‘software’<br />
may be corrected by pharmacological<br />
intervention. By correcting aberrant cytosine<br />
methylation in the DNA sequence, or incorrect<br />
acetylation, methylation, phosphorylation,<br />
ubiquitination or sumoylation modifications<br />
of histone proteins, such agents should, in<br />
principle, mitigate abnormal patterns of gene<br />
expression associated with disease.<br />
In the past 15 years, researchers have<br />
begun to identify a battery of enzymes, which<br />
George S. Mack is a freelance writer based in<br />
Columbia, South Carolina.<br />
Unmarking the genome. An artist rendition of HDAC inhibitors removing marks from chromatin.<br />
(Source: Marc Phares, Photo Researchers)<br />
promote or reverse modifications to histones,<br />
and chemical changes to DNA, which<br />
do not alter the nucleotide sequence (Fig. 1).<br />
It is now known that many of these marks<br />
are heritable from gametes to offspring and<br />
from somatic to daughter cells, but at the<br />
same time, demonstrate a level of plasticity<br />
that could be exploited in therapeutic settings.<br />
Epigenetic modifications can also be<br />
environmentally (and perhaps behaviorally)<br />
Table 1 Approved drugs that target epigenetic mechanism<br />
induced, and these, too, may be passed down<br />
through meiosis and mitosis.<br />
Although chromatin interactions affecting<br />
transcription are known to occur at even<br />
greater chromosomal scales, until now, much<br />
of the research on epigenetic machinery has<br />
focused on nucleosomes—nuclear structures<br />
in which the DNA is wrapped around a core<br />
of two copies each of the H2A, H2B, H3 and<br />
H4 histone proteins. It is now known that the<br />
Company Target Agent Indications<br />
Merck HDAC Zolinza (vorinostat) (suberoylanilide<br />
hydroxamic acid, aka SAHA)<br />
Celgene HDAC Istodax (romidepsin) (formerly FK228, a<br />
cyclic peptide principally active against<br />
class 1 HDACs); nanomolar potent<br />
Cutaneous T-cell<br />
lymphoma (CTCL)<br />
CTCL<br />
Development<br />
stage<br />
FDA approved<br />
Oct. 2006<br />
FDA approved<br />
Nov. 5, 2009<br />
Celgene DNMT Vidaza (5-azacitidine) MDS FDA approved<br />
May 2004<br />
Eisai<br />
Tokyo;<br />
sublicensed to<br />
Johnson & Johnson<br />
DNMT Dacogen (decitabine) MDS FDA approved<br />
May 2006<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1259
feature<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Figure 1 The ever-expanding<br />
family of enzymes that control<br />
the structure of chromatin<br />
and gene transcription.<br />
Modifications of the histone<br />
protein tail and DNA are<br />
shown: changes in acetylation<br />
(Ac) by acetyltransferases and<br />
deacetylases, phosphorylation<br />
(PO 4 ) by kinases, ubiquitylation<br />
(Ub) by ligases and changes<br />
in methylation (Me) by<br />
methyltransferases and<br />
demethylases. Methylation of<br />
CpG on DNA is also shown.<br />
(Source: Epizyme)<br />
Ac<br />
Ub<br />
stability of nucleosome occupancy at specific<br />
genomic loci is affected by the presence of particular<br />
variants of core histone proteins, such<br />
as H3.3 and H2A.Z.<br />
Whereas our knowledge of some of the<br />
chromatin modifications associated with<br />
certain diseases is progressing, little is yet<br />
understood about the mechanisms by which<br />
such marks influence pathogenesis of disease.<br />
Epigenetic modifications can either upregulate<br />
or downregulate genes; for example, it<br />
has been assumed with some clinical vindication<br />
that deacetylated histone conditions<br />
bring some tumor suppressor genes to<br />
unfavorably diminished states of expression,<br />
thereby preventing apoptosis and leading to<br />
cell proliferation and advancing disease. In<br />
addition, hypermethylation with 5-methylcytosine<br />
of DNA within CpG islands is seen in<br />
many cancers, and this phenomenon invites<br />
repressive complexes, consisting of methyl-<br />
CpG-binding-domain proteins and DNA<br />
methyltransferases (DNMTs) for maintaining<br />
DNA methylation, resulting in a relatively<br />
durable silencing of the same gene region 1 .<br />
First-generation DNMT and histone<br />
deacetylase (HDAC) inhibitors are largely<br />
indiscriminate in their activity. DNMTs apply<br />
methylation marks to any cytosine on a DNA<br />
strand that is accessible, but the factors governing<br />
which methyl groups are removed by<br />
DNA demethylases remain unclear. It is also<br />
now recognized that cytosine can be marked<br />
with hydroxymethyl groups as well, but the<br />
significance of this modification is currently<br />
only conjecture.<br />
HDAC inhibitors are classified into four<br />
categories: hydroxamic acid derivatives<br />
(including the fungal antibiotic trichostatin<br />
A), short chain fatty acids (e.g., valproic acid),<br />
cyclic peptides (e.g., depsipeptides) and phenylene<br />
diamines. Pharmacological inhibition<br />
of HDAC activity can affect many cellular<br />
processes other than histone function. For<br />
example, HDAC classes I and II target many<br />
K<br />
K<br />
Me<br />
R<br />
K<br />
S<br />
PO 4<br />
96 Histone methyltransferases (HMTs)<br />
44 Arginine methyltransferases (RMTs)<br />
52 Lysine methyltransferases (KMTs)<br />
Me<br />
CpG<br />
nonhistone proteins, including transcription<br />
factors and proteins that regulate cell division,<br />
migration and death, all of which are<br />
presumably affected by HDAC inhibitors.<br />
Class III HDACs make up the so-called sirtuin<br />
family, which has also been the subject<br />
of intense focus in designing treatments for<br />
age-associated diseases, such as activation of<br />
SIRT1 in type 2 diabetes.<br />
Histone acetylation takes place on lysine<br />
residues that reside on the histone tails, and<br />
as far as researchers can tell the histone acetyltransferases<br />
that apply these acetyl marks and<br />
the HDACs that take them off don’t show any<br />
preference for which lysine residue or which<br />
histone is affected, or even for histones, for that<br />
matter. Some researchers now refer to them as<br />
lysine deacetylases rather than HDACs, which<br />
better describes them by their function, as<br />
opposed to defining them by a target.<br />
A greater level of specificity might be<br />
achievable by targeting histone methyltransferases<br />
(HMTs) and demethylases<br />
(HDMs), where a great deal of new research<br />
has been focused over the past eight years.<br />
Nessa Carey, scientific director at epigenetics<br />
startup company CellCentric (Cambridge,<br />
UK), notes that, compared with HDACs<br />
and DNMTs, some HMTs “are very, very<br />
precise.” For example, the histone lysine<br />
Box 1 Noncoding RNAs as targeted epigenetic drugs?<br />
Evidence is growing that natural antisense transcripts and other long, interdispersed,<br />
noncoding RNAs (lincRNAs) are involved in inducing chromatin modifications (and to a<br />
much lesser extent, alteration of sense promoter DNA methylation). The exact mechanisms<br />
by which noncoding RNAs mediate these activities are being worked out and may involve<br />
promotion of DNA methylation by binding to the promoter region of a corresponding gene<br />
sequence (in the sense orientation) in the genome or providing a template onto which<br />
histone-modifying enzymes are then recruited.<br />
Although mechanisms remain unclear, the association of lincRNAs with disease,<br />
particularly cancer, is convincing. The lincRNA H19, one of the first to be discovered,<br />
is associated with several tumor types, including breast, bladder and liver. Furthermore,<br />
under conditions favoring tumorigenesis, such as hypoxia, knocking H19 down affects<br />
the expression of genes involved in angiogenesis, survival and tumorigenesis, suggesting<br />
a central role for the RNA in tumorigenesis 4 . Similarly, recent studies have shown that a<br />
specific lincRNA called HOTAIR, a part of the HOX locus, which is dysregulated in breast<br />
cancer and foreshadows metastasis, appears to do so by reprogramming cells to a more<br />
embryonic state. Overexpression of HOTAIR in epithelial cells causes repositioning of the<br />
Polycomb repressive complex 2 and alterations of methylation at H3 lysine 27 (ref. 5).<br />
Startup CURNA (Jupiter, FL, USA) is attempting to translate this knowledge into<br />
oligonucleotide therapies based on work from the laboratory of Claes Wahlestedt at<br />
the Scripps Research Institute (Jupiter, FL, USA). According to Wahlestedt, of the two<br />
possible natural antisense mechanisms—concordant, in which inhibiting an antisense<br />
transcript directed against a lincRNA has the effect of repressing gene expression, and<br />
discordant, where it has the opposite effect of derepressing or activating genes—CURNA<br />
is focused on discordant effects (that is, upregulation of gene expression). These are<br />
more attractive from a commercial standpoint because most conventional oligonucleotide<br />
therapeutics, monoclonal antibodies or small molecules act by inhibiting the target<br />
rather than activating it. Upregulating proteins in vivo has other advantages in that it gets<br />
around the problem of having to produce proteins synthetically and of delivering them<br />
intracellularly, says Wahlestedt.<br />
CURNA has an exclusive license for the technology from Scripps on natural antisense<br />
and noncoding RNA as drug targets and has filed additional patents on roughly 100 genes<br />
and inhibitor sequences for upregulating pharmaceutically important targets, seven of<br />
which they have now tested in vivo. In all cases, according to Wahlestedt, the oligos had<br />
the desired effect of upregulating their target gene.<br />
1260 volume 28 number 12 DECEMBER 2010 nature biotechnology
feature<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Table 2 Deals centering around epigenetics products or technology<br />
Epigenetics company Partnering company (location) Date Agreement<br />
MethylGene EnVivo Pharmaceuticals March 2004 EnVivo entered into an exclusive agreement to develop an isotypic HDAC inhibitor for<br />
neurodegenerative diseases for $1.1 million upfront<br />
Syndax Bayer Schering Pharma April 2007 Syndax acquired worldwide development and commercial rights to entinostat<br />
(SNDX-275, formerly MS-275)<br />
Celgene Pharmion November 2007 Celgene acquired rights to Vidaza (DNMT inhibitor) by acquiring Pharmion for $2.9 billion<br />
MGI Pharma Eisai January 2008 Eisai acquired rights to Dacogen by acquiring MGI Pharma for $3.9 billion. MGI had<br />
licensed Dacogen from SuperGen in 2004<br />
4SC<br />
Nycomed<br />
(Planegg-Martinsried, (Zurich)<br />
Germany)<br />
Cronos Therapeutics<br />
(Oxfordshire, UK)<br />
Aphios<br />
(Woburn, MA, USA)<br />
Celleron Therapeutics<br />
(Oxon, UK)<br />
SuperGen<br />
(Dublin, CA, USA)<br />
ValiRx<br />
(London)<br />
VivaCell Biotechnology España<br />
(Cordoba, Spain)<br />
AstraZeneca<br />
(London)<br />
GSK<br />
(Brentford, UK)<br />
methyltransferase KMT1E will only methylate<br />
H3K9 (lysine nine on the H3 tail); so<br />
far, it has not been determined to methylate<br />
any of the other possible lysines. “To a<br />
greater or lesser extent,” says Carey, “these<br />
histone methylases and demethylases will<br />
only look at very specific residues on a very<br />
specific histone tail.” Even so, such enzymes<br />
are still working genome-wide rather than<br />
targeting DNA or chromatin modifications<br />
at particular genes. This is because chromatin<br />
modification enzymes are recruited to<br />
specific genes by DNA-binding proteins or<br />
co-factors or RNAs that localize the enzyme<br />
complex to a specific stretch of sequence. In<br />
themselves, they have no inherent specificity<br />
for a particular stretch of DNA.<br />
Hence, one last intriguing possibility for<br />
epigenetic therapeutic intervention is RNA or<br />
cofactors that direct traffic to specific genes.<br />
New evidence continues to be found that antisense<br />
RNAs transcribed from the genome play<br />
a role in DNA methylation, chromatin modifications<br />
and monoallelic expression (e.g.,<br />
in X-chromosome-inactivation–linked or<br />
imprinting disorders). Such transcripts might<br />
be targeted using oligos or oligos designed to<br />
mimic natural antisense mechanisms (Box 1).<br />
July 2008<br />
July 2008<br />
March 2009<br />
May 2009<br />
October 2009<br />
4SC acquired eight oncology programs from Nycomed for €14 million ($17.8 million in<br />
today’s dollars), including the HDAC inhibitors resminostat (4SC-201) and 4SC-202,<br />
plus two other HDAC inhibitor programs<br />
Cronos became a wholly owned subsidiary of ValiRx<br />
Collaborative R&D agreement to develop a combination therapy for treatment of HIV<br />
latency, consisting of bryostatins (PKC modulators) and HDAC inhibitors<br />
Celleron licensed the HDAC inhibitor AZD9468 (now CDX101) from AstraZeneca.<br />
AstraZeneca was granted the option to discuss reacquisition of commercialization rights<br />
in the future<br />
GSK and SuperGen entered into a collaborative agreement where GSK could license<br />
or acquire as many as four discovery-stage and preclinical agents following proof-ofconcept<br />
experiments. Upfront payment of $5 million with potential of $375 million in<br />
royalties, milestones<br />
Celgene Gloucester Pharmaceuticals December 2009 Celgene acquired rights to Istodax (HDAC inhibitor) by acquiring the company for $340<br />
cash plus potential milestones and royalties worth $300 million<br />
Spectrum<br />
Pharmaceuticals<br />
(Irvine, CA, USA)<br />
Topotarget February 2010 Co-development deal for belinostat (HDAC inhibitor). Spectrum licensed rights for North<br />
America and India for upfront payment of $30 million and potential milestones up to<br />
$320 million<br />
CellCentric Takeda February 2010 CellCentric exclusively outlicensed one of its cancer programs to Takeda<br />
Pharmaceutical, which could potentially be worth more than $200 million to CellCentric<br />
over its course<br />
Cellzome GSK March 2010 GSK gains exclusive license to Cellzome’s Episphere technology as applied to immunoinflammatory<br />
disease for an upfront payment of €33 million, with royalties that could equal<br />
an additional €475 million<br />
Greater specificity?<br />
One key question for epigenetic therapy is,<br />
What actually provides target specificity for<br />
a particular HMT, HDM or, for that matter,<br />
any other chromatin-altering enzyme? “That’s<br />
the question I asked when I first looked at<br />
this job,” says Mark Goldsmith, who joined<br />
Constellation Pharmaceuticals (Cambridge,<br />
MA, USA) as CEO in August 2009. “The<br />
complexity of factors upregulating or downregulating<br />
genes is daunting,” he says, “but the<br />
problem is at least becoming more approachable.”<br />
A few puzzle pieces are now starting to<br />
come together to produce the beginnings of a<br />
blueprint that describes where certain marks<br />
cluster within or around genes according to<br />
their activity status. For example, if a histone is<br />
methylated at position H3K9, it tends to be on<br />
genes that are inactive; however, H3K4 methylation<br />
seems to be on active genes.<br />
Some unpublished research in the laboratories<br />
at Constellation Pharmaceuticals has<br />
shown a change of expression in “tens of genes”<br />
when histone methylation enzymes were<br />
inhibited, “not hundreds or thousands” as with<br />
HDAC inhibitors, according to Goldsmith.<br />
“The fundamental premise and strategy of our<br />
company and the field of epigenetics right now<br />
is the belief that these enzymes are much more<br />
selective,” says Goldsmith, whose company is<br />
currently surveying an array of more than 50<br />
HMTs and more than 30 HDMs. “They are<br />
much more selective in what their substrates<br />
are as well as how they are regulated,” he says.<br />
“Most importantly they are selective in their<br />
cumulative effect on target genes. But we are<br />
still in very early-stage discovery.”<br />
Up until now no company has been willing<br />
to show its hand on exactly which enzymes<br />
might be targeted next, but it appears that<br />
one of the first of the new-generation epigenetic<br />
agents advancing to clinical trials will be<br />
a histone demethylator. “I think that over the<br />
next 18 months you’ll see some histone methyltransferase<br />
inhibitors getting into the clinic,”<br />
says Will West of CellCentric. “We know of a<br />
number of parties who are actively working on<br />
that, and we ourselves are, of course, working<br />
in that space as well.” West won’t confirm any<br />
specific targets his company may be going after,<br />
but in late February CellCentric licensed one of<br />
its cancer programs to Asia’s largest drug maker<br />
Takeda Pharmaceutical (Osaka, Japan), which is<br />
looking to develop new oncology products out<br />
of this deal, which could be worth more than<br />
$200 million to CellCentric (Table 2).<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1261
feature<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Histones<br />
DNA<br />
PRC1<br />
PRC2/3 EZH2<br />
DNMTs<br />
Methylation of<br />
histone tails<br />
Methylation<br />
of DNA<br />
Repressed genes<br />
Repressive<br />
complexes<br />
Takeda oncology research division leader<br />
Osamu Nakanishi says his company is looking<br />
into other unnamed epigenetic enzymes<br />
as well as the histone methylases CellCentric<br />
has identified. “Epigenetics controls so many<br />
things,” he says. “And it has decisive power<br />
for inducing cancer.” Although his work at<br />
Takeda is limited to oncology, Nakanishi ponders<br />
other possibilities, and he suggests that<br />
it would be interesting to look into targeting<br />
epigenetic enzymes for indications beyond<br />
cancers, particularly in metabolic disease,<br />
such as diabetes, and central nervous system<br />
(CNS) disorders, such as schizophrenia and<br />
depression (Box 2). The CNS connection is<br />
noteworthy because neurons are rarely mitotic<br />
after they are in place, and thus gene mutation<br />
would not be a likely culprit in disease etiology<br />
as it often is in cancers. “But epigenetic<br />
mutation or changes are certainly a possibility<br />
in CNS disease,” says Nakanishi. He sums up<br />
Takeda’s perspective on next-generation epigenetic<br />
therapies: “HDACs are so nonselective<br />
that even if we had a very specific HDAC<br />
inhibitor, we would still see toxicity, and that<br />
would limit dose,” he says. “Also, it is still not<br />
clear that even very selective HDAC inhibitors<br />
could be applied to specific diseases.”<br />
Although the new crop of epigenetics companies<br />
has been understandably coy about<br />
what newly identified targets they might be<br />
pursuing, some candidates have been receiving<br />
Figure 2 Making marks. The Polycomb group protein EZH2 adds methyl groups to histone proteins,<br />
primarily to lysine 27 of histone H3 (small blue circles). EZH2 recruits DNA methyltransferases<br />
(DNMTs, small green circles) to certain target genes, either setting up new methyl marks during<br />
development or maintaining the marks after each round of cell division. These marks could recruit<br />
further repressive complexes, such as PRC1 or HDAC co-repressors, through methyl-binding<br />
domain proteins. (Reprinted with permission from Taghavi, P. and van Lohuizen, M., Nature 439,<br />
794–795, 2006)<br />
wide attention. In the histone methylation pathway,<br />
there is a group of polycomb proteins, one<br />
of which is the enhancer of zeste homolog 2<br />
(EZH2), which trimethylates H3K27 and is<br />
currently drawing lots of interest from investigators<br />
(Fig. 2). Certain mutations occurring<br />
in the gene encoding the HDM UTX result in<br />
copious EZH2 expression, and there is strong<br />
in vivo evidence that inhibiting EZH2 function<br />
reduces histone methylation at the one particular<br />
lysine residue (H3K27) and significantly<br />
reduces tumor cell proliferation in various<br />
models 2 . Furthermore, when EZH2 expression<br />
is high in solid tumors, it tends to be a poor<br />
indicator of prognosis.<br />
Aside from these observations, one other<br />
reason for the interest is that EZH2 overexpression<br />
seems to cause gene silencing<br />
in parallel with DNA hypermethylation 3 .<br />
Translational investigator and medical<br />
oncologist Jean-Pierre Issa of the University<br />
of Texas M.D. Anderson Cancer Center<br />
(Houston) has been working with EZH2, and<br />
he believes it’s going to be an important target<br />
if an agent can be developed to counteract it.<br />
“We have evidence that the EZH2 mechanism<br />
is used in the same way DNA methylation is<br />
used to turn off tumor suppressor genes,” he<br />
says. “If you inhibit its expression, it abolishes<br />
the growth of the tumor and induces senescence<br />
in some cancers. I suspect that this will<br />
be the first target that companies are developing<br />
drugs for.<br />
Box 2 Marks in the brain<br />
Epigenetics company Epizyme (Cambridge,<br />
MA, USA) may have a leg up with EZH2<br />
because it holds an exclusive license for this as<br />
well as a second HMT, DOT1L, based on work<br />
done by Yi Zhang at the University of North<br />
Carolina at Chapel Hill. According to Zhang—<br />
who in 2007, along with Robert Horvitz of MIT,<br />
co-founded Epizyme—the company has rights<br />
to the two best characterized HMTs, EZH2<br />
and HOT1L. Epizyme CEO Robert Gould says<br />
that RNA interference (RNAi) screens carried<br />
out in the company to downregulate HMTs<br />
associated with cancer have demonstrated that<br />
such changes influence both epigenetic marks<br />
and invasive phenotypes. Epizyme is currently<br />
screening for small molecules that recapitulate<br />
the same effect as RNAi and optimizing the<br />
resulting candidate drugs in-house.<br />
Expanding indications<br />
There has been considerable debate as to why<br />
epigenetic inhibitors have been more successful<br />
in hematologic cancers than in other<br />
malignancies; indeed, solid tumor results with<br />
HDAC and DNMT inhibitors have been lackluster<br />
at best. Investigators have been looking<br />
to the future, anticipating that more selective<br />
enzyme antagonists would improve results.<br />
Researchers suspect that there are likely hundreds<br />
of epigenetic modifiers, and therefore<br />
the surface has hardly been scratched in determining,<br />
which are the best enzymes to inhibit<br />
or modulate in some way.<br />
Several neurodegenerative diseases have their etiology linked to epigenetics. Rett’s<br />
syndrome, the first for which the link was made in 1999, is caused by a mutation in<br />
methyl-CpG-binding-domain protein. Several syndromes that have an expansion of<br />
repeats, such as Fragile X, in which a CGG triplet is repeated, and Friedrich’s ataxia, in<br />
which GAA X TCC triplets are expanded within an intron in the gene encoding Frataxin,<br />
show silencing of critical genes due to hypermethylation. In the case of Friedrich’s<br />
ataxia, silencing of Frataxin is accompanied by hypoacetylation of histones 3 and 4 and<br />
trimethylation of H3L9. Researchers at Scripps Clinic (La Jolla, CA, USA) developed<br />
a class of HDAC inhibitors that reversed the Frataxin silencing 6 , which Repligen<br />
(Waltham, MA, USA) has licensed and is taking to the clinic. The company filed an<br />
investigational new drug application for one drug, RG2833, and was granted orphan<br />
drug status by the FDA in May of this year.<br />
Based on a Drosophila model of neurodegenerative disease, EnVivo (Watertown, MA,<br />
USA) claims to have developed the only CNS-penetrant HDAC inhibitor, which it reported<br />
in 2008. According to its website, the company is exploring this drug for CNS indications,<br />
like Alzheimer’s and Parkinson’s disease.<br />
1262 volume 28 number 12 DECEMBER 2010 nature biotechnology
feature<br />
One reason why HDAC and DNMT inhibitors<br />
may have been less successful in solid<br />
tumors is that, like any other new, experimental<br />
therapy, they have often been tested in individuals<br />
with advanced disease, who have already<br />
been through several cycles of treatment with<br />
chemotherapies or with targeted cytostatic<br />
agents and are therefore very sick. In solid<br />
tumors, HDAC inhibitors have been tested as<br />
fifth- and sixth-line therapies, in patients who<br />
have already become multi-drug resistant. In<br />
contrast to solid tumors, there were no good<br />
first-line therapies available for myelodysplastic<br />
syndromes (MDS) and, as a result, DNMT<br />
inhibitors were used in people naive to therapy.<br />
“This was just an accident of drug development,”<br />
says Jean-Pierre Issa who was principle<br />
investigator for the pivotal trial that preceded<br />
FDA approval of DNMT inhibitor Dacogen<br />
(decitabine) in May 2006. “To my knowledge,<br />
no one has really tested these drugs [DNMT<br />
inhibitors] in solid tumors as first-line therapies<br />
to determine if their activity is actually substantially<br />
lower than it is in MDS.” Issa designs<br />
clinical trials, and sees patients who have newly<br />
diagnosed disease that is far advanced. He says<br />
he would like to see oncologists speak frankly<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Table 3 Select epigenetic inhibitors in various stages of development<br />
Company Target Agent Indications Development stage<br />
HDAC inhibitors<br />
Merck HDAC Zolinza (vorinostat; suberoylanilide hydroxamic<br />
acid, SAHA). Nonselective, micromolar potent<br />
NSCLC<br />
Multiple myeloma<br />
Mesothelioma<br />
GBM<br />
Latent HIV infection in T cells<br />
Celgene HDAC Istodax (romidepsin) (formerly FK228) Peripheral T-cell lymphoma<br />
Selective for class I HDACs; nanomolar potent Multiple myeloma<br />
Pfizer<br />
New York<br />
HDAC1,<br />
3, 6 and 8<br />
CI-994 (N-acetyldinaline, also tacedinaline)<br />
Originally developed as an anti-convulsive<br />
NSCLC<br />
Pancreatic cancer<br />
MM<br />
Topotarget HDAC Belinostat (PXD101). Hydroxamic acid-type Peripheral T-cell lymphoma (monotherapy)<br />
CUP (BelCaP)<br />
Ovarian cancer (mono and with carboplatin)<br />
Hepatocarcinoma (monotherapy)<br />
Soft tissue sarcoma (mono and with doxorubicin)<br />
NSCLC (mono and in combination)<br />
AML and MDS<br />
Novartis HDAC Panobinostat (LBH589). Nonselective, very<br />
potent. Discovered in house by Peter Atadja<br />
Italfarmaco<br />
Milan<br />
Syndax Pharmaceuticals<br />
MethylGene<br />
Montreal<br />
Multiple myeloma<br />
CTCL<br />
CML<br />
Hodgkin’s lymphoma<br />
Metastatic melanoma<br />
Prostate cancer<br />
HDAC Givinostat (ITF2357) Juvenile idiopathic arthritis<br />
Hodgkin’s lymphoma<br />
Polycythemia vera<br />
HDAC1,<br />
2 and 3<br />
HDAC<br />
Hos2/<br />
HDAC<br />
Entinostat (SNDX-275; formerly MS-275)<br />
Selective for HDACs 1, 2 and 3<br />
Mocetinostat (MGCD0103)<br />
Lymphoma<br />
Selective for class I HDACs<br />
AML<br />
MGCD290 selective for fungal HDAC enzymes Fungal infections<br />
EnVivo Pharmaceuticals HDAC EVP-0334 (central nervous system<br />
(CNS) penetrant)<br />
Sirtris<br />
Cambridge, MA, USA<br />
(Unit of GlaxoSmithKline)<br />
Curis<br />
Cambridge, MA, USA<br />
Johnson & Johnson<br />
New Brunswick, NJ, USA<br />
Pharmacyclics<br />
Sunnyvale, CA, USA<br />
HDAC<br />
(SIRT1)<br />
HDAC<br />
EGFR/<br />
ErbB1<br />
HER2/<br />
neu<br />
HDAC<br />
HDAC<br />
HDAC8<br />
SRT501<br />
Selective for Class III HDACs<br />
Phase 3<br />
Phase 3<br />
Phase 2<br />
Discovery<br />
Phase 2<br />
Phase 2<br />
Phase 3<br />
Phase 2<br />
Phase 2<br />
Phase 2<br />
Phase 2<br />
Phase 2<br />
Phase 2<br />
Phase 1/2<br />
Phase 1/2<br />
Phase 1/2<br />
NSCLC (in combination with Tarceva (erlotinib)) Phase 2<br />
Breast cancer (in combination with aromatase Phase 2<br />
inhibitors)<br />
Leukemia and MDS (in combination with Vidaza Phase 2<br />
(5-azacitidine))<br />
Alzheimer’s disease and other CNS indications<br />
Schizophrenia<br />
Colon cancer<br />
Multiple myeloma<br />
Phase 3<br />
Phase 3<br />
Phase 3<br />
Phase 3 (not open yet)<br />
Phase 2<br />
Phase 2<br />
Phase 2<br />
Phase 2<br />
Phase 2<br />
Phase 2<br />
Phase 2<br />
Phase 1<br />
Phase 1<br />
Preclinical<br />
Phase 2<br />
Phase 2<br />
CUDC-101<br />
(a hydroxamic acid moiety integrated into a<br />
quinazoline receptor tyrosine kinase inhibition<br />
pharmacophore)<br />
Advanced refractory tumors Phase 1<br />
JNJ-26481585<br />
Selective for class I HDACs<br />
Potential for enhanced HSP70 upregulation<br />
and Bcl-2 downregulation<br />
PCI24781 (broad range multi-isoform HDAC<br />
inhibitor)<br />
PCI-34051<br />
Selective for HDAC8<br />
MDS<br />
CML<br />
CLL<br />
B-cell lymphomas<br />
BCR-ABL positive leukemia<br />
Sarcomas<br />
Lymphomas, MM, CLL<br />
T-cell lymphomas<br />
Pediatric neuroblastoma<br />
Autoimmune disease<br />
Phase 1<br />
Phase 1 and 2<br />
Phase 1<br />
Preclinical<br />
(continued)<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1263
feature<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Table 3 Select epigenetic inhibitors in various stages of development (continued)<br />
Company Target Agent Indications Development stage<br />
DNMT inhibitors<br />
Celgene DNMT Vidaza (5-azacitidine) AML Phase 2<br />
Eisai<br />
DNMT Dacogen (decitabine) CML<br />
Phase 3<br />
Tokyo<br />
AML<br />
Phase 3<br />
Sublicensed by Eisai to<br />
AML (elderly)<br />
Phase 2<br />
Johnson & Johnson<br />
Ovarian cancer<br />
Phase 2<br />
HMT inhibitors<br />
Takeda Pharmaceutical<br />
Osaka<br />
(Licensed from CellCentric)<br />
HMTs<br />
Histone<br />
ubiquitinrelated<br />
enzymes<br />
with patients about using investigational drugs<br />
in late-stage disease before traditional cytotoxic<br />
agents have been tried.<br />
There is a generation of newer HDAC<br />
inhibitors coming online, including Celgene’s<br />
(Summit, NJ, USA) Istodax (romidepsin),<br />
which shows highest activity against class 1<br />
and was approved in November 2009. There<br />
is also the very potent nonselective HDAC<br />
HMT inhibitors Cancers Discovery<br />
CellCentric<br />
HMTs HMT inhibitors Cancers Discovery<br />
Histone<br />
ubiquitinrelated<br />
enzymes<br />
HDM<br />
Cellzome HMTs HMT inhibitors Inflammatory and autoimmune diseases, Discovery<br />
such as rheumatoid arthritis, multiple sclerosis,<br />
and inflammatory bowel disease<br />
Epizyme HMTs HMT inhibitors Cancers Discovery<br />
Constellation<br />
HMTs HMT inhibitors Cancers Discovery<br />
Pharmaceuticals<br />
HDAC, histone deacetylase; DNMT, DNA methyltransferase; HMT, histone methyltransferase; HDM, histone demethylase; CUP, cancers of unknown primary origin; GBM,<br />
glioblastoma multiforme; MM, multiple myeloma; NSCLC, non-small cell lung cancer; CLL, chronic lymphocytic leukemia; CML, chronic myelogenous leukemia; MDS,<br />
myelodysplastic syndromes; AML, acute myeloid leukemia.<br />
inhibitor panobinostat (LBH589) being developed<br />
by Novartis (Basel), which is in phase 3<br />
trials for cutaneous T-cell lymphoma (CTCL),<br />
as well phase 2 for prostate cancer and metastatic<br />
melanoma. “I don’t think we’ve heard the<br />
end of histone deacetylase inhibitors, perhaps<br />
not as single agents but in combination with<br />
other drugs,” says Issa (Box 3). “I’m confident<br />
we’re going to find more applications for them.”<br />
Takeda’s Nakanishi doesn’t want to shut the door<br />
on existing drugs either. He proposes that there<br />
may be a way to make HDAC inhibitors work<br />
more effectively by combination with other epigenetic<br />
modulators. “That would be a potential<br />
strategy for regulating more specific, narrower<br />
gene sets,” he says (Table 3).<br />
Several biotechs as well as some pharma<br />
companies have programs in clinical trials<br />
Box 3 It takes two (or maybe more)<br />
Several companies are looking beyond single agents to drug<br />
combinations. Topotarget (Copenhagen) has a pan-HDAC inhibitor,<br />
belinostat, in multiple clinical trials as a single agent (Table 3),<br />
and in combination with a number of chemotherapeutic agents,<br />
including carboplatin, a DNA-modifying agent, and Taxol (paclitaxel),<br />
which interacts with microtubules. The rationale is that if you open<br />
up the chromatin with an HDAC inhibitor, there will be greater<br />
access for DNA-damaging agents. In the case of the belinostat and<br />
Taxol combination, the two drugs have shown synergy in preclinical<br />
studies that looked at tubulin acetylation and antiproliferative<br />
activity. Acetylation of tubulin stabilizes microtubule formation and<br />
thus adversely affects the ability of cancer cells to divide. Henri<br />
Lichtenstein, chief development officer at Topotarget, says their<br />
dose escalation studies in solid tumors show that belinostat could be<br />
combined safely with therapeutic doses of the two chemotherapies.<br />
They looked further into ovarian and bladder cancer and had good<br />
results even with platinum-resistant tumors, which has gotten the<br />
attention of the gynecological group at the US National Cancer<br />
Institute (Bethesda, MD, USA), which is sponsoring further clinical<br />
studies in gynecological cancers. The company is also completing<br />
a proof-of-concept study with belinostat plus carboplatin and Taxol<br />
(BelCaP) in cancers of unknown primary origin (CUP). “What’s also<br />
very interesting [with CUP] is that we can go into first-line patients<br />
since there is no effective treatment,” says Lichtenstein.<br />
Elsewhere, quasi-virtual biotech company Syndax (Waltham,<br />
MA, USA) is focused entirely on combination therapies in solid<br />
tumors, where the need for good therapies is the greatest. Based on<br />
work from Ron Evans’ laboratory at the Salk Institute (La Jolla, CA,<br />
USA) on regulating the expression of nuclear hormone receptors,<br />
Syndax is attempting to reverse drug resistance, in particular<br />
resistance to hormone therapy that develops in some breast cancers.<br />
According to CEO Joanna Horobin, “[Combining HDAC inhibitors<br />
with conventional chemotherapy] has a well-trodden regulatory<br />
path, but it was not clear to us that it was playing to the specific<br />
science.” Instead, they are trying combinations that “leverage the<br />
role of modulating genes that interfere with cell growth in cancer,<br />
particularly in [a] solid tumor setting,” she says. The company has<br />
two phase 2 programs in metastatic breast cancer in combination<br />
with aromatase inhibitors, and in lung cancer combining Tarceva<br />
(erlotinib) and Vidaza (5-azacitidine).<br />
1264 volume 28 number 12 DECEMBER 2010 nature biotechnology
feature<br />
© 2010 Nature America, Inc. All rights reserved.<br />
targeting HDACs (Table 3). Furthermore,<br />
the number of HDAC inhibitor programs in<br />
all stages of development greatly outnumbers<br />
any other class of epigenetic targets (Fig. 3).<br />
MethylGene (Montreal), which was among<br />
the first companies to focus on epigenetic<br />
targets and the first to commit to identifying<br />
isoform-selective inhibitors, according to<br />
Jeffrey Besterman, CSO at MethylGene, currently<br />
has a class 1 ‘selective’ HDAC inhibitor,<br />
mocetinostat (MGCD-0103), an orally active<br />
phenylene diamine derivative currently in<br />
phase 2 trials for follicular lymphoma. The<br />
company also has another HDAC inhibitor,<br />
MGCD-290, that is a specific inhibitor for the<br />
fungal HDAC HOS2. Results of a phase 1 clinical<br />
trial of this compound were reported at the<br />
Interscience Conference on Antimicrobial<br />
Agents and Chemotherapy in September and<br />
Box 4 Diagnostic test development<br />
There has been explosive growth in<br />
commercial operations offering tests<br />
for detecting methylation patterns in<br />
certain malignancies. Basic epigenetic<br />
knowledge can be more rapidly<br />
translated into diagnostic products<br />
because the path to marketing is<br />
abbreviated compared with drug<br />
development, and usually only Clinical<br />
Laboratory Improvement Amendments<br />
certification is required for companies<br />
to set up assays in reference pathology<br />
laboratories in the United States.<br />
On the other hand, tests often have<br />
shortened life cycles due to the<br />
introduction of new and improved<br />
versions by competitors.<br />
Two colorectal cancer (CRC) blood<br />
assays based on DNA hypermethylation<br />
are in late-stage development and<br />
clinical use (Table 4) to noninvasively<br />
check for nascent disease or to see if<br />
an individual is at risk for cancer. One<br />
is being developed to determine loss of<br />
imprinting of insulin-like growth factor 2<br />
(IGF2) as a potential risk factor for<br />
colorectal cancer in younger people by<br />
Orion Genomics (St. Louis), based on<br />
technology Orion exclusively licensed<br />
from Johns Hopkins University (JHU;<br />
Baltimore). Loss of imprinting creates<br />
imbalance through increased expression,<br />
and indeed, loss of IGF2 imprinting<br />
has been shown to occur in nearly every<br />
human tumor that has been examined.<br />
Colon cancer is a particular instance<br />
where both copies of the gene are<br />
being expressed inappropriately. “There<br />
probably should be no IGF2 production<br />
Figure 3 Programs in epigenetics by phase of<br />
development. (Source: CellCentric)<br />
showed the drug was well tolerated, with no<br />
drug-drug interactions at the highest dose.<br />
Looking forward<br />
Researchers anticipate that greater selectivity<br />
will be a major factor for improvements<br />
in clinical responses to epigenetic inhibitors.<br />
Although drug development programs<br />
often seek to increase therapeutic index by<br />
increasing target selectivity, the latter doesn’t<br />
always yield the desired results.<br />
In a collaboration that took place during<br />
2005 and 2006 between Boehringer Ingelheim<br />
(Ingelheim am Rhein, Germany) and the<br />
Research Institute of Molecular Pathology<br />
Number of programs<br />
Table 4 Select epigenetics diagnostics companies<br />
60<br />
40<br />
20<br />
0<br />
Discovery<br />
Company Technology Products or development stage<br />
Epigenomics<br />
Epiontis<br />
(Berlin)<br />
Spin-off from<br />
Epigenomics in<br />
July 2003<br />
Exact Sciences<br />
(Madison, WI, USA)<br />
Orion Genomics<br />
Oncomethylome<br />
Sciences<br />
(Liege, Belgium)<br />
Diagnostics and<br />
biomarkers by<br />
means of DNA<br />
methylation<br />
analysis<br />
Diagnostics and<br />
biomarkers via<br />
DNA methylation<br />
analysis<br />
Multiplex DNA<br />
assay<br />
Diagnostics and<br />
biomarkers via<br />
DNA methylation<br />
analysis<br />
Diagnostics and<br />
biomarkers via<br />
DNA methylation<br />
analysis<br />
Phase 1<br />
HDAC<br />
DNMT<br />
Histone demethylase<br />
Histone methyltransferase<br />
Histone ubiquitin ligase<br />
Histone-ubiquitin-specific proteases<br />
Phase 2<br />
Phase 3<br />
Launched<br />
(IMP, Vienna), which is principally supported<br />
by Boehringer, more than 125,000<br />
compounds were assayed by means of highthroughput<br />
screens to find suitable inhibitors<br />
Main product is mSept9, a test based on methylation of SEPT9 for<br />
early detection of colorectal cancer via blood sample. CE marked<br />
in October 2009 and marketed in the EU. Licensed to Abbott<br />
Molecular for its m2000 real-time PCR automated system. Abbott<br />
will submit a PMA in late 2010. Quest Diagnostics will offer the test<br />
in the US<br />
Lung cancer via bronchial lavage sample expected to be launched in<br />
EU in 2010<br />
Prostate cancer via urine sample in development<br />
Foxp3 marker for regulatory T cells for cancer diagnostics and<br />
screening in development<br />
T-cell marker for cancer screening in development<br />
CD4 T cell (naive and memory) assay in development for immune<br />
monitoring<br />
Collaboration with Genzyme’s cartilage repair product, Carticel, to<br />
develop biomarker assays for cell purity and quality of product<br />
Stool-based DNA test that combines mutation testing, methylation<br />
marker analysis and immunochemical tests<br />
Blood test for detection of loss of imprinting of the pro-growth,<br />
anti-apoptotic IGF2 gene as an assay for colorectal cancer risk.<br />
Technology is licensed from Johns Hopkins University and is in<br />
development<br />
Bladder, lung and ovarian cancer assays in development<br />
Prostate cancer via urine and tissue samples. Partnered with<br />
LabCorp and Johnson & Johnson. Late-stage development<br />
Other cancers partnered with LabCorp, Schering-Plough (now Merck<br />
& Co.) and Merck in development.<br />
• Colorectal cancer screening via blood and stool in development,<br />
partnered with LabCorp<br />
• Bladder cancer testing via urine sample in development<br />
• Lung cancer in early stage development<br />
• Pharmacogenomic tests in early stage development with partners<br />
Abbott Molecular and GSK Biologicals<br />
Valirx (London) Cancer diagnostics • Nucleosomics cancer detection test measures specific histone<br />
modifications that correlate with certain cancers<br />
Sequenom<br />
(San Diego)<br />
Enzo Life Sciences<br />
(Farmingdale, NY,<br />
USA)<br />
Tools for research,<br />
including software<br />
Tools for research<br />
• EpiPanels for cancers and imprinting; marketed<br />
Tools for histone- and DNA-modifying enzymes, for protein methylation<br />
and demethylation, and development of kits for HDACs and<br />
lysine-specific demethylases<br />
PMA, premarket approval; CE, meets requirements of applicable European directive.<br />
(continued)<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1265
feature<br />
Box 4 Diagnostic test development (continued)<br />
© 2010 Nature America, Inc. All rights reserved.<br />
in colon cells—or at least you should have production from<br />
one allele only,” says epigenetics researcher and gene imprint<br />
expert Randy Jirtle of Duke University (Durham, NC, USA).<br />
Researchers from Orion and JHU are collaborating to improve<br />
Orion’s IGF2 assay, and independent of Orion, JHU is conducting<br />
large-scale prospective trials to establish the lifetime cancer<br />
risk. Prior published retrospective studies involving more than<br />
200 patients have demonstrated that CRC patients are 21 times<br />
more likely to harbor the IGF2 biomarker in their blood compared<br />
with appropriate cancer-free subjects 7 . “When clinical trials are<br />
completed, we will market the Orion CRC Risk Test to the 81.6<br />
million people in their twenties and thirties in the United States<br />
and to the primary care physicians who treat them,” says Nathan<br />
Lakey, Orion’s CEO.<br />
Another CRC test based on loss of imprinting of SEPT9 is<br />
already being marketed in the European Union (Brussels) by<br />
Epigenomics (Berlin and Seattle). The Epigenomics CRC test<br />
has not yet been approved for use in the United States, but the<br />
company’s nonexclusive licensee Quest Diagnostics (Madison,<br />
NJ, USA) is set to market those laboratory services to physicians<br />
and patients once the test has been approved. If CRC can be<br />
identified at an early stage while still localized, the 5-year<br />
survival rate is high, ~90%. Yet only 40% of colorectal cancers<br />
are diagnosed as early stage disease, one reason being that there<br />
is resistance to colonoscopy exams due to fear of discomfort as<br />
well as economic concerns. The SEPT9 biomarker accurately<br />
picks up 62.75% of tumors, which is more precise than current<br />
methods of testing fecal samples.<br />
for the H3K9me2 HMT G9a, which is implicated<br />
in such cancers as liver and prostate.<br />
Investigators ended up with seven confirmed<br />
hits and ultimately identified a single smallmolecule<br />
compound BIX-01294 (a diazepinquinazolin-amine<br />
derivative), which was<br />
specific for G9a. “It had only modest potency,”<br />
says Boehringer’s medicinal chemist and<br />
senior principal scientist Michael August.<br />
“At the time, the project didn’t fit any of our<br />
therapeutic program objectives,” he says, but<br />
Boehringer did make the compound available<br />
to academic and translational researchers<br />
through a material transfer agreement.<br />
“There has been a lot of interest in the compound<br />
from the academic community,” he<br />
says. “We sent out a lot of samples to various<br />
groups.” Post-doctoral fellow Stefan Kubicek<br />
of the Broad Institute (Cambridge, MA, USA)<br />
was principal laboratory investigator on the<br />
screening project at IMP when he was a graduate<br />
student; he has found BIX-01294 to be<br />
“very selective,” with regulation of as few as 50<br />
genes, 20 of which are upregulated “in a common<br />
biochemical pathway.” But toxicity has<br />
been “a major problem in preclinical mouse<br />
models,” he says.<br />
The future will undoubtedly bring greater<br />
selectivity and precision in epigenetic therapies.<br />
In the meantime, there are nearer-term<br />
commercial opportunities in molecular diagnostics<br />
(Box 4 and Table 4). What is clear is<br />
that advances in our understanding of chromatin<br />
remodeling processes and their control<br />
is unlikely to translate into panaceas, especially<br />
in cancers that seem to escape and go around<br />
Early diagnosis can give patients the best prognosis, no matter<br />
the disease. With autism spectrum disorders (ASDs), there is no<br />
specific blood test or imaging study that can deliver a definitive<br />
diagnosis so that young children can begin treatment when it is<br />
likely to be most effective. Because autism spectrum disorders<br />
(ASD) occur three to six times more frequently in males than in<br />
females, clinical genetics researchers Julie Jones and<br />
Michael Friez at the Greenwood Genetic Center (GGC; Greenwood,<br />
SC, USA) are working on the hypothesis that the syndrome is<br />
X linked 8 . With chip developer Oxford Genome Technologies<br />
(OTG; Oxford, UK), they have developed a microarray system to<br />
analyze expression at GGC and so far have identified a subset<br />
of eight genes on the X chromosome that are differentially<br />
overexpressed or underexpressed in a cohort of 17 boys with<br />
autism. A larger study is under way, and so far, the trend of<br />
dysregulation continues in the new data set, but the eight genes<br />
do not stand out to the degree they did in the first round of data.<br />
Given the heterogeneity of patients, screening so few genes<br />
to obtain a definitive diagnosis might not be possible. What is<br />
apparent is that there is a greater degree of gene dysregulation in<br />
the autism cohort compared with the controls. “We are working<br />
with OGT to come up with the most appropriate methylation<br />
array design based on this most recent data set,” says Friez. The<br />
etiology of ASDs has been daunting for investigators. “You could<br />
argue that it’s really a chromosome- or genome-wide dysregulation<br />
of methylation and that it may just be a vulnerable location in<br />
males only because there’s not a second copy of the<br />
X chromosome,” says Friez. “We just don’t know yet.”<br />
other modalities. “So far, these drugs [HDAC<br />
and DNMT inhibitors] have been helpful but<br />
not curative,” says Issa. “I don’t believe there is<br />
anything magical about epigenetic therapy as<br />
opposed to other kinds of therapies. Resistance<br />
often develops, and this is the nature of cancer,”<br />
he says. “I think we just have to deal with<br />
it and look for other epigenetic approaches<br />
because multiple modality therapy is the way<br />
to overcome resistance in general.”<br />
1. Bostick, M. et al. Science 317, 1760–1764 (2007).<br />
2. Sooryanarayana, V. Nature 419, 624–629 (2002).<br />
3. Kondo, Y. et al. Nat. Genet. 40, 741–750 (2008).<br />
4. Matouk, I.J. et al. PLoS ONE 2, e845 (2007).<br />
5. Gupta, R.A. et al. Nature 464, 1071–1076 (2009).<br />
6. Herman, D. et al. Nat. Chem. Biol. 2, 551–558<br />
(2006).<br />
7. Cui, H. et al. Science 299, 1753–1755 (2003).<br />
8. Jones, J.R. et al. Am. J. Med. Genet. A. 146A, 2213–<br />
2220 (2008).<br />
1266 volume 28 number 12 DECEMBER 2010 nature biotechnology
patents<br />
On firm ground: IP protection of therapeutic<br />
nanoparticles<br />
Paul Burgess, Peter Barton Hutt, Omid C Farokhzad, Robert Langer, Scott Minick & Stephen Zale<br />
The development of generic therapeutic nanoparticles faces a combination of scientific, patent and regulatory<br />
challenges.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Over the past decade, nanotechnology has<br />
shown the potential to dramatically affect<br />
several fields. In the healthcare sector, there has<br />
been an intense investment in the development<br />
of therapeutic nanoparticles for treatment of<br />
common diseases such as cancer, inflammatory<br />
disorders, infectious disease and cardiovascular<br />
disease. Despite the significant potential<br />
benefits of these products to patients and their<br />
families, no authoritative article has addressed<br />
the intellectual property (IP), regulatory, scientific,<br />
clinical and commercial advantages for<br />
therapeutic nanoparticles. This article outlines<br />
the reasons why innovative pharmaceutical<br />
products based on therapeutic nanoparticles<br />
offer particular advantages in IP protection<br />
from generic competition.<br />
What is a therapeutic nanoparticle?<br />
Parenterally administered therapeutic nanoparticles<br />
typically comprise an active ingredient<br />
together with organic or inorganic biomaterials,<br />
and range from 50 to 200 nm in diameter.<br />
Several classes of therapeutic nanoparticles,<br />
commercially available or in development,<br />
include liposomes, albumin-drug complexes<br />
and polymeric nanoparticles. The earliest therapeutic<br />
nanoparticles were liposomal drugs. For<br />
example, Doxil, which was approved by the US<br />
Food and Drug Administration (FDA) in 1995<br />
for treatment of Kaposi’s sarcoma, is a long-<br />
Paul Burgess, Scott Minick and Stephen<br />
Zale are at BIND Biosciences, Cambridge,<br />
Massachusetts, USA; Peter Barton Hutt is at<br />
Covington & Burling, Washington, DC, USA;<br />
Omid C. Farokhzad is at Brigham and Women’s<br />
Hospital and Harvard Medical School, Boston,<br />
Massachusetts, USA; and Robert Langer is<br />
at Massachusetts Institute of Technology,<br />
Cambridge, Massachusetts, USA.<br />
e-mail: paulburgess@bindbio.com<br />
Formulation patents<br />
Conventional<br />
drug product<br />
circulating liposome containing doxorubicin 1 .<br />
More recently, there has been interest in therapeutic<br />
nanoparticles composed of biodegradable<br />
polymers, which offer long circulation<br />
times characteristic of liposomes together with<br />
controllable drug-release kinetics mediated<br />
by polymer biodegradation 2 . After parenteral<br />
administration, these therapeutic nanoparticles<br />
can selectively accumulate in particular tissues<br />
or body locations, thereby enhancing the delivery<br />
of the drug payload to the site of disease.<br />
Many therapeutic nanoparticles access diseased<br />
tissue through the enhanced permeability and<br />
retention (EPR) effect. The EPR effect occurs<br />
in tumors, sites of inflammation and other diseased<br />
body tissues where the blood vessels are<br />
either disrupted or not fully formed, and as a<br />
result are leakier than normal vessels. The EPR<br />
effect allows nanoparticles to pass readily from<br />
the blood vessels into the tumor or inflamed<br />
tissue and to be differentially retained while the<br />
drug is released from the particles. It has been<br />
demonstrated that, as a result of this effect, therapeutic<br />
nanoparticles accumulate in a particular<br />
target tissue location and deliver more of the<br />
drug to the disease site over a longer period of<br />
time than a conventional drug product administered<br />
as a solution.<br />
Currently marketed therapeutic nanoparticles<br />
use passive targeting; that is, they rely exclu-<br />
Nanoparticle<br />
therapeutic<br />
Platform IP<br />
Process patents<br />
Formulation patents<br />
Manufacturing trade secrets<br />
PK study does not assess<br />
drug at site of action<br />
Figure 1 Market exclusivity: therapeutic nanoparticle versus conventional drug product.<br />
PK, pharmacokinetic.<br />
sively on the EPR effect to enhance delivery of<br />
drugs to the site of disease. Actively targeted<br />
nanoparticles now under development are<br />
surface-functionalized with targeting ligands<br />
that impart even greater specificity by binding<br />
to receptors located on diseased cells or tissues.<br />
The targeting ligand enhances the binding and<br />
retention of the nanoparticle to the target tissue<br />
and can also mediate the uptake or ingestion of<br />
the nanoparticle (and its therapeutic payload)<br />
by a target cell type. Whether a disease site is<br />
actively or passively targeted, therapeutic nanoparticles<br />
can increase the concentration of drugs<br />
in diseased cells or tissues and reduce concentration<br />
where the drugs can cause undesirable<br />
side effects. Moreover, therapeutic nanoparticles<br />
enable the drug developer to modulate the trafficking<br />
of the active pharmaceutical ingredient<br />
within the body without altering its molecular<br />
structure, and hence its inherent pharmacological<br />
activity. The decoupling of drug disposition<br />
from molecular structure affords the possibility<br />
of developing better drugs with fewer tradeoffs<br />
than is possible with conventional drugs, where<br />
both biodistribution and pharmacological<br />
activity depend on molecular structure and<br />
cannot be manipulated independently. This<br />
interdependence often forces undesirable compromises,<br />
or worse, abandonment of otherwise<br />
promising new drug candidates.<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1267
patents<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Table 1 Currently approved nanoparticle or liposomal products on the US market<br />
Trade name Generic name Date approved Patents listed Par IV filing Generic equivalent<br />
Doxil Doxorubicin 1995 No No No<br />
Depocyt Cytarabine 1999 Yes No No<br />
Ambisome Amphotericin B 1997 Yes No No<br />
Abelcet Amphotericin B 1995 No No No<br />
Amphotec Amphotericin B 1996 No No No<br />
DepoDur Morphine 2004 Yes No No<br />
Daunoxome Daunorubicin 1996 No No No<br />
Abraxane Paclitaxel 2005 Yes No No<br />
A successful therapeutic nanoparticle must<br />
be precisely optimized to prevent its removal<br />
from the circulation by the body’s defense systems<br />
before it has a chance to reach its target,<br />
maximize trafficking to the desired location and<br />
minimize accumulation at sites where the drug<br />
may cause side effects. Like conventional pharmaceutical<br />
products, it is necessary to develop<br />
a robust and well-controlled manufacturing<br />
process that produces nanoparticles of uniform<br />
quality from batch to batch; however, nanoparticle<br />
manufacturing has additional complexities<br />
to sort through to uniformly and precisely<br />
control the critical parameters that lead to their<br />
therapeutic benefits.<br />
Therapeutic nanoparticles can encompass<br />
both existing and new active pharmaceutical<br />
ingredients. Indeed, when manufacturers use<br />
validated active pharmaceutical ingredients, the<br />
risk of drug development can be substantially<br />
lower as compared to development of a new<br />
chemical entity. The end product will still be a<br />
substantially differentiated product and retain<br />
market exclusivity as well as or better than a new<br />
chemical entity (Fig. 1). This exclusivity is likely<br />
to extend beyond patent expiration as a result<br />
of additional barriers to generic competition<br />
discussed below that are unique to therapeutic<br />
nanoparticles, thereby potentially reducing or<br />
avoiding the sudden revenue fall-off for successful<br />
proprietary drugs that confronts the<br />
pharmaceutical industry today (Fig. 2).<br />
Patenting of nanoparticles<br />
A therapeutic nanoparticle is a complex, yet<br />
uniformly precise, platform for developing<br />
new drugs. Because of the complexity of these<br />
nanoparticles, opportunities abound for obtaining<br />
meaningful patent protection. For example,<br />
a polymeric nanoparticle can vary by polymer<br />
type, polymer size, drug, mix of polymers, surface<br />
characteristics, targeting ligands, content<br />
of drug encapsulated, concentration of nanoparticles<br />
or manufacturing process. Each one of<br />
those factors can have a considerable effect on<br />
the behavior of the nanoparticle in a biological<br />
system. Polymeric nanoparticle characteristics<br />
critical to its safety and effectiveness, such as its<br />
target tissue accumulation, pharmacokinetics<br />
(clearance and volume of distribution), stability<br />
and drug release kinetics, are dependent on<br />
numerous formulation and process parameters.<br />
The combination of many factors leading to<br />
many outcomes provides numerous grounds<br />
for patentability. Opportunities exist to patent<br />
(i) particular aspects of therapeutic nanoparticles<br />
critical for effective use, (ii) classes of<br />
therapeutic nanoparticle compositions, (iii)<br />
specific therapeutic nanoparticle compositions,<br />
(iv) methods of using therapeutic nanoparticles,<br />
(v) methods of manufacturing therapeutic<br />
nanoparticles, (vi) therapeutic nanoparticles of<br />
particular drugs and (vii) the pharmacokinetic<br />
profile of particular nanoparticles.<br />
As they do for new chemical entities, patenting<br />
opportunities exist in the discovery phase of<br />
identifying new therapeutic nanoparticles. Later<br />
in the product development cycle, additional<br />
opportunities exist to patent clinical-stage uses<br />
and dosing schedules. These patent opportunities<br />
can be combined with existing IP to protect<br />
a new drug entity or can be the basis of a new<br />
patent estate protecting an existing active pharmaceutical<br />
ingredient or product. In addition,<br />
many aspects of developing and manufacturing<br />
nanoparticle products involve extensive knowhow<br />
to optimize the product.<br />
Manufacturing challenges<br />
Therapeutic nanoparticles present many of<br />
the same scientific challenges for generic drug<br />
entry as complex biological drugs. Very small<br />
changes to the manufacturing process of a<br />
therapeutic nanoparticle can result in a different<br />
nanoparticle altogether, both with respect to<br />
its chemical composition and structure as well<br />
as to its ultimate therapeutic effects. Changes<br />
to the manufacturing process of a therapeutic<br />
nanoparticle can affect the drug content of<br />
the particle and the size and chemical makeup<br />
of the polymer component of the particle.<br />
Structural parameters, including the nanoparticle<br />
surface properties and size, and the spatial<br />
arrangement of the polymer, drug and targeting<br />
ligand, are determined by the manner in which<br />
the nanoparticle assembles. Changing any of<br />
these factors may alter critical properties, such<br />
as the distribution of the particles in the body or<br />
the rate at which the drug is released from the<br />
particle, thereby resulting in a different clinical<br />
outcome 3 . Because the analytical tools necessary<br />
to characterize therapeutic nanoparticles<br />
and their specific biological effects are not yet<br />
well established, a generics company may not<br />
be able to demonstrate the degree of sameness<br />
of a nanoparticle product required for FDA or<br />
other regulatory approval. As such, the manufacturer<br />
of the trade or ‘pioneer’ drug could<br />
have great exclusivity advantages by keeping<br />
their final manufacturing process a trade secret<br />
or by obtaining process patents. If a generics<br />
company cannot practice the same manufacturing<br />
process, the generic product probably<br />
does not have the same therapeutic nanoparticle.<br />
Simply put, because the FDA requires that<br />
a generic product be the “same,” then without<br />
the same process, the generics company will<br />
face considerable difficulties producing the<br />
same product.<br />
Regulatory challenges<br />
Stringent regulatory requirements exist for<br />
the entry of generic therapeutic nanoparticles.<br />
Table 1 lists the currently approved nanoparticle<br />
or liposomal products on the US market.<br />
Despite the fact that many of these products<br />
lack patent protection today, and that several<br />
generate worldwide sales of $250–750 million,<br />
none of these products has a generic equivalent<br />
4 . In fact, the FDA has never approved a<br />
parenterally administered generic therapeutic<br />
nanoparticle. No generics company has even<br />
attempted to seek approval for a generic form<br />
of a patent-protected nanoparticle product,<br />
and thus there has not been any patent litigation<br />
resulting from a generics challenge (that<br />
is, a “paragraph IV challenge”). In essence,<br />
the technical and manufacturing complexity,<br />
difficulties in demonstrating bioequivalence<br />
and regulatory challenges result in longer and<br />
more expensive development pathways, which<br />
are incompatible with most generic drug<br />
business models.<br />
The two major regulatory difficulties for the<br />
development of generic therapeutic nanoparticles<br />
are (i) showing bioequivalence and (ii)<br />
FDA requirements for parenteral drug products.<br />
For a generic drug company to avoid<br />
doing full clinical trials of safety and effectiveness,<br />
they must show that, among other things,<br />
the generic product is bioequivalent to the<br />
pioneer product that it references and relies on<br />
for approval. For typical, oral, small-molecule<br />
products, this bioequivalence is easily shown by<br />
dosing healthy volunteers and measuring and<br />
comparing the mean area under the plasma<br />
concentration curve and maximum plasma<br />
1268 volume 28 number 12 DECEMBER 2010 nature biotechnology
patents<br />
© 2010 Nature America, Inc. All rights reserved.<br />
and complex analytical methods. Even with all<br />
that data, the FDA might still require a clinical<br />
trial. The FDA discussed some of these complex<br />
comparison issues in 2001 (ref. 6) and 2007<br />
(ref. 7) and has yet to decide how to address or<br />
solve these issues for therapeutic nanoparticles<br />
as a whole. A recent draft guidance relating specifically<br />
to abbreviated new drug applications<br />
for Doxil, a passively targeted nanoparticle, sets<br />
a very exacting standard 8 . The guidance states<br />
that the generics company should conduct<br />
clinical studies in cancer patients and assess<br />
bioequivalence based on analysis of both free<br />
doxorubicin and liposome-encapsulated doxorubicin.<br />
It further states that the pivotal clinical<br />
study should use test product produced using<br />
the proposed commercial scale manufacturing<br />
process—a far more stringent requirement<br />
than the 1:10 scale conventionally employed<br />
for bioequivalence studies. Moreover, it is recommended<br />
that the proposed generic product<br />
contain the same lipid excipients produced by<br />
the same synthetic route as the reference product,<br />
and that the generic product be manufactured<br />
using the same process as the reference<br />
product. Lastly, the generic product should<br />
be equivalent with respect to a broad array of<br />
physicochemical characteristics, including<br />
liposome composition, physical state of the<br />
encapsulated drug, the liposome internal environment,<br />
morphology, lipid bilayer fluidity,<br />
size distribution, surface chemistry, electrical<br />
potential and charge, and in vitro drug leakage<br />
under a variety of conditions 9 . The FDA will<br />
likely review each therapeutic nanoparticle on<br />
its own and determine the recommended tests<br />
on an individual basis. For therapeutic nanoconcentration<br />
of the drug and then calculating<br />
the 90% confidence interval for the ratio of<br />
those mean responses 5 . If that confidence interval<br />
is within 80–125% of the pioneer product<br />
for those parameters, the generic product is said<br />
to be bioequivalent 5 .<br />
In the case of therapeutic nanoparticles<br />
and other locally acting drugs, the standard<br />
approach to the evaluation of bioequivalence<br />
is not sufficient to show sameness with respect<br />
to rate and extent of absorption of the active<br />
pharmaceutical ingredient at the site of action.<br />
A therapeutic nanoparticle circulates in the<br />
plasma and targets specific tissue locations. It<br />
is this tissue localization that is responsible for<br />
nanoparticles’ enhanced effectiveness and/or<br />
safety. A standard bioequivalence test suitable<br />
for most small-molecule generic products only<br />
assesses plasma drug concentrations in healthy<br />
volunteers, and does not reveal whether a<br />
generic product has the same tissue distribution<br />
and drug concentration in disease sites as a pioneer<br />
product. To directly demonstrate equivalent<br />
drug concentration at the site of action, a<br />
prospective generic company would have to<br />
confront several issues. First, it would need to<br />
test the generic product in individuals with the<br />
target indication because the EPR effect and the<br />
product biodistribution would be different in<br />
healthy individuals. Moreover, for most anticancer<br />
drugs, testing in healthy volunteers is not<br />
feasible because of the drugs’ toxicity. Second,<br />
noninvasive assays for quantifying active pharmaceutical<br />
ingredients or nanoparticles in most<br />
tissues do not exist. No assay currently exists to<br />
measure the drug release that occurs at the tissue<br />
site as opposed to other areas in the body. In<br />
principle, the possibility exists of sampling the<br />
target tissue and analyzing it for the concentration<br />
of nanoparticles and drug. This possibility<br />
is only theoretical because the FDA has never<br />
accepted this type of data for a generic product.<br />
Even if a company did conduct a study testing a<br />
generic product’s localization by conducting an<br />
invasive procedure, there would be a low likelihood<br />
the company would successfully identify<br />
a generic product with comparative localization<br />
on the first attempt. Many tissues simply would<br />
not be available for sampling without an invasive<br />
procedure such as surgery or percutaneous<br />
biopsy, which may create an ethical dilemma.<br />
Assuming that the necessary techniques to<br />
do this type of work could be developed (as<br />
these techniques do not currently exist), the<br />
cost and timing to conduct such testing could<br />
be prohibitive for most generic manufacturers<br />
or create a cost structure for a generic drug that<br />
would make it difficult to set the price below the<br />
proprietary medicine. A company would have<br />
to recruit patients, possibly conduct an invasive<br />
procedure on the patients, and develop new<br />
Cash flow<br />
Profit<br />
Loss<br />
Shorter, less expensive<br />
path to approval<br />
and launch<br />
Traditional drug economics<br />
Nanoparticle drug economics<br />
Longer-term peak sales<br />
Figure 2 Drug development economics: traditional versus therapeutic nanoparticle drugs.<br />
particles that localize to a greater extent than<br />
Doxil, the FDA may also require a showing<br />
of drug and nanoparticle concentration at the<br />
site of action. The combination of these factors<br />
makes for an uncertain, expensive and daunting<br />
development route for generic equivalents<br />
to therapeutic nanoparticles.<br />
A generic therapeutic nanoparticle would<br />
also have to meet the FDA’s stringent requirements<br />
for drug products intended for parenteral<br />
use. The FDA regulations state:<br />
“Generally, a drug product intended for<br />
parenteral use shall contain the same inactive<br />
ingredients and in the same concentration<br />
as the reference listed drug identified by the<br />
applicant under paragraph (a)(3) of this section.<br />
However, an applicant may seek approval<br />
of a drug product that differs from the reference<br />
listed drug in preservative, buffer, or antioxidant<br />
provided that the applicant identifies<br />
and characterizes the differences and provides<br />
information demonstrating that the differences<br />
do not affect the safety or efficacy of the proposed<br />
drug product” 9 .<br />
Essentially, this requirement means that<br />
a generics company seeking to make a drug<br />
product for parenteral use, such as a therapeutic<br />
nanoparticle, must develop the same<br />
formulation, both qualitatively and quantitatively,<br />
as that of the pioneer drug; that is, with<br />
certain exceptions, it must contain precisely<br />
the same ingredients in precisely the same<br />
amounts. The recent guidance from the FDA<br />
on liposomal doxorubicin suggests that even<br />
changes in preservative, buffer or antioxidant<br />
may not be accepted for a generic therapeutic<br />
nanoparticle.<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1269
patents<br />
© 2010 Nature America, Inc. All rights reserved.<br />
This level of sameness contrasts with typical<br />
oral formulations. For generic oral products, a<br />
generics company can substitute a wide variety<br />
of excipients and change their concentrations<br />
as compared to the pioneer product. Generics<br />
companies often change the excipients in a<br />
formulation because the manufacturer of the<br />
pioneer drug has a formulation patent and the<br />
generics company is looking to circumvent this<br />
patent. A generics company’s ability to switch<br />
excipients in oral products often limits the value<br />
of formulation patents for oral products. The<br />
issue for the pioneer drug company is typically<br />
that broad formulation claims can be invalidated<br />
in litigation and narrow formulation<br />
claims are often circumventable.<br />
In contrast, because of the FDA’s requirements<br />
for parenteral drug products, patents<br />
for a therapeutic nanoparticle have a much<br />
greater value than formulation patents for<br />
an oral product. For example, a pioneer drug<br />
company could have a patent claim narrowly<br />
covering a nanoparticle formulation, which<br />
a generics company could not circumvent.<br />
Importantly, however, if this claim covered<br />
an FDA-approved therapeutic nanoparticle, a<br />
generics company would have to develop an<br />
identical product with the exact same polymers<br />
and the exact same drug, all in the exact same<br />
proportions. Based on the FDA guidance on<br />
liposomal doxorubicin, it is now apparent that,<br />
at a minimum, the generic nanoparticle would<br />
also have to match the pioneer product in state<br />
of encapsulated drug, nanoparticle internal<br />
environment, nanoparticle morphology, nanoparticle<br />
size distribution, polymer orientation,<br />
electrical surface potential and drug release 9 .<br />
Regulations for more complex nanoparticles,<br />
such as polymeric nanoparticles, could have<br />
even more rigorous regulatory requirements<br />
including clinical outcome data. An attempt to<br />
circumvent the claim by substituting a different<br />
polymer or slightly reducing the concentration<br />
of a particular polymer would not be allowed<br />
for a generic therapeutic nanoparticle. Thus, the<br />
combination of even a very narrow composition<br />
or process patent claim with the FDA’s general<br />
requirements for parenteral drug products and<br />
the heightened requirements for therapeutic<br />
nanoparticles can result in a significant and<br />
long exclusivity period. Even broader patent<br />
claims could also be obtained for additional<br />
market protection, as described above.<br />
A generics company could attempt to circumvent<br />
the FDA’s requirements for parenteral<br />
drug products by filing a Section 505(b)<br />
(2) new drug application (NDA). Such a filing<br />
would not require the ‘same’ formulation. Of<br />
course by switching formulations, a generics<br />
company would not be able to fully rely on the<br />
pioneer drug company’s previous clinical findings.<br />
At best, a 505(b)(2) NDA would be able<br />
to rely on the toxicology data for the active<br />
pharmaceutical ingredient. A 505(b)(2) applicant<br />
would need to conduct new effectiveness<br />
trials in order to ensure that the new formulation<br />
would deliver the same quantity of drug to<br />
disease sites over the same time course, or produce<br />
the same drug exposure to nondiseased<br />
tissues. Because conducting effectiveness<br />
trials is the most expensive part of product<br />
development, this 505(b)(2) regulatory route<br />
is unlikely to be economically viable for the<br />
generics drug industry. In addition, products<br />
approved through the 505(b)(2) route are not<br />
directly substitutable in the pharmacy for the<br />
branded product. Hence, a 505(b)(2) generic<br />
product would need to be marketed and sold<br />
by a sales force, which most generics companies<br />
do not have.<br />
Conclusions<br />
Therapeutic nanoparticles offer the potential<br />
to dramatically improve the effectiveness and<br />
side-effect profile of new and existing drugs. In<br />
addition, this new drug class presents equivalent<br />
or possibly greater challenges for generic drug<br />
entry than other types of pharmaceutical products,<br />
including biologics. The combination of scientific,<br />
patent, know-how and regulatory<br />
issues makes the development of a generic<br />
therapeutic nanoparticle very difficult and<br />
inconsistent with the generics business model.<br />
Pioneer drug companies have the opportunity<br />
to obtain strong patent protection, which can<br />
maintain market exclusivity for the full patent<br />
term and potentially beyond patent expiration<br />
due to the scientific, manufacturing and regulatory<br />
hurdles confronting the development of<br />
generic therapeutic nanoparticles. If successful,<br />
this opportunity will fundamentally change the<br />
pharmaceutical business model by reducing or<br />
avoiding the sudden revenue fall-off for successful<br />
proprietary drugs that confronts the<br />
industry today (Fig. 2).<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare competing financial interests:<br />
details accompany the full-text HTML version of the<br />
paper at http://www.nature.com/naturebiotechnology/.<br />
1. Doxil Product Information. Ortho Biotech Products, LP<br />
(2008).<br />
2. Farokhzad O. & Langer, R. ACS Perspective 3, 16–20<br />
(2009).<br />
3. Davis, M. Nanotechnol. Law & Bus. 255–262<br />
(September 2006).<br />
4. US Food and Drug Administration. Orange Book:<br />
Approved Drug Products with Therapeutic Equivalence<br />
Evaluations <br />
5. US Food and Drug Administration. Guidance<br />
for Industry: Bioavailability and Bioequivalence<br />
Studies for Orally Administered Drug Products<br />
(March 2003).<br />
6. US Food and Drug Administration. Advisory Committee<br />
For Pharmaceutical Science Proceedings <br />
(July 20, 2001).<br />
7. US Food and Drug Administration. Critical Path<br />
Opportunities for Generic Drugs <br />
(May 1, 2007).<br />
8. US Food and Drug Administration. Draft guidance on doxorubicin<br />
hydrochloride (February 2010).<br />
9. 21 CFR 314.94(a)(iii).<br />
1270 volume 28 number 12 DECEMBER 2010 nature biotechnology
patents<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Recent patent applications in pharmacogenomics<br />
Patent number Description Assignee Inventor<br />
WO 2010115154,<br />
US 20100273219<br />
US 20080038810,<br />
US 7816121<br />
WO 2010115044<br />
US 20100210025<br />
WO 2008144345,<br />
EP 2145021<br />
US 20100178655,<br />
WO 2010083250<br />
WO 2010075570<br />
WO 2010048497<br />
KR 2010011719<br />
JP 4403376<br />
US 20090117583<br />
A method of amplifying target nucleic acids in<br />
samples involving preparing an amplification<br />
mixture of forward and reverse primers comprising<br />
the target-specific portion and a barcode<br />
primer, comprising nucleotides, and subjecting<br />
the mixture to amplification.<br />
A droplet microactuator comprising a substrate<br />
with electrodes for conducting droplet operations<br />
and temperature control elements arranged near<br />
the electrodes to transport nucleic acid droplets on<br />
electrodes; useful for nucleic acid amplification.<br />
A method for depleting a nucleic acid sample<br />
of nontarget nucleic acids involving denaturing<br />
the sample nucleic acids in a reaction mixture,<br />
contacting the denatured nucleic acids with at<br />
least one target-specific primer pair under suitable<br />
annealing conditions, conducting a first cycle of<br />
extension of any annealed target-specific primer<br />
pairs and conducting nuclease digestion of singlestranded<br />
nucleic acid sequences in the reaction<br />
mixture.<br />
A system comprising assigning modules to a<br />
genome, assigning a value or weight to a module<br />
for a given profile, analyzing a genomic sequence<br />
to identify modules and assigning a profile to the<br />
sequence; useful, e.g., for profiling a genomic<br />
sequence.<br />
A method of predicting the likelihood of a<br />
therapeutic response to insulin growth factor-1<br />
receptor (IGF1R) modulator as a cancer treatment,<br />
comprising measuring levels of biomarker genes or<br />
proteins before and after exposing sample to IGF1R<br />
modulator.<br />
Genotyping a single cell, comprising preamplifying<br />
an amplified genome to produce a<br />
pre-amplification reaction mixture comprising<br />
amplicons specific for target nucleic acids, and<br />
amplifying and detecting the amplicons.<br />
A computer-accessible medium used for<br />
assembling at least one haplotype sequence or<br />
genotype sequence of at least one genome, with<br />
stored computer-executable instructions.<br />
A method of diagnosing predisposition to, or<br />
progression of, Alzheimer’s disease, comprising<br />
determining a genetic profile comprising at least<br />
one marker in a specified candidate region and<br />
correlating the genetic profile with a reference<br />
profile.<br />
A single-nucleotide polymorphism provided for<br />
calretinin gene, retinoic acid receptor beta gene<br />
and pre-B-cell leukemia transcription factor 1<br />
gene; for use in biochips for predicting the risk<br />
of obesity.<br />
A unit for detecting interaction between<br />
substances (e.g., hybridization) that has a<br />
reaction region, water-absorptive gel in the<br />
reaction region, an electrode and water-holding<br />
portion; useful in a substrate for bioassays for<br />
analyzing mutations of genes.<br />
A new epidermal growth factor–like variant in<br />
skin (ELVIS)-2 polypeptide; useful for screening<br />
assays, detection assays, predictive medicine and<br />
methods of treatment.<br />
Fluidigm<br />
(S. San Francisco, CA,<br />
USA)<br />
Anderson M, Chen P,<br />
Kaper F, May A, Wang J<br />
Paik PY, Pamula VK, Paik PY, Pamula VK,<br />
Pollack MG,<br />
Pollack MG<br />
Duke University<br />
(Durham, NC, USA),<br />
Advanced Liquid Logic<br />
(Research Triangle Park,<br />
NC, USA)<br />
Fluidigm (S. San<br />
Francisco, CA, USA)<br />
Priority<br />
application date<br />
Publication date<br />
4/2/2009 10/7/2010,<br />
10/28/2010<br />
4/18/2006 2/14/2008,<br />
10/19/2010<br />
Zimmermann BG 4/2/2009 10/7/2010<br />
Victor Chang Cardiac George R, Wouters M 8/15/2006 8/19/2010<br />
Research Institute<br />
(Darlinghurst, Australia)<br />
Bristol-Myers Squibb<br />
(Princeton, NJ, USA)<br />
Fluidigm<br />
(S. San Francisco, CA,<br />
USA)<br />
New York University<br />
(New York)<br />
Genizon Biosciences<br />
(St. Laurent,<br />
Quebec, Canada),<br />
TechnoSynapse<br />
(Blainville, Quebec,<br />
Canada)<br />
Attar RM, Carboni JM,<br />
Chen J, Dongre AR,<br />
Gottardis MM,<br />
Hafezi R, Han X,<br />
Huang F, Hurlburt W,<br />
Robinson DM,<br />
Wittenberg GM<br />
Hamilton A, Lin M,<br />
Mir A, Pieprzyk M<br />
5/17/2007 11/27/2008,<br />
7/22/2010<br />
1/13/2009 7/15/2010,<br />
7/22/2010<br />
Mishra B, Narzisi G 12/24/2008 7/1/2010<br />
Belouchi A, Croteau P,<br />
Debrus S, Kebache S,<br />
Keith T, Little RD,<br />
Paquin B, Poirier J,<br />
Raelson JV, Segal J,<br />
van Eerdewegh P<br />
Jeong J Gang S, Jeong J,<br />
Jeong M, Kim S<br />
Sony (Tokyo) Mamine T, Sakamoto Y,<br />
Segawa Y, Yuo K<br />
Busfield SJ,<br />
Gearing DP,<br />
Millennium<br />
Pharmaceuticals<br />
(Cambridge, MA, USA)<br />
Busfield SJ,<br />
Gearing DP<br />
10/24/2008 4/29/2010<br />
7/25/2008 2/3/2010<br />
10/9/2003 1/27/2010<br />
11/19/1998 5/7/2009<br />
Source: Thomson Scientific Search Service. The status of each application is slightly different from country to country. For further details, contact Thomson Scientific,<br />
1800 Diagonal Road, Suite 250, Alexandria, Virginia 22314, USA. Tel: 1 (800) 337-9368 (http://www.thomson.com/scientific).<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1271
news and views<br />
Megabases for kilodollars<br />
Mikkel Algire, Radha Krishnakumar & Chuck Merryman<br />
Two approaches for selecting oligonucleotides from complex mixtures improve the fidelity and scalability of<br />
DNA synthesis.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
The potential applications of synthetic DNA<br />
are virtually unlimited, ranging from the<br />
design of new genetic circuits and biosynthetic<br />
pathways to the creation of complete<br />
bacterial ge nomes 1–3 . Two reports in this issue<br />
from Church and colleagues bring us a step<br />
closer to an era of abundant synthetic DNA by<br />
addressing the hurdles of high cost and error<br />
rates. Kosuri et al. 4 and Matzas et al. 5 describe<br />
approaches that allow relatively inexpensive<br />
DNA oligo nucleotides (oligos), synthesized<br />
on a microarray, to be used to assemble long<br />
DNA molecules encoding genes and other<br />
genomic elements (Fig. 1).<br />
Currently, the price of long synthetic DNA<br />
is dominated by the price of the initial oligos,<br />
~$0.10 per nucleotide. The second most<br />
costly aspect of constructing synthetic DNA is<br />
the identification of assembled products that<br />
have the correct sequence, a step that requires<br />
sequencing. Sequencing costs vary tremendously<br />
depending on the degree of automation<br />
and other factors, but under the best of<br />
circumstances, they are 10–20% of oligo costs.<br />
The methods of Kosuri et al. 4 and Matzas et al. 5<br />
produce synthetic DNA at a cost of
news and views<br />
© 2010 Nature America, Inc. All rights reserved.<br />
considerably. Presumably, assembly is more<br />
robust because longer tracts of DNA are left<br />
after the amplification primers are removed.<br />
Thus, fewer starting fragments are needed to<br />
assemble each product. The method of Kosuri<br />
et al. 4 is still subject to the error rate of input<br />
oligos, although errors are reduced by using<br />
new, more accurate microarrays and by performing<br />
an enzymatic error correction step.<br />
The article from Matzas et al. 5 directly<br />
addresses the problem of errors in oligonucleotide<br />
sequences and the labor needed to deal<br />
with them. They use a 454 sequencing instrument<br />
to identify microarray oligos with the<br />
correct sequence and then recover these oligos<br />
from the sequencer. In essence, this provides<br />
cheap oligos preverified by cheap sequencing.<br />
454-beads bearing correct sequences are<br />
located with microscope cameras and picked<br />
using a micropipette. The oligos are subsequently<br />
assembled into larger products. Because<br />
the DNA assembly step can be virtually error<br />
free (even during the construction of whole<br />
genomes 1,2 ), presequenced and clonal oligos<br />
have the potential to produce near-perfect<br />
assemblies. Even though Matzas et al. 5 find an<br />
error in one of the eight ~220-bp constructs<br />
made from their microarray, this error rate<br />
is likely to decrease upon further optimization.<br />
Although the starting oligos, polymerase<br />
fidelity and accuracy of 454 sequencing all<br />
contribute errors, the experiments indicate<br />
that major gains are to be had by improving<br />
bead localization and retrieval. Automated<br />
bead extraction is under development, and the<br />
authors hope to achieve recovery rates of two<br />
or three beads per minute. The error rate of<br />
the current system is estimated to reach about<br />
1 in 21,000 bp.<br />
These papers demonstrate that the use of<br />
inexpensive microarray oligonucleotides to<br />
produce long synthetic DNA is becoming more<br />
practical. Assembly technology is also undergoing<br />
rapid development; for example, oligos<br />
produced by any method can be assembled<br />
into larger sequences by a fast and essentially<br />
labor-free method 3 . At the same time, projects<br />
involving intensive DNA synthesis are<br />
gene rating advances such as empirically determined<br />
codon-optimization rules 8 , new insights<br />
into host metabolism 9 and the first microbe<br />
controlled by a synthetic genome 2 .<br />
It is reasonable to wonder about how to<br />
make effective use of megabases of userdefined<br />
sequence. But with low-cost oligos<br />
and a simple means of assembling them, we<br />
can safely predict that our ability to understand<br />
and manipulate biology will only continue<br />
to improve, perhaps leading eventually<br />
to advances such as microbes engineered to be<br />
attenuated vaccines or to synthesize biofuels.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
1. Gibson, D.G. et al. Science 319, 1215–1220 (2008).<br />
2. Gibson, D.G. et al. Science 329, 52–56 (2010).<br />
3. Gibson, D.G. et al. Nat. Methods 7, 901–903 (2010).<br />
4. Kosuri, S. et al. Nat. Biotechnol. 28, 1295–1299 (2010).<br />
No refuge for insect pests<br />
Kongming Wu<br />
Transgenic cotton and corn expressing insecticidal<br />
proteins from the bacterium Bacillus<br />
thuringiensis (Bt) have been cultivated on >200<br />
million ha worldwide over the past 15 years 1 ,<br />
reducing the use of chemical insecticides and<br />
increasing farmers’ profits 2,3 . But the environmental<br />
and economic advantages of these crops<br />
are threatened by the tendency of insect pests<br />
to develop resistance to Bt toxins. A report in<br />
this issue by Tabashnik et al. 4 demonstrates a<br />
new approach for suppressing the emergence<br />
of resistance. Field trials in Arizona over four<br />
growing seasons show that releasing sterile<br />
male pink bollworm moths, which can mate<br />
with resistant females, succeeded in almost<br />
completely eradicating a major cotton pest. The<br />
study establishes that sterile insect release is a<br />
viable alternative to the so-called refuge strategy<br />
for preventing the emergence of resistance and<br />
can enhance the sustainability of Bt crops.<br />
Continuous monoculture of crop varieties<br />
producing Bt toxins provides strong selective<br />
pressure for Bt-resistant insect pests (Fig. 1a).<br />
Indeed, although Bt crops remain effective<br />
against most targeted insect populations, several<br />
pests have evolved resistance 5 . The most<br />
promising strategy for delaying the emergence<br />
of resistance involves planting Bt crops in close<br />
proximity to ‘refuges’, which contain plants that<br />
do not express Bt toxin. Such refuges maintain<br />
populations of insects susceptible to Bt toxin,<br />
which are likely to mate with the rare, resistant<br />
insects. If the mode of inheritance of resistance<br />
is recessive, Bt plants will kill the hybrid progenies<br />
produced by such mating and thereby<br />
delay the evolution of Bt resistance 6 .<br />
Certain governments limit the proportion<br />
of crop that each farmer can plant to Bt<br />
Kongming Wu is at the State Key Laboratory<br />
for Biology of Plant Diseases and Insect Pests,<br />
Institute of Plant Protection, Chinese Academy<br />
of Agricultural Sciences, Beijing, China.<br />
e-mail: kmwu@ippcaas.cn<br />
5. Matzas, M. et al. Nat. Biotechnol. 28, 1291–1294 (2010).<br />
6. LeProust, E.M. et al. Nucleic Acids Res. 38,<br />
2522–2540 (2010).<br />
7. Xu, Q. et al. Proc. Natl. Acad. Sci. USA 106,<br />
2289–2294 (2009).<br />
8. Welch, M. et al. PLoS ONE 4, e7002 (2009).<br />
9. Warner, J. et al. Nat. Biotechnol. 28, 856–862 (2010).<br />
The sterile insect technique offers an alternative to the refuge strategy for<br />
managing resistance to Bt toxins.<br />
crops and require a minimum proportion of<br />
non-Bt crop to serve as a refuge. For example,<br />
the United States and Australia mandate this<br />
approach for Bt cotton that produces only one<br />
Bt toxin 5 . Resistance monitoring data from<br />
these countries have suggested that the refuge<br />
strategy can significantly delay the evolution of<br />
insect resistance to Bt crops 7 . It is also beneficial<br />
to non-Bt crops in the refuges, which are<br />
exposed to fewer pests owing to the control<br />
exerted by neighboring Bt crops 2 . Moreover,<br />
as nontransgenic seed is cheaper, there is an<br />
economic incentive for farmers to adopt the<br />
refuge strategy.<br />
Despite the benefits of the refuge strategy, it<br />
has proved difficult to implement in developing<br />
countries such as China and India, primarily<br />
because of the challenges associated with training<br />
and monitoring millions of smallholder farmers.<br />
During 2009, China planted 3.7 million ha<br />
(~70% of total cotton production) and India<br />
planted 8.4 million ha (~90% of total cotton<br />
production) to Bt cotton. In certain cases, however,<br />
it has been put in place unintentionally. In<br />
China, the primary pest targeted by Bt cotton<br />
is Helicoverpa armigera, which also feeds on<br />
other crops, including corn, soybean, vegetables<br />
and peanuts. These nontransgenic host plants<br />
are frequently planted near Bt cotton and serve<br />
as refuges for H. armigera. Thus far, continuous<br />
resistance monitoring has not detected any resistance<br />
of H. armigera populations to Bt toxin 7 .<br />
In contrast to the situation with H. armigera<br />
in China, reliance on noncotton host plants as<br />
refuges is not an option for controlling pink<br />
bollworm (Pectinophora gossypiella), which<br />
feeds almost exclusively on cotton in some<br />
regions 4 . This pest has evolved resistance<br />
to Bt cotton producing one toxin in western<br />
India, where farmers did not follow regulations<br />
requiring them to plant non-Bt cotton refuges 8 .<br />
In Arizona, where pink bollworm is an alien<br />
species first detected about a century ago, the<br />
efficacy of Bt cotton has been sustained for<br />
nature biotechnology volume 28 number 12 december 2010 1273
news and views<br />
© 2010 Nature America, Inc. All rights reserved.<br />
considerably. Presumably, assembly is more<br />
robust because longer tracts of DNA are left<br />
after the amplification primers are removed.<br />
Thus, fewer starting fragments are needed to<br />
assemble each product. The method of Kosuri<br />
et al. 4 is still subject to the error rate of input<br />
oligos, although errors are reduced by using<br />
new, more accurate microarrays and by performing<br />
an enzymatic error correction step.<br />
The article from Matzas et al. 5 directly<br />
addresses the problem of errors in oligonucleotide<br />
sequences and the labor needed to deal<br />
with them. They use a 454 sequencing instrument<br />
to identify microarray oligos with the<br />
correct sequence and then recover these oligos<br />
from the sequencer. In essence, this provides<br />
cheap oligos preverified by cheap sequencing.<br />
454-beads bearing correct sequences are<br />
located with microscope cameras and picked<br />
using a micropipette. The oligos are subsequently<br />
assembled into larger products. Because<br />
the DNA assembly step can be virtually error<br />
free (even during the construction of whole<br />
genomes 1,2 ), presequenced and clonal oligos<br />
have the potential to produce near-perfect<br />
assemblies. Even though Matzas et al. 5 find an<br />
error in one of the eight ~220-bp constructs<br />
made from their microarray, this error rate<br />
is likely to decrease upon further optimization.<br />
Although the starting oligos, polymerase<br />
fidelity and accuracy of 454 sequencing all<br />
contribute errors, the experiments indicate<br />
that major gains are to be had by improving<br />
bead localization and retrieval. Automated<br />
bead extraction is under development, and the<br />
authors hope to achieve recovery rates of two<br />
or three beads per minute. The error rate of<br />
the current system is estimated to reach about<br />
1 in 21,000 bp.<br />
These papers demonstrate that the use of<br />
inexpensive microarray oligonucleotides to<br />
produce long synthetic DNA is becoming more<br />
practical. Assembly technology is also undergoing<br />
rapid development; for example, oligos<br />
produced by any method can be assembled<br />
into larger sequences by a fast and essentially<br />
labor-free method 3 . At the same time, projects<br />
involving intensive DNA synthesis are<br />
gene rating advances such as empirically determined<br />
codon-optimization rules 8 , new insights<br />
into host metabolism 9 and the first microbe<br />
controlled by a synthetic genome 2 .<br />
It is reasonable to wonder about how to<br />
make effective use of megabases of userdefined<br />
sequence. But with low-cost oligos<br />
and a simple means of assembling them, we<br />
can safely predict that our ability to understand<br />
and manipulate biology will only continue<br />
to improve, perhaps leading eventually<br />
to advances such as microbes engineered to be<br />
attenuated vaccines or to synthesize biofuels.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
1. Gibson, D.G. et al. Science 319, 1215–1220 (2008).<br />
2. Gibson, D.G. et al. Science 329, 52–56 (2010).<br />
3. Gibson, D.G. et al. Nat. Methods 7, 901–903 (2010).<br />
4. Kosuri, S. et al. Nat. Biotechnol. 28, 1295–1299 (2010).<br />
No refuge for insect pests<br />
Kongming Wu<br />
Transgenic cotton and corn expressing insecticidal<br />
proteins from the bacterium Bacillus<br />
thuringiensis (Bt) have been cultivated on >200<br />
million ha worldwide over the past 15 years 1 ,<br />
reducing the use of chemical insecticides and<br />
increasing farmers’ profits 2,3 . But the environmental<br />
and economic advantages of these crops<br />
are threatened by the tendency of insect pests<br />
to develop resistance to Bt toxins. A report in<br />
this issue by Tabashnik et al. 4 demonstrates a<br />
new approach for suppressing the emergence<br />
of resistance. Field trials in Arizona over four<br />
growing seasons show that releasing sterile<br />
male pink bollworm moths, which can mate<br />
with resistant females, succeeded in almost<br />
completely eradicating a major cotton pest. The<br />
study establishes that sterile insect release is a<br />
viable alternative to the so-called refuge strategy<br />
for preventing the emergence of resistance and<br />
can enhance the sustainability of Bt crops.<br />
Continuous monoculture of crop varieties<br />
producing Bt toxins provides strong selective<br />
pressure for Bt-resistant insect pests (Fig. 1a).<br />
Indeed, although Bt crops remain effective<br />
against most targeted insect populations, several<br />
pests have evolved resistance 5 . The most<br />
promising strategy for delaying the emergence<br />
of resistance involves planting Bt crops in close<br />
proximity to ‘refuges’, which contain plants that<br />
do not express Bt toxin. Such refuges maintain<br />
populations of insects susceptible to Bt toxin,<br />
which are likely to mate with the rare, resistant<br />
insects. If the mode of inheritance of resistance<br />
is recessive, Bt plants will kill the hybrid progenies<br />
produced by such mating and thereby<br />
delay the evolution of Bt resistance 6 .<br />
Certain governments limit the proportion<br />
of crop that each farmer can plant to Bt<br />
Kongming Wu is at the State Key Laboratory<br />
for Biology of Plant Diseases and Insect Pests,<br />
Institute of Plant Protection, Chinese Academy<br />
of Agricultural Sciences, Beijing, China.<br />
e-mail: kmwu@ippcaas.cn<br />
5. Matzas, M. et al. Nat. Biotechnol. 28, 1291–1294 (2010).<br />
6. LeProust, E.M. et al. Nucleic Acids Res. 38,<br />
2522–2540 (2010).<br />
7. Xu, Q. et al. Proc. Natl. Acad. Sci. USA 106,<br />
2289–2294 (2009).<br />
8. Welch, M. et al. PLoS ONE 4, e7002 (2009).<br />
9. Warner, J. et al. Nat. Biotechnol. 28, 856–862 (2010).<br />
The sterile insect technique offers an alternative to the refuge strategy for<br />
managing resistance to Bt toxins.<br />
crops and require a minimum proportion of<br />
non-Bt crop to serve as a refuge. For example,<br />
the United States and Australia mandate this<br />
approach for Bt cotton that produces only one<br />
Bt toxin 5 . Resistance monitoring data from<br />
these countries have suggested that the refuge<br />
strategy can significantly delay the evolution of<br />
insect resistance to Bt crops 7 . It is also beneficial<br />
to non-Bt crops in the refuges, which are<br />
exposed to fewer pests owing to the control<br />
exerted by neighboring Bt crops 2 . Moreover,<br />
as nontransgenic seed is cheaper, there is an<br />
economic incentive for farmers to adopt the<br />
refuge strategy.<br />
Despite the benefits of the refuge strategy, it<br />
has proved difficult to implement in developing<br />
countries such as China and India, primarily<br />
because of the challenges associated with training<br />
and monitoring millions of smallholder farmers.<br />
During 2009, China planted 3.7 million ha<br />
(~70% of total cotton production) and India<br />
planted 8.4 million ha (~90% of total cotton<br />
production) to Bt cotton. In certain cases, however,<br />
it has been put in place unintentionally. In<br />
China, the primary pest targeted by Bt cotton<br />
is Helicoverpa armigera, which also feeds on<br />
other crops, including corn, soybean, vegetables<br />
and peanuts. These nontransgenic host plants<br />
are frequently planted near Bt cotton and serve<br />
as refuges for H. armigera. Thus far, continuous<br />
resistance monitoring has not detected any resistance<br />
of H. armigera populations to Bt toxin 7 .<br />
In contrast to the situation with H. armigera<br />
in China, reliance on noncotton host plants as<br />
refuges is not an option for controlling pink<br />
bollworm (Pectinophora gossypiella), which<br />
feeds almost exclusively on cotton in some<br />
regions 4 . This pest has evolved resistance<br />
to Bt cotton producing one toxin in western<br />
India, where farmers did not follow regulations<br />
requiring them to plant non-Bt cotton refuges 8 .<br />
In Arizona, where pink bollworm is an alien<br />
species first detected about a century ago, the<br />
efficacy of Bt cotton has been sustained for<br />
nature biotechnology volume 28 number 12 december 2010 1273
news and views<br />
© 2010 Nature America, Inc. All rights reserved.<br />
more than a decade. From 1996 to 2005, farmers<br />
who grew Bt cotton also planted non-Bt cotton<br />
in compliance with the refuge strategy, and the<br />
susceptibility of pink bollworm to Bt did not<br />
decrease 4 . A disadvantage of this approach,<br />
however, is that populations of the insect pest<br />
must be maintained.<br />
Tabashnik et al. 4 report that superimposing<br />
the sterile insect technique (SIT) on Bt cotton<br />
plants offers a compelling alternative to the<br />
refuge strategy while helping to eradicate<br />
pink bollworm. The authors’ program, which<br />
spanned ~100,000 ha in Arizona from 2006 to<br />
2009, showed that susceptibility to Bt cotton<br />
did not decrease, pink bollworm population<br />
density declined dramatically and insecticide<br />
sprays against this pest were eliminated. The<br />
SIT, a method of pest control using area-wide<br />
inundative releases of sterile insects to reduce<br />
fertility of a field population of the same species,<br />
was first developed in the United States<br />
more than 50 years ago 9 . The repeated release<br />
of sterile insects into the environment in numbers<br />
10- to 100-fold in excess of the size of the<br />
native population can eventually drive the<br />
native population to extinction if the majority<br />
of native female insects mate with sterile males.<br />
Although the SIT has been successful in some<br />
cases, such as the screwworm fly (Cochliomyia<br />
hominivorax) eradication program, its application<br />
has been limited 9 . One challenge has<br />
been the cost of generating a sufficiently high<br />
ratio of sterile to wild insects at the start of a<br />
SIT program (Fig. 1b). Using another control<br />
method to lower the pest’s population density<br />
can make it easier to attain a suitable ratio, but<br />
intensive use of conventional insecticides raises<br />
concerns about harm to nontarget organisms,<br />
including people.<br />
The report of Tabashnik et al. 4 reveals how<br />
the SIT and Bt transgenic technology can be<br />
used in a complementary manner (Fig. 1c).<br />
On the one hand, expression of the Bt toxin<br />
suppresses the size of the native insect population<br />
enough to jump-start the efficacy of the<br />
SIT while increasing its economic feasibility.<br />
On the other hand, when female moths that<br />
have evolved resistance to the Bt toxin mate<br />
with sterile insects, the failure to produce<br />
fertile progeny delays the establishment of<br />
resistance based on dominant inheritance. More<br />
importantly, the combination of the SIT and<br />
Bt crops allows farmers to reduce or eliminate<br />
planting of non-Bt refuges. Meanwhile, because<br />
mating with sterile insects does not produce<br />
fertile progeny, the approach could delay pest<br />
resistance to Bt crops that is based on either<br />
dominant or recessive inheritance (Fig. 1c).<br />
The stunning success of the Arizona<br />
program can be attributed to several favorable<br />
factors, including long-term public<br />
a<br />
Bt cotton<br />
Low level of damage with high risk of<br />
resistance in the absence of refuges<br />
b<br />
c<br />
Non-Bt cotton<br />
Severe damage from<br />
wild insects<br />
Bt cotton<br />
Low level of damage with high risk of<br />
resistance in the absence of refuges<br />
investment in SIT against pink bollworm. Since<br />
1967, sterile pink bollworm moths have been<br />
released over cotton fields in the San Joaquin<br />
Valley of California to prevent establishment of<br />
this pest by moth immigration from southern<br />
California 9 . The Arizona program also benefits<br />
from the extremely high efficacy of Bt cotton<br />
against pink bollworm. Results from Arizona<br />
show that Bt cotton producing either one or<br />
two Bt toxins kills virtually 100% of susceptible<br />
pink bollworm larvae 4 . Coordinated contributions<br />
from cotton farmers, the US Department<br />
Emergence of<br />
resistance alleles in the<br />
absence of refuges<br />
Release of<br />
sterile moths<br />
Release of<br />
sterile moths<br />
Severe change from<br />
resistant insects<br />
Low level of damage with costly<br />
release of large numbers of sterile moths<br />
Extremely low level of damage with<br />
reduced chance of resistance and less<br />
extensive release of sterile insects<br />
Figure 1 Use of the SIT together with transgenic cotton expressing the Bt transgene suppresses the<br />
growth of the pink bollworm population and facilitates management of resistance to Bt toxin. Pink<br />
bollworm feeds only on cotton bolls and does not damage other tissues. (a) Sustainable use of Bt cotton<br />
to control pink bollworm populations is threatened by the emergence of resistance. (b) Although costly,<br />
repeated release of sterile pink bollworm moths (red) in vast excess to the number of wild moths (brown)<br />
can suppress the growth of pink bollworm populations. (c) Combined use of Bt cotton and SIT ensures<br />
that the release of fewer sterile moths can suppress the growth of pink bollworm populations while<br />
preventing the emergence of resistance to Bt toxin.<br />
of Agriculture and university scientists have<br />
also been critical.<br />
Despite the impressive results reported by<br />
Tabashnik et al. 4 and the potential of this combinatorial<br />
strategy for many situations, several<br />
alternatives to conventional Bt crops also show<br />
considerable promise. For instance, new insectresistant<br />
transgenic crops may decrease native<br />
insect populations more efficiently than current<br />
Bt crops. They include proteins modified to<br />
counteract resistance development or the expression<br />
of two or more distinct toxins targeting<br />
Katie Vicari<br />
1274 volume 28 number 12 december 2010 nature biotechnology
news and views<br />
© 2010 Nature America, Inc. All rights reserved.<br />
the same pest species 10 . In another approach,<br />
the US Environmental Protection Agency has<br />
approved sales of mixtures of corn seeds with<br />
and without Bt toxins that kill corn rootworms.<br />
This ensures that farmers comply with the<br />
refuge strategy. However, its use is limited to<br />
the pest species whose larvae do not migrate<br />
between plants. Finally, there are also transgenic<br />
strategies to improve SIT efficiency by<br />
expressing dominant lethal genes, promoting<br />
refractoriness of the sterile insects to disease or<br />
facilitating strain marking, genetic sexing and<br />
molecular sterilization 11 . There thus seems<br />
considerable scope to further enhance the sustainability<br />
of Bt crops and develop new strategies<br />
for integrated pest management.<br />
Nanoparticles in the lung<br />
Wolfgang G Kreyling, Stephanie Hirn & Carsten Schleh<br />
Little is known about the fate of nanoparticles<br />
that enter the lungs, either deliberately through<br />
medical treatments 1 or incidentally through<br />
air pollution 2,3 and occupational exposure<br />
in the workplace. As inhaled nanoparticles<br />
can cross the air-blood barrier into the circulation<br />
and accumulate in secondary organs<br />
and tissues, a biokinetic analysis would be the<br />
first step in a dose-response study as part of<br />
a comprehensive risk assessment (Fig. 1). In<br />
this issue, Frangioni, Tsuda and colleagues 4<br />
undertake such a biokinetic analysis in rats,<br />
using near-infrared imaging to follow the<br />
fate of intratracheally instilled nanoparticles<br />
that are varied systematically in size, surface<br />
modification and core composition. The<br />
authors study the parameters that control<br />
whether nanoparticles remain in the lung,<br />
are transported to regional lymph nodes and<br />
the bloodstream or are cleared from the body<br />
through the kidneys. Their findings should<br />
benefit several areas of research, including<br />
the design of nanoparticles for drug delivery<br />
and the evaluation of the toxicity of particulate<br />
air pollution.<br />
Wolfgang G. Kreyling, Stephanie Hirn and<br />
Carsten Schleh are at the Comprehensive<br />
Pneumology Center, Institute of Lung Biology<br />
and Disease, and Focus Network Nanoparticles<br />
and Health, Helmholtz Zentrum München,<br />
German Research Center for Environmental<br />
Health, Neuherberg/Munich, Germany.<br />
e-mail: kreyling@helmholtz-muenchen.de<br />
COMPETING FINANCIAL INTERESTS<br />
The author declares no competing financial interests.<br />
1. James, C. Global Status of Commercialized Biotech/<br />
GM Crops: 2009 (ISAAA 41, Ithaca, New York,<br />
USA, 2009).<br />
2. Hutchison, W. et al. Science 330, 222–225 (2010).<br />
3. Wu, K. et al. Science 321, 1676–1678 (2008).<br />
4. Tabashnik, B. et al. Nat. Biotechnol. 28, 1304–1307<br />
(2010).<br />
5. Tabashnik, B.E. et al. J. Econ. Entomol. 102, 2011–<br />
2025 (2009).<br />
6. Gould, F. Annu. Rev. Entomol. 43, 701–726 (1998).<br />
7. Liu, C. et al. Sci. China Life Sci. 53, 934–941<br />
(2010).<br />
8. Bagla, P. Science 327, 1439 (2010).<br />
9. Klassen, W. & Curtis, C. in Sterile Insect Technique<br />
(eds. Dyck, V., Hendrichs, J. & Robinson, A.) 3–36<br />
(Springer, The Netherlands, 2005).<br />
10. Tabashnik, B. Science 330, 189–190 (2010).<br />
11. Alphey, N., Bonsall, M.B. & Alphey, L. J. Econ. Entomol.<br />
102, 717–732 (2009).<br />
An imaging study begins to define the parameters that control the<br />
biodistribution of nanoparticles after pulmonary delivery.<br />
The lungs represent a promising route for<br />
drug delivery because of their accessible,<br />
large surface area and because the air-blood<br />
barrier is rather thin 1 . After crossing the airblood<br />
barrier, inhaled nanoparticles have been<br />
shown to reach detectable levels in the blood<br />
circulation and to enter specific cells and<br />
tissues 5 (Fig. 1). Translocation and accumulation<br />
in tissues seem to depend on physicochemical<br />
properties of the nanoparticle core<br />
and surface 5–7 . Thus far, however, few studies<br />
have described nanoparticle transport across<br />
the lungs quantitatively 5,6 .<br />
Frangioni, Tsuda and colleagues 4 find that<br />
noncationic nanoparticles smaller than ~34 nm<br />
in diameter that do not bind serum proteins<br />
reach the regional lymph nodes within 30 min.<br />
Nanoparticles larger than ~34 nm are consistently<br />
retained in the lungs. When the diameter<br />
falls below ~6 nm and the charge is zwitterionic,<br />
about half of the nanoparticles rapidly<br />
enter the bloodstream from the alveolar airspaces<br />
and are mostly cleared from the body by<br />
means of renal filtration. The authors also show<br />
that nanoparticle behavior depends strongly on<br />
the surface coating, which affects binding to<br />
proteins in body fluid.<br />
Biodegradable nanoparticles are currently<br />
considered to be the best choice for targeted<br />
drug delivery, but biopersistent nanoparticles<br />
might also be used if they are cleared from the<br />
body quickly enough to minimize exposure<br />
and retention 8 . As demonstrated previously<br />
by the Frangioni laboratory, intravenously<br />
administered nanoparticles can be rapidly<br />
excreted through the kidneys if their physicochemical<br />
properties are carefully chosen 9,10 .<br />
Efficient renal clearance represents one elegant<br />
way of keeping the accumulation and, hence,<br />
the dose to secondary organs, low 9,10 . The present<br />
study 4 indicates how renal clearance could<br />
be achieved following deposition in the lungs.<br />
The new work 4 also defines the requirements<br />
for targeting nanoparticles to regional<br />
lymph nodes, which may be useful for therapies<br />
directed to the immune system. However,<br />
further quantitative investigation is needed to<br />
differentiate between transport to the blood<br />
through lymphatic drainage and direct translocation<br />
across the air-blood barrier to the<br />
blood. Additional challenges will arise in extrapolating<br />
from the rather simple nanoparticles<br />
used in this study to highly functionalized<br />
nanosystems bearing targeting, therapeutic<br />
and diagnostic molecules.<br />
Finally, the results of Frangioni, Tsuda and<br />
colleagues 4 may help in understanding the<br />
Interstitium<br />
Regional<br />
lymph<br />
nodes<br />
Kidneys<br />
Urine<br />
Epithelium<br />
lining fluid<br />
Nano<br />
particles<br />
AM<br />
Fetus<br />
Circulation<br />
Heart<br />
Liver<br />
Bronchial<br />
epithelium<br />
Alveolar<br />
epithelium<br />
Neutrophil<br />
Monocyte<br />
Immune<br />
system<br />
Brain<br />
Figure 1 Schematic of nanoparticle translocation<br />
from the lung epithelium to regional lymph<br />
nodes and blood circulation. Once circulating<br />
in blood, nanoparticles may accumulate in various<br />
secondary organs of the body. The study by<br />
Frangioni, Tsuda and colleagues 4 highlights<br />
two important pathways of nanoparticle<br />
biodistribution: the pathway of lymphatic drainage<br />
toward regional lymph nodes and renal clearance<br />
from blood to urine. AM, alveolar macrophages.<br />
Katie Vicari<br />
nature biotechnology volume 28 number 12 december 2010 1275
news and views<br />
© 2010 Nature America, Inc. All rights reserved.<br />
the same pest species 10 . In another approach,<br />
the US Environmental Protection Agency has<br />
approved sales of mixtures of corn seeds with<br />
and without Bt toxins that kill corn rootworms.<br />
This ensures that farmers comply with the<br />
refuge strategy. However, its use is limited to<br />
the pest species whose larvae do not migrate<br />
between plants. Finally, there are also transgenic<br />
strategies to improve SIT efficiency by<br />
expressing dominant lethal genes, promoting<br />
refractoriness of the sterile insects to disease or<br />
facilitating strain marking, genetic sexing and<br />
molecular sterilization 11 . There thus seems<br />
considerable scope to further enhance the sustainability<br />
of Bt crops and develop new strategies<br />
for integrated pest management.<br />
Nanoparticles in the lung<br />
Wolfgang G Kreyling, Stephanie Hirn & Carsten Schleh<br />
Little is known about the fate of nanoparticles<br />
that enter the lungs, either deliberately through<br />
medical treatments 1 or incidentally through<br />
air pollution 2,3 and occupational exposure<br />
in the workplace. As inhaled nanoparticles<br />
can cross the air-blood barrier into the circulation<br />
and accumulate in secondary organs<br />
and tissues, a biokinetic analysis would be the<br />
first step in a dose-response study as part of<br />
a comprehensive risk assessment (Fig. 1). In<br />
this issue, Frangioni, Tsuda and colleagues 4<br />
undertake such a biokinetic analysis in rats,<br />
using near-infrared imaging to follow the<br />
fate of intratracheally instilled nanoparticles<br />
that are varied systematically in size, surface<br />
modification and core composition. The<br />
authors study the parameters that control<br />
whether nanoparticles remain in the lung,<br />
are transported to regional lymph nodes and<br />
the bloodstream or are cleared from the body<br />
through the kidneys. Their findings should<br />
benefit several areas of research, including<br />
the design of nanoparticles for drug delivery<br />
and the evaluation of the toxicity of particulate<br />
air pollution.<br />
Wolfgang G. Kreyling, Stephanie Hirn and<br />
Carsten Schleh are at the Comprehensive<br />
Pneumology Center, Institute of Lung Biology<br />
and Disease, and Focus Network Nanoparticles<br />
and Health, Helmholtz Zentrum München,<br />
German Research Center for Environmental<br />
Health, Neuherberg/Munich, Germany.<br />
e-mail: kreyling@helmholtz-muenchen.de<br />
COMPETING FINANCIAL INTERESTS<br />
The author declares no competing financial interests.<br />
1. James, C. Global Status of Commercialized Biotech/<br />
GM Crops: 2009 (ISAAA 41, Ithaca, New York,<br />
USA, 2009).<br />
2. Hutchison, W. et al. Science 330, 222–225 (2010).<br />
3. Wu, K. et al. Science 321, 1676–1678 (2008).<br />
4. Tabashnik, B. et al. Nat. Biotechnol. 28, 1304–1307<br />
(2010).<br />
5. Tabashnik, B.E. et al. J. Econ. Entomol. 102, 2011–<br />
2025 (2009).<br />
6. Gould, F. Annu. Rev. Entomol. 43, 701–726 (1998).<br />
7. Liu, C. et al. Sci. China Life Sci. 53, 934–941<br />
(2010).<br />
8. Bagla, P. Science 327, 1439 (2010).<br />
9. Klassen, W. & Curtis, C. in Sterile Insect Technique<br />
(eds. Dyck, V., Hendrichs, J. & Robinson, A.) 3–36<br />
(Springer, The Netherlands, 2005).<br />
10. Tabashnik, B. Science 330, 189–190 (2010).<br />
11. Alphey, N., Bonsall, M.B. & Alphey, L. J. Econ. Entomol.<br />
102, 717–732 (2009).<br />
An imaging study begins to define the parameters that control the<br />
biodistribution of nanoparticles after pulmonary delivery.<br />
The lungs represent a promising route for<br />
drug delivery because of their accessible,<br />
large surface area and because the air-blood<br />
barrier is rather thin 1 . After crossing the airblood<br />
barrier, inhaled nanoparticles have been<br />
shown to reach detectable levels in the blood<br />
circulation and to enter specific cells and<br />
tissues 5 (Fig. 1). Translocation and accumulation<br />
in tissues seem to depend on physicochemical<br />
properties of the nanoparticle core<br />
and surface 5–7 . Thus far, however, few studies<br />
have described nanoparticle transport across<br />
the lungs quantitatively 5,6 .<br />
Frangioni, Tsuda and colleagues 4 find that<br />
noncationic nanoparticles smaller than ~34 nm<br />
in diameter that do not bind serum proteins<br />
reach the regional lymph nodes within 30 min.<br />
Nanoparticles larger than ~34 nm are consistently<br />
retained in the lungs. When the diameter<br />
falls below ~6 nm and the charge is zwitterionic,<br />
about half of the nanoparticles rapidly<br />
enter the bloodstream from the alveolar airspaces<br />
and are mostly cleared from the body by<br />
means of renal filtration. The authors also show<br />
that nanoparticle behavior depends strongly on<br />
the surface coating, which affects binding to<br />
proteins in body fluid.<br />
Biodegradable nanoparticles are currently<br />
considered to be the best choice for targeted<br />
drug delivery, but biopersistent nanoparticles<br />
might also be used if they are cleared from the<br />
body quickly enough to minimize exposure<br />
and retention 8 . As demonstrated previously<br />
by the Frangioni laboratory, intravenously<br />
administered nanoparticles can be rapidly<br />
excreted through the kidneys if their physicochemical<br />
properties are carefully chosen 9,10 .<br />
Efficient renal clearance represents one elegant<br />
way of keeping the accumulation and, hence,<br />
the dose to secondary organs, low 9,10 . The present<br />
study 4 indicates how renal clearance could<br />
be achieved following deposition in the lungs.<br />
The new work 4 also defines the requirements<br />
for targeting nanoparticles to regional<br />
lymph nodes, which may be useful for therapies<br />
directed to the immune system. However,<br />
further quantitative investigation is needed to<br />
differentiate between transport to the blood<br />
through lymphatic drainage and direct translocation<br />
across the air-blood barrier to the<br />
blood. Additional challenges will arise in extrapolating<br />
from the rather simple nanoparticles<br />
used in this study to highly functionalized<br />
nanosystems bearing targeting, therapeutic<br />
and diagnostic molecules.<br />
Finally, the results of Frangioni, Tsuda and<br />
colleagues 4 may help in understanding the<br />
Interstitium<br />
Regional<br />
lymph<br />
nodes<br />
Kidneys<br />
Urine<br />
Epithelium<br />
lining fluid<br />
Nano<br />
particles<br />
AM<br />
Fetus<br />
Circulation<br />
Heart<br />
Liver<br />
Bronchial<br />
epithelium<br />
Alveolar<br />
epithelium<br />
Neutrophil<br />
Monocyte<br />
Immune<br />
system<br />
Brain<br />
Figure 1 Schematic of nanoparticle translocation<br />
from the lung epithelium to regional lymph<br />
nodes and blood circulation. Once circulating<br />
in blood, nanoparticles may accumulate in various<br />
secondary organs of the body. The study by<br />
Frangioni, Tsuda and colleagues 4 highlights<br />
two important pathways of nanoparticle<br />
biodistribution: the pathway of lymphatic drainage<br />
toward regional lymph nodes and renal clearance<br />
from blood to urine. AM, alveolar macrophages.<br />
Katie Vicari<br />
nature biotechnology volume 28 number 12 december 2010 1275
news and views<br />
© 2010 Nature America, Inc. All rights reserved.<br />
health effects of air pollution as the nanoparticles<br />
studied may share common properties with the<br />
nano-fraction of particulate air pollutants. Most<br />
previous work in this field has used inhalation<br />
of nanoparticle aerosols as the ‘gold standard’<br />
method of exposure because the particles will<br />
deposit throughout the entire bronchiolar and<br />
alveolar epithelium in a physiological manner.<br />
However, such studies are rather elaborate and<br />
expensive, and the deposited nanoparticle dose<br />
is usually difficult to determine. Intratracheal<br />
instillation of small nanoparticle suspension<br />
volumes, as used in the present study 4 , is widely<br />
accepted as an alternative method of delivery<br />
to the lungs. Although the dose administered<br />
by this method is well controlled, lung deposition<br />
is less uniform and far from physiological 7 .<br />
Moreover, the dose is frequently high in order<br />
to observe adverse short-term effects.<br />
The daily dose of inhaled nanoparticles in<br />
humans is limited to a maximal nanoparticle<br />
aerosol concentration of 10 12 m –3 (beyond that,<br />
coagulation changes the concentration rapidly<br />
within one minute under ambient air conditions),<br />
and the daily volume of air inhaled by<br />
an adult is ~15 m 3 . Assuming a nanoparticle<br />
deposition probability of 0.3, the daily deposition<br />
amounts to ~5 × 10 12 nanoparticles.<br />
Frangioni, Tsuda and colleagues 4 used doses of<br />
~10 15 nanoparticles per animal, which exceed<br />
human daily doses to the lungs. Their results<br />
must therefore be interpreted with caution<br />
when attempting to extrapolate to the effects<br />
of particulate air pollutants in humans.<br />
In the context of a field that lacks elementary<br />
knowledge about the short- and longterm<br />
transport of nanoparticles delivered to<br />
the lungs, the present study 4 represents an<br />
important advance. Ultimately, a comprehensive<br />
mechanistic understanding of the interactions<br />
of nanoparticles with biomolecules<br />
and proteins in body fluids or with cellular<br />
membranes and of the biokinetics across the<br />
air-blood barrier, the blood-brain barrier, the<br />
placenta, etc. can be expected to allow predictable,<br />
safe and sustained application of nanoparticles<br />
in biomedicine.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
1. Mansour, H.M., Rhee, Y.S. & Wu, X. Int. J.<br />
Nanomedicine 4, 299–319 (2009).<br />
2. Oberdörster, G., Oberdörster, E. & Oberdörster, J.<br />
Environ. Health Perspect. 113, 823–839 (2005).<br />
3. Maynard, A.D. et al. Nature 444, 267–269 (2006).<br />
4. Choi, H.S. et al. Nat. Biotechnol. 28, 1300–1303<br />
(2010).<br />
5. Semmler-Behnke, M. et al. Small 4, 2108–2111<br />
(2008).<br />
6. Geiser, M. & Kreyling, W. Part. Fibre Toxicol. 7,<br />
2 (2010).<br />
7. Oberdorster, G. J. Intern. Med. 267, 89–105<br />
(2010).<br />
8. Kostarelos, K. Nanomed. 5, 341–342 (2010).<br />
9. Choi, H.S. et al. Nat. Nanotechnol. 5, 42–47 (2010).<br />
10. Choi, H.S. et al. Nat. Biotechnol. 25, 1165–1170<br />
(2007).<br />
1276 volume 28 number 12 december 2010 nature biotechnology
esearch highlights<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Rett’s-like neurons from iPS cells<br />
Induced pluripotent stem (iPS) cells offer<br />
the possibility of creating disease-specific<br />
cells that model pathology and can be used<br />
for screening therapies. Using skin biopsies<br />
from Rett syndrome patients, Marchetto<br />
et al. create iPS cells, from which they are<br />
able to create both neuronal progenitor<br />
cells and mature neurons. Whereas the<br />
progenitor cells appear normal, the Rett’slike<br />
neurons are both morphologically and<br />
functionally abnormal. This recapitulates<br />
the onset of disease, as symptoms don’t<br />
appear until as long as 18 months after<br />
birth (when progenitors have differentiated).<br />
The authors show that knockdown of methyl-CpG–binding protein<br />
(MeCpG), which is mutated in Rett’s patients, leads to cells with<br />
fewer synapses than neurons derived from iPS cells generated from<br />
normal individuals. Finally, the application of two drugs—insulinlike<br />
growth factor, which reverses the phenotype in a Rett’s<br />
mouse model, and gentamicin, which reverses premature protein<br />
termination—partially rescues the Rett’s-like neurons. By staining<br />
cell populations for proteins involved with X inactivation—of<br />
interest as MeCP2 is X linked—the authors could select iPS cells<br />
that had two active X chromosomes. Even so, during neuronal<br />
differentiation, re-inactivation of the X chromosomes was not<br />
random, suggesting the presence of some remnant of X inactivation<br />
after reprogramming. (Cell 143, 527–539, 2010)<br />
LD<br />
Personalizing epigenetic therapy<br />
Unlike the genetic changes that contribute to the progression of liver<br />
cancer, the epigenetic modifications responsible for driving the development<br />
of the disease can be reversed by pharmacological intervention.<br />
Nonetheless, there is little understanding of the factors that influence<br />
whether the therapeutic effects of broad-spectrum chemical inhibitors<br />
of DNA methylation (e.g., reactivation of epigenetically silenced tumor<br />
suppressor genes) over-ride any detrimental consequences. Working with<br />
the cytidine analog zebularine, Anderson et al. use transcriptomic and<br />
epigenomic profiling to reveal a response signature that classifies liver<br />
cancer cell lines as being either sensitive or resistant to the drug. The ability<br />
of zebularine to promote apoptosis in sensitive cell lines correlates with<br />
reduced growth and metastasis of drug-sensitive tumors and prolonged<br />
survival of mice bearing xenografts of sensitive human tumors. In contrast,<br />
the drug increased tumor growth rates and decreased survival rates<br />
of mice bearing resistant xenografts—consistent with zebularine’s ability to<br />
upregulate oncogenic pathways in cell lines predicted to be resistant to the<br />
drug. The ability of the signature to predict clinical outcome with 84–96%<br />
accuracy in a relatively small cohort of liver cancer patients suggests its<br />
value for identifying individuals most likely to benefit from methyltransferase<br />
inhibitors. (Sci. Transl. Med. 2, 54ra77, 2010)<br />
PH<br />
The ’omics of protein–small molecule<br />
interactions<br />
Metabolites are important regulators of many proteins. Nonetheless, the<br />
systematic analysis of small molecule–protein interactions has lagged<br />
Written by Kathy Aschheim, Laura DeFrancesco, Markus Elsner,<br />
Peter Hare & Craig Mak<br />
behind efforts to characterize, for example, the entirety of proteinprotein<br />
or protein-DNA interactions. Now, Li et al. perform a largescale<br />
analysis of hydrophobic metabolites that bind the enzymes of the<br />
ergosterol synthesis pathway and protein kinases in yeast. They affinity<br />
purify tagged versions of proteins of interest and identify proteinbound<br />
molecules by mass spectrometry after methanol extraction. In<br />
the ergosterol pathway, the authors uncover many previously unknown<br />
interactions between intermediate products and enzymes at other levels<br />
of the synthetic cascade. This suggests more integrated control of sterol<br />
synthesis than previously appreciated, although a detailed functional<br />
analysis of the significance of the interactions was not performed.<br />
Among the 103 kinases studied, 21 bind a total of ten different metabolites.<br />
The authors go on to demonstrate that the binding of ergosterol<br />
stabilizes Ssk22 and activates the highly conserved Ypk1 kinase. It seems<br />
likely that the approach could be adapted for hydrophilic molecules.<br />
(Cell 143, 639–650, 2010)<br />
ME<br />
Efficient discovery of rare genetic variants<br />
With sequencing single genomes now commonplace, efforts are shifting<br />
to gather data from many individuals. The resulting increase in statistical<br />
power should identify rare genetic variants, which may underlie<br />
disease. Initial results toward this goal are reported in Nature by the<br />
1000 Genomes Project Consortium, with analytic algorithms detailed in<br />
Genome Research. Three pilot sequencing efforts demonstrate efficient<br />
uses of sequencing to survey genetic variation. First, by reducing the average<br />
number of times any given region of the genome was sequenced, called<br />
‘low coverage’ sequencing, the project analyzed 179 genomes. This identified<br />
most common single-nucleotide variants present in >5% of individuals<br />
and many less-common variants. Second, targeted sequencing<br />
of a subset of the genome allowed 8,140 exons to be analyzed across 697<br />
individuals. Third, sequencing related individuals—in this case a mother,<br />
father and child from two families—enabled de novo germline mutations<br />
to be identified. In Science, Sudmant et al. used single, unique nucleotides<br />
found within repetitive regions of the genome to discover remarkable<br />
plasticity in copy number variation in duplicate genes, thereby making<br />
them amenable to genetic association studies. (Nature 467, 1061–1073,<br />
2010; Genome Res., published online 27 October 2010, doi:10.1101/<br />
gr.1123267.110, doi:10.1101/gr.111120.110, doi:10.1101/gr.113084.110;<br />
Science 330, 641–646, 2010)<br />
CM<br />
lincRNAs in reprogramming<br />
The reprogramming of somatic cells to induced pluripotent stem cells<br />
(iPSCs) is accompanied by widespread changes in gene expression and<br />
epigenetic marks. A recent study by Loewer et al. explores whether it<br />
also involves changes in the expression of long intervening noncoding<br />
RNAs (lincRNAs), a class of RNAs with diverse roles, including<br />
regulation of gene expression and of the epigenome. The authors began<br />
by comparing the expression of ~900 lincRNAs in human fibroblasts,<br />
in iPSCs derived from the fibroblasts and in human embryonic stem<br />
cells. This analysis identified 133 upregulated and 104 downregulated<br />
lincRNAs in the pluripotent cells compared with the fibroblasts.<br />
Twenty-eight of the upregulated lincRNAs were expressed more highly<br />
in iPSCs than in embryonic stem cells, suggesting a possible involvement<br />
in reprogramming. To identify a more generic association with<br />
reprogramming, the authors looked for lincRNAs upregulated in iPSCs<br />
derived both from fibroblasts and from CD34 + cells. This approach<br />
yielded ten candidate lincRNAs, one of which was shown experimentally<br />
to influence reprogramming. (Nat. Genet. published online,<br />
7 November 2010; doi:10.1038/ng.710) KA<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1277
A n a ly s i s<br />
Large-scale in silico modeling of metabolic interactions<br />
between cell types in the human brain<br />
Nathan E Lewis 1 , Gunnar Schramm 2,5 , Aarash Bordbar 1 , Jan Schellenberger 3 , Michael P Andersen 1 , Jeffrey K Cheng 1 ,<br />
Nilam Patel 1 , Alex Yee 1 , Randall A Lewis 4 , Roland Eils 2,5 , Rainer König 2,5 & Bernhard O Palsson 1<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Metabolic interactions between multiple cell types are difficult<br />
to model using existing approaches. Here we present a workflow<br />
that integrates gene expression data, proteomics data and<br />
literature-based manual curation to model human metabolism<br />
within and between different types of cells. Transport reactions<br />
are used to account for the transfer of metabolites between<br />
models of different cell types via the interstitial fluid. We apply<br />
the method to create models of brain energy metabolism that<br />
recapitulate metabolic interactions between astrocytes and<br />
various neuron types relevant to Alzheimer’s disease. Analysis<br />
of the models identifies genes and pathways that may explain<br />
observed experimental phenomena, including the differential<br />
effects of the disease on cell types and regions of the brain.<br />
Constraint-based modeling can thus contribute to the study<br />
and analysis of multicellular metabolic processes in the human<br />
tissue microenvironment and provide detailed mechanistic<br />
insight into high-throughput data analysis.<br />
Constraint-based reconstruction and analysis of genome-scale<br />
microbial metabolic networks has matured over the past decade.<br />
This approach has a wide range of applications, such as providing<br />
insight into evolution, aiding in metabolic engineering, and providing<br />
a mechanistic bridge between genotypes and complex phenotypes 1,2 .<br />
Computational methods 3 and detailed standard operating procedures 4<br />
have been outlined for the reconstruction of high-quality prokaryotic<br />
metabolic networks, and many methods can be deployed for their<br />
analysis 5,6 . Constraint-based modeling of metabolism entered a new<br />
phase with the publication of the human metabolic network (Recon 1) 7 ,<br />
based on build 35 of the human genome. Methods allowing tissuespecific<br />
model construction have followed 8–10 .<br />
Many tissue metabolic functions rely on interactions between cell<br />
types. Thus, methods are needed that integrate the metabolic activities of<br />
multiple cell types. Here, using Recon 1, we analyze and integrate ’omics<br />
1 Department of Bioengineering, University of California, San Diego, La Jolla,<br />
California, USA. 2 Department of Bioinformatics and Functional Genomics,<br />
Institute of Pharmacy and Molecular Biotechnology and Bioquant, University<br />
of Heidelberg, Heidelberg, Germany. 3 Bioinformatics Program, University of<br />
California, San Diego, La Jolla, California, USA. 4 Department of Economics,<br />
Massachusetts Institute of Technology, Cambridge, Massachusetts, USA.<br />
5 Department of Theoretical Bioinformatics, German Cancer Research Center<br />
(DKFZ), Heidelberg, Germany. Correspondence should be addressed to B.O.P.<br />
(palsson@ucsd.edu).<br />
Published online 21 November 2010; doi:10.1038/nbt.1711<br />
data with information from detailed biochemical studies to build multicellular,<br />
constraint-based models of metabolism. We demonstrate this<br />
process by constructing and analyzing models of human brain energy<br />
metabolism, with an emphasis on central metabolism and mitochondrial<br />
function in astrocytes and neurons. Moreover, we provide three detailed<br />
examples, demonstrating the use of models to guide experimental work<br />
and provide biological insight into the metabolic mechanisms underlying<br />
physiological and pathophysiological states in the brain.<br />
RESULTS<br />
Building metabolic models of multiple cell types<br />
‘Omics data sets can be difficult to analyze owing to their size.<br />
However, such data sets can be used to construct large mechanistic<br />
models for specific tissues and cell types 8,9 that serve as a context for<br />
further analysis. The workflow for generating multicellular models<br />
(Fig. 1) consists of the following four steps.<br />
Step 1. Reconstruct a metabolic network for the organism of interest<br />
from genome annotation, lists of biomolecular components and<br />
the literature 4 . Metabolic pathways and associated gene products<br />
are not completely known for any species. Thus, a reconstruction is<br />
refined through iterations of manual curation, hypothesis generation,<br />
experimental validation and incorporation of new knowledge. Recon 1<br />
has been through five iterations 7 .<br />
Step 2. Identify reactions specific to the tissue of interest. Many gene<br />
products are not expressed in all cells at any given time 11 . Therefore,<br />
the presence of a gene product, inferred from ’omics data, such as gene<br />
expression or mass spectrometry, is mapped to the metabolic network<br />
using associations linking genes to proteins and to reactions. This<br />
process may be performed manually or algorithmically 8,9 .<br />
Step 3. Partition the reconstruction, using manual curation of the<br />
literature, into compartments representing different cell types and<br />
organelles. The different cell types are linked with transport reactions,<br />
as supported by literature and experimental data. Initial context-specific<br />
reconstructions are incomplete and may contain false positives owing<br />
to contamination from proximal tissue. Moreover, few high-throughput<br />
data sets are cell-type specific. Thus, the initial reconstruction represents<br />
the union of metabolic networks from various cell types. After the<br />
reconstruction is partitioned using manual curation, it is converted into<br />
a model by specifying inputs, outputs and relevant parameters, and by<br />
representing the network mathematically 12 . Details of proper manual<br />
curation have been described 4 .<br />
Step 4. Simulation and analysis 1,2 . Once the network is accurately<br />
reconstructed and converted into an in silico model, it is used to<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1279
A n a ly s i s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Figure 1 A workflow for bridging the genotypephenotype<br />
gap with the use of high-throughput<br />
data and manual curation for the construction of<br />
multicellular models of metabolism. The models<br />
can be used to (i) predict disease-associated<br />
genes, such as glutamate decarboxylase;<br />
(ii) analyze high-throughput data in the network<br />
context to identify sets of genes that change<br />
together and affect specific pathways, such as<br />
the brain-region-specific suppression of central<br />
metabolism in Alzheimer’s disease patients;<br />
(iii) analyze physiological data in the context of<br />
the model, thereby enabling, for example, the<br />
identification of tissue properties relevant to<br />
disease treatment, such as the calculation of<br />
the percentage of the brain that is cholinergic.<br />
generate hypotheses and to obtain insight into<br />
systems-level biological functions.<br />
In contrast to previous approaches that<br />
reconstructed tissue–specific models 8–10 , the<br />
approach described here accounts for interactions<br />
between cell types within the tissue. First,<br />
published experimental results (e.g., immunohistochemistry<br />
staining of brain tissue 11 ) are<br />
used to refine the reconstructed network into<br />
individual cell type networks. Subsequently, interactions between<br />
these cell types are included as transporters that transfer metabolites<br />
between the different cell type models. This process results in a<br />
tissue-specific model that accounts for metabolic interactions between<br />
cells and the separation of metabolic functions into the relevant cell<br />
types. Thus they provide a more accurate view into tissue metabolism.<br />
Previous work had not taken these steps to generate multicellular<br />
tissue-specific models of this scale.<br />
This workflow was used to build three different multicellular models<br />
of brain energy metabolism. Each model represents one canonical<br />
neuron type (that is, glutamatergic, γ-aminobutarate (GABA)ergic<br />
or cholinergic), its interactions with the surrounding astrocytes and<br />
the transport of metabolites through the blood-brain barrier (Fig. 2).<br />
This reconstruction focuses on the core of cerebral energy metabolism,<br />
including central metabolism, mitochondrial metabolic pathways, and<br />
pathways relevant to anabolism and catabolism of the neurotransmitters<br />
glutamate, GABA and acetylcholine. Briefly, we began with the manually<br />
curated human metabolic reconstruction Recon 1 (ref. 7) (step 1).<br />
Next, we extracted relevant brain-specific reactions by mapping them<br />
to proteins expressed or localized to the brain as reported in the Human<br />
Protein Reference Database (release 5) 13 , H-inv (version 4.3) (HINV) 14<br />
or The Human Proteome Organisation (HUPO) brain proteome<br />
project 15 ; additional reactions were added as dictated by biochemical<br />
data from the literature (step 2). Reactions and pathways were manually<br />
Measured CMR of metabolites<br />
Endothelium/blood<br />
Genome<br />
Step 1: Build species-specific<br />
reconstruction<br />
Experimental<br />
data<br />
EC<br />
HIP<br />
SFG<br />
VCX<br />
Brain region<br />
Literature<br />
Add new<br />
knowledge<br />
Relative<br />
intensity<br />
Step 2: Find context-specific<br />
subnetwork<br />
m/z<br />
Step 3: Construct a<br />
manually curated model<br />
Compartmentalization<br />
and manual<br />
curation<br />
Formulation<br />
into model<br />
Step 4: Simulation and analysis for predictions and biological insight<br />
(i) Gene identification<br />
(ii) 'Omics data analysis<br />
(iii) Physiological data analysis<br />
GAD1<br />
GAD2<br />
Context-specific<br />
data<br />
Suppressed<br />
pathways<br />
Genes<br />
EC<br />
HIP<br />
MTG<br />
PC<br />
SFG<br />
VCX<br />
Down<br />
Up<br />
14 CO 2 release<br />
curated to verify their presence in the human brain, to determine celltype<br />
specificity and to add reactions unique to the different neuron<br />
types (step 3) (see Online Methods for complete details). Thus, the<br />
three models contain the high-flux pathways and important reactions<br />
in neuron and astrocyte metabolism. To our knowledge, these models<br />
currently represent the largest and most detailed models of brain<br />
energy metabolism 16–18 (Supplementary Notes).<br />
Our models of glutamatergic, GABAergic and cholinergic neurons<br />
contain 1,066, 1,067 and 1,070 compartment-specific interactions (that<br />
is, reactions, transformations and exchanges), respectively, involving 983,<br />
983 and 987 compartment-specific metabolites. In total, the three models<br />
are associated with 403 genes. Lists of reactions, genes, citations and<br />
parameters used to constrain the models are detailed in Supplementary<br />
Tables 1–5. The validity of these models is demonstrated through comparisons<br />
to physiological data. Specifically, our models predict ATP production<br />
rates within 8% of the average published value, and internal<br />
flux measurements are consistent with experimentally measured values<br />
(Supplementary Notes). Moreover, three analyses using the models are<br />
detailed here (step 4). Most of these analyses cannot be done on previous<br />
brain models or on Recon 1 (ref. 7) as published. Thus, our models can<br />
provide novel insight into brain energy metabolism.<br />
Identifying genes behind cell type–specific metabolic phenotypes<br />
Alzheimer’s disease is characterized by dementia and is diagnosed postmortem<br />
by histopathological features such as neurofibrillary tangles and<br />
β-amyloid plaques. Notably, metabolic rates of various brain regions<br />
decrease years before the onset of dementia 19 . Experiments have suggested<br />
that glutamatergic and cholinergic neurons are more affected in<br />
moderate stages of Alzheimer’s disease 20 , whereas most GABAergic cells<br />
ChAT flux<br />
Exp. data<br />
Simulation<br />
Astrocyte<br />
Mito<br />
Int<br />
Mito<br />
Glutamatergic<br />
neuron<br />
GABAergic<br />
neuron<br />
Cholinergic<br />
neuron<br />
Figure 2 General structure of the models. Three models were built from the<br />
brain reconstruction. Each model consists of various compartments including<br />
the endothelium/blood, astrocytes, astrocytic mitochondria, neurons, neuronal<br />
mitochondria and an interstitial space between the cell types. Each neuron<br />
metabolic network was tailored to represent a specific neuron type, containing<br />
genes and reactions generally accepted to be unique to the neuron type. Mito,<br />
mitochondrion; Int, interstitial space; CMR, cerebral metabolic rate.<br />
1280 VOLUME 28 NUMBER 12 DECEMBER 2010 nature biotechnology
a n a ly s i s<br />
are relatively unaffected until later stages 21 . Genes responsible for these<br />
differences are not known. Thus, we first analyzed the models to identify<br />
genes that potentially underlie these cell type–specific effects.<br />
Analysis of our models yielded results consistent with known<br />
metabolic changes in Alzheimer’s. Several central metabolic enzymes<br />
exhibit altered expression or activity in Alzheimer’s disease, such as<br />
pyruvate dehydrogenase (PDHm), α-ketoglutarate dehydrogenase<br />
(AKGDm) and cytochrome c oxidase 22–24 . The activities of these<br />
enzymes are affected by the Alzheimer’s disease–related proteins<br />
β-amyloid and Tau kinase 25,26 . In silico, as the activities of PDHm<br />
and cytochrome c oxidase decrease, neurons demonstrate impaired<br />
metabolic capacity (Supplementary Notes and Fig. 1a,b therein), and<br />
deficiencies in PDHm activity leads to a decreased cholinergic neurotransmission<br />
capacity (Supplementary Notes and Fig. 1c therein).<br />
The in vitro AKGDm activity shows the greatest impairment in brains<br />
from individuals with Alzheimer’s disease examined postmortem<br />
(57% decrease compared to normal) 23 . An in silico analysis shows that<br />
this deficiency impairs the metabolic rate in glutamatergic and cholinergic<br />
neurons (Fig. 3a) because it limits oxidative phosphorylation<br />
capacity in these neurons (Fig. 3b,c). Such impairment of oxidative<br />
phosphorylation leads to neuronal apoptosis 27 . However, oxidative<br />
phosphorylation is not impaired in the GABAergic neuron model<br />
(Fig. 3d), consistent with the different phenotypes of GABAergic<br />
compared with glutamatergic and cholinergic neurons. Therefore,<br />
the models were further interrogated to identify a mechanism that<br />
allows GABAergic neurons to absorb the perturbation, thereby leading<br />
to the cell type–specific effects.<br />
Simulations show that GABAergic neurons absorb the AKGDm<br />
perturbation through the GABA shunt (Fig. 3e,f), a pathway that<br />
uses 4-aminobutyrate transaminase and succinate-semialdehyde<br />
© 2010 Nature America, Inc. All rights reserved.<br />
a<br />
b<br />
c<br />
d<br />
e<br />
0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2<br />
Rate of cerebral CO 2 release (µmol min −1 g wet brain −1 )<br />
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0<br />
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0<br />
0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4<br />
Cytochrome c oxidase flux (µmol min −1 g wet brain −1 )<br />
10 −4 10 −3 10 −2 10 −1 10 0<br />
GABA shunt flux (µmol min −1 g wet brain −1 )<br />
Normal glutamatergic<br />
AKGDm inhib. glutamatergic<br />
Normal cholinergic<br />
AKGDm inhib. cholinergic<br />
Normal GABAergic<br />
AKGDm inhib. GABAergic<br />
g<br />
Normalized non-AD GAD NMN expression<br />
f<br />
coa[m]<br />
pyr[m]<br />
atp[m]<br />
coa[m]<br />
amp[m]<br />
ppi[m]<br />
CO [m] 2<br />
nad[m] nadh[m]<br />
PDHm<br />
1.0<br />
0.8<br />
0.6<br />
0.4<br />
0.2<br />
0<br />
ac[m]<br />
cit[m] ACSm<br />
CSm<br />
accoa[m]<br />
h[m]<br />
ACONTm<br />
TCA cycle<br />
nad[m]<br />
icit[m]<br />
coa[m]<br />
h[m]<br />
H 2<br />
O[m]<br />
oaa[m]<br />
mal-L[m]<br />
EC HIP MTG PC SFG VCX<br />
Brain region<br />
ICDHxm<br />
nadp[m]<br />
nadh[m]<br />
MDHm<br />
nad[m]<br />
FUMm<br />
ICDHyrm<br />
H 2<br />
O[m]<br />
fum[m]<br />
fadh 2<br />
[m]<br />
nadh[m]<br />
CO 2<br />
[m]<br />
nadph[m]<br />
glu-L[m]<br />
GABA<br />
shunt<br />
h[m]<br />
nadh[m]<br />
akg[m]<br />
CO 2<br />
[m]<br />
sucsal[m]<br />
fad[m]<br />
SUCD1m<br />
h<br />
Brain region<br />
succ[m]<br />
atp[m]<br />
EC<br />
HIP<br />
MTG<br />
PC<br />
SFG<br />
VCX<br />
nad[m]<br />
coa[m]<br />
4abut[m]<br />
ABTArm<br />
nad[m]<br />
h2o[m]<br />
SSALxm<br />
SUCOASm<br />
coa[m]<br />
GAD1<br />
adp[m]<br />
AKGDm<br />
nadh[m]<br />
succoa[m]<br />
GAD2<br />
pi[m]<br />
DLX<br />
co 2<br />
[m]<br />
1.5<br />
1.0<br />
0.5<br />
0<br />
−0.5<br />
−1.0<br />
−1.5<br />
1.0<br />
0.5<br />
0<br />
−0.5<br />
−1.0<br />
log 2 (Flux fold-change in AD)<br />
log 2 (GAD NMN expression change in AD)<br />
Figure 3 Decrease in AKGDm activity associated with Alzheimer’s disease (AD) shows cell-type and regional effects in silico consistent with<br />
experimental data. (a–e) Kernel density plots show the distribution of feasible fluxes for various reactions. An in silico reduction of AKGDm flux from<br />
normal activity (solid lines) to Alzheimer’s disease brain activity (dashed) decreases the oxidative metabolic rate for glutamatergic and cholinergic<br />
neurons, but not GABAergic neurons (a). This results from a decrease in the feasible fluxes for oxidative phosphorylation (e.g., cytochrome c oxidase)<br />
for both glutamatergic (b) and cholinergic neurons (c), but not GABAergic cells (d). (e,f) This cell-type-specific protection from the AKGDm deficiency<br />
results from an increased flux through the GABA shunt in GABAergic cells (e), by bypassing the damaged AKGDm (f). GABAergic cells maintain a higher<br />
GABA shunt flux because of the expression of glutamate decarboxylase (GAD). Neuroprotective properties of GAD are supported by gene expression.<br />
(g) Severely damaged brain regions in Alzheimer’s disease patients have lower GAD NMN expression in control brain, whereas high GAD NMN regions<br />
(SFG and VCX) show little damage. (h) In Alzheimer’s disease brain, severely affected regions (HIP and entorhinal cortex) show an increase in<br />
GAD NMN and the GAD-inducing DLX family, suggesting that non-GAD expressing neurons may be lost in Alzheimer’s disease. EC, entorhinal cortex;<br />
HIP, hippocampus; MTG, middle temporal gyrus; PC, posterior cingulate cortex; SFG, superior frontal gyrus; VCX, visual cortex; NMN, neuron marker<br />
normalized; inhib., inhibited. All reaction and metabolite abbreviations are defined in Supplementary Tables 1 and 2.<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1281
A n a ly s i s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
dehydrogenase to bypass part of the tricarboxylic acid (TCA) cycle.<br />
However, our models suggest that glutamatergic and cholinergic neurons<br />
cannot, despite carrying a small flux through the shunt enzymes<br />
(Fig. 3e). Support for these results includes recent evidence that suggests<br />
that cerebellar granule neurons, which have higher levels of<br />
GABA, can absorb perturbations to AKGDm through this shunt 28 .<br />
To identify the mechanism allowing only GABAergic neurons to<br />
use the GABA shunt to absorb the AKGDm perturbation, an in silico<br />
analysis was performed to identify contributing genes. Briefly, each<br />
reaction was removed from the GABAergic model, followed by an<br />
assessment of the correlation of flux between AKGDm and oxidative<br />
phosphorylation (Supplementary Notes). This analysis suggested<br />
that the two isoforms of glutamate decarboxylase (GAD) could provide<br />
the cell type–specific neuroprotection.<br />
GAD allows the GABA shunt to carry a higher flux following the<br />
AKGDm perturbation in GABAergic neurons; however, the lack of<br />
GAD in other neuron types greatly limits the use of the GABA shunt<br />
in silico (Fig. 3e). Therefore, by fueling the GABA shunt, GAD may<br />
play a neuroprotective function, thus sparing of GABAergic systems<br />
in earlier Alzheimer’s disease 21 .<br />
Certain populations of glutamatergic and cholinergic cells tend to<br />
be lost earlier in Alzheimer’s disease but others survive. Interestingly,<br />
although GAD is canonically encoded by a GABAergic gene, it occasionally<br />
shows low expression in other neuron types, including<br />
glutamatergic and cholinergic cells 29 . Therefore, such populations of<br />
non-GABAergic, GAD-expressing neurons would also be protected.<br />
Thus, we follow with an analysis of the correlation of GAD expression<br />
and Alzheimer’s disease pathology for further validation.<br />
If GAD has neuroprotective capacity in vivo, we would expect that<br />
brain regions with less GAD per neuron will be more affected in<br />
Alzheimer’s disease, whereas regions with abundant GAD will be<br />
spared. To test this hypothesis, we used a compendium of published<br />
microarrays of neurons without tangles from six brain regions in<br />
Alzheimer’s patients and age-matched, non-Alzheimer’s controls 30 .<br />
In controls, GAD expression levels among the brain regions is consistent<br />
with the extent of neuron loss found in Alzheimer’s disease<br />
patients; that is, brain regions with more neuron loss in Alzheimer’s<br />
disease (e.g., the entorhinal cortex and hippocampus) have lower<br />
GAD expression in control patients, whereas relatively unaffected<br />
regions in Alzheimer’s disease (e.g., superior frontal gyrus and visual<br />
cortex) show much higher levels of GAD expression (Fig. 3g).<br />
In addition, if GAD is neuroprotective, neurons with low GAD<br />
expression should be lost in Alzheimer’s disease, resulting in an<br />
increase in expression of GAD per neuron in histopathologically<br />
affected regions. In microarrays, we observed a significant increase<br />
in the expression of the brain-specific GAD2 in the entorhinal cortex<br />
and hippocampus in Alzheimer’s disease (P = 0.0050 and 0.018,<br />
respectively; SAM test) (Fig. 3h). Expression of all other neuronspecific<br />
genes were tested as controls and showed no correlation with<br />
Alzheimer’s disease pathology (Supplementary Notes and Fig. 2<br />
therein), except the genes encoding DLX2 and DLX5, which induce<br />
GAD expression in the brain (Fig. 3h and Supplementary Notes and<br />
Fig. 3 therein) 31 . Therefore, these results lend additional support to<br />
the possibility that GAD is providing a neuroprotective effect and<br />
that this effect is correlated with the regional specificity of Alzheimer’s<br />
disease. Moreover, the model was able to guide the identification of a<br />
gene and the mechanism for its role in Alzheimer’s disease.<br />
Pathway-based analysis of multicellular models<br />
Atrophy and the cell type–specific effects investigated above cannot<br />
fully explain the decreased metabolic rate in Alzheimer’s disease in<br />
many brain regions 32 . Therefore, using the models as a context for analyzing<br />
gene expression data from Alzheimer’s patients and age-matched<br />
controls, we searched for downregulated metabolic pathways within<br />
surviving cells in metabolically suppressed brain regions. This was done<br />
using PathWave 33 , a method that takes each model and maps microarray<br />
data to the metabolic pathways, which have been optimally arranged on<br />
a two-dimensional grid. Haar Wavelet transforms detected concerted<br />
regulation of neighboring enzymes in the models. Thus, we could identify<br />
groups of differentially expressed genes that are not solely related<br />
by annotation but that are connected by metabolic pathway functions,<br />
thereby providing a mechanistic view of the effects of differential gene<br />
expression (see Online Methods and Supplementary Notes).<br />
This analysis revealed that brain regions showed distinct changes<br />
in metabolic pathways (Supplementary Table 6). The visual cortex<br />
and superior frontal gyrus lack any differentially expressed pathways,<br />
consistent with previous work that shows little change in metabolic<br />
rate in Alzheimer’s disease in these regions 30 . However, the posterior<br />
cingulate cortex (PC) and middle temporal gyrus (MTG) have the<br />
most significantly differentially expressed pathways (23 and 18, respectively).<br />
These two regions show significantly decreased metabolic rates<br />
in Alzheimer’s disease but show fewer histopathological effects 30 . Both<br />
the entorhinal cortex and hippocampus also show decreases in expression<br />
of nine metabolic pathways, though the number may be lower<br />
since these regions suffer a high amount of neuron loss and only histopathologically<br />
healthy neurons were expression profiled. Therefore,<br />
more affected neurons may already have been lost or not profiled.<br />
PathWave also revealed that the four brain regions showing substantially<br />
lower metabolic rates in Alzheimer’s disease (PC, MTG,<br />
hippocampus and entorhinal cortex) 32 also show a significant suppression<br />
of glycolysis and the TCA cycle (P < 5 × 10 −4 ) (Fig. 4). In<br />
addition, the hippocampus, MTG and PC show a suppression of the<br />
malate-aspartate shuttle and oxidative phosphorylation. Individual<br />
regions also show a suppression of other pathways, such as heme<br />
biosynthesis (MTG, PC), ethanol metabolism (entorhinal cortex, PC)<br />
and several amino acid metabolism pathways. Thus, using pathway<br />
topology of our models as a context for microarray analysis, we find<br />
that the decreased metabolic rate in specific regions in Alzheimer’s<br />
disease is associated with the downregulation of central metabolic<br />
gene expression in histopathologically normal neurons.<br />
Identifying metabolic properties relevant to treatment<br />
Deficiencies in cholinergic neurotransmission have long been believed<br />
to contribute to Alzheimer’s disease. Thus, treatment of Alzheimer’s<br />
patients often includes efforts to enhance the cholinergic system.<br />
Although relevant to treatment, it is still not clearly understood which<br />
pathways are used to synthesize acetylcholine and what percentage of<br />
the brain participates in cholinergic neurotransmission.<br />
Studies have demonstrated that cytosolic acetyl-CoA, which is<br />
used to synthesize the neurotransmitter acetylcholine, comes from<br />
the acetyl-CoA formed in the mitochondria. This tight coupling of<br />
acetylcholine to mitochondrial metabolism allows treatments that<br />
increase glucose uptake in the brain to improve cognitive functions<br />
in rats 34 and humans with severe cholinergic cognitive pathologies,<br />
such as Alzheimer’s disease and trisomy 21 (ref. 35). Pathways that<br />
transport acetyl-CoA carbon to the cytosol have been suggested; however,<br />
the mechanism is still not clear 36 .<br />
Constraint-based modeling helped identify two pathways that could<br />
indirectly transport acetyl-CoA into the cytosol and provided insight into<br />
needed complementary pathways. To identify these pathways, reaction<br />
sets were identified by randomly removing reactions from Recon 1 until a<br />
minimum set was determined that couples the mitochondrial and cytosolic<br />
1282 VOLUME 28 NUMBER 12 DECEMBER 2010 nature biotechnology
a n a ly s i s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
a<br />
GLCt2r<br />
G3PD2m<br />
HEX1<br />
PGI<br />
PFK<br />
TPI<br />
FBA<br />
GAPD<br />
PGK<br />
ACYP<br />
EC HIP MTG PC SFG VCX EC HIP MTG PC SFG VCX<br />
ACYP<br />
ABTArm<br />
DPGM<br />
ACONTm<br />
DPGase<br />
AKGDm<br />
ENO<br />
PDHm<br />
CSm<br />
PCm<br />
FBA<br />
FUMm<br />
G3PD2m<br />
CSm<br />
GLUDxm<br />
GAPD<br />
ACONTm<br />
GLUDym<br />
GLCt1r<br />
MDHm<br />
ICDHxm<br />
GLCt2r<br />
ICDHxm ICDHyrm<br />
HEX1<br />
ICDHyrm<br />
MALtm<br />
PFK<br />
MDHm<br />
PGI<br />
SUCD1m<br />
ME2m<br />
PGK<br />
PDHm<br />
PGM<br />
SSALxm<br />
PYK<br />
SSALxm ABTArm<br />
SUCCt2m<br />
TPI<br />
SUCOAS1m<br />
SUCD1m<br />
Regulation of surrounding reactions SUCOASm<br />
SUCOAS1m<br />
Down Up<br />
AKGDm<br />
SUCOASm<br />
PGM ENO PYK<br />
>6 3 0 3 >6<br />
Regulation of surrounding reactions<br />
–log 10 (P value)<br />
GLUDxm GLUDym<br />
Down Up<br />
acetyl-CoA pools. This was repeated until more than 21,000 unique minimal<br />
reaction sets were identified. Singular value decomposition was then<br />
used to identify dominant sets of reactions that frequently co-occur.<br />
The first singular vector is dominated by reactions that frequently<br />
co-occur in the reaction sets (e.g., water transport across cell membranes).<br />
However, the second and third singular vectors are dominated<br />
by reactions that usually co-occur or never co-occur (Fig. 5).<br />
These reactions cluster into three distinct pathways, providing<br />
hypotheses of pathways coupling acetylcholine synthesis and mitochondrial<br />
metabolism. After re-examining the literature and the<br />
’omics data used in the reconstruction, we were able to eliminate<br />
the pathway using cytosolic acetyl-CoA synthetase (Fig. 5a) and<br />
b<br />
MALtm<br />
ME2m<br />
SUCCt2m<br />
FUMm<br />
ME1m<br />
>6 3 0 3 >6<br />
–log 10 (P value)<br />
Figure 4 Analysis of metabolic pathways in different regions of brains affected by Alzheimer’s. (a,b) Pathway analysis demonstrates that<br />
histopathologically normal cells from regions of the brain that are metabolically affected (EC, hippocampus, MTG, and PC) demonstrate a significant<br />
suppression of central metabolic pathways, such as glycolysis (a) and the TCA cycle and surrounding reactions (b). Regions that are metabolically less<br />
affected (SFG and VCX) show no significant suppression. Reaction suppression shown here is a composite expression of the reaction-associated genes<br />
and the genes of closely connected reactions. Only significantly changed reactions are shown (FDR = 0.05). EC, entorhinal cortex, HIP, hippocampus,<br />
MTG, middle temporal gyrus, PC, posterior cingulate cortex, SFG, superior frontal gyrus, VCX, visual cortex. All reaction and metabolite abbreviations<br />
are defined in Supplementary Tables 1 and 2.<br />
validate pathways using ATP-citrate lyase (ACITL) or cytosolic<br />
acetyl-CoA C-acetyltransferase (ACACT1r), which transport<br />
acetyl-CoA from the mitochondria to the cytosol on citrate or<br />
acetoacetate, respectively (Fig. 5b,c and Supplementary Notes).<br />
These two pathways were included in our cholinergic model, contributing<br />
to a correlation between the flux through mitochondrial<br />
pyruvate dehydrogenase and choline acetyltransferase (r = 0.45,<br />
P = 3 × 10 −247 ), consistent with the experimentally observed coupling<br />
between mitochondrial metabolism and acetylcholine production.<br />
ACITL and ACACT1r also correlate with choline acetyltransferase<br />
flux (P < 3 × 10 −90 ). Moreover, it has been reported that the inhibition<br />
of ACITL reduces the acetylcholine production rate by 30% 36 .<br />
Third singular vector<br />
0.4<br />
0.3<br />
0.2<br />
0.1<br />
a<br />
ASPGLUm<br />
b<br />
ACS<br />
ASPNATm<br />
GLUt2m NACASPtm<br />
NACASPAH<br />
AKGMALtm<br />
0<br />
ASPTAm<br />
ASPTA<br />
SUCOASm<br />
cHMGLm/HMGCOASim<br />
−0.1 CSm<br />
OCOAT1m ACACT1rm<br />
ACITL<br />
CITtam<br />
ACACT1r AACOAT<br />
ACACt2m<br />
−0.2<br />
−0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4<br />
Second singular vector<br />
accoa<br />
d e<br />
accoa<br />
f<br />
accoa Mitochondrion<br />
H2O oaa<br />
asp-L<br />
h coa H2O ACACT1rm<br />
asp-L<br />
ASPTAm<br />
coa<br />
CSm<br />
HMGCOASim<br />
aacoa<br />
ASPNATm<br />
atp<br />
coa glu-L akg<br />
hmgcoa<br />
succ coa<br />
coa<br />
cit h<br />
HMGLm OCOAT1m SUCOASm<br />
h glu-L<br />
akg<br />
acac<br />
adp<br />
Nacasp<br />
h CITtam<br />
mal-L<br />
h glu-L<br />
h<br />
pi<br />
succoa<br />
NACASPtm ASPGLUm h GLUt2m<br />
AKGMALtm<br />
h<br />
Nacasp<br />
ACACt2m<br />
mal-L<br />
h glu-L<br />
acac atp<br />
glu-L<br />
akg<br />
coa<br />
NACASPAH H2O<br />
ASPGLUm<br />
cit<br />
AACOAT<br />
asp-L<br />
coa<br />
amp<br />
ac atp<br />
glu-L akg<br />
ppi<br />
atp<br />
coa<br />
aacoa<br />
ACITL<br />
ACS<br />
adp<br />
coa<br />
amp<br />
ppi<br />
pi oaa<br />
ASPTA<br />
ACACT1r<br />
asp-L<br />
accoa<br />
accoa<br />
accoa<br />
Cytosol<br />
Figure 5 Singular value decomposition (SVD) of feasible pathways elucidates potential pathways that allow for coupling of mitochondria acetyl-CoA<br />
metabolism and cytosolic acetylcholine production. We computed 21,000 unique feasible reaction sets, each showing transport of mitochondrial acetyl-CoA<br />
carbon to the cytosol in human metabolism. (a–c) Dominant loading from the SVD of a matrix of all 21,000 pathways represented the reactions making<br />
up three primary pathways (d–f) that allow this coupling of mitochondrial metabolism to acetylcholine production, by carrying the acetyl-CoA carbon on<br />
N-acetyl-l-aspartate (a,d), citrate (b,e) or acetoacetate (c,f). As shown by the second singular vector, reactions in the pathway with citrate tend to be missing<br />
from pathways when the reactions for the acetoacetate pathway are included. The third singular vector shows a similar relationship of the N-acetyl-laspartate<br />
pathway. The ’omics data and known enzyme localization only support the usage of citrate and acetoacetate as potential carriers in neurons.<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1283
A n a ly s i s<br />
a<br />
Cholinergic percentage of brain<br />
5%<br />
4%<br />
3%<br />
2%<br />
1%<br />
[1− 14 C]<br />
[2− 14 C]<br />
Carbon label<br />
position<br />
b c d<br />
ChAT flux<br />
(µmol * g wet brain −1 * min −1 )<br />
4.0 × 10−3 × 10 −3 × 10 −3<br />
2.2<br />
Exp. data<br />
0 mM<br />
3.5 Simulation<br />
2.6<br />
3.0<br />
1.8<br />
2.2<br />
2.5<br />
0.1 mM<br />
2.0<br />
0.2 mM<br />
1.8<br />
1.4<br />
1.5<br />
0.3 mM<br />
1.0<br />
0.4 mM<br />
1.4<br />
1.0<br />
Exp. data<br />
0.5 1.0 mM<br />
Bromopyruvate 1.0<br />
0.6<br />
Simulation<br />
0.5 mM<br />
0<br />
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.42 0.46 0.50 0.54 0.58<br />
0.14 0.18 0.22 0.26 0.30<br />
14 CO 2 release (µmol * g wet brain −1 * min −1 )<br />
ChAT flux<br />
(µmol * g wet brain −1 * min −1 )<br />
2-oxo-3-methyl<br />
butanoate<br />
2-oxo-3-methyl<br />
pentanoic acid<br />
Leucine<br />
14 CO 2 release (µmol * g wet brain −1 * min −1 )<br />
ChAT flux<br />
(µmol * g wet brain −1 * min −1 )<br />
2-oxo-3-methyl<br />
pentanoic acid<br />
2-oxo-3-methyl<br />
butanoate<br />
Leucine<br />
14 CO 2 release (µmol * g wet brain −1 * min −1 )<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Figure 6 Model-aided prediction of cholinergic contribution is consistent with experimental acetylcholine production. Percent brain cholinergic<br />
neurotransmission was predicted based on 14 sets of experimental data in which brain minces were fed [1- 14 C]-pyruvate or [2- 14 C]-pyruvate, followed<br />
by measurement of 14 C-labeled CO 2 and acetylcholine. (a) For each experiment, the feasible amount of the brain that can generate the experimental<br />
response was computed, centering at 3.3%. (b) This parameter was used in the analysis, and the updated model predictions were consistent with<br />
experimental data, such as seen in the case of treating the brain minces with [1- 14 C]-pyruvate and increasing levels of the pyruvate-dehydrogenase<br />
inhibitor bromopyruvate. (c,d) Moreover, the updated model predictions were consistent with measured 14 C-labeled CO 2 and acetylcholine production<br />
for brain minces that were treated with three PDHm inhibitors withheld from previous computations for both supplementation with [1- 14 C]-pyruvate<br />
(c) and [2- 14 C]-pyruvate (d). Error bars on the simulation results represent 25 th and 75 th percentiles. ChAT, choline acetyltransferase.<br />
The in silico inhibition of ACITL reduces acetylcholine production<br />
by 7.3%. It is possible that the in silico decrease is smaller because the<br />
model can immediately adapt to the perturbation, whereas in vivo<br />
regulatory responses would take time to adapt. Notably, the in silico<br />
inhibition of ACACT1r reduces acetylcholine production by 39%.<br />
Thus, cholinergic neurotransmission depends on redundant pathways,<br />
and acetoacetate may play a more dominant role in transporting<br />
mitochondrial acetyl-CoA to the cytosol.<br />
Our analysis provides further insight into the coupling of mitochondrial<br />
metabolism to acetylcholine synthesis, which aids in the<br />
treatment of cholinergic disorders. However, knowledge of the abundance<br />
of cholinergic neurotransmission also aids in this purpose. It is<br />
difficult to identify cholinergic neurons, based solely on cell morphology,<br />
because cholinesterases and immunohistochemical markers for<br />
cholinergic neurons are also found in noncholinergic neurons and<br />
other tissues 37 . Therefore, it is unknown what percentage of all neurotransmission<br />
is cholinergic.<br />
Using our cholinergic model, we compute the percent contribution<br />
of cholinergic neurotransmission based on published data 38 . The<br />
data were obtained from rat brain minces, incubated in solutions<br />
containing [1- 14 C]pyruvate or [2- 14 C]pyruvate. Both acetylcholine<br />
and radiolabeled CO 2 were measured at various titrations of several<br />
pyruvate dehydrogenase inhibitors.<br />
The cholinergic model was subjected to similar levels of pyruvate<br />
dehydrogenase inhibition. The simulations successfully reproduced<br />
the experimental linear relationship between acetylcholine production<br />
and metabolic rate, and acetylcholine production was correlated<br />
with CO 2 release (r = 0.68).<br />
The fraction of cholinergic neurotransmission for the brain was<br />
computed by randomly choosing points from both the distributions<br />
of experimental data and distributions predicted by the simulations.<br />
A scaling factor was subsequently found that reconciles the two. This<br />
was repeated for 14 different combinations of pyruvate labeling and<br />
pyruvate dehydrogenase inhibitors 38 , yielding a median predicted<br />
cholinergic portion of total brain neurotransmission of 3.3% (Fig. 6a).<br />
After adding this new parameter to the model, the predictions correspond<br />
well with experimental data sets (Fig. 6b), including data<br />
representing three pyruvate dehydrogenase inhibitors withheld from<br />
the previous computations (Fig. 6c,d). Thus, the model was used in<br />
conjunction with experimental data to gain insight into physiological<br />
observations and derive physiological parameters, which are dependent<br />
on systems-level activity and are relevant to treatment.<br />
DISCUSSION<br />
In this study we presented a workflow for generating tissue-specific,<br />
multicellular metabolic models. Through the analysis and integration<br />
of ’omics data, followed by manual curation, we used this workflow to<br />
build a first-draft, manually curated, multicellular, metabolic reconstruction<br />
of brain energy metabolism. Three models were generated<br />
from this reconstruction, representing different types of neurons coupled<br />
to astrocytes. We used these models in three distinct analyses,<br />
in which we made predictions and gained systems-level insights into<br />
Alzheimer’s disease and cholinergic neurotransmission.<br />
As experimental methods and data resolution improve, the accuracy<br />
of these models and their predictions should also improve.<br />
Improvements in neuroimaging and metabolomics will allow for<br />
more precise quantification of metabolite flow through the bloodbrain<br />
barrier, which is of interest because dysfunction of this system<br />
accompanies many neurological disorders and injuries 39 . In addition,<br />
improvements in transcriptomics and proteomics will provide higherresolution<br />
quantification of cell- and organelle-specific genes and<br />
proteins. These data will allow models to account for neuron groups<br />
in specific brain regions, subcellular heterogeneity within cells and the<br />
inclusion of less abundant glial cells. For example, higher-resolution<br />
models may provide insight into metabolic changes in specific cell<br />
populations, such as the structures closely related to the olfactory<br />
system, which are affected early on in Alzheimer’s disease 40 .<br />
Insight into mammalian tissue-specific metabolism may be gained<br />
as more multicellular models are constructed. Our models demonstrate<br />
metabolic coupling and synergistic activities that more coarse-grained<br />
models miss, as the three analyses presented here were not possible<br />
using Recon 1 or the previous brain metabolism models. The compartmentalization<br />
of metabolic processes within cells 41 , between cells 42<br />
and in host-pathogen interactions 43 has an important role in normal<br />
physiology. Therefore, such models may provide greater insight and<br />
more accurately predict both true cellular functions and responses to<br />
medical interventions 44 . This insight will be attained as many analytical<br />
methods, in addition to those demonstrated in this work, will be<br />
1284 VOLUME 28 NUMBER 12 DECEMBER 2010 nature biotechnology
a n a ly s i s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
developed and deployed in multicellular reconstructions ranging from<br />
constraint-based analyses 5 to topological studies 45 .<br />
This study serves as an example of how mechanistic relationships between<br />
genotype and phenotype can be built through the difficult task of multi-omic<br />
data integration 46 . From the genotype one can begin to reconstruct the network<br />
for an organism. The integration of high-throughput data and careful<br />
manual curation can add context-specific mechanistic network structure<br />
to genomics information. Thus, this network becomes a representation<br />
of complex genetic interactions and biochemical mechanisms underlying<br />
observed phenotypes. This complex, but mechanistic, relationship between<br />
the genotype and phenotype can be used as a foundational structure upon<br />
which additional high-throughput data can be analyzed and predictive<br />
simulations can be conducted, thus leading to improved understanding,<br />
testable hypotheses and increased knowledge 1,2,44 .<br />
Methods<br />
Methods and any associated references are available in the online version<br />
of the paper at http://www.nature.com/naturebiotechnology/.<br />
Note: Supplementary information is available on the Nature Biotechnology website.<br />
Acknowledgments<br />
The authors thank G. Gibson at Cornell University, I. Thiele at the University of<br />
Iceland and M. Abrams, M. Mo and C. Barrett at UCSD for suggestions pertaining to<br />
this work. This work was funded in part by a Fulbright fellowship, a National Science<br />
Foundation IGERT Plant Systems Biology training grant (no. DGE-0504645),<br />
US National Institutes of Health grants 2R01GM068837_05A1 and RO1 GM071808<br />
and the Helmholtz Alliance on Systems Biology and the BMBF by the NGFN+<br />
neuroblastoma project ENGINE.<br />
AUTHOR CONTRIBUTIONS<br />
N.E.L., J.K.C., A.Y., N.P., M.P.A. and B.O.P. conceived and designed the model.<br />
N.E.L., J.K.C., G.S., R.K., R.E., J.S., A.B. and R.A.L. performed data analyses. The<br />
manuscript was written by N.E.L., G.S., J.S., A.B. and B.O.P.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
Published online at http://www.nature.com/naturebiotechnology/.<br />
Reprints and permissions information is available online at http://npg.nature.com/<br />
reprintsandpermissions/.<br />
1. Feist, A.M. & Palsson, B.O. The growing scope of applications of genome-scale metabolic<br />
reconstructions using Escherichia coli. Nat. Biotechnol. 26, 659–667 (2008).<br />
2. Oberhardt, M.A., Palsson, B.O. & Papin, J.A. Applications of genome-scale metabolic<br />
reconstructions. Mol. Syst. Biol. 5, 320 (2009).<br />
3. Breitling, R., Vitkup, D. & Barrett, M.P. New surveyor tools for charting microbial<br />
metabolic maps. Nat. Rev. Microbiol. 6, 156–161 (2008).<br />
4. Thiele, I. & Palsson, B.O. A protocol for generating a high-quality genome-scale<br />
metabolic reconstruction. Nat. Protoc. 5, 93–121 (2010).<br />
5. Lewis, N.E., Jamshidi, N., Thiele, I. & Palsson, B.O. Metabolic systems biology. in<br />
Encyclopedia of Complexity and Systems Science (ed. Meyers, R.A.) 5535–5552<br />
(Springer, New York, 2009).<br />
6. Orth, J.D., Thiele, I. & Palsson, B.O. What is flux balance analysis? Nat. Biotechnol.<br />
28, 245–248 (2010).<br />
7. Duarte, N.C. et al. Global reconstruction of the human metabolic network based on<br />
genomic and bibliomic data. Proc. Natl. Acad. Sci. USA 104, 1777–1782 (2007).<br />
8. Becker, S.A. & Palsson, B.O. Context-specific metabolic networks are consistent<br />
with experiments. PLOS Comput. Biol. 4, e1000082 (2008).<br />
9. Shlomi, T., Cabili, M.N., Herrgard, M.J., Palsson, B.O. & Ruppin, E. Network-based<br />
prediction of human tissue–specific metabolism. Nat. Biotechnol. 26, 1003–1010<br />
(2008).<br />
10. Jerby, L., Shlomi, T. & Ruppin, E. Computational reconstruction of tissue-specific<br />
metabolic models: application to human liver metabolism. Mol. Syst. Biol. 6, 401<br />
(2010).<br />
11. Ponten, F. et al. A global view of protein expression in human cells, tissues, and<br />
organs. Mol. Syst. Biol. 5, 337 (2009).<br />
12. Palsson, B.O. in Systems Biology: Properties of Reconstructed Networks 322<br />
(Cambridge University Press, Cambridge/New York, 2006).<br />
13. Mishra, G.R. et al. Human protein reference database–2006 update. Nucleic Acids<br />
Res. 34, D411–D414 (2006).<br />
14. Fujii, Y., Imanishi, T. & Gojobori, T. H-Invitational Database: integrated database<br />
of human genes. Tanpakushitsu Kakusan Koso 49, 1937–1943 (2004).<br />
15. Reidegeld, K.A. et al. The power of cooperative investigation: summary and<br />
comparison of the HUPO Brain Proteome Project pilot study results. Proteomics 6,<br />
4997–5014 (2006).<br />
16. Chatziioannou, A., Palaiologos, G. & Kolisis, F.N. Metabolic flux analysis as a tool<br />
for the elucidation of the metabolism of neurotransmitter glutamate. Metab. Eng.<br />
5, 201–210 (2003).<br />
17. Cakir, T., Alsan, S., Saybasili, H., Akin, A. & Ulgen, K.O. Reconstruction and flux<br />
analysis of coupling between metabolic pathways of astrocytes and neurons:<br />
application to cerebral hypoxia. Theor. Biol. Med. Model. 4, 48 (2007).<br />
18. Occhipinti, R., Puchowicz, M.A., LaManna, J.C., Somersalo, E. & Calvetti, D.<br />
Statistical analysis of metabolic pathways of brain metabolism at steady state. Ann.<br />
Biomed. Eng. 35, 886–902 (2007).<br />
19. Reiman, E.M. et al. Functional brain abnormalities in young adults at genetic risk for<br />
late-onset Alzheimer’s dementia. Proc. Natl. Acad. Sci. USA 101, 284–289 (2004).<br />
20. Ginsberg, S.D., Che, S., Counts, S.E. & Mufson, E.J. Single cell gene expression<br />
profiling in Alzheimer’s disease. NeuroRx 3, 302–318 (2006).<br />
21. Lai, M.K.P., Ramirez, M.J., Tsang, S.W.Y. & Francis, P.T. Alzheimer’s disease as a<br />
neurotransmitter disease in Neurobiology of Alzheimer’ Disease (eds. Dawbarn, D.<br />
& Allen, S.J.) 245–281 (Oxford University Press, New York, 2007).<br />
22. Fukui, H., Diaz, F., Garcia, S. & Moraes, C.T. Cytochrome c oxidase deficiency in<br />
neurons decreases both oxidative stress and amyloid formation in a mouse model<br />
of Alzheimer’s disease. Proc. Natl. Acad. Sci. USA 104, 14163–14168 (2007).<br />
23. Bubber, P., Haroutunian, V., Fisch, G., Blass, J.P. & Gibson, G.E. Mitochondrial<br />
abnormalities in Alzheimer brain: mechanistic implications. Ann. Neurol. 57,<br />
695–703 (2005).<br />
24. Gibson, G.E. et al. Alpha-ketoglutarate dehydrogenase in Alzheimer brains bearing<br />
the APP670/671 mutation. Ann. Neurol. 44, 676–681 (1998).<br />
25. Casley, C.S., Canevari, L., Land, J.M., Clark, J.B. & Sharpe, M.A. Beta-amyloid<br />
inhibits integrated mitochondrial respiration and key enzyme activities.<br />
J. Neurochem. 80, 91–100 (2002).<br />
26. Hoshi, M. et al. Regulation of mitochondrial pyruvate dehydrogenase activity by tau<br />
protein kinase I/glycogen synthase kinase 3beta in brain. Proc. Natl. Acad. Sci.<br />
USA 93, 2719–2723 (1996).<br />
27. Gorman, A.M., Ceccatelli, S. & Orrenius, S. Role of mitochondria in neuronal<br />
apoptosis. Dev. Neurosci. 22, 348–358 (2000).<br />
28. Santos, S.S. et al. Inhibitors of the alpha-ketoglutarate dehydrogenase complex<br />
alter [1–13C]glucose and [U-13C]glutamate metabolism in cerebellar granule<br />
neurons. J. Neurosci. Res. 83, 450–458 (2006).<br />
29. Hassel, B., Johannessen, C.U., Sonnewald, U. & Fonnum, F. Quantification of the<br />
GABA shunt and the importance of the GABA shunt versus the 2-oxoglutarate<br />
dehydrogenase pathway in GABAergic neurons. J. Neurochem. 71, 1511–1518 (1998).<br />
30. Liang, W.S. et al. Altered neuronal gene expression in brain regions differentially<br />
affected by Alzheimer’s disease: a reference data set. Physiol. Genomics 33, 240–256 (2008).<br />
31. Stuhmer, T., Anderson, S.A., Ekker, M. & Rubenstein, J.L. Ectopic expression of<br />
the Dlx genes induces glutamic acid decarboxylase and Dlx expression. Development<br />
129, 245–252 (2002).<br />
32. Ibanez, V. et al. Regional glucose metabolic abnormalities are not the result of<br />
atrophy in Alzheimer’s disease. Neurology 50, 1585–1593 (1998).<br />
33. Schramm, G. et al. PathWave: discovering patterns of differentially regulated<br />
enzymes in metabolic pathways. Bioinformatics 26, 1225–1231 (2010).<br />
34. Ragozzino, M.E., Pal, S.N., Unick, K., Stefani, M.R. & Gold, P.E. Modulation of<br />
hippocampal acetylcholine release and spontaneous alternation scores by<br />
intrahippocampal glucose injections. J. Neurosci. 18, 1595–1601 (1998).<br />
35. Watson, G.S. & Craft, S. Modulation of memory by insulin and glucose:<br />
neuropsychological observations in Alzheimer’s disease. Eur. J. Pharmacol. 490,<br />
97–113 (2004).<br />
36. Cooper, J.R. Unsolved problems in the cholinergic nervous system. J. Neurochem.<br />
63, 395–399 (1994).<br />
37. Karczmar, A.G. Cholinergic cells and pathways in Exploring the Vertebrate Cholinergic<br />
Nervous System 686 (Springer, New York, 2006).<br />
38. Gibson, G.E., Jope, R. & Blass, J.P. Decreased synthesis of acetylcholine<br />
accompanying impaired oxidation of pyruvic acid in rat brain minces. Biochem. J.<br />
148, 17–23 (1975).<br />
39. Abbott, N.J., Ronnback, L. & Hansson, E. Astrocyte-endothelial interactions at the<br />
blood-brain barrier. Nat. Rev. Neurosci. 7, 41–53 (2006).<br />
40. Thompson, M.D., Knee, K. & Golden, C.J. Olfaction in persons with Alzheimer’s<br />
disease. Neuropsychol. Rev. 8, 11–23 (1998).<br />
41. Schryer, D.W., Peterson, P., Paalme, T. & Vendelin, M. Bidirectionality and<br />
compartmentation of metabolic fluxes are revealed in the dynamics of isotopomer<br />
networks. Int. J. Mol. Sci. 10, 1697–1718 (2009).<br />
42. Serres, S., Raffard, G., Franconi, J.M. & Merle, M. Close coupling between astrocytic<br />
and neuronal metabolisms to fulfill anaplerotic and energy needs in the rat brain.<br />
J. Cereb. Blood Flow Metab. 28, 712–724 (2008).<br />
43. Bordbar, A. et al. Insight into human alveolar macrophage and M. tuberculosis<br />
interactions via metabolic reconstructions. Mol. Syst. Biol. 6, 422 (2010).<br />
44. Chang, R.L., Xie, L., Xie, L., Bourne, P.E. & Palsson, B.O. Drug off-target effects<br />
predicted using structural analysis in the context of a metabolic network model.<br />
PLoS. Comput. Biol. 6, e1000938 (2010).<br />
45. Lee, D.S. et al. The implications of human metabolic network topology for disease<br />
comorbidity. Proc. Natl. Acad. Sci. USA 105, 9880–9885 (2008).<br />
46. Palsson, B.O. & Zengler, K. The challenges of integrating multi-omics data sets.<br />
Nat. Chem. Biol. 6, 787–789 (2010).<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1285
© 2010 Nature America, Inc. All rights reserved.<br />
ONLINE METHODS<br />
Reconstruction of iNL403. This work focuses on the core of cerebral energy<br />
metabolism and the pathways that play a critical role in cell type–specific<br />
functions in the brain. The pathways in this work include mitochondrial metabolic<br />
pathways, central metabolic pathways closely tied to mitochondrial<br />
function and additional pathways that are needed for modeling neuron and<br />
astrocyte functions. To reconstruct these pathways, a list of known human<br />
mitochondrial, glycolytic and transport reactions were extracted from the<br />
manually curated human metabolic reconstruction Recon 1 (ref. 7). From<br />
this list, reactions were directly added to the brain reconstruction if brainlocalized<br />
protein or gene expression was suggested by the Human Protein<br />
Reference Database (release 5) 13 or HINV 14 , both of which provide tissue<br />
expression presence calls for each gene. Proteomics data from live human<br />
brain, acquired for the HUPO brain proteome project 15 , were also used (see<br />
Supplementary Table 7 for accession numbers). Additional reactions were<br />
added as dictated by biochemical data from the literature (Supplementary<br />
Table 1). This reconstruction is designated iNL403 because it contains<br />
403 genes. Reactions and pathways were manually curated to verify their presence<br />
in the human brain and to determine cell-type localization, thus yielding<br />
a first-draft metabolic reconstruction of the brain metabolic network.<br />
Reactions unique to the different neuron types were determined from the<br />
literature (see notes in Supplementary Table 1), and consist largely of the<br />
reactions needed to make and metabolize their associated neurotransmitters.<br />
A list of all reactions, supporting data, citations and a comparison with previous<br />
brain metabolism models can be found in Supplementary Tables 1–4 and 8.<br />
Three SBML models representing the union of reactions from the neuron-type<br />
specific models can be found as Supplementary Models 1–3. Neuron-type<br />
specific models in SBML format and model updates can be obtained from<br />
http://systemsbiology.ucsd.edu/In_Silico_Organisms/Brain.<br />
Constraint-based modeling. Constraint-based modeling and analysis of<br />
metabolic networks have been previously described 5,12 . Briefly, all of the<br />
reactions are described mathematically by a stoichiometric matrix, S, of size<br />
m × n, where m is the number of metabolites and n is the number of reactions,<br />
and each element is the stoichiometric coefficient of the metabolite in<br />
the corresponding reaction. The mass balance equations at steady state are<br />
represented as<br />
S • v = 0,<br />
where v is the flux vector 12 . Maximum and minimum fluxes and reaction<br />
reversibility, when known, are placed on each reaction, further constraining<br />
the system as follows<br />
vmin ≤ v ≤ vmax .<br />
(1)<br />
(2)<br />
At this point the model can then be used with many constraint-based methods 5<br />
to study network characteristics.<br />
The S matrix was constructed with the mass and charge balanced reactions<br />
from the reconstruction. Select metabolites, known to cross the blood-brain<br />
barrier, were added as exchange reactions, allowing those metabolites to leave<br />
or enter the extracellular space in the model. A few metabolites from network<br />
gaps were allowed to enter or leave the system from the cytosol or mitochondria.<br />
This was only done when transporter mechanisms or subsequent pathway<br />
steps were not known, and when their entrance or removal from the system<br />
was necessary for model function. When available, cerebral metabolic rates<br />
were used from published data to constrain the upper and lower bounds of the<br />
exchange reactions 45,46 . All parameters are detailed in Supplementary Tables 1<br />
and 5. Parameters derived from rat data were varied to demonstrate that results<br />
presented here were robust and therefore relevant to human metabolism, and<br />
constraint-based methods were further employed to assess and compare the<br />
model to metabolic functions of the brain (Supplementary Notes).<br />
Monte Carlo sampling. Monte Carlo sampling was used to generate a set of<br />
feasible flux distributions (points). The method is based on the artificially<br />
centered hit-and-run algorithm with slight modifications. Initially, a set of<br />
nonuniform pseudo-random points, called warm-up points, is generated. In a<br />
series of iterations, each point is randomly moved, always remaining within the<br />
feasible flux space. This is done by (i) choosing a random direction, (ii) computing<br />
the limits of how far one can travel in that direction and (iii) choosing<br />
a new random point along this line. After many iterations, the set of points is<br />
mixed and approaches a uniform sample of the solution space, thus providing<br />
a distribution for each reaction that represents the range and probability of<br />
the flux for each reaction, given the network topology and model constraints.<br />
For more detail, see the Supplementary Notes.<br />
Simulating enzyme deficiencies. Enzyme deficiencies were obtained from the<br />
literature 23 . To simulate each deficiency, the distribution for all candidate flux<br />
states was determined using Monte Carlo sampling. From this distribution, the<br />
most probable flux was found and the reaction upper bound was reduced by the<br />
fraction reported in the literature. All candidate states were then recomputed<br />
and compared with normal candidate flux states.<br />
Alzheimer’s disease microarray analysis. Microarrays were obtained from<br />
the Gene Expression Omnibus (GSE5281). Arrays consist of 161 Affymetrix<br />
Human Genome U133 Plus 2.0 Arrays that profile the gene expression from<br />
laser-capture, microdissected, histopathologically normal neurons from six<br />
different brain regions of Alzheimer’s disease patients and age-matched controls.<br />
These arrays were not used in model construction.<br />
Arrays were normalized using the GC Robust Multi-array Average (gcrma)<br />
normalization function in the bioconductor package for R. Pearson’s correlation<br />
coefficients were computed for all array pairs, and arrays with r < 0.8 were<br />
discarded (that is, GSM119643, GSM119661, GSM119666 and GSM119676).<br />
Different arrays had different levels of glial contamination. Therefore, to<br />
assess the amount of GAD (neuron-specific), the GAD1 and GAD2 levels on<br />
each array were normalized as follows. For each array, the relative amount of<br />
neuron material was determined by computing a ratio for four neuron-specific<br />
genes to the median level across all arrays. Neuron-specific genes were chosen to<br />
represent different neuron parts, including the soma, axon and synaptic bouton<br />
(TUBB3 (ref. 47), NeuN 48 , SYN1 (ref. 49) and ACTL6B 50 ). These were summed<br />
to compute a relative amount of neuron material (NM) for each array, j,<br />
gi j<br />
NM j = ∑ , ,<br />
g<br />
i i<br />
(3)<br />
for each neuron marker gene g i . Because GAD genes are neuron specific in the<br />
central nervous system, these were normalized for each array by the associated<br />
relative amount of neuron material, thus termed GAD NMN for neuron-marker<br />
normalized GAD. It is assumed in this study that the four neuron marker<br />
transcripts used here do not change their expression level between Alzheimer’s<br />
patients and age-matched controls, in the cell populations sampled for microarray<br />
analysis. This assumption is made since there are no published studies<br />
that demonstrate that these genes change expression in healthy cells through<br />
the progression of Alzheimer’s disease and because efforts were made to only<br />
expression profile histopathologically normal neurons. It is possible that there<br />
is downregulation of some neuron marker transcripts among neurons bearing<br />
neurofibrillary tangles, as synapse loss is a hallmark of Alzheimer’s disease 20 .<br />
However, the arrays used in this study profile histopathologically normal neurons<br />
and the surrounding glial cells. Therefore, it is not expected that there will<br />
be significant changes in the expression of these key neuronal genes in the data<br />
used here. Lastly, the inclusion of multiple genes from different cell regions<br />
aims to minimize the effects from expression changes not attributable to glial<br />
cell contamination. The results presented in this work are robust to the removal<br />
of each neuron marker gene (Supplementary Notes and Fig. 4 therein).<br />
PathWave analysis. PathWave allows for the elucidation of pathways that<br />
significantly change together. Its advantage over other methods, such as Gene<br />
Set Enrichment Analysis, is that it takes metabolic network connectivity into<br />
account to identify changes in pathways, thereby identifying significantly<br />
differentially regulated pathways in which the gene products are mechanistically<br />
connected.<br />
PathWave was used as published previously 33 . The reactions in each model<br />
were subdivided into biologically relevant functional pathways. Reactions that<br />
nature biotechnology<br />
doi:10.1038/nbt.1711
© 2010 Nature America, Inc. All rights reserved.<br />
were involved in multiple pathways were added to each associated pathway.<br />
Reactions and their pathways are shown in Supplementary Table 9.<br />
PathWave analyzes microarray data using metabolic reaction connectivity.<br />
For each model, pathways were simplified by removing all metabolites<br />
with connectivity >8 in the metabolic network. Exceptions are listed in<br />
Supplementary Table 10.<br />
For each of these simplified metabolic pathways from each model, reactions<br />
were laid into a two-dimensional, regular square lattice grid. To optimally preserve<br />
neighborhood relations of the reactions, adjacent nodes of the network<br />
were placed onto the grid as close to each other as possible. We mapped each<br />
expression data set, obtained from the Gene Expression Omnibus (GSE5281),<br />
onto the corresponding reactions of the transcribed enzymes. If a reaction<br />
was catalyzed by a complex of proteins, the average expression was taken. The<br />
resulting expression values of each reaction were z-transformed. Haar wavelet<br />
transforms on the optimized grid representation of each pathway were performed<br />
to explore every possible expression pattern of neighboring reactions<br />
and to define groups of reactions within a pathway that showed significant<br />
differences between samples of different conditions.<br />
To obtain significance values, the sample labels were permutated (n = 10,000)<br />
and scores were calculated for each wavelet and permutation. The scores represent<br />
the absolute value of the logarithm of the P-value for each wavelet feature,<br />
calculated by t-tests. For the best hit (highest score of the nonpermutated wavelet<br />
features) a P-value was obtained from the reference distributions and represented<br />
the significance for the corresponding pathway. The P-value for each pathway<br />
was corrected for multiple testing (false-discovery rate (FDR) = 0.05) 51 . Only<br />
pathways with more than three significantly differentially regulated reactions were<br />
further considered (FDR = 0.05). To obtain local patterns in the pathways, all<br />
wavelet features were statistically tested applying t-tests and corrected for multiple<br />
testing. Statistically significant features contained those subgraphs of the metabolic<br />
network that showed differentially regulated patterns. Reconstructing these<br />
subgraphs allowed us to directly detect the regions of interest in the metabolic<br />
network (Supplementary Table 11).<br />
Identifying pathways for acetylcholine synthesis. A flux balance analysisderived<br />
approach was used to identify all possible pathways coupling the mitochondrial<br />
and cytosolic acetyl-CoA pools using known reactions in human<br />
metabolism. First, the potential pathways were identified using Recon 1<br />
(ref. 7). A reaction that supplies mitochondrial acetyl-CoA was added to the<br />
model. A second reaction was added to remove cytosolic acetyl-CoA from the<br />
model. Lastly, all other metabolite uptake and secretion constraints were opened.<br />
Reactions were randomly removed until a minimum pathway was identified,<br />
capable of carrying flux between mitochondrial and cytosolic acetyl-CoA.<br />
This was repeated until >21,000 unique sets of reactions were identified. An<br />
r × p binary matrix was then built with the p unique reaction sets consisting of<br />
r reactions. Each element (i,j) of this matrix was 0 if reaction i was absent from<br />
pathway j or 1 if reaction i was in pathway j. Rows for all reactions that were<br />
never necessary were subsequently removed from the matrix. Singular value<br />
decomposition was then used, followed by varimax factor rotation of the first<br />
five singular vectors. Singular vector loadings demonstrated the dominant sets<br />
of reactions, and their major dependencies, that could be used to couple mitochondrial<br />
acetyl-CoA metabolism and cytosolic acetylcholine metabolism.<br />
Predicting cholinergic neurotransmission. The percentage cholinergic neurotransmission<br />
was computed based on published data 38 . The previously published<br />
data were obtained from rat brain minces that were incubated in solutions containing<br />
[1- 14 C]pyruvate or [2- 14 C]pyruvate. Both acetylcholine and radiolabeled<br />
CO 2 were measured at various titrations of several different inhibitors of pyruvate<br />
dehydrogenase (PDHm) (see Supplementary Notes for all inhibitors).<br />
Simulations were conducted using the cholinergic model. The models were<br />
allowed to take up the same substrates provided experimentally 38 , at rates<br />
consistent with the data (Supplementary Table 5). Monte Carlo sampling<br />
was used to identify all feasible flux states. This was done for various levels of<br />
PDHm inhibition, ranging from 0 to 90% inhibition. The percentage cholinergic<br />
neurotransmission was computed by randomly selecting a feasible flux<br />
state from each level of PDHm inhibition and computing the slope of the<br />
sum of labeled CO 2 -producing fluxes and choline acetyltransferase for the<br />
different simulations. A similar slope was computed from randomly sampled<br />
points from the reported experimental distributions. The ratio of these<br />
slopes represents a feasible percentage cholinergic neurotransmission. This<br />
was repeated 1,000 times and the median value was reported. Comparisons<br />
with the experimental data were done by suppressing the in silico pyruvate<br />
dehydrogenase flux until the measured CO 2 release rate was obtained. At this<br />
level of suppression, the resulting predicted acetylcholine production rate was<br />
compared with the experimentally measured rates. See Supplementary Notes<br />
for more details.<br />
47. Lying-Tunell, U., Lindblad, B.S., Malmlund, H.O. & Persson, B. Cerebral blood flow<br />
and metabolic rate of oxygen, glucose, lactate, pyruvate, ketone bodies and amino<br />
acids. Acta Neurol. Scand. 62, 265–275 (1980).<br />
48. Lying-Tunell, U., Lindblad, B.S., Malmlund, H.O. & Persson, B. Cerebral blood flow<br />
and metabolic rate of oxygen, glucose, lactate, pyruvate, ketone bodies and amino<br />
acids. Acta Neurol. Scand. 63, 337–350 (1981).<br />
49. Tischfield, M.A. et al. Human TUBB3 mutations perturb microtubule dynamics,<br />
kinesin interactions, and axon guidance. Cell 140, 74–87 (2010).<br />
50. Kim, K.K., Adelstein, R.S. & Kawamoto, S. Identification of neuronal nuclei (NeuN)<br />
as Fox-3, a new member of the Fox-1 gene family of splicing factors. J. Biol. Chem.<br />
284, 31052–31061 (2009).<br />
51. De Camilli, P., Cameron, R. & Greengard, P. Synapsin I (protein I), a nerve terminalspecific<br />
phosphoprotein. I. Its general distribution in synapses of the central and<br />
peripheral nervous system demonstrated by immunofluorescence in frozen and<br />
plastic sections. J. Cell Biol. 96, 1337–1354 (1983).<br />
52. Olave, I., Wang, W., Xue, Y., Kuo, A. & Crabtree, G.R. Identification of a polymorphic,<br />
neuron-specific chromatin remodeling complex. Genes Dev. 16, 2509–2517 (2002).<br />
53. Benjamini, Y. & Yekutieli, D. The control of the false discovery rate in multiple<br />
testing under dependency. Ann. Stat. 29, 1165–1188 (2001).<br />
doi:10.1038/nbt.1711<br />
nature biotechnology
B r i e f c o m m u n i c at i o n s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
A robust system for production<br />
of minicircle DNA vectors<br />
Mark A Kay 1,2 , Cheng-Yi He 1,2 & Zhi-Ying Chen 1,2<br />
Minicircle DNA vectors allow sustained transgene expression in<br />
quiescent cells and tissues. To improve minicircle production,<br />
we genetically modified Escherichia coli to construct a<br />
producer strain that stably expresses a set of inducible<br />
minicircle-assembly enzymes, ΦC31 integrase and I-SceI<br />
homing endonuclease. This bacterial strain produces purified<br />
minicircles in a time frame and quantity similar to those of<br />
routine plasmid DNA preparation, making it feasible to use<br />
minicircles in place of plasmids in mammalian transgene<br />
expression studies.<br />
A minicircle episomal DNA vector is a circular expression cassette<br />
devoid of the bacterial plasmid DNA backbone. These vectors have<br />
been used for years in preclinical gene transfer research because of<br />
their 10- to 1,000-fold enhancement compared with regular plasmids<br />
in long-term transgene expression in quiescent tissues in vivo 1,2 and<br />
in vitro 3 . The mechanism of enhanced transgene expression is unclear<br />
but may result from eliminating heterochromatin formation induced<br />
by the plasmid backbone 4 and/or from reducing the death of transfected<br />
cells from inflammation due to CpG responses when plasmids<br />
are delivered by lipid carriers 5 .<br />
The major obstacle to widespread use of minicircles has been their<br />
time-consuming, labor-intensive production. In our previous minicircle<br />
production schemes (Fig. 1a), the minicircle producer plasmid<br />
contained a transgene expression cassette flanked with attB and attP,<br />
a set of inducible enzyme genes (a gene encoding homing endonuclease<br />
I-SceI and two copies of the gene encoding ΦC31 integrase) and<br />
an I-SceI recognition site 6 . The attB and attP sites are the bacterial<br />
and phage attachment sites of ΦC31 integrase, and the ΦC31 and<br />
I-SceI genes are regulated by the l-arabinose–inducible araCBAD<br />
system. Minicircle DNA is generated by recombination between the attB<br />
and attP sites, and I-SceI initiates the destruction of the plasmid DNA<br />
backbone circle by cutting through the engineered I-SceI site (Fig. 1a).<br />
Although the yields from this protocol were ~1 mg of minicircle<br />
DNA from 1 liter of overnight culture, the preparations still contained<br />
~3–15% of the input minicircle producer plasmid plus the<br />
plasmid backbone circle as contaminants. Including CsCl equilibrium<br />
gradient centrifugation to remove these unwanted DNAs,<br />
the production procedure is four labor-intensive days longer than<br />
routine plasmid production protocols. Other groups have made<br />
minicircle DNA vectors using different recombinases, such as the<br />
bacteriophage λ integrase 7 or Cre recombinase 8 . Limitations of these<br />
approaches include low yields, high contamination or the need for<br />
expensive, labor-intensive techniques to isolate minicircles from<br />
bacterial lysates 9 .<br />
We present a system that allows simple, rapid and inexpensive production<br />
of a high-quality form of minicircles (Fig. 1b,c). We reasoned<br />
that the impurity plasmid DNAs in our earlier minicircle production<br />
schemes (Fig. 1a) resulted largely from an all-or-none phenomenon 10 ,<br />
such that in a subset of bacterial cells, the l-arabinose transporter<br />
AraE is not expressed, resulting in the failure of ΦC31 integrase–<br />
directed formation of minicircle and I-SceI–mediated degradation of<br />
the plasmid backbone circle. To overcome this limitation, we replaced<br />
the E. coli ‘Top10’ cells with the BW27783 bacterial strain 10 , in which<br />
the araE gene was driven by the constitutive promoter, cp8. When<br />
used in combination with previous minicircle producer plasmids,<br />
the conversion to minicircle DNA was more complete (Fig. 2) even<br />
with the addition of a very small amount (0.001%) of l-arabinose<br />
to the culture (Fig. 2a). There was variable but substantial plasmid<br />
DNA degradation in these preparations (Fig. 2a,c). We hypothesized<br />
this was due to the presence of endonuclease A, which was confirmed<br />
when we knocked the gene out (Fig. 2b,c), resulting in strain<br />
BWΔendA (Fig. 2b and Supplementary Fig. 1).<br />
Trace amounts of impurity DNAs became visible after knocking<br />
out the gene encoding endonuclease A (Fig. 2c), suggesting that<br />
there was still a lack of complete recombination after l-arabinose<br />
induction. Thus, we engineered the bacterial cells to simultaneously<br />
express a second l-arabinose transporter, LacY A177C 11 and created<br />
the strain 2T (Supplementary Figs. 1 and 2a–d). Its wild-type counterpart,<br />
LacY, is a glucose transporter, which gains an l-arabinose<br />
transporter function with the described missense mutation. This<br />
further reduced but did not completely eliminate the contaminating<br />
DNAs (data not shown).<br />
To eliminate any possible contamination of the genes encoding<br />
ΦC31 integrase and I-SceI from the small amount of parental plasmids<br />
used to generate minicircle DNAs, we relocated both genes from the<br />
minicircle producer plasmid to the bacterial genome. Accordingly, we<br />
integrated three copies of the BAD.I-SceI cassette using the bacteriophage<br />
λ Red homologous recombination system 12 and created the<br />
strain 3S2T (Supplementary Figs. 1 and 3a). We found that the integrated<br />
BAD.I-SceI was fully functional, as evident by the efficient<br />
destruction of a plasmid with eight consecutive I-SceI sites after<br />
l-arabinose induction (Supplementary Fig. 3b,c).<br />
Integration of the BAD.ΦC31 gene was less straightforward. First,<br />
the BAD.ΦC31 cassette could not be integrated using our original Red<br />
system without destroying the integrated BAD.I-SceI gene. Second, we<br />
attempted to integrate the BAD.ΦC31 cassette from a plasmid through<br />
ΦC31 integrase–mediated recombination between the attB site in the<br />
plasmid and attP pre-integrated into the genome. However, this integrant<br />
was unstable (data not shown), most likely because of a reverse<br />
1 Departments of Pediatrics and Genetics, Stanford University School of Medicine, Stanford, California, USA. 2 The Center for Clinical Science Research, Stanford<br />
University School of Medicine, Stanford, California, USA. Correspondence should be addressed to M.A.K. (markay@stanford.edu) or Z.-Y.C. (zychen98@stanford.edu).<br />
Received 21 June; accepted 15 October; published online 21 November 2010; doi:10.1038/nbt.1708<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1287
i e f c o m m u n i c at i o n s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Figure 1 Comparison of the present and previous<br />
minicircle systems. (a) An earlier version of the<br />
minicircle production system. (i) Structure of<br />
the previous minicircle producer plasmid. BAD<br />
and araC, the promoter and the repressor gene<br />
of the inducible l-arabinose-araC.BAD system;<br />
ΦC31, bacteriophage ΦC31 integrase gene; attB<br />
and attP, the bacterial and phage attachment<br />
sites of the ΦC31 integrase; I-SceI, I-SceI<br />
homing endonuclease gene; I-SceIs, the I-SceI<br />
recognition site; AmpR, ampicillin resistance<br />
gene; ColE1, DNA replication origin. (ii) E. coli<br />
strain Top10, original strain used to produce<br />
minicircle. (iii) Flow chart showing the minicircle<br />
production protocol. Each box represents a<br />
major step and the starred boxes represent the<br />
steps required in addition to a routine plasmid<br />
production protocol. (b) The present minicircle<br />
system. (i) Diagram of a minicircle producer<br />
plasmid and its conversion to minicircle DNA.<br />
pMC.ApoE.hFIX, minicircle producer plasmid;<br />
hFIX, human factor IX; sApoE, promoter/<br />
enhancer, as described previously 1 ; KanR,<br />
kanamycin resistance gene. Upon l-arabinose<br />
induction, ΦC31 is expressed to mediate the<br />
formation of minicircle and plasmid backbone<br />
circle and I-SceI to induce the destruction<br />
of plasmid backbone circle. (ii) The genetic<br />
Transgene attP<br />
p2ΦC31.Transgene<br />
2BADΦC31 araC AmpR ColE1 I-SceIs<br />
modifications of the minicircle producing bacterial strain ZYCY10P3S2T. 10P3S2T stands for (1) ten copies of BAD.ΦC31 cassette, that were integrated in<br />
three loci of the bacterial genome: two tandem copies at the ΔendA locus (Supplementary Fig. 4b), and four copies at the araD (Supplementary Fig. 5a) and<br />
galK (Supplementary Fig. 6a) each; (2) three tandem copies of BAD.I-SceI cassette, which were integrated at UMU locus (Supplementary Fig. 3a) and (3)<br />
two genes constitutively expressing l-arabinose transporter, one was araE gene driven by an artificial promoter cp8, which presented in strain BW27783<br />
(ref. 10); the other was the bla-lacY A177C cassette, which was integrated at the lacY locus (Supplementary Fig. 2); bla, beta-galactosidase gene promoter;<br />
lacY A177C, the missense mutant of lacY gene. (iii) Flow chart showing the present minicircle production protocol. (c) Stepwise genetic modification of the<br />
bacterial genome to make the current ZYCY10P3S2T strain.<br />
reaction mediated by leaky ΦC31 integrase expression in concert with<br />
an uncharacterized cofactor 13 . Third, because many more copies of<br />
the BAD.ΦC31 were needed to efficiently generate minicircles, we<br />
designed a strategy using ΦC31 integrase and a second recombinase.<br />
After making the desired integrant by the ΦC31 integrase–mediated<br />
reaction, the resulting attL site was removed by the second recombinase,<br />
eliminating the possibility of the reverse excision reaction<br />
(Supplementary Figs. 4a–c, 5a,e and 6a,d).<br />
Using the above approach, we generated a bacterial host containing<br />
two copies of the BAD.ΦC31 inserted into the deleted endA site<br />
(ΔendA), called the 2P3S2T strain (Supplementary Figs. 1 and 4a–c).<br />
We found that ΦC31 integrase mediated the formation of the minicircle,<br />
and I-SceI mediated destruction of the unrecombined minicircle<br />
producer plasmid and plasmid backbone circle in a minicircle producer<br />
plasmid encoding a 2.2-kb transgene and 32 tandem copies of<br />
the I-SceI site in the plasmid backbone (Supplementary Fig. 4d,e).<br />
However, the minicircle yield was low, suggesting that most of the<br />
minicircle producer plasmid was destroyed by I-SceI before minicircle<br />
formation, and that additional ΦC31 integrase activity was needed to<br />
mediate greater minicircle formation.<br />
We therefore used another targeting plasmid to integrate four copies<br />
of the BAD.ΦC31 cassette into the araD locus (Supplementary<br />
Figs. 1 and 5), and four additional copies into the gene encoding<br />
galK (Supplementary Figs. 1 and 6), resulting in strains 6P3S2T and<br />
ZYCY10P3S2T carrying six and ten copies of BAD.ΦC31, respectively.<br />
These two present strains, in concert with the minicircle producer<br />
plasmid described earlier (Supplementary Fig. 4d), enabled three<br />
improvements over the previous system (Fig. 1a) 6 . First, the procedure<br />
was greatly simplified and, compared to a routine plasmid preparation,<br />
required only an additional temperature change and 5-h incubation<br />
a<br />
b<br />
(i)<br />
(iii)<br />
(i)<br />
(ii)<br />
(iii)<br />
attB<br />
Overnight<br />
culture<br />
attB<br />
Restrict<br />
contaminated<br />
plasmid<br />
backbone circle<br />
Overnight<br />
culture<br />
Pellet the<br />
bacteria<br />
via centrifugation<br />
*<br />
ApoE.hFIX attP<br />
pMC.ApoE.hFIX<br />
32I-SceIs KanR<br />
ColE1<br />
BADI-SceI<br />
* Resuspend * *<br />
bacteria in LB<br />
containing<br />
1% L-arabinose<br />
Isolate<br />
minicircle via<br />
CsCl gradient<br />
centrifugation<br />
Add minicircle<br />
induction mix<br />
comprising 1 volume<br />
of LB, 0.04 volume of<br />
1N NaOH and<br />
0.02% L-arabinose<br />
ApoE.hFIX<br />
*<br />
*<br />
MC.ApoE.hFIX +<br />
attR<br />
(ii)<br />
E. coli strain Top 10<br />
Incubate at<br />
32 °C, 250<br />
r.p.m. for 2 h<br />
Remove ethidium<br />
bromide via<br />
isopropane<br />
extraction<br />
The genetic modifications in bacterial strain ZYCY10P3S2T:<br />
Incubate<br />
at 32 °C/<br />
250 r.p.m.<br />
for 5 h<br />
Plasmid backbone circle<br />
32I-SceIs KanR ColE1<br />
*<br />
Isolate<br />
minicircle<br />
by plasmid<br />
purification<br />
column<br />
* *<br />
Isolate<br />
minicrcle by<br />
plasmid<br />
purification<br />
column<br />
Remove<br />
isopropane<br />
via ethanol<br />
precipitation<br />
Cp8.araE ∆endA Bla.lacY A177C 3BAD.I-SceI 2BAD.ΦC31 4BAD.ΦC31 4BAD.ΦC31<br />
attL<br />
Degradation<br />
after addition of l-arabinose (Fig. 1b). Second, the yield of three<br />
minicircles, with a size range of 2.2–6.0 kb, was 3.4–4.8 mg/1,000 ml<br />
of overnight culture, making it ~3- to 5-times higher (Fig. 2e) than our<br />
previous minicircle producing system (Fig. 1a). In addition, compared<br />
with the previous minicircle production protocol, there were 10-times<br />
fewer contaminating plasmid DNAs, ranging from 0.4% to 1.5% of the<br />
input minicircle preparation as determined by qPCR (Online Methods,<br />
qPCR section) 6 . On a molar scale, the yield of minicircle was 20–70%<br />
higher than the minicircle producer plasmid (Fig. 2e). Third, the cost<br />
of minicircle production was similar to that of a standard plasmid.<br />
These production improvements, together with their superior<br />
expression profiles, make it feasible for minicircle DNA vectors to be<br />
used in place of plasmid DNAs in mammalian expression studies.<br />
Our efforts to optimize minicircle production led us to develop an<br />
improved bacterial genome modification strategy. The enhancements<br />
include: (i) the use of a circular integrating plasmid instead of linear<br />
DNA, allowing repeated integration of the same or different DNA<br />
sequences of up to 10 kb in selected targets (Supplementary Figs. 4b,<br />
5a and 6a); (ii) inclusion of a second recombinase, allowing selective<br />
removal of unwanted sequences, such as the hybrid sequences responsible<br />
for the reverse recombinase reaction, along with the useless or<br />
harmful plasmid backbone DNAs, from the genome (Supplementary<br />
Figs. 4b, 5a and 6a); and (iii) the use of the TPin/9attB.9attP recombination<br />
system (Supplementary Figs. 5a and 6a), eliminating the inherent<br />
problems with the FLP/FRT system, in which the uncontrolled reaction<br />
between the substrate and product FRTs makes repeated integration<br />
virtually impossible. Although these bacterial gene modifications<br />
would be difficult to achieve with conventional homologous recombination<br />
methods, there are other technologies using dual site-specific<br />
recombinases for targeted integration of circular DNA into the bacterial<br />
c<br />
Strain BW27783<br />
Figure 2<br />
Knockout endA,<br />
resulting in strain<br />
BW∆endA<br />
Supplementary<br />
Figure 2<br />
Integrate lacY A177C,<br />
resulting in strain<br />
2T<br />
Supplementary<br />
Figure 3<br />
Integrate 3XBADI-Scel,<br />
resulting in strain<br />
3S2T<br />
Supplementary<br />
Figure 4<br />
Integrate 2xBADΦC31,<br />
resulting in strain<br />
2P3S2T<br />
Supplementary<br />
Figure 5<br />
Integrate 4xBADΦC31,<br />
resulting in strain<br />
6P3S2T<br />
Supplementary<br />
Figure 6<br />
Integrate 4xBADΦC31,<br />
resulting in strain<br />
10P3S2T<br />
1288 VOLUME 28 NUMBER 12 DECEMBER 2010 nature biotechnology
i e f c o m m u n i c at i o n s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Figure 2 Improvement in minicircle quality<br />
and quantity. (a) Strain BW27783 produced<br />
minicircle (MC) with enhanced purity. Minicircles<br />
were produced according to the protocol<br />
described previously 6 . 2Φ31.hFIX, minicircle<br />
producer plasmid (PP); BglII+EcoN1, two<br />
restriction enzymes used to cleave MC before<br />
electrophoresis; l-arab(%), percent of l-arabinose<br />
in the minicircle induction reaction; PB, plasmid<br />
backbone circle. (b) Strategy for inactivation<br />
of endA. pK a nR.endA, the plasmid used to<br />
generate the Pme1-restricted targeting DNA<br />
fragment; KanR, kanamycin-resistance gene; attB<br />
and attP, the bacterial and phage attachment<br />
sites of bacteriophage ΦC31 integrase; boxed<br />
endA, PCR-generated 329- and 754-bp end A<br />
fragments; pBAD.Red, a plasmid expressing the<br />
bacteriophage λ homology recombination complex<br />
(Red) under the control of araC.BAD (BAD);<br />
p2ΦC31, a complementing plasmid encoding<br />
two copies of BAD.ΦC31 gene, one copy of BAD.<br />
I-SceI gene and one I-SceI site; BWΔendA, a<br />
strain derived from BW27783 with the endA<br />
interrupted. (c) Minicircle DNA integrity before<br />
and after disruption of the endA gene. Before<br />
disrupting endA, we observed repeatedly large<br />
variations in the degree of plasmid degradation<br />
as shown in a and c. Because the endonuclease<br />
A is a membrane-bound enzyme, it was possible<br />
that its membrane release and activation varied<br />
during plasmid preparation. 32 °C and 37 °C, the<br />
incubation temperature. All reactions contained<br />
1% l-arabinose. (d) Quality of the minicircle<br />
determined by gel analyses. Minicircle was<br />
made according to the simplified protocol outlined in Figure 1b and Supplementary Protocol; DNAs were cleaved before electrophoresis. (e) Yield of<br />
minicircle producer plasmids and minicircle vector DNAs. The yield was derived from triplicate 400-ml overnight cultures; PP, minicircle producer plasmid;<br />
MC, minicircle. Wilcoxon rank sum test comparing the yield of minicircle and its minicircle producer plasmid: (I), P < 0.05; (II), P > 0.05. We used the<br />
following formula to convert the yield from mg/l to mol/l: mol/l = [yield (mg/l) × 10E-3 g/l]/[size (kb) × 1,000 × 330 × 2 g/mol], where 330 is the average<br />
molecular weight of dNTP. The minicircle producer plasmid, pMC.RSV.hAAT is schematically illustrated in Supplementary Figure 4d; the pMC.CMV.LGNSO<br />
is described in the Constructs section of Online Methods.<br />
genome and creation of a marker-less bacterial strain 14 . However, in our<br />
approach we were able to remove one of the two recombination hybrids<br />
(either attL or attR; Supplementary Figs. 5 and 6), disabling the reverse<br />
reaction and resulting in a stable genomic insertion. Recent studies<br />
have shown that repeated gene sequences inserted into the bacterial<br />
genome are stable for at least 80 generations 15 . This type of site-specific<br />
integration will broaden the ability to make stable genetic modifications<br />
in prokaryotic and eukaryotic genomes.<br />
Methods<br />
Methods and any associated references are available in the online<br />
version of the paper at http://www.nature.com/naturebiotechnology/.<br />
Note: Supplementary information is available on the Nature Biotechnology website.<br />
Acknowledgments<br />
The authors would like to thank P. Valdmanis for critical review of the manuscript.<br />
This work was supported by the US National Institutes of Health - HL064274<br />
(M.A.K.). Bacterial strain BW27783 was a gift of Jay D. Keasling of the University<br />
of California at Berkeley. Plasmid placY A177C was obtained from John E. Cronan<br />
at the University of Illinois.<br />
AUTHOR CONTRIBUTIONS<br />
Z.-Y.C. planned and carried out the experiments. C.-Y.H. conducted a number<br />
of the experiments. M.A.K. and Z.-Y.C. discussed and planned the experimental<br />
strategies. M.A.K. and Z.-Y.C. wrote the manuscript.<br />
a<br />
c<br />
e<br />
Parental p2Φ31.hFIX<br />
Enzyme Bgl II+EcoNI<br />
Strain Top10<br />
BW27783<br />
L-arab (%) 0 1.0 0.1 0.01 0.001 0 1.0 0.1 0.01 0.001<br />
Plasmid p2Φ31.hFIX<br />
Enzyme Bgl II + EcoNI<br />
Strain BW∆endA1 BW27783<br />
32 °C (h) 0 2 2 0 2<br />
37 °C (h) 0 0 2 0 2<br />
PP<br />
PB<br />
MC<br />
d Producer pMC<br />
pMC.<br />
pMC.<br />
Plasmid RSV.hAAT ApoE.hFIX CMV.LGNSO<br />
DNA type PP MC PP MC PP MC<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare no competing financial interests.<br />
b<br />
ColE1 AmpR F1<br />
pK a nR.endA<br />
endA attB KanR attP endA<br />
PP<br />
PB Pme I<br />
Pme 1- Pme I<br />
MC<br />
cutting<br />
endA attB KanR attP endA<br />
+<br />
EndA Strain BW27783<br />
pBAD.Red<br />
endA attB KanR attP endA<br />
p2øC31<br />
Targeting<br />
DNA fragment<br />
Strain<br />
BW∆endA.KanR<br />
endA attR endA Strain BW∆endA<br />
Size (kb) Yield (mg/liter) Weight Yield (E-9 mole/liter) Molar<br />
Minicircle producer<br />
ratio<br />
ratio<br />
plasmid<br />
PP MC PP<br />
MC (MC/PP) PP MC (MC/PP)<br />
pMC.RSV.hAAT 6.1 2.2 5.84 ± 0.41 3.42 ± 0.87(I) 0.59 1.60 ± 0.11 2.59 ± 0.66(I) 1.63<br />
pMC.ApoE.hFIX 8.1 4.2 5.45 ± 0.38 4.83 ± 0.60(II) 0.89 1.12 ± 0.08 1.91 ± 0.24(I) 1.71<br />
pMC.CMV.LGNSO 9.8 6.0 5.57 ± 0.15 3.40 ± 0.35(I) 0.61 0.95 ± 0.03 1.07 ± 0.11(II) 1.18<br />
Published online at http://www.nature.com/naturebiotechnology/.<br />
Reprints and permissions information is available online at http://npg.nature.com/<br />
reprintsandpermissions/.<br />
1. Chen, Z.Y., He, C.Y., Ehrhardt, A. & Kay, M.A. Mol. Ther. 8, 495–500 (2003).<br />
2. Huang, M. et al. Circulation 120, S230–S237 (2009).<br />
3. Jia, F.J. et al. Nat. Methods 7, 197–199 (2010).<br />
4. Riu, E., Chen, Z.Y., Xu, H., He, C.Y. & Kay, M.A. Mol. Ther. 15, 1348–1355<br />
(2007).<br />
5. Tan, Y., Li, S., Pitt, B.R. & Huang, L. Hum. Gene Ther. 10, 2153–2161<br />
(1999).<br />
6. Chen, Z.Y., He, C.Y. & Kay, M.A. Hum. Gene Ther. 16, 126–131 (2005).<br />
7. Darquet, A.M., Cameron, B., Wils, P., Scherman, D. & Crouzet, J. Gene Ther. 4,<br />
1341–1349 (1997).<br />
8. Bigger, B.W. et al. J. Biol. Chem. 276, 23018–23027 (2001).<br />
9. Mayrhofer, P., Blaesen, M., Schleef, M. & Jechlinger, W. J. Gene Med. 10,<br />
1253–1269 (2008).<br />
10. Khlebnikov, A., Datsenko, K.A., Skaug, T., Wanner, B.L. & Keasling, J.D. Microbiology<br />
147, 3241–3247 (2001).<br />
11. Morgan-Kiss, R.M., Wadler, C. & Cronan, J.E. Jr. Proc. Natl. Acad. Sci. USA 99,<br />
7373–7377 (2002).<br />
12. Yu, D. et al. Proc. Natl. Acad. Sci. USA 97, 5978–5983 (2000).<br />
13. Thorpe, H.M. & Smith, M.C. Proc. Natl. Acad. Sci. USA 95, 5505–5510<br />
(1998).<br />
14. Minaeva, N.I. et al. BMC Biotechnol. 8, 63 (2008).<br />
15. Sallam, K.I., Tamura, N., Imoto, N. & Tamura, T. Appl. Environ. Microbiol. 76,<br />
2531–2539 (2010).<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1289
© 2010 Nature America, Inc. All rights reserved.<br />
ONLINE METHODS<br />
Materials. Bacterial strain BW27783 was a gift of J.D. Keasling 10 . Luria-<br />
Bertani broth (LB) powder was purchased from MP Biomedicals and Terrific<br />
Broth (TB) powder from Invitrogen. Plasmid DNA purification kits were<br />
from Qiagen.<br />
Constructs. Plasmid pK a nR.endA (Fig. 2b and Supplementary Glossary) was<br />
made by inserting the following DNA elements: the kanamycin-resistance gene<br />
(KanR) flanked by the bacterial (attB) and phage (attP) attachment sites of the<br />
Streptomyces bacteriophage ΦC31 integrase and PCR-generated 329- and 734-bp<br />
fragments of endonuclease A gene (endA) into the pBlueScript (Stratagene)<br />
plasmid. The plasmid p2ΦC31 (Fig. 2b and Supplementary Figs. 1, 2b, 3a<br />
and Supplementary Glossary) was made by removing the human factor IX<br />
(hFIX) transgene and flanking attB and attP sites from p2ΦC31.hFIX 6 ,<br />
the BAD.I-SceI cassette and I-SceI site were retained. To make the plasmid<br />
p3BAD.I-SceI (Supplementary Fig. 3a and Supplementary Glossary),<br />
the araC repressor gene together with three tandem copies of BAD.I-SceI gene<br />
were inserted downstream of the attB site of the plasmid pK a nR.endA (Fig. 2b<br />
and Supplementary Glossary), followed by replacement of endA with the<br />
737- and 647-bp UMU fragment generated by PCR. The plasmid pBS.8I-SceIs<br />
(Supplementary Fig. 3b and Supplementary Glossary) was constructed by<br />
inserting eight consecutive copies of the 18-bp I-SceI site into pBlueScript. The<br />
plasmids pBAD.Red (Fig. 2b and Supplementary Figs. 1, 2a,c, 3a, 4a, 5a and<br />
6a and Supplementary Glossary) 16 and pcI587.FLP (Supplementary Figs.<br />
1, 4a,b and Supplementary Glossary) 17 , which both carry the temperaturesensitive<br />
A101 origin of replication, were obtained from the E. coli Genetic<br />
Resource Center of Yale University. Plasmid pFRT.KanR.attB (Supplementary<br />
Figs. 1, 4a and Supplementary Glossary) was generated by inserting the attB<br />
site and the KanR flanked with FRT sites of flipase into pBlueScript using DNA<br />
oligonucleotides. Plasmid p2ΦC31.attP.FRT (Supplementary Figs. 1, 4b and<br />
Supplementary Glossary) was constructed using our previous minicircleproducing<br />
plasmid p2ΦC31.hFIX 6 as starting material. The sApoE.hFIX and<br />
BAD.I-SceI cassettes and the I-SceI site were eliminated, the ColE1 origin<br />
and the AmpR were replaced with the KanR, the FRT and attP sites and the<br />
plasmid replication origin R6K plus zeocin-resistance gene (Zeo) derived from<br />
the plasmid pCpG-mcs (InvivoGen). The plasmid pMC.ApoE.hFIX (Figs. 1b,<br />
2d,e and Supplementary Glossary) was made by stepwise replacement of the<br />
AmpR and the F1 origin in the plasmid pBS.8I-SceI (Supplementary Fig. 3b<br />
and Supplementary Glossary) with KanR and sApoE.hFIX and the flanking<br />
attB and attP sites from the plasmid p2ΦC31.hFIX 6 , followed by the insertion<br />
of an additional 24 consecutive I-SceI recognition sites. The plasmid<br />
pMC.CMV.LGNSO (Fig. 2d,e) was made by replacing the attB.RSV.hAAT.attP<br />
fragment in the plasmid pMC.RSV.hAAT (Fig. 2d,e, Supplementary Fig. 4d<br />
and Supplementary Glossary) with the fragment flanked with the attB and<br />
attP derived from the plasmid p2ΦC31.LGNSO, as described previously 3 . The<br />
plasmid placY.TetR (Supplementary Figs. 1, 2a and Supplementary Glossary)<br />
was made by inserting the TetR sequence from pACYC184 (New England<br />
Biolabs) and flanked with PCR-generated 425 and 227 bp of the z and a<br />
lactose operon genes, respectively. Plasmid pbla.lacY A177C (Supplementary<br />
Figs. 1, 2a and Supplementary Glossary) was made by inserting the betalactosidase<br />
gene promoter (bla) derived from pBlueScript fused with the lacY<br />
A177C cDNA derived from plasmid placY A177C obtained from J.E. Cronan 11 .<br />
Plasmid p9attP.TetR.attB (Supplementary Figs. 5a, 6a and Supplementary<br />
Glossary) was made by replacing the z and a fragments in the plasmid<br />
placY.TetR with the attachment site (9attP) from bacteriophage TP901-1<br />
(ref. 18) and attB. Plasmid p4ΦC31.attP.9attB (Supplementary Figs. 5a,<br />
6a and Supplementary Glossary) was generated by replacing the FRT and<br />
ZEO.R6K sequences of p2ΦC31.attP.FRT (Supplementary Figs. 1, 4b and<br />
Supplementary Glossary) with the bacterial attachment site 9attB of bacteriophage<br />
TP901-1 and A101 from plasmid pBAD.Red 16 , followed by insertion<br />
of two additional copies of the BAD.ΦC31 sequence.<br />
Insertional inactivation of the endonuclease A gene in BW27783 (Fig. 2b).<br />
This was achieved by Red-mediated homologous recombination between a<br />
linear DNA and the endA gene. To do this, BW27783 cells were transformed<br />
with pBAD.Red. Cells from one transformed colony were used to make competent<br />
cells, as described 16 . Briefly, the cells were cultured in 25 ml of low salt<br />
LB containing ampicillin (50 μg/ml) and 1% l-arabinose and incubated at<br />
32 °C with shaking at 250 r.p.m. until the OD600 was 0.5. The competent cells<br />
were immediately transformed with 50-ng of the Pme1-restricted targeting<br />
DNA fragment prepared from pK a nR.endA, cultured at 32 °C for 1 h, spread<br />
onto a plate containing kanamycin (Kan, 25 μg/ml) and incubated at 43 °C<br />
overnight. To select the KanR + colonies free of pBAD.Red, eight KanR + colonies<br />
were cultured in 500 μl LB with Kan at 43 °C for 30 min and 1 μl each<br />
was loaded onto the antibiotic-free, Kan + , and Amp + plates, which were incubated<br />
at 43 °C overnight. To remove the KanR gene in the strain BWΔendA.<br />
KanR, competent cells were prepared from one KanR + -colony, transformed<br />
with p2ΦC31 and spread onto a Amp + plate; subsequently, eight cultures were<br />
begun from eight p2ΦC31-transformed colonies in 500-μl antibiotic-free LB<br />
containing 1% l-arabinose at 32 °C for 2 h to induce the removal of KanR via<br />
ΦC31 integrase-mediated recombination between the attB and attP flanking<br />
the KanR, followed by 4 additional hours at 37 °C to induce I-SceI–mediated<br />
destruction of p2ΦC31. The cells were spread onto an antibiotic-free plate.<br />
Cells from eight colonies were cultured in 500 μl antibiotic-free LB at 37 °C<br />
for 30 min, and 1 μl each was loaded onto the antibiotic-free, Kan + and Amp +<br />
plates to select the AmpR − /KanR − colonies. This resulted in the BWΔendA<br />
strain. The disruption of the endA gene was confirmed by DNA sequencing of<br />
the locus-specific PCR products generated from selected colonies.<br />
Integration of the 2 nd l-arabinose transporter lacY A177C and three copies of<br />
the BAD.I-SceI gene (Supplementary Fig. 2). Replacement of the y gene (lacY)<br />
of the lactose operon with the mutant lacY A177C (Supplementary Fig. 2) and<br />
integration of three copies of the I-SceI gene into the UMU locus (Supplementary<br />
Fig. 3) were achieved by the same Red-mediated homology recombination protocol<br />
as described for the endA gene disruption (Fig. 2b). To integrate the lacY A177C,<br />
however, knockout of the wild-type lacY was performed before knocking in the<br />
lacY A177C. Both were accomplished using the same Red-mediated homologous<br />
recombination. We used a DNA fragment containing TetR flanked with fragments<br />
of the z and a genes to knock out lacY, and selected the intermediate, BWΔendA.<br />
TetR on the Tet + (6 μg/ml)-plate. This allowed the selection of the next integrant<br />
using differential KanR/TetR selection (Supplementary Figs. 2a, 5a and 6a).<br />
Integration of the BAD.ΦC31 cassettes. Three experiments were conducted<br />
to integrate ten copies of BAD.ΦC31 into the bacterial genome: two, four<br />
and four copies into the ΔendA locus (Supplementary Fig. 4), araD gene<br />
(Supplementary Fig. 5) and galK gene (Supplementary Fig. 6), respectively.<br />
All three integration events were achieved by ΦC31 integrase-mediated, sitespecific<br />
recombination between the attB or attP site in an integrating plasmid,<br />
and the attB or attP in the targeted genomic sites as described for the endA<br />
gene disruption (Fig. 2b).<br />
In an earlier attempt, we found that the integrated BAD.ΦC31 was lost soon<br />
after its integration. This is probably the result of a reverse reaction mediated by<br />
an uncharacterized excisionase or cofactor that worked in concert with ΦC31<br />
integrase produced by the leaky expression from the integrated BAD.ΦC31.<br />
To stabilize the integrant, we designed a double recombination strategy. The<br />
ΦC31 integrase was used to mediate an integration event followed by FLP- or<br />
TPin-mediated recombination to remove the attL. This would eliminate the<br />
possibility of a reverse reaction between the attL and attR. For integration of the<br />
2BAD.ΦC31 (Supplementary Fig. 4), we used the Red-mediated homologous<br />
recombination strategy with a linear DNA containing an attB site, a KanR<br />
gene and two flanking FRTs for insertion into the ΔendA locus. To remove the<br />
KanR, cells from one 3S2T.KanR colony were transformed with pcI587.FLP,<br />
and cells from a transformed colony were grown in 5 ml of antibiotic-free LB at<br />
42 °C for 8 h. This induced the expression of FLP to mediate the recombination<br />
between the two FRTs, resulting in the elimination of KanR followed by the loss<br />
of the temperature-sensitive pcI587.FLP. Subsequently, the cells named 3S2T.<br />
attB were transformed with p2ΦC31.attP.FRT. Cells from one transformed<br />
colony were grown in 2-ml of LB containing 1% l-arabinose at 32 °C for 2 h,<br />
which induced the expression of ΦC31 integrase to mediate the recombination<br />
between the attB in the ΔendA locus and the attP in the plasmid. The<br />
same FLP-mediated FRT-FRT recombination was conducted to mediate the<br />
removal of the attL and R6K.Zeo and KanR, resulting in the strain of 2P3S2T<br />
(Supplementary Fig. 4b). The integrated 2BAD.ΦC31 was confirmed by<br />
DNA sequencing of the locus-specific PCR product (Supplementary Fig. 4c)<br />
nature biotechnology<br />
doi:10.1038/nbt.1708
© 2010 Nature America, Inc. All rights reserved.<br />
and by demonstrating the function in mediating minicircle formation<br />
(Supplementary Fig. 4e). Integration of the four copies of the BAD.ΦC31<br />
gene into the araD (Supplementary Fig. 5) and galK (Supplementary<br />
Fig. 6) loci, respectively, was also mediated by ΦC31 integrase; however,<br />
the removal of the attL was mediated by bacteriophage TP901-1 integrase<br />
(TPin). To do this, we used the Red recombination system to integrate a DNA<br />
fragment encoding TetR flanked with an attB and a phage attachment site of<br />
TPin (9attP) into the araD or galK site and confirmed the integrant by DNA<br />
sequencing of the integrant-specific PCR product (Supplementary Figs. 5b<br />
and 6b). We integrated p4ΦC31.attP.9attB into the modified araD or galK<br />
locus following the same procedure as used for integrating the p2ΦC31.attP.<br />
FRT (Supplementary Fig. 4) with modifications. We selected the colonies with<br />
integrants using plates containing both Tet and Kan, transformed the selected<br />
intermediates 6P3S2T.KanR.TetR or 10P3S2T.KanR.TetR with plasmid pBAD.<br />
TPin, and screened the resulting colonies using triple antibiotic resistance (Tet,<br />
Kan and Amp). We grew the cells from one colony in 2 ml of antibiotic-free<br />
LB containing 1% l-arabinose, with shaking (250 r.p.m.) at 43 °C for 6–8 h<br />
before spreading onto an antibiotic-free plate and incubated at 43 °C overnight.<br />
l-arabinose induced expression of both recombinases, however, the incubation<br />
temperature allowed significant TPin integrase activity 18 but little ΦC31<br />
integrase activity, resulting in the TPin integrase-mediated removal of attL<br />
and selection of colonies with the desired integrant, 6P3S2T (Supplementary<br />
Fig. 5a,d) and ZYCY10P3S2T (Supplementary Fig. 6a,c).<br />
Minicircle production protocol. Our present minicircle producing system,<br />
the strain ZYCY10P3S2T plus the minicircle producer plasmid pMC.hFIX<br />
or pMC.RSV.hAAT, allowed for a greatly simplified minicircle production<br />
protocol (Fig. 1b). On day one, we inoculated cells from one transformed<br />
colony in 5 ml of TB (pH 7.0) with Kan (50 μg/ml) and incubated at<br />
37 °C with shaking at 250 r.p.m. Later that day, we amplified the bacteria by<br />
combining 100 μl of culture to every 400 ml TB containing Kan (50 μg/ml) and<br />
continued incubation for 16 to 18 h. For the yield comparison study (Fig. 2e),<br />
a 400-ml overnight culture was used to prepare intact plasmid DNA. At the<br />
end of the culture period the A 600 was 3.5– 4.2 with a pH of ~6.5. On day 2,<br />
we prepared a minicircle induction mix comprising 400 ml fresh LB, 16 ml<br />
1N sodium hydroxide and 0.4 ml 20% l-arabinose, and combined it with a<br />
400 ml overnight culture, and incubated the culture at 32 °C with shaking at<br />
250 r.p.m. for an additional 5 h. We used Qiagen plasmid purification kits to<br />
isolate minicircle from bacterial lysates following the manufacturer’s protocol<br />
with modifications. For every 400-ml overnight culture, we used a 2,500<br />
column and 100-ml each of buffers P1, P2 and P3 to ensure complete resuspension<br />
and lysis of the bacteria and a high yield of minicircle DNA vector.<br />
A step-by-step protocol is provided in Supplementary Protocol.<br />
qPCR determination of impurity DNAs. The presence of impurity DNAs, the<br />
unrecombined minicircle producer plasmid and plasmid backbone circle, in<br />
the three minicircle preparations was determined as described in the legends<br />
of Figures 2d,e by qPCR. To optimize DNA-polymerase activity, all the DNA<br />
templates were digested with AflII plus XhoI flanking the ColE1 origin before<br />
qPCR. The minicircle producer plasmid (Supplementary Fig. 7) without<br />
a transgene expression cassette was used as a standard. The PCR primers,<br />
5′-TCCTGTTACCAGTGGCTGCT and 5′-AGTTCGGTGTAGGTCGTTCG,<br />
were specific for ColE1 DNA origin shared by all templates. qPCR was<br />
performed using the Qiagen Quant SYBR Green RT-PCR kit with a RCorbett<br />
system (Corbett). The program included a denaturing step of 94 °C for 10 min,<br />
followed by 25 cycles at 94 °C for 20 s, 55 °C for 15 s, and 68 °C for 20 s each.<br />
16. Datsenko, K.A. & Wanner, B.L. Proc. Natl. Acad. Sci. USA 97, 6640–6645<br />
(2000).<br />
17. Cherepanov, P.P. & Wackernagel, W. Gene 158, 9–14 (1995).<br />
18. Stoll, S.M., Ginsburg, D.S. & Calos, M.P. J. Bacteriol. 184, 3657–3663 (2002).<br />
doi:10.1038/nbt.1708<br />
nature biotechnology
l e t t e r s<br />
High-fidelity gene synthesis by retrieval of<br />
sequence-verified DNA identified using<br />
high-throughput pyrosequencing<br />
Mark Matzas 1,5 , Peer F Stähler 1,5 , Nathalie Kefer 1 , Nicole Siebelt 1 , Valesca Boisguérin 1 , Jack T Leonard 1 , Andreas Keller 1 ,<br />
Cord F Stähler 1 , Pamela Häberle 1 , Baback Gharizadeh 2 , Farbod Babrzadeh 2 & George M Church 3,4<br />
© 2010 Nature America, Inc. All rights reserved.<br />
The construction of synthetic biological systems involving<br />
millions of nucleotides is limited by the lack of high-quality<br />
synthetic DNA. Consequently, the field requires advances in<br />
the accuracy and scale of chemical DNA synthesis and in the<br />
processing of longer DNA assembled from short fragments.<br />
Here we describe a highly parallel and miniaturized method,<br />
called megacloning, for obtaining high-quality DNA by using<br />
next-generation sequencing (NGS) technology as a preparative<br />
tool. We demonstrate our method by processing both chemically<br />
synthesized and microarray-derived DNA oligonucleotides with<br />
a robotic system for imaging and picking beads directly off of<br />
a high-throughput pyrosequencing platform. The method can<br />
reduce error rates by a factor of 500 compared to the starting<br />
oligonucleotide pool generated by microarray. We use DNA<br />
obtained by megacloning to assemble synthetic genes. In<br />
principle, millions of DNA fragments can be sequenced,<br />
characterized and sorted in a single megacloner run, enabling<br />
constructive biology up to the megabase scale.<br />
Current de novo gene construction 1–4 rests on 1990’s technology for<br />
chemical oligonucleotide synthesis, which is costly and has error rates<br />
of 1 in 300 base pairs (bp). Errors are typically avoided by manually<br />
selecting the best Sanger sequences using electrophoretic automation.<br />
Recent innovations in programmable array technology 5–8 offer the<br />
possibility to synthesize pools of thousands to millions of sequences<br />
per array with lengths comparable to conventional synthesis. The<br />
technology thus provides an extremely rich source of DNA oligonucleotides<br />
with great flexibility and superior efficiency regarding<br />
throughput and cost per bp. However, the error rate of microarrayderived<br />
oligonucleotides is typically higher compared to conventional<br />
synthesis, making error avoidance or correction necessary.<br />
Furthermore it is challenging to divide the derived oligonucleotide<br />
pools, containing vast amounts of species, into subpools—necessary,<br />
for example, to guide the assembly of synthetic genes, chromosomal<br />
regions or whole pathways in synthetic biology.<br />
Megacloning turns NGS from a previously purely analytical<br />
method into a preparative tool, and represents a tremendous source<br />
of sequence-verified DNA where the yield from one NGS run is<br />
equivalent to that from hundreds to thousands of Sanger-sequence<br />
runs. It therefore addresses the challenge of error reduction for both<br />
conventional and microarray-derived DNA oligonucleotides. The<br />
method yields high-quality DNA libraries containing perfect parts<br />
with desired and correct sequences in adjustable ratios useful for a<br />
wide range of (bio-)technological applications.<br />
Here we present a proof-of-concept study aimed at the retrieval<br />
of clonal DNA with known sequence from an NGS platform after<br />
sequencing (Fig. 1). The workflow comprises the input of DNA of<br />
short length, an NGS run to generate sequence-verified DNA clones,<br />
the identification of DNA with desired sequence on the sequencer’s<br />
substrate and the retrieval of the clones of choice. The sources for<br />
the input DNA are for the most part independent of the megacloning<br />
step. For the present work, input DNA was derived from conventional<br />
oligonucleotide synthesis and from DNA microarrays. We used<br />
the NGS platform GS FLX from Roche 454 Life Sciences 9,10 . Owing<br />
to its open-top architecture, accessibility of the beads and the bead<br />
size, this platform is well suited for a pick-and-place approach using<br />
micropipettes to retrieve specific beads from the 454-Picotiterplate<br />
(PTP) and transfer them into conventional multi-well plates for further<br />
processing.<br />
First, we established a technical setup for the controlled extraction<br />
of beads. The PTP at this stage contained a natural sample from<br />
human DNA, and extraction was done using a micropipette controlled<br />
by a microactuator device (Supplementary Data). To assess the fidelity<br />
of our setup, we compared the reads coming from the GS FLX<br />
platform with Sanger-derived sequences of DNA amplified from<br />
extracted beads. The alignment of Sanger sequences to the NGS reads<br />
matched 99.9%. Only two mismatches were obtained in 2,410 bp.<br />
Both were putative insertions in the GS FLX reads occurring at<br />
homopolymer stretches and therefore have a high likelihood of being<br />
platform-specific, base-calling artifacts 9 (Supplementary Data).<br />
Next we collected a set of 319 beads with DNA clones from a microarray-derived<br />
pool initially containing 3,918 sequences. The beads for<br />
extraction were selected to ensure that their GS FLX reads perfectly<br />
matched sequences in our starting pool. The obtained DNA and the<br />
1 febit group, Heidelberg, Germany. 2 Stanford Genome Technology Center, Stanford University, Palo Alto, California, USA. 3 Harvard Medical School, Boston,<br />
Massachusetts, USA. 4 Wyss Institute for Biologically Inspired Engineering, Boston, Massachusetts, USA. 5 These authors contributed equally to this work.<br />
Correspondence should be addressed to G.M.C. (gmc@harvard.edu).<br />
Received 8 June; accepted 19 October; published online 28 November 2010; doi:10.1038/nbt.1710<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1291
l e t t e r s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Data<br />
Data<br />
Oligonucleotide design<br />
Write<br />
(oligonucleotide synthesis)<br />
Microarrays<br />
Natural sources<br />
Conventional<br />
oligonucleotide synthesis<br />
Microarray synthesis<br />
-Oligo elution<br />
-Amplification<br />
Read<br />
Next-generation<br />
sequencing<br />
Roche/454<br />
ABI/SOLiD<br />
Illumina<br />
GS FLX<br />
Roche/454 Sort<br />
&<br />
select<br />
Next-generation<br />
sequencing<br />
Pick<br />
&<br />
place<br />
untreated pool were compared after being sequenced independently<br />
on a Genome Analyzer II (Illumina GAII). We mapped 3.1% of reads<br />
from the initial (nonenriched) DNA pool without errors to the set<br />
of 319 selected sequences. In the enriched pool the fraction of reads<br />
mapping perfectly to the target sequences was 84.3%. The increase<br />
by a factor of 27.2 shows clearly a successful enrichment of selected<br />
and correct sequences (Fig. 2a,b). Also the analysis of reads on the<br />
level of single-target sequences shows that for 94% of the sequences<br />
in the selected pool, 50% or more of the reads were correct (Fig. 2c).<br />
Error-prone sequences contained a high number of different species<br />
likely to be caused by known sequence variations on the GAII, as<br />
reported previously 11 .<br />
To test the assembly of gene fragments based on megacloned<br />
oligonucleotides stemming from a microarray, we assembled two<br />
gene fragments, each ~220 bp in length, combining either nine<br />
or ten megacloned, bead-derived amplicons in a PCR-based gene<br />
assembly reaction 12,13 . The obtained assemblies were cloned and<br />
Sanger sequenced. Seven out of eight clones matched the target<br />
sequence perfectly. Interestingly, one clone showed insertions and<br />
deletions all located within a region 23 bp wide. Errors in assemblies<br />
originating from inaccuracies in the starting material could<br />
be expected to be distributed evenly over the entire construct. As<br />
this sequence was otherwise free of errors, these defects were likely<br />
caused by misassembly rather than errors in the building blocks used<br />
(Supplementary Data).<br />
To further evaluate the capabilities of the megacloning approach to<br />
generate biologically functional genes, we applied the method to DNA<br />
fragments 274–394 bp in length and extracted 32 beads from the PTP<br />
carrying putatively correct sequences. These DNA fragments were<br />
the product of gene assembly reactions 12 using overlapping 40-mer<br />
oligonucleotides synthesized using conventional phosphoramidite<br />
chemistry and could be assembled into a model gene encoding β-dglucuronidase<br />
(uidA) 14 (2,080 bp).<br />
Three Sanger sequences obtained from the bead DNA were totally<br />
unrelated to the expected sequence and were probably caused by<br />
wrong bead extraction or contamination. The remaining 29 sequences<br />
PTP<br />
Select<br />
Pick & place<br />
Optical release<br />
Copy off surface<br />
Multiwell plate<br />
Amplify<br />
PCR<br />
Amplify<br />
PCR<br />
Perfect part<br />
Figure 1 Coalescence of DNA reading and writing. The general approach begins with DNA from<br />
a variety of sources. Here we used oligonucleotides synthesized from microarrays as well as from<br />
conventional sources. Then, next-generation sequencing is used to read and identify oligonucleotides<br />
with desired sequences. Here we used the GS FLX platform (454/Roche). Finally, the DNA is sorted<br />
and retrieved selectively, in this case with a microactuator-controlled micropipette guided by two<br />
microscope cameras. The technologies used for retrieval depend on the sequencing platform.<br />
covered 7,195 bp and matched without<br />
errors to the expected target sequences<br />
(Supplementary Data).<br />
We then assembled the model gene out of<br />
nine DNA fragments from the set of 29 matching<br />
beads. The full-length gene construct<br />
was again checked by Sanger sequencing for<br />
absence of errors, and the biological functionality<br />
of the gene was tested in an enzymatic<br />
assay based on the conversion of X-Glc<br />
(5-bromo-4-chloro-3-indolyl-β-glucoside)<br />
substrate into blue dye 15 (Supplementary<br />
Data). Besides the proof of feasibility of<br />
generating biological functional genes, this<br />
experiment further mimics other applications<br />
of our technology, such as the use of sheared<br />
natural DNA and its subsequent sorting<br />
and reordering.<br />
The absence of errors in 7,195 bp of DNA<br />
obtained from 29 extracted beads raised the<br />
question of achievable error rates from the<br />
megacloner process. Therefore we explored<br />
the potential of megacloning using a statistical<br />
model. This model considers two main sources<br />
of error—namely, wrong sequencing calls and<br />
polymerase errors during DNA amplification 16 . The calculations estimated<br />
the chance of finding one error in our extracted sequence space<br />
of ~7,200 bp to be 29%, which is in line with our experimental findings.<br />
The theoretical error rate of bead amplicons after megacloning using<br />
the setup employed in this study was estimated to be 1 error in 21 kbp<br />
(Supplementary Data). Compared with the error rate in the starting<br />
material of 1 error in 40 bp (determined from GAII data of the initial<br />
microarray pool), this equals a 500-fold error reduction.<br />
We further calculated the expected amount of reads from NGS<br />
that match the target sequences of a given pool without errors. These<br />
numbers are crucial to estimate the complexity of pools that can be<br />
processed in one megacloner run. The resulting efficiency and cost<br />
structure are influenced mainly by three parameters: the error rate of<br />
the starting pool, the sequencing accuracy and the length of the variable<br />
sequence (Supplementary Data). With an error rate of 1 error<br />
in 40 bp and an average sequencing accuracy of 99.9% in the GS FLX,<br />
we expect a five- to tenfold cost reduction in producing DNA fragments<br />
(compared to conventional oligonucleotide synthesis) that can<br />
be achieved now with the prototype device (Supplementary Data).<br />
Because these fragments are largely free of errors, further savings can<br />
be expected in gene synthesis because the cost of subsequent sequencing<br />
for final quality control will be lower.<br />
In this work we demonstrated the targeted retrieval of bead-bound<br />
DNA from a high-throughput sequencer without major modifications<br />
to the sequencing process. Previous methods for error correction in<br />
DNA pools 7,17–21 do not adequately handle collections of closely related<br />
oligonucleotide sequences that occur during assembly of repetitive<br />
sequences or multi-gene family libraries. They also do not enable hierarchical<br />
assembly strategies, which are made possible by the ordered<br />
selection and physical separation of clonal DNA described here.<br />
The megacloner process has been proven to be useful for retrieval<br />
and sorting of correct and functional sequences and to increase the<br />
portion of error-free sequences in a sample substantially. This technology<br />
allows the processing of DNA from microarrays but also from<br />
a variety of other sources, such as conventional oligonucleotide synthesis<br />
or natural DNA fragments.<br />
1292 VOLUME 28 NUMBER 12 DECEMBER 2010 nature biotechnology
l e t t e r s<br />
Megacloning could be optimized beyond the estimates in this work<br />
of one error in 21 kbp from input DNA having an error rate of 1 in<br />
40 bp. Although such raw material can be obtained by state-of-the-art<br />
microarray technologies, the quality of input DNA could be increased<br />
further by addressing the amplification step of bead-bound DNA—for<br />
example, with higher fidelity polymerases, as the predicted contribution<br />
of the polymerase to the error rate is 4.7-fold higher than the<br />
expected error rate of the megacloner itself (Supplementary Data).<br />
Another accessible parameter for optimizing the overall process in<br />
terms of error rates is improvement in the quality of the DNA starting<br />
material. Also, optimization of sequencing accuracy could be a way to<br />
improve the ability to select correct parts after NGS. This is, however,<br />
the subject of ongoing optimization in the scope of NGS development,<br />
including ligase-based methods with improved accuracy 22 .<br />
The pool used in our conceptual study contained ~4,000<br />
sequences. According to our results and extrapolations, this can be<br />
increased to ~30,000 sequences per pool with the described setup.<br />
As the bead extraction is generally independent of the pool complexity,<br />
it is mainly limited by the NGS platform and the quality of<br />
the starting material (Supplementary Data). More advanced microarray<br />
formats are able to deliver libraries with even higher complexity<br />
and of sufficient quality to fit into a gene assembly process 23 .<br />
Therefore, with an appropriate degree of automation that reaches<br />
an extraction frequency of two or three beads per minute, which is<br />
achievable with state-of-the-art robotics, the work-up of one PTP<br />
becomes possible within days, resulting in > 10 6 bp per plate. Hence,<br />
the downstream process (amplification, cleanup, assembly) will<br />
represent the next bottleneck.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
a<br />
c<br />
Percent of total read count<br />
Percent of correct matching reads<br />
100<br />
80<br />
60<br />
40<br />
20<br />
100<br />
90<br />
80<br />
70<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
Set 1 Set 2 Set 3 Set 4<br />
98.45<br />
96.53<br />
85.33<br />
84.28<br />
69.38<br />
Reads matching with<br />
max 3 errors in<br />
whole pool<br />
Before megacloning<br />
33.51<br />
Reads with<br />
perfect match<br />
in whole pool<br />
6.07<br />
Reads matching<br />
with max 3 errors<br />
in selected<br />
sequences<br />
After megacloning<br />
3.07<br />
Reads with<br />
perfect match<br />
in selected<br />
sequences<br />
b<br />
Number of sequences<br />
90<br />
80<br />
70<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
0<br />
1–10<br />
11–20<br />
21–30<br />
31–40<br />
41–50<br />
51–60<br />
61–70<br />
71–80<br />
81–90<br />
91–100<br />
Before megacloning<br />
After megacloning<br />
Reads/filtered reads in parts per million<br />
101–200<br />
201–300<br />
301–400<br />
401–500<br />
501–600<br />
601–700<br />
701–800<br />
801–900<br />
901–1,000<br />
1,001–2,000<br />
2,001–3,000<br />
3,001–4,000<br />
4,001–5,000<br />
5,001–6,000<br />
6,001–7,000<br />
7,001–8,000<br />
8,001–9,000<br />
9,001–10,000<br />
More<br />
0<br />
0 50 100 150 200 250 300<br />
100<br />
Percent of total read count<br />
80<br />
60<br />
40<br />
20<br />
0<br />
0 50 100 150<br />
Selected oligos<br />
200 250 300<br />
Figure 2 NGS-based comparison of untreated and megacloned oligonucleotide pools from microarray. (a) Comparison of the initial microarray<br />
oligonucleotide pool (blue) and the pool enriched with the megacloner technology (red) based on the results of the Illumina GAII runs. The bars in<br />
set 1 represent the fraction of reads that could be mapped allowing up to three errors. Bars in set 2 show the fractions of perfectly matching reads to<br />
the sequence set of the initial pool (3,918 sequences). The difference between the blue and the red bar in set 2 represents the enrichment of correct<br />
sequences by megacloning. The bars in set 3 and set 4 show the fractions of reads mapping to sequences from the selected pool of 319 sequences.<br />
The difference between blue and red bars in set 3 shows the enrichment of a selected 319 sequences before megacloning compared with after. Blue<br />
and red bars in set 4 represent the enrichment of sequences that are in the set of 319 selected sequences and that are correct. (b) Histogram of read<br />
counts in the Illumina GAII data of the initial pool (blue) and the enriched megacloned sample (red). Only reads mapping without errors to one of the<br />
319 selected target sequences have been taken into account. To compare the two NGS runs on the basis of read counts, we converted the numbers<br />
into parts-per-million (p.p.m.) from the total number of filtered reads. (c) Composition of reads from the Illumina GAII data including 319 selected<br />
sequences in the initial pool (top) and the enriched pool (bottom). The oligonucleotides are sorted by the fraction of correct reads. Green, correct reads;<br />
red, error-prone reads (compartments in the red bars represent single sequences with a read count of 0.1% or more of total reads for the particular<br />
sequence); light blue, sum of nonunique error-prone reads where each sequence represents less than 0.1% of total reads for the particular sequence;<br />
blue, unique reads. In the Illumina GAII data set from the enriched sample, just 315 out of 319 selected sequences could be detected.<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1293
l e t t e r s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Our next focus in the present context is improvement and<br />
automation of physical bead extraction. The workflow used in this<br />
study still involved a considerable number of manual steps and some<br />
human intervention, which was identified as the most important<br />
source of error in terms of extraction of unwanted beads. Therefore,<br />
the success rate of ~90% (29 beads out of 32) has to be increased for<br />
the bead localization and retrieval process.<br />
The method described here holds the potential to decrease production<br />
cost for synthetic DNA by one or more orders of magnitude. This<br />
source of high-quality DNA could aid the field of synthetic biology, as<br />
well as the production of libraries for antibodies or enzyme variants.<br />
In addition to synthetic sources, the sorting of natural DNA could<br />
enable the quick reconstruction or combination of DNA fragments<br />
to assemble genes, chromosomes or genomes, while simultaneously<br />
including synthetic parts of DNA.<br />
The principle that we applied here using the GS FLX technology<br />
should also be generally applicable to other available NGS platforms<br />
such as Illumina’s GAII, SOLiD, the Polonator or others. In the<br />
present context, the advantage of the GS FLX platform is the robotaccessible<br />
platform architecture and the comparably large size of the<br />
beads. Owing to different architectures of the other platforms, such<br />
as partially closed systems and substantially smaller DNA carriers,<br />
harvesting DNA from those will require a different mechanism, such<br />
as optical approaches including photosensitive and cleavable linkermolecules.<br />
The advantage of these platforms is a considerably higher<br />
number of DNA clones, which potentially could increase the capacity<br />
and throughput of the technology up to the gigabase level.<br />
Methods<br />
Methods and any associated references are available in the online version<br />
of the paper at http://www.nature.com/naturebiotechnology/.<br />
Note: Supplementary information is available on the Nature Biotechnology website.<br />
Acknowledgments<br />
We thank B.A. Roe, F.Z. Najar and D.D. White for sequencing support, J. Jäger for<br />
technical consulting, and D. Summerer, T. Brefort, S. Kosuri and D. Levner for<br />
discussions and comments.<br />
AUTHOR CONTRIBUTIONS<br />
M.M., P.F.S. and G.M.C. conceptualized the megacloning method and wrote the<br />
manuscript; M.M. designed and lead the study, wrote all algorithms for sequence<br />
design, data analysis, image conversion, image processing and microactuator<br />
control; M.M., N.K., N.S. acquired the used technology, set up the microactuator<br />
device and optical systems; N.S. designed the uidA genetic model; M.M., N.K.,<br />
N.S., V.B. and P.H. designed and optimized molecular biological methods; C.F.S.<br />
and J.T.L. contributed to bead picking and engineering concepts; A.K. set up the<br />
statistical models and calculations; J.T.L. contributed to the design of molecular<br />
biological steps and the acquisition of sequencing samples; B.G. and F.B. evaluated<br />
and implemented necessary changes into the sample preparation and the<br />
sequencing process on the 454/Roche platform.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare competing financial interests: details accompany the full-text<br />
HTML version of the paper at http://www.nature.com/naturebiotechnology/.<br />
Published online at http://www.nature.com/naturebiotechnology/.<br />
Reprints and permissions information is available online at http://npg.nature.com/<br />
reprintsandpermissions/.<br />
1. Endy, D. Foundations for engineering biology. Nature 438, 449–453 (2005).<br />
2. Menzella, H.G. et al. Combinatorial polyketide biosynthesis by de novo design and<br />
rearrangement of modular polyketide synthase genes. Nat. Biotechnol. 23,<br />
1171–1176 (2005).<br />
3. Gibson, D.G. et al. Creation of a bacterial cell controlled by a chemically synthesized<br />
genome. Science 329, 52–56 (2010).<br />
4. Carr, P.A. & Church, G.M. Genome engineering. Nat. Biotechnol. 27, 1151–1162<br />
(2009).<br />
5. Gao, X. et al. A flexible light-directed DNA chip synthesis gated by deprotection<br />
using solution photogenerated acids. Nucleic Acids Res. 29, 4744–4750 (2001).<br />
6. Singh-Gasson, S. et al. Maskless fabrication of light-directed oligonucleotide<br />
microarrays using a digital micromirror array. Nat. Biotechnol. 17, 974–978<br />
(1999).<br />
7. Tian, J. et al. Accurate multiplex gene synthesis from programmable DNA microchips.<br />
Nature 432, 1050–1054 (2004).<br />
8. Porreca, G.J. et al. Multiplex amplification of large sets of human exons. Nat.<br />
Methods 4, 931–936 (2007).<br />
9. Margulies, M. et al. Genome sequencing in microfabricated high-density picolitre<br />
reactors. Nature 437, 376–380 (2005).<br />
10. Wicker, T. et al. 454 sequencing put to the test using the complex genome of<br />
barley. BMC Genomics 7, 275 (2006).<br />
11. Willenbrock, H. et al. Quantitative miRNA expression analysis: comparing microarrays<br />
with next-generation sequencing. RNA 15, 2028–2034 (2009).<br />
12. Stemmer, W.P., Crameri, A., Ha, K.D., Brennan, T.M. & Heyneker, H.L. Single-step<br />
assembly of a gene and entire plasmid from large numbers of oligodeoxyribonucleotides.<br />
Gene 164, 49–53 (1995).<br />
13. Richmond, K.E. et al. Amplification and assembly of chip-eluted DNA (AACED): a method<br />
for high-throughput gene synthesis. Nucleic Acids Res. 32, 5011–5018 (2004).<br />
14. Jefferson, R.A., Burgess, S.M. & Hirsh, D. beta-Glucuronidase from Escherichia<br />
coli as a gene-fusion marker. Proc. Natl. Acad. Sci. USA 83, 8447–8451<br />
(1986).<br />
15. Couteaudier, Y., Daboussi, M.J., Eparvier, A., Langin, T. & Orcival, J. The GUS gene<br />
fusion system (Escherichia coli beta-D-glucuronidase gene), a useful tool in studies<br />
of root colonization by Fusarium oxysporum. Appl. Environ. Microbiol. 59,<br />
1767–1773 (1993).<br />
16. Cline, J., Braman, J.C. & Hogrefe, H.H. PCR fidelity of pfu DNA polymerase and<br />
other thermostable DNA polymerases. Nucleic Acids Res. 24, 3546–3551 (1996).<br />
17. Carr, P.A. et al. Protein-mediated error correction for de novo DNA synthesis. Nucleic<br />
Acids Res. 32, e162 (2004).<br />
18. Smith, J. & Modrich, P. Removal of polymerase-produced mutant sequences from<br />
PCR products. Proc. Natl. Acad. Sci. USA 94, 6847–6850 (1997).<br />
19. Bang, D. & Church, G.M. Gene synthesis by circular assembly amplification.<br />
Nat. Methods 5, 37–39 (2008).<br />
20. Fuhrmann, M., Oertel, W., Berthold, P. & Hegemann, P. Removal of mismatched<br />
bases from synthetic genes by enzymatic mismatch cleavage. Nucleic Acids Res.<br />
33, e58 (2005).<br />
21. Binkowski, B.F., Richmond, K.E., Kaysen, J., Sussman, M.R. & Belshaw, P.J.<br />
Correcting errors in synthetic DNA through consensus shuffling. Nucleic Acids Res.<br />
33, e55 (2005).<br />
22. McKernan, K.J. et al. Sequence and structural variation in a human genome<br />
uncovered by short-read, massively parallel ligation sequencing using two-base<br />
encoding. Genome Res. 19, 1527–1541 (2009).<br />
23. Kosuri, S. et al. Scalable gene synthesis by selective amplification of DNA pools from<br />
high-fidelity microchips. Nat. Biotechnol. advance online publication, doi:10.1038/<br />
nbt.1716 (28 November 2010).<br />
1294 VOLUME 28 NUMBER 12 DECEMBER 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.<br />
ONLINE METHODS<br />
Oligo synthesis, sequence design, adaptors. Oligonucleotides used for this<br />
work were synthesized on programmable microarray synthesizers using lightdirected<br />
synthesis methods 5 . Conventional oligonucleotides used for gene<br />
assembly were obtained from Sigma Aldrich. Harvesting of oligonucleotides<br />
from microarray surfaces was performed by chemical cleavage of succinateester<br />
bonds using ammonia hydrochloride solution.<br />
Amplification of microarray-derived oligonucleotide pools by emulsion<br />
PCR. Microarray-derived oligonucleotide pools were amplified before NGS<br />
using emulsion PCR 24 . Therefore universal terminal sequences were attached<br />
during synthesis and served as primer regions. Amplification primers contained<br />
adaptors for sequencing on the Illumina GAII platform and/or the 454<br />
GS FLX (Supplementary Data).<br />
Sequencing on the 454 GS FLX. The sample preparation for the PCRamplified<br />
oligonucleotides was done according to the manufacturer’s protocols<br />
(Roche/454). To keep the DNA intact after sequencing, we exchanged the<br />
bleaching cleaning buffer with TE buffer before the sequencing run to avoid<br />
degradation of DNA during the final cleaning steps of the Roche sequencer.<br />
Data analysis of 454 data and image conversion. NGS reads obtained from<br />
the GS FLX sequencer were aligned to the target sequences in the oligonucleotide<br />
pool to find the best matching sequence for every read and to perform<br />
further analysis, such as error rate estimation. Perfect matching sequences<br />
were selected and localized in the sequencer image by using the coordinates<br />
attached to every read sequence. For sequence data analysis, we used various<br />
Python scripts using the BioPython package. The images from the GS FLX<br />
sequencer were converted into the TIFF standard format using the Python<br />
Imaging Library.<br />
Bead localization and extraction. After aligning the GS FLX reads to the<br />
set of target sequences, we selected reads that perfectly matched one of the<br />
desired oligonucleotide sequences in the pool. For localization of beads we<br />
located the corresponding chemiluminescent signals in the converted raw<br />
image from the GS FLX platform using the x- and y-coordinates that were<br />
included in the NGS raw data. To locate beads in the PTP, we identified reference<br />
points in the raw image and their corresponding positions in the PTP<br />
using suitable patterns of light signals. Based on these reference points the<br />
bead positions on the PTP were calculated using an algorithm for scaling and<br />
rotation. The extraction was performed with a micropipette with an outer<br />
diameter of 28 μm. For pipette handling we used a three-axis microactuator<br />
(Supplementary Data). Before extraction of beads the PTP was stored under<br />
a water layer to prevent desiccation and shrinking of beads. After picking, the<br />
beads were transferred immediately into a PCR vial and stored under water<br />
until further processing.<br />
Amplification of DNA from beads. Amplification of bead-bound DNA<br />
was performed with the primers 454-A and 454-B, targeting the Roche/454<br />
adaptors, or ‘slx-fw-long’ and ‘slx-rev-long’ for Illumina adaptors. For amplification<br />
of fragments with 40-mer variable regions, primers were 5′-biotinylated<br />
to facilitate subsequent removal of primer regions on a streptavidin<br />
matrix. PCR conditions: 20 mM Tris-HCl (pH 8.8), 10 mM ammoniumsulfate,<br />
10 mM potassium chloride, 2 mM magnesium-sulfate, 0.1%<br />
Triton X-100, 200 μM each dNTP, 2% (vol/vol) DMSO, 1 μM each primer,<br />
50 U/ml native pfu polymerase (Fermentas). Cycling: initial denaturation<br />
96 °C (2 min); then 30 cycles of 96 °C (30 s), 63 °C (30 s), 72 °C (30 s) and<br />
final elongation 72 °C (3 min). After amplification, all PCR products were<br />
analyzed on PAGE (Supplementary Data) to check specificity and yield.<br />
For generation of the subpool containing 319 sequences, we estimated the<br />
concentration on the basis of the gel analysis and mixed the amplicons in<br />
equimolar concentrations.<br />
Illumina sequencing and data analysis. As the sample contained suitable<br />
adaptors all steps regarding adaptor ligation have been omitted. All other steps<br />
were done according to the protocols from Illumina.<br />
The NGS raw data obtained from Illumina GAII were processed by the<br />
following steps.<br />
1. Truncation of reads to the length of the variable regions (40 bp).<br />
2. Filtering out reads containing ambiguities (filtered reads).<br />
3. Group reads with similar sequences (bins).<br />
Subsequently for each read we identified the best matching target sequence<br />
from the oligonucleotide pool by mapping all reads to a pseudo-genome using<br />
rapid alignment of small RNA reads (razerS) (http://www.seqan.de/projects).<br />
The pseudo-genome was generated by concatenation of the variable parts<br />
of pool sequences separated by 40-mer poly-T stretches. The corresponding<br />
target sequence could then be determined by the matching position in the<br />
pseudo-genome. Alignments from the razerS output were used to determine<br />
insertions, deletions and substitutions. To compare the two GAII runs based<br />
on the number of correct reads, we converted the read counts into parts-permillion<br />
units (p.p.m.), taking the number of filtered reads before the matching<br />
procedure (after step 2) as a basis.<br />
Assembly of gene fragments from conventional oligonucleotides. Gene fragments<br />
> 200 bp were assembled from conventionally synthesized 40-mer oligonucleotides<br />
having a constant overlap region of 20 nucleotides to the adjacent<br />
oligomer. Primer regions for 454 sequencing and restriction sites for primer<br />
removal were included during assembly. The assembly reaction contained<br />
5 nM of each construction oligonucleotide and 200 nM of terminal primers.<br />
PCR conditions: 1× KOD polymerase buffer (Novagen), 1.25 mM MgSO 4 ,<br />
40 μM each dNTP, 5 U/ml KOD Hot Start Polymerase (Novagen). Cycling<br />
for gene assembly: initial denaturation 96 °C (4 min); then 30 cycles of 96 °C<br />
(10 s), 55–40 °C touchdown (30 s), 72 °C (10 s). For subsequent amplification:<br />
96 °C (10 s), 55 °C (30 s), 72° (30 s), final elongation 72 °C (3 min).<br />
Assembly of genes from >200 bp fragments. Gene assembly up to 2 kbp<br />
were performed according to the protocol used for assembly of > 200 bp from<br />
oligonucleotides.<br />
Primer removal and cleanup of bead amplicons before gene assembly. For<br />
removal of primer regions amplicons were incubated with LguI restriction endonuclease<br />
in 1× Tango buffer (Fermentas) for 3 h at 37 °C. For > 200 bp fragments,<br />
small restriction fragments containing primer regions were removed by PCR<br />
purification columns (GenElute PCR Clean-Up, Sigma Aldrich). For cleanup<br />
of microarray-derived fragments, we used 40-mer variable region biotinylated<br />
primers during bead DNA amplification and removed restriction products containing<br />
biotin residues using streptavidin matrix. The 40-mer fragments were<br />
ethanol precipitated and dissolved in water before further processing.<br />
Assembly of genes from 40-mer double-stranded DNA fragments. For the<br />
assembly of genes from 40-mer dsDNA we used a two-stage assembly protocol<br />
including a primerless PCR followed by a PCR for amplification of the<br />
resulting products described previously 13 .<br />
24. Williams, R. et al. Amplification of complex gene libraries by emulsion PCR.<br />
Nat. Methods 3, 545–550 (2006).<br />
doi:10.1038/nbt.1710<br />
nature biotechnology
l e t t e r s<br />
Scalable gene synthesis by selective amplification<br />
of DNA pools from high-fidelity microchips<br />
Sriram Kosuri 1,2,6 , Nikolai Eroshenko 1,3,6 , Emily M LeProust 4 , Michael Super 1 , Jeffrey Way 1 ,<br />
Jin Billy Li 2,5 & George M Church 1,2<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Development of cheap, high-throughput and reliable gene<br />
synthesis methods will broadly stimulate progress in biology<br />
and biotechnology 1 . Currently, the reliance on columnsynthesized<br />
oligonucleotides as a source of DNA limits further<br />
cost reductions in gene synthesis 2 . Oligonucleotides from<br />
DNA microchips can reduce costs by at least an order of<br />
magnitude 3–5 , yet efforts to scale their use have been largely<br />
unsuccessful owing to the high error rates and complexity of<br />
the oligonucleotide mixtures. Here we use high-fidelity DNA<br />
microchips, selective oligonucleotide pool amplification,<br />
optimized gene assembly protocols and enzymatic error<br />
correction to develop a method for highly parallel gene<br />
synthesis. We tested our approach by assembling 47 genes,<br />
including 42 challenging therapeutic antibody sequences,<br />
encoding a total of ~35 kilobase pairs of DNA. These<br />
assemblies were performed from a complex background<br />
containing 13,000 oligonucleotides encoding ~2.5 megabases<br />
of DNA, which is at least 50 times larger than in previously<br />
published attempts.<br />
The synthesis of DNA encoding regulatory elements, genes, pathways<br />
and entire genomes provides powerful ways to both test biological<br />
hypotheses and harness biology for our use. For example, from the<br />
use of oligonucleotides in deciphering the genetic code 6,7 to the recent<br />
complete synthesis of a viable bacterial genome 8 , DNA synthesis has<br />
engendered tremendous progress in biology. Currently, almost all<br />
DNA synthesis relies on the use of phosphoramidite chemistry on<br />
controlled-pore glass (CPG) substrates. The synthesis of gene-sized<br />
fragments (500–5,000 base pairs (bp)) relies on assembling many<br />
CPG oligonucleotides together using a variety of gene synthesis techniques<br />
2 . Technologies to assemble verified gene-sized fragments into<br />
much larger synthetic constructs are now fairly mature 8–12 .<br />
The price of gene synthesis has fallen drastically over the last decade.<br />
However, the current commercial price of gene synthesis, ~$0.40–<br />
1.00/bp, has begun to approach the relatively stable cost of the CPG<br />
oligonucleotide precursors (~$0.10–0.20/bp) 1 , suggesting that oligonucleotide<br />
cost is limiting. At these prices, the construction of large gene<br />
libraries and synthetic genomes is out of reach to most. There are many<br />
ongoing efforts to lower the cost of gene synthesis that focus on reducing<br />
the cost of the oligonucleotide precursors. For example, microfluidic<br />
oligonucleotide synthesis can reduce reagent cost by an order of magnitude<br />
and has been used for proof-of-concept gene synthesis 13 .<br />
Another promising route is to harness existing DNA microchips,<br />
which can produce up to a million different oligonucleotides on a single<br />
chip, as a source of DNA. Previous efforts have demonstrated that<br />
genes can be synthesized from DNA microchips 3–5,14 . Thus far it has<br />
not been possible to scale up these approaches for at least three reasons.<br />
First, the error rates of oligonucleotides from DNA microchips<br />
are higher than traditional column-synthesized oligonucleotides.<br />
Second, the assembly of gene fragments becomes increasingly difficult<br />
as the diversity of the oligonucleotide mixture becomes larger. Finally,<br />
the potential for cross-hybridization between individual assemblies<br />
imposes strong constraints on the sequences that can be constructed<br />
on an individual microchip.<br />
Recently, the quality of microchip-synthesized oligonucleotides was<br />
improved by controlling depurination during the synthesis process 15 .<br />
These arrays produce up to 55,000 200-mer oligonucleotides on a<br />
single chip and are sold as a ~1–10 picomole pools of oligonucleotides,<br />
termed OLS pools (oligo library synthesis). Several groups have<br />
used OLS pools in DNA capture technologies, promoter analysis and<br />
DNA barcode development 16–20 . We have previously shown that individual<br />
oligonucleotides in a 55,000 150-mer OLS pool were evenly<br />
distributed 18 . We reanalyzed this data set to provide an estimate of<br />
the frequency of transitions, transversions, insertions and deletions<br />
in this OLS pool (Online Methods) and found the overall error rate<br />
to be ~1/500 bp both before and after PCR amplification, suggesting<br />
that OLS pools can be used for accurate large-scale gene synthesis<br />
(Supplementary Table 1).<br />
To test whether OLS pools could be used for DNA microchipbased<br />
gene synthesis, we designed two pools (OLS pools 1 and 2)<br />
of different lengths, each containing ~13,000 130-mer or 200-mer<br />
oligonucleotides, respectively. Figure 1 is a general schematic of our<br />
methods for using OLS pools to perform gene synthesis. Briefly, we<br />
designed oligonucleotides that were then printed on DNA microchips<br />
and recovered as a mixed pool of oligonucleotides (OLS pool).<br />
Next, we took advantage of the long oligonucleotide lengths to<br />
1 Wyss Institute for Biologically Inspired Engineering, Boston, Massachusetts, USA. 2 Department of Genetics, Harvard Medical School, Boston, Massachusetts, USA.<br />
3 Harvard School of Engineering and Applied Sciences, Cambridge, Massachusetts, USA. 4 Agilent Technologies, Santa Clara, California, USA. 5 Present address:<br />
Department of Genetics, Stanford University, Stanford, California, USA. 6 These authors contributed equally to this work. Correspondence should be addressed to<br />
S.K. (sri.kosuri@wyss.harvard.edu) or N.E. (eroshenk@wyss.harvard.edu).<br />
Received 3 August; accepted 25 October; published online 28 November 2010; doi:10.1038/nbt.1716<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1295
l e t t e r s<br />
a<br />
b<br />
c<br />
DNA microchip<br />
OLS pool<br />
Figure 1 Schematic for scalable gene synthesis from OLS pool 2.<br />
(a,b) Pre-designed oligonucleotides (no distinction is made between<br />
dsDNA and ssDNA in the figure) are synthesized on a DNA microchip (a)<br />
and then cleaved to make a pool of oligonucleotides (b). (c) Plate-specific<br />
primer sequences (yellow or brown) are used to amplify separate plate<br />
subpools (only two are shown), which contain DNA to assemble different<br />
genes (only three are shown for each plate subpool). (d) Assembly-specific<br />
sequences (shades of blue) are used to amplify assembly subpools that<br />
contain only the DNA required to make a single gene. (e) The primer<br />
sequences are cleaved using either type IIS restriction enzymes (resulting<br />
in dsDNA) or by DpnII/USER/λ exonuclease processing (producing<br />
ssDNA). (f) Construction primers (shown as white and black sites flanking<br />
the full assembly) are then used in an assembly PCR reaction to build<br />
a gene from each assembly subpool. Depending on the downstream<br />
application, the assembled products are then cloned either before or after<br />
an enzymatic error correction step.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
d<br />
e<br />
f<br />
Plate subpools<br />
Assembly subpools<br />
Processed subpools<br />
Assembled genes<br />
independently PCR amplify and process only those oligonucleotides<br />
required for a given gene assembly. For the 200-mer OLS pool 2, we<br />
first amplified a ‘plate subpool’ that contained DNA to construct<br />
up to 96 genes, and then amplified individual ‘assembly subpools’<br />
to separate the oligonucleotides for an individual gene. For the<br />
130-mer OLS pool 1, we directly amplified assembly subpools,<br />
foregoing the plate subpool step. Next, the primers used for these<br />
amplification steps were removed by either type IIS restriction endonucleases<br />
to form double-stranded DNA (dsDNA) fragments (OLS pool 2),<br />
or a combination of enzymatic steps to form single-stranded DNA<br />
(ssDNA) fragments (OLS pool 1). Finally, we constructed full-length<br />
genes using PCR assembly, performed enzymatic error correction to<br />
improve error rates if necessary, and, finally, cloned and characterized<br />
the constructs.<br />
Obtaining subpools of only those DNA fragments required for any<br />
particular assembly is crucial for robust gene synthesis in very complex<br />
DNA backgrounds. In addition, isolating subpools relieves constraints<br />
on sequence similarity inherent in past approaches. To facilitate the<br />
partitioning of OLS pools into smaller subpools, we designed 20-mer<br />
PCR primer sets with low potential cross-hybridization (‘orthogonal’<br />
primers) derived from a set of 244,000 25-mer orthogonal sequences<br />
developed for barcoding purposes 21 . Two separate orthogonal primer<br />
sets were constructed for the different OLS pools because of their<br />
varying requirements for downstream processing. Both sets were<br />
screened for potential cross-hybridization, low secondary structure<br />
and matched melting temperatures to construct large sets of orthogonal<br />
PCR primer pairs.<br />
To construct genes from the OLS pools, we developed algorithms to<br />
split the sequence into overlapping segments with matching melting<br />
temperatures such that they could be later assembled by PCR. Genes<br />
on OLS pool 1 and 2 were designed differently to test the effect of<br />
different overlap lengths. We designed genes on OLS pool 1 such<br />
that the processed ssDNA pools fully overlapped to form a complete<br />
dsDNA sequence. In OLS pool 2, the processed dsDNA fragments<br />
partially overlapped by ~20 bp and could be assembled into a contiguous<br />
gene sequence using PCR. We initially constructed a set of<br />
fluorescent proteins to test the efficacy of the gene synthesis methods<br />
on both OLS pools (Fig. 2).<br />
For OLS pool 1, we designed two independent ‘assembly subpools’<br />
that encoded GFPmut3b plus flanking orthogonal primer sequences<br />
that were later used for PCR assembly (construction primers). The<br />
two assembly subpools, GFP43 and GFP35, differed in the average<br />
overlap length (43 and 35 bp, respectively), total length (82–90 and<br />
64–78 bases, respectively) and number of oligonucleotides (18 and<br />
22, respectively). We also designed two subpools, control subpools 1<br />
and 2, containing ten and five 130-mer oligonucleotides, respectively,<br />
to test amplification efficacy. The other eight subpools, containing a<br />
total of 12,945 130-mer sequences, were constructed on the same chip<br />
but were not used in this study except to provide potential sources<br />
of cross-hybridization. Each of these 12 subpools was flanked with<br />
independent orthogonal primer pairs (assembly-specific primers).<br />
As a control, we used these same algorithms to design a set of shorter<br />
column-synthesized oligonucleotides (20 bp average overlap; 35–45<br />
bases in length; and 39 total oligonucleotides) encoding GFPmut3b<br />
and obtained them from a commercial provider (IDT). These oligonucleotides<br />
were combined to form a third pool (GFP20) that was<br />
also tested. (All synthesized oligonucleotides used in the study can<br />
be found in Supplementary Sequences).<br />
Each of the four subpools (GFP43, GFP35, control 1 and control 2)<br />
were PCR amplified from the synthesized OLS pool using modified<br />
primers that facilitated downstream processing (Supplementary<br />
Figs. 1 and 2a) 18 . The oligonucleotides were then processed to remove<br />
primer sequences (Supplementary Figs. 2b and 3). Briefly, lambda<br />
exonuclease was used to make the PCR products single stranded, and<br />
then uracil DNA glycosylase, endonuclease VIII and DpnII restriction<br />
endonuclease were used to cleave off the assembly-specific primers.<br />
The resultant gel shows that although the reaction was efficient,<br />
unprocessed oligonucleotide still remained. In addition, we observed<br />
spurious cleavage by DpnII that was likely due to the extensive overlap<br />
within the subpool that is inherent in the gene synthesis process.<br />
We assembled the GFP43, GFP35 and GFP20 subpools using PCR,<br />
which resulted in GFP-sized products as well as many incorrect low<br />
molecular weight products (Fig. 2a).<br />
1296 VOLUME 28 NUMBER 12 DECEMBER 2010 nature biotechnology
l e t t e r s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Figure 2 Gene synthesis products.<br />
(a) Results of PCR assembly of GFPmut3<br />
from two different assembly subpools<br />
(GFP43 and GFP35) that were amplified<br />
from OLS pool 1. Full-length GFPmut3 is<br />
expected to be 779 bp and is indicated<br />
with an asterisk (*). Other bands show lower<br />
molecular weight misassembled products.<br />
(b) Gel purification and re-amplification<br />
of the full-length assembled GFPmut3.<br />
(c) Results of assembling three fluorescent<br />
proteins using the longer oligonucleotides in<br />
OLS pool 2 and a PCR assembly protocol<br />
that did not require gel isolation. (d) Results<br />
of assembling 42 variable regions of singlechain<br />
antibody fragments that contained<br />
challenging GC-rich linkers. Of the<br />
42 assemblies, all but two (7 and 24)<br />
resulted in strong bands of the correct size.<br />
We gel isolated and re-amplified these two,<br />
850 bp<br />
650 bp<br />
200 bp<br />
100 bp<br />
resulting in bands of the correct size (Supplementary Fig. 10b). The antibody that corresponds to each number is given in Supplementary Table 3<br />
and the amino acid sequence of the linker region used is given above each gel with differing amino acids in red.<br />
We gel isolated, digested and then cloned the assembly products<br />
into an expression vector (Fig. 2b and Supplementary Fig. 4). We<br />
used flow cytometry tests, manual colony counts and sequencing<br />
of individual clones to measure the error rates (Supplementary<br />
Fig. 5a,b). All three of the assays correlated well, and the error rates<br />
determined through sequencing were 1/1,500 bp, 1/1,130 bp and<br />
1/1,350 bp for the GFP43, GFP35 and GFP20 synthesis reactions,<br />
respectively (Fig. 3 and Supplementary Table 2).<br />
These results illustrate a number of important points. First, our<br />
subpool assembly primers were sufficiently well-designed to provide<br />
stringent subpool amplification of as few as 5 oligonucleotides<br />
a<br />
Fluorescent (percent)<br />
b<br />
Bp/error<br />
100<br />
90<br />
80<br />
70<br />
60<br />
50<br />
40<br />
30<br />
20<br />
10<br />
0<br />
2,000<br />
1,800<br />
1,600<br />
1,400<br />
1,200<br />
1,000<br />
800<br />
600<br />
400<br />
200<br />
0<br />
GFP20<br />
GFP43<br />
GFP35<br />
GFP43 ErrASE<br />
GFP35 ErrASE<br />
mTFP1<br />
mCitrine<br />
mApple<br />
a<br />
d<br />
650 bp<br />
650 bp<br />
mTFP1 ErrASE<br />
GFP43<br />
MW<br />
GFP35<br />
*<br />
out of a 12,995 oligonucleotide background. Second, the relative<br />
quantities of the oligonucleotides in the assembly subpools were sufficient<br />
to allow PCR assembly. Third, the error rates from 130-mer<br />
OLS pools were sufficiently low to construct gene-sized fragments<br />
(717 bp) such that >50% of the sequences were perfect. In fact, the<br />
error rates from both the GFP43 and GFP35 assemblies were indistinguishable<br />
from the column-synthesized GFP20 assemblies. Fourth,<br />
our data show that the level of fluorescence of our gene assemblies<br />
correlated with the number of constructs with perfect sequence, providing<br />
a useful screen to test fluorescent gene assemblies in OLS pool 2<br />
(Supplementary Fig. 6). Finally, although PCR assembly was able<br />
to generate full-length product, many smaller misassembled products<br />
were also formed, requiring the use of difficult-to-automate gel<br />
isolation steps.<br />
In OLS pool 2, we designed 836 assembly subpools split into<br />
11 plate subpools, encoding 2,456,706 bases of oligonucleotides that<br />
could potentially result in 869,125 bp of final assembled sequence.<br />
We first constructed three fluorescent proteins to test assembly protocols<br />
in OLS pool 2: mTFP1, mCitrine and mApple. We found that<br />
the PCR assembly protocols developed for ssDNA subpools in OLS<br />
pool 1 only produced short (
l e t t e r s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
isolation (Fig. 2c and Supplementary Fig. 7a,b). Cloning followed<br />
by flow cytometry screening showed that 6.8%, 7.5% and 6.8% of the<br />
cells were fluorescent for mTFP1, mCitrine and mApple assemblies,<br />
respectively (Fig. 3a).<br />
Assuming 6% correct sequence per construct and no selection<br />
against errors in the assembly process, the error rate was ~1/250 bp<br />
for 200-mer OLS pool 2, significantly above that of the estimates for<br />
130-mer OLS pool 1 (~1/1,000 bp) and the sequenced 55,000 150-<br />
mer OLS pool (~1/500 bp). This is not completely unexpected, as the<br />
amount of depurination is dependent upon the number of deprotection<br />
steps during synthesis and thus the oligonucleotide length.<br />
Despite the higher error rate, there were several advantages to the<br />
200-mer OLS pool 2. First, the extensive overlaps designed in OLS<br />
pool 1 caused spurious processing of the primers from the assembly<br />
subpools. The use of type IIs restriction endonucleases to process<br />
primers to form dsDNA resulted in more robust processing. Second,<br />
the use of two amplification steps conserves chip-eluted DNA to<br />
allow for future scaling of the gene synthesis process (Supplementary<br />
Note). Third, the assemblies of OLS pool 1 produced many smaller<br />
bands and required lower-throughput gel isolation procedures. This<br />
could be due to mispriming during PCR assembly because of the long<br />
overlap lengths used in the design process. The assemblies in OLS<br />
pool 2 used much shorter overlap lengths and resulted in no smaller<br />
molecular weight misassembled products.<br />
To improve the error rates of the genes assembled from OLS pool 2,<br />
we used ErrASE, a commercially available enzyme cocktail that<br />
detects and corrects mismatched base pairs, to remove errors in the<br />
assembled fluorescent proteins. For each gene, we applied ErrASE<br />
at six different stringencies, reamplified the constructs, cloned the<br />
PCR products and rescreened the cloned genes using flow cytometry.<br />
Improvement of the level of fluorescence progressively increased<br />
with greater ErrASE stringency. At the highest levels of error correction,<br />
the fluorescence levels were 31%, 49% and 26% for mTFP1,<br />
mCitrine and mApple respectively (Fig. 3a and Supplementary<br />
Fig. 8). We also performed the ErrASE procedure on our GFP43<br />
and GFP35 pools from OLS pool 1, resulting in fluorescence levels<br />
of 89% and 92%, respectively (Fig. 3a and Supplementary Fig. 5c).<br />
We sequenced clones of GFP43 and GFP35 and found three errors in<br />
21,510 (1/7,170 bp) and four errors in 20,076 (1/5,019 bp) sequenced<br />
bases, respectively.<br />
As a more challenging test for our DNA synthesis technology,<br />
we designed and synthesized oligonucleotides in OLS pool 2 for<br />
42 genes encoding the variable regions of single-chain antibody<br />
fragments (scFv) regions corresponding to a number of well-known<br />
antibodies. We have previously had trouble synthesizing these<br />
genes using commercial gene synthesis companies. This might be<br />
partly due to the prototype (Gly 4 Ser) 3 linker, which is designed<br />
to maximize flexibility and allow the heavy and light V regions to<br />
assemble 22 . The repetitive nature and high GC content of the linkerencoding<br />
sequences often represents a challenge for accurate DNA<br />
synthesis. We therefore tested three different linker sequences that<br />
varied in GC content and repetitive character of the linker encoding<br />
sequence. In addition, the presence of high sequence homology in<br />
the antibody backbones and linkers represented a potential source of<br />
cross-hybridization that could interfere with assembly (61% average<br />
sequence identity).<br />
As expected, the antibody sequences did not assemble as robustly<br />
as the fluorescent proteins, and thus we further optimized the conditions<br />
during pre- and post-assembly (Supplementary Figs. 7c, 9<br />
and 10a). Under the best protocol, 40 of the 42 constructs assembled<br />
to the correct size (Fig. 2d and Supplementary Table 3). The two<br />
misassembled genes displayed faint bands at the correct size, which<br />
were gel isolated and reamplified to produce strong bands of the correct<br />
size. We sequenced 15 antibodies including representatives from<br />
all three linker types. We performed enzymatic error correction using<br />
ErrASE, gel isolated the product and finally cloned the constructs<br />
into an expression vector. One of the 15 antibodies did not clone,<br />
and another had a deleted linker region in all 21 sequenced clones.<br />
Both of these antibodies were encoded with the highest GC content<br />
linker. The average error rate of the 14 antibodies that did clone was<br />
1/315 bp (Fig. 3b and Supplementary Table 2); this was considerably<br />
higher than the GFP assemblies, but still sufficient for construction<br />
of genes of this size (~10% of clones should be perfect, on average).<br />
In addition, the high levels of sequence similarity between the antibodies,<br />
combined with the successful assembly and sequencing (which<br />
showed no instances of cross-contamination) further validates that<br />
the selective amplification is at least stringent enough to make highly<br />
related protein sequences.<br />
Our results show the assembly of gene-sized DNA fragments<br />
totaling ~35,000 bp from oligonucleotide pools of more than<br />
50 kilobases. A number of key features are important to make the<br />
process work, including the use of low-error starting material, wellchosen<br />
orthogonal primers, subpool amplification of individual<br />
assemblies, optimized assembly methods and enzymatic error<br />
correction. Together, these features enabled gene assembly from<br />
oligonucleotide pools containing at least 50 times more sequences<br />
than previously reported (Supplementary Note). We describe two<br />
separate OLS pool lengths and assembly methods, which have their<br />
own advantages and disadvantages (Supplementary Fig. 1). The<br />
shorter, 130-mer OLS pool 1 assemblies have lower error rates, but<br />
because there are no plate amplifications, will be harder to scale as<br />
we begin to utilize larger OLS pools. The longer 200-mer OLS pool 2<br />
is easier to scale, but contained higher error rates. The costs of<br />
oligonucleotides in both processes are
l e t t e r s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare competing financial interests: details accompany the full-text<br />
HTML version of the paper at http://www.nature.com/naturebiotechnology/.<br />
Published online at http://www.nature.com/naturebiotechnology/.<br />
Reprints and permissions information is available online at http://npg.nature.com/<br />
reprintsandpermissions/.<br />
1. Carr, P.A. & Church, G.M. Genome engineering. Nat. Biotechnol. 27, 1151–1162<br />
(2009).<br />
2. Tian, J., Ma, K. & Saaem, I. Advancing high-throughput gene synthesis technology.<br />
Mol. Biosyst. 5, 714–722 (2009).<br />
3. Tian, J. et al. Accurate multiplex gene synthesis from programmable DNA microchips.<br />
Nature 432, 1050–1054 (2004).<br />
4. Richmond, K.E. et al. Amplification and assembly of chip-eluted DNA (AACED): a<br />
method for high-throughput gene synthesis. Nucleic Acids Res. 32, 5011–5018<br />
(2004).<br />
5. Zhou, X. et al. Microfluidic PicoArray synthesis of oligodeoxynucleotides and<br />
simultaneous assembling of multiple DNA sequences. Nucleic Acids Res. 32,<br />
5409–5417 (2004).<br />
6. Nirenberg, M.W. & Matthaei, J.H. The dependence of cell-free protein synthesis in<br />
E. coli upon naturally occurring or synthetic polyribonucleotides. Proc. Natl. Acad.<br />
Sci. USA 47, 1588–1602 (1961).<br />
7. Söll, D. et al. Studies on polynucleotides, XLIX. Stimulation of the binding of<br />
aminoacyl-sRNA’s to ribosomes by ribotrinucleotides and a survey of codon<br />
assignments for 20 amino acids. Proc. Natl. Acad. Sci. USA 54, 1378–1385<br />
(1965).<br />
8. Gibson, D.G. et al. Creation of a bacterial cell controlled by a chemically synthesized<br />
genome. Science 329, 52–56 (2010).<br />
9. Gibson, D.G. Synthesis of DNA fragments in yeast by one-step assembly of<br />
overlapping oligonucleotides. Nucleic Acids Res. 37, 6984–6990 (2009).<br />
10. Li, M.Z. & Elledge, S.J. Harnessing homologous recombination in vitro to generate<br />
recombinant DNA via SLIC. Nat. Methods 4, 251–256 (2007).<br />
11. Bang, D. & Church, G.M. Gene synthesis by circular assembly amplification.<br />
Nat. Methods 5, 37–39 (2008).<br />
12. Shao, Z., Zhao, H. & Zhao, H. DNA assembler, an in vivo genetic method for rapid<br />
construction of biochemical pathways. Nucleic Acids Res. 37, e16 (2009).<br />
13. Lee, C.-C., Snyder, T.M. & Quake, S.R. A microfluidic oligonucleotide synthesizer.<br />
Nucleic Acids Res. 38, 2514–2521 (2010).<br />
14. Kim, C. et al. Progress in gene assembly from a MAS-driven DNA microarray.<br />
Microelectron. Eng. 83, 1613–1616 (2006).<br />
15. LeProust, E.M. et al. Synthesis of high-quality libraries of long (150mer)<br />
oligonucleotides by a novel depurination controlled process. Nucleic Acids Res. 38,<br />
2522–2540 (2010).<br />
16. Patwardhan, R.P. et al. High-resolution analysis of DNA regulatory elements by<br />
synthetic saturation mutagenesis. Nat. Biotechnol. 27, 1173–1175 (2009).<br />
17. Schlabach, M.R. et al. Synthetic design of strong promoters. Proc. Natl. Acad. Sci.<br />
USA 107, 2538–2543 (2010).<br />
18. Li, J.B. et al. Multiplex padlock targeted sequencing reveals human hypermutable<br />
CpG variations. Genome Res. 19, 1606–1615 (2009).<br />
19. Li, J.B. et al. Genome-wide identification of human RNA editing sites by parallel<br />
DNA capturing and sequencing. Science 324, 1210–1213 (2009).<br />
20. Porreca, G.J. et al. Multiplex amplification of large sets of human exons.<br />
Nat. Methods 4, 931–936 (2007).<br />
21. Xu, Q. et al. Design of 240,000 orthogonal 25mer DNA barcode probes. Proc. Natl.<br />
Acad. Sci. USA 106, 2289–2294 (2009).<br />
22. Huston, J.S. et al. Medical applications of single-chain antibodies. Int. Rev.<br />
Immunol. 10, 195–217 (1993).<br />
23. Carr, P.A. et al. Protein-mediated error correction for de novo DNA synthesis. Nucleic<br />
Acids Res. 32, e162 (2004).<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1299
© 2010 Nature America, Inc. All rights reserved.<br />
ONLINE METHODS<br />
Reanalysis of OLS pool error rates. We reanalyzed a previously published<br />
data set for determining sequencing error rates 18 . Briefly, the data set was<br />
derived from high-throughput sequencing using the Illumina Genome<br />
Analyzer platform of a 53,777 150-mer OLS pool. Two sequencing runs were<br />
performed, the first before any amplification, and the second after two rounds<br />
of ten cycles of PCR (20 cycles total). As our previous analyses were mostly<br />
looking for distribution effects, we reanalyzed these existing data to get an estimate<br />
of error rates before and after PCR amplification. We realigned the data<br />
set using Exonerate to allow for gapped alignments and analysis of indels 24 .<br />
Specifically, we used an affine local alignment model that is equivalent to the<br />
classic Smith-Waterman-Gotoh alignment, a gap extension penalty of −5,<br />
and used the full refine option to allow for dynamic programming–based<br />
optimization of the alignment. These reads were solely mapped on base calls<br />
by the Illumina platform. We used these alignments to count mismatches,<br />
deletions and insertions as compared to the designed sequences. However,<br />
as base-calling can be more error prone on next-generation platforms than<br />
traditional Sanger-based approaches, we filtered the results based only on<br />
high-quality base-calls (Phred scores of ≥30 or >99.9% accuracy). This was<br />
accomplished by converting Illumina quality scores to Phred values using<br />
the Maq utility sol2sanger 25 and only using statistics from base calls of Phred<br />
30 or higher. All error rate analysis scripts were implemented in Python and<br />
are available upon request. Although this method provides an estimate for<br />
error rates, unmapped reads may have higher error rates, thus underestimating<br />
the total average error rate. In addition, base-calling errors might still overestimate<br />
the error rate. Finally, using only high-quality base calls, which usually<br />
occur only in the first ten bases of a read, might only reflect error rates on the<br />
5′ end of the synthesized oligonucleotide.<br />
Design and synthesis of OLS pools. The 13,000 oligos in the first OLS library<br />
(OLS pool 1) were broken up into 12 separately amplifiable subpools (assembly<br />
subpools). Each assembly subpool was defined by unique 20 bp priming<br />
sites that flanked each of the oligos in the pool. The priming sites were<br />
designed to minimize amplification of oligos not in the particular assembly<br />
subpool. This was done by designing set of orthogonal 20-mers (assemblyspecific<br />
primers) using a set of 240,000 orthogonal 25-mers 21 as a seed. From<br />
these sequences we selected 20-mers with 3′ sequence ending in thymidine<br />
or GATC for the forward and reverse primers, respectively. We screened for<br />
melting temperatures of 62–64 °C and low primer secondary structure. After<br />
the additional filtering, 12 pairs of forward and reverse primers were chosen<br />
to be the assembly-specific primers. The 13,000 oligos in the second OLS<br />
library (OLS pool 2) were broken up into 11 subpools corresponding to 11<br />
sets of up to 96 assemblies (plate subpools), which were further divided into a<br />
total of 836 assembly subpools. A new set of orthogonal primers were designed<br />
similarly to the previous set (without the GATC and thymidine constraints)<br />
but further filtered to remove type IIS restriction sites, secondary structure,<br />
primer dimers and self-dimers. The final set of primer pairs was distributed<br />
among the plate-specific primers, assembly-specific primers and construction<br />
primers. See Supplementary Methods for more detailed design information<br />
and primer sequences.<br />
OLS pools were synthesized by Agilent Technologies and are available upon<br />
signing a Collaborative Technology Development agreement with Agilent.<br />
Costs of OLS pools are a function of the number of unique oligos synthesized<br />
and of the length of the oligos (
alemtuzumab, ranibizumab, cetuximab, efungumab, pertuzumab, tadocizumab<br />
and trastuzumab; see Fig. 2d and Supplementary Table 3). Using the same<br />
methods as with the first set of cloned antibodies, this second set was errorcorrected,<br />
gel-isolated, cloned and sequenced. See the Supplementary<br />
Methods for more detailed protocols.<br />
24. Slater, G.S. & Birney, E. Automated generation of heuristics for biological sequence<br />
comparison. BMC Bioinformatics 6, 31 (2005).<br />
25. Li, H. Maq: mapping and assembly with qualities. Welcome Trust Sanger Institute<br />
(2010). Available at: .<br />
26. Cormack, B.P., Valdivia, R.H. & Falkow, S. FACS-optimized mutants of the green<br />
fluorescent protein (GFP). Gene 173, 33–38 (1996).<br />
© 2010 Nature America, Inc. All rights reserved.<br />
doi:10.1038/nbt.1716<br />
nature biotechnology
l e t t e r s<br />
Rapid translocation of nanoparticles from the lung<br />
airspaces to the body<br />
Hak Soo Choi 1 , Yoshitomo Ashitate 1 , Jeong Heon Lee 1 , Soon Hee Kim 1 , Aya Matsui 1 , Numpon Insin 2 ,<br />
Moungi G Bawendi 2 , Manuela Semmler-Behnke 3 , John V Frangioni 1,4,6 & Akira Tsuda 5,6<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Nano-size particles show promise for pulmonary drug<br />
delivery, yet their behavior after deposition in the lung<br />
remains poorly understood. In this study, a series<br />
of near-infrared (NIR) fluorescent nanoparticles were<br />
systematically varied in chemical composition, shape, size and<br />
surface charge, and their biodistribution and elimination were<br />
quantified in rat models after lung instillation. We demonstrate<br />
that nanoparticles with hydrodynamic diameter (HD) less than<br />
≈34 nm and a noncationic surface charge translocate rapidly<br />
from the lung to mediastinal lymph nodes. Nanoparticles of<br />
HD < 6 nm can traffic rapidly from the lungs to lymph nodes<br />
and the bloodstream, and then be subsequently cleared by the<br />
kidneys. We discuss the importance of these findings for drug<br />
delivery, air pollution and carcinogenesis.<br />
Nanoparticles have been proposed as diagnostic, therapeutic and theranostic<br />
agents for a wide variety of human diseases 1–3 . Delivery of nanoparticles<br />
through the lung is receiving increased attention owing to the<br />
large lung surface area available and the minimal anatomical barriers<br />
limiting access to the body 4 . The behavior of nanoparticles in the lung<br />
is also important for understanding the health effects of air pollution.<br />
Recent toxicological studies have confirmed that nano-size particles<br />
reach deep into the alveolar region of the lungs 5,6 and can cause severe<br />
inflammatory reactions because of their large surface area–to-mass<br />
ratio 6 . Inhalation of nanoparticles is increasingly recognized as a major<br />
cause of adverse health effects and has an especially strong influence<br />
on the cardiovascular system and hemostasis, leading to increased<br />
cardiovascular morbidity and mortality 6–8 . Inhaled nanoparticles can<br />
pass from the lungs into the bloodstream and extrapulmonary organs 9 .<br />
Alveolar macrophage-mediated translocation from the lung surface<br />
to the regional tracheobronchial lymph nodes has been shown for<br />
micrometer-sized particles 10 . However, this process takes a relatively<br />
long time, from several hours to weeks 10–12 .<br />
Here we study the behavior of nanoparticles in the first hour<br />
after administration to the rat lung and attempt to define the key<br />
parameters that control their translocation to draining lymph nodes,<br />
the bloodstream and the rest of body, and subsequent elimination<br />
(that is, clearance). The standard approach for studying the translocation<br />
of inhaled nanoparticles and ultrafine air pollutants from the<br />
lungs to extrapulmonary compartments in animals is to perform postmortem<br />
analysis of tissues after inhalation of carbon-based particles 13 ,<br />
radiotracers 9 or neutron-activated metal particles 11,14,15 . Compared to<br />
these methods, real-time NIR fluorescence imaging provides several<br />
advantages for quantifying the process of nanoparticle trafficking,<br />
including a low autofluorescence background, real-time imaging,<br />
real-time overlay with anatomy and highly sensitive detection of targets<br />
several millimeters deep in tissue 16 . We systematically varied the<br />
chemical composition, size, shape and surface charge of NIR fluorescent<br />
nanoparticles to determine how their physicochemical properties<br />
affect trafficking across the alveolar surface barrier into tissue,<br />
and once in the tissue, their biodistribution and clearance. Enabling<br />
this study is our intraoperative fluorescence-assisted resection and<br />
exploration (FLARE) imaging system, which permits two independent<br />
channels (700 nm and 800 nm) of NIR fluorescence images to be<br />
acquired simultaneously with color video images in real time 17,18 .<br />
Two distinct families of nanoparticles were engineered with varying<br />
chemical composition, shape, size and surface charge (Table 1<br />
and Fig. 1a). Inorganic/organic hybrid nanoparticles (INPs) emitted<br />
at 800 nm. Quantum dots with various core(shell) and organic coating<br />
ligands were used to produce INPs of various sizes and charges<br />
(INP1–INP5; Supplementary Fig. 1). Gel-filtration chromatography<br />
was used to measure the final HD, which ranged from 5 to 23 nm in<br />
PBS (Supplementary Fig. 2). It is important to note that the surfacecharged<br />
INPs with either carboxylic (INP3 and INP5) or amino<br />
(INP4) groups showed high protein adsorption, which contributed<br />
to a significant increase in final HD when exposed to 100% serum 19 .<br />
To study the effect of surface charge in detail, we engineered INP1<br />
and INP2 to have zwitterionic and polar coatings, respectively, which<br />
eliminated serum protein binding and resulted in a constant HD in<br />
PBS and serum 19,20 . To engineer INPs with HD > 50 nm, we used<br />
silica nanospheres doped with quantum dots and conjugated NIR<br />
fluorophore CW800 on the silica surface (INP6–INP9). These silica<br />
nanoparticles did not associate with serum proteins; thus, the original<br />
size was maintained after 1 h serum incubation at 37 °C. Organic<br />
1 Division of Hematology/Oncology, Department of Medicine, Beth Israel Deaconess Medical Center, Boston, Massachusetts, USA. 2 Department of Chemistry,<br />
Massachusetts Institute of Technology, Cambridge, Massachusetts, USA. 3 Institute of Lung Biology and Disease, Helmholtz Center München—German Research<br />
Center for Environmental Health, Neuherberg/Munich, Germany. 4 Department of Radiology, Beth Israel Deaconess Medical Center, Boston, Massachusetts, USA.<br />
5 Molecular and Integrative Physiological Sciences, Harvard School of Public Health, Boston, Massachusetts, USA. 6 These authors contributed equally to this work.<br />
Correspondence should be addressed to J.V.F. (jfrangio@bidmc.harvard.edu) or A.T. (atsuda@hsph.harvard.edu).<br />
Received 16 April; accepted 5 October; published online 7 November 2010; doi:10.1038/nbt.1696<br />
1300 VOLUME 28 NUMBER 12 DECEMBER 2010 nature biotechnology
l e t t e r s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Table 1 Chemical and physical properties of inorganic/organic hybrid nanoparticles (INPs) and organic nanoparticles (ONPs)<br />
HD a (nm)<br />
Translocation at 1 h<br />
Inorganic/organic<br />
hybrid nanoparticles<br />
(INPs)<br />
Organic nanoparticles<br />
(ONPs)<br />
Nanoparticle<br />
Core(shell) Organic coating In PBS In serum Surface charge<br />
nanoparticles (ONPs) were designed to emit 700 nm NIR fluorescence,<br />
which permitted comparison of the effect of nanoparticle chemical<br />
composition with 800 nm–emitting INPs.<br />
To investigate the role of HD in the translocation of nanoparticles from<br />
the lung to extrapulmonary compartments of the body, we systematically<br />
increased the size of fluorescent nanoparticles from 5 to 300 nm.<br />
Translocation of INP5 (38 nm HD) and larger particles was severely<br />
restricted at 30 min after administration (Supplementary Fig. 3),<br />
which indicates that the size threshold for rapid translocation of nanoparticles<br />
from lungs to lymph nodes corresponds to an HD
l e t t e r s<br />
a<br />
SBR (normalized)<br />
25<br />
15<br />
5<br />
0<br />
Translocation of INP1<br />
(HD = 5 nm, Zwitterionic)<br />
LN<br />
Blood<br />
Urine<br />
10 20 30 40 50 60<br />
Time (min)<br />
b<br />
Lymph node Lung<br />
Mu<br />
LN+ LN–<br />
Mu<br />
LN+ LN–<br />
c<br />
%ID/g<br />
0.4 Inset<br />
0.3<br />
0.2<br />
0.1<br />
0<br />
0 10 20 30 40 50 60<br />
Time (min)<br />
–<br />
TcO 4<br />
Tc-INP1<br />
Tc-INP3<br />
Tc-INP4<br />
0.3<br />
0.2<br />
0.1<br />
Inset<br />
0<br />
0 5 10<br />
Lag<br />
SBR (normalized)<br />
25<br />
15<br />
5<br />
0<br />
Translocation of INP3<br />
(HD = 27 nm, Anionic)<br />
LN<br />
Blood<br />
Urine<br />
10 20 30 40 50 60<br />
Time (min)<br />
Liver Kidney<br />
%ID/g<br />
2.5<br />
2.0<br />
1.5<br />
1.0<br />
0.5<br />
0<br />
–<br />
TcO 4<br />
Tc-INP1<br />
Tc-INP3<br />
Tc-INP4<br />
TcO 4<br />
– Tc-INP1 Tc-INP3 Tc-INP4<br />
%ID<br />
100<br />
75<br />
50<br />
25<br />
0<br />
Urine<br />
Lung<br />
Body<br />
Total<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Figure 2 Biodistribution, clearance and histological analysis of INPs in Sprague-Dawley rats. (a) Translocation into lymph node, blood and urinary<br />
excretion of INP1 (left) and INP3 (right) using real-time NIR fluorescence imaging. Each point represents the mean ± s.d. of n = 3 animals. SBR,<br />
signal-to-background ratio. (b) Frozen sections obtained from resected organs of INP1-administered Sprague-Dawley rats at 1 h after instillation.<br />
From top to bottom are representative images of lung, lymph node, kidney (arrow: cortex; arrowhead: calyces), and liver. Mu, muscle; LN+, posterior<br />
mediastinal lymph node; LN−, negative para-aortic lymph node. Scale bars, 5 mm. Shown are color video and NIR fluorescence of intact specimens<br />
(left two panels, respectively) along with representative histological images from the same organ/tissue (H&E, NIR, right two panels, respectively). Green<br />
dotted circle, bronchiole; blue dotted circle, glomerular basement membrane; red dotted circle, portal area (portal vein, hepatic artery and bile duct).<br />
Scale bars, 200 μm. All NIR fluorescence images (λ Exc = 760 ± 20 nm and λ Em = 795 nm longpass) have identical exposure times and normalizations.<br />
(c) Quantitative biodistribution and clearance using 99m Tc-conjugated INPs administered intratracheally into Sprague-Dawley rats. The small molecule<br />
TcO 4<br />
− was used as a control. Translocation from lung to blood over time (top). Translocation from lung to regional lymph nodes (bottom left) 1 h after<br />
injection. Recovery of injected dose in urine, lung, body (without lungs) and total (bottom right). Each data point represents the mean ± s.d. of n = 3<br />
animals. All values in blood curves are statistically different (ANOVA) from each other at 1 h.<br />
translocation of nanoparticles into the mediastinal lymph nodes.<br />
A previous study has shown the importance of surface properties<br />
of nanoparticles in hemostasis-inducing effects, such as platelet<br />
activation, which may lead to thrombosis 13 . On the other hand, the<br />
behavior of highly net-charged nanoparticles (charge-to-volume<br />
ratio >±1) was more complicated. As both anionic and cationic<br />
charged molecules quickly adsorb endogenous proteins in the lungs,<br />
similar migration results were expected among the batches 19,21 . The<br />
translocation behavior of nanoparticles in the alveolar lining fluid<br />
was, however, very different. Carboxylic group–modified INP3<br />
was found in the lymph nodes within 30 min after administration,<br />
whereas no significant translocation was observed for the cationic<br />
INP4 (Fig. 1c, bottom), even though they were engineered from the<br />
same batch of CdTe(CdSe) nanocrystals. As it is well known that<br />
cationic-charged materials can be taken up by cells 22 , sometimes<br />
causing acute cytotoxicity, it is likely that cationic particles (INP4)<br />
were rapidly taken up by pulmonary macrophages and/or epithelial<br />
cells shortly after administration, making them unavailable for<br />
translocation to lymph nodes. As cationic nanoparticles remain in<br />
pulmonary cells for a long time, they may cause severe lung injury 23 .<br />
Our results are also consistent with recent reports on pulmonary<br />
drug delivery systems that suggest that surface charge affects translocation<br />
efficiency 24 .<br />
The majority of the nanoparticles administered in this study<br />
remained in the lungs, with small amounts rapidly appearing in the<br />
mediastinal lymph nodes within 30 min after administration. Even<br />
more rapid translocation (in minutes) from the site of deposition to<br />
lymph nodes was only observed for smaller nanoparticles. In particular,<br />
for the smallest nanoparticles (INP1, 5 nm), migration to the mediastinal<br />
lymph nodes was very fast (within 3 min), and there was measurable<br />
nanoparticle accumulation in the kidneys and excretion into urine<br />
at 30 min after administration (Fig. 2a and Supplementary Video 1).<br />
On the other hand, INP3 (27 nm) showed slower accumulation into<br />
lymph nodes (≈10 min) and blood (≈20 min). Most of our ultra-small<br />
nanoparticles were found in the kidneys, with the smallest ones also<br />
found in the urine (Fig. 2b and Supplementary Fig. 4). Because of the<br />
zwitterionic surface coating and small HD 19,25 , there were virtually no<br />
nanoparticles trapped in the hepatic clearance route, such as liver, bile<br />
and intestine (Fig. 2b and Supplementary Video 2).<br />
To confirm the qualitative observations made using NIR fluorescence,<br />
we labeled INPs with 99m Tc-MAS 3 (MAS 3 is S-acetylmercaptoacetyltriserine)<br />
and repeated experiments to better quantify<br />
nanoparticle translocation from the lung airspaces to lymph node,<br />
blood and eventually urine (Fig. 2c and Supplementary Figs. 5 and 6).<br />
Blood clearance was measured by intermittent sampling of the tail<br />
vein. The small-molecule sodium pertechnetate (TcO 4 − ) was used as<br />
a control and showed rapid uptake into blood, consistent with direct<br />
translocation through the alveolar air-blood barrier 9 , excretion into<br />
urine and moderate uptake in lymph nodes at 1 h post-instillation. On<br />
the other hand, INP1 and INP3 appeared in the blood after a short lag<br />
and at concentrations inversely proportional to HD. Because lymph<br />
node uptake of nanoparticles was more pronounced than TcO 4 − ,<br />
this suggests that there could be another pathway for translocation<br />
of nanoparticles from lung airspaces into blood, one involving transient<br />
passage through lymph nodes.<br />
In this study, we have systematically engineered various nanoparticles<br />
to investigate the effect of particle characteristics on the ability<br />
of nanoparticles to translocate from the lungs into the lymph nodes<br />
and bloodstream, and on their fate in vivo. Although nanoparticles are<br />
stable and show a narrow size distribution in PBS, this study suggests<br />
that the final HDs as measured in serum, along with surface charge,<br />
are critical in determining the ultimate migration, biodistribution and<br />
clearance of nanoparticles deposited in the lungs. The key findings<br />
of our study are: (i) a size threshold of ≈34 nm determines whether<br />
there is rapid transepithelial translocation of nanoparticles from the<br />
alveolar luminal surface into the septal interstitium, followed by quick<br />
1302 VOLUME 28 NUMBER 12 DECEMBER 2010 nature biotechnology
l e t t e r s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
translocation to the regional draining lymph nodes where further<br />
translocation into the bloodstream could occur; (ii) below this size<br />
threshold of ≈34 nm, surface charge is a major factor that influences<br />
translocation, with zwitterionic, anionic and polar surfaces being<br />
permissive and cationic surfaces being restrictive; and (iii) when<br />
HD is 100 nm) cationic liposomes provide an improved therapeutic<br />
window to treat lung diseases. That is, systemic absorption is very<br />
low, presumably because these nanoparticles exceed the size and<br />
charge thresholds defined above (with whatever systemic absorption<br />
is observed probably due to liposome instability). Second, and more<br />
importantly, engineering nanoparticle-based drugs to be zwitterionic<br />
and 34 nm) and cationic could potentially minimize<br />
airspace-to-body translocation and provide time for mucociliaryand/or<br />
macrophage-mediated clearance. Once in the body and lodged<br />
in lymph nodes (that is 6 nm ≤ HD ≤ 34 nm), noncationic nanoparticles<br />
could cause inflammation and/or contribute to carcinogenesis. Smaller<br />
nanoparticles, such as those ≈5 nm in HD, are of even more concern<br />
for carcinogenesis and distal inflammation because they are capable<br />
of traveling from the lung to the bloodstream, and once in the bloodstream,<br />
can potentially reach every tissue and organ in the body 26 .<br />
Methods<br />
Methods and any associated references are available in the online<br />
version of the paper at http://www.nature.com/naturebiotechnology/.<br />
Note: Supplementary information is available on the Nature Biotechnology website.<br />
Acknowledgments<br />
We thank R. Oketokoun and S. Gioux for help with developing the FLARE imaging<br />
system and E.P. Lunsford of the Longwood Small Animal Imaging Facility for<br />
assistance with SPECT/CT imaging. The Biophysical Instrumentation Facility for<br />
the Study of Complex Macromolecular Systems (National Science Foundation-<br />
0070319 and US National Institutes of Health (NIH) GM68762) is gratefully<br />
acknowledged. We thank W. Liu and B.I. Ipe for providing quantum dots, L. Moffitt<br />
for editing, and L. Keys and E. Trabucchi for administrative support. This work<br />
was supported in part by NIH grant HL054885 (A.T.), HL070542 (A.T.), HL074022<br />
(A.T.) and R01-CA-115296 (J.V.F.).<br />
AUTHOR CONTRIBUTIONS<br />
H.S.C., Y.A., J.H.L., S.H.K., A.M., N.I. and A.T. performed the experiments. H.S.C.,<br />
M.G.B., M.S.-B., A.T. and J.V.F. reviewed, analyzed and interpreted the data. H.S.C.,<br />
A.T. and J.V.F. wrote the paper. All authors discussed the results and commented on<br />
the manuscript.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare competing financial interests: details accompany the full-text<br />
HTML version of the paper at http://www.nature.com/naturebiotechnology/.<br />
Published online at http://www.nature.com/naturebiotechnology/.<br />
Reprints and permissions information is available online at http://npg.nature.com/<br />
reprintsandpermissions/.<br />
1. Janib, S.M., Moses, A.S. & Mackay, J.A. Imaging and drug delivery using theranostic<br />
nanoparticles. Adv. Drug Deliv. Rev. 3, 1–2 (2010).<br />
2. Zrazhevskiy, P., Sena, M. & Gao, X. Designing multifunctional quantum dots for<br />
bioimaging, detection, and drug delivery. Chem. Soc. Rev. 3, 1–2 (2010).<br />
3. Choi, H.S. & Frangioni, J.V. Nanoparticles for biomedical imaging: Fundamentals<br />
of clinical translation. Mol. Imaging 9, 291–310 (2010).<br />
4. Courrier, H.M., Butz, N. & Vandamme, T.F. Pulmonary drug delivery systems: recent<br />
developments and prospects. Crit. Rev. Ther. Drug Carrier Syst. 19, 425–498<br />
(2002).<br />
5. Tsuda, A., Rogers, R.A., Hydon, P.E. & Butler, J.P. Chaotic mixing deep in the lung.<br />
Proc. Natl. Acad. Sci. USA 99, 10173–10178 (2002).<br />
6. Oberdorster, G. Pulmonary effects of inhaled ultrafine particles. Int. Arch. Occup.<br />
Environ. Health 74, 1–8 (2001).<br />
7. Mills, N.L. et al. Adverse cardiovascular effects of air pollution. Nat. Clin. Pract.<br />
Cardiovasc. Med. 6, 36–44 (2009).<br />
8. Wichmann, H.E. et al. Daily mortality and fine and ultrafine particles in Erfurt,<br />
Germany part I: role of particle number and particle mass. Res. Rep. Health Eff.<br />
Inst. November, 5–86, discussion 87–94 (2000).<br />
9. Moller, W. et al. Deposition, retention, and translocation of ultrafine particles from<br />
the central airways and lung periphery. Am. J. Respir. Crit. Care Med. 177, 426–432<br />
(2008).<br />
10. Harmsen, A.G., Muggenburg, B.A., Snipes, M.B. & Bice, D.E. The role of<br />
macrophages in particle translocation from lungs to lymph nodes. Science 230,<br />
1277–1280 (1985).<br />
11. Semmler-Behnke, M. et al. Biodistribution of 1.4- and 18-nm gold particles in rats.<br />
Small 4, 2108–2111 (2008).<br />
12. Geiser, M. et al. The role of macrophages in the clearance of inhaled ultrafine<br />
titanium dioxide particles. Am. J. Respir. Cell Mol. Biol. 38, 371–376<br />
(2008).<br />
13. Nemmar, A. et al. Ultrafine particles affect experimental thrombosis in an in vivo<br />
hamster model. Am. J. Respir. Crit. Care Med. 166, 998–1004 (2002).<br />
14. Kreyling, W.G. et al. Translocation of ultrafine insoluble iridium particles from lung<br />
epithelium to extrapulmonary organs is size dependent but very low. J. Toxicol.<br />
Environ. Health A 65, 1513–1530 (2002).<br />
15. Semmler, M. et al. Long-term clearance kinetics of inhaled ultrafine insoluble<br />
iridium particles from the rat lung, including transient translocation into secondary<br />
organs. Inhal. Toxicol. 16, 453–459 (2004).<br />
16. Frangioni, J.V. In vivo near-infrared fluorescence imaging. Curr. Opin. Chem. Biol.<br />
7, 626–634 (2003).<br />
17. Gioux, S. et al. High-power, computer-controlled, light-emitting diode-based light<br />
sources for fluorescence imaging and image-guided surgery. Mol. Imaging 8,<br />
156–165 (2009).<br />
18. Troyan, S.L. et al. The FLARE intraoperative near-infrared fluorescence imaging<br />
system: a first-in-human clinical trial in breast cancer sentinel lymph node mapping.<br />
Ann. Surg. Oncol. 16, 2943–2952 (2009).<br />
19. Choi, H.S. et al. Renal clearance of quantum dots. Nat. Biotechnol. 25, 1165–1170<br />
(2007).<br />
20. Choi, H.S. et al. Tissue- and organ-selective biodistribution of NIR fluorescent<br />
quantum dots. Nano Lett. 9, 2354–2359 (2009).<br />
21. Burns, A.A. et al. Fluorescent silica nanoparticles with efficient urinary excretion<br />
for nanomedicine. Nano Lett. 9, 442–448 (2009).<br />
22. Zhang, L.W. & Monteiro-Riviere, N.A. Mechanisms of quantum dot nanoparticle<br />
cellular uptake. Toxicol. Sci. 110, 138–155 (2009).<br />
23. Brown, D.M., Wilson, M.R., MacNee, W., Stone, V. & Donaldson, K. Size-dependent<br />
proinflammatory effects of ultrafine polystyrene particles: a role for surface area<br />
and oxidative stress in the enhanced activity of ultrafines. Toxicol. Appl. Pharmacol.<br />
175, 191–199 (2001).<br />
24. Duncan, R. Polymer conjugates as anticancer nanomedicines. Nat. Rev. Cancer 6,<br />
688–701 (2006).<br />
25. Liu, W. et al. Compact cysteine-coated CdSe(ZnCdS) quantum dots for in vivo<br />
applications. J. Am. Chem. Soc. 129, 14530–14531 (2007).<br />
26. Choi, H.S. et al. Design considerations for tumour-targeted nanoparticles.<br />
Nat. Nanotechnol. 5, 42–47 (2010).<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1303
© 2010 Nature America, Inc. All rights reserved.<br />
ONLINE METHODS<br />
Synthesis of inorganic/organic nanoparticles (INPs). CdSe(ZnCdS) nanocrystals<br />
were used to produce INP1 and INP2 with cysteine and dihydrolipoic<br />
acid (DHLA)-conjugated polyethylene glycol spacer (DHLA-PEG,<br />
MW 1 kDa) coating, respectively, via modification of previously reported<br />
procedures (Supplementary Methods) 19,25 . 800 nm NIR-emitting fluorophore<br />
CW800 (LI-COR) was conjugated on the quantum dot surface in PBS,<br />
pH 7.8 (labeling ratio ≈1.4). CdTe(ZnS) core(shell) NIR quantum dots were<br />
purchased from Invitrogen (QDot 800 ITK) with carboxylated PEG (INP3)<br />
or amino PEG (INP4) ligand coating 27 . To engineer INP5, α-amino-ωcarboxylate<br />
PEG 3.4 kDa (Nektar) was conjugated through conventional EDC<br />
chemistry. INP6 to INP9 silica nanoparticles were prepared by incorporating<br />
CdSe(ZnS) quantum dots in the silica shells and conjugating CW800 fluorophores<br />
on the surface of nanospheres to engineer larger sizes of nanoparticles<br />
(Supplementary Methods) 28 .<br />
Synthesis of organic nanoparticles (ONPs). ONP1 was prepared by a reaction<br />
of Cy5.5-NHS (Amersham Biosciences) with human serum albumin (HSA;<br />
American Red Cross) in PBS at pH 7.8, followed by purification by gel-filtration<br />
chromatography (GFC) using an Econo-Pac P6 cartridge (Bio-Rad). The labeling<br />
ratio was 2.6, estimated using the extinction coefficients of HSA (ε 280nm =<br />
32,900 M −1 cm −1 ) and Cy5.5 (ε 675nm = 19,000 M −1 cm −1 ). ONP2 was obtained<br />
by simply mixing two equivalents of Cy5.5-NHS with amino mPEG 20 kDa<br />
(Nektar) in PBS at pH 7.8, followed by GFC purification (labeling ratio = 1).<br />
A size series of nanospheres based on polystyrene and polyacrylate block<br />
copolymers (ONP3–ONP7) were purchased from Phosphorex and conjugated<br />
covalently to Cy5.5 (5–10 fluorophores per mole of polyacrylate) to produce<br />
700 nm NIR fluorescence emission (Supplementary Methods).<br />
Double-lumen balloon catheter for nanoparticle administration into lungs.<br />
Administration of nanoparticles into the lung without backflow entering the<br />
trachea and obscuring mediastinal lymph nodes imaging was a major challenge.<br />
To solve this problem, we developed a custom-built, double-lumen<br />
balloon catheter that blocked mucociliary flow into the trachea while animals<br />
were being ventilated (Supplementary Fig. 3). A custom-made 20 cm<br />
double-lumen balloon catheter was developed by Pelham Plastics with the<br />
following specifications: OD = 1.5 mm; a thru lumen = 0.71 mm for ventilation;<br />
a balloon lumen = 0.5 mm; inflation balloon OD = 2.5 mm. The catheter<br />
was inserted through a trachea cannula (OD = 2.1 mm, ID = 1.8 mm) into<br />
the right mainstem bronchus and the outer balloon was inflated. The inflated<br />
balloon remained in position for 1 h to block backflow of deposited nanoparticles.<br />
The thru lumen was used for administration of nanoparticle solutions<br />
deep into the lung followed by a puff of air during an inspiration phase to<br />
ensure delivery of a nanoparticle bolus deep into the lung. Both side lungs<br />
were ventilated for 60 min.<br />
Animal models. All animals were used under the supervision of an approved<br />
institutional protocol. 500 to 550 g Sprague-Dawley (SD) male rats from Charles<br />
River Laboratories were anesthetized with 65 mg/kg intraperitoneal pentobarbital<br />
and were mechanically ventilated by tracheostomy using an MRI-1<br />
ventilator (CWE) at 80 breaths per min using room air. A specially designed<br />
double-lumen balloon catheter was placed in the right mainstem bronchus as<br />
described above, and a minimal volume (50–100 μl) of nanoparticle solution<br />
(a mixture of 800 nm INPs and/or 700 nm ONPs) in PBS was then administered<br />
through the thru lumen into lungs at a dose of 10 pmol/g of animal weight. After<br />
each study, animals were euthanized by intraperitoneal injection of 200 mg/kg<br />
pentobarbital, a method consistent with the recommendations of the Panel on<br />
Euthanasia of the American Veterinary Medical Association.<br />
Real-time intraoperative NIR fluorescence imaging. The dual-channel intraoperative<br />
NIR fluorescence imaging system (FLARE) optimized for animal<br />
surgery has been described in detail previously 17,18 . Excitation fluence rate for<br />
white light, 700 nm NIR excitation light and 800 nm NIR excitation light were<br />
20,000 lux, 3.5 mW/cm 2 and 10 mW/cm 2 , respectively. The translocation of<br />
nanoparticles into lymph nodes and bladder was measured using the following<br />
method: the fluorescence (Fl) and background (BG) intensity of a region<br />
of interest over each organ was quantified using custom FLARE software at<br />
the following time points: 0, 1, 2, 5, 10, 20, 30, 40, 50 and 60 min. Signal-tobackground<br />
ratio (SBR), defined as SBR = (Fl–BG)/BG, was then plotted as<br />
a function of time. For measuring nanoparticle concentration in blood, ~20<br />
μl of blood was collected from the tail vein using glass capillary tubes, and<br />
the SBR was measured over time. At least three animals were analyzed at each<br />
time point. Statistical analysis was carried out using the unpaired Student’s<br />
t-test or one-way analysis of variance (ANOVA). Results were presented<br />
as mean ± s.d. and curve fitting was performed using Prism version 4.0a<br />
software (GraphPad).<br />
99m Tc-labeling of nanoparticles and radioscintigraphic imaging. 99m Tc-conjugated<br />
nanoparticles were prepared using high-specific-activity N-hydroxysuccinimide<br />
(NHS) ester of 99m Tc-MAS 3 , as described previously 19,20,26 .<br />
50–100 μl of 99m Tc-conjugated nanoparticles (100–300 μCi) were administered<br />
intratracheally into Sprague-Dawley rats. Sodium pertechnetate (TcO 4 − ) was<br />
used as a control. Measurement of blood clearance was performed by intermittent<br />
sampling of the tail vein. Rats were euthanized at 1 h after administration.<br />
To measure total urinary excretion, the bladder was removed en masse<br />
and combined with excreted urine before measurement of radioactivity in<br />
a dose calibrator. The radioactivity of each resected major organ was measured<br />
on a Wallac Wizard (PerkinElmer) 10-detector gamma counter. Gamma<br />
radioscintigraphy was performed with a Research Digital Camera (Isocam<br />
Technologies) equipped with a 1/2″ NaI crystal, 86 photomultiplier tubes and<br />
high-resolution (1 mm) lead collimator.<br />
Histological analysis of tissues. Tissues were resected 1 h after administration,<br />
fixed in 10% formalin for 3 h, molded with Tissue-Tek OCT compound (Fisher<br />
Scientific) and frozen in liquid nitrogen. Frozen sections were cut to a 10 μm<br />
thickness and stained with hematoxylin and eosin (Histotec Laboratories).<br />
Lung, liver, spleen, kidney, pancreas, intestine, muscle and lymph nodes<br />
were examined by a 4-channel fluorescence microscope. The excitation and<br />
emission filters used were 650 ± 22 nm and 710 ± 25 nm for NIR 700 nm, and<br />
760 ± 20 nm and 790 nm longpass for NIR 800 nm, respectively.<br />
27. Kim, S. et al. Near-infrared fluorescent type II quantum dots for sentinel lymph<br />
node mapping. Nat. Biotechnol. 22, 93–97 (2004).<br />
28. Insin, N. et al. Incorporation of iron oxide nanoparticles and quantum dots into<br />
silica microspheres. ACS Nano 2, 197–202 (2008).<br />
nature biotechnology<br />
doi:10.1038/nbt.1696
l e t t e r s<br />
Suppressing resistance to Bt cotton with sterile<br />
insect releases<br />
Bruce E Tabashnik 1 , Mark S Sisterson 2 , Peter C Ellsworth 3 , Timothy J Dennehy 1,4 , Larry Antilla 5 ,<br />
Leighton Liesner 5 , Mike Whitlow 5 , Robert T Staten 6 , Jeffrey A Fabrick 7 , Gopalan C Unnithan 1 , Alex J Yelich 1 ,<br />
Christa Ellers-Kirk 1 , Virginia S Harpold 1 , Xianchun Li 1 & Yves Carrière 1<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Genetically engineered crops that produce insecticidal toxins<br />
from Bacillus thuringiensis (Bt) are grown widely for pest<br />
control 1 . However, insect adaptation can reduce the toxins’<br />
efficacy 2–5 . The predominant strategy for delaying pest<br />
resistance to Bt crops requires refuges of non-Bt host plants to<br />
provide susceptible insects to mate with resistant insects 2–7 .<br />
Variable farmer compliance is one of the limitations of this<br />
approach. Here we report the benefits of an alternative strategy<br />
where sterile insects are released to mate with resistant insects<br />
and refuges are scarce or absent. Computer simulations<br />
show that this approach works in principle against pests with<br />
recessive or dominant inheritance of resistance. During a largescale,<br />
four-year field deployment of this strategy in Arizona,<br />
resistance of pink bollworm (Pectinophora gossypiella) to Bt<br />
cotton did not increase. A multitactic eradication program that<br />
included the release of sterile moths reduced pink bollworm<br />
abundance by >99%, while eliminating insecticide sprays<br />
against this key invasive pest.<br />
Transgenic cotton and corn that produce proteins from Bacillus<br />
thuringiensis (Bt) for insect control have been planted on a cumulative<br />
total of >200 million ha worldwide since their commercial introduction<br />
in 1996 (ref. 1). Although Bt crops remain effective against most<br />
pest populations, several pests have evolved resistance 2–5 . The main<br />
strategy for delaying pest resistance to Bt crops promotes survival of<br />
susceptible insects by providing host plants that do not produce Bt<br />
toxins 6,7 . These are commonly referred to as ‘refuges’. Ideally, most of<br />
the rare, resistant insects emerging from Bt crops will mate with the<br />
relatively abundant susceptible insects from nearby refuges. If resistance<br />
is inherited as a recessive trait, the Bt crops will kill the hybrid<br />
progeny produced by such matings and evolution of resistance will<br />
be substantially slowed 6,7 .<br />
Retrospective evaluations of global resistance monitoring data<br />
suggest that refuges have delayed pest resistance to Bt crops 3,7 . In<br />
particular, theoretical and empirical analyses imply that refuges<br />
have delayed resistance in pink bollworm (Pectinophora gossypiella),<br />
one of the world’s most destructive pests of cotton 7–10 . This invasive<br />
insect, which was first detected in the United States in 1917, feeds<br />
almost exclusively on cotton in some parts of the southwestern United<br />
States, including Arizona 9,10 . Field and greenhouse data show that<br />
transgenic cotton that produces Bt toxin Cry1Ac (Bt cotton) kills<br />
essentially 100% of susceptible pink bollworm larvae 11–14 . However,<br />
laboratory selection with Cry1Ac quickly produced several resistant<br />
strains of pink bollworm from Arizona that could survive on Bt cotton<br />
plants 11,12,15 . Furthermore, pink bollworm resistance to Bt cotton<br />
has been reported in the field in India, where farmer compliance with<br />
the refuge strategy has been low 5,16 . In contrast, compliance with the<br />
refuge strategy is considered a primary reason why pink bollworm<br />
susceptibility to Bt cotton did not decrease in the field in Arizona<br />
from 1997 to 2005 (refs. 8,17).<br />
As part of a coordinated, multitactic effort to eradicate pink bollworm<br />
from the southwestern United States and northern Mexico, a<br />
new strategy that replaced refuges with season-long releases of sterile<br />
pink bollworm moths was initiated in Arizona in 2006 (refs. 18,19)<br />
(Online Methods). Under this new strategy, Arizona cotton growers<br />
were permitted to plant up to 100% transgenic cotton that produces<br />
either one Bt toxin (Cry1Ac) or two Bt toxins (Cry1Ac and<br />
Cry2Ab) 18,19 . The concept underlying this alternative approach is that<br />
if enough sterile moths are released, resistant moths will mate mainly<br />
with sterile moths, rather than with fertile, wild moths that are either<br />
resistant or susceptible to Bt. In principle, this approach has several<br />
advantages over the refuge strategy. First, farmers could greatly reduce<br />
or eliminate planting of refuges and thus avoid associated complications<br />
and yield losses. Moreover, because matings with sterile insects<br />
do not produce fertile progeny, this approach could delay resistance<br />
that is based on either recessive or dominant inheritance. Unlike the<br />
refuge strategy, this approach does not require maintenance of pest<br />
populations. It is thus compatible with eradication efforts. To test<br />
the idea of delaying resistance with sterile insect releases, we conducted<br />
computer simulations and analyzed more than a decade of<br />
field data from before and after deployment of this strategy statewide<br />
in Arizona.<br />
In the computer simulations, sterile moth releases suppressed<br />
resistance to Bt cotton by decreasing the pest’s population size and<br />
1 Department of Entomology, University of Arizona, Tucson, Arizona, USA. 2 USDA-ARS, San Joaquin Valley Agricultural Sciences Center, Parlier, California, USA.<br />
3 Department of Entomology, University of Arizona, Maricopa Agricultural Center, Maricopa, Arizona, USA. 4 Monsanto Company, St. Louis, Missouri, USA. 5 Arizona<br />
Cotton Research & Protection Council, Phoenix, Arizona, USA. 6 USDA-APHIS, retired. 7 USDA-ARS, US Arid Land Agricultural Research Center, Maricopa, Arizona,<br />
USA. Correspondence should be addressed to B.E.T. (brucet@cals.arizona.edu).<br />
Received 17 August; accepted 8 October; published online 7 November 2010; doi:10.1038/nbt.1704<br />
1304 VOLUME 28 NUMBER 12 DECEMBER 2010 nature biotechnology
l e t t e r s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Years to resistance<br />
>20<br />
15<br />
10<br />
5<br />
*<br />
0<br />
0<br />
5<br />
Steriles<br />
Low<br />
Very low<br />
None<br />
15 20<br />
reducing the probability of mating between resistant moths (Fig. 1,<br />
Supplementary Fig. 1 and Supplementary Methods). In the simulations,<br />
when sufficient numbers of sterile moths were released, pest<br />
populations did not persist and resistance did not occur over a 20-year<br />
period (Fig. 1 and Supplementary Fig. 1). Based on experimental<br />
evidence of pink bollworm responses to Bt cotton 11–13,15,17 , we first<br />
modeled recessive inheritance of resistance with a fitness cost and<br />
incomplete resistance (Fig. 1 and Supplementary Table 1). With no<br />
refuges, resistance evolved in 3 years without releases of sterile moths,<br />
but populations did not persist and resistance did not occur with<br />
weekly ‘low’ releases of 0.62 sterile moths per ha of Bt cotton (Fig. 1).<br />
With refuges accounting for 2 to 20% of the total area planted to<br />
cotton, resistance evolved more slowly with increases in the number<br />
of sterile moths released, increases in the refuge percentage, or both<br />
(Fig. 1). With 20% of cotton planted to non-Bt cotton refuges, resistance<br />
did not occur in 20 years, even without sterile releases (Fig. 1).<br />
Because of fitness costs associated with pink bollworm resistance to<br />
Bt cotton, higher refuge percentages not only reduced the proportion of<br />
the population exposed to selection for resistance but also increased<br />
selection against resistance 8,20 . In a hypothetical worst-case scenario<br />
with dominant inheritance of resistance and no refuges, resistance<br />
evolved in 1 year with no sterile moths, but populations did not persist<br />
and resistance did not occur with weekly releases of 78 sterile moths<br />
per ha of Bt cotton (Supplementary Fig. 1).<br />
The mean release rate of sterile pink bollworm moths achieved from<br />
2006 to 2009 in Arizona was >600 times higher than the simulated rate<br />
that suppressed recessive resistance to Bt cotton for >20 years without<br />
refuges (Fig. 1). For each year from 2006 to 2009, the mean number<br />
of sterile moths released per ha per week was 380 (range = 170–830)<br />
for Bt cotton and 4,400 (range = 3,900–5,200) for non-Bt cotton, with<br />
10<br />
Refuge (%)<br />
Figure 1 Computer simulations of effects of sterile moth releases on<br />
evolution of resistance to Bt cotton. We used a stochastic, spatially<br />
explicit model with a region of 400 cotton fields of 15 ha each<br />
(Supplementary Methods). Two alleles (‘r’, resistant; ‘s’, susceptible)<br />
at a single locus controlled larval survival, which was 0% for ss and rs<br />
(recessive inheritance), and 15% for rr on Bt cotton; 20.8% for ss and<br />
rs, and 17.7% for rr on non-Bt cotton. The initial r allele frequency was<br />
0.018. The criteria for resistance were an r allele frequency >0.50 and a<br />
population size >10% of the initial population size. Each point represents<br />
the median of ten simulations. The simulated sterile moth release rate<br />
was 10 times higher in non-Bt cotton fields than in Bt cotton fields to<br />
mimic field releases. We simulated three sterile release rates: none, very<br />
low and low. Sterile release rates (in moths per ha per week) were 0.16 in<br />
Bt cotton and 1.6 in non-Bt cotton for the very low release rate, and 0.62<br />
in Bt cotton and 6.2 in non-Bt cotton for the low release rate. The actual<br />
release rates in the field were >600 times higher than the low release<br />
rate (see text). With no refuges and the low release rate, the regional<br />
population size decreased to 0 after 2 to 4 years (indicated by asterisk).<br />
In the cases with refuges where resistance did not evolve in 20 years, the<br />
regional population persisted but the r allele frequency remained 25% in all<br />
years, with a mean of 37.4% for 1997 to 2005 (Supplementary Fig. 2).<br />
With the onset of sterile releases, the statewide refuge percentage<br />
declined to 15.4% in 2006, 8.4% in 2007, 2.3% in 2008 and 3.1% in<br />
2009 (Supplementary Fig. 2).<br />
Consistent with the simulation results (Fig. 1), monitoring of pink<br />
bollworm field populations showed no net decrease in susceptibility<br />
to Cry1Ac from 1997 to 2005, when the refuge percentage was<br />
>25% every year, or from 2006 to 2009, when sterile insects were<br />
released and the mean refuge percentage was 7.3% (Fig. 2 and<br />
Supplementary Fig. 2). DNA screening for the three mutations in<br />
the cadherin gene that are linked with pink bollworm resistance to<br />
Cry1Ac did not identify any resistant alleles during 2006 to 2009<br />
(n = 2,499) (Online Methods). Based on larval survival on diet treated<br />
with Cry1Ac, bioassays detected a single resistant individual during 2006<br />
(n = 3,822), but no resistant individuals were found during 2007 or 2008<br />
(n = 3,602) (Fig. 2) (Online Methods). Bioassays also detected no larvae<br />
resistant to Cry2Ab during 2007 or 2008 (n = 2,572). As detailed<br />
below, in 2009, this pest was so scarce in Arizona that we could not<br />
collect enough individuals to conduct bioassays.<br />
Since the eradication program began in 2006, pink bollworm populations<br />
have declined dramatically (Fig. 3). In 2009, only two pink<br />
bollworm larvae were found in 16,600 bolls of non-Bt cotton screened<br />
statewide. This yields an infestation rate of 0.012%, which represents<br />
a 99.9% decline from the 15.3% infestation rate in 2005 (Fig. 3A).<br />
Likewise, the number of wild male pink bollworm moths caught per<br />
trap per week dropped from 26.7 in 2005 to 0.0054 in 2009, a 99.98%<br />
decrease (Fig. 3b). The decrease in pink bollworm populations during<br />
the eradication program was steeper than the decline observed<br />
with the planting of Bt cotton before the eradication program began<br />
in Arizona (Fig. 3) 21 and the declines in other target pests associated<br />
with planting of Bt crops in other regions 22–26 .<br />
Along with declines in pink bollworm populations, insecticide<br />
sprays against this pest fell to historic lows (Fig. 4). The mean<br />
number of sprays per ha per year targeting pink bollworm in Arizona<br />
was 2.7 from 1990 to 1995, which dropped to 0.64 from 1996 to 2005<br />
with use of Bt cotton, before the eradication program (Fig. 4) 10,27 .<br />
Under the eradication program, this mean decreased to 0.14 in 2006,<br />
0.013 in 2007, 0.0029 in 2008 and 0 in 2009 (Fig. 4). The mean<br />
yearly cost of pink bollworm to Arizona cotton growers, including<br />
Resistance allele frequency<br />
0.30<br />
0.25<br />
0.20<br />
0.15<br />
0.10<br />
0.05<br />
0<br />
Before eradication program<br />
Eradication<br />
program<br />
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008<br />
Year<br />
Figure 2 Pink bollworm resistance allele frequency (with 95% confidence<br />
intervals) in Arizona from 1997 to 2008, as estimated from laboratory<br />
bioassays with Cry1Ac (Online Methods). Data from 1997 to 2004 were<br />
reported previously 8 .<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1305
l e t t e r s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
a<br />
Infested non-Bt bolls (%)<br />
(log scale)<br />
b<br />
Moths/trap/week<br />
(log scale)<br />
100<br />
10<br />
1<br />
0.1<br />
Before eradication program<br />
0.01<br />
1997 1998 1999 2000 2001 2002 2003<br />
100<br />
10<br />
1<br />
0.1<br />
0.01<br />
Year<br />
Before eradication program<br />
Eradication<br />
program<br />
2004 2005 2006 2007 2008 2009<br />
Eradication<br />
program<br />
0.001<br />
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009<br />
Year<br />
yield losses and insecticide sprays, was $18 million for 1990 to 1995,<br />
$5.4 million from 1996 to 2005 and only $172,000 for 2006 to 2009<br />
(ref. 28). By using Bt cotton as one component of a comprehensive<br />
integrated pest management program, Arizona growers also greatly<br />
reduced insecticide use against all cotton pests, including those not<br />
killed by Bt cotton, saving a cumulative total of $200 million from<br />
1996 to 2009 (refs. 10,27).<br />
The estimated yearly cost of pink bollworm to US cotton growers<br />
before the eradication program was $32 million per year 19 , which is<br />
$2 million more than the mean annual cost of the eradication program<br />
in the United States and northern Mexico from 2006 to 2009 (ref. 29).<br />
These estimates do not include the value of the indirect advantages<br />
of reduced insecticide use associated with the eradication program,<br />
such as conservation of natural enemies that control pests other<br />
than pink bollworm, and benefits to the environment and human<br />
health 27 . Although increasing economic gains are expected if pink<br />
bollworm remains scarce and program costs decline, eradication of<br />
pink bollworm from the United States and northern Mexico remains<br />
challenging because this invasive pest is widespread and resilient 9,10 .<br />
Nevertheless, our results show that pink bollworm resistance to<br />
Figure 3 Pink bollworm abundance in Arizona before and during the<br />
eradication program. (a) Larval infestation of non-Bt cotton bolls from<br />
1997 to 2009. Analysis of covariance (Online Methods) shows that<br />
infestation (log [% infested non-Bt cotton bolls]) was significantly<br />
affected by year, treatment (before versus during the eradication<br />
program), and a year-by-treatment interaction (P ≤ 0.0001 for each<br />
factor and their interaction, r 2 = 0.97). Linear regression shows that<br />
the slope, which indicates the decrease in infestation per year, was 18<br />
times steeper from 2006 to 2009 (−0.81, r 2 = 0.97, P = 0.012) than<br />
from 1997 to 2005 (−0.044, r 2 = 0.42, P = 0.059). Data from 1997<br />
to 2005 were reported previously 14 . (b) Wild male pink bollworm moths<br />
trapped in Bt cotton fields from 1998 to 2009. Analysis of covariance<br />
shows that number of moths caught per trap per week (log transformed)<br />
was significantly affected by year, treatment (before versus during the<br />
eradication program), and a year-by-treatment interaction (P < 0.0001<br />
for each factor and their interaction, r 2 = 0.95). Linear regression shows<br />
that the slope, which indicates the change in moths trapped per year,<br />
was significantly negative from 2006 to 2009 (−1.0, r 2 = 0.92,<br />
P = 0.04), but did not differ significantly from 0 from 1998 to 2005<br />
(0.017, r 2 = 0.071, P = 0.52).<br />
Bt cotton in Arizona did not increase from 2006 to 2009, despite the<br />
low abundance of non-Bt cotton refuges. Although simulation results<br />
suggest that sterile releases alone can delay resistance to Bt crops<br />
(Fig. 1 and Supplementary Fig. 1), the dramatic decline in Arizona’s<br />
pink bollworm population probably reflects the combined effects of<br />
the sterile releases, high adoption and sustained efficacy of transgenic<br />
cotton producing either one or two Bt toxins and other control tactics<br />
used in the eradication program 18,19 .<br />
Although the sterile insect technique has been known for decades<br />
and used successfully in some cases 30,31 , the program described here<br />
is, to our knowledge, the first large-scale effort using this approach<br />
to suppress pest resistance to a transgenic crop. This program has<br />
benefitted from a strong grower commitment, public investment in<br />
sterile insect technology, a well-developed infrastructure for monitoring<br />
pink bollworm resistance and population density, virtually 100%<br />
efficacy of Bt cotton against pink bollworm, and this pest’s nearly<br />
exclusive dependence on cotton in Arizona 9,10,18,19 . We do not know<br />
whether the success of this program can be replicated with other pests<br />
or even with pink bollworm in other parts of the world. Analyses<br />
of mathematical models imply that refinements of the sterile insect<br />
technique, such as release of transgenic insects carrying a dominant<br />
lethal gene, could be more widely applicable for suppression of pests<br />
that harm crops or transmit pathogens 31–33 . Our results suggest that<br />
further exploration of such tactics could help to enhance the sustainability<br />
of Bt crops. The results reported here also illustrate the idea<br />
that Bt crops are likely to be most useful when combined with other<br />
tactics for integrated control of pests.<br />
Figure 4 Mean number of insecticide sprays per ha per year targeting<br />
pink bollworm on cotton in Arizona from 1996 to 2009. The asterisk<br />
indicates 0 sprays in 2009. Analysis of covariance (Online Methods)<br />
shows that insecticide use (log [sprays + 0.0001]) was significantly<br />
affected by year, treatment (before versus during the eradication<br />
program), and a year-by-treatment interaction (P < 0.0001 for each<br />
factor and their interaction, r 2 = 0.96). Linear regression shows that the<br />
slope, which indicates the change in sprays per year, was significantly<br />
negative from 2006 to 2009 (−1.0, r 2 = 0.98, P = 0.01), but did not<br />
differ significantly from 0 from 1996 to 2005 (−0.035, r 2 = 0.17,<br />
P = 0.17). The regression lines are not plotted because the 0 sprays<br />
for 2009 cannot be represented on the log scale used here and the<br />
regressions were calculated on a different scale (log [sprays + 0.0001]).<br />
Data from 1996 to 2005 were reported previously 27 .<br />
Sprays per year<br />
(log scale)<br />
10<br />
1<br />
0.1<br />
0.01<br />
0.001<br />
0.0001<br />
1996<br />
1997<br />
Before eradication program<br />
1998<br />
1999<br />
2000<br />
2001<br />
2002<br />
2003<br />
2004<br />
2005<br />
Eradication<br />
program<br />
2006<br />
2007<br />
2008<br />
*<br />
2009<br />
Year<br />
1306 VOLUME 28 NUMBER 12 DECEMBER 2010 nature biotechnology
l e t t e r s<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Methods<br />
Methods and any associated references are available in the online version<br />
of the paper at http://www.nature.com/naturebiotechnology/.<br />
Note: Supplementary information is available on the Nature Biotechnology website.<br />
Acknowledgments<br />
We thank N. Alphey, D. Crowder, F. Gould, W. Hutchison, S. Naranjo,<br />
G. Rosenthal, R.L. Smith and K.M. Wu for their thoughtful comments and<br />
assistance; and E. Miller, D. Parker, M. Krueger, N. Manhardt and Arizona Cotton<br />
Research and Protection Council (ACRPC) staff members for their contributions<br />
to the eradication program. This work was supported by funding from the USDA-<br />
National Institute of Food and Agriculture program, ACRPC, Arizona Cotton Growers<br />
Association, Cotton Foundation, Cotton Inc., National Cotton Council, Western IPM<br />
Center, Arizona Pest Management Center, and Monsanto and Dow AgroSciences.<br />
AUTHOR CONTRIBUTIONS<br />
M.S.S. conducted computer simulations; P.C.E. collected and summarized<br />
insecticide use data; L.A., L.L., M.W. and R.T.S. directed the eradication program<br />
and contributed to its design; J.A.F., G.C.U., A.J.Y., C.E.-K., and V.S.H. collected data;<br />
B.E.T., M.S.S., P.C.E., T. J. D., L.A., R.T.S., J.A.F., G.C.U., X. L. and Y.C. contributed<br />
to research design; Y.C. analyzed data. B.E.T. wrote the paper. All authors discussed<br />
the results and commented on the manuscript.<br />
COMPETING FINANCIAL INTERESTS<br />
The authors declare competing financial interests: details accompany the full-text<br />
HTML version of the paper at http://www.nature.com/naturebiotechnology/.<br />
Published online at http://www.nature.com/naturebiotechnology/.<br />
Reprints and permissions information is available online at http://npg.nature.com/<br />
reprintsandpermissions/.<br />
1. James, C. Global status of commercialized biotech/GM crops: 2009. ISAAA Briefs<br />
41 (ISAAA, Ithaca, New York, USA, 2009).<br />
2. Kruger, M.J., Van Rensburg, J.B.J. & Van den Berg, J. Perspective on the development<br />
of stem borer resistance to Bt maize and refuge compliance at the Vaalharts irrigation<br />
scheme in South Africa. Crop Prot. 28, 684–689 (2009).<br />
3. Tabashnik, B.E., Van Rensburg, J.B.J. & Carrière, Y. Field-evolved insect resistance to<br />
Bt crops: definition, theory, and data. J. Econ. Entomol. 102, 2011–2025 (2009).<br />
4. Storer, N.P. et al. Discovery and characterization of field resistance to Bt maize:<br />
Spodoptera frugiperda (Lepidoptera: Noctuidae) in Puerto Rico. J. Econ. Entomol.<br />
103, 1031–1038 (2010).<br />
5. Bagla, P. Hardy cotton-munching pests are latest blow to GM crops. Science 327,<br />
1439 (2010).<br />
6. Gould, F. Sustainability of transgenic insecticidal cultivars: integrating pest genetics<br />
and ecology. Annu. Rev. Entomol. 43, 701–726 (1998).<br />
7. Tabashnik, B.E., Gassmann, A.J., Crowder, D.W. & Carrière, Y. Insect resistance to<br />
Bt crops: evidence versus theory. Nat. Biotechnol. 26, 199–202 (2008).<br />
8. Tabashnik, B.E., Dennehy, T.J. & Carrière, Y. Delayed resistance to transgenic cotton<br />
in pink bollworm. Proc. Natl. Acad. Sci. USA 102, 15,389–15,393 (2005).<br />
9. Henneberry, T.J. & Naranjo, S.E. Integrated management approaches for pink<br />
bollworm in the southwestern United States. Integrated Pest Management Rev. 3,<br />
31–52 (1998).<br />
10. Naranjo, S.E. & Ellsworth, P.C. Fourteen years of Bt cotton advances IPM in Arizona.<br />
Southwest. Entomol. 35, 439–446 (2010).<br />
11. Tabashnik, B.E. et al. Frequency of resistance to Bacillus thuringiensis in field populations<br />
of pink bollworm. Proc. Natl. Acad. Sci. USA 97, 12,980–12,984 (2000).<br />
12. Tabashnik, B.E. et al. Association between resistance to Bt cotton and cadherin<br />
genotype in pink bollworm. J. Econ. Entomol. 98, 635–644 (2005).<br />
13. Carrière, Y. et al. Cadherin-based resistance to Bt cotton in hybrid strains of<br />
pink bollworm: fitness costs and incomplete resistance. J. Econ. Entomol. 99,<br />
1925–1935 (2006).<br />
14. Dennehy, T.J. et al. Susceptibility of Southwestern Pink Bollworm to Bt toxins<br />
Cry1Ac and Cry2Ab2 in 2005. in Arizona Cotton Report (P-151) (Cooperative<br />
Extension, Agricultural Experiment Station, The University of Arizona, Tucson,<br />
USDA, 2007) .<br />
15. Liu, Y.B., Tabashnik, B.E., Dennehy, T.J., Patin, A.L. & Bartlett, A.C. Development<br />
time and resistance to Bt crops. Nature 400, 519 (1999).<br />
16. Stone, G.D. Biotechnology and the political ecology of information in India. Hum.<br />
Organ. 63, 127–140 (2004).<br />
17. Carrière, Y. et al. Long-term evaluation of compliance with refuge requirements for<br />
Bt cotton. Pest Manag. Sci. 61, 327–330 (2005).<br />
18. Antilla, L. & Liesner, L. Program advances in the eradication of pink bollworm,<br />
Pectinophora gossypiella, in Arizona cotton. in Proceedings of the 2008 Beltwide<br />
Cotton Conferences, Nashville Tennessee, January 8–11, 2008, 1162–1168<br />
(National Cotton Council of America, Memphis, Tennessee, USA, 2008).<br />
19. Grefenstette, B., El-Lissy, O. & Staten, R.T. Pink Bollworm Eradication Plan in<br />
the US. (USDA, 2009). .<br />
20. Gassmann, A.J., Carrière, Y. & Tabashnik, B.E. Fitness costs of insect resistance<br />
to Bacillus thuringiensis. Annu. Rev. Entomol. 54, 147–163 (2009).<br />
21. Carrière, Y. et al. Long-term regional suppression of pink bollworm by<br />
Bacillus thuringiensis cotton. Proc. Natl. Acad. Sci. USA 100, 1519–1523<br />
(2003).<br />
22. Adamczyk, J.J. Jr. & Hubbard, D. Changes in populations of Heliothis virescens (F.)<br />
(Lepidoptera: Noctuidae) and Helicoverpa zea (Boddie) (Lepidoptera: Noctuidae) in<br />
the Mississippi Delta from 1986 to 2005 as indicated by adult male pheromone<br />
traps. J. Cotton Sci. 10, 155–160 (2006).<br />
23. Micinski, S., Blouin, D.C., Waltman, W.F. & Cookson, C. Abundance of Helicoverpa zea<br />
and Heliothis virescens in pheromone traps during the past twenty years in<br />
northwestern Louisiana. Southwest. Entomol. 33, 139–149 (2008).<br />
24. Storer, N.P., Dively, G.P. & Herman, R.A. Landscape effects of insect-resistant<br />
genetically modified crops. in Integration of Insect-Resistant Genetically Modified<br />
Crops within IPM Programs, (eds. Romeis, J., Shelton, A.M. & Kennedy, G.G.)<br />
273–302 (Springer, The Netherlands, 2008).<br />
25. Wu, K.M., Lu, Y.H., Feng, H.Q., Jiang, Y.Y. & Zhao, J.Z. Suppression of cotton<br />
bollworm in multiple crops in China in areas with Bt toxin-containing cotton. Science<br />
321, 1676–1678 (2008).<br />
26. Hutchison, W.D. et al. Areawide suppression of European corn borer with Bt maize<br />
reaps savings to non-Bt maize growers. Science 330, 222–225 (2010).<br />
27. Naranjo, S.E. & Ellsworth, P.C. Fifty years of the integrated control concept: moving<br />
the model and implementation forward in Arizona. Pest Manag. Sci. 65, 1267–1286<br />
(2009).<br />
28. Ellsworth, P.C., Fournier, A. & Smith, T.D. Cotton Insect Losses (The University<br />
of Arizona, Cooperative Extension, 2010). .<br />
29. Anonymous. Pink Bollworm Eradication Current Status. (US Department of<br />
Agriculture, February, 2009). < http://www.aphis.usda.gov/plant_health/plant_pest_<br />
info/cotton_pests/downloads/PBW_prog_update2009.pdf>.<br />
30. Krafsur, E.S. Sterile insect technique for suppressing and eradicating insect<br />
populations: 55 years and counting. J. Agric. Entomol. 15, 303–317 (1998).<br />
31. Gould, F. & Schliekelman, P. Population genetics of autocidal control and strain<br />
replacement. Annu. Rev. Entomol. 49, 193–217 (2004).<br />
32. Thomas, D.D., Donnelly, C.A., Wood, R.J. & Alphey, L.S. Insect population control<br />
using a dominant, repressible, lethal genetic system. Science 287, 2474–2476<br />
(2000).<br />
33. Alphey, N., Bonsall, M.B. & Alphey, L. Combining pest control and resistance<br />
management: synergy of engineered insects with Bt crops. J. Econ. Entomol. 102,<br />
717–732 (2009).<br />
nature biotechnology VOLUME 28 NUMBER 12 DECEMBER 2010 1307
© 2010 Nature America, Inc. All rights reserved.<br />
ONLINE METHODS<br />
Pink bollworm eradication program. The goal of this ongoing program is to<br />
eradicate pink bollworm from the United States and northern Mexico 18,19 . The<br />
program includes: (i) mapping cotton field locations, sizes and types (Bt or<br />
non-Bt); (ii) measuring pink bollworm abundance by checking cotton bolls for<br />
larval infestation and by trapping adult males; (iii) monitoring pink bollworm<br />
resistance to Bt toxins Cry1Ac and Cry2Ab; and (iv) controlling pink bollworm<br />
using Bt cotton, sterile moth releases, cultural practices, mating disruption<br />
with pheromone in non-Bt cotton and minimal insecticide applications. The<br />
grower-sponsored Arizona Cotton Research and Protection Council (ACRPC)<br />
proposed a plan to allow Arizona cotton growers to plant up to 100% Bt cotton<br />
that produces either one or two Bt toxins, with sterile insect releases instead<br />
of non-Bt cotton refuges for delaying resistance 18 . The US Environmental<br />
Protection Agency (EPA) convened a scientific advisory panel to review the<br />
plan 34 and the EPA subsequently approved the plan.<br />
The eradication program began in Texas, New Mexico and Mexico in 2001<br />
and has expanded each year since 19 . It was initiated in phases in Arizona,<br />
starting in 2006 with eastern and central Arizona (Cochise, Graham, Greenlee,<br />
Maricopa, Pima and Pinal Counties), which accounted for 83% of Arizona’s<br />
cotton acreage that year. The program extended to northwestern Arizona<br />
(La Paz and Mohave Counties) and southern California in 2007, and to southwestern<br />
Arizona (Yuma County) and Baja California in 2008. Funding for the<br />
binational program is provided by growers (80%) and the USDA-Animal Plant<br />
Health Inspection Service (APHIS) (20%) 29 .<br />
Computer simulations. To assess the potential effects of sterile moth releases<br />
on evolution of resistance to Bt cotton, we used a previously described<br />
stochastic, spatially explicit model of pink bollworm resistance to Bt cotton 35<br />
with some modifications. We modeled scenarios where resistance would evolve<br />
with limited refuges and no sterile releases, as reported for pink bollworm<br />
resistance to Bt cotton producing Cry1Ac in India 5 and for some cases with<br />
other pests 3 . As detailed in the Supplementary Methods, we based assumptions<br />
primarily on empirical data for pink bollworm. However, to conservatively<br />
test the potential for sterile moth releases to delay resistance, we also<br />
used some assumptions that overestimate the rate of resistance evolution.<br />
Supplementary Table 1 summarizes the parameter values that we examined.<br />
Sterile moth releases. The sterile moths released were from the APHIS strain<br />
of pink bollworm 36 maintained at the USDA-APHIS Pink Bollworm Rearing<br />
Facility in Phoenix, Arizona. This strain originated from Arizona and was<br />
infused yearly with wild insects until 2000. Moths were marked internally by<br />
rearing larvae on artificial diet containing a fat soluble dye 37 (oil red dye<br />
no. 2114, Passaic Color and Chemical Co.). Moths were irradiated with 200 Gy<br />
in a Shepherd 484R Cobalt-60 irradiator and stored in containers in groups<br />
of 2 million at 4 °C for 1–2 d until release. Moths were released from small<br />
airplanes such as Cessna model 206. Each plane had a tube underneath for<br />
releasing moths and a device that controlled the release rate. Moths were usually<br />
released from dawn to 11 a.m. at an altitude of roughly 150 m and a speed<br />
of ~180 km per hour. Throughout each cotton-growing season from 2006 to<br />
2009, 1.7–2.1 billion sterile pink bollworm moths were released over Arizona<br />
cotton fields. From May to October, each cotton field received releases two or<br />
three times per week. For each year from 2006 to 2009, the mean number of<br />
sterile moths released per ha per week was 380 for Bt cotton (range = 170–830)<br />
and 4,400 (range = 3900–5200) for non-Bt cotton. The release rate was higher<br />
for non-Bt cotton than for Bt cotton because larval survival and emergence of<br />
wild moths was expected to be higher in non-Bt cotton.<br />
Whereas spatial separation between refuges and Bt crops could limit the efficacy<br />
of the refuge strategy, sterile insect releases were made directly into Bt and<br />
non-Bt cotton fields. Also, although temporal asynchrony in moth emergence<br />
between refuges and Bt cotton fields could reduce the effectiveness of the refuge<br />
strategy 15 , sterile releases were made frequently throughout the season, so that<br />
sterile moths were available consistently for mating with wild moths.<br />
Refuge percentage. For 1998 to 2009, we determined the total area of cotton<br />
(Gossypium hirsutum (upland cotton) and G. barbadense (Pima cotton))<br />
planted and the area planted to non-Bt cotton in Arizona using methods<br />
similar to those described previously 17 . We calculated the refuge percentage<br />
as the area of non-Bt cotton divided by the total area of cotton, multiplied by<br />
100%. Thus, our estimate of the refuge percentage includes non-Bt cotton<br />
planted by growers who planted no Bt cotton. In each year, an experienced<br />
field crew trained by the ACRPC mapped the position of cotton fields and<br />
collected information from producers on cotton type (Bt or non-Bt) throughout<br />
Arizona. In each year, data collected by field crews were mapped with<br />
Geographic Information System (GIS) software and validated by comparing<br />
cotton field locations between field-generated paper maps and computergenerated<br />
maps. Enzyme-linked immunosorbent assays for Cry1Ac from<br />
a randomly chosen subset of fields showed that all of the fields tested had<br />
been correctly identified on the GIS map 17 . We analyzed the GIS maps using<br />
ArcView software to calculate the area planted to each type of cotton in each<br />
year. The statewide Bt cotton percentage for 1997 was reported previously 38 .<br />
For 1997 to 2009, Arizona’s yearly mean total area planted to cotton was<br />
98,000 ha (range = 58,000–123,000 ha). In addition to cotton, pink bollworm<br />
larvae in Arizona feed on okra (Abelmoschus esculentus), which typically grew<br />
on less than 150 ha per year in Arizona, approximately 1/500th (0.2%) or less<br />
of the area planted to cotton. We did not include okra in our calculations of<br />
non-Bt cotton refuges, but sterile moths were released on okra at the same<br />
rate used on non-Bt cotton.<br />
Resistance monitoring. We monitored pink bollworm resistance to Bt toxins<br />
Cry1Ac and Cry2Ab using bioassays 8,11 . We also used DNA screening to monitor<br />
resistance to Cry1Ac 8,12,13,39–41 . The bioassays can detect resistance caused<br />
by any mechanism. Because pink bollworm resistance to the diagnostic concentrations<br />
of Cry1Ac and Cry2Ab used in bioassays is recessive 11–13,15,17,42 ,<br />
however, the bioassays do not distinguish between homozygous susceptible<br />
larvae and heterozygous larvae carrying only one copy of a resistance gene. The<br />
DNA screening detects any of the three mutations in a cadherin gene that are<br />
tightly linked with resistance to Cry1Ac in several laboratory-selected strains<br />
of pink bollworm from Arizona that survive on Bt cotton plants 12,13,39–41 .<br />
Although the DNA screening can identify single resistance alleles in heterozygous<br />
insects, it detects only the three known cadherin resistance alleles.<br />
Bioassays. To monitor pink bollworm resistance to Cry1Ac and Cry2Ab, we<br />
used previously described methods for field sampling, laboratory bioassays<br />
and data analysis 8,11,42 . To monitor resistance to Cry1Ac, each year from 1997<br />
to 2008, we tested an average of 2,730 larvae from an average of 11.6 cotton<br />
fields in Arizona. The progeny of field-collected pink bollworm from each site<br />
were reared and tested separately. Neonates were tested individually for 21 d<br />
on artificial diet without toxin (control) or on diet with 10 μg Cry1Ac per ml<br />
diet, which kills susceptible homozygotes and heterozygotes but not resistant<br />
homozygotes 11 . Based on recessive inheritance of resistance, the Cry1Ac<br />
resistance allele frequency for each site was estimated as the square root of the<br />
frequency of survivors after adjustment for control mortality 11 . We calculated<br />
the 95% confidence interval for each yearly statewide mean resistance allele<br />
frequency using the bootstrap method with 10,000 repetitions 8 . Bioassay data<br />
from 1997 to 2004 were reported previously 8 . To monitor resistance to Cry2Ab,<br />
we used methods similar to those described above for Cry1Ac. Neonates were<br />
tested individually for 21 d on artificial diet without toxin (control) or on diet<br />
with a diagnostic concentration of 10 μg Cry2Ab per ml diet 42 . The numbers of<br />
larvae tested at the diagnostic concentration of Cry2Ab were 2,052 larvae from<br />
nine cotton fields in 2007 and 520 larvae from two cotton fields in 2008. The<br />
sample size was smaller in 2008 because the scarcity of pink bollworm in that<br />
year made it difficult to collect enough live individuals to conduct bioassays.<br />
DNA screening. We used previously described field sampling procedures and<br />
allele-specific PCR methods to screen for three cadherin mutations linked<br />
with pink bollworm resistance to Cry1Ac 8,12,13,39–41 . Details of the methods<br />
and results of DNA screening from 2001 to 2005 were reported previously 41 .<br />
DNA screening was completed here for the following numbers of wild pink<br />
bollworm individuals collected from the field from Arizona: 1,033 in 2006,<br />
884 in 2007, 364 in 2008 and 218 in 2009 (total = 2,499).<br />
Pink bollworm abundance. Pink bollworm abundance was measured with<br />
two complementary methods: checking bolls of non-Bt cotton for larvae and<br />
capturing male moths in Bt cotton fields with pheromone-baited traps.<br />
nature biotechnology<br />
doi:10.1038/nbt.1704
© 2010 Nature America, Inc. All rights reserved.<br />
Bolls. Pink bollworm abundance in bolls of non-Bt cotton plants in Arizona<br />
was determined from 1997 to 2009 by sampling bolls from commercial cotton<br />
fields during August to November and cutting them open to check for larvae as<br />
described previously 14 . The mean sample size per year was 18,200 bolls (range<br />
= 2,900–54,300) from 7 to 44 cotton fields.<br />
Pheromone traps. Male pink bollworm moths were captured in Bt cotton<br />
fields from 1998 to 2009 using delta traps baited with septa impregnated with<br />
4 mg of the pink bollworm female sex pheromone gossyplure (1:1 mixture<br />
of the Z,E-7,11 and Z,Z-7,11 isomers of hexadecadienyl acetate, Shin-Etsu<br />
Corporation) 43 . Traps were near field edges, usually at 0.8 m high (range =<br />
0.5–1.5 m). Traps and lures were changed every week. Traps were brought into<br />
the laboratory where moths were identified under magnification. During the<br />
eradication program when sterile moths were released, sterile males in traps<br />
were identified visually by the red dye used in their larval diet. When wild pink<br />
bollworm males were rare in traps, particularly during 2008 and 2009, pink<br />
bollworm males that did not readily show dye were subjected to additional<br />
testing as follows: each male was ground individually with a glass rod in a<br />
glass shell vial containing several milliliters of acetone. A strip of Whatman<br />
no. 4 filter paper trimmed to a point at the top was put in the glass vial. After<br />
the acetone rose to the top of the filter paper and evaporated, the male was<br />
deemed sterile if red dye appeared at the top of the filter paper and wild if no<br />
red dye was visible.<br />
We analyzed data on wild males caught in traps that were collected from<br />
the field each year from 15 April to 15 June because data for this period were<br />
available for all years from 1998 to 2009. The mean number of traps per year<br />
was 10,400 (range = 1,498–29,928) with a seasonal total of up to nine traps<br />
(one per week) at each monitoring site. More than 150 Bt cotton fields were<br />
monitored with traps each year.<br />
A previous analysis based on data from 1992 to 2001 (before the eradication<br />
program) showed significant declines in pink bollworm males captured<br />
in pheromone traps in regions of Arizona where abundance of Bt cotton was<br />
high but not in regions of Arizona where abundance of Bt cotton was low 21 .<br />
Compared with the data reported and analyzed here (Fig. 3b), the previous<br />
analysis differs in terms of the years studied (1992 to 2001 before; 1998 to 2009<br />
here), the criteria for standardizing the time period examined within years<br />
(accumulation of degree-days before; calendar dates here); spatial scale (15<br />
regions of Arizona analyzed separately before; statewide data pooled here),<br />
and the distribution of pheromone traps (traps near all cotton fields before;<br />
only traps in Bt cotton fields here). The analysis here shows a steep statewide<br />
decline in males captured in pheromone traps in Bt cotton fields during the<br />
eradication program (2006 to 2009) but not before the eradication program<br />
(1998 to 2005) (Fig. 3b).<br />
Insecticide sprays. Insecticide sprays targeting pink bollworm in Arizona<br />
cotton from 2006 to 2009 were calculated by adding the number of sprays<br />
made by growers and by the ACRPC as part of the eradication program. The<br />
ACRPC sprayed when larval infestation reached or exceeded 5% of bolls.<br />
Growers sprayed based on their own criteria. The number of sprays made<br />
by growers against pink bollworm was estimated as described previously 28 .<br />
The mean number of sprays per ha per year made by growers against<br />
pink bollworm was 0.068 in 2006, 0.0095 in 2007, 0 in 2008 and 0 in 2009<br />
(ref. 10). The mean number of sprays per ha per year made by the ACRPC<br />
against pink bollworm was 0.070 in 2006, 0.0035 in 2007, 0.00029 in 2008 and<br />
0 in 2009. Data for 1996 to 2005, which reflect only sprays made by growers,<br />
were reported previously 27 .<br />
Analysis of data on pink bollworm abundance and insecticide sprays. We<br />
calculated the percentage decrease in abundance as: 100% − [(final abundance/<br />
initial abundance) × 100%]. For example, infestation of non-Bt cotton bolls<br />
was 0.012% in 2009 (final abundance) and 15.3% in 2005 (initial abundance).<br />
The percentage decrease in abundance was 100% − [(0.012%/15.3%) × 100%],<br />
which equals 99.9%.<br />
We tested for effects of the eradication program on three response variables:<br />
(i) infestation of non-Bt cotton bolls, (ii) wild males caught per trap per<br />
week in Bt cotton fields, and (iii) insecticide sprays per ha per year targeting<br />
pink bollworm. To test for effects of the eradication program on infestation<br />
of non-Bt cotton bolls, we compared trends for two time periods: years with<br />
Bt cotton before the eradication program (1997–2005) and years during the<br />
eradication program (2006–2009). For each period, simple linear regression<br />
was used to evaluate the association between the percentage of infested bolls<br />
(log transformed) and year. We also used covariance analysis to evaluate the<br />
effects of year, treatment (before versus during the eradication program), and<br />
their interaction on the percentage of infested bolls (log transformed). In<br />
this analysis a significant interaction term indicates that the slope before the<br />
eradication program differs from the slope during the eradication program.<br />
For the covariance analysis of the boll infestation data, years before the eradication<br />
program were coded as 1 (1997) to 9 (2005) and during the eradication<br />
program as 1 (2006) to 4 (2009). As with the boll infestation data, we used<br />
simple linear regression and covariance analysis to evaluate the effects of the<br />
eradication program on the number of wild males caught per trap per week in<br />
Bt cotton fields (log transformed). For this analysis, years with available trap<br />
data before the eradication program were 1998–2005 and were coded as 1–8.<br />
We used the same approach to analyze the effects of the eradication program<br />
on the number of insecticide sprays targeting pink bollworm. Because the<br />
spray data included a zero (for 2009), we added 0.0001 to the number of sprays<br />
before performing the log transformation 44 . For sprays, years analyzed before<br />
the eradication program were 1996 to 2005 and were coded as 1–10. Analyses<br />
were performed in JMP 45 .<br />
34. U.S. Environmental Protection Agency. Scientific Advisory Panel, October 24–26,<br />
2006: Evaluation of the resistance risks from using 100% Bollgard and Bollgard II<br />
cotton as part of a pink bollworm eradication program in the state of Arizona<br />
.<br />
35. Sisterson, M.S., Antilla, L., Carrière, Y., Ellers-Kirk, C. & Tabashnik, B.E. Effects<br />
of insect population size on evolution of resistance to transgenic crops. J. Econ.<br />
Entomol. 97, 1413–1424 (2004).<br />
36. Liu, Y.B. et al. Effects of Bt cotton and Cry1Ac toxin on survival and development<br />
of pink bollworm (Lepidoptera: Gelechiidae). J. Econ. Entomol. 94, 1237–1242<br />
(2001).<br />
37. Graham, H.M. & Mangum, C.L. Larval diets containing dye for tagging pink bollworm<br />
moths internally. J. Econ. Entomol. 64, 376–379 (1971).<br />
38. Sims, M.A. et al. Arizona’s multi-agency resistance management program for Bt<br />
cotton: sustaining the susceptibility of pink bollworm. in Proceedings of the 2001<br />
Beltwide Cotton Conferences, January 9–13, 2001, Anaheim, California, 2,<br />
1175–1179 (National Cotton Council of America, Memphis, Tennessee, USA,<br />
2001).<br />
39. Morin, S. et al. Three cadherin alleles associated with resistance to Bacillus<br />
thuringiensis in pink bollworm. Proc. Natl. Acad. Sci. USA 100, 5004–5009<br />
(2003).<br />
40. Morin, S. et al. DNA-based detection of Bt resistance alleles in pink bollworm.<br />
Insect Biochem. Mol. Biol. 34, 1225–1233 (2004).<br />
41. Tabashnik, B.E. et al. DNA screening reveals pink bollworm resistance to Bt cotton<br />
remains rare after a decade of exposure. J. Econ. Entomol. 99, 1525–1530<br />
(2006).<br />
42. Tabashnik, B.E. et al. Asymmetrical cross-resistance between Bacillus thuringiensis<br />
toxins Cry1Ac and Cry2Ab in pink bollworm. Proc. Natl. Acad. Sci. USA 105,<br />
11,889–11,894 (2009).<br />
43. Tabashnik, B.E. et al. Dispersal of pink bollworm (Lepidoptera: Gelechiidae) males<br />
in transgenic cotton that produces a Bacillus thuringiensis toxin. J. Econ. Entomol.<br />
92, 772–780 (1999).<br />
44. Sokal, R.R. & Rohlf, F.J. Biometry (W.H. Freeman and Company, San Francisco,<br />
1969).<br />
45. SAS Institute. JMP 8.0, Cary, North Carolina, USA (2008).<br />
doi:10.1038/nbt.1704<br />
nature biotechnology
errata and corrigenda<br />
Erratum: GM alfalfa—who wins?<br />
Jeffrey L Fox<br />
Nat. Biotechnol. 28, 770 (2010); published online 9 August 2010; corrected after print 7 December 2010<br />
In the version of this article initially published, Geertson Seed Farms was spelled “Geerston.” The error has been corrected in the HTML and PDF<br />
versions of the article.<br />
Erratum: South-South entrepreneurial collaboration in health biotech<br />
Halla Thorsteinsdóttir, Christina C Melon, Monali Ray, Sharon Chakkalackal, Michelle Li, Jan E Cooper, Jennifer Chadder,<br />
Tirso W Saenz, Maria Carlota de Souza Paula, Wen Ke, Lexuan Li, Magdy A Madkour, Sahar Aly, Nefertiti El-Nikhely, Sachin Chaturvedi,<br />
Victor Konde, Abdallah S Daar & Peter A Singer<br />
Nat. Biotechnol. 28, 407–416 (2010); published online 7 May 2010; corrected after print 7 December 2010<br />
In the version of this article initially published, a line was missing connecting India and China in Figure 3. The error has been corrected in the<br />
HTML and PDF versions of the article.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Corrigendum: The BioPAX community standard for pathway data sharing<br />
Emek Demir, Michael P Cary, Suzanne Paley, Ken Fukuda, Christian Lemer, Imre Vastrik, Guanming Wu, Peter D’Eustachio,<br />
Carl Schaefer, Joanne Luciano, Frank Schacherer, Irma Martinez-Flores, Zhenjun Hu, Veronica Jimenez-Jacinto, Geeta Joshi-Tope,<br />
Kumaran Kandasamy, Alejandra C Lopez-Fuentes, Huaiyu Mi, Elgar Pichler, Igor Rodchenkov, Andrea Splendiani, Sasha Tkachev,<br />
Jeremy Zucker, Gopal Gopinath, Harsha Rajasimha, Ranjani Ramakrishnan, Imran Shah, Mustafa Syed, Nadia Anwar, Özgün Babur,<br />
Michael Blinov, Erik Brauner, Dan Corwin, Sylva Donaldson, Frank Gibbons, Robert Goldberg, Peter Hornbeck, Augustin Luna,<br />
Peter Murray-Rust, Eric Neumann, Oliver Reubenacker, Matthias Samwald, Martijn van Iersel, Sarala Wimalaratne, Keith Allen,<br />
Burk Braun, Michelle Whirl-Carrillo, Kei-Hoi Cheung, Kam Dahlquist, Andrew Finney, Marc Gillespie, Elizabeth Glass, Li Gong,<br />
Robin Haw, Michael Honig, Olivier Hubaut, David Kane, Shiva Krupa, Martina Kutmon, Julie Leonard, Debbie Marks, David Merberg,<br />
Victoria Petri, Alex Pico, Dean Ravenscroft, Liya Ren, Nigam Shah, Margot Sunshine, Rebecca Tang, Ryan Whaley, Stan Letovksy,<br />
Kenneth H Buetow, Andrey Rzhetsky, Vincent Schachter, Bruno S Sobral, Ugur Dogrusoz, Shannon McWeeney, Mirit Aladjem,<br />
Ewan Birney, Julio Collado-Vides, Susumu Goto, Michael Hucka, Nicolas Le Novère, Natalia Maltsev, Akhilesh Pandey, Paul Thomas,<br />
Edgar Wingender, Peter D Karp, Chris Sander & Gary D Bader.<br />
Nat. Biotechnol. 28, 935–942 (2010); published online 09 September 2010; corrected after print 7 December 2010<br />
In the version of this article initially published, the affiliation for Ken Fukuda was incorrect. The correct affiliation is Computational Biology<br />
Research Center, National Institute of Advanced Industrial Science and Technology, Tokyo, Japan. The error has been corrected in the HTML and<br />
PDF versions of the article.<br />
Corrigendum: Analysis of a genome-wide set of gene deletions in the fission<br />
yeast Schizosaccharomyces pombe<br />
Dong-Uk Kim, Jacqueline Hayles, Dongsup Kim, Valerie Wood, Han-Oh Park, Misun Won, Hyang-Sook Yoo, Trevor Duhig,<br />
Miyoung Nam, Georgia Palmer, Sangjo Han, Linda Jeffery, Seung-Tae Baek, Hyemi Lee, Young Sam Shim, Minho Lee, Lila Kim,<br />
Kyung-Sun Heo, Eun Joo Noh, Ah-Reum Lee, Young-Joo Jang, Kyung-Sook Chung, Shin-Jung Choi, Jo-Young Park, Youngwoo Park,<br />
Hwan Mook Kim, Song-Kyu Park, Hae-Joon Park, Eun-Jung Kang, Hyong Bai Kim, Hyun-Sam Kang, Hee-Moon Park, Kyunghoon Kim,<br />
Kiwon Song, Kyung Bin Song, Paul Nurse & Kwang-Lae Hoe<br />
Nat. Biotechnol. 28, 617–623 (2010); published online 16 May 2010; corrected after print 7 December 2010<br />
In the version of this article initially published, the address of one of the authors, Young-Joo Jang, was incorrect. The correct address is Laboratory<br />
of Cell Cycle & Signal Transduction, WCU Department of NanoBioMedical Science, Institute of Tissue Regeneration Engineering, Dankook<br />
University, Cheonan, Korea. The error has been corrected in the HTML and PDF versions of the article.<br />
1308 volume 28 number 12 DECEMBER 2010 nature biotechnology
careers and recruitment<br />
Catching the wave in China<br />
Connie Johnson Hambley<br />
China is presenting the pharmaceutical world with unprecedented opportunity. Here is what is happening there and<br />
how you can be a part of it.<br />
© 2010 Nature America, Inc. All rights reserved.<br />
Catching a trend at its earliest stage and<br />
reaping its rewards as it advances is easy<br />
only in hindsight. It takes more than being<br />
able to identify a trend to ensure job security:<br />
it takes understanding the factors behind the<br />
trend to gauge its strength, longevity, direction<br />
and sustainability, as well as your place<br />
in its potential opportunities. Armed with<br />
this information, you can best assess your<br />
career options.<br />
Seeds of a boom<br />
In hindsight, it is easy to see the factors that<br />
created the biotech boom in the United<br />
States. The confluence of world-class academic<br />
institutions, human talent, innovative<br />
ideas, abundant investment capital, economic<br />
incentives and enjoyment of profits<br />
combined to create a magnet for the world’s<br />
brightest who wanted to be educated and<br />
work in the United States. Sporadic layoffs in<br />
big pharma were not inherently destabilizing<br />
because biotech companies were being created<br />
fast enough to absorb the talent. Many<br />
say there was no better place to be during the<br />
boom years than in Cambridge, San Diego or<br />
Research Triangle Park.<br />
During the boom, having a PhD or handson<br />
experience was a guarantee to lifelong<br />
employment. It was easy to manage a career<br />
when you were in demand and had a pick<br />
of jobs. But now, those who can find jobs<br />
are often relocating families, taking on less<br />
desirable or part-time positions or taking<br />
pay cuts. This comes as a surprise to many<br />
who feel that they have finally arrived at<br />
the station only to find that their train has<br />
already left.<br />
In addition, venture capital investment<br />
in biotech startups has declined, with some<br />
Connie Johnson Hambley is at Steele Executive<br />
Search, Boston, Massachusetts, USA.<br />
e-mail: connie@steelesearch.com<br />
venture capitalists citing ‘investor fatigue’—<br />
characterized as a reluctance to ‘re-up’<br />
financing of laboratories and talent as a drug<br />
moves closer to the clinic. As a result, companies<br />
cut research staff and run virtually,<br />
meaning they consist of management and<br />
project managers who coordinate with contract<br />
research organizations (CROs). Drug<br />
development is increasingly moving offshore<br />
through outsourcing and big pharmas such<br />
as GlaxoSmithKline and Merck are creating<br />
R&D outposts in China, which are taking the<br />
jobs with them. In the new reality, a global<br />
industry must now include Asia, not just the<br />
United States and Europe.<br />
China is poised for a<br />
prolonged period of<br />
sustainable growth<br />
and that can translate<br />
into job longevity and<br />
career advancement,<br />
says Connie Johnson<br />
Hambley.<br />
As a result, opportunities for pharmaceutical<br />
professionals are exploding in places like<br />
India and China. However, it is China that is<br />
poised to enjoy the greatest growth. The ZRG<br />
Partners Global Life Science Hiring Index 1<br />
states that in the third quarter of 2010 the Asia<br />
Pacific region enjoyed a 20% increase in hiring<br />
over the second quarter, whereas the Americas<br />
saw a 2% decline in hiring in the same period.<br />
Even before the recent decline, the seeds for<br />
the boom in China were being sown.<br />
From ‘Made in China’ to ‘Created in China’<br />
“A tidal wave of change is coming to China,<br />
and the goal for many is to ride the wave<br />
rather than be buried by it,” says Jason Mann,<br />
who advises biopharmaceutical firms on<br />
emerging market strategy with a focus on<br />
China. “As Americans we often view the<br />
Chinese through the viewpoint of our own<br />
lifetime, as the sudden rise of a new power,<br />
but many Chinese view themselves through<br />
a historical lens of thousands of years.”<br />
According to Mann, we are now seeing the<br />
impact of reforms in China over the last 30<br />
years. He notes that some observe an unspoken<br />
contract at work between the people and<br />
the government in post-Mao China: “Give<br />
the people an ever-rising standard of living<br />
through economic reform and they will focus<br />
less on political reform.” This emphasis on<br />
economic reform combined with China’s<br />
scale means that the government can mobilize<br />
tremendous resources behind favored<br />
sectors, with an increasing focus on life science<br />
investment and growth.<br />
The Chinese government created a<br />
complex series of reforms and initiatives<br />
designed to move China from an economy<br />
based on manual labor and manufacturing<br />
to one of innovation and creation, as articulated<br />
in their most recent five-year plan.<br />
Biotechnology is a key focus, and the biotech<br />
industry in China has grown to nearly $9<br />
billion from $3 billion barely five years ago,<br />
with much of that growth heavily supported<br />
with government investment. An estimated<br />
additional $1.5 billion investment into drug<br />
development has been announced.<br />
Barriers to entrepreneurial growth have<br />
been removed. Intellectual property laws<br />
were revised and high-profile enforcement<br />
of those laws has given investors and scientists<br />
greater confidence. And with the reemergence<br />
of the stock exchange barely 20<br />
years ago, the longstanding Chinese tradition<br />
of private enterprise has been enhanced by<br />
strengthening the mechanisms for distributing<br />
ownership over a larger number of people.<br />
Chinese companies are now mastering<br />
that truly entrepreneurial holy grail—the initial<br />
public offering. All of these factors show<br />
that China is poised for a prolonged period<br />
nature biotechnology volume 28 number 12 DECEMBER 2010 1309
careers and recruitment<br />
© 2010 Nature America, Inc. All rights reserved.<br />
of sustainable growth and that can translate<br />
into job longevity and career advancement.<br />
The human factor<br />
Restructuring laws and funneling money into<br />
infrastructure are helpful supports, but it is the<br />
human talent that gets the work done. In drug<br />
discovery and development, Western-trained<br />
pharmaceutical professionals have an edge.<br />
Wang Huiyao, visiting fellow at the<br />
Brookings Institution and director general<br />
of the Center for China and Globalization,<br />
states that China has had 1.62 million students<br />
go overseas to get their education since<br />
1978. Only 497,000 have returned to China<br />
and only 8% of science and engineering PhD<br />
graduates have returned. As China builds its<br />
industry, the lack of seasoned professionals<br />
is strongly felt.<br />
A hallmark of Western R&D culture is to<br />
question what is being presented and not to<br />
merely accept information. Scientific rigor<br />
is the intellectual process of innovation<br />
and creativity where peers and superiors<br />
engage in a dynamic dialog of questions<br />
and critiques requiring scientists to push<br />
back with rebuttals to support their findings.<br />
Weaknesses are identified, unusual<br />
relationships are discovered and creative<br />
problem-solving results. This is in contrast<br />
to traditional Chinese education, which has<br />
a more top-down approach, where challenging<br />
an authority can be uncomfortable. This<br />
cultural ethic produces high-quality skills<br />
uniquely suited to drug discovery.<br />
In 2008, China launched its ‘Thousand<br />
Talents’ program designed to change China’s<br />
economy from an investment-driven to a<br />
more human resources–driven one. With its<br />
large population viewed as a valuable natural<br />
resource, the Thousand Talents program is<br />
designed to invest in and cultivate human<br />
capital. Business entrepreneurs, technical<br />
professionals and highly skilled workers<br />
make up three of the six targeted categories<br />
and each is essential for creating and<br />
sustaining a biotech and life science boom.<br />
The plan lures highly trained professionals<br />
and researchers to its shores with financial<br />
incentives and then uses those professionals<br />
to further train local talent.<br />
China is also pouring money and resources<br />
into revising and deepening their educational<br />
systems to address the needs of their population<br />
as a whole and to supply the human<br />
capital that their targeted industries require.<br />
The Thousand Talents program acknowledges<br />
that it is not only book training that<br />
young minds need, but experienced professionals<br />
to guide them and impart knowledge.<br />
And these professionals need not just be ‘sea<br />
turtles’, or Chinese returnees—the plan specifically<br />
allows for the recruitment of foreign<br />
nationals. Western-trained pharmaceutical<br />
talent receives the most interest.<br />
“The young scientists in China are hungry<br />
for knowledge. They want to learn and<br />
they want to learn from the best,” says Yun<br />
He, chief scientific officer of BioDuro, the<br />
Beijing subsidiary of US-based CRO giant<br />
PPD. “The most important factor for success<br />
is to have the right talent. The people part is<br />
hard. [The] money part is easier.” Focusing<br />
on cultivating talent in-house, BioDuro<br />
launched the BioDuro Learning Institute in<br />
2009 using professional instructors and their<br />
own senior management to provide instruction<br />
on everything from management skills<br />
to good laboratory practices to English.<br />
Limited time of opportunity<br />
Language is not a barrier to working in<br />
China. “Most scientists are surprised when<br />
they walk into a lab in Beijing and sit down<br />
and interact with a team,” states John Oyler,<br />
president and CEO of BeiGene, a startup<br />
cancer research company in Beijing. English<br />
is the predominant language in most laboratories<br />
because senior and middle managers<br />
received essential scientific training in the<br />
United States. But Oyler cautions there could<br />
be “a short window of time” for a non-Chinese<br />
speaking employee to reap all of the rewards<br />
of the current climate.<br />
Both He and Oyler roughly estimate the true<br />
unequalled opportunity horizon at five years.<br />
“The quality that already exists in China makes<br />
it almost too late for some synthetic chemistry<br />
and perhaps structural biology professionals,”<br />
continues Oyler, citing the more systematic<br />
and formulaic disciplines more suited to current<br />
Chinese educational standards. “Medicinal<br />
chemistry scientists have a shorter window.<br />
Folks in development or other more specialized<br />
areas of biology may have ten years and translational<br />
research may be even longer given the<br />
sheer number of patients and the importance<br />
of proximity to the clinic.” These time horizon<br />
estimates are supported by both the five-year<br />
plan and the Thousand Talents program goals.<br />
Certainly stellar talent knows no geographical<br />
bounds. If you are preeminent in your field, you<br />
will have opportunities regardless of country.<br />
So what does this mean for you?<br />
In China, Western-trained scientists have<br />
an edge most are not aware of, and it comes<br />
in the very essence of their training. The<br />
pharmaceutical industry has always shifted<br />
workers around the globe to meet its needs<br />
and China is the new frontier. Its growth is<br />
not a passing trend but a major force in its<br />
early stages. Individual employees may be<br />
powerless to avoid being laid off, but they<br />
are not powerless to take advantage China’s<br />
opportunities.<br />
COMPETING FINANCIAL INTERESTS<br />
The author declares competing financial<br />
interests: details accompany the full-text HTML<br />
version of the paper at http://www.nature.com/<br />
naturebiotechnology/.<br />
ACKNOWLEDGMENTS<br />
C.J.H. would like to thank Michael Russo at the<br />
Bruckner Group for his help in the preparation of<br />
this article.<br />
1. ZRG Partners. Global Life Science Hiring Index,<br />
http://www.zrgpartners.com/Documents/lifescience<br />
indexq32010.pdf<br />
1310 volume 28 number 12 DECEMBER 2010 nature biotechnology
people<br />
© 2010 Nature America, Inc. All rights reserved.<br />
PolyTherics (London) has announced the appointment of John Burt<br />
(left) as chief business officer. He was most recently CEO and cofounder<br />
of Thiakis. His previous posts include positions at Imperial<br />
Innovations, GlaxoSmithKline and Vernalis.<br />
“We are delighted that John has joined us and are sure that his<br />
experience, notably with Thiakis and GSK, will be invaluable,”<br />
says PolyTherics CEO Keith Powell. “We now have a management<br />
team which, together with our chairman, Ken Cunningham, has<br />
the breadth of experience with both biotech and large pharma<br />
companies to take PolyTherics to the next stage in its evolution.”<br />
Vical (San Diego) has announced the appointment<br />
of Igor P. Bilinsky as senior vice president,<br />
corporate development. Bilinsky was<br />
previously at Halozyme Therapeutics, where he<br />
served most recently as vice president, business<br />
development and special operations.<br />
Aastrom Biosciences (Ann Arbor, MI, USA) has<br />
named Ronald Cresswell to its board of directors.<br />
Cresswell is the former senior vice president<br />
and CSO of Warner-Lambert and has served on<br />
the board of directors of Esperion Therapeutics,<br />
Allergan, Curagen and Vasogen.<br />
Milind S. Deshpande has been promoted to<br />
president of R&D and retains his position as<br />
CSO, and Mary Kay Fenton has been promoted<br />
to senior vice president and retains her<br />
position as CFO at Achillion Pharmaceuticals<br />
(New Haven, CT, USA). Deshpande previously<br />
served as the company’s executive vice<br />
president of R&D and Ms. Fenton served as<br />
vice president.<br />
Jeffrey D. Edelson has been appointed chief<br />
medical officer of Palatin Technologies<br />
(Cranbury, NJ, USA). Edelson has more than<br />
15 years of clinical development experience<br />
in the pharma and biotech industries, most<br />
recently serving as executive vice president<br />
of R&D and chief medical officer of Ikano<br />
Therapeutics. Prior to that, he was vice president<br />
and therapeutic area head of novel therapeutics<br />
for Johnson & Johnson Pharmaceutical<br />
Research and Development. In addition, Palatin<br />
announced the resignation of Trevor Hallam<br />
as executive vice president of R&D, effective<br />
December 31. His departure results from the<br />
company’s decision last quarter to eliminate<br />
research and discovery activities, and focus on<br />
advancing its phase 2 clinical candidates.<br />
Ambit Biosciences (San Diego) has named<br />
biotech industry veteran Faheem Hasnain as<br />
chairman of the board and Alan Fuhrman<br />
as CFO. From December 2008 until its<br />
acquisition by Abbott Laboratories in April,<br />
Hasnain was president, CEO and a director<br />
of Facet Biotech. Previously he was with PDL<br />
BioPharma, Biogen Idec, Bristol-Myers Squibb<br />
and GlaxoSmithKline. Fuhrman joins Ambit<br />
from Naviscan, where he served as vice president<br />
and CFO. He also held positions at Sonus<br />
Pharmaceuticals and Integrex.<br />
James Patrick Kelly has been appointed senior<br />
vice president, CFO, treasurer and secretary of<br />
Vanda Pharmaceuticals (Rockville, MD, USA).<br />
Kelly has more than 18 years of financial and<br />
operating experience in the healthcare industry,<br />
most recently as vice president, controller<br />
at MedImmune.<br />
The board of directors of Athersys (Cleveland,<br />
OH, USA) has elected Ismail Kola as a director<br />
of the company. Kola currently serves as executive<br />
vice president of Belgium-based pharma<br />
company UCB and as president of UCB New<br />
Medicines, UCB’s discovery research through<br />
proof-of-concept organization since November<br />
2009. He currently serves on the board of<br />
directors of Astex Pharmaceuticals, Synosia<br />
and Ondek, and has previously served on the<br />
board of directors for companies such as Eragen<br />
Biotech, Promega and Ingene.<br />
Regeneron Pharmaceuticals (Tarrytown, NY,<br />
USA) has elected Christine A. Poon to fill<br />
a new seat on its expanded board of directors.<br />
Poon currently serves as dean of Ohio<br />
State University’s Fisher College of Business.<br />
She was formerly vice chairman of the board<br />
of directors of Johnson & Johnson and<br />
worldwide chairman of the Johnson & Johnson<br />
Pharmaceuticals Group.<br />
Vestaron (Kalamazoo, MI, USA) has announced<br />
the election of John Sorenson to the position of<br />
chairman and Elin Miller to the position of vice<br />
chairman of the company’s board of directors.<br />
Both were previously members of the board.<br />
Former Chairman Eli Thomssen remains on<br />
the board. Sorenson was formerly president of<br />
Syngenta Biotechnology and before that president<br />
of Syngenta Seeds. Miller previously held<br />
senior management positions with Dow, Dow<br />
AgroSciences and ArystaLifeScience.<br />
Amarin (Dublin, Ireland, and Mystic, CT,<br />
USA) has announced the appointment of two<br />
replacements for president, CEO and director<br />
Colin Stewart, who resigned in November to<br />
address personal matters. Joseph S. Zakrzewski,<br />
who has served as Amarin’s chairman of the<br />
board since January 2010, has been named<br />
CEO while continuing to serve as chairman.<br />
He most recently served as CEO of Xcellerex.<br />
Additionally, John F. Thero has been appointed<br />
president of Amarin. Since November 2009,<br />
Thero had been the company’s CFO.<br />
Gentris (Morrisville, NC, USA) has appointed<br />
Sam C. Tetlow executive chairman of the<br />
company’s board of directors. He currently<br />
serves as vice president of business development<br />
for Calvert Research and as a director<br />
of Immunologix. In addition, Gentris<br />
has named Howard McLeod as the company’s<br />
first chief scientific advisor. McLeod<br />
is the Fred N. Eshelman Distinguished<br />
Professor and director of the Institute for<br />
Pharmacogenomics and Individualized<br />
Therapy at the University of North Carolina<br />
at Chapel Hill. He serves as principal investigator<br />
for the CREATE Pharmacogenomics<br />
Research Network and is a member of the<br />
FDA Committee on Clinical Pharmacology.<br />
Timothy P. Walbert has been elected to the<br />
board of directors of antibody therapeutics<br />
developer XOMA (Berkeley, CA, USA). A<br />
seasoned executive with experience launching<br />
therapeutic products including Abbott<br />
Laboratories’ Humira (adalimumab), Walbert<br />
currently serves as chairman, president and<br />
CEO of privately held Horizon Pharma.<br />
1312 volume 28 number 12 DECEMBER 2010 nature biotechnology