Nature Biotechnology 02 2010 (magazine journal; February 2010)

volume 28 number 2 february 2010

e d i tor i a l 103 PhRMA wants you! 103 Impossible fixes

© 2010 Nature America, Inc. All rights reserved.

Transmission electron micrograph of hepatitis C viruses. Rice and colleagues present a method to detect individual hepatitis C virus–infected cells using a fluorescent reporter molecule (p 167). Credit: James Cavallini/Photo Researchers, Inc.

news 105 PML problems loom for Rituxan 107 Pfizer stakes a claim in plant cell–made biopharmaceuticals 109 Amylin’s $1 billion heavyweight deal 110 Peruvian GM advocate faces criminal charges 110 EC convenes crisis talks on European biotech sector 112 Irish bioethics council axed 112 Amgen trumps Roche 112 Report blames GM crops for herbicide spike, downplays pesticide reductions 114 Industry gains on money-back schemes 114 $2 million rice verdict against Bayer 114 Biorefineries’ stimulus win 115 Purpose-built chromosome 115 FDA balks on MedImmune’s cell-grown flu vaccine 116 data page: 2009: Turning the corner 117 News feature: The HER2 testing conundrum

B i oe n trepre n e u r B u i l d i n g a b u s i n ess 120 US increased herbicide use linked with GM crops, p 112

Coming to terms David H Oden, Jeffrey A Wolfson & Christina W Marshall

op i n i o n a n d comme n t 123 126 128

C O R R E S P O ND E N C E Fab-arm exchange An economic and technical evaluation of microalgal biofuels Ontology engineering

Nature Biotechnology (ISSN 1087-0156) is published monthly by Nature Publishing Group, a trading name of Nature America Inc. located at 75 Varick Street, 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, 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 (corporate institution). Canada add 5% GST #104911595RT001; Euro-zone: €202 (personal), €2,795 (institution), €3,488 (corporate institution); Rest of world (excluding China, Japan, Korea): £130 (personal), £1,806 (institution), £2,250 (corporate institution); Japan: Contact NPG Nature Asia-Pacific, Chiyoda Building, 2-37 Ichigayatamachi, Shinjuku-ku, Tokyo 162-0843. Tel: 81 (03) 3267 8751, Fax: 81 (03) 3267 8746. POSTMASTER: Send address changes to Nature Biotechnology, Subscriptions Department, 342 Broadway, PMB 301, New York, NY 10013-3910. Authorization to photocopy material for internal or personal 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 (CCC) Transactional Reporting Service, provided the relevant copyright fee is paid direct to CCC, 222 Rosewood Drive, Danvers, MA 01923, USA. Identification code for Nature Biotechnology: 1087-0156/04. Back issues: US$45, Canada add 7% for GST. CPC PUB AGREEMENT #40032744. Printed by Publishers Press, Inc., Lebanon Junction, KY, USA. Copyright © 2010 Nature Publishing Group. Printed in USA.

i

volume 28 number 2 february 2010 C O M M E N TA R Y 131

case study: Never again Christopher Scott

133

Going to ridiculous lengths—European coexistence regulations for GM crops Koreen Ramessar, Teresa Capell, Richard M Twyman & Paul Christou

feat u re pate n ts Regulations to prevent outcrossing of GM pollen in the EU, p 133

137

141

Changing the rules of the game: addressing the conflict between free access to scientific discovery and intellectual property rights Miriam Bentwich Recent patent applications in antibody fragments

© 2010 Nature America, Inc. All rights reserved.

N E W S A ND V I E W S 142

ChIPs and regulatory bits Xin He & Saurabh Sinha

144

From genomics to crop breeding Richard Flavell

145

Spilling the beans on legume biology Peter Hare

146

Systematic tracking of cell fate changes Jonghwan Kim & Stuart H Orkin

148

Research highlights

comp u tat i o n a l b i o l ogy Mining the maize genome, p 144

R eso u rce 149

Rational association of genes with traits using a genome-scale gene network for Arabidopsis thaliana Insuk Lee, Bindu Ambaru, Pranjali Thakkar, Edward M Marcotte & Seung Y Rhee

research B R I E F C O M M UNI C AT I O N S 157

Enhanced antibody half-life improves in vivo activity J Zalevsky, A K Chamberlain, H M Horton, S Karki, I W L Leung, T J Sproule, G A Lazar, D C Roopenian & J R Desjarlais

Computational discovery of plant gene function, p 149

nature biotechnology

iii

volume 28 number 2 february 2010 l etters 161

Expansion and maintenance of human embryonic stem cell–derived endothelial cells by TGFb inhibition is Id1 dependent D James, H-s Nam, M Seandel, D Nolan, T Janovitz, M Tomishima, L Studer, G Lee, D Lyden, R Benezra, N Zaninovic, Z Rosenwaks, S Y Rabbany & S Rafii

167

Real-time imaging of hepatitis C virus infection using a fluorescent cell-based reporter system C T Jones, M T Catanese, L M J Law, S R Khetani, A J Syder, A Ploss, T S Oh, J W Schoggins, M R MacDonald, S N Bhatia & C M Rice

172

Rational design of cationic lipids for siRNA delivery S C Semple, A Akinc, J Chen, A P Sandhu, B L Mui, C K Cho, D W Y Sah, D Stebbing, E J Crosley, E Yaworski, I M Hafez, J R Dorkin, J Qin, K Lam, K G Rajeev, K F Wong, L B Jeffs, L Nechev, M L Eisenhardt, M Jayaraman, M Kazem, M A Maier, M Srinivasulu, M J Weinstein, Q Chen, R Alvarez, S A Barros, S De, S K Klimuk, T Borland, V Kosovrasti, W L Cantley, Y K Tam, M Manoharan, M A Ciufolini, M A Tracy, A de Fougerolles, I MacLachlan, P R Cullis, T D Madden & M J Hope

178

CORRIGENDA AND ERRATA

© 2010 Nature America, Inc. All rights reserved.

Endothelial cells from hESCs, p 161

careers a n d recr u i tme n t 179

Fourth quarter lag in biotech hiring Michael Francisco

180

people

Systematicly improving siRNA delivery, p 172

ADVERTISEMENT Biotech in China A special report analyzing China’s emerging biotech and pharmaceutical industries. The main feature investigates the local strategies of western and Chinese organisations, assessing massive investment programs and a returning western trained skilled work force. Will China achieve its ambition to develop and innovative drug discovery industry to compliment its generics industry? How long do the experts believe it will be before China has its own home grown innovative new drug approved for the global markets? What are activities of western drug companies in China? The “Biotech in China” special report follows Letters on page 176 and is produced with the commercial support from the organizations featured in the Advertorial Partnering Profiles.

nature biotechnology

v

in this issue

© 2010 Nature America, Inc. All rights reserved.

Endothelial cell recipe Endothelial cells derived from pluripotent stem cells might one day provide the raw material for engineering or repairing blood vessels. To develop an improved method for vascular differentiation, Rafii and colleagues generate a human embryonic stem cell line that expresses green fluorescent protein under the control of the endothelial cell–specific VE-cadherin promoter. Using this reporter cell line to screen molecules involved in early developmental signaling, they find that inhibition of transforming growth factor (TGF)β beginning at day 7 of differentiation increases the yield of endothelial cells and maintains the cells’ vascular phenotype for up to ten cell divisions. Mechanistic investigation identifies the transcription factor Id1 as a key mediator of the effects of TGFβ inhibition. [Letters, p. 161] KA

Predicting plant gene function Despite extensive mutant screening, the functions of many plant genes are still unknown. Lee et al. predict gene function in the model plant Arabidopsis thaliana by gauging the likelihood that pairs of genes are involved in the same biological processes. Each pair of genes is assigned a score that combines many types of experimental and computational data gathered in Arabidopsis. The score also incorporates data from other organisms, such as yeast, worm, fly and humans, on genes that show substantial sequence similarity to Arabidopsis genes. Then, the function of an Arabidopsis gene is predicted based on the scores linking it to other genes with known function. To demonstrate the utility of the approach, Lee et al. predicted and validated the roles of genes in seed pigmentation, lateral root development and drought sensitivity. By integrating multiple sources of data using methods customized for plants, Lee et al. predict gene function with greater confidence than by using only a single source of data. This study provides a resource for identifying genes that influence agriculturally and economically important plant traits. [Resource, p. 149] CM Written by Kathy Aschheim, Markus Elsner, Michael Francisco, Peter Hare, Craig Mak & Lisa Melton

nature biotechnology volume 28 number 2 february 2010

Potent siRNA delivery Empirical screening has revealed novel lipid nanoparticle formulations that have substantially enhanced in vivo delivery of therapeutic small interfering (si)RNAs. Now Semple et al. have set a new potency standard for siRNA delivery to the liver by adopting a more rational approach to the design of cationic lipids. They refined an empirically identified cationic lipid (1,2-dilinoleyloxy-3-dimethylaminopropane), widely regarded as the benchmark for use in lipid nanoparticles, by dividing the structure into three functional elements and then systematically testing modifications of each element in isolation. This strategy to reveal structure-activity relationships was guided by the putative role of cone-shaped lipids to induce nonbilayer phases, such as the hexagonal HII phase illustrated here. When formulated to silence hepatic gene expression, the best-performing lipid variant conferred in vivo activity at siRNA doses as low as 0.01 mg/kg in rodents and 0.1 mg/kg in nonhuman primates. [Letters, p. 172] PH

Lighting up HCV infection Protocols for detecting hepatitis C virus infection without the need for complex manipulation of clinical samples or the use of genetically engineered viruses is of prime importance for many applications in basic research and drug development. Rice and colleagues have now developed a fluorescent reporter system that allows the detection of individual cells infected by wild-type viruses. The reporter molecule is based on a known target of a viral protease, interferon-β promoter stimulator protein 1 (IPS-1). The C-terminal part of IPS-1, including the mitochondrial targeting sequence and the protease cleavage site, is fused to a fluorescent protein and a nuclear localization sequence. Upon expression of the viral protease the construct is cleaved and the fluorescent protein relocalizes from the mitochondria to the nucleus. The authors use their reporter molecule to study viral propagation in living cells and the development of stress responses in cells after infection. They also demonstrate that primary hepatocytes can be infected with hepatitis C viruses in vitro. [Letters, p. 167] ME

v

i n t h is issu e

Linking antibody half-life to efficacy Mutations that improve the affinity of antibodies for the neonatal Fc receptor (FcRn) are known to enhance antibody longevity in vivo. Nonetheless, it has never been demonstrated that prolonged

Patent roundup Bringing a five-year patent dispute to an end, a federal court in Boston has ruled that Roche’s Mircera (methoxy polyethylene glycolepoetin beta) does, indeed, infringe on Amgen’s patents and slapped the Swiss drugmaker with a permanent injunction banning sales in the US market. Roche has agreed to a limited license agreement with Amgen that will allow it to sell Mircera in the US until 2014. [News, p. 112] LM

© 2010 Nature America, Inc. All rights reserved.

To solve the conflict between free access to scientific discovery and intellectual property rights, Bentwich proposes a provisional patented paper application procedure that could promote earlier disclosure of novel scientific knowledge and justify the requirement to grant inexpensive licenses for using inventions for the advancement of other research. [Patent Article, p. 137] MF Recent patent applications in antibody fragments. [New patents, p. 141]

vi

MF

exposure to Fc-engineered therapeutic antibodies necessarily enhances in vivo activity. Desjarlais and colleagues use cynomolgus monkeys and a humanized transgenic mouse model to show that the reduction in antibody clearance caused by increased antibody affinity for FcRn enhances the antitumor activities of antibodies targeted against either an internalizing surface receptor (epidermal growth factor receptor) or a cytokine (vascular endothelial growth factor). The enhanced pharmacokinetics associated with Fc-engineered variants may translate to greater convenience for patients, reduced costs and higher efficacy. [Brief Communications, p. 157] PH

Next month in • Screening for drugs that reconfigure metabolism •L  eukemia stem cell quiescence and resistance to chemotherapy •D  irected evolution of an MRI contrast agent for dopamine •A  utophagy harnessed to clear mutant huntingtin protein

volume 28 number 2 february 2010 nature biotechnology



www.nature.com/naturebiotechnology

EDITORIAL OFFICE [email protected] 75 Varick Street, Fl 9, New York, NY 10013-1917 Tel: (212) 726 9200, Fax: (212) 696 9635 Chief Editor: Andrew Marshall Senior Editors: Laura DeFrancesco (News & Features), Kathy Aschheim (Research), Peter Hare (Research), Michael Francisco (Resources and Special Projects) Business Editor: Brady Huggett Associate Business Editor: Victor Bethencourt News Editor: Lisa Melton Associate Editors: Markus Elsner (Research), Craig Mak (Research) Editor-at-Large: John Hodgson Contributing Editors: Mark Ratner, Chris Scott Contributing Writer: Jeffrey L. Fox Senior Copy Editor: Teresa Moogan Managing Production Editor: Ingrid McNamara Senior Production Editor: Brandy Cafarella Production Editor: Amanda Crawford Senior Illustrator: Katie Vicari Illustrator/Cover Design: Kimberly Caesar Senior Editorial Assistant: Ania Levinson

© 2010 Nature America, Inc. All rights reserved.

MANAGEMENT OFFICES NPG New York 75 Varick Street, Fl 9, New York, NY 10013-1917 Tel: (212) 726 9200, Fax: (212) 696 9006 Publisher: Melanie Brazil Executive Editor: Linda Miller Chief Technology Officer: Howard Ratner Head of Nature Research & Reviews Marketing: Sara Girard Circulation Manager: Stacey Nelson Production Coordinator: Diane Temprano Head of Web Services: Anthony Barrera Senior Web Production Editor: Laura Goggin NPG London The Macmillan Building, 4 Crinan Street, London N1 9XW Tel: 44 207 833 4000, Fax: 44 207 843 4996 Managing Director: Steven Inchcoombe Publishing Director: Peter Collins Editor-in-Chief, Nature Publications: Philip Campbell Marketing Director: Della Sar Director of Web Publishing: Timo Hannay NPG Nature Asia-Pacific Chiyoda Building, 2-37 Ichigayatamachi, Shinjuku-ku, Tokyo 162-0843 Tel: 81 3 3267 8751, Fax: 81 3 3267 8746 Publishing Director — Asia-Pacific: David Swinbanks Associate Director: Antoine E. Bocquet Manager: Koichi Nakamura Operations Director: Hiroshi Minemura Marketing Manager: Masahiro Yamashita Asia-Pacific Sales Director: Kate Yoneyama Asia-Pacific Sales Manager: Ken Mikami DISPLAY ADVERTISING [email protected] (US/Canada) [email protected] (Europe) [email protected] (Asia) Global Head of Advertising and Sponsorship: Dean Sanderson, Tel: (212) 726 9350, Fax: (212) 696 9482 Global Head of Display Advertising: Andrew Douglas, Tel: 44 207 843 4975, Fax: 44 207 843 4996 Asia-Pacific Sales Director: Kate Yoneyama, Tel: 81 3 3267 8765, Fax: 81 3 3267 8746 Display Account Managers: Global Account Development Manager: Graham Combe, Tel: 44 207 843 4914, Fax: 44 207 843 4749 New England: Sheila Reardon, Tel: (617) 399 4098, Fax: (617) 426 3717 New York/Mid-Atlantic/Southeast: Jim Breault, Tel: (212) 726 9334, Fax: (212) 696 9481 Midwest: Mike Rossi, Tel: (212) 726 9255, Fax: (212) 696 9481 West Coast South: George Lui, Tel: (415) 781 3804, Fax: (415) 781 3805 West Coast North: Bruce Shaver, Tel: (415) 781 6422, Fax: (415) 781 3805 Germany/Switzerland/Austria: Sabine Hugi-Fürst, Tel: 41 52761 3386, Fax: 41 52761 3419 United Kingdom/Ireland: Jeremy Betts, Tel: 44 207 843 4968, Fax: 44 207 843 4749 Scandinavia/Iceland/Spain/Portugal: Evelina Rubio-Hakansson, Tel: 44 207 843 4079, Fax: 44 207 843 4749 France/Belgium/The Netherlands/Italy/Israel/Eastern Europe: Nicola Wright, Tel: 44 207 843 4959, Fax: 44 207 843 4749 Asia-Pacific Sales Manager: Ken Mikami, Tel: 81 3 3267 8765, Fax: 81 3 3267 8746 Greater China/Singapore: Gloria To, Tel: 852 2811 7191, Fax: 852 2811 0743 NATUREJOBS [email protected] (US/Canada) [email protected] (Europe) [email protected] (Asia) US Sales Manager: Ken Finnegan, Tel: (212) 726 9248, Fax: (212) 696 9482 European Sales Manager: Dan Churchward, Tel: 44 207 843 4966, Fax: 44 207 843 4596 Asia-Pacific Sales Manager: Ayako Watanabe, Tel: 81 3 3267 8765, Fax: 81 3 3267 8746 SITE LICENSE BUSINESS UNIT Americas: Tel: (888) 331 6288 Asia/Pacific: Tel: 81 3 3267 8751 Australia/New Zealand: Tel: 61 3 9825 1160 India: Tel: 91 124 2881054/55 ROW: Tel: 44 207 843 4759

[email protected] [email protected] [email protected] [email protected] [email protected]

CUSTOMER SERVICE www.nature.com/help Senior Global Customer Service Manager: Gerald Coppin For all print and online assistance, please visit www.nature.com/help Purchase subscriptions: Americas: Nature Biotechnology, Subscription Dept., 342 Broadway, PMB 301, New York, NY 10013-3910, USA. Tel: (866) 363 7860, Fax: (212) 334 0879 Europe/ROW: Nature Biotechnology, Subscription Dept., Macmillan Magazines Ltd., Brunel Road, Houndmills, Basingstoke RG21 6XS, United Kingdom. Tel: 44 1256 329 242, Fax: 44 1256 812 358 Asia-Pacific: Nature Biotechnology, NPG Nature Asia-Pacific, Chiyoda Building, 2-37 Ichigayatamachi, Shinjuku-ku, Tokyo 162-0843. Tel: 81 3 3267 8751, Fax: 81 3 3267 8746 India: Nature Biotechnology, NPG India, 3A, 4th Floor, DLF Corporate Park, Gurgaon 122002, India. Tel: 91 124 2881054/55, Tel/Fax: 91 124 2881052 REPRINTS [email protected] Nature Biotechnology, Reprint Department, Nature Publishing Group, 75 Varick Street, Fl 9, New York, NY 10013-1917, USA. For commercial reprint orders of 600 or more, please contact: UK Reprints: Tel: 44 1256 302 923, Fax: 44 1256 321 531 US Reprints: Tel: (617) 494 4900, Fax: (617) 494 4960

E d i t o ria l

PhRMA wants you! The biotech brand is in danger of being sullied by the blurring of pharma and biotech boundaries.

© 2010 Nature America, Inc. All rights reserved.

E

arly last month, the Pharmaceutical Research and Manufacturers of America (PhRMA) trumpeted that its drive to diversify membership had led to the recruitment of seven biotech firms. The drive began last July, shortly after Roche announced that it was leaving PhRMA and joining the Biotechnology Industry Organization (BIO), characterizing itself as “the world’s largest biotechnology company.” Tit-for-tat membership battles of healthcare lobby groups are usually of little importance. However, the status that Roche attaches to its biotech identity and the pursuit of biotech firms by PhRMA suggest that the ‘biotech’ brand has a distinct and desirable cachet. The pharmaceutical industry does have a public relations problem. A Harris poll from August 2009 found that 84% of US adults blame the pharma industry for the problems with the US healthcare system. Trust ratings for pharma are also low, with only 7–14% of US adults willing to believe company statements as true (Harris Polls, 2003–2007). In contrast, attitudes toward the biopharmaceutical sector are generally positive. An April 2009 survey from the state industry association, MassBio, for instance, found 51% of voters had positive feelings towards biotech. So ‘biotech’ is young, thrusting, optimistic and tackling healthcare through research while ‘pharma’ is gigantic, profiteering, world-worn and tackling healthcare by increasing market share. In reality, of course, no such clear-cut divisions exist. The biotech and pharma sectors form parts of a continuum, indivisible by the presence or absence of biologics, size, research concentration, clinical competence, greed or vulnerability to acquisition. Readers may be surprised to learn that Roche is more of a biologics company than a drug company: in 2009, 54% of its revenues came from biologics and only 24% from small molecules. Other pharmaceutical companies have been developing and acquiring biologics competence at an increasing rate over the past decade or so. Pfizer’s acquisition of Wyeth last October was motivated partly by the need to have more biologics. Conversely, a significant portion of the pipeline of Amgen and Genzyme is small molecules. The blurring of the biotech-pharma boundary has led to convergent evolution of lobby groups. PhRMA’s current mission is advocacy on behalf of “pharmaceutical/biotechnology research companies” while BIO’s is to be “the advocate for its member organizations—both large and small.” Little to choose between the two. There is, however, one very important difference. Whereas PhRMA has 34 full members, BIO has 1,200 members, most of which pay only nominal fees. BIO is thus the only advocate for the smaller, younger, nonrevenuedriven companies. Its voice on behalf of smaller firms may not always be loud and clear, but it is surely a voice. By donning the ‘biotech mantle’, PhRMA may be hoping to capture for its members some of the public and political sympathies enjoyed by BIO. In the face of this direct competition, BIO needs to make a choice. It can out-PhRMA PhRMA, becoming a better advocate for large pharma, trying

nature biotechnology volume 28 number 2 february 2010

to be all things to all companies. Or it could differentiate itself by becoming an out-and-out advocate for small to medium-sized biotech, deemphasizing the large company agenda and rededicating itself to the innovative edge of the industry. We hope they choose the second path.

Impossible fixes Impractical solutions to European biotech financing don’t help anyone.

A

couple of months ago, the European Commission and the industry trade association European Biopharmaceutical Enterprises held a closed meeting in Brussels to discuss a survey from ECORYS (Rotterdam), the Danish Technological Institute (DTI: Taastrup) and three other consultants undertaken as part of a €2.99 ($4.22) million contract from the Commission (see p. 110). The survey pointed to three (obvious) funding gaps during biotech development. It also estimated that 40% of European biotechs needed more cash before the end of 2010. Urgent problems, indeed, requiring urgent solutions. So what did the report recommend? First, it proposed that the European Commission should consider sector-specific policy measures targeting the biopharmaceutical sector. In a rational world, this is plausible and sensible. But the European Union is not rational. It is an aggregation of formerly warring nation states each seeking economic advantages while acting out historical resentments through petty and circular administrative fine-tuning. To counter the nationalistic tendencies, the European Commission’s competitiveness legislation forbids governments from instituting sector-specific incentives for industry. Nations can support companies that are small or young or research intensive, but they cannot single out specific sectors, like IT or biotech, for specific help. Short of overhauling the entire philosophy of the European endeavor, this recommendation, therefore, is a non-starter. A second proposal in the report suggests that national and European policy makers should reform the financial markets across Europe to allow venture capitalists to operate across national borders. Again, nice idea, but a pipe dream. In the main, the European Union has 27 distinct financial systems with 27 national regulatory systems and asynchronicity throughout its multiple economic cycles. This, too, is a solution for the next decade or beyond. Thankfully, by the time it went to press, the report had been stripped of suggestions that European biotech should be financed using gold bars from the croc at the end of the rainbow or getting Sir Bob Geldof and Bono to organize a Bio-Aid telethon. European biotech financing is a tough problem in need of imaginative solutions. But solutions must be practical—not outside the realms of reality.

103

news in this section Pfizer moves on plant cell manufacturing

EC surveys European industry funding gap p110

GM crops and rising herbicide use p112

p107

Last October, Roche-Genentech and the US Food and Drug Administration (FDA) issued a “Dear Doctor” letter notifying rheumatologists about a case of progressive multifocal leukoencephalopathy (PML) in an arthritis patient treated with Rituxan (rituximab). This, the third PML case to emerge, is prompting concerns because the affected patient is the first to contract the condition on Rituxan without any prior treatment with another immune suppressant. Elsewhere, news that once again, cases of PML continue to be a concern for multiple sclerosis (MS) patients receiving Biogen Idec’s Tysabri (natalizumab) is again prompting regulators to scrutinize current monitoring and control of this brain malady. As reports of PML cases mount, drug developers are understandably anxious that the potentially fatal brain infection could seriously compromise their most valuable antibody franchises. When Cambridge, Massachusetts–based Biogen Idec launched Tysabri in late 2004, the monoclonal antibody (mAb) was one of biotech’s brightest lights. It was the first IgG4 mAb directed against the α4 subunit of the integrin VLA-4 (very late antigen 4) adhesion complex expressed on activated lymphocytes, monocytes and other leukocytes. In addition, its safety profile appeared favorable and it was about twice as efficacious as existing treatments, such as interferon β and Copaxone (glatiramer acetate). In 2005, Biogen voluntarily withdrew the drug after PML, a rare, life-threatening brain disease, was diagnosed in two patients on Tysabri (Nat. Biotechnol. 23, 397–398, 2005). But the company was able to get the drug back on the market in 2006 after implementing a stringent PML patient monitoring program (Fig. 1; Nat. Biotechnol. 27, 986, 2009). The latest 23 additional documented cases of PML in Tysabri patients prove these safety problems have not gone away. What’s more, cases of the disease have also turned up in patients on two other immune modulating products, both from S. San Francisco, California–based Genentech. Last June, PML cases prompted the company to withdraw its anti-CD11a mAb Raptiva (efalizumab)

Developing B cell

Bone marrow bri

a Tys

Stroma cells

VCAM-1

Tysabri

Blood vessel

α4 β1

Polyoma JC virus

Hematopoietic progenitor cell

Marina Corral

© 2010 Nature America, Inc. All rights reserved.

PML problems loom for Rituxan

Binding of Tysabri to α4 integrins is thought to prevent hematopoietic progenitor and developing B cells from attaching to vascular cell adhesion molecule 1(VCAM-1) promoting these precursor cells to migrate into the circulation. If JC virus is residing in the bone marrow in a latent state, it would also migrate out into circulation. JC virus uses B cells and their DNA-binding proteins to initiate viral replication. (Adapted from N. Engl. J. Med. 348, 68–71, 2003).

from the psoriasis market; and at the end of last year, more cases were associated with its anti-CD20 blockbuster Rituxan, which is marketed for rheumatoid arthritis and nonHodgkin’s lymphoma. Initially, the association of PML with Tysabri had suggested only drugs with mechanisms of action related to VLA-4 would be affected. But the finding of PML in patients on Raptiva and Rituxan, which bind two unrelated targets, has added urgency to the quest to better understand the reactivation process of John Cunningham (JC) virus, the human polyomavirus that causes the disease. Now, researchers are trying to determine exactly why these particular drugs make people more susceptible to this otherwise extremely rare condition, and how this effect can be avoided. Biogen Idec has been collaborating on the PML problem with Elan of Dublin, the com-

nature biotechnology volume 28 number 2 february 2010

pany with which it codeveloped Tysabri. Al Sandrock, senior vice president of neurology R&D at Biogen Idec says they will also be working with Roche-Genentech, which copromotes Rituxan in the US market and in Europe. “All three companies have signed a letter of intent. We have ideas on what to work on but we will nail that down early in the new year,” Sandrock says. A Genentech spokesperson confirmed by e-mail that “there have been discussions between several companies, including Biogen, Genentech and Roche, in forming a consortium dedicated to PML” but added that “to date no contracts have been finalized.” Biogen and Elan have the advantage of access to thousands of banked blood and urine samples from patients who have taken Tysabri, according to Ted Yednock, the Elan researcher who invented the antibody and

105

NEWS

Raptiva is voluntarily withdrawn by Genentech in June, after four confirmed PML cases and three deaths

2004

June

Tysabri approved 2005

2006

Tysabri pulled from the market after two PML cases identified

2007

2008

October

Tysabri goes back on the market mid-year, with new safety and reporting measures in place

“Dear Doctor” letter warns that patients on Rituxan are at an increased risk of developing PML. EMA begins a new review of Tysabri

2009

PML cases are confirmed in patients taking Rituxan and Raptiva

© 2010 Nature America, Inc. All rights reserved.

Figure 1 Timeline for the reintroduction of Tysabri and PML cases.

is now working on PML. Biogen’s scientists have already developed an assay that identifies carriers of JC virus. It is estimated that upwards of 50% of people are infected with this opportunistic virus, which mostly lies latent indefinitely. PML is extremely rare in anyone but those with impaired immune systems, such as people with AIDS, B-cell malignancies or systemic lupus erythematosus (SLE). Once JC virus is activated it infiltrates the central nervous system (CNS), causing neurobehavioral symptoms, which can become life threatening. Fortunately, Tysabri’s effects are rapidly reversible, so if patients at risk of PML can be identified, their normal immune function can quickly be restored and most patients recover. The assay Biogen has developed “only tells you whether the person has been infected or not,” says Yednock. “We also need to be able to tell the level of viral replication or whether it has transitioned from latent to activated virus [capable of infecting brain cells].” “This is such a ubiquitous virus, something must happen to enable it to grow in glial cells,” says Joseph Berger of the Department of Neurology at the University of Kentucky College of Medicine. One line of research has suggested that the structure of Tysabri itself might be to blame. Researchers at Genmab in Copenhagen have reported that IgG4 antibodies like Tysabri, which are bi specific and contain a flimsy backbone, can exchange the Fab-arm portion of the antibody with endogenous IgG4s with the wrong specificity (Nat. Biotechnol. 27, 767–771, 2009). If Tysabri molecules recombined with endogenously produced anti–JC virus antibodies, the researchers postulated, the resultant bispecific mAb might capture the virus and carry it into the CNS. Scientists at PanGenetics in Utrecht, The Netherlands, who have an IgG4 mAb in

106

development, have challenged this theory, pointing out that clinical data suggest there is very little circulating JC virus to be captured and no mechanism by which Tysabri could transport virus into the CNS (see p. 123; SciBX 2(32); doi:10.1038/scibx.2009.1231). Also, mAbs are not the only drugs that can cause PML. Basel-based Roche’s CellCept (mycophenolate mofetil, which targets inosine monophosphate dehydrogenase 1), for example, has also been linked to the disease. Most experts appear to be leaning toward other theories. Recent studies suggest the antibodies can specifically affect the movement of mature and immature lymphocytes in the body. By binding VLA-4, Tysabri prevents lymphocytes from binding to the vascular cell adhesion molecule-1 (VCAM-1), and if hematopoietic stem cells and pre-B cells are unable to bind VCAM1, they migrate out from the bone marrow. “We have plenty of data now that show these progenitor cells are found at higher than normal rates in the peripheral system in patients on this drug,” says Gene Major of the Laboratory of Molecular Medicine and Neuroscience at the National Institute of Neurological Discoveries and Stroke in Bethesda, Maryland. Some of these B cells contain JC virus that may replicate prompted by DNA-binding proteins (New Engl. J. Med. 361, 1041–1043, 2009). This mechanism could also explain how Rituxan can cause PML because this mAb reduces the number of CD20+ B cells in peripheral blood, which also stimulates the release of progenitor cells from the marrow. It is also thought that Tysabri limits the passage of cytotoxic T cells into the brain because they need to bind VCAM-1 to migrate out of the circulation. Because these are precisely the cells that would defend

against the JC virus, the drug could be handicapping patients’ native immune defense. In the meantime, regulators are keeping a close eye on Tysabri and Rituxan. The European Medicines Agency (EMA) was reviewing the data about Tysabri and PML as Nature Biotechnology went to press. Crystal Rice, an FDA spokesperson, wrote in an e-mail, “We have been receiving and continue to receive reports of PML cases in real time, are monitoring the incidence of PML both in the United States and worldwide on an ongoing basis, and are continuing to assess the issue to determine the need for further regulatory action.” With Tysabri, risk of PML has currently been calculated at one in a thousand, but that could change as more people take the drug for extended periods. For Rituxan, the picture is less clear. PML occurs in rheumatoid arthritis and SLE patients sometimes even when they are not taking Rituxan. “It’s not clear how much additional risk Rituxan imparts,” says Leonard H. Calabrese of the Cleveland Clinic in Cleveland. On the basis of current data, he believes the risk is so low the number is not even calculable. Nonetheless, rheumatologists are very concerned about PML. Calabrese says he and colleagues recently conducted a survey and found that “the level of concern doesn’t coincide with the available risk data.” Many arthritis drugs, he points out, have serious potential side effects. But with Rituxan and PML, “there is a disconnect between knowledge and fear.” It’s also unclear whether other immunemodulating drugs that mobilize hematopoietic stem cells and pre-B cells may encounter problems similar to those Tysabri and Rituxan have. Copenhagen-based Genmab’s Arzerra (ofatumumab), approved in October for chronic lymphocytic leukemia, Genentech’s ocrelizumab (in late-stage trials for rheumatoid arthritis and SLE) and Biolex’s BIX-301 non-Hodgkin’s B cell lymphoma, all target CD20. As yet, there are no reports of PML cases associated with these drugs. “We are perturbing very specific arms of a highly complex system,” Berger says. That power can clearly backfire if you don’t know all the drug’s effects. As a result, Biogen’s Sandrock expects risk management plans are “here to stay.” Berger says he’s been trying to convince drug companies to establish registries for drugs with novel mechanisms. “As a community, we need to be prepared to identify these problems as they arise,” he says. Malorye Allison Acton, Massachusetts

volume 28 number 2 february 2010 nature biotechnology

news

On December 1, Pfizer became the first big pharma to commit to take to market a latestage biologic drug produced in plant cells. It acquired rights to taligurase alfa, a form of the enzyme glucocerebrosidase in development for the treatment of Gaucher’s disease, from Protalix Biotherapeutics in Carmiel, Israel. Protalix has completed phase 3 studies and has submitted a new drug application for the drug, also known as prGCD, eyeing US marketing approval in 2010. At the request of the US Food and Drug Administration (FDA) last year, the company has already begun supplying prGCD to patients in the US under an expanded access program and similarly to patients in the EU under a compassionateuse protocol. This apparent comfort level of regulators, along with the interest of a major drug company, signals a new level of recognition of plant cell–based manufacturing as a viable and potentially less expensive alternative to mammalian and bacterial production of biopharmaceuticals, including biosimilar versions of existing drugs. Protalix has already collected $65 million from the deal, which gives New York–based Pfizer worldwide rights to prGCD, excluding Israel, and could earn another $50 million in milestones. Protalix will continue to manufacture the drug, which it produces in carrot

cells, pay 40% of all expenses going forward and receive the same percentage of revenues in return. The company’s prGCD will compete with Genzyme’s Ceredase (alglucerase), a form of the enzyme beta-glucocerebrosidase purified from human placental tissue that is modified to be terminated with mannose, and Cerezyme (imiglucerase), a recombinant human beta-glucocerebrosidase with a His495→Arg substitution and the same sugar modification. Both of Genzyme’s products are indicated for the treatment of Gaucher’s disease, a rare lysosomal storage disorder resulting from a hereditary deficiency in the glucocerebrosidase enzyme. Gaucher’s disease is the most prevalent among the group of lysosomal storage disorders, which have been a historic focus for Genzyme in Cambridge, Massachusetts. This is Pfizer’s first move into the area of rare and neglected diseases, the result of a process the company began a year ago to identify such opportunities. “Protalix’s name and technology platform and their work in Gaucher’s disease came to the top of that list. We approached them in the middle of last year and things moved fairly quickly,” says Andrew Curtis, biosimilar and orphan drugs director for Pfizer’s established products business.

Protalix

© 2010 Nature America, Inc. All rights reserved.

Pfizer stakes a claim in plant cell–made biopharmaceuticals

Protalix’s bioreactor plant cell system. The GMP-approved system is set up to manufacture a range of proteins, including antibodies, complex enzymes and plant-derived pharmaceuticals.

nature biotechnology

volume 28 number 2 february 2010

107

© 2010 Nature America, Inc. All rights reserved.

NEWS “This was a unique opportunity to take advantage of their plant cell technology platform with this lead drug to perhaps provide a more cost-effective therapy,” he says, adding that treatments for lysosomal storage diseases have typically been among the most expensive, costing several hundreds of thousands of dollars annually. Protalix’s phase 2 data and interim reads on its phase 3 study, including available reports from data safety and monitoring boards, were reassuring, as was the fact that regulators had asked Protalix to initiate early access programs. “Typically, companies are not asked to do that unless FDA or EMEA [the European Medicines Agency] or both have a good level of comfort around the safety and efficacy of the drug,” says Curtis. “All those things came to bear on the comfort level we had with Protalix, even though it would be the first plant cell–based therapy to be approved.” Curtis is also quick to point out that although there may be future opportunities to leverage the Protalix platform, this was a product-oriented transaction. “This deal was born from the commercial arm of Pfizer. We see this obviously as an opportunity to work more closely with them going forward, to maximize their technology platform, but the technology platform trailed behind. We got the drug and along the way discovered how wonderful the platform is.” Beyond prGCD, the Protalix technology platform could potentially provide opportunities for Pfizer to work in other rare genetic diseases and even beyond that, in the production of biosimilars, to lower the cost of goods, says Curtis. Protalix also has programs in preclinical development for a biosimilar version of the anti-tumor necrosis factor alpha fusion receptor protein Enbrel (etanercept) and for enzyme replacement to treat Fabry’s disease, another lysosomal storage disorder. In plant biotech, as is true for other biotech deals, the trend “seems to have gone, over the last ten years, away from companies that have platform technology much more towards product-driven investments,” says Charles Arntzen of the Biodesign Institute at Arizona State University in Tempe. “Protalix never goes around talking that they have a cell-culture system, except at plant meetings. They focus on the enzyme they are producing and the product niche it will fit into.” Pittsboro, North Carolina–based Biolex Therapeutics is the only other company with a plant-derived compound in late-stage trials. Their controlled-release form of interferon α-2b made in aquaculture, currently

108

Box 1 Green antibody farming Viral vectors have enough going for them that the technique could become a staple for protein production in whole plants; they can achieve high levels of expression in a very short time. “You can use agrobacterium, for example, to essentially bathe the interior surface of the leaf,” says Arntzen. “Once you get one copy of the merged virus [it is usually introduced in two parts along with a recombinase delivered as a separate gene] in the cell, boom, off it goes. You adjust conditions so you have a coordinated assembly so every cell in the leaf is simultaneously getting the effect of having a virus starting its replication in that cell.” Icon was among the first to successfully use these viral vectors, and there are at least a dozen other companies that have gone that way. “It looks to me like the wave of the future,” Arntzen says. The biggest advantage may come in producing monoclonal antibodies, where the best work is being done using two separate viruses: one produces the light chain, the other the heavy chain. Essentially, two RNA species are introduced into a single cell at the same time. “They assemble beautifully and it’s routine now to get about half a gram of antibody out of a kilogram of plant leaves,” says Arntzen. “There are enough data, not necessarily all published, especially for immunoglobulins of IgG type made in plants, and enough information and knowledge in several large companies to say that all these antibodies are fully active and pharmacologically as good as those produced in other systems,” says Gleba, adding that Genentech of S. San Francisco, California, and others have shown that their existing molecules are less effective because of fucose added to the glycans. “If you remove it, the effector function improves dozens of times and that should directly translate into therapeutic effect, so you need less of the molecule,” he says. MR

in phase 2b for the treatment of hepatitis C, received $60 million in venture funding from a group led by Clarus Ventures and Orbimed Advisors, both known for late-stage, product-oriented investments. It’s long been believed that plant cell– based manufacturing has the potential to be less expensive than mammalian or Chinese hamster ovary cell–based methods, in part because plants produce protein with a glycosylation pattern closer to human, says Arntzen (Box 1). That’s apparently the case with prGCD. To produce Ceredase and Cerezyme, Genzyme has to clip off additional sugars to expose terminal mannose residues, says David Aviezer, Protalix’s president and CEO. “With plant technology, by performing ER [endoplasmic reticulum] retention of the protein during the glycosylation process, we can obtain the correct mannose glycosylation pattern directly made by the cell.” That reduces processing costs—prGCD could be an order of magnitude less expensive to produce than either Cerezyme or another potential competitor such as Shire’s velaglucerase. Chineham, UK–based Shire is expecting an FDA approval decision for their product by the end of February 2010. Yet some in the field see Protalix’s platform as a somewhat crude first-generation technology that only touches on the potential for plant-based pharmaceutical produc-

tion. “You increase the cost of production by going back to fermenters,” says Yuri Gleba of Icon Genetics, now part of the Bayer Innovation Group in Halle, Germany. What’s more, expression is not exceptionally high in the Protalix cell line, and the company is using expression cassettes that have been in circulation for a generation, he says. “Protalix has shown these fermenters are cheaper, so you can have 20–30% cheaper product as a result, which could justify the new business case. But with other technologies you can get even higher expression levels and lower cost of goods,” he adds. On the other hand, cell culture brings the advantages of a plant organism in “the same regulatory environment FDA has learned and set for the last 20 years” for biopharmaceuticals, says Aviezer—in clean rooms and under the same standard operating procedures for growing mammalian cells and purifying proteins. And Icon’s Gleba acknowledges that Protalix’s success in product development with prGCD shows they have a keen business sense. “If you are not strong on one side you have to compensate by being excellent on another, and by all accounts, they are,” he says. The deal with Pfizer and the approval of prGCD “should open the floodgates, in my opinion,” he says. “It is by far the most significant development in the plant-made pharmaceuticals arena right now.” Mark Ratner Cambridge, Massachusetts

volume 28 number 2 february 2010 nature biotechnology

news Amylin’s $1 billion heavyweight deal

© 2010 Nature America, Inc. All rights reserved.

Amylin

In November, Amylin announced a $1 billion partnership with Osaka, Japan–based Takeda to codevelop and commercialize obesity treatments. Takeda paid the San Diego–based biotech $75 million upfront for Symlin/metreleptin combination as part of an agreement that could exceed $1 billion if certain development and sales-dependent milestones are hit. The deal also includes Amylin’s amylinomimetic compound davalintide, which is currently only in phase 2 studies. But, as Stephen O’Rahilly, director of the Metabolic Research Laboratories at the University of Cambridge in the UK, points out, Symlin (pramlintide) is already approved and is “used by a lot of type I diabetes patients to smooth out control and prevent the weight gain that happens when on insulin.” The agreement comes amid a surge of deals in metabolic disease, particularly for diabetes treatments that have potential weight loss benefits for the obese. For example, on December 23, Paris-based Sanofi-Aventis paid €100 ($143) million for a 19.9% stake in Zealand Pharma. This Amylin has gained a strong position in the already crowded metabolic disease marketplace. Copenhagen-based biotech is developing a peptide analog of Amylin’s Byetta (glucagon-like peptide 1 (GLP-1)/exendin 4) for type 2 diabetes, which has also shown efficacy in promoting weight loss. In what has become a crowded market, products will likely gain a competitive edge if they can fight both metabolic disease and obesity, the latter with a potential market of 300 million people worldwide. Amylin already looks to have consolidated its position. According to Collins Stewart analyst Salveen Kochnover, “They may not be first to market, but I think Amylin signed a nice partnership, so they’re well-positioned if any of their drugs work.” The nearest competitor, she says, is Vivus, the Mountain View, California–based company, whose once daily capsule Qnexa (phentermine and topiramate) for the diabetes market achieved 15% weight loss in phase 3 trials. Elsewhere, Novo Nordisk has launched a long-acting GLP-1 analog Victoza (liraglutide; Nat. Biotechnol. 27, 682–685, 2009) in Europe as a type 2 diabetes treatment. At present, the Copenhagen-based biotech still awaits a decision from the US Food and Drug Administration. Recent studies show Victoza to be more effective in helping people shed weight than the anti-obesity pill orlistat, marketed as Xenical by Roche and Alli by London-based GlaxoSmithKline. But Amylin’s foothold in the marketplace is further strengthened by its two first-in-class synthetic gut hormone drugs approved for type 2 diabetes—the insulin boosting glucagon-like peptide 1 (GLP-1) analog Byetta (exenatide), and Symlin. Symlin is a synthetic version of amylin, a neuroendocrine hormone secreted by pancreatic beta cells, that boosts insulin action and helps regulate appetite, food intake The path to approval for and glucose control. In diabetes sufferers, sensitivity to Symlin is often reduced over Amylin and its competitors time, in much the same way that insulin sensitivity is lost. By combining Symlin with metreleptin, a recombinant version of human leptin (a hormone secreted by fat cells won’t be easy. that acts on the hypothalamus to regulate food intake), the company hopes to treat obesity. Thus far, the drug combo has shown impressive weight loss in animal and human trials, and Amylin is expecting to announce favorable results from phase 2 trials following positive top-line data. The other drug included in the Takeda deal is davalintide, a second-generation amylin receptor agonist, which mimics the action of amylin. The path to approval for Amylin and its competitors won’t be easy. Safety issues that led to the withdrawals of the cannabinoid receptor antagonist Acomplia (rimonabant) and, before that, appetite suppressant Redux (dexfenfluramine), mean that novel products are likely to face intense scrutiny. In a recent study, the Symlin/metreleptin combination produced impressive weight loss in a broad population of obese people, but a subanalysis showed less dramatic effects in severely obese individuals. And, according to Tom Hughes, CEO of obesity therapeutics company Zafgen, headquartered in Cambridge, Massachusetts, although Amylin’s injectible is more expensive and inconvenient than an oral drug, the ‘hormone replacement’ approach avoids the idiosyncratic toxicity inherent to small molecules. “I think that may be at the heart of why this deal got as much as it did at the stage that it’s at,” he says. “It is appealing on a number of levels to think that you’re giving back something that is important or missing, which is restoring the balance to the patient and leading to the effect.” However, John Wilding, a clinical researcher at the University of Liverpool, UK, says there’s only a nugget of truth in the claim. Pharmacological doses are orders of magnitude higher than physiological concentrations, and in type 2 diabetes, amylin levels actually tend to be high. Hayley Birch

nature biotechnology volume 28 number 2 february 2010

109

NEWS

Peruvian GM advocate faces criminal charges

© 2010 Nature America, Inc. All rights reserved.

RPP

A molecular biologist could face a prison sentence for criticizing a report on transgenic gene spread. Ernesto Bustamante Donayre, vice president of the Peruvian College of Biologists, a professional organization, stands Ernesto Bustamante on accused of defamation, Peruvian RPP radio. a criminal offense, which in Peru can carry a prison term or fine. What triggered the suit was his public criticism of a report prepared by Antonietta Ornella Gutiérrez Rosati, a biologist at the La Molina National Agricultural University in Lima, identifying a P34S promoter and NK603 and BT11 transgenes in 14 of 42 maize samples from the Barranca region. Gutiérrez sent summaries of her findings to both the National Agricultural Research Institute and El Comercio newspaper in 2007 calling for a moratorium on transgenic crops until biosafety regulations are in place to prevent the spread to human food. Bustamante, a frequent contributor to radio and print, with no financial links to crop companies, described the alleged detection of three simultaneous transgenic events from two firms as “absurdly improbable” in his newspaper column and called for her claims to be peer reviewed. “The main point of my criticism,” Bustamante says, “was her going to the press instead of to her peers.” After Bustamante refused to retract his statements, Gutiérrez filed a suit for defamation. She later presented her findings to the Peruvian Genetic Society of which she is president, but would not comment on the case, except to say that “you must use respect” in scientific discussion and that her critics have “polarized” the debate. Although Peruvian farmers already import transgenic products for animal feed, several interest groups oppose their widespread introduction, which they label a foreign intrusion and threat to Peruvian biodiversity. An ongoing investigation is seeking to replicate Gutiérrez’s findings, but the government lacks the regulations to enforce its biosafety laws even if it does find transgenic crop outcrossing. The criminal case, however, threatens to stifle all scientific discussion. “Regardless of whether he gets sentenced or not I don’t think anyone is going to criticize anything,” says plant scientist Wayne Parrott, from the University of Georgia, a regular visitor to Peru. Bustamante’s colleague and supporter Luis Destefano Beltrán of the Cayetano Heredia Peruvian University agrees that “many people have tried to avoid taking sides.” Peru retains criminal defamation laws, which the Inter-American Commission on Human Rights concluded in 1995 are incompatible with the American Convention on Human Rights. Bustamante, who expects a ruling early this year, says, “The point is not whether I’m right or wrong. It’s the fact that for criticizing somebody on scientific grounds I’m being tried in criminal court.” Lucas Laursen

110

EC convenes crisis talks on European biotech sector Last December, officials of the European Commission (EC), together with Emmanuel Chantelot, executive director of the industry trade association European Biopharmaceutical Enterprises, convened a closed meeting to discuss the plight of the European biotech sector. Held at the EC headquarters in Brussels, the meeting was attended by policymakers, CEOs from small-to-medium-sized (SME) companies, national biotech associations, venture capitalists (VCs) and big pharma representatives. The working group discussed the findings of an EC-commissioned survey carried out by the Danish Technological Institute (DTI) in Taastrup on the problems of access to finance faced by the biopharma industry. According to this study, lack of access to capital is threatening innovation and competitiveness in the sector, with 40% of companies facing extinction by the end of the year without a further cash injection. On the basis of the group’s discussions, several policy recommendations were put forward with the potential to increase sustainability of the European sector (Box 1). The DTI’s findings portray European biotech as a rapidly deteriorating sector: 7% of the region’s biotech SMEs need capital immediately, 40% must raise capital within a year and nearly 75% over the next 2 years. “Some SMEs are going to go out of business, and many are stretching resources and cutting back on programs” says Thomas Saylor, chair of the SME platform of the European Association for Bioindustries (EuropaBio). The data were gathered through a survey of 87 biopharma companies in Europe throughout May and June 2009, desk research of reports and interviews with experts. The survey is deemed to be representative of the state of European biopharma, although there is an intentional bias toward smaller and younger companies, following the EC’s requested sampling criterion. “The core of the problem is that there is less venture capital money for small biotechs,” says Chantelot, and lack of cash creates funding gaps in the chain from startup to initial public offering. The most severe gaps are at the early, high-risk stages, making it hard for fledgling companies to get off the ground or even stay afloat. The key reason for this gap in Europe, says Ivica Cerina, a partner at NGN Capital in Heidelberg, Germany, is pressure over the past five years to “de-risk,” pushing investors to focus on later stages and avoid risky startups. Private sources of equity account, on average, for some 60% of all SME funding.

To address this problem, the DTI report recommends increasing public co-investments in micro-fund and business angels to provide the seed money needed to get an SME rolling, while simultaneously creating tax incentives for doing so. As the former type of funding tends to operate on different timescales from venture funding, and with different skill sets and strategic views, several of the VCs at the meeting also pushed for additional money. As such, Cerina would prefer to see more funds go into, or side-byside with, existing venture capital funds “either as dedicated early stage vehicles or directly into specialized, early stage–savvy VCs”. Enhancing access to existing pools of money could also support the recovery process, says Saylor, citing the European Union’s 7th Research Framework (FP7). “Funding has been cumbersome to apply for, and for small companies, in particular, it can be a huge administrative burden to take on the reporting requirements under the framework,” Saylor says. He would also like to see followon funding of the sort established by the UK’s Technology Strategy Board, which makes a

AP Photo/Virginia Mayo

in brief

Flying the biotech flag. Policymakers, investors and companies gathered at the Commission’s headquarters in Brussels to discuss how to overcome funding shortages faced by European biotechs.

volume 28 number 2 february 2010 nature biotechnology

© 2010 Nature America, Inc. All rights reserved.

news company eligible “for a bigger pot of money” on meeting certain milestones. Another example is the €2 ($2.9) billion fund available to SMEs until 2013 through the RiskSharing Finance Facility (RSFF) set up by the EC and the European Investment Bank. This is a great idea in principle, but the eligibility criteria tend to rule out biotech SMEs. To apply, companies must be profitable, which many biotechs are not for several years as they plough their money into R&D. “For biotech SMEs, the RSFF criteria need to be changed,” says Dirk Carrez, director of Industrial Biotech at EuropaBio. Others would like to see the Luxembourgbased European Investment Bank (EIB) and European Investment Fund (EIF) take bolder steps to support fledgling biopharma enterprises. Flexible loans and other financing instruments for companies close to self-sufficiency are currently made available by EIB and EIF, typically for late-stage co-investments of over €20–30 ($29–43) million a year. “We’d like to see a bigger share of that funding going to life science companies,” says Chantelot, who suggests the EC draw up a new mandate to give the sector priority in funding. Chantelot is also keen for big pharma to expand its corporate venture capital programs for early stage SMEs—a trend already underway (Nat. Biotechnol. 27, 403– 404, 2009). Public funds could help offset the risk, he adds. “For late-stage biotech companies in partnership with big pharma, EIB could lend or co-invest some money to minimize the risk for the other parties.” A key issue is determining whether the lack of capital currently stifling many companies reflects a problem with the financial instruments currently administered by the EIB and EIF or a problem with the companies themselves. In other words, are deserving companies being let down by the current system, or should these companies not be receiving funding at all as they are unlikely to become sustainable enterprises? “The EC is looking carefully at this question,” says Giulia Del Brenna, head of the EC unit on competitiveness in the pharmaceuticals industry and biotechnology, speaking in a nonofficial capacity. European officials are also considering another option, and that is to set up a European Biopharmaceutical Innovation Fund specifically dedicated to biotech startups. VCs argue that this could indeed be useful, particularly for early stage companies at risk, but stress the need for the fund to be administered with the same due diligence and market considerations that typically constrain private financing. Some VCs disagreed with the DTI report’s conclusions that capital supply was part of the problem. “Capital is available in EU, but it’s been

Box 1 Recommendations for European biotech The report commissioned by the EC (Directorate-General Enterprise and Industry) from The Danish Technological Institute is entitled The Financing of Biopharmaceutical Development in Europe (http://ec.europa.eu/enterprise/sectors/biotechnology/documents/ index_en.htm). It provides several suggestions for increasing the access companies have to capital: • The EC should undertake an in-depth analysis of the effects of different tech transfer models used within and outside Europe (good practice) to improve the effectiveness of biopharmaceutical R&D and commercialization and ensure that the sector is competitive and able to attract private funding. • Early stage investments should be encouraged to ensure that innovative companies continue their development activities, perhaps by supporting micro-funds and investments by business angels in early stage biopharmaceutical companies through public co-investments and tax incentives. • Establish a European Biopharmaceutical Innovation Fund focused on investing in biotech companies while considering the geographical reach of the existing funding mechanisms at European and national levels to ensure that global funding opportunities are exploited. • Improve the framework conditions for both biopharmaceutical companies and the venture capital industry in Europe.

kept dry,” says NGN’s Cerina. “While in better times most companies would have found a safe harbor, now there are too many companies that are not compelling enough chasing limited capital resources, and only the best ones will find investors.” Michiel de Haan, a VC and general partner of Aescap Venture Management, Amsterdam, agrees. “Out of 100 various proposals we look at we only invest in one or two. It’s a very strong selection process, and that doesn’t make us popular. But for high-quality propositions there is healthy competition and enough VCs around to invest.” Despite their reservations, VCs would still welcome moves to make investing in biotech SMEs more attractive, such as tax breaks and other forms of risk sharing with public investments. Programs that help companies with subsidies, guaranteed loans, and technology and innovation loans on which interest is paid back according to the success of the product work very well, says de Haan, who cites France’s tax-credit system and Holland’s Technology Start-Up Programme (a specialized seed fund), as illustrative examples. Indeed, de Haan argues that there’s a need to “look at those countries

where these types of financial instruments are very good for the biotech startup, and learn from them.” One other recommendation from the DTI report is to improve the quality of new venture propositions through better technology transfer from universities. For Marja Marakow, currently the chief executive of the European Science Foundation and former vice president of the University of Helsinki, a key issue is professionalizing technology transfer. “Public universities typically cannot hire first-class professionals with the requisite expertise to run tech transfer offices.” Instead, they are frequently staffed by civil servants who lack the relevant research and private-sector experience. The EC recognizes that biotech SMEs are important to Europe’s economy, but whether the biotech sector should receive targeted help remains an open question, and one to be explored further as a new commission comes into office this month. What is clear, says the EC’s Del Brenna, is that the EC is listening to the biotech sector and digesting the meeting’s recommendations. Dan Jones Brighton, UK

New product approvals Actemra (tocilizumab)

nature biotechnology volume 28 number 2 february 2010

Genentech (S. San Francisco, California)

The US Food and Drug Administration approved Genentech’s Actemra to treat rheumatoid arthritis. Actemra is the first US-approved drug to target interleukin-6 and is aimed at patients who do not respond to older tumor necrosis factor alpha inhibitors. Actemra is also approved in the EU, India, Brazil, Switzerland and Australia.

111

NEWS

© 2010 Nature America, Inc. All rights reserved.

Irish bioethics council axed Stem cell research in Ireland has been thrown into a state of confusion, after a recent government decision to cut all funding for the Irish Council for bioethics at the end of the year. Paradoxically, the move coincides with a recent Supreme Court decision that removes some of the legal uncertainties surrounding human embryonic stem cell research in the country. The judges denied a woman the right to proceed with in vitro fertilization without the consent of her estranged husband. In doing so, the court ruled that embryos outside the womb are not protected by the country’s constitutional protection of the unborn. Although this ruling affects human embryonic stem cell (hESC) research by providing clarification on the status of pre-implanted embryos, scientists remain wary of proceeding until a supporting framework is in place. “I’m going to behave responsibly. It’s going to be done by the book,” says Barry Moore at University College Cork (UCC), who has already received clearance to carry out hESC research from UCC’s research ethics committee. Ireland has no laws governing human stem cell research and scientists have been operating in a legal limbo. “The lack of an independent bioethics board will have serious repercussions for how Ireland is seen as a hub for medical research, and that will have to be addressed as a matter of urgency,” says scientific director Stephen Sullivan of the newly formed Irish Stem Cell Foundation, which is calling on the government to reinstate the council. Cormac Sheridan

Amgen trumps Roche A 5-year patent dispute between Roche and Amgen over the anti-anemia drug Mircera (methoxy polyethylene glycolepoetin beta) has ended. Roche of Basel acknowledged in court that Mircera, its pegylated-erythropoietin, infringed on Amgen’s erythropoietin patent and would drop its challenges. The ruling ensures that Mircera sales are barred and Roche is kept out of the US market until mid-2014, when Amgen’s patents expire. Amgen currently dominates the US market with erythropoiesisstimulating agents (ESAs)—Epogen (epoetin alfa) and Aranesp (darbepoetin alfa)—which together generated $5.6 billion in sales last year. However, Thousand Oaks, California–based Amgen may now have to contend with US Food and Drug Administration (FDA) regulations, as a panel of outside experts expected to meet in 2010 will re-examine safety concerns over ESAs (Nat. Biotechnol. 25, 607–608, 2007). Writing in January in the New England Journal of Medicine (doi:10.1056/NEJMp0912328), FDA officials are urging proper dosing of ESAs in individuals with chronic kidney disease, as certain regimens appear to increase the risk of cardiovascular events and death. The panel may impose regulations on the ESA market or decide that additional clinical trials are needed. The outcome of this meeting, says Eric Schmidt, a biotech analyst at Cowen and Company in New York, is that it may bring down sales, as drug companies may no longer be allowed to push high-dose regimens. Nazlie Latefi

112

Report blames GM crops for herbicide spike, downplays pesticide reductions A recent report published by the Organic Center, an organic farming advocacy organization headquartered in Foster, Rhode Island, claims that the use of herbicides in weed control has risen sharply since transgenic crops’ commercial introduction in 1996. Increasing cultivation of glyphosate (N-phosphonomethyl glycine)-tolerant transgenic crops, particularly soybean, has led to an aggregate increase in herbicide use of 383 million pounds over the past 13 years, on top of what the Organic Center’s chief scientist Charles Benbrook models suggest would have been applied had the technology never been deployed (http://www.organic-center.org/science.pest.php?action=view&report_id=159). The report also downplayed that transgenic corn and cotton have delivered reductions in insecticide use totaling 64.2 million pounds over the same time period. The report’s findings on herbicides are in stark contrast to the standard agrochemical industry line that transgenic crops have reduced the chemical load on the environment. Several critics have questioned the assumptions underlying the analysis and any significance that can be drawn from it, particularly as the report comes from an advocacy group seeking to “communicate the verifiable benefits of organic farming and products to society.”

Rising glyphosate resistance is a plausible explanation for the increasing use of herbicides, however. Among plant scientists, there is little disagreement on the problem of glyphosate-resistant weeds. “It certainly is fair to point out the failure in glyphosate stewardship, that the threat of resistance wasn’t appreciated, that more diverse management wasn’t used to try to prevent or delay resistance emerging,” says Chris Boerboom, extension weed scientist at the University of Wisconsin in Madison. The issue of herbicide resistance has already become acute in some US states. Report author Benbrook claims that the cotton and soy industries in the Southeast are on “the brink of collapse” because of the cost of dealing with glyphosate-resistant weeds. Benbrook goes on to argue that increasing reliance on herbicides paired with more expensive, engineered tolerance traits will erode farmers’ profitability, while compounding environmental and public health risks (through increased chemical exposure). The report’s other main finding—that insect-resistant transgenic crops have helped cut pesticide use—was downplayed by Benbrook, who claims the increase in the volume of herbicides applied “swamps” the

Greg Gardnes/istockphoto

in brief

Crop spraying on the up. Glyphosate-resistant weeds may be driving an increased reliance on herbicide use.

volume 28 number 2 february 2010 nature biotechnology

© 2010 Nature America, Inc. All rights reserved.

news benefits of decreased insecticide use attributable to corn and cotton expressing genes that encode one or more Bacillus thuringiensis (Bt) insect toxins. Bt crops could have a brighter future than herbicide-resistant transgenic varieties, the report states, “but if, and only if, [insect] resistance is prevented.” The report is based on extrapolations of pesticide use survey data compiled by the US Department of Agriculture’s (USDA) National Agricultural Statistics Service (NASS). Benbrook relies on annual trait acreage data compiled by St. Louis–based Monsanto to disaggregate transgenic crops from the total crop acreage. However, no NASS data on corn or soy are available for 2007 or 2008, years for which Benbrook posits unusually large pesticide increases of 20% and 27%, respectively. The main uncertainties stem from gaps in NASS data, which, since 2001, have only been gathered intermittently, and from that data’s failure to distinguish between pesticide use on transgenic crop varieties and on their conventional counterparts. Benbrook postulates the emergence of glyphosate resistance has fueled a sharp upswing in the use of other herbicides on glyphosatetolerant crops, whereas levels of herbicide used on conventional crops have fallen because of ongoing improvements in potency. But Janet Carpenter, formerly of the Washington, DC–based US National Center for Food and Agricultural Policy and now an independent agricultural biotech consultant, disagrees. “That’s all extrapolation,” she says. “The bottom line is we don’t know what has happened to pesticide use in the last couple of years.” Benbrook says that additional data from future surveys can be factored into his model when it becomes available. “The valid criticism—or valid question—is these are all average numbers,” he says. “I would place a fair amount of confidence in these averages as a reflection of what’s going on out there.” In a published critique of the report, Dorchester, UK–based consultancy PG Economics argues that Benbrook overestimates herbicide application rates for biotech crops and underestimates them for conventional crops (http://www.pgeconomics.co.uk/ pdf/OCreportcritiqueNov2009.pdf). It cites a new study from the US Geological Survey, which found that concentrations of several major pesticides either declined or remained constant in US corn belt rivers and streams during 1996–2006 (http://pubs.usgs.gov/ sir/2009/5132/). However, the study period does not include the two most recent years, during which Benbrook claims the greatest increase in herbicide use has occurred. PG

Economics, which also published a lengthy study on the global socioeconomic and environmental impacts of transgenic crops in May last year, has drawn on two sources: pesticide use data from a commercial source, DMR Kynetec, of St. Louis, which Benbrook says is in general agreement with his own findings; and what he describes as ‘faulty’ simulation data generated by the Washington DC–based National Center for Food and Agricultural Policy, based on exercises run with university extension weed scientists. “It’s impossible to reconcile their estimates with the NASS data,” Benbrook says. In the meantime, several scientists have voiced support for the general thrust of the study. “There’s nothing surprising there,” says Matt Liebman, who holds the H.A. Wallace chair for Sustainable Agriculture at Iowa State University in Ames. Dealing with glyphosate-resistant weeds will require alterations in cropping systems that rely solely on the marriage of the herbicide-tolerance trait and the associated herbicide to control weeds. Widespread convergence on a narrow range of options, such as the rotation of glyphosate-resistant corn and glyphosateresistant soybean, has been a significant factor, says Liebman. “You have good conditions for rapid selection of herbicide resistance.” Monsanto and its competitors are responding to the problem by offering farmers subsidies to include third-party herbicides in their weed control systems. They are also stacking additional tolerance traits that can be paired with other herbicides, such as dicamba (3,6-dichloro-2-methoxybenzoic acid), glufosinate (phosphinothricin) and 2,4-d (dichlorophenoxyacetic acid). External factors have hampered progress, however. “The biggest contributor to weed resistance has been the European Union’s slow approval process for new biotech-enhanced seeds. After many years of delays, the EU finally granted approval of Liberty Link [phosphinothricin-acetyltransferase] soybeans, which are resistant to a different active ingredient [l-phosphinothricin],” says Bob Callanan, communications director of the American Soybean Association, located in St. Louis. Critics argue that more diversified approaches will be needed, such as alternative crop rotations, novel herbicides—it’s 20 years since a new mechanism of action was commercialized, notes Boerboom—and alternative weed control methods. “If you want to keep this tool available and effective there has to be some way, short of fallowing a field, of delaying the development of resistant weeds,” says Robert Kremer, of the USDA’s Agricultural Research Service at

nature biotechnology volume 28 number 2 february 2010

Columbia, Missouri. The market dominance of transgenic crop varieties limits some of the options, however. “It’s very difficult to go and find nontransgenic soybean,” he says. “Conventional corn rotated with Roundup Ready [glyphosate-resistant] soybeans would be very logical,” says Boerboom. “We have an excellent selection of conventional herbicides we can use in corn.” That Monsanto’s Roundup Ready cropping system has been a major hit with farmers is not in dispute. “The simplicity, the high efficacy and the perceived low cost have been very attractive,” says Liebman. In this respect, even Benbrook agrees: “The weed management systems that Roundup Ready [crops] replaced were unforgiving and required a high level of skill and management to get the benefits out of them,” he says. What’s more, glyphosate, which inhibits 5-enolpyruvylshikimic acid-3-phosphate synthase, a plant enzyme involved in amino acid biosynthesis (the engineered trait comprises a bacterial form of the enzyme, which is unaffected), has a relatively benign environmental profile in comparison with many other herbicides. Moreover, it has allowed many crop growers to shift to no-till agriculture, which reduces fossil fuel inputs. The problems farmers are encountering now are not new. “The selection for glyphosate resistance is not unique. We’ve selected for a whole lot of other herbicide families as well,” says Aaron Hager, weed science extension specialist at the University of Illinois at Urbana-Champaign, Illinois. “There’s a plant somewhere in the world that’s resistant to an herbicide that hasn’t even been discovered yet. That’s how selection occurs.” As glyphosate has played a central role in US crop production over the past decade, it can be argued the technology has become a victim of its own success. For many farmers, weed control will, however, soon become more complex. Some of the alternatives offer less favorable environmental profiles. Dicamba, a synthetic auxin or plant hormone, can drift off-site and interfere with flowering plants, for example. “There will be objections raised to it by the environmental community because of nontarget effects,” says Liebman. Nevertheless, US agriculture is not facing a doomsday scenario, according to Boerboom. “I don’t think it’s like we’re going into some dark age of chemical use on the landscape,” he says. But a new phase in the molecular arms race between biotech and nature is getting underway. Cormac Sheridan Dublin

113

NEWS

in brief

© 2010 Nature America, Inc. All rights reserved.

Industry gains on moneyback schemes Risk-sharing agreements that assess innovative drugs based on long-term cost effectiveness may not be helping governments save money, a new study suggests. “In the short term, it’s been to [industry’s] advantage,” says lead investigator Mike Boggild, a neurologist at The Walton Centre in Liverpool. In 2002, the UK government entered a ‘risk-sharing’ agreement over five multiple sclerosis drugs that the UK’s National Institute for Health and Clinical Excellence (NICE) had deemed too expensive. NICE reversed its decision after drug makers dropped their prices and agreed to reimburse government if the drugs did not prove cost effective in the long term. The study results based on two-years’ data suggest that the drugs are not cost effective, although Boggild warns it is too early to draw firm conclusions. “The cost effectiveness of the drugs can go in either direction, depending on which assumptions we use,” he says. This type of scheme is inherently difficult to run, adds Jon Nicholl, director of the Medical Care Research Unit at Sheffield University, UK, because stakeholders have conflicting interests: the state wants to reduce costs, whereas industry wants to maximize profits. A different approach, in which firms refund treatment costs for nonresponsive patients, may be a better way to improve cost effectiveness, he says. Asher Mullard

$2 million rice verdict against Bayer A St. Louis district court has ordered Bayer CropScience to pay over $2 million in compensatory damages to two Misssouribased rice farmers whose crops cross-bred with the company’s genetically modified (GM) LibertyLink during field testing. When the unwanted presence of transgenic rice was discovered in 2006, several countries halted US rice imports, which led to farmers’ economic loss and prompted more than 1,000 similar lawsuits against Bayer CropScience, whose US operations are based in The Research Triangle Park, North Carolina. This first trial, whose verdict was issued last December, has been called a bellwether case. “We are studying the court’s award of monetary damages in detail and are considering our options,” says Richard Breum, corporate spokesperson for Bayer CropScience in Monheim, Germany. “Since each case is different, we evaluate each separately. Last year the court ruled against the plaintiffs in their efforts to obtain class action status in the litigation, noting overall differences in individual plaintiff’s situations and claims.” In 2007 the US Department of Agriculture (USDA) decided against pursuing enforcement action against the company. It noted that investigators within the Animal and Plant Health Inspection Service (APHIS) at USDA were “unable to make any definitive determinations” about the inadvertent release, during field trials, of two varieties of LibertyLink rice that then mixed with commercial rice crops in Missouri and several neighboring states. Jeffrey L Fox

114

Biorefineries’ stimulus win Nineteen start-ups have landed the bulk of federal stimulus funding earmarked for industrial biofuel and biomass programs. The US Department of Energy (DOE) in December announced $564 million in funding towards the building and operating of facilities that convert nextgeneration feedstocks such as switchgrass and wood chips into fuels and products. Grants range from $2.5 million to $81.1 million each (Table 1), which dwarf funds allocated to related areas such as plant genomics research. Small-scale or pilot facilities will receive up to $25 million, demonstration scale $50 million, and one company, Bluefire Ethanol in Irvine, California, more than $81.1 million to build a commercial plant. Amyris Biotechnologies, for example, will add its $25 million to the $165 million investment money it has accrued over the last 7 years. The Emeryville, California–based company will use the stimulus grant to expand its pilot facility, explore feedstocks for making renewable hydrocarbons and scale-up production of both fuel and biobased chemicals, says Kinkead Reiling, cofounder. But the money is not intended to cover all biorefinery building costs—the DOE expects grant winners collectively to match prize funds with at least $700 million in nonfederal investment. “[The grants] can boost investor confidence in those projects and allow companies to attract the full amount of the funding needed to get the project done,” says Paul Winters, a spokesperson for the Biotechnology Industry Organization in Washington, DC. Adds Reiling, “It’s an excellent shot in the arm for the industry, but compared to the size of the problem [energy crisis], it’s small.” The stimulus bill, known as the American Recovery and Reinvestment Act, was passed in February 2009. Emily Waltz

Table 1 Selected biofuel companies receiving US stimulus funds Grant ($ million) Project description

Company /location Bluefirea/California

81.1

To construct a facility that produces ethanol fuel from woody biomass, mill residue and sorted municipal solid waste. The facility will have the capacity to produce 19 million gallons of ethanol per year and will be in Fulton, Mississippi.

BioEnergy Internationalb/ Lake Providence, Louisiana

50.0

To produce succinic acid from sorghum. The biological process being developed displaces petroleum-based feedstocks and uses less energy per ton of succinic acid produced than its petroleum counterpart.

Enerkemb/ Pontotoc, Mississippi

50.0

Located at an existing landfill, this project will use feedstocks such as woody biomass and biomass removed from municipal solid waste to produce ethanol and other green chemicals through gasification and catalytic processes.

INEOS New Planet BioEnergyb/ Vero Beach, Florida

50.0

This project will cultivate algae in ponds that will ultimately be converted into green fuels, such as jet fuel and diesel, using the Dynamic Fuels refining process.

Sapphire Energyb/ Columbus, New Mexico

50.0

To cultivate algae in ponds that will ultimately be converted into green fuels, such as jet fuel and diesel, using the Dynamic Fuels refining process.

Algenol Biofuelsc/ Freeport, Texas

25.0

To produce ethanol directly from carbon dioxide and seawater using algae. The facility will have the capacity to produce 100,000 gallons of fuel-grade ethanol per year.

UOPc/ Kapolei, Hawaii

25.0

To integrate existing technology from Wilmington, Delaware– based biofuels firm Ensyn and UOP to produce green gasoline, diesel and jet fuel from agricultural residue, woody biomass, dedicated energy crops and algae.

ZeaChemc/ Boardman, Oregon

25.0

To use purpose-grown hybrid poplar trees to produce fuel‐ grade ethanol using hybrid technology. Additional feedstocks such as agricultural residues and energy crops will also be evaluated in the pilot plant.

HALDOR TOPSOEc /Des Plaines, Illinois

25.0

To convert wood to green gasoline by fully integrating and optimizing a multi‐step gasification process. The pilot plant will have the capacity to process 21 metric tons of feedstock per day.

ICMc/ St. Joseph, Montana

25.0

To modify an existing corn‐ethanol facility to produce cellulosic ethanol from switchgrass and energy sorghum using biochemical conversion processes.

Amyris Biotechnologiesc

25.0

To produce a diesel substitute through the fermentation of sweet sorghum. The pilot plant will also have the capacity to coproduce lubricants, polymers and other petrochemical substitutes.

aIncreased

funding to existing biorefinery projects. bDemonstration scale. cPilot scale. Source: US Department of Energy

volume 28 number 2 february 2010 nature biotechnology

news

in brief

Purpose-built chromosome

© 2010 Nature America, Inc. All rights reserved.

“It’s functional, and also a very good metaphor for what the center is trying to achieve.” Larry Malcic, one of the architects of London’s UK Centre for Medical Research and Innovation (UKCMRI), says scientists exclaimed, “that’s a chromosome,” when he presented the building designs without knowing its symbolic significance. The new $978 million UKCMRI is being built in central London as a partnership between University College London, Cancer Research UK, the Medical Research Council and the Wellcome Trust. It will house four leading science organizations to conduct biomedical research on genetics, stem cells and common diseases, and is expected to open in 2015. (Times, December 8, 2009)

in their words “They just wait until WHO [World Health Organization] says ‘pandemic’ and activate the contracts.” Wolfgang Wodarg, a member of the German Social Democratic Party and chair of PACE health committee, conveniently shifts blame for Germany’s surplus H1N1 vaccine stocks on to the companies that redirected resources and expertise to make a product available in just a few months. (Pharma Times, January 4, 2010) “These sweetheart deals are being done on the backs of consumers. From the perspective of the Federal Trade Commission, [they] are one of the worst abuses across the board in healthcare and

should be stopped.” Federal Trade Commission (New York) chairman Jon Leibowitz will press for a provision in the healthcare reform bill to end deals in which brand-name drugmakers pay generic producers to delay copycat versions of best-selling meds. (New York Times, January 12, 2010) “The pharmaceutical industry has destroyed so much institutional knowledge over the last decade that it makes the Taliban, blowing up temples, look like high school pranksters.” Anonymous blogger. (In the Pipeline, January 12, 2010) “Cannibalism is rife within the biotech industry!” Barry Canton, a cofounder of Ginkgo Bioworks (Boston), on how his and other companies are acquiring equipment castoffs from universities and other companies from online auctioneers. (The Boston Globe, January 4, 2010)

FDA balks on MedImmune’s cell-grown flu vaccine The shift towards new cell culture–based flu vaccine production has been dealt a blow as MedImmune of Gaithersburg, Maryland, puts its manufacturing efforts on hold. The AstraZeneca subsidiary took this step after the US Food and Drug Administration (FDA) requested follow-on studies that would substantially increase the cost and time to market beyond what the company expected. In its contract with the Department of Health and Human Services (HHS), MedImmune proposed an efficacy study comparing immune responses in volunteers receiving cellproduced with those receiving egg-produced vaccines, considering them genetically identical, followed by a large safety trial. But the FDA termed cell-grown vaccine a new product, requesting Medimmune conduct a clinical trial during an influenza season, as well as demonstrate efficacy in adults before vaccinating children. The plan “became cumbersome and complicated and did not address significant scientific and medical issues we thought we needed to address to advance this vaccine,” says George Kemble, vice president of vaccine R&D at MedImmune. “I don’t think there is any deliberate delay,” says Anthony Fauci, director of the National Institute of Allergy and Infectious Diseases, noting the move is due to safety and efficacy data gathering. Jose Romero, member of the FDA vaccine advisory committee, comments in an unofficial capacity, “General FDA concerns include exposing humans to adventitious agents that might be lurking in cell lines or the remote possibility of transmitting an oncogene that could create cancer in a human host.” Elsewhere, last November, Novartis of Basel inaugurated a $1 billion cell culture flu vaccine manufacturing facility in partnership with the HHS. The plant in Holly Springs, North Carolina, is the first large-scale cell culture flu vaccine and adjuvant production facility in the US. Wendy Wolfson

Selected research collaborations Partner 1

Partner 2

$ (millions) Details

Alopexx Pharmaceuticals Sanofi-Aventis (Paris) (Cambridge, Massachusetts)

375

Sanofi-Aventis will pay Alopexx for rights to codevelop a monoclonal antibody (mAb) for treating Escherichia coli, Staphylococcus aureus and other infections. Alopexx receives an upfront payment, research funding and is eligible for milestone payments that could reach $375 million in total, plus royalties. Sanofi will have the option to license the product, which will be in phase 1 trials in 2010.

Seattle Genetics (Bothell, Washington)

Millennium/ Takeda (Osaka, Japan)

290

Millennium will pay $60 million upfront, plus milestones that could exceed $230 million, to codevelop Seattle Genetics’ brentuximab vedotin (SGN-35). The antibody drug conjugate composed of an anti-CD30 mAb and monomethyl auristatin E is currently in a pivotal phase 2 trial to treat relapsed and refractory Hodgkin’s lymphoma. Under the agreement, the Takeda Group keeps commercial rights to the drug outside the US and Canada where Seattle Genetics retains full rights.

Athersys (Cleveland)

Pfizer (New York)

111

Pfizer will pay Athersys $6 million initially and up to $105 million in the future for rights to develop Athersys’s stem cells to treat ulcerative colitis and Crohn’s disease. The product, MultiStem, consists of multipotent adult progenitor cell, and is in early clinical trials for heart attacks and in cancer patients receiving bone marrow transplants.

Syngenta (Basel)

CSR Sugar (Melbourne, Australia)

*

Syngenta has acquired exclusive global rights, excluding Australia, to CSR Sugar’s SugarBooster, a transgenic technology to develop cane plants with high sugar content. The license agreement includes milestone payments and royalties on product sales to CSR Sugar. The terms of the deal were not disclosed.

*Financial details not disclosed.

nature biotechnology volume 28 number 2 february 2010

115

data page

2009: Turning the corner Walter Yang Although initial public offerings showed signs of resuscitation (at least 13 more companies are now in the queue), follow-on financings came in above $6 billion—the second-best year over the past decade.

Stock market performance

Global biotech industry financing

Although biotech indices were up ~16% last year, they underperformed other major indices.

The boost in partnership promises to US biotechs and follow-on financings pushed industry funding to $61.3 billion, up 82% from 2008.

1,500

Swiss Market

S&P 500

1,400

NASDAQ Biotech

Dow Jones

1,300

NASDAQ

BioCentury 100

Year

Venture money fell slightly last year, as private companies raised $5.1 billion versus $5.3 billion in 2008. 107 1,436 5,241

8,000 7,000

4,000

40 797 3,151

78 1,451 3,871

69 1,213 4,045

58 1,206 4,315

Asia-Pacific 55 1,145 4,070

53 1,130 3,947

Europe Americas

3,000 2,000 1,000 0

2003

2004

2003 Americas 184 Europe 77 Asia-Pacific 5

2005 2004 190 82 6

2005 182 95 9

2006 Year 2006 204 81 8

2007 2007 223 105 9

2008 2008 212 101 6

2009

2009 196 87 4

Table indicates number of VC investments and includes rounds where the amount raised was not disclosed. Source: BCIQ: BioCentury Online Intelligence

17.3 6.1 5.4 2.7 4.8 1.9 10.9 8.8 5.3 2.9 3.3 2.6 8.9 9.1 4.0 2.2 3.9 0.5

10

20 30 40 50 Amount raised ($ billions)

60

IPOs

70

Partnership figures are for deals involving a US company. Source: BCIQ: BioCentury Online Intelligence, Burrill & Co. PIPEs, private investments in public equity; IPOs, initial public offerings

Global biotech initial public offerings (IPOs) The North American IPO market showed signs of life, with four companies raising $705 million versus only one raising $6 million in 2008. 3,500 Amount raised ($ millions)

12/09

11/09

10/09

9/09

8/09

7/09

6/09

5/09

Month ending

Global biotech venture capital (VC) investment

VC amount raised ($ millions)

© 2010 Nature America, Inc. All rights reserved.

4/09

2003 0

3/09

800 2/09

2004

1/09

2005

Partnering Debt and other Venture capital PIPEs Follow-ons

19.8 11.9 5.6 4.7 5.6 2.0

2006

900

5,000

22.4 11.7 6.8 4.7 4.4 3.0

2007

1,000

6,000

20.0 3.2 5.3 3.1 1.9 0.1

2008

1,100

700

36.9 10.0 5.1 2.2 6.0 0.9

2009

1,200

12/08

Index

Despite the shaky start to 2009, the biotech sector regained its financial footing. Biotech indices were up, as were offerings and partnership monies. Excluding collaborations, the sector raised a total of $24.3 billion.

585 1,055 1,309

230 474 1,852

3,000 2,500

15 930 913

2,000

Asia-Pacific Europe

98 869 1,063

Americas 65 158 705

1,500 1,000 500 0

43 19 483

2003

12 115 6

2004

2003 Americas 8 Europe 1 Asia-Pacific 5

2005 2004 35 12 6

2005 18 24 3

2006 Year 2006 26 21 3

2007 2007 23 21 7

2008 1 3 2

2008

2009

2009 4 3 3

Table indicates number of IPOs. Source: BCIQ: BioCentury Online Intelligence

Notable 2009 deals IPOs

Company (lead underwriters) Talecris (Morgan Stanley, Goldman Sachs, Citigroup, JPMorgan) Movetis (Credit Suisse, KBC) Cumberland (UBS, Jefferies, Wells Fargo) Omeros (Deutsche Bank) China Nuokang (Jefferies)

Percent Amount change in raised stock price ($ millions) since offer $550.0 17% $146.0 $85.0 $68.2 $40.7

3% –5% –30% –13%

Date completed 30-Sep 3-Dec 10-Aug 7-Oct 9-Dec

Mergers and acquisitions Target Genentech Sepracor Medarex CV Therapeutics Cougar Biotechnology Ovation Pharmaceuticals Proteolix

Acquirer Roche Dainippon Sumitomo Bristol-Myers Squibb Gilead Sciences Johnson & Johnson H. Lundbeck Onyx Pharmaceuticals

Value ($ millions) $46,800 $2,600 $2,400 $1,400 $1,000 $900 $851

Date announced 12-Mar 3-Sep 22-Jul 12-Mar 21-May 09-Feb 12-Oct

Licensing /collaboration Venture capital Company (lead investors) Clovis Oncologya (Domain, New Enterprise Associates, Versant, Aberdare, Abingworth, Frazier, ProQuest, company management) Zogenix (Clarus, Domain) BioVex (Forbion, Morningside, Ventech, MVM) Pacific Biosciencesb (Deerfield, Intel) Hyperion (Bay City Capital, Panorama Capital) NovImmune (BZ Bank) Sopherion (Zoticon Bioventures) aLead

investor not available. bSeries E extension.

116

Amount raised ($ millions) $145.0

Round number NA

Date closed 21-May

$71.0 $70.0 $68.0 $60.0 $56.4 $55.0

2 6 5 3 NA 3

07-Dec 10-Nov 12-Aug 30-Jun 12-May 18-Feb

Value Researcher Investor ($ millions) Deal description PTC Roche $1,924 Develop small molecules against four central nervous Therapeutics system disease targets Nektar AstraZeneca $1,505 Worldwide rights to NKTR-118 for opioid-induced constipation and NKTR-119 for pain without constipation Incyte Novartis $1,310 Ex-US rights to INCB18424; in phase 3 for myelofibrosis; worldwide rights to preclinical INCB28060 Targacept AstraZeneca $1,240 Worldwide rights to develop and commercialize major depressive disorder compound TC-5214 Exelixis Sanofi>$1,161 Exclusive, worldwide rights to XL147 and XL765 in aventis phase 1b/2 to treat cancer ZymoGenetics Bristol-Myers $1,105 Codevelop and commercialize ZymoGenetics’ phase 1 Squibb hepatitis C virus compound PEG-interferon lambda Amylin Takeda >$1,075 Codevelop and commercialize therapeutics for obesity and related indications Alder Bristol-Myers $1,069 Worldwide rights to ALD518 for all indications, except Squibb cancer

volume 28 number 2 february 2010 nature biotechnology

ne w s feat u re

The HER2 testing conundrum

© 2010 Nature America, Inc. All rights reserved.

Problems in interpreting diagnostic tests for HER2 may be compromising patient access to effective treatments. As new versions of therapies targeting HER2 work their way through clinical trials, will the situation get even murkier? Malorye Allison investigates. A recent study from the University of California, San Francisco, reveals that one in five HER2 tests gives the wrong answer1. Furthermore, the article, which reviews the medical literature, reports that as many as two-thirds of breast cancer patients who should be tested for HER2 are not, and consequently a significant fraction of women treated with Genentech’s Herceptin (trastuzumab) have never been tested for HER2 overexpression. The health benefit provider Wellpoint, of Indianapolis, might dispute that finding. According to Genentech staff scientist Mark Sliwkowski, the insurer has data showing that 98% of its breast cancer patients are tested. However, doctors differ in their views on testing before prescribing Herceptin. “Some doctors don’t know how to interpret test results,

they prefer just to prescribe it and assess the patient’s progress,” says Michael Liebman of the patient stratification company Strategic Medicine of Kennett Square, Pennsylvania. More than a decade after the drug received US Food and Drug Administration (FDA) approval, the personalized medicine paradigm clearly has holes. Many experts are frustrated and troubled by the state of HER2 testing, especially as new opportunities for tests are on the horizon. And as trials testing Herceptin at earlier stages and in combination with other drugs continue, experts are starting to wonder what besides HER2 overexpression might be influencing an individual’s response to the drug. These questions promise to not only spur the development of a range of new tests to guide breast cancer therapy but also fundamentally

change understanding of this disease, lead to new treatments and potentially have an impact on treatment of other cancers. Testing tempest Personalized medicine proponents point to Herceptin as a paradigm changer: the monoclonal antibody targeting HER2 (also referred to as HER2/neu and ERBB2) evens the playing field for breast cancer patients overexpressing HER2, whose tumors are typically more aggressive. But testing was problematic from the start, due to either sloppy execution or complex tumor biology. “Giving Herceptin early improves outcome so dramatically that it is an absolute tragedy to miss patients who should be getting it,” says Jeffrey Ross, from Albany Medical College in Albany, New York, who helped develop a fluorescent in situ hybridization (FISH) HER2 test marketed by Downers Grove, Illinois–based Vysis (Table 1). FISH tests are currently considered the gold standard. As more data become available and the HER2 story evolves, it’s becoming clear that some pieces don’t fit together quite as well as they might. For example, patients whose tumors have progressed on the drug, sometimes respond to Herceptin when it is given later with chemotherapy. Furthermore, fewer than 50% of HER2-positive metastatic breast

Table 1 Selected HER2 tests Company Location

Name of test Status

Technology

Biogenex San Ramon, California

InSite HER2/neu CB11 FDA approved

Immunohistochemistry assay using a monoclonal antibody directed against the internal domain of HER2/neu available either in automated or manual formats

Dako Glostrup, Denmark

HER2 FISH pharmDx Kit FDA approved

FISH assay to determine HER2 gene amplification in formalin-fixed, paraffin-embedded breast cancer specimens. Gene amplification is determined from the ratio between the number of signals from the hybridization of the HER2 gene probe and the number of signals from the hybridization of the reference chromosome 17 probe (green signals)

Dako

HercepTest FDA approved

Semi-quantitative immunohistochemistry assay for determination of HER2 protein overexpression in breast cancer tissues routinely processed for histological evaluation

Genomic Health

Oncotype DX CLIA validated

RT PCR–based assay analyzes the expression of a panel of 21 genes, among them HER2. Oncotype DX predicts disease recurrence and assesses benefit from certain types of chemotherapy

Invitrogen Carlsbad, California

SPOT-Light HER2 CISH Kit FDA approved

Chromogenic in situ hybridization (CISH) using a DNA probe. Quantifiable results are visualized under a standard brightfield microscope.

Monogram Biosciences

HERmark Breast Cancer Assay CLIA-validated

Proximity-based assay, which provides direct quantitative measurements of HER2 total protein and HER2 homodimer levels

Siemens Healthcare Diagnostics HER2/neu ELISA Erlangen, Germany FDA approved

Sandwich enzyme immunoassay using mouse monoclonal for capture and a different biotinylated mouse monoclonal antibody for the detection of human HER2/neu protein. Detection is by direct chemiluminescence. Protein is quantified by spectrophotometry

Ventana-Roche Tucson

Fully automated silver in situ hybridization assay for HER2 and chromosome 17 Inform HER2 Silver in situ Hybridization Approved in Europe and elsewhere but not detection. Chromogenic signals are detected through the use of silver deposition technology. Results and morphological significance can be interpreted using convenby FDA tional brightfield microscopy

Ventana-Roche

Pathway anti-HER2/neu (Clone CB11) FDA approved

Semiquantitative immunohistochemistry assay using a monoclonal antibody for the detection of c-erbB-2 (HER2) antigen using Ventana’s family of automated instrument platforms

Vysis (Abbott)

PathVysion HER2 DNA Probe Kit FDA approved

Fluorescence in situ hybridization (FISH) assay to determine HER2 amplification, using LSI HER2 probe, which spans HER2, and CEP 17 probe, which hybridizes to the alpha satellite DNA located at the centromere of chromosome

CLIA, Clinical Laboratory Improvement Amendment; ELISA, enzyme-linked immunosorbent assay.

nature biotechnology volume 28 number 2 february 2010

117

© 2010 Nature America, Inc. All rights reserved.

N E W S feat u re tumors respond to Herceptin alone2. “It’s not positive, but he admits that the discrepancy just a simple translation of gene overamplifi- could just reflect problems with the tests. His cation to susceptibility to the drug,” says Larry group is doing further microarray analysis of Norton, of Memorial Sloan-Kettering Cancer the NSABP trial samples. If the earlier findings are confirmed, a trial could be launched in the Center in New York. Others wonder if the test is even necessary. summer of 2010 to test Herceptin in patients Studies from two clinical trials (NSABP B31 with HER2-negative tumors. “We have an NCIand NCCTG N9831) presented at the American approved protocol,” Paik writes in an e-mail. Society of Clinical Oncologists (ASCO) annual meeting in 2007 (ref. 3) suggested that some Abundance of riches individuals with HER2-negative tumors From the time Herceptin was launched, experts can benefit from Herceptin. In these trials, have warned that existing tests have problems. The most comwhich compared monly used test, chemotherapy an immunohisalone with che- a 17q11.2–q12 region tochemistrymotherapy plus Centromere Telomere based assay, Herceptin, only Her2 (ERBB2) happens to be women who were the least dependHER2 positive able, especially could participate. ~190 kb when performed Upon retesting, LSI HER2 in laboratories however, some that do only tumor samples b occasional tests. came up negaThe immunohistive. Nonetheless, tochemistry test some women measures protein with negative levels, whereas test results benthe newer FISHefited from the based tests meadrug, which has sures gene copy spurred a now number and long-running are believed to debate. Although be more relimany experts able, especially believe this findin expert hands. ing to be an artiSome think fact of variation the FISH assay in test accuracy, should be the others think this Figure 1 FISHing for HER2. (a) Probe map shows the standard, but may be another relative size of the Vysis LSI HER2 probe and the gene. clearly, it makes important clue. (b) An example of a FISH test for HER2 amplification a big difference “It’s easy to shows multiple copies of the HER2 gene (red clusters who is doing the dismiss a find- signals) compared to chromosome number (green signals). testing. In 2006, ing you can’t Source: Abbott. ASCO and the explain, but this is forcing us to reexamine our notions of what College of American Pathologists released being HER2 positive or negative means,” says stricter guidelines which, according to Ross, forced many laboratories that had low test volNorton. The controversy around these particular trial umes to send the samples to laboratories with findings may be resolved soon. Samples from higher volumes and more experience. But some the NCCTG N9831 trial are being retested in a experts were dismayed that the new guidelines round-robin fashion by three different groups. did not recommend FISH over immunohisResults will then be sent to a central monitor- tochemistry. Others, including Norton, are skeptical of all ing group to identify any discrepancies and to try to pinpoint their cause. Soonmyung Paik available tests. The FISH probe, he points out, of the National Surgical Adjuvant Breast and is large (190 kilobases in the case of the Vysis Bowel Project (a National Cancer Institute probe), spanning the gene and then some (Fig. 1). (NCI)-sponsored cooperative based at the “When you see changes in HER2 at the gene University of Pittsburgh) postulates that indi- copy number level, is that a reflection of HER2 viduals with HER2-negative primary tumors itself or of generalized genomic instability?” he may have circulating tumor cells that are HER2 asks. Genomic instability, he points out, is not

118

a random event and the region where HER2 is found is a ‘hot spot’. Norton thinks that the prescribing of HER2-targeting drugs won’t be improved until we understand how specific mutations influence a tumor’s susceptibility and sequencing can be routinely done on biopsies. Comparative genome hybridization (CGH) studies done in his laboratory suggest that simple ‘amplified’ and ‘nonamplified’ readings available from FISH do not adequately reflect the complex changes that can occur in this region. For example, CGH studies revealed that in some samples that were HER2 positive on FISH, the amplified area was actually adjacent to HER2 and not within it4. It may end up that higher resolution methods like CGH are needed to get the right information about a tumor’s status. Meanwhile, companies like Genomic Health in Redwood City, California, and Labcorp’s Monogram Sciences of S. San Francisco, California, are jumping in with new approaches to testing breast cancer patients. Genomic Health claims that an advantage of its quantitative PCR-based test, OncotypeDX, is its accuracy. “Our test is more than 95% concordant with reference labs’ assessment by FISH,” says Steve Shak, chief medical officer at Genomic Health. HER2 is one of 21 genes included in OncotypeDX, which is used to quantify the risk of recurrence of early breast cancer and the response to particular types of chemotherapy. Starting in 2008, the company began including estrogen receptor, progesterone receptor and HER2 status in every report it provides. Monogram’s HERMark test measures HER2 total protein as well as functional homodimers in a dual-antibody format. The company claims the test has advantages over FISH because it is a direct measurement of the protein, and that it is seven to ten times more sensitive than immunohistochemistry testing. Albany Medical College’s Ross counters, “Monogram has no prospective randomized data to support that their test is better.” Neither company’s test is FDA approved for use with Herceptin. For now, the tests’ use may be confined to confirming or clarifying results obtained using other tests. “We need not just technical accuracy but to know if these tests are actually clinically relevant,” says Edith Perez, of the Mayo Clinic Florida in Jacksonville. But Genomic Health seems optimistic. “We have extremely positive feedback on the value of being able to look at that result, especially in those cases when the results for HER2 testing are uncertain,” says Shak. The company is doing additional studies of the test’s ability to predict whether particular individuals will benefit from HER2-targeted therapy.

volume 28 number 2 february 2010 nature biotechnology

© 2010 Nature America, Inc. All rights reserved.

ne w s feat u re Others wonder whether HER2 is even the right thing to test. Strategic Medicine’s Liebman, who was at Vysis (now part of Abbott Laboratories of Abbott Park, Illinois) when the first FISH test was developed, says that for certain patients, the immunohistochemistry and FISH results never agree. “Just because you have a change in gene copy number, that doesn’t mean it’s expressed,” he says. The fact that a significant fraction of HER2-positive patients with metastatic disease fail to respond to the drug also suggests that it is not a causal marker but a surrogate. There may be room for yet more tests. Microarray studies indicate that between 20 to 30 distinct classes of breast cancer exist, according to Charles Perou, of the Lineberger Comprehensive Cancer Center in Chapel Hill, North Carolina. A few of those make up the most clinically relevant subtypes, but there are enough differences between those types that more and better tests are urgently needed. “Our array data show that there are at least two kinds, and maybe many more, of patients with HER2-positive disease,” he says. The difference can be seen in how the patients respond to chemotherapy, with 80% responding in one group and only 30% in the other. Son of Herceptin One other HER2-targeting drug, the smallmolecule Tykerb (lapatinib from London-based GlaxoSmithKline), is now approved for use in breast cancer, but a bevy of next-generation versions of Herceptin and new combinations with the drug are nearing the market, potentially giving oncologists even more choices when deciding which drug to use and when to use it. Genentech currently has two new HER2targeting drugs in phase 3 trials. Pertuzumab is a HER2 dimerization inhibitor that binds to a different epitope on HER2 than Herceptin. The drug inhibits HER2 dimer formation with other HER family members, such as HER3 and HER1. Genentech is currently studying a combination of both HER2 inhibitors in breast cancer. “With this approach, we are addressing the question of what happens when you have more complete HER2/neu blockade,” says Genentech’s Sliwkowski. The drug has already shown promise in early trials. In one study, about a quarter of women whose disease had progressed while they were taking Herceptin had their tumors shrink by >50% when pertuzumab was added to their treatment regimen. The company is also optimistic about T-DM1 (trastuzumab-DM1), a drug conjugate

that combines Waltham, Massachusetts–based ImmunoGen’s antimitotic maytansinederivative DM1 cancer-killing agent with Herceptin. In earlier studies, this drug made tumors shrink even in some women with advanced breast cancer who had been treated with a median of seven different drugs. The drug is being “moved up the line,” according to Sliwkowski, and will be tested in a randomized phase 2 trial comparing TDM-1 versus Herceptin plus chemotherapy. “We think it is so active that it’s important to try this,” says Sliwkowski. To be able to use a targeted therapy without chemotherapy is the dream and it may be very close to realization. Dennis Slamon, a University of California, Los Angeles, oncologist who was part of the team that developed Herceptin back in the 1980s, is enthusiastic about combining it with Genentech’s vascular endothelial growth factor (VEGF) inhibitor, Avastin (bevacizumab). Slamon points out that the two pathways are linked. When HER2 is amplified, “one of the pathways that consistently goes up is VEGF,” he says. Trials of the combination (Avastin plus Herceptin) have been encouraging. In phase 2 trials, “the two antibodies alone, with no chemo, are giving objective response rates in 54% of women with metastatic disease.” Used at earlier stages, Slamon believes the combination could be even more powerful and he is optimistic that this regimen will eventually be tested in a phase 3 trial without concomitant chemotherapy. Growing understanding of the pathways relating to HER2 are also leading to new drug targets. Phosphatidyl inositol 3-phosphate (PI3) kinase and PTEN (phosphatase and tensin homolog) are two other players that seem connected to HER2. Mutations in PI3 kinase occur in 30% of breast cancers and cause the enzyme to be turned on all the time. “It’s a classic oncogenic activating mutation,” says William Sellers, head of oncology research at Novartis Institutes for Biomedical Research in Cambridge, Massachusetts. PTEN, meanwhile, is a tumor suppressor gene which can act by blocking the activation of PI3 kinase. Mutations in that gene can again lead to overactivation. “In many instances, the amplification of HER2 leads to signaling through this pathway,” Sellers says. This has become the number one pathway of interest in cancer. The most advanced new compounds targeting the pathway are mTOR inhibitors, but these work far downstream, and researchers would like to hit it earlier on. Novartis’ BKM120 is a selective PI3 kinase inhibitor that the company

nature biotechnology volume 28 number 2 february 2010

hopes will do just that. All these targets, including HER2, are found in other cancers, which means they could have wider use. Herceptin is being tested in stomach cancer, for example. If these drugs are used more broadly, new tests will be needed. The question then, Sellers says, will be, “How do we know which therapeutic to use?” Improving outcomes Controversy about testing has dogged Herceptin from the beginning. “Slamon battled for ten years to prove HER2 was important,” recalls Shak, who worked on Herceptin’s development while at Genentech. “He had to fight that hard because so many groups were doing their tests without quality control.” But experts are adamant that testing for HER2 must be improved. “In the adjuvant setting, you are giving a woman absolutely substandard care if you are wrongly denying her the drug,” says Ross. Norton concurs: “We must do better.” Although standardizing immunohistochemistry seems to be the most obvious next step, it may not necessarily be the best way to improve outcomes. “Even when countries with a national health service do everything they can to standardize this test, we still see unacceptable margins of error,” says Liebman. Strategic Medicine is working with The Mayo Clinic and Thompson Reuters, headquartered in New York, to build data models that will reveal which steps are most likely to improve the quality of HER2 testing. Most laboratories are using immunohistochemistry, and the cost of making improvements across the entire community, he points out, would likely be prohibitive. “Our goal is to find the weakest points in the system, whether it is an issue with reimbursement, diagnostic development or education. What should be the priority fix that gets us the biggest impact in improving patient care?” Others think the answer clearly lies in applying some newer technologies. “I think sequencing will give us the final answer, once we have inexpensive-enough techniques,” says Norton. Malorye Allison, Acton, Massachusetts 1. Phillips, K.A. et al. Cancer 115, 5166–5174 (2009). 2. McArthur, H.L. & Hudis, C. Clin. Cancer Res. 15, 6311–6313 (2009). 3. Perez, E.A. et al. J. Clin. Oncol. 25, 512, Suppl. 18S (2007). 4. McArthur, H.L. et al., abstract 1005, presented at European CanCer Organization 15 and 34th European Society for Medical Oncology Multidisciplinary Congress, Berlin, Sept. 20–24, 2009.

119

building a business

Coming to terms David H Oden, Jeffrey A Wolfson & Christina W Marshall Before taking other people’s money to finance your venture, it pays to fully educate yourself about the strings attached.

© 2010 Nature America, Inc. All rights reserved.

Y

ou’ve found an investor who’s willing to make a substantial investment in your biotech company—that’s great news. But after the handshake, the next thing is to negotiate the term sheet outlining the structure of the transaction to ensure a true meeting of the minds. Term sheets should always be used in complex investment transactions—especially those involving venture capital investors or other institutional investors. The term sheet sets forth the key terms of the proposed transaction. A good rule of thumb is that the term sheet should address any provision that could kill the deal. If you skip on drawing up a term sheet, then during the drafting and negotiation of the investment documents there may be no clear record of the parties’ understandings on key issues. In the long run, this will cause confusion and discord, and any subsequent documents will probably take more time and cost more to draft and negotiate because the participating parties may be unwittingly using the definitive documents to negotiate—or renegotiate—key terms. Worse still, well into the process, it may become apparent that you are unable to reach agreement on one or more deal-killer terms and the transaction may collapse (Box 1). In the following article, we guide you through the key steps in drawing up a term sheet. Getting this right is important to ensure you remain in control of your company and receive your share of returns.

David Oden is a partner and Christina Marshall is an associate at Haynes and Boone LLP, Richardson, Texas, USA. Jeff Wolfson is a partner at Haynes and Boone LLP, Washington, DC, USA. e-mail: [email protected], [email protected] and [email protected]

120

Box 1 Potential deal killers During negotiations with an investor, you can encounter several hitches. These issues kill more deals than the U.S. Securities and Exchange Commission. • Company technology undervalued by investor(s) or overvalued by founder(s). • Valuation too dependent on issuance of meaningful patent protection. • Partner(s) in joint development arrangements insist on absolute control of patent rights. • Licensing exclusivity in which the partner or licensee in market is not incentivized to commercialize. • Investor(s) or partner(s) insist on control of bet-the-company litigation. • Founder(s) will lose too much control of the company. • Deal requires clinical milestones that are realistically unreachable. •F  uture company flexibility is too limited, particularly in partnering and/or development deals. • Overly cumbersome approval process by investor(s) or partner(s) that could hinder rapid market response. • Liability for clinical trials or indemnification in partnering or joint development deals.

 efore the money B Although some lucky companies are approached by numerous venture capital funds, many have only one investor at a time. The availability and interest of venture capital often depends on the boom and bust cycles of the biotech industry and the economy as a whole (Box 2). At times, companies have been lucky to locate a single interested investor, whereas at other times they have had to fend off multiple investors or limit investment. If your transaction is with only one investor, it may be a bit simpler, faster and less expensive, though not by much. The downside of having only one investor is that there will be fewer pockets to reach into for the next financing. And, if the sole investor declines to participate in the next round, you will be in the position of starting from scratch to attract new ones. If your transaction includes multiple investors, more money and expertise may be available to you. Additionally, there is a much greater

likelihood that at least one investor familiar with the company and its technology will be able to participate in subsequent financings. In this case, the investors will generally select one to be the ‘lead’—the party primarily in charge of due diligence, negotiations and preparation of the definitive investment agreements. During due diligence, the lead investor may examine multiple aspects of your company, including the technical expertise of the founders and key scientific employees, the market conditions and competition, the patent and trademark/branding positions of your company and clearance over any third-party intellectual property (IP) in the space, the R&D pipeline and future patent protection, the status and estimated cost of upcoming clinical trials, the status of US Food and Drug Administration (Rockville, Maryland) interaction and approval, the in-license and out-license agreements the company holds, the agreements with employees and consultants such as contract research organizations, and many other issues.

volume 28 number 2 FEBRUARY 2010 nature biotechnology

© 2010 Nature America, Inc. All rights reserved.

b u ildi n g a b u s i n e s s That’s a lot to handle, so to ensure a smooth diligence process with the lead investor, you should have your legal counsel (preferably independent from your regular IP counsel) pre-evaluate the portfolio and IP-related agreements to help identify and remedy any potential roadblock issues (like ownership of technology) before seeking investment. The lead investor is usually the one investing the most money, and that group should be the main contact for you and your counsel. In this situation, a term sheet is absolutely essential, and all investors should participate in the drafting and negotiation of the term sheet. When all are comfortable with the terms, the other investors should step back and let the lead investor negotiate the rest of the documents based on the term sheet. The next step is dealing with the type of security your investors will be purchasing in return for their financing: common stock, preferred stock, a promissory note (normally convertible into equity) or some combination of these (Box 3). But perhaps the most important thing for you is the valuation the investors assign to your company. Before consummation of the deal, the investor and your firm will go through a very detailed evaluation to determine what portion of the company’s total equity the investor will purchase. This valuation will involve looking at the company and its prospects, the values for comparable companies, the current investment climate and the general economic conditions. Valuation is a combination of art and science and therefore is open to substantial disagreement and negotiation. It can be particularly dependent on the results of a thorough due diligence investigation. It’s important because the total value will determine what percentage of the company the investor will purchase in exchange for the investment. To use a simple example, if the investor is investing $1 million in a company with a pre-money valuation of $1 million, then the investor will own 50% of the company after the investment (assuming that the company will have a value of $2 million post-money). If the company has a pre-money value of $4 million, then the investor will own 20% of the company post-money ($1 million being 20% of a $5 million post-money value). This determination of value is a key area of conflict between founders and investors. Not surprisingly, founders usually want a higher valuation and investors typically seek a lower one. Living with investors Most founders are familiar with vesting—the concept that stock options will become exercis-

Box 2 Term sheet trends During the past boom for biotech companies (about 9–10 years ago), companies could not have asked for more advantageous term sheets. At that time, investors were more fearful of missing out on a great opportunity than of losing their investment. But the pendulum has inevitably swung back to reflect market conditions, so today term sheets tend to be very investor friendly. Biopharma venture capital funding has substantially decreased since the boom days, squeezed by conditions in the financial markets and, more recently, burned by the global economic downturn. With less biotech venture funding available, companies have had to give up more. Unless your company is an unusually attractive investment opportunity, do not expect much negotiating power at the term sheet stage.

able (that is, they will ‘vest’) over time. Vesting is also typical in a venture capital investment, but in a different way: the founder will typically be asked to put his or her equity ownership at risk of being repurchased by the company in the event that the founder is no longer associated with the company for any reason. The rationale behind vesting is that the venture investor is really betting on people (you and your team) as well as the company and the technology. If you leave, retire, decide to go in a different direction or get fired, then you’ll no longer be in a position to push the company forward. And if you still own a substantial portion of the company, this is untenable for your investors. For protection, an investor will typically ask you, the founder, to enter a vesting agreement, whereby all your stock is subject to repurchase by the company at a nominal price per share (typically, the price originally paid by the founder). The company’s right to repurchase the stock will be triggered if the founder leaves the company for any reason, including the termination of employment. This right of repurchase generally decreases over time, so that at some point none of your stock is subject to repurchase. For example, in a five-year vesting (which is fairly typical), the company will have the right (but not the obligation) to repurchase 100% of the founder’s stock for the first year after the investment, 80% in year two, 60% in year three and so on. After five years, none of the founder’s stock will be subject to repurchase. As a founder, your risk is the concern over being ousted by investors, perhaps to bring on a more business-savvy CEO. This often occurs even if you’re performing well as chief executive. Many founders will seek provisions guaranteeing their position for a sufficiently long time, ensuring immediate vesting of rights or other protective measures like specifying a reasonable repurchase price for their stock if involuntarily or unexpectedly separated from the company. Also up for discussion is the amount of control investors will have over the daily operations and major decisions of the company. Specifically, particular attention in negotiations should be

nature biotechnology volume 28 number 2 FEBRUARY 2010

paid to whether the investor gains a seat on the company’s board, the power the investor has on the board and the voting rights the investor may have as a stockholder. It is fairly normal for an investor to obtain one or more seats on the company’s board of directors if the investment is a substantial amount of money and especially if the investor or a designee has expertise that will be helpful to the founders. The rationale here is that the investor wants the right to help control the company (and, in turn, try to protect his or her investment) and you want professional assistance in running the company. Venture investors specialize in running and growing companies—most founders do not. A venture firm’s presence on the board can really help those companies that need assistance with business aspects. When properly arranged, this can provide founders with a renewed opportunity to focus on what may be their core competency—the technology or science. Still, the issue remains of how many board seats the investor is entitled to and the total size of the board. It would be common and expected that a large investor would be entitled to at least one board seat but uncommon to give the investor enough seats to control the board. Investors normally require an agreement with the company and the other stockholders regarding the investors’ rights as a stockholder. These voting agreements usually contain provisions permitting the investor to designate board members and prohibiting the company from taking certain actions without the investor’s approval. Remember to heavily negotiate these aspects at the term sheet stage of the transaction as they will restrict your ability to run the company as you see fit. Exit strategies Because an investor’s primary goal is to obtain a substantial return on his or her initial investment, the term sheet will include multiple provisions focused on how the investor will get the money back—the ‘exit strategy’. These rights may include a liquidation preference,

121

© 2010 Nature America, Inc. All rights reserved.

b u ildi n g a b u s i n e s s redemption of the securities purchased by the investor and registration rights. The type of security (Box 3) that the investor will purchase is directly related to its exit strategy. For example, investors may use a promissory note to try to protect their investment in the event that the company is sold or dissolved by having a ‘liquidation preference’ (liquidation includes being sold). Essentially, the liquidation preference says that if the company is sold or dissolved for whatever reason, the investor’s investment (or a multiple thereof) is paid back in full before any funds are paid to other stockholders. This should be of special concern to you because it represents an amount of money that will be paid out before you, as founder, get one dime of the proceeds. You should try to negotiate the most narrow liquidation preference possible to maximize the amount of money that will go to you and other stockholders. Tensions may arise only upon liquidation because the liquidation preference can often reveal diverging views between an investor, who might have little incentive to seek additional revenue for the founders at exit, and the founders, who would like to finally share in a payday after years of underappreciated efforts. In the case of a strictly failed biotech company (not taken public or acquired, for example), the investors will typically take any available cash to recover their lost investment when assets are sold off to the highest bidder. The most valuable assets are often the patent rights and in-licensed rights, and they can be accompanied by trade secret information, such as clinical data from patient trials or even a Food and Drug Administration drug approval, as well as real estate, furniture and the like. Here you could often receive little or nothing due to the liquidation preference, but it may be possible to negotiate around the liquidation preference and obtain a share of any cash proceeds raised by asset liquidation. Preferred stock that is ‘redeemable’ means that the stock must be repurchased by the company upon the happening of a specified event, such as the passage of time, an insufficient level of cash, a failed drug trial, poor clinical study results, criminal accusations over patient consent or merely at the option of the investor. The company will normally have to purchase the stock back at the investor’s purchase price plus any accrued but unpaid dividends. Redemption is a feature of preferred stock that is generally demanded by investors in the current market. Registration rights provide an investor with the power to register the shares of stock he or

122

Box 3 Defining stock All types of stock are not equal. The main types of stock that you will encounter fall into three categories: Common stock. This is the normal type of stock that all companies issue, and the rights of common stockholders are set forth in the corporation laws of the company’s state of formation. Common stock is usually owned by the founders. Preferred stock. This is usually demanded by most professional investors. Preferred stock is created by amending the company’s certificate of incorporation to include the type and amount of preferred stock issuable and the rights and privileges of the preferred stockholders. Preferred stock normally has preference over common stock when issuing dividends and distributing assets upon the liquidation or sale of the company. The terms of the preferred stock are typically heavily negotiated and should be discussed in detail in the term sheet to ensure the parties agree on this fundamental point. Promissory note. This can take the place of stock and is usually convertible to common or preferred stock upon the occurrence of a certain event (for example, meeting one or more commercial milestones like successful phase 1, 2 or 3 trials), the passage of time or at the option of the investor. The terms of the promissory note are also heavily negotiated and should be addressed in the term sheet. The investor may prefer a promissory note because in the event of liquidation, noteholders typically recover their investment before any stockholders, even preferred stockholders. Convertible promissory note deals are common in very early stage investing or in so-called ‘bridge’ financings (short-term loans made in anticipation of subsequent equity financings).

she owns during the company’s initial public offering (IPO) or after the company has completed its IPO. Registered stock is freely transferable. Even so, it should be noted that although agreements regarding registration rights are enforceable, the underwriter may restrict or eliminate such rights at the time of an IPO depending on both the respective registration rights of other investors and the market conditions. Conclusions Regardless of whether your transaction involves an investment, an asset purchase, a joint development project or a more complex structure, it is crucial for the parties to enter a term sheet—it will substantially increase the chances of successfully closing a deal. Also, having a written agreement that outlines the terms of the transaction will minimize the potential for confusion, costly negotiation and disagreement between the parties during the drafting and negotiation of the investment documents. Depending on your need for capital and the relative attractiveness of your company to investors, the terms of a financing transaction may or may not be negotiable. If you do not immediately need funds and the investor

finds your firm attractive, you will have more leverage negotiating financing terms than if you face an immediate cash crisis. Either way, you should pay particular attention to a few key terms of the investment. Specifically, try to negotiate advantageous positions regarding the percentage of equity the investor will purchase in the transaction, the amount of control the investor will have over the company’s daily operations and major decisions and the amount of money the investor will receive upon the sale or liquidation of the company. These terms will directly affect the control you and the other founders have over the company post-investment, as well as your share of the investment returns. Money can be hard to find right now, but according to a survey conducted by the US National Venture Capital Association (Washington, DC) in December 2008, (National Venture Capital Association, 2009 Venture Capital Predictions Survey Results, Dec. 17, 2008), the biotech and life science sectors are viewed as the second most promising areas for increasing venture investment. If that’s correct, close scrutiny of term sheets in biotech ventures is going to become even more important than before.

To discuss the contents of this article, join the Bioentrepreneur forum on Nature Network:

http://network.nature.com/groups/bioentrepreneur/forum/topics

volume 28 number 2 FEBRUARY 2010 nature biotechnology

correspondence

© 2010 Nature America, Inc. All rights reserved.

Fab-arm exchange To the Editor: In a recent Letter, Labrijn et al.1 reported that therapeutic wild-type IgG4s engage in Fab-arm exchange with endogenous human IgG4 in vivo. The work presented embellishes a theme that was revived by van der Neut Kolfschoten et al.2, who confirmed in an excellent paper previous hypotheses and findings3,4, that is, that IgG4s are dynamic molecules that exchange Fab arms by swapping a heavy chain and attached light chain (half molecule) with a heavylight chain pair from another molecule, resulting in bispecific antibodies. Whereas van der Neut Kolfschoten et al.2 suggest that future studies should address the contribution of IgG4 Fab-arm exchange to in vivo activity of therapeutic monoclonal IgG4 antibodies, Labrijn et al.1 demonstrate that Fab-arm exchange between natalizumab (Tysabri) and endogenous human IgG4 can indeed be observed in blood samples from natalizumab-treated individuals. It is unfortunate that Labrijn et al.1 do not address the intriguing suggestion raised by van der Neut Kolfschoten et al.2, and no data are presented on how the phenomenon of Fab-arm exchange may affect the therapeutic activity of therapeutic wild-type IgG4. Instead, Labrijn et al.1 repeat what was already postulated by van der Neut Kolfschoten et al.2; that is, that Fab-arm exchange could have biological consequences in that the binding to the cognate antigen could change in time from an avidity to an affinity interaction, thereby possibly decreasing binding strength and changing homologous cross-linking to non-cross-linking behavior. Indeed, this is a significant take-home message and it contributes to the overall knowledge of an important subclass of therapeutic antibodies. The awareness of Fab-arm exchange and the subsequent undesired introduction of unpredictability for human immunotherapy,

as created by van der Neut Kolfschoten et al.2 and confirmed by Labrijn et al.1, is of substantial value for antibody development companies who choose IgG4 as the preferred subclass for their products. We therefore agree with the conclusion that mutations that prevent Fab-arm exchange in vivo should be considered when designing therapeutic IgG4. In a Perspectives article that accompanied the original Fab-arm exchange paper by van der Neut Kolfschoten et al.2 in Science, Burton and Wilson5 rightly drew attention to the significance of the work. They concluded by stating, “[In instances where wildtype IgG4 molecules have been used in clinical trials] the possibility that Fab arm exchange could contribute to adverse effects in IgG4 therapy[6,7] should be explored immediately,” thereby referring to papers presenting the cytokine storm precipitated by TGN1412 (ref. 6) or the occurrence of progressive multifocal leukoencephalopathy (PML) in natalizumab-treated patients7. This concluding statement introduced an alleged link between wild-type IgG4 and adverse events. Labrijn et al.1 refer to exactly this statement in the last part of their letter where they recite that the potential exchange with preexisting IgG4 with undesired specificity raises the possibility that Fab-arm exchange could have contributed to some of the adverse events reported for wild-type IgG4. It is regrettable that, in contrast to Burton and Wilson’s5 suggestion, Labrijn et al.1 do not further explore the possibility that Fabarm exchange could have contributed to either the occurrences of PML or cytokine storm experimentally. Instead, Labrijn et al.1 briefly summarize the current theory of reduced immune surveillance, which plausibly explains the occurrence of PML in natalizumab-treated patients, but go on to counter that JC virus (JCV)–very late activation antigen 4 (VLA4; or alpha4

nature biotechnology volume 28 number 2 february 2010

beta1 chain integrin) bispecific antibodies mediate transport of JCV into the central nervous system (CNS) to cause PML. Not only have they failed to provide supporting experimental data but they offer no substantiating evidence from the literature. The mechanism by which JCV-VLA4 bispecifics might mediate transfer of virus to the CNS is presumably based on the following assumptions. First, sufficient amounts of anti-JCV IgG4 should be present in the circulation to form JCV-VLA4 bispecifics. Labrijn et al.1 do not report data on potential detection of JCV-VLA4 bispecific antibodies in natalizumab-treated patients, nor do they predict what the chances of formation of these would be. In a recent study, Egli et al.8 measured the prevalence of JCV infection and replication in 400 healthy donors. They report the IgG seroprevalence for JCV to be 58%. However, it is not further specified what proportion of the JCV IgG was IgG4. But assuming that IgG4 anti-JCV will be present in a proportion of healthy donors, JCV-VLA4 bispecifics are only expected to exist transiently, given the dynamic nature of Fab-arm exchange2. Second, if JCV-VLA4 bispecifics were “mediating the capture of JCV,” as postulated by Labrijn et al.1, free viral particles would need to be in the circulation or in tissues. Egli et al.8 report that JCV DNA could not be detected in any of the 400 blood samples from healthy donors. In addition, Iacobaeus et al.9 analyzed the cerebrospinal fluid, cerebrospinal fluid cells and blood from 217 patients with multiple sclerosis (MS) and 212 controls for detection of JCV DNA. They reported a low copy number of JCV DNA in only four samples (two MS and two controls), none in the other 425. These four individuals had no sign or symptom of PML nor did they develop the disease during follow-up. The combined publications by Egli et al.8 and Iacobaeus et al.9 thus demonstrate that free virus particles could not be detected in plasma or cerebrospinal fluid samples of the majority of healthy donors and MS patients. This is in line with previous studies that report residence of JCV in the kidney in an asymptomatic

123

correspondence

© 2010 Nature America, Inc. All rights reserved.

Table 1 IgG4 monoclonals that have been in trials or on the market Company (location)

Antibody (brand name)

Stabilized hingea Type

Target

Stage

Wyeth (Madison, NJ, USA) UCB (Brussels)

Gemtuzumab (Mylotarg)

YES

Humanized

CD33

Market

Biogen-Idec (Cambridge, MA, USA), Elan (Dublin)

Natalizumab (Tysabri)

NO

Humanized

VLA-4 (CD49d)

Market

Biogen-Idec, UCB

CDP 571 (Humicade)

No

Humanized

Tumor necrosis factor alpha (TNFalpha)

Discontinued after phase 3

AstraZeneca (London) with Medimmune-Cambridge Antibody Technology (CAT) (Gaithersburg, MD, USA)

CAT-152/lerdelimumab

N/A

Human

Transforming growth factor (TGF)-beta 2

Discontinued after phase 3

PDL Biopharma (Redwood City, CA, USA) with Biogen-Idec

Volociximab

N/A

Chimeric

α5β1 integrin

Phase 2

Bristol Myers Squibb (Princeton, NJ, USA)

BMS-663513

N/A

Human

CD137 (4-1BB)

Phase 2

Genzyme (Cambridge, MA, USA) with AstraZeneca (Medimmune-CAT)

GC-1008

N/A

Human

TGF-beta (1,2,3)

Phase 2 Phase 2

Tanox (Houston) with Biogen-Idec

TNX-355/ibalizumab

No

Humanized

CD4

AstraZeneca (Medimmune-CAT)

CAT-354

No

Human

Interleukin 13

Phase 2

iCo Therapeutics (Vancouver, BC, Canada) with Medimmune-CAT

iCo-008/CAT 213)/bertilimumab

No

Human

Eotaxin

Phase 2

Altor Bioscience (Palm Beach, FL, USA) with Tanox

ALT-836/TNX-832

N/A

Chimeric

Tissue factor

Phase 2b

Biogen-Idec

IDEC-151/clenoliximab

YES

Chimeric

CD4

Discontinued after phase 2

Genzyme, AstraZeneca (Medimmune-CAT)

CAT-192/metelimumab

No

Human

TGF-beta 1

Discontinued after phase 2

Biotest (Dreieich, Germany)

BT-062

N/A

Humanized

Syndecan-1 (CD138)

Phase 1

Innate Pharma (Marseille, France)

IPH 2101

Yes

Human

Natural killer inhibitory receptor

Phase 1

Human Genome Science (Rockville, MD, USA)/ Kirin (Tokyo)

HGS-TR2J

No

Human

TNF-related apoptosisinducing ligand receptor 2 (TRAIL-2R)

Phase 1

Human Genome Science

HGS004

No

Human

CC-motif chemokine receptor 5 (CCR5)

Phase 1

PanGenetics

PG102

NO

Humanized

CD40

Phase 1

PanGenetics

ch5D12

NO

Chimeric

CD40

Phase 1

Johnson & Johnson (New Brunswick, NJ, USA)

hOKT3γ4

N/A

Humanized

CD3

Discontinued after phase 1

GPC Biotech (Martinsreid, Germany)

1D09C3

No

Human

Human leukocyte antigen (HLA)-DR

Discontinued after phase 1

TeGenero Immuno Therapeutics (now closed; Wurzburg, Germany)

TGN1412

NO

Humanized

CD28

Discontinued after phase 1

aUppercase

data (YES/NO) confirmed; sentence case data (Yes/No) based on patent/literature information; N/A, not available.

state and the tropism of the virus for (pre-)B cells and CD34+ hematopoietic progenitor cells (reviewed in ref. 10). So in the event that JCV-VLA4 bispecifics were to be circulating in natalizumab-treated patients, the chance that these bispecifics would indeed mediate capture of free virus particles would be vanishingly low. Lastly, Labrijn et al.1 postulate that JCV, captured by JCV-VLA4 bispecifics, is transported into the CNS by infiltrating activated (VLA4+) leukocytes. This last step would require an active infiltration of leukocytes into the CNS of natalizumab-treated patients. Although the influx of leukocytes into areas of disease activity is a pathological hallmark of MS, it is exactly this feature that is inhibited by natalizumab. Numerous studies, including both animal and human data, have demonstrated that antibodies

124

against α4-integrins effectively prevent the accumulation of leukocytes in the CNS. More specifically, it has been demonstrated that compared with patients with MS not treated with natalizumab, cerebrospinal fluid from natalizumab-treated patients has significantly fewer white blood cells, CD4+ T cells, CD8+ T cells, CD19+ B cells and CD138+ plasma cells. These levels remain low, even 6 months after cessation of natalizumab (reviewed in ref. 11). On the basis of the above evidence, we conclude the following: first, JCV-VLA4 bispecifics in natalizumab-treated patients have not been demonstrated and, if they were to exist, would be transient; second, the chance for JCV-VLA4 bispecifics to capture free virus particles is infinitesimal; and third, any facilitation of transport of JCV by activated leukocytes is prevented by an

overall natalizumab-mediated inhibition of leukocyte entry into the CNS. We wish to emphasize that it is wellknown that PML tends to arise in chronically immunosuppressed patients. CD4+ and CD8+ T lymphopenia resulting from HIV infection, chemotherapy or immunosuppressive therapy are the primary risk factors. In addition to the cases reported for natalizumab treatment, PML has been reported to occur in patients treated with IgG1-based biologicals, including rituximab (Rituxan), efalizumab (Raptiva) and alemtuzumab (Campath)10,12,13. Furthermore, we would like to point out that Labrijn et al.1 fail to explain how Fabarm exchange might have caused cytokine storm in TGN1412-treated patients. However, the authors do prominently mention that the kinetics of these adverse events are

volume 28 number 2 february 2010 nature biotechnology

© 2010 Nature America, Inc. All rights reserved.

correspondence compatible with the first detection of Fabarm exchange in their study. We wish to stress that none of the many papers published that studied the mechanism for TGN1412associated cytokine storm consider Fab-arm exchange. In contrast, almost all point to specific CD28 target biology (e.g., see refs. 14,15). Although we acknowledge that a Letter can have room for some speculation, the repeated insinuated link between Fab-arm exchange and adverse events to natalizumab and TGN1412 is misleading. We believe that Fab-arm exchange poses no generic safety issue but recognize that a major disadvantage will be the unpredictability for human immunotherapy due to the dynamics of the transient existence of specific Fab-arm combinations and reduced ability to crosslink the originally targeted antigen. The absence of any generic safety issue is supported by the number of independent clinical studies performed with monoclonal IgG4 to date (Table 1). We fear that, through repetition, the notion that Fab-arm switching causes adverse events will become accepted, which will undermine the efforts of biopharmaceutical companies developing wild-type IgG4 candidate drugs. This concern is substantiated by the observation that, since the publication of the previously mentioned Perspectives article by Burton and Wilson5, the alleged link between Fab-arm exchange and adverse events to wild-type IgG4 did reach biotechnology stakeholders. One biopharmaceutical company announced the discontinuation of its antibody program mentioning the “General concerns of Fab-arm exchange,” although the drug candidate had not raised any unexpected or unacceptable safety concerns in initial clinical testing (Supplementary Note). On top of that, the Distillery section of the newsletter SciBX16, which each week summarizes the most essential scientific findings in techniques of commercial interest, summarizes Labrijn et al.1 by reciting “Fabarm exchange occurs when the arm fragment of a therapeutic antibody exchanges with the arm fragment of an endogenous plasma antibody. The result is a bispecific antibody that reduces the binding affinity of the therapeutic antibody and potentially leads to side effects such as progressive multifocal leukoencephalopathy (PML).” A later issue of SciBX17 amplified this summary. This demonstrates that reinforcement of the unfounded assertion has already begun. In conclusion, we are not aware of any evidence or theory to support a generic safety issue for Fab-arm switching and would like to

alert the antibody community on misleading ‘arm-waving’. Note: Supplementary information is available on the Nature Biotechnology website. COMPETING INTERESTS STATEMENT The authors declare competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/ naturebiotechnology/.

Ellen Broug1, Philip A Bland-Ward2, John Powell2 & Kevin S Johnson2 1PanGenetics BV, Utrecht, The Netherlands.

2PanGenetics BV, Royston, United Kingdom. email: [email protected]

1. Labrijn, A.F. et al. Nat. Biotechnol. 27, 767–771 (2009). 2. van der Neut Kolfschoten, M. et al. Science 317, 1554–1557 (2007). 3. Schuurman, J. et al. Immunology 97, 693–698 (1999). 4. Aalberse, R.C. & Schuurman, J. Immunology 105, 9–19 (2002). 5. Burton, D.R. & Wilson, I.A. Science 317, 1507–1508 (2007). 6. Suntharalingam, G. et al. N. Engl. J. Med. 355, 1018– 1028 (2006). 7. Kleinschmidt-DeMasters, B.K. & Tyler, K.L. N. Engl. J. Med. 353, 369–374 (2005). 8. Egli, A. et al. J. Infect. Dis. 199, 837–846 (2009). 9. Iacobaeus, E. et al. Mult. Scler. 15, 28–35 (2009). 10. Carson, K.R. et al. Lancet Oncol. 10, 816–824 (2009). 11. Stüve, O. et al. J. Neurol. 255 suppl.6, 58–65 (2008). 12. Carson, K.R. et al. Blood 113, 4834–4840 (2009). 13. Waggoner, J., Martinu, T. & Palmer, S.M. et al. J. Heart Lung Transplant. 28, 395–398 (2009). 14. Schraven, B. & Kalinke, U. Immunity 28, 591–595 (2008). 15. Gogishvili, T. et al. PLoS ONE 4, e4643 (2009). 16. Anonymous. SciBX 2, doi:10.1038/scibx.2009.1195. 17. Edelson, S. SciBX 2, doi:10.1038/scibx.2009.1231.

Labrijn et al. reply: Our papers test the hypothesis of IgG4 Fabarm exchange1,2. In a series of in vitro and in vivo experiments, we provide convincing experimental evidence for the occurrence of Fab-arm exchange and show that this mechanism represents an intrinsic activity of IgG4 antibodies that affects treatment with IgG4-based therapeutics in patients. Broug et al. criticize our approach and suggest that we are reiterating previous insights. Surely, however, Broug et al. understand that we are employing the timehonored approach of hypothesis-driven research. Thus, a hypothesis should not be accepted until rigorous testing shows it to be true, which then leads to novel hypotheses. IgG4 Fab-arm exchange thereby has only recently become scientific fact1,2. The hypothesis that states that Fab-arm exchange of therapeutically administered and endogenous IgG4 molecules may generate bispecific IgG4 with undesired specificity that induce adverse events3

nature biotechnology volume 28 number 2 february 2010

still requires (clinical) testing. Broug et al. specifically object to a discussion in which we visit the idea that Fab-arm exchange of natalizumab with patient IgG4 might play a role in the pathogenesis of PML2. Examining this notion, we draw the conclusion (albeit not very plausible under normal conditions) that our analysis cannot fully exclude the possibility that cross-linking under certain (pathological) conditions may be sufficient for undesired biological effects. This limitation, however, also applies to the discussion brought forward by Broug et al., despite their comprehensive citing of literature. We fully acknowledge the absence of reported evidence for generic safety issues with IgG4 molecules and do not claim that natalizumab is ‘unsafe’. As our pharmacokinetic modeling already suggests, the frequency of IgG4 with undesired cross-linking ability is probably very low1. Studies with large patient populations will be required to investigate their presence and effect. In the case of natalizumab, the potential safety-risk, furthermore, has already been mitigated by extensive screening for JCV. Does our research undermine efforts of biopharmaceutical companies developing wild-type IgG4 candidate drugs, as argued by Broug et al.? The mechanism of Fab-arm exchange, in our view, makes the development of nonstabilized IgG4 antibody therapeutics highly problematic as it complicates manufacturing, affects pharmacokinetics and makes pharmacodynamics unpredictable. The last of these problems is brought home by observations with the therapeutic IgG4 anti-HLA DR antibody 1D09C3, which was not discontinued because of unsubstantiated concerns, as suggested by Broug et al., but instead was found to become inactive following Fab-arm exchange4. With the demise of 1D09C3, it can be learned from Table 1 in Broug et al. that Pangenetics’ CD40 antibodies ch5D12 and PG102 remain the only two confirmed nonstabilized IgG4 antibodies in clinical development. Biotechnology companies, of course, are responsible for plotting their own strategy. Nevertheless, as discussed by De Rubertis et al., successful companies should combine strong science with a capability to absorb failure, adapt and move on5. It can certainly be argued that the impact of the unpredictable behavior of nonstabilized, wild-type, IgG4 antibodies in terms of manufacturing, consistency,

125

correspondence pharmacology and clinical safety can be minimized by installing appropriate counter measures. The costs and efforts involved as well as the disadvantages in positioning relative to stabilized competitor drugs, however, strongly suggest to us that such development routes should no longer be considered. COMPETING INTERESTS STATEMENT The authors declare competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/ naturebiotechnology/.

Aran F Labrijn, Janine Schuurman, Jan G J van de Winkel & Paul W H I Parren Genmab, Utrecht, The Netherlands. e-mail: [email protected] 1. Labrijn, A.F. et al. Nat. Biotechnol. 27, 767–771 (2009). 2. van der Neut Kolfschoten, M. et al. Science 317, 1554– 1557 (2007). 3. Burton, D.R. & Wilson, I.A. Science 317, 1507–1508 (2007). 4. Hansen, K. et al. Mol. Immunol. 46, 3269–3277 (2009). 5. De Rubertis, F., Fleck, R. & Lanthaler, W. Nat. Biotechnol. 27, 595–597 (2009).

© 2010 Nature America, Inc. All rights reserved.

An economic and technical evaluation of microalgal biofuels To the Editor: In her News Feature “Biotech’s green gold”1, Emily Waltz details the ‘hype’ being propagated around emerging microalgal biofuel technologies, which often exceeds the physical and thermodynamic constraints that ultimately define their economic viability. Our calculations (Supplementary Box 1) counter such excessive claims1,2 and demonstrate that 22 MJ m–2 d–1 solar radiation supports practical yield maxima of ~60 to 100 kl oil ha–1 y–1 (~6,600 to 10,800 gal ac–1 y–1) and an absolute theoretical ceiling of ~94 to 155 kl oil ha–1 y–1, assuming a maximum photosynthetic conversion efficiency of 10% (ref.3) (results summarized in Table 1). To evaluate claims and provide an accurate analysis of the potential of microalgal biofuel systems, we have conducted industrial feasibility studies and sensitivity analyses based on peer-reviewed data and industrial expertise. Given that microalgal biofuel research is still young and its development still in flux, we anticipate that the stringent assessment of the technology’s economic potential presented below will assist R&D investment and policy development in the area going forward. If sustainable and profitable processes can be developed, the potential benefits of these technologies for the common good appear compelling and include the production on nonarable land of biodiesel, methane, butanol, ethanol, aviation

126

fuel and hydrogen, using waste or saline water, as well as CO2 from industrial or atmospheric sources. We have examined industrial feasibility models of microalgal systems to identify the key economic drivers and provide an industrially relevant update on previous economic analyses4,5. Two of our models are described here as ‘base case’ (that is, integrating current technology) and ‘projected case’, which is considered achievable but has not yet been demonstrated at commercial scale. The 30-year internal rate of return (IRR; Fig. 1 and Supplementary Fig. 1) and net present value (NPV; Supplementary Fig. 2) are used as a measure of the profitability of different production scenarios. IRR values of 15% and above are considered to indicate the potential for economic viability. Importantly, all subsidies including carbon credits have been deliberately excluded, as have financial optimization techniques, to provide a substantial financial contingency. The base case is intended to represent an emerging scenario from the industry and involves the following assumptions: (i) production of microalgal biomass using 500 ha of microalgal production systems; (ii) the extraction of oil; (iii) the co-production and extraction of a high value product (HVP; e.g., β-carotene at 0.1% of biomass, $600/kg); and (iv) the sale of the remaining biomass as feedstock (e.g., soymeal or

fishmeal substitute). In contrast, the projected case is intended to represent the microalgal biofuel industry at maturity and no longer incorporates the co-production of HVPs. The base case is essentially a selfsubsidizing, co-production model. Although it produces ~100 times more oil than HVP on a per tonnage basis, the revenue from HVPs is ~10 times that of oil due to their difference in value. In reality, deployment of this co-production approach will require the servicing of a diversity of HVP markets, as HVP markets are small and easily saturable. Consequently, a major consideration is that the technical developments required for the commercialization of individual HVPs can be as challenging as those required for biofuel production. Therefore, the existence of one or more suitable market-ready HVPs represents a central decision point for would-be biofuel producers. Improved microalgal productivity approaching the targets identified in the projected case will reduce the reliance on co-production (Table 1) as the industry matures. All assumptions (variable settings) in this model are detailed in the Supplementary Data and are based on what are considered to be realistically achievable and published peer-reviewed values. The key findings of this model are summarized in Figure 1 (also see Supplementary Figs. 1 and 2) as sensitivity analysis plots in which individual or multiple settings (e.g., biomass productivity and construction costs) are varied to evaluate their effect on the IRR. For example, as construction costs are reduced (Fig. 1a), the IRR increases. The appropriateness of using the IRR as a measure for profitability in this study is demonstrated in the Supplementary Data and Supplementary Figure 2. Detailed figures for NPV are also provided. This model deliberately does not discriminate between open pond and closed bioreactor systems, the pros and cons of which are hotly debated; instead, it compares construction costs versus yield (Fig. 1e,f) as this is the critical factor (that is, low cost/low yield and high cost/high yield reactors can theoretically be equally profitable). We illustrate how key factors affect the IRR of both case studies (Fig. 1). These were (i) capital costs for construction of the ponds/reactors (Fig. 1a) and (ii) the biomass productivity (g m–2 day–1) (Fig. 1b). In the base case, the third key factor was the role of HVPs (Fig. 1c), whereas in the projected case, the corresponding effect was oil price (Fig. 1d), but biomass oil content (see Supplementary Fig. 1d) was

volume 28 number 2 february 2010 nature biotechnology

correspondence reduction and biomass productivity may provide an alternative route to successful stand-alone fuel systems. From a policy and investment perspective, important conclusions can be drawn from our analysis. First, despite exaggerated claims, our economic analyses suggest that the ~400% increase in investment in microalgal biofuels observed during 2007– 2008 (ref. 6; which has continued to increase in 2009) is sensible, given the potential to meet an IRR of 15% and the future potential to achieve higher returns as biotech and process improvements are made (Fig. 1). This raises the question of why economically viable microalgal biofuel production systems have not yet been demonstrated. In our view, this can be explained in two ways: first, existing pilot and demonstration plants (at <5 ha) are well below the size threshold for economic viability, and second, insufficient time has passed for the industry to evolve from recent capital injection (2006–2007) through to large-scale commercial production. Thus, the most appropriate and cost-effective mix of technologies are yet to be successfully integrated and optimized, and even realistic, viable enterprises are still in the commercial development phase. It is important to note that several external factors are likely to increase both the need for, and the viability of, these

Table 1 Practical and theoretical yield maxima for microalgal biomass and oil productiona Photosynthetic conversion efficiency (%)

Biomass energy

Oil

Biomass prod. Biomass yield

Oil yield

Residual biomass

(GJ ha–1 yr–1) (MJ kg–1) (%) (g m–2 d–1) (T ha–1 yr–1) (L ha–1 yr–1) (T ha–1 yr–1)

2.1

1,677

22.98

25

20.0

73

19,837

6.4

5,101

27.95

50

50.0

183

99,390

92

6.5

5,220

22.98

25

62.2

227

61,400

170

6.5

5,220

27.95

50

51.2

187

100,943

93

8.0

6,424

22.98

25

76.7

280

75,570

210

55

8.0

6,424

27.95

50

63.0

230

124,237

115

10.0

8,030

22.98

25

95.9

350

94,462

262

10.0

8,030

27.95

50

78.6

287

155,297

143

Supplementary Box 1 for full details on calculations and assumptions.

of NPV (15%), with the steepest NPV slopes representing the most influential variables within the target range. Several other production models that met a profitability criterion of 15% IRR were also identified, confirming the flexibility of process development under current market conditions (Supplementary Table 1). Reduced reliance on HVP co-production is realistic at increasing productivities and commodity prices. Although companies pursuing such coproduction strategies already exist, this need not be viewed as a linear progression from HVP- to oil-dominated business plans. Sustained initial investment into cost

a

b

Base case

40

Base case

80

Projected case

Projected case 60 IRR (%)

IRR (%)

30 20

0 50

c

100 150 200 250 Construction costs (1,000 $ ha–1)

0 10

300

d

Base case

80

Projected case IRR (%)

60 40 20 0

30 40 50 60 70 80 Biomass yield (g m–2 d–1)

90 100

Base case

25

Projected case

20 15 10

500 1,000 1,500 2,000 2,500 3,000 HVP (0.1% yield) price ($ kg–1)

0 25

f

120 100 80 60 40 20 0

20

5 0

Base case Cost 1,000 $ ha–1 50 150 100 90 80 70 60 50 40 30 20 Biomass yield (g m–2 d–1)

10

IRR (%)

e

40 20

10

IRR (%)

only a minor factor. In each case, HVPs or oil represent the dominant revenue streams at ~60% and ~50%, respectively. The diversification of products (HVPs, biomass and oil) in the base case has the potential to protect against oil price fluctuations and to subsidize oil prices. Production area scales nonlinearly in our analysis (Supplementary Fig. 1a) with sizes <200 ha showing markedly reduced profitability. With small facility size, the profitability of a complete business model is quite low due to the high capital costs for establishment and low revenue stream. IRR rises rapidly with increased sizes but soon stabilizes. This biphasic trend reflects poorly scalable factors, such as harvesting/ processing machinery and the staff complement. This indicates that economies of scale exist but only up to a point (~200 ha in our model). Total annual operating cost was also examined (Supplementary Fig. 1e) and reflects a resilience of the model to operating cost variations in comparison to other critical factors. We conclude that the minor contribution of each operating cost individually (e.g., nutrient, power or labor costs) would not culminate in major effects on IRR. Figure 1e (base case) and Figure 1f (projected case) demonstrate that as technological developments simultaneously improve multiple key variables (e.g., bioreactor costs and net biomass productivity), even modest improvements synergize to generate substantial improvements in economic viability. This suggests considerable economic potential for microalgal biofuel technologies in the longer term. To rank the degree of influence of each of the factors on the base case (Supplementary Fig. 2e) and the projected case (Supplementary Fig. 2f), each factor was varied several-fold (e.g., 25%, 50%, 100%, 200%, 400% of the set value) as a function

IRR (%)

© 2010 Nature America, Inc. All rights reserved.

aSee

Biomass energy

250

60 50 40 30 20 10 0

50

75 100 125 Oil price ($ bbl–1)

150

Projected case Cost 1,000 $ ha–1 50 150 100 90 80 70 60 50 40 30 20 Biomass yield (g m–2 d–1)

10

250

Figure 1 Sensitivity analysis. Using the parameters described in the text, an industrial feasibility study was constructed to model the effect of varying specific interconnected parameters on the internal rate of return (IRR). (a–d) The effect of these variables on the base case and projected case scenarios. (e,f) Potential gains from simultaneous advancement in the key factors of construction cost and biomass productivity.

nature biotechnology volume 28 number 2 february 2010

127

© 2010 Nature America, Inc. All rights reserved.

correspondence systems. Examples include the possibility of a resurgence in oil price from $30/bbl recently (as of December 2008; http:// tonto.eia.doe.gov/dnav/pet/hist/rwtcd. htm) back to and beyond the previous high of $147/bbl (July 2008) in the near future (modeled here as $100 per barrel in both cases), the introduction of more stringent CO2 emissions targets and carbon trading schemes (potentially rising to ~$200 per ton CO2 by 2050; refs. 7,8) and the increased demand for food and fuel by a population rising from 6.8 billion in 2009 to 9.4 billion in 2050 (refs. 9,10). All of these factors appear to be strong drivers for the economic viability of this technology. This analysis suggests that although microalgal biofuel systems remain in an early stage of development, they are now approaching profitability if the co-production systems in the base case, and/or the increased productivities in the projected case can be attained. A recent report by Huntley and Redalje11 estimates that current technology could produce oil for $84/bbl (with no value attributed to the non-oil fraction), with reasonable advancements in technology reducing this cost to $50/bbl or less. This supports our conclusion that co-production is required in the short term and that at increased oil prices (that is, $100 in this model) an IRR of 15% could be obtained. Considerable synergies also exist between microalgae biofuel production and a wide range of other industries, including human and animal food production, veterinary applications, agrochemicals, seed suppliers, biotech, water treatment, coal seam gas, material supplies and engineering, fuel refiners and distributors, bio-polymers, pharmaceutical and cosmetic industries, as well as coal-fired power stations (CO2 capture) and transport industries, such as aviation. Sound opportunities therefore exist for the development of a rapidly expanding sustainable industry base whose productivity is independent of soil fertility and less dependent on water purity. Thus, these technologies can conceivably be scaled to supply a substantial fraction of oil demand without increasing pressure on water resources while potentially contributing to food production. Furthermore, as this study was conservatively modeled on published data, excluding subsidies (which are actually commonly used to develop other renewable energy sectors, for example, photovoltaics) and proprietary technologies, it follows that strategic partnerships and government policy decisions will play a large

128

part in facilitating a coordinated scale-up and deployment of these technologies to contribute to future energy security. Note: Supplementary information is available on the Nature Biotechnology website. ACKNOWLEDGMENTS The authors gratefully acknowledge the support of the Australian Research Council, IMBcom, and the economic advice of Liam Wagner. COMPETING INTERESTS STATEMENT The authors declare competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/naturebiotechnology/.

Evan Stephens1, Ian L Ross1, Zachary King2, Jan H Mussgnug3, Olaf Kruse3, Clemens Posten4, Michael A Borowitzka5 & Ben Hankamer1 1The University of Queensland, Institute for Molecular Bioscience, St. Lucia, Queensland, Australia. 2IMBcom, St. Lucia, Queensland, Australia. 3University of Bielefeld, Department of Biology, AlgaeBioTech Group, Bielefeld, Germany. 4University of Karlsruhe, Institute of Life Science Engineering, Bioprocess Engineering, Karlsruhe,

Germany. 5Murdoch University, School of Biological Sciences and Biotechnology, Algae R&D Center, Murdoch, Western Australia, Australia. e-mail: [email protected] Waltz, E. Nat. Biotechnol. 27, 15–18 (2009). Mascarelli, A. Nature 461, 460–461 (2009). Melis, A. Plant Sci. 177, 272–280 (2009). Weissman, J.C. & Goebel, R.P. Design and Analysis of Pond Systems for the Purpose of Producing Fuels (Solar Energy Research Institute, Golden, Colorado, SERI/STR231–2840, 1987). 5. Benemann, J.R. & Oswald, W.J. Systems and Economic Analysis of Microalgae Ponds for Conversion of CO2 to Biomass. Final Report to the Pittsburgh Energy Technology Center. Grant no. DE-FG22–93PC93204 (1996). 6. Huggett, B. Nat. Biotechnol. 26, 1208–1209 (2008). 7. McFarland, J.R., Reilly, J.M. & Herzog, H.J. Energy Econ. 26, 685–707 (2004). 8. Zhu, Z., Graham, P., Reedman, L. & Lo, T.A. Decis. Econ. Finance 32, 35–48 (2009). 9. Anonymous. World Population Data Sheet (Population Reference Bureau, Washington, DC, 2009). 10. Anonymous. Soziale und Demographische Daten zur Weltbevölkerung (Deutsche Stiftung Weltbevölkerung, Hannover, Germany, 2009). 11. Huntley, M.E. & Redalje, D.G. Mitig. Adapt. Strategies Glob. Change 12, 573–608 (2007).

1. 2. 3. 4.

Ontology engineering To the Editor: Gene Ontology (GO)1 and similar biomedical ontologies are critical tools of today’s genetic research. These ontologies are crafted through a painstaking process of manual editing, and their organization relies on the intuition of human curators. Here we describe a method that uses information theory to automatically organize the structure of GO and optimize the distribution of the information within it. We used this approach to analyze the evolution of GO, and we identified several areas where the information was suboptimally organized. We optimized the structure of GO and used it to analyze 10,117 gene expression signatures. The use of this new version changed the functional interpretations of 97.5% (P < 10–3) of the signatures by, on average, 14.6%. As a result of this analysis, several changes will be introduced in the next releases of GO. We expect that these formal methods will become the standard to engineer biomedical ontologies. Every year, over 400,000 new articles enter the biomedical literature2, creating an unprecedented corpus of knowledge that is impossible to explore with traditional means of literature consultation. This situation motivated the development of biomedical ontologies, structured information

repositories that organize biomedical findings into hierarchical structures and controlled vocabularies. GO is arguably the most successful example of a biomedical ontology. GO is a controlled vocabulary to describe gene and gene product attributes in any organism and includes 26,514 terms organized along three dimensions: molecular function, biological process and cellular component. GO has become even more intensively used with the introduction of high-throughput genomic platforms because of its ability to categorize large amounts of information using a controlled vocabulary to group objects and their relationships1,3,4. Today, GO and other biomedical ontologies are the result of a painstaking, costly and slow process of manual curation that requires reaching a consensus among many experts to implement a change. Furthermore, the topology of GO has become critically important because of the introduction of gene set enrichment methods. These methods have allowed investigators to characterize the results of a high-throughput experiment in terms of coherent, knowledge-defined sets of genes (e.g., pathways, functional classes or chromosomal locations) rather than in terms of anecdotal evidence about

volume 28 number 2 february 2010 nature biotechnology

© 2010 Nature America, Inc. All rights reserved.

correspondence systems. Examples include the possibility of a resurgence in oil price from $30/bbl recently (as of December 2008; http:// tonto.eia.doe.gov/dnav/pet/hist/rwtcd. htm) back to and beyond the previous high of $147/bbl (July 2008) in the near future (modeled here as $100 per barrel in both cases), the introduction of more stringent CO2 emissions targets and carbon trading schemes (potentially rising to ~$200 per ton CO2 by 2050; refs. 7,8) and the increased demand for food and fuel by a population rising from 6.8 billion in 2009 to 9.4 billion in 2050 (refs. 9,10). All of these factors appear to be strong drivers for the economic viability of this technology. This analysis suggests that although microalgal biofuel systems remain in an early stage of development, they are now approaching profitability if the co-production systems in the base case, and/or the increased productivities in the projected case can be attained. A recent report by Huntley and Redalje11 estimates that current technology could produce oil for $84/bbl (with no value attributed to the non-oil fraction), with reasonable advancements in technology reducing this cost to $50/bbl or less. This supports our conclusion that co-production is required in the short term and that at increased oil prices (that is, $100 in this model) an IRR of 15% could be obtained. Considerable synergies also exist between microalgae biofuel production and a wide range of other industries, including human and animal food production, veterinary applications, agrochemicals, seed suppliers, biotech, water treatment, coal seam gas, material supplies and engineering, fuel refiners and distributors, bio-polymers, pharmaceutical and cosmetic industries, as well as coal-fired power stations (CO2 capture) and transport industries, such as aviation. Sound opportunities therefore exist for the development of a rapidly expanding sustainable industry base whose productivity is independent of soil fertility and less dependent on water purity. Thus, these technologies can conceivably be scaled to supply a substantial fraction of oil demand without increasing pressure on water resources while potentially contributing to food production. Furthermore, as this study was conservatively modeled on published data, excluding subsidies (which are actually commonly used to develop other renewable energy sectors, for example, photovoltaics) and proprietary technologies, it follows that strategic partnerships and government policy decisions will play a large

128

part in facilitating a coordinated scale-up and deployment of these technologies to contribute to future energy security. Note: Supplementary information is available on the Nature Biotechnology website. ACKNOWLEDGMENTS The authors gratefully acknowledge the support of the Australian Research Council, IMBcom, and the economic advice of Liam Wagner. COMPETING INTERESTS STATEMENT The authors declare competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/naturebiotechnology/.

Evan Stephens1, Ian L Ross1, Zachary King2, Jan H Mussgnug3, Olaf Kruse3, Clemens Posten4, Michael A Borowitzka5 & Ben Hankamer1 1The University of Queensland, Institute for Molecular Bioscience, St. Lucia, Queensland, Australia. 2IMBcom, St. Lucia, Queensland, Australia. 3University of Bielefeld, Department of Biology, AlgaeBioTech Group, Bielefeld, Germany. 4University of Karlsruhe, Institute of Life Science Engineering, Bioprocess Engineering, Karlsruhe,

Germany. 5Murdoch University, School of Biological Sciences and Biotechnology, Algae R&D Center, Murdoch, Western Australia, Australia. e-mail: [email protected] Waltz, E. Nat. Biotechnol. 27, 15–18 (2009). Mascarelli, A. Nature 461, 460–461 (2009). Melis, A. Plant Sci. 177, 272–280 (2009). Weissman, J.C. & Goebel, R.P. Design and Analysis of Pond Systems for the Purpose of Producing Fuels (Solar Energy Research Institute, Golden, Colorado, SERI/STR231–2840, 1987). 5. Benemann, J.R. & Oswald, W.J. Systems and Economic Analysis of Microalgae Ponds for Conversion of CO2 to Biomass. Final Report to the Pittsburgh Energy Technology Center. Grant no. DE-FG22–93PC93204 (1996). 6. Huggett, B. Nat. Biotechnol. 26, 1208–1209 (2008). 7. McFarland, J.R., Reilly, J.M. & Herzog, H.J. Energy Econ. 26, 685–707 (2004). 8. Zhu, Z., Graham, P., Reedman, L. & Lo, T.A. Decis. Econ. Finance 32, 35–48 (2009). 9. Anonymous. World Population Data Sheet (Population Reference Bureau, Washington, DC, 2009). 10. Anonymous. Soziale und Demographische Daten zur Weltbevölkerung (Deutsche Stiftung Weltbevölkerung, Hannover, Germany, 2009). 11. Huntley, M.E. & Redalje, D.G. Mitig. Adapt. Strategies Glob. Change 12, 573–608 (2007).

1. 2. 3. 4.

Ontology engineering To the Editor: Gene Ontology (GO)1 and similar biomedical ontologies are critical tools of today’s genetic research. These ontologies are crafted through a painstaking process of manual editing, and their organization relies on the intuition of human curators. Here we describe a method that uses information theory to automatically organize the structure of GO and optimize the distribution of the information within it. We used this approach to analyze the evolution of GO, and we identified several areas where the information was suboptimally organized. We optimized the structure of GO and used it to analyze 10,117 gene expression signatures. The use of this new version changed the functional interpretations of 97.5% (P < 10–3) of the signatures by, on average, 14.6%. As a result of this analysis, several changes will be introduced in the next releases of GO. We expect that these formal methods will become the standard to engineer biomedical ontologies. Every year, over 400,000 new articles enter the biomedical literature2, creating an unprecedented corpus of knowledge that is impossible to explore with traditional means of literature consultation. This situation motivated the development of biomedical ontologies, structured information

repositories that organize biomedical findings into hierarchical structures and controlled vocabularies. GO is arguably the most successful example of a biomedical ontology. GO is a controlled vocabulary to describe gene and gene product attributes in any organism and includes 26,514 terms organized along three dimensions: molecular function, biological process and cellular component. GO has become even more intensively used with the introduction of high-throughput genomic platforms because of its ability to categorize large amounts of information using a controlled vocabulary to group objects and their relationships1,3,4. Today, GO and other biomedical ontologies are the result of a painstaking, costly and slow process of manual curation that requires reaching a consensus among many experts to implement a change. Furthermore, the topology of GO has become critically important because of the introduction of gene set enrichment methods. These methods have allowed investigators to characterize the results of a high-throughput experiment in terms of coherent, knowledge-defined sets of genes (e.g., pathways, functional classes or chromosomal locations) rather than in terms of anecdotal evidence about

volume 28 number 2 february 2010 nature biotechnology

correspondence

7

8

9

10

11

12

13

14 Bits Ganglion mother cell fate determination

6

Terpene metabolism

5

Fructose metabolism

4

Heterocycle metabolism

3

Cell proliferation

2

Defense response

1

Macromolecule metabolism

Biological process

0

single genes5,6. GO has become a primary provider of these gene sets and researchers use its graphical structure to identify the specificity of a gene class so that they will compare classes of the same specificty7. Previous studies have found that the structure of GO does not conform to expected intuitions regarding the structure and distributions of ontology terms8,9. Gene enrichment methods typically use the structure of ontologies as a proxy for the specificity of a term10,11 or, in some cases, use automated procedures to identify structural biases and to compensate for them in the analysis7,8,12. Unfortunately, in some cases, even these compensative methods are unable to reach the same conclusions of a well-calibrated ontology (Supplementary Notes 1). The approach we advocate here aims to solve the problem at its root by optimizing the structure of the ontology so that it will indeed be an accurate representation of the informational specificity of any term in the ontology. This approach would not only avoid the necessity to compensate for biases but also improve the semantic transparency of the ontology structure. To do so, we introduce an automated method for engineering the structure of GO based on the information content of each single term. The intuition behind this method is that ontologies are information systems and, as such, they can be optimized using the well-established mathematics of information theory. Given its mathematical nature, this optimization process can be automated, thus producing a principled and scalable architecture to engineer GO and, analogously, other biomedical ontologies. Our approach starts from the quantification of information contained in the terms of the ontology. The information content of a term is computed from the amount of annotation available for it relative to all other terms, and it is a

measure of the surprise caused by labeling a gene with this term rather than with any other term (Supplementary Notes 2). For instance, if a term contains all genes, then it is not surprising for a given gene to be labeled with it, so this term does not contain much information. Thus, the more genes or gene products associated with a term, the less specific the term is and the less information is conveyed by it. This ‘surprise factor’ is called ‘self-information’, and information theory provides a formal definition for it13 (Fig. 1). Using information theory, we analyzed the evolution of the information content of GO across time, examining 2 million genes across all the organisms encoded in the ontology annotations. This process highlighted information biases and inefficiencies that may affect the usage of GO and identified those areas of the ontology that were suboptimally organized. The analysis identified three types of information inefficiencies in the structure of GO.

2.9 × 101, 7.7 × 10−2, 6.1 × 10−1

0.61

3.0 × 101, 7.9 × 10−2, 5.9 × 10−1

0.6

Topological metric

© 2010 Nature America, Inc. All rights reserved.

Figure 1 Spectrum of GO terms: examples ranging from 1 to 14 bits.

The first type of inefficiency arises from the variability of the information content among the terms within a given ontology level. By the principle of maximum entropy, an even a priori distribution of information (where all terms in a level are equally specific and hence equally informative) is most efficient because a random experiment is most informative if the probability distribution over outcomes is uniform13. Furthermore, gene set enrichment methods often use GO level (that is, distance from the top of the graph) as a proxy for degree of specificity7,10,11; this strategy implicitly relies on within-level uniformity of information content. Optimally, then, all the terms in a given level would have equal specificity and, therefore, the same information content. Our analysis revealed that the original version of GO contained a large degree of such intralevel variability of information content. For example, the term ‘pilus retraction’ was originally at level 2, at the same level of terms like ‘cell cycle’ and ‘cell development’ that are actually much more general. The second type of structural inefficiency, inter-level variability, arises from deviations in information content between levels. In general, terms become more specific as the information content of a level increases with depth in the graph. In some areas of GO, however, the mean information content decreases from one level to the next, creating an information bottleneck. In this case, most of the annotation information of the previous level is transmitted to the next through only a few terms. The larger the decrease in information content, the more severe the bottleneck. The presence of these

3.0 × 101, 7.6 × 10−2, 6.0 × 10−1 3.2 × 101, 7.7 × 10−2, 5.9 × 10−1

0.59 0.58

2.9 × 10−1, 7.6 × 102, 5.9 × 10−1

2.5 × 101, 7.6 × 10−2, 5.9 × 10−1

3.0 × 101, 7.6 × 10−2, 5.8 × 10−1

0.57

2.5 × 101, 7.5 × 10−2, 5.7 × 10−1

0.56 0.55

2.5 × 101, 7.4 × 10−2, 5.6 × 10−1

2.0 × 101, 7.6 × 10−2, 5.6 × 10−1

0.55 0.078 0.074 0.072

Inter-level metric

0.07

20

22

24

26

28

30

32

Inter-level metric

Figure 2 Three-dimensional evolution of GO over ten releases from 2005 to 2007 along the three dimensions of structural inefficiency. An ontology with no inefficiency across these metrics would be at the origin (0,0,0).

nature biotechnology volume 28 number 2 february 2010

129

© 2010 Nature America, Inc. All rights reserved.

correspondence areas of suboptimal information distribution violate the assumption of gene set enrichment analysis methods7,12 that the specificity in GO terms effectively increases from one level to the next (Supplementary Notes 3). The third type of structural inefficiency, topological variability, arises from the suboptimal organization of the branches. The principle of maximum entropy dictates that the closer a topological structure is to uniform, the greater is the information that experiments can derive from it8. We used entropy rate to quantify the uniformity of the GO branch structure (Supplementary Notes 4) so that a higher entropy rate indicates that the ontology structure is closer to uniform. We analyzed the evolution of GO along these three dimensions of structural inefficiency using ten releases of GO containing over 2 million unique genes14. Figure 2 plots their structural inefficiencies for each release of GO and illustrates how they have been decreasing over time (Supplementary Notes 5). For instance, with time point 8 (February 1, 2007), interlevel variability and topological variability saw substantive improvements, coinciding with introduction of the [‘is_a complete’] property in GO15. In contrast, intra-level variability saw comparatively modest improvements over the evolution of GO. One of the greatest dangers of structural inefficiencies in GO is the impact they can have on the functional interpretation of the results of high-throughput experiments. We thus optimized the information distribution of GO by introducing single-level changes and modifying 1,001 relationships and 11% of GO terms, thus significantly improving the overall intra-variability (P < 10−3) (Supplementary Notes 6). We used this optimization method to create a modified, improved GO and we compared it to the current GO in the interpretation of 10,117 gene expression signatures from DNA microarray experiments16. Each signature contains genes differentially expressed between two biological conditions, and we compared the results of gene enrichment analysis of these signatures obtained by the original and the modified GO. We found that these changes significantly affected the functional

130

interpretations of 97.5% (P < 10−3) of the experimental gene signatures and altered the resulting set of GO categories by 14.6% on average (Supplementary Notes 7). On the basis of this analysis, we presented 14 recommendations to the GO Consortium and most of these new annotations (12) will be introduced in the next release of GO (Supplementary Notes 8). Finally, as a result of our analysis, we applied this approach to more complicated multi-level structural changes. We suggested the GO Consortium move 12 terms. The terms all underwent the standard curatorial validation of the GO consortium, and 11 of them are now included in the current release of GO. The twelfth term, pigmentation (GO:0043473) had few annotations at the time but was not moved as it was expected that many more genes would be annotated with that term in the future. The most striking result of our experiment was to show the convergence of mathematical optimality and biological validity and that a formal, automated analysis is able to uncover sound biological information hidden in the structure of the ontology. By altering the ontology itself, our approach improves gene enrichment results in ways that cannot be obtained by simply changing the underlying gene enrichment method (Supplementary Notes 1). Our analysis also reveals that GO contains more information than is currently used. By optimizing the distribution of information within GO, our method can be used to aid the design of more efficiently organized knowledge repositories—leading to a more effective use of biological information. This method is already being used to achieve this aim by the GO Consortium and other ontologies, such as the Phenotypic Quality Ontology (PATO)17 in the OBO Foundry18. We expect that formal and automated methods will become the standard for the engineering of biomedical ontologies. Note: Supplementary information is available on the Nature Biotechnology website. ACKNOWLEDGMENTS This work was supported in part by the National Library of Medicine (NLM/NIH) under grants 1K99LM009826 and 5T15LM007092 and by the National Human Genome Research Institute

(NHGRI/NIH) under grants 2P41HG02273, 1R01HG003354, and 1R01HG004836. The authors are grateful to the anonymous reviewers for their helpful suggestions. COMPETING INTERESTS STATEMENT The authors declare no competing financial interests.

Gil Alterovitz1–3, Michael Xiang1,2, David P Hill4, Jane Lomax5, Jonathan Liu6, Michael Cherkassky2, Jonathan Dreyfuss1,2, Chris Mungall7, Midori A Harris5, Mary E Dolan4, Judith A Blake4 & Marco F Ramoni1,2 1Children’s

Hospital Informatics Program, Harvard-MIT Division of Health Sciences and Technology, Harvard Medical School, Boston, Massachusetts, USA. 2Partners Healthcare Center for Personalized Genetic Medicine, Boston, Massachusetts, USA. 3Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA. 4Jackson Laboratory, Bar Harbor, Maine, USA. 5EMBL-EBI, Wellcome Trust Genome Campus, Hinxton, UK. 6Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA. 7Lawrence Berkeley National Laboratory, Berkeley, California, USA. e-mail: [email protected] or [email protected]. 1. Ashburner, M. et al. Nat. Genet. 25, 25–29 (2000). 2. Davis, D.A., Ciurea, I., Flanagan, T.M. & Perrier, L. Med. J. Aust. 180, S68–S71 (2004). 3. Camon, E. et al. Nucleic Acids Res. 32, D262–D266 (2004). 4. Harris, M. et al. Nucleic Acids Res. 32, D258–D261 (2004). 5. Subramanian, A. et al. Proc. Natl. Acad. Sci. USA 102, 15545–15550 (2005). 6. Doniger, S.W. et al. Genome Biol. 4, R7 (2003). 7. Al-Shahrour, F., Diaz-Uriarte, R. & Dopazo, J. Bioinformatics 20, 578–580 (2004). 8. Alterovitz, G., Xiang, M., Mohan, M. & Ramoni, M.F.G.O. Nucleic Acids Res. 35, D322–D327 (2007). 9. Ogren, P.V., Cohen, K.B. & Hunter, L. Pac. Symp. Biocomput. 174–185 (2005). 10. Dennis, G. Jr. et al. Genome Biol. 4, 3 (2003). 11. Zhou, M. & Cui, Y. In Silico Biol. 4, 323–333 (2004). 12. Raychaudhuri, S., Chang, J.T., Sutphin, P.D. & Altman, R.B. Genome Res. 12, 203–214 (2002). 13. MacKay, D.J.C. Information Theory, Inference, And Learning Algorithms, xii (Cambridge University Press, Cambridge, U.K.; New York, 2003). 14. Wu, C.H. et al. Nucleic Acids Res. 34, D187–D191 (2006). 15. The Gene Ontology Consortium. Nucleic Acids Res. 36, D440–D444 (2008). 16. Yi, Y., Li, C., Miller, C. & George, A.L. Jr. Genome Biol. 8, R133 (2007). 17. Gkoutos, G.V. et al. Comp. Funct. Genomics 5, 545– 551 (2004). 18. Smith, B. et al. Nat. Biotechnol. 25, 1251–1255 (2007).

volume 28 number 2 february 2010 nature biotechnology

case study

Never again

c o m m e n ta r y

Christopher Scott In 2008, Roche acquired Genentech, ending the most successful symbiotic business relationship the biotech/ pharma sector has ever seen. Morphing biotech business models, pharma management’s short-termism and dwindling investor patience means we’ll never see the like of it again.

© 2010 Nature America, Inc. All rights reserved.

W

hen Roche’s management acquired a $2.1 billion stake in Genentech in 1990, it eyed not only a deep pipeline of drugs, but also a team of dedicated, independent scientists. At the same time, Roche executives also saw a financial market buzzed about Genentech stock. So the Swiss company bought a majority stake, with an option to buy up the rest later, and left Genentech as a standalone unit. In 1999, the Basel-based pharma exercised that option, taking Genentech off the market for a few brief months before placing it back on the New York Stock Exchange in another public offering, taking advantage of the bullish environment. Roche then reduced its ownership by selling off shares twice more, bringing in nearly $8 billion for the three offerings (each public sale at a higher pershare price) and Roche got to keep the majority of product revenue every year, as a primary stakeholder. By all accounts, this relationship was mutually beneficial. The pharma’s stake in Genentech was essentially a gift that kept on giving. For its part, Genentech retained product marketing rights inside the United States and was able to foster its science in a ‘small company’ culture free from interference by its Swiss partner. But Roche couldn’t resist taking the final bite: in March last year the pharma acquired the remaining 44% stake it didn’t already own in a $46.8 billion deal. Analysts fretted that the entrepreneurial culture would yield to a staid Swiss lockstep. Even Roche chairman Franz Humer, who drove the deal, had previously said that a buyout would “never work, because if we owned all of Genentech we would kill it.” In market share alone, Roche-Genentech is now the seventh largest US pharmaceutical company, projected to generate $17 billion in revenue yearly. In the months since the merger, Roche pulled out of the Washington, DC, lobbying group Pharmaceutical Research and Manufacturers of America, saying it would work instead through the biotech equivalent, the Washington, DC–based Biotechnology Industry Organization. And the June American Society of Clinical Oncology meeting was chock-full of rosy reports from Roche-Genentech’s growing oncology portfolio, with ten new molecular entities in clinical development. In August, the US Food and Drug Administration approved Avastin (bevacizumab) for kidney cancer. New uses for established drugs, such as Tarceva (erlotinib), Lucentis (ranibizumab) and Avastin, are also moving through confirmatory trials. The union produced its share of changes in senior management. Founder Arthur Levinson moved up to the board of directors and the president of product development, Susan Desmond-Hellmann, departed. Sandra Horning, a longtime Stanford academic with no pharma experience, took over as senior vice president of global clinical development. The chief of operations, Richard Scheller, and his top lieutenants stayed. Time will tell whether the new structure can sustain the track record of the independent biotech, but even before the buyout, Genentech’s business had become more reliant on extensions Christopher Scott is Contributing Editor at Nature Biotechnology.

nature biotechnology volume 28 number 2 february 2010

of its existing product franchise (similar to any other big company) rather than production of new therapeutic entities. Few pharma-biotech relationships look anything like the RocheGenentech courtship. Consider Johnson & Johnson (J&J) of New Brunswick, New Jersey, locking up Centocor for $4.9 billion in 1999. Centocor was another biotech with an early blockbuster and good science to back it up. But J&J treated it like any other merger; though it left the biotech as a stand-alone unit, the latter was not public and had no scientific autonomy. In 2008, J&J merged Centocor with Ortho Biotech, slashed 400 jobs at Ortho and deployed the remaining resources to support Centocor’s high-selling anti-tumor necrosis factor alpha compound Remicade (infliximab). Today, J&J partially owes its standing as a leading drug maker to Remicade: in 2008, it had worldwide revenue of $5.3 billion. That’s a big hitter in J&J’s lineup, but it has made Centocor look an awful lot like a one-trick pony. Genentech, it was not. More diversifying was Basel-based Novartis’ purchase of Emeryville, California–based Chiron for $5.1 billion in 2006. Here again, unlike Roche’s hands-off approach with Genentech, the acquisition immediately sent executives from the biotech packing and put scientists and their projects on the bubble. That year, Novartis launched a three-year reorganization plan, and Chiron went inside Novartis’ vaccine and diagnostics division, which last year accounted for just 4% of its $42.6 billion in sales. Though Chiron’s vaccine expertise has been important in developing chicken egg–free flu vaccine production methods, Novartis did not buy Chiron to play parent; it bought it to hoard the valuables. Is there another young Genentech and nurturing pharma parent still out there? Staring across today’s biotech landscape (economic downturn or not), the answer seems to be no. Another Genentech would need not only to have a promising and established pipeline, but also to be surrounded by a free-wheeling scientific culture. That is awfully hard to find these days, in part because biotech companies are no longer built with flexible business plans designed to chase big questions down dark alleys, as Genentech was. Firms are designed for a quick and total buyout and a return on venture money. Any burgeoning Genentech wannabe would be cherry-picked long before its first product launch. What’s more, once consummated, big mergers today tend to isolate resources for the probable winning compounds, leaving the rest of the science to wither on the vine. The priorities for today’s top management are more myopic: how does an acquisition and its compounds affect the balance sheet; science is secondary. One final reason that Genentech-Roche is a once-in-a-lifetime relationship: luck. Though Genentech was a clear biotech leader in 1990, and its vaunted pipeline has long been discussed as one of the industry’s best, the business of making drugs is scored by failure and serendipity alike. It took the financial stability of Roche, the genius of Genentech’s scientists, an inspiring intellectual culture and providence to spin out product after product. But lightning does not strike twice.

131

commentary

Going to ridiculous lengths—European coexistence regulations for GM crops Koreen Ramessar, Teresa Capell, Richard M Twyman & Paul Christou

G

enetically modified (GM) crops now cover over 100 million hectares of arable land in >20 countries, and this trend toward increased uptake and deployment is growing at a steady rate1. Inevitably, GM and non-GM crops of the same species will be grown near each other, a concept defined by the term ‘coexistence’2. There has been an extraordinary and sustained campaign mainly in the European Union (EU; Brussels) that has united certain stakeholders, including organic producers, certification bodies and environmental groups, against GM/non-GM coexistence. The escalating battle has drawn in producers, retailers, governments, regulatory bodies, scientists and, of course, the general public. The outcome in the EU is a mess: a haphazard and inconsistent set of rules that has no rational scientific underpinning, which obstructs GM producers, misleads the public and adds unnecessary layers of complexity to international trade. GM/non-GM coexistence is now a loaded term, used by opponents as a de facto criticism of GM agriculture and a self-fulfilling reason to impose restrictions. Is there any way to encourage a rational approach to the coexistence debate? Koreen Ramessar and Teresa Capell are at the Departament de Producció Vegetal I Ciència Forestal, University of Lleida, Lleida, Spain; Richard M. Twyman is at the Department of Biological Sciences, University of Warwick, Coventry, UK; and Paul Christou is at the Departament de Producció Vegetal I Ciència Forestal, University of Lleida and the Institució Catalana de Recerca i Estudis Avançats, Passeig Lluís Companys, Barcelona, Spain. e-mail: [email protected]

ANDREW SYRED/SCIENCE PHOTO LIBRARY

© 2010 Nature America, Inc. All rights reserved.

Even if a GM crop can surmount Europe’s excessive product registration process, any farmer hoping to plant it must then navigate tortuous, arbitrary and scientifically unjustifiable coexistence regulations.

Special treatment required? Keeping GM corn pollen grains (like this one pictured at a magnification of 795×) segregated from conventional corn is one of the purposes of Europe’s coexistence regulations.

Adventitious presence The basis of the campaign against GM/nonGM coexistence is “adventitious presence,” which is defined (in the context of GM agriculture) as the presence of unwanted GM material in non-GM commodities. The adventitious presence of GM material can occur in many ways (Fig. 1), but most often through outcrossing, the growth of volunteer plants from stray seeds and admixture after harvest3. The adventitious presence of GM material in non-GM commodities is often presented as disastrous by opponents of GM technology and described using terms such as ‘contamination’ and ‘adulteration’. However, it is important to recognize that the reasons it is thus regarded

nature biotechnology volume 28 number 2 february 2010

differ according to different stakeholders. Environmental pressure groups are keen to promote uncertainties about the impact of GM crops on human health and the environment and oppose coexistence on the basis that the adventitious presence of GM material is a safety issue, even though the safety of GM crops must be demonstrated to regulators before licensing for commercial production. Organic producers, on the other hand, oppose coexistence because they fear their organic status and associated organic price premium may depend on the absence of GM material, prompting legal challenges and lobbying against GM agriculture both within the EU and elsewhere2,4. The European Commission (EC; Brussels) has confirmed that coexistence is purely an economic issue by defining it as “…issues relating to the economic consequences of adventitious presence of material from one crop in another and the principle that farmers should be able to cultivate freely the agricultural crops they choose, be it GM crops, conventional or organic crops...”5. Intimately intertwined with the political coexistence debate is the European public’s antipathy to GM products and preference for non-GM products. Public uncertainty about the safety of GM products is exaggerated by environmental pressure groups and some parts of the media, thus helping to create the preference. Many suppliers and retailers, responding to consumer pressure, have therefore imposed restrictions on the use of GM material and its presence in food products, encouraging producers to segregate GM and non-GM crops. A vicious circle has been created.

133

© 2010 Nature America, Inc. All rights reserved.

C O M M E N TA R Y Coexistence practices in non-GM agriculture What often gets forgotten in the heat of the GM/non-GM coexistence debate is that different varieties of the same crop species have coexisted for generations and that adventitious presence is recognized as an inevitable consequence of coexistence that can be minimized but not entirely eliminated. Therefore, almost all traded agricultural commodities anticipate some degree of inadvertent mixing, and thresholds exist that are recognized in laws, regulations and/or voluntary standards. Such thresholds have resulted in the development of a series of measures that are applied during cultivation, harvest, transport and storage to minimize outcrossing, the growth of volunteer plants and inadvertent mixing3. These best practices were established decades ago and have evolved to deliver high purity seed and grain to support the production, distribution and trade of products from different agricultural systems. The principles of these coexistence practices are dependent on context (which crops and where they are grown), consistent, proportionate to need, fair and practical. Examples of successful coexistence practices in non-GM agriculture include production systems for certified seeds (e.g., hybrid seed), organic crops coexisting with conventional crops and commodity crops coexisting with specialty crops (e.g., field corn with sweet corn and/or popcorn, and specialty corns such as high-amylose, high-oil, white, waxy, hard endosperm and nutritionally dense varieties)6. Perhaps one of the best-studied examples of coexistence in conventional agriculture is standard rapeseed varieties and specialty high erucic acid rapeseed (HEAR) varieties for industrial use, particularly because HEAR is regarded as antinutritional and undesirable in food (and therefore constitutes an actual risk rather than a consumer preference, as is the case for GM crops). Contracts for growing HEAR crops require that only certified HEAR seed is used, equipment should be cleaned and segregated and that there should be an isolation distance of between 50 m (e.g., in the UK) and 100 m (e.g., in Germany) from other rapeseed crops. The admixture threshold for HEAR in food rapeseed is 2% although recorded levels are usually much lower. For example, the 100-m separation distance in Germany generally delivers seed lots with HEAR levels <0.2%, and only a few seed lots contain >0.5%. In the UK, coexistence research shows that separation distances as low as 9 m still provide bulk rapeseed harvests containing <0.5% HEAR2.

134

Growing phase Seedbed preparation; start material

a

Sowing

Growing

Harvest

Post-harvest

Storage; processing; transport

Supply chain

Potential avenues for on-farm mixing between gentically modified (GM) and non-GM crops Volunteers from GM pre-cultures; dispersal of GM seeds via straw and/ or manure

Seeds; mixing in machinery during sowing

Crossfertilization; Dispersal of GM seeds via straw and/or manure

Mixing in machinery during harvest

Volunteers

Mixing during on-farm storage and processing; mixing during transport to collection point

b On-farm coexistence measures to ensure the purity of a crop Seed purity

Spatial isolation (isolation distances); field characteristics; pollen barriers; temporal isolation (period of flowering, rotation)

Cleaning of machinery; removal of bolters

Cleaning of machinery; maneuver space for machinery

Control of volunteers; specific tillage operations; application of herbicides and/or weeding

Cleaning of storage and processing rooms; cleaning of transport vehicles

Figure 1 The many routes to adventitious presence of GM material in a conventional crop. (a) Stages where on-farm adventitious mixing between GM and non-GM crops can occur. (b) On-farm co-existence measures to ensure crop purity during production (reproduced from ref. 3 with permission from EDP Sciences).

Coexistence practices in GM agriculture The United States and Canada have embraced GM agriculture, and their agricultural systems support the coexistence of conventional, organic and GM crops under a common set of practical guidelines7. In contrast, GM crops in the EU are treated very differently from other specialty crops. Because of the precautionary approach adopted by EU regulators, they are subjected to more extensive and stringent safety testing than their conventional counterparts, even though the safety of GM crops and products is demonstrated before they are given approval to enter the agricultural production system. Once approved for commercial release and marketing, there should be no grounds for treating the coexistence of conventional and GM crops any differently from, for example, the coexistence of conventional and HEAR varieties of rapeseed, but the prescribed practices set at national and regional levels are much more strictly regulated, with lower adventitious-presence thresholds, larger isolation distances, harsh economic liability provisions on the producer and the proposal of mandatory additional preventive measures3. GM adventitious-presence thresholds in the EU are the strictest in the world8,9. In the United States, Canada and Japan, non-GM products may contain up to 5% GM material before they must be labeled as GM. Other countries have lower tolerance thresholds (e.g., 1% in Australia, New Zealand, South Africa, Brazil and China). The EU has a twotier tolerance policy (Regulation (EC) no.

1830/2003; ref. 10), with a 0.9% limit applied to approved products and a zero tolerance threshold applied to unapproved products, replacing the temporary 0.5% second-tier limit previously approved by the European Food Safety Authority (Parma, Italy). Additionally, Recommendation 2003/556/EC (ref. 5) provides guidelines for the development of national coexistence strategies and best practices that, where necessary, can be applied to prevent non-GM products exceeding the labeling threshold, which means coexistence is officially a matter of ‘national competence’ where each member state is responsible for the establishment of a legislative framework on a crop-by-crop basis. Some EU member states (Austria, Belgium, Bulgaria, Czech Republic, Denmark, Germany, Hungary, Latvia, Luxemburg, Portugal, Romania and the Slovak Republic) have already started adopting regulations governing the planting and handling of GM crops, whereas others are still in the process of developing their regulations. The lack of overarching regulation means that the crops covered in each member state’s regulations, and the minimum isolation distances that are imposed, vary greatly (Supplementary Table 1). The adoption dynamics of GM crops in Europe differ among and even within member states, as discussed below. At the current time, the only GM crop cultivated in the EU is Bacillus thuringiensis (Bt) corn expressing the insecticidal protein Cry1Ab, and this accounts for <2% of the total EU corn output, compared with 75% in the United States11.

volume 28 number 2 february 2010 nature biotechnology

© 2010 Nature America, Inc. All rights reserved.

C O M M E N TA R Y Spain is arguably the most enthusiastic adopter of GM agriculture in the EU, allowing the cultivation of GM crops without a complete regulation regime. The establishment of coexistence rules has been prevented by disputes between the Spanish Ministry of Agriculture (influenced by farmers’ lobbies) and the Ministry of Environment (influenced by ecological lobbies). Coexistence is currently determined by seed company guidelines together with some specific regulations12, but there are no compulsory training courses, no specific liability rules and 50-m isolation distances are standard13. Despite successful coexistence in Spain, market forces have created region-by-region segregation. In the productive agricultural regions of Catalonia and Aragon, 55% and 42% of corn, respectively, is GM14. In contrast, Asturias and the Basque Country have declared themselves GM free with the support of regional governments and some farmers’ associations. Several EU member states require farmers to gain official approval before they are allowed to plant GM crops. In Austria, farmers need approval from local authorities for each field and crop (similar procedures are being considered in Hungary, Ireland and the Slovak Republic). Austria has the strictest regime, and even though there are some coexistence measures (zero-risk seed purity regulation, compulsory training courses and strict liability policies), the Austrian authorities are against GM crops and strive to avoid coexistence instead of promoting it15. Austrian provinces have approached the EU to establish GM-free regions, but in September 2007 the European Court of Justice finally rejected general statutory regional bans on GM crops, arguing that a statutory ban is a denial of the freedom of choice for farmers and consumers16. Poland and Belgium are also seeking to avoid the deployment of GM crops (120 communities in Belgium have already declared themselves GM free). Portugal has a complete system of regulation (established before commercial planting) with compulsory training courses, strict anti-cross-pollination measures and a public compensation fund. Even so, this still allows some flexibility in isolation measures depending on voluntary agreements among neighbors. This kind of collective initiative avoids complicated anti-cross-pollination measures and expensive double farm facilities. How far is far enough? EU coexistence guidelines (Recommendation 2003/556/EC) state that “…Management

measures for coexistence should reflect the best available scientific evidence on the probability and sources of admixture between GM and non-GM crops…”5, but it is quite clear that this recommendation is being ignored in many EU member states. Some countries require vast isolation distances that bear no relationship to the underpinning scientific evidence. For example, Luxemburg requires 800 m between GM and non-GM corn and 3 km between GM and non-GM rapeseed. Latvia requires 4 km between GM and conventional non-GM rapeseed and 6 km if the non-GM rapeseed is organic (Supplementary Table 1). Such isolation distances impose immense costs on GM farmers because they have to negotiate with a much larger number of neighboring farms and, in practical terms, simply remove their choice in relation to adopting GM crops17. The minimum isolation distances imposed on GM producers in the EU should be those that are sufficient to maintain the adventitious presence of GM material below 0.9% in neighboring organic and conventional plots. The current isolation distances were based on the assessment of biological and physical processes that affect outcrossing3,4,18,19, and these tend to differ between studies if factors such as pollen viability; male sterility; flowering synchrony; wind speed and direction; weather conditions; field size and shape; and distance, topography and vegetation between the pollen donor and recipient fields are not standardized. Maize pollen is released in very large quantities, between 4.5 and 25 million pollen grains per plant over a typical 5- to 8-day period20, but is larger (90–125 µm) and heavier than the pollen of most other wind-pollinated plants, and therefore dispersal is limited to about 10% of the range covered by other species, often settling within a few hundred meters of its source. A research study conducted by the Spanish Institute for Agriculture & Food Research and Technology (Madrid) demonstrated that in field trials, the average presence of the Bt gene in conventional maize separated from Bt maize by just 2–10 m is <0.9%21. In another study, the maximum distance over which any cross-pollination between GM and non-GM maize occurred was 200 m (a single kernel event), with further events also observed at 150 m and 100 m (that is, a total of three pollen grains from a 4,000-m2 plot of GM maize22). Monitoring of gene flow between adjacent GM and non-GM maize fields by the Portuguese Ministry of Agriculture between 2006 and 2009 has shown that in 80% of cases, the mean level of cross-pollination was <0.3%, with the highest level being 0.7%23.

nature biotechnology volume 28 number 2 february 2010

A meta-analysis of existing cross-fertilization studies4 concluded that an isolation distance of just 50 m between GM and non-GM seed production fields would be sufficient to maintain cross-fertilization levels below 0.5% at the border of the recipient corn field, showing that the 800-m separation required in such countries as Bulgaria, Hungary and Luxembourg is completely unjustified on scientific grounds. In 2001, the EC Scientific Committee on Plants (Brussels) proposed tolerance thresholds of 0.3% for crosspollinating crops and 0.5% for self-pollinating and vegetatively propagated crops24. Toward successful coexistence in Europe GM crops could enhance European agriculture through higher productivity, better use of resources and reduced environmental impact, but the acceptance of GM technology by European consumers remains low. The principle of coexistence predates the deployment of GM crops and has been used successfully for many years to preserve the purity of seed stocks and specialty conventional crops, allowing different varieties of sexually compatible, outcrossing species to be grown in close proximity. Isolation is the primary method to reduce adventitious presence, and isolation distances for different varieties of conventional crops have been defined according to scientific investigations of gene flow. This scientific process appears to have been discarded by the EU and its member governments in the case of GM agriculture. Not only are the thresholds for adventitious presence far stricter than for conventional crops, but the isolation distances implemented to achieve such thresholds are arbitrary, excessive and appear to be politically motivated rather than to reflect scientific reality. Insisting on such inflated and capricious isolation distances places economic and regulatory pressure on farmers, who face stern punitive measures if outcrossing (or ‘contamination’) occurs with nearby conventional crops. The US and Canadian systems place more emphasis on the balance between GM and non-GM crops, the isolation distances are based on scientific principles and both GM and non-GM farmers have a stake in preventing adventitious presence. These practices have enabled the successful coexistence of GM and non-GM (including organic) crops outside Europe for many years without government involvement, and there should be no rational objection to the adoption and standardization of such practices throughout the EU.

135

C O M M E N TA R Y Note: Supplementary information is available on the Nature Biotechnology website. ACKNOWLEDGMENTS Research in our laboratory is funded by MICINN (Ministry of Science and Innovation), Spain (BFU2007-61413); European Union Framework 6 Program – The Pharma-Planta Integrated Project LSH2002-1.2.5-2; European Union Framework 7 ProgramSmartCell Integrated Project 222716; European Union Framework 7 European Research Council IDEAS Advanced Grant (to P.C.) Program-BIOFORCE; Acciones Complementarias (MICINN) BIO200524826-E. Centre CONSOLIDER on Agrigenomics funded by MICINN, Spain.

© 2010 Nature America, Inc. All rights reserved.

COMPETING INTERESTS STATEMENT The authors declare no competing financial interests. 1. James, C. International Service for the Acquisition of Agri-biotech Applications Brief No 37 (ISAAA, Ithaca, NY, USA, 2008). 2. Brookes, G. & Barfoot, P. Co-existence in North American Agriculture: Can GM Crops Be Grown with Conventional and Organic Crops? (PG Economics Ltd., Dorchester, UK; 2004). 3. Devos, Y. et al. Agron. Sustain. Dev. 29, 11–30 (2009). 4. Sanvido, O. et al. Transgenic Res. 17, 317–335 (2008).

136

5. European Commission. Off. J. Eur. Commun. L189, 36–47 (2003). 6. Brookes, G. Coexistence of GM and Non-GM Crops. Current Experience and Key Principles (PG Economics Ltd., Dorchester, UK; 2004). 7. Kalaitzandonakes, N. in Proceedings to the Second International Conference on Co-existence Between GM and non-GM-based Agricultural Supply Chains, Montpellier, France, November 14–15, 2005 (ed. Messéan, A.) 29–30 (Agropolis Productions, 2005). 8. Ramessar, K., Capell, T., Twyman, R.M., Quemada, H. & Christou, P. Nat. Biotechnol. 26, 975–978 (2008). 9. Ramessar, K., Capell, T., Twyman, R.M., Quemada, H. & Christou, P. Mol. Breed. 23, 99–112 (2009). 10. European Commission. Off. J. Evr. Commun. L268, 24–28 (2003). 11. Abbott, A. & Schiermeier, Q. Nature 450, 928–929 (2007). 12. Asociación Profesional de Empresas Productoras de Semillas Selectas (APROSE). Guía 2007 de Buenas Prácticas para el Cultivo de maíz Bt. (Asociación Profesional de Empresas Productoras de Semillas Selectas, Madrid, Spain; 2007). 13. Corti-Varela, J. in Proceedings to the 2nd Annual Cambridge Conference on Regulation, Inspection and Improvement, Cambridge, UK, September 11–12, 2007. 14. Binimelis, R. J. Agric. Environ. Ethics 21, 437–457 (2008). 15. Beckmann, V., Soregaroli, C. & Wesseler, J. Am. J. Agric. Econ. 88, 1193–1199 (2006).

16. EuropaBio. EU Court rejects Austrian biotech ban—supports right to choose biotech crops (EuropaBio, Brussels) (14 September 2007). 17. Messean, A., Angevin, F., Gomez-Barbero, M., Menard, K. & Rodriguez Cerezo, E. in Technical Report Series of the Joint Research Center of the European Commission (Institute for Prospective Technological Studies, Sevilla, Spain; 2006). 18. Aylor, D.E. Agric. For. Meteorol. 123, 125–133 (2004). 19. Devos, Y., Reheul, D. & de Schrijver, A. Environ. Biosafety Res. 4, 71–87 (2005). 20. Paterniani, E. & Stort, A.C. Euphytica 23, 129–134 (1974). 21. Brookes, G. et al. GM Maize: Pollen Movement and Crop Co-existence. (PG Economics Ltd., Dorchester, UK; 2004). 22. Luna, V.S. et al. Crop Sci. 41, 1551–1557 (2001). 23. Cruz de Carvalho, P. (coordinator). Coexistência Entre Culturas Geneticamente Modificadas e Outros Modos de Produção Agrícola. Relatórios de Acompanhamento. (Ministério da Agricultura, do Desenvolvimento Rural e das Pescas/Direcção-Geral de Agricultura e Desenvolvimento Rural, Lisbon, 2009). 24. Scientific Committee on Plants (SCP). Opinion of the Scientific Committee on Plants Concerning the Adventitious Presence of GM Seeds in Conventional Seeds (European Commission, Brussels, Belgium; 2001).

volume 28 number 2 february 2010 nature biotechnology

p at e n t s

Changing the rules of the game: addressing the conflict between free access to scientific discovery and intellectual property rights Miriam Bentwich A provisional patented paper application procedure could promote earlier disclosure of novel scientific knowledge.

© 2010 Nature America, Inc. All rights reserved.

A

major constraint on the advancement of scientific research in biotech is the growing barriers to free exchange of scientific knowledge associated with the increased number of patents linked to intellectual property rights (IPR). This matter, which can be tagged as the ‘secrecy threat’, has been addressed in numerous publications1–7, and can be described as essentially revolving around two issues. One issue concerns the possible reluctance of patent applicants and owners to fully disclose their acquired novel scientific knowledge through early and easily accessible scientific publications. The other issue was broadly termed the “tragedy of the anticommons” by Eisenberg and Heller and has been further developed by other scholars8–10. It involves the possible constraining of other scientists from using patented novel knowledge for further scientific research for economic reasons. Indeed, numerous empirical studies have shown that scientists delay scientific publication of research they expect to patent and are reluctant to freely share research results and materials with other scientists11–15. Other studies and judicial decisions have shown that, contrary to its intended purpose6,16,17, the patent system does not necessarily offer a practical incentive for patent applicants to disclose much information regarding their inventions in a comprehendible manner18–21. In fact, the notion of prior art, so crucial in determining the eligibility of a patent application with respect to its novelty, at best, significantly limits the grace period in

Miriam Bentwich is in the Department of Political Science, The Hebrew University, Mt. Scopus, Jerusalem, Israel. e-mail: [email protected]

which potential patentees may publicly disclose information regarding their invention before a corresponding patent application is submitted22–24. Finally, additional case-inpoint research illuminates instances such as that of Myriad Genetics’ policy regarding its patented breast cancer genes and their mutations, where patentees exploited their IPR to increase their profits at the direct expense of sharing novel scientific knowledge10,25–27. Admittedly, there have already been extensive and contrasting attempts to address the challenge at hand3,28–30, the bulk of which revolves around practical steps that may be taken to mitigate the secrecy threat and its negative consequences for scientific research without deserting the patent system20,31–33. However, these practical steps appear to be flawed with respect to the two main concerns associated with the secrecy threat. Thus, suggested practical measures against the tragedy of the anticommons, such as patent pools and governmental compulsory licensing for use of patents, lack a theoretical justification from an IPR advocate’s viewpoint. Meanwhile, other suggested practical steps, like extending the grace period for ‘self prior art’ or using the US provisional patent application (PPA) option more extensively, could allow potential patentees to publish their scientific knowledge at an earlier stage and in a fuller manner. Yet these practical steps have so far neither obligated nor stimulated potential patentees to do so in practice. We present here an alternative and, we hope, more successful approach to tackle the secrecy threat without undermining the patent system. This is achieved by a revised mandatory version of PPAs, designed for patents in biotech and administered by leading

nature biotechnology volume 28 number 2 february 2010

scientific journals, which would be officially recognized by the US Patent and Trademark Office (USPTO) and subsequently by other patent offices as well. From optional PPAs to mandatory PPPAs Since 1995, the USPTO has offered the option to submit provisional patent applications34 that are easier and substantially cheaper to file than a full-fledged patent application35–37. Submitting a PPA is easier because, unlike regular patent applications, PPAs do not require the inclusion of the patent’s claims, which define the extent of the protection sought in a patent application in detailed technical terms. Additionally, PPAs do not necessitate discussion of relevant prior art or use of applicable legal terms. Instead, PPAs need only adequately describe the invention and its scope38,39. As a result, this type of patent application lacks fundamental ingredients of any full-fledged patent application, and therefore, patents cannot evolve directly from PPAs. However, PPAs allow potential patentees to secure an earlier valid ‘priority date’ (that is, official filing date) for their future patented invention before the latter is fully prepared, thereby decreasing the chances of possible competition over their inventions36. Such an advantage, though, is granted only if two conditions are met. First, the nonprovisional patent application must be submitted no later than 12 months past the filing date of the PPA. Second, the provisional application must adequately provide a written description of the full scope of the invention explicated in the later full application36,39. Additionally, when these two conditions are met, and a corresponding non-US patent application is submitted within 12 months

137

pat e n ts How the PPPA procedure works Step 1: Creating required components for PPPA through online paper submission system

Optional data files

Submitted manuscript

Generate DTS certification(s)

Charge applicable funds

Send author DTS certification files

Step 2: Full patent application submission

Full patent application

Proof of article publication

Submitted manuscript & optional data files

DTS certification files

© 2010 Nature America, Inc. All rights reserved.

Step 3: PPPA validity determination by USPTO

Condition 1: FPA due no later than 12–18 months past DTS certification date

Condition 2: Submitted manuscript and optional data files sufficiently disclose the invention described in FPA

Figure 1 A graphical outline demonstrating how the PPPA procedure can be practically employed. PPPA, provisional patented paper application; DTS, digital time-stamping service; FPA, full patent application; USPTO, US Patent and Trademark Office.

after the filing date of the US PPA, in principle, the filing date of the PPA may also serve as the priority date for the non-US patent application. Consequently, US PPAs can be used in conjunction with other patent offices (e.g., the European Patent Office and Japan Patent Office), thereby potentially expanding the grant of IPR beyond US soil. As already mentioned, although PPAs theoretically enable their submitters to share their novel knowledge with other researchers, they do not obligate patent applicants to do so. In fact, because they are not full patent applications, PPAs constitute an anti-incentive for sharing novel scientific knowledge outside the patent system. For instance, if a patent applicant is unable to submit the full patent application within the allowed oneyear period, the applicant will lose the ability to patent the invention, at least insofar as it is explicated in the early publication as the latter now constitutes patent-blocking prior art22. The importance of the PPA procedure is that it opens the door to a contingent acknowledgment of underdeveloped and informally submitted patent applications.

138

Hence, PPAs are informally submitted in the sense that, for themselves, they are neither published nor examined by the USPTO; and they are underdeveloped (by design) because they purposely accept patent applications lacking fundamental ingredients like claims and the discussion of prior art, as long as these applications contain a sufficient description of the invention and its scope. Consequently, in principle, a scientific paper that includes a description of an invention’s purpose and the method(s) used to attain it may qualify as a PPA. What’s more, insofar as the content and submission date of such a paper can be verified, the paper need not be physically submitted to USPTO before the filing of the full patent application, because other than recording their existence, USPTO does not execute any action with regard to PPAs. From this perspective, therefore, scientific publications would not constitute an external prior art, undermining the eligibility of patent applications. Instead, such publications, if recognized as provisional patented paper applications (PPPAs), would provide the very same protection that traditional PPAs grant a future intended patent, including their

supposed compatibility with the regulations of non-US patent offices. At the same time, contrary to the existing model of PPAs, the suggested PPPA alternative version will obligate patent applicants to share their novel knowledge with colleagues through scientific publications (Fig. 1). Thus, given the proper tools for date and content verification, PPPAs may be entirely administered by peer-reviewed academic journals instead of the USPTO, thereby further binding such provisional patent applications to scientific publications, while significantly lowering the fee charged for these applications. Moreover, because the scientific publication constitutes either the very PPPA or at least amounts to a significant integral portion of it (as will be explained later), there is a direct incentive for potential patentees to fully disclose the information concerning their inventions in the corresponding scientific publications. Otherwise, they risk their eligibility to use the PPPA as the basis for the full patent application, thereby undermining their ability to be granted with the desired patent and its conferred IPR. Similarly, potential patentees would be interested in disclosing that information as early as possible because any delay in the submission of PPPAs naturally increases the risk of potential competition over their inventions. Meanwhile, as the suggested PPPA procedure is inseparable from scientific publications, the potential patentee might very well choose not to use this procedure. Therefore, we suggest that the PPPA be defined as an obligatory step on the way to filing a full patent application in biotech. This being the case, it might also be advisable to extend the deadline for filing the full patent application for, say, another six months, so that potential patentees would have a larger safety net, covering instances where the paper is not immediately accepted. Moreover, paradoxically, the supposed reluctance of potential patentees to use PPPAs, should this be left to their choice, appears to be grounded on a sort of prisoner’s dilemma, where the assumed least risky choice is conflated with and mistakenly preferred over the best outcome choice, due to lack of cooperation40,41. Hence, lacking knowledge regarding the decisions of other potential patentees concerning the use of PPPAs and their entailed extra risks, a potential patentee would not want to take these extra risks. For, by taking such extra risks, she might put her invention in a more vulnerable position than competitors who choose not to use PPPAs. Yet, theoretically, if all potential patentees agreed (that is, cooperated among

volume 28 number 2 february 2010 nature biotechnology

pat e n ts themselves) to use PPPAs before their full patent applications were submitted, such an agreement would result in a much better outcome for all involved parties. By reaching this agreement, all PPPA submitters would be granted the following benefits: 1. Elimination of the extra risks associated with PPPAs, as the latter will entail now merely standard risks shared by all potential patentees. 2. Preliminary professional assessments of the invention’s validity through the journal’s peer-review process, thereby indicating whether the invention is worth investing in.

© 2010 Nature America, Inc. All rights reserved.

3. An earlier ‘certificate of quality’, should the scientific paper describing the invention be accepted, thereby potentially increasing the ability to raise more funds. 4. Possible advice for prepatenting improvements to the invention, leading to a potential increase in the patent’s value. Mandatory PPPAs are, therefore, justifiable, even from the viewpoint of potential patentees, while being an acceptable form of provisional patent application (PPA). At the same time, PPPAs address the first portion of the secrecy threat, regarding the possible reluctance of potential patentees to publicly share new scientific knowledge in a full and easily accessible manner. From theory to practice Assuming the employment of mandatory PPPAs is logically plausible, even from a potential patentee’s standpoint, it is still necessary to explain whether and how they can be applied in practice. Essentially, enabling valid and effective PPPAs necessitates that scientific publications provide papers’ submitters with trustworthy certifications of the date and content of the relevant information that is needed to constitute such PPPAs. That way, though PPPAs are not administered by a patent office, but rather by academic journals, PPPAs would be amply reliable for use by both potential patentees and the USPTO. Accordingly, to supply such trustworthy certifications, academic journals must comply with the following three requirements: 1. Provide authors, upon their request, with the necessary certification of the paper submitted for publication that describes the invention for which a patent may be sought.

2. Grant authors, upon their request, a separate certification for additional optional and unpublished data that they wish to include as part of the PPPA. Such certified but unpublished data contain materials that are not suitable for a scientific paper, but might be technically important for securing the full patent application (e.g., certain claims or additional figures). 3. Ensure that the granted certifications can be independently verified by the USPTO in a simple manner, once full patent applications are submitted. That is, allow the USPTO to (i) check the authenticity of these certifications and (ii) ensure that, insofar as the content of the certified manuscript is concerned, it is adequately covered by the finalized published paper. All of the aforementioned requirements can be realized by a combination of two existing and widely accessible technologies. One technology is the digital time-stamping service, issuing a trusted third-party time-stamp that associates a date and time with a digital document in a cryptographically strong way. The digital time-stamp can be used at a later date (e.g., by USPTO) to independently verify that an electronic document with a particular content existed at the time stated on its timestamp, thereby certifying the document’s content and creation date42,43. In this respect, it is important to note two points. First, a digital time-stamp, understood as a specific type of digital signature44, is legally supported and recognized by applicable legislations in many countries worldwide, and specifically by the United States45. Second, commercial digital time-stamping services46 are already available on the Internet with costs as low as 40 cents per use, making the financial barriers negligible to end-users. The other technology is the widely employed online paper submission processing software (e.g., Editorial Manager, Manuscript Central) already used, according to a recent study, by 76% of the peer-reviewed academic journals, and still growing47. The significance of these web-based computer programs is twofold. First, they can provide the option of date and content certification within the paper submission process, through seamlessly embedding the already available digital time-stamping toolkits that were designed to be set-in within a third-party software environment. Second, these programs include a bundle of features like financial transaction and automatic e-mail communication with the submitters of papers48. As such, these features provide further necessary support

nature biotechnology volume 28 number 2 february 2010

for employing PPPAs, like the ability to send authors their needed certified documents and their finalized articles’ proofs, as well as collecting the applicable fees for USPTO and the digital time-stamping service. PPPAs limit the effects of the tragedy of the anticommons Once used, PPPAs can also provide an adequate theoretical justification, from a patents’ advocate viewpoint, for significantly limiting the negative effects of the tragedy of the anticommons. This theoretical rationalization is based on patentees’ interest in securing the main benefit that is gained by their patents, namely IPR, as patents effectively grant patentees exclusive property rights over their inventions, even if for a limited period of time. The original notion of exclusive property rights and its justification is usually attributed to John Locke, the forefather of classical liberalism, though he simply used the term ‘property rights’49. In fact, Locke’s framing of property rights is still influential among contemporary libertarians and neoliberals, who usually perceive themselves as his successors, and therefore Locke’s thought is still relevant for contemporary discussions of exclusive property rights50–52. According to Locke, the initial justification for private exclusive property rights stems from an individual’s “mixing of his labor” in a previously unowned object. Thus, an individual, naturally, has exclusive ownership rights over his own labor, namely, it is his property. Consequently, by mixing his labor, it becomes inextricably part of the object, thereby rendering anyone else who would then use that object as effectively infringing the property that the first person has ‘invested’ in the object through his labor. By annexing her labor to the object, therefore, the individual restricts others’ right to use it, making it exclusively her own49,53. Acquiring a previously unowned object, then, necessitates that the new owner be the sole investor of labor in this object (be it directly from his own work, or indirectly through transfer of other people’s labor to his disposal). Inversely put, so long as other people have invested their direct or indirect labor in the object at hand as well, without being compensated for their loss of proportionate investment, no one can claim exclusive property rights over the object. From this perspective, therefore, a patent, pertaining per se to a new invention, involves an ownership of a previously unowned object, and may be granted to the patent applicant only if at the time the patent was granted, the applicant (and his party) could be

139

© 2010 Nature America, Inc. All rights reserved.

pat e n ts considered as the direct or indirect sole laborer(s) over the invention. However, by using PPPAs, potential patentees enjoy the labor that the anonymous referees have invested or mixed in the knowledge underlying the inventions intended to be patented. Similar to Locke’s example of a person’s mixing his labor in a natural and unowned land, thereby cultivating it and rendering it as more profitable, the anonymous referees’ work, as already indicated above, also assists in increasing the value of the knowledge at the basis of the intended patent. Consequently, in principle, a patent applicant has to ‘purchase’ the referees’ labor in enhancing the knowledge on which the invention is based, so that the applicant would be the only party to have invested her labor in the invention she is seeking to patent. Yet because the referees are anonymous, the patent applicant cannot transfer the payment to a particular person or group of people. Moreover, the referees themselves are able to perform their job (that is, labor) based on the current relevant scientific knowledge, acquired through the previous labor of other scientists, who have gained their knowledge on the grounds of their predecessors, and so on. In other words, only through compensating the scientific community as a whole can the potential patentee buy the labor, invested by other parties that are affiliated with this community in the theoretical knowledge, underlying the invention for which a patent is sought. This compensation would be made by unequivocally guaranteeing the community’s members an inexpensive license to use the theoretical knowledge behind the invention to which the community has contributed. Consequently, such a guarantee would be applicable, for example, to theoretical knowledge concerning particular mapped genes and their mutations (e.g., BRCA1 and BRCA2), but not to specific devices and services that further use this knowledge, as in the case of a breast cancer diagnostic kit. Still, it was precisely with regard to the type of patent associated with BRCA1 and BRCA2 and its potential damage to the free flow of novel scientific knowledge that the concern about the tragedy of the anticommons was expressed in the first place. Meanwhile, notice that the argument advanced here is not only theoretically plausible, but also practically allowable. Thus, the requirement to furnish other scientists with cheap licenses to use the theoretical

140

knowledge underlying a patented invention decreases the patentee’s income from such licensing. However, it also decreases the same patentee’s expenditures on getting similar licenses to other patentees’ theoretical knowledge. Moreover, the justified compensation scope is merely limited to theoretical knowledge used only within the conduct of research. Therefore, the pledge to inexpensive licensing for research endeavors neither covers using services and devices that use this theoretical knowledge, nor permits selling them without the explicit agreement of the patentee, thereby preserving the overall integrity of patentees’ IPR and their conferred privileges. Conclusions We have presented a novel and relatively easy solution to the secrecy threat entailed in patents and their undesired effects on the free access to scientific knowledge and to the advancement of biotechnological research. By suggesting a mandatory PPPA procedure, largely based on the already available and internationally recognized PPA procedure, we have demonstrated how the two main concerns associated with the secrecy threat can be addressed. We have shown how PPPAs may promote earlier and fuller disclosures of novel scientific knowledge and provide an ample justification for requiring patentees to grant inexpensive licenses for use of their inventions by other parties conducting research, while protecting the integrity of patents and IPR. COMPETING INTERESTS STATEMENT The author declares no competing financial interests. ACKNOWLEDGMENTS I am grateful to Zvi Bentwich and Isaac (Zahon) Bentwich for their invaluable comments on a previous version of this article. I also thank Moshe Maor and Avner de-Shalit for preliminary discussions regarding the article’s theme. Finally, I would like to express my gratitude to the Lady Davis Trust at the Hebrew University for its generous financial support that enabled the pursuit of this study. 1. Andrews, L.B. Nat. Rev. Genet. 3, 803–808 (2002). 2. Caulfield, T., Cook-Deegan, R.M., Kieff, F.S. & Walsh, J.P. Nat. Biotechnol. 24, 1091–1094 (2006). 3. Cook-Deegan, R., Chandrasekharan, S. & Angrist, M. Nature 458, 405–406 (2009). 4. Cukier, K.N. Nat. Biotechnol. 24, 249–251 (2006). 5. David, P.A. J. Inst. Theor. Econ. 160, 9–34 (2004). 6. Eisenberg, R.S. Yale LJ 97, 177–231 [225] (1987). 7. Holman, C.M. Trends Biotechnol. 25, 539–543 (2007). 8. Heller, M.A. & Eisenberg, R.S. Science 280, 698–701 (1998). 9. Shapiro, C. in Innovation Policy and the Economy, vol. 1 (eds. Jaffe, A.B., Lerner, J. & Stern, C.) 577–579 (MIT

Press, Cambridge, Massachusetts, 2001). 10. Merz, J.F., Kriss, A.G., Leonard, D.G.B. & Cho, M.K. Nature 415, 577–579 (2002). 11. Blumenthal, D., Campbell, E.G., Anderson, M.S., Causino, N. & Louis, K.S. J. Am. Med. Assoc. 277, 1224–1228 (1997). 12. Schissel, A., Merz, J.F. & Cho, M.K. Nature 402, 118 (1999). 13. Campbell, E.G. et al. J. Am. Med. Assoc. 287, 473– 480 (2002). 14. Grushcow, J.M. J. Leg. Stud. 33, 59–84 (2004). 15. Murray, F. & Stern, S. J. Econ. Behav. Organ. 63, 648–687 (2007). 16. 35 US Code § 112, 113. 17. Walsh, J.P., Cho, C. & Cohen, W.M. Science 309, 2002–2003 (2005). 18. Brenner v. Manson 383 US (1996), 519, 534. 19. Vitronics Corp v. Conceptronic Inc. 90 F3d (Fed. Cir. 1996), 1576, 1583. 20. Fromer, J.C. Iowa Law Rev. 94, 539–606 (2009). 21. Long, P. Univ. Chic. Law Rev. 69, 625–680 (2002). 22. 35 US Code § 102(b). 23. Tokkyo Ho [Patent Law] § 30(31) & 30(33). 24. Anonymous. European Patent Convention. Article 54 (European Patent Office, Munich, 1973). 25. Benowitz, S. J. Natl. Cancer Inst. 94, 80–81 (2002). 26. Borger, J. Rush to patent genes stalls cures for disease. Guardian (London) 15 December 1999, p.1. 27. Williams-Jones, B. Health Law J. 10, 123–146 (2002). 28. Caulfield, T., Gold, E.R. & Cho, M.K. Nat. Rev. Genet. 1, 227–231 (2000). 29. Fabrizio, K.R. & Di Minin, A. Res. Policy 37, 914–931 (2008). 30. Petherbridge, L. Maine Law Rev. 59, 339–384 (2007). 31. Bagley, M.A. Boston Coll. Law Rev. 47, 217–274 (2006). 32. Ebersole, T.J., Guthrie, M.N. & Goldstein, J.A. Intellect. Prop. Technol. Law J. 17, 6–13 (2005). 33. Franzoni, C. & Scellato, G. Proc. Acad. Innov. Entrepreneurship 2008, 388–401 (2008). 34. Anonymous. Federal Register. 73, 47535, 47540 (2008). 35. http://www.uspto.gov/web/offices/pac/provapp.htm 36. 35 US Code § 119(e), 120. 37. 35 US Code § 102(e). 38. 35 US Code § 111(b). 39. 35 US Code § 112 (first paragraph). 40. Axelrod, R. Am. Polit. Sci. Rev. 75, 306–318 (1981). 41. Rapoport, A. in Game Theory as a Theory of Conflict Resolution (ed. Rapoport, A.) 17–34 (Springer-Verlag, New York, 1974). 42. Cipra, B. Science 261, 162–163 (1993). 43. Skevington, P.J. & Hart, T.P. BT Technol. J. 15, 39–44 (1997). 44. http://www.digistamp.com/FAQts.htm#legality 45. Feng, H. & Wah, C.C. Inf. Manage. Comput. Secur. 10, 159–164 (2002). 46. http://www.guardtime.com; http://www.digistamp.com 47. Tananbaum, G. & Holmes, L. Learn. Publ. 21, 300– 306 (2008). 48. Wood, D. Learn. Publ. 14, 151–158 (2001). 49. Locke, J. in Two Treatises of Government (ed. Lasslet, P.) Second Treatise §5 (Cambridge University Press, Cambridge, 1967). 50. Machan, T.R. in Liberty for the Twenty-First Century: Contemporary Libertarian Thought (ed. Machan, T.R.) 209–226 (Rowman and Littlefield, Lanham, Maryland, 1995). 51. Narveson, J. in Liberty for the Twenty-First Century: Contemporary Libertarian Thought (ed. Machan, T.R.) 19–40 (Rowman and Littlefield, Lanham, Maryland, 1995). 52. Nozick, R. Anarchy, State and Utopia (Basic Books, New York, 1974). 53. Hailwood, S.A. Exploring Nozick: Beyond Anarchy, State and Utopia (Avebury, Aldershot, UK, 1996).

volume 28 number 2 february 2010 nature biotechnology

patents

© 2010 Nature America, Inc. All rights reserved.

Recent patent applications in antibody fragments Priority application date

Publication date

Patent number

Description

Assignee

Inventor

WO 2009126730, WO 2009126730

A method of preparing nucleotides of single-chain variable fragments encoding an antigen-specific binding domain by amplifying the variable regions of the antibody’s heavy chain and the lambda and kappa light chains using PCR with a set of primers.

University of Pennsylvania (Philadelphia)

Mason N

4/9/2008 10/15/2009, 12/30/2009

US 20090297439

An immuno-imaging agent for the detection of a tumor in a subject, comprising an anti-Met monoclonal antibody, its fragment and a genetically engineered/humanized antibody containing the epitope binding region or complementarity-determining regions of the antibody.

Metheresis Translational Rsesearch (Lugano, Switzerland)

Carminati P, Comoglio PM, van Dongen G

6/2/2008

FR 2931481, WO 2009141458

New isolated antibodies or at least one of their functional fragments, which is specific to an epitope comprising at least one lysyl compound useful, e.g., for detecting epitope in a sample in vitro or in vivo.

Covalab (Villeurbanne, France)

Ceylan I, El Alaoui EBS, Thomas V

5/23/2008 11/27/2009, 11/26/2009

WO 2009138714

A method for the separation of a fragment antibody, e.g., single-chain variable fragment, involving contacting a medium containing the fragment antibody with a synthetic affinity ligand attached to a support matrix under conditions where the fragment antibody binds to the ligand.

Avecia Biologics (Manchester, UK)

Liddell JM

5/16/2008 11/19/2009

FR 2929519, WO 2009136031

Use of a monoclonal antibody secreted by a hybridoma or its functional fragments to prepare a medicament to inhibit the growth of a primary tumor for early cancer treatment where the cancer is, e.g., colon, lung prostate cancer.

Haeuw JF Pierre Fabre Medicament (Boulogne, France)

US 20090269277

A method for delivering an agent useful for diagnosing or treating, e.g., cancer or cardiovascular disease, by administering a hexameric stably tethered structure comprising the agent, an IgG antibody and antibody fragments or cytokines to the subject.

Chang C, IBC Goldenberg DM, Pharmaceuticals (Morris Plains, NJ, Rossi EA USA)

WO 2009129521

An antigen composition for early detection of Mycobacterium tuberculosis or immunizing against infection, comprising, e.g., M. tuberculosis proline threonine repetitive protein fragments having specific sequences that bind antibody specific for protein.

New York University (New York)

WO 2009127046

A new antibody or an antigen-binding fragment comprising a complementarity-determining region Gly-X1e-X2e-X3e-X4e-X5e-X6e-X7e-X8e-His (SEQ ID NO. 65); useful for reducing the growth of prostate tumor cells.

ProScan Rx Cuello AC, Pharma (Montreal) Gold P, Melancon D, Moffett S, Saragovi HU

WO 2009099545, US 20090202557

A method of preparing crystals of an antigen-binding fragment (Fab) of an antibody, comprising mixing the Fab with a reservoir solution comprising polyethylene glycol (PEG) and a buffer.

Abbott Bioresearch (Worcester, MA, USA) Argiriadi MA, Borhani DW, Ghayur T, Wu C, Xiang T

Argiriadi MA, Borhani DW, Ghayur T, Wu C, Xiang T

1/30/2008 8/13/2009, 8/13/2009

WO 2009092014

Purifying a nonaggregated antibody or an immunoreactive antibody fragment from an impure preparation containing the antibody or antibody fragment, comprising contacting the impure preparation with an apatite chromatography support.

Gagnon PS

Gagnon PS

1/18/2008

7/23/2009

KR 2009011215

A surface expression vector expressing short-chain variable fragments of porcine epidemic diarrhea virus (PEDV)-neutralizing antibody encoded by a gene having a defined sequence of 777 amino acids (Seq. id. no. 1) on membrane surface of bacteria; useful for expressing short-chain variable fragments of PEDV-neutralizing antibody on the surface of Escherichia coli, where the bacteria is useful in composition for preventing or treating diarrhea.

Republic of Korea Ministry of Agriculture and Forestry (Seoul)

Cho S, Hyun B, Kim I, Kim S, Pyo H, Song J

7/25/2007

2/2/2009

Laal S, Zolla-Pazner S

4/4/2008

12/3/2009

10/9/2009, 11/12/2009

10/19/2005 10/29/2009

4/19/2008 10/22/2009

4/14/2008 10/22/2009

Source: The status of each application is slightly different from country to country. For further details, contact Thomson Scientific, 1800 Diagonal Road, Suite 250, Alexandria, Virginia 22314, USA. Tel: 1 (800) 337-9368 (http://www.thomson.com/scientific).

nature biotechnology volume 28 number 2 february 2010

141

news and views

ChIPs and regulatory bits Xin He & Saurabh Sinha

© 2010 Nature America, Inc. All rights reserved.

Machine learning reveals combinatorial patterns of transcription factor binding that drive gene expression. Gene expression patterns are orchestrated largely by promoters and enhancers, which integrate a multitude of signaling and transcriptional inputs1. But exactly how these cis-regulatory modules control transcription on a global scale has been devilishly hard to decipher. A recent study by Zinzen et al.2 in Nature marks exciting progress on this problem. The authors’ key innovation is a technique to predict the expression activity driven by any genomic segment using transcription factor occupancy data generated from high-throughput chromatin immunoprecipitation (ChIP) assays. The approach led to the discovery of dozens of novel cis-regulatory modules involved in mesoderm and muscle development in the Drosophila melanogaster embryo, while providing new insights into the underlying regulatory code. In parallel with experimental studies, there have been two main computational directions for explaining the mode of action of cis-regulatory modules—that is, how several transcription factor binding sites in a module integrate the intracellular concentrations of transcription factors to determine gene expression. Approaches based on fundamental physical-chemical principles have made significant headway in predicting function directly from sequence3,4 but have been limited by gaps in our understanding of the underlying molecular mechanisms, such as the combined effects of simultaneous protein-DNA and protein-protein interactions. In addition, statistical approaches have been used to associate specific expression states with the patterns of binding site occurrence shared by the corresponding regulatory sequences5. A common hurdle in both paradigms is the nontrivial nature of predicting transcription factor occupancy from sequence alone. Xin He and Saurabh Sinha are at the University of Illinois at Urbana-Champaign, Urbana, Illinois. e-mail: [email protected] or [email protected]

142

Zinzen et al.2 adopted a pragmatic solution that circumvents this challenge by using ChIP-chip technology to directly measure the occupancy levels of five transcription factors known to be involved in mesoderm specification (Twi, Bin, Tin, Mef2 and Bap). They identified candidate cis-regulatory modules as clusters of ChIP-binding peaks and tackled the problem of mapping quantitative measurements of transcription factor occupancy within these modules to one of five predetermined classes of expression patterns, each of which corresponded to a specific tissue or developmental stage (Fig. 1). A machine learning technique called support vector machine was used to achieve high accuracy in this classification task, sidestepping the need for mechanistic details of the regulatory process. Loosely speaking, this technique treats each module’s occupancy profile as a point in space and learns how to best draw boundaries that separate points corresponding to known modules with different expression patterns. Then, any candidate module’s expression pattern can be predicted based on where it lies in space relative to these boundaries. The authors experimentally validated 35 of 36 predicted cis-regulatory modules for mesoderm specification, demonstrating a remarkable success rate. The method of Zinzen et al.2 uses transcription factor occupancy information on an entire genomic segment rather than at individual binding sites. Its success therefore seems to support the ‘information display’ model of cis-regulatory module function6, which contends that regulatory function depends on the number and types of binding sites in a genomic segment and not on their precise arrangement. However, arrangements of sites may have an important role in determining net transcription factor occupancy or in fine-tuning the expression activity; the paper’s findings do not rule out either possibility.

It is also worth noting that the transcription factor occupancy information used by Zinzen et al.2 was obtained from whole-embryo measurements. That the method does not require spatial information on the concentrations of regulatory proteins is a practical advantage but also raises questions about its generalizability. For example, would it be possible to model the expression patterns of genes involved in anterior-posterior axis specification without reading the spatial patterns of the transcription factors themselves? Or is there a fundamental difference between models appropriate for regulatory networks that respond to a morphogen gradient versus more downstream networks that impart tissue-specific expression? Notably, Zinzen et al.2 found that different transcription factor occupancy profiles may yield the same regulatory activity. Although this has been implicit in existing models of cisregulatory logic4, definitive examples such as those highlighted here are rare. Moreover, the authors observed that diverse cis-regulatory modules with similar activity were bound by a key common regulator (e.g., Twi for mesoderm and Bin for visceral muscle), with other transcription factors acting on specific cis-regulatory modules to modulate the gene expression pattern. This may prove to be a general design principle for achieving expression patterns that exhibit the same tissue specificity while allowing for minor differences. The novelty of the authors’ approach lies in predicting expression solely from quantitative transcription factor occupancy values. This raises the possibility of incorporating direct occupancy measurements (from ChIPchip or ChIP-Seq experiments) into previous, sequence-based models of expression4, which attempt to predict transcription factor occupancies and use these to explain the expression pattern. Such a combination might lead to greater predictive power compared with approaches based on sequence or occupancy alone. Future modeling efforts are also likely

volume 28 number 2 february 2010 nature biotechnology

news and views

Transcription factor

ChIP

Candidate cis-regulatory modules

Twi Tin Mef2 Bin Bap

or mp e T Extract occupancy profiles of known modules

Extract occupancy profile of candidate module Profiles Twi Tin Mef2 Bin Bap

Expression patterns Mesoderm (M)

Somatic muscle (SM)

5− 7 8 10 −9 − 12 11 −1 14 3 −1 5 © 2010 Nature America, Inc. All rights reserved.

e

tag

s al

Visceral muscle (VM)

Stage

Mesoderm and somatic muscle (MSM) Mesoderm and visceral muscle (MVM) Others

Train classifier Support vector machine Predict expression of candidate modules M

SM

VM

MSM MVM Others

In vivo validation

Figure 1 Pipeline for discovery of cis-regulatory modules involved in mesoderm specification. ChIP-chip assays provide genome-wide occupancy information for each of five relevant transcription factors at five different temporal stages of embryonic development. Clusters of ChIP peaks are designated as candidate cis-regulatory modules. Transcription factor occupancy profiles are generated for each candidate module (left). The same ChIP-chip data are used to generate occupancy profiles of previously identified cisregulatory modules (right). These profiles, together with experimentally determined expression patterns driven by each module, which are curated from the literature, are used to train a support vector machine classifier. The classifier is used to predict the expression pattern (visceral muscle in this example) driven by the candidate cis-regulatory module. The prediction is verified in vivo by a transgenic reporter assay. Reporter results reprinted from ref.2, with permission of the authors.

expression patterns are easier to come by8; thus, adapting the authors’ approach to work with gene, rather than module, expression patterns as training data would go a long way toward ensuring broader application. The new method may also be useful in synthetic biology. Whether for ab initio design of a sequence that drives a desired tissue-specific pattern9 or for the refinement of an existing sequence to be used in a synthetic circuit10, the utility of quantitative models of expression is well recognized. The working model proposed here could help to identify several endogenous sequences with the same regulatory function and could even suggest the variants (by specifying targets of mutation) that are best suited for the specific engineering goal. As genome-wide assays of transcription factor–DNA binding become more common, tools that interpret the resulting data to elucidate combinatorial gene regulation will be needed. The study by Zinzen et al.2 offers an innovative approach to building such tools and sets the stage for more in-depth explorations of regulatory networks. COMPETING INTERESTS STATEMENT The authors declare no competing financial interests.

to involve whole-genome assays of chromatin state, such as nucleosome occupancy or various histone modifications4. There are some practical considerations in applying the proposed strategy more broadly. First, the method relies on prior knowledge of all relevant transcription factors, which in the case of mesoderm specification was available from extensive prior work. For studies of other regulatory networks, this requirement might be mitigated using existing statistical

techniques7 that identify binding sites overrepresented in known cis-regulatory modules of the network, thus inferring the relevant transcription factors. Second, the model has a ‘training phase’ that requires expression measurements on a large number of cis-regulatory modules—the authors used 139 modules with previously characterized expression in mesoderm and/or muscle. Such data are not available for most regulatory systems and are difficult to generate. On the other hand, gene

nature biotechnology volume 28 number 2 february 2010

1. Davidson, E.H. The Regulatory Genome: Gene Regulatory Networks in Development and Evolution (Academic Press, 2006). 2. Zinzen, R.P., Girardot, C., Gagneur, J., Braun, M. & Furlong, E.E. Nature 462, 65–70 (2009). 3. Janssens, H. et al. Nat. Genet. 38, 1159–1165 (2006). 4. Segal, E. & Widom, J. Nat. Rev. Genet. 10, 443–456 (2009). 5. Beer, M.A. & Tavazoie, S. Cell 117, 185–198 (2004). 6. Arnosti, D.N. & Kulkarni, M.M. J. Cell. Biochem. 94, 890–898 (2005). 7. Warner, J.B. et al. Nat. Methods 5, 347–353 (2008). 8. Tomancak, P. et al. Genome Biol. 8, R145 (2007). 9. Venter, M. Trends Plant Sci. 12, 118–124 (2007). 10. Haseltine, E.L. & Arnold, F.H. Annu. Rev. Biophys. Biomol. Struct. 36, 1–19 (2007).

143

news and views

From genomics to crop breeding Richard Flavell

© 2010 Nature America, Inc. All rights reserved.

New insights into the maize genome must be incorporated into breeding programs to realize the potential of crop genomics. A collection of recent papers in Science and PLoS Genetics1–7 represents a landmark in plant biology and an indispensable resource for efforts to improve maize by breeding. The papers include the draft genome sequence of an inbred maize line1; analyses of the chromosomal organization of genes, microRNAs and transposable elements1,3; comparisons between the genomes of some agriculturally relevant inbred lines4; and transcriptional profiles of some of the same inbred lines and of hybrids generated from crosses between them2,5. The wealth of information in these reports should accelerate breeding projects aimed at generating superior varieties of corn and other crops. Over 1 billion people eat maize or meat from maize-fed livestock. Maize also provides the raw materials for manufactured products ranging from coatings for paper and cloth to biodegradable plastics and biofuel. Average maize yields over the past 40 years have doubled in the United States, although not elsewhere. This success was due in part to breeding of better-performing hybrids, which are generated by combining the genomes of inbred plants from different genetic groups. There remains much scope for continuing to improve maize yields by exploiting the yield gain in hybrids relative to their inbred parents, a phenomenon known as heterosis. But to support the world’s growing population, it will be necessary to enhance the rate of increase in the productivity of maize and other crops, especially in inhospitable climates. This challenge will likely be addressed through better farming, more reliable seed supplies and more stable markets, as well as by the application of genomics technologies to breeding of superior varieties. As most commercially relevant plant phenotypes depend on the interactions of large numbers of genes and also on the positions of genes with respect to one another and to sites of recombination, plant breeding is an activity that involves whole genomes. Any plant breeder would like to know, first, how every chromosomal segment, gene or allele— Richard Flavell is at Ceres, Inc., Thousand Oaks, California, USA. e-mail: [email protected]

144

alone and in combination with others—contributes to specific traits, and, second, how to alter the genome to manipulate traits at will. Historically, these goals remained out of reach because knowledge of gene–trait associations was limited by the need to measure traits in the field and to define their genetic basis by the inadequate trait mapping procedures of classical and statistical genetics (right side of Fig. 1). With the advent of genomics technologies, breeders can characterize the allelic content of their particular germplasm in exquisite detail throughout the breeding program and so preserve the most valuable allele combinations (left side of Fig. 1). Progress in identifying gene–trait associations for maize is being achieved by constructing and analyzing “nested association mapping” lines. In this approach, chromo-

some segments from diverse inbreds are recombined into sets of inbreds by a very large number of recombination events, and the plants are compared in ways that allow the roles of individual chromosome segments to be inferred6,7. The chromosome segments are also being defined by their molecular polymorphisms and by the genes they carry, allowing them to be tracked throughout the breeding process. However, a very large number of defined polymorphisms will be required to uniquely mark all alleles. This has recently become possible with wholegenome sequencing technologies1,3. Integration of the new genomics technologies with traditional breeding strategies will also empower breeders in their efforts to design and select the best combinations of chromosome segments, genes and alleles available in the species to meet commercial

Diverse population of plants/chromosomes

Gene-trait associations Genome-wide genotyping

Trait analysis in fields at multiple locations

Selection of potential inbred parents of higheryielding hybrids

Crossing of inbreds to make hybrids

Selection and evaluation of heterotic hybrids for commerce

Figure 1 A highly simplified scheme illustrating how genomics can contribute to steps in a maize-hybrid breeding pipeline. Plants selected from parent inbred improvement programs are used in crosses to make F1 hybrids that are then evaluated. F1 hybrids with improved performance are then adopted for commercialization. Historically, traits could not be managed efficiently as the causal DNA sequences remained largely unknown and their existence was recognized only through the use of extensive field trials (right side). New genomics technologies to determine complete genomes, DNA polymorphisms, whole genome expression patterns and chromosomal haplotype blocks (reflecting recombination patterns) now make it possible to build detailed gene-trait associations and manage such associations throughout the breeding program, including the selection of combinations of alleles (left side). The combined use of trait assessments in the field and genomics technologies increases the efficiency of crop breeding.

volume 28 number 2 february 2010 nature biotechnology

criteria for crop improvement (Fig. 1). The recent papers1–7 will guide selection of parents by clarifying the substantial differences in active gene contents between different maize inbreds. Assuming that a major basis of heterosis is the complementation of these differences in inbreds1,2, the new information provides pointers as to which parents should complement well in heterotic hybrids. Interestingly, the genome-wide expression studies of Swanson-Wagner et al.5 suggest that there is much differential gene expression between alleles in hybrids relative to the inbred parents and that this is driven primarily by the paternal allele of trans-acting expression quantitative trait loci—genes that exert their effect by regulating the expression of other genes, often elsewhere in the genome. This provocative finding raises the question of the extent to which heterosis depends also on epigenetic and parent-oforigin imprinting effects, and of whether breeders should therefore focus on these loci, in particular when making crosses in specific directions. The positions and frequencies of recombination events relative to the positions of diverse alleles influence tremendously the efficiencies of improved plant production because they determine how readily new combinations of alleles are created. Gore et al.4 revealed that some 21% of genes lie in low-recombination regions around the centromeres and, thus, that variation in these genes is difficult to exploit in breeding, except by means of new chromosome

combinations and hybrids. Vielle-Calzada et al.3 defined >100 regions characterized by low genetic diversity among maize lines. These regions may therefore be associated with traits that contributed to domestication (such as a larger number of kernels that remain attached to the cob and lack a stony casing) and consequently have been selected

The wealth of information in these reports should accelerate breeding projects aimed at generating superior varieties of corn and other crops. in all modern maize breeding programs. Other regions are highly polymorphic and may therefore be associated with adaptation to different geographic regions and sources of variation that breeders select and elaborate. The largest maize-breeding companies have probably already sequenced many corn genomes, or parts of them, and they have much more accurate information on phenotypes of lines in a wide array of environments than the public sector does. Nevertheless, the studies1–7 will provide these companies with much new data and analyses that will be used to drive their breeding programs. Unfortunately, most smaller companies lack the resources (e.g., databases and information

Spilling the beans on legume biology Soybean is the most recent addition to the rapidly growing list of crops for which a high-

quality draft genome is now available. Writing in Nature, Schmutz et al.1 report that the 1.1-gigabase soybean genome—the largest shotgunsequenced plant genome—is predicted to encode 46,000 genes. Two genome duplication events are likely Roy Kaltschmidt

© 2010 Nature America, Inc. All rights reserved.

news and views

nature biotechnology volume 28 number 2 february 2010

technology systems) to fully exploit the new genomic information, suggesting that the gap between those who have this capability and those who do not will continue to widen. Those who stand to benefit especially from these reports1–7 are the breeders in the public sector and small companies that are seeking to provide improved lines for poorer societies. Foremost among them is the Consultative Group on International Agricultural Research and its tropical maize breeding efforts spearheaded by the International Maize and Wheat Improvement Center (Centro Internacional de Mejoramiento de Maíz y Trigo; CIMMYT). CIMMYT has struggled to fully embrace genomics and has lagged behind the leading private sector companies in exploiting genomics in its breeding program, on which so many depend. Let us hope that these publications will prove sufficiently compelling to inspire their government funders and other publicsector breeders to expedite the application of genomics in crop breeding. COMPETING INTERESTS STATEMENT The author declares no competing financial interests. 1. Schnable, P.S. et al. Science 326, 1112–1115 (2009). 2. Springer, N.M. et al. PLoS Genet. 5, e1000734 (2009). 3. Vielle-Calzada, J.-P. et al. Science 326, 1078 (2009). 4. Gore, M.A. et al. Science 326, 1115–1117 (2009). 5. Swanson-Wagner, R.A. et al. Science 326, 1118– 1120 (2009). 6. McMullen, M.D. et al. Science 325, 737–740 (2009). 7. Buckler, E.S. et al. Science 325, 714–718 (2009).

to account for the observation that ~75% of these genes are found in multiple copies. Although the importance of soybean as a source of protein and oil alone testifies to the potential implications of understanding its genetic makeup, this genome will also serve as the reference for ~20,000 leguminous species that play a critical ecological role through their unique ability to fix nitrogen with the help of rhizobial bacteria. Availability of the genome should accelerate the association of quantitative

trait loci of nutritional, economic and ecologically important traits with the causal DNA sequences from soybean in the near future. In the longer term, the genome will likely also be leveraged to improve the way in which a range of leguminous subsistence crops are used to both replenish soil nitrogen through crop rotation and meet the expanding needs of developing nations for protein and energy. Peter Hare 1. Schmutz, J. et al. Nature 463, 178– 183 (2010).

145

criteria for crop improvement (Fig. 1). The recent papers1–7 will guide selection of parents by clarifying the substantial differences in active gene contents between different maize inbreds. Assuming that a major basis of heterosis is the complementation of these differences in inbreds1,2, the new information provides pointers as to which parents should complement well in heterotic hybrids. Interestingly, the genome-wide expression studies of Swanson-Wagner et al.5 suggest that there is much differential gene expression between alleles in hybrids relative to the inbred parents and that this is driven primarily by the paternal allele of trans-acting expression quantitative trait loci—genes that exert their effect by regulating the expression of other genes, often elsewhere in the genome. This provocative finding raises the question of the extent to which heterosis depends also on epigenetic and parent-oforigin imprinting effects, and of whether breeders should therefore focus on these loci, in particular when making crosses in specific directions. The positions and frequencies of recombination events relative to the positions of diverse alleles influence tremendously the efficiencies of improved plant production because they determine how readily new combinations of alleles are created. Gore et al.4 revealed that some 21% of genes lie in low-recombination regions around the centromeres and, thus, that variation in these genes is difficult to exploit in breeding, except by means of new chromosome

combinations and hybrids. Vielle-Calzada et al.3 defined >100 regions characterized by low genetic diversity among maize lines. These regions may therefore be associated with traits that contributed to domestication (such as a larger number of kernels that remain attached to the cob and lack a stony casing) and consequently have been selected

The wealth of information in these reports should accelerate breeding projects aimed at generating superior varieties of corn and other crops. in all modern maize breeding programs. Other regions are highly polymorphic and may therefore be associated with adaptation to different geographic regions and sources of variation that breeders select and elaborate. The largest maize-breeding companies have probably already sequenced many corn genomes, or parts of them, and they have much more accurate information on phenotypes of lines in a wide array of environments than the public sector does. Nevertheless, the studies1–7 will provide these companies with much new data and analyses that will be used to drive their breeding programs. Unfortunately, most smaller companies lack the resources (e.g., databases and information

Spilling the beans on legume biology Soybean is the most recent addition to the rapidly growing list of crops for which a high-

quality draft genome is now available. Writing in Nature, Schmutz et al.1 report that the 1.1-gigabase soybean genome—the largest shotgunsequenced plant genome—is predicted to encode 46,000 genes. Two genome duplication events are likely Roy Kaltschmidt

© 2010 Nature America, Inc. All rights reserved.

news and views

nature biotechnology volume 28 number 2 february 2010

technology systems) to fully exploit the new genomic information, suggesting that the gap between those who have this capability and those who do not will continue to widen. Those who stand to benefit especially from these reports1–7 are the breeders in the public sector and small companies that are seeking to provide improved lines for poorer societies. Foremost among them is the Consultative Group on International Agricultural Research and its tropical maize breeding efforts spearheaded by the International Maize and Wheat Improvement Center (Centro Internacional de Mejoramiento de Maíz y Trigo; CIMMYT). CIMMYT has struggled to fully embrace genomics and has lagged behind the leading private sector companies in exploiting genomics in its breeding program, on which so many depend. Let us hope that these publications will prove sufficiently compelling to inspire their government funders and other publicsector breeders to expedite the application of genomics in crop breeding. COMPETING INTERESTS STATEMENT The author declares no competing financial interests. 1. Schnable, P.S. et al. Science 326, 1112–1115 (2009). 2. Springer, N.M. et al. PLoS Genet. 5, e1000734 (2009). 3. Vielle-Calzada, J.-P. et al. Science 326, 1078 (2009). 4. Gore, M.A. et al. Science 326, 1115–1117 (2009). 5. Swanson-Wagner, R.A. et al. Science 326, 1118– 1120 (2009). 6. McMullen, M.D. et al. Science 325, 737–740 (2009). 7. Buckler, E.S. et al. Science 325, 714–718 (2009).

to account for the observation that ~75% of these genes are found in multiple copies. Although the importance of soybean as a source of protein and oil alone testifies to the potential implications of understanding its genetic makeup, this genome will also serve as the reference for ~20,000 leguminous species that play a critical ecological role through their unique ability to fix nitrogen with the help of rhizobial bacteria. Availability of the genome should accelerate the association of quantitative

trait loci of nutritional, economic and ecologically important traits with the causal DNA sequences from soybean in the near future. In the longer term, the genome will likely also be leveraged to improve the way in which a range of leguminous subsistence crops are used to both replenish soil nitrogen through crop rotation and meet the expanding needs of developing nations for protein and energy. Peter Hare 1. Schmutz, J. et al. Nature 463, 178– 183 (2010).

145

news and views

Systematic tracking of cell fate changes Jonghwan Kim & Stuart H Orkin

Deciphering the regulation of eukaryotic gene expression is a formidable challenge because of the multilayered nature of regulatory mechanisms. In an effort to decode the complexity of molecular processes governing cell fate changes in mouse embryonic stem (ES) cells, Lu et al.1, in a recent issue of Nature, integrated multiple ‘omics’ data sets and systematically monitored temporal changes in some of the key regulatory events (Fig. 1). In offering this broad view, their systems approach to understanding dynamic fate changes provides insight into how to deconstruct complex networks in other cellular contexts, such as lineage specification, differentiation and somatic cell reprogramming. Cells modulate gene expression in response to external and/or internal stimuli. Owing to the complexity of regulatory mechanisms, efforts to date have focused largely on one aspect at a time. For example, investigators have studied epigenetic modification, transcription, post-transcriptional modification, translation or post-translational modification occurring within a sequence of events, rather than all processes in aggregate. The recent development of high-throughput technologies—including gene expression profiling; global mapping of protein-DNA interactions; mapping of histone modification by microarray or sequencing; proteinprotein interaction mapping and protein abundance measurement by mass spectrometry; and gene knockdown by RNA interference—offers the potential to observe biological phenomenon at a global level2. As each of these methods generates a wealth of data, data handling and analysis tools become rate limiting in the interpretation. This is the challenge addressed by Lu et al.1. ES cells are distinguished by their capacity for perpetual self-renewal and pluripotency (the ability to differentiate into all tissues). Nanog is one of the key transcription factors in ES cells and is known to be required for maintenance of pluripotency in mouse ES Jonghwan Kim and Stuart H. Orkin are at the Children’s Hospital Boston and Dana Farber Cancer Institute, Boston, Massachusetts, USA. e-mail: [email protected]

146

ES cells

Differentiated cells Nanog downregulation

Day 0 Day 1

Day 3

Day 5

Temporal regulation Different regulatory levels

© 2010 Nature America, Inc. All rights reserved.

High-throughput measurements across several regulatory levels provide a comprehensive view of ES-cell differentiation.

Interactive analysis and visualization of data using GATE software

Histone modification RNA polymerase II occupancy mRNA abundance

Data acquisition and integration

Nuclear protein abundance

and/ or

Heat map movie

Dynamic scatter plot

Figure 1 Systematic monitoring of differentiating ES cells. Upon knockdown of the key transcription factor Nanog, changes in histone acetylation, RNA polymerase II occupancy, mRNA abundance and nuclear protein abundance were measured over five days of differentiation by systems biology tools. Acquired data sets were integrated, analyzed and visualized by GATE software.

cells3,4. In their elegant experiments, Lu et al.1 sought to introduce a single genetic perturbation by knocking down Nanog with an inducible small hairpin RNA and to examine the consequences of this perturbation in a global fashion. Upon knockdown of Nanog, ES cells exit the stem cell state and differentiate. During the accompanying dynamic changes, Lu et al.1 measured the outcomes at four different regulatory layers: (i) histone H3 lysine 9 and 14 acetylation by ChIP-chip, indicating an active epigenetic signature, (ii) RNA polymerase II occupancy by ChIPchip, indicating active gene transcription, (iii) mRNA transcript abundance by expression microarray, indicating an outcome of transcription, and (iv) nuclear protein abundance by mass spectrometry. They collected data for each regulatory layer at four time points (days 0, 1, 3 and 5 after Nanog knockdown) to monitor the temporal sequence of events. By comparing changes in each regulatory layer, Lu et al.1 observed that, in general, changes between different gene expression steps are moderately well correlated. However, they also found discordant regu-

lation for a large number of genes. Notably, ~42–53% of all proteins whose levels change substantially did not correlate with changes in their respective mRNA abundances. This discrepancy has been observed in lower organisms, such as yeast, but has been rather unclear in the mammalian context owing to either technical limitations or improper experimental settings5. Although Lu et al.1 did not address this issue further, their results clearly imply that additional layers of regulation beyond transcriptional control, such as translational and post-translational modifications, are of critical importance in cell fate decisions. Using gene-ontology analysis, Lu et al.1 also observed that chromatin-modifying enzymes are regulated by RNA polymerase II occupancy but not by the other three regulatory layers. This suggests that although chromatin remodeling is known to be important in ES cell fate decision, primary control is exerted through transcriptional regulation via transcription-factor occupancy. A challenge in handling such large, diverse data sets is the difficulty of visualizing the data without adequate graphical software.

volume 28 number 2 february 2010 nature biotechnology

© 2010 Nature America, Inc. All rights reserved.

news and views The authors provide a tool, Grid analysis of time-series expression (GATE), which uses a correlation-based clustering algorithm for the comparison and visualization of multilayered time-course data sets as interactive images or movies6. Using GATE, Lu et al.1 conducted temporal and dynamic analyses among different regulatory layers with clustering of genes that are regulated in a similar pattern, revealing connections between the regulatory mechanisms underlying ES cell fate changes. Their analysis reveals the temporal order of regulatory network configurations after a single perturbation, the downregulation of Nanog. The study by Lu et al.1 does not plumb the full complexity of cell fate determination. Rather, it should be viewed as a first step in what will have to be a series of attempts to incorporate multiple data sets into a comprehensive understanding of gene regulation. Lu et al.1 introduced an artificial perturbation in ES cells by knocking down a single critical factor. Surely, during normal development, changes in multiple regulators may occur

simultaneously. The necessity to oversimplify the regulatory problem is evident in other ways. For example, different kinds of histone modifications involved in positive or negative gene regulation that are likely to be functionally relevant were not examined. Moreover, several forms of regulation were not studied, including post-transcriptional/ translational regulation by microRNAs7, post-translational modification of proteins, and modulation of protein localization. A truly comprehensive accounting of the dynamics of fate changes will require consideration of these additional regulatory layers. Tracking cell-state transitions by multiple high-throughput assays, as well as the integration of such observations in a systematic fashion, is a monumental task. Ultimately, one would like to develop ways to use the kinds of large data sets analyzed by Lu et al.1 to predict the specific outcomes in ES cells of other regulatory perturbations. Given the heterogeneity in gene expression among cells8,9, it is likely that additional

nature biotechnology volume 28 number 2 february 2010

technologies will be needed, such as quantitative monitoring of gene expression and of epigenetic modifications at the single-cell level. The development of readily accessible databases for storing large data sets from various platforms, as well as user-friendly analysis and visualization tools, will also be necessary to facilitate the comprehensive understanding of multilayered gene regulatory networks during dynamic cellular processes. COMPETING INTERESTS STATEMENT The authors declare no competing financial interests. 1. Lu, R. et al. Nature 462, 358–362 (2009). 2. MacArthur, B.D., Ma’ayan, A. & Lemischka, I.R. Cold Spring Harb. Symp. Quant. Biol. 73, 211–215 (2008). 3. Mitsui, K. et al. Cell 113, 631–642 (2003). 4. Chambers, I. et al. Cell 113, 643–655 (2003). 5. de Sousa Abreu, R., Penalva, L.O., Marcotte, E.M. & Vogel, C. Mol. Biosyst. 5, 1512–1526 (2009). 6. MacArthur, B.D., Lachmann, A., Lemischka, I.R. & Ma’ayan, A. Bioinformatics 26, 143–144 (2010). 7. Marson, A. et al. Cell 134, 521–533 (2008). 8. Singh, A.M., Hamazaki, T., Hankowski, K.E. & Terada, N. Stem Cells 25, 2534–2542 (2007). 9. Chambers, I. et al. Nature 450, 1230–1234 (2007).

147

research highlights

© 2010 Nature America, Inc. All rights reserved.

Adaptive optics in microscopy Light microscopy of whole tissues is often complicated by distortions caused by the optical inhomogeneities in the biological specimen. By borrowing and adapting approaches from astronomy, Ji et al. develop adaptive optics to correct aberrations. The correction is achieved using a ‘spatial light modulator’, an active optical element that permits adjustment of the tilt and phase of the light passing through more than 100 individual segments at the rear pupil of the objective lens. The extent of the modulation for each of the segments is determined by an algorithm that first measures and corrects the spatial deflections caused by the sample inhomogeneities and then corrects errors in the phase for each of the segments. The improvements in signal strength, image fidelity and resolution that can be achieved using adaptive optics are demonstrated by imaging fluorescent beads and neurons in 300- to 500-µm thick brain slices. The technique can be used to improve the performance of many wide-field and point scanning microscopy technologies, including the latest super resolution techniques that are especially sensitive to optical aberrations. (Nat. Methods advance online publication, December 27, 2009, doi:10.1038/nmeth.1411) ME

Digestible plant walls Rigid cell walls give plants strength, but they also confound attempts by plant genetic engineers to convert woody plants to biofuels. Creating rigid cell walls requires intermolecular cross-linking of pectins, facilitated by de-methyl-esterified homogalacturonans (HGA). Releasing fermentable sugars from plant cell walls, on the other hand, requires environmentally unfriendly chemicals or high temperatures. Now Lionetti et al. show that plants transformed with enzymes that inhibit the de-esterification of HGA polymers are more accessible to enzymatic degradation. Twice as much sugar was released from the leaves of transgenic Arabidopsis thaliana plants overexpressing fungal polygalacturonase, 60% more when plants were transformed with a pectin methylesterase inhibitor (PMEI). To show that pectin architecture was being modified, the researchers reacted the transgenic leaves with an antibody that binds to blocks of de-esterified HGA and found reduced binding to the transgenic plants. They were able to replicate these findings in wheat (Triticum durum), an industrially important plant. Finally, they found that Arabidopsis expressing PMEI had more biomass than the control plants (due to cell expansion). This contrasts with polygalacturonase-expressing transgenic plants that generally have less biomass. The group suggests that regulating polygalacturonase expression in time and space might prevent the loss of biomass. (Proc. Natl. Acad. Sci. USA 107, 616–621, 2010) LD

Microarray SNP detection heats up Gresham et al. have discovered new rules to enhance the accuracy of DNA microarrays. These rules substantially improve the ability Written by Kathy Aschheim, Laura DeFrancesco, Markus Elsner & Craig Mak

148

of probes on the array to bind the correct sequence when similar sequences are present in a sample, which is particularly useful when identifying variation at the level of single nucleotide polymorphisms (SNPs). The authors varied different experimental parameters and discovered that a probe is best able to discriminate SNPs when its melting temperature (Tm) is ~2–5 °C below the temperature used to hybridize samples to the array. This knowledge was used to design microarrays with probes between 16 and 35 nt in length—in contrast to the common practice of making uniform-length probes— such that all probes have the same melting temperature, and thus all achieve optimal performance at the same hybridization temperature, ~2–5 °C above the Tm. These so called isothermal microarrays outperformed conventional arrays at identifying heterozygous SNPs. The use of an optimum hybridization temperature in tandem with uniform Tm probes (varying in length) may be useful for designing arrays for other applications. (Proc. Nat. Acad. Sci. USA, published online January 8, 2010, doi:10.1073/pnas.0913883107) CM

Co-workers in transcription factories Chromosome conformation capture on a chip (4C) is capable of detecting remote chromatin interactions on a genome-wide scale. Schoenfelder et al. use a modification of this method to analyze the genome-wide repertoire of transcriptional interactions associated with globin genes in erythroid tissues. Their technique, dubbed enhanced ChIP-4C, first cross-links proteins and DNA to generate a snapshot of the spatial organization of the nucleus. However, the 4C assay is then modified to incorporate an RNA polymerase–recognizing antibody that identifies DNA that is near to, but not necessarily on the same chromosome as, actively transcribed copies of the ‘bait’ gene (that is, globin). Subsequently, a biotinylated bait-specific DNA probe is used to enrich for ‘prey’ sequences, which are cross-linked to the bait. The DNA in the enriched sample is then identified by microarray analysis. Analysis of the promoters of actively transcribed globin genes in mouse cells reveals a transcription factor, Klf1, required for regulating genes in globin-containing ‘transcription factories’ in the nucleus. The approach should facilitate understanding of how genes are brought together in the nucleus to regulate their expression. (Nat. Genet. 42, 53–61, 2010) CM

Platelet ally Biotech has no shortage of new ideas on how to staunch bleeding. Where traditional therapy amounts to little more than the application of pressure or absorbent material, research in the past decade has sought to enhance the body’s intrinsic mechanisms of coagulation. Allogeneic platelets, recombinant clotting factors, red blood cells displaying the cell-adhesive RGD sequence, self-assembling peptides, liposomes and a block copolymer of hemoglobin and fibrinogen are some of the strategies that have been tried, but none has demonstrated adequate safety and efficacy. Now Bertram et al. have proposed to control bleeding with “synthetic platelets,” or nanoparticles consisting of poly(lactic-co-glycolic acid)-poly-l-lysine block copolymer cores carrying polyethylene glycol chains that are capped with RGD sequences. Working with a rat model of major injury to the femoral artery, the authors found that the nanoparticles bind to platelets and boost clot formation more effectively than existing therapies. Moreover, the nanoparticles were rapidly cleared from the circulation, and no adverse effects were observed. (Sci. Transl. Med., published online December 16, 2009, doi:10.1126/scitransl med.3000397) KA

volume 28 number 2 february 2010 nature biotechnology

resource

Rational association of genes with traits using a genome-scale gene network for Arabidopsis thaliana

© 2010 Nature America, Inc. All rights reserved.

Insuk Lee1,2,5, Bindu Ambaru4,5, Pranjali Thakkar4, Edward M Marcotte2,3 & Seung Y Rhee4 We introduce a rational approach for associating genes with plant traits by combined use of a genome-scale functional network and targeted reverse genetic screening. We present a probabilistic network (AraNet) of functional associations among 19,647 (73%) genes of the reference flowering plant Arabidopsis thaliana. AraNet associations are predictive for diverse biological pathways, and outperform predictions derived only from literature-based protein interactions, achieving 21% precision for 55% of genes. AraNet prioritizes genes for limited-scale functional screening, resulting in a hit-rate tenfold greater than screens of random insertional mutants, when applied to early seedling development as a test case. By interrogating network neighborhoods, we identify AT1G80710 (now DROUGHT SENSITIVE 1; DRS1) and AT3G05090 (now LATERAL ROOT STIMULATOR 1; LRS1) as regulators of drought sensitivity and lateral root development, respectively. AraNet (http://www.functionalnet.org/aranet/) provides a resource for plant gene function identification and genetic dissection of plant traits. Manipulating plant traits that affect the production of food, fiber and renewable energy has important agricultural and economic con­ sequences. What is the best approach for identifying genes for impor­ tant plant traits? Forward genetics is limited as mutations in many genes may generate only moderate or weak phenotypes. Similarly, although reverse genetics allows for directed assay of gene perturba­ tions1, saturated phenotyping for many plant traits is impractical. A pragmatic near-term solution is the computational identification of likely candidate genes for desired traits, allowing for focused, efficient use of reverse genetics. This solution is not unlike rational drug design in which computer-assisted and expert knowledge are combined with targeted screening for the desired drug, or in this case, trait. One emerging approach for prioritizing candidate genes is networkguided guilt by association. In this approach, functional associations are first determined between genes in a genome on the basis of exten­ sive experimental data sets, encompassing millions of individual obser­ vations. Such a map of functional associations is often represented as a graph model and referred to as a functional gene network2. Probabilistic functional gene networks integrate heterogeneous biological data into a single model, enhancing both model accuracy and coverage. Once a suitable network is generated, new candidate genes are proposed for traits based upon network associations with genes previously linked to these traits. Such network-guided screen­ ing has been successfully applied to unicellular organisms3,4 and Caenorhabditis elegans5,6, and is a proposed strategy for identifying human disease genes4,5,7–10. We demonstrate here that this approach successfully identifies genes affecting specified traits for a reference flowering plant, A. thaliana, and we introduce a genome-wide, functional gene network for

Arabidopsis suitable for prioritizing candidate genes for a wide variety of traits of economic and agricultural importance. RESULTS Reconstruction of an Arabidopsis gene network We integrated diverse functional genomics, proteomics and compara­ tive genomics data sets into a genome-wide functional gene network, using data integration and benchmarking methods customized for Arabidopsis genes (Supplementary Methods). The data sets included mRNA co-expression patterns measured from DNA microarray data sets (Supplementary Table 1), known Arabidopsis protein-protein interactions11–14, protein sequence features including sharing of protein domains, similarity of phylogenetic profiles15–17 or genomic context of bacterial or archaebacterial homologs18–20, and diverse gene-gene associations (mRNA co-expression, physical protein interactions, multiprotein complexes, genetic interactions, litera­ ture mining) transferred from yeast21, fly11,22–24, worm5 and human genes based on orthology25 (Supplementary Table 2). In total, 24 distinct types of gene-gene associations, encompassing >50 ­million individual experimental or computational observations, were scored for their ability to correctly reconstruct shared membership in Arabidopsis biological processes. Then, these were incorporated into a single integrated network model, dubbed AraNet. AraNet contains 1,062,222 functional linkages among 19,647 genes (~73% of the total Arabidopsis genes), with each linkage weighted by the log likelihood of the linked genes to participate in the same biological processes. Integrating data improves network coverage and accuracy, as tested by recovery of known functional associations (Fig. 1a and Supplementary Fig. 1). AraNet extends substantially beyond

1Department of Biotechnology, College of Life Science and Biotechnology, Yonsei University, Seodaemun-gu, Seoul, Korea. 2Center for Systems and Synthetic Biology, and 3Department of Chemistry and Biochemistry, Institute for Cellular and Molecular Biology, University of Texas at Austin, Austin, Texas, USA. 4Department of Plant Biology, Carnegie Institution for Science, Stanford, California, USA. 5These authors contributed equally to this work. Correspondence should be addressed to I.L. ([email protected]) or E.M.M. ([email protected]) or S.Y.R. ([email protected]).

Received 10 June 2009; accepted 23 December 2009; published online 31 January 2010; doi:10.1038/nbt.1603

nature biotechnology  VOLUME 28  NUMBER 2  FEBRUARY 2010

149

a

b

Likelihood ratio for linked genes to participate in same biological process (cumulative likelihood ratio, log scale)

Coverage of 27,029 protein-coding genes (%)

resource

© 2010 Nature America, Inc. All rights reserved.

an

no

ta Re Al tion liab l a o le nn nly ot at io Ar n aN et

Figure 1  Construction, accuracy and coverage of AraNet, a functional AraNet gene network for Arabidopsis. (a) Pairwise gene linkages derived from 24 AT-CX 148 80 AT-DC AT-GN diverse functional genomics and proteomics data sets, representing >50 AT-LC 90 AT-PG million experimental or computational observations, were integrated into CE-CC 60 AT-LC CE-CX a composite network with higher accuracy and genome coverage than any 55 CE-GT accuracy CE-LC individual data set. The integrated network (AraNet) contains 1,062,222 level CE-YH 33 DM-PI 40 functional linkages among 19,647 (73%) of the 27,029 protein-coding HS-CX HS-DC Arabidopsis genes. The plot x axis indicates the log-scale percentage of HS-LC 20 HS-MS HS-YH the 27,029 protein-coding genes13 covered by functional linkages derived 20 SC-CC 12 SC-CX from the indicated data sets (curves); the y axis indicates predictive SC-DC SC-GT quality of the data sets, measured as the cumulative likelihood ratio of 7 SC-LC 0 SC-MS linked genes to share GO biological process term annotations, tested using SC-TS 4 SC-YH 0.632 bootstrapping and plotted for successive bins of 1,000 linkages 1 10 100 each (symbols). Data sets are named as XX-YY, where XX indicates species Coverage of 27,029 protein-coding genes (%) of data origin (AT, A. thaliana; CE, C. elegans; DM, D. melanogaster; HS, H. sapiens; SC, S. cerevisiae) and YY indicates data type (CC, co-citation; CX, mRNA co-expression; DC, domain co-occurrence; GN, gene neighbor; GT, genetic interaction; LC, literature-curated protein interactions; MS, affinity purification/mass spectrometry; PG, phylogenetic profiles; PI, fly protein interactions; TS, tertiary structure; YH, yeast two-hybrid. Relative contribution of each data type and combining different evidences for inferring function is discussed in Supplementary Discussion (Supplementary Figs. 13–15). (b) AraNet spans ~73% of the protein-coding genes, far in excess of current GO biological process annotations for Arabidopsis, for which ~12.2% of genes are annotated by reliable experimental evidence (GO evidence codes IDA, IMP, IGI, IPI, IEP) or traceable author statements (GO evidence code TAS), or ~45% annotated by any evidence including computational inferences or sequence homology. The subset of AraNet linkages stronger than the likelihood ratio for literature-curated protein interactions (AT-LC, corresponding to a likelihood ratio of 35:1) covers 55% of the genes.

well-characterized Arabidopsis genes (Fig. 1b): 23,720 Arabidopsis genes are unannotated with Gene Ontology (GO) biological process annota­ tions by reliable experimental evidence13. AraNet includes more than half (7,465) of genes lacking even sequence homology–based annota­ tions (14,847 genes). These genes’ functions can now be hypothesized based upon their network neighborhoods. AraNet implicates specific processes for 4,479 uncharacterized genes using guilt by association. There are 2,986 uncharacterized genes associated only with other uncharac­terized genes in AraNet, suggesting many still-uncharacterized cellular processes in plants. Evaluating the accuracy of AraNet To verify the reliability of functional associations in AraNet, we tested their consistency with known Arabidopsis gene annotations by applying guilt by association in AraNet to identify genes associ­ ated with specific biological processes. Each gene in the genome was scored for association with a particular process by summing network edge weights connecting that gene to known genes in that process. A gene’s resulting score corresponds to the naive Bayes estimate for the gene to belong to that process given network evidence (Fig. 2a). Performing cross-validation of this test allows us to assess predictive power with a receiver-operator characteristic (ROC) curve, measur­ ing the true-positive prediction rate versus false-positive prediction rate as a function of prediction score. We use the area under the ROC curve (AUC) to summarize performance. AUC values of ~0.5 and 1 indicate random and perfect performance, respectively. Using cross-validation, we tested AraNet’s ability to correctly associate genes with each GO biological process, observing signifi­ cantly better than random predictability for the majority of biologi­ cal processes (P < 10−53; Wilcoxon signed rank test unless noted other­wise) (Fig. 2b). AraNet incorporates data from other organ­ isms; we correspondingly observed higher predictability for evolu­ tionarily conserved processes than for GO processes annotated only with plant genes (P < 10−24, Wilcoxon rank sum test) (Figs. 2c,d). Nonetheless, genes were correctly associated with plant-specific processes at significantly higher rates than expected by chance (P < 10−28) (Fig. 2d). For example, many important plant traits showed high predictability, including abiotic stress responses (Fig. 2e), organ development (Fig. 2f), biotic stress responses (Supplementary

150

Fig. 2a) and hormonal signaling (Supplementary Fig. 2b). Tests on two additional independent data sets—the set of reliable GO cellular component annotations describing 86 subcellular locations or pro­ tein complexes of Arabidopsis proteins13 and the KEGG definitions of 82 Arabidopsis biochemical pathways26—show similarly high predictive power (P ≤ 10−16 and P ≤ 10−14, respectively, compared to chance) (Fig. 3). We find that AraNet shows far stronger predict­ ability than previous smaller-scale networks of Arabidopsis genes (Supplementary Table 3 and Supplementary Fig. 3)27–30, stemming at least in part from greater coverage, which allows for stronger guilt by association due to higher query gene coverage. Thus, AraNet is strongly predictive for many Arabidopsis processes, including those specific to plants. Contributions of plant versus non-plant data to AraNet Many AraNet linkages derive from orthology to animals and yeast, organisms evolutionarily distant from Arabidopsis. Therefore, we asked the extent to which non-plant–derived linkages contribute to AraNet’s accuracy. A version of AraNet composed only of links from yeast and animal data sharply reduces predictive power (P < 10−4) (Fig. 2b). A version of only plant-derived links performs substan­ tially better (P ≤ 10−11), implying that plant-derived data underlies much of AraNet’s predictive power (Fig. 2b). This plant-derived data dependence is stronger when predicting plant-specific than conserved processes (P ≤ 10−19, Wilcoxon rank sum test) (Fig. 2c,d). Any method for linking genes to plant traits must perform well for plant-specific processes, so we examined evidence supporting these cases. Notably, even for well-predicted plant-specific pathways (AUC scores ranging from ~0.7 to 1, Fig. 4), supporting evidence did not derive entirely from plants. For example, photorespiration genes were identified by combining evidence from Arabidopsis, human and C. elegans. Similarly, trichome differentiation genes were recovered using predominantly human and yeast-derived evidence. Genes of abscisic acid–mediated signaling drew support from all organisms, including fly protein interactions. Thus, although plant-derived links provide most of AraNet’s predictive power, non-plant–derived linkages help ­substantially in associating genes with plant-specific processes, as proc­ esses unique to plants nonetheless often involve conserved genes with conserved interactions.

VOLUME 28  NUMBER 2  FEBRUARY 2010  nature biotechnology

resource a

True-positive rate

Figure 2  Predictive power of AraNet for conserved and plant-specific biological processes. AraNet’s 1 predictive capacity was measured using crossvalidated receiver operator characteristic (ROC) curve analysis, as illustrated in (a). For a given process, each gene in the Arabidopsis genome is rank-ordered by the sum of its network linkage 0 0 1 scores to the set of ‘bait’ genes already associated False-positive rate with that process (omitting each bait gene from the bait set for purposes of evaluation). High-scoring GO BP (conserved) GO BP (plant-specific) GO BP (all) genes are most tightly connected to the bait set and 1.0 1.0 1.0 are the most likely new candidates to participate 0.9 0.9 0.9 in that process. This trend is evident in a ROC plot 0.8 0.8 0.8 measuring recovery of bait genes as a function of 0.7 0.7 0.7 rank, calculating the true-positive prediction rate 0.6 (sensitivity; TP/(TP+FN)) versus the false-positive 0.6 0.6 prediction rate (1−specificity; FP/(FP+TN)). If bait 0.5 0.5 0.5 genes are highly interconnected (red circles), unlike 0.4 0.4 0.4 random genes (blue circles), additional genes connected to the bait genes (green circles) are more likely to be involved in the same process. The area under the cross-validated ROC curve (AUC) provides Organ development Abiotic response Random (0.50) Random (0.50) a measure of predictability, ranging from ~0.5 for 1.0 1.0 Root development (0.62) Response to water random expectation (blue curve) to 1 for perfect Cuticle deprivation (0.73) 0.8 0.8 predictions (red curve). (b) Distributions of AUC development (0.61) Response to hydrogen Stamen values are plotted for network-based identification peroxide (0.73) 0.6 0.6 development (0.64) Cold acclimation (0.72) of genes for each of the 318 GO biological process Stomatal complex 0.4 0.4 Response to heat (0.80) morphogenesis (0.71) terms with annotations, (c) for each of the 151 Trichome Response to high biological process terms with annotations shared 0.2 0.2 morphogenesis (0.73) light intensity (0.79) between plant and animal or between plant and Carpel Response to development (0.75) 0 0 oxidative stress (0.82) yeast and (d) for each of the 167 biological process 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 Ovule terms with annotations found in plants but absent development (0.81) False-positive rate False-positive rate from animals and fungi. In bar-and-whiskers plots, the central horizontal line in the box indicates the median AUC and the boundaries of the box indicate the first and third quartiles of the AUC distribution. Whiskers indicate the 10th and 90th percentiles, and circles indicate individual outliers. AraNet specifically identified genes associated with (e) plant abiotic stress response genes and (f) organ developmental processes, as annotated by GO. AUC values are indicated in parentheses.

Linked genes share cell type–specific expression patterns Many traits in multicellular organisms pertain to specific tissues or cell types. The predictive strength shown by AraNet for such proc­ esses raises the question of how a global gene network, incorporat­ ing diverse samples and data from orthologs, can correctly identify genes for cell type– and tissue-specific processes. Using measure­ ments of transcript observations in 20 root cell types31 that were not used in building AraNet, we measured the extent to which genes linked in AraNet were spatiotemporally co-expressed in these cells. We find that linked genes show strong cell-specific co-expression in Arabidopsis (Fig. 3c)—indeed, far stronger than in previous networks of Arabidopsis genes (Supplementary Table 3)27–30—with linked genes four times more likely to be expressed in the same cell types than expected by chance. Thus, although different individual networks were not constructed for each cell type, such cell and tissue specificity is nonetheless at least in part implicitly encoded in AraNet linkages. This correlation between functional association and spatiotemporal co-expression of genes likely enhances prediction strength for many traits, and is evident even for linkages between characterized and uncharacterized genes (Fig. 3c), supporting applicability of AraNet to uncharacterized genes. Associating genes with specific mutant phenotypes Because linked genes in AraNet tend to operate in the same processes (Figs. 1–4), we might expect that they often affect the same phenotypic traits3,5. This allows association of new candidate genes with traits of interest based on network connections. To test this, we used results from large-scale mutant seed phenotyping32 and analyzed genes

nature biotechnology  VOLUME 28  NUMBER 2  FEBRUARY 2010

do m A r Ar aN aN e e da t p t ta la Ar on nt pl aN ly an et t d no at a

Area under ROC curve

do m A aN raN e e da t p t ta la Ar on nt pl aN ly an et t d no at a Ar

R an

f

True-positive rate

do m Ar A r aN aN da et p et ta la Ar on nt pl aN ly an et t d no at a

R an

True-positive rate

e

© 2010 Nature America, Inc. All rights reserved.

d

Area under ROC curve

Area under ROC curve

c

R an

b

whose disruption induced embryonic lethality or changes in seed (embryo) pigmentation. Genes involved in each trait were interlinked significantly more often compared to chance (p < 10−31 for embryonic lethality and P < 10−10 for seed pigmentation, normal distribution) (Fig. 3d). Unlike AraNet, previous Arabidopsis gene networks27–30 do not significantly predict either phenotype (Supplementary Fig. 4). Thus, AraNet offers a feasible approach for selecting genes likely to be associated with specific plant traits. Tenfold enrichment for seed pigmentation genes To experimentally test the association of new genes with a trait, we used 23 known seed pigmentation genes (Supplementary Table 4) to search AraNet for new pigmentation genes. Genes in this phenotypic class generally affect chloroplast development or photomorphogenesis, and mutant seedlings show early developmental defects, with albino, pale green, purple or variegated leaves33. From AraNet’s top 200 candidate genes, we screened all genes with available homozygous T-DNA insertional mutant lines (Supplementary Table 5). We screened 90 candidate genes (repre­ sented by 118 mutant lines), of which 14 genes (represented by 17 lines) exhibited color and morphology defects in young seedlings, reminiscent of seed pigmentation mutants (Supplementary Tables 6 and 7). This represents a tenfold enrichment in the discovery rate of the mutant phenotype (P ≤ 10−12, binomial distribution) over that observed during screens of T-DNA insertional lines33 (see Online Methods). This discovery rate compares well to animal networks, for example, in C. elegans 16 tumor suppressor effectors were identified from 170 candidates5.

151

resource

0.7 0.6 0.5

0.8 0.7 0.6 0.5

do m Ar aN pl an et Ar td aN at a et on no ly pl an td at a Ar aN et

m do

Ar aN et

R an

Ar aN pl an et Ar td aN at a et o no nl y pl an td at a

0.4

True-positive rate

0.8

d

4.5 4.0 3.5 3.0 2.5 2.0 1.5

1.0 0.8 0.6

Random Embryonic lethality (0.71) Seed pigment (0.68)

0.4 0.2 0

1.0

0

0.5

0.2 0.4 0.6 0.8 False-positive rate

1.0

do m i In Mu lti zed te ro ne lo t gs wo rk G net w G or M ne k tw or k At PI Ar D Kn aN A ow et raN no et nsu un A bs kn T-C et ow X of n Ar lin aN ks et

0.9

Area under ROC curve

0.9

c

Likelihood ratio of linked genes co-expressing across 20 root cell types

Isozyme-free KEGG 1.0

0.4

Figure 3  Validation of AraNet by independent data sets. (a) The predictive power of AraNet for 86 GO cellular component terms. (b) AraNet predictive power for 82 KEGG metabolic pathways (excluding isozymes). Bar-and-whiskers plots are as in Figure 2b–d. (c) Co-occurrence of mRNA transcripts across 20 root cell-types31 for network-linked genes. Genes linked in AraNet (AraNet, lane 6) were more co-expressed in each root cell-type than gene pairs from random networks (repeating the calculation for 100 randomized networks and plotting the distribution of the 100 resulting odds ratios (randomized, lane 1; see Supplementary Methods for details). This trend cannot be explained simply by the incorporation of Arabidopsis mRNA expression data into AraNet, as a version of the network lacking this data shows similarly high cell-type specificity (AraNet no AT-CX, lane 7). Compared to other previous Arabidopsis networks (lanes 2–5, Supplementary Table 3), AraNet shows 1.6- to 4-fold higher cell-type specificity. Subset of AraNet composed of only known-unknown genes shows high cell type–specificity as well, supporting predictability of AraNet using cell-specific expression data (lane 8). (d) AraNet shows predictability for genes affecting embryonic lethality or changes in seed pigmentation, as identified in the SeedGenes database 32. AUC values are indicated in parentheses.

Of the 14 genes with mutant phenotypes, 3 genes (AT5G45620, AT4G26430 (also known as CSN6B) and AT5G50110) exhibited the phenotypes in two alleles, 6 genes in only one of the two alleles, and 5 genes were tested in only one allele (Fig. 5a and Supplementary Table 7). The 6 genes in which only one of the two alleles showed pheno­type are likely to be untagged and were not characterized further. Expressivity of the phenotypes of the 11 lines representing 8 genes (6 lines for 3 genes and 5 lines for 5 genes) varied among indi­ vidual plants within the homozygous population, ranging from delayed or failed germination, arrested or delayed development, anthocyanin accumulation, clear or white patches on the shoot to pale green shoot. As expected from known seed pigmentation mutants, survival rate in soil was <100% in most lines (Supplementary Table 8). To determine how these genes are associated with seed pigmentation, we examined linkages among the known and newly identified (3 sup­ ported by two alleles and 5 by one allele) genes (Fig. 5b). These genes Figure 4  AraNet correctly associates genes with many processes unique to plants, nonetheless relying at least in part on data from animals and yeast, which contribute evidence for linkages among genes that are broadly conserved but whose roles in Arabidopsis are in plant-specific processes. The performance at associating genes with each of 29 biological processes specific to plants (that is, annotated only with plant genes in GO and known to occur only in plants) is summarized as the area under a crossvalidated ROC curve (AUC). Even though these processes are absent in animals or fungi, the associated genes often have orthologs in these taxa, and AraNet draws upon data from these orthologs in making the associations. Each gray square demarks the support provided by a data set, measured as percentage of a sum of log likelihood scores contributing to that process, with darker gray indicating higher support. Data sets are labeled as in Figure 1.

152

form five network components, two belonging to photomorphogenesis and three to chloroplast development (Supplementary Table 9). The largest component includes members of the COP9 signalosome complex (CSN), an evolutionarily conserved post-translational regu­ latory complex involved in cell proliferation, response to DNA damage and gene expression34. In plants, the CSN complex is essential for photomorphogenesis35. Two of the three genes supported by two inde­ pendent alleles belong to this component, AT4G26430 and AT5G45620. AT4G26430 (CSN6B) encodes a subunit of CSN, CSN6. Using a single allele, CSN6B has been shown to be genetically redundant to another gene (CSN6A) under white light, though only partially redundant in dark and blue light36. AT5G45620 encodes a protein with sequence similarity to a subunit of the lid subcomplex of 26S proteasome, but its biological role is unknown13. Supporting evidence for its prediction comes from protein domain co-occurrence with FUS5, FUS6 and COP8, and among their human orthologs (Supplementary AT-CX AT-DC AT-GN AT-LC AT-PG CE-CC CE-CX CE-GT CE-LC CE-YH DM-PI HS-CX HS-DC HS-LC HS-MS HS-YH SC-CC SC-CX SC-DC SC-GT SC-LC SC-MS SC-TS SC-YH

© 2010 Nature America, Inc. All rights reserved.

b

R an

GO CC 1.0

R an

Area under ROC curve

a

Negative regulation of photomorphogenesis Glucosinolate biosynthetic process Cytokinin mediated signaling Negative regulation of abscisic acid mediated signaling Photosystem II assembly Photosynthesis, light reaction Ovule development Response to high light intensity Photosynthesis, light harvesting in photosystem II Chlorophyll biosynthetic process Regulation of seed germination Chloroplast fission Carpel development Regulation of stomatal movement Negative regulation of flower development Trichome differentiation Lignin biosynthetic process Phenylpropanoid biosynthetic process Response to water deprivation Trichome morphogenesis Photorespiration Phototropism Cold acclimation Jasmonic acid biosynthetic process Stomatal complex morphogenesis Abscisic acid mediated signaling Systemic acquired resistance Chlorophyll catabolic process Embryonic development ending in seed dormancy

0.5 0.6 0.7 0.8 0.9 1.0 AUC

0

100 Contribution (%)

VOLUME 28  NUMBER 2  FEBRUARY 2010  nature biotechnology

Table 5). The CSN and the lid subcomplex of 26S proteasome com­ plexes share structural and functional similarities 34, suggesting involvement of other protein degradation machineries in photo­ morphogenesis and early seedling development. The self-pollinated progeny of mutants that survived to make seeds did not show the seedling defects under standard growth conditions as their progeni­ tors (data not shown). However, when grown in dark or under blue light, the mutants showed slight (5–25%) but significant (P < 0.01, paired t-test, see Online Methods) differences in hypocotyl length (Fig. 5c,d). CSN6B mutants showed reduced hypocotyl length in dark but slightly increased hypocotyl length when grown under blue light compared to wild type. The other two mutants had longer hypocotyls than wild type under both dark and blue light conditions. All three genes have paralogs in the genome. As double mutants of CSN6A and CSN6B show a constitutive photomorphogenic phenotype, severe seedling dwarfism and high levels of anthocyanin accumulation36, it is possible that AT4G26430 and AT5G45620 may also exhibit genetic redundancy with their respective paralogs. The remaining hits from our screen link to each other and known seed pigmentation genes in three components relevant to thylakoid biogenesis and chlorophyll biosynthesis, processes affecting chloro­ plast development and function (Fig. 5b). Supplementary Table 9 details possible roles for the newly discovered genes. These results con­ firm that AraNet can efficiently associate new genes with a specific phenotypic trait.

previously uncharacterized genes. We selected three uncharacterized genes (AT1G80710, AT2G17900 and AT3G05090) based on several criteria: (i) no known biological process assigned; (ii) predicted by AraNet to be involved in developmental regulatory processes; and (iii) exist as single-copy genes. These represent extremely stringent tests of the network-based association method, and are all cases in which prediction based on sequence homology has failed. AraNet predicts GO biological process annotations, ordering pre­ dictions by the sum of the log likelihood scores linking a gene to genes already annotated by each term (Supplementary Table 10). For the three genes selected, we tested for morphological and physiological phenotypes in the top ten predicted processes. Two control genes, AT1G15772 and AT2G34170, were chosen randomly from genes lack­ ing AraNet functional predictions. Mutant plants were confirmed for homozygosity (Supplementary Table 11) and lack of detectable transcripts (data not shown). Self-pollinated progeny of homozygous plants were subjected to a bank of phenotypic assays based on the top ten predictions (see Online and Supplementary Methods). Of the three mutants, two exhibited phenotypes in the predicted processes. AT1G80710 is a regulator of drought sensitivity AraNet implicated the gene AT1G80710 in the response to water dep­ rivation, among other processes, drawing support from affinity puri­ fications of yeast orthologs (SC-MS)37,38 (Supplementary Table 10). This gene is expressed in all tissues examined, with highest abundance in flowers (Supplementary Fig. 5). We asked whether the ability to retain water differed in the mutants. Under drought, mutant plants retained ~80% of the water of wild type (P ≤ 0.001, unpaired t-test, Fig. 6a). Reduced water retention was not observed in control mutants (Supplementary Fig. 6).

Discovering functions for uncharacterized Arabidopsis genes Given that AraNet can successfully associate genes with traits of interest, we wished to test hypothesized roles for uncharacterized Arabidopsis genes in planta. AraNet predicts biological roles for 4,479

Wild type

AT5G45620-1

AT5G45620-2

AT4G26430-1

AT4G26430-2

AT5G50110-1

c

AT5G50110-2

AT4G AT4G AT5G AT5G AT5G Wild type 26430-1 26430-2 45620-1 45620-2 15610-1

Dark

a

CSN8

AT5G04110

CSN7

Known seed pigmentation genes

CHLG

New seed pigmentation genes (two alleles tested)

AT5G50110 CSN1

AT5G45620 COP1

AT4G26430

CHLI1

ALB3

AT3G20160

CHLD CSN4

AT5G15610 DET1

Linkage by plant-derived data CHLH

DXS

Linkage by animal-derived data Linkage by plant + animal data

Chloroplast development

d Difference in hypocotyl length (%)

Photomorphogenesis

AT5G11480

New seed pigmentation genes (one allele tested)

Blue light

b

30 Dark Blue light

25

*

*

20

*

15

*

nature biotechnology  VOLUME 28  NUMBER 2  FEBRUARY 2010

0-

1

02

G

15

61

62

AT 5

45

62

01

G

AT 5

45

AT 5

G

26 43 02

G

AT 4

G

26 43 01

Figure 5  Discovery of seed (embryo) pigmentation defective genes predicted by AraNet guilt by * 10 association. (a) Seedling pigmentation defects are apparent in each of two independent alleles for * 5 * the genes AT5G45620, AT4G26430 (CSN6B) and AT5G50110, all predicted based on network 0 connections to known pigmentation defect genes. (b) Eight new seed pigmentation defective genes –5 * * are organized into five network components by connections to known seed pigmentation genes, with – 10 evidence for the connections coming both from plant- and animal-derived data sets. (c) Mutants linked to known CSN genes show longer hypocotyl length than wild type in dark and under blue light, except CSN6B mutants, which show slightly shorter hypocotyls in dark. Scale bar, 1.3 cm. (d) Most of the differences in hypocotyl length of these mutants are slight (5–25%) but significant. Significant differences from wild type are indicated by asterisks (P < 0.01, paired t-test, n = 26 (dark), 32 (blue light)). n indicates the number of plates, each plate containing 7–8 plants of wild type and mutant genotype. Results are from seven (dark) or eight (blue light) independent experiments, plotting mean ± s.e.m. AT 4

© 2010 Nature America, Inc. All rights reserved.

resource

153

resource

0.6 0.5 0.4 0.3 0.2 0.1 0

Watered Drought

b 12

c 12

0 µM ABA 2.5 µM ABA 5 µM ABA 10 µM ABA

10 8 6 4 2 0

Wild type 0

5

10 15 Time (h)

20

e 14

0 µM ABA 2.5 µM ABA 5 µM ABA 10 µM ABA

10 8 6 4 2 0

Wild type Irs1-1

12 Number of LR

Wild type drs1-1

Water loss (% fresh weight)

0.7

Water loss (% fresh weight)

Relative water content (Fw – Dw)/(Tw – Dw)

a

10 8 6 4 2

drs1-1 0

5

10 15 Time (h)

20

0

Total LR

LR Emerged Elongated primordia LR LR before meristems emergence (stage VIII) (stages IV-VII)

Figure 6  Discovery of regulators of drought sensitivity and Irs1-1 Wild type lateral root development from previously uncharacterized Wild type Irs1-1 (pGWB2-LRS1) (pGWB2-LRS1) 1 nM lAA 10 nM lAA genes using AraNet. (a) Plants carrying a T-DNA insertion Wild type Irs1-1 Wild type Irs1-1 (drs1-1) in a previously uncharacterized gene, AT1G80710, retained significantly less water than wild type under drought. Relative water loss was calculated as (Fw − Dw)/(Tw − Dw) (Fw, fresh weight; Dw, dry weight; Tw, turgor weight). Significant differences were observed between the relative water loss of wild type and mutant plants (P ≤ 0.001, unpaired t-test, n = 15) and between watered and drought conditions of the same genotype (P ≤ 0.0001, unpaired t-test, n = 15). (b,c) Transpiration was reduced in wild type plants in the presence of abscisic acid (ABA) in a dosage dependent manner (b), whereas mutant plants were insensitive to ABA (c). (d) The number of lateral roots is strongly reduced in lines carrying a T-DNA insertion (lrs1-1) in another previously uncharacterized gene AT3G05090. This phenotype can be complemented by reintroduction of the functional gene. When additional copies of the gene are expressed in a wild type strain, lateral roots increase, whereas the primary root decreases in length. Six other independent transformants in each background gave similar phenotypes (data not shown). 1 nM Auxin (IAA) increases the number and length of lateral roots in both the wild-type and mutant seedlings. Contrarily, 10 nM IAA severely reduces the primary root length in both genotypes. Scale bar, 1.4 cm. (e) Different stages of the lateral root formation are affected in the lrs1-1 mutant. Wild-type lateral roots are distributed fairly evenly among lateral root primordia, emerged lateral root and elongated lateral root. The mutant has reduced numbers of the lateral roots in all of these stages, though the reduction is more severe in the emerged and elongated lateral root than that in the lateral root primordia. LR, lateral root. Error bars indicate s.e.m.

© 2010 Nature America, Inc. All rights reserved.

d

Drought response is mediated by several signaling pathways in Arabidopsis, including the hormone abscisic acid (ABA), transcription factor DREB2A, ERD1 and E3 SUMO ligase SIZ139,40. To determine whether the reduced water retention upon drought stress is mediated by ABA, we examined the effect of ABA on transpiration of detached leaves. The mutant was insensitive to ABA on water loss, whereas the wild type lost significantly less water in the presence of 10 µM ABA (P ≤ 0.0004, unpaired t-test, Fig. 6b,c). At this ABA concentration, mutant leaves lost 30% more water than wild type (P ≤ 0.04, unpaired t-test, Fig. 6b,c). ABA showed no effect on germination rate in the mutant (data not shown), indicating that not all ABA-mediated proc­ esses are affected in the mutant. Both the water retention and ABA-insensitive water transpiration response segregated as a single recessive Mendelian locus and linked to the T-DNA insertion (Supplementary Table 12 and Supplementary Fig. 7). We designate AT1G80710 as DRS1 (DROUGHT SENSITIVE 1) and the T-DNA insertion allele (Salk_001238C) as drs1-1. An independ­ ent T-DNA allele (Salk_149366C) that we designate as drs1-2 exhibited the same phenotypes in relative water content after drought and ABAinsensitive water transpiration (Supplementary Fig. 8), confirming that the phenotypes are linked to mutations in DRS1. DRS1 is a WD-40 repeat family protein containing a DWD (DDB1 binding WD-40) motif 41. Some DWD-containing proteins are substrate receptors for DDB1-Cul4 ubiquitin ligase machinery in humans, yeast and Arabidopsis41,42. Combination of AraNet prediction and experimental testing thus demonstrates that DRS1 promotes tolerance to drought stress, possibly mediated by ABA, and suggests involvement of DDB1Cul4–mediated protein degradation in drought response. Given that the a priori odds of selecting a gene affecting the response to water deprivation are ~1 in 318 (currently only 85 of 27,029 Arabidopsis

154

genes are annotated for response to water deprivation), these tests strongly support the network-based approach to rationally associate even entirely uncharacterized genes with plant traits. AT3G05090 is a regulator of lateral root development The second candidate gene, AT3G05090, was implicated in cell pro­ liferation and meristem organization, drawing support from phylo­ genetic profiling of bacterial homologs of Arabidopsis proteins and domain co-occurrence patterns of yeast orthologs (Supplementary Table 10). We examined both shoot and root development in at3g05090-1 seedlings. We did not observe shoot phenotypes, but the number of lateral roots was significantly reduced (P ≤ 10−37, unpaired t-test, Fig. 6d,e and Supplementary Fig. 9). This pheno­ type segregated as a single recessive Mendelian locus linked to the T-DNA insertion (Supplementary Table 12 and Supplementary Fig. 10a). The length of the primary root was shorter than in wild type (Fig. 6d) but this phenotype was unlinked to the T-DNA insertion (Supplementary Fig. 10b), showing that the lateral root phenotype is separable and independent from the primary root phenotype. We designate AT3G05090 as LRS1 (LATERAL ROOT STIMULATOR 1) and the at3g05090-1 allele as lrs1-1. Homozygous lines transformed with a wild-type coding sequence driven under a 35S CaM virus promoter complemented the lateral root phenotype (Fig. 6d and Supplementary Fig. 9). To determine if the lateral root formation is blocked before the lateral root meristem emergence, we examined the number of lateral root primordia and meristems (lateral root stages IV–VIII43). Wild-type lateral roots are distributed fairly evenly among lateral root primordia, emerged lateral root and elongated lateral root (Fig. 6e). The mutant has reduced numbers of lateral roots at all of these stages, though the reduction is more severe in the

VOLUME 28  NUMBER 2  FEBRUARY 2010  nature biotechnology

© 2010 Nature America, Inc. All rights reserved.

resource emerged and elongated lateral root than in the lateral root primordia (Fig. 6e). Transforming wild-type lines with the 35S::LRS1 construct did not increase the number of lateral roots, but we observed a dra­ matic increase in the length of the lateral roots and a decrease in the primary root length (Fig. 6d and Supplementary Fig. 9). Regulation of root architecture and function, modulated by both intrinsic and extrinsic signals, is critical for efficient nutrient and water use for plants. Auxin, a plant hormone, is a key regulator for lateral root development, including lateral root initiation, primor­ dium development and emergence44. The reduction in number of lateral roots in the mutant and the increase in lateral root length concomitant with the decrease in the primary root length in the over­ expressed lines evoke defects in auxin accumulation or perception44. We thus asked whether exogenous auxin could alleviate the phenotype by growing plants in the presence of native auxin, indole acetic acid (IAA). IAA increased the lateral root number in the mutant (Fig. 6d and Supplementary Fig. 11a,b), demonstrating that auxin perception was not altered in the mutant and suggesting that auxin accumulation is compromised in the mutant. Auxin accumulation can be altered by changing synthesis, degradation, sequestration or transport45. To test for auxin transport defects, we examined effects of an auxin trans­ port inhibitor, N-(1-naphthyl)phthalamic acid (NPA) on root growth. NPA decreases both the number and length of lateral roots in both genotypes (Supplementary Fig. 11a,b). LRS1 encodes another DCAF protein41, suggesting involvement of DDB1-Cul4–mediated protein degradation in lateral root development. These results demonstrate that the lrs1-1 mutant is defective in lateral root development and sug­ gest roles for DDB1-Cul4–mediated protein degradation in regulating auxin accumulation during lateral root primordium development and lateral root meristem emergence, consistent with its hypothesized roles in cell proliferation and meri­stem organization. DISCUSSION We demonstrate here that genes can be rationally associated with plant traits through guilt by association in a gene network. For this pur­ pose, we created AraNet, a genome-wide gene network for A. thaliana, a reference organism for flowering plants, including many crops. AraNet is the most extensive gene network for any plant thus far; gene annotations derived by network guilt by association extend substan­ tially beyond current gene annotations. We validated the network’s predictive power by cross-validation tests, independent pathway and phenotype data sets, cell-specific expression data sets, and by experi­ ments on computationally selected candidate genes. AraNet generates at least two main types of testable hypotheses. The first type uses a set of genes known to be involved in a specific process as bait to find new genes involved in that process. This test is useful if the bait genes are well connected (that is, high AUC). We used the set of genes conferring seed pigmentation defects (AUC = 0.68) as bait and found a tenfold enrichment in identifying mutants with comparable phenotypes. Of the 318 GO biological processes with more than five genes, ~43% have AUCs of at least 0.68 (Supplementary Table 13), sug­ gesting that AraNet will be useful in identifying new genes in nearly half of these biological processes. Similar distributions of AUCs are found for GO cellular component terms and KEGG pathways (Supplementary Tables 14 and 15). In practice, this translates into identifying a small set of new genes from a relatively limited-scale screen of the top network-predicted candidates (e.g., computer simulations suggest finding an average of 4–7 novel genes from tests of the top 200 candi­ dates for biological processes with AUC >0.6; Supplementary Fig. 12). The second type of hypothesis involves predicting functions for uncharacterized genes. We assayed predicted phenotypes for three

nature biotechnology  VOLUME 28  NUMBER 2  FEBRUARY 2010

uncharacterized genes, two of which showed phenotypes in the pre­ dicted processes, response to drought and meristem development. There are 4,479 uncharacterized genes in AraNet (16% of protein-coding genes) with links to characterized genes, suggesting broad utility for AraNet in identifying candidate functions. Both of these modes of operation can be easily performed on the AraNet website. Although AraNet currently shows high accuracy for many processes (Figs. 2–4), there are nonetheless specific processes that are poorly represented, with this trend stronger among plant-specific processes (Fig. 2d). This trend manifested in our experimental validation of only two of three tested candidate genes, although these intentionally rep­ resented challenging cases lacking any current functional annotation and for which sequence homology approaches had failed. Although we observed that non-plant–derived data sets helped identify genes for plant-specific processes, it is clear that more plant data sets will enhance the utility of gene networks for finding trait-relevant genes. Three major causes underlie such cases of poor predictive ­performance. First, our current knowledge of genetic factors for a process may be so sparse that AraNet cannot link them efficiently. Second, AraNet may lack linkages or data relevant to the poorly predicted processes. These two trends likely explain the lower performance among plant-specific pro­cesses relative to more broadly studied, evolutionarily conserved ­pro­cesses. Additional plant-specific data sets, such as protein interactions, should help here, as should considering both indirect and direct network linkages for ranking candidates. Third, strongly implicated candidate genes that nonetheless test negative for a trait, resulting in apparent false positives, might be masked by epistatic effects, thus actually representing true predictions and false-negative assay results. This trend may be ­reasonably common and has been previously observed in yeast46. AraNet represents a step toward the goal of computationally iden­ tifying gene-trait associations in plants. This work suggests that gene networks for food and energy crops will facilitate manipulation of traits of economic importance and crop genetic engineering. Methods Methods and any associated references are available in the online version of the paper at http://www.nature.com/naturebiotechnology/. Note: Supplementary information is available on the Nature Biotechnology website. Acknowledgments We are grateful to M. Ahn, A. Noorani and V. Bakshi for technical assistance, J. Shin for assistance on AraNet web design, T. Nakagawa (Shimane University, Japan) for providing pGWB2, K. Barton for providing lab space and D. Meinke, M. Running, W. Briggs, Z. Wang and K. Dreher for helpful discussions. This work was supported by Carnegie Institution for Science (B.A., S.Y.R.), a grant from the National Science Foundation (MCB-0520140) to S.Y.R. and by the National Research Foundation of Korea (NRF) grant funded by the Korean government (no. 2009-0063342, 20090070968) and Yonsei University (no. 2008-7-0284, 2008-1-0018) to I.L. and from the National Science Foundation, National Institutes of Health, and Welch (F1515) and Packard Foundations to E.M.M. AUTHOR CONRIBUTIONS I.L. and S.Y.R. conceived the project, I.L. created AraNet using approaches developed with E.M.M., B.A. performed the experimental tests, P.T. assisted in seed pigmentation mutant analysis, S.Y.R. supervised the experimental tests, I.L., E.M.M. and S.Y.R. analyzed AraNet and wrote the manuscript. COMPETING INTERESTS STATEMENT The authors declare no competing financial interests. Published online at http://www.nature.com/naturebiotechnology/. Reprints and permissions information is available online at http://npg.nature.com/ reprintsandpermissions/. 1. Alonso, J.M. et al. Genome-wide insertional mutagenesis of Arabidopsis thaliana. Science 301, 653–657 (2003).

155

© 2010 Nature America, Inc. All rights reserved.

resource 2. Marcotte, E.M., Pellegrini, M., Thompson, M.J., Yeates, T.O. & Eisenberg, D. A combined algorithm for genome-wide prediction of protein function. Nature 402, 83–86 (1999). 3. McGary, K.L., Lee, I. & Marcotte, E.M. Broad network-based predictability of Saccharomyces cerevisiae gene loss-of-function phenotypes. Genome Biol. 8, R258 (2007). 4. Fraser, H.B. & Plotkin, J.B. Using protein complexes to predict phenotypic effects of gene mutation. Genome Biol. 8, R252 (2007). 5. Lee, I. et al. A single gene network accurately predicts phenotypic effects of gene perturbation in Caenorhabditis elegans. Nat. Genet. 40, 181–188 (2008). 6. Zhong, W. & Sternberg, P.W. Genome-wide prediction of C. elegans genetic interactions. Science 311, 1481–1484 (2006). 7. Lage, K. et al. A human phenome-interactome network of protein complexes implicated in genetic disorders. Nat. Biotechnol. 25, 309–316 (2007). 8. Franke, L. et al. Reconstruction of a functional human gene network, with an application for prioritizing positional candidate genes. Am. J. Hum. Genet. 78, 1011–1025 (2006). 9. Linghu, B., Snitkin, E.S., Hu, Z., Xia, Y. & Delisi, C. Genome-wide prioritization of disease genes and identification of disease-disease associations from an integrated human functional linkage network. Genome Biol. 10, R91 (2009). 10. Huttenhower, C. et al. Exploring the human genome with functional maps. Genome Res. 19, 1093–1106 (2009). 11. Hermjakob, H. et al. IntAct: an open source molecular interaction database. Nucleic Acids Res. 32, D452–D455 (2004). 12. Alfarano, C. et al. The Biomolecular Interaction Network Database and related tools 2005 update. Nucleic Acids Res. 33, D418–D424 (2005). 13. Swarbreck, D. et al. The Arabidopsis Information Resource (TAIR): gene structure and function annotation. Nucleic Acids Res. 36, D1009–D1014 (2008). 14. de Folter, S. et al. Comprehensive interaction map of the Arabidopsis MADS Box transcription factors. Plant Cell 17, 1424–1433 (2005). 15. Huynen, M., Snel, B., Lathe, W. III & Bork, P. Predicting protein function by genomic context: quantitative evaluation and qualitative inferences. Genome Res. 10, 1204–1210 (2000). 16. Pellegrini, M., Marcotte, E.M., Thompson, M.J., Eisenberg, D. & Yeates, T.O. Assigning protein functions by comparative genome analysis: protein phylogenetic profiles. Proc. Natl. Acad. Sci. USA 96, 4285–4288 (1999). 17. Wolf, Y.I., Rogozin, I.B., Kondrashov, A.S. & Koonin, E.V. Genome alignment, evolution of prokaryotic genome organization, and prediction of gene function using genomic context. Genome Res. 11, 356–372 (2001). 18. Bowers, P.M. et al. Prolinks: a database of protein functional linkages derived from coevolution. Genome Biol. 5, R35 (2004). 19. Dandekar, T., Snel, B., Huynen, M. & Bork, P. Conservation of gene order: a fingerprint of proteins that physically interact. Trends Biochem. Sci. 23, 324–328 (1998). 20. Overbeek, R., Fonstein, M., D’Souza, M., Pusch, G.D. & Maltsev, N. The use of gene clusters to infer functional coupling. Proc. Natl. Acad. Sci. USA 96, 2896–2901 (1999). 21. Lee, I., Li, Z. & Marcotte, E.M. An improved, bias-reduced probabilistic functional gene network of baker’s yeast, Saccharomyces cerevisiae. PLoS ONE 2, e988 (2007). 22. Breitkreutz, B.J. et al. The BioGRID Interaction Database: 2008 update. Nucleic Acids Res. 36, D637–D640 (2007). 23. Chatr-aryamontri, A. et al. MINT: the Molecular INTeraction database. Nucleic Acids Res. 35, D572–D574 (2007).

156

24. Giot, L. et al. A protein interaction map of Drosophila melanogaster. Science 302, 1727–1736 (2003). 25. Remm, M., Storm, C.E. & Sonnhammer, E.L. Automatic clustering of orthologs and in-paralogs from pairwise species comparisons. J. Mol. Biol. 314, 1041–1052 (2001). 26. Ogata, H. et al. KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res. 27, 29–34 (1999). 27. Cui, J. et al. AtPID: Arabidopsis thaliana protein interactome database an integrative platform for plant systems biology. Nucleic Acids Res 36, D999–D1008 (2007). 28. Geisler-Lee, J. et al. A predicted interactome for Arabidopsis. Plant Physiol. 145, 317–329 (2007). 29. Gutierrez, R.A. et al. Qualitative network models and genome-wide expression data define carbon/nitrogen-responsive molecular machines in Arabidopsis. Genome Biol. 8, R7 (2007). 30. Ma, S., Gong, Q. & Bohnert, H.J. An Arabidopsis gene network based on the graphical Gaussian model. Genome Res. 17, 1614–1625 (2007). 31. Brady, S.M. et al. A high-resolution root spatiotemporal map reveals dominant expression patterns. Science 318, 801–806 (2007). 32. Meinke, D., Muralla, R., Sweeney, C. & Dickerman, A. Identifying essential genes in Arabidopsis thaliana. Trends Plant Sci. 13, 483–491 (2008). 33. McElver, J. et al. Insertional mutagenesis of genes required for seed development in Arabidopsis thaliana. Genetics 159, 1751–1763 (2001). 34. Wei, N., Serino, G. & Deng, X.W. The COP9 signalosome: more than a protease. Trends Biochem. Sci. 33, 592–600 (2008). 35. Peng, Z., Serino, G. & Deng, X.W. Molecular characterization of subunit 6 of the COP9 signalosome and its role in multifaceted developmental processes in Arabidopsis. Plant Cell 13, 2393–2407 (2001). 36. Gusmaroli, G., Figueroa, P., Serino, G. & Deng, X.W. Role of the MPN subunits in COP9 signalosome assembly and activity, and their regulatory interaction with Arabidopsis Cullin3-based E3 ligases. Plant Cell 19, 564–581 (2007). 37. Gavin, A.C. et al. Proteome survey reveals modularity of the yeast cell machinery. Nature 440, 631–636 (2006). 38. Krogan, N.J. et al. Global landscape of protein complexes in the yeast Saccharomyces cerevisiae. Nature 440, 637–643 (2006). 39. Catala, R. et al. The Arabidopsis E3 SUMO ligase SIZ1 regulates plant growth and drought responses. Plant Cell 19, 2952–2966 (2007). 40. Shinozaki, K. & Yamaguchi-Shinozaki, K. Gene networks involved in drought stress response and tolerance. J. Exp. Bot. 58, 221–227 (2007). 41. Lee, J.H. et al. Characterization of Arabidopsis and rice DWD proteins and their roles as substrate receptors for CUL4-RING E3 ubiquitin ligases. Plant Cell 20, 152–167 (2008). 42. Jin, J., Arias, E.E., Chen, J., Harper, J.W. & Walter, J.C. A family of diverse Cul4Ddb1-interacting proteins includes Cdt2, which is required for S phase destruction of the replication factor Cdt1. Mol. Cell 23, 709–721 (2006). 43. Casimiro, I. et al. Dissecting Arabidopsis lateral root development. Trends Plant Sci. 8, 165–171 (2003). 44. Fukaki, H., Okushima, Y. & Tasaka, M. Auxin-mediated lateral root formation in higher plants. Int. Rev. Cytol. 256, 111–137 (2007). 45. Vanneste, S. & Friml, J. Auxin: a trigger for change in plant development. Cell 136, 1005–1016 (2009). 46. Li, Z. et al. Rational extension of the ribosome biogenesis pathway using networkguided genetics. PLoS Biol. 7, e1000213 (2009).

VOLUME 28  NUMBER 2  FEBRUARY 2010  nature biotechnology

ONLINE METHODS

© 2010 Nature America, Inc. All rights reserved.

All analyses are based on the set of 27,029 predicted protein coding loci of Arabidopsis (genome release version TAIR7)13. Reference and benchmark sets, raw data sets, and the construction and computational validation of AraNet are described in Supplementary Methods (Supplementary Figs. 16–18). Targeted reverse genetics screening for seed pigmentation mutants. We searched AraNet with 23 confirmed seed pigmentation mutants (Supplementary Table 4) from the SeedGenes database32 as bait. We retrieved the homozygous T-DNA insertional lines for the top 200 candidates from the SIGnAL database1 and obtained the stocks from the Arabidopsis Biological Resource Center (Supplementary Tables 5 and 6). Seven to 9 seeds for each line were sterilized as described below (Plant material). Seeds were stratified at 4 °C for 2 d in the dark and grown in Murashige and Skoog (MS) medium with 1% sucrose under continuous illumination of 50–80 µmol/m2 s at 22 °C. Seedlings were observed under a dissecting microscope (Leica MZ125) 6 d after germination and followed up 10–12 d after germination. For each of the lines where sufficient seeds were available, the assay was conducted at least twice. T-DNA insertions and genotypes of the seeds for the 11 lines with the mutant phenotypes described in this study (6 lines representing two alleles of three genes and 5 lines representing single alleles of five genes) were confirmed using PCR (Supplementary Methods and Supplementary Table 16). To determine the significance of the discovery rate, we used the results of a large-scale screening of T-DNA insertional mutants for embryo defective or seed pigmentation mutants32 as the background rate. This study found ~1,260 seed pigmentation mutants from screening 120,000 T-DNA lines. Because this was a forward-genetics screening whereas our screen was reverse-genetic (that is, preselected for intragenic insertions), we adjusted the total number of lines in the background to 84,000 based on the genome-wide distribution of T-DNA insertion sites of 70% insertion events in intragenic regions1. Candidate selection for uncharacterized genes and experimental validation of mutant phenotypes. To test the predictive power of AraNet in planta, we analyzed mutant phenotypes of genes of unknown function, whose biologi­ cal roles were inferred by the annotations of the neighbors of these genes in AraNet. Of the 27,029 protein-encoding genes in Arabidopsis, 14,847 have no information about the biological processes in which they are involved. More than half of these uncharacterized genes (7,465 genes) are included in AraNet. Of these, 4,479 genes are inferred to be associated with specific biological processes based upon annotated AraNet neighbors (using only IDA, IMP, IGI, IPI, IEP and TAS evidence). To test the accuracy of such inferences made by AraNet, we chose three genes to characterize experimentally. These were cho­ sen on the basis of available homozygous knockout lines, absence of paralogs and AraNet inferences of involvement in specific biological processes. The genes chosen were AT3G05090, AT1G80710 and AT2G17900, whose top ten AraNet predictions are shown in Supplementary Table 10. From the predic­ tions, we assayed all of the phenotypes that could be measured with available resources at Carnegie. For AT1G80710, the following were tested: response to water deprivation (rank 3); trichome differentiation (rank 8); leaf development (rank 9). For AT3G05090, the following phenotypes were tested: trichome differentiation (rank 1); leaf development (rank 3); cell proliferation (rank 5); meristem organization (rank 8); and regulation of flower development (rank 10). For AT2G17900, the following were tested: meristem organization (rank 2); leaf morphogenesis (rank 3); hyperosmotic salinity response (rank 4); brassinosteroid mediated signaling (rank 5); multidimensional cell growth (rank 6); response to auxin stimulus (rank 7); detection of brassinosteroid stimulus (rank 8). In addition, we selected two genes randomly, AT1G15772 and AT2G34170, which were included in AraNet but were neighbors of other uncharacterized genes, to test specificity of all observed phenotypes. Plant material. Seeds of homozygous T-DNA knockout mutants were obtained from the Arabidopsis Biological Resource Center. The stock numbers for the 118 seed pigmentation candidate genes are listed in Supplementary Table 6. Seeds for the five uncharacterized genes were SALK_059570C (AT3G05090), SALK_001238C (AT1G80710), SALK_127952C (AT2G17900), Salk_118634C (AT1G15772) and Salk_099804C (AT2G34170). For experiments conducted in soil, seeds were sown in soil (Premier Pro-mix) supplemented with fertilizer

doi:10.1038/nbt.1603

(Osmocote Classic, Hummert International). Seeds were stratified at 4 °C in the dark for 2 d and grown under 16/8 h of light/dark (90–100 µmol/m2 s) and 30% humidity at 22 °C. For experiments conducted in agar plates, seeds were surface sterilized with 15% commercial bleach (6.25% sodium hypochlorite), containing a few drops of Tween-20 detergent and rinsed with sterile water five times. Seeds were sown on agar plates containing 0.43% MS salts, 0.5% 2-(N-Morpholino)ethanesulfonic acid (MES), 0.5% sucrose, 0.8% agar, pH 5.7. Plants on agar plates were grown under constant illumination of 50–80 µmol/m2 s at 22 °C. For root assays, 50 ml of the MS medium was prepared poured into 100 × 100 × 15 mm square plates (Fisher Scientific) 1 d before planting to minimize plate-to-plate variability. T-DNA insertions were confirmed, and genetic linkage, complementation, and overexpression tests were performed as described in Supplementary Methods (Supplementary Tables 16 and 17). Visible phenotype assays. The following traits were observed by naked eye and using dissecting (Leica MZ125) and compound (Nikon Eclipse E600 with Nomarski optics) microscopes throughout the life cycle of the mutant plants: trichome differentiation (observations made on rosette and cauline leaves and sepals), leaf development and morphogenesis, cell proliferation, meristem organization and multidimensional cell growth (observations made on leaf, floral, inflorescence and root organs). To detect phenotypes in the regulation of flower development, we observed floral organs and flowering time under long days (16/8 of light/dark) and short days (8/16 of light/dark). Hypocotyl length measurements. Seeds were germinated and grown verti­ cally for 4 d in dark or under 4 µM/m2 s continuous blue light. Seven to eight seeds of mutant and wild-type Col-0 (a reference strain that is the genetic background for the T-DNA mutants) were planted per plate and two plates per genotype were tested in each experiment. Each condition was tested in 7–8 independent experiments. Hypocotyl length was measured using ImageJ on photographs of the plates after 4 d of growth. The average hypocotyl length of each genotype was determined from each plate and the difference in hypocotyl length between wild type and mutant was determined using one-tailed, paired t-test. Root length and number measurements. Seeds were germinated and grown vertically. Root measurements were taken 10–11 d after germination. Lateral roots were counted using a dissecting microscope or from digital images of plants using ImageJ. Different stages of the lateral roots were determined using a compound microscope. The root length was measured by tracing the length of the root using ImageJ on digital images of the seedlings. Auxin response. Auxin and auxin transport inhibitor treatments were car­ ried out as described47. Seeds were sown on MS agar medium and grown under continuous light. After 4 d, seedlings were transferred to either MS agar medium (control) or MS agar medium containing 1 nM, 5 nM, 10 nM or 30 nM indole acetic acid (IAA, Sigma) or 1 nM, 10 nM, 100 nM or 1 µM of naphthyl­ phthalamic acid (NPA, Chem Service). Both wild type (Col-0) and mutant seedlings were transferred to the same plates. On the tenth or eleventh day of growth, the primary root length, number of lateral roots and length of lateral roots were measured as described above. Significant differences were deter­ mined by unpaired t-test. Each experiment was conducted with 2–3 plates of 7–10 plants each of wild type and mutant per plate. At least three independent experiments were carried out for each hormone assay. Abscisic acid (ABA) response. The effect of ABA on detached leaf transpi­ ration was determined as described48 with some modifications: plants were grown in soil under long-day conditions (16/8 h light/dark) under white light of 90–100 µM m−2 s−1 at 22 °C. The largest, fully-open rosette leaves of 4–week-old plants were excised at the bottom of the petioles and were placed into a Parafilm-sealed 1.5 ml centrifuge tubes containing 1.4 ml of 0, 2.5 µM, 5 µM and 10 µM of ABA in an artificial xylem sap solution (15 mM KNO3, 1 mM CaCI2, 0.7 mM MgSO4, and 1 mM (NH4)2HPO4, with pH adjusted to 5.0 with 1 M phosphoric acid49). Transpiration was measured by weigh­ ing total weight of the tubes at times 0, 2, 4, 6 and 22 h. All of the excisions took place between 10 am and noon (4–6 h after the onset of illumination).

nature biotechnology

For the F2 linkage test of drs1-1, four leaves were excised from each plant and two were treated with the sap solution only and the other two were treated with 10 µM of ABA in the sap solution. Details on the linkage test are found in Supplementary Methods. For wild-type and mutant comparisons, each experiment used two to four leaves from three to four plants per genotype at each time point and was conducted in triplicate. Four independent experiments were conducted.

47. Cho, H.T. & Cosgrove, D.J. Regulation of root hair initiation and expansin gene expression in Arabidopsis. Plant Cell 14, 3237–3253 (2002). 48. Munns, R. & King, R.W. Abscisic acid is not the only stomatal inhibitor in the transpiration stream of wheat plants. Plant Physiol. 88, 703–708 (1988). 49. Goodger, J.Q., Sharp, R.E., Marsh, E.L. & Schachtman, D.P. Relationships between xylem sap constituents and leaf conductance of well-watered and water-stressed maize across three xylem sap sampling techniques. J. Exp. Bot. 56, 2389–2400 (2005). 50. Giraud, E. et al. The absence of ALTERNATIVE OXIDASE1a in Arabidopsis results in acute sensitivity to combined light and drought stress. Plant Physiol. 147, 595–610 (2008).

© 2010 Nature America, Inc. All rights reserved.

Drought response assay. Response to water deprivation was determined by measuring relative water content as described50. Plants were grown in soil under long-day conditions (16/8 h light/dark) under white light of 90–100 µM m−2 s−1 at 22 °C for 4–5 weeks. Watering was stopped for the drought treatment and relative water content was measured on day 0, 4, 7 and 10 of droughting. Control plants were watered every 2–3 d. To measure relative water content, we excised plants at the shoot/root junction, removed any bolts and weighed rosettes to determine the fresh weight (Fw). The rosettes were then completely submerged in water for 4 h and weighed to determine the turgid weight (Tw). Rosettes were then dried overnight at

80 °C and weighed to obtain the dry weight (Dw). Three plants from each genotype for each condition were measured. Relative water content was calculated as (Fw − Dw)/(Tw − Dw) and the significance of differences was determined by unpaired t-test. Three plants of each genotype were used for each time point per experiment. Four independent experiments were conducted. An interactive web tool for AraNet-based candidate gene selection is avail­ able at http://www.functionalnet.org/aranet/.

nature biotechnology

doi:10.1038/nbt.1603

B r i e f c o m m u n i c at i o n s

Improved affinity for the neonatal Fc receptor (FcRn) is known to extend antibody half-life in vivo. However, this has never been linked with enhanced therapeutic efficacy. We tested whether antibodies with half-lives extended up to fivefold in human (h)FcRn transgenic mice and threefold in cynomolgus monkeys retain efficacy at longer dosing intervals. We observed that prolonged exposure due to FcRn-mediated enhancement of half-life improved antitumor activity of Fc-engineered antibodies in an hFcRn/Rag1–/– mouse model. This bridges the demand for dosing convenience with the clinical necessity of maintaining efficacy. The well-established role of FcRn in IgG turnover has been the foundation for Fc engineering efforts aimed at improving the pharmaco­ kinetic properties of therapeutic antibodies1,2. Despite contrary results about the relationship between FcRn affinity and half-life 3,4, several such efforts at pharmacokinetic engineering in nonhuman primates, whose FcRn is similar to that of humans, have demonstrated that engineered antibody variants have a prolonged half-life5–8. Yet, although the successful extension of half-life in pharmacokinetic experiments bodes well for the prospect of improving clinical dosing, a critical gap remains. For half-life extension technologies to be of practical use, efficacy of a biotherapeutic with longer half-life must be preserved at longer dosing intervals. Although the relationship between drug exposure and efficacy is well-established, this correlation has not thus far been established for antibodies engineered for longer half-life. We coupled rational design methods with high-throughput protein screening to engineer a series of Fc variants with greater affinity for human FcRn. Variants were constructed in the context of the humanized anti-vascular endothelial growth factor (VEGF) IgG1 antibody bevacizumab9 (Avastin), which is currently approved for the treatment of colorectal, lung, breast and renal cancers. A description of the construction, production and binding studies of the antibodies is provided in Supplementary Methods. As FcRn binds IgG at the lower pH of the early endosome (pH 6.0–6.5) but not at the higher pH of blood (pH 7.4), we used Biacore to screen antibodies for binding to human FcRn at pH 6.0. Our engineered variants demonstrated between 3- and 20-fold greater binding to FcRn at pH 6.0, with ­improvements

a 100

Xtend-VEGF IgG1 Bevacizumab

10 1 0.1 0 10 20 30 40 50 60 70 80 90 Days

b Concentration (µg/ml)

© 2010 Nature America, Inc. All rights reserved.

Jonathan Zalevsky1,3, Aaron K Chamberlain1, Holly M Horton1, Sher Karki1, Irene W L Leung1, Thomas J Sproule2, Greg A Lazar1, Derry C Roopenian2 & John R Desjarlais1

due almost exclusively to slower off-rate (koff ) (Supplementary Fig. 1 and Supplementary Table 1). A lead variant, M428L/N434S, sub­ sequently selected principally based on its pharmacokinetic performance (see below), provided an 11-fold improvement in FcRn affinity at pH 6.0. We refer to this double substitution in the context of bevacizumab as Xtend-VEGF. Details of a pharmacokinetic study in cynomolgus monkeys (Macaca fascicularis) to evaluate the capacity of the variants to improve serum half-life are provided in the Supplementary Methods. Binding improvements of the variants to monkey FcRn at pH 6.0 were comparable to improvements for human FcRn, and the rank order of the variants in FcRn affinity was the same (data not shown). When three monkeys per group were injected intravenously with 4 mg/kg variant or native IgG1 anti-VEGF antibody, we observed a large improvement in half-life for the variants relative to native IgG1 (Supplementary Fig. 2a). Fitted parameters for the full set of variants (Supplementary Table 2) indicated increases in β-phase half-life, area under the curve (AUC) measurements and the rate of antibody clearance from serum. The observed 9.7-d half-life for native IgG1 bevacizumab agrees with the published value (9.3 d) for a slightly lower (2 mg/kg) dose10. Among the engineered antibodies that were tested, the Xtend double variant performed best (Fig. 1a). It prolonged halflife from 9.7 to 31.1 d, a 3.2-fold improvement in serum half-life relative to native IgG1 (Supplementary Table 2). Simple allometric scaling extrapolations suggest that such improvement can potentially translate into human half-lives >50 d. We then sought to further challenge the applicability of pharmacokinetic engineering by targeting an internalizing cell-surface antigen that potentially provides a competing sink for antibody clearance.

Concentration (µg/ml)

Enhanced antibody half-life improves in vivo activity

Xtend-EGFR IgG1 Cetuximab

100

10

1

0

5

10 Days

15

20

Figure 1  Increasing antibody affinity to FcRn promotes half-life extension in cynomolgus monkeys. (a) Log-linear changes in serum concentrations for anti-VEGF (bevacizumab) antibodies in cynomolgus monkeys. All antibodies were administered by single 60-min intravenous infusion at 4 mg/kg and serum antibody concentrations were determined using a VEGF antigendown immunoassay. Results are means ± s.e.m. (n = 2 for bevacizumab and n = 3 for variants). (b) Log-linear changes in serum concentrations for anti-EGFR antibodies in cynomolgus monkeys. Monoclonal antibodies were administered by single 30-min intravenous infusion at 7.5 mg/kg and serum antibody concentrations were determined using an EGFR antigen-down immunoassay. Results are means (n = 2 animals per test article).

1Xencor, Inc., Monrovia, California, USA. 2The Jackson Laboratory, Bar Harbor, Maine, USA. 3Present address: Takeda San Diego, Inc., San Diego, California, USA. Correspondence should be addressed to J.R.D. ([email protected]).

Received 20 October 2009; accepted 14 December 2009; published online 17 January 2010; doi:10.1038/nbt.1601

nature biotechnology  volume 28  number 2  February 2010

157

b r i e f c o m m u n i c at i o n s b

10 1 Xtend-VEGF IgG1 Bevacizumab

0.1 0

5

Mean tumor volume (mm3)

c

10

15 Days

20

10 1 0.1 0.01

25

Xtend-EGFR IgG1 Cetuximab 0

800

*

600 400 200 35

45

55

65

5

10

d

Vehicle IgG1 Bevacizumab Xtend-VEGF

1,000

0

100

1,000

1,200

Day

© 2010 Nature America, Inc. All rights reserved.

Concentration (µg/ml)

100

Mean tumor volume (mm3)

Concentration (µg/ml)

a

75

85

15 Days

20

25

Vehicle IgG1 Cetuximab Xtend-EGFR

800 600

*

400 200 0

10

20

30

40

Day

Figure 2  Improved antibody half-life translates to greater in vivo efficacy. (a) Log-linear changes in serum concentrations of anti-VEGF antibodies in hFcRn mice. All antibodies were administered via single intravenous bolus at 2 mg/kg, and serum antibody concentrations were determined using a human immunoglobulin recognition immunoassay. Results are means ± standard errors (n = 6). For some data points, errors are smaller than can be indicated. (b) Log-linear changes in serum concentrations of anti-EGFR antibodies in hFcRn mice. The study design was identical to that described in a, except that serum concentrations were measured with an EGFR antigen-down immunoassay. (c) Xenograft study in hFcRn/Rag1–/– mice comparing activity of native IgG1 and Xtend variant versions of bevacizumab against established SKOV-3 tumors. Tumor volume is plotted against day after tumor cell injection. Antibodies were dosed at 5 mg/kg every 10 d starting on day 35 (indicated by the arrows). n = 8 mice/group. *, P = 0.028 at 84 d. (d) Xenograft study in hFcRn/Rag1–/– mice comparing activity of anti-EGFR antibodies against established A431 tumors. Tumor volume is plotted against day after tumor cell injection. Antibodies were dosed 5 mg/kg every 10 d starting on day 10 (indicated by the arrows). n = 9 mice/group. *, P = 0.005 at 35 d.

Several studies have demonstrated that antibodies to ­ epidermal growth factor receptor (EGFR) are internalized. Moreover, nonlinear dose-dependent clearance has been observed in monkeys and humans, leading to the hypothesis that receptor-dependent internalization makes a major contribution to clearance of anti-EGFR antibodies11,12. The M428L/N434S Xtend variant was constructed in a humanized version (huC225) of the anti-EGFR antibody cetuximab (C225)13 (Erbitux), which is approved for the treatment of colorectal and head and neck cancers. We refer to this pharmacokinetically enhanced anti-EGFR antibody as Xtend-EGFR. The improvement in affinity for human FcRn resembled that observed for anti-VEGF; binding to human EGFR antigen was unperturbed, and both cetuximab and humanized cetuximab cross-react with cynomolgus EGFR14 (data not shown). The 7.5 mg/kg dose chosen for this study is in a range where the dose-clearance relationship is nonlinear14. In our hands cetuximab had a half-life of 1.5 d (Supplementary Table 2), similar to previously published data at the same dose (2.7–3.1 d)14. Consistent with the bevacizumab results, the Xtend variant antiEGFR increased half-life to 4.7 d, reflecting a 3.1-fold improvement (Fig. 1b and Supplementary Table 2). We have thus demonstrated pharmacokinetic improvements conferred by Fc engineering of an internalizing antibody, even when it is dosed within the nonlinear clearance regime. We performed pharmacokinetic experiments in C57BL/6J (B6)background mice that are homozygous for a knockout allele of murine FcRn and heterozygous for a human FcRn transgene (mFcRn–/–,

158

hFcRn+)15, referred to here as hFcRn mice. A description of these experiments is provided in the Supplementary Methods. Serum concentration data for native IgG1 and Xtend anti-VEGF antibodies showed a dramatic enhancement in half-life for the variant relative to native IgG1 (Fig. 2a), improving half-life fourfold from ~3–12 d (Supplementary Table 2). In the anti-EGFR context, the Xtend variant improved half-life to 13.9 d relative to 2.9 d for ­cetuximab, resulting in an enhancement of about fivefold (Fig. 2b and Supplementary Table 2). The IgG1 version of huC225 also had a relatively short half-life of 2 d (data not shown). We observed a general correlation between antibody half-life and FcRn affinity at pH 6.0 across two anti-VEGF studies and one anti-EGFR hFcRn pharmaco­ kinetic study. The pharmacokinetic results for individual variants and native IgG1 were consistent and reproducible between the three studies (Supplementary Fig. 2b–c and Supplementary Table 2). To test whether the slower clearance of our pharmacokineticengineered antibodies results in improved exposure-related pharmacology, we developed an hFcRn transgenic, Rag1–/– immunodeficient mouse strain (Supplementary Methods and Supplementary Fig. 3). For VEGF, SKOV-3 tumors were established to 25–60 mm3 and then treated with either vehicle or 5 mg/kg native IgG1 or Xtend variant bevacizumab every 10 d. This dosing schedule approximated the halflife of the Xtend variant, but was three to four half-lives longer than the half-life of the native IgG1 version (Supplementary Table 2). A statistically greater level of tumor reduction (P = 0.028 at study termination) was observed for the Xtend variant relative to the native IgG1 version (Fig. 2c). A similar study in hFcRn/Rag1–/– mice comparing Xtend-EGFR to a native IgG1 version showed similar improvements in tumor reduction (P = 0.005) against established A431 epidermoid carcinoma tumors (Fig. 2d). Consistent with the pharmacokinetic results in hFcRn mice (Fig. 2a–b), the variants reduced clearance in the hFcRn/Rag1–/– mice (Supplementary Fig. 3a–b), demonstrating an inverse correlation between tumor volume and serum concentration of antibody at study termination. These results indicate that the slower clearance of the variant antibodies leads to higher drug exposure and consequently superior tumor-suppressing pharmaco­ logy. Additional studies comparing various dosing intervals of the Xtend variants and parent antibodies will be necessary to precisely define dosing regimens for optimal clinical benefit. However, the results described here firmly establish a positive correlation between pharmaco­kinetic enhancement and in vivo efficacy. Despite the reasonably long half-lives of monoclonal antibodies, market pressures for higher patient convenience and compliance continue to drive antibody drug programs toward less frequent dosing schedules. Yet, because of the potential loss in efficacy when the dosing frequency is not justified by the pharmacokinetics of the drug, the critical issue of whether slower antibody clearance through Fc engineering leads to superior exposure-dependent efficacy has remained unresolved. Our results indicate that, for at least some therapies, efficacy can be preserved with extended dosing intervals enabled by pharmacokinetic engineering. This work thus paves the way for a new generation of antibody therapies and biologically superior versions of approved antibody drugs that deliver finer control over dosing while providing greater convenience to patients. Note: Supplementary information is available on the Nature Biotechnology website.

Acknowledgments We thank The Jackson Laboratory JAX West and SNBL USA for carrying out pharmacokinetic experiments, B. Dahiyat for helpful discussions, and A. Eivazi, D.-H.T. Nguyen, H. Herman, J.M. Jacinto and U.S. Muchhal for technical contributions.

VOLUME 28 NUMBER 2  FEBRUARY 2010  nature biotechnology

b r i e f c o m m u n i c at i o n s AUTHOR CONTRIBUTIONS J.Z., A.K.C., H.M.H., G.A.L., D.C.R. and J.R.D. designed the research, J.Z., A.K.C., H.M.H., S.K., I.W.L.L. and T.J.S. carried out experiments, and J.Z., G.A.L. and J.R.D. wrote the manuscript. COMPETING INTERESTS STATEMENT The authors declare competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/naturebiotechnology/. Published online at http://www.nature.com/naturebiotechnology/. Reprints and permissions information is available online at http://npg.nature.com/ reprintsandpermissions/.

© 2010 Nature America, Inc. All rights reserved.

1. Roopenian, D.C. & Akilesh, S. Nat. Rev. Immunol. 7, 715–725 (2007). 2. Presta, L.G.. Curr. Opin. Immunol. 20, 460–470 (2008). 3. Datta-Mannan, A., Witcher, D.R., Tang, Y., Watkins, J. & Wroblewski, V.J. J. Biol. Chem. 282, 1709–1717 (2007).

4. Gurbaxani, B., Dela Cruz, L.L., Chintalacharuvu, K. & Morrison, S.L. Mol. Immunol. 43, 1462–1473 (2006). 5. Dall′Acqua, W.F., Kiener, P.A. & Wu, H.. J. Biol. Chem. 281, 23514–23524 (2006). 6. Hinton, P.R. et al. J. Biol. Chem. 279, 6213–6216 (2004). 7. Hinton, P.R. et al. J. Immunol. 176, 346–356 (2006). 8. Yeung, Y.A. et al. J. Immunol. 182, 7663–7671 (2009). 9. Presta, L.G. et al. Cancer Res. 57, 4593–4599 (1997). 10. Lin, Y.S. et al. J. Pharmacol. Exp. Ther. 288, 371–378 (1999). 11. Fan, Z., Lu, Y., Wu, X. & Mendelsohn, J. J. Biol. Chem. 269, 27595–27602 (1994). 12. Lammerts van Bueren, J.J. et al. Cancer Res. 66, 7630–7638 (2006). 13. Naramura, M., Gillies, S.D., Mendelsohn, J., Reisfeld, R.A. & Mueller, B.M. Cancer Immunol. Immunother. 37, 343–349 (1993). 14. Imclone Systems, Inc Biologic License Application 125084, Erbitux (Cetuximab) (US Food and Drug Administration, Feb. 12, 2004). 〈http://www.accessdata.fda. gov/drugsatfda_docs/bla/2004/125084_ERBITUX_PHARMR_P2.PDF〉. 15. Petkova, S.B. et al. Int. Immunol. 18, 1759–1769 (2006).

nature biotechnology  VOLUME 28 NUMBER 2  FEBRUARY 2010

159

l e tt e rs

Expansion and maintenance of human embryonic stem cell–derived endothelial cells by TGFb inhibition is Id1 dependent

© 2010 Nature America, Inc. All rights reserved.

Daylon James1, Hyung-song Nam2,7,8, Marco Seandel1,3,8, Daniel Nolan1, Tyler Janovitz1, Mark Tomishima4, Lorenz Studer4, Gabsang Lee4, David Lyden1, Robert Benezra2, Nikica Zaninovic5, Zev Rosenwaks5, Sina Y Rabbany1,6 & Shahin Rafii1 Previous efforts to differentiate human embryonic stem cells (hESCs) into endothelial cells have not achieved sustained expansion and stability of vascular cells. To define vasculogenic developmental pathways and enhance differentiation, we used an endothelial cell–specific VE-cadherin promoter driving green fluorescent protein (GFP) (hVPr-GFP) to screen for factors that promote vascular commitment. In phase 1 of our method, inhibition of transforming growth factor (TGF)β at day 7 of differentiation increases hVPr-GFP+ cells by tenfold. In phase 2, TGFβ inhibition maintains the proliferation and vascular identity of purified endothelial cells, resulting in a net 36-fold expansion of endothelial cells in homogenous monolayers, which exhibited a transcriptional profile of Id1highVEGFR2highVE-cadherin+ ephrinB2+. Using an Id1-YFP hESC reporter line, we showed that TGFβ inhibition sustains Id1 expression in hESC-derived endothelial cells and that Id1 is required for increased proliferation and preservation of endothelial cell commitment. Our approach provides a serum-free method for differentiation and long-term maintenance of hESC-derived endothelial cells at a scale relevant to clinical application. Human embryonic stem cells (hESCs), which self-renew indefinitely1, offer a plentiful source of endothelial cells for therapeutic revascularization. However, few studies have identified specific developmental stimuli sufficient to support the specification and maintenance of large numbers of functional and vascular-committed endothelial cells from hESCs2–7. Although small numbers of hESC-derived endothelial cells have been generated in short-term cultures, these cells have not been subjected to sustained expansion, angiogenic profiling or interrogated as to the stability of vascular fate. As a result, molecular pathways that maintain vascular identity and long-term expansion of hESC-derived endothelial cells remain unknown. To detect the emergence of endothelial cells from differentiating hESCs in real time, we generated a cell line for endothelial cell–specific lineage

tracing. We cloned a 1.5-kilobase fragment from a bacterial artificial chromosome (BAC) containing the genomic locus of the human endothelial cell–specific gene VE-cadherin (CDH5). The promoter sequence of this gene, encompassing a region upstream of exon 1, was inserted into a lentiviral vector upstream of GFP (hVPr-GFP; Fig. 1a). Human endothelial cells transduced with this vector showed robust expression of GFP, in contrast to transduced human mesenchymal and fibroblastic cells, which did not express GFP (Supplementary Fig. 1a–c). Endothelial-specific expression of the reporter was also evident in transduced, spontaneously differentiating hESCs (RUES1 line; Fig. 1b and Supplementary Fig. 1d–j): hVPr-GFP+ cells were organized into capillary-like structures expressing endothelial cell markers, including VE-cadherin, CD31 and CD34 (Supplementary Figs. 1d–g and 2a,b), and were negative for alpha smooth muscle actin (α-SMA) and CD45, a marker of hematopoietic cells (Supplementary Figs. 1h–j and 2c). Using the hVPr-GFP hESC reporter line, we tracked the chronology and geometry of vasculogenic differentiation in differentiating embyroid bodies by time-lapse confocal microscopy. Beginning at day 5, we observed the specification and emergence of hVPr-GFP+ cells (Supplementary Video 1 and Supplementary Fig. 3), and by day 8, hVPr-GFP+ cells co-expressing vascular endothelial growth factor receptor (VEGFR)2 and CD31 (Fig. 1c) formed motile vessel-like structures (Supplementary Video 2). These data validated the ability of the hVPr-GFP reporter construct to specifically identify and track hESC-derived nascent endothelial cells. We used the reporter line to develop a chemically defined, serum-free method for enhancing vascular differentiation. In phase 1, heterogenous embryoid body cultures of hVPr-GFP hESCs were sequentially stimulated with bone morphogenetic protein (BMP)4, activinA, fibroblast growth factor (FGF)-2 and VEGF-A8–10 (Fig. 1d). Although these growth conditions promoted formation of hVPr-GFP+ structures (Supplementary Fig. 4 and Supplementary Videos 3 and 4), the yield of dissociated hVPr-GFP+ endothelial cells obtained by fluorescence-activated cell sorting (FACS) was low, and these few isolated endothelial cells could not be expanded without the majority of cells assuming a non-endothelial cell phenotype

1Howard Hughes Medical Institute, Ansary Stem Cell Institute, Department of Genetic Medicine, Weill Cornell Medical College, New York, New York, USA. 2Program in Cancer Biology and Genetics, Memorial Sloan-Kettering Cancer Center, New York, New York, USA. 3Division of Medical Oncology, Department of Medicine, Memorial Sloan-Kettering Cancer Center, New York, New York, USA. 4Developmental Biology Program, Memorial Sloan-Kettering Cancer Center, New York, New York, USA. 5Ronald O. Perelman and Claudia Cohen Center for Reproductive Medicine, New York, New York, USA. 6Bioengineering Program, Hofstra University, Hempstead, New York, USA. 7Present address: Weill Cornell Medical College, New York, New York, USA. 8These authors contributed equally to this work. Correspondence should be addressed to S.R. ([email protected]).

Received 1 December 2009; accepted 8 January 2010; published online 17 January 2010; doi:10.1038/nbt1605

nature biotechnology volume 28 number 2 february 2010

161

l e tt e rs a Intron 1 12.5 kb Exon 1 64 bp

LTR U3 U5

105

104

Exon 2 226 bp

eGFP

GA RRE

c

hVPr-GFP

α-CD31

1.5 kb hVPr

b

ATG

U3 U5

WPRE

103

102 0 −102 −100

−102 0 102 −203

d

Suspension

103

104

105

α-VEGFR2

Adherent Isolation

0

1

2

4 BMP-4 Activin A

14

7

FGF-2

VEGF-A

SB431542

Phase 1 differentiation

e

hVPr-GFP

f

Nuc VECad hVPr-GFP

103

Propidium iodide

© 2010 Nature America, Inc. All rights reserved.

Mesodermal and vascular specification

g

Phase 2 isolation

EC amplification

h

−SB

Day 14

i

+SB

Grouped

Isolated

2.0 102

% hVPr-GFP+ cells

Day (−1)

1.6

101

P < .001

P < .01

1.2 0.8

100

0.15% 100

101

102

1.8% 103

hVPr-GFP

100

101

102

0.4

103

d0+

N

d7+

N

d7+

Figure 1 Sequential TGFβ activation followed by inhibition during phase 1 differentiation promotes a tenfold expansion of hVPr-GFP + hESC-derived cells. (a) A 1.5-kb fragment of the putative human VE-cadherin promoter (hVPr) region was isolated from a BAC clone and placed upstream of GFP in a lentiviral expression vector (hVPr-GFP). (b) Spontaneously differentiating embryoid bodies exhibited expression of hVPr-GFP in tubular structures. Inset, merge of GFP and brightfield views. (c) Flow cytometric analysis showed hVPr-GFP+ cells were positive for the vascular markers CD31 and VEGFR2. (d) Schematic diagram showing the sequence in which BMP4, activinA, FGF-2, VEGF-A and SB431542 were added and removed from differentiation cultures. EC, endothelial cell. (e) Adherent hVPr-GFP cultures stimulated with SB431542 (10 µM) resulted in differentiation of hESCs into monolayers of hVPr-GFP+ adherent cells. Inset, hVPr-GFP+ cells alone. (f) Human VPr-GFP+ cells were immuno-positive for VE-cadherin. Blue, nuclear counterstain. Inset, hVPr-GFP + cells alone. (g,h) The proportion of hVPr-GFP+ cells was measured by flow cytometry at day 14 after culture in the absence (–SB; g) and presence (+SB; h) of SB431542. (i) Measurement of hVPr-GFP+ cells at day 14 when embryoid bodies were cultured either in groups or as isolated embryoid bodies and SB431542 was added at day 0, day 7 or not at all (N). Error bars represent s.d. of experimental values performed in triplicate. Scale bars, 100 µm.

(data not shown). We therefore screened for bioactive small molecules that would improve vascular differentiation. Screening of >20 molecules associated with early developmental signaling pathways (Supplementary Table 1) showed that the TGFβ-inhibitory molecule SB431542 (ref. 11) reproducibly increased the yield of hVPr-GFP+ cells. Adding SB431542 to differentiation cultures at day 7 resulted in the formation of hVPrGFP+ VE-cadherin+ monolayers (Fig. 1e,f), which, upon dissociation, yielded tenfold more hVPr-GFP+ endothelial cells than cultures stimulated by cytokines alone (Fig. 1g–i). No hVPr-GFP+ cells were generated if SB431542 was added at the onset of differentiation (day 0), suggesting that vascular commitment depends on active TGFβ/activin/nodal signaling before day 7. Kinetic analysis of differentiation suggested a shift from a pluripotent phenotype (Oct3/4+; Fig. 2a) to a vascular phenotype (CD31+; Fig. 2b,c) through a mesodermal intermediate (brachyury+; Fig. 2a). Addition of SB431542 to differentiating hESC cultures at day 7 accelerated the reduction of Oct3/4 and brachyury and increased the number

162

of hVPr-GFP+CD31+ cells beginning at about day 9, while reducing expression of α-SMA (Fig. 2b,c). After isolation from heterogenous cultures by FACS, endothelial cells grown in the absence of TGFβ inhibition retained high expression of CD31 but also expressed α-SMA (Supplementary Video 5), indicating that these endothelial cell–like cells had not assumed a terminally committed vascular fate. Expression of α-SMA in hESC-derived endothelial cells suggested a degree of plasticity that is not present in terminally differentiated endothelial cells (human umbilical vein endothelial cells; HUVEC, Fig. 2b). Indeed, extended culture (>10 d after FACS isolation) of hESCderived endothelial cells in the absence of TGFβ inhibition yielded a substantial number of cells co-expressing VE-cadherin and α-SMA (Fig. 2d). One explanation for the increased percentage of endothelial cells in SB431542-stimulated cultures is maintenance of the vascularcommitted state after specification. To test the capacity of TGFβ inhibition to promote expansion of pure populations of hESC-derived endothelial cells, we dissociated day 14 differentiation cultures, isolated

volume 28 number 2 february 2010 nature biotechnology

l e tt e rs

1,000 100 10 1 0

4

7

9

12

αSMA − SB αSMA + SB

600

CD31 − SB CD31 + SB

500 400

10k

300 200

5k

100

14

−1

0

4

Day Nuc SMA VECad

7

9

12

c

hVPr-GFP+ − SB 1.0

hVPr-GFP+ + SB

2.5

0.8

CD31 − SB CD31+ + SB

2.0

+

0.6

1.2

0.4

1.0

0.2

0.5

14

7

8

9

Day

e

Nuc VECad −SB

f

11

14

Day

g

Nuc VECad +SB

12

Cell number (k)

© 2010 Nature America, Inc. All rights reserved.

d

15k

Total cells CD31+ cells

h

i

4.0

140

Total number of αSMA+ cells

−1

b

Percent CD31+ cells

Relative expression (a.u.)

Oct3/4 − SB Oct3/4 + SB

120

10

100

8 6 4 2

−SB

+SB

80 60 40 20

−SB

+SB

Percent of PHH3+ cells

Brachyury − SB Brachyury + SB

Relative expression (a.u.)

a

10,000

In aggregate, TGFβ inhibition in phase 1 and 2 resulted in a 36-fold expansion in the total number of vascular-committed hESC-derived endothelial cells, with 7.4 CD31+ VE-cadherin+ endothelial cells generated from every one hESC input over 20 d, compared with 0.2 endothelial cells per input hESC derived from control culture conditions (Fig. 2j). Similar levels of expansion of hESC-derived endothelial cells were achieved with four additional hESC lines and one induced pluripotent stem cell line using the same protocol except that either SB431542 or soluble TGFβRII

Percent hVPr-GFP+ cells

hVPr-GFP+ cells by FACS and expanded them for an additional 5 d with or without SB431542 (Phase 2, Fig. 2e–i). SB431542-treated cultures yielded more cells in the 5-d period, and a higher percentage of the population retained an α-SMA−CD31+VE-cadherin+ phenotype (Fig. 2e–h). In addition to preserving the vascular phenotype, SB431542 also increased cell proliferation, as indicated by a higher percentage of phospho-histoneH3+ (PHH3) mitotic endothelial cells (Fig. 2i and Supplementary Videos 6 and 7).

3.0 2.0 1.0

−SB

+SB

j Undifferentiated hESCs Non-vascular cells Smooth muscle cells Endothelial cells

<10k ECs

(−SB)

1:0.2 ~10k ECs

−SB

50k hESCs ~60k ECs

+SB

Day 0

1:7.4

(+SB)

Day 7 (+SB)

~360k ECs Day 14 (purification)

Phase 1 Heterogenous differentiation culture

Day 20

Ratio of ESCs/ECs

Phase 2 Homogenous isolated culture

Figure 2 TGFβ inhibition after endothelial cell isolation during phase 2 increases yield and preserves vascular identity of purified endothelial cells. (a–c) Human VPr-GFP hESCs were sequentially stimulated with cytokines (–SB) and SB431542 (+SB) (Fig. 1d and Online Methods) and cultures were assessed for the prevalence of pluripotency (Oct3/4) and mesodermal transcripts (brachyury) (a), CD31 and α-SMA transcripts (b) and endothelial cell markers hVPr-GFP and CD31 (c) at multiple time points during differentiation. The secondary axis in b shows values for cells shown in solid bars. (d) Isolated endothelial cells that were cultured in the absence of SB431542 were stained for both VE-cadherin and α-SMA and showed rare cells that were positive for both markers (arrowhead in the inset). Inset, α-SMA alone. (e–i) Human VPr-GFP + cells were isolated from differentiation cultures at day 14 by FACS and further cultured in the absence (e) or presence (f) of SB431542. (g) Flow cytometric assessment of CD31 was performed after 5 d of isolated culture (total cells are shown in white and CD31 + cells are shown in black in the bar graph). (h) After isolation and 5 d of culture in the presence or absence of SB431542, the incidence of α-SMA+ cells was measured. (i) After 5 d of culture following isolation, unstimulated cultures showed reduced incidence of cells positive for phospho-histoneH3 (PHH3 +), relative to SB431542-stimulated cultures. The mean incidences of α-SMA and phospho-histoneH3 positive cells were obtained by counting positively stained cells in multiple parallel wells. (j) The yield of endothelial cells (ECs) from hESCs is schematized relative to a 50,000 hESC input at day 0. The relative difference in endothelial cell (ECs) number is indicated at day 14 (upon isolation from differentiation cultures), and day 20 (after expansion in isolated conditions). The ratio of input hESCs to committed hESC-derived endothelial cells after 20 d is also shown. Relative transcript abundance was measured by QPCR and normalized to the housekeeping gene β-actin (ACTB). Error bars in (a–c and g–i) represent s.d. of experimental values performed in triplicate. Scale bars, 100 µm.

nature biotechnology volume 28 number 2 february 2010

163

l e tt e rs

Expression

d

103

102

103

104

low

+

3 Days (+SB)

Genes

104

102

CD31 ld1high

CD31 ld1

: Below detection threshold

105

ld1 promoter: YFP

ld1:YFP

−SB

+SB

−SB

+SB

4 Days ld1low(−SB)

ld1low(+SB) ld1high(−SB)

III

f

ld1:YFP MFI (k)

α-CD31 antibody

© 2010 Nature America, Inc. All rights reserved.

105

ld1:YFP

II +

IV

30

ld1high(+SB)

V

ld1:YFP MFI

VI

CD31 MFI

35

24

28

18

21

12

14

6

7

5k

5k

11.2k

21.7k

18.3k

53.8k

I

II

III

IV

V

VI

CD31 MFI (k)

c CD31+ id1high

CD31+ id1low

d14 id1:YFP FACS

2 3 4 5

CXCR4 EphrinB2 EphB4 Lyve-1 Podoplanin Prox-1 VEGFR1 (Flt1) VEGFR2 VEGFR3 (Flt4) CD34 CD133 (Prominin1) E-Selectin P-Selectin PECAM (CD31) Thrombomodulin VE-cadherin vonWillebrand factor HOXA9 ld1 ld2 ld3 α-Smooth muscle actin

b

e I

0.0 0.5 1.0

Cord blood CD34+ cells

Endothelium (HUVEC)

hVPr-GFP EBs d14 (−SB)

Sorted hVPr-GFP+ d24 (+SB)

Umbilical cord derived

Sorted hVPr-GFP+ d14 (+SB)

hESC derived

Smooth muscle cells

a

Total cells

Figure 3 Molecular profiling of hESC-derived endothelial cells reveals a signature defined by high Id1 expression. Human VPr-GFP embryoid bodies and highly purified hVPr-GFP+ cells were compared to mature vascular cells by microarray analysis. (a) RNA was extracted for microarray analysis from human VPr-GFP embryoid bodies cultured in the presence of recombinant cytokines alone until day 14; isolated endothelial cells (99.8% pure) from hVPr-GFP embryoid bodies cultured in the presence of recombinant cytokines and the TGFβ inhibitor SB431542 until day 14; isolated endothelial cells (>95% pure) from hVPr-GFP embryoid bodies cultured in the presence of recombinant cytokines and the TGFβ inhibitor SB431542 until day 14, followed by 10 d additional culture in the presence of cytokines and SB431542; HUVECs; human umbilical vein smooth muscle cells; and CD34 + umbilical cord blood cells. Expressed factors are displayed in an ordered array, as labeled, with highly expressed factors shown in red, minimally expressed factors shown in blue and factors for which transcripts are below a significant expression level shown in gray. (b–f) Following the endothelial cell differentiation protocol (Fig. 1d), Id1YFP hESC-derived cells were sorted by FACS, separating the CD31+ population into Id1-YFPhigh-expressing cells (b (green) and c) and Id1-YFPlow-expressing cells (b (red) and d). Insets, brightfield views on the day after isolation of Id1-YFPhigh (c) and Id1-YFPlow (d) cells. (e) After 3 d culture in the presence of SB431542, both populations were transferred to conditions with and without SB431542 for an additional 4 d (+SB and –SB, respectively). (f) Total cells and mean fluorescence intensity (MFI) measurements of Id1:YFP (black) and CD31+ (white) were measured for: CD31+Id1low (I) and CD31+Id1high (II) populations upon isolation; and for four populations following culture conditions (as shown in e, III–VI). Scale bars, 100 µM.

receptor decoys was used interchangeably to inhibit activation of the activin/nodal branch of TGFβ superfamily signaling (Supplementary Fig. 5a–e). These results demonstrate that the effect of TGFβ inhibition shown for the RUES1 line is applicable to other pluripotent cell lines. To define the vasculogenic transcriptional signature of hESC-derived endothelial cells at different time points during phases 1 and 2, we carried out Affymetrix microarray analyses of several hESC-derived populations and mature cell types (Fig. 3a). The yield of freshly isolated phase 1 endothelial cells in the absence of TGFβ inhibition was insufficient for microarray analyses, underscoring the value of our approach for generating sufficient expanding (phase 1) and vascular-committed (phase 2) endothelial cells for molecular profiling. Phase 1 hESC-derived endothelial cells showed increased levels of factors typical of arterial-like endothelial cells (VEGFR2, VEGFR1, Id1, CD31, CD34, VE-cadherin, vWF, thrombomodulin, ephrinB2 and E-selectin) but

164

not of lymphatic endothelial cells (Prox1 and podoplanin). Markers of vascular progenitor cells, including CD133 and Id1 (refs. 12–17), were also highly expressed in phase 1 endothelial cells and downregulated upon in vitro culture. Transcription factors expressed primarily in committed endothelial cells, including HoxA9 (ref. 18), were not expressed in phase 1 endothelial cells. Accordingly, we defined a comprehensive vasculogenic expression profile of the hESC-derived endothelial cell population as VE-cadherin +VEGFR2highId1highthrombomodulinhighephrinB2+CD133+HoxA9−, whereas mature endothelial cells were identified by a VE-cadherin+VEG FR2lowId1lowephrinB2+CD133−HoxA9+ phenotype. Id1 was one of numerous transcription factors upregulated in phase 1 endothelial cells. Because it has been shown to modulate differentiation and maintenance of vascular cell fate19, we focused on Id1 as a potential mediator of the pro-angiogenic effect of TGFβ-inhibition observed in our study. To track Id1 expression in live hESC differentiation cultures, we used

volume 28 number 2 february 2010 nature biotechnology

l e tt e rs c hVPrGFP 0.56%(+/− 0.2)

104

103

hVPrGFP

hVPrGFP

VEGFR2 14.6%(+/− 2.6)

102

−140

0 102

103

104

105

VEGFR2 6.2%(+/− 1.4) −170

0 102

103

104

d

d14+8 (+SB)

f

Total cells CD31+ cells

150

100

Scr

105

VE-cadherin

50k

25k

50

α-VEGFR2

e

75k

200

Cell number

α-CD31

10

b hVPrGFP 1.31%(+/− 0.3)

Relative Id1 expression

a 5

g

Id1

−

Scr

Id1

HUVEC

hVprGFP

Nuc

hVPr-GFP

Scr

Id1

Scr

−SB

h

Id1 +SB

hVPr-GFP

GIB4

© 2010 Nature America, Inc. All rights reserved.

i

Figure 4 TGFβ inhibition upregulates Id1 expression and is necessary for the increased yield of functional endothelial cells capable of in vivo neoangiogenesis. (a,b) Human VPr-GFP hESCs that were stably transduced with control (a) or Id1-specific (b) shRNAs were differentiated according to the protocol shown in Figure 1d and assessed at day 14 for the prevalence of VEGFR2+ (blue) and hVPr-GFP+ (green) cells. The insets show plots of side scatter on the y axis and hVPr-GFP on the x axis. (c) Control and Id1-specific shRNAs were added to HUVEC or freshly isolated (at day 14) hVPr-GFP + cells, and the relative Id1 transcript levels were measured after 3 d. *, P < 0.05. Error bars, s.d. of experimental values performed in triplicate. (d) Control and Id1-specific shRNAs were added to freshly isolated hVPr-GFP + cells, which were cultured in the absence or presence of SB431542. After 5 d, the total cell number and proportion of CD31 + cells was measured by flow cytometry. Error bars, s.d. of experimental values performed in triplicate. Scr, scrambled control shRNA. (e–g) Human VPr-GFP+ cells were isolated by FACS at day 14 and expanded in monolayer culture (e) for 8 d while retaining expression of both the endogenous VE-cadherin (f) and the hVPr-GFP transgene (g). Panel e shows a mosaic view of one well of a 24-well dish. A magnified view of the boxes in e and f are shown in f and g, respectively. (h,i) Expanded cells were injected in Matrigel plugs into immunodeficient mice and excised after 10 d following intravital labeling of functional vasculature with lectin (GIB4, blue). h, View of hVPr-GFP+ cells alone; i, view of hVPr-GFP+ cells merged with GIB4+ cells. Scale bars, 100 µM.

a stable BAC transgenic hESC line20 containing yellow fluorescent protein driven by the Id1 promoter (Id1-YFP) (Fig. 3b–f) (Nam, H.S. and Benezra, R., unpublished data). Differentiated endothelial cells were isolated at day 14 from Id1-YFP cultures (Fig. 1d), sub-fractionating the CD31+ population into Id1-YFP high-expressing (Fig. 3c) and low-expressing (Fig. 3d) cells, and these populations were serially expanded for 7 d with or without the TGFβ inhibitor (Fig. 3e,f). Flow cytometric analysis of these cells revealed a direct relationship between upregulation of Id1 expression and TGFβ inhibition. Notably, although SB431542 increased the percentage of the CD31+ population, the mean fluorescence intensity of CD31 on these cells was lower than that of unstimulated cells. These data suggested that TGFβ inhibition increased expansion of hESC-derived endothelial cells by maintaining high levels of Id1 expression and preserving an immature proliferative phenotype. To determine the requirement for Id1 in mediating endothelial cell commitment, we transduced hVPr-GFP+ cells with lentiviral short hairpin (sh)RNA targeted against the Id1 transcript (Fig. 4a,b). In the presence of SB431542, knockdown of Id1 reduced the numbers of both VEGFR2+ vascular progenitors and hVPr-GFP+ cells at day 14. When the Id1 shRNA construct was introduced after isolation of the hVPr-GFP+ fraction (Fig. 4c), it elicited a marked decrease in CD31+ endothelial cells after 5 d of SB431542 treatment (Fig. 4d). These results identified TGFβ inhibition– mediated Id1 upregulation as a primary effector in promoting endothelial cell expansion and maintaining long-term vascular identity.

nature biotechnology volume 28 number 2 february 2010

To demonstrate that our cultured endothelial cells could form functional vessels, we grew purified hVPr-GFP+ cells from day 14 differentiation cultures for an additional 8 d in the presence of SB431542. These endothelial cells showed high proliferative potential (up to ten cell divisions) and generated homogenous hVPr-GFP+VE-cadherin+ monolayers (Fig. 4e–g) with retention of hVPr-GFP fluorescence at the single-cell level (arrowheads in Fig. 4g). These cells were subcutaneously injected in Matrigel plugs into nonobese (NOD)/severe combined immunodeficient (SCID) mice and 10 d later extracted after intravenous injection of lectin into live animals. In Matrigel plugs, hVPr-GFP+ cells co-localized with lectin+ cells, forming chimeric vessels along with host cells (Fig. 4h–i and Supplementary Videos 8 and 9). These data indicated that the endothelial cells generated by our methods could function in vivo. A prerequisite to therapeutic vascularization using hESC-derived cells is generation of abundant durable endothelial cells that upon expansion maintain their angiogenic profile without differentiating into nonendothelial cell types. Here, we show that differentiation of hESCs into a large number of stable and proliferative endothelial cells can be achieved by early-stage TGFβ-mediated mesoderm induction followed by TGFβ inhibition beginning at day 7 (phase 1) and after isolation at day 14 (phase 2). Using this approach, we achieved a 36-fold net expansion of committed endothelial cells. The increased yield allowed transcriptional analysis, which revealed a molecular signature that sheds light on the regulatory influences that govern embryonic vasculogenesis. Indeed, genes encoding

165

© 2010 Nature America, Inc. All rights reserved.

l e tt e rs factors associated with vascular progenitor identity (Id1high, VEGFR2high, CD133)12–17,19 as well as vascular markers (PECAM, VE-cadherin, ephrinB2) were highly expressed in hESC-derived endothelial cells and, among these factors, Id1 was found to act downstream of TGFβ inhibition to increase endothelial cell yield by promoting proliferation and preserving vascular commitment. These studies establish TGFβ modulation of Id1 expression as a determinant of hESC-derived endothelial cell identity and set the stage for large-scale generation of authentic long-lasting human endothelial cells for therapeutic vascularization. Our use of vascular-specific hVPr-GFP and Id1-YFP hESC reporter lines in small-molecule screens allowed the discovery of the TGFβ inhibitor SB431542 as a key stimulus for human endothelial cell differentiation and proliferation in serum-free conditions. In murine ESCs, TGFβ and serum factors promote smooth muscle cell differentiation, whereas inhibition of this pathway promotes formation of CD31+ cells21. Our data show that stage-specific TGFβ inhibition, beginning on day 7 at a point following TGFβ-mediated mesoderm induction, increases the mitotic index and maintenance of hESC-derived endothelial cells by upregulation of Id1 expression. Differentiation of hVPr-GFP hESCs with TGFβ inhibition generated endothelial cells at yields tenfold greater than those of cells differentiated with angiogenic factors alone, and after purification, TGFβ inhibition supported endothelial cell expansion for up to ten cell divisions while retaining the angiogenic surface phenotype. The ability of TGFβ inhibition to increase endothelial cell yield in both differentiating (phase 1) and purified (phase 2) cultures resulted in a 36-fold increase in the absolute number of hESC-derived endothelial cells, with 95% of the population maintaining endothelial cell identity. As such, we have established a means of generating a homogenous population of stable endothelial cells in ratios that greatly exceed hESC input and are relevant to therapeutic vasculoplasty. Expression of Id1 has been shown to inhibit cell differentiation and growth arrest in multiple cell types22. The TGFβ signaling pathway, through the effectors Smad3 and ATF3, has been shown to repress Id1 promoter activity23. The link between TGFβ signaling, Id1 and preservation of proliferation and phenotypic identity of hESC-derived endothelial cells provides insight into the molecular mechanisms that regulate vascular ontogeny during human development. Indeed, these results point toward a biphasic role for TGFβ signaling during vasculogenesis, whereby early activation of this pathway is required for specification of mesodermal progenitors, and inhibition after vascular commitment functions to increase mitotic index and prevent the loss of endothelial identity. Our approach for vascular monitoring and differentiation may enable the identification of as-yet unrecognized vasculogenic and angiogenic modulators for preclinical studies aimed at the cell-based therapeutic revascularization of ischemic tissues. METHODS Methods and any associated references are available in the online version of the paper at http://www.nature.com/naturebiotechnology/. Accession codes. GEO: GSE19735. ACKNOWLEDGMENTS We thank A. Brivanlou for providing the RUES1 hESC line. D.J., M.S. and G.L. are Fiona and Stanley Druckenmiller Fellows of the New York Stem Cell Foundation. S.R. is supported by Howard Hughes Medical Institute; Ansary Stem Cell Institute; Anbinder and Newmans Own Foundation; National Heart, Lung, and Blood Institute R01 grants HL075234 and HL097797; Qatar National Priorities Research Program; and Empire State Stem Cell Board and New York State Department of Health, NYS C024180.

166

Author Contributions D.J. designed and performed the experiments and wrote the manuscript. H.-s.N. and R.B. designed and created the Id1-YFP BAC transgenic vector. M.S. performed experiments and contributed to the manuscript. D.N. performed flow cytometric experiments. T.J. performed molecular cloning. M.T. and L.S. generated the Id1YFP BAC transgenic hESC line. L.S. and G.L. generated the FD iPSC line. N.Z. and Z.R. generated the hESC lines WMC2, WMC8 and WMC9. D.L. and S.Y.R. designed experiments and performed data analysis. S.R. designed experiments and wrote the manuscript. Note: Supplementary information is available on the Nature Biotechnology website. COMPETING INTERESTS STATEMENT The authors declare no competing financial interests. Published online at http://www.nature.com/naturebiotechnology/. Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/. 1. Thomson, J.A. et al. Embryonic stem cell lines derived from human blastocysts. Science 282, 1145–1147 (1998). 2. Yamahara, K. et al. Augmentation of neovascularization in hindlimb ischemia by combined transplantation of human embryonic stem cells-derived endothelial and mural cells. PLoS ONE 3, e1666 (2008). 3. Sone, M. et al. Pathway for differentiation of human embryonic stem cells to vascular cell components and their potential for vascular regeneration. Arterioscler. Thromb. Vasc. Biol. 27, 2127–2134 (2007). 4. Lu, S.J. et al. Generation of functional hemangioblasts from human embryonic stem cells. Nat. Methods 4, 501–509 (2007). 5. Goldman, O. et al. A boost of BMP4 accelerates the commitment of human embryonic stem cells to the endothelial lineage. Stem Cells 27, 1750–1759 (2009). 6. Nourse, M.B. et al. VEGF induces differentiation of functional endothelium from human embryonic stem cells: implications for tissue engineering. Arterioscler. Thromb. Vasc. Biol. 30, 80–89(2009). 7. Bai, H. et al. BMP4 regulates vascular progenitor development in human embryonic stem cells through a smad-dependent pathway. J. Cell Biochem. published online, doi:10.1002/jcb.22410 (30 November 2009). 8. Huber, T.L., Kouskoff, V., Fehling, H.J., Palis, J. & Keller, G. Haemangioblast commitment is initiated in the primitive streak of the mouse embryo. Nature 432, 625–630 (2004). 9. Levenberg, S., Zoldan, J., Basevitch, Y. & Langer, R. Endothelial potential of human embryonic stem cells. Blood 110, 806–814 (2007). 10. Yang, L. et al. Human cardiovascular progenitor cells develop from a KDR+ embryonicstem-cell-derived population. Nature 453, 524–528 (2008). 11. Inman, G.J. et al. SB-431542 is a potent and specific inhibitor of transforming growth factor-beta superfamily type I activin receptor-like kinase (ALK) receptors ALK4, ALK5, and ALK7. Mol. Pharmacol. 62, 65–74 (2002). 12. Gehling, U.M. et al. In vitro differentiation of endothelial cells from AC133-positive progenitor cells. Blood 95, 3106–3112 (2000). 13. Kelly, M.A. & Hirschi, K.K. Signaling hierarchy regulating human endothelial cell development. Arterioscler. Thromb. Vasc. Biol. 29, 718–724 (2009). 14. Peichev, M. et al. Expression of VEGFR-2 and AC133 by circulating human CD34(+) cells identifies a population of functional endothelial precursors. Blood 95, 952–958 (2000). 15. Rafii, S. & Lyden, D. Cancer. A few to flip the angiogenic switch. Science 319, 163–164 (2008). 16. Gao, D. et al. Endothelial progenitor cells control the angiogenic switch in mouse lung metastasis. Science 319, 195–198 (2008). 17. Lyden, D. et al. Impaired recruitment of bone-marrow-derived endothelial and hematopoietic precursor cells blocks tumor angiogenesis and growth. Nat. Med. 7, 1194–1201 (2001). 18. Rossig, L. et al. Histone deacetylase activity is essential for the expression of HoxA9 and for endothelial commitment of progenitor cells. J. Exp. Med. 201, 1825–1835 (2005). 19. Ruzinova, M.B. & Benezra, R. Id proteins in development, cell cycle and cancer. Trends Cell Biol. 13, 410–418 (2003). 20. Placantonakis, D.G. et al. BAC transgenesis in human embryonic stem cells as a novel tool to define the human neural lineage. Stem Cells 27, 521–532 (2009). 21. Watabe, T. et al. TGF-beta receptor kinase inhibitor enhances growth and integrity of embryonic stem cell-derived endothelial cells. J. Cell Biol. 163, 1303–1311 (2003). 22. Jankovic, V. et al. Id1 restrains myeloid commitment, maintaining the self-renewal capacity of hematopoietic stem cells. Proc. Natl. Acad. Sci. USA 104, 1260–1265 (2007). 23. Kang, Y., Chen, C.R. & Massague, J. A self-enabling TGFbeta response coupled to stress signaling: Smad engages stress response factor ATF3 for Id1 repression in epithelial cells. Mol. Cell 11, 915–926 (2003).

volume 28 number 2 february 2010 nature biotechnology

l e tt e rs ONLINE METHODS Human ESC culture. The experiments delineated in this report were performed primarily with the recently approved RUES1 hESC (generous gift from A. Brivanlou24), and corroborated using WMC2, WMC7, WMC8, generated at Weill Cornell Medical College (courtesy of Z. R./N.Z.) and H9 (Id1-YFP, courtesy of R.B./H.-s.N. and L.S./M.T.) and IPSc (courtesy L.S./G.L.). The permissions for use of these cell lines were obtained after comprehensive review by the CornellRockefeller-Sloan Kettering Institute ESC research oversight committee. The funding for execution of these studies was secured from approved non-federal funding resources. Human ESC culture medium consisted of Advanced DMEM/F12 (Gibco) supplemented with 20% knockout serum replacement (Invitrogen), 1× non-essential amino acids (Gibco), 1× l-glutamine (Invitrogen), 1× penicillin/ streptomycin (Invitrogen), 1× β-mercaptoethanol (Gibco) and 4 ng/ml FGF-2 (Invitrogen). Human ESCs were maintained on Matrigel using hESC medium conditioned by mouse embryonic fibroblasts (Chemicon).

© 2010 Nature America, Inc. All rights reserved.

Lentiviral vectors and transduction. Supernatants containing infectious particles were collected 40 and 68 h after transfection of HEK 293T with hVPr-GFP along with accessory vectors as previously described25. Viral supernatants were concentrated by ultracentrifugation and used to transduce undifferentiated RUES1 hESCs. After two passages, hESCs were disaggregated by accutase to form single cells, which were isolated and expanded to form multiple parallel cultures, each containing a relatively consistent level of viral incorporation. After expansion, these cultures were differentiated as adherent embryoid bodies and screened for the presence of GFP+ cells. Id1-YFP hESC reporter line and lentiviral Id1 shRNA knockdown. A bacterial artificial chromosome (BAC) was modified to place YFP under control of the endogenous Id1 promoter locus. This construct was electroporated into the H9 hESC line, selected for BAC integration using antibiotic resistance and subcloned. Clones were assessed and selected based on expression of YFP in Id1 hESC derivatives after spontaneous differentiation. The Id1 and control shRNA lentiviral constructs were obtained from Open Biosystems and viral particles were assembled according to the manufacturer’s recommendations (pLKO Lentiviral Packaging System). Embryoid bodies. Human VPr-GFP hESCs were grown to confluence on Matrigel (BD Biosciences) and then incubated in 5 units/ml dispase (Gibco) until colonies were completely detached from the substrate. Human VPr-GFP embryoid bodies were washed and cultured in hESC medium on ultra-low attachment plates (Corning) and cultured in the conditions described, with replacement of cytokine-supplemented medium every 48 h. Embryoid bodies were fixed in 4% paraformaldehyde and frozen for cryosectioning and staining. Endothelial differentiation protocols. Embryoid bodies were generated and cultured in base hESC medium, supplemented with the cytokines as shown. Sequential administration of cytokines was implemented (Fig. 1d). Briefly, embryoid bodies were generated in hESC base medium without FGF-2. On the morning after generation of embryoid bodies (day 0), medium was supplemented with 20 ng/ml BMP4 (R&D Systems) (removed at day 7); on day 1, medium was supplemented with 10 ng/ml activinA (R&D Systems) (removed at day 4); on day 2, medium was supplemented with 8 ng/ml FGF-2 (Peprotech) (remained for the duration of culture); on day 4, embryoid bodies were transferred to adherent conditions on Matrigel-coated plates and medium was supplemented with 25 ng/ml VEGF-A (Peprotech) (remained for the duration of culture); on day 7, SB431542 (Tocris) was added at 10 µM concentration and remained for indicated duration. Cultures were dissociated using 0.5% Trypsin/EDTA (Gibco) or Accutase (eBioscience). Absolute yield as well as ratio of input hESCs to differentiated endothelial cells was calculated from the number of live cells recovered from differentiation cultures at days 0, 14 and 20. Purified endothelial cells could be frozen and thawed in 10% DMSO with >90% recovery. Quantitative PCR. Total RNA was prepared from cultured cells using the RNeasy extraction kit (Qiagen) and reverse transcribed using Superscript II reverse transcriptase (Invitrogen) according to the manufacturer’s instructions. Relative quantitative PCR was performed on a 7500 Fast Real Time PCR System (Applied Biosystems) using either TaqMan PCR mix along with Id1 and β-actin

nature biotechnology volume 28 number 2 february 2010

primer pairs, or SYBR Green PCR mix (Applied Biosystems). Human-specific SYBR green primer pairs used were: PECAM – f, 5′-tctatgacctcgccctccacaaa–3′, r, 5′ gaacggtgtcttcaggttggtatttca-3′; Oct3/4 - f, 5′-aacctggagtttgtgccagggttt-3′, r, 5′-tgaacttcaccttccctccaacca-3′; Brachyury – f, 5′-cagtggcagtctcaggttaagaagga-3′, r, 5′-cgctactgcaggtgtgagcaa-3′; and a-SMA, f, 5′-aatactctgtctggatcggtggct-3′, r, 5′-acgagtcagagctttggctaggaa-3′. Cycle conditions were: one cycle at 50 °C for 2 min followed by 1 cycle at 95 °C for 10 min followed by 40 cycles at 95 °C for 15s and 60 °C for 1 min. Primers were checked for amplification in the linear range and primer dissociation and verified. Threshold cycles of primer probes were normalized to the housekeeping gene β-actin (ACTB) and translated to relative values. Endothelial cell isolation and flow cytometry. Endothelial cells were isolated from differentiation cultures using Magnetic Activated Cell Sorting (MACS; Miltenyi Biotech) with an antibody against CD31 conjugated to magnetic microbeads. Alternatively, cells were isolated by virtue of the expression of GFP/YFP or a fluorophore conjugated antibody to human CD31 or VEGFR2 (BD) using a FACSAriaII (BD). Microarray analysis. The Affymetrix Human Genome U133 2.0 array was used to analyze gene expression. In brief, using Qiagen RNeasy kits, total RNA was extracted from: Human VPr-GFP embryoid bodies that were cultured in the presence of recombinant cytokines alone until day 14; MACS-sorted endothelial cells isolated from hVPr-GFP embryoid bodies cultured in the presence of recombinant cytokines alone until day 14; MACS-sorted endothelial cells isolated from hVPr-GFP transduced embryoid bodies cultured in the presence of recombinant cytokines and the TGFβ inhibitor SB431542 until day 14; MACS-sorted endothelial cells isolated from hVPr-GFP embryoid bodies cultured in the presence of recombinant cytokines and the TGFβ inhibitor SB431542 until day 14, followed by 10 d additional culture in the presence of cytokines and SB431542; human umbilical vein endothelial cells; human umbilical vein smooth muscle cells; and CD34+ umbilical cord blood cells. The Superscript choice kit (Invitrogen) was used to make cDNA with a T7-(dT)24 primer incorporating a T7 RNA polymerase promoter. The biotin-labeled cRNA was made by in vitro transcription (Enzo Diagnostics). Fragmented cRNA was hybridized to the gene chips, washed, and stained with streptavidin phycoerythrin. The probe arrays were scanned with the Genechip System confocal scanner and Affymetrix Microarray suite 4.0 as used to analyze the data. Matrigel plug. Human VPr-GFP embryoid bodies were differentiated for 14 d by our differentiation protocol followed by expansion in the presence of SB431542 for 10 d and injected subcutaneously into NOD/SCID mice in a suspension of Matrigel. After 2 weeks, Griffonia simplificolia IB4 lectin and/or Ulex europus agglutinin lectin were administered intra-vitally to Matrigel plug–bearing mice and plugs were harvested, fixed overnight in 4% paraformaldehyde and equilibrated in 30% sucrose before freezing and cryosectioning. Immunofluorescence. Cryosections were immunocytochemically stained as previous described24. Briefly, samples were permeabilized in PBST and blocked in 5% donkey serum. Samples were incubated for 2 h in primary antibodies blocking solution, washed 3 times in PBS and incubated in CY3-conjugated secondary antibodies (Jackson Laboratories) for 1 h. After washing, some sections were counterstained for nucleic acids by TO-PRO3 (Invitrogen) before mounting and imaging by confocal microscopy. Primary antibodies included CD31 (DAKO), CD34 (DAKO), Phospho-HistoneH3, Smooth Muscle Actin (DAKO) and VE-cadherin (R&D). All imaging was performed using a Zeiss 510 META confocal microscope. Live imaging and 3D rendering. Human VPr-GFP embryoid bodies were cultured in a TOKAI-HIT live cell-imaging chamber on a Zeiss 510 META confocal microscope. Laser intensity and interval were optimized to ensure viability of cells for the duration of the experiments. Three-dimensional reconstruction and rendering of optical z-stacks was performed using Improvision Volocity software. 24. James, D., Noggle, S.A., Swigut, T. & Brivanlou, A.H. Contribution of human embryonic stem cells to mouse blastocysts. Dev. Biol. 295, 90–102 (2006). 25. Naldini, L. et al. In vivo gene delivery and stable transduction of nondividing cells by a lentiviral vector. Science 272, 263–267 (1996

1

letters

Real-time imaging of hepatitis C virus infection using a fluorescent cell-based reporter system

© 2010 Nature America, Inc. All rights reserved.

Christopher T Jones1, Maria Teresa Catanese1, Lok Man J Law1, Salman R Khetani2,5, Andrew J Syder1,5, Alexander Ploss1, Thomas S Oh1, John W Schoggins1, Margaret R MacDonald1, Sangeeta N Bhatia2–4 & Charles M Rice1 Hepatitis C virus (HCV), which infects 2–3% of the world population, is a causative agent of chronic hepatitis and the leading indication for liver transplantation1. The ability to propagate HCV in cell culture (HCVcc) is a relatively recent breakthrough and a key tool in the quest for specific antiviral therapeutics. Monitoring HCV infection in culture generally involves bulk population assays, use of genetically modified viruses and/or terminal processing of potentially precious samples. Here we develop a cell-based fluorescent reporter system that allows sensitive distinction of individual HCVinfected cells in live or fixed samples. We demonstrate use of this technology for several previously intractable applications, including live-cell imaging of viral propagation and host response, as well as visualizing infection of primary hepatocyte cultures. Integration of this reporter with modern image-based analysis methods could open new doors for HCV research. For over two decades, advances in HCV assay systems have been hardwon. Methodologies have ranged from adapted selectable genomes and detection methods that require fixation or cell lysis, such as immunostaining and quantitative RT PCR, to the use of infectious reporter viruses2,3. Broadening the scope of HCV research, however, will require versatile new assays that allow sensitive single-cell analysis of infection events using unmodified viral genomes. To construct a cellular marker of HCV infection, we adapted a known substrate of the HCV NS3-4A protease4–6, the mitochondrially tethered interferon (IFN)-β promoter stimulator protein 1 (IPS-1; ref. 7), also termed MAVS8, VISA9 or Cardif 4. The C-terminal region of IPS-1, encompassing the NS3-4A recognition site and a ­mitochondrial targeting sequence, was fused to green fluorescent protein (EGFP-IPS, Fig. 1a) or to the red fluorescent proteins (RFPs) mCherry or TagRFP. We also introduced an SV40 nuclear localization sequence (NLS) between the RFP variant and IPS-1 segment (RFP-NLS-IPS, Fig. 1a). Human hepatoma (Huh-7.5) cells stably transduced with lentiviruses encoding EGFP-IPS or RFP-NLS-IPS exhibited punctate fluorescence consistent with mitochondrial localization of the reporter, which was confirmed by colocalization with native IPS-1 (Fig. 1b).

We determined the reporter phenotype of the EGFP-IPS or RFPNLS-IPS constructs in the presence of NS3-4A by transduction into Huh-7.5 cells stably expressing an autonomously replicating HCV subgenome10 (JFH-1 strain, SG-JFH). Replicon-harboring cells expressing EGFP-IPS showed diffuse fluorescence, whereas an NS3-4A cleavage–resistant form11 of the reporter (EGFP-IPS(C508Y); Fig. 1c) exhibited a punctate pattern. Similarly, replicon-containing Huh7.5 cells expressing RFP-NLS-IPS, but not RFP-NLS-IPS(C508Y), showed nuclear translocation of fluorescence (Fig. 1c). Both reporters displayed a punctate pattern in the absence of the HCV replicon (Fig. 1c). These results indicate that cleavage of EGFP-IPS and RFPNLS-IPS are dependent on an intact NS3-4A recognition site and that HCV-dependent fluorescence relocalization (HDFR) can be used as a marker of viral replication. HCV exists as multiple genotypes, which exhibit extensive sequence divergence as well as differences in pathogenesis and treatment susceptibility12. Evasion of the innate immune response by cleavage of native IPS-1, however, is likely to be a conserved feature of HCV infection. In addition to JFH-1 (genotype 2a), Huh-7.5 cells harboring H77 (genotype 1a) or Con1 (genotype 1b) subgenomes10 were transduced with EGFP-IPS or EGFP-IPS(C508Y). Regardless of the HCV strain, EGFP-IPS transduction resulted in diffuse fluorescence, and EGFPIPS(C508Y) expression led to punctate EGFP (Fig. 1c). Whereas the lack of replicon systems for other genotypes precludes comprehensive analysis, these results indicate that cleavage of EGFP-IPS can be used as a marker of several diverse HCV strains. In contrast, replication of other positive-strand RNA viruses, such as yellow fever virus or Venezuelan equine encephalitis virus, did not lead to fluorescence relocalization (Supplementary Fig. 1a). These results suggest that the HDFR reporter system achieves a high level of HCV specificity combined with genotype independence. Although replicon-containing cells constitutively express the viral proteins, monitoring authentic virus infection is important for analyses of HCV biology and therapeutic inhibition. To determine the ability of HDFR to detect infection, we inoculated Huh-7.5 cells expressing RFP-NLS-IPS with an HCVcc reporter virus expressing secreted Gaussia luciferase, Jc1FLAG2(p7-nsGluc2A)13, followed by

1Center

for the Study of Hepatitis C, Laboratory of Virology and Infectious Disease, The Rockefeller University, New York, New York, USA. 2Division of Health Sciences and Technology, Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA. 3Howard Hughes Medical Institute, 4Division of Medicine, Brigham & Women’s Hospital, Boston, Massachusetts, USA. 5Present addresses: Hepregen Corporation, Medford, Massachusetts, USA (S.R.K.) and iTherX Pharmaceuticals, San Diego, California, USA (A.J.S.). Correspondence should be addressed to C.M.R. ([email protected]). Received 17 August 2009; accepted 4 January 2010; published online 31 January 2010; doi:10.1038/nbt.1604

nature biotechnology  VOLUME 28  NUMBER 2  FEBRUARY 2010

167

letters b CARD IPS-1

PRO

1

Reporter

Merge

c

TM

WT

C508Y

Huh-7.5

508 462

IPS-1

Huh-7.5

a

540

EGFP-IPS

RFP-NLS-IPS

SG-Con1

RFP

6 Huh-7.5

Log10 RLU

7

G

81 D

α-

C

-β

Ig

N

PB

S

4

PBS

IFN-β

IgG

α-CD81

RFP-NLS-IPS

Hoechst

5

IF

© 2010 Nature America, Inc. All rights reserved.

+HCVcc

Uninfected

8

SG-JFH

d

SG-H77

NLS RFP-NLS-IPS

SG-JFH

EGFP-IPS

EGFP-IPS

Figure 1  An IPS-1-based reporter system for detection of HCV infection. (a) Schematic of IPS-1 and derivative reporter constructs. The caspase recruitment domain (CARD) and proline-rich (PRO) domains of IPS-1 are indicated. The HCV NS3-4A protease cleaves IPS-1 at C508 (arrow). The C-terminal transmembrane domain (TM) directs IPS-1 to the outer membrane of mitochondria. EGFP-IPS encodes EGFP fused to residues 462–540 of IPS-1. RFP-NLS-IPS encodes a red fluorescent protein (mCherry or TagRFP) and an SV40 nuclear localization signal (NLS, PKKKRKVG) fused to residues 462–540 of IPS-1. (b) EGFP-IPS and RFP-NLS-IPS localize to mitochondria in Huh-7.5 cells. Native IPS-1, detected by immunofluorescent staining (IPS-1), and EGFP or RFP autofluorescence (Reporter) were visualized in untransduced (Huh-7.5) or transduced (EGFP-IPS or RFP-NLS-IPS) cells by confocal microscopy. Merge images also depict Hoechst nuclear dye (blue). (c) EGFP-IPS and RFP-NLS-IPS relocalize in response to HCV replication. Huh-7.5 cell lines harboring subgenomic (SG) neomycin-selectable replicons were transduced with lentiviruses expressing wild type (WT) or mutant (C508Y) EGFP-IPS or RFP-NLS-IPS. H77, genotype 1a; Con1, genotype 1b; JFH-1, genotype 2a. Wide-field fluorescence images of unfixed cells are shown. (d) RFPNLS-IPS relocalizes in HCV-infected cells. Huh-7.5 cells expressing RFP-NLS-IPS were infected with secreted Gaussia luciferase HCVcc reporter virus, Jc1FLAG2(p7-nsGluc2A), in the presence of PBS, IFN-β, blocking antibody (α-CD81) or isotype control (IgG). Luciferase activity in the culture supernatants (left) and reporter (RFP) or nuclear dye (Hoechst) fluorescence (right) were monitored at 48 h post-infection. Wide-field fluorescence images of fixed cells are shown. Scale bars, 20 µm. RLU, relative light units.

incubation for 48 h. Uninfected cells showed punctate fluorescence, whereas HCV-infected cultures displayed a distinct nuclear signal (Fig. 1d). Inoculation in the presence of IFN-β largely abolished the fluorescence translocation phenotype. Similarly, cells infected in conjunction with a monoclonal antibody targeting a known HCV entry factor (α-CD81) did not show nuclear fluorescence. Detection of Gaussia luciferase in the culture supernatants yielded corresponding results (Fig. 1d). Staining for viral replicase protein NS5A in infected EGFP-IPS–expressing cells supported the correlation between fluorescence relocalization and HCV replication at the single-cell level (Supplementary Fig. 1b). Monitoring infection by fluorescence relocalization does not require cells to be fixed, lysed or processed. These advantages suggested the possibility of real-time visualization of HCV infection in live cells. Huh-7.5 cells expressing RFP-NLS-IPS and a constitutive mitochondrial marker (EGFP-cytochrome c oxidase subunit VIII fusion protein; mito-EGFP) were inoculated with Jc1FLAG2 (p7-nsGluc2A) and monitored by live-cell microscopy beginning at

168

6 h post-infection (Fig. 2a,b; DMSO). Translocation of RFP-NLS to the nucleus could be detected as early as 10–12 h post-inoculation, with complete cleavage by 16–18 h (Fig. 1b and Supplementary Video 1a). In contrast, cells infected in the presence of a viral RNAdependent RNA polymerase inhibitor (2′CMA)14 showed very limited nuclear fluorescence (Fig. 2b and Supplementary Video 1b). We then investigated whether drug treatment of cells with established HCV infection could lead to observable reconstitution of mitochondrially localized fluorescence. RFP-NLS-IPS reporter cells were infected with Jc1FLAG2(p7-nsGluc2A) for 24 h before treatment with VX-950, an inhibitor of the NS3-4A15 protease, or DMSO as a vehicle control and imaged for an additional 24 h (Fig. 2a,c). Over the time course of the experiment, RFP-NLS-IPS localization in DMSO cells remained unchanged, whereas steady reconstitution of punctate fluorescence was seen in the majority of infected cells treated with the protease inhibitor. These results indicate that the reporter system can be used to visually monitor NS3-4A inhibition in real time (Supplementary Video 1c,d).

VOLUME 28  NUMBER 2  FEBRUARY 2010  nature biotechnology

letters Figure 2  Time-lapse live-cell imaging of HCVcc infection. (a) Schematic of live-cell imaging time course. Huh-7.5 cells stably expressing RFP-NLS-IPS and a mitochondrially targeted EGFP-cytochrome c oxidase subunit VIII fusion protein (mito-EGFP) were infected with HCVcc reporter virus, Jc1FLAG2(p7-nsGluc2A) (time = 0 h). (b) Cells were infected in the presence of DMSO or HCV RNA-dependent RNA polymerase inhibitor 2′CMA. (c) Cells were infected for 24 h before removal of the inoculum and addition of imaging medium containing DMSO or the NS3-4A protease inhibitor VX-950. Images were captured every 30 min starting at 6 h (b) or 24.5 h (c) post-infection. RFP fluorescence is shown in grayscale. Time (h) from the start of infection (b) or drug addition (c) are indicated. Scale bar, 20 µm. See Supplementary Video 1a–d for the full time course.

a

b

Time (h)

6

b

c

0

24

48

9

12

15

18

21

24

4

8

12

16

20

24

27

DMSO

2′CMA

c 2

28

© 2010 Nature America, Inc. All rights reserved.

DMSO

The availability of spectrally distinct HDFR reporters (EGFP-IPS and RFP-NLS-IPS) suggested the possibility of discerning infection VX950 in two separate cell populations simultaneously. We applied this advantage to visualize the recently described phenomenon of CD81-independent HCV infection. Circulating HCV enters hepatocytes through a complex pathway involving multiple co-receptors. CD81, SCARB1 and two tight-junction proteins, CLDN1 and OCLN, have been shown to be essential for this process16–19. Recent reports, however, suggest a second, CD81independent route of virus entry, which may entail particle transfer through close cell-cell contacts20,21. This transmission mode may be highly biologically relevant in the context of chronic infection, and the development of inhibitors targeting this entry pathway necessitates a reliable method of detection. To monitor routes of HCV spread, we employed cells expressing RFP-NLS-IPS and EGFP-IPS as producer and target populations, respectively. EGFP-IPS target cells were engineered to stably express a short hairpin (sh)RNA targeting CD81 (EGFP-IPS/CD81−) or an irrelevant sequence (EGFP-IPS/IRR), and tested for permissiveness to cell-free virus using an adapted HCVcc (J6/JFH clone 2), which exhibits superior titers to J6/JFH 22. Cells expressing CD81 shRNA had undetectable levels of CD81 protein (Supplementary Fig. 2a). At 48 h post-infection, the majority of EGFP-IPS/IRR cells exhibited diffuse EGFP, whereas EGFP-IPS/CD81− cells were largely nonpermissive (Fig. 3a). Fluorescence-activated cell sorting (FACS) analysis of fixed samples stained with an NS5A antibody supported these observations, indicating that <1% of EGFP-IPS/ CD81− cells were infected, compared to ~80% of the EGFP-IPS/IRR targets (Supplementary Fig. 2). To investigate HCV transmission in a mixed cell population, we pre-infected RFP-NLS-IPS cells for 36 h before co-culturing with EGFP-IPS/IRR or EGFP-IPS/CD81− target cells. Mixing uninfected RFP-NLS-IPS cells with either target population did not result in EGFP relocalization. In contrast, the majority of EGFP-IPS/IRR cells exhibited diffuse fluorescence upon co-culture with infected producers, presumably as a result of both CD81-dependent and CD81-independent infection routes. Culture of infected producer cells with EGFP-IPS/CD81− targets also led to EGFP-IPS cleavage, correlating with CD81-independent infection of 10–15% of cells (Fig. 3a and Supplementary Fig. 2). Use of the HDFR system in a mixed population can therefore provide a rapid visual assay for CD81independent HCV spread. In addition to exploiting a number of cellular factors during virus uptake, HCV affects host pathways during replication and

nature biotechnology  VOLUME 28  NUMBER 2  FEBRUARY 2010

­ athogenesis. The ability to correlate infection with altered cell biology p on a single-cell level would be invaluable to detecting virus-induced phenotypes. We therefore examined whether HDFR could be multi­ plexed with fluorescent markers of host processes. Stress pathways are charged with protecting the cell against various environmental insults, including heat shock, oxidative stress and virus infection. In response to stress, phosphorylation of translation factor eIF2α leads to the appearance of cytoplasmic stress granules, in which mRNAs are triaged and translation is stalled (reviewed in ref. 23). Not surprisingly, many viruses have evolved mechanisms to modulate the stress response and subvert translational suppression (reviewed in ref. 24); the effect of HCV on stress granule formation is unknown. We examined the stress response in Huh-7 cells expressing RFP-NLS-IPS and EGFP-tagged Ras-Gap-SH3 domain binding protein (EGFP-G3BP), a marker of stress granule formation25. Transduced cells were infected with Jc1FLAG2(p7-nsGluc2A) and subsequently monitored by live cell imaging; HCV replication was readily observed starting at 14–16 h post-infection. EGFP-G3BP exhibited a cytoplasmic distribution until ~30 h post-infection, when stress granule formation commenced in a fraction of HCV-positive cells. Interestingly, stress granules often appeared to be transient and, in some cases, formed and dissolved multiple times within a single cell (Fig. 3b). Stress granule formation was not observed in neighboring uninfected cells, nor in infected ­cultures treated with 2′CMA (Supplementary Video 2a–c), ­suggesting a dependence on HCV replication. Although the mechanisms under­ lying this phenomenon are still obscure, these observations ­support the utility of single-cell analysis and reporter multiplexing for ­discovery and dissection of virus-host interactions. Although we have demonstrated that HDFR can be used to detect infection of highly permissive hepatoma cell lines, we also sought to monitor HCV uptake by primary human hepatocytes. Primary cells are arguably the most relevant culture system in which to study HCV biology, but they have traditionally posed substantial challenges. Primary hepatocytes show low permissiveness for both viral entry and RNA replication, leading to poor expression of HCV-specific antigens, and making standard immuno­fluorescence techniques unreliable for visualizing infection. We reasoned that the sensitivity of HDFR might circumvent these

169

letters EGFP-IPS

a

CD81–

IRR

RFP-NLS-IPS + IRR

c

RFP-NLS-IPS + – CD81

WT+ 2′CMA

WT+ DMSO

C508Y+ DMSO

–HCVcc

+HCVcc

RFP-NLS-IPS

Co-culture 32.5

33.5

34.5

36.5

37

RFPNLS-IPS © 2010 Nature America, Inc. All rights reserved.

39.5

48

16 Infected cells/island (number)

32

EGFPG3BP

Merge

b

14 12 10 8 6 4 2 0 WT+ DMSO

WT+ C508Y+ 2′CMA DMSO

Figure 3  Use of the IPS-1-based reporter to expand HCVcc culture systems. (a) Co-culture of spectrally distinct HCV reporter cell lines for visualizing CD81-independent infection. Wide-field fluorescence images of unfixed mono- and co-cultures of Huh-7.5 cell lines expressing RFP-NLS-IPS or EGFP-IPS in the presence (+HCVcc) or absence (−HCVcc) of J6/JFH clone 2. EGFP-IPS cells stably express shRNA targeting CD81 (CD81 −) or an irrelevant sequence (IRR). Monocultures were infected for 72 h before imaging. In co-culture experiments, RFP-NLS-IPS cells were infected with HCVcc for 36 h before mixing with uninfected EGFP-IPS/IRR or EGFP-IPS/CD81 − cells. Co-cultures were incubated for an additional 48 h before imaging. (b) Multiplexing the HCV reporter with a marker of the stress response. An Huh-7 cell line expressing RFP-NLS-IPS and an EGFP-tagged stress granule marker (EGFP-G3BP) was infected with Jc1FLAG2(p7-nsGluc2A). Live-cell imaging was initiated at 6 h post-infection with images captured every 30 min. Montage shows selected time points beginning at 32 h post-infection; times (h) from the start of infection are indicated. See Supplementary Videos 2a–c. (c) Visualization of HCVcc infection in primary hepatocytes. Primary human hepatocytes maintained as MPCCs were transduced with lentiviruses expressing wild type (WT) or mutant (C508Y) RFP-NLS-IPS. At 24 h post-transduction, MPCCs were infected with Jc1FLAG2(p7-nsGluc2A). After 12 h, virus was removed and MPCC medium containing DMSO or 2′CMA was added. Unfixed MPCCs were imaged by wide-field fluorescence microscopy at 48 h post-infection. Representative phase contrast (top row) and corresponding RFP fluorescence images (middle row) are shown. Enlarged fluorescence images (bottom row) correspond to area denoted by white dotted box (middle row). The number of cells per MPCC island exhibiting nuclear RFP at 48 h post-infection is plotted for each condition. For WT RFP-NLS-IPS+DMSO, n = 40; WT RFP-NLS-IPS+2′CMA, n = 30; C508Y RFP-NLS-IPS+DMSO, n = 35. Bar, mean number of positive cells/island. Scale bars, 20 µm (a and b), 200 µm (c, top row), 10 µm (c, lower row).

difficulties, and explored its use for detecting HCV infection in the recently developed micropattern co-culture (MPCC) system26. MPCCs consist of primary adult human hepatocytes seeded on islands of collagen and surrounded by mouse fibroblast ‘feeder’ cells. These conditions allow hepatocytes to be maintained for extended periods without the rapid decline in cellular functions seen in conventional monocultures or random co-cultures 27. To visualize HCV infection in MPCC hepatocytes, we transduced cultures with RFPNLS-IPS or RFP-NLS-IPS(C508Y) and infected them 48 h later with Jc1FLAG2(p7-nsGluc2A), allowing MPCC infection to be monitored in parallel by Gaussia luciferase secretion28 (Supplementary Fig. 3). MPCC islands were examined by live cell microscopy and the cells per island exhibiting nuclear RFP were enumerated (Fig. 3c). In cultures transduced with RFP-NLS-IPS followed by DMSO treatment, ~98% of the islands observed contained cells with nuclear RFP, averaging four cells/island. In the RFP-NLS-IPS population treated with 2′CMA, only 6% of islands exhibited infected cells, corresponding to an average of 0.06 cells/island. Cells transduced with RFP-NLS-IPS(C508Y) did not show nuclear RFP in any of the islands examined (Fig. 3c). These results indicate that infection of primary hepatocytes can be readily detected using the HDFR system.

170

To our knowledge, visualization of HCV infection in live primary hepatocytes has not been demonstrated previously. Although systems for tracking HCV replication in culture have expanded rapidly in recent years, robust detection methods applicable to imaging of individual live cells have not been available. We describe a sensitive HCV reporter that allows easy distinction of infected and uninfected cells in live or fixed cultures by standard fluorescence microscopy. The robust signal of the reporter system derives from the efficiency of NS3-4A cleavage and the constitutive high-level expression of the substrate; nuclear translocation increases visualization, as the reporter becomes concentrated in a region with low autofluorescence, a particular advantage when working with hepatocytes. Although reporter cleavage does not occur in the absence of an active protease, transient signal may be expected in the presence of a polymerase inhibitor—this high sensitivity may have to be factored in as ‘background’ for some applications. The HDFR system does not require genetic modification of the viral genome and showed efficient detection of all HCV genotypes tested. This raises the possibility of using the reporter to identify novel infectious isolates directly from patient samples, potentially expanding the HCVcc system beyond the currently available genotype 2a strain. Coaxing

VOLUME 28  NUMBER 2  FEBRUARY 2010  nature biotechnology

letters

© 2010 Nature America, Inc. All rights reserved.

HCVcc to infect ­biologically relevant primary cell types may also be key to ­understanding authentic viral processes and patient-specific responses. The low level of replication observed in these cultures may reflect heterogeneity between individual cells or viral genomes, and underscores the value of single-cell analysis in dissecting the often subtle or variable phenotypes associated with chronic infection. Combining HDFR-based visualization with laser capture micro­scopy and analysis of neighboring infected and uninfected cells could begin to unravel the determinants of pathogenesis or virus control. The value of single-cell analysis was further illustrated by multiplexing HCV detection with a fluorescent marker of cellular stress, allowing direct visual correlation of viral and host events. Recent advances in automated microscopy and ‘high-content’ screening have made a large number of cellular phenotypes, including drug toxicity profiles, accessible to interrogation in a multiparametric format (reviewed in ref. 29). Addition of a robust fluorescent trans­location assay requiring minimal sample processing has the potential to integrate HCV research into this burgeoning field. We anticipate that the ability of HDFR to increase the flexibility and diversity of HCV culture systems will be important for basic virology and antiviral drug development. Methods Methods and any associated references are available in the online version of the paper at http://www.nature.com/naturebiotechnology/. Note: Supplementary information is available on the Nature Biotechnology website. Acknowledgments We acknowledge the expert support of The Rockefeller University Bioimaging Core Facility, with special thanks to A. North, S. Galdeen and S. Bhuvanendran. We thank The Rockefeller University Flow Cytometry Resource Center, supported by the Empire State Stem Cell Fund through NY State Department of Health (NYSDOH) contract no. C023046; opinions expressed here are solely those of the authors and do not necessarily reflect those of the Empire State Stem Cell Fund, the NYSDOH, or the State of NY. We are grateful to C. Stoyanov (The Rockefeller University) for YF17D(5′C25Venus2AUbi), J. Tazi for G3BP (Institut de Génétique Moléculaire de Montpellier) and I. Frolov (UTMB) for Venezuelan equine encephalitis virus-EGFP. We thank M. Holz, A. Forest, M. Panis and A. Webson for laboratory support and C. Murray for critical reading of the manuscript. This work was supported by Public Health Service grants R01 AI075099 (C.M.R.) and R01 DK56966 (S.N.B.). This work was also funded by the Office of the Director/National Institutes of Health (NIH) through the NIH Roadmap for Medical Research, Grant 1 R01 DK085713-01 (C.M.R. and S.N.B.). Information on this Roadmap Transformative R01 Program can be found at http://grants.nih. gov/grants/guide/rfa-files/RFA-RM-08-029.html. Additional funding was provided by the Greenberg Medical Research Institute and the Starr Foundation (C.M.R.). S.N.B. is an Howard Hughes Medical Investigator investigator. C.T.J. was supported by National Research Service Award DK081193; M.T.C. was supported by a Women & Science Fellowship; L.M.J.L. is supported by a Natural Sciences and Engineering Research Council of Canada fellowship. A.P. is a recipient of a Kimberly LawrenceNetter cancer research discovery fund award. AUTHOR CONTRIBUTIONS C.T.J. and C.M.R. designed the project, analyzed results and wrote the manuscript. C.T.J., M.T.C., L.M.J.L., A.J.S., S.R.K., T.S.O., A.P. and J.W.S. performed the experimental work. S.R.K., J.W.S., T.S.O., M.R.M. and S.N.B. contributed reagents and technical expertise. COMPETING INTERESTS STATEMENT The authors declare competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/naturebiotechnology/.

nature biotechnology  VOLUME 28  NUMBER 2  FEBRUARY 2010

Published online at http://www.nature.com/naturebiotechnology/. Reprints and permissions information is available online at http://npg.nature.com/ reprintsandpermissions/.

1. Shepard, C.W., Finelli, L. & Alter, M.J. Global epidemiology of hepatitis C virus infection. Lancet Infect. Dis. 5, 558–567 (2005). 2. Bartenschlager, R. & Sparacio, S. Hepatitis C virus molecular clones and their replication capacity in vivo and in cell culture. Virus Res. 127, 195–207 (2007). 3. Bartenschlager, R. Hepatitis C virus molecular clones: from cDNA to infectious virus particles in cell culture. Curr. Opin. Microbiol. 9, 416–422 (2006). 4. Meylan, E. et al. Cardif is an adaptor protein in the RIG-I antiviral pathway and is targeted by hepatitis C virus. Nature 437, 1167–1172 (2005). 5. Foy, E. et al. Control of antiviral defenses through hepatitis C virus disruption of retinoic acid-inducible gene-I signaling. Proc. Natl. Acad. Sci. USA 102, 2986–2991 (2005). 6. Li, X.D., Sun, L., Seth, R.B., Pineda, G. & Chen, Z.J. Hepatitis C virus protease NS3/4A cleaves mitochondrial antiviral signaling protein off the mitochondria to evade innate immunity. Proc. Natl. Acad. Sci. USA 102, 17717–17722 (2005). 7. Kawai, T. et al. IPS-1, an adaptor triggering RIG-I- and Mda5-mediated type I interferon induction. Nat. Immunol. 6, 981–988 (2005). 8. Seth, R.B., Sun, L., Ea, C.K. & Chen, Z.J. Identification and characterization of MAVS, a mitochondrial antiviral signaling protein that activates NF-kappaB and IRF 3. Cell 122, 669–682 (2005). 9. Xu, L.G. et al. VISA is an adapter protein required for virus-triggered IFN-beta signaling. Mol. Cell 19, 727–740 (2005). 10. Tscherne, D.M. et al. Superinfection exclusion in cells infected with hepatitis C virus. J. Virol. 81, 3693–3703 (2007). 11. Loo, Y.M. et al. Viral and therapeutic control of IFN-beta promoter stimulator 1 during hepatitis C virus infection. Proc. Natl. Acad. Sci. USA 103, 6001–6006 (2006). 12. Simmonds, P. Genetic diversity and evolution of hepatitis C virus-15 years on. J. Gen. Virol. 85, 3173–3188 (2004). 13. Marukian, S. et al. Cell culture-produced hepatitis C virus does not infect peripheral blood mononuclear cells. Hepatology 48, 1843–1850 (2008). 14. Carroll, S.S. et al. Inhibition of hepatitis C virus RNA replication by 2′-modified nucleoside analogs. J. Biol. Chem. 278, 11979–11984 (2003). 15. Lin, K., Perni, R.B., Kwong, A.D. & Lin, C. VX-950, a novel hepatitis C virus (HCV) NS3–4A protease inhibitor, exhibits potent antiviral activities in HCV replicon cells. Antimicrob. Agents Chemother. 50, 1813–1822 (2006). 16. Pileri, P. et al. Binding of hepatitis C virus to CD81. Science 282, 938–941 (1998). 17. Scarselli, E. et al. The human scavenger receptor class B type I is a novel candidate receptor for the hepatitis C virus. EMBO J. 21, 5017–5025 (2002). 18. Evans, M.J. et al. Claudin-1 is a hepatitis C virus co-receptor required for a late step in entry. Nature 446, 801–805 (2007). 19. Ploss, A. et al. Human occludin is a hepatitis C virus entry factor required for infection of mouse cells. Nature 457, 882–886 (2009). 20. Timpe, J.M. et al. Hepatitis C virus cell-cell transmission in hepatoma cells in the presence of neutralizing antibodies. Hepatology 47, 17–24 (2008). 21. Witteveldt, J. et al. CD81 is dispensable for hepatitis C virus cell-to-cell transmission in hepatoma cells. J. Gen. Virol. 90, 48–58 (2009). 22. Diamond, D.L. et al. Temporal proteome and lipidome profiles reveal HCV-associated reprogramming of hepatocellular metabolism and bioenergetics. PLoS Pathog. 6, e1000719 (2010). 23. Anderson, P. & Kedersha, N. RNA granules. J. Cell Biol. 172, 803–808 (2006). 24. Schutz, S. & Sarnow, P. How viruses avoid stress. Cell Host Microbe 2, 284–285 (2007). 25. Tourriere, H. et al. The RasGAP-associated endoribonuclease G3BP assembles stress granules. J. Cell Biol. 160, 823–831 (2003). 26. Khetani, S.R. & Bhatia, S.N. Microscale culture of human liver cells for drug development. Nat. Biotechnol. 26, 120–126 (2008). 27. Bhatia, S.N., Balis, U.J., Yarmush, M.L. & Toner, M. Effect of cell-cell interactions in preservation of cellular phenotype: cocultivation of hepatocytes and nonparenchymal cells. FASEB J. 13, 1883–1900 (1999). 28. Ploss, A. et al. Persistent hepatitis C virus infection in microscale primary human hepatocyte cultures. Proc. Natl. Acad. Sci. USA (in the press). 29. Wollman, R. & Stuurman, N. High throughput microscopy: from raw images to discoveries. J. Cell Sci. 120, 3715–3722 (2007).

171

ONLINE METHODS

© 2010 Nature America, Inc. All rights reserved.

Cell culture. Huh-7 and Huh-7.5 cells were cultured at 37 °C, 5% CO 2 in Dulbecco’s Modified Eagle Medium (DMEM, Invitrogen) containing 10% FBS and 0.1 mM nonessential amino acids (NEAA) (complete media), unless otherwise noted. For time-lapse imaging, cells were maintained in CO2-independent media (Invitrogen) containing 10% FBS, 0.1 mM NEAA, 1 mM sodium pyruvate and 2 mM l-glutamine (imaging media). Huh-7.5 cell lines harboring selectable subgenomic replicons10 were grown in complete media containing 0.5 mg/ml G418. Huh-7.5 cells stably expressing the pLenti-3′-U6-EC-EP7 vector encoding an shRNA against CD81 (nt 268-288, cDNA numbering) or a predicted nontargeting sequence (IRR) have been previously described21 and were grown in complete media containing 6 µg/ml blasticidin. MPCC cultures were generated as previously described 26 and maintained in high glucose DMEM, 10% FBS, 0.5 U/ml insulin, 7 ng/ml glucagon, 7.5 µg/ml hydrocortisone and 1% penicillin-streptomycin. For HCV inhibition, culture media was supplemented with 0.2% DMSO, 14 mM 2′CMA, 24 mM VX-950, or 1,000 U/ml IFN-β (Peprotech). For neutralization experiments, HCVcc infections were performed in the presence of 10 µg/ml antibody directed against CD81 (BD Biosciences) or an isotype control (IgG1κ, BD Biosciences). Virus stocks. Jc130 and Jc1FLAG2(p7-nsGluc2A)13 are fully infectious HCVcc viruses that have been previously described. J6/JFH clone 2 is a passaged derivative of J6/JFH31 that contains a number of adaptive mutations that increase infectious titers22. Bi-Ypet-Jc1FLAG2 is a bicistronic reporter virus in which the HCV IRES drives expression of Ypet, an EGFP variant with enhanced brightness, in the first cistron; the EMCV IRES drives expression of the second cistron, which encodes the Jc1 polyprotein with a FLAG epitope at the N terminus of E2. YF17D(5′C25Venus2AUbi) is a monocistronic yellow fever reporter virus (kindly provided by C. Stoyanov, The Rockefeller University) encoding the Venus fluorescent protein, a yellow-shifted variant of EGFP. Venezuelan equine encephalitis virus-EGFP (kindly provided by I. Frolov, UTMB) is a double subgenomic EGFP reporter virus derived from the TC83 vaccine strain of Venezuelan equine encephalitis. Virus stocks were generated by electroporation of in vitro transcribed RNAs into the appropriate cell lines, as described previously31–33. Plasmid constructs. Constructs were created by standard methods; plasmid and primer sequences are available upon request. IPS-1–based reporters and subcellular localization markers were constructed in a lentivirus backbone derived from TRIP-EGFP34. Residues 462–540 of IPS-1 (IPS) were obtained by PCR from a human hepatocyte cDNA library (Ambion) and inserted into the BsrGI/XhoI sites of TRIP-EGFP to generate TRIP-EGFP-IPS. IPS-1 mutation C508Y was generated by overlap PCR mutagenesis. TRIP-mCherry and TRIP-TagRFP were used to construct TRIP-RFP-NLS-IPS plasmids. TRIPmCherry was derived from the TRIP-mCherry-CLDN1 plasmid18. TagRFP sequence was obtained from pTagRFP-C (Evrogen). TRIP-RFP-NLS-IPS plasmids encode the SV40 nuclear localization signal (NLS, PKKKRKVG) and IPS fused to the C terminus of RFP. TRIP-mito-EGFP, encodes the mitochondrial targeting sequence of human cytochrome c oxidase subunit VIII fused to the N terminus of EGFP. TRIP-EGFP-G3BP encodes the Ras-Gap-SH3 domain binding protein (G3BP, kindly provided by J. Tazi25) fused to the C terminus of EGFP. Generation of lentivirus pseudoparticles and transductions. Pseudoparticles (pp) were generated by co-transfection of 293T cells with TRIP provirus, HIV gag-pol, and vesicular stomatitis virus envelope protein G (VSV-G) plasmids using a weight ratio of 1:0.8:0.2, as described previously18. Huh-7 and Huh7.5 cells were transduced by incubation for 6 h at 37 °C with TRIPpp diluted 1:3 in complete media supplemented with 4 µg/ml polybrene and 20 mM

nature biotechnology

HEPES. In some cases, transduced hepatoma cell populations were enriched using a FACSAria II high-speed flow cytometry cell sorter (BD Biosciences). For transduction of MPCC, cultures were first treated for ~20 s with 0.025% EDTA/Trypsin, before washing and overnight incubation with 1:3 diluted TRIPpp stocks. Immunofluorescence staining and FACS analysis. For NS5A immunostaining, cells grown on glass coverslips were washed with PBS before fixation in formaldehyde (3.7% wt/vol in PBS) and incubation in blocking buffer (3% BSA, 0.2% saponin in PBS). After overnight incubation at 4 °C with monoclonal antibody 9E10 (ref. 31) (1:2,000 in blocking buffer) and 1 h incubation at 25 °C with AlexFluor-594-conjugated secondary antibody to mouse (Invitrogen, 1:1,000 in blocking buffer), cell nuclei were stained with Hoechst dye (Thermo Scientific). Coverslips were mounted using ProLong Gold Antifade reagent (Invitrogen). For IPS-1 immunostaining, cells grown on glass bottom multiwall plates (MatriCal) were washed with PBS and fixed with 2% paraformaldehyde. After incubation in blocking buffer for 30 min, and polyclonal anti-IPS-1 antibody (Cell Signaling Technology, 1:50 in blocking buffer) for 2 h, AlexaFluor-555 or AlexaFluor-488 conjugated anti-rabbit IgG secondary antibodies (Cell Signaling Technology, 1:500 in blocking buffer) were added for 1 h at 25 °C. Cells were then stained with Hoechst nuclear dye followed by application of ProLong Gold Antifade reagent. For FACS analysis, cells were harvested using AccuMax (eBioscience) and fixed using Fixation/Permeabilization buffer (BD Biosciences) for 10 min at 4 °C. Fixed cells were washed with BD Perm/ Wash buffer (BD Biosciences), incubated 30 min at 25 °C with AlexaFluor-647conjugated 9E10 antibody (1:4000 in BD Perm/Wash buffer), washed twice with BD Perm/Wash buffer and once with FACS buffer (PBS/3% FBS) before analysis using a BD LSR II flow cytometer and BD FACSDiva software. Analysis of FACS data was performed using FlowJo software. Microscopy. Wide-field fluorescent images were captured using an Eclipse TE300 (Nikon) inverted microscope and SPOT imaging software or the Discovery-1 system and MetaXpress software (Molecular Devices). Confocal imaging of fixed samples was performed using an inverted Axiovert 200 laser scanning microscope (Zeiss). For long-term live cell imaging, cells were grown on rat-tail collagen–coated (BD Biosciences) no. 1.5 Lab-Tek II 4-chambered (Thermo Fisher Scientific) coverslips. Live cells maintained at 37 °C in imaging media were imaged using a Zeiss Axiovert 200 inverted microscope equipped with an UltraView spinning disk confocal head (Perkin-Elmer), an Orca ERcooled CCD camera (Hamamatsu), a 20×/0.75 N.A. Plan-Apochromat objective, and an environmental chamber (Solent Scientific). Solid-state 491 and 561 nm lasers (Spectral Applied) and ET 530/50 and ET 605/70 emission filters (Chroma) were used for excitation and emission of EGFP and RFP fluorescence, respectively. Alternatively, time-lapse images were captured using an Olympus IX71 inverted microscope equipped with an Orca ER cooled CCD camera, a 20×/0.75 N.A. UPlan SApo objective and an environmental ­chamber. Image acquisition was performed using Metamorph (Molecular Devices) and processing was performed using ImageJ64. 30. Pietschmann, T. et al. Construction and characterization of infectious intragenotypic and intergenotypic hepatitis C virus chimeras. Proc. Natl. Acad. Sci. USA 103, 7408–7413 (2006). 31. Lindenbach, B.D. et al. Complete replication of hepatitis C virus in cell culture. Science 309, 623–626 (2005). 32. Lindenbach, B.D. & Rice, C.M. Trans-complementation of yellow fever virus NS1 reveals a role in early RNA replication. J. Virol. 71, 9608–9617 (1997). 33. Petrakova, O. et al. Noncytopathic replication of Venezuelan equine encephalitis virus and eastern equine encephalitis virus replicons in mammalian cells. J. Virol. 79, 7597–7608 (2005). 34. Zennou, V. et al. HIV-1 genome nuclear import is mediated by a central DNA flap. Cell 101, 173–185 (2000).

doi:10.1038/nbt.1604

letters

Rational design of cationic lipids for siRNA delivery

© 2010 Nature America, Inc. All rights reserved.

Sean C Semple1,6, Akin Akinc2,6, Jianxin Chen1,5, Ammen P Sandhu1, Barbara L Mui1,5, Connie K Cho1, Dinah W Y Sah2, Derrick Stebbing1, Erin J Crosley1, Ed Yaworski1, Ismail M Hafez3, J Robert Dorkin2, June Qin2, Kieu Lam1, Kallanthottathil G Rajeev2, Kim F Wong3, Lloyd B Jeffs1, Lubomir Nechev2, Merete L Eisenhardt1, Muthusamy Jayaraman2, Mikameh Kazem3, Martin A Maier2, Masuna Srinivasulu4, Michael J Weinstein2, Qingmin Chen2, Rene Alvarez2, Scott A Barros2, Soma De2, Sandra K Klimuk1, Todd Borland2, Verbena Kosovrasti2, William L Cantley2, Ying K Tam1,5, Muthiah Manoharan2, Marco A Ciufolini4, Mark A Tracy2, Antonin de Fougerolles2, Ian MacLachlan1, Pieter R Cullis3, Thomas D Madden1,5 & Michael J Hope1,5 We adopted a rational approach to design cationic lipids for use in formulations to deliver small interfering RNA (siRNA). Starting with the ionizable cationic lipid 1,2-dilinoleyloxy-3dimethylaminopropane (DLinDMA), a key lipid component of stable nucleic acid lipid particles (SNALP) as a benchmark, we used the proposed in vivo mechanism of action of ionizable cationic lipids to guide the design of DLinDMA-based lipids with superior delivery capacity. The best-performing lipid recovered after screening (DLin-KC2-DMA) was formulated and characterized in SNALP and demonstrated to have in vivo activity at siRNA doses as low as 0.01 mg/kg in rodents and 0.1 mg/kg in nonhuman primates. To our knowledge, this represents a substantial improvement over previous reports of in vivo endogenous hepatic gene silencing. A key challenge in realizing the full potential of RNA interference (RNAi) therapeutics is the efficient delivery of siRNA, the molecules that mediate RNAi. The physicochemical characteristics of siRNA— high molecular weight, anionic charge and hydrophilicity—prevent passive diffusion across the plasma membrane of most cell types. Therefore, delivery mechanisms are required that allow siRNA to enter cells, avoid endolysosomal compartmentalization and ­localize in the cytoplasm where it can be loaded into the RNA-induced Figure 1  Proposed mechanism of action for membrane disruptive effects of cationic lipids and structural diagram of DLinDMA divided into headgroup, linker and hydrocarbon chain domains. In isolation, cationic lipids and endosomal membrane anionic lipids such as phosphatidylserine adopt a cylindrical molecular shape, which is compatible with packing in a bilayer configuration. However, when cationic and anionic lipids are mixed together, they combine to form ion pairs where the cross-sectional area of the combined headgroup is less than that of the sum of individual headgroup areas in isolation. The ion pair therefore adopts a molecular ‘cone’ shape, which promotes the formation of inverted, nonbilayer phases such as the hexagonal HII phase illustrated. Inverted phases do not support bilayer structure and are associated with membrane fusion and membrane disruption9,21.

s­ ilencing complex. To date, formulation in lipid nanoparticles (LNPs) ­represents one of the most widely used strategies for in vivo delivery of siRNA1,2. LNPs represent a class of particles comprised of different lipid compositions and ratios as well as different sizes and structures formed by different methods. A family of LNPs, SNALP3–6, is charac­ terized by very high siRNA-encapsulation efficiency and small, uniformly sized particles, enabled by a controlled step-wise dilution methodology. LNPs, including SNALP, have been successfully used to silence therapeutically relevant genes in nonhuman primates6–8 and are currently being evaluated in several clinical trials. An empirical, combinatorial chemistry–based approach recently identified novel materials for use in LNP systems7. A key feature of this approach was the development of a one-step synthetic strategy that allowed the rapid generation of a diverse library of ~1,200 compounds. This library was then screened for novel materials capable of mediating efficient delivery of siRNA in vitro and in vivo. Here, we instead used a medicinal chemistry (that is, structure-activity relationship) approach, guided by the putative in vivo mechanism of action of ionizable cationic lipids, for rational lipid design. Specifically, we hypothesized that after endocytosis, the cationic lipid interacts with naturally occurring anionic phospholipids in the endosomal membrane, forming ion pairs that adopt nonbilayer structures and disrupt membranes (Fig. 1)9–12. We previously advanced the concept +

–

Headgroup

+–

Linker

Cylindrical shape supports bilayer structure

Bilayer

Hydrocarbon chains

Cone shape disrupts bilayer structure

Hexagonal HII

DLinDMA

1Tekmira Pharmaceuticals, Burnaby, British Columbia, Canada. 2Alnylam Pharmaceuticals, Cambridge, Massachusetts, USA. 3Department of Biochemistry and Molecular Biology and 4Department of Chemistry, University of British Columbia, Vancouver, British Columbia, Canada. 5Present address: Alcana Technologies, Vancouver, British Columbia, Canada. 6These authors contributed equally to this work. Correspondence should be addressed to S.C.S. ([email protected]) or A.A. ([email protected]).

Received 16 September 2009; accepted 17 December 2009; published online 17 January 2010; doi:10.1038/nbt.1602

172

VOLUME 28  NUMBER 2  FEBRUARY 2010  nature biotechnology

letters b

120 100 80 60 40 20 0 0.01 0.1 1 10 100 Factor VII siRNA dose (mg/kg)

c

Relative serum factor VII protein (%)

Relative serum factor VII protein (%)

a

100 80 60 40 20 0 0.01 0.1 1 10 Factor VII siRNA dose (mg/kg) O

Me2N

O

Me2N

R

O R

O

R R

DLinDAP O

Me2N

DLin-K-DMA

R

O R

O O

DLinDMA O Me2N

120

O

R R

DLin-KC2-DMA

O Me2N

O

R R

DLin-KC3-DMA

O Me2N

O

R R

DLin-KC4-DMA

Figure 2  In vivo evaluation of novel cationic lipids. (a) Silencing activity of DLinDAP ( ), DLinDMA ( ), DLin-K-DMA ( ) and DLinKC2-DMA ( • ) screening formulations in the mouse Factor VII model. All LNP-siRNA systems were prepared using the preformed vesicle (PFV) method and were composed of ionizable cationic lipid, DSPC, cholesterol and PEG-lipid (40:10:40:10 mol/mol) with a Factor VII siRNA/total lipid ratio of ~0.05 (wt/wt). Data points are expressed as a percentage of PBS control animals and represent group mean (n = 5) ± s.d., and all formulations were compared within the same study. (b) Influence of headgroup extensions on the activity of DLin-K-DMA. DLin-K-DMA ( ) had additional methylene groups added between the DMA headgroup and the ketal ring linker to generate DLin-KC2-DMA ( • ), DLin-KC3-DMA ( ) and DLin-KC4-DMA ( ). The activity of PFV formulations of each lipid was assessed in the mouse Factor VII model. Data points are expressed as a percentage of PBS control animals and represent group mean (n = 4) ± s.d. (c) Chemical structures of novel cationic lipids.

© 2010 Nature America, Inc. All rights reserved.

R=

of using ionizable cationic lipids with pKas < pH 7.0 to efficiently formulate nucleic acids at low pH and maintaining a neutral or low cationic surface charge density at pH 7.4 (ref. 13). This strategy should provide better control of the circulation properties of these systems and reduce nonspecific disruption of plasma membranes. As positive charge density is minimal in the blood but increases substantially in the acidic environment of the endosome, this should activate the membrane-destabilizing property of the LNP. Although these attributes may account for the activity of these systems upon internalization by hepatocytes, they do not necessarily explain the high levels of hepatic biodistribution observed for many LNPs, including SNALP. Although these LNPs do not specifically include a targeting ligand to direct them to hepatocytes after systemic administration, it is possible that these LNPs associate with one or more proteins in plasma that may promote hepatocyte endocytosis. The ionizable cationic lipid DLinDMA has proven to be highly effective in SNALP, has been extensively tested in rodents and nonhuman primates, and is now being evaluated in human clinical trials. Therefore, we selected it as the starting point for the design and synthesis of novel lipids. We chose the mouse Factor VII model 7, as the primary in vivo screening system to assess functional LNPmediated delivery to hepatocytes. Briefly, C57BL/6 mice received a single dose of LNP-formulated Factor VII siRNA through bolus tail vein injection and serum was collected from animals 24 h after administration to analyze Factor VII protein level. The initial screening of LNP-siRNA systems was conducted using LNPs prepared by a preformed vesicle method14 and composed of ionizable cationic lipid, distearoylphosphatidylcholine (DSPC), cholesterol and PEGlipid (40:10:40:10 mol/mol), with a Factor VII siRNA/total lipid ratio of ~0.05 (wt/wt). Although not a bilayer-destabilizing lipid, a small amount of phosphatidylcholine was incorporated into the LNP to help stabilize the LNP both during formulation and while it was in circulation. A short acyl chain PEG-lipid was incorporated into the LNP to control particle size during formulation, but is designed to leave the LNP rapidly upon intravenous injection. As our goal was to identify novel ionizable cationic lipids for use in LNPs, we aimed to minimize other effects by using a single robust composition and set of formulation conditions suitable for all novel lipids tested. The preformed vesicle method employing the composition listed above provides a convenient platform for such testing, but uses a different formulation process, a different lipid composition and a different siRNA/lipid ratio than SNALP. The structure of DLinDMA can be divided into three main regions: the hydrocarbon chains, the linker and the headgroup (Fig. 1). A detailed structure-function study to

nature biotechnology  VOLUME 28  NUMBER 2  FEBRUARY 2010

investigate the impact of increasing the number of cis double bonds in the hydrocarbon chains found the linoleyl lipid containing two double bonds per hydrocarbon chain (DLinDMA) to be optimal 15. We therefore maintained the linoleyl hydrocarbon chains present in DLinDMA as an element in our lipid design, and focused on optimizing the linker and headgroup moieties. The linker region in a bilayer structure resides at the membrane interface, an area of transition between the hydrophobic membrane core and hydrophilic headgroup surface. Our approach to linker modification of DLinDMA involved introducing groups expected to exhibit different rates of chemical or enzymatic stability and to span a range of hydrophilicity. A variety of these rationally designed lipids were made, characterized and tested (Supplementary Syntheses 1 and Supplementary Table 1). LNPs based on the ester-containing lipid DLinDAP showed substantially reduced in vivo activity compared to LNPs based on the alkoxy-containing lipid DLinDMA (Fig. 2). Further, LNPs based on DLin-2-DMAP, a lipid with one alkoxy linkage and one ester linkage, yielded activity intermediate between DLinDAP- and DLinDMA-based LNPs (Supplementary Table 1). Although it is uncertain why the ester-containing lipids are considerably less active in vivo, we speculate that the diester lipid (DLinDAP) is relatively inactive because it is more readily hydrolyzed in vivo than the alkoxy analog (DLinDMA), and therefore, unable to either protect the siRNA adequately before release from the endosome and/or survive long enough in the endosome to disrupt the membrane. These hypotheses are being investigated. LNPs based on lipids ­containing carbamate or thioether linkages also resulted in dramatically reduced in vivo activity. Interestingly, the introduction of a ketal ring linker into DLinDMA resulted in LNPs that were ~2.5-fold more potent in reducing serum Factor VII protein levels relative to the DLinDMA benchmark, with an ED50 (that is, dose to achieve 50% gene silencing) of ~0.4 mg/kg versus 1 mg/kg, respectively (Fig. 2). Given the importance of positive charge in the mechanismof-action hypothesis guiding the lipid design, the effects of structural changes in the amine-based headgroup were investigated in the context of DLin-K-DMA as the new benchmark lipid. A series of headgroup modifications were made, characterized and tested to explore the effects of size, acid-dissociation constant and number of ionizable groups (Supplementary Syntheses 2 and Supplementary Table 2). Piperazino, morpholino, trimethylamino or bis-dimethylamino modifi­ cations tested were not better than the benchmark dimethylamino headgroup of DLin-K-DMA. As an additional parameter, the distance between the dimethylamino group and the dioxolane linker was varied by introducing additional methylene groups. This parameter can

173

letters Table 1  Biophysical parameters and in vivo activities of LNPs containing novel lipids Apparent LII to HII phase transition temperature (°C)b lipid pKaa

Cationic lipid

In vivo ED50 (mg/kg)

DLinDMA

6.8 ± 0.10

27

~1

DLinDAP

6.2 ± 0.05

26

40–50

DLin-K-DMA

5.9 ± 0.03

19

~0.4

DLin-KC2-DMA

6.7 ± 0.08

20

~0.1

DLin-KC3-DMA

7.2 ± 0.05

18

~0.6

DLin-KC4-DMA

7.3 ± 0.07

18

>3

values ± s.d. (n = 3 to 9). bLII to HII phase transition was measured at pH 4.8 in equi­ molar mixtures with DSPS, using differential scanning calorimetric, repeat scans reproducible to within 0.1 °C.

affect both the pKa of the amine headgroup as well as the distance and flexibility of the charge presentation relative to the lipid bilayer interface. Inserting a single additional methylene group into the headgroup (DLin-KC2-DMA) produced a dramatic increase in potency relative to DLin-K-DMA. The ED50 for this lipid was ~0.1 mg/kg, making it fourfold more potent than DLin-K-DMA and tenfold more potent than the DLinDMA benchmark when compared head-to-head in the Factor VII model (Fig. 2a). Further extension of the tether with additional methylene groups, however, substantially decreased activity, with an ED50 of ~0.6 mg/kg for DLin-KC3-DMA and >3 mg/kg for DLin-KC4-DMA (Fig. 2b). As changes in lipid design and chemistry may affect the pharmaco­ kinetics, target tissue accumulation and intracellular delivery of LNP formulations, we investigated the relative importance of these para­ meters on LNP activity at an early stage in this research program. Several of the novel lipids were incorporated into LNP-siRNA formulations containing cyanine dye (Cy3)-labeled siRNA. Plasma, liver and spleen levels of siRNA were determined at 0.5 and 3 h after injection at siRNA doses of 5 mg/kg, and the results are presented in Supplementary Table 3. In general, formulations that were the most active in the mouse Factor VII screens achieved the highest liver levels of siRNA at 0.5 h; however, delivery of siRNA to the target tissue was not the primary factor responsible for activity. This is supported by the observations that most formulations accumulated in the liver and spleen quite quickly and that some formulations with similar liver levels of siRNA had large differences in activity. Moreover, plasma pharmacokinetics alone did not predict activity. For example, although DLin-KC2-DMA and DLinDMA had virtually indistinguishable blood pharmacokinetic profiles in mice (data not shown), the activity of DLin-KC2-DMA in LNPs is approximately tenfold greater than the

a

b 1.4

•

174

Relative liver TTR/GAPDH mRNA levels

Figure 3  Efficacy of KC2-SNALP in rodents and nonhuman primates. (a) Improved 120 efficacy of KC2-SNALP relative to the 100 initial screening formulation tested in mice. The in vivo efficacy of KC2-SNALP 80 () was compared to that of the unoptimized DLin-KC2-DMA screening (that is, PFV) 60 formulation ( ) in the mouse Factor VII model. Data points are expressed as a percentage 40 of PBS control animals and represent group mean (n = 5) ± s.d. (b) Efficacy of KC2-SNALP 20 in nonhuman primates. Cynomolgus monkeys 0 (n = 3 per group) received a total dose of either 0.001 0.01 0.1 1 10 0.03, 0.1, 0.3 or 1 mg/kg siTTR, or 1 mg/kg Factor VII siRNA dose (mg/kg) siApoB formulated in KC2-SNALP or PBS as 15-min intravenous infusions (5 ml/kg) through the cephalic vein. Animals were euthanized 48 h after administration. TTR mRNA levels relative to liver samples. Data points represent group mean ± s.d. *, P < 0.05; **, P < 0.005. Relative serum factor VII protein (%)

© 2010 Nature America, Inc. All rights reserved.

apK a

same formulation with DLinDMA. Taken together, these results led us to conclude that rapid target tissue accumulation was important, but not sufficient, for activity. Moreover, other parameters were more critical for maximizing the activity of LNP-siRNA formulations. Two important parameters underlying lipid design for SNALPmediated delivery are the pKa of the ionizable cationic lipid and the abilities of these lipids, when protonated, to induce a nonbilayer (hexagonal HII) phase structure when mixed with anionic ­ lipids. The pKa of the ionizable cationic lipid determines the surface charge on the LNP under different pH conditions. The charge state at ­ physiologic pH (e.g., in circulation) can influence plasma protein adsorption, blood clearance and tissue distribution behavior16, whereas the charge state at acidic pH (e.g., in endosomes) can influence the ­ ability of the LNP to combine with endogenous anionic lipids to form ­endosomolytic nonbilayer structures9. Consequently, the ability of these lipids to induce HII phase structure in mixtures with anionic lipids is a ­measure of their bilayer-destabilizing capacity and relative endosomolytic potential. The fluorescent probe 2-(p-toluidino)-6-napthalene sulfonic acid (TNS), which exhibits increased fluorescence in a hydrophobic environment, can be used to assess surface charge on lipid bilayers. Titrations of surface charge as a function of pH can then be used to determine the apparent pKa of the lipid in the bilayer (hereafter referred to as pKa) of constituent lipids17. Using this approach, the pKa values for LNPs containing DLinDAP, DLinDMA, DLin-K-DMA, DLin-KC2-DMA, DLin-KC3-DMA and DLin-KC4-DMA were determined (Table 1). The relative ability of the protonated form of the ionizable cationic lipids to induce HII phase structure in anionic lipids was ascertained by measuring the bilayer-to-hexagonal HII transition temperature (TBH) in equimolar mixtures with distearoylphosphatidylserine (DSPS) at pH 4.8, using 31P NMR18 and differential scanning calorimetric analyses19. Both techniques gave similar results. The data presented in Table 1 indicate that the highly active lipid DLin-KC2-DMA has pKa and TBH values that are theoretically favorable for use in siRNA delivery systems. The pKa of 6.4 indicates that LNPs based on DLin-KC2-DMA have limited surface charge in circulation, but will become positively charged in endosomes. Further, the TBH for DLin-KC2-DMA is 7 °C lower than that for DLinDMA, suggesting that this lipid has improved capacity for destabilizing ­bilayers. However, the data also demonstrate that pKa and TBH do not fully account for the in vivo activity of lipids used in LNPs. For ­example, although DLin-KC3-DMA and DLin-KC4-DMA have identical pKa and TBH values, DLin-KC4-DMA requires a more than fivefold higher dose to achieve the same activity in vivo. Moreover,

1.2 1.0 0.8 0.6 0.4

*

* **

0.2 0

1 0.03 0.1 0.3 1 mg/kg mg/kg mg/kg mg/kg mg/kg PBS siApoB siTTR

GAPDH mRNA levels were determined in

VOLUME 28  NUMBER 2  FEBRUARY 2010  nature biotechnology

letters Table 2  Clinical chemistry and hematology parameters for KC2-SNALP–treated rats Vehicle

siRNA dose (mg/kg)a

PBS KC2-SNALP

1

ALT (U/L)

AST (U/L)

Total Bilirubin (mg/dl)

BUN (mg/dl)

RBC (× 106/µl)

Hemoglobin (g/dl)

WBC (× 103/µl)

PLT (× 103/µl)

56 ± 16

109 ± 31

2±0

4.8 ± 0.8

5.5 ± 0.3

11.3 ± 0.4

11 ± 3

1,166 ± 177

58 ± 22

100 ± 14

2±0

4.4 ± 0.6

5.6 ± 0.2

11.6 ± 0.6

13 ± 2

1,000 ± 272

KC2-SNALP

2

73 ± 9

KC2-SNALP

3

87 ± 19

81 ± 10 100 ± 30

2.2 ± 0.4

4.3 ± 0.6

5.9 ± 0.3

11.6 ± 0.3

13 ± 4

1,271 ± 269

2±0

5.0 ± 0.8

6.0 ± 0.2

11.9 ± 0.4

15 ± 2

958 ± 241

aNontargeting,

© 2010 Nature America, Inc. All rights reserved.

luciferase siRNA. Sprague-Dawley rats (n = 5) received 15-min intravenous infusions of KC2-SNALP formulated siRNA at different dose levels. Blood samples were taken 24 h after administration. ALT, alanine aminotransferase; AST, aspartate aminotransferase; BUN, blood urea nitrogen; RBC, red blood cells; WBC, white blood cells; PLT, platelets.

DLin-KC2-DMA and DLin-KC4-DMA, which have very similar pKa and TBH values, exhibit a >30-fold difference in in vivo activity. This result suggests that other parameters, such as the distance and flexibility of the charged group relative to the lipid bilayer interface, may also be important. Thus, although the biophysical parameters of pKa and TBH are useful for guiding lipid design, the results presented in Table 1 support the strategy of testing variants of lead lipids, even ones with very similar pKa and TBH values. The lipid composition chosen for the initial formulation and screening of novel ionizable cationic lipids (cationic lipid/DSPC/ cholesterol/PEG-lipid = 40:10:40:10 mol/mol, siRNA/total lipid ~ 0.05 wt/wt) was useful for determining the rank-order potency of novel lipids, but is not necessarily optimal for in vivo delivery. In addition, the in vivo activity of resultant LNP-siRNA formulations is affected by the formulation process employed and the resulting particle structure. Improvements in activity were possible with the preformed vesicle process by modifying and optimizing lipid ratios and formulation conditions (results not shown). However, we chose to further validate DLin-KC2-DMA activity specifically in the context of the SNALP platform, currently the most advanced LNP formulation for delivery of siRNA in vivo. We therefore tested in vivo a version of SNALP (termed KC2-SNALP), which uses less PEG lipid than reported previously6 and in which DLinDMA was replaced with DLinKC2-DMA. The incorporation of DLin-KC2-DMA into SNALP led to a marked improvement in potency in the mouse Factor VII model; the measured ED50 decreased from ~0.1 mg/kg for the unoptimized screening formulation to ~0.02 mg/kg for the KC2-SNALP formulation (Fig. 3a). KC2-SNALP also exhibited similar potency in rats (data not shown). Furthermore, after a single administration in rats, KC2-SNALP–mediated gene silencing was found to persist for over 10 d (Supplementary Fig. 1). In addition to efficacy, tolerability is another critical attribute of a suitable LNP-siRNA delivery system for human use. We therefore studied the single-dose tolerability of KC2-SNALP in rats—a ­popular rodent model for assessing the toxicology of siRNA and nucleic acid– based therapeutics. As doses near the efficacious dose level were found to be very well tolerated (data not shown), single-dose escalation studies were conducted starting at doses ~50-fold higher (1 mg/kg) than the observed ED50 of the formulation. To understand formulation ­toxicity in the absence of any toxicity or pharmacologic effects resulting from target silencing, we conducted the experiments using a ­nontargeting control siRNA sequence directed against luciferase. KC2-SNALP containing luciferase siRNA was prepared in the exact same manner as that containing Factor VII siRNA, and the resultant size, lipid composition and entrapped siRNA/lipid ratio were similar. Clinical signs were observed daily and body weights, serum chemistry and hematology parameters were measured 72 h after dosing. KC2-SNALP was very well tolerated at the high dose levels examined (relative to the observed ED50 dose) with no dose-dependent, clinically significant changes in key serum chemistry or hematology parameters (Table 2).

nature biotechnology  VOLUME 28  NUMBER 2  FEBRUARY 2010

Given the promising activity and safety profile observed in rodents, studies were initiated in nonhuman primates to investigate the translation of DLin-KC2-DMA activity in higher species. For these studies, we chose to target transthyretin (TTR), a hepatic gene of high therapeutic interest20. TTR is a serum protein synthesized primarily in the liver, and although amyloidogenic TTR mutations are rare, they are endemic to certain populations and can affect the peripheral nerves, leading to familial amyloidotic polyneuropathy, and the heart, leading to familial amyloid cardiomyopathy. Currently, the only disease-modifying therapy is liver transplantation. We treated cynomolgus monkeys with a single 15-min intravenous infusion of KC2-SNALP–formulated siTTR at siRNA doses of 0.03, 0.1, 0.3 and 1 mg/kg. Control animals received a single 15-min intravenous infusion of PBS or KC2-SNALP–formulated ApoB siRNA at a dose of 1 mg/kg. Tissues were harvested 48 h after administration and liver mRNA levels of TTR were determined. A clear dose response was obtained with an apparent ED50 of ~0.3 mg/kg (Fig. 3b). A ­toxicological analysis indicated that the treatment was well tolerated at the dose levels tested, with no treatment-related changes in ­animal appearance or behavior. No dose-dependent, clinically significant alterations in key clinical chemistry or hematological parameters were observed (Supplementary Table 4). In summary, we applied a rational approach to the design of novel cationic lipids, which were screened for use in LNP-based siRNA delivery systems. Lipid structure was divided into three main functional elements: alkyl chain, linker and headgroup. With DLinDMA as a starting point, the effect of each of these elements was investigated in a systematic fashion, by holding the other two constant. First, the alkyl chains were established, then linker was varied and, finally, different headgroup structures were explored. Using this approach, important structure-activity considerations for ionizable cationic lipids were described and lipids with improved activity relative to the DLinDMA benchmark were identified. A SNALP formulation of the best-performing lipid (DLin-KC2-DMA) was well-tolerated in both rodent and nonhuman primates and exhibited in vivo activity at siRNA doses as low as 0.01 mg/kg in rodents, as well as silencing of a therapeutically significant gene (TTR) in nonhuman primates. Although the scope of the current work has been limited to hepatic delivery in vivo, the TTR silencing achieved in this work (ED50 ~ 0.3 mg/kg) represents a substantial improvement in activity relative to previous reports of LNP-siRNA mediated silencing in nonhuman primates. Methods Methods and any associated references are available in the online version of the paper at http://www.nature.com/naturebiotechnology/. Note: Supplementary information is available on the Nature Biotechnology website. Acknowledgments The authors thank K. McClintock for assistance with animal studies. The authors also thank the Centre for Drug Research and Development at the University

175

letters of British Columbia for use of the NMR facilities and M. Heller for his expert assistance in setting up the 31P-NMR experiments. AUTHOR CONTRIBUTIONS J.C., M.A.C., P.R.C., T.D.M., M.J.H. and K.F.W. designed and advised on novel lipids. J.C., K.F.W. and M.S. synthesized novel lipids. M.J.H., T.D.M., J.C., K.F.W., M.M., K.G.R., M.A.M., M.T. and M.J. analyzed and interpreted lipid data. T.D.M., M.J.H. and M.A.T. co-directed novel lipid synthesis and screening program. S.C.S. designed and directed rodent in vivo studies. S.C.S., S.K.K., B.L.M., K.L., M.L.E., M.K., A.P.S., Y.K.T., S.A.B., W.L.C., M.J.W. and E.J.C. generated rodent in vivo data, including Factor VII and tolerability analyses. L.N., V.K., T.B., R.A., Q.C. and D.W.Y.S. developed novel siRNAs targeting TTR. R.A. and A.A. designed and directed NHP in vivo studies. S.C.S., S.K.K., A.A., B.L.M., I.M., A.P.S., Y.K.T., R.A., T.B., D.W. Y. S., S.A.B., J.Q., J.R.D. and A.d.F. analyzed and interpreted in vivo data. B.L.M., K.L., A.P.S., S.K.K., S.C.S. and E.J.C. generated and characterized preformed vesicle formulations with novel lipids. D.S. and C.K.C. developed methods and designed and conducted HPLC lipid analyses of preformed vesicle formulations. E.Y. and L.B.J. prepared SNALP formulations. P.R.C. directed biophysical studies and advised on methods. A.P.S., I.M.H., S.D. and K.W. performed biophysical characterization studies (pKa, NMR, differential scanning calorimetric) of novel lipids and formulations. M.J.H., P.R.C., T.D.M., A.P.S., I.M.H. and K.F.W. analyzed biophysical data. S.C.S., M.J.H., A.A. and P.R.C. co-wrote the manuscript. T.D.M., M.M., M.A.M., M.A.T. and A.D.F. reviewed and edited the manuscript. S.C.S., M.J.H., A.A., P.R.C., I.M. and A.D.F. were responsible for approval of the final draft.

© 2010 Nature America, Inc. All rights reserved.

COMPETING INTERESTS STATEMENT The authors declare competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/naturebiotechnology/. Published online at http://www.nature.com/naturebiotechnology/. Reprints and permissions information is available online at http://npg.nature.com/ reprintsandpermissions/.

1. de Fougerolles, A.R. Delivery vehicles for small interfering RNA in vivo. Hum. Gene Ther. 19, 125–132 (2008). 2. Whitehead, K.A., Langer, R. & Anderson, D.G. Knocking down barriers: advances in siRNA delivery. Nat. Rev. Drug Discov. 8, 129–138 (2009). 3. Judge, A.D. et al. Confirming the RNAi-mediated mechanism of action of siRNA-based cancer therapeutics in mice. J. Clin. Invest. 119, 661–673 (2009).

176

4. Judge, A.D. et al. Sequence-dependent stimulation of the mammalian innate immune response by synthetic siRNA. Nat. Biotechnol. 23, 457–462 (2005). 5. Morrissey, D.V. et al. Potent and persistent in vivo anti-HBV activity of chemically modified siRNAs. Nat. Biotechnol. 23, 1002–1007 (2005). 6. Zimmermann, T.S. et al. RNAi-mediated gene silencing in non-human primates. Nature 441, 111–114 (2006). 7. Akinc, A. et al. A combinatorial library of lipid-like materials for delivery of RNAi therapeutics. Nat. Biotechnol. 26, 561–569 (2008). 8. Frank-Kamenetsky, M. et al. Therapeutic RNAi targeting PCSK9 acutely lowers plasma cholesterol in rodents and LDL cholesterol in nonhuman primates. Proc. Natl. Acad. Sci. USA 105, 11915–11920 (2008). 9. Hafez, I.M., Maurer, N. & Cullis, P.R. On the mechanism whereby cationic lipids promote intracellular delivery of polynucleic acids. Gene Ther. 8, 1188–1196 (2001). 10. Xu, Y. & Szoka, F.C. Jr. Mechanism of DNA release from cationic liposome/DNA complexes used in cell transfection. Biochemistry 35, 5616–5623 (1996). 11. Zelphati, O. & Szoka, F.C. Jr. Mechanism of oligonucleotide release from cationic liposomes. Proc. Natl. Acad. Sci. USA 93, 11493–11498 (1996). 12. Torchilin, V.P. Recent approaches to intracellular delivery of drugs and DNA and organelle targeting. Annu. Rev. Biomed. Eng. 8, 343–375 (2006). 13. Semple, S.C. et al. Efficient encapsulation of antisense oligonucleotides in lipid vesicles using ionizable aminolipids: formation of novel small multilamellar vesicle structures. Biochim. Biophys. Acta 1510, 152–166 (2001). 14. Maurer, N. et al. Spontaneous entrapment of polynucleotides upon electrostatic interaction with ethanol-destabilized cationic liposomes. Biophys. J. 80, 2310–2326 (2001). 15. Heyes, J., Palmer, L., Bremner, K. & Maclachlan, I. Cationic lipid saturation influences intracellular delivery of encapsulated nucleic acids. J. Control. Release 107, 276–287 (2005). 16. Semple, S.C., Chonn, A. & Cullis, P.R. Interactions of liposomes and lipid-based carrier systems with blood proteins: Relation to clearance behaviour in vivo. Adv. Drug Deliv. Rev. 32, 3–17 (1998). 17. Bailey, A.L. & Cullis, P.R. Modulation of membrane fusion by asymmetric transbilayer distributions of amino lipids. Biochemistry 33, 12573–12580 (1994). 18. Cullis, P.R. & de Kruijff, B. The polymorphic phase behaviour of phosphatidyl­ ethanolamines of natural and synthetic origin. A 31P NMR study. Biochim. Biophys. Acta 513, 31–42 (1978). 19. Epand, R.M., Robinson, K.S., Andrews, M.E. & Epand, R.F. Dependence of the bilayer to hexagonal phase transition on amphiphile chain length. Biochemistry 28, 9398–9402 (1989). 20. Sekijima, Y., Kelly, J.W. & Ikeda, S. Pathogenesis of and therapeutic strategies to ameliorate the transthyretin amyloidoses. Curr. Pharm. Des. 14, 3219–3230 (2008). 21. Cullis, P.R., Hope, M.J. & Tilcock, C.P. Lipid polymorphism and the roles of lipids in membranes. Chem. Phys. Lipids 40, 127–144 (1986).

VOLUME 28  NUMBER 2  FEBRUARY 2010  nature biotechnology

ONLINE METHODS Synthesis of cationic and PEG-lipids. A detailed description of the cationic lipid syntheses is available in the Supplementary Syntheses 1 and 2. The synthesis of N-[(methoxy poly(ethylene glycol)2000)carbamoyl]-1,2-dimyristyloxlpropyl-3-amine (PEG-C-DMA) was as described 22. The synthesis of R-3-[(ω-methoxy poly(ethylene glycol)2000)carbamoyl)]-1,2-dimyristyloxlpropyl-3-amine (PEG-C-DOMG) was as described7. These lipids were interchangeable in the formulation without substantially affecting activity (data not shown), and are collectively referred to as PEG-lipid.

© 2010 Nature America, Inc. All rights reserved.

siRNA synthesis. All siRNAs were synthesized by Alnylam and were charac­ terized by electrospray mass spectrometry and anion exchange highperformance liquid chromatography (HPLC). The sequences for the sense and antisense strands of Factor VII, ApoB and control siRNAs have been reported7. The sequences for the sense and antisense strands of the TTR siRNA is as follows: siTTR sense: 5′-GuAAccAAGAGuAuuccAudTdT-3′; antisense: 5′-AUGG AAuACUCUUGGUuACdTdT-3′. 2′-O-Me–modified nucleotides are in lowercase. siRNAs were generated by annealing equimolar amounts of complementary sense and antisense strands. Preformed vesicle method to formulate LNP-siRNA systems. LNP-siRNA systems were made using the preformed vesicle method14. Cationic lipid, DSPC, cholesterol and PEG-lipid were solubilized in ethanol at a molar ratio of 40:10:40:10, respectively. The lipid mixture was added to an aqueous buffer (50 mM citrate, pH 4) with mixing to a final ethanol and lipid concentration of 30% (vol/vol) and 6.1 mg/ml, respectively, and allowed to equilibrate at 22 °C for 2 min before extrusion. The hydrated lipids were extruded through two stacked 80 nm pore-sized filters (Nuclepore) at 22 °C using a Lipex Extruder (Northern Lipids) until a vesicle diameter of 70–90 nm, as determined by dynamic light scattering analysis, was obtained. This generally required 1–3 passes. The siRNA (solubilized in a 50 mM citrate, pH 4 aqueous solution containing 30% ethanol) was added to the pre-equilibrated (35 °C) vesicles at a rate of ~5 ml/min with mixing. After a final target siRNA/lipid ratio of 0.06 (wt/wt) was reached, the mixture was incubated for a further 30 min at 35 °C to allow vesicle reorganization and encapsulation of the siRNA. The ethanol was then removed and the external buffer replaced with PBS (155 mM NaCl, 3 mM Na2HPO4, 1 mM KH2PO4, pH 7.5) by either dialysis or tangential flow diafiltration. Preparation of KC2-SNALP. siRNA were encapsulated in SNALP using a controlled step-wise dilution method process as described23. The lipid ­constituents of KC2-SNALP were DLin-KC2-DMA (cationic lipid), dipalmitoylphosphatidylcholine (DPPC; Avanti Polar Lipids), synthetic cholesterol (Sigma) and PEG-C-DMA used at a molar ratio of 57.1:7.1:34.3:1.4. Upon formation of the loaded particles, SNALP were dialyzed against PBS and filter sterilized through a 0.2 µm filter before use. Mean particle sizes were 75–85 nm and 90–95% of the siRNA was encapsulated within the lipid particles. The final siRNA/lipid ratio in formulations used for in vivo testing was ~0.15 (wt/wt). In vivo screening of cationic lipids for Factor VII activity. Eight- to 10-week-old, female C57BL/6 mice were obtained from Harlan. Mice were held in a pathogen-free environment and all procedures involving animals were performed in accordance with local, state and federal regulations, as applicable, and approved by the Institutional Animal Care and Use Committee (IACUC). LNP-siRNA systems containing Factor VII siRNA were diluted to the appropriate ­concentrations in sterile PBS immediately before use and the formulations were ­administered intravenously through the lateral tail vein in a total volume of 10 ml/kg. After 24 h, animals were anesthetized with ketamine/xylazine and blood was collected by cardiac puncture and processed to serum (microtainer serum separator tubes; Becton Dickinson). Serum was tested immediately or stored at −70 °C for later analysis for Factor VII levels. Measurement of Factor VII protein in serum. Serum Factor VII levels were determined using the colorimetric Biophen VII assay kit (Aniara) 7. Briefly, serially diluted pooled control serum (200–3.125%) and appropriately

doi:10.1038/nbt.1602

diluted serum samples from treated animals (n = 4–5 animals per dose level) were analyzed in 96-well, flat bottom, nonbinding polystyrene assay plates (Corning) using the Biophen VII kit according to manufacturer’s instructions. Absorbance was measured at 405 nm and a calibration curve was generated using the serially diluted control serum to determine levels of Factor VII in serum from treated animals, relative to the saline-treated control animals. ED50 values for each formulation were derived from linear interpolation of the Factor VII activity profile, and included data points within 10–90% residual Factor VII activity (typically three to six points). Formulations containing novel lipids were always screened with one or more benchmark formulations to control and assess assay variability over time, and formulations with promising activity were repeated, with an expanded number of dose levels. In situ determination of pKa using TNS. The pKa of each cationic lipid was determined in LNPs using TNS17 and preformed vesicles composed of cationic lipid/DSPC/cholesterol/PEG-lipid (40:10:40:10 mol%) in PBS at a concentration of ~6 mM total lipid. TNS was prepared as a 100 µM stock solution in distilled water. Vesicles were diluted to 100 µM lipid in 2 ml of buffered solutions containing 1 µM TNS, 10 mM HEPES, 10 mM 4-morpholineethanesulfonic acid , 10 mM ammonium acetate, 130 mM NaCl, where the pH ranged from 2.5 to 11. Fluorescence intensity was monitored in a stirred, thermostated cuvette (25 °C) in a PerkinElmer LS-50 Spectrophotometer using excitation and emission wavelengths of 321 nm and 445 nm. Fluorescence measurements were made 30 s after the addition of the lipid to the cuvette. A sigmoidal best fit analysis was applied to the fluorescence data and the pKa was measured as the pH giving rise to half-maximal fluorescence intensity. Differential scanning calorimetry. Analyses were performed using the same samples used for 31P NMR, on a TA Instruments Q2000 calorimeter using a heat/cool/heat cycle and a scan rate of 1 °C/minute between 10 °C and 70 °C. Repeat scans were reproducible to within 0.1 °C. The temperature at the peak amplitude of the endo- and exotherms was measured for both the heating and cooling scans, and the TBH values observed correlated closely with the phase transition temperatures measured using 31P NMR. Determination of siRNA plasma levels. Plasma levels of fluorescently labeled Cy3 siRNA were evaluated at 0.5 and 3 h after intravenous injection of selected LNP (preformed vesicle) formulations, administered at an siRNA dose of 5 mg/kg, in C57BL/6 mice. Blood was collected in EDTA-containing Vacutainer tubes, processed to plasma at 2–8 °C, and either assayed immediately or stored at –30 °C. An aliquot of the plasma (100 µl maximum) was diluted to 500 µl with PBS (145 mM NaCl, 10 mM phosphate, pH 7.5); methanol (1.05 ml) and chloroform (0.5 ml) were added; and the sample was vortexed to obtain a clear, single-phase solution. Additional water (0.5 ml) and chloroform (0.5 ml) was added and the resulting emulsion was sustained by periodic mixing. The mixture was centrifuged at 500g for 20 min and the aqueous phase containing the Cy3-labeled siRNA was collected and the fluorescence measured using an SLM Fluorimeter at an excitation wavelength of 550 nm (2 nm bandwidth) and emission wavelength of 600 nm (16 nm bandwidth). A standard curve was generated by spiking aliquots of plasma from untreated animals with the formulation containing Cy-3-siRNA (0 to 15 µg/ml), and the resulting standards were processed as indicated above. Determination of siRNA biodistribution. Tissue (liver and spleen) levels of siRNA were evaluated at 0.5 and 3 h after intravenous injection in C57BL/6 mice after administration of LNP (preformed vesicle) formulations containing selected novel lipids. After blood collection, animals were perfused with saline to remove residual blood from the tissues; liver and spleen were then collected, weighed and divided into two pieces. Portions (400–500 mg) of liver or whole spleens were weighed into Fastprep tubes and homogenized in 1 ml of Trizol using a Fastprep FP120 instrument. An aliquot of the homo­ genate (typically equivalent to 50 mg of tissue) was transferred to an Eppendorf tube and additional Trizol was added to achieve a final volume of 1 ml. Chloroform (0.2 ml) was added and the solution was mixed and incubated for 2–3 min, before being centrifuged for 15 min at 12,000g. An aliquot (0.5 ml) of the aqueous phase was diluted with 0.5 ml of PBS and the

nature biotechnology

sample fluorescence was measured as described above. The data were expressed as the percent of the injected dose (in each tissue).

Statistical analysis. P-values were calculated for comparison of K2C-SNALP– treated animals with PBS-treated animals using analysis of variance (ANOVA, single-factor) with an alpha value of 0.05. P < 0.05 was considered significant. 22. Heyes, J., Hall, K., Tailor, V., Lenz, R. & MacLachlan, I. Synthesis and characterization of novel poly(ethylene glycol)-lipid conjugates suitable for use in drug delivery. J. Control. Release 112, 280–290 (2006). 23. Jeffs, L.B. et al. A scalable, extrusion-free method for efficient liposomal encapsulation of plasmid DNA. Pharm. Res. 22, 362–372 (2005).

© 2010 Nature America, Inc. All rights reserved.

In vivo nonhuman primate experiments. All procedures using cynomolgus monkeys were conducted by a certified contract research organization using protocols consistent with local, state and federal regulations, as applicable, and approved by the IACUC. Cynomolgus monkeys (n = 3 per group) received either 0.03, 0.1, 0.3 or 1 mg/kg siTTR, or 1 mg/kg siApoB (used as control) formulated in KC2-SNALP as 15-min intravenous infusions (5 ml/kg) through the cephalic vein. Animals were euthanized 48 h after administration, and a 0.15–0.20 g sample of the left lateral lobe of the liver was collected and snap-frozen in liquid nitrogen. Prior studies have established uniformity of silencing activity throughout the liver6. TTR mRNA levels, relative to GAPDH

mRNA levels, were determined in liver samples using a branched DNA assay (QuantiGene Assay)6. Clinical chemistry and hematology parameters were analyzed before and 48 h after administration.

nature biotechnology

doi:10.1038/nbt.1602

c o r r i g e n d a & e r r ata

Corrigendum: Microdroplet-based PCR enrichment for large-scale targeted sequencing Ryan Tewhey, Jason B Warner, Masakazu Nakano, Brian Libby, Martina Medkova, Patricia H David, Steve K Kotsopoulos, Michael L Samuels, J Brian Hutchison, Jonathan W Larson, Eric J Topol, Michael P Weiner, Olivier Harismendy, Jeff Olson, Darren R Link & Kelly A Frazer Nat. Biotechnol. 27, 1025–1031 (2009); published online 1 November 2009; corrected after print 11 November 2009 In the version of this article initially published, the email address for K.A.F. should have been [email protected]. The error has been corrected in the HTML and PDF versions of the article.

Corrigendum: The valuation high ground Jeffrey J Stewart & Ben Bonifant Nat. Biotechnol. 27, 980–983 (2009); published online 24 September 2009; corrected online 6 November 2009; pdf corrected 5 February 2010

© 2010 Nature America, Inc. All rights reserved.

In the version of this article initially published, the email address for Ben Bonifant was incorrect. The email address is bbonifant@campbellalliance. com. The error has been corrected in the HTML and PDF versions of the article.

Corrigendum: Receptor-binding specificity of pandemic influenza A (H1N1) 2009 virus determined by carbohydrate microarray Robert A Childs, Angelina S Palma, Steve Wharton, Tatyana Matrosovich, Yan Liu, Wengang Chai, Maria A Campanero-Rhodes, Yibing Zhang, Markus Eickmann, Makoto Kiso, Alan Hay, Mikhail Matrosovich & Ten Feizi Nat. Biotechnol. 27, 797–799 (2009); published online 9 September 2009; corrected after print 5 February 2010 In the version of this article initially published, two acknowledgments were inadvertently omitted: NCI Alliance of Glycobiologists for Detection of Cancer and Cancer Risk; and Biotechnology and Biological Sciences Research Council. The error has been corrected in the HTML and PDF versions of the article.

Corrigendum: Small but not simple Markus Elsner Nat. Biotechnol. 28, 42 (2010); published online 8 January 2010; corrected after print 5 February 2010 In the version of this article initially published, the organisms in question were incorrectly identified as Mycobacterium pneumoniae and Mycobacterium genitalium. The correct names are Mycoplasma pneumoniae and Mycoplasma genitalium, respectively. The error has been corrected in the HTML and PDF versions of the article.

Erratum: A nuclear magnetic resonance technique for determining hybridoma cell concentration in hollow fiber bioreactors Anthony Mancuso, Erik J. Fernandez, Harvey W. Blanch & Douglas S. Clark Biotechnology 8, 1282–1285 (1990); corrected online 5 February 2010 In the version of this article initially published online, a graph published in print as Figure 2 was erroneously duplicated and appeared as both Figure 1 and Figure 2. The original Figure 1 has been restored in the online PDF version of the article.

Erratum: Can web 2.0 reboot clinical trials? Malorye Allison Nat. Biotechnol. 27, 895–902 (2009); published online 8 October 2009; corrected after print 5 February 2010 In the version of this article initially published, Sharib Khan was incorrectly identified as the CEO of TrialX. He is cofounder. The error has been corrected in the HTML and PDF versions of the article.

178

volume 28 number 2 february 2010 nature biotechnology

careers and recruitment Fourth quarter lag in biotech hiring Michael Francisco

© 2010 Nature America, Inc. All rights reserved.

T

ougher times returned for the fourth quarter of 2009, as the slight hiring increase seen in the previous quarter at the 25 largest biotechs and 10 largest pharma companies (Nat. Biotechnol. 27, 1056, 2009) did not hold, according to three representative job databases (Tables 1 and 2). Big biotech and pharma companies across the board posted fewer job listings with the exception of Bayer (Leverkusen, Germany), whose US job listings jumped from 4 to 78 on Monster.com. The landscape for existing jobs is mixed as well, with Seattlebased biotech VLST reportedly reducing its workforce by 30% as part of a restructuring. Contract research organization Charles River Laboratories (Wilmington, MA, USA) is cutting 300 jobs and plans to suspend operations at a preclinical services facility by the middle of 2010, after the completion of ongoing projects. On the pharma side, Pfizer has informed New Jersey officials of its plans to lay off 400 employees at Wyeth’s old research center, as part of its acquisition. In addition, Merck will eliminate 500 jobs—mostly sales and administrative positions—in conjunction with its takeover of Schering-Plough. The company, however, plans to safeguard 500 jobs as part of its acquisition of Avecia (Billingham, UK). Merck will acquire all of Avecia’s assets, including all process development and scale-up, manufacturing, quality and business support operations. And Lonza Group of Basel, Switzerland, a large supplier of active pharmaceuticals ingredients, is closing three of its manufacturing sites in 2010. One hundred and seventy-five jobs are expected Table 1 Who’s hiring? Advertised openings at the 25 largest biotech companies Companya

Number of employees

Monsanto

Number of advertised openingsb Monster

Biospace

Naturejobs

21,700

0

0

112

Amgen

16,800

35

45

2

Genentech

11,186

11

22

73

Genzyme

11,000

54

0

105

9,700

33

23

Life Technologies

to be lost. The sites are located in Conshohocken, Pennsylvania, Shawinigan, Canada and Wokingham, UK. “The economic pressures of the past 18 months have clearly accelerated the cost reduction efforts of the pharmaceutical industry,” according Lonza’s press release. Adds Lonza CEO Stefan Borgas, “The closure of the three sites will help to optimize our global operational network and further increase the competitiveness for our customers.” The company says it will increase its platform in Asia. In a sign of confidence of biotech’s role in growing the economy, the governor of the state of Missouri embraced a proposal to “direct tens of millions of tax dollars to Missouri’s biotechnology industry,” according to the Associated Press. The plan would divert an annual portion of the new tax revenues generated by biotech companies to a special state fund, from which incentives could be given to new or expanding entrepreneurs in the same field, perhaps in the form of aid to startups, providing infrastructure to lure existing out-of-state firms and subsidizing college-based training for the employees of biotech companies. The move is expected to result in a sizable number of biotech jobs. Missouri already is home to some top university and private-sector researchers in the life sciences. But economic development officials say Missouri is lagging when it comes to converting that research into commercial ventures. Nature Biotechnology will continue follow hiring and firing trends throughout 2010. Table 2 Advertised job openings at the ten largest pharma companies Companya

Number of employees

Number of advertised openingsb Monster

Biospace

Naturejobs

Johnson & Johnson

119,200

203

1

0

Bayer

106,200

78

30

1

GlaxoSmithKline

103,483

1

3

4

Sanofi-Aventis

99,495

0

0

0

1

Novartis

98,200

3

30

60

PerkinElmer

7,900

16

0

0

Pfizer

86,600

39

33

8

Bio-Rad Laboratories

6,600

5

10

0

Roche

78,604

8

17

4

Biomerieux

6,140

2

0

0

Abbott Laboratories

68,697

30

0

0

Millipore

5,900

4

22

0

AstraZeneca

67,400

18

7

5

IDEXX Laboratories

4,700

11

1

0

Merck & Co.

59,800

0

0

0

Biogen Idec

4,700

35

28

0

aData

Gilead Sciences

3,441

2

12

0

WuXi PharmaTech

3,172

0

0

0

Qiagen

3,041

0

0

0

Cephalon

2,780

0

0

0

Biocon

2,772

0

0

0

Celgene

2,441

11

2

0

Biotest

2,108

7

5

0

Actelion

2,054

1

2

0

Amylin Pharmaceuticals

1,800

21

8

0

Elan

1,687

5

1

0

Illumina

1,536

26

0

10

Albany Molecular Research

1,357

0

0

0

Vertex Pharmaceuticals

1,322

19

33

4

CK Life Sciences

1,315

0

0

0

defined in Nature Biotechnology’s survey of public companies (27, 710–721, 2009). bAs searched on Monster.com, Biospace.com and Naturejobs.com, January 10, 2010. Jobs may overlap.

obtained from MedAdNews. bAs searched on Monster.com, Biospace.com and Naturejobs.com, January 10, 2009. Jobs may overlap.

aAs

nature biotechnology volume 28 number 2 february 2010

Michael Francisco is Senior Editor, Nature Biotechnology

179

people

© 2010 Nature America, Inc. All rights reserved.

Makefield Therapeutics (Newtown, PA, USA) has announced the appointment of Jim Ballance (left) as CSO. Most recently a corporate strategic consultant to the biotech industry, Ballance brings more than 20 years of experience in the development and manufacturing of novel therapeutics. His experience includes such positions as vice president, technology development at BioRexis Pharmaceutical, director of biotech evaluation at Aventis Behring, chief technology officer at Genesis Therapeutics and head of R&D at Delta Biotechnology.

Inovio Biomedical (San Diego) has named Mark L. Bagarazzi as chief medical officer. Bagarazzi joins Inovio from Merck & Co., where he was director of worldwide regulatory affairs for vaccines and biologics. Before joining Merck in 2001, he was director of the HIV/ AIDS program for St. Christopher’s Hospital for Children in Philadelphia. IRX Therapeutics (New York) has named Neil L. Berinstein CSO, succeeding company founder John W. Hadden, who remains a member of the board of directors. Berinstein previously served as assistant vice president and global program leader at Sanofi Pasteur, where he was in charge of leading the development of the company’s cancer vaccines both strategically and operationally. Genzyme (Cambridge, MA, USA) has named Ron Branning as its new senior vice president of global product quality. Branning is tasked with ensuring the quality of all Genzyme products manufactured at 17 sites around the world. He brings 30 years of experience in product quality and regulatory affairs at biotech and pharma companies including Johnson & Johnson, Gilead Sciences and Genentech as well as companies acquired by Baxter, Wyeth and Pfizer. Ian Brown has been named CEO of BioCeramic Therapeutics (London) as part of a succession plan developed by the board of directors to succeed the company’s outgoing founding CEO, Daniel Green. Brown has held senior executive positions with Chromogenix (previously Kabi Pharmacia), Instrumentation Laboratory, Cordlife, SDP Technology and Avanti Capital. Green will remain as a company director. CombinatoRx (Cambridge, MA, USA) has announced changes to its senior management

180

team following its recently completed merger with Neuromed. Mark Corrigan, former executive vice president of R&D at Sepracor and a member of the CombinatoRx board of directors, will assume the role of president and CEO of CombinatoRx and Christopher Gallen, former CEO of Neuromed, will serve as executive vice president of R&D. CombinatoRx’s interim president and CEO Robert Forrester has decided to leave his position at the company to pursue other opportunities. Roger Hickling has been appointed R&D director and a board director at Phytopharm (Godmanchester, UK). He previously worked at SmithKline Beecham where he oversaw both in-house and partnered early stage neuroscience development projects, and was most recently R&D director and a board member at Alizyme Therapeutics. Prana Biotechnology (Melbourne, Australia) has named Paul Marks as a director of the company. Marks was previously vice president of foreign exchange with PrudentialBache Securities and senior FX strategist with National Australia Bank. He also serves as director of Conquest Mining and several private companies. BIND Biosciences (Cambridge, MA, USA) has announced the appointment of Scott Minick as president and CEO. Minick was formerly a managing director at ARCH Venture Partners and previously president and COO of Sequus Pharmaceuticals/Liposome Technology. Biogen Idec (Cambridge, MA, USA) has announced that James Mullen will retire as president and CEO as of June 8. Mullen will also retire from Biogen’s board at the end of his current term as a director at the company’s 2010

annual shareholder meeting. The company has started a search for Mullen’s successor. CrystalGenomics (Seoul, S. Korea) has chosen Eric M. Nelson to serve as vice president of business development for its US subsidiary, Emeryville, California-based CG Pharmaceuticals. Nelson brings 22 years of experience in business development and licensing for pharma and biotech companies. Most recently, he was global head of business development at Advinus Therapeutics. Proteonomix (Mountainside, NJ, USA) has announced the following management changes: Joel Pensley has resigned as secretary, director and general counsel and Roger Fidler has joined the company as director and general counsel; and Steven Byle, presently a director, has assumed the role of secretary. Pensley resigned as an officer and director due to health concerns. Fidler has been the sole director, president, CEO and CFO of Global Agri-Med Technologies. Eric Ruby has joined Presidio Pharmaceuticals (San Francisco) as vice president of regulatory affairs. He served previously as senior director, regulatory affairs at Alnylam Pharmaceuticals, where he was responsible for regulatory filings to support clinical trials in the US and Europe. Protagen (Dortmund, Germany) has appointed two new members to its executive board: Peter Schulz-Knappe, former CSO at Proteome Sciences, will head the company’s diagnostics business unit, and co-founder and COO Martin Blüggel will manage the protein services unit. Pieris (FreisingWe i h e n s t e p h a n , Germany) has announced the appointment of Stephen S. Yoder (left) as CEO. He succeeds interim CEO and co-founder Claus Schalper, who will remain CFO. Yoder joins Pieris from MorphoSys, where he served as general counsel and head of licensing & IP.

volume 28 number 2 february 2010 nature biotechnology