volume 28 number 3 march 2010
e d i tor i a l S
© 2010 Nature America, Inc. All rights reserved.
181 182
Colored scanning electron micrograph of cardiac muscle fibrils (green) and mitochondria (orange). Mootha and colleagues identify drugs that reconfigure energy metabolism and demonstrate the protective capacity of an approved drug in models of cardiac and cerebral ischemia (p 249). Credit: Steve Gschmeissner/Photo Researchers, Inc.
America’s got talent—can it keep it? H1N1dsight is a wonderful thing
ne w s 183 184 185 187 188 189 189 189 191 191 192 192 192 193 197
Ark’s gene therapy stumbles at the finish line Monsanto’s alfalfa reaches Supreme Court GSK/Sirtris compounds dogged by assay artifacts US biodefense contracts continue to lure biotechs Ride ‘n Drive on government waste Chinese institute makes bold sequencing play Melanoma vaccine for dogs Biotechs go virtual RNAi delivery shop Brazil boosts bioscience Patent income tax slashed Abbott hit with record fine Resuscitated deCODE refocuses on diagnostics News feature: One year in—Obama’s biotech scorecard News feature: The lengthening handshake
B i oentrepreneur B u i l d i ng a bus i ness 200
Seeking the biotech eBay Nuala Moran
op i n i on an d comment Ark’s glioma gene therapy runs into trouble, p 183
CORRESPONDENCE 203 Oversulfated chondroitin sulfate is not the sole contaminant in heparin 212 Why FDA recruitment of ‘critics’ is a problem 212 Genetic exceptionalism
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 3 march 2010 v3
feature 214 Lost in migration George S Mack and Andrew Marshall patents v1
Optimal solution
230
Patenting biotech beyond the central dogma George Wu
234
Recent patent applications in DNA diagnostics
v2
© 2010 Nature America, Inc. All rights reserved.
Modelling metabolite flow, p 245
NEWS AND VIEWS 235
Genetic therapy for spinal muscular atrophy see also p 271 Alex MacKenzie
237
Targeting leukemia stem cells Hanna K A Mikkola, Caius G Radu & Owen N Witte
239
Cellular targets for influenza drugs Ji-Young Min & Kanta Subbarao
241
Navigating genomic maps of cancer cells Marcel P van der Brug & Claes Wahlestedt
242
Grass genomics on the wild side Craig Mak
243
Research highlights
see also p 275
computat i ona l b i o l og y pr i mer 245
What is flux balance analysis? Jeffrey D Orth, Ines Thiele & Bernhard Ø Palsson
MRI dopamine sensor, p 264
research ARTICLES 249
Nutrient-sensitized screening for drugs that shift energy metabolism from mitochondrial respiration to glycolysis V M Gohil, S A Sheth, R Nilsson, A P Wojtovich, J H Lee, F Perocchi, W Chen, C B Clish, C Ayata, P S Brookes & V K Mootha
256
Harnessing chaperone-mediated autophagy for the selective degradation of mutant huntingtin protein P O Bauer, A Goswami, H K Wong, M Okuno, M Kurosawa, M Yamada, H Miyazaki, G Matsumoto, Y Kino, Y Nagai & N Nukina
264 Directed evolution of a magnetic resonance imaging contrast agent for noninvasive imaging of dopamine M G Shapiro, G G Westmeyer, P A Romero, J O Szablowski, B Küster, A Shah, C R Otey, R Langer, F H Arnold & A Jasanoff
Gene therapy for neuromuscular disease, p 271
nature biotechnology
iii
volume 28 number 3 march 2010 l etters 271
Rescue of the spinal muscular atrophy phenotype in a mouse model by early postnatal delivery of SMN K D Foust, X Wang, V L McGovern, L Braun, A K Bevan, A M Haidet, T T Le, see also p 235 P R Morales, M M Rich, A H M Burghes & B K Kaspar
275 Induction of cell cycle entry eliminates human leukemia stem cells in a mouse model of AML Y Saito, N Uchida, S Tanaka, N Suzuki, M Tomizawa-Murasawa, A Sone, Y Najima, see also p 237 S Takagi, Y Aoki, A Wake, S Taniguchi, L D Shultz & F Ishikawa Overcoming chemotherapy resistance, p 275
281 Isotopic labeling of terminal amines in complex samples identifies protein N-termini and protease cleavage products O Kleifeld, A Doucet, U auf dem Keller, A Prudova, O Schilling, R K Kainthan, A E Starr, L J Foster, J N Kizhakkedathu & C M Overall
© 2010 Nature America, Inc. All rights reserved.
careers an d recru i tment 289
The importance of foreign-born talent for US innovation YeonJi No & John P Walsh
292
people
Polymers to identify N-terminal peptides, p 281
nature biotechnology
v
i n t h i s i ss u e
Following up their earlier report on a viral vector that crosses the blood-brain barrier, Kaspar and coworkers present therapeutic results in a mouse model of spinal muscular atrophy (SMA). SMA, caused by homozygous mutation of the survival motor neuron gene, is a neuromuscular degenerative disease that is often fatal in childhood. A year ago the Kaspar group showed that intravenous injection of self-complementary (sc) AAV9 in neonatal mice enables widespread transduction of neurons (Foust, K.D. et al. Nat Biotechnol. 27, 42–44, 2009). Now, working with SMA researcher Arthur Burghes, they use scAAV9 to deliver the survival motor neuron gene to newborn SMA mice, achieving rescue of motor function and a dramatic extension of survival. SMA mice replicate much of the pathology seen in human patients and without treatment survive for only ~15 days after birth. A previous experimental therapy had increased life span to ~45 days, but mice receiving the scAAV9 gene therapy are still alive at 250 days. To begin to investigate whether the approach might work in human infants, the authors inject scAAV9-GFP into a newborn cynomolgus monkey and find efficient transduction of spinal motor neurons (pictured), demonstrating potential for clinical translation. [Letters, p. 271; News and Views, p. 235] KA
Adaptor peptides against mutant huntingtin Huntington’s disease (HD) is a dominantly inherited neurodegenerative disorder caused by expression of a mutant allele encoding huntingtin (HTT) protein containing >35 glutamine residues in its N-terminal polyglutamine (polyQ) tract. In part because of the challenge in specifically destroying mutant HTT or its mRNA without compromising expression of the normal HTT allele, there are no approved therapies for HD. However, several studies have shown the potential of strategies involving RNA interference or augmentation of the roles that macroautophagy and proteasomal degradation normally play in turning over mutant HTT. Using a different approach, Nukina and colleagues harness chaperone-mediated autophagy—a process distinct from macroautophagy insofar as it requires heat shock cognate protein-70 (HSC70) proteins (rather than autophagosomes) to target proteins for lysosomal destruction. They Written by Kathy Aschheim, Markus Elsner, Michael Francisco, Peter Hare, Brady Huggett & Lisa Melton
nature biotechnology volume 28 number 3 march 2010
show that intrastriatal injection of an adeno-associated virus 1 (AAV1) encoding two copies of the polyQ-binding peptide-1 (previously shown to bind an expanded (but not a normal) polyQ tract in HTT) ameliorates the disease phenotype in a HD mouse model. What’s more, by fusing two HSC70-binding motifs to the polyQ-binding peptides in the AAV1 construct, the authors were able to enhance these effects—prolonging survival, improving motor performance and achieving gains in mouse body weight. The extension of life span exceeds that reported by other single-drug therapies tested using the R2/6 mouse model of HD. The principle of expressing adaptor peptides comprising HSC70 binding motifs and a moiety specific for a misfolded protein may find broader application in a range of diseases where selective degradation of mutant gene products is desirable. [Articles, p. 256] PH
Sensitizing leukemia stem cells Initial chemotherapy to treat leukemia frequently succeeds in eliminating the vast majority of cancer cells. Nevertheless, the seemingly successful therapy is often followed by a relapse of the disease and subsequent death. Saito et al. describe a strategy to sensitize the small population of drug-resistant leukemia stem cells thought to be responsible for disease relapse after standard treatment. Using a mouse model derived from human leukemia samples, the authors show that pretreatment with granulocyte colony-stimulating factor (G-CSF) induces the normally quiescent leukemia stem cells to enter the cell cycle. The increased proliferation is correlated with an increase in apoptosis and cell death upon treatment with the chemotherapy drug cytarabine. The G-CSF/cytarabine combination therapy was associated with an ~100-fold reduction in the number of leukemia stem cells in the mouse model. What’s more, mice injected with leukemia cells purified from animals treated with the drug combination survived longer than those injected with cells from animals treated with chemotherapy alone. [Articles, p. 275; News and Views, p. 237] ME
Cellular energetics in the cross-hairs Many diseases, ranging from cancer to neurodegenerative disorders, are characterized by shifts in cellular energy metabolism. Accordingly, drugs that can be safely titrated to activate or inhibit mitochondrial respiration relative to glycolysis might be useful for treating cancer and ischemic diseases, respectively. Mootha and colleagues screen for such compounds by taking advantage of the fact that cells fed galactose rely almost exclusively on glutamine-driven oxidative phosphorylation. They
vii
Lauren Solomon/Shreya Sheth
© 2010 Nature America, Inc. All rights reserved.
Muscle medicine
i n t h i s i ss u e use a quantitative, nutrient-sensitized growth assay to test the abilities of >3,500 small molecules to selectively inhibit growth and proliferation in galactose- relative to glucose-containing media. As the screen suggested that meclizine (Antivert), an over-the-counter drug approved for treating nausea and vertigo but never linked to energy metabolism, redirects energetic flux away from mitochondria, the authors evaluate its role in preventing cellular damage caused by ischemia. Pretreatment with meclizine protects isolated, perfused hearts from injury in an ex vivo model of myocardial infarction and prevents cerebral damage in a mouse model of transient focal stroke. Although the exact mechanism(s) responsible for these protective effects remain to be established, the authors provide evidence supporting a novel target of meclizine. [Articles, p. 249] PH
© 2010 Nature America, Inc. All rights reserved.
Dopamine detection by MRI The use of modern imaging methods such as magnetic resonance imaging (MRI) or positron emission tomography has increased our understanding of the physiological and signaling processes in the brain. A more detailed analysis of the workings of the brain will require methods with high temporal and spatial resolution that make molecular events such as the secretion of neurotransmitters accessible. Using directed evolution techniques, Shapiro et al. develop an MRI contrast agent sensitive to the neurotransmitter dopamine. The contrast agent is based on the bacterial cytochrome P450 BM3, which contains a paramagnetic iron atom that changes its MRI signal upon binding
Patent roundup From April 2013, the UK government plans to slash taxes on revenue generated from research-related patents in a strategy designed to attract potential investment. [News, p. 192] LM A Texas jury has ordered Abbott Laboratories (Deerfield, IL) to pay a record $1.8 billion to Centocor (Horsham, PA) over a patent infringement case concerning the monoclonal antibody drug Humira (adalimumab) treatment for rheumatoid arthritis. [News, p. 192] LM Like most other things, intellectual property (IP) can now be shopped for online. What kind of challenges do internet IP exchanges face? [Building a Business, p. 200]
viii
On the TAILs of orphan proteases Simple and reliable approaches to characterizing the N termini of proteins in complex samples could provide insight into the regulation and functions of gene products; they could also identify the cleavage products generated by proteases with poorly defined specificities. Substrates have yet to be defined for the majority of human proteases. Addressing this need will be key to understanding the physiological roles of proteases and fully exploiting their value in drug development. Nonetheless, current approaches to identifying protease cleavage sites in native proteins (as opposed to synthetic peptides) cannot survey all N-terminal fragments while simultaneously identifying N-terminal fragments generated by a protease of interest. Overall and colleagues introduce terminal amine isotopic labeling of substrates (TAILS) to address this challenge. The approach involves labeling control and proteasetreated samples with different isotopes and then using amino-reactive dendritic polyglycerol aldehyde polymers to enrich for all N-terminal fragments. When coupled with fairly exacting bioinformatics criteria, a single mass spectrometric analysis enables high-confidence identification of substrates for proteases without known consensus cleavage sites. TAILS also characterizes post-translational protein modifications, such as acetylation and cyclization. The authors use the approach to identify known, as well as new, substrates for two metalloproteases implicated in metastasis, as well as dipeptidyl peptidase IV, a target relevant to treating diabetes. [Letters, p. 281] PH
Next month in BH
Several recent US federal court decisions have set new interpretations for rules governing novelty, nonobviousness and patentable subject areas. Wu alerts biotech inventors and patent practitioners alike to what is considered patentable when drafting claims. [Patent Article, p. 230] MF Recent patent applications on RNA diagnostics. [New patents, p. 234]
of its natural ligand, arachidonic acid. The authors iteratively evolve the protein by selecting mutants with a decreased affinity for arachidonic acid and an increased affinity for dopamine. After five rounds of mutagenesis and selection, they isolate two proteins with a high selectivity and sensitivity for dopamine. The sensor can detect the release of dopamine triggered by high potassium ion concentrations both in cell culture and in the brains of living mice. [Articles, p. 264] ME
BH
• Qualification of biomarkers for kidney toxicity • Real-time neuroimaging in 3D • Derivation of human ES cells in suspension • Synthetic urate homeostasis network • Plant defense with pattern-recognition receptors
volume 28 number 3 march 2010 nature biotechnology
www.nature.com/naturebiotechnology
© 2010 Nature America, Inc. All rights reserved.
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 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
Editorial
America’s got talent—can it keep it? To remain competitive in biotech, policymakers should pay more attention to retaining skilled foreign workers than to fixating on illegal immigration.
© 2010 Nature America, Inc. All rights reserved.
A
survey published on p. 289 provides data that one of America’s greatest strengths is the contribution of foreign nationals to the intellectual property underpinning entrepreneurial efforts in the life sciences. The study is the first to examine systematically the involvement of immigrants in patented biotech inventions. It finds that nearly 31% of lead inventors in biotech are foreign-born. This is a disproportionate contribution considering that only ~11% of the general US population and ~22% of the college-educated US workforce are foreign-born. The study analyzes a sample of >1,900 US-based inventors with patents not only granted in the United States, but also filed in Japan and Europe from 2000 to 2003. This result aligns with a study published last March (Issues Sci. Technol. 45–52, 2009), showing that foreign nationals residing in the United States were inventors or co-inventors in one-quarter of all World Intellectual Property Organization patent applications filed in 2006 emanating from the United States. The apparent overrepresentation of immigrant talent is also reflected in entrepreneurial data; over a quarter of US companies in all industrial sectors have foreign nationals as founders. For biotech startups, the number is 20%—somewhat lower, but still disproportionately high compared with the proportion of immigrants in the US population as a whole. This kind of data has been fueling the recent debate on the consequences of US immigration policy for innovation. With rising unemployment and a stuttering economy, some US politicians have called for a more restrictive stance on immigration: not quite ‘American jobs for Americans’, but perhaps headed in that direction. Critics of that policy direction point out that measures such as the Grassley-Sanders amendment in the 2009 American Recovery and Reinvestment Act restricts any firm receiving stimulus money from hiring immigrants on H-1B visas for one year. They also claim that the visa application process for foreign nationals coming to work in the United States is particularly dysfunctional, arbitrary and overly complex, forcing some applicants to wait up to a year for visa approval. Such measures, critics say, threaten US competitiveness by cutting off the very lifeblood of innovation, especially when other countries are becoming more attractive to mobile skilled workers. The UK, for instance, has introduced a points system that allows entry to some workers, even before they get a job. And, starting in May 2009, the European Union’s blue card program not only admits skilled workers but also allows them to bring their families and enables their spouses to work. India and China, of course, now increasingly offer much better commercial and research opportunities to would-be returnees, who might have remained in the United States in earlier decades. So, has US immigration policy in recent years impaired the ability of the country’s biotech sector to recruit and retain skilled immigrants? The answer, surprisingly perhaps, is no. The reality is that it will take much more than recent administrative adjustments to immigration procedures to dissuade skilled researchers
nature biotechnology volume 28 number 3 march 2010
from coming to the United States for professional or educational development. There is actually very little evidence that US immigration policy has harmed the country’s innovation capacity at all. Indeed, the specter of the ‘reverse brain drain’ appears to be based largely on myths. The first of these myths is that technologically skilled immigrants contribute disproportionately more to reinventing the US economy than the US-born. Yes, it is striking that 20% of CEOs of US biotech firms are foreign-born (25% in non-biotech startups) and that 31% of important patents filed from the United States have a foreign-born first inventor. Nevertheless, the immigrant supergeek is largely a phantom created by misperception. As the study we publish shows, foreign-born researchers are no more likely to invent than their US-born colleagues. The percentage of inventions originating from foreign-born researchers is higher because the percentage of US immigrants educated to the PhD level is much higher than the percentage of PhDs in the US population as a whole. The National Science Foundation (NSF) publishes a regular survey of PhD recipients in the United States. Its most recent study, published last year (http://www.nsf.gov/statistics/srvydoctorates/), shows that in 2008, 33% of all doctoral recipients in the United States were foreigners, a figure that rose to 48% in the physical sciences and 60% in engineering. In the life sciences, the proportion of foreign doctoral recipients was 29%. And the number of foreign-born PhDs is probably higher than even these figures because in the NSF survey, green-card holders were categorized as US citizens. When at least 29% of the US-trained life science PhDs are foreignborn, it is somehow less surprising that 31% of biotech inventors are foreign-born. The second myth underlying fears about the ‘reverse brain drain’ is that recent US policy has discouraged the influx and retention of technologically skilled immigrants. Here again, the NSF data on doctoral recipients shows that over the George W. Bush years, the number of foreigners receiving a US life sciences PhD increased from 2,158 (25% of the total) in 1998 to 3,246 (29%) in 2008—and the proportion of foreigners receiving US PhDs in all disciplines increased from 33% to 42%. What this means is that skilled immigrants in the biotech sector do make a significant contribution to US economic prosperity, but not a disproportionate one. Until now, America has been able to supplement homegrown talent with foreign talent lured by a business-supportive environment and competitive investment in R&D. But that does not justify complacency going forward. As competition for skilled labor increases around the world, it would be wise to streamline visa processing times and encourage foreign researchers to stay by offering more permanent resident visas (green cards) to newly minted PhDs from US universities. The problem for knowledge-based economies is not too many immigrants—it is retaining too few skilled ones. The sooner policy makers recognize that, the better.
181
editorial
H1N1dsight is a wonderful thing Criticisms of the response of governments and of the pharmaceutical industry to the threat of the H1N1 epidemic are wide of the mark.
© 2010 Nature America, Inc. All rights reserved.
T
he pandemic of influenza strain H1N1 has not yet manifested itself as the pestilence of biblical proportions that many had anticipated. According to the World Health Organization (WHO), swine flu is present in 212 countries and, to mid-February, has caused nearly 16,000 deaths worldwide. This is substantially lower than the death toll for seasonal influenza, which the US Centers for Disease Control (CDC) puts at around 30,000 a year in the United States, and a hundredfold lower than some initial estimates suggesting that one million deaths might be expected across the population of developed nations. The lack of virulence of H1N1 has left many countries with a stockpile of unused H1N1 vaccine. This is now prompting accusations that the WHO was guilty of scaremongering, with the complicity of drug companies. In reality, there’s no reason to think that the WHO or drug companies should or could have acted differently. And the fact that a much larger proportion of the world’s population is now immune to H1N1 than before the pandemic represents a great step forward in protecting the health of the world’s population. One downside of the relative mildness of H1N1 is that European governments now find themselves with vaccine stockpiles and are looking to trim their orders: Germany cancelled 30% of its order for GlaxoSmithKline (GSK)’s Pandemrix; the French health minister, Roselyne Bachelot, has cancelled 50 million doses; and the Belgian government has cancelled around a third of its 12.6 million–dose order. The UK has not bought up an option it had on vaccine from Baxter, and it is currently sitting on a vaccine stockpile of around 20 million doses with the prospect of buying another 30 million from GSK under a contract that is not reversible. The US government, too, has a huge vaccine stockpile because only 61 million US citizens have been vaccinated. All this has created a finger-pointing environment in which governments are now being called to task for overreacting to the threat from swine flu. On December 18, a Socialist representative in the German Bundestag, Wolfgang Wodarg, and 13 other European national parliamentarians signed a motion claiming that “pharmaceutical companies have influenced scientists and official [public health] agencies” leading them to “squander tight health care resources [on] inefficient vaccine strategies and needlessly exposed millions of healthy people to the risk of unknown side-effects of insufficiently tested vaccines.” This motion resulted in a hearing at the end of January convened by the Parliamentary Assembly of the Council of Europe to discuss the issue of “fake pandemics” (as the parliamentary group describes it). These events raise several issues. It is disturbing that a rogue politician with a good sense of the zeitgeist can create such mischief for the WHO and drug industry, despite virtually no evidence of wrongdoing. Wolfgang Wodarg was the prime mover behind the fake pandemic outcry. He is a physician and a self-proclaimed expert in lung disease, who left medical practice in 1994. Back in August 2009, early in the swine flu outbreak, he
182
was already talking up the possible side-effects of the vaccine. Wodarg has a history of dubious positioning with respect to biotech. Back in 2005, for instance, he was responsible for a report to the European Parliament on genetically modified organisms (GMOs) that called for stricter regulation of labeling, liability, good farming practice and GMO-free zones. The fact that the media continue to give credence to his kind ought to trouble the life science industry. Beyond the machinations of antibiotech gadflies, however, there are more serious matters. Deeper reflection suggests that it would have been very difficult both politically and economically for any developed nation’s government not to have invested significantly in vaccine and flu drugs. Faced with the certainty of a new influenza virus to which a large proportion of the world’s population was immunologically naive, and the uncertainty of the predictive epidemiological models, governments had little political choice but to act, anticipating something close to the worst case scenario. And the options they had at hand were somewhat limited. Closing national borders is an option, although this is increasingly difficult given the volumes of travel and international trade that characterize our global village. Another option is to close schools. Although this would restrict spread, it comes with a substantial cost. In September 2009, a report from the Brookings Institution found that closing all schools in the United States for four weeks might cost between $10 billion and $47 billion in parent absenteeism from work. A Swedish study published in the same month (Eurosurveillance 14, 5–11, 17 September 2009) showed that even without school closure, work absenteeism would represent a substantial economic burden as workers became ill themselves or took time from work to tend sick children. The authors estimated that vaccination against pandemic H1N1 influenza would be highly cost-effective, saving ~€250 ($340) million. Other economic forecasters painted still gloomier pictures; in late 2009, the World Bank estimated that a worldwide pandemic could wipe out 0.7–4.8% of global gross domestic product. Thus, it is difficult to conclude, even now, that governments that bought into preventative strategies for H1N1 made the wrong decision. Although the nature of the threat may have been overstated, the WHO, CDC and other authorities had little scientific evidence at the beginning of the H1N1 pandemic to discount the most dire predictions of fatalities. The only question that remains is whether the purchase of vaccine orders by different countries should be better coordinated on an international scale. At the moment, different nations essentially compete with one another to grab a share of the vaccine supply from manufacturers. But is it justifiable that those nations that widely overestimate their demand— presumably taking supply away from other countries—can then return their stockpile to the drug companies that they contracted to supply vaccine in the first place?
volume 28 number 3 march 2010 nature biotechnology
news in this section Artifact controversy over GSK/Sirtris compounds p185
Biotechs continue pursuit of biodefense dollars
China storms into world genomics p189
p187
Gene therapy trailblazer Ark Therapeutics was dealt a blow last December, when the European Medicines Agency (EMEA) rejected the London-based company’s marketing application for its adenoviral gene therapy Cerepro (sitimagene ceradenovec) in malignant glioma. According to EMEA, Cerepro’s trial data was statistically underpowered and failed to show sufficient efficacy in terms of postponing death or re-intervention. The decision sent Ark’s share price into a nosedive, plunging 43% to barely above the value of its cash assets, and is widely viewed as a setback for gene therapy as a whole, with Cerepro having represented the field’s best hope for approval in the West. All may not be lost, however. Ark is disputing the agency’s estimate of Cerepro’s overall survival benefit to glioma patients. The company still hopes the EMEA decision may be reversed on appeal, which is in process and likely to be considered next month. The EMEA’s negative opinion is disappointing because previous announcements relating to Ark’s Cerepro clinical data suggested that it could be a useful adjunct to surgical treatment of glioma. Ark is one of the firms at the forefront of commercial gene therapy (Nat. Biotechnol. 26, 1057–1059, 2008) and has several other gene therapy products well advanced in its pipeline—some of them, according to financial analysts, even more promising than Cerepro, which had anticipated worldwide annual sales of $200 million. Cerepro consists of a herpes simplex virus gene for thymidine kinase encased in an adenoviral vector from which the E1 and part of the E3 regions have been deleted to prevent viral replication. The thymidine kinase is capable of converting Basel, Switzerland–based Roche’s prodrug ganciclovir into its highly toxic form, deoxyguanosine triphosphate. Indicated for people with operable, high-grade malignant glioma, Cerepro treatment involves injection of the gene therapy vector into the cavity that remains after tumor resection followed by administration of ganciclovir. After taking up the thymidine kinase adenovirus package, any residual tumor cells left after surgery are killed when they convert ganciclovir into its toxic form.
R. Bick, B. Poindexter, UT Medical School/Science Photo Library
© 2010 Nature America, Inc. All rights reserved.
Ark’s gene therapy stumbles at the finish line
Glioblastoma, a malign form of glioma shown here as a computer model, remains a hard-to-treat brain tumor. This type of brain tumor has the worst prognosis of any central nervous system malignancy.
In April 2009, Ark provided an update on positive phase 3 trial results reported the previous year (Nat. Biotechnol. 26, 1057–1059, 2008), announcing that Cerepro treatment resulted in statistically significant improvement of the primary endpoints of death or re-intervention from around four months after surgery. The EMEA’s negative opinion of Cerepro therefore came as something of a surprise. According to the agency’s Committee for Medicinal Products for Human Use (CHMP), Ark’s phase 3 clinical data did not prove Cerepro’s efficacy in terms of postponing death or re-intervention. What’s more, the treatment increased the risk of serious side effects such as paralysis of one side and seizures. Thus, the decision to reject was ostensibly made on a purely routine risk-benefit determination.
nature biotechnology volume 28 number 3 march 2010
There may be more to it than that, however. Sam Fazeli, senior research analyst at broker Piper Jaffray in London, points out that the CHMP marked down its estimate of Cerepro’s overall survival benefit to only 16–24%; earlier data presented by the lead investigator of the phase 3 trial, Zvi Ram of Tel Aviv Medical Center in Israel, at last October’s US Society for Neuro-Oncology meeting in New Orleans had suggested a sustained 57% improvement over standard care. One reason given by the CHMP for their lower estimate was that some patient subgroup analyses were not statistically significant. That the trial was conducted open label was an additional complicating factor, as investigator bias cannot be fully ruled out. “Physician subjectivity in the re-intervention decision could have favored Cerepro—that is, Cerepro patients may have been left longer by surgeons prior to
183
NEWS
in brief Monsanto’s alfalfa reaches Supreme Court
© 2010 Nature America, Inc. All rights reserved.
Bo Insogna/istockphoto
In April, the US Supreme Court will hear Monsanto’s case for why it should be cleared to resume reselling Roundup Ready alfalfa seeds. The verdict, which is expected to affect the regulation Alfalfa is one of of other biotech the most important crops, including legumes in agriculture. genetically modified (GM) sugar beets, could make it easier for GM crops to stay on the market, as it will no longer be possible to ban a crop, once approved, without a full hearing. Monsanto’s GM alfalfa was approved by the United States Department of Agriculture (USDA, Washington, DC) in 2005, but the Center for Food Safety in February 2006 sued the USDA for not properly investigating the impact of the GM seeds on the environment. The United States District Court for the Northern District of California in 2007 banned the GM alfalfa seeds nationwide, pending a draft environmental impact statement (EIS) from the USDA. Monsanto appealed, and the case has now worked its way to the US Supreme Court. Peter McHugh, deputy general counsel at the Biotechnology Industry Organization (Washington, DC), says he disagreed with the process applied in the lower courts, adding that if Monsanto wins, in the future the farmers, growers and seed producers of agbio “will have an opportunity to have a full and fair evidentiary hearing before there’s an injunction.” In short, the ruling will determine whether a product can be banned without a hearing after it has been given the agency’s blessing. Drew Kershen, a professor of law at the University of Oklahoma, says it’s “important to set the standard when injunctions can be used, when the argument is that USDA’s Animal and Plant Health Inspection Service (APHIS) needs to stop and prepare an EIS.” If the Supreme Court overturns the ban on alfalfa, it would mean that producers and users of GM seeds facing an injunction do not need to stop selling and planting their GM crops immediately, if at all. Either way, the ruling should affect other agbiotech court cases, specifically a case due to begin in March, also filed by the Center for Food Safety, against Monsanto’s GM sugar beets. There is more at stake where beets are concerned (Nat. Biotechnol. 27, 970, 2009), because whereas Roundup Ready alfalfa seeds make up only 1% of the market, sugar beets were deregulated in 2005, and today 95% of sugar beets sold are from Roundup Ready GM seeds. Boonsri Dickinson
184
intervention,” explains Fazeli. This would have further weakened the significance of the study findings, and it is on this point that Ark is seeking to change the committee’s mind. The firm believes it can demonstrate that investigator bias could not have been an important factor and so the analyses were probably statistically significant. However, the data could be compromised by a much more fundamental problem, namely the small numbers treated. Only 119 patients received the therapy in the phase 3 trial, far fewer than would be accepted in a pivotal trial for any small-molecule drug. “The crucial question is, has there been enough patient exposure?” says Robin Davison, analyst at Edison Investment Research in London. He suspects that the EMEA may be reluctant to authorize such a novel therapy based on no more than 300plus patients in total, even for a condition with very short life expectancy. “That must have been a factor, because Ark is pioneering a regulatory pathway here,” he notes. “Regulators are faced with a fine balance, knowing that granting the first approval for gene therapy based on a relatively small exposure might open the door to similar approvals. That must be weighing on their minds—they may feel they may be opening Pandora’s box.” These issues are particularly acute for gene therapy products because of the field’s checkered history in terms of clinical successes and media hype over adverse events. The death 11 years ago of a patient in an adenoviral gene therapy trial at the University of Pennsylvania (Nat. Biotechnol. 23, 519–521, 2005) and more recent reports of toxicity in patients receiving adeno-associated virus (AAV) gene therapies (Nat. Biotechnol. 25, 949, 2007) are likely to have attuned regulators to safety issues. Companies with gene therapies in human trials, however, remain upbeat. Another European gene therapy company, Amsterdam Molecular Therapeutics (AMT) of Amsterdam, has filed a Marketing Authorization Application with EMEA for its lipoprotein lipase (LPL) deficiency treatment Glybera (a recombinant AAV vector expressing the Ser447X variant of the human LPL gene). “I cannot suspect any hidden agenda at the EMEA,” says AMT chief executive officer Jörn Aldag. “We don’t think the negative opinion has anything to do with gene therapy in general because EMEA’s reasons are very clearly documented as benefit versus risk.” The company is seeking approval for Glybera in Canada and expects to file in the US in 2012. The therapy could be available as early as next year.
An Ark spokesman says gene therapies, such as Cerepro, are “clearly” open for approval based on clinical results in the same way as established pharmaceuticals. The company notes that the Cerepro setback is “not entirely surprising,” given the newness of the therapeutic approach and the notorious difficulties of trial designs around malignant glioma. In fact, AMT’s Aldag’s perception is that gene therapy is seen increasingly by regulators as a favorable treatment mode: “The skepticism we have experienced in the past has actually declined.” He says AMT was positively encouraged by regulators to file marketing applications for Glybera. Here again, however, numbers in the clinic are small—studies so far have involved 27 patients. The EMEA’s opinion of Cerepro could sway thinking the other side of the Atlantic. According to Fazeli at Piper Jaffray, US physicians have been viewing the European regulatory process for Cerepro with “intense interest.” If EMEA approval occurs, this may be a “gating factor” for a US marketing deal, he says. There is little doubt that the US Food & Drug Administration (FDA) is watching Europe’s gene therapy licensing policy very closely. The FDA holds regular bimonthly discussions with the EMEA’s Committee for Advanced Therapies, the body ultimately responsible for the negative opinion on Cerepro. And according to Martyn Ward, head of clinical trials at the UK’s Medicines & Healthcare Products Regulatory Agency (MHRA), FDA officials have in the past taken their lead from Europe when granting permission for early stage gene therapy trials. Companies in the business know that the FDA is hanging back and plan their regulatory strategy accordingly. “Ark wanted to get EMEA approval so as to force FDA to come to a position,” says Edison’s Robin Davison. “Once a gene therapy has been approved in Europe, the FDA would feel political and patient pressure to consider it,” he says. “And they wouldn’t be able to rewrite the rules [for a clinical study] from scratch.” But now, he says, nobody knows what the prospects for FDA submission might be. As Nature Biotechnology went to press, an announcement from EMEA on the re-examination of Cerepro data was imminent. But the prospects are sobering. In the past four years, only ~30–40% of appeals on EMEA decisions have been successful—Roche’s successful appeal for Tarceva (erlotinib) in pancreatic cancer in December 2006 is a notable example. From that perspective, the odds of the first gene therapy getting a green light from Europe’s regulators look slightly worse than those of flipping a coin. Peter Mitchell London
volume 28 number 3 march 2010 nature biotechnology
news
In January, scientists at Pfizer Global Research and Development, in Groton, Connecticut, reported that small molecules developed by Cambridge, Massachusetts–based Sirtris Pharmaceuticals do not activate the sirtuin pathway that has been linked to longevity. The disconcerting discovery, published in The Journal of Biological Chemistry (published online, doi:10.1074/jbc.M109.088682, 8 January 2010), is not the first to cast doubt on compounds that target sirtuins. In a paper published last November (Chem. Biol. Drug Des. 74, 619–624, 2009), researchers at Amgen of Thousand Oaks, California, also showed that a purported anti-aging compound in red wine, resveratrol, doesn’t act on the pathway either. These findings are fueling skepticism not just concerning Sirtris and its resveratrol-like compounds but also about the due diligence process at Londonbased GlaxoSmithKline, which purchased the Cambridge, Massachusetts–based biotech in April 2008 for an eye-popping $720 million (Nat. Biotechnol. 26, 595, 2008). The controversy over Sirtris drugs reached a tipping point in January with a publication by Pfizer researchers led by Kay Ahn showing that resveratrol activates SIRT1 only when linked to a fluorophore. Although Ahn declined to be interviewed by Nature
Biotechnology, a statement issued by Pfizer says the group’s findings “call into question the mechanism of action of resveratrol and other reported activators of the SIRT1 enzyme.” Most experts, however, say it’s too soon to write off Sirtris’ compounds altogether, assuming they’re clinically useful by mechanisms that don’t involve sirtuin binding. And for its part, GSK won’t concede that Sirtris’ small molecules don’t bind the targets. In an e-mailed statement, Ad Rawcliffe, head of GSK’s WorldWide Business Development group, says, “There is nothing that has happened to date, including the publication [by Pfizer,] that suggests otherwise.” The evidence Sirtris brought to the table came with a complicated history, which GSK claims to have been well aware of. Scientists have for years expressed skepticism about the company’s core premise: that sirtuins emulate the anti-aging benefits of calorie restriction, and that by activating sirtuins with drugs, it’s possible to treat age-related diseases (Nat. Biotechnol. 26, 371–374, 2008). In 2005, Matt Kaeberlein, of the University of Washington, Seattle, published the first data showing that calorie restriction doesn’t activate sirtuins in yeast (Science 310, 1193–1196, 2005). Those data have since been replicated in othera
Dr David Becker/Wellcome Images
© 2010 Nature America, Inc. All rights reserved.
GSK/Sirtris compounds dogged by assay artifacts
In yeast and worms as the C. elegans pictured above, sirtuins can extend lifespan by up to 70 percent. Sirtris hopes to develop a pill that might do the same for humans, or at least ward off the diseases of aging.
nature biotechnology volume 28 number 3 march 2010
185
© 2010 Nature America, Inc. All rights reserved.
NEWS laboratories, he says. In mice, calorie restriction triggers the SIRT1 enzyme, but the company’s small molecules target SIRT1 in only some tissues and not all, Kaeberlein adds. Also in 2005, Kaeberlein and John Denu, from the University of Wisconsin, Madison, independently showed that resveratrol activates SIRT1 only when the sirtuin is bound to a fluorescent label. Pfizer scientists also report that Sirtris’ small molecules, including SRT501—currently in phase 2 clinical trials to treat type 2 diabetes—bind SIRT1 only in the presence of this fluorophore. These drugs do, however, bind something: unlabeled native proteins, which presumably may influence sirtuin pathways in the cell. “We still don’t know what resveratrol or the Sirtris compounds actually bind to in cells,” says Kaeberlein. In an e-mailed response, GSK’s Rawcliffe states that the company knew SIRT1 modulation in vitro is complicated and was aware of the controversies surrounding the fluorescence-based assay and the precise mechanism of action of the published compound(s). Nonetheless, Rawcliffe says GSK remains confident that the activity seen in cell-based and animal studies are acting through an SIRT1-dependent mechanism. As to GSK’s due diligence in the Sirtris deal, Rawcliffe remains confident that their process did an “excellent job in allowing us to understand the full view of the scientific field and to place Sirtris in perspective.” Even so, unsubstantiated comments on Derek Lowe’s ‘In the Pipeline’ blog (http://pipeline.corante.com/) allege that the deal went through against the advice of some internal GSK scientists and that similar due diligence processes on the Sirtris compounds at other companies, such as Basel, Switzerland–based Novartis and Amgen, had raised flags about artifacts in the assays. According to Uwe Schoenbeck, Pfizer’s chief science officer for external R&D and innovation (who did not comment directly on either GSK or Sirtris), due diligence is essentially a risk assessment. “You’re looking to make the best educated guess about a company with the data available to you,” he says. Pharmaceutical companies are generally looking for a strategic fit, he explains, and trying to identify opportunities that neither they nor the acquired firms can exploit in isolation. The first step is a nonconfidential exchange of data and information. Assuming both parties share mutual interest in a deal, the due diligence then moves on to an exchange of confidential information coordinated in part through a materials transfer agreement. This gives both parties access to what’s known as a confidential data package
186
containing sensitive information—raw data and chemical structures of lead compounds, for instance—that showcase a biotech’s competitive advantage. Due diligence investigators represent cross-sectional areas of expertise—discovery and development, toxicology and intellectual property, among others—assembled under a single lead, Pfizer’s Schoenbeck says. According to another due diligence expert, who requested not to be identified, GSK’s aim in buying Sirtris would have been to secure access to lead compounds, with proof of concept for therapeutic utility. That view was confirmed by Rawcliffe, who said the promise that sirtuin biology could yield “transformational medicines” was apparent to GSK during the acquisition. The fact that GSK paid $720 million for Sirtris ($22 a share, as compared with the market’s valuation of $12 a share) also suggests that GSK was competing with other suitors, which drove up the offer price. How each company assesses the promise and value of a biotech’s assets, however, varies “philosophically” from pharma company to pharma company, Schoenbeck says. Pharma scientists engaged in due diligence may try to replicate a biotech firm’s results in a pilot study. This is especially true when the data are uncertain, as would have been the case with Sirtris. Peter DiStefano, chief scientific officer with Cambridge, Massachusetts–based Elixir Pharmaceuticals, claims the commercial Fluor-de-Lys fluorometric detection assay kit from Enzo Lifesciences that Sirtris relies on for binding evidence can be unreliable, given that compounds often bind to the fluorophore itself. Results from that test should be confirmed with counterscreens, and more expensive and cumbersome radiolabeling assays that provide more definitive conclusions, he says. GSK’s Rawcliffe would not say whether the company had used counterscreens in its due diligence. And when asked whether the company had done its own pilot study of the Sirtris compounds, he appeared to answer in the negative. “As part of this diligence, we investigated these controversies in a number of ways,” he wrote. “[That included] speaking to many people on both sides of the argument. We are satisfied with the outcome of that process.” Sources interviewed for this article speculate that GSK and Sirtris might have more convincing data that they haven’t yet shared with the public. And for his part, Rawcliffe claims GSK/Sirtris are planning to publish results that, he says, elucidate how their small molecules might activate SIRT1 and how that relates to disease processes. But Brian Kennedy, also from the University of Wisconsin, counters that Sirtris has been
threatening to publish these data for years. When Kennedy was a postdoctoral student at the Massachusetts Institute of Technology, he studied under Leonard Guarente, who is now an advisor to Sirtris. He’s since published findings showing that resveratrol does not bind the yeast sirtuin SIR2 (J. Biol. Chem. 280, 17038– 17045, 2005). And together with Kaeberlein, he found that calorie restriction does a good job at extending yeast lifespan even in a SIR2-gene knockout strain, suggesting the two pathways are unrelated. Kennedy says he’s perplexed when Sirtris scientists—notably the company’s cofounder David Sinclair—claim possession of compounds that bind SIRT1 more effectively than resveratrol, when data suggest that resveratrol doesn’t bind SIRT1 in the first place. “It’s possible that Sirtris and GSK have information that resolves these issues,” he concedes. “I haven’t seen it, so at this time I remain skeptical. I agree their motivation isn’t necessarily to enlighten the public, but you have to wonder by this point why they’re holding it back.” At the same time, Kennedy acknowledges that resveratrol and other purported sirtuin activators do seem to confer metabolic benefits in rodents, even if they don’t extend lifespan. “That’s a conundrum,” he says. These metabolic benefits were recently refuted in the new Pfizer paper, which found that the Sirtris compounds don’t lower plasma glucose in obese mice fed high-fat diets, as reported earlier by Sirtris scientists. GSK spokesperson Janet Morgan attributes those contradictory findings in part to impurities in Pfizer’s prepared versions of the Sirtris compounds. And Kaeberlein says that of all Pfizer’s new findings, these appear to be the least robust. “[Pfizer] didn’t seem to make a strong case for this,” he says. “And if they had, that would have been a surprise, because of all the things resveratrol might do, its benefits in diabetes and obesity seem to be the most believable.” Both Kennedy and Kaeberlein suggest resveratrol and resveratrol-like compounds might yield health benefits through other pathways. Thomas Hughes, president and CEO of Cambridge, Massachusetts–based Zafgen, says he’s not surprised that GSK went after Sirtris, despite the controversial nature of its research. “The whole field of drug discovery has been incredibly energized by this prospect of anti-aging biology and its influence on new pathways that could work in diseases like Alzheimer’s and diabetes,” he says. “And when you have something so potentially disruptive, it’s rare to have a situation where you can’t pull equal stacks of papers that support or refute the idea.” Charlie Schmidt Portland, Maine
volume 28 number 3 march 2010 nature biotechnology
news
The US Department of Health and Human Services (HHS) has signed a contract worth up to $143 million with Elusys Therapeutics to develop and manufacture the Pine Brook, New Jersey–based biotech’s anti-anthrax toxin monoclonal antibody (mAb), Anthim. The Biomedical Advanced Research and Development Authority (BARDA) will pay Elusys $16.8 million in the first year, with options for additional funding over the next four years, to develop and commercialize the high-affinity mAb. As countermeasures for an anthrax attack are still lacking, the US government—still smarting from stinging criticism of its bioterrorism preparedness in a January report from the bipartisan Commission on the Prevention of Weapons of Mass Destruction Proliferation and Terrorism— continues to throw money at any biotech companies prepared to chance the vagaries of being a government contractor. In February, the Obama administration announced $135 million increase in BARDA contracts, earmarked for next-generation anthrax and acute radiation syndrome countermeasures. Project BioShield, set up in 2004 to accelerate medical countermeasures for national security and administered by BARDA, has left a trail of plummeting stock prices as contracts have been canceled, agreements rescinded and terms renegotiated. John Clerici, a founding principal at consultancy Tiber Creek Partners, says his group advises companies to avoid biodefense work as a sole business model for a company. “If you want to create a business that’s solely built around being a government contractor and developing drugs only for that government market, we believe that’s destined for failure, and it’s played out badly for those companies that have tried to do it,” says Clerici, whose firm specializes in advising biotechs on non dilutive capital, particularly for biodefense and medical countermeasures contracts. Therapies against anthrax are a case in point. The latest Elusys deal follows two failed attempts by the government to procure secondgeneration anthrax vaccines for its strategic stockpile. First, a contract worth $877.5 million with South San Francisco, California–based VaxGen was canceled in 2006, after the vaccine proved too unstable. VaxGen stock, which had once been valued at more than $30 a share, dropped to less than a dollar and never recovered. And last December, BARDA backed off from a request for proposals (RFP) after several companies failed to meet the requirement that they be ready for licensure within eight years. On the heels of this cancellation, shares of two Maryland-based biotech companies working on anthrax vaccines, Emergent BioSolutions
in Rockville and PharmAthene in Annapolis, both plummeted. Had the RFP gone through, it would have resulted in a contract worth $600 million for the winning company, and the government would ultimately have purchased 25 million doses. Other notable victims in the biodefense roulette include Acambis (now part of Sanofi-Pasteur of Paris), whose contract for smallpox vaccine was canceled by BARDA in 2007, and HollisEden Pharmaceuticals of San Diego, which suffered the same fate over a treatment for acute radiation syndrome. Although bureaucracy and poor communication have played a part in these debacles, the blame does not lie entirely at the door of BARDA or HHS. In the case of the withdrawn RFP affecting PharmAthene and Emergent BioSolutions, all sides agreed that none of the submitted proposals met the requirements for funding under the BioShield Legislation Act, which specifies that BARDA should support development of products that are in late-stage development (that is, close to use in humans). Likewise, BARDA is not accountable for the failure of several products to meet clinical testing benchmarks. Even so, there is concern that the emphasis on late-stage development has left a financing lacuna for the risky early stages of product development. BioShield legislation was written with the expectation that at least some of that risk would be shared by contracting companies with the size and experience to manage it. But
big pharma has consistently shunned bioterrorism countermeasures, leaving small biotechs to step in. For many fledgling firms, a single biodefense contract could represent all or most of its market value. BARDA is well aware of the risks facing smaller companies and the dire consequences of contract cancellations. In December, HHS secretary Kathleen Sebelius announced a comprehensive countermeasure review that is currently ongoing. Robin Robinson, director of BARDA, expects this review to highlight the best solutions moving forward. “Certainly, we know from our meetings with the White House that the President is very much in favor of us making big and bold movements forward with this,” he says. Perhaps wiser from past experience, BARDA has converted its anthrax vaccine RFP to a modified broad area announcement (BAA) intended to support development of recombinant protective antigen (rPA)-based anthrax vaccines. In addition, they offered PharmAthene the opportunity to negotiate a modification of a pre-existing development contract for its rPA vaccine candidate, SparVax. PharmAthene president and CEO David Wright says, “At the end of the day, BARDA and the government will be having the same commitment, and putting the same amount of money or even more money in, to ensure a less risky road for the government in procurement.”
Photo/Tom Gannam
© 2010 Nature America, Inc. All rights reserved.
US biodefense contracts continue to lure biotechs
In the US, anthrax still rates as the number one bioterror threat. Here, a hazardous materials team disposes of material taken from a student at University of Missouri-Rolla in 2007, who claimed to possess anthrax and a bomb.
nature biotechnology volume 28 number 3 march 2010
187
NEWS
Company
Product
Stage of development
BARDA contract
Human Genome Sciences
ABthrax (raxibacumab), a human mAb against anthrax protective antigen (PA)
Biologic license application
$151 million
Emergent BioSolutions
AV-7909, a combination of BioThrax (aluminum-adsorbed cellfree filtrates of unencapsulated Bacillus anthracis) and Coley Pharmaceutical’s VaxImmune (an unmethylated CpG-motif oligonucleotide that acts as an agonist of Toll-like receptor 9)
Phase 2
$447.6 million
Anthrax immune globulin (AIGIV), polyclonal antibodies raised against BioThrax
Phase 1/2
$13 million
Elusys Therapeutics
Anthim (ETI-204), a humanized mAb against PA
Phase 1
Up to $143 million
PharmAthene
SparVax, an injectable rPA absorbed on to hydrogel
Phase 2
$3.9 million NIAID
DynPort Vaccine
Anthrax vaccine, an injectable rPA vaccine
Phase 1
NA
Medarex, a subsidiary of Bristol-Myers Squibb
Valortim (MDX-1303), a fully human mAb against PA
Phase 1
$1 million from the US Department of Defense (DoD) payable to partner PHarmAthene
Advanced Life Sciences
Restanza, a once-daily oral ketolide cethromycin that inhibits B. anthracis protein synthesis
Preclinical
$3.8 million from DoD
Sources: Sagient Research, BiomedTracker and BARDA. NA, Not available.
Even with government support, PharmAthene and other companies working under biodefense contracts face the stark reality of drug development: 30% of all candidates that reach phase 3 clinical trials within the eight-year BARDA procurement time are likely to fail. According to some policy experts, this ought to be a good reason for the government to contract with large pharmaceutical companies. “There’s a systemic problem, which is that most of the contracts for developing new biodefense measures are with small biotech companies,” says Gregory Koblentz, an assistant professor at George Mason University and deputy director of the Biodefense Graduate Program, “and these companies tend to be undercapitalized, understaffed and don’t have the depth and breadth of experience to take a drug from the research and development phase through clinical trials, scale-up to large-scale production and licensure.” But pharma has so far has been notably absent from biodefense contract competitions. One possible explanation is the perception that government is a bad customer. At least this was the experience for Bayer (Leverkusen, Germany) during the anthrax attacks of 2001. As the demand for Bayer’s antibiotic Cipro (ciprofloxacin) soared, a faction led by Senator Charles Schumer (D-NY) suggested that the government should use existing law to issue a compulsory purchase order suspending Bayer’s patent. This would have allowed other manufacturers to make generic Cipro and charge a lower price. In the end, this threat was never enacted, and experts doubt it could have been under the circumstances. However, the incident is well remembered (and resented) and raises the question of whether any company with a patent covering an important bioterrorism countermeasure could risk its patent being threatened at a later date.
188
For the US government, anthrax treatment remains a critical element in its biodefense strategies. Emergent Solutions currently manufactures the only approved vaccine for anthrax based on protective antigen (PA). This is an older ‘legacy’ vaccine and, although considered safe and effective, requires six injections over 18 months to be fully effective. The hope for the next-generation, or rPA, anthrax vaccines is that they will provoke a stronger immune response and not require such a cumbersome vaccination schedule. Antibiotics and therapeutics are also important pieces of the anthrax countermeasure puzzle because an anthrax infection has different phases and because it can be so rapidly lethal. Antibiotics are effective against active bacterial infection, for example, but not spores and not a latestage infection. Emergent BioSolutions is responding to the withdrawn RFP by refocusing on development of its original anthrax vaccine (BioThrax; aluminum-adsorbed cell-free filtrates of unencapsulated Bacillus anthracis), seeking FDA approval for a modified, four-dose regimen. Meanwhile, the company’s $400 million
procurement contract with the Centers for Disease Control and Prevention of Atlanta, Georgia, to manufacture and deliver 14.5 million doses of BioThrax for the strategic national stockpile remains unchanged. “This is just one of many starts and stops along the way that we’ve seen in the past four years. We understand that process well, as we have been a player in this space for more than a decade,” says Daniel Abdun-Nabi, president and COO of Emergent BioSolutions. The BioShield budget announced last month is more flexible than those of previous years, as it can allocate resources for R&D and for companies running projects at early stages of development. But there is no ready market for biodefense countermeasures other than the government. As Elizabeth Posillico, president and CEO of Elusys, points out, “There’s one customer and the company has little control over the decision to purchase the drug. It’s not driven by the market so much as the customer’s needs.” So chasing government contracts alone, however lucrative, will continue to be a risky strategy for biotech. Catherine Shaffer Ann Arbor, Michigan
Ride ‘n Drive on government waste Danish enzyme manufacturer Novozymes invited journalists to burn up official waste by taking a spin on the flex-fuel Chevy HHR truck at the Washington Auto Show in January. The engine was powered by paper discarded by White House offices. Novozymes
© 2010 Nature America, Inc. All rights reserved.
Table 1 Anthrax countermeasures in development
volume 28 number 3 march 2010 nature biotechnology
news
China’s BGI has become the largest nextgeneration genome sequencing center in the world after it purchased 128 new HiSeq 2000 genome sequencers from Illumina. The deal was announced in January by Jay Flatley, president and CEO of San Diego–based Illumina, at the 28th Annual J.P. Morgan Healthcare Conference in San Francisco (Box 1). The announcement comes amid a boom of scientific productivity in China centered around next-generation sequencing technology, which has resulted in the publication of three landmark papers by Chinese researchers in the past two months alone: the sequencing of the cucumber (Nat. Genet. 41, 1275–1281, 2009) and giant panda (Nature 463, 311–317, 2010) genomes and the human pan-genome (Nat. Biotechnol. 28, 57–63, 2010). BGI, which started life as the Beijing Genomics Institute but then moved to Shenzhen, China, will install most of the newly acquired Illumina sequencers at its new genome center at Hong Kong Science Park throughout 2010, says BGI executive director Jun Wang. “We will use these instruments to help build research and application platforms for sustainable development in agriculture, bioenergy, personalized health care and related fields in China.” The emergence of BGI as a sequencing powerhouse could have a significant upside
for Hong Kong’s biotech sector. “Having the largest DNA sequencing facility in the world in Hong Kong will hopefully contribute towards attracting talent in genomics and bioinformatics and stimulate collaboration with other local research groups, such as those in the two medical schools,” says Dennis Lo, director of the Li Ka Shing Institute of Health Sciences (LiHS), a translational medicine research institute of The Chinese University of Hong Kong (CUHK). Lo, who is also associate dean for research of the Faculty of Medicine at CUHK, adds that the facility’s international visibility might also “trigger more investment from the Hong Kong government and the commercial sector to develop biotech in Hong Kong.” LiHS recently installed 10 Illumina Genome Analyzers, eight of which are part of a joint CUHK-BGI genome research center that intends to conduct collaborative “projects in the fields of cancer, diabetes and plant genomics,” he says. BGI’s vice president, Xiuqing Zhang, also points to a deliberate international agenda. “BGI’s investment in Illumina’s new HiSeq 2000 system is an important step in our effort to develop a premier sequencing facility that serves scientists globally,” he says. The HiSeq 2000 platform is capable of generating two billion paired-end reads and 200 gigabases of quality-filtered data in a single run, allowing
BGI
© 2010 Nature America, Inc. All rights reserved.
Chinese institute makes bold sequencing play
China’s BGI, now in Shenzhen, has become Illumina’s largest customer overnight.
nature biotechnology volume 28 number 3 march 2010
in brief Melanoma vaccine for dogs A canine melanoma vaccine has received a full license from the US Department of Agriculture, the first therapeutic cancer vaccine approved for human or animal use. The San Diego–based Vical sees the approval of its DNA vaccine Oncept as an indicator of potential success for its human therapeutic vaccine currently in development for metastatic melanoma. Others are more cautious. “[It] is quite an achievement, but I don’t believe that Oncept’s approval has brought us any closer to a human therapeutic cancer vaccine, as researchers have seen cures in animal models of melanoma for quite some time,” says Martin Bachmann, of Cytos, a company in Schlieren, Switzerland. Oncept contains a gene encoding human tyrosinase, an enzyme associated with skin pigmentation, which stimulates an immune response against canine tyrosinase in melanoma cells. “Canine melanoma is similar in disease course and spread to human melanoma, so the results could be relevant for predicting response in humans for a similar DNA vaccine. However, the study was not randomized, and it’s hard to forecast how the human immune system will respond,” says Christian Ottensmeier, a Cancer Research UK investigator at Southampton University. Oncept will be commercialized by Vical’s licensee, Merial, the animal health subsidiary of Paris-based Sanofi-aventis. Suzanne Elvidge
Biotechs go virtual Outsourcing the early stages of R&D is a growing trend among young biotech firms in the UK, a new report reveals. Researchers at Cass Business School in London tracked 68 university and public service laboratory spin-outs as part of a larger Engineering and Physical Sciences Research Council (EPSRC) project on high-tech business organization. The study reveals that up to one-third of these firms have embraced an innovative ‘virtual biotech’ business model to help reduce the time taken to reach clinical trials and build up a pipeline of early stage products. According to Dzidziso Samuel Kamuriwo, the report’s author, this business model has flourished among fledgling biotechs thanks to a combination of local policies that favor the industrialization of public science, multiple sources of funding and high-quality science conducted in public labs. The advantages of going ‘virtual’ include flexibility and few or no capital costs, which help slow cash burn (Nat. Biotechnol. 27, 886–888, 2009). Companies that adopt this virtual approach tend to outsource R&D to their founding institutions or to specialist service firms and are dependent on strong project management to succeed. The virtual model works best across several therapeutic areas, while the integrated model—keeping everything in house—thrives on fewer areas. In the long run, Kamuriwo points out, integrated companies are more likely to succeed, but after experiencing the short-term gains of the virtual approach, companies find it hard to change their organization model. Susan Aldridge
189
NEWS
© 2010 Nature America, Inc. All rights reserved.
Box 1 A human genome in a week The BGI deal represents Illumina’s largest single order for sequencers so far and is estimated to be worth over $88 million, given that the retail list price of the HiSeq 2000 is $690,000, although BGI may have received a significant volume discount. BGI signed an agreement last month with the China Development Bank under which BGI will receive $1.5 billion in ‘collaborative funds’ over the next ten years, which the institute plans to use for infrastructural development and to cover running costs. However, BGI is not the only institute sequencing genes in China, so the country’s overall annual equipment expenditure is likely to be in excess of $1 billion. “There are other scientists in China involved in gene sequencing,” says Jun Wang. “There are another three genome centers and more labs with [sequencing] instruments.” Although the initial cost of the Illumina sequencers may be high, technical innovations “take the cost of sequencing a human genome below $10,000,” says Illumina’s Jay Flatley, noting that just three years ago, sequencing a human genome cost $1 million, whereas the new HiSeq 2000 sequencing system could “recreate the International Human Genome Project in a week.” According to the manufacturer, the HiSeq 2000 is capable of generating 200 gigabases of data per sequencing run and 25 gigabases of data per day. It uses two flow cells and an innovative dual-surface imaging method, enabling new levels of sequencing output as well as experimental flexibility. “With the rapid reduction in the costs associated with [gene] sequencing, this approach will soon be cost-effective enough to be used on a routine clinical basis,” Lo predicts. JF
researchers to obtain 30-fold coverage of two human genomes in a single run. “Our goal is to build partnerships and collaborations around the world that contribute to our global society,” says Zhang. “Creating solutions that enhance agriculture and food production, for example, are a key focus for us, [as are] more region-specific programs, such as the development of the personal genomics field in China.” “[BGI’s investment] aligns well with some of (China’s) strategies in a number of areas, like reducing medical costs and improving plant- and animal-based industry,” says Charles Cantor, chief scientific officer of San Diego– based Sequenom. The company’s technology is commonly used to follow up whole-genome studies to confirm newly discovered alleles and phenotypic associations, and Cantor expects China to be an important market. China has been “very successful” in efforts to develop genome analysis software, he adds. BGI has also been establishing its own technical platforms based on large-scale genome sequencing, efficient bioinformatics analysis and innovative genetic healthcare initiatives, with these achievements having contributed considerably to the development of genomics in China and elsewhere. BGI’s Wang, who is also a professor in the Department of Biology at the University of Copenhagen, notes that genomic sequencing platforms would also open the door for further collaborative projects worldwide, including the 1,000 plant and animal reference project, to which BGI has pledged $100 million; the 10,000 microbial genomes project; and the
190
Chinese Cancer Genome project. BGI has already established itself as an important genome center on the world map. It successfully sequenced 1% of the human genome for the International Human Genome Project; contributed 10% of the International Human HapMap Project, with which CUHK was also involved; played a key role in the SinoBritish Chicken Genome Project; completely sequenced the rice and silkworm genomes; and completed the first diploid human genome sequence of an Asian. As reference genomes for certain organisms are now relatively common, “different strains are now being sequenced to discover the variations related to certain traits in those species,” says Wang, noting that “large-scale studies in exomics, metagenomics, epigenomics and transcriptomics” have suddenly all become realistic propositions. In terms of applications, Lo’s group at CUHK has pioneered the use of next-generation
DNA sequencing as a molecular diagnostics tool, using noninvasive prenatal diagnosis as an example. CUHK researchers have also been involved in viral and bacterial genomic sequencing. For example, “my group was one of the first two Asian groups to report the complete sequencing of the SARS [severe acute respiratory syndrome] coronavirus and to use genomic information to research the molecular epidemiology and evolution of this virus,” says Lo. In China, “the challenges are getting more resources, [especially] funding, samples, and collaborative agreements from various groups that study different subjects,” says Wang. “We also need more young talent to work in this field, in particular to analyze the data.” Finally, Wang stresses that although the Illumina deal may have made BGI the largest sequencing center in the world, “genomics cannot be done alone and must be performed on an international basis.” One of the best examples concerns the 1000 Genomes Project, an international research effort to establish the most detailed catalog of human genetic variation by sequencing the genomes of at least 1,000 individuals from different ethnic backgrounds, in which BGI is collaborating with genome centers in Germany, the UK and the USA. The size of China’s population could be a boon to international efforts at unraveling how genome variations contribute to disease. Indeed, China has also made a commitment to cancer research. “They are doing work to describe such patients, and one can only expect it’s going to accelerate as they become more sophisticated and [Chinese people] get better and better healthcare,” says Richard Cotton, who is head of the Genomics Disorders Research Centre at the University of Melbourne in Australia and convenor of the Human Variome Project, which focuses on collecting and curating human genetic variations that affect health. Cotton’s efforts have focused on single-gene disorders, and he is excited about the potential for new research to emerge from China’s
New product approvals Victoza (liraglutide [rDNA origin] injection)
Novo Nordisk (Copenhagen)
The US Food and Drug Administration approved the new drug application for Victoza, the first once-daily human glucagon-like peptide-1 (GLP-1) analog for type 2 diabetes. Victoza is indicated as an adjunct to diet and exercise to improve blood sugar control in adults with type-2 diabetes mellitus.
Xiaflex (collagenase clostridium histolyticum)
BioSpecifics Technologies (Lynbrook, New York) and Auxilium Pharmaceuticals (Malvern, Pennsylvania)
The US Food and Drug Administration approved Xiaflex for adults with Dupuytren’s contracture with a palpable cord. Xiaflex consists of two microbial collagenases in a defined mass ratio, Collagenase AUX-1 and Collagenase AUX-II, which are isolated and purified from the fermentation of Clostridium histolyticum bacteria.
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
news bulked-up resources. “I can see a renaissance in single gene disorder work because the sequencers can be put in there to find something useful for patients. There’s no question in my mind that these sequencers will be put to work in the Chinese population,” adds Cotton. But that won’t necessarily be enough, he says. Healthcare infrastructure will also be critical. “They’ve got to (obtain) the material to put in them. Single gene disease research is not highthroughput. A patient comes in the door of a doctor’s office and then leaves, but that rate is not very high. If they get themselves organized in China, they could get a lot of samples coming in,” adds Cotton. Along with its continued strengthening of expertise in sequencing and analysis, “China is now producing more well-trained scientists than any other country,” says Cantor. As long as this trend continues, China’s impact on progress in research and technology is likely to continue to rise proportionally. “It is inevitable,” he adds. And in terms of both investment (Nature
463, 282, 2010) and scientific output, China’s genome centers are already beginning to rival some genome centers in Europe and North America, although today the emphasis is more on collaborative research between many international centers than on competition. China’s efforts could well spur development elsewhere. “I think there is a tremendous opportunity for any country or funding agency to really empower that discovery by making some investment. I assume that China understands that. They have a good group at BGI, which has a pretty good track record of making this kind of thing work,” says Richard Wilson, director of The Genome Center at the Washington University School of Medicine, St. Louis, Missouri. “If countries such as the US, or the UK, or anyone else sees that as a good reason to build up their own genetic sequencing infrastructure, I think that’s a good thing. The more the better.” John Fox Hong Kong and Jim Kling Bellingham, Washington
in their words “Women would not even know they had [a] BRCA gene if it weren’t discovered under a system that incentivizes patents.” Defense attorney Brian Poissant pitches the importance of gene patents in motivating commercialization of discoveries at a hearing in the lawsuit on Myriad’s BRCA1 and BRCA2 claims opposed by the American Civil Liberties Union, a coalition of civil rights, research and women’s health groups. (GenomeWeb, 2 February 2010) “Over the last 30 years, we as a nation have spent $9 per American per year on cancer research. Enough to buy you a couple of lattes.” Francis Collins recalibrates the US public’s understanding of how much has been spent on the ‘war on cancer’ declared by President Nixon almost 40 years ago. (CBS Evening News, 28 Jan 2010) “His members think he gave away the farm for nothing. So he was really tossed because of a falling out with the board over miscalculating how to negotiate.” An unnamed industry source gives the inside view on Billy Tauzin’s decision to resign as chairman of PhRMA, which spent $26,150,520 in 2009 on lobbying, according to the nonpartisan Center for Responsive Politics. (ABC News, 12 February 2010) “I now believe it is time I move on and hand the mantle of leadership of this great organization
to others as passionate as myself and to explore the many other interests I would like to pursue in this special second-chance life that I have been given.” Billy Tauzin, cancer survivor and president of pharma industry trade group PhRMA resigns amid criticisms over the group’s involvement in proposed US healthcare reform. (ABC News, 12 February 2010) “There’s no doubt there’s risk in pharmaceuticals, and there should be! If you make 30% returns it should be a risky business. If you don’t want risk, go be a grocery store and make 6%.” Andrew Witty, GlaxoSmithKline’s CEO, comments on pharmaceutical R&D in the light of controversy over anti-aging compounds developed by their 2008 acquisition Sirtris. (Forbes, 25 January 2010) “There are physicians earning so much money [from drug makers] that they would give up their jobs…It’s a shocking story. Normally, you’d give up the [company] honoraria.” Steve Nissen, head of cardiovascular medicine at the Cleveland Clinic Foundation, comments on Lawrence DuBuske’s decision to leave Harvard rather than forgo payments from industry (which had totaled $99,375 for 40 talks). (The Boston Globe, 23 January 2010) “Arrays are a hundred times cheaper, a hundred times faster and materially more accurate than sequencing will be over the next couple of years.” Illumina’s Jay Flatley told delegates at the J.P. Morgan conference in San Francisco in January that arrays are likely to hold the upper hand for genomewide association studies over next-generation sequencers. (GenomeWeb, 15 January 2009)
nature biotechnology volume 28 number 3 march 2010
in brief RNAi delivery shop Silence Therapeutics of London and Intradigm Corporation of Palo Alto, California, have merged, in a deal designed to boost their competitiveness as providers of RNA interference (RNAi) delivery solutions. The two firms have developed separate technologies to enhance RNAi delivery and stability, currently the biggest challenge in RNAi-based therapeutics. By merging, the new company hopes to offer a set of systems to overcome these problems. Silence will contribute the AtuPlex delivery platform designed to stabilize siRNA within a liposome, whereas Intradigm’s system is a biodegradable, synthetic peptide–based polymer that allows any tissue in the body to be targeted by adding a ligand. The deal, which took place as a reverse merger and for which Silence issued close to 80 million shares to acquire Intradigm, has been valued at about £20 million ($32.6 million). According to Simos Simeonidis, an analyst at Rodman & Renshaw, the combined company (which retains the name Silence Therapeutics) may become a more attractive partner for big pharma than its predecessors were because it has more platforms to offer. But he doesn’t think the merger makes Silence any more competitive compared with market leaders Alnylam and Sirna because pharma can always partner with multiple biotechs, each offering different technologies. Besides, said Simeonidis, “no one has all the answers, not even Silence with their multiple delivery technologies.” Nazlie Latefi
Brazil boosts bioscience Brazil’s national economic and social development bank BNDES has signed an agreement to invest in a selection of innovative bioscience and infrastructure projects at the state-owned Oswaldo Cruz Foundation, part of the Brazilian Ministry of Health (Nat. Biotechnol. 27, 1063–1064, 2009). The Rio de Janeiro–headquartered foundation, also known as Fiocruz, is planning a range of R&D projects that would require an estimated R$1 billion (US$536 million). BNDES will cover part of these costs, and Fiocruz hopes to get the rest through partnerships with the private sector. Fiocruz has already received the first R$40 million ($21.4 million) installment, which is being used to finish the facilities of a new Center for Technological Development in Health (CDTS) and to fund several projects at the Immunobiological Technology Institute in Rio de Janeiro (known as Bio-Manguinhos). Among the schemes selected for funding is the production of recombinant epoetin alpha, recombinant human alpha-interferon and PEG-interferon. The recently launched Integrated Center for Prototypes, Biodrugs and Diagnostic Reagents in Rio de Janeiro will partner to develop new bacterial and viral vaccines. This year, Fiocruz hopes to start producing 50 million doses of recombinant human insulin per year, thanks to a technological exchange with the Ukrainian Indar Institute. Brazil now imports around 170 million doses of insulin a year. Ricardo Bonalume Neto
191
NEWS
© 2010 Nature America, Inc. All rights reserved.
Patent income tax slashed To make UK biopharma more attractive to potential investors, the UK government plans to cut taxes on revenue generated from researchrelated patents. As from April 2013, the socalled ‘patent box’ will charge a 10% corporate tax rate, rather than the usual 28%, on revenues stemming from products patented and manufactured in the UK. Industry players and analysts are enthusiastic. “We are completely behind the government’s announcement,” says Joseph Wildy, joint head of external relations for the London-based BioIndustry Association. Similar schemes that are already in place in other countries such as Belgium and Spain have already yielded encouraging results, he says: “We can only imagine it will work here, too.” The initial signs are positive; GlaxoSmithKline, a London-based drug firm, has said it will invest £500 million in the UK to capitalize on the new initiative. The government will disclose details of the patent box—such as what, exactly, companies must do to qualify for relief—and solicit feedback during a threemonth consultation with industry. It will then finalize the scheme in time for inclusion in next year’s Finance Bill. Although the future of the scheme depends on government support, and New Labour is up for re-election this year, Wildy says that the opposition party has provided encouraging signs that it, too, would back the scheme. Asher Mullard
Abbott hit with record fine A Texas jury has ordered Abbott Laboratories to pay $1.8 billion—the largest award in a patent infringement case to date. The case over breakthrough rheumatoid arthritis drug Humira pitted Centocor Ortho Biotech, of Horsham, Pennsylvania, and New York University (NYU) against Illinois-based drug giant Abbott Labs and its subsidiaries Abbott Bioresearch Center and Abbott Biotechnology. The dispute was over US Patent No. 7,070,775, issued to Centocor and NYU in July 2006, with Centocor as the exclusive licensee. At a jury trial in June 2009, the plaintiffs alleged that Abbott, in developing Humira (adalimumab) infringed their monoclonal antibody and antibody fragments patent. Worldwide sales of Humira topped $1.66 billion last year, making it by far Abbott’s biggest product. The jury deliberated for five hours before ruling for the plaintiffs, finding that Abbott had infringed four claims of the ‘775 patent. Abbott was found not guilty of willful infringement, however. In December, US district judge T. John Ward ordered Centocor and NYU to recover from Abbott $1,168,466,000 in lost profits and $504,128,000 in reasonable royalties and awarded an additional $175,641,661 in prejudgment interest. Abbott has appealed the judgment to the Federal Circuit Court of Appeals. “While there are many issues still to be addressed by the Federal Circuit on the appeal,” says Thomas Kowalski of Frommer Lawrence & Haug, “the case shows that biotech patents are being upheld as valid, enforceable and infringed and serious defenses are being carefully considered.” Michael Francisco
192
Resuscitated deCODE refocuses on diagnostics A group of US investors has acquired the assets of deCODE genetics, the Icelandic genomics company that filed for bankruptcy in November (Nat. Biotechnol. 28, 5–6, 2010). Saga Investments—formed by Polaris Ventures of Waltham, Massachusetts, and ARCH Venture Partners, in Seattle, Washington—has put up $13.9 million to bring deCODE through the bankruptcy process and recapitalize the firm. Going forward, deCODE expects to make deals around its genomics resources, which are built around a detailed genetic profiling of the Icelandic population and ancestral data. But unlike that of companies (including deCODE) set up in the 1990s offering genomics services, the focus of the new firm will now be on diagnostics, not drug discovery. deCODE’s cofounder Kári Stefánsson will continue to lead the research team, which will remain in Reykjavik. Business operations will now be run out of Boston, Massachusetts and led by Earl Collier Jr., a former strategist (exec VP) for Cambridge, Massachusetts–based Genzyme, who takes over the CEO reins from Stefánsson. Despite its financial difficulties, deCODE has a clear track record of scientific productivity, says Collier. “We’ve been deepening our abilities and broadening them at a time when the field itself is moving forward.” But although its genetic research flourished, the company found its products in development difficult to market and service contracts difficult to execute. “The last conversations deCODE was able to have in a serious way about commercial things were almost two years ago,” Collier says. Now, with the reorganization, “the commercial aspects can be put back in balance with the scientific assets,” he says. deCODE expects to benefit from increasing awareness that genetic stratification of patients can improve the likelihood of clinical success as well as patient management and treatment. The long-term nature of genomics-based drug development was one reason the partnerships cut by the first generation of genomics companies—Millennium Pharmaceuticals (now wholly owned by Takeda of Osaka, Japan), Human Genome Sciences (Rockville, Maryland) and Incyte Pharmaceuticals (Wilmington, Delaware), for example, as well as deCODE—were unsuccessful. “This is the model for the second decade” of genomics, says Terry McGuire of Polaris Ventures, one of the venture capital firms that formed Saga. deCODE expects to ink partnerships in which disease-related genetic discoveries will be applied to risk assessment and pharmacogenomics tests. “It’s fair to say we will be more focused in terms of commercialization of intellectual property in the diagnostics direction, not so much in the drug development direction,” says Collier. As part of the reorganization, deCODE shelved three drugs in development, which are now owned by Saga. It had already shuttered its Emerald Biosciences and Emerald Biostructures drug discovery operations in Bainbridge Island, Washington, and its structural chemistry site in Woodridge, Illinois, in 2009, prior to the bankruptcy filing. Mark Ratner Cambridge, Massachusetts
AFP PHOTO/Leon Neal
in brief
Iceland not only has natural beauty but also a population with detailed genealogical information and medical records that has enabled deCODE to uncover genetic causes of common diseases.
volume 28 number 3 march 2010 nature biotechnology
n e ws f e at u r e
© 2010 Nature America, Inc. All rights reserved.
One year in—Obama’s biotech scorecard
standalone legislation, Waxman, with White House backing, might prove more determined to shorten that 12-year period, a provision that he did not fight while it was bundled into the healthcare reform package. President Obama “should leave it alone,” says Greenwood of BIO. As healthcare reform falters, the biotech industry awaits the fate of The Waxman approach is “not a way of movbiosimilars and tax credits. Jeff Fox reports. ing [biosimilars] reform forward,” according to Peter Pitts, president of the Center for Medicine in the Public Interest (CMPI) in New York. More Although the Obama administration’s first year merly of the Durham, North Carolina venture practically, Waxman “doesn’t have the votes of in office was consumed with healthcare reform firm Pappas Ventures and now based in New his colleagues.” Nonetheless, even though the and with kick-starting the faltering economy, York: “Some of the draconian provisions that FDA is “working to develop a regulatory pathsome of its earliest decisions affecting science were scary to pharma and biotech went by the way for biosimilars, the agency is still waiting for legislative authority,” he adds. policy bode well for the biotech community and wayside in the healthcare reform bills,” he says. Meanwhile, the biotech industry could also for research in general. Among those decisions Moreover, the mere fact that both political lose ground if the tax are the lifting of the ban on federal funding of parties and both the credits, known under human embryonic stem (ES) cells, the recon- Senate and the House of the rubric “Therapeutic figuring of former president Bush’s Council on Representatives reached Tax Credit,” were to go Bioethics and the appointment of biotech-savvy agreement on biosimidown with the pendofficials to key posts, including Steven Chu at the lars was important, ing healthcare reform Department of Energy and Roger Beachy at the says Jim Greenwood, bills, Greenwood says. National Institute of Food and Agriculture. president of the “If Congress pivots However, following a special election in Biotechnology Industry and turns its attentions Massachusetts that replaced the late Senator Organization (BIO) in to a jobs bill, we will Edward Kennedy, a champion of healthcare Washington, DC. If the try to get the discovery reform, with the moderate Republican Scott pending bills were to Brown, who opposes it, the fates of key provi- pass, those provisions It’s lonely at the top. President Obama’s failure to tax credit into that; it belongs there because sions in the healthcare bills dealing with follow- would provide “a path- push healthcare reform through Congress leaves on, or biosimilar, therapeutics and of sizable way for biosimilars that biotech issues unresolved. (Source: White House) it would help to keep companies alive.” That tax credits to biotech companies for discovery is balanced, while also drug research are now up in the air. At the same protecting innovators,” adds Michael Werner, a credit would make companies eligible for fedtime, the future of several potentially onerous partner with Washington, DC–based law firm eral funds to offset R&D costs, and thus would ‘sunshine’ provisions stipulating that physician Holland & Knight. “Early in 2009, we weren’t sure be particularly helpful for companies without researchers disclose the sources of their research how this would play out, but Congress arrived products and revenue streams—which is true support, honoraria, consulting fees and royalties at a balance. So patients will benefit because it of the majority of biotechs. “Certainly, we were also appears less clear. Amid that uncertainty, contains standards for safety and yet drugs will poised to celebrate this significant accomplishment. We still have a shot to get it done because the administration’s plans to emphasize com- ultimately get into the market.” parative effectiveness research, which could The provisions for biosimilars are “rea- it has broad support. And it would be sad to have affect whether federal programs pay for many sonably good,” agrees Gregory Conko of the it evaporate,” he says. Yet another part of the massive healthcare of biotech’s high-end therapeutic products, also Competitive Enterprise Institute (CEI) in may be thwarted. Washington, DC. They not only mandate 12 reform legislation—known as “Physician Another sign of shifting political currents years of “exclusivity,” thus protecting the intel- Payment Sunshine Provisions”—now seems was the surprise announcement in February lectual property (IP) investments of innova- unlikely to move forward, according to that Billy Tauzin, CEO of the Pharmaceutical tor companies, he says, but also lay out an Washington, DC–based Thomas Sullivan, the Research and Manufacturers of America “approval pathway. The [pending] legislation founder of the website Policy and Medicine and (PhRMA) in Washington, would be stepping also calls for clinical results but doesn’t spec- president of the medical education company down this June. Tauzin says that “it is time I ify how much testing to do, giving HHS [the Rockpointe in Columbia, Maryland. Although move on,” and that he plans to “explore…other Department of Health and Human Services] the House and Senate versions vary, both sought interests”—much of which sounds like code for that authority and allowing FDA [the Food to require detailed reporting by physicians doing an involuntary departure. Quick to put PhRMA’s and Drug Administration] to waive some clin- research and consulting, for payments as small considerable weight behind healthcare reform, ical test procedures. It’s not a perfect approval as $10, and would have imposed steep fines for failing to disclose such work. his announced departure could mean that the path, but it’s generally pretty good.” pharmaceutical industry will seek another With healthcare legislation stalled, the biostance on reform better suited to the altered similar provisions could be “pushed through A new FDA regime political landscape. on their own” as separate legislation, according Since her appointment as FDA commissioner to Conko. However, neither the White House last May (Box 2), Margaret Hamburg has already Healthcare reform fallout nor Representative Henry Waxman (D-CA), garnered some positive feedback. As Alan Even if healthcare reform legislation collapses who chairs the House Energy and Commerce Goldhammer, deputy vice president of regula(Box 1, Fig. 1), there are lessons to be drawn from Committee, is happy with the 12-year exclusiv- tory affairs at PhRMA in Washington, DC, puts the negotiations, suggests Arthur Klausner, for- ity period. If biosimilars were to be handled in it, “I’m not ready to say the new management
nature biotechnology volume 28 number 3 march 2010
193
NE W S f e at u r e
Box 1 Key provisions of proposed health insurance legislation With the passage of comprehensive healthcare reform now looking increasingly unlikely, the final legislation that achieves bipartisan consensus will probably look very different from the bills that were before the House and Senate at the beginning of the year (see Fig. 1). Some of the key provisions in that legislation included the following: • 30 million of the USA’s 40 million uninsured would gain coverage • Elimination of exclusions for pre-existing conditions • Employers not providing health insurance could be fined • Insurance exchanges would be set up to provide competition among commercial insurers for covering individuals and small businesses
© 2010 Nature America, Inc. All rights reserved.
• ‘Cadillac’ health plans (those in which premiums exceed $23,000 for a family) would be taxed (Senate version only)
is good, bad or indifferent, but they’re making good progress.” Thomas Murray, president of the Hastings Center in Garrison, New York, is more effusive: “She’s extraordinarily bright, focused, young and energetic, yet shows maturity and gravitas.” Even BIO is warming to Hamburg: “We are comfortable with the new leadership at FDA…Hamburg is bright, understands the industry and is not hostile to optimizing FDA,” says Greenwood. “FDA is standing up for consumers rather than being [only] a strictly regulatory agency,” says Pitts of CMPI. “Hamburg stands up for the culture and beliefs of the staff, and she also stood up to Senator [Byron] Dorgan [D-ND] on drug importation.” Dorgan sought to revive interest in importing drugs to lower their prices, a move that Hamburg resisted, arguing in December that doing so raised “safety concerns.” If stats are anything to go by, the FDA still has room for improvement; last year, the agency approved a little more than two dozen new medical entities—only slightly more than the historic average. But PhRMA’s Goldhammer is generally upbeat: “The bigger issue is that the FDA is better staffed and financial support for the agency is solid,” he says. (In the latest budget proposal, the agency would receive a 23% increase (Box 3).) Despite increased resources, the FDA is “still struggling to develop the biotech tools for doing twenty-first century regulatory science,” Pitts says. “And the ‘Critical Path Initiative’ [a national strategy for driving innovation in drug discovery and development] is not being properly funded. There are still a whole lot of problems from being understaffed and underfunded. Also, the research department is not robust for looking at twenty-first century molecules, taking away from the efficiency and raising the cost of FDA reviews.” One concern during the past year was the higher-than-average number of last-minute glitches due to safety issues that caused delays in reviews of new therapeutics. Furthermore,
194
according to statistics compiled by San Diego, California–based consultant Sagient Research, the agency missed Prescription Drug User Fee Act (PDUFA) dates on more than three times as many products in 2008 and 2009 as it did in 2006 and 2007, going from 4% to 14%. “Ideally, we want a system for discussing safety issues earlier during development, not when the license application comes in. Such last-minute delays are counterproductive,” says Goldhammer. Another concern is how the FDA will handle risk evaluation and mitigation strategies (REMS) for those therapeutic products that raise safety questions but nonetheless are being licensed and marketed. Also as yet undetermined is how the FDA and other stakeholders will adopt social electronic media tools, such as Twitter, for communicating the risks and benefits in using particular therapeutic products, including those subject to REMS. Food and fuels Although some FDA watchers, including several members of Congress, suggested early last year that its food regulatory responsibilities might be pulled out and consolidated in a separate agency, there seems to be little appetite for doing so at the moment, according to Conko of CEI. “The
June 20, 2009. PhRMA agrees to spend $80 billion on improving drug benefits for seniors on Medicare in return for expanded market August 7, 2009. On her Facebook page, Sarah Palin claims Obama administration planning ‘death panels’ in legislation September 16, 2009. Senate Finance Committee Chairman Max Baucus releases his healthcare proposal October 13, 2009. Senate Finance Committee approves Joint Healthcare Bill 14-9 December 8, 2009. Senate Democrats drop public option from healthcare proposal January 19, 2010. Democrats lose 60-vote majority needed to pass legislation after Scott Brown wins Ted Kennedy’s Senate seat in Massachusetts
Administration’s plate looks pretty full for the next couple of years, and the food safety legislation already working its way through Congress will likely take away some of the impetus for splitting the FDA in two,” he says. Nonetheless, the FDA is taking food safety issues seriously. In January, the agency appointed Michael Taylor deputy commissioner for foods, a new position that accords higher prominence to safety and will entail closer coordination of food-related programs throughout the agency. Earlier in his career, Taylor worked on food safety at both the FDA and the US Department of Agriculture (USDA; Washington, DC), and he also spent time in the agbiotech industry working for Monsanto of St. Louis. Meanwhile, several weeks ago, USDA secretary Tom Vilsack named Elisabeth Hagen as undersecretary for food safety. Hagen has been chief medical officer at the USDA and before that was a senior executive for the Food Safety Inspection Service. Although the focus is not food safety, a comprehensive review of how genetically modified (GM) crops are evaluated is under way at the USDA. In a separate development, the US Supreme Court in January agreed to hear ‘Monsanto v. Geertson Seed Farms,’ which deals with GM alfalfa. Earlier, the federal district court required the USDA to undertake an environmental impact statement on this same crop, marking the first such analysis for any GM crop. Biofuels is another area where there is bipartisan support. DOE secretary Chu in January announced $80 million for biofuels research projects under the American Recovery and Reinvestment Act, much of it going to study algae-based and other biofuels. More than half, $44 million, will go to the National Alliance for Advanced Biofuels and Bioproducts, which is led by the Donald Danforth Plant Science Center in St. Louis. Another nearly $34 million is awarded to the National Advanced Biofuels Consortium, directed by the National February 24, 2009. President Obama announces plan for comprehensive healthcare reform in joint session of Congress July–August 2009. Obama and leading Democrats receive increasingly critical reception for healthcare reforms at Town Hall meetings August 15, 2009. Obama suggests public option could be dropped from healthcare reform October 12, 2009. US insurance industry releases report predicting $4,000 increase in yearly premium if proposed legislation in Senate enacted November 7, 2009 Health Care Reform Bill passes the House of Representatives December 24, 2009. Senate votes 60–39 in favor of adopting the $871 million healthcare reform measure, opening way for reconciliation of bills in final legislation
Figure 1 Healthcare refom timeline. Source: Timelines.com
volume 28 number 3 march 2010 nature biotechnology
n e ws f e at u r e
© 2010 Nature America, Inc. All rights reserved.
Renewable Energy Laboratory and Pacific Northwest National Laboratory. “We’re still pressing for federal funding for biorefineries,” says Greenwood of BIO. However, the administration budget request seeks funding for both USDA and DOE programs supporting biofuels R&D as well as $500 million in loan guarantees at DOE in this field, he points out. “If the US is to make a serious move, particularly [toward] biofuel for motor vehicles, we have to go down this road and will need federal subsidies to prove its viability.” Positive movement At the National Institute of Health (NIH; Bethesda, Maryland) last December, director Francis Collins announced the approval of the first 13 human ES cell lines. These lines were approved under guidelines for ES cell research that were adopted last July, bringing the number of approved lines up to 42, with 90 more “pending.” In early February, one of the most widely used lines from the “Bush registry” (H-1) was added to the approved list. This is the first and only Bush-era line to be approved, and as such is a cause of consternation to researchers who have based their research programs on yet-tobe-approved lines. Collins explains, “In accordance with the guidelines, these stem cell lines were derived from embryos that were donated under ethically sound informed consent processes,” he says. “More lines are under review now, and we anticipate continuing to expand this list of responsibly derived lines eligible for NIH funding.” Last November, President Obama established a new Presidential Commission for the Study of Bioethical Issues, appointing Amy Gutmann, who is president of the University of Pennsylvania, as its chair and James W. Wagner, president of Emory University, as vice chair. The commission agenda is uncertain, but administration sources count among the issues needing to be addressed the creation of stem cells by novel means, intellectual property (IP) issues regarding gene sequencing, biomarkers and other screening tests, protections for humans who participate as subjects in research projects, scientific integrity and conflicts of interest involving researchers. “I have heard it [the commission] will be more pragmatically focused on solving problems than was the [President George W.] Bush Council on Bioethics,” says Murray of the Hastings Center. One “welcome twist” is that Gutmann’s expertise as a political scientist is in “democratic deliberation,” he adds. “I have high hopes that the commission will do a more sophisticated job at being a deliberative body and will use electronic tools to make that a reality.” Working within an
Box 2 Key appointments affecting science policy Many of the appointments made by the Obama administration have received positive reception from the research community. Francis Collins. Confirmed as NIH director last August, Collins is the former director of the National Center for Human Genome Research (1993–2008), a scientist and best-selling author of The Language of God: A Scientist Presents Evidence for Belief (Free Press, 2006). Margaret Hamburg. Confirmed as FDA commissioner last May, Hamburg comes to the position from the Nuclear Threat Institute, where she had served as senior scientist since 2001. Hamburg may be best known for innovative programs she developed in the 1990s to control the spread of AIDS and tuberculosis while serving as New York City’s commissioner of health (1992–1997). Kathleen Sebelius. Confirmed as Secretary of Health and Human Services (HHS) last April, Sebelius is a former governor of Kansas. As a past insurance commissioner and politician, she is known for her advocacy for consumers and for repeated (failed) attempts to institute healthcare reform in her home state. Tom Vilsack. Confirmed as Secretary of Agriculture in January 2009, Vilsack is former governor of Iowa. In 2001, Vilsack was named governor of the year by BIO for “his support of the industry’s economic growth and agricultural biotechnology research.” Although the appointment was met with derision from environmentalists and ‘slow food’ advocates, some say the secretary has shown himself to be a “staunch supporter” of both the farmer and the environment, his position in favor of corn-based biofuel notwithstanding. Steven Chu. Confirmed as Secretary of Energy in January 2009, Chu is a former head of the Lawrence Berkeley National Laboratory in Berkeley, California, a physicist and a Nobel laureate. He is also an advocate of alternative energy, including nuclear energy. Roger Beachy. Appointed first director of the USDA National Institute of Food and Agriculture in September, Beachy comes most recently from the Donald Danforth Plant Science Center in St. Louis, MO. He has carried out pioneering research on engineering viral resistance in plants.
administration that is especially adept at harnessing the Internet, for instance, the commission may experiment with using electronic social networking tools to involve a broader range of people in addressing issues on its agenda, he says. No announcements about who will be sitting on the commission panel. Looming threats Several patent policy questions that received little attention last year will soon come clamoring forth, according to Robert Cook-Deegan, director of the Duke Institute for Genome Sciences and Policy in Durham, North Carolina. For instance in February the Secretary’s Advisory Committee on Genetics, Health, and Society, which advises HHS secretary Kathleen Sebelius, released a revised draft of its “Report on Gene Patents and Licensing Practices and Their Impact on Patient Access to Genetic Tests,” which includes IP-unfriendly recommendations easing access to genetic diagnostic testing. Meanwhile, in February a hearing before US district judge Robert W. Sweet in New York considered a lawsuit brought by the American Civil Liberties Union and others against Myriad Genetics of Salt Lake City, Utah. It, too, seeks
nature biotechnology volume 28 number 3 march 2010
to ease patient access to genetic tests—in this case, tests that are a principal source of income for Myriad—that evaluate mutations associated with an individual’s risk for developing breast cancer. The outcome of that court case “has implications across the board that could affect other uses of gene patents,” Cook-Deegan says. This legal dispute “probably won’t be over in 2010, and it will go as high as the courts will accept. It’s a high-stakes event,” he says. On the legislative side, although the main action is in the Senate, patent reform is no longer on a fast track, according to David Forman, a patent attorney with the law firm Finnegan Henderson, Farabow, Garrett & Dunner in Washington. However, he says, “Proponents are still working on it, and there is no word that it has been completely pushed aside.” How it might affect the biotech industry also remains uncertain. Other high-stakes events affecting healthcare are gaining momentum. Last year substantial federal funds were allocated for comparative effectiveness research through the stimulus package, with a new proposal to spend $286 million in the 2011 fiscal year’s budget. Subsequently, healthcare reform legislation incorporated a
195
NE W S f e at u r e Box 3 The Presidential Budget Request for 2011 The overall federal civilian R&D budget for fiscal year 2011 is for $61.6 billion, which is $3.7 billion, or 6.4%, higher than the previous year’s. Several federal programs within this overall budget affect the biotech sector: • NIH: $1 billion increase, with a total of budget of $32.1 billion, including more than $6 billion for 30 cancer drug trials in 2011, double the number of novel compounds in phase 1–3 clinical trials by 2016 and support completion of a comprehensive catalog of cancer mutations for the 20 most common malignancies. • FDA: $2.5 billion budget plus $4 billion in user fees, an overall increase of 6% for expanding postmarket safety surveillance and making safety data more accessible. • HHS: more than $400 million to enhance the development of medical countermeasures, including those that target biological threat agents.
© 2010 Nature America, Inc. All rights reserved.
• DOE: $4.7 billion for clean energy technology, including $220 million for biofuels and biomass R&D; and $5.1 billion for the Office of Science, including $1.8 billion for basic energy sciences. • USDA: $54 million to promote agricultural production and biotech exports; $429 million, the highest funding level ever, for competitive research grants through the Agriculture and Food Research Initiative. • NSF: $7.4 billion, an 8% increase over the 2010 enacted level, to drive creation of industries and jobs by doubling funding for multidisciplinary research targeted at next-generation information and biological technologies. • US Patent and Trademark Office (Washington, DC): The office has been granted full access to its fee collections (previously some went into a general fund) as well as a 15% fee increase on patent services. • A federal tax credit of as much as $5,000 for small businesses that add workers, as well as a tax break for increasing salaries for those who earn less than $100,000 per year.
framework for drawing on the results of such research when deciding what medical products to pay for, according to Conko. “Adding to the body of such research is a good goal,” he says. “The concern going forward is whether costcutting bodies will take those results to establish a process for treating patients.” The use of comparative effectiveness research in this way would “create a legitimate concern because of the need to trim billions of dollars from Medicare outlays,” he continues. “If taxpayers are footing the bill, we want to be sure not to pay for ineffectual treatments. But it’s still a black box how commissions will work, and we may never know if a study in the hands of a cost-cutting body will inappropriately deny treatment options. We want to keep our eyes on this process.” CMPI’s Pitts says that the US is “behind Europe” in conducting comparativeness effectiveness research and that it is critical to use new resources properly: “It can’t be merely cost-effectiveness research but has to empower doctors to develop more effective treatments quicker.” Furthermore, although the stimulus legislation stipulated that the results cannot be used for making medical cost reimbursement decisions, US federal officials still seem to be “going in the wrong direction on this,” he adds. “It should not be about one drug being better than another but which is better for which patients.”
196
Bad marks for bioterror preparedness In December, the Obama Administration issued a new “National Strategy for Countering Biological Threats,” and it is continuing to work on “harmonizing” biological security policies and practices among federal entities, according to a senior official. Key elements of that strategy include promoting “global health security” and reinforcing “norms of safe and responsible conduct.” What’s more, during his speech before Congress in January, President Obama alluded to an “initiative that will give us the capacity to respond faster and more effectively to bioterrorism or an infectious disease—a plan that will counter threats at home and strengthen public health abroad.” In part, Obama was responding to a stinging report from the bipartisan Commission on the Prevention of Weapons of Mass Destruction Proliferation and Terrorism. “The United States is failing to address several urgent threats, especially bioterrorism,” said former senator Bob Graham, who chairs the commission, when the report was released in January. “Each of the last three administrations has been slow to recognize and respond to the biothreat.” Although the commission “gave the administration an ‘F’ for preparedness, Obama included funding for it in his budget request, and that’s a positive sign,” says BIO’s Greenwood. “Biotech companies look forward to participating in these efforts.”
Although administration plans are far from complete, they outline solid principles and put an important White House “stamp” on these issues, says Gerald Epstein, director of the Center for Science, Technology and Security Policy at the American Association for the Advancement of Science in Washington. Moreover, the December policy statement “addresses disease threats in the real world”—including those that arise from natural sources, something that the whole world faces—and does not restrict its scope to “deliberately delivered” threats. This new strategy is “not an appropriations bill,” and the federal government “still needs ways to engage companies” in these efforts, he adds. Brighter economic outlook While Congress recovers from its wearisome healthcare reform negotiations, signs of a recovering economy and reawakening risktaking on Wall Street are reviving optimism regarding the economic prospects for the biotech sector. “A lot of people thought 2009 would be a blood bath, and we were a lot more worried a year ago,” says Willy De Greef, secretary general of the European Association for Bioindustries (EuropaBio) in Brussels. “But a year later, it is more a puddle. Some biotech companies have gone under, and it was a tough year, with some bankruptcies. But the wholesale wiping out of companies that some feared hasn’t happened. The biotech sector held up rather well.” New York–based Klausner also points out that to some extent the biotech sector is insulated from the vagaries of the market compared with other sectors. “Because 95% of biotech companies don’t yet sell anything, the bad economy did not have much effect on them,” he says. And patients do not defer from filling their prescriptions because of economic climate; “for those [companies] that sell medicines, purchases can’t be put off.” Of course, companies that are not selling products also have “an insatiable appetite for money,” he continues. A year ago, investors were in no position to feed that appetite, but now they are showing “more confidence.” At the federal level, the last year was “very encouraging,” says Greenwood of BIO. “Work spent explaining biotechnology to Congress has borne fruit,” he adds, referring to the compromises worked out on biosimilars and tax credits. Of course, there are lingering frustrations, too, including stalled efforts, particularly in the Senate, to allow federal Small Business Innovation Research (SBIR) funding for venture capital–backed biotech companies. “We maybe could get that resolved in 2010,” he says. There is “hope.” Jeffrey L. Fox, Washington DC
volume 28 number 3 march 2010 nature biotechnology
ne w s feat u re
The lengthening handshake
Big deals Stealing the drug industry’s M&A show in 2009 was a trio of megadeals: Basel, Switzerland– based Roche sealed up its almost $47 billion buyout of Genentech, of South San Francisco (initiated the previous year); Pfizer, of New York, completed its $68 billion takeover of Madison, New Jersey–based Wyeth; and Merck of Whitehouse Station, New Jersey, paid $41 billion for Kenilworth, New Jersey–based Schering-Plough. By early 2010, plans were disclosed to fix overlaps and nail down synergies. Pfizer said it would send hundreds of Wyeth employees packing; Merck did the same with Schering redundancies; and all parties— under pressure from investors to make good on the transactions—were scrutinizing R&D for any overlapping capabilities or unnecessary personnel. Most pundits forecast at least a moderate rise in M&A activity, as pharma beefs up its flagging R&D capabilities. PricewaterhouseCoopers believes the year will feature strategic deals and “mergers of productivity,” but PwC partner Tracy Lefteroff says he never expected the takeover boom in 2009 that some claimed would be driven by biotechs’ cash desperation. “I’ve seen how difficult it is to kill these small biotechs,” he says. “They’ve always managed to leverage their assets to get the funding they need, even if it’s at a much slower and scaled-back pace.” Chris Dokomajilar, senior consultant at Deloitte Recap, concurs. “The fire sale was back in 2008,” he says. “Companies are either
holding off, or looking for other avenues. The M&A route is the last resort.” Tony Gibney, managing director of Leerink Swann in New York, finds at least one trend from 2009 that is heartening for biotech and will continue. M&A involving middle marketcap firms was “unprecedented,” with clinical development–stage companies snatched up despite the risk, and “you’re seeing more than just one party show up” at the table to bid, he says. New Brunswick, New Jersey–based Johnson & Johnson (J&J) bought Cougar Biotechnology, of Los Angeles, for $970 million. Paris-based Sanofi Aventis took over South San Francisco, California–based BiPar Sciences in a deal worth up to $500 million. Vertex Pharmaceuticals, of Cambridge, Massachusetts, took over Laval, Quebec–based ViroChem Pharma in a $377 million cash-andstock transaction. Gibney says 60–65% of the M&A deals have been of this kind, compared to around 20% previously. Pharma tended to pay impressive one-day premiums to share prices, but still paid prices near or below the 52-week trading prices, though deals above $400 million showed lower premiums because the mid-caps Sales by business area
Medium diversification (50–70% from primary care/ specialty drugs)
High diversification (<50% from primary care/ specialty drugs) Estimated 2009 sales split
The desire to find low-risk ways to generate revenue as patent expirations loom has driven M&A activity in the biotech and pharmaceutical sectors in recent years. But in this respect, 2009 was not a banner year. According to Walnut Creek, California–based consulting firm Deloitte Recap, only 35 pharma-to-biotech mergers worth more than $20 million took place; and for biotech-to-biotech takeovers, the number was only slightly higher—38. But more important than the M&A activity level are the reasons companies are buying and the clever ways in which buyers are minimizing risk and sellers are maximizing upside. What has emerged is a new kind of merger—structured deals involving takeovers that take place in stages or buyouts with milestones.
Estimated 2014 sales split
© 2010 Nature America, Inc. All rights reserved.
Although mergers and acquisitions (M&As) failed to hit the heights some analysts had predicted in 2009, a new type of tiered transaction rose to prominence—the structured deal. Randy Osborne reports.
had not been trounced as severely as others. And M&A deals have been more carefully structured, tying the ultimate payout to milestones of clinical and regulatory success (as in the Sanofi-BiPar transaction) so that the risk-reward ratio satisfies both sides (Table 1). Chadds Ford, Pennsylvania–based Endo Pharmaceuticals paid $370 million in cash for Indevus Pharmaceuticals, of Lexington, Massachusetts, with as much as $267 million more tied to milestones. Similarly, The Medicines Co., of Parsippany, New Jersey, staged its buyout of Cambridge, Massachusetts– based Targanta Therapeutics from $42 million to $138 million. As in many deals, Medicines paid a high premium (72%) to Targanta’s stock price, agreeing to a value that the lackadaisical market had not recognized. “You’ve seen more [milestone-dependent takeovers] historically with private companies, and that continues to be a big part of private deals,” Gibney says, but in the year ahead the trend will pertain more and more to acquisitions of public firms as well. “Those deals almost mimic M&A,” Gibney says, citing the deal by Basel, Switzerland–based Novartis with Wilmington, Delaware–based Incyte for the latter’s phase 3 JAK1/2 inhibitor against myelofibrosis and an earlier-stage compound in the cMET inhibitor class. The collaboration and licensing arrangement drew $150 million up front, plus $60 million as an immediate development milestone, with a total of more than $1 billion in payments possible over time. PTC Therapeutics, of South
100%
12%
22% 75%
64% 54%
50%
46%
37%
17%
12%
49%
51%
31%
16%
10%
13%
15%
20%
69%
69%
70%
PFE
GSK
MRK
18%
18% 11%
25%
34%
24%
25%
RHHBY
J&J
36%
57%
Low difersification >75% from primary care/ specialty drugs) 5%
7% 8%
16%
15% 5%
79%
80%
85%
LLY
BMS
AZN
0% ABT (1)
BAY (2) NVS (3) SNY (4)
100%
14%
23% 75% 50%
68%
56%
13%
0%
11% 18% RHHBY
21%
39% 39%
22%
22%
12% 21%
21%
15%
15%
59%
62%
63%
67%
GSK
PFE
BMS
MRK
6% 20%
16% 10%
15%
59%
25%
46%
20% J&J
20%
31%
34%
NVS (3)
ABT (1)
Primary care/specialty drugs
46%
46%
BAY (2) SNY (4)
Biologicals/vaccines
74%
LLY
74%
AZN
Other/nondrug
Figure 1 Company strategies focused on diversification. Roche, Novartis and Johnson & Johnson have made a strong case for diversification. RHHBY, Roche; J&J, Johnson & Johnson; ABT, Abbott; BAY, Bayer; NVS, Novartis; SNY, Sanofi-Aventis; PFE, Pfizer; GSK, GlaxoSmithKline; MRK, Merck; LLY, Lilly; BMS, Bristol-Meyers Squibb; AZN, AstraZeneca. Source: Bank of America/Merrill Lynch, New York.
nature biotechnology volume 28 number 3 march 2010
197
N E W S feat u re Table 1 Selected structured mergers and acquisitions Date
Acquirer
Target
Terms
January 13, 2009
Cephalon (Frazer, Pennsylvania) (Nasdaq: CEPH)
Ception Therapeutics (Malvern, Pennsylvania) (private)
An option, not an outright sale; $350 million potential deal; $100 million upfront to Ception for an option to buy the company
January 5, 2009
Endo Pharmaceuticals (Nasdaq: ENDP)
Indevus Pharmaceuticals $370 million cash ($4.50 per Indevus share) for Indevus; potential $267 million (Nasdaq: IDEV) cash ($3.00 per Indevus share) for regulatory and sales milestones
January 12, 2009
The Medicines Co. (Nasdaq: MDCO)
Targanta Therapeutics
Staged buyout from $42 million to $138 million
January 13, 2009
Cephalon (Nasdaq: CEPH)
Ception Therapeutics (Malvern, Pennsylvania) (private)
An option, not an outright sale; $350 million potential deal; $100 million upfront to Ception for an option to buy the company
Incyte (Nasdaq: INCY)
$150 million up front; $60 million immediate development milestone; possible $1 billion in payments over time
© 2010 Nature America, Inc. All rights reserved.
November 25, 2009
Novartis (NYSE: NVS)
Plainfield, New Jersey, garnered $100 million up front in a partnership valued as high as $437 million with Cambridge, Massachusetts–based Genzyme, to co-develop lead program PTC124, a small-molecule drug that targets genetic disorders caused by nonsense mutations, such as some cases of cystic fibrosis and Duchenne’s muscular dystrophy. “Structured deals are here to stay,” agrees Sherrill Neff, a partner with Philadelphiabased venture capital (VC) firm Quaker BioVentures. “It’s very effective where the large acquiring company doesn’t want to destroy the advantages [in top personnel] of the small company,” he says. “Sometimes, the last thing they want to do is lose all that management firepower.” Although such arrangements seem to favor the acquirer, Stephen Bloch, of the VC company Canaan Partners, says the “staged earn-out” model can work well for both sides. Canaan saw two exits from its portfolio travel that route last year. San Diego–based Calixa Therapeutics was bought for $92 million by Cubist Pharmaceuticals, of Lexington, Massachusetts, which agreed to pay up to $310 million more to Calixa shareholders if milestones are met. Canaan also had BiPar, grabbed by Sanofi. “In the old days, you saw a lot of deals with sales milestones, but those take too long. We’re seeing tangible, reasonably near-term clinical milestones. The ‘bio-dollars’ are possible for everybody to reach.” We saw an awful lot more [structured M&A in 2009], including even public-company situations,” says Glen Giovanetti, head of global technology for Ernst &Young (E&Y). “From an acquirer’s standpoint, it’s the way to go, but when there are lots of potential buyers, you’re not going to see it.” Although venture capitalists may be “increasingly comfortable” with such a setup, “it’s not a home run. They still
198
If option is exercised: • $250 million to acquire 100% of Ception • Ception stakeholders would receive clinical and regulatory milestone payments
If option is exercised: • $250 million to acquire 100% of Ception • Ception stakeholders would receive clinical and regulatory milestone payments
want some home runs, but this is a ‘bird-inthe-hand’ environment.” ‘Pivotal point’ reached Structured or not, M&As will stay near 2009’s levels, as pharma continues to be ultraselective, in Bloch’s view. “This is a debate we have all the time at my shop,” he says. “A lot of small companies out there are vulnerable without a capital market, but the reality is that the big [pharma] guys can only digest so much. If you’re investing [in a biotech], you have to be very careful that the company has enough leverage and white space around it to be attractive to pharma”—in other words, that the biotech’s pipeline fits as near to perfectly as possible with the pharma’s strategy. “We need to think about it from the science up, but also from the commercial down, and sometimes the ‘commercial down’ gets forgotten.” “I think we’re at a pivotal point’ in M&A, Neff says. “What’s different is that [pharma is] getting serious about cutting deep in their own organizations to make way for an increased spend on the acquired companies.” Previously, scientists in the pharma behemoths “didn’t like research that wasn’t their own— there was a big uphill battle every time,” Neff says, but the senior ranks lately have changed, and are “populated by people who really know what they’re doing.” He points to Jeremy Levin, senior vice president of strategic transactions at New York–based Bristol-Myers Squibb, who “comes out of our world and knows his way around biotech. He’s unafraid of finding good science within the walls of BMS or somewhere else.” Bank of America/Merrill Lynch of New York seems to agree, noting that low-diversity BMS, which pulled about 80% of revenues from small molecules—more specifically, primary care and specialty drugs—will reduce that number to 63% by 2014.
Pharma firms are becoming more practical about R&D, Neff says, “cutting deep in advance” of M&A so that they can dedicate money to the acquired firm for new experiments. “If you’ve got a $5 billion research budget allocated in the traditional way, and you have a guy [doing research] who knows ‘X’ billion is his for this year, it’s very difficult to convince that guy to give up on of his pet projects” after the takeover of an outside company, he says. “Sanofi, BMS, GlaxoSmithKline—they’re all very consciously going out there and saying, ‘We want to spend a much higher R&D dollar outside of these doors.’ And this year, they’re scrambling all over the place, looking for something to launch in 2012 or 2013.” Chip Gillooly, global vice president with the capital group at Durham, North Carolina– based consulting firm Quintiles Transnational, is less sanguine about would-be M&A deals overcoming the hurdles cited by Neff. “I haven’t seen [the idea of bringing new blood and its research aboard] succeeding phenomenally well,” he says. Uppermost management often welcomes newcomers, “and frankly, many of them are in their roles because their predecessors didn’t. So the barriers are coming down, but this is like an ocean liner. It takes a long time to turn.” Trying new ways to get R&D done as speedily as possible and with as few ruffled feathers is an ongoing effort, Gillooly says, and will influence the rate of M&A in the coming year. Pharma firms are paying much more heed to “how we can build virtual infrastructures so that we can accelerate and disband research as soon as it’s no longer relevant, and build another infrastructure as we need it,” he says. As an example of what Gillooly calls “the beautiful insanity of our industry, M&A talks evolve into [a situation where the pharma firm says], ‘I don’t want infrastructure, I want this core team of brilliant
volume 28 number 3 march 2010 nature biotechnology
ne w s feat u re
© 2010 Nature America, Inc. All rights reserved.
people, and I want these three or five assets.” Gillooly says three-way and four-way deals, short of M&A, will become more common. Diversify or die Leerink Swan’s Gibney finds “absolutely a trend toward diversification. In the early 1990s, pharma wanted to do ‘purification’, with med-tech spinouts and carve-outs of non-core assets,” he says. This push lasted into 2006, when Pfizer, of New York, made “the last of the purification plays” by selling its consumer health division to J&J. PwC’s Lefteroff notes that J&J is one of the few pharmas “that stuck with the old model—it’s what all of them used to be.” Now, the pendulum has begun to swing back. On the spectrum of diversity in revenue gainers, Roche sits at the top, with 54% of sales coming from biologics and vaccines (Fig. 1). London-based AstraZeneca is the least, with 88% of its income derived from small molecules. Bank of America/Merrill Lynch project that Roche’s biologics and vaccines share will rise to 59% by 2014, and AstraZeneca by then will be gaining 79% of its sales from small molecules, as the firm casts meanwhile for biotech buyouts to ease the imbalance. That’s potentially good news for biotechs, but Gibney says the still weak market in early 2010 also makes pharma lean toward reducing risk, especially given the uncertainty around US healthcare reform. Hence the entry by pharmas into over-the-counter drug sales, with Novartis’ exercise of its option to buy the remaining shares in Alcon, of Hünenberg, Switzerland, for $39 billion. Novartis had purchased 25% of Alcon in 2008, and decided at the start of this year to take the rest. At the end of 2009, Sanofi made known its plan to buy the over-the-counter drug specialist Chattem, of Chattanooga, Tennessee, for $1.9 billion. Such deals “don’t really impact [M&A chances for] biotechs all that much,” Lefteroff says. “These companies aren’t walking away from or defocusing their pharmaceutical pipelines. They’re looking for a way to steady their earnings.” Anyway, says E&Y’s Greene, the favorable glow that consumer-oriented companies seem to emit “may just be an artifact of the math—you may not be adding shareholder value, although [pharma’s purchase of such a firm] reduces exposure to the vagaries of R&D over time,” so their appeal may not persist. Reading tea leaves In general, “an increase, but not monumental” in M&A between pharma and biotech is what Gillooly expects during 2010. “There will be a fair amount of activity around asset acquisi-
tions, where they sort of acquire the company, but not all of it,” he says. The trend toward such pacts is gathering steam, and “there are more to be done,” Gibney says, although some smaller companies may be reluctant because such deals—by tying up much of the firm with a partner—“take away the [option for a full] M&A outcome for some period of time.” Paratek Pharmaceuticals, of Boston, was able to land such a deal with Novartis valued at up to $485 million to develop the late-stage antibiotic PTK 0796 for life-threatening infections, after Merck ended a licensing arrangement with Paratek for the candidate. Carmiel, Israel–based Protalix BioTherapeutics entered a profit-sharing deal with Pfizer for the phase 3 Gaucher’s disease drug Uplyso (taliglucerase alfa), even though Protalix’s stock slid on the news because investors had hoped Pfizer would buy the whole company. If there’s anything to hold down biotech M&A in 2010, it might be the improving capital market. “There’s less of a need to do the deals, and there’s the perpetual confidence of a smaller company to get to the next milestone,” Gibney says. Interest remains strong in the “sleeper” mid-cap category, too—those companies valued between $1 billion and $5 billion, whose worth has stayed level over the past 12 months. It creates some interesting conversations between such firms and wouldbe partners or acquirers as they try to hash out terms. More threatening for pharma-to-biotech M&A may be the drift toward an emphasis on patient outcomes. The goal of coming up with a better result with a drug—rather than what appears in the laboratory to be a “better” drug—will skew the way deals are done and with whom, he says. “As the overall healthcare environment or ecosystem shifts toward patient outcomes, all the players need to refocus on what creates value,” Greene says, and the future may find pharma turning its eyes to companies devoted to services, devices, monitoring and information and away from innovative biotechs. Stephen Kaldor, former executive at Takeda San Diego, led the Japanese firm’s $270 million buyout of Syrrx, of La Jolla, California, in 2005. He has maintained a keen interest in the M&A space ever since. Like others, he forecasts “at least a moderate uptick for privately and publicly held biotechs. If you look at pharmapharma [deals], there will be some of those as well, but most of that has already occurred.” For the past year and half or so, M&A has been largely product focused, Kaldor notes, as firms struggle to fill near-term revenue gaps brought about by patent expirations and
nature biotechnology volume 28 number 3 march 2010
flagging sales. “Some of that will continue— it’s kind of the baseline of activity,” he says. But Kaldor, now president and CEO of San Diego–based Ambrx, sees “an uptick in gaining access to capabilities, and novelty on top of capabilities.” “Pharma has finally placed the bet on biologics,” he notes. “A few have been late to come to the table, but they have pretty much all placed a stake in the ground.” Along with the push for access to platforms as well as products, Kaldor says, the twinned movements will continue toward outsourcing early-stage development and toward establishing business units for further research—“autonomous units looking to insert ‘plug and play’ capabilities.” (The latter model was pioneered by J&J, with units such as Centocor, Janssen Pharmaceutica and OrthoMcNeil Pharmaceutical.) E&Y’s Giovanetti doubts the draw of platforms will be that powerful for pharma in M&A, “unless it’s such a novel technology that they feel they have to control it” entirely through ownership. Instead, the pharma firms will go for collaborations that give access. Negotiations between pharma and product companies often start as talks about alliance, and end in takeovers, he says, but this is harder for a platform-based firm to pull off. Smaller, product-based biotech companies may have a harder time getting the usual type of M&A done. “We’ve got a broken model there,” according to Gillooly. “Still very little capital is available, and the structure, more often than not, that these companies assume is to build a ‘mini-pharma,’ in hopes they can control the levers associated with getting a product to market.” Such industries as finance, aviation and insurance “realized a few decades ago that a successful company cannot be master of all, but has to create networks and partnerships that allow them to be nimble and responsive.” Kaldor, though, predicts that biotechs in 2010 will have more leverage in conventional M&A, helped by an improving economic climate at the start of the year, and that more M&A will occur. “People are, without being silly, looking at initial public offerings as a preferred track,” he says. In January, ten filed IPOs waited in the queue, and hopes were bolstered by Cambridge, Massachusetts–based Ironwood Pharmaceuticals’ decision to raise its price range from $172.5 million to a level that could go as high as $267.2 million. Although Ironwood disappointed Wall Street by pricing 16.7 million shares at $11.25 each for $188 million, well below the range’s top end, the linedup IPOs bode well for biotech. The industry “was trying to fake things a year ago,” Kaldor says. “Now it’s credible again.” Randy Osborne, Atlanta
199
building a business
Seeking the biotech eBay Nuala Moran
© 2010 Nature America, Inc. All rights reserved.
Internet exchanges suggest an easy route to sourcing and licensing technology, but can biotech intellectual property be packaged up and sold in this way?
T
here are sites for buying and selling most things on the internet, and intellectual property (IP) is no exception. The earliest of these IP exchanges sprang up at the height of the dot-com boom, when there was a rush to set up electronic marketplaces to ease business-to-business transactions of all sorts. Whereas selling paper clips and office stationery is one thing, it remains unclear—even a decade after the first IP sites were launched— if biotech patents and knowledge can be packaged, bought and sold in this way. In the past few years, there has been a renewed effort to make this option practical. After all, startups, collaborations and partnerships are the engine of the biotech industry, and each involve IP transactions of some sort. At the same time, there is now an imperative to translate publicly funded research into economic growth, which has increased pressure on the technology transfer offices of universities and other research institutions to patent any resulting discoveries and find new ways to market them. The first IP sites were set up by large corporations in such sectors as electronics, software and chemicals to allow companies to trade discoveries from their in-house research departments that they did not intend to exploit themselves. But other IP exchange sites have since been established by universities as a new avenue for technology transfer. Many of these sites, in both Europe and the United States, are financed by public or charitable funding in support of innovation policies. Given that such a high proportion of public funding goes to life sciences, most of these sites carry biotech-related IP. How do IP exchanges work? There are several different approaches to showcasing IP on the internet (Table 1).
Nuala Moran is a freelance writer based in Cheshire, UK.
200
Typically, technology transfer offices list technology available for licensing on their own website. But an increasing number of sites aggregate offerings from several institutions. Such sites basically operate as dating agencies—they are places to find technology and to express your interest to its owner. Some are free to use, whereas others charge subscription fees. The level of sophistication in terms of uploading and searching information varies from site to site, as does the level of customer support. There are also third-party sites (particularly in the United States) that go one step further and try to help foster the actual transaction, allowing entrepreneurs to not only find intellectual property but also buy the rights online. Some of these sites will also provide brokers to act as intermediaries. Biotech IP comes in many shapes and sizes, and some types are more amenable than others to being sold online. These are generally tools of some sort, and include materials, biomarkers, animal models and bioinformatics software. For commoditized products such as these, standard licenses can be filled in and processed online. However, most biotech IP is far more complex to use, apply and commercialize, and it cannot be packaged neatly in this cookie-cutter way. One recent trend has been the consolidation of different IP exchange sites, as demonstrated by the recent agreement between B-Bridge Technology Transfer (Mountain View, California), which has mainly American and Japanese users, and the European site Innoget (Barcelona, Spain). Last October, the two said they will provide full access to each other’s websites. The demand from users for one-stop shopping makes further consolidation likely. Gauging effectiveness The idea of an eBay for IP, where it would be possible to locate goods and buy them, seems
attractive. But those with experience in the biotech industry doubt whether life science IP can be parcelled up and sold like old baseball cards or a bicycle. Glyn Edwards, CEO of Antisoma (London), says his company “might use [IP exchanges] to look for stuff ” but admits it is “not a perfect market.” The company has spent more than a decade building up its network of personal contacts and combing the world’s publicly funded research bodies for IP to in-license. Edwards says this particular skill is where a lot of his company’s value resides, adding that Antisoma is “well-networked, and we know the people.” Edwards believes Antisoma gets access to research findings at an early stage through this networking. Tom Hockaday, managing director of Isis Innovation, the technology transfer arm of Oxford University, agrees that the exchanges work best as a shop window. Although Isis does list technology for sale, it uses only exchange services that are free. “It lets people know you are out there as a source of opportunities,” says Hockaday. “But if we’re required to pay, we won’t do it. That’s a commercial decision; we exist to market [Oxford University’s] technology—that’s what we are good at, so we don’t pay anyone else to do it.” Hockaday adds that his organization sells early stage technology, which means it “needs work to get it to market.” That points to a problem: in general, when a technology is in-licensed, tons of data need to be assimilated and know-how acquired to make the technology successful. That ancillary material and knowledge is not guaranteed when in-licensed through an exchange site. Besides, Hockaday says, “we want to know the development plans of people taking on our technologies.” In biotech, there is a long path from the university lab to the market. Typically, Isis and its counterparts don’t get
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
building a business paid for technology—instead, they take an equity stake or are entitled to development milestones. In situations like that, it’s important to know how the licensee intends to take a project forward and that there is funding to do so. Like Hockaday, others want a personal relationship with the partners they are trading IP with. On the sell side, this means knowing there is a credible development plan, and on the buy side it means knowing that the expertise required to make the technology function will be on hand. Another issue is negotiations—what is paid and when, including up-front payments, clinical milestones, research costs, royalties on sales and so forth. This generally cannot be conducted over exchange services. Most sites are only as sophisticated as having a buy or don’t buy option. Additionally, even with a platform technology or a drug target, there are many different ways of slicing up rights—dividing them by geography and markets, by different indications or by usage (say, as a research tool or a diagnostic). This involves complex negotia-
tions. For example, if you are taking rights in one region, it is important to know who has the license in another. If you want to develop a product in one disease area now, and you are successful, will you have the freedom to move into other indications in the future? IP exchanges offer a one-size-fits-all approach that, as yet, cannot handle these subtleties. Justine Lalonde, an executive at F. Hoffmann La-Roche’s Pharma Partnering Group (Basel, Switzerland), says her company doesn’t use IP exchanges, though they do use a subscription drug discovery and development database to check out the compounds of potential partners or assess the competition in particular areas. A business development and partnering specialist at another large pharmaceutical company gave a similar response. Although pharmaceutical companies are more and more reliant on partnering with biotechs to fill their pipelines, IP exchanges are not seen as the place to go courting. Another factor is time, according to biotech executive and investor Danny Green, former CEO of the startup BioCeramic Therapeutics (London), who remains a director of the com-
pany. “The CEO of a young startup doesn’t have the time to go searching on exchanges,” he says. “You should look to a specialist— it’s important to stick to your priorities and not to spend time thinking about something when other people out there are experts.” Green is not alone in reaching this conclusion. Five other biotechs asked for their views on the value of IP exchanges in partnering say they do not use them. The reservations about the value of these sites are shared across the industry, from technology transfer offices and startups to more mature biotech and pharmaceutical companies. Everyone can understand the appeal of an eBay for biotech, but as yet no single exchange provides that critical mass. And biotech IP is not old junk from the garage—in general, it cannot be photographed. Instead of being a tangible object, IP usually consists of immense data files that require insight and know-how if anything is to be extracted from them or any value built around them. There is also a general disdain over what type of material gets posted. Given the protective nature of the biotech industry, some
Table 1 Selected websites offering IP that include life-science technologies URL
Description
http://www.b-BridgeTechnologies.com/
This site provides a searchable database. More information can be obtained by posting messages to IP owners on a private bulletin board. The site, which launched in April 2009, lists technology from US and Japanese universities.
http://www.ibridgenetwork.org/
This site lists 10,820 technologies from 104 organizations, mainly US universities. There is an online licensing facility that provides the ability to enter into licenses with research labs directly from the site. Although the not-for-profit site, sponsored by the Kauffman Foundation (Washington, DC), covers all technologies, it has a high number of biomedical, biotech and drug discovery listings.
http://www.innoget.com/
This site advertises itself as a portal for open innovation. As well as listing and requesting technologies, users can also pose challenges or problems for which they are seeking answers. Users pay an annual fee for listing and requesting technologies, but posing a challenge is free. The Spanish site offers a telephone support line.
http://www.switt.ch/
This free-to-use site run by the Swiss Technology Transfer Association (Basel, Switzerland) lists technology for hire from universities across Switzerland.
http://www.university-technology.com/
This site carries IP listings from 13 Scottish universities across all fields of technology. Expressions of interest are passed on to the relevant institution, but for some technologies it is possible to fill out a standard license form online.
http://www.tynax.com/
This full-service trading exchange charges no fees but takes a commission on completed deals. It features more than 10,000 patents and technologies and brokers transactions via a network of agents and its own staff at its headquarters in Silicon Valley.
http://www.knowledgeexpress.com/
This site is run by the information services company UTEK (Tampa, Florida) and provides access to a range of technology listings and patent databases. The subscription fee ranges from $7,950 for a single user to $100,000 for a global license. It incorporates another former exchange site, TechEx.
http://www.TechTransferonline.com/
This site lays claim to the world’s longest list of IP, with 95,000 technologies available for sale or license. Although there are no fees to list IP, there is an annual fee of $250 to view listings. Technology owners are alerted every time their technology is viewed and can contact the viewer regardless of whether the viewer expressed an interest.
http://www.theintellectualproperty.net/
This site is run by Manchester University Intellectual Property and is funded by the Northwest Development Agency with backing from corporate sponsors. The site, which is free to use, went live in December 2009 and has listings from 20 UK universities. Currently, it is operating on a modest scale, with 400 technologies on offer and 5 requests for technologies listed.
http://www.yet2.com/
This site was set up by corporate heavyweights including Dow Chemical (Midland, Michigan), DuPont (Wilmington, Delaware), Monsanto (St. Louis) and 3M (St. Paul, Missouri) at the height of the dot-com boom in 1999. It claims to represent over 42% of the world’s R&D capacity. The site has more than 100 registered users.
http://www.flintbox.com/
This site was set up in 2003 by UBC Research Enterprise, the technology commercialization arm of the University of British Columbia. Most of its listings and users are from North America, though it does have some in Denmark. The site offers three grades of membership at fees of C$600 (US$568), C$2,400 (US$2,272) and C$5,000 (US$4,733). In November 2009, the website was acquired by Wellspring (Pittsburgh), a supplier of software that enables technology transfer offices to manage and market IP. As a result, starting in May 2010, it will be free for users to post technology on the site, and Wellspring is also promising to incorporate new features.
nature biotechnology volume 28 number 3 march 2010
201
building a business
© 2010 Nature America, Inc. All rights reserved.
suggest that the IP listed on these websites as technology for hire has been turned down by industry leaders—in other words, advertising technology is a desperate move. As BioCeramic’s Green puts it, the material posted has already been “hawked around.” Looking forward Perhaps IP exchanges will build enough critical mass to replace existing channels. Chris Haley, marketing services manager of Imperial Innovations (London), has assessed 15 or so IP websites. Imperial Innovations is interested in such exchanges from the perspective of both out-licensing technology from its parent institution, Imperial College London, and building up the IP foundations of its spin-out companies. He says there are “more and more appearing, and my overall view is they potentially could be good [routes for out-licensing].” In particular, he pointed to the London Technology Network, which carries listings of technology available for licensing from institutions in the city. The site is attractive because it is publicly funded through London’s regional development agency and is free to use. The IP exchanges do have another potential use: as an avenue for startups to explore or locate related technology. One patent alone is not sufficiently attractive to gain funding, and using exchanges to pool technologies could strengthen the portfolio and allow a startup to make a more compelling case to investors. However, even in that preferred route, exchanges might be helpful but not essential. It’s likely that IP exchanges could shape up
202
to be a passive channel, enabling companies to off-load technology they do not have the resources to develop in-house. But perhaps the most important aspect of IP exchanges thus far is the way they have opened up the market for small and medium enterprises (SMEs), says Mark Thompson, head of market development at the University of Manchester Intellectual Property. “The big companies have people to [scout for interesting technology]. SMEs don’t have the time, or much idea of how, to access technology coming out of universities.” In fact, after an 18-month pilot involving 30 universities and more than 100 companies, University of Manchester Intellectual Property launched last December what it claims is the first free-to-all technology trading portal. “This is the first system available which is suitable for SMEs to use to either market their technology or seek out innovations. It’s as easy to use as sites such as eBay but is of course free,” says Thompson. He’s encouraged that SMEs are starting to use the site, and he is planning more marketing to attract them. Exchange sites do have the benefit of being discreet. There are times when a bioentrepreneur needs to add value by acquiring rights to a particular piece of IP property but, for competitive reasons, does not want to advertise that fact. IP exchanges offer a way of quietly accessing a technology. This can be done without letting the seller know
the reasons for your interest or the potential value of the IP. And it can allow companies to make a pre-emptive strike, acquiring rights to a technology to block a competitor. Conclusions There is widespread appreciation of the potential of buying and selling IP on the internet. Those who have tested these sites are generally sellers rather than buyers. But there is a feeling on both sides that biotech IP is not amenable to being packaged and sold. Despite these reservations, no one dismisses IP exchange sites out of hand: it’s clear they provide a way for users to locate technologies and get a sense of what is actually out there. In fact, most think they will become more useful over time, given some consolidation and a greater number of users. If the websites are to work, most feel they would be most useful for buying and selling some categories of technology, such as materials, culture media, mouse models of diseases or biomarkers—areas in which the product is relatively simple, the rights are not exclusive and one size fits all. Although networking and personal contacts are at the heart of in-licensing, IP exchanges should be an increasingly important tool. “I would definitely keep an eye on these exchanges,” says Imperial Innovation’s Haley. “There’s no obvious eBay out there yet, but if someone could crack it, that would enhance their value.”
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 3 march 2010 nature biotechnology
correspondence
© 2010 Nature America, Inc. All rights reserved.
Oversulfated chondroitin sulfate is not the sole contaminant in heparin To the Editor: Contaminated heparin was associated with at least 149 deaths in 2007 and 2008, according to the information published at the US Food and Drug Administration (FDA) website (http://www.fda.gov/ Drugs/DrugSafety/PostmarketDrug SafetyInformationforPatientsandProviders/ ucm112669.htm). Sasisekharan and his colleagues1–3 have analyzed 6 lots of heparin associated with the contamination event and 28 lots of heparin produced by a heparin manufacturer from 2004 to 2007 (ref. 2). Using NMR analysis, they report that the contaminants in heparin include an impurity, specifically dermatan sulfate, and a contaminant, oversulfated chondroitin sulfate (OSCS) that is presumed to be derived from animal cartilage1. Work in our laboratory analyzing the same 28 heparin samples reveals the presence of other contaminants in heparin—some oversulfated and some undersulfated—all of which, like dermatan sulfate and OSCS, are found in, or can be made from, heparin by-product (a highly variable waste product comprising heparan sulfate, dermatan sulfate and chondroitin sulfate generated during heparin purification from crude heparin). With the exception of dermatan sulfate and OSCS, which can be definitively detected by NMR, our results suggest these additional contaminants cannot be distinguished from heparin in samples because of their similar NMR profiles2 and similar rates of migration under capillary electrophoresis and anion exchange highperformance liquid chromatography (HPLC). Taken together, our findings suggest the need for further work to unambiguously identify and characterize additional contaminants present in different batches of heparin. Heparin is the most highly sulfated naturally occurring glycosaminoglycan (GAG). It is enriched in porcine, ovine and bovine intestines or bovine lung entrails along with less sulfated GAGs, including
heparan sulfate, dermatan sulfate and chondroitin sulfate. These GAGs are made by all animal cells and are present in all tissues4. Pharmaceutical-grade heparin is prepared from crude heparin by removal of the less sulfated GAGs, the so-called heparin by-product, from heparin. As a result of its high degree of sulfation, heparin has higher anticoagulation activities than its less sulfated heparin by-product. Indeed, danaparoid (also referred to as Orgaran), an anticoagulant derived from porcine mucosa heparin byproduct containing 70–80% heparan sulfate and 20–30% chondroitin sulfate and/or dermatan sulfate5, was removed from the US market in 2002 due to its unsatisfactory anticoagulation activities and difficulties in obtaining heparin by-product that had a constant heparan sulfate to chondroitin sulfate/dermatan sulfate ratio5,6. In this context, we examined whether the OSCS observed by Sasisekharan1–3 in contaminated heparin might be derived by oversulfation of heparin by-product, rather than from animal cartilage chondroitin sulfate. Similarly, we sought to determine whether heparan sulfate, dermatan sulfate, chondroitin sulfate and oversulfated forms of these GAGs were also present in adulterated heparin samples. As heparin contains no, or vanishingly low, levels of galactosamine, we commenced our analysis by quantifying the glucosamine/ galactosamine ratio in each of the heparin lots (derived as previously reported3) (Supplementary Methods). This assay7,8 has passed a double-blind test designed by the FDA with a set of heparin and galactosamine-containing GAG mixtures. We found that the 28 heparin samples contained 0–37% galactosamine-containing GAGs, with an average value of 14.8% (Fig. 1). The most contaminated heparin lot (2007-29) contained 37% galactosamine, consistent with the 39% galactosamine found by a recently published assay9. We also analyzed 28 ‘uncontaminated’ heparin lots produced by different heparin manufacturers
nature biotechnology volume 28 number 3 march 2010
in the United States and found that 20 out of 28 heparin lots had no galactosaminecontaining GAGs (Fig. 1, far right). Finally, we confirmed our findings using a different assay in which the samples were digested with a mixture of heparin lyases I, II and III. These enzymes degrade heparin and heparan sulfate to component disaccharides generating a chromophore at 232 nm. The adulterated preparations did not yield the expected optical density values compared with control, unadulterated heparin (Supplementary Fig. 1). Oversulfated GAGs can inhibit these enzymes, so the decrease in optical density could result from contaminating chondroitin sulfate, other impurities or oversulfated material. In contrast to our findings, Sasisekharan and colleagues3 previously reported these same 28 heparin lots to contain 0–27% (average, 6.4%) galactosamine-containing GAGs, including the impurity, dermatan sulfate and the previously identified contaminant, OSCS (Supplementary Table 1 in ref. 3). Their calculation of the amount of dermatan sulfate and OSCS in contaminated heparin was based on NMR analysis using two assumptions: first, each of the 15 protons in heparin, dermatan sulfate and OSCS measured by NMR produces the same signal intensity by integration; and second, dermatan sulfate has 100% iduronic acid and 0% glucuronic acid residues. However, we assert that these assumptions are questionable. First, the NMR proton signal intensity by integration is not equal for each proton of OSCS and heparin1,10 (Supplementary Fig. 2). Second, two proton shifts (A1 and U5) of OSCS are invisible in the contaminated heparins tested (Supplementary Fig. 2), indicating that either the contaminants were not OSCS or OSCS had different proton shifts when mixed with heparin or other unidentified contaminants. Indeed, in Figure 4 of their Nature Biotechnology paper1, the two-dimensional (2D) NMR profiles of isolated heparin contaminant reported by
203
40
Analysis of 28 suspect heparin lots
30
Analysis of 28 heparin lots from other heparin manufacturers
20
10
a b
© 2010 Nature America, Inc. All rights reserved.
H qu ex an os tif am ic in at e io n
H qu ex an os tif am ica in tio e n
0
P re ubl su ish lts ed
Galactosamine-containing GAGs (%)
c orre s p on d en c e
Figure 1 Galactosamine-containing GAGs in heparin. We analyzed 28 suspect heparin lots by NMR (Supplementary Table 1; ref. 3) and by hexosamine quantification. We also analyzed another set of 28 heparin lots from different heparin manufacturers not suspected of contamination. Of this last set of heparins, one lot (‘a’) with a high level (7.5%) of galactosamine-containing GAGs was from a heparin manufacturer whose heparins were reported in 2008 to be contaminated in Europe, and another lot (‘b’), previously obtained from Glycomed in 1992, contained 3.9% galactosamine-containing GAGs. Of the remaining unadulterated heparin lots, 6 contained <1.6% galactosamine-containing GAGs (in a range from 0.5–1.6%) and 20 had no detectable galactosamine-containing GAGs. The other 6 heparin lots contained 0.5–1.6% galactosamine-containing GAGs.
Sasisekharan and colleagues do not match those of cartilage OSCS, especially both proton and 13C shifts at position 5 of uronic acid residue that distinguishes glucuronic acid from iduronic acid. The content of iduronic acid in dermatan sulfate sourced from porcine mucosa changes based on subspecies, age, diet and the environment in which pigs are raised10,11. For example, two porcine mucosa heparin lots produced in 2000 were contaminated with chondroitin sulfate and dermatan sulfate, respectively, based on NMR analysis10. A dermatan sulfate containing 50% iduronic acid and 50% glucuronic acid would produce a 50% iduronic acid–related peak at 2.08 p.p.m. and a 50% glucuronic acid–related peak at 2.04 p.p.m. because only N-acetyl galactosamine next to iduronic acid produces the unique 2.08 p.p.m. peak12. However, the 2.04 p.p.m. peak contributed by glucuronic acid in chondroitin sulfate and heparin are indistinguishable and thus cannot be used to estimate the amount of contaminant10. This partially explains the discrepancy between our galactosamine quantification data and published results. Glucosamine/galactosamine assays do not distinguish heparan sulfate from heparin because both contain glucosamine. To determine if heparan sulfate was present in contaminated heparin, we systematically
204
analyzed the contaminated heparin lot (2006-49) where 3.2% dermatan sulfate was reported previously. We found that this heparin lot had 16% galactosaminecontaining GAGs (Supplementary Table 1). NMR and capillary electrophoresis analysis were performed on this lot along with heparin (Supplementary Fig. 3). We observed two elution peaks by capillary electrophoresis analysis. The first peak belonged to heparin and the second peak was a contaminant less sulfated than heparin. The two peaks had a ratio 11:9. Less than half of the second peak could be assigned to the reported 3.2% dermatan sulfate in this contaminated lot based on NMR analysis (3.2% × 6 (response factor for dermatan sulfate) = 19.2%). No heparan sulfate NMR proton signals were visible in the contaminated heparin lot compared to heparin, which was consistent with the published report3. We purified the contaminant by both anion exchange chromatography and ethanol precipitation. Both methods separate heparin from less sulfated GAGs. The purified low-sulfated contaminant behaved as a homogenous preparation eluting at 6.5 min by capillary electrophoresis analysis (Supplementary Fig. 3d,e). Very little material migrated like heparin (Supplementary Fig. 3b). Hexosamine quantification assays
showed that the low-sulfated contaminant purified by anion exchange HPLC contained 52% heparan sulfate and 48% chondroitin/ dermatan sulfate, whereas the low-sulfated contaminant purified by ethanol precipitation contained 56% heparan sulfate and 44% chondroitin/dermatan sulfate (Table 1). NMR analysis showed a signature that highly resembled porcine mucosa heparan sulfate proton signals and differed considerably from heparin (Supplementary Fig. 3). Bacterial GAG lyases, such as chondroitinases and heparin lyases, can also be used to distinguish heparin, heparan sulfate, chondroitin sulfate and dermatan sulfate. We first used a chondroitinase ABC assay (which measures degradation of both chondroitin sulfates and dermatan sulfate) to analyze the composition of the low-sulfated contaminant prepared by anion exchange chromatography or ethanol precipitation. Our analysis suggested that the contaminant comprised heparin-like material (56–58%) and chondroitin sulfate/ dermatan sulfate (42–44%; Table 1). The GAG composition values obtained by enzymatic digestion were similar to those obtained by hexosamine quantification. We next demonstrated that 44% of the ethanol-soluble low-sulfated contaminant was susceptible to digestion by chondroitinase ABC. In contrast, we found that 32% ethanolsoluble low-sulfated contaminant was susceptible to digestion by chondroitinase B, which digests chondroitin sulfate B (dermatan sulfate). These results indicate that the ethanol-soluble low-sulfated contaminant contained 73% dermatan sulfate (32% chondroitinase B digestible dermatan sulfate/ 44% chondroitinase ABC digestible GAG) and 27% chondroitin sulfate (Table 1). By using the same chondroitinase digestion scheme, we found that HPLC-purified low sulfated contaminant contained 95% (40/42) dermatan sulfate and 5% chondroitin sulfate (Table 1). Their dermatan sulfate compositions were similar to porcine mucosal dermatan sulfate that contained 87% (84/97) dermatan sulfate and 13% chondroitin sulfate (Table 1). Note that the heparin byproduct also contained substantial amounts of galactosamine-containing GAG, mostly dermatan sulfate (83% = 38/46) (Table 1). We also investigated the susceptibility of samples to a heparin lyase I, II and III assay, which distinguishes heparin from less-sulfated heparan sulfate. Both heparin and heparan sulfate are digested by heparin lyases I, II and III, whereas only low-sulfated and/or unsulfated domains characteristic of heparan sulfate are preferentially cleaved
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
c orre s p on d en c e by heparin lyase III—in our experiments, ~18% for normal heparin compared with 62% of a commercial preparation of porcine mucosal heparan sulfate (Table 1). Depending on the method of sample preparation, we found that as much as 48–57% of the heparin-like material in the low-sulfated contaminant was susceptible to heparin lyase III degradation (Table 1). Thus, four separate types of analyses of our samples suggest that the low-sulfated contaminant consists of substantial amounts of chondroitin sulfate, dermatan sulfate and heparan sulfate, which resembles heparin by-product. Importantly, neither heparan sulfate nor chondroitin sulfate in the contaminated heparin lot is detected by one dimensional (1D) NMR analysis. We reasoned that if NMR does not detect heparan sulfate in contaminated heparin, it might not be able to detect oversulfated heparan sulfate owing to their structural similarities. To test this possibility, we chemically sulfated porcine intestine mucosa heparan sulfate and studied it using 1D NMR analysis. Oversulfated heparan sulfate showed many novel proton peaks (Fig. 2a) compared with authentic heparin (Fig. 2c); however, when the oversulfated heparan sulfate was admixed with heparin at a ratio of 30:70, only the typical heparin-like proton 1D NMR profile was observed (Fig. 2). To make matters worse, we found that oversulfated heparan sulfate migrates with heparin by both capillary electrophoresis and anion exchange HPLC, making it impossible to use the latter methods to detect such a contaminant. To confirm these observations, we chemically sulfated heparin by-product (Fig. 2d, NMR profile of heparin by-product or GAG waste provided by FDA). Oversulfated heparin by-product (Fig. 2b) shows oversulfated heparan sulfate, dermatan sulfate and chondroitin sulfate proton peaks (Fig. 2d) compared with authentic heparin (Fig. 2c). Even so, when the oversulfated heparin by-product is admixed with heparin at a ratio of 1:1, a proton NMR profile that resembles contaminated heparin is observed, with peaks at 2.14 p.p.m. and 2.11 p.p.m. for OSCS and oversulfated dermatan sulfate, respectively (Fig. 2e). Again oversulfated heparan sulfate is invisible by 1D-NMR analysis in a 1:1 mixture of heparin and oversulfated heparin by-product. Multiple proton signals in the 2.14 p.p.m. and 2.00 p.p.m. region are visible in all six published contaminated heparin NMR profiles (Supplementary figures in ref. 1), indicating that dermatan sulfate and OSCS are not the only contaminants in heparin.
a
* * * * *
*
*
*
b
Impurity
c
d OSDS *
*
e
** *
*
* *
OSCS *
*
OSHS * *
f g
5.5
5.0
4.5
4.0
3.5
3.0
2.5
p.p.m.
2.14
2.08
p.p.m.
Figure 2 Proton chemical shifts of oversulfated heparan sulfate were eliminated in the presence of heparin. Proton NMR profiles: (a) Oversulfated heparan sulfate. (b) 30% oversulfated heparan sulfate + 70% heparin. (c) Heparin. (d) Contaminated heparin lot 2004-5, which had both oversulfated heparan sulfate and 5% galactosamine-containing GAGs that were undetectable by NMR. (e) Oversulfated heparin by-product. (f) A mixture of 50% oversulfated heparin by-product and 50% heparin. (g) Contaminated heparin lot 2007-26. OSHS, heparin oversulfated by chemical sulfation; OSDS, dermatan oversulfated by chemical sulfation; OSCS, chondroitin oversulfated by chemical sulfation; *, oversulfated heparan sulfate.
The relative signal strength for OSCS (2.14 p.p.m.) versus oversulfated dermatan sulfate (2.11 p.p.m.) might be a reflection of specific heparin by-products used for adulteration, which is demonstrated by the variation of OSCS levels but not overall galactosaminecontaining GAGs in 28 contaminated heparin lots (Supplementary Table 1). Indeed, the relative amount of heparan sulfate, chondroitin sulfate and dermatan sulfate in heparin by-product varies not only among subspecies of pigs11,13 but also among different animal tissues14. For example, crude heparin from camel intestine contains only heparin and heparan sulfate15, whereas chondroitin sulfate is enriched in crude heparin from bovine lung14. This heterogeneity complicates the analysis and evaluation of the deleterious side effects of each component of oversulfated heparin
nature biotechnology volume 28 number 3 march 2010
by-product. Clearly, further analysis would benefit from the use of 2D NMR methods to resolve these findings. By employing liquid chromatography/mass spectrometry analyses, we further confirmed the presence of chemically sulfated/desulfated GAGs and chemically oversulfated heparan sulfate in several batches of tested heparins16. As dermatan sulfate and OSCS detected by CE (Supplementary Table 1) and NMR3 could not explain most of the galactosaminecontaining GAGs found in the 28 heparin samples (Fig. 1), we performed detailed anion exchange HPLC and hexosaminequantification analysis of a contaminated heparin sample (Supplementary Fig. 4). We also performed CE and NMR analysis of purified heparin contaminants (Supplementary Fig. 5) without removing heparin/heparan sulfate–like materials by
205
c orre s p on d en c e able 1 Composition of GAG standards and low-sulfated contaminants purified from T contaminated heparin lot 2006-49 GAG sample
Hexosamine analysis
Enzymatic analyses
Glucosamine Heparan sulfate content (%)a and heparinb Normal heparinf Crude heparin
98 ± 2%g 83 ±
7%i
Heparan sulfatec
Chondroitin sulfate and dermatan sulfated
Dermatan sulfatee
100%h
18%h
0%h
ND
87%
27%
13%
ND
Heparin by-product
52 ± 0.2%
54%
37%
46%
38%
HPLC-purified LSC
52 ± 3%
58%
57%
42%
40%
Ethanol-soluble LSC
56 ± 0.1%
56%
48%
44%
32%
Porcine mucosal heparan sulfate
95 ± 2%
95%
62%
ND
ND
ND
ND
97%
84%
Porcine mucosal dermatan sulfate
3 ± 0.1%
LSC, low-sulfated contaminant; ND, not determined.
© 2010 Nature America, Inc. All rights reserved.
aPercentage
of total hexosamine (glucosamine plus galactosamine). bSusceptibility to heparin lyases I, II and III. cSusceptibility to heparin lyase III. dSusceptibility to chondroitinase ABC. eSusceptibility to chondroitinase B. fNormal heparin lots were defined by their nearly stoichiometric susceptibility to enzymatic depolymerization by heparin lyases, low galactosamine content and normal capillary electrophoresis and NMR profiles. gAverage value for 20 preparations of pure heparin, analyzed twice9. hSigma heparin (H-4784, lot no. 104K1177). iAverage value for three lots of crude heparin, analyzed twice.
low pH nitrous acid degradation. Overall, our data indicate that heparin had been contaminated with chemically modified heparin by-products since 2004. In a final set of assays, we investigated the capacity of each oversulfated form of chondroitin sulfate, dermatan sulfate and heparan sulfate to activate coagulation and the contact system (OSCS has previously been shown by Sasisekharan and colleagues to activate the contact system3). We found that chemical sulfation enhanced the anticoagulant activities of each component in the by-product as measured by a human plasma clotting assay (that is, the activated partial thromboplastin time assay, which measures the time necessary to induce clot formation in pooled normal human plasma). In the absence of GAGs, 0.5 ml of
plasma took 30.2 s to clot using the standard protocol (Table 2). Adding 0.6 µg of heparin prolonged the clotting time to 78.1 s, whereas adding 0.6 µg of OSCS, oversulfated dermatan sulfate or oversulfated heparan sulfate prolonged the clotting time up to 41.8 s. In contrast, tenfold more chondroitin sulfate, dermatan sulfate or heparan sulfate was needed to achieve comparable prolongation of clotting times. These findings confirm that chemical sulfation could turn each component of heparin byproduct into a much better anticoagulant, as reported previously17,18. Oversulfated heparin contaminants induce contact system activation by producing anaphylactic toxins, complement factors C3a, C5a and bradykinin3,19. As OSCS from cartilage has been suggested to be responsible
able 2 Enhanced anticoagulation and kallikrein activation activities by oversulfated T forms of heparan sulfate, dermatan sulfate and chondroitin sulfate Compounds
aPTT assay (seconds for clotting to occur)
Kallikrein activity OD405nm induced by 20 µg/ml GAG at 3 min
Kallikrein activity OD405nm induced by 20 µg/ml GAG at 40 min
Heparin
78.1 ± 1.4 (0.6 µg)
0.209 ± 0.008
0.275 ± 0.030
Heparan sulfate
36.2 ± 0.3 (6 µg)
0.033 ± 0.002
0.214 ± 0.042
Dermatan sulfate
38.5 ± 0.4 (6 µg)
0.067 ± 0.001
0.236 ± 0.042
Chondroitin sulfate
35.7 ± 0.2 (6 µg)
0.029 ± 0.001
0.409 ± 0.025
OSHS
40.9 ± 0.1 (0.6 µg)
0.253 ± 0.022
0.433 ± 0.020
OSDS
40.8. ± 0.1 (0.6 µg)
0.444 ± 0.021
0.521 ± 0.022
OSCS
41.8 ± 0.5 (0.6 µg)
0.521 ± 0.012
0.568 ± 0.007
No GAG control
30.2 ± 0.1
0.029 ± 0.005
0.345 ± 0.013
aPTT, activated partial thromboplastin time. OSHS, heparin oversulfated by chemical sulfation; OSDS, dermatan oversulfated by chemical sulfation; OSCS, chondroitin oversulfated by chemical sulfation. No GAG control, buffer added at the same volume used for GAG samples.
206
for anaphylactic reactions in patients3,19, we directly compared the anaphylactic effects of both natural and chemically sulfated heparan sulfate, chondroitin sulfate and dermatan sulfate with those of heparin (Table 2), using a heparin contaminant–induced contact system activation assay (that is, kallikrein activity assay19). We incubated human plasma with GAGs at 37 °C and measured the kinetics of kallikrein activity induction colorimetrically by kallikrein substrate cleavage at two time points. Because the plasma control exhibited a 7 min delay in plate surface–induced kallikrein generation in the presence of 2 mM CaCl2, we chose optical density (OD) readings at 3 min as a measure of the rate of kallikrein induction by GAGs and OD readings at 40 min as a measure of overall kallikrein induction by both GAGs and plate surface (Table 2). We found that heparan sulfate (OD, 0.033), dermatan sulfate (OD, 0.067) and chondroitin sulfate (OD, 0.029) did not induce kallikrein activities compared with the control (OD, 0.029) at 3 min. In contrast, heparin (OD, 0.209), oversulfated heparan sulfate (OD, 0.253), oversulfated dermatan sulfate (OD, 0.444) and OSCS (OD, 0.521) induce substantial kallikrein activities. At 40 min, all of the preparations have high kallikrein activities. Our data indicate that oversulfated forms of heparan sulfate, dermatan sulfate and chondroitin sulfate are potent contact system activators. Thus, each oversulfated heparin by-product component could pose a safety risk if added to heparin. In conclusion, our data suggest a likely source of heparin contamination is chemically oversulfated/modified heparin by-product, of which OSCS is an important component. Recent published reports indicate that there are two kinds of heparin by-products. One has the closest resemblance to heparin based on NMR analysis2 and the other could be enriched in chondroitin sulfate14, dermatan sulfate2 or heparan sulfate15, depending on the source of crude heparin. Therefore, the two types of heparin by-products provide a readily accessible source of material for adulteration compared with cartilage chondroitin sulfate, which has to be purchased, transported and processed. Further studies are needed to definitively confirm the presence of these heparin by-products, which have safety implications for patients20. Note: Supplementary information is available on the Nature Biotechnology website. ACKNOWLEDGMENTS The authors would like to thank H. Ye and J. Reepmeyer of the FDA for purification of heparin
volume 28 number 3 march 2010 nature biotechnology
c orre s p on d en c e contaminants and capillary electrophoresis analysis of heparin samples. The authors would like to thank A. d’Avignon of Washington University for the analysis of samples by NMR. The authors also thank J. Esko, E. Conrad, S. Kornfeld, J. Baenziger and D. Tollefsen for their suggestions, comments and critiques of the manuscript. L.Z. thanks E. Unanue, S. Santoro, J. Ladenson and N. Brown for their support in establishing a GAG structure/function laboratory at Washington University. L.Z. thanks J. Metz for critical reading of the original manuscript. This work is supported in part by US National Institutes of Health grant R01GM069968 to L.Z. and a St. Louis Children’s Discovery Institute Research Fund to L.Z.
OSCS 1H spectrum, pH 7
COMPETING INTERESTS STATEMENT The authors declare no competing financial interests.
Jing Pan1,2, Yi Qian1,2, Xiaodong Zhou1,2, Andrew Pazandak1, Sarah B Frazier1, Peter Weiser1, Hong Lu1 & Lijuan Zhang1
© 2010 Nature America, Inc. All rights reserved.
1The Department of Pathology and Immunology,
Washington University School of Medicine, St. Louis, Missouri, USA. 2These authors contributed equally to this work. e-mail:
[email protected] 1. Guerrini, M. et al. Nat. Biotechnol. 26, 669–675 (2008). 2. Guerrini, M. et al. Proc. Natl. Acad. Sci. USA 106, 16956–16961 (2009). 3. Kishimoto, T.K. et al. N. Engl. J. Med. 358, 2457– 2467 (2008). 4. Hovingh, P., Piepkorn, M. & Linker, A. Biochem. J. 237, 573–581 (1986). 5. Acostamadiedo, J.M., Iyer, U.G. & Owen, J. Expert Opin. Pharmacother. 1, 803–814 (2000). 6. Anonymous. Danaparoid.
(accessed 17 February 2010). 7. Frazier, S.B., Roodhouse, K., Hourcade, D.E. & Zhang, L. Open Glycosci. 1, 31–39 (2008). 8. Studelska, D.R., Giljum, K., McDowell, L.M. & Zhang, L. Glycobiology 16, 65–72 (2006). 9. Volpi, N., Maccari, F. & Linhardt, R.J. Anal. Biochem. 388, 140–145 (2009). 10. Zhang, Z. et al. J. Pharm. Sci. 98, 4017–4026 (2009). 11. Liu, H., Zhang, Z. & Linhardt, R.J. Nat. Prod. Rep. 26, 313–321 (2009). 12. Li, F. et al. Glycoconj. J. 25, 603–610 (2008). 13. Fareed, J. et al. Int. Angiol. 27, 457–461 (2008). 14. Volpi, N. J. Chromatogr. B Biomed. Appl. 685, 27–34 (1996). 15. Warda, M., Gouda, E.M., Toida, T., Chi, L. & Linhardt, R.J. Comp. Biochem. Physiol. C Toxicol. Pharmacol. 136, 357–365 (2003). 16. Pan, J. et al. Glyobiol. Insights 2, 1–12 (2010). 17. Maruyama, T., Toida, T., Imanari, T., Yu, G. & Linhardt, R.J. Carbohydr. Res. 306, 35–43 (1998). 18. Toida, T. et al. Int. J. Biol. Macromol. 26, 233–241 (1999). 19. Blossom, D.B. et al. N. Engl. J. Med. 359, 2674–2684 (2008). 20. Ahmad, S. Front. Biosci. 12, 3312–3320 (2007).
Marco Guerrini, Zachary Shriver, Annamaria Naggi, Benito Casu, Robert J Linhardt, Giangiacomo Torri & Ram Sasisekharan reply: Our paper in Nature Biotechnology reported the structural identification of a major contaminant in suspect heparin lots1. Notably, the structural assignment of oversulfated chondroitin sulfate (OSCS) was made by four independent laboratories (and has since been confirmed by others2,3).
OSCS 1H spectrum, pH 5
5.5
5.0
4.5
4.0
3.5
3.0
2.5
p.p.m.
Figure 1 Comparison of the proton NMR spectra of synthesized OSCS at pH 7 and pH 5. The chemical shift of the H5 proton shifts as a function of pH. GalNAc, N-acetylgalactosamine; GlcA, glucuronic acid.
In their conclusion and in agreement with our original findings, Pan et al. conclude that OSCS is an important contaminant in heparin. The identification of OSCS as a major contaminant within heparin has had a number of ramifications, including providing an underlying logic for the original screening methods developed by the FDA and since implemented into pharmacopeias. Widespread adoption of such methods as screening tools has helped secure the heparin supply and, most importantly, decreased to baseline adverse reactions associated with administration of heparin. To extend this analysis, we have systematically investigated factors that may influence the signatures associated with OSCS in the proton NMR spectrum. As demonstrated previously, the identity of the counter-ion4 influences the chemical shifts associated with OSCS. We also find that minor differences between the chemical shifts of the authentic standard and the isolated contaminant displayed in ref. 1, especially the C5 and the H5 of the uronic acid, are likely due to pH differences between the samples (Fig. 1). Such pH shifts are not unusual given that samples are typically exchanged and analyzed in unbuffered D2O. Regardless of minor chemical shift differences, inspection of Table 2 in ref. 1 indicates that the complete chemical shift assignment for OSCS made by two of the laboratories agrees very well with one another and with the original assignment made by Maruyama et al.5. As such, the structure of OSCS was correctly assigned in Guerrini et al.1 Additionally, we have completed studies to robustly define how changing
nature biotechnology volume 28 number 3 march 2010
the nature of the contaminant, either the monosaccharide identity, linkage or overall sulfation pattern, affects the ability of a variety of analytical techniques to detect a given contaminant. As one part of this investigation, we examined the effect of differential sulfation on the ability of various NMR techniques to detect oversulfated chondroitins. To this end, chondroitins with various degrees of sulfation (ds), ranging from 2.4 to 4.0, were prepared and analyzed by proton NMR. Whereas OSCS (ds = 4.0) is readily detected in the proton NMR in the 2.0–2.2 region, materials with lower ds values are not distinguishable within the same region (Fig. 2). Depending on the level of contamination, these materials would not be detected by proton NMR analysis, even upon inspection of the entire spectrum. However, multidimensional NMR is able to distinguish differentially sulfated chondroitin from heparin, owing to several structural signatures, including those of 2, 3-O-sulfoglucuronic acid as well as disulfated N-acetylgalactosamine (Fig. 3). Furthermore, multidimensional NMR can readily detect the presence of oversulfated heparin/heparan sulfate due to several characteristic signals, including 2, 3-O-sulphouronic acid (Fig. 4 and Supplementary Table 1). In conclusion, as the acetyl signal of partially sulfated heparin/heparin sulfate or chondroitin show similar proton chemical shifts to that of heparin, a proton NMR method might fail to detect such a contaminant if present in a heparin preparation. Indeed, we have scanned a wide variety of potential sulfated polysaccharide contaminants to
207
c orre s p on d en c e
a
© 2010 Nature America, Inc. All rights reserved.
5.4
5.2
5.0
4.8
4.6
4.4
4.2
4.0
3.8
3.6
3.4
p.p.m.
b
2.3
2.2
2.1
p.p.m.
Figure 2 NMR spectra of sulfated chondroitins. (a) NMR spectra of sulfated chondroitins with various degrees of sulfation of the disaccharide repeat. (b) Zoom of the N-acetyl region from 1.9–2.4 p.p.m. and comparison of the various chondroitin products with heparin.
address which are detectable by proton NMR and which are not6,7. Taken together, these results reinforce the importance of using orthogonal approaches, including multidimensional NMR (for example, heteronuclear single quantum coherence (HSQC) methods), such as were used in our original paper1 to ensure accurate identification of contaminants as well as ensure, to as great an extent as is practical, their absence in heparin. Within this framework, Pan et al. argue in their Correspondence that the major contaminant present in heparin could be oversulfated heparin by-products (or components thereof), presumably obtained through chemical sulfonation. At the outset, we find the authors’ claim that their data support the presence of other contaminants, beyond OSCS, in
208
suspect heparin lots analyzed in early 2008 unsubstantiated. We would note that Pan et al. alternatively use the phrase “NMR” in a way to mean either onedimensional (1D) proton NMR or the approach used by us1. In their letter, they present only 1D proton data, whereas we used both 1D and two-dimensional (2D) NMR experiments to definitively identify major impurities and contaminants in suspect heparin lots. Furthermore, with no experimental substantiation, they claim that, in the context of signals associated with ‘native’ heparin, contaminants are not always apparent upon inspection of the NMR spectrum and, accordingly, NMR is not useful as a quality-control test. However, in contradiction of this assertion, we and others have shown that multidimensional NMR can detect
and quantify contamination that might arise from sulfonation of numerous sidestream components, either alone or mixed with one another8–10. Indeed, other persulfonated polysaccharide components could serve a similar role to OSCS, and thus an appropriate testing strategy for heparin should account for not just OSCS but also other potential persulfonated polysaccharide derivatives. This theme is something that has been explored by us (either individually or in collaboration) since the initial structural elucidation of OSCS6,7,11. Purification of heparin is a multistep process12. Previous studies indicate that (1) the composition and identity of the waste at each purification step is different; and (2) different manufacturing processes can generate waste streams that are structurally and compositionally distinct from one another9,12. Pan et al. use these points to claim that heparin by-product may contain several major components, including heparan sulfate(HS) and/or chondroitin sulfate (CS) and/or dermatan sulfate (DS). Unfortunately, they present no quantitative analysis of the percentages of these components, but rather cite data from analysis of heparin purified from camel intestine and bovine lung (refs. 15 and 17 of Pan et al.). However, pharmaceutical heparin, used clinically, is derived from porcine intestine (or intestinal mucosa). As such, findings regarding products from species other than pig have no real relevance to the identity of sidestream products produced as part of the purification of pharmaceutical heparin. Furthermore, Pan et al. attempt to support their argument by stating that the content and sequence of component glycosaminoglycans present in porcine mucosa can change based on subspecies, age and environment, without providing supporting data. In addition, the references they cite (refs. 11, 12 and 14 within Pan et al.) do not support this point. In follow-up studies, completed after initial publication of the structure of OSCS, we demonstrate that for one porcine-based process, dermatan sulfate is a major impurity and sidestream product produced from crude heparin9. Conversely, very little CS is found as a waste product in the manufacture of heparin from porcine intestines. Persulfonation of this sidestream material, enriched in DS, would accordingly lead to primarily oversulfated DS, which would be detected in NMR analysis and is readily differentiated from OSCS (Figure S8 and Figure 3 of ref. 9).
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
c orre s p on d en c e Notably, the quality of the NMR spectra and results reported by Pan et al is poor (including, but not limited to, a noticeable lack of splitting of the H2 glucosamine signal and low resolution of glucosamine and iduronic acid anomeric signals). We, and others, have found that accurate quantification of proton signals is strongly related to the quality and resolution of the spectrum9,13. Proper magnetic field (≥500 MHz), sample concentration and solution pH value, together with the absence of multivalent cations are conditions that must be carefully controlled to ensure appropriate NMR measurements. For example, in Supplementary Figure 5b of Pan et al., there is significant line broadening of the H1 proton of 2-O-sulfo iduronic acid (5.20–5.30 p.p.m.), indicative of suboptimal sample preparation and/or analysis conditions. Significant and selective line broadening, such as is observed in Supplementary Figure 5 of Pan et al., is typically observed in samples contaminated by multivalent cations. Furthermore, the authors assert that proton signals of A-2, A-3 and A-4 (ring positions of N-sulfoglucosamine) are much higher than those of the rest of the protons. As shown in previous studies using HSQC analysis13, only the H2 of N-sulfoglucosamine is usually sufficiently resolved from the rest of protons to be quantified accurately in the 1D proton spectra. Conversely, the H3 and H4 protons severely overlap with other signals belonging to unsulfated uronic acid (H2 of iduronic acid and H3, H4 and H5 of glucosamine) and thus show an overall higher intensity in the corresponding 1D spectrum10,14. In addition, at several points in their Correspondence, Pan et al. state that different data sets, derived from Supplementary Table 1 in the present Correspondence, support one another and thus confirm their results, citing specific examples. However, looking at the data as a whole, there are some puzzling findings that raise important questions. For example, analysis of the data for the 32 samples that have both hexosamine and heparinase results, indicates that the heparinase data do not correlate with the galactosamine data (Supplementary Fig. 1). The data clearly form two clusters, on the basis of whether there is an early-migrating peak (ostensibly OSCS) in capillary electrophoresis; however, in neither cluster is there a significant correlation between increasing galactosamine content and decreased
a
b
p.p.m.
p.p.m.
92
55
94
60
96 98
65
100
70
102
75
104
80
106
85
108 110 5.8
5.6
5.4 5.2
5.0 4.8
4.6 4.4 p.p.m.
5.2 5.0 4.8 4.6 4.4 4.2 4.0 3.8 3.6 3.4 3.2 p.p.m.
Figure 3 Two-dimensional HSQC analysis of a 10% wt/wt mixture of sulfated chondroitin (ds = 3.1) in heparin compared to a reference standard of sulfated chondroitin. (a) The anomeric region. (b) Signals from outside the anomeric region. Clearly identifiable signals associated with 2,3-di-O-sulfoglucuronic acid (G 2S,3S) and sulfated N-acetylgalactosamine (GalNAc) are distinct from those of heparin. G2S, 2-O-sulfoglucuronic acid; G, glucuronic acid; black, heparin; red, reference standard of sulfated chondroitin.
susceptibility to heparinase digestion (r2 values of 0.098 and 0.014 for early and late migrating peaks, respectively). On the basis of this analysis, the data sets do not, in fact, support one another, nor do they support the actual presence of another major contaminant in heparin. The enzymatic studies presented by Pan et al. are confounded, as evidenced by the fact that in their Supplementary Figure 1 (and Supplementary Table 1 of Pan et al.), the authors use inhibition of heparinases as a readout of contamination/ impurity levels. Nevertheless, in Table 1 of their Correspondence, they use enzyme susceptibility to determine the percentage of heparin-like components and chondroitin/ dermatan sulfate–like components in partially purified impurities/contaminants (for example, so-called low-sulfated contaminants). How could they use such an enzymatic assay to determine the percentage of individual components if the enzymes used to benchmark such determinations are inhibited by the presence of other components? Indeed, enzymatic digestion is a valuable tool for both characterization and quality control of heparin material; however, the data presented by Pan et al. are not supportive of their conclusions. Finally, interwoven in their argument around heparin by-product, they also attempt to advance an argument that oversulfated heparan sulfate could be present in heparin. First, as mentioned above, persulfonated heparan sulfate, if present, would have characteristic structural signatures that would be readily detected,
nature biotechnology volume 28 number 3 march 2010
especially in a multidimensional NMR experiment. Indeed, a recent study by us demonstrated that, in contradiction to the assertion of Pan et al., a well-controlled proton NMR spectrum, or, preferably, a multidimensional NMR experiment, can readily distinguish oversulfated heparin/ heparan sulfate in the context of a mixture (respectively, Supplementary Fig. S10 and Fig. 5 of ref. 9). Second, there is also ambiguity in the definition of “heparan sulfate” as used by Pan et al. In the context of the biology, Pan et al. seem to refer to heparan sulfate as porcine intestinal heparan sulfate that has a degree of sulfation of ~0.3 sulfates per disaccharide15. However, in the context of the precipitation experiments, “heparan sulfate” likely refers to a different set of compounds; namely, undersulfated chains of heparin. Because heparin is a mixture of polysaccharide chains, which differ in terms of overall sulfation and sulfation pattern, isolated components of the mixture (for example, by molecular mass, sulfation density) have been shown to have different properties from the whole mixture12. Purification of OSCS from heparin is a rather difficult process, given the overall polydispersity of heparin. As such, one plausible explanation for the glucosamine/galactosamine content of the ‘purified’ contaminant presented by Pan et al. is that the contaminant is only partially purified and thus still contains chains of heparin with less sulfate density. This interpretation would also explain the data presented in Supplementary Figure 5 of their Correspondence. Here, the major
209
c orre s p on d en c e a
b
© 2010 Nature America, Inc. All rights reserved.
Figure 4 Two-dimensional HSQC analysis of oversulfated heparin waste compared to authentic heparin. (a) The HSQC spectrum excluding the anomeric region. Clearly observed signals from oversulfated heparin/heparan sulfate include disulfated uronic acid. (b) HSQC spectrum of the anomeric region.
resonances due to polysaccharides in spectrum B labeled “Purified Oversulfated Heparin Contaminant from Lot 2007-23” arise from just two sets of signals, one of which can clearly be assigned to heparin and the other of which can be assigned to OSCS. Pan et al. conclude their Correspondence stating, “This heterogeneity complicates the analysis and evaluation of…each component of oversulfated heparin
by-product. Clearly, further analysis would benefit from the use of 2D NMR methods to resolve these findings.” Such an approach is precisely the type of approach that was completed, first in 2008 (ref. 1) and later in 2009 (ref. 9). In these papers, multidimensional NMR, run in multiple laboratories, allowed us to identify or confirm signatures that arise from oversulfated glycosaminoglycans (GAGs), either in isolation or in the
presence of heparin. To extend our results, we also reanalyzed our HSQC data presented in ref. 1 to determine the monosaccharide composition of samples C1–C2 and S1–S6 based upon a published methodology (Table 1)13,16. Consistent with our original findings, quantification of the monosaccharide composition of the heparin-like component (which would include heparan sulfate or oversulfated heparin/heparan sulfate, if present) indicates no appreciable differences between samples C1–C2 and S1–S6. If, for example, heparan sulfate was a significant impurity, we would expect to see a higher level of N-acetylglucosamine in one or more of samples S1–S6 and a correspondingly lower level of overall sulfation, including 6-O-sulfoglucosamine and 2-O-sulfoiduronic acid. These signatures are not observed. Finally, in none of the samples are the signatures observed for 2,3-di-O-sulfouronic acid, a key signature for oversulfated heparin-like components. In addition, we also would like to comment on and clarify a set of statements made by the authors with regards to how
Table 1 Quantitative HSQC analysis of samples C1, C2 and S1–S6 Constituent (mole%) Glucosaminesa
Sample C1
C2
S1
S2
S3
S4
S5
S6
58.0
62.7
61.9
61.5
66.2
59.3
59.4
66.4
9.1
9.3
8.7
12.9
8.3
10.8
9.7
8.4
N-sulfoglucosamine–glucuronic acid disaccharide
10.7
6.9
8.9
9.7
6.3
10.0
8.4
8.4
N-acetylglucosamine
N-sulfoglucosamine–2-O-sulfoiduronic acid disaccharide N-sulfoglucosamine–iduronic acid disaccharide
14.6
13.6
14.6
11.8
13.1
12.3
14.2
10.6
3-O-sulfo, N-sulfoglucosamine
6.7
6.6
5.9
4.1
6.1
7.6
8.2
5.6
N-sulfoglucosamine at the reducing end of the GAG chain
1.0
0.8
0.0
0.0
0.0
0.0
0.0
0.6
77.0
76.0
84.8
87.2
82.1
83.0
84.0
81.4
6-O-sulfoglucosaminea Uronatesb 2-O-sulfoiduronic acid
73.6
66.1
73.2
71.1
70.7
67.0
68.8
71.7
Iduronic acid–6-O-sulfo, N-acetyl/sulfoglucosamine disaccharide
7.6
8.1
6.8
5.6
8.1
6.1
6.1
6.3
Iduronic acid–N-acetyl/sulfoglucosamine disaccharidec
2.1
1.4
4.2
3.7
5.2
4.9
2.3
2.7
Glucuronic acid–N-sulfoglucosamine disaccharide
8.5
11.2
8.6
7.7
7.6
8.3
8.2
8.3
Glucuronic acid–3-O-sulfo, N-sulfoglucosamine
2.7
4.5
2.7
3.2
1.6
3.4
3.6
2.1
Glucuronic acid–N-acetyl glucosamine disaccharide
5.4
5.9
4.4
5.2
6.6
7.4
6.3
7.2
Galacturonic acid
0.0
2.8
0.0
1.8
0.0
1.4
2.6
1.2
Epoxide
0.0
0.0
0.0
1.7
0.0
1.5
2.0
0.5
Linkage region
5.3
2.6
4.4
3.2
4.3
3.6
3.5
3.4
aNormalized
mole percentage of various glucosamine moieties. Measurement of total 6-O sulfation is completed separately and itself is normalized to 100%. Also, quantification of 6-O sulfoglucosamine in heavily contaminated samples is increased owing to the signal being proximate to an additional signal at 4.3/69.0 p.p.m. bNormalized mole percentage of various uronic acid moieties. Measurement of the linkage region is completed separately and is not included in the overall normalization. cQuantification of iduronic acid linked to N-sulfoglucosamine or N-acetylglucosamine in heavily contaminated samples is increased owing to the signal being close to another signal at 4.9/104.2 p.p.m.
210
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
c orre s p on d en c e they frame and/or interpret the data and conclusions from previous studies1,17. First, they repeatedly attribute to us a claim as to the source of OSCS, namely porcine cartilage. OSCS may very well originate from persulfonation of porcine cartilage; however, we have not asserted such, nor have we presented data that can be used to definitively assign source. In our set of analyses, persulfonation removes any information, such as composition, that could be obtained regarding the source of the chondroitin sulfate. Furthermore, they claim that in the study of Kishimoto et al.17, quantification of the amount of dermatan sulfate and OSCS requires two assumptions: first, each of the 15 protons in each of the disaccharide units of the polysaccharides produces signals of the same intensity; and second, dermatan sulfate consists of 100% iduronic acid and N-acetylgalactosamine residues. The calculation does not require making either of these assumptions. The first assumption, namely, equivalence in signal intensity, is most assuredly not the case, owing to signal overlap and sequence heterogeneity. Second, several studies have identified that dermatan sulfate has a major N-acetyl signal at ~2.08 p.p.m., arising from the N-acetyl galactosamine attached to iduronic acid18. Previously, it has been shown that porcine intestinal mucosa dermatan sulfate has >90% iduronic acid19, a value that agrees well with the 87% calculated by Pan et al. Furthermore, we conducted proton NMR analysis of dermatan sulfate to confirm this fact (Supplementary Fig. 2). Finally, the calculation presented in ref. 17 is well-founded and is similar to one previously used by Perlin and colleagues20 to calculate the amount of dermatan sulfate in a heparin preparation. We have confirmed the quantitative analysis of samples through spike and recovery experiments (Supplementary Fig. 3). In addition, it is important to note two points on this analysis. First, the small separation in proton NMR of the –CH3 signal of the N-acetyl peak of heparin and dermatan sulfate requires a high-field instrument and careful experimental control to ensure accurate results. Second, although Pan et al. equate not determined (ND) to ‘0.0’ in the “published results”17 column of Table 1 in their Correspondence, this need not be the case, owing to the limit of quantification for both the proton NMR method and the galactosamine method of Pan et al. As such, direct comparison of these results to one another is not warranted. Therefore, based on the data presented in our original paper1, the scientific
literature on the topic and data presented herein, we conclude the following: • OSCS is a major contaminant in suspect heparin lots collected in February/March of 2008. • Analysis of heparin samples using orthogonal analytical techniques, including high-resolution capillary electrophoresis and multidimensional NMR, indicates that oversulfated GAG mixtures are not a major contaminant in tested samples1,9,17. Nevertheless, analytical control tests for heparin should be assessed for their ability to detect oversulfated GAG mixtures, including individual components, most pertinently oversulfated dermatan sulfate. We have shown that a combination of 1D and 2D NMR, in conjunction with other analytical procedures, can readily detect such compounds, if they are present. • Oversulfation of heparin/heparin sulfate has been shown to lead to signature monosaccharides, including 3-O sulfoiduronic acid, 2,3-di-O-sulfouronic acid, and potentially N-desulfonation10,21 (Fig. 4 and Supplementary Table 1). Multidimensional NMR can readily detect signatures for oversulfated heparin (including the aforementioned 2,3-di-Osulfouronic acid), if they are present in a heparin preparation9. We observed none of these signals in any of the tested samples, indicating that oversulfated heparin/ heparan sulfate is not a major contaminant. • In 1D NMR experiments, certain signals may be masked if heparin contains a heparan sulfate impurity or is contaminated with oversulfated heparan sulfate. This is not a result of an interaction between polysaccharide species, it is simply because these species share close chemical identity. This is an important argument as to why orthogonal analytical approaches, including multidimensional NMR measurements1, should be used for the identification of contaminants and/or impurities in heparin. In conclusion, the most important result of the efforts in late 2007 and early 2008 of many individuals and organizations, including scientists from government, industry and academia, is an improvement in clinical outcomes for heparin treatment: since capillary electrophoresis and NMR tests to screen for OSCS were introduced by the Food and Drug Administration,
nature biotechnology volume 28 number 3 march 2010
the number of adverse events associated with heparin administration has been reduced to baseline once again. As such, starting with the introduction of screening methods, through the structural and biological work to identify and understand OSCS, to the monograph revisions instituted by the various pharmacopeias, the heparin supply chain has been secured, though constant vigilance is warranted. Note: Supplementary information is available on the Nature Biotechnology website. Note added in proof: since the submission of this letter, another study has confirmed the findings in Guerrini et al.1 and Kishimoto et al.17 that OSCS is a major contaminant in heparin (McKee et al. Structure Elucidation and Biological Activity of the Oversulfated Chondroitin Sulfate Contaminant in Baxter Heparin, J. Clin. Pharm. (epub ahead of print, 10 February 2010, doi:10.1177/0091270009355158) ACKNOWLEDGEMENTS The authors thank G. Cassinelli, M. Nasr and L. Buhse for discussions and the US National Institutes of Health for grant HL101721 to R.S. and R.J.L. to support this work. 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/. 1. Guerrini, M. et al. Nat. Biotechnol. 26, 669–675 (2008). 2. Trehy, M.L., Reepmeyer, J.C., Kolinski, R.E., Westenberger, B.J. & Buhse, L.F. J. Pharm. Biomed. Anal. 49, 670–673 (2009). 3. Viskov, C. et al. Clin. Appl. Thromb. Hemost. 15, 395–401 (2009). 4. McEwen, I. et al. J. Pharm. Biomed. Anal. 49, 816– 819 (2009). 5. Maruyama, T., Toida, T., Imanari, T., Yu, G. & Linhardt, R.J. Carbohydr. Res. 306, 35–43 (1998). 6. Zhang, Z. et al. J. Pharm. Sci. (in the press). 7. Guerrini, M. et al. Thromb. Haemost. 102, 907–911 (2009). 8. Toida, T. et al. Int. J. Biol. Macromol. 26, 233–241 (1999). 9. Guerrini, M. et al. Proc. Natl. Acad. Sci. USA 106, 16956–16961 (2009). 10. Casu, B. et al. Carbohydr. Res. 263, 271–284 (1994). 11. Li, B. et al. Biochem. Pharmacol. 78, 292–300 (2009). 12. Griffin, C.C. et al. Carbohydr. Res. 276, 183–197 (1995). 13. Guerrini, M., Naggi, A., Guglieri, S., Santarsiero, R. & Torri, G. Anal. Biochem. 337, 35–47 (2005). 14. Yates, E.A. et al. Carbohydr. Res. 294, 15–27 (1996). 15. Toida, T. et al. Biochem. J. 322 (Pt 2), 499–506 (1997). 16. Bisio, A. et al. Thromb. Haemost. 102, 865–873 (2009). 17. Kishimoto, T.K. et al. N. Engl. J. Med. 358, 2457– 2467 (2008). 18. Linhardt, R.J. et al. Biochem. Pharmacol. 42, 1609– 1619 (1991). 19. Sudo, M. et al. Anal. Biochem. 297, 42–51 (2001). 20. Neville, G.A., Mori, F., Holme, K.R. & Perlin, A.S. J. Pharm. Sci. 78, 101–104 (1989). 21. Yates, E.A. et al. Carbohydr. Res. 329, 239–247 (2000).
211
c orre s p on d en c e
© 2010 Nature America, Inc. All rights reserved.
Why FDA recruitment of ‘critics’is a problem To the Editor: The News article by Catherine Shaffer in the December issue1 entitled “FDA recruits prominent critics” contends that the “the general response” to the appointment of antiindustry zealot Peter Lurie of Public Citizen “is positive, even among those who don’t necessarily agree with Lurie’s positions.” As a former US Food and Drug Administration (FDA) official myself (from 1979 to 1994), I find it difficult to comprehend how Shaffer came up with such a misrepresentation. But given the scant number of sources (a senior FDA staffer, a PR specialist and a representative of a nonprofit (anti-corporate) lobbying organization that she quotes in the rest of her piece), perhaps Shaffer’s biased analysis and lack of balance simply reflect a low standard of reporting. The vast majority of FDA employees are civil servants. Unlike at some other federal agencies, there are only a handful of political appointees, and contrary to the thrust of Shaffer’s piece, most of President Barack Obama’s choices for them have been woefully inappropriate rather than “positive.” Apart from Meghan Scott (of the union-backed group WakeUpWalmart.com of Washington, DC) and Lurie, Shaffer fails to mention the dubious choices now brought into the fold: Joshua Sharfstein, deputy FDA commissioner, who in effect directs all dayto-day operations of the agency, has a history of anti–drug industry bias that dates from his days in medical school. His fingerprints are already evident on various costly, antiinnovative and excessive regulatory actions taken by the FDA. Ralph S. Tyler, newly appointed general counsel, whose main qualification seems to be that he is a crony of Sharfstein. Tyler, whose last job was insurance commissioner of Maryland, lacks any professional experience with FDA-related legal issues. Lynn Goldman, as a part-time consultant to the FDA’s lead scientist. While a senior Environmental Protection Agency official in the Clinton administration, Goldman never met a regulation she didn’t like and oversaw some of the most radical, unscientific policy- and decision-making imaginable (including toward agbiotech)—another inside-the-Beltway illustration that no bad deed goes unrewarded. Shaffer concludes her piece with a quote from Diana Zuckerman, president of the National Research Center for Women and Families, Washington, DC. Zuckerman, who
212
claims that FDA employees are promoted for their “willingness to please” regulated industry and that “industry is getting their way more often than the science would merit,” evidently believes that regulation is insufficiently stringent. Nowhere in this piece is it acknowledged that recruitment of Lurie and company will further exacerbate what is generally considered to be the increasingly stringent and stultifying regulation that has been imposed on drug makers over the past decade and which has increased the time and costs of drug development, diminished competition and slowed approvals to a trickle. An increasingly risk-averse US Congress has granted the FDA additional powers that place new restrictions on the prescribing, distribution, sale and advertising of drugs; and at the same time, regulators have imposed new criteria in addition to the statutory requirements for safety and efficacy, in order for drug sponsors to obtain even those limited approvals. What are these new criteria? Seemingly arbitrarily, the FDA sometimes demands that new drugs are not merely effective but are actually superior to existing therapies, a new standard that is often difficult and extremely costly to meet. In April 2007, the agency announced what appears to be a landmark policy decision: although the law requires that to be marketed, a drug must simply be shown to be safe and effective, by denying approval of Merck’s (Whitehouse Station, NJ, USA) new drug, Arcoxia (etoricoxib), a cyclooxygenase (COX)-2 inhibitor for the relief of arthritis pain, the FDA said that Arcoxia needed to be shown to be superior to existing drugs to obtain approval. Robert Meyer, director of the FDA office that oversees arthritis drugs [director of the Office of New Drug Evaluation II, Center for Drug Evaluation and Research], claimed that the agency’s advisory committee had sent a clear message that “simply having
another drug on the market…didn’t appear to be sufficient reason” for approval2. But whether or not the advisory committee meant to convey that (and in any case, advisory committee recommendations are not binding), it is specious reasoning. In addition, post-marketing studies as a condition of approval are tantamount to a new, fourth criterion for approval. Whereas they were once rare and the subject of discussions between FDA and drug sponsors, now they are required in more than threequarters of approvals, and FDA dictates what shall be done. In addition to the imposition of the new criteria to obtain approval to market new drugs, the FDA is now empowered to demand Risk Evaluation and Mitigation Strategies, some elements of which are so draconian that arguably they amount to limited approvals. They constrain physicians’ prescribing practices, corporate advertising and pharmacy practices, and have the potential to reduce drastically the potential market for new drugs. Finally, as measured by numerous metrics— number of clinical studies and patients to support a New Drug Application, number of black-box warnings on labels and the length of time required for and expense of clinical trials, for example—risk aversion at the FDA is high and escalating. If any of the above looks like “companies are still getting their way more often than the science would merit,” then the industry should expect dire times ahead indeed. COMPETING INTERESTS STATEMENT The author declares no competing financial interests.
Henry I Miller The Hoover Institution, Stanford University, Stanford, California, USA. e-mail: [email protected] 1. Shaffer, C. Nat. Biotechnol. 28, 7–8 (2010). 2. Richwine, L. FDA panel rejects Merck’s Vioxx successor. http://www.reuters.com/article/ idusn1234670420070413 (2007).
Genetic exceptionalism To the Editor: September’s Editorial1 contained a fascinating contradiction that illustrates a basic problem with our attitudes to genetic information. Consider the following two quotes, from the end of the first paragraph
and the start of the third. The context is Stephen Quake’s publication of his own genome sequence in the same issue2. “Like scientific pioneers before him, Quake is heroically self-experimenting—testing the risks in publishing identifiable personal
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
c orre s p on d en c e information of the most intimate kind.” “On its own, the sequence of letters in a human genome is uninformative. Its power for good arises only from associations with medical histories, behavioral characteristics, physical descriptions and environmental influences.” It is widely understood that, for most people, differences in health and longevity are influenced by a large number of allelic differences that indeed could be detected by sequencing their genomes and then comparing them to thousands of other genome sequences. Such comparisons are the goal of large-scale individual genome sequencing. Another technology can also detect them. We might call it ‘whole genome integral phenotype screening’, or we might call it ‘are your parents alive?’ This is more informative than the genome sequence is now and is likely to be as informative as the genome sequence will ever be for individual health outcomes, because these are modified by behavior, environment and the epigenetic knock-ons of early upbringing. For younger people, whether their grandparents are alive is more usefully diagnostic. Autobiographers and genealogists have routinely put this information into the public domain for decades, without expressions of awe at their daring or agonized debates about publishing information “of the most intimate kind.” The same people who have no qualms about telling life insurance companies whether their parents are still alive regard with horror the idea that they might reveal the results of genetic tests or genome sequences to those same companies. Why this response to gene sequence data? It is not that it is ‘fundamental’, as the second quote illustrates; it is fundamental only in the sense that it is the foundation on which any number of individuals could be constructed. It is not that you cannot change it. I cannot change when my grandparents died either. I suspect that the reason, as usual, is money. In the mid-1990s groups seeking venture capital funding for the growing area of genomics had to hype the importance, the primacy of the genome over all other aspects of biology. People wanting to promote their own shiny genome center (and their own promotion to its director) did the same to academic funders. I was at one conference in 1998 where a senior academic, who subsequently ended up running a major genome center, said “The genome project is the most exciting research project we shall ever see. ... The genome project is the blueprint for humanity” (a remark I
thought so egregious that I wrote it down verbatim). The venture capitalists bought it, as did the public markets and journalists five years later. The project was an outstanding technical success but, for reasons captured in the second quote, something of a biomedical flop. But the idea that DNA is our very essence remained. Individual genome sequences will prove scientifically valuable, although they will probably prove far less valuable than the current enthusiasm suggests: again, if history teaches us anything, it is that the latest breakthrough idea or technology is always going to solve everything and always disappoints. We will find some unexpected genes are involved in human strength and weakness. We will, for example (and here is a firm prediction) find that genes expressed solely in the mesolimbic system are causally linked to diabetes. But they are not our essence, not personal descriptions “of the most intimate kind.” They are, in fact, not much more informative than a short family history and probably less intimately revealing about our likely future health and longevity than combining genealogy (without any formal genetics) with the routine general practitioner questions about smoking, drinking, diet, exercise, marital status and so on. The morals of releasing any information are the province of debate in wider society. But what information can tell us, what information is trivial, useful, confidential or “personal information of the most intimate kind” is actually a technical issue, and it is our responsibility to present the nature of the knowledge that can be generated from that information in a realistic light. The genome, genetic information in general, is personal. So are weight, eating habits, residential zip-code and so on. Starting with journals such as those in the Nature group, scientists should make a concerted effort to roll back this stillgrowing belief that genetic information is qualitatively different from all other sorts of information about us, that the genome has an almost sacred centrality to the human condition. In short, it is time to can the hype. COMPETING INTERESTS STATEMENT The author declares no competing financial interests.
William Bains Rufus-Scientific, Cambridge, UK. e-mail: [email protected] 1. Anonymous. Nat. Biotechnol. 27, 777 (2009). 2. Pushkarev, D., Neff, N.F. & Quake, S.R. Nat. Biotechnol. 27, 847–850 (2009).
nature biotechnology volume 28 number 3 march 2010
213
F E AT U R E
Lost in migration George S Mack and Andrew Marshall
© 2010 Nature America, Inc. All rights reserved.
Combinations of cytostatic treatments and chemotherapies currently in clinical practice offer limited hope for patients whose cancers have spread. But increasing understanding of the processes underlying metastasis may one day provide other therapeutic options.
I
n the United States alone, over 565,000 people will die from cancer this year according to the American Cancer Society. As invasive and expansive as primary tumor masses might be, they account for only one out of ten cancer fatalities. For those diagnosed with solid tumors, occult metastases in such organs as the bone marrow, lung, liver or brain account for the vast majority of cancer deaths. Over the past century, drug developers and clinicians have honed surgical techniques and introduced an increasing array of poisons, including radiotherapies or cytotoxic chemotherapies, as well as molecularly targeted cytostatic drugs in the battle against cancer. After diagnosis, if a metastasis has not already occurred, in many cases removal of the primary tumor can restore a patient to long-term health. But all too frequently, metastatic colonies, after multiple rounds of standard-ofcare therapies, can reactivate and become life threatening—sometimes months, sometimes years after the initial diagnosis. Although cancer biology has recently made progress in understanding how cancer spreads, our understanding of the molecular mechanisms of metastasis remains rudimentary. Basic research lacks animal models that closely resemble the course and pattern of clinical disease in different cancers. Controversy continues concerning the origin and contribution in various tumors of cells from the primary mass and disseminated tumor cells (DTCs, which, because of their progenitor-like qualities, have been termed ‘cancer stem cells’). And the molecular similarities between primary tumors and their metastases remain unclear; indeed, the combination of genomic instability and differing selective pressures and environGeorge S. Mack is a freelance writer based in Columbia, South Carolina. Andrew Marshall is editor, Nature Biotechnology.
214
mental cues at sites distant from the primary tumor (metastatic speciation) can make secondary tumors a moving target for therapeutic intervention. The clinical complexities of selecting cancer patients who respond to treatment with novel targeted agents, the advanced stage of disease that many of these patients are in and the lack of definitive regulatory criteria to measure the effectiveness of treatments all combine to make drug development against m etastatic cancer a daunting challenge. An emerging picture The metastasis field has arrived at an important crossroads. Once a murky and incomprehensible disease process, cancer is beginning to yield to recent advances in cancer biology that throw light on not only the mechanisms by which primary tumors disseminate through the body and then colonize a secondary organ but also the interactions between tumor cells, stromal cells and immune cells involved throughout this metastatic cascade. Cancers spread when tumor cells escape from the primary cancer into the surrounding tissue, invade the vasculature (a process termed intravasation), voyage through the vascular system (blood or lymph) to a distant location and, reversing the process, exit from the vasculature (extravasation) into an organ remote from the original site and establish a secondary tumor (Fig. 1). It is these secondary tumors that cause the majority of cancer deaths, either through direct organ compromise or through complications as a result of last-ditch treatments. The metastatic process is by its very nature inefficient—and ideas on which part of the cascade is the rate-limiting step have varied over the years. Whereas it was once assumed that metastasis occurred only after the primary tumor had achieved a certain size (that is, the ‘linear progression model’), recent
work tracking tumor cells as they disseminate throughout the body suggests that at least for some cancers (e.g., ductal carcinoma), seeding of cells into distant organs happens early on1. What’s more, single tumor cells disseminated to the bone marrow and peripheral blood can often be detected years before the occurrence of clinically detectable metastases. In other cases (e.g., breast cancer), a significant fraction of patients with detectable DTCs never develop distant metastases. Conversely, in 5% of the hospitalized cancer population, patients have metastatic disease without diagnosis of the primary tumor (termed CUP; cancer of unknown primary). This suggests that in certain cancer types, DTCs sometimes acquire the capacity to proliferate and colonize early on, even if epithelial tumor cells at the primary site fail. What’s more, colonization of the distant tissue to establish a secondary tumor, which requires both a permissive microenvironment as well as a migratory malignant cell receptive to the growth-promoting signals of the destination organ—otherwise known as the ‘soil and seed’ hypothesis first promulgated by the late 19th century British surgeon Stephen Paget—appears to be a key, rate-limiting step. The emergence of a secondary cancer is so rare, at least relative to the number of tumor cells set free into circulation, that as few as 0.01% of circulating cells ultimately survive to form new viable distant colonies of tumor tissue, according to cancer biologist Isaiah (Josh) Fidler of the University of Texas MD Anderson Cancer Center (Houston). A one-cubic-centimeter tumor mass can contain a billion cells and shed a million cells per day into circulation. “And if only ten cells survive, that means there are ten chances that a metastasis will occur every day,” he says. However, excluding dormant secondary lesions that may never be detected or do any harm, statistics tell investigators that at the
volume 28 number 3 march 2010 nature biotechnology
feature Primary tumor Tumor-associated macrophage RHoC LOX MET MMP-9 NEDD-g
a
Platelets
DTC
Endothelial progenitor cells (EPC)
b
c
© 2010 Nature America, Inc. All rights reserved.
Hematopoietic progenitor cell (HPC)
Mesenchymal stem cell (MSC) Perivascular fibroblasts
Secreted fibronectin
TNFα TGFβ MMP-9 SDF1 P-selectin E-selectin CD44 LOX VEGFA PlGF Angiopoietin-like protein 4
Pre-metastatic niche
VEGFA PlGF RANKL CXCR4 IL-11 ET-1
Micrometastasis
Macrometastasis
Figure 1 Steps to metastasis formation. (a) In response to factors secreted by the primary tumor, including RHoC, LOX, MMP-9, inflammatory S100 chemokines and serum amyloid A3 (SAA3) are upregulated in pre-metastatic sites leading to clustering of bone marrow–derived hematopoietic progenitor cells (HPCs). HPCs secrete a variety of pre-metastatic factors, including tumor necrosis factor α (TNF-α), MMP9 and TGFβ. Activated fibroblasts, possibly derived from mesenchymal stem cells (MSCs), secrete fibronectin, an important adhesion protein in the niche and LOX expression is increased, modifying the local ECM. (b) DTCs engraft the niche to populate micrometastases. The site-specific expression of adhesion integrins on activated endothelial cells such as P-selectin and E-selectin may enhance DTC adhesion and extravasation at these sites, and cell-cell interactions such as CD44 ligation in the metastatic niche may promote DTC survival and enable proliferation. (c) Recruitment of endothelial progenitor cells (EPCs) to the early metastatic niche mediates the angiogenic switch and enables progression to macrometastases and colonization. Adapted from ref. 28.
time of solid tumor diagnosis, roughly half of all patients have a metastatic lesion already in place, and some of these may have been growing for 2 to 3 years. Another intriguing clinical observation is that some cancers (e.g., lung and pancreas) metastasize very soon after primary tumor cells progress through the tumor boundary,
whereas others spread much later—as much as decades later in the case of breast cancer, prostate cancer and ocular melanoma. In addition, particular cancers tend to metastasize to particular organs (Table 1, Box 1 and Fig. 2), which cannot be fully explained by organ accessibility or vascular traffic patterns. Thus, although the latency period for breast cancer
Table 1 Tropisms of common cancers Tumor type
Principal sites of metastasis
Breast
Bones, lungs, liver and brain
Lung adenocarcinoma
Brain, bones, adrenal gland and liver
Skin melanoma
Lungs, brain, skin and liver
Colorectal
Liver and lungs
Pancreatic
Liver and lungs
Prostate
Bones
Sarcoma
Lungs
Uveal melanoma
Liver
Source: ref. 2.
nature biotechnology volume 28 number 3 march 2010
metastases in the brain (which is protected by the blood-brain barrier) is longer than that for bone marrow (where there are large fenestrations in blood vessels), lung adenocarcinoma and other highly invasive cancers are still capable of rapidly spreading to both locations irrespective of accessibility. The molecular events underlying these different proclivities remain poorly understood, although molecular signatures associated with tropism to different distant sites are coming into focus2 (Box 2). Vital imaging has allowed researchers to see tumor cells move throughout the body of experimental animals (Box 3 and Fig. 3), and some of the properties of the ‘pre-metastatic niche’ are coming to light. Breaking out of the basement Once a primary tumor has acquired initiating mutations that proffer unlimited proliferation potential, repress apoptotic mechanisms, redirect cell metabolism and accelerate genomic instability and aneuploidies, the next step in
215
feature
© 2010 Nature America, Inc. All rights reserved.
Box 1 Bone—a model for drugs targeting metastases in specific organs? Most cancer treatments in clinical practice Infiltration and latency currently focus on the primary tissue of origin, but Bone marrow Bone matrix Circulating another approach is to treat malignancies from the cancer cells Stromal perspective of the site of metastasis. This would niche include both the migrating tumor cells and the stromal cells at the organ to which they spread; thus, breast cancer cells that metastasize to lungs SDF1 express such factors as angiopoietin-like 4, COX-2 and MMP-1, whereas those that spread to bone express parathyroid hormone-related peptide, G0 (dormant) tumor necrosis factor α, IL-6 and IL-11 and respond to stromal-cell derived factor 1 produced in the bone marrow. A current example of a cancer therapy where the concept of targeting a specific metastatic niche is being realized is the use of agents that Growth Death Micrometastasis inhibit osteoclast activity in bone metastasis factors signals (indolent growth) (Fig. 2). The balance between osteoblast activity and osteoclast activity is important Colonization competence (years to decades) Odanacatib in bone metastasis as bone matrix must be TGFβ, IGF1 and destroyed to create a niche for tumor cell BMPs colonies to grow. Novartis’ bisphosphonate Zometa (zoledronic acid), a small molecule Cathepsin K Osteolytic macrometastasis that binds to hydroxyapatite in the bone matrix, Bone was approved in early 2002 and marketed lysis for multiple myeloma and to manage or delay ACE-011 bone metastasis from lung, breast and prostate Osteoclasts Activin cancers (as well as to treat hypercalcemia of receptor malignancy). Bisphosphonates promote apoptosis RANKL PTHRP, IL-6 Prolia Activin of osteoclasts, and, accordingly, Zometa delays IL-11 and TNFα Osteoprotegerin Myeloid or prevents skeletal-related events, such as bone progenitor cells deterioration, fractures and metastatic bone DKK-1 BHQ880 pain. Amgen’s humanized monoclonal antibody Osteoblasts Prolia (denosumab; formerly AMG 162) binds to receptor activator of NF-κB ligand (RANKL) on Stimulate Inhibit the cell membrane of osteoclasts and prevents RANKL interaction with its receptor RANK, which Figure 2 Cellular pathways associated with bone metastases and associated therapies. Drugs reduces bone resorption by inhibiting osteoclast are shown in red. Adapted from ref. 2. activity. In a head-to-head phase 3 trial, Amgen released data last July showing that Prolia BHQ-880, an intravenous neutralizing human mAb against demonstrates a clear advantage over Zometa the soluble endogenous Wnt inhibitor Dickkopf-1 (DKK1) that on postponement of time to the first metastatic skeletal-related stimulates osteoblastogenesis, in phase 1b/2 trial of myeloma events. The initial disease indication for Prolia is post-menopausal patients with bone metastases. And Acceleron is developing osteoporosis, and the product could be a $2.5 billion–$3.0 otatercept (ACE-011), a fusion protein containing a soluble form billion per year product, according to Wells Fargo (San Francisco) of the activin IIa receptor and an Fc fragment of IgG1, which was senior biotech analyst Aaron Reames, who anticipates no major reported in December to result in remission in 7 out of 22 multiple roadblocks to FDA approval. myeloma patients with osteolytic lesions. Three other companies with programs in proof-of-concept Therapies specifically targeting lung, liver and brain metastases trials targeting the bone metastatic niche are Merck, Novartis are further in the future, although several startups are attempting to and Acceleron Pharmaceuticals (Cambridge, MA, USA). Merck address the difficulty of getting cancer drugs to penetrate the bloodis investigating a small molecule targeting lysosomal cysteine brain barrier to access secondary tumors there. Thus, by conjugating protease cathepsin K, an enzyme expressed by osteoclasts Taxol to angiopep-2, a 19-mer peptide mimetic of the N-terminal during bone resorption and metastasis. In a randomized, doublesequence of aprotinin that binds the lipoprotein receptor-related blind phase 2 trial of 43 patients with breast cancer and bone protein on the blood-brain barrier, AngioChem (Montreal) hopes to metastases, the drug, odanacatib (a [2,2,2-trifluoro-1-(biphenylfacilitate delivery to malignant gliomas—a phase 1 trial is currently 4-yl)ethyl]-4-fluoroleucine derivative of 4-fluoroleucineis), has ongoing. And, in The Netherlands, to-BBB Technologies (Leiden) is shown robust, sustained and reversible antiresorptive activity, with developing liposomes coated with glutathione-conjugated PEG to no demonstrable effect on off-target cathepsins. Novartis (under deliver Adriamycin to brain tumors. GM and AM license from Morphosys (Martinsreid, Germany)) is investigating
216
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
feature cancer progression involves the loss of cellular polarity and detachment from the basement membrane. In addition, alterations are induced in the interactions between tumor epithelial cells and the extracellular matrix (ECM). Enzymes, such as matrix metalloproteinases (MMP-1, MMP-2, MMP-3 and MMP-13), epilysin and the transmembrane serine protease TMPRSS4, are secreted from or upregulated on tumor epithelial cell surfaces and into the ECM to trigger detachment, allowing cells to migrate away from the primary tumor mass. MMPs are zinc-dependent endopeptidases capable of breaking down the elements of ECM (that is, collagens, fibronectin, laminin, elastin and basement membrane glycoproteins), which interact with tumor cell surface proteins, such as β4, α5β1 and αVβ6 integrins and cadherins, and they are involved in making space for the angiogenesis and lymphogenesis that enables nourishment and progression of tumor cells outside the tumor margins. From the drug developer’s point of view, the accessibility of the ECM and cell membrane proteins compared with intracellular targets could be advantageous, particularly in terms of dose-limiting toxicities. Thus far, however, little success has been achieved in targeting these enzymes. In the 1990s, British Biotech Pharmaceuticals (Oxford, UK) developed two MMP inhibitors, batimastat for injection and marimastat, for oral use, both of which are hydroxamic acid derivatives. Because these were nonselective MMP inhibitors, patients receiving the drugs experienced severe muscle and joint pain, and the reduced dosages employed lacked efficacy in trials for glioblastoma, pancreatic, gastric, breast and ovarian cancers. Several other MMP inhibitor flops in cancer followed, including Bayer’s (Leverkusen, Germany) tanomastat, Novartis’s (Basel) MM1 270 and Agouron’s prinomastat (AG3340). These early disappointments in targeting MMPs have had the effect of discouraging other drug developers from targeting these enzymes. “Let’s say they were broad spectrum and not specific for certain MMPs,” says molecular cell biologist Reuven Reich of Hebrew University of Jerusalem. “That’s the reason these kinds of drugs failed, we think.” Recently Reich and his collaborators Amnon Hoffman and Eli Breuer, both also of Hebrew University, published findings from in vivo work they had been doing in murine models with a new MMP inhibitor, cis-2-aminocyclohexylcarbamoylphosphonic acid (cis-ACCP)3. They believe their small orally bioavailable molecule could be useful as an antimetastatic agent. It’s a more specific compound than marimastat targeting only MMP-2, and it has less affinity for iron than
the hydroxamic acid products, that in all probability combined with iron from myoglobin to precipitate out into the tissues and cause severe pain in those taking the British Biotech drug marimastat. But because of its specificity, cis-ACCP is a much milder inhibitor and is soluble in water, which is ideal for extracellular targeting and also elimination through the kidneys. Thus far, their research group has gathered data using two cancer models in mice—murine melanoma cells injected into the tail veins and an orthotopic human prostate cancer cell model. The data show solid tumor metastases in mouse models reduced by ~90%. There was no toxicity detectable, and in rats it was shown that 84% of intravenously administered cis-ACCP was eliminated, unchanged, in the urine. Reich is now looking to “partner with a pharmaceutical company to help us to take it to a higher level.”
Time will tell whether the early excitement in MMPs as drug targets in metastasis is justified. What is clear is that the expression of MMP subtype varies with cancer stage and type. Thus, non–small cell lung cancer (NSCLC) overexpresses MMP-11 and MMP14 rather than MMP-2, which likely means that agents, such as cis-ACCP, are unlikely to be efficacious in this disease. With this in mind, further advances in our understanding the complexities of MMP functions in vivo might be required before targeted agents become a reality. Another family of ECM proteins linked with metastatic processes is the S100 proteins, especially S100 calcium-binding protein A4 (S100A4, formerly known as MTS1). Evidence is accumulating that S100A4, when upregulated in the ECM, disrupts cell-to-cell adhesion and facilitates the process of break-
Box 2 Early diagnosis? Ultimately, the ability to keep one step ahead of cancer may depend on having the ability to detect metastases earlier in the process, rather than attempting to treat advanced, established disease. As it is the ethos of clinicians to insist on seeing evidence of disease before treating a patient, improved diagnostic tests, both molecular and imaging, that exploit increasing understanding of the molecular events that predict critical mitogenic and metastatic alterations will be critical to improvement in cancer patient outcomes. In 2006, molecular cell biologists Bruce Zetter and Marsha Moses, both of Children’s Hospital Boston and Harvard Medical School, co-founded Predictive Biosciences (Lexington, MA, USA) to develop noninvasive tests that can make sense of certain biomarkers, notably MMPs and a disintegrin and metalloproteinases, both of which have been shown to contribute to ECM degradation and associated tumor cell detachment. The idea is to develop algorithms around these enzymes so patients can be monitored for recurrence and progression of diverse types of malignancies, somewhat like the prostate-specific antigen test that warns against advancing prostate cancer. Because these enzymes are detectible in urine, patients can be tested as frequently as necessary. Predictive Biosciences has a urine bank of over 8,500 samples that they are using to validate a library of biomarkers licensed from Children’s Hospital of Boston and affiliated institutions. The company’s first target is bladder cancer survivors, who undergo painful cystoscopies every few months to look for the recurrent disease. A recent study of 530 patients, 84 with bladder cancer, shows that levels of MMP-9 are effective in discriminating disease-free patients from those suffering a recurrence 42% of the time15. In addition, the Moses laboratory has described a biomarker, lipocalin 2, which is associated with progressive breast cancer and which can be measured in urine16. Another way to get at metastases early is looking for the presence of DTCs. Although originally accomplished by identifying cells with epithelial markers present in the bone marrow that is predominantly mesenchymal in origin, this approach is now being tested in the peripheral blood. One such system has been offered since 2004 by Veridex (a Johnson & Johnson company located in Raritan, NJ, USA). The technology, called CellSearch, pulls circulating tumor cells from the blood of colon, breast and prostate cancer patients using magnetic nanoparticles coated with mAbs against EpCAM. Researchers at Massachusetts General Hospital (Boston) have also developed a microfluidic device, the CTC (circulating tumor cell) chip, which has 78,000 antibodycoated microposts on a surface the size of a business card that can trap DTCs expressing EpCAM from whole blood. Mehmet Toner, professor of biomedical engineering at HarvardMIT Division of Health and Science Technology, and his colleagues have shown that their device is capable of detecting DTCs in the blood of patients with five different tumor types with 99% accuracy17. Commercialization has been handed off to a startup, On-Q-ity (Waltham, MA, USA). Laura DeFrancesco and GM
nature biotechnology volume 28 number 3 march 2010
217
feature
© 2010 Nature America, Inc. All rights reserved.
Box 3 Modeling metastasis Almost all drugs tried in humans work against subcutaneous xenografts in mice. The problem is this hardly ever translates to the clinic. According to a 2004 study, as few as 3.8% of patients in phase 1 cancer drug trials between 1991 and 2002 achieved an objective clinical response18. Placing tumor cells under the skin, although commonly used in drug testing settings because the tumors are easy to establish and measure, does not take into account tumor tropism—the predilection for tumors to grow in only certain environments. In seminal work conducted over 20 years ago describing the behavior of tumor cells Figure 3 AntiCancer’s MetaMouse allows visualization of tumors in whole animals following in different microenvironments, MD Anderson’s orthotopic implantation of human pancreatic tumor cells expressing red fluorescent protein. Fidler confirmed this principle and went on to The mice in the different panels received different treatments. Reprinted with permission develop orthotopic models for studying metastasis, from ref. 29. in which cells derived from a variety of human to see where tumors developed, researchers can see the tumors tumors were implanted into correct anatomical and their progress in real time with an imaging system (Fig. 2). sites in nude mice19. The difference is night and day. Measuring Taking a different approach to dissecting metastasis, the metastasis from tumors planted subcutaneously, Fidler reported Canadian company Innovascreen (New Glasgow, Newfoundland, zero successes in 700 tries, whereas orthotopic placement of Canada) has developed an in vivo system to observe and quantify tumor tissue in mice produced metastases in every tumor type tumor cell behavior and treatment response using a shell-less (ex attempted. ovo) avian embryo system. Innovascreen founders John Lewis, the Echoing that sentiment is Robert Hoffman, president and CEO, and CSO Andries Zijlstra originated an intravital imaging CEO of AntiCancer (San Diego) and professor of surgery at system while at Scripps Clinic (La Jolla, CA, USA)22, which is the University of California at San Diego. “An intact tumor sold as a service by Innovascreen. By injecting tumor cells into microenvironment is necessary for a good cancer model,” he the chorioallantoic membrane where they form a tumor, or into says. Hoffman’s company has commercialized the concept the vasculature of the chorioallantoic membrane, they are able to by creating a set of mice in which tumor tissue is implanted monitor the migration of cells away from the primary tumor, the orthotopically at various sites to follow tumor progression and invasion of the vasculature by tumor cells as well as extravasion dissemination20. Hoffman, along with his then post-doc Takashi from the vessels using three-dimensional time-lapse photography. Chisima, in 1996 had the idea to make the tumors fluorescent The imaging system can visualize as many as six different so that imaging could be used to follow tumor cells to distant molecules, including fluorescently labeled therapeutics that target tissues and organs. It would be the first time cancer metastases vasculature and fluorescent proteins transduced into the tumor would be observed through expression of green fluorescent cells. Using this direct observation approach, preclinical drug protein21. Today, the company offers a set of organ-specific candidates can be quickly assessed for their ability to affect each animal models, called MetaMouse, which have tumor tissue step in metastasis. “These assays are all designed to be completed implanted in different organs (e.g., breast, brain, prostate), that within 3 weeks,” Lewis says. “It allows for flexibility in planning allow stromal cells and tumor cells to be labeled with different and refinement of dosing and other experimental parameters to fluorescent labels so that interactions between the two cell types evaluate investigational drugs.” LD and GM can be studied in vivo. Instead of killing and opening the animal
ing through basement membranes. Although S100A4 has been accepted for years as a useful prognostic biomarker—when it is not expressed aberrantly, many breast cancer patients can survive to live out a full life4— until recently, it hasn’t attracted much attention as a therapeutic target. This is set to change as Supratek Pharma (Montreal) pushes ahead with a S100A4targeting drug that was originally approved two decades ago. Supratek has modified the antiallergy product azaxanthone with an excipient that enhances solubility to create an orally bioavailable drug, SP-MET-X1. The intention is to administer SP-MET-X1 in larger doses than labeled for the allergy indi-
218
cation, but as the safety of azaxanthone has been long established, the company hopes that it will be a good candidate for long-term chronic treatment, administered several times daily, potentially for years, as a prophylaxis and maintenance therapy. Thus far, supporting data from preclinical animal models show that the drug inhibits or delays metastasis formation, implantation and progression in animal models. Moreover, Supratek investigators believe SP-MET-X1 can be synergistic with other forms of therapy, such as chemotherapy. The company is planning to begin a phase 1 trial with SP-MET-X1 this year, but the first indication for which the company will file is not yet settled.
Inhibiting urokinase Another enzyme that is able to proteolytically degrade the ECM and basement membrane around primary tumors is the secreted 54 kDa serine protease urokinase-type plasminogen activator (uPA). In research that originated at the Technical University of Munich two decades ago, investigators Manfred Schmitt, Viktor Magdolen, Nadia Harbeck and Olaf Wilhelm were searching for reasons why a subset of post-surgical breast cancer patients did particularly poorly in terms of survival. By studying patient plasma and tumor samples, they discovered that levels of uPA are inversely correlated with survival. Their findings ultimately revealed that uPA can trigger
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
feature a plasminogen proteolytic sequence in the ECM that under normal conditions serves to prevent clotting in the alveoli of the lungs and glomerular apparatus of the kidneys. But uPA can also degrade ECM, interact with MMPs and erode into the microvasculature, thereby allowing tumor cells to exit from a primary site and entry into the lymph or bloodstream. These DTCs also can generate uPA, enabling them to escape from the narrow lumina of microvasculature into new host sites where they break down connective tissues to create a nidus for potential metastatic colonies. It is now known that uPA is produced and secreted as a single-chain proenzyme (prouPA) that binds to the cell surface uPA receptor (uPAR). On receptor binding, pro-UPA is cleaved (primarily by plasmin but also by kallikrein, blood coagulation factor XIIa and cathepsin B) into its two-chain, active form, targeting activity to areas of the cell surface containing uPAR, where it cleaves ECM components, such as fibronectin and laminin receptor (integrin α6β1). Besides its role in proteolysis, uPAR also regulates many cell surface proteins, such as integrins, growth factor receptors and G protein–coupled receptors, and activates the signaling of metastasispromoting factors, such as basic fibroblast growth factor, vascular endothelial growth factor (VEGF), transforming growth factor beta (TGF)β and hepatocyte growth factor/scatter factor (HGF/SF). The proteolytic activity of uPA is regulated by plasminogen activator inhibitor-1 (PAI-1) and PAI-2, the former of which induces internalization of the enzyme–receptor complex, leading to uPA degradation. Developing agents that target uPA has been difficult, in part due to the problem of achieving specificity versus other serine proteases, such as thrombin or plasmin. Back in the 1990s and early 2000s, several companies were working on uPA inhibitor programs, including Abbott Laboratories (Abbott Park, IL, USA), GlaxoSmithKline (Brentford, UK) and Pfizer (New York), as well as smaller companies, such as Corvas International (acquired by Dendreon of Seattle in 2003) and 3-Dimensional Pharmaceuticals (acquired by Johnson & Johnson (New Brunswick, NJ, USA) in 2003). One reason these programs were largely ineffective is that compounds need to be very basic to bind uPA’s active site, which almost invariably creates pharmacokinetic problems in making the drug bioavailable. One biotech company, however, founded by Wilhelm, Schmitt and Magdolen, is continuing development of a small-molecule pro-drug that binds uPA. Wilex (Munich) is currently testing Mesupron (formerly
WX-671) in combination with chemotherapy in phase 2 trials as a first-line treatment for individuals with HER2-negative metastatic breast cancer (Table 2). Wilex researchers have addressed the pH problem by creating a pro-drug Mesupron, which has excellent bioavailability and can be given orally but is pharmacologically inactive. In the liver and other tissues, Mesupron is metabolized to the company’s original formulation WX-UK1, a potent inhibitor of uPA. Because the drug is orally bioavailable and has shown acceptable toxicity in phase 1 studies in healthy volunteers, the company suggests their agent could potentially be used to prevent or depress metastasis formation on a chronic basis after tumor resection or other gold standards of care. Indeed, the US Food and Drug Administration (FDA), which typically wants to see new chemical entities tested in latestage disease after other therapies have failed, approved a phase 2 trial of Mesupron with the chemotherapy Gemzar (gemcitabine), which began in July 2008, for locally advanced, nonoperable, nonmetastatic pancreatic cancer to determine if there is improvement in response rate and progression-free survival and to see if time to metastasis and overall survival can be improved. Thus far, the results are promising; in some of these patients, mortality is lower and therapy is being continued longer than originally planned. “Our approach is somewhat different,” says Wilex head of R&D Paul Bevan. “We’re not looking to get rid of metastases—we’re looking to stop them from forming.” The receptor for uPA (uPAR) also represents an interesting target. With the exception of neutrophils and monocytes, quiescent cells almost never express uPAR. An initial report establishing that it is particularly enriched in metastatic lesions compared with primary tumors has since been confirmed by several other studies5. Moreover, immunohistochemistry of primary tumor masses shows that uPAR is more concentrated on tumor cells located at the invasive front or leading edge, which is consistent with the assertion that the receptor could be a metastatic driver. Investigators at the now-out-of-business biotech company Attenuon, which was in San Diego, and collaborators at the University of Texas MD Anderson Cancer Center have validated uPAR alone as a blockade target for preventing metastasis through a pleiotropic group of events that include downregulated expression of c-MET (mesenchymal-epithelial transition factor) and suppression of insulinlike growth factor 1 (IGF-1)-dependent tumor cell migration and invasion in both colon and pancreatic cancer cells. Former Attenuon CSO
nature biotechnology volume 28 number 3 march 2010
Andrew Mazar, now entrepreneur in residence and member of the Robert H. Lurie Cancer Center at Northwestern University in Evanston, Illinois, has recently found a licensing partner for his former company’s preclinical humanized monoclonal antibody (mAb) huATN658, which targets uPAR. The antibody has not been tested in humans as yet, but mouse studies have demonstrated significantly suppressed human pancreatic tumor growth and liver metastases along with complete repression of retroperitoneal invasion. “I think the product still has a future,” says Mazar. “Clearly, uPAR is an important and very selective target for a variety of different tumor types, but until now companies have never really figured out how to go after this particular receptor.” Today, Mazar is working on a new project to inhibit metastasis by targeting the uPA using nanoparticles invented at Northwestern. “I think this approach also has the potential to broadly target a broad spectrum of tumor types,” he says, “And in fact, the profile of tumors to be targeted will be similar to those targeted with the anti-uPAR antibody.” It is worth noting that many clinicians view PAI-1 as one of the most informative prognostic markers in several cancers, high levels of the protein being associated with poor prognosis. Cancer in transition Alterations in the interactions between the primary tumor and ECM are also accompanied by the conversion of epithelial tumor cells into mesenchymal cells—the so-called epithelialto-mesenchymal transition (EMT). This is associated with the expression of several small noncoding regulatory RNA molecules (Box 4). EMT is thought to be a pivotal event in metastasis, endowing tumor cells with their migratory, invasive and stem-like properties, their ability to suppress apoptosis and senescence and their capacity to dampen immune responses. It has long been known to be involved in embryogenesis, but its role in the spread of human cancer has become widely appreciated only recently, mainly because evidence of its occurrence in clinical samples was overlooked—individual mesenchymal cells originating during EMT are hard to differentiate from stromal cells or other tumorassociated fibroblasts in the vicinity. An important initial step in EMT is the degradation by MMPs (e.g., MMP-3 or MMP-13) of the protein E-cadherin (epithelial calciumdependent adhesion molecule), a molecule that maintains adhesion between cells and acts as a tumor suppressor in normal cells. When E-cadherin expression diminishes, the appearance of N-cadherin (neuronal calciumdependent adhesion molecule) on tumor cell
219
feature
Table 2 Select agents in late-stage development that target aggressive/metastatic cancers Company
Target
Agent
Oncology indications
Development stage
Novartis
Hydroxyapatite of bone matrix
Zometa
Bone metastases in multiple myeloma and breast-to-bone metastasis
Marketed (February 2002)
Amgen
RANKL (RANK ligand)
Denosumab (humanized mAb)
Breast-to-bone metastasis and prostateto-bone metastasis
Phase 3
Angiopoietin 2
AMG-386 (Fc fragment linked to 20-res- Breast, ovarian and RCC idue peptide that binds angiopoietin-2)
© 2010 Nature America, Inc. All rights reserved.
Exelixis
Phase 2
IGF-1R
AMG-479 (a fully human mAb)
Advanced solid tumors
Phase 2
c-MET
Rilotumumab (a fully human IgG2 mAb)
Metastatic colon cancer
Phase 2
pan-RTKs
XL-184 (small-molecule RTK inhibitor)
Thyroid cancer
Phase 3
pan-RTKs
XL-184
Advanced solid tumors
Phase 2
OSI Pharmaceuticals
IGF-1R
OSI-906 (small-molecule RTK inhibitor)
Metastatic adrenocortical carcinoma
Phase 3
Adherex Technologies
N-cadherin
ADH-1 (a cyclic pentapeptide)
Melanoma
Phase 2
Antisense Pharmaceuticals
TGFβ
Trabedersen (a phosphorothioate oligodeoxynucleotide)
Glioblastoma
Phase 2
ArQule
c-MET
ARQ-197 (a small-molecule RTK inhibitor)
Various advanced solid tumors
Phase 2
Bristol-Myers Squibb
SRC-family protein-RTKs
Dasatinib (orally bioavailable small molecule)
Breast-to-bone metastasis, breast cancer (triple negative), colorectal cancer and liver
Phase 2
Centocor (Johnson & Johnson)
αvβ3/αvβ5 integrin
Intetumumab (fully human mAb)
Metastatic melanoma
Phase 2
Genentech
c-MET
MetMAb (a humanized monovalent 5D5 Fab mAb)
NSCLC
Phase 2
SMO
GCD-0449 (small-molecule inhibitor)
Metastatic basal cell carcinoma, colorectal cancer and ovarian cancer
Phase 2
Malignant melanoma
Global TransBiotech
Heparanase
PI-88 (phosphomannopentaose)
GlaxoSmithKline
c-MET, VEGF-R2, AXL RTKs
Foretinib (small-molecule RTK inhibitor) Metastatic squamous cell carcinoma and gastric cancer
Phase 2 (completed) Phase 2
Facet Biotech
α5β1 integrin
Volociximab (a chimeric IgG4 mAb)
Various solid tumors
Phase 2
MethylGene
c-MET, VEGFR 1,2,3, Tie 2 and Ron RTKs
MGCD265 (small-molecule RTK inhibitor)
NSCLC
Phase 2
Merck
IGF-1
Dalotuzumab (MK-0646, a humanized mAb)
Metastatic colon cancer and others
Phase 2
Cathepsin K
Odanacatib, a [2,2,2-trifluoro-1(biphenyl-4-yl)ethyl]-4-fluoroleucine derivative of 4-fluoroleucine
Breast and bone metastases
Phase 2
Wilex
uPA
WX-UK1 (serine protease inhibitor)
Breast cancer and other solid tumors
Phase 2b
uPA and other serine proteases
Mesupron (pro-drug of WX-UK-1)
Pancreatic, breast, and head and neck cancer
Phase 2
surfaces is known to be a survival factor for melanoma cells and doubtless other tumor cells as well6. Indeed, with N-cadherin expression, epithelial tumor cells acquire a mesenchymal cell phenotype, more like fibroblasts, with an ability to migrate and navigate across basement membranes, pass between endothelial cells and enter into vascular lumina. Recent work by cancer biologist Robert Weinberg, director of the Ludwig Center for Molecular Oncology at MIT (Cambridge, MA, USA), shows that the entire EMT program can be triggered by any one of at least seven early embryonic transcription factors (e.g., Snail1/2, survival of motor neuron protein interacting protein (SIP1), ZEB1, ZEB2, Slug, KLF8, E12/E47A-E2A). Several other proteins also repress E-cadherin indirectly, including Twist, Goosecoid, E2.2 and Foxc2. “It seems
220
increasingly clear that human cancers can turn on these long-silent genes, and in so doing the cells acquire in one fell swoop most all of the attributes that they need in order to execute the steps of the invasion metastasis cascade,” says Weinberg. So far, only one company has attempted to clinically exploit EMT by specifically targeting N-cadherin. Adherex Technologies (Durham, NC, USA) has been conducting a phase 2 trial for melanoma patients with ADH-1, a cyclic pentapeptide containing the N-cadherin extracellular domain cell adhesion recognition motif (histidine-alanine-valine). Cadherins normally bind to each other from cell to cell, but if any one of those three amino acids is disrupted, intermolecular binding is prevented. ADH-1 blocks the adhesion sites and hence the binding. Surgical oncologist Douglas Tyler
of Duke Medical Center (Durham, NC, USA), principal investigator in the ADH-1 trials, had previously demonstrated marked improvements in tumor responses in a rat model of melanoma using systemic ADH-1 plus locally administered melphalan (l-phenylalanine mustard) compared with melphalan alone. These preclinical studies led the way for initiation of clinical trials with systemic ADH-1 plus melphalan for patients with advanced extremity melanoma. After a positive phase 1 trial, phase 2 results with overall response rates of 58% versus 40% with chemotherapy alone were presented at the American Society for Clinical Oncology (ASCO) meeting in late May 2009 in Orlando, Florida. Ultimately, Adherex would like to evaluate the durability of these results beyond 3 months, but the product is currently on hold due to lack of resources and
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
feature the switching of priorities to monetize more mature projects. Tyler says, “We would like to learn more about what ADH-1 is doing in the tumor that makes it more sensitive to the chemotherapy, but we’re actually very excited. This is the only drug that has really targeted the N-cadherin molecule, which is important in melanoma and some other malignancies.” One other cadherin-like molecule currently under investigation is receptor tyrosine kinase (RTK) anaplastic lymphoma kinase (ALK). BerGenBio (Bergen, Norway), founded in 2007 by researchers from the University of Bergen, is evaluating small-molecule and mAb inhibitors of RTK AXL, which has an extracellular domain comprising fibronectin III and immunoglobulin motifs similar to cadherin-type adhesion molecules. Preclinical studies carried out by BerGenBio in three-dimensional cell culture and animal models of angiogenesis, metastasis and survival have shown that RTK AXL is essential for breast cancer metastasis but not for angiogenesis. Another protein associated with EMT is autocrine motility factor (AMF), a C-X-X-C motif cytokine also known as neuroleukin. As yet, no companies have taken any therapeutic agents into the clinic that specifically target AMF, although it is known that Genentech/ Roche’s (S. San Francisco, CA, USA) humanized mAb Herceptin (trastuzumab), which primarily targets the epidermal growth factor (EGF) family receptor (EGFR) HER2, inhibits AMF expression and augments the activity of specific simple sugar inhibitors of AMF, such as erythrose 4-phosphate and d-mannose 6-phosphate7. Developmental signaling pathways An important function of E-cadherin is interaction (by means of its cytoplasmic domain) with β-catenin, which together with α-catenin, anchors the cadherin complex to the cell’s actin cytoskeleton. β-catenin comprises a multiprotein complex (including AXIN, adenomatous polyposis coli (APC) and glycogen synthase kinase–3β) that is part of the Wnt signal transduction pathway. When Wnt binds to its receptor (frizzled), disheveled (Dsh/Dvl) protein then stabilizes the β-catenin complex, which then travels to the nucleus where it triggers the expression of a range of mitogenic and EMT-related genes; in Wnt’s absence, excess cytoplasmic β-catenin is targeted for degradation in the proteosome. In malignancies, such as colon adenocarcinoma and squamous cell carcinomas, however, the Wnt pathway is upregulated. Mutations in certain components (e.g., APC or AXIN) of the β-catenin multiprotein complex become constitutively active, allowing β-catenin to
localize to the nucleus and begin transcription of proliferative and/or cell cycle genes, such as c-Myc (also known as MYC) and cyclin D1. Apart from a few nonsteroidal anti-inflammatory drugs (NSAIDs; e.g., aspirin, Indocin (indomethacin) and OSI Pharmaceuticals’ (Melville, NY, USA) Aptosyn (exisulind)) that have shown limited activity in suppressing β-catenin transcription, few compounds specifically targeting the Wnt pathway have been developed—partly because β-catenin regulation involves protein-protein interactions, which are hard to block using small molecules. Even so, in September, a team at Novartis identified two new enzymes that can regulate β-catenin stability: tankyrase TRF1interacting ankyrin-related ADP-ribose poly-
merase (TNKS) and TNKS2 (ref. 8). These kinases help break up the multiprotein complex, facilitating ubiquitin-mediated degradation of β-catenin. According to Novartis, small-molecule inhibitors of TNKS and TNKS2 are currently under development. Other developmental signaling pathways linked with EMT are the Hedgehog/ Smoothened pathway and the Notch pathway. Three companies currently have compounds targeting the Hedgehog cascade in active development. In September, Genentech reported that its orally active small-molecule GDC-0449 showed potent antitumor activity in 8 out of 18 patients in a phase 1 trial for metastatic skin cancer (basal cell carcinoma), with 1 unconfirmed partial response,
Box 4 Noncoding RNAs and metastasis One area of research yet to be fully exploited by industry is the vast group of ‘noncoding’ or regulatory RNAs that include microRNAs (miRNAs), nucleolar RNAs, piwi-interacting RNAs, medium RNAs that range from 30–300 nt and the very large RNAs that can extend up to 100 kb. Although some commercial efforts are currently targeting miRNAs, as yet very little work has been done on how to translate knowledge concerning other noncoding RNAs into therapeutic programs. For noncoding RNAs, “It will take time and resources to evaluate and find which transcripts in these groups are really functional and which ones are important in the pathogenesis of diseases, such as cancer,” says cancer biologist Fabricio Costa of Children’s Memorial Research Center at Northwestern University’s Feinberg School of Medicine. “In some cases, expression of certain noncoding RNAs can predict how aggressive and prone to metastasis a cancer might be,” he says. Examples include the MALAT-1 (metastasis-associated lung adenocarcinoma transcript-1) gene—a large (8 kb) noncoding RNA that is overexpressed in lung cancers and which has been associated with lung cancer progression23—and DD3, a large (2–4 kb depending on splice variants) noncoding RNA overexpressed in prostate cancers24. “I would love to see pharma and biotech take a look at these two in particular and this new field of research in general,” says Costa, who has been looking for possible therapeutic candidates based on noncoding RNAs. “This group could play a huge role in the management of cancer progression and metastasis,” he says. In the area of miRNAs, several candidates have already been associated with metastasis. For example, miR-200 and miR-205 inhibit the EMT-inducing transcription factors ZEB1 and ZEB2. In breast carcinoma, loss of miR-200 correlates with a decrease in E-cadherin. Acting in the opposite direction, miR-21 is upregulated in many cancers and facilitates TGFβ-induced EMT. Other miRNAs, such as miR-10b, miR-373 and miR-520c, have also been associated with progression in breast carcinoma. And in lung adenocarcinomas, miR-NLET7a2 precursor miRNA expression also correlates with poor survival. Several companies are investigating either the reconstitution of tumor-suppressive miRNAs or sequence-specific knockdown of oncogenic miRNAs by antagomirs. Regulus Therapeutics (Carlsbad, CA, USA) has described preclinical data showing that miR-296 induces angiogenesis in glioblastoma, whereas miR-451 increases the efficacy of Gleevec in the same indication. At a Keystone Symposium in January, the company also presented data that it had successfully downregulated miR-34a using an antagomir strategy in human hepatocellular carcinoma cells. The company has also studied the activity of three prometastatic miRNAs—miR-21, miR-10b and miR-122—in glioma, breast cancer and liver carcinoma, respectively. Elsewhere, Asuragen (Austin, TX, USA) subsidiary Mirna Therapeutics has reported preclinical data on miR-Rx34, which suppresses metastasis in nude mice xenograft models of NSCLC, lung cancer, prostate cancer and pancreatic cancer. The company is also developing antagomirs against miR-NLET7. GM and AM
nature biotechnology volume 28 number 3 march 2010
221
© 2010 Nature America, Inc. All rights reserved.
feature 7 patients achieving stable disease and 2 experiencing disease progression. The company is also currently testing the drug in a phase 2 trial of metastatic colon cancer together with chemotherapy plus the company’s antiVEGF mAb Avastin (bevacizumab). Elsewhere, Bristol-Myers Squibb (Princeton, NJ, USA)/ Exelixis (S. San Francisco, CA, USA) and Infinity Pharmaceuticals (Cambridge, MA, USA)/Mundipharma (Göteborg, Sweden) are carrying out phase 1 testing of small-molecule inhibitors of the Smoothened receptor in individuals with metastatic solid tumors. Aberrant Notch pathway activation is associated with brain, breast and lung cancers. Although drug development programs targeting Notch and its ligand delta-like ligand 4 (DLL4) are even less mature than those for Hedgehog, some compounds have already reached the clinic. For example, Merck (Whitehouse Station, NJ, USA) is developing MK-0752, a small molecule that targets the protease γ-secretase—an enzyme that cleaves Notch, enabling it to translocate to the nucleus and activate transcription of neoplastic and pro-angiogenic factors. Since 2005, the pharma company has initiated several phase 1 trials of MK-0752 (either singly or in combination with chemotherapy) in patients with metastatic breast cancer or refractory pediatric, central nervous system cancer. Drawbacks to inhibiting γ-secretase are the possible toxic effect on normal stem cells in the body (potentially narrowing MK-0752’s therapeutic window) and the observation that mutations in signal transduction downstream of Notch (e.g., in the phosphatase and tensin homolog (PTEN)–phosphoinositide 3 kinase– Akt pathway) may give rise to drug-resistant cells. One way around the narrow therapeutic window might be to target the Notch ligand DLL4 instead. OncoMed Pharmaceuticals (Mountain View, CA, USA), which in December 2007 struck a worldwide strategic alliance with GlaxoSmithKline for four mAbs aimed at cancer stem cells, is targeting DLL4 with a humanized mAb (OMP-21M18) that is currently in a phase 1 trial for solid tumors; if the antibody proceeds through to phase 2 trials, GlaxoSmithKline can exercise its option to take over development. Last year, in a partnership with Sanofi-aventis (Paris), Regeneron (Tarrytown, NY, USA) also initiated testing a fully human anti-DLL4 mAb (REGN421) in solid tumors. Targeting IGF and TGFβ If the Notch pathway has only recently risen on to the radar of drug developers, enthusiasm for the association of the IGF pathway with cancer signaling has waxed and waned several
222
times over the past two decades. Again, IGF is known to induce rapid internalization of E-cadherin from the membrane to the cytoplasm, inducing EMT through the nuclear factor κB (NF-κB)-Snail axis in mammary epithelial cells or by upregulating the transcription factor Zeb in prostate carcinoma cells. Pfizer and OSI Pharmaceuticals currently have the most advanced IGF-targeting programs. Pfizer’s fully human IgG2 mAb, which binds IGF-1 receptor’s (IGF-1R) extracellular domain, is currently in phase 3 trials for nonsmall cell lung cancer; OSI has initiated phase 3 testing of OSI-906, a small-molecule inhibitor of the RTK domain of IGF-1R, for patients with metastatic adrenocortical carcinoma. The same receptor is also being targeted by Merck, which has a humanized IgG1 antibody (MK4606) currently in phase 2 trials for metastatic colon cancer, and by Amgen (Thousand Oaks, CA, USA), which is combining a fully human anti-IGF-1R mAb (AMG479) with its antiEGFR human mAb Vectibix (panitumumab) or chemotherapy in a phase 2 trial for patients with advanced solid tumors. Two other companies, ImClone (New York, a wholly owned subsidiary of Eli Lilly) and Sanofi-aventis, also have anti-IGF-1R mAbs in early clinical development. One company, Silence Therapeutics (London), is even developing a small interfering RNA (siRNA) that inhibits IGF-1R for the potential treatment of prostate cancer. Another growth factor that could provide opportunities for therapeutic targeting is TGFβ, which has contrasting roles in tumor initiation and metastasis. During the early phase of tumorigenesis, TGFβ inhibits tumor epithelial cell growth by inducing cell cycle arrest and apoptosis. Paradoxically, in later stages of tumor progression—as mutations in cell cycle regulators and constitutive activation of Ras circumvent its inhibitory activity—the secreted cytokine confers a pro-metastatic effect on tumor cells and the surrounding stromal cells, impairing immune surveillance and promoting invasion (e.g., via angiopoietin-like 4 induction), angiogenesis (via VEGF induction) and migration to distant sites. When overexpressed on breast, colon, liver, lung, prostate and gastric cancers, TGFβ has been shown to promote EMT via the Smad pathway (inducing such transcription factors as Snail and Slug) and elevates the expression of many of the same proliferative/cell cycle genes as Wnt. One of the most advanced therapies against TGFβ in cancer is Antisense Pharma’s (Regensberg, Germany) trabedersen, a phosphorothioate oligodeoxynucleotide (oligo) specific for the 5′-cggcatgtcta-ttttgta-3′ sequence of the mRNA encoding the
TGFβ2 isoform. As TGFβ2 is the most highly expressed isoform in astrocytoma cells, the company is developing the drug for the treatment of malignant brain cancers, where it has to be delivered intracranially by pump. A pivotal phase 3, multicenter clinical trial for recurrent or refractory grade III anaplastic astrocytoma is currently underway to compare the 2-year survival rates and tumor responses of patients receiving trabedersen with those receiving standard chemotherapy. An even more ingenious route to the inhibition of TGFβ is being taken by NovaRx (San Diego). The company currently has brought Lucanix (belagenpumatucel-L), a therapeutic cancer vaccine (Box 5), into phase 3 trial for patients with advanced and metastatic NSCLC. Other companies less advanced in the clinic have also been pursuing TGFβ as a target, albeit with more conventional approaches. For example, Genzyme (Cambridge, MA, USA) is developing GC-1008, a human mAb against TGFβ. In August 2008, the company began a phase 1/2 dose-escalation study in individuals with advanced metastatic melanoma or renal cell carcinoma, the results of which are awaited this year. Other companies, such as Johnson & Johnson subsidiary Scios (Fremont, CA, USA) and Eli Lilly (Indianapolis) were developing small-molecule inhibitors of the TGFβ type I receptor kinase, but news of either compound has not been forthcoming since 2007, suggesting that development of these inhibitors is not a priority. Heparin-binding growth factors Heparin-binding growth factors (HBGFs), such as HGF and heparin-binding EGF-like growth factor, are another group of proteins widely implicated in metastases. One lesserknown HBGF, pleiotrophin, is known to be elevated in aggressive solid tumors (e.g., glioblastoma, melanoma and pancreatic cancer) and pleiotrophin serum levels often can be seen to drop precipitously after tumor resection. In the late 1990s, Washington, DC–based Georgetown University investigator Anton Wellstein found a receptor for pleiotrophin by screening a human cDNA phage display library against immobilized pleiotrophin. A short protein fragment was isolated that was part of the extracellular domain of the RTK ALK. “We discovered the soft spot on the ALK receptor where it binds the ligand,” says Wellstein. “And I thought hitting directly at the ligand-receptor interaction would be the best target for future drug development.” In his recent work, Wellstein has developed a high-affinity human single-chain variable fragment (scFv) antibody to target the
volume 28 number 3 march 2010 nature biotechnology
feature
© 2010 Nature America, Inc. All rights reserved.
Box 5 Treatments of last resort? The cancer field is littered with failures of cancer vaccines in advanced, metastatic cancer. Much of the reason for the high attrition rate is the fact that these treatments are often tested in patients that have undergone multiple rounds of treatment and are consequently very sick, in the end stages of disease. That said, many cancer vaccines applied to latestage cancers have not specifically targeted antigens associated with metastatic events. There are, however, some notable exceptions. The poster child for cancer vaccines is Dendreon’s Provenge (sipuleucel-T), which is indicated for individuals with metastatic hormone-resistant prostate cancer and awaits a final decision from the FDA on May 1. This patient-specific vaccine is produced by incubating a patient’s own blood, enriched for dendritic cells and other antigen-presenting cells, with a recombinant fusion protein composed of prostatic acid phosphatase and granulocyte-macrophage colony-stimulating factor. Prostatic acid phosphatase is normally present only in vanishing amounts in the blood, but often (though not exclusively) it is elevated in people with metastatic prostate cancer. Another antigen that is highly expressed in metastatic disease is melanoma-associated antigen (MAGE A3). By packaging MAGE-A3 in liposomes, GlaxoSmithKline Biologicals (Brussels) has created astuprotimut-r, which is now in phase 3 testing for patients with metastatic MAGE-A3-positive melanoma. Elsewhere, NovaRx (San Diego), is developing Lucanix (belagenpumatucel-L), a therapeutic cancer vaccine comprising DNA tumor cell lines genetically modified to express antisense DNA specific for TGFβ. Here the idea is to suppress TGFβ-mediated immune tolerance of the tumor cells in the vaccine and stimulate a more efficacious T-cell response to the cancer. In August 2008, NovaRx initiated a multicenter, randomized phase 3 trial in patients with advanced and metastatic NSCLC. AM
ligand-binding domain in ALK that blocks binding of pleiotrophin and at the same time does not appear to agonize the signal cascade that pleiotrophin clearly stimulates in tumor and stromal cells. In vivo murine studies have shown that this scFv antibody will inhibit invasion of aggressive human glioblastoma cells (U87MG) into an endothelial monolayer, which is one of the initial steps in the metastatic cascade. “It blocks the pleiotrophin effect on invasion and thus can inhibit metastasis,” says Wellstein. He hopes human trials can begin within the next 12 to 18 months. Another HBGF that induces and supports tumor progression is HGF, the natural ligand for the c-MET RTK. HGF binding causes dimerization of c-MET, leading to increases not only in the expression and/or secretion of MMP2, MMP7, MMP9 and uPA, but also in the phosphorylation and activation of multiple downstream pathways, including E-cadherin (which is ubiquinated and targeted to lysosomes), Erk1/2 (cell proliferation/differentiation), Akt (cell survival/apoptosis), protein kinase C, paxillin/focal adhesion kinase (FAK) (cytoskeletal functions involved in migration and adhesion) and retinoblastoma protein (Rb/Rb1; cell cycle). The c-MET receptor is inappropriately activated in many cancers through diverse mechanisms, including overexpression and a variety of mutations. For example, somatic mutations of c-MET are selected for during the metastatic spread of head and neck squamous carcinoma; and
although MET gene amplifications are relatively infrequent in primary tumors, they are often found in advanced solid tumors in gastric, lung or esophageal cancer. Early work in the laboratory of Toshikazu Nakamura at Osaka University identified a fragment of the HGF α-chain that contains an N-terminal hairpin domain (N-domain) and four kringle domains that inhibit HGF signaling. Several years ago, the fragment, termed NK4, was inlicensed by Kringle Pharma (Osaka, Japan). Since that time, however, no progress has been reported, and the Japanese biotech is currently seeking to outlicense commercialization rights to the peptide. Several other drugs targeting the HGF/c-MET axis are making further headway in the clinic. For example, Exelixis’ small-molecule pan-RTK inhibitor (XL-184), which blocks the ATP binding site of c-MET, is currently in late-stage trials. Preliminary results from a pivotal phase 3 test of the drug in individuals with metastatic medullary thyroid cancer were reported at January’s JP Morgan Healthcare Conference in San Francisco. According to the company, antitumor activity was observed, with tumor shrinkage (30% or more) observed in 90% of the subjects and 29% of patients experiencing a partial response. If these initial results pan out, Exelixis plans to file a new drug application for the second half of 2011. The company is currently co-developing the molecule with Bristol-Myers Squibb, which licensed exclusive worldwide rights to XL-184 in December 2008.
nature biotechnology volume 28 number 3 march 2010
Two other ATP-competitive inhibitors of c-MET currently under development are MethylGene’s (Monteal) MGCD-265 and GlaxoSmithKline’s foretinib (XL-880). MethylGene’s compound is an orally delivered N-(3-fluoro-4-(2-arylthieno[3,2-b]pyridin-7yloxy)phenyl)-2-oxo-3-phenylimidazolidine1-carboxamide that binds not only c-MET, but also VEGF receptors (VEGFR)-1, VEGFR-2 and VEGFR-3, as well as Tie 2 and Ron RTKs. Last September, MethylGene initiated a phase 2 trial for MGCD-265 in combination with Tarceva (erlotinib) or the chemotherapy Taxotere (docetaxel) in advanced metastatic NSCLC. The GlaxoSmithKline drug, which inhibits the c-MET, VEGFR2 and AXL RTKs, is also currently in a phase 2 trial for recurrent or metastatic squamous cell cancer of the head and neck and a phase 2 trial for metastatic gastric cancer. Finally, ArQule (Woburn, MA, USA; formerly Cyclis Pharmaceuticals), partner Daiichi Sankyo (Tokyo) and Asian licensee Kyowa Hakko Kirin (formerly Kyowa Hakko Kogyo) are developing ARQ-197, an orally administered small molecule that inhibits c-MET by binding outside the ATP-binding pocket of the kinase. In October 2007, the program entered phase 2 trials for microphthalmia transcription factor tumors and pancreatic cancer in the United States and Eastern Europe, respectively; and last October, a phase 2 trial in hepatocellular carcinoma was also initiated. In a recent announcement, ArQule said it expected to have data from a phase 1/2 NSCLC trial initiated in March 2008 in the first half of this year. Yet more companies are attempting a different tact: targeting the extracellular portion of the c-MET receptor rather than inhibiting its intracellular RTK activity. For example, Genentech/Roche is developing MetMAb (RG-3638), a humanized, monovalent 5D5 Fab antibody that binds the extracellular domain of c-MET, preventing HGF binding and subsequent receptor activation. In April 2009, Genentech initiated a phase 2 trial evaluating MetMAb in combination with OSI Pharmaceutical’s Tarceva for second- and third-line metastatic non–small cell lung cancer. And Amgen’s rilotumumab (AMG-102), a fully human IgG2 class anti-c-MET mAb, is also under development for the potential treatment of various types of advanced-stage cancers. Although results from a phase 2 trial in glioblastoma multiforme indicated that rilotumumab has only limited efficacy when administered as a monotherapy, the company is continuing a phase 2 study combining AMG-102 with platinum-based chemotherapy for metastatic colorectal cancer.
223
© 2010 Nature America, Inc. All rights reserved.
feature One other HBGF co-receptor, glypican-1, is being pursued as a drug target. Cancer researcher Murray Korc and his group at Dartmouth Medical School (Hanover, NH, USA) have demonstrated that glypican-1 is expressed on both epithelial tumor cells and stromal cells together and is critical for wellorganized proliferation, angiogenesis, invasion and metastasis of human pancreatic cancers. Korc has isolated endothelial cells from glypican-1 knockout mice and added VEGF in vitro. A greatly diminished mitogenic effect is observed on the endothelial cells of the glypican-1 knockouts versus the proliferative endothelial cells isolated from wild-type mice9. So, in addition to expression of glypican-1 from tumor epithelium, its presence in the tumor stromal/endothelial cell microenvironment appears to be important in making the process work. “We think it actually all fits,” says Korc. “You have overexpression of multiple ligands that are heparin binding and that activate multiple types of receptors, and you have overexpression of glypican-1, which seems to be very important.” Korc is now seeking how best to downregulate glypican-1. “We need to come up with a mouse trap that prevents glypican-1 from doing its thing,” he says. Because the glypican-1 is a large molecule and resides on cell surfaces and in the ECM as well as in connective tissues, Korc is considering monoclonal or polyclonal antibodies as potential therapies. Hitting hypoxia Because tumor masses grow faster than their blood supply, and because tumor-associated vasculature is leaky and chaotic in configuration, a low oxygen tension environment (hypoxia) is the rule rather than the exception in solid tumors. In recent years, evidence has also been accumulating that hypoxic tumor cells have an increased invasive and metastatic potential. Hypoxia stabilizes hypoxia-inducible factor (HIF1-α), which, by targeting the transcription factors Twist and Snail, induces EMT as well as changes in tumor cell metabolism (upregulating hexokinase), angiogenesis, invasion and apoptotic potential. HIF1-α also boosts levels of an enzyme, lysyl oxidase (LOX), which cross-links collagen in the ECM. High levels of LOX have been correlated with shorter metastasis-free survival and poor prognosis in head and neck cancer and estrogen receptor–negative breast cancer. In 2004, work by Amato Giaccia and postdoctoral fellow Janine Erler, both of Stanford University School of Medicine (Stanford, CA, USA), revealed that tumor cell–secreted LOX is required for hypoxia-induced invasion through increased cell-matrix interactions
224
and FAK activation, thus enabling metastatic dissemination. Continuing her research at the Institute of Cancer Research in London, Erler went on to show that LOX secreted by hypoxic tumor cells plays an important role in generating a pre-metastatic niche through recruitment of bone marrow–derived cells, which prepare tissue for the incoming metastasizing tumor cells10. The latter phenomenon could be due to LOX-mediated ECM modifications, which Erler is currently in the process of characterizing in her laboratory. Giaccia has co-founded Arresto BioSciences (Palo Alto, CA, USA), where a humanized mAb is now in preclinical development to inhibit LOX. “Our data show that if extracellular LOX is inhibited, we can inhibit the metastatic process,” says Giaccia, who prefers an antibody rather than a small molecule for targeting LOX. “If you are going after the metastatic process, you really want to go after a target that’s involved in metastasis and not a target that’s necessarily involved with primary tumor growth,” he adds. “When we target LOX, we’re seeing very little effect in the primary tumor, but instead we are seeing our effect on metastasis.” Other companies investigating the potential of hypoxia-related targets are Threshold Pharmaceuticals (Redwood City, CA, USA), Enzon Pharmaceuticals (Bridgewater, NJ, USA) and Oncothyreon (Seattle WA, USA). Researchers at Threshold have been developing the pro-drug TH302, a 2-nitroimidazoletriggered bromo analog of ifosfamide that is converted into its DNA-alkylating active form, dibromo isophosphoramide mustard, under hypoxic conditions. In January, the company presented data at the JP Morgan Healthcare Conference indicating efficacy in a phase 1/2 trial for soft tissue sarcoma: of 20 evaluable patients, 5 (25%) had partial responses, 12 (60%) had stable disease and 3 (15%) had progressive disease. The drug has also shown activity in a preclinical model of metastatic prostate cancer. Elsewhere, Enzon (in collaboration with Santaris Pharma of Horsholm, Denmark) has a locked nucleic acid antisense oligonucleotide (EZN-2968) targeting HIF1-α for the treatment of advanced cancers, such as renal cell or colorectal cancers, in phase 1 testing. And finally, Oncothyreon is carrying out phase 1 safety testing of PX-478 (the N-oxide of melphalan), a small-molecule inhibitor of HIF1-α, for advanced solid cancers. Moving out of the neighborhood Another way in which tumor epithelial cells maintain their contacts with ECM components, such as fibronectin, collagen and laminin, is through integrins. By interacting in the
cytoplasm in a complex with FAK and SRC family kinases, integrins mediate attachment to the actin cytoskeleton. In turn, calciumdependent GTPases, such as RhoC in breast cancer, are involved in the induction of cytoplasmic extensions (filopodia) in tumor cells, enabling them to migrate out of the primary cancer and into distal sites. Several companies have attempted to target integrins as a means of controlling advanced cancers. At last May’s ASCO meeting, Johnson & Johnson subsidiary Centocor Ortho Biotech (Malvern, PA, USA) announced results of a randomized, phase 2 trial of its anti-αvβ3/ αvβ5 integrin fully human mAb intetumumab (CNTO-95) in 129 individuals with metastatic melanoma with or without chemotherapy. The trial showed a trend toward prolongation of progression-free survival, overall survival and disease control and comparable activity to decarbazine chemotherapy. Two other integrin-targeted molecules, Facet Biotech’s (Redwood City, CA, USA) volociximab (chimeric IgG4 mAb targeting the AAB1 component of α5β1 integrin) and MedImmune/AstraZeneca’s (Gaithersburg, MD, USA) Abegrin (etaracizumab; a humanized anti-αvβ3 integrin) have gotten as far as phase 2 testing but have shown less impressive efficacy in single agent/chemotherapy combination trials for various metastatic solid tumors. Andrew Mazur, now of Northwestern University, was also involved in an α5β1 integrin antagonist program when at Attenuon. The drug, ATN-161, is a small peptide derived from the synergy region of fibronectin that blocks α5β1 integrin binding. Phase 2 trials in which ATN-161 was administered by intravenous infusion in renal cell carcinoma demonstrated preliminary hints of beneficial activity at the highest dose before funding was terminated. Mazur says the peptide is being licensed out along with his former company’s antibody huATN-657 to a still undisclosed pharma as an anti-metastic agent or a therapy for Crohn’s disease. As yet, industry has not announced any clinical programs specifically targeting RhoC. Complegen (Seattle) disclosed in 2008 that it has identified selective inhibitors with nanomolar affinity for RhoC using its highthroughput screen in yeast; initial testing of the compounds has also been carried out in xenograft transplantation studies for metastasis. But until now, most interest has focused on compounds that inhibit farnesylation/prenylation of RhoC and other RAS family members, either by intervening in the mevalonate pathway involved in the biosynthesis of farnesyl diphosphate (using approved HMGCoA reductase inhibitors, such as statins,
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
feature or bisphosphonates and isoprenoids) or by the development of specific prenylation inhibitors. This line of therapeutic attack, however, has also fallen out of favor. As of January, Janssen Pharmaceutica (a Johnson & Johnson subsidiary in Beerse, Belgium) and Merck/Schering-Plough, the two companies most actively pursuing prenylation inhibitors, had discontinued development of their prenylation inhibitors in advanced cancers owing to lack of efficacy and toxicity concerns. Commercial interest has also centered on targeting components further down the signal transduction pathway. Last year, Bristol-Myers Squibb received FDA approval for its Src inhibitor Sprycel (dasatinib). Sprycel, which also inhibits the BCR-ABL tyrosine kinase that interacts with FAK, is currently approved for the treatment of the hematological malignancies acute lymphoblastic leukemia and chronic myelogenous leukemia. Bristol-Myers Squibb has also been testing the compound in latestage trials for castration-resistant prostate cancer and late-stage (androgen-resistant) breast cancer that has spread to the bone. Two other companies, AstraZeneca and Kinex Pharmaceuticals (Buffalo, NY, USA), have also been developing small-molecule inhibitors of Src; as yet, however, compelling evidence of efficacy in terms of extending survival or delaying progression of advanced solid tumors remains to be shown for the latter programs. Into the circulation Once tumor cells have undergone EMT and then migrated from the primary site, they often aggregate with platelets and fibrin in the circulation, embolize in capillaries or directly attach to endothelial cells by means of integrins, P-selectin and the epithelial cell adhesion molecule (EpCAM), which through its intracellular domain associates with β-catenin, FHL2 and Lef-1, leading to the initiation of cellular proliferation programs11. One molecule present in the bloodstream that has been shown to interfere with this process is heparin. It is thought to act in two distinct ways: first, heparin inhibits endoglycosidase heparanase, an enzyme that is often secreted from cells, such as stimulated platelets and leukocytes, aiding the degradation of ECM heparin sulfate and facilitating extravasation across the vascular endothelium to sites of inflammation; and second, heparin binds to P-selectin on endothelial cells, which is used by metastasizing tumor cells to anchor on to the vascular wall. No commercial programs have yet focused on heparin’s ability to block P-selectin, but reports have appeared describing the ability of 2,6-O-disulfated dermatan sulfate to inhibit
platelet–tumor cell association in vivo and suppress metastatic growth in mouse models12. One company seeking to take advantage of heparin’s inhibitory effect on heparanase is Global TransBiotech (Monterey Park, CA, USA). Last year, the company in-licensed PI-88 (phosphomannopentaose), an oligosaccharide inhibitor of heparanase from Progen (Toowong, Australia), which had previously been developing the drug for hepatocelllular carcinoma. Global TransBiotech is focusing on further development and registration of the PI-88 in Taiwan, China, Hong Kong and Singapore; last September, it announced that it had completed a phase 2 trial of the drug in metastatic melanoma, with registration trials expected this year. Another heparin mimetic is the smallmolecule suramin, a polysulfated naphylurea currently under investigation by Optimum Therapeutics (Columbus, OH, USA). Suramin has been found to enhance the activity of chemotherapeutic Taxotere more than tenfold in human NSCLC xenograft models. The company is currently taking suramin into phase 1/2 trial in patients with second- or third-line NSCLC tumors. Unlike the heparin-mimetics, therapeutic approaches to address EpCAM have proceeded considerably further in the clinic. In 1995, Centocor (in partnership with GlaxoSmithKline) developed and launched in Germany the anti-EpCAM murine mAb Adjuqual (edrecolomab) for the treatment of residual colorectal cancer after surgery. Although this drug was withdrawn in 2001, last year, Fresnius Biotech (Hamburg, Germany, in collaboration with TRION Pharma of Munich) received marketing approval from the European Medicines Agency (EMEA; London) for another anti-EpCAM molecule. Removab (catumaxomab) is a bispecific rat/ mouse hybrid antibody comprising a Fab fragment of a mouse IgG2a antibody specific to human EpCAM and a Fab fragment of a rat IgG2b antibody specific to human CD3 on T cells, indicated for intraperitoneal treatment of malignant ascites (excess fluid in the peritoneal cavity) in EpCAM-positive carcinoma patients. The mAb recognizes EpCAM-positive tumor cells and CD3+ T-cells by means of its bispecific binding arms and also binds Fcγ receptor I/III-positive immune cells through its Fc domain. Three other anti-EpCAM agents are also in clinical development, including a fully human IgG1 mAb (adecatumumab), a recombinant fusion protein comprising human anti-EpCAM mAb linked to interleukin (IL)-2 (tucotuzumab celmoleukin) and a humanized anti-EpCAM scFv fused to recombinant Pseudomonas exotoxin (VB4-845) sponsored by
nature biotechnology volume 28 number 3 march 2010
MicroMet (Bethesda, MD, USA), Merck Serono (Darmstadt, Germany) and Viventia Biotech (Mississauga, ON, Canada), respectively. Elsewhere, Galapagos (Mechelen, Belgium) has also been working on a small-molecule RGD integrin receptor modulator, now designated GLPG0187, which it acquired from ProSkelia (Paris, a spinout from Sanofi-aventis’ skeletal disease group) in 2006. In January (2010), Galapagos announced that a phase 1 trial with GLPG0187 had been completed in healthy volunteers and that the proposed disease indication would be expanded from bone metastasis to a broader range of metastatic cancers. The company says the product demonstrated biological activity as indicated by certain biomarkers in healthy subjects as well as a good safety profile. A new phase 1 trial will begin later this year in cancer patients. According to Galapagos senior vice president for drug discovery Graham Dixon, the company’s data from murine models suggest that GLPG0187 may not only prevent tumor cells from gaining ingress into bone tissue but also prevent osteoclasts from attaching to bone matrices, where they would ordinarily begin osteolytic activity to set up a metastatic niche. Dixon says the drug has been engineered to bind to a range of integrins in the RGD family, which may account for other opportune properties, such as an antiangiogenic effect in primary and distal sites as well as apoptosis of osteoclasts. As yet, the molecular components that mediate initial engraftment and by which tumor cells extravasate from the circulation remain an area of active investigation. What is clear is that of the millions of cancer cells that enter the circulation, only a few successfully engraft, and even fewer proceed to proliferate at secondary sites. In some cases, cancer cells divide within the occluded lumen of vessels until the tumor mass becomes so large it obliterates the vessel, pushing aside endothelial cells, pericytes and smooth muscle cells and invading the surrounding tissue. Countering vessel growth To proceed from a micrometastasis comprising a few tumor cells to a larger, clinically significant macrometastasis, the incipient tumor must recruit a blood supply. Macrophages and other bone marrow–derived cells recruited to the tumor by inflammatory mediators are thought to potentiate the angiogenic stimulus by expression of VEGF-A, fibroblast growth factor 2, platelet-derived growth factor (PDGF) and angiopoietins. These factors accelerate the recruitment of other immune cell effectors and proteases (e.g., MMPs and uPA) to facilitate ECM remodeling.
225
© 2010 Nature America, Inc. All rights reserved.
feature One means of interfering with this process is to downregulate pro-inflammatory signals, thereby blocking the influx of cells that ramp up the immune response. Inhibition of prostaglandin synthesis by inactivation of cyclooxygenases 1 and 2 (COX-1 and COX-2) by means of NSAIDs, such as aspirin and indomethacin, has been shown to slow the progression of colorectal cancer in animal models as well as upregulate expression of the metastasis repressor NM23 (see ‘Colonization’ below). Indeed, a phase 2 trial of Pfizer’s Celebrex (celecoxib) plus interferon-α in 20 individuals with renal cell carcinoma showed two partial responses and a minor benefit for time to progression. It has also been shown that COX-2 expression in the stroma of intestinal tumors can induce VEGF production and angiogenesis through elevated prostaglandin E2 levels. A wealth of evidence has now accrued for the pivotal role of VEGF-A (together with angiopoietins, PDGF and other factors) in pathological blood vessel growth. The prototypical VEGF inhibitor is Avastin, which is now approved for use in a host of cancers—colon, lung, metastatic breast, glioblastoma and renal cell carcinoma, mostly in conjunction with chemotherapy. As well as biologics, such as Regeneron’s Aflibercept (a recombinant decoy receptor comprising portions of VEGFR-1 and VEGFR-2), tens of small-molecule VEGF RTK inhibitors, as well as pan-RTK inhibitors, are in clinical development or registered for marketing in advanced cancers. Many of these angiogenesis inhibitors are now being tested in combination trials with chemotherapies in patients with metastasizing tumors. As it has been shown that resistance to Avastin often emerges as a result of pericyte recruitment by the tumor vasculature via PDGF receptor (PDGFR) signaling, several recently approved angiogenesis inhibitors also inhibit this pathway. For example, Bayer (Leverkusen, Germany) and Onyx (S. San Francisco, CA, USA) have launched Nexavar (sorafenib), an oral, small-molecule inhibitor of VEGFR, PDGFR and Raf for the treatment of advanced renal cell carcinoma and unresectable hepatocellular carcinoma. And Pfizer has launched Sutent (sunitinib; SU-11248), an oral small-molecule inhibitor of VEGFR2, PDGFRbeta, c-Kit, Flt3 and DDR1 (discoidin domain receptor tyrosine kinase 1) for the treatment of gastrointestinal stromal tumors after disease progression or intolerance to Novartis’ Gleevec (imatinib mesylate) and of advanced renal cell carcinoma. Programs that are in earlier stages of development include Amgen and Takeda’s project to develop an inhibitor (AMG-386) of the proangiogenic factors angiopoietin-1 (Tie-2) and
226
angiopoietin-2. A related protein, angiopoietin-like 4, has also been shown to be important for mediating extravasation of DTCs into lung tissue. Currently in phase 2 testing in recurrent and metastatic breast cancer, ovarian, peritoneal or fallopian tube tumor and renal cell carcinoma, AMG-386 is a ‘peptibody’ Fc fragment linked to a peptide that the companies plan to take into a multicenter phase 3 trial in ovarian cancer this year. Elsewhere, Pfizer subsidiary CovX (San Diego) is developing a fusion protein comprising an angiopoietin-2 binding peptide with an antibody scaffold (which is used to confer favorable pharmacokinetics to the molecule). Last September, the company initiated a phase 1/2 trial in patients with renal cell carcinoma. Yet another mode of attack—developing a drug that mimics the antiangiogenic molecule thrombospondin-1—is under development by Abbott (Deerfield, IL, USA). The company is testing a thrombospondin-1 mimetic peptide (ABT-510) that causes endothelial cells to apoptose through activation of Fas/Fas ligand (FasL) and the Src-related kinase p59 Fyn. The mimetic is currently in a phase 1 trial for solid tumors. Vascular biologist Randolph Watnick of Children’s Hospital Boston has recently identified another factor, the highly conserved glycoprotein prosaposin, which upregulates thrombospondin-1. Secretion of prosaposin into the ECM initiates p53-dependent activity of thrombospondin-1. According to Watnick, injecting prosaposin into mice with highly metastatic human cell lines results in 80% fewer metastases to the lung and none at all to lymph nodes, with overall survival increased by 30% over controls. On the other hand, when prosaposin expression is inhibited in tumor cells, much greater metastasis formation is observed13. The fact that prosaposin activates expression of p53 gives Watnick confidence that there must certainly be other antimetastatic factors in the process beyond just inhibition of new vascular formation. “Thrombospondin-1 is one of the major players,” he says, “but a lot of other things are going on and contributing to antimetastatic activity.” Indeed, Watnick’s work has established that oncogenes RAS (also known as HRAS) and c-MYC are suppressors of thrombospondin-1. The future of prosaposin as an antimetastatic therapy is unclear, but the Children’s Hospital Boston technology transfer team is actively looking to license the idea to commercial interests. Watnick envisions prosaposin or a similar agent being used for prophylaxis after tumor resections. “Now the next step is obviously to show whether or not it has any activity on existing metastases. We’re going down those roads now,” he says.
Colonization Assuming that EMT is a key process in metastasis, it follows that DTCs on arrival at a distant organ must in turn shed their mesenchymal phenotype to establish macrometastases containing rapidly proliferating epithelial tumor cells—the mesenchymal-epithelial transition. It is possible that such sites lack the EMT-inducing factors present at the primary tumor; however, currently the molecular mechanisms and stromal factors involved in signaling the mesenchymal-epithelial transition remain an area of active investigation. One set of targets where there has been slow progress is the metastasis suppressor genes (not mentioned above). These are factors that have been shown to have activity in inhibiting metastatic colonization. Suppressor gene research has been notoriously difficult to translate to the clinic, partly because of a lack of understanding as to how these genes mediate their antimetastatic effects. Of several genes that have been identified to date, two have been the focus for drug development. KiSS-1, expression of which is significantly downregulated in secondary tumors, is thought to mediate its activity by enhancing the activity of I-κB, which inhibits the binding of NF-κB to promoters of genes encoding pro-inflammatory and pro-metastatic factors. It encodes a 54 amino acid peptide, metastin, that acts as a ligand for the human placental orphan G protein–coupled receptor OT7T175. Researchers led by Tetsuya Ohtaki at Takeda Pharmaceuticals (Tokyo) reported 9 years ago that metastin attenuates pulmonary metastasis in a mouse xenograft model using a melanoma cell line, with no effect on the primary tumor, even after cancer at secondary sites has been established14. Importantly, although KiSS-1 expression becomes downregulated in certain cancers, OT7T175 continues to be expressed, suggesting that exogenous application of the peptide would have activity. Although Takeda was considering metastin as an antimetastatic drug and had commenced studies on sustained-release formulations of the peptide and screening of small-molecule compounds that mimic it, the company does not appear to have undertaken further development of the drug. More recently, Patricia Steeg and her colleagues at the National Cancer Institute (Bethesda, MD, USA) have been pursuing translational approaches involving the histidine kinase Nm23-H1. The gene encoding Nm23-H1 was the first metastasis suppressor gene discovered and its expression in human tumors often is associated with poor patient survival. As well as binding many proteins, Nm23-H1 is thought to partly mediate its
volume 28 number 3 march 2010 nature biotechnology
feature � �
Cixutumumab OSI-906 Figitumumab XL-184 MGCD-265 Foretinib MetMAb AMG102 AMG479
Cytoplasm ��
OMP21M18 REGN421
proMMPs
cis-ACCP
DLL4
MK-0752
E-cadhein
Chemokines hormones transmitters (e.g., interleukins and AMF)
Survival factors (e.g., IGF-1, HGF and pleotrophin)
RTK
GPCR PLC
Grb2/SOS G-protein
PI3K
PI-88
Akt
Heparanase
© 2010 Nature America, Inc. All rights reserved.
Plasmin
CNTO-95 Volociximab Abegrin GLPG0187
RTK
Notch
Heparan sulfate
Adenylate cyclase
NF-κB
IκB
HIF1α
PKA
Raf
EZN-2968 PX-478
Cytokines (e.g., EPC)
STAT3,5
TGF-β
Frizzled Wnt
NM23
GSK-3β CK1α Axin TNKS
APC
Novartis small molecules
β-Catenin
MAPK ERK
Patched
Mad: Max
Nucleus ERK
JNKs
Fos
Jun
Caspase 9
SMO
β-Catenin: TCF
CyclD CDK4
p16
Gli
GDC-0499 IPI-926 XL-139
p15
E2F
Gene regulation
Apoptosis
CyclE CDK2
ARF mdm2
FADD
Hedgehog
TCF
CREB
Caspase 8
BHQ-880
TNIK
Cytochrome C
SMAD4
DKK1
Dasatinib
Rb TGFR
Ep-CAM
Dishevelled
LOX
Myc: Max SMAD3 SMAD2
WX-UK1 huATN-658
FAK Src
Bcl-xL
Trabedersen Lucanix GC-1008
proUPA
uPA
MEK
MEKK JAKs
uPA
Fyn/Shc
MKK Cytokine receptor
Removab Adecatumumab Tucotuzumumab
ECM
Integrins
cdc42
Ras
PKC
Akkα
Plasminogen
Growth factors (e.g., EGF, PDGF and VEGF)
Metastin
γ secretase
MMPs
Aflibercept Avastin RTK inhibitors
Bcl-2
p53
Cell proliferation
p27
p21
Bad Mt Abnormality sensor
Fasr
Bax
Cytoplasm
Bim LOX
ABT-510
FasL
Arresto mAb
Figure 4 Potential cellular targets of drugs associated with metastasis. Drugs are shown in red. Adapted from ref. 30.
effects by phosphorylating three targets: ATP citrate lyase, aldolase C and kinase suppressor of Ras (KSR). The first substrate is involved in lipogenesis and glucose metabolism (the energetics of cancer metabolism is an emerging area for therapies), aldolase C has a role in hypoxic stress and KSR is central to efficient mitogenic signaling (through Ras, Raf, MEK and Erk). Nm23-H1 also downregulates expression of several metastasis-associated cell surface receptors and growth factors, including c-MET, Smoothened, pleiotrophin, endothelial differentiation gene product EDG2 (the receptor for serum lysophosphatidic acid) and L1 cell adhesion molecule. As yet, no treatments specifically targeting Nm23-H1 have been reported, but several NSAIDs (including aspi-
rin and indomethacin), steroid hormones and their mimetics (e.g., estradiol and medroxyprogesterone acetate (MPA)) and nuclear receptor ligands (e.g., all-trans retinoic acid) have been shown to upregulate the protein. Steeg’s group has been investigating the potential of MPA to stimulate Nm23-H1 expression in breast cancer cells; MPA combined with chemotherapy has been in phase 2 testing for advanced metastatic breast cancer. The clinical conundrum Despite improvements in surgical techniques and chemotherapy, current clinical practice leaves people who already have metastatic cancer virtually no treatment options. Oncologists now have an increasing armamentarium of
nature biotechnology volume 28 number 3 march 2010
targeted therapies at their disposal for treating primary cancers, including agents that target the EGF family of receptors (including HER2), numerous other RTK inhibitors, drugs that target intracellular signaling pathways (e.g., Akt, phosphoinositol 3 kinase, protein kinase C, mTOR and MAPK), drugs that target metabolism (e.g., hexokinase, pyruvate dehydrogenase kinase or ATP citrate lyase), cell cycle inhibitors, apoptosis inhibitors, DNA repair inhibitors (including drugs that target poly(ADP-ribose) polymerase), histone deacetylase inhibitors and drugs that target the ubiquitin system. And although some of these agents have improved the prognosis for patients diagnosed with certain types of solid tumors, they are paradoxically exacerbating
227
feature
© 2010 Nature America, Inc. All rights reserved.
Box 6 Cancer stem cell controversy One particularly troublesome property of metastatic cancers is resistance to chemotherapy, which has been attributed to drug resistance mechanisms in DTCs (cancer stem cells). Along these lines, Markus Frank and George Murphy at the Children’s Hospital and Brigham and Women’s Hospital in Boston have characterized a subgroup of melanoma cells that express the P-glycoprotein family member ABCB5 (ATP-binding cassette, sub-family B (MDR/TAP), member 5), which is a chemoresistance mediator. First cloned in the Frank laboratory, ABCB5 operates, at least in part, as an Adriamycin efflux transporter, and based on that discovery Frank shifted his research into the area of tumor stem cell biology25. His interest in ABCB5 was tweaked by its close association with CD166 or ALCAM (activated leukocyte cell adhesion molecule), a known biomarker correlated with progression of primary melanoma, which is also expressed in mesenchymal progenitor skin cells of humans and mice26. It was with these underpinnings that Frank and his colleagues proposed that ABCB5 might be a marker identifying melanoma and perhaps other DTCs. In a xenograft-transplanted mouse model of human melanoma, Frank and Murphy established that the relative intensity of ABCB5 expression was threefold greater in lymph node metastases, compared to thin primary melanomas. But in benign melanocytic nevi, ABCB5 showed close to zero expression. When they targeted ABCB5-positive cells with an anti-ABCB5 mAb, they found that tumor formation and growth were significantly inhibited, versus controls where tumor formation and growth occurred in 100% of the subjects. Although these presumed stem cells account for a distinct minority in tumor tissue, they appear to drive cancer progression. “What’s really astonishing is that by targeting 10% of the cells in an established melanoma xenograft, the tumor growth is halted,” says Frank. However, the group has demonstrated that not every ABCB5-positive cell is a melanoma initiator, but the result of their work in targeting the antigen suggests to them that it is much more than a biomarker. “I would say that ABCB5 is quite intriguing as a chemoresistance mediator which identifies the melanoma stem cells,” says Frank, adding, “They could be the cells left behind after chemotherapy in patients.” The work has not been without controversy. In a packed hotel conference room during the 2009 International Melanoma Congress in November, organized by Massachusetts General Hospital chief
the clinical problem of metastases. Thus, whereas breast patients are now living longer on Herceptin, many more of those patients are relapsing with drug-resistant brain metastases. To make matters worse, certain mechanisms of drug resistance might also render tumors more competent for metastasis. For example, lung adenocarcinomas resistant to EGF RTK inhibitors often acquire amplifications of c-MET, with prometastatic consequences. The current vogue in oncology is to combine targeted agents with each other or with cytotoxic chemotherapies, such as Taxol (paclitaxel) and Adriamycin (doxorubicin), for the treatment of advanced and metastatic cancers. This approach is predicated on accumulating evidence that redundancy in cancer signaling
228
of dermatology David Fisher, a session was specifically dedicated to the issue of ABCB5-positive cells and their potential as cancer stem cells. On the program were Frank and University of Michigan (Ann Arbor) researcher Sean Morrison, who has conducted research with Frank’s antibody and believes his data show that that there is no significant difference between ABCB-positive and ABCB-negative cells in their ability to initiate tumor formation in mice. Moreover, Morrison finds that cells with tumor-initiating capabilities are rather common—as many as one in four cells taken from human primary and metastatic melanoma tumors compared to Frank’s observation that as few as 1 in 1,090,000 human metastatic melanoma cells formed tumors within 8 weeks of transplantation into mice27. But specific technical differences exist between the work done by the two researchers. Frank used nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice, whereas Morrison used a much more severely immunocompromised mouse (NOD/SCID interleukin-2 receptor gamma chain null (Il2rg2/2)). Frank argues that, of course, it is easier to initiate tumors in mice with virtually no immune system. However, because cancer patients typically have some immune capabilities, it may not be valid to use a mouse that is so profoundly immunosuppressed. Morrison counters that the nature of the immunity in partially suppressed immune mice that are receiving a cross-species graft is completely different from tumor-specific immunity in a human patient who is recognizing and mounting a response to his own tumor. Furthermore, Frank takes issue with the Morrison method of injecting the human tumor cells mixed with a gel medium containing growth factors and nutrients (Matrigel) that could reinforce the tumor cells’ vitality. Morrison responds that no matter what is done to the cells, if they do not have the genome of a tumor-initiating cell, they will not engraft and form a tumor mass. Fisher, who is not working with ABCB5, says “I would say the pendulum was swinging much more in somewhat of a skeptical direction about the question of whether there is a stem cell at all and secondly, whether ABCB5 would be the marker of that stem cell if it exists.” But Fisher has raised another issue that could have farreaching ramifications in translational research. “I think that people are finding some reason to be concerned that a mouse with some intact immunity is going to be informing responses in patients better than a mouse with less intact immunity,” he says. GM
pathways (e.g., between EGF receptor and KRAS or Notch and Akt) often enables cancer cells to circumvent inhibition and become resistant to single, targeted agents. And yet, all too often, even combinations of drugs are proving unsuccessful or merely capable of extending survival by a matter of months. One part of the problem is the inherent drug resistance of many advanced cancers. Many tumor cells that have undergone EMT to acquire progenitor-like qualities very often become resistant to targeted therapies that inhibit mitogenic activity or evolve such mechanisms as chemoresistance regulators (Box 6). And as anyone with cancer likely has a large number of DTCs, only a small number of which are required to regenerate a tumor,
to be effective any drug must eliminate nearly all these cells to avoid relapse. Thus, even if DTCs are as susceptible to therapy as epithelial tumor cells in the primary tumor, the survival of just a few cells might still lead to relapse. A second problem is that cells in secondary cancers are a moving target (and a different moving target from cells in the primary tumor epithelium). Genomic instability in DTCs, together with the selective pressure mounted by a different set of signals in the surrounding stroma of a secondary organ, results in tumor cells that accumulate genetic lesions that diverge from those present in tumor cells at the primary site. According to Paul Workman of the UK’s Institute of Cancer Research, “the full benefits of the [combination] approach
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
feature will only come to fruition when we can really apply genetic stratification and pathway-activation profiling,” together with a comprehensive systems biology perspective. To account for cancer dissemination, targeted treatments will have to take into account not only the primary tumor cell profiling but also the profiling of DTCs in remote sites. This is an area that is still in its infancy, and compared with primary tumors, much more effort is required to collect and characterize material from clinical biopsies of metastatic lesions—not a straightforward task in itself as such lesions can be hard to locate and access in the body. If researchers and clinicians ultimately hope to treat or prevent metastasis, several issues will have to be addressed. One significant dilemma in the development of metastasis inhibitors is that investigators are currently handicapped from the very beginning because safety-conscious regulators have by and large required potential cancer drugs, especially new chemical entities, to be tested in patients only during late-stage disease as a last resort after several therapies have failed and where safety is not the all-consuming issue anymore. “That’s really a difficult problem to tackle,” says Reuven Reich. “The moment you have metastases, a metastasis inhibitor would not help at all. You would have to do clinical trials at the very early stages, not at the end.” Another issue is the continuing emphasis in clinical trial design of complete and partial response (of the primary tumor) to cancer drugs as the main endpoints for human testing. As evident from clinical experience, the relationship between degree of shrinkage of a primary tumor and permanence of effect and cancer patient survival is weak at best. Clearly, the endpoints of greater relevance for metastasis treatments are progression-free survival (time from patient randomization until objective tumor progression or death) and time to progression (time from when patient was randomized to tumor progression). A key question in designing treatments is where in the metastatic process—EMT, migration from tumor, intravasation, dissemination through the circulation, embolism and adhesion, extravasation into a remote organ as a micrometastasis or colonization as a macrometastasis—are the most effective points for therapeutic intervention. The fact that cancers in most patients have already metastasized at the time of initial diagnosis suggests that the
early stages of the process might not be effectively treated (unless biomarkers can be identified that are accessible to diagnosis; see Box 3). And as many DTCs appear to disseminate but never blossom into a full-blown macrometastasis, agents that target colonization at a distal site might represent a good starting point for drug development. In addition, therapies that interfere with the local autocrine and paracrine signaling mechanisms in organs that keep dormant micrometastases alive could also prove useful. Because of the nature of DTCs and the ability of a very few cells to go on and seed a metastasis, administration would probably have to be carried out over a lifetime after initial cancer diagnosis, much in the same way as those at risk for cardiovascular disease take statins and antihypertensive drugs. The necessity of early drug use and potential lifetime use combine to make long-term drug safety trials in some ways pivotal for potential metastasis inhibitors. “Unfortunately, that’s the kind of clinical trial companies will never do,” says Northwestern University entrepreneur-inresidence Mazar, who spent more than 7 years at Abbott Laboratories, where he was involved in the preclinical development of two biologic compounds that would ultimately complete phase 3 trials. “It would be a long trial. Phase 3 could be many years, and without having some indication that this thing’s really going to work, they would be reluctant to go into it,” he says. On the other hand, one might reasonably think that if metastatic cancer is lethal, then industry would be attracted to the field because first-to-market players are typically the big winners. One might also think that like statins, everyday use of antimetastatic agents would amount to huge annuities for pharmas. “Pharma has not been as interested in the process of dissemination as they have the primary tumor field,” says cancer biologist Bruce Zetter of Children’s Hospital Boston, who has been founder of two cancer-related companies (Box 3). “Half of cancer patients already have metastases in place at the time of diagnosis, which makes the solution to the problem very difficult,” he says. “And the other half is cured by surgery. So you don’t have to treat them.” For all these reasons, it seems that the agents against metastatic cancer most likely to be developed by industry are those with at least some efficacy in treating epithelial tumor cells
nature biotechnology volume 28 number 3 march 2010
in the primary cancer. There is no doubt that cancer biologists are unearthing targets that do have overlapping roles in tumor initiation and/or proliferation as well as metastasis. And there is also little doubt that the study of metastasis as a process is becoming more manageable. “If you had asked me that 10 or 20 years ago, I would have said it’s impossibly complex and well beyond our capabilities to attack and to understand the process in a conceptually simple fashion,” says Weinberg of MIT. “But it’s not complex beyond measure anymore. One can begin to understand it in terms of a finite number of signaling pathways. It’s something that’s within reach of current experimental techniques, and I think that within the next 5 to 10 years, we should have a very clear and detailed understanding of how metastasis occurs.” 1. Klein, C. Nat. Rev. Cancer 9, 302–312 (2009). 2. Nguyen, D.X., Bos, P.D. & Massague, J. Nat. Rev. Cancer 9, 274–284 (2009). 3. Hoffman, A., Breuer, E. & Reich, R., J. Med. Chem. 51, 1406–1414 (2008). 4. Rudland, P.S. et al. Cancer Res. 60, 1595–1603 (2000). 5. Heiss, M.M. et al. Nat. Med. 1, 1035–1039 (1995). 6. Nakajima, S. et al. Clin. Cancer Res. 10, 4125–4133 (2004). 7. Talukder, A.H. et al. Clin. Cancer Res. 8, 3285–3289 (2002). 8. Huang, S. et al. Nature 461, 614–620 (2009). 9. Korc, M. et al. J. Clin. Invest. 118, 89–99 (2007). 10. Erler, J.T. et al. Cancer Cell 15, 35–44 (2009). 11. Maetzel, D. et al. Nat. Cell Biol. 11, 162–171 (2009). 12. Yamada, S. & Sugahara, K. Curr. Drug Discov. Technol. 5, 289–301 (2008). 13. Kang, S.Y. et al. Proc. Natl. Acad. Sci. USA 106, 12115–12120 (2009). 14. Ohtaki, T. et al. Nature 411, 613–617 (2001). 15. Fernández, C.A. et al. J. Urol. 182, 2188–2194 (2009). 16. Yang, J. et al. Proc. Natl. Acad. Sci. USA 106, 3913– 3918 (2009). 17. Nagrath, S. et al. Nature 450, 1235–1239 (2007). 18. Chen, E.X. et al. J. Am. Med. Assoc. 292, 2150–2151 (2004). 19. Fidler, I.J. Cancer Res. 50, 6130–6138 (1990). 20. Hoffman, R.M. Invest. New Drugs 17, 343–359 (1999). 21. Chishima, T. et al. Proc. Natl. Acad. Sci. USA 94, 11573–11576 (1997). 22. Zijlsta, A. et al. Cancer Cell 13, 221–234 (2008). 23. Costa, F.F. Gene 357, 83–94 (2005). 24. Costa, F.F. Gene 386, 1–10 (2007). 25. Frank, N.Y. et al. J. Biol. Chem. 278, 47156–47165 (2003). 26. Frank, N.Y. et al. Cancer Res. 65, 4320–4333 (2005). 27. Quintana, E. et al. Nature 456, 593–598 (2008). 28. Psaila, B. & Lyden, D. Nat. Rev. Cancer 9, 285–293 (2009). 29. Katz, M.H. et al. Cancer Res. 64, 1828–1833 (2004). 30. Lodish, H. Molecular Cell Biology (New York, WH Freeman, 2003).
229
p at e n t s
Patenting biotech beyond the central dogma George Wu
© 2010 Nature America, Inc. All rights reserved.
Biotech inventors and patent practitioners alike need to be aware of new interpretations of what is considered patentable, and draft claims that extend beyond biological principles.
B
iotech is developing at a rapid pace worldwide. In part to address the growing demand for patent protection in the biotechnological realm, the US Court of Appeals for the Federal Circuit (CAFC) in recent years has set forth new interpretations for rules governing novelty, non-obviousness and patentable subject areas. The central dogma of molecular biology, which provides that sequential information from DNA to protein is deterministic, was first articulated by Francis Crick in 1958. By the process of transcription, a stretch of DNA for a gene dictates the sequence for its correlative mRNA, and then the same mRNA serves as a template for the translation of amino acids into a sequence of polypeptides, which form a protein. Conversely, although a polypeptide is not known to serve as a template for either RNA or DNA elongation, there is a faithful relationship between polypeptide, RNA and DNA sequences and thus, knowing a polypeptide sequence would provide, by corollary, information on the sequence for the source RNA and DNA. Thus, the central dogma reveals a predictable chain of information from DNA to protein, whereupon sequential information for any one component (for example, DNA, RNA or protein) offers determinative sequence information for the other components. Further, the deterministic predictability of the central dogma could be expanded to purified and isolated components alone, without sequential information because molecular techniques, including cloning and sequencing, are commonplace and make it possible for any person skilled in the art to determine sequence information and isolate other correlative components (for example, DNA).
George Wu is at Neural Essence, New York, New York, USA. e-mail: [email protected]
230
The CAFC clearly appreciates deterministic predictability as defined by the central dogma. Several recent decisions in this and other courts have redefined the standard by which patents may protect biotech in the United States, and an upcoming district court ruling could shake up the legal scope of patentability. Specifically, in In re Gleave1, In re Kubin2 and In re Bilski3, the CAFC sets forth a new framework defining that which is novel and non-obvious, and in particular, what is patentable in terms of methods or processes in the field of biotech. Under US patent law, an invention must, among other things, be both novel4 and non-obvious5 to be patentable; that is, the invention for which a patent is sought must not already exist and it must not be an obvious modification of an existing invention. Additionally, a patent is awarded only to an invention that falls within a patentable subject area6. Finally, and perhaps more sensationally, the case ACLU v. Myriad7 will be tried in federal court in the Southern District of New York later this year, which may well redefine whether a gene itself is patentable. In re Gleave: novelty One of the requirements for patentability is that an invention be novel (that is, it must be new compared to prior art). In In re Gleave, decided in March 2009, the CAFC held that a reference to sense oligonucleotide sequence that does not disclose antisense function can anticipate composition claims for the corresponding antisense oligonucleotide sequence, so long as a person having ordinary skill in the art (that is, a molecular biologist in this case) would recognize how to make the antisense. This case is distinguished from In re Wiggins8, a case decided by the CAFC’s predecessor, the Court of Customs and Patent Appeals, that had held that a listing, by name, of compounds having
potential or theoretical existence, does not anticipate a subsequent claim to one of the compounds. However, the CAFC noted that there is a key difference in that in Gleave, a person having ordinary skill in the art can make the invention, whereas that person in Wiggins could not, and that difference makes anticipation possible. In other words, the CAFC recognized that such a person, familiar with the central dogma, would know that sequence information intrinsic to a sense oligonucleotide sequence is deterministic for its correlative antisense oligonucleotide sequence, and with current technology in oligonucleotide synthesis, antisense oligonucleotides could be readily made. Hence, although a particular antisense oligonucleotide sequence in a patent application is novel because it has never been previously articulated or expressed, it would not be considered new for the purpose of patent protection if its correlative sense oligonucleotide sequence is already known, and a person having ordinary skill in the art could synthesize (make) it. In re Kubin: non-obviousness Another requirement for patentability is that an invention be non-obvious (that is, if a person in the art would not have known how to solve the problem at which the invention is directed by using exactly the same mechanism). In In re Kubin, decided in April 2009, the CAFC held that when a surface protein (Natural Killer cell activation–inducing ligand, or NAIL) and antibody specific to this protein were already known in prior art and given current knowledge of molecular cloning techniques, the DNA sequence correlative to this protein would be obvious and therefore not patentable. Otherwise stated, knowledge of standard molecular cloning techniques, which rely upon Crick’s central dogma, would make it possible for persons
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
pat e n t s familiar with the central dogma to determine the DNA sequence that gave rise to the known surface protein. Hence, a claim to that DNA sequence would be unpatentable. This decision reversed the CAFC’s previous holding in In re Deuel9, in which the Court had held that an invention can satisfy the requirement for non-obviousness, even if it is “obvious to try.” In In re Deuel, the CAFC asserted that if a methodology to make a biomolecule (for example, a DNA sequence) exists, a claim to the biomolecule would not necessarily be obvious, unless prior art renders the biomolecule itself obvious. The CAFC felt compelled to reverse this decision after the 2007 decision by the US Supreme Court, KSR v. Teleflex10, which stands for the premise that whether an invention is “obvious to try” for a person of ordinary skill in the art should be a strong factor in determining non-obviousness. Hence, for In re Kubin, the CAFC reasoned that cloning a gene after the correlative protein is already known would be obvious for anyone having ordinary skill in the art. Cloning, as the CAFC sees it, is not difficult and can proceed without undue experimentation. Undue experimentation may serve as another bar to obtaining a patent. Specifically, when a patent provides an inexact disclosure which would require persons skilled in the relevant art to engage in substantial experimentation to replicate the invention, it comes into conflict with 35 USC §112 (ref. 11), which requires that the patent disclosure be enabling. In a manner of speaking, the CAFC has deemed molecular cloning techniques to be so standard and common that once a protein is known, the cloning of the DNA obviously follows. What is startling about this decision for biotech is that along with In re Gleave, it expands the reaches of the central dogma in that prior disclosures without known sequence information could make an invention with sequence information obvious. Again, in this instance, a prior disclosure of a protein makes obvious a claim to its correlative DNA because it would be obvious to try to isolate the DNA given standard techniques, and by a similar extrapolation, a disclosure of any component of the DNA-protein axis would make obvious the other members of the axis given the expansive reach of current techniques (for example, cloning and sequencing) in molecular biology. In re Bilski: patentable subject areas for method/process claims The CAFC in In re Gleave and In re Kubin has addressed the standard of novelty and non-obviousness, respectively, for biotech, as
they relate to composition claims. Claims are short descriptions that define the metes and bounds of an invention in a patent. Claims must be novel and non-obvious, and they can be drafted in several ways. A claim can recite a method, which encompasses various steps to a process; enumerate a composition, which includes elements or parts of an apparatus; or describe a product by process. Now, even if an invention could be defined through method and/or process claims (as opposed to composition claims) and possibly circumvents the standard of novelty and non-obviousness raised by In re Gleave and In re Kubin, inventors and practitioners still must be cognizant of another critical case: In re Bilski. In re Bilski, decided by the CAFC in October 2008, addresses what can be considered as patentable subject matter for method/process claims. According to section 101 (ref. 6) of the US patent law, “any new and useful process, machine, manufacture or composition of matter” is eligible for patent protection. In In re Bilski, the CAFC further clarified this rule by establishing “the machine or transformation test” to determine what qualifies as a method/process claim. Specifically, a method/process claim is patentable if, (i) it is tied to a particular machine or apparatus or (ii) it transforms a particular article into a different state or thing. Any process that can be performed entirely in the human mind would not be patentable. As for the transformed articles, “a process for a chemical or physical transformation of physical objects or substances is [definitely] patent-eligible subject matter,” and in addition, a transformed article must be “physical objects or substances [or] representative of physical objects or substances” (e.g., X-ray attenuation data and the representations were of physical objects). Further, the use of a specific machine or transformation of an article must impose meaningful limits on the claim’s scope and use of such machine or transformation must not be merely an “insignificant extra-solution activity.” The Bilski decision, as it currently stands, is having a significant reverberation in the patenting world for its limitation on general business methods, as well as in the biotech world for the same limitation placed on biotechnical or medical diagnostic methods. However, clarity regarding “the machine or transformation test” remains an issue, particularly with regards to two aspects: (i) the characteristic of the machine and the nature of its association to the process that would qualify the process as patentable; and (ii) the degree to which the machine or the transformation can
nature biotechnology volume 28 number 3 march 2010
be involved in or associated with the process so as not to be considered “insignificant extrasolution activity.” Two relatively recent decisions, Classen Immunotherapies, Inc. v. Biogen IDEC12 and Prometheus Laboratories, Inc. v. Mayo Collaborative Services13, have begun to apply the test. In Classen, the decision centers on the patent eligibility of a claim to a “method of determining whether an immunization schedule affects the incidence or severity of a chronic immune-mediated disorder in a treatment group of mammals.” The method in question has a step in which one or more immunogens are given to a group of mammals and then a marker is compared between a control and a treatment group. Presumably, giving the immunogen to a group of mammals would cause an alteration in the level of a marker within a particular mammal and the level would be reflective of the effectiveness of the immunogen. Thus, transformation of an article could arguably be present and the transformed article is a physical object (immunogen). Further, it could be stated that the transformation step is an integral step in the process because otherwise, a correlation would not exist between the marker and the immunogen. On its face, this step would not seem to be an “insignificant extrasolution activity.” Nonetheless, the CAFC did not find the claim to be patentable because it was “neither tied to a particular machine or apparatus,” nor did it “transform a particular article into a different state or thing.” More recently in Prometheus, the CAFC, referring to Bilski, explained that the question for patentability centers on whether “a claim is drawn to a fundamental principle or an application of a fundamental principle,” with the latter being patentable. The patent claim concerns a method of “optimizing therapeutic efficacy” by first administering a particular drug to a subject and then using the subject’s metabolite level to adjust future drug doses. The District Court had ruled the claim to be unpatentable “because the correlations resulted from a natural body process.” However, the CAFC found that because “[t]he transformation is of the human body following administration of a drug and the various chemical and physical changes of the drug’s metabolites that enable their concentrations to be determined,” these steps in Prometheus were in essence “method of treatment” steps, which, according to the CAFC, “are always transformative when a defined group of drugs is administered to a body to alleviate the effects of an undesired condition”13. Interestingly, looking at the two aforementioned cases, the “method of treatment” steps
231
© 2010 Nature America, Inc. All rights reserved.
pat e n t s in Prometheus were considered to be transformative because a physical change of the drug’s metabolites had occurred within the human body, but similar steps in Classen, which included administering immunogens to a group of mammals resulting in a correlative change in a systemic marker, were not considered transformative. Further, in Prometheus, if the process in the human body is considered transformative, how is it considered not an “insignificant extra-solution activity” that the District Court deemed “a natural body process?” Both cases illustrate the need for further clarity for the test enunciated in Bilski, which will happen soon—In re Bilski was argued before the US Supreme Court in November 2009 and a decision is expected this Spring. Nonetheless, even though the test is still undergoing refinement, inventors and practitioners should be mindful of Bilski and purposefully draft method/process claims that as much as possible satisfy “the machine or transformation test.” Furthermore, practitioners and inventors alike must realize that a method/process claim must fundamentally extend beyond the central dogma or any other biological theories or techniques to a degree that either the claim is significantly associated with a machine or includes appreciably a transformation of a physical article. Although the CAFC ostensibly does not require technology as a precondition to patentability, the test itself suggests an underlying technological requirement. For the machine prong of the test, an obliged association of a machine clearly indicates a technological requirement. As for the transformation prong, although not necessarily expressing an overt technological requirement, it does require that a physical article be transformed, which upon closer analysis, is a definition of technology. In fact, technology is “[t]he application of knowledge to facilitate the obtaining and transformation of natural materials”14. Both prongs in essence reflect attributes of technology, and thus, the test at a minimum suggests a technological bias for patentability, if not an overt requirement. ACLU v. Myriad: patenting of human genes In contrast to Bilski, which has limited patentable subject matter in regards to methods and processes and its necessary association with a machine or a transformation of an article, an upcoming case in the Southern District of New York, colloquially known as ACLU v. Myriad7, is a full frontal assault on patenting human genes. Myriad Genetics is the exclusive licensee of several patents that cover both the gene and the polypeptide for BRCA1 and
232
BRCA2, which are linked to breast and ovarian cancer, as well as methods for screening mutations in BRCA1 and BRCA2. Hence, along with the challenge to the patenting of a gene, the suit also seeks to overturn patents on methods for screening genetic mutations for BRCA1 and BRCA2. Currently, any research or diagnostic test involving BRCA1 and BRCA2 requires the prior approval of Myriad. The American Civil Liberties Union (ACLU) and several other plaintiffs are challenging both the validity and the constitutionality of seven patents on the basis that they violate the ‘products of laws of nature limitation’ in the Patent Act7 and Article I, Section 8 of the Constitution, which grants patents to “inventors” on their “discoveries.” Moreover, the method claims in the patents are also faulted for infringing upon the First Amendment. In addition to Myriad, other defendants in the suit include the United States Patent and Trademark Office and the University of Utah. The US Supreme Court stated in Diamond v. Chakrabarty15 that “anything under the sun that is made by man” is eligible for patent protection, if it meets the other requirements of patentability such as novelty and non-obviousness. Since then, purified and isolated genes (and polypeptides) have been patentable because they are considered human-made, unlike natural human genes. Isolated genes are simply not found in nature. However, the ACLU considers that its current suit is distinguished from Diamond because the invention in Diamond concerns a genetically engineered bacteria that eats oil for use in oil spill cleanups, whereas the inventions in the Myriad patents are not engineered or changed by man, according to the ACLU. Fundamentally, the ACLU argues that the claims for BRCA1 and BRCA2, although for isolated genes, should not be considered any different from natural human genes based on the concept that isolated genes are only of interest because they have the same function and contain the same sequence information as their natural counterparts. The crux of the question is whether isolation or purification makes an isolated gene sufficiently “humanmade.” Again, the qualitative deterministic information that lies within a sequence of DNA (a gene), as expanded from the principle of the central dogma, lies at the center of the controversy because an isolated gene retains the sequence information from the natural gene. The ACLU in its suit asserts that all of Myriad’s patents violate the ‘products of nature’ limitation that the Supreme Court has read in section 101 of the Patent Act6. Moreover, according to article I, Section 8 of
the Constitution, which authorizes Congress to grant patents to “inventors” on their “discoveries,” the ACLU maintains that isolated genes should not be considered discoveries. Also remarkably, the patents are being challenged for violating the First Amendment, because the ACLU contends the Myriad patents “limit scientific research, learning and the free flow of information.” The First Amendment bars the ownership of a particular idea or thought, and the ACLU contends that some of the claims in the Myriad patents, particularly the method claims that compare mutations of BRCA1 with a wild-type BRCA1 are merely abstract ideas that would block the free flow of more ideas. There are certain mutations that correlate with increased breast and ovarian cancer. The ACLU contends that this idea should not be patented. Also notably, in light of the “machine or transformation test” enunciated in Bilski, the ACLU could also argue that the Myriad method claims comprising comparison of mutations of BRCA1 lack the element of machine or transformation, and thus, fail to meet the requirement for patentability. Lastly, as an appeal to public sentiment, the ACLU argues that because of the monopoly afforded by the patents, individuals who cannot afford Myriad’s test for BRCA1 and BRCA2 would be prevented from seeking alternative tests for these genes because Myriad’s patents have effectively blocked entrepreneurial entry into the marketplace. Although the ACLU and the other plaintiffs are not likely to prevail at the district level, or at the circuit level with the CAFC, simply because Diamond v. Chakrabarty has been authoritative for approximately 30 years, anything would be possible if the case were to reach the Supreme Court. The Court could very well reassess whether an isolated, purified gene is considered “human-made” and patentable, potentially shifting radically how biotech will be protected. The future for biotech patents The overarching principle enunciated by Francis Crick in the central dogma, and other biological principles and common molecular techniques, are being considered more and more in rules governing novelty, non-obviousness and patentable subject area. As discussed above, in In re Gleave and In re Kubin, the CAFC has limited composition claims for biotech essentially by considering how biological principles and molecular techniques make putative inventions anticipated or obvious, and thus, resetting the standards for novelty and non-obviousness, respectively. Further, in In re Bilski, method/process
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
pat e n t s claims must exceed biological principles, and must include a necessary association with either a machine or a transformation of an article that is either physical or representative of a physical item. Although the topography for the patenting of biotech is still changing and much uncertainty remains in the patent world, it is apparent from the recent decisions (which collectively may act as a barometer of judicial sentiment) that patentability will depend less on biology and more on technology. This means that because biological principles, such as the central dogma, and standard biological techniques are being considered in the rules that govern novelty and non-obviousness, scientists and inventors cannot simply rely on biological novelty and non-obviousness. The weather vane of change may be pointing to the European rule, according to which an invention must have a technical character to be awarded a patent16. Although the CAFC in In re Bilski overtly rejected a technology test, three judges (Mayer, Dyk and Linn) did consider technology to be an indispensable condition for patentability. Moreover, upon further analysis, both prongs of “the machine or transformation test” suggest technical innovation. The two prongs may just be two aspects of a broad definition of technology. As such, technology is inevitably the key, and what is needed is for the courts (the CAFC or the Supreme Court) to refine and clarify
the meaning of “technology.” Of course, the definition has to be sufficiently broad so as not to exclude biological technology as evidenced by Prometheus, where the invention lies in the substantive manipulation of biological articles. ACLU v. Myriad challenges a fundamental part of patenting related to biotech, the patenting of a gene. Whether substantive change will occur as a result of the challenge by the ACLU remains to be seen. Nonetheless, it is particularly interesting to note that this suit by the ACLU follows the European Patent Office’s decision last year to limit a similar set of patents owned by Myriad in Europe to diagnostic testing, excluding the claims to the BRCA genes. If the case reaches the Supreme Court, the Court may very well find inspiration in the European decision. The challenge could be reflective of public sentiment, and if so, one wonders whether, even if the ACLU fails in this case and gene protection is maintained, a future ruling by the Supreme Court will eventually lead to substantive change. Lastly, for medical diagnostic companies like Myriad, their business interest would be protected if either claims for the gene itself or claims for genetic testing continue to be permitted. Patenting of the gene would offer protection of a greater scope, but the protection afforded by the method/process claims for the company’s diagnostic testing should suffice to protect their business interest, as the chief
nature biotechnology volume 28 number 3 march 2010
interest for the medical diagnostic companies would be the continued commercialization and marketing of their diagnostic services. However, since Bilski, Myriad should also be concerned whether their method/process claims would meet the “machine or transformation test.” And although from a commercial perspective, the allowance of their method/ process claims for medical diagnostics should protect their immediate interest, any blockage of gene patenting would ripple along the deterministic track of the central dogma pointing to future impasse in the patenting of other biological products, from proteins to stem cells, potentially altering the entire industry. COMPETING INTERESTS STATEMENT The author declares no competing financial interests. 1. In re Gleave, 560 F.3d 1331 (Fed. Cir. 2009). 2. In re Kubin, 561 F.3d 1351 (Fed. Cir. 2009). 3. In re Bilski, 545 F.3d 943 (Fed. Cir. 2008). 4. 35 USC §102. 5. 35 USC §103. 6. 35 USC §101. 7. Association for Molecular Pathology et al. v. US Patent & Trademark Office, et al. Docket No. 09CV4515. 8. In re Wiggins, 488 F.2d 538 (CCPA 1973). 9. In re Deuel, 51 F.3d 1552 (Fed. Cir. 1995). 10. KSR v. Teleflex, 550 US 398 (2007). 11. 35 USC §112. 12. Classen Immunotherapies, Inc. v. Biogen IDEC, 304 Fed. Appx. 866 (Fed. Cir. 2008). 13. Prometheus Laboratories, Inc. v. Mayo Collaborative Services, No. 2008–1403 (Fed. Cir. 2009). 14. Darvill, T. The Concise Oxford Dictionary of Archaeology. (Oxford University Press, 2002). 15. Diamond v. Chakrabarty, 447 US 303 (1980). 16. European Patent Convention. Article 52(1).
233
patents
© 2010 Nature America, Inc. All rights reserved.
Recent patent applications in DNA diagnostics Priority Publication application date date
Patent number
Description
Assignee
Inventor
WO 2008127199, EP 2146740
A method of preventing, inhibiting, arresting or reversing tumorigenesis in a cell, comprising modulating the amount or the activity of a Rab binding protein.
Agency for Science, Technology and Research (Singapore)
Lim B, Miller LD, Zhang J
4/13/2007
10/23/2008, 1/27/2010
US 20090325813
A method of determining the level of target oligonucleotide suspected of being present in a sample for diagnosing, e.g., cancer, by measuring hybridization of the target and reference oligonucleotide to a test and calibration microarray and comparing the results.
Curry BU, Wang H
Curry BU, Wang H
6/26/2008
12/31/2009
WO 2009147519
A method of determining delivery rates and/or efficiency National Center for of a small interfering RNA (siRNA), microRNA (miRNA) Research Sciences or related molecule to/in target organs or cells, compris- (Paris) ing measuring levels, in exosomes or vesicles of target organs or cells of siRNA and/or miRNA.
Gibbings D, Voinnet O
6/6/2008
12/10/2009
WO 2009143624
University of British A method of determining the acute allograft rejection Columbia (Vancouver, status of a subject by determining the nucleic acid expression profile of nucleic acid markers or proteomic BC, Canada) expression profile of proteomic markers in a biological sample from the subject.
Balshaw R, Cohen Freue G, Gunther O, Keown P, McManus B, McMaster R, Meredith A, Mui A, Ng R, Scherer A
5/30/2008
12/3/2009
WO 2009100029
A diagnostic method aiding in the detection of disease or other medical condition, comprising isolating a microvesicle fraction from a biological sample and detecting the presence or absence of a biomarker within the microvesicle fraction.
General Hospital Corp. (Boston)
Breakefield XO, Brown D, Miranda KC, Russo LM, Skog J
2/1/2008
8/13/2009
WO 2009036513
Identifying an epigenetic and genetic profile in a subject associated with a mental health condition, comprising detecting DNA methylation of the serotonin transporter (5HTT) promoter region and/or the level of expression of the 5HTT gene.
Murdoch Childrens Research Institute (Parkville, Victoria, Australia)
Craig JM, Olsson CA, Saffery R
9/21/2007
3/26/2009
WO 2009033185
A method of identifying druggable targets comprising obtaining a cell or organism with an RNA interference (RNAi) pathway, infecting the cell or organism with a virus, and assaying for a change in expression of the microRNAs in the cell or organism.
University of Massachusetts (Boston)
Kowalik TF, Stadler BM
9/6/2007
3/12/2009
WO 2008113773
A nucleic acid molecule comprising a 7SL smallRNA– derived sequence comprising at least the binding domain to srp9 and srp14 proteins of the 7SL ribonucleocomplex, a sequence identical or complementary to a target sequence and a pol III type III promoter; useful for diagnosing an agerelated pathology, e.g., Alzheimer’s disease.
Biorigen (Genoa, Italy)
Pagano A
3/16/2007
9/25/2008
WO 2008103135
A new isolated microRNA (miRNA) polynucleotide, useful for diagnosing and treating a colorectal cancer in a subject, and for screening for test agents which that affect miRNA generation.
Johns Hopkins University (Baltimore, MD, USA)
Cummins J, Kinzler KW, Velculescu V, Vogelstein B
2/16/2007
8/28/2008
WO 2008097926
A new linear double-stranded DNA template comprising an open reading frame encoding a therapeutic, prophylactic or diagnostic polypeptide or nucleic acid molecule; useful for in vitro transcription of RNA and treating or inhibiting disorders.
Yale University (New Haven, CT, USA)
Bahceci E, Komarovskaya ME, Rabinovich PM, Weissman SM
2/2/2007
8/14/2008
Source: Thomson Scientific Search Service. 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/).
234
volume 28 number 3 march 2010 nature biotechnology
news and views
Genetic therapy for spinal muscular atrophy Alex MacKenzie
© 2010 Nature America, Inc. All rights reserved.
A severe inherited neuromuscular disease is corrected in mice by intravenous gene delivery. In the world of inherited pediatric disorders, the case of spinal muscular atrophy is particularly poignant. Most afflicted infants and children, while largely neurologically and completely cognitively intact, grow progressively weaker over time, with many ultimately succumbing to respiratory failure at a young age. Thus, research in this issue by Kaspar and colleagues1 reporting a gene therapy rescue of the disease phenotype in a mouse model of spinal muscular atrophy is welcome news—all the more so given the authors’ preliminary data suggesting that the approach could work in primates. This study, combined with the possibility of disease screening in newborns, raises, for the first time, hope of real therapeutic progress against this as yet untreatable disorder. The autosomal recessive 5q spinal muscular atrophies (so called as the disease gene maps to chromosome 5q13.1) are characterized by a loss of motor neurons, resulting in weakness of all volitional muscles and often an ultimately unsustainable respiratory failure2. Although the outlook for patients with spinal muscular atrophy type I, the most common and severe form of the disease, has improved with better nutritional and particularly better respiratory care (at least in the developed world)3, it is still one of the leading inherited causes of infant mortality. The genetics of 5q spinal muscular atrophy are complex. Instead of the discrete disabling intragenic mutations that underlie most recessive disorders, a portion or, most frequently, the entirety of the survival motor neuron (SMN)-1 gene is usually homozygously deleted. The ubiquitously expressed and evolutionarily conserved SMN protein is involved with many aspects of RNA metabolism. Alex MacKenzie is at the Children’s Hospital of Eastern Ontario Research Institute, University of Ottawa, Ottawa, Ontario, Canada. e-mail: [email protected]
Spinal cord
scAAV9-SMN
SMA mouse
SMA mouse
Spinal cord
Motor neuron
Neurites die back
More robust neurite arborization
Atrophic muscle fibers
mou use Dead SMA mouse
Rescued SMA mouse
Figure 1 Gene therapy for mice with spinal muscular atrophy (SMA). SMA mice (null for the murine SMN gene and homozygous for variants of human SMN transgenes) are born with a normal motor neuron complement. However, the motor neurons undergo rapid attrition, likely a result of synaptic failure and denervation with attendant muscular atrophy. The mice become wasted and succumb at two weeks of age (left), analogous to an untreated mild human type I SMA. Injection of scAAV9-SMN into the facial vein of day-old SMA pups results in SMN expression in ~40% of motor neurons, normalization of synaptic electrophysiology and an extension of life span to >250 days, albeit at half the size of unaffected mice (right).
Unsurprisingly, in all metazoan species save humans, complete ablation of SMN is embryonically lethal. In humans, however, loss of SMN1 is offset by the presence of a variable number of copies of the human-specific gene SMN2, which makes a small amount of SMN protein, permitting initial survival of the organism but not of all motor neurons4.
nature biotechnology volume 28 number 3 march 2010
Why motor neurons are the cell type most severely affected by loss of SMN has been a central conundrum in the field. Although many aspects of the molecular pathogenesis are still unclear, recent evidence points to a synaptopathy possibly devolving from a deficient presynaptic transcriptome, resulting in denervation and early motor neuron
235
© 2010 Nature America, Inc. All rights reserved.
n e w s an d vi e w s attrition4. In distinction to the better known amyotrophic lateral sclerosis, spinal muscular atrophy seems to be a truly cell autonomous disorder: antenatal transgenic motor neuron SMN repletion is essentially curative in spinal muscular atrophy mice5 (and presumably in humans). Whether postnatal SMN restoration would have a similar benefit was unknown until the present work. The greater the SMN2 copy number—both in infants and children with spinal muscular atrophy and in mouse models—the milder the disease. This observation has made robust pharmacologic inducers of SMN2 a holy grail of translational researchers. The best result reported so far was achieved using trichostatin, a potent histone deacetylase inhibitor (a perennial drug class favorite for transcript modulation in monogenic disease treatment), combined with nutritional supplementation (also a popular choice when optimizing treatment of pediatric disease). The approach roughly tripled the 15-day lifespan of genetically engineered mice with severe spinal muscular atrophy3. This comparatively modest success made welcome a 2009 report from the Kaspar group6 describing a self-complementary adeno-associated virus (scAAV)-9 vector that crosses the blood-brain barrier after systemic administration. Because scAAV9 transduced motor neurons in neonatal mice (although not in adult mice), it seemed well suited to gene therapy for spinal muscular atrophy. The self-complementarity of the scAAV genome is a crucial aspect of its potency, halving the AAV payload (happily, SMN is only 300 amino acids) but, by virtue of the dimeric invertedrepeat genomic structure, often allowing for a marked increase in transgene expression7. Now, the Kaspar group, working with spinal muscular atrophy pioneer Arthur Burghes, details by far the most successful rescue yet reported of a genetically and physiologically faithful mouse model of severe spinal muscular atrophy1. To achieve this landmark, the authors injected scAAV9 carrying SMN1 into the facial vein of mice pups on postnatal day 1 (Fig. 1). The result: transduction of 40% of motor neurons and an extension of longevity from 2 weeks to 250 days and counting, combined with normal motor function and almost normalized neuromuscular electrophysiology. Intriguingly, the mice are roughly half the size of their wild-type siblings. The diminutive stature of humans with spinal muscular atrophy is well known; whether the small size of the mice results from rescue of less than half of the motor neurons and/or from a missing ‘extra-motor-neuron’ growth-determining role of SMN is unclear.
236
An obvious question with such work is whether it can be extended to primates. Anticipating this issue, the authors studied systemic injection of scAAV9-GFP in a cynomolgus macaque at 1 day of age. Four weeks later, the level of GFP transduction in spinal motor neurons was similar to that seen in mice, auguring well for application to humans. This is heartening news on several levels. Along with recent encouraging reports of AAV gene therapy of retinal disease, it marks the further rehabilitation of gene therapy as a credible therapeutic alternative for neurologic diseases. In addition, a significant worry confronting spinal muscular atrophy research-
“Along with recent encouraging reports of AAV gene therapy of retinal disease, (this study) marks the further rehabilitation of gene therapy as a credible therapeutic alternative for neurologic disease.” ers—given reports of ‘fetal’ disease, rapid postnatal motor neuron attrition observed even in the condition’s milder forms and the rather limited success achieved with pharmacologic SMN2 induction—was that a postnatal intervention may just be too late and that motor neurons were doomed irrespective of any rapid SMN repletion in newborns. Although extrapolation from mouse data to humans is never certain, the dramatic results reported here put pay to concerns that postnatal therapy shall be ineffective. So, in addition to the obvious issues of clinical safety and cross-species efficacy, what are the hurdles to clinical introduction of this mode of gene therapy? Three come to mind immediately. First, can sufficient quantities of GMPgrade, unrecombined AAV9-SMN be generated, especially as 1014 viral genomes are likely to be needed to treat one infant? It seems that encouraging progress has been made on this front through work by several academic and biotech groups. The accepted wisdom is that large-scale GMP AAV production is now a tractable goal8,9. Clearly, the substantial cost of a single treatment (on the order of several tens of thousands of dollars) would pale in comparison to the psychosocial and monetary costs of untreated spinal muscular atrophy. Second, might there be an immune response to the AAV capsid, effectively neutralizing its
impact? (A response to SMN itself should not be a concern given that all infants with spinal muscular atrophy have low levels of SMN protein encoded by SMN2.) This is perhaps the most imponderable and possibly insuperable of the potential barriers. It may be that we get lucky and this is not an issue; certainly, a highly efficient expression cassette permitting therapeutic levels of expression at relatively low vector doses may allow nonimmunogenic long-term expression after systemic administration. If an immune response does occur, induction of tolerance to the capsid, genetic modification of capsid antigen, transient immunosuppression, or, as recently shown, proteasome inhibition to block presentation of capsid antigen10 would all be credible approaches to circumventing the problem. Third, is the presymptomatic identification of infants with spinal muscular atrophy possible and likely to be undertaken? Kaspar and colleagues6 found that scAAV9 transduction of motor neurons is most effective in neonates, and, obviously, this is when the proverbial horse is still in the barn as regards motor neuron loss in spinal muscular atrophy. The efficacy of AAV9-SMN diminishes if mice are treated on postnatal day 5, and administration at postnatal day 10 has no effect1. Comparable metrics will have to be determined in primates, but clearly the testing, diagnosis and treatment must happen in rapid succession, in keeping with the 2 weeks that often elapse between birth and treatment in other newborn disorders now being screened. The arcane genetics of spinal muscular atrophy work in our favor in this regard. In distinction to most other mutationally heterogeneous monogenic disorders, the homozygous absence of a single SMN1 exon-7 cytosine is found in most instances of classic 5q spinal muscular atrophy. Progress in detecting this singlenucleotide polymorphism in the high-throughput fashion required for a diagnostic screen has been reported from several groups11, 12. We are now on the cusp of combining DNAbased universal newborn screening with mass spectrometry–based organic and protein analyte detection methods. The general expectation is that when the first such DNA-based screens are launched, spinal muscular atrophy will be among the disorders included. Despite these issues, we seem to have a perfect storm: a relatively common (at least in the rarefied world of orphan diseases of childhood) and at present untreatable disorder; a tractable means of presymptomatic diagnosis; and, with this report, the possibility of correcting the causative deficiency. Spinal muscular atrophy, which has been a proving ground for several advances in orphan disease research, including
volume 28 number 3 march 2010 nature biotechnology
n e w s an d vi e w s therapeutic pharmacologic modulation of gene expression, high-throughput drug discovery and large-scale government–academic collaborations, seems poised to again serve as an instructive case study—one that may actually cure infants of a fatal disorder. COMPETING INTERESTS STATEMENT The author declares no competing financial interests. 1. Foust, K.D. et al. Nat. Biotechnol. 28, 271–274 (2010). 2. Crawford, T. in Neuromuscular Disorders of Infancy, Childhood, and Adolescence: A Clinician’s Approach (Jones R.H., De Vivo, D.C. & Darras B.T., eds.) 145–
166 (Butterworth Heinemann, Philadelphia, 2003). 3. Narver, H.L. et al. Ann. Neurol. 64, 465–470 (2008). 4. Burghes, A.H. & Beattie, C.E. Nat. Rev. Neurosci. 10, 597–609 (2009). 5. Gavrilina, T.O. et al. Hum. Mol. Genet. 17, 1063–1075 (2008). 6. Foust, K.D. et al. Nat. Biotechnol. 27, 59–65 (2009). 7. McCarty, D.M. Mol. Ther. 16, 1648–1656 (2008). 8. Smith, R.H., Levy, J.R. & Kotin, R.M. Mol. Ther. 17, 1888–1896 (2009). 9. Clement, N., Knop, D.R. & Byrne, B.J. Hum. Gene Ther. 20, 796–806 (2009). 10. Finn, J.D. et al. Mol. Ther. 18, 135–142 (2009). 11. Kao, H.Y. et al. Clin. Chem. 52, 361–369 (2006). 12. Pyatt, R.E., Mihal, D.C. & Prior, T.W. Clin. Chem. 53, 1879–1885 (2007).
© 2010 Nature America, Inc. All rights reserved.
Targeting leukemia stem cells Hanna K A Mikkola, Caius G Radu & Owen N Witte Acute myeloid leukemia stem cells can be made susceptible to chemotherapy by inducing them to divide. The goal in leukemia treatment is to permanently eradicate all leukemic cells while preserving the body’s reservoir of hematopoietic stem cells. In many cases, however, a seemingly successful treatment ends in disease relapse owing to the survival of a small population of dormant leukemia stem cells that are resistant to chemotherapy. In this issue, Ishikawa and colleagues1 describe a strategy for targeting such stem cells in a mouse model of human acute myeloid leukemia. Treatment with granulocyte colonystimulating factor (G-CSF), a cytokine that induces mobilization and cell cycle entry of hematopoietic stem cells, causes the transplanted human leukemia stem cells to proliferate and makes them susceptible to the chemotherapeutic cytarabine (Fig. 1). This finding supports the hypothesis that leukemia could be eradicated at its roots by developing treatments focused on the unique properties of the rare leukemia stem cells. Although not all malignancies fit into the cancer stem cell model2, acute myeloid leukemia is one of the prototype diseases for which there is solid evidence of stem cell– like behavior3. Acute myeloid leukemia stem cells and normal hematopoietic stem cells Hanna K A Mikkola is in the Department of Molecular, Cell and Developmental Biology, Caius G Raduis is in the Department of Medical Molecular Pharmacology and Owen Witte is an investigator with the Howard Hughes Medical Institute, University of California, Los Angeles, California. e-mail:[email protected]
share many surface molecules, use common self-renewal mechanisms and are predominantly in a quiescent state. Hematopoietic stem cells must be quiescent to avoid exhaustion and to minimize the risk of oncogenic events. Similarly, as shown in studies of acute promyelocytic leukemia, a subtype of acute
“This finding supports the hypothesis that leukemia could be eradicated at its roots by developing treatments focused on the unique properties of rare leukemia stem cells.” myeloid leukemia, quiescence protects leukemia stem cells from excessive DNA damage and exhaustion, increasing their self-renewal capacity and survival4. To maintain quiescence, both hematopoietic and leukemia stem cells interact with components of the bone marrow niche. Building on their previous work5, Ishikawa and colleagues1 now show that leukemia cells residing near the endosteal surface are quiescent, in contrast to those located more centrally in the bone marrow cavity. Chemotherapy with cytarabine alone killed the bulk of the leukemic cells but left the leukemia stem cells adjacent to the endosteum intact. After treatment with G-CSF, however, these latter cells began to proliferate and became susceptible to cytarabine. Although
nature biotechnology volume 28 number 3 march 2010
the response was variable between the mice engrafted with cells from seven different patient samples, the increased sensitivity was observed in every case, causing apoptosis and an average 100-fold drop in leukemia stem cell frequency. Although the inherent toxicity of the chemotherapy made it impossible to follow the survival of the primary transplant recipients, mice that received leukemia cells from animals treated with a combination of G-CSF and cytarabine were much less likely to develop the disease and die of it than mice receiving leukemia cells from animals treated with chemotherapy alone. A similar approach has been shown in one clinical trial to improve disease-free survival6, whereas little benefit was observed in others 7,8. This difference might be explained by the fact that in the trials where no improvement was seen, the patients had a more unfavorable prognosis based on age, cytogenetic abnormalities or response to previous treatment. The study by Löwenberg et al.6 indicates that the standard risk group, which excludes individuals with unfavorable prognoses, is most likely to benefit from the therapy. Future studies will better define the subgroup of patients that will respond and whether the failure to respond relates to cytogenetic differences, acquired mechanisms of drug resistance or perhaps differences in bone marrow microenvironments. Notably, the behavior of leukemia stem cells was not assessed directly in the clinical studies. Not all individuals respond equally to G-CSF during hematopoietic stem cell mobilization, and the response is especially weak in individuals with prior chemo- or radiotherapy. Alternative approaches for hematopoietic stem cell mobilization are being investigated. For example, interferon-α induces cycling of hematopoietic stem cells, rendering them more susceptible to cytotoxic agents9. It would be interesting to determine whether such approaches would also induce cycling of leukemia stem cells in patients who respond poorly to G-CSF. In addition to ‘nonspecific’ treatments to manipulate leukemia stem cell behavior, it may be possible to affect leukemia stem cells directly by targeting molecules required for their self-renewal and interactions with the bone marrow niche. Hematopoietic stem cells are attracted to the niche by the chemokine CXCL12 (SDF1) and its receptor CXCR4 and by the adhesion molecules VCAM-VLA4 and angiopoietin-Tie2 (ref. 10). The same molecules are also involved in leukemia stem cell–niche interactions, and elevated levels of CXCR4 and VLA4 have been associated with poor response to
237
n e w s an d vi e w s therapeutic pharmacologic modulation of gene expression, high-throughput drug discovery and large-scale government–academic collaborations, seems poised to again serve as an instructive case study—one that may actually cure infants of a fatal disorder. COMPETING INTERESTS STATEMENT The author declares no competing financial interests. 1. Foust, K.D. et al. Nat. Biotechnol. 28, 271–274 (2010). 2. Crawford, T. in Neuromuscular Disorders of Infancy, Childhood, and Adolescence: A Clinician’s Approach (Jones R.H., De Vivo, D.C. & Darras B.T., eds.) 145–
166 (Butterworth Heinemann, Philadelphia, 2003). 3. Narver, H.L. et al. Ann. Neurol. 64, 465–470 (2008). 4. Burghes, A.H. & Beattie, C.E. Nat. Rev. Neurosci. 10, 597–609 (2009). 5. Gavrilina, T.O. et al. Hum. Mol. Genet. 17, 1063–1075 (2008). 6. Foust, K.D. et al. Nat. Biotechnol. 27, 59–65 (2009). 7. McCarty, D.M. Mol. Ther. 16, 1648–1656 (2008). 8. Smith, R.H., Levy, J.R. & Kotin, R.M. Mol. Ther. 17, 1888–1896 (2009). 9. Clement, N., Knop, D.R. & Byrne, B.J. Hum. Gene Ther. 20, 796–806 (2009). 10. Finn, J.D. et al. Mol. Ther. 18, 135–142 (2009). 11. Kao, H.Y. et al. Clin. Chem. 52, 361–369 (2006). 12. Pyatt, R.E., Mihal, D.C. & Prior, T.W. Clin. Chem. 53, 1879–1885 (2007).
© 2010 Nature America, Inc. All rights reserved.
Targeting leukemia stem cells Hanna K A Mikkola, Caius G Radu & Owen N Witte Acute myeloid leukemia stem cells can be made susceptible to chemotherapy by inducing them to divide. The goal in leukemia treatment is to permanently eradicate all leukemic cells while preserving the body’s reservoir of hematopoietic stem cells. In many cases, however, a seemingly successful treatment ends in disease relapse owing to the survival of a small population of dormant leukemia stem cells that are resistant to chemotherapy. In this issue, Ishikawa and colleagues1 describe a strategy for targeting such stem cells in a mouse model of human acute myeloid leukemia. Treatment with granulocyte colonystimulating factor (G-CSF), a cytokine that induces mobilization and cell cycle entry of hematopoietic stem cells, causes the transplanted human leukemia stem cells to proliferate and makes them susceptible to the chemotherapeutic cytarabine (Fig. 1). This finding supports the hypothesis that leukemia could be eradicated at its roots by developing treatments focused on the unique properties of the rare leukemia stem cells. Although not all malignancies fit into the cancer stem cell model2, acute myeloid leukemia is one of the prototype diseases for which there is solid evidence of stem cell– like behavior3. Acute myeloid leukemia stem cells and normal hematopoietic stem cells Hanna K A Mikkola is in the Department of Molecular, Cell and Developmental Biology, Caius G Raduis is in the Department of Medical Molecular Pharmacology and Owen Witte is an investigator with the Howard Hughes Medical Institute, University of California, Los Angeles, California. e-mail:[email protected]
share many surface molecules, use common self-renewal mechanisms and are predominantly in a quiescent state. Hematopoietic stem cells must be quiescent to avoid exhaustion and to minimize the risk of oncogenic events. Similarly, as shown in studies of acute promyelocytic leukemia, a subtype of acute
“This finding supports the hypothesis that leukemia could be eradicated at its roots by developing treatments focused on the unique properties of rare leukemia stem cells.” myeloid leukemia, quiescence protects leukemia stem cells from excessive DNA damage and exhaustion, increasing their self-renewal capacity and survival4. To maintain quiescence, both hematopoietic and leukemia stem cells interact with components of the bone marrow niche. Building on their previous work5, Ishikawa and colleagues1 now show that leukemia cells residing near the endosteal surface are quiescent, in contrast to those located more centrally in the bone marrow cavity. Chemotherapy with cytarabine alone killed the bulk of the leukemic cells but left the leukemia stem cells adjacent to the endosteum intact. After treatment with G-CSF, however, these latter cells began to proliferate and became susceptible to cytarabine. Although
nature biotechnology volume 28 number 3 march 2010
the response was variable between the mice engrafted with cells from seven different patient samples, the increased sensitivity was observed in every case, causing apoptosis and an average 100-fold drop in leukemia stem cell frequency. Although the inherent toxicity of the chemotherapy made it impossible to follow the survival of the primary transplant recipients, mice that received leukemia cells from animals treated with a combination of G-CSF and cytarabine were much less likely to develop the disease and die of it than mice receiving leukemia cells from animals treated with chemotherapy alone. A similar approach has been shown in one clinical trial to improve disease-free survival6, whereas little benefit was observed in others 7,8. This difference might be explained by the fact that in the trials where no improvement was seen, the patients had a more unfavorable prognosis based on age, cytogenetic abnormalities or response to previous treatment. The study by Löwenberg et al.6 indicates that the standard risk group, which excludes individuals with unfavorable prognoses, is most likely to benefit from the therapy. Future studies will better define the subgroup of patients that will respond and whether the failure to respond relates to cytogenetic differences, acquired mechanisms of drug resistance or perhaps differences in bone marrow microenvironments. Notably, the behavior of leukemia stem cells was not assessed directly in the clinical studies. Not all individuals respond equally to G-CSF during hematopoietic stem cell mobilization, and the response is especially weak in individuals with prior chemo- or radiotherapy. Alternative approaches for hematopoietic stem cell mobilization are being investigated. For example, interferon-α induces cycling of hematopoietic stem cells, rendering them more susceptible to cytotoxic agents9. It would be interesting to determine whether such approaches would also induce cycling of leukemia stem cells in patients who respond poorly to G-CSF. In addition to ‘nonspecific’ treatments to manipulate leukemia stem cell behavior, it may be possible to affect leukemia stem cells directly by targeting molecules required for their self-renewal and interactions with the bone marrow niche. Hematopoietic stem cells are attracted to the niche by the chemokine CXCL12 (SDF1) and its receptor CXCR4 and by the adhesion molecules VCAM-VLA4 and angiopoietin-Tie2 (ref. 10). The same molecules are also involved in leukemia stem cell–niche interactions, and elevated levels of CXCR4 and VLA4 have been associated with poor response to
237
n e w s an d vi e w s
Leukemia stem cell
© 2010 Nature America, Inc. All rights reserved.
Disease relapse
Chemotherapy kills all leukemia cells
Sustained remission
No pretreatment
Induction of leukemia stem cell cycling Bone marrow niche cell
Chemotherapy kills non-stem cells
Leukemia cell
Figure 1 Eradication of leukemia by induction of leukemia stem cell cycling before chemotherapy. Standard chemotherapy kills leukemia cells that are in the cell cycle, but spares leukemia stem cells that are in a quiescent state in the bone marrow endosteal niche. The inability to kill leukemia stem cells ultimately leads to relapse of the disease. Induction of cell cycle entry before chemotherapy renders leukemia stem cells susceptible to chemotherapy, facilitating their eradication and sustained disease remission. Proliferation of leukemia stem cells can be induced by cytokines such as granulocyte colony-stimulating factor and possibly by blocking leukemia stem cell–niche interactions. Colored bars indicate molecules required for leukemia stem cell–niche interactions, and green hexagons represent the chemotherapeutic drug.
chemotherapy and unfavorable prognosis in acute myeloid leukemia. Combining an inhibitor of the CXCR4–CXCL12 interaction (AMD1300) with G-CSF has been shown to potentiate hematopoietic stem cell mobilization and may be effective in inducing leukemia stem cell cycling. Leukemia stem cell–niche interactions can also be targeted by antibodies specific for antigens expressed on the surface of the stem cells, such as CD44 and CD123 (interleukin-3 receptor), which are required for homing and survival10. Using these antibodies together with G-CSF might prevent reacquisition of niche contacts. Small-molecule inhibitors may be useful for blocking key self-renewal pathways that sustain leukemia stem cells, such as Wnt and Shh signaling10. However, targeting these pathways alone might not affect the quiescent cells. By combining novel approaches for inducing leukemia stem cell cycling and eradication with indicators for predicting treatment response and relapse, it may be possible to develop effective, individualized treatment plans. At the same time, it will be important to unravel the mechanisms that leukemia stem cells may acquire to resist mobilization or the effects of cytotoxic drugs. One possibility is reduced activity of deoxycytidine kinase (dCK), the rate-limiting enzyme for deoxyribonucleoside salvage metabolism and for the activation of cytarabine and other
238
common chemotherapeutic agents11. dCK activity can be affected by mutations, genomic rearrangements or alternative splicing, and reduction of its activity has been associated with drug resistance in several cancers. dCK is also required for efficient development of lymphocyte precursors12, but its involvement in mobilization and selfrenewal of hematopoietic stem cells and leukemia stem cells should be assessed. Future studies will reveal whether modulation of the deoxyribonucleoside salvage metabolism or other resistance mechanisms might improve leukemia treatment. A key goal of modern cancer research is to minimize the toxicity of therapies. As leukemia stem cells share many properties with hematopoietic stem cells, it is crucial to assess whether a therapy affects normal hematopoiesis. Ishikawa and colleagues1 showed that treatment with G-CSF before cytarabine did not increase apoptosis of normal human hematopoietic stem cells, raising the question of whether hematopoietic stem cells and leukemia stem cells respond differentially to G-CSF. Investigation of hematopoietic stem cell cycling and function will be needed to fully evaluate the therapeutic index of the treatment. Notably, the clinical trials that added G-CSF to the treatment regimen have not reported adverse effects with G-CSF. It is important to remember that leukemia stem cells are not always a rare population
of quiescent cells. Different oncogenes influence which cells can act as leukemia stem cells, how they behave and whether they rely on niche interactions. The phenotype, frequency and functional properties of these cells may also change during disease progression. Understanding how leukemia stem cells thrive at different stages of the disease may enable development of a repertoire of therapies for teasing them out of their niches and killing them so as to prevent disease relapse. 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/. 1. Saito, Y. et al. Nat. Biotechnol. 28, 275–208 (2010). 2. Quintana, E. et al. Nature 456, 593–598 (2008). 3. Dick, J.E. Ann. NY Acad. Sci. 1044, 1–5 (2005). 4. Viale, A. et al. Nature 457, 51–56 (2009). 5. Ishikawa, F. et al. Nat. Biotechnol. 25, 1315–1321 (2007). 6. Löwenberg, B. et al. N. Engl. J. Med. 349, 743–752 (2003). 7. Estey, E.H. et al. Blood 93, 2478–2484 (1999). 8. Milligan, D.W., Wheatley, K., Littlewood, T., Craig, J.I. & Burnett, A.K. Blood 107, 4614–4622 (2006). 9. Essers, M.A. et al. Nature 458, 904–908 (2009). 10. Lane, S.W., Scadden, D.T. & Gilliland, D.G. Blood 114, 1150–1157 (2009). 11. Staub, M. & Eriksson, S. in Cancer Drug Discovery and Development: Deoxynucleoside Analogs in Cancer Therapy (ed. Peters, G.J.) 29–52 (Humana Press Inc., 2006). 12. Toy, G. et al. Proc. Natl. Acad. Sci. USA, published online, doi:10.1073/pnas.0913900107 (31 December 2009).
volume 28 number 3 march 2010 nature biotechnology
n e w s an d vi e w s
Cellular targets for influenza drugs Ji-Young Min & Kanta Subbarao
Seasonal influenza causes hundreds of thousands of deaths annually, and the potential for vastly higher mortality from an influenza pandemic remains an ever-present threat to global health. Two recent papers in Cell1,2 and two in Nature3,4 describe a significant advance in the search for new influenza drugs. The studies reveal hundreds of human genes that are required for the influenza replicative cycle, at least some of which may emerge as novel drug targets. Like almost all mammalian viruses, influenza virus has evolved strategies for exploiting host cell factors to promote its replication and suppress antiviral immune responses. Identification of these host factors would expand the number of potential drug targets far beyond the 11 proteins encoded in the viral genome. To discover and validate human genes implicated in the influenza life cycle, all four research teams1–4 transfected cultured human cells with pools of short interfering (si)RNAs designed to silence the majority of human genes and monitored the effects of knocking down individual genes on viral infectivity. Initial hits from each study were used to search databases of protein-protein interactions, allowing prediction of hostcell pathways that are likely to be needed either for the viral replicative cycle or for the immune response to viral infection. High-throughput siRNA screens of human cells have been used previously to identify host genes important for infection by the HIV5–7, hepatitis C8,9 and West Nile10 viruses. For influenza virus, however, this approach was applied only in Drosophila melanogaster cells11 (Table 1). The relevance of the 110 factors identified in the Drosophila study to infection in humans remains unclear. Because flies are not a natural host of influenza virus, the authors had to use a modified virus. Moreover, flies do not express all of the proteins required for influenza virion assembly and infectivity. The four recent studies1–4 identified new genes and pathways associated with virtually every step in the influenza virus life cycle (Fig. 1). The early steps include attachment to the appropriate receptor on the surface of Ji-Young Min and Kanta Subbarao are at the Laboratory of Infectious Diseases, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, Maryland, USA. e-mail: [email protected]
Binding to the target cell Hemagglutinin
Release Endocytosis
Zanamivir Oseltamivir
Budding
Neuraminidase Amantadine Rimantadine
Cell adhesion
GTP-binding proteins Endosome trafficking
M2
Assembly RNP formation
H+
Protein synthesis
Proton transport Fusion
Golgi Translation initiation WNT signaling Nuclear import
Proteasomal degradation Ubiquitination
Vesicular transport
Splicing mRNA synthesis cRNA synthesis
Mitochondrial metabolism
Nucleus
Replication Kim Caesar
© 2010 Nature America, Inc. All rights reserved.
High-throughput RNAi screens in human cells suggest new approaches to curb influenza virus infection.
Figure 1 Cellular targets for anti-influenza drugs in the context of the replication cycle of influenza virus. Stages of influenza A virus replication are in green. Cellular pathways shown by siRNA screens to be essential for completion of the viral replication cycle are shown in red. The influenza A virus protein hemagglutinin binds to sialylated glycoprotein receptors on the host-cell surface, and the virus enters the cell by receptor-mediated endocytosis. Following internalization and endosomal acidification, which permits fusion of the host and viral membranes by altering the conformation of hemagglutinin, viral ribonucleoproteins (RNPs; dark blue) are released in the cytoplasm. In the nucleus of infected cells, the viral RNAs are transcribed into mRNAs and replicated by the viral RNA–dependent RNA polymerase complex. The newly synthesized viral RNPs are exported into the cytoplasm and, after assembly, mature virions bud from the cell surface. Currently, the viral M2 ion channel protein and neuraminidase are the only two targets of influenza antiviral drugs (gray boxes) licensed by the US Food and Drug Administration. Adamantane drugs, which include amantadine and rimantadine, block the action of the viral M2 protein during uncoating of the virus. Zanamivir and oseltamivir target neuraminidase, which is required for release of progeny virus from the cell surface. Adapted from ref. 13.
the host cell, virus entry through endocytosis, fusion of the endosomal and viral membranes to allow uncoating of the virus particle, and transport of the viral components to the nucleus, where transcription and replication take place. The late events include export of the viral ribonucleoprotein complex and RNA into the cytoplasm, where translation and viral assembly occur, and release of progeny virions from the cell surface. Brass et al.1 used fluorescence microscopy to score the abundance of the influenza envelope glycoprotein hemagglutinin as a marker of infection with influenza A/PR/8/34 (PR8) virus in siRNA-treated osteosarcoma cells and identified 312 human genes that
nature biotechnology volume 28 number 3 march 2010
regulated susceptibility or resistance to influenza PR8 virus infection. Notably, they demonstrated that interferon-inducible transmembrane proteins inhibit an early step of influenza replication. König et al.3 infected siRNA-transfected cells with a recombinant influenza A/WSN/33 (WSN) virus in which the coding region of hemagglutinin was replaced by a luciferase reporter gene. The absence of functional hemagglutinin protein in this screen limited it to identifying genes involved in early stages of infection (viral entry, uncoating and nuclear import), as well as viral RNA transcription and translation. Late events, such as virus assembly, budding and release, could not be scored. These authors found 295
239
n e w s an d vi e w s Table 1 siRNA-based screens to identify host factors associated with influenza virus infection and replication Host cell line
Virus used for screen
Drosophila Recombinant D-Mel2 WSN
Readout
Events in viral life cycle
Networks analyzed
Pathways
Number of Number of genes screened filtered hits Ref.
Luciferase activity
Early only
NDa
NDa
13,071
110
11
Early and late
17,877 HPRDc, BINDd, Endosomal acidification, vesicular BioGRIDe trafficking, mitochondrial metabolism, RNA splicing, mRNA transport
312
1
Human U2OS
PR8b
HA expression
Human HBEC
PR8
Amount of infectious Early and late virus
BioGRIDe, BINDd, IntActf
WNT, NF-κB, MAPK and p53 signaling, 1,745g apoptosis
Human A549
Recombinant WSN
Luciferase activity
Early only
Reactomeh, BINDd, MINTi, HPRDc
19,628 Kinase-regulated signaling, ubiquitination, phosphatase activity, endosomal acidification, actin organization and function, sumoylation, autophagy, nucleocytoplasmic transport
295
3
Human A549
WSN
NP expression and luciferase activity
Early and late
STRINGj, Reactomeh
Endosome trafficking, RNA splicing, CLK1 and p27 signaling
287
4
22,843
2
© 2010 Nature America, Inc. All rights reserved.
aND, not determined. bA subset of the results were evaluated with seasonal or pandemic influenza viruses. cHuman Protein Reference Database (http://www.hprd.org/). dBiomolecular Interaction Network Database (http://www.bind.ca/). ehttp://www.thebiogrid.org/. fhttp://www.ebi.ac.uk/intact/. g1,745 candidate genes identified from yeast two-hybrid and genome-wide microarray analyses were subjected to an siRNA screen for validation. hhttp://www.reactome.org/. iMolecular INTeraction database (http://mint.bio.uniroma2.it/mint/). jhttp://string.embl.de/
CLK1, CDC-like kinase 1; HA, hemagglutinin; HBEC, human bronchial epithelial cells; NP, influenza virus nucleoprotein; PR8, A/PR/8/34; WSN, A/WSN/33
genes whose knockdown by at least two siRNAs reduced luciferase expression, and knockdown of 219 of these genes inhibited WSN virus replication in multiple-cycle growth. Some of the identified host factors were shown to be important for replication of the swine-origin 2009 pandemic H1N1 virus. Moreover, the authors demonstrated that small-molecule inhibitors of several factors, including the vacuolar ATPase and calcium/calmodulin-dependent protein kinase 2β, inhibited influenza virus replication. Karlas et al.4 used a two-step approach in which A549 cells transfected with an siRNA library were infected with the WSN virus and cells were immunostained for viral nucleoprotein protein expression as a marker of virus infection. In the second step, the culture supernatants from the A549 cells were transferred to 293T human embryonic kidney reporter cells bearing an inducible influenza virus–specific luciferase construct. This approach identified 287 host factors that affected replication of WSN virus. The authors then confirmed the involvement of a majority of these host factors in the replicative cycles of other influenza viruses, including the 2009 pandemic H1N1 virus, by showing that transfection with siRNAs targeting the host factors reduced virus titers. Notably, they provided in vivo evidence for the importance of the p27 gene by studying p27 knockout mice. In contrast to these fluorescence-based approaches1,3,4, Shapira et al.2 used a combination of yeast two-hybrid and genome-wide microarray analyses to define a physical and regulatory map of host-virus interactions comprising 1,745 candidate genes involved in influenza PR8 infection. Using siRNA technology, they then validated the roles of all 1,745 candidate genes.
240
Notwithstanding the value of these screens to identify candidate genes for further analysis using molecular biology and virology tools, it must be noted that all siRNA screens have inherent limitations. First, because siRNAs deplete only a single gene at a time, genes whose functions are redundant with other genes might not be found. Second, unannotated genes and genes that are crucial for virus infection but are lethal or toxic to the viability of host cells will probably be excluded. Lastly, the data set gathered from each screen seems to depend strongly on the conditions and readouts of the screen. Indeed, although each of the five studies1–4,11 had the same goal, they did not identify the same set of host genes; a close comparison of the five studies shows that fewer than 20 genes were recovered more than once. This discrepancy can likely be explained by differences in the cell lines, assay designs and siRNA libraries used in the four studies (Table 1). Whereas Brass et al.1 worked with an osteosarcoma cell line (U2OS) and PR8 virus, König et al.3 and Karlas et al.4 both used a human alveolar basal epithelial cell line (A549), and used recombinant and wildtype WSN viruses, respectively. The conditions for siRNA treatment and virus infection also differed between the studies. Brass et al.1 treated cells with siRNA for 72 h before infection and scored viral infectivity 12 h after infection. In contrast, König et al.3 and Karlas et al.4 treated cells with siRNA for 48 h and scored readouts 12 h after infection. Thus, the screens may have identified host genes that affect different stages of infection. The hundreds of new host factors identified by the four studies1–4 should be welcome news to drug developers. Current influenza drugs target only two proteins, and both are viral
gene products: the M2 ion channel (the target of amantadine (Symmetrel) and its derivative rimantadine) and neuraminidase (the target of zanamivir (Relenza) and oseltamivir (Tamiflu)). Unfortunately, resistance to both classes of drugs is now widespread, and new antiviral drugs are urgently needed to combat influenza epidemics and pandemics. Host-cell factors have been successfully targeted in the case of other viral infections (e.g., the CCR5 coreceptor of HIV-1), and they have the distinct advantage of lacking the high mutation rates of influenza virus genes, which have enabled the development of resistance to current drugs. Finally, it is worth noting that viruses from different families may interact with the same cellular factors with widely disparate effects. For example, CCR5 deficiency is a risk factor for early clinical manifestations of West Nile virus infection12. Although the extent to which the data from these screens will influence drug development remains to be seen, they unquestionably extend our understanding of influenza virus biology and suggest potentially fruitful leads for the rational design of broad-spectrum antiviral drugs. COMPETING INTERESTS STATEMENT The authors declare no competing financial interests. Brass, A.L. et al. Cell 139, 1243–1254 (2009). Shapira, S.D. et al. Cell 139, 1255–1267 (2009). König, R. et al. Nature 463, 813–817 (2010). Karlas, A. et al. Nature 463, 818–822 (2010). Brass, A.L. et al. Science 319, 921–926 (2008). König, R. et al. Cell 135, 49–60 (2008). Zhou, H. et al. Cell Host Microbe 4, 495–504 (2008). Cherry, S. et al. Genes Dev. 19, 445–452 (2005). Li, Q. et al. Proc. Natl. Acad. Sci. USA 106, 16410– 16415 (2009). 10. Krishnan, M.N. et al. Nature 455, 242–245 (2008). 11. Hao, L. et al. Nature 454, 890–893 (2008). 12. Lim, J.K. et al. J. Infect. Dis. 201, 178–185 (2010). 13. von Itzstein, M. Nat. Rev. Drug Discov. 6, 967–974 (2007).
1. 2. 3. 4. 5. 6. 7. 8. 9.
volume 28 number 3 march 2010 nature biotechnology
n e w s an d vi e w s
Navigating genomic maps of cancer cells Marcel P van der Brug & Claes Wahlestedt
© 2010 Nature America, Inc. All rights reserved.
What can we learn from the first genome sequences obtained from cancerous cells? The human genome is often viewed as fixed and unchanging—a perspective borne out by the availability of many complete sequences of individuals, all of which are essentially static snapshots of the human sequence. The reality is less absolute, however, with the genomes in individual cells or groups of cells changing over time, accumulating somatic mutations and rearrangements, the vast majority of which are benign. When alterations do result in malignancy, the process of gradual change can be dramatically accelerated. Four recent papers in Nature1–4 report genomic maps of different cancers and seek to identify a unifying architecture at the genome level for each malignancy. Taken together, these papers showcase the wealth and variety of data we can expect to see in abundance from future cancer genome sequencing and/or characterization efforts, including unbiased whole genome cataloging of somatic mutations, identification of larger structural rearrangements and insights into the evolutionary history of each mutation. But a remaining problem is how to translate the abundance of data into a full understanding of disease. As these large and detailed maps contain the sum total of the mutational events over the lifetime of the individual and the lifetime of the tumor, it is a considerable challenge to separate benign events from those that drive pathogenesis. Stephens et al.1 provide a detailed view of somatic rearrangements across 24 breast cancer genomes, with two key observations that merit highlighting: the increased frequency of rearrangements across coding regions and the nonrecurrent nature of most architectural changes. In another paper, Dalgliesh et al.2 examine a partial coding genome from 101 cases of clear cell renal cell carcinoma and draw attention to mutations within the epigenetic machinery. Finally, two papers by Pleasance et al.3,4 cover similar ground in small-cell lung carcinoma and melanoma, but instead examine the genome at a single nucleotide resolution. These last two Marcel P. van der Brug & Claes Wahlestedt are in the Departments of Neuroscience and Molecular Therapeutics, The Scripps Research Institute, Jupiter, Florida, USA. e-mail: [email protected]
Tumor cell population
Genome A Transcriptome A Epigenome A
Phenotype 1
Common drivers Genome B Transcriptome B Epigenome B
Phenotype 2
+ Treatment Y
Treatment X
+ Treatment Z
Figure 1 ‘The hope’. Under ideal conditions, in a future scenario, one could conceivably compare the entire genome of many clonal cell populations within an individual. With complementary functional data in hand, one may then gradually be able to define a critical set of driver mutations that define the given tumor type and proceed to ‘Treatment X’, thought to be most efficacious for this catalog of mutations. Examination of the remaining genome diversity within the individual tumors could perhaps allow one to further tailor individual treatment by, for example, assessing risk of relapse or presence of known therapeutic resistance, prompting ‘Treatment X’ to be complemented with ‘Treatment Y’ or ‘Treatment Z’, etc.
studies ably demonstrate that whole genome sequencing is on track to become as commonplace as large-scale single nucleotide polymorphism genotyping. Ideally, increased use of genome sequencing will be accompanied by a greater reliance on genome-wide sequence data in cancer diagnostics and therapy (Fig. 1). With the massive increase in genomic data, it has become possible to infer the evolutionary history of many of the mutations in cancer genomes and to construct a clonal history for a population of cells. The patterns in somatic substitution, such as the preponderance of C→T and G→A alterations common to ultraviolet light exposure, and the more general observation of the imbalance in mutation rate between transcribed and untranscribed regions, are fascinating insights from the COLO-829 melanoma cell line1. It is evident that mutagenic processes have preferential modes of action and may even be enriched in certain genomic regions, in this case, dependent on the levels of transcription and the strand being transcribed. This discovery presents us with a conundrum. If we try to delineate functional mutations among variations by searching for commonalities between the different cancers, the trend of substitutions is reproducible, whereas the majority of individual substitutions are not. This makes it very difficult to
nature biotechnology volume 28 number 3 march 2010
distinguish driver mutations (those that are causally implicated in disease) from passenger mutations (those that are not directly implicated in disease). We must still exercise caution in interpreting apparent signatures of positive selection as driving cancer development. A related caveat is the need for efficient access to these stored data. Realistically, we may expect that hundreds of thousands of genomes will be generated over the next decade. Management of these data would clearly benefit from the establishment of an industry standard that would enhance data portability between platforms. Although the structural approaches taken by Stephens et al.1 are informative ways to characterize different types of cancer, an important challenge that remains to be fully addressed is the heterogeneity of cancer cells within tumors. Potentially, each primary tumor has a reservoir of cells with a heterogenic signature that may vary in number and distinctiveness as a function of time. It is this pool of heterogeneous cells from which drugresistant clones may be drawn. Uncovering such mutational diversity is the promised first step to fine-tuning diagnosis and predicting the outcome of therapy, as well as explaining the spectrum of clinical phenotypes. All four studies agree that vastly greater numbers of cancer genomes must be examined to determine the common rearrangements that
241
result in pathology. In addition, it is likely that individuals who have secondary tumors will have subsets of cells exposed to different stromal factors and under different selective pressures than those at the primary tumor. As practical as whole genome sequencing is becoming, it may still be a considerable length of time before it is practical to profile thousands of individuals and thousands of cells from each tumor. Even with this level of resolution within each patient, it is not clear when we will capture enough variants of each tumor genome to confidently predict therapeutic resistance (or lack thereof) with a high degree of confidence. In addition to identifying driver mutations, it would also be desirable to decipher the phenotypic contribution of each mutation or combinations of mutations. In the short term, perhaps the greatest potential for understanding the catalog of somatic variants identified in these studies will come not from adding more genomic data on other cancers or on different clonal populations but from
mining complementary information, such as whole genome epigenetic and transcriptomic data. This was already undertaken in the study of clear cell renal cell carcinoma by Dalgliesh et al.2, and hints of this promise are also apparent in the recent work of Verhaak et al.5. In the latter, a correlation of copy number variation, mutational events and expression profiles allowed classification of subtypes of glioblastoma that, critically, differ in their response to radiation and drug treatments. Rather than complicating interpretation of the data sets, such approaches should help us understand pathology by unraveling the functions of driver mutations6. Certainly, the promise of a complete mutational catalog of cancer is attractive. However, the spectrum of observed mutations is likely to be enormous given that many are under very little selection pressure. As clear mechanistic insights have yet to come from these data, the paramount question for therapeutic efforts remains: what are the high-priority
Grass genomics on the wild side Grasses are, in human terms, perhaps the most economically important of all plant families.
food, feed and fuel. Genome sequences are currently available for crop members of the Ehrhartoideae (rice) and Panicoideae (sorghum and maize). Now, writing in Nature, an international consortium reports the genomic sequence of the wild grass Brachypodium distachyon—a first for a member of the Pooideae1. Many crops have unintentionally been bred to a high level of ploidy after centuries of domestication. This makes genome sequencing and genetic analysis particularly challenging for species such as bread wheat (17,000 Mb). Although it is of no intrinsic commercial importance, B. distachyon has emerged as a valuable model for BrachyTAG programme (John Innes Centre, Norwich, UK)
© 2010 Nature America, Inc. All rights reserved.
n e w s an d vi e w s
And of the twelve grass subfamilies, the Ehrhartoideae, Panicoideae and Pooideae include many of the species with the greatest potential to address humankind’s need for
242
therapeutic targets, and why are they better than our current favorites? After all, it can be argued that industry is already awash in seemingly attractive targets and early drug discovery projects. Unfortunately, until a large number of cancer genomes is sequenced and functional mutations are discovered, cancer genome sequencing will likely provide little benefit to cancer therapy or patient care. As we are now in the midst of an explosive increase in cancer genome sequencing, a case can be made for greater focus on the development of new approaches for functional analysis of the expanding mutational catalog. COMPETING INTERESTS STATEMENT The authors declare no competing financial interests. 1. Stephens, P.J. et al. Nature 462, 1005–1010 (2009). 2. Dalgliesh, G.L. et al. Nature 463, 360–363 (2010). 3. Pleasance, E.D. et al. Nature 463, 191–196 (2010). 4. Pleasance, E.D. et al. Nature 463, 184–190 (2010). 5. Verhaak, R.G. et al. Cancer Cell 17, 98–110 (2010). 6. Greenman, C. et al. Nature 446, 153–158 (2007).
understanding other grasses, in large part because of its far more streamlined genome (272 Mb). When compared with wheat (right), the much smaller stature of mature B. distachyon (center) compares favorably with that of the better-known model plant, Arabidopsis thaliana (left). This makes it easier to cultivate large numbers of lines in a relatively small space. And like A. thaliana, its rapid life cycle and amenability to genetic manipulation further enhance its experimental tractability relative to most crops. The new genome sequence enables cross-species comparison of the genomes of the three major subfamilies of grasses, generating results that should prove useful for analyzing members of the Pooideae with larger genomes, such as wheat, oats, barley,
rye and several forage grasses. And coupled with a wealth of mutant strains, detailed genetic maps, and resources for marker-assisted breeding, the B. distachyon genome sequence promises to provide valuable insights into how grasses from other subfamilies, such as the bioenergy species switchgrass and Miscanthus, might be improved to enhance their potential for sustainable fuel production. Genes in B. distachyon bear a closer resemblance to their counterparts in rice and sorghum than to the relevant genes in their more distant relative A. thaliana, suggesting that B. distachyon has a valuable part to play in accelerating progress in grass functional genomics. Craig Mak 1. The International Brachypodium Initiative Nature 463, 763–768 (2010)
volume 28 number 3 march 2010 nature biotechnology
result in pathology. In addition, it is likely that individuals who have secondary tumors will have subsets of cells exposed to different stromal factors and under different selective pressures than those at the primary tumor. As practical as whole genome sequencing is becoming, it may still be a considerable length of time before it is practical to profile thousands of individuals and thousands of cells from each tumor. Even with this level of resolution within each patient, it is not clear when we will capture enough variants of each tumor genome to confidently predict therapeutic resistance (or lack thereof) with a high degree of confidence. In addition to identifying driver mutations, it would also be desirable to decipher the phenotypic contribution of each mutation or combinations of mutations. In the short term, perhaps the greatest potential for understanding the catalog of somatic variants identified in these studies will come not from adding more genomic data on other cancers or on different clonal populations but from
mining complementary information, such as whole genome epigenetic and transcriptomic data. This was already undertaken in the study of clear cell renal cell carcinoma by Dalgliesh et al.2, and hints of this promise are also apparent in the recent work of Verhaak et al.5. In the latter, a correlation of copy number variation, mutational events and expression profiles allowed classification of subtypes of glioblastoma that, critically, differ in their response to radiation and drug treatments. Rather than complicating interpretation of the data sets, such approaches should help us understand pathology by unraveling the functions of driver mutations6. Certainly, the promise of a complete mutational catalog of cancer is attractive. However, the spectrum of observed mutations is likely to be enormous given that many are under very little selection pressure. As clear mechanistic insights have yet to come from these data, the paramount question for therapeutic efforts remains: what are the high-priority
Grass genomics on the wild side Grasses are, in human terms, perhaps the most economically important of all plant families.
food, feed and fuel. Genome sequences are currently available for crop members of the Ehrhartoideae (rice) and Panicoideae (sorghum and maize). Now, writing in Nature, an international consortium reports the genomic sequence of the wild grass Brachypodium distachyon—a first for a member of the Pooideae1. Many crops have unintentionally been bred to a high level of ploidy after centuries of domestication. This makes genome sequencing and genetic analysis particularly challenging for species such as bread wheat (17,000 Mb). Although it is of no intrinsic commercial importance, B. distachyon has emerged as a valuable model for BrachyTAG programme (John Innes Centre, Norwich, UK)
© 2010 Nature America, Inc. All rights reserved.
n e w s an d vi e w s
And of the twelve grass subfamilies, the Ehrhartoideae, Panicoideae and Pooideae include many of the species with the greatest potential to address humankind’s need for
242
therapeutic targets, and why are they better than our current favorites? After all, it can be argued that industry is already awash in seemingly attractive targets and early drug discovery projects. Unfortunately, until a large number of cancer genomes is sequenced and functional mutations are discovered, cancer genome sequencing will likely provide little benefit to cancer therapy or patient care. As we are now in the midst of an explosive increase in cancer genome sequencing, a case can be made for greater focus on the development of new approaches for functional analysis of the expanding mutational catalog. COMPETING INTERESTS STATEMENT The authors declare no competing financial interests. 1. Stephens, P.J. et al. Nature 462, 1005–1010 (2009). 2. Dalgliesh, G.L. et al. Nature 463, 360–363 (2010). 3. Pleasance, E.D. et al. Nature 463, 191–196 (2010). 4. Pleasance, E.D. et al. Nature 463, 184–190 (2010). 5. Verhaak, R.G. et al. Cancer Cell 17, 98–110 (2010). 6. Greenman, C. et al. Nature 446, 153–158 (2007).
understanding other grasses, in large part because of its far more streamlined genome (272 Mb). When compared with wheat (right), the much smaller stature of mature B. distachyon (center) compares favorably with that of the better-known model plant, Arabidopsis thaliana (left). This makes it easier to cultivate large numbers of lines in a relatively small space. And like A. thaliana, its rapid life cycle and amenability to genetic manipulation further enhance its experimental tractability relative to most crops. The new genome sequence enables cross-species comparison of the genomes of the three major subfamilies of grasses, generating results that should prove useful for analyzing members of the Pooideae with larger genomes, such as wheat, oats, barley,
rye and several forage grasses. And coupled with a wealth of mutant strains, detailed genetic maps, and resources for marker-assisted breeding, the B. distachyon genome sequence promises to provide valuable insights into how grasses from other subfamilies, such as the bioenergy species switchgrass and Miscanthus, might be improved to enhance their potential for sustainable fuel production. Genes in B. distachyon bear a closer resemblance to their counterparts in rice and sorghum than to the relevant genes in their more distant relative A. thaliana, suggesting that B. distachyon has a valuable part to play in accelerating progress in grass functional genomics. Craig Mak 1. The International Brachypodium Initiative Nature 463, 763–768 (2010)
volume 28 number 3 march 2010 nature biotechnology
researc h h i g h l i g h ts
© 2010 Nature America, Inc. All rights reserved.
Fluorescent proteins in a lifetime Traditionally, the search for ever brighter fluorescent proteins has been conducted by determining the brightness of individual cells or bacterial colonies expressing mutated versions of a parent protein. Although this approach has created many improved variants, properties other than the molecular brightness—such as expression level, maturation efficiency and colony morphology—often influence the results. Goedhart et al. screen mutant libraries for intrinsically brighter cyan fluorescent proteins by measuring the fluorescence lifetime of bacterial colonies. Fluorescence lifetime is correlated with the quantum yield of the fluorophore and can be evaluated independent of confounding factors. Starting with the brightest available cyan fluorescent protein, SCFP3A, the authors find a variant, mTurquoise, which is ~50% brighter than the commonly used mCerulean and ~30% brighter than the parent protein. The high quantum yield also makes this protein an excellent donor for Förster resonance energy transfer (FRET) experiments. The comparatively long lifetime extends the range of cyan fluorescent proteins that can be imaged simultaneously by lifetime unmixing, and the monoexponential decay of the fluorescence simplifies the quantitative analysis of FRET by lifetime imaging. (Nat. Methods 7, 137–139, 2010) ME
Avoiding vaccine-vector antibodies Despite the potential of adenovirus-5 and vaccinia virus for use as vaccine vectors, the utility of these platforms can be limited by the magnitude of the CD8+ T-cell responses elicited (vaccinia) and preexisting antibody immunity (adenovirus). Furthermore, both platforms elicit vectorspecific antibody immunity, which interferes with booster vaccinations. Flatz et al. show that replication-defective lymphocytic choriomeningitis virus (LCMV) may be an attractive alternative vaccine delivery platform. They generate replication-defective LCMV by deleting the envelope glycoprotein gene and replacing it with diverse antigens, including a foreign viral envelope protein and tumor antigens. LCMV is known to stimulate a weak neutralizing antibody response, and in this study even multiple doses of LCMV vector did not elicit inhibitory antibody levels. In mice as well as in cultured human blood cells, LCMV vectors selectively target and activate dendritic cells, thereby eliciting strong CD8+ T-cell responses. In mice, LCMV vector-induced CD8+ T cells are more effective at destroying solid tumors than those induced by recombinant adenovirus-5 or vaccinia virus, and they confer long-lived antibacterial protection as well as antiviral immunity. What’s more, LCMV vectors elicit protective antibody immunity against vaccine antigens. (Nat. Med. advance online publication, doi:10.1038/nm.2104, 7 February 2010) CM
Synthetic enhancer diversity explored Natural transcriptional enhancers discovered in organisms such as cytomegalovirus are workhorses of biological research. Elledge and colleagues Written by Markus Elsner, Laura DeFrancesco & Craig Mak
nature biotechnology volume 28 number 3 march 2010
synthesize ~50,000 artificial enhancer sequences, screen them for activity in six mammalian cell lines and discover cell type–specific enhancers. They started by synthesizing a pool of synthetic enhancers comprising all possible 10-nt motifs repeated ten times in 100-nt sequences. They then cloned each enhancer upstream of a green fluorescent protein reporter on a retroviral vector, transfected cells with the virus, recovered cells with the strongest signal by fluorescence-activated cell sorting and subsequently sequenced the inserts. In many cases, the recovered enhancers contain motifs that match known transcription factor binding sites. Notably, the enhancers recovered span a wide range of strengths up to twice as strong as the wild-type cytomegalovirus enhancer. Other enhancers are active only in specific cell lines. These results suggest that the method may be useful for generating highly optimized experimental reagents. (Proc. Nat. Acad. Sci. USA 107, 2538–2543, 2010) CM
Nanofactories disrupt bacterial communication Interfering with quorum sensing—a process involved with biofilm formation, bioluminescence and virulence—could control bacterial growth without eliciting drug resistance. Bentley and colleagues design a multifunctional protein assembly (a ‘nanofactory’) that can trigger quorum-sensing behavior in bacteria. The self-assembling nanofactory comprises four modules: a targeting module (bacteria-specific antibody), a sensing module, a synthesis module (a fusion protein of the enzymes that synthesize the quorum sensing molecule AI-2) and an assembly module (protein G, which attaches the assembly module to the targeting module). The researchers show that the nanofactory can both pick out specific bacteria from a mixture and induce quorum-sensing behavior. Finally, they show that after functionalizing Escherichia coli with a specific nanofactory, unrelated species can sense their presence owing to the production of the quorum-sensing molecule by the functionalized E. coli. Bacteriostatic nanofactories could combat specific pathogens by inducing biofilm formation before the concentration of bacteria is high enough to damage the host, potentially avoiding the emergence of drug resistance. (Nat. Nanotechnol. advance online publication, doi:10.1038/nnano.2009.457, 17 January 2010) LD
Biofuel-producing bacterial factories Complex chemical processes are now required to convert expensive plant and animal oils into fuels and chemicals. But in a tour de force of microbial engineering, Keasling and colleagues, along with scientists at LS9 (San Carlos, CA), demonstrate that it is feasible to consolidate many key reactions into a single strain of Escherichia coli that can then convert inexpensive plant biomass into fatty acid–based fuels and chemicals. They introduce or delete ~10 microbial enzymes so that a single bacterial strain liberates simple sugars from plant biomass and, when grown on glucose alone, overproduces free fatty acids and synthesizes the fatty acids into fatty acid ethyl esters (FAEE; a component of biodiesel) or medium chain fatty alcohols (which are commercially important). Although the function of each enzyme had been characterized previously and each of these steps can be accomplished by adding purified enzymes to a reaction vessel, performing the steps in a single organism has the potential to simplify processing and to reduce costs. An optimized strain produces 647 mg l–1 of FAEE when grown on 2% glucose, which is 9.4% of the theoretical yield and is within an order of magnitude of that required for commercial production. The remaining necessary improvements are conceivably achievable with existing strain and process engineering strategies. (Nature 463, 559–562, 2010) CM
243
primer
What is flux balance analysis? Jeffrey D Orth, Ines Thiele & Bernhard Ø Palsson
© 2010 Nature America, Inc. All rights reserved.
Flux balance analysis is a mathematical approach for analyzing the flow of metabolites through a metabolic network. This primer covers the theoretical basis of the approach, several practical examples and a software toolbox for performing the calculations.
F
lux balance analysis (FBA) is a widely used approach for studying biochemical networks, in particular the genome-scale metabolic network reconstructions that have been built in the past decade1-4. These network reconstructions contain all of the known metabolic reactions in an organism and the genes that encode each enzyme. FBA calculates the flow of metabolites through this metabolic network, thereby making it possible to predict the growth rate of an organism or the rate of production of a biotechnologically important metabolite. With metabolic models for 35 organisms already available (http://systemsbiology.ucsd.edu/ In_Silico_Organisms/Other_Organisms) and high-throughput technologies enabling the construction of many more each year5-7, FBA is an important tool for harnessing the knowledge encoded in these models. In this primer, we illustrate the principles behind FBA by applying it to predict the maximum growth rate of Escherichia coli in the presence and absence of oxygen. The principles outlined can be applied in many other contexts to analyze the phenotypes and capabilities of organisms with different environmental and genetic perturbations (a Supplementary Tutorial provides ten additional worked examples with figures and computer code). Flux balance analysis is based on constraints The first step in FBA is to mathematically represent metabolic reactions (Box 1 and Fig. 1). Jeffrey D. Orth and Bernhard Ø. Palsson are at the University of California San Diego, La Jolla, California, USA. Ines Thiele is at the University of Iceland, Reykjavik, Iceland. e-mail: [email protected]
The core feature of this representation is a tabulation, in the form of a numerical matrix, of the stoichiometric coefficients of each reaction (Fig. 2a,b). These stoichiometries impose constraints on the flow of metabolites through the network. Constraints such as these lie at the heart of FBA, differentiating the approach from theory-based models dependent on biophysical equations that require many difficultto-measure kinetic parameters8,9. Constraints are represented in two ways, as equations that balance reaction inputs and outputs and as inequalities that impose bounds on the system. The matrix of stoichiometries imposes flux (that is, mass) balance constraints on the system, ensuring that the total amount of any compound being produced must be equal to the total amount being consumed at steady state (Fig. 2c). Every reaction can also be given upper and lower bounds, which define the maximum and minimum allowable fluxes of the reactions. These balances and bounds define the space of allowable flux distributions of a system—that is, the rates at which every metabolite is consumed or produced by each reaction. Other constraints can also be added10. From constraints to optimizing a phenotype The next step in FBA is to define a phenotype in the form of a biological objective that is relevant to the problem being studied (Fig. 2d). In the case of predicting growth, the objective is biomass production—that is, the rate at which metabolic compounds are converted into biomass constituents such as nucleic acids, proteins and lipids. Biomass production is mathematically represented by adding an artificial ‘biomass reaction’— that is, an extra column of coefficients in the
nature biotechnology volume 28 number 3 march 2010
matrix of stoichiometries—that consumes precursor metabolites at stoichiometries that simulate biomass production. The biomass reaction is based on experimental measurements of biomass components. This reaction is scaled so that the flux through it is equal to the exponential growth rate (µ) of the organism. Now that biomass is represented in the model, predicting the maximum growth rate can be accomplished by calculating the conditions that result in the maximum flux through the biomass reaction. In other cases, more than one reaction may contribute to the phenotype of interest. Mathematically, an ‘objective function’ is used to quantitatively define how much each reaction contributes to the phenotype. Taken together, the mathematical representations of the metabolic reactions and of the objective define a system of linear equations. In flux balance analysis, these equations are solved using linear programming (Fig. 2e). Many computational linear programming algorithms exist, and they can very quickly identify optimal solutions to large systems of equations. The COBRA Toolbox11 is a freely available Matlab toolbox for performing these calculations (Box 2). Suppose we want to calculate the maximum aerobic growth of E. coli under the assumption that uptake of glucose, and not oxygen, is the limiting constraint on growth. This calculation can be performed using a published model of E. coli metabolism12. In addition to metabolic reactions and the biomass reaction discussed above, this model also includes reactions that represent glucose and oxygen uptake into the cell. The assumptions are mathematically represented by setting the maximum rate of glucose uptake to a physiologically realistic level (18.5 mmol
245
p r ime r
© 2010 Nature America, Inc. All rights reserved.
Box 1 Mathematical representation of metabolism Metabolic reactions are represented as a stoichiometric matrix (S) of size m × n. Every row of this matrix represents one unique compound (for a system with m compounds) and every column represents one reaction (n reactions). The entries in each column are the stoichiometric coefficients of the metabolites participating in a reaction. There is a negative coefficient for every metabolite consumed and a positive coefficient for every metabolite that is produced. A stoichiometric coefficient of zero is used for every metabolite that does not participate in a particular reaction. S is a sparse matrix because most biochemical reactions involve only a few different metabolites. The flux through all of the reactions in a network is represented by the vector v, which has a length of n. The concentrations of all metabolites are represented by the vector x, with length m. The system of mass balance equations at steady state (dx/dt = 0) is given in Fig. 2c26: Sv = 0 Any v that satisfies this equation is said to be in the null space of S. In any realistic large-scale metabolic model, there are more reactions than there are compounds (n > m). In other words, there are more unknown variables than equations, so there is no unique solution to this system of equations. Although constraints define a range of solutions, it is still possible to identify and
glucose gDW–1 h–1; DW, dry weight) and setting the maximum rate of oxygen uptake to an arbitrarily high level, so that it does not limit growth. Then, linear programming is used to determine the maximum possible flux through the biomass reaction, resulting in a predicted exponential growth rate of 1.65 h–1. Anerobic growth of E. coli can be calculated by constraining the maximum rate of uptake of oxygen to zero and solving the system of equations, resulting in a predicted growth rate of 0.47 h–1 (see Supplementary Tutorial for computer code). As these two examples show, FBA can be used to perform simulations under different conditions by altering the constraints on a model. To change the environmental conditions (such as substrate availability), we change the bounds on exchange reactions (that is, reactions representing metabolites flowing into and out of the system). Substrates that are not available are
246
v3
v3
v3 Optimization maximize Z
Constraints 1) Sv = 0 2) a i < v i < b i
v1
v1 Unconstrained solution space v2
v1
Allowable solution space v2
Optimal solution v2
Figure 1 The conceptual basis of constraint-based modeling. With no constraints, the flux distribution of a biological network may lie at any point in a solution space. When mass balance constraints imposed by the stoichiometric matrix S (labeled 1) and capacity constraints imposed by the lower and upper bounds (ai and bi) (labeled 2) are applied to a network, it defines an allowable solution space. The network may acquire any flux distribution within this space, but points outside this space are denied by the constraints. Through optimization of an objective function, FBA can identify a single optimal flux distribution that lies on the edge of the allowable solution space.
analyze single points within the solution space. For example, we may be interested in identifying which point corresponds to the maximum growth rate or to maximum ATP production of an organism, given its particular set of constraints. FBA is one method for identifying such optimal points within a constrained space (Fig. 1). FBA seeks to maximize or minimize an objective function Z = cTv, which can be any linear combination of fluxes, where c is a vector of weights indicating how much each reaction (such as the biomass reaction when simulating maximum growth) contributes to the objective
constrained to an uptake rate of 0 mmol gDW–1 h–1. Constraints can also be tailored to the organism being studied, with lower bounds of 0 mmol gDW–1 h–1 used to simulate reactions that are irreversible in some organisms. Nonzero lower bounds can also force a minimal flux through artificial reactions (like the biomass reaction) such as the ‘ATP maintenance reaction’, which is a balanced ATP hydrolysis reaction used to simulate energy demands not associated with growth13. Constraints can even be used to simulate gene knockouts by limiting reactions to zero flux. FBA does not require kinetic parameters and can be computed very quickly even for large networks. This makes it well suited to studies that characterize many different perturbations such as different substrates or genetic manipulations. An example of such a case is given in example 6 in Supplementary Tutorial, which explores the effects on
function. In practice, when only one reaction is desired for maximization or minimization, c is a vector of zeros with a value of 1 at the position of the reaction of interest (Fig. 2d). Optimization of such a system is accomplished by linear programming (Fig. 2e). FBA can thus be defined as the use of linear programming to solve the equation Sv = 0, given a set of upper and lower bounds on v and a linear combination of fluxes as an objective function. The output of FBA is a particular flux distribution, v, which maximizes or minimizes the objective function.
growth of deleting every pairwise combination of 136 E. coli genes to find double gene knockouts that are essential for survival of the bacteria. FBA has limitations, however. Because it does not use kinetic parameters, it cannot predict metabolite concentrations. It is also only suitable for determining fluxes at steady state. Except in some modified forms, FBA does not account for regulatory effects such as activation of enzymes by protein kinases or regulation of gene expression. Therefore, its predictions may not always be accurate. The many uses of flux balance analysis Because the fundamentals of flux balance analysis are simple, the method has found diverse uses in physiological studies, gap-filling efforts and genome-scale synthetic biology3. By altering the bounds on certain reactions, growth on different media (example 1 in Supplementary Tutorial) or of bacteria with multiple gene
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
p r ime r
knockouts (example 6 in Supplementary Tutorial) can be simulated14. FBA can then be used to predict the yields of important cofactors such as ATP, NADH, or NADPH15 (example 2 in Supplementary Tutorial). Whereas the example described here yielded a single optimal growth phenotype, in large metabolic networks, it is often possible for more than one solution to lead to the same desired optimal growth rate. For example, an organism may have two redundant pathways that both generate the same amount of ATP, so either pathway could be used when maximum ATP production is the desired phenotype. Such alternate optimal solutions can be identified through flux variability analysis, a method that uses FBA to maximize and minimize every reaction in a network16 (example 3 in Supplementary Tutorial), or by using a mixed-integer linear programming–based algorithm17. More detailed phenotypic studies can be performed such as robustness analysis18, in which the effect on the objective function of varying a particular reaction flux can be analyzed (example 4 in Supplementary Tutorial).
A B+C B + 2C D
Genome-scale metabolic reconstruction
Reactions
b
1 2
Mathematically represent metabolic reactions and constraints
A B C D
...
n
Bi o Gl mas uc s Ox ose yg en
a
Met abolit es
Figure 2 Formulation of an FBA problem. (a) A metabolic network reconstruction consists of a list of stoichiometrically balanced biochemical reactions. (b) This reconstruction is converted into a mathematical model by forming a matrix (labeled S), in which each row represents a metabolite and each column represents a reaction. Growth is incorporated into the reconstruction with a biomass reaction (yellow column), which simulates metabolites consumed during biomass production. Exchange reactions (green columns) are used to represent the flow of metabolites, such as glucose and oxygen, in and out of the cell. (c) At steady state, the flux through each reaction is given by Sv = 0, which defines a system of linear equations. As large models contain more reactions than metabolites, there is more than one possible solution to these equations. (d) Solving the equations to predict the maximum growth rate requires defining an objective function Z = cTv (c is a vector of weights indicating how much each reaction (v) contributes to the objective). In practice, when only one reaction, such as biomass production, is desired for maximization or minimization, c is a vector of zeros with a value of 1 at the position of the reaction of interest. In the growth example, the objective function is Z = vbiomass (that is, c has a value of 1 at the position of the biomass reaction). (e) Linear programming is used to identify a flux distribution that maximizes or minimizes the objective function within the space of allowable fluxes (blue region) defined by the constraints imposed by the mass balance equations and reaction bounds. The thick red arrow indicates the direction of increasing Z. As the optimal solution point lies as far in this direction as possible, the thin red arrows depict the process of linear programming, which identifies an optimal point at an edge or corner of the solution space.
...
*
–1 –1
m
d
Mass balance defines a system of linear equations
= 0
vn vbiomass vglucose voxygen
–v1 + ... = 0 v1 – v2 + ... = 0 v1 – 2v2 + ... = 0 v2 + ... = 0 etc.
Define objective function (Z = c1* v1 + c2* v2 ... )
To predict growth, Z = v biomass
v2
e
...
Fluxes, v
Stoichiometric matrix, S
c
v1 v2
–1 1 –1 1 –2 1
Reaction 1 Reaction 2 ... Reaction n
Z Point of optimal v
Calculate fluxes that maximize Z
Solution space defined by constraints
v1
A more advanced form of robustness analysis involves varying two fluxes simultaneously to form a phenotypic phase plane19 (example 5 in Supplementary Tutorial). All genome-scale metabolic reconstructions are incomplete, as they contain ‘knowledge gaps’ where reactions are missing. FBA is the basis for several algorithms
that predict which reactions are missing by comparing in silico growth simulations to experimental results20-22. Constraint-based models can also be used for metabolic engineering where FBA-based algorithms, such as OptKnock23, can predict gene knockouts that allow an organism to produce desirable compounds24,25.
Box 2 Tools for FBA FBA computations, which fall into the category of constraint-based reconstruction and analysis (COBRA) methods, can be performed using several available tools 27-29. The COBRA Toolbox11 is a freely available Matlab toolbox (http://systemsbiology.ucsd.edu/ Downloads/Cobra_Toolbox) that can be used to perform a variety of COBRA methods, including many FBA-based methods. Models for the COBRA Toolbox are saved in the Systems Biology Markup Language (SBML) 30 format and can be loaded with the function ‘readCbModel’. The E. coli core model used in this Primer is available at http://systemsbiology.ucsd.edu/Downloads/E_coli_Core/. In Matlab, the models are structures with fields, such as ‘rxns’ (a list of all reaction names), ‘mets’ (a list of all metabolite names) and ‘S’ (the stoichiometric matrix). The function ‘optimizeCbModel’ is used to perform FBA. To change the bounds on reactions, use the function ‘changeRxnBounds’. The Supplementary Tutorial contains examples of COBRA toolbox code for performing FBA, as well as several additional types of constraint-based analysis.
nature biotechnology volume 28 number 3 march 2010
247
p r ime r Ultimately, FBA produces predictions that must be verified. Experimental studies are used as part of the model reconstruction process and to validate model predictions. Studies have shown that growth rates of E. coli on several different substrates predicted by FBA agree well with those obtained by experimental measurements14. Model-based predictions of gene essentiality have also been shown to be quite accurate2. This primer and the accompanying tutorials based on the COBRA toolbox (Box 2) should help those interested in harnessing the growing cadre of genome-scale metabolic reconstructions that are becoming available.
© 2010 Nature America, Inc. All rights reserved.
ACKNOWLEDGMENTS This work was supported by National Institutes of Health grant no. R01 GM057089. 1. Duarte, N.C. et al. Proc. Natl. Acad. Sci. USA 104, 1777–1782 (2007). 2. Feist, A.M. et al. Mol. Syst. Biol. 3, 121 (2007). 3. Feist, A.M. & Palsson, B.O. Nat. Biotechnol. 26,
248
659–667 (2008). 4. Oberhardt, M.A., Palsson, B.O. & Papin, J.A. Mol. Syst. Biol. 5, 320 (2009). 5. Thiele, I. & Palsson, B.O. Nat. Protoc. 5, 93–121 (2010). 6. Feist, A.M., Herrgard, M.J., Thiele, I., Reed, J.L. & Palsson, B.O. Nat. Rev. Microbiol. 7, 129–143 (2009). 7. Durot, M., Bourguignon, P.Y. & Schachter, V. FEMS Microbiol. Rev. 33, 164–190 (2009). 8. Covert, M.W. et al. Trends Biochem. Sci. 26, 179– 186 (2001). 9. Edwards, J.S., Covert, M. & Palsson, B. Environ. Microbiol. 4, 133–140 (2002). 10. Price, N.D., Reed, J.L. & Palsson, B.O. Nat. Rev. Microbiol. 2, 886–897 (2004). 11. Becker, S.A. et al. Nat. Protoc. 2, 727–738 (2007). 12. Orth, J.D., Fleming, R.M. & Palsson, B.O. in EcoSal —Escherichia coli and Salmonella Cellular and Molecular Biology (ed. Karp, P.D.) (ASM Press, Washington, DC, 2009). 13. Varma, A. & Palsson, B.O. Biotechnol. Bioeng. 45, 69–79 (1995). 14. Edwards, J.S., Ibarra, R.U. & Palsson, B.O. Nat. Biotechnol. 19, 125–130 (2001). 15. Varma, A. & Palsson, B.O. J. Theor. Biol. 165, 477– 502 (1993). 16. Mahadevan, R. & Schilling, C.H. Metab. Eng. 5, 264–276 (2003).
17. Lee, S., Phalakornkule, C., Domach, M.M. & Grossmann, I.E. Comput. Chem. Eng. 24, 711–716 (2000). 18. Edwards, J.S. & Palsson, B.O. Biotechnol. Prog. 16, 927–939 (2000). 19. Edwards, J.S., Ramakrishna, R. & Palsson, B.O. Biotechnol. Bioeng. 77, 27–36 (2002). 20. Reed, J.L. et al. Proc. Natl. Acad. Sci. USA 103, 17480–17484 (2006). 21. Kumar, V.S. & Maranas, C.D. PLoS Comput. Biol. 5, e1000308 (2009). 22. Satish Kumar, V., Dasika, M.S. & Maranas, C.D. BMC Bioinformatics 8, 212 (2007). 23. Burgard, A.P., Pharkya, P. & Maranas, C.D. Biotechnol. Bioeng. 84, 647–657 (2003). 24. Feist, A.M. et al. Metab. Eng. published online, doi:10.1016/j.ymben.2009.10.003 (17 October 2009). 25. Park, J.M., Kim, T.Y. & Lee, S.Y. Biotechnol. Adv. 27, 979–988 (2009). 26. Palsson, B.O. Systems Biology: Properties of Reconstructed Networks (Cambridge University Press, New York, 2006). 27. Jung, T.S., Yeo, H.C., Reddy, S.G., Cho, W.S. & Lee, D.Y. Bioinformatics 25, 2850–2852 (2009). 28. Klamt, S., Saez-Rodriguez, J. & Gilles, E.D. BMC Syst. Biol. 1, 2 (2007). 29. Lee, D.Y., Yun, H., Park, S. & Lee, S.Y. Bioinformatics 19, 2144–2146 (2003). 30. Hucka, M. et al. Bioinformatics 19, 524–531 (2003).
volume 28 number 3 march 2010 nature biotechnology
Articles
Nutrient-sensitized screening for drugs that shift energy metabolism from mitochondrial respiration to glycolysis
© 2010 Nature America, Inc. All rights reserved.
Vishal M Gohil1–3,7, Sunil A Sheth1–3,7, Roland Nilsson1–3, Andrew P Wojtovich4,5, Jeong Hyun Lee6, Fabiana Perocchi1–3, William Chen1–3, Clary B Clish2, Cenk Ayata6, Paul S Brookes4,5 & Vamsi K Mootha1–3 Most cells have the inherent capacity to shift their reliance on glycolysis relative to oxidative metabolism, and studies in model systems have shown that targeting such shifts may be useful in treating or preventing a variety of diseases ranging from cancer to ischemic injury. However, we currently have a limited number of mechanistically distinct classes of drugs that alter the relative activities of these two pathways. We screen for such compounds by scoring the ability of > 3,500 small molecules to selectively impair growth and viability of human fibroblasts in media containing either galactose or glucose as the sole sugar source. We identify several clinically used drugs never linked to energy metabolism, including the antiemetic meclizine, which attenuates mitochondrial respiration through a mechanism distinct from that of canonical inhibitors. We further show that meclizine pretreatment confers cardioprotection and neuroprotection against ischemia-reperfusion injury in murine models. Nutrient-sensitized screening may provide a useful framework for understanding gene function and drug action within the context of energy metabolism. Virtually all cells exhibit metabolic flexibility and are capable of shifting their reliance on glycolysis relative to mitochondrial respiration. Such shifts can occur at different timescales through a variety of mechanisms, allowing cells to cope with prevailing nutrient availability or energetic demands. There is mounting evidence of the therapeutic potential of targeting such shifts. For example, many cancer cells rely on aerobic glycolysis (the Warburg effect)1, and a recent study has shown that pharmacological agents that shift their metabolism toward mitochondrial respiration can retard tumor growth2. Conversely, studies in animal models have shown that attenuating mitochondrial respiration can prevent the pathological consequences of ischemiareperfusion injury in myocardial infarction and stroke3–7. These observations motivate the search for agents that can safely induce shifts in cellular energy metabolism in humans. Promising work in this area has focused on hypoxia-inducible factor (HIF)8, a well-studied transcriptional regulator of genes involved in the cellular adaptation to hypoxia9,10. HIF inhibitors and activators have been identified through both academic and pharmaceutical drug screens and exhibit preclinical efficacy in treating cancer11 and in ischemic disease12, respectively. Other approaches to treat ischemic injury include induced hypothermia, which has met with mixed results13. New classes of agents, that can be titrated to safely shift energy metabolism, may yet provide important therapeutic value in several human diseases. Here we use a nutrient-sensitized screening strategy to identify drugs that shift cellular energy metabolism based on their selective effect on cell growth and viability in glucose- versus galactose-containing
media. Nutrient-sensitized screening is based on the evidence that mammalian cells redirect their energy metabolism in response to the available sugar source14. Culturing cells in galactose as the sole sugar source forces mammalian cells to rely on mitochondrial oxidative phosphorylation (OXPHOS) and was previously used to diagnose human mitochondrial disorders or drug toxicity 15,16. By screening our chemical library for drugs that selectively inhibit cell growth and proliferation in galactose- relative to glucose-containing media, we identify several FDA-approved compounds that redirect oxidative metabolism to glycolysis. We pursue the mechanism and potential clinical utility of one drug, meclizine, which is available without prescription, crosses the blood-brain barrier and has never, to our knowledge, been linked to energy metabolism. RESULTS A metabolic state–dependent growth and viability assay Consistent with previous studies focused on other cell types14,17, we found that human MCH58 skin fibroblasts grown in glucose derive ATP from both aerobic glycolysis and mitochondrial glutamine oxidation (Fig. 1). However, when these cells are grown in galactose they exhibit a five- to sixfold decrease in the extracellular acidification rate (ECAR)18, reflecting decreased glycolysis, and a twofold increase in the oxygen consumption rate (OCR), consistent with a switch to glutamine oxidation14 (Fig. 1b,c). Moreover, cells grown in galactosecontaining medium maximize mitochondrial ATP production by using a larger fraction of mitochondrial respiration for ATP synthesis (Supplementary Fig. 1).
1Center
for Human Genetic Research, Massachusetts General Hospital, Boston, Massachusetts, USA. 2Broad Institute of Massachusetts Institute of Technology and Harvard, Cambridge, Massachusetts, USA. 3Department of Systems Biology, Harvard Medical School, Boston, Massachusetts, USA. 4Department of Pharmacology and Physiology and 5Department of Anesthesiology, University of Rochester Medical Center, Rochester, New York, USA. 6Stroke and Neurovascular Regulation Laboratory, Department of Radiology, Massachusetts General Hospital, Charlestown, Massachusetts, USA. 7These authors contributed equally to this work. Correspondence should be addressed to V.K.M. ([email protected]). Received 23 November 2009; accepted 12 January 2010; published online 14 February 2010; doi:10.1038/nbt.1606
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
249
Articles a
b
c
Positive Sglu/gal scores indicate drugs that were selectively lethal or inhibited growth in galac40 Glucose Lactate Lactate tose-containing medium, such as inhibitors 30 ETC ETC of OXPHOS. Negative Sglu/gal scores may arise from inhibition of glycolysis or from inhibie– O2 H O e– O2 H2O 20 2 tion of proliferation, as fibroblasts cultured in TCA ATP TCA ATP Galactose 10 cycle glucose grow more rapidly (Supplementary cycle Glutamate Glutamate Fig. 3b). For most drugs, Sglu/gal is close to 0 0 50 100 150 200 zero, indicating similar effects on growth and OCR (pmol/min/30 k cells) Glutamine Glutamine viability in glucose- and galactose-containing media (Fig. 2b and Supplementary Table 1). Figure 1 Metabolic plasticity of human fibroblasts. (a,b) Schematic representation of cellular Reassuringly, the upper tail of the Sglu/gal energy metabolism pathways. (a) Cells grown in glucose-rich medium derive ATP from glycolysis as well as from glutamine-driven respiration. TCA, tricarboxylic acid; ETC, electron transport chain. distribution (Fig. 2b) is highly enriched (b) Replacing glucose with galactose forces cells to generate ATP almost exclusively from glutaminefor known respiratory chain and OXPHOS driven oxidative metabolism14. (c) Measurement of ECAR, a proxy for the rate of glycolysis, and inhibitors: the top 25 compounds include OCR, a proxy for mitochondrial respiration, of fibroblasts grown in media containing 10 mM glucose 20 compounds previously known to disrupt or 10 mM galactose for 3 d. Data are expressed as mean ± s.d. (n = 5). respiration by directly interrupting mitochondrial respiration or uncoupling it from The metabolic flexibility of fibroblasts permits screening for com- ATP synthesis (Supplementary Table 2). Conversely, the lower tail is pounds that retard growth or are lethal to cells only in a given metabolic enriched for known antineoplastic agents (Fig. 2b): 14 of the 25 lowest state. In a pilot experiment, we confirmed nutrient-dependent drug sen- Sglu/gal scores correspond to known chemotherapeutic agents that are sitivity of fibroblasts to known inhibitors of OXPHOS (Supplementary likely to retard the growth and viability of cells rapidly proliferating Fig. 2). To screen a library of chemicals, we designed a high- in glucose (Supplementary Table 3). throughput microscopy-based growth assay to identify compounds We next asked if any clinically used drugs exhibit high Sglu/gal scores. that differentially affect growth and viability in galactose- relative Among the top 2% of the Sglu/gal distribution (83 compounds), we to glucose-containing media (Fig. 2a and Supplementary Fig. 3a). identified 25 agents that have been used clinically (Supplementary Because the proliferation rates are higher for cells provided with Table 4). Previous reports provide evidence that 9 out of these 25 drugs glucose instead of galactose as their sole sugar source (Supplementary (papaverine, phenformin, artemisinin, pentamidine (NebuPent), Fig. 3b), we considered the normalized cell number in each of the two clomiphene (Clomid), pimozide (Orap), niclosamide, fluvastatin nutrient conditions. By measuring growth and survival over a 3-day (Lescol), carvedilol (Coreg)) can directly inhibit or uncouple the mitoperiod, we were able to increase our ability to discover compounds chondrial respiratory chain (Supplementary Table 4). This list includes with even subtle effects on energy metabolism. two antimalarial drugs (mefloquine (Lariam) and artemisinin), the latter of which has been reported to require mitochondrial resA small-molecule screen for agents that shift energy metabolism piration in the malarial parasite for its action20. The remaining 16 We screened a library of 3,695 chemical compounds in duplicate. clinically used agents cover a broad range of indications and diverse The library has been previously described19 and consists of two com- mechanisms of action and, to our knowledge, have never been linked mercially available compound collections that span nearly half of all to energy metabolism. We were particularly interested in identifying FDA-approved drugs, as well as other bioactives and natural products. compounds that induce subtle metabolic shifts, as they may represent We analyzed the glucose and galactose results jointly to assign each particularly safe drugs with which to manipulate energy metabolism. drug a score, Sglu/gal, defined as the log ratio of the normalized cell To this end, we focused on commercially available drugs exhibiting number in glucose divided by the normalized cell number in galac- low to intermediate, positive Sglu/gal scores (0.15–0.45). We carried tose. The full table of results is provided in Supplementary Table 1. out secondary assays of OCR, ECAR and cell viability and confirmed Glucose
ATP Pyruvate
Galactose
Glucose
Galactose
72 h
3,695 compounds
b 1.0 0.8 0.6
Sglu/gal
0.4
No change
0.2 High Low No change
c
Antimycin Isorotenone Pentamidine Mefloquine Artemisinin Meclizine Fluvastatin Niclosamide Sertraline
Bottom 250
0 –0.2 –0.4 –0.6 –0.8
Top 250 All data (n = 3,695) Known OXPHOS inhibitors (n = 29) Antineoplastic drugs (n = 140)
Vinblastine Floxuridine Methotrexate Mitoxanthrone
OCR/ECAR (% of DMSO)
a
Sglu/gal
© 2010 Nature America, Inc. All rights reserved.
ECAR (mpH/min/30 k cells)
ATP Pyruvate
120
Ascorbate
Bisacodyl Clemastine Vinpocetine
DMSO
100
Sertraline
80
Nonoxynol-9 Meclizine Menadione Nifuroxazide Mefloquine Thonzonium
60 40
Chlorhexidine
20 0 0
0.2
Clofilium Alexidine
0.4 0.6 Sglu/gal
Antimycin
0.8
1
Figure 2 A nutrient-sensitized screen to discover agents that shift energy metabolism. (a) Schematic of the drug screen. MCH58 fibroblasts grown in 96-well plates in glucose- or galactose-containing media are exposed to a chemical library of 3,695 compounds for 72 h. The logarithm of the normalized cell number in glucose versus galactose provides a summary statistic (Sglu/gal) for each compound. (b) Results from a nutrient-sensitized screen. Sglu/gal is plotted for top and bottom 250 compounds. (c) Secondary assays to evaluate compounds with modest yet positive Sglu/gal scores. The OCR/ECAR ratio of selected compounds is plotted against the compounds’ corresponding Sglu/gal score from b. OCR and ECAR measurements were made on MCH58 fibroblasts grown in glucose and are normalized to cell viability. Compounds indicated by red symbols exhibited a statistically significant decrease in the OCR/ECAR ratio based on at least three independent replicates (P < 0.05; two-sided t-test).
250
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
Articles
40 20 20
40 60 80 Meclizine (µM)
*
300 250 200
*
150 100 50
0 – + Meclizine (50 µM) 293
e
–
+
OCR (% of baseline)
40
80
STHdh
Oligomycin (1 µM) Antimycin (1 µM) 120
40
*
30
**
10 0
50
f
160
Time (min)
*
20
40
0
10 25 Meclizine (µM)
*
50
*
*
30 20 10
0 – + Meclizine (50 µM) 293
+ Q7/7
HeLa
Rotenone (1 µM) Meclizine (50 µM) 20 0 –20 –40 –60 –80 –100 –120
–
50
–
+
–
+ Q7/7
HeLa
STHdh
HeLa 6h DMSO + – – DFO (100 µM) – – + Meclizine (50 µM) – + – HlF1-α
96 kDa
β-actin
42 kDa
HlF2-α
96 kDa
β-actin
42 kDa
that eight of these agents induce statistically significant (P < 0.05) metabolic shifts (Fig. 2c). Of these eight clinically used drugs, we were particularly interested in meclizine (Antivert), which has been approved for the treatment of nausea and vertigo for decades, is available over the counter, has a favorable safety profile, and likely penetrates the blood-brain barrier given its efficacy in disorders of the central nervous system21. Meclizine attenuates respiration in intact cells In secondary assays, we replicated our screening result and confirmed that galactose-grown cells are more sensitive to increasing doses of meclizine (Fig. 3). In agreement with our secondary screening assay (Fig. 2c), treatment with meclizine reduced the OCR in a dosedependent manner in cells cultured in glucose-rich medium (Fig. 3b). Meclizine-induced reduction in OCR and concomitant increase in 50 µM meclizine DMSO
60 G/M ADP
55 50 45
CCCP
40 35
Antimycin
30 25 20
b O2 in solution (nmol)
a O2 in solution (nmol)
Rotenone S ADP 60
50 µM meclizine DMSO
50 40
CCCP
30
Antimycin
20 10
0
100
200 Time (s)
300
400
ECAR occurred in all cell types tested, including immortalized mouse striatal cells, human embryonic kidney cells (HEK293) and HeLa cells (Fig. 3c–d and Supplementary Fig. 4). Although meclizine is classified as a histamine receptor (H1) antagonist and a weak muscarinic acetylcholine receptor antagonist22, the other 64 annotated H1 receptor antagonists and 33 annotated antimuscarinic antagonists in our chemical library did not exhibit elevated Sglu/gal scores (anticholinergic P = 0.26, anti-H1 P = 0.77; Mann-Whitney rank sum test). We tested two classic antihistamines—pyrilamine and pheniramine as well as two well-characterized antimuscarinic agents—atropine and scopolamine for their ability to inhibit OCR. Unlike meclizine, these agents did not inhibit cellular OCR (Supplementary Fig. 5). These results suggest that meclizine’s effect on energy metabolism occurs by means of a mechanism not involving cholinergic or histamine receptors. Meclizine’s suppression of cellular oxygen consumption occurred with much slower kinetics than canonical inhibitors of OXPHOS that directly target the respiratory chain or ATP synthase (Fig. 3e). The slow kinetics suggest that it takes time for meclizine to accumulate in mitochondria or alternatively, that it might act indirectly. To distinguish between these alternatives, we studied the effect of meclizine on isolated mitochondria. Using glutamate and malate, pyruvate and malate or succinate as fuel substrates, we found no effect of meclizine on respiration of isolated mitochondria (Fig. 4a–c). Meclizine did not have
c
d
50 µM meclizine DMSO
P/M
55
0.8
45 40
CCCP
35 30
Antimycin
25
100
200 300 Time (s)
400
500
CCCP
Mec/DMSO
0.6
0
100
200 300 Time (s)
400
0.5
500
50 µM meclizine DMSO
Figure 4 Effect of meclizine on bioenergetics of isolated mitochondria. (a–c) Acute effect of meclizine on oxygen consumption in isolated mitochondria using glutamate/malate (a), succinate (b) or pyruvate/ malate (c) as substrates. Traces are representative of five independent measurements. (d) Acute effect of meclizine on mitochondrial membrane potential measured with tetramethyl rhodamine methyl ester (TMRM) in isolated mitochondria. Traces are representative of five independent measurements. (e) Acute effect of meclizine on mitochondrial NADH in isolated mitochondria. Traces are representative of five independent measurements. Mitochondria (Mito), glutamate and malate (G/M), succinate (S), pyruvate/ malate (P/M), meclizine (Mec) or DMSO, ADP and carbonyl cyanide m-chlorophenyl hydrazone (CCCP) were added at indicated time points.
0
50
100
150
200
250
Time (s)
e
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
ADP
0.7
20 15
0
Mito G/M 0.9
ADP
50
0.40
Mito Mec/DMSO ADP
0.35 NADH (au)
© 2010 Nature America, Inc. All rights reserved.
d
*
Figure 3 Effects of meclizine on cellular energy metabolism. (a) Cell viability of MCH58 fibroblasts cultured in glucose- or galactose-containing media with varying doses of meclizine for 3 d. Data are expressed as mean ± s.d. (n = 5). (b) OCR in MCH58 fibroblasts cultured in glucose-containing medium with varying doses of meclizine for 200 min. Data are expressed as mean ± s.d. (n = 3). (*, P < 0.05; **, P < 0.005; two-sided t-test). (c,d) OCR (c) and ECAR (d) in multiple cell types cultured in glucose-containing medium with 50 µM meclizine or DMSO for 200 min. Data are expressed as mean ± s.d. (n ≥ 3). (*, P < 0.05; two-sided t-test). (e) Time course of meclizine (50 µM)-mediated OCR reduction over DMSO baseline compared to other inhibitors of OXPHOS (1 µM each) in HEK293 cells. Data are expressed as mean ± s.d. (n ≥ 3). (f) HIF-1α and HIF-2α detection by western blot analysis of protein extract from HeLa cells after 6-h treatment with 0.1% DMSO, 100 µM deferoxamine (DFO) or 50 µM meclizine. The complete immunoblot is provided as Supplementary Figure 7b.
60
100
ECAR (mpH/min/30 k cells)
OCR (pmol/min/30 k cells)
0
b
∆ψm (au)
Percent viability (RLU)
60
0
c
Glucose Galactose
80
O2 in solution (nmol)
100
OCR (pmol/min/30 k cells)
a
0.30
50 µM meclizine DMSO CCCP
G/M
0.25 0.20
0
50
100 150 Time (s)
200
250
251
Articles a
T (min)
Meclizine pretreatment confers protection against ischemic injury Previous studies have clearly established that brief, nonlethal episodes of ischemia can confer prophylaxis against subsequent stroke or myocardial infarction, and studies in model systems have shown that chemical inhibition of mitochondrial respiration can mimic this protection, a process coined “chemical preconditioning”6. Having shown that meclizine, an over-the-counter drug that crosses the blood-brain barrier can attenuate mitochondrial respiration, we sought to determine whether this drug is cardioprotective and neuro protective in cellular and animal models. First, we tested meclizine in an adult rat ventricular cardiomyocyte model of simulated ischemia-reperfusion injury (Fig. 5a). A 20-min meclizine pre-incubation followed by wash-out before ischemia elicited a dose-dependent protection of cardiomyocytes against cell death, whereas other antihistamines (pyrilamine and pheniramine) and antimuscarinic agents (scopolamine and atropine) did not
252
50
110
Veh.
Normoxia
Veh.
Ischemia
Drug
Ischemia
* *
140
Reperfusion Reperfusion
c
100
Cardiomyocyte VO2 (% of control)
Cell viability (% of control)
80 60 40 20
t M Ve rol ec hi cl ( M 0.1 e ec µ (0 M) M .5 µ ec M M (1 ) e µM At c (5 ) ro µ Ph p ( M) en 1 µ ir M ) Py (1 µ M r Sc il (1 ) op µ o M) (1 µM )
0
100 80 60 40
*
*
*
20 0
e 60
Control Meclizine
0
50
100 Time (min)
150
80
* * *
60 40 20 0
SIR
Infarct/area at risk (%)
120
Ischemia
Rate pressure product (% of initial)
© 2010 Nature America, Inc. All rights reserved.
30
b 100
d a qualitative impact on membrane potential or redox potential during respiratory state transitions of isolated mitochondria (Fig. 4d,e). Furthermore, meclizine treatment had no effect on mitochondrial morphology, membrane potential, mitochondrial (mt) DNA copy number or the expression of mtRNAs in intact cells (Supplementary Fig. 6). Collectively, these observations demonstrate that unlike classic inhibitors or uncouplers such as rotenone, antimycin, oligomycin or carbonyl cyanide m-chlorophenyl hydrazone, meclizine does not itself directly inhibit and/or uncouple the OXPHOS machinery in isolated mitochondria, and does not reduce mitochondrial biogenesis in intact cells. Instead it may act through novel signaling or transcriptional mechanisms. Activation of HIF1-α or HIF2-α is known to induce transcriptional rewiring of energy metabolism from mitochondrial respiration to glycolysis23. However, unlike with deferoxamine, a known inducer of the HIF pathway, we did not observe HIF1-α or HIF2-α stabilization after meclizine treatment (Fig. 3f and Supplementary Fig. 7a,b). Moreover, the kinetics of meclizine’s OCR inhibition argue against a transcriptional mechanism, as meclizine treatment resulted in inhibition within 2 h, whereas OCR inhibition mediated by deferoxamine only became apparent after 12 h (Supplementary Fig. 7c). In addition, we examined the effect of meclizine treatment on HIF-responsive genes using an HIF response element-luciferase reporter construct and recorded no induction of luciferase activity after a 6-h treatment (Supplementary Fig. 7d). Collectively, these studies suggest that meclizine inhibits mitochondrial respiration indirectly, in an HIF-independent manner that does not involve histaminergic or muscarinic receptor signaling.
0 10
Control SIR SIR+drug
C on
Figure 5 Meclizine is cardioprotective in cellular and ex vivo models of cardiac ischemia. (a) Protocol for the simulated ischemia-reperfusion (SIR) injury model. (b) Viability of adult rat cardiomyocytes subjected to SIR injury, in the presence of indicated concentrations of meclizine (Mec), atropine (Atrop), pheniramine (Phenir), pyrilamine (Pyril) and scopolamine (Scopo). (c) Respiration of cardiomyocytes after exposure to indicated concentrations of meclizine. (d,e) Langendorff perfused rat hearts were subjected to 25 min of global ischemia followed by 2 h of reperfusion. Meclizine treatment comprised infusion of 1 µM meclizine from a port above the aortic cannula for 20 min, followed by a 1-min wash-out before ischemia. (d) Rate pressure product (heart rate × left ventricular developed pressure) is expressed as a percentage of the initial value throughout the ischemia-reperfusion (IR) protocol. (e) Following IR, hearts were stained with TTC and infarct size was quantified. All data are means ± s.e.m. from four to six individual experiments. (*, P < 0.05, ANOVA in d, Student’s t-test in e).
0 1 5 10 Meclizine (µM)
*
40
20
0
IR alone IR + 1 µM Mec
provide protection (Fig. 5b). These results are consistent with the hypothesis that mild OXPHOS inhibition is cytoprotective in ischemic injury. As with the other cell types, meclizine inhibited oxygen consumption in cardiomyocytes in a dose-dependent manner (Fig. 5c and Supplementary Fig. 8a,b) but not in isolated cardiac mitochondria (Supplementary Fig. 8c,d). Next, we tested whether meclizine protects isolated, perfused rat hearts from ischemia-reperfusion injury in an ex vivo model of ischemic injury. Meclizine preserved heart pump function after the ischemic event (Fig. 5d) and significantly (P < 0.05) reduced the infarct area of Langendorff perfused rat hearts subjected to 25 min of global ischemia (Fig. 5e). Chemical preconditioning has also been shown to be protective in animal models of cerebral ischemia6,12,24–26. To determine whether meclizine might similarly be useful in this context, we first established safety and pharmacokinetic parameters for an intraperitoneal dosing regimen (see Online Methods). We found that mice tolerate daily intraperitoneal injections of 100 mg/kg meclizine without any weight loss or behavioral changes even after four consecutive days. Six hours after a single intraperitoneal dose, the plasma concentration is in the 3–5 µM range, a concentration sufficient to attenuate mitochondrial respiration of primary mouse neurons (Supplementary Fig. 9). We then tested whether meclizine protects against cerebral ischemia by pretreating mice with two intraperitoneal injections of 100 mg/kg meclizine, or an equal volume of vehicle at 17 h and 3 h before a 1 h transient middle cerebral artery occlusion (Fig. 6). We found that total infarct volume was significantly reduced by 23% in meclizinetreated animals (P = 0.03, Fig. 6c). In addition, meclizine significantly reduced the area of infarction in brain slices with the greatest area of infarct (P = 0.02, Fig. 6d,e). The in vivo protective effect of meclizine is likely independent of its antihistamine or antimuscarinic property, as treatment with pyrilamine or scopolamine did not decrease infarct volume (Fig. 6c) or reduce the area of infarction in brain slices (Fig. 6d,e). Furthermore, meclizine-pretreated animals tended toward having preserved neurological function compared to controls (P = 0.07, Kruskal-Wallis non-parametric ANOVA).
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
Articles
–17 h
25 h Reperfusion
Ischemia
60 40 20
20
in e
in e am r il
–10 0 10 20 30 40 50 60 70 80 Time (minutes) Vehicle Meclizine Pyrilamine Scopolamine
20 15 10 5 0
Sc
Py
Vehicle
op ol am
in e
le
ec liz M
Ve hi c
© 2010 Nature America, Inc. All rights reserved.
40
e 25
0
d
Vehicle Meclizine Pyrilamine Scopolamine
60
0
*
80
80
Infarct area (mm2)
Infarct volume (mm3)
c 100
CBF (% of baseline)
–3 0 h h 1 h
Time
100
Reperfusion
120 1st 2nd injection injection
MCA occlusion
b
CCA occlusion Baseline
a
Slice numbers 1 – 10 Meclizine Pyrilamine
1
2
3
4 5 6 7 8 Slice number
9 10
Scopolamine
The cerebral blood flow deficit (Fig. 6b) and the amount of post operative weight loss did not differ between the groups. DISCUSSION Recent studies have shown that changes in cellular energy metabolism can accompany a range of human diseases, and that targeting energy metabolism may hold therapeutic potential. However, we lack an arsenal of clinically safe and useful agents that target energy metabolism by distinct mechanisms. In this study, we have introduced a facile, nutrient-sensitized screening strategy aimed at identifying small molecules that shift cellular energy metabolism from mitochondrial respiration to glycolysis. We have identified several FDA-approved drugs that exhibit such activity and may have potential for helping to abrogate ischemia-reperfusion injury in the heart, brain and perhaps other sensitive organ systems such as the lung or kidney. Focusing on one specific hit from our screen, meclizine, we have demonstrated that it suppresses OXPHOS by means of a mechanism distinct from classic inhibitors or uncouplers, and that it confers protection against cardiac and cerebral ischemia in cellular and animal models. A large body of literature demonstrates that agents that blunt mitochondrial respiration can offer prophylaxis against cell death after ischemia and reperfusion in the heart4,5,27 or brain3,6,26. This effect is thought to occur through suppressing oxidative injury, and may be related to protection conferred by ischemic preconditioning, although the precise molecular mechanism is not known. Notably, redirecting energy metabolism toward glycolysis can reduce oxidative damage and suppresses apoptosis28–30. Interestingly, switching to anaerobic metabolism appears to be a natural adaptation to reduced oxygen availability31, and activation of the HIF pathway provides one such strategy for redirecting energy metabolism toward glycolysis9,32. Recent studies using genetic and chemical approaches of activating the HIF pathway have shown promising results in various models of ischemia-reperfusion injury. For example, myofibers of prolyl
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
Figure 6 Meclizine is neuroprotective in a mouse model of stroke. (a) Protocol for the murine model of stroke. Male C57BL/6 mice were treated with two intraperitoneal injections of 100 mg/kg meclizine, 20 mg/kg pyrilamine and 0.5 mg/kg scopolamine or vehicle at 17 h and 3 h before 1 h transient middle cerebral artery occlusion followed by 24 h of reperfusion. (b) Cerebral blood flow (CBF) measured at baseline and after occlusion of the common carotid artery (CCA) and middle cerebral artery (MCA) upon treatment with meclizine, scopolamine, pyrilamine or vehicle. Data represent mean ± s.d. (c) Infarct volume measured on TTC-stained 1-mm thick coronal slices obtained from mice treated with meclizine, scopolamine, pyrilamine or vehicle. Data points refer to independent experiments, and the solid line represents their mean. (*P < 0.05 versus vehicle and scopolamine, P < 0.01 versus pyrilamine; one-way ANOVA followed by Tukey’s multiple comparison test). (d) Representative images of TTC-stained 1-mm thick coronal brain sections (slice 1–10). (e) Infarct area in the rostrocaudal extent of the brain (slice 1–10) upon treatment with meclizine, scopolamine, pyrilamine or vehicle. Data points represent the mean area of infarction in individual slice levels ± s.d. in mm2 (n = 14 for vehicle, n = 8 for meclizine, n = 8 for pyrilamine, n = 5 for scopolamine, *P < 0.05).
hydroxylase 1 knockout mice are resistant to acute ischemia because of reduced generation of oxidative stress33. In preclinical studies, prolyl hydroxylase inhibitors confer protection in models of myocardial infarction34, stroke35 and renal ischemia36. However, as HIF regulates the expression of a plethora of genes, and unwanted side effects have remained a concern37, it might be useful to expand the arsenal of agents that shift energy metabolism. Our screen has identified a new metabolic activity for meclizine, an over-the-counter drug that has been in use in the United States for > 40 years for treatment of nausea and vertigo. We found that 1 µM meclizine pretreatment provided cytoprotection in in vitro and ex vivo models of cardiac ischemia-reperfusion injury (Fig. 5). In addition, we showed that prophylaxis with meclizine significantly reduced infarct volume in an in vivo model of cerebral ischemia (Fig. 6). The utility of pretreatment paradigms described in this study arises in clinical settings in which ischemic insults can be anticipated. Examples of such situations include patients undergoing high-risk surgical procedures and the large cohort of patients that suffer from diseases of recurrent ischemia, such as unstable angina or recurrent transient ischemic attacks38. An open question is whether currently approved doses of meclizine achieve the required blood concentrations required for cardioprotection or neuroprotection. Post-marketing surveillance data support the safe nonprescription use of meclizine, and published studies in animals, including nonhuman primates, have shown that higher doses can be tolerated39,40. However, because the potency of meclizine in attenuating mitochondrial respiration appears to vary across cell types (Figs. 3c and 5c and Supplementary Fig. 9), pre clinical studies of efficacy and toxicity are required to rigorously determine optimal dosing and safety regimens before evaluating the utility of meclizine for new clinical indications in humans. Our detailed studies on the effects of meclizine on cellular energy metabolism clearly show that it attenuates mitochondrial respiration in a manner distinct from other drugs of known mechanism of action and independent of the HIF pathway (Fig. 3f and Supplementary Fig. 7). In contrast to canonical inhibitors, meclizine does not directly target the OXPHOS machinery in isolated mitochondria (Fig. 4) and can be titrated over a broad range of concentrations to achieve inhibition of cellular OCR by 10–60% (Fig. 3b). Our data suggest that meclizine acts independently of the muscarinic or histamine receptors, as drugs affecting these two receptors did not inhibit OCR (Supplementary Fig. 5) and they do not confer neuroprotection or cytoprotection in our models (Figs. 5 and 6). We do not know the precise molecular target
253
© 2010 Nature America, Inc. All rights reserved.
Articles of meclizine responsible for this effect on energy metabolism, but one possibility is a metabolic target outside the mitochondrion whose subsequent impact is to direct metabolism away from OXPHOS. Alternatively, meclizine may undergo a bio-transformation into a product that directly inhibits the OXPHOS machinery. We cannot exclude the possibility that meclizine, like many clinically used drugs, hits multiple cellular targets to affect cellular energetics and confer cytoprotection. Nutrient-sensitized screening, as we have presented it, builds on previous studies that have shown that many cultured cells generate their ATP from either glycolysis or glutamine oxidation14,17. However, the screening strategy may not necessarily work in other cell types, for example, cells with limited metabolic flexibility, cells that do not have pathways for glutamine oxidation, or postmitotic cells in which a growth assay is not possible. Another limitation of our approach is that the compounds that emerge from the screen may act not only on energy-related pathways, but also potentially on other properties influenced by the switch in nutrients. For example, we have noted that cells grown in glucose tend to proliferate more quickly, and for this reason, drugs from the right side of the tail (Fig. 2b) could either be blunting glycolytic metabolism or affecting rapid proliferation. Hence, secondary assays are still required to confirm the energetic consequences of a drug identified by our screening assay. Our screen contained only a few thousand compounds and has already shown high sensitivity for identifying drugs that target cellular pathways of energy metabolism. As we have shown here, compounds emerging from the upper tail of the distribution (Fig. 2b and Supplementary Tables 2 and 4) could serve as valuable lead compounds for prophylaxis against heart attack and stroke, but they may also find broader application for preventing or treating a wide variety of other diseases involving oxidative damage, such as neurodegenera tive disorders. The opposite tail includes dozens of compounds already used as chemotherapeutic agents, perhaps due to their selective toxicity in more rapidly proliferating cells, and may include additional, clinically safe agents that could be useful as adjuvant therapies. The results of our screen may help to pinpoint clinical benefits (or toxicity) of drugs that are not readily attributable to their known targets or mechanism of action. We anticipate that this strategy can be extended to other nutrients—such as fatty acids or ketone bodies. The nutrient-sensitized assay can also be used to screen a much larger library of compounds or even genome-wide RNA interference perturbations to systematically understand drug action and gene function within the broader context of cellular energy homeostasis. Methods Methods and any associated references are available in the online version of the paper at http://www.nature.com/naturebiotechnology/. Availability of raw data. Screening data in the form of cell count per well are available at ChemBank for galactose plates (http://chembank. broadinstitute.org/assays/view-assay.htm?xn=1020.0120) and glucose plates (http://chembank.broadinstitute.org/assays/view-assay. htm?xn=1020.0121). Note: Supplementary information is available on the Nature Biotechnology website. Acknowledgments We thank E. Shoubridge for the MCH58 cell line; M. MacDonald for immortalized striatal cells; R. Xavier for the HRE luciferase construct; S. Norton, B. Wagner and the Broad Chemical Screening Platform for assistance in compound arraying; J. Evans of the Whitehead Institute/MIT BioImaging Center for assistance with high-throughput microscopy; C. Belcher-Timme for technical assistance; T. Kitami for assistance with mitochondrial imaging; M. Mehta for assistance reviewing drug
254
toxicity data; S. Calvo, A. Chess, R. Gould, E. Lander, A. Ting, S. Vafai and members of Mootha lab for valuable discussions and comments. This work was supported by fellowships or grants from the United Mitochondrial Disease Foundation (V.M.G.); Howard Hughes Medical Institute (S.A.S. and V.K.M.); National Institutes of Health (RO1 HL-071158 to P.S.B.); Deane Institute for Integrative Research in Stroke and Atrial Fibrillation (C.A.); American Heart Association (#0815770D to A.P.W.); the Burroughs Wellcome Fund (V.K.M.); the Center for Integration of Medical and Innovative Technology (V.K.M.); and the American Diabetes Association/Smith Family Foundation (V.K.M.). AUTHOR CONTRIBUTIONS V.M.G. and V.K.M. conceived the project; V.M.G., S.A.S., J.H.L., W.C., F.P., C.B.C. and A.P.W. performed experiments; V.M.G., S.A.S., J.H.L., R.N., F.P., C.A., P.S.B. and V.K.M. performed statistical and data analysis; V.M.G., S.A.S. and V.K.M. wrote the paper. 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. Warburg, O. On the origin of cancer cells. Science 123, 309–314 (1956). 2. Bonnet, S. et al. A mitochondria-K+ channel axis is suppressed in cancer and its normalization promotes apoptosis and inhibits cancer growth. Cancer Cell 11, 37–51 (2007). 3. Huber, R., Spiegel, T., Buchner, M. & Riepe, M.W. Graded reoxygenation with chemical inhibition of oxidative phosphorylation improves posthypoxic recovery in murine hippocampal slices. J. Neurosci. Res. 75, 441–449 (2004). 4. Burwell, L.S., Nadtochiy, S.M. & Brookes, P.S. Cardioprotection by metabolic shutdown and gradual wake-up. J. Mol. Cell. Cardiol. 46, 804–810 (2009). 5. Chen, Q., Camara, A.K., Stowe, D.F., Hoppel, C.L. & Lesnefsky, E.J. Modulation of electron transport protects cardiac mitochondria and decreases myocardial injury during ischemia and reperfusion. Am. J. Physiol. Cell Physiol. 292, C137–C147 (2007). 6. Riepe, M.W. et al. Increased hypoxic tolerance by chemical inhibition of oxidative phosphorylation: “chemical preconditioning”. J. Cereb. Blood Flow Metab. 17, 257–264 (1997). 7. Piantadosi, C.A. & Zhang, J. Mitochondrial generation of reactive oxygen species after brain ischemia in the rat. Stroke 27, 327–331 (1996). 8. Kaelin, W.G. Jr. & Ratcliffe, P.J. Oxygen sensing by metazoans: the central role of the HIF hydroxylase pathway. Mol. Cell 30, 393–402 (2008). 9. Kim, J.W., Tchernyshyov, I., Semenza, G.L. & Dang, C.V. HIF-1-mediated expression of pyruvate dehydrogenase kinase: a metabolic switch required for cellular adaptation to hypoxia. Cell Metab. 3, 177–185 (2006). 10. Fukuda, R. et al. HIF-1 regulates cytochrome oxidase subunits to optimize efficiency of respiration in hypoxic cells. Cell 129, 111–122 (2007). 11. Semenza, G.L. Defining the role of hypoxia-inducible factor 1 in cancer biology and therapeutics. Oncogene published online, doi:10.1038/onc.2009.441 (30 November 2009). 12. Fraisl, P., Aragones, J. & Carmeliet, P. Inhibition of oxygen sensors as a therapeutic strategy for ischaemic and inflammatory disease. Nat. Rev. Drug Discov. 8, 139–152 (2009). 13. Hoesch, R.E. & Geocadin, R.G. Therapeutic hypothermia for global and focal ischemic brain injury–a cool way to improve neurologic outcomes. Neurologist 13, 331–342 (2007). 14. Reitzer, L.J., Wice, B.M. & Kennell, D. Evidence that glutamine, not sugar, is the major energy source for cultured HeLa cells. J. Biol. Chem. 254, 2669–2676 (1979). 15. Robinson, B.H., Petrova-Benedict, R., Buncic, J.R. & Wallace, D.C. Nonviability of cells with oxidative defects in galactose medium: a screening test for affected patient fibroblasts. Biochem. Med. Metab. Biol. 48, 122–126 (1992). 16. Marroquin, L.D., Hynes, J., Dykens, J.A., Jamieson, J.D. & Will, Y. Circumventing the Crabtree effect: replacing media glucose with galactose increases susceptibility of HepG2 cells to mitochondrial toxicants. Toxicol. Sci. 97, 539–547 (2007). 17. DeBerardinis, R.J. et al. Beyond aerobic glycolysis: transformed cells can engage in glutamine metabolism that exceeds the requirement for protein and nucleotide synthesis. Proc. Natl. Acad. Sci. USA 104, 19345–19350 (2007). 18. Wu, M. et al. Multiparameter metabolic analysis reveals a close link between attenuated mitochondrial bioenergetic function and enhanced glycolysis dependency in human tumor cells. Am. J. Physiol. Cell Physiol. 292, C125–C136 (2007). 19. Wagner, B.K. et al. Large-scale chemical dissection of mitochondrial function. Nat. Biotechnol. 26, 343–351 (2008). 20. Golenser, J., Waknine, J.H., Krugliak, M., Hunt, N.H. & Grau, G.E. Current perspectives on the mechanism of action of artemisinins. Int. J. Parasitol. 36, 1427–1441 (2006). 21. The Food and Drug Administration Antiemetic drug products for over-the-counter human use; final monograph. Fed. Regist. 52, 15866–15893 (1987).
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
Articles 32. Lu, C.W., Lin, S.C., Chen, K.F., Lai, Y.Y. & Tsai, S.J. Induction of pyruvate dehydrogenase kinase-3 by hypoxia-inducible factor-1 promotes metabolic switch and drug resistance. J. Biol. Chem. 283, 28106–28114 (2008). 33. Aragones, J. et al. Deficiency or inhibition of oxygen sensor Phd1 induces hypoxia tolerance by reprogramming basal metabolism. Nat. Genet. 40, 170–180 (2008). 34. Philipp, S. et al. Stabilization of hypoxia inducible factor rather than modulation of collagen metabolism improves cardiac function after acute myocardial infarction in rats. Eur. J. Heart Fail. 8, 347–354 (2006). 35. Siddiq, A. et al. Hypoxia-inducible factor prolyl 4-hydroxylase inhibition. A target for neuroprotection in the central nervous system. J. Biol. Chem. 280, 41732–41743 (2005). 36. Bernhardt, W.M. et al. Preconditional activation of hypoxia-inducible factors ameliorates ischemic acute renal failure. J. Am. Soc. Nephrol. 17, 1970–1978 (2006). 37. Brahimi-Horn, M.C. & Pouyssegur, J. Harnessing the hypoxia-inducible factor in cancer and ischemic disease. Biochem. Pharmacol. 73, 450–457 (2007). 38. Dirnagl, U., Becker, K. & Meisel, A. Preconditioning and tolerance against cerebral ischaemia: from experimental strategies to clinical use. Lancet Neurol. 8, 398–412 (2009). 39. Giurgea, M. & Puigdevall, J. Experimental teratology with Meclozine. Med. Pharmacol. 15, 375–388 (1966). 40. Lione, A. & Scialli, A.R. The developmental toxicity of the H1 histamine antagonists. Reprod. Toxicol. 10, 247–255 (1996).
© 2010 Nature America, Inc. All rights reserved.
22. Brunton, L.L., Lazo, J.S. & Parker, K.L.. Goodman & Gilman’s The Pharmacological Basis of Therapeutics, edn. 11 (The McGraw-Hill Companies, 2006). 23. Papandreou, I., Cairns, R.A., Fontana, L., Lim, A.L. & Denko, N.C. HIF-1 mediates adaptation to hypoxia by actively downregulating mitochondrial oxygen consumption. Cell Metab. 3, 187–197 (2006). 24. Gidday, J.M. Cerebral preconditioning and ischaemic tolerance. Nat. Rev. Neurosci. 7, 437–448 (2006). 25. Sugino, T., Nozaki, K., Takagi, Y. & Hashimoto, N. 3-Nitropropionic acid induces ischemic tolerance in gerbil hippocampus in vivo. Neurosci. Lett. 259, 9–12 (1999). 26. Ratan, R.R. et al. Translation of ischemic preconditioning to the patient: prolyl hydroxylase inhibition and hypoxia inducible factor-1 as novel targets for stroke therapy. Stroke 35, 2687–2689 (2004). 27. Lesnefsky, E.J. et al. Blockade of electron transport during ischemia protects cardiac mitochondria. J. Biol. Chem. 279, 47961–47967 (2004). 28. Jeong, D.W., Kim, T.S., Cho, I.T. & Kim, I.Y. Modification of glycolysis affects cell sensitivity to apoptosis induced by oxidative stress and mediated by mitochondria. Biochem. Biophys. Res. Commun. 313, 984–991 (2004). 29. Hunter, A.J., Hendrikse, A.S. & Renan, M.J. Can radiation-induced apoptosis be modulated by inhibitors of energy metabolism? Int. J. Radiat. Biol. 83, 105–114 (2007). 30. Vaughn, A.E. & Deshmukh, M. Glucose metabolism inhibits apoptosis in neurons and cancer cells by redox inactivation of cytochrome c. Nat. Cell Biol. 10, 1477–1483 (2008). 31. Ramirez, J.M., Folkow, L.P. & Blix, A.S. Hypoxia tolerance in mammals and birds: from the wilderness to the clinic. Annu. Rev. Physiol. 69, 113–143 (2007).
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
255
ONLINE METHODS All experiments were done in accordance with the national and institutional guidelines for animal welfare, adhering to protocols approved by the institutional subcommittee on research animal care.
© 2010 Nature America, Inc. All rights reserved.
Cell culture. Immortalized MCH58 human diploid fibroblasts containing the pLKO.1 vector were grown in DMEM high-glucose medium (Invitrogen) with 10% FBS (Sigma), 1× penicillin, streptomycin and glutamine (Invitrogen), 2 µl/ml puromycin and 50 µg/ml uridine at 37 °C and 5% CO2. The highglucose medium was replaced with 10 mM glucose or 10 mM galactose wherever indicated. All media contained 1 mM pyruvate and 4 mM glutamine. MCH58 fibroblasts (without pLKO.1), HeLa and HEK293 cells were grown in DMEM high-glucose medium with 10% FBS at 37 °C and 5% CO2; STHdhQ7/7 mouse striatal cells were grown at 33 °C. Measurement of cellular OCR and ECAR. OCR and ECAR measurements were carried out as previously described18 with minor modifications. Briefly, MCH58 fibroblasts were seeded in XF24-well cell culture microplates (Seahorse Bioscience) at 30,000 cells/well in 10 mM glucose- or 10 mM galactose– containing media and incubated at 37 °C and 5% CO2 for ~20 h. Before the measurements were made, the growth medium was replaced with ~925 µl of assay medium (with 10 mM glucose or 10 mM galactose as the sugar source) and cells were incubated at 37 °C for 60 min. The OCR and ECAR measurements on HeLa, HEK293 and STHdhQ7/7 mouse striatal cells were carried out by growing them in XF24 plates at 40,000/well (HeLa and STHdhQ7/7 mouse striatal cells) or 50,000/well (HEK293 cells) for ~20 h under their regular growth conditions, as described in the previous section. The assay medium was the same as above except 25 mM glucose was used as the sugar source. Three baseline measurements were recorded before the addition of compounds. For the secondary screening assay on selected compounds listed in Figure 2c, MCH58 fibroblasts were grown in 25 mM glucose-containing medium in the presence of a compound for 16–20 h. Each compound was tested at screening concentration—antimycin (3.65 µM), ascorbate (10 µM), clofilium tosylate (5.9 µM), nifuroxazide (7.27 µM), meclizine (10 µM), menadione (11.62 µM), clemastine (5.82 µM), vinpocetine (5.71 µM), bisacodyl (5.53 µM), nonoxynol-9 (10 µM), sertraline (10 µM), thonzonium (3.91 µM), chlorhexidine (3.96 µM), mefloquine (10 µM) and alexidine (3.93 µM). OCR measurements on the mouse primary cortical neurons obtained from day E14–15 embryos were carried out in their regular growth medium. Cell viability assays. MCH58 fibroblasts were seeded in 96-well plates at 5,000 cells/well in DMEM high-glucose medium. After ~20 h, cells were washed in PBS and the growth medium was replaced with 10 mM glucose- or 10 mM galactose-containing media containing different concentrations of meclizine, rotenone, antimycin, oligomycin or DMSO (0.1%). The cells were then grown for 72 h and cell viability was assayed by the CellTiter-Glo Luminescent Viability assay (Promega). Chemical screening and high-throughput assay of cell number quantification. MCH58 fibroblasts were seeded at 5,000 cells/well using the robotic MultiDrop Combi (ThermoFisher Scientific) dispenser into 96-well plates (PerkinElmer) at 100 µl per well in DMEM high-glucose medium. Twenty-four hours later, cells were washed twice in PBS and medium was replaced with either 10 mM glucose- or 10 mM galactose-containing media. Approximately 100 nl of each compound was pin-transferred in duplicate into the plates with a steel pin array using the CyBi-Well robot (CyBio). The compound collection of 3,695 drugs includes two commercially available libraries (Spectrum and Prestwick). Compound-treated plates were incubated at 37 °C for 72 h. Cells were then washed once with PBS, stained with 10 µM Hoechst 33342 (Invitrogen) and fixed in 3.7% formaldehyde solution for 15 min. Wells were then washed once and stored in PBS. Cell culture plates were stored at 4 °C until the time of imaging, which was at most 24 h after fixation. Imaging was performed by Arrayscan VTi automated microscope (ThermoFisher Scientific) with the use of an automated plate stacker at the Whitehead Institute Biomedical Imaging Center. Four nonoverlapping images at 5× magnification (NA 0.25) were acquired for a field of view of 1.3 mm × 1.3 mm per image. Image analysis was performed using the freely available open-source software
nature biotechnology
package CellProfiler41. Images were analyzed individually by first identifying nuclei using the Hoechst staining signal. Total number of nuclei per field were recorded, and the sum of nuclei from four images per well gave cell count per well. Computation was performed using a UNIX computer cluster. The unprocessed data from our screen in the form of cell count per well are available in ChemBank. Please note that the ChemBank data sets include statistical calculations that were not used in our final analysis. For a detailed description of our statistical methods, please refer to the section below “Statistical analysis of screening data,” which outlines the method used to calculate our summary statistic, Sglu/gal. The Sglu/gal value for every compound included in our screen can be found in Supplementary Table 1. Assays of mitochondrial physiology. Mitochondria were isolated from C57BL/6 mouse kidneys by differential centrifugation and resuspended in experimental buffer as described previously42, to a final concentration of 0.5 mg/ml. State 2 respiration was initiated with the addition of 12.5 mM glutamate and 12.5 mM malate, or 20 mM pyruvate and 1 mM malate or 12 mM succinate and 3 µM rotenone. State 3 respiration was initiated with the addition of 0.2 mM ADP, and uncoupled respiration was initiated with the addition of 5 µM carbonyl cyanide m-chlorophenyl hydrazone. Mitochondria were incubated for 5 min in respiration buffer pH 7.4 supplemented with either meclizine (50 µM) or DMSO. O2 consumption was monitored with a Fiber Optic Oxygen sensor probe (Ocean Optics) at 25 °C, and NADH (endogenous, 370 ± 7 nm excitation, 440 ± 4 nm emission) and membrane potential (1.25 µM TMRM, 546 ± 7 nm excitation, 590 ± 4 nm emission) were measured with a Perkin-Elmer LS50B luminescence spectrometer. Drug testing in cardiac ischemia-reperfusion injury. Adult rat ventricular cardiomyocytes were isolated by endotoxin-free collagenase perfusion, and the simulated ischemia-reperfusion injury was performed as previously described43. Briefly, ischemia comprised 1 h in anoxic glucose-free buffer at pH 6.5, and reperfusion comprised 30 min in normoxic glucose-replete buffer at pH 7.4. Cell viability was monitored by Trypan blue exclusion. Rat cardiac mitochondria were isolated, and respiration of both cells and mitochondria was measured using a Clark oxygen electrode, as previously described 44. Isolated rat hearts were retrograde reperfused in Langendorff mode under constant flow as described previously45. Drug testing in neuronal ischemia-reperfusion injury. Plasma concentrations of meclizine were determined after intraperitoneal injections in C57BL/6 mice. Injections of 100 mg/kg meclizine or vehicle control were followed by cardiac puncture blood draws at 1 h or 6 h. Absolute concentrations of meclizine in plasma were measured using liquid chromatography tandem mass spectrometry against a purified standard. To check the protective effect of the tested compounds, male C57BL/6 mice were treated with two intraperitoneal injections of 100 mg/kg meclizine, 20 mg/kg pyrilamine and 0.5 mg/kg scopolamine or vehicle, 17 h and 3 h before ischemia. Drug doses for pyrilamine and scopolamine were chosen based on previous literature evidence of in vivo brain bioavailability46,47. The experimenter was blinded to treatment groups. Mice were anesthetized with isoflurane (2.5% induction, 1.5% maintenance, in 70% N2O/30% O2), and subjected to 1 h transient middle cerebral artery occlusion using an intraluminal filament inserted through the external carotid artery. Regional cerebral blood flow was monitored using a laser Doppler probe placed over the core middle cerebral artery territory. Rectal temperature was controlled at 37 °C by a servo-controlled heating pad. Total infarct volume was calculated on 2,3,5-triphenyltetrazolium chloride (TTC)-stained 1-mm thick coronal sections by integrating the infarct areas in each of ten slice levels. Infarct volume was calculated using the ‘indirect method’, that is, contralateral hemisphere minus ipsilateral non-infarcted volume. Data were expressed as mean ± s.d. One-way ANOVA followed by Tukey’s multiple comparison test was used for analysis of values between groups. Neurological deficit scores were analyzed by Kruskal-Wallis nonparametric ANOVA followed by Dunn’s multiple comparisons test. P < 0.05 was considered statistically significant. Western blot analysis. HIF1-α and HIF2-α stabilization was assessed in MCH58 fibroblasts and HeLa cells by carrying out western blot detection using anti-HIF1-α (Cell Signaling) and anti-HIF2-α (Novus Biologicals) antibodies.
doi:10.1038/nbt.1606
Protein was extracted from cells pretreated with either 0.1% DMSO, 50 µM meclizine or 100 µM deferoxamine (Sigma). SDS PAGE was performed on 15 µg/lane protein using NuPAGE 4–12% Bis-Tris gel from Invitrogen. Western blot analysis was performed as per the standard procedures. β-actin was used as a loading control.
© 2010 Nature America, Inc. All rights reserved.
HIF reporter assay. Luciferase activity in HeLa cells transiently transfected with HIF response element (HRE)/luciferase construct was measured by the Dual-Luciferase Reporter Assay System (Promega). The drug treatment was initiated ~24 h after the transfection and the treatment was continued for 6 h with 0.1% DMSO, 100 µM deferoxamine or 50 µM meclizine. HeLa cells were transfected using the Fugene reagent (Roche) as per the manufacturer’s instruction. HRE reporter construct was a generous gift from R. Xavier, Massachusetts General Hospital. Assays of mitochondrial abundance, morphology and membrane potential in intact cells. Mitochondrial (mt) morphology, membrane potential, mtDNA content and mtDNA expression were measured in HeLa cells treated with 50 µM of meclizine over a 6-h period. MitoTracker CMXRos (Invitrogen) and Hoechst 33342 staining of live cells was performed as per manufacturer’s recommendation. The stained cells were observed at 20× magnification with an Olympus CKX41 microscope and fluorescent images were captured with a QiCAM camera. The mitochondrial membrane potential was measured with a ratiometric dye, JC-1 by measuring relative fluorescent units at 590 nm and 535 nm using a fluorescent plate reader. The mtDNA copy number and mtRNA transcripts were measured by quantitative real-time PCR assays as described previously48. Drug annotation. All FDA-approved drugs were annotated after confirming their entry in the Orange book, 29th edition 2009. For other clinically used compounds, we used the corresponding PubMed and PubChem entries. Statistical analysis of screening data. Normalized cell counts were computed separately for each 96-well plate by dividing the cell count of each
doi:10.1038/nbt.1606
well with the trimmed mean (the average after discarding the largest and smallest value) of the cell counts for the 16 DMSO-treated wells on each plate. A single 96-well plate exhibited very low counts in all wells and was discarded. We then removed measurements that failed to replicate by requiring that the ratio of normalized counts was <1.5 between replicates; this excluded 49 wells. The remaining replicate measurements were averaged, and fold changes were computed as the ratio between the averaged normalized counts for the glucose and galactose screens. The Sglu/gal for each compound was calculated by taking the log10 of the normalized and averaged fold change in glucose divided by galactose. To evaluate statistical significance, we computed Z-scores for each well against the DMSO (null) distribution and averaged Z-scores across replicates. Wells with average Z-score > 2.5 were considered significant.
41. Carpenter, A.E. et al. CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol. 7, R100 (2006). 42. Mootha, V.K., Arai, A.E. & Balaban, R.S. Maximum oxidative phosphorylation capacity of the mammalian heart. Am. J. Physiol. 272, H769–H775 (1997). 43. Wojtovich, A.P. & Brookes, P.S. The complex II inhibitor atpenin A5 protects against cardiac ischemia-reperfusion injury via activation of mitochondrial KATP channels. Basic Res. Cardiol. 104, 121–129 (2009). 44. Wojtovich, A.P. & Brookes, P.S. The endogenous mitochondrial complex II inhibitor malonate regulates mitochondrial ATP-sensitive potassium channels: implications for ischemic preconditioning. Biochim. Biophys. Acta 1777, 882–889 (2008). 45. Nadtochiy, S.M., Tompkins, A.J. & Brookes, P.S. Different mechanisms of mitochondrial proton leak in ischaemia/reperfusion injury and preconditioning: implications for pathology and cardioprotection. Biochem. J. 395, 611–618 (2006). 46. Miyazaki, S., Imaizumi, M. & Onodera, K. Effects of thioperamide, a histamine H3-receptor antagonist, on a scopolamine-induced learning deficit using an elevated plus-maze test in mice. Life Sci. 57, 2137–2144 (1995). 47. Toyota, H. et al. Behavioral characterization of mice lacking histamine H(3) receptors. Mol. Pharmacol. 62, 389–397 (2002). 48. Baughman, J.M. et al. A computational screen for regulators of oxidative phosphorylation implicates SLIRP in mitochondrial RNA homeostasis. PLoS Genet. 5, e1000590 (2009).
nature biotechnology
Articles
Harnessing chaperone-mediated autophagy for the selective degradation of mutant huntingtin protein
© 2010 Nature America, Inc. All rights reserved.
Peter O Bauer1, Anand Goswami1, Hon Kit Wong1, Misako Okuno1, Masaru Kurosawa1, Mizuki Yamada1, Haruko Miyazaki1, Gen Matsumoto1, Yoshihiro Kino1, Yoshitaka Nagai2 & Nobuyuki Nukina1 Huntington’s Disease (HD) is a dominantly inherited pathology caused by the accumulation of mutant huntingtin protein (HTT) containing an expanded polyglutamine (polyQ) tract. As the polyglutamine binding peptide 1 (QBP1) is known to bind an expanded polyQ tract but not the polyQ motif found in normal HTT, we selectively targeted mutant HTT for degradation by expressing a fusion molecule comprising two copies of QBP1 and copies of two different heat shock cognate protein 70 (HSC70)– binding motifs in cellular and mouse models of HD. Chaperone-mediated autophagy contributed to the specific degradation of mutant HTT in cultured cells expressing the construct. Intrastriatal delivery of a virus expressing the fusion molecule ameliorated the disease phenotype in the R6/2 mouse model of HD. Similar adaptor molecules comprising HSC70–binding motifs fused to an appropriate structure-specific binding agent(s) may have therapeutic potential for treating diseases caused by misfolded proteins other than those with expanded polyQ tracts. HD is an autosomal dominant neurodegenerative disorder caused by expansion of an N-terminal polyQ tract in one of the two alleles encoding HTT protein1. A prominent feature of this disorder is progressive neurodegeneration, with the intranuclear and cytoplasmic accumulation of aggregated mutant HTT inside neurons2,3. The expanded polyQ tract in mutant HTT forms a β-sheet structure, which causes formation of fibrillar and nonfibrillar aggregates4,5 and mediates aberrant interactions with transcription factors, disrupting the regulation of transcription6–9. Although the pathological significance of the expanded polyQ tracts in mutant HTT has been clearly established, treatments that can prevent physical and mental decline associated with HD have not yet been approved. The therapeutic potential of downregulating expression of the mutant HTT allele was first demonstrated in a tetracycline-regulated mouse model of HD10. Whereas nuclear inclusions and behavioral abnormalities appeared after expression of the mutant HTT was induced, the inclusions disappeared and the behavioral phenotype was ameliorated when expression of the mutant protein was terminated. Subsequent efforts have focused on either specifically inhibiting expression of the mutant HTT allele by RNA interference (RNAi) mediated by small interfering RNAs (siRNAs)11,12 or short hairpin RNAs (shRNAs)13,14, or by enhancing the degradation of mutant HTT. The latter category of approaches has involved augmenting the role that macroautophagy normally plays in the clearance of mutant HTT15,16, enhancing the proteasomal degradation of mutant HTT17 or simultaneous activation of both macroautophagy and proteasomal degradation18. In this study, we investigated the possibility of recruiting the selective type of autophagy, chaperone-mediated autophagy (CMA), to
s pecifically promote degradation of mutant HTT without affecting levels of normal HTT encoded by the second allele. Macroautophagy involves the formation of double membrane vesicular structures, autophagosomes, which nonspecifically sequester parts of the cytoplasm and ultimately fuse with protease-containing lysosomes19. In CMA, on the other hand, specific proteins are recognized by HSC70, which together with other chaperones and cochaperones, channels them to the surface of lysosomes. After binding to Lamp2a, the targeted proteins are then translocated across the lysosomal membrane and degraded20,21. We show that expression of a 46 amino acid–peptide adaptor molecule comprising two copies of the polyglutamine-binding peptide 1 sequence22 (in this study, we called this peptide as QBP1 or Q) and two different HSC70-binding motifs (abbreviated as HSC70bm or H) directs HTT with an expanded polyQ tract to the CMA machinery for degradation. Viral delivery of our QBP1-HSC70bm construct to the striatum in the R6/2 mouse model of HD strongly reduces polyQ aggregation in the transduced areas, ameliorates symptoms and extends the life span. Adaptations of this strategy may be applicable to other neurological diseases caused by expanded polyQ tracts, as well as to disorders such as Alzheimer’s disease and Parkinson’s disease, which are also caused by misfolded proteins. RESULTS Incorporation of HSC70bm into polyQ proteins enhances their degradation As only proteins that are specifically recognized by the HSC70containing chaperone complex are degraded by CMA, we first assessed whether two HSC70bm sequences were able to target proteins with an expanded polyQ tract for lysosomal degradation. We introduced
1Laboratory
for Structural Neuropathology, RIKEN Brain Science Institute, Wako-shi, Saitama, Japan. 2Department of Degenerative Neurological Diseases, National Center of Neurology and Psychiatry, Kodaira, Tokyo, Japan. Correspondence should be addressed to N.N. ([email protected]). Received 25 September 2009; accepted 25 January 2010; published online 28 February 2010; doi:10.1038/nbt.1608
256
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
Articles in mouse neuroblastoma (Neuro2a) cells and compared the effects with those observed 60Q Venus Hoechst PolyQ Merge with the other fusions to monomeric red K F ERQ fluorescent protein (mRFP, hereafter referred to as R) listed in Figure 2a. When transfected 60Q P = 0.0062 c P = 0.0264 into 150Q Neuro2a, a stable Neuro2a cell 60Q P = 0.0343 60Q-HSC70bm line with ecdysone-inducible expression of 1.2 1.0 a tNHTT-150Q fusion to enhanced green 0.8 fluorescent protein (tNHTT-150Q-EGFP), 60Q-HSC70bm 0.6 the QBP1-containing molecules (RQ and 0.4 RHQ; Fig. 2a) co-localized with inclusions 0.2 containing tNHTT-150Q-EGFP (Fig. 2b). 0 No 3-MA Pepstatin Leupeptin However, expression of RHQ efficiently treatment reduced the accumulation of soluble and Figure 1 Inclusion of two HSC70bm sequences within the protein carrying expanded polyQ tracts insoluble aggregates of tNHTT-150Q-EGFP decreases its aggregation. (a) Schematic representation of 60Q-HSC70bm. (b) PC12 cells transfected compared to RQ and other tested constructs with either 60Q or 60Q-HSC70bm. The presence of two HSC70bm motifs in 60Q-HSC70bm reduced the number of inclusions as compared to 60Q (scale bar, 20 µm). (c) Quantification of the PC12 cells (Fig. 2c,d). The 16Q Neuro2a cell line stably with inclusions 24 h after transfection with either 60Q or 60Q-HSC70bm. 10 mM 3-methyladenine expresses an ecdysone-inducible copy of (3-MA), 10 µM pepstatin A and 10 µM leupeptin were added to inhibit macroautophagy, the proteases, tNHTT fused to 16 glutamine residues, the cathepsins D and E, and cathepsin A, respectively. Bars in c represent relative mean values ± s.e.m. number encoded by a normal HTT allele. Levels from four independent experiments, with levels of aggregation observed for 60Q normalized to a value of tNHTT-16Q-EGFP in 16Q Neuro2a cells of 1. Quantifications were performed by ArrayScan. were not affected by RQ or RHQ (Fig. 2c,e). Expression of RHQ reduced the aggregation a copy of the HSC70-binding motif of α-synuclein (VKKDQ), which of both tNHTT-150Q-EGFP and tNHTT-150Q with nls (tNHTTis known to be degraded by CMA23, and a copy of the consensus 150Q-nls-EGFP) relative to expression of RQ (Fig. 2f,g). Besides HSC70-binding motif (KFERQ) (together referred to as HSC70bm), reducing the number of inclusions in Neuro2a cell lines expressing between an N-terminally truncated HTT (tNHTT, the sequence tNHTT-150Q-EGFP and tNHTT-150Q-nls-EGFP, RHQ also reduced encoded by the first 90 N-terminal residues of exon 1 of the human the average size of the inclusions (Supplementary Fig. 2). We investigated the effect of the QBP1-containing molecules HTT gene) sequence fused to 60 tandem glutamine residues (60Q) and the fluorescent reporter protein Venus. We then transiently (Q and HQ) on polyQ-related cytotoxicity by using a fluorescent expressed this construct (60Q-HSC70bm) and a control lacking marker for cell toxicity, propidium iodide (PI) and by analyzing the HSC70bm (60Q) in rat pheochromocytoma (PC12) cells. The 60Q- cleavage of procaspase-3. As both mRFP and PI have red fluorescence, HSC70bm construct (Fig. 1a) generated fewer aggregates (42.4% we used constructs without mRFP in these experiments. The Q and reduction) than cells expressing 60Q (Fig. 1b,c). In the case of 60Q HQ constructs were able to decrease the aggregation and number of PIcarrying a nuclear localization signal (nls), inclusion of two copies positive cells compared to their control counterparts, scrambled QBP1 of HSC70bm (60Q-HSC70bm-nls) reduced nuclear aggregation by (S) and HSC70bm-scrambled QBP1 (HS), respectively in both 150Q 36.2% of that seen with 60Q-nls expression, with decreased trans- and 150Q-nls Neuro2a cell lines. Scrambled QBP1 was produced by location of the protein into nuclei (Supplementary Fig. 1). Whereas rearranging the 22 amino acids in QBP1 (for the nucleotide sequence pepstatin A–mediated inhibition of the two primary lysosomal pro- see Online Methods). Schematic representations of all four constructs teases, cathepsins D and E, alleviated the effect of incorporating used in this set of experiments are shown in Supplementary Figure 3a HSC70bm into the primary sequence, inhibition of macroautophagy and the amino acid sequences are provided in Supplementary Table 1. (autophagosome formation) by 3-methyladenine did not. Under HQ decreased the cell death associated with expression and aggregaphysiological conditions, cathepsin A is located at the lysosomal tion of tNHTT-150Q-EGFP and tNHTT-150Q-nls-EGFP, which was membrane, where it triggers cleavage of Lamp2a and negatively accompanied by reduced caspase-3 activation as compared to HS and Q regulates CMA24. Treatment with the cathepsin A inhibitor, leupeptin, (Supplementary Fig. 3b–g). substantially accentuated the differences in aggregation observed after transfection with either 60Q or 60Q-HSC70bm (Fig. 1c). These The RHQ adaptor targets proteins with an expanded polyQ experiments suggested that the presence of HSC70bm could enhance tract for degradation by CMA the degradation of a protein with an expanded polyQ tract in a We next examined the mechanism that accounts for the enhanced capacity of RHQ to inhibit aggregation of tNHTT with 60 glutamine CMA-dependent manner. residues relative to the weaker effects seen using RQ. Expression of QBP1 inhibits toxic conformational transition of the expanded polyQ A QBP1-HSC70bm peptide reduces polyQ aggregation The synthetic QBP1 peptide binds specifically to expanded polyQ stretch in mutant HTT, which is thought to be a trigger for its oligo tracts in several proteins and inhibits their aggregation by prevent- merization and aggregation25,26. This probably increases the accessiing their oligomerization22,25. To test whether we could specifically bility of the mutant protein to degradation systems. As the addition target HTT proteins with an expanded polyQ tract for CMA-mediated of HSC70bm intensified the effect of QBP1 on polyQ aggregation degradation, we designed a 46 amino acid–peptide adaptor molecule and protein levels, we examined and compared the rates of degracomprising the same HSC70bm (hereafter referred to as H) used in dation caused by RQ and RHQ. We performed a chase experiment 60Q-HSC70bm fused to two copies of QBP1 (hereafter referred to as using Neuro2a cells with ecdysone-inducible expression of tNHTTQ). We then explored the effect of this construct on the aggregation EGFP with 60Q (60Q Neuro2a) and found that RHQ enhanced the of fusions of tNHTT to either 150 or 16 tandem glutamine repeats degradation of the soluble tNHTT-60Q-EGFP protein more than R, RS,
© 2010 Nature America, Inc. All rights reserved.
b
P = 0.0011
P = 0.0024
V K KDQ
P = 0.0019
Level of aggregation in transfected PC12 cells
a
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
257
a
R
mRFP
RH
mRFP mRFP
Sc QBP1
mRFP
Hsc70bm
RQ
mRFP
QBP1
RHQ
mRFP
RS aggregated tNHTT-150Q-EGFP RHS
Anti-GFP kDa kDa
100 75
37
25 β-tubulin 50
50
Anti-RFP
β-tubulin
RQ
tNHTT-60Q-EGFP β-tubulin tNHTT-60Q-EGFP
RHQ
RHQ
RS
RHS
RQ
RHQ
g
1.4 1.2 1.0 0.8 0.6 0.4 0.2 0 R
RH
RS
3-MA
0 12 24 36 48 12 24 36 12 24 36
NH4Cl 12 24 36
RHS
RQ
RHQ
P = 0.0024
35 30 25 20 15 10 5 0
P = 0.000001 P = 0.00004
R
RH
RS
RHS P = 0.019
25 20
RQ
RHQ
P = 0.00081 P = 0.00097
15 10 5 0 R
RH
RS
RHS
RQ
RHQ
proteasomal degradation and macroautophagy by 5 µM MG132 and 10 mM 3-methyladenine (Supplementary Fig. 4), respectively, had almost no effects on RHQ-mediated degradation of tNHTT-60Q-EGFP after 12, 24 and 36 h of exposure to these inhibitors (Fig. 3). These
b 12 h 1.2 1.0 0.8 0.6 0.4 0.2 0
1.2 1.0 0.8 0.6 0.4 0.2 0
24 h
36 h
48 h
No drug
R
RS
RHS
RQ
3MA
R
RS
RHS
RHQ
RQ
RHQ
1.2 1.0 0.8 0.6 0.4 0.2 0
1.2 1.0 0.8 0.6 0.4 0.2 0
P = 0.045 P = 0.0025 P = 0.0011
RH
Percent of transfected 150Q Neuro2a cells with inclusions
R
P = 0.007
P = 0.04
f
Percent of transfected 150Q-nls Neuro2a cells with inclusions
e
P = 0.031 P = 0.0019 P = 0.047
Insoluble Soluble
1.4 1.2 1.0 0.8 0.6 0.4 0.2 0
16Q
P = 0.013
d
RQ
Soluble tNHTT-polyQ -EGFP
150Q
tNHTT-60Q-EGFP β-tubulin
R RH RS RHS RQ RHQ
25
Relative levels of tNHTT-60Q-EGFP
RHS
tNHTT-60Q-EGFP
Merge
RH
R RH RS RHS RQ RHQ
Relative levels of tNHTT-60Q-EGFP
RS
β-tubulin
mRFP
QBP1
Hsc70bm
tNHTT-60Q-EGFP R
tNHTT-150Q -EGFP
Sc QBP1
P = 0.039 P = 0.0011 P = 0.00072 P = 0.00011
MG132
Hoechst R
P = 0.037 P = 0.00047 P = 0.0024
No drug h
Hsc70bm
RS
RHS and RQ 12, 24, 36 and 48 h after the promoter shut-off (removal of the transgene expression inducer) (Fig. 3). The RHQ-mediated degradation of tNHTT-60Q-EGFP was inhibited by ammonium chloride, a classical inhibitor of lysosomal degradation. However, inhibition of
a
b
RHS
c
Relative levels of tNHTT-150Q-EGFP (WB)
Figure 2 RHQ inhibits polyQ aggregation more efficiently than RQ. (a) Schematic representations of all constructs tested. R, mRFP; RH, mRFP fused to HSC70bm (VKKDQ and KFERQ); RS, mRFP fused to scrambled QBP1 (Online Methods); RHS, mRFP fused to HSC70bm and scrambled QBP1; RQ, mRFP fused to QBP1; RHQ, mRFP fused to HSC70bm and QBP1 (Sc, scrambled). (b) Confocal images of the 150Q Neuro2a cells transfected with R, RH, RS, RHS, RQ and RHQ. Molecules containing QBP1 co-localized with the tNHTT150Q-EGFP inclusions (bar, 5 µm). Cells with inclusions were chosen to illustrate the colocalization of QBP1-containing molecules with tNHTT-150Q-EGFP. (c) The expression of RQ and RHQ in 150Q Neuro2a cells decreased the accumulation of insoluble tNHTT-150Q-EGFP on the gel top. Note the shift of tNHTT-150QEGFP from an insoluble to a soluble form when RQ was co-expressed. The levels of tNHTT-16QEGFP were not affected. Full-length western blots are presented in Supplementary Figure 13. (d) RHQ reduced the levels of aggregated and soluble forms of tNHTT-150Q-EGFP by 81.8% and 33.6%, respectively. WB, western blot. (e) Quantification of tNHTT-16Q-EGFP protein levels. (f) ArrayScan analysis revealed 28% and 56% reductions in inclusion formation in 150Q Neuro2a cells by RQ and RHQ, respectively, after 24 h of differentiation and induction. (g) In 150Q-nls Neuro2a cells, expression of RQ and RHQ reduced the number of tNHTT-150Q-nlsEGFP inclusion by 23% (RQ) and 53% (RHQ) after 48 h of differentiation and induction compared to RS and RHS, respectively. Bars in d and e represent mean values ± s.e.m. from four independent experiments. Quantifications were performed by ArrayScan.
Relative levels of tNHTT-16Q-EGFP (WB)
© 2010 Nature America, Inc. All rights reserved.
Articles
MG132
R
RS
RHS
RQ
RHQ
RQ
RHQ
NH4Cl
R
RS
RHS
β-tubulin
Figure 3 RHQ enhances the degradation of tNHTT-60Q-EGFP. (a) Western blot analysis of the chase experiment in 60Q Neuro2a cells. Full-length western blots are presented in Supplementary Figure 13. (b) RHQ enhanced the tNHTT-60Q-EGFP degradation during the chase period. 5 µM MG132, 10 mM 3-methyladenine and 15 mM ammonium chloride (NH4Cl) were used to inhibit proteasome, macroautophagy and lysosomal degradation, respectively. NH4Cl had no effect on proteasome activity and did not affect cell viability (Supplementary Fig. 4a,b). Bars in b represent relative mean values ± s.e.m. from three independent experiments. Protein levels at time 0 represent the control condition (arbitrarily set as 1).
258
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
Articles a
tNHTT16Q-EGFP Lysotracker Merge
b
tNHTT150Q-EGFP Lysotracker Merge
R
RS
e
RQ RHS RHQ
PNS
S
S
c
EM48
shRNA pSil only HSC70 Lamp2a
Lyso
RS
Anti-CathD
Lyso PNS HQ
1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0
PNS (EM48/tubulin) Lyso (EM48/cathepsin D)
R
RS
RQ
RHS RHQ
f Percent of RFP-positive 150Q Neuro2a cells with aggregates
RHS
Anti-RFP
RQ
Lyso pSil Hsc70 35 30 25 20 15 10 5 0
Lamp2a Hsc70+Lamp2a P = 0.0006
RHQ
g
P = 0.0000009 P = 0.00003 P = 0.037
pSil only HSC70 shRNA Lamp2a shRNA
+ – –
– + –
R RS RHS RQ RHQ R RS RHS RQ RHQ
Anti-GFP
– – +
– + +
R RS RHS RQ RHQ R RS RHS RQ RHQ
Soluble tNHTT-150Q -EGFP
Anti-Lamp2a Anti-HSC70 R
RS RHS RQ RHQ
Anti-RFP Anti-β-tubulin
h
pSil HSC70 Lamp2a HSC70+Lamp2a Figure 4 RHQ induces the degradation of tNHTT with expanded polyQ tracts by recruiting the P = 0.0001 CMA machinery. (a,b) Constructs identical to those described in Figure 2a, but lacking mRFP P = 0.0006 1.2 (representations of S, HS, Q, HQ are shown in Supplementary Fig. 3a) were co-transfected with 1.0 tNHTT-16Q-EGFP (a) and tNHTT-150Q-EGFP (b) constructs into Neuro2a cells (bar, 4 µm). (c) Dot 0.8 blot analysis of the lumenal fraction of purified lysosomes from Neuro2a cells co-transfected with 0.6 0.4 tNHTT-60Q-EGFP and R, RS, RHS, RQ or RHQ. PNS, post-nuclear supernatant. (d) Quantification of 0.2 HTT translocation into lysosomes. Levels of HTT detected by EM48 (anti-huntingtin antibody) were 0 normalized to levels of β-tubulin (in PNS) and levels of cathepsin D (in lysosomes). (e) Compound R RS RHS RQ RHQ images generated by ArrayScan illustrating inclusion formation in 150Q Neuro2a cells transfected with R, RS, RHS, RQ and RHQ and shRNA specific for HSC70 and/or Lamp2a. (blue, Hoechst 33258; green, tNHTT-150Q-EGFP; red, tested molecule). (f) RNAi of HSC70, Lamp2a or both alleviated the inhibitory effect of RHQ on the formation of inclusions containing tNHTT-150Q-EGFP inclusions. (g) Western blot analysis of soluble polyQ protein in 150Q Neuro2a cells transfected with tested constructs and shRNA for HSC70 and/or Lamp2a. Full-length western blots are presented in Supplementary Figure 13. (h) Silencing of HSC70, Lamp2a or both alleviated the inhibitory effect of RHQ on levels of soluble tNHTT-150Q-EGFP. Bars in d and f represent mean values ± s.e.m. from three independent experiments. Bars in h represent relative mean values ± s.e.m. from three independent experiments. For h, values for the R construct were arbitrarily set to 1. Relative levels of soluble tNHTT-150Q-EGFP (WB)
© 2010 Nature America, Inc. All rights reserved.
d
– + +
PNS
Q
HQ
– – +
β-tubulin
Lyso Q
– + –
R
PNS
HS
HS
+ – –
results suggested that the effects of RHQ expression on polyQ degradation were mediated primarily by CMA under the conditions we used. To confirm the enhanced degradation of the polyQ protein, we used a Neuro2a cell line system stably expressing tNHTT tagged with kikume green (kikGR) fluorescent tag and containing 62 tandem glutamine residues (tNHTT-62Q-kikGR). As the kikGR protein is truncated by specific cleavage after irradiation at a wavelength of ~400 nm 27, this cleaved form of kikGR provides an extremely reliable indication of protein degradation. We irradiated tNHTT-62Q-kikGR Neuro2a cells at 400 nm for 5 min and analyzed the levels of tNHTT-62Q-kikGR in these cells 12 h later. Whereas RHQ decreased the levels of the cleaved form of tNHTT-62Q-kikGR protein by 45.6%, the full-length (continuously expressed) form of tNHTT-62Q-kikGR showed a 24% decrease relative to RQ (Supplementary Fig. 5). These results demonstrated the effectiveness of RHQ in enhancing tNHTT-polyQ degradation as compared to the effects of RQ and the control molecules. To investigate whether HQ targets tNHTT fused to either 16 or 150 glutamine residues to lysosomes, we co-expressed S, HS, Q and HQ constructs with tNHTT-16Q-EGFP or tNHTT-150Q-EGFP in Neuro2a cells for 16 h. In the cells expressing tNHTT-16Q-EGFP, none of the co-transfected constructs had any effect on the subcellular distribution of the green fluorescence (Fig. 4a and Supplementary Fig. 6a).
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
On the other hand, when tNHTT-150Q-EGFP was co-expressed with HQ, the fluorescence intensity decreased and redistributed to co-localize partly with the lysosomal marker lysotracker-rhodamine28 (Fig. 4b and Supplementary Fig. 6b). To confirm this observation, we co-transfected the R, RS, RHS, RQ and RHQ constructs with tNHTT-60Q-EGFP into Neuro2a cells and purified intact lysosomes 16 h later. After disrupting the lysosomes and removing the membranes by centrifugation, we analyzed the lumenal fraction for the presence of tNHTT-60Q-EGFP. We observed a dramatic shift in the distribution of soluble tNHTT60Q-EGFP from the cytosol to lysosomes (Fig. 4c,d). To assess the role of CMA in mediating RHQ- and HQ-induced degradation, we silenced HSC70 and/or Lamp2a with shRNA in 150Q Neuro2a cells (Fig. 4e–h and Supplementary Fig. 7). All three combinations of HSC70- and Lamp2a-specific shRNAs reduced the inhibitory effects of RHQ on the aggregation of tNHTT-150Q-EGFP (Fig. 4e,f). Moreover, the levels of soluble tNHTT-150Q-EGFP were reduced by >50% as compared to other tested molecules. Knockdown of the CMA machinery components ablated the effect of RHQ on the levels of soluble tNHTT-150Q-EGFP (Fig. 4g,h). Taken together, these data suggest that the HQ and RHQ adaptor molecules enhance the degradation of tNHTT containing either 60 or 150 glutamine residues by targeting them for degradation by CMA.
259
Articles a
c
Mouse
Virus:
Q
R
Q
HQ
R
HQ
TG Mouse:
1
Virus:
R
2 Q
R
3 Q
R
1 Q
Q
2 HQ
Q
3 HQ
Q
1 HQ
2
R HQ
R
3 HQ
R
1
1
1C2
1
Anti-RFP 2
2
2
3
3
3
Injected rAAV
β-tubulin AGERA
* **
***
R Q
Q HQ
R HQ
R/Q
Q/HQ
R/HQ
f
R
e
1.4 1.2 1.0 0.8 0.6 0.4 0.2 0
* **
Q
***
R Q
Q HQ
R HQ
R/Q
Q/HQ
R/HQ
Q
HQ
R
Relative levels of soluble HTT-polyQ in R6/2 striata (WB)
d
Relative HTT-polyQ aggregation in R6/2 striata (WB)
1.2 1.0 0.8 0.6 0.4 0.2 0
HQ
1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0
* *** ** R Q
Q HQ
R HQ
R/Q
Q/HQ
R/HQ
g
The RHQ adaptor inhibits aggregation of proteins with an expanded polyQ tract in vivo We next assessed the effect of RHQ in two mouse models of HD. Intrastriatal injections of recombinant adeno-associated virus (rAAV) encoding R, RQ or RHQ (rAAV-R, rAAV-Q and rAAV-HQ) were performed in 4-week-old R6/2 mice3. Mice expressing tNHTT with 128 to 135 glutamine residues (tNHTT-polyQ) were used in this study. We injected the mice with three different contralateral combinations (R/Q, Q/HQ or R/HQ), and dissected the striata and prepared the lysates 4 weeks later. The rAAVs were widely distributed in the striatum (Supplementary Fig. 8). The analysis of the SDSinsoluble tNHTT-polyQ aggregates by filter trap assay revealed that the injection of rAAV-HQ reduced tNHTT-polyQ aggregation by 78.4% relative to the contralaterally injected rAAV-Q, and by 87.2% relative to rAAV-R in the R/HQ striata, whereas rAAV-Q reduced tNHTT-polyQ aggregation in the R/Q striata by 40% (Fig. 5a,b). The western blot analysis of the lysates was consistent with the filter trap assay results, where rAAV-Q decreased the aggregation by 25.6% in R/Q, rAAV-HQ by 83% in Q/HQ and by 90.8% in R/HQ brains (Fig. 5c,d). Furthermore, use of agarose gel electrophoresis to resolve the aggregates29 revealed a dramatic reduction in diffuse smearing staining, which represents larger complexes of tNHTT-polyQ, in HQ but not in Q and R samples (Fig. 5c). Levels of soluble polyQ protein increased by 21.7% in rAAV-Q–injected striata compared to the contralateral rAAV-R–injected striata, suggesting the aggregatedto-soluble shift of the protein resulted from aggregation inhibition
260
Relative amount of inclusions in the transduced areas (R6/2)
EM48
Anti-RFP
Figure 5 Detection of HTT levels and 1.2 the effect of the rAAV-HQ on inclusion 1.0 formation in R6/2 mouse brains. * 0.8 (a) Filter trap assay analysis of HTT-polyQ aggregation in nine R6/2 mouse brains 0.6 injected with three different combinations 0.4 ** of rAAV. (b) Quantification of the SDS*** 0.2 insoluble material detected by filter trap 0 assay (*, P = 0.037; **, P = 0.0083; ***, Q HQ R Q R HQ P = 0.0015). (c) Western blot analysis of HTT aggregation (gel top, EM48 antibody), soluble HTT (1C2 antibody), and higher molecular polyQ complexes by agarose gel electrophoresis for resolving aggregates (AGERA). Full-length western blots are presented in Supplementary Figure 13. (d) Quantification of the gel top-detected aggregation (*, P = 0.093; **, P = 0.0027; ***, P = 0.0045). (e) Quantification of soluble HTT-polyQ (*, P = 0.034; **, P = 0.00003; ***, P = 0.0001). (f) Brain sections stained with anti-RFP, EM48 and anti-ubiquitin antibodies (bar, 40 µm). (g) Numbers of ubiquitin-positive inclusions in the transduced areas (*, P = 0.0026; **, P = 5.535 × 10−8; ***, P = 2.009 × 10−8). Bars in b, d, e and g represent the relative mean values ± s.e.m. from three brains injected with different rAAV in each striatum. R in R/Q and R/HQ, and Q in Q/HQ brains represent the control value of 1. Anti-ubiquitin
© 2010 Nature America, Inc. All rights reserved.
Relative HTT-polyQ aggregation in R6/2 striata (FTA/tubulin)
EM48
b
HQ Aggregated HTT-polyQ Soluble
EM48
by rAAV-Q (Fig. 5c,e). Injected rAAV-HQ reduced the levels of soluble tNHTT-polyQ by 70% in Q/HQ and by 26.6% in R/HQ brains (Fig. 5c,e). When we counted the ubiquitin-positive inclusions in the transduced areas of the striata, we observed a 21% decrease in the rAAV-Q–injected site compared to the contralateral rAAV-R–injected site, whereas injection of rAAV-HQ reduced the amount of inclusions by 75.9% compared to rAAV-Q, and by 81.4% compared to rAAV-R (Fig. 5f,g). Furthermore, we observed less atrophy of the striata injected with rAAV-HQ than with the contralateral striata injected with rAAV-R or rAAV-Q (Supplementary Fig. 9). To test the effect of rAAV-HQ in another HD model, we injected HD190Q-EGFP mice30 with the same contralateral combinations of the viruses used for R6/2 at the age of 6 weeks and prepared brain sections from these animals 10 weeks later. When we counted the EGFP-positive inclusions in the fresh frozen sections, we observed reductions of 20.5%, 45.8% and 57.3% in R/Q, Q/HQ and R/HQ brains, respectively (Supplementary Fig. 10). The results from the two different HD mouse models clearly demonstrate the beneficial effect of including HSC70bm because rAAV-HQ was able to decrease the expanded HTT aggregation in vivo more efficiently than rAAV-Q and the decrease of HTT aggregation was accompanied by reduced levels of soluble polyQ. A virally expressed HQ adaptor ameliorates the HD phenotype in a mouse model We further evaluated the therapeutic potential of rAAV-HQ on the phenotype of R6/2 mice. We examined changes in body weight, clasping
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
Articles and by reduced fluorescence of the redistri buted tNHTT-150Q-EGFP (Fig. 4b and 80% 1.5 Supplementary Fig. 6). Taken together, our 70% ** * 3 1.4 60% *** results suggest that our adaptor molecule 2 1.3 50% **** 1 1.2 40% comprising QBP1 and HSC70bm motifs 0 1.1 30% specifically targets proteins with expanded 20% 1.0 10% 0.9 polyQ tracts for lysosomal degradation. 0% 0.8 Thus far, most experimental therapies for 0.7 HD have delivered small molecules or nucleic 14 4 6 8 10 12 12 6 8 10 acids to target aggregate formation, mutant Age (weeks) Age (weeks) HTT degradation, or protein interactions and c 200 d 1.0 cellular events disrupted by mutant HTT31,32. HQ/HQ R/R 0.9 * Q/Q 180 Q/Q To compare the effects of new HD treatments 0.8 R/R 160 HQ/HQ 0.7 with previous studies, we used the R6/2 mouse 140 ** *** 0.6 120 model, which is the most widely used HD 0.5 100 *** 0.4 mouse model for experimental therapeutic 80 0.3 60 interventions and it is especially suitable for 0.2 40 0.1 comparing the increases in life span ena20 0 0 bled by the different approaches17,33. To our 90 100 110 120 130 140 150 6 8 10 12 Day knowledge, the 29.6% extension of life span Age (weeks) in R6/2 mice using our strategy exceeds that reported for other single-drug therapies. For Figure 6 Effect of rAAV-HQ on the HD phenotype in R6/2 mice. (a) Relative body weight changes in mice after bilateral injection with rAAV-R, rAAV-Q and rAAV-HQ at 6–14 weeks of age (n = 11 for instance, treatments such as Congo Red or all three groups until 12 weeks; at 14 weeks, n = 7 for R/R, n = 10 for Q/Q, n = 11 for HQ/HQ) trehalose33,34, which target HTT aggregates (*, P = 0.0401; **, P = 0.0081; ***, P = 0.0182; ****, P = 0.0225). (b) Clasping scores. n = 11 or aggregate formation, increased survival by for all three groups (scale: 0, no clasping; 1, clasping of the forelimbs only; 2, clasping of both fore 16.4% and 11.3%, respectively. Another comand hind limbs once or twice; 3, clasping of both fore and hind limbs more than 3 times or more pound used in experimental HD treatment is than 5 s). (c) Rotarod performance; n = 11 for all three groups (*, P = 0.0055; **, P = 0.002; cystamine, which may reduce HTT aggrega***, P = 0.0455; ****, P = 0.0242). (d) Survival curves for mice after bilateral injection with rAAV-R, rAAV-Q, and rAAV-HQ; n = 11 for all groups. Median survival (days): R/R-108; Q/Q-123; HQ/ tion by inhibition of transglutaminase (TG), HQ-140. Mean survival (days): R/R-105.5 ± 3.9; Q/Q-122.3 ± 4.8; HQ/HQ-136.2 ± 4.4. which is thought to cross-link expanded Log-rank test, P < 0.0001; Wilcoxon test, P < 0.0001. polyQ proteins and facilitate their aggregation. Increased TG activity is observed in the brains of individuals with HD35. In two studies score, rotarod performance and life span in mice injected in both stri- with cystamine treatments starting at ages 7 weeks or 3 weeks, the ata with the same virus at the age of 4 weeks. The loss of body weight survival of R6/2 mice increased by 12% and 19.5%, respectively36,37. in both R/R (both striata injected with rAAV-R) and Q/Q (both striata We recently reported that activation of proteasomal degradation by the injected with rAAV-Q) mice was substantially more severe than that amiloride derivative, benzamil, increased the life span of R6/2 mice by of the HQ/HQ (both striata injected with rAAV-HQ) mice, beginning 10%17. Single-agent treatments, such as sodium butyrate38, co-enzyme at 8 weeks of age (Fig. 6a). HQ/HQ mice exhibited markedly lower Q10 (ref. 39), clioquinol40, caspase inhibitor zVAD-fmk41, FGF-2 clasping scores at all time points from 6 to 12 weeks, compared to (ref. 42) and sertraline43, increase the survival of R6/2 mice by 20.8%, both R/R and Q/Q mice. In Q/Q mice, the limb clasping posture was 21.3–25.3%, 20%, 25%, ~20% and 20%, respectively. Nonetheless, the ameliorated only at the ages of 6 and 8 weeks (Fig. 6b). Consistent strongest single-drug effect in the R6/2 mouse (a 29.1% extension of with the improvement in clasping scores, we also found that HQ/HQ life span) was reported using mithramycin, a guanosine-cytosine–rich mice showed markedly better performance than R/R or Q/Q mice on DNA binding antitumor antibiotic that modulates transcription44. the rotarod at 6–12 weeks (Fig. 6c). Most importantly, the life span of The combination of mithramycin and cystamine enabled a 40% HQ/HQ mice increased substantially with a median survival time of increase in R6/2 mice survival45. Protein- and peptide-based strate140 d compared to 108 d for R/R and 123 d for Q/Q mice (Fig. 6d), gies have involved QBP1 or intrabodies that inhibit aggregation of making the intrastriatal delivery of rAAV encoding HQ one of the HTT containing expanded polyQ46,47. However, the effect in an HD most effective experimental therapies reported to increase the life mouse model could not be studied well, owing to limited delivery of span of R6/2 mice31,32. the purified QBP1 to the brain46. Adenoviral delivery of Happ1 intrabody, which recognizes the polyproline and proline-rich domains of DISCUSSION HTT and decreases its aggregation, improved motor and cognitive Our proof-of-concept experiments suggest that a peptide compris- performance as well as neuropathology in four HD transgenic mouse ing HSC70–binding motifs and a domain capable of recognizing a models, including R6/2. However, it improved the body weight and misfolded protein could provide a general way to specifically target prolonged the life span only in N171-82Q mice47. mutant proteins for degradation by CMA. When co-expressed with Our in vivo study revealed beneficial effects of QBP1, as well as the RHQ or HQ, tNHTT containing 150Q or 60Q co-localized with a adaptor molecule comprising QBP1 and HSC70bm on the pathology lysosomal marker (Fig. 4b) and was relocated to the lysosomal frac- and phenotype of HD model mice. When the polyQ-binding motif tion of the cytoplasm (Fig. 4c,d), respectively. This translocation was QBP1 (rAAV-Q) was delivered to R6/2 mice, the aggregation of accompanied both by enhanced degradation of tNHTT-60Q-EGFP, as mutant HTT decreased by 21–25.6%, rotarod performance improved observed in the chase experiments (Fig. 3 and Supplementary Fig. 5), by 11% and 30% at 6 and 8 weeks of age, respectively, and life span
© 2010 Nature America, Inc. All rights reserved.
b 100% 90%
HQ/HQ
R/R
Q/Q
HQ/HQ
R/R
Q/Q
R/R
Q/Q HQ/HQ
HQ/HQ
R/R
Q/Q
Percent of mice
R/R Q/Q HQ/HQ
Surviving
Seconds on rotarod
Body weight change
a
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
261
© 2010 Nature America, Inc. All rights reserved.
Articles was extended by 13.9%. These effects are comparable to those reported for treatments that primarily target aggregation of HTT protein bearing expanded polyQ tracts33,34,36,37. However, when the adaptor molecule rAAV-HQ was transduced, the therapeutic effect in R6/2 mice was much stronger, with 81.4–90.8% inhibition of aggregation, a 25% increase in body weight, 48–70% improvement in rotarod performance and 29.6% increase in survival. The differences in phenotypes of R6/2 mice expressing RQ and RHQ in terms of HTT aggregation, phenotype and survival suggest that blocking protein aggregation alone may not be sufficient for effective treatment of polyQ diseases. Several gene therapies using the R6/2 model have also been reported. Intraventricular injection of siRNA to R6/2 at an early postnatal period inhibited HTT transgene expression and extended the survival of the mice by 16% 11. Although human mutant HTT gene expression could be specifically knocked down experimentally by RNAi against the human mutant HTT transgene without affecting endogenous normal mouse HTT expression48–50, specific inhibition of mutant gene expression might not be feasible owing to ablation of the normal gene in humans 51. Nonetheless, this problem may be overcome in the future if polymorphisms linked to expanded polyQ tracts in mutant HTT are identified and confirmed52. Brain-derived neurotrophic factor and Noggin expression by adenoviral transduction to R6/2 mouse brain induced neurogenesis and extended life span by 16.8%53. Thus, our approach of bilateral injection of rAAV encoding an adaptor molecule appears to be one of the most effective single-agent therapies in the R6/2 mouse model of HD. The effect on the aggregation of mutant HTT with expanded polyQ tracts was not restricted to R6/2 mice as the adaptor molecule was also able to strongly inhibit the aggregation of mutant HTT in HD190Q-EGFP mice30 (Supplementary Fig. 10). We also investigated the applicability of adaptor molecule RHQ to other human proteins with expanded polyQ tracts, such as ataxin-3 and androgen receptor. RHQ was able to decrease the accumulation and aggregation of these proteins in cellular models of these diseases (Supplementary Fig. 11). This suggests the potential of our approach for developing therapeutics for diseases caused by conformational changes in proteins other than those with expanded polyQ tracts, such as tauopathies (e.g., certain forms of frontotemporal dementia) and synucleopathies (e.g., Parkinson’s disease). The development of peptides or intrabodies with even higher binding affinity to abnormal proteins should enhance the potential of our strategy. For example, intrabodies which are able to inhibit aggregation and toxicity of mutant HTT and α-synuclein have been extensively studied54–56. Tagging such an intrabody with HSC70bm might increase its therapeutic effect. Virus vectors may offer potential for future clinical use, although certain safety concerns represent a drawback of this therapeutic approach. In this study, however, the intrastriatal injection of rAAV-R, rAAV-Q and rAAV-HQ did not affect the body weight and rotarod performance of wild-type mice (Supplementary Fig. 12). The discovery of small compounds capable of conjugating disease-related proteins with HSC70 potentially represents another therapeutic solution for diseases caused by misfolded proteins. We also believe that using adaptor molecules similar to those described here may provide unique ways of therapeutic or experimental regulation of endogenous protein levels by enhancing their degradation. Methods Methods and any associated references are available in the online version of the paper at http://www.nature.com/naturebiotechnology/.
262
Note: Supplementary information is available on the Nature Biotechnology website. Acknowledgments This work was supported by a Grant-in-Aid for Scientific Research on Priority Areas (Research on Pathomechanisms of Brain Disorders) from the Ministry of Education, Culture, Sports, Science and Technology of Japan (17025044) and by a Grant-in-Aid for the Research on Measures for Intractable Diseases from the Ministry of Health, Welfare and Labor, Japan. This work was supported partly by a grant from the Japan Society for the Promotion of Science (JSPS). P.O.B. was a JSPS postdoctoral fellow. We thank D.W. Chapmon for stylistic revision of the manuscript. Monomeric KikGR (mKikGR) was a kind gift of A. Miyawaki, RIKEN BSI Japan. AUTHOR CONTRIBUTIONS P.O.B. raised the hypothesis and designed the experiments. Experimental work was performed by P.O.B., A.G., H.K.W., H.M., M.O., M.K. and M.Y. The tNHTT-62QkikGR Neuro2a cell line system was prepared by G.M. The EGFP-AR constructs were prepared by Y.K. The manuscript was written by P.O.B. and N.N. Y.N. provided some information about treatment using QBP1. N.N. supervised the project. All authors discussed results and commented on 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. Zoghbi, H.Y. & Orr, H.T. Glutamine repeats and neurodegeneration. Annu. Rev. Neurosci. 23, 217–247 (2000). 2. DiFiglia, M. et al. Aggregation of huntingtin in neuronal intranuclear inclusions and dystrophic neurites in brain. Science 277, 1990–1993 (1997). 3. Davies, S.W. et al. Formation of neuronal intranuclear inclusions underlies the neurological dysfunction in mice transgenic for the HD mutation. Cell 90, 537–548 (1997). 4. Perutz, M.F., Pope, B.J., Owen, D., Wanker, E.E. & Scherzinger, E. Aggregation of proteins with expanded glutamine and alanine repeats of the glutamine-rich and asparagine-rich domains of Sup35 and of the amyloid beta-peptide of amyloid plaques. Proc. Natl. Acad. Sci. USA 99, 5596–5600 (2002). 5. Tanaka, M. et al. Expansion of polyglutamine induces the formation of quasiaggregate in the early stage of protein fibrillization. J. Biol. Chem. 278, 34717–34724 (2003). 6. Dunah, A.W. et al. Sp1 and TAFII130 transcriptional activity disrupted in early Huntington’s disease. Science 296, 2238–2243 (2002). 7. Nucifora, F.C. Jr. et al. Interference by huntingtin and atrophin-1 with cbp-mediated transcription leading to cellular toxicity. Science 291, 2423–2428 (2001). 8. Yamanaka, T. et al. Mutant Huntingtin reduces HSP70 expression through the sequestration of NF-Y transcription factor. EMBO J. 27, 827–839 (2008). 9. Bauer, P.O. & Nukina, N. The pathogenic mechanisms of polyglutamine diseases and current therapeutic strategies. J. Neurochem. 110, 1737–1765 (2009). 10. Yamamoto, A., Lucas, J.J. & Hen, R. Reversal of neuropathology and motor dysfunction in a conditional model of Huntington’s disease. Cell 101, 57–66 (2000). 11. Wang, Y.L. et al. Clinico-pathological rescue of a model mouse of Huntington’s disease by siRNA. Neurosci. Res. 53, 241–249 (2005). 12. DiFiglia, M. et al. Therapeutic silencing of mutant huntingtin with siRNA attenuates striatal and cortical neuropathology and behavioral deficits. Proc. Natl. Acad. Sci. USA. 104, 17204–17209 (2007). 13. Harper, S.Q. et al. RNA interference improves motor and neuropathological abnormalities in a Huntington’s disease mouse model. Proc. Natl. Acad. Sci. USA. 102, 5820–5825 (2005). 14. Rodriguez-Lebron, E., Denovan-Wright, E.M., Nash, K., Lewin, A.S. & Mandel, R.J. Intrastriatal rAAV-mediated delivery of anti-huntingtin shRNAs induces partial reversal of disease progression in R6/1 Huntington’s disease transgenic mice. Mol. Ther. 12, 618–633 (2005). 15. Ravikumar, B. et al. Inhibition of mTOR induces autophagy and reduces toxicity of polyglutamine expansions in fly and mouse models of Huntington disease. Nat. Genet. 36, 585–595 (2004). 16. Sarkar, S. et al. Small molecules enhance autophagy and reduce toxicity in Huntington’s disease models. Nat. Chem. Biol. 3, 331–338 (2007). 17. Wong, H.K. et al. Blocking acid-sensing ion channel 1 alleviates Huntington’s disease pathology via an ubiquitin-proteasome system-dependent mechanism. Hum. Mol. Genet. 17, 3223–3235 (2008). 18. Bauer, P.O. et al. Inhibition of rho kinases enhances the degradation of mutant huntingtin. J. Biol. Chem. 284, 13153–13164 (2009). 19. Mizushima, N., Ohsumi, Y. & Yoshimori, T. Autophagosome formation in mammalian cells. Cell. Struct. Funct. 27, 421–429 (2002). 20. Cuervo, A.M. & Dice, J.F. A receptor for the selective uptake and degradation of proteins by lysosomes. Science 273, 501–503 (1996). 21. Agarraberes, F.A. & Dice, J.F. A molecular chaperone complex at the lysosomal membrane is required for protein translocation. J. Cell Sci. 114, 2491–2499 (2001).
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
Articles 22. Nagai, Y. et al. Inhibition of polyglutamine protein aggregation and cell death by novel peptides identified by phage display screening. J. Biol. Chem. 275, 10437–10442 (2000). 23. Cuervo, A.M., Stefanis, L., Fredenburg, R., Lansbury, P.T. & Sulzer, D. Impaired degradation of mutant alpha-synuclein by chaperone-mediated autophagy. Science 305, 1292–1295 (2004). 24. Cuervo, A.M., Mann, L., Bonten, E.J., d’Azzo, A. & Dice, J.F Cathepsin A regulates chaperone-mediated autophagy through cleavage of the lysosomal receptor. EMBO J. 22, 47–59 (2003). 25. Nagai, Y. et al. Prevention of polyglutamine oligomerization and neurodegeneration by the peptide inhibitor QBP1 in Drosophila. Hum. Mol. Genet. 12, 1253–1259 (2003). 26. Nagai, Y. et al. A toxic monomeric conformer of the polyglutamine protein. Nat. Struct. Mol. Biol. 14, 332–340 (2007). 27. Tsutsui, H., Karasawa, S., Shimizu, H., Nukina, N. & Miyawaki, A. Semi-rational engineering of a coral fluorescent protein into an efficient highlighter. EMBO Rep. 6, 233–238 (2005). 28. Rodriguez-Enriquez, S., Kim, I., Currin, R.T. & Lemasters, J.J. Tracker dyes to probe mitochondrial autophagy (mitophagy) in rat hepatocytes. Autophagy 2, 39–46 (2006). 29. Weiss, A. et al. Sensitive biochemical aggregate detection reveals aggregation onset before symptom development in cellular and murine models of Huntington’s disease. J. Neurochem. 104, 846–858 (2008). 30. Kotliarova, S. et al. Decreased expression of hypothalamic neuropeptides in Huntington disease transgenic mice with expanded polyglutamine-EGFP fluorescent aggregates. J. Neurochem. 93, 641–653 (2005). 31. Beal, M.F. & Ferrante, R.J. Experimental therapeutics in transgenic mouse models of Huntington’s disease. Nat. Rev. Neurosci. 5, 373–384 (2004). 32. Li, J.Y., Popovic, N. & Brundin, P. The use of the R6 transgenic mouse models of Huntington’s Disease in attempts to develop novel therapeutic strategies. NeuroRx 2, 447–464 (2005). 33. Tanaka, M. et al. Trehalose alleviates polyglutamine-mediated pathology in a mouse model of Huntington disease. Nat. Med. 10, 148–154 (2004). 34. Sanchez, I., Mahlke, C. & Yuan, J. Pivotal role of oligomerization in expanded polyglutamine neurodegenerative disorders. Nature 421, 373–379 (2003). 35. Karpuj, M.V. et al. Transglutaminase aggregates huntingtin into nonamyloidogenic polymers, and its enzymatic activity increases in Huntington’s disease brain nuclei. Proc. Natl. Acad. Sci. USA 96, 7388–7393 (1999). 36. Karpuj, M.V. et al. Prolonged survival and decreased abnormal movements in transgenic model of Huntington disease, with administration of the transglutaminase inhibitor cystamine. Nat. Med. 8, 143–149 (2002). 37. Dedeoglu, A. et al. Therapeutic effects of cystamine in a murine model of Huntington’s disease. J. Neurosci. 22, 8942–8950 (2002). 38. Ferrante, R.J. et al. Histone deacetylase inhibition by sodium butyrate chemotherapy ameliorates the neurodegenerative phenotype in Huntington’s disease mice. J. Neurosci. 23, 9418–9427 (2003). 39. Smith, K.M. et al. Dose ranging and efficacy study of high-dose coenzyme Q10 formulations in Huntington’s disease mice. Biochim. Biophys. Acta 1762, 616–626 (2006).
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
40. Nguyen, T., Hamby, A. & Massa, S.M. Clioquinol down-regulates mutant huntingtin expression in vitro and mitigates pathology in a Huntington’s disease mouse model. Proc. Natl. Acad. Sci. USA 102, 11840–11845 (2005). 41. Ona, V.O. et al. Inhibition of caspase-1 slows disease progression in a mouse model of Huntington’s disease. Nature 399, 263–267 (1999). 42. Jin, K. et al. FGF-2 promotes neurogenesis and neuroprotection and prolongs survival in a transgenic mouse model of Huntington’s disease. Proc. Natl. Acad. Sci. USA 102, 18189–18194 (2005). 43. Peng, Q. et al. The antidepressant sertraline improves the phenotype, promotes neurogenesis and increases BDNF levels in the R6/2 Huntington’s disease mouse model. Exp. Neurol. 210, 154–163 (2008). 44. Ferrante, R.J. et al. Chemotherapy for the brain: the antitumor antibiotic mithramycin prolongs survival in a mouse model of Huntington’s disease. J. Neurosci. 24, 10335–10342 (2004). 45. Ryu, H. et al. ESET/SETDB1 gene expression and histone H3 (K9) trimethylation in Huntington’s disease. Proc. Natl. Acad. Sci. USA 103, 19176–19181 (2006). 46. Popiel, H.A., Nagai, Y., Fujikake, N. & Toda, T. Delivery of the aggregate inhibitor peptide QBP1 into the mouse brain using PTDs and its therapeutic effect on polyglutamine disease mice. Neurosci. Lett. 449, 87–92 (2009). 47. Southwell, A.L., Ko, J. & Patterson, P.H. Intrabody gene therapy ameliorates motor, cognitive, and neuropathological symptoms in multiple mouse models of Huntington’s disease. J. Neurosci. 29, 13589–13602 (2009). 48. Harper, S.Q. et al. RNA interference improves motor and neuropathological abnormalities in a Huntington’s disease mouse model. Proc. Natl. Acad. Sci. USA 102, 5820–5825 (2005). 49. Rodriguez-Lebron, E., Denovan-Wright, E.M., Nash, K., Lewin, A.S. & Mandel, R.J. Intrastriatal rAAV-mediated delivery of anti-huntingtin shRNAs induces partial reversal of disease progression in R6/1 Huntington’s disease transgenic mice. Mol. Ther. 12, 618–633 (2005). 50. DiFiglia, M. et al. Therapeutic silencing of mutant huntingtin with siRNA attenuates striatal and cortical neuropathology and behavioral deficits. Proc. Natl. Acad. Sci. USA 104, 17204–17209 (2007). 51. Huang, B. et al. High-capacity adenoviral vector-mediated reduction of huntingtin aggregate load in vitro and in vivo. Hum. Gene Ther. 18, 303–311 (2007). 52. Zhang, Y., Engelman, J. & Friedlander, R.M. Allele-specific silencing of mutant Huntington’s disease gene. J. Neurochem. 108, 82–90 (2009). 53. Cho, S.R. et al. Induction of neostriatal neurogenesis slows disease progression in a transgenic murine model of Huntington disease. J. Clin. Invest. 117, 2889–2902 (2007). 54. Southwell, A.L. et al. Intrabodies binding the proline-rich domains of mutant huntingtin increase its turnover and reduce neurotoxicity. J. Neurosci. 28, 9013–9020 (2008). 55. Emadi, S., Barkhordarian, H., Wang, M.S., Schulz, P. & Sierks, M.R. Isolation of a human single chain antibody fragment against oligomeric alpha-synuclein that inhibits aggregation and prevents alpha-synuclein-induced toxicity. J. Mol. Biol. 368, 1132–1144 (2007). 56. Lynch, S.M., Zhou, C. & Messer, A. An scFv intrabody against the nonamyloid component of alpha-synuclein reduces intracellular aggregation and toxicity. J. Mol. Biol. 377, 136–147 (2008).
263
ONLINE METHODS
© 2010 Nature America, Inc. All rights reserved.
Materials. The macroautophagy inhibitor 3-methyladenine (3MA) and the lysosomal inhibitor NH4Cl were purchased from Sigma. The cathepsin D and E inhibitor pepstatin A and the cathepsin L inhibitor leupeptin were from Nacalai Tesque and the proteasome inhibitor MG132 (Z-Leu-Leu-Leu-aldehyde) was from Wako Chemicals. Hoechst 33258 and Lysotracker-rhodamine were obtained from Molecular Probes. Mouse monoclonal antibody specific for N-terminal of htt (EM48), mouse monoclonal anti-polyglutamine-expansion (1C2) and rat monoclonal anti-β-tubulin antibodies were from Chemicon. Anti-GFP, anti-RFP, anti-LC3 and anti-p62 antibodies were from MBL. Anti-ubiquitin was purchased from Dako, anti-cathepsin D, anti-HA, goat polyclonal anti-HSC70 and rabbit polyclonal anti-caspase-3 antibodies were from Santa Cruz Biotechnology. Anti-Lamp2a rat monoclonal antibody was obtained from Abcam. The plasmid d1EGFP-N1 was purchased from Clontech. Rapamycin was purchased from Calbiochem. Mice. Two HD mouse models were used in this study. Heterozygous HTT exon 1 transgenic female mice of the R6/2 strain (145 CAG repeats; (HD exon 1)) were originally obtained from the Jackson Laboratory. Mice that were used for our in vivo study carried 128–135CAG repeats. The HD190Q-EGFP transgenic female mice harbor mutant truncated N-terminal HTT containing 190 CAG repeats fused with EGFP in its genome. These animals show progressive motor abnormalities, shorter life spans and neuropathology such as formation of inclusions in the brain30. All the experiments with mice were approved by the Animal Experiment Committee of the RIKEN Brain Science Institute. Plasmids. The construction of the N-terminal fragments of human huntingtin exon1 (tNHTT) encoding 16, 60 and 150Q repeats and EGFP was previously described57. Plasmids for the transient transfection were prepared by the introduction of the coding sequences into pcDNA3.1 vector (Invitrogen). The tNHTT-60Q fragment was fused to the N-terminal of a variant of yellow fluorescent protein (Venus) and inserted into the pcDNA3.1 vector (60Q). Two Hsc70 binding motifs (HSC70bm) (in bold: 5′-GTTAAGAAGGATCAA GCTGGAGCCGCTGCACCG-AAGTTCGAACGTCAA-3′) were inserted between the Cfr13I and BamHI cutting sites of 60Q. Complementary oligonucleotides encoding two copies of the active (forward: 5′-TCGAACTGGAA GTGGTGGCCAGGTATCTTCGACTCGAACTGGAAGTGGTGGCCAGGTAT CTTCGAC-3′) and scrambled (forward: 5′-TGGGGATGGCCTAATGACTT CGACTGGAAGGGTTGGTTCAGCCCTTGGAAGATTAGCTGGATCAAC-3′) QBP1 were synthesized (Operon), annealed, fused to C terminus of mono meric red fluorescent protein (mRFP) and inserted into the pcDNA3.1 vector. HSC70bm were amplified with primers containing SalI (New England Biolabs) in the forward and XmaI cutting site in the reverse primer from the 60Q-HSC70bm construct. The PCR product was introduced between mRFP and scrambled or active QBP1. To produce molecules without an mRFP tag, all constructs were amplified using primers containing BamHI in the forward and XbaI in the reverse primer, respectively, and introduced into the pcDNA3.1 vector. The schemes and abbreviations of the constructs are listed in Figure 2a and Supplementary Figure 3a. The amino acid sequences of the expressed molecules are given in Supplementary Table 1 (without an mRFP tag). The construction of plasmids encoding human truncated or full-length ataxin-3 containing 20 or 130Q (in pEGFP-N1 vector) was described previously58. Human androgen receptor with 23Q (AR23Q) was amplified from human brain cDNA library by PCR using a set of primers BglII-AR-Fw (5′-AAAA GATCTATGGAAGTGCAGTTAGGGCT-3′) and SalI-AR-Rv (5′-AAAAAAG TCGACCTGGGTGTGGAAATAGATGG-3′), cleaved by BglII and SalI, and introduced into the BglII-SalI sites of the pEGFP-C1 vector (EGFP-AR23Q). The CAG repeat tract of EGFP-AR23Q was expanded by a method described previously59. A primer set, BglII-AR-Fw and MmeI-AR-exp-Rv (5′-CATCCT CACCCTGCTGCTGCTCCAACTGCCTGGGG-3′) was used to amplify the 5′ coding sequence including the CAG repeat tract. Another primer set, MmeIAR-exp-Fw (5′-AGGCCGCGAGCGCAGCACCTTCCGACGCCAGTTTG-3′) and AR-630Rv (5′-TCTCCCGCTGCTGCTGCCTT-3′), was used to amplify the CAG tract and its 3′ flanking region. These two fragments were digested by MmeI (New England Biolabs), gel-purified, and treated with T4 DNA ligase to connect them at their CAG repeat tracts. The ligated fragment was gelpurified and amplified by PCR using BglII-AR-Fw and AR-630Rv primers,
nature biotechnology
and digested by BglII and AflII (New England Biolabs). The resulting fragment was ligated with EGFP-AR. By two cycles of expansion, EGFP-AR45Q and EGFP-AR99Q were obtained. The N-terminal fragments of AR23Q and AR99Q were amplified by PCR using primers BglII-AR-Fw and SalI-AR-396Rv (5′-AAAAAAGTCGACGACGCAACCTCTCTCGGGGT-3′), cleaved by BglII and SalI, and subcloned into the BglII-SalI sites of the pEGFP-C1. Monomeric KikGR (mKikGR) was a kind gift of A. Miyawaki. KikGR is cleaved and photo-converted by irradiation of ~350–420 nm 20. It contains additional eleven mutations in the KikGR tetramer (BAD95669), such as A17S, G32R, I37T, N39T, C116T, V126T, N161E, Q167E, H219Y, L222T and P223Y. The pFRT-KikGR was constructed by insertion of mKikGR fragment PCR-amplified using a specific primer set, 5′-CCGAATTCATGCTAGCACC ATGGATCCTAGTGTGATTACATCAGAAATG-3′ as forward and 5′-TTTTA GATCTTATCCGGACTTGGCTTCAAATTCATACTTGGCGCC-3′ as reverse primer, into NheI and BamHI sites of pcDNA5/FRT/TR vector (Invitrogen). The pFRT-tNHTT-62Q-KikGR was then generated by inserting the HTT exon 1 with 62Q (HD62) into EcoRI and BamHI sites of pFRT-mKikGR vector. pFRT-HD62-HA-KikGR was then constructed by the introduction of HA epitope into the BamHI site using the following double-stranded oligoDNA, 5′-GATCTATACCCATACGATGTTCCAGATTACGCG-3′ and 5′-GA TCCGCGTAATCTGGAACATCGTATGGGTATA-3′. All constructs were verified by sequencing. Cell culture, transient transfection and treatments. Mouse neuroblastoma (Neuro2a) cells were maintained in Dulbecco’s modified Eagle’s medium (Sigma) supplemented with 10% heat-inactivated FBS (Sigma), 100 U/ml penicillin and 100 µg/ml streptomycin (Invitrogen) at 37 °C in an atmosphere containing 5% CO2 and 95% air. Establishment of stable Neuro2a cell lines with the ecdysone-inducible mammalian expression system (Invitrogen), that express tNHTT-16Q-EGFP (16Q Neuro2a cells), tNHTT-60Q-EGFP (60Q Neuro2a cells), tNHTT-150Q-EGFP (150Q Neuro2a cells) and tNHTT-150Q-nls-EGFP (150Qnls Neuro2a cells) has been described earlier57,60. Neuro2a cells were differentiated with 5 mM dbcAMP (N6,2′-O-dibutyryladenosine-3′,5′-cyclic monophosphate sodium salt) (Nacalai Tesque) and induced to express tNHTTpolyQ with 2 µM ponasterone A (ponA; Invitrogen) for indicated times. PC12 cells were grown in the same conditions as Neuro2a cells, except for the serum composition of 5% of fetal bovine and 10% of horse serum (Sigma). Neuro2a/FRT/TR cell line was generated as described in the protocol for the Flp-in/T-Rex system (Invitrogen). Neuro2a tNHTT-62Q-kikGR cell line was generated by transfecting Neuro2a/FRT/TR cells with the pFRT-HD62KikGR and pOG44 (Invitrogen) constructs using Lipofectamine 2000 reagent and selected with 200 µg/ml hygromycin. The expression of the pFRT-tNHTT62Q-KikGR was induced by doxycycline treatment of the cells. For the photoinduced kikGR cleavage, cells were exposed to ~400 nm wavelength for 5 min. All transient transfections were performed with Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instruction. Cell death and cell viability assays. For quantification of cell death, 5 µg/ml each of Hoechst 33342 and propidium iodide (PI) were added to differentiated and induced Neuro2a cells transiently transfected with either of the tested constructs (Supplementary Fig. 3a). After 10 min at 37 °C, the number of cells with PI uptake over the total number of cells was calculated by ArrayScan. Cell viability (5 × 103 cells/well of a 96-well plate) was determined using MTT assay as described previously57. Isolation of intact lysosomes. Neuro2a cells grown in 10-cm dishes were co-transfected with tNHTT-60Q-EGFP and tested constructs, and after 16 h of incubation, cells were collected for the experiment. We used a previously described method for the purification of lysosomes from Chinese hamster ovary cells61. Metrizamide used in the previous study was replaced by Histodenz (Sigma). Purified lysosomes were permeabilized with 0.01% Triton X-100 and the samples were centrifuged at 20,000g for 10 min. Supernatant containing the lumenal fraction of the lysosomes was analyzed. RNAi. Each sense and anti-sense template short hairpin (sh) RNA for Lamp2a and HSC70 was purchased from Operon, annealed and ligated into pSilencer1.0 vector with U6 promoter according to the manufacturer’s instructions (Ambion).
doi:10.1038/nbt.1608
© 2010 Nature America, Inc. All rights reserved.
The target sequences were as follows: Lamp2a, 5′-AACCATTGCAGTACCTGA CAA-3′; HSC70, 5′-AACTGGAGAAAGTCTGCAACC-3′. The plasmids containing shRNA were sequence verified. Plasmids were co-transfected with DNA coding for tested molecules into Neuro2a cells using Lipofectamine 2000. After 2 d of silencing, cells were differentiated and induced. We performed RT-PCR and western blot analysis to verify the knockdown efficiency (Fig. 4g and Supplementary Fig. 7). ArrayScan quantification. For the quantification of the inclusions, cells were grown in 24-well plates for indicated periods, fixed in 4% paraformaldehyde, washed and incubated with Hoechst 33258 at 1:1,000 dilution in PBS. Cells were analyzed by ArrayScanVTI High Content Screening (HSC) Reader (Cellomics) using Target Activation BioApplication (TABA). TABA analyzes images acquired by an HSC Reader and provides measurements of the intracellular fluorescence intensity and localization on a cell-by-cell basis. In each well, at least 10,000 cells were counted and quantified for the presence of the inclusions. Nuclei stained with Hoechst 33285 provided the autofocus target and their count gave the exact number of the quantified cells. The screening itself consisted of two scans using Hoechst, FITC (for GFP) and TRITC (for RFP) fluorescence. First, the number of inclusions in transfected cells was calculated when fluorescent spots were at least 5 pixels (magnification 20× for cytoplasmic and 40× for nuclear aggregates) with average GFP intensity >1,500 in the RFP background. Second, nuclei were defined as the objects of interest and the cells with average intensity >50 within 3 pixels from the nucleus were selected for the analysis. The percentage of the cells with aggregates was then calculated. When the constructs without mRFP were used for transfection, the procedure was the same, except the RFP intensity was not measured. Scanning was performed with triplicate or quadruplicate in each experimental condition. Data were generated from the quantification of >250,000 cells in each experimental setup. Chase experiments. To determine whether tNHTT-polyQ-EGFP degrades faster in the presence of RHQ, chase experiments were performed. 60Q Neuro2a cells were first induced to express tNHTT-60Q-EGFP for 20 h and then transfected with the tested constructs. PonaA was removed 4 h later; cells were washed with PBS and incubated in the medium containing dbcAMP (for differentiation) for 12–48 h. Cells were lysed and the levels of tNHTT-60QEGFP were analyzed by western blot analysis. Neuro2a tNHTT-62Q-kikGR cells were induced for 24 h, and then the 62QkikGR was cleaved by 5 min irradiation of the cells. Chase phase lasted for 12 h before the cells were collected and analyzed. Construction and stereotaxic injection of rAAV-R, rAAV-Q and rAAV-HQ. The viral expression constructs rAAV1/2-CAG-(R; Q; HQ)-WPRE (WPRE, woodchuck post-translational regulatory element) were prepared by sub cloning of mRFP (R), mRFP-QBP1 (Q) and mRFP-HSC70bm-QBP1 (HQ) into an adeno-associated (serotype-2) viral (rAVET) cassette. Viral vectors were packaged and affinity purified (GeneDetect) for high expression in mouse brain tissue. The stereotaxic injections of rAAV into R6/2 mouse striata were performed in 4-week-old mice. The animals were first anesthetized by intraperitoneal injection of pentobarbital and placed in a stereotaxic apparatus. rAAV was injected into the right and left striatum through burr holes in the skull using a 5 µl Hamilton syringe mounted on the stereotaxic apparatus. Injections were placed 0.5 mm anterior to the bregma, 1.5 mm lateral to the sagittal suture and 2 mm below the skull surface. The rate of injection was 0.3 µl/min with total volume of 3 µl (equivalent to 3.6 × 109 genomic particles). Mice were euthanized at 8 weeks of age and the level of aggregation in the striatum was analyzed by western blot analysis, filter trap assay and immuno histochemistry. Another group of R6/2 mice was injected bilaterally with the same rAAV for phenotype analysis. HD190Q-EGFP mice were injected at 6 weeks of age and the striata analyzed 10 weeks later. Immunoblotting. Cells were washed twice with ice-cold PBS, scraped and resuspended in lysis buffer (0.5% Triton X-100 in PBS, 0.5 mM phenylmethylsulfonyl fluoride, complete protease inhibitor mixture (Roche)). After incubating on ice for 30 min lysates were briefly sonicated. Equal amounts of protein were boiled for 5 min in 2× SDS-sample buffer, separated by 5–12% gradient
doi:10.1038/nbt.1608
SDS-PAGE and electrophoretically transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore). The membranes were blocked in 5% skim milk in 0.05% Tween 20/Tris-buffered saline (TBST) and incubated with primary antibody (dilutions in accordance to manufacturer’s recommendations) overnight at 4 °C. Then the membranes were washed three times in TBST and incubated for 1 h with horseradish peroxidase-conjugated secondary antibody (dilution 1:5,000). Immunoreactive proteins were detected with enhanced chemiluminescence reagents (Amersham). Full scans of western blot data are presented in Supplementary Figure 13. Dot blots were prepared by blotting 10 µg of total proteins onto nitrocellulose membranes using a vacuum manifold and were processed as described for western blots. Images were made either by exposing the membranes to Hyperfilm MP (Amersham) or by Intelligent Dark Box II using Image Reader Las-1000Lite (both Fujifilm). The quantifications were performed using Multi Gauge software (Fujifilm). Filter trap assay. The assay was done using a Hybri-Dot manifold (Bio-Rad) and cellulose acetate membrane filter with pore size of 0.2 µM (Advantec). The cell lysates were prepared as for western blot analysis. The same amount of protein from each experimental condition was diluted to 100 µl in PBS with 2% SDS and applied onto the membrane. Soluble proteins were removed by vacuum suction whereas the SDS-resistant aggregates stay trapped. Wells were washed three times with 2% SDS/PBS and suction was maintained for 20 min to allow complete and tight trapping of SDS insoluble material. Membranes were subsequently blocked with 5% skim milk and immunoblot was performed. Agarose gel electrophoresis for resolving aggregates. Agarose gel electrophoresis provides a simple and sensitive biochemical detection method for quantitative and qualitative investigations of aggregate formation in HD models29. Briefly, 1.5% agarose gels (BMBio) were prepared in 375 mmol/l Tris–HCl, pH 8.8 buffer and brought to boiling in a microwave oven. After melting, SDS was added to a final concentration of 0.1%. Samples were diluted in 2× nonreducing Laemmli sample buffer (150 mmol/L Tris–HCl pH 6.8, 33% glycerol, 1.2% SDS and bromophenol blue) and incubated for 5 min at 95 °C. After loading, gels were run in Laemmli running buffer (192 mmol/l glycine, 25 mmol/l Tris-base, 0.1% SDS). Semi-dry electroblotter Trans-blot s.d. Cell (Bio-Rad) was used to blot the gels on PDVF membranes in the transfer buffer containing 192 mmol/l glycine, 25 mmol/l Tris-base, 0.1% SDS, and 15% methanol. After transfer, starting with the blocking step, immunoblot membranes were processed as described in immunoblotting. Histology. Serial-cut 20-µm sections were used for immunohistochemistry after fixation in 4% paraformaldehyde. Sections were treated with antibodies against RFP, ubiquitin and HTT (EM48) followed by detection using ABC Elite kit (Vector Laboratories). The images were taken by Olympus DP50 microscope. For direct RFP and GFP fluorescence, frozen sections were observed using Biozero (Keyence). In situ hybridization. For in situ hybridization, serial-cut 40-µm sections and nonradioactive digoxigenin-labeled cRNA probe against DARPP-32 were used as described previously30. IMAGE 2938032 clone (Invitrogen) was used to synthesize the probe against DARPP-32. The in vivo study of rAAV-HQ in R6/2 mice. To address the beneficial effect of (R)HQ in vivo, we employed the R6/2 mouse model in which the progressive HD pathology is well characterized and has been extensively used for preclinical drug testing31. Viral vectors encoding R, Q and HQ were injected to the left and right striatum at 4 weeks of age. Body weight was measured every second week from 4 to 14 weeks of age. The clasping and rotarod performance was tested at the same time animals were weighed starting from 4 and 6 weeks, respectively. For the clasping score, mice were suspended by the tail for 30 s and the clasping phenotype was graded to a particular level according to the following scale: 0, no clasping; 1, clasping of the forelimbs only; 2, clasping of both fore- and hind limbs once or twice; 3, clasping of both fore- and hind limbs more than 3 times or more than 5 s. Before each rotarod testing, mice were first trained on a rotating rod moving at 4 r.p.m. for 5 min. The testing itself was done on the rotating rod with linearly increasing speed from 4 r.p.m. up to 45 r.p.m.
nature biotechnology
in 300 s. Mice of 6–12 weeks of age were all subjected to rotarod test with the same moving speed. For the survival distribution, the number of days each mouse survived was recorded and the data collected (R/R, n = 11; Q/Q, n = 11; HQ/HQ, n = 11) were subjected to Kaplan-Meier analysis.
© 2010 Nature America, Inc. All rights reserved.
Statistical analysis. We used unpaired student’s t-test for comparison between two sample groups. One-way ANOVA Fisher’s test followed by Tukey’s HSD test was used for multiple comparisons with a 95% confidence level. We generated these data with XLSTAT software. For survival rate we plotted the survival distribution curve with the Kaplan-Meier method followed by log-rank and Wilcoxon testing (JMP Statistical Discovery software, SAS Institute). We considered the difference between comparisons to be significant when P < 0.05 for all the statistical analysis.
57. Wang, G.H. et al. Caspase activation during apoptotic cell death induced by expanded polyglutamine in N2a cells. Neuroreport 10, 2435–2438 (1999). 58. Wang, G.H., Sawai, N., Kotliarova, S., Kanazawa, I. & Nukina, N. Ataxin-3, the MJD1 gene product, interacts with the two human homologs of yeast DNA repair protein RAD23, HHR23A and HHR23B. Hum. Mol. Genet. 9, 1795–1803 (2000). 59. Peters, M.F. & Ross, C.A. Preparation of human cDNas encoding expanded polyglutamine repeats. Neurosci. Lett. 275, 129–132 (1999). 60. Zemskov, E.A. et al. Pro-apoptotic protein kinase C delta is associated with intranuclear inclusions in a transgenic model of Huntington’s disease. J. Neurochem. 87, 395–406 (2003). 61. Madden, E.A., Wirt, J.B. & Storrie, B. Purification and characterization of lysosomes from Chinese hamster ovary cells. Arch. Biochem. Biophys. 257, 27–38 (1987).
nature biotechnology
doi:10.1038/nbt.1608
Articles
Directed evolution of a magnetic resonance imaging contrast agent for noninvasive imaging of dopamine
© 2010 Nature America, Inc. All rights reserved.
Mikhail G Shapiro1,5, Gil G Westmeyer2,5, Philip A Romero3, Jerzy O Szablowski1, Benedict Küster2, Ameer Shah1, Christopher R Otey3, Robert Langer1, Frances H Arnold3 & Alan Jasanoff1,2,4 The development of molecular probes that allow in vivo imaging of neural signaling processes with high temporal and spatial resolution remains challenging. Here we applied directed evolution techniques to create magnetic resonance imaging (MRI) contrast agents sensitive to the neurotransmitter dopamine. The sensors were derived from the heme domain of the bacterial cytochrome P450-BM3 (BM3h). Ligand binding to a site near BM3h’s paramagnetic heme iron led to a drop in MRI signal enhancement and a shift in optical absorbance. Using an absorbance-based screen, we evolved the specificity of BM3h away from its natural ligand and toward dopamine, producing sensors with dissociation constants for dopamine of 3.3–8.9 M. These molecules were used to image depolarization-triggered neurotransmitter release from PC12 cells and in the brains of live animals. Our results demonstrate the feasibility of molecular-level functional MRI using neural activity–dependent sensors, and our protein engineering approach can be generalized to create probes for other targets. MRI is a uniquely valuable tool for studying the brain because MRI scans are noninvasive and can provide information at relatively high spatial resolution (< 100 µm) and temporal resolution (~1 s) from living specimens. Functional imaging (fMRI) of brain activity is possible with MRI methods sensitive to cerebral hemodynamics1. The most common fMRI technique, blood oxygen level–dependent (BOLD) fMRI, is based on oxygenation of hemoglobin, an endogenous oxygen-sensitive MRI contrast agent present in the blood2. Although BOLD fMRI has had a tremendous impact in neuroscience, the method provides only a slow and indirect readout of neural activity, owing to the complexity of neurovascular coupling3. Considerably more precise measurements of brain function would be possible with MRI sensors that were directly and rapidly responsive to neurochemicals involved in the brain’s information processing4. The challenging process of developing sensors for next-generation neuroimaging could be greatly accelerated using advanced molecular engineering techniques. Directed evolution is a molecular engineering method that employs successive rounds of mutagenesis and selection to generate proteins with novel functionality, starting from a molecule with some of the desired properties of the end product5. This technique could be applied to evolve MRI sensors from proteins that are magnetically active (for example, paramagnetic) and have tunable ligand-binding or catalytic properties. The flavocytochrome P450-BM3 (BM3), a fatty acid hydroxylase from Bacillus megaterium, contains a paramagnetic iron atom embedded in a solvent-accessible substrate-binding pocket, suggesting that it could produce ligand-dependent MRI signal changes. BM3’s binding specificity is also highly tunable, as demonstrated by previous efforts to identify novel enzymatic activities through directed evolution of
this protein6–9. If BM3 variants could be engineered to act as MRI sensors, they would be genetically encodable, an added advantage over synthetic molecular imaging agents. We sought to apply directed evolution of BM3 to develop MRI sensors for a key signaling molecule in the brain, the neurotransmitter dopamine. To our knowledge, no MRI contrast agent for sensing dopamine (or any other neurotransmitter) currently exists, but there is considerable interest in measuring dopamine-related activity by MRI10. Dopamine is of particular significance because of its roles in learning, reward and motor coordination11, and because the dysfunction of dopaminergic systems underlies addiction12 and several neurodegenerative diseases13. Existing techniques for measuring dopamine in vivo are either invasive point-measurement methods14–16 or positron emission tomography procedures17 with low spatial and temporal resolution. MRI could be used successfully for dopamine measurement if combined with an imaging agent capable of responding quickly, reversibly and specifically to extracellular dopamine fluctuations from <1 µM to tens of micromolar18,19. To be comparable with established functional brain imaging techniques, interaction of dopamine with the probe should also produce image signal changes on the order of 1% or more in vivo20. Here we show that directed evolution of BM3 is capable of producing dopamine sensors that largely meet these specifications. RESULTS P450 BM3 reports ligand binding in MRI To evolve dopamine probes for MRI, we focused on the heme domain of BM3 (BM3h), a 53-kDa moiety that is catalytically inactive in the absence of the full protein’s reductase domain21. BM3h contains a
1Department of Biological Engineering and 2Department of Brain & Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA. 3Division of Chemistry & Chemical Engineering, California Institute of Technology, Pasadena, California, USA. 4Department of Nuclear Science & Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA. 5These authors contributed equally to this work. Correspondence should be addressed to A.J. ([email protected]).
Received 25 June 2009; accepted 27 January 2010; published online 28 February 2010; doi:10.1038/nbt.1609
264
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
Articles b
1.6
0.30 1/T1 (s–1)
1.2 0.8 0.4
0.28 PBS
AA
DA
OD
c 0.10 0.06 0.02
© 2010 Nature America, Inc. All rights reserved.
330
0
400 µM 1 mM
50
480
d 0.05
e
0
330
150
200
250
[AA] (µM)
–0.05 380 430 λ (nm)
100
Rel. change
0
0.29
∆OD
r1 (mM–1 s–1)
a
380 430 λ (nm)
480
1.0 0.5 0 –8
–6 –4 –2 Log(conc.)
single iron(III) atom (mixed spin 1/2 and 5/2)22 bound to a hemin prosthetic group and axially coordinated by residue Cys400 on the protein. In the absence of substrates, the remaining coordination site is filled by a water molecule23. Interaction of the heme iron with exchanging water molecules at this axial site promotes T1 relaxation in aqueous solutions24 and is therefore predicted to modulate MRI contrast. To determine the extent of this effect, we used a spin echo pulse sequence in a 4.7-T MRI scanner to measure the proton relaxation rate as a function of protein concentration in PBS; the slope of this relationship (T1 relaxivity, or r1) provides a standard measure of the strength of a contrast agent. For BM3h in the absence of ligands, an r1 value of 1.23 ± 0.07 mM−1 s−1 was obtained. Addition of a saturating quantity of the natural BM3 substrate, arachidonic acid (400 µM concentration), resulted in an r1 of 0.42 ± 0.05 mM−1 s−1 (Fig. 1a). This ligand-induced decrease in relaxivity, probably arising from the displacement of water molecules at the BM3h heme, enabled quantitative sensing of arachidonic acid using MRI (Fig. 1b) and suggested that BM3h could serve as a platform for molecular sensor engineering. We next tested whether dopamine or related compounds could serve as unnatural ligands to BM3h when applied at high enough concentrations. As measured by MRI, addition of 1 mM dopamine to BM3h in fact induced a drop in r1 to 0.76 ± 0.03 mM−1 s−1 (Fig. 1a). Binding of arachidonic acid is known to induce a change (blue shift) in BM3h’s optical absorbance spectrum because of perturbation of the electronic environment of the heme chromophore25. To determine whether the relaxation change induced by dopamine also reflects interaction with the BM3h heme, we measured optical spectra of the protein in the presence and absence of 1 mM dopamine. The interaction produced a small but clearly discernable red shift of λmax, from 419 to 422 nm (Fig. 1c), indicative of ligand coordination to the heme iron25. This suggests that dopamine (at 1 mM) directly replaces water as an axial metal ligand in the BM3h substrate-binding pocket and that directed evolution of BM3h binding specificity could therefore improve the protein’s relative affinity for dopamine. In addition to providing mechanistic insight, the correspondence between optical and MRI measurements of ligand binding to BM3h implied that either modality could be used to obtain quantitative binding parameters. We monitored the difference between absorption at two wavelengths as a function of ligand concentration to determine binding isotherms for arachidonic acid and dopamine (Fig. 1d,e). For BM3h, the apparent Kd for arachidonic acid was 6.8 ± 0.5 µM; the Kd for dopamine was 990 ± 110 µM. Goals for the production of BM3h-based MRI sensors thus included decreasing the affinity
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
Figure 1 Ligand binding to the BM3 heme domain changes MRI contrast and optical absorption in a concentration-dependent manner. (a) T1 relaxivity (r1) of BM3h in PBS solution and in the presence of 400 µM arachidonic acid (AA) or 1 mM dopamine (DA); inset shows T1-weighted spin echo MRI image intensity (TE/TR = 10/477 ms) of microtiter plate wells containing 240 µM BM3h in PBS alone (left) or in the presence of 400 µM arachidonic acid (middle) or 1 mM dopamine (right). (b) T1 relaxation rates (1/T1) measured from solutions of 28.5 µM BM3h incubated with 0–250 µM arachidonic acid. (c) Optical absorbance spectra of 1 µM BM3h measured alone (blue) and after addition of 400 µM arachidonic acid (gray) or 1 mM dopamine (orange). OD, optical density. (d) Difference spectra showing the change in BM3h absorbance as a function of wavelength upon addition of 400 µM arachidonic acid (gray) or 1 mM dopamine (orange). (e) Normalized titration curves showing binding of BM3h to arachidonic acid (gray) or dopamine (orange). We computed the optical signals used for titration analysis by subtracting the minimum from the maximum of difference spectra (arrowheads in d) under each set of conditions. Error bars in a, b and e reflect s.e.m. of three independent measurements (errors in e were smaller than the symbols).
for arachidonic acid, increasing the dopamine affinity by at least two orders of magnitude and maintaining or enhancing the relaxivity changes observed upon ligand binding. Directed evolution of dopamine-responsive BM3h variants To create an MRI sensor for dopamine using directed evolution, we developed a customized screening methodology (Fig. 2a). Results shown in Figure 1 suggested that either MRI-based or optical assays could be used to distinguish BM3h mutants with differing ligand affinities. We chose an absorbance assay for our screen because lower protein concentrations (~1 µM) could be used in this format. Input to each round of screening consisted of a library of BM3h mutants, each with an average of one to two amino acid substitutions, generated by error-prone PCR from the wild-type (WT) gene or a previously selected mutant. We transformed DNA libraries into Escherichia coli. We grew and induced approximately 900 randomly selected clones in microtiter plate format, then prepared cleared lysates for optical titration with dopamine and arachidonic acid in a plate reader. Titration data were analyzed to determine Kd values for both ligands. An average of 79% of assayed mutants had sufficient protein levels (absorbance signal > 30% of parent) and clean enough titration curves (r 2 > 0.8) for Kd estimation. Mutant affinities appeared to be distributed randomly about the dissociation constant measured for the corresponding parent protein, but we were able to identify individual clones with desired affinity changes in each round (Fig. 2b). From each screen, we chose eight to ten mutants on the basis of their estimated Kds, purified them in bulk, re-titrated them to obtain more accurate estimates of their dopamine and arachidonic acid affinities, and examined them with MRI to ensure that robust ligand-induced changes in r1 could be detected. On the basis of these assays, we chose as a parent for the next round of evolution the mutant showing the best combination of relaxivity changes, improved dopamine affinity and decreased affinity for arachidonic acid. After carrying out the screening strategy over multiple rounds, we found a steady trend in the distribution of Kd values toward greater affinity for dopamine and less affinity for arachidonic acid (Fig. 2b–d). Little change in binding cooperativity was observed, and changes in partial saturation generally occurred over 100-fold ranges of dopamine concentrations. Five rounds of evolution yielded a BM3h variant with eight mutations (Fig. 2e), four near the ligand-binding pocket and four at distal surfaces of the protein. One residue (Ile263) was first mutated to threonine (third round), then to alanine (fourth round). The clones selected from rounds 1, 3 and 5 had two new mutations each. We did not determine the individual contributions of
265
Articles a
Figure 2 Screen-based isolation of BM3h mutants with enhanced dopamine affinity. (a) Schematic of the directed evolution approach, including (left to right) generation of a mutant DNA library, transformation into E. coli and growth in multiwell plate format, spectroscopic analysis of each mutant’s ligand binding affinities, and detailed MRI and optical characterization of selected mutant proteins. (b) Histograms of mutant dopamine dissociation constants determined during each round of directed evolution, comparing each mutant protein’s relative dopamine affinity (measured in plate format) to the Kd of the parent protein (measured in bulk). Kd distributions for screening rounds 1 (black), 2 (green), 3 (red), 4 (cyan) and 5 (purple) are labeled with numbers in circles. Color-coded arrowheads indicate the measured Kds of parent proteins used to create the library of mutants at each round; yellow arrowhead indicates the Kd of the mutant protein selected after round 5. (c) Dissociation constants for dopamine (DA; orange) and arachidonic acid (AA; gray) for WT BM3h and mutant BM3h variants isolated at each round of screening; progressive increases in dopamine affinity and attenuation of arachidonic acid affinity are evident. Colored arrowheads indicate correspondence with data in b. Error bars denote s.e.m. of three independent measurements. (d) Titration analysis of dopamine binding to WT BM3h and to proteins selected after each round of directed evolution (colored as in b). Mutant proteins identified by rounds 4 (8C8) and 5 (B7) were considered to be end products of the screening procedure. (e) X-ray crystal structure34 of WT BM3h (gray; heme group shown in orange) bound to palmitoleic acid (black), indicating the locations of amino acid substitutions accumulated during directed evolution of enhanced dopamine binding affinity. Each mutation’s location is marked with a blue sphere and a label color-coded according to the parent protein in which the substitution was first identified (see legend for b). The previously characterized I366V mutation (asterisk) was incorporated between screening rounds 4 and 5 to improve the thermostability of the engineered proteins.
H105D10 EP3C8 H103D8 H104D3 H106B5 H104B3 H103D1 H105D9 C810H10 H106H5
10
11
12
H
5
30
Relative change
c 2
10
1.0 0.8
–2
DA
–3 1
–4
AA
–5 –5
–4 Log(Kd )
–3
–6
–2
e
0
V286E
1 2 3 4 Rounds of evolution
T268A
I366V*
5
Q189R
B7 8C8
0.6 0.4 0.2
Y305H
WT
0 –6
–4 Log([DA])
F81L
I263A
–2
L75P
I263T
protein) that could be fitted by binding isotherms with estimated Kd values of 4.9 ± 2.7 µM for BM3h-8C8 and 2.7 ± 2.9 µM for BM3h-B7 (Fig. 3b,c). For both BM3h variants, the stability, reversibility and rate of dopamine binding were established using spectroscopic assays (Supplementary Figs. 1 and 2). We investigated the reporting specificities of BM3h-8C8 and BM3h-B7 for dopamine by measuring MRI signal changes that resulted from incubation of 28.5 µM of each protein with 30 µM of either dopamine or one of eight other neuroactive molecules: norepinephrine (a neurotransmitter formed by catalytic hydroxylation of dopamine), 3,4-dihydroxy-l-phenylalanine (DOPA, the biosynthetic precursor to dopamine), serotonin, glutamate, glycine, γ-aminobutyric acid (GABA), acetylcholine and arachidonic acid (Fig. 3d). Of these potential ligands, only dopamine, norepinephrine and serotonin elicited
these mutations to the observed changes in affinity. We introduced the mutation I366V by site-directed mutagenesis before the fifth round to enhance thermostability and tolerance of BM3h to further mutation26,27; it did not noticeably affect dopamine binding affinity. The mutant proteins selected after the fourth and fifth rounds of evolution, denoted BM3h-8C8 and BM3h-B7, had optically determined dissociation constants of 8.9 ± 0.7 µM and 3.3 ± 0.1 µM, respectively, for dopamine, and 750 ± 140 µM and 660 ± 80 µM, respectively, for arachi donic acid. The T1 relaxivity of BM3h-8C8 was 1.1 ± 0.1 mM−1 s−1 in the absence of ligand and 0.17 ± 0.03 mM−1 s−1 in the presence of 400 µM dopamine (Fig. 3a). For BM3h-B7, the corresponding r1 values were 0.96 ± 0.13 mM−1 s−1 and 0.14 ± 0.04 mM−1 s−1. Both sensor variants showed a dopamine concentration–dependent decrease in T1-weighted MRI signal (up to 13% with 28.5 µM
WT
0.8
B7 8C8
0.4
8C8 B7
0
PBS
DA
8
16
31
63
d
0.30 0.29
0.01
0.28
0
∆(1/T1) (s–1)
1.2
c
[DA] (µM) 0
1/T1 (s–1)
b
1.6
–1
–1
s )
a r1 (mM
© 2010 Nature America, Inc. All rights reserved.
3
20
0 –6
d
4
Log(Kd )
Relative frequency
b
0.27 0.26 0.25
0
50
100
150
[DA] (µM)
500
–0.01 –0.02 –0.03
Gly
9
Glu
8
GABA
7
G
ACh
6
F
AA
5
E
5HT
4
D
NE
3
C
DOPA
2
B
DA
1
A
DA NE 5HT 0 0.1 0.2 0.3 Ka (µM–1)
Figure 3 Selected sensor proteins produce strong and specific MRI signal changes in response to dopamine. (a) Relaxivity values measured from BM3h-B7 (yellow bars) and BM3h-8C8 (purple bars) in PBS alone or in the presence of 400 µM dopamine (DA). Inset, T1-weighted MRI signal (TE/TR = 10/477 ms) obtained from 195 µM BM3h-B7 or BM3h-8C8, each incubated in microtiter plate wells with or without 400 µM dopamine (wells ordered left to right as in the bar graph). (b) MRI image showing signal amplitudes measured from wells containing 28.5 µM WT BM3h, BM3h-8C8 or BM3h-B7, each incubated with increasing dopamine concentrations (0–63 µM, left to right). The image was obtained using a T1-weighted pulse sequence (TE/TR = 10/477 ms). (c) Relaxation rates (1/T1 values) measured from solutions of 28.5 µM WT BM3h (black), BM3h-B7 (yellow) or BM3h-8C8 (purple), as a function of total dopamine concentration. Curves were fitted using a ligand-depleting bimolecular association model. (d) Changes in 1/T1 relative to ligand-free protein for 28.5 µM BM3h-B7 (yellow) or BM3h-8C8 (purple) incubated with 30 µM dopamine, serotonin (5HT), norepinephrine (NE), DOPA, arachidonic acid (AA), acetylcholine (ACh), GABA, glutamate or glycine. Inset, spectroscopically determined affinities (Ka = 1/Kd) of BM3hB7 and BM3h-8C8 for dopamine, serotonin and norepinephrine. Error bars in panels a, c and d denote s.e.m. of three independent measurements.
266
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
Articles
© 2010 Nature America, Inc. All rights reserved.
substantial changes in the T1 relaxation rate (1/T1). For BM3h-8C8, the 1/T1 reductions produced by norepinephrine and serotonin were 0.0076 ± 0.0023 s−1 and 0.0041 ± 0.0020 s−1, respectively, compared to 0.0182 ± 0.0006 s−1 for dopamine; for BM3h-B7, norepinephrine and serotonin induced 1/T1 decreases of 0.0112 ± 0.0024 s−1 and 0.0171 ± 0.0005 s−1, respectively, compared to 0.0208 ± 0.0002 s−1 for dopamine. We measured the affinities of BM3h-based dopamine sensors for these competitors spectroscopically (Fig. 3d, inset). For BM3h-8C8, measured Kds were 44 ± 3 µM and 80 ± 8 µM for norepi nephrine and serotonin, respectively, and for BM3h-B7 the Kd values were 18.6 ± 0.4 µM and 11.8 ± 0.1 µM, respectively. Although both BM3h-8C8 and BM3h-B7 show substantially higher affinity for dopamine than for norepinephrine (fivefold and sixfold, respectively) or for serotonin (ninefold and fourfold, respectively), the BM3h-8C8 variant is more specific for sensing dopamine at concentrations above 10 µM. In settings where dopamine is known to be the dominant neurotransmitter, BM3h-B7 may provide greater overall sensitivity. The specificity data also provided a possible indication of the geometry of dopamine binding to the evolved BM3h proteins. Only monoamines showed affinity for BM3h-8C8 and BM3h-B7, whereas two catechols that lack primary amines, epinephrine and 3,4-dihydro phenylacetic acid, showed no measurable affinity (data not shown). Combined with the spectral evidence that dopamine directly coordinates the BM3h heme (Fig. 1c), the titration results therefore suggest that the dopamine amine may serve as an axial ligand to the BM3h heme in the sensor-analyte complexes we examined. BM3h-based sensors detect dopamine released from PC12 cells We asked whether BM3h mutants produced by directed evolution could sense dopamine release in a standard cellular model of dopaminergic function. We applied an established protocol28 to test the ability of our sensors to measure dopamine discharge from PC12 cells stimulated with extracellular K+ (Fig. 4a). Cells were cultured in serum-free medium supplemented with dopamine to promote packaging of the neurotransmitter into vesicles. After pelleting and washing, we resuspended cells in a physiological buffer containing 32 µM BM3h-B7 and either 5.6 or 59.6 mM K+ (cells in the low-K+ condition were osmotically balanced with Na+). T1-weighted MRI images (spin echo TE/TR = 10/477 ms) obtained with BM3h-B7 showed a 4.0 ± 0.5% reduction in signal intensity in the supernatant of K+-stimulated cells, compared with cells for which isotonic Na+ was used as control (Fig. 4b). This corresponded to a 54 ± 4% decrease in sensor r1 (Fig. 4c). Given the dopamine dissociation constant of BM3h-B7 and its relaxivities under ligand-free and dopamine-saturated conditions, and assuming negligible dilution of the sensor after mixing with cells, we estimated supernatant dopamine concentrations of 60.3 ± 7.9 µM for stimulated cells and 22.2 ± 1.1 µM for controls. These estimates were in reasonable agreement with an independent quantification of dopamine release measured using an enzyme-linked immunosorbent
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
b
c
d
∆(1/T1) (s–1)
[DA] (µM)
a
Relative signal
Figure 4 BM3h-based sensors measure dopamine release in cell culture. 0.020 80 0.185 MRI BM3h (a) PC12 cells depolarized by addition of 54 mM K+ were stimulated to release ELISA 0.180 dopamine (DA) into supernatants containing a BM3h-based sensor; cells did 0.015 60 DA + not release dopamine after addition of 54 mM Na . (b) T1-weighted spin echo 0.175 0.010 40 MRI signal amplitudes (TE/TR = 10/477 ms) measured from the supernatants 0.170 + of PC12 cells incubated with 32 µM BM3h-B7 in the presence of K (stimulus) 20 0.005 0.165 or Na+ (control). Inset, MRI image of microtiter wells under corresponding 0.160 0 0 conditions. (c) Relaxation rates measured from the samples in b, minus K+ Na+ K+ Na+ K+ Na+ K+ Na+ the relaxation rate of buffer not containing BM3h-based sensors. Given the approximate concentration of BM3h variants in these samples, the ∆(1/T1) values presented here can be converted to apparent relaxivities of 0.23 and 0.50 mM−1 s−1 in K+ and Na+ incubation conditions, respectively. (d) Data from c were used to estimate the concentrations of dopamine present in samples treated with K+ and Na+ (dark bars). We independently measured the concentrations of dopamine under equivalent conditions using ELISA (light bars).
assay (ELISA), which yielded concentrations of 54 ± 9 µM and 13 ± 2 µM for stimulated and control cells, respectively (Fig. 4d). We were also able to use BM3h-8C8 to image dopamine release from PC12 cells. Under experimental conditions similar to above, BM3h-8C8 had a 37 ± 2% reduction in r1 in the supernatant of K+-stimulated cells relative to Na+ controls (Supplementary Fig. 3). Dopamine detection in the brain of living rats As an initial test of the ability of BM3h-based sensors to measure dopamine concentrations in intact animals, we injected BM3h-8C8 in the presence or absence of exogenous dopamine into the brains of anesthetized rats. We chose this simple experimental protocol for validation of the sensor because it guaranteed the presence of reproducible and unambiguous micromolar-level dopamine concentrations, suitable for evoking robust responses from our sensors in vivo. We obtained T1-weighted MRI scans (fast spin echo TE/TR 14/277 ms, 8.9 s per image) continuously during 0.5-µl-min−1 paired infusions of 500 µM BM3h-8C8 with and without 500 µM dopamine, via cannulae implanted stereotaxically into the left and right striatum. Dopamine-dependent contrast changes were apparent in images obtained during and after the injection period (Fig. 5a). We quantified MRI changes across multiple trials in striatal regions of interest (ROIs) that were reliably (though inhomogeneously) filled by convective spread of the contrast agent from the cannula tips (~1.5 mm radius). Consistent with results obtained in vitro, addition of dopamine dampened the observed MRI intensity enhancement by approximately 50% (Fig. 5b); the effect was significant (t-test, P = 0.003, n = 7). We performed the same paired infusion procedure with WT BM3h, which has very low affinity for dopamine (Kd ~1 mM). As expected, the time course of the MRI signal during and after the WT BM3h injection period (Fig. 5c) was not significantly affected by the presence or absence of dopamine (t-test, P = 0.8, n = 5), indicating that the dopamine-dependent signal differences shown in Figure 5b require the presence of a micromolar-affinity dopamine sensor and cannot be explained by physiological or biochemical effects of dopamine itself. Moreover, infusion of 500 µM dopamine alone into the brain produced no noticeable signal changes in an equivalent experiment (data not shown). Histological analysis showed minimal evidence of toxicity due to these procedures (Supplementary Fig. 4). Using relaxivity values measured for BM3h-8C8 in vitro, we estimated maximal concentrations of 89 ± 19 µM BM3h-8C8 and 75 ± 28 µM dopamine from the data of Figure 5b, averaged across the striatal ROIs. The ability to quantify BM3h-8C8 concentration on the basis of its T1 enhancement in the absence of elevated dopamine represents an advantage of this sensor’s ‘turn-off ’ mechanism. To test whether BM3h-8C8 could detect release of endogenous neurotransmitters in the rat brain, we acquired MRI data during co-infusion of the dopamine sensor with elevated concentrations of K+, a depolarizing chemical stimulus shown previously to release large amounts of
267
Articles dopamine into the striatum29,30. We chose K+ over pharmacological stimuli to obviate potential solubility- or viscosity-related artifacts in the experimental paradigm. K+ itself had no effect on r1 of the BM3h variants (data not shown). In the stimulation experiments, three 5-min blocks of high-K+ (153 mM) infusion alternated with 10-min ‘rest’ periods during which we administered a low-K+ solution (3 mM, osmotically balanced with Na+). Both high- and low-K+ solutions were delivered at a rate of 0.2 µl min−1 and also contained 500 µM BM3h-8C8, ensuring that a relatively constant concentration of dopamine sensor was present throughout the procedure. We acquired T1-weighted MRI scans continuously as for the exogenous dopamine infusion experiments. To control for effects unrelated to neurotransmitter sensing by the contrast agent (potentially including K+-induced edema or hemodynamic responses incompletely suppressed by the T1-weighted spin echo pulse sequence), we paired each striatal injection of BM3h-8C8 with an injection of WT BM3h into
b
c
Percent change
Percent change
Figure 5 BM3h-8C8 reports dopamine in BM3h-8C8 WT BM3h injected rat brains. (a) Top, coronal MRI image 30 30 (0.7 mm anterior to bregma, averaged over the –DA injection period) from a single rat injected with 20 20 500 µM BM3h-8C8 in the presence (orange dashed circle) or absence (blue dashed circle) %∆ of equimolar dopamine; the image contrast was 10 10 25 linearly adjusted for display. MRI hyperintensity is noticeable near the tip of the dopamine-free 0 0 +DA cannula. The circles indicate approximate ROIs 0 5 10 15 20 25 30 0 5 10 15 20 25 30 (~1.5 mm around cannula tips) over which –25 Time (min) Time (min) image intensity was averaged for quantitative analyses. Bottom, map of percent signal –%SD change (%∆) for the same animal, computed 1.1 by comparing pre- and post-injection MRI –logP signal. Areas corresponding to both high- and 5.7 low-dopamine co-injections (+DA and −DA) are delineated by apparent signal changes, but the strong difference between the two conditions is clear. (b) Time courses of relative signal change observed during injection of BM3h-8C8 −DA BM3h-8C8 WT BM3h 0 (blue) or +DA (orange), averaged over multiple 0 animals (n = 7) in ROIs denoted in a. Gray shading denotes the 20-min injection period. 0.04 (c) Corresponding time courses of a control * injection in which WT BM3h was introduced 3 0.02 instead of the dopamine sensor (n = 5). 2 (d) Statistical parametric map of t-test 0.0 1 significance values (color scale) for correlation –0.02 0 of MRI intensity with low- and high-K+ conditions in an individual rat, overlaid on –1 –0.04 a corresponding T1-weighted coronal slice Low K+ –2 (grayscale) showing injection cannulae used –0.06 High K+ –3 BM3h-8C8 for BM3h-8C8 infusion (left, purple dashed –0.08 20 30 40 50 60 70 circle) and WT BM3h control infusion (right, 8C8 WT Time (min) black dashed circle). (e) Maps of percent signal difference (SD) between high- and low-K+ conditions observed in 2.7-mm-diameter ROIs centered around BM3h-8C8 sensor (left) and WT BM3h control (right) injection sites, after spatial coregistration and averaging across multiple animals (n = 6); ROIs correspond approximately to the color-coded circles in d. Voxels outlined in green are those that showed the most significant correlation with the K + stimulus regressor in the group analysis (Student’s t-test, P < 0.01); these generally showed ~1% mean signal change. Gray cross-hatching indicates approximate locations of the infusion cannulae. (f) Mean MRI signal change from baseline observed during high-K+ (dark bars) and low-K+ (light bars) periods in ROIs centered around infusion sites for BM3h-8C8 (purple) and WT BM3h (gray) proteins. ROIs were cylinders 2.7 mm in diameter and extending over three 1-mm-thick slices registered around the infusion sites; signal was averaged in unbiased fashion over all voxels, regardless of correlation with the stimulus. The signal difference in the presence of BM3h-8C8 was statistically significant (P = 0.0008, asterisk). (g) Graph shows the mean time course of MRI signal in voxels within the BM3h-8C8–infused ROI and identified as correlated (P < 0.05) with the stimulus, averaged over animals and binned over 1.5-min intervals (shaded area denotes s.e.m., n = 6; individual traces are shown in Supplementary Fig. 6 online). Gray vertical bars denote periods associated with highest K + stimulation, accounting for delays due to convective spreading of K+ from the cannulae tips and the dead time of the injection apparatus. Arrowheads indicate the timing of pump switches associated with transitions from low to high (up) and from high to low (down) K + infusion conditions. Panels above the graph depict ‘snapshots’ of signal change spaced throughout the first K+ stimulation cycle, as indicated by the dotted lines. The ROI corresponds to the left side of e, and the color scale denotes 0% (black) to 3% (yellow) signal change from baseline at each voxel and time point.
d
268
e
g
Percent change
f
Percent change
© 2010 Nature America, Inc. All rights reserved.
a
the opposite hemisphere, following the same blocked K+ stimulation paradigm for both injections. As in conventional ‘block design’ fMRI, we performed a t-test analysis to evaluate the correspondence of each voxel’s intensity time course with the alternating periods of low and high K+. We determined an appropriate temporal shift for the stimulus-related analysis windows with respect to infusion buffer switches by observing the time courses of similarly switched mock infusions into 0.6% agarose phantoms31 and by comparing these with statistical results as a function of offset (Supplementary Fig. 5 and Online Methods). As additional controls for MRI effects unrelated to dopamine sensing, we examined MRI signal change in response to K+ stimulation and again in response to dopamine infusion, both in the absence of contrast agents (data not shown). We also continuously monitored blood oxygen levels and heart rate. In no case were stimulus-associated changes observed.
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
Articles Figure 5d shows the distribution of voxels with significant (t-test, P < 0.01) MRI signal decreases in response to K+ stimulation in a single rat. We performed a group analysis by combining data from all subjects (n = 6) over geometrically defined ROIs centered around the injection cannula tips in each animal. In three slices spanning the infusion site, seven voxels within 0.75 mm of the BM3h-8C8 injection cannula, but only one voxel near the WT cannula, showed strong correlation (P < 0.01) with the stimulus. We mapped mean signal decreases over 2.7-mm-diameter ROIs corresponding to the BM3h-8C8 and WT BM3h injection sites in the group analysis (Fig. 5e). Again, dopamine sensor–dependent responses were apparent. The signal difference between low- and high-K+ periods averaged across the entire BM3h-8C8 ROI (all voxels within a 2.7-mm-diameter by 3-mm-long cylinder, regardless of modulation by K+) was 0.07%, whereas the signal difference averaged across the control ROI was −0.02% (Fig. 5f). The high- versus low-K+ signal difference observed near the BM3h-8C8 infusion site was significant (t-test, P = 0.0008) and consistent with the expected suppression of MRI signal by dopamine release under high-K+ conditions. The mean time course of all stimulus-correlated voxels (P < 0.05) showing K+-induced MRI signal changes near the BM3h-8C8 injection site, averaged over animals, is shown in Figure 5g. Discernable signal decreases of up to 3% were produced during each K+ stimulation block. The first K+ block evoked the largest response (presumably because of partial dopamine depletion over subsequent blocks32) and elicited a clear spatiotemporal pattern of mean MRI signal change from baseline over the course of the stimulation period (Fig. 5g, top panels). DISCUSSION These results demonstrate the feasibility of developing molecularlevel fMRI sensors and serve as a proof of principle that BM3h-based probes can be used to monitor dopamine signaling processes in vivo. With the experimental conditions and estimated sensor concentrations (34 ± 4 µM) used for our K+ stimulation experiments, MRI signal changes of ~3% would be evoked by the rewarding brain stimuli reported in previous studies to release large amounts of dopamine18,19. This amplitude is reasonably large by functional imaging standards, and it could be used in the near term to map phasic dopamine release at high resolution across the striatum, or more generally to study meso limbic dopamine dynamics in animal models of reward processing and neurological conditions that can be probed with strong stimuli. Sensitivity gains will be possible using repeated stimulation and statistical analysis techniques, as in conventional fMRI, and by optimizing the imaging approach itself. For instance, higher-field scanners and faster alternatives to the T1-weighted spin echo pulse sequences we used here may offer improved signal-to-noise ratios. Directed evolution or rational modification of BM3h variants for substantially higher relaxivity is possible as well (unpublished data). Sensors with higher relaxivity will produce larger MRI signal changes, and could have the added benefit of reducing the potential for dopamine buffering, because they may be used at lower concentrations in vivo: with 35 µM sensor and 35 µM total dopamine present, for example, ~60% of the dopamine would be bound to the sensor, but with 15 µM sensor present, only ~30% dopamine would be sequestered. Protein engineering techniques could also be used to improve the dopamine affinity and specificity of the first-generation sensors described here. Our method for producing dopamine sensors represents a general paradigm for the development of molecular probes for MRI. Sensors may be evolved for targets inside or outside the brain; the diversity of potential targets is exemplified by the contrast between WT BM3h, which produces MRI signal changes in response to
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
long-chain fatty acids, and BM3h-8C8 and BM3h-B7, which respond to a catecholamine. Contrast agents engineered to detect dopamine and other signaling molecules in the brain will permit functional neuroimaging based on direct detection of neuronal events rather than hemodynamic changes. Exogenous delivery of macromolecules such as BM3h to large regions of animal brains should be possible using a variety of techniques33. Because BM3h is a protein, it might also be possible to deliver variants via expression from transfected cells in vivo or in transgenic subjects. Preliminary evidence that BM3h can be expressed to 1% protein content in mammalian cells supports the feasibility of this approach (Supplementary Results). Because of their small size, BM3h-based dopamine sensors might sample synaptic dopamine better than voltammetry or microdialysis probes, and with appropriate targeting could potentially become synapse specific. Dopamine sensor-dependent MRI would offer a combination of spatial coverage and precision inaccessible to other methods and uniquely suited to studies of dopaminergic function in systems neuroscience research. 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 thank V. Lelyveld for helpful discussions and assistance with in vitro measurements, N. Shah for help with MRI procedures and W. Schulze for help with automated analysis methods. We are grateful to C. Jennings and D. Cory for comments and suggestions about the manuscript, and to D. Vaughan for consultation regarding histology. We thank P. Caravan and again D. Cory for access to low-field relaxometers. M.G.S. thanks the Fannie and John Hertz Foundation and the Paul and Daisy Soros Fellowship for generous support. This work was funded by a Dana Foundation Brain & Immuno-Imaging grant, a Raymond & Beverley Sackler Fellowship and US National Institutes of Health (NIH) grants R01-DA28299 and DP2-OD2441 (New Innovator Award) to A.J., NIH grant R01-GM068664 and a grant from the Caltech Jacobs Institute for Molecular Medicine to F.H.A. and NIH grant R01-DE013023 to R.L. AUTHOR CONTRIBUTIONS M.G.S. conceived and performed the directed evolution and in vitro assessment of dopamine sensors; G.G.W. designed and conducted the in vivo experiments; P.A.R. performed directed evolution screening for BM3h variants; J.O.S. assisted with screening and in vitro experiments; B.K. assisted with data analysis for in vivo experiments; A.S. assisted with in vivo experiments; C.R.O. worked with M.G.S. to establish BM3h screening methods; R.L. provided consultation and essential materials; F.H.A. supervised the directed evolution work; A.J. established research direction, supervised the project overall and co-wrote the paper with M.G.S. and G.G.W. 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. Buxton, R.B. Introduction to Functional Magnetic Resonance Imaging: Principles and Techniques (Cambridge University Press, New York, 2001). 2. Ogawa, S., Lee, T.M., Kay, A.R. & Tank, D.W. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc. Natl. Acad. Sci. USA 87, 9868–9872 (1990). 3. Logothetis, N.K. What we can do and what we cannot do with fMRI. Nature 453, 869–878 (2008). 4. Jasanoff, A. MRI contrast agents for functional molecular imaging of brain activity. Curr. Opin. Neurobiol. 17, 593–600 (2007). 5. Bloom, J.D. et al. Evolving strategies for enzyme engineering. Curr. Opin. Struct. Biol. 15, 447–452 (2005). 6. Li, Q.S., Schwaneberg, U., Fischer, P. & Schmid, R.D. Directed evolution of the fatty-acid hydroxylase P450 BM-3 into an indole-hydroxylating catalyst. Chemistry (Easton) 6, 1531–1536 (2000).
269
© 2010 Nature America, Inc. All rights reserved.
Articles 7. Glieder, A., Farinas, E.T. & Arnold, F.H. Laboratory evolution of a soluble, selfsufficient, highly active alkane hydroxylase. Nat. Biotechnol. 20, 1135–1139 (2002). 8. Meinhold, P., Peters, M.W., Chen, M.M., Takahashi, K. & Arnold, F.H. Direct conversion of ethane to ethanol by engineered cytochrome P450 BM3. ChemBioChem 6, 1765–1768 (2005). 9. Otey, C.R., Bandara, G., Lalonde, J., Takahashi, K. & Arnold, F.H. Preparation of human metabolites of propranolol using laboratory-evolved bacterial cytochromes P450. Biotechnol. Bioeng. 93, 494–499 (2006). 10. Knutson, B. & Gibbs, S.E. Linking nucleus accumbens dopamine and blood oxygenation. Psychopharmacology (Berl.) 191, 813–822 (2007). 11. Schultz, W. Multiple dopamine functions at different time courses. Annu. Rev. Neurosci. 30, 259–288 (2007). 12. Hyman, S.E., Malenka, R.C. & Nestler, E.J. Neural mechanisms of addiction: the role of reward-related learning and memory. Annu. Rev. Neurosci. 29, 565–598 (2006). 13. Damier, P., Hirsch, E.C., Agid, Y. & Graybiel, A.M. The substantia nigra of the human brain. II. Patterns of loss of dopamine-containing neurons in Parkinson’s disease. Brain 122, 1437–1448 (1999). 14. Young, A.M., Joseph, M.H. & Gray, J.A. Increased dopamine release in vivo in nucleus accumbens and caudate nucleus of the rat during drinking: a microdialysis study. Neuroscience 48, 871–876 (1992). 15. Garris, P.A. & Wightman, R.M. Different kinetics govern dopaminergic transmission in the amygdala, prefrontal cortex, and striatum: an in vivo voltammetric study. J. Neurosci. 14, 442–450 (1994). 16. Gubernator, N.G. et al. Fluorescent false neurotransmitters visualize dopamine release from individual presynaptic terminals. Science 324, 1441–1444 (2009). 17. Lindsey, K.P. & Gatley, S.J. Applications of clinical dopamine imaging. Neuroimaging Clin. N. Am. 16, 553–573 (2006). 18. Ewing, A.G., Bigelow, J.C. & Wightman, R.M. Direct in vivo monitoring of dopamine released from two striatal compartments in the rat. Science 221, 169–171 (1983). 19. Michael, A.C., Ikeda, M. & Justice, J.B. Jr. Mechanisms contributing to the recovery of striatal releasable dopamine following MFB stimulation. Brain Res. 421, 325–335 (1987). 20. Duong, T.Q., Kim, D.S., Ugurbil, K. & Kim, S.G. Spatiotemporal dynamics of the BOLD fMRI signals: toward mapping submillimeter cortical columns using the early negative response. Magn. Reson. Med. 44, 231–242 (2000).
270
21. Munro, A.W. et al. P450 BM3: the very model of a modern flavocytochrome. Trends Biochem. Sci. 27, 250–257 (2002). 22. Macdonald, I.D., Munro, A.W. & Smith, W.E. Fatty acid-induced alteration of the porphyrin macrocycle of cytochrome P450 BM3. Biophys. J. 74, 3241–3249 (1998). 23. Ravichandran, K.G., Boddupalli, S.S., Hasermann, C.A., Peterson, J.A. & Deisenhofer, J. Crystal structure of hemoprotein domain of P450BM-3, a prototype for microsomal P450′s. Science 261, 731–736 (1993). 24. Modi, S. et al. NMR studies of substrate binding to cytochrome P450 BM3: comparisons to cytochrome P450 cam. Biochemistry 34, 8982–8988 (1995). 25. Lewis, D.F.V.. Guide to Cytochrome P450 Structure and Function (Taylor & Francis, New York, 2001). 26. Fasan, R., Chen, M.M., Crook, N.C. & Arnold, F.H. Engineered alkane-hydroxylating cytochrome P450(BM3) exhibiting nativelike catalytic properties. Angew. Chem. Int. Edn Engl. 46, 8414–8418 (2007). 27. Bloom, J.D., Labthavikul, S.T., Otey, C.R. & Arnold, F.H. Protein stability promotes evolvability. Proc. Natl. Acad. Sci. USA 103, 5869–5874 (2006). 28. Ohnuma, K., Hayashi, Y., Furue, M., Kaneko, K. & Asashima, M. Serum-free culture conditions for serial subculture of undifferentiated PC12 cells. J. Neurosci. Methods 151, 250–261 (2006). 29. Ewing, A.G., Wightman, R.M. & Dayton, M.A. In vivo voltammetry with electrodes that discriminate between dopamine and ascorbate. Brain Res. 249, 361–370 (1982). 30. Gerhardt, G.A., Rose, G.M. & Hoffer, B.J. Release of monoamines from striatum of rat and mouse evoked by local application of potassium: evaluation of a new in vivo electrochemical technique. J. Neurochem. 46, 842–850 (1986). 31. Chen, Z.J. et al. A realistic brain tissue phantom for intraparenchymal infusion studies. J. Neurosurg. 101, 314–322 (2004). 32. Michael, A.C., Ikeda, M. & Justice, J.B. Jr. Dynamics of the recovery of releasable dopamine following electrical stimulation of the medial forebrain bundle. Neurosci. Lett. 76, 81–86 (1987). 33. Vykhodtseva, N., McDannold, N. & Hynynen, K. Progress and problems in the application of focused ultrasound for blood-brain barrier disruption. Ultrasonics 48, 279–296 (2008). 34. Li, H. & Poulos, T.L. The structure of the cytochrome p450BM-3 haem domain complexed with the fatty acid substrate, palmitoleic acid. Nat. Struct. Biol. 4, 140–146 (1997).
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
ONLINE METHODS Animal care. We performed all experiments involving vertebrate animals with approval of the Massachusetts Institute of Technology Committee on Animal Care.
© 2010 Nature America, Inc. All rights reserved.
Library construction. We constructed BM3h mutant libraries in accordance with a previously published protocol7. The starting parent for evolution was the WT heme domain of BM3 with a C-terminal hexahistidine tag, housed in the pCWori vector35. We produced mutant libraries through error-prone PCR using the primers 5′-GAAACAGGATCCATCGATGCTTAGGAGGTCAT-3′ (forward) and 5′-GCTCATGTTTGACAGCTTATCATCG-3′ (reverse) and Taq polymerase (AmpliTaq, Applied Biosystems) with 25 µM MnCl2, producing ~1–2 mutations per gene. Between the fourth and fifth rounds of evolution, we introduced the mutation I366V into BM3h-8C8 via overlap extension PCR to improve protein thermostability27. Protein expression and high-throughput screening. We inoculated mutant colonies into deep-well 96-well plates containing 0.4 ml Luria broth (LB) medium and grew them overnight. On each plate, we included the parent clone and up to three previous parents in triplicate. We then transferred 0.1 ml of each culture to new plates containing 1.2 ml fresh terrific broth (TB) medium per well, supplemented with 100 µg ml−1 ampicillin, 0.2 mM isopropyl β-d1-thiogalactopyranoside (IPTG) and 0.5 mM δ-aminolevulinic acid (ALA). We stored remaining LB cultures with glycerol at −80 °C. After 20–30 h of protein expression at 30 °C, we pelleted cultures and lysed the pellets in 0.65 ml PBS containing 0.75 mg ml−1 hen egg lysozyme (Sigma-Aldrich) and 5 µg ml−1 DNase I (Sigma-Aldrich). We recorded absorbance spectra of 200 µl of cleared lysate from each well in a multiwell plate reader (Spectramax Plus, Molecular Devices) before and after addition of successively more concentrated dopamine or arachidonic acid. We analyzed the resulting absorbance spectra in Matlab (Mathworks) using a custom routine that calculated the absorbance difference spectra for each acquisition relative to ligand-free lysate, computed the difference between maximum and minimum of each difference spectrum, plotted each value as a function of ligand concentration and, for each well, fitted a non-ligand-depleting bimolecular association function to estimate the corresponding Kd. We subsequently compared mutant Kd values to those of the parents within each plate and chose eight to ten mutants showing the greatest decrease in Kd for dopamine and/or the greatest increase in Kd for arachidonic acid for bulk expression and analysis. Bulk expression and titrations. To produce selected proteins in bulk, we began by inoculating frozen LB cultures of candidate mutants into 30 ml TB medium containing 100 µg ml−1 ampicillin. We induced the cultures at log phase with 0.6 mM IPTG, supplemented them with 0.5 mM ALA and 50 µg ml−1 thiamine and shook them for an additional 20–25 h to express protein. We then lysed pelleted cells with BugBuster and Lysonase (EMD Chemicals) and purified BM3h mutants over Ni-NTA agarose (Qiagen). We exchanged buffer to PBS over PD-10 desalting columns (GE Healthcare), and measured protein concentration using a carbon monoxide binding assay36. To characterize ligand affinities of the purified variants, we titrated protein samples with dopamine, arachidonic acid, serotonin, norepinephrine, pyrocatechol, 3,4-dihydroxy-l-phenylalanine (DOPA), 3,4-dihydroxyphenylacetic acid (DOPAC), homovanillic acid (HVA), 3-methoxytyramine (3MT), acetylcholine, glutamate, glycine, GABA and epinephrine (all from Sigma Aldrich) and analyzed the results using Matlab as described above. We performed all measurements at room temperature (~21 °C). 3MT had a Kd of 73 ± 13 µM for BM3h-B7 and 183 ± 28 µM for BM3h-8C8. HVA and pyrocatechol showed no measurable affinity. In vitro magnetic resonance imaging. To assess magnetic relaxation behavior of the proteins, we arrayed BM3h samples (60–100 µl) into microtiter plates and placed them in a 40-cm-bore Bruker Avance 4.7 T MRI scanner, equipped with a 10-cm-inner-diameter birdcage resonator radiofrequency coil and 26 G cm−1 triple-axis gradients. We filled unused wells of the microtiter plates with PBS and performed imaging at ~21 °C on a 2-mm slice through the sample. We used a T1-weighted spin echo pulse sequence; echo time (TE) was 10 ms, and repetition times (TR) were 73, 116, 186, 298, 477, 763 ms, 1.221, 1.953,
doi:10.1038/nbt.1609
3.125 and 5.000 s. Data matrices consisted of 512 × 128 points, zero-filled to 1,024 × 512 points, where the second dimension corresponds to the phaseencoding direction; the field of view (FOV) was 16 × 8 cm. We reconstructed and analyzed images using custom routines running in Matlab and adjusted contrast to optimize MRI images presented in the figures. We calculated relaxation rates by exponential fitting to the image data, using an equation of the form I = k[1 – exp(−TR/T1)], where I was the observed MRI signal intensity and k was a constant of proportionality. We then determined values of r1 by linear fitting to a plot of R1 against protein concentration for six to eight BM3h concentrations in the range from 0 to 240 µM. We also performed low-field relaxivity measurements using benchtop spectrometers operating at 21 °C with proton resonance frequencies of 20 MHz and 60 MHz (Bruker Minispec NMS120 and mq60). Samples of 150 µL containing 50–100 µM BM3h-8C8 in PBS in the absence or presence of 500 µM dopamine yielded 20-MHz relaxivity measurements of 1.0 or 0.25 mM−1 s−1, respectively, and 60-MHz relaxivities of 1.1 or 0.23 mM−1 s−1, respectively. Dopamine release from PC12 cells. We grew PC12 cells in suspension in F-12K medium supplemented with 15% (vol/vol) horse serum and 2.5% (vol/vol) FBS (ATCC). In preparation for dopamine release experiments, we incubated 50-ml cell cultures for 1 h in medium supplemented with 1 mM dopamine and 1 mM ascorbic acid, pelleted the cells and washed them twice with Locke’s buffer (154 mM NaCl, 5.6 mM KCl, 3.6 mM NaHCO3, 2.3 mM CaCl2, 5.6 mM d-glucose and 5 mM HEPES pH 7.4). To Locke’s buffer missing 54 mM NaCl and containing or not containing the sensor, we added a 1:50 dilution of 2.7 M KCl (stimulus) or NaCl (control). We resuspended the washed PC12 cell pellets in 200 µl of either K+- or Na+-supplemented buffer, with or without sensor. After 30–60 min incubation at ~21 °C, we pelleted cells and imaged the supernatant in an MRI scanner as described above. We estimated dopamine release by calculating sensor saturation level from observed r1, then solving the quadratic equation describing bimolecular equilibrium binding with a known Kd, and assuming 32 µM of sensor for ligand concentration. We made independent measurements of dopamine release using the Dopamine EIA Kit (LDN). Brain injection of sensors with exogenous dopamine. For injection experiments testing the effect of exogenous dopamine on BM3h-8C8 and WT BM3h (Fig. 5a–c), we anesthetized adult male Lewis rats with 1–2% isoflurane. We stereotaxically inserted plastic guide cannulae (Plastics One) bilaterally into the striatum and secured them in place with dental cement (coordinates with respect to bregma: +0.7 mm anterior, 3 mm lateral, 6 mm below the surface of the skull). We connected tubing filled with silicone oil to an MRI-compatible dual channel syringe pump (Harvard Apparatus), attached it to internal cannulae and back-filled the cannulae with contrast agent solution, 500 µM BM3h-8C8 or WT BM3h, with or without equimolar dopamine (chemicals from Sigma-Aldrich). We ran the pump in infusion mode for a few seconds to ensure that no air entered the system and then lowered the internal cannulae (connected to the pump) into the implanted guide cannulae and fixed them in place with dental cement. After the implantation was complete, we transferred the animal to a plastic positioning device (Ekam Imaging) for imaging and placed it into a 4.7-T Bruker Avance scanner. We acquired fast spin echo (FSE) MRI scans (TE/TR 14/277 ms, 8.9 s per scan, 0.3 × 0.3 × 1.0 mm resolution, 3.8 × 3.8 cm FOV, data matrix 128 × 128) before infusion (seven scans) and continuously during and after bilateral infusion of paired solutions, each injected for 20 min at 0.5 µl min−1. We monitored heart rate continuously during the infusions using a Nonin Medical 8600V pulse oximeter equipped with a nonmagnetic sensor. We digitized raw oximetry readings using a National Instruments USB-6008 interface and converted them to heart rate using a Matlab code. Values were stable at around 350 beats per minute ± 40 (s.d.). We analyzed MRI data from these experiments using custom routines running in Matlab. We detrended image signal time courses, converted them to percent change with respect to the preinjection baseline and averaged them over striatal ROIs. ROIs were chosen to approximate the maximal volumes reliably filled with the contrast agent, and were defined by a five-voxel in-plane radius (2.7 mm diameter) around the cannula tips over three image slices centered rostrocaudally around the implantation position, excluding voxels with notable signal dropout due to the cannulae themselves. We produced data for group analyses by combining data from ROIs defined separately with respect
nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
to the injection cannulae tips in each individual. We computed maximal MRI signal changes and signal change maps by averaging the image intensity at the end of the injection period (scans 121–140) and subtracting and normalizing it to the pre-injection intensity (scans 1–7). Protein and dopamine quantification based on in vivo imaging data. We estimated absolute concentrations of contrast agent and dopamine under the assumptions that minimal endogenous dopamine was present, that the relaxivity and dopamine affinity of BM3h variants were the same in vivo and in vitro, and that the MRI acquisition procedure satisfied a strong T1-weighting requirement, where TR << T1. Under these assumptions, the fractional MRI signal change ∆I/I0 is approximately equal to ∆R1/R10, the fractional change in R1 (equal to 1/T1). R10 is the basal value of R1, measured as 0.55 ± 0.01 s−1 from curve fitting to multiple FSE images obtained with different TR values. We estimated the maximal total concentration of BM3h-8C8 by determining the corresponding ∆R1 averaged over multiple injections of BM3h-8C8 in the absence of dopamine and dividing it by the relaxivity of the unliganded protein. We determined total dopamine concentration from the value of ∆R1 observed during injection of BM3h-8C8 plus dopamine, the relaxivities of liganded and unliganded BM3h-8C8, the previously determined BM3h-8C8 concentration and the mass action relationships governing binding of the sensor to dopamine. In in vitro measurements, BM3h-8C8 had a T2 relaxivity of 4 mM−1 s−1; addition of 1 mM dopamine did not noticeably perturb this value significantly, suggesting that T2 effects in conjunction with appropriate imaging methods might be able to provide a basis for protein quantification similar to the approach we describe here. Histological analysis. After MRI contrast agent injection experiments using the paradigm described above, we placed rats under terminal anesthesia with ketamine and xylazine and transcardially perfused them with phosphate buffer containing heparin (Hospira) and then with 4% wt/vol paraformaldehyde (Sigma-Aldrich). We removed brains and obtained coronal cryosections of 10 µm thickness at 100-µm intervals across a range extending ~1 mm anterior and posterior to the injection cannula insertion site. We used standard protocols for hematoxylin and eosin staining. Terminal deoxynucleotidyl transferase–mediated dUTP nick-end labeling (TUNEL) staining was performed using the DeadEnd Colorimetric TUNEL system from Promega with visualization enhanced by the DAB Substrate Kit from Vector Laboratories. Histological procedures were implemented by Wax-it Histology Services. In vivo potassium stimulation experiments. For in vivo K+ stimulation experiments (Fig. 5d–g), we used isoflurane-anesthetized male Lewis rats. We fitted internal cannulae with Y-connectors and positioned them through bilateral guide cannulae at coordinates 0.8 mm anterior to bregma, 2.8 mm lateral to the midline and 7.8 mm below the skull surface. Each two-channel cannula delivered a given BM3h variant (BM3h-8C8 or WT, paired on opposite hemispheres of the brain; sides were randomized). On each two-channel injection
nature biotechnology
cannula, we loaded one arm with protein in standard artificial cerebrospinal fluid (aCSF, containing 150 mM Na+ and 3 mM K+) and the other arm with protein in high-K+ modified aCSF containing no Na+ and 153 mM K+. Two infusion pumps (Harvard Apparatus) drove the infusions; one pump controlled the standard aCSF (low-K+) infusions on both BM3h-8C8 and WT control sides, and the other pump controlled high-K+ infusions on both sides. We programmed the two pumps and synchronized them with the MRI experiment so as to acquire a preinfusion image baseline for 2 min, followed by continuous scanning over three stimulation cycles consisting of 10 min low K+ alternating with 5 min high K+, followed by a further 10 min of low K+, followed by up to 30 min of post-injection signal acquisition. During these experiments, we continuously recorded heart rate and found it to be 355 beats per minute ± 45 (s.d.); blood oxygen saturation levels were 94.1 ± 5.8%. We acquired T1-weighted multislice MRI scan series as for the dopamine injection experiments described above. We imported raw data into Matlab, processed them with spatial smoothing over nearest neighbors (in-plane) and converted them to percent signal change with respect to a fitted third-order poly nomial baseline. Scans from the initial protein-only injection period (<15 min) were excluded from the analysis. To statistically analyze data acquired during the three cycles of K+ stimulation, we used a procedure analogous to classical fMRI methods, by performing a t-test on intensity values associated with high and low K+ conditions. We considered voxels showing lower signal during the 5-minute intervals corresponding to K+ stimulation to be consistent with the expected effect of K+-evoked dopamine release on MRI signal in the presence of BM3h-8C8. We estimated that the delay between infusion pump switching and actual changes to K+ concentration in the brain was roughly 8–9 min. We derived this estimate by recording the time required for spreading of Trypan blue to a radius of 0.75 mm (comparable to ROIs used for most of the analyses presented) from an injection cannula embedded in 0.6% agarose, in a switched injection paradigm equivalent to the K+ stimulation paradigm applied in vivo. We chose a delay of 9 min for analyses presented in the text, but delays ranging from 7 to 12 min produced qualitatively similar statistical results, all with elevated numbers of voxels near the BM3h-8C8 infusion site showing the expected MRI signal decrease upon K+ stimulation and far fewer (if any) voxels near the WT BM3h control cannula showing significant (P < 0.01) effects (Supplementary Fig. 5). We performed ROI-wide computations on cylindrical regions of 1.5 or 2.7 mm diameter in-plane extending over three (1 mm thick) image slices, centered about the BM3h-8C8 and WT BM3h infusion cannula tips, excluding from the calculations voxels showing substantial signal dropout due to the cannulae themselves. We performed group analyses by combining data from ROIs defined separately with respect to injection cannulae in each animal, without further anatomical coregistration. 35. Barnes, H.J., Arlotto, M.P. & Waterman, M.R. Expression and enzymatic activity of recombinant cytochrome P450 17 alpha-hydroxylase in Escherichia coli. Proc. Natl. Acad. Sci. USA 88, 5597–5601 (1991). 36. Omura, T. & Sato, R. The carbon monoxide-binding pigment of liver microsomes. I. Evidence for its hemoprotein nature. J. Biol. Chem. 239, 2370–2378 (1964).
doi:10.1038/nbt.1609
letters
Rescue of the spinal muscular atrophy phenotype in a mouse model by early postnatal delivery of SMN
© 2010 Nature America, Inc. All rights reserved.
Kevin D Foust1, Xueyong Wang2,6, Vicki L McGovern3,6, Lyndsey Braun1, Adam K Bevan1,4, Amanda M Haidet1,4, Thanh T Le3, Pablo R Morales5, Mark M Rich2, Arthur H M Burghes3,4 & Brian K Kaspar1,3,4 Spinal muscular atrophy (SMA), the most common autosomal recessive neurodegenerative disease affecting children, results in impaired motor neuron function1. Despite knowledge of the pathogenic role of decreased survival motor neuron (SMN) protein levels, efforts to increase SMN have not resulted in a treatment for patients. We recently demonstrated that self-complementary adeno-associated virus 9 (scAAV9) can infect ~60% of motor neurons when injected intravenously into neonatal mice2–4. Here we use scAAV9-mediated postnatal day 1 vascular gene delivery to replace SMN in SMA pups and rescue motor function, neuromuscular physiology and life span. Treatment on postnatal day 5 results in partial correction, whereas postnatal day 10 treatment has little effect, suggesting a developmental period in which scAAV9 therapy has maximal benefit. Notably, we also show extensive scAAV9-mediated motor neuron transduction after injection into a newborn cynomolgus macaque. This demonstration that scAAV9 traverses the blood-brain barrier in a nonhuman primate emphasizes the clinical potential of scAAV9 gene therapy for SMA. Proximal SMA results in motor neuron death in the spinal cord. SMA is caused by loss of survival motor neuron gene 1 (SMN1) and retention of SMN2, resulting in reduced levels of SMN, a ubiquitously expressed protein important in the assembly of ribonucleoprotein complexes1,5–7. Neuronal expression of SMN appears essential8. Recent work using a double transgenic knockout mouse model of SMA showed that postnatal lentiviral-mediated delivery of SMN to motor neurons increased survival by 3–5 d in an animal that normally survives ~13 d9. Pharmacological approaches have increased survival up to ~40 d10,11. We and others recently demonstrated that intravenous injection of scAAV9 into 1-d-old (postnatal day 1, P1) mice and cats infects ~60% of motor neurons, indicating the potential of this approach in treating SMA2,12. Here, we report that scAAV9-mediated SMN gene replacement (with scAAV9SMN) in SMA mice results in an unprecedented improvement in survival and motor function13. We also show that scAAV9–green fluorescent protein (GFP) crosses the blood-brain barrier in a nonhuman primate and transduces motor neurons, supporting the possibility of translating this treatment option to human patients.
To determine transduction levels in SMA mice (SMN2+/+; SMN∆7+/+; Smn−/−), we injected 5 × 1011 genomes of scAAV9-GFP or scAAV9-SMN (n = 4/group) under control of the chicken-βactin hybrid promoter into the facial vein on P1. We found that 42 ± 2% of lumbar spinal motor neurons expressed GFP (Fig. 1a and Supplementary Table 1) 10 d after injection. The levels of SMN in the brain, spinal cord and muscle in scAAV9-SMN–treated animals are shown in Figure 1b. SMN levels were increased in brain, spinal cord and muscle in treated animals, but were still lower than controls (SMN2+/+; SMN∆7+/+; Smn+/–) in neural tissue (Supplementary Fig. 1). Spinal cord immunohistochemistry demonstrated expression of SMN within choline acetyl transferase (ChAT)-positive cells after scAAV9-SMN injection (Supplementary Fig. 2). We next evaluated whether scAAV9-SMN treatment of SMA animals improved motor function14. SMA animals treated with scAAV9-SMN or scAAV9-GFP on P1 were assessed for the ability to right themselves compared to control and untreated animals (n = 10/group). Control animals could right themselves quickly, whereas the SMN- and GFP-treated SMA animals showed difficulty at P5. However, by P13, 90% of SMN-treated animals could right themselves compared with 20% of GFP-treated controls and 0% of untreated SMA animals, suggesting that SMN-treated animals improved (Fig. 1c). At P18, SMNtreated animals were larger than GFP-treated animals but smaller than controls (Fig. 1d). Locomotive ability of the SMN-treated animals was nearly identical to that of controls as assayed by x, y and z plane beam breaks (open field testing) and wheel running (Supplementary Figs. 3 and 4 and Supplementary Movie). Age-matched untreated SMA animals were not available as controls for open field or running wheel analysis owing to their short life span. We next examined survival of SMN-treated SMA animals (n = 11) compared with GFP-treated SMA animals (n = 11). No GFP-treated control animals survived past P22, with a median life span of 15.5 d (Fig. 1e). We analyzed body weight in SMN- or GFP-treated animals compared to wild-type littermates. The GFP-treated animals’ weights peaked at P10 and then precipitously declined until death. In contrast, SMN-treated animals showed a steady weight gain to approximately P40, where the weight stabilized at 17 g, half the weight of controls (Fig. 1f). The smaller size of corrected animals is likely related to
1Center for Gene Therapy, The Research Institute at Nationwide Children’s Hospital, Columbus, Ohio, USA. 2Wright State University, Dayton, Ohio, USA. 3Department of Molecular and Cellular Biochemistry and 4Integrated Biomedical Graduate Program, The Ohio State University, Columbus, Ohio, USA. 5The Mannheimer Foundation, Inc., Homestead, Florida, USA. 6These authors contributed equally to this work. Correspondence should be addressed to B.K.K. ([email protected]).
Received 29 July 2009; accepted 28 January 2010; published online 28 February 2010; doi:10.1038/nbt.1610
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
271
letters a
Brain
b
Untreated SMN treated Control
SMN Actin Spinal cord Untreated SMN treated Control
SMN Actin Quadriceps Untreated SMN treated Control
c Righting time (sec)
Control Untreated
d
30 20
SMN treated
10 5
7
n = 11
9 11 Age (d)
13
f
n=6
80 60 40
GFP treated SMN treated
20 0
0
50
100 150 Time (d)
200
GFP treated
P18
Body weight (g)
0
e 100 Percent survival
© 2010 Nature America, Inc. All rights reserved.
GFP treated SMN treated
40
SMN GAPDH Lane 1 2 3 4 5 6 7 8 9
Figure 1 Phenotypic correction of SMA mice injected on P1. (a) Injection of scAAV9-GFP in SMA animals results in GFP expression (green) within dorsal root ganglia and motor neurons (ChAT staining in red) in the lumbar spinal cord 10-d post-injection. (b) Western blots from tissues of control, scAAV9-SMN–treated and untreated SMA animals show elevated levels of SMN expression in SMN-treated animals compared to control animals, although levels are still lower than those of control littermates. Quantifications of western blots are available in the Supplementary Figures 1 and 7. (c) Righting ability shows that SMN-treated animals can right themselves similarly to control animals by P13. (d) SMA animals treated with scAAV9-SMN are larger than GFP-treated animals. (e) scAAV9-SMN treatment of SMA animals results in greatly extended survival over GFP treatment. (f) Body weight assessments show an increase in animals treated with scAAV9-SMN versus those treated with GFP. Scale bars, 200 µm (a); 50 µm (a inset).
250
40
Control SMN treated
Control
GFP treated Untreated
30 20 10 0
1 10 20 30 40 50 60 70 80 90 Age (d)
the tropism and incomplete transduction of scAAV9, resulting in a ‘chimeric’ animal in which some cells are not transduced. Additionally, the smaller size suggests an embryonic role for SMN. Notably, no deaths occurred in the SMN-treated group until P97. Furthermore, this death appeared to be unrelated to SMA as the mouse died after trimming of long extensor teeth. We euthanized four animals (P90–99) for electrophysiology of neuromuscular junctions (NMJs). The remaining six animals were still alive as of resubmission in November 2009 and had surpassed 250 d of age. A recent report demonstrated that neuromuscular transmission is abnormal in SMA mice15. To determine whether the reduction in endplate currents (EPCs) was corrected with scAAV9-SMN, we recorded EPCs from the tibialis anterior (TA) muscle16. P9–P10 animals were evaluated to ensure the presence of the reported abnormalities. Control mice had an EPC amplitude of 19.1 ± 0.8 nA versus 6.4 ± 0.8 nA in untreated SMA animals (P = 0.001), confirming published results15. Notably, scAAV9-SMN–treated SMA animals had a significant improvement at P10 over age-matched untreated SMA animals (8.8 ± 0.8 versus 6.4 ± 0.8 nA, P < 0.05). However, gene therapy treatment had not restored normal EPC at P10 when comparing scAAV9-SMN– treated SMA animals with controls (19.1 ± 0.8 versus 8.8 ± 0.8 nA, P = 0.001). At P90–P99, there was no difference in EPC amplitude
between controls and SMA mice that had been treated with scAAVSMN (Fig. 2a). Thus, treatment with scAAV9-SMN fully corrected the reduction in synaptic current. P90–P99 age-matched untreated SMA animals were not available as controls owing to their short life span. The amplitude of EPCs is determined by the number of synaptic vesicles released after nerve stimulation (quantal content) and the amplitude of the muscle response to the transmitter released from a single vesicle (quantal amplitude). Untreated SMA mice have a reduction in EPC primarily because of reduced quantal content15. In our P9–P10 cohort, untreated SMA animals had a reduced quantal content compared with wild-type controls (5.7 ± 0.6 versus 12.8 ± 0.6, P < 0.05), but scAAV9-SMN–treated animals were again improved over the untreated animals (9.5 ± 0.6 versus 5.7 ± 0.6, P < 0.05), but not to the level of wild-type animals (9.5 ± 0.6 versus 12.8 ± 0.6, P < 0.05). At P90–P99, the quantal content in treated SMA mice was
a
Treated SMA
Control
Endplate current
20 nA 3 ms
b Miniature endplate current
0.5 nA 3 ms
c 50 Hz repetitive stimulation
20 nA 20 ms
Figure 2 Effects of SMN treatment at P1 on NMJs of adult SMA mice. Untreated SMA mice do not survive to adulthood. (a) scAAV9-SMN treatment restores endplate currents (EPC) in ~90-d-old SMA animals. In control mice, the mean EPC amplitude was 82.6 ± 3.5 nA, and in treated SMA mice, it was 83.4 ± 4.1 nA (P = 0.89, n = 4 mice for each group). (b) Affected animals treated with scAAV9-SMN had an increase in miniature endplate currents. (c) Both control and treated SMA endplate currents had a similar degree of depression during 50 Hz nerve stimulation. (d–i) Representative sections from the transverse abdominis (TVA), a proximal muscle with innervation abnormalities in SMA mice 2, shows normal innervation in both wild-type (d–f) and SMN-treated (g–i) animals. Scale bars, 10 µm.
272
d
NF
g
NF
e
btx
f
Merge
h
btx
i
Merge
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
a
b
c
d
e
f
g
h
i
15 scAAV9-SMN (P10) scAAV9-SMN (P5) scAAV9-SMN (P2) scAAV9-GFP (P1)
80 60 40 20 0
0
20 40 Age (d)
60
Weight (g)
100
Figure 3 Systemic injection of scAAV9-GFP into SMA mice of varying ages. (a–c) Animals injected on P2 have a transduction pattern identical to P1-injected animals, with motoneuron transduction in lumbar spinal cord. (d–f) P5-injected animals have more glial transduction and less motoneuron (f inset, arrow) transduction than younger animals in lumbar spinal cord analysis. (g–i) The pattern of increasing glial transduction continues in P10-injected animals. GFP (green), ChAT (red, a motoneuron marker) and merged (yellow). (j–k) scAAV9-SMN injection on P2 in SMA animals rescues life span and increases body weight (n = 6), whereas P5 scAAV9-SMN delivery in SMA animals only partially rescues life span and body weight (n = 4) compared with control scAAV9-GFP–treated (n = 10). No increase in life span or body weight was seen in mice treated with scAAV9-SMN on P10 (n = 4). (l–q) Systemic injection of scAAV9-GFP into a cynomolgus macaque on P1 results in a similar transduction pattern within the spinal cord as previously shown in P1-injected mice. GFP (l,o), ChAT (m,p) and merged (n,q) images from thoracic spinal cord demonstrate motor neuron transduction. A representative longitudinal section is shown in (l–n), indicating transduction along the neuraxis. Transverse sections (o–q) mimic the pattern of dorsal root ganglia and motor neuron transduction seen in P1-injected mice. Inset scale bars, 50 µm; c,f,i, 100 µm; n,q, 200 µm.
scAAV9-SMN (P10) scAAV9-SMN (P5) scAAV9-SMN (P2) scAAV9-GFP (P1)
k
j Percent survival
© 2010 Nature America, Inc. All rights reserved.
letters
10 5 0
1
10
l
m
n
o
p
q
20 Age (d)
30
slightly reduced (control = 61.3 ± 3.5; SMA-treated = 50.3 ± 2.6, P < 0.05) but was compensated for by a statistically significant increase in quantal amplitude (Fig. 2b; control = 1.39 ± 0.06; SMA-treated = 1.74 ± 0.08, P < 0.05). Quantal amplitudes in young animals had no significant differences (control = 1.6 ± 0.1, untreated SMA = 1.3 ± 0.1, treated SMA = 1.1 ± 0.1 nA, P = 0.28). The reduction in vesicle release in untreated SMA mice was due to a decrease in probability of vesicle release, demonstrated by increased facilitation of EPCs during repetitive stimulation15. Both control and treated SMA EPCs were reduced by close to 20% by the 10th pulse
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
of a 50 Hz train of stimuli (Fig. 2c, 22 ± 3% reduction in control versus 19 ± 1% reduction in treated SMA, P = 0.36). This suggests that the reduction in probability of release was corrected by replacement of SMN. During electrophysiologic recording, no evidence of denervation was noted. Furthermore, all adult NMJs analyzed showed normal morphology and full maturity (Fig. 2d–i). P9–P10 transverse abdominis immunohistochemistry showed the typical neurofilament accumulation in untreated SMA NMJs15,17–19, whereas treated SMA NMJs showed a marked reduction in neurofilament accumulation (Supplementary Fig. 5). A recent study using a histone deacetylase inhibitor to extend survival of SMA mice reported necrosis of the extremities and internal tissues20. In our study, mice developed necrotic pinna between P45–P70 (Supplementary Fig. 6). Pathological examination of the pinna revealed vascular necrosis, but necrosis was not found elsewhere. We previously demonstrated that vascular endothelium was among the cell types transduced after systemic scAAV9 delivery2. Lack of necrosis in the tail and hind-paws could be due to treatment of vascular tissue, whereas the development of the pinna after P1 precludes correction of this tissue owing to loss of recombinant vector genomes in dividing cells21–23. To explore the therapeutic window in SMA mice, we performed systemic scAAV9-GFP injections at varying postnatal time points to evaluate the pattern of transduction of motor neurons and astrocytes. scAAV9GFP systemic injections in mice on P2, P5 or P10 showed distinct differences in the spinal cord. There was a shift from neuronal transduction in P2-treated animals toward predominantly glial transduction in older, P10 animals, consistent with our previous studies and knowledge of the developing blood-brain barrier in mice (Fig. 3a–i)2,24. To determine the therapeutic effect of SMN delivery at these various time points, small cohorts of SMA-affected mice were injected with scAAV9-SMN on P2, P5 and P10 and evaluated for changes in survival and body weight (Fig. 3j–k). P2-injected animals were rescued and indistinguishable from animals injected with scAAV9-SMN on P1. However, P5-injected animals showed a more modest increase in survival of ~15 d, whereas P10-injected animals were indistinguishable from GFP-injected SMA pups. These findings support previous studies demonstrating the importance of increasing SMN levels in neurons of SMA mice8. Furthermore, these results suggest a finite period during development in which intravenous injection of scAAV9 can target neurons in sufficient numbers for benefit in SMA.
273
© 2010 Nature America, Inc. All rights reserved.
letters To assess the potential for clinical translation of this approach, we investigated whether scAAV9 can traverse the blood–brain barrier in nonhuman primates25. We intravenously injected a male cynomolgus macaque on P1 with 1 × 1014 particles (2.2 × 1011 particles/g of body weight) of scAAV9-GFP and euthanized it 25 d after injection. Examination of the spinal cord revealed robust GFP expression within the dorsal root ganglia and motor neurons along the entire neuraxis (Fig. 3l–q), as seen in P1-injected mice. This finding demonstrates that early systemic delivery of scAAV9 can efficiently target motor neurons in a nonhuman primate. In conclusion, we report here the most robust postnatal rescue of SMA mice to date, with correction of motor function, neuromuscular electrophysiology and survival after a one-time delivery of SMN. Intravenous scAAV9 treats neurons, muscle and vascular endothelium, all of which have been proposed as target cells for treatment2. Although this study did not attempt to dissect the roles of different cell types in SMA, our P10 data show that SMN replacement in astrocytes is not effective in delaying disease, consistent with previous results using transgenic approaches8. We have also defined a window of opportunity for targeting motor neurons in neonates. Future studies in nonhuman primates will further elucidate a therapeutic window more relevant to human therapy. Advances in vector design, such as AAV capsid modification, mutagenesis or gene shuffling, may expand the opportunity to target neurons in the adult26–28. Although SMA children are often asymptomatic at birth, newborn screening that can detect SMA has been developed, supporting the feasibility of delivering scAAV9-SMN to affected children29. Additionally, we have demonstrated widespread transduction within the spinal cord of a nonhuman primate species. We are continuing to advance this delivery system in nonhuman primates and evaluating immunological consequences to the SMN gene and AAV capsid in order to set the stage for human clinical trials of scAAV9SMN in SMA. Given that SMA is a disease of low versus no protein, we do not anticipate an immune response against the SMN transgene. Further, we expect gene delivery to newborn patients to occur prior to wild-type AAV infection, thereby lowering the chances of preexisting immunity to the AAV capsid. 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 This work was supported by NIH/NINDS R21NS064328 to B.K.K., NINDS R01NS038650 to A.H.M.B., NINDS core P30-NS045758, RC2 NS069476-01 and Miracles for Madison Fund to B.K.K. and A.H.M.B. and NINDS P01NS057228 to M.M.R. We thank R. Levine and E. Nurre for expert technical assistance and J. Ward for pathology services. AUTHOR CONTRIBUTIONS K.D.F., M.M.R., A.H.M.B. and B.K.K. designed and executed experiments and wrote the manuscript. V.L.M., X.W, L.B., A.M.H., A.K.B., P.R.M. and T.T.L. contributed to experiments. 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/.
274
1. Burghes, A.H. & Beattie, C.E. Spinal muscular atrophy: why do low levels of survival motor neuron protein make motor neurons sick? Nat. Rev. Neurosci. 10, 597–609 (2009). 2. Foust, K.D. et al. Intravascular AAV9 preferentially targets neonatal neurons and adult astrocytes. Nat. Biotechnol. 27, 59–65 (2009). 3. Gao, G. et al. Clades of adeno-associated viruses are widely disseminated in human tissues. J. Virol. 78, 6381–6388 (2004). 4. McCarty, D.M. et al. Adeno-associated virus terminal repeat (TR) mutant generates self-complementary vectors to overcome the rate-limiting step to transduction in vivo. Gene Ther. 10, 2112–2118 (2003). 5. Lefebvre, S. et al. Identification and characterization of a spinal muscular atrophydetermining gene. Cell 80, 155–165 (1995). 6. McGovern, V.L., Gavrilina, T.O., Beattie, C.E. & Burghes, A.H. Embryonic motor axon development in the severe SMA mouse. Hum. Mol. Genet. 17, 2900–2909 (2008). 7. MacKenzie, A.E. & Gendron, N.H. Tudor reign. Nat. Struct. Biol. 8, 13–15 (2001). 8. Gavrilina, T.O. et al. Neuronal SMN expression corrects spinal muscular atrophy in severe SMA mice while muscle-specific SMN expression has no phenotypic effect. Hum. Mol. Genet. 17, 1063–1075 (2008). 9. Azzouz, M. et al. Lentivector-mediated SMN replacement in a mouse model of spinal muscular atrophy. J. Clin. Invest. 114, 1726–1731 (2004). 10. Avila, A.M. et al. Trichostatin A increases SMN expression and survival in a mouse model of spinal muscular atrophy. J. Clin. Invest. 117, 659–671 (2007). 11. Hastings, M.L. et al. Tetracyclines that promote SMN2 exon 7 splicing as therapeutics for spinal muscular atrophy. Sci. Transl. Med 1, 5–14 (2009). 12. Duque, S. et al. Intravenous administration of self-complementary AAV9 enables transgene delivery to adult motor neurons. Mol. Ther. 17, 1187–1196 (2009). 13. Le, T.T. et al. SMNDelta7, the major product of the centromeric survival motor neuron (SMN2) gene, extends survival in mice with spinal muscular atrophy and associates with full-length SMN. Hum. Mol. Genet. 14, 845–857 (2005). 14. Butchbach, M.E., Edwards, J.D. & Burghes, A.H. Abnormal motor phenotype in the SMNDelta7 mouse model of spinal muscular atrophy. Neurobiol. Dis. 27, 207–219 (2007). 15. Kong, L. et al. Impaired synaptic vesicle release and immaturity of neuromuscular junctions in spinal muscular atrophy mice. J. Neurosci. 29, 842–851 (2009). 16. Wang, X. et al. Decreased synaptic activity shifts the calcium dependence of release at the mammalian neuromuscular junction in vivo. J. Neurosci. 24, 10687–10692 (2004). 17. Cifuentes-Diaz, C. et al. Neurofilament accumulation at the motor endplate and lack of axonal sprouting in a spinal muscular atrophy mouse model. Hum. Mol. Genet. 11, 1439–1447 (2002). 18. Kariya, S. et al. Reduced SMN protein impairs maturation of the neuromuscular junctions in mouse models of spinal muscular atrophy. Hum. Mol. Genet. 17, 2552–2569 (2008). 19. Murray, L.M. et al. Selective vulnerability of motor neurons and dissociation of pre- and post-synaptic pathology at the neuromuscular junction in mouse models of spinal muscular atrophy. Hum. Mol. Genet. 17, 949–962 (2008). 20. Narver, H.L. et al. Sustained improvement of spinal muscular atrophy mice treated with trichostatin A plus nutrition. Ann. Neurol. 64, 465–470 (2008). 21. Clark, K.R. et al. Gene transfer into the CNS using recombinant adeno-associated virus: analysis of vector DNA forms resulting in sustained expression. J. Drug Target. 7, 269–283 (1999). 22. Duan, D. et al. Circular intermediates of recombinant adeno-associated virus have defined structural characteristics responsible for long-term episomal persistence in muscle tissue. J. Virol. 72, 8568–8577 (1998). 23. Nakai, H. et al. Extrachromosomal recombinant adeno-associated virus vector genomes are primarily responsible for stable liver transduction in vivo. J. Virol. 75, 6969–6976 (2001). 24. Saunders, N.R., Joakim Ek, C. & Dziegielewska, K.M. The neonatal bloodbrain barrier is functionally effective, and immaturity does not explain differential targeting of AAV9. Nat. Biotechnol. 27, 804–805, author reply 805 (2009). 25. Kota, J. et al. Follistatin gene delivery enhances muscle growth and strength in nonhuman primates. Sci. Transl. Med 1, 6–15 (2009). 26. Koerber, J.T. et al. Molecular evolution of adeno-associated virus for enhanced glial gene delivery. Mol. Ther. 17, 2088–2095 (2009). 27. Maheshri, N., Koerber, J.T., Kaspar, B.K. & Schaffer, D.V. Directed evolution of adeno-associated virus yields enhanced gene delivery vectors. Nat. Biotechnol. 24, 198–204 (2006). 28. Asokan, A. et al. Reengineering a receptor footprint of adeno-associated virus enables selective and systemic gene transfer to muscle. Nat. Biotechnol. 28, 79–82 (2010). 29. Pyatt, R.E., Mihal, D.C. & Prior, T.W. Assessment of liquid microbead arrays for the screening of newborns for spinal muscular atrophy. Clin. Chem. 53, 1879–1885 (2007).
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
ONLINE METHODS
© 2010 Nature America, Inc. All rights reserved.
Review board approval. All animal procedures were approved by Nationwide Children’s Hospital Institutional Animal Care and Use Committee, Wright State Institutional Animal Care and Use Committee, The Mannheimer Foundation Animal Care and Use Committee and Ohio State University Animal Care and Use Committee. Animals. SMA parent mice (Smn+/−, SMN2+/+, SMN∆7+/+) were time mated30. Cages were monitored 18–21 d after visualization of a vaginal plug for the presence of litters. Once litters were delivered, the mother was separated from pups, pups were given tattoos for identification and tail samples were collected. Tail samples were incubated in lysis solution (25 mM NaOH, 0.2 mM EDTA) at 90 °C for 1 h. After incubation, tubes were placed on ice for 10 min then received an equal volume of neutralization solution (40 mM Tris pH5). After the neutralization buffer, the extracted genomic DNA was added to two different PCR reactions for the mouse Smn allele (Forward 1: 5′-TCCAGCTC CGGGATATTGGGATTG, Reverse 1: 5′-AGGTCCCACCACCTAAGAAAGCC, Forward 2: 5′-GTGTCTGGGCTGTAGGCATTGC, Reverse 2: 5′-GCTG TGCCTTTTGGCTTATCTG) and one reaction for the mouse Smn knockout allele (Forward: 5′-GCCTGCGATGTCGGTTTCTGTGAGG, Reverse: 5′-CCAGCGCGGATCGGTCAGACG). After analysis of the genotyping PCR, litters were culled to three animals. Affected animals (Smn−/−, SMN2+/+, SMN∆7+/+) were injected as previously described with 5 × 10 11 particles of scAAV9-SMN or scAAV9-GFP2. Ultrasound-guided intracardiac delivery of scAAV9. In older mice we used ultrasound-guided intracardiac injections to efficiently deliver the gene therapy vector. Animals were anesthetized using 1–2.5% isofluorane (in O2 gas) throughout the procedure. Animals were secured with tape to a heated platform and fitted with a nose cone. A Vevo 2100 ultrasound was used to visualize the animal’s heart and monitor vitals. For scAAV9-GFP injections into wild-type mice, P5 animals were injected with 7E+11 vector genomes (vg, in 70 µl), P10 animals with 3.5E+11vg (in 70 µl), into the left ventricle. Affected mice were similarly injected with scAAV9-SMN on P5 and P10 with 3E+11vg (in 60 µl). Upon successful vector delivery, the animals were monitored for signs of cardio-pulmonary distress, disconnected from the anesthetic machine and placed in an oxygen recovery chamber for a short period of time before being returned to its cage. Nonhuman primate subjects. The selected experimental subject was a 1-d-old cynomolgus macaque (Macaca fascicularis) born in December 2009, at The Mannheimer Foundation, an Association for Assessment and Accreditation of Laboratory Animal Care International-accredited facility. The dam and sire of this neonate were both part of the cynomolgus breeding colony at the Foundation, housed in an outdoor enclosure. All cynomolgus macaques at the Foundation are fed a commercial diet (Harlan/Teklad 2050), supplemented with seeds, fruits/produce and provided with fresh water ad libitum. All uses and procedures were approved and in accordance with the Institutional Animal Care and Use Committee (IACUC) of the Mannheimer Foundation. Both dam and newborn were negative for selected retroviruses (simian immunodeficiency virus, simian retrovirus type D and simian T-cell lymphotropic virus type 1) and B-virus (cercopithecine herpes virus 1). AAV9 administration to nonhuman primates. As soon as the newborn subject was identified and selected in the outdoor enclosure, it was brought into an indoor hospital room on the same birth-day and pair-housed along with its dam. Within 24 h of the birth and just before the AAV9 delivery both dam and newborn were sedated with ketamine HCL (at a dose of 10 mg/kg intramuscular) and briefly separated from each other to perform the initial procedures. Baseline blood samples were collected from both dam and newborn through femoral venipunctures. After blood samples were collected, the dam was placed back in its cage; the newborn was positioned on sternal recumbency, and one of its legs shaved/disinfected in preparation for the intravenous injection of the AAV9. A 24-gauge intravenous catheter was placed into the saphenous vein. Vector solution was drawn into a 12 cc syringe pre-wetted with saline solution (0.9% NaCl). The intravenous catheter was flushed and its patency verified by using ~1 cc of the same saline solution. A total volume of 10 cc of the AAV9
doi:10.1038/nbt.1610
was delivered intravenously (medium dose of 1-5E+12 vg/g) in a bolus fashion. After injection, the newborn was recovered in a controlled temperature isolette (set at 95 °F) and returned to its dam after full recovery was achieved. Perfusion-fixation procedures and organ collections. 25 d post-AAV9 injection dam and infant were accessed once more by sedation with ketamine HCL (at a dose of 10 mg/kg) and Telazol (at a dose of 5 mg/kg), respectively. A follow-up blood sample was collected from the dam; the infant was separated and taken to the necropsy room for its terminal collections and procedures. A 24-gauge intravenous catheter was placed into a saphenous vein to facilitate subsequent doses of the anesthetic drug and the delivery of the euthanasia solution. Once the infant was confirmed to be in a deep anesthetic plane, a blood sample was collected, and then both thoracic and abdominal cavities were exposed to proceed with the perfusion process. A 20-gauge needle was inserted intra-cardially into the left ventricle, the right auricle was sectioned and the perfusion started first with 0.9% NaCl and then with 4% paraformaldehyde solution. A total of ~1.5 liters of the solution were perfused by gravity flow. Upon completion of the perfusion, several sections of major organs were harvested and fixed by immersion using the same paraformaldehyde solution. Viral vector. AAV9 was produced by transient transfection procedures using a double-stranded AAV2-ITR–based CB-GFP vector, with a plasmid encoding Rep2Cap9 sequence, as previously described3 along with a adenoviral helper plasmid; pHelper (Stratagene) in HEK293 cells. Our serotype 9 sequence was verified by sequencing and identical to that previously described3. Virus was purified by two cesium chloride density gradient purification steps, dialyzed against PBS and formulated with 0.001% Pluronic-F68 to prevent virus aggregation and stored at 4 °C. All vector preparations were titered by quantitativePCR using Taq-Man technology. Purity of vectors was assessed by 4–12% SDS-acrylamide gel electrophoresis and silver staining (Invitrogen). Behavior. Pups were weighed daily and tested for righting reflex every other day from P5–P13. Pups were placed on their sides and time to right was recorded, with a maximum of 30 sec allowed14. Every 5 d between P15 and P30, animals were tested in an open field analysis (San Diego Instruments). Animals were given several minutes within the testing chamber before the beginning of testing, then activity was monitored for 5 min. Beam breaks were recorded in the x, y and z planes, averaged across groups at each time point, then graphed. Immunofluorescence. Whole mount tissue. Whole mount TVA muscle was blocked in 10% Tween-20 (Sigma), 4% goat serum (Sigma) and PBS for 30 min. Whole mount tissue was incubated with goat anti-mouse neurofilament 160, (1:500, Chemicon) in 10% Tween-20, 0.4% goat serum, PBS overnight and incubated with Alexa Fluor-488 anti-mouse secondary antibody (1:1,000, Molecular Probes) for 2 h, and Alexa Fluor594 alpha-bungarotoxin (1:1,000, Molecular Probes) for 30 min. Tissues were mounted in Vectashield (Vector Labs). Tissue sections. Whole mount TVA muscle were post-fixed in 4% paraformaldehyde then stained with chicken anti-mouse neurofilament heavy chain, (1:1000, EnCor Biotechnology) in 10% Tween-20, 0.4% goat serum, PBS for 2 h and incubated with Alexa Fluor 488 anti-chicken secondary antibody (1:1000, Molecular Probes) and AlexaFluor594 alpha-bungarotoxin (1:1,000) for 30 min. Tissue sections were mounted in Vectashield. Confocal microscopy. All images were captured with the Leica TCS_SL scanning confocal microscope system using an inverted Leica DMIRE2 microscope and PMT detectors. Images were captured at 25°C with the following objective: 63× HCX Plan Apo CS oil, NA = 1.4.0. A Z-Galvo stage was used to obtain z-series stacks of ~30 images each. Image acquisition, overlays, scale bars and measurements were produced with the Leica Confocal Software v2.61, and subsequent image processing was performed with Adobe Photoshop CS2. Cell counts. For both GFP and Gem quantifications, spinal cords from fixed animals were removed from the carcass. Lumbar enlargements were blocked
nature biotechnology
and sliced into 40-µm thick sections on a vibratome (Leica). Sections were collected in order in a 96-well plate. Every 12th section was sampled for a total of eight sections spanning ~3.8 mm. The sections were fluorescently labeled using immunohistochemistry against choline acetyl transferase, green fluorescent protein or survival motor neuron. Sections were counted using the 63× objective on a confocal microscope. Greater detail on the extent of central nervous system transduction by scAAV9 following P1 injection is shown in previous work30.
© 2010 Nature America, Inc. All rights reserved.
Western blot analysis. 100 mg of tissue was homogenized in Tissue Protein Extraction Reagent (Pierce). The sample was mixed with an equal volume of loading buffer (62.5 mM Tris, pH 6.8, 20% glycerol, 200 mM DTT, 0.2% bromophenol blue) and run on a 12.5% polyacrylamide gel. Samples were transferred to Immobilon-P (Millipore). The blot was blocked in 5% milk powder, 0.5% BSA in PBS-Tween for 1 h, and then incubated for 1 h with a primary antibody cocktail of MANSMA 2, 7, 13 and 19. Bound primary antibody was detected by horseradish peroxidase conjugated secondary antibody followed by chemiluminescence (ECL Western Blotting Detection Reagents, Amersham Biosciences). The blots were then stripped and reprobed with a β-actin monoclonal antibody (clone AC-15, Sigma-Aldrich) or a GAPDH monoclonal antibody (Millipore) to control for protein loading. Electrophysiology. The recording chamber was continuously perfused with Ringer’s solution containing the following (in mmol/l): 118 NaCl, 3.5 KCl,
nature biotechnology
2 CaCl2, 0.7 MgSO4, 26.2 NaHCO3, 1.7 NaH2PO4 and 5.5 glucose, pH 7.3–7.4 (20–22 °C, equilibrated with 95% O2 and 5% CO2). Endplate recordings were performed as follows. After dissection, the tibialis anterior muscle was partially bisected and folded apart to flatten the muscle. After pinning, muscle strips were stained with 10 µM 4-Di-2ASP [4-(4-diethylaminostyryl)-N-methyl pyridinium iodide] (Molecular Probes) and imaged with an upright epi fluorescence microscope. At this concentration, 4-Di-2ASP staining enabled visualization of surface nerve terminals as well as individual surface muscle fibers. All of the endplates were imaged and impaled within 100 µm. We used two-electrode voltage clamp to measure endplate current (EPC) and miniature EPC (MEPC) amplitude. Muscle fibers were crushed away from the endplate band and voltage clamped to −45 mV to avoid movement after nerve stimulation. Statistics. Statistical analyses were performed using Graph Pad Prizm software. Means were represented with s.e.m. Student t-tests were performed to compare groups using a 95% confidence level. Kaplan Meier Survival analysis was performed.
30. Monani, U.R. et al. The human centromeric survival motor neuron gene (SMN2) rescues embryonic lethality in Smn−/− mice and results in a mouse with spinal muscular atrophy. Hum. Mol. Genet. 9, 333–339 (2000).
doi:10.1038/nbt.1610
letters
Induction of cell cycle entry eliminates human leukemia stem cells in a mouse model of AML
© 2010 Nature America, Inc. All rights reserved.
Yoriko Saito1, Naoyuki Uchida2, Satoshi Tanaka3, Nahoko Suzuki1, Mariko Tomizawa-Murasawa1, Akiko Sone1, Yuho Najima1, Shinsuke Takagi1,2, Yuki Aoki1, Atsushi Wake2, Shuichi Taniguchi2, Leonard D Shultz4 & Fumihiko Ishikawa1 Cancer stem cells have been proposed to be important for initiation, maintenance and recurrence of various malignancies, including acute myeloid leukemia (AML)1–3. We have previously reported4 that CD34+CD38− human primary AML stem cells residing in the endosteal region of the bone marrow are relatively chemotherapy resistant. Using a NOD/SCID/IL2rnull mouse model of human AML, we now show that the AML stem cells in the endosteal region are cell cycle quiescent and that these stem cells can be induced to enter the cell cycle by treatment with granulocyte colony-stimulating factor (G-CSF). In combination with cell cycle-dependent chemotherapy, G-CSF treatment significantly enhances induction of apoptosis and elimination of human primary AML stem cells in vivo. The combination therapy leads to significantly increased survival of secondary recipients after transplantation of leukemia cells compared with chemotherapy alone. Despite increasing understanding of the pathogenesis and biology of AML, patient outcomes remain poor, with a median survival of ~1 year with standard treatment due to high rates of disease relapse5. Recent studies have demonstrated that leukemia stem cells (LSCs) play a central role in AML pathogenesis, suggesting that failure to eradicate these cells is an important factor in patient outcomes6. Resembling normal hematopoietic stem cells (HSCs) in their ability to engraft and to produce progeny and self-renew continuously in vivo7, AML stem cells home to the endosteal region of the bone marrow, where they are resistant to chemotherapy. In addition, an enrichment of AML stem cells in the G0 phase of cell cycle suggests cell cycle quiescence as a mechanism of their chemotherapy resistance. Therefore, we aimed to determine the role of cell cycle status in chemotherapy resistance of AML stem cells and to develop a therapeutic strategy against LSCs by functionally modifying them in vivo. We generated mouse models of human AML by engrafting newborn nonobese diabetic/severe combined immunodeficient/interleukin (NOD/SCID/IL) 2rγnull mice with purified human (h)CD34+CD38− LSCs from seven AML patients. The mean peripheral blood and bone marrow hCD45+ cell engraftment levels of recipients were
77.6 ± 5.9% and 95.0 ± 1.8% (mean ± s.e.m., n = 48), respectively, indicating that the recipient bone marrow was nearly completely replaced by human AML cells. Flow cytometry plots of bone marrow from representative recipients engrafted with cells from each case of AML are shown in Figure 1a (cases 3 and 5) and Supplementary Figure 1a (cases 1, 2, 4, 6 and 7). All hCD45+ cells in the recipient peripheral blood and bone marrow were also hCD33 +, indicating that all CD45+ human hematopoietic cells in these recipients are human AML cells (Supplementary Fig. 1b). The injection of purified hCD34+CD38− cells resulted in repopulation of recipient bone marrow with hCD34+hCD38−, hCD34+CD38+ and hCD34− cells (Fig. 1a and Supplementary Fig. 1c). Successful secondary engraftment (Supplementary Fig. 1d) demonstrated the self-renewal capacity of hCD34+hCD38− cells, thereby satisfying the criteria for malignant stem cells (long-term engraftment capacity, capacity to develop leukemia in vivo, the generation of non-stem AML cells and self-renewal capacity). Although there was case-dependent variability, the majority of recipient bone marrow LSCs was in G0 and G0/G1 phases of the cell cycle. The frequency of cells in G0 and G0/G1 was significantly higher in the hCD34+CD38− than in the hCD34+CD38+ population (Supplementary Table 1). Next we examined the relationship between LSC cell cycle status and the cytotoxic effect of the chemotherapeutic agent cytosine arabinoside (Ara-C), a key chemotherapeutic agent used both in remission induction and post-remission therapy for AML patients. Mean pretreatment peripheral blood engraftment level as measured by percent hCD45+ cells was similar between Ara-C–treated and control AML-engrafted recipients (74.9 ± 4.2% and 67.6 ± 7.1%, respectively, n = 15 for each group, P = 0.4041 by two-tailed t-test). Representative cell cycle analyses after BrdU incorporation are shown in Supplementary Figure 1e. When the AML-engrafted mice were given Ara-C, bone marrow CD34 +CD38− AML cells in the S-phase of the cell cycle were preferentially eliminated (S = 0.1 ± 0.1% in Ara-C–treated group and S = 10.6 ± 0.9% in control group, respectively, n = 15 for each group, P < 0.0001 by two-tailed t-test). This was accompanied by the enrichment of quiescent LSCs in the G0/G1 phases of the cell cycle (G0/G1 = 91.7 ± 2.3% in the Ara-C–treated group and G0/G1 = 80.9 ± 1.5% in controls, n = 15 for each group,
1Research Unit for Human Disease Models, RIKEN Research Center for Allergy and Immunology, Yokohama, Japan. 2Department of Hematology, Toranomon Hospital, Tokyo, Japan. 3Nippon Becton Dickinson Company, Tokyo, Japan. 4The Jackson Laboratory, Bar Harbor, Maine, USA. Correspondence should be addressed to F.I. ([email protected]).
Received 29 July 2009; accepted 14 January 2010; published online 14 February 2010; doi:10.1038/nbt.1607
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
275
7AAD(–)MNCs
CD34 CD38
LSC:0.3%
Case 5 (post-cytokine) G0:12.4%
hCD45
hCD38
Pyronin Y hCD45
99.4%
+
+
–
CD34 CD38
LSC:1.2%
G0:32.0%
hCD45
Hoechst
Hoechst
hCD38
Pyronin Y
hCD45
96.1%
–
–
G0:77.8%
hCD34
7AAD(–)MNCs
+
+
CD34 CD38
hCD45 LSC:1.1%
99.5%
hCD45
Hoechst
+
+
7AAD(–)MNCs
hCD34
Case 5 (baseline)
Relative cell number
Pyronin Y hCD38
hCD34
Relative cell number
Case 3 (post-cytokine)
CD34+CD38–
G0:49.9%
hCD45
b
+
Relative cell number
hCD45 LSC:0.8%
99.2%
Pyronin Y
Case 3 (baseline)
7AAD(–)MNCs
hCD34
a
Relative cell number
letters
hCD38
Hoechst
Percent G0 cells within LSCs
© 2010 Nature America, Inc. All rights reserved.
c
Baseline Figure 1 Quiescent LSCs enter cell cycle after in vivo G-CSF 100 Post-cytokine treatment. (a,b) Representative contour plots of hCD34+CD38− LSCs in bone marrow of recipients engrafted with LSCs isolated from case 3 80 and 5. In a and b, histograms demonstrate nearly complete replacement of the recipient bone marrow with human AML cells (left panels). The 60 phenotypic isolation of hCD34+CD38− LSCs is shown in the middle panels. At baseline, the majority of bone marrow LSCs are quiescent 40 G0 cells, as demonstrated by Hoechst 33342-low PyroninY-negative phenotype (a). After in vivo G-CSF treatment, LSCs enter cell cycle 20 as indicated by increase in PyroninY-positive cells (indicating RNA synthesis), Hoechst 33342-high cells (indicating DNA synthesis), resulting in decreased frequency of cells in G0 phase (b). (c) Recipient Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 Case 7 bone marrow LSCs in G0 phase decreased with G-CSF treatment (open circles) compared with control recipients (filled circles). Horizontal bars indicate means + s.e.m. Number of control and experimental (G-CSF) recipients, respectively, from AML case 1 (n = 7, n = 6, P = 0.0010); case 2 (n = 9, n = 6, P < 0.0001); case 3 (n = 10, n = 5, P < 0.0001); case 4 (n = 6, n = 4, P = 0.0016); case 5 (n = 5, n = 6, P = 0.0043); case 6 (n = 5, n = 5, P = 0.0050); case 7 (n = 3, n = 5, P = 0.0004); each by two-tailed t-test.
P = 0.0011 by two-tailed t-test). These findings demonstrate that within hCD34+CD38− LSC population, cell cycle quiescence is associated with Ara-C resistance. Therefore, we hypothesized that entry of quiescent LSCs into the cell cycle would increase their susceptibility to chemotherapeutic agents. To test this hypothesis, we examined whether granulocyte colonystimulating factor (G-CSF) induces human LSCs to enter the cell cycle in vivo. Although induction of human and mouse normal HSC cell cycle entry by G-CSF has been described, the effect of G-CSF on LSCs has not been investigated8,9. Hoechst/PyroninY (Fig. 1a,b) and BrdU incorporation (Supplementary Fig. 1e) assays were used to measure the induction of the cell cycle in LSCs. In all seven primary AML cases, the LSCs in the bone marrow of AML-engrafted recipients treated with 300 µg/kg G-CSF daily for 5 d showed a significant (P values in legend to Fig. 1) reduction in the fraction of cells in G0 phase with a conco mitant increase of LSCs in S and G2/M phases (Fig. 1c). We have previously demonstrated that chemotherapy-resistant CD34+CD38− LSCs are enriched within the bone marrow endosteal region, whereas CD38+ AML cells resided mainly in the central region of the bone marrow4. Therefore, we performed Ki67 labeling in primary human AML-engrafted recipient bone sections to directly examine the cell cycle status of LSCs within the bone marrow endosteal region, adjacent to the bone endosteum lined with osteopontin-positive osteoblasts (Supplementary Fig. 2). Ki67 is a nuclear protein expressed in proliferating cells during late G1, S, G2 and M phases but not in quiescent cells in the G0 phase of the cell cycle. Figure 2 shows bone sections from a representative AML-engrafted recipient at steady state and after G-CSF treatment. In an untreated recipient, recipient (bone marrow human AML chimerism of 98.5%) human leukemic cells in the central zone of the bone marrow were strongly Ki67+, indicating that they are highly proliferative, whereas
276
the majority of AML cells abutting the endosteum were in G0 phase as evidenced by the negative Ki67 labeling (Fig. 2a,b). AML cells in the perivascular region— one of the physiological niches for normal murine HSCs10,11—were actively cycling as demonstrated by Ki67+ staining (Fig. 2d,e). Z-stack images showing vertical slices of the bone section further confirmed the relative paucity of Ki67+ cells in the region adjacent to the endosteum (Fig. 2c,f). This endosteal enrichment of G0 AML cells combined with our previous observation that chemotherapy-resistant LSCs preferentially localize within the bone marrow endosteal region suggest that AML relapse may be attributable to cell cycle quiescence within the bone marrow niche. We then examined whether in vivo cytokine treatment induces quiescent LSCs within the endosteal niche to enter the cell cycle. Figure 2g–i and Supplementary Figure 3a–c show bone sections from a representative human AML-engrafted recipient treated with G-CSF (bone marrow human AML engraftment of 97.8%). After 5 d of G-CSF administration, human CD45+ AML cells in the endosteal region became Ki67+, indicating their cell cycle entry (Fig. 2g–i). The central zone of the bone marrow from the same recipient also had numerous Ki67+CD45+ human AML cells (Supplementary Fig. 3a–c). Three-dimensional reconstruction of the bone section clearly demonstrated cell cycle entry by human CD45+ AML cells within the endosteal region after cytokine treatment (Fig. 2j,k and Supplementary Movies 1 and 2). A great majority of bone marrow cells in the AML-engrafted recipients examined were hCD45+CD33+AML cells (93.2 ± 1.3% for G-CSF-treated group, n = 37; 96.2 ± 0.6% for control group, n = 45). Co-labeling for human CD34 and Ki67 demonstrated that human CD34+ AML cells in the endosteal region enter the cell cycle in response to in vivo cytokine treatment (Supplementary Fig. 3d,e,g,h). The results of BrdU staining were consistent with Ki67 labeling, showing cell cycle induction of human AML cells within the endosteal region in cytokine-treated recipients
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
letters (Supplementary Fig. 3f,i). Quantification of human CD45+Ki67+ cells within the endosteal region demonstrated a statistically significant increase (P < 0.0001) in cycling LSCs after in vivo cytokine treatment (Supplementary Table 2). Taken together, these findings indicate that in vivo administration of G-CSF induces quiescent LSCs residing within the endosteal niche to enter the cell cycle. Next, to demonstrate that cell cycle entry increases the susceptibility of human LSCs to chemotherapy in vivo, we developed an in vivo treatment model using AML-engrafted recipients. The recipients received either Ara-C 1 g/kg daily for 2 d or G-CSF 300 µg/kg daily for 5 d combined with Ara-C 1g/kg daily on days 4 and 5. After chemotherapy (Ara-C) alone or chemotherapy following cell cycle induction by G-CSF, the recipient bone marrow was evaluated through (i) flow cytometric assessment of LSC apoptosis as evidenced by active caspase 3 expression, (ii) histological examination of bone marrow in situ to directly examine LSC apoptosis within the bone marrow niche by TUNEL staining and (iii) functional determination of the frequency and AML-initiating capacity of the remaining LSCs by limiting-dilution serial transplantation. Figure 2 Quiescent human AML cells within the bone marrow endosteal region enter the cell cycle after in vivo G-CSF treatment. (a–f) Immunofluorescence labeling for human CD45 and Ki67 in the peripheral (a–c) and the central zones (d–f) of bone marrow from a primary humal AML cell recipient with bone marrow human AML chimerism of 98.5%. The peripheral zone contains both the endosteal and perivascular regions and the central zone contains the perivascular region, framed by white rectangles on the images. (a–c) CD45+ human AML cells in the endosteal region are largely Ki67− (cell cycle quiescent), whereas in the perivascular region, CD45+ human AML cells are Ki67+ (cycling). Serial vertical sectional images through the bone section confirm that the endosteal region is nearly devoid of Ki67+ nuclei. (d–f) In contrast, the central zone of the bone marrow, including the perivascular region, is enriched for Ki67+ cycling human CD45+ AML cells, again confirmed by serial vertical sectional images. (g–i) Immunofluorescence labeling for human CD45 and Ki67 in the peripheral zone of bone marrow from a recipient with bone marrow human AML chimerism of 97.8%. The peripheral zone contains both the endosteal and perivascular regions, framed by white rectangles on the images. CD45+ human AML cells in the endosteal region enter cell cycle after in vivo cytokine treatment, as demonstrated by their Ki67 expression. Serial vertical sectional images through the bone section confirm that nuclei of human CD45+ AML cells are adjacent to the endosteum express Ki67. (j,k) Three-dimensional reconstruction of recipient bone sections was performed using serial sectional images. At steady state, the AML cells adjacent to the bone marrow endosteum are cell cycle quiescent (Ki67−) (j). After in vivo G-CSF treatment, AML cells within the bone marrow endosteal region became Ki67+, indicating that they had entered the cell cycle. CD45 (red), Ki67 (green), DAPI (blue) and merged images are shown (k). (a,d,g) 20× magnification, scale bars, 50 µm. (d,e,h) 40× magnification, scale bars, 20 µm.
Mean pretreatment human AML engraftment as measured by percent of peripheral blood hCD45+ cells was similar between groups receiving chemotherapy alone and chemotherapy with cell cycle induction (31.6 ± 4.1% with a range of 10.6–67.5% and 28.5 ± 4.3% with a range of 9.9–78.4%, respectively, P = 0.6031 by two-tailed t-test). The degree of apoptosis in LSCs was quantified by flow cytometric measurement of active caspase 3 expression within bone marrow CD34+CD38− cells (Fig. 3a and Supplementary Fig. 4). Chemotherapy with cell cycle induction significantly (P values in legend to Fig. 3) reduced the frequency of viable nonapoptotic hCD34+CD38− LSCs in recipients engrafted with AML originating from each of the seven primary AML cases examined. Although this effect is variable among the seven AML cases reflecting biological heterogeneity, it is statistically significant in each case. In situ examination of the bone marrow demonstrated that in recipients treated with chemotherapy alone, TUNEL-negative viable AML cells remained at the endosteum (Fig. 3b, upper panels). In contrast, chemotherapy with cell cycle induction resulted in increased apop tosis in the endosteal region (as well as the central region), as evidenced Low magnification DAPI Ki67
a
b
High magnification DAPI Ki67 Endosteal
c
Z-stack images DAPI CD45 Ki67
f
DAPI
i
Z-stack images DAPI CD45 Ki67
Perivascular
Peripheral zone
Steady state
CD45
Merge
Low magnification
d
DAPI
Ki67
CD45
e
Merge
High magnification
DAPI
Ki67
Z-stack images
CD45
Ki67
Central zone
Perivascular
CD45
g
Merge
Low magnification DAPI Ki67
CD45
h
Merge
High magnification DAPI Ki67 Endosteal
Peripheral zone
Post-cytokine
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
j
CD45
Merge
CD45
k
Steady state
DAPI Ki67 CD45
Merge Post-cytokine
DAPI Ki67 CD45
Endosteal Bone
Endosteal Bone
277
letters a
b
60
*
*
*
*
*
*
* *
TUNEL staining (Ara-C post cytokine)
20
Case 2
Case 3
Case 4
Case 5
Cell dose: 2 × 102
+ + + + + + + +
Case 7
Cell dose: 2 × 103
100
100
100 Percent engrafted cells
Percent engrafted cells
Case 6
80 60 40 20
5/18 6 9 12 15 18 21 24 18/18 Weeks post-transplantation No engraftment at 24 weeks
80 60 40 20
Ara-C post-cytokine
60% 40%
Ara-C alone
20%
4
8 12 16 20 Weeks post-transplantation
24
2 2 × 10 (n = 18) 2 × 103 (n = 55) 2 × 104 (n = 38) 2 × 105 (n = 14)
100
80 60 40 20
e
Cell dose: 2 × 105
80 60 40 20
Relapse
Chemotherapy
Dead AML cells
Resistant AML blasts
f
Chemotherapy
Bone Putative endosteal niche
+ + + + + +
0/14 6 9 12 15 18 21 24 5/15 Weeks post-transplantation No engraftment at 24 weeks
No engraftment at 24 weeks
2 × 102 (n = 18) 2 × 103 (n = 45) 2 × 104 (n = 31) 5 2 × 10 (n = 15)
80%
Cell dose: 2 × 104
0/38 6 9 12 15 18 21 24 16/31 Weeks post-transplantation
4/55 6 9 12 15 18 21 24 33/45 Weeks post-transplantation No engraftment at 24 weeks
d 100%
+
Percent engrafted cells
c
Percent survival
* * * * * * *
HE staining (Ara-C post cytokine)
40
Percent engrafted cells
Percent caspase-3– cells within LSCs
80
Case 1
© 2010 Nature America, Inc. All rights reserved.
TUNEL staining (Ara-C alone)
HE staining (Ara-C alone)
100
Relapse
Cytokine
Chemotherapy Quiescent LSCs Cycling AML cells
Cycling LSCs
Figure 3 Cell cycle entry potentiates chemotherapy-induced apoptosis of LSCs in vivo, reduces LSC frequency and leads to superior survival of secondary recipients. (a) Reduced LSC survival as measured by percent active caspase-3 − cells in bone marrow hCD34+CD38− LSCs from recipients after chemotherapy with cell cycle induction (open circles) compared with chemotherapy alone (filled circles). Horizontal bars indicate means + s.e.m. Number of control and experimental (G-CSF) recipients, respectively, from AML case 1. Number of recipients receiving chemotherapy alone or chemotherapy with cell cycle induction, respectively, from AML case 1 (n = 3 each, P = 0.0127); case 2 (n = 4 each, P = 0.0006); case 3 (n = 7, n = 4, P = 0.0001); case 4 (n = 4 each, P < 0.0001); case 5 (n = 4 each, P = 0.0005); case 6 (n = 5 each, P < 0.0001); case 7 (n = 5 each, P = 0.0003); each by two-tailed t-test. (b) HE and TUNEL staining of bone sections from recipients after chemotherapy alone reveal apoptosis in the central region of the bone marrow whereas cells abutting the endosteum remain viable (*). In contrast, bone marrow of recipient following chemotherapy with cell cycle induction shows apoptosis in the endosteal region (+) in addition to the central zone. Scale bars, 10 µm. (c) Peripheral blood hCD45 + AML cell engraftment time course of serial transplant recipients at graft doses of 2 × 10 2, 2 × 103, 2 × 104 and 2 × 105 hCD34+ AML cells, from 6–24 weeks post-transplantation. The findings are summarized in Supplementary Table 4. Blue and red symbols and lines, respectively, indicate recipients of bone marrow hCD34 + cells from Ara-C alone- and Ara-C post-cytokine–treated AML-engrafted mice. Case 1, [ l]; case 2, [n]; case 3, [s]; case 4, [t]; case 5, [©]; case 6, [l]; case 7, [n]. (d) From AML-engrafted mice receiving either chemotherapy alone or chemotherapy with cell cycle induction, bone marrow hCD34 + cells were harvested and retransplanted at doses of 2 × 10 2, 2 × 103, 2 × 104 and 2 × 105 into new recipients that were followed for 24 weeks. Recipients of bone marrow hCD34 + cells from AML-engrafted mice treated with chemotherapy with cell cycle induction (red) demonstrated higher overall survival compared with recipients of mouse bone marrow hCD34 + cells (blue) treated with chemotherapy alone, as estimated by the Kaplan-Meier method (P < 0.0001 for comparison of recipients within a given graft dose and for all recipients combined). (e,f) Central role of cell cycle-quiescent human primary AML LSCs in AML relapse and induction of cell cycle entry as a therapeutic strategy targeting LSCs. (e) In the past, disease relapse was attributed to the inability to completely eliminate AML cells by chemotherapy. (f) The current model maintains that chemotherapy-resistant LSCs that survive and repopulate the bone marrow prompt AML relapse. One mechanism through which LSCs successfully evade chemotherapy is their cell cycle quiescence within the bone marrow endosteal niche. Induction of cell cycle entry sensitizes these LSCs to chemotherapy, leading to reduced potential for relapse.
278
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
letters by decreased cellularity by hematoxylin and eosin (HE) staining and positive TUNEL staining of the residual cells (Fig. 3b, lower panels). After treatment, a large majority of the bone marrow cells was hCD45+CD33+ (95.9 ± 0.8% for chemotherapy alone, n = 32; 91.4 ± 1.3% for chemotherapy with cell cycle induction, n = 29). Therefore, the majority of the cells present in the histological sections shown are human AML cells, not residual mouse hematopoietic cells. Taken together, these findings indicate that chemotherapy-induced LSC apoptosis within the endosteal niche is enhanced by cell cycle induction of LSCs in vivo. We found significant reductions in both the proportion and number of viable hCD34+ cells in the bone marrow of recipients after chemotherapy with cell cycle induction with the absolute number of hCD34+ bone marrow cells decreasing 1.5- to sevenfold compared with those receiving chemotherapy alone (Supplementary Table 3). Although the reduction in hCD34+ cells indirectly demonstrates the reduction of LSC frequency and number, in vivo evaluation of AML initiation is required to definitively prove the reduction in LSC frequency and function. Therefore, we carried out limiting-dilution serial transplantation of purified viable bone marrow hCD34+ cells, at doses of 2 × 102, 2 × 103, 2 × 104 and 2 × 105 cells per mouse, the mice being AML-engrafted recipients treated with either chemotherapy alone or chemotherapy with cell cycle induction. CD34+ cells, not CD34+CD38− cells, were used as the graft for serial transplantation given the possibility that the CD38 expression level may alter with exposure to G-CSF. Peripheral blood engraftment time course and overall engraftment efficiency of serial transplant recipients at each human CD34+ LSC graft dose are shown in Figure 3c and Supplementary Table 4. At the graft dose of 200 cells, none of the 18 recipients of LSCs from cytokine combined with Ara-C–treated AML-engrafted mice developed AML at 24 weeks post-transplantation. In contrast, 13/18 (72.2%) recipients of 200 LSCs from Ara-C only– treated AML-engrafted mice developed AML. At the dose of 2 × 104 and 2 × 105, all recipients of LSCs from Ara-C only–treated AML-engrafted recipients developed AML by week 24 (n = 38 and n = 14, respectively) whereas only 15/31 and 10/15 recipients of LSCs from G-CSF+Ara-C– treated AML-engrafted mice developed AML. Although the engraftment kinetics varied among LSC samples, at each dose level, the frequency of recipient mice not developing AML was significantly (P < 0.0001) higher in the group treated with G-CSF followed by Ara-C compared with the group treated with Ara-C alone (Fig. 3c and Supplementary Table 4). The frequency of LSCs was estimated by Poisson statistics using the method of maximum likelihood, a standard methodology used in the estimation of HSC frequencies in limiting-dilution transplantation assay12,13. Estimated LSC frequency within bone marrow hCD34+ cells was significantly lower in the recipients after chemotherapy with cell cycle induction (Supplementary Table 5, overall frequencies 1 in 55,076 versus 1 in 560, P = 0.0001 by two-tailed t-test). Furthermore, 24 weeks after transplantation, 7.2% (9 out of 125) of recipients of bone marrow hCD34+ cells from AML-engrafted recipients treated with chemotherapy alone survived whereas 74.3% (81 out of 109) of the recipients of bone marrow hCD34+ cells from AML-engrafted recipients exposed to cytokine before chemotherapy survived (Fig. 3d). Taken together, these findings indicate that in vivo cell cycle entry potentiates the elimination of chemotherapy-resistant human LSCs. Although increased elimination of LSCs correlated with increased long-term survival of the serial transplantation recipients, it was not possible to directly assess the long-term survival of in vivo-treated recipients, owing to the inherent toxicity of the treatments leading to severe complications and frequent death of the recipients shortly after treatment. Our data demonstrate that primary AML LSCs are cell cycle quiescent within the endosteal niche of the bone marrow of a newborn NOD/SCID/IL2rγnull xenotransplantation model. These quiescent
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
LSCs within the bone marrow endosteal region survive in vivo chemotherapy despite efficient elimination of cycling AML cells, consistent with the gap between successful remission induction and overall survival in AML. Furthermore, cell cycle quiescence of human AML LSCs can be modified by in vivo G-CSF administration, leading to substantially enhanced chemotherapy sensitivity of human AML LSCs in vivo. These findings provide proof of concept that cell cycle inducers such as G-CSF can directly modify the functional behavior of LSCs in vivo, increasing their responsiveness to chemotherapy (Fig. 3e,f). The role of G-CSF as an inducer of myeloid maturation, differentiation and proliferation has been successfully exploited to accelerate granulocyte recovery in the setting of myelosuppressive chemotherapy and its function as a mobilizer of HSCs has proven effective for the collection of peripheral blood hematopoietic stem cells14. G-CSF was also found to metabolically activate AML and chronic myelogenous leukemia (CML) cells, enhancing the effects of chemotherapy in vitro. This observation led to clinical studies examining G-CSF during induction chemotherapy for AML15,16. Statistically significant improvements in the rate of relapse and disease-free survival at 4 years in all patients was reported in a large, randomized, controlled study of 640 newly diagnosed AML patients17. Subsequently, five other randomized, controlled studies examining the addition of G-CSF during induction chemotherapy reported equivocal results18–22. The effects of cytokinecombined chemotherapy on LSCs were unevaluated in these studies. It is difficult to interpret the results of these studies as a composite because of different the treatment regimens and patient populations. Two studies evaluate treatment efficacy in relapsed/refractory patients whereas three trials comprised elderly patients. LSC-targeted therapy is unlikely to affect the rates of remission induction substantially, as LSCs are not expected to constitute a substantial portion of the total body AML burden at the time of induction. Therefore, LSC-targeted therapy likely has the greatest positive impact in the setting of post-remission therapy, whether as a part of consolidation therapy or as pretransplant conditioning. This is consistent with uncontrolled single-center experiences reporting improved patient outcomes associated with pretransplant conditioning regimen containing G-CSF23,24. We observed considerable variability in the response of LSCs to cell cycle induction and the frequency of residual functional LSCs after chemotherapy with cell cycle induction, reflecting biological hetero geneity of AML. Further studies are required to determine if LSC response to cell cycle modifying agents correlates with patient characteristics such as age, cytogenetics, presence of specific mutations and response to prior therapies. In addition, the LSC responsiveness to cell cycle induction itself as a prognostic indicator must be investigated. Agents that modify cell cycle or disrupt stem cell–niche inter actions are also expected to affect normal HSCs. To begin to address this issue, we examined in vivo survival of normal HSCs after Ara-C alone compared with Ara-C following G-CSF in NOD/SCID/IL2rγnull recipients reconstituted with normal human hematopoiesis25. We found that the rate of apoptosis in hCD45+CD34+CD38− human HSCs was not significantly increased after cytokine-combined chemotherapy compared with chemotherapy alone (Supplementary Fig. 5). These findings are consistent with the lack of excess hematologic toxicity in patients receiving cytokine-combined chemotherapy in previously published clinical studies17–22. More extensive evaluation of the effects of the proposed therapy including long-term repopulation and selfrenewal capacity of human HSCs is required. Normal HSC rescue may be required to optimize the effect of cytokine-combined anti-LSC therapy and to assure patient safety. In a model of murine leukemia, restriction of cell cycle entry appears to contribute to LSC maintenance, further supporting the
279
© 2010 Nature America, Inc. All rights reserved.
letters essential role of cell cycle regulation in LSC function 26. Therefore, agents that promote LSC cell cycle entry and/or disrupt the inter actions between LSCs and the bone marrow niche, including modifiers of CXCL12/CXCR4, VLA-4/fibronectin and Groβ/CXCR2 axes, may effectively alter LSC chemotherapy responsiveness27–29. Recently, it was demonstrated that a CXCR4 antagonist mobilizes murine leukemic cell lines and human AML cells30. Similarly, mobilization and chemosensitization of murine leukemia with a CXCR4 antagonist has also been shown31. In addition, recent studies have elegantly demonstrated that interferon-α induces the proliferation of murine normal HSCs and that the suppression of interferon-α signaling protects murine normal HSCs from exhaustion through the maintenance of cell cycle quiescence32,33. These findings have led to the hypothesis that the mechanism of interferon-α action against CML may be related to its effects on the cell cycle status of CML stem cells34. Whether these agents are effective in altering the cell cycle status and chemotherapy responsiveness of LSCs must be investigated in various hematologic malignancies. Here we demonstrated that functional modification of human primary AML LSCs facilitates their effective elimination in vivo. Recent studies have also demonstrated that cell surface markers expressed on LSCs may be effective targets for anti-LSC therapy35–37. Multifaceted investigational approaches that integrate functional modification of LSCs with discovery of LSC-specific target molecules will be required for the development of effective LSC-targeted therapies for AML. 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 thank T. Tanaka and M. Yamaguchi for assistance with confocal microscopy. We thank T. Kanabayashi for assistance with histological studies. We thank M. Narita for assistance with patient data collection. This work was supported through grants from the Ministry of Education, Culture, Sports, Science and TechnologyJapan, National Institute of Biomedical Innovation, Takeda Science Foundation and Uehara Memorial Foundation to F.I. and National Institutes of Health grant CA20408 to L.D.S. AUTHOR CONTRIBUTIONS Y.S., L.D.S. and F.I. designed the study, analyzed the data and wrote the manuscript. N.U., A.W. and S. Taniguchi provided clinical samples, information and discussion. Y.S., S. Tanaka, M.T.-M., N.S., A.S. and F.I. performed the experiments. S. Takagi and Y.A. analyzed the data. Y.N. analyzed the data 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. Bonnet, D. & Dick, J.E. Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat. Med. 3, 730–737 (1997). 2. Jordan, C.T., Guzman, M.L. & Noble, M. Cancer stem cells. N. Engl. J. Med. 355, 1253–1261 (2006). 3. Lapidot, T. et al. A cell initiating human acute myeloid leukaemia after transplantation into SCID mice. Nature 367, 645–648 (1994). 4. Ishikawa, F. et al. Chemotherapy-resistant human AML stem cells home to and engraft within the bone-marrow endosteal region. Nat. Biotechnol. 25, 1315–1321 (2007). 5. Estey, E. & Dohner, H. Acute myeloid leukaemia. Lancet 368, 1894–1907 (2006). 6. Dick, J.E. Stem cell concepts renew cancer research. Blood 112, 4793–4807 (2008). 7. Clarke, M.F. et al. Cancer stem cells–perspectives on current status and future directions: AACR Workshop on cancer stem cells. Cancer Res. 66, 9339–9344 (2006). 8. Lapidot, T. & Petit, I. Current understanding of stem cell mobilization: the roles of chemokines, proteolytic enzymes, adhesion molecules, cytokines, and stromal cells. Exp. Hematol. 30, 973–981 (2002).
280
9. Wilson, A. et al. Hematopoietic stem cells reversibly switch from dormancy to selfrenewal during homeostasis and repair. Cell 135, 1118–1129 (2008). 10. Sugiyama, T., Kohara, H., Noda, M. & Nagasawa, T. Maintenance of the hematopoietic stem cell pool by CXCL12-CXCR4 chemokine signaling in bone marrow stromal cell niches. Immunity 25, 977–988 (2006). 11. Sipkins, D.A. et al. In vivo imaging of specialized bone marrow endothelial microdomains for tumour engraftment. Nature 435, 969–973 (2005). 12. Conneally, E., Cashman, J., Petzer, A. & Eaves, C. Expansion in vitro of transplantable human cord blood stem cells demonstrated using a quantitative assay of their lympho-myeloid repopulating activity in nonobese diabetic-scid/scid mice. Proc. Natl. Acad. Sci. USA 94, 9836–9841 (1997). 13. Wang, J.C., Doedens, M. & Dick, J.E. Primitive human hematopoietic cells are enriched in cord blood compared with adult bone marrow or mobilized peripheral blood as measured by the quantitative in vivo SCID-repopulating cell assay. Blood 89, 3919–3924 (1997). 14. Smith, T.J. et al. 2006 update of recommendations for the use of white blood cell growth factors: an evidence-based clinical practice guideline. J. Clin. Oncol. 24, 3187–3205 (2006). 15. Jorgensen, H.G. et al. Intermittent exposure of primitive quiescent chronic myeloid leukemia cells to granulocyte-colony stimulating factor in vitro promotes their elimination by imatinib mesylate. Clin. Cancer Res. 12, 626–633 (2006). 16. te Boekhorst, P.A., Lowenberg, B., Vlastuin, M. & Sonneveld, P. Enhanced chemosensitivity of clonogenic blasts from patients with acute myeloid leukemia by G-CSF, IL-3 or GM-CSF stimulation. Leukemia 7, 1191–1198 (1993). 17. Lowenberg, B. et al. Effect of priming with granulocyte colony-stimulating factor on the outcome of chemotherapy for acute myeloid leukemia. N. Engl. J. Med. 349, 743–752 (2003). 18. Amadori, S. et al. Use of glycosylated recombinant human G-CSF (lenograstim) during and/or after induction chemotherapy in patients 61 years of age and older with acute myeloid leukemia: final results of AML-13, a randomized phase-3 study. Blood 106, 27–34 (2005). 19. Buchner, T., Berdel, W.E. & Hiddemann, W. Priming with granulocyte colonystimulating factor–relation to high-dose cytarabine in acute myeloid leukemia. N. Engl. J. Med. 350, 2215–2216 (2004). 20. Estey, E.H. et al. Randomized phase II study of fludarabine + cytosine arabinoside + idarubicin +/− all-trans retinoic acid +/− granulocyte colony-stimulating factor in poor prognosis newly diagnosed acute myeloid leukemia and myelodysplastic syndrome. Blood 93, 2478–2484 (1999). 21. Milligan, D.W., Wheatley, K., Littlewood, T., Craig, J.I. & Burnett, A.K. Fludarabine and cytosine are less effective than standard ADE chemotherapy in high-risk acute myeloid leukemia, and addition of G-CSF and ATRA are not beneficial: results of the MRC AML-HR randomized trial. Blood 107, 4614–4622 (2006). 22. Ohno, R. et al. A double-blind controlled study of granulocyte colony-stimulating factor started two days before induction chemotherapy in refractory acute myeloid leukemia. Kohseisho Leukemia Study Group. Blood 83, 2086–2092 (1994). 23. Mori, T. et al. Long-term follow-up of allogeneic hematopoietic stem cell transplantation for de novo acute myelogenous leukemia with a conditioning regimen of total body irradiation and granulocyte colony-stimulating factor-combined highdose cytarabine. Biol. Blood Marrow Transplant. 14, 651–657 (2008). 24. Ooi, J. et al. Unrelated cord blood transplantation for adult patients with de novo acute myeloid leukemia. Blood 103, 489–491 (2004). 25. Ishikawa, F. et al. Development of functional human blood and immune systems in NOD/SCID/IL2 receptor γ chain(null) mice. Blood 106, 1565–1573 (2005). 26. Viale, A. et al. Cell-cycle restriction limits DNA damage and maintains self-renewal of leukaemia stem cells. Nature 457, 51–56 (2009). 27. Matsunaga, T. et al. Interaction between leukemic-cell VLA-4 and stromal fibronectin is a decisive factor for minimal residual disease of acute myelogenous leukemia. Nat. Med. 9, 1158–1165 (2003). 28. Christopher, M.J., Liu, F., Hilton, M.J., Long, F. & Link, D.C. Suppression of CXCL12 production by bone marrow osteoblasts is a common and critical pathway for cytokine-induced mobilization. Blood 114, 1331–1339 (2009). 29. Fukuda, S., Bian, H., King, A.G. & Pelus, L.M. The chemokine GRObeta mobilizes early hematopoietic stem cells characterized by enhanced homing and engraftment. Blood 110, 860–869 (2007). 30. Zeng, Z. et al. Targeting the leukemia microenvironment by CXCR4 inhibition overcomes resistance to kinase inhibitors and chemotherapy in AML. Blood 113, 6215–6224 (2009). 31. Nervi, B. et al. Chemosensitization of AML following mobilization by the CXCR4 antagonist AMD3100. Blood 113, 6206–6214 (2009). 32. Sato, T. et al. Interferon regulatory factor-2 protects quiescent hematopoietic stem cells from type I interferon-dependent exhaustion. Nat. Med. 15, 696–700 (2009). 33. Essers, M.A. et al. IFNα activates dormant haematopoietic stem cells in vivo. Nature 458, 904–908 (2009). 34. Sipkins, D.A. Rendering the leukemia cell susceptible to attack. N. Engl. J. Med. 361, 1307–1309 (2009). 35. Jin, L. et al. Monoclonal antibody-mediated targeting of CD123, IL-3 receptor alpha chain, eliminates human acute myeloid leukemic stem cells. Cell Stem Cell 5, 31–42 (2009). 36. Majeti, R. et al. CD47 is an adverse prognostic factor and therapeutic antibody target on human acute myeloid leukemia stem cells. Cell 138, 286–299 (2009). 37. Saito, Y. et al. Identification of therapeutic targets for quiescent, chemotherapyresistant human leukemia stem cells. Sci. Transl. Med. 2, 17ra9 (2010).
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
ONLINE METHODS Patient samples. All experiments were performed with authorization from the Institutional Review Board for Human Research at RIKEN RCAI. Following written informed consent, AML cells were collected from patients with FrenchAmerican-British (FAB) classification system subtype M1 (case 4), M2 (cases 3, 6, 7), M4 (cases 1, 2) and MDS/AML (case 5). Bone marrow mononuclear cells (BMMNCs) were isolated using density-gradient centrifugation.
© 2010 Nature America, Inc. All rights reserved.
Mice. NOD.Cg-PrkdcscidIl2rgtmlWjl/Sz (NOD/SCID/IL2rγnull) mice were developed at The Jackson Laboratory by backcrossing a complete null mutation at the Il2rg locus onto the NOD.Cg-Prkdcscid (NOD/SCID) strain. Mice were bred and maintained under defined flora with irradiated food and acidified water at the animal facility of RIKEN and at The Jackson Laboratory according to guidelines established by the Institutional Animal Committees at each respective institution. Xenogeneic transplantation. Newborn (within 2 d of birth) NOD/SCID/ IL2rγnull recipients received 150 cGy total body irradiation using a 137Cssource irradiator, followed by intravenous injection of AML cells within 2 h. For primary transplantation (to generate human AML-engrafted mice to be used for subsequent experiments), 103 – 5 × 104 sorted 7AAD− lineage (hCD3/hCD4/hCD8)−hCD34+hCD38− AML patient bone marrow cells per recipient were injected, as described previously4. For secondary transplantation (to determine the frequency of AML-initiating cells remaining after Ara-C alone or Ara-C following G-CSF), 2 × 102, 2 × 103, 2 × 104 or 2 × 105 sorted 7AAD−hCD45+hCD34+ bone marrow cells from treated AML-engrafted recipients were transplanted into newborn NOD/SCID/IL2rγnull recipients following the same irradiation and injection procedure as described above. For fluorescence-activated cell sorting (FACS), AML patient BMMNCs were labeled with fluorochrome-conjugated mouse anti-hCD3, anti-hCD4, antihCD8, anti-hCD34 and anti-hCD38 monoclonal antibodies (BD Biosciences) and recipient BMMNCs were labeled with mouse anti-hCD45, anti-hCD34 and anti-hCD38 monoclonal antibodies (BD Biosciences) followed by cell sorting using FACSAria (BD Biosciences). Doublets were excluded by analysis of FSC/SSC-height and FSC/SSC-width. The purity of hCD34+hCD38− and hCD34+ cells was >98% after sorting. Ara-C and G-CSF treatments. For experiments comparing no treatment, G-CSF alone, Ara-C alone and Ara-C following G-CSF, primary human AMLengrafted recipients at 12–24 weeks post-transplantation were used. For each experiment comparing treatment groups, recipient pairs were chosen from the same litter, transplanted with the same primary AML sample at the same dose on the same day, to control for variability among litters and variability due to differences in engraftment levels. Recombinant human G-CSF (PeproTech) was given as follows: 300 µg/kg subcutaneously once a day for 5 d. Ara-C (Biogenesis) was given as follows: 1g/kg intraperitoneally once a day for 2 d. Ara-C following G-CSF treatment was given as follows: G-CSF 300 µg/kg subcutaneously once a
doi:10.1038/nbt.1607
day for 5 d with concurrent administration of Ara-C 1g/kg intraperitoneally once a day on treatment days 4 and 5. The recipients were euthanized 16 h after the final injection. For recipients undergoing cell cycle analysis by BrdU incorporation, 1.5 mg BrdU (BD Biosciences) per recipient was injected intraperitoneally immediately after the final injection of G-CSF and/or Ara-C. Flow cytometry. Recipients were bled retro-orbitally every 3 weeks starting week 6 post-transplantation for the evaluation of human AML engraftment. For the analysis of Ara-C alone and Ara-C following G-CSF–treated recipients, bone marrow was harvested by flushing two tibiae and one femur and bone marrow cell count was performed manually and by automated blood cell analyzer (Celltac α, Nihon Kohden), to estimate the absolute number of bone marrow nucleated cells from each recipient. The absolute number of human CD34+ cells per mouse (from two tibiae and one femur) was determined by multiplying percent bone marrow 7AAD−hCD45+hCD34+ cells with total bone marrow nucleated cell count thus obtained. For quantification of cells in G0/G1, S and G2/M phases of cell cycle, in vivo BrdU incorporation was measured using the BrdU flow kit (BD Pharmingen). For quantification of cells in G0 phase of cell cycle, cells were labeled with Hoechst 33342 and Pyronin Y followed by surface staining using standard procedures. Intracellular staining for active caspase-3 using rabbit anti-active caspase-3 monoclonal antibody (BD Pharmingen) was performed for the quantification of apoptotic cells. For surface labeling, mouse anti-human CD45, CD34 and CD38 monoclonal antibodies were used (BD Biosciences). The analyses were performed using FACSAria and FACSCanto II (BD Biosciences). Histological analysis. Paraformaldehyde-fixed, decalcified, paraffin-embedded sections were prepared from femurs of primary human AML-engrafted recipients. Primary antibodies used for labeling were mouse anti-human CD45 monoclonal (Immunotech), anti-human CD34 monoclonal (Beckman Coulter), rabbit anti-human Ki67 polyclonal (Spring Bioscience), mouse anti-BrdU monoclonal (DAKO) and rabbit anti-mouse osteopontin polyclonal (IBL) antibodies. HE staining was performed using standard methodology. TUNEL staining was performed according to standard procedures using ApopTag peroxidase in situ apoptosis detection kit (Intergene). Light microscopy was performed using Zeiss Axiovert 200 (Carl Zeiss Microimaging). Laser-scanning confocal imaging was obtained using Zeiss LSM 710 (Carl Zeiss). Three-dimensional reconstructions of bone sections were performed by analyzing serial sectional confocal images using the IMARIS software (BITPLANE). Statistical analysis. Unless otherwise noted, numerical data are presented as mean ± s.e.m. and differences are examined by two-tailed t-test (GraphPad Prism, GraphPad). The difference in survival was analyzed by the log-rank (Mantel-Cox) test (GraphPad Prism, GraphPad). LSC frequency was estimated through Poisson statistics using the method of maximum likelihood and analyzed using two-tailed t-test (StemSoft Software).
nature biotechnology
letters
© 2010 Nature America, Inc. All rights reserved.
Isotopic labeling of terminal amines in complex samples identifies protein N-termini and protease cleavage products Oded Kleifeld1,2,4–6, Alain Doucet1,2,4,6, Ulrich auf dem Keller1,2,4,5, Anna Prudova1,2,4, Oliver Schilling1,2,4,5, Rajesh K Kainthan3,4, Amanda E Starr1,4, Leonard J Foster1, Jayachandran N Kizhakkedathu3,4 & Christopher M Overall1,2,4 Effective proteome-wide strategies that distinguish the N-termini of proteins from the N-termini of their protease cleavage products would accelerate identification of the substrates of proteases with broad or unknown specificity. Our approach, named terminal amine isotopic labeling of substrates (TAILS), addresses this challenge by using dendritic polyglycerol aldehyde polymers that remove tryptic and C-terminal peptides. We analyze unbound naturally acetylated, cyclized or labeled N-termini from proteins and their protease cleavage products by tandem mass spectrometry, and use peptide isotope quantification to discriminate between the substrates of the protease of interest and the products of background proteolysis. We identify 731 acetylated and 132 cyclized N-termini, and 288 matrix metalloproteinase (MMP)-2 cleavage sites in mouse fibroblast secretomes. We further demonstrate the potential of our strategy to link proteases with defined biological pathways in complex samples by analyzing mouse inflammatory bronchoalveolar fluid and showing that expression of the poorly defined breast cancer protease MMP-11 in MCF-7 human breast cancer cells cleaves both endoplasmin and the immunomodulator and apoptosis inducer galectin-1. Characterization of protein N-termini is an essential aspect of the functional annotation of any proteome. Protein isoforms and N-terminal post-translational modifications, such as proteolytic truncations and acetylation1–3 influence the localization and activity of many proteins. Several strategies to selectively isolate protein N-terminal peptides, also known as N-terminomics, have been described4–7, but until combined fractional diagonal chromatography (COFRADIC)4, such positional proteomics approaches were not reported for global protease cleavage site analysis, also known as degradomics. Because analysis of N-terminal peptides reveals information about both the protein targets and sites of cleavage of proteases, new approaches to distinguish the N-termini of mature proteins from protease-generated
(neo)-N termini have recently gained traction (Supplementary Discussion). Positive-selection techniques to identify the N-termini generated by digestion with a protease of interest modify the free α-amines of proteins and the neo-N-termini of protease cleavage products for enrichment and exploit natural N-terminal acetylation to prevent co-purification of these blocked peptides 4,6,8–10. However, because techniques avoiding diversely blocked peptides exclude the simultaneous analysis of most of the protein N termini present before exposure to a protease of interest, they are not also suitable for comprehensive analysis and annotation of the N-termini of proteins in complex proteomes. Despite recent advances in identifying the acetylated N termini of mature proteins2,11, using single peptides to identify proteins12 and discover protease substrates with low false-positive rates remains especially problematic for proteases with broad or unknown specificity where manual parsing of the data for known cleavage sites cannot be performed. To tackle these challenges we developed TAILS, a 3-d, three-step quantitative proteomics approach for labeling and isolating N-terminal peptides before and after exposure to a protease of interest (Fig. 1a). After proteome-wide proteolysis, the protease is inactivated and the sample is denatured and reduced. To quantify cleavage events specific to the protease of interest and to distinguish these from proteolysis products present in an untreated sample, we introduced stable isotopes to determine the relative abundances of peptides in the protease-treated and control samples. Reductive dimethylation of primary amines using (d(2)C13)-formaldehyde (protease-treated, heavy) or (d(0)C12)formaldehyde (control, light), catalyzed by sodium cyanoborohydride13, simultaneously labels and blocks all lysine ()-amines, as well as the free (α)-amino groups of N termini on all proteins and their protease cleavage products. The labeled proteomes are combined and trypsinized. As lysine dimethylation precludes trypsin cleavage, by cleaving with arginyl peptidase (ArgC)-like specificity after arginines only, the resultant generation of longer peptides by trypsin aids the identification of the protease-shortened neo-N-terminal peptides generated by cleavage.
1Department
of Biochemistry and Molecular Biology, 2Department of Oral Biological and Medical Sciences and 3Department of Pathology and Laboratory Medicine, for Blood Research, University of British Columbia, Vancouver, British Columbia, Canada. 5Present addresses: Israel Institute of Technology, Faculty of Biology, Haifa, Israel (O.K.); Institute of Cell Biology, ETH Zürich, Zürich, Switzerland (U.a.d.K.), University of Freiburg, Institute for Molecular Medicine and Cell Research, Freiburg, Germany (O.S.). 6These authors contributed equally to this work. Correspondence should be addressed to C.M.O. ([email protected]) or J.N.K. ([email protected]). 4Centre
Received 16 November 2009; accepted 28 January 2010; published online 7 March 2010; doi:10.1038/nbt.1611
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
281
Proteome NH2
NH2
K
K
K
K
Time (h) 0
Proteolysis NH2 NH2
Control NH2
NH2 K
K K
K
K
NH2 K
NH2 K
Isotopic labeling of amine groups
16
20
Sample mixing
NH2
Tryptic digestion NH2 NH2
NH2 NH2
NH2 36
K
NH2
NH2 NH2
NH2
Capture of internal peptides with amine-reactive polymer NH
HN
282
a
NH
N
Figure 1 Schematic diagram of TAILS and the polymer developed for proteomics. (a) The TAILS workflow and scheme for bioinformatics analysis. Unfractionated proteins from control (blue)- and protease (green)-treated samples are labeled on the primary amines of the N-termini (NH 2) and lysines (K) by dimethylation using light (d(0)C12)-formaldehyde (blue spheres) or heavy (d(2)C13)-formaldehyde (green spheres), respectively. Gray pentagons represent a naturally blocked protein N-terminus. Protease cleavage (scissors) generates fragments cleaved on the nonprime (P) side (for naturally blocked peptides; shown in orange) and fragments cleaved on the prime (P′) side (neo-N-terminal peptides; shown in red). The high protease/control ratio peptides (≥3.0 or heavy singleton) correspond to protease-generated neo-N-termini (peak C). These are distinguished from background proteolysis products and original mature N-terminal peptides that have an isotope ratio of ~1.0 because they are equally abundant in both samples. Two examples of peptide pairs with 6.0 ∆ m/z (bracketed) due to the differential labeling are shown. Mature protein N-terminal peptides or background proteolysis products are shown in peak A. In peak B, naturally blocked N-terminal peptides (Ac, but this includes all forms of blocked N-termini) are labeled if lysines (K) are present. Cleavage close to a labeled N-terminus results in loss of such tryptic peptides and hence in a low isotope ratio (≤3.0 or light singleton), as the tryptic peptide that spans the protease cleavage site is only present in the control sample (peak D). Cleavage in naturally blocked peptides can also be so analyzed when lysine labeled (appearing like peak D). A time line indicating approximate duration of the various steps in a typical TAILS protocol is shown. Data analysis time varies based on the number of samples, labeling strategy and sample requirements. (b) Chemical structure of HPG-ALD polymer.
validation approach that involves further evaluation of only N-terminally dimethylated peptides within 5 p.p.m. of their expected mass for high-confidence N-terminal peptide identification. Identification of naturally blocked N-terminal peptides with the same
HN H
NH
NH NH
52
Negative selection of peptides with blocked N-termini by centrifugation
MS/MS analysis of N-terminal peptides Mature N-termini Protease labeled Ac+ labeled K -generated -depleted N-termini 1.0 1.0 Protease/control high low
53
Intensity
Next, TAILS uses a new class of highly water-soluble (concentrations of ~100 mg/ml are routinely feasible) polymers to selectively enrich the blocked N-terminal peptides by negative selection. We synthesized a series of hyperbranched polyglycerols (HPGs) with molecular weights in the range of 100–600 kDa and narrow polydispersity14 and converted them to dendritic HPG-aldehydes (ALDs) by periodate oxidation of their 1,2-diol groups using HIO4 (Fig. 1b). The HPG-ALD polymer readily reacts with all unblocked internal and C-terminal tryptic peptides through their free N-termini in sodium cyanoborohydride. However, along with the dimethylated lysines, acetylated, cyclized and isotopically labeled protein N-terminal peptides and the neo-N-terminal peptides of their cleavage products are unreactive, remaining unbound for recovery by ultrafiltration. Binding of tryptic peptides occurs at up to 2.5 mg peptide/mg polymer, a more than tenfold improvement in capacity over amine-reactive resins, with no nonspecific interactions with peptides (Supplementary Results 1). This reduces the number of false-positive peptides identified and improves sample recovery, thereby reducing the sample size for analysis to 100 µg, compared to >50 mg in a recently described procedure10. Due to the massive sample simplification and high recovery, proteome coverage using TAILS is excellent, even without sample prefractionation before a single tandem mass spectrometry (MS/MS) analysis using a high mass accuracy LTQ-Orbitrap, unlike other approaches which can require up to 150 analyses per sample11. However, coverage can be increased with prefractionation by strong cation exchange chromatography and ten MS/MS analyses. Confident protein assignment from a single peptide is critical for the successful application of a N-terminal positional proteomics technique. However, since protein identification confidence increases with the detection of two or more peptides per protein and this is not always possible in a successful N-terminal positional proteomics experiment, we applied a combination of highly accurate MS mea surements, orthogonal validation and stringent data analysis criteria to address this issue. Using Mascot and X! Tandem to search for peptide sequences against protein and peptide databases15, those identified by both search engines are compiled. Next, we used an orthogonal
HN
© 2010 Nature America, Inc. All rights reserved.
letters
A 56
B
C
m/z
D
Bioinformatics analysis of three biological replicates measured by high mass accuracy MS/MS Peptide identification: database (IPI protein + peptide) searches of MS/MS data by Mascot + X! Tandem Peptide validation: selection of peptides with ≥99% confidence based on PeptideProphet probability Protein identification and N-terminome description: From >2 different unique peptides (original mature, neo-N-terminal) and/or double identification of peptides in different biological replicates, charge states or modifications
181
Hierarchical substrate winnowing substrate identification: isotope ratio-based classification of high-confidence individual cleavage sites from neo-N peptides (C) or indirectly from peptide loss (D) Biologically relevant candidate substrate selection: cellular localization + literature reports O
b
H
H
HO O
O
O
O O
O
OH
O
O
O
O
O O
O H
H
O H
O O
HO
O O
H H
O O
O
HO
O
O O
O
O
HO
O
O O
O
H O
OH
O
O
H
H
O
O
O
OH
O
H H
O
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
letters stringency simply requires changing the search parameters. We then validated individual search results using the Trans-Proteomic Pipeline with a PeptideProphet false-discovery rate of <1%16. We further reduced the chance of false protein identification by prescribing that any N-terminal peptide must be identified twice if it is to be included in further analyses. These identifications can be from biological replicates, different charge states, or of the same peptide bearing different natural modifications, such as oxidation of methionine or deamidation of glutamine or asparagine. Hence, all peptide numbers reported are very conservative. The iProphet tool of the Trans-Proteomic Pipeline16 then combines all peptide search results for high-confidence protein identification. If the N-terminal peptide of the intact protein is also identified with one or several neo-N-terminal peptides (because of multiple cleavage sites) in the same sample, this also enables high-confidence protein identification by two or more different and unique peptides. This is an additional advantage of negative selection approaches, including COFRADIC, that also analyze mature N-terminal peptides in the same sample. The last step in the bioinformatics component of TAILS analysis involves a process we refer to as hierarchical substrate winnowing, whereby high-confidence biologically relevant candidate substrates are selected from all possible candidates identified from neoN-terminal peptides. This commences with analyzing isotope-labeled peptide abundance ratios to identify peptides either generated (high protease/control ratio, heavy singletons) or lost (low protease/control ratio, light singletons) after proteolysis (Fig. 1a). Peptide quantification is key to reduce the substrate false-discovery rate. Hence, specific protease products are distinguished from those of other proteases against the background of both samples and original mature protein N-termini with isotope ratios of ~1.0. Other criteria for hierarchical substrate winnowing are described below. Matrix metalloproteinases (MMPs) are a family of 23 extracellular proteases important for maintaining homeostasis in humans in the extracellular environment and when expressed in elevated amounts or new locations contribute to pathology. Of these, MMP-2 is one of the four signature proteins involved in breast cancer metastasis17 with pleotrophic activites18 including mobilizing VEGF19 and regulating innate immunity20. We used a six-protein mixture incubated with this broad-specificity protease21 (Supplementary Results 2), to identify 56 labeled original mature N-terminal and MMP-2-cleaved neoN-terminal peptides. We identified both the N-terminal peptide of the chemokine human monocyte chemoattractant protein 3 (CCL7) (pyr1QPVGINTSTTCCYR) and the heavy-labeled CCL7 neo-N-terminal peptide (5INTSTTCCYR), confirming cleavage site results obtained using Edman sequencing20. In silico N-terminal–labeled peptide selection22 before polymer enrichment identified only 15 peptides (Supplementary Results 2), demonstrating the analytical advantage of negative selection over analyses without selection. To test the capacity of TAILS to analyze a complex proteome, we subjected fibroblast secretome samples (n = 3) from mouse to partial digestion in their native states with the endoproteinase GluC, which cleaves after glutamate and aspartate residues and thus enables manual data parsing for validation. Of the 1,501 peptides identified before TAILS, only 73 were high-ratio GluC generated, compared with 531 after TAILS. This was coupled to a decrease in tryptic peptides from 1,177 to 42 (Table 1a and Supplementary Results 3), thereby increasing neo-N-terminal peptides from 4.8% to 62%. The efficiency of the dimethylation and labeling reaction is reflected by the demonstration that only seven peptides were cleaved at lysine, which occurs only if its side chain is unmodified. However, it is possible that these peptides are proteolysis products of other proteases and so might not reflect missed labeling.
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
In other approaches, manual parsing easily identifies cleaved peptides generated by highly specific proteases such as caspases9,10,15,22,23 and GluC21, but this is not possible for proteases with unknown or broad specificity. We overcame this problem in TAILS by using heavy/ light isotope (protease/control) ratios to classify peptides. Whereas high-ratio, semi-tryptic peptides represent proteolytic products generated by the protease of interest, peptides with ratios ~1.0 (including mature protein N-terminal peptides and background proteolysis products from other proteases in the sample before or during harvesting, or during assay independent of the test protease) are unaffected by the test protease. Low-ratio peptides are those eliminated through cleavage by the protease of interest. To identify cleavage sites from ratio analysis, we quantified the distribution of those mature protein N-terminal– labeled peptides that lack a glutamate and naturally acetylated N-terminal peptides containing labeled lysine (Supplementary Results 3). From their average ratio of 0.93 ± 0.39, we set the cutoff ratio at 3 (mean ± 5 × s.d.) for peptides that are significantly different. Indeed, of the 551 GluC cleavage products, 531 had ratios >3 (Table 1a). For the 20 peptides with ratios <3, quantification of spectral noise or co-elution and co-fragmentation of unrelated peptides that overlap with the isotopic couples (∆6 m/z) reduces the ratios from true singletons. However, if desired, prefractionation can help to resolve this by improving peptide separation. Procollagen C-endopeptidase enhancer-1 is one of the proteins we show to be cleaved by GluC. Before TAILS, we identified 11 tryptic peptides (which the polymer then removed), but no neo-N-terminal peptides (Table 1b). One tryptic peptide spanning a GluC cleavage site was only in the undigested control as a light singleton before TAILS and its loss was coincident with the generation of its GluC neoN-terminal product identified only after TAILS as a heavy singleton. A second high-ratio GluC peptide was also found only after TAILS. Together, quality LTQ-Orbitrap spectra (Supplementary Results 3) and ∆m/z of 0.001 Da enabled unambiguous peptide identification and confirmation of cleavage by GluC after a glutamate by Mascot, X! Tandem and de novo sequencing. We next assessed the capabilities of TAILS by applying it to study the broad-specificity metalloprotease MMP-2. We used Mmp2−/− fibro blast secretomes (n = 3), as these are MMP-2 naive proteomes, performing at least two or four technical replicates in each experiment or labeling swap controls. Supplementary Results 4 lists 1,223 peptides from 546 proteins identified by Mascot and X! Tandem. The data were also searched against a semi-ArgC peptide-centric database15 (1,079 peptides, 415 proteins) (Supplementary Results 5). Our observation that many of the peptides are identified in all biological and technical replicates and in every search (e.g., neo-N-terminal peptides of bone morphogenetic protein-1 (BMP-1) and Protein SET isoform-1) demonstrates the reproducibility of the approach. The neopeptide (NH2-4INTSTTCCYR-COOH, m/z = 1,160.50) from MMP-2 cleavage of 100 pmol human CCL7 at 4Gly-Ile spiked into samples as a positive control underscores the sensitivity of TAILS. High-confidence neo-N-terminal peptides with ratios >3 were compiled and further validated by iProphet16 to generate a list of substrate candidates identified from identical peptides in two or more biological replicates as the next step in hierarchical substrate winnowing (Supplementary Results 6). Substrates were also often identified by multiple peptide forms (~30%) and ~50% of the substrates were identified by two or more different peptides arising from multiple neopeptides per protein, as well as the original mature protein N-terminal peptide. One hundred forty-six high-confidence MMP-2 substrates, identified by 288 cleaved peptides, were selected (Fig. 2a and Supplementary Results 6). Since MMP-2 is an extracellular
283
letters other neo-N-terminal peptides in Supplementary Results 4 and 5, which did not meet our strict criteria, might represent real cleavage sites, although this requires validation. To confirm the accuracy and sensitivity of TAILS, we examined the susceptibility of the cysteine protease inhibitor cystatin C and the chemokine fractalkine (CX3CL1; this protein is naturally present at <100 pg/ml) to MMP-2 cleavage. In both cases, the cleavage site determined by TAILS exactly matches the in vitro MMP-2 cleavage site that we determined by Edman sequencing18,19 (Fig. 2b). N-terminal truncation of cystatin C reduces its inhibitory activity towards cysteine proteases19 and MMP-2 cleavage of CX3CL1 coverts it to a receptor antagonist and sheds the chemokine from the cell surface18. We selected six other MMP-2 candidate substrates (Table 1c) for validation by cleavage assays. Full-length proBMP-1, a metalloprotease involved in procollagen maturation, was cleaved in the propeptide by MMP-2 (Fig. 2c). We used biochemical assays to confirm that fibulin-2, extracellular matrix protein-1 (ECM-1), biglycan, sulfated
© 2010 Nature America, Inc. All rights reserved.
protease that is activated on the cell surface, 34 of the cleaved proteins that are extracellular or bound to the cell membrane are likely to be of biological relevance (see tables in Supplementary Discussion); 20 more are known substrates, representing 33% of validated MMP-2 protein substrates in the MEROPS database; four more are substrates of other MMPs and eight are homologous to proteins cleaved by other MMPs (and thus seem strong candidates for being MMP-2 substrates). Finally, 82 of the other proteins we identified have been shown in other proteomics screens to be affected by MMP-2 activity, but not validated by determining their cleavage sites18,19. The ability of TAILS to identify a total of 113 known MMP-2 substrates (Fig. 2a), validates its capacity to identify substrates of broad specificity proteases from complex samples. Further, the consistency of P4-P4′ subsite specificity in the known and candidate substrates supports their identification as substrates (see figures in Supplementary Discussion). The most biologically relevant new substrates are shown in Table 1c. Nonetheless, it is worth noting that
Table 1 TAILS analysis of mouse fibroblast secretome digested by GluC and MMP-2 a. Number of peptides identified after secretome digestion with GluC Before TAILS Sample
Internal tryptic
1 2 3 Total
776 252 149 1,177
Protein mature GluC H/L > 3 GluC Others N-term Others 42 20 11 73
6 14 13 33
30 10 11 51
112 34 21 167
Total
Sample
966 330 205 1,501
1 2 3 Total
Internal tryptic 26 6 10 42
After TAILS
GluC Protein mature GluC N-term Others H/L > 3 Others 286 160 85 531
16 3 1 20
64 18 22 104
122 24 17 163
Total 514 211 135 860
b. Peptides identified for procollagen C-endopeptidase enhancer-1 (IPI00120176) Before TAILS After TAILS Protease/ control
Position
Mass
P1
Peptide
80–90 91–106 110–129 116–129 140–148 202–213 214–230 275–291 411–431 432–445 446–458
1389.6654 1656.8019 2186.1763 1417.8296 1097.5854 1513.7177 1872.9206 1856.9387 2445.2658 1585.9173 1556.8592
R R R R R F R R R R R
VFDMELHPSCR YDALEVFAGSGTSGQR FCGTFRPAPVVAPGNQVTLR PAPVVAPGNQVTLR GFLLWYSGR GKFDVEPDTYCRKGa YDSVSVFNGAVSDDSKRa DAVEKESALSPGEDVQRa QMPPMKKGASYLLMGQVEENRa GPILPPESFVVLYR SNQDQILNNLSKRa
Position 71–79 281–291
Mass
Mass difference
1027.5886 0.0012 1191.6323 0.0015
Peptide P1 E E
Protease/ control
GQTVSLSFR 16.67 SALSPGEDVQR H Singleton
0.88 0.85 L Singleton 0.78 0.85
c. Extracellular, secreted and membrane MMP-2 candidate substrates Extracellular matrix protein 1 (IPI00469845)b Sulfated glycoprotein 1 (IPI00321190)b Bone morphogenic protein 1 (IPI00469541)b Fibulin 2 isoform b (IPI00830803)b,c Peptidyl-prolyl cis-trans isomerase A (cyclophilin A) (IPI00554989)b,c Biglycan (IPI00123194)b,c Insulin-like growth factor–binding protein 4 (IPI00112487)c Heat-shock 70 kDa protein 4 (IPI00331556) Peptidyl-prolyl cis-trans isomerase NIMA-interacting 1 (Pin-1) (IPI00132093) Enolase alpha (IPI00122684) Dickkopf-related protein 3 (IPI00131904) Peroxiredoxin-6 (IPI00555059) Cathepsin B (IPI00113517) Cathepsin L1 (IPI00128154) Calsyntenin-1 isoform 1 (IPI00470000) Macrophage migration inhibitory factor (IPI00230427)
Rab GDP dissociation inhibitor beta Isoform 1 (IPI00122565) Brain acid soluble protein 1 (IPI00129519) Chondroitin sulfate proteoglycan 4 isoform 1 (IPI00128915) Thioredoxin-like protein 1 (IPI00266281) Alpha-mannosidase 2 (IPI00114044) Calreticulin (IPI00123639) 78 kDa glucose-regulated protein (IPI00319992) Heat-shock protein HSP 90-beta (IPI00554929) 40 kDa peptidyl-prolyl cis-trans isomerase (Cyclophilin-40) (IPI00132966) Peptidyl-prolyl cis-trans isomerase FKBP1A (IPI00266899) Carboxypeptidase E (IPI00119152) Coatomer subunit delta (IPI00313558) Cysteine-rich protein 2 (IPI00121319) Dynactin subunit 2 (IPI00116112) Asparaginyl-tRNA synthetase (IPI00223415) Enolase beta (IPI00228548)
(a) TAILS of GluC-digested fibroblast native secretome. The numbers of nonredundant peptides identified from three biological replicates are shown before and after TAILS. TAILS reveals many GluC-generated neo-N-terminal peptides (67%) that were previously masked in the original sample (~6%) by tryptic peptides. 96% of the GluC-generated neo-Nterminal peptides are of high ratio (>3). (b) Procollagen C-endopeptidase enhancer-1 peptides identified before and after TAILS of GluC-digested secretome. Peptide DAVEKESALSPGEDVQR was cleaved at an internal E (bold) in the protease-treated sample, thereby explaining its low abundance in the GluC sample (light (L) singleton). Its cleavage yielded the peptide SALSPGEDVQR found only after TAILS to give a heavy (H) singleton. (c) The most biologically relevant and highest-confidence novel substrates of MMP-2, which are secreted, extracellular matrix or cell membrane proteins that passed the hierarchical substrate winnowing criteria of high-confidence, high-ratio peptides that were identified two or more times. ~30% of the peptides were identified in multiple charge states or multiple forms (methionine oxidation). Alpha-mannosidase 2 was identified after loss of the pyroglutamine N-terminal peptide following MMP-2 cleavage. aQuantifiable
284
internal tryptic peptides with dimethylated K (bold). bCandidate substrates that were biochemically confirmed in in vitro assays. cKnown substrates of other MMPs.
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
letters a
c
Novel MMP-2 substrates 34
Substrates of other MMPs 4
– + 0
+ + 0
+ + 1
+ + 4
+ + 16
50 66
37
45
25
5 pyr-QHLG↓ MTKC 9
KPPR↓ LVGG murine
Galectin-1 + + MMP-11 – +
Galectin-1 15 10
9
GPRM↓ LGAP human
Endoplasmin MMP-11
+ + ++ 4
+ + + 4
ECM-1
75
45
50
35
37
25
25
18.4
+ + – – ++ ++
Endoplasmin 116 66 45 35 25 18
Cyclophilin A + MMP-2 – Time (h) 16
+ ++ 16
Biglycan – + + + MMP-2 ++ – + ++ Time (h) NA 4 4 4
– ++ 16
Cyclophilin A
66 45
Biglycan
35
*
25 18.4
e
100 80 60
LIX (1-92) 9,855 LIX (3-92) 9,689
40 20 0 4,000
1.1E+4
4,841 4,924 7,000 Mass (m/z)
11,000
LIX (3-92) 9,688
100
8,340
80 60
LIX (1-92) 9,858
40 20
Intensity
25 20
– +
QHLG↓ 5MTKC
ECM-1 – + MMP-2 + – Salt NA + Time (h) 0 4
Percent intensity
Fibulin-2 + + + + MMP-2 – – + ++ Time (h) 0 12 12 12 Fibulin-2 150
Intensity
© 2010 Nature America, Inc. All rights reserved.
+ – 0
Sulfated glycoprotein 1 MMP-2 Time (h) 75 Sulfated glycoprotein 1
35 Cleavage product + 1:100 enzyme substrate ratio; ++,1:10 ratio
Cleavage site N-terminal sequencing
d
+ + + + + + + + ++ ++ – – 4 16 0 1 0 16
116
TAILS
Cystatin C
– + ++ + 0 0
proBMP-1
b
Fractalkine
+ – 0
Percent intensity
Reported but unconfirmed MMP substrates 90
Known MMP-2 substrates 20
BMP-1 MMP-2 Time (h)
4,845 4,927
0 4,000
7,000 11,000 Mass (m/z)
Figure 2 Substrates of MMP-2, MMP-11 and in BALF identified by TAILS. (a) The numbers of MMP-2 substrates falling into the four indicated categories were compiled from proteins identified from 288 cleavage sites (146 proteins) or the loss of their acetylated N-terminal peptide by cleavage. (b) Alignment of cleavage sites of mouse fractalkine and cystatin C, two novel mouse substrates of MMP-2 identified by TAILS, with the MMP-2 cleavage sites in human fractalkine and human cystatin C, as determined by Edman sequencing. The positions of the cleavage sites are identical even though the primary sequences of the mouse and human proteins differ slightly. The cyclized original mature protein N-terminal peptide of fractalkine was present as a low-ratio peptide and only after TAILS (Supplementary Results 7). (c) Biochemical validation of MMP-2 substrates identified using TAILS. Candidate MMP-2 substrates were incubated with active human MMP-2 at the molar ratios (+ or ++) and for the times (h) indicated and then analyzed by SDS-PAGE and silver staining. Arrows, proteolytic fragments. ECM-1 proteolytic fragments of 30 and 28 kDa were found when assayed in 251 mM ionic strength (++), whereas a major initial fragment of 44 kDa is generated in 6 mM ionic strength (+). (d) Biochemical validation of MMP-11 substrates identified using TAILS. Recombinant human galectin-1 and endoplasmin were incubated with p-aminophenylmercuric acetate–activated MMP-11. Reaction products were separated on 15% acrylamide Tris-tricine gels and silver stained. (e) MALDI-TOF spectra of in vitro cleavage products of full-length LIX (amino acid residues 1-92) incubated in the absence (black) or presence (red) of DPPIV. The predicted m/z for singly charged LIX (1-92) is 9,854 and for LIX (3-92) is 9,686. Left panel, DPPIV (8 mU) completely processes LIX (1-92) to a 9,689 m/z product, deconvoluted as LIX (3-92) (∆m/z 3); double-charged ions are also evident. Right panel, at 1/6,000 wt/wt DPPIV (8 µU)/LIX, both the full-length (1-92) and DPPIV cleavage product (3-92) of LIX (∆m/z 2) are evident.
glycoprotein-1 and cyclophilin A are substrates of MMP-2 in vitro (Fig. 2c). Although most of these are extracellular matrix proteins found interstitially or in basement membranes, gene ontology analysis of cyclophilin A suggests that it is intracellular. However, in addition to proteins released upon cell death, numerous intracellular proteins in the secretome are now recognized as having bona fide extracellular roles24. For example, whereas extracellular cyclophilin A acts as a chemoattractant and binds immunosuppressants, it acts as a peptidylprolyl cis-trans isomerase A within the cell. It is therefore of interest that the related peptidyl-prolyl cis-trans isomerase FK506-binding protein-1A was also identified as a substrate. These intracellular proteins apparently constitute a new group of MMP substrates in the extracellular compartment. Analysis of MMP-2-cleaved mouse secretomes revealed that 28% of N-terminal peptides were labeled and hence unblocked in the original sample (Fig. 3a). The removal of the N-terminal methionine depended on the amino acid at position 2, with this process being preferentially associated with valine, glycine, alanine and serine residues (Fig. 3b).
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
Acetylation occurred on 731 original mature protein N-terminal peptides but at the initiator methionine in only 153 of these instances. In 578 cases, acetylation was at position 2 in the protein after removal of 1Met, with alanine, serine and methionine being the preferred acetylated residues (Fig. 3c). Although most original mature protein N-terminal peptides had ratios centered on 1.0, a number of N-terminal peptides had a low protease/control ratio, such as the acetylated N-terminal peptide of galectin-1 (Supplementary Results 7). This indicated that loss of these peptides occurred in the protease-treated samples following cleavage in the proteins within these peptide sequences, resulting in a low ratio. Hence, analysis of the full N-terminome by TAILS can provide additional evidence for new substrates as indicated by loss of substrate that is not possible in other approaches. We also identified 132 different cyclized peptides having pyroGlu or 5-oxothiomorpholine-3-carboxylic acid (Supplementary Results 7) derived from cyclization of a N-terminal N-carboxyamidomethyl-cysteine, itself formed after cysteine alkylation by iodoacetamide, as routinely performed in sample preparation
285
letters
Met intact 17%
Unblocked 28%
Signal peptide removed 46% Met intact Propeptide Met 17% removed removed 2% 35%
Met removed 90% Acetylated 49%
Met intact 10%
Unblocked 51%
Signal peptide removed 39% Propeptide Met intact removed 10% Met removed 3% 48%
30 20 10 ACDEGKLMNPQSTV N-terminal amino acid at position 2
60 50 40 30 20 10
60
ACDEGKLMNPQSTV N-terminal amino acid at position 2
60 50 40 30 20 10 0
f
60 50 40 30 20 10 0
i
A CD EGK L MN PQS T V Acetylated N-terminal amino acid
A CD EGK L MN PQS T V Acetylated N-terminal amino acid
60
286
Occurrence (%)
50 50 91% for proteomics analyses. The cyclized mature 40 40 1 N-terminal peptide of fractalkine pyr- QHLG Acetylated Met intact 9% 30 30 53% MTKCEIMCDKMTSR was a low-ratio pepUnblocked 20 20 Signal peptide 47% removed tide, whereas the underlined neopeptide is a 10 10 60% Met 0 0 high-ratio peptide generated by MMP-2 cleavMet intact removed Propeptide A CD EGK L MN PQS T V A CD EGK L MN PQS T V removed 5% 30% 5% age (Fig. 2b and Supplementary Results 7). N-terminal amino acid at position 2 Acetylated N-terminal amino acid Such reciprocal peptide ratios provide very convincing confirmation of cleavage. Cyclized N-termini of many MMP-2 cleavage products were also Nonetheless, although it cannot cleave extracellular matrix comenriched by TAILS (Supplementary Results 7). Twenty-eight of ponents28, its substrates in breast cancer remain unkown. MMP-11 these with labeled lysines were heavy singletons, 17 of which were in correlates with bad prognosis and outcome and its absence is an multiple samples or in multiple states and met all hierarchical sub- important prognostic indicator of disease-free survival29. Therefore, strate winnowing criteria. These include substrates such as plectin-1, our identification of several substrate candidates for MMP-11, of protein SET isoform-1, sulfated glycoprotein-1, endoplasmin, gelsolin which the current cancer drug targets galectin-1 and endoplasmin30 and filamin A. Fifteen were identified by two or more different pep- were biochemically validated, could be of clinical relevance. Galectin-1 tides (Supplementary Results 6). Two new substrates, protein DJ-1 is associated with tumor cell migration, and stimulates inflammaand alpha-mannosidase 2, were only found from cyclized peptides, tion and apoptosis of activated T cells, thereby contributing to tumor with DJ-1 being confirmed as an MMP-2 substrate25. This emphasizes aggressiveness and host evasion31. Hence, loss of function of galectin-1 the need for unbiased peptide selection and comprehensive search by cleavage at multiple sites may contribute to the pleiotropic activistrategies for all forms of peptides, both at the natural N-terminus ties of MMP-11, for example, in abrogating apoptosis of tumor and and on neo-N-termini, to maximize substrate coverage. immune cells induced by galectin-127. We next applied TAILS in more complex settings. Most mature We also performed TAILS on plasma-rich bronchoalveolar lavage protein N-termini in the secretome of an MCF-7 human breast cancer fluid in the context of a mouse model of lipopolysaccharide-induced cell line were unaffected by expression of either recombinant MMP-11 lung inflammation. Among the 855 N-terminal peptides we identified, activated intracellularly by furin or its catalytically inactive mutant many were original mature protein N-termini from lung-specific pro(E216A) (Fig. 3d)(Supplementary Results 8). The distribution of teins and low-abundance inflammatory mediators (Supplementary amino acid residues at the N-terminus after removal of the initiator Results 9). Once again, we observed that only a low percentage of methionine (Fig. 3e) differed from those in the murine secretomes, proteins retained the initiator methionine (Fig. 3g) and 60% displayed perhaps indicating perturbed specificity for N-terminal methionine evidence of cleavage of signal sequences and propeptides (Fig. 3h) removal in human cancer cells. However, like the mouse fibroblast (Supplementary Results 9). Many proteins showed processive proteosecretomes, the distribution for acetylated residues favored alanine lytic processing of the N-termini (ragging) (Supplementary Results 9). (37%), serine (18%) and methionine (17%) (Fig. 3c,f). As with MMP-2, Fifty-three percent of the proteins were acetylated (Fig. 3g,i), which galectin-1 was identified from a low-ratio peptide, indicating deple- is a lower percentage than that found in the secretome analyses tion from the MMP-11–expressing cells. The ability of MMP-11 to (Fig. 3a), where intracellular proteins, which have a higher acetylation digest galectin-1 in vitro was confirmed using a biochemical assay, as content1, accumulate in the culture medium upon cell lysis. Notably, was the likelihood of heat shock protein-90β1 family member endo- 612 dimethylated inflammatory protease-generated cleavage prodplasmin being an in vivo target of MMP-11(Fig. 2). Galectin-1 and ucts were identified (Supplementary Results 9). A broad dynamic endoplasmin are the first cellular substrates to be discovered since range of >6 log orders of magnitude (down to pM concentrations) the cloning of this important, but enigmatic, protease 20 years ago26. was covered, ranging from serum proteins (e.g., fibrinogen, hemoEndoplasmin is another intracellular protein that can be mislocalized globin and apolipoprotein A1) to cytokines CXCL2 and CXCL5/LIX to the extracellular environment, where it promotes tumorigenesis24. (chemokine (C-X-C motif) ligands 2 and 5) present at ~3 ng/ml32 Presumably, colocalization of the two proteins in the extracellular and receptors for macrophage colony-stimulating factor–1, tumor compartment places endoplasmin under the regulation of MMP-11. necrosis factor–α and epidermal growth factor. In addition to the MMP-11 is expressed by the reactive stroma surrounding tumors and N-terminal peptide 1APSSVIAATELR of LIX, a cleaved, activated form promotes tumor establishment by inhibiting cancer cell apoptosis27. identified by the neo-N-terminal peptide 2SSVIAATELR, indicative Occurrence (%)
© 2010 Nature America, Inc. All rights reserved.
Met removed
40
0
h
g
50
0
e
d
c
60
Occurrence (%)
Acetylated 82%
b
Occurrence (%)
Met removed 83%
Occurrence (%)
a
Occurrence (%)
Figure 3 Characterization of original mature protein N-terminal peptides. (a–i) Annotation of the original mature protein N-termini in mouse fibroblast secretomes treated with MMP-2 (a–c), secretomes of human MCF-7 breast carcinoma cells transfected with MMP11 or inactive mutant MMP-11 (E216A) (d–f), and mouse bronchoaleolar lavages collected 24 h after lipopolysaccharide induction of lung inflammation (g–i). (b,e,h) Frequency distribution of amino acids found at the protein N-terminal after initiator methionine removal. (c,f,i) Frequency distribution of naturally acetylated amino acids at the protein N-termini. Met, methionine. Single-letter amino acid coding is used for amino acid occurrences.
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
letters of canonical cleavage by dipeptidyl peptidase IV (DPPIV), was identified in bronchoalveolar lavage fluid (BALF) and validated in vitro (Fig. 2e). As for MMP-8 cleavage of LIX at its N-terminus, cleavage by DPPIV in this instance is also an activating event that stimulates polymorphonuclear neutrophil chemotaxis 33. DPPIV regulates bioactive molecules, including cytokines and chemokines, through the proteolytic processing of dipeptides from the N-termini of proteins, predominantly those with a proline residue at the P1 position. Elevated levels of DPPIV are associated with inflammation, and DPPIV inhibitors have proven useful in treating diabetes. N-terminal positional proteomics has been proposed for MS sample simplification, although proteome-wide coverage has usually been limited5–8. The ability to identify large numbers of proteins in BALF, with abundances spanning a range greater than six orders of magnitude, underscores the potential of TAILS to tackle the dynamic range analysis problem posed by complex proteomes. Our success is in large part attributed to the dendritic HPG-ALD polymers used in TAILS. The major impediment of current dendrimer-based technology is the lack of versatile reagents resulting from the complex and time-consuming synthesis and limitations posed by dendrimer size (typically up to ~60 kDa), and poor solubility, along with the restricted number and type of the functionalities. The polymers we characterized here and used in TAILS lack detectable nonspecific interactions and the aldehyde functionality (~3,200 functional groups/molecule) is the highest reported. The >600 kDa mass difference of the peptides and polymers used greatly improves peptide enrichment over dendrimers34. Finally, the polyglycerol polymer scaffold prepared for aldehyde derivatization is synthesized by a single reaction, unlike dendrimers, which require multistep syntheses and purification. Overall, TAILS achieves broad protease substrate coverage and can readily identify post-translationally modified N-termini in the same analysis by simply changing search parameters. Negative selection has the unique advantage in detecting cleavage products with N-terminal cyclized glutamine, glutamate and cysteine at P1′, as these will be unreactive in positive-selection techniques. Amino acid bias is apparent in a positive selection approach that relies on subtiligase biotinylation of α-amine groups of neo and mature N-terminal peptides10. The subtiligase used in this approach does not recognize proline and shows low preferences for glutamate and aspartate35. Notwithstanding these limitations, no enrichment of any kind results in protease cleavage fragments remaining buried in the background leading to low coverage. This lack of enrichment can be partially overcome by in silico selection22 or in PROTOMAP gels23, which display stable cleavage fragments within the resolution limits of gels, often analyzing multiple peptides from each fragment by MS/MS (Supplementary Discussion). However, the absence of cleavage site information reduces confidence in substrate assignments that resulted in only a 20–25% overlap in PROTOMAP and subtiligase data sets for caspases in Jurkat cell apoptosis10,23,36. TAILS negative selection is also advantageous as the variety of original mature protein N-terminal-blocked peptides each present distinct chemical challenges for their enrichment by positive-selection strategies. We characterized 964 acetylated N-terminal peptides in mouse and human secretomes that were acetylated preferentially on methionine, alanine and serine, as also recently found by others2,11. In addition, we identified 342 cyclized peptides in mouse secretomes, revealing the prevalence of this modification. When analyzing proteases with broad or unknown specificities, approaches without quantification cannot reliably distinguish protease-specific cleavage fragments from background proteolysis products. On the other hand, isotopic labeling
nature biotechnology VOLUME 28 NUMBER 3 MARCH 2010
enables profiling of broad-specificity proteases, as shown for MMP-2, or proteases of unknown specificity, such as MMP-11, where data parsing for known cleavage sites cannot be performed. Finally, as shown by identifying >600 inflammatory proteolytic cleavage peptides in BALF, biologically relevant protease substrates can be detected by TAILS in vivo at physiological concentrations of enzyme and substrate. Methods Methods and any associated references are available in the online version of the paper at http://www.nature.com/naturebiotechnology/. Accession information. All LC-MS/MS data (RAW files) are deposited in the open access public repository Tranche (http://tranche. proteomecommons.org/) as listed in Supplementary Results 10. The polymers for proteomics described here are available through Flintbox (http://www.flintbox.ca/). Note: Supplementary information is available on the Nature Biotechnology website. Acknowledgments O.K. was supported by the Centre for Blood Research (CBR) (University of British Columbia), Canadian Institute for Health Research/Heart and Stroke Foundation of Canada (CIHR/HSFC) Strategic Training Program in Transfusion Science research fellowship. A.D. acknowledges the Fonds Quebecois de la Recherche sur la Nature et les Technologies and the Michael Smith Foundation for Health Research (MSFHR) for research fellowships. U.a.d.K. was supported by a Deutsche Forschungsgemeinschaft (DFG) research fellowship. A.P. acknowledges the support from the CBR Strategic Training Program in Transfusion Science. O.S. acknowledges support from the DFG and the MSFHR. A.E.S. is supported by Natural Sciences and Engineering Research Council of Canada, the MSFHR and CIHR Strategic Training Program STP-53877. L.J.F. is the Canada Research Chair in Organelle Proteomics, a Michael Smith Foundation Scholar and a Peter Wall Institute for Advanced Studies Early Career Scholar. J.N.K. is the recipient of a Canadian Blood Services (CBS)/CIHR new investigator award in transfusion science. C.M.O. is supported by a Canada Research Chair in Metalloproteinase Proteomics and Systems Biology. This work was supported by grants from the CIHR and from a program project grant in Breast Cancer Metastases from the Canadian Breast Cancer Research Alliance with funds from the Canadian Breast Cancer Foundation and The Cancer Research Society as well as with an Infrastructure Grant from the Canada Foundation for Innovation (CFI) and the MSFHR. We thank W. Chen and the UBC Centre for Blood Research Mass Spectrometry Suite (supported by the CFI and the MSFHR) for proteomics analysis, G. Butler for analysis of cyclophilin A cleavage by MMP-2 and V. Goebeler for helpful suggestions regarding endoplasmin assays. D.E. Brooks is thanked for his encouragement and use of facilities. The authors thank the LMB Macromolecular Hub at the UBC Centre for Blood Research for the use of their research facilities, which is supported by the CFI and the MSFHR. MCF7 cells, stably transfected with either full-length human MMP-11 or inactive control (MMP-11 (E216A)) were a kind gift from B. Mari (University of Nice, France). Recombinant human proBMP-1, human fibulin-2, human ECM-1a (truncation of 123 amino acid at the N-terminus), mouse sulfated glycoprotein 1 and DPPIV were kindly provided by S. Walter (Johannes Gutenburg University), T. Sasaki (Oregon Health and Science University), J. Merregaert (University of Antwerp), S. Koochekpour (LSU-Health Sciences Center) and C. McIntosh (University of British Columbia), respectively. AUTHOR CONTRIBUTIONS O.K. participated in the project design, developed TAILS, performed proof of principal and MMP-2 experiments, did the bioinformatics data analysis and drafted the manuscript. A.D. participated in the project design, performed TAILS on GluC, MMP-2 and BALF and in vitro assays and revised the manuscript. U.a.d.K. and A.P. participated in the project design, performed the MMP-11 analyses and revised the manuscript. O.S. participated in the project design and the bioinformatics data analysis. A.S. performed the LIX and DPPIV experiments. R.K.K. produced the HPG-ALD polymers. L.J.F. performed the mass spectrometry analyses on his Orbitrap mass spectrometer. J.N.K. is the senior author for the HPG-ALD chemistry and synthesis. He participated in the project design, engineered the HPG-ALD polymer series and participated in the manuscript writing. C.M.O. is the overall senior author for the degradomics and proteomics
287
letters section of the manuscript. He conceived the project and design and was responsible for project supervision, data interpretation and manuscript writing and provided grant support. COMPETING INTERESTS STATEMENT The authors declare no competing financial interests.
© 2010 Nature America, Inc. All rights reserved.
Published online at http://www.nature.com/naturebiotechnology/. Reprints and permissions information is available online at http://npg.nature.com/ reprintsandpermissions/.
1. Brown, J.L. & Roberts, W.K. Evidence that approximately eighty per cent of the soluble proteins from Ehrlich ascites cells are Nalpha-acetylated. J. Biol. Chem. 251, 1009–1014 (1976). 2. Dormeyer, W., Mohammed, S., Breukelen, B., Krijgsveld, J. & Heck, A.J. Targeted analysis of protein termini. J. Proteome Res. 6, 4634–4645 (2007). 3. Doucet, A., Butler, G.S., Rodriguez, D., Prudova, A. & Overall, C.M. Metadegradomics: toward in vivo quantitative degradomics of proteolytic post-translational modifications of the cancer proteome. Mol. Cell. Proteomics 7, 1925–1951 (2008). 4. Gevaert, K. et al. Exploring proteomes and analyzing protein processing by mass spectrometric identification of sorted N-terminal peptides. Nat. Biotechnol. 21, 566–569 (2003). 5. Kuhn, K. et al. Isolation of N-terminal protein sequence tags from cyanogen bromide cleaved proteins as a novel approach to investigate hydrophobic proteins. J. Proteome Res. 2, 598–609 (2003). 6. McDonald, L., Robertson, D.H., Hurst, J.L. & Beynon, R.J. Positional proteomics: selective recovery and analysis of N-terminal proteolytic peptides. Nat. Methods 2, 955–957 (2005). 7. McDonald, L. & Beynon, R.J. Positional proteomics: preparation of amino-terminal peptides as a strategy for proteome simplification and characterization. Nat. Protoc. 1, 1790–1798 (2006). 8. Ji, C., Guo, N. & Li, L. Differential dimethyl labeling of N-termini of peptides after guanidination for proteome analysis. J. Proteome Res. 4, 2099–2108 (2005). 9. Timmer, J.C. et al. Profiling constitutive proteolytic events in vivo. Biochem. J. 407, 41–48 (2007). 10. Mahrus, S. et al. Global sequencing of proteolytic cleavage sites in apoptosis by specific labeling of protein N termini. Cell 134, 866–876 (2008). 11. Staes, A. et al. Improved recovery of proteome-informative, protein N-terminal peptides by combined fractional diagonal chromatography (COFRADIC). Proteomics 8, 1362–1370 (2008). 12. Gupta, N. & Pevzner, P.A. False discovery rates of protein identifications: a strike against the two-peptide rule. J. Proteome Res. 8, 4173–4181 (2009). 13. Hsu, J.L., Huang, S.Y., Chow, N.H. & Chen, S.H. Stable-isotope dimethyl labeling for quantitative proteomics. Anal. Chem. 75, 6843–6852 (2003). 14. Kainthan, R., Muliawan, E., Hatzikiriakos, S. & Brooks, D. Synthesis, characterization, and viscoelastic properties of high molecular weight hyperbranched polyglycerols. Macromolecules 39, 7708–7717 (2006). 15. Van Damme, P. et al. Caspase-specific and nonspecific in vivo protein processing during Fas-induced apoptosis. Nat. Methods 2, 771–777 (2005). 16. Keller, A., Eng, J., Zhang, N., Li, X.J. & Aebersold, R. A uniform proteomics MS/MS analysis platform utilizing open XML file formats. Mol. Syst. Biol. 1, 2005.0017 (2005).
288
17. Minn, A.J. et al. Genes that mediate breast cancer metastasis to lung. Nature 436, 518–524 (2005). 18. Dean, R.A. & Overall, C.M. Proteomics discovery of metalloproteinase substrates in the cellular context by iTRAQ labeling reveals a diverse MMP-2 substrate degradome. Mol. Cell. Proteomics 6, 611–623 (2007). 19. Dean, R.A. et al. Identification of candidate angiogenic inhibitors processed by matrix metalloproteinase 2 (MMP-2) in cell based proteomic screens: disruption of vascular endothelial growth factor (VEGF)/heparin affin regulatory peptide (pleiotrophin) and VEGF/connective tissue growth factor angiogenic inhibitory complexes by MMP-2 proteolysis. Mol. Cell. Biol. 27, 8454–8465 (2007). 20. McQuibban, G.A. et al. Inflammation dampened by gelatinase A cleavage of monocyte chemoattractant protein-3. Science 289, 1202–1206 (2000). 21. Schilling, O. & Overall, C.M. Proteome-derived, database-searchable peptide libraries for identifying protease cleavage sites. Nat. Biotechnol. 26, 685–694 (2008). 22. Enoksson, M. et al. Identification of proteolytic cleavage sites by quantitative proteomics. J. Proteome Res. 6, 2850–2858 (2007). 23. Dix, M.M., Simon, G.M. & Cravatt, B.F. Global mapping of the topography and magnitude of proteolytic events in apoptosis. Cell 134, 679–691 (2008). 24. Butler, G.S. & Overall, C.M. Proteomic identification of multitasking proteins in unexpected locations complicates drug targeting. Nat. Rev. Drug Discov. 8, 935–948 (2009). 25. Butler, G.S., Dean, R.A., Tam, E.M. & Overall, C.M. Pharmacoproteomics of a metalloproteinase hydroxamate inhibitor in breast cancer cells: Dynamics of matrix metalloproteinase-14 (MT1-MMP) mediated membrane protein shedding. Mol. Cell. Biol. 28, 4896–4914 (2008). 26. Basset, P. et al. A novel metalloproteinase gene specifically expressed in stromal cells of breast carcinomas. Nature 348, 699–704 (1990). 27. Andarawewa, K.L. et al. Dual stromelysin-3 function during natural mouse mammary tumor virus-ras tumor progression. Cancer Res. 63, 5844–5849 (2003). 28. Pei, D., Majmudar, G. & Weiss, S.J. Hydrolytic inactivation of a breast carcinoma cell-derived serpin by human stromelysin-3. J. Biol. Chem. 269, 25849–25855 (1994). 29. Ahmad, A. et al. Stromelysin 3: an independent prognostic factor for relapse-free survival in node-positive breast cancer and demonstration of novel breast carcinoma cell expression. Am. J. Pathol. 152, 721–728 (1998). 30. Fu, Y. & Lee, A.S. Glucose regulated proteins in cancer progression, drug resistance and immunotherapy. Cancer Biol. Ther. 5, 741–744 (2006). 31. He, J. & Baum, L.G. Presentation of galectin-1 by extracellular matrix triggers T cell death. J. Biol. Chem. 279, 4705–4712 (2004). 32. Tateda, K. et al. Chemokine-dependent neutrophil recruitment in a murine model of Legionella pneumonia: potential role of neutrophils as immunoregulatory cells. Infect. Immun. 69, 2017–2024 (2001). 33. Tester, A.M. et al. LPS responsiveness and neutrophil chemotaxis in vivo require PMN MMP-8 activity. PLoS One 2, e312 (2007). 34. Tao, W.A. et al. Quantitative phosphoproteome analysis using a dendrimer conjugation chemistry and tandem mass spectrometry. Nat. Methods 2, 591–598 (2005). 35. Chang, T.K., Jackson, D.Y., Burnier, J.P. & Wells, J.A. Subtiligase: a tool for semisynthesis of proteins. Proc. Natl. Acad. Sci. USA 91, 12544–12548 (1994). 36. Simon, G.M., Dix, M.M. & Cravatt, B.F. Comparative assessment of large-scale proteomic studies of apoptotic proteolysis. ACS Chem. Biol. 4, 401–408 (2009).
VOLUME 28 NUMBER 3 MARCH 2010 nature biotechnology
ONLINE METHODS
© 2010 Nature America, Inc. All rights reserved.
Polymer synthesis and characterization. The dendritic HPG-ALD polymers were synthesized by a periodate oxidation of high molecular weight HPGs in water14. For a typical reaction high molecular weight HPG-670 kDa (1 g) was dissolved in 10 ml of ddH2O, followed by the addition of 0.9 g of HIO4 (HPG-ALD-IV). The solution was stirred for 3 h at 22 °C and dialyzed against ddH2O using 1,000 MWCO membrane for 48 h with 8 changes of ddH2O. The aldehyde content in the polymer was manipulated by changing the ratio of HIO4 to 1,2-diols used during the synthesis. Polymer concentration in the solution was determined by drying 250 µl of polymer solution at 75 °C for 72 h and the dry weight of the material was determined. HPG-ALD polymers were characterized by both 1H NMR spectroscopy (in D2O) and aldehyde content determination37 (Supplementary Results 1). Polymer binding capacity. Polymer capture efficiency and capacity were determined by titration with an unblocked synthetic peptide analog of the neo-N-terminus of human chemokine CCL7 produced by MMP-2 cleavage20 (NH2-INTSTTCCYR-COOH, m/z = 1160.50). The aldehyde groups on the polymer were reacted with the N-terminal α-amino groups of the peptide by an aldamine reaction and the resulting Schiff base was reduced to a secondary amine with NaCNBH338,39. HPG-ALD polymers (12.5–25 µg) were reacted with 25-80 µg of peptide in 50 µl of 50 mM HEPES buffer pH 7.0, 20 mM NaCNBH3 for 16 h, 37 °C. HPG-ALD III (100 µg) was reacted with 0–300 µg of peptide in the same conditions. Separation of polymers and unbound peptide was achieved using Microcon spin filter devices with a 10-kDa cutoff (Millipore). The filtrate flow-through was collected and peptide titration performed in 96-well plates by mixing 20 µl of peptide solution and 200 µl of O-phthalaldehyde solution (Pierce). The fluorophore was excited at 320 nm and emission measured at 460 nm for 20 min using a BMG PolarStar fluori meter. For additional details see Supplementary Results 1. Determination of aldehyde content. The different HPG-ALD polymers (50–300 µg) were reacted with 50 mM 2,4-dinitrophenylhydrazine (DNP) in 400 µl solution containing 850 mM H2SO4 and 4.5% ethanol for 30 min at 22 °C. The insoluble polymer-DNP complex was centrifuged for 10 min at 17,000g and the supernatant was discarded. Pellets were solubilized in DMSO and the absorbance was measured at 360 nm. Formaldehyde-DNP complex solutions ranging from 0.05 to 1 mM were used as standards for the titration. For additional details see Supplementary Results 1. Peptide binding capacity. HPG-ALD-II and V (12.5 µg) or HPG-ALD-I and IV (25 µg) were reacted with 25–80 µg of peptide A (NH2-INTSTTCCYRCOOH) in 50 µl of 50 mM HEPES buffer pH 7.0, 20 mM NaCNBH3 for 16 h, 37 °C. HPG-ALD-III (100 µg) was reacted with 0–300 µg of peptide in the same conditions. Separation of polymers and unbound peptide is achieved using Microcon spin filter devices with a 10-kDa cut-off. The filtrate flow-through was collected and peptide titration performed in 96-well plates by mixing 20 µl of peptide solution and 200 µl of O-phthalaldehyde solution. Fluorophore was excited at 320 nm and emission measured at 460 nm for 20 min using a BMG PolarStar fluorimeter. For additional details see Supplementary Results 1.
were grown in Dulbecco’s modified Eagle’s medium, 10% cosmic calf serum, 2 mM l-glutamine, 70 mM Xanthine, 1× HT supplement (100 µM sodium hypoxanthine and 16 µM thymidine) (Invitrogen). At 70% confluency the cells were washed extensively with PBS to remove serum proteins and grown overnight serum free. Cells were then washed again and incubated in serum-free, phenol-free Dulbecco’s modified Eagle’s medium. MCF7 cells, stably transfected with either full-length human MMP-11 or inactive control (MMP-11 (E216A)) were a kind gift from B. Mari. Cells were grown in Dulbecco’s modified Eagle’s medium, 10% fetal calf serum, 2 mM l-glutamine, 10,000 U/ml penicillin, 10,000 µg/ml streptomycin and 0.8 mg/ml G418. MCF7 transfectants were grown to 80% confluency, washed extensively with PBS and incubated in serum-free, phenol-free DMEM. Secretome collection and concentration. Mmp2 −/− fibroblast–conditioned medium proteins (secretome) were collected at 24 or 48 h with protease inhibitors (1 mM EDTA, 1 mM PMSF) immediately added. The proteins were clarified by centrifugation (5 min, 500g), filtration (0.22 µm) and additional centrifugation (30 min, 8,000g). The proteins were concentrated ×100 by ultrafiltration using Amicon Ultra-15 centrifugal filter units (3 kDa cut-off, Millipore). The sample buffer was exchanged to 50 mM HEPES, 150 mM NaCl, 10 mM CaCl2 by five cycles of dilution and concentration within the same concentrating device. Protein concentration was determined by BCA assay (Pierce) and Bradford assay (BioRad). MCF7 secretome was collected at 40 h with protease inhibitors (1 mM EDTA, 1 mM PMSF) immediately added. The proteins were clarified as described above. The conditioned medium was acidified with 0.4% trifluoroacetic acid vol/vol and applied to a C4 solid phase extraction cartridge (VYDAC) and washed with 5% acetonitrile, 0.1% trifluoroacetic acid. Concentrated proteins were eluted with 75% acetonitrile, 0.1% trifluoroacetic acid. Samples were dried under vacuum and resuspended in 100 mM HEPES buffer, pH 7.5. Protein concentration was determined by Bradford assay. Bronchioalveolar fluid collection. Lung inflammation was induced in mice by intranasal instillation of 10 µg LPS in sterile saline. Mice were euthanized by an overdose of pentobarbital 24 h later and their lungs were lavaged six times with 500 µl of sterile PBS. Cells were removed by centrifugation (2,000g, 15 min) and serum albumin was depleted with a QProteome albumin depletion kit (Qiagen). Proteins were then precipitated with 15% trichloroacetic acid, resuspended in 0.5 M guanidine hydrochloride, 100 mM HEPES, pH 7.0 and the protein concentration was measured by bicinconinic acid assay (Pierce). Preparation of GluC-digested secretomes. Three different preparations of human fibroblast secretome were each split into halves. One-half was partly digested by GluC (endoproteinase derived from Staphylococcus aureus protease V8; Sigma) (1:100 wt/wt) under native conditions (Supplementary Results 3) in ammonium bicarbonate buffer pH 7.8 to favor hydrolysis after glutamoyl bonds and then labeled with d(2),13C-formaldehyde (heavy). The other half was incubated with buffer alone (control) and labeled with d(0),12Cformaldehyde (light). The heavy- and light-labeled samples of each proteome preparation were mixed and digested with trypsin. Each sample was analyzed by MS/MS before and after TAILS negative selection using HPG-ALD-III polymer.
CCL7 N-terminal peptide enrichment. All steps were conducted in 100 mM HEPES buffer. Synthetic human chemokine CCL7 (5 µg) presenting a N-terminal pyroglutamate14 was reduced and alkylated. CCL7 was digested with trypsin (Promega) in a ratio 1:100 wt/wt at pH 7.8 for 16 h, 37 °C. Half of the solution was incubated with HPG-ALD-V in a fivefold excess with 20 mM NaCNBH3 at pH 7.0 for 16 h, 37 °C. Unbound peptides were separated as described above. Peptide samples were desalted, subjected to MS analysis and peptide mass fingerprinting was performed using Aldente proteomic tools from Expasy 40 (http://ca.expasy.org/tools/aldente) (version May 10th 2007) using cysteine carboxyamidomethylation and N-terminal pyroglutamate modifications. Proteome data analysis was conducted against the UniProtKB/SwissProt database (Release 52.4) restricted to the Homo sapiens taxon using 0.2 Da as the maximum spectrometer mass shift and trypsin as the digestive enzyme.
Preparation of MMP-2-digested secretomes. Recombinant proMMP-2 was expressed and purified41 and activated with p-aminophenylmercuric acetate (APMA) (1 mM, 30 min). A mixture of six proteins (Supplementary Results 2), including the known MMP-2 substrate, CCL7 (100 pmol) were incubated with sufficient enzyme (1:100 enzyme/protein) to ensure adequate cleavage over 16 h, 37 °C for experimental proof of concept of TAILS. Control samples were from the same batch incubated without enzyme. CCL7 (100 pmol) was also added to two samples (100 µg) of concentrated conditioned medium protein before incubation. In other experiments, CCL7 (5.4 µg) was first incubated with 100 ng of active MMP-2 in 100 mM HEPES, 200 mM NaCl, 20 mM CaCl2, pH 7.8 for 16 h, 37 °C and spiked into the samples at different amounts. CCL7 cleavage was confirmed by MALDI-TOF MS.
Cell lines and cells cultures. Immortalized Mmp2−/− fibroblasts were selected and maintained as previously described18,41. For proteomics analyses, the cells
Dimethylation by formaldehyde. Protein (100 µg) from protease-exposed samples and controls were denatured by addition of guanidinium hydrochloride
doi:10.1038/nbt.1611
nature biotechnology
(2–4 M final) and reduced with 5 mM DTT at 65 °C for 30 min and alkylated with 12.5 mM iodoacetamide for 2 h in the dark. Any remaining proteases are inactivated by this treatment. Iodoacetamide was then quenched by addition of 30 mM DTT and the pH of the samples was adjusted to between 6 and 7 with 1 N HCl, which ensures that the α-amine N-terminus is charged. The samples were amine-labeled by addition of 20 mM isotope-containing heavy (d(2)C13) (Cambridge Isotopes) or light (d(0)C12) formaldehyde in the presence of 10 mM NaCNBH3 (ALD reagent, Sterogene Bioseparations) for 16 h, 37 °C. Labeling reagents were quenched before tryptic digest by addition of ammonium bicarbonate 100 mM.
© 2010 Nature America, Inc. All rights reserved.
Tryptic digestion. Formaldehyde amine-labeled sample pairs were combined (protease treated and control) and after acetone precipitation overnight at −80 °C were resuspended in 100 µl of 50 mM HEPES pH 8.5. The proteins were digested for 18 h at 37 °C with proteomics grade modified trypsin (Promega) (1:30). Digests were desalted by Sep-Pak C18 (Waters) or C18 OMIX tips (Varian) before MS analysis. Amine-terminal blocked peptide enrichment. Trypsin-digested mixtures of heavy and light amine-labeled proteins were reacted with HPG-ALD polymer HPG-ALD II or III (at a concentration five times its binding capacity) in solution containing 50 mM HEPES pH 7.0 and 20 mM NaBH3CN for 16 h, 37 °C. Following incubation, unreacted aldehyde groups on the polymer were blocked by addition of 100 mM ammonium bicarbonate or glycine for 10 min. Separation of polymers and N-terminal blocked peptides (unbound) was achieved using Microcon filter devices with a 10-kDa cut-off. The peptidespolymer mixtures were filtered through Microcon spin filters followed by one 50-µl wash with 50 mM HEPES. Elution and wash samples were pooled together and desalted using Sep-Pak C18. Liquid chromatography-MS/MS and data analysis. Peptide samples were desalted on STop And Go Extraction tips42 then separated on a 15-cm long, 75-µm inner diameter fused silica emitter (8 µm diameter opening, pulled on a P-2000 laser puller (Sutter Instruments)) packed with 3-µm diameter ReproSil Pur C18 beads on an Agilent 1100 Series nanoflow HPLC in-line with an LTQ-Orbitrap (ThermoFisher Scientific) as described43. The peptides were then injected in the mass spectrometer with a nanospray ionization source (Proxeon Biosystems). Buffer A consisted of 0.5% acetic acid, and buffer B consisted of 0.5% acetic acid and 80% acetonitrile. Gradients were run from 6% B to 30% B over 60 min, then 30% B to 80% B in the next 10 min, held at 80% B for 5 min, and then dropped to 6% B for another 15 min to recondition the column. The LTQ-Orbitrap was set to acquire a full-range scan at 60,000 resolution from 350 to 1500 Th in the Orbitrap and to simultaneously fragment the top five peptide ions in each cycle in the LTQ. Because Orbitrap data were to be used for quantification, blank gradients where buffer B was injected were interspersed between analytical gradients to eliminate carryover. Orbitrap RAW data files were converted to mzXML in profile mode using the TPP from the Systems Biology Institute in Seattle16. For Mascot searches mzXML files were processed to Mascot generic format using the mz2search program of the TPP. These peak lists were searched against Mouse IPI protein database version 3.24 (Release date 12/2006) composed of 52,326 predicted proteins using Mascot version 2.1 or 2.2 (Matrix Science) and X! Tandem 2007.07.01 release with the k-score plug-in. This database was supplemented with the protein sequences of the following proteins: human CCL7, BSA, chicken lysozyme, chicken ovalbumin, horse myoglobin and human MMP-2. Mascot and X! Tandem searches for Orbitrap data of heavy- and light-labeled peptides were performed separately (one search for light-labeled and one for heavy-labeled peptides) using the following criteria: cleavage specificity was set for one Pseudomonas aeruginosa arginyl peptidase (ArgC) cleavage at the termini of the peptides (semi-ArgC), with up to three missed cleavages, cysteine carbamidomethyl, peptide N-terminal and lysine reductive dimethylation (both set as heavy or light) were set as fixed modifications, methionine oxidation was set as a variable modification, peptide tolerance and MS/MS tolerance were set at 10 p.p.m. (peptides only further considered if <5 p.p.m.) and 0.8 Da, respectively, and the scoring scheme used was ESI-TRAP. Similar criteria were used for mapping all proteins in samples before TAILS peptide enrichment except that peptide N-terminal reductive dimethylation (heavy or light) was
nature biotechnology
set as a variable modification. Searches for N-terminal acetylated or cyclized N-terminal peptides were performed in a similar manner by substituting peptide N-terminal reductive dimethylation with acetylation or searched where peptide N-terminal glutamine and glutamate are converted to pyroglutamate and N-terminal cysteines to 5-oxothiomorpholine-3-carboxylic acid. Nonspecific interaction measurement. BSA was denatured, reduced and alkylated as described. The sample was then acetone precipitated and the pellet washed and resuspended in 100 mM HEPES, pH 8.0. Trypsin gold was added (ratio 1:50, wt/wt) and the sample was incubated for 16 h, 37 °C. The peptide solution was split in two and each half was labeled using either heavy or light formaldehyde. The peptides were cleaned by reversed-phase chromatography (Waters C18 Sep-pak) and acetonitrile was removed by Speed-Vac evaporation. Heavy-labeled peptides (20 µg) were incubated with 70 µg of HPG-ALD polymer in 100 mM HEPES pH 7.0 in a final volume of 50 µl for 2 h, 37 °C. Light-labeled peptides (20 µg) were incubated in the same conditions, but without polymer. The solutions were filtered using a spin-filter with a cut-off of 10 kDa. Peptides were recovered in the filtrate whereas polymer was retained on the filter. The filter was washed three times with 50 µl 100 mM HEPES, pH 7.0, three times with 50 µl of 1 M NaCl and once with 20% acetonitrile. Equal volumes of the filtrates of each step coming from the heavy- and lightlabeled peptide samples were pooled and analyzed by MALDI-TOF MS using Voyager-DE STR (Applied Biosystems) in reflector mode. Peptide abundance ratio. Data were analyzed using the TPP. Quantitative information was obtained for the matched peptides by reconstructing individual ion chromatograms of the two isotopic species and integrating the outline of the respective peaks using either the XPRESS44 or ASAPRatio software tool45. If a peptide was submitted to collision-induced dissociation in more than one charge state these ratios were not combined, but analyzed separately in order to obtain quantification data for the same peptide in different charge states for hierarchical substrate winnowing. Biochemical validation of substrates. Recombinant human proBMP-1, human fibulin-2, human ECM-1a (truncation of 123 amino acid at the N-terminus) and mouse sulfated glycoprotein 1 were kindly provided by S. Walter, T. Sasaki, J. Merregaert and S. Koochekpour, respectively. Recombinant human galectin-1 was from Research Diagnostics Inc. Recombinant endoplasmin (Grp94 Recombinant Canine Protein) was from Assay Designs. Recombinant cyclophilin A was from BIOMOL. MMP-2 was activated by 1 mM APMA for 15 min before substrate was added. Porcine kidney DPPIV (Sigma) was generously provided by C. McIntosh. Synthetic LIX was prepared and characterized as described33. BMP-1 (3.7 µM) was incubated with MMP-2 (37 and 370 nM) in assay buffer: 50 mM HEPES, 100 mM NaCl, 10 mM CaCl2, 0.015% Brij-35, pH 7.8 at 37 °C for up to 16 h. Digests were electrophoresed on a 7.5% acrylamide SDSPAGE and stained with silver nitrate. Fibulin-2 (670 nM) was incubated with MMP-2 in assay buffer at 37 °C, 12 h. Samples (1 µg) were analyzed by 10% acrylamide SDS-PAGE. ECM-1 (2 µM) was incubated with 100 nM MMP-2 in Tris-assay buffer (50 mM Tris, 200 mM NaCl, 1 mM CaCl2, 0.01% Brij-35, pH 7.4) at 37 °C, 4 h or in low-salt buffer (5 mM Tris, 1 mM CaCl2, 0.01% Brij-35, pH 7.4)46 at 37 °C, 4 h and analyzed by 12% SDS-PAGE. Biglycan (5 µM) was incubated with 50 and 500 nM MMP-2 in assay buffer at 37 °C for 4 h and 1 µg of protein/sample was analyzed on a 12% SDS-PAGE. Cyclophilin A was incubated at 37 °C, 18 h with 1:10 molar ratio MMP-2. Products were electrophoresed on 12.5% Tris-tricine gels and silver stained. Sulfated glycoprotein 1 (1.67 µM) was incubated with 16.7 and 167 nM MMP-2 at 37 °C for 0 to 16 h in assay buffer and analyzed on a 12% SDS-PAGE. For MMP-11 assays, galectin-1 was incubated with APMA-activated catalytic domain of MMP-11 in 50 mM HEPES, 100 mM NaCl, 10 mM CaCl 2, 0.05% Brij-35, pH 7.4 at 37 °C, 18 h. Reaction products (1 µg) were separated on a 15% acrylamide Tris-tricine gel and visualized by silver staining. Endoplasmin was incubated in Tris-assay buffer with APMA-activated catalytic domain of MMP-11 and analyzed on a 12% Tris-glycine gel and visualized by silver staining. LIX chemokine cleavage by DPPIV and MALDI-TOF analysis of products were performed as described47. LIX (1-92) (1 µg) was incubated with 0–8 mU DPPIV in 100 mM Tris, pH 8.0 at 37 °C, 16 h; cleavage products
doi:10.1038/nbt.1611
were analyzed in SPA matrix by MALDI-TOF mass spectrometry on a Voyager-DE STR. The specific activity of the DPPIV was 49 mU/µg protein where 1 unit is defined as the rate of appearance (µmole/min) of p-nitroaniline from Gly-l-Pro p-nitroanilide in 100 mM Tris-HCl, pH 8.0 at 37 °C. A detailed protocol for the TAILS approach is available in ref. 48.
© 2010 Nature America, Inc. All rights reserved.
37. LeDissez, C., Wong, P. & Brooks, D. Analysis of surface aldehyde functions on surfactant-free polystyrene/polyacrolein latex. Macromolecules 29, 953–959 (1996). 38. Leitner, A. & Lindner, W. Chemistry meets proteomics: the use of chemical tagging reactions for MS-based proteomics. Proteomics 6, 5418–5434 (2006). 39. Veh, R., Corfield, A., Sander, M. & Schauer, R. Neuraminic acid-specific modifications and tritium labeling of gangliosides. Biochim. Biophys. Acta 486, 145–160 (1976). 40. Gasteiger, E. et al. Protein identification and analysis tools on the ExPASy server. in The Proteomics Protocols Handbook (ed. Walker, J.M.) 571–607 (Humana Press, Totowa, New Jersey, USA, 2005). 41. Butler, G.S., Tam, E.M. & Overall, C.M. The canonical methionine 392 of matrix metalloproteinase 2 (gelatinase A) is not required for catalytic efficiency or structural integrity: probing the role of the methionine-turn in the metzincin metalloprotease superfamily. J. Biol. Chem. 279, 15615–15620 (2004).
42. Rappsilber, J., Ishihama, Y. & Mann, M. Stop and go extraction tips for matrixassisted laser desorption/ionization, nanoelectrospray, and LC/MS sample pretreatment in proteomics. Anal. Chem. 75, 663–670 (2003). 43. Chan, Q.W., Howes, C.G. & Foster, L.J. Quantitative comparison of caste differences in honeybee hemolymph. Mol. Cell. Proteomics 5, 2252–2262 (2006). 44. Han, D.K., Eng, J., Zhou, H. & Aebersold, R. Quantitative profiling of differentiationinduced microsomal proteins using isotope-coded affinity tags and mass spectrometry. Nat. Biotechnol. 19, 946–951 (2001). 45. Li, X.J., Zhang, H., Ranish, J.A. & Aebersold, R. Automated statistical analysis of protein abundance ratios from data generated by stable-isotope dilution and tandem mass spectrometry. Anal. Chem. 75, 6648–6657 (2003). 46. Fujimoto, N. et al. Extracellular matrix protein 1 inhibits the activity of matrix metalloproteinase 9 through high-affinity protein/protein interactions. Exp. Dermatol. 15, 300–307 (2006). 47. Starr, A.E. & Overall, C.M. Chapter 13. Characterizing proteolytic processing of chemokines by mass spectrometry, biochemistry, neo-epitope antibodies and functional assays. Methods Enzymol. 461, 281–307 (2009). 48. Kleifeld, O., Doucet, A., Kizhakkedathu, J. & Overall, C.M. System-wide proteomic identification of protease cleavage products by terminal amine isotopic labeling of substrates. Nat. Protoc. advance online publication, doi:10.1038/nprot.2010.30 (7 March 2010).
doi:10.1038/nbt.1611
nature biotechnology
careers and recruitment
The importance of foreign-born talent for US innovation Yeonji No & John P Walsh
© 2010 Nature America, Inc. All rights reserved.
A survey suggests that foreign-born scientists and engineers play a major role in scientific and innovation output in the United States.
A
s noted in a recent Nature editorial, the United States has traditionally relied heavily on immigrants to complement existing science and engineering talent1. Recent debates on both immigration and innovation policy have brought renewed interest in the role of non-US born science and engineering (S&E) personnel in the American innovation system2–5. As the economy has worsened, there have been increasing calls to restrict immigrant workers, including high-skilled S&E workers. For example, the Grassley-Sanders amendment in the American Recovery and Reinvestment Act restricts firms receiving stimulus funding from hiring immigrants on H-1B visas for one year. The Export Administration Regulations limit the research fields and information access of non-US scientists. The PATRIOT Act of 2001, the Enhanced Border Security and Visa Entry Reform Act of 2002 and the Visa Mantis program have created a “chilly climate” for the non-US born. H-1B visas for high-skilled workers and international applications to American graduate programs decreased significantly after 2001 (refs. 6,7). These policies are causing concerns about possible adverse effects on America’s competitiveness because of the significant contributions foreign-born S&E workers make to US science and technology3–5,8,9. Employers in high-tech industries such as the software, computer science and engineering sectors have been lobbying to have the cap on H-1B visas raised. The recent extension of the period of Optional Practical Training for international students is an attempt to respond to these concerns. The editorial in Nature calls for streamlining the visa process and reducing barriers to entry1. Given concerns about the economic downturn and the importance of innovation to Yeonji No and John P. Walsh are at the Georgia Institute of Technology, Atlanta, Georgia, USA. e-mail: [email protected]
economic growth, it is crucial to get a more detailed understanding of the role of non-US talent in contributing to technological advance and commercialization of inventions. Recent work suggests that entrepreneurs born outside of the US are disproportionately represented among high-tech startups3,10. And there is some evidence that they generate a disproportionate number of patents4,11. However, until now, there have been few systematic data tracing the commercialization outcomes of a large sample of patents or papers, although recent studies are efforts in that direction4,12. To further our understanding of the impact of the foreign born on the US innovation system, we used a unique data set based on a nationally representative survey of over 1,900 US-based inventors on ‘triadic’ patents (those granted in the United States and also applied for in Japan and Europe) to examine the percentage of foreign-born inventors, their country of origin and the relations between country of origin (United States or other) and our various measures of patent quality (Supplementary Methods). We focus across the innovative landscape, not limiting our analysis to high-tech sectors or to particular industries such as information technology (IT). For example, our data include a significant number of biotech, biomedical and related inventions (266, or about 14% of the sample)—representing fields have received less attention in this debate than have IT and other engineering-related fields. Our data allow us to combine the inventor-, project- and company-level information from the survey with invention-level information on a particular named patent to examine the value and outcomes of each invention. Therefore, unlike in prior work, we can estimate patent-level models of invention commercialization, controlling for detailed field-, firm-, inventor- and project-level characteristics, which allows us to specify more accurately the differences between US- and foreign-born inventors in regard to the
nature biotechnology volume 28 number 3 march 2010
likelihood of their inventions becoming commercially available innovations. We also have multiple measures of patent quality, including self-reports on whether the patent ranks in the top 10% either in terms of technical significance in its field or in terms of its economic value, and the number of forward citations. Our data can also distinguish high-skilled (PhD-level) and less-skilled immigrants, a distinction that has important implications for immigration policy. Finally, we can highlight biomedical inventions to investigate the contribution of the non-US born to advancing biotech and related industries. These new results complement and move beyond previous work looking at immigrant entrepreneurs and comparing different categories of immigrants (and native born) on overall measures of patent productivity3–5. One limitation of our data set is that we are asking a representative inventor to respond on behalf of the whole project team. About 70% of the inventions have multiple inventors (with a mean of 2.7 inventors). Our sampling strategy targets the first-listed US inventor. In over 95% of the cases, this is the first inventor (and in 27% of the cases the only inventor). Thus, we are interpreting our sample as representing the ‘lead inventor’ on the patent and are arguing that this lead inventor has significant influence on the outcome of the project, such that the lead inventor’s characteristics (including country of origin, education background, firm size) can be used to characterize the invention. For example, we will code patents as being generated by foreign-born inventors if the lead inventor (respondent) is foreign born. There may be additional US-born inventors listed on the patent (and vice versa for patents coded as being by US-born inventors), which adds some measurement error to our classification. A check of subsamples limited to first inventors and solo inventors shows that the descriptive statistics and the distinctions between USand foreign-born individuals are robust with
289
ca r e e r s an d r e c r u i tm e nt 2% 2%
1% 1%
3% 4% 5%
11%
© 2010 Nature America, Inc. All rights reserved.
71%
United States Others China India UK
Taiwan Canada Germany Russian Federation
Figure 1 Country of origin for triadic patent inventors, 2000–2003.
respect to these differing subsets of representative inventors (Supplementary Methods). Results We find that almost 30% of lead inventors are non-US born, compared to only about 11% of the overall US population and about 22% of the college-educated S&E workforce7,13. Among biotech-related inventions, 31% of the lead inventors were foreign born. China and India each account for about 15% of the foreign-born respondents (Fig. 1). If we include Hong Kong and Taiwan, over 25% of all inventors born outside of the US (and 7% of all US inventors) are Chinese. We also find significant numbers from the United Kingdom, Canada, Germany and the Russian Federation. Most inventors born outside of the US received their highest degree in the US, but a significant minority (over 30%) was educated overseas. Inventors born outside of the US are also much more likely to have a PhD (68% versus 37%, P < 0.01). The distribution across the type of organizations is very similar across the two groups. The foreign born are under-represented in mechanical and miscellaneous technology sectors. They are more likely to have a degree in chemistry and less likely to have a social science or humanities degree (Supplementary Methods). Some of this high rate of patenting is thus likely to be due to the over-representation of the foreign-born among the most educated (especially PhDs) as well as to the concentration of the non-USborn in S&E fields14. Consistent with prior studies, we found that non-US-born inventors make a disproportionate contribution to the US innovation system (Table 1)8,15. The average number of invention
290
disclosures in the last three years (10 versus 9) and publications in the last three years (6 versus 2) were significantly higher for the non-US born. In addition, the percentages of patents self-reported as being in the top 10% in technical significance (22% versus 13%) or economic value (16% versus 11%) were also significantly higher for the non-US born lead inventors’ patents than for those of the US born, while the mean number of forward citations (3.2 versus 3.1) and percent of patents commercialized (63% versus 60%) were at least as high. To specify the contribution of the foreignborn status, controlling for other correlates of innovative performance such as firm, project and invention characteristics, we conducted a series of regression analyses. We controlled for the lead inventor having a PhD degree, and also tested models adding an interaction term for foreign-born status with PhD status (Table 2, Model 3 in Supplementary Methods), to see whether the high-skilled foreign-born inventors are more likely to produce high-quality patents than US-born PhDs (or foreign-born individuals without PhDs) (Table 2, Model 1 in Supplementary Methods). shows the effects of the lead inventor being foreign born on each of our output measures, without controlling for other factors, and Table 2, Model 2 in Supplementary Methods show models with appropriate control variables for each outcome. We began with measures of productivity (publications and disclosures)4,8,11. We controlled for having a PhD, time spent on invention-related activity, technology class and type of organization (large firm, small firm or public research organization). Consistent with prior work, the expected probability of having one or more publications was 26% for the US born and 38% for the foreign born (for this and other estimates of expected probabilities, we set the other variables in the logistic model to the mean or the reference category)8. The foreign born also had higher expected publication counts (Supplementary Methods). However, again consistently with prior work, we found that, controlling for technology class, education and
time spent on inventing, the foreign born were not more likely to have any invention disclosures or have higher numbers of disclosures4. As expected, we also found that PhDs (both US and foreign born) have more publications and more disclosures. The interaction effects between PhD and foreign-born status were not significant. Next, we examined the quality of the patented inventions using our three measures (self-rated technical significance, self-rated economic value and counts of forward citations). For these models, we controlled for organization type (as above), size of the project (inventor months and number of inventors), project goal (enhancing the technology base of the firm or generating a new line of business, with developing the existing business as the excluded category) and technology class. The expected probability of having a top 10% patent (technical significance) was 21% for the patents of US-born lead inventors and 37% for the foreign-born (net of the control variables). This gap was even greater among the PhD-level lead inventors, producing a 20% higher probability of a foreign-born lead inventor with a PhD having a top 10% invention, compared to a US-born lead inventor with a PhD. When we added the interaction term, the main effect of being foreign born (that is, for those foreign-born individuals without a PhD) was not statistically significant. We found similar results for self-reported economic value. The percent of top 10% patents (economic value) was 12% for the US-born lead inventors but 21% for the foreign-born. When we added the foreign-born with PhD interaction term, foreign-born lead inventors still had a significantly higher chance of a top 10% patent (economic value), whereas those with a PhD (either US born or foreign born) were not significantly more likely to have patents of top 10% economic value (in part because they are more likely to work on more upstream projects that may have high technical significance but may not be as commercially valuable). We also examined forward citations to these patents. When estimating the number of forward citations, we include additional controls
Table 1 Productivity descriptive statistics: US versus foreign-born primary inventor response only US-born
Foreign-born
Mean number of invention disclosures, last 3 years
9
10
9**
Mean number of publications, last 3 years
2
6
3***
Mean number of forward citations to the patent
Total
3
3
3
Top 10% in technical significance (% yes)
13
22
16***
Top 10% in economic value (% yes)
11
16
13***
Any commercialization (% yes)
60
63
61
72 (1,276)
28 (502)
100 (1,778)
Total sample size (%, N) *P < 0.1, **P < 0.5, ***P < 0.01. Weighted by sampling weight.
volume 28 number 3 march 2010 nature biotechnology
© 2010 Nature America, Inc. All rights reserved.
ca r e e r s an d r e c r u i tm e nt for patent characteristics that are likely to drive citations, as well as education, organization, project and technology class (Supplementary Methods). Patents by non-US-born lead inventors have about 9% more forward citations. Again, we found a negative relation between the lead inventor having a PhD and the number of forward citations to the patent (similar to economic value) both overall and when we examined the non-US-born and PhD interaction. Finally, we examined differences in the probability of commercializing the invention, controlling for other predictors of commercialization, such as the type of organization, technology class, project goal, size of the project and technical significance of the patent. We found that having a non-US-born lead inventor increases the expected commercialization rate (net of other variables) from 78% to 83%. Again, we found that inventions by PhDs were less likely to be commercialized, controlling for the technical significance and types of projects (that is, whether more upstream or closer to market). Thus, not only do foreign-born lead inventors produce more valuable inventions, but, net of their value, these inventions are also more likely to be commercialized, suggesting that inventions with foreign-born lead inventors are over-represented among successful innovations. When we separated the foreign born into those whose highest degree was obtained in the United States and those who finished their education overseas, we found that both groups’ patents had higher average value (compared to those of the US born) across all three measures (Supplemental Methods), although only the US-educated foreign-born lead inventors’ patents had a significantly higher probability of being commercialized. Thus, both foreignborn lead inventors educated in the United States (likely first arriving on student visas) and those trained overseas (likely arriving on H-1B visas) make a disproportionate contribution to US innovation4. Finally, we tested to see whether biotech inventions differed from other inventions and whether the role of the foreign born differed from those in other technologies, controlling for inventor, firm and project characteristics (Supplemental Methods). Compared to other technologies, biotech is characterized by both an especially close link with science and an especially strong dependence on patents. In addition, commercialization lags tend to be longer, in part because of the extensive testing related to the regulatory process. We found that biotech inventors had above-average numbers of publications, suggesting closer links between science and commercial activity in this field. Biotech inventions had, on average, fewer forward
citations, which likely reflects citation practices and the extent of prior art. Commercialization rates were also lower, probably due to the greater uncertainty of biotech inventions (such that many inventions are patented on the hope that some will be successfully commercialized) and longer commercialization lags. When we compared the foreign born working in biotech to the foreign born in our sample from non-biotech sectors, we found few differences (although the foreign born in biotech were less likely to publish than the foreign born in other fields or the US born in biotech). Thus, overall, we found that the foreign-born lead inventors in biotech generally produce inventions that are as valuable as those in other fields (such as IT) that have been the focus of the debates on high-skilled immigration. Discussion These findings suggest that scientists and engineers born outside of the US play a major role not only in our scientific output but also our innovation output. Although many of the measures we assessed are self-reported and subject to potential biases, the consistency across several kinds of measures, and with prior work using different measures, suggests that some confidence can be placed in the results. As has been observed in previous studies, we find that foreign-born individuals are no more likely to invent, once we control for field and degree4. However, the quality of the patents by lead inventors born outside of the US are higher on average (whether assessed on the basis of self-reported quality, forward citation or the probability of commercialization), even after controlling for technology class, education level, and firm and project characteristics. Moreover, the contribution of foreign-born inventors in the US could be even larger if we took into account possible spillover effects14. These findings suggest that the combination of self-selection by those coming to the US, and screening by universities and firms as well as by immigration officials, are leading to significant numbers of foreign-born inventors who are producing inventions of above-average quality and making an important contribution to the US innovation system. Those engaged in current policy debates should thus be wary of negatively affecting the participation of the foreign-born scientists and engineers in the US innovation system, with the restrictions on stimulus package funding and complications in the visa process being examples of policies that may adversely affect the ability of the US economy to benefit from foreign born inventors1. In particular, we should continue to allow high-skilled immigrants, especially those with degrees from US graduate schools, the opportunity to work in this country in order
nature biotechnology volume 28 number 3 march 2010
to capture the returns from US investment in training those with graduate degrees (often in the form of research assistantships funded by the federal government). Growing competition from other countries and a less inviting US environment may lead these mobile inventors to relocate to more hospitable locations, resulting in a significant loss to the US innovation system at a time when US competitiveness depends heavily on technological advance. Recent policies that prevent firms from hiring the most qualified workers, regardless of country of birth, can only reduce US competitiveness and retard economic recovery. Note: Supplementary information is available on the Nature Biotechnology website. Acknowledgment We would like to thank Japan’s Research Institute of Economy, Trade and Industry and the Georgia Research Alliance for funding for this research. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. 1. Editorial. US visa nightmares. Nature 461, 12 (2009). 2. National Research Council Policy. Implications of International Graduate Students and Postdoctoral Scholars in the United States (National Academy of Science, Washington, DC, USA, 2005). 3. Wadhwa, V., Saxenian, A.L., Rissing, B. & Gereffi, G. America’s new immigrant entrepreneurs: part I. Duke Science, Technology & Innovation Paper 23 (2007). 4. Hunt, J. Which Immigrants Are Most Innovative and Entrepreneurial? Distinctions by Entry Visa. National Bureau of Economic Research Working Paper w14920 (National Bureau of Economic Research, Cambridge, Massachusetts, USA, 2009). 5. Kerr, W.R. & Lincoln, W.F. The Supply Side of Innovation: H-1B Visa Reforms and US Ethnic Invention. Harvard Business School Working Paper 09-005 (Harvard Business School, Boston, Massachusetts, USA, 2008). 6. National Academy of Sciences. Policy Implications of International Graduate Students and Postdoctoral Scholars in the United States, (National Academy Press 2005). http://www.nap.edu/catalog.php?record_ id=11289#toc 7. National Science Board. Science and Engineering Indicators (US Government Printing Office, Washington, DC, USA, 2006). 8. Levin, S.G. & Stephan, P.E. Science 285, 1213–1214 (1999). 9. National Science Board. Science and Engineering Indicators (US Government Printing Office, Washington, DC, USA, 2008). 10. Hsu, D.H., Roberts, E.B. & Eesley, C.E. Entrepreneurs from Technology-Based Universities. Research Policy 36, 768–788 (2007). 11. Kerr, W.R. Harvard Business School Working Paper The Ethnic Composition of US Inventors, (Cambridge, Massachusetts, USA, 2007). 12. Kling, J. & DeFrancesco, L. Nat. Biotechnol. 25, 1217– 1223 (2007). 13. US Census Bureau http://factfinder.census.gov/servlet/ QTTable?_bm=y&-geo_id=D&-qr_name=DEC_2000_ SF4_U_QTP14&-ds_name=D&-_lang=en&redoLog=false (US Census Bureau, Washington, DC, 2003). 14. Hunt, J. & Gauthier-Loiselle, M. How Much Does Immigration Boost Innovation? National Bureau of Economic Research Working Paper w14312 (National Bureau of Economic Research, Cambridge, Massachusetts, USA, 2008). 15. Wadhwa, V. Issues Sci. Technol. Spring, 45–52 (2009).
291
© 2010 Nature America, Inc. All rights reserved.
people
Next-generation antibody developer Xencor (Monrovia, CA, USA) has announced the appointment of Edgardo Baracchini (left) as chief business officer. Baracchini brings more than 15 years of experience to Xencor, most recently as senior vice president of business development at Metabasis Therapeutics until its merger with Ligand Pharmaceuticals in 2009. Previously, he was vice president of business development at Elitra Pharmaceuticals, director of business development at Agouron Pharmaceuticals and assistant director of business development at Isis Pharmaceuticals. “As evidenced by the company’s string of partnerships last year, including Merck and Pfizer, Xencor’s antibody engineering technology provides a unique offering to partners, particularly at a time when there’s increasing competition in biologics,” says Baracchini. “I look forward to working with the Xencor team to further utilization of this broad-based technology.”
Felix J. Baker has been named to Ardea Biosciences’ (San Diego) board of directors. He is a managing partner of Baker Bros. Advisors, the manager and advisor to the funds of Baker Brothers Investments, Ardea’s largest shareholder. Lisa Costantino (left) has been named CFO of EMD Serono’s (Rockland, MA, USA) US organization, with responsibilities including the oversight and strategic direction-setting of finance and controlling, tax, accounting, procurement and information services. Costantino has more than 25 years of experience at EMD Serono, most recently serving as vice president of finance.
Pharmaceuticals as vice president clinical, medical and regulatory affairs for Europe, and in 2006 he became vice president global medical affairs inflammation, worldwide at UCB. Neuraltus Pharmaceuticals (Palo Alto, CA, USA) has announced that Andrew Gengos has been named president and CEO. Gengos joins Neuraltus after more than seven years at Amgen, where he served as vice president, strategy and corporate development. Before joining Amgen, he was vice president, CFO, and chief business officer of Dynavax Technologies.
NextBio (Cupertino, CA, USA) has appointed Kevin G. Cronin vice president of sales. He was most recently senior vice president of sales at Symyx Technologies.
Stem Cell Therapy International (Tampa, FL, USA) has named Hoon Han chairman of the company’s board of directors effective immediately and has announced the resignation of Lixian (John) Jiang from all positions with the company. Han is president and CEO of HistoStem Co., which established the first cord blood bank, the first bone marrow bank and the first stem cell bank in Korea. Jiang will continue assisting the company in an advisory capacity when the company forms its medical advisory committee.
Prime Therapeutics (St. Paul, MN, USA) has announced the appointment of Stacey Fahrner as vice president, government affairs. Fahrner previously served as a senior policy manager at the Blue Cross and Blue Shield Association in Washington, DC. Before that, she was a senior policy manager at Abbott Laboratories.
Lonza Group (Basel, Switzerland) has announced the hiring of Jonathan Knowles in a part-time role as scientific advisor, mainly to support the company’s LIFT (Lonza Innovation for Future Technologies) initiative. Knowles was formerly a member of the Roche executive committee and head of group research.
TxCell (Valbonne, France) has named Miguel Forte to the newly created position of chief medical officer. In 2004, he joined Nabi
Aton Pharma (Lawrenceville, NJ, USA) has named Wayne S. Mulcahy vice president, clinical operations. He has more than 30 years of
292
experience, most recently serving as vice president of R&D at Alita Pharmaceuticals. Aton also announced that Michael H. Richardson has joined the company as vice president, commercial operations for the newly formed orphan drug business unit. Richardson was previously a group director of marketing and senior director of worldwide product planning for Cephalon. Viroblock (Geneva) has announced the hiring of Jamie Paterson as CEO. Paterson has 20 years of experience in the pharmaceutical industry, most recently as chief commercial officer at Neurim Pharmaceuticals. Robert Phelps has joined SuppreMol (Martinsried/Munich, Germany) as head of business development and licensing. Most recently, he was director, business development and licensing at PARI Pharma. William H. Rastetter has been elected to the board of directors of Neurocrine Biosciences (San Diego), effective immediately. He currently serves as chairman and CEO of Receptos and chairman of Illumina and is a partner in the venture capital firm Venrock. Rastetter retired as executive chairman of Biogen Idec at the end of 2005. Billy Tauzin, president of the Pharmaceutical Research and Manufacturers of America (Washington, DC), has announced his resignation after taking criticism for his support of President Obama’s plan to overhaul health care. A former congressman from Louisiana, Tauzin will officially step down on June 30. Jan van Heek has been named to the board of Amarin (Dublin and Mystic, CT, USA) as a non-executive director. As part of his board role, van Heek has been appointed chairman of the company’s audit committee. He previously spent more than 18 years at Genzyme, most recently as senior advisor to the CEO and senior management team. Karo Bio (Stockholm) has announced the resignation of its president and CEO Per Olof Wallström to pursue other opportunities. Wallström, who has served in his position for the past five years, will stay with the company until his successor has been appointed.
volume 28 number 3 march 2010 nature biotechnology