Вы находитесь на странице: 1из 107

volume 30 number 9 september 2012

e d i tor i a l
805 Will the floodgates open for gene therapy?
806 Further confirmation needed

Visualization of mass cytometry cell


n ews
signaling data from 14 cell types 807 First gene therapy nears landmark European market authorization
(y axis) identified in primary human 808 US funds vaccine centers for biodefense
peripheral blood mononuclear cells
treated with 27 kinase inhibitors 810 Bayer acquisition spotlights biopesticides
(x axis) and 14 stimuli (z axis). 810 Regulatory fog lifts on obesity drugs
Bodenmiller et al. describe an approach
for multiplexing samples for mass
812 Amylin’s three-party good-bye
© 2012 Nature America, Inc. All rights reserved.

cytometry, which facilitates such large- 812 Industry cautiously welcomes Supreme Court decision on healthcare overhaul
scale analyses (p 858). 814 Compulsory license bandwagon gains momentum
Credit: Erica Savig, Nolan lab.
814 Biotechs opt for alternative floatation strategy
815 GSK buys partner Human Genome Sciences
815 India’s biosimilar regulations
815 Myriad’s patents redux
816 NIH injects $275 million into undiagnosed diseases and RNA research
816 China’s key R&D programs behind schedule
816 Around the world in a month
817 data page: Drug pipeline: Q212
818 News feature: Patient power

B i oe n trepre n eur
Gene therapy nears the finish line,
p 807
B u i l d i n g a bus i n ess
npg

821 How much risk are you prepared to take?


Mark Van Dyke

op i n i o n a n d comme n t

C O R R E S P O ND E N C E
825 To be or not to be transgenic
826 Broad consent in biobanking
826 An accelerated workflow for untargeted metabolomics using the METLIN database
828 Successful suppression of a field mosquito population by sustained release
Foundations get busy, p 818 of engineered male mosquitoes

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$4,677 (institution), US$5,382
(corporate institution). Canada add 5% GST #104911595RT001; Euro-zone: €202 (personal), €3,713 (institution), €4,634 (corporate institution); Rest of world
(excluding China, Japan, Korea): £130 (personal), £2,400 (institution), £2,990 (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, 75 Varick Street, 9th Floor, New York, NY 10013-1917. 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 © 2012 Nature America, Inc. All rights reserved. Printed in USA.

i
volume 30 number 9 september 2012

feature
WNT
pate n ts
831 Teva v. AstraZeneca and secret prior art under 102(g)(2)
Sandra Lee & Michael Knierim
834 Recent patent applications in biosensors

Frizzled
N E W S A ND V I E W S
835 Modulating WNT receptor turnover for tissue repair
Therapeutic targeting of adult stem Arie Abo & Hans Clevers
cells, p 835
836 RNA-mediated programmable DNA cleavage
Rodolphe Barrangou
838 Silicon dreams of cells into symbols
Jeremy Gunawardena
841 Research highlights
© 2012 Nature America, Inc. All rights reserved.

computat i o n a l b i o l og y

perspect i ve
842 Navigating cancer network attractors for tumor-specific therapy
Pau Creixell, Erwin M Schoof, Janine T Erler & Rune Linding

research
Understanding teratoma risk from cell
perspect i ve
therapies, p 849
849 Lessons from human teratomas to guide development of safe stem cell therapies
Justine J Cunningham, Thomas M Ulbright, Martin F Pera & Leendert H J Looijenga

ARTICLES
858 Multiplexed mass cytometry profiling of cellular states perturbed by small-
npg

molecule regulators
Bernd Bodenmiller, Eli R Zunder, Rachel Finck, Tiffany J Chen, Erica S Savig,
Robert V Bruggner, Erin F Simonds, Sean C Bendall, Karen Sachs, Peter O Krutzik
& Garry P Nolan
868 Combinatorial discovery of polymers resistant to bacterial attachment
Andrew L Hook, Chien-Yi Chang, Jing Yang, Jeni Luckett, Alan Cockayne,
Steve Atkinson, Ying Mei, Roger Bayston, Derek J Irvine, Robert Langer,
Daniel G Anderson, Paul Williams, Martyn C Davies & Morgan R Alexander

Discovering bacteria-resistant
materials, p 868

nature biotechnology iii


volume 30 number 9 september 2012

l etters
876 Directed differentiation of human pluripotent stem cells into mature airway
epithelia expressing functional CFTR protein
Amy P Wong, Christine E Bear, Stephanie Chin, Peter Pasceri, Tadeo O Thompson,
Ling-Jun Huan, Felix Ratjen, James Ellis & Janet Rossant
883 Polyethyleneimine is a potent mucosal adjuvant for viral glycoprotein antigens
Frank Wegmann, Kate H Gartlan, Ali M Harandi, Sarah A Brinckmann,
Margherita Coccia, William R Hillson, Wai Ling Kok, Suzanne Cole, Ling-Pei Ho,
Teresa Lambe, Manoj Puthia, Catharina Svanborg, Erin M Scherer, George Krashias,
Adam Williams, Joseph N Blattman, Philip D Greenberg, Richard A Flavell,
Amin E Moghaddam, Neil C Sheppard & Quentin J Sattentau
Differentiating stem cells to lung
epithelia, p 876 889 Engineering phosphorus metabolism in plants to produce a dual fertilization and
weed control system
Damar Lizbeth López-Arredondo & Luis Herrera-Estrella

careers a n d recru i tme n t


894 Theory to practice: real world case-based learning for management degrees
Maria Theodosiou, Jean-Philippe Rennard & Arsia Amir-Aslani
© 2012 Nature America, Inc. All rights reserved.

896 people

Polyethyleneimine as a mucosal
adjuvant, p 883
npg

Dual fertilizer and weed control using


phosphite, p 889

nature biotechnology v
in this issue

27 kinase inhibitors. The unprecedented scale of the study allowed


Combinatorial discovery of bacteria- the authors to assess the kinase- and cell type–selectivity of inhibi-
resistant materials tors and differential responses of cells collected from eight human
Bacterial infections donors. In principle, the approach could be used to multiplex tens
associated with medical of thousands of samples, facilitating new uses of mass cytometry as a
devices such as catheters, readout for drug screens and genome-wide RNA interference studies.
heart valves and prosthetic [Articles, p. 858] CM
joints pose a substantial
burden of disease, so Engineering fertilization and weed control
preventing such infections
through the use of Phosphorus is an essential and non-
materials that reduce renewable element that is included in
bacterial attachment, and commercial fertilizers to improve crop
© 2012 Nature America, Inc. All rights reserved.

subsequent formation of yields, because most soils lack suffi-


biofilms and growth, is cient phosphate for maximum plant
an attractive option for growth. But as phosphorus is nonre-
intervention. Alexander and colleagues present a high-throughput newable, new approaches to the provi-
microarray approach to screen hundreds of polymeric materials sion of phosphorus for plant growth are
for those that restrict the adherence and growth of bacteria. They required. Phosphite is a reduced form of
find a new group of structurally related materials comprising ester phosphorus that is more soluble than
and cyclic hydrocarbon moieties that dramatically reduce the orthophosphate, but it is not used as
attachment of the pathogenic species Pseudomonas aeruginosa, a fertilizer because it inhibits plant growth. López-Arredondo and
Staphylococcus aureus and Escherichia coli to coated surfaces. Herrera-Estrella report the production of transgenic Arabidopsis
The authors show that the discovered materials, when used to and tobacco that express a bacterial phosphite-oxidoreductase
coat silicone catheters, outperform a commercial silver hydrogel gene, enabling these plants to use phosphite as the sole source of
coating in a mouse infection model, revealing the potential of phosphorus. The transgenic plants can metabolize phosphite sup-
these materials for clinical translation. [Articles, p. 868] SJ plied in sterile and nonsterile soil, as well as when it is applied as
a foliar fertilizer. Because they lack the phosphite-oxidoreductase
gene, weeds are inhibited by phosphite, and the authors show that
Multiplexing mass cytometry phosphite could hold promise as a dual fertilization and weed control
system for plants transgenic for this trait. Adoption of this approach
npg

Mass cytometry can has the potential to reduce the amount of nonrenewable phosphorus
currently measure 34 ors
used in commercial farming. [Letters, p. 889] SJ
ibit
parameters on cells. 27
inh

But samples must be


analyzed one at a time,
Mature lung epithelial cells from stem cells
which hampers large- Until now, cystic fibrosis
scale studies of many research has been ham-
samples across several pered by the lack of a
conditions. Nolan and ready supply of lung epi-
colleagues describe the thelial cells expressing
use of seven metal ion tags to label cells in each well of a 96-well the mutated forms of
plate with a unique combination of tags. This allows the entire plate the cystic fibrosis trans-
to be analyzed in a single instrument run, thereby increasing sample membrane conductance
throughput while leaving 27 parameters to characterize the cells. The regulator (CFTR) gene that give rise to the disease. Rossant and col-
authors harness this multiplexing capability to quantify 14 phosphor- leagues provide a solution by devising a protocol for differentiating
ylation sites in 1,344 treated samples of primary human peripheral human pluripotent stem cells to mature CFTR-expressing proximal
blood mononuclear cells. The cells are from time-course or dose- airway epithelia and applying it to induced pluripotent stem cells
response experiments using different combinations of 12 stimuli and (iPSCs) derived from cystic fibrosis patients. The protocol retraces
the pathway of in vivo lung development. Pluripotent stem cells are
converted stepwise into definitive endoderm, anterior foregut (by
Written by Kathy Aschheim, Michael Francisco, Susan Jones, treatment with fibroblast growth factor (FGF)2 and sonic hedgehog)
Jason Kreisberg & Craig Mak and proximal lung progenitor cells (by treatment with FGF7, FGF10,

nature biotechnology volume 30 number 9 SEPTEMBER 2012 vii


i n t h is iss u e

FGF18 and of bone morphogenetic protein 4). Finally, an air-liquid mixed with viral glycoproteins, PEI
interface culture is used to differentiate the proximal lung progenitor forms large, spherical complexes that
cells into mature epithelial cells expressing the appropriate mark- are taken up by dendritic cells, result-
ers, including CFTR, which is properly localized to the apical plasma ing in the production of high-avidity
membrane and shows cAMP-regulated channel activity. To dem- antibodies and a primarily TH2-
onstrate the utility of their approach in drug screening, the authors biased response in mice. Complexes
generate CFTR-expressing cells from induced pluripotent stem cells of PEI with either a herpes simplex
derived from patients with the most common cystic fibrosis mutation virus type-2 glycoprotein or influ-
(F508del) and show that the phenotypic defect is partly rescued by an enza hemagglutinin administered
experimental cystic fibrosis drug. [Letters, p. 876] KA intranasally are shown to protect
mice against an otherwise lethal viral
challenge. Unlike other experimental
Polyethyleneimine as a mucosal adjuvant mucosal adjuvants, such as cholera
Adjuvants can greatly increase the immunogenicity of vaccines. toxin or α-galactosylceramide, PEI
Although many important pathogens gain entry at mucosal surfaces, triggers the release of double-stranded DNA from dying cells, leading
there are no mucosal adjuvants licensed for use in people. Sattentau and to activation of cytoplasmic DNA sensors and the transcription factor
colleagues now report that polyethyleneimine (PEI), a cationic polymer Irf3, a key regulator of interferon alpha and beta. Future studies will
used for DNA delivery in vitro and in vivo, has the unexpected property determine whether PEI is an effective adjuvant for a wider range of
of functioning as a mucosal adjuvant for viral protein antigens. When antigens in humans. [Letters, p. 883]  JK
© 2012 Nature America, Inc. All rights reserved.

Patent roundup
Next month in
A recent Federal Circuit decision concerning an oft-neglected
prior art provision could have a substantial impact on the
patentability of life science inventions. Lee and Knierim break • Mammalian cells as building blocks in synthetic
down Teva v. AstraZeneca and the future of Section 102(g)(2). biology
[Patents, p. 831] MF • Detecting newly synthesized secreted proteins
• TRICEPS pulls down receptors
Recent patent applications in biosensors.
[New Patents, p. 834] MF • CRISPR processing for predictable gene expression
npg

viii volume 30 number 9 SEPTEMBER 2012 nature biotechnology


Editorial

Will the floodgates open for gene therapy?


In a matter of days, a momentous event will occur: a gene therapy will, for the first time anywhere in the Western
hemisphere, be available commercially with full marketing approval.

T owards the end of July, the European Medicines Agency’s Committee


for Medicinal Products for Human Use (CHMP) recommended the
approval in the European Union of a treatment called Glybera (alipogene
That is difficult, leading to life-threatening pancreatitis in a high propor-
tion of patients. Glybera addresses the underlying LPL deficiency through
multiple intramuscular injections of an adeno-associated virus vector that
tiparvovec) from a small Dutch company, uniQure biopharma. Unless mat- delivers functional LPL genes to muscle cells.
ters get untypically administratively complicated, it should be only a matter Getting Glybera to market has been a struggle. The company that had
of days before the European Commission endorses this recommendation, started its development—a now-defunct Dutch startup called Amsterdam
© 2012 Nature America, Inc. All rights reserved.

making the approval official. Molecular Therapeutics (AMT)—submitted its approval file in December
Admittedly, the event will probably not compare in terms of impact on 2009. The gene therapy protocol, which includes repeated administration
the public’s imagination with the first manned moon landing, the first test of the functional gene and immunosuppression to prevent ‘rejection’ of the
tube baby or even the sequencing of the human genome. But within its own treatment didn’t go down well with regulators.
limited firmament, the first gene therapy to be sold legally in Europe will be After nearly two years of head scratching, see-sawing and scientific
an affair of some significance. bu­reaucracy, the CHMP declared the treatment “non-approvable” in October
Gene therapy enthusiasts are comparing Glybera’s imminent approval 2011, causing AMT to halve its workforce to sit out the process. In February,
to the US approval in 1986 of Orthoclone OKT3, the first monoclonal as AMT was irreversibly bound for liquidation, Dutch investor Forbion
antibody (mAb) approved for use in humans. Back in those simpler times, put forward €6 ($7.5) million to found uniQure biopharma as a vehicle to
enthusiasts then thought OKT3 would open the floodgates for a plethora acquire AMT’s assets, including Glybera.
of therapeutic antibodies. It subsequently became clear that there was some wiggle room that would
And in a way, it did. But first there was only a trickle. The flood came much allow Glybera onto the market. The CHMP recommended marketing autho-
later. In that sense, the comparison between Glybera and OKT3 may well be rization only under “exceptional circumstances,” under which uniQure will
appropriate. OKT3 was very much a prototypic antibody therapeutic. It was have to monitor patient outcomes and feed the information straight to the
the first clinical product of the unpatented Köhler and Milstein technology European Medicines Agency.
that had been described some 11 years earlier. Researchers had developed Jörn Aldag, the CEO of uniQure (and before that the CEO of AMT),
mAbs against a vast range of targets and had already shown, preclinically maintains that the approval of Glybera puts gene therapy “where [mAbs]
at least, the exquisite specificity of many of them. Boosters of the technol- were in the late 1990s, when they were just visible, but tiny.” He predicts a
npg

ogy fondly, and correctly, imagined a procession of mAbs for all sorts of “steep growth curve” (p. 807).
conditions marching upon a market desperate for drugs targeting highly He may well be right. The relatively rich pipeline of gene therapy candi-
defined molecular markers. In the end, it did happen…only the timing dates already in human trials suggests there may be a surge in the number
was a bit wrong. The problem was, OKT3 sidestepped one fundamental of gene therapies approved over the next few years.
deficiency of 1986 murine antibody drugs. What is less certain, however, is whether there will be a noticeable
OKT3 was approved in an era when clinicians had not defined the limita- flood of patients treated. Many of the gene therapies in clinical develop-
tions that the human-anti-mouse-antibody (HAMA) effect would impose ment are treatments for very rare, single-gene deficiencies: Leber’s con-
on the use of mouse-derived antibodies. The basic problem seems obvious genital amaurosis, severe combined immunodeficiency, Wiskott-Aldrich
now: in response to mouse antigens, humans mount an immune reaction syndrome and the like. Among these Mendelian disorders, the hemophilias—
that tends, at the least, to severely limit repeated dosing of the same (or themselves hardly common conditions—represent a pinnacle of frequency.
other) mouse antibodies. Not only that, but the immune response itself Even for these kinds of conditions, there may be HAMA moments in the
can be dangerous. future—potential immune attenuation of efficacy, safety issues related to
OKT3 conveniently circumvented this drawback because its target was constructs or vectors, not to mention competition from existing enzyme
CD3 on T cells, the binding of which leads to apoptosis. As a treatment for replacement therapies that have been available for many years. For more
the suppression of organ transplant rejection, the murine mAb not only was complex illnesses—and gene therapy is being explored for heart disease
used in patients who already were immunosuppressed but also was given as and cancer, too—it remains far from clear that the technical challenges
an acute treatment for the first 7 days after transplantation. Thus, neutral- will be quickly overcome.
izing antibodies were not a problem. HAMA? SchmAMA! OKT3 was OK. So Glybera’s approval is unlikely to augur a flood of gene therapy drugs
Glybera may be OK, too. It is a gene therapy for lipoprotein lipase (LPL) for blockbuster markets that transform the drug business in the manner
deficiency, an ultra-rare inherited disorder that affects no more than two of mAbs. But even if gene therapies turn out to be nichebusters rather
people per million in the general population. Affected people can’t break than blockbusters, their successful commercialization is a remarkable
down fat and try to manage their disease by strictly limiting fat in their diet. achievement nonetheless.

nature biotechnology volume 30 number 9 september 2012 805


editorial

Further confirmation needed


A new mechanism for independently replicating research findings is one of several changes required to improve the
quality of the biomedical literature.

O ver the past year, the reputation of the biomedical literature has
taken a bit of a beating. Controversy has centered around studies by
researchers at Bayer Pharmaceuticals and Amgen, which independently
returned, the initiative’s advisory board determines whether the study
has earned a ‘certification of reproducibility’ (authors cannot appeal the
board’s decision).
have shown that a troublingly high number of papers in certain areas It is entirely the author’s prerogative as to whether the findings are writ-
of translational research cannot be replicated. Last month, as a response ten up; if they are, they can be published in a special section of PLoS
to this problem, the Palo Alto, California–based Science Exchange ONE. Nature Publishing Group and Rockefeller University Press have also
announced the launch of the Reproducibility Initiative (http://www. agreed to link from the original publication to the PLoS ONE paper, and
reproducibilityinitiative.org/). This initiative is one way in which online Figshare (http://figshare.com/) will host the data from the verification.
© 2012 Nature America, Inc. All rights reserved.

platforms can facilitate rapid and independent corroboration of published Whether authors will elect to publish findings that fail to replicate their
results. But major progress in improving the reproducibility of research original results is unclear—if they don’t, then the arrangement clearly fails
will likely require more sweeping changes to the way in which science is to address the problem of research reproducibility.
published and validated. Another big question is, who pays for the work? According to Science
Published research findings are often modified or refuted by subse- Exchange, initially authors will be paying. But this is less than optimal,
quent evidence. There is nothing unusual about this; it is the way scientific given the competing interest and the scarcity of research funding.
knowledge progresses. But there is an increasing concern that publication A different scenario would be for funding agencies to bankroll the
bias toward positive results, rising competition to rush findings into print, effort. Alternatively, tech transfer offices—or some other, richer part of
an overemphasis on publishing conceptual breakthroughs in high-impact the author’s institution—could support validation work as a means of
journals and a lack of incentives for academic researchers to retract irre- making academic assets more attractive to potential licensing partners.
producible findings may be increasing the incidence of false claims in the With the current vogue for ‘capital efficiency’, perhaps venture capitalists
literature. All of which has implications for translational research. and pharma companies will use the service if it’s cheaper than replicating
Misleading papers result in considerable expenditure of time, money work in-house.
and effort by researchers following false trails. This affects the careers So what about the other papers, validation of which won’t find support
of postdocs and academics. It affects companies and investors, present- from investors, companies, institutions or funders?
ing yet another barrier for the translation of academic discoveries into Online commenting on papers—available on Nature for the past 2.5
new medicines by diverting funds away from real advances. And most years but yet to be rolled out to other Nature research journals—remains
troublingly, it can result in patients being exposed to drugs on the basis of an easy way for the community to highlight problematic studies. More
npg

wrong information. In the past year, two studies have brought into sharp could also be done during the peer review process in terms of rejecting
focus just how bad the problem may be. papers in which only ‘representative’ data are reported or proper statistical
In September 2011, a team of researchers at Bayer provided a retro- analysis is not used (Nat. Neurosci. 14, 1105–1107, 2011).
spective survey of four years of work in oncology, women’s health and Journal editors and reviewers can also do more to ensure that the rel-
cardiovascular target validation (Nat. Rev. Drug. Discov. 10, 712, 2011). evant information is captured about an experimental protocol and condi-
They asked 23 of their R&D scientists to tally papers they’d acted upon tions, instrument settings and parameters. Too often, there is more than
and whether or not the findings had panned out. The analysis revealed a little secret sauce to the procedures followed in a laboratory. Detailed
that only ~20–25% of the relevant published data could be corroborated description and the publication of full protocols (together with videos
internally. depicting experiments) could help.
Five months later, a collaboration involving Amgen scientists published Perhaps most importantly, different research communities need to
the results of their efforts to replicate findings from recent publications in come together to address particularly troublesome research questions in
the clinical oncology literature (Nature 483, 531–533, 2012). The data were their field; in oncology, for example, a clear need exists for guidance on
disturbing. Of 53 papers, only 6 (11%) were reproducible. A particularly the preclinical cancer models to use in particular settings and to address
troubling aspect was the disclosure that in return for cooperation, several particular questions.
of the authors required the company to sign a confidentiality agreement Clearly, more should be done to increase the quality of published work.
preventing the identity of their paper from being revealed. The Reproducibility Initiative, and other efforts like it (e.g., http://www.
Which brings us to the Reproducibility Initiative. sciencecheck.org/), will help by validating results and ensuring that sup-
The effort takes advantage of Science Exchange’s existing network of porting data are placed in openly accessible repositories. Greater atten-
>1,000 core facilities and contract research organizations. After authors of tion also needs to be paid during peer review to the completeness of the
an original publication submit their study design, the initiative matches the experimental protocol disclosed, the supporting data and the robustness of
work to qualified facilities (the identity of which is masked from authors), the authors’ analysis. Most of all, a change in publication culture is needed.
which then attempt to replicate the studies for a fee. Once the results are Sometimes replication is as important as discovery.

806 volume 30 number 9 September 2012 nature biotechnology


news
in this section
Industry cautiously Biotechs opt for GSK buys partner
welcomes Supreme alternative floatation Human Genome
Court decision on strategy p814 Sciences p815
healthcare
overhaul p812

First gene therapy nears landmark European market authorization


Some 22 years after the first US Food and and clinically important reduction in the
Drug Administration–approved trial of occurrence of acute pancreatitis, the most
a gene therapy, a genetic medicine for the insidious symptom of LPL deficiency. Given
ultra-rare inherited disorder lipopro- the length of response, Aldag believes there
tein lipase (LPL) deficiency is poised for should be annualized payments in the
final approval in Europe. If the European same range as enzyme-replacement thera-
Medicines Agency (EMA) gives its formal pies, which patients receive on a consistent
authorization at the end of this month, basis. Currently in Europe, those types of
the LPL deficiency treatment, alipogene medicines are priced at €150,000–450,000
tiparvovec (Glybera), will be the first gene ($187,000–561,000) per annum. (LPL defi-
© 2012 Nature America, Inc. All rights reserved.

therapy to be approved for sale in a Western ciency cannot be treated with a replacement
market. Although the challenges of admin- therapy because the enzyme’s half-life is too
istering and monitoring novel gene thera- short.)
pies will likely limit clinical applications to Similar pricing would make alipogene the
indications poorly served by conventional most expensive approved medicine of all time
medicines in the near term, the decision for a single treatment, so it bears watching
potentially opens the way for a renaissance how this will play out with pricing and reim-
in genetic medicines. bursement bodies. In addition, there are
Approval will also be a watershed moment onerous post-marketing requirements that
from a regulatory standpoint, as the EMA’s will ensure patients who have received ali-
assessment of alipogene has followed pro- pogene continue to be followed. uniQure will
Fat particles (pictured) accumulate in the
tracted deliberations in which the file was have to maintain a patient registry and put in blood stream of people with lipoprotein lipase
examined and voted on four times, and the place a risk management plan under which deficiency. Source: uniQure
conventions—if not the rules—governing no more than two patients per month will be
EMA procedures were thrown out of the treated, at a limited number of expert centers.
window (Box 1). The company intends to follow-up its cost uniQure dear, but it also, “represents a
“It’s been a long ride, but an important success with the EMA by submitting the very substantial investment on the regulatory
one, and the outcome is terrific,” says Jörn alipogene file to the US Food and Drug side,” which has parallels with the amount of
npg

Aldag, CEO of Amsterdam-based uniQure, Administration (FDA) and Health Canada. time and effort the FDA and other agencies
the developer of alipogene tiparvovec. Based on talks with those regulatory agen- put into establishing the safety of monoclo-
But guiding the gene therapy through cies, Aldag says he’s “pretty certain they will nal antibodies (mAbs) and biopharmaceu-
the byzantine and meticulous European accept the same data.” ticals. “That work by the regulators is very
regulatory process has been expensive. If alipogene can clear the final hurdle of much paying off in terms of getting approvals
uniQure (formerly Amsterdam Molecular formal market authorization this month, it [of mAbs] now,” Ostrove says.
Therapeutics) invested €50 ($62) million will be a landmark for others in the gene Sander van Deventer, co-founder of
on the development of alipogene, including therapy field (Table 1), providing valida- uniQure’s forerunner, Amsterdam Molecular
€15 ($19) million spent after the product tion that a gene therapy can indeed be com- Therapeutics, and who earlier in his career
was first submitted to EMA in December mercialized. According to Aldag, the field is was the first to administer an anti-tumor
2009. Though the extremely low incidence primed for growth. “Gene therapy is at the necrosis factor (TNF)-a mAb, to patients
of LPL deficiency (1–2 afflicted per million position where monoclonal antibodies were with Crohn’s disease, is in a unique position
people) would seem to leave uniQure with in the late 1990s, when they were just visible, to compare the regulators’ views of mAbs
little chance of recouping this investment, but tiny. It will now follow a similar, very in the 1990s versus today’s gene therapies.
the many profitable approved products for steep, growth curve,” he predicts. That pilot antibody, Remicade (infliximab),
rare diseases have blazed a pricing path ali- Jeff Ostrove, president and CEO of a secured FDA approval in 1998, but “the
pogene can follow, though it needs tweaking. US gene therapy specialist, Ceregene, of attitude of many people to [mAbs] prior to
Alipogene, a recombinant adeno-associated San Diego, agrees with this assessment. 1995 was exactly the same as to gene ther-
viral (AAV) vector expressing the Ser447X “[Alipogene] is a very positive dry run for apy today. Now that’s going to change,” van
variant of human LPL, is given in a single the entire field and really in many ways is Deventer says.
sitting by means of several injections into the very positive tip of an iceberg.” Ostrove The road to approval for alipogene has not
the leg muscles, resulting in a long-term notes that reaching this point has not only been easy. After the EMA’s second rejection

nature biotechnology volume 30 number 9 SEPTEMBER 2012 807


NEWS

in brief Table 1 Selected companies with gene therapies in clinical development


Company Product Indication Clinical status
US funds vaccine centers for uniQure Alipogene (Glybera) Lipoprotein lipase Pending approval
biodefense deficiency
AMT-060 Hemophilia B Phase 1/2
The US Department of Health and Human
Services (HHS) has allocated $400 million Advantagene (Auburndale, ProstAtak (AAV expressing herpes Prostate cancer Phase 3
for three Centers for Innovation in Advanced Massachusetts) simplex virus thymidine kinase Glioma Phase 2
combined with valacyclovir)
Development and Manufacturing for vaccine Pancreatic cancer Phase 1/2
production, but some are concerned that
GlaxoSmithKline (London) 2696273 (GIADAl retroviral vector ADA deficiency in Phase 3
excessive focus on influenza may leave the encoding adenosine deaminase severe combined
nation less prepared against other threats. (ADA) for ex vivo hematopoietic immunodeficiency
A 2010 report prepared by HHS proposed stem cell therapy) (SCID)
establishing such centers to accelerate Vical (San Diego) Allovectin (4,853-bp DNA plasmid Melanoma Phase 3
development of medical countermeasures encoding human MHC class I
against chemical, biological, radiological HLA-B7 and chimpanzee b-2-
and nuclear threats, develop new vaccine microglobulin plus cytofectin
production technologies, and enable rapid DMRIE/DOPE)
manufacture of existing vaccines. On June 19, Applied Genetic AGTC-0106 (AAV vector carrying Alpha-1 antitrypsin Phase 2
the HHS Biomedical Advanced Research and Technologies (Alachua, alpha-1 antitrypsin gene) deficiency
Development Authority announced allocations Florida)
of $163 million for a center in Maryland, Bluebird Bio Lenti-D (lentiviral vector encod- Adrenoleukodystrophy Phase 2
$60 million for a center in North Carolina ing human ATP binding cassette
and $176 million for a center in Texas. Each subfamily D for ex vivo therapy of
center partners with companies—Novartis’s
© 2012 Nature America, Inc. All rights reserved.

hematopoietic stem cells)


vaccine facility in Holly Springs, North Carolina, Lentiglobin (lentiviral vector encod- Beta thalassemia Phase 2
GlaxoSmithKline’s Vaccines of Marietta, ing human beta-globin gene or a
Pennsylvania, and Emergent BioSolutions hybrid A-gamma/beta-globin for
Manufacturing in Baltimore, with researchers ex vivo hematopoietic stem cell
at Texas A&M (College Station), Michigan State gene therapy)
University (Flint), Duke (Durham) and North Celladon (San Diego) Mydicar (AAV1 encoding Cardiomyopathy Phase 2
Carolina State (Raleigh)—to accelerate response sarcoplasmic reticulum ATPase 2a)
to future pandemics. Ceregene Cere-120 (AAV2 encoding Parkinson’s disease Phase 2
Maj. Gen. (ret.) Philip Russell, former director neurturin)
of the Walter Reed Army Institute of Research,
Cere-110 (AAV2 encoding nerve Alzheimer’s Phase 2
an expert consulted by HHS, expressed growth factor) disease
disappointment in the implementation.
Genzyme (Cambridge, AAV-hAADC-2 (AAV2 encoding Parkinson’s disease Phase 2
“They combined influenza preparedness with
Massachusetts) the human aromatic l-amino acid
biodefense against very strong recommendations
decarboxylase gene)
from me and a lot of other people,” he says.
“They’re totally unrelated problems, and Oxford BioMedica ProSavin (equine infectious anemia Parkinson’s disease Phase 2
virus encoding aromatic l-amino (complete)
totally unrelated issues.” When this plan
acid decarboxylase, tyrosine hydrox-
was devised, more funds were available for
ylase and GTP cyclohydrolase I
preventing flu outbreak than for bioterrorism,
making this emphasis an expedient strategy. Shenzhen SiBiono Gendicine (replication-defective Lung cancer Phase 2
GeneTech (Shenzen, China) adenovirus vector encoding human
npg

However, he believes that supporting efforts


P53)
to mass produce and distribute flu vaccine
The US National Institutes of Health Clinicaltrials.gov website lists 2,965 gene therapy trials, of which 1,103 are listed as
will do little to facilitate R&D against novel recruiting. The vast majority of these are being run by academic investigators. AAV, adeno-associated virus; MHC, major histo-
threats. “We envisioned a single organization compatibility complex; DMRIE, 1,2-dimyristyloxypropyl-3-dimethyl-hydroxyethylammonium bromide; DOPE, dioleyl phosphati-
totally dedicated to advanced development and dyl ethanolamine.
manufacturing of biodefense vaccines,” says
Russell. “But it seems like technical capabilities
were secondary to lobbying power and decision of the gene drug in November 2011, uniQure [gene therapy] doses are so tiny, and this means
making.” Michael Eisenstein stopped all investment in the product and you won’t need the same multimillion dollar
abandoned the clinical and manufacturing scale of investment in manufacturing facilities.”
in their words infrastructure. The company is now rebuild-
ing, and Aldag says he hopes the first patient
Alipogene is most obviously a role model
for other gene therapies for monogenic
“It could be the to receive a commercial gene therapy will be inherited disorders, but it also paves the
biggest upside treated in the first half of next year. way for products in broader, multifactorial,
surprise in the history
On a broader level, there remains more noninherited diseases, such as ProSavin,
of the pharmaceutical
and biotech industry.”
spadework to do in terms of agreeing upon developed by Oxford, UK–based Oxford
Following the flop reimbursement, convincing physicians and BioMedica, which delivers genes for the three
of Pfizer/J&J’s others to push through the wider adoption enzymes that are needed to generate dopa-
bapineuzumab’s of gene therapy, admits Ena Prosser, partner mine, the neurotransmitter that is depleted
in August, ISI at Fountain Healthcare Partners of Dublin, a in Parkinson’s disease. “We are basking in
Group analyst Mark
venture capital firm that has invested in gene reflected sunshine because there are now
Schoenebaum ponders
the likely aftermath were Eli Lilly’s Alzheimer’s therapy companies. Even so, she notes one more ticks in the box for gene therapy,” says
anti-b-amyloid antibody, solanezumab, advantage some gene therapies will have—com- Stuart Naylor, the company’s CSO.
approved. (Bloomberg, 8 August 2012) pared with mAbs and in terms of uptake, “the Although alipogene is targeted at an

808 volume 30 number 9 SEPTEMBER 2012 nature biotechnology


news

Box 1 A regulatory marathon


The alipogene file was submitted in December 2009 and suffered its first rejection
by the EMA’s Committee for Medicinal Products for Human Use (CHMP) on June 23,
2011, after experts on another EMA body, the Committee for Advanced Therapeutics
(CAT) questioned its clinical efficacy, as only 27 patients had been treated, though it
acknowledged that the product’s AAV vector and associated manufacturing protocol met
safety standards. When uniQure requested a reexamination of the file, CAT suggested
its concerns could be addressed by post-marketing surveillance, but the CHMP was not
convinced, and rejected alipogene for a second time in November 2011.
At this point, uniQure considered the product dead, especially given the failures of
other companies, such as Ark Therapeutics and its adenoviral gene therapy sitimagene
ceradenovec (Cerepro) in malignant glioma, to negotiate the EMA approval process. But
in January 2012, the European Commission, with no official public explanation, told
the committee to think again. These deliberations resulted in a 16–15 majority in favor
of approving alipogene when the CHMP voted for the third time in April. Although the
vote was favorable, because 1 of the 32 CHMP members was absent on the day of the
vote and the rule book requires an absolute majority of 17 positives, alipogene again
got a thumbs down.
At this point, the EMA rewrote the rules, and rather than referring this decision to the
European Commission, told both CAT and the CHMP to take another look. Rules were
© 2012 Nature America, Inc. All rights reserved.

further breached because even though the official EMA line is that it won’t look at new
information when reexamining a file, it did weigh further follow-up data. This extended
the patient data considered from 2.4 years to more than 4 years when the file was
scrutinized for the fourth time in July. NM

extremely small group of patients, Naylor uniQure was able to negotiate repeated
believes the level of internal debate about rejections of alipogene by the EMA only
the treatment among the EMA’s various because its investors kept the faith; in truth,
committees and expert groups means the the gene therapy field has never been flush
agency is better suited to provide guidance with cash. That is beginning to change, how-
on the clinical development of other gene ever, as clinical trials deliver compelling
therapies. “This means better reassurance evidence of long-lasting effects, as seen with
that you don’t invest money in the wrong ProSavin, where Parkinson’s disease patients
place,” he says. have seen sustained improvements in symp-
Similarly, Nigel Parker, CEO of gene toms and in some cases were able to reduce
therapy startup, FKD their doses of L-dopa.
Therapies OY of Kuopio, This increased finan-
npg

Finland, says, “It’s clear uniQure was able to cial interest was exempli-
now that if you have got negotiate repeated fied in July when Bluebird
the right clinical data, rejections of alipogene Bio, of Cambridge,
all the technical barriers Massachusetts, raised
and safety concerns over
by the EMA only $60 million in an oversub-
gene-based medicines because its investors scribed Series D financ-
[have been addressed] kept the faith; in truth, ing.The backers are a roll
and the products are the gene therapy field call of blue chip investors,
approvable.” with five new funds join-
In addition, alipogene has never been flush ing the four existing ven-
is important in high- with cash. ture capitalists, and—in a
lighting the potentially sign that gene therapy is
superior therapeutic power of gene therapy edging further toward becoming a commer-
over existing treatments, Parker says. For cial proposition—Shire, of Dublin, made a
example, FKD’s phase 2-ready bladder can- strategic investment also.
cer treatment Instiladrin delivers the gene With many imponderables remaining—pric-
encoding interferon-a2b, prompting the ing, partnering and the number of patients—it
endogenous generation of the anti-cancer is too soon to say what the financial returns will
drug for seven days after a single treatment. be, but says Aldag, “We can tell people we’re
In comparison, when interferon-a2b is No. 1; we’ve done it. Gene therapy is something
injected directly, it is cleared from bladder the EMA permits us to bring to patients.”
cells within 24 hours. Nuala Moran, London

nature biotechnology volume 30 number 9 SEPTEMBER 2012 809


NEWS

in brief Regulatory fog lifts on obesity drugs


Bayer acquisition spotlights In the span of a summer month, the US Food metic inhibitor of prohibitin-1 receptor, adi-
biopesticides and Drug Administration (FDA) approved its potide, in phase 1 testing. “When we licensed
The July acquisition of AgraQuest, a 16-year-old first two obesity drugs in 13 years, a period adipotide from MD Anderson a year and a
biopesticide company, by Bayer CropSciences in which it rejected any weight loss drug put half ago, big pharma didn’t want to talk about
(BCS), provides validation to a sector that has before it and withdrew products from the [partnering] because of regulatory struggles.”
attracted interest from major agbiotech firms. market due to psychiatric and cardiovascular There is a striking difference “between now
The Monheim am Rhein, Germany–based
safety concerns. The recent approvals have and even six months ago as to how big pharma
agriculture giant agreed to acquire AgraQuest
for $425 million plus milestones. This follows rejuvenated a field that had been thin on hope companies are approaching the obesity mar-
a series of deals between the biotech and large and brought pharma and investors back into ket,” he says.
agriculture firms, such as BASF and Monsanto, the space (Table 1). The increased attention is welcome, but
and Bayer’s 2010 acquisition of the Israeli That’s not to say safety concerns around for obesity drugs to thrive, they must move
biopesticide firm AgroGreen. The news is
obesity drugs have been put to rest. Indeed beyond the mistakes and modalities of the
welcome to the Davis, California, area, where
the company’s R&D site is expected to expand.
the two drugs—Belviq (lorcaserin) from past. Fenfluramine and dexfenfluramine—two
The acquisition of AgraQuest gives BCS San Diego-based Arena Pharmaceuticals, nonspecific serotonin (5-hydroxytryptamine;
access to the company’s ‘green’ product lines, approved June 27, and Qsymia (topira- 5-HT) receptor antagonists—were pulled from
which will be integrated into its platforms for mate/phentermine) from Mountain View, the market in 1997 after being associated with
crop protection and pest control. AgraQuest California–based Vivus and partner Tokyo- valvular heart disease.
touts its line of fungicides and insecticides as
based Eisai Pharmaceuticals, approved July Belviq is a derivative of nor-dexfenflura-
comparatively immune to resistance problems.
Standard single-chemical agents attack only 17—were delayed by nearly two years, in part mine. Unlike its parent molecule, it selec-
© 2012 Nature America, Inc. All rights reserved.

one pathway, whereas microbiological agents owing to concerns about cardiovascular side tively activates 5-HT2C receptors rather than
produce a variety of active compounds. effects. But Thomas Hughes, president and the 5-HT2A and 5-HT2B receptors, which are
AgraQuest’s flagship product, Serenade, CEO of Cambridge, Massachusetts–based thought to be involved in hallucinogenic side
contains a patented strain of Bacillus subtilis
Zafgen, which is developing a drug for obesity, effects and heart valve defects, respectively.
that produces more than 30 active lipopeptides,
according to company literature. These protect
says the “fog has lifted.” Qsymia is an extended release formulation
against a variety of fungi and bacteria, including “There was a period of many years of fail- of two existing drugs, topiramate and phen-
fire blight, botrytis, sour rot, rust, sclerotinia, ure in the pharma space, unwillingness in the termine. Both Belviq and Qsymia target the
powdery mildew, bacterial spot and white mold. regulatory space to approve obesity drugs, central nervous system (CNS) and are attached
Mark Faust, a consultant and principal and there were companies struggling to get to extensive post-marketing cardiovascular
of Cincinnati-based Echelon Management
approval. I would meet people who said they studies. Another drug, Contrave (naltrexone/
International, sees value in companies that
establish synergies among approaches versus would never invest in obesity,” Hughes says, bupropion) from La Jolla, California–based
those trying to create magic bullets. “Magic adding that now the investor climate is “abso- Orexigen Therapeutics, also targets the CNS,
bullets don’t work forever. Evolution happens. lutely different.” and the FDA required a cardiovascular out-
Things change. I think these synergistic Big pharma began showing interest in the comes trial to consider approvability because
approaches tend to last longer and be more
winter, after the FDA Advisory Committee of an increase in blood pressure and pulse rate
resistant to evolution,” he says. Jim Kling
voted 20–2 on February 22 in favor of approv- seen in its clinical program. The company
ing Qsymia, says Christopher Anzalone, expects to resubmit its New Drug Application
president and CEO ofObesityPasadena, trends
CA–basedamond
(NDA)U.S.basedadults
on an interim analysis of that
npg

in their words Arrowhead Research, which has BRFSS, 1990, 2000, 2010 second half of 2013.
a peptidomi- safety trial in the
(*BMI ≥30, or about 30lbs. overweight for 5´4˝ person)
“We employ 5,000
1990 2000
people in China
already, a few years
ago there were only a
few hundred. …And
that’s good for Basel.”
Severin Schwan,
Roche’s CEO offers
scant consolation for 2010
the thousand made
redundant at Roche’s
shuttered R&D center in Nutley, New Jersey.
(Sonntags Zeitung, 29 July 2012)
“The point is not for doctors to castigate people,
but to understand how people are responding
to their treatments.” George Savage, co-founder
of Proteus Digital Health of Redwood City,
Calif., describes the potential of his company’s
No data <10% <10–14% 15–19% 20–24% 25–29% ≥30%
recently approved pill/sensor, which signals
a smart phone after ingestion, to impact
compliance.. (International Science Times,
Obesity trends among US adults in 1990, 2000 and 2010. Obese is characterized as a body mass
1 August 2012)
index (BMI) >30. Source: Center for Disease Control, Atlanta

810 volume 30 number 9 SEPTEMBER 2012 nature biotechnology


news

Table 1 Selected compounds in development for obesity—phase 2 and beyond


Company name Product (modality) Molecular target Latest stage of development
Orexigen Therapeutics Contrave naltrexone SR/bupropion SR Mu opioid receptor (OPRM1) (MOR) Registration
(small-molecule antagonist)
Novo Nordisk (Bagsvaerd, Denmark) Victoza liraglutide (peptide agonist) GLP-1 receptor Phase 3
Shionogi (Osaka) Velneperit (small-molecule antagonist) Neuropeptide Y receptor Y5 (NPY5R) Phase 2b (in Japan)
TransTech Pharmaceuticals, High Point, NC HPP404 (small-molecule antagonist) Histamine H3 receptor (HRH3) Phase 2
NT Life Sciences (Kadmon & Nano Terra KD026 (small-molecule antagonist)* Microsomal triglyceride transfer protein (MTP) Phase 2
joint venture)
7TM Pharma (Horsholm, Denmark) Obineptide (small-molecule Neuropeptide Y receptor Y2 (NPY2R) and Phase 1/2
antagonist)* Neuropeptide Y receptor Y4 (NPY4R)
TM30339 (small-molecule Neuropeptide Y receptor Y4 (NPY4R) Phase 1/2
antagonist)*
*Not in active development.

To some degree these side effects are to the production of new fatty acid molecules by Tartaglia, who also is a partner at Third
be expected in drugs targeting the CNS. To the liver and helping convert stored fats into Rock Ventures in Boston expects to see more
move beyond that “the next step tends to be energy. Combined data from two phase 1 stud- investment, including from his firm, in the
taking lower doses of those drugs and com- ies in 28 obese women showed that beloranib obesity space. Beyond Ember, Third Rock
bining them together to increase efficacy resulted in 4.3 kg weight loss versus 0.6 kg has invested in Zafgen and Boston-based
© 2012 Nature America, Inc. All rights reserved.

while decreasing side effects—we saw that in weight gain for placebo over 25 days. Rhythm Pharmaceuticals, whose RM-493,
hypertension and it was a step forward,” says Zafgen’s Hughes says the weight loss seen a small-peptide melanocortin 4 receptor
Frank Greenway, professor and chief of the in those trials, as well as preclinical studies (MC4R) agonist, is in phase 1 testing. The
outpatient clinic at Pennington Biomedical in dogs, was about 1 kg/week and was linear. trick to investing in these companies, he
Research Center in Baton Rouge, Louisiana. The company plans to start a phase 2 trial of says, is to “derisk early.” The firm spent about
“That’s kind of where we are now with obesity the compound in 100–120 obese patients in “one and a half years using seed money to
drugs. Where we want to get to is drugs that September. Zafgen, however, has the addi- do studies to make sure there wouldn’t likely
act on the periphery, where the disease is, like tional hurdle of being a first-in-class medi- be cardiovascular risks before committing
angiotensin receptor blockers in hyperten- cine, and it will need to blaze a regulatory $34 million in a series A round last December”
sion.” path of its own. “So we need to figure out to Ember, Tartaglia says.
The newer obesity drugs still in devel- how to do that without spending half a bil- There are safety risks beyond cardiovascular,
opment plan to do just that. Arrowhead lion dollars and taking seven years,” he notes. of course. 7TM Pharma of Lungby, Denmark,
Research’s peptide adipotide has two func- Steven Smith, co-director of transla- is focusing on TM38837, an antagonist of
tional domains: one targets prohibitin-1 tional research at the Sanford-Burnham peripheral cannabinoid receptor type 1 (CB1).
receptor, which is found on the surface of Medical Research Institute’s Diabetes and At least four CB1 antagonists—including two
adipose vascular endothelial cells but not Obesity Center in Orlando, Florida, says that had completed phase 3 testing and a third
on the surface of other cells, and the sec- MetAP2 “is a target I have a lot of enthusi- that was in phase 3 testing—were shelved after
ond causes apoptosis by disrupting mito- asm around right now,” though he also likes the approach was shown to increase risk of
npg

chondrial membranes inside the cells. Ember Therapeutics’ irisin variants. Irisin is psychiatric disorders, including suicidality.
The compound thus selectively destroys a naturally occurring hormone that is pro- Christian Elling, vice president of development
the blood supply that supports the growth duced in muscle cells during exercise. The for 7TM, says the psychiatric effects were due
of unhealthy white fat. In one preclinical Boston-based company’s founders, including to the compounds’ actions of CB1 in the brain,
study, adipotide decreased the body weight Bruce Spiegelman at Harvard, Cambridge, whereas the weight loss effects were due to both
of obese mice by 30% after 28 days of daily Massachusetts, discovered that irisin also CNS and peripheral receptors. TM38837 is
injections, Anzalone says, but there was no converts white fat cells into brown fat, which currently in phase 1 trials.
effect on lean animals, “so there is no con- burns calories into heat. When fed a high- When TM38837 moves into phase 2 will
cern about wasting away if a skinny person fat diet, mice with overexpressed irisin had depend on financing, which is the company’s
takes the drug.” Data from the phase 1 trial lower weight gain than mice with normal iri- biggest challenge for developing the product.
in 39 obese prostate cancer patients, testing sin levels (Nature 481, 463–358, 2012). Elling is hopeful that the two recent approv-
tumor aggressiveness against weight loss, are Ember’s interim CEO Louis Tartaglia says als will benefit 7TM. “For our compound, we
expected in mid-2013. the company aims to begin phase 1 trials of have the whole CB1 history to contend with
Zafgen’s small-molecule beloranib also an irisin variant in 2014, and he thinks the and we need to convince investors that this
circumvents the CNS. The compound is an way to battle obesity is through polyphar- is a good idea,” he comments. The approv-
inhibitor of methionine aminopeptidase 2 macology. Ember’s drugs work on energy als “will not necessarily make it easy to get
(MetAP2). In obese people, the liver normally expenditure, and “a drug working on energy the financing, but they will help. We are still
packages circulating lipids as fat and trans- expenditure like our drug plus an appetite far away from achieving standard-of-care in
ports that fat to adipose tissue. MetAP2 inhi- suppressant [such as Qsymia or Belviq] will obesity.”
bition directly affects metabolism by reducing ultimately be the way to treat obesity,” he says. Aaron Bouchie, Ithaca, New York

nature biotechnology volume 30 number 9 SEPTEMBER 2012 811


NEWS

in brief Industry cautiously welcomes Supreme


Amylin’s three-party Court decision on healthcare overhaul
good-bye
This summer’s ruling by the US Supreme Court the same time, PPACA now at least provides
The buyout of Amylin Pharmaceuticals by
to uphold President Barack Obama’s healthcare a legal framework for the US Food and Drug
Bristol-Myers Squibb (BMS) completed in
August called for BMS to pay $31 per share
reform law has been welcomed by the drug Administration (FDA) to come up with a path-
and gather up ownership of Amylin for about industry. Industry executives say The Patient way for biosimilars manufacturers.
$5.3 billion. But the uniquely structured Protection and Affordable Care Act (PPACA), How much the drug market will expand in
deal also brought in a third party (London- which requires most Americans to obtain 2014 when PPACA comes into force is unclear.
based AstraZeneca) and added a $1.7-billion health insurance and
payout by BMS to cover both Amylin debt and includes a handful of
a contractual obligation to Eli Lilly, putting provisions that affect
the total deal value at $7 billion.
drug development,
New York–based BMS immediately
packaged Amylin’s products together into
isn’t perfect and may
a new collaboration with AstraZeneca, in not ultimately ben-
which the latter pays the now wholly-owned efit biotechs’ bottom
subsidiary Amylin $3.4 billion, and any line, but the ruling
forthcoming profits and losses are split at least ends a highly

Associated Press, Getty Images


equally between AstraZeneca and BMS. fractionalized and
The assets in the collaboration are Amylin’s protracted battle.
© 2012 Nature America, Inc. All rights reserved.

GLP-1 agonists, Byetta (exenatide) and


“The biggest enemy
Bydureon (exenatide extended release);
metreleptin, being reviewed by the FDA for
of the economy
diabetes and/or hypertriglyceridemia in rare and our industry is
forms of inherited or acquired lipodystrophy; uncertainty,” says Ron
and Symlin (pramlintide acetate), approved Cohen, founder of
Although the general population remains generally split on the healthcare
for type 1 and 2 diabetes in patients with Acorda Therapeutics reform, most in biotech industry view it as a positive.
inadequate glycemic control already taking in Ardsley, New York.
meal-time insulin. The three approved “It’s good that we have closure.” At the same More than 53 million Americans are unin-
products combined to sell about $830
time, companies are scrambling to ensure that sured, and the healthcare reform law aims to
million worldwide in 2011.
The side payments to Indianapolis-based
they will be compliant with the law’s provisions cover more than half of them. To that end, the
Eli Lilly in the buyout stem from Amylin’s on reporting physician payments. law expands Medicaid to include people at or
initial development and commercialization The PPACA legislation was enacted in 2010, below 133% of the federal poverty level. This
partnership for Byetta with Eli Lilly, signed but was promptly challenged by 26 states and a bodes well for companies like GlycoMimetics
in 2002, with a total potential value of $300 trade group for small businesses. In a 5–4 deci- in Gaithersburg, Maryland, which is develop-
million. But in May 2011, Amylin sued Lilly sion on June 28 this year, the Supreme Court ing a small-molecule drug to treat vasoocclusive
for engaging in anticompetitive acts and for deemed the law constitutional, upholding it crisis of sickle cell disease. Most people with
breaching that original 2002 agreement,
almost in its entirety. The court based its deci- the disease access treatment through Medicaid
as Lilly earlier in the year had formed a
sion on the US Congress’s power to impose and Medicare, says Rachel King, CEO of
npg

collaboration with Boehringer Ingelheim in


Ingelheim, Germany. The two companies were taxes. The law requires individuals to obtain GlycoMimetics. An expansion of Medicaid
to jointly develop and commercialize two health insurance by 2014 or pay a penalty, and would be a way to give more people access to
oral diabetes agents, Boehringer Ingelheim’s that financial penalty “may reasonably be char- the company’s drug candidate, if it is approved.
linagliptin and BI10773, and Lilly’s two acterized as a tax,” Chief Justice John Roberts But although the Medicaid expansion was
basal insulin analogs, LY2605541 and wrote in the majority opinion. “Because the initially mandatory for states, the Supreme
LY2963016, with an option to co-develop Constitution permits such a tax, it is not our Court in its decision deemed the provision
and co-commercialize Lilly’s anti-TGF-b
role to forbid it, or to pass upon its wisdom or overly coercive, and effectively made the pro-
monoclonal antibody. Lilly planned to use the
fairness,” he wrote. gram voluntary for states. Kings says she is
same sales force to sell both exenatide and
the direct competitor Boehringer Ingelheim’s A few key provisions stand out for innovative concerned that if states opt out, that will limit
linagliptin, and this is the detail that caused biotech companies. Most notably, drug makers’ patients’ access. Indeed, several state gover-
Amylin to file suit. The issue was resolved markets should expand as more Americans nors have already said publicly they would
when Lilly and Amylin ended the alliance become insured and gain access to medicines. not participate in the expansion, even though
and Amylin regained rights to exenatide. Small companies developing therapeutics the federal government would pay for 100%
Having full rights made Amylin an attractive also receive grants and tax credits for their of the additional costs through 2020. The US
acquisition target, except that the breakup
projects, which may prove crucial for cash- Congressional Budget Office in July estimated
called for Amylin to, among other things, pay
starved startups (Box 1). What’s more, inno- that in 2022, 3 million fewer people would be
15% of global sales of exenatide until the
sum reached $1.2 billion, plus interest. Any vative drugs will receive 12 years of exclusivity insured due to state opt-outs.
potential suitor needed to take on that long- against competition from biosimilars. The lan- Also offsetting the benefits of a bigger ros-
term financial commitment, and thus, BMS’s guage demarcating extra exclusivity for biolog- ter of insured Americans are the $80 billion
multiparty buyout deal, with money flowing in ics should provide innovative companies, and in rebates and fees drug makers must pay on
three directions. Brady Huggett the investors who place their money in them, their commercialized products to help fund the
with more certainty, say industry executives. At plan. “In that sense, it costs us money. But on

812 volume 30 number 9 SEPTEMBER 2012 nature biotechnology


news

June 2012 October 2012 January 2013 January 2014 January 2015 January 2016

Patient-Centered The Independent New funding provided PPACA in force. US One year into Federal support to
Outcomes Research Payment Advisory to state Medicaid citizens able to buy implementation states ends, which
Institute issues first Board formed aimed at programs that choose coverage directly in biopharma industry may lead certain
grants extending the life of to cover preventive affordable insurance gets first indications of cash-strapped states
Medicare Trust Fund services for patients exchanges offering how much new to opt out of Medicare
begins operation. choice of health plans business healthcare
PhRMA has stated reform will provide
elimination of board as
a top priority

the other hand, bringing additional people into transthyretin (TTR)-mediated amyloidosis, sures are meant to help spot improper industry-
the system is helpful,” says Cohen at Acorda, hemophilia and beta thalassemia. Maraganore doctor relationships, but the paperwork could
which has two commercialized products. “It says that some biotech companies could expect prove challenging for small companies with few
may end up being a wash, and that’s fine.” a market increase of 20% due to the elimination resources, Cohen says.
Biotechs are continuing to dig through the of the caps, but it will vary depending on the The law’s 2.3% tax on revenue for device
more obscure sections of the law and finding type of drug and the age of the patient popula- manufacturers is proving difficult for the
measures that affect their businesses. Sections tion. smaller companies in that sector, particularly
3001 and 3008, for example, includes incen- The law eliminates other caps as well. those that have revenues but aren’t yet profit-
tives for hospitals to reduce infections. That’s Insurance companies can’t exclude people for able. Congress included the tax in the law to
good for S. San Francisco–based KaloBios having a preexisting condition, nor can they help pay for the expansion of health coverage.
© 2012 Nature America, Inc. All rights reserved.

Pharmaceuticals, which is developing a thera- discontinue coverage after someone receives a But it is unclear yet whether the presumed mar-
peutic monoclonal antibody to prevent pneu- bad diagnosis. And children can now stay on ket expansion will offset that tax, says Samuel
monia, an infection that often occurs in people their parents’ policies until they are 26 years old. Lynch, founder of BioMimetic Therapeutics in
who are on ventilators in hospitals. “The gov- This is a key reform for companies like Sarepta Franklin, Tennessee, which makes recombinant
ernment is pushing to reduce preventable infec- Therapeutics in Cambridge, Massachusetts, human platelet-derived growth factor products
tions,” says David Pritchard, CEO of KaloBios. which is developing a therapy for people with that are regulated as devices. Lynch says the
“And we’ve got a drug that prevents infection.” Duchenne muscular dystrophy, who suffer the tax will likely force him to pass along some of
Biotech executives working in orphan, or worst effects of the disease in their teens and the additional costs to hospitals in the form of
rare, diseases say the law’s insurance market early twenties, and rarely live to see age 30. higher prices for his products. Lynch’s overall
reforms are key for them. The law bans insur- It is not all good news, however. Likely head- impression of the Act: “I’m not a fan.”
ance companies from putting annual or lifetime aches for industry in the law relate to what are In one of the more indirect effects of the law,
limits on individual policies—a boon to people known as the ‘sunshine provisions’. The Act biotech companies should expect the momen-
with continually high healthcare costs and, requires companies to report all payments to tum to pick up on comparative effectiveness,
by extension, the companies that make treat- doctors and hospitals that are over $10. “It’s the practice of governments and insurance pro-
ments for them. “Patients with rare diseases a huge administrative burden,” says Cohen, viders judging drugs based on relative health
often hit these caps early in life,” says John which hired new people as a result. “Every sin- or cost benefits. “The flip side of some of these
Maraganore, CEO of Alnylam Pharmaceuticals gle time you buy [a doctor] a lunch you have a changes and the removal of the caps is there
in Cambridge, Massachusetts, which is devel- reporting obligation,” and separate reports have will be a more active debate around health
npg

oping treatment for rare diseases, such as to be filed with each state, he says. The mea- outcomes and comparative effectiveness,” says
Maraganore. Companies developing early-stage
drugs should not only collect clinical data but
Box 1 Where are they now? also data around the economic benefits of
their candidates, he says. The Act established a
As a measure of Obama’s PPACA, legislators in 2010 allocated $1 billion for grants government vehicle for conducting compara-
and tax credits to small companies developing therapeutic products. But the sum was tive effectiveness research called the Patient-
spread out across nearly 3,000 companies and 4,600 projects, leaving most with awards Centered Outcomes Research Institute, which
of less than $1 million. For the larger of these awardees, the grant “might have been a issued its first grants in June.
percentage on one month’s burn rate,” says Glen Giovannetti, global life sciences leader Many provisions in the Act will surely be
at Ernst & Young. tweaked over time (Table 1), and the possibil-
But for Remedy Pharmaceuticals in New York, the grant money was a godsend. The
ity that parts of the law could be repealed by
$733,437 the company received through the program “moved the whole company
legislators after the November elections still
forward,” says co-founder Sven Jacobson. “It came at a time when very little money
looms. In the meantime, some small biotechs
was flowing around and people were scared to invest.” After investors partly matched
will wait to see how the larger companies com-
the award, Jacobson was able to design a small pilot study of his product. According
ply with the law. “There’s an election coming
to Jacobson, the results of this pilot helped him raise another $3.7 million from angel
up and things may change again,” says Lynch at
investors and to plan a larger study.
Biomimetic. “As a small company you want to
The Therapeutic Discovery awards were a one-shot deal, but the Biotechnology Industry
wait to the extent that you can to see how these
Organization is pushing for Congress to extend and expand the program because it turned
things are being implemented by larger compa-
out to be so popular. If there is a next time, the organization has proposed that the
nies so that [you] don’t get too far out in front.”
government should grant larger awards to fewer companies. EW
Emily Waltz, Nashville, Tennessee

nature biotechnology volume 30 number 9 SEPTEMBER 2012 813


NEWS

in brief Biotechs opt for alternative floatation strategy


Compulsory license The Jumpstart Our Business Startups
bandwagon gains momentum (JOBS) Act, passed into law in the US
India’s decision in March to grant its first-ever in April, has revitalized the practice
compulsory license, which allows a company
known as a Form 10 filing. This type
to make and market a drug that the patent
holder has not been able to make sufficiently of public floatation is considered
affordable and accessible, has drawn cheers more efficient and less risky than
from healthcare activists and opprobrium from a traditional initial public offering
the pharma sector. The licensee, Hyderabad- (IPO), and though it existed before the
based Natco Pharma, will sell its generic version JOBS Act, there has been heightened
of Bayer’s liver and kidney cancer drug Nexavar
interest in the process since the
(sorafenib) for 3% of the patented drug’s price
in return for paying 6% royalty on sales to
legislation passed. “The JOBS Act put
Leverkusen-based Bayer. Nata Menabde, India’s into the public consciousness that
representative to the World Health Organization, it’s important to find new ways to do
told CNBC-TV18, “India has taken a good IPOs,” says William Hicks, a partner
political stand on compulsory licenses and in the New York office of the law firm
we respect that move. [A compulsory license Mintz Levin, and that has led to a
is] an important tool that governments have
renewed focus on Form 10, adding, “It
in their arsenal and they should be using it
as appropriate, as per national legislation, could become the new normal for” for
and keep public health interests above any emerging growth companies.
© 2012 Nature America, Inc. All rights reserved.

other interests.” Pharma companies and In a Form 10 fundraising, a


Western patent offices such as the US Patent company prepares a prospectus
and Trademark office have registered their document to go public, but raises
opposition, and Bayer appealed the Indian The rising popularity of Form 10 filings is
letting companies consider raising funds and
money through a private placement
government’s decision. Court arguments were
heard in August. However, until the decision is going public in a stepped process. before filing the prospectus. Because
reversed, the precedent means that other Indian the private placement is predicated
companies may seek similar arrangements on also going public—and therefore
with the government. Compulsory licenses eventually having a publicly traded stock—the process broadens the investor base
have previously been used in Brazil, Thailand for the private placement by allowing the participation of so-called ‘crossover’
and South Africa, and just last month China
investors who would normally invest only in public companies that have greater
amended its intellectual property laws to allow
for compulsory licenses. Michael Francisco liquidity than those in the biotech sector.
A key characteristic of the Form 10 process is that the company decides when
its stock begins trading, as opposed to a traditional IPO, where a company goes
through a filing and disclosure process, gains momentum and then prices its shares

in their words at the end—when perhaps the climate for the deal has changed. “People are using
Form 10 as part of a conscious de-risk strategy,” says Hicks. “If the market’s not
good, with a traditional IPO you can have a horrible deal and not get what you
“We cannot on the one
thought you were going to get. With this process, you raise the money first, then you
npg

hand be marketing
our green image and go public.” (Though, much like with a shelf offering for public companies, firms
on the other hand are not required to sell shares on the open
market.)
growing GM crops.
European consumers Form 10 can also be used to create a
Using the Form 10
will not be fooled by public shell company into which a biotech process for going
this duality.” Gillian
Westbrook, manager
firm can merge—a strategy used successfully public “could become
of the Irish Organic by Los Angeles-based Cougar Biotechnology the new normal.”
Farmers and Growers (now a division of Johnson & Johnson) in
Association, after the Irish Agriculture and 2006. This is cleaner than a reverse merger
Food Development Authority approved testing with an existing company, which may carry the risk of litigation or other liabilities.
of a blight-resistant transgenic potato. (The The strategy also may also give companies more leverage in acquisition
Independent, 2 August 2012)
discussions. “Now, the threshold for doing a traditional IPO is so high,” it
“I personally believe those [biologics] are the isn’t a viable exit option for investors and potential buyers know it, says Hicks.
future of drug development in our industry,
Ovascience, a fertility company in Cambridge, Massachusetts; KaloBios
they just need time to continue to germinate.”
Following AstraZeneca’s $15.6 billion Pharmaceuticals, an antibody developer in San Francisco; and Fresh Medical
acquisition and doubling of the workforce Laboratories, a diagnostics company in Salt Lake City, Utah, are among the
at MedImmune, CEO Peter Greenleaf of the companies with recent Form 10 filings. If those and others are successful, Form
Gaithersburg, Md., biotech emphasizes the 10 could provide a consistent, alternative exit path for investors, and help small
need for a long term perspective, given just one biotechs put pressure on buyers in negotiations.
approved product (a swine flu vaccine) and no Mark Ratner, Cambridge, Massachusetts
drug programs in late development. (Washington
Post, 5 August 2012)

814 volume 30 number 9 SEPTEMBER 2012 nature biotechnology


news

GSK buys partner Human Genome Sciences


in brief
London-based GlaxoSmithKline
India’s biosimilar regulations
New guidelines for “similar biologics”
(GSK) acquired for $3.6 billion
launched August 15 will help address local
long-time collaborator Human patient access to expensive drugs while serv-
Genome Sciences (HGS), ing to attract international manufacturers to
the developer of Benlysta India. The Department of Biotechnology and
(belimumab), a human Central Drugs Standard Control Organisation
antibody that targets BLyS developed the regulatory framework with input
from industry and academia. Under the new
(B lymphocyte stimulator)
guidelines, manufacturers must prove similar-
approved in 2011 as the first ity to a reference biologic already approved
new drug for lupus in almost in India or licensed and sold for at least four
50 years (Nat. Biotechnol. 29, years in a regulated market. The biosimilar
292, 2011). The partnership, should be comparable to the innovator drug in
which began in 1993, has safety, efficacy and quality as demonstrated
by analytical and clinical trials and the pre-
generated two other compounds
clinical and clinical studies should also be
now in late-stage trials at comparative in nature. Approval “without
GSK: darapladib, an inhibitor The buyout of HGS continues a run on big biotechs
involved clinical trials” is possible if manu-
of lipoprotein-associated by pharma over the last several years. Source: Human
facturers prove close similarity to its reference
Genome Sciences
phospholipase A2 discovered by product, and show consistency in production
GSK using HGS technology, for process. Anurag Rathorem at Indian Institute
© 2012 Nature America, Inc. All rights reserved.

treating cardiovascular diseases including atherosclerosis, and albiglutide, a glucagon- of Technology in New Delhi, who contributed
to the guidance says, “We took both the US
like peptide-1 (GLP-1) receptor antagonist for type 2 diabetes, created by HGS using
and European guidelines into consideration
albumin-fusion technology and licensed to GSK in 2004. GSK initially launched a $13 while drafting.” The guidelines have received
per share tender offer for HGS in April 2012. HGS’ management advised shareholders a cautious welcome from industry. “The
to reject that bid, claiming among other things that it did not adequately reflect the requirement for comparative clinical trials
value to HGS of the three products, though it also cited synergies GSK would obtain in will certainly have an impact on the compa-
the buyout and the benefit of HGS’ $2.6 billion in net operating loss carryforwards and ny’s budget allocation,” says Geena Malhotra,
chief of research at Mumbai-based Cipla.
R&D tax credits. HGS opened the door to competing offers from other companies, but
Recent deals between Mumbai-based Emcure
GSK then upped its bid to $14.25 per share, or the final $3.6-billion price. “It was a Pharma and Basel-based Roche and between
foregone conclusion that HGS was going to be acquired by GSK,” given the terms of Merck Serono and Dr. Reddy’s Laboratories in
their partnership, says David Brindley of The European Centre for Accelerating Medical Hyderabad are a testament to India’s growing
Innovations at the University of Oxford. “From an investors’ point of view, I’d go as attractiveness as a biosimilars manufacturing
far as to say it was formulaic.” Had HGS been as valuable to anyone else, “there’s no hub. Killugudi Jayaraman
way that there would not have been a bidding war,” he says. With its global marketing
infrastructure, GSK may be in better position to maximize the value of Benlysta. With net Myriad’s patents redux
sales of $52.3 million in 2011 and $31.2 million in first-quarter 2012, HGS’ launch of The biotech industry breathed a collective
the drug had fallen below analyst estimates. sigh of relief with the news that on August
If not for GSK’s support, Benlysta might never have seen the light of day. The 16 the US Court of Appeals for the Federal
npg

Court in New York upheld the Myriad patents


companies began working together in 1993 when GSK (then SmithKline Beecham)
on BRCA1 and BRCA2. The court came to the
committed up to a whopping $125 million for rights to HGS’ mRNA-based discovery same 2–1 decision in 2011, but was asked
process, among the first industry models for a broad-based genomics approach linking to revisit the case by the Supreme Court,
protein discovery with disease. But over the next decade, despite the funding from GSK following its ruling on a related case. (In
and a series of large public stock offerings, HGS’ own product development efforts Mayo Collaborative Sciences v. Prometheus
foundered. Then, in 2005, GSK became HGS’ partner for Benlysta’s development, Labs, the Supreme Court found a certain
remaining stalwart even after a failed Phase 2 trial a few months later threatened the diagnostic to represent a law of nature, and
hence not be eligible for patent.) As before,
compound’s future and initiating the first of two dramatically successful Phase 3 studies
the appeals court reversed two earlier district
in 2007 (Nat. Biotechnol. 27, 779, 2009). Mark Ratner, Cambridge, Massachusetts court findings—that Myriad’s DNA test are
products of nature, and that a method for
screening potential cancer therapeutics
employs basic scientific principles. Judge
in their words Kimberly Moore wrote that Congress has
authorized an “expansive scope of patentable
“I never thought this at the university for developing chimeric T cell subject matter…and the USPTO [US Patent
would happen, that cancer therapy. (Bloomberg Businessweek, and Trademark Office] has allowed patents on
the pharma industry 7 August 2012) isolated DNA sequences for decades,” saying
would get into ultra- “Out of our research in France, we haven’t really it is a matter of policy, not for the courts to
personalized therapy.” developed a new molecule in 20 years.” decide. Dan Vorhaus, editor of The Genomics
The University of Chris Viehbacher, Sanofi’s CEO, explains why Law Report, says “Those are the types of extra-
Pennsylvania’s Carl he is firing workers in Toulouse and Montpellier legal policy-oriented questions that are at the
June after Novartis much to the chagrin of French politicians, heart of the Myriad litigation and will not be
penned a $20 million labor leaders and researchers. (Bloomberg decided by litigation...[but] by policy makers.”
deal to fund a center Businessweek, 9 August 2012) he says. Laura DeFrancesco

nature biotechnology volume 30 number 9 SEPTEMBER 2012 815


NEWS

in brief
NIH injects $275 million gain understanding of the mechanisms underlying yuan ($95 billion) in three stages by 2020 to
into undiagnosed diseases these conditions, and improve their diagnosis and attain the world’s leadership in key fields. In
treatment. An estimated 6% of the US population a report presented on June 27 to the National
and RNA research suffers from such conditions. The Extracellular People’s Congress, China’s legislature, Auditor
In July, the US National Institutes of RNA Communication program, meanwhile, General Jiayi Liu described an audit of 8 of the
Health (NIH)’s Common Fund announced will explore how RNA is released from cells 16 S&T schemes, although he did not reveal
pledges of $145 million for research into and packaged for transport, and then interacts which projects had been audited. According to
undiagnosed diseases and $130 million to with and influences specific cell types and Liu’s report, 134 projects failed to meet their
study extracellular RNA communication. functions. This offers a chance to “transform our milestones, yet continued to receive funding for
These areas represent “key roadblocks understanding of endocrinology and intercellular their next stages. Furthermore, 582 of the
in biomedical research and…emerging communication,” Wilder says, as well as provide 2,401 sets of research results randomly
scientific opportunities ripe for Common opportunities for diagnostics and therapeutics. examined by CNAO had been replaced by results
Fund investment,” according to Elizabeth This initiative will launch in fiscal 2013 and run from other projects. For the drug development
Wilder, director of the NIH’s Office of Strategic for five years. Further details on the Undiagnosed scheme, the central government invested nearly
Coordination, which manages the program. Diseases Program are expected in fiscal year 6 billion yuan ($882.5 million) in support of
The Common Fund, established in 2006, is 2014. Malorye Allison more than 900 drug development projects
a $545-million pot of discretionary money, to and technology platforms by the end of 2011.
support research that cuts across institutes. A Tianjin-based bio-entrepreneur, Zailin Yu,
The NIH’s Undiagnosed Diseases Program China’s key R&D programs president of SinoBiotech, whose project was
(UDP) joined the National Human Genome
Research Institute, the NIH Office of Rare
behind schedule among those audited by CNAO, says “The
reason [that so many projects failed to reach
Diseases Research and the NIH Clinical Center Ninety three per cent of projects that comprise their milestones] is because the scheme’s goal
© 2012 Nature America, Inc. All rights reserved.

together in an effort to serve patients with China’s National Key Science and Technology to develop the world’s leading innovative drugs
mystery diseases that have eluded diagnoses. (S&T) Schemes had not finished their tasks by is unrealistic.” Rigid financial management and
Since its inception in 2008, more than 200 2011, the end of the first stage, according to late allotment of funding also contribute to the
cases have been tackled by the UDP. With the Chinese National Audit Office (CNAO) in failure to finish tasks, he adds. Liu reported that
the new funding, UDP will go extramural (the Beijing. Launched in 2009 and funding a range institutions and researchers involved in the
projects up until now have been intramural) to of activities from innovative drug development audit have promised to correct their problems
establish a network of medical research centers to airplane manufacturing, China’s 16 key and improve management of their research
that will leverage advances in genomics to S&T schemes plan to spend some 600 billion funding. Hepeng Jia

Around the world in a month

KAZAKHSTAN
Scientists at the National
Biotechnology Center in Kazakhstan
develop carrots producing a recombinant
tuberculosis vaccine. Laboratory tests in animals
npg

are said to show eating the carrots improves


tuberculosis immunity.
CHINA
Researchers from the Beijing
University of Agriculture report
breeding two female cloned calves with
adipocyte fatty acid binding protein in their
meat. The binding protein is supposed to
enhance the taste of the meat.

MALAYSIA
Researchers in Malaysia and Japan describe
a method of making ethanol from the empty
fruit bunches left over from palm oil production. The
empty bunches are an abundant waste product of palm
BRAZIL oil production in Southeast Asia, with 15 million metric
Vienna-based See Algae Technology tons collected annually in Malaysia alone.
plans to construct what it says will be
the world's first seaweed-to-biofuel production
plant in the Brazilian state of Pernambuco. The
$9.8-million facility will be built next year and will
use carbon emissions from ethanol production
to grow the seaweed.

816 volume 30 number 9 SEPTEMBER 2012 nature biotechnology


data page

Drug pipeline: Q212


Laura DeFrancesco

In the first two quarters of 2012, there have been 12 approvals agent for Alzheimer’s, Amyvid, were of note. Sodium glucose trans-
(NME and biologics) compared with 30 for 2011. The appovals of porter 2 inhibitors continue to impress in the clinic. But trials of
two obesity drugs, Qsymia and Belviq, a plant produced replace- Pfizer’s flagship Alzheimer’s therapy bapineuzumab were finally
ment enzyme for Gaucher’s disease, Elelyso, and a PET imaging suspended.

Notable regulatory approvals (Q2 2012)


Drug/company Indication Approval and drug information
FDA approvals by lead indication area
Amyvid (florbetapir Alzheimer’s 4/6/12, FDA. 18F-labeled small molecule allows detec-
70
F-18)/Eli Lilly imaging disease tion of amyloid beta in patients through positron
emission tomography
SOM230/Novartis Cushing’s 4/25/12, Europe. Cyclohexapeptide somatostatin analog 60
syndrome with unique ability to bind somatostatin receptor subtypes

Number of FDA approvals


1-5 50
Stendra (avanafil)/ Erectile 4/12/12, FDA. Phosphodiesterase 5 inhibitor with faster
Vivus dysfunction onset than marketed drugs 40
Elelyso (taliglucerase- Gaucher’s 5/1/12. Human glucocerebrosidase expressed in carrot
alpha)/Pfizer disease cell line 30
Perjeta (pertuzumab)/ Breast cancer 6/8/12, FDA. Humanized mAb that inhibits HER2
Roche dimerization by binding different epitope from Herceptin 20
Belviq (lorcaserin Obesity 6/27/12, FDA. Small-molecule selective agonist of only
hydrochloride)/Arena 5-hydroxy tryptamine 2C receptor subtype
10
Pharmaceuticals
© 2012 Nature America, Inc. All rights reserved.

Myrbetriq Overactive 6/28/12, FDA. Small-molecule beta 3 adrenergic agonist,


(mirabegron)/ bladder relaxing smooth muscle 0

11
95

96

97

98

99

00

01

02

03

04
05

06

07
08

09
10

*
12
Astellas Pharma

20
19

19

19

19

19

20

20

20

20

20
20

20

20
20

20
20

20
Qsymia (phentermine Obesity 7/12/12, FDA. Amphetamine appetite suppressor and Year
and topiramate)/Vivus anticonvulsant with unknown mechanism in obesity
Source: BioMedTracker, a service of Sagient Research (http://www.biomedtracker.com/). FDA, US
■ Infectious disease (108) ■ Cardiovascular (47) ■ Ophthalmology (20) ■ Obstetrics/gynecology (13)
Food and Drug Administration; EMA, European Medicines Agency.
■ Oncology (83) ■ Endocrine (39) ■ Psychiatry (20) ■ Dermatology (9)
■ Autoimmune/immunology (47) ■ Hematology (21) ■ Gastroenterology (17) ■ Rheumatology (4)
Notable regulatory setbacks (Q2 2012) ■ Respiratory (16) ■ Other (98)
■ Neurology (54) ■ Metabolic (24)
Drug/company Indication Setback summary
Setipiprant/ Allergic rhinitis
4/2/12. Phase 3 study of small-molecule antagonist of
Actelion chemoattractant receptor-homologous molecule expressed Source: FDA; BioMedTracker, a service of Sagient Research (http://www.biomedtracker.com/).
on TH2 cells (CRTH2) failed to show efficacy *2012 partial year from Jan. 1 to Aug. 11. Numbers in parentheses after legends are total
Vyndaqel (tafamidis Familial amyloid 6/18/12. FDA issued a complete response letter requesting approvals since 1995.
meglumine)/Pfizer polyneuropathy more efficacy data, despite EMA approval. Small molecule,
originally developed by FoldRx, stabilizes transthyretin,
inhibiting fibril formation
BMS-986094/ Hepatitis C 8/1/10. Phase 2b trial of nucleoside inhibitor of
Bristol-Myers RNA-dependent RNA polymerase (NS5b), originally
Squibb developed by Inhibitex, halted due to safety issues Notable trial results (Q2 2012)
Bapineuzumab/ Alzheimer's 8/6/12. Phase 3 trial of anti-beta amyloid humanized mAb Drug/company Indication Results
Pfizer and Johnson disease discontinued after failing to meet co-primary endpoints Canagliflozin/Johnson Diabetes Phase 3 results of this sodium glucose transporter 2
& Johnson & Johnson inhibitor improved glycemia in patients with moder-
Talactoferin/ Non-small cell 8/6/12. Phase 3 trial of dendritic cell activator (recombi- ate renal impairment (Am. Diabetes Assoc., Abstract
Agennix lung cancer nant human lactoferrin) suspended when overall survival of 41LB, 2012)
patients did not improve Dapagliflozin/Bristol- Diabetes Phase 3 results of this sodium glucose transporter 2
npg

Cytofab Sepsis 8/8/12. Company halted further development on this sheep Myers Squibb inhibitor demonstrated that 10% of TD2 patients with
(AZD9773)/ polyclonal anti-TNF alpha antibody after phase 2b failed to cardiovascular disease achieved three endpoints (Am.
AstraZeneca show improvement over placebo Diabetes Assoc., Abstract 1114-P, 2012)
Ganitumab (AMG Pancreatic 8/8/12. Company halted phase 3 trial when human mAb Amatuximab/Esai Mesothelioma Phase 2 results of chimeric mAb against mesothelin dem-
479)/Amgen cancer against insulin-like growth factor-1 was unlikely to meet onstrated 39% with partial response, 51% with stable
overall survival endpoint disease with mean progression free survival of 6.1 months
Source: BioMedTracker, a service of Sagient Research (http://www.biomedtracker.com/). FDA, US (ASCO, Abstract 7030, 2012)
Food and Drug Administration; EMA, European Medicines Agency. AMG386/Amgen Renal cell carci- Phase 2 results of fusion protein (peptibody) targeting
noma angiopoietins demonstrated overall response rate of
58-59% (ASCO, Abstract 4606, 2012)
Notable upcoming regulatory decisions (Q4 2012) GLPG0634/Galapagos Rheumatoid Phase 2 results of selective JAK1 inhbitor showed over-
Drug/company Indication Upcoming catalyst arthritis all improvement in ACR20 in 83% of patients receiving
Glybera (alipogene Lipoprotein 9/25/12, EMA. Glybera, a Ser447X variant of human drug versus 33% receiving placebo (EUNETHYDIS Int.
tiparvovec/UniQure lipase deficiencylipoprotein lipase gene delivered by AAV vector Conf., Abstract OP0263, 2012)
Linaclotide/Ironwood Irritable bowl 9/8/12, PDUFA. Linaclotide, a 14- mer peptide Ipragliflozin/Astellas Diabetes Phase 2 trial of this sodium glucose transporter2 inhibi-
syndrome agonist of guanylate cyclase-C that increases fluid tor shows that patients reduce A1C and body weight
secretion into the intestine (Am. Diabetes Assoc., Abstract 1046-P, 2012)
Metreleptin/Amylin Lipodystrophy 10/3/12, PDUFA. Metreleptin is recombinant human Miravirsen/Santaris Hepatitis C Phase 2 trial of oligonucleotide targeting MIR-122
leptin, which affects blood glucose and triglycerides demonstrated prolonged dose dependent reduction in
Jetrea/Thrombogenics Thyroid cancer 10/17/12, PDUFA. Jetrea is recombinant microplas- HCV RNA (Eur. Assoc. Study Liver, Abstract 58, 2012)
min, a truncated stable form of plasmin PF-332991/Pfizer Sarcoma Phase 2 results of cyclin dependent kinase 4/6 inhibi-
Enzalutamide/Astellas Prostate cancer 11/21/12, PDUFA. Enzalutamide, an androgen recep- tor demonstrated 70% progression free survival at
tor antagonist of higher affinity than bicalutamide 12 weeks (ASCO, Abstract 10002, 2012).
Cabozantinib/Exelixis Thyroid cancer 11/29/12, PDUFA. Cabozantinib inhibits hepatocyte Yttrium-90 clivatu- Pancreatic Phase 2 trial of mAb targeting mucin antigen labeled
growth factor receptor MET, RET and vascular endo- zumab tetraxetan/ cancer with yttrium-90 demonstrated disease control in 58%
thelial growth factor receptor 2 Immunomedics of patients in 7.7 months (Soc. Nucl. Med., Abstract
Raxibacumab (ABthrax)/ Anthrax 12/15/12, PDUFA. Abthrax, a human monoclonal anti- 495, 2012)
Human Genome Sciences body that blocks anthrax from binding to cell surface Source: BioMedTracker, a service of Sagient Research (http://www.biomedtracker.com/). FDA, US
BG-12 (dimethyl Multiple sclerosis 12/28/12, PDUFA. BG-12, a second-generation Food and Drug Administration; EMA, European Medicines Agency.
fumarate)/Biogen Idec fumarate that is immunomodulatory
Kynamro (mipomersen Dyslipidemia/ 12/31/12, EU. Mipomersen, 2ʹ-O-(2-methoxy)
sodium)/Sanofi hypercholester- ethyl-modified ribose antisense oligo against
olemia apoB-100
Source: BioMedTracker, a service of Sagient Research (http://www.biomedtracker.com/). FDA, US
Food and Drug Administration; EMA, European Medicines Agency. Laura DeFrancesco is Senior Editor, Nature Biotechnology.

nature biotechnology volume 30 number 9 SEPTEMBER 2012 817


N E W S fe at u re

Patient power ‘Why can’t we do it? Maybe it’s time we became


more aggressive’,” says Beall.

Patient foundations are not only exploring new funding models but Breaking with tradition
Traditionally, foundations have focused on
also catalyzing translational research, with notable successes. patient education and support, allocating a rela-
Jim Kling reports. tively small portion of their budgets to research.
“We don’t behave quite like the legacy organi-
zations,” says MRF’s Johnson. When legacy
A few weeks ago, the US Food and Drug neurological space, but they realized that the foundations handed out research money, they
Administration (FDA) approved Kyprolis pathway overlapped with work we were doing. followed the US National Institutes of Health
(carfilzomib), an irreversible proteasome inhib- So they came to us to test compounds in our (NIH) model, accepting applications for basic
itor developed by Onyx Pharmaceuticals of assays,” says Scott Johnson, president, CEO and research funds but then backing off from financ-
S. San Francisco, California. The success of founder of the MRF. ing further development work. But now, some
this drug program owes a great deal to the The MRF doesn’t stop at R&D. It also ensures foundations see their role as not just shepherd-
Multiple Myeloma Research Foundation that discoveries have ing drug candidates
(MMRF) of Norwalk, Connecticut, which patent protection to along the tangled path
supported carfilzomib from its earliest stages. encourage further to drug development
It is indicative of the growing credibility of investment. Johnson but actually changing
an increasing number of patient foundations says the foundation the way translational
that are galvanizing translational research, has encountered little research and clini-
© 2012 Nature America, Inc. All rights reserved.

particularly in areas poorly supported by the resistance from mem- cal development are
private sector. Unwilling to wait for academic ber universities. Most done. “What’s dif-
groups to move basic research into the clinic, have limited budgets, ferent about us and
or for industry to swoop in, these foundations forcing university pat- a few other, newer
are bankrolling preclinical development and ent offices to prioritize organizations is that
beyond. Not only are they compensating for their efforts in areas, The new philanthropist. Kathy Giusti, founder of we look at the entire
gaps in existing traditional funding sources, such as hardware and the MMRF, participates in a plenary session at a continuum all the
Partnering for Cures meeting in November, 2011.
but they are also exploring new models to software, that have way from discovery
Source: MMRF.
coordinate research efforts, improve clinical faster payoffs than biology that might
trial recruitment, capture intellectual property drug research. The be done in academic
(IP), overhaul biobanking and counter data foundation pays for patent submissions and labs, to thinking about how to cross the transla-
hoarding by academics and companies. splits royalties with the university. “It’s a no- tion gap, to working with pharmaceutical com-
brainer. It’s almost like new money that they panies to move drugs forward,” says Johnson.
Found in translation wouldn’t have gotten at all,” says Johnson. Formerly a serial entrepreneur and business
Like the MMRF, the Myelin Repair Foundation The inspiration for these activities derives consultant, Johnson founded the MRF in 2002
(MRF) of Saratoga, California, is taking a more in part from the success of the Cystic Fibrosis with a patient’s perspective; he has been living
hands-on approach than a traditional patient- (CF) Foundation of Bethesda, Maryland, which with MS for over 30 years. When he became
centered foundation would. The foundation invested about $75 million in Kalydeco (iva- more active in MS research, he was surprised
npg

brought together the leading research groups caftor), the Vertex Pharmaceuticals drug that to find that no one had a research plan. “This
in the field to form a network—the Accelerated was approved by the FDA in January. In the was one of the most stunning things to me.
Research Collaboration—with a goal of getting 1990s, the foundation recognized that indus- Coming from business, you have a business
a drug into the clinic by 2014. Just last June, it try wouldn’t get serious about the field without plan. I assumed if you were doing research that
announced the initiation of a phase 1 clinical more incentives, so it took several steps: it cre- you had a research plan to attack this disease,”
trial—an autologous stem cell therapy devel- ated a network of patients at clinical sites in the he says. Under his guidance, the MRF has devel-
oped by Robert Miller at Case Western Reserve Seattle area to assist with the organization of oped its own plan. “We sat down our experts and
University in Cleveland, Ohio—putting them clinical trials; and it established a screening proj- defined what is known, and prioritized ques-
several years ahead of schedule. ect for oral drug candidates that could open up tions to answer, and that became the basis for
In addition to the $45 million that the MRF the abnormal ion channels that plague patients the initial experiments that we funded.”
has raised, some 80% of which has been handed with CF. In addition to the $75 million in sup- In fact, these agencies are being run like for-
out to the eight academic centers compris- port of Kalydeco, the foundation spent an addi- profit agencies, with deals structured more like
ing their network, the foundation unveiled in tional $70 million in partnership with Vertex those of a business than a charity. As the CF
January its own Translational Medicine Center, to develop a second-generation drug. “Vertex Foundations’ Beall describes it: “We don’t just
where myelin repair assays can be performed would tell you that they wouldn’t be in the cystic give them the money: we have milestones and
on promising drug candidates coming from fibrosis area if it were not for the foundation de- they have incentives to perform. If they don’t
anywhere, including industry. For example, in risking their early efforts,” says Robert J. Beall, perform we can walk away, and we have reach-
February, the foundation signed a deal with president and CEO of the CF Foundation. in rights we can use [if we walk away]. We get
a company, ENDECE Neural of Mequon, “A lot of these [patient foundation] boards royalties back, and use them to fund other ven-
Wisconsin, to test its compound libraries for the are run by parents. They see that the CF ture philanthropy.”
ability to reverse myelin damage. “They were a Foundation [has helped develop] a drug that Another nonprofit, the CHDI Foundation,
cancer company with no thought of entering the changes the course of a disease, and they ask, which works with Huntington’s disease

818 volume 30 number 9 September 2012 nature biotechnology


news fe at u re

Table 1 How foundations are catalyzing R&D progress


clinical progress and to assess therapeutic
efficacy. The Michael J. Fox Foundation is
Foundation Initiative Description
addressing this with a five-year, $45 mil-
MMRF CoMMpass Provides comprehensive tracking of genomic changes in
1,000 patients with multiple myeloma from bone marrow lion longitudinal biomarker study called the
and peripheral blood collected at baseline, at suspected Parkinson’s Progression Markers Initiative
complete response, and at relapse or progression (PPMI). Launched in 2010, the study includes
CHDI Foundation PLoS Currents Ensures data from CHDI projects are rapidly lodged in 400 untreated patients with Parkinson’s dis-
Huntington’s Disease open repositories under a Creative Commons license.
(http://currents.plos.org/)
ease and 200 controls, each of whom will
undergo a battery of tests and assessments. All
Parent Project PLoS Currents Promotes rapid communication of new research results
Muscular Dystrophy Muscular Dystropy and operational analyses derived from the study or man- data will be freely available in ‘real time’ on a
agement of all types of muscular dystrophy (http://cur- portal (http://www.PPMI-info.org); research-
rents.plos.org/md/) ers need not be funded by the Michael J. Fox
The Addi & Cassi Cyclodexin IND Promotes clinical testing of an approved drug re-purposed funding to access the samples. As novel bio-
Fund (Reno, NV) for a rare condition neglected by commercial sponsors. In
2009, FDA approval given for compassionate-use investi-
markers are discovered, they may be incorpo-
gational new drug (IND) request to use hydroxypropyl beta- rated into the study.
cyclodextrin in Niemann-Pick Type C disease The project was modeled in part after the
Alzheimer’s Disease NeuroImaging (ADNI)
researchers, has also established its own trans- subgroups and then say ‘Hey, we’ve got a clinical study, according to Todd Sherer, CEO of The
lational unit to perform screening, absorp- trial for you’. There will be hypothesis-driven tri- Michael J. Fox Foundation. Begun in 2003, with
tion, distribution, metabolism and excretion als that we can conduct within our clinical net- backing from the NIH’s National Institute on
(ADME) characterization, pharmacokinetics work,” says Quinn Young. Aging and the National Institute of Bioimaging
© 2012 Nature America, Inc. All rights reserved.

and other assays. The foundation approaches Cambridge, Massachusetts–based Millennium and Bioengineering, ANDI was able to engage
companies working on mechanisms or targets Pharmaceuticals and Onyx Pharmaceuticals the pharmaceutical industry as partners in
that could affect the neurological processes have made a multiyear, multimillion-dollar the effort to screen patients with Alzheimer’s
related to Huntington’s disease. It then pays commitment, according to Quinn Young, even disease using various imaging modalities and
for compounds to be tested. “We tell them the though the results will not be patentable. “The to conduct genetic and cognitive assessments.
results and they own the IP. It’s hard for people benefit to the companies is that they will have “They slowly got companies to come on board
to say no to that if they’re confident that you’ll exclusive access to the data for six months. At and get comfortable with this precompetitive
respect confidentiality and IP and that you’ll the end of the sixth month, it goes into the pub- arrangement. That was one of the things in
do high-quality work,” says Robi Blumenstein, lic domain in a portal that we’re building,” says our favor—companies had had exposure to the
president at CHDI. Quinn Young. Participating academic centers Alzheimer’s initiative. The other was our posi-
also have data access during the sixth month. tion as a neutral body. There was a high trust
Tackling clinical trials In addition, companies will have information factor that we had no motive in our coordinat-
Clinical trials present many challenges for drug on their own drug, in real time, as well as their ing role,” says Sherer. Companies contributing
developers, from recruiting patients to validat- competitors’ drugs. “I think that the data are of to PPMI include multinationals, such as Pfizer
ing clinical endpoints and finding suitable bio- interest on a number of levels, from translational of New York, GlaxoSmithKline of London,
markers. Foundations are taking steps to address medicine to clinical development, all the way to Merck of Whitehouse Station, New Jersey,
these problems. marketing,” Quinn Young opines. Genentech/Roche of Basel, GE Healthcare of
In June, the MMRF began its most ambi- The program is expected to cost more than Waukesha, Wisconsin, Covance of Princeton,
npg

tious clinical program, called CoMMpass, $40 million, largely because of the costs of New Jersey, and Abbott of Deerfield, Illinois, as
which will track 1,000 patients newly diag- assays and sequencing. “No one else would well as biotechs like Biogen-Idec of Cambridge,
nosed with multiple myeloma, recording all have the incentive to do something like this, and Massachusetts, and Avid Radiopharmaceuticals
clinical data, including biopsies to be per- no single center would see enough patients to of Philadelphia.
formed at every relapse over five years. “We get this done. From a funding perspective, it’s Along similar lines, CHDI is conducting an
want to understand the profile of patients that tough to get these types of trials funded by the observational study to characterize patients in
are truly at high risk [of relapse]. About 20% government, and a single company wouldn’t be anticipation of clinical trials. It will collect clini-
of patients who initially respond to therapy motivated enough to do it,” says Quinn Young. cal and personal information from patients, as
relapse very quickly and have a much poorer The availability of a patient database is accel- well as polymorphisms and other genetic data.
prognosis. We want to understand what that erating the pace of clinical development. Nearly Twelve thousand subjects are currently enrolled.
population looks like,” says Anne Quinn 40 trials have opened to date, and Quinn Young The data will be available for future clinical tri-
Young, vice president of strategic alliances at says they have started 60% faster than other als. “We’ll have [patients] preregistered and
the Multiple Myeloma Research Foundation. oncology trials. “That’s based on a single-site characterized so that we can enroll quickly for
Researchers can also track responses in those experience at Vanderbilt, but it has involved trials,” says Blumenstein. The foundation has
patients to various treatments. dozens of trials. Before we refined the process also taken steps to expedite informed consent.
The program includes 41 academic and com- there, the rate at which we opened trials was very “It’s less scientific content than organization
munity medical centers, and testing will be per- similar to [that of others in the field],” Quinn content,” he adds.
formed by the Translational Genomics Research Young adds.
Institute (Phoenix, Arizona). Participants will Other foundations are chipping away at Working with regulators
also be screened for various known mutations, different problems with clinical trials. One Moving beyond clinical trials, foundations are
and newly identified mutations will be added as issue for Parkinson’s disease drug develop- also engaging with the FDA to work through
they are discovered. “We may be able to identify ment is the dearth of biomarkers to track regulatory hurdles, such as establishing

nature biotechnology volume 30 number 9 september 2012 819


N E W S fe at u re

appropriate outcomes for clinical trials. The the agony of writing up negative results,” says into making sure that we collect really good data
Michael J. Fox Foundation, for example, will CHDI’s Blumenstein. Such papers would be associated with the samples,” says McBurney. He
begin an effort to validate clinical outcome hard to place in most journals, anyway. envisions the program as a potential model for
measures for Parkinson’s disease cognition, Foundations are in a position to change other diseases. The approach “doesn’t have to be
which in turn could be used for late-stage that. The scientific advisory board of MRF’s limited to MS,” he says.
clinical trials. Such efforts have the potential Accelerated Research Collaboration meets with
to streamline the regulatory process. principal investigators two to three times a year. Brain trusts
Companies can then use the protocols in their “There’s a huge amount of intellectual firepower Foundations are also becoming a repository
clinical trials, rather than developing their own. to discuss [projects]. The more you open up, the of expertise. Many have made substantial
“When they go to the FDA, they can refer to a more benefit it is to [the principal investigator] investments in internal scientific acumen, in
foundation-funded study for why they selected as well as the team. And it prevents them from the form of advisory boards and employees.
a particular outcome measure,” says Sherer. The doing research that’s already been done,” says “Companies come to foundations to get insight
agency is more likely to accept the outcome Johnson. into what models to use, or to get a gut check
measure at face value if a neutral foundation Openness hasn’t always been an easy sell. on the potential of a therapy. We see hundreds
developed it, he adds. Johnson says that it took a couple of years to of applications every year, and we interact with
Other foundation efforts are broader in scope. convince researchers to share data immediately, hundreds of thought leaders external to the
In the field of artificial pancreas systems, the “but now some of our PIs have joint lab meet- foundation, so we have a very good bird’s-eye
Juvenile Diabetes Research Foundation (JDRF), ings,” he says. view of what’s going on in Parkinson’s disease
headquartered in New York, recognized that CHDI has a similar attitude. “We’re in a good research. We see that as a resource beyond
there were no FDA-established guidelines for position to see [negative results] because we funding that research foundations are bringing
approval of such systems. Last June, the foun- fund so much work in Huntington’s disease. to the table,” says Sherer.
© 2012 Nature America, Inc. All rights reserved.

dation consulted with companies and academ- We know if we already funded someone to do Many of those partnerships involve bio-
ics and produced a draft guidance document, an experiment and it didn’t work. We can put tech companies looking to de-risk early-stage
which they then presented to the FDA to kick- those investigators in touch with each other,” programs, hoping for funds to fill gaps in an
start its internal process. The JDRF secured a says Blumenstein. CHDI encourages patenting early-stage development program. But the
commitment from the FDA, which issued its of therapeutic candidates, but animal models, companies are often looking for more than
own draft guidance in December 2011. “We assays and other tools developed must be made just money, according to Insel. “We have a lot
probably accelerated the process by several publicly available to other researchers. “We’ve of insight and knowledge of type 1 diabetes.
years,” believes Richard Insel, chief scientific tried to institutionalize the concept of sharing,” We fund over $100 million a year in research,
officer at the JDRF. says Blumenstein. and about one-third of our funding goes out-
Foundations are in a unique position to gain a The Accelerated Cure Project for multiple side the United States. We have a pretty big
regulatory agency’s confidence, agrees Timothy sclerosis (ACPMS; Waltham, Massachusetts), overview of the type 1 diabetes landscape, as
Coetzee, chief research officer at the National a foundation started in 2001 by tech entre- well as insights into the unmet medical needs
Multiple Sclerosis Society. “We don’t have a preneur Art Mellor, provides researchers of patients.”
commercial stake… the FDA has engaged with with research tools and highly characterized Pharma companies are on board with this.
patient advocate organizations, saying, ‘We want biological samples from about 3,000 subjects Through its relationship with the National
to partner with you to bring people together’. from ten multiple sclerosis clinical centers in Multiple Sclerosis Society’s Fast Forward initia-
They need to hear what the scientific com- the United States. Upon request, aliquots of the tive, EMD-Serono of Rockland, Massachusetts,
munity is thinking, what the stakeholders are samples are sent out to researcher; in return, all has funded a variety of early-stage companies
npg

thinking. Foundations have the ability to touch data must be submitted to the repository data- and academic groups that are working on a
all those worlds in a manner that really isn’t base for general use. Over 60 research studies range of biological mechanisms and targets.
biased,” says Coetzee. have used the resources to date. “We certainly “I don’t know if it would be possible to get
“[Foundations] are really well connected and couldn’t have afforded to fund those studies this sort of diversity in another setting,” says
in tune with the needs of patients and the needs directly, but we greatly reduced the cost and Shearman.
of R&D organizations,” says Mark Shearman, time to do them,” says Robert McBurney, CEO The research is also well targeted, accord-
vice president of global research and early devel- of ACPMS. ing to Shearman. “[The foundation] really
opment at EMD-Serono, a contributor to the The open-source nature of the project may has the patient in mind, so there’s no risk that
National MS Society’s Fast Forward initiative. have turned some companies off, but there is we’re going to be side-tracked by doing aca-
This initiative, started in 2007, provides fund- evidence that it hasn’t been much of a barrier. demic research for research’s sake,” says Steve
ing for early-stage biotechs and academics with Of 60 studies, about 20 were done by companies, Arkinstall, senior vice president and head of
technology ready to license. according to McBurney. “It wasn’t as big a prob- US research and global external innovation at
lem as we might have anticipated.” EMD-Serono.
Promoting sharing Researchers have an exclusivity period to In fact, a big-picture view is critical to the
Fierce competition among academic and indus- secure IP before turning over the data. The success of foundations, Johnson says: “I think
try researchers lead groups to keep a tight hold open source–style system puts a premium on the real secret sauce here is driving compounds
on their data. Furthermore, negative studies data quality, and as a result, research applica- through the entire process, not haphazardly
often don’t see the light of day. “It’s understand- tions are judged not on the hypothesis but the funding in unrelated areas.”
able. It’s hard (for an) investigator to go through quality of the study design. “We put a lot of effort Jim Kling, Bellingham, Washington

820 volume 30 number 9 September 2012 nature biotechnology


building a business

How much risk are you prepared to take?


Mark Van Dyke
Any faculty member wishing to commercialize a discovery or technology should contemplate the pros and cons of
pursuing the entrepreneurial route or licensing to an outside entity.

L ike life and business, academic research


is all about risk. Risk is involved when
deciding which students to take into a facul-
If you have a good idea of where your tech-
nology sits relative to the rest of the world and
know you want to move your idea into a for-
pursued if the university principals have a very
good understanding of the value of the tech-
nology, the potential market and a substantial
ty’s research group, whom to collaborate with, profit entity, you might find that universities amount of data about the asset(s).
© 2012 Nature America, Inc. All rights reserved.

what grants to submit and when and where to often default to the standby commercializa- The problem for you, of course, is that most
publish. It also comes in play with the direc- tion strategy of contacting companies they universities are not experienced in getting
tion of research projects and potential rewards. know (or perhaps you know) and shopping the research to this stage and typically do not have
There is similar risk involved with taking technology as a standard patent license. You the expertise, infrastructure or funding to do
one’s research out of the lab and into a com- should be aware that this strategy rarely works, it. Moreover, if capabilities do exist within your
mercial entity—a complicated process. When as most companies will view your technology university for limited clinical testing, there
Winston-Salem, North Carolina–based as risky academic research, unproved by the may be only enough capacity to advance one
KeraNetics, where I am CSO, was formed in standards they use for internal evaluation of technology at a time, so you might be forced to
2008, we spent nearly a year meeting and talk- new technologies and for in-licensing deci- compete for scarce internal resources in an envi-
ing with the university’s Office of Technology sions (though it helps if you are employed at a ronment where commercialization decision-
Asset Management, the Legal Office, the university with a big name in biotech, such as making experience is typically lacking.
Office of Research, the Dean of the School of Harvard University, Massachusetts Institute of Often the best option is for you to become
Medicine and the Conflict of Interest (COI) Technology or Stanford University). After all, the commercialization champion, in which
Management Director. Each group had dif- for more than a decade many established com- case starting a company moves to the top of
ferent goals and concerns, and each indepen- panies have followed the strategy of acquiring your list of options. But be warned: launching
dently gauged the risk involved in launching late-stage technologies that have already been a startup will require a high level of commit-
KeraNetics. Here I will use that experience substantially ‘de-risked’ (usually in another ment, so consider all the personal and profes-
to provide an idea of how you should evalu- smaller commercial venture) through the clinic, sional cons before making such an important
npg

ate the risk involved in pursuing the different the US Food and Drug Administration (FDA) decision. At the onset, you should ask yourself
paths to commercializing your technology or or even the market. Notwithstanding the uphill what your ultimate goal is: do you want to stay
discovery. challenge, this strategy can only be successfully in academia and have a commercial outlet for

Considering a startup
Technology transfer offices (TTOs) often ask a Box 1 The pointed questions
few basic questions (Box 1) when approached
by faculty with an idea, experimental results or When you approach your technology transfer office with a new technology or asset to
an invention disclosure. If you cannot answer commercialize, the officers will likely have a list of questions to ask you. Be prepared. Ask
these questions, then you are not ready to yourself:
What is the invention?
consider a commercialization plan. But even
What is it used for?
before you go to the TTO, the first questions
Why is it useful?
need to be ‘why do I want to launch a startup?’
What is the current state for development?
and ‘is my technology ready?’.
Are there other applications or uses for the invention?
What competitive technologies exist?
Mark Van Dyke (formerly of Wake Forest
What companies or other users of the invention may be interested in further developing,
University School of Medicine) is currently an
licensing or buying the invention?
Associate Professor at Virginia Polytechnic
What have you published or presented publically that relates to this invention?
Institute and State University in Blacksburg,
What unpublished data do you have?
Virginia, USA, and CSO at KeraNetics LLC,
What research plans do you have for further development, and do you have the funding
Winston Salem, North Carolina, USA.
to pursue these plans?
e-mail: mvandyk5@vt.edu

nature biotechnology volume 30 number 9 september 2012 821


building a business

your research, or do you want to create a new of the marketplace. For example, when start- interest. If there is an entrepreneurial nature to
career for yourself if the company is successful? ing a company, you must consider the end: is your work, it is a good idea to keep an eye out
Depending on your answers, your thought the goal for you and your investors to exit the for these types of students and recruit them
processes, strategies and actions may change. startup by being acquired or through an initial to your lab.
Above all, be honest with yourself about your public offering (IPO)? If so, you must consider The choice of a CEO is one of the most
intentions and with those you involve in the what will motivate a favorable decision among important initial decisions a startup com-
process. the merger and acquisition (M&A) team or pany will make. Many investors will attest to
Here is one big con: done correctly, com- potential investors and build the company in the value of an experienced CEO when deal-
mercializing your research will easily double that direction. ing with the uncertainties of a startup. In
your workload. If you are the inventor, you will It is mostly true that universities are bureau- basic terms, you need to consider whether a
need to help develop the business plan, identify cratic entities with many political agendas in high-powered technical or scientific expert
market opportunities, assess competitive tech- play at any one time. It is a difficult place to is the best choice or someone with the abil-
nologies, educate potential investors, evaluate overlay business principles and derive the ity to manage the business and raise money.
initial hires, translate many lab processes into degree of predictability that a startup needs Obviously someone with both abilities would
an industrial setting, begin the early scientific to be successful. The unpredictability of uni- be ideal, but these individuals are rare and not
activities of the company and, in some cases, versity politics, their inexperience in dealing easy for most startups to find. If the company’s
act as the company’s first CEO or (like me) the with startup businesses and lack of urgency in business plan calls for additional research to
CSO. In fact, creating a startup often comes with decision-making all increase the risk around be funded largely through competitive grants
80-hour work weeks, which drag on for three startups. and resources provided through the university,
or more years, all while you are often trying to I found that when research from my lab was a scientific CEO may make the most sense.
teach classes, run a research group, keep your lab being considered for commercialization, it was However, if the startup plans to raise most of
© 2012 Nature America, Inc. All rights reserved.

funded and remain active in the scientific com- helpful to connect commercialization criteria its funding through equity investment, a CEO
munity. And most of this often happens during to those things most familiar to typical aca- with business expertise and connections to
non-regular business hours, and is accompanied demic faculty—the five ‘P’s: people, patents, capital is more desirable. At KeraNetics, an ini-
by an atmosphere of constant urgency. papers, preliminary data and ‘phunding’. tial emphasis was placed on finding someone
You should also expect that the university will with an ability to raise money but ultimately a
have concerns about your COI and conflict of People. The human factor is one of the most senior scientist was quickly hired who could
commitment. Many universities have strict pol- critical assets in this process. Although the guide the company scientifically as well as lead
icies that help keep the focus of faculty on more value of management expertise cannot be early grant efforts.
scholarly activities. Universities tend to adopt overstated, technical prowess is equally impor- You may be asking yourself, ‘where do I find
one of two approaches: either avoidance of con- tant. This is where full commitment from the such people and how can I, an academic scien-
flict (or even the appearance of conflict) or strict faculty champion and/or inventor—whether tist, find an entrepreneurial network of such
management of conflict. If you are at an institu- that is you or otherwise—is crucial. However, individuals?’. A good place to start is at your
tion that practices the former and you want to your initial hires in all areas are important, and own university. Your TTO is the obvious place
be actively engaged in entrepreneurial endeav- sometimes a pool of talent is available from the as they should be familiar with the regional
ors, you will need to find a new job; conflict of faculty’s own lab. Many academic labs have a business community. If they have extended
interest is inevitable. If your institution practices wealth of smart, eager, talented graduate stu- their professional network through organi-
the latter approach, you can simply be mindful dents and postdocs who can rise to the chal- zations such as the Association of University
of the university policies, address stakeholder lenge of a startup venture. Their knowledge Technology Managers, they may be able to
npg

concerns, be transparent in business dealings can be extensive because they often have con- put you in contact with more experienced
that involve university interests and protect the tributed greatly to the development and test- individuals and perhaps even possible CEO
interests of the trainees entrusted to your care. ing of the technology, and their enthusiasm candidates. Other potential startup manage-
In the case of my involvement in KeraNetics, for building from the ground up is typically ment may be found through your university’s
a progressive COI management plan had been inexhaustible and infectious. Although they School of Business. Some of the faculty may
developed that included a person tasked with are not generally CEO or CSO material, expe- be active or formerly active entrepreneurs, or
oversight who was essentially embedded into rienced postdocs can be critical early hires to will likely have an extensive network of con-
my research program and reported through help transition the technology into the com- tacts. Schools of Business are also a great place
the university’s Office of Research. His role was pany and begin building in-house expertise. to get assistance in business plan development
to attend our group meetings; review our data, However, it is worth noting that good scien- from either faculty or students. During the
especially anything presented in public forums; tists do not always make the best initial hires early stages of KeraNetics, students from the
ensure that disclosure of potential COI was for a startup company. Skills at the bench Wake Forest University Schools of Business
given in public forums; meet with trainees to are not the sole criteria for startup success, did several class-related projects to evaluate
review their research progress and to ensure no so graduate students and postdocs with a the company’s potential products, markets and
barriers to scholarship were present; and inde- predisposition for business, especially those business development strategies.
pendently report to the university and attest to who have taken additional business-related A third option is to speak with your scien-
our following of the management plan and rules coursework, or preferably the occasional MD tific colleagues that have traveled the entre-
of the institution. or PhD-MBA student, have high potential preneurial road. Many scientific societies have
in a startup. If your university has a business groups within them that focus on industrial
Building it school, you will often find these types of grad- science or the commercialization of research.
For most academics it is hard to separate passion uate students either enrolled in dual programs National meetings are great venues to meet and
for their research from the practical realities or taking a few business classes out of pure speak with members of these groups.

822 volume 30 number 9 september 2012 nature biotechnology


building a business

Patents. Patents are the most important pri- would think. Having a robust, well-funded While no funding source should be over-
mary asset of the company, and you will need academic research program (in addition to looked, the startup may find that non-SBIR
proper protection of the technology as well as the funding the startup itself will need) within or non-STTR mechanisms offer longer-term
freedom to operate in an opportune area of the university that can support and expand grant periods and higher maximum funding
the market. In most instances, the process of the capabilities of your startup can enhance caps.
conducting a legal and/or business analysis of your ability to address the unexpected or take Beyond grants and the initial funding, it is
the patent or patents in your research area is an advantage of new opportunities. This helps best to raise additional equity only when there
excellent place to start. reduce risk. In the US, such funding takes is demonstrable progress, such as completing
If the claims of the patent(s) behind your the form of traditional grants from agencies, a pivotal preclinical study. Raising money on
critical technology give exclusive access to such as the Department of Defense (DoD), the bad news can be extremely detrimental to a
a large unmet medical need and freedom National Science Foundation (NSF) and the startup, so if funding is to be raised in several
to operate, then you should feel confident. National Institutes of Health (NIH). tranches, it is best if it happens after success-
Investors will require you to show that the On the company funding side, it will be ful binary events. But in general, always raise
technology is indeed ‘first in class’, so you will key for you to structure the startup so that it more money than you think you will need.
need to understand the current state of the is competitive for nondilutive grant funding. KeraNetics has gone through three funding
art and incorporate competitive technologies The role of grant reviewer should already be rounds and fortunately has always had good
into research plans. KeraNetics had a substan- familiar to you, so it can be relatively easy to news on which to raise money. A fairly steady
tial portfolio of 20 initial patent applications evaluate a startup’s ability to hit the goals of a stream of incoming grant funding has helped
when it started, and an extensive freedom to grant. The problem exists in the nature of a in this regard, as well as favorable data from
operate analysis was conducted soon after the startup. Such early-stage ventures often have pivotal studies.
company was formed. no track record of commercialization, no fed- You must have critical mass in all these areas
© 2012 Nature America, Inc. All rights reserved.

eral funding, no assets, no technical employees before launching your startup. But what con-
Papers. Peer-reviewed reports, especially of their own, and no research space indepen- stitutes critical mass? Is one patent, one paper
those in high-impact journals, legitimize your dent of the founder’s academic lab, so what or one grant enough? This will differ depend-
research and give credibility to the technology would make them competitive for funding? ing on the technology. Again, it is best to begin
behind your startup to investors. They also help If the research is the only novelty, why would with the end in mind and look to other suc-
support the patents and are often the basis for the research fit better as a company-funded cessful academic startups for guidance. For
regulatory activities, such as establishing clas- grant rather than a traditional NIH grant to example, if the university has a track record
sification of the technology via its intended the principal investigator? of building successful startups based on single
mode of action. Key papers, particularly those The answer is that your company must offer patents or narrowly focused technologies,
using large-animal models, can serve as safety a high potential for commercialization of the chances are they can help find management
and efficacy data for obtaining FDA approval technology. This can happen only with the and early investors.
for human clinical trials. A solid understand- infrastructure that comes with a substantive
ing of the basic science related to the technol- startup, such as working capital, independent Focus and expecting change
ogy is vital to these and other processes, such lab space, equipment and a few employees on These five Ps are just guidelines—the one truth
as product and/or process optimization and the payroll. Fortunately, many universities now in biotech startups is that things will change.
developing a pipeline of future products. At offer incubator space that can be rented at low Having a well thought-out, carefully consid-
KeraNetics, for example, large-animal data cost and give access to core facilities that would ered business plan is important for maintain-
were in place for three different products otherwise be beyond the financial means of a ing the focus and direction of the company.
npg

when the company was formed, as well as half typical startup. This is also another reason why Academics are naturally curious and are often
a dozen key papers either published or under some initial capital is essential, not only for tempted to follow new discoveries. This can be
review. paying the rent but also for hiring employees a mistake in an academic startup; it is critical
and perhaps acquiring a few pieces of essential that a company stay on track and be aware of
Preliminary data. The more data behind your research or production equipment. necessary changes to the business plan or new
technology or asset, the better. It is important When most people think of startup grant opportunities. Again, experienced management
because it helps establish a technology platform funding, they naturally gravitate to the ubiq- can anticipate the need to adjust the business
that will be the future of the startup company. uitous Small Business Innovation Research plan before change is forced upon the startup.
More importantly, preliminary data fuel grant (SBIR) and Small Business Technology Funding helps mitigate risk, but the distrac-
submissions, which can provide critical, non- Transfer (STTR) programs at many govern- tion of raising money, whether from investors
dilutive funding to a startup while the company ment agencies, but there are many other or grants, can often divert the company from
is operating under negative cash flow. funding agencies that support effective execution of its business plan. The key is to
When KeraNetics was founded, three years industry-academic partnerships. Many DoD raise enough money initially, and if additional
of unpublished data were available, mostly Program Announcements, for example, do funding is required, raise it as infrequently as
in cell culture and rodent models, and this not exclude for-profit entities from eligibility. possible to avoid dilution or the dreaded ‘down
was used for several early grant applications. To the extent that the startup can demonstrate round’ in which funding is more costly to the
Although slow to start, the company was an effective, functional partnership, an ability current investors, including you as a founder.
awarded more than $8 million in grants in its to meet the needs of the DoD better than an Critical mass in the five Ps will help immensely
first three years. academic group alone and preferably a track when raising initial capital and will make that
record of having worked through the issues money some of the cheapest to obtain.
Phunding. The importance of funding is obvi- associated with such arrangements with the However, contingencies are difficult to
ous, but perhaps for more reasons than some university, the company will be successful. envision and execute if the technology does

nature biotechnology volume 30 number 9 september 2012 823


building a business

not lend itself to change or new opportuni- one, and academic faculty need to carefully offers wonderful educational opportunities to
ties. Many biotech companies have survived consider their own professional and personal graduate students and postdocs, it can greatly
the long and risky road to regulatory approval goals before venturing into this often chaotic expand the translational potential of one’s
by executing contingencies in their business space. Many faculty, especially those with research, and ultimately it can lead to commer-
plan. These can range from orphan status for a history of strong institutional support for cialization of technologies that are of benefit
a narrow application of a drug to development their research, may be uncomfortable with to patients—a worthwhile goal for the ultimate
of diagnostic tools based on the company’s the notion that the startup company may dispensation of your life’s work.
research. Startup companies will often consider only have months’, weeks’ or even days’ worth
COMPETING FINANCIAL INTERESTS
non-biomedical uses for their technology so of operating capital in the bank. If a startup
The author declares competing financial interests:
that the uncertainty of regulatory approval can seems like an unattractive option at this point, details are available at http://www.nature.com/
be avoided for these products. These strategies you may want to consider a more arms-length doifinder/10.1038/nbt.2330.
can range from research-use-only products to relationship with the startup, a simple licens-
veterinary and even non-medical applications. ing deal with an existing company or handing For more content on bioentrepreneurism,
Finally, it is always best to maintain a resil- the technology off to someone with a higher visit our Trade Secrets blog.
ient demeanor and instill this in the company’s risk tolerance, such as a close colleague or http://blogs.nature.com/trade_secrets/
employees. The startup game is not for every- trusted postdoc. However, a startup company
© 2012 Nature America, Inc. All rights reserved.
npg

824 volume 30 number 9 september 2012 nature biotechnology


correspondence

To be or not to be transgenic
To the Editor: can be expected to contain about 1.8 million soybean coding regions are SNPs, or one SNP
It is sad and ironic that even though much novel SNPs. per 2,000 bp in coding regions; the frequency
progress has been made in deciphering Not only are SNPs commonplace but in noncoding regions is 0.5%, or 1 per
the genetic content of food plants and techniques that create SNPs have a long history 191 bp6. This is similar to the level of SNPs
modifying their genomes for the betterment of of safe use by breeders. Before the advent of the in the human genome. Such crops as maize
humankind, many of the principles of modern techniques described by Waltz, the only tool are much more diverse; in this cereal, SNPs
plant genetics, firmly established decades available to breeders to alter DNA sequences account for as many as 1.3% of base pairs7.
ago, are now so easily was the use of radiation Tenaillon et al.8 have estimated that any two
forgotten or ignored. Such and chemical mutagens. alleles of a maize gene for a 300–400-amino-
is the case with many of the The Food and Agricultural acid protein would differ by 3.5 amino
© 2012 Nature America, Inc. All rights reserved.

alarmist arguments raised Organization (Rome) and acids due to SNP accumulation. Within a
in the News Feature by International Atomic Energy diverse population, there are likely to be
Emily Waltz1 in the March Agency maintain a database 15–20-amino-acid differences between
issue entitled “Tiptoeing (http://www-infocris.iaea. proteins from two alleles of a single maize
around transgenics.” Waltz org/MVD/) that currently gene. It is therefore not surprising that
focuses on the controversy lists 2,543 known plant attempts at protein engineering have not
surrounding the regulation varieties developed through converted enzymes into toxins, as toxic
of modern (and, in mutagenesis, including proteins have substantial structural differences
fact, not-so-modern) many common or widely from other proteins and need to perform
biotechnological techniques, grown and consumed crop specific physiological roles to act as toxins9.
such as those that alter single plants, of which 14% were SNPs are thus really minor variations
base pairs by replacing one derived with chemical compared with the larger-scale changes
nucleotide with another (that is, create single- mutatagens3. Chemical mutagens are still that have accumulated in crops during
nucleotide polymorphisms or SNPs). We feel it used to create the same kind of SNPs4 that are domestication and breeding. A case in point
is important to stress that such genetic changes cited as a cause for concern in the Waltz News are the elongated tomatoes on today’s market
must be viewed in a historical and biological Feature. Although the genetic basis and extent (which could fall under the category of
context to understand why calls for new layers of SNPs for the mutant phenotype are usually cisgenics, another technology highlighted by
of regulation over technologies that introduce unknown when mutagenesis is employed, the Waltz). However, in the tomato’s case, its DNA
npg

SNPs and other changes are unwarranted. resulting crops are considered as safe as others got copied and moved to another location
The most relevant counterargument and are not subject to premarket regulatory in the genome through naturally occurring
to the need for regulation is the fact that review. mechanisms10, most probably after the
mutations normally happen. Mutations Breeders depend on sheer luck to find an tomato’s arrival in Spain11.
occur spontaneously in nature, and their alteration in the gene encoding their trait The Waltz article also discusses the
rate can be increased by the use of mutagens. of interest when they employ mutagenesis, new technology developed by Pioneer
On the whole, mutation is a good thing, for and they must accept random alterations HiBred (Des Moines, IA, USA) in which
without mutation, we would still be biofilm elsewhere in the genome whenever these do transgenic plants produce nontransgenic,
on the bottom of the ocean. Although typical not affect crop growth, performance and yield male sterile plants that are used in hybrid
mutation rates are quite low when calculated to an unacceptable point. Today, by using in production. The argument is made that
on a gene or base-pair basis, they are high vitro techniques, breeders have the ability to although these plants are not transgenic
enough that new mutations are the rule rather target the gene of interest, and not introduce per se, they should be viewed as such. But if
than the exception. For example in Arabidopsis unintended and unwanted mutations this same ‘sins of the fathers’ argument were
thaliana, the mutation rate per base pair per elsewhere in the genome. If anything, applied elsewhere in agriculture, humans
generation is estimated to be 7 per billion base therefore, modern techniques should decrease should not consume modern-day tomatoes
pairs2. Given that there are 125,000,000 base concerns for safety, not raise them. because their parents contained a toxin.
pairs in the A. thaliana genome, 1.75 new SNP This leads us to the question of whether This nonsensical argument is not applied to
mutations are expected per generation per SNPs alter protein safety. SNPs accumulate conventional plant varieties and, therefore,
diploid plant. Although SNPs appear to occur in plants and animals. One simple means of there is no reason why it should be applied to
at about the same rate in all plants, crop plants quantifying SNP formation is by comparing transgenics.
have larger genomes, and thus more SNPs. Just SNP differences between pairs of genotypes Despite millennia of plant genetic
one average hectare of 240,000 soybean plants or varieties5. For example, 0.05% of bases in modification, thus far we have not found

nature biotechnology volume 30 number 9 september 2012 825


correspondence

a single verified report whereby breeding 3University of Florida, Gainesville, Florida, USA. Although relatively new compared with
or radiation and/or chemical mutagenesis e-mail: wparrott@uga.edu its genomic and proteomic predecessors,
resulted in a toxin, allergen or other hazard research in the field of metabolomics has
1. Waltz, E. Nat. Biotechnol. 30, 215–217 (2012).
that was not known to exist before. These 2. Ossowski, S. et al. Science 327, 92–94 (2010). already led to the discovery of biomarkers for
facts support the conclusion that DNA 3. Ahloowalia, B.S., Maluszynski, M. & Nichterlein, K. disease, fundamental insights into cellular
insertions and other types of mutations Euphytica 135, 187–204 (2004). biochemistry and clues related to disease
4. Balyan, H.S., Sreenivasulu, N., Riera-Lizarazu, O.,
do not pose unreasonable risks to the Azhaguvel, R. & Kianian, S.F. in Advances in Agronomy, pathogenesis1,2.
environment or to human and animal Vol. 98 (ed. Sparks, D.L.) 357–414 (2008). The success of metabolomics over the
5. Tajima, F. Genetics 105, 437–460 (1983).
health, regardless of how they came about. 6. Huan, N.V., Sugimoto, H. & Harada, K. Breed. Sci. 55,
past decade has relied largely on advances in
COMPETING FINANCIAL INTERESTS 441–446 (2005). mass spectrometry instrumentation, which
The authors declare competing financial interests: 7. Ross-Ibarra, J., Morrell, P.L. & Gaut, B.S. Proc. Natl. make it possible to detect thousands of
Acad. Sci. USA 104, 8641–8648 (2007).
details are available at http://www.nature.com/ 8. Tenaillon, M.I. et al. Proc. Natl. Acad. Sci. USA 98, metabolites simultaneously from a biological
doifinder/10.1038/nbt.2347. 9161–9166 (2001). sample. Coupled with developments in
9. Pariza, M.W. & Cook, M. Regul. Toxicol. Pharmacol. 56, bioinformatic tools such as XCMS Online
Wayne A Parrott1, Joseph M Jez2 & 332–342 (2010).
L Curtis Hannah3 10. Xiao, H., Jiang, N., Schaffner, E., Stockinger, E.J. & (https://xcmsonline.scripps.edu/)3, it
1University of Georgia, Athens, Georgia, USA. van der Knaap, E. Science 319, 1527–1530 (2008). has now become relatively routine to
11. Rodriguez, G.R. et al. Plant Physiol. 156, 275–285
2Washington University, St. Louis, Missouri, USA.
(2011).
comprehensively compare the intensities of
thousands of metabolite peaks in one sample
group to those in another in an untargeted
manner. This approach, called untargeted
Broad consent in biobanking metabolomics, has the potential to implicate
© 2012 Nature America, Inc. All rights reserved.

unexpected pathways with a unique


phenotype or disease process.
To the Editor: consent can be burdensome and impede Despite the attractiveness of having a
The Feature in the February issue by Scott research. comprehensive and unbiased approach
et al.1 on the policy challenges of biobanking This experience suggests to us that broad for profiling metabolites that is analogous
characterizes broad specimen donor consent is ethically responsible, provided to those used in the other ‘omic’ sciences,
informed consent as “ethically contentious.” there is comprehensive oversight and a robust an overwhelming proportion of the
A survey of public attitudes is cited. This informed consent process. With the continued metabolomic community exclusively uses a
same survey found that a significant support of donors, we look forward to targeted platform in which only a specified
percentage of individuals are prepared “to applying this model in biobanking efforts. list of metabolites is measured. The benefit
consent broadly to future research use and to COMPETING FINANCIAL INTERESTS of such a targeted platform is speed. Unlike
forego additional choices as a result”2. The authors declare no competing financial interests. the untargeted platform, after the targeted
With our perspectives in patient advocacy mass spectrometry methods are established,
Chris Hempel1, Geoffrey Lomax2 &
or at research centers aimed at bringing new minimal effort and resources are required
Steve Peckman3
regenerative therapies to patients, we have to profile these specific metabolites over a
1Addi & Cassi Fund, Reno, Nevada, USA.
consistently emphasized the value of research large number of samples. In contrast, the
2California Institute for Regenerative Medicine,
donors’ perspectives. In the context of major bottleneck of untargeted metabolomics
San Francisco, California, USA. 3Eli and Edythe
protocols for creating immortalized cell lines has been the challenge of determining
npg

Broad Center of Regenerative Medicine, University


for banking and distribution, we have also the identities of the peaks found to be
of California, Los Angeles, California, USA.
witnessed support for broad consent. Indeed, e-mail: glomax@cirm.ca.gov dysregulated in the untargeted profiling data.
enthusiasm is even more pronounced among Traditionally, the untargeted
1. Scott, C.T. et al. Nat. Biotechnol. 30, 141–147
those touched by disease, and patient donors (2012).
metabolomic platform involves multiple
actually express concern that study-specific 2. Simon, C.M. et al. Genet. Med. 13, 821–831 (2011). steps (Fig. 1). The first step is acquiring
global mass spectrometry data for each of
the samples. Next, these data are analyzed
using bioinformatic software that performs
quantitative analyses to find peaks that
An accelerated workflow for are significantly different between sample
groups. The investigator then typically
untargeted metabolomics using the searches the mass-to-charge (m/z) ratios of
the peaks of interest manually in metabolite

METLIN database databases. Searches that return hits within


the mass accuracy of the instrument are
considered to be putative identifications.
To the Editor: from genes, transcripts and proteins. The To confirm the identifications, tandem
Metabolites, which are typically recognized as high correlation between metabolites and mass spectrometry (MS/MS) data from the
small molecules that are involved in cellular phenotype has created a surge of interest in research sample are then compared to the
reactions, provide a functional signature the field that is reflected in the number of MS/MS data of a commercial standard. To
of phenotype that is complementary to the metabolomic publications growing from just obtain the MS/MS data, a targeted MS/MS
upstream biochemical information obtained a few articles in 1999 to over 5,000 in 2011. analysis is typically performed on one of

826 volume 30 number 9 September 2012 nature biotechnology


correspondence

Autonomous metabolomic workflow database that has advanced functionality


Step 2 to automate metabolite identification and
Step 1
reduce the labor-intensive bottleneck that has
Data acquistion Quantitative comparative analysis
MS and MS/MS and metabolite identification traditionally been associated with untargeted
metabolite profiling. Instead of manually
XCMS Online comparing the MS/MS data from research
samples to the MS/MS data of commercial
Traditional metabolomic workflow standards, the new version of METLIN
allows metabolomic investigators to upload
Step 1 Step 2 Step 3 Step 4 Step 5 Step 6
their MS/MS data to the METLIN database
MS Comparative Manual MS Manual Targeted MS/MS Manual Comparison
acquisition analysis database aquisition MS/MS spectra
ID so the comparisons can be performed in an
search automated way. By using automated MS/
MS data matching, metabolite identities
can be confirmed much more efficiently
Figure 1 Schematic representation of the traditional metabolomic workflow involving six steps
and the new METLIN-based workflow with only two steps. In the two-step autonomous workflow,
and quickly compared with the traditional
mass spectrometry (MS) and MS/MS data are acquired simultaneously during profiling and untargeted metabolomic workflow. The
searched in the METLIN database for automated identification, thereby reducing the time of the quality of the match between the MS/MS data
workflow from days or weeks to minutes or hours. from the research sample and the MS/MS
data from the METLIN library is measured
the research samples for which the peak of physiologically relevant metabolites in by a newly introduced METLIN scoring
was determined to be upregulated. The METLIN and the other three largest databases system, which is based on a modified version
© 2012 Nature America, Inc. All rights reserved.

fragmentation pattern of the MS/MS data available (HMDB, MassBank and LipidMaps), of the established X-Rank scoring system9. To
is then manually compared with that of the metabolites were isolated from Escherichia evaluate the correlation of METLIN MS/MS
MS/MS data from a commercial standard coli and standard human serum using defined data to MS/MS data acquired using different
(however, not all commercial standards, for protocols8. Samples were analyzed in both instrument platforms, we performed a
example, stereoisomers, can be resolved by the positive and negative modes with an ESI- comparative experiment was performed
MS/MS data alone). QTOF mass spectrometer (Supplementary using 23 metabolite standards. The
To facilitate identification of metabolites Methods). Each peak detected (excluding compounds were measured on five different
in the untargeted workflow, we launched a isotopes) was searched in each of the four instruments, and the resulting spectra were
freely accessible metabolite database called databases. Figure 2 shows the number of hits matched against the METLIN database.
METLIN4 in 2004 (http://metlin.scripps.edu/) for each database and also the subset of the Based on the modified X-Rank scoring
that incorporates MS/MS data from model hits for which MS/MS data were available to system, the correct result was returned as the
compounds. Recently, other metabolite confirm the metabolite identification. first hit for 90 out of the 101 spectra (89.1%;
databases such as the Human Metabolome In addition to its increased size, here Supplementary Table 1 and Supplementary
Database (HMDB)5, MassBank6 and we describe a new version of the METLIN Figs 1–27).
LipidMaps7 have also begun incorporating
MS/MS data for standard compounds. These
Database hits for E. coli and human serum
repositories allow investigators to compare
MS/MS data from their research samples Total
npg

6,000

METLIN
to MS/MS data from model compounds Total
METLIN Human
catalogued in the database and thereby
5,000

E. coli serum
improve the speed, efficiency and cost Hits with MS/MS Hits with MS/MS
effectiveness of untargeted studies. Hits without MS/MS Hits without MS/MS
4,000
Number of hits

Over the past 7 years, our objective has


been to generate a sufficiently large MS/MS LipidMaps LipidMaps
3,000

data library that can be used in an automated HMDB HMDB


manner to revise the traditional untargeted
2,000

metabolomic workflow (Fig. 1). Since we MassBank


originally reported the establishment of MassBank
1,000

METLIN in 2005, we have increased the


number of MS/MS spectra that are included
0

in the database by a factor of 150. As of


April 2012, METLIN contains MS/MS data Figure 2 Estimate of the physiological relevance of metabolite coverage in metabolomic databases.
on >10,000 distinct metabolites at four Metabolites from human serum and E. coli were isolated and analyzed in both the positive and negative
different collision energies. These data were modes by ESI-QTOF mass spectrometry, and the mass of each metabolite was searched with a tolerance
collected using an electrospray ionization/ of 5 parts per million in the METLIN, LipidMaps, HMDB and MassBank databases. LipidMaps contains
data primarily on lipids, which is only a subset of the metabolome, but was included in the comparison
quadrupole time-of-flight (ESI-QTOF)
for the sake of completeness. For human serum, 12,170 features were detected and searched, and for
mass spectrometer in both the positive and E. coli, 11,641 features were detected and searched. The number of hits on the basis of accurate mass
negative detection modes, representing a are shown in light blue and light red for E. coli and human serum, respectively. The subset of those
total number of >48,000 high-resolution hits that also contained MS/MS data are shown in dark blue and dark red for E. coli and human serum,
spectra. To estimate the current coverage respectively.

nature biotechnology volume 30 number 9 september 2012 827


correspondence

Some classes of metabolites produce performing high-throughput, untargeted Ralf Tautenhahn1,2, Kevin Cho1,2,
characteristic fragments or neutral losses metabolomics using this type of accelerated Winnie Uritboonthai1,2, Zhengjiang Zhu1,2,
in their MS/MS spectra that can be used as workflow. Moreover, we have shown that the Gary J Patti3–5 & Gary Siuzdak1,2
signatures for unique chemical functional coverage of the METLIN database enables 1Department of Chemistry, Center for
groups. For example, the MS/MS spectra the characterization and identification
Metabolomics, The Scripps Research Institute,
of phosphatidylcholines are characterized of thousands of naturally occurring La Jolla, California, USA. 2Department of
by a fragment at m/z 184.07. For instances metabolites in biological samples. Thus, the Molecular Biology, The Scripps Research Institute,
in which the MS/MS data uploaded by a new METLIN database has the potential La Jolla, California, USA. 3Department of
user do not match any compound in the to expedite the workflow for untargeted Chemistry, Washington University School of
database, the new version of the METLIN metabolomics as more investigators obtain Medicine, St. Louis, Missouri, USA. 4Department
database will search the MS/MS data for mass spectrometry instrumentation that of Genetics, Washington University School of
characteristic fragments that can be used can produce high-quality MS/MS data with Medicine, St. Louis, Missouri, USA. 5Department
for molecular classification. The search can increasing speed and sensitivity. of Medicine, Washington University School of
also be performed manually by accessing Medicine, St. Louis, Missouri, USA.
Note: Supplementary information is available at http:// e-mail: gjpattij@wustl.edu or siuzdak@scripps.edu
the ‘fragment search’ or ‘neutral loss
www.nature.com/doifinder/10.1038/nbt.2348.
search’ options. These tools provide a new
1. Yanes, O. et al. Nat. Chem. Biol. 6, 411–417 (2010).
mechanism by which unknown metabolites ACKNOWLEDGMENTS 2. Patti, G.J. et al. Nat. Chem. Biol. 8, 232-234 (2012).
can be chemically classified, and they take This work was supported by the California Institute of 3. Smith, C.A., Want, E.J., O’Maille, G., Abagyan, R. &
Regenerative Medicine (TR1-01219), the US National Siuzdak, G. Anal. Chem. 78, 779–787 (2006).
advantage of the large amount of MS/MS data
Institutes of Health (R24 EY017540, P30 MH062261, 4. Smith, C.A. et al. Ther. Drug Monit. 27, 747–751
in the library. RC1 HL101034 and P01 DA026146) and the US (2005).
To highlight the new database National Institutes of Health National Institute on 5. Wishart, D.S. et al. Nucleic Acids Res. 37, D603–D610
© 2012 Nature America, Inc. All rights reserved.

(2009).
functionalities, we performed MS/MS on Aging L30 AG0 038036 (G.J.P.). Financial support
6. Horai, H. et al. J. Mass Spectrom. 45, 703–714 (2010).
select peaks from the metabolite extracts of was also received from the US Department of Energy
7. Sud, M. et al. Nucleic Acids Res. 35, D527–D532
(grants FG02-07ER64325 and DE-AC0205CH11231).
E. coli and human serum. These data were (2007).
8. Yanes, O., Tautenhahn, R., Patti, G.J. & Siuzdak, G.
uploaded to the METLIN database, and COMPETING FINANCIAL INTERESTS Anal. Chem. 83, 2152–2161 (2011).
fragment matching was performed using The authors declare no competing financial interests. 9. Mylonas, R. et al. Anal. Chem. 81, 7604–7610 (2009).
the automated feature described above.
Representative examples of metabolites
identified on the basis of the mass
spectrometry and MS/MS data using this
method are shown in Supplementary Figures
Successful suppression of a field
28–32. The compounds identified ranged
from lipids to smaller, polar metabolites. mosquito population by sustained
Additionally, representative examples of
unknown compounds that were classified by release of engineered male
characteristic fragments are also shown.
With the combination of the METLIN mosquitoes
functionalities described here and the
increasing speed of QTOF instrumentation
npg

for performing MS/MS, there is the potential To the Editor: successfully control several agricultural
to reduce the untargeted metabolomic Our paper published last year described the pest insects5. Large numbers of sterile
workflow to just two steps (Fig. 1). Using results of preliminary release experiments insects are released to mate with their wild
high–scan-speed QTOF instruments, showing that engineered counterparts and thereby
mass spectrometry and MS/MS data can sterile male mosquitoes reduce their reproductive
be acquired simultaneously in a single could mate with females potential. However,
run. Quantitative information can then in a wild population in despite its attractive
be extracted from the data using the the Cayman Islands1. This features, this technique
bioinformatic software XCMS Online, and trial was supported by is not in operational use
metabolites can be identified simultaneously simple simulation models against mosquitoes, in
by matching the MS/MS data with MS/MS indicating that sustained part because of damaging
data in the METLIN database in an release of sufficient numbers effects of sterilizing doses
automated fashion, an approach that is self- of such males should of radiation on the released
directed or autonomous in nature. substantially suppress a mosquitoes6–8. Following a
With this truncated workflow, the time target population within similar principle, we have
needed to perform untargeted profiling and a few weeks or months2–4. proposed that engineered
the subsequent metabolite identification In the following letter, we males carrying a dominant
may be reduced to minutes or hours as describe a field release lethal transgene could
compared to the days or weeks needed with experiment testing this proposition. be released to mate with wild females; the
the traditional workflow. The results shown The sterile insect technique is an resulting progeny would die as a result of the
here from automated MS/MS matching environmentally friendly, species-specific lethal effect of the transgene. We named this
highlight the applicability of the method for method of pest control that is used to system RIDL (release of insects carrying a

828 volume 30 number 9 September 2012 nature biotechnology


correspondence

dominant lethal gene)9,10. The Aedes aegypti a


RIDL strain, OX513A, has a single transgenic
sequence encoding a red fluorescent marker
and tetracycline-repressible late-acting
dominant lethality3.
Approximately 3.3 million engineered
OX513A males (male mosquitoes do not
bite) were released in a 23-week period George ●
East
in 2010 in a field site in Grand Cayman, a ● Town
End
British Overseas Territory in the Caribbean. ● Bodden 5 km
The preliminary release experiments Town

showed that OX513A males could mate b


with wild females1 and, together with
simulation models3, indicated a minimum
release rate of 3,150 males per hectare (ha) Area A
per week to induce population collapse in
the absence of immigration. We aimed for Area B
twin targets of the release of >4,000 males
per ha per week and a 10:1 ratio of sterile to
wild males in the field. Area C
Releasing large numbers of engineered
© 2012 Nature America, Inc. All rights reserved.

sterile males into a wild mosquito population 500 m


should have several measurable effects on
the wild population, including, in temporal c Release rate/ha/week

order, an increase of the male-to-female ratio, 16,000


Period 1 Period 2 Period 3
females mating with engineered males and 14,000
12,000 Pupal release
suppression of the target population. The 10,000
8,000 Adult release
mosquito population—OX513A and wild— 6,000
4,000
was monitored using adult traps and ovitraps, 2,000
which mimic natural oviposition sites 0
ay ay n n l l l g g p p ct
(Supplementary Notes). Adult traps were M M Ju Ju Ju Ju Ju Au Au Se Se O
5 19 3 17 31 9
also used to monitor the numbers of OX513A 8 22 14 28 11 25
males in the field, as an input of males should
Figure 1 Field site and mosquito releases in 2010. (a) Map of Grand Cayman showing the locations of
change the sex ratio of the population. Larvae
East End and Bodden Town, which were the treatment and external control sites, respectively, and the
hatched from field-collected eggs from capital, George Town. (b) Aerial photograph of East End showing areas A, B and C. (c) Weekly numbers
ovitraps were screened for fluorescence to of adult males released per hectare from direct adult release (solid bars) and emerging from pupal
determine paternity (fluorescent larvae had deployment (shaded bars). Releases occurred 3 times per week; release numbers shown for each week
an OX513A father, and nonfluorescent larvae are the sum of these three releases. During period 1, all treatment areas received treatment, during
had a wild-type father). These monitoring period 2, areas A and B were treated, and during period 3, only area A was treated.
npg

methods were used to evaluate the field trial


endpoints: sterile-to-wild male ratio ≥10:1, males per ha per week for the first 3 weeks in areas C (likelihood ratio (LR) P = 0.63)
fluorescence ratio ≥50% and statistically and later reduced this to ~7,700 (95% CI and B (LR P = 0.13) but were significantly
significant population suppression relative to 6,900–8,500) adult males per ha per week, lower (~40%) than those in an external
control site(s). which was supplemented with ~4,900 (95% untreated control site (LR P = 0.026). After
Initial releases of OX513A males across CI 3,800–6,000) adults eclosing from ~5,600 that time point, the index in area A was
a 55-ha area (areas A, B and C, period 1; (95% CI 4,500–6,800) pupae deployed in highly significantly lower than those in all
Fig. 1) and then across a reduced 32-ha area field. This gave a mean OX513A-to-wild other areas (P < 0.0001 for each pairwise
(areas A and B, period 2) were restricted by male ratio in the release area of 25.2:1 (95% comparison).
production difficulties to an average of 1,400 bootstrap CI 17.8:1 to 34.9:1, n = 3,155) Over the last 7 weeks of the release period,
(95% CI 990–1,800) males per ha per week in the first 4 weeks of period 3. We also the mean ovitrap index in the untreated
and 3,900 (95% CI 2,600–5,300) males per found an increase in the proportion of the areas was 49% (95% CI 43–55%; Fig. 2c). In
ha per week, respectively (Fig. 1c). These field-collected eggs carrying the fluorescent contrast, the mean ovitrap index in area A
release rates also did not achieve the target marker, with a peak of 88% (Fig. 2b), was 10% (95% CI 7–14%), which is an 80%
OX513A-to-wild male ratios, reaching a 1.9:1 implying that the majority of wild females reduction relative to the untreated areas,
ratio (95% bootstrap CI 1.2:1 to 2.8:1, n = were mating with OX513A males. indicating strong population suppression
967) in period 1 and a 4.8:1 OX513A-to-wild We used the ovitrap index—the in the treated area during this period. The
ratio (95% bootstrap CI 2.6:1 to 7.9:1, proportion of ovitraps in each area with one degree of suppression that is possible in
n = 1,994) in period 2 (Fig. 2a). We therefore or more eggs after 1 week—as our primary such a trial is limited by immigration of wild
further reduced the release area to 16 ha measure of population density (Fig. 2c). females from surrounding areas, such as area
(area A, period 3). In this final period, we Until early in period 3, the weekly ovitrap B, as well as, potentially, from eggs laid at an
released ~14,000 (95% CI 13,700–14,500) indices for area A were very similar to those earlier period.

nature biotechnology volume 30 number 9 september 2012 829


correspondence

a Period 1 2 3 e Figure 2 Effect of periodic release of OX513A male mosquitoes on a wild


population. Release of engineered males should lead to an increase in the
Estimated OX513A:WT adult male ratio OX513A fluorescence profile
male-to-female ratio of the field population, deposition of hybrid eggs after
100
mating of released males with wild females and, finally, suppression of the
target population if sufficient sterility or mortality is induced. (a) The ratio of
10 OX513A to wild-type (WT) adult males estimated from the sex ratio of adults
caught in BG-Sentinel traps. Treatment periods 1–3 are indicated (see also
1 Fig. 1). Increasing the input rate (males per ha per week) and, later, declining
b Percentage of fluorescent larvae wild population numbers resulted in an increase in this ratio over time.
100
(b) Percentage of larvae recovered from ovitraps in treated areas with the
80
RIDL transgene as detected by fluorescence. This percentage is plotted
60
only into September 2010, as the number of eggs collected after that time
40
became too low to act as a reliable measure because of suppression of the
20
target population. (c) Ovitrap index in East End (areas A–C) and the external
0
c Percentage ovitrap index
untreated control site, Bodden Town. Area C received low-level treatment
100 Area A early in the trial (period 1) for a duration and level that were not considered
Area B effective. Because of its close proximity and ecology to area A, area C
80
Area C provided a largely ‘untreated’ internal control that was highly comparable to
60 Bodden Town
area A. From the beginning of August, the ovitrap index in area A declined
40
relative to all control areas. All populations declined later, which is typical of
20 seasonal variation driven by rainfall. However, even after some rainfall from
0 March 2011, by June, the population in area A was still below that in control
d Weekly rainfall in East End (mm) area C. (d) Weekly rainfall in East End in 2010 (mm). (e) Red fluorescence
200
150
of OX513A larvae. Four larvae are shown, of which the upper two are wild-
type and the lower two are OX513A. The larvae were imaged under normal
© 2012 Nature America, Inc. All rights reserved.

100
50 illumination (left) and epifluorescence (right). Red fluorescence of the
0
OX513A larvae is caused by expression of DsRed2 (ref. 3).
Ap
r ay n
Ju
l g p ct r
ov Dec Jan Feb Ma Ap
r ay un
M Ju
1 Au Se O N 1 M 1J
1 1 1 1 1 1 1 1 1 1 1 1

The time delay inherent in sterile-male– ACKNOWLEDGMENTS 1Mosquito Research and Control Unit, Grand

release methods (the population reduction We thank L. New and J. Renmant for assistance in egg Cayman, Cayman Islands. 2Vector Group,
production, Z. Ebanks and E. Ebanks for technical Liverpool School of Tropical Medicine, Liverpool,
occurs in the progeny of released insects),
help in Grand Cayman, T. Matthews and E. Moxon UK. 3Oxitec Limited, Oxford, UK. 4Department
combined with female monogamy, means for administrative support and all staff and students
that we were releasing a sufficient number of Zoology, University of Oxford, Oxford, UK.
at the Mosquito Research Control Unit and Oxitec 5Medical Entomology Unit, Infectious Diseases
of males of sufficient quality to achieve for their help and support during this study. We
thank G. Labbé and P. Gray for comments on the Research Centre, Institute for Medical Research,
suppression at least 4–6 weeks before the
manuscript and the Lands and Survey Department Kuala Lumpur, Malaysia. 6Medical Research
point at which population suppression was of the Cayman Islands Government for permission Council Centre for Outbreak Analysis and
detected (5 August 2010; Fig. 2c). This to use imagery and data. A.F.H. thanks Adapco, Modelling, Department of Infectious Disease
indicates that suppression was achieved by Bayer and Central Life Sciences for supporting her Epidemiology, Faculty of Medicine, Imperial
releasing approximately 3,500 males per ha PhD studentship. I.B. and A.C. are PhD students College London, St Mary’s Campus, London, UK.
supported by Biotechnology and Biological Sciences e-mail: luke.alphey@oxitec.com
per week, which gave a male-to-female ratio
Research Council Collaborative Awards in Science
in adult traps of 3:1, corresponding to an
npg

and Engineering. C.A.D. acknowledges the UK 1. Harris, A.F. et al. Nat. Biotechnol. 29, 1034–1037
overflooding ratio of ~5:1. Medical Research Council for Centre funding (2011).
This trial was conducted in an area support. 2. Atkinson, M.P. et al. Proc. Natl. Acad. Sci. USA 104,
9540–9545 (2007).
with no conventional control targeting 3. Phuc, H.K. et al. BMC Biol. 5, 11 (2007).
A. aegypti and a relatively high initial COMPETING FINANCIAL INTERESTS
4. Alphey, N., Alphey, L. & Bonsall, M.B. PLoS ONE 6,
The authors declare competing financial interests:
population density; larviciding and removal details are available at http://www.nature.com/
e25384 (2011).
5. Dyck, V.A., Hendrichs, J. & Robinson, A.S. Sterile
of breeding sites within an integrated doifinder/10.1038/nbt.2350. Insect Technique: Principles and Practice in Area-Wide
program would greatly reduce the number Integrated Pest Management (Springer Netherlands,
of sterile mosquitoes required. The positive 2005).
Angela F Harris1,2, Andrew R McKemey3, 6. Alphey, L. et al. Vector Borne Zoonotic Dis. 20,
outcome and successful demonstration of Derric Nimmo3, Zoe Curtis3, Isaac Black3,4, 295–311 (2010).
population suppression is encouraging for Siân A Morgan3, Marco Neira Oviedo3, 7. Helinski, M., Parker, A. & Knols, B.G.J. Malar. J. 5, 41
(2006).
genetic control strategies in general and, Renaud Lacroix3, Neil Naish3, Neil I Morrison3, 8. Benedict, M.Q. & Robinson, A.S. Trends Parasitol. 19,
in particular, validates the potential of Amandine Collado3,4, Jessica Stevenson3, 349–355 (2003).
OX513A RIDL mosquitoes for population Sarah Scaife3, Tarig Dafa’alla3, 9. Alphey, L., Nimmo, D., O’Connell, S. & Alphey, N. in
suppression. Guoliang Fu3,4, Caroline Phillips3, Transgenesis and the Management of Vector-Borne
Disease Vol. 627 (ed. Aksoy, S.) 93–103 (Landes
Andrea Miles3, Norzahira Raduan3,5, Bioscience, 2008).
Note: Supplementary information is available at http:// Nick Kelly1, Camilla Beech3, Christl A Donnelly6, 10. Thomas, D.D., Donnelly, C.A., Wood, R.J. & Alphey, L.S.
www.nature.com/doifinder/10.1038/nbt.2350. William D Petrie1 & Luke Alphey3,4 Science 287, 2474–2476 (2000).

830 volume 30 number 9 September 2012 nature biotechnology


p at e n t s

Teva v. AstraZeneca and secret prior art under


102(g)(2)
Sandra Lee & Michael Knierim
A recent Federal Circuit decision and the reasoning behind it could have a significant impact on the patentability of
other life science inventions, even after changes in the patent law.

O n December 1, 2011, the United States Background this was the case in Teva v. AstraZeneca. The
© 2012 Nature America, Inc. All rights reserved.

Court of Appeals for the Federal Circuit Section 102(g) was originally intended to codify accused product was Crestor (rosuvastatin),
issued its opinion in Teva v. AstraZeneca1 con- the law of interferences4. Nevertheless, courts a well-known stabilized statin formulation
cerning an oft-neglected prior art provision— have long interpreted it as capable of being used for the treatment of high cholesterol. Statins
35 USC § 102(g)(2). On its face, the opinion as a prior art provision, not limited to interfer- are widely known to treat high cholesterol,
reiterates well-established principles relating to ences5. In 1999, Congress amended Section but are inherently unstable. To be medically
102(g)(2) prior art, which is often referred to 102(g), separating it into two parts to clarify its viable, statins must be manufactured in sta-
as ‘secret’ prior art because it lacks a publicity dual role as both an interference provision and a bilized formulations.
requirement. Thus, something not yet publicly prior art provision6. Subsection 102(g)(2), being AstraZeneca first produced the accused
disclosed as required by Section 102(a) and the prior art provision, presently reads, “A per- product in mid-1999. The product contained
(b) can be prior art under Section 102(g)(2). son shall be entitled to a patent unless…before the following ingredients: (i) a statin; (ii) cro-
However, lurking within the language of the such person’s invention thereof, the invention spovidone, which is an amido group containing
opinion’s text, there are two easily overlooked, was made in this country by another inventor polymeric compound; and (iii) tribasic calcium
but nevertheless striking, facets to this case: the who had not abandoned, suppressed, or con- phosphate, which has a stabilizing effect on
court’s broad characterization of the invention, cealed it. In determining priority of invention statin formulations. At the time of the first pro-
and the fact that AstraZeneca actually invali- under this subsection, there shall be considered duction, AstraZeneca understood that tribasic
dated Teva’s claim without meeting each and not only the respective dates of conception and calcium phosphate was a stabilizer. However,
every limitation. reduction to practice of the invention, but also AstraZeneca failed to appreciate that crospo-
The decision came just months after passage the reasonable diligence of one who was first to vidone had a stabilizing effect, but instead was
npg

of the America Invents Act (AIA), the long- conceive and last to reduce to practice, from a adding it as a disintegrant.
awaited patent reform law which will effectively time prior to conception by the other.” The claim at issue was directed to a stabilized
remove Section 102(g) from Title 35 of the US Section 102(g)(2) has two main requirements. statin formulation, where the stabilizing com-
Code on March 16, 2013 (ref. 2). Although First, the inventive activity must occur in the pound is either an amido-group or amino-group
some might be tempted to dismiss this opinion United States. Second, the invention must be containing polymeric (AGCP) compound or
as obviated by the new provisions of the AIA, the continuously used and “not abandoned, sup- combination thereof. Teva filed a provisional
decision thus will have relevance at least until pressed, or concealed….” Thus, the Section US patent application on April 10, 2000. It fol-
the current statute is replaced. Moreover, cer- 102(g)(2) analysis is different from that of the lowed up a year later with a utility application,
tain continuing applications claiming priority other provisions of Section 102. Instead of first which issued in May of 2003. A reissue patent
before the effective date will also be governed establishing that a reference is prior art, and then followed in 2007. Teva filed suit in 2008. The
by the existing language of Section 102 (ref. 3). determining if the reference contains each and claim at issue requires (i) a statin; (ii) at least one
With the long pendency times currently expe- every limitation of the claim, Section 102(g)(2) AGCP compound; and (iii) no other stabilizing
rienced at the US Patent and Trademark Office, requires a determination that an inventor has compound.
and the years before pending applications will made “the invention.” AstraZeneca filed a motion for summary
be involved in litigation, the existing language judgment of invalidity under 102(g)(2). For
of Section 102 will be relevant for many years to Teva v. AstraZeneca purposes of the motion, AstraZeneca stipulated
come, both in prosecution and litigation. Prior art under Section 102(g)(2) often arises infringement. The motion was successful, and
when a party has reduced to practice a prod- Teva appealed to the Federal Circuit. On appeal,
Sandra Lee is a partner and Michael Knierim uct before the effective filing date of a patent. Teva raised four arguments. The first and central
is an associate at Baker Botts, New York, New It should be noted that although this gives argument was that Section 102(g)(2) required
York, USA. Section 102(g)(2) the flavor of prior use, AstraZeneca to show an appreciation at the time
e-mail: sandra.lee@bakerbotts.com 102(g)(2) is not a prior use defense. Indeed, of invention that crospovidone had a stabilizing

nature biotechnology volume 30 number 9 SEPTEMBER 2012 831


pat e n t s

effect. In support of this argument, Teva relied prove appreciation of what it had made, not the claim, as the claim includes the negative
on case law holding that accidental discovery why or how it worked. Additionally, the inven- limitation. Yet, the court still held that pro-
does not give rise to prior inventive activity for tor need not understand the invention in the duction of Crestor by AstraZeneca in 1999
purposes of Section 102(g)(2) (ref. 7). same terminology a plaintiff subsequently was sufficient inventive activity to invalidate
Teva’s second argument was that the US uses to claim it. Applying this distinction to Teva’s claim.
District Court for the Eastern District of Teva’s remaining arguments, the court held The language of the opinion provides two
Pennsylvania had implicitly construed the that Teva’s second argument regarding broad plausible justifications for this outcome. On
claim in an overbroad fashion to include a claim construction is tantamount to argu- one hand, the court stated that “because of
product containing a plurality of stabilizing ing that AstraZeneca needed to understand AstraZeneca’s limited concession of infringe-
compounds. This argument seems counter- its invention in the same terms used in the ment, there is no question that the amount of
intuitive, particularly in light of the fact that asserted claims. As to Teva’s third argument, crospovidone AstraZeneca’s drug contained
neither party disputes that Crestor contains the court noted that this case does not raise falls within the scope of the asserted claims
tribasic calcium phosphate, a stabilizer that is issues of inherency. Rather, AstraZeneca’s con- as defined by the limitation ‘stabilizing effec-
not an AGCP compound. Closely paralleling cession of infringement established that the tive amount’”1. This raises the consideration
its second argument, Teva’s third argument was accused product is an embodiment within the that but for AstraZeneca’s concession, the
that the district court had improperly applied scope of the claims. Finally, the court rejected court would not be bound to accept that
inherency precedent from the 102(b) context Teva’s fourth argument, which presupposed Crestor is an embodiment within the scope
without requiring AstraZeneca to prove appre- that AstraZeneca needed to appreciate the sta- of the claims, and thus perhaps not Section
ciation. And finally, Teva’s fourth argument was bilizing effect of crospovidone, as that issue had 102(g)(2) prior art.
that, assuming AstraZeneca was in fact first to already been decided to the contrary. On the other hand, the court characterized
invent, AstraZeneca suppressed or concealed “the invention” as a broad concept sufficient
© 2012 Nature America, Inc. All rights reserved.

its invention by failing to disclose that crospo- Implications to invalidate a narrower claim. That is, the
vidone stabilized the formulation. The court’s holding and reasoning appears to court considered the invention as a formula-
The court dealt with Teva’s arguments in turn. be in line with precedent, as it is not a new idea tion including an AGCP stabilizer, and Teva
The court held that to establish prior invention, that Section 102(g) requires appreciation of simply expressed the invention in a differ-
the party asserting it must prove that it appreci- what was made, but not why it works. In 1972, ent way by claiming an additional negative
ated what it had made. The prior inventor does limitation. This justification is buttressed by
not need to know everything about how or why Section 102(g)(2) will be the court’s extensive citation to the proposi-
the invention worked, or conceive of its inven- tion that a prior inventor need not describe
tion in the same words as the patentee would applicable to many cases for his invention in the same language as subse-
later use to claim it1. Thus, the court held that years to come. However, the quently claimed.
AstraZeneca needed only to appreciate that its
invention was stable and to identify the compo- Federal Circuit opinion in Teva v. The America Invents Act
nents of the invention. AstraZeneca provides an early As noted previously, Section 102(g)(2) will be
To arrive at this position, the court cited applicable to many cases for years to come.
Dow Chemical Co. v. Astro-Valcour, Inc.8, reference point from which to However, the Federal Circuit opinion in Teva v.
Mycogen Plant Sciences v. Monsanto Co.9 and compare the changes brought AstraZeneca provides an early reference point
Invitrogen Corp. v. Clontech Labs10. In Dow from which to compare the changes brought
and Mycogen, the court found prior inventive
about by the America Invents Act. about by the AIA.
npg

activity sufficient for Section 102(g)(2) even For claims filed after March 16, 2013, the
though the prior inventor did not appreciate the Court of Customs and Patent Appeals, the new Section 102 will apply. There is, of course,
certain aspects of the invention. In Dow, the Federal Circuit’s predecessor court, albeit in the an exception for certain continuing applica-
prior inventor did not appreciate the patent- interference context, held that “[a]n inventor tions claiming priority before the effective
ability of his invention, but recognized what need not understand precisely why his inven- date of the AIA. Notwithstanding the excep-
it was and that it was beneficial. In Mycogen, tion works in order to achieve an actual reduc- tion, the new Section 102 lacks a direct analog
the prior inventor had also appreciated what tion to practice”11. to old 102(g). However, the AIA does provide
he had made, but described his invention as a Claim 1 has three main limitations. The a limited substitute to Section 102(g)(2). The
process of modifying the frequency of nucleo- claim requires a statin, a stabilizing effective prior user rights defense, formerly limited to
tides instead of modifying the frequency of amount of at least one AGCP compound, business methods, has been expanded and
codons. In Invitrogen, which Teva had relied and “wherein said stabilized pharmaceuti- reinvigorated with the passage of the AIA. The
heavily upon, the court held that prior inven- cal composition does not contain a stabiliz- “prior commercial use defense,” as it is now
tive activity consisting of randomly altering a ing effective amount of another stabilizer or referred to, expands the defense to all catego-
panel of mutant genes falls within a category a combination or other stabilizers.”12 Crestor ries of patentable subject matter. Although the
of accidental duplication cases insufficient for contains a statin, and thus the first limitation prior commercial use defense in some respects
prior inventive activity under 102(g)(2). is met. It also contains crospovidone, which is similar to Section 102(g)(2), there are some
Notwithstanding the differences in out- is an AGCP compound and has a stabilizing notable differences.
come, the court held that “Dow, Mycogen Plant effect. Thus, the second limitation is met. A major difference is that the prior com-
Sciences, and Invitrogen are consistent applica- Finally, it contains tribasic calcium phos- mercial use defense is not a prior art pro-
tions of the same rule”1. The court noted that phate, which is a stabilizing compound, but vision—successfully asserting it will not
the distinguishing characteristic to establish is not an AGCP compound. Thus, Crestor invalidate a patent13. However, although this
prior invention is that an inventor needs to does not meet each and every limitation of is a personal defense, the sale or disposition

832 volume 30 number 9 SEPTEMBER 2012 nature biotechnology


pat e n t s

of a useful end result by a person entitled to with precedent that holds, for purposes of 4. Rich, G.S. Speech to the New York Patent Law
assert the defense shall exhaust the patent Section 102(g)(2), that a prior inventor need Association (Nov. 6, 1951).
5. See Thomson, S.A. v. Quixote Corp., 166 F.3d 1172,
owner’s rights with respect to the purchaser. not describe his prior invention in the same n. 3 (Fed. Cir. 1999).
To raise the prior commercial use defense, a words as subsequently claimed, the court has 6. S. 1948 § 4806 (1999).
defendant must commercially use the subject invalidated a patent with Section 102(g)(2) 7. See Invitrogen Corp. v. Clontech Labs., Inc., 429 F.3d
1052 (Fed. Cir. 2005).
matter that would otherwise infringe a patent secret prior art that failed to meet each and 8. Dow Chemical Co. v. Astro-Valcour, Inc., 267 F.3d
in the United States at least one year before the every limitation of the claim. In light of the 1334 (Fed. Cir. 2001).
earlier of the effective filing date or Section plausible justifications for its outcome, this 9. Mycogen Plant Sciences v. Monsanto Co., 243 F.3d
1316 (Fed. Cir. 2001).
102(b) public disclosure date of the patent case’s impact on other provisions of Section 10. Invitrogen Corp. v. Clontech Labs., 429 F.3d 1052
being asserted. Finally, if the defense is raised 102 remains uncertain, but could prove to (Fed. Cir. 2005).
without a reasonable basis, the case will be be significant. 11. Parker v. Frilette, 462 F.2d 544, 547 (CCPA 1972).
12. US Patent No. RE39,502, Claim 1 (emphasis added)
made exceptional for the purpose of awarding 13. H.R. 1249 at § 5 (amending 35 USC § 273(g) to recite
attorneys fees14. COMPETING FINANCIAL INTERESTS “[a] patent shall not be deemed to be invalid under
The authors declare no competing financial interests. section 102 or 103 solely because a defense is raised
or established under this section.”).
Conclusions 14. H.R. 1249 at § 5 (amending 35 USC § 273(f) to recite
1. Teva v. AstraZeneca. No. 2011–1091 (Fed. Cir., Dec.
Although the central holding that Section 1, 2011). “[i]f the defense under this section is pleaded by a
102(g)(2) requires an appreciation of what 2. H.R. 1249, 112th Cong. § 3 (2011). person who is found to infringe the patent and who
3. Chisum, D.S. Priority among competing patent appli- subsequently fails to demonstrate a reasonable basis
was made, but not why or how it works, for asserting the defense, the court shall find the case
cants under the American Invents Act, § II.A (2011).
the Federal Circuit has taken this line of <http://papers.ssrn.com/sol3/papers.cfm?abstract_ exceptional for the purpose of awarding attorney fees
precedent one step further. In conjunction id=1969592> under section 285.”).
© 2012 Nature America, Inc. All rights reserved.
npg

nature biotechnology volume 30 number 9 SEPTEMBER 2012 833


pat e n t s

Recent patent applications in biosensors


Priority
Patent number Description Assignee Inventor application date Publication date
WO 2012098758 A detection apparatus for a biosensor with National Institute of Akiyama S, 1/20/2011 7/26/2012
an optical prism incident surface and a Advanced Industrial Fujimaki M,
detection plate adhesion surface that are set Science and Technology Nagata K
at a specific angle; useful in, e.g., the medical (Tokyo)
field.
MX 2010012079 A regenerable biosensor for determining the Autonomous University Salas BV, 1/13/2011 7/19/2012
enzyme activity of protease. The biosensor is of Baja California Stoytcheva MS,
provided with a resonator that is provided with (Mexicali, Mexico) Zlatev RK
a quartz microbalance unit, where the quartz
microbalance unit is provided with a coercive
power of 10,000–40,000 units, ensuring a
simple and efficient biosensor.
WO 2012096162 A sensor chip for measuring properties of Panasonic Hashimotodani K, 1/13/2011 7/19/2012
substances, e.g. cells, comprising a diaphragm (Osaka, Japan) Nakano Y,
provided with two different surfaces and Nakatani M, Oka T,
through-hole(s) formed between surfaces, Ushio H, Yamada Y,
where a portion of the surfaces and through- Yamamoto T
hole is covered with a silicon layer.
US 20120178178 A biosensor cartridge, e.g., mass-based sensor, Samsung Electronics Choi YS, Do Jae P, 1/6/2011 7/12/2012
optical sensor, electrical sensor, quartz crystal (Suwon, S. Korea) Han KY, Kim SK,
microbalance, cantilever sensor and surface Lee HJ, Lee JH,
acoustic wave sensor, for automating biosensor Lee JN, Lee SS,
© 2012 Nature America, Inc. All rights reserved.

processes to detect biological material from a Lee YH


biological sample, e.g., blood.
US 20120168306 A biosensor system for performing affinity- University of Texas Hassibi A, Jang B, 11/17/2008 7/5/2012
based detection to detect an analyte in a System (Austin, TX, Manickam A
biological sample. Uses include but are not USA)
limited to toxin, hormone, DNA strand, protein
and bacteria.
JP 2012122744 A method for manufacturing a biosensor Murata Manufacturing Tanigawa M 12/6/2010 6/28/2012
involving adhering an electrode layer and a Co. (Kyoto, Japan)
cover layer on the surface of a spacer layer by
hot melt adhesive.
KR 2012057808 A biosensor useful for detecting gonyautoxin Korea Ocean R&D Jae HL, Lee TK, 11/29/2010 6/7/2012
(GTX) from phytoplankton, comprising anti- Institute (Ansan, Man C
GTX antiserum separated from a combination S. Korea)
of protein and GTX from immunized mouse.
KR 2012057596 A sensibility diagnostic system comprising Yonsei University Hyo IJ, Jung HL 12/30/2011 6/5/2012
a sensibility diagnostic chip equipped with Industrial-Academic
a biosensor, where the sensibility diagnostic Co-operative Foundation
unit converts emotions into an emotional (Seoul)
quotient based on the received signals from
the subject. The system is useful for diagnosis
npg

of sensibility, e.g., stress.


KR 2012055009 A biosensor useful for multiplexed diagnostics, Electronics and Bong PH, Moon YJ, 11/22/2010 5/31/2012
comprising a plane light source, specific Telecommunications Park JW
antibody, detection area, color charge-coupled Research Institute
device camera and long-pass filter. (Daejon, S. Korea)
CN 102477416 A high-salt lipase useful in the printing, Zhang B Fang J, Liu G, Su D, 9/5/2011 5/30/2012
petrochemical and environmental industries, Zhang B
single-cell protein production, cosmetic
production, biological material, biosensor
and biomedical fields. The high-salt lipase is
cost effective and has strong selectivity and
specificity with improved enzyme activity,
stability and practicability.
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).

834 volume 30 number 9 September 2012 nature biotechnology


news and views

Modulating WNT receptor turnover for tissue repair


Arie Abo & Hans Clevers
New mechanistic insight into the regulation of WNT signaling supports efforts to target this pathway in adult stem
cells for regenerative medicine.

Experimental cell therapies often involve cell a b


transplantation, but for some conditions it may
be advantageous to follow a simpler strategy of
Enterocytes
enhancing tissue repair by coaxing endo­genous RSpo
RSpo
stem cells to proliferate and differentiate. This WNT
© 2012 Nature America, Inc. All rights reserved.

approach has shown promise in animal models


Internalization Ubiquitnation
of disease, where potentiation of the Wnt
pathway in stem cells by the RSpondin (RSpo)
family of secreted proteins stimulates tissue TA
compartment LGR5 ZNRF3 LRP5/6 Frizzled
regeneration and healing. In a recent study in
Nature, Hao et al.1 reveal important new infor-
mation about the mechanism by which RSpo Membrane clearance Degradation Activation
BMI1+
proteins boost WNT signaling. They show that +4 position
activation of the WNT receptor is inhibited by Paneth cell

its interaction with ZNRF3, a transmembrane LGR5+


crypt stem cell
E3 ubiquitin ligase. RSpo proteins effectively
sequester ZNRF3 by inducing dimerization of
ZNRF3 with the LGR4 receptor, thereby activat- Figure 1 Modulating LGR5+ crypt stem cells with RSpo proteins for regeneration of the gastrointestinal
ing WNT signaling (Fig. 1). Although Hao et epithelium. (a) The architecture of the crypt. LGR5+ cells (yellow) reside between the Paneth cells
(green) at the bottom of the crypt, and BMI1+ cells (red) reside at position +4. LGR5+ cells are WNT
al.1 worked with HEK293 cells rather than with
dependent and can give rise to all cell lineages of the gastrointestinal epithelium. (b) Molecular
stem cells, expression of LGR receptors in adults mechanism by which RSpo proteins activate LGR5+ stem cells by activating WNT. In the absence of
appears to be restricted to stem cells. Thus, their RSpo proteins, ZNRF3 ubiquitinates the WNT receptor Frizzled, targeting it for degradation. In the
study has broad implications for regenerative
npg

presence of RSpo proteins, ZNRF3 dimerizes with LGR5 receptor, leading to removal of ZNRF3 from
medicine as it identifies a new RSpo-protein tar- the cell surface, accumulation of WNT receptor complex on the membrane and enhancement of WNT
get that may be expressed in adult stem cells and signaling. TA, transit-amplifying cells.
further validates the importance of WNT com-
plex turnover for therapeutic intervention. Wnt3a delivered in liposomes was shown to recent studies implicating the stem cell markers
The Wnt signaling pathway regulates the stimulate skeletal mesenchymal stem cells and Lgr4 and Lgr5 in RSpo protein–mediated Wnt
expression of genes that are critical for embryonic promote bone repair2. In another strategy, acti- signaling7. Lgr4 and Lgr5 are homologous
development and tissue homeostasis. However, vation of Wnt signaling with antibodies against orphan G-protein-coupled receptors that have
the possibility of exploiting this pathway for the Wnt pathway molecule Dkk1 was used to been identified as specific markers of adult
regenerative medicine has been hampered by enhance bone formation in a mouse model of stem cells. In the gastrointestinal tract, they
difficulties in producing recombinant WNTs in rheumatoid arthritis3. RSpo proteins were the are expressed only on stem cells, which reside
suf­ficient quantities for systemic administration. first Wnt signaling modulators to be synthesized among the Paneth cells at the bottom of the
In one recent study using local administration, in relatively large amounts and administered crypt (H.C. and colleagues)8 (Fig. 1). Crypt
systemically to animals (A.A. and colleagues)4–6. stem cells play a critical role in homeostasis and
Arie Abo is at the California Institute for This work demonstrated the regenerative poten- regeneration of the gastrointestinal epithelial
Regenerative Medicine, San Francisco, tial of RSpo proteins in models of rheumatoid lining. Moreover, single crypt stem cells isolated
California, USA, and Hans Clevers is at the arthritis, oral mucositis, and experimental colitis from the mouse intestine can reconstitute the
Hubrecht Institute, Developmental Biology and showed that tissue regeneration is a direct crypt-villus structure ex vivo when appropriate
and Stem Cell Research, Royal Netherlands result of potentiating the Wnt pathway in stem growth factors are provided7. If RSpo proteins
Academy of Arts and Sciences Utrecht, 3584CT cells, which triggers expansion and regeneration are added to cultured mouse Lgr5+ cells, the
Utrecht, The Netherlands. of the epithelial layer and bone. cells form large crypt-villus organoids7, simi-
e-mail: aabo@cirm.ca.gov or The missing molecular link between RSpo lar to intestinal crypt and villi structures, that
h.clevers@hubrecht.eu proteins and stem cells emerged with our contain fully polarized enterocytes with mature

nature biotechnology volume 30 number 9 september 2012 835


news and v iews

brush borders, Goblet cells and enteroendocrine identify ZNRF3 as a potential target for thera- activation has been reported in numerous can-
cells. All RSpo family members (RSpo1–4) bind peutic intervention.  Although the authors did cers, including breast and colon cancer. One
with high affinity to LGR4 and LGR5 receptors not study stem cells, they demonstrated the role of the first mutations in colon cancer is in the
and thereby potentiate WNT signaling (H.C. of ZNRF3 in regulating the WNT pathway not tumor suppressor adenopolyposis coli gene, a
and colleagues)9,10. However, the mechanisms simply in cultured cells but also by knockdown mutation that leads to constitutive activation
underlying this activity of RSpo proteins have experiments in zebrafish and mice, providing of the WNT pathway, and tumorigenesis may
not been well understood. important in vivo validation. be accelerated by chronic inflammation in the
The WNT receptor complex is a heterodimer Whichever WNT-signaling targets and drugs colon.  Therefore, in inflammatory bowel dis-
of the LRP6 and Frizzled receptors. RSpo pro- ultimately prove most effective, enhancing tis- ease patients, treatment with agents such as
teins were previously shown to activate WNT sue repair by activating endogenous stem cells RSpo proteins may reduce the risk of tumor­
signaling by blocking LRP6 internalization is likely to be beneficial for a diverse group of igenesis by repairing damaged mucosa and
through a mechanism that involves inhibition of disorders. In the gastrointestinal tract, several thereby  minimizing exposure to inflamma-
the interaction of DKK1 and Kremen (A.A. and diseases are associated with injuries of the epi- tory reactive oxidants5. More generally, WNT
colleagues)11. Dkk1 inhibits Wnt signaling by thelial lining, including inflammatory bowel dis- signaling modulators such as RSpo proteins
binding to and inducing dimerization of Lrp6 ease (comprising Crohn’s disease and ulcerative may be safer than approaches that use WNT
with the transmembrane proteins Kremen1 and colitis), which involves destruction of mucosal ligands because they potentiate the regenera-
Kremen2, followed by rapid endocytosis of Lrp6 integrity and chronic exposure to gut microbial tion of endogenous damaged tissue without
from the plasma membrane12. The new work by flora.  Current therapies for inflammatory bowel violating the order of events essential for tis-
Hao et al.1 reveals a second, similar mechanism disease all work through an anti-inflammatory sue restoration. Thus, WNT modulators could
for regulating WNT receptor complex turnover, mechanism using agents such as nonsteroidal be used to potentiate WNT-dependent repair
which operates not through DKK1/Kremen but anti-inflammatory drugs or biologics that target in specific tissues rather than throughout the
© 2012 Nature America, Inc. All rights reserved.

through the ZNRF3 axis (Fig. 1). the pro-inflammatory cytokine tumor necrosis body, avoiding potential side effects caused by
The authors began by carrying out a genomic factor alpha. However, additional therapies are growth factors such as WNT ligands.
analysis to discover genes that are upregulated clearly needed to treat primary and secondary,
in primary colorectal tumors exhibiting hyper- nonresponsive patients. Although the role of COMPETING financial INTERESTS
The authors declare no competing financial interests.
activity of the WNT pathway, which identified stem cells in gastrointestinal disease has not yet
the candidate genes ZNRF3 and RNF43. These been elucidated, targeting epithelial stem cells to 1. Hao, H.X. et al. Nature 485, 195–200 (2012).
genes were found to be functional homologs accelerate mucosal repair should be beneficial 2. Minear, S. et al. Sci. Transl. Med. 2, 29–30 (2010).
of each other that negatively regulate WNT in some cases. So far, the function of Lgr5+ stem 3. Diarra, D. et al. Nat. Med. 13, 156–163 (2007).
4. Zhao, J. et al. Gastroenterology 132, 1331–1343
signaling. Knocking down ZNRF3 in HEK293 cells has been studied mostly in stomach, small (2007).
cells led to higher cell-surface levels of the intestine, colon and hair follicle, but this marker 5. Zhao, J. et al. Proc. Natl. Acad. Sci. USA 106,
LRP6 and Frizzled receptors, whereas over­ has also been detected in mammary gland, 2331–2336 (2009).
6. Krönke, G. et al. Arthritis Rheum. 62, 2303–2312
expressing ZNRF3 promoted ubiquitinylation brain and, after tissue damage, in pancreas and (2010).
of the Frizzled receptor and its subsequent liver.  The roles of these cells in tissue regenera- 7. Sato, T. et al. Nature 459, 262–265 (2009).
8. Barker, N. et al. Nature 449, 1003–1007 (2007).
degradation. Experiments with RSpo proteins tion is the subject of ongoing research (H.C.). 9. Carmon, K.S. et al. Proc. Natl. Acad. Sci. USA 108,
showed that RSpo 2, 3 and 4 increased cell- In developing strategies for translating WNT 11452–11457 (2011).
surface levels of the LRP6 and Frizzled signaling biology to regenerative medicine, we 10. de Lau, W. et al. Nature 476, 293–297 (2011).
11. Binnerts, M.E. et al. Proc. Natl. Acad. Sci. USA 104,
receptors and that RSpo1 decreased ubiquit- must not overlook the potential risks associated 14700–14705 (2007).
inylation of the Frizzled4 receptor. with overactivation of this pathway. WNT
npg

12. Mao, B. et al. Nature 417, 664–667 (2002).


Hao et al.1 also investigated the roles of LGR4
and LGR5. RSpo protein–mediated turnover
of Frizzled receptors required LGR receptors.
Moreover, ubiquitinylation and cell-surface
expression of Frizzled receptors 4, 5 and 8 were
RNA-mediated programmable DNA
abolished when LGR4 receptor was knocked
down with short interfering RNA or when the cleavage
cells were treated with RSpo proteins. In addi-
tional experiments to unravel the links between Rodolphe Barrangou
LGR receptors, RSpo proteins and ZNRF3, RSpo
proteins were found to mediate interaction of Genome engineering has a new tool—endonucleases involved in bacterial
LGR4 and ZNFR3 and to induce membrane adaptive immunity that can be reprogrammed with customizable small,
clearance of ZNRF3.  Furthermore, dimeriza-
noncoding RNAs.
tion of LGR4 and ZNRF3 either artificially or
by RSpo proteins led to endocytosis of ZNRF3
and accumulation of WNT receptor complex Genome engineering methods rely on site- cleavage site. Enzymes that are currently
on the cell surface. specific endonucleases that trigger sequence widely used for genome engineering include
Enhancing WNT signaling to promote stem modification by DNA-repair systems at the zinc finger nucleases (ZFNs) and transcription
cell proliferation and differentiation is emerg- activator-like effector nucleases (TALENs),
ing as a promising strategy to promote regen- Rodolphe Barrangou is at DuPont Nutrition both of which can be designed to target spe-
eration of a variety of tissues. The results of and Health, Madison, Wisconsin, USA. cific genome sequences. These proteins have
Hao et al.1 further validate this approach and e-mail: rodolphe.barrangou@dupont.com to be engineered for each application to

836 volume 30 number 9 september 2012 nature biotechnology


news and v iews

a Cas9 genes (cas) constitute the CRISPR-Cas system.


RuvC
CRISPR-Cas provides RNA-mediated adap-
tive immunity against viruses and plasmids in
bacteria and archaea2. Immunity is acquired
Target DNA PAM through the integration of invasive genetic ele-
ment sequences as ‘spacers’ into the CRISPR
crRNA locus of the host. Arrays of CRISPR spacers are
HNH transcribed and processed into small interfer-
tracrRNA
ing CRISPR RNAs (crRNAs) that guide Cas
proteins to mediate cleavage of homologous
Cas9 i dsDNA sequences2–4.
b RuvC Among the three types of CRISPR-Cas
RuvC+
HNH– systems that have been identified based on
cas content and sequences5, type I- and type
Target DNA PAM ii III-mediated immunity rely on a large and
multiprotein Cas complex, whereas type II
RNA chimera systems solely (and conveniently) rely on a
RuvC–
HNH+ single Cas protein, Cas9. Although the target
HNH
sequence is defined by the CRISPR spacer,
iii
RuvC+ a strict requirement in many CRISPR-Cas
HNH+ systems is the presence of a proto-spacer–
© 2012 Nature America, Inc. All rights reserved.

c associated motif (PAM) sequence, typically


2–4 nucleotides (nt), located in the immedi-
FokI ate vicinity of the spacer target. Most cur-
rent applications of CRISPR-Cas systems
Target DNA focus on increasing the breadth of phage
resistance in industrial bacteria, leveraging
FokI CRISPR spacer hypervariability for genotyp-
TAL-effector
DNA binding ing and epidemiological surveys and exploit-
domain ing iterative spacer additions for analysis of
host-virus dynamics6. The ability of func-
tional CRISPR-Cas systems to be trans-
ferred across even distant bacterial genera
d and species had previously set the stage for
FokI
ZFPs bind DNA
heterologous applications7.
Jinek et al.1 show that the Cas9-crRNA
Target DNA ribonucleoprotein complex mediates specific
FokI
DNA cleavage for CRISPR-encoded immunity
in bacteria. A trans-activating CRISPR RNA
npg

Figure 1 crRNA-guided DNA cleavage by Cas9. (a) The Cas9 protein (light blue) combines with (crRNA), termed tracrRNA, pairs with another
crRNA (red) and tracrRNA (orange) to form a ribonucleoprotein interfering complex. The crRNA distinct crRNA that guides Cas9 to homo­l­
sequence guides the interference complex to a complementary sequence in the target DNA (dark ogous target sequences (Fig. 1a). In another
blue). Once an R-loop is formed (see open structure), the HNH and RuvC Cas9 domains nick the
series of experiments, the authors demon-
complementary and noncomplementary strands, respectively, 3 nt away from the PAM sequence.
(b) Reprogramming of DNA cleavage by an RNA chimera molecule that includes partial tracrRNA linked strate that the endonuclease HNH and RuvC
to a target-specific customized crRNA (purple). Single-stranded DNA nicking can be achieved by mutation (a Holliday junction resolvase) domains of
of either the HNH motif (i) or the RuvC motif (ii), whereas double-stranded cleavage can be achieved Cas9 nick the complementary and noncomple-
using the two wild-type domains (iii). (c) TALEN DNA cleavage. TAL effector DNA binding domains on the mentary DNA strands, respectively, generating
left and on the right recognize the left and right target sequence, respectively, and trigger dsDNA cleavage a dsDNA blunt cleavage. The ability to nick or
through the FokI nuclease domain. (d) ZFN DNA cleavage. Zinc finger proteins (ZFPs) recognize target
cleave DNA at defined and discrete locations in
DNA sequences and trigger dsDNA cleavage through the FokI nuclease domain.
a DNA sequence opens new avenues for edit-
ing and assembling DNA sequences in vitro.
reprogram the cleavage site to specifically tar- synthetic RNA-mediated reprogramming of In an elegant proof-of-concept experiment, the
get the DNA sequence of interest. In a recent cleavage specificity. Several immediate applica- authors demonstrate that Cas9 DNA cleavage
paper in Science, Jinek et al.1 find that two RNA tions of this approach can be envisioned, nota- specificity (minimally defined by 13-nt perfect
molecules direct DNA cleavage in a ribonucleo- bly customized single-strand DNA nicking and base-pairing between crRNA and target DNA)
protein complex that forms part of the adaptive double-stranded (ds)DNA cleavage for genome can be reprogrammed to target a green fluores-
immune system of bacteria, and that a single engineering and editing. cent protein gene using a chimeric RNA fusion
RNA chimera can be engineered to repro- The molecular machinery in the Jinek et al.1 of the crRNA 3′ end with the tracrRNA 5′ end
gram sequence specificity. This synthetic tour study is part of the clustered regularly inter- (Fig. 1b). Arguably, this combines the flex-
de force introduces a new endonuclease family spaced short palindromic repeats (CRISPR), ibility of RNA interference with the power of
to the genome engineering arsenal that enables which, together with CRISPR-associated DNA restriction enzymes to generate single- or

nature biotechnology volume 30 number 9 september 2012 837


news and v iews

double-stranded DNA breaks in any sequence Although immediate applications of this COMPETING FINANCIAL INTERESTS
for genome targeting or editing. new tool include customized DNA nicking The author declares competing financial interests:
details are available in the online version of the paper.
Several studies have shown that ZFNs and/or cleavage in bacteria, there are intriguing
and TALENs (Fig. 1c,d) can be engineered possibilities for genome editing and genome 1. Jinek, M. et al. Science advance online publication,
to induce mutations at specific locations in engineering of eukaryotes. This will require doi:10.1126/science.1225829 (28 June 2012).
2. Wiedenheft, B., Sternberg, S.H. & Doudna, J.A. Nature
genome sequences8,9 in plants (Zea mays), testing whether crRNA-Cas systems can effi- 482, 331–338 (2012).
bacteria (Escherichia coli), model animals ciently cleave chromatin DNA in vivo and be 3. Deltcheva, E. et al. Nature 471, 602–607 (2011).
4. Garneau, J.E. et al. Nature 468, 67–71 (2010).
(Drosophila melanogaster), human embryonic readily transferred into organisms of inter-
5. Makarova, K.S. et al. Nat. Rev. Microbiol. 9, 467–477
stem cell lines and induced pluripotent stem est, notably yeast and fungi, but also plants, (2011).
cell lines. However, these proteins have to be for crop and agricultural applications, and 6. Andersson, A.F. & Banfield, J.F. Science 320,
1047–1050 (2008).
redesigned for every different DNA sequence human cells, for medical purposes. Only 7. Sapranauskas, R. et al. Nucleic Acids Res. 39,
targeted. The 3-nt recognition motif of ZFNs the future will tell whether this program- 9275–9282 (2011).
can be promiscuous and generate off-target mable molecular scalpel can outcompete 8. Cermak, T. et al. Nucleic Acids Res. 39, 7879–7889
(2011).
cleavage, and target frequency has been esti- ZFN and TALEN DNA scissors for precise 9. Miller, J.C. et al. Nat. Biotechnol. 29, 143–148 (2011).
mated at ~1 in every 500 bp8. The ability to genomic surgery. 10. Kim, E. et al. Genome Res. 22, 1327–1333 (2012).
modulate TALEN specificity by modification
of the repeat variable di-residues within their
amino acid sequence has been an advantage
of TALENS compared with ZFNs9. However,
the nonspecific nuclease domain of FokI in Silicon dreams of cells into
TALENs can generate nonspecific and non­
symbols
© 2012 Nature America, Inc. All rights reserved.

targeted cleavage. Also, the sole product is


strictly dsDNA cleavage, and target frequencies
have been estimated at ~1 in every 35 bp8. Jeremy Gunawardena
Chimeric RNA-Cas9 systems could provide
several advantages over ZFNs and TALENs, Diverse mathematical models combine to create a comprehensive whole-cell
including sequence-recognition specificity (by
computational model of a human pathogen.
customized crRNA sequences), PAM dinucleo­
tides that can be frequent (GG in the case of
Jinek et al.1), the convenience of redesign- Writing in Cell, a team from Stanford Entelos, which models dynamic interactions
ing nucleotides in nucleic acid sequences as University and the J. Craig Venter Institute among organ systems in diseases like diabetes6.
opposed to modifying amino acids in protein has reported a whole-cell simulation of The origins of the M. genitalium simulation
sequences, and the ability to nick either or both Mycoplasma genitalium, a urogenital parasite lie in the constraint-based models of whole-cell
DNA strands. The latter provides a molecular adored by synthetic biologists for its reduced metabolism developed by Bernhard Palsson7.
basis for precise genome editing by a program- genome, a mere snip at 525 genes. This auda- Such a constraint-based model implements one
mable DNA nicking enzyme10. Because Cas9 cious accomplishment by Karr et al.1 of what of the 28 processes into which Karr et al.1 have
has two distinct domains (RuvC and HNH), has been called “a grand challenge of the 21st partitioned the cell’s operations (Fig. 1). These
which each nick a specific DNA strand, wild- century”2 goes beyond previous attempts3 in its include processes that track exchanges with
type Cas9 and functional domain mutants comprehensiveness and especially in capturing the extracellular medium and all the metabolic
npg

can generate either dsDNA breaks or single- the elusive and subtle chicken-and-egg recur- fluxes, the state of the supercoiled chromosome,
stranded DNA nicks, as desired (Fig. 1b). sion through which a cell creates itself. transcription of all active genes, processing of all
Single-stranded breaks are repaired by error- Models of living processes have a long mRNAs, translation of all proteins, formation of
free homologous recombination at the nicking history in science and fiction. From Jacques de all macromolecular complexes including RNA
site, producing a precise mutation, as opposed Vaucason’s automata to The Terminator, our polymerases and ribosomes, and progress of
to error-prone, nonhomologous, end-joining creations are rooted in a yearning to control the cytokinesis and FtsZ polymerization. It would
repair for dsDNA cleavage, which frequently uncontrollable. In the afterglow of cybernetics, be a nightmare to tie all of these processes
creates insertions and deletions after ZFN or Arthur Guyton built a model of the systemic together without a beautiful idea: between
TALEN cleavage. blood circulation4 that still inspires awe, if timesteps of one second, the processes operate
Molecular analysis of Cas9 cleavage has not actual use. Physiologists and bioengineers independently of each other. What links them
shown that both linearized and super- have long been assembling a Virtual Heart5 are 16 variables that record the overall cellular
coiled DNA can be cleaved, suggesting that and have even been aspiring to a Virtual Rat state (e.g., counts of all mRNAs, protein mono-
multiple distinct chimeric RNAs can be con- (http://www.systemscenters.org/centers/ mers, protein complexes). The variables are read
currently or sequentially used to process DNA virtual-physiological-rat/). Industrial appli- by the processes at the start of each timestep
at multiple target sites using a single enzyme, cations include the computational platform and written back at the end, in a loop that
opening avenues for genome stacking and PhysioLab, developed by the biotech company repeats until the cell divides. This takes slightly
shuffling. Moreover, specific cleavage sites can longer than M. genitalium itself does, 10 h
be engineered to generate DNA molecules that on a 128-node cluster that runs the publicly
have custom extremities for iterative genome Jeremy Gunawardena is in the Department available, object-oriented Matlab code.
build-up. Accordingly, crRNA-Cas–directed of Systems Biology, Harvard Medical School, The one-second time-scale separation
nicking and cleavage set the stage for more Boston, Massachusetts, USA. enables two vital features. First, it allows pain-
precise DNA surgery and genome editing. e-mail: jeremy@hms.harvard.edu less implementation of the subtle recursion

838 volume 30 number 9 september 2012 nature biotechnology


news and v iews

16 Cell variables Mathematical models 28 Cell processes


Translation

RNA DNA
Chromosome Replication
Condensation
Probabilistic binding
Supercoiling
Transcript Replication initiation
RNA Transcriptional regulation
Markov model /
Transcription
Polypeptide probabilistic binding
Segregation
Protein monomer Terminal organelle
Boolean
Protein activation
Protein

Complex
Host interaction
RNA pol Aminoacylation
Start Ribosome Mass action
Complexation
Protein folding Cell Yes Finish
FtsZ ring Ribosome assembly division?
Protein modification
Metabolite

Metabolic reaction Protein translocation


Protein processing I and II
Metabolite RNA modification
No
Stochastic Poisson
RNA decay
process
Geometry Protein decay
DNA damage
Host DNA repair
Other

RNA processing
Mass
© 2012 Nature America, Inc. All rights reserved.

Stimulus ODE FtsZ polymerization

Constraint-based models Metabolism


Time
Geometric analysis Cytokinesis

Repeat

Figure 1 Whole-cell simulation of M. genitalium. The functionality of the cell was divided into 28 processes (right column). Each process was independently
modeled using the most appropriate mathematical representation (middle column). The background color of each process matches the background color of
the mathematical representation used to model it. The simulation is broken down into one-second timesteps. At the start of each timestep, the processes read
the values of the 16 cell variables (left column), carry out their computations independently of each other and then update the cell variables at the end of the
timestep. This procedure integrates the different processes, which influence each other only through the cell variables, and is carried out repeatedly until the
simulation terminates at cell division, as measured by the progress of FtsZ polymerization. RNA pol, RNA polymerase; ODE, ordinary differential equation.

mentioned above, by which proteins make the complexity, all that can be deduced from a Because models are only as good as their
the ingredients out of which they themselves model is implicit in its assumptions. Implicit assumptions, they can fall short of expectations.
are made. Robert Rosen struggled with the does not mean obvious: the assumptions Constraint-based models are sometimes touted
profound challenges of such “metabolic underlying the whole numbers, 1, 2, 3,…, as requiring no assumptions beyond reaction
closure”8,9, and successful implementation have kept mathematicians on their toes since stoichiometries. Such models do well at pre-
npg

of it distinguishes this model from any other. Diophantus’s Arithmetica. The assumptions dicting exchange fluxes with the environment
Second, it allows each of the 28 processes to of the M. genitalium model are not so pro- but, as for the internal fluxes, they may not get
be independently implemented by the most found. However, they have not been brought even the directions correct. If thermodynam-
appropriate method (e.g., constraint-based together in this way before. The novelty arises ics is not put in, it does not magically emerge.
models, stochastic processes, discrete models, from looking inwards, from learning whether Similarly, the M. genitalium model cannot pro-
computer code, ordinary differential equa- we really understand what we think we know vide insights into lysine acetylation11 because
tions) (Fig. 1). The right horse can be chosen about how metabolism, transcription, transla- that is not part of its program. Models cannot
for the right course. The 1,900 resulting param- tion and replication recursively create a new supplant the kind of experiments undertaken
eters were estimated using experimental data cell. Indeed, the model yields a very interesting by Luis Serrano and colleagues, whose compre-
garnered from a variety of organisms, from prediction of this kind. Simulations show that hensive unraveling of the metabolome12, tran-
Mycoplasma pneumoniae to Escherichia coli. replication initiation, and replication itself, vary scriptome13 and proteome14 of M. pneumoniae
Karr et al.1 had to wade through over 900 pub- substantially in duration between stochastically revealed how noncoding RNAs and multifunc-
lications in the process. The model is neither different cells, whereas cell cycle durations are tional enzymes can compensate for a reduced
simple nor elegant but it is effective. Knockout tightly clustered around nine hours. According genome. The expectation that, with enough
of individual genes in silico identifies essential to the model, cells that take longer to initiate details, a model will miraculously spring to
genes with 80% accuracy compared to data replication accumulate a greater pool of the life and make such experiments unnecessary
from M. genitalium. dNTPs needed to make DNA so they replicate is the stuff of fiction.
Although such accuracy is encouraging, faster, making up for lost time. Metabolism This is not to say that models cannot tell us
what we always demand of a model is emergent coordinates the cell cycle, independently of about things we do not yet know exist. Physicists
novelty. What we often forget is that a model genetic regulation. It is a beguiling prediction, have spent the gross domestic product of a
is not a description of reality; it is a descrip- but whether artifact or real biology remains to small country confirming the physical exis-
tion of our assumptions about reality10. Despite be tested. tence of something conjured up conceptually

nature biotechnology volume 30 number 9 september 2012 839


news and v iews

in Peter Higgs’ 1964 theoretical paper. Much if whole-cell simulation is to aspire to COMPETING FINANCIAL INTERESTS
the same conjuring has happened in biology. mammalian complexity. The author declares no competing financial interests.
Michaelis and Menten conjured up enzyme- Having pulled off a tour-de-force of com-
substrate complexes more than 30 years before putational biology, it is a shame that Karr 1. Karr, J.R. et al. Cell 150, 389–401 (2012).
2. Tomita, M. Trends Biotechnol. 19, 205–210
Britton Chance showed they existed, using a et al.1 succumb to genocentrism in their title. (2001).
simple model now familiar to all biochemists10. The model no more “predicts phenotype from 3. Ishii, N., Robert, M., Nakayama, Y., Kanai, A. &
Tomita, M. J. Biotechnol. 113, 281–294 (2004).
Mendel conjured up the discrete particles later geno­type” than DNA makes RNA makes pro- 4. Guyton, A.C., Coleman, T.G. & Granger, H.J. Annu. Rev.
called genes using high school–level algebra. tein. DNA does not make anything. For that you Physiol. 34, 13–44 (1972).
These kinds of models trade detail for abstrac- need a cell. How strange it is that cell theory, 5. Noble, D. Physiology (Bethesda) 19, 191–197 (2004).
6. Shoda, L. et al. Clin. Exp. Immunol. 161, 250–267
tion; they focus on a particular question to the the first true theory in biology, predating both (2010).
exclusion of all else. They require more inspira- Darwin’s evolution and Mendel’s genetics, is so 7. Edwards, J.S., Covert, M. & Palsson, B.O. Environ.
tion, less perspiration. readily ignored in our current fixation with the Microbiol. 4, 133–140 (2002).
8. Rosen, R. Life Itself: a Comprehensive Enquiry into
Karr et al.1 already have their sights set on genome, only one of the cell’s many components. the Nature, Origin and Fabrication of Life. Complexity
E. coli, a more experimentally tractable organ- The irony is that the paper itself is profoundly in Ecological Systems (Columbia University Press,
2005).
ism of interest to a broad range of biologists. antigenocentric; metabolic closure could hardly 9. Letelier, J.-C. & Soto-Andrade, J. J. Theor. Biol. 238,
With 4,288 genes, a division time of 30 min be otherwise8. Moreover, the real paper—the 949–961 (2006).
and far more copious data, computational supplementary information, all 120 pages 10. Gunawardena, J. Mol. Biol. Cell 23, 517–519
(2012).
power may become the limiting resource. of it—is modern cell bio­logy compiled into 11. van Noort, V. et al. Mol. Syst. Biol. 8, 571 (2012).
One wonders whether specialized hardware, instructions, symbols and formulas. The authors 12. Yus, E. et al. Science 326, 1263–1268 (2009).
like that which David Shaw and colleagues have published a new kind of molecular and cell 13. Güell, M. et al. Science 326, 1268–1271 (2009).
14. Kühner, S. et al. Science 326, 1235–1240 (2009).
have exploited so spectacularly in molecu- biology textbook—one that is executable. Our 15. Dror, R.O., Dirks, R.M., Grossman, J.P., Xu, H. &
© 2012 Nature America, Inc. All rights reserved.

lar dynamics15, will eventually be required, students will want to download the app. Shaw, D.E. Annu. Rev. Biophys. 41, 429–452 (2012).
npg

840 volume 30 number 9 september 2012 nature biotechnology


perspective

Navigating cancer network attractors for tumor-


specific therapy
Pau Creixell1, Erwin M Schoof1, Janine T Erler2 & Rune Linding1

Cells employ highly dynamic signaling networks to drive dynamics of signaling networks and ultimately cellular phenotype. Next,
biological decision processes. Perturbations to these signaling we describe five general properties of cancer signaling networks (Fig. 1)
networks may attract cells to new malignant signaling and and define five challenges in cancer network biology and propose
phenotypic states, termed cancer network attractors, that strategies to overcome them (Fig. 2). By meeting these challenges,
result in cancer development. As different cancer cells reach network biology may fundamentally advance not only basic biology
© 2012 Nature America, Inc. All rights reserved.

these malignant states by accumulating different molecular but also patient treatment. Finally, we describe how a combination
alterations, uncovering these mechanisms represents a grand of relatively new technologies could become a potent cocktail for the
challenge in cancer biology. Addressing this challenge will discovery of network drugs, and we discuss the practical implementation
require new systems-based strategies that capture the intrinsic of personalized and tumor-specific cancer therapy.
properties of cancer signaling networks and provide deeper
understanding of the processes by which genetic lesions From genomic lesions to functional network perturbations
perturb these networks and lead to disease phenotypes. Tumor cells often harbor hundreds to thousands of genetic lesions. But
Network biology will help circumvent fundamental obstacles based on the observation that some of these genetic lesions are repeat-
in cancer treatment, such as drug resistance and metastasis, edly observed in several cancers (e.g., BRAF V600E, present in >50%
empowering personalized and tumor-specific cancer therapies. of all malignant melanomas5), it has been hypothesized that only a few
genetic lesions are causally implicated in cancer development (‘drivers’),
Cells are constantly computing decisions based on the integration of whereas the majority have no functional consequences (‘passengers’)6.
different cues that reach them at various times. In contrast to single- Although this classification has had some use in identifying
cell organisms, in multicellular organisms, cellular decisions should, mutations that are highly prevalent, it is now apparent that a tumor
ultimately, benefit the organism as a whole, even if that implies that an is not, under any circumstances, a static and uniform population of
individual cell will have to decide to commit suicide. In line with this malignant cells. Rather, it is a dynamic ensemble of subpopulations
unique feature, signaling networks have evolved during multicellular with different abnormalities undergoing molecular evolution7–9.
evolution to allow cells to integrate cues and make decisions that ensure Two fundamental principles of cancer signaling networks can explain
npg

cooperative behavior between them. By hijacking these mechanisms, why a binary driver/passenger classification may be too simplistic to
cancer cells escape cooperative rules and transition from a game gov- accommodate the complex dynamic nature of tumors. First, different
erned by Nash equilibria1,2 between all cells into a new scenario where tumors can develop similar phenotypes by acquiring mutations in
cancer cells decide their behavior purely based on their own benefit, different proteins10, in what we term analogous mutations (Fig. 1a).
or as phrased by Hanahan and Weinberg3, “become masters of their Second, it has been shown that two different mutations not capable
own destinies.” Given the central role played by signaling networks in of causally driving cancer by themselves are able to do so when they
the integration of cues to compute any cellular responses, we argue that appear in combination within the same cells or even within two
cancer is not simply a disease with a genetic basis, but is one ultimately neighboring cells11, in what could be described as two passengers
driven by perturbations at the signaling network level, and that both the becoming drivers or, as we refer to them, synthetic oncogenes (Fig. 1b).
‘cue-signal-response’ rules of cellular decision-making and the switch Thus, patient-to-patient heterogeneity can be driven by the presence
in strategy from cooperative to selfish are major, hitherto understudied, of different mutations in the same or in different proteins that lead to a
hallmarks of cancer3,4. similar signaling state and phenotypic outcome.
In this article, we dissect the strategies cancer cells use to become Altogether, the intrinsic heterogeneity of tumors makes it a pressing
‘selfish’ and drive disease. We first review how genetic lesions can lead to challenge for cancer network biologists to develop tools to identify
altered protein function, which can result in changes to the structure and the extent to which combinations of cancer mutations affect protein
function and cellular and phenotypic states (Fig. 2a,b). Even though
several such tools have been developed (reviewed in ref. 12), existing
1Cellular Signal Integration Group (C-SIG), Center for Biological Sequence
methods are mainly based on protein structure and/or sequence
Analysis (CBS), Department of Systems Biology, Technical University of Denmark
(DTU), Lyngby, Denmark. 2Biotech Research & Innovation Centre (BRIC), conservation. This is at odds with recent findings that show that cancer
University of Copenhagen, Copenhagen, Denmark. Correspondence should be mutations tend not to cluster on the most conserved protein regions.
addressed to J.T.E. (janine.erler@bric.ku.dk) or R.L. (linding@cbs.dtu.dk). In kinases, for example, mutations typically hit the kinase activation
Published online 10 September 2012; doi:10.1038/nbt.2345 segment, a functional, yet largely nonconserved protein region13.

842 volume 30 number 9 SEPTEMBER 2012 nature biotechnology


perspecti v e

An insightful example of how to explore


b this sequence-function relationship in protein
domains was carried out by researchers in
the Ranganathan and Yaffe laboratories who,
using methods from statistical mechanics,
a Mutation in A =
healthy cell
Mutation in B =
healthy cell
generated synthetic WW domains de novo
that maintained fold and function17,18.
Synthetic
oncogenes Further supporting a complex sequence-
function relationship, additional studies from
Different mutations but the Ranganathan laboratory demonstrated
same phenotype
Mutation in A + Mutation in B = that, in addition to protein architecture
cancer cell c described as combinations of modules such
as globular domains and linear motifs19–21,
e protein domains themselves often have well-
Response A

Properties of cancer Activation of Activation of defined sectors formed by sparse networks of


signaling networks kinase 1 leads kinase 1 leads
Cue A to one outcome to another outcome residues often linking spatially distant regions
(anti-apoptotic) (pro-apoptotic)
that contribute cooperatively but unequally
to its function22,23. Although some targeted
Cue A studies analyzing several cancer mutations in
Baseline network a single kinase have been conducted24, similar
ate
B A d Network state A Network state B
approaches to those used for WW domains
ate
© 2012 Nature America, Inc. All rights reserved.

ll f ll f
Ce Ce should be pursued to generate high-throughput
experimental studies of cancer mutations in the
Dynamic context of signaling networks. These would
networks
help gain a better understanding of which
amino acid residues can be changed freely
without affecting the protein and network
function and, most importantly, which cannot.
Timepoint 1 Timepoint 2

From network perturbations to cellular


Figure 1 Properties of cancer signaling networks. (a) Analogous mutations. Two different tumors phenotypes
may achieve the same signaling and phenotypic outcome with two different mutations (b) Synthetic The characterization of cellular signaling pro-
oncogenes. Mutations that are not oncogenic on their own can cooperate when appearing together cesses has largely focused on identifying the
to drive tumor formation11; by analogy to synthetic lethality, we call the genes harboring cooperative function of individual genes and proteins. A
mutations, synthetic oncogenes. (c) Multivariate nature of signaling networks. The response of a cell to notable exception is a landmark study25 on the
a specific cue depends on, and can only be predicted by taking into account, the state of the cellular
context dependence of the Jun-activated kinase
signaling networks25. This dependency, known as the multivariate nature of signaling networks, is
often neglected when classifying mutations and genes as oncogenes or tumor suppressors and cancer
(JNK) in apoptosis. Before this work, para-
drivers or passengers. (d) Dynamic networks. Although signaling networks are often represented as doxical results suggested that JNK had a pro-
apoptotic function26, an anti-apoptotic func-
npg

static, it is clear that they are highly dynamic entities. Given that the role of signaling networks in
computing cellular responses is highly dependent on it, and that cancer mutations will perturb it, this tion27 or even a lack of involvement in apopto-
dynamic nature is a critical property of cancer signaling networks. (e) Signaling network landscapes. sis28. The systematic approach undertaken by
The different states that a signaling network occupies can be represented as a landscape (with stable Janes et al.25 revealed that the phosphorylation
steady states or attractors represented as valleys and unstable steady states represented as hills), where
the cell constantly gets pushed by signaling cues31,32,39,40. These states drive cellular and disease
status of JNK (and thus its catalytic activity)
phenotypes and represent network drug targets. was not sufficient to determine apoptotic com-
mitment; instead, activation of JNK could lead
to both apoptosis and proliferation depending
Because cancer cells would obtain the greatest fitness advantage on the cellular signaling network state at the time of activation. Thus, this
from mutations that target the most-functional residues, we reason work demonstrated that a protein’s cellular role is not a static property
that a better understanding of the functionality of protein residues but rather can only be defined dynamically—that is, its role depends
would allow more accurate predictions of the consequences of cancer on the context of the network it is operating within. Similar context
mutations. Functional residues have been defined as those residues dependencies have been confirmed for other kinases, such as Erk and
required for a protein to perform its molecular function(s), in the MK2. Because of this, which is referred to as the multivariate property
sense that they cannot be freely changed without directly affecting of signaling networks (Fig. 1c), we suggest that it is essential to study
the role(s) of the protein14. Here we extend this definition to include cellular context at the systems level.
a more fine-grained and precise definition of protein function as Although these multivariate molecular networks seem to have evolved
an ensemble of protein features that together describe the different a complex structure that makes them robust against deletion of a few
functional capabilities of proteins (e.g., ATP binding, substrate proteins29, they are highly dynamic. Thus, a more accurate description
specificity, protein activation or phospho-tyrosine binding). This new of signaling networks should take into account the fact that a single static
definition would not only adapt well to current studies of sequence- network does not exist unchanged over time. Instead, a cell contains
function associations15,16, but also lead to a better description of the a dynamic ensemble of networks whose different permutations are
effects of a mutation affecting such residues (Fig. 2a,b). manifested in the cell depending on the different cues the cell is presented

nature biotechnology volume 30 number 9 SEPTEMBER 2012 843


perspecti v e

with (Fig. 1d). This dynamic nature of signaling networks could, at least We postulate that network-attacking mutations affect the cell not by
in part, explain why all mutant proteins do not seem to be expressed at a perturbing how the signaling landscape is projected to the phenotypic
given point in time30, if a substantial part of the proteome is so dynamic dimension, but by changing the ensemble of dynamic networks that
that it is expressed only when the cell senses a specific cue. can be manifested in a cell and, in consequence, the number and
Moreover, according to a general principle of complex systems stability of steady states in the signaling landscape, thus creating new
introduced in the 1980s31,32, dynamic cellular networks can only exist attractor states that only cancer cells can occupy, also known as cancer
in a finite number of states, owing to the constraints that interactions network attractors (Fig. 3). This has additional implications for other
between nodes impose on one another. These network states can be mechanisms, such as oncogene and non-oncogene addiction41, where
represented as landscapes, where most-probable and least-probable cancer cells would be trapped in cancer attractor states and could
states are represented as valleys and mountains, respectively (Fig. 1e). escape from them by reverting the genomic aberration that initially
Cells are continuously exploring this landscape
and are pushed from one state to another by
different environmental or intracellular cues.
b
A
Implications for cancer research Feature 1 P P
B C
The multivariate nature of signaling networks has
profound implications for cancer research. Just as a Disease genome
DP P P
E

Feature 2 Disease networks


it is inaccurate to assign a static function (e.g.,
apoptotic or anti-apoptotic) to a single protein,
it is clear that static interpretations of mutations, Feature 3
that is, driver or passenger mutations, are also
© 2012 Nature America, Inc. All rights reserved.

misleading. For example, given that the pheno- Protein sequence Disease proteome
typic role of JNK strongly depends on network c
state, it is clear that a mutation in JNK (and thus e
probably any other mutation) should not be

Response B
Challenges in cancer
statically labeled as a driver or passenger or as network biology Cue A
an oncogene or tumor suppressor, as such clas-
sifications are context dependent (e.g., disease
or cell-type specific). Several examples, such Drug A Drug B Cue A
as Myc33 or WT1 (ref. 34) gene products that
act as both tumor suppressors and oncogenes,
support this idea. These results underscore the
Rewiring dTumor subpopulations Ce
ll fat
e
B

Ce
ll fat
eA

Immune system
importance of assessing mutations based on their
effects on signaling networks and of developing
novel classification methods to do so. Along these
lines, MAP2K4 (one of the protein kinases that
can phosphorylate and activate JNK) has been
shown to be recurrently lost or mutated in sev- Surrounding Matrix
tissue
eral cancers35–38. These represent prime examples
npg

of mutations that may display ambivalent pheno-


typic impact similar to JNK.
Motivated by the example of MAP2K4 and Figure 2 Challenges in cancer network biology. (a) Functional consequences of cancer mutations.
many other mutated kinases38, we maintain Using an ensemble of protein-function features (e.g., ATP binding, substrate specificity, activation
that mutations capable of affecting signaling of the protein kinase or phospho-tyrosine binding), which together represent a comprehensive
description of a protein’s molecular functions, will enable more accurate and predictive evaluation
networks—which we call network-attacking
of cancer mutations. (b) Modeling of disease networks. Although experimental and computational
mutations (Fig. 2c)—are more likely to affect tools for modeling molecular networks exist, creating more comprehensive, sensitive and
phenotype than other mutations. Thus, we discuss accurate new tools especially designed to model disease-associated networks still represents a
a general strategy in which mutations in individual big challenge in network biology. (c) Network-attacking mutations and cancer network attractors.
cancers are assessed based on, first, the likelihood Network-attacking mutations are mutations that lead to a new cellular phenotype by perturbing
they will affect protein function, and second, signaling networks either at the network structure or the network dynamics level. Network-
the cellular role of the signaling network that attacking mutations transform signaling networks, generating new possible network states by
changing the number and/or stability of steady states in the signaling landscape 31,32,39,40. These
they are operating within (Fig. 3). Our strategy acquired signaling capabilities lead to alterations in the cell’s normal ‘cue-signal-output’ flow
extends the concepts introduced by Waddington and thereby drive disease phenotypes (see Fig. 3 for further details). (d) Tumor subpopulations
and elaborated by Kauffman and Huang et and micro-environment. The field is only beginning to comprehend the complex interactions that
al.31,32,39,40, where cancer mutations are turned exist between different co-evolving tumor cell subpopulations and between those cells and the
into perturbations capable of reshaping these tumor microenvironment, both of which strongly influence tumor progression. (e) Network-aware
landscapes. We represent the cellular response and temporal drugs. As predicted by R.L. and Pawson66 several years ago, new pharmaceutical
strategies that target networks instead of single proteins are becoming available 47,48. We predict
or phenotype as another dimension where each
this trend will not only continue, but also include recent advances that highlight the possibility to
network state (every point in the landscape) is ‘cure’ networks using time- and order-dependent therapies68. In coming years, the discovery of
constantly projected to and translated into a resistant, metastatic, tissue or cell-specific networks could lead to an even greater advance in the
cellular decision or phenotypic outcome. field of network medicine (Fig. 5).

844 volume 30 number 9 SEPTEMBER 2012 nature biotechnology


perspecti v e

Wild type Network-attacking mutation Network-attacking mutation


caused the perturbed landscape. Given affecting network dynamics affecting network structure
the high degree of determinism that exists
No mutation Mutation A Mutation B
between signaling networks, landscapes and
phenotypes, we argue that network-attacking Genome
mutations are at the heart of all new decision-
making capabilities acquired by cancer cells. Kinase A A A
Consequently, in our view, the study of No
P
cue Substrates B C B C B C
both network-attacking mutations and new Signaling
attractor states acquired by cancer cells, that network
A A A
is, cancer network attractors, deserves the Cue P P
B C B C BP C
highest priority from the field. Such studies
should be performed through systematic and
Cue
quantitative sampling of cell dynamics at Signaling Cue
multiple levels (e.g., genomic or epigenetic, landscape Cue Cancer
Cancer attractor
proteomic and phenotypic), followed by attractor

B
A

A
nonlinear interpolation and integrative Phenotypic

te

te

te

te

te

te
fa

fa

fa

fa

fa

fa
ll

ll

ll

ll

ll

ll
outcome

Ce

Ce

Ce

Ce

Ce

Ce
computational modeling (Fig. 4).
The first network-attacking cancer mutation,
Figure 3 Network-attacking cancer mutations. Proteins are the key elements of signaling networks as a
described more than 15 years ago42, was a point result of their ability to integrate external cues and direct the information flow toward a specific cellular
mutation in the kinase domain of RET (M918T), outcome (e.g., epidermal growth factor (EGF) leading to proliferation or tumor necrosis factor alpha
which leads to a switch in peptide specificity. In (TNF-a) leading to apoptosis). Network-attacking mutations affect the ‘cue-signal-output’ cellular
© 2012 Nature America, Inc. All rights reserved.

line with their importance, network-attacking information flow by affecting either the dynamics (middle), for example, by keeping proteins
mutations have attracted more attention in constitutively active, or the structure (right), by affecting protein specificity, of the signaling networks.
recent years43–48. Moreover, information has Signaling networks can be represented as a landscape with the most likely network states represented
as valleys (stable steady states or attractors) and the least likely network states as mountains (unstable
been accumulating steadily about how specific- steady states). Network-attacking mutations dysregulate signaling networks by perturbing the number
ity in signaling networks and modular protein and/or stability of steady states in the landscape, effectively creating new cancer-specific attractors that
domains emerges49–51, leading to the defini- only cancer cells will be able to reach.
tion of determinants of specificity in protein
domains52,53. These determinants, sometimes referred to as specificity- Despite the fact that the number of known cancer network-attacking
determining residues, are residues that can lead to substrate specificity mutations is still relatively low, recent findings suggest that in-frame
changes after mutation. Notably, direct mutagenesis of these determinants mutations are enriched on interaction interfaces57, which implies
of specificity has been used to rewire the entire histidine kinase signal- they are also likely to affect determinants of specificity. Moreover,
ing system in bacteria in a predictive manner54. Recent follow-up work many fusion proteins have been discovered that likely directly rewire
indicates that mutations in determinants of specificity prevent cross-talk or create new network states58. Given the rate at which cancer muta-
and allow protein family expansions55, in a process similar to the one tions are being reported and the development of new computational
powered by negative selection over Src homology 3 (SH3) protein domains methods for systematically identifying these mutations (Fig. 2b),
that show similar specificity56. We propose that similar studies in human we predict a steep increase in the number of network-attacking muta-
signaling networks, coupled with mapping of cancer mutations on these tions that will be uncovered in the coming years.
determinants of specificity, would shed new light on whether signaling
npg

rewiring is a general principle of oncogenesis and tumor progression, Personalized cancer network biology
knowledge of which would in turn be critical as molecular therapies tar- Led by recent advances in sequencing technologies, the amount of data
get proteins and their networks and not genes. on cancer genome mutations is growing exponentially59. Current efforts

Sampling strategy - Interpolation -


Figure 4 Traditional versus network biology Comprehensive Linear
approaches. In more traditional biological
approaches, where only one or a few genes
or proteins are sampled across a limited
[Stimuli y]

[Stimuli y]

set of conditions, there has been limited Traditional biology


approach
success in deriving predictive models across Initial data set
conditions or cell types that would require
High
comprehensive sampling. In contrast, apoptosis [Stimuli x] [Stimuli x]
network biology relies on systematic
[Stimuli y]

sampling across combinations of states


that result in increased performance of a
network model. Unlike classic approaches, Sampling strategy - Interpolation -
in which the system is stimulated with Low Model-driven Nonlinear
single specific cellular cues (e.g., growth [Stimuli x] apoptosis
factor), in the network biology approach, the
Network
[Stimuli y]

[Stimuli y]

multivariate nature of signaling networks and


Network biology model
the nonlinear relationship between signaling approach
input and output can be successfully
elucidated by interrogating the system with
[Stimuli x] [Stimuli x]
multiple orthogonal cues.

nature biotechnology volume 30 number 9 SEPTEMBER 2012 845


perspecti v e

from the Cancer Genome Atlas and Cancer


A
Genome Project, now under the umbrella of the P
P
International Cancer Genome Consortium60, Single-cell B C
will facilitate the annotation and collection of DNA sequencing
DP P E
P
cancer genome data. We foresee similar waves
Interpretation
of technological progress and the generation
Best combination
of new consortiums in the cancer proteomics CyToF/mass spectrometry
fields in the near future. The establishment
of the Clinical Proteomic Tumor Analysis
Consortium (http://proteomics.cancer.gov/pro- Cell culture
grams/cptacnetwork), and the implementation
Network drugs Resistance
of new approaches61 and labeling techniques62 follow-up
Early diagnostic Phenotyping
optimized for patient samples are encouraging
advances in this direction.
These advances, however, will need to
be coordinated with new algorithmic and
experimental high-throughput methods (e.g., Xenograft model
high-content screening) capable of interpreting
this flood of information because the functional Figure 5 Personalized cancer network biology. The goal of personalized cancer network biology is to
interpretation of the data is currently the main be able to treat each tumor with the best combination of drugs tailored to that tumor. Ideally, early
bottleneck in the field of personalized cancer diagnosis should be followed by the development of tumor-specific cell lines and xenograft models,
© 2012 Nature America, Inc. All rights reserved.

cancer genome sequencing, and proteomic and phenotypic analysis. Combinations of network drugs
network biology. Computational integration should then be tried in the tumor-specific cell line and xenograft model and eventually transferred
of large quantitative data sets is also becoming back to the patient. Continuing to treat the tumor-specific cell culture with the same network drug
increasingly important, and thus there is a combination as is used in the patient may be useful for understanding potential resistance and/or
growing requirement for supercomputing metastasis.
infrastructure with large algorithmic dynamic
range (e.g., next-generation large shared memory systems). Benchmarking high-throughput profiling of phenotypic cell states in the tumor and
and validation of systematic workflows and algorithms is already receiving design of patient-specific combinations of network drugs with resistance
increasing attention through initiatives, such as the DREAM challenge63 follow-up (Fig. 5). Relatively new techniques, such as single-cell and high-
and IMPROVER64. depth sequencing70,71, imaging72 and cytometry time-of-flight73, could
Two emerging areas in network biology that are likely to contribute to prove especially valuable for monitoring the number, properties and
the future of cancer research are the study of cell-cell interactions (Fig. 2d) behavior of different tumor subclones (Fig. 2d). Ideally, network drugs,
and drugs specifically designed to interfere with diseased network such as the aforementioned order- and time-dependent combination68,
dynamics (that is, network drugs; Fig. 2e). should then be chosen based on the interpretation of sequencing as well
R.L. and collaborators65 studied cell-cell interactions by isotopically as the proteomic and phenotypic analysis of tumor cells and tested on
labeling two distinct subpopulations of cells, one expressing ephrin-B1+ the tumor-specific cell lines and xenograft model. The best-performing
and the other Eph-B2+, and carrying out a comprehensive phospho- combination should ultimately be transferred back to the patient (Fig. 5).
proteomic analysis. This strategy facilitated the first measurements This whole process should take the shortest time possible to avoid
of phosphorylation events during the interaction of two cell the evolution of the tumor in the patient and the consequent loss of
npg

subpopulations. The proliferative behavior of cancer cells is still poorly relationship between the primary tumor and the cell line. Tumor-
understood in part because it is difficult to experimentally study the specific cell lines would be kept and treated with the same drugs used
transmission of proliferative factors from one cell to its neighbors3. in the patient to monitor tumor evolution and treat for resistance
Therefore, we argue that a similar isotopic labeling strategy could and/or metastasis as soon as there is enough evidence of it (Fig. 5).
be used to investigate the cooperation between cells with different Ideally, every patient and paired xenograft or cell line should have a
oncogenic lesions that together (that is, synthetic oncogenes; Figs. 1b complete electronic record showing the treatment history to facilitate
and 2d) lead to tumor formation11. retrospective and cross-disease studies74,75.
Combination drugs that interfere with disease networks (so-called
network medicine66) have been shown to lead to a better response than Conclusions
single-hit therapies by causing secondary perturbations to signaling Although we have highlighted some of the challenges that still exist in
networks47,48,67. Recent work by the Yaffe laboratory represents a clear cancer network biology, substantial progress is also being made. For
leap forward within the field of network medicine68,69. Following network example, the usage of patient-derived tumor tissue in animal xenograft
modeling, Yaffe and colleagues68 managed to decode the signaling models to test the response to particular drugs aimed at developing
network dynamics that drive resistance to DNA-damaging chemotherapy. new personalized cancer therapy is rapidly becoming an established
This information was used to sensitize otherwise resistant triple-negative technology76. Surgical orthotopic implantation to transplant tumors
breast cancer cells to conventional DNA-damaging chemotherapy by taken directly from the patient to the corresponding organ of immu-
administering doxorubicin (Adriamycin, Doxil) and erlotinib (Tarceva) nodeficient mice77 is currently one of the most promising methods to
in an order- and time-dependent fashion. This could be considered the enable drug screening in patients. In addition, new clinical trials, such
first example of temporal network drugs (Figs. 2e and 5). as the MD Anderson T9 project78, are under way in which patients are
We predict that personalized or even tumor-specific cancer therapy will given therapy that targets tumor-specific aberrations. Nevertheless,
become a reality in the foreseeable future, starting from early diagnosis of the implementation of the strategy depicted in Figure 5 would benefit
the disease, followed by next-generation sequencing, proteomic analysis, from further developments in technology, funding and legislation. For

846 volume 30 number 9 SEPTEMBER 2012 nature biotechnology


perspecti v e

example, generating models for cancer research that represent human 21. Seet, B.T., Dikic, I., Zhou, M.M. & Pawson, T. Reading protein modifications with
interaction domains. Nat. Rev. Mol. Cell Biol. 7, 473–483 (2006).
patient diversity79 and mimicking the complexity of tumor microenvi- 22. Halabi, N., Rivoire, O., Leibler, S. & Ranganathan, R. Protein sectors: Evolutionary
ronments (J.T.E. and collaborators)80 remain extraordinary challenges units of three-dimensional structure. Cell 138, 774–786 (2009).
(Fig. 2), and further research efforts and investments are required. 23. Reynolds, K.A., McLaughlin, R. & Ranganathan, R. Hot spots for allosteric regulation
on protein surfaces. Cell 147, 1564–1575 (2011).
As cancer biology becomes a ‘big data’ science, similar to physics, 24. Wan, P.T. et al. Mechanism of activation of the RAF-ERK signaling pathway by
we expect to see more systematic, data-driven research efforts that oncogenic mutations of B-RAF. Cell 116, 855–867 (2004).
will uncover and confront many of the tumor complexities that have 25. Janes, K.A. et al. A systems model of signaling identifies a molecular basis set for
cytokine-induced apoptosis. Science 310, 1646–1653 (2005).
remained elusive so far. 26. Lei, K. & Davis, R.J. JNK phosphorylation of Bim-related members of the Bcl2 family
Despite recent predictions of >13 million cancer deaths in 2030 induces Bax-dependent apoptosis. Proc. Natl. Acad. Sci. USA 100, 2432–2437 (2003).
27. Lamb, J.A. et al. JunD mediates survival signaling by the JNK signal transduction
(ref. 81), as discussed in this Perspective, we foresee that within this pathway. Mol. Cell 11, 1479–1489 (2003).
timeframe tumor-specific medicine will become a reality, thanks to 28. Abreu-Martin, M.T. et al. Fas activates the JNK pathway in human colonic epithelial
a new generation of cancer network biologists who will hopefully cells: lack of a direct role in apoptosis. Am. J. Physiol. 276, G599 (1999).
29. Jeong, H., Mason, S.P., Barabasi, A.L. & Oltvai, Z.N. Lethality and centrality in protein
overcome these challenges, positively contributing to the battle networks. Nature 411, 41–42 (2001).
against this devastating disease and the significant reduction of patient 30. Shah, S.P. et al. The clonal and mutational evolution spectrum of primary triple-
suffering. negative breast cancers. Nature advance online publication, doi:10.1038/nature10933
(4 April 2012).
31. Kauffman, S. & Levin, S. Towards a general theory of adaptive walks on rugged
ACKNOWLEDGMENTS
landscapes. J. Theor. Biol. 128, 11–45 (1987).
We apologize to our colleagues whose work could not be cited due to space 32. Kauffman, S.A. & Weinberger, E.D. The NK model of rugged fitness landscapes and its
limitations. We thank all members of the C-SIG (DTU), the ErlerLab (BRIC), application to maturation of the immune response. J. Theor. Biol. 141, 211–245 (1989).
M. Yaffe (MIT) and N. Brunner (KU) for critical input on this manuscript. R.L. is 33. Uribesalgo, I., Benitah, S.A. & Di Croce, L. From oncogene to tumor suppressor: The
a Lundbeck Foundation Fellow and is supported by a Sapere Aude Starting Grant dual role of Myc in leukemia. Cell Cycle 11, 1757–1764 (2012).
from The Danish Council for Independent Research and a Career Development 34. Yang, L., Han, Y., Saurez Saiz, F. & Minden, M.D. A tumor suppressor and oncogene:
© 2012 Nature America, Inc. All rights reserved.

Award from Human Frontier Science Program. J.T.E. is supported by a Hallas the WT1 story. Leukemia 21, 868–876 (2007).
Møller Stipend from the Novo Nordisk Foundation. Visit http://www.networkbio. 35. Ellis, M.J. et al. Whole-genome analysis informs breast cancer response to aromatase
inhibition. Nature 486, 353–360 (2012).
org/, http://www.lindinglab.org/ and http://www.erlerlab.org/ for more information
36. Curtis, C. et al. The genomic and transcriptomic architecture of 2000 breast tumours
on cancer-related network biology. reveals novel subgroups. Nature 486, 346–352 (2012).
37. Kan, Z. et al. Diverse somatic mutation patterns and pathway alterations in human
AUTHOR CONTRIBUTIONS cancers. Nature 466, 869–873 (2010).
All correspondence should be addressed to both J.T.E. and R.L. 38. Greenman, C. et al. Pattern of somatic mutation in human cancer genomes. Nature
446, 153–158 (2007).
Competing FINANCIAL interests 39. Waddington, C.H. The Strategy of the Genes: a Discussion of Some Aspects of
The authors declare no competing financial interests. Theoretical Biology (Allen & Unwin, 1957).
40. Huang, S. & Ingber, D.E. Shape-dependent control of cell growth, differentiation, and
Published online at http://www.nature.com/doifinder/10.1038/nbt.2345. apoptosis: switching between attractors in cell regulatory networks. Exp. Cell Res.
Reprints and permissions information is available online at http://www.nature. 261, 91–103 (2000).
com/reprints/index.html. 41. Luo, J., Solimini, N.L. & Elledge, S.J. Principles of cancer therapy: oncogene and
non-oncogene addiction. Cell 136, 823–837 (2009).
1. Nash, J.F. Equilibrium points in N-person games. Proc. Natl. Acad. Sci. USA 36, 42. Songyang, Z. et al. Catalytic specificity of protein-tyrosine kinases is critical for
48–49 (1950). selective signaling. Nature 373, 536–539 (1995).
2. Nash, J.F. Non-cooperative games. Ann. Math. 54, 286–295 (1951). 43. Zhong, Q. et al. Edgetic perturbation models of human inherited disorders. Mol. Syst.
3. Hanahan, D. & Weinberg, R.A. Hallmarks of cancer: the next generation. Cell 144, Biol. 5, 321 (2009).
646–674 (2011). 44. Dreze, M. et al. ‘Edgetic’ perturbation of a C. elegans BCL2 ortholog. Nat. Methods
4. Hanahan, D. & Weinberg, R.A. The hallmarks of cancer. Cell 100, 57–70 (2000). 6, 843–849 (2009).
5. Davies, H. et al. Mutations of the BRAF gene in human cancer. Nature 417, 949–954 45. Pe’er, D. & Hacohen, N. Principles and strategies for developing network models in
(2002). cancer. Cell 144, 864–873 (2011).
6. Stratton, M.R., Campbell, P.J. & Futreal, P.A. The cancer genome. Nature 458, 719– 46. Vidal, M., Cusick, M.E. & Barabási, A.-L.L. Interactome networks and human disease.
npg

724 (2009). Cell 144, 986–998 (2011).


7. Ding, L. et al. Clonal evolution in relapsed acute myeloid leukaemia revealed by whole- 47. Schoeberl, B. et al. Therapeutically targeting ErbB3: a key node in ligand-induced
genome sequencing. Nature 481, 506–510 (2012). activation of the ErbB receptor-PI3K axis. Sci. Signal. 2, ra31 (2009).
8. Gerlinger, M. et al. Intratumor heterogeneity and branched evolution revealed by 48. Huang, P.H. et al. Quantitative analysis of EGFRvIII cellular signaling networks reveals
multiregion sequencing. N. Engl. J. Med. 366, 883–892 (2012). a combinatorial therapeutic strategy for glioblastoma. Proc. Natl. Acad. Sci. USA 104,
9. Hou, Y. et al. Single-cell exome sequencing and monoclonal evolution of a JAK2- 12867–12872 (2007).
negative myeloproliferative neoplasm. Cell 148, 873–885 (2012). 49. Miller, M.L.L. et al. Linear motif atlas for phosphorylation-dependent signaling. Sci.
10. Cancer Genome Atlas Research Network. Integrated genomic analyses of ovarian Signal. 1, ra2+ (2008).
carcinoma. Nature 474, 609–615 (2011). 50. Linding, R. et al. Systematic discovery of in vivo phosphorylation networks. Cell 129,
11. Wu, M., Pastor-Pareja, J.C. & Xu, T. Interaction between RasV12 and scribbled clones 1415–1426 (2007).
induces tumour growth and invasion. Nature 463, 545–548 (2010). 51. Mok, J. et al. Deciphering protein kinase specificity through large-scale analysis of yeast
12. Ng, P.C. & Henikoff, S. Predicting the effects of amino acid substitutions on protein phosphorylation site motifs. Sci. Signal. 3, ra12 (2010).
function. Annu. Rev. Genomics Hum. Genet. 7, 61–80 (2006). 52. Brinkworth, R.I., Breinl, R.A. & Kobe, B. Structural basis and prediction of substrate
13. Dixit, A. et al. Sequence and structure signatures of cancer mutation hotspots in protein specificity in protein serine/threonine kinases. Proc. Natl. Acad. Sci. USA 100, 74–79
kinases. PLoS ONE 4, e7485 (2009). (2003).
14. Pazos, F. & Bang, J.-W. Computational prediction of functionally important regions in 53. Turk, B.E. Understanding and exploiting substrate recognition by protein kinases. Curr.
proteins. Curr. Bioinform. 1, 15–23 (2006). Opin. Chem. Biol. 12, 4–10 (2008).
15. Fowler, D.M. et al. High-resolution mapping of protein sequence-function relationships. 54. Skerker, J.M. et al. Rewiring the specificity of two-component signal transduction
Nat. Methods 7, 741–746 (2010). systems. Cell 133, 1043–1054 (2008).
16. Jensen, L.J. et al. Ab initio prediction of human orphan protein function from post- 55. Capra, E.J., Perchuk, B.S., Skerker, J.M. & Laub, M.T. Adaptive mutations that prevent
translational modifications and localization features. J. Mol. Biol. 319, 1257–1265 crosstalk enable the expansion of paralogous signaling protein families. Cell 150,
(2002). 222–232 (2012).
17. Socolich, M. et al. Evolutionary information for specifying a protein fold. Nature 437, 56. Zarrinpar, A., Park, S.H. & Lim, W.A. Optimization of specificity in a cellular protein
512–518 (2005). interaction network by negative selection. Nature 426, 676–680 (2003).
18. Russ, W., Lowery, D., Mishra, P., Yaffe, M. & Ranganathan, R. Natural-like function in 57. Wang, X. et al. Three-dimensional reconstruction of protein networks provides insight
artificial WW domains. Nature 437, 579–583 (2005). into human genetic disease. Nat. Biotechnol. 30, 159–164 (2012).
19. Puntervoll, P. et al. ELM server: a new resource for investigating short functional sites 58. Brehme, M. et al. Charting the molecular network of the drug target Bcr-Abl. Proc. Natl.
in modular eukaryotic proteins. Nucleic Acids Res. 31, 3625–3630 (2003). Acad. Sci. USA 106, 7414–7419 (2009).
20. Lim, W.A. & Pawson, T. Phosphotyrosine signaling: evolving a new cellular 59. Wong, K.M.M., Hudson, T.J. & McPherson, J.D. Unraveling the genetics of cancer: genome
communication system. Cell 142, 661–667 (2010). sequencing and beyond. Annu. Rev. Genomics Hum. Genet. 12, 407–430 (2011).

nature biotechnology volume 30 number 9 SEPTEMBER 2012 847


perspecti v e

60. Ledford, H. Big science: the cancer genome challenge. Nature 464, 972–974 72. Pedersen, M.W. et al. Sym004: a novel synergistic anti-epidermal growth factor receptor
(2010). antibody mixture with superior anticancer efficacy. Cancer Res. 70, 588–597 (2010).
61. Bensimon, A., Heck, A.J.R. & Aebersold, R. Mass spectrometry-based proteomics and 73. Bendall, S.C. & Nolan, G.P. From single cells to deep phenotypes in cancer. Nat.
network biology. Annu. Rev. Biochem. 81, 379–405 (2012). Biotechnol. 30, 639–647 (2012).
62. Geiger, T., Cox, J., Ostasiewicz, P., Wisniewski, J.R. & Mann, M. Super-SILAC mix for 74. Roque, F.S. et al. Using electronic patient records to discover disease correlations
quantitative proteomics of human tumor tissue. Nat. Methods 7, 383–385 (2010). and stratify patient cohorts. PLOS Comput. Biol. 7, e1002141 (2011).
63. Prill, R.J. et al. Towards a rigorous assessment of systems biology models: the DREAM3 75. Jensen, P.B., Jensen, L.J. & Brunak, S. Mining electronic health records: towards
challenges. PLoS ONE 5, e9202 (2010). better research applications and clinical care. Nat. Rev. Genet. 13, 395–405
64. Meyer, P. et al. Verification of systems biology research in the age of collaborative (2012).
competition. Nat. Biotechnol. 29, 811–815 (2011). 76. Blumenthal, R.D. & Goldenberg, D.M. Methods and goals for the use of in vitro and
65. Jørgensen, C. et al. Cell-specific information processing in segregating populations of in vivo chemosensitivity testing. Mol. Biotechnol. 35, 185–197 (2007).
Eph receptor ephrin-expressing cells. Science 326, 1502–1509 (2009). 77. Hoffman, R.M. Orthotopic mouse models expressing fluorescent proteins for cancer
66. Pawson, T. & Linding, R. Network medicine. FEBS Lett. 582, 1266–1270 (2008). drug discovery. Expert Opin. Drug Discov. 5, 851–866 (2010).
67. Chandarlapaty, S. et al. AKT inhibition relieves feedback suppression of receptor tyrosine 78. Gonzalez-Angulo, A.M., Hennessy, B.T. & Mills, G.B. Future of personalized medicine
kinase expression and activity. Cancer Cell 19, 58–71 (2011). in oncology: a systems biology approach. J. Clin. Oncol. 28, 2777–2783 (2010).
68. Lee, M.J. et al. Sequential application of anticancer drugs enhances cell death by rewiring 79. Hunter, K.W. Mouse models of cancer: does the strain matter? Nat. Rev. Cancer 12,
apoptotic signaling networks. Cell 149, 780–794 (2012). 144–149 (2012).
69. Erler, J.T. & Linding, R. Network medicine strikes a blow against breast cancer. Cell 149, 80. Cox, T.R. & Erler, J.T. Remodeling and homeostasis of the extracellular matrix:
731–733 (2012). implications for fibrotic diseases and cancer. Dis. Model. Mech. 4, 165–178
70. Navin, N. et al. Tumor evolution inferred by single-cell sequencing. Nature 472, 90–94 (2011).
(2011). 81. WHO. World health organization fact sheet 297 (2012). http://www.who.int/
71. Nik-Zainal, S. et al. The life history of 21 breast cancers. Cell 149, 994–1007 (2012). mediacentre/factsheets/fs297/en/
© 2012 Nature America, Inc. All rights reserved.
npg

848 volume 30 number 9 SEPTEMBER 2012 nature biotechnology


perspective

Lessons from human teratomas to guide development


of safe stem cell therapies
Justine J Cunningham1, Thomas M Ulbright2, Martin F Pera3 & Leendert H J Looijenga4
The potential for the formation of teratomas or other neoplasms practical terms it may be difficult to guarantee the complete absence of
is a major safety roadblock to clinical application of pluripotent undifferentiated cells in production at this scale. Third, although undif-
stem cell therapies. Preclinical assessment of the risk of tumor ferentiated pluripotent cells carry the greatest concern, it is uncertain
formation in this context poses considerable scientific and what risks are posed by other (progenitor) populations. Fourth, unlike
regulatory challenges, especially because animal xenograft drugs, which have half-lives on the order of hours, cellular therapeutics
models may not properly reflect the long-term tumorigenic may be designed to persist for a large fraction of the patient’s life span,
© 2012 Nature America, Inc. All rights reserved.

potential of human cells. A better understanding of the biology of and current technology for monitoring the long-term fate of grafted
spontaneously occurring teratomas and related tumors in humans cells in animals or humans has serious limitations.
can help to guide efforts to assess and minimize the potential Lessons learned from the biology of spontaneous human teratomas
hazards of embryonic stem cell or induced pluripotent stem cell offer a unique opportunity to gain insight into potential complications
therapeutics. Here we review the features of teratomas derived of hES cell therapies and to answer key questions for the stem cell
experimentally from human pluripotent stem cells and argue field. In this Perspective, we survey the features of teratomas formed
that they most closely resemble spontaneous benign teratomas by pluripotent stem cell grafts in animals and compare these experi-
that occur early in both mouse and human life. The natural mental tumors to the various classes of teratomas seen clinically. We
history and pathology of these spontaneously occurring teratomas highlight those features of spontaneously occurring teratomas that
provide important clues for preclinical safety assessment and can help guide efforts to identify, understand and mitigate risks of
patient monitoring in trials of stem cell therapies. pluripotent stem cell therapies.

The risk that therapeutics derived from human embryonic stem Study of tumors from human pluripotent stem cells
(hES) cells or induced pluripotent stem (hiPS) cells will give rise The assessment of the tumorigenic potential of human pluripotent
to neoplasms, particularly teratomas (Box 1) and other embryonal stem cells in animal models is carried out for two different objectives.
tumors, is a central safety hurdle to the application of these tech- The most common goal is to assess the developmental potential of
nologies in regenerative medicine. Eliminating this risk requires a particular cell line, specifically, to determine whether the isolate is
npg

extinguishing the intrinsic features of hES and hiPS cells— pluripotent. Indeed the formation of teratomas in animals has become
pluripotency and indefinite self-renewal—before transplantation so the gold standard for hES cell pluripotency. However, this relates only
the cells undergo differentiation in a uniform, predictable manner. to the cells’ capacity to generate somatic lineages. A smaller number
With current awareness that many cancers may originate from a of studies have examined teratoma formation from the standpoint of
stem cell component, there is concern that hES cell–derived trans- preclinical safety assessment1,2.
plants might be contaminated with residual undifferentiated (that is, The frequency of teratoma formation by human pluripotent stem
pluripotent) cells capable of forming benign neoplasms or even of cell grafts can be influenced by a number of factors, including the site
giving rise to aggressive invasive or ‘malignant’ teratomas. of injection3–6, the cell inoculum6,7, the addition of excipients such as
Preclinical safety assessment of the potential for tumor forma- Matrigel5,6 and inoculation of other cell populations into the graft8.
tion in stem cell therapy poses challenges for several reasons. First, However, few of these variables have been convincingly shown to
animal xenograft models may not accurately predict the fate of grafted influence the histological makeup of teratomas, with the interesting
cells in humans. Second, for some applications it may be necessary to exception of co-inoculation of human fetal cells (see below). Studies of
repeatedly treat patients with doses containing billions of cells, and in human cancer stem cell transplants in mice have demonstrated clearly
that the degree to which the host animal is immunocompromised
1Department of Drug Safety Evaluation, Allergan Inc., Irvine, California, USA. can profoundly influence the relationship between the size of the cell
2Department of Pathology & Laboratory Medicine, Indiana University School of inoculum and the probability of tumor formation9,10. Although the
Medicine and Indiana University Health Partners, Indianapolis, Indiana, USA. influence of host immune phenotype on tumor formation has not
3The University of Melbourne, Walter and Eliza Hall Institute of Medical Research,

Florey Neurosciences Institutes, Melbourne, Victoria, Australia. 4Department


been systematically investigated for human pluripotent stem cells, it
of Pathology, Erasmus MC-University Medical Centre Rotterdam (Daniel den is likely important for these cells as well.
Hoed Cancer Centre), Josephine Nefkens Institute, Rotterdam, The Netherlands. In general, when ~1 × 106 undifferentiated diploid human pluripotent
Correspondence should be addressed to J.J.C. (cunningham_justine@allergan.com). stem cells are inoculated into immunocompromised mice, tumors arise
Published online 10 September 2012; doi:10.1038/nbt.2329 in a high proportion of recipients within 4 to 6 weeks3–6. The histogenesis

nature biotechnology  VOLUME 30  NUMBER 9  SEPTEMBER 2012 849


Perspective

Box 1  Glossary of terms


Carcinoma. A malignant tumor derived from epithelial tissues such as the skin or epithelial components of other organs, such as the gut
and lungs.
Choriocarcinoma. A tumor comprising malignant trophoblastic tissues; may be seen in aberrant gestations (e.g., after hydatidiform mole)
or as a form of type II GCT.
Dysgerminoma. A primitive malignant tumor resembling primordial germ cells that occurs in the ovary; seminoma is the counterpart in
the testis and germinoma in extragonadal sites.
Dysplasia. Literally, bad or abnormal development. In pathology, a term most often used to describe early cellular abnormalities (nuclear
enlargement, hyperchromasia, cellular disorganization) that may progress to a fully malignant, invasive growth. Usually describes a
premalignant cellular condition.
Embryonal carcinoma. A tumor resembling the undifferentiated, highly proliferative, primitive epithelial tissues of the embryo with very
limited differentiation toward either embryonic or extra-embryonic lineages. The cells are characterized by expression of embryonic genes
OCT3/4, NANOG and SOX2.
Extragonadal. Any region of the body that is outside of the gonads, for example, head, chest, spine.
Germinoma. See Dysgerminoma.
Neoplasia. Literally, new development or growth. In pathology, a term used to describe a clonal proliferation of cells that most often
results in the formation of a mass (tumor). Resulting neoplasms may be either benign (nonspreading) or malignant (capable of metasta-
sizing to other sites).
© 2012 Nature America, Inc. All rights reserved.

Nonseminoma. A class of (testicular) GCTs that includes all types other than seminoma, that is, embryonal carcinoma, teratoma,
choriocarcinoma and yolk sac tumor (type II GCT). The distinction between seminoma and nonseminoma reflects clinical differences,
including age of onset, survival and response to treatment. The majority of nonseminomas are malignant.
-oma. Greek, group or mass. In medicine, refers to a neoplasm or tumor.
Sacrococcygeal. The ′tail end′ of the spine that includes the bones of the posterior pelvis, the sacrum and coccyx.
Seminoma. See Dysgerminoma.
Terato-. Greek, wonder, expression of divine power.
Teratos- or Teras-. Greek, deformity or monster.
Teratoma. A monstrous mass/tumor consisting of tissues from all three embryonic germ layers (ectoderm, mesoderm and endoderm), but
pathologists also recognize ‘bidermal’ and ‘monodermal’ teratomas. Simply, they are lesions foreign to their anatomical sites, such as
growth of hair and sebaceous glands within the lungs. These tumors are named for appearance rather than behavior. Most are benign,
whereas those that are malignant have defined histological characteristics.
Teratocarcinoma. An old term for a tumor with elements of teratoma and EC-like cells, the latter representing a malignant, pluripotential
cell population. More accurately, this class of tumor is termed a nonseminoma, which is by definition malignant.
Yolk sac tumor (endodermal sinus tumor). A neoplasm of germ or stem cell origin that resembles yolk sac elements of either extra-
embryonic or embryonic (gut epithelium) type.
npg

of these tumors is poorly understood, but there is evidence that the cells formed by mouse ES14 and iPS cells15, which contain undifferenti-
recapitulate post-implantation embryonic development at an early stage ated tumor stem cells capable of metastasis.
after engraftment11. The neoplasms that ultimately develop typically con- One report examined the effect of transplanting human fetal
tain a range of somatic tissues displaying varying degrees of maturation. tissue in conjunction with hES cells8. Strikingly, the tumors from
Some neoplasms show substantial organotypic differentiation, with vari- these co-transplanted cells consisted not of teratomas but of primitive
ous cell types organized into structures resembling fetal or adult tissues,
but often they contain more primitive structures representing early stages
of development, such as neural rosettes (Fig. 1). Commonly there are a b c
abundant cystic lesions of uncertain tissue origin. Although spontane-
ously arising neoplasms that consist largely of teratoma may also contain
elements representative of fetal membranes, such as yolk sac or placenta,
only one study has described these tissues in teratomas derived from the
implantation of diploid hiPS cells into animals12.
Importantly, a very common class of spontaneous tumors in Figure 1  Teratoma from hES cells showing varying degrees of maturation.
adults that may contain teratoma, type II testicular germ cell tumors This teratoma from an hES cell transplant in immunocompromised mice
(GCTs)13 (Table 1), frequently contains undifferentiated malignant shows a range of somatic tissues displaying varying degrees of maturation.
cells called embryonal carcinoma (EC) cells. EC cells are the stem (a) Primitive neuroectodermal tissue consisting of neural tube–like
structures (arrow). (b) Classic mesodermal tissues can be observed
cells of these tumors and are characterized by expression of pluri­ in many hES cell–derived teratomas, including cartilage (arrow).
potency genes such as OCT3/4 (also known as POU5F1) and SOX2. (c) Endodermal tissues resembling mucosal structures of the gut
They have not been described in xenografts of diploid hES cells but surrounded by smooth muscle (arrow). H&E-stained paraffin sections of
have been found in hiPS cell grafts 12. This is in contrast to tumors human teratoma from the collection of M.F.P.

850 VOLUME 30  NUMBER 9  SEPTEMBER 2012  nature biotechnology


perspective

Table 1  The five types of human GCTs including teratomas


Type I II III IV V
Appearance

Phenotype Teratoma/yolk-sac tumor Seminoma/nonseminoma/ Spermatocytic seminoma Dermoid cyst Gestational trophoblastic
dysgerminoma/germinoma neoplasia (hydatidiform
mole & choriocarcinoma)
Elements of teratoma Yes Yes No Yes No
Incidence 0.12/100,000 6–11/100,000 0.1/100,000 3/100,000 7.5/100,000
Gender bias Females Males Males (exclusively) Females Females (exclusively)
Usual age of onset Neonates and children ~12–35 years >50 years Children/adults Fertile women
(adolescents to adults)
Cell of origin Stem cell/PGC/gonocyte PGC/gonocyte Spermatogonia/spermatocyte Primary/secondary oocyte Empty ovum fertilized
by sperm
Imprinting status Biparental/partially erased Erased Male Maternal Paternal
Genotype Diploid/aneuploid Tri-tetraploid/aneuploid Diploid/aneuploid Diploid/tri-tetraploid Diploid
Columns I, II and IV represent those GCTs that may contain elements of teratoma. Images are hematoxylin and eosin (H&E)-stained sections of human GCTs from the collection of
T.M.U. Table adapted from Oosterhuis and Looijenga13. Incidence, gender bias and usual age of onset were derived from the International Agency for Cancer Research82.
© 2012 Nature America, Inc. All rights reserved.

undifferentiated cells with a diploid karyotype that lacked markers of however, whether such lesions should be regarded as neoplasms dis-
pluripotent stem cells. Although the precise nature of these tumors tinct from teratomas or, rather, as ‘monodermal’ teratomas, that is,
and their biological significance remain unknown, this finding sug- lineage-restricted proliferations that are otherwise similar to classic
gests that the microenvironment of the graft can profoundly influence teratomas (see below). Importantly, any form of malignant or benign
the histology of the tumor. This is also demonstrated in type II GCTs, neoplasm that arises from a graft presents a safety hazard. Unwanted
the histology of which is influenced by the anatomical localization ectopic graft tissue can impinge on vital structures directly or indi-
of the gonad, whereas mainly the seminoma phenotype occurs in rectly (secrete neurotransmitters, hormones or induce inflammation,
nonscrotal testis (as is the case for the ovarian counterpart). fibrosis or malignant transformation). However, patients with benign
During propagation in vitro, human pluripotent stem cells can spontaneous teratomas, even large ones, are generally cured by surgi-
acquire characteristic genetic abnormalities, such as overrepresenta- cal resection alone25,26.
tion of chromosome 12 or 17 (refs. 16,17). Many of these abnormali- The risk of tumor formation associated with any pluripotent stem
ties are also found in some classes of spontaneously arising type II cell–derived therapy requires the development of strategies to assure
GCTs that may contain a component of teratoma. Several reports have the absence of residual stem cells. The challenge occurs in trying to
shown that cell lines bearing these characteristic genetic abnormali- guarantee that a product is completely free of undifferentiated cells
ties form tumors with fewer differentiated elements, and at a lower or other partially differentiated cells that could generate a tumor.
cell inoculum, compared with diploid cells18–20. There are also some Below we discuss features of spontaneous teratomas in humans that
indications that tumors from genetically abnormal hES cell cultures may illuminate efforts to estimate the risks of pluripotent stem cell–
npg

can contain elements of EC21,22, although this is as yet an infrequent derived therapies.
observation.
The primary risk of pluripotent stem cell therapy is often consid- Spontaneous teratomas in humans
ered to be teratoma formation, but there is also the potential that Most spontaneous teratomas belong to a subgroup of a broader and
other types of embryonal neoplasms corresponding to specific tis- complex class of neoplasms referred to as GCTs. Recently, a new clas-
sue lineages might arise after engraftment. Hypothetically, these sification of GCTs was proposed based on genetics and site of origin
could develop from progenitor populations that possess varying as well as morphology13,27. Five types were identified (Table 1), three
degrees of replicative capacity. Some examples include the forma- of which— types I, II and IV—may contain elements of teratoma.
tion of cystic lesions in rats inoculated with oligodendrocyte precur- The predominantly gonadal distribution and origin of types II and
sors23 and the formation of primitive neural tissue after engraftment IV teratomas indicate that they likely arise out of multiple defective
of hES cell–derived dopaminergic neurons24. It remains debatable, processes during germ cell development. In contrast, the origin of

Table 2  Similarities between Type I teratomas and hES cell–derived teratomas


Type Phenotype Cytological features Cell of origin Imprinting status Genotype Common aneuploidies
Type I teratoma Mature or immature ter- Immature organoid PGC or pluripotent Biparental or partially Diploid Loss of (parts) of 1p, 4, 6q;
atoma (unless complicated differentiation with stem cell erased gain of (parts) of 1q, 12p13,
by yolk sac tumor) or without yolk 20q, 22 (only observed in yolk
sac elements sac tumor)
hES cell teratoma Mature or immature Immature organoid Pluripotent cell from Biparental Diploid 3, 5, 8, 12, 13, 17, 18, 20,
teratoma (with or without differentiation with or ICM of blastocyst (monoallelic) 21, X
cystic elements) without cysts
ICM, inner cell mass; PGC, primordial germ cell.

nature biotechnology  VOLUME 30  NUMBER 9  SEPTEMBER 2012 851


Perspective

Figure 2  Genomic imprinting and developmental origin of GCTs. Germ cell tumors (I–IV)
Schematic of normal (black) and aberrant (red) germ cell development. Normal development
Embryogenesis starts with fertilization and formation of a population
GI Fertilization GI
of pluripotent stem cells (epiblast cells), which give rise to the various
differentiation lineages (somatic and extra-embryonic) in vivo and to ES +
cells in culture. All pluripotent stem cells have a bi-parental pattern of
genomic imprinting (GI) (yellow), resulting from a pure male (blue) and Proliferation

Not present

Not present
female (purple) mature germ cell. GI is an epigenetic mechanism that ES
differentially marks the maternal and paternal alleles of a specific set of Differentiation
~100 genes, such that only one parental allele of an imprinted gene is ES
expressed. Imprinting marks inherited from either parent must be erased TE/YST
Specification
TE/YST
and reset during development of the germ line, an event that occurs once (I) Germ cell Compartment
(I)
the primordial germ cells have migrated to the gonads. The different types PGC
Migration
of GCTs originate from different stages of stem or germ cell development.
Erasement
Type I are teratomas (TE) and yolk sac tumors (YST) found in neonates Gonocyte
SE/NS DG/NS
and infants that originate from a stem or primordial germ cell with a bi- Sex determination
(II) (II)
parental or partially erased pattern of GI. Type II are the seminomas (SE)/ (Pre-)
dysgerminomas (DG) and nonseminomas (NS) diagnosed in adolescents Spermatogonium Oogonium
and young adults that originate from an erased germ cell. Type III are the (A,B)
spermatocytic seminomas (SS), predominantly in elderly males. Type IV Prim. oocyte
are dermoid cysts (DC). Type V are hydatidiform moles (MH). The timing of
birth, start of meiosis and puberty are indicated.
SS Prim. spermatocyte DC
(III) (IV)
© 2012 Nature America, Inc. All rights reserved.

Sec. spermatocyte Sec. oocyte


type I tumors is less clear; they may arise either from pluripotent cells

Germ cell

Germ cell
Spermatid Ovum

Soma

Soma
of the peri-implantation embryo that do not undergo normal pattern-
Testis Ovary
ing or from an early germ cell precursor. Thus, the term ‘teratoma’
designates a generic class of neoplasms with similar histopathological
appearance but potentially diverse origins. MH
(V)

Teratomas formed by pluripotent stem cells resemble type I GCTs Sperm Mature oocyte
Histologic and genomic analysis suggests that teratomas in the Birth Start meiosis Start puberty
type I GCT category, but not type II or IV, are the most plausible clini-
cal equivalent of an hES cell–derived teratoma. Human pluripotent
stem cells are phenotypically similar to the human EC cells found in tetraploid (Table 1). The exception are type IV teratomas, which
type II GCTs, and some previous reports have suggested that experi- frequently show loss of heterozygosity for polymorphic markers,
mental teratomas derived from karyotypically abnormal hES cells supporting their derivation from a diploid, post-meiotic I germ
mimic type II GCTs16. Certainly, some features of type II GCTs are cell, with any given chromosome being either of mostly maternal or
similar to teratomas derived from abnormal hES cells, including an paternal origin32.
immature tissue phenotype, appearance of carcinoma-like or EC cells
and aneuploidies (gains or losses in chromosomes X, 12 and 17)16. Similarities in differentiation patterns. Differentiation within
But teratomas derived from diploid hES or hiPS cells do not resemble type I and hES cell teratomas results in the development of mature,
npg

type II GCTs. In fact, type II GCTs originate from a malignant organoid cellular structures. Type I teratomas characteristically
­intratubular germ cell28 with an aneuploid DNA content29. In contrast, ­consist of mature tissues having organ-like relationships to each other,
type I mature teratomas generally have no malignant potential. including, for example, smooth muscle encircling, in an ­intestinal-
Type IV teratomas should not be equated with pluripotent stem like fashion, enteric-type epithelium resting on a ­supporting ­lamina
cell teratomas either, despite their content of mature, organoid propria, or pancreatic tissue having normal ductal and islet cell
tissue types. They are almost exclusively found in the ovaries and arise components admixed with acinar tissue. Remarkably, teratomas
from an oocyte that has undergone parthenogenesis. These lesions derived from karyo­typically normal hES cells show an organoid
are usually cystic and form skin-like structures (dermoid cysts or pattern of differentiation that extends to the development of tissue
mature cystic teratomas). niches within the tumor. This suggests that developing structures
Type I GCTs are divided into two classes: teratomas, which are dip- within hES cell teratomas retain their normal interactive signals to
loid, and yolk sac tumors, which are typically aneuploid. They can co-instruct the specification of surrounding regions33,34, mimicking
arise in a wide variety of body regions, with the anatomical site playing embryonic development. We (L.H.J.L) and others have seen OCT3/4+
a part in their pathogenesis. Sacrococcygeal teratomas are the most cells in type I teratomas, but they are negative for CD30 (refs. 35–37),
common extra-gonadal type I GCT30 and are the likely parallel of hES indicating that they do not contain EC cells.
cell–derived teratomas. That these tumors might originate from an
early germ cell is supported by experiments in mice31. Evidence sup- Shared cell of origin? Type I teratomas and hES cell–derived terato-
porting the close relationship between type I GCTs and teratomas from mas may have a similar cell of origin—a pluripotent stem cell from
pluripotent stem cells is discussed below and summarized in Table 2. the inner cell mass or an early germ cell. hES cells are generated by
culturing the inner cell mass. However, Zwaka and Thomson specu-
Karyotypic similarities. Both type I and hES cell teratomas contain lated that during ES cell establishment, the pluripotent self-renewing
diploid, karyotypically normal cells. By contrast, most teratomas population arises from an obligate early germ cell intermediate
within the other types of GCTs are typically aneuploid, triploid or that differentiates from the inner cell mass38. Recent evidence from

852 VOLUME 30  NUMBER 9  SEPTEMBER 2012  nature biotechnology


perspective

mouse studies supports this hypothesis39,40. Table 3  Histologic features associated with malignancy in human teratomas
Likewise, the sacrococcygeal teratoma is Malignant features Description Tissue types of concern Examples
hypothesized to originate from a stem cell Primitive Cellular components with Primitive neuroectodermal
from the inner cell mass or a remnant of the ­embryonic-type densely packed chromatin rosettes or tubules; primitive
epiblast (a derivative of the inner cell mass), elements showing both apoptosis and nephrogenic tissues
mitotic activity
or from a very early primordial germ cell.
These stem cells are thought to have escaped
normal regulatory control and become mis-
placed during embryogenesis41,42. It has been
suggested that sacrococcygeal teratomas Extra-embryonic Spider web-like network of Yolk sac elements
express pluripotent markers without indica- tissues cells with vacuolated cyto-
plasm and abundant basement
tions of malignant transformation36, but this membrane deposits; may also
has to be confirmed in independent studies. have tubular, cystic, papillary
Further support for a very early plur­i­ and solid patterns of cells
potent or germ cell origin of sacrococcygeal
teratomas comes from studies of genomic
imprinting (Fig. 2). Type I teratomas show a Vascular space Endothelial-lined spaces Intravascular thrombi
invasion containing clumps of
pattern of imprinting that is either somatic or
neoplastic cells
partially erased, and therefore consistent with
an origin in the peri-implantation epiblast or
early primordial germ cell13,27,43. By contrast,
© 2012 Nature America, Inc. All rights reserved.

type II teratomas have a completely erased


genomic imprint, consistent with a gonocyte
origin, whereas type IV teratomas show com- Phenotype of Any of a variety of neoplasms Established pathological
plete maternal imprinting, consistent with a malignancy as seen at numerous hallmarks of malignancy
13,27,43 body sites
post-meiotic germ cell origin . These
data suggest that type I tumors originate
from pluripotent stem cells or early primor-
dial germ cells. In hES and hiPS cell cultures,
imprinting is variable, with partial erasure of
imprinting at some loci in some cell lines44. Images are H&E-stained sections of human teratomas from the collection of T.M.U.
Although this variability is often attributed
to growth under artificial conditions in vitro,
it could also be consistent with a developmental status between that genetic, epigenetic and histological features of these usually benign
of an epiblast cell and one that has embarked on the pathway to germ tumors that are strongly predictive of progression to malignancy.
cell formation and erasure of imprinting. The appearance of such features in hES or hiPS cell–derived terato-
mas would therefore be suggestive of the presence of transformed
Similarities in dysregulation. Finally, cells giving rise to both type I cells in the cultures.
teratomas and hES cell–derived teratomas might share similar chro-
npg

mosomal aberrations and cellular hallmarks when malignant progres- Histological hallmarks of malignancy in teratomas. Malignant
sion is observed16,45–48. Morphologically, these aberrations manifest changes in human teratomas in general correlate with the pres-
as distinct features in sacrococcygeal teratomas, including the pres- ence of several key histological hallmarks (Table 3 and Fig. 3). The
ence of pockets of OCT3/4+ cells, yolk-sac elements and, in some grading of teratomas was originally established for ovarian cases
instances, neuroectoderm. Not surprisingly, disabling p53 improves and ranges from grade 0 (absence of embryonic-type elements) to
hiPS cell generation but also leads to enhanced tumor formation in grade 3 (substantial amounts of immature elements)50. Malignant
chimeric mice49. These topics are discussed further below. behavior is almost exclusively confined to those tumors that are
grade 2 or 3 (‘high’ grade)51 and is based on the selection of the
Lessons from teratomas for stem cell translation slide that contains the greatest amount of immature neuroectoderm.
The demonstrated potential of cultured human pluripotent stem Similar criteria have also been applied to sacrococcygeal teratomas
cells to accumulate genetic lesions characteristic of malignant (type I) and found to correlate with clinical behavior52 but are not
transformation has been a major concern for the field. Advances in useful for type II teratomas. Karyotypic abnormalities (gain of auto-
genomics technology have enabled genome-wide sequencing of hES somal and X chromosomes) in immature ovarian teratomas have
and hiPS cells, and, as a consequence, in addition to gross karyotypic correlated with large volumes of embryonic-type neuroectoderm,
changes, there are increasing reports of small amplifications, dele- with normal chromosomal findings in tumors lacking or having
tions and point mutations occurring during prolonged periods of small amounts of such tissue53,54. In addition, in sacrococcygeal
culture16,17. The emerging challenge for safety assessment lies in the teratomas, high-grade immaturity correlates with the development
interpretation of the significance of these changes and their potential of a yolk sac tumor component, which is considered to be the ‘true’
effects on cell behavior. Biological assays that enable the detection indicator of malignant behavior in this disease 45,52. Karyotypic
of a malignant phenotype in pluripotent stem cells are therefore studies of sacrococcygeal teratomas lacking a yolk sac tumor com-
of enormous value. Here, reference to spontaneously occurring ponent have yielded normal results, with chromosomal abnormali-
teratomas can inform our understanding because there are certain ties correlating only with the evolution of yolk sac tumor 46,54–58.

nature biotechnology  VOLUME 30  NUMBER 9  SEPTEMBER 2012 853


Perspective

Figure 3  Histologic features characteristic of malignant progression


in human teratomas. (a) This benign teratoma in a child shows an a b
organoid replication of pancreatic tissue, including large ducts, lobules
of acinar tissue and occasional islets. Scale bar, 500 µm. (b) Primitive
neuroectodermal tissue (arrow, upper left field) consisting of neural-
type tubules lined by stratified columnar cells is interspersed with more
mature tissues, including a cellular fibrous stroma and cystic gland
Scale bar, 200 µm. (c) Overgrowth of primitive neuroectodermal tissue
in this testicular teratoma from a young man results in a primitive
neuroectodermal tumor (PNET). Scale bar, 200 µm. (d) Endodermal yolk-
sac elements appear as loosely arranged cords of cells (arrow, upper field)
and represent the classic hallmark of a malignant change in an otherwise
pure teratoma. Scale bar, 100 µm. H&E-stained paraffin sections of
human teratoma from the collection of T.M.U.

Importantly, the presence of primitive embryonic tissue components,


in the absence of the other criteria, classifies a lesion as an imma- c d
ture teratoma and potentially malignant.
Primitive neuroectodermal elements have been observed in
grafts of pluripotent stem cells (Fig. 1), as have yolk sac elements12,
consistent with their similarity to the type I GCT. It is not clear
whether yolk sac elements are rare or whether they have been under-
© 2012 Nature America, Inc. All rights reserved.

reported because they were unrecognized. Thus, accurate assessment


of the malignant potential of teratomatous lesions requires a trained
pathologist who has extensive experience in the spectrum of human
germ cell neoplasia.

Genetic and epigenetic changes predictive of malignancy. Previous


studies have noted that frequently observed karyotypic abnormalities
in pluripotent stem cells, in particular overrepresentation of chromo-
somes 12, 17 and X16, or functional inactivation of the TP53 pathway,
are characteristic of malignant type II GCTs13,59. Malignancy in spon-
taneous type I human teratomas is most often correlated with yolk sac Preclinical and clinical assessment of safety
elements that contain similar and other aneuploidies (Table 2)46,55–58. The preclinical safety assessment of cellular therapeutics currently
The remainder of the type I teratomas contain diploid cells with no relies on in vitro and in vivo tests. The key genetic and histological
current genetic data to explain pathogenesis. Several studies have features that characterize malignant progression of type I teratomas
demonstrated altered growth and differentiation properties of aneu- in humans are discussed above and should be considered in safety
ploid hES cells60,61. Only one study has shown that genetic aberrations assessment when examining cell lines or teratoma xenografts derived
in hES cell lines alter the behavior of cells derived from them. In this from them. Currently, the main objectives of in vitro safety testing are
report, differentiated astrocytes derived from an aneuploid hES cell to assess the presence of stem cells, the purity of the differentiated
npg

line had properties analogous to astrocytoma and glioblastoma62. In population and the presence of cells with genetic alterations related
principle, any genetic alterations could confer a metastatic potential to malignancy. Highly sensitive approaches to achieving the first two
and should be used as an exclusion criterion. of these aims are discussed below. High-throughput assays to monitor
Could epigenetic change account for the appearance of a benign genetic or epigenetic alterations across panels of oncogenes or tumor
type I teratoma? There may be subtle changes in the epigenome inde- suppressors, or losses or gains of genetic material, are available now
pendent of DNA promoter methylation, possibly relating to polycomb but are limited in their ability to pick up changes in minority cell
genes, that silence tumor suppressor gene activity in cells within ter- populations. There is a need to develop more- sensitive in vitro assays
atomas63. In pluripotent stem cells, polycomb genes act as repressors of tumorigenic potential. These might include biological assays of
of transcription factors that drive differentiation64. Compared with cell behavior, such as changes in self renewal, growth or cell survival,
normal hES cells, EC cells show two specific repressive polycomb or tests to measure expression of protein markers characteristic of
marks in their chromatin: dimethylated H3K9 and trimethylated transformed cells.
H3K9 (ref. 65). These epigenetic marks may function to inhibit dif- In vivo assays are aimed at assessment of tumor formation in immuno­
ferentiation and are associated with hypermethylation of DNA and compromised animals. These tests are limited by two factors. First, as
inactivation of tumor suppressors in adult malignancies. Although in assays of cancer stem cells70, insufficient sensitivity may underes-
these observations were limited to one EC cell line derived from a timate tumorigenic potential (owing to residual innate or acquired
type II malignant GCT, similar epigenetic changes may be predictive immunity in immunocompromised animals or possible requirements
of malignant behavior in hES or hiPS cells. It is reported that EC cells for species-specific microenvironmental factors to support tumor
in type II GCTs have an intermediate DNA methylation pattern 66,67 growth). Second, the relatively short life span of rodent test species
related to activation of the de novo methyltransferase DNMT3L68. precludes long-term monitoring of grafts that may persist in patients
No data are available so far regarding the type I GCT. In this context, for decades. It is difficult to see how these limitations can be overcome
analysis of additional markers, including SSEA-5 in hES and hiPS in routine practice, so the development of surrogate markers for cell
cells69, is of interest. transformation is a priority.

854 VOLUME 30  NUMBER 9  SEPTEMBER 2012  nature biotechnology


perspective

Table 4  Methods for assaying residual pluripotent stem cells in clinical product or patient monitoring
Molecular Potential stage of application
Method of detection compartment Sensitivity Limitations Preclinical CMC Clinical
Methods with accepted clinical utility
ASO-qRT-PCR Tumor DNA 0.001%, that is, 1 CTC in Requires large number of samples for repeated Yes Yes Yes
100,000 normal83 testing to assure statistical certainty
Flow cytometry Tumor cell 0.01%, that is, 1 CTC in Requires four- to six-color flow, necessitating Yes Yes Yes
10,000 normal84 multiple cell surface markers
ELISAs Tumor protein Ultrasensitive assays detect Requires unique protein expression & correlation of Yes Yes Yes
in sub-pg/ml range85 protein signal with cell number
MRI Tumor cell Masses >0.3 cm Unknown effect of imaging labels on stem cell Yes No Yes
phenotype or genotoxic potential
FDG-PET Tumor size Masses >1 cm (ref. 86) Poor spatial resolution No No Yes

Methods with emerging evidence


qRT-PCR Tumor miRNA Limit of detection down to Requires identification of miRNAs with known Yes Yes Yes
10 copies of miRNA87 association with pluripotent cells
Immuno-PCR (TPA) Tumor protein Limit of detection in Requires unique protein expression & correlation of Yes Yes Yes
femtogram range88 protein signal with cell number
Fluorescent nanocrystals Tumor miRNA Limit   of detection in Requires identification of miRNAs with known Yes Yes Yes
& cation exchange ­femtomole range89 association with pluripotent cells
Nanoparticle surface Tumor miRNA Limit of detection in Requires identification of miRNAs with known Yes Yes Yes
© 2012 Nature America, Inc. All rights reserved.

plasmon resonance attomole range90 association with pluripotent cells


Bioluminescence (BLI) Tumor cell Limit of detection to be Requires demonstration that vectors used to label Yes No No
determined91,92 cells have no effect on cell product profile
CMC, product chemistry, manufacturing and controls; CTC, circulating tumor cell; ELISA, enzyme-linked immunosorbent assay; miRNA, microRNA; MRI, magnetic resonance
imaging; PET, positron emission tomography; TPA, TaqMan protein assay.

Thresholds in safety assessment. A key hurdle is determining the ­ roliferating cell populations such as hES or hiPS cells72, and whose
p
threshold level of residual stem cells that poses a risk for teratoma persistence in vivo may be limited. Serum biomarkers could provide
development. Because there is so much variation among protocols a more sensitive and simpler method for monitoring the safety and
for teratoma formation in the stem cell literature, estimates of the predicting the risk of teratoma formation in preclinical studies and
threshold for teratoma formation range anywhere from 2 to >10,000 in patients receiving hES cell–derived therapies, as they do in the
cells6,7,71. Results from xenograft data, even in immunocompromised diagnosis and monitoring of cancer.
or humanized animals, would probably overestimate the true value. Certain genes that are expressed mainly during development and
The natural history of spontaneous type I teratomas suggests that this not in adult tissues are often reexpressed by transformed cancer cells.
threshold is more likely to be within the low end of the range, maybe Although no definitive teratoma biomarker currently exists, proteins
even <10 cells. As noted above, type I teratomas are thought to arise such as alpha-fetoprotein, beta human chorionic gonadotropin and
either from epiblast cells that have failed to undergo differentiation placental alkaline phosphatase are clinically approved markers for
or from early primordial germ cells. Either of these cell populations some histologically defined forms of GCT. The sensitivity and selectiv-
npg

is quite small, and founder cells of the tumors likely represent only a ity of some of these marker assays are impressive; measurement of cir-
small subpopulation within them, as type I tumors are observed in culating beta human chorionic gonadotropin in pregnancy can detect
patients who had normal development of the soma and germ line. If the presence of 102 cells producing the hormone. Alpha-fetoprotein,
the founder population of type I GCTs is within the range of 10–100 in particular, has become a useful diagnostic for the development and
cells, then detection of contamination in stem cell therapeutics must recurrence of yolk sac elements in patients with type I teratomas52,73.
be accurate in the range of 1 in 107 to 1 in 108 cells for applications A plethora of secreted and cell-surface molecules74–76, transcription
requiring an inoculum of 1 × 109 cells. Highly sensitive technologies factors77–79 and micro RNAs80,81 that are characteristic of pluripotent
will be required to achieve such levels of detection (Table 4). Because stem cells or their early differentiated progeny have now been identi-
a spontaneous tumor is similar to an autograft, these considerations fied, and it is easy to envision that sensitive multiplex assays for such
would apply most closely to hiPS cell therapy. In hES cell therapy, the markers could be used in preclinical or phase 1 safety assessment.
use of allografts combined with immunosuppression further compli- Such assays could be used to detect the presence of pluripotent stem
cates the assessment of threshold values. cells or lineage-committed progenitors with proliferative capacity and
neoplastic potential (Table 4). In addition to conventional immuno­
Methods to evaluate and monitor pluripotent cell–derived trans- assays for serum proteins, new and highly sensitive assay technologies
plants. Despite the marked progress in imaging technologies, today’s that can exploit these opportunities are emerging. Furthermore, with
science falls short of readily visualizing a single or small cluster of the advent of ultrasensitive methods for monitoring preneoplastic
neoplastic cells in patients, a level of sensitivity that may be neces- mutations from circulating DNA in plasma samples, it would be
sary for monitoring stem cell transplants. Positron emission tomog- feasible to detect metastatic spread of a transplant.
raphy (PET) typically resolves masses ≥1 cm, a size that could contain
millions of cells. Tracking cells with magnetic resonance imaging CONCLUSIONS
(MRI) requires labeling them with contrast agents whose physico- Assessment of the risk of tumor formation by pluripotent stem cell–
chemical properties have an as yet unknown genotoxic potential for based therapies is an essential part of safety evaluation. Unfortunately,

nature biotechnology  VOLUME 30  NUMBER 9  SEPTEMBER 2012 855


Perspective

at present the basic scientific knowledge underlying efforts to develop 21. Adewumi, O. et al. Characterization of human embryonic stem cell lines by the
a sound regulatory framework in this field is rudimentary. Studies of International Stem Cell Initiative. Nat. Biotechnol. 25, 803–816 (2007).
22. Andrews, P.W. et al. Embryonic stem (ES) cells and embryonal carcinoma (EC)
the origin, pathology, genetics and clinical management of terato- cells: opposite sides of the same coin. Biochem. Soc. Trans. 33, 1526–1530
mas in humans provide rich opportunities for reverse translational (2005).
23. Akst, J. Cysts stall stem cell trial. Scientist (Aug). 27 (2009).
research to drive future laboratory studies. Although parallels exist 24. Roy, N.S. et al. Functional engraftment of human ES cell-derived dopaminergic
between hES cells, hiPS cells and the EC cells of malignant human neurons enriched by coculture with telomerase-immortalized midbrain astrocytes.
teratomas that commonly occur in adolescents or young adults, the Nat. Med. 12, 1259–1268 (2006).
25. De Backer, A., Madern, G.C., van de Ven, C.P., Tibboel, D. & Hazebroek, F.W.
tumors that arise from cultured pluripotent stem cells have more in Strategy for management of newborns with cervical teratoma. J. Perinat. Med. 32,
common with benign (type I) GCTs of early childhood. The natural 500–508 (2004).
history of these spontaneously occurring tumors provides guidelines 26. De Backer, A. et al. Study of the factors associated with recurrence in children
with sacrococcygeal teratoma. J. Pediatr. Surg. 41, 173–181 (2006).
for preclinical safety assessment of hES or hiPS cell lines and for the 27. Oosterhuis, J.W., Looijenga, H.J., van Echten, J. & de Jong, B. Chromosomal
development of measures to monitor stem cell grafts in preclinical constitution and developmental potential of human germ cell tumors and teratomas.
Cancer Genet. Cytogenet. 95, 96–102 (1997).
studies and in phase 1 trials. 28. Simmonds, P.D. et al. Primary pure teratoma of the testis. J. Urol. 155, 939–942
(1996).
Acknowledgments 29. de Graaff, W.E. et al. Ploidy of testicular carcinoma in situ. Lab. Invest. 66,
J.J.C. would like to thank T. Dervieux for critical review and feedback on the 166–168 (1992).
manuscript. L.H.J.L. would like to thank J. Wolter Oosterhuis for the wonderful 30. Gobel, U. et al. Germ-cell tumors in childhood and adolescence. GPOH MAKEI and
discussions of the last decades, significantly contributing to the work performed. the MAHO study groups. Ann. Oncol. 11, 263–271 (2000).
31. Runyan, C., Gu, Y., Shoemaker, A., Looijenga, L. & Wylie, C. The distribution and
behavior of extragonadal primordial germ cells in Bax mutant mice suggest a novel
COMPETING FINANCIAL INTERESTS origin for sacrococcygeal germ cell tumors. Int. J. Dev. Biol. 52, 333–344
The authors declare no competing financial interests. (2008).
32. Vortmeyer, A.O. et al. Microdissection-based analysis of mature ovarian teratoma.
Published online at http://www.nature.com/doifinder/10.1038/nbt.2329.
© 2012 Nature America, Inc. All rights reserved.

Am. J. Pathol. 154, 987–991 (1999).


Reprints and permissions information is available online at http://www.nature.com/ 33. Gertow, K. et al. Organized development from human embryonic stem cells after
reprints/index.html. injection into immunodeficient mice. Stem Cells Dev. 13, 421–435 (2004).
34. Blum, B. & Benvenisty, N. Clonal analysis of human embryonic stem cell
differentiation into teratomas. Stem Cells 25, 1924–1930 (2007).
1. Laflamme, M.A. et al. Cardiomyocytes derived from human embryonic stem cells in 35. Busch, C., Oppitz, M., Wehrmann, M., Schweizer, P. & Drews, U. Immunohistochemical
pro-survival factors enhance function of infarcted rat hearts. Nat. Biotechnol. 25, localization of Nanog and Oct4 in stem cell compartments of human sacrococcygeal
1015–1024 (2007). teratomas. Histopathology 52, 717–730 (2008).
2. Lu, B. et al. Long-term safety and function of RPE from human embryonic stem cells 36. Busch, C. et al. Isolation of three stem cell lines from human sacrococcygeal
in preclinical models of macular degeneration. Stem Cells 27, 2126–2135 (2009). teratomas. J. Pathol. 217, 589–596 (2009).
3. Przyborski, S.A. Differentiation of human embryonic stem cells after transplantation 37. Rijlaarsdam, M.A. et al. Specific detection of OCT3/4 isoform A/B/B1 expression
in immune-deficient mice. Stem Cells 23, 1242–1250 (2005). in solid (germ cell) tumours and cell lines: confirmation of OCT3/4 specificity for
4. Cooke, M.J., Stojkovic, M. & Przyborski, S.A. Growth of teratomas derived from germ cell tumours. Br. J. Cancer 105, 854–863 (2011).
human pluripotent stem cells is influenced by the graft site. Stem Cells Dev. 15, 38. Zwaka, T.P. & Thomson, J.A. A germ cell origin of embryonic stem cells? Development
254–259 (2006). 132, 227–233 (2005).
5. Prokhorova, T.A. et al. Teratoma formation by human embryonic stem cells is site- 39. West, J.A., Park, I.H., Daley, G.Q. & Geijsen, N. In vitro generation of germ cells
dependent and enhanced by the presence of Matrigel. Stem Cells Dev. 18, 47–54 from murine embryonic stem cells. Nat. Protoc. 1, 2026–2036 (2006).
(2009). 40. Chu, L.F., Surani, M.A., Jaenisch, R. & Zwaka, T.P. Blimp1 expression predicts
6. Hentze, H. et al. Teratoma formation by human embryonic stem cells: evaluation embryonic stem cell development in vitro. Curr. Biol. 21, 1759–1765
of essential parameters for future safety studies. Stem Cell Res. (Amst.) 2, 198–210 (2011).
(2009). 41. O’Rahilly, R. & Muller, F. Neurulation in the normal human embryo. Ciba Found.
7. Lee, A.S. et al. Effects of cell number on teratoma formation by human embryonic Symp. 181, 70–82 (1994).
stem cells. Cell Cycle 8, 2608–2612 (2009). 42. Sano, K. Pathogenesis of intracranial germ cell tumors reconsidered. J. Neurosurg.
8. Shih, C.C., Forman, S.J., Chu, P. & Slovak, M. Human embryonic stem cells are 90, 258–264 (1999).
prone to generate primitive, undifferentiated tumors in engrafted human fetal tissues 43. Oosterhuis, J.W., Stoop, H., Honecker, F. & Looijenga, L.H. Why human extragonadal
npg

in severe combined immunodeficient mice. Stem Cells Dev. 16, 893–902 (2007). germ cell tumours occur in the midline of the body: old concepts, new perspectives.
9. Kelly, P.N., Dakic, A., Adams, J.M., Nutt, S.L. & Strasser, A. Tumor growth need Int. J. Androl. 30, 256–264 (2007).
not be driven by rare cancer stem cells. Science 317, 337 (2007). 44. Blum, B. & Benvenisty, N. The tumorigenicity of human embryonic stem cells.
10. Quintana, E. et al. Efficient tumour formation by single human melanoma cells. Adv. Cancer Res. 100, 133–158 (2008).
Nature 456, 593–598 (2008). 45. Heifetz, S.A. et al. Immature teratomas in children: pathologic considerations: a
11. Gertow, K. et al. Early events in xenograft development from the human embryonic report from the combined Pediatric Oncology Group/Children’s Cancer Group.
stem cell line HS181–resemblance with an initial multiple epiblast formation. Am. J. Surg. Pathol. 22, 1115–1124 (1998).
PLoS ONE 6, e27741 (2011). 46. Veltman, I.M., Schepens, M.T., Looijenga, L.H., Strong, L.C. & van Kessel, A.G.
12. Griscelli, F. et al. Malignant germ cell-like tumors, expressing Ki-1 antigen (CD30), Germ cell tumours in neonates and infants: a distinct subgroup? APMIS 111,
are revealed during in vivo differentiation of partially reprogrammed human-induced 152–160 (2003).
pluripotent stem cells. Am. J. Pathol. 180, 2084–2096 (2012). 47. Spits, C. et al. Recurrent chromosomal abnormalities in human embryonic stem
13. Oosterhuis, J.W. & Looijenga, L.H. Testicular germ-cell tumours in a broader cells. Nat. Biotechnol. 26, 1361–1363 (2008).
perspective. Nat. Rev. Cancer 5, 210–222 (2005). 48. Golas, M.M., Gunawan, B., Raab, B.W., Fuzesi, L. & Lange, B. Malignant
14. Rossant, J. & Papaioannou, V.E. The relationship between embryonic, embryonal transformation of an untreated congenital sacrococcygeal teratoma: a amplification
carcinoma and embryo-derived stem cells. Cell Differ. 15, 155–161 (1984). at 8q and 12p detected by comparative genomic hybridization. Cancer Genet.
15. Fu, W. et al. Residual undifferentiated cells during differentiation of induced Cytogenet. 197, 95–98 (2010).
pluripotent stem cells in vitro and in vivo. Stem Cells Dev. 21, 521–529 49. Hong, H. et al. Suppression of induced pluripotent stem cell generation by the
(2011). p53-p21 pathway. Nature 460, 1132–1135 (2009).
16. Baker, D.E. et al. Adaptation to culture of human embryonic stem cells and 50. Norris, H.J., Zirkin, H.J. & Benson, W.L. Immature (malignant) teratoma of the
oncogenesis in vivo. Nat. Biotechnol. 25, 207–215 (2007). ovary: a clinical and pathologic study of 58 cases. Cancer 37, 2359–2372
17. Harrison, N.J., Baker, D. & Andrews, P.W. Culture adaptation of embryonic stem (1976).
cells echoes germ cell malignancy. Int. J. Androl. 30, 275–281 (2007). 51. O’Connor, D.M. & Norris, H.J. The influence of grade on the outcome of stage I
18. Fazeli, A. et al. Altered patterns of differentiation in karyotypically abnormal human ovarian immature (malignant) teratomas and the reproducibility of grading. Int. J.
embryonic stem cells. Int. J. Dev. Biol. 55, 175–180 (2011). Gynecol. Pathol. 13, 283–289 (1994).
19. Tavakoli, T. et al. Self-renewal and differentiation capabilities are variable between 52. Gonzalez-Crussi, F., Winkler, R.F. & Mirkin, D.L. Sacrococcygeal teratomas in infants
human embryonic stem cell lines I3, I6 and BG01V. BMC Cell Biol. 10, 44 and children: relationship of histology and prognosis in 40 cases. Arch. Pathol.
(2009). Lab. Med. 102, 420–425 (1978).
20. Moon, S.H. et al. Effect of chromosome instability on the maintenance and 53. Surti, U., Hoffner, L., Chakravarti, A. & Ferrell, R.E. Genetics and biology of human
differentiation of human embryonic stem cells in vitro and in vivo. Stem Cell Res. 6, ovarian teratomas. I. Cytogenetic analysis and mechanism of origin. Am. J. Hum.
50–59 (2011). Genet. 47, 635–643 (1990).

856 VOLUME 30  NUMBER 9  SEPTEMBER 2012  nature biotechnology


perspective

54. Baker, B.A. et al. DNA content of ovarian immature teratomas and malignant germ 74. Lebkowski, J.S. et al. Human embryonic stem cells: culture, differentiation, and
cell tumors. Gynecol. Oncol. 71, 14–18 (1998). genetic modification for regenerative medicine applications. Cancer J. 7 (suppl. 2),
55. Mostert, M. et al. Comparative genomic and in situ hybridization of germ cell tumors S83–S93 (2001).
of the infantile testis. Lab. Invest. 80, 1055–1064 (2000). 75. Kelly, O.G. et al. Cell-surface markers for the isolation of pancreatic cell types derived
56. Schneider, D.T. et al. Genetic analysis of childhood germ cell tumors with from human embryonic stem cells. Nat. Biotechnol. 29, 750–756 (2011).
comparative genomic hybridization. Klin. Padiatr. 213, 204–211 (2001). 76. Roche, S. et al. Comparative analysis of protein expression of three stem cell
57. Schneider, D.T. et al. Genetic analysis of mediastinal nonseminomatous germ cell populations: models of cytokine delivery system in vivo. Int. J. Pharm. advance
tumors in children and adolescents. Genes Chromosom. Cancer 34, 115–125 online publication, doi: 10.1016/j.ijpharm.2011.12.041 (20 January 2012).
(2002). 77. Chambers, I. et al. Functional expression cloning of Nanog, a pluripotency sustaining
58. Schneider, D.T. et al. Molecular genetic analysis of central nervous system germ factor in embryonic stem cells. Cell 113, 643–655 (2003).
cell tumors with comparative genomic hybridization. Mod. Pathol. 19, 864–873 78. Boyer, L.A. et al. Core transcriptional regulatory circuitry in human embryonic stem
(2006). cells. Cell 122, 947–956 (2005).
59. Voorhoeve, P.M. et al. A genetic screen implicates miRNA-372 and miRNA-373 as 79. Assou, S. et al. A meta-analysis of human embryonic stem cells transcriptome
oncogenes in testicular germ cell tumors. Cell 124, 1169–1181 (2006). integrated into a web-based expression atlas. Stem Cells 25, 961–973
60. Herszfeld, D. et al. CD30 is a survival factor and a biomarker for transformed human (2007).
pluripotent stem cells. Nat. Biotechnol. 24, 351–357 (2006). 80. Suh, M.R. et al. Human embryonic stem cells express a unique set of microRNAs.
61. Werbowetski-Ogilvie, T.E. et al. Characterization of human embryonic stem cells Dev. Biol. 270, 488–498 (2004).
with features of neoplastic progression. Nat. Biotechnol. 27, 91–97 (2009). 81. Morin, R.D. et al. Application of massively parallel sequencing to microRNA profiling
62. Gopalakrishna-Pillai, S. & Iverson, L.E. Astrocytes derived from trisomic human and discovery in human embryonic stem cells. Genome Res. 18, 610–621
embryonic stem cells express markers of astrocytic cancer cells and premalignant (2008).
stem-like progenitors. BMC Med. Genomics 3, 12 (2010). 82. Curado, M.P. et al. Cancer Incidence in Five Continents, Vol IX (IARC Scientific
63. Furukawa, S. et al. Yolk sac tumor but not seminoma or teratoma is associated with Publications No. 160, Lyon, France, 2007).
abnormal epigenetic reprogramming pathway and shows frequent hypermethylation 83. Sarasquete, M.E. et al. Minimal residual disease monitoring in multiple myeloma:
of various tumor suppressor genes. Cancer Sci. 100, 698–708 (2009). a comparison between allelic-specific oligonucleotide real-time quantitative
64. Sauvageau, M. & Sauvageau, G. Polycomb group genes: keeping stem cell activity polymerase chain reaction and flow cytometry. Haematologica 90, 1365–1372
in balance. PLoS Biol. 6, e113 (2008). (2005).
65. Ohm, J.E. et al. A stem cell-like chromatin pattern may predispose tumor suppressor 84. Mason, K.D. & Juneja, S. Go with the flow for monitoring response in myeloma
genes to DNA hypermethylation and heritable silencing. Nat. Genet. 39, 237–242 with minimal residual disease. Leuk. Lymphoma 49, 177–178 (2008).
(2007). 85. St. Ledger, K. et al. Analytical validation of a highly sensitive microparticle-based
© 2012 Nature America, Inc. All rights reserved.

66. Netto, G.J. et al. Global DNA hypomethylation in intratubular germ cell neoplasia immunoassay for the quantitation of IL-13 in human serum using the Erenna
and seminoma, but not in nonseminomatous male germ cell tumors. Mod. Pathol. immunoassay system. J. Immunol. Methods 350, 161–170 (2009).
21, 1337–1344 (2008). 86. Eriksson, L. et al. An investigation of sensitivity limits in PET scanners.
67. Wermann, H. et al. Global DNA methylation in fetal human germ cells and germ Nucl. Instrum. Methods Phys. Res. A 580, 836–842 (2007).
cell tumours: association with differentiation and cisplatin resistance. J. Pathol. 87. Schmittgen, T.D. et al. Real-time PCR quantification of precursor and mature
221, 433–442 (2010). microRNA. Methods 44, 31–38 (2008).
68. Minami, K. et al. DNMT3L is a novel marker and is essential for the growth of 88. Gillis, A.J. et al. Expression and interdependencies of pluripotency factors LIN28,
human embryonal carcinoma. Clin. Cancer Res. 16, 2751–2759 (2010). OCT3/4, NANOG and SOX2 in human testicular germ cells and tumours of the
69. Tang, C. et al. An antibody against SSEA-5 glycan on human pluripotent stem testis. Int. J. Androl. 34, e160–e174 (2011).
cells enables removal of teratoma-forming cells. Nat. Biotechnol. 29, 829–834 89. Li, J., Schachermeyer, S., Wang, Y., Yin, Y. & Zhong, W. Detection of microRNA by
(2011). fluorescence amplification based on cation-exchange in nanocrystals. Anal. Chem.
70. Visvader, J.E. & Lindeman, G.J. Cancer stem cells in solid tumours: accumulating 81, 9723–9729 (2009).
evidence and unresolved questions. Nat. Rev. Cancer 8, 755–768 (2008). 90. Zhou, W.J., Chen, Y. & Corn, R.M. Ultrasensitive microarray detection of short RNA
71. Lawrenz, B. et al. Highly sensitive biosafety model for stem-cell-derived grafts. sequences with enzymatically modified nanoparticles and surface plasmon resonance
Cytotherapy 6, 212–222 (2004). imaging measurements. Anal. Chem. 83, 3897–3902 (2011).
72. Crabbe, A. et al. Effects of MRI contrast agents on the stem cell phenotype. 91. Cao, F. et al. Noninvasive de novo imaging of human embryonic stem cell-derived
Cell Transplant. 19, 919–936 (2010). teratoma formation. Cancer Res. 69, 2709–2713 (2009).
73. Havranek, P. et al. Sacrococcygeal teratoma in Sweden: a 10-year national 92. Su, W. et al. Bioluminescence reporter gene imaging characterize human embryonic
retrospective study. J. Pediatr. Surg. 27, 1447–1450 (1992). stem cell-derived teratoma formation. J. Cell. Biochem. 112, 840–848 (2011).
npg

nature biotechnology  VOLUME 30  NUMBER 9  SEPTEMBER 2012 857


Articles

Multiplexed mass cytometry profiling of cellular states


perturbed by small-molecule regulators
Bernd Bodenmiller1,5,6, Eli R Zunder1,6, Rachel Finck1,6, Tiffany J Chen1–3, Erica S Savig1,4,
Robert V Bruggner1,2, Erin F Simonds1, Sean C Bendall1, Karen Sachs1, Peter O Krutzik1 & Garry P Nolan1

Mass cytometry facilitates high-dimensional, quantitative analysis of the effects of bioactive molecules on human samples at
single-cell resolution, but instruments process only one sample at a time. Here we describe mass-tag cellular barcoding (MCB),
which increases mass cytometry throughput by using n metal ion tags to multiplex up to 2n samples. We used seven tags to
multiplex an entire 96-well plate, and applied MCB to characterize human peripheral blood mononuclear cell (PBMC) signaling
© 2012 Nature America, Inc. All rights reserved.

dynamics and cell-to-cell communication, signaling variability between PBMCs from eight human donors, and the effects of 27
inhibitors on this system. For each inhibitor, we measured 14 phosphorylation sites in 14 PBMC types at 96 conditions, resulting
in 18,816 quantified phosphorylation levels from each multiplexed sample. This high-dimensional, systems-level inquiry allowed
analysis across cell-type and signaling space, reclassified inhibitors and revealed off-target effects. High-content, high-throughput
screening with MCB should be useful for drug discovery, preclinical testing and mechanistic investigation of human disease.

High-throughput in vitro screening has accelerated the discovery of providing a comprehensive view of the cellular state. Third, inter-
drug candidates, but paradoxically coincides with a steep decline in cellular communication and emergent systems properties should be
the approval rate for novel molecular entities1,2. The enormous rate evaluated. And, lastly, measurements should be performed with high
of attrition as drug candidates move through clinical development throughput. A screening approach with these features would enable
can be partly attributed to the disconnect between human physiology the identification of compounds with in vivo efficacy against the tar-
and the in vitro screening regimen, which cannot measure efficacy in geted disease and low toxicity at the earliest stage of drug discovery.
heterogeneous tissues or detect off-target toxicities2–4. If the original Some methods have attempted to implement these features. Parallel
screening regimen more closely reflected human physiology by enzymatic or phage display assays offer exceptional in vitro selectivity
using human samples, such as PBMCs or cancer biopsies, efficacy profiling13–17, but do not provide in vivo data. Cellular assays based
and toxicity could be identified earlier in the development process. on proliferation, apoptosis or expression of reporter proteins approxi-
High-dimensional analysis of cellular signaling networks can pro- mate in vivo activity18, but drug selectivity, mechanism of action and
vide a detailed representation of cellular state5,6; it is often presumed
npg

signaling network responses cannot be determined. Gene expression


that additional biologically informative measurements of markers of analysis19,20 and liquid chromatography coupled to tandem mass
pathways would be a desirable outcome for high-throughput screen- spectrometry6,21,22 measure thousands of parameters, but lack high
ing. Compounds that target certain signaling molecules can lead to throughput and single-cell resolution23,24. High-throughput micro-
successful therapeutic outcomes7, but many compounds that target scopy offers deep characterization of single cells23–25, but the limited
known oncogenic lesions lack clinical efficacy8. As such, the in vitro number of surface and signaling molecules measured simultaneously
targets of a drug candidate cannot be used to accurately predict effi- restricts the breadth of analysis.
cacy in vivo owing to signaling network complexity and differences Fluorescence-based flow cytometry (FBFC) permits measurement
between patients or cell subpopulations of the same patient 6,7,9–12. of up to 12 molecules on a single cell simultaneously26–28, allowing
Therefore, high-content, single-cell analysis of signaling networks cell subpopulations and their signaling network states to be deter-
in human samples during drug development could provide welcome mined at the same time29. Drug-screening applications for FBFC
insight into the manifold effects of drugs on cellular systems. have been implemented by hardware30,31 or by sample multiplexing
We propose that an ideal drug-screening approach should, first, with fluorescent cell barcoding (FCB)32,33. With these adaptations,
be based on primary human samples, with systemic behavior that FBFC has become a powerful tool for drug screening and preclinical
resembles normal physiology and the targeted disease state. Second, ana­lysis. FBFC falls short of the ideal drug screening method, how-
subpopulation-specific, system-wide signaling networks and their ever, because the number of simultaneously measured parameters is
correlation to cell and disease phenotypes should be quantified, limited owing to spectral overlap27, hampering the comprehensive

1Baxter Laboratory in Stem Cell Biology, Department of Microbiology and Immunology, Stanford University, Stanford, California, USA. 2Biomedical Informatics

Program, Stanford University, Stanford, California, USA. 3Department of Computer Science, Stanford University, Stanford, California, USA. 4Cancer Biology Program,
Stanford University, Stanford, California, USA. 5Present address: Institute of Molecular Life Sciences, University of Zurich, Zurich, Switzerland. 6These authors
contributed equally to this work. Correspondence should be addressed to G.P.N. (gnolan@stanford.edu).

Received 30 April; accepted 2 July; published online 19 August 2012; corrected online 23 August 2012; doi:10.1038/nbt.2317

858 VOLUME 30  NUMBER 9  SEPTEMBER 2012  nature biotechnology


Articles

Figure 1  Mass-tag cell barcoding. (a) Cells a O


C C
O O
C C
O

were covalently labeled with a bifunctional O N N O SH


SH
SH O N N O M
3+
3+
M
M
3+

compound, maleimido-mono-amide-DOTA M3+ + SH


M
3+
3+ 3+
O N N O SH O N N O M M 3+
SH M
(mDOTA). This compound can be loaded C O SH
SH
C O M
3+
M
3+
O HN O HN S
with a lanthanide(III) isotope ion, and reacts N HS N

covalently with cellular thiol groups through O O

the maleimide moiety. (b) Seven unique


lanthanide isotopes were used to generate b c 10 3
BC1+
128 combinations, enough to barcode each
sample in a 96-well plate. The seven lanthanide La

BC channel 1
BC1+
139 102 BC2+
isotopes, their masses and their locations on Well no. 94
Pr Barcoding Analyte
the 96-well plate are shown. (c) A density dot masses masses 101
141 105
plot of barcoded cells is shown with the y-axis
100
and x-axis plot showing barcoding (BC) channel 1 Nd
146 104 BC