Sept. 9, 2002 | One of the most seductive lures of the genomic revolution is the promise of personalized medicine. The rapid identification of tens of thousands of human genes and hundreds of thousands of DNA variations that might influence disease susceptibility has spawned a new field — pharmacogenomics. Dozens of companies have sprung up over the past few years, quantifying and cataloguing human genetic variation and using algorithms to tease out correlations among markers, genes, diseases, and drug response.
"Pharmacogenomics is the necessary beginning for the entry into personalized medicine," says Kari Stefansson, CEO of deCODE genetics in Reykjavik, Iceland.
But today, most pharmaceutical companies are more concerned with weeding potentially dangerous compounds out of their pipelines than with finding the ideal drug for the right patient. For some, this is a stunning disappointment, but for the industry as a whole, this diversion on the path to personalized medicine provides an exciting new way to combat the serious issue of toxicity-related drug failure — one of the industry's biggest problems.
Scattered throughout the human genome are millions of discrete, one-letter variations known as SNPs (single nucleotide polymorphisms). Most SNPs are benign, with absolutely no effect on gene structure or expression. But a subset of these variations provides crucial links to disease-causing genes, either because they directly alter a gene's activity or because they help pinpoint the location of such a disease-related gene.
Their abundance and facile identification make SNPs the new markers of choice for genetic studies, particularly for those seeking to unravel complex diseases like Type 2 diabetes, caused by the interplay of multiple genes and environmental factors. SNPs are also found in genes for drug-metabolizing enzymes, influencing individuals' ability to process a drug properly.
Many companies have compiled large collections of SNPs with a view to developing diagnostic and prognostic tests, as well as to guide the development of a new generation of drugs that would target genetically determined subsets of patients. Companies including Genset SA (recently acquired by Serono SA), DNAPrint
|Tools of the Trade: Advances in Genotyping
|One of the major problems confounding pharmacogenomics from the beginning has been the quality and quantity of the genotyping tools.
genomics, deCODE genetics, Genaissance Pharmaceuticals Inc., and Oxagen Ltd., as well as a few pharmaceutical giants such as GlaxoSmithKline and Novartis AG, invested heavily in the field. Dozens of tool companies also sprang up offering a range of technologies for SNP detection and genotyping (see "Tools of the Trade,"
Although the research, clinical, and investment communities loved the idea at first, pharmacogenomics is a prime example of how new biotechnologies are often far more complicated than expected (recall monoclonal antibodies and antisense, for example). Only in this case, people underestimated both the business and the scientific issues.
A Bumpy Ride
"The early story was, 'We'll find genetic variation that we can relate to drug effects, whether it is toxicity or efficacy, create mechanisms for assaying that, and license that out,'" recalls R. Mark Adams, vice president of bioinformatics at Variagenics Inc., based in Cambridge, Mass. "Finding SNPs wasn't the hard part. In fact, that proved to be easier than anyone anticipated."
In 1999, The SNP Consortium Ltd., made up of pharma companies and funding organizations, began identifying and publishing on the Web more than 1 million SNPs, thus preventing any pharmacogenomic company from claiming a monopoly. The result was an "early one-two punch," says Adams. "First the data became plentiful, then it wasn't even clear if you could patent it."
Another problem was the pharmaceutical industry's concern that by subdividing patient populations, pharmacogenomics would segment the multibillion-dollar markets that they depend on. "The industry is not sure whether this is friend or foe. Will it cause people to take more of their drug or less?" says William Evans, chairman of pharmaceutical sciences at St. Jude's Children's Hospital in Memphis, Tenn.
As a result, pharmacogenomics companies hoping for licensing arrangements found they were waiting much longer than anticipated. "They didn't see a value proposition, so it was much harder than we thought to get those Big Pharma deals," says Jay Mohr, Variagenics' new president and chief business officer.
"Like everyone else, we anticipated that the early demand for this would be much bigger than it turned out to be," says Richard Judson, senior vice president of informatics at New Haven, Conn.-based Genaissance.
As a result, many business plans have been hastily rewritten — some even scrapped. DNAPrint, for example, is developing forensic as well as pharmacogenomic tests, while Variagenics is focusing on their own
"The early story was, 'We'll find genetic variation that we can relate to drug effects, whether it is toxicity or efficacy, create mechanisms for assaying that, and license that out.'"
R. Mark Adams, Variagenics Inc.
oncology diagnostics. Genaissance is still pursuing its groundbreaking STRENGTH (Statin Response Examined by Genetic HAP Markers) study, seeking markers associated with the effects of therapy using the blockbuster cholesterol-lowering statin drugs. But the company is also diversifying, doing gene expression studies and marketing the HAP database and DecoGen informatics platform
Large pharmaceutical firms, meanwhile, are hedging their bets. Mohr cites figures from clinical research firm Covance Inc. estimating that 80 percent of Covance's large pharmaceutical clients are banking DNA samples from patients enrolled in clinical trials, even if they are not actually doing association studies. "Even companies like GlaxoSmithKline and Bristol-Myers Squibb, which are most heavily involved, are doing many fewer studies relative to the rest of the field," says Adams.
So what are the big companies doing with all of that DNA? And how are they actually using pharmacogenomics?
A prime concern of the pharmaceutical industry is avoiding drugs that fail in clinical trials or, even worse, after approval. "The cost of drug development is shaped like the outline of a brontosaurus," says venture capitalist Jerry Karabelas of Princeton, N.J.-based Care Capital. The tip of the dinosaur's tail marks the discovery starting point, while its lofty head represents the high cost of completing clinical trials. Once the drug is on the market, the human and financial toll of unforeseen side effects becomes even greater.
The lessons of several spectacular drug withdrawals, including Fen-Phen, Rezulin, and Baycol, have fueled demand for improved means to test prospective drugs for toxicity. The mantra "Fail fast, fail cheap" was probably first coined by Michael Pavia, former chief technology officer for Millennium Pharmaceuticals Inc., in describing how the company planned to secure its pipeline. Pavia's phrase quickly caught on as more genomics firms vied to demonstrate the quality of their drug selection process.
SNPs in action. At top, the DNA sequence of the gene for methyl guanine methyl transferase (MGMT), an important DNA repair enzyme. The start of the coding sequence is shown in bold (atg). Two non-synonymous SNPs that result in amino-acid substitutions in the MGMT protein, are highlighted in red (ctt to att; atc to gtc). Below, the resulting amino-acid changes in MGMT indicated as red amino acid residues: leucine to phenylalanine (position 84), and isoleucine to valine (position 143). Also shown are the respective frequencies of each allele. The light blue sphere shows the active site of the protein, close to the Ile-to-Val polymorphism.
Identifying better drug candidates to begin with is good, but what about the hundreds of candidate drugs already in development, many of which have years of research invested in them and could reach the market in just a few more years? Pharma companies are desperate to pick through those or improve their chances of success. Pharmacogenomics could help with both problems by opening an early window into a compound's effects on drug metabolism and toxicity.
For example, about 60 percent of all marketed drugs are broken down in the body by the cytochrome P450 family of enzymes. Individuals vary greatly in the efficiency of their P450 enzymes: Some people are poor metabolizers, while others metabolize very quickly. An individual's CYP450 profile could thus predict if he or she will experience side effects when given a particular drug or not respond at a particular dose.
Of the 50 or so pharmacogenomics-related new drug applications and investigative new drug applications received by the FDA in recent years, two-thirds involve screening patients for drug-metabolizing enzymes. Investigators want to find out early about potential population-specific toxicities and dosing requirements.
What about the banked DNA? The FDA may someday demand genotype information as part of the drug application process, and genotyping costs are likely to decrease significantly in the coming years. But there is another possible use for these samples. "You can run your trials, bank the DNA, and then not worry about it unless something comes up," explains Colin Dykes, a consultant and former research director at Variagenics. If a cluster of patients emerges with an unusual response to the drug, the DNA can be tested retrospectively, and perhaps a new trial initiated or the labeling adjusted. By genotyping only when necessary, companies avoid doing high-cost association studies.
Financial restraint isn't the only reason for the highly selective interest in genotyping. Some researchers think that it's also scientifically sensible. "Today, people are more realistic about how genetics will impact drug development," says Brian Spear, director of pharmacogenetics at Abbott Laboratories in Abbott Park, Ill. "Previously, the idea was that we would identify the right patient population, we would create new drugs, and we would sell them specifically for those patients. People have done enough work to determine how difficult it is and how infrequently this will happen. What people are learning is that there are a lot of ways to use genetics within pre-clinical and clinical development that are turning out to be quite useful."
Toxicogenomics, a fledgling offshoot of pharmacogenomics, is proving a lot easier for the industry to swallow. It doesn't break up blockbuster markets or require a complete redesign of the pharmaceutical business model. Instead, it's a fresh new tool that could help solve one of the industry's biggest headaches.
Companies like Gene Logic Inc., Phase-1 Molecular Toxicology Inc., and Iconix Pharmaceuticals Inc. are creating gene expression databases filled with signatures of toxic responses in humans or traditional animal models. Chip manufacturers, including Affymetrix Inc., Motorola Inc., and BD Biosciences Clontech, are marketing ADME (absorption, distribution, metabolism, elimination/excretion) and toxicology chips to screen for drug toxicity or determine if patients need a dose adjustment. Although the emphasis is currently on gene expression, SNP chips are also coming into use. (Screening tests based on protein and metabolite signatures are not far off.)
"We have chips with a whole range of genes, including some that are specific for toxic responses, such as genes in the liver and kidneys," says Phillip Stafford, group leader in statistical informatics at Motorola Life Sciences. The company has developed the CodeLink expression arrays including chips containing human, mouse, and rat genes. (The Codelink business line was recently acquired by Amersham Biosciences.) "Some drugs have a perfectly good response on the targets, but a latent response elsewhere," Stafford explains. "Expression chips can tell you if, two or three months down the road, you've got a latent toxic response in the kidney."
In the clinic, gene expression or SNP chips can be used to zero in on subpopulations that are at risk for either a bad response or no response. "Researchers at the Mayo Clinic used our P450 SNP chip to screen patients in a trial for a psychiatric drug. They had one patient who had an unusual response, and the chip indicated it was a metabolizing enzyme problem that was addressed by changing the dose," Stafford says. "People are using the chips in discovery, but also to decide which drugs will go into clinical trials."
Nothing to Fear but the FDA
Originally, pharmacogenomics was meant to encompass tests for drugs already on the market, new targets gleaned from studies of population subsets, and drug/diagnostic pairings for personalized prescriptions — therapeutics prescribed exclusively based on results from a particular diagnostic test. But linking drug and diagnostic development is tricky. No one wants to impede the launch of a drug because they are waiting for a partner diagnostic. Tying the two products together heightens the risk that nothing will be approved — at least not in time for the developer to profit fully.
"One reason people may think it will be difficult to get a test approved is because of what happened with Herceptin," said Lawrence Lesko, director of the FDA's office of clinical pharmacology and biopharmaceutics at the Center for Drug Evaluation and Research, in an interview with Bio·IT World in June (page 45). (Herceptin is the breast cancer drug from Genentech Inc. that is typically prescribed following a diagnostic test.) "That was the first time the agency ever looked at a drug and a diagnostic test together, and there may have been some rough edges. But we have learned a lot from that experience, and now, we expect some submissions will require co-review of a drug and kit." However, no one is rushing to be the second test case.
"We're very interested in how the FDA proceeds," says Variagenics' Mohr. "We are just waiting to see if they issue a guidance with a capital 'G' or a little 'g.'" Lesko told Bio·IT World that the agency is seeking to champion the field, launching a variety of print and educational initiatives.
"Interest in pharmacogenomics is increasing overall, primarily because the FDA is becoming interested," says Abbott Labs' Spear. "The FDA is concerned that there are identifiable populations that will respond to drugs in ways that are not currently addressed on the drug label."
Indeed, the FDA wants to know about genotype-specific effects, and it wants doctors to understand them, too. So, even if companies don't want to develop and market the diagnostic tests themselves, they may have to do the studies and fork over the information. "The role of genetics will become more prominent in the labeling of certain drugs," Spear says. "That means that there will be testing, even if no one mandates it."
If no test is available, it could be more difficult to get approval. "Once the FDA sees the technology is reliable and cheap, companies won't have a choice about it," says Michael Liebman, director of computational biology at the University of Pennsylvania and chief scientific officer for Philadelphia-based ProSanos Corp. "Companies that take a proactive position will be in a leadership role; otherwise, they'll have to respond reactively."
Labeling the drugs opens another can of worms. Perched at the tip of this information pyramid are doctors who need to interpret the test results without necessarily knowing every genetic variation associated with a drug response. "One of the big challenges for everyone is educating the physicians about the role of genetics with certain drugs," says Spear. But IT may offer a solution. "If the information is part of the label and, as a result, ends up in something like the physician's Palm Pilot, you may be able to make this kind of change even without education."
So there are signs that real pharmacogenomics could also catch on.
"There is no question you can find markers for drug response and they can be clinically useful," says consultant Dykes. "There are data
|St. Jude's Test Makes It Better
|After a pair of leukemia patients at St. Jude's Children's Hospital in Memphis, Tenn., reacted badly to the drug, William Evans decided to do something about it.
already available that would allow the development of molecular diagnostics tests." He anticipates a push to find markers for some drugs already on the market. One promising example of this is the St. Jude's TPMT (Thiopurine Methyltransferase) test (see "St. Jude's Test Makes It Better,"
More would help. "A big success story is always important, and that's why we are doing the STRENGTH trial," says Judson at Genaissance. The company is trying to find markers of efficacy for the highly successful cholesterol-lowering statin drugs. Despite enrolling only a few hundred patients, powerful associations have emerged from the trial. "The number of patients you need depends largely on the strength of the associations," says Judson. "Doing it this way wasn't a shot in the dark, but we could have been unlucky if the associations turned out to be much weaker than we expected."
"Today, commercializing and making pharmacogenomics real means developing molecular diagnostics," says Variagenics' Mohr. His company is betting on multiple lines of information, including gene expression, SNPs, and other factors like loss of heterozygosity. The early focus is colorectal cancer and developing a pipeline of diagnostics for the standard chemotherapies used to treat the disease.
Pharmacogenomic companies aren't the only ones looking into this. Compugen Ltd.'s Michal Preminger, vice president of new research directions, says the Tel Aviv, Israel-based company has several agreements to perform data mining and molecular analysis for firms whose "drugs are less successful than others," to determine if there are patient subpopulations who gain specific benefit from the drug.
The Rumblings of Giants
Meanwhile, companies that pay health-care bills — employers, HMOs, health insurers, and so on — are also understandably concerned about drug safety and utility, and some are starting to do something about it.
Orchid GeneShield and Merck-Medco (both based in New Jersey) recently announced a collaboration to identify genetic variations that predict response to asthma drugs. Interleukin Genetics Inc., based in Waltham, Mass., is teaming up with UnitedHealth Group's Center for Health Care Policy and Evaluation to study the influence of genetic variation in treatment of inflammatory diseases. Interleukin also has a deal with Kaiser Permanente's Center for Health Research to examine how genetic variation influences the risk of diabetes-related heart disease.
Compugen just announced several confidential agreements with health-care providers including one HMO, and more are expected (see July Bio·IT World, page 26). The bioinformatics company is mining the clinical information in these records and may add genotype data later.
"The reimbursers have a simple
|Answering the Billion-Dollar Question
|Before they start using SNPs to guide treatment, doctors will want to know that these tests are worth the necessary time and expense.
calculation to make," says Dykes. "We know how much we have to pay for this, and how often we pay it. If there is a test that cuts hospitalization in half, they can easily figure out what it's worth." Outcomes studies will therefore be necessary to encourage this trend (see "Answering the Billion-Dollar Question,"
right). Providers are one of the few groups that could power pharmacogenomics through to the clinic, so these studies will be pivotal for the future of the field. The Public Effort
The public initiative in this field continues to thrive. By next spring, the Human Genome Project will be complete. The public/private SNP Consortium has already exceeded its goals, putting more than 1.5 million SNPs into the public domain.
Francis Collins, director of the National Human Genome Research Institute, is championing the creation of a public haplotype map — a map including a critical mass of some 400,000 common SNPs to ensure that most genes are represented by a SNP either within its coding sequence or nearby. Such a map, costing an estimated $100 million, should accelerate association research by reducing the number of SNPs that have to be studied.
Several divisions of the National Institutes of Health (NIH) have already pledged about $32 million,
and international partners say they will contribute as well. But with the U.S. government earmarking resources for biodefense, Collins is turning for help from the pharmaceutical industry. "Pharma companies are interested in seeing it generated," Collins says, "but it has not been an easy year for them, given what has happened in the stock market."
For Collins, there's also the nagging question of "whether markers for drug response will diminish the markets for widely prescribed drugs." But Collins and his team are persevering. "If there is a shortfall in funding, we can just extend the timeline," says NIH Program Officer Lisa Brooks.
The "Hap Map" doesn't have to be comprehensive; it must merely have sufficient SNP coverage to provide good markers for the most useful genes. "Considering how huge our ignorance is now," Collins says, "if we found haplotypes across 85 to 90 percent of the genome, that would be an incredible treasure of information."
Collins is unimpressed by the hubbub that has shaken the industry lately. "In some quarters there was a misunderstanding, or naivete, about how having the sequence was going to solve everything. And there were some business models built solely upon the notion of quick profits, particularly selling subscription databases."
He dismisses talk about a foundering industry. "I think that every pharmaceutical company is still expecting that genomics will be the platform upon which they will build the next generation of drugs," says Collins. Others echo Collins' perspective. "We will change the treatment of cancer," says Variagenics' Adams. And there is no hint of doubt in his voice.
But there are clearly challenges ahead. "The sequencing of the human genome was a definable milestone that was very clear-cut, and that is hard to replicate with any of the other parts of this science that are necessary to understand the genome."
IN PART TWO: The impact of SNPs on disease gene identification, drug discovery, and informatics, plus a Q&A with Klaus Lindpaintner of Roche.
PHOTO CREDITS: PERSONALIZED MEDICINE PHOTOGRAPH BY ANN ELLIOTT CUTTING; R. MARK ADAMS BY FURNALD & GRAY