Variagenics: Lower-Risk Pharmacogenomics

Despite its presumed intuitive appeal, the concept of applying pharmacogenomics to predict drug response remains largely unproven, and the handful of companies trying for a foothold in this market face challenges on several fronts: finding drug companies who've bought into the concept and are willing to collaborate, negotiating deals without the leverage of issued patents linking SNPs and haplotypes to drug response, and convincing investors both that the technology won't be commoditized, and that pharmaceutical firms won't bring it in-house and leave them in the cold. Variagenics is assuming that drug companies with whom it partners will use data from clinical trials that incorporate pharmacogenomic assays to obtain a label linking the sale of the drug and the use of the tests, which Variagenics will supply or license. As it seeks partners, Variagenics is also raising its visibility by collaborating with CROs and offering expertise and services to those drug companies and biotechs who have yet to devote resources to in-house pharmacogenomics programs but who want a taste of the technology.

Variagenics is piggy-backing on the efforts of an array of players, including CROs and life-sciences instrument companies, to make its pharmacogenomics dreams into a commercial reality.

by Mark L. Ratner

• Variagenics wants to use pharmacogenomics tools with mid- to late-stage development candidates or drugs already on the market to better identify responders and select out from trials those patients at risk for serious side effects.

• Because it gathers the bulk of its data using known and presumptively "druggable" gene targets, the company believes its approach is lower risk than that of companies who are trying to uncover novel genes associated with disease.

• Variagenics is assuming that drug developers with whom it partners will use data from clinical trials that incorporate pharmacogenomic assays to obtain a label linking the sale of the drug and the use of the tests, which Varigenics will supply or license. However, before such FDA approvals and a demonstrated ability to command significant royalties on sales of drugs and diagnostics, the success of the strategy will remain problematic.

• Because most potential pharmaceutical partners have yet to work out the internal dynamics of developing diagnostics in tandem with their therapeutics, Variagenics knows adoption will take time. Thus, as it seeks such partners, Variagenics is also raising its visibility by collaborating with CROs and offering expertise and services to those drug companies and biotechs who have yet to devote resources to in-house pharmacogenomics programs but want a taste of the technology.

• Should it encounter setbacks in its partnering programs, Variagenics may yet decide that the fastest albeit higher-risk route to pharmacogenomics success is via in-licensing and rescuing a failing or failed drug.

Despite its presumed intuitive appeal, the concept of applying pharmacogenomics to predict drug response remains largely unproven. The handful of companies trying for a foothold in this market face challenges on several fronts: finding drug companies who've bought into the concept and are willing to collaborate, negotiating deals without the leverage of issued patents linking SNPs and haplotypes to drug response, and convincing investors that the technology won't be commoditized, that pharmaceutical firms won't bring it in-house and leave them in the cold, and that the business is therefore sustainable.

During the year 2000 biotech IPO boom, two companies, Variagenics Inc. and Genaissance Pharmaceuticals Inc. , went public with a pharmacogenomics model, raising $65 million [See Deal], and $83 million [See Deal], respectively. But others have already fallen by the wayside. Gemini Genomics PLC bought Eurona Medical in December 1999 for $5 million to get access to data and clinical samples from Swedish populations, then sharply curtailed Eurona's operations. [See Deal] PPGx Inc. , a pharmacogenomics service-oriented joint venture between PPD Inc. and Axys Pharmaceuticals Inc. [See Deal] that suffered from a lack of commitment and funding from its backers and never got off the ground, was acquired by DNA Sciences Inc. in December 2000. [See Deal]

Meanwhile, more established platform companies have been using pharmacogenomics for some time as a tool for discovering new gene-based targets, including the largest functional genomics players—Millennium Pharmaceuticals Inc. , Celera Genomics Group , a division of Applera Corp. , Gene Logic Inc. , and CuraGen Corp. , to name a few—as well as newer, more specialized platform companies, many of which also belong to the year 2000 IPO class, such as Sequenom Inc. [See Deal], Orchid BioSciences Inc. [See Deal], and specialist population genetics companies deCode genetics Inc. [See Deal] and Gemini Genomics [See Deal] (see Exhibit 1).

Variagenics and Genaissance, however, are unique in confining their research largely to the identification of variations in known genes between individuals that result in differences in drug response. They're aiming to predict optimal drug combinations, dosage, and duration of treatment; stratify patients for response; and select out of trials those patients at risk for side effects. "It's applied, pragmatic genomics," says Taylor Crouch, president and CEO of Variagenics, distinguishing its strategy from that of companies looking for new genes and markers of disease, on the basis that Variagenics' starting point is a gene target that is already presumptively correlated with disease and around which drugs have been or are currently being developed.

Drug developers have long been aware that pharmacogenomics tests could identify ultra- high or low-responders. But such testing was viewed in the context of safety or regulatory issues, Crouch points out, not as a marketing opportunity. "It's not new, but the philosophy around it has changed from safety to getting drugs to patients that otherwise wouldn't have them prescribed."

Reach Out

One of Crouch's favorite slides depicts a Venn diagram of the stakeholders in pharmacogenomics: drug firms, diagnostics companies, clinical research organizations, and life-sciences instrumentation companies, with pharmacogenomics at the interlocking center of those four core constituencies. The slide then builds to show another layer of stakeholders orbiting the core to whom Crouch must also attend: patients, the public, health care providers, payors, regulators, and academia.

Biotech companies have often used similar slides to show the potent reach of a nascent technology. But Variagenics' early outreach to so many of these groups is unique, and it has implemented a diversified partnering strategy. As it waits for drug companies to determine the structure of their pharmacogenomics initiatives, and for patents to issue on what it believes will be high-margin tests that drug companies will want to link to their therapeutics, Variagenics is selling expertise.

Instead of offering a subscription to its database and informatics system and having the user do the work, Variagenics is offering to drug companies the capability to identify haplotypes (marker sets comprising patterns of SNPs [single-nucleotide polymorphisms] that are predictive of drug response), validate the haplotypes, and rapidly turn that information into clinical assays, as a way of showcasing its capabilities, while at the same time it retains control over the IP derived from data gathered while working with the partner. Once patents have issued, the company expects to then be able to use its IP to negotiate royalties on sales of drugs whose label links its use to a test derived from the proprietary marker.

In addition to partnering with pharmaceutical companies, Variagenics has also established relationships with CROs Covance Inc. (assay co-development and sale for use in clinical trials) and Quintiles Transnational Corp. (preferred provider co-marketing arrangement), and instrument suppliers Waters Corp. (reagent supply) and Bruker Daltonics Inc. (mass spectrometry systems for use with Variagenics' DNA testing and analysis platform).

Crouch didn't bring this broad vision of its mission to Variagenics, however. Before he joined the company in April 1999, its board had already realized that adoption of pharmacogenomics would take several years, and that the best way to sustain a business would be to develop products—clinically validated assays and a DNA testing and analysis platform—while maintaining its emphasis on selling pharmacogenomics information and services as a tool for drug development.

To accomplish that second goal—selling pharmacogenomics programs—meant making deals with the development side of pharmaceutical companies, which in turn meant getting a buy-in from their clinical, medical, and marketing functions, executives who are far less familiar with technology deal making than discovery managers. The Variagenics board therefore wanted its management team to be familiar with the perspective and priorities of drug development.

So it turned to Crouch, who joined Variagenics from Parexel International Corp. , where he was SVP, worldwide marketing and strategic development. Prior to Parexel, he'd also spent ten years in the pharmaceutical and medical device industries in new product development marketing, including a stint at Schering-Plough Corp. , where he reported to Anthony Wild, PhD, later president of Warner-Lambert's global pharmaceuticals sector, who recently joined Variagenics' board of directors. While at Schering, Wild recalls, Crouch's thinking often ranged beyond marketing. "Taylor always thought in terms of the clinical profile, not just the promotional profile," he says.

By all accounts, Crouch genuinely believes in pharmacogenomics and that building bridges across clinical, regulatory, and marketing functions can only improve the odds of everyone's success. And along with Crouch, almost all of the company's senior executives have a range of experience in the CRO industry as well as pharmaceutical or diagnostics development.

Slow Pitch

In the last 3-4 years, the pharmaceutical industry has come to understand the potential of pharmacogenomics technology, which compares genetic variations, base pair by base pair, between individuals who have been treated with a drug, and catalogs the patterns of these differences to determine which patients will be better responders and which ones are at high risk for developing serious side effects. More broadly, pharmacogenomics tools are also being developed as prognostics of disease propensity—part of Genaissance's approach, for example. The challenge has been finding a sustainable business model in the face of an understandably slow technology integration process within the pharmaceutical industry, and an even slower, if equally understandable, rate of adoption in diagnostics.

Nor is the power of the technology assumed, and drug companies understandably won't spend on the technology until given a reason to do so. Validation could come via prospective clinical studies that show a correlation between sets of markers and drug response, or a pronouncement from regulators that efficacy data based on use of a pharmacogenomic test will speed or improve the chances of a drug's approval. But none of these events has happened yet. In fact, the first prospective clinical trials (where informed consent is secured up front and tests are run on patient samples during the trial, thus providing the quality of information likely to be accepted by regulatory agencies) are only now beginning. And only a handful of peer-reviewed publications exist that look at pharmacogenomics and drug response retrospectively; that is, by analyzing data and patient samples after the trial has been completed.

One paper widely cited is Genaissance's September 2000 paper in the Proceedings of the National Academy of Sciencesshowing the predictive value of haplotypes in the beta-2 adrenergic receptor associated with use of the beta agonist, albuterol. But there's little beyond that. A published study of Abbott Laboratories Inc. 's asthma drug, zileuton, a 5-lipooxygenase inhibitor, showed that all patients with a particular form of the 5-LO promoter had no response to the drug (these patients comprise 15% of all non-responders). Other studies have shown variances that predict response to leukotriene inhibitors and different drug effects due to variances in the thiopurine methyltransferase (TPMT) gene, which metabolizes the leukemia drugs 6-mercaptopurine and azothioprine.

"The only other data, which are not peer reviewed, are based on our statin trial," notes Genaissance EVP and CFO, Kevin Rakin. In that study, the company showed in a small cohort that a particular haplotype difference could be used in a cost effective way to choose among three statin compounds for improved lowering of LDL cholesterol. Rakin, however, also acknowledges the difficulty of making the case absent peer-reviewed data showing that pharmacogenomics works, as well as the frustration at the slow pace of validation of the technology.

On the other hand, "in the context of the typical curve of deal flow for early-stage technology, I doubt we're very different," Rakin suggests. "You've got to look a the cycle within pharma: ten companies put together the SNP Consortium, they've hired in pharmacogenomics at the VP and director level and are setting budgets and strategy, and three companies have deals. The pieces are being put into place. Pharmacogenomics is following the pattern of other technologies that have become an integral part of the pharma environment."

Both Genaissance and Variagenics believe that validation will come from company-sponsored proof of concept studies, and anticipate a drug approval tied to a diagnostic in 2-3 years. Variagenics is establishing a number of proof of concept trials with academic groups and companies in respiratory diseases, cardiovascular diseases, and oncology using its pharmacogenomics markers either to select for an enriched patient population or to reduce side effects. It expects to have results in the second quarter from one study, a retrospective analysis conducted in conjunction with a Boston-area academic group, of the ability of a proprietary marker set to select from among several statins the one that best lowers patients' cholesterol levels. A second study, in biphasic disorder through an academic network coordinated by a Boston-area psychiatry group, will see if a Variagenics marker set can identify the subset of patients whose weight will balloon (some patients add up to 20 pounds in one month following the beginning of therapy, other gain 10-15 pounds in a year) and to analyze which patients respond to traditional neuroleptics versus novel drugs in the class. A third study will use its markers for the colon cancer drug 5-FU, to determine whether that drug should be used alone (40-50% of colon cancer patients will respond to a single agent) or as part of a cocktail.

For its part, Genaissance expects to start a prospective trial of statins in February, and has plans over all to start trials encompassing 19 drugs in five disease areas this year.

If one is looking for validation from the large drug firms, the strongest advocate for pharmacogenomics has been GlaxoSmithKline PLC , which now has an internal genotyping unit operating under Good Laboratory Practices and a policy that every clinical trial in progress or proposed go through a pharmacogenomics evaluation process to prioritize the need for such tests. "The inevitable consequence of this effort is the use of diagnostics in tandem with a drug," states Colin Dykes, PhD, VP, research and genomics at Variagenics and formerly director of Glaxo's UK genetics division. Glaxo has also formed a predictive medicine division.

Industry executives say pharmacogenomics programs at both Glaxo and SmithKline Beecham (which recently merged with Glaxo [See Deal]) separately received a boost when they validated pharmacogenomic effects in a trial. But the companies are unlikely to disclose or publish proof of concept results. Moreover, some companies remain reluctant to talk about pharmacogenomics because the notion exists that it is mostly used in clinical trials to uncover the reason underlying a negative occurrence (side effects or low response rates). On the other hand, Taylor Crouch says just as many companies now look at pharmacogenomics as a potential strategic marketing tool, and may be reticent to reveal their planning for that reason.

Along with Glaxo, Roche and Pfizer Inc. also openly talk about their pharmacogenomics programs. "They wouldn't be doing that if they were worried about a negative association," Crouch says. Plus, he notes, several large companies have purchased genotyping platforms. Nonetheless, the kind of details that will truly validate the use of pharmacogenomics will likely be revealed only after an NDA submission which links use of the drug and an assay, or if the FDA makes stronger statements in support of the technology before that.

The latter is unlikely, given the agency has already said that if pharmacogenomics data support a NDA and some patient populations would suffer adverse effects, a drug would have a better chance of approval. Where FDA has been silent is on the efficacy side—the area of greatest interest to drug marketers. And given the agency's reputation for being more reactionary than proactive, Crouch is among the few that anticipate additional proactive FDA proclamations. His rationale is that "genomics keeps getting into the public discourse," and FDA representatives are always an invited presence at meetings centered on ethical and policy issues. "It doesn't evoke an ethical debate they way that stem cell research does, but pharmacogenomics does raise the issue of whether we're approving drugs with one eye closed, and whether we can do more."

Adoption by Big Pharma has also been slowed by the fact that pharmacogenomics spans both discovery and development. In discovery, companies may be using pharmacogenomics to search for genes and genetic markers associated with disease. The starting point is usually the identification of SNPs and haplotypes, which researchers can now look for by screening DNA from blood samples using high-throughput methodologies and looking for variations at the DNA level. Coupled with patient histories, or using selected populations (homogeneous ethnic groups, sets of twins, or others with shared risk factors), they hope to then make associations between the mapped SNPs and haplotypes, and disease.

To succeed in making a discovery deal usually means finding a champion within research. But when it comes to using pharmacogenomics to predict drug response, implementation must come from the development side of an organization, not discovery. And many of the newly designated VPs and directors of pharmacogenomics, who are scientists out of research, are just now learning the intricacies of building consensus among clinical, regulatory, and marketing functions, which often operate within tight budgets and timelines to approval, as opposed to a more open-ended discovery timetable.

At large drug companies, because of the reference to genomics, pharmacogenomics became embedded in research, Crouch points out, which has slowed implementation. "There's been little discussion of medical need and the payor decision-making process," he says. "Pharmacogenomics needs to move out of discovery and into development. In fact, I believe it really belongs in marketing. When I was at Schering-Plough, for example, marketing was a key driver of drug development."

Further muddying the waters is the fact that pharmacogenomics trials require the testing of drugs and diagnostics in tandem. Because of the differences in their competitive advantages—margins, intellectual property, stand-alone value, distribution channels, and durability of the franchise—drug developers have not been interested in, and know little about, diagnostics development. Even if they see value in a diagnostics discovery, they lack the machinery for development and an understanding of the different regulatory issues, including how to handle data. And because of diagnostics' lower margins, they won't divert focus from the drug development opportunity until the potential for a bigger payback from diagnostics is established.

But progress is being made, says Taylor Crouch, as evidenced by the fact that, in last couple of months, more than a half dozen of the biggest pharma and biotech companies he'd previously contacted, having thought through pharmacogenomics in-house, have come back to Variagenics to talk about collaborating. One company came back to Variagenics six months ago with a drug for ADHD that has a thousand-fold dose-response effect, to evaluate whether to consider launching it in tandem with a test. "They're still struggling with the issue because talking about higher dosages in children requires a lot of physician education."

Because of the anticipated slow adoption curve by its core client base, Variagenics early on decided to sell product—assays and systems—as well as information, to reduce its risk. But the company had to evolve to reach that point in its decision-making. Variagenics was founded in 1992 as KO Technologies Inc. by the Massachusetts Institute of Technology 's David Housman, PhD, to identify cancer drug targets based on loss of heterozygosity (LOH, the deterioration of chromosomes in cancer cells), mostly in genes involved in cellular metabolism, as opposed to traditional oncogenes. However, despite the scientific appeal of applying LOH analysis, high-throughput tools for comparing heterozygous versus homozygous genes did not yet exist. Eventually, the company's CEO, Fred Ledley, MD, pointed Variagenics towards a higher-throughput means for validating a variety of disease markers—not just cancer targets—which brought it into the accumulation and evaluation of data on SNPs, a key pharmacogenomics tool.

The company's move away from cancer drug discovery and into pharmacogenomics accelerated with its acquisition of Avitech Diagnostics Inc. in September 1997 [See Deal]. Ledley had heard about Avitech's enzyme mutation detection (EMD) technology, which was faster than using single-stranded complementary polymorphism, the industry standard at the time. Having sold Avitech's life sciences business to FMC Bioproducts (now the BioWhittaker Molecular Applications division of Cambrex Corp. ) [See Deal], and later, the life-sciences portion of EMD to Amersham Pharmacia Biotech Ltd., a division of Nycomed Amersham PLC [See Deal], the company's board had decided to liquidate it.

Ledley convinced Avitech's Anne Bailey to join Variagenics to design a system to run pharmacogenomics using EMD, which the company used until about 16 months ago, when it fully converted to ABI 3700sequencers, once that technology finally began to outpace its prior platform. Bailey, now VP, diagnostic and process development, points out that "We've been proactive in adopting best-in-class technologies, both off the shelf and internally developed. But in all cases we have made significant improvements in acquired technologies to streamline and integrate high-throughput SNP discovery processes."

Although Ledley was the visionary—he also saw the opportunity to use CROs to pave the way to pharma—he lacked drug development experience, and the Variagenics board did not believe he was the best point person to deal with CROs, or pharma companies, for that matter, and replaced him with Crouch. (Ledley did, however, manage to raise $12 million in a critical venture round in July 1997 [See Deal], a time when pharmacogenomics was not yet on the radar screen as an important technology.)

The Variagenics Solution

The side of Variagenics' current business that most people think of as pharmacogenomics is its VIP(Variagenics Impact Program) partnering effort with pharmaceutical companies, under which the company applies its suite of technologies to identify markers of drug response, usually starting with drugs in Phase II trials. The goal of the program is, on the pharma side, to develop the cleanest drug possible by identifying patients at risk for serious side effects and to reduce the size of clinical trials and improve efficacy by prospectively identifying non-responders. For Variagenics, it's a way to get access to clinical samples and patient histories, build its IP portfolio, fund assay development, receive milestones for better enabling a drug's development and, ultimately, get a royalty on drug sales if the drug's use is tied to use of a molecular diagnostic under Variagenics' patents. Its pitch to partners is the ability to develop markers that correlate with drug response, and to rapidly develop assays. To accomplish this, Variagenics uses five discovery tools that enable it to identify the critical genetic information relevant to a drug's activity: pathway analysis, gene expression profiling, haplotype analysis, molecular modeling, and LOH analysis.

It currently has four VIPdeals: with Amgen Inc. , just signed in January 2001 [See Deal], Boehringer Ingelheim Pharmaceuticals Inc. [See Deal], and two undisclosed partners—a top-tier pharma and a biotech, according to the company. It also expects to enter into 6-8 VIP partnerships in 2001, which may include expansion to new products with existing clients as well as penetration into new companies. In three of its four VIP partnerships, Variagenics expects to look at clinical outcomes data this year in trials where it has already obtained consent.

The VIPdeal structures are similar, with fully loaded funding as a minimum commitment from the partner, presumably leading to more significant development milestones, and royalties, which will be negotiated as programs hit go/no-go milestones. Variagenics holds on to its intellectual property, which centers around proprietary markers that, once clinically validated, presumably will be the basis for proprietary molecular diagnostic products it will license to diagnostics companies.

That none of its four VIPpartnerships includes guaranteed royalties could be seen as a lack of commitment by the partner. In two cases there were extensive discussions on the subject, and the agreement was signed without agreement on a rate: it was more important to the company, Crouch says, that the collaboration get going and its partner get a taste of what it can do.

"We offer a flexible partnering approach that puts validation ahead of longer-term economic negotiations," he asserts. To preserve the opportunity to receive royalties from several companies with drugs in the same class that could benefit from the same test, Crouch also doesn't want to license rights to a test to anyone exclusively. And each VIPagreement reiterates that Variagenics owns the markers.

Crouch is confident enough in the eventual value of the company's intellectual property to hold out for low single-digit royalties. "I think that from the drug company's perspective, enablement would be worth that much. In fact, our board member, Tony Wild, said his former company, Warner-Lambert, would have been willing to pay up to 12 percent for that kind of enablement." But Crouch also acknowledges that it's unrealistic to expect any kind of substantial royalty absent proof of concept.

Moreover, by breaking programs into discrete go/no-go segments, VIPs are intentionally structured to have modest entry investments by the partner. In some cases Variagenics will already have looked at a pathway and the effect of a drug on it in patients with different characteristics (physical characteristics, ethnicity, risk factors, or environmental influences, for example). It often does this on its own dime prior to approaching a pharmaceutical company, using publicly available and proprietary information about the target and genes around the target.

In the first step of a VIP, Variagenics prepares a report on genetic variance discovery, either drawing on data already in its proprietary SNP database or by newly sequencing genes. (The intellectual property on relevant SNPs or haplotypes would have already been filed by Variagenics.) Partners reimburse its fully burdened costs, including the costs for finding those genes already in the database that have been tied to response to the partner's drug. Such funding typically ranges from several hundred thousand to over one million dollars, depending on the scope and complexity of the pathway and the degree of characterization required.

Once a partner buys into Variagenics' target-based theory of why some patients suffered side effects or didn't respond to the partner's drug, it provides Variagenics access to patient samples to prove the hypothesis. This transfer begins the second phase of a VIP—determining which SNPs and haplotypes will be most predictive of drug response. "It's this ‘winnowing' part of the process—starting with, for example, 15 genes and associated SNPs and getting down to a high probability of obtaining a result that will lead to a prescribing action, then developing an assay to test it—that adds the most value for our partners," notes Variagenics' CFO, Rick Shea. The partners will then take the assay into clinical trials, usually in Phase II. "For now, most of our customers tend to want to use pharmacogenomics retrospectively, with the expectation of prospective genotyping as part of a Phase III pivotal trial," he adds. That's when the milestone payments would kick in: for proceeding to Phase III with an assay, on FDA filing for approval by the partner, with labeling that includes use of the Variagenics diagnostic, on approval, and when the drug reaches specified sales levels.

Variagenics has also established a co-marketing, preferred provider arrangement with Quintiles under which Quintiles' worldwide business development group will incorporate Variagenics' SNP discovery and clinical design services into its selling cycle. But that arrangement has yet to bear fruit, Crouch acknowledges. "It's not tangible to them, yet," he says. "The collaboration should heat up as more trials are conducted using pharmacogenomics."

That will take some time because pharmacogenomics isn't widely used in clinical trials. In the past two years, drug companies have lived through the investor and biotech hype and a period of "kicking the tires," according to Crouch, and are just now reaching the point where they have anticipated pharmacogenomics in their clinical protocols and obtained patient consent and collected blood samples. "The reality is very few genotyping platforms are in place," to enable economically feasible patient testing: the cost of identifying SNPs, while declining, is still much too high for the clinical setting (see "Pharmacogenomics Reality Check,"START-UP, March 2000 [A#2000900048). And while some experiments using assays developed in academia could be going on below the industry's radar screen, it has not reached the level of full-scale testing under the auspices of GCP, he suggests.

For the most part, the assays drug companies use in clinical trials test for drug metabolizing enzymes such as the CYP family—the kind of diagnostics that have been conducted in some form for decades, before traditional pharmacogenetics testing morphed into pharmacogenomics. And these are not validated assays suitable for FDA filings, although industry watchers say some companies, notably Glaxo and Bristol-Myers Squibb Co. , are developing validated assays in-house. However, for use in a prospective clinical trial to collect "FDA-able" data, companies may want to outsource development of such data to a GLP lab, notes Anne Bailey.

Who Wants to Partner?

Because the largest companies have already designated in-house groups to shepherd pharmacogenomics technology internally, they are less likely to turn to collaborators at an early stage, Crouch believes. Mid-sized companies, on the other hand, are more likely to turn to outside advisors to help them use pharmacogenomics to identify and validate marker sets using those companies' banked samples, which plays to the service side of Variagenics' business. And Crouch puts large, product-driven biotechs in that same category. "We would not have expected a year ago the enthusiasm we're seeing from biotechs," he says. In the case of Amgen, for example, Variagenics had come in to talk with them generally and discovered they had a drug that could potentially benefit from Variagenics' discovery approach—in an area where Variagenics had already invested in its own research. Although the drug hasn't been disclosed, it's in clinical testing for a chronic indication, as are all of the VIPs signed to date.

With the Quintiles co-marketing deal along with its own business development effort, the company next turned to the laboratory testing side of the pharmacogenomics business, both as another route into pharma and as a revenue source, through licensing its assays. Quintiles passed on participating in the lab testing opportunity, opening the door for Covance, the largest lab analyzer of patient blood samples for clinical trials worldwide.

Covance gives Variagenics an outlet with established infrastructure it can tap into. "We want to establish proprietary content and technology and plug that in to industries that are already operating in that space," Crouch explains. "For the pharmacogenomics industry as a whole, we're talking about changing 350 drugs that will go into clinical research this year, and thousands more as throughput increases. That requires a process you can scale up, to address pharmacogenomics on a global scale."

In 1999, Variagenics partnered with Covance [See Deal]. It is developing assays, for which Covance reimburses Variagenics on an FTE basis for the ones it selects. Covance will then use the assays in the trials it runs for its clients, paying Variagenics a royalty on assay sales. Although it is obligated to supply only Covance within the CRO market, Variagenics retains the right to license a VIPpartner on a non-exclusive basis, and the partner would be free to contract with another lab.

Two years ago, when it was looking for a pharmacogenomics partner, Covance looked at more than 20 companies before selecting Variagenics. "We wanted a technology attuned to clinical trials and drug development to build into Phase II and III trials, not diagnostic tests," explains Bill Campbell, PhD, director, genomics, Covance Central Laboratories Services, "and a company interested in discovery, not just service. Variagenics showed the ability to rapidly respond to customized test development with a platform we could easily transfer to our own."

While the funding level from Covance may be modest in biotech deal terms—amounting to less than $1 million annually, according to Variagenics' Rick Shea—it represents an unusually large commitment to R&D for a CRO. "Covance may have over $800 million of top-line revenues," Taylor Crouch says, "but it is very judicious with its R&D expenditures, and this collaboration represents a substantial financial commitment from them." To date, Variagenics has delivered 33 assays, many of them assays for metabolizing enzymes that Covance offers to a range of its clients.

"We've never had a partnership like this," says Campbell. "The program with Variagenics is strategic and is the largest investment that Central Labs has ever agreed to. We've transferred technology in-house assay by assay, but never in a broad program."

The partnership works more or less in the following manner: customers come to Covance with a drug and ask what tests should be performed. Variagenics will look at the drug and its presumed target, suggest 5-10 or so candidate genes and design tests for them, which they transfer to Covance. The process usually takes 2-3 months, depending on the number of genes known and the number of genetic variations that may be predictive. The development time also depends on the ability to run multiple tests at the same time on the same sample (multiplexing), which becomes more important as the number of genes increases.

The most widely requested tests among those already delivered by Variagenics are for the CYP450 family of drug metabolizing genes; in particular 2D6, 2C9, and 2C19, according to Campbell. Clients are also requesting ApoE genotyping for Alzheimer's patients, he says. (Variagenics has an issued patent, licensed from now-disbanded Nova Molecular Inc. [See Deal], claiming a method for treating patients with neurodegenerative disorders by predicting response using its ApoE marker. It also holds patents on the TPMT gene and the methylene tetrahydrofolate reductase [MTHFR] gene, for cardiovascular and cancer pharmacogenomics applications.)

Because pharmacogenomics is still an emerging technology, Covance hasn't yet tried to market its capabilities extensively. But it wants to do so in 2001, Campbell says, and began stepping up its DNA collection efforts in preparation. "Our DNA banking grew ten-fold last year. In the middle of last year we began an aggressive sales and marketing endeavor to ask clients to consider the benefits of collecting DNA. But many of them don't perceive the benefits or know whether it'll be of benefit to a particular drug." Nonetheless, Covance now has DNA from 16 of 20 drug company clients, some collected by policy for every patient across all trials, others across a therapeutic area, some for a particular drug in specific trials. And while only a small number of clients are actually doing DNA testing, four times as many were testing in the second half of 2000 as did in the first half of the year. Taylor Crouch adds that three years ago, only one in ten drug companies were banking DNA in clinical trials.

"Talking to pharma companies confirms that many of them have made large investments in pharmacogenomics internally," Campbell notes, and are developing research tests internally and through consortia. "Eventually," he believes, "they will outsource their tests to be clinically validated." Working through a CRO that is already collecting samples as a mainstay of its business is appealing to the pharmaceutical industry, he says, because investigators won't be asked to divide samples and ship them to various places. Plus, Covance is already working under GLP conditions.

The third component of Variagenics' revenues will come from sales of DNA testing and analysis systems for pharmacogenomics that incorporate its NuCleavetechnology, including the sale of reagent kits, which have been developed by Variagenics and are being sold by Waters [See Deal]. The company designed the NuCleave assay with clinical diagnostic applications in mind, where accuracy and reagent cost are critical and measuring large numbers of SNPs at the same time is a lesser consideration (see Exhibit 2).

The core of the NuCleavesystem is a mass spectrometer instrument drop-shipped by its manufacturer, Bruker—a Biflex III MALDI-TOF, the same instrument used by Variagenics' competitor, Sequenom, in that company's industrial genomics applications. Hamilton Instruments supplies robotics for sample handling, Variagenics provides informatics software, and Waters, Variagenics' other licensee, makes reagent kits. Variagenics announced the commercial availability of NuCleave in January 2001.

"NuCleaveintroduces extra internal quality control parameters" to clinical genotyping, explains Anne Bailey, and reduces reagent costs as compared to other mass spec systems. It is a chemical cleavage technology developed partially with Harvard University that modifies bases in amplified gene fragments at specific locations (for example, replacing each A base on the strand with a modified, chemically cleavable A), allowing for genotype identification based on both the length and mass of the fragments, and confirmation based on the mass of the internal fragments of the complementary strand.

The Waters relationship grew out of Variagenics' desire to license a sample purification process it had used and improved for use on the back end of the NuCleaveprocess. For its part, Waters had been targeting the genotyping and SNP analysis markets to leverage its mass spec product offerings. The licensing discussions eventually broadened, and in June 2000 the parties established a strategic alliance combining NuCleave and the purification process for mass spec applications. Waters bought $7.5 million of Variagenics stock in its July 2000 IPO [See Deal] and paid a $3 million up-front technology transfer payment. Variagenics may receive up to an additional $4 million in development-based and sales milestones, plus royalties on kit sales.

"We're not expecting high initial demand for systems," Shea points out, which will cost $350,000-500,000. "Our interest is to get one or more systems into Covance, and to initially sell to research institutions as well as pharmaceutical companies." In January 2001, it placed its first NuCleavesystem, at Covance Central Laboratories Services. "It's important to us that Covance gets comfortable with NuCleave," Shea says, because it is not used to making large, up-front bets on equipment systems. The recent slowdown in Phase III trials outsourced to CROs also makes it a difficult time for Covance's management to introduce new technology in-house.

On the other hand, while Shea, from his CFO's perspective, may be reluctant to factor significant revenues from systems sales into Variagenics' business model, sell-side analysts and others at the company say they expect NuCleaveproduct sales to be a considerable source of income. And Variagenics could leverage the value of the NuCleave platform by developing it as a high-throughput discovery research tool, presumably with a partner.

Diagnostics on the Horizon

Variagenics is comfortable with analysts' 2001 revenue projections of $7-8 million, almost all from existing and assumed new collaborations. "Our challenge comes in 2002," Shea suggests, when Wall Street projects revenues to more than double over 2001, mostly based on progress in the VIPprograms. "That'll require moving downstream in our VIP programs to the milestone stage, and/or broadening into molecular diagnostics."

Indeed, it is the molecular diagnostics opportunity and the prospect of high-margin products that tantalizes Taylor Crouch. He acknowledges that Variagenics' first deal licensing to a diagnostics company the rights to develop clinical molecular diagnostics to run on its platform is probably at least a year away. But he's focused on the opportunity, for which he thinks Variagenics could command royalties in the 15-25% range, or even profit sharing.

"Diagnostics is validated, it's global and operationally oriented and well placed to take advantage of pharmacogenomics when it comes out," Crouch observes. But to capture that opportunity, he believes, diagnostics developers need to be earlier investors in new technologies and not sit back, as they have historically, and wait for a product that is a year or less from the market to present itself. "They have to plug into the engine," he says.

Anne Bailey agrees. "It will no longer be a matter of saying "put it in the box, call us when it's ready, and we'll go sell it for you," as is the current practice in many companies. "But for them to commit to a development program early in a Phase III clinical trial will be a tremendous change. Senior management must move away from the commodity mindset and work, as pharma companies do, with payors to show validation."

Although diagnostics companies don't generate a lot of free cash to pour into R&D, Crouch thinks genomics could change that, just as it has enabled instrumentation companies to reinvent themselves as life-sciences companies, gaining access to both cash and equity. "Cash isn't the issue, P&L funding is," he suggests. "In that sense, they face the same issue as pharma companies: how to invest P&L dollars to fuel growth while the pipeline is weak."

"Diagnostics companies are receptive to pharmacogenomics," adds Nick Dracopoli, PhD, executive director, pharmacogenomics, for Bristol-Myers, which a little over a year ago entered into a collaboration with Millennium subsidiary Millennium Predictive Medicine Inc. to identify and validate pharmacogenomic markers in cancer. [See Deal] "We're in negotiations with several diagnostics providers to find a common technology platform to utilize our markers." They're receptive, he says, but since such a deal entails more technology development risk than diagnostics companies are used to, the markers must be amenable to multiplexing for the deal to be commercially viable. Market size is also an issue for them, he points out. "No one marker or set of markers for a given drug will be enough to drive the market." But Dracopoli is convinced of the potential for the technology. "It's already happening, in a stepwise way. The technology is now in place to assess a small number of markers, and we're moving toward a multiplex approach."

Intellectual property will also be key to attracting diagnostics partners, he believes. "Companies need some level of IP to develop a diagnostic, or we'll end up with a lot of home brews that will not drive the market forward. On the other hand, the more markers you add to the mix, the more complicated the IP becomes."

At Bristol-Myers, sample and tissue banking, with appropriate consent, is built into the clinical trials infrastructure, but is not routinely done. "We won't consider it routine until it's written into every clinical protocol," Dracopoli suggests. "Now it's usually an amendment to a protocol." He shares the majority view that, in the end, speed of acceptance and widespread incorporation will be dictated by data. "No doubt, the technology will be reduced to practice," he says. But for validation, "we need to first pick off the low-hanging fruit."

It Might Take a Rezulin

When pharma's interest in pharmacogenomics accelerates, will it decide to take the technology in-house, leaving Variagenics out in the cold? One can compare the technology to combinatorial chemistry, which failed to become the foundation of an independent business largely for that reason. "Those companies had an opportunity," Crouch says, "but they allowed themselves to become commoditized by pharma before they could establish momentum." On the other hand, if the combi-chem opportunity first presented itself today, he believes the now-improved financing climate for platform technology companies could have enabled those companies to independently develop their libraries, screens, and high-throughput methodologies and build value prior to partnering.

But should Variagenics suffer setbacks in its partnering programs or if credible validation fails to materialize as expected, a fallback could be for it to obtain one of the otherwise efficacious drugs whose side effects have caused its withdrawal from the market or at least limited its marketability. Referring to the Warner-Lambert compound pulled from the market because of unexpected liver toxicity in a small group of patients, Tony Wild has said, somewhat hyperbolically, that Variagenics should take a license to troglitazone (Rezulin), develop a pharmacogenomics profile to identify patients at risk for side effects, and sell the drug themselves.

Indeed, down the line, in-licensing drugs that can be rescued via pharmacogenomic testing is a part of the company's plan—especially for compounds with a long remaining patent life, as is the case with Rezulin. Another example Wild cites is Felbatol(felbamate), a Carter-Wallace Inc. anti-epileptic drug from the late '80s with potentially severe hematological side effects. The product is still on the market, but carries a stern black-box warning. With a pharmacogenomics label, "it could be a several hundred million-dollar product," Wild says.

Pharmacogenomics-based drug rescue businesses have plenty of unknowns. Most such compounds will not have long shelf lives remaining, raising the question of how to create exclusivity for such a "rediscovered" compound. They've already been killed once, and it's hard to pull nails out of the coffin and resurrect them. But perhaps the major risk is the FDA and other countries' regulatory bodies: there have been few, if any, examples of drugs given black-box warnings that were re-marketed with a diagnostic for identifying the right patients. And despite FDA's declaration that it would look more favorably at approval if an NDA included pharmacogenomics safety data, the agency might also require a tremendously expensive physician-education program in order to make sure the doctors ordered the diagnostic, potentially destroying much of the underlying profitability, as well as complicating the reimbursement program. Thus, for the time being, Variagenics wants to stick with its stepwise, lower-risk approach to building its business.

But some say that drug rescue is precisely what it will take to make drug companies adopt pharmacogenomics as an everyday part of development. Just as the drug industry didn't accept the idea of selling chirally pure versions of its drugs until Aventis SA made a great success out of the Seldane metabolite, Allegra, drug companies might not embrace the power of pharmacogenomics until someone else proves the case.