Pharmacogenomics

Pharmacogenomics

by Jeffrey M. Perkel

Among the many justifications for the Human Genome Project was the promise of personalized medicine. By sequencing the human genome, the rationale went, scientists could begin to suss out how it varies from person to person. That variation, in turn, would explain the diversity of responses in patient populations to therapeutics -- the incidence of side effects or drug efficacy, for instance. The inevitable result: one day in the not-too-distant future, drugs would be selected and dosed based on genetic information. The era of "guesstimated" medicine was on the wane.

A decade or so later, hyperbole has met reality. Drug-receptor interactions turned out to be "messier" (that is, less exclusive) than expected, which in turn makes tying polymorphism to phenotype more complicated, says Julio Licinio, director of the John Curtin School of Medical Research at the Australian National University and editor of The Pharmacogenomics Journal. As a result, though a relative handful of drugs now list genetic markers in their prescribing information datasheets, most do not.

"It's a field that suffered from tremendous hype in the beginning," Licinio says. "The Journal has more submissions than ever, but it's more sobering, with feet on the ground, than in the beginning."

Pharmacogenomics is the name given to that branch of genetics that was to be the driver of that medical sea-change. Less a technology than an experimental philosophy, pharmacogenomics is, Licinio says, "the area of medicine or pharmacology that looks at gene variation that can predict drug response."

By that definition, despite its -genomics suffix, pharmacogenomics is not limited to analyses of DNA, but rather also includes RNA and protein biomarkers. That means that the tools of pharmacogenomics run the gamut from immunohistochemistry to sequencing, microarrays to PCR. Whatever the nature of the assay, though, the idea is the same: to predict an individual's likelihood of responding, or having an adverse reaction to, a drug.

"It's the use that's classifying it, not the word genomic," explains Harry Glorikian, managing partner at Scientia Advisors, an industry consulting firm.

The breast cancer antibody therapeutic Herceptin and its companion diagnostic, Herceptest (based on immunohistochemical identification of the HER2/neu receptor on tumor tissue), for instance, qualify as pharmacogenomics products because the test predicts whether an individual's tumor will respond to the drug. So, too, do assays measuring genetic polymorphisms in metabolic enzymes, which can influence a person's ability to process and eliminate pharmaceuticals. Indeed, one of the first pharmacogenomics products to be approved by the US Food and Drug Administration wasRoche's AmpliChip, an AffymetrixGeneChip-based assay for polymorphisms in the cytochrome P450 drug metabolizing genes, which was green-lit in late 2004. More recently, the FDA has recommended using polymorphisms in the genes CYP2C9 and VKORC1to help guide dosing of the widely administered anticoagulant, warfarin (Coumadin). (For a complete list of FDA-recognized pharmacogenomic biomarkers, see "Table of Valid Genomic Biomarkers in the Context of Approved Drug Labels".)

According to Licinio, warfarin dosing is a tricky business: under-dose, and the patient can have a stroke; over-dose, and the patient can hemorhage. Given the stakes, the drug is a bellwether of how likely doctors are to embrace genetic testing for routine drug dosing problems, he says. "If people don't do it for this drug, I don't see them doing it for any drug."

Not included under the pharmacogenomics umbrella, according to Licinio's definition, are nucleic acid-based molecular diagnostics that address, for instance, risk prediction, disease diagnosis, and monitoring, as none of these influences drug treatment decisions. Thus tests like Myriad Genetics' BRACAnalysis (a DNA sequencing-based assay) and deCODE Genetics' deCODE ProstateCancer test (based on the Centaurus SNP detection assay) don't fit the definition, because they report disease predisposition.

On the other hand, contract work by those same companies for pharmaceutical clients often does meet the definition of pharmacogenomics. Richard Leach, vice president of scientific services at deCODE Genetics, says his company's CLIA- and CAP-accredited lab in Reykjavik, Iceland sometimes works with pharmaceutical companies to help them map clinical trial outcomes, such as drug efficacy, to genetic variation -- the very essence of pharmacogenomics.

According to Rick Hockett, Chief Medical Officer at Affymetrix, genetics can inform three different aspects of the drug-patient interaction: drug efficacy, metabolism, and side effects. And tools exist to target studies at every level, from genome-wide fishing expeditions, to more focused, targeted studies. The company's Genome-Wide Human SNP Array 6.0 and Axiom Genotyping Solution, for instance, enable the former application, scanning up to 1.8 million genetic markers genome-wide; its DMET Plus Arrays focus specifically on 1,936 markers in just 225 drug metabolizing genes. Custom arrays, says Hockett, can handle the "intermediate confirmatory space."

In January, researchers at Brigham and Women's Hospital, in Boston, and collaborators used the Affymetrix DMET array to demonstrate that efficacy of the drug clopidogrel (Plavix) was tied to polymorphisms in the cytochrome P450 gene, CYP2C19. The drug as administered is actually inactive; the 2C19 enzyme is required to convert it to its active form.1

"2C19 variants decreased the exposure to the active metabolite of that drug, which led to no apparent clinical benefit for patients who carried at least one 2C19 variant," Hockett explains.

Applied Biosystems, part of the Life Technologies Corp., also targets the pharmacogenomics market with a sizable panel of TaqMan® Drug Metabolism (DME) genotyping assays. With some 2,700 TaqMan® genotyping assays, the panel represents variants in the regulatory and protein coding regions of 221 drug metabolism and transporter genes.

Sejal Desai, a product manager for the genomic assays business at Life Technologies, says the TaqMan® DME assay collection allows detection of single-nucleotide polymorphisms (SNPs), insertion-deletions (in/dels), and multinucleotide polymorphisms (MNPs) in these genes. Users can obtain the assays either individually or as a set.

"That's the beauty of TaqMan® assays, the flexibility is really important for customers," says Desai. "They can buy assays as a panel, they can buy them individually, however they want."

Such assays can measure more than just genetic polymorphisms. The AlloMap test, from molecular diagnostics firm XDx, uses quantitative real-time PCR to measure the expression of a panel of genes that can predict the likelihood of cardiac transplant rejection at the time of testing.

According to Dirk Lammerts, vice president of marketing and corporate development at XDx, that application makes AlloMap a molecular diagnostic, but not a pharmacogenomics product. But the firm is doing some pharmacogenomics work: In collaboration with Bristol-Myers Squibb (BMS), the company is searching for gene expression biomarkers of systemic lupus erythematosus, which will help BMS advance the development of its drug Orencia, a rheumatoid arthritis therapeutic.

Though he declined to disclose details of the collaboration, Lammerts says in general it will follow the model the company used to develop AlloMap. Starting with a whole-transcriptome arrays (from Agilent Technologies) that could probe expression of 30,000 genes, the firm selected 252 candidates that seemed linked to cardiac rejection. Then, using RT-PCR, the company whittled the list of significant transcripts to 68, and ultimately to 11. Nine additional genes were added as controls to produce the final set of 20, whose weighted expression is used to compute a numerical measure of risk.

Pharmacogenomics, says Lammerts, is not just about invariant genomic markers like SNPs. Sometimes it requires more real-time information, such as gene expression or protein abundance.

"In a disease [such as cancer or immune rejection] where the biology is not well understood and is dynamically changing, especially when you want to know if a drug is working better or less effectively, you have to resort to gene expression, because that's how you measure the current state," he says.

For Licinio, the pharmacogenomics dream that was hyped a decade ago is not dead, but delayed. If nothing else, that's because the required clinical studies are unlike anything seen today. For instance, he says, drug studies almost always consider patients with a single condition and a single drug. But that doesn't reflect the reality of the modern patient population.

According to the informal survey by one of Licinio's colleagues at a major research hospital in California, "Each patient takes on average 14 different medications. These combinations have never been tested, even in a rat," he says. "So each person, when they go to the doctor, they're like a guinea pig in an experiment of n equals 1."

Collecting the required data would require enormous studies and concomitantly massive costs. As a result, he says, perhaps 50 or 100 years from now the hype will come to pass, "But it is in the vastly far future."

References:
1Mega JL et al., "Cytochrome P-450 Polymorphisms and Response to Clopidogrel," New England Journal of Medicine,360:354-62, 2009.

  • <<
  • >>

Comments