Personalized medicine, pharmacogenomics, and companion diagnostics

One particularly powerful outcome of the Human Genome Project has been the appreciation of how much small-scale genetic diversity, in the form of single nucleotide polymorphisms (SNPs), exists across the apparently normal general population. Large and still growing databases of these variations are behind the topic of this month’s examination of the growing fields of personalized medicine and companion diagnostics (CDx).

Genetics and pharmacogenomics

Note the use of the word “apparently” in the paragraph above. Let’s consider a simple example in the form of an SNP known as rs671. Residing on Chromosome 12 at chromosomal nucleotide position 111803962, this occurs within the gene for aldehyde dehydrogenase 2 (ALDH2). The base at this locus commonly exists as either an adenine (A) or guanine (G), with fairly strong correlation to ethnic background: people of central European descent are essentially 100% certain to have the homozygous G:G genotype, while people of Asian heritage are roughly 40% likely to be A:G heterozygotes and 10% likely to be A:A homozygotes.

(Two quick reminders in case anyone is puzzled: while DNA is double-stranded, we generally only refer to the sequence of one strand in any given sequence location, as the other strand is defined through Watson-Crick base pairing; thus the locus is really either a G:C or A:T base pair, not just a lone floating G or A. Second, since Chromosome 12 is an autosome, everyone has two copies and thus we need to describe both individual loci; thus the “G:G” or similar shorthand notation.)

In most situations, whether a person is of the G:G, G:A, or A:A genotype has little known significance. However, this single apparently innocuous nucleotide change has a very obvious and immediate physical manifestation (phenotype) when people carrying the differing genotypes are exposed to alcohol. ALDH2 is the second enzyme on the breakdown pathway for ethanol where ethanal (the aldehyde coming from ethanol in the first step of the path) is converted to ethanoic acid, and there’s a difference in the kinetics, or functional speed, of the enzyme as coded for by the different genotypes. The G form ALDH2 is a fast enzyme, catalyzing the rapid breakdown of ethanal before it can build up, whereas the A form ALDH2* is slower, and as a result, ethanal levels build up. The result is a rapid—and often very highly visible—flushing of the face, known sometimes as “Asian Flush” because of its most common appearance in people of Asian heritage. As one might guess, the phenotype is most apparent in people with the A:A genotype (all slow enzyme) and less so in G:A heterozygotes, where about half the available enzyme is the fast isoform. In addition to the flushing, people with the A:G or A:A genotype also tend to suffer from very severe hangovers from even moderate alcohol consumption; not surprisingly, many of these people learn to avoid alcohol altogether.

While this particular example is not of much medical significance, it serves to demonstrate how a single seemingly innocuous change in the genome can have a dramatic impact on how a specific substance such as a drug may be metabolized differently in different people. It’s this link between genotype and drug metabolism—pharmacogenomics—which is leading the drive into applications of personalized medicine and companion diagnostics.

A familiar example: warfarin

A clinically relevant example comes from warfarin (Coumadin) dosing. In use since the 1950s as a clotting inhibitor through indirect action on vitamin K processing, warfarin is prescribed for many people at risk of thrombotic episodes such as strokes or deep venous thrombosis, or recurrent myocardial infarction. While use of warfarin is generally thought to be highly beneficial in these cases, it’s critical to adjust the dosing level such that while unwanted clotting doesn’t occur, neither is the patient put at unacceptable risk for uncontrolled bleeding. It’s been known for a long time that different people seemed to have very different sensitivities to this drug, requiring careful trial and error testing of patients to establish a therapeutic dose level.

The reason for this diversity of response lies primarily in polymorphisms in two genes: VKORC1 (vitamin K epoxide reductase 1; also on the vitamin K processing pathway) and CYP2C9 (cytochrome P, which acts to clear a wide range of circulating drugs including warfarin). Just as in our ALDH2 example, SNPs in these two genes (rs9923231 for VKORC1, and most commonly rs1799853 and rs1057910 in CYP2C9) impact drug kinetics and cause large variations in what constitutes a therapeutic dose for warfarin. The molecular ability to genotype relevant patients for these critical alleles prior to introduction of warfarin therapy was hailed as a large help in estimating appropriate initial dose ranges and avoiding relatively common adverse reactions from over- or under-dosing prior to establishing individual patient response data. In fact, genotyping for warfarin sensitivity was indicated as early as 2007 by the U.S. Food and Drug Administration (FDA), making this one of the first widely applied and best known pharmacogenomic tests. At present, the jury is still out on the true utility in this specific example; while numerous case studies have indicated availability of genotypic data has helped prevent adverse reactions to warfarin dosing, it’s also been reported that in this particular case, physicians have such a long familiarity with having to control dosing through trial and error that little true benefit is added by addition of genetic data. (Readers interested in one example of a recent review of these conflicting conclusions may see reference 1.)

Regardless of whether the warfarin example will stand the test of time with regard to true utility, other applications of pharmacogenomics are clearly apparent. In particular, companies active in drug discovery and development are increasingly interested in knowing genotypic profiles of subjects enrolled in clinical trials. It is not hard to imagine scenarios where in aggregate across an entire test population, a drug in testing may not show statistical efficacy; however, analysis of responding versus non-responding patient populations might demonstrate a genetic basis for identification of particular genetic backgrounds where efficacy is observed. Similarly, adverse drug reactions may be identified as being associated with a particular genotype. This increased granularity, as it were, of the patient populations promises to both more frequently identify useful drugs (in appropriate patient genotypes) and also identify genetic-based contraindications.

Companion diagnostics

This concept of personalized medicine, treatment tailored to the genotype of the patient, is taken a step further in oncology. Here, rather than focusing on the genotype of the patient, it is the genotype of the cancerous cells which is of interest. In particular types of cancers, there are frequent commonalities with respect to particular cell growth signalling pathways being deregulated. Mutation of a gene along such a pathway can lead to a form of signalling molecule in an “always on” form, constantly signalling for growth when it should not be. Elucidation of some of these common oncogenic pathways has allowed for the development of examples of highly effective and specific drugs which can block their respective target pathways. A key concept, however, is that these drugs are only effective at blocking growth signals which are in effect “upstream” of the drug in the signal path; if an ectopic signal is being created “downstream” of the drug action point, the drug has no effect.

Perhaps the best-known example of this relates to the epidermal growth factor receptor (EGFR) pathway and the KRAS and BRAF genes. Oncogenic mutations upstream of the receptor can be effectively blocked or inhibited by drugs like Erbitux (cetuximab) or Vectabix (panitumumab), but mutations in KRAS or BRAF—effectively downstream of the receptor—are not responsive to these drugs. Knowledge of this allows for significant cost savings to the medical system through avoidance of non-productive use of these costly medicines, and also saves time by allowing immediate progression to alternate second-line therapies more likely to be effective.

Examples of this type of personalized medicine, wherein a specific genetic test is tied to the expected efficacy of a drug, are referred to as companion diagnostics (CDx), in the sense that they are a “companion” to a particular drug; the test is employed in assessing the likely utility of the drug treatment. A review of the FDA website “List of Cleared or Approved Companion Diagnostic Devices (In Vitro and Imaging Tools)” has a total of 19 entries (at this time of writing), with five being PCR-based, and seven being based on in-situ hybridization molecular methods. (Other methods in this list include immunohistochemistry methods and one magnetic resonance imaging [MRI]-based method). Of the 19, a total of 10 drugs are covered, out of which 18 are related to cancer. Of these 18, a full 10 CDx tests relate to use of just one drug (Herceptin).

These small and focused numbers, compared to the total number of drugs on the market, should give the reader reason consider just how much room for growth there is in the CDx field. The significance of this has not been lost on drug development (or molecular diagnostics) companies, which are actively expanding in this direction. One recent report predicted a growth rate in excess of 18% for this field between 2013 and 2019.² As this growth proceeds, clinicians will increasingly have access to highly specific molecular-based tools not just to diagnose the patient, but to select the most effective drug treatments, to guide therapeutic dose selections, and to avoid adverse drug reactions.