Gene differences lead to variations in drug response

The term “pharmacogenomics” came into vogue roughly 15 years ago or so, when the dream of the Human Genome Project had just completed its first release. It was becoming apparent that novel sequencing technologies were going to make it feasible to sequence, if not entire genomes of patients, at least selected highly informative loci at a low enough cost and fast enough turn around time to have potential for front-line clinical utility. Many drugs have well understood and narrowly defined targets (such as a particular cellular receptor or intracellular enzyme) and may also have known specific interactions with enzymes responsible for activation (conversion of prodrug forms to active forms) and/or enzymes responsible for breakdown and clearance of the drug.

This creates three obvious steps where individual variations in genetic sequence can alter response to the drug:

Target molecule – changes in binding efficiency and degree of activity modulation;
activation pathway – changes in kinetics of active drug availability; and
degradation pathway – changes in clearance rates, impacting steady state levels and duration of impact.

There are additional possible chances for interplay between individual genetic variation and pharmacodynamics/pharmacokinetics, such as specific transport molecules, but the general gist remains the same. Genetic differences can lead to individual variations in dose responsiveness to a given drug, and for drugs with a narrow therapeutic window (the dosing level which maximizes benefits and minimizes side effects), knowledge of this should be applicable in determining aappropriate dosing regimens.

Warfarin—variations in target and clearance

The simpler a given drug’s metabolism, the better understood the impact of various genetic variation on this metabolism, and the narrower the therapeutic window the more this pharmacogenomic approach seems attractive. Warfarin (the name, as many readers may know, comes from Wisconsin Alumni Research Foundation WARF; also known under trademark as “Coumadin”) provided a convenient intersection of these attributes.

Warfarin, and in particular its “S” stereoisomer form, acts indirectly as an anticoagulant by inhibiting the activity of the enzyme VKORC1, the vitamin K epoxide reductase. Reduced vitamin K is needed in the clotting cascade to convert inactive Factors II, VII, IX, and X to their active forms during clot formation, so a lack of reduced vitamin K slows the clotting process. The “S” enantiomer of warfarin is in turn degraded by a cytochrome CYP2C9, a member of the Cytochrome P450 family which is active in degrading a number of drugs (such as NSAIDs and angiotensin II receptor blockers).

Not all VKCOR1 genes are identical. In fact, there are quite a few known allelic variations in this gene but two in particular—1173 C>T and 1639 G>A—are associated with less translation of the VKORC1 mRNA into protein, leading to lower levels of enzyme available. (A quick aside here, those codes aren’t as mysterious as they look. They’re the number of the nucleotide in the gene changed, the wild type or “normal” base found there, and the mutated form.) None of us need special training in enzyme kinetics or pharmacology to grasp the key point here: If you have less enzyme around, and your goal is to partially block or slow down clotting (not just stop it altogether, which would be very dangerous), then you want less enzyme inhibitor in the system than you would want in the case of someone starting off with more enzyme. There’s additional nuances related to whether people are homozygous or heterozygous for these mutations to factor in as well, but to a first approximation we have a pretty good idea of how much VKORC1 is present in each of these genetic scenarios, and thus, some idea of how much we’d like to suppress it in order for all of these cases to end up with the same optimal therapeutic window of VKCOR1 activity.

We also need to think about genetic variations in CYP2C9, though, because these will influence what the duration and effective level of warfarin is in the system. A faster version of the enzyme will require more frequent or larger doses to maintain the same drug level then a sluggish version of the enzyme, where a dose will linger around longer. It turns out there’s a whole bunch of known CYP2C9 single nucleotide genetic variants which impact speed of warfarin breakdown, but two are particularly significant. CYP2C9*2 (these mutations get their own special names; in the coding system referenced above, this would be 430 C>T) has only about 70 percent of normal activity for warfarin clearance, and CYP2C9*3 (aka 1075 A>C) is a dismal 20 percent of normal rate. The caveats above regarding homozygous vs. heterozygous apply here as well, but the clear bottom line is if you blindly give the same (optimal) dose you’d give on a CYP2C9 wild type individual to someone homozygous for CYP2C9*3, their effective steady state warfarin levels would be exceedingly high, and you’d be supressing coagulation more than intended (or a safe balance). Imagine now if that person also had one of the VKCOR1 mutations described above, and you’d have a recipe for very poor outcomes.

Warfarin has however been around and in use for treating clotting risk since before individualized genomic medicine was anything but science fiction, so clinicians were well used to doing a careful, individualized dose titration process with patients. Starting out at low doses and measuring clotting responses allowed for effective phenotypic determination of a drug dose yielding the desired outcome range. This was however not a very fast process, meaning that significant time—days to weeks—might be spent at doses below the therapeutic window, during the titration process. Clearly, if you could start off right about at the optimal range, you’d be able to achieve benefits in patients at risk for unwanted clotting faster than by trial and error.

Warfarin—the perfect scenario?

Warfarin therefore provided an appealing trial case for pharmacogenomics; the majority of genetic influence is from a small number of known variations, the therapeutic window is narrow, and guided knowledge of effective dosing from outset of treatment is expected to have tangible benefits.

It’s no surprise then that utility of genetic testing involving the VKORC1 and CYP2C9 genes as part of initial dose determination was first formally indicated by the U.S. FDA in 2007, further updated in 2010 to tables of suggested doses based on various combinations of alleles of these two genes. Multiple commercial assays were released at around this time with clearance for testing these particular alleles, and there was significant enthusiasm for their uptake.

The foregoing is however all preamble to the focus of this month’s Primer; namely, where are we today with regard to this grandfather of pharmacogenomic tests, and are there any lessons to be learned from it? This topic has been the subject of intensive study and publication by specialists in the topic, so for those looking for highly detailed answers a trip to the primary literature is both recommended (and readily fruitful).

Within the constraints of brevity in force here, the general consensus however can be summarized as, “it’s not really as helpful as we expected,” or to quote just one recent review,¹ “the findings still present disparities …further studies should be encouraged to try to demonstrate the benefits.” This certainly hasn’t been the unmitigated triumph of new technology that proponents of molecular testing were hoping for, and perhaps even expecting—why?

What can we take away from all of this?

The answers to this are multifactorial. One significant reason likely has to do with the fact that the pre-existing approach to dose titration was well established, simple, relatively cheap, and effective.

A second factor is that while the genetic markers mentioned above (and target of most tests) are the most significant mutations known, they are not an exhaustive list.

Even when guided by pharmacogenomic data, it remains critical to confirm that the end physiological result is what’s intended—thus, all patients are taking part in traditional testing anyway. Because rare outlier cases can occur where testing might not uncover less common genetic reasons for enhanced sensitivity to warfarin, there’s also a risk mitigation incentive to start dosing and response monitoring at the low end of the dose spectrum regardless of genetic results. If, however, a first low dose is administered and results seen are in agreement with expectation from genetic testing, then a faster movement to predicted optimal dose range would seem warranted.

In essence then the issue is that even in this poster child case for pharmacogenomics, phenotypic impact remains the truly meaningful endpoint by which results are judged and best practices use genetics as a supporting tool rather than standalone replacement.

If this is a model case, where are we likely to find pharmacogenomics of use in the future? The requirements for a simple well described uptake/action/drug decay pathway, with a limited number of genetic variations of known impact, is essential but likely possible in many scenarios. Situations where phenotypic response testing is expensive, complex, invasive, poorly available, or otherwise high risk will also be most likely to benefit from tools which allow faster achievement of correct dosing.

As larger data sets of genetic variation and allele combinations vs. response to particular drugs become available, more cases which meet the first of these will be uncovered; and in the cases where something from the second group of requirements are also met, there will be a utility for these tests. Warfarin testing may not have lived up to its initial and perhaps overly ambitious expectations for revolutionizing drug dosing, but it has provided a wealth of information on the real-life application of pharmacogenomics. Our consideration of this suggests that within limitations, personalized medicine of this format will be increasingly applicable as an aid in effective drug dosing.