Pharmacogenomics: Challenges and Opportunities

The concept that genetic variation contributes to variability in disease phenotypes and in drug responses is widely accepted and validated in many research settings. For some drugs, there are clear implications of genetic information for drug therapy to avoid toxicity and to optimize response (1, 2). In addition, understanding genetic contributors to variability in drug response provides a new tool in drug development that carries the hope of decreasing the risk for unexpected toxicities, identifying patients most likely to respond, and streamlining drug development (3). The English physiologist Archibald Garrod (4) proposed that similar genetic factors might underlie inborn errors of metabolism and variable responses to drugs. The field of pharmacogenetics developed after variant drug responses due to large single gene effects were described (1, 58), as Garrod suggested. Genetic science is now increasingly turning its attention to how variation in large gene sets, in complex biological pathways (see Glossary), or in the whole genome, contributes to variable phenotypes, such as disease susceptibility (9). This evolution from genetics to genomics is paralleled by progress in understanding the genetic contribution to variable drug responses, from pharmacogenetics to pharmacogenomics (Table 1) (Figure 1). This review will outline progress in the field by describing mechanisms underlying variable drug responses, the potential role of genetic factors in their causes, and contemporary and evolving approaches to identifying these genetic factors. Examples are presented throughout, although it is not our goal to review these in a comprehensive fashion; rather it is our intent to identify the challenges that must be overcome and their potential solutions if genetic and genomic information is to be integrated into drug prescribing. Table 1. Challenges in Pharmacogenomics Figure 1. The concept of pharmacogenetics. Pharmacokinetics focuses on large clinical effects of single gene variants in small numbers of patients. However, the concept of pharmacogenomics examines many genomic loci, including large biological pathways and the whole genome, to identify variants that together determine variability in response to drug therapy. Mechanisms Underlying Variable Drug Responses Clinicians and the lay public accept the notion that not all patients respond to drug therapy in the same fashion. An overarching challenge in contemporary therapeutics is to define the mechanisms underlying such variability. Occasionally, the distribution of drug responses across a population is clearly bimodal, suggesting a predominant role for a single variable that is often genetic (Figure 2, top). More commonly, drug responses in patient populations show a broad distribution (Figure 2, bottom). Studies with twins done as early as the 1960s support the idea that this pattern of responses may also include a prominent genetic component (1214). Figure 2. Two types of variability in drug action. Top. Volunteers received 10 mg of the CYP2D6 substrate debrisoquine, and the ratio of urinary concentrations of the parent drug and its 4-hydroxy metabolite in urine were determined. This experiment identifies at least 2 distinct populations, extensive and poor metabolizers, separated at the antimode (arrow). Redrawn with permission from reference 10. Bottom. Change in FEV1 in 1117 participants in 3 different trials of antiasthmatic therapy (inhaled steroids). Although the responses vary markedly, from an apparently deleterious drug effect to a highly beneficial one, there is no antimode. Redrawn with permission from reference 11. Two distinct processes, either of which can be influenced by genetic factors, underlie the generation of a clinical drug action: delivery to and removal from target sites in plasma on cell surfaces, or within cells (pharmacokinetics) (see Glossary) and interaction with the targets to generate a cellular effect that is translated to clinical effect (pharmacodynamics). Thus, the starting point for many contemporary pharmacogenetic studies is identification of variable drug responses in an individual patient or across a population. Then, an understanding of the underlying pharmacokinetics or pharmacodynamics can be used to identify individual candidate genes (see Glossary) in which variant function may explain the variable drug response. An alternate approach, to interrogate many candidate genes or even the whole genome (see Glossary), has emerged more recently and will be discussed further. Genetically Determined Pharmacokinetics Some of the earliest findings in pharmacogenetics involve variations in single genes encoding drug-metabolizing enzymes, which can underlie aberrant responses to substrate drugs (Table 2). The highest likelihood of aberrant drug responses occurs when genetically determined reduced function of 1 drug elimination pathway is coupled with the absence of alternate pathways that can readily subserve the same function (15). Individual cases were first identified in patients with clinically dramatic phenotypes (see Glossary) because these patients were homozygous (see Glossary) for loss of function in such alleles (see Glossary). Table 2. Examples of Associations between Drug Response and Genetic Variants* Coding-Region Variants Changes in DNA sequence that occur in regions that encode protein may lead to changes in the primary amino acid sequence and protein function. A well-studied example that is entering routine clinical practice is the thiopurine methyltransferase (TPMT) gene, whose protein product is responsible for bioinactivation of thiopurines, such as azathioprine or mercaptopurine (1, 16, 17). Rare individuals who are homozygous for loss of function variants are at high risk for bone marrow aplasia during therapy with standard doses, and this is stated in the package label. Ten percent of persons carry a single abnormal allele and are also at increased risk for bone marrow toxicity (18, 19). Conversely, standard doses of mercaptopurine that are used in the 90% of patients with functional alleles mutations (see Glossary) may in fact be inadequate for achieving an optimal antileukemic effect (20). The most common mechanism for drug elimination is metabolism by members of the cytochrome P450 (CYP) superfamily. Common coding-region CYP variants that affect drug elimination and responses have now been described. The frequency of many variants varies by ethnicity, and this may be one factor determining ethnic-specific drug responses. Up to 10% of white and African-American persons are homozygous for loss of activity of a cytochrome P450 isoform, which is termed CYP2D6. Persons with this poor-metabolizer genotype have drug accumulation and increased side effects with some antidepressants (21). In addition, persons who are poor metabolizers do not metabolize codeine to its active metabolite morphine and thus have reduced analgesia (22). An important implication of the identification of highly variable CYP2D6 activity is that new drug candidates that are eliminated predominantly by this enzyme are often not further developed (23, 24). In contrast to CYP2D6, the poor metabolizer trait for a different CYP, CYP2C19, is more common in Asian persons, and persons with this genotype have higher drug concentrations and a greater cure rate of Helicobacter pylori infections during therapy with the CYP2C19 substrate omeprazole (25). DNA Variants in Noncoding Regions Only a small fraction of the human genome encodes proteins. One important role of noncoding DNA is to regulate the amount of messenger RNA (mRNA) transcribed, and thus protein generated, in the basal state or in response to many environmental stimuli. Sequence variants in regulatory regions that result in altered amounts of otherwise normally functioning protein can underlie abnormal drug responses. A good example is the repeat polymorphism (see Glossary) in the promoter of UGT1A1, which encodes the glucuronosyltransferase responsible for conjugation of bilirubin and many drugs. The most common hypofunctional allele, termed UGT1A1*28, is an insertion of 2 extra base pairs (TA) in a key regulatory region of the gene, resulting in decreased protein expression. Impaired elimination of bilirubin by this mechanism is the cause of the Gilbert syndrome. UGT1A1 is responsible for the metabolism of SN-38, the active metabolite of the anticancer drug irinotecan, and persons who are homozygous for UGT1A1*28 are at increased risk for serious adverse effects of the drug (26). This effect is of sufficient clinical importance that it is now described on the irinotecan product label (27, 28). Commercial tests for variants in TPMT, CYP2D6, CYP2C19, and UGT1A1 are now available (29). Variable Drug Transport Drug entry into and removal from cells are often active processes, accomplished by specific drug transport molecules (30, 31), and variants in the genes encoding these transporters have been implicated in variable drug responses. Thus, normal function of the drug efflux transporter P-glycoprotein is required for the biliary excretion of digoxin, and a common P-glycoprotein polymorphism has been associated with variable serum digoxin (32, 33). Similarly, polymorphisms in an organic anion (uptake) transporter have been implicated in the efficacy and some adverse effects of statins (34). Genetically Determined Pharmacodynamics Variability in the Genes Encoding Drug Targets Drugs can produce highly variable effects, even in the absence of substantial variability in drug concentrations at target sites. This pharmacodynamic variability tends to be drug- or disease-specific, in contrast to pharmacokinetic variability that often extends across many drugs and disease processes. One obvious set of genes in which variants might account for such pharmacodynamic variability is those encoding drug targets. Thus, a plausible candidate gene for modulating variability in response to 2-agonists in a