Principles in Pharmacogenetics

It is well recognized that drug therapy produces highly variable effects among individuals. The discipline of clinical pharmacology has defined multiple mechanisms underlying this variability, and they may be conveniently divided into two broad classes. One mechanism is pharmacokinetic: that is, variability in any one of a number of processes that determine drug and metabolite concentrations in plasma and tissue determines variability in the outcome of drug therapy. The other mechanism is pharmacodynamic: here, individuals display variable drug effects despite equivalent drug concentrations. As in all other branches of medicine, understanding these mechanisms has moved over the past 20 years from a descriptive exercise to identification of specific molecules that modulate variability in drug outcome through these mechanisms. Thus, whereas the notion that unusual responses to drugs may be genetically determined was suggested in the late nineteenth century and the term pharmacogenetics formally coined in the late 1950s (1), it has been only in the last decade that the specific genetic defects leading to such unusual responses have been defined. Moreover, with an increasing appreciation of the density of polymorphisms in the human genome has come the understanding that some, and perhaps much, variability in drug action may reflect interindividual variability at the genomic level. This, in turn, has defined the nascent field of pharmacogenomics. Two types of allelic variants have been described. Mutations are rare variants that are very frequently associated with changes in encoded amino acids in congenital diseases such as cystic fibrosis, the congenital long-QT syndromes discussed further later, or the familial epilepsies (2). The term polymorphism is used to describe allelic variants that are detected in >1% of a given population. The most common types of polymorphisms are single nucleotide polymorphisms, known as SNPs. Other types of polymorphisms include dinucleotide or trinucleotide repeat motifs, and insertion or deletion of larger portions of DNA. Polymorphisms may be present in introns or exons. When they are present in exons, they may result in an amino acid change (nonsynonymous) or no change (synonymous). When they are present in introns, they may still regulate gene expression. It has been estimated that there are >3,000,000 SNPs in the human genome and that ≤500,000 may reside in coding regions. By the 1950s, two defects in drug disposition accounting for variability in the outcome of drug therapy had been described. The first was the rare syndrome of prolonged paralysis after administration of succinylcholine, found to be due to familial defects in pseudocholinesterase, the enzyme responsible for removal of succinylcholine (3). The second was the increased incidence of isoniazid neurotoxicity in patients with unusually rapid N-acetylation (4). N-acetylation was determined, in twin studies, to be familial, but it was only in the early 1990s that the molecular basis for the defect was defined. It is now recognized that the activity of N-acetyl transferase is determined by two enzymes, NAT1 and NAT2. NAT1 is constitutively expressed, and individuals lacking NAT1 activity have not been described. Genetic variants in NAT2 leading to loss of enzymatic activity are present in slow acetylators, who metabolize only via NAT1, in contrast to rapid acetylators, who have both enzymes available for this reaction (5). In the 1970s, workers in England reported profound hypotension in a small number of subjects treated with the investigational antihypertensive debrisoquine (6). Subsequent studies identified debrisoquine 4-hydroxylation as the major (and indeed the only significant) pathway for debrisoquine bioinactivation. Individuals with defective debrisoquine 4-hydroxylation developed very high concentrations of parent drug after single drug doses, thereby explaining hypotension. At the same time, a similar scenario unfolded in Germany with the description of rare individuals with defective N-oxidation of sparteine, an oxytocic drug (7). Subsequent studies demonstrated co-inheritance of defective debrisoquine 4-hydroxylation and sparteine N-oxidation, indicating that the same genetic abnormality was responsible for these two pharmacogenetic defects. The formal name for the affected enzyme is cytochrome P-450 2D6 (CYP2D6). CYP2D6 is not particularly abundant in human liver, but it is now recognized that multiple drugs, including antiarrhythmics, β blockers, phenformin, tricyclic antidepressants, antipsychotics, and codeine are metabolized by this enzyme (8). CYP2D6 is expressed also in brain, and some studies have implicated variability in CYP2D6 activity in Parkinsons disease, although this is controversial (9). The clinical importance of CYP2D6-mediated metabolism for an individual drug (and an individual patient) depends on whether alternate pathways for drug elimination are present. Thus, for example, the reason that sparteine and debrisoquine were identified early as CYP2D6 substrates is that alternate pathways are not present, so markedly increased drug concentrations (and exaggerated drug effects) were the first clue to the presence of the defect. In the case of codeine, CYP2D6-mediated biotransformation leads to formation of a more active metabolite, morphine. Hence, CYP2D6 poor metabolizers (PMs) demonstrate reduced narcotic and analgesic effects after codeine administration (10). The CYP2D6 PM phenotype is present in 5–10% of whites and African Americans, and is much rarer (perhaps 1%) in Asian populations. Thus, most subjects are extensive metabolizers (EMs). It is in EMs, however, that the most serious manifestations of the CYP2D6 polymorphism may arise, as it is in these subjects that CYP2D6-based drug interactions may occur. A number of drugs in common use, including quinidine, tricyclic antidepressants, and some selective serotonin reuptake inhibitors (fluoxetine, paroxetine), are potent CYP2D6 inhibitors. As a result, an EM subject receiving a CYP2D6 substrate drug will display the phenomenon of phenocopying when an inhibitor is added; that is, they will behave as though they were a PM. Thus, although PMs may be subject to unusual drug effects, these are usually manifest within several doses of initiation of the drug. EM subjects conversely may be stabilized with prolonged drug therapy and then display an unusual drug effect when the inhibitor is added. If a drug can be eliminated by multiple pathways, the CYP2D6 polymorphism may assume less clinical importance. The antiarrhythmic agent flecainide is eliminated both by CYP2D6-mediated bioinactivation to inactive metabolites as well as by renal excretion. Thus, PM subjects may receive flecainide and display no difference in pharmacologic effects compared with EM subjects. However, should renal failure develop in a PM subject, or in an EM subject receiving a CYP2D6 inhibitor, marked flecainide accumulation with life-threatening toxicity can ensue (11). The molecular defects underlying variable CYP2D6 activity have now been defined. Loss-of-function mutations in the CYP2D6 gene are present in ∼25% of whites and African Americans, and the PM trait develops in homozygotes. More than 50 individual such mutations have been described, complicating genetic diagnosis. The activities of CYP2D6 in heterozygotes and in homozygotes for functional alleles are not distinguishable. Individuals with CYP2D6 gene duplication, hyperextensive metabolizers, have been described, particularly in the context of tricyclic antidepressant therapy, in which very large doses may be required to obtain therapeutic plasma concentrations (12,13). The principles defined in clinical and molecular studies of CYP2D6 apply equally to other pathways of drug metabolism. Clinically important allelic variants have been described for two other members of the cytochrome P-450 super family, CYP2C9 and CYP2C19. CYP2C9 is responsible for the disposition of S-warfarin (the active enantiomer of the anticoagulant warfarin) and the antihypertensive losartan (14). CYP2C19 metabolizes the antiseizure drugs (AEDs) mephenytoin (MHT) and diazepam (DZP) (13), and the gastric proton-pump inhibitor omeprazole. Polymorphisms in a transferase, thiopurine methyltransferase, are important in predicting both the benefits and side effects during therapy with azathioprine and 6-mercaptopurine in rheumatoid arthritis and childhood leukemias (15,16). The reader is referred to reviews for discussion of more details of specific drugs and adverse or beneficial drug effects linked to these polymorphisms (8,17). The principles defined by these studies apply equally to other pathways of drug disposition in which polymorphisms have not (yet) been identified. The most important drug-metabolizing enzyme by weight and by number of substrate drugs is CYP3A4, which is expressed in the liver, intestine, and kidney. Polymorphisms in the coding region of the CYP3A4 gene have not been identified, although individuals vary substantially in their ability to biotransform CYP3A4 substrates. CYP3A4 substrate–based drug interactions can assume great importance for substrates in which alternate pathways of drug elimination are absent. This has been a special problem for the antihistamines terfenadine and astemizole and the gastric prokinetic agent cisapride (18). All of these compounds are extraordinarily potent blockers of a specific repolarizing current in the heart, and through this mechanism may confer a risk for potentially fatal drug-related arrhythmias. Fortunately, these compounds ordinarily undergo very extensive bioinactivation by CYP3A4 in intestine and liver before their entry into the systemic circulation. Indeed, the antihistaminic effects of terfenadine are mediated largely by the non–QT-prolonging metabolite fexofenadine, which has now been marketed as an antihistamine in its own right as Allegra. Administration of CYP3A4 inhibitors, such as erythromycin or ketoconazole, results in a marked increase in plasma concentrations of terfenadine, astemizole, or cisapride, with attendant QT prolongation and risk of fatal arrhythmias (discussed further later) (19,20). Within the past 5 years, it has become obvious that drug uptake and drug efflux from important cellular sites of action are not passive processes, but rather are mediated by the activity of specific drug-transport molecules. The best studied of these is P-glycoprotein (Pgp), which is expressed on the luminal aspect of enterocytes, on the biliary aspect of hepatocytes, on the luminal aspect of renal tubular cells, and on the luminal aspect of capillary cells making up the blood–brain barrier. The widely used cardiac glycoside digoxin provides a good example of the role Pgp can play in drug disposition; the drug is not metabolized, but undergoes biliary and renal excretion. In the late 1970s, clinical studies demonstrated that quinidine predictably doubles plasma digoxin concentrations, with an attendant increase in both cardiovascular and neurologic toxicity (21). Subsequently, a multitude of structurally and apparently mechanistically unrelated drugs have been shown to produce a similar effect: amiodarone, verapamil, itraconazole, erythromycin, and cyclosporine are examples. In in vitro studies, we and others showed that digoxin is a Pgp substrate and that quinidine, amiodarone, and other drugs elevating digoxin concentrations inhibit Pgp-mediated digoxin transport (22–25). Moreover, we found that although quinidine elevated digoxin concentrations in wild-type mice, this effect was absent in mice in which the homolog of MDR (the gene encoding Pgp in humans) was knocked out (25). Taken together, these findings demonstrate the importance of Pgp as a transport/elimination mechanism for digoxin and form the basis of its clinically important drug interactions. Another group of drugs that are clinically important Pgp substrates are human immunodeficiency virus (HIV) protease inhibitors. We have shown that these compounds are actively excluded from the central nervous system (CNS) by Pgp-mediated efflux (26). In mice in which the homolog of Pgp has been disrupted, marked elevation of CNS concentrations of HIV protease inhibitors (≤40-fold) was observed, suggesting that Pgp inhibition might be a reasonable therapeutic strategy for enhancing the effectiveness of these drugs, particularly to attack the virus at sanctuary sites such as the CNS. A synonymous SNP in exon 26 of the MDR gene, encoding Pgp, has been identified as a major modulator of the disposition of digoxin (27). The mechanism whereby a synonymous SNP could exert such an effect has not been elucidated, but likely reflects either a role for this region in expression of the Pgp gene or linkage dysequilibrium between this SNP and another polymorphic site, as yet unidentified, that actually regulates Pgp function. These pharmacokinetic concepts are important for understanding variability in drug actions among individuals and may be important in the drug-development process. For example, Pgp inhibitors are under consideration to enhance the efficacy of HIV protease inhibitors, and many large pharmaceutical companies will not further pursue compounds that undergo exclusive metabolism by a single pathway, particularly one that is genetically determined, such as the CYP2D6. Available AEDs do not display marked interindividual variability as a function of such genetically determined pharmacokinetics. However, carbamazepine (CBZ) and phenytoin (PHT) are well recognized as inducers of specific pathways (usually CYP3A4) and thus can cause drug interactions to which physicians and others caring for patients with epilepsy should be alert. DZP and MHT are substrates for CYP2C19, whereas CBZ, DZP, and likely lamotrigine (LTG) and PHT are CYP3A4 substrates. Our laboratory has a long-standing interest in drug-induced long-QT syndrome (18). In this entity, occasional patients exposed to QT-prolonging drugs develop exaggerated responses, culminating in a potentially fatal ventricular tachycardia, torsades de pointes. Implicated drugs include antiarrhythmics such as quinidine and sotalol, as well as a wide range of noncardiovascular agents, including antipsychotics (such as thioridazine), some tricyclic antidepressants, antibiotics (such as high-dose erythromycin or trimethoprim-sulfamethoxazole, or pentamidine), and other agents, such as antihistamines or cisapride. In some cases (e.g., terfenadine, astemizole, cisapride), elevation of plasma drug concentrations, because of the use of high dosages or coadministration of inhibitors of drug metabolism, can be implicated. Other risk factors have been identified, including hypokalemia and female gender. Nevertheless, the development of torsades de pointes during therapy with these agents is sufficiently unpredictable as to suggest as yet unidentified modulators of this effect. An unusual familial syndrome, the congenital long-QT syndrome, has very similar electrocardiographic manifestations (even in the absence of drugs) and, in some kindreds, a relatively high incidence of sudden death due to torsades de pointes (28,29). Thus, it becomes reasonable to hypothesize that as yet unidentified genetic factors may modulate the development of the drug-associated long-QT phenotype. As discussed later, this may be mutations in the same genes in which mutations cause the congenital syndrome or in others. The cause of the congenital long-QT syndrome is mutations in genes encoding cardiac ion channel proteins. To date, mutations have been identified in three pore-forming proteins (α subunits), including the cardiac sodium channel (SCN5A) and two potassium channels, KvLQT1 (also known as KCNQ1) and HERG (also known as KCNH2)(30–32). Disease-causing mutations have also been identified in two function-modulating proteins (β subunits), minK and MiRP1, also known as KCNE1 and KCNE2(33,34). Interestingly, mutations causing familial epilepsy have also been identified in ion channels. These include KCNQ2 and KCNQ3, members of the same subfamily as KvLQT1(35–37), and the sodium channel β-subunit gene (SCN1B) associated with the generalized epilepsy with seizures plus syndrome (38). Thus, the congenital long-QT syndrome and genetic epilepsies are both examples of ion channelopathies causing disturbances in excitable tissues manifest by highly symptomatic rapid cellular discharges. In the cardiology world, we call these arrhythmias, whereas in the neurology world, they are termed epilepsies. Whether the analogies between the two entities extend beyond the superficial is an interesting question. Most drugs that cause marked QT prolongation and torsades de pointes block a specific cardiac potassium current, IKr. In heterologous systems, expression of HERG results in IKr, and MiRP1 is a function-modifying gene for this channel complex. Based on this understanding alone, one could postulate three mechanisms whereby variability in the interaction between an IKr blocking drug and its pharmacologic target might arise, thus providing pharmacodynamic mechanisms for the development of torsades de pointes. These are Variability in the drug target. This could arise by variable expression of the drug target or by mutations that modulate its sensitivity to drug block. Indeed, mutations in MiRP1 that render the channel complex more sensitive to drug block than wild-type complexes have been described (34–39). External factors might modulate the extent to which a drug might block IKr. Hypokalemia is a well-recognized risk factor for drug-induced torsades de pointes. IKr gating, and sensitivity to drug block (which may be related processes), are exquisitely sensitive to changes in extracellular potassium within the physiologic range. For example, decreasing extracellular potassium from 8 to 1 mM (the extremes of the physiologic range), resulted in a 40-fold increase in affinity of channels for blocking drug (40). From a genetic perspective, any factor altering extracellular potassium would therefore be expected to modulate the development of drug-induced torsades de pointes. Alteration in other component(s) of the complex biologic system within which drugs act. Repolarization of the cardiac action potential is accomplished by a time-dependent decrease in inward currents (primarily through calcium currents) and a time-dependent increase in currents, including IKr and others, notably IKs, the potassium current formed by co-assembly of KvLQT1, and minK. The redundant nature of these repolarizing mechanisms provides considerable protection against drug-induced torsades de pointes. Indeed, some loss-of-function mutations in the genes encoding IK do not demonstrate any important clinical manifestations in the absence of drug challenge, presumably because a robust IKr still is present to accomplish repolarization. However, in this situation administration of an IKr blocker then exposes the subclinical IKs defect, and marked QT prolongation and torsades de pointes ensue (41,42). In addition, we have identified a polymorphism in the minK gene that is overrepresented among patients with drug-associated torsades de pointes. In vitro studies demonstrate that this polymorphism results in subtle changes in IKs gating. In addition, in follow-up computer modeling studies, these changes become manifest as arrhythmias in the setting of block of IKr.(43). The considerable genetic diversity that distinguishes us as individuals is only now becoming tractable to study. Individual allelic variants have been described that strikingly modulate drug actions, through either pharmacokinetic or pharmacodynamic mechanisms. An alternative scenario is that individual allelic variants may contribute only in a small degree to variable drug actions, but that it may be possible to identify collections of dozens, hundreds, or even thousands of SNPs that, together, identify a patient at high or low risk for beneficial or toxic drug effects (44). Thus, a major advance that the human genome project may bring is new approaches to optimize therapy with available agents and to develop new ones.