Science Wiki

Pharmacogenomics is the study of how an individual's genome influences the response to drugs. It lies at the intersection of pharmacology and genomics. Its goal is to use genetic information to choose drugs and doses that are more effective and less likely to cause harm than the standard regimen for the average patient. The older term pharmacogenetics usually refers to the effect of single genes, and pharmacogenomics refers to the effect of many genes or of the whole genome. In practice the two terms are often used interchangeably.

Clinical dosing regimens are usually derived from clinical trials in which the average response is measured. Individuals differ widely in how they handle a drug, and part of this variation is inherited. Pharmacogenomics relates genetic polymorphisms to differences in pharmacokinetics (absorption, distribution, metabolism, and excretion) and pharmacodynamics (the effect of the drug on its targets). These associations help to predict efficacy and to prevent some adverse drug reactions. Genes are only one of several influences on drug response. Age, body weight, organ function, drug interactions, diet, and adherence also matter, and genotype explains a variable part of the differences between patients.

Early Enzymopathies, CYP2D6 Discovery, and High-Throughput Genotyping

Variation in the response to drugs was recognized in the 1950s. Inherited deficiency of plasma cholinesterase was linked to prolonged paralysis after the muscle relaxant suxamethonium, hemolysis after the antimalarial primaquine was linked to glucose-6-phosphate dehydrogenase deficiency, and the rate of acetylation of isoniazid was shown to be inherited, with "slow" and "fast" acetylators. Arno Motulsky proposed in 1957 that inherited factors underlie differences in drug response, and Friedrich Vogel coined the term "pharmacogenetics" in 1959. In the 1970s, studies of the metabolism of debrisoquine and sparteine led to the discovery of the polymorphism of CYP2D6. The sequencing of the human genome and the development of high-throughput genotyping in the 2000s led to the term "pharmacogenomics" and to genome-wide association studies of drug response. Since about 2010, consortia have published guidelines for the clinical use of genotype data, and several drug labels now carry pharmacogenomic information.

Drug-metabolizing enzymes

The liver is the main site of drug metabolism, and the intestine, kidney, and other tissues also contribute. In Phase I metabolism, enzymes introduce or expose polar groups on a drug by oxidation, reduction, or hydrolysis. The main enzymes are the cytochrome P450 (CYP) superfamily. They are membrane-bound, heme-containing monooxygenases located mostly in the endoplasmic reticulum of hepatocytes. A small number of CYP enzymes, mainly CYP3A4, CYP3A5, CYP2D6, CYP2C9, CYP2C19, and CYP1A2, account for most of the oxidative metabolism of drugs. CYP enzymes are involved in the Phase I metabolism of about 70 to 80 percent of drugs that undergo metabolism, and not of all drugs. CYP2D6 metabolizes about a fifth to a quarter of commonly used drugs, even though it makes up only a few percent of the hepatic CYP protein.

Phase II enzymes conjugate drugs or their Phase I metabolites with polar groups, which makes them easier to excrete. Polymorphic enzymes include the UDP-glucuronosyltransferases (for example UGT1A1), thiopurine S-methyltransferase (TPMT), N-acetyltransferase 2 (NAT2), and the glutathione S-transferases. Genes for transporters, such as SLCO1B1 (the hepatic uptake transporter OATP1B1) and ABCB1 (P-glycoprotein), also contribute to variation in drug disposition.

Enzyme kinetics and clearance

Many drug-metabolizing enzymes follow Michaelis-Menten kinetics. The rate of the reaction v at substrate concentration [S] is:

At therapeutic concentrations, which are usually far below K_m, the rate is approximately proportional to the concentration, with a slope equal to the intrinsic clearance:

Genetic variants can lower V_max, for example by producing an unstable protein, by altering splicing, or by deleting the gene. Other variants change the active site and increase K_m (reduce the affinity for the substrate). Both changes reduce CL_int. In the well-stirred model of the liver, the hepatic clearance CL_H depends on the hepatic blood flow Q_H, the unbound fraction of the drug in blood f_u, and the intrinsic clearance:

For drugs with a low extraction ratio, CL_H is approximately f_u CL_int, and a change in enzyme activity changes the clearance proportionally. For drugs with a high extraction ratio, clearance is limited by blood flow, and a change in the enzyme has a smaller effect on the clearance of intravenous doses but a larger effect on the bioavailability of oral doses.

The exposure to a drug is measured by the area under the concentration-time curve (AUC), which depends on the dose and the clearance:

where F is the oral bioavailability. At steady state, the average concentration with a dose D given every τ hours is C_ss = F D / (CL τ). The elimination half-life is t_1/2 = ln 2 × V_d / CL, where V_d is the volume of distribution.

The effect of a polymorphism depends on the fraction f_m of the total clearance that goes through the affected pathway. If the pathway is absent in a poor metabolizer and the other pathways are unchanged, the clearance is reduced from CL to _CL_(1 − f_m), and the exposure is increased by a factor of:

For example, if a drug has f_m = 0.8 through a polymorphic enzyme, then a poor metabolizer has a fivefold higher AUC than a normal metabolizer on the same dose. If f_m = 0.3, the increase is only about 1.4-fold, so the same genotype matters much less. The clinical effect therefore depends on the drug as well as the genotype, and on the therapeutic index of the drug.

Nomenclature

Genetic variants of pharmacogenes are named using the "star" (*) allele system. Each star allele is a haplotype that is defined by one or more variants, such as single-nucleotide polymorphisms, insertions and deletions, or copy number variants. The *1 allele is the reference allele with normal function in most genes. A diplotype such as _CYP2D6_*1/*4 is the combination of the alleles on the two chromosomes. The PharmVar database maintains the allele definitions, and PharmGKB curates the clinical annotations of gene-drug pairs. Some alleles have no function, such as _CYP2D6_*4 (a splicing defect that is the commonest non-functional allele in people of European ancestry, with a frequency of about 20 percent), _CYP2D6_*5 (a whole-gene deletion), and _CYP2C19_*2 and *3. Some alleles have reduced function, such as _CYP2D6_*10 (common in East Asian populations) and *41. Some alleles have increased function, such as _CYP2C19_*17 (a promoter variant) and duplications of CYP2D6.

Activity score and phenotype categories

A genotype is converted to a predicted phenotype. For CYP2D6, an activity score system assigns each allele a value (for example 1 for normal function, between 0 and 1 for decreased function, and 0 for no function), and the score of the diplotype is the sum of the two values. For duplicated alleles, the value of the allele is multiplied by the number of copies. For example, _CYP2D6_*1/*4 has an activity score of 1 + 0 = 1, and _CYP2D6_*1/*1×3 has a score of 1 + 3 = 4. In the consensus terms of the Clinical Pharmacogenetics Implementation Consortium (CPIC), the categories are:

Phenotype categories are predictions from genotype. They are approximate, and measured enzyme activity within a group varies widely, because it also depends on factors such as age, drugs, and disease. For some genes, a genotype test does not detect rare variants or structural variants, so a "normal" result does not exclude unusual function.

Frequency in populations

The frequency of the alleles differs among populations. About 5 to 10 percent of people of European ancestry are CYP2D6 poor metabolizers, compared with about 1 percent of East Asians and a few percent of people of African ancestry. The frequency of ultrarapid metabolizers through gene duplication is about 1 to 2 percent in Northern Europe and is higher (up to about 10 percent or more in some studies) in parts of North Africa and the Arabian Peninsula. About 2 to 5 percent of people of European ancestry and 10 to 20 percent of East Asians are CYP2C19 poor metabolizers. The frequency of a recessive phenotype is given by the Hardy-Weinberg principle. If the poor metabolizer genotype frequency is q² = 0.07, the allele frequency is q = √0.07 ≈ 0.26, and the fraction of heterozygous carriers is 2_pq_ = 2 × 0.74 × 0.26 ≈ 0.39. Because populations are genetically diverse, ancestry is not a reliable replacement for genotype in an individual. Genotyping is more accurate than the use of race or ethnicity categories.

The clinical effect of a metabolizer phenotype depends on whether the drug is active or is a prodrug.

Active drugs

For most drugs, the parent compound is the active form, and metabolism terminates or reduces its effect.

• In poor metabolizers, the clearance is reduced, and standard doses can lead to higher exposure and a greater risk of dose-related adverse effects. Examples include the antidepressants and antipsychotics metabolized by CYP2D6 and the proton pump inhibitors, voriconazole, and some antidepressants metabolized by CYP2C19.
• In ultrarapid metabolizers, the clearance is faster, and standard doses can produce concentrations below the effective range, so that treatment may fail. The effect depends on the therapeutic index and on the contribution of other elimination routes.

Prodrugs

A prodrug must be converted to its active form, often by the same enzymes. The relationship is then reversed.

• Clopidogrel is an antiplatelet prodrug that is activated in two oxidation steps, with CYP2C19 making the main contribution. Patients who are CYP2C19 poor or intermediate metabolizers form less of the active metabolite and have less inhibition of platelet aggregation. After percutaneous coronary intervention with stent placement, they have a higher rate of major adverse cardiovascular events, including stent thrombosis. Guidelines recommend an alternative antiplatelet drug, such as prasugrel or ticagrelor, for such patients when it is suitable, and the label of clopidogrel carries a boxed warning. The magnitude of the effect differs among studies and depends on the clinical situation.
• Codeine is converted by CYP2D6 to morphine, which is responsible for most of its analgesic effect. Poor metabolizers have little conversion, and codeine provides little pain relief. Ultrarapid metabolizers can form higher concentrations of morphine, and there are reports of life-threatening or fatal respiratory depression, in particular in children after tonsillectomy and in infants of breastfeeding mothers who were ultrarapid metabolizers. Regulators have restricted the use of codeine in children for this reason. Tramadol is activated by CYP2D6 in a similar way.
• Tamoxifen is converted by CYP2D6 and other enzymes to the more active metabolite endoxifen. Studies on whether CYP2D6 genotype predicts the outcome of breast cancer treatment have given inconsistent results, and guidelines do not agree on the use of genotype for tamoxifen.

High-Evidence Germline and Somatic Pharmacogenomic Biomarkers

The best-known gene-drug pairs include:

Pharmacogenomic markers also guide cancer therapy. In this setting it is important to distinguish germline variants, which are inherited and present in all cells (as in the examples above), from somatic variants, which are present only in the tumor (for example EGFR mutations that predict response to EGFR inhibitors, or HER2 amplification and trastuzumab). Tumor genotyping is often considered a separate field, precision oncology, although the concepts overlap.

Hapten, p-i, and Altered Peptide Repertoire Models of T-Cell Activation

Some variants affect the target of a drug or the host response to it. Examples are VKORC1 (the target of warfarin), ADRB2 (the receptor for beta-agonists), and CFTR. Severe immune-mediated reactions to drugs are often related to variants of the human leukocyte antigen (HLA) genes. HLA molecules, which in humans are encoded by the major histocompatibility complex, present peptides to T cells. These reactions are idiosyncratic, meaning that they are unpredictable by the usual pharmacology of the drug and are not strongly related to the dose. They are typically delayed T cell-mediated hypersensitivity reactions.

Several mechanisms have been proposed for how a small drug molecule triggers an HLA-restricted T cell response. In the hapten or prohapten model, the drug or a metabolite binds covalently to a peptide. In the pharmacological interaction (p-i) model, the drug binds non-covalently to the HLA molecule or to the T cell receptor and stimulates T cells directly. In the altered peptide repertoire model, the drug binds in the peptide-binding groove of the HLA molecule and changes the set of peptides that are presented. Abacavir is the best-documented example of the last mechanism.

The best-validated associations are:

• _HLA-B_*57:01 and abacavir, an antiretroviral drug. Carriers have a high risk of a hypersensitivity syndrome with fever, rash, and gastrointestinal and respiratory symptoms. Testing for the allele before starting treatment is standard practice, and treatment is avoided in carriers.
• _HLA-B_*15:02 and carbamazepine, an anticonvulsant. The allele is strongly associated with Stevens-Johnson syndrome and toxic epidermal necrolysis (SJS/TEN) in patients of Han Chinese and Southeast Asian ancestry, in whom the allele is common, and it is rare in people of European ancestry. In these severe reactions, necrosis and detachment of the epidermis occur at the dermal-epidermal junction. Testing before starting is recommended in populations at risk. Most carriers who take the drug do not develop the reaction, so the genotype indicates risk and does not predict that a reaction will occur. _HLA-A_*31:01 is associated with hypersensitivity reactions to carbamazepine in people of European and Japanese ancestry.
• _HLA-B_*58:01 and allopurinol. The allele is associated with severe cutaneous adverse reactions, and it is more frequent in some Asian and African-descent populations.

Enzyme Inhibition, Transcriptional Induction, and Inflammatory Downregulation

The phenotype of an individual may differ from the phenotype that is predicted from the genotype, a situation called phenoconversion. The main causes are:

• Inhibition by other drugs. A drug that inhibits an enzyme can make a genotypic normal metabolizer behave like a poor metabolizer. For example, fluoxetine, paroxetine, and bupropion are strong inhibitors of CYP2D6. The effect reflects the inhibition of the enzyme by the drug or its metabolite, and its onset and offset depend on the half-lives of these compounds. Because fluoxetine and its metabolite norfluoxetine have long half-lives, inhibition develops over days to weeks and persists for weeks after the drug is stopped.
• Induction. Drugs such as rifampicin, carbamazepine, and phenytoin, and some herbal products (St John's wort), induce the expression of enzymes such as CYP3A4, which increases clearance.
• Disease and inflammation. Systemic inflammation, infection, cancer, and liver disease can reduce the expression of several CYP enzymes, in part through cytokines such as interleukin 6. Some changes may involve epigenetic mechanisms such as DNA methylation, but the evidence for the role of epigenetics in drug metabolism is limited, and it is mostly not proved.
• Other factors. Age, pregnancy, smoking (which induces CYP1A2), diet, and the gut microbiome can all affect drug handling.

Because of phenoconversion, genotype should be interpreted together with the list of current drugs and clinical status. Some guidelines consider the effect of strong inhibitors when predicting the phenotype.

Guidelines and databases

Expert groups evaluate the evidence for gene-drug pairs and publish recommendations for how to use a test result, on the assumption that the result is already available. The best known are the Clinical Pharmacogenetics Implementation Consortium (CPIC), which is based in the United States, and the Dutch Pharmacogenetics Working Group (DPWG). These guidelines are peer-reviewed and graded by evidence, and they are not legal requirements. They do not usually address whether a test should be ordered. The United States Food and Drug Administration maintains a table of pharmacogenetic associations in drug labels, and the European Medicines Agency and other regulators also give information. PharmGKB and PharmVar provide curated data. The labels differ in whether testing is required, recommended, or only informative.

Testing strategies

Testing can be done in response to a prescription (reactive or point-of-care testing for a single gene-drug pair) or in advance (preemptive testing), in which a panel of many genes is genotyped once and the results are stored in the health record, where they can be used by clinical decision support at later prescriptions. Genotyping arrays and targeted panels are most common. Sequencing of the whole exome or genome can also be used, but some pharmacogenes are difficult to analyze because of their structure. CYP2D6 is highly polymorphic, has a nearby pseudogene, and has copy number variation. Long-read sequencing and specialized algorithms help to resolve such regions. Results must be reported in a standard way, and, because they are relevant for life, they need to be linked to the record in a form that the clinician sees at the time of prescribing.

Evidence of benefit

For some gene-drug pairs, such as _HLA-B_*57:01 and abacavir, _HLA-B_*15:02 and carbamazepine in at-risk populations, and TPMT, NUDT15, and DPYD with their drugs, genotyping is widely accepted as useful for preventing severe toxicity. For other pairs, such as the use of genotype to select antidepressants or to dose warfarin, results of randomized trials have been mixed, and the benefit depends on the setting, the drug, and the outcome. A large European cluster-randomized crossover trial of a preemptive panel of 12 genes, in which patients received genotype-guided prescriptions, reported a reduction of about 30 percent in the odds of clinically relevant adverse drug reactions. The wide use of preemptive testing is still the subject of research and of debate about cost-effectiveness. Pharmacogenomics is therefore established for some applications in oncology, transplantation, infectious diseases, and cardiology, and its use in other areas, such as psychiatry and pain management, is in different stages of adoption. Statements that it is the standard of care in a whole specialty are not accurate.

Challenges

• Evidence. Many associations come from small studies, and effect sizes differ among populations and studies.
• Diversity. Most genetic studies have used participants of European ancestry, so the variants that are important in other populations are less completely known. This can lead to inequities in the accuracy of tests.
• Complexity. Most drug responses depend on several genes, as well as on non-genetic factors, and polygenic models are still under development.
• Infrastructure and cost. Electronic health records need to store results and to give advice at the time of prescribing. Reimbursement varies by country and test.
• Education. Prescribers and pharmacists need training to interpret results.
• Ethics and privacy. Genetic data are sensitive, and they can reveal information about relatives and about other conditions, so consent, storage, and data protection are important.

Why does CYP2D6 ultrarapid metabolizer status cause opposite clinical outcomes for codeine versus tricyclic antidepressants?

Codeine is an inactive prodrug that requires CYP2D6-catalyzed _O_-demethylation to form its active analgesic metabolite, morphine; consequently, CYP2D6 ultrarapid metabolizers (carrying gene duplications like _CYP2D6_*1/*1×N) generate toxic systemic morphine concentrations that can trigger fatal respiratory depression. Conversely, tricyclic antidepressants (such as nortriptyline) are administered as active parent compounds that CYP2D6 inactivates into polar hydroxy-metabolites, so ultrarapid metabolizers clear the drug rapidly and suffer subtherapeutic treatment failure.

How does the abacavir–HLA-B*57:01 interaction alter the self-peptide repertoire presented to T cells?

Unlike covalent drug haptens, abacavir binds non-covalently deep within the F-pocket of the HLA-B*57:01 antigen-binding cleft (specifically interacting with Ser116 and Asp114). This steric remodeling shifts the chemical specificity of the binding groove, causing the MHC class I molecule to load and display a novel repertoire of endogenous self-peptides (favoring C-terminal tryptophan or phenylalanine over smaller aliphatic residues) that host CD8⁺ cytotoxic T cells misrecognize as foreign, triggering systemic hypersensitivity.

What is phenoconversion in clinical pharmacogenomics?

Phenoconversion is a clinical mismatch in which a patient's actual in vivo drug-metabolizing capacity diverges from their genetic star-allele prediction due to extrinsic physiological or pharmacological modifiers. For example, a patient with a normal-metabolizer _CYP2D6_*1/*1 genotype who is co-prescribed a potent CYP2D6 mechanism-based inhibitor such as fluoxetine or paroxetine undergoes functional conversion into a phenotypic poor metabolizer until enzyme turnover restores baseline hepatic activity.

• Pharmacogenetics
• Personalized medicine
• Cytochrome P450
• Pharmacokinetics
• Pharmacodynamics
• Adverse drug reaction
• Drug metabolism