CYP2C19 genotype-phenotype correlation: current insights and unanswered questions

Introduction

Pharmacogenetics as a molecular tool for precision medicine is already a reality in many countries. With prescribing decisions being supported by the patient’s genotype, pharmacogenetics improves both efficacy and safety outcomes [1]. However, this is not the scenario for all countries. While countries with developed economies are currently concerned about adopting electronic health records and other emerging resources to integrate pharmacogenomics into routine care practice [2], 3], resource-limited environments still experience insufficient knowledge. This disparity means that most of the allele frequency data and their associated evidence for clinical utility come from populations other than their own [4], 5].

An illustrative instance is the CYP2C19 enzyme, a member of the cytochrome P450 family. As it is implicated in the metabolism of about 10 % of clinically used drugs [6], 7], this enzyme is of particular interest in the field of clinical pharmacology. As a reflection of its relevance, five of the 27 Clinical Pharmacogenetics Implementation Consortium (CPIC) guidelines currently available specifically provide recommendations regarding CYP2C19-drug interactions: voriconazole [8], proton pump inhibitors [9], antiplatelet agents [10], and antidepressants [11], 12].

And this is precisely where the issue lies. The 36 alleles currently recognized by the Pharmacogene Variation Consortium were identified from research in populations of predominantly European ancestry. Consequently, the evidence available for clinical pharmacogenetics primarily applies to this specific group.

CYP2C19 predicted phenotype

To enable the clinical integration of pharmacogenetics information into the concept of precision medicine, it is required to investigate beyond an individual’s genotype and also comprehend their metabolizer phenotype. However, phenotype analyses based on monitoring the concentration ratio of the probe drug to its corresponding CYP-specific metabolite demand significant costs and analysis time [13]. Therefore, as the genome remains invariant throughout the course of life, the process of translating genotype data into a phenotype prediction was proposed by CPIC as a simpler approach [14]. According to the genotyping test results, the predicted phenotype indicates the expected enzymatic activity for each patient, which can be used as therapeutic guidance. This prediction, however, was developed based on European individuals’ evidence.

As previously observed in studies with Latin-American populations, there is a remarkable discordance between the genotype-estimated (predicted) phenotype and the measured drug metabolism (metabolic) phenotype for several participants from Mexico, Nicaragua, and Ecuador [15], [16], [17]. In these population groups, only poor metabolizers exhibited correspondence between their genotype-predicted phenotype and the actual oxidation capacity. Substantial overlap was found among the other metabolizing phenotypes, especially the ultrarapid ones [15], [16], [17], suggesting that genotype alone may not reliably predict the CYP2C19 metabolizer status for all other phenotypes except poor metabolizers.

Genotype-phenotype discordance

The patient’s metabolic profile is multifactorial and, therefore, influenced by conditions beyond genetics, potentially altering the expected phenotype. These other factors include drug interactions, epigenetic factors, dietary habits, and pathological conditions that may influence enzyme expression (Figure 1).

Figure 1:

Possible reasons for discordance between the predicted phenotype and the metabolic phenotype.

In the first case, drug interactions can lead to phenoconversion, a phenomenon characterized by a mismatch between the genotype-based prediction of phenotype and the actual capacity of CYP450-mediated drug-metabolizing, attributed to extrinsic nongenetic factors [18]. While some drugs are recognized as inhibitors of CYP2C19 (for example, fluconazole, fluoxetine, and fluvoxamine), potentially phenoconverting individuals to intermediate or poor metabolizers [19], there are also drugs known to be inducers of CYP2C19 (rifampin, and efavirenz), which in turn may stimulate the metabolism of drugs that are substrates of this enzyme. These interactions are irrelevant for CYP2C19 poor metabolizers, who have absent enzyme activity [20]. Therefore, the extent of drug interaction is mainly determined by the drugs used concomitantly, the patient’s metabolizer phenotype and the corresponding contribution of CYP2C19 to the metabolism of the administered treatment [20].

Phenoconversion is of particular concern among polypharmacy patients, and although it should be included in pharmacogenetic results interpretation, its effects are often underestimated in clinical practice [18]. Among other factors, negligence is a consequence of insufficient information on phenoconversion and guidance for the clinical interpretation of phenotypes after this interaction has been considered.

In terms of epigenetic mechanisms, they modulate gene expression without altering the DNA sequence, but instead by recruiting regulatory machinery [21]. The main epigenetic mechanisms are represented by histone modifications (often acetylation), DNA methylation, and non-coding RNAs [22]. Histone acetylation, which consists of adding acetyl groups to histone tails, can activate or inactivate the adjacent genes by increasing DNA accessibility to transcription factors. The combined effect of a DNA methyltransferase inhibitor, 5-aza-2′-deoxycytidine, and an inhibitor of histone deacetylases, trichostatin A, on the expression of CYP2C19 has been assessed in an in vitro study. Post-treatment results indicated CYP2C19 mRNA was significantly up-regulated in cell lines [23].

On the other hand, mechanisms such as DNA methylation and microRNAs (miRNAs) mostly behave as epigenetic silencers [22]. Methylation of CpG islands of a target gene can influence the binding affinity of transcription factors to the promoter region [22], while miRNAs may post-transcriptionally regulate protein expression usually by binding to the 3′-untranslated region of the target mRNA and thereby inhibiting protein translation [24]. For example, the expression of four CYP450 isoenzymes (CYP2C8, CYP2C9, CYP2C19, and CYP3A4) was decreased by miRNAs miR-452–5p and miR-224–5p during inflammation [25].

Pathological processes may also contribute to interindividual variability in drug clearance through downregulation of CYP450 enzymes. Indeed, patients with an acute or chronic inflammatory disease may experience the phenomenon of phenoconversion, attributed to cytokines that impair enzyme expression [26], [27], [28]. The severity of this suppression is directly associated with the extent of inflammation and the cytokine profile [28]. Infections caused by agents have also been already evaluated; the SARS-CoV-2 infection was shown to modulate the CYP activity in an isoform-specific manner [29], while the Plasmodium vivax malaria altered CYP2C19 metabolic phenotypes, probably through increasing circulating proinflammatory cytokines [30].

CYP2C19 and under-represented populations

Differences in allele frequencies could also explain disparities in precision medicine and lack of replication across populations, as phenotype prediction mostly depends on the alleles examined and their respective frequencies within the population of interest.

As previously reported [1], 31], there is considerable inter-ethnic heterogeneity in pharmaco-alleles frequencies, with certain variants being ethnic-specific. Therefore, when investigating groups of individuals sharing common ethnicity or ancestry, the frequency of genetic variants, as well as their combination within a haplotype, may be unpredictable in admixed populations compared to parental populations. Without extensive genotyping or whole genome sequencing approaches, the less frequent polymorphisms of underrepresented populations may remain undetected, while the standard ones may be overestimated.

The predominance of research focused on European ancestry has resulted in an inadequate representation of worldwide genetic diversity [32], 33]. When rarely studied, ethnically diverse populations are frequently gathered in meta-analyses, therefore missing particular genetic patterns. Although combining data is an excellent analytical strategy to identify variants with consistent effects across populations, it may underlook variants that are rare in other populations but prevalent in these diverse groups [32]. Therefore, the discordance between the genotype and phenotype may also be attributed to either insufficient comprehensive genotyping or unknown variants [5].

Extrapolating data from one ethnic group to another is most challenging for functionally significant alleles that exhibit considerable differences in ethnic frequencies, in which cases more significant interethnic disparities in drug outcomes may be observed [1]. Identifying alleles exclusive of under-represented populations would assist in prioritizing pharmacogenetic variants with significant clinical impact in those populations and, therefore, optimize the allocation of limited resources. Otherwise, admixed populations will always remain on the sidelines.

Furthermore, the construction of pharmacogenetics clinical guidelines currently available mainly were based on European data. Consequently, these guidelines could inadequately predict outcomes in different ethnicities, and patients who carry variants not addressed by such guidelines could experience more harm than benefit. Therefore, there is an urgent need for analyzing populations other than those already well characterized, also employing next-generation sequencing to identify new clinically actionable variants. Without sufficient evidence, it is not possible to establish guidelines appropriate to these groups.

CYP2C-TG haplotype

A recent notable advancement in CYP2C19 pharmacogenetics is a new haplotype described, referred to as “CYP2C:TG”. It is encoded by two non-coding variants (rs2860840C>T and rs11188059G>A) in the CYP2C18 gene, located in the CYP2C gene cluster, and was associated with increased CYP2C19 enzyme activity. In patients previously classified as CYP2C19 normal metabolizers, the metabolic rate of sertraline and escitalopram levels, both CYP2C19 substrates, were at a similar extent as CYP2C19*17 carriers (rapid or ultrarapid metabolizers) [34]. It means although a patient may be classified as having predicted normal CYP2C19 metabolism according to current guidelines, the CYP2C:TG haplotype can increase the metabolism to rapid or ultrarapid [34]. Therefore, these individuals would need special attention, possibly requiring dose adjustment or switching to alternative medications, as with CYP2C19*17 carriers.

Shortly afterward, another study indicated that this same haplotype was also associated with treatment failure of omeprazole in gastroesophageal reflux disease. Those findings may suggest that refractory patients carrying CYP2C:TG haplotypes could benefit from increased omeprazole doses [35].

It is pertinent to mention that both investigations included European cohorts [34], 35]. Therefore, while further research in larger cohorts is required to replicate and validate these findings, it is also recommended that such investigations also include diverse populations. It would also be interesting to evaluate the response of individuals carrying this haplotype to other substrates of this enzyme.

Prediction of CYP2C19 ultrarapid metabolism, ethnicity and clinical implications

Determining the exact relationship between CYP2C19 genotype and phenotype is very important. This enzyme is involved in the metabolism of therapeutically relevant diseases and drugs, such as risk of depression [36], 37] or suicide [38], 39], or treatment failure with antidepressants [40], antipsychotics [41], 42] or antiepileptics [43]. CYP2C19 genetic polymorphism has also been involved in the synergy between clopidogrel and calcium-channel blockers for regulation of blood pressure [44].

Despite the importance of understanding inter-ethnic and inter-individual variability in drug response, there are few studies of genetic polymorphisms of drug metabolizing, transporter, and receptor enzymes in the various ethnic groups in Latin America [45], Therefore, in a previous study we have analyzed the frequency of genetic polymorphisms of drug-metabolizing enzymes in the Latin American populations [46] of the Dominican Republic [47], Nicaragua [16], Ecuador [15], and Mexico [17]. Additionally, the existence of ‘phenocopies’ should be taken into consideration [48].

Conclusions and future perspectives

While some countries are already considering the possibility and potential benefits of preventive pharmacogenetics testing [49], low- and middle-income countries may still be at earlier stages of clinical pharmacogenetics implementation [4]. Furthermore, as the historical context of many of these countries has resulted in diverse admixed populations, an additional source of variation may influence the association of polymorphisms to the phenotype of interest. The predominance of research conducted on populations of European ancestry represents a bias, which effectively translates into inadequate phenotypic prediction for individuals of under-represented ancestries.

Supposing that pharmacogenetic studies continue to deprioritize these groups, identifying rare alleles in populations of European ancestry (but of significant relevance in underrepresented populations) will become unfeasible. There is an urgent need to include ethnically diverse groups in human genetic studies to promote equity in human pharmacogenomics and improve precision in medical care.

Extrapolating data from genetically homogeneous ethnic groups to admixed populations is not recommended. To precisely predict phenotype from genotype, the patient must undergo testing for variant alleles pertinent to their ethnic origin. Taking advantage of the technological advances in genomics, a growing number of unidentified and clinically actionable variants could be identified. It is also expected that results from ongoing or future studies could contribute significantly to the existing knowledge gap by identifying new ethno-specific variants of clinical relevance, as well as validating previous findings from studies in other populations.

It will require, for sure, a coordinated and intentional effort by both the international research communities and funding agencies to increase representation in genetic databases. This effort involves including ethnically diverse populations in further studies and prioritizing financial support for research in historically neglected nations. It is essential to ensure that genomic data reflects ethnic diversity, allowing for more precise clinical application and reducing global health inequities. So, in the long term, it will be feasible to propose a panel of allelic variants specific to each population.