Role of Pharmacogenomics in Dialysis and Transplantation

Abstract

Purpose of Review

Pharmacogenomics is the study of differences in drug response based on individual genetic background. With rapidly advancing genomic technologies and decreased costs of genotyping, the field of pharmacogenomics continues to develop. Application to patients with kidney disease provides growing opportunities for improving drug therapy.

Recent Findings

Pharmacogenomics studies are lacking in patients with chronic kidney disease and dialysis but are abundant in the kidney transplant field. A clinically actionable genetic variant exists in the CYP3A5 gene, with the initial tacrolimus dose selection optimized based on CYP3A5 genotype. Though many pharmacogenomics studies have focused on transplant immunosuppression pharmacokinetics, an expanding literature on pharmacodynamic outcomes like calcineurin inhibitor toxicity and new onset diabetes is providing new information on patients at risk.

Summary

Appropriately powered pharmacogenomics studies with well-defined phenotypes are needed to validate existing studies and unearth new findings in patients with kidney disease, especially the chronic kidney disease and dialysis population.

Introduction

Pharmacogenomics provides insight into the inter-individual variability in drug response based on one’s genetic background. The term pharmacogenetics is used to describe a single variation, or polymorphism, in the genome and how this affects drug response. When multiple genetic variants or even the entire human genome are considered, as has been made possible with our current advanced technologies, the term pharmacogenomics is used. Increasingly more complex scenarios in drug response are being considered, including combinations of genetic variants or gene-gene interactions, the influence of the microbiome, and epigenetics.

An obvious application of pharmacogenomics is to avoid serious adverse drug reactions. Additional goals include selecting the best drug when choices are available, optimal dose of a drug, and cost-savings for ineffective or potentially harmful therapy. Identifying genetic variation in drug response also may shed light on new targets for drug development.

Intense research in pharmacogenomics has been undertaken in certain fields (1), but less has been done in patients with kidney disease, with the exception of transplant. Additionally, pharmacogenetic studies of commonly used drugs often exclude patients with significant kidney disease, even though patients with kidney disease have a high burden of medical comorbidities that leads to many drug exposures. This review will provide the pertinent background and update of the most recent pharmacogenomics studies relevant to patients with kidney disease, while highlighting the needs for future research.

General pharmacogenomics principles

When prescribing a drug to a patient, many factors are considered that affect drug response, including patient age, sex, comorbid diseases, other medications, dietary supplements, and nutritional status. Pharmacogenomics incorporates another valuable factor by considering an individual’s genetic background. While a single variation in an individual’s genome may have a relatively large effect on drug response, in many cases multiple variants with small effects may contribute a large cumulative effect. One challenge of pharmacogenomics has been the ability to measure these multiple variants to explain an observed drug response, often because study cohorts are underpowered.

Genetic variation can affect both pharmacokinetics and pharmacodynamics of a drug, and these outcomes are used as the phenotypes in pharmacogenomics. Pharmacokinetics refers to actions of the body on a drug, including absorption, distribution, metabolism, and excretion (ADME) of an administered drug that are carried out by proteins like membrane transporters and metabolizing enzymes, including the cytochrome P450 system. Common pharmacokinetic phenotypes include drug concentration, dose, clearance, and area under the curve. Pharmacodynamics is action of a drug on the body and includes the intended drug binding target as well as unintended targets that may cause unwanted side effects.

When studying genetic variation, one may take a hypothesis-driven approach, where one or more candidate genes are selected based on known or suspected pharmacokinetic or pharmacodynamic mechanisms. But, with high throughput genomic technologies, a hypothesis-free discovery approach is used in which large numbers of genetic variants (thousands to millions) are examined simultaneously for a chosen drug phenotype, like for genome-wide association studies (GWAS) or whole exome sequencing studies. Important considerations in pharmacogenomics studies include: 1) accounting for ethnic background, since frequency of genetic variants vary by race; 2) study power, as many are underpowered especially with multiple comparisons adjustments for genomics studies; 3) definition of the phenotype of interest, which can vary widely from study to study making comparisons difficult; and, 4) presence of validation studies.

Pharmacogenomics of dialysis and chronic kidney disease

To date, pharmacogenomics has not been intensely applied to chronic kidney disease (CKD) and dialysis patients. While recent genomic studies of these populations exist, they are centered on kidney disease risk or progression, associated dialysis complications like arteriovenous fistula failure, or patient survival (2–4) rather than phenotypes related to drug response. But, a few potentially important pharmacogenomics associations have been explored. Notable are the renin-angiotensin system (RAS) blocker drugs, angiotensin converting enzyme inhibitors (ACEI) or angiotensin receptor blockers (ARBs), first line agents for the treatment of high blood pressure in CKD. Several genetic variants in RAS have been described, including the well-studied angiotensin converting enzyme insertion deletion (ACE I/D, rs4646994) that affects ACE levels. Early pharmacogenetic studies supported an effect of ACE I/D genotype on blood pressure response in patients taking ACEI and ARBs, though later studies have not, as recently reviewed (5). As RAS polymorphisms have not been consistently associated with antihypertensive response, currently no indication for clinical implementation exists (6).

A potential area of application would be pharmacogenomics of immunosuppression for glomerulonephritis (GN) based on several studies in kidney transplantation. Such studies should be repeated in the GN populations given other confounding factors exist, including lack of co-therapy with calcineurin inhibitors and differences in kidney function. For example, in a study of 39 patients with GN from lupus nephritis or ANCA-associated vasculitis taking mycophenolate, UGT2B7, UGT1A, and ABCB1 genotypes were not associated with mycophenolate pharmacokinetics (7) as has been reported in kidney transplant (8).

Clearly the field is wide open for exploration in CKD and dialysis. Looking at extreme phenotypes, like dialysis patients who do and do not respond to erythropoietin, may be a good starting place, as it would be reasonable to think genetic variation in drug response might explain some of the variability. For example, Derebail et al. recently described presence of hemoglobinopathy and need for higher erythropoietin dose requirement in African Americans on dialysis (9). Pharmacogenomics studies admittedly are complex in this patient population with many confounders, including drug interactions and change in pharmacokinetics with renal failure. To date, there does not appear to be any specific drug recommendations to be made for CKD and dialysis patients based on pharmacogenomics studies.

Pharmacogenomics of kidney transplant

Modern immunosuppressive therapy for kidney transplantation, including calcineurin inhibitors, mycophenolate, mammalian target of rapamycin (mTOR) inhibitors, with or without steroids has allowed graft survival rates of 97% and acute rejection incidence of less than 10% at one year (10). Clinical use of these drugs, however, has been complicated by their narrow therapeutic index, where underexposure may increase acute rejection risk and overexposure may lead to toxicity (11). Often large pharmacokinetic inter-individual variability is noted. As a result, therapeutic drug monitoring is routinely used, with dose adjustments based on blood concentrations (11). Though clearly useful, therapeutic drug monitoring limitations include unavailability for some drugs and inability to use for optimal starting dose. Furthermore, even when target drug concentrations are achieved, levels do not directly correlate with efficacy or toxicity. This was recently described by Bouamar et al., who did not find an association between tacrolimus trough blood concentrations and acute rejection incidence during 6 months post kidney transplant (12). This may be partially explained by the fact that whole blood concentrations do not accurately reflect important pharmacodynamics effects of a drug, for example intracellular lymphocyte concentrations for an immunosuppressive medication. Given these multiple caveats, investigators have used pharmacogenomics approaches in hopes of better understanding and using transplant immunosuppression, with the majority of studies focused on calcineurin inhibitors.

Calcineurin inhibitors and pharmacokinetic phenotypes

The calcineurin inhibitors cyclosporine and tacrolimus are substrates for the ABCB1 drug efflux transporter encoded by ABCB1 gene and are metabolized by the cytochrome (CYP) P450 3A family enzymes, particularly CYP3A4 and CYP3A5. In the CYP3A5 gene, a single nucleotide polymorphism (SNP) may occur (rs776746, 6986G>A) that affects CYP3A5 expression. In individuals with at least one *1 allele, functional CYP3A5 protein is present (CYP3A5 expressers), whereas those homozygous for *3 have no CYP3A5 expression (CYP3A5 non-expressers) (13). For tacrolimus, it has been well established through both candidate gene and larger genomic studies that CYP3A5 non-expressers have higher dose-adjusted trough blood concentrations and lower dose requirements than CYP3A5 expressers, with findings consistent across race groups and adult and pediatric kidney transplant recipients (14–19). Recent studies confirming this include a study of 177 kidney transplant recipients in Norway, 206 recipients from northern Spain, as well as a one of Korean recipients in which both tacrolimus and its metabolites were associated with CYP3A5 status, among others (20–24). In addition to high inter-individual variability, some transplant recipients also have marked intra-individual variability with tacrolimus, which has been identified as a risk factor for nephrotoxicity. Recipient CYP3A5 has not been associated with tacrolimus intra-individual variability in kidney transplant recipients (25), reinforced in a recent study of 118 kidney transplant recipients followed 1 year post transplant (26).

With 39% of tacrolimus inter-individual variability explained by CYP3A5 (15) investigators have sought other contributing variants. Past studies have focused on the ABCB1 polymorphisms, rs1045642 (3435C>T), rs2032582 (2677G>A/T), and rs1128503 (1236C>T), which are in linkage disequilibrium and inherited as a haplotype (13). Results of these SNPs on tacrolimus pharmacokinetics are conflicting and generally appear not be relevant when considered individually (23, 27). A recent study of 108 Brazilian kidney transplant recipients did show recipients homozygous for the T allele (TTT/TTT) at all 3 SNPs had higher dose-normalized trough blood concentrations than controls (CGC/CGC or CGC/TTT), even when stratified by CYP3A5 status (28). Newer studied variants include CYP3A4*22, POR*28, and PPARA. In CYP3A4*22(rs35599367, C>T), the T allele is associated with decreased tacrolimus dose requirement (23, 29). Elens et al. showed this effect was independent and additive when stratifying patients by CYP3A5 status. However, others have been unable to confirm this finding (22, 30, 31). Another recently highlighted variant, important in CYP-mediated drug oxidation, is POR*28 (rs1057868, C>T), with the T allele associated with increased tacrolimus dose requirement. In these studies, POR*28 was only relevant in patients who were also CYP3A5 expressers (21, 32–34). The gene PPARA encodes the perixosome proliferator-activated receptor alpha (PPAR-α), and variants rs4253728, G>A and rs4823613, A>G affect CYP3A activity. Initial studies of PPARA and tacrolimus show conflicting results (21, 35). These newly described variants will need further validation across populations before deciding their clinical importance. In addition, further studies assessing the interaction among multiple genetic variants will be needed to realize the full potential of pharmacogenomics, such as the relevance of POR*28 only in CYP3A5 expressers above. Please see Table 1 for a summary of variants related to transplant pharmacokinetics.

Table 1.

Summary of discussed pharmacogenetic variants related to pharmacokinetics of kidney transplant immunosuppression^*

Drug	Genotype (rs number)	Phenotype	Reference
Tacrolimus	CYP3A5*3 (rs776746)	*3 homozygotes (CYP3A5 non-expressers) have higher dose-adjusted trough blood concentrations and lower dose requirements compared to *1 carriers; evaluated in a randomized control trial with no effect on 3 month outcomes	(13–24, 36)
	ABCB1 3435C>T (rs1045642) 2677G>A/T (rs2032582) 1236C>T (rs1128503)	No clear effect on pharmacokinetics	(23, 27)
	CYP3A4*22 (rs35599367)	May explain additional variability in dose-adjusted blood concentration in combination with CYP3A5*1/*3	(22, 23, 29–31)
	POR*28 (rs1057868)	Associated with lower dose-adjusted blood concentration but only in CYP3A5 expressers	(21, 32–34)
	PPARA (rs4253728 and rs4823613)	Associated with dose adjusted blood concentration but results conflicting	(21, 35)
Cyclosporine	CYP3A5*3 (rs776746)	No clear effect on pharmacokinetics	(13)
	ABCB1 3435C>T (rs1045642) 2677G>A/T (rs2032582) 1236C>T (rs1128503)	No clear effect on pharmacokinetics	(27)
	CYP3A4*22 (rs35599367)	Associated with higher dose-adjusted blood concentration	(21, 29, 30)
	POR*28 (rs1057868)	Associated with lower dose-adjusted blood concentration in CYP3A5 non-expressers	(32)
Mycophenolate	UGT1A9 (rs6714486) (rs17868320)	Associated with lower MPA exposure	(8, 37)
Sirolimus	CYP3A5*3 (rs776746)	Associated with higher dose-adjusted blood concentration	(38)
Everolimus	CYP3A5*3 (rs776746)	No clear effect on pharmacokinetics	(30, 39)

^*List not exhaustive of all published literature but reflects recent publications or well established relationships

Abbreviations: MPA, mycophenolic acid

Despite the strong association of CYP3A5 with tacrolimus pharmacokinetics, it remains unclear if dosing by genotype will improve transplant outcomes, especially since therapeutic drug monitoring allows quick adjustment of dose. Randomized controlled trials showing benefit are needed, but only one is completed. Thervet, et al. showed genotyped-guided dosing did achieve more kidney transplant recipients reaching target tacrolimus level by day 3 compared to controls, but no differences were seen in patient/graft survival, nephrotoxicity, or acute rejection over 3 months follow up (36). More studies like this are needed with longer duration of follow up and carefully phenotyped populations to see if certain sub groups benefit, like a high versus low risk group for acute rejection. In addition, validated dosing algorithms incorporating clinical and genetic factors that work across multiple cohorts are needed. This was exemplified by a published equation for tacrolimus dosing, which validated in one patient cohort but was not able to be validated in another (40–42). Lack of a universal algorithm probably reflects in part our lack of full understanding of factors that contribute to tacrolimus drug variability (43).

The impact of polymorphisms on cyclosporine pharmacokinetics has been less clear due to conflicting reports, as recently reviewed (13). A meta-analysis of CYP3A5*3 in 1821 kidney transplant recipients concluded that CYP3A5*3 may have a small effect on cyclosporine pharmacokinetics but is not likely clinically relevant (44). Likewise, a recent review of ABCB1 determined a lack of consistent effect on cyclosporine pharmacokinetics, though it might be important in nephrotoxicity (discussed below) (27). Unlike for tacrolimus, CYP3A4*22 has been consistently associated with higher dose-adjusted trough blood concentrations for cyclosporine, indicating need for lower dose (21, 29, 30). Cyclosporine treated patients who are CYP3A5 non-expressers and homozygous for POR*28 may have higher CYP3A4 activity, leading to lower dose-adjusted trough blood concentrations and need for higher dose (32), but further confirmatory studies are needed. Based on studies to date, dosing of cyclosporine based on genetic variants is not warranted.

Calcineurin inhibitors and pharmacodynamic phenotypes

Blood drug concentrations may not directly correspond with pharmacodynamic outcomes, as observed clinically when a patient experiences acute rejection or toxicity despite drug levels being in therapeutic range. Thus pharmacogenetic study of pharmacodynamic outcomes in kidney transplant is of increasing interest, though overall less well studied. A summary of these are in Table 2. Acute rejection is commonly investigated, but studies to date have not shown a strong association of pharmacogenes with acute rejection in kidney transplant recipients treated with calcineurin inhibitors, as lately reviewed (13, 44). A recent meta-analysis confirmed that CYP3A5 and ABCB1 have minimal effect on acute rejection in tacrolimus treated patients (24).

Table 2.

Summary of discussed pharmacogenetic variants related to pharmacodynamics of kidney transplant immunosuppression^*

Pharmacodynamic outcome	Genotype (rs number)	Result	Reference
Acute rejection	CYP3A5*3 (rs776746)	No clear effect with calcineurin inhibitors	(13, 24, 36, 44)
	ABCB1 3435C>T (rs1045642) 2677G>A/T (rs2032582) 1236C>T (rs1128503)	No clear effect with calcineurin inhibitors	(13, 24)
	UGT1A9 (rs6714486) (rs17868320)	May be associated with acute rejection as a result of MPA underexposure	(8, 37)
	IMPDH1	May be associated with acute rejection in mycophenolate treated patients	(8, 45)
Calcineurin inhibitor nephrotoxicity	ABCB1 3435C>T (rs1045642) 2677G>A/T (rs2032582) 1236C>T (rs1128503)	No clear effect based on recipient genotype Related to nephrotoxicity based on donor genotype	(23, 27, 46)
New onset diabetes	POR*28 (rs1057868)	Associated with increased risk of NODAT	(47)
	PPARA (rs4253728 and rs4823613)	Associated with increased risk of NODAT	(47)
	Multiple SNPs in genes related to β-cell apoptosis	Associated with increased risk of NODAT	(48)
Sirolimus-related hyperlipidemia	ABCB1 3435C>T	Associated with higher total and low-density lipoprotein level-C cholesterol	(49)

^*List not exhaustive of all published literature but reflects recent publications

As a major contributor to chronic allograft failure, calcineurin inhibitor nephrotoxicity remains an important modifiable cause. Notably, blood levels may not reflect local accumulation of calcineurin inhibitors in renal tissue and risk of nephrotoxicity. For example, Zheng et al. showed that in healthy volunteer CYP3A5 expressers and nonexpressers, cyclosporine oral clearance was similar, but CYP3A5 expressers had 50% higher area under the concentration time curves of active cyclosporine metabolites, with decreased urinary clearance, suggesting increased intra-renal accumulation that could contribute to nephrotoxicity (50). Another study of cyclosporine treated kidney transplant recipients showed that even though cyclosporine blood levels were not associated with ABCB1 polymorphisms, those homozygous for 3435TT had an increased odds of nephrotoxicity (OR 4.2, 95% CI 1.3–13.9) and gingival hyperplasia, again suggesting blood levels may not fully reflect exposure at target sites (51). Overall, however, pharmacogenetic studies of recipient CYP3A5 and ABCB1 in calcineurin inhibitor toxicity have shown mixed results, making it difficult to draw conclusions.

On the other hand, donor genotype may be important in nephrotoxicity. This review so far has focused on recipient genotype, but CYP3A5 and ABCB1 and other pharmacogenes are expressed in the kidney, so that the donor kidney genotype may differ from that of the recipient, affecting levels of calcineurin inhibitors at the target organ. A study of 4,471 white kidney transplant recipients on calcineurin inhibitors by Moore et al. showed an association of donor ABCB1 3435CC and increased long term graft failure, though this was in contrast to smaller studies that had shown the 3435TT genotype to be worse (46). Further pharmacogenomics studies considering donor-recipient genotypes are warranted (52).

New onset diabetes after transplantation (NODAT) is another toxicity observed with tacrolimus but also seen with cyclosporine and mTOR inhibitors. In addition to well established clinical factors like age, gender, and weight gain, genetics likely play a role with many variants recently described but not necessarily validated, reflecting small if real effects. One recent paper by McCaughan described 7 SNPs in genes related to β-cell apoptosis (48). Other recent studies have implicated POR*28 and PPARA (47), adiponectin gene (rs1501299, 276G>T) (53), mitochondrial haplogroup H (54), and Fok1 vitamin D receptor polymorphisms. Many of the described variants have plausible biologic connections to glucose metabolism pathways and should be further studied to validate findings and explain the pathogenesis of NODAT (55).

Pharmacogenomics provides novel ways for monitoring drug efficacy. For example, Yoon and colleagues have showed that in monocytes from kidney transplant recipients homozygous for CYP3A5*3, the HLA/DR+ mean fluorescent intensity was significantly lower compared to those from CYP3A5*1 carriers, with significant negative correlations with dose-adjusted trough blood concentrations, suggesting the HLA/DR expression on monocytes might be explored as a tool for monitoring of tacrolimus toxicity (56). In another study, Vafadari et al. a showed the pharmacodynamic effect of tacrolimus on inhibiting IL-2 production depended on ABCB1 3435C>T genotype (57).

Other immunosuppressive drugs: mycophenolate and mTOR inhibitors

Mycophenolate mofetil (MMF) is a prodrug that is hydrolyzed to the active form mycophenolic acid (MPA). The pharmacokinetics of MPA are complex, including hepatic conjugation by uridine diphosphate glucuronosyl transferases (UGTs), producing active and inactive metabolites that may enter enterohepatic circulation. Most studies in kidney transplant recipients have focused on polymorphisms in the UGT isoforms, particularly in UGT1A9 and UGT2B7, as well as genes encoding organic anion transporters (OATs), like SLCO1B3, which have been recently reviewed (8, 37). Several studies have reported UGT1A9 SNPs (rs6714486, 275T>A and rs17868320, 2152C>T) are associated with lower MPA exposure though conflicting evidence exists. Interestingly, in some studies, the effect of UGT polymorphisms appears to depend on which calcineurin inhibitor is present (58). Genetic variation in the target of MPA, inosine 5-monophosphate dehydrogenase (IMPDH), has also been studied with some but not all studies showing association with acute rejection (8, 45). The commonly recognized toxicities of MPA including leukopenia, anemia, and diarrhea have been examined with mixed results (8, 59). In summary, though provocative, pharmacogenetic variants are not yet indicated for clinical implementation for mycophenolate.

The mTOR inhibitors sirolimus and everolimus have been the least studied drugs in transplant pharmacogenomics. Like the calcineurin inhibitors, they are substrates of the ABCB1 drug efflux transporter and CYP3A4/CYP3A5 enzymes (45). Several studies have shown that patients homozygous for CYP3A5*3 have higher dose-adjusted sirolimus levels (38), but this has not been as widely validated as for tacrolimus. CYP3A5*3 has not been associated with everolimus pharmacokinetics (30, 39). A common adverse effect of sirolimus therapy is hyperlipidemia, and one study group has shown an association of ABCB1 3435C>T with increased sirolimus blood levels as well as with higher total cholesterol and low-density lipoprotein level-C in kidney transplant recipients (49).

Conclusion

Pharmacogenomics has been limited in the CKD and dialysis population, whereas a growing wealth of information is available for kidney transplantation. The only actionable genetic variant to date appears to be the CYP3A5 genotype for initial tacrolimus dosing in kidney transplant. But, the long term impact of adjusting tacrolimus dose based on CYP3A5 genotype requires further study. Though pharmacogenomics in kidney disease has not proven to benefit clinical practice as quickly as once hoped, value exists in pushing ahead in this exciting field. Further studies assessing the interaction among multiple genetic variants will be needed to realize the full potential of pharmacogenomics. Some investigators have begun such inquiries, such as the relevance of POR*28 only in CYP3A5 expressers taking tacrolimus (32), while other efforts to combine genotypes have not improved results (30). For transplant, dual consideration of donor and recipient genotype is vital. As the focus moves to donor-recipient gene interactions, gene-gene interactions, and gene-environment interactions in the setting of increasingly sophisticated genomic technologies and computational analysis, much is yet to be discovered and tested to improve our management of patients with kidney disease when it comes to drug therapy.

Key points.

• Pharmacogenomics studies in CKD and dialysis patients are lacking, with investigation needed both for drugs specific to kidney therapy (like erythropoietin) as well as for drugs commonly used in these populations due to common comorbidities.

• To date, pharmacogenomics recommendations for ACEI and ARBs and blood pressure response based on RAS polymorphisms cannot be made.

• Adjusting the initial tacrolimus dose in kidney transplant recipients based on CYP3A5 genotype may be useful, but studies on best dose and beneficial effect on long term outcome are still warranted. A higher dose is needed for CYP3A5 expressers.

• Newly described variants important to tacrolimus pharmacokinetics (CYP3A4*22, POR*28, and PPARA) as well as multiple variants for pharmacodynamic outcomes like calcineurin inhibitor toxicity and NODAT await validation.

• For clinical implementation, pharmacogenetic variants must be validated in appropriate populations and show consistent evidence, preferably in the form of randomized control trials, that implementation affects outcomes.