Identification of a Functional CYP2C8 Variant Allele that Alters Splicing, Reduces Protein Expression, and Increases Drug Exposure

Abstract

This study investigated genetic determinants of the pharmacokinetics of the CYP2C8 index drugs repaglinide and gemfibrozil, and their interaction in healthy participants. Sequencing data from a study with montelukast revealed a novel functional CYP2C8 allele (rs2071426, CYP2C8*19), predicted to create an intronic splice donor site. In human liver samples, CYP2C8*19 associated with transcript‐specific changes in CYP2C8 mRNA expression, reduced CYP2C8 protein expression, and decreased enzyme activity. Consistently, participants with the CYP2C8*19/*19 genotype had 45% greater area under the plasma repaglinide concentration–time curve from time zero to infinity (AUC_0‐∞) than participants with CYP2C8*1/*1 (P = 1.6 × 10⁻⁴). Participants with CYP2C8*1/*3 had 26% smaller AUC_0‐∞ (P = 0.0033) and those with CYP2C8*1/*4 had 51% greater AUC_0‐∞ (P = 8.2 × 10⁻⁴). The fold increase in repaglinide AUC_0‐∞ caused by gemfibrozil was 36% (P = 1.3 × 10⁻⁴) smaller in CYP2C8*19/*19 participants than in CYP2C8*1/*1 participants. In a genome‐wide association study (GWAS), SLCO1B1 c.521 T>C (rs4149056) associated with increased repaglinide AUC_0‐∞ (P = 4.5 × 10⁻¹⁵; n = 172) and SLCO1A2 variants associated with decreased AUC_0‐∞ (P < 10⁻⁸). In a GWAS of repaglinide after gemfibrozil pretreatment, SLCO1C1 variants associated with decreased AUC_0‐∞ (P < 1.6 × 10⁻⁸; n = 66). Participants with the poor function SLCO1B1 genotype showed a 32% smaller fold increase in repaglinide AUC_0‐∞ following gemfibrozil than participants with the normal function SLCO1B1 genotype (P = 0.0045). This study characterizes CYP2C8*19 as a novel decreased function allele and shows that CYP2C8 and SLCO1B1 genotypes affect the gemfibrozil–repaglinide interaction.

Study Highlights.

WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Repaglinide is an index CYP2C8 substrate. Considerable variability exists in repaglinide pharmacokinetics, and genetic variation explains part of the variability. Gemfibrozil is an index CYP2C8 inhibitor, which increases the AUC_0‐∞ of repaglinide ~ 7–8‐fold.

WHAT QUESTION DID THIS STUDY ADDRESS?

What are the most important genetic variants influencing the pharmacokinetics of repaglinide and gemfibrozil and their interaction?

WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

This study characterized a novel CYP2C8 allele, CYP2C8*19, which results in an alternative CYP2C8 transcript, reduces the protein expression of CYP2C8, and increases the exposure to CYP2C8 substrates. In addition, genome‐wide significant associations were observed between SLCO1B1 and SLCO1A2 variants and repaglinide pharmacokinetics without concomitant medications, and between SLCO1C1 variants and repaglinide pharmacokinetics after gemfibrozil pretreatment.

HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

This study describes a novel decreased function CYP2C8 allele and elucidates the roles of SLCO1B1 and CYP2C8 in interindividual variability of the pharmacokinetics of repaglinide and the gemfibrozil–repaglinide interaction. Pharmacogenetic variation should be considered in the interpretation of drug–drug interaction studies.

Repaglinide and gemfibrozil are recommended as an index substrate and inhibitor, respectively, for investigating cytochrome P450 (CYP) 2C8‐mediated drug–drug interactions in humans. ¹ Repaglinide undergoes extensive metabolism primarily by CYP2C8 to inactive metabolites, with a minor contribution by CYP3A4. ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ Repaglinide is also a substrate for organic anion‐transporting polypeptide (OATP) 1B1, encoded by SLCO1B1, an influx transporter which transports repaglinide from blood into the hepatocytes. ⁸ , ⁹ Gemfibrozil is metabolized by the UDP‐glucuronosyltransferase 2B7 enzyme into gemfibrozil 1‐O‐β‐glucuronide, which is a strong metabolism‐dependent inhibitor of CYP2C8. ³ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ Moreover, gemfibrozil is a weak OATP1B1 inhibitor. ³ , ²¹ Concomitant gemfibrozil use causes a 7–8‐fold increase in the AUC_0‐∞ of repaglinide. ⁴ , ¹⁰ , ¹¹ , ¹² , ²² , ²³ , ²⁴

Genetic variation influences the pharmacokinetics of repaglinide. It is well established that the SLCO1B1 c.521 T>C (p.Val174Ala, rs4149056) no function variant of OATP1B1 increases exposure to repaglinide. ⁸ , ²⁵ , ²⁶ However, the role of genetic variability in CYP2C8 activity is unclear. Some studies have reported that the CYP2C8*3 allele, defined by c.416G>A (p.Arg139Lys, rs11572080) and c.1196A>G (p.Lys399Arg, rs10509681), is associated with reduced plasma concentrations of repaglinide. ³ , ⁸ , ²⁷ In contrast, the impact of genetic variation on gemfibrozil pharmacokinetics has not been investigated.

Genetic variation may confound the assessment of drug–drug interactions. The aim of this study was to identify genetic variants affecting the pharmacokinetics of the CYP2C8 index drugs repaglinide and gemfibrozil. A genome‐wide association study (GWAS) was used as a hypothesis‐free method to identify variants affecting the pharmacokinetics of repaglinide and gemfibrozil. Because pharmacogenetic variation of CYP2C8 is not well established, genetic variation of CYP2C8 was further investigated with a candidate gene analysis. Data from a previous pharmacogenetic study with another CYP2C8 substrate, montelukast, and high throughput sequencing of the CYP2C locus was used to identify potential novel functional CYP2C8 variants.

METHODS

Participants

A total of 252 unrelated healthy white Finnish individuals were included from previous pharmacokinetic studies, with 172 participants from studies with repaglinide and 158 from studies with gemfibrozil (Table S1 , Table S2 ). Of the participants, 66 were from gemfibrozil–repaglinide interaction studies. The participants' mean ± standard deviation age was 23.1 ± 2.9 years and weight 72.0 ± 11.4 kg. The subjects were ascertained to be healthy by medical history, clinical examination, and laboratory tests before entering the studies. None was a tobacco smoker or used any continuous medication.

Genotyping

During the pharmacokinetic studies, a whole blood sample was collected from each participant into an ethylenediaminetetraacetic acid (EDTA)‐containing tube for DNA extraction. For this study, genomic DNA was extracted using the Maxwell 16 LEV Blood DNA Kit on a Maxwell 16 Research automated nucleic acid extraction system (Promega, Madison, WI) and genome‐wide genotyping was carried out at the Institute for Molecular Medicine Finland (Helsinki, Finland) using Illumina HumanCoreExome‐24v1‐1_A BeadChip (Illumina, San Diego, CA). Hardy–Weinberg equilibrium P > 10⁻⁵ and proportion missing ≤ 0.03 were employed as quality thresholds for including genotype data in statistical analysis. A total of 260,995 single‐nucleotide variations (SNV) with MAF ≥ 0.05 were included in the genome‐wide analyses. First‐ and second‐degree relatives were identified with identity by descent analysis and excluded from further analyses. A principal component analysis was carried out to identify genetic outliers, who were excluded from analyses. To supplement missing data for the candidate gene analysis, the participants were genotyped for selected variants with TaqMan genotyping assays on a QuantStudio 12 K Flex Real‐Time PCR system (Thermo Fisher Scientific, Waltham, MA). GWAS top hit and known or suspected functional variants of SLCO1B1 and CYP2C8 with a MAF ≥ 0.01 were included in the candidate gene analyses (Table S3 ). Haplotypes were computed with PHASE v2.1.1. ²⁸ , ²⁹ For determination of functional SLCO1B1 genotype groups, SLCO1B1 haplotypes were defined based on the c.388A>G (rs2306283), c.463C>A (rs11045819), c.521 T>C (rs4149056), and c.1929A>C (rs34671512) SNVs according to Pharmacogene Variation Consortium. ³⁰ SLCO1B1 genotype function was assigned as described previously. ³¹

Pharmacokinetics

Repaglinide and gemfibrozil plasma concentrations were obtained from the previous pharmacokinetic studies (Table S1 , Table S2 ). Studies on repaglinide were either drug–drug interaction studies where repaglinide was the victim drug or pharmacogenetic studies. The participants had ingested either a 0.25‐mg (n = 151) or a 0.5‐mg (n = 21) dose of repaglinide (NovoNorm, Novo Nordisk, Bagsværd, Denmark) with 150 mL water in the morning between 8.00 and 9.30 a.m. after an overnight fast. Timed blood samples were drawn before and up to 7–12 h after repaglinide ingestion for the determination of repaglinide plasma concentrations. If a participant had participated in multiple studies, studies with the 0.25 mg dose of repaglinide, gemfibrozil–repaglinide interaction studies, and the most recent studies were prioritized in data selection.

Studies on gemfibrozil were drug–drug interaction studies where gemfibrozil was the perpetrator drug (Table S1 , Table S2 ). In most of the studies, gemfibrozil (Lopid 600 mg tablet; Gödecke, Freiburg, Germany) was administered at a dose of 600 mg b.i.d. for 2 days prior to the study day (i.e., five doses in total). In the gemfibrozil–repaglinide interaction studies, gemfibrozil was administered 1 h before repaglinide, except for one study in which it was administered at the same time with repaglinide. ¹¹ In two studies, the gemfibrozil dose was 900 mg ¹¹ , ²⁴ and in two studies gemfibrozil was administered as a single dose. ¹¹ , ¹² Gemfibrozil 1‐O‐β glucuronide concentrations were available for 98 of the 158 participants. Studies with a gemfibrozil dose of 600 mg, gemfibrozil–repaglinide interaction studies, and studies with a total of five doses of gemfibrozil were prioritized in data selection.

The plasma concentrations of repaglinide and gemfibrozil had been quantified by use of an API 3000 or a 5500 Qtrap liquid chromatography–tandem mass spectrometry system (AB Sciex, Toronto, ON, Canada), as described previously (Table S2 ). The lower limit of quantification was 0.01–0.1 ng/mL for repaglinide and 2.5–250 ng/mL for gemfibrozil and gemfibrozil 1‐O‐β‐glucuronide (Table S2 ). The day‐to‐day coefficient of variation (CV) was below 15% at relevant concentrations for all analytes.

Pharmacokinetic variables were recalculated with standard noncompartmental methods (Phoenix® WinNonlin®, version 6.4; Certara, Princeton, NJ). Repaglinide area under the plasma concentration–time curve from time zero to infinity (AUC_0‐∞) was calculated without concomitant medication and after gemfibrozil pretreatment. Repaglinide AUC_0‐∞ values for participants from studies with the 0.5 mg dose of repaglinide were linearly adjusted to the 0.25 mg dose. For gemfibrozil and gemfibrozil 1‐O‐β glucuronide, the between dose AUC (AUC_0‐12h) was calculated for participants from studies with multiple doses of gemfibrozil, and the AUC_0‐∞ was calculated for participants from single‐dose studies. For studies with a 900 mg dose of gemfibrozil, gemfibrozil and gemfibrozil 1‐O‐β glucuronide AUC values were adjusted to the 600 mg dose as described previously. ¹¹ AUC values were calculated using the linear up—logarithmic down method, with extrapolation to infinity by dividing the last predicted concentration by the elimination rate constant.

Statistical analyses for repaglinide and gemfibrozil

The data were analyzed using the statistical programs JMP Pro 17.2 and 18.0 (SAS Institute, Cary, NC) and IBM SPSS Statistics for Windows 29.0 and 30.0 (Armonk, NY). Before statistical analysis, body weight and pharmacokinetic variables were log‐transformed. Covariates were identified using a forward stepwise linear regression analysis. The tested variables were sex and body weight for all pharmacokinetic variables, unadjusted AUC_0‐12h of gemfibrozil and gemfibrozil 1‐O‐β‐glucuronide for repaglinide after gemfibrozil pretreatment, and number of gemfibrozil doses for gemfibrozil and gemfibrozil 1‐O‐β‐glucuronide. P‐value thresholds of 0.05 and 0.10 were employed for entry into and removal from the model.

Associations of genetic variants with the pharmacokinetic variables of repaglinide and gemfibrozil were investigated in a GWAS using stepwise linear regression analysis, with covariates set as fixed factors. In case of missing genotype data, cases were excluded pairwise. A P‐value of < 5 × 10⁻⁸ was considered genome‐wide significant.

A general linear model was used to analyze the associations of SLCO1B1, CYP2C8, and SLCO1C1 genotype groups with the pharmacokinetic variables of repaglinide without concomitant medication and after gemfibrozil pretreatment. Genotype groups with less than three participants were excluded from the analyses. Pharmacokinetic variables were set as dependent variables with appropriate covariates and genotype groups as fixed factors. Pairwise comparisons were performed with the Fisher's least significant difference method, and a P‐value < 0.05 was considered statistically significant.

OATP1A2 transport

The transport of repaglinide by OATP1A2 was investigated in MDCKII cells stably transfected with human OATP1A2, essentially as described previously. ³² Repaglinide concentrations were quantified by use of a 4000 Qtrap liquid chromatography–tandem mass spectrometry system (AB Sciex).

OATP1B1 and OATP1B3 biomarkers

Associations of the SLCO1C1 rs10841611 SNV with the OATP1B1 biomarker glycochenodeoxycholate 3‐O‐glucuronide (GCDCA‐3G) and the OATP1B3 biomarker glycochenodeoxycholic acid 3‐O‐sulfate (GCDCA‐S) were investigated in 356 healthy participants with previously determined biomarker and genotype data. ³³ , ³⁴ Statistical comparisons were carried out using a general linear model, with SLCO1B1 genotype as a fixed factor for GCDCA‐3G. ³¹

Identification of functional CYP2C8 variants

To discover novel functional variants, genetic variation of CYP2C8 was investigated in 191 healthy participants from a previous pharmacogenetic study with the CYP2C8 substrate montelukast and high throughput sequencing of the CYP2C locus. ³⁵ Associations of CYP2C8 variants (MAF > 5%) with montelukast AUC_0‐∞ were investigated with linear regression analysis as described previously. ³⁶ The impact of the CYP2C8 variant that showed the strongest association with montelukast AUC_0‐∞, rs2071426, was further investigated in a stepwise linear regression analysis with other genes affecting the pharmacokinetics of montelukast, as described previously. ³⁶

Human liver samples

The associations of CYP2C8 rs2071426 with CYP2C8 messenger RNA (mRNA) expression, and CYP2C8 protein expression and activity were investigated in a human liver cohort collected from 150 patients. In the same cohort, the associations of SLCO1C1 rs10841611 were investigated with mRNA expression of SLCO1B1, SLCO1B3, and SLCO1C1 and protein expression of OATP1B1 and OATP1B3. CYP2C8 genotypes, mRNA, and in vitro amodiaquine N‐desethylation activity were determined as described previously. ³⁷ CYP2C8 protein was quantified by targeted proteomics using mass spectrometry. ³⁸ OATP1B1 and OATP1B3 proteins were determined with immunoblot analyses. ³⁹ Statistical comparisons were univariate Wilcoxon or Kruskall–Wallis tests with P < 0.05 considered statistically significant. Analyses were performed in R (version 4.5.1).

The associations of CYP2C8 rs2071426 with the expression levels of different CYP2C8 mRNA transcripts and SLCO1C1 rs10841611 with the expression levels of SLCO1B1, SLCO1B3, and SLCO1C1 transcripts were investigated in bulk RNA‐sequencing data from 265 human liver samples obtained from patients undergoing laparoscopic gastric bypass operation at the Kuopio University Hospital, as part of the Kuopio Obesity Surgery Study. ³⁵ Transcript‐level expression was quantified as transcripts per million (TPM) from kallisto pseudoalignment against the GRCh38 RefSeq transcriptome. Using the corresponding NCBI GTF annotation (GCF_000001405.40_GRCh38.p14), we restricted the analysis to protein‐coding RefSeq transcripts annotated with “NM_” identifiers and retained only transcripts with a mean TPM ≥ 10 across all samples. Expression values were transformed as log₂(TPM) prior to analysis. Participants were genotyped using the Illumina OMNI Exome Express array. Analysis of covariance (ANCOVA) was used to assess the association between genotype and transcripts, with the log₂‐transformed expression as the dependent variable; genotype was set as the main factor and RNA integrity number, age, sex, and body mass index were set as covariates. Analyses were performed in R (version 4.5.2).

Ethics statement

Each participant gave a written informed consent before participation. The pharmacokinetic studies were approved by the Ethics Committee of the Hospital District of Helsinki and Uusimaa and the Finnish Medicines Agency Fimea. The use of human liver samples and clinical information was approved by the ethics committees of the University Medical Center Charite Berlin and the University Hospital Tübingen. The Kuopio Obesity Surgery Study was approved by the Ethics Committee of the Northern Savo Hospital District.

RESULTS

Discovery of a novel functional CYP2C8 allele

To discover functional CYP2C8 alleles, targeted sequencing data of the entire CYP2C8 gene and pharmacokinetic data of the CYP2C8 substrate montelukast were reanalyzed from a previous study (n = 191). ³⁶ An intronic CYP2C8 SNV (rs2071426) showed the strongest association with montelukast AUC_0‐∞ (Figure 1 ). The rs2071426 variant allele was found to reside in a common CYP2C8 haplotype with no missense variants, herein named as CYP2C8*19 (Figure S1 ). When accounting for other genetic variants affecting montelukast, CYP2C8*19 associated with a 13% (90% CI: 7–20%; P = 3.7 × 10⁻⁴) increase in the AUC_0‐∞ of montelukast per copy of the variant allele (Table S4 ).

Figure 1.

Discovery of CYP2C8 variants associated with montelukast pharmacokinetics and impact of rs2071426 on CYP2C8 expression and activity in human liver samples. (a) gene‐level association plot of CYP2C8 variants with montelukast AUC_0‐∞ in 191 healthy participants; (b) associations of rs2071426 with CYP2C8 mRNA and protein expression, and in vitro activity measured by amodiaquine metabolism; (c) associations of rs2071426 with transcript‐specific CYP2C8 expression. In (a) dark purple indicates variants strongly linked (r ² > 0.96) with rs2071426. In box and whiskers plots (b, c) box lines indicate median with interquartile range, whiskers extend to smallest or largest values within 1.5 × interquartile range from the hinge, and dots represent individual values.

CYP2C8 rs2071426 is located in the first intron of the main protein‐coding transcript of CYP2C8 (NM_000770.3), and it causes a splice donor site resulting in an alternative transcript with an additional exon (NM_001198855.1) (Figure 2 ). In the first human liver sample cohort, rs2071426 associated with decreased CYP2C8 protein expression and activity measured with amodiaquine N‐desethylation (Figure 1 ). In the second cohort, rs2071426 associated with altered expression of different CYP2C8 transcripts (Figure 1 ). Expression of NM_001198855.1 containing the additional exon was 63% (90% CI: 47–80%, P = 9.7 × 10⁻¹⁴) and 119% (90% CI: 78–170%, P = 8.4 × 10⁻⁹) higher in rs2071426 heterozygotes and homozygotes than in noncarriers. In contrast, the expression of the main protein‐coding transcript (NM_000770.3) was 12% (90% CI: 6–19%, P = 0.0094) and 24% (90% CI: 11–35%, P = 0.011) lower in rs2071426 heterozygotes and homozygotes than in noncarriers. Moreover, rs2071426 associated with decreased expression of NM_001198853.1 and increased expression of NM_001198854.1 (Figure 1 ).

Figure 2.

Location of rs2071426 in CYP2C8 transcripts and protein structure. Data obtained from NCBI Genome Data Viewer (GDV) https://www.ncbi.nlm.nih.gov/gdv/. ⁴⁰

GWAS of repaglinide and gemfibrozil pharmacokinetics

The AUC_0‐∞ of repaglinide without concomitant medication showed large, 16.5‐fold, between‐subject variability, which was reduced to only 3.7‐fold after gemfibrozil pretreatment (Table S5 ). In a GWAS, 21 SNVs in the SLCO1 region associated genome‐wide significantly with the AUC_0‐∞ of repaglinide (Table 1 , Figure 3 ). The strongest association was observed between the SLCO1B1 c.521 T>C (rs4149056, p.Val174Ala) no function variant and an increased AUC_0‐∞ of repaglinide (P = 4.5 × 10⁻¹⁵). After accounting for SLCO1B1 c.521 T>C, six nearly completely linked SNVs in the SLCO1A2 gene associated genome‐wide significantly with a decreased AUC_0‐∞ of repaglinide, with a 3′UTR‐variant (rs11045916) showing the strongest association (P = 3.3 × 10⁻⁹). The SLCO1A2 SNVs were also in a linkage disequilibrium with SLCO1B1 c.463C>A (p.Pro155Thr, rs11045819), which defines the SLCO1B1*14 increased function allele (Figure S2 ). No genome‐wide significant associations were observed after accounting for both SLCO1B1 c.521 T>C and SLCO1A2 rs11045916 (Figure S3 ).

Table 1.

Significant associations from genome‐wide association studies

Trait	Analysis	rsID	Nearest gene	Variant type	Nucleotide change	Amino acid change	Effect (90% CI)	P	MAF
Repaglinide AUC0‐∞	Step 1	rs4149056	SLCO1B1	Missense	c.521 T>C	p.Val174Ala	41% (32%, 50%)	4.5 × 10−15	0.25
Repaglinide AUC0‐∞	Step 1	rs4363657	SLCO1B1	Intron	c.1498‐1331 T>C	‐	33% (25%, 42%)	1.5 × 10−11	0.33
Repaglinide AUC0‐∞	Step 1	rs11045863	SLCO1B1	Intron	c.1497 + 2901C>T	‐	−32% (−38%, −26%)	3.0 × 10−11	0.13
Repaglinide AUC0‐∞	Step 1	rs11045916	SLCO1A2	3′UTR	c.*3042 T>A	‐	−37% (−44%, −30%)	4.3 × 10−11	0.10
Repaglinide AUC0‐∞	Step 1	rs11045891	SLCO1B1	3′UTR	c.*449A>C	‐	−33% (−40%, −27%)	9.0 × 10−11	0.13
Repaglinide AUC0‐∞	Step 1	rs71446763	SLCO1A2	3′UTR	c.*4084C>T	‐	−37% (−43%, −29%)	9.2 × 10−11	0.10
Repaglinide AUC0‐∞	Step 1	rs11045917	SLCO1A2	3′UTR	c.*2534C>T	‐	−37% (−43%, −29%)	9.2 × 10−11	0.10
Repaglinide AUC0‐∞	Step 1	rs11045918	SLCO1A2	3′UTR	c.*1770G>T	‐	−37% (−43%, −29%)	9.2 × 10−11	0.10
Repaglinide AUC0‐∞	Step 1	rs6487215	SLCO1A2	Intron	c.1610 + 107C>T	‐	−37% (−43%, −29%)	9.2 × 10−11	0.10
Repaglinide AUC0‐∞	Step 1	rs10841795	SLCO1A2	Missense	c.38 T>C	Ile13Thr	−37% (−43%, −29%)	9.2 × 10−11	0.10
Repaglinide AUC0‐∞	Step 1	rs11045906	SLCO1A2	Intergenic	A>G	‐	−30% (−36%, −24%)	2.5 × 10−10	0.16
Repaglinide AUC0‐∞	Step 1	rs11045819	SLCO1B1	Missense	c.463C>A	p.Pro155Thr	−32% (−39%, −25%)	7.1 × 10−10	0.11
Repaglinide AUC0‐∞	Step 1	rs10841781	SLCO1A2	3′UTR	c.*3087 T>C	‐	−28% (−34%, −22%)	1.9 × 10−9	0.16
Repaglinide AUC0‐∞	Step 1	rs4149006	SLCO1A2	3′UTR	c.*712C>A	‐	−28% (−34%, −22%)	2.7 × 10−9	0.16
Repaglinide AUC0‐∞	Step 1	rs1304539	SLCO1B3	Intron	c.85‐1847 T>G	‐	30% (21%, 40%)	1.4 × 10−8	0.26
Repaglinide AUC0‐∞	Step 1	rs4149117	SLCO1B3	Missense	c.334G>T	p.Ala112Ser	30% (21%, 40%)	1.5 × 10−8	0.26
Repaglinide AUC0‐∞	Step 1	rs7311358	SLCO1B3	Missense	c.699A>G	p.Ile233Met	30% (21%, 40%)	1.5 × 10−8	0.26
Repaglinide AUC0‐∞	Step 1	rs10841684	SLCO1B3	Intron	c.728‐3330A>T	‐	30% (21%, 40%)	1.5 × 10−8	0.26
Repaglinide AUC0‐∞	Step 1	rs11045797	SLCO1B1	Intron	c.84 + 15122 T>C	‐	−31% (−38%, −23%)	1.7 × 10−8	0.11
Repaglinide AUC0‐∞	Step 1	rs11045681	SLCO1B3‐SLCO1B7	Intron	c.1866‐20164A>G	‐	−30% (−37%, −23%)	3.7 × 10−8	0.12
Repaglinide AUC0‐∞	Step 1	rs11045676	SLCO1B3‐SLCO1B7	Intron	c.1866‐23829 T>A	‐	−30% (−37%, −23%)	4.4 × 10−8	0.12
Repaglinide AUC0‐∞	Step 2	rs11045916	SLCO1A2	3′UTR	c.*3042A>T	‐	−30% (−37%, −23%)	3.3 × 10−9	0.10
Repaglinide AUC0‐∞	Step 2	rs71446763	SLCO1A2	3′UTR	c.*4084C>T	‐	−30% (−36%, −22%)	9.5 × 10−9	0.10
Repaglinide AUC0‐∞	Step 2	rs11045917	SLCO1A2	3′UTR	c.*2534C>T	‐	−30% (−36%, −22%)	9.5 × 10−9	0.10
Repaglinide AUC0‐∞	Step 2	rs11045918	SLCO1A2	3′UTR	c.*1770G>T	‐	−30% (−36%, −22%)	9.5 × 10−9	0.10
Repaglinide AUC0‐∞	Step 2	rs6487215	SLCO1A2	Intron	c.1610 + 107C>T	‐	−30% (−36%, −22%)	9.5 × 10−9	0.10
Repaglinide AUC0‐∞	Step 2	rs10841795	SLCO1A2	Missense	c.38 T>C	Ile13Thr	−30% (−36%, −22%)	1.0 × 10−8	0.10
Repaglinide AUC0‐∞ during gemfibrozil	Step 1	rs972505	SLCO1C1	Intron	c.1799‐374C>A	‐	−21% (−25%, −16%)	1.1 × 10−8	0.48
Repaglinide AUC0‐∞ during gemfibrozil	Step 1	rs10841611	SLCO1C1	Intron	c.1916 + 31 T>C	‐	−21% (−25%, −16%)	1.1 × 10−8	0.48
Repaglinide AUC0‐∞ during gemfibrozil	Step 1	rs6487138	SLCO1C1	Synonymous	c.1927C>T	p.Leu643Leu	−20% (−25%, −16%)	1.6 × 10−8	0.48
Repaglinide AUC0‐∞ during gemfibrozil	Step 1	rs953001	SLCO1C1	3′UTR	c.*165C>T	‐	−20% (−25%, −16%)	1.6 × 10−8	0.48
Repaglinide AUC0‐∞ during gemfibrozil	Step 1	rs953002	SLCO1C1	3′UTR	c.*235C>T	‐	−20% (−25%, −16%)	1.6 × 10−8	0.48
Repaglinide AUC0‐∞ during gemfibrozil	Step 1	rs10444412	SLCO1C1	3′UTR	c.*528 T>C	‐	−20% (−25%, −16%)	1.6 × 10−8	0.48

Repaglinide AUC_0‐∝ without concomitant medications was analyzed in 172 participants. Repaglinide AUC_0‐∞ after gemfibrozil pretreatment was analyzed in 66 participants. Step 1 refers to the first part of the linear regression analysis, which included investigated variant with covariate, while in step 2 also the most significant variant of step 1 was included in the model.

AUC_0‐∞, area under the plasma concentration–time curve from time zero to infinity; CI, confidence interval; MAF, minor‐allele frequency; rsID, reference SNP cluster ID.

Figure 3.

Manhattan plots of (a) repaglinide AUC_0‐∞ in 172 participants, (b) repaglinide AUC_0‐∞ after adjusting for rs4149056, and (c) repaglinide AUC_0‐∞ after gemfibrozil pretreatment in 66 participants. Repaglinide AUC_0‐∞ values (d) by rs4149056 genotype, (e) by rs4149056 and rs11045916 genotypes, and (f) after gemfibrozil pretreatment by rs10841611 genotype. In (a–c) horizontal lines indicate the genome‐wide significance level of 5 × 10⁻⁸. In (d–f) horizontal lines indicate geometric means with 90% confidence intervals, and circles indicate individual AUC_0‐∞ values.

In a GWAS of repaglinide AUC_0‐∞ after gemfibrozil pretreatment, six nearly completely linked SNVs in the SLCO1C1 gene associated with a decreased AUC_0‐∞ of repaglinide (Table 1 , Figure S2 ). The strongest associations were observed with two completely linked (r ² = 1, D′ = 1) SNVs, rs10841611 and rs972505, and a decreased AUC_0‐∞ of repaglinide (P = 1.1 × 10⁻⁸). The six SLCO1C1 SNVs were not linked with the SLCO1 SNVs associated with repaglinide AUC_0‐∞ without concomitant medication or with any SLCO1B1 missense variant. No genome‐wide significant associations were observed after accounting for SLCO1C1 rs10841611. The fold increase in repaglinide AUC_0‐∞ caused by gemfibrozil, gemfibrozil AUC, gemfibrozil 1‐O‐β‐glucuronide AUC, or gemfibrozil 1‐O‐β‐glucuronide/gemfibrozil AUC ratio showed no genome‐wide significant associations (Figure S3 ).

Mechanistic studies of SLCO1A2 and SLCO1C1 associations

Since SLCO1A2 SNVs associated strongly with a decreased AUC_0‐∞ of repaglinide, repaglinide transport by OATP1A2 was investigated in vitro; however, MDCKII cells overexpressing OATP1A2 showed no significant uptake of repaglinide (Figure S4 ). SLCO1C1, which showed genome‐wide significant associations with repaglinide pharmacokinetics, has not been previously reported to influence pharmacokinetics. Because SLCO1C1 is located near SLCO1B1 and SLCO1B3, which both transport repaglinide, associations of SLCO1C1 rs10841611 were investigated in two human liver sample cohorts. In the first cohort, SLCO1C1 rs10841611 showed no significant associations with OATP1B1 or OATP1B3 protein expression nor with mRNA expression of SLCO1B1, SLCO1B3, or SLCO1C1 (Figure S5 ). In the second cohort, the expression of SLCO1B3 main transcript (NM_019844.4) was 18% (90% CI: 6–31%; P = 0.038) and 33% (90% CI: 18–51%; P = 3.6 × 10. ⁴ ) higher in SLCO1C1 rs10841611 heterozygotes and homozygotes, respectively, than in noncarriers (Figure S5 ). Furthermore, we investigated the associations of SLCO1C1 rs10841611 with the OATP1B1 biomarker GCDCA‐3G and the OATP1B3 biomarker GCDCA‐S in healthy participants, but SLCO1C1 rs10841611 showed no significant associations with these compounds (Table S6 ). SLCO1C1 rs10841611 did not associate with the AUC of gemfibrozil or gemfibrozil 1‐O‐β‐glucuronide (Table S7 ).

Effects of SLCO1B1 , CYP2C8 , and SLCO1C1 genotypes on repaglinide pharmacokinetics without concomitant medication

Based on the above findings, we next investigated the effects of SLCO1B1, CYP2C8, and SLCO1C1 genotype groups on repaglinide pharmacokinetics. The AUC_0‐∞ of repaglinide without concomitant medication was 105% (90% CI: 74–143%; P = 8.7 × 10⁻¹¹) and 22% (90% CI: 7–38%; P = 0.011) greater in participants with the poor and decreased function SLCO1B1 genotypes, respectively, than in those with the normal function SLCO1B1 genotype (Table 2 , Figure  S6 ). Conversely, the AUC_0‐∞ was 26% (90% CI: 7–42%; P = 0.034) and 21% (90% CI: 9–31%; P = 0.0055) lower in participants with the highly increased and increased function SLCO1B1 genotypes, respectively, than in those with the normal function SLCO1B1 genotype. In participants with the CYP2C8*19/*19 genotype, the AUC_0‐∞ of repaglinide was 45% (90% CI: 24–70%; P = 1.6 × 10⁻⁴) greater than in participants with the CYP2C8*1/*1 genotype. Furthermore, the AUC_0‐∞ of repaglinide was 26% (90% CI: 13–37%; P = 0.0033) smaller in participants with the CYP2C8*1/*3 genotype and 51% (90% CI: 24–84%; P = 8.2 × 10⁻⁴) greater in those with the CYP2C8*1/*4 genotype, than in participants with the CYP2C8*1/*1 genotype. Compared with SLCO1C1 rs10841611 noncarriers, the AUC_0‐∞ of repaglinide was 18% (90% CI: 7–27%; P = 0.01) and 26% (90% CI: 15–35%; P = 4.7 × 10⁻⁴) lower in SLCO1C1 rs10841611 heterozygotes and homozygotes, respectively.

Table 2.

Associations of SLCO1B1 genotype groups and CYP2C8 and SLCO1C1 genotypes with the AUC_0‐∞ of repaglinide (n = 166) without concomitant medication

Trait	Genotype group (n, %)	Geometric mean (90% CI)	GMR (90% CI)	P
Repaglinide AUC0‐∞ (ng × h/mL)	SLCO1B1
	Highly increased function (5, 3%)	3.2 (2.5, 3.9)	0.74 (0.58, 0.93)	0.034
	Increased function (30, 18%)	3.4 (3.0, 3.8)	0.79 (0.69, 0.91)	0.0055
	Normal function (64, 39%)	4.3 (4.0, 4.6)	1
	Decreased function (50, 30%)	5.2 (4.7, 5.7)	1.22 (1.07, 1.38)	0.011
	Poor function (17, 10%)	8.8 (7.6, 10.2)	2.05 (1.74, 2.43)	8.7 × 10−11
	CYP2C8
	*1/*3 (15, 9%)	3.1 (2.7, 3.6)	0.74 (0.63, 0.87)	0.0033
	*3/*19 (3, 2%)	3.2 (2.4, 4.3)	0.76 (0.57, 1.03)	0.13
	*1/*1 (56, 34%)	4.2 (3.9, 4.6)	1
	*1/*19 (55, 33%)	4.7 (4.3, 5.2)	1.13 (0.99, 1.28)	0.13
	*1/*4 (10, 6%)	6.4 (5.3, 7.6)	1.51 (1.24, 1.84)	8.2 × 10−4
	*4/*19 (5, 3%)	4.5 (3.6, 5.7)	1.07 (0.84, 1.37)	0.65
	*19/*19 (22, 13%)	6.1 (5.4, 7.0)	1.45 (1.24, 1.70)	1.6 × 10−4
	SLCO1C1 rs10841611 T>C
	Noncarrier (42, 25%)	5.5 (5.0, 6.0)	1
	Heterozygote (86, 52%)	4.5 (4.2, 4.9)	0.82 (0.73, 0.93)	0.010
	Homozygote (38, 23%)	4.1 (3.7, 4.5)	0.74 (0.65, 0.85)	4.7 × 10−4

Note: The data are estimated geometric marginal means from a general linear model analysis. SLCO1B1 results are adjusted for CYP2C8 and SLCO1C1, CYP2C8 results are adjusted for SLCO1B1 and SLCO1C1, and SLCO1C1 results are adjusted for SLCO1B1 and CYP2C8. Two participants had SLCO1B1 genotype with undefined function and they were excluded from the analysis. CYP2C8*3/*3 and CYP2C8*4/*4 genotypes were observed in only two participants, and therefore, they were excluded from the analysis.

AUC_0–∞, area under the plasma concentration–time curve from time zero to infinity; CI, confidence interval; GMR, geometric mean ratio to reference group.

Effects of SLCO1B1 , CYP2C8 , and SLCO1C1 genotypes on the gemfibrozil–repaglinide interaction

The AUC_0‐∞ of repaglinide after gemfibrozil pretreatment was 35% (90% CI: 24–45%; P = 5.6 × 10⁻⁵) and 17% (90% CI: 5–27%; P = 0.029) lower in SLCO1C1 rs10841611 homozygotes and heterozygotes, respectively, than in SLCO1C1 rs10841611 noncarriers (Table 3 , Figure S6 ). The AUC_0‐∞ of repaglinide after gemfibrozil pretreatment was 32% (90% CI: 9–61%; P = 0.020) greater in participants with the poor function SLCO1B1 genotype than in participants with the normal function SLCO1B1 genotype. The fold increase in repaglinide AUC_0‐∞ caused by gemfibrozil was 36% (90% CI: 24–47%; P = 1.3 × 10⁻⁴) lower in participants with the CYP2C8*19/*19 genotype than in those with the CYP2C8*1/*1 genotype (Table 3 , Figure 4 ). Compared with participants with the normal function SLCO1B1 genotype, the fold increase in repaglinide AUC_0‐∞ by gemfibrozil was 32% (90% CI: 16–45%; P = 0.0045) lower in participants with the poor function SLCO1B1 genotype.

Table 3.

Associations of SLCO1B1 genotype groups and CYP2C8 and SLCO1C1 genotypes with the pharmacokinetics of repaglinide (n = 58) after gemfibrozil pretreatment

Trait	Genotype group (n, %)	Geometric mean (90% CI)	GMR (90% CI)	P
Repaglinide AUC0‐∞ after gemfibrozil pretreatment (ng × h/mL)	SLCO1B1
	Increased function (4, 7%)	31.7 (26.2, 38.3)	0.95 (0.77, 1.17)	0.67
	Normal function (26, 45%)	33.4 (30.7, 36.3)	1
	Decreased function (23, 40%)	34.5 (31.6, 37.8)	1.03 (0.91, 1.17)	0.64
	Poor function (5, 9%)	44.2 (37.1, 52.6)	1.32 (1.09, 1.61)	0.020
	CYP2C8
	*1/*3 (4, 7%)	42.0 (35.2, 50.1)	1.16 (0.94, 1.42)	0.23
	*1/*1 (17, 29%)	36.2 (32.7, 40.1)	1
	*1/*19 (27, 47%)	34.5 (31.4, 37.8)	0.95 (0.83, 1.09)	0.55
	*19/*19 (10, 17%)	33.0 (29.3, 37.3)	0.91 (0.78, 1.07)	0.34
	SLCO1C1 rs10841611 T>C
	Noncarrier (15, 26%)	42.7 (38.5, 47.3)	1
	Heterozygote (28, 48%)	35.6 (32.6, 38.7)	0.83 (0.73, 0.95)	0.029
	Homozygote (15, 26%)	27.5 (24.5, 31.0)	0.65 (0.55, 0.76)	5.6 × 10−5
Fold increase in repaglinide AUC0‐∞ caused by gemfibrozil	SLCO1B1
	Increased function (4, 7%)	8.4 (6.7, 10.4)	1.12 (0.88, 1.43)	0.42
	Normal function (26, 45%)	7.5 (6.8, 8.2)	1
	Decreased function (23, 40%)	7.1 (6.4, 7.9)	0.95 (0.83, 1.10)	0.58
	Poor function (5, 9%)	5.1 (4.2, 6.1)	0.68 (0.55, 0.84)	0.0045
	CYP2C8
	*1/*3 (4, 7%)	10.1 (8.2, 12.4)	1.26 (1.00, 1.60)	0.10
	*1/*1 (17, 29%)	8.0 (7.1, 8.9)	1
	*1/*19 (27, 47%)	7.4 (6.6, 8.3)	0.93 (0.79, 1.09)	0.45
	*19/*19 (10, 17%)	5.1 (4.4, 5.8)	0.64 (0.53, 0.76)	1.3 × 10−4
	SLCO1C1 rs10841611 T>C
	Noncarrier (15, 26%)	8.0 (7.1, 8.9)	1
	Heterozygote (28, 48%)	6.7 (6.1, 7.4)	0.84 (0.72, 0.98)	0.059
	Homozygote (15, 26%)	6.4 (5.6, 7.4)	0.81 (0.67, 0.97)	0.054

The data are estimated geometric marginal means from a general linear model analysis. SLCO1B1 results are adjusted for CYP2C8 and SLCO1C1, CYP2C8 results are adjusted for SLCO1B1 and SLCO1C1, and SLCO1C1 results are adjusted for SLCO1B1 and CYP2C8. Only two participants had highly increased function SLCO1B1 genotype or CYP2C8*3/*19 genotype, thus they were excluded from the analyses. Each of the CYP2C8*1/*4, CYP2C8*3/*3, CYP2C8*4/*4, or CYP2C8*4/*19 genotypes was observed in only one participant, and therefore, those participants were excluded from the analyses.

AUC_0–∞, area under the plasma concentration–time curve from time zero to infinity; CI, confidence interval; GMR, geometric mean ratio to reference group.

Figure 4.

Effects of SLCO1B1, CYP2C8, and SLCO1C1 genotypes on the gemfibrozil–repaglinide interaction. Individual repaglinide AUC_0‐∞ values during the placebo and gemfibrozil phases are connected with gray lines (n = 58). Colored lines indicate geometric means of repaglinide AUC_0‐∞ in the different genotype groups.

DISCUSSION

This study investigated pharmacogenetic variation of repaglinide and gemfibrozil disposition, as well as their interaction. We identified a novel functional CYP2C8 allele, CYP2C8*19, which was associated with increased AUC_0‐∞ of repaglinide and another CYP2C8 substrate, montelukast. CYP2C8*19 is defined by the rs2071426 variant, which introduces an additional splice site, leading to transcript‐specific alterations in CYP2C8 mRNA expression and reduced CYP2C8 protein expression and activity. In the GWAS of repaglinide without concomitant medication, the SLCO1B1 c.521 T>C no function allele showed a strong association with increased AUC_0‐∞ of repaglinide, whereas SLCO1A2 variants were associated with decreased repaglinide AUC_0‐∞. However, as OATP1A2 did not transport repaglinide in vitro, the observed SLCO1A2 associations are likely attributable to linkage disequilibrium with the increased function SLCO1B1*14 allele. Additionally, six SLCO1C1 variants showed genome‐wide significant associations with repaglinide AUC_0‐∞ after gemfibrozil pretreatment.

The CYP2C8*19 defining variant (rs2071426) is located in intron 1 of CYP2C8 and creates a splice donor site, resulting in the inclusion of an additional exon through alternative splicing. In human liver samples, CYP2C8*19 associated with increased expression of a transcript with the additional exon (NM_001198855.1) and reduced expression of the main protein‐coding transcript (NM_000770.3), resulting in decreased CYP2C8 protein expression and enzyme activity. Consistent with our results, splicing quantitative trait locus data from the Genotype‐Tissue Expression project shows a strong association of rs2071426 with reduced splicing of the first intron of CYP2C8 in the liver (P = 6.8 × 10⁻³³). ⁴¹ The alternatively spliced NM_001198855.1 encodes a protein (NP_001185784.1) that is 70 amino acids shorter than the canonical isoform, resulting in the loss of the N‐terminal α‐helix responsible for anchoring the enzyme to the endoplasmic reticulum. ⁴² The CYP2C8*19 allele had a moderate effect on repaglinide pharmacokinetics, as CYP2C8*19 homozygotes had 45% greater repaglinide AUC_0‐∞ than noncarriers, consistent with a 33% decrease in repaglinide clearance. This is the first study that describes the CYP2C8*19 allele, shows that it influences the pharmacokinetics of CYP2C8 substrates, and provides a mechanistic explanation for the associations.

The strong association of SLCO1B1 c.521 T>C with increased repaglinide AUC_0‐∞ observed in the GWAS is consistent with previous studies ⁸ , ²³ , ²⁵ , ²⁶ and is explained by loss of OATP1B1 function. ⁸ , ⁹ Although SLCO1A2 variants associated with decreased repaglinide AUC_0‐∞, OATP1A2 did not transport repaglinide in vitro. Therefore, a likely explanation is linkage disequilibrium with the increased function SLCO1B1*14 allele, which increases hepatic OATP1B1 expression. ³⁹ , ⁴³ The mechanism behind increased OATP1B1 expression remains unclear, but it may involve altered microRNA binding in the SLCO1B1 3′UTR, as multiple 3′UTR variants are present in the SLCO1B1*14 haplotype (Figure S7 ).

Associations between SLCO1C1 variants and repaglinide AUC_0‐∞ have not been reported previously, nor has SLCO1C1 been implicated in the pharmacokinetics of other drugs. SLCO1C1 is predominantly expressed in the brain and pituitary gland, unlike the liver‐enriched SLCO1B1 and SLCO1B3. ⁴⁴ , ⁴⁵ The SLCO1C1 variants were not in linkage disequilibrium with known functional variants of the nearby SLCO1B1 and showed no associations with OATP1B1 or OATP1B3 biomarker levels. While SLCO1C1 rs10841611 associated with slightly increased SLCO1B3 mRNA expression in one liver cohort, no corresponding effects on protein expression were observed. Moreover, gemfibrozil did not modify the impact of SLCO1C1 variants on repaglinide pharmacokinetics, supporting an OATP1B1‐independent mechanism. Further studies are needed to elucidate the role of SLCO1C1 in pharmacokinetics and the underlying biological mechanisms.

In addition to CYP2C8*19, the CYP2C8*3 and CYP2C8*4 alleles associated with repaglinide pharmacokinetics. Similar to previous studies with repaglinide, montelukast, and cinitapride, the CYP2C8*3 allele associated with decreased exposure to repaglinide, consistent with increased metabolic clearance. ³ , ⁸ , ³⁶ , ⁴⁶ The mechanism behind increased enzyme activity related to CYP2C8*3 is unclear and for some other drugs unchanged or decreased metabolism has also been reported suggesting a substrate specific effect. ³ , ⁴⁷ , ⁴⁸ Opposite to CYP2C8*3, CYP2C8*4 associated with a slightly increased AUC_0‐∞ of repaglinide, consistent with the results of previous studies with other CYP2C8 substrates. ³⁶ , ⁴⁶ Our data suggest that CYP2C8*3 increases and CYP2C8*4 decreases repaglinide metabolism, but it is important to keep in mind that the effects may be substrate specific.

Gemfibrozil increased repaglinide AUC_0‐∞ only fivefold in participants with the CYP2C8*19/*19 genotype, while the average increase was 7.6‐fold. This can be explained by a smaller fraction of repaglinide metabolized by CYP2C8 in participants with the CYP2C8*19/*19 genotype than in those with CYP2C8*1/*1. Similarly, gemfibrozil increased repaglinide AUC_0‐∞ fivefold in participants with the poor function SLCO1B1 genotype. Because these participants lack OATP1B1 activity, this effect largely reflects the contribution of CYP2C8 inhibition to the gemfibrozil–repaglinide interaction. The finding that gemfibrozil increased repaglinide AUC_0‐∞ by approximately one‐third less in participants with the poor function SLCO1B1 genotype than in those with the normal function genotype suggests that gemfibrozil inhibits OATP1B1 activity by ~ 50%.

This study demonstrates that reduced‐function CYP2C8 and SLCO1B1 genotypes may lead to underestimation of CYP2C8‐ or OATP1B1‐mediated drug–drug‐interactions with repaglinide. The ICH M12 guideline recommends excluding individuals with poor function genotypes or evaluating drug–drug interactions separately by genotype when investigating substrates or inhibitors of polymorphic enzymes or transporters. ¹ Although CYP2C8*19/*19 is only a decreased function genotype, the magnitude of the gemfibrozil–repaglinide interaction was considerably smaller in participants with this genotype. In contrast, the interaction appeared greater in CYP2C8*3 carriers, but this association was not statistically significant, likely due to the small number of CYP2C8*3 carriers. The relatively high frequency of the CYP2C8*19/*19 genotype supports genotyping of participants in repaglinide interaction studies to enable stratification of results.

No significant associations were observed in the gemfibrozil GWAS. Although the sample size was relatively large for a pharmacokinetic study, it was likely underpowered to detect rare variants or variants with small effects. The data originated from drug–drug interaction studies with gemfibrozil as the perpetrator, but the victim drugs are not known to affect gemfibrozil pharmacokinetics, and variability in gemfibrozil pharmacokinetics was low between studies. Overall, this study suggests the absence of common genetic variants with large effects on gemfibrozil pharmacokinetics, while effects of rare variants cannot be excluded.

This study included only white Finnish participants. Allele frequencies of pharmacogenetic variants vary across populations. Although the functional impact of these variants on repaglinide pharmacokinetics is expected to be consistent across populations, the proportion of variability attributable to each variant depends on its population‐specific frequency. In the Finnish population, the CYP2C8*19 allele was the most prevalent CYP2C8 variant allele, with a frequency of 28%. The frequency of CYP2C8*19 is similar among European and South Asian populations, but notably lower in East Asians and Africans (Table S8 ). In sub‐Saharan African and African American populations, rs2071426 is strongly linked (r ² = 0.61–0.79, D′ = 0.97–1.00) with the CYP2C8 c.805A>T missense variant (rs11572103, p.Ile269Phe), defining the CYP2C8*2 allele (Table S8 ). ⁴⁹ In one study, CYP2C8*2 associated with decreased metabolism of the CYP2C8 substrate pioglitazone in African Americans. ⁵⁰ Further studies are needed to determine if rs2071426 contributes to the reduced activity of CYP2C8*2.

In conclusion, these data indicate that genetic variation in SLCO1 and CYP2C8 play important roles in the pharmacokinetics of repaglinide and in explaining interindividual variability in the gemfibrozil–repaglinide interaction. Moreover, we characterized a novel CYP2C8 allele, CYP2C8*19, which is relatively frequent and results in an alternative CYP2C8 transcript, reduces the protein expression of CYP2C8, and increases the exposure to CYP2C8 substrates. Our data also demonstrated previously unidentified genome‐wide significant associations of SLCO1C1 variants with repaglinide pharmacokinetics after gemfibrozil pretreatment. The mechanisms of these associations remain unclear but appear to be independent of OATP1B1 and OATP1B3. Overall, the results indicate that pharmacogenetic variation should be considered in the interpretation of drug–drug interactions studies.

FUNDING

This study was supported by grants from the Sigrid Jusélius Foundation (Helsinki, Finland) and State funding for university‐level health research (Helsinki University Hospital, Finland). MS and KK were in part supported by the Robert Bosch Stiftung Stuttgart, Germany. JIK was supported by the Paulo Foundation (Helsinki, Finland) and the Finnish Cultural Foundation (Helsinki, Finland). DK was supported by the Research Council of Finland (contract 316458). The Kuopio Obesity Surgery Study was supported by the Finnish Diabetes Research Foundation (Tampere, Finland), State funding for university‐level health research (Kuopio University Hospital, Finland), the Research Council of Finland (grant no. 138006), the Finnish Cultural Foundation, and the University of Eastern Finland Spearhead Funding.

CONFLICT OF INTEREST

The authors declared no competing interests for this work.

AUTHOR CONTRIBUTIONS

A.J.H.M., P.H., and M.Ni. wrote the manuscript; A.J.H.M. and M.Ni. designed the research; A.J.H.M., P.H., K.K., D.K., M.G., T.T., J.I.K., M.Ne., V.T.M., J.P., M.T., M.S., A.T., J.T.B., and M.Ni. performed the research; A.J.H.M. and M.Ni. analyzed the data.

Supporting information

Data S1.