Impacts of pregnane X receptor and cytochrome P450 oxidoreductase gene polymorphisms on trough concentrations of apixaban in patients with non-valvular atrial fibrillation

We examined the impact of polymorphisms in genes encoding cytochrome P450 (CYP) 3A5 (gene code CYP3A5), P-glycoprotein (ABCB1), breast cancer resistance protein (ABCG2), cytochrome P450 oxidoreductase (POR), and pregnane X receptor (PXR; NR1I2) on the daily dose-adjusted steady-state trough concentrations (C0h/D) of apixaban. The analyses included 104 patients with non-valvular atrial fibrillation (NVAF) undergoing AF catheter ablation. The CYP3A5*3; ABCG2 421C > A; ABCB1 1236C > T, 2677G > A/T, 3435C > T, and 2482-2236G > A; NR1I2 11156A > C, 11193T > C, and 8055C > T; and POR*28 genotypes were determined. The combination of the noted NR1I2 genotypes determined the PXR*1B haplotype. Multiple linear regression analyses demonstrated that decreased creatinine clearance (Ccr) and the PXR*1B/*1B haplotype correlated with increased C0h/D of apixaban, while the presence of the POR*28 allele correlated with decreased C0h/D of apixaban (partial R2 = 0.168, 0.029, and 0.044, all P < 0.05). The mean (95% CI) of estimated marginal means of apixaban C0h/D calculated using Ccr as a covariate was the highest in POR*28 noncarriers with PXR*1B/*1B (23.5 [21.0–25.9] ng/mL/[mg/day]) and lowest in POR*28 carriers with other haplotypes (16.6 [15.5–17.7] ng/mL/[mg/day]). The PXR*1B haplotype and POR*28 genotype statuses, which involve genes that impact the expression of multiple drug-metabolizing enzymes and drug-transporters, may have modest effects on the C0h/D of apixaban, but these effects were found to be small.


Introduction
Apixaban, a direct oral anticoagulant and reversible inhibitor of factor Xa (FXa), is widely used for the prevention of venous thrombosis or thrombus formation in non-valvular atrial fibrillation (NVAF) and in the treatment of venous thromboembolism [1]. In the ARISTOTLE trial, apixaban was shown to more effectively reduce the risk of stroke, systemic thromboembolism, and major bleeding in patients with NVAF, as compared to warfarin [2]. Therefore, apixaban is increasingly prescribed to patients newly diagnosed with NVAF [3].
The dose of apixaban is typically determined based on the patient's age, weight, and serum creatinine level without routine anticoagulant testing or plasma drug concentration monitoring [1]. However, achieving accurate dosing of apixaban is important, as it has been reported that increased plasma concentrations and areas under the plasma concentration-time curves (AUC) are associated with bleeding risk and thromboembolic events [4][5][6]. In addition, large interindividual variabilities in the pharmacokinetics (PK) of apixaban have been observed [7,8]. For example, Gulilat et al. reported that the coefficient of variations of peak concentration (C max ) and trough concentration (C 0h ) at the steady-state were 55.0% and 47.3%, respectively, in NVAF patients receiving 5 mg of apixaban twice daily [9]. These variations may be due to variables impacting the clearance pathways of apixaban. Orally administrated apixaban is absorbed mainly in the small intestine, where the bioavailability is approximately 50%. Apixaban is then metabolized by cytochrome P450 (CYP) in the liver and is excreted through the kidneys. Renal excretion as the unchanged form accounts for 27% of the total clearance, and excretion as metabolites in the urine and feces accounts for 25% of the administered dose. The elimination half-life of apixaban is reported to be approximately 12 h [10].
Apixaban is mainly metabolized by CYP3A4/5, with minor contributions from CYP1A2, CYP2C8/9/19, and CYP2J2 [11]. Furthermore, apixaban is also a substrate of the efflux transporter P-glycoprotein (P-gp; gene code ABCB1) and breast cancer resistance protein (BCRP; gene code ABCG2) [12]. These transporters are involved in absorption from the small intestine, in excretion from hepatocytes into bile, and in renal tubular secretion of substrate drugs [13].
Although several studies have investigated the effects of polymorphisms of genes encoding drug metabolic enzymes and drug transporters on the PK of apixaban, the results have been inconsistent. Ueshima et al. reported that the oral clearance of apixaban was lower in Japanese AF patients with the CYP3A5*3 allele than those with the CYP3A5*1/*1 genotype and was lower in those with ABCG2 421A/A genotype than those with ABCG2 421C allele [14,15]. Dimatteo et al. reported that the C max , not the C 0h , was higher in Caucasian patients taking apixaban with ABCB1 2482-2236A/A genotype than those with the G allele [16]. On the other hand, Roşian et al. reported that there were no significant differences in the trough or peak plasma concentrations of apixaban between patients with the genotype ABCB1 3435C > T and ABCB1 2482-2236G > A [17]. In addition, Lenoir et al. reported that there were no significant differences in the AUC 0-6 h of apixaban among patients with the ABCB1 genotypes 1236C > T, 2677G > A/T, and 3435C > T or among patients with various CYP3A genotypes, including CYP3A5*3 [18].
Several systems are known to be involved in the induction of drug-metabolizing enzymes or transporters, and variations in these genes may help to further explain differences in the PK of apixaban. For example, the nuclear receptor, pregnane X receptor (PXR), regulates the transcription of genes encoding several drug-metabolizing enzymes, such as CYP2 and CYP3A, and drug transporters, such as P-gp, and it thus facilitates the elimination of xenobiotics from the body [19,20]. Accordingly, polymorphisms in PXR (also known as NR1I2) have been found to affect the induction of its target genes [21]. The PXR*1B haplotype cluster, which is common in Asian populations [22], has been characterized by the combination of NR1I2 8055C > T, 11156A > C, and 11193T > C genotypes [22,23] (Supplementary Table 1).
Another relevant system is P450 oxidoreductase (POR), which transfers electrons from NADPH oxidase to CYP enzymes, increasing CYP activity and affecting the metabolism of drug substrates [24,25]. Among several single nucleotide polymorphisms (SNPs) of POR, the most common variant is POR*28 (c.1508C > T, rs1057868), and its allele frequency in the Japanese population is approximately 40% [26]. According to an in vitro study, this SNP is associated with increased activity of multiple CYPs, including CYP1A2, CYP2C19, and CYP3A4/5 [27,28]. Therefore, the statuses of the PXR*1B haplotype and the POR*28 genotype may affect blood levels of drugs that are substrates for the above-mentioned metabolic enzymes and transporters, but the impacts of these polymorphisms on the PK of apixaban have not been investigated.
The first purpose of this study was to investigate the impact of polymorphisms of genes encoding the major metabolic enzymes and transporters, CYP3A5, ABCG2, and ABCB1, that directly affect the PK of apixaban, on daily dose-adjusted steady-state C 0h (C 0h /D) of apixaban in Japanese patients with NVAF. The second goal was to investigate the effects of polymorphisms of PXR and POR, which affect the expression or activity of CYPs and/or ATP-binding cassette (ABC) transporters, on this C 0h /D.

Patients
This single-center study enrolled 104 patients undergoing AF catheter ablation at the Hirosaki University Hospital from June, 2018, through October, 2020. These patients had been taking 2.5 mg or 5 mg apixaban twice daily. Patients who received verapamil or rifampicin during the study period were excluded from the analyses. A pharmacist or nurse confirmed the time of day of each apixaban dose. Apixaban was skipped only in the morning of the day of AF ablation. Blood collection for measurement of C 0h was performed at 10 to 12 h after the last apixaban administration on the morning of the day of AF ablation. The study protocol was approved by the Ethics Committee of Hirosaki University Graduate School of Medicine (project identification code: 2018-011-02), and all patients provided written informed consent before the study. With the exception of amiodarone (AMD), patients who were taking drugs that affect the PK of apixaban were excluded from the analysis due to small sample sizes (verapamil, n = 1, and rifampin, n = 1) [1,29].

Analysis of plasma concentrations of apixaban
Blood samples were centrifuged at 3500 rpm for 10 min at 4 °C, and the separated plasma was stored at −30 °C until analysis. Plasma concentrations of apixaban were measured by ultra-performance liquid chromatography (UPLC) tandem mass spectrometry using an ACQUITY UPLC System (Waters, MA, USA). Plasma (100 µL) was mixed with 150 µL of acetonitrile and 10 µL of an internal standard (IS), which consisted of 1000 ng/mL rivaroxaban. The mixture was vortexed for 30 s and centrifuged at 13,500 rpm for 5 min at room temperature. The supernatant (100 µL) was diluted with 100 µL MilliQ water. The sample was transferred to an autosampler vial, and 5 µL was then injected into an ACQUITY UPLC phenyl column (1.7 µm, 75 mm × 2.1 mm) at 40 °C. The mobile phase consisted of (A) MilliQ water with 0.1% formic acid and (B) acetonitrile with 0.1% formic acid. The flow rate was 0.4 mL/min. Gradient conditions were as follows: 0-1.0 min, held in 5% B; 1.0-6.0 min, linear from 5 to 95% B; 6.0-7.0 min, held in 95% B; 7.0-7.1 min, linear from 95 to 5% B; and 7.1-10.0 min, held in 5% B.
The analyte and IS were ionized and detected using a Xevo TQD mass spectrometer (Waters). Positive electrospray ionization was performed in the multiple reaction monitoring mode. Ion transitions of apixaban and IS were m/z 460.2 → 443.0 and 436.1 → 144.9. Cone voltages and collision energies were 40 V and 20 eV for apixaban and 40 V and 30 eV for IS. The calibration curve was linear in the range of 5 to 400 ng/mL. If the plasma concentration of the analyte exceeded the upper limit of the calibration curve, the plasma sample was diluted twice with an equal volume of blank human plasma and remeasured. Calibration curves showed good linearity, with R 2 > 0.99. The intra-and inter-day accuracy values (CV%) were all within ± 15%, and precision values (CV%) were all less than 15% within the range of each calibration curve.

Statistical procedures
The distributions of continuous variables were evaluated using the Shapiro-Wilk test and histograms. Continuous data from demographic and clinical information were presented as means ± standard deviations (SDs), minima and maxima. Allele frequencies of gene polymorphisms were evaluated according to the Hardy-Weinberg equilibrium using chi-square tests. One-way analysis of variance or Student's t-tests and correlation tests with Pearson's correlation coefficient (r) values were used to determine differences and correlations between groups with normally distributed continuous variables. Kruskal-Wallis tests or Mann-Whitney U tests and correlation tests with Spearman's ρ values were used to determine differences and correlations between groups with non-normally distributed continuous variables. In comparisons among multiple groups, if the P values were less than 0.05, the difference between the two groups was analyzed by Student's t-tests or Mann-Whitney U tests with Bonferroni correction. In multiple linear regression analyses using the forward-backward stepwise selection method, all of the factors considered in the univariate analysis were entered as explanatory variables, and factors with independence were selected. The polymorphisms of PK-related genes considered in this study involve SNPs that have been reported to lead to functional effects on drug metabolic enzymes and transporters; specifically, these SNPs have either been reported to affect the PK of apixaban [11,12] or to affect the activities of the proteins [22,28]. Therefore, these SNPs were added to the explanatory variables that were found to be independent according to the analysis described above. In addition, a multivariate analysis using the all-possible-regression method was also carried out. The regression line that resulted in P values for all explanatory variables of less than 0.05 and that had the highest coefficient of determination (R 2 ) was determined to be the final model. For genetic polymorphisms that were classified into three genotypes, each genotype was converted into a dummy variable (0 or 1), and the dummy variables for heterozygous genotypes were excluded as factors in this analysis. To assess the validity of the results of the multiple linear regression analysis, the normality of the distribution of residuals between measured and predicted values was investigated. The percent variation that could be explained by this multiple regression equation was expressed as R 2 . An interaction between categorical variables (PXR*1B haplotype and POR*28 genotype) identified as independent variables affecting the dependent variable (C 0h /D of apixaban) in the multiple linear regression analysis was evaluated using the analysis of covariance (ANCOVA). Here, continuous variables (creatinine clearance, Ccr), which were also identified as independent variables, were entered into ANCOVA as covariates, and the interaction between PXR*1B haplotype and POR*28 genotype and the differences of C 0h /D of apixaban between genotypes for each polymorphism were evaluated based on estimated marginal means (EMMs). Differences with P values of less than 0.05 were considered statistically significant. Statistical analysis was performed with SPSS 28.0 for Windows (SPSS IBM Japan Inc., Tokyo, Japan).

Patient characteristics
The demographic and clinical characteristics of the patients with NVAF taking apixaban are listed in Table 1. The means ± SDs of age, body weight, and Ccr as estimated by the Cockcroft and Gault equation [30] were 68.6 ± 9.1 years, 64.0 ± 12.1 kg, and 71.5 ± 26.1 mL/min, respectively. There were no patients with serious hepatic dysfunction, as noted by a Child-Pugh score of greater than 2. During the study period, 19 patients were taking AMD.

Relationships of apixaban C 0h /D to patient characteristics
The values of C 0h /D of apixaban that were calculated from single blood draws from each of 104 patients did not follow a normal distribution (P < 0.001). Comparisons and  correlations between apixaban C 0h /D and patient demographic and clinical characteristics are listed in Table 2. In univariate analyses, the C 0h /D of apixaban showed linear positive and negative correlations with age and Ccr, respectively (ρ = 0.357 and −0.396, respectively; both P < 0.001) ( Supplementary Fig. 1a, b). On the other hand, there were no significant differences or correlations in other patient factors. The distributions of apixaban C 0h /D among patients with genotypes that include various polymorphisms in PKrelated genes are shown in Supplementary Figs. 2 through 4. There were no significant differences found among all genotypes (all P > 0.05) ( Table 3). In the forward-backward stepwise selection multiple linear regression analysis, Ccr, PXR*1B/*1B, and POR*28 carrier were found to be independent factors influencing apixaban C 0h /D (partial R 2 = 0.168, 0.029, and 0.044, respectively; all P < 0.05) ( Table 4). A multiple linear regression analysis using the all-possibleregression method was again performed, with the genotypes of all PK-related genes and Ccr entered as explanatory variables. The results of this analysis were similar to those of the previous analysis. Specifically, the residuals between actual and predicted values of apixaban C 0h /D were normally distributed (P = 0.121) (Supplementary Fig. 5a). In addition, no significant correlations were identified between the residuals of apixaban C 0h /D and Ccr (ρ = −0.022, P = 0.826) (Supplementary Fig. 5b) or between the residuals of of apixaban C 0h /D and PXR*1B haplotypes and POR*28 genotypes (P = 0.473 and 0.680, respectively) ( Supplementary Fig. 5c, d). Furthermore, the residuals for each group were evenly distributed around the zero line. According to an ANCOVA analysis, there was no interaction between the effects of the PXR*1B haplotype status and the POR*28 genotype status on the C 0h /D of apixaban (P = 0.488). Figure 1 shows differences in the EMMs of the C 0h /D of apixaban between POR*28 genotypes and PXR*1B haplotypes calculated using Ccr as a covariate. In POR*28 carriers, the mean (95% CI) of the EMMs was higher in subjects with the PXR*1B/*1B haplotype than in subjects with other haplotypes (19.8 [17.9-21.6] vs. 16.6 [15.5-17.7] ng/mL/ [mg/day], P followed by Bonferroni correction = 0.016). On the other hand, in POR*28 noncarriers, the difference between subjects with these PXR*1B haplotypes was not significant (23.5 [21.0-25.9] vs. 20.5 [19.4-21.6] ng/mL/ [mg/day], P followed by Bonferroni correction = 0.104). In PXR*1B/*1B carriers, the differences in the means (95% CI) of EMMs between POR*28 genotype carriers and noncarriers were not significant (19.8 [17.9-21.6] vs. 23.5 [21.0-25.9] ng/mL/[mg/day], P followed by Bonferroni correction = 0.068). Conversely, in other haplotypes, this value was significantly higher in POR*28 noncarriers than in POR*28 carriers (20.5 [19.4-21.6] vs. 16.6 [15.5-17.7] ng/mL/[mg/day], P followed by Bonferroni correction < 0.001).

Effect of plasma concentration of apixaban on prothrombin time
A significant correlation was observed between the C 0h of apixaban and prothrombin time, as measured with Thromborel S (ρ = 0.519, P < 0.001) (Supplementary Fig. 6).

Discussion
The results of our analyses suggest that PXR and POR polymorphisms affect the steady-state C 0h /D of apixaban. A decrease in Ccr and the PXR*1B/*1B genotype correlated with an increased C 0h /D of apixaban, while the presence of the POR*28 allele correlated with a decreased C 0h /D in Japanese patients with NVAF. However, the effects of these gene polymorphisms on the C 0h /D of apixaban were smaller than that of Ccr. Therefore, further followup studies in larger populations are needed to clarify the interactions between the effects of multiple polymorphisms of PK-related genes on renal function and the PK of apixaban. It has been suggested that PXR polymorphisms may affect the expression level of the genes encoding CYP1A2, CYP2C8/9/19, CYP3A4, and P-gp, as well as CYP3A5 and BCRP [21]. In addition, POR polymorphisms have also been suggested to potentially affect the activities of CYP1A2, CYP2C19, and CYP3A4/5 [27,28]. Because it is known that multiple CYPs and ABC transporters are involved in the PK of apixaban [10], it is possible that PXR and POR polymorphisms, which affect the activities of these multiple clearance pathways, may have affected the C 0h /D of apixaban. In previous studies, CYP3A5*3 and ABCG2 421C > A polymorphisms have been reported to affect the apparent clearance of orally administered apixaban [14,15]. In addition, it has been reported that the dose-and weight-adjusted C 0h of tacrolimus, a substrate of CYP3A5, was different between POR*28 genotypes in renal transplant recipients with the CYP3A5*1 allele but not in those with the CYP3A5*3/*3 genotype [31]. On the other hand, in our study, these gene polymorphisms had no effect on the C 0h /D of apixaban. The differences in apixaban C 0h /D between PXR*1B haplotypes and/or POR*28 genotypes became apparent only when the EMMs calculated using Ccr as a covariate were used as an indicator. In other words, the impact of these genetic polymorphisms on the C 0h /D of apixaban (partial R 2 = 0.073) was smaller than that of Ccr (partial R 2 = 0.168). However, to the best of our knowledge, the impact of PXR and POR polymorphisms on the PK of apixaban has not been investigated previously. Further PK and pharmacogenomics studies involving larger numbers of patients with NVAF are needed in order to fully characterize these relationships.
The patient factor that was found to most strongly influence the C 0h /D of apixaban in this study was Ccr. The significance of Ccr to the PK of apixaban is consistent with the Cockcroft-Gault equation, which is used to determine the appropriate dosage of apixaban for patients with NVAF in the USA and Japan and which takes into account age, weight, and serum creatinine [1]. Although the contribution of renal excretion to overall apixaban elimination is limited to approximately 27% [10], renal function remains the most important factor for predicting apixaban C 0h /D. However, it has been reported that in cases of renal dysfunction, the presence of mutant alleles of ABCB1, ABCG2, and CYP3A5 can lead to impairment of multiple apixaban elimination pathways, in turn substantially increasing its exposure [32]. Furthermore, it has been reported that the uremic toxins that accumulate in the plasma of patients with reduced renal function may reduce the activity of drug-metabolizing enzymes and transporters [33]. This mechanism may represent an additional factor leading to decreased clearance of apixaban. These factors may explain why the contribution of Ccr to the C 0h /D of apixaban was much greater than that of the POR*28 and PXR*1B polymorphisms. It thus will be necessary to conduct further investigations into the relationship between polymorphisms of PK-related genes and plasma levels of uremic toxins on the C 0h /D of apixaban in patients with renal deficiencies.
In addition, it is important to note that we did not identify a significant effect of the concomitant use of AMD on the C 0h /D of apixaban. This result was in agreement with a report by Ueshima et al. [14], but Gulilat et al. reported a dose-dependent effect of AMD on plasma concentrations of apixaban [9]. However, Gulilat et al. also noted that these effects were minor in patients taking less than 400 mg of AMD per day. Because the maximum dose of AMD in the present study was 100 mg/day, our finding of no significant effect supports these previous results.
There were several limitations in this study. First, although O-desmethyl apixaban (M2) is the predominant metabolite produced by CYP3A4/5 [11], we did not measure the plasma concentrations of M2. Therefore, the effect of CYP3A5 polymorphisms on apixaban C 0h /D could not be fully evaluated. By examining the effects of PXR*1B and POR*28 polymorphisms on the ratio of M2 to apixaban, it would have been possible to partially assess the effect of CYP3A4/5 activity on the clearance of apixaban. Second, changes in plasma concentration levels of apixaban over time were not monitored in this study, and the effects of PXR and POR polymorphisms on steady-state C max and AUC of apixaban could not be determined. Considering that blood concentration of apixaban and anti-FXa activity are expected to fluctuate within a dosing interval, in contrast to the blood concentration of warfarin and the international normalized ratio [34], the influence of polymorphisms of PK-related genes examined in this study on C max should also be examined. In particular, the ABCG2 421C > A and ABCB1 2482-2236G > A genotypes may affect the absorption of apixaban from the small intestine [14][15][16] and should be further investigated. Despite these limitations, however, this study identified new polymorphisms of PK-related genes that affect the C 0h /D of apixaban and that are easily monitored in patients with NVAF. Therefore, not only should the well-known effects of polymorphisms in CYP and ABC genes be considered but also PXR and POR polymorphisms should also be investigated as factors potentially affecting the PK of apixaban.
In conclusion, the genotype status of POR*28 and the haplotype status of PXR*1B correlate with the C 0h /D of apixaban in NVAF patients. However, the impacts of these SNPs on the C 0h /D of apixaban were found to be smaller than that of Ccr. Therefore, further studies are needed to clarify the usefulness of these genotype analyses in the development of individualized dosages of apixaban in the clinical setting.