A pharmacogenetic pilot study reveals MTHFR, DRD3, and MDR1 polymorphisms as biomarker candidates for slow atorvastatin metabolizers

A randomized pilot study was carried out in 60 healthy Mexican volunteers to determine ATV pharmacokinetic parameters. A single dose of 80 mg ATV was administered. The clinical study complied with Good Clinical Practice standards, the guidelines of the Declarations of Helsinki and Tokyo, and the Mexican regulations on Bioavailability and Bioequivalence Studies (NOM-177-SSA1-1998) [22]. Furthermore, the protocol was approved by the Research and Ethics Committee of the pharmacokinetic study center, Ipharma S.A. (Monterrey, Mexico). The clinical study has been registered at the Australian New Zealand Clinical Trials Registry (registration number: ACTRN12614000851662).

Sixty four healthy male candidates from northeastern Mexico were recruited and a written informed consent was obtained. Inclusion criteria were: non-smoker, 18-to-45-year old, weight ≥ 50 kg, body mass index (BMI) of 20–26 kg/m², availability for completing the study, being healthy. Since ATV is classified as a pregnancy category X drug, only males were considered for the study. Candidates were excluded for: any abnormal lab result, significant personal or family medical history of angioedema o allergies, the existence of concurrent disease, use of prescription or over-the-counter medication or alcohol before enrollment, history of smoking, alcohol or drug abuse, and incompliance or non-willingness to complete the study. Four candidates were excluded because of the consumption of alcohol and/or substances or an altered blood pressure. The health status of the volunteers was confirmed by a medical history, a physical examination, an electrocardiogram (ECG), laboratory tests (blood count, blood chemistry, liver function tests, and urinalysis), and seronegativity for human immunodeficiency virus and hepatitis B and C viruses.

After an overnight fast at the study center (Ipharma, S. A.), each subject was given a single dose of 80 mg of ATV-coated tablets (Pfizer Pharmaceuticals LLC, Caguas Site, Caguas, PR). The volunteers were under direct medical supervision at the study site. Venous blood (4 mL) was collected in K₂EDTA-coated Vacutainers^TM (BD Diagnostics, Franklin Lakes, NJ, US), before ATV administration (time 0), and at the following time points after drug administration: 0.25, 0.5, 0.75, 1, 1.5, 2, 2.5, 3, 3.5, 4, 5, 6, 8, 12, 24, 36 and 48 h. Plasma was separated by centrifugation (15 min at 1600 g at 4 °C) and stored in cryovials at −80 °C until analysis, using a method validated by Ipharma S. A. [3, 23, 24].

Leukocytes were obtained from the buffy coat, and genomic DNA was extracted by the alkaline lysis method [25]. Seven multiplex polymerase chain reactions (PCRs) amplified the desired gene regions, following a validated protocol [26]. Screening for gene polymorphisms was performed using the PHARMAchip® microarray (Progenika, Derio, ES). This pharmacogenetic genotyping device detects, with a 99.9 % specificity and sensitivity, 85 gene polymorphisms in 34 genes involved in drug metabolism and response, including those encoding cytochrome P450 enzymes, phase II metabolism enzymes, receptors, and transporters. Amplified products were fractionated with DNAse according to a validated protocol [26], followed by fluorescent labeling and hybridization of the microarray, in an automated TECAN HS4800PRO platform (Ventana Medical Systems Inc., Tucson, AZ, US). The hybridization pattern was revealed using the Innoscan 710 scanner (Innopsys S.A., Carbonne, FR). Polymorphic variants were determined using PHARMAchip software V.3.2.9 [26]. Two additional polymorphisms not included in the PHARMAchip, rs2231142 (C__15854163_70) and rs4149056 (C__30633906_10), in ABCG2 and SLCO1B1 respectively, were included in the study and analyzed by Real-Time PCR system using validated Genotyping Assays (Applied Biosystems, Foster City, CA, US) according to the manufacturer’s instructions. Typed polymorphisms were only includedin subsequent association studies after having passed three quality control tests: the genotype call rate (>0.90 completeness to obtain 99.8 % accuracy), the Hardy-Weinberg equilibrium (HWE) test (P-value > 0.05), and the minor allele frequency (MAF) criterion (>0.01).

Proteins were eliminated from the plasma samples by adding 4 volumes of acetonitrile to 100-μL samples, vortexing (70 rpm, 4 min.), and precipitating by centrifugation (9600 g, 10 min., 10 °C). Protein-free supernatant (300 μL) was recovered and 5-μL samples were injected into an Agilent 1100 high-performance liquid chromatographer (HPLC; equipped with an autosampler and a binary pump), which was connected to an Agilent 6410 tandem mass spectrometer (MS/MS) with a triple quadrupole detector (Agilent Technologies, Santa Clara, CA, US) to measure ATV calcium levels. A C₁₈ pre-column and a Synergi^TM Fusion-RP column (4 μm, 80 Å, 50 × 2 mm; Phenomenex, Torrance, CA, US) formed the solid phase, whereas the mobile phase consisted of 0.03 % formic acid/70 % acetonitrile in analytical grade water. The column temperature was 40 °C, the flow rate 0.4 mL/min, and the auto-sampler temperature 20 °C. The detection system used an ESI MS/MS precursor ion (+) 559.3 m/z and a product ion (+) 440.3 m/z. Under these conditions, interday linearity was assessed by performing calibration curves from 0.5 to 100 ng/mL (0.5, 2.5, 5, 10, 25, 50 and 100); intraday quality control was evaluated by using eight ATV control samples of 1.7, 7.5, 35, and 75 ng/mL each.

WinNonlin® professional software V.5.3 (Pharsight Corp., Mountain View, CA, US) was used for pharmacokinetic analysis. The maximum plasma concentration (C_max) and the time to reach C_max (T_max) were calculated from the observed concentration-time data in plasma. Pharmacokinetic parameters were estimated with the non-compartmental method after oral administration of a single dose of ATV and were as follows: 1) the area under the plasma concentration-time curve from time 0 to the time of the last measurement (AUC_0-t), calculated using the logarithmic-linear trapezoidal rule, 2) the area under the curve from time 0 to the time extrapolated to infinity (AUC_0-∞), 3) the apparent clearance of the fraction dose absorbed (Cl/F), 4) the elimination rate constant in the terminal phase (Ke), and 5) the half-life in the terminal phase of the drug (T_1/2).

For sample size calculation, it was assumed that the coefficient of variation (CV) was 45 % for the C_max and AUC of ATV. Considering a significance level of 5 %, a minimum power of 80 %, an Ω of 0.25, and a confidence interval of 90 %, a sample size of 58 would suffice. The metabolizer phenotypes classification was made using a multivariate analysis of the combined pharmacokinetics parameters C_max and AUC_0-t. To minimize the effect of scale differences, before calculating the distance matrix, these variables were standardized. Next, the individual values of C_max and AUC_0-t were subjected to hierarchical cluster analysis (HCA) using the Ward linkage method and the interindividual Manhattan distances were computed. The standardization, HCA, and the hierarchical clustering dendogram were made using Minitab 16 demo software (Minitab Inc., State College, PA, US) [27]. We identified the participants of each cluster and calculated the geometric means of all pharmacokinetic parameters of each cluster. According to the geometric means of the pharmacokinetic parameters of the clusters they were classified into metabolizer phenotypes. Next, one-way ANOVA and the Kruskal-Wallis H test were used to validate the classification model. The HWE was determined by comparing the genotype frequencies with the expected values using the maximum likelihood method [28]. To detect significant differences between 2 groups, Student’s t-test or the Mann Whitney U test were used for parametric and non-parametric distributions, respectively. Differences between more than 2 groups were assessed by one-way ANOVA or the Kruskal-Wallis H test for parametric and non-parametric distributions, respectively. Post hoc tests (LSD and Tamhane’s T2) were used for pairwise comparisons. To evaluate the contribution of genetic factors to the variability of the pharmacokinetic parameters linear regression analysis was done. Possible associations between genotypes or genotype combinations and phenotypes were assesed using contingency tables Χ² statistics and Fisher’s exact tests. The linear regression analysis and association studies were performed under three different models (dominant, recessive, and additive). Odds ratios were estimated with 95 % confidence intervals. The model for prediction was confirmed using stepwise multiple linear regression analysis. Aforementioned analyses were performed with SPSS for Windows, V.20 (IBM Corp., Armonk, NY, US). All P-values were two-tailed. The corrected P (Pc) values were adjusted according to Bonferroni’s correction for multiple comparisons and the Benajmini-Hochberg procedure was applied to exclude spurious associations [29]. A P-value ≤ 0.05 was considered statistically significant.

Sixty male subjects completed the study. Volunteers were of mestizo descent, most of them students (73 %) from the state of Nuevo Leon (83 %). Other demographic characteristics did not display significant variability (Table 1).

Despite controlling physiological and environmental conditions the pharmacokinetic parameters were highly variable (Table 2). The geometric mean ± SD for the pharmacokinetic parameters obtained were: C_max = 41.44 ± 23.35 ng/mL, AUC_0-t = 141.88 ± 86.78 ng/mL*h, AUC_0-∞ = 157.12 ± 87.24 ng/mL*hr, Cl/F = 509.20 ± 265.57 L/h, T_1/2 = 9.81 ± 6.58 h, and the Ke = 0.071 ± 0.035.

C_max and AUC_0-t were used for HCA classification of pharmacokinetic profiles, because C_max tends to best reveal differences in pharmacokinetic profiles and AUC_0-t is considered to be the best parameter to evaluate a drug’s interindividual pharmacokinetic variation [30]. The HCA, based on centroid distance, revealed three main clusters. Which we identified as slow metabolizers (30.00 %), normal metabolizers (41.66 %), and rapid metabolizers (28.33 %), as shown in Fig. 1. The geometric means of the pharmacokinetics parameters were significantly different among the three clusters (P ≤ 0.016), except for T_1/2 (Table 2). The mean concentration-time profile and the geometric mean pharmacokinetic parameters of ATV obtained for each metabolizer phenotype are shown in Fig. 2a and Table 2, respectively. We observed a > 9-fold difference in ATV pharmacokinetic paratmeters between the fastest metabolizer indivual (C_max = 10.94 ng/mL and AUC_0-t = 55.23 ng/mL*h) and the slowest metabolizer individual (C_max = 101.85 ng/mL and AUC_0-t = 454.41 ng/mL*h). The distribution of phenotypes regarding C_max-AUC_0-t values, are presented in Fig. 2b.

The allele and genotype frequencies of the gene polymorphisms with a potential impact on drug metabolism are presented in Additional file 1. The HWE applied to most of the gene polymorphisms, with the exception of the gene deletions of GSTM1 and GSTT1, because the heterozygous variants were not detected [31]. The polymorphism rs1800896 in the IL10 was not in HWE equilibrium. The polymorphisms in CYP2D6, NAT2, TPMT, and TYMS were below the call rate threshold of 0.9. The SNPs in DPYD, rs1799807 in BCHE, and rs28399504 and rs41291556 in CYP2C19 had a MAF < 0.01. The aforementioned polymorphisms were excluded from subsequent analyses; a total of 30 SNPs remained for statistical analysis.

The various pharmacokinetic parameters were affected differentially by the different genetic loci;i.e. a certain polymorphism had an effect on C_max but not on AUC_0-t or the other way round, while there were also polymorphisms that affected both and/or other parameters (Table 3). The effect of MTHFR-rs1801133 on C_max was statistically significant. Heterozygous variant (C/T) carriers and homozygous variant (T/T) carriers had lower C_max values compared to homozygous wild-type (C/C) carriers (P = 0.018 and 0.004, respectively). Carriers of the variant genotype (C/T or T/T) showed significant lower values of C_max (P = 0.006), AUC_0-t (P = 0.050) and AUC_0-∞ (P = 0.044) but statistically significant higher values of Cl/F (P = 0.044) as compared to homozygous wild-type subjects (C/C). The genotypes resulting from the DRD3-rs6280 (Ser9Gly) polymorphism had a significant impact on ATV pharmacokinetics. First, the homozygous wild-type (C/C) carriers had lower T_1/2 values (P = 0.003) and higher Ke values (P = 0.008) as compared to homozygous variant (T/T) carriers. Second, when comparing T/T with (C/T) genotypes, there were significant differences for AUC_0-t, AUC_0-∞, Cl/F, T_1/2 and Ke values (P were 0.027, 0.024, 0.024, 0.027, and 0.041, respectively). Third, the presence of the wild-type allele, combination (C/C + C/T), had a significant influence on all pharmacokinetics parameters (P values: C_max = 0.050, AUC_0-t = 0.026, AUC_0-∞ = 0.016, Cl/F = 0.016, T_1/2 = 0.004, and Ke = 0.007). GSTM3- rs1799735 had a significant effect on AUC_0-∞ and Cl/F when *A/*A and *A/*B were compared (P = 0.041 for both parameters). Regarding TNF-rs1800629, the A allele carriers had significantly higher AUC_0-t, AUC_0-∞, and T_1/2 values (P = 0.035, 0.030, and 0.025, respectively) and lower Cl/F and Ke values (P = 0.030 and 0.025). The homozygous variant (A/A) carrier was not found. MDR1 (ABCB1) and OATP1B1 (SLCO1B1) were the only ATV transport-related genes with an effect on ATV pharmacokinetics. The variant allele (T) of MDR1-rs1045642 produced a significant increase of C_max (P = 0.037) when the combination of the heterozygous and homozygous variant (C/T + T/T) is compared with the homozygous wild-type (C/C). The C allele of the SLCO1B1-rs4149056 polymorphism significantly affected AUC_0-t, AUC_0-∞, and Cl/F values (P = 0.004 for all three parameters) in homozygous wild-type allele and heterozygous (C/C + C/T) carriers. None of the other 24 polymorphisms tested had a significant impact on ATV pharmacokinetics. The influences of polymorphisms on the ATV pharmacokinetics are shown in Table 3.

No individual genotype correlated with any metabolizer phenotype after Bonferroni’s correction for multiple testing. However, MDR1-rs1045642 behavior was remarkable in this aspect, as no homozygous wild-type (C/C) was a slow metabolizer.

Of the six polymorphisms with an effect on ATV pharmacokinetics, three polymorphisms associated with the slow metabolizer phenotype considering genetic models. The C/T or T/T genotype of MTHFR-rs1801133, the T/T genotype of DRD3-rs6280, and the C/T or T/T genotype of MDR1-rs1045642 were significantly associated with slow metabolizer phenotype using dominant, recessive, and dominant models, respectively. This association remained statistically significant after adjusting for multiple testing using Bonferroni’s correction (P < 0.05; Table 4).

Linear regression analysis using aforementioned genetic models confirmed that these five polymorphisms affected the variability of pharmacokinetic parameters of ATV, except for the TNF-rs1800629 polymorphism (Additional file 2).

Next, we analyzed genotype combinations of the six polymorphisms that individually had a significant effect on ATV pharmacokinetic parameters (Table 3 and Additional file 3): cluster A (subjects with genotypes related to normal metabolism), cluster B (subjects with only 1 genotype related to decreased metabolism), and cluster C (subjects with 2 or more genotypes related to decreased metabolism). As the MTHFR, MDR1, and DRD3 genotypes effected pharmacokinetics most, clusters that only considered these genes were formed: Cluster D (subjects with MTHFR, MDR1 and DRD3 genotypes related to normal metabolism), cluster E (subjects in which either MTHFR, MDR1 or DRD3 gentoype related to decreased metabolism), and cluster F (subjects in which all MTHFR, MDR1 and DRD3 genotypes were related to decreased metabolism). The analysis of genotype combinations revealed that cluster C, i.e. subjects with 2 or more genotypes related to decreased metabolism, had a significant higher C_max (P ≤ 0.016), AUC_0-t (P ≤ 0.011) and AUC_0-∞ (P = 0.011), but significantly lower Cl/F values (P = 0.011) when compared with cluster A and B. The genotype combination analysis limited to MTHFR, MDR1 and DRD3 showed that cluster F was significantly different from clusters D and E; a higher C_max (P = 5.3 × 10⁻⁵), AUC_0-t (P = 3.61 × 10⁻⁴), and AUC_0-∞ = (4.22 × 10⁻⁴), but lower Cl/F (P = 1.35 × 10⁻⁴). The influences of the clusters on ATV pharmacokinetics are shown in Table 3.

The association analysis between clusters and phenotypes displayed a mutual dependency and association (P ≤ 0.05). The C and F cluster were associated with slow metabolizers as shown in Table 5. The stepwise multiple regression analysis showed that the combination of MTHFR, DRD3 and MDR1 polymorphisms are related to ATV slow metabolizers. The combination of these three polymorphisms contributed to the pharmacokinetic variability prediction with an R² = 0.295, and adjusted R² = 0.257 with a P = 2.26 × 10⁻⁴.

ATV was well tolerated by all subjects, since no volunteer showed any adverse effects during and at the end of the pharmacokinetic study. No clinically significant changes from baseline were observed in the physical examination or the ECG during the study, and no clinically significant mean changes from baseline were observed for any laboratory parameters.