Distribution of the cytochrome P450 *alleles for CYP2C9 and CYP2C19 in a cohort of the Danish Blood Donor Study determined by using the Illumina Infinium Global Screening Array

Cytochromes P450 (CYP450) drug metabolizing enzymes are the key enzymes in catalysing the oxidative biotransformation of 70%–80% of all drugs in clinical use to either inactive metabolites or active substances.^{1, 2} Polymorphism of genes encoding the CYP450 family of enzymes has attracted considerable attention as the major targets for pharmacogenomics (PGx) testing since they are highly polymorphic and thereby determining for drug response and adverse drug reactions (ADRs).^3-5 CYP2C9 and CYP2C19 are both members of the CYP2C superfamily and located on chromosome 10. Both genes are highly polymorphic, and a large number of alleles have been identified in the human population.^{6, 7} These alleles are defined by specific single nucleotide polymorphisms (SNPs), which may affect enzyme function and thereby drug metabolism. CYP2C9 is involved in the metabolism of commonly used drugs, such as warfarin and the non-steroidal anti-inflammatory drugs (NSAIDs), for example, diclofenac and ibuprofen, whereas CYP2C19 is involved in the metabolism of, for example, clopidogrel, citalopram and proton pump inhibitors (PPIs).⁸ The most frequent CYP2C9 and CYP2C19 alleles beside the *1 allele found in the European population are CYP2C9*2 (430C > T, rs1799853) leading to decreased enzymatic function and CYP2C9*3 (1075A > C, rs1057910)⁹ with no function⁷ and for CYP2C19 the alleles are CYP2C19*2 (681G > A, rs4244285) with no function and CYP2C19*17 (−806C > T, rs12248560) having increased enzymatic function, due to enhanced gene expression.⁹ The Danish Blood Donor Study (DBDS) is a national prospective research cohort and biobank initiated in 2010 and became nationwide in 2015. The cohort consists of healthy blood donors for which questionnaire data and blood samples are collected upon inclusion.^{10, 11} The aim of this study was to compare and validate genotype data obtained with the Illumina Infinium Global Screening Array for CYP2C9 and CYP2C19 from 100 180 participants in the DBDS cohort against a random sample of 65 individuals of the same DBDS cohort, where the genotypes were additionally determined by using the ViennaLab CYP2C9 mpx RealFast PCR genotyping assay and 67 individuals were determined by using the Luminex xTAG CYP2C19 Kit v3 genotyping assay.

In this study, CYP2C19*2 (c.681G > A, rs4244285), CYP2C19*17 (c.−806C > T, rs12248560), CYP2C9*2 (c.430C > T, rs1799853) and CYP2C9*3 (1075A > C, rs1057910) were measured in 100 180 individuals (DBDS freeze 20 210 503). The genome-wide genotyping was performed by deCODE (Reykjavik, Iceland) using Illumina's Infinium Global Screening Array v2.0 (Illumina, San Diego, California, USA), henceforth Illumina GSA, a high-density array with 654 027 fixed markers as well as a number of custom markers. The genotyping was based on DNA extraction from blood samples. For quality control, variants that did not satisfy Hardy–Weinberg equilibrium (HWE) at the cutoff p-value > 0.0001 and samples with a low genotype calling rate (<98%) were excluded. Detailed information on genotyping and quality control has been described in Hansen et al., Gudbjartsson et al. and subsequent DBDS genomics publications.^12-15 The CYP2C9 and CYP2C19 variants included in this study were all featured directly on the Illumina GSA. Whole blood was randomly collected from 67 blood donors, from the same cohort, by using EDTA tubes and subsequently stored at −80°C until further processing. DNA was extracted from 100 μL of whole blood using MagNA Pure Compact Nucleic Acid Isolation Kit 1 (Roche, Mannheim, Germany) and MagNA Pure Compact (Roche, Mannheim, Germany) according to the manufacturer's instructions. DNA was stored at −20°C until further analysis. CYP2C19 genotyping was determined using xTAG CYP2C19 Kit v3 (Luminex Corp., Austin, Texas, USA) according to the manufacturer's instructions, and data was analysed using the TDAS CYP2C19 v 1.01 software (Luminex Corp., Austin, Texas, USA). CYP2C9 genotyping was performed using the CYP2C9 mpx RealFast Assay (ViennaLab Diagnostics, Vienna, Austria). Real-time PCR was performed using the CFX96 system (Bio-Rad, Hercules, California, USA), and PCR conditions were 95°C for 3 min, followed by 39 cycles of 95°C for 15 s and 60°C for 60 s. Data collection and genotype determination were performed using CFX Manager 3.1 software (Bio-Rad, Hercules, California, USA). Descriptive statistics was applied to examine the distribution of genotypes for the entirety of the DBDS cohort as well as make comparisons between the xTAG CYP2C19, CYP2C9 Real Fast Assay and genotyping array data. Concordance rates, that is, the percentage of identically determined genotypes across different platforms, were calculated for each combination of genotyping platforms. The comparison of CYP2C19 genotypes was performed for 67 individuals and CYP2C9 genotypes for 65 individuals with genetic data available in the DBDS cohort. Four primer sets for amplification and sequencing of regions containing relevant SNPs for CYP2C19*2 and *17 and CYP2C9*2 and *3 were designed using the Primer3 software, and specificity was investigated using UCSC In-Silico PCR (https://genome.ucsc.edu/cgi-bin/hgPcr). The primer sets are listed in Table S1. PCR amplification was performed using Phusion™ High-Fidelity DNA Polymerase (Thermo Fisher Scientific, Waltham, Massachusetts, USA), 500 nM of each primer and 100 ng genomic DNA. PCR amplification was performed using the CFX96 (Bio-Rad, Hercules, California, USA) using the following conditions: 98°C for 30 s, followed by 35 cycles of 98°C for 10 s, 60°C for 10 s and 72°C for 30 s. PCR products were purified using a 1% agarose gel, and bands, corresponding to the size listed in Table S1, were excised, purified using the GeneJET Gel Extraction Kit (Thermo Fisher Scientific, Waltham, Massachusetts, USA) and sequenced using the same forward primer as in the PCR amplification and sequenced by Eurofins genomics (Cologne, Germany). Sanger sequencing data was analysed using Snapgene 7.1 software (GSL Biotech LLC, Boston, Massachusetts, USA). Sanger sequencing data with a quality level less than 99 and a quality base calling less than 40 were excluded from the study. Sanger sequencing was used to solve any disputes.

Haldane exact tests were used to determine whether genotypes obtained from Sanger sequencing, Luminex xTAG or ViennaLab Realfast PCR were in HWE. A chi-squared test was used for analysing HWE in the entire DBDS cohort (Illumina GSA). For the smaller sample size exact tests, variants with p > 0.05 were deemed to be in HWE, whereas a cutoff of p > 0.0001 was used for the chi-squared tests.

All statistical analyses were performed using R v4.0.0.¹⁶

Genotype data were retrieved from 100 180 individuals in the DBDS cohort, and the prevalence of CYP2C19*2, CYP2C19*17, CYP2C9*2 and CYP2C9*3 was calculated and is shown in Table 1. Genetic variability of CYP2C19 and CYP2C9 is dominated by the CYP2C19*17 and CYP2C9*2 alleles compared to the CYP2C19*2 and CYP2C9*3 alleles, respectively, for both heterozygous and homozygous expression of the CYP2C19 and CYP2C9 genes. All genotyped variants were observed to be in HWE, except for CYP2C9*3, determined by Sanger sequencing (p = 0.047). All p-values from the exact and chi-squared tests for HWE are included in Tables 1 and 2.

Table 2 shows the allele calls for specific CYP2C19 and CYP2C9 alleles on an individual level in a study population of 65 individuals for CYP2C9 and 67 individuals for CYP2C19, when comparing the genetic data obtained by the Sanger sequencing, the Illumina microarray and two commonly used PCR-based platforms, namely, the xTAG CYP2C19 assay and CYP2C9 RealFast PCR. One sample was excluded from the xTAG CYP2C19 dataset due to missing genotype calls. Two samples were excluded from the RealFast PCR CYP2C9 dataset due to missing genotype calls. As can be seen, only the Sanger sequencing identified homozygous star alleles for CYP2C9*3.

When comparing the concordance between the platforms on an individual level, differences between allele callings of each allele investigated are displayed in Table 3. The highest degree of concordance was seen between the Illumina GSA and the Sanger sequencing, ranging from 98.4% to 100% when compared to the Illumina GSA and Luminex xTAG/RealFast PCR or the Sanger sequencing and Luminex xTAG/RealFast PCR.

In this study, we explored existing genotype data for four CYP450 gene variants from the DBDS cohort, obtained using the Illumina GSA. The genotype distributions across the entire DBDS cohort was reported, consistent with previously reported findings for individuals of European ancestry.⁹ For example, the variant frequency for the effect allele for CYP2C9*2 (c.430C > T, rs1799853) was 12.27% in DBDS, while the gnomAD reported frequency is 13.19% for non-Finish Europeans and 11.42% for Finns.¹⁸ The genotypes were also compared to findings from different genotype protocols, Luminex xTAG (CYP2C19) and ViennaLab Realfast PCR (CYP2C9) and Sanger sequencing (CYP2C9 and CYP2C19), as shown in Table 3. Overall, we found the concordance rates for genotypes obtained by either Luminex xTAG or ViennaLab Realfast protocols and the Illumina GSA-based genotypes of DBDS to be low, 67.7%–94.0%, depending on SNP. It is worth noting that most of the disagreements arise from heterozygous variants being called homozygous and vice versa: 34/36 samples (94.4%) for Luminex and Illumina and 22/23 samples (95.7%) for ViennaLab and Illumina. The remaining few disagreements are called wildtypes on one platform but homozygous for the effect allele on the alternative platform. Sanger sequencing was used as a reference, and we assumed these genotypes to be the true genotypes for each sample, with one notable exception, as the Sanger sequencing results for CYP2C9*3 did not meet HWE (Table 2). Our results showed almost complete concordance between the genotypes obtained by Sanger sequencing and the Illumina GSA; only for CYP2C9*3 (1075 C/C) did we see a single sample called differently on the two platforms (Table 2). Interestingly, this specific sample was called homozygous for the effect allele using Sanger sequencing but as a wildtype on both the ViennaLab and Illumina platforms. Due to the low occurrence of the effect allele for CYP2C9*3 in conjunction with the low sample size, this homozygous call was likely the reason why HWE was not observed for the CYP2C9*3 genotyping results obtained by Sanger sequencing.

Our findings suggest that the Illumina GSA obtained the correct genotypes, while the Luminex and ViennaLab platforms got some level of erroneous calls, especially with regards to distinguishing homozygous and heterozygous samples. For Illumina GSA, this is consistent with previous reports showing the platform to have high concordance with existing and known data, with slight variations depending on the exact DNA input, methodology and algorithm for calling the genotypes.^{19, 20} We believe the current study to be important for highlighting the possible use of existing genotype data for pharmacogenetics. However, it should be noted that this study features a very modest sample size and that this is a major constraint. While the evidence is compelling, larger sample sizes may be needed to substantiate all the observations described in this manuscript. In conclusion, the distribution of four CYP450 gene variants, CYP2C19*2 (c.681G > A, rs4244285), CYP2C19*17 (c.−806C > T, rs12248560), CYP2C9*2 (c.430C > T, rs1799853) and CYP2C9*3 (1075A > C, rs1057910), in the DBDS cohort is consistent with known genotype distributions. When comparing genotyping results from Illumina GSA and Luminex xTAG or ViennaLab Realfast PCR using Sanger sequencing as a reference, the Luminex and ViennaLab platforms performed the worst, with several incorrect calls. These inconsistent genotype calls were more likely to stem from erroneous calling of heterozygous versus homozygous genotypes. This means that using the Illumina GSA could be a valuable tool in the DBDS to identify clinically relevant drug–gene and drug–drug–gene interactions based on clinical outcome.

All authors contributed to the study's conception and design according to journal author guidelines and the ICMJE definition of authorship.

The authors declare no competing interests.