New Pharmacogenetic Markers to Predict the Risk of Bleeding During Taking of Direct Oral Anticoagulants Pharmacogenetic Markers of Bleeding

Aim. To search for new pharmacogenetic biomarkers of bleeding risk in patients taking rivaroxaban and dabigatran for different indications: atrial fibrillation, endoprosthesis of large joints of lower limbs. Material and methods. The study enrolled 29 patients (17 patients received dabigatran and 12 – rivaroxaban), who had hemorrhagic complications during taking direct oral anticoagulants. To find new pharmacogenetic biomarkers of bleeding risk, a next generation sequencing (NGS) was performed for selected candidate genes. Results. Among the patients with bleeding who received dabigatran, 13 variants of the nucleotide sequence showed statistically significant deviation from the population values: 11 in the CES1 gene and 2 in the ABCB1 gene. Among the patients with bleeding who received rivaroxaban, 7 variants of nucleotide sequence showed significant deviation: 4 in the ABCG2 gene, 2 in the CYP3A4 gene, and 1 in the ABCB1 gene. Conclusion. The identified in this study polymorphisms of candidate genes ABCB1, ABCG2, CES1, CYP3A4 were associated with the risk of bleeding in patients taking rivaroxaban and dabigatran. It makes an important contribution to the pharmacogenetics of direct oral anticoagulants and require additional assessment of clinical significance in further studies.

after knee and hip replacement and for prophylaxis and treatment of deep vein thrombosis (DVT) and pulmonary embolism (PE). In the USA since 2011 to 2014 prescription of rivaroxaban increased from 0.13% to 13.87%, and dabigatran from 1.3% to 7.6% [1]. The proportion of DOACs among oral anticoagulants in 2017-2018 was 56.5% and 31% in the UK and the USA, respectively [2,3]. In 2018 rivaroxaban and apixaban were among the top 10 best-selling medicines in the U.S.: 10th and 2nd place, respectively [4].
The increase in DOACs use which have a number of benefits in comparison with warfarin (faster onset of action, no need for routine control of pharmacodynamic response, predictable pharmacokinetics, fixed dosage regimen etc.) is associated with the most common adverse eventbleeding including some of them requiring emergency medical care [5]. For example, in the UK, each increase in DOACs prescription by 10% in routine general practitioners practice results in 0.9% increase in hemorrhagic events [5]. At the same time, with the begining of DOACs use 4,929 additional cases of emergency hospitalization due to hemorrhagic events during anticoagulant therapy were registered in this country between 2011 and 2016 [5].
Different clinical and demographic factors (age, renal function impairment, race and ethnicity, gender, smoking, drug interactions, diet, etc.) as well as genetic factors (polymorphism of genes encoding cytochrome P-450 isoenzymes and drug transporters, etc.) contribute to the variability of pharmacological response to DOACs. At the same time, despite the rapid growth of DOACs prescription and the increase in hemorrhagic events during therapy, peculiarities of DOACs pharmacogenetics remain insufficiently studied [6]. Today, there is a limited number of pharmacogenetic studies of the relationship of gene polymorphism of the drug transport system and drug metabolism (ADME) with a response to DOACs: single-nucleotide polymorphisms of CES1 and ABCB1 genes affecting pharmacokinetic parameters and clinical outcomes during dabigatran therapy and single-nucleotide polymorphisms of ABCB1 gene affecting pharmacokinetic parameters of rivaroxaban and apixaban have been detected [7]. However, there is a significant lack of data on the search for new genetic biomarkers through genome sequencing (exom, full genome) or individual candidate genes in patients with an unsatisfactory response and hemorrhagic events due to DOACs.
The aim of the present study was to search for new pharmacogenetic biomarkers of bleeding risk in patients with atrial fibrillation or after hip and knee replacement taking rivaroxaban and dabigatran.

Material and methods
Written informed consent to participate in the study has been received from all participants. The study was approved by the local Ethical committee of Russian Medical Academy of Continuous Professional Education.

Patients
The study included 29 patients ( DOACdirect oral anticoagulants, GFRglomerular filtration rate

Justification for selecting candidate genes for next-generation sequencing (NGS)
The initial sample of patients was divided into two groups: a) patients who received dabigatran (n=17) b) patients who received rivaroxaban (n=12). In each group the search for potentially significant variants in target genes was done. According to the literature search on the specialized resource PharmGKB [8] and known pharmacokinetic features of the DOACs from Summary of product characteristics [9] ABCB1 and CES1 for dabigatran, ABCB1, ABCG2 and CYP3A4 for rivaroxaban were selected to look for novel pharmacogenetic markers. ABCB1 gene encodes Pglycoprotein or multiple drug resistance protein, a substrate for which are both rivaroxaban and dabigatran [10]. Rivaroxaban's pharmacokinetics may also be affected by another ABCG2 transporter that encodes the breast cancer resistance protein (BCRP) [10]. Gene CES1 encodes carboxylesterase-1, which participates in metabolism of dabigatran to form dabigatran etaxylate.
Finally, the gene CYP3A4, encodes the isoenzyme of cytochrome system P-450 -CYP3A4, a substrate for which is rivaroxaban [10].
DNA extraction. DNA was extracted from whole peripheral venous blood samples by selective prescription using high concentrations of salts ("salting-out").
Sequencing. Mutations in patients were screened by high-performance semiconductor parallel DNA sequencing using Ion S5 (Thermo Fisher). In the process of sample preparation we used the standard protocol of preparation byAmpliSeq libraries (Thermo Fisher) with the help of Ion AmpliSeq Library Kit 2.0 reagents and user panel of primers, which includes exons and adjacent intron areas of ABCB1, ABCG2, CES1, CYP3A4 genes, as well as their 5' and 3' non-coding areas.
Analysis of sequencing results was carried out using the following software: 2) ANNOVAR (annotation of the functional value of genetic variants, filtration of known polymorphisms using the GnomAD database); 3) IGV (Integrative Genomic Viewer) for expert filtration of sequencing artifacts and sequence alignment results in which mutations are detected by automatic analysis tools.
To determine the potentially significant variants for each found variant, the occurrence in the investigated group was compared with the occurrence in the sample of 15708 complete genomes of healthy people, separated by NGS method, from the GnomAD database [11]. The p-value was calculated using the continuity corrected χ-square (Statistica program), FDR correction was used as a multiple test correction.

Measuring of rivaroxaban and dabigatran plasma concentrations
Determination of rivaroxaban and dabigatran plasma concentration was carried out by high-
At the same time, 4 of them (rs3826193, rs3826192, rs2307240, rs62028647) lie in the functionally significant area and are non-synonymous (i.e. they lead to the change in amino acid sequence) and 4 (rs3826191, rs3826194, rs74019278, rs76828834) are synonymous (i.e. they do not lead to the change in amino acid sequence). Of the remaining variants of the nucleotide sequence found in the CES1 gene, 2 (rs375970897, rs62028646) belong to the intronic region and 1 to the 5'-nontranslated region (rs761128900). Among the found significant substitutions for the ABCB1 gene, 1 (rs9282564) is in the exonic region and is not synonymous; another 1 (rs41297348) substitution is in the underlying cis-regulating region.

Rivaroxaban
In patients received rivaroxaban with bleeding events significant deviations showed 7 variants of nucleotide sequence: 4 in ABCG2 gene, 2 in CYP3A4 gene and 1 in ABCB1 gene (Table 3).

Discussion
Patients with bleeding events on DOACs (dabigatran and rivaroxaban) were included in this study.
Among patients receiving dabigatran (110 and 220 mg two times a day), only 4 (23.5%) of 17 patients had minimum equilibrium plasma concentrations of the drug exceeding the previously described therapeutic limits [12]. Among patients taking rivaroxaban (10-20 mg once a day), 2 (16.6%) ones had a level of the minimum equilibrium plasma concentration of the drug above the previously described therapeutic limits. The data obtained may indicate that the "gold standard"assessment of DOAC concentration is not an ideal prognostic factor of bleeding risk for patients taking DOACs; and therefore studies for additional markers of bleeding for dabigatran and rivaroxaban are required.
As mentioned above, dabigatran is a substrate of P-glycoprotein encoded with the ABCB1 gene. In addition, hepatic carboxylesterase-1 (CES1), encoded by the CES1 gene plays an important role in the pharmacokinetics of dabigatran, under the influence of which dabigatran metabolite M2 is converted into the active form of the drug, dabigatran etexilate. [13]. Several polymorphic variants of the CES1 gene associated with lower dabigatran plasma concentrations and lower risk of bleedings have been identified in earlier studies, including full genome association studies [14]. For example, the carriage of minor alleles by the polymorphic marker rs2244613 of the CES1 gene may be associated with lower dabigatran concentrations in plasma and lower risk of bleeding events on dabigatran [14]. Another single-nucleotide polymorphism of the same rs8192935 gene demonstrated association with minimum and maximum dabigatran concentration, but did not affect the risk of bleedings [14,15]. The association with reduction of activation rate of dabigatran is shown also for polymorphic marker rs71647871 of CES1 gene [13]. Single-nucleotide polymorphisms rs4148738, rs1045642 and rs2032582 of the ABCB1 gene, which have been associated with the dabigatran plasma concentration and the risk of bleeding on dabigatran should be highlighted [14,16,17]. At the same time, the variants identified in this study have not been previously described as associated with the risk of bleeding during therapy with dabigatran. This requires investigating the clinical relevance of the identified variants in larger studies. Moreover, the polymorphic variant rs62028647 has shown a strong non-equilibrium linkage with previously well described [18] variant rs2244613, and these variants can be allocated as haplotype.
The polymorphism of ABCB1, ABCG2 genes encoding the efflux transporters, and the polymorphism of CYP3A4 gene encoding the eponymous isoenzyme of cytochrome P-450 system, may have a significant impact on the pharmacokinetics of rivaroxaban and the bleeding risk on rivaroxaban. Previously published studies have described 4 single-nucleotide polymorphisms of the ABCB1 gene (rs2032582, rs1045642, rs4148738, rs1128503), which may be associated (mainly in haplotype) with higher plasma concentrations of rivaroxaban and, consequently, higher risk of bleeding events [17,19,20]. Some allelic single-nucleotide polymorphisms of ABCG2 and CYP3A4 genes, which affect significantly the pharmacokinetic features of rivaroxaban and the risk of bleeding during this treatment, are not described. The polymorphic markers identified in this study for the ABCB1, ABCG2 and CYP3A4 genes were not previously described as associated with the risk of bleeding during rivaroxaban therapy, which also requires an assessment of their clinical significance in larger studies.

Study limitations
The used NGS approach has some technological features; the variants selected according to the NGS results in subsequent studies will be confirmed by classical sequencing using the Senger method after the accumulation of data in larger samples. It should also be noted that the different size of the compared samples introduces additional statistical flaws and errors. Thus, the variants lying at the distribution edges of the smaller sample have a greater representation for their group and, accordingly, may show false significance. This points to the need to confirm the significance of the identified substitutions by comparison with a comparable control sample.

Conclusion
The polymorphic variants of candidate genes ABCB1, ABCG2, CES1, CYP3A4 identified in this study, associated with the risk of bleeding during rivaroxaban or dabigatran treatment, make an important contribution to the pharmacogenetic study of DOACs and require additional assessment of clinical significance in larger studies.
Disclosures. All authors have not disclosed potential conflicts of interest regarding the content of this paper.
Financial support: The study was funded by Russian Science Foundation, project № 16-15-00227