In the last 65 years, warfarin has been the most widely prescribed oral anticoagulant for treating and preventing thromboembolic events (Barnes et al., 2015). However, due to its narrow therapeutic index and high variance of pharmacokinetics and pharmacodynamics, warfarin interpatient responses are largely variable. Doses of warfarin administered out of the optimal therapeutic range commonly lead to serious adverse events, such as hemorrhagic and thromboembolic events in cases of over and under dosage, respectively. In addition to demographic and clinical factors, such as age, weight and use of concurrent medication, genetics is well established to play an important role in influencing warfarin maintenance dosage among patients (Klein et al., 2009; Johnson et al., 2011).

The key genetic variants that affect warfarin therapeutic response are presented in the VKORC1 and CYP2C9 genes, which encode the warfarin target enzyme and the primary warfarin metabolizer, respectively (Lee and Klein, 2013). However, as both these factors can explain 35–50% of the warfarin dose variability (Lee and Klein, 2013), there has been increasing interest in investigating additional clinically important genetic polymorphisms that may further explain interindividual warfarin dose variability among different populations.

Recently, studies have implicated the nonsynonymous variant presented in the CYP4F2 gene (c.1297G>A, p.Val433Met, rs2108622) in warfarin sensitivity (Caldwell et al., 2008; Cooper et al., 2008; Borgiani et al., 2009; Takeuchi et al., 2009; Cen et al., 2010; Wei et al., 2012). The CYP4F2 gene codes for the enzyme CYP4F2 (cytochrome P450 4F2), which is responsible for the primary metabolization of vitamin K in the liver. Warfarin mechanism of action consists of blocking the vitamin K redox cycle in order to diminish the concentration of the vitamin K reduced form, which is a coactivator of several coagulation factors (van Gorp and Schurgers, 2015). Thus, variations in the vitamin K bioavailability are speculated to interfere in the final response of treatment with warfarin.

Another genetic variant that has been investigated for influencing warfarin maintenance dosage is the silent polymorphism ABCB1 c.3435C>T (rs1045642) (Wadelius et al., 2004; De Oliveira Almeida et al., 2011; Kim et al., 2013). The ABCB1 gene (also called MDR1—Multidrug Resistance 1 Gene) codes for the well-known multidrug efflux pump P-glycoprotein (P-gp), responsible for restricting the absorption of many xenobiotics, as well as excreting them and their metabolites through the kidneys and liver. The referred ABCB1 polymorphism has been extensively studied and associated with the variability dosage of many drugs, including warfarin, which is one of the many substrates of P-gp (Lepper et al., 2005; Salama et al., 2006).

In this context, the aim of this study was to assess the association of ABCB1 and CYP4F2 genotypes with warfarin dose requirements in Brazilian patients.

In the present study, we found the warfarin patients who carry ABCB1 c.3435T variant alleles (CT and TT genotypes) need, on average, fewer warfarin stable maintenance doses in comparison to the individuals that are wild-type CC carriers. Similar data were obtained in the study of Wadelius et al. They found that an ABCB1 haplotype containing only the 3435T allele as variant (called D haplotype) was over-represented [p(χ2) = 0.0242] among the low warfarin dose patients (<0.33 mg/kg, 60%) subgroup in comparison to the medium dose (0.33–0.46 mg/kg, 25%) and the high dose (>0.46 mg/kg, 15%) subgroups (p = 0.0094). On average, the patients that carried one 3435T variant allele for the D haplotype (n = 20) needed 24% lower warfarin dose and dose/body weight (BW) than the other patients (n = 181) who did not carry this haplotype (Wadelius et al., 2004).

In addition, Kim et al. found that patients with ABCB1 c.3435TT genotypes needed, descriptively, the lowest dose of warfarin compared with those with the other 2 genotypes (warfarin dose (mg/week): CC: 59.5 (n = 70); CT: 56.9 (n = 90); TT: 55.6 (n = 36). Nevertheless, no statistical differences between the genotypic subgroups were found. However, according to race, it is possible to verify that the African, Hispanic, and Asian Americans (the non-white subgroup in Kim's study), when analyzed together presented a decreasing pattern of mean warfarin doses according to the addition of the 3435T allele (Kim et al., 2013). The authors considered that the study had little statistical power due to a small number of patients in some genotypic subgroups. Likewise, other studies have also failed to find statistically significant association between ABCB1 c.3435C>T and warfarin dose requirements (Takeuchi et al., 2009; Issac et al., 2014).

Contrary to our result, a Brazilian study of De Oliveira Almeida et al. found that the ABCB1 c.3435T variant allele contributes to higher warfarin doses in thrombophilic patients (De Oliveira Almeida et al., 2011; de Oliveira Almeida et al., 2014). This finding was exclusively seen for the subgroup of patients who needed up to 70 mg of weekly warfarin dose. It is valid to point out that in the study of De Oliveira Almeida et al. the frequency of the T allele was very high (66%) in comparison to the reported ABCB1 c.3435C>T polymorphism frequency in other Brazilian studies, including the present one (around 40%) (Scheiner et al., 2010; Santos et al., 2011a). Furthermore, the patients were categorized into two subgroups: those receiving more and those receiving less than 70 mg warfarin weekly dose, which differs from the methodology of analysis of the present study.

In the literature, the studies that aimed to investigate the effect of the ABCB1 c.3435C>T polymorphism on the codified P-glycoprotein (P-gp) are quite controversial (Fromm, 2002; Sakaeda et al., 2003; Lepper et al., 2005). Some studies have reported pronounced efflux of substrate and higher mRNA expression of P-gp in the gastrointestinal tract in 3435CC carriers in comparison to 3435CT or 345TT (Greiner et al., 1999; Hoffmeyer et al., 2000), while other studies have found that these parameters are significantly lower in the wild-type individuals (Moriya et al., 2002; Nakamura et al., 2002; Morita et al., 2003). In other words, some studies have observed that the ABCB1 c.3435C>T polymorphism leads to augmented P-gp function or expression, although other studies have found that this variant causes decreased or loss of P-gp activity (Hoffmeyer et al., 2000; Hitzl et al., 2001; Kim et al., 2001; Sakaeda et al., 2003; Kimchi-Sarfaty et al., 2007). However, these studies were not performed with the same substrates and populations. So, it is likely that the effect of the ABCB1 c.3435C>T polymorphism on the P-gp activity depends on the substrate-specificity and environmental factors (Perloff et al., 2001; Lepper et al., 2005).

Moreover, the biological assumptions about how the ABCB1 c.3435C>T polymorphism affect P-gp are diverse. Despite the fact that this polymorphism is a silent mutation, as it does not lead to change of the codified isoleucine amino acid at position 1145, Kimchi-Sarfaty et al. have hypothesized that this variant may lead to a conformational P-gp change, which may affect the substrate specificity, explaining the observed affected efflux-pump function (Kimchi-Sarfaty et al., 2007). This conformational change could be due to distinct timing of cotranslational folding of the codon containing the variant. Furthermore, other so far unidentified variants that are in linkage disequilibrium with the c.3435C>T polymorphism, presented in control expression regions (e.g., promoter, enhancer) of the ABCB1 gene, may influence P-gp expression (Hoffmeyer et al., 2000; Frittitta et al., 2001). Additionally, as the ABCB1 c.3435C>T is in strong linkage disequilibrium with the non-synonymous c.2677T>G (p.Ser893Ala) and the synonymous c.1236T>C (p.Gly412=) variants, some authors hypothesize that the c.3435C>T itself does not directly influence the P-gp activity, but this common haplotype does (Salama et al., 2006; Saraeva et al., 2007).

We suggest that a possible biological explanation for our finding regarding the impact of ABCB1 c.3435C>T polymorphism in the warfarin dose requirements is that the rs1045642 variant allele leads to an altered P-gp fold, affecting the warfarin site of recognition, and therefore leads to decreased warfarin efflux activity. It would cause augmented warfarin bioavailability and therefore lower warfarin doses would be required for those patients who are CT and TT carriers.

Our study presents the following limitation: we did not assess the genetic ancestry of the patients. Instead, we used the self-declared race/color parameter, categorized according to the Brazilian Census IBGE (Brazilian Institute of Geography and Statistics) criteria. Self-declared race/color does not correlate completely with genetic ancestry in Brazilians (Pena et al., 2011), as the self-perception of color/race in Brazil is primarily accounted for by phenotypic characteristics, as well as socio, cultural and geographic factors into which the subject is inserted (Santos et al., 2009). However, this parameter is useful in the real-life setting.

The association between ABCB1 c.3435C>T and warfarin dose requirements, the main finding of this study, was observed for the overall stable group and non-white self-declared subgroup of patients, but not for those who self-declared their race/color as white. So, the above cited limitation might be, at least partially, contributing to this lack of association. The c.3435T variant allele in the ABCB1 gene is frequently rarer in brown and black than in white people (Santos et al., 2011a; Soares et al., 2012). However, in the study of Kim et al. the P-gp function was found to be similar, on average, in European Americans and African Americans, for example (Kim et al., 2001). Thus, further studies are needed to elucidate the specific impact of the ABCB1 c.3435C>T polymorphism in non-whites.

Regarding the lack of association of the CYP4F2 c.1297G>A polymorphism with warfarin dose requirements in the present study, when this variant was included as a covariate in a multiple-regression analysis along with other relevant clinical (age, gender, BMI, self-declared race, amiodarone use) and genetic factors (VKORC1 and CYP2C9 genotypes), it did not contribute significantly to the model, even when self-declared race/color subgroups were assessed. Nevertheless, we verified a descriptive bias of augmented dose in individuals that carry the polymorphic allele A, but with no statistical significance (corresponding to GA and AA genotypes): GG = 27.7 ± 0.4 (n = 156), GA = 2 9.3 ± 0.5 (n = 126), AA = 28.9 ± 0.9 (n = 27).

Through a genome wide association study (GWAS), Caldwell et al. implicated the CYP4F2 c.1297G>A polymorphism as a minor significant influencer of warfarin dose variability. The authors found that in three independent cohorts of white patients who had a stable warfarin dose (n = 1,009), the CYP4F2 c.1297G>A variant contributed to approximately 1 mg/day more in homozygous AA individuals, when compared to wild-type (Caldwell et al., 2008). Other studies reported similar results (Zhang et al., 2009; Perini et al., 2010; Botton et al., 2011).

In conclusion, our results suggest ABCB1 c.3435C>T polymorphism may marginally influence warfarin dose requirements in Brazilian patients, when associated with other genotypic, demographic and clinical factors.

This study was approved by the ethics committee and all included subjects provided written informed consent (SDC: 4519/17/019, CAPPesq: 0804/10).

All authors have contributed substantially to the conception and design of this paper. LT performed the experiments, collected and analyzed the data, and wrote the paper. LM recruited the patients, collected data and critically revised the manuscript. RS recruited the patients. JK and AP provided the facilities. PS analyzed the data and critically revised the manuscript.