Association of ABCB1 and FLT3 Polymorphisms with Toxicities and Survival in Asian Patients Receiving Sunitinib for Renal Cell Carcinoma

Sunitinib is a tyrosine kinase inhibitor used as first-line treatment for metastatic renal cell carcinoma (mRCC). Asian ethnicity has been previously associated with lower clearance and greater toxicities for sunitinib treatment, relative to Caucasian ethnicity. Research focusing on identifying corresponding biomarkers of efficacy and toxicity has been hitherto conducted in Caucasian populations, and few of the reported associations have been externally validated. Our work thus aims to investigate candidate biomarkers in Asian patients receiving sunitinib, comparing the observed genotype effects with those reported in Caucasian populations. Using data from 97 Asian mRCC patients treated with sunitinib, we correlated 7 polymorphisms in FLT3, ABCB1, VEGFR2, ABCG2 and BIM with patient toxicities, response, and survival. We observed a stronger association of FLT3 738T genotype with leucopenia in our Asian dataset than that previously reported in Caucasian mRCC patients (odds ratio [OR]=8.0; P=0.03). We observed significant associations of FLT3 738T (OR=2.7), ABCB1 1236T (OR=0.3), ABCB1 3435T (OR=0.1), ABCB1 2677T (OR=0.4), ABCG2 421A (OR=0.3) alleles and ABCB1 3435, 1236, 2677 TTT haplotype (OR=0.1) on neutropenia. Primary resistance (OR=0.1, P=0.004) and inferior survival (progression-free: hazard ratio [HR]=5.5, P=0.001; overall: HR=5.0, P=0.005) were associated with the ABCB1 3435, 1236, 2677 TTT haplotype. In conclusion, ABCB1 and FLT3 polymorphisms may be helpful in predicting sunitinib toxicities, response and survival benefit in Asian mRCC patients. We have also validated the association between FLT3 738T and sunitinib-induced leucopenia previously reported in Caucasian populations, but have not validated other reported genetic associations.


Introduction
Sunitinib is a tyrosine kinase inhibitor that targets vascular endothelial growth factor receptors (VEGFR1, VEGFR2 and VEGFR3), platelet-derived growth factors (PDGFRα and PDGFβ), Fms-like tyrosine kinase 3 (FLT3) and the RET protein [1][2][3][4]. It is used as a standard treatment of metastatic renal cell carcinoma (mRCC) in the first-line setting. Although sunitinib has demonstrated benefits in comparison with interferon therapy [4], clinical outcomes including best radiological response, survival and toxicities are heterogeneous, with 25% of patients achieving complete or partial response and 57% exhibiting severe adverse effects in the recent COMPARZ trial [5]. Sunitinib-associated toxicities include diarrhea, hand-foot syndrome, mucositis, hypertension, leucopenia, neutropenia and thrombocytopenia, as well as abnormalities in hepatic, renal, pancreatic and left ventricular function [4]. In the landmark phase 3 trial, toxicities led to dose interruption in 38% and dose reduction in 32% of patients [4]. Asian patients have been noted to experience higher toxicities from sunitinib therapy. For instance, the incidences of grade 3 to 4 thrombocytopenia (37.7%), neutropenia (29.5%) and anemia (21.9%) reported in Korean patients [6] were more than double of the incidences reported in Western patients [4,7,8]. This might be related to a previous observation that Asian ethnicity is associated with decreased sunitinib clearance as compared to Caucasians [9].
We identified three candidate polymorphisms, 1236C/T, 3435C/T and 2677G/T of ABCB1 for their demonstrated effect on the functionality of the multi-specificity transporter encoded [22,23]. Sunitinib is a substrate of ABCB1 and another efflux transporter encoded by ABCG2; brain accumulation of sunitinib has been observed to increase significantly in ABCB1 knockout and ABCB1/ABCG2 double knockout mice, despite bioavailability after oral dosing remaining similar to that of wild type mice [24]. Recently, ABCB1 1236C/T and 2677G/T were found to be associated with the clearance of sunitinib in a study involving 114 cancer patients in the Netherlands [25]. Currently, the three ABCB1 polymorphisms have been associated with hand-foot syndrome and survival in sunitinib receivers in exploratory studies [11,13,19] in Europe. Given these findings and known interethnic allele frequency variations (for instance, the ABCB1 1236 T allele was found in 71.9% and 41% of a Chinese [26] and German population [27] respectively), we were interested in investigating the correlation of ABCB1 polymorphisms with sunitinib treatment outcomes in Asian patients.
Recently, a 2,903-base-pair deletion polymorphism in intron 2 of the BIM gene was found to be associated with unfavorable outcomes upon treatment with multiple tyrosine kinase inhibitors (TKIs). For example, inferior imatinib response in chronic myelogenous leukemia and shorter progression-free survival in EGFR-mutated non-small-cell lung cancer treated with gefitinib or erlotinib was observed in an Asian population [28]. The likely underlying mechanism is alternate splicing leading to loss of the pro-apoptotic BCL2-homology domain 3 (BH3) [28]. The involvement of BIM in sunitinib activity has been suggested by several prior animal and in vitro studies-Naik et al. demonstrated that destruction of tumor vasculature by VEGF-blocking antibodies was BIM-dependent [29] and Yang et al. noted that there was upregulation of BIM along with other proapoptotic genes in human medulloblastoma cell lines treated with sunitinib [30]. The investigation of the association of BIM deletion with outcomes in sunitinib-receiving patients is therefore a subject of interest to us.
This study aimed to evaluate genetic polymorphisms to investigate their association with sunitinib toxicities and survival benefits in Asian renal cancer patients. We have selected candidate polymorphisms based on previously reported effects in Caucasian patients in order to compare the genotype effects seen in each ethnicity. We anticipated that the increased prevalence of high-grade toxicities in Asians would yield increased statistical power in determining relevant genetic markers.

Patients and treatment
A total of 97 mRCC patients who received sunitinib between 2006 and 2014 at the National Cancer Centre Singapore (NCCS) were included in this retrospective study. The study was approved by the Institutional Review Board (Singapore Health Services) and written informed consent was obtained from each patient. Sample size estimation is detailed in S5 Table. The majority of patients (79/97) received sunitinib at a starting dose of 37.5 mg daily over 4 consecutive weeks followed by a 2 week break. This attenuation and deviation from the drug labelrecommended dosage of 50mg daily was established as routine at NCCS after severe to lifethreatening toxicities were frequently noted when sunitinib was initiated at 50mg daily. Efficacy outcomes as determined through a national retrospective analysis have been comparable [31]. 12 patients in this study received a starting dose of 50mg daily. 6 patients received a starting dose of 25mg daily due to advanced age or an aversion to the expected toxicities.

Follow-up and data collection
Sunitinib toxicities and best radiological response were evaluated based on CTCAE version 3.0 [32] and RECIST criteria version 1.1 [33]. Laboratory assessments of serum creatinine, total bilirubin, albumin, aspartate transaminase (AST), alanine transaminase (ALT), hemoglobin, leucocytes and platelets and clinical examinations for hand-foot syndrome and diarrhea were conducted at baseline (before starting sunitinib) and at two time points in each cycle: after 4 weeks of daily sunitinib and after 2 weeks of sunitinib-free rest (before starting the next cycle). Patient characteristics including age, gender, self-reported ethnicity, body weight and height and Eastern Cooperative Oncology Group (ECOG) performance status were also collected. Memorial Sloan-Kettering Cancer Center (MSKCC) prognostic score [34] was calculated for each patient with the available data. All collected data was de-identified by a third party before being used in statistical analysis. The follow-up period ended at the end of April, 2014.

Survival endpoint definition
Progression-free survival (PFS) was defined as the time from the date of sunitinib initiation to the date of sunitinib termination when sunitinib was terminated due to radiological or clinical evidence of progressive disease (PD), severe toxicities or death, with termination due to PD and death due to PD as events. Dose reduction did not count as an endpoint for PFS. Overall survival (OS) was defined as the time from the date of sunitinib initiation to the date of death or to the date of the last follow-up for censored cases.
Germline DNA was obtained from the buffy coat or from formalin-fixed tissue of benign kidney obtained from nephrectomy. The labeling on blood tubes and tissue slides were deidentified by a third party before they were used for DNA extraction. Genotyping was done by PCR amplification of the flanking region of each SNP followed by direct sequencing.

Statistical analysis
Genotype associations with toxicity events or best radiological response were first analyzed using univariate logistic regression. Genotypes generating P<0.20 were further analyzed using multivariate logistic regression including patient age, gender, baseline ECOG status and starting dose as covariates. PFS and OS were estimated using the Kaplan-Meier method [37]. Univariate associations of genotypes and patient characteristics with PFS and OS were analyzed using either a two-tailed log rank test [38] or a Cox proportional hazard test depending on the property of the variable. Genotypes generating P<0.20 were further analyzed using a multivariate Cox regression model by including patient characteristics which had univariate P values of less than 0.05 as covariates and PFS or OS as the depending variable. Only patients for whom sunitinib was the first line treatment for mRCC were included in PFS and OS analyses. In all analyses, missing data were kept missing except for baseline ECOG status, which was replaced with the median value. With an exploratory purpose, multiple testing correction was not done.

Patient characteristics and genotype frequencies
The demographic and baseline clinical characteristics of the 97 patients included in this study are listed in Table 1. The polymorphism frequencies of the 6 SNPs and the BIM deletion are listed in Table 2. Hardy-Weinberg equilibrium held for all 6 SNPs and the BIM deletion (P>0.05) [39]. After verifying pairwise linkage disequilibrium for ABCB1 3435C/T, ABCB1 1236C/T and ABCB1 2677G/TA using a Chi-square test (P< 0.05 in each pair) and phasing with PLINK [40], haplotype TTT was found to be the most common haplotype. It was found in 51 patients, among whom 8 were homozygous carriers. A complete list of haplotypes and their frequencies is provided in S2 Table. Correlation of genotypes to toxicities Univariate and multivariate logistic regression analyses for associations between genetic markers and clinical outcomes are listed in Table 3 (non-significant results are provided in S3 Table). It is noteworthy that the FLT3 738 TT genotype was associated with an 8.0-fold increase in the risk of leucopenia (P = 0.03) and a 2.7-fold increase in the risk of neutropenia (P = 0.04). The ABCB1 1236 T allele, ABCB1 3435 T allele, ABCB1 2677 T allele, ABCB1 3435, 1236, 2677 TTT haplotype and the ABCG2 421 A allele were correlated with a 3-fold (P = 0.03), 10-fold (P = 0.01), 3-fold (P = 0.04), 10-fold (P = 0.03) and 3-fold (P = 0.03) decrease in the risk of neutropenia respectively. The ABCB1 1236 T and ABCB1 3435 T alleles were correlated with a 25-fold (P = 0.0005) and 3-fold (P = 0.02) decrease in the risk of diarrhea respectively. No genotypes were correlated with thrombocytopenia, hepatotoxicity or hand-foot syndrome. The VEGFR2 1191C/T genotype and BIM deletion were not associated with the toxicity endpoints.

Correlation of genotypes with best radiological response and patient survival
Primary sunitinib resistance, defined as the condition in which progressive disease is the best radiological response observed, was more common in carriers of the ABCB1 3435 TT genotype (P = 0.02), ABCB1 2677 TT genotype (P = 0.01) and the ABCB1 3435, 1236, 2677 TTT haplotype (P = 0.004) ( Table 4). Median PFS of the 81 patients who received sunitinib as the firstline therapy was 8.1 months and median OS was 19.5 months. As shown in Table 5 (non-significant results are provided in S4 Table), after including starting dose as a covariate based on univariate P<0.05, the ABCB1 3435, 1236, 2677 TTT haplotype was correlated with inferior PFS (P = 0.001) and OS (P = 0.005) (survival curves are provided in Fig 1).

Discussion
We observed that the FLT3 738 TT genotype predisposed to sunitinib-related leucopenia, an association which had previously been previously reported by van Erp et al. in Caucasian patients [11]. The effect size we observed i.e. an 8.0-fold increase in risk was greater than the 2.4-fold increase previously reported [11]. This may be related to interethnic differences in allele frequencies of other potentially leucopenia-predisposing genotypes such as the CYP1A1 2455A/G (the G allele is present in 3% of Caucasians and 26% of Chinese based on NCBI data) noted by van Erp et al. [11] but not included in this study. Houk et al. also correlated Asian ethnicity with a 13% decrease in sunitinib clearance and 15% increase in peak serum sunitinib concentration and area under curve compared to a control group that was composed of >85% Caucasians [9]. The increased drug exposure, for which interethnic differences in polymorphism frequencies could potentially play a role, may have an influence on the effect sizes of genotype-toxicity associations. The ABCB1 2677T allele was associated with reduced neutropenia risk and inferior radiological response and the ABCB1 1236 T allele was associated with reduced risk of neutropenia and diarrhea. A trend was observed for the association of the ABCB1 1236 T allele with inferior PFS (P = 0.09) and OS (P = 0.07), which was previously reported in Caucasians [13]. This association appears to be in accordance with the findings of Diekstra et al., whose study correlated ABCB1 1236 TT and ABCB1 2677 TT to increased clearance of sunitinib and its active metabolite in 114 cancer patients using univariate analyses that did not include demographic covariates [25]. It is also congruent with Beuselinck et al.'s findings that mRCC patients who received sunitinib as first-line therapy and carried ABCB1 1236 TT or ABCB1 2677 TT/TA require fewer dose reductions due to toxicities compared to carriers of other genotypes [21]. One plausible hypothesis is that increased clearance leads to decreased drug exposure, reduced toxicity and inferior response. However, Diekstra et al. noted that the effect size of a single genetic polymorphism on clearance is much smaller than that of inter-individual variability and is thus inadequate to directly guide dosing [25]. Therefore, the discovery of a panel of genetic markers that collectively offers adequate predictive power and the addition of non-genetic (eg. demographic) markers into the model remain to be investigated.
Our observation that the ABCG2 421 AA genotype was associated with reduced risk of neutropenia (which we defined as <2000/μL being equivalent to grade 1 and above as described in CTCAE version 3.0 [32]) appears to be inconsistent with the observation of Kim et al. [20] that grade 3 or grade 4 neutropenia is significantly more common in carriers of this genotype. In comparison with the Korean cohort (n = 65) studied by Kim et al. [20], among whom 61.5% were first-line sunitinib receivers, 83.5% of our mostly Chinese cohort of patients were firstline sunitinib receivers. Furthermore, 81.4% of our patients started treatment with a reduced dose (37.5mg daily) from the standard course (50mg daily). Although further studies are required for clarification, these differences may possibly explain the discordant observations. The limitations of this study include the retrospective nature of our data collection and the attenuated dosing regimens adopted in Singapore to reduce toxicity. Indeed, we observed lower toxicity incidences as compared to that of the recent COMPARZ trial [5]. For example, 13%, 49%, 46% and 21% of our cohort developed leucopenia, thrombocytopenia, neutropenia and diarrhea respectively. However, the survival outcome we observed (median PFS: 8.1 months; median OS: 19.5 months) is similar to that observed previously by van der Veldt et al. [19] (median PFS: 10.0 months; median OS: 16.3 months), whose study of a cohort of 136 mRCC patients employed the standard 50mg daily dose and calculated PFS and OS from the day of sunitinib initiation. Furthermore, we included starting dose in the multivariate analyses for each genotype correlation with toxicities, response and survival to avoid confounding effect produced by uneven dosing in the genotype models.

Conclusion
Based on our findings, ABCB1 and FLT3 polymorphisms may be helpful in predicting sunitinib toxicities, response and survival benefit in Asian mRCC patients. We have validated the predisposition to leucopenia associated with FLT3 polymorphism as has been previously reported in Caucasian populations.
Supporting Information S1 Table. Previously reported SNPs with effect on outcomes of sunitinib treatment. (DOC)