CYP2A6 and ERCC1 polymorphisms correlate with efficacy of S-1 plus cisplatin in metastatic gastric cancer patients

Background: We evaluated the association between polymorphisms of cytochrome P450 2A6 (CYP2A6)/excision repair cross-complementation group 1 (ERCC1)/X-ray repair cross-complementing group 1(XRCC1) and treatment outcomes of metastatic gastric cancer (MGC) patients treated with S-1/cisplatin. Methods: Among MGC patients (n=108), who received S-1 (40 mg m−2 b.i.d., days 1–14) and cisplatin (60 mg m−2, day 1) every 3 weeks, we analysed the wild-type allele (W) and variants (V) of CYP2A6 (*4, *7, *9, *10), and the polymorphisms of ERCC1 (rs11615, rs3212986) and XRCC1 (rs25487). Results: Patients having fewer CYP2A6 variants had better response rates (W/W vs W/V other than *1/*4 vs V/V or *1/*4=66.7 vs 58.3 vs 32.3% P=0.008), time to progression (TTP) (7.2 vs 6.1 vs 3.5 months, P=0.021), and overall survival (23.2 vs 15.4 vs 12.0 months, P=0.004). ERCC1 19442C>A (rs3212986) was also associated with response rate (C/C, 46.7% vs C/A, 55.3% vs A/A, 87.5%) (P=0.048) and TTP (4.4 vs 7.6 vs 7.9 months) (P=0.012). Patients carrying both risk genotypes of CYP2A6 (V/V or 1/*4) and ERCC1 19442C>A (C/C) vs those carrying none showed an adjusted odds ratio of 0.113 (P=0.004) for response, and adjusted hazard ratios of 3.748 (P=0.0001) for TTP and 2.961 (P=0.006) for death. Conclusion: Polymorphisms of CYP2A6 and ERCC1 19442C>A correlated with the efficacy of S-1/cisplatin.

Gastric cancer is the second most common cause of cancer death worldwide and the most common cancer in Korea (Parkin et al, 2005;Shin et al, 2007). As unresectable or metastatic gastric cancer (MGC) patients have a dismal prognosis with a median survival of less than 1 year despite chemotherapy, more effective treatment is urgently needed.
Until recently, the most common combination chemotherapies for MGC were based on infusional fluorouracil (5-FU) and/or cisplatin, with cisplatin/5-FU and epirubicin/cisplatin/5-FU regarded as reference treatments. Interest in oral fluoropyrimidines has been increasing, however, because of the convenience and safety they offer, and a novel oral fluoropyrimidine, S-1 is being actively investigated. In recent phase III trial, S-1 alone, or combined with cisplatin, showed promising activity as both a palliative and as an adjuvant therapy in advanced gastric cancer (Sakuramoto et al, 2007;Koizumi et al, 2008;Boku et al, 2009;Ajani et al, 2010). Now that several phase III trials have demonstrated that both capecitabine and S-1 are not inferior to 5-FU in the setting of a platinum-containing combination, oral fluoropyrimidine plus platinum combinations are being widely used in clinical practice or as novel reference regimens in clinical trials (Cunningham et al, 2008;Kang et al, 2009;Ajani et al, 2010). In addition, S-1 plus cisplatin shows a better safety profile than 5-FU plus cisplatin (Ajani et al, 2010). On the basis of these efficacy and safety data, S-1 plus cisplatin has become one of the most commonly used regimens in MGC, yet little is known about which subset of patients is most likely to benefit from the therapy. Identification of predictive markers for efficacy and/or toxicity could lead to more tailored therapy and, ultimately, improved treatment outcomes. An oral fluoropyrimidine, S-1, consists of 5-chloro-2,4-dihydroxypyridine, potassium oxonate, and tegafur, which is converted to 5-FU in the liver mainly by cytochrome P450 2A6 (CYP2A6) (Shirasaka et al, 1996;Ikeda et al, 2000). The enzyme CYP2A6 has polymorphic variants; CYP2A6*2, *4, *5, and *20 show no enzyme activity, whereas CYP2A6*6, *7, *9, *10, *11, *12, *17, *18, and *19 show reduced activity (http://www.cypalleles. ki.se). Recent pharmacokinetic studies showed the plasma concentrations and clearances of 5-FU and/or tegafur differed according to the CYP2A6 polymorphisms in patients treated with S-1 (Fujita et al, 2008;Kaida et al, 2008;Kim et al, 2009). Here we hypothesise that CYP2A6 polymorphisms affect the clinical outcomes of patients who are undergoing S-1-containing chemotherapy for MGC -poorer efficacy and/or lower toxicity in patients with defective variant alleles.
Cisplatin is cytotoxic mainly through formation of DNA adducts that cause inter-or intrastrand crosslinking. Nucleotide excision repair (NER) and base excision repair (BER) systems are involved in the repair of such damage. Key and rate-limiting enzymes include excision repair cross-complementation group 1 (ERCC1) in NER and X-ray repair cross-complementing group 1 (XRCC1) in BER, and they could lead to less cisplatin-induced DNA damage and thus to a poor drug response (Bosken et al, 2002;Furuta et al, 2002;Reed, 2005;Olaussen et al, 2006). Polymorphisms of those DNA repair enzymes are associated with altered functional activity and variations in clinical outcome in patients treated with cisplatin-based chemotherapy (de las Penas et al, 2006;Krivak et al, 2008;Bradbury et al, 2009;Kalikaki et al, 2009;Shim et al, 2010). Although there are other important factors in DNA repair pathways, we selected ERCC1 rs11615 and rs3212986 and XRCC1 rs25487 for genotyping, based on their frequencies of minor alleles 40.1 in Asian populations according to the Single-Nucleotide Polymorphism database (http://www.ncbi.nlm.nih.gov/projects/ SNP), and previous findings of associations with treatment outcomes of gastric cancer patients treated with platinum (Liu et al, 2007;Goekkurt et al, 2009;Huang et al, 2009).
In this study, we investigated association between CYP2A6, ERCC1, and XRCC1 polymorphisms and clinical outcomes of MGC patients who received S-1 plus cisplatin chemotherapy.

Study population and treatment
We prospectively collected clinical data on 134 consecutive MGC patients who had received palliative S-1 plus cisplatin as first-line chemotherapy in the Center for Gastric Cancer of the National Cancer Center, Korea, between April 2006 and April 2010. Of those, we excluded 10 patients who did not have measurable lesions, 9 who were lost to follow-up during the first cycle, and 7 in whom blood samples were not available, leaving 108 patients eligible for analysis. Other eligibility criteria included Eastern Cooperative Oncology Group (ECOG) performance status p2, age X18 years, no concurrent uncontrolled medical illness, and adequate haematological (absolute neutrophil count X1500 per ml, platelet count X100 000 per ml), hepatic (aminotransferase p2.5 Â the upper limit of normal (ULN)) (p 5 Â ULN in the presence of liver metastases), total bilirubin p1.5 Â ULN), and renal (creatinine p1.5 Â ULN) function.
Treatment consisted of 40 mg m À2 oral S-1 twice daily (within the hour following morning and evening meals) on days 1 to 14, and a 15-min intravenous infusion of cisplatin 60 mg m À2 on day 1 of a 3-week cycle. Prophylactic administration of granulocytecolony stimulating factor was not allowed. To prevent nausea and vomiting, 5-hydroxytryptamine-3 receptor antagonists and dexamethasone and/or aprepitant were administered before chemotherapy. Treatment was continued in the absence of disease progression or unacceptable toxicity.
All patients provided written informed consent, and the study was approved by the Institutional Review Board at the Research Institute and Hospital, National Cancer Center. All information was obtained with appropriate Institutional Review Board waivers.

Assessment of efficacy and toxicity
Computed tomography scans were performed every 2 -3 cycles or if clinically indicated to evaluate tumour response to treatment, which was assessed according to Response Evaluation Criteria in Solid Tumors (Therasse et al, 2000). Objective responses were confirmed by a second evaluation 4 to 6 weeks later. The response rate and time to progression (TTP) were assessed by investigators. A complete blood cell count with differential, serum chemistry profile, and electrolyte level analysis were performed every 3 weeks. Toxicity was graded according to National Cancer Institute Common Toxicity Criteria (version 3.0).

CYP2A6, ERCC1, and XRCC1 genotyping
Using genomic DNA extracted from 3 ml peripheral blood with a Qiagen DNA extraction kit (Qiagen, Hilden, Germany), we identified the common variant alleles that affect CYP2A6 activity or expression in Asian populations (CYP2A6*4, *7, *9, and *10), as well as the wild-type allele (CYP2A6*1), as previously described (Kong et al, 2009). Briefly, we used polymerase chain reaction (PCR) -restriction fragment length polymorphism, sequencing, and primer extension methods to determine three polymorphic sites ( -48T4G, 6558T4C, and 6600G4T) and deletion of the CYP2A6 gene. We used a GeneAmp PCR system 9700 thermal cycler (Applied Biosystems, Foster City, CA, USA) and performed electrophoresis with an ABI Prism 3100 analyser (Applied Biosystems).

Statistical analysis
We assessed associations between variables using the Pearson w 2test or the Fisher's exact test for categorical variables. We performed multivariate logistic regression analyses to ascertain whether the genetic polymorphisms are independently associated with treatment responses after adjusting for other relevant variables. We defined TTP as the time from the initiation of treatment to the date of documented disease progression, and defined overall survival (OS) as the time from the initiation of treatment to the date of death from any cause or the last follow-up visit. We used the Kaplan -Meier method and the log-rank test to estimate and compare survival distribution, and used Coxregression models for survival multivariate analysis. We used trend tests to assess statistical significance of changes in the relationship between treatment efficacy and genetic polymorphism. We categorised genetic polymorphisms on an ordinal scale, according to genotypes (0 for homozygous non-risk allele, 1 for heterozygous risk allele, and 2 for homozygous risk allele), with risk alleles defined as those associated with poorer treatment efficacy. We considered the results as statistically significant when two-sided P-values were o0.05. Table 1 shows the clinical characteristics of the 108 eligible patients. Their median age was 57 years (range, 26 -72), and the median follow-up period was 19.9 months (range, 1.4 -59.1 months). Most (88.9%) of the patients had an ECOG performance status of 0 -1, and all had metastatic disease, with 24 (22.2%) of them having recurrent metastatic disease after previous curative gastrectomy. Among the characteristics listed in Table 1, ECOG performance status (0 -1 vs 2) was significantly associated with tumour response rate (Pearson w 2 -test, P ¼ 0.008), TTP (log-rank Po0.001), and OS (log-rank Po0.001); sex was significantly associated with tumour response rate (Pearson w 2 -test, CYP2A6/ERCC1 polymorphisms and the S-1/CDDP efficacy SR Park et al P ¼ 0.004) and TTP (log-rank P ¼ 0.041); and number of organs with metastases (o3 vs X3) was significantly associated with TTP (P ¼ 0.029). Table 2 shows the frequencies of the various genotypes. The allelic frequencies were 0.57 for CYP2A6*1, 0.13 for CYP2A6*4, 0.07 for CYP2A6*7, 0.20 for CYP2A6*9, and 0.03 for CYP2A6*10, and were comparable to those previously reported in Asian populations (Kwon et al, 2001;Schoedel et al, 2004;Mwenifumbo et al, 2005;Nakajima et al, 2006). To analyse the effect of the CYP2A6 polymorphisms on treatment outcomes, we classified CYP2A6*4, CYP2A6*7, CYP2A6*9, and CYP2A6*10 as variant alleles. We assigned patients who carried *1/*7, *1/*9, or *1/*10 to a wild-type/ variant (W/V) group and those who carried two variant alleles to a variant/variant (V/V) group. However, because CYP2A6*4 leads to deletion of the entire gene and thus loss of enzyme activity (patients homo-or heterozygous for the *4 allele might have lower enzyme activity or expression than those with other allele variants), we sorted the genotypes into three groups -W/W, W/V other than *1/*4, and V/V or *1/*4.

Genotype frequencies
The frequencies were 0.76 and 0.24 for the C and T alleles, respectively, of rs11615 (Asn118Asn) of ERCC1, 0.75 and 0.25 for the C and A alleles at 19442C4A of ERCC1 (rs3212986), and 0.63 and 0.37 for the G and A alleles of rs25487 (Arg399Gln) of XRCC1. We found no significant association between any polymorphisms and age, sex, ECOG performance status, disease status, histology, tumour location, or number of organs with metastases (data not shown).

Association between genotypes and TTP and OS
Time to progression varied significantly with CYP2A6 genotype (median TTP, 7.2 months for W/W patients vs 6.1 months for W/V other than *1/*4 vs 3.5 months for V/V or *1/*4) (log-rank P ¼ 0.032; trend test P ¼ 0.021) ( Table 3, Figure 1A), and the CYP2A6 genotype was a significant independent risk factor for TTP after adjustment for ECOG performance status, sex, and number of organs with metastases (  Figure 1B), and the CYP2A6 genotype was a significant independent predictor for OS after adjustment for ECOG performance status. The adjusted HRs of patients with W/V other than *1/*4 and those with V/V or *1/*4, relative to patients with W/W, were 1.806 (95% CI, 0.906 -3.598; P ¼ 0.093) and 3.118 (95% CI, 1.483 -6.558; P ¼ 0.003), respectively (Table 4). Among the ERCC1 and XRCC1 polymorphisms, ERCC1 19442C4A was significantly associated with TTP, with risk in the order of C/C4C/A4A/A (median TTP, 4.4 months for C/C patients vs 7.6 months for C/A vs 7.9 months for A/A) (log-rank P ¼ 0.041; trend test P ¼ 0.012) (Table 3, Figure 2A). In multivariate analysis, the ERCC1 19442C4A genotype was an independent predictor for TTP with borderline significance after adjustment for ECOG performance status, sex, and number of organs with metastases (adjusted HR for patients carrying C/C compared with patients carrying A/A genotype as a reference, was 2.365 (95% CI, 0.909 -6.155; P ¼ 0.078) ( Table 4). The adjusted HR of patients with A/A or A/C relative to patients with C/C was 1.546 (95% CI, 0.990 -2.413; P ¼ 0.055).

DISCUSSION
In this retrospective study of MGC patients given S-1 plus cisplatin as first-line chemotherapy, we demonstrated that CYP2A6 and ERCC1 19442C4A polymorphisms correlated with treatment efficacy, and that CYP2A6 polymorphism was an especially strong independent predictor for all efficacy endpoints -response rate, TTP, and OS. Response, TTP, and OS were significantly poorer in patients with two variant alleles or *1/*4 whose enzyme product has reduced or no activity and that could result in a reduced conversion rate of tegafur to 5-FU. Our finding that patients with a V/V or *1/*4 genotype compared with those with a W/W genotype had a probability of response of only 0.22 and a 2.29-fold risk of progression, and a 3.12-fold risk of death are consistent with finding in MGC patients treated with S-1 plus docetaxel (Kong et al, 2009). In addition, our present results are supported by the findings that CYP2A6*4 reduces plasma 5-FU concentration and increases the area of under the concentration-time curve (AUC) and C max for tegafur in non-small cell lung cancer patients treated with S-1 alone or in combination with cisplatin (Kaida et al, 2008). Moreover, in advanced biliary cancer patients, treated with S-1 plus oxaliplatin, the AUC and C max for 5-FU and the metabolic ratio (exposure ratio of 5-FU to tegafur) are significantly higher in patients homozygous for wild-type CYP2A6 than in those with one or two variant alleles (*4, *7, or *9) , again suggesting that CYP2A6 genotype is a strong predictor of the efficacy of S-1-based chemotherapy.
Log-rank P = 0.032 Median ( CYP2A6/ERCC1 polymorphisms and the S-1/CDDP efficacy SR Park et al and thymidine phosphorylase (Ichikawa et al, 2006;Choi et al, 2010;Koizumi et al 2010). Results from those studies, however, have not been consistent. In addition, genotyping peripheral blood cells, which we did in the present study, is more optimal than evaluating tumour expression levels of mRNA or protein, in which there may not be clinical accessibility or the assay may not be applicable because of arbitrary cutoff levels, lack of standard criteria, or observer variation.

A/A A/C C/C
Log-rank P = 0.041   CYP2A6/ERCC1 polymorphisms and the S-1/CDDP efficacy SR Park et al 2 ¼ CYP2A6*1/*9 or *1/*12 vs group 3 ¼ CYP2A6*1/*2, *1/*4, *9/*12, *9/*4, or *9/*9), the fractional clearance of nicotine to cotinine was about 80% in group 2, and about 50% in group 3 compared with group 1 (Benowitz et al, 2006). The mean total plasma clearance of nicotine was 18.8±6.0, 15.5±4.9, and 11.7±5.1 ml min À1 kg À1 in groups 1, 2, and 3, respectively. These data suggest that CYP2A6*1/*4 results in lower enzyme activity than other W/V genotypes. Although ERCC1 19442C4A was significantly associated with response rate, the exact functional consequences of this polymorphism have yet to be elucidated. The 3 0 untranslated region might affect mRNA stability and result in lower expression levels of the enzyme (Chen et al, 2000); the lower DNA repair capacity would increase the damage done by platinum agents and hence, increase their efficacy. In clinical studies, however, the ERCC1 19442C4A polymorphism has provided variable results. Our finding that the C/C genotype, compared with the C/A or A/A genotype, was associated with poorer treatment efficacy is consistent with studies of cisplatin-treated oesophageal or nonsmall cell lung cancer patients (Bradbury et al, 2009;Kalikaki et al, 2009). In contrast, other studies reported that the A allele was associated with shorter progression-free survival and/or OS in stage III non-small cell lung cancer or epithelial ovarian cancer patients treated with platinum-based chemotherapy (Zhou et al, 2004;Krivak et al, 2008). These mixed results might be attributable to variation in patient or tumour characteristics and treatments delivered across studies.

A/A A/C C/C
Our finding that the number of risk genotypes of CYP2A6 and ERCC1 19442C4A (V/V or *1/*4 and C/C) was significantly associated with treatment efficacy, and especially, the group with both showed very poor clinical outcomes (response rate 16.7%, median TTP 3.2 months, and median OS 7.7 months) suggests that CYP2A6 polymorphisms predict primarily efficacy of S-1-based chemotherapy and ERCC1 19442C4A polymorphisms have an additive predictive role in the S-1 plus cisplatin setting. Further investigation is warranted to determine whether patients with CYP2A6 and ERCC1 19442C4A risk genotypes would have better outcomes if they were treated with fluoropyrimidines that do not require CYP2A6 activation, such as 5-FU itself or capecitabine, and/or non-platinum agents.
The present study had the following limitations: (1) Although we collected clinical data prospectively, this was a retrospective analysis, so our findings should be validated in prospective studies; (2) As pharmacokinetic data were not available, we could not evaluate the association between CYP2A6 genotype and the pharmacokinetic parameters of S-1 components and their active metabolites; (3) As the current study is not a prospective, randomised trial, it cannot be excluded that CYP2A6 gene polymorphism could be a prognostic factor of gastric cancer independent of the effect of S-1; (4) We had no functional data for the ERCC1 19442C4A polymorphisms; (5) We did not analyse polymorphisms in other genes, such as thymidylate synthase that might influence 5-FU efficacy.
In conclusion, our study demonstrated that the CYP2A6 and ERCC1 19442C4A genotypes correlated with the efficacy of S-1 plus cisplatin in MGC patients. These results are consistent with the findings of our previous study, which reported for the first time that the CYP2A6 genotype was associated with the treatment efficacy of S-1 plus docetaxel in MGC patients (Kong et al, 2009). Large-scale randomised prospective studies are warranted to validate our findings, which might provide useful information for selecting appropriate candidates for S-1-based chemotherapy, ultimately leading to a more tailored approach to chemotherapy.