Pharmacogenomics of platinum-based chemotherapy response in NSCLC: a genotyping study and a pooled analysis

Published data showed inconsistent results about associations of extensively studied polymorphisms with platinum-based chemotherapy response. Our study aimed to provide reliable conclusions of these associations by detecting genotypes of the SNPs in a larger sample size and summarizing a comprehensive pooled analysis. 13 SNPs in 8 genes were genotyped in 1024 NSCLC patients by SequenomMassARRAY. 39 published studies and our study were included in meta-analysis. Patients with GA or GG genotypes of XRCC1 G1196 had better response than AA genotype carriers (Genotyping study: OR = 0.72, 95%CI: 0.53-0.96, P = 0.028; Meta-analysis: OR = 0.74, 95%CI: 0.62-0.89, P = 0.001). Patients carrying CT or TT genotypes of XRCC1 C580T could be more sensitive to platinum-based chemotherapy compared to patients with CC genotype (OR = 0.54, 95%CI: 0.37-0.80, P = 0.002). CC genotype of XRCC3 C18067T carriers showed more resistance to platinum-based chemotherapy when compared to those with CT or TT genotypes (OR = 0.69, 95%CI: 0.52-0.91, P = 0.009). Our study indicated that XRCC1 G1196A/C580T and XRCC3 C18067T should be paid attention for personalized platinum-based chemotherapy in NSCLC patients.


INTRODUCTION
Lung cancer is a serious health problem in the whole world for many decades with its highest mortality [1]. It is histologically consisted of small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC). NSCLC accounts for about 85% of the total cases. Platinum-based chemotherapy is the standard chemotherapy for first line treatment of NSCLC, especially for advance stage patients [2,3]. However, the response to platinum-based chemotherapy is greatly variable among individuals, and drug resistance is easily occurred by intrinsic or acquired.
A number of studies suggested single nucleotide polymorphisms (SNPs) may affect chemotherapy response [4][5][6][7][8][9]. Most of the studies focused on the SNPs of genes in DNA repair pathways or transporters, such as ERCC1, XPD, XRCC1 and MDR1 [10][11][12][13][14][15][16][17][18]. However, inconsistent results came from different studies on the same issue. For example, Huang et al. and Zhao et al. reported that ERCC1 C8092A may be useful predictive markers for response to platinum-based chemotherapy [19,20], but some other studies showed the contradictory results [21][22][23]. The same situations were also existed in MDR1 C3435T, XPD A2251C and other SNPs. Although several reviews and meta-analyses summarized the pharmacogenomics of platinum-based chemotherapy response in NSCLC patients [24], they still not reached the consistent conclusion. Moreover, the published meta-analyses were not comprehensive as they usually analyzed only one or several SNPs in their studies [25,26].
In this study, we investigated the relationships between widely studied SNPs and platinum-based chemotherapy response in a larger NSCLC sample size. We also provided a comprehensive meta-analysis on pharmacogenomics of the platinum-based chemotherapy response. Understanding the genetic variants contributed Clinical Research Paper to platinum-based chemotherapy will provide guide for individualized chemotherapy and benefit for NSCLC patients.

MATERIALS AND METHODS Subjects
A total of 1024 NSCLC patients were enrolled in our genotyping study. They were from Shanghai Chest, Shanghai Zhongshan, or Shanghai Changhai Hospitals  Table 1). The patients to be eligible for the study had to meet the following criteria: (1) histologically or cytologically confirmed NSCLC, and primary tumor in the lung; (2) Patients received platinum-based chemotherapy for at least two cycles, and had no surgery or radiotherapy before. Exclusion criteria included (1) pregnancy or lactation, (2) active infection, (3) symptomatic brain or leptomeningeal metastases, and (4) previous or concomitant other malignancies. Chemotherapy response was evaluated by Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 [27]. Responders were consisted of complete responders (CR) and partial responders (PR) while non-responders including stable disease (SD) and progressive disease (PD). All subjects provided written informed consent, incompliance with the code of ethics of the World Medical Association (Declaration of Helsinki) before this study. The study protocol was approved by the ethics committee of Xiangya School of Medicine, Central South University (registration number: CTXY-110008-2). We applied for clinical admission to the Chinese Clinical Trial Registry (registration number: ChiCTR-RO-12002873).

SNP selecting, DNA extracting and genotyping
By considering the meta-analysis of pharmacogenomics of platinum-based chemotherapy in NSCLC, we searched the publications which reported the associations of SNPs with platinum-based chemotherapy response of NSCLC in PubMed database, ISI Web of Knowledge and Cochrane Library. As shown in Figure  1, we found 13 SNPs were widely studied in different publications after systematic literature review. In order to get more reliable results by included more studies and with larger total sample size in the meta-analysis, we did genotyping study about these 13 SNPs in our samples. Information of the 13 SNPs was summarized in Table 2.
Genomic DNA of all subjects was isolated from a 5 mL peripheral blood sample using the FlexiGene DNA Kit (Qiagen, Hilden, Germany) and stored at -20C until use. Genotyping was conducted by the Sequenom MassARRAY system (Sequenom, San Diego, CA, USA).

Publication search and inclusion criteria
We did a systematic literature search in PubMed database, ISI Web of Knowledge and Cochrane Library. The identified articles were reviewed carefully to find more relevant articles. The included articles were published before November 23 th 2015. Keywords for searching the related publications were platinum (platinum, cisplatin, carboplatin, oxaliplatin) and polymorphism (polymorphism, SNP, mutation, variation, single nucleotide polymorphism) and lung cancer.
The inclusion criteria of publications were as follows: (1) studies about platinum-based chemotherapy response; (2) Patients with NSCLC; (3) the data of genotypes in responders and nonresponders could be obtained. Studies were excluded by any one of the following conditions: (1) the data of genotypes could not be provided; (2) articles involved in patients received surgery or radiotherapy.

Data extraction
All data were extracted independently by two investigators (JC and ZW) using the same data recording form, but they were blind to each other during the whole extracting process. The discrepancies of the extracted data were discussed and resolved with consensus. The following information were collected from each study: first author's name, publication year, ethnicity, country, sample size, polymorphisms, alleles of the investigated polymorphism, genotyping methods, disease stage, chemotherapy regimen, and the numbers of responders and non-responders in different genotypes.

Statistical analysis
In our genotyping study, the chi-square and Student t tests were used to determine the differences in sex, age, smoking status and histology between responders and nonresponders. Unconditional logistic regression was performed to estimate the association of the polymorphisms with chemotherapy response by calculating odds ratios and their 95% confidence intervals with adjustments. The association study was analyzed in additive, dominant and recessive models. The P value was 2 sided, and P < 0.05 was considered statistically significant. The aforementioned statistical analyses were performed by PLINK 1.07 [28] and SPSS 18.0 (IBM, Armonk, NY, USA).
In the meta-analysis, the pooled odds ratio (OR) and associated 95% confidence interval (95% CI) were calculated by using the Z test. The genetic model was chosen by logistic regression [29]. The heterogeneity of publications in each meta-analysis was assessed by using Q statistic test, it with a significance level of P < 0.05. We selected the random-effect model to get the results with a wider CIs if P < 0.05. Otherwise, the fixed-effect model was used to calculate the pooled ORs and P values [30,31]. To further evaluate the extent of heterogeneity between publications, I 2 statistic test was also employed, its values of 25%, 50% and 75%were considered as low, moderate and high heterogeneity respectively [32]. The publication bias was examined by the inverted funnel plots, Begg's test [33]and Egger's test [34]. All calculations were conducted by Stata 12.0 (StataCorp LP, College Station, USA). The P value was 2 sided, and P < 0.05 was considered statistically significant.

Associations of the Polymorphisms with platinumbased chemotherapy response in genotyping study
1024 NSCLC patients were enrolled in our genotyping study and their clinical characteristics were summarized in Table 1. All of the patients received platinum-based chemotherapy at least two cycles. 237 of them showed good response while 787 had poor response to the treatment. 13 SNPs attempted to be genotyped by Sequenom's MassARRAY system, but 3 (XRCC1 C580T, CDA A79C, XRCC3 C18067T) of the SNPs were failed in primer design since primers of these 3 SNPs would form heterodimers with other primers. Additionally, 2 SNPs (MDR1 G2677T/A, XPD G934A) were not genotyped successfully in all samples, their genotyping results failed in Hardy-Weinberg equilibrium test. The results of associations between 8 SNPs and platinumbased chemotherapy were shown in Table 3 and Table  S1. XRCC1 G1196A was significantly related to the platinum-based chemotherapy response. Patients with GA or GG genotypes were more sensitive to platinum-based chemotherapy. We also conducted subgroup analyses which samples selected by age (55 years old), sex, smoking status, histology or chemotherapy regimen. The results of subgroup analyses were summarized in Table 4. In patients with <55 years old, GSTP1 A313G and XPG G3310C were related to the chemotherapy response. In patients with ≥55 years old, ERCC1 C354T was associated with chemotherapy response. MDR1 C3435T, G2677T/A and XPD A2251C showed significant associations in patients of females. XRCC1 G1196A was related to drug response in smoking patients. In AC subgroup, ERCC1 C354T and XPG T138C were associated with platinum sensitivity. In patients with VP treatment, XRCC1 G1196A and MDR1 C3435T were correlated with platinum-based chemotherapy response.

Characteristics of eligible studies
Overall 4014 studies were selected during the first step of systematic literature review about platinum and lung cancer. With further reviewed, there were 475 studies were involved in single nucleotide polymorphisms. After     reviewing the abstracts, 32 reviews or meta-analyses and 306 irrelevant studies were excluded. After reading the full texts of the 137 articles which left for reviewed in next step, we found that 41 articles focused on prognosis or toxicity of platinum-based chemotherapy, 21 lacked enough information, 19 were in vitro studies, 9 were about small cell lung cancer, 7 involved in patients with surgery or radiotherapy, and 1 was duplicated publication. Finally, there were 39 publications and our genotyping study included in meta-analysis. The publications included 13 SNPs in 8 genes (Figure 1). The characteristics of these studies were summarized in Table 5. Funnel plot, Begg's test and Egger's test were used to estimate publication bias among the included studies. Visual inspection of the funnel plot of SNPs revealed a symmetrical inverted V shape ( Figure S1).

Stratified analyses of chemotherapy response
Since moderate or high heterogeneity was showed in pooled analysis of several SNPs, and for considering that ethnic differences and different chemotherapy regimen may contribute to the drug response, we conducted stratified analyses of Asian, Caucasian and cisplatinbased chemotherapy. It was interesting to found that XPD A2251C was significantly correlated with platinum response in Asian population by pooling 12 studies [7,21,22,35,[42][43][44][45][46][47][48][49] included 2570 subjects (OR = 1.29, 95%CI: 1.03-1.62, P = 0.026) ( Figure 3A). Patients of Asians with XPD A2251C AA genotype could be more sensitive to platinum-based chemotherapy. XRCC3 C18067T was significantly related to drug response in Caucasian population (OR = 0.50, 95%CI: 0.30-0.85, P = 0.011) ( Figure 3B). Patients of Caucasians with CT and TT genotypes of this polymorphism had better drug response. By only pooling the investigations that patients received cisplatin-based chemotherapy, MDR1 C3435T and MDR1 G2677T/A were significantly correlated to  Figure 3C and 3D). Patients carrying CC genotype of MDR1 C3435T or GG genotype of MDR1 G2677T/A had better chemotherapy response. There were 11 publications studied association of XRCC1 G1196A [7, 22, 36-40, 46, 50-52], but high heterogeneity (P<0.001, I 2 = 74.3%) and publication bias exist among the studies (Begg's test P = 0.024; Egger's test P = 0.002). Moreover, it was hardy to determine about the association of XPD A2251C with chemotherapy response for its 95%CI was 0.99-1.44. Thus, we made quality assessment about the publications by using quality scoring criteria from Wu's study [26], but it was modified by considering high (or low) quality if the score >12 (or ≤12) in our analyses. We did quality stratified analyses in SNPs which reported in more than 10 studies. The results showed that XPD A2251C was more likely related to drug response in pooled analysis of high quality publications which included 2625 patients (OR = 1.22, 95%CI: 1.00-1.50, P = 0.05) ( Figure 4A), and XRCC1 G1196A was associated with chemotherapy response in high quality publications with low heterogeneity. (OR = 0.74, 95%CI: 0.62-0.89, P = 0.001) ( Figure 4B).

Quality assessment of meta-analysis results
The reliability of our meta-analysis results were assessed by the criteria described in Table S2. Score was determined by the following 4 factors: number of included publications, sample size of patients, heterogeneity and publication bias. Total score ranged from 2 to 12. Metaanalysis results were considered low or medium or high reliability if the score were 2-6 or 7-9 or 10-12. The quality assessment results were described in Table 6. We were exciting to notice that high reliability showed about XRCC1 G1196A pooled analysis result in high quality studies, while the total pooled analysis result was in low reliability. Moreover, our genotyping study result of XRCC1 G1196A was consistent with the high reliability result.

DISCUSSION
In this study, we investigated the associations of widely studied SNPs (13 polymorphisms in 8 genes) with platinum-based chemotherapy response in NSCLC patients. We also conducted a comprehensive metaanalysis of these SNPs. Our results showed that XRCC1 G1196A/C580T, and XRCC3 C18067T were significantly correlated with platinum-based chemotherapy. XPD A2251C, MDR1 C3435T and MDR1 G2677T/A, GSTP1 A313G, XPG G3310C, ERCC1 C354T were correlated to platinum-based chemotherapy response in different subgroups.
Platinum-based chemotherapy was widely used for treatment of advanced NSCLC, but pharmacogenomic differences between individuals may affect drug response. In the last several decades, gene polymorphisms were revealed to play an important role in chemotherapy response [24]. Therefore, most studies focused on the polymorphisms of genes involved in DNA repair pathway, transporters, metabolism and detoxification. In this study, we selected 9 SNPs in DNA repair pathway genes, 2 SNPs of transporter genes and 2 from metabolism and detoxification genes. They were extensively studied by researchers but results were not consistent. We used a larger sample size (n = 1024) to detect relationships between these SNPs and platinum-based chemotherapy response in NSCLC. Moreover, we conducted a metaanalysis to verify the results. DNA repair was one of the classical platinum resistance mechanism [53]. Polymorphisms in genes of DNA repair pathway could affect DNA repair capacity and alter sensitivity to platinum-based chemotherapy. In this study, XRCC1 G1196A was significantly correlated to platinum-based chemotherapy response both in our genotyping study and meta-analysis. XRCC1 C580T and XRCC3 C18067T were also related to the drug response from the meta-analysis. XRCC1 and XRCC3 were both DNA base excision repair (NER) genes. XRCC1 G1196A/ C580T and XRCC3 C18067T were all non-synonymous SNPs. They could directly contribute to gene expression and activity. Increased NER capacity would lead to decrease sensitivity of patients to chemotherapy [54]. A recent meta-analysis conducted by Gao et.al [55] about XRCC1 G1196A and C580T showed consistent result with our findings about XRCC1 C580T, but no significantly association showed about XRCC1 G1196A in their study. We carefully compared the included studies in the two meta-analyses. Most of studies included in   their meta-analysis were small sample size studies, and they also included the studies with patients received radiotherapy which we excluded [56,57]. For DNA repair gene polymorphisms, additionally, we also found that XRCC3 C18067T showed contribution to drug response in Caucasians and XPD A2251C was related to the response in Asian population. Allele frequency usually different between races, and racial differences were important factors for drug response [58,59]. We did not find any statistical evidence for associations between the other SNPs (ERCC1 C354T/C8092A, XPD G934A and XPG G3310C/T138C) in DNA repair pathway and drug response in the meta-analysis. In our genotyping study, these polymorphisms also showed no significant correlations to platinum-based chemotherapy response in the overall analysis.
Transporters involved in drug resistance by decrease uptake or efflux of the drugs by the proteins known as ATP binding cassette transporters [60]. MDR1 also named ABCB1, it encodes P-glycoprotein which is an ATPdependent drug efflux. It is responsible for decreased drug accumulation and often mediates the development of resistance to anticancer drug [61][62][63]. We found that both MDR1 C3435T and G2677T/A were associated with cisplatin-based chemotherapy response from the results of meta-analysis. However, the numbers of subjects pooled for these two SNPs in cisplatin-based chemotherapy subgroup were small. We considered that these associations maybe need further investigations. Additionally, there were several findings in our genotyping study about the SNPs. Both MDR1 C3535T and G2766T/A were associated with platinum-based chemotherapy response in females, and MDR1 C3435T also related to platinum response in AC patients.
GSPT1 and CDA were metabolism and detoxification pathway genes. GSTP1 was a member of GST family. It mediates formation of platinum-glutathione adduct [64] which decreased chemotherapy sensitivity. 7 studies were included in GSTP1 A313G pool analysis [51,52,[65][66][67], but no significant correlation between this polymorphism and chemotherapy response. However, we found GSTP1 A313G related to platinum-based chemotherapy response of patients with <55 year old in our genotyping study. CDA is an enzyme that metabolic inactivation of gemcitabine [68]. The patients in 4 studies of CDA A79C were received platinum-gemcitabine treatment [7,41,45,69], but no associations were found in the meat-analysis of CDA A79C.
We conducted this study to get comprehensive conclusions about SNPs contributed to platinum-based chemotherapy response in NSCLC patients, but there were several possible limitations. It was unfortunately about the genotyping of XRCC1 C580T and XRCC3 C18067T failed in our samples, thus meta-analyses of these two SNPs were calculated without the results from our genotyping study. In meta-analyses of ERCC1 and MDR1, we did not include the data from the studies by Krawczyk et al. [70] and Du et al. [71] because they classified SD patients as responders. Our meta-analysis mainly used unadjusted estimates because not all publications presented their adjusted estimates. Finally, we attempted to conduct meta-analysis in each subgroup, but the other publications did not conduct subgroup analysis, or when they did but not in the same way. Thus, we cannot do pool analyses in the subgroups.
It was our first attempt to evaluate the quality of meta-analysis results. Only ERCC1 C3435T/C8092A, XPD A2251C/G934A, XRCC1 G1196A and XRCC3 C18067T showed high reliability. It means the other SNPs may need more investigations, though most of them showed medium reliability. In conclusion, contributions of XRCC1 G1196A/C580T and XRCC3 C18067T to drug response were further confirmed by our large sample size genotyping study and comprehensive meta-analysis. It is meaningful for personalized platinum-based treatment of NSCLC. Studies investigate associations of these SNPs with platinum-based chemotherapy response in NSCLC could be less meaningful after our study.