Characterization of Functional Excision Repair Cross-Complementation Group 1 Variants and Their Association with Lung Cancer Risk and Prognosis

Lung cancer is widespread all over the world, with high incidence rate and leading lethality. Small cell lung cancer (SCLC) accounts for 15% to 25% of total lung cancer, tending to be more aggressive than non-SCLC (NSCLC). Chemotherapy is the mainstay of treatments for SCLC (1). Platinum-based regimens, with good clinical responsivity, have been used as standard first-line chemotherapy of SCLC patients. However, the intrinsic or treatment-induced resistance still occurs in some patients and thus limits the therapeutic efficacy of the drugs.

The DNA repair capacity, especially nucleotide excision repair activity, has been linked to genetic predisposition of lung cancer or platinum chemotherapy outcome (2–4), because of its important role in the removal of the adducts formed by environmental etiologic factors, such as tobacco smoke, or chemotherapy drugs, such as platinum-based regimens. The excision repair cross-complementation group 1 (ERCC1), combined with xeroderma pigmentsum group F (5, 6), acting as a heterodimeric, structure-specific endonuclease, plays a rate-limiting role in nucleotide excision repair by catalyzing the incision on the 5′-side relative to the damaged site (7). In addition, ERCC1/xeroderma pigmentsum group F exerts important nucleotide excision repair–independent roles in the DNA interstrand crosslink repair (8), recombination processes (9), and genome integrity by shaping the telomeres (10).

In the last 20 years, many efforts were made to investigate the roles of ERCC1 in cancer susceptibility or clinical resistance (11–16). This project was highlighted in the last year by the work of Olaussen et al. (17), who showed that the NSCLC patients with ERCC1-negative tumors had a longer survival time than those with ERCC1-positive tumors. Subsequently, Zheng et al. (18) showed a joint effect of ERCC1 and RRM1 expression as determinants of survival in the patients with early-stage NSCLC. These solid evidences proposed that, as pivotal components of nucleotide excision repair, heightened expression of ERCC1/xeroderma pigmentsum group F means low risk of lung cancer development but high probability of poor chemotherapy prognosis. In addition, it has been shown that the ERCC1 RNA levels are highly correlated with nucleotide excision repair activity in blood lymphocytes (19) and lung cancer patients have lower ERCC1 RNA levels in their blood cells than cancer-free controls (20), although inconsistent result has also been reported (21).

Single-nucleotide polymorphisms (SNP) in DNA repair genes, which may contribute to interindividual diversity in DNA repair capacity, may be the more efficient predictive factors for cancer susceptibility and platinum-chemotherapy outcome compared with the mRNA levels. Numerous studies have investigated the association between ERCC1 SNPs, e.g., Asn118Asn or 8092C>A, and susceptibility to survival status of several types of human cancer (22–26). However, most of these studies investigated only one or two SNPs but not all SNPs in ERCC1, and the results are not all consistent. Furthermore, little or nothing has been known about the functional significance of the most investigated SNPs in the gene. Thus, to identify all variants that influence ERCC1 expression and/or its protein function and examine the effects of these SNPs on susceptibility to carcinogenesis and chemotherapy outcome are highly warranted.

Here, we reported a comprehensive analysis of ERCC1 polymorphisms and their association with risk for developing lung cancer and outcome of SCLC patients treated with platinum-based chemotherapy. We identified five SNPs by resequencing the 5′-flanking region, all exons with their surrounding introns, and the 3′-untranslated region (UTR) of the ERCC1 locus and found two functional SNPs located in the 5′-flanking region that are associated with lung cancer susceptibility and prognosis of lung cancer patients treated with platinum-based chemotherapy.

SNP detection. Thirty DNA samples derived from blood of randomly selected healthy subjects (all were Han Chinese) were used to search for SNPs within the ∼2-kb promoter region, 5′-UTR, coding regions, and 3′-UTR of ERCC1 (Genbank accession no. AF512555). These samples included 60 chromosomes, providing at least a 95% confidence level to detect alleles with frequencies of >5%. In reference to the human ERCC1 gene sequence, 13 sets of PCR primers were designed for SNP screening (available upon request). The 13 fragments from each subject were amplified, and SNPs were identified by directly sequencing the PCR products with ABI PRISM Dye Terminator Sequencing kits (Applied Biosystems) and loading the samples onto an ABI 3730 sequencer. Finally, we used the Mutation Explorer program (Todaysoft, Inc.) to identify SNP candidates that were then confirmed by two independent observers. We further confirmed these SNP positions and individual genotypes by reamplifying and resequencing the SNP site from the opposite strand.

Study subjects. This study recruited two sets of subjects. The case-control analysis consisted of 1,000 patients with lung cancer and 1,000 cancer-free control subjects. All subjects were genetically unrelated ethnic Han Chinese. The characteristics of the study subjects have been described previously (27, 28). In brief, the eligible patients with newly diagnosed, histopathologically confirmed, and previously untreated (by radiotherapy or chemotherapy) primary lung cancer were consecutively accrued between January 1997 and June 2002 at Cancer Hospital, Chinese Academy of Medical Sciences. There were no age, sex, stage, or histology restrictions; however, patients with previous cancer or metastasized cancer from other organs were excluded. The control subjects were randomly selected from a pool of cancer-free subjects recruited from a nutritional survey conducted in the same region during the same time period as the cases were collected. The selection criteria include no prior history of cancer, and controls were frequency matched to the cases by age (± 5 y) and sex. At recruitment, informed consent was obtained from each subject who was then interviewed for detailed information on demographic characteristics and lifetime history of tobacco use. Of the 1,000 patients and 1,000 control subjects defined in the previous study, only 988 patients and 986 control subjects were successfully genotyped in the present study, because the DNA samples of the rest of the subjects was no longer available.

For analysis of chemotherapy outcome, 162 SCLC patients were accrued between July 2000 and July 2005. All patients were ethnic Han Chinese diagnosed as SCLC by pathologic or cytologic examination. Patients were first evaluated by a group of experts consisting of surgeons, medical oncologists, pathologists, and image diagnosticians, and those who were eligible for chemotherapy were treated with carboplatin plus etoposide (CE) regimen. Tumor was measured by image diagnosticians, and response was evaluated by medical oncologists according to WHO criteria for solid tumors. Response evaluation was repeated every two treatment cycles and lasted six to eight cycles or until occurrence of disease progression. Time to progression (TTP) was defined as from beginning of treatment to documented progression. When patients experienced disease recurrence during CE treatment, they were given second-line chemotherapy. Survival time was calculated from the date when patient received the CE chemotherapy to the date of last follow-up or death. Dates of death were obtained from (a) inpatient and outpatient records, (b) patient's family, or (c) local Public Security Census Register Office. Patients who were not departed were censored at the last date they were known to be alive based on the date of last contact. In this study, follow-up was terminated when patients survived >5 y.

This study was approved by the Institutional Review Board of the Chinese Academy of Medical Sciences Cancer Institute and Hospital.

Genotyping of ERCC1 polymorphisms. Five ERCC1 SNPs identified by DNA sequencing, i.e., −433T>C, 262G>T, 3525C>T, 4855C>T, and 14443C>A were genotyped using PCR-based RFLP methods, done without knowledge of case-control status of the subjects. The primers and PCR conditions for amplifying DNA fragments containing these polymorphic sites are available upon request. The PCR products were digested with restriction enzyme (New England Biolabs) TaqI (for 262G>T), MspI (for 4855C>T), MboII (for 14443C>A), or BstuI (for −433T>C or 3525C>T) to determine genotypes, and each PCR product contains only one respective restriction site. The genotypes identified by PCR-RFLP methods were further confirmed by direct DNA sequencing of the PCR products. A 10% masked random sample of cases and controls was tested twice by different persons, and the results were concordant for all the masked duplicate sets.

Construction of reporter plasmids. Five DNA fragments corresponding to the enhancer region, basic region, or 5′-UTR of the ERCC1 5′-flanking region were generated by PCR (primers are available upon request) and subcloned into pGL3-Basic vector (Promega). The resultant plasmids, designated p-T-T, p-P-T, p-B, p-U-T, or U-T, respectively, were sequenced to confirm containing exclusively wild-type alleles at −433 and 262 positions relative to transcriptional start site. The p-T-T construct (contains −433T/262T) was then site-specifically mutated to create construct p-C-T, p-T-G, or p-C-G, which contains −433C/262T, −433T/262G, or −433C/262G, respectively. All the four constructs were identical, except for the different allele at −433 or 262 polymorphic site. All constructs used in this study were restriction mapped and sequenced to confirm their authenticity.

Transient transfections and luciferase assays. HeLa and A549 cells were used for luciferase assays. Cells were plated in a multiwell plate and grown to 60% to 70% confluence. The constructed reporter plasmid (50 ng) or the empty pGL3-Basic vector (50 ng) was cotransfected with pRL-SV40 (1 ng; Luciferase Assay System, Promega) to cells, respectively, using LipofectAMINE reagent (Life Technologies, Inc.) according to the manufacturer's protocol. The pRL-SV40, containing Renilla reniformis luciferase, was used to standardize transfection efficiency. For each plasmid construct, three independent transfection experiments were done, and each was done in triplicate. The empty pGL3-Basic vector cotransfected with pRL-SV40 plasmid served as a reference. Fold increase was calculated by defining the activity of empty pGL3-Basic vector as 1. Differences were determined by t test, and P < 0.01 was considered significant.

Electrophoretic mobility shift assays. For electrophoretic mobility shift assays (EMSA), synthetic double-stranded and 5′ biotin-labeled oligonucleotides corresponding to the ERCC1 −433T, −433C, 262T, or 262G sequences and HeLa cell nuclear extract (Promega) were incubated at 25°C for 20 min using the Light Shift Chemiluminescent EMSA kit (Pierce). The reaction mixture was separated on 6% PAGE, and the products were detected by stabilized streptavidin–horseradish peroxidase conjugate (Pierce). Unlabeled oligonucleotides at 200-fold molar excess were added to the reaction for competition.

Real-time analysis of ERCC1 RNA. Total RNA was isolated from surgically removed normal lung tissues adjacent to the tumors of 34 individual patients and then converted to cDNA using oligo (dT)₁₅ primer and Superscript II (Invitrogen). ERCC1 RNA was measured by real-time quantitative reverse transcription–PCR using the ABI Prism 7000 sequence detection system (Applied Biosystems) based on the SYBR-Green method. Each assay was done in triplicate, and the expression of individual ERCC1 RNA measurements was calculated relative to expression of β-actin using a modification of the method described by Lehmann and Kreipe (29). The primers and probes used for detecting ERCC1 RNA are available upon request.

Haplotype construction and statistical analysis. The associations between ERCC1 genotypes and risk of developing lung cancer were assessed by odds ratios (OR) and their 95% confidence intervals (95% CI), which were computed by logistic regression models using Statistical Analysis System software (version 9.0; SAS Institute). We tested the null hypotheses of multiplicative gene-smoking interaction and evaluated departures from multiplicative joint effect by including main effect variables and their product terms in the logistic regression model. A more than multiplicative joint effect (30) was suggested when OR₁₁ > OR₁₀ × OR₀₁. The homogeneity test was done to compare the difference between smoking-related ORs among different genotypes or between the product of related ORs and joint effect OR. Linkage disequilibrium was analyzed using Haploview software (31). We used Haplo.score approach to test the association of statistically inferred haplotypes with lung cancer and Haplo.glm to calculate adjusted ORs and 95% CIs for each haplotype (32, 33). Both Haplo.score and Haplo.glm were implemented in the Haplo.stats software compiled by the R language. This program has the advantage that adjustment for covariates and computation of simulation P values for global and each haplotype can be done. The number of simulations for empirical P values was set as 1,000. The associations between overall survival of SCLC patients and ERCC1 genotypes were estimated using Kaplan-Meier method and assessed using the log-rank test. Cox proportional hazard models were used to adjust for disease stage and treatment, and genotypes were treated as indicator variables. These statistical analyses were also done using Statistical Analysis System software. All statistical tests were two-sided, and P < 0.05 was considered significance.

SNP identification and linkage disequilibrium analysis. Five SNPs, −433T>C, 262G>T, 3525C>T, 4855C>T, and 14443C>A, were identified by resequencing ERCC1 5′-flanking region, coding region, and 3′-UTR in DNAs from 30 subjects. These SNPs are located in the promoter region (−433T>C, rs3212930), intron 1 (262G>T, rs2298881), exon 2 (3525C>T, rs11615), the intron immediately 3′ to the exon 3 (4855C>T, rs3212961), and 3′-UTR (14443C>A, rs3212986), respectively, and have been recorded in the National Center for Biotechnology Information SNP database. All the SNPs except −433T>C were common, with the minor allele frequencies in 986 controls being 0.456, 0.257, 0.489, and 0.298 for the 262T, 3525T, 4855T, and 14443A allele, respectively. The −433C allele is 0.047, relatively rare in our study population (Supplementary Table S1). D′ and r² values calculated by Haploview software revealed that the five SNPs were in linkage disequilibrium (Supplementary Table S2).

Effects of −433 and 262 SNPs on ERCC1 promoter activity. To investigate the preliminary role of the 5′-flanking region of ERCC1, a set of deletion constructs were generated (Fig. 1A) and transfected into HeLa and A549 cells. Intriguingly, the reporter gene expression driven by p-U-T, containing just the 5′-UTR, was nearly 3-fold higher than that driven by p-B, containing the basic transcriptional region. The p-T-T construct retaining the full-length 5′-flanking region displayed a 2-fold higher activity compared with p-P-T, carrying the same promoter region as p-T-T but lacking the 5′-UTR (Fig. 1B). Similar results observed in both HeLa and A549 cells highlight the importance of the 5′-UTR in transcriptional activation of ERCC1. However, although transcriptional activity of p-U-T was comparable with that of the p-T-T in A549 cells, it was significantly different in HeLa cells, implying a minor, possible negative, transcriptional regulation role of the promoter region between −1048 and −420 (Fig. 1B). We next examined whether the −433T>C or 262G>T change has functional effect on transcriptional activity of the ERCC1 promoter. The results showed that in both HeLa and A549 cells, p-T-T, p-T-G, and p-C-T produced comparable reporter gene expression; but the p-C-G produced a significantly lower reporter gene expression, indicating a synergic effect of the −433C and 262G SNPs on reduction of the ERCC1 promoter activity (Fig. 1C).

EMSA was then done to examine whether the −433C or 262G variant diminishes its ability to bind transcriptional factors. Under our experiment conditions, HeLa cell nuclear extract was able to bind both −433T and −433C or 262T and 262G probes (Fig. 2A and B). However, the affinity of the protein to bind 262G probe was considerably lower than that to bind the 262T probe (Fig. 2B, lane 2 versus lane 6). In addition, this DNA-protein interaction could be eliminated by competition assays with both unlabeled 262T or 262G probe but not nonspecific probe, indicating that the binding is sequence-specific (Fig. 2B, lanes 3-5 and lanes 8-10).

Effects of ERCC1 SNPs on ERCC1 RNA expression. We also evaluated the association between −433T>C or 262G>T genotype and ERCC1 RNA levels in surgically removed normal lung tissues adjacent to the tumors of 34 lung cancer patients. Real-time reverse transcription–PCR analyses showed that 262TT genotype carriers had significantly higher mean (±SE) ERCC1 RNA levels than the 262GG genotype carriers [0.00337 ± 0.00086 (n = 12) versus 0.00098 ± 0.00036 (n = 10); P = 0.011]. The 262GT genotype carriers had a level of 0.00173 ± 0.00057 (n = 12) that is lower than 262TT genotype carriers but higher than 262GG genotype carriers, although the differences did not reach statistical significance. These results further suggested that the 262G>T change is associated with decreased ERCC1 transcription. However, no meaningful results were obtained for the −433T>C SNP due to rarity of the variant genotypes. No association between ERCC1 RNA levels and genotypes of the other three SNPs, i.e., 3525C>T, 4855C>T, and 14443C>A, were observed (Supplementary Table S3).

ERCC1 variants and risk of developing lung cancer. A case-control panel of 988 patients and 986 control subjects were genotyped to assess the association between detected ERCC1 SNPs and risk of developing lung cancer. The selected characteristics of study subjects are shown in Supplementary Table S4. The genotype distributions of all the SNPs in patients and control subjects conformed to the Hardy-Weinberg equilibrium. We found that, among the five SNPs, only the −433 site displayed allelic and genotype frequencies that were significantly different between patients and control subjects (Supplementary Table S1). Multivariate logistic regression analysis showed that subjects with the −433CC or TC genotype had a 1.84-fold (95% CI, 1.37-2.45) increased risk for developing lung cancer compared with those with the −433TT genotype (Table 1).

Because DNA damage caused by tobacco carcinogens is believed to be involved in lung carcinogenesis, we therefore examined the joint effects of ERCC1 SNPs and smoking. It was found that, among smokers who smoked >32 pack-years, a significant joint effect was seen for both −433T>C and 262G>T SNPs but not others (Supplementary Table S5). We therefore further examined this effect between combinations of −433T>C and 262G>T genotypes and smoking (Table 2). Because the −433CC genotype was extremely rare, this genotype was combined with the TC genotype for analysis. Logistic regression analysis showed that the −433T>C and 262G>T polymorphisms were not associated with increased risk of lung cancer among nonsmokers. However, among both light (≤32 pack-years) and heavy (>32 pack-years) smokers, subjects carrying at least one copy of the −433C and one copy of the 262G allele were at significantly elevated risk of developing lung cancer, with the adjusted ORs of 2.67 (95% CI, 1.26-5.65) and 2.83 (95% CI, 1.30-6.16) for light smokers and 8.86 (95% CI, 2.77-28.36) and 8.13 (95% CI, 2.78-24.10) for heavy smokers. Because the OR (8.18; 95% CI, 2.78-24.10) for heavy smokers with the 262GG and −433TC + CC genotypes is more than the product of OR for heavy smokers with the 262TT and −433TT genotypes and the OR for nonsomkers with the 262GG and −433TC + CC genotypes (i.e., 3.50 × 1.43 = 5.01), a multiplicative joint effect between the SNPs and smoking in intensifying risk of lung cancer is suggested (30). Similar effect was also evident among light smokers (Table 2).

Interactions of multiple SNPs within a haplotype may have effect on biological phenotype; we thus proceeded to examine the effect on risk of lung cancer by these five SNPs in ERCC1 in the context of haplotypes. The haplotype frequencies were computed using Haplo.stats software, and the results are presented in Supplementary Table S6. A significant difference in haplotype frequencies was observed between patients with lung cancer and controls (χ² = 49.64; P < 0.0001; degrees of freedom, 18). It was shown that all the C₋₄₃₃-G₂₆₂–containing haplotypes were associated with significantly elevated risk of lung cancer compared with the TTCTC haplotype, with the ORs of 1.75 (95% CI, 1.73-1.78), 5.89 (95% CI, 5.79-5.98), 2.12 (95% CI, 0.98-4.59), and 1.70 (95% CI, 1.22-2.37) for the CGCTA, CGTCC, CGCCC, and CGCCA haplotypes. Also, it seemed that only the haplotypes containing both −433C and 262G alleles had the effect, suggesting an interaction between the −433C and 262G alleles in the context of haplotype although the numbers of CGCTA and CGTCC haplotypes were small in analysis. In addition, we found that the TTTCA haplotype was associated with a significantly decreased risk of developing lung cancer (OR, 0.58; 95% CI, 0.56-0.60), whereas the TTCCC haplotype was associated with an increased risk (OR, 2.32; 95% CI, 1.42-3.79) compared with the TTCTC haplotype.

We also examined whether there were differences in genotype and haplotype frequencies of the five SNPs among different histologic subtypes of lung cancer, such as SCLC and NSCLC, and found that the association is not dependent on histologic subtype of the cancer (data not shown).

ERCC1 variants and clinical outcome of CE chemotherapy. Because ERCC1 is involved in resistance to platinum-based chemotherapy (17, 18), we additionally examined the effect of ERCC1 variants on clinical outcome of SCLC patients treated with CE regimen. Baseline clinical characteristics of the 162 SCLC patients are summarized in Supplementary Table S7. The median age of these patients was 56 years (range, 27-82 years), and they consisted of 120 (74.1%) limited SCLC and 42 (25.9%) extensive SCLC. Forty-five (27.8%) patients underwent surgery before chemotherapy. Among 117 patients without surgery and thus having measurable tumor, 103 (88.0%) achieved complete response or partial response, whereas 14 (12.0%) displayed stable disease or progressive disease. The overall median TTP was 9 months (range, 1-60 months), and the overall median survival time was 16 months (95% CI, 12-25).

Genotype frequencies of the five ERCC1 SNPs in 162 SCLC patients treated with CE were similar to those in 988 lung cancer patients of the case-control panel. Because of rarity, the −433T>C SNP was excluded for the analysis. We found that although other SNPs, i.e., 3525C>T, 4855C>T, and 14443C>A were not associated with clinical outcome of patients treated with CE (data not shown), the 262G>T genotypes were significantly associated with response to CE chemotherapy and TTP of the disease after treatment. As shown in Supplementary Table S8, among responders (complete response + partial response), 42.7% of patients carried the 262GG genotype; however, among nonresponders (stable disease + progressive disease), none of patients carried this genotype (P = 0.005). Similarly, 46.3% of patients who carried the 262GG genotype had a median TTP of ≥9.0 months, whereas 25.5% of patients with this genotype had median TTP of <9.0 months (P = 0.04). Patients carrying the 262GG genotype had a longer median TTP compared with those with the GT or TT genotype [16 months (95% CI, 12-29) versus 7 months (95% CI, 7-14) or 9 months (95% CI, 8-14); log-rank test, P = 0.024 and 0.012, respectively]. However, genotype frequencies of the 262G>T SNP were not significantly different in terms of sex, age, tumor stage, and surgical treatment of the patients (Supplementary Table S8).

We further investigated whether there was significant difference in median survival time among patients carrying different ERCC1 genotype after treatment with CE. Although no significant association was detected between the other four SNPs and patients' survival time (data not shown), we found that patients carrying the 262GG genotype had significantly longer median survival time than patients carrying the 262GT or TT genotype [30 months (95% CI, 27-49) versus 19 months (95% CI, 15-22) or 17 months (95% CI, 14-24); P = 0.025 and 0.005, respectively; trend test, P = 0.0002; Table 3 and Fig. 3A]. Besides ERCC1 genotype, disease stage was also associated with patient survival time, with the median survival time of patients who had LD-SCLC being significantly longer than those who had ED-SCLC [24 months (95% CI, 19-30) versus 16 months (95% CI, 12-20); P = 0.0001; Table 3 and Fig. 3B]. However, age and sex had no effect on patient survival time. Cox proportional model analyses showed that the adjusted hazard ratios of death for the 262GT and 262TT genotypes were 1.58 (95% CI, 0.96-2.62; P = 0.074) and 1.98 (95% CI, 1.13-3.47; P = 0.017), respectively, compared with that of the 262GG genotype, indicating that this SNP may be an independent prognostic factor. SCLC stage was also independently associated with patient prognosis, with a hazard ratio of death for limited disease patients being 0.50 (95% CI, 0.33-0.77) compared with extensive disease patients.

In the present study, we investigated genetic polymorphisms in the ERCC1 gene and examined their associations with risk of developing lung cancer and with clinical outcome of lung cancer patients treated with platinum-based chemotherapy. We identified five SNPs located in the 5′-flanking region, coding regions, and 3′-UTR of the ERCC1 gene by resequencing DNA samples from a Chinese Han population. Although these SNPs are already documented in the National Center for Biotechnology Information SNPs database, to our knowledge, the −433T>C, 262G>T, and 4855C>T SNPs were studied for the first time. We found that the −433T>C and 262G>T changes substantially alter ERCC1 transcription and significantly associated, alone and in a gene-smoking joint effect manner, with risk of developing lung cancer. In addition, we also observed that the 262G>T SNP was associated with response and survival time in SCLC patients treated with CE chemotherapy. Our results showing that the 262T allele is associated with reduced risk for the development of lung cancer but poor drug response and clinical outcome in patients treated with platinum-based drug compared with the 262G allele are in agreement with the notion that DNA repair system is a two-sided sword in human cancer (34).

Previous studies have characterized the crucial regulatory region of ERCC1 and identified a few transcriptional factors, including AP1, RAS, and MZF1, that participate in transcriptional regulation of the gene (35–37). However, most previous studies focused on the promoter region and the function of the 5′-UTR remains to be elucidated. In the present study, we explored the role of the 5′-UTR in the transcriptional regulation of ERCC1 and the results from reporter gene assays revealed that the 5′-UTR harbors a greater promoter activity compared with the basic region, manifesting that the 5′-UTR may house an internal promoter. In addition, our results indicated that the region between −1,048 and −420, usually thought to be enhancer, may act as minor or even recessive regulator. In light of the role of the regions between −1,048 and −420 and 5′-UTR, a hypothesis that the two SNPs located in these regions might affect ERCC1 transcription was examined by various biochemical assays. The 262G>T was shown to enhance the promoter activity, most likely to be due to higher binding activity of the T allele toward unknown transcriptional factors. Although no meaningful results for the −433T>C SNP were obtained from EMSA and real-time mRNA analysis due to rarity of the CC genotype, a strong joint effect between −433T>C and 262G>T polymorphisms on the promoter activity of ERCC1 was indicated by luciferase assays. Together, these findings indicate that the −433T>C and 262G>T SNPs are of functional significance. These findings also provide biological plausibility in explanation of our observations that these two SNPs are associated with individual's susceptibility to lung cancer and outcome of CE chemotherapy.

We did not find any novel nonsynonymous polymorphisms in the coding region of ERCC1 except for the two reported SNPs, 3525C>T and 14443C>A (also designated Asn118Asn and 8092C>A, respectively, in literature). The associations between 3525C>T and 14443C>A SNPs and lung cancer risk have been reported in several case-control studies, but the results are inconclusive (19–23, 38, 39). In the present study, we did not find a significant association between risk of developing lung cancer and 3525C>T or 14443C>A SNPs in our study population, either in genotype or in haplotype analysis. The inconsistent results among studies might reflect the difference in genetic background and carcinogen exposure in different populations. In addition, most previous studies had relatively small sample size, which often results in spurious genetic associations (40). Notably, the biological functions of the 3525C>T and 14443C>A SNPs are not clear yet, although the former was previously supposed to impair ERCC1 translation (41). In contrast, we found that the −433T>C and the 262G>T SNPs were associated with risk of lung cancer and the −433C and 262G alleles act cooperatively in the context of both genotypes and haplotypes. These results are in agreement with reporter gene assays showing a synergic effect of the −433C and 262G changes on reduction of the ERCC1 promoter activity. Logically, our findings suggest that low ERCC1 expression is a risk factor for developing lung cancer. However, it has previously been reported that low ERCC1 RNA level in blood lymphocytes is not a risk factor of lung cancer (21). This discrepancy raises questions whether the ERCC1 RNA level in blood lymphocytes reflects that in target lung tissue and whether the ERCC1 RNA level may necessarily reflect ERCC1 protein level in different tissues. Further studies are needed to address these questions.

Although evidence has been accumulated to associate the expression or genetic variants of ERCC1 with outcome of platinum-based chemotherapy in certain cancers, little is known about the effect of ERCC1 variants on SCLC chemotherapy. Platinum-based treatments have currently been used as standard first-line chemotherapy of SCLC patients because of their benefit over other regimens in patient survival time and quality of life (42, 43). However, it is a well-known observation that not all SCLC patients response to platinum-based treatment and the clinical outcome of the treatment is of great heterogeneity. Anticancer activity of platinum drugs is mainly mediated via DNA damages caused by platinum, including DNA adduct formation, cross-linking, and oxidative damage, which block replication and consequentially result in cell death. However, these damages can be removed by DNA repair system, resulting in drug resistance. Our results demonstrating that SCLC patients carrying the 262G allele, which is associated with decreased ERCC1 promoter activity, had a better response to CE chemotherapy, longer median TTP, and longer survival time are consistent with previous reports showing lower expression of ERCC1 in tumor is correlated with longer survival time of NSCLC patients (17, 18). However, no significant results were observed for the −433T>C SNP, apparently because of limited statistical power due to the rarity of the SNP in our study population. In addition, disease stage was also associated with patient survival time, which is generally in agreement with previous studies (44, 45).

The present study has some limitations. Although we characterized functional significance of the −433T>C and 262G >T changes, the precise cooperative mechanism of these two SNPs in modulating ERCC1 expression was not clearly elucidated. The analysis of the association between ERCC1 variants and clinical outcome of SCLC patients treated with CE was based on relatively small sample size. Large studies are under way to validate these results.

In conclusion, we identified two functional SNPs (−433T>C and 262G>T) in the 5′-flanking region of the ERCC1 locus by whole-gene resequencing and they are associated with susceptibility to lung cancer and efficacy of platinum-based anticancer treatment in the opposite direction. If confirmed in further perspective studies, the ERCC1 SNPs might serve as simple and less invasive biomarkers for personalized chemotherapy of platinum-based anticancer drugs.