Comprehensive analysis of DNA repair gene polymorphisms and survival in patients with early stage non‐small‐cell lung cancer

Abstract

This study was conducted to analyze a comprehensive panel of single nucleotide polymorphisms (SNP) in DNA repair genes to determine the relationship between polymorphisms and the survival outcome of patients with early stage non‐small‐cell lung cancer (NSCLC). Three hundred and ten consecutive patients with surgically resected NSCLC were enrolled. Forty‐eight SNP in 27 DNA repair genes were genotyped and their associations with overall survival (OS) and disease‐free survival (DFS) were analyzed. Individually, six SNP exhibited significant associations with survival outcome. When the six SNP were combined, OS and DFS decreased as the number of bad genotypes increased (P _trend < 0.0001 for both). Patients with three, and four or five bad genotypes had a significantly worse OS and DFS compared with those carrying zero or one bad genotypes (adjusted hazard ratio [aHR] for OS = 3.53, 95% confidence interval [CI] = 1.25–9.97, P = 0.02, and aHR for DFS = 3.31, 95% CI = 1.41–7.76, P = 0.006; and aHR for OS = 5.47, 95% CI = 1.87–16.00, P = 0.002, and aHR for DFS = 4.42, 95% CI = 1.82–10.74, P = 0.001, respectively). These findings suggest that the six SNP identified can be used as prognostic markers for patients with surgically resected early stage NSCLC. (Cancer Sci 2010; 101: 2436–2442)

Lung cancer, predominantly non‐small‐cell lung cancer (NSCLC), is the most common cause of death worldwide.^{( 1 )} Despite advances made in the diagnosis and treatment of NSCLC in the last few decades, the prognosis for lung cancer is still very poor, with a 5‐year survival rate of ∼15%. In addition, even among patients with resectable NSCLC, the majority still die from disease recurrence. Although the TNM staging system is the best prognostic index for operable NSCLC, patients with the same pathological stage of disease display marked variability in recurrence and survival.^{( 2 )} Therefore, research is currently in progress to identify the molecular markers that are associated with the prognosis of patients with lung cancer. Single nucleotide polymorphisms (SNP) are the most common form of human genetic variations. Several studies have demonstrated that some of these SNP affect the expression or activities of enzymes, and are therefore associated with the risk of cancer.^{( 3} , ^{4 )} In addition, genetic polymorphisms have become increasingly studied as potential prognostic markers in a variety of cancers, including lung cancer.^{( 5} , ⁶ , ^{7 )}

DNA repair systems are essential for the maintenance of genome integrity, and therefore the deregulation of DNA repair genes can dramatically increase the risk of cancer.^{( 8 )} In humans, more than 125 genes are involved in the DNA damage response and several kinds of DNA repair pathways exist, such as base excision repair, nucleotide excision repair, mismatch repair and double strand break repair.^{( 9 )} Although the effects of DNA repair gene polymorphisms on the risk of cancer have been widely investigated, only a few studies have evaluated the prognostic significance of DNA repair gene polymorphisms in patients with cancer. We have hypothesized that individual variation in DNA repair capacity conferred by DNA repair gene polymorphisms influences the survival outcome of patients with lung cancer. To test this hypothesis, we analyzed a comprehensive panel of polymorphisms in DNA repair genes to determine the relationship between polymorphisms and the survival outcome of patients with early stage NSCLC.

Materials and Methods

Patients.  The study included all patients (n = 310) with stage I, II or IIIA (micro‐invasive N2) NSCLC who underwent curative surgical resection at the Kyungpook National University Hospital (Daegu, Korea) between September 1998 and August 2006. Patients who underwent chemotherapy or radiotherapy prior to surgery were excluded to avoid the effects on DNA. All of the patients included in this study were ethnic Koreans. The histological type of lung cancers was as follows: 189 cases of squamous cell carcinomas (SCC); 116 cases of adenocarcinomas; and five cases of large cell carcinomas. The pathological staging of tumors, which was determined according to the International System for Staging Lung Cancer,^{( 10 )} was as follows: stage I, 177 patients; stage II, 46 patients; and stage IIIA, 87 patients. Written informed consent was obtained from all patients prior to surgery and this study was approved by the Institutional Review Board of the Kyungpook National University Hospital.

Polymorphism selection and genotyping.  We searched the public database and related literature to identify potentially functional polymorphisms in DNA repair genes. We selected 48 SNP in 27 DNA repair genes. The SNP selection favored those with a minor allele frequency of at least 5%, those located in the promoter or untranslated region or coding region of the gene, those previously evaluated in relation to cancer, or those with evidence of functional significance. The selected SNP were genotyped using a sequenome mass spectrometry‐based genotyping assay. Of the set of 48 SNP selected, seven SNP that were monomorphic (APEX L104R, E126D, R237A and D283G; and MSH2 rs17224094, rs17217681 and rs17217877), one SNP (RAD52 rs11226) with a genotyping call rate <90.0%, and one SNP (ATR rs2227930) that was deviated from the Hardy–Weinberg equilibrium (HWE; P < 0.05) were excluded from further analysis. The remaining 39 SNP in 26 genes were evaluated for association analysis. The SNP identification numbers, genotype and minor allele frequencies, genotyping call rates and P‐values for HWE of the 39 SNP are shown in Table 1. For quality control, the genotyping analysis was blinded with respect to the patients. In addition, approximately 10% of the samples were randomly selected to be genotyped again by PCR‐RFLP or DNA sequencing.

Table 1.

 Genotype frequencies of DNA repair gene polymorphisms analyzed in patients (n = 310) with non‐small‐cell lung cancer

DNA repair	Gene		Polymorphism†		Genotype‡, n						MAF in healthy populations
	Symbol	Name	ID no.	Base (amino acid) change	1/1	1/2	2/2	CR (%)	MAF	HWE P	Asian†	European†	African†
Base excision repair	PARP1/ADPRT	ADP‐ribosyltransferase	rs1136410	T>C (V762A)	97	143	66	98.7	0.449	0.330	0.456	0.167	0.008
	APEX/APE1	Apurinic endonuclease I	rs1130409	T>G (D148E)	103	136	59	96.1	0.426	0.248	0.322	0.508	0.275
	LIG3	Ligase III	rs3135967	IVS3+579A>G	167	106	23	95.5	0.257	0.288	0.233	0.467	0.067
	MBD4	Methyl‐CpG binding domain protein 4	rs140693	G>A (E346K)	129	131	36	95.5	0.343	0.758	0.444	0.217	0.183
	OGG1	8‐Oxoguanine DNA glycosylase	rs1052133	G>C (C326S)	84	153	59	95.5	0.458	0.478	0.477	0.776	0.856
	XRCC1	X‐ray repair cross‐complementing group 1	rs3213245	−77T>C	234	55	3	94.2	0.104	0.907	0.114	0.403	0.460
			rs25489	G>A (R280H)	231	60	5	94.8	0.112	0.680	0.041	0.033	0.025
			rs25487	G>A (R399Q)	160	96	25	90.6	0.260	0.061	0.279	0.431	0.100
Direct reversal	MGMT/AGT	O 6 ‐methylguanine‐DNA methyltransferase	rs1625649	−485G>T	139	142	28	99.7	0.320	0.331	–	–	–
			rs12917	C>T (L115F)	244	56	7	99.0	0.114	0.089	0.178	0.125	0.175
Nucleotide excision repair	XPA	Xeroderma pigmentosum group A	rs3176629	−501C>T	252	50	1	97.7	0.086	0.367	0.057	0.000	0.375
			rs1800975	114A>G	83	152	74	99.7	0.485	0.787	0.534	0.586	0.763
	XPC	Xeroderma pigmentosum group C	rs2607775	−27C>G	272	22	1	95.2	0.041	0.445	0.089	0.491	0.407
			rs1870134	G>C (V16L)	154	105	27	92.3	0.278	0.149	0.182	0.000	0.000
			rs2228000	C>T (A499V)	157	119	20	95.5	0.269	0.689	0.456	0.283	0.058
			rs2228001	A>C (K939Q)	119	129	54	97.4	0.392	0.070	0.356	0.400	0.258
	ERCC1	Excision repair cross‐complementing group 1	rs11615	C>T (N118N)	175	100	19	94.8	0.235	0.362	0.289	0.650	0.017
	XPD/ERCC2	Xeroderma pigmentosum group D	rs1799793	G>A (D312N)	262	38	3	97.7	0.073	0.231	0.102	0.314	0.068
			rs1052555	C>T (D711D)	247	38	1	92.3	0.070	0.717	0.034	0.306	0.080
			rs13181	T>G (K751Q)	246	49	2	95.8	0.089	0.795	0.081	0.333	0.183
	XPF/ERCC4	Xeroderma pigmentosum group F	rs1799801	T>C (S835S)	187	100	18	98.4	0.223	0.348	0.318	0.250	0.110
	XPG/ERCC5	Xeroderma pigmentosum group G	rs17655	G>C (D1104H)	94	141	57	94.2	0.437	0.752	0.478	0.733	0.467
Mismatch repair	MLH1	MutL (E. coli) homolog 1	rs1800734	−93A>G	87	143	68	96.1	0.468	0.530	0.489	0.800	0.917
			rs1799977	A>G (I219V)	280	18	0	96.1	0.030	0.591	0.078	0.333	0.042
	MSH2	MutS (E. coli) homolog 2	rs1863332	−434A>C	182	91	16	93.2	0.213	0.306	0.178	0.058	0.183
			rs3732183	IVS10+12A>G	110	142	54	98.7	0.408	0.487	0.400	0.658	0.283
Double strand break repair	BRCA1	Breast cancer 1	rs799917	C>T (P871L)	138	135	26	96.5	0.313	0.384	–	–	–
	BRCA2	Breast cancer 2	rs144848	T>G (N372H)	159	124	15	96.1	0.258	0.139	0.307	0.292	0.125
	LIG4	Ligase 4	rs1805388	C>T (T9I)	228	66	7	97.1	0.133	0.399	0.056	0.167	0.158
	NBS1	Nijmegen breakage syndrome	rs1063045	C>T (L34L)	98	134	61	94.5	0.437	0.228	0.455	0.283	0.292
			rs1805794	C>G (Q185E)	101	137	63	97.1	0.437	0.193	0.400	0.292	0.150
	RAD51	Rad51 (S. cerevisiae) homolog	rs1801320	170G>C	233	55	5	94.5	0.111	0.409	0.146	0.067	0.250
	XRCC3	X‐ray repair cross‐complementing group 3	rs1799796	IVS6‐14A>G	143	120	40	97.7	0.330	0.069	0.289	0.317	0.125
DNA damage response	ATM	Ataxia‐telangiectasia mutated	rs664143	IVS61+60C>T	74	152	72	96.1	0.497	0.728	0.556	0.508	0.325
	ATR	Ataxia‐telangiectasia and Rad3‐related	rs2227928	T>C (M211T)	102	144	53	96.5	0.418	0.860	0.422	0.583	0.833
			rs2229032	G>A (R2425Q)	236	61	5	97.4	0.118	0.646	0.122	0.125	0.033
	CHEK1/CHK1	Check point kinase 1	rs521102	IVS11+35C>T	115	133	56	98.1	0.403	0.114	0.413	0.500	0.214
	CHEK2/CHK2	Check point kinase 2	rs2267130	IVS8‐200T>C	164	111	30	98.4	0.280	0.087	0.271	0.396	0.239
	TP53	Tumor protein p53	rs1042522	G>C (R72P)	106	136	60	97.4	0.424	0.176	0.409	0.233	0.669

†Information about single nucleotide polymorphisms (SNP) and SNP identification and MAF in other ethnic populations (Asian, European and African) were obtained from the NCBI database (http://ncbi.nih.gov). The transcription start site was counted as +1 in the reference sequences. ‡Wild‐type allele is denoted by 1 and the polymorphic allele by 2. CR, call rate; HWE, Hardy–Weinberg equilibrium; MAF, minor allele frequency.

Statistical analysis.  The demographic and clinical information was compared across genotype and stage using chi square tests for categorical variables. The HWE was tested using a goodness‐of‐fit chi square test with one degree of freedom. The degree of linkage disequilibrium between polymorphisms was measured using Haploview.^{( 11 )} The haplotypes and haplotype frequencies were estimated based on a Bayesian algorithm using the Phase program.^{( 12 )} Overall survival (OS) was measured from the day of surgery until the date of death or the date of the last follow up. Disease‐free survival (DFS) was calculated from the day of surgery until recurrence or death from any cause. The survival estimates were calculated using the Kaplan–Meier method. The differences in OS or DFS across different genotypes were compared using the log‐rank test. Hazard ratios (HR) and 95% confidence intervals (CI) were estimated using multivariate Cox proportional hazards models, with adjustment for age (≤64 versus >64 years), gender (male versus female), smoking status (never‐ versus ever‐smoker), pathological stage (I versus II–IIIA) and adjuvant therapy. All analyses were performed using the Statistical Analysis System for Windows, version 9.1 (SAS Institute, Cary, NC, USA).

Results

Patient characteristics and clinical predictors.  The clinical and pathological characteristics of patients and the association with OS and DFS are shown in Table 2. There were 114 deaths (36.8%). The estimated 5‐year OS and DFS for all the patients was 53.3% (95% CI = 45.8–60.2%) and 44.0% (95% CI = 36.9–50.9%). Upon univariate analysis, the pathological stage was significantly associated with OS and DFS (P < 0.0001 for OS and DFS).

Table 2.

 Univariate analysis for overall survival by demographics, smoking status, histological type and pathological stage

Variables	No. cases	Overall survival			Disease‐free survival
		No. deaths (%)†	5Y‐OSR (%)‡	Log‐rank P	No. events (%)†	5Y‐DFSR (%)‡	Log‐rank P
Overall	310	114 (36.8)	53	–	148 (47.7)	44	–
Age (years)
≤64	147	50 (34.0)	55	0.44	68 (46.3)	45	0.79
>64	163	64 (39.3)	52		80 (48.1)	43
Gender
Male	245	98 (40.0)	51	0.17	121 (49.4)	43	0.74
Female	65	16 (24.6)	67		27 (41.5)	51
Smoking status
Never	61	16 (26.2)	67	0.24	27 (44.3)	47	0.97
Ever	249	98 (39.4)	51		121 (48.6)	44
Pack‐years§
≤39	150	60 (40.0)	48	0.99	74 (49.3)	41	0.58
>39	99	38 (39.4)	56		47 (47.5)	48
Histological type¶
Squamous cell carcinoma	189	70 (37.0)	54	0.50	88 (46.6)	46	0.50
Adenocarcinoma	116	42 (36.2)	52		58 (50.0)	40
Pathological stage
I	177	45 (25.4)	61	<0.0001	59 (33.3)	56	<0.0001
II	46	23 (50.0)	44		29 (63.0)	31
IIIA	87	46 (52.9)	43		60 (69.0)	28
Adjuvant therapy††
No	62	36 (58.1)	42	0.57	42 (67.7)	27	0.75
Yes	71	33 (46.5)	46		47 (66.2)	31

†Row percentage. ‡Five‐year overall survival rate (5Y‐OSR) and 5‐year disease free survival rate (5Y‐DFSR): proportion of survival derived from Kaplan–Meier analysis. §In ever‐smokers. ¶Five large‐cell carcinomas were excluded from this analysis. ††In pathological stage II + IIIA: 55 cases received chemotherapy; five cases received radiotherapy; and two cases received chemotherapy and radiotherapy.

Genotype frequencies and effect on OS and DFS.  Thirty‐three SNP and their haplotypes were not significantly associated with survival outcome. Six SNP (PARP1 rs1136410T>C [V762A], XRCC1 rs25489G>A [R280H], MGMT rs12917C>T [L115F], XPC rs2228000C>T [A499V], XPD rs1799793G>A [D312N] and MSH2 rs3732183A>G [IVS10+12]) were significantly associated with survival outcome in multivariate analysis. Patients with the PARP1 rs1136410 CC genotype had a significantly better OS and DFS compared with those carrying the TT or TC genotype (adjusted HR [aHR] for OS = 0.50, 95% CI = 0.28–0.88, P = 0.02; and aHR for DFS = 0.48, 95% CI = 0.23–0.78, P = 0.003). The XRCC1 rs25489, XPC rs2228000 and MSH2 rs3732183 SNP exhibited a better survival outcome under a dominant model for the variant allele for each SNP. The MGMT rs12917C>T exhibited a worse OS under a dominant model for the variant T allele, and the XPD rs1799793G>A exhibited a worse OS under a recessive model for the variant allele (Table 3). The association of MGMT rs12917C>T genotypes with the survival of patients was further examined after categorizing the patients by smoking status. The rs12917 CT or TT genotype exhibited a significantly worse OS in ever‐smokers (aHR for OS = 1.91, 95% CI = 0.11–3.29, P = 0.02), but not in never‐smokers (aHR = 1.10, 95% CI = 0.47–2.58, P = 0.82). The genotype and polymorphic allele frequencies of the six polymorphisms were not significantly different according to patient‐ or tumor‐related factors, such as age, gender, smoking status, histological subtype, pathological stage or adjuvant therapy (data not shown).

Table 3.

 Overall survival and disease‐free survival according to the polymorphisms in DNA repair genes

Polymorphism/Genotype	No. cases (%)†	Overall survival				Disease‐free survival
		No. deaths (%)‡	5YSR (%)§	HR (95% CI)¶	P¶	No. events (%)‡	5YSR(%)§	HR (95% CI)¶	P¶
PARP1/ADPRT rs1136410††
TT	97 (31.7)	40 (41.2)	48	1.00	–	52 (53.6)	37	1.00	0.97
TC	143 (46.7)	56 (39.2)	52	0.94 (0.62–1.42)	0.76	73 (51.1)	41	0.99 (0.69–1.42)	0.005
CC	66 (21.2)	14 (21.2)	68	0.47 (0.26–0.87)	0.02	19 (28.8)	66	0.47 (0.28–0.80)	–
TT + TC	240 (78.4)	96 (40.0)	51	1.00	–	115 (52.1)	40	1.00	0.003
CC	66 (21.2)	14 (21.2)	68	0.50 (0.28–0.88)	0.02	19 (28.8)	66	0.48 (0.23–0.78)	–
XRCC1 rs25489††
GG	231 (78.6)	87 (37.7)	51	1.00	–	115 (48.8)	41	1.00	–
GA	60 (20.4)	16 (26.7)	68	0.57 (0.33–0.97)	0.04	21 (35.0)	61	0.58 (0.36–0.92)	0.02
AA	3 (1.0)	1 (33.3)	67	0.78 (0.10–6.19)	0.82	1 (33.3)	67	0.54 (0.07–4.12)	0.55
GG	231 (78.6)	87 (37.7)	51	1.00	–	115 (48.8)	41	1.00	–
GA + AA	63 (21.4)	17 (27.0)	68	0.61 (0.36–1.03)	0.06	22 (34.9)	61	0.59 (0.37–0.94)	0.03
MGMT rs12917††
CC	244 (79.5)	84 (34.4)	57	1.00	–	113 (46.3)	47	1.00	–
CT	56 (18.2)	24 (42.9)	43	1.63 (1.02–2.58)	0.04	29 (51.8)	35	1.38 (0.91–2.09)	0.13
TT	7 (2.3)	3 (42.9)	30	1.75 (0.54–5.88)	0.35	3 (42.9)	51	1.41 (0.44–4.59)	0.59
CC	244 (79.5)	84 (34.4)	57	1.00	–	113 (46.3)	47	1.00	–
CT + TT	63 (20.5)	27 (42.9)	42	1.67 (1.07–2.61)	0.02	32 (50.8)	35	1.39 (0.93–2.08)	0.11
XPC rs2228000††
CC	157 (53.0)	63 (40.1)	49	1.00	–	79 (50.3)	42	1.00	–
CT	119 (40.2)	37 (31.1)	60	0.62 (0.41–0.93)	0.02	50 (42.0)	50	0.68 (0.48–0.96)	0.03
TT	20 (6.8)	5 (25.0)	60	0.71 (0.28–1.79)	0.47	9 (45.0)	43	1.02 (0.51–2.04)	0.97
CC	157 (53.0)	63 (40.1)	49	1.00	–	79 (50.3)	42	1.00	–
CT + TT	139 (47.0)	42 (30.2)	61	0.66 (0.44–0.99)	0.04	59 (42.5)	49	0.74 (0.53–1.05)	0.09
XPD rs1799793††
GG	262 (86.5)	99 (37.8)	51	1.00	–	128 (48.9)	42	1.00	–
GA	38 (12.5)	10 (26.3)	76	0.52 (0.26–1.05)	0.07	15 (38.5)	61	0.64 (0.38–1.14)	0.13
AA	3 (1.0)	2 (66.7)	0	5.79 (1.33–25.26)	0.02	2 (66.7)	0	3.15 (0.74–13.38)	0.12
GG + GA	300 (99.0)	153 (39.5)	54	1.00	–	143 (47.7)	44	1.00	–
AA	3 (1.0)	2 (66.7)	0	5.95 (1.36–26.00)	0.02	2 (66.7)	0	3.10 (0.73–13.19)	0.13
MSH2 rs3732183††
AA	110 (36.0)	45 (40.8)	47	1.00	–	56 (50.9)	44	1.00	–
AG	142 (46.4)	45 (31.7)	57	0.64 (0.42–0.96)	0.03	61 (43.0)	47	0.73 (0.51–1.05)	0.08
GG	54 (17.6)	22 (40.7)	50	0.75 (0.45–1.29)	0.28	29 (53.7)	35	0.96 (0.81–1.51)	0.85
AA	110 (36.0)	45 (40.8)	47	1.00	–	56 (50.9)	44	1.00	–
AG + GG	196 (64.0)	67 (34.2)	55	0.68 (0.48–0.99)	0.05	90 (45.9)	44	0.79 (0.59–1.11)	0.18

†Column percentage. ‡Row percentage. §Five‐year survival rate (5YSR): proportion of survival derived from Kaplan–Meier analysis. ¶Hazard ratio (HR), 95% confidence interval (CI) and their corresponding P‐values were calculated using multivariate Cox proportional hazard models, adjusted for age, gender, smoking status, tumor histology, pathological stage and adjuvant therapy. ††Genotype failure: four cases for rs1136410; 16 cases for rs25489; three cases for 12917; 14 cases for 2228000, seven cases for rs1799793; and six cases for rs3732183.

Combined effect of DNA repair gene SNP on OS and DFS.  Because the rs1136410 TT or TC, rs25489 GG, rs12917 CT or TT, rs2228000 CC, rs1799793 AA and rs3732183 AA genotypes were associated with worse survival outcomes, we considered these six genotypes as bad genotypes and then evaluated their combined effects by grouping the patients based on the number of bad genotypes. As shown in Table 4 and Figure 1, OS and DFS were decreased as the number of bad genotypes increased (P _trend < 0.0001 for OS and DFS). Based on multivariate analysis, patients carrying three, and four or five bad genotypes had significantly worse OS and DFS compared with those carrying zero or one bad genotype (aHR for OS = 3.53, 95% CI = 1.25–9.97, P = 0.02, and aHR for DFS = 3.31, 95% CI = 1.41–7.76, P = 0.006; and aHR for OS = 5.47, 95% CI = 1.87–16.00, P = 0.002, and aHR for DFS = 4.42, 95% CI = 1.82–10.74, P = 0.001, respectively).

Table 4.

 Combined effects of the six polymorphisms in the DNA repair genes on overall survival and disease‐free survival

No. bad genotypes	No. cases (%)†	Overall survival				Disease‐free survival
		5YSR (%)‡	Log‐rank P	HR (95% CI)§	P§	5YSR (%)‡	Log‐rank P	HR (95% CI)§	P§
0 or 1¶	28 (10.0)	81	0.001	1	–	73	0.001	1	–
2	94 (33.4)	62	–	1.99 (0.69–5.78)	0.21	50	–	2.11 (0.88–5.03)	0.09
3	111 (39.5)	50	–	3.53 (1.25–9.97)	0.02	41	–	3.31 (1.41–7.76)	0.006
4 or 5	48 (17.1)	38	–	5.47 (1.87–16.00)	0.002	30	–	4.42 (1.82–10.74)	0.001
P trend	–	–	–	<0.0001			–	<0.0001	–

†Twenty‐nine cases, in which genotyping failed for any of the six single nucleotide polymorphisms, were excluded from this analysis. The number in parenthesis is the column percentage. ‡Five‐year survival rate (5YSR): proportion of survival derived from Kaplan–Meier analysis. §Hazard ratio (HR), 95% confidence interval (CI) and their corresponding P‐values were calculated using multivariate Cox proportional hazard models, adjusted for age, gender, smoking status, tumor histology, pathological stage and adjuvant therapy. ¶The number of patients with zero bad genotypes was six.

Figure 1.

 Kaplan–Meier survival estimates of overall survival (a) and disease‐free survival (b) according to the number of bad genotypes of the six single nucleotide polymorphisms (rs1136410, rs25489, rs12917, rs2228000, rs1799793 and rs3732183). P‐values are from the log‐rank test.

Discussion

Given that the prognosis of cancer patients likely involves multistep, multigenic pathways, it is unlikely that any single polymorphism would have a dramatic effect on survival outcome, and it is important to undertake a pathway‐based analysis that assesses the combined effects of polymorphisms that interact in the same pathway. Therefore, we used a multigenic approach to evaluate the associations of a panel of SNP in DNA repair genes with the survival outcome of patients with surgically‐resected NSCLC. In the present study, six SNP (rs1136410, rs25489, rs12917, rs2228000, rs1799793 and rs3732183) and their combined genotypes were associated with the survival outcome of patients. These findings suggest that the SNP in DNA repair genes, particularly their combined genotypes, can be used as prognostic factors for patients with surgically resected NSCLC.

The most significant finding of this study was that combined analysis of multiple SNP in the same or relevant pathways had a much higher potential value in terms of unraveling polymorphisms of prognostic importance. Although six SNP were associated with survival outcome in individual SNP analysis, considering the borderline CI and multiple comparisons, the impact of any individual SNP on survival outcomes would be minimal. However, when these six SNP were combined, OS and DFS decreased as the number of bad genotypes increased (P _trend < 0.0001 for OS and DFS). Matullo et al. ^{( 13 )} reported a dose‐dependent relationship between the number of adverse alleles in DNA repair gene SNP and the level of DNA adducts in peripheral blood cells, suggesting a stepwise decrease in DNA repair capacity (DRC) as the number of adverse alleles increases, and providing biological plausibility and validity to our approach.

This study suggests that DNA repair gene polymorphisms may influence malignant phenotypes of NSCLC. There is a growing realization that genetic polymorphisms strongly influence not only the development of cancer, but also cancer progression and prognosis.^{( 14 )} DNA repair gene polymorphisms have been shown to be associated with DRC, and genetic and epi‐genetic alterations lead to clonal expansion of abnormal cells through the acquisition of a proliferative advantage.^{( 13} , ¹⁵ , ¹⁶ , ^{17 )} Therefore, it is possible that functional polymorphisms in DNA repair genes may affect the natural history of lung cancer, such as tumor histopathology including the stage or grade of disease, the rate of disease progression, or the propensity for metastasis, thereby influencing the survival outcome.

In the present study, six SNP were associated with survival outcome. The XPC rs2228000C>T (A499V) SNP exhibited a better survival outcome under a dominant model for the variant T allele. The XPC protein plays a key role in nucleotide excision repair (NER) that removes bulky DNA adducts that are typically generated by polycyclic aromatic hydrocarbons in tobacco smoke. The XPC protein binds with HR23B to form a complex that recognizes damaged DNA and initiates NER through the recruitment of other components of the repair apparatus, such as XPA, transcription factor IIH (TFIIH) and XPG. Because the A499V SNP is located in the region of the HR23B interaction domain of XPC, the Ala‐to‐Val change may affect the XPC‐HR23B interaction, thereby influencing NER capacity.^{( 18 )} Consistent with our finding, it has been reported that subjects with the rs2228000 CT or TT genotype had a higher DRC compared with those with the rs2228000 CC genotype.^{( 19} , ^{20 )} For the XPD rs1799793G>A SNP, patients with homozygous variant A alleles exhibited worse survival outcome compared with those carrying the GG or GA genotype. The XPD protein is an ATP‐dependent helicase, a subunit of TFIIH that is essential for NER. Because the rs1799793G>A [D312N] SNP is located in the domain of interaction between the XPD protein and its helicase activator, p44 protein, inside the transcription factor IIH complex, it has been suggested that this SNP may thwart XPD‐p44 protein interaction, thereby leading to reduced NER capacity.^{( 21 )} In agreement with this suggestion, the rs1799793 AA genotype has been reported to be associated with reduced DRC^{( 15 )} and an increased risk of lung cancer in a meta‐analysis.^{( 22 )} Therefore, our finding of an association of the XPD rs1799793 AA genotype with a worse survival outcome is biologically plausible. In the present study, MGMT rs12917C>T was associated with a worse survival outcome under a dominant model of the variant T allele. MGMT removes mutagenic O ⁶ ‐alkylguanine DNA adducts originating from exposure to alkylating agents, such as nitrosamines present in tobacco smoke.^{( 23 )} It has been suggested that the rs12917C>T [L115F] SNP may reduce the stability of the MGMT protein, thereby reducing MGMT‐mediated repair of DNA alkyl adducts.^{( 23} , ^{24 )} Hill et al. ^{( 25 )} reported that individuals with the rs12917 CT or TT genotype had a significantly higher level of tobacco‐specific nitrosamine‐induced chromosome aberrations compared with those with the CC genotype. Therefore, our finding of an association of the rs12917 CT or TT genotype with a worse survival outcome is also biologically plausible. In addition, it is conceivable that the MGMT rs12917C>T SNP may play an important role in determining the prognosis of lung cancer in ever‐smokers. For the MSH2 rs3732183A>G (IVS10+12A>G) SNP, we found that the AG or GG genotype was associated with a better survival outcome compared with the AA genotype. The hMSH2 gene is one of the mismatch repair genes. It has been reported that multiple alternatively spliced isoforms of hMSH2 mRNA are present in normal human tissues. Some of the isoforms would produce truncated proteins that lack exon 2–8 or exon 13, and the others are in‐frame deletions lacking exon 5 or exon 2–7. The relative level of expression of each splicing variant has shown considerable interindividual variation.^{( 26 )} Although the functional relevance of this SNP is still unknown, this splicing site polymorphism may influence the alternative splicing of hMSH2, resulting in interindividual variation in the levels of expression of the hMSH2 splicing variants.^{( 27 )} Our finding is in agreement with the results of previous studies, in which this SNP was related to a significantly decreased risk of lung cancer and a better survival outcome in patients with colorectal cancer in a dominant model for the variant G allele.^{( 27} , ^{28 )}

PARP1 plays a critical role in base excision repair (BER), which is a major DNA repair pathway for small base adducts produced by oxidation, methylation and environmental carcinogens.^{( 29 )} PARP1 acts as a sensor of DNA streak breaks during BER. In addition to its role as a nick sensor, PARP1 also interacts with the scaffold protein, XRCC1, and may accelerate the recruitment of DNA repair proteins to strand interruption. The PARP1 rs1136410T>C [V762A] SNP, located within the carboxy‐terminal catalytic domain, has been reported to influence PARP activity, thereby affecting BER capacity.^{( 30 )} A Chinese study group reported that the rs1136410T>C SNP was associated with a significantly increased risk of esophageal SCC and lung cancer.^{( 31} , ^{32 )} Lockett et al. ^{( 30 )} also reported that this SNP was associated with a significantly increased risk of prostate cancer in American Caucasians. However, in contrast to these studies, Li et al. ^{( 33 )} reported that this SNP was associated with a significantly decreased risk of SCC of the head and neck in American Caucasians. In addition, the rs1136410T>C SNP has been reported to be associated with a non‐significantly reduced risk of melanoma and breast cancer in American Caucasians and Chinese women, respectively.^{( 34} , ^{35 )} Our finding that the rs1136410T>C SNP was associated with better survival outcome is in agreement with some reports,^{( 33} , ³⁴ , ^{35 )} but in disagreement with others.^{( 30} , ³¹ , ^{32 )} XRCC1 acts as a facilitator or coordinator in BER through its interaction with PARP1, DNA polymerase β and DNA ligase III. Provided that the rs25489G>A [R280H] SNP is located in the vicinity of the PARP1 interaction domain of XRCC1, this SNP may alter XRCC1‐PARP1 interaction and thereby influence BER capacity.^{( 36 )} Discrepancy exists with respect to the association between the rs25489G>A SNP and lung cancer; indeed, both an increased and decreased risk have been reported.^{( 37 )} In the present study, the XRCC1 rs25489G>A SNP was associated with a better survival outcome in the dominant model for the variant A allele. Although it is difficult to decipher the mechanism(s) for the discrepant finding from different studies, including an oppositely directed (flip‐flop) association, the different genetic backgrounds in the study populations or interactive effects of multiple polymorphisms might contribute to the discrepancy.^{( 38 )} Another possible explanation for the discrepancy between different studies might be that different cell types of cancer have different etiologies and different carcinogenesis pathways.

In the present study, six SNP in DNA repair genes were found to be associated with the survival outcome of patients with early stage NSCLC, and the combined genotypes of these six SNP to be independent prognostic markers for those patients. Consequently, in addition to the pathological stage, testing for the presence of these six SNP may help identify patient subgroups at high risk of a poor disease outcome, thereby helping to refine therapeutic decisions in the treatment of NSCLC. However, larger studies are required to confirm our findings in other ethnic populations. In addition, further studies are needed to determine the effects of these SNP on the survival outcome of other types of cancer because different cell types of cancer have different etiologies and different carcinogenesis pathways.

Disclosure Statement

The authors have no conflict of interest.