Functional Variants of the NEIL1 and NEIL2 Genes and Risk and Progression of Squamous Cell Carcinoma of the Oral Cavity and Oropharynx

Abstract

Purpose

Human DNA glycosylases NEIL1 and NEIL2 participate in oxidized base excision repair and protect cells from DNA damage. NEIL1 (MIM:608844) and NEIL2 (MIM:608933) variants may affect their protein functions, leading to altered cell-death and carcinogenesis. To date, only one reported study investigated the association between NEIL1 and NEIL2 polymorphisms and cancer risk.

Experimental Design

Genotype and haplotypes of the NEIL1 NT_010194.16:g.46434077G>T (rs7182283) and g.46438282C>G (rs4462560) and NEIL2 NT_077531.3:g.4102971C>G (rs804270) polymorphisms were determined for 872 patients with newly diagnosed SCCOOP and 1,044 cancer-free non-Hispanic white control subjects frequency matched by age and sex. Crude and adjusted odds ratios (OR) and 95% confidence intervals (95% CI) were calculated using multivariate logistic regression, and false-positive report probabilities were also calculated.

Results

We found no overall differences in the frequencies of alleles, genotypes, and haplotypes of NEIL1 g.46434077G>T and NEIL1 g.46438282C>G polymorphisms between cases and controls. However, the NEIL2 g.4102971CC genotype was associated with a significantly increased risk of SCCOOP (adjusted OR, 1.30; 95% CI, 1.02–1.65); this increase in risk was the highest among current alcohol drinkers (adjusted OR, 1.87; 95% CI, 1.28–2.72), particularly in patients with oropharyngeal cancer (adjusted OR, 1.35; 95% CI, 1.04–1.76). The NEIL2 g.4102971CC genotype also was significantly associated with SCCOOP of advanced stages.

Conclusions

Polymorphisms of the NEIL2 gene may be markers for risk and progression of SCCOOP, particularly in patients with oropharyngeal cancer. Larger studies are needed to confirm our findings.

Introduction

Head and neck cancers are the sixth most common cancers, with a prevalence of more than 1.6 million cases worldwide (1). They account for approximately 2.8% of all malignancies in the United States (1, 2). Despite slight decrease in its incidence rate (~ 4% since 1980) and a modest improvement in 5-year survival (from 54.4% to 59.4% over the last 20 years), these cancers continue to be a clinical challenge (1, 2).

The squamous cell carcinomas of the oral cavity and oropharynx (SCCOOP) are an etiologically similar subgroup of head and neck cancers, and it is estimated that 34,360 new cases of SCCOOP and 7,550 deaths from it will have occurred in the United States alone in 2007 (2, 3). Even with the use of modern therapeutic options such as surgical management, radiation treatment, and chemotherapeutic intervention, about 50% of all patients will ultimately die of this disease, especially for those patients who were diagnosed with advanced or relapsed disease, which is almost uniformly fatal (2).

The importance of an individual’s lifestyle, particularly the use of tobacco and alcohol, has been well recognized in the etiology of SCCOOP. However, despite these risk factors, relatively few people actually develop these diseases; likewise, some individuals develop SCCOOP in the absence of such habits or other identifiable lifestyle and environmental factors. In this case, genetic susceptibility may play a role in the initiation and development of SCCOOP. For instance, genetically determined DNA repair capacity may contribute to the variation in susceptibility to head and neck cancers (4–6). Identifying genetic factors that modulate the risk of SCCOOP thus likely will help determine at-risk subgroups that can benefit from primary prevention programs.

DNA repair systems play an important role in the maintenance of genomic integrity and stability in response to environmental exposure, replication errors, and cumulative effects of aging. In humans, more than 100 genes are involved in the five major DNA repair pathways: direct reversal, basal excision repair (BER), nucleotide excision repair (NER), mismatch repair, and recombination repair (7). NER targets bulky, helix-distorting adducts, such as benzo(a)pyrene-guanine adduct, whereas BER removes smaller altered bases produced by oxidation, methylation, radiation, and other minor base modifications (8). In a series of association studies, we found that variants in genes involved in both BER and NER contributed to an individual’s susceptibility to head and neck cancers (9–13).

In mammalian cells, the repair of DNA bases that have been damaged by reactive oxygen species (ROS) is initiated primarily by a series of DNA glycosylases, including two that belong to a class of DNA glycosylases homologous to the bacterial Fpg/Nei family, NEIL1 (nei endonuclease VIII–like 1) and NEIL2 (nei-like 2). These glycosylases initiate the first step in BER by cleaving bases damaged by ROS and introducing a DNA strand break via the associated lyase reaction (14–17). NEIL1 and NEIL2 protect cells from radiation-mediated cell death, and functional variants of NEIL1 and NEIL2 may affect the protein functions, leading to altered cell-death probability and carcinogenesis potential.

The human NEIL1 gene is located on chromosome 15q23, and mutations in NEIL1 have been shown to cause increased risk of developing primary gastric cancer in a Japanese cohort (18). NEIL1 has been resequenced by the National Institute of Environmental Health Sciences Environmental (NIEHS) Genome Project (EGP)¹, and at least 62 single nucleotide polymorphisms (SNPs)² have been reported; four of those are nonsynonymous SNPs (nsSNPs): p.Ser82Cys, p.Gly83Asp, p.Cys136Arg, and p.Asp252Asn, with very low minor allele frequencies (all = 0.01)³. Two of those polymorphic variants, p.Ser82Cys and p.Asp252Asn, showed near wild-type enzyme specificity and kinetics, whereas p.Gly83Asp was devoid of glycosylase activity and p.Cys136Arg may be glycosylase deficient (19). NEIL2 is located on chromosome 8p23.1, and at least 250 SNPs in this gene are reported in the dbSNP database⁴. Four of these SNPs were confirmed as non-synonymous (nsSNPs) (p.Thr70Ser, p.Arg103Gln, p.Arg257Leu, and p.Pro304Thr) in the NIEHS EGP SNP database⁵, and an additional one, P123T, was reported recently (20). All these nsSNPs have been found rare in both healthy control participants and colon cancer patients (20).

To date, no reports have been published about the possible association between variants of NEIL1 and NEIL2 genes and the risk of head and neck cancers. Because of the role of the NEIL1 and NEIL2 genes in regulating DNA repair and because decreased DNA repair capacity has been associated with increased risk of head and neck cancers, we hypothesized that NEIL1 and NEIL2 polymorphisms contribute to genetic susceptibility to SCCOOP. To test this hypothesis, we conducted a hospital-based case-control study to identify associations between SCCOOP and newly reported common (i.e., with a minor allele frequency of ≥ 0.05) functional SNPs of the NEIL1 gene, NT_010194.16:g.46434077 (rs7182283) and NT_010194.16:g.46438282 (rs4462560), and the NEIL2 gene, g.4102971 (rs804270).

Materials and Methods

Study subjects and data collection

The recruitment of our study participants has been previously described (21, 22). Briefly, the study population consisted of 872 patients with SCCOOP and 1044 cancer-free control participants recruited from May 1, 1995, through September 30, 2006. All patients had newly diagnosed, untreated SCCOOP that was histologically confirmed at The University of Texas M. D. Anderson Cancer Center. Twenty four percent of all SCCOOP patients were eligible, and 90% of the eligible patients contacted chose to participate in this study. Only non-Hispanic white patients were included in the final analysis because genotype frequencies can vary between ethnic groups and relatively few minority patients were seen at M. D. Anderson. Among the 872 SCCOOP patients with primary tumors included in the analysis, 299 (34.3%) had cancers of the oral cavity and 573 (65.7%) cancers of the oropharynx. Patients with second oral cavity and oropharyngeal tumors; primary tumors of the hypopharynx, nasopharynx, or sinonasal tract; primary tumors outside the upper aerodigestive tract; cervical metastases of unknown origin; or histopathologic diagnoses other than SCCOOP were excluded from the analysis.

The regional lymph node involvement of SCCOOP was defined as N₀ to N₃ as follows (23): N₀, no regional node metastasis; N₁, metastasis in a single ipsilateral lymph node, ≤3 cm in the greatest dimension; N₂, metastasis in a single ipsilateral lymph node, >3 cm but <6 cm in the greatest dimension; or in multiple ipsilateral lymph nodes, none ≥6 cm in the greatest dimension; or in any bilateral or contralateral lymph node, <6 cm in the greatest dimension; N₃, metastasis in any lymph node, ≥6 cm in the greatest dimension. The extent of the primary SCCOOP was defined as T₁ to T₄ as follows: T₁, tumor ≤2 cm at the greatest dimension; T₂, tumor >2 cm but <4 cm in the greatest dimension; T₃, tumor ≥4 cm in the greatest dimension; T₄, tumor invading adjacent structures.

Cancer-free control participants were recruited from persons who were not hospital patients or seeking health care but who had accompanied the case patients visiting the clinics; controls could not be genetically related to any cases or controls already selected. We first surveyed potential control participants at the clinics by using a short questionnaire to determine their willingness to participate in our research studies and to obtain information about demographic features, smoking and alcohol drinking status (current, former, or never), and personal history of cancer. Our control participants were required to have no previous history of cancer and were not under medical care or receiving treatment for any known disease. They were frequency matched to the case patients by age (± 5 years) and sex. Among the visitors to our institution who were screened for possible participation as controls, 73% were eligible for and were therefore offered participation. Of these eligible potential controls, 85% agreed to and ultimately did participate. We interviewed each enrolled control subject to obtain data about their personal history of exposure to known etiologic factors in SCCOOP, such as tobacco smoking and alcohol use. After signing the informed consent form, each participant donated 30 mL of blood, of which 1 mL was used for genomic DNA extraction. The protocol for this research study was approved by the M. D. Anderson Cancer Center Institutional Review Board.

Genotyping

We extracted genomic DNA from the buffy-coat fraction of the blood samples by using a blood DNA mini kit (Qiagen Inc., Valencia, CA) according to the manufacturer’s instructions. The selected NEIL1 and NEIL2 polymorphisms were determined using the polymerase chain reaction (PCR)–restriction fragment length polymorphism method, as previously described (21). We performed the PCRs with a PTC-200 DNA engine (Peltier thermal cycler, M J Research, Inc., Waltham, MA) in 10 μL of PCR mixture. This PCR mixture included approximately 20 ng of genomic DNA, 0.1 mM deoxynucleotide triphosphate, and 1× PCR buffer (50 mM KCl, 10 mM Tris HCl, and 0.1% Triton X-100); 1.5 mM MgCl₂, 0.5 units of Taq polymerase (Denville Scientific Inc.; Metuchen, NJ); and 2 pmol of each primer. The following primers were used to amplify the target fragments containing these four polymorphisms (mismatch bases are underlined): 5′-CAACCTCCTGATTAACTGGAACCACA -3′ (forward) and 5′-TCACTTCAGCCCAGGAGACCAG -3′ (reverse) for NEIL1 NT_010194.16:g.46434077G>T (rs7182283); 5′-GTCTCTTCACTGGCTTTTGGGG -3′ (forward) and 5′-TCCCAGGTATTTGGTGGGTAGG -3′ (reverse) for NEIL1 NT_010194.16:g.46438282C>G (rs4462560); and 5′-ACCCCCCACCTCGGGCACTCGG -3′ (forward) and 5′-AAGATGCCGCGCCCACCCGC -3′ (reverse) for NEIL2 NT_077531.3:g.4102971C>G (rs804270). There were 119-bp, 118-bp, and 131-bp amplified PCR products for the NEIL1 g.46434077, NEIL1 g.46438282, and NEIL2 g.4102971 polymorphisms, respectively. The NlaIII, HaeIII, and SacII restriction enzymes (New England Biolabs, Beverly, MA) were used to distinguish the NEIL1 g.46434077, NEIL1 g.46438282, and NEIL2 g.4102971 polymorphisms, respectively, which resulted in 91-bp and 28-bp fragments in the presence of the NEIL1 g.46434077T allele; 96-bp and 22-bp fragments in the presence of the NEIL1 g.46438282C allele; and 112-bp and 19-bp fragments in the presence of the NEIL2 g.4102971C allele. More than 10% of the samples were randomly selected for confirmation, and the results were 100% concordant.

Statistical analysis

The χ² test was used to evaluate differences between cases and controls in the frequency distributions of selected demographic variables, smoking status, alcohol use, and each allele and genotype of the three NEIL1 and NEIL2 polymorphisms. Unconditional univariate and multivariate logistic regression analyses were performed to obtain the crude and adjusted odds ratios (ORs) and their 95% confidence intervals (CIs) for cancer risk, assuming a genetic recessive model. Multivariate adjustment was conditional on the effect of age, sex, smoking status, and alcohol use. Subjects who had smoked > 100 cigarettes in their lifetimes were categorized as ever smokers, and others were never smokers. Ever smokers who had quit smoking > 1 year previously (prior to disease diagnosis for the cases and before the time of questionnaire administered for the controls) were categorized as former smokers, and the other smokers were categorized as current smokers. Similarly, subjects who had drunk alcoholic beverages at least once a week for > 1 year previously were categorized as ever drinkers, and others were never drinkers. Ever drinkers who had quit drinking > 1 year previously were categorized as former drinkers, and the other drinkers were categorized as current drinkers. We further stratified the genotype data by subgroups of age, sex, smoking, alcohol drinking, and tumor site and assessed any trend in risk in multivariate logistic regression models. Hardy–Weinberg equilibrium (HWE) of the NEIL1 and NEIL2 genotypes was tested by performing a goodness-of-fit χ²-test. The linkage disequilibrium D′ value, logarithm of odds (LOD) score, and r² value were computed for two-locus models. We estimated NEIL1 haplotypes using unphased genotypes. The haplotype of the highest frequency was used as the reference group to calculate odds ratios for haplotypes associated with SCCOOP. Logistic regression was used to evaluate the association between the SCCOOP and NEIL1 haplotypes. Haplotypes were reconstructed using PHASE version 2 software (24)⁶

We used the false-positive report probability (FPRP) to test for false-positive associations (25). For all significant genetic effects observed in our study, we calculated FPRP with prior probabilities of 0.0001, 0.001, 0.01, 0.1 and 0.25. The OR was set close to the observed value obtained in our study, and a probability of < 0.2 was considered noteworthy. We analyzed our data by using SAS software (Version 8e; SAS Institute Inc., Cary, NC); P < 0.05 was considered statistically significant, and all statistical tests were two sided,.

Results

Characteristics of non-genetic risk factors for SCCOOP in the study population

The frequency distributions of selected characteristics of the cases and controls are presented in Table 1. The cases and controls were well matched for age and sex, with a mean age of 56.5 years for the cases (± 11.2 years; range, 18–85 years) and 56.4 years for the controls (± 11.0 years; range, 20–85 years) (P = 0.975); there was also no difference in the distribution of age groups (≤ 50, 51–64, and ≥ 65) nor in sex groups between the case and control groups, with men constituting 77.2% of patients and 76.8% of controls (P = 0.853). However, there were more current smokers (33.6%) and current drinkers (51.7%) among the cases than among the controls (15.3% and 40.7%, respectively), and this differences were statistically significant (P < 0.001). There was an increased risk of SCCOOP in former and current smokers (OR, 1.62; 95% CI, 1.31–2.01) and OR, 3.55; 95% CI, 2.78–4.53, respectively) and in former or current alcohol drinkers (OR, 2.06; 95% CI, 1.59–2.68, and OR, 2.01; 95% CI, 1.64–2.47, respectively) compared with non-smokers and non-drinkers, respectively (Table 1). Therefore, we made further adjustment for these covariates in the multivariate logistic regression analysis.

Table 1.

Characteristics and odds ratios of covariate variables in the study

Variable	Cases (n = 872)	Controls (n = 1,044)	Odds Ratio*	P†
	n (%)	n (%)
Mean age (± SD)‡	56.5 (±11.2)	56.4 (±11.0)		0.975
Age
≤ 50	245 (28.1)	303 (29.0)	Reference	0.898
51–64	410 (47.0)	487 (46.7)	1.04 (0.84–1.29)
≥ 65	217 (24.9)	254 (24.3)	1.06 (0.83–1.35)
Gender
Female	199 (22.8)	242 (23.2)	Reference	0.853
Male	673 (77.2)	802 (76.8)	1.02 (0.82–1.26)
Smoking status
Never	259 (29.7)	502 (48.1)	Reference	<0.001
Former	320 (36.7)	382 (36.6)	1.62 (1.31–2.01)
Current	293 (33.6)	160 (15.3)	3.55 (2.78–4.53)
Alcohol use
Never	238 (27.3)	451 (43.2)	Reference	<0.001
Former	183 (21.0)	168 (16.1)	2.06 (1.59–2.68)
Current	451 (51.7)	425 (40.7)	2.01 (1.64–2.47)
Tumor site
Oral cavity	299 (34.3)
Oropharynx	573 (65.7)

^*Odds ratio was calculated for categorical variables of age, gender, smoking, and alcohol use.

^†P value is from the χ² test for categorical variables. The P value for mean age was from the Wilcoxon rank test.

^‡Mean age (± SD) was the mean and standard deviation of variable age.

Allele and genotype distributions of the NEIL1 and NEIL2 polymorphisms among the cases and controls

As shown in Table 2, the frequencies of minor alleles of NEIL1 g.46434077T and g.46438282G and NEIL2 g.4102971C were 47.2%, 31.5% and 44.4%, respectively, for the cases and 46.4%, 31.8% and 43.0%, respectively, for the controls, but none of the differences between the cases and controls was statistically significant (P = 0.611, 0.809 and 0.359, respectively). All distributions of the frequencies of observed genotypes in the controls were consistent with those obtained from the Hardy-Weinberg equilibrium model (P > 0.05). The NEIL1 g.46434077TT genotype was more frequent in the cases (22.3%) than in the controls (20.6%) but was not associated with risk of SCCOOP compared with either the GG genotype (OR, 0.99; 95% CI, 0.76–1.29) or the combined GG+GT genotypes (OR, 1.03; 95% CI, 0.82–1.29). The NEIL1 g.46438282GG genotype was less frequent in the cases (8.9%) than in the controls (9.5%), and similarly was not associated with risk of SCCOOP compared with either the CC genotype (OR, 0.99; 95% CI, 0.71–1.39) or the combined CC+CG genotypes (OR, 0.97; 95% CI, 0.70–1.34). The NEIL2 g.4102971CC genotype was more frequent in the cases (20.7%) than in the controls (17.1%), and although this genotype was not associated with risk of SCCOOP compared with the NEIL2 g.4102971GG genotype (adjusted OR, 1.18; 95% CI, 0.90–1.56), it was associated with a significant elevation in risk when compared with the combined GG+CG genotypes (OR, 1.30; 95% CI, 1.02–1.65; P < 0.05, assuming a recessive genetic model).

Table 2.

Genotype frequencies and odds ratios of NEIL1 g.46434077 and g.46438282 and NEIL2 g.4102971 variants among SCCOOP cases and controls in a non-Hispanic white population

SNP	Cases	Controls	Crude Odds ratio	Adjusted Odds ratio*	P†
	n (%)	n (%)
NEIL1 g.46434077G>T‡
GG	243 (28.0)	291 (27.9)	Reference	Reference	0.624
GT	432 (49.7)	537 (51.5)	0.96 (0.78–1.19)	0.94 (0.76–1.17)
TT	194 (22.3)	215 (20.6)	1.08 (0.83–1.4)	0.99 (0.76–1.29)
GG+GT	675 (77.7)	828 (79.4)	Reference	Reference	0.364
TT	194 (22.3)	215 (20.6)	1.11 (0.89–1.38)	1.03 (0.82–1.29)
T allele	820 (47.2)	967 (46.4)			0.611
NEIL1 g.46438282C>G§
CC	392 (46.0)	477 (45.8)	Reference	Reference	0.902
CG	385 (45.1)	465 (44.7)	1.01 (0.83–1.22)	1.05 (0.87–1.28)
GG	76 (8.9)	99 (9.5)	0.93 (0.67–1.3)	0.99 (0.71–1.39)
CC+CG	777 (91.1)	942 (90.5)	Reference	Reference	0.656
GG	76 (8.9)	99 (9.5)	0.93 (0.68–1.27)	0.97 (0.70–1.34)
G allele	537 (31.5)	663 (31.8)			0.809
NEIL2 g.4102971G>C||
GG	273 (31.9)	322 (31.1)	Reference	Reference	0.069
CG	405 (47.4)	538 (51.9)	0.89 (0.72–1.09)	0.86 (0.69–1.06)
CC	177 (20.7)	177 (17.1)	1.18 (0.91–1.54)	1.18 (0.90–1.56)
GG+CG	678 (79.3)	860 (82.9)	Reference	Reference	0.044
CC	177 (20.7)	177 (17.1)	1.27 (1.01–1.6)	1.30 (1.02–1.65)
C allele	759 (44.4)	892 (43.0)			0.359

^*Adjusted odds ratio: adjusted for age, gender, smoking, and alcohol use.

^†P value is from the χ² test for the frequency distribution of genotypes in each SNP.

^‡NEIL1 g.46434077G>T: missing genotype information in 3 cases and 1 control.

^§NEIL1 g.46438282C>G: missing genotype information in 19 cases and 3 controls.

^||NEIL2 g.4102971G>C: missing genotype information in 17 cases and 7 controls.

Association between the combined genotypes and Haplotype of the NEIL1 gene and the risk of SCCOOP

Further analysis suggested that there was a linkage disequilibrium between the two NEIL1 polymorphisms (D′ = 1.0, r² = 0.304, and P < 0.001). Since haplotype preserves the joint linkage disequilibrium structure, we reconstructed haplotypes using the unphased genotypes based on the observed NEIL1 g.46434077 and g.46438282 genotypes to further assess their association with risk of SCCOOP. There were four haplotypes: TC, GG, GC and TG; haplotype TC was the most common (occurring in 43.02% of the cases and 43.44% of the controls) and haplotype TG was the least frequent, with a frequency < 5% in both cases and controls. The frequency distributions of these four haplotypes were comparable between cases and controls (χ₂ test: P = 0.87). Therefore, neither the genotypes nor haplotypes of NEIL1 g.46434077 and NEIL1 g.46438282 were associated with risk of SCCOOP (data not sown).

Stratified analysis of the associations between NEIL1 and NEIL2 polymorphisms and risk of SCCOOP

We further stratified the data by age, sex, smoking status, alcohol drinking status, and cancer site. To facilitate the analysis, we evaluated the risk of SCCOOP by estimating the ORs associated with the combined NEIL1 g.46434077 (GG+GT), NEIL1 g.46438282 (CC+CG), and NEIL2 g.4102971 (GG+CG) genotypes compared with their homozygous genotypes (NEIL1 g.46434077TT, NEIL1 g.46438282GG, and NEIL2 g.4102971CC, respectively), with adjustment for the aforementioned variables. As shown in Table 3, although none of the genotypes of the NIEL1 gene was associated with cancer risk in all subgroups examined, the OR was virtually always elevated (except for former alcohol drinkers), sometimes significantly so, for the CC genotype of NEIL2 g.4102971 compared with the combined GG+CG genotypes. In former smokers and current drinkers, the OR the CC genotype of NEIL2 g.4102971 were 1.51 (95% CI, 1.03–2.23; P = 0.082) and 1.87 (95% CI, 1.28–2.72; P = 0.003), respectively, compared with either the GG or CG genotype, and this risk was more evident for oropharynx (OR, 1.35; 95% CI, 1.04–1.76; P = 0.02) than for oral cavity (OR, 1.19; 95% CI, 0.83–1.69; P = 0.545), although this difference was not statistically significant (data not shown).

Table 3.

Odds ratios of SCCOOP associated with genotypes in the stratified analysis

Stratified variables	n (case/control)*	NEIL1 g.46434077G>T		P‡	n (case/control)§	NEIL1 g.46438282C>G		P‡	n (case/control)||	NEIL2 g.4102971G>C		P‡
		Percentage (case/control)	Adjusted OR (95% CI)†			Percentage (case/control)	Adjusted OR (95% CI)†			Percentage (case/control)	Adjusted OR (95% CI)†
		TT	TT vs. GG + GT			GG	GG vs. CC + CG			CC	CC vs. CG + GG
Age
≤ 50	243/303	21.0/20.5	0.98 (0.64–1.50)	0.88	237/303	10.5/7.6	1.47 (0.80–2.68)	0.231	236/301	19.9/16.9	1.25 (0.80–1.95)	0.376
51 to 64	409/487	24.0/20.5	1.06 (0.75–1.48)	0.218	404/485	9.7/11.1	0.89 (0.56–1.42)	0.473	404/484	20.8/18.2	1.27 (0.90–1.81)	0.327
≥ 65	217/253	20.7/20.9	0.92 (0.58–1.46)	0.955	212/253	5.7/8.7	0.63 (0.30–1.32)	0.211	215/252	21.4/15.1	1.40 (0.85–2.28)	0.077
Gender
Male	671/801	23.4/20.5	1.11 (0.86–1.43)	0.176	662/800	8.6/9.1	1.00 (0.69–1.45)	0.731	659/797	21.1/17.8	1.27 (0.97–1.66)	0.115
Female	198/242	18.7/21.1	0.76 (0.46–1.25)	0.533	191/241	9.9/10.8	0.86 (0.44–1.66)	0.776	196/240	19.4/14.6	1.49 (0.87–2.54)	0.181
Smoking status
Never	258/502	18.6/18.7	0.98 (0.67–1.44)	0.968	252/502	7.5/10.6	0.68 (0.39–1.18)	0.184	251/499	19.5/17.6	1.13 (0.76–1.66)	0.528
Former	320/381	20.9/22.6	0.91 (0.63–1.31)	0.602	315/379	10.5/8.4	1.35 (0.81–2.27)	0.36	317/380	21.8/16.6	1.51 (1.03–2.23)	0.082
Current	291/160	27.1/21.9	1.29 (0.80–2.06)	0.218	286/160	8.4/8.8	0.96 (0.46–1.97)	0.897	287/158	20.6/16.5	1.41 (0.83–2.39)	0.292
Alcohol use
Never	237/451	21.5/20.6	1.05 (0.72–1.55)	0.783	234/450	9.8/11.1	0.87 (0.52–1.47)	0.606	235/449	20.9/19.6	1.10 (0.74–1.63)	0.698
Former	183/168	19.7/16.1	1.26 (0.72–2.20)	0.38	179/168	5.6/10.1	0.51 (0.23–1.16)	0.115	181/166	18.8/18.7	0.97 (0.56–1.69)	0.979
Current	449/424	23.8/22.4	0.98 (0.70–1.37)	0.618	440/423	9.8/7.6	1.43 (0.87–2.38)	0.25	439/422	21.4/13.7	1.87 (1.28–2.72)	0.003
Tumor site
Oral cavity	297/1043	22.6/20.6	1.04 (0.75–1.44)	0.468	294/1041	7.8/9.5	0.83 (0.50–1.36)	0.375	296/1037	18.6/17.1	1.19 (0.83–1.69)	0.545
Oropharynx	572/1043	22.2/20.6	1.03 (0.80–1.33)	0.455	559/1041	9.5/9.5	1.04 (0.73–1.49)	0.985	559/1037	21.8/17.1	1.35 (1.04–1.76)	0.020

The result is bolded if the 95% confidence interval does not include 1 or P < 0.05.

^*No. (cases/controls): 3 cases and 1 control had missing values for NEIL1 g.46434077G>T.

^†Adjusted OR, odds ratios were adjusted for all covariates (age, gender, smoking status, and alcohol use), excluding the stratified variable.

^‡P value was from the χ² test statistics from comparisons of genotype frequency in cases and controls.

^§No. (cases/controls): 19 cases and 3 controls had missing values for NEIL1 g.46438282C>G.

^||No. (cases/controls) 17 cases and 7 controls had missing values for NEIL2 g.4102971G>C.

Association between the NEIL2 g.4102971 polymorphisms and progression of SCCOOP

Because only the NEIL2 g.4102971 polymorphism was associated with the risk of SCCOOP, we further evaluated its association with progression of SCCOOP. We found a statistically significant association between the NEIL2 g.4102971CC genotype and advanced stages of primary SCCOOP tumor (T) as well as regional lymph node metastasis at diagnosis (N). The genotype frequency distributions were statistically significantly different between cases and controls for tumor stages T₂ and T₃ and for node stages N₂ and N₃. Using the 1044 controls as the reference group, the adjusted ORs for the CC genotype of NEIL2 g.4102971 were calculated for each stratum by T or N stage. ORs for the CC genotype were significantly increased as the stages increased with the highest risk for T₃ and N₃ stage after adjustment for other covariates. For example, the OR for the genotype CC of NEIL2 g.4102971 in SCCOOP patients with N₃ was 3.06 (95% CI, 1.49–6.26). When we grouped SCCOOP patients into two groups according to their T and N status (i.e., a group with less-advanced disease [N₀ or N₁ and T₁ or T₂] and a group with more-advanced disease [N2, N3, T3, or T4], the OR for the genotype CC vs. other genotypes was 1.39 (95% CI, 1.07–1.81) in SCCOOP patients with more-advanced disease, compared with those patients with less-advanced disease, after adjusting for other covariates (Table 4).

Table 4.

Associations between the combined genotypes of the NEIL2 g.4102971 polymorphism and progression of SCCOOP^*

Stratified variables*	NEIL2 g.4102971G>C			Crude OR	Adjusted OR †	P‡
	n (case/control)	CC	GG+CG
Primary tumor (T)
T1	216/1037	16.2/17.1	83.8/82.9	0.94 (0.63–1.40)	0.98 (0.65–1.46)	0.758
T2	315/1037	22.5/17.1	77.5/82.9	1.41 (1.04–1.93)	1.48 (1.07–2.03)	0.028
T3	159/1037	24.5/17.1	75.5/82.9	1.58 (1.06–2.35)	1.75 (1.16–2.64)	0.023
T4	165/1037	19.4/17.1	80.6/82.9	1.17 (0.77–1.78)	1.19 (0.76–1.87)	0.464
Lymph node metastasis
N0	277/1037	19.5/17.1	80.5/82.9	1.18 (0.84–1.65)	1.24 (0.87–1.77)	0.346
N1	127/1037	15.7/17.1	84.3/82.9	0.91 (0.55–1.50)	1.01 (0.60–1.69)	0.708
N2	413/1037	21.8/17.1	78.2/82.9	1.35 (1.02–1.80)	1.40 (1.05–1.88)	0.036
N3	38/1037	34.2/17.1	65.8/82.9	2.53 (1.27–5.03)	3.06 (1.49–6.26)	0.007
Combine N and T
N0,1 and T1,2	260/1037	18.8/17.1	81.2/82.9	1.13 (0.79–1.60)	1.19 (0.83–1.72)	0.500
N2,3 or T3,4	595/1037	21.5/17.1	78.5/82.9	1.33 (1.03–1.72)	1.39 (1.07–1.81)	0.027

Note: The result is bolded if the 95% confidence interval does not include 1 or P < 0.05.

^*Stratified variables: T: The extent of the primary SCCOOP. T₁: tumor 2 cm at the greatest dimension; T₂ to T₄: increasing greatest dimensions. N: regional lymph node involvement. N₀: no regional lymph nodes involved; N₁ to N₃: increasing involvement of regional lymph nodes.

^†Adjusted OR: odds ratios were adjusted for age, gender, smoking status, and alcohol use.

^‡P value from the χ² test of different frequencies of NEIL2 g.4102971 genotypes in cases and controls.

Because most of the significant findings were in the subgroup analysis, we calculated the FPRP value for all the significant associations. As shown in Table 5, when the assumption of prior probability was 0.1, the NEIL2 g.4102971CC genotype was still associated with an elevated risk of SCCOOP in current alcohol drinkers and in those with tumor stage N₃, with FPRP = 0.05 and 0.1, respectively. However, if the prior probability was set to 0.01, all significant associations we found in this study became false-positive findings.

Table 5.

FPRP for significant associations of NEIL2 g.4102971 and SCCOOP

Subgroup	Positive OR95%CI*	P†	OR‡	Statistical power§	Prior probability
					0.25	0.1	0.01	0.001	0.0001
All participants	1.27 (1.01–1.60)	0.044	1.2	0.69	0.16	0.36	0.86	0.98	1.00
Current alcohol users	1.71 (1.19–2.45)	0.003	1.7	0.51	0.02	0.05	0.38	0.86	0.98
Oropharynx cancers	1.36 (1.05–1.76)	0.02	1.3	0.63	0.09	0.22	0.76	0.97	1.00
T2	1.41 (1.04–1.93)	0.028	1.4	0.53	0.14	0.32	0.84	0.98	1.00
T3	1.58 (1.06–2.35)	0.023	1.5	0.6	0.10	0.25	0.79	0.97	1.00
N2	1.35 (1.02–1.80)	0.036	1.3	0.61	0.15	0.35	0.85	0.98	1.00
N3	2.53 (1.27–5.03)	0.007	2.5	0.51	0.04	0.10	0.56	0.93	0.99
N2,3 or T3,4	1.33 (1.03–1.72)	0.027	1.3	0.58	0.12	0.29	0.82	0.98	1.00

^*OR 95% CI calculated from the study sample by univariate logistic regression analysis.

^†P value calculated by the χ² test in comparisons of genotype distributions in cases and controls.

^‡OR was the odds ratio closest to the observed odds ratio in each stratified subgroup used to calculate FPRP.

^§Statistical power was calculated using the number of observations in the subgroup and the OR and P values in this table.

Discussion

In this study, we investigated the associations between three common, potentially functional polymorphisms of NEIL1 and NEIL2 genes and the risk of SCCOOP in a hospital-based case-control study. When we evaluated each polymorphism separately, only the NEIL2 g.4102971 polymorphism, but not NEIL1 g.46434077 or g.46438282, was associated with risk of SCCOOP. In addition, the NEIL2 g.4102971 polymorphism appeared to be associated with more advanced SCCOOP, particularly for oropharyngeal cancer. Given the role of the NEIL genes in the BER pathway, it is biologically plausible that functional NEIL2 polymorphisms may modulate the risk and/or progression of cancer.

Several groups have reported that NEIL1 and NEIL2 sequence variants are associated with the risk of such diseases as colorectal cancer, gastric cancer, and metabolic syndrome as well as brain ontogeny in different ethnic groups (20, 26, 27). One study suggested that mutations in NEIL1 may reduce the gene expression and protein activities in a subset of gastric cancers (18), but another study reported that some sequence variations in both NEIL1 and NEIL2 were reportedly not associated with the risk of colorectal cancer (20). However, no previous studies investigated associations between NEIL1 and NEIL2 polymorphisms and the risk of SCCOOP. Our current large case-control study support the notion that selected variants in NEIL2 may contribute to the etiology of SCCOOP.

Several lines of evidence of molecular mechanisms support our findings. Oxidized DNA base lesions such as thymine glycol and 8-hydroxyguanine are often toxic, mutagenic or even carcinogenic (12). In mammalian cells, the repair of DNA bases that have been damaged by ROS is primarily initiated by a series of DNA glycosylases that include OGG1, NTH1, NEIL1 and NEIL2. It has been shown that in vitro translated mouse or human NEIL1 can remove thymine glycol and 5-hydroxyuracil much more efficiently than 8-oxoG in double- and single-strand DNA (28, 29). With the loss of the NEIL1 functions, NEIL1 knockout mice develop metabolic syndrome, manifesting as severe obesity, dyslipidemia, and fatty liver disease (26). The genetic variants of NEIL1 selected in the present study may not have been sufficient to cause severe loss of NEIL1 functions; alternatively, our study may not have had enough power to detect small differences in terms of cancer risk, if any existed. NEIL2 primarily functions to excise oxidative products of cytosine, with its greatest activity for 5-hydroxyuracil, but it shows negligible or undetectable activity for 8-oxoG (14). Our finding of an association between the selected NEIL2 variant and risk of SCCOOP could be due to a function loss; alternatively, this variant may be in linkage disequilibrium with other causal variants. These hypotheses should be tested in future mechanistic studies.

In the present study, we observed an increased risk associated with the variant NIEL2 genotype in former smokers but not current smokers, and this finding is likely to be a chance finding. Likewise, we also observed a significantly increased risk of SCCOOP among current drinkers of alcohol, suggesting that a gene-environment interaction may be involved in the development of SCCOOP; however, our study did not have enough statistical power to confirm any such gene-environment interaction. Ethanol reportedly induces oxidative stress via metabolic activation, leading to oxidative DNA damage and a decrease in hepatic antioxidant defense (27, 30), which may explain the putative role of the NEIL2 g.4102971 polymorphism in alcohol-induced SCCOOP. However, this finding may also have occurred by chance, owing to the small number of observations in our stratified analysis.

More interesting is our finding that NEIL2 g.4102971CC genotype was significantly associated with more-advanced SCCOOP, suggesting that the NEIL2 g.4102971CC genotype may be associated with SCCOOP progression, although it is also possible that our results may be due to selection bias commonly occurring in hospital-based case-control studies. However, we can speculate that because NEIL2 is involved in the repair of oxidative damage to DNA either accumulated in the aging process or as a result of fast-growing tumors due to their enhanced metabolic activities (31), an altered function of NEIL2 due to the variant may lead to excessive oxidative damage to DNA, causing additional mutations or a mutator phenotype with genomic instability that may promote tumor progression (32). This hypothesis needs to be tested in additional mechanistic studies.

It is difficult to compare our genotyping data with those from other studies, because few studies on NEIL1 and NEIL2 polymorphisms have been published. Additionally, because other less common functional variants of these two genes were not assayed in the present study, our finding must be confirmed in studies that enroll larger numbers of patients with SCCOOP with genotyping data based on dense gene maps, such as the HapMap database (33).

In conclusion, we found that the NEIL2 g.4102971CC genotype was associated with a significantly increased risk of SCCOOP, particularly advanced SCCOOP or oropharygeal cancer, compared with other genotypes. However, the FPRP values for all the significant findings in our study were greater that 0.2 when the prior probability was set at 0.01, suggesting that these findings could all be false positives. Therefore, our findings should validated in future population-based studies that include larger numbers of patients with oropharyngeal cancer, more detailed data on environmental exposure, more SNPs in more genes in the same biologic pathway, and survival data. Because the majority of patients did not have enough follow-up time or death events yet, we will evaluate the role of NEIL2 g.4102971CC genotype on disease prognosis in this patient cohort in the future follow-up study.