In this study, we investigated associations between four selected potentially functional SNPs of ERCC4
and risk of SCCHN in a non-Hispanic white population. Our data showed that the variant GG genotype of rs2276466 was significantly associated with a decreased risk of SCCHN and that the variant TT genotype of rs3136038 showed a borderline significant decreased risk with SCCHN in recessive models. Such protective effects were more evident in oropharyngeal cancer, older subjects, non-smokers and alcohol drinkers. But no significant interaction effect between any of the variables analyzed was found. No significant associations were observed between the other two SNPs (i.e., rs1799798 and rs1800067) and risk of SCCHN. In addition, the observed significant associations are supported by data that the variant homozygous genotypes of rs2276466 and rs3136038 were associated with increased mRNA expression levels of ERCC4
in EBV-transformed lymphoblastoid cell lines derived from Caucasians. Our findings suggested that rs2276466 and rs3136038 in ERCC4
may be functional and contribute to SCCHN susceptibility. Although the published genome-wide association study (GWAS) have not identified ERCC4
variants as susceptibility loci 
, it was not possible to evaluate the effects of the genotypes for those potentially functional SNPs in ERCC4
in the GWAS, in which no enough samples nor information about exposures were available for analysis of such gene-environment interactions. In the present study, the data were available for us to evaluate the potential interaction effects between ERCC4
polymorphisms and selected variables on SCCHN risk, and no significant interaction effects were found, suggesting that no risk effects were modified by these variables under investigation. However, our findings need to be confirmed by large-scale studies.
A few studies have investigated the association between ERCC4
SNPs and risk of SCCHN, which focused on rs1800067 (Arg415Gln) with inconsistent results. Only three published studies with small sample sizes have examined rs1800067 in association with risk of laryngeal cancer in Polish and German populations, but no significant associations were found in these studies 
, which are similar to our results. No published studies, so far, have reported the association between rs2276466, rs3136038 and rs1799798 and SCCHN risk.
rs2276466 and rs3136038 reside in the predicted miRNA-binding site of the 3′UTR region and the transcription factor binding site of the 5′UTR region of ERCC4
, respectively. Although functional significance of these two SNPs remains unclear, a growing body of evidence indicated that SNPs located at miRNA-binding sites or transcription factor binding sites are likely to affect gene expression levels by modifying miRNA targeting activity or by altering DNA binding properties of transcription factors and thus may contribute to susceptibility to cancer 
. Interestingly, we found that the homozygous variant genotypes of both rs2276466 and rs3136038 were indeed associated with increased mRNA expression levels of ERCC4
in EBV-transformed lymphoblastoid cell lines derived from Caucasians, which supports the observed associations of these two homozygous variant genotypes with a decreased risk of SCCHN. Our previous studies also showed that decreased expression of ERCC4
mRNA and protein in lymphocytes was associated with an increased risk of SCCHN 
. Recently, Vaezi et al. reported that low expression of ERCC4 was associated with longer progression-free survival in patients with SCCHN treated with DNA damaging agents 
. Both in vitro
and in vivo
studies from XPF-deficient cells and mice showed that lower XPF expression was associated with an exquisite sensitivity to DNA damaging agents 
. Taken together, these two potentially functional SNPs (rs2276466 and rs3136038) may alter mRNA expression levels of ERCC4
, thereby affecting DRC and modulating cancer susceptibility. However, the exact molecular mechanisms underlying the observed effects need further investigation.
In summary, we report that two potentially functional ERCC4 SNPs, i.e., rs2276466 and rs3136038, may influence mRNA expression levels of ERCC4 and thus contribute to genetic susceptibility to SCCHN. Nevertheless, some limitations in the current study should be considered. Firstly, since this was a hospital-based case-control study, the issue of selection bias was unavoidable. However, the hypothesis tested in this study was a genotype-driven rather than an environment-driven, and it is unlikely to have improperly selected individuals related to genotype. Secondly, ERCC4 may play an important role in NER involved in removal of a variety of bulky DNA adducts such as those formed by tobacco carcinogens PAHs. Therefore, potential residual confounding may exist, if only the smoking status but not the smoking dose and duration data were available in our current study. For example, we observed a protective effect of the homozygous variant genotypes of rs2276466 and rs3136038 in non-smokers but not smokers, which is difficult to explain. Hence, detailed information for smoking or drinking habits (dose and duration) should be taken into account in future studies. Thirdly, although the sample size of our study was relatively large, the statistical power was still limited in the analyses of some groups with small sample sizes. For example, in the analyses by tumor sites, the ERCC4 rs2276466 and rs3136038 were only found to be associated with risk of oropharyngeal cancer but not for cancers of oral cavity, hypopharynx and larynx. In addition, no significant associations were observed between the other two SNPs (i.e., rs1799798 and rs1800067) and risk of SCCHN in the recessive model. This lack of significance could be either because there was no such effect or because the small sample size in these groups (only 307 oral cavity cancer patients and 203 larynx and hypopharynx patients, and only eight cases and eight controls had the homozygous variant genotype for rs1799798 and rs1800067, respectively) limited the statistical power to detect a modest effect on cancer risk. Hence, our findings should be interpreted with caution. Fourthly, due to the multiple comparisons being made, we cannot rule out the possibility of false-positive associations. When correcting our estimates for multiple testing according to Bonferroni, no significant associations were found. Therefore, our findings need to be replicated, and more profound functional studies are necessary to explore the functional relevance of these SNPs and the underlying molecular mechanisms for the observed associations.