NAD(P)H:Quinone Oxidoreductase 1 (NQO1) Pro187Ser Polymorphism and the Risk of Lung, Bladder, and Colorectal Cancers: a Meta-analysis

NAD(P)H:quinone oxidoreductase 1 (NQO1), formerly known as diphtheria toxin diaphorase (DT-diaphorase), is a cytosolic flavoenzyme that catalyzes the two-electron reduction of quinoid compounds into hydroquinones (1). Two competing quinone metabolism pathways exist in human cells. In one pathway, quinones undergo the one-electron reduction catalyzed by phase I enzymes, such as cytochrome b5 reductase and P450 reductase. This process leads to the formation of alkylating species and free radicals as the metabolite semiquinone auto-oxidizes under aerobic conditions (2). Alternatively, NQO1 protects cells from oxidative damage by preventing quinones from entering the one-electron reduction and by catalyzing a two-electron reduction, which leads to the less toxic hydroquinones that are readily excreted when conjugated. In vitro evidence suggests that NQO1 activity can specifically reduce the formation of benzo(a)pyrene quinone-DNA adducts generated by cytochrome P450 reductase (3). Thus, NQO1 has been described as an anticancer enzyme.

In addition to its protective role in carcinogenesis, NQO1 has also attracted wide attention as a drug-metabolizing enzyme for antitumor therapy. Many cytotoxic agents containing the quinone moiety, such as the benzoquinone group, can be activated by NQO1 through the reduction into hydroquinone [which in this case can auto-oxidize and lead to the formation of free radicals and alkylating species; (4)]. In addition to its capacity to activate cytotoxic compounds, NQO1 is also known to bioactivate certain environmental procarcinogens, such as nitroaromatic compounds and heterocyclic amines, present in tobacco smoke and certain processed foods (5, 6). The dual function of detoxification and the bioactivation of xenobiotics renders the overall role of NQO1 in carcinogenesis uncertain.

The NQO1 gene is located on chromosome 16q22. Many single nucleotide polymorphisms (SNP) have been discovered in this gene (7). Figure 1 shows the NQO1 gene structure and the positions of currently validated SNPs. A nonsynonymous SNP has been described by Traver et al. at nucleotide position 609 (8). The variant is a C-to-T transition and results in a proline to serine amino acid substitution at codon 187 in the protein (dbSNP ID: rs1800566). The variant allele results in reduced enzymatic activity according to in vitro studies (9, 10). Compared with the wild type (C/C), the homozygous variant has only 2% to 4% of the quinone reductase activity, whereas the heterozygote variant has a 3-fold decrease in enzyme activity level (11). While the Pro187Ser polymorphism has been linked to benzene toxicity (7), it has also been hypothesized to affect cancer susceptibility by modifying the internal exposure to bioactivated carcinogens. A wide variation of the allele frequency has been observed across ethnic groups. The homozygous variant genotype is as rare as 2% in white populations but as frequent as 20% in Asian populations (12). This variation may have important implications for chemoprevention and chemotherapy.

Because NQO1 activity can be up-regulated by chemical or dietary inducers (13-15), determining the role of this enzyme in individual susceptibility has important implications for cancer chemoprevention. As common environmental exposure to quinone arises from the oxidative metabolite of benzo(a)pyrene [benzo(a)pyrene 3,6-quinones] present in tobacco smoke, the effect of the NQO1 polymorphism, if any, may be more evident among smoking-related cancers. However, epidemiologic studies examining the effect of the NQO1 Pro187Ser polymorphism on smoking-related cancer risk have reported inconsistent findings. Several studies have examined the effect of the NQO1 the polymorphism on lung cancer susceptibility. The odds ratios (OR) reported seem to be in different directions in different ethnic groups. A study by Xu et al. also suggested that the effect of the genotype on lung cancer risk was strongly modified by environmental exposure [i.e., smoking behaviors; (16)].

The NQO1 genetic polymorphism has also been hypothesized to modulate the risk of bladder cancer, a largely environmentally induced cancer. Major risk factors for bladder cancer are cigarette smoking and occupational exposures to chemicals, such as aromatic amines and polycyclic aromatic hydrocarbons (17). A recent study found that NQO1 expression is elevated in bladder tumor tissues in a subset of patients (50% had a greater than four times increase in NQO1 level), suggesting that NQO1 plays a role in bladder carcinogenesis (18).

The risk of sporadic colorectal cancer has been linked to exposure to carcinogenic compounds in food, such as heterocyclic amines in fried meat (19, 20). It has been hypothesized that NQO1 expressed in the intestinal tract plays an important role in detoxifying dietary carcinogenic compounds, and the loss of enzyme activity may increase susceptibility to colorectal cancer.

In this study, we conducted meta-analyses to examine the hypotheses that the NQO1 Pro187Ser polymorphism modifies the risk of lung, bladder, and colorectal cancer. To further study the hypotheses that the association of NQO1 Pro187Ser polymorphism on cancer risk varies between ethnic groups and smoking status, we conducted stratified analyses to examine the stratum-specific OR estimates. Because lung cancers of different histologic types may have different etiologies, we also examined whether the association of the variant genotype could be more profound for certain subtypes of lung cancer.

Epidemiologic studies examining the associations of the NQO1 Pro187Ser polymorphism with lung, bladder, and colorectal cancer risk published before January 2006 were identified by searching the PubMed and the ISI web of knowledge databases using the following key words in the title/abstract text: “quinone oxidoreductase,” “NQO1,” “DT-diaphorase,” “DTD,” “quinone reductase,” “NAD(P)H dehydrogenase (quinone),” in combination with “lung cancer,” “lung carcinoma,” “colorectal,” “colon,” “rectum,” “CRC,” “bladder,” “transitional cell carcinoma,” and “urothelial.” In addition, we reviewed the literature cited by each article identified from the database search. We requested from investigators unpublished risk estimates from studies that we were aware of that had not yet been published. Studies were eligible for inclusion if at least one of the exposures included the NQO1 Pro187Ser genotypes, and at least one of the outcomes examined was lung, bladder, or colorectal cancer. We excluded studies if neither the NQO1 Pro187Ser genotype frequency among study subjects nor the crude effect estimates (ORs) nor 95% confidence intervals (95% CI) were reported. We confirmed that in all studies, cases were pathologically confirmed, and that the genotype distribution in controls (in each ethnic group if the study included several ethnic groups was multiethnic) was in Hardy-Weinberg equilibrium. We did not impose exclusion criteria based on sample size or language. Only case-control studies were included in this analysis because no other study design was identified.

Two independent researchers extracted the following information for each study: publication date, country where the study was conducted, source for controls, method of matching controls to cases, ethnicity of the study population, genotyping method, material for DNA sample, number of cases and controls, genotype frequency for cases and controls, adjusted ORs for the variant genotype, and adjustment factors. Disagreements were solved by consensus. We also extracted, whenever available, the number and genotype frequency for cases and controls by ethnicity and by ever smoking status. For lung cancer, where possible, we extracted the stratified genotype distribution by histologic type. We requested the stratified genotype distribution from the corresponding author, when smoking status or lung cancer histology was available in the study but was not reported in the article.

For each study, we examined whether the genotype distribution in controls was in Hardy-Weinberg equilibrium using the χ² test. We then calculated the crude OR and 95% CIs for lung, bladder, and colorectal cancer. Summary estimates for crude as well as adjusted ORs were calculated with the statistical program STATA [version 8; (21)]. For each analysis, data were combined using inverse variance weighing and using both fixed effects and random effects models. Summary estimates [relative to wild type (C/C)] were obtained separately for the heterozygous genotype (C/T), the homozygous variant genotype (T/T), and for carrying at least one variant allele (C/T or T/T). A stratified analysis was conducted to examine the effect estimates within subgroups defined by ethnicity, smoking status, or histology. Some studies included subjects from multiple ethnic groups; these subjects were extracted into separate substudies within the respective studies and analyzed as independent studies in the stratified analysis where possible. We estimated summary ORs when, within cancer site, there were at least three studies reporting the stratified genotype frequencies or stratified effect estimates.

Heterogeneity between studies was examined using the Mantel-Haenszel test for heterogeneity. Due to the small number of studies in some strata, P < 0.10 were used as an indication for the presence of heterogeneity. The random effects model was used for the estimate if heterogeneity was suggested by the test (P < 0.1). If heterogeneity was detected after stratification by ethnicity, two approaches were used to explore the source of heterogeneity. We did a sensitivity analysis based on the magnitude of the Q statistic to assess the contribution of each study to the heterogeneity detected (22, 23). In addition, the analysis was further stratified based on the region (e.g., America, Europe, and China) of the study to explore the source of heterogeneity.

Publication bias was assessed with a visual inspection of the funnel plot, by the rank correlation test of Begg and Mazumdar (24) and the regression asymmetry test of Egger et al. (25). A significance level of 0.10 was used as an indication for the potential presence of publication bias. We also conducted influence analyses, where each study was excluded one at a time to determine the magnitude of influence on the overall summary estimate. We considered a study to be influential if the exclusion of the study changes the effect estimate by 20%.

Twenty-one case-control studies examining the effect of NQO1 Pro187Ser polymorphisms on the risk of lung cancer were identified (16, 26-45); one study was only identified from citation (34), and one was an unpublished study (44). We excluded one study due to the lack of genotype information (37). Study subjects in Lin et al. (35) seemed to overlap with the subjects in Lin et al. (36); therefore, the former was excluded. The remaining 19 studies included a total of 6,980 cases and 8,080 controls that had NQO1 Pro187Ser genotype information. Of the 19 studies, eight studies were based on white populations, seven studies were based on Asian populations, and four studies included multiple ethnic/racial groups. Selected study characteristics are summarized in Table 1. The genotype distributions in controls were in agreement with Hardy-Weinberg equilibrium for all studies. Subjects in the pooled study by Skuladottir et al. (39), focusing on lung cancer risk for subjects under 60 years, seemed to overlap by about 100 cases and 100 controls with the subjects in Sorensen's study (40). Therefore, we only included subjects over 60 years from Sorensen's study in the overall meta-analysis to avoid double counting. In the study by Hamajima et al. (30), 20% of the controls from one of the two control sources were expected to have cancer. This source of control was thus excluded from our meta-analyses.

When all studies were included in the meta-analyses, the variant allele did not seem to be associated with lung cancer risk, as shown in Table 2. The summary OR for carrying one variant allele and the homozygous variant genotype was 1.04 (95% CI, 0.92-1.19) and 1.07 (95% CI, 0.98-1.16), respectively. Strong heterogeneity between studies was detected for the effect of carrying the variant allele.

When stratified on ethnicity, heterogeneity between studies seemed to have been reduced in white and African-American populations. In the white population, there was no clear association between the variant genotypes and the risk for lung cancer, although a modestly increased risk was suggested for the homozygotes (OR C/T versus C/C, 1.07; 95% CI, 0.98-1.16; OR T/T versus C/C, 1.19; 95% CI, 0.94-1.50; Table 2). The meta-analysis showed no effect of carrying at least one variant allele in African-American populations.

No association was observed for all variant genotypes in Asians (Table 2). However, strong heterogeneity between studies was detected within the Asian populations. With an attempt to minimize heterogeneity, we removed one study in our sensitivity analysis that contributed the most to heterogeneity (34). In contrast to the results for whites, after the exclusion of the outlier study, reduced lung cancer risk was observed in the Asian population for carrying the variant allele [OR C/T versus C/C, 0.85; 95% CI, 0.69-1.05; OR T/T versus C/C, 0.81; 95% CI, 0.52-1.24 (random effects model) and OR C/T + T/T versus C/C, 0.87; 95% CI, 0.68-1.11 (random effects model)]. Stratification on region did not entirely explain the heterogeneity detected, as heterogeneity persisted within the studies conducted in China.

We conducted meta-analyses restricted to studies that presented adjusted ORs to assess whether adjustment for potential confounders may affect the results (Table 3). Most studies were adjusted for age, sex, and smoking (Table 1). For whites, the results were similar to those of the meta-analyses of crude estimates including all studies (Table 2). The variant genotypes were found to be associated with reduced lung cancer risk in Asians in these analyses, reflecting the exclusion of the outlier study (34) due to the nonavailability of the adjusted estimates from this study. We did not find significant differences between the adjusted and crude effect estimates. Egger's test P was small (P = 0.02) in the meta-analysis for the adjusted OR in whites (Table 3; C/T versus C/C). However, we did not find evidence for publication bias from examining Begg's funnel plots.

Effect modification by smoking status seemed to be minimal in Asian populations (Table 4). In general, a protective effect was indicated, especially for ever smokers (OR, 0.78; 95% CI, 0.59-1.02). In the white population, there seemed to be a slightly elevated risk for carrying the variant allele among ever smokers (OR, 1.08; 95% CI, 0.96-1.22) but not among never smokers [OR, 0.97; 95% CI, 0.62-1.54 (random effects model)]. The study by Saldivar et al. (38) seemed to be influential for the effect of the NQO1 polymorphism among White never smokers. Removing this study from the pool resulted in a summary OR of 0.76 (95% CI, 0.54-1.08; random effects model).

When we stratified the analysis based on lung cancer histology, no effect of the NQO1 polymorphism on adenocarcinoma or small cell lung cancer in the White populations was observed (Table 4). There is possibly an increased risk for squamous cell carcinoma for the T allele carriers (OR, 1.11; 95% CI, 0.92-1.33). In the Asian population, the protective effect of the NQO1 variant genotype was only suggestive for adenocarcinoma (OR, 0.88; 95% CI, 0.68-1.15) but not for squamous cell carcinoma (OR, 1.28; 95% CI, 0.88-1.88).

We identified nine studies examining the effect of the NQO1 Pro187Ser polymorphism on the risk of bladder cancer (46-54); one study (51) was excluded due to the lack of genotype information. The remaining eight studies included a total of 1,410 cases and 1,485 controls. Of the eight studies, six studies were based on white populations, and the other two were based on Asian and Hispanic subjects (Table 1). Genotype distribution in controls was in agreement with Hardy-Weinberg equilibrium for all eight studies.

The summary OR suggested that the T/T genotype increased the risk of bladder cancer when we restricted our analysis to whites (OR, 1.20; 95% CI, 1.00-1.43; Table 2). The adjusted ORs showed a similar pattern of association as the summary crude estimates (Table 3). Analyses further stratified by ever smoking status in white populations suggested a possible effect modification by ever smoking on the effect of the NQO1 polymorphism, although the CIs overlapped (Table 4). The study by Park et al. (50) seemed to be influential for the effect of the NQO1 polymorphism on bladder cancer among white never smokers. Removing this study from the pool resulted in a summary OR of 2.67 (95% CI, 1.48-4.84; random effects model) for never smokers. Although Egger's test P was small (<0.05) in some meta-analyses in whites (Table 2, C/T versus C/C; Table 3, C/T + T/T versus C/C), we did not find evidence for publication bias from examining Begg's funnel plots.

We identified six studies examining the effect of the NQO1 Pro187Ser polymorphism on the risk of colorectal cancer (30, 55-59). Of the six, five studies were based on white populations, and one was based on a Japanese population (Table 1). Genotype distribution in controls was in agreement with Hardy-Weinberg equilibrium for all studies. The six studies included 1,781 cases and 2,494 controls that had genotype information.

Overall, the heterozygous genotype was associated with a modestly elevated risk for colorectal cancer (OR, 1.15; 95% CI, 1.01-1.32), whereas the homozygous variant genotype was not associated with the risk (Table 2). When we excluded the Japanese study (30), carriers of the T allele had an increased risk for colorectal cancer in the white population (OR, 1.18; 95% CI, 1.02-1.35). When restricted to the subset of studies reporting adjusted risk estimates, the overall association was not observed for the adjusted or the crude ORs (Table 3). Adjusted risk estimates for Whites and stratified analysis by anatomic site were not possible for colorectal cancer due to the limited number of studies. No evidence of publication bias was found for the NQO1 polymorphism and colorectal cancer risk based on Egger's test and Begg's funnel plot.

We found that the NQO1 Pro187Ser polymorphism seems to modify the risk of lung, bladder, and colorectal cancer, but this effect may vary by the predominant ethnicity of the studied group. A clear association between the polymorphism and lung cancer risk was not indicated in whites. It seems that in Asian populations, the polymorphism may possibly decrease the risk of lung cancer. An inverse dose-response was observed with increasing numbers of the T allele, providing further support of the causal association between the NQO1 Pro187Ser polymorphism and lung cancer in Asians. For bladder cancer, the presence of the variant genotype seems to increase risk, and indeed, the relative risk is greater among those who are homozygous than those who are heterozygous. With respect to colorectal cancer, the presence of the NQO1 Pro187Ser variant genotype seems to modestly increase risk. However, we are less confident about this conclusion because the heterozygous genotype seems to have higher risk compared with the homozygous variant genotype. Our analyses for colorectal and bladder cancer were only done on whites because there are insufficient studies in other racial groups to examine these outcomes.

Although we did not observe a clear association between lung cancer risk and the variant genotypes in whites, a positive association was suggested for the T/T genotype. Because the T/T genotype is very rare in the white population, we might simply not have enough power to detect the association. For this same reason, we could not conduct a separate analysis for the T/T genotype in the stratified analyses. Thus, the possibility that the T/T genotype, if examined separately, may interact with the environmental exposure on modifying lung cancer risk cannot be excluded. Further studies with sufficient power are needed to study lung cancer risk in the T/T carriers in the white population.

There was no overall association observed between the NQO1 variant genotype and lung cancer risk in Asians when all Asian studies were included. However, this conclusion of null association was highly sensitive to the inclusion of the study by Lin et al. (34). After we excluded this outlier study in the sensitivity analysis to minimize heterogeneity, the results suggesting a protective role of the variant genotype were no longer sensitive to the inclusion/exclusion of any other study. Therefore, our interpretation for Asian populations was mainly based on the results obtained after the exclusion was made.

Our lung cancer results are consistent with a previous meta-analysis on the NQO1 Pro187Ser polymorphism and lung cancer risk, which included 13 studies published before August 2004 [OR C/T + T/T versus C/C for white, 1.10; 95% CI, 0.96-1.26 based on five studies in the white population and OR C/T + T/T versus C/C for Asians, 0.78; 95% CI, 0.64-0.94 based on five Asian studies (60)]. Our meta-analysis included 19 studies, and we were able to conduct stratified analyses based on smoking status and histology type for each ethnic group. The comparison between the overall and smoking/histology–stratified OR estimate provides further insight for causal inference and gene-environment interaction.

The potential protective effect of the variant genotype in Asians may be explained by the failure of the variant enzyme to bioactivate procarcinogens, such as nitrosamine in tobacco. Such reduced activity in activating environmental procarcinogens seems to be more important than the reduced detoxifying activity in Asian populations in modifying lung cancer risk, especially for adenocarcinoma. This is consistent with the finding that tobacco-specific nitrosamines [4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone] seem to have some specificity for adenocarcinoma (61, 62). The overall protective effect observed may be due to the fact that adenocarcinoma has replaced squamous cell carcinoma as the most common type of lung cancer in Asia (63). It is possible that the effect of the SNP differs by histology, although we did not observe a significant difference in risk of lung adenocarcinoma and squamous cell carcinoma for Asians.

The only Asian study (Korean) that examined the relationship between the NQO1 Pro187Ser polymorphism and bladder cancer found a protective effect of the variant allele (47). Similarly, the only Asian study (Japanese) that examined the relationship between the variant allele and colorectal cancer found a significant protective effect (30). Although the findings from Asian studies for bladder and colorectal cancer need to be confirmed by additional studies, there seems to be some consistency in the protective effect of the variant genotype in Asians.

If the protective effect observed in our study was truly unbiased, several possibilities may explain why such an effect was only detected in Asians but not in whites. First, because the NQO1 enzyme possesses the dual function of detoxifying and bioactivating different carcinogenic compounds, the relevant environmental exposures in Asia may differ from Western countries (e.g., smoking of low-tar cigarettes). Second, there may be other genetic mechanisms present in the Asian population that compensate more effectively for the loss of the detoxifying activity of the NQO1 variant type. Furthermore, the Pro187Ser SNP may be in linkage disequilibrium with other functional SNPs in Asians. Further studies involving different ethnic/racial groups from the same geographic areas (therefore similar environmental exposure) may help to answer this question.

Our study should be interpreted in light of a number of weaknesses. This study is a meta-analysis of case-control studies, most of which were hospital based. Thus, the controls may not always be truly representative of the underlying source populations, especially when the polymorphism is also expected to affect the risk for other diseases. In addition, the definition for ever smoking and the adjustment factors for the adjusted ORs varied between studies. A pooled analysis is more ideal to examine effect estimates by strata defined by multiple variables (e.g., sex, race, and age) and to examine pooled estimates adjusting for the same set of covariates. Although we did not find evidence for publication bias, we could not completely exclude such a possibility. In particular, our results for the Asian populations should be interpreted in a conservative manner, keeping in mind the different degree of publication bias operating in the local literature and the language bias in PubMed-indexed Asian studies (64-66).

To our knowledge, this study is the most comprehensive meta-analysis to date to have assessed the relationship between the NQO1 Pro187Ser polymorphism and lung cancer and the first systematic review for the NQO1 Pro187Ser polymorphism and the risk of bladder and colorectal cancer. We have found a fairly consistent association pattern with the variant genotypes across all three types of cancers within a given ethnicity. Furthermore, whether we used the crude or the adjusted risk estimates, the overall inference we obtained remained the same. In summary, the variant genotype was associated with a modest increase in risk for bladder and colorectal cancer in whites, whereas a protective role of the variant genotype was suggested among Asian populations. Future analyses should be conducted in large-scale cohorts or case-control studies and should study potential effect modification by the predominant exposures in different populations.

We thank the following individuals for providing unpublished genotype data for our analyses: Dr. Cathryn Bock (Karmanos Cancer Institute, Wayne State University School of Medicine), Dr. Karin Broberg (Lund University Hospital, Sweden), Dr. Nobuyuki Hamajima (Nagoya University Graduate School of Medicine, Japan), Dr. Rayjean Hung (IARC, France), Dr. Pinpin Lin (National Health Research Institutes, Taiwan), and Dr. Mette Sorensen (Institute of Cancer Epidemiology, the Danish Cancer Society, Denmark).