Alcohol Dehydrogenase 3 and Risk of Squamous Cell Carcinomas of the Head and Neck

Alcohol dehydrogenase 3 (ADH3) gene loci carry two alleles coding respectively for the γ1 and γ2 subunits of the dimeric ADH3 enzyme. γ1 and γ2 subunits are different by one single amino acid, arginine for γ1 and glutamine for γ2 at 271. The various isoenzymes formed by combinations of these two different allelic products had different in vitro kinetic properties with respect to the rate of ethanol oxidation; isoenzymes encoded by the ADH3*1 allele activated ethanol 2.5 times as fast as those encoded by the ADH3*2 allele (1). All five classes of ADH enzymes are responsible for the oxidation of ethanol to acetaldehyde, which mainly occurs in the liver, and the ADH3 genes belongs to ADH class I (2). Although pure ethanol has never been shown to be carcinogenic in laboratory experiments, it has been consistently shown to be one of the major risk factors for head and neck cancer (3-7 ). Acetaldehyde, the metabolite of alcohol oxidation, is implicated in ethanol-induced cell damage, and production of free radicals and DNA hydroxylated bases. It is also mutagenic and carcinogenic in experimental animals, in short term cell culture assays (8-12).

Based on these experiment results, it has been hypothesized that individuals who are homozygous for the fast allele of ADH3 are at greater risk for alcohol-related cancers. Due to the anatomic location of the head and neck, and the different catalytic function of ADH3 on alcohol metabolism, this hypothesis seemed reasonable, which is why it has drawn much attention. Thus far, seven studies have examined the association between ADH3 genotypes and the risk of head and neck cancer.

In a study on French alcoholics, Coutelle et al. (13) found a significantly elevated occurrence of the ADH3^1-1 genotype among French Caucasian men with laryngeal or oropharyngeal cancer. Harty et al. (5) also reported that the ADH3^1-1 genotype increased the risk of oropharyngeal cancers among Puerto Rican heavy alcohol drinkers. However, Bouchardy, Olshan, Sturgis, and Zavras and their coworkers found no evidence to support the main effect of ADH3 genotypes, and no joint effects or interactions were detected between ADH3 genotypes and the amount or duration of alcohol consumption (6, 7, 14, 15). Schwartz et al. (16) found that individuals with ADH3^2-2 had increased risk of head and neck squamous cell carcinomas (HNSCC) compared with those with ADH3^1-2 and ADH3^1-1 genotype. In a pooled analysis of data from these seven case-control studies, significantly increased risk of head and neck cancer was not observed among different ADH3 genotypes (17).

Because of these inconsistent findings, we conducted a case-control study to examine the association between the ADH3 genotypes and risk of HNSCC, and to determine whether ADH3 genotypes modified HNSCC risk increased by alcohol drinking and whether ADH3 genotypes were of prognostic significance for HNSCC. Because an association between the presence of human papillomavirus (HPV) and the development of head and neck cancer is widely accepted, and because we also found that HPV had an interaction effect with alcohol in our previous study (18), we considered it important to include the HPV data in this study to evaluate whether the interaction between HPV and alcohol consumption was associated with a specific ADH3 genotype. Therefore, we examined the interaction between HPV and the ADH3 genotype.

The details of subject recruitment were described previously (19). Briefly, cases were newly diagnosed or previously diagnosed HNSCCs enrolled in two periods: from 1994 to 1997 and from 2000 to 2002 at the University of Iowa Hospitals and Clinics (UIHC) and the Iowa City Veterans Affairs Medical Center (VAMC). All controls were being seen for annual or disease screening visits in the Departments of Family Medicine and Internal Medicine at the UIHC and at the College of Dentistry, University of Iowa, during the same time period as the cases. Criteria for inclusion required that the control not be seriously chronically or acutely ill. They were also excluded on the basis of never having a diagnosis of a HNSCC by past medical history, questionnaire, and a review of the Iowa Surveillance, Epidemiology, and End Results (SEER) database. All participants signed a University or VAMC–approved human subject consent form before enrollment. Of the eligible head and neck cancer cases, 2.5% were excluded because of mental incapacity or a language barrier, 5.5% did not complete the questionnaire or specimen collections procedures prior to treatment, and 5% refused to participate. Less than 1.0% of the controls were excluded because of prior head and neck cancer, 2.0% did not complete the specimen collection requirements, and 5.0% refused to participate. A total of 348 HNSCC cases and 330 controls were included in the study.

Data regarding demographics, risk factors for HNSCC, and risk related to HPV exposure were collected in a self-administered questionnaire by trained clinical staff. An exfoliated oral cell rinse was collected on all participants. This was done using 10 cc of normal saline which was swished in the mouth for 30 seconds, expectorated into a pill cup, and transferred to a snap cap tube for freezing at −80°C. All specimens were reviewed prior to freezing using a hemocytometer to verify and estimate the number of nucleated squamous cells. The cell count information was used in preparing the PCR reaction to assure a minimum number of squamous cells were included in each reaction.

After proteinase-K digestion, DNA extraction was done by phenol-chloroform from the samples collected between 1994 and 1999, and by a QIAamp DNA blood mini kit from samples collected from 2000 to 2002 according to the manufacturer's instructions (Qiagen, Valencia, CA). All extracted DNA were stored at −20°C. The DNA was used for both ADH3 and HPV analyses.

A 162-bp fragment of the ADH3 gene was amplified by PCR by using 0.1 g genomic DNA. Primers used were modified from Olshan et al. (6), ADH 3ex 8F (AAT AAT TAT TTT TCA GGC TTT AAG AGT AAA TAT TCT GT) and ADH 3ex 8R (TGT GTG TGT AAT CTA CCT CTT TCC AGA GC). The nine-nucleotide TGT GTG TGT “tail” was added to the reverse primer to improve the resolution of the restriction digest fragments of the PCR amplification products. PCR amplification was done in a thermocycler (PE Biosystems, Foster City, CA). Genomic DNA was denatured at 94°C for 10 minutes; then 35 cycles were carried out, each comprising denaturation for 30 seconds at 94°C, annealing for 30 seconds at 55°C, and extension for 45 seconds at 72°C. Each set of PCR amplifications included a sample with a known ADH3 type as a positive control and a tube with all the PCR reagents, but without any DNA, used as a contamination control.

The PCR products were digested with SspI at 37°C for more than 6 hours. The digested PCR products were then run on a polyacrylamide gel, stained by the silver stain method (20). Then, the gels were dried by the gel dryer and results were recorded. The ADH3^1-1 allele was characterized by the presence of fragments of 68 and 63 bp, whereas the ADH3^2-2 allele was characterized by a single 131 bp fragment, the ADH3^1-2 allele was determined by showing the three 131, 68, and 63 bp bands. Positive (samples of known genotype) and negative (PCR reagents only) controls were included in each batch of PCR.

Materials and protocols used for HPV detection and typing, and the definitions of high-risk and low-risk HPV types were essentially the same as previously reported (19). Briefly, PCR was done with primers of MY09/MY11 to amplify HPV and primers of β-globin to verify the presence of intact target DNA and the adequacy of PCR amplifications. A 5 μL aliquot PCR product was further analyzed with dot blot hybridization for a more sensitive detection of HPV. The samples positive by the dot blot assay underwent hemi-nested PCR with MY09/GP5+ or nested PCR with GP5+/GP6+ primers to provide an adequate specimen for sequence analysis, which could identify a broad range of both low-risk and high-risk HPV types.

Total cumulative tobacco exposure was expressed in pack-years and included exposure to cigarettes, cigars, pipe smoking, snuff, and chewing tobacco. Alcohol consumption was evaluated as the lifetime average number of alcoholic drinks consumed per week, which was based on the self-reported data for each type of beverage. One drink was considered equivalent to 1.5 ounces of liquor, 4 ounces of wine, or 12 ounces of beer. A dose-duration measure for alcohol was also examined, which was calculated as the average number of drinks per week multiplied by duration of regular usage in years, but results were similar to those reported for average number of drinks per week. Never tobacco or alcohol use was defined as not having used these substances on a regular basis for 1 year or more during their lifetime. Former use was defined as not having used tobacco or alcohol at least 1 year prior to diagnosis or interview. The Wilcoxon rank sum test was used to compare quantitative variables between two groups of patients, and the χ² or Fisher's exact test was used to compare categorical variables among groups. The χ² goodness-of-fit test was used to determine whether the observed ADH3 distribution among cases or controls was different from the expected distribution under the Hardy-Weinberg equilibrium.

Multiple logistic regression was used to estimate the odds ratios (OR). ORs examining the association between HNSCC and demographic factors were adjusted for age as a continuous variable, tobacco pack-years (never, 0-30, >30) and average number of drinks per week (never, 1-21, >21). The cutpoints examined for tobacco and alcohol usage were based on the distributions among controls. Lack of linearity for continuous variables in the logistic regression models was assessed as discussed by Hosmer and Lemeshow (21). When analyzing alcohol usage as a continuous variable, we truncated the average drinks per week at 140 to reduce the influence of extreme values. Effect modification between ADH3 genotype and alcohol or tobacco usage was examined on a multiplicative scale by including the appropriate interaction term into the logistic regression models.

Complete follow-up data were obtained from the National Cancer Institute Iowa SEER Cancer and UIHC Tumor Registries for the cases enrolled between 1994 and 1997. Therefore, time-to-event analyses were restricted to newly diagnosed cases enrolled during this period; HNSCCs that were recurrent at study entry were not included. Survival was measured in years from the date of diagnosis until death or until the date the patient was last known to be alive. Patients who died of other causes than HNSCC were considered censored observations in the disease-specific survival analysis. Time to recurrence was measured in years from the date of diagnosis until disease recurrence or until the patient was last known to be alive. Patients that were never disease-free were not included in the “time-to-disease recurrence” analyses and patients dying before disease recurrence were treated as censored observations. Hazards ratios (HR) measuring the association between ADH3 genotype and survival or recurrence were estimated from the Cox proportional hazards models (22) and were adjusted for factors previously found to be prognostically significant in HNSCC (23). These included age, gender, stage of disease at initial diagnosis, tumor grade, treatment, tobacco pack-years, average drinks per week, and HPV high-risk (HPV-HR) status. The proportional hazards assumption was assessed for each covariate, including ADH3 genotype, by the Wald statistic for the interaction term between the prognostic factor and time to event (24). Survival curves were estimated by the Kaplan-Meier method (25) and were created in Splus 2000 (26). Statistical analyses were done using SAS version 8.2 (27).

Calculations of false-positive report probabilities (FPRP) were conducted as discussed by Wacholder et al. (28) using prior probabilities that ranged from 0.00001 to 0.25 and HRs of 1.2, 1.5, 2.0, and 3.3 (the observed HR). Power estimates that were needed in the FPRP calculations were based on a two-sided log rank test (29), at a significance level of 0.05, and assumed a 5-year survival rate of 70% for the group with better survival (23), an a priori estimate.

We evaluated 348 HNSCC cases (39.9% in the oropharynx and 64.1% in the oral cavity) and 330 controls. Most participants were Caucasian (Table 1). Compared to controls, case participants were more likely to be current smokers, drink more heavily, or do both.

ADH3 genotypes and allele frequencies were similar between cases and controls (Table 2). There was no association between carriers of ADH3*1 and risk of HNSCC after adjusting for tobacco usage, alcohol consumption, and age. The genotype distribution in cases satisfied the Hardy-Weinberg equilibrium (X² = 0.65; df = 2; P = 0.42) as did the controls (X² = 0.053; df = 2; P = 0.974). The presence of at least one ADH3*1 allele and ADH3^1-1 genotype frequency were also not significantly different between cases and controls (54.7% versus 60.3%, P = 0.10, and 31.0% versus 36.1%, P = 0.17, respectively). Genotyping results were 100% concordant for duplicate assays in 5 control and 35 case samples. Positive control from each batch always had the same ADH3 genotype.

Alcohol consumption was significantly associated with risk of cancer after adjusting for age, tobacco usage, and ADH3 genotype. Relative to participants who never drank, those who consumed less than 21 drinks per week on average had 1.7 times the risk of developing HNSCC, and those who drank more than 21 drinks per week on average had 3.5 times the risk. When alcohol consumption was modeled as a continuous variable, there was a lack of linearity (P < 0.001). Therefore, average drinks per week was examined as a categorical term. There was no significant interaction between ADH3 genotypes and alcohol usage (P = 0.8). When the association of ADH3 genotype with HNSCC was stratified by alcohol usage (Table 3), there was a statistically nonsignificant reduced risk for ADH3^1-1 (OR, 0.6) and ADH^1-2 (OR, 0.5) among light drinkers. These associations were stronger among those with tumors in the oral cavity than they were for those with cancer in the oropharynx. Compared with ADH3^2-2, those with ADH3^1-1 or ADH3^1-2 who reported drinking on average 1 to 21 drinks per week had a lower risk of cancer of the oral cavity (OR, 0.4, 0.2-1.0 and 0.4, 0.2-0.9, respectively). These associations were not seen among those with oropharyngeal carcinomas (ADH3^1-1: OR, 0.9, 0.3-2.8; ADH3^1-2: OR, 0.8, 0.3-2.0, respectively). Alcohol consumption was also evaluated by dose-duration but similar conclusions as those reported for average drinks per week were observed.

After adjusting for age, alcohol consumption and ADH3 genotypes, HNSCC risk was elevated among those who smoked more than 30 packs per year (OR, 1.5, 0.97-2.5) compared with never tobacco users. ADH3 genotypes did not modify HNSCC risk among smokers overall (P = 0.8) or among cases with oropharyngeal cancer (P = 1.0). However, compared with ADH3^2-2, there was a reduced risk of oral cavity squamous cell carcinomas for ADH3^1-2 (OR, 0.3, 0.1-0.9) among those who smoked 1 to 30 packs per year.

HPV DNA was detected in 25.0% of case participants and 17.3% of controls. Five cases were infected with multiple types, all of which included HPV-16. The high-risk types in cases were HPV-16 (17.7%), HPV-18 (2.0%), HPV-31 (0.6%), and HPV-33 (1.1%); in controls HPV-16 was also the most prevalent type (8.8%), along with HPV-18, 31, and 58 (0.3% each). HPV-HR infection was an independent risk factor for HNSCC after adjusting for age, alcohol, tobacco usage, and ADH3 genotype (OR, 2.5; CI, 1.5-3.9) but ADH3 genotype did not modify the HNSCC risk associated with HPV-HR infection (P = 0.53). Four cases, but no controls, were excluded from HPV analysis for globin negativity.

To determine whether the ADH3 genotype had prognostic significance in HNSCCs, we analyzed overall survival, disease-specific survival, and time to recurrence. Of the 189 HNSCCs enrolled during 1994 to 1997, 136 were newly diagnosed cases with ADH and HPV data. Three cases were lost to follow-up within less than a month and therefore were not included in the analyses. Cause of death was unknown for 3 cases and 23 cases were never disease-free (ADH3^1-1, 11%; ADH3^1-2, 23%; ADH3^2-2, 13%). Of those last known to be alive, the median follow-up time was 6.8 years (ranging from 4.3 to 8.5 years). Overall survival, disease-specific survival, and time-to-recurrence curves by ADH3 genotype are shown in Fig. 1, and 2- and 5-year survival and recurrence rates are shown in Table 4. In multivariate analyses, adjusting for age, gender, stage of disease, treatment modality, tumor grade, tobacco pack-years, average drinks per week, and HPV high-risk status based on oral exfoliated cells, there were significant differences in survival and recurrence among the ADH3 genotypes with ADH3^1-2 having poorer prognosis than each of the other genotypes and being more likely to experience disease recurrence than HNSCCs with ADH3^1-1 (see Table 4 for adjusted HRs and P values).

Given that we had no a priori hypothesis concerning the association between ADH3 genotype and prognosis, we evaluated the FPRP for the overall survival comparison between ADH3^1-2 versus ADH3^1-1. Calculations were conducted as suggested by Wacholder et al. (28). If the prior probability that there is a true difference in survival between ADH3^1-2 versus ADH3^1-1 is >10%, then the FPRP would be <3%, and a statistically significant finding would have at least a 97% chance of representing a true association. If the prior probability is around 1%, then there is at least a 79% chance that a statistically significant finding represents a true association. In contrast, if the prior probability is 0.01% or less, then the FPRP would be >69%, with at most a 31% chance of representing a true association. Interestingly, the power calculation for this comparison of detecting a HR of at least 3.3 was at least 0.91 at a significance level of 0.05.

This is the first study, to our knowledge, that shows that the ADH3 genotypes may be an independent prognostic factor in HNSCC. In multivariate analyses, adjusting for age, gender, stage of disease, treatment, tumor grade, tobacco pack-years, average drinks per week, and HPV-HR status, cases with ADH3^1-1 and ADH3^2-2 had significantly better survival than those with ADH3^1-2 (HRs ranged between 0.3 and 0.4). Moreover, a higher percentage of never disease-free patients was observed among the ADH3^1-2 patients and cases with ADH3^1-1 were significantly less like to recur than those with ADH3^1-2 (HR, 0.2). Survival and recurrence analyses provided extensive follow-up duration on patients last known to be alive (ranging from 4.3 to 8.5 years). At this point, it is unclear why cases with ADH3^1-2 had worse prognosis but, as illustrated by examination of FPRP and post hoc power calculations, this possible novel finding warrants further investigation for verification. As complete follow-up data becomes available on the patients enrolled between 2000 and 2002, it will supply further insight into the associations described here between ADH3 genotypes and survival or disease recurrence.

We also showed that alcohol consumption is a significant independent risk factor for HNSCC. This result confirmed findings from other studies concerning alcohol consumption as an independent risk factor for head and neck cancer (5, 7, 16, 19). However, our data did not show the risk of HNSCC to be significantly associated with the genotypes of ADH3 after adjusting for age, race, tobacco, and alcohol usage. This lack of association with ADH3 genotypes was also noted by other investigators (6, 7, 14, 15). Our results were in contrast with two other reports which showed that participants with the ADH3^1-1 genotype had a much higher risk of head and neck cancer (5, 13), and another study showing that participants with ADH3^1-1 had borderline higher risk for oral cancer only (17). In addition, unlike one investigation (5), we did not find that the risk of HNSCC associated with alcohol consumption was modified by the ADH3 genotype in those who were heavy alcohol drinkers.

There are several potential explanations for these inconsistent results. First, alcohol metabolism in humans involves five classes of ADHs and seven classes of ADH isoenzymes. It has not been clearly determined by which class of ADH enzymes alcohol is catalyzed in oral mucosa. Only ADH class III and class IV ADHs were present from gingival and lingual mucosa, class I ADH enzyme, which included ADH3, were not detectable (30). Class I ADHs enzymes are of the highest concentration in liver and adrenals, with lower levels in kidney and lung, but not in brain and heart (2). Therefore, class III and IV ADH, not class I, might play a significant role in generating acetaldehyde in the head and neck tissues, and might thereby influence the risk for HNSCC related to excessive alcohol consumption. Furthermore, the polymorphic variation in ADH2 and ALDH2 could also result in different acetaldehyde levels in the upper aerodigestive tract through either faster production or slower removal of acetaldehyde. One case-control Japanese study on the association between ADH2 and ALDH2 and risk of head and neck cancer showed that higher risk of head and neck cancer occurred in people with ADH2^1-1 and ALDH2^1-2 genotypes (31). Because ADH2*2 and ALDH2*2 in Caucasians are low (32), no case-control study of Caucasians and the association between ADH2 or/and ALDH2 and risk of head and neck cancer has been done.

Second, sample size is an important issue in genetic studies. Interestingly, the sample sizes for the two studies showing that ADH3^1-1 had significantly increased risk of head and neck cancer were much smaller than the ones that did not find the association (5-7, 13-15). In addition, the controls of these two studies were either alcoholics or heavy drinker (≥57drinks per week). For drinkers consuming 15 to 56 drinks per week on average, the ADH3^2-2 had the highest risk instead of ADH3^1-1 as in the ≥57drinks per week group (5). The Schwartz et al. study with a larger sample size found heavy alcohol drinkers with ADH3^2-2 genotype who consumed more than 29 drinks per week was of highest risk compared with those with the genotypes of ADH3^1-1 and ADH3^1-2 but not overall and in other drink groups. This conclusion was opposite to that of Harty and Coutelle and was also contrary to the hypothesis that the fast allele of ADH3*1 was of higher risk than ADH3*2. Three other studies (6, 7, 14) had larger sample sizes, together with one with a smaller sample size (15) and hospital recruited controls, all showed no significant difference in head and neck cancer risk from different ADH3 genotypes. When the original data from the seven studies on ADH3 were combined (17), no significant increased cancer risk was observed among all head and neck cancer sites, but a slightly increased risk for oral cavity cancer cases was noted among individuals with ADH3^1-1.

A third explanation for the diverse results may be due to the fact that the presence of the ADH3*1 allele frequencies varies among different races (17). ADH3*1 allele ranged from 53.9% to 69.1% in a few studies of Caucasians from different geographies (7, 13-15), and ranged from 58.0% to 62.4% from studies of mainly Caucasian but with other races included (5, 6, 16). African-Americans had a much higher rate of the ADH*1 gene of 76% and 90% from two studies (6, 34). The study of Harty et al. included mixed race controls (69.9% white, 6.9% black, 16.4% mestizo, others 6.9%). Thus, the observed association between increased risk of head and neck cancer and ADH3^1-1 in the Harty study could possibly be due to differences in the contributions of subpopulations, rather than by a physiologic effect of the genetic variant (35). The pooled analysis (17) showed that ADH3 genotypes of controls in three other studies (13, 14, 16) departed from Hardy-Weinberg equilibrium, which could also affect the analytic results. Our study included 95.2% Caucasian for controls, and 95.7% for cases. The joint effects of alcohol use and ADH3 genotypes among Caucasians only were quite similar to the overall result. The presence of at least one ADH*1 allele in these Caucasians was 59.7%, consistent with reports of similar racial groups (6, 13, 14, 16). The genotype distribution among all controls (P = 0.97) and in Caucasian controls only (P = 0.90) satisfied the Hardy-Weinberg equilibrium.

Experiments showed that oral microflora were responsible for a majority of the acetaldehyde production in the oral cavity, especially among tobacco and alcohol users (36-39). Both tobacco use and heavy alcohol consumption independently increased salivary acetaldehyde production by 60% to 75% and tobacco use together with heavy alcohol consumption (>40 g pure alcohol per day) can increase salivary acetaldehyde production by about 100% as compared with nonsmokers and lower dose alcohol consumers. Such evidence suggests that all classes of ADH should have only minor influence on oral acetaldehyde production, and thus would not mainly be associated with an increased risk of HNSCC from alcohol consumption. Increased salivary acetaldehyde production by microorganisms also could be an explanation concerning the much higher HNSCC risk observed among both heavy tobacco and alcohol users from this and other studies (40, 41).

HPV-HR-type infection has been consistently shown to be related to the development of head and neck cancer (19, 41-43). In this study, we again showed that HPV-HR infection was an independent risk factor for HNSCC after adjusting for age, alcohol, tobacco usage, and ADH3 genotype. Although a significant interaction between HPV-HR infection and heavy alcohol use was observed from our previous study (18), ADH3 genotypes did not modify the HNSCC risk associated with HPV-HR infection in this analysis (P = 0.55).

In conclusion, our data did not show that HNSCC risk was significantly associated with ADH3 genotypes, but alcohol consumption, tobacco usage, and HPV-HR infection were three independent risk factors for HNSCC. ADH3 genotypes did not modify the associated risks with these factors. Patients with ADH3^1-2 seemed to have worse survival compared with other two ADH3 genotypes and less recurrence of HNSCC than those with ADH3^1-1.