Alcohol Consumption and Cigarette Smoking Increase the Frequency of p53 Mutations in Non-Small Cell Lung Cancer¹

Lung cancer is the leading cause of cancer deaths among males and females in the United States (1). Epidemiological evidence suggests that most cases of lung cancer are directly attributable to cigarette smoking (2). Only 5–10% of all of the lung cancer cases occur in patients without a prior history of cigarette smoking (3). Compared with nonsmokers, smokers have a 10-fold greater risk of dying from lung cancer, and in heavy smokers this risk increases to 15- to 25-fold (2). In contrast,the relationship between alcohol consumption and lung cancer is less well defined. Alcohol consumption is closely related to cigarette smoking, which makes it difficult to assess the independent risk of alcohol on lung cancer alone (4). Furthermore, minor alleles in the D2 dopamine receptor gene, previously implicated in chronic alcoholism, have recently been reported to play a role in determining nicotine addiction (5). Several recent studies have controlled for tobacco use and have demonstrated an increased incidence of lung cancer in heavy alcohol drinkers, with one study(4) reporting an adjusted odds ratio of 1.8 for alcohol drinkers (one or more drinks of alcohol per day) compared with infrequent drinkers (4, 6, 7, 8).

Molecular epidemiological studies have begun to link specific environmental carcinogens with mutational events in cancer progression(9, 10). The p53 gene has been a common target of these mutational studies because it is the most frequently mutated gene in human cancer (10). These studies have also shown that the p53 mutational spectrum of smoking-related neoplasms is distinctly different from that present in non-smoking-related cancers. Recently, benzo(a)pyrene diol epoxide, the active form of benzopyrene and a potent carcinogen in cigarette smoke, has been shown to bind preferentially to select regions of the p53 gene (11, 12). Formation of these benzopyrene adducts likely reflects the high frequency of certain p53 mutations in smoking-associated tumors and correlates with several known mutational hotspots in the p53 gene(12). In addition, exposure to other environmental agents such as ethanol may further augment the mutational effect of tobacco smoke (13). In head and neck squamous cell cancer, a cancer long-associated with both tobacco and alcohol use in epidemiological studies, the frequency of p53 mutations and overexpression is significantly higher in patients who smoked cigarettes and drank alcohol than in patients who only smoked cigarettes (13, 14). Although heavy alcohol consumption increases lung cancer risk, the effect of cigarette smoking and ethanol intake on the frequency of p53 gene mutations in lung cancer is not known. In the present study, we studied the effect of alcohol consumption and cigarette smoking on mutation of the p53tumor suppressor gene in patients with non-small cell lung cancer.

One hundred forty-five consecutive patients undergoing surgical resection of non-small cell lung cancer at The Johns Hopkins Hospital or the Johns Hopkins Bayview Medical Center were entered into a prospective study examining the prognostic significance of p53 mutations in non-small cell lung cancer. For the present analysis, which examined the effect of alcohol consumption and smoking on p53 mutations, we elected to continue accruing patients until at least 10 nonsmokers with lung cancer had been identified. Five patients were excluded because a smoking history could not be obtained(three patients) or they had received preoperative chemotherapy (two patients). Of the remaining 140 patients, 130 (93%) had a history of cigarette smoking.

Demographic data were collected from patient interviews and a review of the hospital charts and from The Johns Hopkins Hospital Tumor Registry. Pathological stage was determined using the revised International System for Staging Lung Cancer (15). Tumors from patients with squamous cell carcinoma or adenocarcinoma of the lung and a history of a second primary squamous cell or adenocarcinoma at a different site underwent careful pathological review of both cancers to exclude a pulmonary metastasis from a nonlung primary tumor. Four patients, in whom the possibility of metastases could not be excluded, were not included in the study. Demographic data were compiled by staff members who had no knowledge of the mutational analysis of the p53 gene in the patient’s tumor. This research protocol was approved by the Joint Committee on Clinical Investigation of The Johns Hopkins School of Medicine, and informed consent was obtained from all of the patients.

The history of alcohol consumption and cigarette smoking was obtained from patient interview and clinic notes or from the tumor registry. Patients were classified as alcohol drinkers if they consumed one or more drinks (one drink being defined as 1 ounce of 86-proof hard liquor, one 3.6-ounce glass of wine containing 12% alcohol, or one 12-ounce can of beer) per day on average during the 20 years before being diagnosed with lung cancer, as occasional drinkers if they consumed alcohol but less than one drink per day on average, and as nondrinkers if they consumed less than one drink per day or abstained from alcohol (13). Nonsmokers were defined as patients who had smoked <100 cigarettes in their lifetime (4). All of the smokers had at least a 10-pack-year history of smoking. Lifetime cigarette consumption was quantitated by averaging the number of packs smoked over the number of total smoking years (pack-years). Smoking patients were also stratified by the number of years they had quit smoking in 5-year intervals before the diagnosis of lung cancer or whether they were still smoking cigarettes.

Portions of the patient’s primary lung cancer were collected and promptly frozen at −80°C. Sections of the primary tumor were stained with H&E and examined by light microscopy. Tumors with low neoplastic cellularity (<70%) were further microdissected to remove contaminating normal cells, and DNA was isolated. A 1.8-kb fragment of the p53 gene (exons 5 through 9) was amplified from the primary tumor from the first 95 patients with a history of cigarette smoking and in all of the 10 nonsmokers. Exons 5 through 9 of the p53 gene were sequenced directly using cycle sequencing(16). In addition, the same DNA isolated from all of the 105 tumors was further sequenced (exons 2 through 11) using the GeneChip p53 assay (Affymetrix Inc., Santa Clara, CA) as described previously (17). Discrepancies between these two techniques (exons 5 through 9) were resolved by a third assay(oligonucleotide-specific hybridization) as described previously(17).

Clinical and pathological characteristics between smokers and nonsmokers were compared using χ² or Fisher’s exact test for categorical variables and Wilcoxon’s nonparametric rank test for continuous variables. The association between multiple clinical and pathological variables and mutation of the p53gene were examined by logistic regression. Statistical analysis was performed using SAS system software.

One hundred and five patients with non-small cell lung cancer undergoing surgical resection were finally included in this study. This group included the 10 nonsmokers and 95 smokers whose tumors had previously undergone p53 sequence analysis. A history of alcohol intake was available from 97 (92%) of the 105 patients. Forty-four % of these patients had an alcohol intake of one or more drinks per day. Twenty-one % of the patients drank alcohol occasionally (less than one drink per day), and 35% abstained from alcohol. No significant differences in age, gender, tumor cell type,tumor stage, tumor size, or tumor grade were observed among patients who drank more than one drink per day and those who drank less than one drink per day.

Clinical characteristics for smokers (n = 95)and nonsmokers (n = 10) with non-small cell lung cancer are shown in Table 1 . Nonsmokers who developed non-small cell lung were significantly older,more often female, and more likely to develop adenocarcinoma and had smaller cancers than smokers. Nonsmokers with primary lung cancer were also more likely to have a prior history of cancer [pancreas, breast,colon, lymphoma, prostate, melanoma (excludes nonmelanoma skin cancer)] than patients with a history of cigarette smoking (60 versus 24%; P = 0.02). A trend toward earlier stage cancers was also observed in the nonsmokers.

Clinical characteristics for alcohol drinkers (n= 43) and nondrinkers (n = 54) with non-small cell lung cancer are shown in Table 2 . No significant differences were observed in any of these clinical and pathological characteristics among the two groups.

Less than one-half of the smokers [46 (48%) of 95] continued to smoke cigarettes at the time of cancer diagnosis. The remaining 49 patients (52%) had quit smoking at least 1 year before being diagnosed with lung cancer. Eight of these patients had quit within 5 years, and an additional 11 patients had quit within 10 years of the diagnosis of lung cancer. Thirty patients had quit smoking more than 10 years before being diagnosed with lung cancer. The duration and intensity of exposure (in pack-years) was available from 74 of the 95 smokers. Thirty-seven patients (50%) smoked less than 50 pack-years, 31 patients (42%) smoked between 50 and 100 pack-years, and only 6 patients (8%) had a >100-pack-year smoking history. In addition,median lifetime cigarette use was similar between alcohol drinkers and nondrinkers (50 versus 40 pack-years; P= 0.28).

p53 sequence analysis was performed on tumor samples from 105 patients with non-small cell lung cancer using direct sequencing and the p53 GeneChip. At least one mutation of the p53 gene was identified in 56 (53%) of 105 tumors. A comparison between these two sequencing techniques has been reported previously (17). One tumor contained two p53mutations for a total of 57 mutations in 105 non-small cell lung cancers (Table 3). The most common p53 mutations were GC→TA transversions and GC→AT transitions. Only one p53 mutation was present among the 10 nonsmokers, which precluded any comparison of mutation spectra among nonsmokers and smokers. GC→AT transitions were more common in alcohol drinkers [6 (14%) of 43] than in nondrinkers [2(4%) of 54]; however, this difference failed to reach statistical significance (P = 0.12). No other statistically significant differences were observed among the mutation spectrum from nondrinkers and the mutation spectrum from patients consuming at least one drink per day (data not shown). Thirty-four %of the p53 mutations included frame-shifts, deletions,splice sites, or termination codons. These mutations would be predicted to encode truncated p53 proteins that immunohistochemical analysis usually fails to detect.

Clinical variables associated with p53 gene mutations were determined using logistic regression analysis (Table 4). p53 mutations were detected in 74% (31 of 42) of the cancers from patients with a history of alcohol use, in 35% (7 of 20)of the patients with occasional alcohol use (less than one drink per day), and in 40% (14 of 35) of the patients who abstained from alcohol. Because the frequency of p53 mutations among occasional and never drinkers was similar (P= 0.71), we pooled these groups together for further comparisons. The odds ratio of having a p53 mutant tumor for patients drinking one or more drinks per day versus less than one drink per day was 4.6 (95% confidence interval, 1.9–11; P = 0.001).

A significant association between cigarette smoking and mutation of the p53 gene was also observed. p53 mutations were present in 58% of the tumors from patients with a history of cigarette smoking (55 of 95), whereas only 10% of the nonsmokers (1 of 10) had tumors with a p53 mutation. The odds ratio for having a tumor with a p53 mutation for smokers versus nonsmokers was 12.4 (95% confidence interval, 1.5–102; P =0.02). p53 mutations were also detected significantly more frequently in squamous cell cancers than in primary adenocarcinoma of the lung (67 versus 38%). The odds ratio of having a p53 mutation in squamous cell cancers versusadenocarcinomas was 3.2 (95% confidence interval, 1.4–7.5; P = 0.005). Patient age, gender, tumor stage,total smoking exposure in pack-years or interval since quitting smoking, and a history of other primary cancers were not significantly correlated with the presence of p53 mutations (Table 4).

Alcohol use, smoking history, and cell type were then examined in a logistic regression model to define independent variables associated with mutation of the p53 gene in non-small cell lung cancer. A history of alcohol intake (≥1 drink per day) was the variable with the strongest statistical association with p53 mutations(odds ratio, 3.9; 95% confidence interval, 1.5–10; P < 0.005). Cell type was also significantly associated with p53 mutations with squamous cell cancers having a greater frequency of mutations than adenocarcinomas (odds ratio, 2.9; 95% confidence interval, 1.2–7.2; P< 0.02). In the multivariate analysis, smoking was associated with the greatest overall risk (odds ratio, 6.5) of harboring a p53 mutation in non-small cell lung cancer. However, smoking was also associated with a large confidence interval, in large part attributable to the low frequency of nonsmokers in the population and the low incidence of p53 mutations and squamous cell histology in this group of patients (95% confidence interval,0.73–58; P = 0.09). Therefore, we further examined the role of smoking in mutations of the p53 gene in patients with only adenocarcinoma of the lung (9 of 10 nonsmokers developed adenocarcinoma). Mutations of the p53 gene were present in 46% of smokers with adenocarcinoma (21 of 46) but in none of the 9 nonsmokers with adenocarcinoma of the lung(P = 0.016).

We also examined the quantitative effect of alcohol consumption and cigarette consumption (in pack-years) on the p53 mutation rate. Smoking intensity was weakly correlated with alcohol intake. The median number of pack-years smoked (50 pack-years) among the alcohol drinkers (≥1 drink per day) was not significantly greater (Pearson correlation coefficient, 0.10; P < 0.39)than the median number of pack-years smoked (40 pack-years) among the nondrinkers (≤1 drink per day). The frequency of p53mutations was significantly greater in alcohol drinkers than in nonsmokers and nondrinkers, and this increase was present in both light(<50 pack-years) and heavy (≥50 pack-years) smokers (Table 5). There was also a trend toward an increased frequency of p53 mutations in light smokers (<50 pack-years) who consumed alcohol when compared with light smokers who consumed less than one drink per day (82 versus 53%; P = 0.07).

Alcohol consumption had a greater effect on the p53mutation rate than pack-years smoked; therefore, we examined the p53 mutation rate in patients without a history of tobacco or alcohol use, in all of the smokers without a history of alcohol intake, and in all of the smokers with a history of alcohol use. Seventy-six % of the cancers from patients who drank alcohol (≥1 drink per day) and smoked cigarettes contained p53 mutations(31 of 41). In contrast, only 42% of tumors from nondrinkers (<1 drink per day) who smoked contained a p53 mutation (20 of 48), and only 14% of the nondrinkers and nonsmokers had a tumor with a p53 mutation (1 of 7). Only one nonsmoker drank alcohol (wild-type p53), and eight patients with an unknown history of alcohol consumption (including two nonsmokers) were excluded from the combined logistic regression analysis. The odds ratio for tobacco use but no alcohol use versus no alcohol or tobacco use was 4.3 (95%confidence interval, 0.8–38.4; P = 0.19),alcohol and tobacco use versus no alcohol or tobacco use was 18.6 (95% confidence interval, 2.0–174; P =0.01), and alcohol and tobacco use versus tobacco use but no alcohol use was 4.3 (95% confidence interval, 1.7–10.8; P < 0.01; Fig. 1).

Distinct mutation patterns in the p53 gene have been linked with certain environmental carcinogens (9, 10). In the present study, p53 gene mutations were present in 53%of the 105 patients with non-small cell lung cancer and were associated with alcohol consumption and tobacco smoking, in addition to tumor cell type.

Alcohol consumption is an independent risk factor for cancers of the esophagus, pharynx, larynx, and liver (18). A strong synergism between alcohol and smoking exists for cancers of the head and neck (18). Ethanol is not itself genotoxic, and neither it nor alcoholic beverages are tumorigenic in animal models(18). However, both tobacco and alcohol use are independently associated with p53 mutations in head and neck squamous cell cancers (13). Several mechanisms have been postulated to explain how alcohol may potentiate the effect of environmental carcinogens. In animal studies, alcohol has been demonstrated to suppress hepatic clearance of low molecular weight nitrosoamines present in cigarette smoke by inhibiting several isoforms of the cytochrome P450 superfamily (18). This inhibition then leads to an increase in the dose-delivery of nitrosoamines to posthepatic tissues and an increase in the formation of DNA adducts(18). Furthermore, acetaldehyde, the initial breakdown product of ethanol inhibits the DNA repair enzyme O⁶-methylguanine DNA methyltransferase(19). If left unrepaired by O⁶-methylguanine DNA methyltransferase, O⁶-methylguanine may pair with thymine during DNA replication. In the present study, a nonsignificant trend toward an increase in GC→AT transitions was observed among alcohol drinkers. The present study demonstrates that alcohol may also play an important role in the pathogenesis of lung cancer.

Prior studies that examined the effect of cigarette smoking on mutations of the p53 gene in various cancer types have yielded conflicting results (20, 21, 22, 23, 24, 25, 26, 27, 28, 29). Several studies have used immunohistochemical analysis to evaluate the p53 protein; this method has substantial false-negative and false-positive rates when compared with direct gene sequencing (21, 22, 23). In the present study, 34% of the 57 mutations would have resulted in a truncated p53 protein, which usually do not stain with labeled anti-p53 antibodies. Several reports have correlated the frequency of p53 mutations with lifetime tobacco exposure and have not included or have included very few nonsmokers to achieve a statistically meaningful result (24, 25, 26). Other investigators have not demonstrated a statistically significant relationship between cigarette smoking and mutation of the p53 gene but have reported a relatively low frequency of p53 mutations in cigarette smokers (27, 28). We identified p53 mutations in 58% of the smokers in the current study using two complementary sequencing methods. Previous work in our laboratory (17) has demonstrated that the use of direct sequencing and the p53 GeneChip assay will detect more mutations than the use of either technique alone.

The pattern of mutations in the p53 gene varies among cancer types and often provides clues to the pathogenesis of the tumor(10). GC→TA transversions were the most common mutation observed in the 95 smokers in this study and have also been commonly reported in other smoking-related cancers (esophagus, head and neck;Refs. 10, 13). These mutations are the hallmark of mutagenesis involving certain types of polycyclic aromatic hydrocarbons, such as benzopyrene (10, 12). Mutations at CpG dinucleotides also occurred commonly in our series (32%) and have been reported in from 9 to 35% of non-small cell lung cancer patients(10, 25). Mutations at CpG dinucleotides are generally presumed to be secondary to the endogenous deamination of methylated cytosine, especially in colon cancer, in which 55% of mutations occur at CpG sites (10, 30). However, preferential binding of several carcinogens, including benzo(a)pyrene diol epoxide,has recently been demonstrated at methylated CpG sites corresponding to mutational hotspots in the p53 gene (30). Benzo(a)pyrene diol epoxide-induced DNA damage at these CpG sites is poorly repaired and likely explains the high incidence of mutations at these sites in non-small cell lung cancer(31).

Non-small cell lung cancer occurs infrequently in nonsmokers and has distinct clinical, pathological, and genetic features from cancer in patients with a history of cigarette smoking. One limitation to this study is the low number of nonsmokers with non-small cell lung cancer available for analysis. Adenocarcinoma is more common in nonsmokers,and, in our series, 90% of the nonsmokers had adenocarcinoma(10, 32). p53 mutations in non-small cell lung cancer are also more common in large-cell and squamous cancers than in adenocarcinoma. However, the frequency of p53 mutations in smokers with adenocarcinoma was still significantly higher than in the nonsmokers with adenocarcinoma (46 versus 0%). In the current series, too few mutations occurred in the nonsmokers to analyze the p53 mutational spectrum in this group. However, other series have reported an increased frequency of GC→AT transitions and deletions in nonsmokers (27). An increased incidence of K-ras mutations and chromosomal loss at the FHITgene locus has also been associated with cigarette smoking (27, 33).

Loss of p53 tumor suppressor gene function allows the evolution of a clonal population of cells with a selective growth advantage that may result in cancer progression. Aberrant p53 expression has been documented in preneoplastic bronchial lesions and may be an important step in the progression of non-small cell lung cancer (34). Moreover, p53 mutations may be associated with more aggressive behavior and shorter survival in non-small cell lung cancer (35). Previous epidemiological data have strongly linked lung cancer with cigarette smoking and to a lesser degree heavy alcohol consumption. The present study did not examine the effect of cigarette smoking or alcohol consumption on the overall risk of developing lung cancer. However, our results demonstrate that alcohol consumption and cigarette smoking together increase the frequency of p53 gene mutations in patients with non-small cell lung cancer more than that observed with cigarette smoking alone. These data suggest that, although not necessarily mutagenic by itself, alcohol may potentiate the effects of other known carcinogens, thereby influencing the development and behavior of lung cancer.