The p53 Codon 72 Polymorphism and Lung Cancer Risk¹

The p53 tumor suppressor gene, located on chromosome 17p13, is one of the most commonly mutated genes in all types of human cancer (1, 2). Recent studies of the function of the wild-type p53 demonstrated that its antiproliferative effect is mediated by stimulation of a 21-kDa protein (p21cip1/waf1) that inhibits cyclin-dependent kinase activity and, thereby, cell division (3, 4). This negative cell cycle controller effect may explain why the wild-type p53 gene can suppress the transformation of cells by activated oncogenes, thereby inhibiting the growth of malignant cells in vitro and suppressing the tumorigenic phenotype in vivo (5, 6). Analysis of somatic tissue from many human cancers has shown that the wild-type p53 allele frequently is lost and a mutant allele retained,providing a growth advantage for malignant cells (7, 8, 9). The mutation of the p53 gene can damage its DNA-binding properties and transcription factor function, inhibiting its normal function in cell cycle control and in cell proliferation (10).

To date, several polymorphisms in the wild-type p53 gene locus have been described. The codon 72 polymorphism on the 4th exon of the p53 gene, which produces variant proteins with an arginine (CGC) or proline (CCC), has been reported to be associated with bladder and lung cancer (11, 12, 13).

An association of the codon 72 p53 polymorphism with lung cancer susceptibility has been reported by several authors. In one study, Weston et al. (13) reported an increased frequency of the proline allele in adenocarcinomas, but a later study by this same group (14) did not confirm this finding in a different set of cancer cases and controls. The homozygous Pro/Pro genotype was found to be overrepresented in a study of Japanese lung cancer, especially in Kreyberg type I but not in adenocarcinoma (15). More recently, an enhanced risk was reported for African Americans with both the Pro/Progenotype and an early onset of lung cancer (16). A Swedish study has also suggested that the codon 72 alleles may not be functionally involved in lung cancer but, rather, may be a marker in linkage dysequilibrium with other cancer susceptibility sites (17). A Spanish study reported that there was no difference in the prevalence of the codon 72 p53polymorphism between lung cancer cases and controls. However, in that study, the Pro allele of the p53 germ-line polymorphism increased slightly the lung cancer risk for the GSTM1-null genotype among smokers (18). Murata et al. (19) reported that, among 191 lung cancer cases, 115 colorectal cancer patients, and 152 controls, there was a statistically significant difference in genotype frequency only in nonsmokers with lung cancer, with the homozygous Arg/Arggenotype overrepresented in that group.

Hence, the literature to date has not been consistent with respect to the association of codon 72 polymorphism with lung cancer susceptibility. In the present study, we conducted a large case-control study of lung cancer patients and controls and examined the genotype frequency of codon 72 of lung cancer patients and controls, using PCR-based genotyping methods to further evaluate the possible relevance of this polymorphism for lung cancer risk.

The present study was a hospital-based case-control study that included 482 lung cancer patients and 510 controls. Eligible cases included all patients with primary lung cancer (stages I and II) presenting for thoracic surgery at the Massachusetts General Hospital between December 1992 and August 1996. Controls (n = 510) were friends or spouses of the following groups: lung cancer (n =190), other cardiothoracic surgery patients (n = 320).

A detailed interviewer-administered questionnaire was completed for each case and control by a trained interviewer. A modified standardized American Thoracic Society respiratory questionnaire (20)with additions on a detailed occupational and environment exposure history was used. The questionnaire included information on average cigarettes smoked daily for current smokers, years smoked, and time since quitting smoking for ex-smokers. As an indication of cumulative smoking exposure, pack-years were defined as the average number of packs smoked per day multiplied by years smoked. We also obtained information on all of an individual’s job titles, job tasks, years and dates of exposure, and family history of cancer in first-degree relatives.

PCR-RFLP analysis of the codon 72 of the p53 gene originally described by Ara et al. (1) was used to identify p53 BstUI genotypes. The two primers were 5′-TTGCCGTCCCAAGCAATGGATGA-3′ and 5′-TCTGGGAAGGGACAGAAGATGAC-3′. Each PCR reaction mixture (50 μl) contained 10 pmol of each primer,2.0 mm MgCl₂, 200 mm each dNTP, 1 unit of Taq polymerase and 100–300 ng of genomic DNA. Reaction mixtures were preincubated for 5 min at 94°C. PCR conditions were 94C° for 30 s and 55°C for 1 min, followed by 72°C for 1 min for 35 rounds. After confirmation of an amplified fragment of the expected size (199 bp) on an agarose gel, the PCR products were digested with 2 units of restriction enzyme BstUI (New England Biolabs, Beverly,MA) at 60°C for 16 h. DNA fragments were electrophoresed through a 2% agarose gel and stained with ethidium bromide (Fig. 1).

Univariate statistics (χ² and ttests) were used first to compare cases and controls for demographic variables and genotype prevalence. Multivariate logistic regression analysis was performed to assess the association between the p53 polymorphism and lung cancer. Potential confounding factors adjusted for included sex, age (years), race, education level,smoking status (current, ex-smoker, nonsmoker) and pack-years. For the purpose of modeling the association between the p53 variant gene frequency and lung cancer, we compared the homozygous(Pro/Pro) variant and the heterozygous genotype(Arg/Pro) with the homozygous (Arg/Arg)genotype, respectively. Because of low number of homozygous Pro/Pro subjects in some subgroups based on pack-years, we combined the homozygous Pro/Pro variant with heterozygous Arg/Pro as a single group to compare with the homozygous Arg/Arg genotype. In this model, we assumed that risk associated with the Arg/Pro genotype would be intermediate between that of the Arg/Arg and the Pro/Pro genotypes. On the basis of this assumption, we coded the data as follows (using logistic regression model): Arg/Arg genotype = 0; Arg/Pro genotype = 0.5; Pro/Pro genotype = 1. The resulting coefficient yields the risk associated with having the Pro/Pro +Arg/Pro versus Arg/Arg genotype,adjusting for covariates.

The distribution of demographic variables for cases and controls is summarized in Table 1. Men are overrepresented in the cases (54.9 versus 45.1%),with females overrepresented in the controls (54.9 versus45.1%). The mean ages were 65.5 years for cases and 61.6 years for controls. Predictably, cases were significantly more likely to be current smokers (41.9 versus 17.9%), to have smoked more cigarettes per day (28.6 versus 15.4), and to have accumulated more pack-years (57.8 versus 22.6) than the controls. There was no statistically significant difference between cases and controls with regard to education and race.

We examined of the distribution of p53 codon 72 genotypes among controls and lung cancer patients by histological subtypes (Table 2). The frequencies of the three genotypes, Arg/Arg, Arg/Pro, and Pro/Pro, were 46.5, 41.6, and 12.0%, respectively, in controls. The crude genotypic frequencies in the lung cancer patients were similar to those of the controls. When lung cancer cases were stratified by histological subtype, the distribution of the three genotypes in adenocarcinoma patients differed from controls: Arg/Arg, Arg/Pro, and Pro/Pro were 39.3, 44.3, and 16.4%, respectively. There was no statistically significant difference in the prevalence of the polymorphism among squamous cell, large cell, small cell, mixed cell,and bronchoalveolar cell subtypes.

We examined the data further by stratifying values by potentially important confounding factors (Table 3). Within lung adenocarcinoma patients and controls, the variant Pro/Pro genotype was strongly associated with cigarette smoking, with a higher prevalence in current smokers. Among cases, but not controls who smoked, the prevalence of the variant genotype appeared to vary by cumulative cigarette consumption category, with a lower prevalence of the Pro/Pro genotype in light smokers but a higher prevalence in moderate and heavy smokers among adenocarcinoma cases but not controls. In older cases (65–71 years),the prevalence of the Pro/Pro genotype was higher than in controls (not statistically significant). The other variables examined were not statistically significant between cases and controls.

We then examined the association of p53 genotype with lung adenocarcinoma, using a multivariate logistic regression model (Table 4). We first compared the Pro/Pro genotype with the Arg/Arg genotype. The crude OR³for Pro/Pro versus Arg/Arg was 1.61(P = 0.03; 95% CI, 1.04–2.49). After adjustment for age, sex, race, education, pack-years, and smoking status, the same covariates, the OR remained elevated 1.59 (P = 0.06;95% CI, 0.97–2.61). For the purpose of this analysis, the Pro/Pro genotype was combined with the Arg/Progenotype because of low prevalence of the Pro/Pro in subgroups of subjects based on pack-years. The combined variant genotype group had a 1.34-fold higher risk for adenocarcinoma than the genotype Arg/Arg (P = 0.06; 95% CI,0.98–1.83). After adjustment for the same covariates, the OR was 1.45(P = 0.04; 95% CI, 1.02–2.06). Because we assumed the genetic contribution to risk might be less at very high doses of smoking (21), we excluded the group that included the top 15% of subjects based on pack-years (84 pack-years) and repeated the same logistic regression model (data not shown). The OR(both crude and adjusted) rose slightly, with an adjusted OR for the combined variant genotype of 1.50 (95% CI = 1.03–2.18) and an adjusted OR for the Pro/Pro genotype of 1.79 (95% CI =1.06–3.02). To avoid the effect of ethnicity, we ran the logistic model two ways (with all races or with only Caucasians), but the results remained almost the same.

The data in Table 4 display the association of the combined contributions of genotype and smoking exposure to lung adenocarcinoma risk, where the Arg/Arg genotype and no pack-years(nonsmokers) was used as the referent group. Pack-years were divided into quartiles, and the Pro/Pro and Arg/Progenotypes (examined) were one category. As expected, lung adenocarcinoma risk increased with increasing cumulative cigarette dose. In addition, at each level of smoking except the light smokers,the risk associated with smoking was higher for persons with the Pro/Pro + Arg/Pro combined genotype. The group with both the heaviest tobacco smoke exposure and the p53combined Pro/Pro + Arg/Pro genotype had the highest risk of lung adenocarcinoma, roughly 39-fold higher (95% CI,10.76–140.51) than the lowest risk group of nonsmokers who were p53 Arg/Arg. The results revealed the same trend when quintiles or sextiles of pack-years were used. Because the number of cases who had never smoked was too low (n = 17), we combined those who had never smoked with subjects who smoked 1–29 pack-years into a single group and reran the logistic regression model;the results showed the same trend. As expected, the adjusted OR for lung adenocarcinoma by p53 and pack-years decreased,and the 95% CI became tighter (Table 5).

There is an expanding body of literature suggesting that host factors, including genetic polymorphisms, may explain some of the individual differences in cancer occurrence (22, 23). We have shown that the codon 72 germ-line polymorphism(Arg-Pro polymorphism) of the common tumor suppressor p53 gene contributes to susceptibility to smoking-induced adenocarcinoma of the lung. The Pro/Prohomozygous genotype occurred more frequently in adenocarcinomas. The prevalence of the Pro/Pro genotype in adenocarcinoma was higher than that of other genotypes and increased with increasing pack-years.

In this study, we examined the prevalence of p53 codon 72 polymorphisms in a Caucasian group of lung cancer patients and controls. The prevalence of the Pro/Pro genotype in adenocarcinoma cases was statistically different from that of the controls (16.4 versus 12.0%). A Japanese study reported that the prevalence of the Pro/Pro variant in patients with adenocarcinoma was 1.2-fold higher, which was not statistically different from controls (15). Weston et al.(13) reported a prevalence of 26% for their pooled control group and 21% in all types of lung cancer combined. Jin et al. (16) reported that the susceptible Pro/Pro genotype was associated with a 1.6-fold higher risk of all types of lung cancer combined in African Americans and 1.9-fold higher risk in Mexican Americans, with neither reaching statistical significance. Weston et al. (14) later reported no association between the allele frequencies of p53 and susceptibility for all lung cancers in a Caucasian and an African-American population. The discordance in these studies may be the result of choice of controls, the small sample sizes, and hence,the inability to stratify by histological type.

In our study, the frequency of the Pro/Pro genotype in adenocarcinoma was much lower than that of the Arg/Arggenotype in the low pack-years stratum. The prevalence of the Pro/Pro genotype rose linearly with pack-years. Our results show that the frequency of the Pro/Pro genotype was low at lower cumulative cigarette levels, in contrast to several other published reports (13, 14, 15). However, the major reason for the low prevalence of the Pro/Pro genotype in our study was that, at light cumulative levels, the number of subjects was small. Because the homozygous Pro/Pro variant frequency is so low in some pack-years quintiles, we reanalyzed the data after combining the heterozygous and homozygous variants and found a statistically significant association with adenocarcinoma. Furthermore, we found that the most important confounding factor for the study appears to be pack-years, which must be adequately adjusted in the analysis of association between p53 genotype and lung adenocarcinoma. The association of the p53 codon 72 variant genotype with increased lung adenocarcinoma risk was modified by gender, race, age,education, smoking status, and pack-years and remained statistically significant after we adjusted for these factors.

As expected, the OR for lung adenocarcinoma increased with dose of cigarette tobacco smoke for both genotypes. In addition, at each level of smoking except for light smokers, subjects with the p53combined variant genotype experienced higher risk of adenocarcinoma than subjects who had the p53 Arg/Arg genotype. Individuals with the heaviest tobacco smoke exposure and p53combined genotype had the highest risk of lung cancer, roughly 38-fold higher than the lowest risk group of nonsmokers who were Arg/Arg. These data suggest that the presence of the p53 gene product exerts a protective effect for smoking-induced lung cancer and that the presence of one variant allele alters this product. Thus, the modifications by p53 and pack-years work independently and increase the risk of the susceptible genotypes for lung adenocarcinoma. Considering the biological role of p53 in carcinogenesis, one of the most plausible interpretations for the distribution of the Pro/Pro genotype is that this heritable polymorphism imparts high risk of developing adenocarcinoma.

Genetic differences in risk may be smaller at high loads of carcinogen exposure, when environmental influences may overcome the association with a genetic predisposition (21, 24, 25). However, in our study, we found that there was little difference between the susceptible genotypes and pack-years at higher doses of smoking.

The reason for the observed tissue-specific difference in the risk conferred by the germ-line p53 polymorphism is unknown. There have been changing trends in the occurrence of adenocarcinoma of the lung over the past two decades. It has been hypothesized that changes in cigarette composition and smoking behavior have produced a histological shift over the past decades. The proportional decrease in lung small cell carcinomas may be related to a reduction in exposure of the central bronchi to polyaromatic hydrocarbons, and the dramatic rise in lung adenocarcinoma may be related to increased exposure of the peripheral lung to tobacco-specific nitrosamines. The hypothesis suggested by Kawajiri et al. (23) is that different carcinogenic processes are involved in the genesis of various tumor types because of the presence of functionally different p53 alleles (Pro- or Arg-type). The functional difference of the p53 polymorphism at codon 72 has been reported (26). A p53Arg/Arg genotype induces apoptosis with faster kinetics and suppresses transformation more efficiently than the p53Pro/Pro genotype.

Our observations provide evidence that cigarette smoking should be appropriately controlled in the analysis of p53 codon 72 polymorphisms and lung cancer and that when adjusted for smoking, the presence of the codon 72 variant allele is associated with increased lung adenocarcinoma risk. The previously reported inconsistent findings for an association of the codon 72 variant in the p53 gene with lung cancer between light and heavy smokers could be attributable to confounding factors, to sample size limitations, and to various choices of controls.

Our study has several potential limitations that might have influenced our results. First, our controls come from friends or spouses referred by lung cancer cases or by cardiothoracic patients. A necessary condition for the validity of our OR estimates of the genotype association is that, conditional on the variables controlled in the analysis, (a) the genotype distribution of the two kinds of controls is the same and (b) represents the distribution in the source population. To evaluate whether the first statement is satisfied, we tested for differences in the genotype frequencies of the two control groups, finding no significant differences(χ² = 0.4; P = 0.5). Given that we found no significant differences in genotype frequencies and that the frequency in the collapsed control group was similar to the estimates observed in past studies of Caucasian populations, it seems reasonable to assume that the two control groups represent a common population. Selecting controls among friends or spouses of lung cancer patients or other surgical patients might have introduced bias by two different mechanisms (27). First, the distribution of exposures among gregarious subjects, i.e., subjects who tend to be named by more than one other person as controls, might be different from the distribution among non-gregarious subjects and,therefore, it might not represent the distribution in the source population that includes both types of subjects. Second, exposure among friends and spouses of patients might be similar to the exposure among the patients themselves, and thus controls may not represent the distribution in the source population. These two factors are likely to be relevant for variables such as age, gender, smoking, or dietary habit. However, within categories of these variables and among the same racial or ethnic group, the genetic polymorphisms in p53 are unlikely to be associated with being gregarious or with having a friend or spouse with lung cancer. Therefore, the genotype-disease association is unlikely to be affected substantially by this type of bias.

On the other hand, the reported estimates for the effect of pack-years on lung cancer are likely to be underestimated. Moreover, our case population includes only surgically treated lung cancer patients(stages I and II), and those with more advanced tumors not eligible for surgery were not included. Thus, differential selection of cases with respect to the genotype could occur if the genotype is related to the stage at which the lung cancer is detected. To our knowledge, one Chinese study reported that patients with the Pro/Progenotype were more than five times more likely to die at early postoperative stages than those with the Arg/Pro genotype (28). This may introduce overestimation of the genotype OR because poorly differentiated tumors and tumor with metastasis are more likely to become inoperable. Finally, ethnicity could be a confounder of the OR for the genotype-disease association. However, in the context of our study, this type of confounding is unlikely to be important because we got almost the same results when we included all races or just Caucasian subjects in the logistic regression model. Therefore,ethnicity is unlikely to be strongly associated either with the genotype, as indicated above, or with lung adenocarcinoma risk.

Our analysis of the possible association between the p53genotype and pack-years revealed that the susceptible genotype is more clearly associated with an increased risk of adenocarcinoma among smokers. This finding is in agreement with some reports (29, 30) and is discordant with others (15), and suggests that genetic susceptibility may play a role in certain medium exposure conditions but may be overpowered by heavy exposure to the carcinogens in high doses of tobacco smoke. Seemingly moderate genetic risk, combined with environmental exposures, can determine an important proportion of lung cancers. However, further study in defined light smoking subgroups is required to substantiate the findings that the Pro/Pro genotype predisposes individuals to adenocarcinoma of the lung.

We gratefully acknowledge the assistance of Linda Lineback,Barbara Bean, Li Su, Lilian Xu, Zheng-fa Zuo, and Lucy-Ann Principe-Hasan.