Relationship of H-ras-1, L-myc, and p53 polymorphisms with lung cancer risk and prognosis.

Proto-oncogenes (H-ras-1 and L-myc) and tumor-suppressor gene (p53) loci have been implicated in lung carcinogenesis. DNA restriction fragment length polymorphisms at these gene loci are being evaluated in a case-control study as markers predictive of risk for cancer or of prognosis when cancer is present. The cases and controls had a cigarette-smoking history of 40 or more pack years or other abnormalities in pulmonary function tests, their ages were closely matched (64 years for cases and 61 years for controls) and the ratio of Caucasians to African Americans was close to unity (cases, 0.95:1.00, controls, 1.00:0.88). The H-ras-1 gene contains an insertion deletion polymorphism. Inheritance of rare H-ras-1 alleles, defined by MspI digestion, confers a relative risk for lung cancer of 2.0 (95% confidence interval, 0.5-7.3) for Caucasians and 3.2 (0.9-11.6) for African Americans (74 cases, 67 controls). The L-myc gene sequence has a restriction site (EcoR1) polymorphism between the second and third exons. Inheritance of restriction site-present alleles was reported to confer poor prognosis (presence of lymph node metastases) in Japanese lung cancer patients. This hypothesis was tested in both case-control study subjects (56 cases, 55 controls) and additional surgical cases (40), but no evidence was found to support the hypothesis in the U.S. population. The p53 gene is a tumor-suppressor gene that can encode either a proline or an arginine in the 72nd residue. No associations was found between the minor allele (proline) and diagnosis of lung cancer (76 cases, 68 controls).(ABSTRACT TRUNCATED AT 250 WORDS)


Introduction
Epidemiological and pharmacogenetic studies have postulated the existence of inherited predisposition for lung cancer (1,2). Furthermore, familial studies suggest a genetic component to this disease, in spite of its now well documented environmental etiology [tobacco smoking (3)]. Few useful genetic markers exist, however, that allow assessment of individual risk for of human lung cancer. The biomarkers evaluated in the current study consist of proto-oncogenes and tumor-suppressor genes that are implicated in the genesis human lung cancer: H-ras-1 (4), L-myc (5), and p53 (6). A growing number ofinvestigations by others have attempted to evaluate various potential biomarkers of individual cancer risk, but most of these suffer from inadequate study design. With respect to lung cancer, case-control studies are of limited value if age and tobacco smoking history are not controlled. The studies reported here were undertaken in the context of a formal case-control design that accounts for age, tobacco smoking history, gender, ethnic background, occupational history, and familial history of cancer.

Accrual of Study Subjects
The design of this epidemiological case-control study of lung cancer is described in detail elsewhere (7). Briefly, patients with histologically confirmed lung cancer who had not received radiation or chemotherapy were recruited at the University of Maryland and Baltimore Veterans Administration Hospitals between 1985 and 1989. Histological diagnosis of lung cancer was confirmed by pathological review. Two control groups were recruited: chronic obstructive pulmonary disease (COPD) patients and patients with cancers at anatomical sites other than the lung and bladder. The COPD patients all had a clinical diagnosis and demonstrated abnormalities in pulmonary function tests (forced expiratory volume in 1 sec [FEV1] of less than 75% predicted normal and/or FEV1/forced vital capacity less than 75% predicted) and/or a cigarette smoking history of 40 pack-years or more. The second control group was recruited from among cancer patients with a variety of malignancies including cancers of the colon, esophagus, stomach, and breast and melanoma.
Study recruits were screened for eligibility or exclusion. Exclusion criteria included: treatment in an intensive care unit, blood pressure < 100/60 mm Hg, inability to take oral medication or to be interviewed, general anesthesia within the last 5 days, severe renal or liver disease (creatinine > 4.0 mg/mL or total bilirubin > 3 IU, or serum glutamicoxaloacetic transaminase or serum glutamic-pyruvic transaminase > 300 IU), previous diagnosis ofseparate primary malignancy other than basal-cell carcinoma, inability or refusal to give informed consent, or physician refusal. An interview was administered to eligible subjects by a trained study nurse, and data were collected on sociodemographic and anthropomorphic characteristics, recent and remote tobacco use, personal medical history, usual and recent diet, current medications, familial history of cancer, alcohol use, and occupational and residential history. Medical records were also reviewed to abstract other relevant information.
Following electrophoresis, gels were treated with alkali (400 mM NaOH; 600 mM NaCl; 25 min) and neutralized (1.5 M NaCl; 500 mM Tris-HCl; pH 7.5; 30 min), and DNA was immobilized by capillary transfer onto nylon membrane support matrices. Samples were hybridized to the full-length, 6.6kb, 32P-labeled, human H-ras-1 BamHIfragment under stringent conditions (NaCl, 100 mM; SDS, 0.5%; dextran sulfate, 10%; 65°C; 16 hr). The membranes were washed under increasingly stringent conditions (final conditions: 37.5 mM NaCl; 3.8 mM HOC [COONa] [CH3COO*Na]2; 0.1% SDS; 65°C; 30 min), and X-ray films (Eastman Kodak Company, Rochester, NY) were used to detect hybridization of the radioactive probe. Allelotyping was based on the original system reported by Krontiris et al. (9); in which common alleles are designated by an integer (i.e., al, a2, a3, and a4), and other alleles are designated by decimals with respect to the closest smaller common allele (e.g., al.1, al.2 for alleles larger than al but smaller than a2, and a2.1, a2.2 for alleles larger than a2 but smaller than a3). Similarly, al.25 is determined to be larger than al.2 but smaller than al.3. Authentic, well-characterized alleles from original studies were frequently used in the analysis of test samples.

Polymerase Chain Reaction-Restriction Enzyme Analysis for L-myc
Polymerase chain reaction-restriction enzyme (PCR-RE) analysis for L-myc was carried out on 56 lung cancer patients (clinical diagnosis), 36 COPD patients, and 19 other cancer patients who had been enrolled in the casecontrol study. Since the allelic frequencies at the L-myc locus were found to differ by race (African American versus Caucasian), further samples were obtained (from trauma victims; 16 Caucasian and 24 African American) to test the hypothesis of an association between race and allelic frequency. Additional lung cancer cases were also obtained at surgery (27 Caucasian and 23 African American) to address rigorously the question of an association between the presence of metastatic disease and the distribution of L-myc alleles in lung cancer. Genomic DNA was amplified using primers that flank the polymorphic EcoRI site in the L-myc gene (5'AGTTCACTCACAGGC-CACAT3'-* and +-5'TGCATATCAGGAAGCTTGAG3').
The PCR was performed using a DNA thermal cycler (Perkin-Elmer Cetus, Emeryville, CA), amplified DNA segments were digested with EcoRI and subjected to electrophoresis in polyacrylamide gels (8%) (New England Biolabs, Beverly, MA). The gels were treated with an ethidium bromide solution and examined under ultraviolet light. A representative subset of samples was analyzed by Southern hybridization with the 1.9-kb EcoRV/SmaI fragment of an L-myc cDNA clone to corroborate the PCR-RE analyses (10).

Polymerase Chain Reaction-Restriction Enzyme Analysis for p53
An easy PCR-RE genotyping test was developed on the basis of previous reports (11,12). The method used here, however, included an important modification: the primer flanking codon 72 in the 5' region contained a single basepair mismatch (resulting in the formation of a new AccII restriction site between exon 2 and exon 3; that is, an A to C change at nucleotide position 11863 of p53) to control for completion of digestion. DNA samples were amplified for 35 cycles using a standard protocol (13) either with primers 1 and A (below) or by a hemi-nested strategy using primers 1

Statistical Analysis
Arithmetic means, standard errors, correlation coefficients, Student's t-test stepwise and linear regression were performed using the SAS statistical package (15).

Analysis of the H-ras-1 Polymorphism
Twenty-three discrete restriction fragments (MspI) were found in 141 study subjects for the H-ras-1 protooncogene. Resolution of 13 of these restriction fragments is shown in Figure 1, after hybridization with an H-ras-1 clone. Common, intermediate, and rare alleles were determined according to the criteria defined above. The allelic distribution of H-ras-1 MspI fragments was different for African Americans and for Caucasians residing in the United States. Notably, an allele designated a3.5, a variant of the common a4 allele, was common among African Americans. An excess of rare alleles was observed in lung cancer patients compared to both COPD and nonpulmonary cancer controls (Table 1). In Caucasians, rare alleles were found to be approximately twice as frequent in lung cancer patients as in the COPD or nonlung cancer control groups (x2 = 6.5, p < 0.05, degrees of freedom (df = 2); the same comparison in African Americans showed a 3-fold excess of rare alleles in the cases (X2 = 26.4, p < 0.0001, df = 2). Similar findings were made when lung adenocarcinomas were excluded (X2 = 9.9, p < 0.01, df = 2, for Caucasians; x2 = 25.9, p < 0.0001, df = 2, for African Americans).
Since allelic frequency could differ by race (8), Caucasians and African Americans were analyzed independently for relative risk for lung cancer ( Table 2). Although small numbers in each group hamper the analysis, an increase in the number of rare alleles for lung cancer (excluding adenocarcinoma) was observed in African Americans. Crude odds ratios for risk based on the number of rare alleles show elevated risk in all individuals with rare alleles and increasing risk based on the number of rare alleles in the four comparisons considered: Caucasians and African Americans, considering all lung cancer cases and excluding cases with adenocarcinoma of the lung. The trend is statistically significant only for nonadenocarcinoma of the lung in African Americans (X2 =  3.9, df = 1, p = 0.05). Adjustment of these results for other risk factors for lung cancer evaluated in this study, for example smoking (pack years), age, family history of lung cancer in first-degree relatives, and the debrisoquine metabolic phenotype, revealed no important confounding factors.

Analysis of the L-myc Polymorphism
Southern blot analysis of human genomic DNA samples that had been digested with EcoRI revealed a DNA restriction fragment length polymorphism consisting of two alleles at the L-myc locus ( Fig. 2A-C). The PCR was also used to reveal this polymorphism in genomic DNA samples when primers that flank the EcoRI restriction site were used to amplify a DNA segment of267 base pairs ( Fig. 2D-F). Initially, the two methods were applied to the same DNA samples, and it was found that the PCR and Southern hybridization of EcoRI-digested materials gave consistent results (Fig. 2).
A highly significant difference in allelic distribution between African Americans and Caucasians was found (Table 3); the distributions for all subjects was consistent with the Hardy-Weinberg equilibrium (16). The casecontrol data show that in the Caucasian population the ratio of restriction site-absent alleles (L-alleles) to restriction site-present (S-alleles) was 0.509:0.491, whereas in the African American population the distribution was 0.277:0.763. Similarly, among a group of trauma victims, allele frequencies were 0.562:0.438 and 0.145:0.855 in Caucasians and African Americans, respectively.
In the lung cancer case-control study, PCR-RE or Southern hybridization analyses for L-myc were con-  aDNA restriction fragment length polymorphism (RFLP) pattern determined by Southern blot or polymerase chain reaction-RFLP analysis, L = 10 kb and S = 6.6 kb EcoRI fragments of the L-myc proto-oncogene.
ducted on a total of 111 subjects; the allelic distributions are given in Table 4. When race and other variables were not controlled for, there was no association between allelic frequency and lung cancer (likelihood ratio x2 = 1.14, p = 0.56, df = 2). In addition, the allelic frequencies were similar in each group (likelihood ratio x2 = 0.12, p = 0.94, df = 2, n = 55). When a variety of other study variables were entered into a multivariate statistical model to predict lung cancer, neither race nor the number of restriction sitepresent alleles was found to be associated with cancer diagnosis. The effect of a number of study variables on the number of alleles was also tested using a multivariate linear regression model: neither lung cancer diagnosis nor any other study characteristic tested predicted the presence of a specific allelotype, with the exception of race (F = 12.79, p = 0.0005). Thus, African Americans are more likely to have restriction site-present alleles than Caucasians. The distribution of L-myc alleles was also unrelated to histologic subtype of lung cancer (data not shown). There does not appear to be any relationship between clinical stage of lung cancer and L-myc alleles in either African Americans or Caucasians. Similarly, within a group of surgically diagnosed Caucasian lung cancer patients with orwithout lymph node metastasis, no association with L-myc genotype was observed. In these latter studies, the number of Caucasian patients was sufficiently large to make valid statistical analyses however, the number of African American patients was too small to evaluate the relationship between lymph node metastasis and the L-myc genotype.
Analysis ofp53 Polymorphism Figure 3 shows the strategy for PCR-RE analysis of exon 4 ofp53. Dideoxy-chain termination DNA sequencing of the PCR products was performed to confirm their authenticity (DNA sequencing analysis was performed on a small subset of samples only). PCR analysis with restriction enzyme digestion was performed on 144 study subjects (76 cases and 68 controls). There was no association between presence of the p53 genotype and diagnosis of lung cancer (small-cell and nonsmall-cell combined); however, the observed allelic distribution was significantly different between African Americans and Caucasians (X2 = 20.8, p< 0.0001, df = 2).

Discussion
Molecular epidemiological studies of potential lung cancer risk factors have been described. A case-control study format has been used to evaluate the relationship between risk of developing lung cancer, or prognosis (when lung cancer is present), and genotype at the H-ras-1, L-myc, and p53 gene loci. The results of these studies show that individuals with H-ras-1 rare alleles are at increased risk for lung cancer (especially for nonadenocarcinoma in African Americans); that the L-myc genotype is not indicative of lung cancer risk or disease outcome in a U.S. population (prognosis, as measured by performance status, tumor stage, or presence of lymph node metastasis); and that the  proline variant of p53 does not appear to predict lung cancer risk. Each ofthese potentially useful biomarkers of lung cancer risk displayed a strong association with race: that is, allelotypic distributions were dependent on the racial origin of the study subjects (African Americans compared to Caucasians). These data indicate that the H-ras-1, L-myc, and p53 polymorphisms should be examined in more detail with regard to their potential as indicators of individual cancer risk, as should the underlying mechanisms that may be responsible for risk of cancer when certain genotypes are inherited. Furthermore, the data indicate the importance of controlling for race as a confounding factor in epidemiological studies.
The underlying mechanistic basis for the contribution of the inheritance of H-ras-1 rare alleles to risk of lung cancer is not known; however, the variable tandem repeat (VTR) region may be involved in somatic-recombinational events, implying that certain types ofVTR are more stable than others (19). This hypothesis may be extended to VTR regions other than that at the H-ras-1 gene locus. Alternately, VTR regions may have gene enhancer activity (20,21), therefore, it follows that certain variants may be more active than others. Notwithstanding the absence of a clear mechanistic basis for the observations reported here, they are supported by the results of a number of other studies (22)(23)(24), and, although the data in the literature are conflicting (25), not all studies of lung cancer have controlled for the confounding factors of tobacco smoking and racial background of the study subjects.
It was reported previously, by Kawashima et al. (26), that the EcoRI restriction site-present allele at the L-myc gene locus (S allele) was an indicator of lymph node metastasis (poor prognosis) in Japanese lung cancer patients. This hypothesis and the hypothesis that inheritance of S alleles is indicative of lung cancer risk have now been tested in two independent studies (10,1 7). Contrary to the results ofthe study in a Japanese population, neither of the two subsequent studies found associations in Caucasian populations.
L-myc genotyping has been conducted in a variety of cancer groups, mainly of Japanese origin. An association was found in the case of lung (surgical diagnosis) (26), renal (18), gastric (27), and bone cancers (28) between increased frequency of the S allele and advanced disease (metastasis); however, no association was found for lung [clinical diagnosis (26)], colorectal (29), or breast cancers (30). The reasons for these apparently contradictory results are unclear, but a genetic difference related to ethnicity could exist between respective lung cancer populations, even though the allelic frequencies in American Caucasians and Japanese controls do not differ significantly (L:S, 0.562:0.438 versus 0.485:0.515). Future studies which examine this question should control strictly for variation due to race and be designed to eliminate potential bias. Given that allelic frequency varies with ethnicity (L:S, 0.277:0.763 and 0.562:0.438 for African Americans and Caucasians, respectively), if one ethnic group consistently seeks medical attention at a later stage, or is com-m posed of an unusual number of prevalent (versus incident) cases, or is associated with a particular histology or exposure, bias could result. In the present study, the prognostic associations considered were similar in Caucasians and African Americans.
For p53, the proline variant of the polymorphism in the coding region is more frequent in African Americans than in Caucasians. No evidence was found, however, for an association between the presence or absence of the proline variant and lung cancer risk.
Tobacco smoke is the major etiological factor in lung cancer, however, not all individuals who smoke cigarettes develop lung cancer, and some individuals with little or no exposure develop lung cancer (31). These observations suggest that other factors may be important in the determination of lung cancer risk. Using a case-control design, we have evaluated three polymorphic genetic loci (H-ras-1, L-myc, and p53) as potential biomarkers of lung cancer risk. Racial variation in allelic distribution was apparent for each of these polymorphisms, which suggests that controlling for racial variation as a confounding factor in epidemiological studies is an important consideration in study design.