Primary Adenocarcinomas of the Lung in Nonsmokers Show a Distinct Pattern of Allelic Imbalance¹

Lung cancer is a common malignancy etiologically related to smoking. Genotoxic tobacco metabolites form bulky DNA adducts in the genome resulting in frequent chromatid breaks and mutations (1). In recent years, it is increasingly recognized that factors not related to direct cigarette smoking, such as passive smoking, toxicity from vaporized cooking oil, indoor fossil fuel combustion, and so forth, might also contribute to lung cancer development, particularly in women (2, 3). From studies in Hong Kong and mainland China, where relatively few (30–40%) female Chinese lung cancer patients smoke (4), it has long been observed that the incidence and mortality rate of female lung cancers are high compared with world standards (5).

The carcinogenic pathway of lung cancer development in nonsmokers is unclear, but probably involves a complex interplay of genetic and environmental mechanisms that lead to progressive accumulation of multiple genetic aberrations. Some of the aberrations may form part of overall genomic damages caused by the carcinogenic events. Others may accumulate from genomic instabilities that occur during tumor progression. However, not all altered loci are expected to harbor genes with determinant roles in tumor development. Thus, it is likely that some genetic alterations constitute “genetic noise” that may mask the essential aberrations. On the other hand, clinical tumors that show less extensive AIs would be expected to harbor less genetic noise and yet include the minimal combination of essential aberrations required for the malignant phenotype. In this study, we aimed to identify the essential molecular targets of lung carcinogenesis in nonsmokers by focusing on these tumors and detecting recurrent genetic aberrations represented by loci with high AI frequency.³ Forty-two adenocarcinomas from nonsmokers (NT-ADs) were first studied by microsatellite analysis using a broad panel of markers covering all nonacrocentric autosomes. To investigate whether the identified targets were unique to nonsmoker lung cancers, a selective panel of markers consisting of those with the highest AI frequencies in NT-ADs were analyzed in 29 T-ADs for comparison.

Seventy-one primary lung adenocarcinomas were collected after informed consent from 42 nonsmokers (41 never-smokers, 1 passive-smoker) and 29 smokers (25 chronic smokers, 4 ex-smokers), including 28 men and 43 women ages 38–81 (mean ± SD, 59.8 ± 10.4). The patients were recruited from the Grantham Hospital, Hong Kong, during the period 1992–1999. All of them were ethnic Chinese, and none had received any preoperative radiation or chemotherapy. Demographic data were obtained through patient interviews conducted by the designated clinician-in-charge (S. W. C.) at the first hospital admission according to a standardized protocol and were verified by a review of the hospital charts recorded in subsequent visits. For never-smokers, only patients who were lifetime nonsmokers and not exposed to smoking spouses were recruited. One patient was originally recruited as a never-smoker, but subsequent information revealed possible environmental tobacco smoke exposure, and she was designated as a passive-smoker. Chronic smokers had 15–60 pack-years of cigarette-smoking history and included those who had stopped smoking for <6 months. The ex-smokers had stopped smoking for <10 years, and those beyond this duration were excluded from the study. Smokers of tobacco products other than cigarettes were also excluded. The nonsmokers were predominantly women, and the smokers were predominantly men (P < 0.0001). Significant differences in age, tumor grade, and pathological stages were not present between the two populations (Table 1). Tumor classification was according to the WHO histological classification of lung tumors (1991), and the study cases included all subtypes of adenocarcinoma but not adenosquamous carcinoma.

The control samples consisted of macroscopically normal lung taken from a portion of the surgical specimen farthest removed from the tumor and/or peripheral blood mononuclear cell pellets obtained before any blood transfusion. Freshly obtained resection specimens were snap-frozen in liquid nitrogen and kept at −70°C until used. The tumor samples were examined histologically before use to ensure at least 80% of tumor by area, and normal lungs were examined to ensure no tumor presence. Selective cases, including those containing heavy normal cell admixture or presenting with interpretative difficulties, were microdissected to obtain pure tumor tissues for repeated analysis to ensure reproducibility of results.

Genomic DNA from the tumors, normal lung, and peripheral leukocytes was extracted using proteinase K digestion, phenol-chloroform extraction, and ethanol precipitation according to standard methods. For NT-ADs, 84 markers covering all nonacrocentric chromosomal arms and spanning common regions of genetic gain or loss in lung and other common carcinomas were selected after literature review and were analyzed using commercial primers (Research Genetics, Huntsville, AL). For T-ADs, a subset of 43 markers including those with high AI frequencies in the NT-ADs, as well as loci in chromosomes 3p, 9p21-p23, and 11q22-q24 that contain well-known hot spots of allelic loss in lung cancers, were analyzed (Table 2; see loci with footnote c citation). PCR was performed with 20- to 50-ng DNA template, 1.5–2.5 mm MgCl₂, forward and reverse primers (0.2 μm each), 200 μm dNTPs, 1.5 unit of Taq polymerase (Platinum Taq; Life Technologies, Inc.) and 1× PCR buffer (20 mM Tris-HCl [ph 8.4], 50 mM KCl). Denaturation (25–30 cycles) at 94°C for 60 s, annealing at 50°C to 63°C for 60 s according to individual markers, and extension at 72°C for 60 s was performed. PCR products were radiolabelled by nucleotide incorporation, resolved in 6% polyacrylamide gel containing urea and formamide, and exposed to X-ray films for 6–48 h. Cases were scored as heterozygous when two alleles were distinguished in the control DNA and as noninformative when only one allele was visualized. Heterozygous cases showing obvious reduction in the relative intensity of one allele in the tumor DNA was scored as AI. Results were recorded independently by at least two investigators (L. P. C. and M. P. W.). The percentages of agreement of results were generally more than 90%. Cases with discrepant interpretation were resolved by consensus and, if necessary, were repeated using DNA from microdissected tumor samples. Six markers (D1S 1597, D1S 518, D10S 2325, D10S 1223, TP53, and D19S 586) were additionally studied by fluorescence-labeled primers using ABI Prism 377 and Genescan fragment analysis software to ensure that criteria applied in visual interpretation of AI were reliable and reproducible.

The AI frequency of an individual marker was calculated as the ratio of the number of tumors showing AI to the total number of tumors showing heterozygosity for that marker. The extent of genetic aberrations of an individual tumor was denoted by its FAL, and was calculated as the ratio of the number of markers showing AI to the number of markers showing heterozygosity for that tumor. To identify the genetic aberrations commonly involved in tumors and, thus, more likely to represent essential alterations distinct from those predominantly altered in tumors with extensive aberrations and more likely to constitute background changes, the tumors were dichotomized and analyzed as the low- and high-FAL groups according to the mean FAL value of NT-ADs or T-ADs, respectively. Chronic and ex-smokers were grouped as smokers; passive and never-smokers were grouped as nonsmokers for statistical analysis. The χ² or Fisher’s exact test (GraphPad InStat, version 3.00) was used for the comparison of AI frequencies between patient groups and the analysis of covariance (SPSS for windows, version 10.1) for correlation of FAL with clinicopathological variables. A two-sided P of <0.05 was taken as showing statistical significance.

Microsatellite analysis of 84 markers was performed in NT-ADs from 39 women and 3 men, and the mean AI frequency was 37.0 ± 12.3% (range, 10–64%). The extent of chromosomal alteration in each NT-AD was evaluated by the FAL, which ranged from 3 to 76% (mean, 40.0 ± 20.4%). No significant correlation was found between the FAL and age, tumor grade, tumor stage, nodal stage, metastasis, or pathological stages. The mean FAL was used as the threshold to designate two groups, each of 21 tumors, with high or low FAL. No significant differences in the distribution of sex and other clinicopathological parameters between the low or high-FAL tumors were found. The low-FAL tumor group was examined for loci of high AI frequency, defined as those with AI of >40% using the approximated overall mean (37%) as the cutoff. The loci of frequent AI comprised 10 loci in 7q31, 8p23.2, 10p14-p15, 13q12.3, 16q24, 17p13.1-p13.3, 17q22, 19q13.3, and Xq11.2-q12 (Table 2; see loci with footnote d citation). These loci also showed AI of 40% or above in the high-FAL tumors, and a significant difference with the low-FAL tumors was observed in 13q12.3 (P = 0.002) only. The remaining 74 loci showed relatively infrequent AI (40% or less) in the low-FAL tumors compared with the high-FAL tumors, with the difference reaching statistical significance in 26 loci (P of <0.0001 to 0.05; χ² test, Table 2).

To investigate the uniqueness of involvement of the frequently altered loci in non-tobacco-related tumors, a selective panel of markers (Table 2; see loci with footnote c citation), including those with AI >40% in NT-ADs, were then analyzed in 29 T-ADs from 25 male and 4 female patients for comparison. The FAL of the T-ADs ranged from 4 to 84% (mean, 44.5 ± 24.7%). No statistical correlation was found between the FAL and the clinicopathological variables of age, tumor grade, tumor stage, nodal stage, metastasis, or pathological stage. Loci in 2q33, 6p23-p24, 10p14-p15, 13q14.1, 16q24, 17q11-q12, 17q22 (P = 0.06), 19p13.2, 19q13.3, 20p12, and 20q12 showed more frequent alterations in NT-ADs than in T-ADs (Fig. 1, solid bars). Although the differences did not reach statistical significance, these loci were relatively infrequently altered in T-ADs: they showed AI frequency lower than the mean level (49.7 ± 14.5%) for T-ADs. Loci in 1q24-q25, 2p24-p25, 3p22.3, 3p14.2, 3q25.2-q26, 7q31, 8p23.2, 9p22, 9q34, 10q26, 11q23.1, 13q12.3, 14q32, 17p13.1-p13.3, 18q23, and 21q22 showed closely similar AI frequencies in both groups (within 10%; Fig. 1, dotted bars), or more frequent alterations in T-ADs (Fig. 1, hatched bars).

The low-FAL tumors of NT-ADs and T-ADs, consisting of 21 NT-ADs and 13 T-ADs both designated according to the mean FAL values calculated from the panel of 43 markers analyzed in both tumor groups as cutoff, were compared for recurrent aberrations. There were more women than men in the NT-ADs (P < 0.0001; Table 3). No significant differences in other clinicopathological parameters were detected. In the low-FAL NT-ADs, the same 10 loci as those identified in the analysis of all of the markers showed AI frequency of >40%. Three of 10 loci showed significantly more frequent AI in the low-FAL NT-ADs than in the T-ADs (16q24, P = 0.03; 17q22, P = 0.04; and 19q13.3, P = 0.04), whereas no significant difference was observed for 7q31, 8p23.2, 10p14-p15, 13q12.3, and 17p13.1-p13.3 (Table 3).

We have analyzed the pattern of AI in a broad spectrum of microsatellite markers in adenocarcinomas from nonsmokers to identify common genetic targets of tumor development. In view of the considerable amount of data that result from studies using genome-wide approaches, we have used two strategies to help us focus on alterations that are most likely to be important. Firstly, we have used the mean AI percentage of all of the markers in NT-ADs (40%) as our threshold to designate loci of frequent AI throughout this study. Secondly, we propose that essential and critical alterations that drive malignant progression could be found in tumors that show fewer genetic changes; therefore, we searched for these essential aberrations by examining tumors with low FAL. No FAL correlation with the patients’ age, tumor grade, tumor, nodal, metastasis, or pathological stages is found in either NT-ADs or T-ADs, implying that differences between tumors with low or high FAL are unlikely to be caused by variations in these clinicopathological parameters. Accordingly, 10 loci in 7q31, 8p23.2, 10p14-p15, 13q12.3, 16q24, 17p13.1-p13.3, 17q22, 19q13.3, and Xq11.2-q12 showed AI of >40% in both low- and high-FAL tumors, which suggests that these loci are important in all lung cancers arising in nonsmokers. The remaining loci with low AI frequencies in the low-FAL group, especially the 26 loci with significantly more frequent AI in the high-FAL tumors, might represent loci that are not critical for tumor development.

To investigate whether frequently altered genetic loci identifie d in NT-ADs were uniquely involved in non-tobacco-mediated carcinogenesis, loci with frequent AI were also analyzed in 29 adenocarcinomas from smokers (T-ADs) for comparison. The results indicated a wide overlap in genetic aberrations when all tumors of both groups were analyzed, including 11 regions (2q33, 6p23-p24, 10p14-p15, 13q14.1, 16q24, 17q11-q12, 17q22, 19p13.2, 19q13.3, 20p12, and 20q12) of frequent AI in NT-ADs but relatively infrequent changes in T-ADs, and 17 regions (1q24-q25, 2p24-p25, 3p22.3, 3p14.2, 3q25.2-q26, 7q31, 8p23.2, 9p22, 9q34, 10q26, 11q23.1, 13q12.3, 14q32, TP53, 17p13.1-p13.3, 18q23, and 21q22) with genetic gain/loss at a similar or higher frequency in T-ADs. Comparison between low-FAL tumors of nonsmokers and smokers revealed that among the 10 loci of recurrent alterations found in NT-ADs, changes in 16q, 17q, and 19q were significantly more frequent than in the low-FAL T-ADs, which indicated that they are likely to be critically and uniquely involved in non-smoking-related cancers. On the other hand, alterations in 7q, 8p, 10p, 13q, and 17p were not found to be significantly different, and they could be related to similarities in tumor development in smokers and nonsmokers, such as exposure to common mutagenic compounds, involvement of common susceptibility factors, disruption of tissue-specific regulatory mechanisms of the lung, and so forth. Interestingly, 17q22 is located in a syntenically conserved region in the mouse genome that contains the mouse pulmonary adenoma resistance gene 1 (PAR1) involved in the genesis of the urethane-induced mouse model of lung adenocarcinoma (6). Because lung cancers developing in this animal model have been shown to share pathogenetic and biological behavior that is similar to that in human lung cancers (7), our finding of frequent AI in this region suggests that important tumor suppressor genes for human lung cancer could be located in 17q22. Fong et al. (8) have also reported 42% alterations of 17q in NSCLC, a frequency similar to that in our findings. 16q23-q24, which spans the common fragile site FRA16D, shows frequent allelic loss or homozygous deletions in cancer cell lines or primary cancers (9). The H-cadherin gene located in this region shows frequent down-regulation and methylation in NSCLC, but mutations have not been identified (10, 11). 10p15 and 19q13.3 are novel regions of genetic alterations identified in a genome-wide screening study of lung cancer cell lines using high-density marker sets (12, 13). Frequent alterations in 7q31 have not been previously reported in allelotype studies; however, karyotyping and comparative genomic hybridization studies have consistently reported polysomy of chromosome 7 or amplification of 7q31 in NSCLC (14, 15). Because amplified copies of a DNA fragment may manifest as AI, our results may be a reflection of the polysomy and may indicate the presence of potential oncogenes in this region.

Because of the small number of male nonsmoking and female smoking patients who present with excisable lung cancers in our population, the effect of gender differences on AIs cannot be independently evaluated in our study. The high proportion of nonsmoking women with lung cancers may reflect potential influences of female-specific factors or differences in susceptibility to common carcinogenic agents. In a study of Chinese subjects, significantly higher DNA adduct levels were found in the nontumor lungs of female nonsmokers than in those of male nonsmokers (16), which suggests that female susceptibility to DNA damage derived from environmental carcinogen exposure may be a confounding factor in lung cancer development. A higher expression level of gastrin-releasing peptide receptor in women, which is involved in the regulation of cellular proliferation and is induced in smokers, has also been proposed as a mechanism for the increased susceptibility (17).

Our findings of frequent genetic changes in NT-ADs, with a mean AI frequency of 40%, and the involvement of an overlapping spectra of loci compared with T-ADs, are different from those reported by Sanchez-Cespedes et al. (18) in a recent study in which AI was infrequent in 18 adenocarcinomas from nonsmokers. None of 54 markers from 28 chromosomal arms showed alteration of >25%. Changes at 9p21, 12p, and 19q13.3, each occurring at 22%, were the commonest alterations found. Our data support frequent involvement of 19q13.3 (48%) and 9p21 (45%) in NT-ADs in general, but we have not found 12p alterations to be a frequent event. Furthermore, 19q13.3 involvement is also frequent in our low-FAL tumors, which suggests that genes that are critically essential for lung cancer oncogenesis in nonsmokers may be linked to this region. The reason for the discrepant findings between the two studies is not clear, but variations in genetic, environmental, and lifestyle factors may be involved. It would be interesting to compare the patient data, particularly sex distribution, of nonsmoking subjects in the reported study.

In summary, we have shown that AI in 16q24, 17q22, and 19q13.3 may represent essential alterations in lung adenocarcinomas arising in nonsmoking patients, particularly for females. They represent useful sites for additional mapping of genetic targets in non-tobacco-associated lung carcinogenesis by high-density microsatellite marker sets.

We thank May Wong, Discipline of Dental Public Health, Dental Faculty, University of Hong Kong, Prince Philip Dental Hospital, Hong Kong, for her help in statistical analysis.