Distinct Epidermal Growth Factor Receptor and KRAS Mutation Patterns in Non–Small Cell Lung Cancer Patients with Different Tobacco Exposure and Clinicopathologic Features

Cigarette smoking is the most important known cause of lung cancer (1–3) with around 10% of chronic smokers eventually developing the disease (4, 5). Although the risk is reduced with cessation of smoking, it does not decrease to the level of nonsmokers (5, 6), suggesting that genotoxic effects of tobacco can be long-lasting. Environmental tobacco smoke is also a recognized carcinogen that increases the susceptibility of passive smokers to lung cancer (3). For never smokers, different etiology and carcinogenic processes are likely to be involved, but current understanding is still limited.

Epidermal growth factor receptor (EGFR) is a receptor tyrosine kinase that plays a crucial role in cell proliferation, survival, and differentiation (7). Ligand-binding leads to receptor dimerization, autophosphorylation of cytoplasmic tyrosine kinase domains, and activation of the EGFR signaling pathway (8–10). Findings of EGFR overexpression in a number of human solid tumors, including lung, breast, prostate, bladder, colon, head and neck, and ovarian carcinomas (11–13), have led to the targeting of EGFR in anticancer drug development. It has been observed that nearly all non–small cell lung cancer patients who show clinical responses to the orally administered EGFR inhibitor gefitinib harbor somatic mutations in the EGFR gene (14–17). In vitro experiments have shown that certain EGFR mutants show increased tyrosine kinase activity and confer higher sensitivity to growth inhibition by tyrosine kinase inhibitors (TKI; refs. 15, 18). Interestingly, response to TKI-based therapy and EGFR mutations are more frequently found in nonsmokers, female patients, adenocarcinomas, and lung cancers showing bronchioloalveolar features (15, 17, 18).

Lung cancer incidence in men is higher in Hong Kong than in many developed countries, whereas the incidence in women is similar (3). However, unlike female lung cancers in North America of which >70% are caused by smoking (2), most of the cases in Hong Kong occur in nonsmokers (3). Distinct molecular genetic pathways underlying lung carcinogenesis in men and women or in patients with different tobacco exposure are postulated. The objective of this study is to evaluate whether this epidemiologic profile and other clinicopathologic characteristics of primary lung adenocarcinoma are related to mutations in the EGFR and KRAS genes.

Sample collection and DNA extraction. The samples were collected from 241 patients who underwent tumor resection in the Grantham Hospital and the Queen Elizabeth Hospital, Hong Kong in 1989 to 2003. They consisted of 215 adenocarcinomas, 11 lymphoepithelioma-like carcinomas (LELC), and 15 squamous cell carcinomas. Aspects of the LELC cases in this study had been reported in a previous study (19). Sample collection for research was undertaken after approval by the Ethics Committee, The University of Hong Kong before 2002 and by the Institutional Review Board of the respective hospitals from 2002 onwards. An informed written consent was obtained from each patient. The patients were all ethnic Chinese and had not received chemotherapy or radiation before surgery. The smoking history was obtained prospectively from the patients according to a standard protocol at the first hospital admission, reviewed, and verified in every subsequent outpatient visit. A summary of the demographic and pathologic data was listed in Table 1. Categories of tobacco exposure were classified as follows: nonsmokers referred to patients who had smoked <100 cigarettes in their lifetime; ex smokers referred to subjects with lifetime consumption of >100 cigarettes but who had stopped smoking for at least 1 year before recruitment; current smokers referred to patients who were still smoking at the time of recruitment; details of cigarette consumption was recorded in 57 of 68 smokers which ranged from 2 to 150 pack-years; passive smokers referred to patients who volunteered a history of regular exposure to environmental tobacco smoke exhaled from current smokers at home or at work places. Tumor typing and grading were independently done by two qualified anatomic pathologists according to the WHO criteria for classification of tumors of the lung (20). Discrepancies in typing and grading were resolved by consultation and mutual agreement. LELC were all undifferentiated tumors, and other non–small cell lung cancer were graded as well differentiated, moderately differentiated, or poorly differentiated. Bronchioloalveolar carcinoma type (BAC type) tumors referred to tumors showing exclusively a lepidic growth pattern along intact alveolar walls. Mucinous differentiation referred to tumors showing predominant diffuse goblet cell–like features. Control DNA of individual patients was obtained from peripheral blood mononuclear cells or nonneoplastic lung collected from an area of the resected specimen farthest away from the tumor. The tumor and corresponding normal lung were snap frozen and kept in liquid nitrogen within 30 minutes of resection. The samples were reviewed histologically before DNA extraction to ensure ≥70% of tumor cell proportion in the tumor samples and absence of tumor cells in the normal lung. DNA was extracted by standard proteinase K digestion and phenol-chloroform extraction method.

EGFR and KRAS mutations screening.EGFR exons 18 to 21 were amplified with intron-based primers previously published by Lynch et al. (15) and Paez et al. (16). DNA fragment containing KRAS mutation hotspots codons 12 and 13 were amplified with primers KRAS-1F (5′-CTGGTGGAGTATTTGATAGTGTATT-3′) and KRAS-1R (5′-ATCTGTATCAAAGAATGGTCCTG-3′); whereas for codon 61 with primers KRAS-2F (5′-GAAGTAAAAGGTGCACTGTAATAAT-3′) and KRAS-2R (5′-CAATTTAAACCCACCTATAATGGT-3′). The nucleotide sequence for EGFR was numbered according to National Center for Biotechnology Information accession no. NM_005228 (coding sequence starts at base 247) and for KRAS was NM_033360 (coding sequence starts at base 182).

PCR products were treated with Exonuclease I and Shrimp Alkaline Phosphatase at 37°C for 15 minutes followed by heating at 80°C for 15 minutes to stop the enzymatic reaction. Treated PCR products were sequenced with Big Dye Terminator v3.1 sequencing kit (Applied Biosystems, Foster City, CA). Cases with novel or double mutations were confirmed by repeated PCR and sequencing, and the corresponding normal DNA was also sequenced to verify that the mutations were somatic.

EGFR mutant cDNA subcloning. For eight of the tumors with double mutations having frozen tissues available, RNA was extracted by Trizol and reverse transcribed into cDNA with SuperScript First-Strand Synthesis System (Invitrogen, Carlsbad, CA). The cDNA was PCR amplified with specific primers EGFR-CF (5′-CAAGCTTCTGGAGGGTGAG-3′) and EGFR-CR (5′-GGGCCATTTTGGAGAATTC-3′) to generate a 1.3-kb fragment containing EGFR exons 18 to 21. The fragment was cloned into pCR4-TOPO vector (Invitrogen) and the plasmids from individual colonies were sequenced to determine the allelic distribution of the mutations.

Statistical analysis. Statistical analysis was done using SPSS 13.0 (SPSS, Inc., Chicago, IL). The χ² test and Fisher's exact test were used to compare the proportion of EGFR or KRAS mutations among different clinicopathologic groups. Mann-Whitney U test was done for comparison of means between mutant or wild-type tumors. To investigate the effects of covariates on gene mutations, multiple logistic regression analysis using a forward stepwise (likelihood ratio) method was done with odds ratio (OR) calculated. Initial testing included smoking history, age, gender, tumor histology, size, differentiation, and pathologic stage. Only variables showing statistically significant association with gene mutations were subjected to final regression analysis. The two-sided significance level was set at P < 0.05.

EGFR mutations and clinicopathologic profile.EGFR mutations were detected in 48% (116 of 241) of tumors with a significant association with histologic type, involving 115 of 215 (54%) adenocarcinomas, 1 of 11 (9%) LELC, and none of 15 squamous cell carcinomas (P < 0.001). There was significant association with the smoking history (P < 0.001), with the highest mutation ratio in nonsmokers (71%) followed by decreasing ratios in passive smokers (55%), ex smokers (29%), and lowest in smokers (16%). Significant associations were also observed with nonsmokers compared with the smokers (P < 0.001) or ex smokers (P < 0.001) but not passive smokers subgroups. When only the 215 adenocarcinoma were considered, similar patterns of significant associations between EGFR mutations and different levels of tobacco exposure were observed (Table 2). The amounts of tobacco consumption were known in 47 smokers, including 8 mutant tumors with mean 40.0 ± 16.3 pack-years smoked and 39 wild types with mean 32.2 ± 27.9 pack-years. No significant difference in cumulative tobacco consumption was observed between the two groups. EGFR mutations were significantly associated with the female gender (P < 0.001). There was decreasing ratios of EGFR mutations from well (64%) followed by moderate (55.7%) and poor (26.8%) tumor differentiation (P < 0.001), with subgroup testing also showing significant association with well-differentiated compared with poorly differentiated tumors (P < 0.001). No association was found between EGFR mutation status and patient's age, tumor size, or pathologic stage. Multiple logistic regression analysis to investigate the effects of smoking history, gender, and tumor differentiation on EGFR mutations in adenocarcinoma showed that smoking history (compared with nonsmokers: smokers OR = 0.092, P < 0.001; ex smokers OR = 0.241, P = 0.004; passive smokers OR = 0.440, P = 0.166) and differentiation (compared with well differentiated: poorly differentiated OR = 0.294, P = 0.014; moderately differentiated OR = 0.861, P = 0.682) but not gender were significant determinants for EGFR mutations (Nagelkerke R², 33.3% overall, 29.8% for smoking history).

EGFR mutations and tumor cell differentiation features. Twenty-four tumors showed growth characteristics of bronchioloalveolar differentiation (BAC-type tumors), including 17 without invasive growth, and satisfied the current WHO criteria for defining bronchioloalveolar carcinomas (20), and seven showing focal invasion of the pleura or intratumoral lymph vascular channels. Fifteen of the 24 (62.5%) BAC-type tumors compared with 100 of 191 (52.4%) of non–BAC-type adenocarcinoma showed EGFR mutations (P = 0.391; Table 3). For BAC-type tumors, all EGFR mutations occurred in nonsmokers (15 of 20), and none was found in patients with other levels of tobacco exposure (one passive smoker and three smokers; P = 0.012). A significant association was found with nonmucinous (15 of 19) compared with mucinous (0 of 5) tumors (P = 0.003; Table 3). In contrast, when non–BAC-type adenocarcinoma were considered, no significant association was observed regarding mucin differentiation. No association between EGFR mutation and the presence of invasion was found for BAC-type tumors. In addition, none of three solid carcinomas with mucin production and three fetal adenocarcinomas showed EGFR mutation.

EGFR mutation patterns. A total of 131 mutations, including 40 types of mutation arranged in different patterns of single and double mutations, were found. They involved 101 tumors of single and 15 of double mutations (Table 4). Most mutations were found in exon 19 (45%) and exon 21 (41%) followed by exon 18 (8%) and exon 20 (6%). Except for one silent change of A755, confirmed to be somatic by sequence comparison with corresponding normal DNA, all the other 130 mutations led to amino acid sequence alterations. There were 70 amino acid substitutions (53%) scattered throughout the four exons. In-frame deletions (56 of 131, 43%) were found only in exon 19 and insertions (5 of 131, 4%) only in exon 20. Most of the amino acid substitutions involved single nucleotide changes, but there were also cases with changes of two to four nucleotides. A dinucleotide change (2400_2401GG>TT) was found in one patient that led to alterations of two codons (CTG>CTT-L718 silent and GGC>TGC-G719C) in exon 18. Another change involved four nucleotides (2487_2490AAGA>CCCG) resulting in L747F and R748P in exon 19. The commonest deletion was delE746_A750 (34 of 56, 61%). The size of the deleted fragments ranged from three to eight amino acids. Loss of L747 and R748 was observed in 55 of the 56 deletions (Fig. 1). The remaining deletion was found immediately downstream at 2499_2522, resulting in delS752_I759. Interestingly, codon K745 positioned immediately upstream of L747 and R748 was unaltered in all patients. In exon 20, most of the insertions involved a valine (80%). Three of the five insertions were an addition of three amino acids between V769 and D770. One mutant had a replacement of two nucleotides by five (2557_2558AA>GGGTT), resulting in the substitution of one amino acid by two (N771 replaced by GF).

Double mutations. Double mutations located either in the same or different exons were found in 15 tumors (Table 5). Five novel alterations were found coupling with another amino acid substitution: G724S (with L861Q), E734K (with L861Q), Q787R (with L858R), V834L (with L858R), and L858M (with L861Q). Notably, when mutation at L858 was coupled with mutations at L861, L858M instead of the more common L858R was found. The two L858M were observed only in the presence of L861Q, whereas both L858R and L861Q could be observed alone or coupling with other point mutations. Sequencing results of EGFR cDNA mutant subclones from eight tumors with double mutations showed either doubly mutated or wild-type alleles only (Table 4).

KRAS mutations and clinicopathologic profile. All KRAS mutations were found in adenocarcinoma and occurred in 21 of 215 tumors (9.8%). Nineteen mutants were found at codons 12 and 2 at codon 61 (Table 4). KRAS mutations were significantly associated with smoking history (P = 0.001), with higher proportions in ex smokers (21.4%) and smokers (19%) than nonsmokers (3.6%) and passive smokers (0%; Table 2). Significant associations were also found with nonsmokers compared with the ex smoker (P = 0.005) or smoker (P = 0.003) but not passive smoker subgroups. KRAS mutations were also significantly associated with the male gender (P = 0.009). When all adenocarcinoma were considered, there was no significant association between KRAS mutation and tumor differentiation, but subgroup testing showed statistically significant association with poorly differentiated compared with well differentiated (P = 0.037; Table 2). No significant association was found between KRAS mutations and age, BAC-type morphology, tumor size, or stage. Notably, tumors with KRAS or EGFR mutations were mutually exclusive with no tumor containing both mutations. Logistic regression analysis to investigate the effects of smoking and gender on KRAS mutations in adenocarcinoma showed that smoking history was the only significant determinant of KRAS mutation (Nagelkerke R² = 16.7%; compared with nonsmokers: for smokers, OR = 6.261, P = 0.003; for ex smokers, OR = 7.295, P = 0.004; for passive smokers, OR = 0, P = 0.998 for passive smokers).

The mutational profile of EGFR and KRAS was studied in three histologic types of lung non–small cell lung cancer and compared with their clinicopathologic characteristics. The tumor types were chosen for their different etiologic background: adenocarcinoma is currently the predominant tumor type in both smokers and nonsmokers; squamous cell carcinoma was the commonest tumor type in smokers before the use of cigarette filters; LELC is an uncommon tumor associated with EBV (21) and comprises 5% to 9% of primary non–small cell lung cancer (19, 22). Consistent with reported findings that EGFR mutations are prevalent in adenocarcinoma, we have found no mutation in squamous cell carcinoma and only 1 in 11 LELC. Compared with other reported studies, the EGFR mutation rates of nonsmokers, ex smokers, and smokers patients in Hong Kong were the highest amongst patients from Japan (nonsmokers, 66%; ever smoker, 22%; ref. 23), Korea (nonsmokers, 46%; ex smokers, 13%; smokers, 4%; ref. 24) and the United States (nonsmokers, 20%; ex smokers, 8%; smokers, 3%; ref. 25). Decreasing proportions of EGFR mutations were observed in never (71%), passive (55%), previous (29%), and current smokers (16%), supporting the observation reported in patients from the United States (25) or Korea (24) that EGFR mutation was negatively correlated with increasing direct exposure to tobacco. The mutations were common in adenocarcinoma showing well to moderate differentiation as well as in those showing nonmucinous BAC features. In view of the latter observation, we also analyzed the adenocarcinoma for immunophenotypic expression of the pneumocytic marker TTF1 but no significant association with EGFR mutations was found (data not shown). Although limited examples were available for evaluation, EGFR mutations were apparently less common in tumors showing other types of differentiation, such as mucinous BAC, fetal adenocarcinomas, or solid tumor with mucinous differentiation. Regression analysis showed that smoking history was the major determinant of EGFR mutations accounting for 30% of R², leaving differentiation accounting for 3.3%. Although mutations were more frequent in women, patient's gender was found to be a confounding factor related to the predominance of nonsmokers in female patients in Hong Kong.

The mutations detected in this study included amino acid substitutions (53%), deletions (43%), and duplications or insertions (4%). About 86% of amino acid substitutions involved four codons, including E709 (4%), G719 (9%), L858 (63%), and L861 (10%). Most of the amino acid substitutions (73%) were located on or very near to conserved kinase motifs, such as the highly conserved glycine-rich GXGXΦG motif (₇₁₉GSGAFG₇₂₄) of the ATP-binding loop of protein kinases (26) and the conserved DFG (codons 855-857) motif in the activation loop of EGFR (16, 17).

All deletions were in frame and involved parts of E746_I759 in exon 19, with a consensual loss of the dinucleotide L747_R748 in >98% of cases. These deleted fragments were very close to K745, a conserved amino acid critical for ATP-binding within the tyrosine kinase domain (17), which was retained in all tumors studied. This interesting observation confirmed the role of K745 for kinase activity, whereas loss of L747_R748 might cause a regulatory defect.

All insertions were in frame and found within region V769 toV774 in exon 20, with 80% showing an insertion of valine. By sequence alignment, this insertion region was consistent with the insertion site previously reported in EGFR and was analogous to that of HER2 in lung adenocarcinomas (27).

Double mutations involving the same or different exons were observed in 15 patients, accounting for 13% of all mutants (15 of 116) and 6% of all patients (15 of 241). Our data showed that rare mutations could couple with another rare mutation, as other investigators had previously noted (28), but could also coexist with common mutants, such as L858R (with E709K, Q787R, T790M, L833V, or V834L) or exon 19 deletion (with G873E or T790M). Together with data from other reports, we have observed three mutations that only existed in double mutations: E709 mutations (14, 23, 25), T790M (23), and L858M. One interesting observation was that when the leucine at both residues 858 and 861 were replaced simultaneously, L858M rather than the more common L858R was the “partner” (cases E18 and E234; Table 5). We also investigated whether the mutations resided on the same allele by subcloning cDNA derived from eight tumors with double mutations. The results showed that all picked colonies were either wild-type or double mutants. In contrast, Kobayashi et al. showed in a recurrent tumor originally harboring an exon 19 deletion, a second T790M mutation that either coupled with wild-type or the preexisting deletion, suggesting that the two mutations might be biallelic or originated from two different cell populations (29). They also showed that the second T790M mutation did not significantly contribute to tumor growth but instead conferred resistance to gefitinib amongst at least four TKIs in the recurrent tumor (29). In our samples, the tumors that showed double mutations (including T790M) occurred de novo without prior TKI treatment.

Gazdar et al. reviewed the common patterns of EGFR mutations and pointed out that exon 19 deletions, exon 20 duplications or insertions, and L858 or L861 substitutions together accounted for about 95% of all mutations identified (28). In this study, those mutations constituted 84%, whereas a large variety of relatively rare mutations, including double and silent mutations, were also observed. Our findings, together with those reported in the literature (14–17, 23, 25, 28), suggested that lung cancers from Oriental patients display greater complexities of EGFR mutational patterns than other ethnicities. Thus far, only a few mutation types had been shown to confer tyrosine kinase activity and sensitivity to growth inhibition by TKI (15) using in vitro transfection systems. The significance of other mutation types is incompletely understood but not all are activating mutations. Instead, they could represent nonfunctional alterations (e.g., A755 silent), changes coselected with a nearby activating mutation, or even a TKI-resistant mutant like T790M (29). Because the mutation profile could carry important therapeutic implications, it would be important to study the functional role of the different single and double EGFR mutants detected in this study for better assessment and prediction of the survival benefit of TKI treatment.

KRAS mutations were found in 21 of 215 adenocarcinoma (9.8%). Similar to other studies, the mutations were significantly associated with previous or current smokers (24) and occurred in mutually exclusive tumors to those containing EGFR mutations (23, 28). An interesting observation was that mutations of EGFR occurred more often in nonsmokers and well-differentiated tumors, whereas KRAS mutations occurred more in smokers and poorly differentiated tumors. Given that KRAS mediates an alternative mechanism for activation of the phosphatidylinositol 3-kinase and AKT pathways, activation of either gene can lead to common signaling abnormalities. To date, EGFR mutations have not been found in cancers of other organs (15, 25) and occurred rarely in other histologic tumor types apart from lung adenocarcinoma. No other receptor tyrosine kinases have shown mutations in lung carcinomas (16). Together with other published studies, our findings indicate that EGFR mutation is one of the most important and specific oncogenic aberrations found in lung adenocarcinoma, especially for nonsmokers, and EGFR and KRAS served as alternative activation points in lung adenocarcinoma development, at different prevalence in patients of different smoking and clinicopathologic background.