Despite recent advances in diagnosis and multimodality therapies for lung cancers, the prognosis still remains unsatisfactory with 5-year survival rates of only 15% for all stages.1 Because of the limitations of conventional chemotherapeutic agents, new strategies based on molecular mechanisms are needed, which will provide more effective clinical benefits for patients with lung cancer. Although several genetic or epigenetic changes in critical genes (oncogenes and tumor suppressor genes) have been identified in human cancers including lung cancers, such discoveries rarely lead to clinical applications immediately. However, successful treatments to inhibit specific protein kinase such as imatinib for the BCR–ABL translocation or c-KIT mutation and trastuzumab for HER2 amplification have encouraged the concept of targeted therapy for human cancers.2, 3 The discovery that mutations of the tyrosine kinase (TK) domain of epidermal growth factor receptor (EGFR) occur in a subset of lung cancers and predict for sensitivity to TK inhibitors (TKIs) has generated enormous interest. This review focuses on the relationship between EGFR mutations and clinico-pathologic features in lung cancers as well as to mutations in other molecules related to EGFR signaling pathways. These findings may lead to the identification of lung cancer subpopulations that will demonstrate the optimal response to targeted therapies, and support the notion that different molecular mechanisms are involved in the pathogenesis of lung cancers in smoker and never-smoker populations.
Somatic mutations in the tyrosine kinase (TK) domain of the epidermal growth factor receptor (EGFR) gene in lung cancers have generated enormous interest, because they predict for sensitivity to TK inhibitors (TKIs). While mutational status is of great importance in determining response to TKIs, it is not the sole factor, and evidence is accumulating that EGFR gene amplification, other members of the EGFR family (HER2, HER3) and genes downstream of EGFR signaling (KRAS, BRAF), may be involved in cancer pathogenesis and the response of TKIs. EGFR mutations occur in highly selected subpopulations of lung cancer patients: adenocarcinoma histology, never-smoker status, East Asian ethnicity and female gender. The recent finding of “a resistance associated” mutation for TKIs also provides new insights into this complicated mechanism. Thus, molecular-based studies to analyze the biological functions and to assess TKI sensitivity depending on the type of mutations are required. Epidemiological studies to identify possible carcinogenic factor(s) affecting different subpopulations are also of interest. In addition, for optimal therapeutic approach a comprehensive understanding of the genes related to EGFR signaling pathway, including RAS/RAF/MAPK and PI3K-AKT pathways, are required. © 2005 Wiley-Liss, Inc.
Type of EGFR mutations and their significance
After the initial publications of EGFR mutations in lung cancers,4, 5 this exciting field has moved rapidly and several studies have been published. Interestingly, EGFR mutations seem to be limited to lung cancers, and no mutations have been confirmed in other types of cancers.4, 6, 7, 8 However, recent studies reported that rare missense mutations (exons 19 and 20) were detected in colorectal cancer,9 and the same deletions (exon 19) as seen in lung cancers were detected in squamous cell carcinoma of head and neck.10 We summarize EGFR mutations reported in 9 published studies,4, 5, 6, 11, 12, 13, 14, 15, 16 and include additional cases analyzed by the authors. A total of over 2,000 non-small-cell lung cancer (NSCLC) samples have been analyzed, and a total of 477 mutations were detected. To date, all mutations are thought to be somatic in origin, and germline mutations have not been detected. The mutations are limited to the first 4 exons (exons 18–21), coding for the TK domain, which consists of 2 roughly globular-shaped N- and C-lobes.17 The mutation-associated exons code for the N-lobe and the 5′ portion of the C-lobe (Fig. 1a). Adenosine triphosphate (ATP) is bound in a cleft between these 2 lobes and sits beneath a N-lobe structure, the phosphate binding loop (P-loop) containing a highly conserved glycine-rich sequence (GXGXXG motif). TKI such as gefitinib and erlotinib bind to the same cleft. A centrally located activation loop (A-loop) in the C-lobe provides a platform for downstream protein activation, and is phosphorylated in the active state. Phosphorylation stabilizes the A-loop in an open and extended state. The most prominent structure in the N-lobe, the αC-helix, helps regulate the angle of the ATP binding cleft.17
The mutations consisted of 3 very different types (deletions, insertions and missense point mutations), and they all target key structures around ATP binding cleft, including the P-loop, the αC-helix and the A-loop. In-frame deletions in exon 19 accounted for 44% of all mutations and are the most predominant type of EGFR mutation (Fig. 1b). Missense point mutations were the second most common mutation spanning all 4 exons, especially single nucleotide substitution L858R in exon 21 (accounting for 41% of all mutations) and occasionally G719X (X indicates A, C or S) in exon 18 (accounting for 4% of all mutations). Other rare missense mutations have been described at multiple sites. In-frame duplications/insertions in exon 20 were also detected accounting for 5% of all mutations. The G719X mutation located on the highly conserved glycine-rich sequence in the P-loop, and the L858R mutation located near the conserved aspartic acid–phenylealanine–glycine sequence (DFG motif) that stabilizes the A-loop. The other two common mutations, in-frame deletions and in-frame duplications/insertions, occur on either side of the αC-helix. We hypothesized that both mutations result in similar configurational changes, causing a shift of the helical axis, narrowing the ATP binding cleft and resulting in both increased gene activation and TKI sensitivity.17 As postulated in a previous study, mutations result in repositioning of critical residues around ATP binding cleft and in stabilizing the interactions with both ATP and TKIs.4 To date, two forms of mutant receptors (deletions and L858R) have been demonstrated to increase the amount and duration of activation compared to those of wild-type receptors.4 The mutations result in activation of downstream pathways, and preferentially activate the antiapoptotic pathways (PI3K/AKT and JAK-STAT), but have less effect on cellular proliferation through ERK/MAPK signaling.18
Do all types of mutation have equal effects in response to TKIs or tumorigenesis? Are all rare mutations functional (i.e., activating)? Mitsudomi et al. demonstrated that gefitinib was more effective in patients with deletions than in patients with other types of mutation.19 Only 3 types of mutant cell lines (in-frame deletions and L858R missense mutation with or without T790M) are available to date.20 However, there are relatively few cases with the rare mutations, we do not know the relation between these mutations and responses to TKIs. We demonstrated that patients with tumors having deletions had a worse survival than those with L858R mutation with borderline significance,6 and these findings suggest that the specific mutation types have different effects not only in sensitivity to TKIs but also in the progression of lung cancers. Because only a few studies of biological function using mutant constructs have been conducted, the biologic effects of other mutant phenotypes (rare missense mutations and duplications/insertions) are unknown.
The mutations apparently target the progenitor cells of the peripheral airways (the “terminal respiratory unit”21), and the resultant tumors are peripherally arising adenocarcinomas, often having features of bronchiolo-alveolar growth patterns and lacking mucin production. Of interest, in cases of mutation-positive adenocarcinomas, potential precusor lesions, namely atypical adenomatous hyperplasias, have been found to harbor the same mutations.22 We also found that mutation-positive tumors demonstrated a field effect, with mutations present in adjacent bronchioloes (Tang et al., in press). No study has demonstrated that EGFR mutations correlated with disease stage. Thus, EGFR mutations appear to be early event in the lung cancer pathogenesis.
Clinical and demographic factors that predispose to EGFR mutations
As we described previously, EGFR mutations strikingly target adenocarcinoma histology (30% – 413 of 1380) compared to other NSCLC histologies (2% – 16 of 993) (p < 0.001; Fig. 1c). Only 16 cases of other NSCLC histologies (7 of which were adeno-squamous cell carcinomas, 6 cases were squamous cell carcinoma, 1 large cell carcinoma and 2 others) had EGFR mutations. The mutations were absent in neuroendocrine lung tumors, including bronchial carcinoids, large cell neuroendocrine carcinomas and small cell carcinomas.6 While responses to gefitinib have been reported to be higher23 and EGFR mutations were preferentially observed in tumors having bronchioloalveolar carcinoma (BAC) features,4, 15 we found no association with the BAC subtype of adenocarcinoma in 97 cases from the United States,6 using the strict criteria as stated by the World Health Organization classification of lung tumors.24
Smoking status was available for 2,128 cases and the mutation frequency was significantly different between never smokers (45%) and ever smokers (7%) (p < 0.001; Fig. 1c). When limited to adenocarcinoma cases (n = 1,082), mutations were present in 54% (232 of 433) of never smokers compared to 16% (101 of 649) in ever smokers (Fig. 1d). For 160 NSCLC cases from the United States with detailed smoking history, in our study, 3% of current smokers, 8% of former smokers and 20% of never smokers had EGFR mutations.6 There was a significant trend effect that smoking was inversely related to the occurrence of EGFR mutations and the same finding was also reported in another series.16 Although there is the possibility of “second-hand smoke” or “passive smoke,” these results clearly showed that tobacco is not a major carcinogen for EGFR mutations. EGFR mutations are the first molecular change that specifically targets tumors arising in never smokers. This suggests that the pathogenesis of lung cancers arising in smoker and never smokers are distinct.
Ethnicity was available for 2,347 cases and the frequency of EGFR mutations was significantly different between East Asian patients (33%) and non-Asian patients (6%) (p < 0.001; Fig. 1c). When limited to adenocarcinoma cases, East Asian patients had a 48% (270 of 563) mutation rate, while other ethnicities had a 12% (63 of 519) rate (Fig. 1d). Of interest, 4 of 5 Asian patients with lung cancer in the United States and Australia, in our study, had EGFR mutations, suggesting that genetics may be more important than geographic factors for affecting EGFR mutations, although the number is small. Several polymorphic variations in EGFR gene including a CA repeat in intron 1 or single nucleotide polymorphisms in the promoter region were identified and analyzed,25, 26 and some of these polymorphisms help regulate EGFR expression level, and their distributions demonstrate ethnic variations. It remains to be determined if these polymorphisms contribute toward higher mutational rates in certain subpopulations.
Gender data were available for 2,252 cases and the frequency of EGFR mutations was also significantly different between females (38%) and males (10%) (p < 0.001) (Fig. 1c). When limited to adenocarcinoma cases, female patients had a 49% rate (203 of 411) and male patients had a 19% rate (130 of 671) (Fig. 1d). In two studies and ours based on Japanese subjects, over 50% of female lung cancer patients had EGFR mutations. The role of gender in the frequency of EGFR mutations is not understood, although it is postulated that sex hormones or environmental factors may be responsible.
All the four features affecting EGFR mutations in lung cancers appear to be important, and these are also known to predict for response to gefitinib treatment.27, 28 Because adenocarcinoma histology is more frequent in NSCLC arising in East Asian countries, never smokers and women, the independent contribution of each of the 4 identified associated factors (histology, ethnicity, smoking history and gender) needs to be evaluated in a large multivariate analysis. We are in the process of performing such a study. Although EGFR mutations are strongly correlated to clinical response to TKIs, the correlation is not absolute and mutations are not sole factors. While one-third of patients in a large phase II clinical trial achieved stable disease (SD) that may contribute to the overall survival,27, 28 we do not know the precise correlation between EGFR mutations and SD. Other mechanisms, including gene amplifications, interactions with other EGFR gene family members, or effects of other downstream signal molecules, may influence response to TKIs. Allelic imbalance of EGFR locus (loss of wild-type allele or selective amplification of mutant allele) is more frequent in cases with EGFR mutation than in cases without mutation (authors' unpublished data). In addition, tumors with EGFR mutation had increased gene copy number.29 Amplification of other members of the EGFR gene family have also been detected in tumors and cell lines (authors unpublished data in collaboration with Marileilla Garcia). Thus, response to TKI is complex and may be regulated by multiple factors.
Mutation status of related genes
The EGFR family consists of 4 molecules including EGFR (HER1 or ERBB1), HER2 (EGFR2 or ERBB2/NEU), HER3 (EGFR3 or ERBB3) and HER4 (EGFR4 or ERBB4). The basic structure is very similar among family members, but each has distinct properties: HER2 has strong kinase activity but has no identified ligand and HER3 lacks kinase activity. Recently, somatic mutations of HER2 TK domain were also reported in lung adenocarcinomas.20, 30 Of interest, HER2 mutations were of the same type and targeted the same region (3′ of the αC-helix) in exon 20 as did EGFR in-frame duplications/insertions20 (Figs. 2a and 2b). Further, HER2 mutations were also associated with East Asian ethnicity, female gender and never-smoker status.20 The remarkable similarities between mutations in these 2 genes are unprecedented.
KRAS encoding a small GTP binding protein is one of the well-documented oncogenes and is frequently activated by missense mutations in many human cancers. KRAS mutations were detected in ∼20% of NSCLC, especially in adenocarcinoma and in smokers. There are 3 published studies that analyzed both KRAS and EGFR mutation status in the same tumors (1,536 cases are available)6, 13, 14 and they indicate that EGFR and KRAS mutations are mutually exclusive. One study demonstrated that KRAS mutations are associated with a lack of TKIs sensitivity in lung adenocarcinoma.31KRAS binds to BRAF, and thus both genes are part of the EGFR family signaling cascade. However, BRAF mutations are rarely detected (0–3%) in lung cancer32, 33, 34 compared to KRAS mutations. We also analyzed BRAF mutations (exon 11 and 15) in over 400 cases, including cell lines and primary tumors, and found BRAF mutations in less than 3% (unpublished data). BRAF is a nonreceptor serine/threonine kinase, but its kinase domain has a structure similar as other protein kinases, including EGFR members (Figs. 2a and 2b). BRAF mutations are also located in the P-loop or A-loop as are some of the EGFR mutations. The V600 mutation in A-loop is the most frequent type of BRAF mutation in human cancers.35 Interestingly, this mutation occurs in a similar position to the L861Q mutation in EGFR, and this mutation is present in the viable germline dark skin 5 mutation in the mouse.36 In our analysis, although the frequency of HER2 or BRAF mutations are relatively low (Fig. 2c), the mutations of these 4 genes (EGFR, HER2, KRAS and BRAF) are mutually exclusive. Of interest, mutations of RET, RAS and BRAF are mutually exclusive in thyroid papillary cancer,37 as are KRAS and BRAF in colorectal cancers.38, 39 These results indicate that simultaneous mutations of multiple genes in the same signaling pathways are not required for lung cancer pathogenesis as well as for other type of cancer, and a single mutation in any of the 4 genes may suffice. The finding that EGFR and HER2 gene mutations target never smokers while KRAS mutations favor smokers suggest that adenocarcinomas in smokers and never smokers arise via different pathogenic pathways.
Mechanisms of TKIs resistance
Despite initial responses, sometimes dramatic, patients treated with TKIs eventually develop recurrent disease resistant to further TKIs therapy. Recent reports demonstrated that a second mutation contributes to this acquired resistance to TKIs.40, 41 They demonstrated that 4 out of 7 relapsed patients had a second mutation, T790M in exon 20 of the TK domain, and mutant constructs with T790M demonstrated resistance to TKIs while preserving the kinase activity in vitro. Interestingly, one resected primary tumor13 and the NCI-H1975 NSCLC cell line had the T790M mutation along with the activating L858R mutation in exon 21. Since both cases had not been treated with TKIs, the T790M mutation in these cases appears to be induced spontaneously rather than following TKIs treatment. These findings raise the question as to which stage in the progression of lung cancer does the T790M mutation occur and what is the other mechanism for resistance in cases lacking this second mutation. More than 10 different mutations within BCR–ABL kinase domain spanning P-loop and A-loop have been detected with resistance to imatinib.42 These results suggest that the possibility that mutations other than T790M and that are not identified as yet in EGFR gene, or gene amplifications (rarely found in BCR–ABL gene correlated with imatinib resistance) contribute to TKI resistance in relapsed lung cancer patients.
Although the limitations of available samples from relapsed patients make it difficult to conduct large scale studies, re-biopsy from relapsed region in patients treated with TKIs should be considered as a standard procedure to elucidate the significance of T790M as well as other possible mutations or genetic alterations.
From the mutation analyses of large numbers of NSCLC samples, several features of mutations of EGFR and related genes are now becoming apparent and they may be correlated to sensitivity and resistance to TKIs therapy. However, several unsolved questions remain and their answers may provide important clinical insights.
Recent findings about somatic mutations in HER2 and resistance-associated mutations in EGFR provide a new basis for the development of TKIs for NSCLC. Because the 2 genes have closely related TK domains and identical mutations are predicted to result in the same cellular effects, targeted therapy for HER2 should be possible in mutated lung cancers as well as EGFR targeted therapy. Further, nearly 50% of adenocarcinoma cases had mutations of either 1 of the 4 genes (Fig. 2d), indicating the possibility to inhibit each gene's activities by specific inhibitors. Routine genetic screening after diagnostic biopsy or surgical resection should be needed for true “targeted therapies” in each lung cancer patients. For a resistance-associated mutation, modified inhibitors that overcome resistance may prove beneficial, as seen in imatinib refractory chronic myeloid leukemia cases.43 Several specific ERBB TKIs, pan-ERBB TKIs, monoclonal antibodies and other type of inhibitors for kinases have also been developed and are in various phases of clinical trials.44, 45 A full understanding of the epidemiology and molecular biology of EGFR and its family members and their interrelationships are essential for the selection of the subpopulations most likely to benefit, including the development of new targeted drugs and for interpreting the molecular pathogenesis of lung cancers.