Abnormalities of epidermal growth factor receptor in lung squamous-cell carcinomas, adenosquamous carcinomas, and large-cell carcinomas

Tyrosine kinase domain mutations are not rare in tumors with an adenocarcinoma component

Authors


Abstract

BACKGROUND.

Tyrosine kinase domain (TKD) gene mutations of the epidermal growth factor receptor gene (EGFR) have proven to be clinically significant in nonsmall-cell lung cancer (NSCLC), particularly in adenocarcinoma. However, TKD mutations together with deletion mutations in the extracellular domain of EGFR (EGFRvIII) have not been fully investigated in NSCLC except for adenocarcinoma. The present study sought to gain further insight into the significance of EGFR mutations in NSCLC by focusing on nonadenocarcinoma NSCLC.

METHODS.

EGFR TKD mutations were investigated using direct sequencing and mutation-specific polymerase chain reaction (PCR), and EGFRvIII mutations were examined using reverse transcriptase-PCR in samples from 42 NSCLC patients and 6 NSCLC cell lines excluding adenocarcinoma.

RESULTS.

EGFR TKD mutations were detected in 1 of 7 (14%) squamous-cell carcinomas with an adenocarcinoma component and 2 of 4 (50%) adenosquamous carcinomas. In contrast, EGFR TKD mutations were not identified in 24 pure squamous-cell carcinomas without any adenocarcinoma component, 7 large-cell carcinomas, or 6 cell lines. EGFRvIII was detected solely in 1 of 7 large-cell carcinomas (14%), but not in 31 squamous-cell carcinomas, 4 adenosquamous carcinomas, or 6 cell lines.

CONCLUSIONS.

These results suggest that EGFR TKD mutations are found in NSCLCs with an adenocarcinoma element. Patients with such lesions are thus considered candidates for molecular therapies targeting EGFR. Cancer 2007 © 2007 American Cancer Society.

Various mutations within the epidermal growth factor receptor gene (EGFR) have recently been found in nonsmall-cell lung cancer (NSCLC).1–17 These mutations tend to cluster in the tyrosine kinase domain (TKD) of the gene. The potential impact of EGFR mutations in NSCLC has attracted substantial attention from researchers and clinicians because the presence of mutations may critically influence the effects of tyrosine kinase inhibitors (TKIs), such as gefitinib and erlotinib, against lung cancer. Numerous studies have suggested that gefitinib is more effective against tumors harboring EGFR mutations than against tumors without such mutations, and similar results have been reported for erlotinib, although these hypotheses have yet to be confirmed in randomized controlled studies.1–3, 6–9, 11

These types of mutations are widely recognized as occurring more frequently in adenocarcinomas than in tumors of other histology.2–17 The frequency of EGFR TKD mutations in NSCLC except for adenocarcinoma is reported to be generally low (Table 1). These results lead clinicians to refrain from using gefitinib on patients with nonadenocarcinoma NSCLC. However, a subset of nonadenocarcinoma NSCLCs has been reported as harboring EGFR mutations in several studies.2–5, 7–11, 14, 16, 17 These findings prompted us to investigate whether nonadenocarcinoma NSCLCs with EGFR mutations display any specific clinical features.

Table 1. Frequency of EGFR TKD Mutations in NSCLC
ADSQLAADSQOthersAuthors
MUTNMUTNMUTNMUTNMUTN
  1. AD indicates adenocarcinoma; NSCLC, nonsmall-cell lung cancer; SQ, squamous cell carcinoma; LA, large cell carcinoma; ADSQ, adenosquamous carcinoma; MUT, number of cases with EGFR TKD mutations.

15 (21%)701 (2%)49Paez et al.
37 (17%)2139 (5%)178Bell et al.
114 (39%)2896 (3%)230Shigematsu et al.
36 (46%)792 (3%)17Suzuki et al.
23 (26%)880 (0%)25Pao et al.
39 (57%)690 (0%)241 (13%)8Huang et al.
29 (67%)434 (40%)100 (0%)1Chou et al.
14 (22%)651 (5%)212 (50%)4Han et al.
115 (54%)2150 (0%)151 (9%)11Tam et al.
12 (21%)581 (5%)210 (0%)12 (22%)9Cappuzzo et al.
60 (56%)1080 (0%)310 (0%)90 (0%)6Sonobe et al.
39 (10%)3750 (0%)4540 (0%)31Marchetti et al.
37 (45%)820 (0%)350 (0%)21 (100%)1Tokumo et al.
29 (40%)720 (0%)450 (0%)20 (0%)1Tomizawa et al.
110 (49%)2240 (0%)350 (0%)91 (20%)50 (0%)4Kosaka et al.
136 (42%)2240 (0%)1021 (4%)270 (0%)11 (25%)4Sugio et al.

Deletion of exons 2–7 in EGFR gene (EGFRvIII), which is often found in glioblastoma,18 has also been identified in a subset of NSCLC patients, although the frequency of this mutation appears to vary (Table 2).19–22 EGFRvIII is also a possible target of TKIs, and revealing the presence of EGFRvIII mutation in NSCLC is thus considered clinically relevant.22 However, EGFRvIII has not been investigated in NSCLC as intensively as EGFR TKD mutations. We previously reported that, similar to patients in Western countries, EGFRvIII is very rare in lung adenocarcinoma among Japanese.23 The present study also investigated EGFRvIII in nonadenocarcinoma NSCLC to elucidate whether EGFRvIII has any contribution to tumorigenesis for this type of lung cancer.

Table 2. Frequency of EGFRvIII in NSCLC
ADSQLAADSQOthersAuthors
MUTNMUTNMUTNMUTNMUTN
  1. AD indicates adenocarcinoma; NSCLC, nonsmall-cell lung cancer; SQ, squamous cell carcinoma; LA, large cell carcinoma; ADSQ, adenosquamous carcinoma; MUT, number of cases with EGFRvIII.

0 (0%)102 (15%)132 (100%)21 (14%)7Garcia de Palazzo et al.
0 (0%)260 (0%)320 (0%)7Jungbluth et al.
19 (41%)4610 (42%)241 (17%)6Okamoto et al.
0 (0%)1233 (5%)56Ji et al.

Furthermore, overexpression of epidermal growth factor receptor (EGFR) has also been shown to be related to EGFR mutations23 and susceptibility to TKIs in NSCLC.11 We therefore investigated EGFR expression in these samples to explore whether this has any impact on nonadenocarcinoma NSCLC.

Overall, this study was intended to reveal whether nonadenocarcinoma NSCLCs contain EGFR abnormalities and can thus be possible targets of molecular therapy. We demonstrate herein that EGFR TKD mutations are not rare in NSCLCs with an adenocarcinoma component and EGFRvIII can be found in patients with large-cell carcinoma.

MATERIALS AND METHODS

Patients

A total of 96 consecutive Japanese patients with NSCLC underwent surgery in the Department of Thoracic Surgery at Kyorin University Hospital between May 2001 and March 2003. Postoperative diagnoses were made by trained pathologists. Patients diagnosed with squamous-cell carcinoma, adenosquamous carcinoma, or large-cell carcinoma were enrolled for further analysis. Pathological diagnosis was based on the criteria of the World Health Organization classification system.24 Briefly, squamous-cell carcinoma comprises tumors with <10% adenocarcinoma component. In particular, squamous-cell carcinoma without a detectable adenocarcinoma component was termed ‘pure squamous-cell carcinoma’ in this study. Adenosquamous carcinoma comprises tumors with a 10% to 90% adenocarcinoma component. Adenocarcinoma comprises tumors with a >90% adenocarcinoma component, and patients with this diagnosis were excluded from the present study. After surgery, some patients underwent chemotherapy and/or radiotherapy with various regimens, including gefitinib therapy in 5 patients with recurrence. Written informed consent to analyze tissue DNA, RNA, and protein was obtained from each patient before operation. Results of molecular analyses on these patients have not previously been reported.

Patient Data

Clinical data were obtained from in- and outpatient medical records. The following criteria were used to classify smoking status: never smoker, patients who had smoked <100 cigarettes in their lifetime; former smoker, patients who had stopped smoking ≥12 months before diagnosis; and current smoker.

Cell Lines and Clinical Samples

Lung squamous-cell carcinoma cell lines EBC-1, LK2, and Sq-1 and large-cell carcinoma cell lines 86-2 and Lu99 were obtained from the Cell Resource Center for Biomedical Research, Tohoku University (Sendai, Japan). Squamous-cell carcinoma cell line RERF-LC-AI was purchased from Riken Bioresource Center (Tsukuba, Japan). Lung adenocarcinoma cell lines NCI-H1650 and H1975, which have been reported to show delE746-A750 and L858R, respectively, were purchased from the American Type Culture Collection (Manassas, VA) and used as positive controls for EGFR TKD mutations.25 The glioblastoma cell line U87MGΔEGFRSH, which has been reported to show EGFRvIII, was kindly donated by Professor Webster K. Cavenee (Ludwig Institute for Cancer Research, San Diego) and was used as a positive control for EGFRvIII mutation.26 Tumor samples and visually normal lung tissues distant from the tumor were immediately frozen after resection and preserved at −80°C. Visually normal lung tissues were confirmed as containing no tumor component on pathological examination. DNA, RNA, and protein were extracted from these samples and cell lines according to methods previously described.23

Direct Sequencing of EGFR TKD

Mutations of EGFR in lung cancer cluster within exons 18–21, a main portion of TKD.1, 2 Direct sequencing analysis of EGFR from exons 18–21 using genomic DNA was therefore performed, as previously reported,23 for all NSCLC samples, including the 8 NSCLC cell lines.

EGFR TKD Mutation-Specific PCR

Lung cancer samples are frequently intermingled with large amounts of normal tissue. In addition, several NSCLC cell lines tested herein reportedly harbor an L858R mutation in only a minor population of cells (1%–10%). The sensitivity of sequencing analysis, which usually needs >30% mutant DNA, was thus insufficient to detect mutations in such contaminated tumor samples. A mutation-specific polymerase chain reaction (PCR) method that we previously developed was thus used to detect the major EGFR TKD mutations with high sensitivity. This method can simultaneously detect delE746-A750 and L858R, which together account for approximately 70% of EGFR TKD mutations, with a low percentage of mutant DNA (2.5% for delE746-A750 and 0.25% for L858R) being detectable. The precise method has been described elsewhere.27 In brief, PCR was performed with 100 ng of extracted DNA and primers specific for delE746-A750 and L858R were added together with the PCR mix. PCR conditions were as follows: 95°C for 5 minutes; then 45 cycles of 95°C for 30 seconds, 56°C for 30 seconds, and 72°C for 30 seconds; followed by a final 10 minutes at 72°C. Consequently, a 153-bp band and a 104-bp band are detected in samples containing delE746-A750 and L858R, respectively. For the detection of delT751-K758, we set up PCR amplifying the area of DNA including this deletion. This PCR can clearly discriminate a short form of DNA with this deletion (241 bp) from germline DNA (265 bp).

Analysis of EGFRvIII

EGFRvIII is generated by total deletion of exons 2–7 (801 bp) in the extracellular domain of the EGFR gene. Reverse transcriptase PCR (RT-PCR) is reportedly useful in detecting this deletion in glioblastoma,18 and was thus used in this study. For detecting the 801-bp deletion, tumor cDNA and primers were subjected to 40 cycles of PCR amplification. PCR primers used were as follows: sense, 5′-GTA TTG ATC GGG AGA GCC G-3′; antisense, 5′-GTG GAG ATC GCC ACT GAT G-3′.23 In addition, direct sequencing was used to confirm the presence of EGFRvIII in samples showing positive results according to RT-PCR. The sense primer for RT-PCR was also used as a primer for sequencing analysis.

Western Blotting

Western blotting was used to analyze expression of EGFR protein rather than immunohistochemistry, as the results of the latter method are considered less quantitative and reproducible than the former.28 For EGFR Western blotting analysis, 100 μg of tumor or normal lung protein was used. Protein samples were subjected to Western blotting analyses using anti-EGFR polyclonal antibody (Santa Cruz Biologicals, Santa Cruz, CA) according to the manufacturer's instructions. The level of expression in each tumor was determined as follows: (0), very weak or no EGFR band (170 kD), similar to normal lung; (1), easily visible band; and (2), very strong band similar to levels of β-actin. An EGFR expression level of 1 or 2 was defined as indicating EGFR overexpression.23

Microdissection

To elucidate which component includes EGFR TKD mutations in tumors containing both adenocarcinoma and squamous-cell carcinoma components, separate mutational analysis of each component was performed using microdissection techniques. For adenosquamous carcinomas and squamous-cell carcinomas with an adenocarcinoma component, a trained expert pathologist performed laser capture microdissection using LM200 (Arcturus, Mountain View, CA) and manual microdissection of both squamous-cell carcinoma and adenocarcinoma component on 8 μm-thick hematoxylin and eosin (H&E)-stained, formalin-fixed, paraffin-embedded histology sections. DNA was extracted from each microdissected tissue by incubation with proteinase K (200μg/mL) for 24 hours at 37°C, and EGFR TKD mutation-specific PCR was performed using DNA extracted from both squamous-cell carcinoma and adenocarcinoma components.

Statistical Analysis

The significance of differences in categorical data was tested using the χ2 test or Fisher exact test.

RESULTS

Patient Characteristics

Postoperative pathological examinations revealed 42 nonadenocarcinoma NSCLCs (31 squamous-cell carcinomas, 7 large-cell carcinomas, and 4 adenosquamous carcinomas) among 96 NSCLC tumors. The 31 squamous-cell carcinomas comprised 24 pure squamous-cell carcinomas (77%) and 7 squamous-cell carcinomas with an adenocarcinoma component (23%). Patient characteristics in the present study are summarized in Table 3. Patients comprised 39 men (93%) and 3 women (7%) with a median age of 73 years (range, 40–84 years), and 30 current smokers (71%), 8 former smokers (19%), and 4 never smokers (10%).

Table 3. Characteristics and EGFR Abnormalities of Patients
CaseAgeSexSmokingHistologyDifferentiationAdenocarcinoma componentEGFR mutationEGFR expressionPostoperative recurrenceGefitinib responsePrognosis
  1. M indicates male; F, female; C, current smoker; F, former smoker; N, never smoker; Number, pack-year smoking; SQ, squamous cell carcinoma; ADSQ, adenosquamous carcinoma; LA, large cell carcinoma; well, well differentiated; mod, moderately differentiated; poor, poorly differentiated; L858R, nt. 2819T>G; del 1, del E746-A750, del nt. 2481–2495; del 6, del T751-K758,del nt. 2496–2519; I759N, nt. 2522A>T; EGFRvIII, deletion of exons 2–7; PD, progressive disease; NE, not evaluable; A, alive; D, dead; U, unknown.

159MC60SQmod  2  A
270MC76.5SQwell  1  A
378MF69SQmod  2yes U
480MC120SQmod  1yes U
555MC30SQwell  0yes D
661FC30SQmod  0yes D
761MC50SQmod  1  A
852MC100SQmod  2  U
975MC100SQ   2yes D
1074MC40SQpoor  2yesPDD
1179MC90SQ   2  A
1269MNSQwell  0  D
1369MC80SQ   1  D
1468MF15SQmod  1  A
1574MC55SQmod  2  A
1670MC75SQmod  1yesPDD
1775MC50SQmod  2  A
1874MC50SQmod  2  U
1977MF159SQpoor  2yes D
2069MC40SQmod  2yes D
2172MF30SQpoor  1yes A
2274MC90SQpoor  2  A
2363MC37.8SQpoor  1  A
2481MC25.2SQmod  0  A
2567MF108SQpooryesL858R2yesNED
2673MC50SQpooryes 1yes U
2754MC30SQmodyes 2  A
2866MC62.5SQmodyes 2  A
2974MF77.5SQwellyes 1yes D
3073MNSQ yes 0yes A
3184MC47.3SQpooryes 0yes D
3279MF50ADSQwellyesdel 11yes U
3374FNADSQwellyesdel 6 and I759N0  A
3477MCADSQpooryes 2  U
3573MC100ADSQ yes 2yes D
3644MC43.5LA  EGFRvIII2  D
3783FF20.5LA   2yes D
3882FNLA   2  A
3940MC22LA   1yesNEU
4054MC52.5LA   1yesPDA
4147MC50.5LA   2  U
4261MC30LA   2yes D

EGFR Mutations in Cell Lines

Sequencing analysis revealed delE746-A750 mutation in H1650 and L858R mutation in H1975, as previously reported.23 In contrast, no squamous-cell carcinoma cell lines or large-cell carcinoma cell lines displayed EGFR TKD mutations even according to mutation-specific PCR (data not shown). The absence of L858R mutations in squamous-cell carcinomas tested in the present study (LK2, EBC-1, Sq-1, and RERF-LC-AI) was inconsistent with the previous study demonstrating that these lines harbor the L858R mutation in various proportions of cells (LK2, 1%; EBC-1 and Sq-1, 10%; RERF-LC-AI, 100%).29 RT-PCR and direct sequencing revealed the EGFRvIII mutation only in the U87MGΔEGFRSH glioblastoma cell line, but not in any lung cancer cell lines (Fig. 1).

Figure 1.

EGFRvIII mutation. (A) Results of reverse-transcriptase polymerase chain reaction (RT-PCR) for detecting EGFRvIII. The U87MGΔEGFRSH glioblastoma cell line showed a short 352-bp band, indicating the presence of EGFRvIII. One large cell carcinoma sample (Case 36) showed a weak 352-bp band together with wildtype allele. (B) Direct sequencing (sense direction) of EGFRvIII in U87MGΔEGFRSH. Mutant allele is dominant. (C) Direct sequencing (sense direction) of EGFRvIII in Case 36. Low signals for mutant DNA are visible.

EGFR TKD Mutations in Patient Samples

EGFR mutations in TKD found in patient samples in this study are summarized in Table 3. Mutations within exons 18–21 were found in 1 of 31 squamous-cell carcinomas (3%; including 7 squamous-cell carcinomas with an adenocarcinoma component), and in 2 of 4 adenosquamous carcinomas by direct sequencing. No cryptic mutations were detected by mutation-specific PCR in tumor samples negative for mutations by direct sequencing. Case 33 had both delT751-K758 and I759N mutations, although the biological significance of the latter mutation is considered unclear. Interestingly, the sole case of squamous-cell carcinoma with TKD mutation displayed an adenocarcinoma component. These findings indicate that all mutated cases were tumors with an adenocarcinoma component, and none of the 24 ‘pure squamous-cell carcinoma,’ cases showed TKD mutation (Table 3). When the squamous-cell carcinoma with an adenocarcinoma component and adenosquamous carcinoma were combined, 3 of 11 such cases (27%) were shown to have TKD mutations. The difference in frequency of TKD mutations between tumors with or without an adenocarcinoma component was significant (P = .007). No large-cell carcinomas contained an adenocarcinoma component or displayed TKD mutations. In addition, no significant correlations between EGFR TKD mutations and clinical features such as patient age (P = .256), gender (P = .145), and smoking status (P = .145) were identified.

Microdissection

Microdissection analysis was performed on 3 samples (2 adenosquamous carcinomas and 1 squamous-cell carcinoma with an adenocarcinoma component) with EGFR TKD mutations. By performing mutation-specific PCR separately for both squamous and adenocarcinoma components, mutations were detected in both components in all 3 cases (Fig. 2).

Figure 2.

Results of microdissection and mutation-specific polymerase chain reaction (PCR). (A-C) Histological presentation of microdissection in 3 cases with EGFR mutations. Left: Pictures representing both squamous-cell carcinoma and adenocarcinoma components. SQ: squamous-cell carcinoma component; AD: adenocarcinoma component. Right: Upper and lower pictures show microdissected adenocarcinoma and squamous-cell carcinoma components, respectively. (D) Mutation-specific PCR for delE746-A750 and L858R in microdissected tissues. A 153-bp band indicating delE746-A750 is present in both squamous-cell carcinoma and adenocarcinoma components in Case 32. A 104-bp band indicating L858R is present in both components in Case 25. H1650: Positive control for delE746-A750; H1975: Positive control for L858R. (E) PCR detection of delT751-K758 in Case 33. A 241-bp band indicating delT751-K758 is present in both squamous and adenocarcinoma components in this case. H1975: Negative control for delT751-K758 showing a wildtype 265-bp band.

EGFRvIII in Patient Samples

RT-PCR of the EGFR extracellular domain in 42 samples revealed that only 1 case of large-cell carcinoma displayed a short 352-bp band indicating EGFRvIII mutation (Table 3). Direct sequencing revealed that this case included a minor population of EGFRvIII (Fig. 1B,C). This case also showed a normal 1153-bp band in addition to a 352-bp band.

EGFR Expression in Patient Samples

EGFR overexpression was present in 25 of the 31 squamous-cell carcinomas (81%), 3 of 4 adenosquamous carcinomas (75%), and 7 of 7 large-cell carcinomas (100%) (Fig. 3, Table 3). EGFR TKD mutations were identified in 3 of the 35 samples with EGFR overexpression (9%) and 1 of 7 samples without overexpression (14%). No significant correlation was noted between EGFR overexpression and mutation (P = .421).

Figure 3.

Analysis of epidermal growth factor receptor (EGFR) protein expression by Western blotting. EGFR protein expression was classified into 3 levels: (2), (1), and (0). Three representative samples (each corresponding to expression level (2), (1), and (0), respectively) together with a normal lung sample representing expression level (0) and the cell line H1650 representing level (2) are shown on the same blot. Expression of β-actin of the corresponding samples is shown below.

DISCUSSION

This study analyzed EGFR gene mutations in detail in nonadenocarcinoma NSCLCs. Most strikingly, EGFR TKD mutations were found exclusively in tumors with an adenocarcinoma component, including adenosquamous carcinomas. Although many studies have reported that EGFR TKD mutations are exceptional in squamous-cell carcinoma, our results suggest that these mutations are not very rare in squamous-cell carcinoma with an adenocarcinoma component and adenosquamous carcinoma. Whereas no reliable data appear to be available regarding the frequency of squamous-cell carcinomas with an adenocarcinoma component, adenosquamous-cell carcinoma reportedly comprises 3% of all NSCLCs.30 Our results suggest that a substantial portion of nonadenocarcinoma NSCLCs may harbor an EGFR mutation, and thorough histologic investigation to identify an adenocarcinoma component or genetic analyses to identify EGFR mutations are warranted for squamous-cell carcinomas.

Previous studies investigating EGFR mutations have not precisely described whether the squamous-cell carcinomas analyzed included an adenocarcinoma component, and many have even failed to separately analyze adenosquamous carcinomas. Actually, concordant with our results, studies discriminating adenosquamous carcinomas from squamous-cell carcinomas have uniformly shown that only the former type of tumors harbor mutations.14, 16 In the present study, microdissection analysis was performed to explore whether these mutations are restricted to an adenocarcinoma component. Consistent with the previous report,31EGFR TKD mutations were found in both adenocarcinoma and squamous-cell carcinoma components. These findings may suggest that squamous-cell carcinomas with an adenocarcinoma component are intrinsically different from pure squamous-cell carcinomas, and that an EGFR TKD mutation might have occurred in common progenitor cells destined to become both cell types in these tumors. However, the possibility remains that these findings are simply the result of artifacts due to contamination between the 2 cell types during analyses. Further studies are required to investigate biological differences between pure squamous-cell carcinomas and squamous-cell carcinomas with an adenocarcinoma component.

In addition to implications in tumorigenic mechanisms of EGFR TKD mutations, these results are also clinically important because patients with EGFR TKD mutations may be susceptible to TKI treatment. To date, no studies have precisely analyzed gefitinib sensitivity for squamous-cell carcinoma with an adenocarcinoma component. Similar to adenocarcinomas with mutations, the squamous-cell carcinoma component of these types of tumors with mutations may be susceptible to gefitinib therapy regardless of histology. Because of the small number of patients treated using gefitinib in this cohort, the efficacy of gefitinib could not be analyzed in our patients. Future studies with large subject populations will clarify this issue.

RT-PCR analysis of the extracellular domain of EGFR revealed that, similar to adenocarcinomas analyzed previously,23 EGFRvIII is very rare in nonadenocarcinoma NSCLCs. In contrast, several reports have demonstrated that as much as 32% of NSCLCs, irrespective of histology, show this type of mutation.19, 21 This discrepancy may be attributable to the methods applied to detect EGFRvIII. The quality of antibody might influence rates of this mutation in previous studies using immunohistochemistry with EGFRvIII antibody. Most recently, a study investigating EGFRvIII in a large number of patients using RT-PCR has shown that, consistent with our results, this mutation is very rare in NSCLCs and found only in nonadenocarcinomas. These findings indicate that EGFRvIII contributes to tumor development in only a minor subset of NSCLCs.

In addition to EGFR mutations, we also analyzed EGFR protein expression in nonadenocarcinoma NSCLCs using Western blotting. EGFR protein overexpression was found in 35 of 42 nonadenocarcinoma NSCLCs (83%). Patients in our cohort thus showed a higher frequency of EGFR protein overexpression than those in a previous meta-analysis,28 which may be because of differences in ethnicity. Whereas we have previously reported that EGFR overexpression is strongly correlated with TKD mutations in adenocarcinomas,23 no such correlation was demonstrated in the present study for nonadenocarcinoma NSCLCs. This may be explained by differences in the histology of the tumors, although a definitive conclusion cannot be reached because of the small number of patients with TKD mutations in the present study. One recent report suggests that EGFR protein overexpression contributes more strongly to TKI susceptibility than EGFR TKD mutations.11 Although gefitinib has failed to show clinical efficacy in nonadenocarcinoma NSCLC patients,32 other TKIs such as erlotinib or anti-EGFR antibody drugs may prove beneficial for tumors with EGFR overexpression.

In conclusion, we clarified that a subset of nonadenocarcinoma NSCLCs contains EGFR mutations. In particular, EGFR TKD mutations are not rare events in NSCLCs with an adenocarcinoma component. These molecular abnormalities will not only provide insights into the tumorigenesis and oncologic properties of these tumors, but will also render them possible targets for molecular therapies.

Acknowledgements

We thank Professor Webster K. Cavenee of the Ludwig Institute for Cancer Research at San Diego for the kind gift of glioblastoma cell line U87MGΔEGFRSH, and Motoo Nagane, MD, PhD, from the Department of Neurosurgery at Kyorin University Hospital, for technical advice. We also thank Yoichi Kameda, MD, PhD, Kazuo Masui, MD, from the Division of Pathology, and Tsutomu Yoshida, BSc, from the Laboratory for Molecular Diagnostics at Kanagawa Prefectual Cancer Center Institute, for technical assistance in performing the laser capture microdissection

Ancillary