Protein overexpression and gene amplification of epidermal growth factor receptor in nonsmall cell lung carcinomas

An immunohistochemical and fluorescence in situ hybridization study


  • Shioto Suzuki M.D.,

    1. Department of Pathology, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Yamanashi, Japan
    Search for more papers by this author
  • Yoh Dobashi M.D.,

    Corresponding author
    1. Department of Pathology, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Yamanashi, Japan
    • Department of Pathology, Faculty of Medicine, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, 1110 Shimokato, Tamaho, Nakakoma, Yamanashi 409-3898, Japan
    Search for more papers by this author
    • Fax: (011) 81 552736727

  • Hiroyuki Sakurai M.D.,

    1. Department of Surgery, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Yamanashi, Japan
    Search for more papers by this author
  • Keiichi Nishikawa M.D.,

    1. Department of Internal Medicine, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Yamanashi, Japan
    Search for more papers by this author
  • Mitsuhiko Hanawa M.D.,

    1. Department of Pathology, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Yamanashi, Japan
    Search for more papers by this author
  • Akishi Ooi M.D.

    1. Department of Pathology, Interdisciplinary Graduate School of Medicine and Engineering, University of Yamanashi, Yamanashi, Japan
    Search for more papers by this author



Recently, molecular therapies targeting epidermal growth factor receptor (EGFR) have been developed for clinical use. The current study was conducted to determine 1) the exact frequency of EGFR protein overexpression, 2) the correlation between protein overexpression and EGFR amplification, and 3) the correlation between the status of the genetic and clinicopathologic features in nonsmall cell lung carcinomas (NSCLC).


In total, 181 NSCLC samples were examined immunohistochemically using an antibody against EGFR, and tumor cells that exhibited overexpression were examined further for EGFR amplification by fluorescence in situ hybridization.


Overexpression of EGFR protein was found in 34% of the tumors. Among these, EGFR amplification was demonstrated in 74%. High-level gene amplification was found exclusively in tumors cells with high protein expression. In most of these tumors, cells that exhibited EGFR overexpression and gene amplification were distributed heterogeneously, even within a single tumor nodule. Statistically, EGFR overexpression was correlated significantly with lymph node metastasis and with a more advanced pathologic stage. Moreover, in adenocarcinomas, gene amplification was correlated significantly with lymph node metastasis and tended to be correlated with a more advanced pathologic stage.


The overexpression of EGFR in NSCLC was accompanied predominantly, but not exclusively, by gene amplification. It is important to evaluate not only protein overexpression but also the EGFR status to design adjuvant therapies for patients with NSCLC, because specimens that exhibit both protein overexpression and gene amplification may predict eventual lymph node metastasis and, possibly, aggressive tumor behavior. Cancer 2005. © 2005 American Cancer Society.

The epidermal growth factor receptor (EGFR) gene, which is located on chromosome 7p12, encodes a 170-kD membrane glycoprotein. Upon binding of specific ligands, such as epidermal growth factor and transforming growth factor-α, the receptor dimerizes, which leads to receptor autophosphorylation and activation of a signal cascade. This results in changes in gene and protein expression that are crucial to tumor progression, including proliferation, decreased apoptosis, angiogenesis, and invasion.

EGFR was purified initially from the human squamous cell carcinoma (SCC) cell line, A431,1 which overexpresses EGFR 2–100-fold due to a commensurate 3–110-fold increase in EGFR copy numbers.2 Since then, it has been shown that many types of epithelial malignancies, including head and neck, ovarian, cervical, bladder, and esophageal carcinomas, express increased levels of EGFR on the cell membrane, as determined mainly by immunohistochemistry (IHC).3 However, there has been much controversy about EGFR expression in nonsmall cell lung carcinoma (NSCLC): The reported frequencies of EGFR overexpression have varied from 34% to 84%.4, 5 In addition, analyses of the correlations between protein overexpression and clinicopathologic factors or patient survival have produced contradictory results: Several different groups reported that EGFR overexpression was associated with either shorter or longer survival.3

Some studies have suggested that the mechanism of EGFR overexpression differs depending on the tumor type: Southern blot analysis has indicated that gene amplification is a major cause of EGFR overexpression in glioblastoma6 and in head and neck carcinomas,7 whereas it is uncommon in lung carcinoma,8 colon carcinoma,9 pancreatic carcinoma,10 and renal cell carcinoma.11 Using fluorescence in situ hybridization (FISH) combined with IHC, we previously reported that gene amplification is responsible primarily for EGFR overexpression, especially high-level overexpression, in gastric and colorectal carcinomas and in soft tissue sarcomas. However, we also found another group of patients in which lower level overexpression occurred without gene amplification.12–14 This led us to speculate that there are two different mechanisms responsible for EGFR overexpression in tumors, i.e., gene amplification and transcriptional/translational enhancement. The coexistence of these two mechanisms may explain why different studies have arrived at contradictory results concerning the frequency of EGFR overexpression and its association with clinicopathologic profiles in NSCLC. In this regard, it seems that FISH combined with IHC may be the best method for detecting gene amplification and/or protein overexpression. However, there has been only one previous study using this strategy to analyze EGFR aberrations in NSCLC,15 which showed that 60% of tumors exhibited overexpression, and 9% of tumors exhibited gene amplification.

EGFR recently has attracted much attention clinically due to the development of targeted therapies. IMC-C225 (cetuximab or Erbitux™; ImClone Systems Inc., Branchburg, NJ), a monoclonal antibody that targets EGFR, is now approved by the United States Food and Drug Administration for use in patients with colorectal carcinomas.16 Among the various small molecule inhibitors of tyrosine kinase, ZD1839 (Iressa™; AstraZeneca, Macclesfield, United Kingdom) has been approved for clinical use in patients with NSCLC17 in the United States. In Japan, where ZD1839 has been used in advance, it has garnered much attention for its dramatic effectiveness in some patients and for its unfortunate side effects, such as interstitial pneumonia in patients with NSCLC.18 Therefore, it is urgently necessary to ascertain the profiles of the target for these types of therapies to restrict treatment, if necessary, to patients who actually would benefit from the therapy.

For this report, we examined NSCLC samples for overexpression of EGFR protein by IHC and for gene amplification by FISH with the objectives of 1) determining the exact frequency of EGFR overexpression, 2) clarifying the relation between protein overexpression and gene amplification, and 3) clarifying the correlation between the status of the EGFR gene and clinical or pathologic features. The ultimate objective of these studies is to facilitate more successful EGFR-targeted therapies against NSCLCs.


Tissue Samples

We examined 181 specimens of NSCLC and concurrently excised lymph node metastases that were obtained from consecutive surgeries performed at the Department of Surgery, University of Yamanashi, between 1983 and 2004. The patients included 111 male patients and 70 female patients with a median age of 69 years (mean age, 67.3 years; age range, 38–85 years). The condition of each patient was assessed according to the system for staging primary tumor/regional lymph nodes/distant metastasis (TNM) described in the American Joint Committee on Cancer Staging Manual.19 The World Health Organization Classification of Tumors20 was used to determine histologic classification. The 181 patients were classified using the TNM classification system as follows: Stage IA, 48 patients; Stage IB, 48 patients; Stage IIA, 6 patients; Stage IIB, 26 patients; and Stage IIIA, 53 patients. Histologically, the specimens were classified into 119 cases of adenocarcinoma (AC), 57 cases of SCC, and 5 cases of large cell (LC) carcinoma. This laboratory study was approved by the Institutional Review Board at the University of Yamanashi, and written informed consent was obtained from all patients.


Resected lung samples were immersed immediately in 20% buffered neutral formalin, fixed overnight, and embedded in paraffin according to standard procedures. Serial sections (4 μm) that had been cut from representative paraffin embedded carcinoma tissues and placed onto silane-coated glass slides were used for hematoxylin and eosin staining, IHC detection of EGFR, and FISH analysis. IHC detection of EGFR was carried out on all primary tumors and on metastatic tumors of the lymph nodes. A monoclonal antibody against the external domain of human EGFR (Novocastra Laboratories, Newcastle, United Kingdom; working dilution, 1:20) was used. The specificity and sensitivity of the antibody against EGFR were verified previously.12, 13 For IHC detection, a high-temperature, antigen-unmasking technique was used, i.e., the section was autoclaved in 0.01 M citrate buffer, pH 7.0, at 121 °C for 10 minutes. Antibody was visualized by avidin-biotin binding to peroxidase-conjugated secondary antibodies (Dako Japan, Kyoto, Japan). In each analysis, a colon carcinoma section with previously confirmed EGFR overexpression13 was included as a positive control. EGFR positivity in the IHC analysis was reviewed by three pathologists (S.S., Y.D., and A.O.), who were not informed about the status of gene amplification. The intensity of reactivity was scored using a four-tier system, which we established in previous studies on EGFR as follows: negative, no discernible staining or background type staining; 1 +, definite cytoplasmic staining and/or equivocal discontinuous membrane staining; 2 +, unequivocal membrane staining with moderate intensity; and 3 +, strong and complete plasma membrane staining.12, 13 Samples that exhibited 2 + or 3 + immunostaining were scored as “overexpression,” because we previously demonstrated that only samples that showed 2 + or 3 + positivity in IHC analysis according to our criteria were associated frequently with gene amplification of EGFR.12–14 The extent (%) of positive staining cells was measured in the representative large section in each tumor.


FISH analysis was undertaken for all primary tumors with 2 + and 3 + staining, for their metastatic lymph nodes, and for 10 samples with 1 + staining. In addition, 10 tumors that scored negative for EGFR staining were selected at random. Gene amplification of EGFR (7p12) was determined using fluorescent-labeled DNA probe sets (LSI EGFR/CEP 7™; Vysis, Downers Grove, IL): A SpectrumOrange™-labelled, EGFR-specific probe and a SpectrumGreen™-labelled, centromeric probe that hybridizes to the centromeric region of the chromosome were used as controls to normalize copy numbers for chromosome 7. FISH was performed using standard methods, which were modified to incorporate an intermittent, short-term microwave treatment during the initial period of hybridization.21 In brief, 4-μm sections were deparaffinized, and this was followed by heating in a 0.01-M citrate buffer, pH 6.0, using a microwave processor (MI-77; Azumaya Company, Tokyo, Japan) for 10 minutes. After treatment in 0.2% pepsin/0.01 N HCl for 10 minutes at 37 °C, the samples were immersed in 0.1% NP-40/2 × standard saline citrate (SSC) for 10 minutes at 37 °C; then, their DNA was denatured by treatment in 70% formamide/2 × SSC for 5 minutes at 85 °C. Ten microliters of the probe solution were placed on a glass slide with a coverslip. The sample slides in the hybridization mixture were put in a microwave processor. Then, they were irradiated for 3 seconds at 2-second intervals (2.45 GHz, 300 Watts) with the temperature sensor set at 42 °C and hybridized at 42 °C overnight. Posthybridization was carried out according to the manufacturer's protocol. The tissue sections were counterstained with 4′,6-diamidine-2′-phenylindole dihydrochloride and p-phenylenediamine in phosphate buffered saline and glycerol (DAPI II) (Vysis) and examined with a fluorescence microscope (Olympus, Tokyo, Japan) equipped with a triple band pass filter set (Vysis) for DAPI II, SpectrumOrange, and SpectrumGreen and with filter sets specific to SpectrumOrange and SpectrumGreen. For positive controls, colon carcinoma tissues with previously confirmed EGFR were used.13 The numbers of EGFR and CEP7 signals were counted in each tumor nuclei and were evaluated according to the following definitions: basically, a cell in which the number of EGFR signals was greater than the number of CEP7 signals was interpreted as positive for amplification. Among those, 1) cells with a definite cluster of signals or with > 10 signals for EGFR were regarded as showing high-level amplification,12, 22, 23 and 2) a cell with 3–10 signals for EGFR was considered to have low-level amplification. In addition, a cell in which both centromeric and EGFR signals were increased equally was scored as polysomic.24 FISH images were taken using a photographic camera and were recorded on film slides.

Statistical Analysis

Agreement among observers in the interpretation of IHC specimens was qualified by κ statistics.25 In accordance with the criteria of Landis and Koch,26 the κ values were divided into several scales to evaluate the strength of agreement: κ < 0.00, poor; 0.00 < κ < 0.20, slight; 0.21 < κ < 0.40, fair; 0.41 < κ < 0.60, moderate; 0.61 < κ < 0.80, substantial; and 0.81 < κ < 1.00, nearly perfect. A chi-square test for independence was used to examine the correlation between the status of EGFR and the several pathologic and clinical features. Patient survival was analyzed by the Kaplan–Meier method with the log-rank test for univariate analysis.



Positive IHC staining for EGFR was confined almost exclusively to the tumor cells, whereas nonneoplastic bronchial and alveolar epithelium and stromal cells were negative. Overexpression (2 + and 3 + staining) of EGFR was found in 61 of 181 NSCLCs (34%), occurred somewhat more frequently in SCC than in AC (25 of 57 in SCCs vs. 36 of 119 in ACs; P = 0.0759), and was not observed at all in patients with LC carcinoma. There was no significant correlation between protein overexpression and histologic type (P = 0.0548). Overall interobserver agreement was nearly perfect (κ = 0.91; 95% confidence interval, 0.87–0.97). In the positively stained samples, heterogeneity of EGFR staining was observed even within a single tumor nodule in all tumors examined (Fig. 1A,B). In tumors that exhibited EGFR overexpression, the populations of positive cells ranged from 5% to 90% (mean ± standard deviation, 54% ± 24%). The populations of positive staining cells in each histologic subtype were as follows; 5–89% in SCCs and 8–90% in ACs. No morphologic difference was found among the tumors with and without overexpression of EGFR, nor were any histologic or cytologic differences observed between overexpressing and nonoverexpressing cells within a single tumor. However, in the SCC samples, a clear topologic difference often was observed in the staining intensity of EGFR between centrally and peripherally located tumor cells of carcinoma nests, with higher intensity observed in peripheral cells.

Figure 1.

(A) Squamous cell carcinoma and (B) adenocarcinoma of the lung immunostained for epidermal growth factor receptor. Carcinoma cells with 3 + immunostaining are distributed heterogeneously in the tumor (original magnification × 100).


FISH analysis was performed for 36 ACs and 25 SCCs from samples that had exhibited EGFR overexpression in IHC and in for 10 additional samples each from tumors that were scored “1 + staining” and “negative.” The EGFR gene was detected clearly as an orange signal in all the tumors examined, except for 7 tumors (5 ACs and 2 SCCs with 2 + staining) in which the signals for both EGFR and the centromere were too faint to count. These latter tumors were excluded from the current analysis. Among the remaining 31 ACs and 23 SCCs (54 samples in total), EGFR amplification was detected in 40 tumors (23%), and the frequency of EGFR gene amplification among the tumors that showed protein overexpression was 74% (Table 1). More specifically, high-level amplification of the EGFR gene was found in 8 ACs and in 9 SCCs (17 tumors; 9.8%), and it was found that 15 of those 17 tumors showed 3 + staining in IHC. Low-level amplification was found in 14 ACs and in 9 SCCs (23 tumors; 13%), and 17 of those 23 tumors were scored as 2 + staining in IHC (Table 1). In 16 of 17 tumors from the high-level amplification group, EGFR gene amplification was detected as 1 or 2 large clusters of orange signals, as shown in Figure 2A. Precise quantification was not possible in most tumor cells due to the tight clustering of signals. In the remaining single sample of AC, FISH analysis revealed > 10 homogeneous, multiple, scattered signals over the nuclei (Fig. 2B).

Table 1. Correlation between Levels of Epidermal Growth Factor Receptor Expression and Gene Amplification
IHCNo. of tumors with gene amplification (%)
  1. IHC: immunohistochemical grading.

 3 + 1 (9.1) 3 (27.3) 7 (63.6)11 (100.0)
 2 + 8 (40.0)11 (55.0) 1 (5.0)20 (100.0)
 1 + or 010 (100.0) 0 (0.0) 0 (0.0)10 (100.0)
 Total no.1914 841
Squamous cell carcinoma    
 3 + 0 (0.0) 3 (27.3) 8 (72.7)11 (100.0)
 2 + 5 (41.7) 6 (50.0) 1 (8.3)12 (100.0)
 1 + or 010 (100.0) 0 (0.0) 0 (0.0)10 (100.0)
 Total no.15 9 933
Figure 2.

Dual-color fluorescence in situ hybridization (FISH) analysis of the epidermal growth factor receptor gene (EGFR) (orange signals) and centromere 7 (green signals). (A) The same field that is shown in Figure 1A. One or 2 clusters of orange signals of EGFR are found in tumor nuclei. (B) FISH shows tumor cells with multiple, scattered signals of EGFR. (C) FISH shows tumor cells with 4 or 5 signals of EGFR. Original magnification ×100 (A) ×250 (B, C).

In 23 tumors from the low-level amplification group, the number of orange signals (EGFR) ranged between 3 and 8, and the numbers of green signals (centromere 7), when identified, ranged from 2 to 4 (Fig. 2C). In ten of those 23 tumors, the centromere signals were too faint to count, even using the filter set specific to SpectrumGreen™. Overall, the identification of tumors that exhibited “polysomy of chromosome 7,” according to the definition used in this study, was not possible, and all 23 tumors were designated arbitrarily as low-level amplification/polysomy (LA/Poly). Nuclei from cancer cells that did not exhibit gene amplification, normal epithelial cells, and nonneoplastic stromal or inflammatory cells generally had the normal two faint signals for EGFR.

Comparison of IHC and FISH

When the results of IHC and FISH were compared, it was found that protein overexpression and gene amplification were correlated significantly (P < 0.0001). Among 20 tumors that did not exhibit overexpression, i.e., the tumors that showed 1 + positivity or negative staining, no gene amplification was observed. High-level amplification was significantly more frequent in tumors with 3 + staining than in tumors with 2 + staining (15 of 22 tumors vs. 2 of 32 tumors; P < 0.0001) (Tables 1 and 2). Compared with the adjacent serial sections that were immunostained for EGFR protein, the areas of tumor cells in which gene amplification could be detected overlapped completely with areas of protein overexpression. This coincidence could be confirmed on a cell-by-cell basis in the tumor cells that exhibited both high-level amplification and high-level overexpression, as shown in Figures 1A and 2A. Conversely, nonamplifying cells also were overexpressed in 14 tumors (13 tumors with 2 + staining and 1 tumor with 3 + staining).

Table 2. Statistical Analysis
VariableNo. of tumorsCorrelation with
% of EGFR overexpression% of high amplificationa
  • EGFR: epidermal growth factor; AC: adenocarcinoma; SCC: squamous cell carcinoma; LC: large cell carcinoma; ND: not done.

  • a

    Analysis was limited to the tumors with overexpression.

  • b

    Fluorescence in situ hybridization analysis was not done for large cell carcinomas, because there were no tumors with overexpression.

Histologic subtype   
 P value 0.0548Unanalyzable
  AC vs. SCC 0.07590.2972
T classification   
 P value 0.14190.7332
  AC 0.15030.4322
  SCC 0.23300.8152
N classification   
 P value  0.081
  N0 vs. N1/N2 0.0028 
  AC 0.00150.0149
  SCC 0.39210.7946
 P value 0.00460.3960
  AC 0.00150.0936
  SCC 0.61000.9377
Immunohistochemical  grading   
 1 +61  
 2 +396.3
 3 +22 68.2
 P value (IHC 2 + vs. 3 +)  < 0.0001
Survival rate (P value) 0.09410.9319

EGFR Status in Lymph Node Metastases

Among the 61 tumors (36 ACs and 25 SCCs) that exhibited overexpression of EGFR in the primary tumors, 35 tumors (22 ACs and 13 SCCs) had lymph nodes metastases, and 10 tumors (5 ACs and 5 SCCs) had EGFR-overexpressing tumor cells in the metastatic lymph nodes. FISH analysis of the lymph nodes revealed a combined pattern of amplified and nonamplified cells in each of the 10 tumors. The type of amplification in the primary and metastatic tumors was identical in 7 of 10 tumors (4 tumors showed high-level amplification, and 3 tumors showed low-level amplification). In the other three tumors, metastatic tumors showed low-level gene amplification despite the fact that their primary tumors showed high-level amplification.

Comparison of EGFR Status and Clinicopathologic Factors

We next examined the correlations between EGFR overexpression or gene amplification and representative clinicopathologic factors, including histologic subtypes, tumor size (T classification), lymph node status (N classification), and pathologic stage (p stage). When all NSCLCs were analyzed altogether, EGFR overexpression was correlated with lymph node metastasis (P = 0.028). When the frequency of lymph node metastasis was evaluated within the tumors with EGFR overexpression, the presence of high amplification tended to correlate with lymph node metastasis, although the correlation was not statistically significant (P = 0.081) (Table 2). However, there was no specific correlation between overexpression/amplification and T classification. Collectively, the p stage was correlated well with EGFR overexpression (P = 0.0046) but not with gene amplification.

Next, the same statistical analyses were performed within each histologic subtype. In the SCCs, no significant correlation between overexpression/amplification and clinicopathologic factors was identified. In contrast, in ACs, lymph node metastasis and higher pathologic stage were correlated significantly with EGFR overexpression (P = 0.0015 and P = 0.0015, respectively). Furthermore, when the correlations were evaluated within the group that showed EGFR overexpression, the presence of high amplification was associated with lymph node metastasis (P = 0.0149), but not with pathologic stage (P = 0.0936), compared with the tumors that showed LA/poly or no amplification (Table 2). The significance of EGFR overexpression and/or gene amplification in the LC carcinomas could not be evaluated due to the limited number of samples.

Next, EGFR overexpression and gene amplification were evaluated for their potential prognostic significance. Univariate analysis revealed no significant difference in survival rates with respect to protein overexpression or gene amplification (Table 2). However, patients who had AC that exhibited high amplification tended to have a shorter survival (data not shown).


In the current study, overexpression of EGFR was found in 61 of 181 specimens of NSCLC (34%), with a somewhat higher incidence in SCCs than in ACs (P = 0.0759). Among these tumors, EGFR gene amplification was detected in 40 tumors (23%), which included 17 tumors (9.8%) that had high-level gene amplification. Thus, among tumors that exhibited protein overexpression, the frequency of EGFR amplification was 74%. Conversely, all of the tumors that exhibited gene amplification were positive for protein overexpression. Overall, protein overexpression and gene amplification were correlated significantly (P < 0.0001). Furthermore, high-level amplification was significantly more frequent in tumors with 3 + staining than in tumors with 2 + staining (P < 0.0001). These results suggest that gene amplification plays a major role in EGFR overexpression, especially in NSCLC tumors with high-level expression.

Previous studies of EGFR gene amplification in NSCLC using Southern blot analysis reported much smaller frequencies, ranging from 0% to 9%27, 28 compared with the current results (23%). This may be explained by the heterogeneous distribution of cells that exhibited EGFR gene amplification in NSCLC observed in the current study. Thus, positive signals from individual cells may be diluted when they are pooled together with lower expression cells. In addition, amplification in a small fraction of tumor cells may have been obscured further due to contamination by tumor cells that lacked gene amplification, and nonneoplastic cells. Therefore, FISH analysis on paraffin embedded tumors may be the best method for clarifying gene amplification in heterogeneous solid tumors like lung carcinoma, because it enables us to determine precise copy numbers on a cell-by-cell basis and the overall amplification pattern within the tissue. However, there has been only one FISH study to date that examined EGFR in NSCLC reported. Using comparative IHC and FISH analyses on 183 microarrayed NSCLC tissues, Hirsch et al.15 found EGFR overexpression in 62% of NSCLCs and overall gene amplification in 9% of NSCLCs.

The significant difference in the rates of overexpression and in amplification/polysomy between those reported by Hirsch et al. (62% vs. 34%, respectively) and those reported in the current study (59% vs. 23%, respectively) may result from different scoring systems. For example, weak cytoplasmic staining was interpreted as “1 +” in our study, whereas the same samples may have been scored higher in the other report. Similarly, in FISH analyses, accurate, consistent identification of low-level amplification and polysomy 7 was not possible, presumably due to nuclear truncation by FISH in cut sections29; thus, these may have been false-negative results in our experiments. However, there was no statistical difference in the high-level amplification rates between the 2 studies (P = 0.1867). Moreover, Hirsch et al. reported gene amplification in a clustered pattern, similar to our observations. This fact indicates that detection of high-level amplification by FISH is unambiguous and objective among researchers. It also shows that analysis of 3 areas of tissue measuring 1.5 mm in greatest dimension per tumor, which was employed in the study by Hirsch et al., is sufficient to detect cells with EGFR amplification. These positive cells typically are localized in a limited area in most tumors, as shown by the current results.

In the current study, clustered signals were observed in all tumors that exhibited a high level of gene amplification, with the exception of one AC in which isolated signals were scattered in the nuclei. It is accepted generally that clustered signals found by FISH correspond to amplified signals in homogeneously staining regions (HSR), and multiple scattered signals correspond to double-minute chromosomes (DM).30 Several FISH studies have reported a different preponderance of these two patterns of EGFR amplification in different types of carcinoma. In gliomas,6 gastric carcinomas,12 and colorectal carcinomas,13 it was found that EGFR amplification was predominantly of the DM type. In contrast, our preliminary data suggest that, in esophageal carcinomas, EGFR amplification is exclusively of the HSR type (unpublished data). The clinical significance of these different amplification patterns remains to be clarified; however, one of the in vitro studies demonstrated that the DM pattern disappeared in response to chemotherapy,31 suggesting that chemotherapy against tumors with EGFR amplification of the DM type may cause down-regulation of EGFR and, subsequently, may abrogate the EGFR-mediated intracellular signal cascade.32

Although there is a good correlation between EGFR gene amplification and protein overexpression, there was no clear evidence of gene amplification in 14 of the tumors in the current study, although most of those tumors showed 2 + immunostaining intensity. Similar findings have been reported not only in pulmonary carcinomas8 but also in renal,11 pancreatic,10 and colon carcinomas.9, 13 Although protein overexpression in these tumors probably is caused by transcriptional or posttranscriptional activation,31 various theories have been proposed to explain the underlying mechanisms. Sheikh et al. demonstrated that, in cultured cell lines, p53 directly activates EGFR expression at the transcriptional level by binding to a p53-responsive site in the EGFR gene promoter.33 Recently, studies have focused on the modulation of EGFR transcription by polymorphic CA repeats34 and by a 140-base pair enhancer region,35 both of which reside within intron 1. Kersting et al.36 assessed the frequency of amplification of the whole EGFR gene, amplification of the first-intron CA repeat, and protein overexpression in 222 invasive breast carcinomas using FISH, Taqman reverse transcriptase-polymerase chain reaction, and IHC analyses, respectively. Those authors found that extension of the first-intron CA repeat with subsequent amplification of this domain was the major cause of protein overexpression and was responsible for 18.7% of all findings of EGFR overexpression, compared with 12.5% of findings in which amplification of the entire EGFR gene was responsible. However, amplification of this CA repeat or 140-base pair enhancer element could result, at most, in a 5-fold or 13-fold enhancement of transcriptional activation, respectively,35, 36 whereas gene amplification, as in A431 cells, may cause up to a 260-fold increase in mRNA expression.36 These facts imply that transcriptional up-regulation caused by expanded CA repeat or enhancer element is minor compared with up-regulation caused by high-level gene amplification. Indeed, our previous study of EGFR expression in soft tissue sarcomas by immunoblot analysis showed that, among those tumors that were positive for EGFR protein overexpression, protein expression was extraordinarily higher in cells that also were positive for gene amplification.14 This remarkable difference in EGFR expression levels may have be responsible for the higher rate of lymph node metastasis in patients who had AC that exhibited gene amplification. Other mechanisms of EGFR overexpression include increased EGFR gene copy numbers by chromosome 7 polysomy. The presence of chromosome 7 polysomy is not uncommon in solid carcinomas and has been suggested to account for overexpression in the absence of amplification.14, 15 However, polysomy was not significant in the 14 tumors studied here that exhibited EGFR overexpression without gene amplification, as discussed above.

In previous IHC studies of NSCLC, no consistent conclusion was reached with regard to the correlation between the EGFR overexpression and clinicopathologic features,37 reflecting the large differences in reported IHC data. Our study showed that primary tumors that were positive for EGFR overexpression were correlated significantly with extensive lymph node metastases (P = 0.0028) and higher pathologic stage (P = 0.0046). Furthermore, 10 of 35 tumors with metastasis to the lymph nodes were positive for EGFR overexpression, and all of those tumors also were positive for gene amplification. These results suggest that targeting of EGFR by the new adjuvant therapies in the treatment of NSCLC may hold much promise, especially because these therapies will include the treatment of metastatic tumors.

Unlike gliomas, in which frequent rearrangements affecting the EGFR extracellular domain have been studied extensively, mutations or rearrangement of EGFR have not been defined in variety of malignancies.38, 39 Recently, however, Lynch et al. identified somatic mutations in the tyrosine kinase domain of the EGFR gene in gefitinib-responsive pulmonary ACs.40 Those authors concluded that an analysis of EGFR mutations could be used to identify the patients with NSCLC who were most eligible for gefitinib therapy, because overexpression of EGFR in the absence of any mutation may signify a gefitinib-resistant tumor. However, it remains unclear whether and how gene amplification accompanying protein overexpression may be predictive of outcome in gefitinib-based therapy.

In the current study, we demonstrated a high concordance between gene amplification and protein overexpression in EGFR in NSCLC, similar to that observed for c-erbB-2 in breast carcinoma. However, unlike c-erbB-2 in breast carcinoma, the detection of gene amplification by FISH has not been approved as a test for deciding the application of this therapy. Thus, future clinical trials are needed urgently to evaluate the responsiveness to treatment in two groups of patients, i.e., patients with tumors that show high-level overexpression of EGFR by gene amplification and patients with tumors that show low-to-moderate overexpression without gene amplification.