Comparative Prognostic Value of Epidermal Growth Factor Quantitative Protein Expression Compared with FISH for Head and Neck Squamous Cell Carcinoma

Head and neck squamous cell carcinoma (HNSCC) is the sixth leading cause of cancer-related deaths worldwide. Advances in molecular biology have led to the identification of molecules that contribute to the malignant phenotype and induce resistance to chemotherapy and radiotherapy. Pharmacologic inhibition of these molecules has become a major strategy in cancer therapy and is often combined with chemotherapy and/or radiotherapy.

Grandis and colleagues first showed the role of the epidermal growth factor receptor (EGFR) as a prognostic factor in patients with HNSCC. The authors showed that overexpression of EGFR is a very common molecular alteration in HNSCC (1). Further work subsequently revealed that treatment with EGFR-targeted therapy such as cetuximab inhibits EGFR signaling, potentiates the cytotoxic effects of chemotherapy or radiation on cancer cells, and increases cure rates (2–4).

A critical question in EGFR-targeted therapy in HNSCC patients has been patient selection because the EGFR protein expression by immunohistochemistry (IHC) has not been closely associated with the response rate and other outcome measures (5–7). An inverse relationship between EGFR content and response to cetuximab was reported in a small randomized trial, where EGFR content was determined by conventional IHC (4). Immunohistochemical studies evaluating EGFR expression status in tumor samples are limited by the technical difficulties inherent in assessing EGFR with this assay such as variability in immunohistochemical techniques, different methods of pathologist-based scoring and the semiquantitative nature of the assay. To overcome this problem, a method of in situ automated quantitative analysis (AQUA) has been developed which allows measurements of protein expression within subcellular compartments that provides a number directly proportional to the number of molecules expressed per unit area (8). Thus, we avoid biases associated with conventional IHC.

We previously analyzed a cohort of 95 oropharyngeal cancers on a tissue microarray (TMA) annotated with long-term follow-up data for EGFR protein expression levels using AQUA and correlated those and found that high EGFR protein levels were associated with significantly greater local recurrence, progression, and death rates (8).

Assessment of EGFR gene copy number has been associated with response to EGFR-targeted therapies in other tumors. In non–small cell lung cancer, EGFR FISH-positive status was associated with high RNA and protein expression levels and with improved outcome following treatment with the EGFR tyrosine kinase inhibitors gefitinib and erlotinib (9, 10). Chung and colleagues (11) reported that increased EGFR gene copy number by gene amplification or high polysomy as determined by FISH was a common genetic alteration in a cohort of HNSCC and was significantly associated with inferior recurrence-free and overall survival.

The underlying mechanisms of increased EGFR protein expression are poorly understood. Nonetheless, previous investigators have not consistently showed a correlation between EGFR gene copy number and protein content, using conventional IHC (11). In the present study, we quantitatively assessed EGFR protein expression levels using AQUA and compared those with EGFR gene copy number by FISH. The aim of this study was to gain insight into the clinical significance of EGFR gene copy number and protein expression levels in HNSCC.

Following Institutional Review Board approval, paraffin-embedded specimens from primary HNSCCs treated at Yale-New Haven Hospital and Aristotle University Hospital were collected. Inclusion criteria were histologically confirmed primary HNSCCs treated at Yale-New Haven Hospital and Aristotle University Hospital between 1992 and 2005, and therapy with either external beam radiotherapy (EBRT) or gross total surgical resection and postoperative radiotherapy. Exclusion criteria included presentation with metastatic or recurrent disease or failure to receive a full course of radiation therapy. Patients with incomplete clinicopathologic data or those lost to follow-up were also excluded. Following Institutional Review Board approval, the TMA was constructed as previously described including 102 cases (12). Tissue cores 0.6 mm in size were obtained from formalin-fixed paraffin-embedded (FFPE) tissue blocks from the archives of Department of Pathology at both Yale University and Aristotle University of Thessaloniki. Hematoxylin- and eosin-stained slides from all blocks were first reviewed by the pathologists (D.K. and T.Z.) to select representative areas of invasive tumor to be cored. The cores were placed on the recipient microarray block using a Tissue Microarrayer (Beecher Instrument). All tumors were represented with at least 2-fold redundancy. Previous studies have showed that the use of TMAs containing 1 to 2 histospots provides a sufficiently representative sample for analysis by IHC (13). The TMA was then cut to yield 5-μm sections and placed on glass slides using an adhesive tape transfer system (Instrumedics Inc.) with UV cross-linking.

TMAs were deparaffinized and stained as previously described (14). In brief, slides were deparaffinized in 2 changes of xylene and rehydrated through changes of ethanol with decreasing concentrations. Slides were then subjected to heat-induced antigen retrieval by pressure cooking in 0.1 mol/L citrate buffer (pH 6.0) for approximately 10 minutes. Endogenous peroxidase activity was blocked by incubating in 0.3% hydrogen peroxide in methanol for 30 minutes. Nonspecific antibody binding was blocked with 0.3% bovine serum albumin for 30 minutes at room temperature. Following these steps, slides were incubated with primary antibody to EGFR (1:500, clone 31G7; Zymed Laboratories) at 4°C overnight. This antibody has been extensively validated in previous studies using IHC and Western blot analysis of neoplastic tissue and tumor cell lines (15, 16). Subsequently, slides were incubated with goat anti-mouse secondary antibody conjugated to a horseradish peroxidase–decorated dextran polymer backbone (Envision, Dako Corporation) for 1 hour at room temperature. Tumor cells were identified by use of anticytokeratin antibody (rabbit anti-pancytokeratin antibody, 1:100, Z0622; Dako Corporation) with subsequent goat anti-rabbit antibody conjugated to Alexa 546 fluorophore (1:100, A11035; Molecular Probes). We added 4′,6-diamidino-2-phenylindole (DAPI) to visualize nuclei (Prolong Gold with DAPI, P36931; Molecular Probes). Fluorescent chromogen Cy-5 tyramide (1:50; Perkin Elmer Corp) was used for target identification. Cy-5 (red) was used because it is well outside the green-orange spectrum of tissue autofluorescence.

TMAs were incubated in a hypotonic solution 0.075 mol/L KCl for 15 minutes at 37°C and fixed in 3:1 methanol/glacial acetic acid. Following dehydration, TMAs were treated for 10 minutes in pepsin (0.01% in 0.01 mol/L HCl) at 37°C and fixed in 1% formaldehyde at room temperature for 10 minutes. The LSI EGFR SpectrumOrange/CEP 7 SpectrumGreen probe (Vysis; Abbott Laboratories) was applied according to the manufacturer's instructions. Codenaturation of probe and target DNA was achieved by incubation at 80°C for 5 minutes. Hybridization was carried out at 37°C for 20 hours and the unbound probe was removed in 3 washes with 50% formamide, 2× SSC and 1 wash with 2× SSC, 0.1% NP-40, each for 5 minutes at 45°C. Chromatin was counterstained with DAPI in Vectashield artifade (Vector).

Mean numbers per cell of EGFR and CEP 7 probes were estimated as well as the EGFR/CEP 7 ratio using an epifluorescence microscope coupled with a triple band pass interference filter (blue/red/green) and single band filters for blue, red, and green (Chroma Technology Corp.). Images were acquired using a cooled CCD camera (SenSys, Photometrics). Samples were classified in 2 strata: FISH negative, with no or low genomic gain (≤4 copies of the gene in >40% of cells), and FISH positive, with a high level of polysomy (≥4 copies of the gene in ≥40% of cells) or gene amplification defined by a ratio of gene/chromosome per cell ≥ 2 or ≥ 15 copies of the gene per cell in 10% or greater of the analyzed cells (10).

Histospots containing less than 5% tumor as assessed by mask area (automated) were excluded from further analysis. Pearson's correlation coefficient (R) was used to assess the correlation between AQUA scores from redundant tumor cores and to assess the association between EGFR FISH and AQUA. Survival analysis was performed at 5-year cutoff points. Progression-free survival and overall survival were assessed by Kaplan–Meier analysis with log-rank score for determining statistical significance. Relative risk was assessed by Cox proportional hazards regression by multivariable analysis. Correlations between clinicopathologic characteristics and survival were assessed by univariate Cox regression. All calculations and statistical analyses were performed by SPSS 15.0 for Windows (SPSS, Inc.).

Our study included 102 patients with histologically confirmed HNSCC. Sixty-four patients had complete EGFR expression data. Correlations between clinicopathologic variables and EGFR protein expression and EGFR gene copy number are summarized in Table 1.

There were 64 patients with primary HNSCC who met inclusion criteria and for whom we had complete EGFR expression data. We excluded, from the analysis, 38 cases (among 102) with missing EGFR expression information. These cases did not differ from the ones included in the analysis with respect to patient gender, tumor site, tumor node metastases (TNM) stage, histologic grade, and progression-free and overall survival as assessed by Fisher's exact test and log-rank test, respectively. The staining pattern for EGFR was mainly membranous and cytoplasmic (Fig. 1A and B) and tumor AQUA scores ranged between 2.44 and 62.9 (Fig. 1C). The regression between scores from double redundant spots was again high (Fig. 1D; R = 0.9), which allowed us to average the 2 scores for each tumor and use the mean score in our analysis. The cohort was divided into high and low EGFR expressers using the median score as the cutoff point (17). Of those tumors successfully evaluated for EGFR protein expression, tumors with high versus low EGFR levels did not differ by patient gender, TNM stage, histologic grade, or tumor site (Table 1).

EGFR gene copy number was assessed by FISH in 64 samples. Eighty-three percent (53/64 samples) had no or low gain of EGFR genomic sequences (EGFR FISH negative) and 17% (11/64 samples) had high levels of EGFR genomic gain (EGFR FISH positive). Figure 2 illustrates tumors categorized as FISH negative (Fig. 2A and B) and FISH positive (Fig. 2C). Sixty-four samples were used for statistical analyses to determine associations between the FISH status and patient demographics. Besides anatomic subsite, patient characteristics did not differ between FISH-positive and FISH-negative samples including sex, clinical stage, and histologic grade (Table 1).

The association between tumor EGFR protein expression (as determined by AQUA) and EGFR and FISH positivity was analyzed by Pearson correlation. There was a significant correlation between EGFR AQUA score and FISH positivity (Pearson ρ = 0.353, P < 0.04). Figure 2 shows a representative histospot with high EGFR protein levels (Fig. 2D) and EGFR gene copy number amplification (Fig. 2E).

The level of EGFR protein expression by AQUA was also evaluated for association with progression-free survival. Kaplan–Meier analysis (Fig. 3A) showed no significant association between either EGFR protein level (P = 0.495) or EGFR FISH status (P = 0.111) and progression-free survival.

EGFR expression was examined in relation to overall survival. Patients with high EGFR protein expression by AQUA had a 5-year overall survival of 27.7%, compared with a 5-year survival of 54% among those with low EGFR protein expression (Fig. 3B). There was no significant association between EGFR FISH status and overall survival (P = 0.201). Results for univariate analysis are summarized in Table 2.

Using the Cox proportional hazards model, we conducted multivariate analysis to assess the independent predictive value of tumor EGFR expression groups for local recurrence and disease-free survival (DFS). Multivariate analysis adjusted for the well-characterized prognostic variables of age, sex, subsite, TNM stage, and histologic grade. In this model, high tumor EGFR protein expression status remained an independent prognostic factor for progression-free survival (P = 0.047), along with grade (P = 0.024). Results for multivariate survival analysis are summarized in Table 3.

In the present study, we sought to compare EGFR genomic copy number by FISH with a highly quantitative method for automated analysis of EGFR protein content and to determine the association of each with outcome, in HNSCC patients curatively treated with adjuvant or definitive radiation. We found that EGFR FISH positivity was not common, with 17.2% of cases meeting the definition of FISH positivity, and was associated with high EGFR protein levels as assessed by AQUA. There was no association between EGFR FISH status and outcome measures, whereas high EGFR protein levels by AQUA correlated with overall survival. To our knowledge, this is the first comparative study between EGFR overexpression by AQUA and gene amplification in cancer.

Chung and colleagues (11), reported increased EGFR gene copy number—gene amplification or high polysomy—as showed with FISH, in 58% of HNSCC. Increased copy number was strongly associated with worse recurrence-free and overall survival. FISH status was not associated with protein overexpression by IHC. The authors did not find correlation between FISH status and RNA or protein expression levels assessed by DNA microarray and IHC. Fourteen of 86 samples were from recurrent tumors and patients had received various treatments that were within the standard of care with curative intent at that time. Temam and colleagues analyzed 134 HNSCC for EGFR gene copy numbers using quantitative real-time PCR reaction (18). Thirty-two (24%) of 134 tumors contained aberrant copy number including 22 (17%) with increased copy number and 10 (7%) with decreased copy number. The authors of this well-conducted study found that patients whose tumors had EGFR copy number alterations (particularly patients with increased copy numbers) had significantly inferior overall, cancer-specific, and DFS compared with patients with normal copy numbers (P < 0.0001). FISH was performed in only 16 tumors and showed complete agreement with quantitative PCR. Possible explanations for difference in results include the following: The French specimens were fresh frozen, whereas U.S. specimens were FFPE, they were collected over a wide time range (1985–2003) and no treatment details are provided. Mrhalova and colleagues (19) analyzed 33 patients with HNSCC for EGFR protein expression and gene copy number by IHC and FISH, respectively. Gene amplification was identified in 11 (17.2%) patients and did not correlate with EGFR protein levels. Braut and colleagues (20) performed IHC and FISH on TMAs composed of 145 laryngeal tissue samples including 38 samples of normal mucosa, 46 samples of hyperplastic lesions, and 61 samples of cancerous lesions. EGFR gene amplification was found in 8 of 50 (16%) samples and correlated with cytoplasmic EGFR protein expression. Koynova and colleagues (21) analyzed a TMA consisting of 1,385 laryngeal cancers for EGFR copy number changes and found amplification in 10.37% of cases. Freier and colleagues (22) analyzed 609 HNSCC, including 511 primary carcinomas of different clinical stage and anatomic localization and 98 recurrent carcinomas, second primary carcinomas, and regional metastases on a TMA. The overall prevalence of amplifications of EGFR was 12.7%. FISH analysis of specimens from the EXTREME study for EGFR gene amplification or polysomy was performed to correlate FISH status with response to cetuximab (23). Forty percent of patients had EGFR gene copy number greater than 3, 12% of patients had an EGFR/CEP7 score of 2 or more, and more than 10% had EGFR clusters present in 25% or more of cells. There was no correlation between EGFR gene copy number and outcome in the cetuximab + chemotherapy treatment arm.

In our study, there was no association between EGFR FISH status and outcome probably due to the low rate of FISH positivity which was consistent with the vast majority of the studies. Intratumor heterogeneity observed in EGFR gene copy number in HNSCC may account for the variation in EGFR gene amplification and copy number reported thus far in HNSCC. Study limitations also include the small number of specimens and the fact that patients did not receive uniform treatment. Patients, however, received standard treatment with curative intent. These cohorts are useful for identification of prognostic biomarkers, whereas discovery of predictive biomarkers requires uniformly treated patient cohorts.

The underlying mechanisms of EGFR protein overexpression are poorly understood. Gene amplification has been proposed as a potential mechanism of EGFR protein upregulation, but studies have revealed additional mechanisms such as transcriptional regulatory mechanisms that control EGFR expression: single nucleotide polymorphisms in EGFR promoter region [i.e., GC (EGFR*1)] or common CA dinucleotide repeat in intron 1 of EGFR affecting EGFR mRNA levels (24). Higher activity of the promoter region and higher mRNA levels have been observed in the non-GC–containing haplotype and this haplotype may predict for greater sensitivity to gefitinib. Shorter number of CA dinucleotide repeats in intron 1 of EGFR was associated with greater in vitro sensitivity to erlotinib in 14 head and neck cancer cell lines (24). The common EGFR-R521K genotype (G/G) was significantly associated with increased skin toxicity (P = 0.024) and showed a trend toward reduced risk of tumor progression (HR = 0.55; 95% CI = 0.27–1.08; P = 0.08) in a single arm phase II study which included fifty-one patients treated with second-line regimen (cetuximab/docetaxel) for recurrent or metastatic HNSCC (25). Therefore, although a strong correlation between EGFR FISH status and protein levels measured by AQUA was found in our study, quantitative assessment of EGFR protein levels provided by AQUA reflects more closely the availability of EGFR per cell and seems to be a more attractive assay for prediction for response to cetuximab in HNSCC patients. As dual EGFR/her 2 inhibitors, irreversible inhibitors, fully humanized antibodies, combinations with other targeted agents, and increased acceptance of on study biopsy for badly needed correlative assays emerge in future studies, the finding that AQUA, which is inexpensive, can be done on small cores of paraffin-embedded material is superior to the more technically demanding and expensive assays may facilitate important research.

In summary, we show that EGFR protein levels assessed by AQUA provide important prognostic information. EGFR gene copy number had no prognostic impact in the present cohort. Furthermore, EGFR FISH status correlates closely with EGFR AQUA score. It seems that, in addition to EGFR gene copy number, EGFR protein levels are regulated by transcriptional and posttranscriptional mechanisms. EGFR AQUA score is a candidate predictive biomarker for response to EGFR-targeted therapies in HNSCC. However, additional studies are required to delineate the biology involving the EGFR pathway and to gain insight into the clinical application of the EGFR AQUA assay as a predictive marker in clinical trials using EGFR-targeted therapies in HNSCC.

D. Rimm is a scientific consultant and cofounder of the AQUA technology. The other authors disclosed no potential conflicts of interest.

This study was funded by Yale School of Medicine Institutional startup funds (AP) and the Virginia Alden Wright Fund (CS).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.