Mcl-1, Vascular Endothelial Growth Factor-R2, and 14-3-3σ Expression Might Predict Primary Response against Radiotherapy and Chemotherapy in Patients with Locally Advanced Squamous Cell Carcinomas of the Head and Neck

Purpose: This study was done to explore whether the expression of a selected set of proteins could predict primary response to radiotherapy or concomitant radiotherapy and chemotherapy in patients with advanced head and neck cancer.

Experimental Design: Forty-three pretreatment tumor biopsies were taken during diagnostic panendoscopy and examined for Mcl-1, vascular endothelial growth factor (VEGF)-R2, CD9, and 14-3-3σ expression by immunohistochemistry. Forty-three patients underwent primary radiotherapy, of which, 29 patients received concomitant chemotherapy (low dose daily cisplatin, mitomycin C bolus). The primary end-point was locoregional tumor control 6 months after completion of radiotherapy. Mcl-1, VEGF-R2, CD9, and 14-3-3σ expression were correlated with patients' primary response to radiotherapy and chemotherapy and with established clinicopathologic variables.

Results: Thirty complete and 13 partial responses were observed in our patient group. High expression levels of Mcl-1 (P = 0.021), VEGF-R2 (P = 0.032), and 14-3-3σ (P = 0.013), but not of CD9, in tumor biopsies was correlated with complete response. Overexpression of at least two of the three aforementioned proteins in pretreatment biopsies predicted—with a likelihood of 80%—whether a patient would achieve complete response to radiotherapy and chemotherapy. However, if only one of these proteins is overexpressed, there is a likelihood of 84.6% that this patient would not completely respond to therapy.

Conclusion: Determining the expression levels of Mcl-1, VEGF-R2, and 14-3-3σ may be helpful in predicting the early clinical response in head and neck tumor patients receiving primary radiotherapy and chemotherapy and may further allow a pretherapeutic selection of patients.

Despite significant improvements in the treatment of squamous cell carcinomas of head and neck cancer, typical 5-year combined survival rates for all stages of disease are still in the range of 22% to 63% (1). Although the therapeutic management of head and neck cancer has benefited from small and incremental improvements over the past few years, the management of advanced head and neck tumors continues to principally depend on tumor localization and on tumor stage (2). The tumor-node-metastasis stage alone, however, is not sufficient to stratify patients according to their likeliness to respond against different therapies (3, 4). In order to improve treatment outcome, a better understanding of the pathophysiology of head and neck cancer, and the identification of potentially clinically relevant molecular markers for early diagnosis and optimal treatment planning are necessary. Several potential prognostic molecular factors have been described during the past few years (5–25). Proteins which have been implicated in the control of tumor cell survival and which have a prognostic potential are, among others, Mcl-1 (11–18), vascular endothelial growth factor (VEGF)-R2 (19–21), CD9 (22), or 14-3-3σ (23–25). Mcl-1 has been chosen in this study because this antiapoptotic protein is widely expressed (∼90%) in squamous cell carcinomas of the head and neck, whereas Bcl-2 is expressed only in 15% of head and neck tumors (26). In addition, in a recent report from our group (27), we have shown that Mcl-1 expression inhibits radiation-induced cell death. Furthermore, our own preliminary data also indicate that down-regulation of Mcl-1 in two head and neck cancer cell lines modulates chemosensitivity.⁸

VEGF-R2 and CD9 were included because they were previously described to be expressed in squamous cell carcinomas of the head and neck (19–21). There are several publications strongly supporting a role for the VEGF receptor signaling system in the response to radiotherapy (28–30). CD9—in a recently published study from our group—was shown to be a strong prognostic factor for overall and for disease-free survival in patients with squamous cell carcinomas of the head and neck (22).

14-3-3σ has been found to be expressed in squamous cell carcinomas of the head and neck as well as playing a role in cell death (23–25). Another reason for investigating this factor are unpublished data from our group obtained from proteomics analysis of head and neck tumor samples which indicated that 14-3-3σ might be overexpressed in the malignant phenotype of these tumors.

To the best of our knowledge, a potential correlation between the expression of this set of tumor-associated proteins and the patient's response to radiotherapy in head and neck cancer has thus far not been investigated. The purpose of this article was to analyze the expression pattern of the combination of all four potent predictive factors Mcl-1, VEGF-R2, CD9, and 14-3-3σ in the pretreatment tumor biopsies of 43 patients with advanced head and neck cancer, and to investigate a possible correlation with the early tumor response to radiotherapy alone or radiotherapy in combination with concomitant chemotherapy.

Patients. Tumor biopsies of 43 untreated patients (9 females, 34 males), with locally advanced head and neck cancer [T₂ (n = 1), T₃ (n = 11); T₄ (n = 31), N₀ (n = 2), N₁ (n = 2), N_2b (n = 16), N_2c (n = 19), and N₃ (n = 4)] were obtained during diagnostic panendoscopy. After panendoscopy, all patients underwent a hyperfractionated and accelerated irradiation schedule (31), with concomitant boost at the Department of Radiation Therapy, Medical University of Vienna, from 1999 to 2002. Twenty-nine of the 43 patients underwent an identical irradiation protocol plus chemotherapy, consisting of low-dose cisplatin (at 6 mg/m², days 1-20 of radiotherapy) and mitomycin C (at 12 mg/m² on day 21 of radiotherapy). Tumor response was evaluated clinically by the head and neck surgeon and by CT and/or MRI scans 3 and 6 months after the end of radiotherapy. In case of conflicting or unclear results, a biopsy was obtained during control panendoscopy. All patients were seen by the involved physicians (radiologist, radiation oncologist, head and neck surgeon, and/or oncologist) in the tumor board. Patients without evidence of disease or with remaining tumor masses in the primary site and/or locoregional lymph nodes were scored as complete or partial responders, respectively. In the complete responder group, an advanced tumor and lymph node staging (T_3-4 and N_2-3) was diagnosed in 97% and 87% of all patients, respectively. In the partial responder group, 100% of all patients had a T_3-4 and N_2-3 staging. Chemotherapy was administered to 21 and 8 patients with complete or partial tumor remission, respectively, and did not affect outcome in a significant manner.

Immunohistochemistry. Dewaxed and microwaved slides were incubated with goat polyclonal anti-Mcl-1 (Dako, Glostrup, Denmark), mouse monoclonal anti-VEGF-R2 (Santa Cruz Biotechnology, Santa Cruz, CA), mouse monoclonal anti-CD9 (Novocastra Laboratories, Newcastle upon Tyne, United Kingdom) and mouse monoclonal anti-14-3-3σ (Research Diagnostics, Inc., Concord, MA) antibodies overnight at room temperature. As a control, slides were exposed to goat preimmune serum (Vector Laboratories, Burlingame, CA) or IgG₁ (Ancell, Bayport, MN) antibody. After washing with TBS thrice, slides were incubated with a multilink antibody (Dako) for 1 hour at room temperature, washed, and again exposed to alkaline phosphatase–conjugated Streptavidin-AP/10% human serum (Dako) for 1 hour at room temperature. Visualization was done with Fast Red TR, 4-chloro-2-methylbenzenediazonium salt (Sigma-Aldrich, St. Louis, MO), counterstained with hemalaun, dehydrated, and mounted. All slides were graded as either positive or negative. The cutoff for positive versus negative immunostaining was set at 33% positive cells per image field. Only a few tumors showed a weak or distinctive protein expression in all cells indicating background staining. The expression pattern of each marker was independently determined in the complete microscopic tumor section on the slide by two investigators (B.M. Erovic and D. Thurnher). Observer bias was minimized by repeating the evaluation without knowledge about clinical data. In 3 out of 43 tumor samples, a difference regarding the interpretation (cutoff of 33%) between the two investigators was found. In these three cases, samples were reanalyzed by an independent reviewer (E. Selzer).

Statistics. To assess the correlation between Mcl-1, VEGF-R2, CD9, and 14-3-3σ expression on one hand, and tumor remission, sex, tumor, and lymph node status on the other hand, univariate and multivariate stepwise logistic analyses were done. Wilcoxon rank-sum test was used to analyze the correlation of lymph node status, chemotherapy, Mcl-1, VEGF-R2, CD9, and 14-3-3σ expression. In order to analyze the quality of possible predictors for remission, we calculated sensitivity, specificity, and the predictive values, assuming a binomial distribution (95% confidence intervals). Confidence intervals were calculated using Clopper-Pearson values. Overall survival curves were calculated using Cox regression analysis and by the Kaplan-Meier method. Statistical significance was assumed if P < 0.05. Statistical analyses were done using SAS and SPSS-8.0 statistical software packages.

Immunohistochemistry. Distribution of Mcl-1, VEGF-R2, CD9, and 14-3-3σ expression in 43 pretreatment tumor biopsies is summarized in Table 1 and Fig. 1.

Prediction of patients' primary response. Expression of Mcl-1 (P = 0.021), VEGF-R2 (P = 0.032), and 14-3-3σ (P = 0.013) in tumor biopsies of untreated head and neck cancer patients significantly correlated by univariate logistic regression analysis to patients' response to primary radiotherapy and chemotherapy. No correlation could be found between the expression of CD9 and patients' primary response to radiotherapy and chemotherapy. Multivariate logistic regression analysis showed that 14-3-3σ (P = 0.02) and VEGF-R2 (P = 0.023) correlated significantly with tumor response (Table 2).

Statistical analyses showed that the combination of Mcl-1, VEGF-R2, and 14-3-3σ expression in each pretreatment biopsy improves the power of the predictive potential of these markers. A pretreatment biopsy which is highly positive for more than two proteins predicts tumor remission with a sensitivity of 80% (confidence intervals, 61.4%, 92.3%) and a specificity of 84.6% (confidence intervals, 54.6%, 98.0%).

Survival data. Six months after therapy, 70% of 43 patients had a complete response, whereas 30% were partial responders; 32% were alive without and 5% with evidence of disease, whereas 63% had died of cancer before the end of the observation period (March 1, 2005). Overall survival and disease-free interval ranged from 3 to 61 months and from 3 to 53 months, respectively. Overall survival did not correlate with Mcl-1, VEGF-R2, CD9, and 14-3-3σ expression in tumor biopsies (data not shown). Median survival time was 22 months.

We retrospectively compared Mcl-1, VEGF-R2, CD9, and 14-3-3σ expression to patients' clinical response 6 months after radiotherapy alone or in combination with chemotherapy. Taken together, our statistical analysis showed that the expression pattern of Mcl-1, VEGF-R2, and 14-3-3σ in untreated tumor biopsies predicted patients' response to radiotherapy and/or combined radiotherapy and chemotherapy.

To improve patients' survival, particularly in advanced tumor stages, a better understanding of cellular pathways and factors, which could serve as predictive markers, is clearly necessary. Using clinically validated predictors, a stratification of patients into treatment groups, such as surgery or radiotherapy and chemotherapy should be possible. Therefore, intense efforts are under way to identify genes or proteins which are differentially expressed in normal and cancerous tissues or in cancer specimens from patients who did or did not respond to treatment (32–37).

VEGF-R2 was found to be overexpressed in patients who completely responded to therapy. This observation is in agreement with the current scientific literature because recent studies have shown that, particularly in squamous cell carcinomas of the head, increased levels of VEGF, and its receptor VEGF-R2, are surrogate markers for tumor remission after radiotherapy (38, 39).

A key role of 14-3-3σ, also known as stratifin, is its control of the G₂ cell cycle checkpoint. In response to DNA damage, 14-3-3σ is induced in a p53-dependent manner facilitating DNA damage repair before cell cycle progression (23–25). Moreover, an essential role of 14-3-3σ in the response against ionizing radiation-induced cell death could be shown by gene knockout experiments in colorectal carcinoma cell lines (40). We noted significantly increased levels of 14-3-3σ in patients who responded completely to radiotherapy and chemotherapy, whereas decreased levels were observed in the partial responder group.

It is surprising to find overexpression of Mcl-1 in patients with a complete tumor remission. Because Mcl-1 is well known as an antiapoptotic protein, it might be expected that Mcl-1 would be expressed at higher levels in patients with a worse response to radiotherapy and chemotherapy (27, 41).

However, there are examples in the literature linking higher levels of expression of an antiapoptotic protein (Bcl-2) with favorable outcomes, specifically in head and neck cancer patients. Two recent publications reported significantly better responses of Bcl-2-positive tumors of the head and neck region in response to radiotherapy and chemotherapy (37, 42). It is indeed very interesting to find that both high Bcl-2 and Mcl-1 expression levels might be associated with better treatment responses. Moreover, in vitro studies with melanoma cell lines could show that Mcl-1 expression is associated with the cell survival response against ionizing radiation (27).

The results of this retrospective study suggest a prospective analysis of Mcl-1, 14-3-3σ, and VEGF-R2 expression in order to confirm these results. In fact, we have now decided to routinely analyze these three proteins, in conjunction with other factors, such as TP53, Bcl-2, and endothelial growth factor receptor expression in pretreatment biopsies of our head and neck patients. We anticipate obtaining sufficient information in the near future to assess the relative importance of the abovementioned factors as predictors for the response against radiotherapy.

By analyzing Mcl-1, 14-3-3σ, and VEGF-R2 expression, it might be possible to stratify patients into different treatment groups. Different protein expression patterns may predict resistance to radiotherapy and chemotherapy and thus may be a helpful guide in choosing between therapeutic strategies, such as intensified combined modality therapy versus surgery alone or novel experimental regimen.

It would therefore be of potential clinical relevance if our observation could be further validated with larger patient numbers. In addition, it will also be of interest to investigate potential correlations between the expression of the abovementioned factors and other proteins, such as the endothelial growth factor receptor, which are potential targets for combination treatments using small molecules and antibodies.

We thank Sieglinde Bauer for excellent laboratory assistance and Prof. Dr. Peter Bauer for statistical advice.