Focal adhesion kinase expression by immunohistochemical analysis in normal mucosa (A) and in tumor tissue samples showing slight (B), moderate (C), or strong (D) immunostaining (hematoxylin, original magnification ×400).
Kaplan-Meier disease-specific survival curves of the patients by E-cadherin and focal adhesion kinase (FAK) expression.
Rodrigo JP, Dominguez F, Suárez V, Canel M, Secades P, Chiara MD. Focal Adhesion Kinase and E-Cadherin as Markers for Nodal Metastasis in Laryngeal Cancer. Arch Otolaryngol Head Neck Surg. 2007;133(2):145-150. doi:10.1001/archotol.133.2.145
To explore the value of E-cadherin and focal adhesion kinase (FAK) expression in the prediction of cervical lymph node metastases in squamous cell carcinoma of the supraglottic larynx.
Immunohistochemical analysis of retrospectively selected cases.
The study population was composed of 95 previously untreated men with squamous cell carcinoma of the supraglottic larynx.
All the patients underwent surgical resection of the tumor and bilateral neck dissection.
Main Outcome Measures
E-cadherin and FAK expression in relation to nodal metastases.
Decreased E-cadherin expression was correlated with the presence of nodal metastases (P = .006). The combination of E-cadherin and FAK expression resulted in a superior accuracy in assessing nodal metastasis (P = .001). Histological grade was also associated with nodal metastases (P = .005). Multivariate analysis confirmed that these parameters were independent predictors of nodal metastases. In addition, the cases with decreased E-cadherin and increased FAK expression presented a significantly reduced disease-specific survival (P = .005).
The combination of the expression of E-cadherin and FAK could increase our ability to identify patients with clinically negative lymph nodes who are at considerable risk for occult metastases.
It is widely accepted that the presence of lymph node metastases is the single most adverse independent prognostic factor in head and neck squamous cell carcinoma (HNSCC).1 It has been reported that the 5-year overall survival is reduced by approximately 50% in patients with cervical lymph node metastases. The preoperative detection of lymph node metastases is crucial to the effective treatment of patients with HNSCC. Cervical lymph node metastasis cannot always be predicted from the size and extent of invasion of the primary cancer. Even patients with a small primary tumor have frequently been found to have cervical lymph node metastasis. Although the most recent techniques of computed tomography, magnetic resonance imaging, ultrasonography, positron emission tomography, and ultrasound-guided fine-needle aspiration biopsy have reached a sensitivity of more than 80% in detecting occult metastases,2 the detection of occult, microscopic metastatic deposits will continue to evade recognition, and the true lymph node status of the neck will remain uncertain. Thus, interest has developed in identifying molecular markers that would enable the selection of patients at risk for lymph node metastasis.
The process of metastasis is a cascade of linked sequential steps involving multiple host-tumor interactions. Both cell-cell interactions and cell-stroma interactions play an important role during the metastatic cascade. Weakening of cell-cell and cell–extracellular matrix adhesion is obviously imperative for tumor cells to metastasize. Reduced cell-cell adhesiveness is associated with loss of contact inhibition of proliferation, thereby allowing escape from growth control signals. Moreover, the suppression of cell-cell adhesiveness may trigger the release of cancer cells from the primary cancer nests and confer invasive properties on a tumor.3 In epithelial tissues, cell-cell adhesion is mediated largely by the members of the cadherin family and in particular by E-cadherin. Many studies documenting loss or reduction of E-cadherin protein expression have been reported in various human cancers, including HNSCC.3 In HNSCC, we and others have reported that abnormal expression of E-cadherin correlates with lymph node metastases.4- 6
In addition to reduced cell-cell adhesiveness, loss of adhesion of the epithelial cells to the extracellular matrix is one of the fundamental pathways that promotes tumor cell migration, invasion, and metastasis. A key factor involved in the control of cell–extracellular matrix interactions is focal adhesion kinase (FAK), an intracellular tyrosine kinase protein that is localized to cellular focal contact sites.7 Focal adhesion kinase is activated and tyrosine phosphorylated in response to integrin clustering.8 Evidence is emerging that aberrant FAK expression, activity, and signaling can potentially contribute to the development of typical features of malignancy.9 Overexpression of FAK protein has been demonstrated in different tumor types,9 and in some of them it has been correlated with the invasive potential of a tumor and poor patient prognosis.10- 13 Recently, we have shown that FAK is also overexpressed in HNSCC and that strong FAK expression is associated with nodal metastasis.14
We investigated FAK expression in a homogeneous group of patients with squamous cell carcinoma of the supraglottic larynx that we had previously examined for E-cadherin expression.4 In that work, reduction of E-cadherin expression was significantly correlated with lymph node metastases. In the present study, we aimed to confirm the correlation of FAK expression with nodal metastasis in this homogeneous group of patients and the effect of the combination of these 2 markers (E-cadherin and FAK) in the prediction of lymph node metastasis.
Of the 101 patients with squamous cell carcinoma of the supraglottic larynx previously reported,4 we obtained adequate tissue samples for the present study in 95 of them, following institutional review board guidelines. Patients were treated between 1988 and 1994. Only male patients were included owing to the exceptionally low proportion of women with this kind of tumor in our environment. All the patients included in our study underwent surgical resection of their tumor and bilateral neck dissection (functional or radical based on surgical findings). All of them had a single primary tumor and had microscopically clear surgical margins, and none had undergone treatment prior to surgery. None of the patients had distant metastases at the time of surgery. A total of 47 patients (50%) received postoperative radiotherapy. As a general rule, this was administered to the patients with histologically N2 or N3 neck lesions, and also in N0 and N1 neck lesions with locally advanced stage (T4). The clinicopathologic data from the patients are given in Table 1. The stage of disease was determined after the surgical resection of the tumor according to the TNM system of the International Union Against Cancer (fifth edition). The histological grade was determined according to the degree of differentiation of the tumor (Broders classification).
The formalin-fixed, paraffin-embedded tissues were cut into 4-μm sections and dried on capillary-gap glass slides (DakoCytomation, Carpinteria, Calif). The sections were deparaffinized with standard xylene and hydrated through graded alcohols into water. Antigen retrieval was performed by heating 10 minutes in a pressure cooker with citrate buffer. Tissue slides were placed for 20 minutes in a 3% hydrogen peroxide blocking medium and then allowed to react with the anti–E-cadherin antibody (clone 36; Transduction Laboratories, Lexington, Ky) at a concentration of 1:2000 for 30 minutes or the mouse IgG anti-FAK monoclonal antibody clone 4.47 (Upstate Biotechnology Inc, Lake Placid, NY) at 1:250 dilution for 1 hour and 30 minutes. A single sample was used for each antibody, and all the cases were stained in the same session. Immunodetection was performed at room temperature on an automatic staining workstation (TechMate 1000; BioTEK Solutions, Santa Barbara, Calif) using the Envision peroxidase mouse system and diaminobenzidine chromogen as substrate (Envision Plus; DakoCytomation). Counterstaining with hematoxylin for 1 minute was the final step. After staining, the slides were dehydrated through graded alcohols and mounted with a coverslip using a standard medium. Negative controls with an omission of the antiserum from the primary incubation were also included. Positive controls (normal laryngeal epithelium for E-cadherin and a sample of colon adenocarcinoma known to express FAK) were used. Normal laryngeal mucosa was obtained from patients without cancer after informed consent.
The slides were analyzed randomly by 3 of the authors (J.P.R., F.D., and V.S.) without knowledge of clinicopathological data. As previously stated,4 the staining of E-cadherin was predominantly membranous with some cytoplasmatic staining. A staining score was given based on the intensity of the membranous stain (0-4) and percentage of cells stained (0%-100%). The 2 components were multiplied for an overall staining score between 0 and 400. A score was given for both tumor and the surrounding normal-appearing epithelium. For the purposes of this study, we classified the patients as low expressers (E-cadherin expression below the mean of the staining scores) and high expressers (E-cadherin expression above the mean).
In the case of FAK, cytoplasmic-positive immunostaining was observed in all tumor cells with equal intensity within a tumor specimen. Differences in the intensity of staining were observed between individual tumor samples. Therefore, tumor immunoreactivity was scored by taking into account the intensity of staining in tissue samples. We were able to define 3 staining categories (Figure 1): slight staining, moderate staining, and strong staining. Because the clinicopathological characteristics of the cases with moderate and strong staining were similar, for the statistical analysis, we grouped the cases that were scored in these categories.
All statistical analyses were performed using SPSS statistical software version 8.0 (SPSS Inc, Chicago, Ill). Correlations between T classification, N classification, histological grade, and FAK and E-cadherin protein expression were computed using the χ2 test. Also, T classification, histological grade, and the combination of E-cadherin and FAK expression were related to the nodal status in contingency tables and analyzed by the χ2 test. To investigate whether a combination of parameters was more predictive for the nodal status than a single parameter, stepwise logistic regression analysis was performed using the variables that showed significant or near significant correlation with metastases in univariate analysis. Survival curves were calculated using the Kaplan-Meier product limit estimate. Deaths from causes other than the index tumor or its metastases were not considered treatment failures, and these patients were censored in all analysis involving the length of survival. Differences between survival times were analyzed by the log-rank method. P<.05 was considered statistically significant.
The pattern of E-cadherin expression in normal mucosa and primary carcinomas was previously described.4 Expression was generally weaker in tumors than in normal epithelium; the mean ± SD staining score for carcinomas was 131 ± 103 (median, 90). In this study, 54 patients (57%) were classified as low expressers and 41 (43%) as high expressers. Focal adhesion kinase expression was identified in the normal mucosa where staining was weak and confined to the basal cell layer of the epithelium with a cytoplasmic expression pattern (Figure 1A). Positive staining was also observed in vascular endothelial cells and the acinar cells of the seromucosa glands, whereas stroma cells showed negligible FAK expression. In contrast to the normal epithelium, FAK expression in the tumor tissue displayed a homogeneous distribution throughout the whole tumor area. The intensity of staining varied among individual tumors: 42 (44%) of the tumors showed a weak immunostaining (Figure 1B), whereas 53 (56%) displayed a moderate to strong FAK expression (Figure 1C and D). The expression of FAK was not correlated with the expression of E-cadherin in our patients (P>.99) (Table 2).
Table 3 and Table 4 present the correlation of E-cadherin and FAK expression, respectively, with T classification, nodal metastases, and pathologic grading. The differences in expression between the different T classifications were not statistically significant for both markers. No relationship was observed between E-cadherin or FAK expression and histopathologic differentiation. In contrast, there were significant differences in the expression of E-cadherin (P = .006) and near significant in the expression of FAK (P = .07) between those tumors with and without nodal metastases: nodal metastases were more frequent in the cases with low E-cadherin expression and in the cases with moderate to strong FAK expression. When combining these 2 markers in a new variable, we found that the cases with low E-cadherin expression and moderate to strong FAK expression have the higher incidence of nodal metastases (P = .001) (Table 5).
Another parameter that was significantly associated with lymph node metastases in univariate analysis was decreased histologic differentiation (P = .005), whereas increased T classification showed a near-significant association (P = .08). The results of stepwise logistic regression analysis showed that T classification (P = .04), histological grade (P = .004), E-cadherin expression (P = .03), and the combination of E-cadherin and FAK expression (P = .045) were statistically significant independent predictors of nodal metastases. Therefore, its combination results in a higher predictive value for nodal metastases. By combining T classification, histological grade, E-cadherin expression, and the combination of E-cadherin and FAK expression, an accuracy of 78% was achieved, with a sensitivity of 74% and a specificity of 83% (Table 6).
During the follow-up period, 5 cases developed local recurrence, 21 regional recurrence, and 3 distant metastases. The cases with decreased E-cadherin (staining score below the mean) and increased FAK expression (moderate to strong staining) had a higher incidence of tumor recurrence, which was near statistically significant (P = .06; Table 7). This seems to be owing to a higher incidence of nodal recurrences and distant metastases (Table 7). In consequence, the cases with decreased E-cadherin and increased FAK expression had a significantly lower disease-specific survival (P = .005; Figure 2).
The presence or absence of metastatic disease in cervical lymph nodes is the single most significant determinant of the subsequent therapy and prognosis for patients with head and neck cancer.1 Although preoperative node status assessment has improved after the advent of computer-assisted imaging techniques, the detection of occult micrometastasis can be challenging and unreliable.2,15 The ability to predict cervical lymph node metastases from a primary tumor biopsy sample would provide tremendous advantages for the determination of optimal therapeutic strategies. Thus, interest has developed in identifying molecular markers that would enable selection of patients at risk for lymph node metastasis.
Based on the assumption that metastasis is mainly determined by properties of the primary tumor and its interaction with surrounding structures, identifying gene products involved in the biological processes of tumor invasion and metastasis may provide markers that predict the presence of metastasis.16 However, since invasion and metastasis are very complicated multistep processes, it is likely that more than 1 marker will be needed to assess an individual patient's risk of nodal metastases. Then, it is not surprising that in this study using 2 markers, we have increased the ability to predict the presence of nodal metastases compared with our previous studies that analyzed only E-cadherin or FAK expression.4,14
Various studies have shown that the presence of metastasis can be predicted based on gene expression patterns present in the primary tumor.17,18 In these works, using complementary DNA microarrays, the authors identified gene expression patterns capable of predicting the presence of lymph node metastases with an overall accuracy as high as 86%, which is better than the current clinical diagnostic accuracy. Furthermore, in the study by Roepman et al,18 the predictor (composed of a set of 102 genes) has a predictive accuracy for N0 status of 100%. Then, no false-negative predictions were made, which is most important for the goal of achieving clinical relevance. These studies suggest that the metastatic state can be deciphered from the primary tumor gene expression pattern. The major drawback of these studies is that they are based in a laborious and complex technology that limits its implementation in routine clinical use. In contrast, we have obtained an overall accuracy in predicting nodal metastasis as high as 78%, using only 2 immunohistochemical markers (E-cadherin and FAK) and 2 classic clinicopathologic parameters (T classification and degree of differentiation). These results are not better than those obtained with the most recent techniques of computed tomography, magnetic resonance imaging, ultrasonography, positron emission tomography, and ultrasound-guided fine-needle aspiration biopsy, which have reached a sensitivity of 80% to 85% in detecting occult metastases.2 However, these determinations can be performed easily on standard paraffin-fixed pathologic specimens, facilitating its inclusion in the diagnostic workup and treatment planning of these tumors. This may be especially true in the treatment of the neck that remains N0 after clinical and imaging evaluation. In these cases E-cadherin and FAK expression may play a role in the decision to treat a N0 neck with a neck dissection or with close follow-up (in addition to other factors like T classification, histological differentiation, age, and comorbidity).
Alterations in cellular adhesion seem to have a fundamental role in regulating the invasive tumor phenotype in HNSCC.19 Among the many structures involved in cellular adhesion, the adherens-type junctions share a common association with the microfilament system. In the adherens-type junctions, distinct cytoplasmic components are associated with 2 types of junctions: the cadherins, catenins, and the associated cytoskeletal proteins, which together form 1 distinct structure (the cadherin system), while talin, vinculin, FAK, and the integrins form another (the integrin system).20 A large body of evidence now demonstrates that the loss of E-cadherin expression is related to more invasive and metastatic cancers,3 as we have also demonstrated in our previous study.4
Focal adhesion kinase is a protein tyrosine kinase that becomes phosphorylated and subsequently activated in response to integrin clustering, cellular adhesion to the extracellular matrix, or cellular transformation with the v-Src oncogene.9 Focal adhesion kinase has been shown to be a survival signal for anchorage-independent cells, and the inhibition of FAK in tumor cells has been shown to result in cell death.9 Thus, it is reasonable that tumor cells up-regulate FAK expression to maintain survival as they grow and invade surrounding tissues. The overexpression of FAK has been documented in many types of human cancers, including HNSCC.14,21,22 In these tumors, we have previously shown that FAK was overexpressed in all the tumors analyzed, as compared with normal mucosa, but with different degrees of overexpression.14 In addition, we have observed that the cases with moderate to strong FAK expression have a higher incidence of nodal metastasis, although the differences did not reach statistical significance. In agreement with our results, FAK overexpression has also been previously correlated with tumor invasiveness and lymph node metastasis in other types of cancer.10- 12
On the other hand, integrin-dependent cell-matrix adhesions and cadherin-mediated cell-cell contacts can communicate with each other. Such crosstalk contributes to an altered balance between the 2 adhesion types and is likely to contribute to cancer progression. Usually, cells that assemble dynamic cell-matrix contacts are more migratory and have less robust cell-cell adhesions, whereas cells that predominantly form cadherin-mediated cell-cell contacts maintain polarized epithelial-cell morphologic structure and typically have less robust focal-adhesion complexes. Focal adhesion kinase has emerged as a mediator of crosstalk between integrin-mediated focal adhesions and intercellular junctions. The effect of FAK on intercellular adhesion might be through regulation of expression or localization of the cadherins themselves.9 In this study, we did not find any relationship between the expression of FAK and the expression of E-cadherin. However, the cases with high FAK expression and low E-cadherin expression have the higher metastatic phenotype. In addition, these cases have a higher incidence of nodal recurrences and distant metastases and a significantly reduced disease-specific survival. These data suggest that these alterations could contribute to cell invasiveness.
In most of the studies, HNSCCs are considered as 1 tumor type. However, although being a histopathological entity, these carcinomas may behave differently according to the different locations within this area. This is reflected by differences in growth pattern, clinical behavior, and prognosis. This variability in the behavior according to the location suggests different intrinsic tumor properties, and, as it was shown in previous studies,23 the genetic alterations responsible for these properties also reflect these differences. Nevertheless, local anatomical circumstances may also play a role. To avoid this problem, in our study we have selected a large number of cases of the same site that were well balanced by tumor stage and grade. However, this also has the limitation that our results may not be applicable to other head and neck sites.
In conclusion, E-cadherin and FAK are abnormally expressed in supraglottic squamous cell carcinomas. The combination of the expression of these molecules in these carcinomas is an independent predictor of lymph node metastases and could increase our ability to identify patients with clinically negative lymph nodes who are at considerable risk for occult metastases. Possibly, with a characteristic expression pattern of a set of invasion factors determined by immunohistochemical analysis, it may be possible to evaluate the nodal status of patients more reliably by studying the biopsy material of the primary tumor, which is relatively easily accessible in most HNSCCs.
Correspondence: Juan Pablo Rodrigo, MD, Servicio de Otorrinolaringología, Hospital Universitario Central de Asturias, 33006 Oviedo, Asturias, Spain (email@example.com).
Submitted for Publication: April 24, 2006; final revision received July 24, 2006; accepted October 8, 2006.
Author Contributions: Drs Rodrigo, Dominguez, Suárez, and Chiara had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Rodrigo, Dominguez, and Chiara. Acquisition of data: Rodrigo, Suárez, Canel, and Secades. Analysis and interpretation of data: Rodrigo, Dominguez, and Chiara. Drafting of the manuscript: Rodrigo and Suárez. Critical revision of the manuscript for important intellectual content: Dominguez, Canel, Secades, and Chiara. Statistical analysis: Dominguez and Canel. Obtained funding: Rodrigo and Chiara. Administrative, technical, and material support: Dominguez, Suárez, Canel, Secades, and Chiara. Study supervision: Rodrigo and Chiara.
Funding/Support: This work was supported by grants PI041537 and PI030463 from the Fondo de Investigaciones Sanitarias, Madrid, Spain. Dr Rodrigo receives funding from Instituto de Salud Carlos III, Madrid, Spain (Plan de intensificación de la actividad investigadora).
Acknowledgment: We thank Aurora Fernández, Inés Argüelles, Olivia García, and Marta Sanchez García for technical assistance.
Financial Disclosure: None reported.