Weak Expression of Focal Adhesion Kinase (pp125FAK) in Patients with Cervical Cancer Is Associated with Poor Disease Outcome

In the past 10 years, new molecular biological approaches in cancer treatment have emerged and significant progress has been made through the discovery of specific inhibitors of target molecules involved in tumor cell signal transduction (1). It is recognized that abnormal cell signal transduction arising from protein tyrosine kinases is implicated in the initiation and progression of a variety of human cancers, including gynecologic malignancies (2–4). Several classes of proteins involved in signal transduction, such as integrins and kinases, are altered in cells possessing invasive or metastatic capabilities (4). Besides receptor tyrosine kinases, pure cytoplasmic tyrosine kinases are also implicated in tumorigenesis and anchorage-independent growth of cells. An early and crucial event in this regard is the phosphorylation of proteins on tyrosine residues. A pertinent downstream target is pp125 focal adhesion kinase (FAK), which binds to the β₁-integrin subunit (5, 6).

FAK represents a 125-kDa nonreceptor tyrosine kinase, which is ubiquitously expressed throughout development, widely detectable in adult tissues, and specifically localized in focal adhesions that represent contact sites for cellular interactions with the extracellular matrix. FAK holds a key position in the signal transduction network and is phosphorylated in response to clustering of integrins, cell spreading, or formation of focal adhesions (5–7). Thus far, at least six tyrosine residues (Y397, Y407, Y576, Y577, Y861, and Y925) have been identified as phosphorylation sites. Autophosphorylation of the major site Y397 leads to recruitment of Src kinase and subsequent stepwise phosphorylation of FAK on the other sites, which in turn enhances FAK kinase activity (5). Due to its unique structure, FAK is able to act as a scaffold for the assembly of multiprotein signaling complexes and triggers important downstream signaling events, such as the activation of mitogen-activated protein kinases via Ras (8).

From a functional point of view, FAK is involved in cell cycle regulation (9) and apoptosis (10). Several studies confirmed a pivotal role of FAK in tumorigenesis and metastasis with elevated expression levels in various human malignancies (e.g., colon, breast, and ovarian cancers; refs. 11–15). These findings were associated with enhanced cell migration (16) and proliferation (17). In lung cancer, an increased phosphorylation of FAK was correlated with nodal involvement and disease-free survival (18). The functional importance of FAK is supported by experiments showing that targeting FAK by antisense oligonucleotides leads to an inhibition of cell migration, invasion, and proliferation as well as an induction of apoptosis and enhanced sensitivity to camptothecins (19, 20). Although the mechanisms underlying the increased expression of FAK in tumor cells are not fully understood, amplification of the FAK gene has been reported in cancer cell lines (11).

Cervical cancer is the second most common malignancy among women worldwide (21). According to the IARC data, ∼30,000 new cases of cervical cancer were diagnosed among women in the European Union in 2004 (22), and cervical cancer is still one of the leading causes of cancer mortality in developing countries (21). Thus far, there are only few reports linking FAK to cervical carcinogenesis (23–25). McCormack et al. described elevated FAK expression and activity in HPV18-immortalized human genital epithelial cells and human papillomavirus–containing carcinoma cells compared with controls (24). The authors suggested a role for FAK-mediated signaling pathways in cervical carcinogenesis. In contrast, Moon et al. found that the expression of FAK was not significantly different in cervical carcinoma cells compared with controls (25).

We therefore aimed to characterize the expression of FAK protein in 162 patients with cervical cancer and correlate our findings with clinicopathologic variables and patient survival.

Patients and tissue samples. Patients (n = 166) with cervical carcinoma who were treated between 1988 and 1995 were enrolled. The tissue samples were obtained at the time of surgery at the Department of Obstetrics and Gynecology, Medical Faculty Hospital (Plzen, Czech Republic). The institutional review board approved the investigation protocol. Paraffin-embedded tissues were retrieved from the pathology archive of the hospital. All patients were treated by radical hysterectomy and pelvic lymphadenectomy, properly staged according to the 1995 International Federation of Gynecology and Obstetrics classification. Ninety-four percent of patients underwent adjuvant pelvic radiotherapy. Histologic slides were reviewed regarding the histologic type, grade, stromal invasion, lymphatic vascular space invasion (LVSI), and pelvic lymph node status by two experienced pathologists (Department of Pathology, Medical Faculty Hospital) blinded to the clinical data. All tissue specimens and slides were reexamined at Freiburg University Medical Center (Freiburg, Germany) by an experienced pathologist (A.z.H.), confirming the previous diagnosis. Serial sections were prepared for H&E staining and immunohistochemistry analysis. Clinical data, including follow-up data and pathologic characteristics of the tumor (e.g., tumor type, tumor size, lymph node status, stromal invasion, tumor grade, LVSI, and GOG score), were provided by the Department of Obstetrics and Gynecology, Medical Faculty Hospital. The GOG score was calculated as proposed by the Gynecologic Oncology Group (26) by multiplying the relative risks for depth of tumor invasion, tumor size, and LVSI. Because only 18 (10.8%) patients in our study had tumor size >40 mm, we used 35 mm as a cutoff to achieve adequate patient numbers for statistical analysis. Characteristics of patients and tumors are summarized in Table 1.

Immunohistochemistry. Immunohistochemistry was done at the Department of Obstetrics and Gynecology, Freiburg University Medical Center. Routinely formalin-fixed and paraffin-embedded specimens were used to study the expression of FAK. To reach this goal, the monoclonal mouse anti-pp125FAK antibody mAb4.47 (Upstate Biotechnology, Charlottesville, VA) in combination with heat induced antigen retrieval and indirect immunoperoxidase technique was applied as described previously (14). mAb4.47 specifically recognizes total (i.e., phosphorylated and nonphosphorylated) FAK on formalin-fixed and paraffin-embedded sections⁶

A score was applied to classify the intensity of immunohistochemical staining as absent (0), weak (1+), moderate (2+), or strong (3+). The percentage of positive cells for each intensity was estimated for the whole tumor on one section. The extent of expression was simplified as focal (when <50% of cells were stained) and diffuse (when >50% of cells were stained). In all samples investigated, a diffuse and widespread staining of the tumor cells was observed. Therefore, the resulting immunohistochemistry score was based exclusively on the intensity of immunohistochemical staining.

The slides were evaluated independently by three investigators blinded to the clinical data.

Statistical analysis. Survival probabilities were calculated by the product limit method of Kaplan and Meier. A log-rank test was applied for assessing statistical differences between survival curves. Multivariate analysis was done using the Cox proportional regression analysis.

The first Cox model included all variables significantly related to patient survival in the univariate analysis (FAK expression, pelvic lymph node metastasis, depth of tumor invasion, GOG score, tumor size, and LVSI). Starting from this model, a backward elimination procedure was done. The nonsignificant variables GOG score, tumor size, and LVSI were removed from the model. The final Cox model included FAK expression, pelvic lymph node metastasis, and depth of tumor invasion. Subsequently, the statistical analysis was also done by means of a forward selection, starting from the null model. Successive adding of the variables mentioned above significantly improved the Cox proportional model in the first three steps.

Correlation between clinicopathologic variables and FAK expression was tested with the χ² test. Ps < 0.05 were considered statistically significant. The SPSS version 13 (SPSS, Inc., Chicago, IL) was used for the calculations.

Patients. A total of 166 patients diagnosed with early-stage cervical cancer were enrolled (Table 1). The mean follow-up time was 72.4 months (95% confidence interval, 68.5-76.2; range, 1-166). The median age at diagnosis was 40.5 years (range, 23-68). The distribution of histologic type, tumor size, histologic grade, LVSI, and pelvic lymph node metastasis was in accordance with previously published clinical trials (28, 29). All patients received appropriate surgical treatment and adjuvant therapy (see Materials and Methods). Overall, 32 patients died of their disease during follow-up and estimated survival rates were 91.6%, 84.9%, 81.3%, and 69% at 1, 3, 5, and 10 years, respectively, which is comparable with published data (21, 26, 28–30).

Immunohistochemistry analysis. Immunohistochemistry analysis was done to determine the expression and subcellular localization of FAK in tumor tissue and results were retrieved in 162 of 166 samples. FAK was expressed in the carcinoma cells of all cervical cancer samples. Barely any FAK was detected in adjacent normal cervical epithelium. A weak and focal staining was observed exclusively in the basal layers of the cervical epithelium. Specific cytoplasmic and occasionally membranous FAK expression was restricted to the dysplastic and invasive carcinoma cells (Fig. 1). The staining pattern was diffuse and widespread, and almost all of the tumor cells were stained for FAK. However, clear differences in staining intensity were observed. Fifty-five of 162 (34%) tumor samples showed weak expression (score 1+), 63 (39%) showed moderate expression (score 2+), and 44 (27%) showed strong expression of FAK (score 3+).

Survival analysis. For survival analysis, data from 162 patients with known FAK status were available. Patients diagnosed with a tumor expressing weak amounts of FAK revealed a similar distribution of age and histologic subtype compared with those with moderate or strong FAK expression (data not shown). Importantly, patients with weak FAK expression (group A) were characterized by a significantly poorer survival compared with those with moderate (group B) and strong (group C) FAK expression (P = 0.002, log-rank test; Fig. 2). There was no difference between the survival curves of patients from groups B and C (P = 0.87, log-rank test). For patients of groups B and C, the mean survival times were 126.9 months (95% confidence interval, 117.1-136.8) and 149 months (95% confidence interval, 135-163), respectively, whereas the mean survival time for patients of group A was 94.2 months (95% confidence interval, 77-111). According to FAK expression, Kaplan-Meier analysis revealed estimated survival rates at 5 years of 88.6% and 87.3% for patients of groups C and B, respectively, compared with 67% for patients of group A (Fig. 2). Estimated survival rates at 10 years were 88.6%, 87.3%, and 42.7% for groups C, B, and A, respectively. In accordance with these data, patients who died of their disease were more likely belonging to group A (P = 0.001, χ² test; Table 2).

Furthermore, the univariate survival analysis revealed significance for the variables pelvic lymph node status (N₀ versus N₁; P = 0.001), LVSI (yes versus no; P = 0.042), tumor size (<35 versus >35 mm; P = 0.044), depth of tumor invasion (<5 versus >5 mm; P = 0.022), and GOG score (<40 versus 40-120 versus >120; P = 0.029) but not for histologic grade (P > 0.05, all log-rank test; Fig. 3).

Relationship between FAK expression and clinicopathologic characteristics.Table 2 summarizes the relationship between different clinicopathologic characteristics and FAK expression in early-stage cervical cancer. A significant inverse correlation was observed between FAK expression and pelvic lymph node metastasis. Patients with tumors expressing weak amounts of FAK were more likely to have positive pelvic lymph nodes (P = 0.026). Furthermore, there was a significant inverse correlation between FAK expression and recurrent disease (P = 0.013). The correlation between FAK expression and the variables histologic grade, depth of tumor invasion, GOG score, LVSI, tumor size, and histologic type (squamous cell carcinoma versus adenocarcinoma) did not reach statistical significance (P > 0.05, all χ² test).

Relationship between different clinicopathologic characteristics. The presence of LVSI was significantly associated with pelvic lymph node metastasis (P = 0.001), histologic grade (P = 0.011), depth of tumor invasion (rs = 0.34; P = 0.004), tumor size (rs = 0.28; P < 0.001), and GOG score (rs = 0.62; P < 0.001). As expected, there was a positive correlation between GOG score and depth of tumor invasion (rs = 0.85; P < 0.001), tumor size (rs = 0.76; P < 0.001), and LVSI (rs = 0.62; P < 0.001) (26). Furthermore, we found a positive correlation between tumor size and depth of tumor invasion (rs = 0.55; P < 0.001) and between tumor size and pelvic lymph node metastasis (rs = 0.26; P = 0.001). Depth of tumor invasion was associated with pelvic lymph node metastasis (rs = 0.31; P < 0.001, all χ² test). There was no relationship between histologic grade and the variables tumor size, depth of tumor invasion, pelvic lymph node metastasis, and GOG score.

The multivariate Cox regression analysis (Table 3) revealed FAK expression and pelvic lymph node metastasis as significant independent predictors of survival, whereas the depth of tumor invasion did not reach statistical significance.

Given the important role of FAK in cell adhesion, cell migration, and signal transduction, which are of major importance in tumor invasion and metastasis, it seemed reasonable to investigate FAK expression in cervical cancer and its effect on patient outcome in detail. Slides from formalin-fixed and paraffin-embedded tumors diagnosed between 1988 and 1995 in combination with the FAK-specific mAb4.47 were used. mAb4.47 seems to be especially suited for this investigation, as this monoclonal antibody was selected to recognize an epitope at the NH₂-terminal part of FAK even after formalin fixation and paraffin embedding of the tissues. Although this antibody recognizes FAK regardless of Y397 phosphorylation⁶ (27), there is a large body of evidence that the total amount of FAK correlates well with the activity of FAK (16, 24, 31, 32). FAK was expressed specifically in the cytoplasm of invasive human cervical carcinoma cells. Although none of the carcinoma specimens (n = 162) investigated stained negative for FAK, clear differences in staining intensity reflecting the amount of FAK were seen. In a first attempt to investigate the relevance of FAK in cervical cancer, the correlation between FAK expression and survival of the patients was analyzed. In addition to already established variables (i.e., pelvic lymph node status, LVSI, tumor size, depth of tumor invasion, and GOG score), FAK was found to be significantly related to patient survival in the univariate analysis. Most importantly, patients with tumors expressing weak amounts of FAK were characterized by a significantly poorer 5- and 10-year survival compared with those diagnosed with tumors expressing moderate and strong amounts of FAK. These considerable differences, according to FAK expression, together with our well-documented and large patient cohort suggest an important functional role of FAK in cervical cancer progression. Moreover, we were able to show that FAK expression was an independent prognostic variable even in the multivariate analysis; therefore, FAK might be a potential diagnostic tool to differentiate “low-risk” from “high-risk” cervical cancer patients. In particular, FAK could serve as a marker identifying patients without pelvic lymph node involvement who are at high risk of recurrence. It is of note that weak FAK expression was correlated with pelvic lymph node metastasis and recurrent disease, but there was no correlation between FAK expression and any other clinicopathologic variable in our study. In accordance with published data, the pelvic lymph node status was an independent predictor of survival in the multivariate analysis but not the other investigated clinicopathologic variables. In a previous study, we were able to show that pelvic lymph node status is the most important prognostic factor in cervical cancer and even a better prognosticator than clinical International Federation of Gynecology and Obstetrics stage (33).

According to the standards of the Medical Faculty Hospital at the time of primary therapy, almost 94% of the patients enrolled in this retrospective study underwent adjuvant pelvic radiotherapy. Even considering that most of the patients have been overtreated, this is without concern on our immunohistochemistry analysis.

Although we detected increased expression of FAK in tumor tissue, which is in accordance with published data (12, 14, 15, 32), the correlation between FAK expression and patient survival was initially in sharp contrast to our expectation. Increased expression of FAK has been described in a variety of human malignancies compared with controls and has also been correlated with poor patient outcome in recent publications (31, 32). However, there is also one report of reduced FAK expression in metastatic liver tumors compared with their matched primary human colorectal adenocarcinoma samples (34), suggesting that FAK might have different roles in different tumors or during distinct stages of tumor progression. A few studies showed an up-regulation of FAK expression in the transformational process from physiologic tissue via in situ carcinoma into invasive cancer (12, 23, 27, 35, 36), suggesting that up-regulation of FAK might be an early event in carcinogenesis. In line with this, precursor lesions adjacent to the invasive tumor component revealed strong FAK immunoreactivity. By contrast, if FAK expression is detectable at all in normal cervical epithelium, it is restricted to the basal cell layers, where proliferation occurs. Comparable results were obtained from studies investigating squamous epithelium and FAK expression patterns (32, 37, 38).

A recently published article by Van Nimwegen et al. showed that FAK is required in the early phase of metastasis formation in an animal model (39). It is still unclear whether continued FAK expression and activity is required for maintenance of the tumor phenotype, but this is an important factor in determining the clinical use of potential FAK inhibitors (40).

In our study, weak FAK expression in cervical cancer tissue was associated with poor patient outcome. Our data are supported by a recent study showing increased carcinoma cell migration after inhibition of FAK in HeLa cells that are derived from cervical cancer (41). A negative role for FAK during the invasion of carcinoma cells in vitro has also been described by Lu et al. (42). In these experiments, tumor cell invasion was initiated by dephosphorylation and down-regulation of FAK. Thus, tumor cells might possess different types of invasive mechanisms, explaining the apparently contradictory role of FAK for tumor cell invasion (42–44). A lost or reduced FAK expression in highly malignant tumor cells has been described by other authors as well (25, 45, 46). Hence, FAK and integrin activation might be features of early-stage or intermediate-stage cancer cells. Further tumor progression may lead to a completely anchorage-independent phenotype where these signals are no longer required. This in turn may lead to down-regulation of FAK expression as observed in our study.

While comparing these results, one should keep in mind that methods to investigate and quantify FAK expression differ considerably (e.g., specificity of the antibodies used for immunohistochemistry and Western blotting, investigation of tissue extracts containing various amounts of tumor and nontumor cells, reverse transcription-PCR, etc.).

In conclusion, we observed a correlation between FAK expression in early-stage cervical cancer and patient outcome. Thus, cancers showing weak FAK expression might be already advanced on their way to “full malignancy” compared with those with moderate and high intratumoral FAK expression. Alternatively, these cancers might represent a subgroup relying on FAK-independent mechanisms for invasion (40, 47). Furthermore, other oncogenic kinases, such as Src, may have a nonnegligible effect on tumor progression, as constitutively active Src was shown to be able to bypass the need for FAK in promoting the turnover of focal contacts (48). Clearly, other molecular-based techniques are required to investigate these issues in more detail.

In summary, the correlation between FAK expression in cervical cancer tissue and the survival of these patients points to an important functional role of this protein in development and/or progression of the disease. Most interestingly, our results show that weak expression of FAK in invasive cervical cancer is a strong independent predictor of poor disease outcome.

We thank Prof. Schulte-Mönting (Institute of Medical Biometry and Medical Informatics, University of Freiburg) for statistical support.