INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Organization of cells in normal tissue and organs requires cells to interact with various extracellular matrix proteins as well as with neighboring cells. These interactions are dependent on cell adhesion molecules (CAMs). Defective interactions between adhesion molecules have a critical role in cancer. Detachment of cells is the first step in the process of metastasis and is dependent on the presence or absence of functioning adhesion molecules. Based on structural and sequence homologies, CAMs can be grouped into four distinct families: integrins, selectins, the immunoglobulin superfamily, and cadherins. Cadherins are transmembrane glycoproteins that function as cell-to-cell adhesion molecules in the presence of calcium and are linked to the cytoskeleton via intermediate proteins, the catenins. Cadherins typically consist of an intracytoplasmatic region, a single membrane-spanning segment and five tandemly repeated extracellular domains. Cadherin interacts with cadherin on another cell through a dimeric zipper constituting a homophilic binding (Figure 1). This interaction involves mainly the histidine-alanine-valine (HAV) sequence of the first extracellular domains. The intracellular domain is also of great importance and this is illustrated by the observation that when partial deletions of the intracellular domain are present, E-cadherin cannot function as a cell-cell adhesion receptor, even though the extracellular domain remains intact. In experimental studies, highly invasive tumor cells have been shown to lose their invasiveness when transfected with a normal E-cadherin gene, localized on chromosome 16q22.1 (1). Furthermore, anti-E-cadherin antibodies can induce invasive behavior in tumor cells and this supports the notion that impaired E-cadherin expression on tumor cells can be associated with malignant behavior, i.e., local invasiveness (2). Clinical studies in patients with a wide variety of human malignancies have shown that reduced E-cadherin expression is associated with dedifferentiation and lymphogenous spread of the tumor (3). In one study by Böhm and coworkers, reduced expression of E-cadherin was found to be associated with moderately and poorly differentiated squamous and small cell carcinoma in a limited number of patients with lung cancer (11).

View larger version (43K): [in this window] [in a new window]   Figure 1.   Schematic representation of cadherin-mediated cell-cell adhesion.

Expression of E-cadherin was also found to be reduced in the metastatic lymph nodes. The main objective of the present study was to further investigate the clinical significance of E-cadherin expression in non-small cell lung cancer (NSCLC).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients and Tumors

The primary tumors of 111 patients who underwent complete resection for T1 and T2 non-small cell lung cancer (65 squamous, 29 adeno-, and 17 large cell anaplastic carcinomas) between January 1989 and July 1996 in the De Wever Hospital, Heerlen, the Netherlands, were studied retrospectively. A mediastinoscopy previously performed was negative prior to surgery in all cases. During pulmonary resection a systematic mediastinal lymph node dissection was performed in all patients, contributing to highly accurate surgical staging (pTNM) (12). Since the majority of patients did not have lymph node metastasis we stopped evaluating patients with N0 stage after 1995 in order to have a balanced number of patients in each subgroup of lymph node stage. Positive lymph nodes were detected in 63 patients. The TNM classification as described by Mountain was used in the post-operative staging of the disease (13). In order to have comparable patient groups, only those patients with a solitary tumor without extrapulmonary extension were selected for the study, i.e., the maximal tumor stage was T2. Patients were grouped according to lymph node staging. Patients with ipsilateral mediastinal lymph node metastases received radiotherapy post-operatively. The histological criteria of the World Health Organization classification for lung cancer were used for typing (adeno-, squamous cell, and large cell undifferentiated carcinoma) and grading (well, moderately, poorly, and undifferentiated carcinoma) of NSCLC (14). Follow-up on all patients was available for at least 10 mo and a maximum of 94 mo with a median of 20 mo.

Immunohistochemistry

The majority of the clinical studies with E-cadherin have been performed on frozen sections. Because there are reports in which it has been shown that some monoclonal antibodies to E-cadherin demonstrated varying immunoreactivity in formalin-fixed, paraffin-embedded tissue, a pilot study was performed to establish if paraffin-embedded tissues showed the same staining results as frozen sections using HECD-1 monoclonal antibody ([IgG1]:1:400; Takara, Otsu, Japan). Eight tissue sections chosen at random from patients with NSCLC were used for this study. The staining procedure of both groups was identical and is described in detail below.

In the frozen sections, four slides showed artifacts so that staining results were unreliable, whereas the pretreated paraffin tissues showed strong expression in all cases. Since paraffin-embedded tissues stained at least as well as frozen tissue, all lung resection specimens and the lymph node metastases were retrieved for the present study. Sections 4 µm thick were cut from the paraffin-embedded blocks, mounted on APES-coated slides and air-dried overnight at 37° C. For immunostaining the sections were deparaffinized in xylene and rehydrated in a descending ethanol series. Endogenous peroxidase activity was blocked by immersion for 10 min in 3% hydrogen peroxidase in methanol, after which the slides were rinsed in phosphate-buffered saline (PBS; pH = 7.2-7.4). The slides were then placed in a 0.1 M citrate buffer (pH = 6.0) and boiled for 10 min in a microwave oven at 750 W. After preincubation with 1% bovine serum albumin (Sigma, Zwijndrecht, The Netherlands)/PBS for 10 min, tissue sections were incubated overnight at room temperature in a humidified chamber with the primary antibody E-cadherin (Clone HECD-1; [IgG1]1:400; Takara). After incubation, the sections were rinsed in PBS and incubated with the secondary biotinylated goat-anti-mouse Ig (1:400; DAKO A/S, Glostrup, Denmark) for 45 min at room temperature. After washing in PBS, the slides were incubated with streptavidin conjugated with horse radish peroxidase (1:600; DAKO A/S). The peroxidase reaction was developed with 3,3-diaminobenzidine/0.002% H₂O₂ solution (Sigma). Finally, the sections were counterstained with Harris' hematoxylin, dehydrated, cleared in xylene and finally embedded in Entellan.

Multitissue slides with epithelial and nonepithelial tissues were used as controls. In addition, normal bronchus epithelium, normal mucosal glands as well as pneumocytes within the tumor sections were used as positive internal controls. In three cases, the normal bronchus epithelium failed to show a clear staining with HECD-1 and these cases were excluded from the study. Because of tumor heterogeneity, results were classified semiquantitatively into three categories according to the proportion of tumor cells which were immunoreactive for E-cadherin. Cytoplasmatic as well as membranous staining was considered positive. Cut off points were determined beforehand. When clear staining was present in less than 10% of the tumor cell population the result was defined as negative (). When more than 10% but less than 50% of the tumor cells stained for E-cadherin the result was defined as weakly positive (+). When more than 50% of the tumor cells stained positive the result was defined as strongly positive (++). When the overall staining intensity was faint the result was considered negative, regardless of the proportion of cells that stained. All tumor slides were examined at random by two different investigators (M.A.S., P.H.M.H.T.). In case of disagreement the slides were reviewed again and a consensus score was noted.

Statistical Methods

For statistical analysis frequency tables, chi-squared test and Pearson's correlation coefficient were used. Life-table probabilities of overall survival (i.e., survival without regard to cause of death) were calculated by the method of Kaplan and Meier, and differences in survival between subgroups were compared with the log-rank test. Overall length of survival was measured starting from the day of surgery. Interobserver reproducibility of the E-cadherin classifications was compared using kappa statistics (15). By using a Cox proportional-hazard regression model we examined the prognostic value of several variables on the hazard rate. The SAS system was used for statistics. For the Cox proportional-hazard regression we used SPSS for Windows.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Clinicopathological Variables

The distribution of sex, age, and tumor type and differentiation among the three groups (N0, N1, N2) did not differ significantly (Table 1). Squamous cell carcinoma was the most predominant subtype in all stages, in agreement with the epidemiology of lung cancer in Western Europe.

Patterns in E-Cadherin Expression

E-cadherin expression was predominantly confined to the cell membrane in some of the tumors, while in others E-cadherin expression was diffusely cytoplasmatic. In the vast majority, however, the staining pattern was membranous as well as cytoplasmatic. Because there was no clear opinion regarding the predominant staining pattern in the individual tissue slides, we have refrained from pattern correlations.

Correlations of E-Cadherin Expression and Tumor Characteristics

E-cadherin expression turned out to be strongly correlated with lymph node stage. The tissue sections with the highest proportion of E-cadherin positive tumor cells were most common in the patients without lymph node metastases (N0) (63%). The percentage of tumors with strong E-cadherin expression declined from 46% (N1) to 8% (N2). E-cadherin expression classified as negative was predominant in patients with mediastinal lymph node metastases, 68% as opposed to 20% in the patients without lymph node metastases (Figure 2) (Pearson correlation coefficient 0.52, p = 0.0001). Furthermore, E-cadherin expression was significantly correlated with increasing tumor differentiation (Pearson correlation coefficient 0.27, p = 0.005). The classification of E-cadherin expression in the metastases usually corresponded with that of the primary tumor, but there were a few exceptions in each direction (Tables 2 and 3) (p < 0.0001). In the evaluation of the tumor sections, there was minor disagreement on E-cadherin expression between the observers in 25 cases, with a major disagreement only in one case. A minor disagreement is defined as a difference of one scoring level, whereas in a major disagreement there is a difference of two scoring levels. The kappa value was 0.63 for the primary tumors and 0.74 for the metastases so that we can conclude that the interobserver agreement for E-cadherin expression was good. There was no significant correlation between E-cadherin expression and the histological tumor cell type.

View larger version (16K): [in this window] [in a new window]   Figure 2.   Correlation between E-cadherin expression and lymph node stage.

E-Cadherin and Clinical Outcome

In the Kaplan and Meier patient survival estimates increased E-cadherin was found to be significantly correlated with improved survival (log rank, p = 0.006). This was demonstrated in the patient group with the highest proportion of E-cadherin positive cells in which 60% of the patients were estimated to be alive at 36 mo, compared to only 32% in the patient group classified as negative for E-cadherin (Figure 3). In multivariate Cox regression analysis E-cadherin expression had a significant effect on survival, whereas sex, age, and tumor type and differentiation failed to show such effects. Since E-cadherin expression and lymph node stage were highly correlated, they were not found to be independent prognostic factors in a multivariate analysis. In calculating the hazard ratio for E-cadherin expression we found that patients who were classified as negative for E-cadherin expression, had a hazard ratio of 1.4 as compared to the E-cadherin weak positive group and a hazard ratio of 2.41 as compared to the patient group classified as strongly positive. As expected, lymph node stage was also significantly correlated with patient survival (log rank, p = 0.03). Additionally we have performed Kaplan and Meier patient survival estimates with stratification for E-cadherin expression on subsets of patients with different lymph node stages. No statistically significant difference in survival was found in these subsets of patients (Table 4). However, a larger number of patients within each lymph node stage is needed to give a definite answer on this subject.

View larger version (13K): [in this window] [in a new window]   Figure 3.   E-cadherin expression and overall patient survival estimates according to Kaplan and Meier.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The present data have shown that reduced expression of E-cadherin in non-small cell lung cancer is significantly correlated with increased lymphogenous metastasis, tumor dedifferentiation and poor survival. We have also noticed a heterogenous distribution of E-cadherin in the tumor sections, which is explained by tumor heterogeneity as well as by instability of E-cadherin expression (16, 17). Semiquantitative classification of E-cadherin expression with immunostaining by two investigators has demonstrated good reproducibility. In the majority of tissue sections we always observed some cytoplasmatic staining as well as membranous staining. This observation was surprising because E-cadherin is linked to the cell membrane but our observation is shared by other investigators (18, 19). A possible explanation is that E-cadherin precursor proteins in the cytoplasm may be recognized by HECD-1. Furthermore, it has been suggested that functional or structural abnormalities of the E-cadherin molecule occur in tumors along with reduced E-cadherin expression, resulting in cytoplasmatic staining (18). Unlike other investigators, we have found no apparent loss of immunoreactivity during formalin fixation and paraffin embedding as compared to the findings in frozen sections. Expression of cadherins changes dynamically during embryogenesis and is considered an important determinant of tissue morphology. The role of cadherins later in life remains crucial because multicellular organisms cannot exist without the association of cells (20). Normal epithelial tissues exhibit stable E-cadherin expression even when inflammation is present. In contrast, E-cadherin expression is known to be heterogenously distributed in carcinoma cells. Moreover, clinical studies have reported that impaired expression of E-cadherin was closely correlated with tumor dedifferentiation and lymphogenous metastasis in several carcinomas. Different mechanisms that lead to inactivation of E-cadherin expression have been proposed in human cancer. Translational disorders or allelic loss of the E-cadherin gene have been reported, but these have only been described in a minority of human cancers (18).

It has been demonstrated that the intracellular part of the E-cadherin molecule is also responsible for the stability and function of the extracellular domain of E-cadherin (1). Although other factors involving the motility of cancer cells without affecting E-cadherin should also be considered, low E-cadherin expression seems to have a crucial role in cell-cell detachment and increased invasiveness (21). It has been postulated that as a result of decreased E-cadherin expression, there is a loss of negative feedback control which is determined by cell-cell interaction resulting in increased proliferation of cells (19). Further investigation is necessary to clarify if reduced E-cadherin expression offers a statistically significant prognostic factor independent of lymph node stage. Identifying relevant prognostic factors in NSCLC is important in predicting clinical outcome and in the future it might enable us to find the most optimal treatment regimen for the individual patient.