A Switch from E-Cadherin to N-Cadherin Expression Indicates Epithelial to Mesenchymal Transition and Is of Strong and Independent Importance for the Progress of Prostate Cancer

There is currently a need for improved prediction of long-term outcome in patients with clinically localized prostate cancer (1, 2). Among important molecular determinants, alterations in cell adhesion molecules are involved in both embryologic development and tumor progression (3). The classic cadherins (E-cadherin, N-cadherin, and P-cadherin) are transmembrane adhesion glycoproteins (4), which link to the actin cytoskeleton by different catenins (5). As a feature of aggressive tumors, epithelial to mesenchymal transition (EMT) is characterized by reduced E-cadherin and increased N-cadherin expression, contributing to a stroma-oriented cellular adhesion profile with increased tumor cell motility and invasive properties, and this molecular profile has recently been reported in several tumors (6–8). However, it is not known whether the EMT phenotype is important for the progress and prognosis of human prostate cancer. Some studies have indicated reduced E-cadherin expression in subgroups of aggressive prostate cancer (9, 10). One study showed N-cadherin positivity in only 5% of prostatic cancers, but with no clinicopathologic correlations (11), whereas another report showed no expression (12). Simultaneous up-regulation of N-cadherin and down-regulation of E-cadherin have been found in more aggressive prostate cancer cell lines (6, 13, 14), but survival data related to the EMT profile are lacking.

P-cadherin is expressed mostly in basal and proliferating layers of epithelia and decreases with cellular differentiation (15). In breast cancer, increased expression has been related to the “basaloid phenotype” and a worse prognosis (16). In benign prostate tissue, P-cadherin has been promoted as a novel basal cell marker (17, 18) but has also been variably detected in subgroups of prostate cancers (12, 18, 19). Prognostic implications of this finding are not known.

β-Catenin and p120^CTN are attached to the cadherin complex (5, 20) and are required for the adhesive properties of classic cadherins (20). Reduced membranous β-catenin and p120^CTN, and increased expression of nuclear β-catenin, have been associated with aggressive prostate cancer (11, 21–24), but no significant survival differences were noted for p120^CTN (11, 24).

On this background, the aim of our study was to examine a panel of cell adhesion molecules (E-cadherin, N-cadherin, P-cadherin, β-catenin, and p120^CTN) with reference to clinicopathologic phenotype and prognostic information in prostate cancer and with special focus on the significance of EMT.

As described previously (25, 26), a consecutive series of 104 men treated by radical prostatectomy for clinically localized prostate cancer during 1988 to 1994, with long and complete follow-up, was included. Clinical stage T1/T2 disease, negative bone scan, and generally good health were the prerequisites for radical retropubic prostatectomy. The majority of cancers in this series are clinical stage T2 and presented before the prostate-specific antigen (PSA) era started in Norway in the mid-1990s. Consequently, the prevalence of adverse prognostic factors, such as capsular penetration, seminal vesicle invasion, and positive surgical margins, is rather high compared with most contemporary series. No patients treated by radical prostatectomy received radiotherapy before biochemical failure or clinical recurrence.

Two separate Gleason scores were recorded: one standard score from radical prostatectomy specimens and one local score on the 1 to 2 cm² tumor area from which tissue microarray samples were punched. Further, WHO histologic grade (26), largest tumor dimension, capsular penetration, seminal vesicle invasion, involvement of surgical margins, presence of lymph node and skeletal metastasis, clinical and pathologic stage, and serum PSA (s-PSA) level before and after surgical treatment were recorded. Tumor cell proliferation (Ki-67; ref. 27), microvessel density (26), vascular proliferation index (25), glomeruloid microvascular proliferation (28), and expression of PTEN (29) and p27 (29) were included from previous studies for comparison.

The entire prostate was cut into 5-mm transverse serial sections. Based on H&E-stained slides, the area of highest histologic tumor grade was identified and cut out of the paraffin blocks, reembedded in paraffin, and sectioned for immunohistochemistry for estimation of Ki-67 (27), microvessel density (26), vessel proliferation by Ki-67/F-VIII double staining (25), and glomeruloid microvascular proliferation (28).

The tissue microarray technique has been described and validated in several studies (30–33). This method has been used in our studies of prostate cancers (25, 29) and expression of cell adhesion markers in other tumors (34, 35). Representative tumor areas were identified on H&E slides, and three tissue cylinders (diameter of 0.6 mm) were punched from the donor block and mounted into a recipient paraffin block (30).

Immunohistochemical staining was done on formalin-fixed and paraffin-embedded tissue using 5-μm sections from tissue microarray blocks as described (25). Western blot analysis confirming the specificity of the antibodies against E-cadherin, N-cadherin, P-cadherin, and β-catenin has been done and described in a recent report from our research group (35).

E-cadherin. After microwave antigen retrieval (boiling for 15 min at 350 W) in citrate buffer (pH 6.0), slides were incubated overnight at 4°C with the monoclonal mouse E-cadherin antibody M3612 (diluted 1:400; DAKO). Staining was done on TechMate 500 automated slide processing equipment (DAKO).

N-cadherin. After boiling in Tris-EDTA buffer (pH 9.0) for 20 min at 350 W, sections were incubated for 1 h at room temperature with the monoclonal mouse antibody M3613 (diluted 1:25; DAKO).

P-cadherin. After boiling for 20 min at 350 W in target retrieval solution (DAKO) buffer (pH 6.0), an autostainer (DAKO) was used for staining, with incubation for 1 h at room temperature with the monoclonal mouse antibody C24120 (diluted 1:100; BD Transduction Laboratories).

β-catenin. After microwave retrieval in citrate buffer (pH 6.0; boiling for 15 min at 350 W), tissue microarray slides were incubated for 25 min at room temperature with the monoclonal mouse antibody (clone 14; diluted 1:800; BD Transduction Laboratories) using TechMate 500 automated slide processing equipment for staining.

p120^CTN. After boiling for 15 min at 500 W in target retrieval solution buffer (pH 6.0), an autostainer was used with incubation for 1 h at room temperature with a monoclonal mouse antibody (clone 98; diluted 1:3,000; BD Transduction Laboratories).

Staining for N-cadherin, P-cadherin, and p120^CTN was done using the EnVision-labeled polymer method, with commercial kits (DAKO). For E-cadherin and β-catenin, the standard avidin-biotin method was used. The peroxidase was localized by the diaminobenzidine tetrachloride peroxidase reaction and counterstained with Mayer's hematoxylin. Negative controls were obtained using isotypic mouse immunoglobulin (IgG1). Samples with known reactivity were used as positive controls (e.g., liver, colon, epidermis, endometrium, and multitissue sections).

E-cadherin stained cell membranes consistently in benign and variably in malignant epithelium (Fig. 1A and B). N-cadherin was negative in benign epithelium; mostly incomplete but distinct membranous staining was found in a subgroup of cancers (Fig. 1C and D). Cytoplasmic N-cadherin staining with variable intensity was also recorded. P-cadherin stained basal cells. In malignant cells, positive cases showed mixed cytoplasmic and membranous staining (Fig. 1F); other cases were weak or negative (Fig. 1E). β-Catenin showed a mixed cytoplasmic and membranous staining in benign epithelium; both membranous and nuclear staining were found in malignant epithelium (Fig. 1G and H), and these were positively correlated (P = 0.008). p120^CTN showed membranous staining particularly in basal cells, and variable membranous staining was recorded in malignant epithelium (Fig. 1I and J). No nuclear staining was observed.

For all factors, a staining index (SI; values 0-9) was calculated as a product of staining intensity (0-3) and proportion of positive cells (0% = 0, 1-10% = 1, 11-50% = 2, >50% = 3; ref. 36). Cutoff points for SI categories were mostly based on median values, considering the frequency distribution for each marker; categories with similar survival estimates were merged when appropriate. E-cadherin, membranous β-catenin, and p120^CTN were categorized by median values as strong (SI > 4) or weak (0-4). Membranous N-cadherin and P-cadherin as well as nuclear β-catenin expression were divided by negative (SI = 0) or positive (SI ≥ 1) staining, and cytoplasmic N-cadherin and cytoplasmic P-cadherin were divided by their median values as strong (SI > 3) or weak (0-3).

Postoperatively, s-PSA, locoregional tumor recurrences, distant metastases, and patient survival were recorded (26). Time from surgery until biochemical failure (defined as persistent or rising s-PSA level of >0.5 ng/mL in two consecutive blood samples) was noted. Further, a tumor in the prostatic fossa or evidence of distant metastasis on bone scan, X-ray, or magnetic resonance imaging was recorded as clinical recurrence. The last time of follow-up was December 2001 (25). Median follow-up time was 95 months (7.9 years). No patients were lost because of insufficient data. Sixty-seven patients experienced biochemical failure, 31 patients had clinical recurrence, and 9 patients died of prostate cancer.

Associations between variables were assessed by Pearson's χ² test or the Mann-Whitney U test. Univariate survival analysis was done by the Kaplan-Meier method (log-rank test), and multivariate survival analysis was done by the proportional hazards method (likelihood ratio test). Model assumptions were examined by log-log plots. The Statistical Package for the Social Sciences statistical package 13.0 (SPSS, Inc.) was used.

Largest tumor dimension (median) was 28 mm (range, 10-45 mm). Fifteen tumors (14.4%) had a standard Gleason score of 0 to 6, 34 tumors (32.7%) had a Gleason score of 3 + 4, 33 tumors (31.7%) had a Gleason score of 4 + 3, and 22 tumors (21.2%) had score of 8 to 10. Standard Gleason score was significantly correlated with local Gleason score (Spearman's rho = 0.65; P < 0.0005). Capsular penetration was shown in 72 cases (69.2%), seminal vesicle invasion in 35 cases (33.7%), and positive surgical margins in 55 cases (52.9%). Pelvic lymph node infiltration was found in seven patients (6.7%) at time of surgery. The median preoperative s-PSA was 11.2 ng/mL (range, 1.8-70.0). More details are available (26).

E-cadherin. Weak membrane expression (36%) was significantly associated with all adverse clinicopathologic variables (Table 1), including poor WHO histologic grade (P = 0.001).

N-cadherin. Positive membranous expression (34%) was associated with poor WHO histologic differentiation (P = 0.030), seminal vesicle invasion (P = 0.006), and pelvic lymph node infiltration (P = 0.029; Table 1). Local Gleason score ≥4 + 3 showed a trend (P = 0.083). Cytoplasmic N-cadherin expression was associated with increased proliferation by Ki-67 (P = 0.012).

P-cadherin. Positive membranous expression (44%) was significantly related to increased local Gleason score ≥4 + 3 (P = 0.002) and to the poorly differentiated group by WHO (P = 0.04).

β-catenin. Weak membranous expression (44%) was significantly related to standard Gleason score ≥4 + 3 (P = 0.008; Table 2), poor WHO histologic grade (P = 0.036), and increased preoperative s-PSA (P = 0.019). Positive nuclear expression (47%) was related to the poorly differentiated group by WHO (P = 0.033) and presence of glomeruloid microvascular proliferations (P = 0.05).

p120^CTN. Weak membranous expression (50%) was significantly associated with standard Gleason score ≥4 + 3 (P = 0.003), pelvic lymph node infiltration (P = 0.013; Table 2), and advanced clinical stage (P = 0.047).

Weak expression of membranous E-cadherin was strongly associated with presence of membranous N-cadherin (P < 0.0005), presence of membranous P-cadherin (P < 0.0005), weak membranous expression of β-catenin (P < 0.0005), and weak expression of membranous p120^CTN (P < 0.0005). A nonsignificant trend was observed between presence of membranous P-cadherin and presence of membranous N-cadherin (P = 0.14). Both presence of membranous N-cadherin (P = 0.002) and presence of P-cadherin (P = 0.006) were related to weak expression of membranous p120^CTN. Weak membrane staining of β-catenin was significantly associated with weak membrane expression of p120^CTN (P < 0.0005).

Significant alterations of p27 and PTEN have previously been reported in this series (29). Weak E-cadherin (P < 0.0005) and weak membranous β-catenin expression (P < 0.0005), as well as positive N-cadherin (P = 0.001) and P-cadherin expression (P = 0.006), were all related to low expression of p27. Weak E-cadherin (P = 0.015) and positive membranous P-cadherin expression (P = 0.039) were associated with low PTEN expression.

The prognostic significance of microvessel density and vascular proliferation index has previously been studied in this material (25, 26), and these angiogenic markers were not associated with any of the five cell adhesion proteins.

E-cadherin. Weak membrane expression was strongly associated with shorter time to biochemical failure, clinical recurrence (Table 3; Fig. 2), locoregional recurrence (P = 0.011), skeletal metastasis (P = 0.003), and cancer-specific death (P = 0.012).

N-cadherin. Positive membrane expression was associated with shorter time to biochemical failure, clinical recurrence (Table 3; Fig. 2), and skeletal metastasis (P = 0.046). No significant differences were found for cytoplasmic expression.

EN-switch. Univariate survival analyses were also done for the EN-switch (weak E-cadherin and positive N-cadherin, n = 23; 22%). This subgroup was strongly associated with shorter time to all five end points (Fig. 1).

P-cadherin. Positive membrane staining was significantly associated with shorter time to skeletal metastases (P = 0.036), and nonsignificant trends with shorter time to biochemical failure and clinical recurrence (Table 3; Fig. 2) were observed. No significant differences were found for cytoplasmic expression.

β-Catenin. No significant survival differences were observed (Table 3; Fig. 2).

p120^CTN. Weak membrane expression was associated with shorter time to biochemical failure, clinical recurrence (Table 3; Fig. 2), locoregional recurrence (P = 0.034), and cancer-specific death (P = 0.023).

Cell adhesion markers (E-cadherin, N-cadherin, P-cadherin, β-catenin, and p120^CTN) with P values of <0.15 in univariate survival analyses were included together with preoperative s-PSA, standard Gleason score (≤3 + 4 versus ≥4 + 3), and pathologic stage.

When including individual variables in a simultaneous model (excluding the combined EN-status), E-cadherin [hazard ratio (HR), 2.5; P = 0.019], N-cadherin (HR, 5.6; P = 0.003), and standard Gleason score (HR, 4.3; P = 0.002) were all independent predictors of time to clinical recurrence, whereas E-cadherin (HR, 1.8; P = 0.02), standard Gleason score (HR, 2.9; P < 0.0005), and pathologic stage (HR, 2.7; P = 0.001) all had an independent prognostic effect on time to biochemical failure. In contrast, P-cadherin, β-catenin, or p120^CTN did not show independent prognostic effect for any of these end points.

In a final simultaneous model, excluding E-cadherin and N-cadherin, the EN-status (weak E-cadherin and strong N-cadherin), as an indication of EMT in prostate cancers, consistently showed an independent prognostic effect, stronger than for E-cadherin and N-cadherin separately, together with Gleason score (≤3 + 4 versus ≥4 + 3) using both biochemical failure and clinical recurrence as end points (Table 4).

The prognosis of patients with clinically localized prostate cancer cannot be accurately predicted by standard variables such as preoperative s-PSA, Gleason score, and pathologic stage alone, and there is a need for supplementary prognostic factors (2). The aim of the present study was to explore the significance of cell adhesion markers in comparison with the above-mentioned “triad.” As a novel finding, increased expression of N-cadherin was a strong and independent predictor of clinical recurrence after radical prostatectomy.

We here show an independent prognostic significance of reduced E-cadherin expression as found by others (9–11, 37–40). Reduced expression of E-cadherin in prostate cancer may be caused by DNA hypermethylation (41) and transcriptional (42) and posttranslational mechanisms (43) but probably not DNA mutations (38). E-cadherin repression seems to be a dynamic and partly reversible process, and reexpression has been observed in metastatic prostate cancer (38, 43).

In our series, E-cadherin was not associated with preoperative s-PSA. One explanation might be that lowered E-cadherin is a marker of biological aggressiveness, whereas s-PSA is only a marker of tumor volume. Indeed, high preoperative s-PSA predicted advanced pathologic stage (pT3) but was without independent prognostic effect in predicting clinical recurrence after surgical treatment.

Importantly, a “cadherin switch” with increased N-cadherin and reduced E-cadherin expression had an independent prognostic effect on time to both biochemical failure and clinical recurrence in multivariate survival analyses, stronger than for E-cadherin or N-cadherin separately. Survival data for N-cadherin have not been previously reported for prostate cancer. N-cadherin expression contributes to a stroma-oriented cellular adhesion profile with increased tumor cell motility and invasive properties, indicating a possible EMT (6–8). Thus, cell adhesion molecules might add prognostic information beyond that presently given by histologic evaluation alone (44). Supporting our findings, soluble E-cadherin and N-cadherin were recently studied as serum biomarkers in prostate cancer, suggesting increased level of N-cadherin as a marker of ongoing EMT and tumor progress (45, 46). The higher frequency of positive N-cadherin staining compared with two other studies (11, 12) may be a result of increased detection using a different antibody validated by Western blot (35) in our lab.

Strong P-cadherin staining was found in the basal cell compartment, and this is in concert with earlier studies (12, 17–19). Further, P-cadherin was found to be reexpressed in a subset of prostate cancers, indicating a worse outcome although without independent prognostic power. This seems to be similar to P-cadherin as a myoepithelial marker in benign breast tissues, being reexpressed in the basaloid phenotype of more aggressive breast cancer (16). According to our demonstration of significant associations between both N-cadherin and P-cadherin with reduced E-cadherin-expression, the term cadherin switch should also include reexpression of P-cadherin in some prostate cancers. However, no independent prognostic effect was observed for P-cadherin, β-catenin, or p120^CTN, and these factors seem to be less important for prognostication.

In the present study, we asked whether there was an association between alterations of cell adhesion molecules, as evidence of EMT, and activation of the vascular system. Interestingly, one study showed a correlation between increased hypoxia-inducible factor and down-regulated E-cadherin in renal cell carcinoma (47), indicating a relationship between these pathways. However, we could not find any evidence for such a relationship, as none of the cadherins were linked to vascular related factors, such as microvessel density or vascular proliferation.

As a novel finding, our present data strongly suggest the importance of EMT (increased N-cadherin and decreased E-cadherin expression) for the progression and patient prognosis of human prostate cancer. Whereas previous findings have indicated reduced E-cadherin expression in prostate cancer subgroups (9, 10), an EMT phenotype and its association with outcome data have not been presented. Because this marker could have significant effect on the care of prostate cancer patients, including prospects of targeted therapy (48, 49), we suggest larger prospective studies to further validate our findings.

We thank Karen Bøhm-Nilsen, Gerd Lillian Hallseth, Bendik Nordanger, and Grethe Waaler for excellent technical assistance.