Importance of P-Cadherin, β-Catenin, and Wnt5a/Frizzled for Progression of Melanocytic Tumors and Prognosis in Cutaneous Melanoma

Adhesion molecules are important for the development and progression of human cancer (1, 2). Among these, the cadherins represent a family of transmembrane proteins promoting calcium-dependent cell-cell adhesion in a tissue-specific manner depending on which isotype is present. E-cadherin is the major subtype in human epithelial cells and is also considered the main adhesion mediator between melanocytes and epidermal cells in the normal skin (1). P-cadherin is strongly expressed by the basal epidermal cells, indicating an association with proliferating cells (3–5). During embryonal development in mice, altered expression of E-cadherin and P-cadherin in the melanoblasts leads to differential interactions with surrounding cells, with a peak expression of E-cadherin just before the entry of melanoblasts into the epidermis (4). In the skin, N-cadherin is expressed on fibroblasts, endothelial cells, and neurons and is proposed to be responsible for the homotypic anchorage of these cells (6, 7).

Melanoma cells are suggested to shift from E-cadherin to N-cadherin expression during tumor development, and expression of N-cadherin might facilitate tumor cell association with stromal fibroblasts and endothelial cells leading to increased invasive capacity (8). Although in vitro data support this model, it has not been confirmed in studies of human melanocytic tumors at different stages (9, 10). Whereas cultured human melanocytes are shown to express both E-cadherin and P-cadherin, expression was found to be reduced in eight (E-cadherin) and four (P-cadherin) of nine investigated melanoma cell lines using flow cytometric detection (11). A few small studies have reported the expression of P-cadherin in melanocytic nevi and melanomas (9, 10), but clinicopathologic associations or patient survival have not been explored previously.

The cytoplasmic domain of E-cadherin links to the cytoskeleton through interactions with β-catenin. In the cytoplasm, β-catenin is associated with the tumor suppressor protein APC, and degradation of β-catenin occurs by glycogen synthase kinase-3β–dependent phosphorylation. Further, nuclear β-catenin functions as a transcriptional activator through complex formation with members of the Tcf/Lef family (12, 13). In addition to playing a critical role in cell adhesion, β-catenin also functions in the Wnt signaling pathway (13). Wnt proteins are secreted glycoproteins that interact with members of the Frizzled receptors, and they are considered to act as oncoproteins by disruption of β-catenin degradation (12) leading to cytoplasmic and nuclear accumulation of this protein. Mutations in the β-catenin gene (CTNNB1) have been reported in some melanoma cell lines (14) but seem to be rare in vivo despite frequent nuclear or cytoplasmic expression of β-catenin by immunohistochemistry (15). In addition, there are no studies to support that nuclear localization of β-catenin indicates an alteration of the Wnt pathway or tumor progression in human melanoma. In fact, down-regulation or loss of β-catenin expression has been reported in subgroups of more advanced melanomas (16, 17).

Gene expression profiling of cutaneous melanoma (cDNA microarray) has shown an increased expression of Wnt5a/Frizzled in subsets of tumors, and it has been speculated that increased Wnt5a might indicate a more aggressive melanoma behavior (18, 19). Recently, it was reported that Wnt exerts its function through at least three different pathways: the canonical pathway involving nuclear accumulation and stabilization of β-catenin, the noncanonical pathway mediating cytoskeletal changes through activation of Rho and Rac, and the Wnt-Ca²⁺ pathway involving a set of cellular mediators, including protein kinase C, phospholipase C, and calcium-calmodulin kinase 2. Wnt5a is considered to act mainly through the noncanonical pathway. Although pilot studies of melanoma cell lines and a few melanoma specimens have indicated that increased expression of Wnt and its receptor Frizzled correlates with histologic features of invasiveness (20, 21), no larger validation studies of these findings concerning clinical outcome and prognosis have been done.

On this background, the aim of our study was to investigate the role of cadherin subtypes, β-catenin, and Wnt5a/Frizzled in the progression of melanocytic tumors and as markers of aggressive melanomas. We also wanted to ascertain a possible relationship between adhesion markers and tumor-associated angiogenesis or lymphangiogenesis, which have been focused in our previous studies (22–24).

Of all melanomas occurring in Hordaland County, western Norway (∼450,000 inhabitants or 10% of the Norwegian population) during 1981 to 1997, 97.5% were diagnosed at The Gade Institute, Section for Pathology, Haukeland University Hospital. There were no significant differences in sex, anatomic site, or stage between these cases and the 2.5% with a diagnosis from other laboratories, although the latter patients were 6 years younger (median age). All cases diagnosed and recorded as malignant melanoma of the nodular type or not otherwise specified were reviewed previously and 202 cases were included (median age, 64.4 years; median thickness, 3.6 mm). The presence of a vertical growth phase and the lack of a radial growth phase (i.e., adjacent in situ or microinvasive component) were used as inclusion criteria. There was no history of familial occurrence. Clinicopathologic characteristics and some survival data have been reported previously, and information on Ki-67 expression, angiogenesis, and lymphangiogenesis was included for comparison (23–28).

In this study, cases with sufficient material left in the tissue microarray blocks (range, 122-133 cases; differences due to dropouts) were examined for the expression of E-cadherin, N-cadherin, P-cadherin, β-catenin, Wnt5a, and Frizzled. In addition, 58 paired metastases (local skin, regional lymph nodes, distant) were examined.

In addition to the series of nodular melanoma, where our purpose was to examine the expression pattern and prognostic effect in this aggressive melanoma type, 32 cases of benign melanocytic nevi (median age, 26.6 years) and 20 cases of invasive superficial spreading melanomas (median age, 49.0 years; median thickness, 1.7 mm) were included to look at associations between these markers and different stages of melanocytic tumor progression. The study was done in accordance to the Helsinki Declaration.

The following variables were recorded: date of histologic diagnosis, sex, age at diagnosis, anatomic site of the primary tumor, and presence of metastases at diagnosis (local, regional, distant). The H&E-stained slides were reexamined previously, and the histologic features included tumor thickness according to Breslow (29), level of invasion according to Clark et al. (30), microscopic ulceration, and vascular invasion (27).

The technique of tissue microarray was introduced previously (31) and validated by independent studies of several tumor markers (32, 33). For tissue microarray construction (31, 33), representative tumor areas were identified on H&E-stained slides generally at the suprabasal areas of the primary tumors. Tissue cylinders with a diameter of 0.6 mm were then punched from selected areas of the donor block and mounted into a recipient paraffin block using a custom-made precision instrument (Beecher Instruments, Silver Spring, MD). Sections of the resulting tissue microarray blocks (5 μm) were then made by standard technique. As recommended (32), three parallel tissue cylinders were sampled from each case. For internal validation, tissue microarray sections from 50 randomly selected cases were stained for Ki-67 as described previously (25), and the labeling index (percent positive tumor cell nuclei) was determined. A highly significant correlation between results from tissue microarray sections and standard slides was present (26).

The immunohistochemical staining was done on thin sections (5 μm) of paraffin-embedded archival tissue using tissue microarray sections as described. Samples were dewaxed with xylene/ethanol before microwave antigen retrieval and antibody incubation using protocols optimized for each antibody used.

E-cadherin. Staining for E-cadherin was done on tissue microarray slides using antigen retrieval as described (10 minutes at 750 W and 15 minutes at 350 W) and boiling in Tris-EDTA buffer (pH 9). The slides were incubated for 1 hour at room temperature with the monoclonal mouse E-cadherin antibody M3612 (DAKOCytomation, Copenhagen, Denmark) diluted 1:100.

N-cadherin. After microwave retrieval in Tris-EDTA buffer (pH 9) and boiling for 10 minutes at 750 W and 20 minutes at 350 W, tissue microarray sections were incubated for 1 hour at room temperature with the monoclonal mouse antibody M3613 (DAKOCytomation) diluted 1:50.

P-cadherin. The P-cadherin protocol included microwave antigen retrieval (10 minutes at 750 W and 20 minutes at 350 W) in Tris-EDTA buffer (pH 9) and incubation for 1 hour at room temperature with the P-cadherin monoclonal mouse antibody C24120 (BD Transduction Laboratories, San Diego, CA) diluted 1:200.

β-Catenin. After microwave retrieval boiling in Tris-EDTA pH 9 (10 minutes at 750 W and 15 minutes at 350 W), tissue microarray slides were incubated for 1 hour at room temperature with the monoclonal β-catenin antibody (BD Transduction Laboratories) diluted 1:800.

Wnt5a. The staining protocol for Wnt5a included microwave antigen retrieval in citrate buffer (pH 6) for 10 minutes at 750 W and 15 minutes at 350 W. The tissue microarray sections were then incubated for 1 hour at room temperature with the polyclonal goat antibody (AF 645; R&D Systems, Minneapolis, MN) diluted 1:100 followed by incubation in 1 hour with a secondary rabbit anti-goat antibody (Z0454; DAKOCytomation) before using the EnVision polymer K4002 (DAKOCytomation).

Frizzled. After microwave retrieval (10 minutes at 750 W and 30 minutes at 350 W) in citrate buffer (pH 6), the sections were incubated for 1 hour at room temperature using the rabbit polyclonal antibody sc-9169 (Santa Cruz Biotechnology, Santa Cruz, CA) diluted 1:50.

The staining procedures were all done using the EnVision-labeled polymer method, with commercial available kits (DAKOCytomation), with 3-amino-9-ethylcarbazole peroxidase as substrate before brief counterstaining with Mayer's hematoxylin. For monoclonal antibodies of mouse origin, negative controls were obtained using isotypic mouse immunoglobulin (IgG1) in the same dilution as the primary antibody of concern, whereas rabbit IgG was used in the same way as negative control for the Frizzled antibody and nonimmune goat serum for the Wnt5a antibody. All control experiments were negative.

The human melanoma cell line WM35 (WM35: CRL-2807) were grown in a 2% tumor medium containing a 4:1 mixture of MCDB 153 medium (American Type Culture Collection, Teddington, Middlesex, United Kingdom) with 1.5 g/L sodium bicarbonate and Leibowitz's L-15 medium with 2 mmol/L l-glutamine (American Type Culture Collection) supplemented with 0.005 mg/mL bovine insulin, 1.68 mmol/L CaCl₂, and 2% fetal bovine serum. The total cell lysate was made by adding 150 mmol/L NaCl, 50 mmol/L Tris-HCl (pH 7.4), 1% Triton X-100, 0.1% SDS, 1% sodium deoxycholate, and 1 μL/mL leupeptin to the cells. The cells were passed several times through a 21-gauge needle, and phenylmethylsulfonyl fluoride (100 mmol/L) was added. After centrifugation at 10,000 × g for 10 minutes, the supernatant was collected. A mixture 1:1 of total cell lysate (the supernatant) and 2× sample buffer [0.25 mol/L Tris-HCl (pH 6.8), 4% SDS, 20% glycerol, 0.05% bromophenol blue] and 5% β-mercaptoethanol were boiled for 5 minutes and the proteins were separated in a 10% acrylamide SDS-PAGE gel using 15 μL/lane and transferred to a nitrocellulose membrane (Bio-Rad, Cambridge, MA; Hybond-P polyvinylidene difluoride, Amersham, Piscataway, NJ) at 100 V in Tris-glycine buffer. The membranes were then blocked with 5% dry milk with 0.05% Tween 20 and incubated with the above-mentioned antibodies against E-cadherin (diluted 1:100), N-cadherin (1:100), β-catenin (1:500), Wnt5a (1:1,000), and Frizzled (1:100), respectively. Labeled polymer-horseradish peroxidase anti-mouse (Envision, DAKOCytomation) for monoclonal antibodies and polymer-horseradish peroxidase anti-rabbit for polyclonal antibodies were added in a 1:500 dilution. Concerning Wnt5a, this step was preceded by incubation with a rabbit anti-goat antibody (Z0454) diluted 1:5,000. The bands were visualized by enhanced chemiluminescence detection. The P-cadherin antibody, diluted 1: 250, was incubated against a A431 lysate prepared from a human epidermoid carcinoma cell line (BD Transduction Laboratories) and added directly to a 10% acrylamide SDS-PAGE gel and processed as described above.

The results indicate that the antibodies are specific as they presented distinct signals corresponding to their expected molecular weights (Fig. 1).

E-cadherin and β-catenin staining was predominantly membranous in the melanomas; for β-catenin, cases with positive nuclei were also seen. Cytoplasmic staining dominated for N-cadherin and P-cadherin, but some degree of membranous staining was present in many cases. Wnt5a and Frizzled showed a nuclear staining pattern, but some cases with cytoplasmic staining were also found.

The immunohistochemical stainings were recorded using a semiquantitative and subjective grading, considering both the intensity of staining and the proportion of tumor cells showing unequivocal positive reaction. Intensity: 0, no staining; 1, weak staining; 2, positive staining; and 3, strong staining. Area: 0, no staining; 1, positive staining in <10% of tumor cells; 2, positive staining in 10% to 50% of tumor cells; 3, positive staining in >50% of tumor cells. Staining index: staining intensity times positive area (34, 35).

In subsequent statistical analyses, cutoff points for staining index categories were mainly based on median values, considering also the frequency distribution curve and size of subgroups for each marker. As a consequence, E-cadherin, nuclear Wnt5a, and cytoplasmic Frizzled staining indices were categorized by their median value as high (>4) or low (0-4), cytoplasmic N-cadherin and cytoplasmic β-catenin staining index as high (>3) or low (0-3), whereas nuclear β-catenin expression and membranous N-cadherin and P-cadherin were divided by absence (staining index = 0) or presence (staining index ≥ 1) of staining. For cytoplasmic P-cadherin expression, the cutoff point was set at staining index = 4 because few cases had higher values than the approximate median value of staining index = 6.

Hospital records were used to obtain follow-up information, including the presence and type of recurrences, as well as the time until recurrence. In some cases, general practitioners responsible for the follow-up of patients were contacted. Information about time and cause of death was obtained from the Cancer Registry of Norway and Statistics Norway. Complete information on patient survival and time and cause of death was available in all 202 cases. Last date of follow-up was December 31, 1999, and median follow-up time for survivors was 89 months (range, 24-221). Clinical follow-up (with respect to recurrences) was not carried out in 14 patients (predominantly older patients), and 21 patients were not treated with complete local excision. Thus, recurrence-free survival was available in 167 patients. During the follow-up period, 72 (36%) patients died of malignant melanoma and 45 (22%) died of other causes. Of the 167 radically treated patients with data on recurrence-free survival, 74 (44%) had recurrent disease.

Analyses were done using the SPSS statistical package version 12.0 (SPSS, Inc., Chicago, IL). Associations between different categorical variables were assessed by Pearson's χ² test. Continuous variables not following the normal distribution were compared between two or more groups using the Mann-Whitney U or Kruskal-Wallis H tests. Wilcoxon signed rank test was used to compare related samples. Univariate analyses of time to death due to malignant melanoma or time to recurrence (recurrence-free survival) were done using the product-limit procedure (Kaplan-Meier method), and differences between categories were estimated by the log-rank test, with date of histologic diagnosis as the starting point. Patients who died of other causes were censored at the date of death. The influence of covariates on patient survival and recurrence-free survival was analyzed by the proportional hazards method and tested by the likelihood ratio test.

Stainings are shown in Fig. 2. Expression of E-cadherin turned out to be stronger in melanomas (P = 0.030) when compared with benign nevi (Table 1). There was also a significantly increased expression of membranous β-catenin (P < 0.0001) in melanomas but reduced nuclear (P < 0.0001) and cytoplasmic (P = 0.024) β-catenin expression. Further, cytoplasmic expression of P-cadherin was increased in melanomas (P < 0.0001), whereas the opposite was found for membranous P-cadherin with loss of expression from nevi to melanomas (P < 0.0001). The same pattern was seen for N-cadherin, with increased cytoplasmic expression (P < 0.0001) and decreased membranous staining in melanomas when compared with nevi (P < 0.0001). The nuclear expression of the ligand Wnt5a and its receptor Frizzled were reduced in melanomas compared with benign nevi (P = 0.042 and P < 0.0001, respectively), whereas the opposite was found for cytoplasmic expression with increased levels of staining in melanomas (P = 0.013 and P < 0.0001).

Standard paraffin sections from five dermal nevi were pretreated and stained as described. For all antibodies, the staining in positive cases was uniform and showed no specific polarity or variation depending on the depth of the melanocytic nests. The staining was comparable with our findings in the tissue microarray series.

When comparing the subtypes of superficial spreading and nodular melanomas, the most noteworthy finding was that cytoplasmic expression of P-cadherin was stronger in the nodular melanomas (P = 0.005), whereas membranous expression showed the opposite pattern (P < 0.0001). Membranous N-cadherin staining was also found to be weaker in nodular melanomas when compared with the superficial spreading subtype (P = 0.005). Expression of E-cadherin was reduced from superficial to nodular melanomas (P = 0.012).

E-cadherin. Low E-cadherin expression (less than median value) was significantly associated with weak β-catenin membrane expression (P < 0.0001) but not with any of the other markers tested in this study or with any of the clinicopathologic features.

N-cadherin. Membranous and cytoplasmic staining for N-cadherin was recorded separately. Cytoplasmic expression was positively associated with levels of cytoplasmic Wnt5a expression (P = 0.007), P-cadherin (P = 0.034), and Frizzled (P = 0.015). Presence of membranous N-cadherin expression was associated with strong membranous β-catenin staining (P = 0.022) but not with any of the other markers or variables in this study.

P-cadherin. As for N-cadherin, membranous and cytoplasmic stainings were recorded separately. P-cadherin expression in the cytoplasm was inversely correlated to nuclear expression of Frizzled (P = 0.014). As mentioned above, strong cytoplasmic expression was associated with strong cytoplasmic N-cadherin staining (P = 0.034). Strong cytoplasmic P-cadherin expression was significantly associated with increased tumor thickness, with median thickness of 4.4 mm compared with a median of 3.0 mm when P-cadherin index was low (P = 0.005). Regarding membranous P-cadherin expression, the opposite was found, as negative staining was associated with increased tumor thickness (median, 4.4 mm) compared with cases with positive membranous P-cadherin (median thickness, 3.2 mm; P = 0.012). The same pattern was seen regarding Clark's level of invasion, as we found a positive association between cytoplasmic P-cadherin expression and Clark's level of invasion V (P = 0.019) but an inverse correlation between membranous P-cadherin expression and Clark's level of invasion (P = 0.018).

β-Catenin. β-Catenin staining was found in both nucleus and cytoplasm as well as in the membranes of tumor cells, and staining indices were recorded for each compartment separately. Reduced membranous expression was significantly associated with lower expression of E-cadherin (P < 0.0001) and N-cadherin (P = 0.022). The subgroup with lack of nuclear β-catenin expression was associated with increased tumor thickness with median 4.2 versus 2.9 mm when nuclear β-catenin expression was present (P = 0.002). Cytoplasmic β-catenin expression was found to be associated with thinner tumors according to Breslow (2.9 versus 4.2 mm; P = 0.021).

Wnt5a and Frizzled. Nuclear and cytoplasmic expression of Wnt5a and its receptor, Frizzled, were recorded separately. Nuclear staining of these markers was strongly associated with each other (P < 0.0001), and nuclear expression of Wnt5a correlated positively with nuclear expression of β-catenin (P = 0.031). Nuclear expression of Wnt5a was stronger in p53-positive cases (P = 0.040) and was associated with lower tumor thickness with median 3.3 versus 4.3 mm in cases with low nuclear Wnt5a staining index (P = 0.013) and with absence of vascular invasion (P = 0.027). Weak nuclear expression of Frizzled showed an association with Clark's level of invasion V (P = 0.050). Cytoplasmic Wnt5a expression correlated positively to high expression of cytoplasmic P-cadherin (P = 0.031) and N-cadherin (P = 0.007) but inversely to membranous P-cadherin expression (P = 0.017). Strong Frizzled cytoplasmic staining index correlated to strong cytoplasmic expression of both P-cadherin (P < 0.0001) and N-cadherin (P = 0.015).

There were no associations between cell adhesion markers and angiogenesis as estimated by microvessel counts. However, there was a significantly higher lymphatic vessel count in cases expressing low levels of E-cadherin (12.5 versus 6.5 vessels/mm²; P = 0.037). A borderline association between high lymphatic vessel count and high levels of nuclear Wnt5a expression (15.6 versus 6.2 vessels/mm²; P = 0.067) was also noted.

Except a nonsignificant trend toward an association between strong cytoplasmic P-cadherin expression and increased proliferation rate assessed by Ki-67 staining (P = 0.08), no association between other adhesion markers and tumor cell proliferation was found.

In the subgroup of nodular melanomas, the 10-year survival rate in our study was 57%, and 36% of all patients died of malignant melanoma during the follow-up period. High levels of cytoplasmic P-cadherin expression were significantly associated with decreased survival rate, with 53% estimated 5-year survival compared with 76% when P-cadherin expression was low (P = 0.047; Fig. 3A). We found a 55% estimated 5-year survival in cases negative for nuclear β-catenin compared with 76% estimated 5-year survival when β-catenin nuclear staining was present (P = 0.0072, log-rank test; Fig. 3B). β-Catenin did not predict recurrence-free survival, whereas 5-year recurrence-free survival was 30% in cases with high P-cadherin cytoplasmic expression compared with 67% in cases with low expression (P = 0.040). No association to patient outcome was seen regarding the other investigated markers E-cadherin, N-cadherin, Wnt5a, or Frizzled (Table 2).

In multivariate analysis of patient survival, loss of nuclear expression of β-catenin was an independent prognostic variable (hazard ratio, 2.8; P = 0.004, likelihood ratio test) when included along with known prognostic factors like tumor thickness, vascular invasion, Clark's level of invasion, and tumor cell proliferation assessed by Ki-67. The final multivariate model for β-catenin is given in Table 3.

Reduced expression from primary tumors to metastases was found as a nonsignificant trend considering membranous E-cadherin staining index (P = 0.067), whereas cytoplasmic P-cadherin tended to be increased in metastatic lesions (P = 0.082). No such trend was seen regarding membranous P-cadherin expression (P = 0.22). In addition, there was significantly higher expression of membranous N-cadherin in metastases when compared with the corresponding primary tumors (P = 0.009, Wilcoxon signed rank test). Regarding nuclear expression of Frizzled and Wnt5a, reduced staining was observed in metastases (P = 0.05 and 0.017, respectively). There were no significant associations between nuclear or membranous β-catenin expression in primary tumors and their corresponding metastases.

In this study, the expression pattern of P-cadherin was significantly altered from benign melanocytic nevi to primary and metastatic melanomas, and we found a significant association with patient survival, which has not been reported previously. These findings indicate that P-cadherin expression might represent a marker of progression and prognosis in melanocytic lesions. Thus, cytoplasmic expression of P-cadherin was significantly increased from benign nevi to malignant melanoma, whereas membranous staining was decreased, and cytoplasmic positivity tended to be further increased in metastatic lesions. Strong cytoplasmic P-cadherin expression in primary tumors was associated with markers of aggressive melanomas, such as increased vertical tumor thickness and level of invasion, and with decreased patient survival. Similarly, increased expression of P-cadherin has been reported in aggressive subgroups of breast (36, 37) and endometrial (38) cancers, further supporting its value as a prognostic marker in different tumors. In our melanoma series, the shift from membranous to cytoplasmic staining should be especially noted in nodular melanomas.

Previous experimental studies have indicated that a “switch” from E-cadherin to N-cadherin expression is taking place during melanoma development and progression (39, 40). In our present study, N-cadherin showed the same shift pattern as P-cadherin from membranous to cytoplasmic expression when comparing benign nevi with melanomas. Increased cytoplasmic expression of N-cadherin tended to be associated with decreased patient survival, although not significant. The finding that high levels of P-cadherin expression were associated with increased tumor thickness and reduced recurrence-free and overall survival may indicate that alterations in E-cadherin and N-cadherin are frequent during earlier stages of melanoma development and that P-cadherin might play an important role in more advanced melanomas of the vertical growth phase. Still, a switch from E-cadherin to N-cadherin expression was indicated when metastases were compared with primary tumors in line with experimental evidence (39, 40). Thus, our findings extend previous observations on the shift in cadherin profile during melanoma development to include alterations in P-cadherin expression, suggesting that a switch from membranous to increased cytoplasmic expression of P-cadherin might be a marker of more aggressive primary melanoma subgroups with reduced survival. The mechanism behind this apparent redistribution of P-cadherin is presently not known.

The dual role of β-catenin in mammalian cells is not completely understood, but nuclear accumulation is thought to activate transcription of target genes, some of which are involved in tumorigenesis, such as c-Myc (41), cyclin D1 (42), and matrix metalloproteinase-7 (43). In this series of vertical growth phase melanomas, loss of nuclear β-catenin expression was associated with more aggressive tumor behavior and significantly reduced patient survival. Consistent with this, benign nevi expressed nuclear β-catenin more often than did melanomas, whereas the opposite was found for membranous staining. Cytoplasmic β-catenin expression showed the same pattern as membranous staining and was found to be reduced in melanomas when compared with nevi. The reason for this shift in β-catenin expression from benign to malignant melanocytic lesions is not clear. It has been shown previously that 6 of 27 melanoma cell lines harbored β-catenin exon 3 mutations (14), rendering β-catenin resistant to degradation and resulting in cytoplasmic and nuclear accumulation. On the other hand, exon 3 mutations were found in only 1 of 50 melanoma specimens in spite of cytoplasmic and nuclear β-catenin expression in one third of the cases (15), indicating that other mechanisms are likely to be involved.

Regarding prognostic importance, an association between stronger nuclear β-catenin expression and favorable prognosis, similar to our present findings, has been reported in other tumors as well, such as ovarian (44) and hepatocellular (45) carcinomas. Correspondingly, absent or weak expression of β-catenin, especially the lack of nuclear staining, was associated with markers of disease progression and tended to predict recurrence of disease and poor prognosis in a series of 91 mainly acral melanomas (17). In another study of 106 superficial spreading and 58 nodular melanomas, loss of cytoplasmic expression of β-catenin correlated with more advanced lesions and reduced disease-free survival, but an influence on overall survival was not seen (16). In contrast to this, we found that loss of nuclear β-catenin showed a significant association with decreased patient survival in our melanoma series, and this has not been reported previously. The prognostic influence persisted in multivariate survival analyses, and β-catenin expression was thus a stronger prognostic marker than P-cadherin status in these tumors. The mechanism behind loss of nuclear β-catenin expression in advanced melanomas needs to be studied in further detail.

Gene expression profiling has indicated that Wnt5a might be a marker of aggressive melanomas (19), and initial studies done on human melanoma cell lines and tumor samples seem to validate these findings (20, 21), but no correlation to patient outcome has been shown to date. In other tumors, the role of Wnt5a is controversial, as it is found to be up-regulated in cancers of the lung, breast, and prostate (46) but down-regulated in pancreatic (47) and urothelial carcinomas, where Wnt5a is found to act as a tumor suppressor (48). This is in line with the findings that Wnt5a signals not only through the canonical pathway by stabilization of β-catenin but also through different noncanonical pathways, which actually promote degradation of β-catenin independent of glycogen synthase kinase-3 (49, 50). In our present study of benign and malignant melanocytic tumors, nuclear expression of both Wnt5a and its receptor Frizzled, which were highly coexpressed, was significantly reduced from benign nevi to melanomas, and this could be consistent with a tumor suppressor role. In established melanomas, Wnt5a expression was significantly associated with nuclear β-catenin expression, and this could support an inhibitory role of Wnt5a for β-catenin degradation as indicated from basic studies (51, 52). The finding of both cytoplasmic and nuclear expression of Wnt5a and Frizzled is interesting, because membrane-associated binding was to be expected. This could possibly be explained by the involvement of Wnt/Frizzled in the different pathways mentioned above, but future studies are needed to clarify if the nuclear expression implies other and distinct biological functions as is the case for other proteins (e.g., basic fibroblast growth factor; refs. 26, 53). In addition, we used a pan-Frizzled antibody, and the possibility that a more specific antibody to Frizzled 5 could have influenced our findings cannot be ruled out.

Further, nuclear Wnt5a expression was found to correlate with thinner melanomas and absence of vascular invasion, again indicating a protective role in tumor progression and suggesting the presence of complex regulatory interactions among these protein pathways. Any association with tumor recurrences or patient survival was, however, not found in this first prognostic study of Wnt5a/Frizzled expression in cutaneous melanoma.

In conclusion, our study indicates that alterations in the expression level and subcellular localization of various adhesion molecules are important in the development and progression of melanocytic lesions, and significant associations were found with prognosis in established melanomas. Strong P-cadherin expression in the cytoplasm and loss of nuclear β-catenin expression were predictive of reduced patient survival. In addition, loss of Wnt5a/Frizzled expression might be involved in the development of melanocytic tumors and should be further studied.

We thank Gerd Lillian Hallseth, Bendik Nordanger, and Randi Nygaard for excellent technical assistance.