MATERIALS AND METHODS

Tissue samples

Our study was based on tissue specimens from untreated patients with Wilms’s tumour obtained from the division of pathology at The Hospital for Sick Children, Toronto, Canada. A series comprising 37 Wilms’s tumours was accrued for our study. Table 1 shows the clinicopathological parameters (tumour stage, histopathology, metastases, and percentage of differentiated elements) of group A, which comprised nine paired cases of primary tumour and the corresponding metastasis; group B, which comprised 10 primary (non-metastatic) tumours; and group C, which comprised eight metastatic tumours. The cases from group C were from patients who initially presented with metastases and therefore no primary was available. All tumours were assessed by staff pathologists and histological stage classifications were determined according to the National Wilms’s study tumour group.

Western blot analysis

Protein was extracted from snap frozen tissue with NP40 lysis buffer (150mM NaCl, 1% NP-40 (vol/vol), 0.5% sodium deoxycholate (wt/vol), 50mM Hepes, pH 7.5, and complete protease tablet (Roche-Boehringer Mannheim, Quebec, Canada), and western blot analysis was performed as described previously.²⁶ Aliquots (50 μg) of tumour protein extracts were analysed for six tumours (WiT 146, WiT 156, WiT 141, WiT 154, WiT 158, and WiT 157). Protein was quantified spectrofluorometrically using the Bradford assay (Bio-Rad, Hercules, California, USA). Blots were incubated with primary antibodies, anti-E-cadherin (1/50 dilution; mouse monoclonal (4A2C7); Zymed, San Francisco, California, USA), anti-β catenin (1/100 dilution; mouse monoclonal (C19220); Transduction Laboratories, Mississauga, Ontario, Canada) for one hour at room temperature, followed by several washes at room temperature in 1% skimmed milk in Tris buffered saline and 0.1% Tween 20. The secondary antibody, goat antimouse IgG alkaline phosphatase Fab fragments (1/1000 dilution; Roche-Boehringer Mannheim) was applied for one hour at room temperature. The membranes were incubated in CDPStar (Roche-Boehringer Mannheim) and exposed to Kodak X-omatic film for five minutes to one hour.

Immunohistochemistry

Immunohistochemistry was performed on paraffin wax embedded tumours after antigen retrieval, as described previously.¹⁸ The primary antibodies against E-cadherin (1/50 dilution; mouse monoclonal (4A2C7); Zymed), β catenin (1/100 dilution; mouse monoclonal (C19220); Transduction Laboratories), γ catenin (1/100 dilution; goat polyclonal (C-20); Santa Cruz), and ezrin (1/50 dilution; goat polyclonal (C-15); Santa Cruz) were incubated overnight at 4°C in antibody diluting buffer (Dimensions Laboratories, Mississauga, Canada). The secondary antibodies used were biotinylated goat antimouse IgG (1/200 dilution; Molecular Probes, Eugene, Oregon, USA) and biotin SP conjugated Affinipure donkey antigoat IgG (1/200 dilution; Jackson Laboratories, West Grove, Pennsylvania, USA), and were applied for one hour at room temperature. Finally, avidin horseradish peroxidase conjugate (1/1000 dilution; Neutravidin; Molecular Probes) was applied for one hour at room temperature. The bound antibody complex was visualised by reaction in 3,3‘diaminobenzidine substrate (Stable DAB; Research Genetics, Huntsville, Alabama, USA), and sections were counterstained with Meyer’s haematoxylin.

Evaluation of immunohistochemical staining

The staining intensity was scored using a semiquantitative scale: −, negligible; +, weak; ++, moderate; +++, strong; and ++++, very strong. The staining was evaluated by two independent observers without knowledge of the clinical outcomes. Tumour samples that revealed membranous (that is, intercellular staining at the cell membrane) patterns for E-cadherin, β catenin, γ catenin, and ezrin were designated as expected (that is, normal). Expression were considered to be abnormal when the proteins displayed positive cytoplasmic (with the exception of β catenin) or nuclear staining. The staining intensities of E-cadherin and β catenin in tumour cells were evaluated using the adjacent normal kidney epithelium as an internal positive control, thus allowing direct comparison within the same tissue specimen. In normal kidney epithelium, E-cadherin and β catenin proteins were strongly expressed at the intercellular borders of the proximal tubules and cell–cell boundaries of the collecting ducts, respectively (fig 1 C,F). Distal tubules showed weak expression of E-cadherin and β catenin. There was no staining at the lumenal surface or in the basal aspects of the cell membrane of tubules. γ Catenin was expressed at the cell membrane in collecting ducts and distal tubules, and ezrin in the distal tubules and glomerular epithelial cells of normal kidney (fig 1 I,L). No nuclear immunoreactivity was found for β catenin, E-cadherin, γ catenin, or ezrin in normal kidney. A negative control slide was prepared by omission of the primary antibody.

Statistical analysis

A nine point ordinal scale was used to analyse the data, with 1 representing the lowest staining category (− or negative) and 9 the highest (++++ or very strong). All other possible combinations in between were scored to represent the ordered nature of the staining categories. Tumours were nested into three groups: primary, metastatic, and primary (non-metastatic). Four measurements representing the crossed effects between proteins (E-cadherin and β catenin) and cell types (blastemal and epithelial) were taken for each tumour. Because these measurements were taken in the same tumour some correlation was assumed, so that analysis of variance for a correlated cross nested design was used to investigate the main effects of protein, cell type, and group, together with their two way and three way interactions. The mixed model theory was used to analyse the data. The PROC MIXED statistical program (SAS, Cary, North Carolina, USA) was used to perform the ANOVA. Table 2 shows the significance of the main effects and two and three way interactions. Individual contrasts were calculated to assess a priori differences of specific combinations of the three main effects (table 3). A p value of 0.05 or less was considered significantly different.

RESULTS

Expression of E-cadherin in primary and metastatic Wilms’s tumours

E-cadherin expression in primary and metastatic tumours was evaluated using immunohistochemistry (summarised in table 4). The expression of E-cadherin was relatively low in the cytoplasm of blastemal cells in seven of 19 of the primary tumours and nine of 16 of the metastatic tumours. Low to moderate E-cadherin expression was seen predominantly in the cytoplasm of epithelial cells and at the apical (lumenal end) of dysplastic tubular cells in 10 of 19 of the primary tumours and 11 of 16 of the metastatic tumours (fig 1 A,B). E-cadherin was not expressed in five metastatic tumours and three primary tumours that were representative of more undifferentiated phenotypes, comprising 95% and 70–85% blastemal cells, respectively. One primary tumour with features of 30% epithelial cells, 60% blastemal cells, and diffuse anaplasia was also negative for E-cadherin.

Most of the E-cadherin staining was found at the cell membrane in primary tumours not undergoing metastasis (group B), and in the cytoplasm of cells in metastatic tumours (groups A and C). Only two metastatic tumours showed nuclear staining (not shown). Although E-cadherin staining patterns varied within the tumours, probably because of variations in cellular differentiation, expression did correlate with increasing histological differentiation.

Expression of β catenin in primary and metastatic Wilms’s tumours

A high degree of expression of the β catenin protein was seen in blastemal and epithelial cells of primary tumours, with only two cases expressing low to moderate amounts. β Catenin was immunolocalised to both the cytoplasm and membrane of blastemal and epithelial cells (fig 1 D,E). Moderate to high β catenin expression was seen in all metastatic tumours, except for one case, which showed weak staining. In the metastatic cases, β catenin was immunolocalised to the cytoplasm and membrane, with seven of 16 cases showing nuclear staining in both blastemal and epithelial cells. In cases of metastatic tumour with distant metastases, an increase in staining intensity was seen in both cell types near the tumour invasive front (not shown).

One particular case (WiT 100), a recurrent metastatic tumour, demonstrated strong expression of both E-cadherin and β catenin in the nuclei of both blastemal and epithelial cells, with no apparent membranous or cytoplasmic staining. The liver metastasis (WiT 108) of WiT 100 showed an absence of E-cadherin and β catenin expression in both cell types. In all tumour cases examined, the stroma was consistently negative for E-cadherin and β catenin in both primary and metastatic Wilms’s tumours.

Both the degree of expression of E-cadherin and β catenin and the pattern of expression showed no apparent association with tumour stage. There were no apparent differences between E-cadherin and β catenin expression between primary tumours from group A and primary tumours from group B (table 4). However, the expression of β catenin was higher than that of E-cadherin in all the tumour groups studied (p < 0.0001; table 2). There was a significant difference between the degree of staining for both the proteins (E-cadherin and β catenin) in the two cell types, with stronger staining in epithelial cells than in blastemal cells (p = 0.0700). A two way interaction between the protein and cell type was also tested (p = 0.0068; table 2). There was no significant interaction on three way analysis noted between group–protein–cell type (p = 0.0817; table 2), and no significant difference between the degree of expression of E-cadherin and β catenin in primary and metastatic tumours for the two cell types (table 3).

Western blot analysis of E-cadherin and β catenin expression in Wilms’s tumours

Classic Wilms’s tumours are designated as homotypic Wilms’s tumours, whereas tumours exhibiting muscle differentiation are designated as heterotypic Wilms’s tumours (table 1). Tumour tissues from six homotypic and heterotypic Wilms’s tumour cases were used to examine E-cadherin and β catenin expression by western blot analysis (fig 2 A,B). Poorly differentiated primary homotypic Wilms’s tumours and one case of metastatic Wilms’s tumour (WiT 157) showed no evidence of the 120 kDa E-cadherin polypeptide by western blot analysis. Follow up of the primary homotypic tumours (WiT 146 and WiT 156) showed progression to a metastatic phenotype. Differentiated primary heterotypic Wilms’s tumours expressed the 120 kDa E-cadherin polypeptide and did not progress to a metastatic state. In general, E-cadherin expression in the tumours was lower than that seen in the normal kidney control.

As measured by western blot analysis, high amounts of the 92 kDa β catenin polypeptide were found in both primary (homotypic and heterotypic) tumours and in a single case of metastatic Wilms’s tumour, with increased amounts being found in primary heterotypic tumours that did not metastasise. No association between the differentiation state of the tumours and the degree of β catenin staining was seen.

In normal kidney, similar amounts of β catenin and E-cadherin were detected by western blot. In Wilms’s tumours, the expression of the β catenin protein was higher than that for E-cadherin using western blot analysis, and this correlated with the immunohistochemical findings (table 5).

Expression of γ catenin and ezrin proteins in primary and metastatic Wilms’s tumours

The expression of γ catenin was high in the cytoplasm of blastemal and epithelial cells in primary tumours from group A (fig 1G). However, the expression of γ catenin was consistently low in both cell types in primary tumours from group B. In metastatic tumours, moderate expression of γ catenin was seen mainly in the epithelial cells of local metastatic tumours and low expression was seen in blastemal cells of distant metastatic tumours (fig 1H). Nuclear staining for γ catenin was seen in some of the tumours from group B, near to the invasive front of tumours undergoing distant metastasis (not shown). This nuclear staining pattern was similar to that noted in colorectal and lung carcinomas.²⁷

Immunostaining was strong and homogeneous for ezrin in blastemal and epithelial cells of primary tumours from group B. Areas of differentiated blastema appearing as pseudoglomeruli showed positive staining for ezrin (fig 1J). In general, the degree of staining was moderate and low in local and distant metastases, respectively (fig 1K). Occasional regions of blastema showed a positive focus of cytoplasmic staining surrounded by negatively stained blastema. In dysplastic tubules situated within regions of blastema, ezrin staining was localised to the cytoplasm and at the lumenal portion of tubular cells residing within areas of stroma. Metastatic tumours showed an increase in staining at the tumour invasive front (not shown). Similar patterns of γ catenin and ezrin expression were found in the primary and metastatic tumours from group A (table 4). Stromal cells were negative for both γ catenin and ezrin.

DISCUSSION

Tumour cell invasion is a multistep process requiring complex alterations in adhesive interactions. One of the molecules implicated in the progress of tumour invasion is E-cadherin, a transmembrane glycoprotein that mediates calcium dependent cell adhesion between epithelial cells.^28,29 Through its extracellular calcium binding domains, E-cadherin forms a molecular zipper with other E-cadherin proteins that are located in the adherens junctions between adjacent cells. The intracellular domain of E-cadherin interacts with other cytosolic proteins such as catenins, ezrin, and actin filaments.¹⁵ Specifically, β catenin and γ catenin are membrane undercoat proteins that bind directly to E-cadherin and link the cytoplasmic terminal tail of E-cadherin to the actin cytoskeleton mediating the function of E-cadherin.³⁰ Met activation by HGF has also been shown to modulate the phosphorylation of catenins, leading to changes in the cellular distribution of adhesion molecules.³¹ These findings demonstrate the importance of the association between HGF/met, the cadherin/catenin complex, and ezrin proteins in the acquisition of a tumour invasive phenotype.

In our present study, E-cadherin showed a loss of membrane-type expression, with a moderate to low degree of staining in the cytoplasm of blastemal and epithelial cells in primary and metastatic tumours. Recently published studies on E-cadherin in Wilms’s tumour support our observations.¹² The loss of expression of E-cadherin in cell membranes, often with relocation to the cytoplasm, has been observed in many malignancies, and is an indicator of abnormal cadherin function.^17,27 Cellular redistribution of E-cadherin from the membrane to the cytoplasm and suppression of its function in vitro have been associated with the ability of HGF to induce tyrosine phosphorylation of β catenin and γ catenin.³¹

“E-cadherin showed a loss of membrane-type expression, with a moderate to low degree of staining in the cytoplasm of blastemal and epithelial cells in primary and metastatic tumours”

The 120 kDa E-cadherin polypeptide was expressed in differentiated heterotypic Wilms’s tumours that did not undergo metastasis, but was absent in poorly differentiated homotypic Wilms’s tumours that progressed to a more metastatic phenotype. Similar findings correlating the loss of E-cadherin expression and either a high metastatic probability or advanced stages of tumour were identified in renal cell carcinomas and in a few cases of Wilms’s tumour by immunohistochemistry only and not by western blot analysis.^12,32 Our study showed an overall decrease in the expression of E-cadherin, which was inversely proportional to the degree of tumour differentiation. Reduced expression of E-cadherin has been shown to induce dedifferentiation and invasiveness in tumour cells.³¹ The mechanisms underlying the reduced or absent expression of E-cadherin in tumours are not fully understood; however, CpG methylation in the E-cadherin promoter region has been proposed as a common mechanism of its inactivation in human tumours.³³ Mutations in E-cadherin have also been found in a variety of carcinomas such as gastric, breast, endometrium, and ovary.¹²

In Wilms’s tumour, we found high amounts of β catenin protein in the cytoplasm and membrane of tumour cells of primary and metastatic tumours, with some metastatic cases demonstrating nuclear expression. Moderate to high amounts of the 92 kDa β catenin polypeptide were found in primary (homotypic and heterotypic) tumours and metastatic tumours relative to normal kidney. Normally, β catenin is localised to the cell membrane and is absent or present in very low amounts within the cytoplasm. Mutations within APC, GSK3B, or axin genes can cause β catenin to accumulate within the cytoplasm, predisposing towards its transfer to the nucleus.³⁴ Nuclear staining for β catenin not only correlates with the possible transcriptional regulation of genes involved in proliferation and invasion, but also with mutations in the β catenin gene as seen in hepatocellular carcinomas, Wilms’s tumours, and prostate carcinomas.^11,17 Our present studies on E-cadherin and β catenin expression patterns confirm recently reported findings in Wilms’s tumour.^12,35

We have recently reported high levels of HGF and met expression in Wilms’s tumour.¹⁸ We therefore postulate that an extracellular signal such as HGF may alter the phosphorylation status of the catenins and release them from the cadherin–catenin complex. This in turn would contribute to tumour metastases not only by disrupting the E-cadherin linkage to the actin cytoskeleton, but also by releasing β catenin into the cytosol, where it could function as a signal transducer to promote cell growth and migration.

In general, the expression of γ catenin was high in the cytoplasm of blastemal and epithelial cells in primary tumours with metastasis and lower in primary tumours not undergoing metastasis. Metastatic tumours showed lower expression of γ catenin expression, with some tumours demonstrating nuclear expression. Higher expression of γ catenin was noted at the tumour invasive front, similar to β catenin, although the degree of expression of γ catenin differed from that of β catenin. It has been suggested that γ catenin may play a different role in adhesion and metastasis and may be regulated by a different mechanism to that controlling β catenin.³⁶

Our studies showed very high expression of ezrin in the cytoplasm of dysplastic tubules and pseudoglomeruli in primary non-metastatic tumours compared with primary tumours undergoing metastasis, and moderate expression in metastatic tumours. In general, ezrin showed a similar degree and pattern of expression to that seen for γ catenin. Occasional foci of blastema showed strong cytoplasmic staining for ezrin, suggesting possible pseudoglomeruli differentiation. In cells with polarised structures, ezrin may function as a membrane–cytoskeletal linker that regulates cell surface morphology and thus is expressed along the cell membrane.¹⁶ However, in cells responding to extracellular stimuli, such as HGF, ezrin may facilitate migration and morphogenetic changes because of its involvement in cytoskeletal reorganisation, thereby changing its localisation to the cytoplasm.²⁵

In summary, metastatic Wilms’s tumours showed lower expression of E-cadherin and γ catenin, with nuclear staining for β catenin, when compared with primary tumours. Low expression of E-cadherin was also associated with a poorly differentiated tumour phenotype. We have obtained preliminary evidence of increased expression of β catenin, γ catenin, and ezrin near the tumour invasive front in metastatic Wilms’s tumours. Several studies in various carcinomas have shown a relation between E-cadherin/catenin expression and clinicopathological variables, implicating these proteins as potential prognostic indicators of tumour progression.^37–39 Further studies would be needed to establish a similar association in Wilms’s tumour. However, we postulate that the changes seen in the expression and subcellular distribution of E-cadherin, β catenin, γ catenin, and ezrin in Wilms’s tumour may reflect the increased expression and potential activity of HGF/met in these tumours. Thus, these proteins may play an important role in the progression of Wilms’s tumour towards the acquisition of an invasive phenotype and may contribute to the development of metastases.