Misregulated E-cadherin expression associated with an aggressive brain tumor phenotype

Abstract

Background: Cadherins are essential components of the adherens junction complexes that mediate cell-cell adhesion and regulate cell motility. During tissue morphogenesis, changes in cadherin expression (known as cadherin switching) are a common mechanism for altering cell fate. Cadherin switching is also common during epithelial tumor progression, where it is thought to promote tumor invasion and metastasis. E-cadherin is the predominant cadherin expressed in epithelial tissues, but its expression is very limited in normal brain.

Methodology/principal findings: We identified E-cadherin expression in a retrospective series of glioblastomas exhibiting epithelial or pseudoepithelial differentiation. Unlike in epithelial tissues, E-cadherin expression in gliomas correlated with an unfavorable clinical outcome. Western blotting of two panels of human GBM cell lines propagated either as xenografts in nude mice or grown under conventional cell culture conditions confirmed that E-cadherin expression is rare. However, a small number of xenograft lines did express E-cadherin, its expression correlating with increased invasiveness when the cells were implanted orthotopically in mouse brain. In the conventionally cultured SF767 glioma cell line, E-cadherin expression was localized throughout the plasma membrane rather than being restricted to areas of cell-cell contact. ShRNA knockdown of E-cadherin in these cells resulted in decreased proliferation and migration in vitro.

Conclusions/significance: Our data shows an unexpected correlation between the abnormal expression of E-cadherin in a subset of GBM tumor cells and the growth and migration of this aggressive brain tumor subtype.

Figure 1. E-cadherin expression correlates with worse outcome for patients with glioblastomas exhibiting epithelial appearance.

A. E-cadherin protein in 8 of the 9 positive cases was focally restricted to discrete nests of tumor cells, reflecting areas of epithelial-like differentiation. In these positive cells E-cadherin was found primarily on the plasma membrane. However, unlike in normal epithelial cells, the localization of E-cadherin was not concentrated at areas of homotypic cell-cell contact, but rather localized uniformly along the plasma membrane (indicated by arrows in Ai). Left: Immunostain for E-cadherin (magnified in Ai to show detail); Right: H&E. The scale bar is 20 µm and applies to all Figure 1 images except Ai and Bi. B. 1 of the 9 E-cadherin positive cases showed an unusually high expression of E-cadherin which was located both on the membrane and cytoplasmically. Left: Immunostain for E-cadherin (magnified in Bi to show detail); Right: H&E of the tumor's epithelial component. C. Two independent examples of β-catenin localization by immunostain. Unlike E-cadherin, β-catenin is distributed throughout the tumor samples. D. Individual cases of glioblastoma exhibiting epithelial or pseudoepithelial differentiation were analyzed by immunohistochemistry for the absence (Negative) or presence (Membraneous/cytoplasmic) of E-cadherin protein. This data was correlated with overall survival using Kaplan-Meier analysis. There is a statistically significant survival difference (p = 0.021).

Figure 2. E-cadherin protein expression occurs in a subset of xenograft glioma lines.

Nineteen glioma cell lines propagated as xenografts in mouse flank were examined by Western Blot for expression of various cadherin and catenin proteins. Actin serves as a loading control. Positive control lysates are from MCF7 cells (E-cadherin, β-catenin), UMRC3 cells (N-cadherin), and MDA231 cells (cadherin-11). Low levels of E-cadherin expression were detected in a small subset of these glioma xenografts.

Figure 3. E-cadherin protein expression correlates with glioma cell invasiveness in an orthotopic xenograft mouse model.

Xenograft cell lines 5, 6, 8, 10, 12, 14, 15, 16, 22, 26, 28, 34, 36, 38, 43, 44, 46, and 59 were examined for relative invasiveness following orthotopic injection into mouse brain. Lines were categorized as highly, moderately, minimally, or non invasive. The level of E-cadherin expression (relative to actin expression) determined by Western blot for each cell line was then plotted vs. relative glioma invasiveness. The lines through the data indicate the median for each invasiveness category; *indicates a statistical difference (one-tailed, unpaired t test) between the two categories of invasiveness at p<0.05.

Figure 4. E-cadherin expression in conventional glioma cell lines is rare.

Nineteen conventionally grown glioma cell lines were examined by Western Blot for expression of various cadherin and catenin proteins. α-tubulin serves as a loading control. Positive control lysates are as in Figure 2. SF767 was the only conventional glioma cell line examined that expressed E-cadherin.

Figure 5. SF767 cells lack junctional organization of actin and have disorganized adhesive structures containing E-cadherin.

A. Immunofluorescence for E-cadherin expression and actin localization was carried out on paraformaldehyde-fixed/Triton X-100-permeabilized SF767 and MCF7 cells (as a control). Arrows indicate areas of cell-cell contact; arrowheads point to areas of the plasma membrane without cell-cell contact. MCF7 cells form compact cell-cell adhesions to which the actin cytoskeleton and E-cadherin tightly localize. In contrast, E-cadherin localization in SF767 is at cell-cell contacts and on the plasma membrane at areas without cell-cell contact. Additionally, the actin cytoskeleton is not properly organized at areas of cell-cell contact. 63X magnification. The scale bar is 10 µm and applies to all images in Figure 5. B. Immunofluorescence for E-cadherin and β-catenin expression was carried out on MeOH-fixed/permeabilized SF767 and MCF7 cells. Arrows and arrowheads are as in A. β-catenin and E-cadherin both localize tightly to adherens junctions between MCF7 cells. In contrast, fewer proper cell-cell junctions exist in the SF767 cells. Both E-cadherin and β-catenin localization is diffusely distributed on the plasma membrane, in addition to a disorganized presence at areas of cell contact. 63X magnification.

Figure 6. E-cadherin depletion inhibits SF767 growth and migration.

A. SF767 cells were infected with lentivirus expressing either non-target shRNA or E-cadherin shRNA on day zero, and then harvested on days 2, 4, 7, and 10 post-infection (p.i.). RNA from the cells was used for quantitative RT-PCR analysis of CDH1 mRNA, normalized to the expression of GAPDH mRNA; data is plotted as fold change vs. the normalized CDH1 levels in wild-type SF767 cells. The shEcad#20 cell line was discarded after 7 days p.i. due to lack of difference with the pLKO-NT cells in terms of growth rate, morphology, and CDH1 mRNA expression. The shEcad#21 cell line grew poorly and was completely harvested by 7 days p.i. B. Additional cells were harvested on days 4, 7, and 10 p.i. for protein analysis. Western blot was used to determine knockdown of E-cadherin; α-tubulin is a loading control. The generally poor condition of SF767-shEcad#21 cells by day 7 p.i. is reflected in the reduced levels of α-tubulin seen on the blot, despite loading equal µg of total protein from each sample. C. Growth rates of E-cadherin-depleted SF767-shEcad#23 vs. control SF767-non target cells were determined over 5 days using an MTT assay. E-cadherin-depleted SF767 cells grew more slowly than control SF767 cells. This difference is statistically significant at 72, 96, and 120 hours in culture (2-way ANOVA followed by Bonferroni post hoc tests; * indicates p<0.05). These two cell lines were generated independently of those displayed in parts A and B. D. Migration of E-cadherin-depleted SF767-shEcad#23 vs. control SF767-non target cells (the same cell lines used for the growth experiment in C) was determined using a Boyden chamber trans-well migration assay. Fewer E-cadherin-depleted SF767 cells than control SF767 cells migrated toward the chemoattractant. The difference is statistically significant (paired t-test; * indicates p<0.05).