Interplay between neural-cadherin and vascular endothelial-cadherin in breast cancer progression

Ep5 and Ep5ExTu cells were cultured as described [26] in Dulbecco's modified Eagle's medium (DMEM-F12; Lonza, Basel, Switzerland) supplemented with 15% fetal calf serum (FCS). 293T cells were kept in DMEM Glutamax (Gibco, Darmstadt, Germany) supplemented with 10% FCS.

Oligonucleotides (Eurogentec, Seraing, Belgium) encoding small interfering RNA (siRNA) molecules specific for mouse VE-cadherin (5'-GUCUCUGAGU ACUUCCUUA-3') or N-cadherin (5'-GGAUGUGCAG GAAGGACAG-3' and 5'-UGUCAAUGGG GUUCUCCAC-3') were designed and verified to be specific for each cadherin by a Blast search (National Center for Biotechnology Information, Bethesda, MD, USA) against the mouse genome. A scrambled oligonucleotide sequence without significant homology to murine sequences (5'-AGUCGCUUAG AAACGAGAA-3') was used as a control. These oligonucleotides were then cloned into the lentiviral vector, pLVTHM, according to the guidelines provided by Tronolab (Laboratory of Virology and Genetics, École Polytechnique Fédérale de Lausanne, Switzerland). Viral particles were produced by transient co-transfection of 293T cells with the recombinant pLVTHM lentivector constructs, the packaging vector psPAX2 and the envelope vector pMD2.G. Ep5ExTu cells were then transduced with the lentiviral particles contained in supernatants of transfected 293T cells, and stable cell lines were selected by fluorescence-activated cell sorting (FACS) on the basis of green fluorescent protein (GFP) expression. Clones of Ep5ExTu cells expressing Sh-VE-cadherin and Sh-N-cadherin were expanded and the expression of VE-cadherin and N-cadherin was monitored by immunoblot. Experiments were approved by the Sächsisches Staatsministerium für Umwelt und Landwirtschaft, Dresden, Germany (Re: 55-8811.72/69).

Human SUM 149 cells were cultured in DMEM-F12 supplemented with 15% FCS. Silencing of human N-cadherin or human VE-cadherin was performed by using SMARTpools (Dharmacon, Lafayette, CO, USA). Cells (1-2 × 10⁵) were seeded in 2 ml DMEM, 15% FCS in 6-well plates 24 h before transfection. The medium was replaced by 2 ml Opti-MEM I (Invitrogen, Karlsruhe, Germany) 1 h before transfection. 100 pmol siRNA was mixed with 500 μl Lipofectamine 2000 (Invitrogen) diluted in a final volume of 1 ml Opti-MEM I and incubated for 30 min at room temperature to allow the formation of complexes. For transfection, the medium was removed and the DNA-Lipofectamine mixture was added to the cells, which were then incubated at 37°C. 1 ml DMEM, 15% FCS was added 6 h after transfection and the cells were cultivated for another 24 h before analysis.

Mouse cDNA encoding VE-cadherin (kindly provided by Prof. D. Vestweber, Münster, Germany) was cloned into the P6NST50 vector (kindly provided by Prof. D. Lindemann, Dresden, Germany). Ep5ExTu (Sh-N-cad2) cells were transduced with the VE-cadherin-encoding virus particles. Cells stably expressing VE-cadherin were then selected by FACS on the basis of their GFP expression.

Ep5ExTu cells (10⁵) were plated and labeled with bromodeoxyuridine (BrdU) in 96-well plates in DMEM-F12 supplemented with 15% FCS. After 24 h, cell proliferation was quantified using a colorimetric immunoassay based on BrdU incorporation according to manufacturer's instructions (Roche, Mannheim, Germany). The amount of BrdU incorporated into the cells was measured by using an ELISA plate reader (Plus MS2 Reader, Titertek, Huntsville, AL, USA). For each independent experiment, six wells per condition were used.

Ep5ExTu cells (2 × 10⁵) were seeded in 4 ml DMEM-F12 supplemented with 15% FCS in 6 cm dishes 24 h before RNA isolation. Total RNA was isolated from cell lysates using a universal RNA Purification Kit according to the manufacturer's protocol (Roboklon, Berlin, Germany). Aliquots of 3 μg of total RNA were reverse transcribed using Superscript II (Invitrogen, Karlsruhe, Germany) and random hexameric primers (Roche, Mannheim, Germany). The sequences of the PCR primers used are shown in Additional file 1. The intensity of the PCR bands was quantified using Bio-Rad densitometer and Quantity One analysis software (Hercules, CA, USA). qRT-PCR analysis for human VE-cadherin and N-cadherin was performed following reverse transcription of total RNA using a Reverse Transcriptase Core Kit (Eurogentec, Seraing, Belgium), by real-time PCR (Mastercycler ep Realplex; Eppendorf, Hamburg, Germany) using QuantiFast SYBR Green PCR Kit (Qiagen, Valencia, CA, USA). All reactions were run in duplicates and Ct values were normalized against the GAPDH gene, using the delta-delta-Ct method. Primer sequences used were: VE-cadherin (CDH5), (forward) 5'-CGT GAG CAT CCA GGC AGT GGT AGC-3', (reverse) 5'-GAG CCG CCG CCG CAG GAA G-3'; N-cadherin (CDH2) (forward) 5'-CCA CCT TAA AAT CTG CAG GC-3', (reverse) 5'-GTG CAT GAA GGA CAG CCT CT-3'; GAPDH, (forward) 5'-CTC CTC TGA CTT CAA CAG CGA CA-3', (reverse) 5'-GAG GGT CTC TCT CTT CCT CTT GT-3'.

Immunoblot analysis was performed as described previously [7, 26, 27]. The primary antibodies used were anti-VE-cadherin (R&D Systems, Wiesbaden, Germany), anti-N-cadherin and anti-β-actin (Sigma-Aldrich, Munich, Germany). The secondary antibodies were horseradish peroxidase-conjugated anti-rabbit immunoglobulin G (IgG) (Novus Biologicals, Littleton, CO, USA), anti-goat IgG (Jackson Immunoresearch, Soham, UK) and anti-mouse IgG (Cell Signaling Technology, Frankfurt, Germany). Band intensity was quantified using Quantity One analysis software (Bio-Rad, Hercules, CA, USA). Antibodies used for detection of human VE-cadherin, N-cadherin and E-cadherin were goat anti-VE-cadherin (Santa Cruz Biotechnology, Santa Cruz, CA, USA), mouse anti-N-cadherin (BD Biosciences, Bedford, MA, USA), mouse anti-E-cadherin (BD Biosciences, Bedford, MA, USA). Membranes were incubated with secondary antibody conjugated with IRDye Infrared Dyes (IRDYE 800CW donkey anti-goat, IRDYE 800CW donkey anti-mouse), and bands were revealed with a LI-COR scanner (LI-COR Biosciences, Lincoln, NE, USA).

Cells (2 × 10⁵) were seeded on glass coverslips. After 24 h, cells were fixed as described [7] and stained with the following primary antibodies: anti-VE-cadherin (R&D Systems, Wiesbaden, Germany), anti-N-cadherin (BD Biosciences, Bedford, MA, USA), anti-E-cadherin (Sigma-Aldrich, Munich, Germany), anti-β-catenin (Cell Signaling Technology, Frankfurt, Germany) and anti-Vimentin (Sigma-Aldrich, Munich, Germany). The secondary antibodies used were goat anti-rat Alexa 594, chicken anti-rabbit Alexa 594 and rabbit anti-goat 594 (Molecular Probes, Leiden, The Netherlands). Antibodies used for detection of human VE-cadherin, N-cadherin and E-cadherin were: goat anti-VE-cadherin (Santa Cruz Biotechnology, Santa Cruz, CA, USA), mouse anti-N-cadherin (BD Biosciences, Bedford, MA, USA), mouse anti-E-cadherin (BD Biosciences, Bedford, MA, USA).

Ep5ExTu cells were cultured in DMEM-F12 supplemented with 15% FCS. Cells were then trypsinized, rinsed twice in 5 ml PBS and resuspended in PBS at a concentration of 1 × 10⁶ cells/ml. Tumor experiments were performed as described previously [27] with eight- to twelve-week-old female BALB/c mice (Taconic, Ejby, Denmark). All tumors were measured with calipers every two to three days and the volume of each measurement was calculated as: (width² × length)/2. Tumors were collected 10 or 15 days after injection, embedded in Tissue Tek (Sakura Finetek, Staufen, Germany) and frozen on dry ice. Animal experimentation was approved by the Landesdirektion Dresden, Germany (Re: 24-9168.11-1/2009-19).

Eight μm frozen sections were cut and air dried. Sections were then fixed in 100% acetone for 10 min at -20°C and air dried. Rehydrated sections were stained with antibodies as described for immunofluorescence staining of cells.

The tissue microarray included formalin-fixed, paraffin-embedded probes of 84 invasive breast cancers. The clinicopathological features are summarized in Additional file 2. Specimens were dewaxed, and immunohistochemical staining was performed using an automated immunostainer according to the manufacturer's protocol (Benchmark; Ventana Medical Systems, Tucson, AZ, USA) as described previously [7]. The primary antibodies were anti-human VE-cadherin and anti-human N-cadherin (Polyclonal; Abcam, Cambridge, UK). The signal was amplified using the VENTANA amplification kit (Benchmark; Ventana Medical Systems, Tucson, AZ, USA) and visualized using avidin-biotin labeling and 3, 3'-diaminobenzidine. Slides were counterstained with hematoxylin. Evaluation of the staining was performed separately for nuclear, cytoplasmic and membrane-associated expression in a semi-quantitative manner. Expression was considered as positive if at least 1% of the tumor cells were stained. Chi-squared test was used for statistical evaluation of the results, and a P value < 0.05 was considered as statistically significant. The study was approved by the Ethics Committee (Ethikkommisssion) of the Faculty of Medicine of the University of Dresden (Re: EK59032007).

We recently reported that VE-cadherin is aberrantly expressed in approximately 73% of human invasive mammary carcinomas (27 out of 37 cases). Expression was observed in a subset of tumor cells, as well as in the tumor vasculature. Here, we have extended this analysis by examining a larger number (n = 84) of invasive breast cancer specimens in a tissue microarray. This array included carcinomas of different tumor stages, histological types and histopathological grades. The clinicopathological features are summarized in Additional file 2. VE-cadherin protein was detected by immunohistochemistry in a subset of tumor cells in 61 of the 79 (77%) invasive carcinomas examined (Table 1). Yet, we did not observe a significant correlation between VE-cadherin expression and clinicopathological factors (data not shown). 82% of the tumors expressed N-cadherin protein, and 62% of the specimens were positive for both VE- and N-cadherin. This result shows for the first time that these two cadherins can be co-expressed in breast cancer (Figure 1). However, the localization of VE- and N-cadherin in tumor cells appears to be partially different (Table 1). VE-cadherin membrane staining was observed in 31 tumors (39%) whereas N-cadherin was localized at the membrane in only 6 breast cancers (7%). Table 2 summarizes the relationship between N-cadherin expression and the clinicopathological features of the breast cancers: membrane localization of N-cadherin in tumors significantly correlated with smaller tumor size (5 out of 36 cases pT1 vs. 1 out of 47 cases pT2, P = 0.04); in contrast, nuclear expression of N-cadherin was observed primarily in poorly differentiated tumors.

To investigate the functional role of N-cadherin in breast cancer cells, N-cadherin expression was silenced by transfection of two different siRNAs into Ep5ExTu cells. Unexpectedly, downregulation of N-cadherin in Ep5ExTu cells resulted in reduced VE-cadherin expression (see Additional file 3). Next, we stably silenced N-cadherin expression by lentiviral transduction with two different short hairpin RNAs (shRNAs), Sh-Ncad1 and Sh-Ncad2. We selected three independent cell clones (Sh-Ncad1.1, Sh-Ncad1.2 and Sh-Ncad2) that displayed reduced N-cadherin expression in comparison to the control cell line that had been transduced with a scrambled shRNA (Sh-Scr) (Figure 2a). In Sh-Ncad1.1 and Sh-Ncad2 cells, N-cadherin protein levels were reduced by approximately 75% as compared to the Sh-Scr cell line, whereas in Sh-Ncad1.2 cells, N-cadherin protein expression was reduced by only 45%. Consistent with the results obtained from the transient transfection with siRNA (see Additional file 3), immunoblot analysis (Figure 2a) and immunofluorescence staining (see Additional file 3) of the cells revealed also a clear reduction in VE-cadherin expression in the N-cadherin knockdown cell lines (Sh-Ncad1.1, Sh-Ncad1.2 and Sh-Ncad2). This result suggests that N-cadherin is required for sustaining the expression of VE-cadherin in Ep5ExTu cells.

In contrast, VE-cadherin silencing by lentiviral transduction of Ep5ExTu cells with shRNA specific for VE-cadherin affected the protein level of N-cadherin only slightly because N-cadherin levels were similar to the Sh-VE-cadherin cell lines (Sh-VEcad1 and Sh-VEcad2) and the control Sh-Scr cell line (Figure 2b). However, as described for endothelial cells [28, 29], VE-cadherin expression affected N-cadherin localization (Figure 2c and [7]). In the Sh-Scr cell line, VE-cadherin was expressed heterogeneously and presented at the cell surface similar to untransfected Ep5ExTu cells, whereas N-cadherin staining was uniformly distributed over the cell and localized at cell-cell contacts in only a few cells (Figure 2c). In cells silenced for VE-cadherin (Sh-VEcad1), N-cadherin was enriched at cell-cell contacts. This shows that silencing of VE-cadherin leads to a redistribution of N-cadherin to cell-cell contacts, without influencing its expression levels significantly.

We also investigated the expression of VE-cadherin and N-cadherin in the human breast carcinoma cell line, SUM 149. By western blot analysis, immunofluorescence staining and qRT-PCR, SUM 149 cells expressed VE-cadherin and N-cadherin, in addition to E-cadherin (see Additional File 4). Knockdown of N-cadherin resulted in reduction of VE-cadherin mRNA levels but not vice versa. Thus, regulation of VE-cadherin by N-cadherin is observed in both mouse and human breast cancer cells.

Ep5ExTu cells exhibit fibroblastoid characteristics, such as elongated shape and relatively weak cell-cell contacts as compared to epithelial cells [7, 25, 26]. We have previously observed that E-cadherin is expressed in epithelial Ep5 cells, but not in fibroblastoid Ep5ExTu cells, in agreement with the observation that E-cadherin is downregulated during EMT [7]. The two Ep5ExTu cell lines in which N-cadherin expression was efficiently silenced by shRNA (Sh-Ncad1.1 and Sh-Ncad2), displayed a cobblestone-like appearance characteristic for epithelial cells, with individual cells abutting each other (Figure 3a). Consistent with the idea that these cells have undergone MET, the epithelial phenotype of these Sh-N-cadherin cell lines was associated with re-expression of E-cadherin (Figure 3b and 3c). Thus, E-cadherin can be re-induced in Ep5ExTu cells by blocking N-cadherin expression. Interestingly, E-cadherin expression was stronger in the Sh-Ncad1.1 and Sh-Ncad2 cell lines than in the Sh-Ncad1.2 cell line (Figure 3b and 3c), in which N-cadherin expression is only moderately suppressed. Similar to Ep5 cells, E-cadherin staining was observed at sites of cell-cell contacts in both Sh-Ncad1.1 and Sh-Ncad2 cell lines (Figure 3d). These results suggest that the re-induction of E-cadherin in Ep5ExTu cells and its localization at cell-cell contacts correlates with the extent of N-cadherin reduction: the lower the N-cadherin levels are, the more pronounced is the epithelial phenotype. Silencing VE-cadherin in Ep5ExTu cells, in contrast, led to only a partial gain of the epithelial morphology and was not associated with re-induction of E-cadherin (see Additional file 5). This shows that N-cadherin has a more pronounced influence on the maintenance of the mesenchymal tumor cell phenotype than VE-cadherin.

β-catenin accumulation in the nuclei of carcinoma cells and transcriptional activation of its target genes are known to take place during malignant transformation, thereby stimulating tumor progression [2, 30‐33]. E-cadherin can sequester β-catenin in epithelial cells, thus lowering the amount of cytoplasmic and nuclear β-catenin, decreasing its transcriptional activity and inhibiting its tumorigenic potential [34]. To investigate whether the localization of β-catenin in Ep5ExTu breast cancer cells is altered by N-cadherin silencing, we performed immunofluorescence staining. In (epithelial) Ep5 cells, β-catenin was localized at the plasma membrane (Figure 4a) where also E-cadherin is expressed (Figure 3d). Likewise, β-catenin was localized at the plasma membrane along with E-cadherin in the two Sh-N-cadherin cell lines that had greatly reduced N-cadherin (Sh-Ncad1.1 and Sh-Ncad2) (Figure 3d, Figure 4a). In contrast, clone Sh-Ncad1.2, which had moderately reduced N-cadherin, and clone Sh-VE-cadherin showed no significant differences in β-catenin localization compared to the Sh-Scr control cell line (Figure 4a and Additional file 6), and we observed localization of β-catenin to cell-cell contacts in only a few cells. Together, these results show that efficient silencing of N-cadherin in mesenchymal murine mammary carcinoma cells leads to the re-expression of E-cadherin and the co-localization of this receptor with β-catenin at the plasma membrane.

Nuclear localization of β-catenin and the expression of vimentin are involved in epithelial-to-mesenchymal transition and correlate with enhanced invasive and migratory properties of cells [24, 35]. Because β-catenin can stimulate vimentin expression [36], we tested whether the recruitment of β-catenin to the cell membrane of the N-cadherin knockdown cell lines is correlated with changes in the expression of EMT-related genes. The expression of the transcription factor Snail was reduced in all Sh-N-cadherin cell lines as compared to the control cell line (Figure 4b and 4d), whereas the expression of Smad interacting protein-1 (SIP1) and vimentin, which are involved in EMT [37], was downregulated only in Sh-Ncad1.1 and Sh-Ncad2 cell lines (Figure 4c-f and Additional file 6). Whether or not vimentin and SIP1 expression is regulated by N-cadherin remains to be determined.

We have previously shown that silencing VE-cadherin leads to reduced Ep5ExTu cell proliferation and tumor growth [7]. To determine if N-cadherin downregulation has a similar effect, we evaluated the proliferation of Sh-N-cadherin cell lines in vitro by BrdU staining. Sh-VEcad2 and Sh-Ncad1.2 cells displayed lower proliferation rates than the control cell line, whereas the proliferation rate of Sh-Ncad1.1 and Sh-Ncad2 cells was not reduced (Figure 5a). To study the effect of N-cadherin on tumor growth, each of the three Sh-N-cadherin cell lines and the control cell line (Sh-Scr) were injected subcutaneously into BALB/c mice. All Sh-N-cadherin cell lines grew significantly slower than the control cell line (Figure 5b). Notably, the Sh-Ncad1.1 and Sh-Ncad2 cell lines, which showed the strongest reduction of N-cadherin expression, hardly grew in vivo. The Sh-Ncad1.2 cell line showed an intermediate growth rate, similar to the Sh-VE-cadherin cell line (Figure 5c). These data show that a strong (approximately 75%) reduction of N-cadherin expression in Ep5ExTu cells is necessary to inhibit tumor growth efficiently. To perform a histological analysis of tumor tissue, we isolated tumors at day 10 post injection. Downregulation of N-cadherin and VE-cadherin expression in the Sh-Ncad1.2 and Sh-Ncad2 tumors was confirmed by immunofluorescence staining (see Additional file 7). E-cadherin staining was only visible in the Sh-N-cadherin tumors (Figure 5d), and was more intense in Sh-Ncad2 than in Sh-Ncad1.2 tumors. Because vimentin expression has been implicated in carcinogenesis [38, 39], we also examined the expression of this intermediate filament in the different experimental tumor groups. Vimentin expression was strongly reduced only in the Sh-Ncad2 cell line (Figure 5d). Taken together, our results suggest that both re-induction of E-cadherin expression and downregulation of vimentin expression contribute to tumor growth inhibition resulting from N-cadherin silencing in Ep5ExTu cells.

Because N-cadherin silencing in Ep5ExTu cells also led to a reduction of VE-cadherin expression, and VE-cadherin silencing inhibited tumor growth, we wanted to determine whether the loss of VE-cadherin contributes to the tumor growth inhibition resulting from silencing N-cadherin. To address this question, we determined whether the forced expression of VE-cadherin in Sh-Ncad2 cells can affect tumor growth of N-cadherin-deficient cancer cells. A lentiviral expression vector encoding VE-cadherin was therefore introduced into Sh-Ncad2 cells. Two independent clones (N-cadVE1 and N-cadVE2) showed VE-cadherin expression levels similar to that of the Sh-Scr cell line (Figure 6a). The forced expression of VE-cadherin had no influence on cell morphology (see Additional file 8), and the expression of N-cadherin and vimentin in N-cadVE1 and N-cadVE2 cells remained unchanged (Figure 6b). These results suggest that expression of VE-cadherin in the absence of N-cadherin cannot induce Sh-Ncad2 cells to undergo EMT.

Next, we explored whether forced expression of VE-cadherin can influence cell proliferation. No significant difference in cell proliferation was observed between the VE-cadherin-expressing cell lines (N-cadVE1 and N-cadVE2) and control cell line (Sh-Ncad2) (see Additional file 8). To evaluate whether forced VE-cadherin expression influences the growth of Sh-N-cad2 tumors, we inoculated cells from the VE-cadherin-expressing lines (N-cadVE1 and N-cadVE2), Sh-N-cad2 or control Sh-Scr cell lines into wild-type BALB/c mice and monitored tumor growth. The VE-cadherin-expressing cell lines (N-cadVE1 and N-cadVE2) and Sh-Ncad2 displayed comparable growth rates and scarcely grew in vivo (Figure 6c). These results show that forced VE-cadherin expression cannot accelerate tumor growth in the absence of N-cadherin expression, indicating that the concomitant loss of VE-cadherin expression does not contribute to the reduction of tumor growth resulting from N-cadherin silencing.

Since the localization of VE-cadherin within adherens junctions is required for its function [5], we examined the possibility that the inappropriate localization of VE-cadherin is responsible for its inability to restore tumor growth of N-cadherin-deficient cells. By immunofluorescence staining, VE-cadherin expression was observed at the cell borders in N-cadVE1 cells, similar to control Sh-Scr cells (Figure 7). E-cadherin was also localized at cell-cell junctions as in the Sh-Ncad2 control cell line (Figure 7). This shows that VE-cadherin cannot displace E-cadherin from the cancer cell surface.