Phenotype-dependent effects of EpCAM expression on growth and invasion of human breast cancer cell lines

MCF-7, T47D and SkBR3 cell lines were purchased from the American Type Culture Collection (ATCC). All three cell lines were isolated from metastasis and display an epithelium-like phenotype. MCF-7, T47D cell lines were cultivated in MEM, Eagle’s medium (PAA Laboratories GmbH), and SKBR 3 in McCoy’s medium (PAA Laboratories GmbH) containing 10% fetal calf serum (PAA Laboratories GmbH) and 100 IU/mL penicillin, 100 μg/mL streptomycin and 2 mM glutamine (all PAA Laboratories GmbH). MCF-7^ns-ctrl, MCF-7^E#2, T47D^ns-ctrl, T47D^E#2, SkBR3^ns-ctrl and SkBR3^E#2 cell lines were generated by lentiviral transfection with the corresponding pGIPZ shRNA mir virus and selected with 3µg/mL puromycin (Invitrogen) for 5 days in standard culture medium.

Hs578t and MDA-MB-231 cells (ATCC) originate from metastasis of human mammary carcinosarcoma and a poorly differentiated adenocarcinoma, respectively. Moreover, the latter cell type is an elongated and a highly invasive, metastatic phenotype. Hs578t and MDA-MB-231 cells were cultivated in RPMI1640 containing 10% tetracycline-free FCS (Clontech). Parental lines of Hs578t and MDA-MB-231 were lentivirally transfected (pLENTI6/TR, Invitrogen) and selected with 1 mg/mL neomycin (Biochrom) to express the tetracycline repressor (TR) protein. Selected Hs578t^TetR and MDA-MB-231^TetR lines were lentivirally transfected (pLenti6.3 EpCAM) and stable cell lines selected with 2 μg/mL blasticidin (Invitrogen). Hs578t^{TetR EpCAM} and MDA-MB-231^{TetR EpCAM} lines were propagated in 10% tetracycline-free FCS (Clontech).

Immortalized non-tumorigenic human mammary epithelial cells (MCF-10A) were obtained from the ATCC and cultivated Dulbecco's modified Eagle's medium F12 (DMEM/F12) supplemented with 5% horse serum (both Invitrogen), 1% penicillin/streptomycin (all PAA Laboratories GmbH), 0.5 μg/mL hydrocortisone, 10 μg/mL insulin and 20 ng/mL recombinant human EGF (all Sigma Biochemicals). Human Mammary Epithelial Cells (HMECs) were purchased from Promocell. HMECs were cultivated in Mammary Epithelial Cell Growth Medium with recommended supplements (Promocell, 0.004 mL/mL Bovine Pituitary Extract, 10 ng/mL epidermal growth factor, 5 μg/mL insulin and 0.5 μg/mL hydrocortisone) on collagen type I (Sigma Biochemicals) -coated ventilated plastic flasks.

Tissue sections were deparaffinized and hydrated in xylene and graded alcohol series. Antigen retrieval was performed in a water bath (95°C) for 20 min with a target retrieval solution (Dako Cytomation), and endogenous peroxidase activity was blocked with 3% H₂O₂/methanol. Sections were incubated in blocking solution containing 10% fetal calf serum (Dako Cytomation) for 45 min and then stained for one hour with primary antibody (mouse anti-human EpCAM, ESA, clone VU-1D9, Novocastra, 1 μg/mL). Moreover, serial sections were incubated with an alpha smooth muscle cell actin (clone 1A4; 1:100, Sigma Biochemicals) or Ki67 (clone MIB-1, Dako Cytomation), or desmin (clone CD33, Dako), or vimentin (clone V9, Dako), or E-cadherin (clone 32A8, Cell Signaling Technology). Primary antiserum was detected after incubation with a biotinylated secondary antibody (biotinylated rabbit anti-mouse IgG, Vector Laboratories Inc.) using the Vectastain Elite ABC Kit (Vector Laboratories Inc.) and the FAST DAB Tablet Set (Sigma Biochemicals). Sections were counterstained with Meyer’s hematoxylin and mounted with Pertex (Medite).

Cells were seeded on eight-well culture chamber slides (Falcon BD Labware) at densities of 20,000 cells/well and incubated overnight in 5% CO₂ at 37°C. After being fixed in 4% paraformaldehyde and permeabilized with 0.2% Triton-X-100 cells were blocked with PBS containing 3% BSA for 45 min at room temperature (RT). E-cadherin- (clone 32A8, Cell Signaling Technology), vimentin- (clone v9, Dako Cytomation), and cytokeratin 18- (clone DC 10, Dako) specific antibodies were applied in a 1:200 dilution at RT for two hours. After washing in PBS cells were incubated with secondary fluorochrome-labeled antibodies (FITC-labeled rabbit anti-mouse, Dako Cytomation) and nuclei were counterstained with DAPI (Molecular Probes). Cells were embedded in fluorescent mounting medium (Dako Cytomation) and viewed with constant laser settings using a fluorescence microscope (Zeiss Axiovert 200M with Axiovision 4.7 Software, Carl Zeiss Optics).

Cells were harvested and lysed in an RIPA buffer (Sigma) containing protease inhibitors (Complete Mini EDTA-free; Roche Applied Science). Total protein (20 μg) was denaturated, separated by a 4%-20% SDS-PAGE (Criterion TGX, Bio-Rad) and transferred to Immuno-Blot^TM polyvinylidene difluoride (PVDF) membrane (Bio-Rad). After blocking the membrane in 5% non-fat milk powder dissolved in phosphate-buffered saline (PBS), membranes were incubated overnight in 0.5% non-fat milk powder at 4°C with primary mouse antibodies. Antibodies used for Western Blot analysis were a mouse monoclonal directed against EpCAM (C-10; Santa Cruz Biotechnology) and a mouse monoclonal against alpha tubulin (clone B5-1-2; Sigma Biochemicals). Afterwards, membranes were incubated for one hour with an HRP-conjugated goat anti-mouse IgG (Dako Cytomation) diluted 1:1,000. After washing, a chemoluminescent substrate (LumiGLO Reagent and Peroxide, Cell Signaling Technology) was added to the membrane, which was then exposed in the Chemidoc XRS station (Biorad Laboratories). Different glycosylated isoforms of EpCAM were analyzed by Western Blot after deglycosylating the protein in RIPA buffer by means of PNGase F and Endo H (both from New England Biolabs). Band intensities were analyzed and calculated using the Quantity One 4.6 software (BioRad Laboratories).

Three lentiviral transfer vectors (pGIPZ shRNAmir) targeting different regions of the EpCAM open reading frame (clone Ids: V2LHS_134162, V2LHS_134158, V2LHS_23265) and validated non-silencing control (ns-ctrl; RHS4346) were ordered from Open Biosystems. pGIPZ clones were propagated in the presence of 25 μg/ml zeocin (Invitrogen) and plasmids purified with the high-speed Plasmid MIDI Kit (Qiagen). Lentiviral particles were produced by transfecting HEK 293FT cells (Invitrogen) with helper plasmid mix (trans-lentiviral packaging system, Open Biosystems) and Lipofectamin 2000 (Invitrogen).

The EpCAM cDNA (NM_002354, Open Biosystems) was subcloned into the Gateway pENTR11dual section vector (Invitrogen). The pENTR11 vector was site-specifically recombined with the pDEST6.3 vector (Invitrogen) using the Gateway LR Clonase II Pus Enzyme Mix (Invitrogen). The resulting pDEST6.3 vector with the EpCAM open-reading frame was transformed and propagated in One-Shot Stabl3 bacteria (Invitrogen). Lentiviruses were produced in HEK 293FT cells (Invitrogen) by transfecting cells with the pDEST6.3 EpCAM vector and helper plasmid mix (ViraPower, lentiviral support kit, Invitrogen) using Lipofectamine 2000. Lentiviral titers were determined by real time PCR and quantification of woodchuck hepatitis virus posttranscriptional response element expression (WPRE-for: 5-actgacaattccgtggtgtt; WPRE-rev: 5-agatccgactcgtctgagg), as described elsewhere [18].

Real time cell proliferation and migration experiments were performed using the RTCA DP instrument (Roche Diagnostics GmbH), which was placed in a humidified incubator maintained at 5% CO₂ and 37°C. For proliferation assays cells were seeded in complete medium in 16-well plates (E-plate 16, Roche Diagnostics GmbH) at a density of 5,000 cells/well. The plate, which contained gold microelectrodes on its bottom, was monitored every ten minutes for four hours (adhesion process), then once every 30 min until the end of the experiment, (72h). Cell migration was performed using special 16-well plates with 8-μm pores (CIM plate 16, Roche Diagnostics GmbH). These plates, resembling conventional transwells, have microelectrodes on the underside of the membrane, which is located between an upper and a lower chamber. Cells were seeded into the upper chamber at a density of 20,000 cells/well in a serum-free medium. Lower chambers were filled with complete medium. The plate was monitored every 15 min for 12 h. Data analysis was performed using the RTCA software 1.2 supplied with the instrument (Roche Diagnostics).

The ice-cold collagen type I solution was obtained by mixing on ice 8.2 volumes of rat-tail collagen type I (2.0 mg/ml, Sigma Aldrich), 1 volume of MEM 10x concentrated medium and 0.8 volumes of sterile 0.2 M NaOH; pH was adjusted to 7.4 with sterile NaOH. 3D collagen drop cultures were generated by resuspending 5.0 × 10⁵ breast cancer cells in 30 μl collagen mix. Hs578t^{TetR EpCAM} and MDA-MB-231^{TetR EpCAM} cells were additionally stimulated with tetracycline (1 μg/mL). Then 30 μl collagen drops were dropped into a sterile 10-cm, Ø cell culture dish (Falcon) coated with sterilized parafilm and allowed to polymerize for 30 min at 37°C in a humidified cell culture incubator. After polymerization of 3D cultures, collagen plugs were either transferred into 96-well plates (1 plug/well) and maintained in a 200 μL standard complete cell culture medium for 3D in vitro cell proliferation experiments or directly grafted on a chicken CAM for in vivo tumor growth studies.

MCF-7^ns/ctrl, MCF-7^E#2, T-47D^ns/ctrl, T-47D^E#2, SkBR3^ns/ctrl, SkBR3^E#2 or Hs578t^{TetR EpCAM} or MDA-MB-231^{TetR EpCAM} were grown in 96-well plates in the presence of 2.5% serum. Hs578t^{TetR EpCAM} or MDA-MB-231^{TetR EpCAM} was cultivated in complete medium with or without 1 μg/mL tetracycline. After three days of in vitro growth a cell proliferation assay was performed (BrdU ELISA kit, GE Healthcare). Cells were pulsed with 5-bromo-2’-deoxyuridine 20 h prior to measurement in ELISA. Microplates with cell plugs were centrifuged at 2,000 g to allow the plugs to attach to the bottom of the plate. Thereafter, cells in plugs were fixed with the fixative solution and the ELISA performed according to the manufacturer instructions.

On embryo development day 0 fertilized chicken eggs (Gallus domesticus) were placed in a 75%-80% humidified 37°C incubator to allow normal embryo development. On day 3 eggs were opened, egg shells removed and embryos were placed in a sterile Petri dish in an egg incubator (Grumbach) to induce CAM development. Embryos were inspected daily until the day of experiment. On day 8, when the chorioallantoic membrane (CAM) and its vasculature were well developed, the experiment was performed. Subsequently, four onplants (collagen cell plugs) per chicken were grafted onto the CAM. Growth of tumor xenografts was inspected on a daily basis using a stereo fluorescence microscope (Olympus SZW 10). On day 6 post-grafting chicken embryos were sacrificed by hypothermia, xenografts/CAM cut out and stored in 4% paraformaldehyde (Sigma Biochemicals) for immunohistochemistry or in TRI reagent (Sigma Biochemicals) for RNA isolation.

Statistical analyses were performed with the GraphPad Prism™ software 5.0 (GraphPad Software, Inc.). All tests of statistical significance were two-sided. Student’s T test and the Mann–Whitney U Test were used to study differences between two groups. Statistical analyses of quantitative PCR data were performed according to the delta Ct method described by Pfaffl et al. [19] and p values were calculated with Student's T test or the two-way ANOVA test.

EpCAM gene and protein expression was analyzed in various breast cancer cell lines, immortalized (MCF-10A) and normal mammary epithelial cells (HMECs). Normal and immortalized cells showed no detectable protein in Western Blot analysis (Figure 1A/B). The same holds true for the mesenchyme-like breast cancer cell lines MDA-MB-231 and Hs578t (Figure 1A/B). By contrast, epithelium-like tumor cell lines MCF-7, T47D, and SkBR3 showed strong expression of the EpCAM protein as basic and glycosylated isoforms of 35 and 40 KDa (Figure 1A). Western Blot results were confirmed by real-time PCR (Figure 1C). Epithelial breast carcinoma cell lines had significantly stronger EpCAM gene expression than did HMECs or immortalized MCF-10A mammary epithelial cells. Mesenchymal MDA-MA-231 and Hs578t displayed low amounts of EpCAM mRNA (Figure 1C). Based on these findings EpCAM^high and EpCAM^low breast cancer cell lines were phenotyped for markers of epithelial or mesenchymal differentiation. MCF-7 and T47D cells with strong EpCAM expression were strongly positive for epithelial markers cytokeratin 18 and E-cadherin and negative for mesenchymal vimentin expression as determined by immunofluorescence analysis (Figure 1D). In contrast, Hs578t and MDA-MB-231 cells with low EpCAM expression were strongly positive for the mesenchymal marker vimentin and negative or weakly positive for E-cadherin and cytokeratin 18. Hs578t cells were more transdifferentiated to mesenchymal cancer cells and more fibroblast-like than MDA-MB-231 (Figure 1D).

Based on our previous analysis of EpCAM expression in epithelium- and mesenchyme-like breast carcinoma cell lines we selected three EpCAM^high cell lines for lentiviral knockdown studies, i.e. T47D with a strongly glycosylated, MCF-7 with basic and glycosylated, SkBR-3 with only basic EpCAM isoform (Figure 1A). Glycosylation patterns were analyzed after deglycosylating all EpCAM isoforms by means of endoglycosidases to a 35 kDa not glycosylated isoform (data not shown). EpCAM-high cell lines were lentivirally transfected to express a non-silencing control (n/s ctrl) or EpCAM-specific shRNA together with turbo GFP and a puromycin resistance gene (Figure 2A). Three different sequences, E#1, E#2 and E#3 targeting the open reading frame (ORF) of EpCAM, were pretested (data not shown) and the E#2 sequence giving the best knockdown was used for further studies. Effective knockdown was controlled by real-time PCR on all three cell lines (Figure 2B) and by Western Blot analysis (Figure 2C). Lentivirally transfected cells were selected for five days with puromycin and subsequently used for analysis of cell proliferation. In comparison to n/s ctrl cells, E#2 knockdown cells displayed a significantly lower capacity to proliferate in vitro after seeding the same starting amount of cells into the xCelligence real-time cell counting system (Figure 2D). These observations were made in all three cell lines under serum-reduced conditions, i.e. 2.5% serum, over the observed time period of three days. The pro-survival effect of EpCAM was less pronounced under full mitotic conditions, i.e. 10% serum in the culture medium (data not shown). Additionally, cell proliferation was analyzed by measuring DNA synthesis, i.e. BrdU incorporation, under 3D culture conditions in collagen drops with tumor cells. All used cell lines displayed a significant reduction in DNA synthesis after knockdown of EpCAM within three days of in vitro growth (Figure 2E).

Based on our in vitro results we aimed to study the effects of EpCAM knockdown in the in vivo situation. For this purpose, we selected the chicken chorioallantoic membrane (CAM) xenograft model that allows us to study human cells in an animal microenvironment [20, 21]. Chicken embryos have only innate immune responses and tolerate growth of human cancer cells for a period of six days when cells are transplanted on the surface of the CAM tissue (Figure 3A). MCF-7 and T47D onplants can be observed daily by stereo-fluorescence microscopy due to expression of turbo GFP (Figure 3B). SkBr3 cells failed to form vascularized tumors that invade the CAM (data not shown). Macroscopically, MCF7 and T47D tumors did not differ in size when analyzing and calculating only surface areas on CAMs. Immunohistochemical analysis of tumor cells by the proliferation marker Ki67 revealed that knockdown of EpCAM significantly affected tumor cell invasion into host tissue (Figure 3C). Knockdown of EpCAM strongly reduced the invasion capacity of MCF-7 and T47D tumor cells in chicken tissue (Figure 3C, yellow line and arrows). Moreover, as determined by staining for mesenchymal cells, such as pericytes (desmin) and smooth muscle cells of chicken blood vessels (ASMA), reactive responses of the chicken stroma to the invading tumor cell front were nearly abolished (Figure 3C).

Based on our previous analysis on EpCAM expression in epithelium- and mesenchyme-like breast carcinoma cell lines (Figure 1) we selected two EpCAM^low cell lines for inducible lentiviral overexpression studies, i.e. MDA-MB-231 and Hs578t breast carcinoma cells, both with a mesenchymal cancer phenotype (Figure 1D). Cells were lentivirally transfected to express the bacterial tetracycline-repressor protein (TET-R) and selected with neomycin for ten days. Thereafter, MDA-MB-231^TetR and Hs578t^TetR cell lines were lentivirally transfected with an EpCAM open reading frame under a tetracycline-repressible CMV promoter (Figure 4A) and short-time selected with blasticidin. Both cell lines, MDA-MB-231^{TetR EpCAM} and Hs578t^{TetR EpCAM}, were tested for tetracycline-induced expression of EpCAM by real-time PCR (Figure 4B). In comparison to cells growing without tetracycline and having low endogenous transcript levels, addition of tetracycline induced a 20- to 30-fold expression of the EpCAM gene. In addition to gene expression, we also analyzed the induction of EpCAM protein expression by Western Blot analysis (Figure 4C). In both cell lines, MDA–MB-231^{TetR EpCAM} and Hs578t^{TetR EpCAM}, we observed a strong induction of EpCAM protein already 24 h after adding tetracycline to the culture medium (Figure 4D). It is noteworthy that both cell lines produced predominantly the glycosylated (40 kDa) isoform and, to a minor extent, also the hyperglycosylated isoform (42 kDa). Tetracycline-inducible EpCAM cell lines were used for analysis of in vitro cell proliferation in the xCelligence system. Induction of EpCAM by tetracycline had no statistically significant effect on cell proliferation within an observed time period of three days (Figure 5A). Since Hs578t as well as MDA-MB-231 cells show a high capacity to migrate, we studied the effects of EpCAM on in vitro cell migration. Cells were pre-incubated for 24 h with tetracycline, counted and then transferred into the xCelligence real-time cell migration system. Induction of EpCAM by tetracycline did not affect in vitro cell migration within an observed period of 12 h in MDA-MB-231 cells (Figure 5B). In contrast, the fibroblast-resembling Hs578t cells were significantly inhibited in cell migration after overexpression of EpCAM (Figure 5B). Analyses of DNA synthesis in the BrdU incorporation assay of 3D collagen drop cultures also did not reveal significant differences in in vitro cell proliferation, three days after induction of EpCAM (Figure 5C).

Depending on cell culture conditions and the respective microenvironment, MDA-MB 231 cells with parentally low EpCAM and E-cadherin expression can induce endogenous transcription of both genes (Figure 6A/B). The MDA-MB-231 cell line, when growing under 3D culture conditions, displayed larger amounts of EpCAM and E-cadherin than under standard 2-D culture conditions. Transcript levels of E-cadherin and EpCAM were even more elevated under in vivo culture conditions in the CAM.

Hs578t^{TetR EpCAM} cells could not be used in the chicken CAM xenograft model, because they did not invade CAM tissue, lacked neovascularization and displayed many apoptotic cells due to lack of oxygen and nutrients (data not shown). Consequently, experiments were performed with the highly invasive MDA-MB-231^{TetR EpCAM} cells. MDA-MB-231^{TetR EpCAM} cells have low endogenous EpCAM expression on the mRNA level, even without induction with tetracycline (Figure 6A). Stimulation with tetracycline in the collagen-graft resulted in an approx. 50-fold induction of EpCAM gene expression, as determined by real time PCR (Figure 6A). Interestingly, in vivo MDA-MB-231 cells re-express E-cadherin, and EpCAM overexpression represses E-cadherin gene expression (Figure 6B). Macroscopically, tumors did not differ in size after six days of in vivo growth (data not shown). We therefore performed immunohistochemical analysis of CAM sections to analyze changes on the cellular and tissue level (Figure 7A-D). MDA-MB-231^{TetR EpCAM} without EpCAM overexpression were highly invasive, growing deeply into host tissue and showed only dispersed cell clusters that were weakly positive for EpCAM (Figure 7A). By comparison, EpCAM-overexpressing tumors were clearly positive for EpCAM protein expression and displayed an irregular invasion front into host tissue (Figure 7A, red line). This front was also visualized by staining for human tumor cells using cadherin (Figure 7B). EpCAM-overexpressing tumors had no compact and homogeneous invasion front into host tissue. Moreover, when analyzing the proliferation marker Ki67 EpCAM^low tumors had a line of highly mitotic cells in the invasion front that was nearly lost in EpCAM^high tumors (Figure 7C). Inside tumors, the number of Ki67-positive cells did not differ significantly. Additionally, we stained mesenchymal human tumor cells for vimentin protein expression (Figure 7D). EpCAM^high tumors showed no differences in vimentin expression, suggesting that EpCAM overexpression did not directly affect epithelial to mesenchymal transition (EMT). Moreover, EpCAM^high tumors showed more innate immune responses from the chicken immune system. Heterophils (granulocytes) of the chicken were recruited to the tumor invasion front (blue line) and invaded in massive clusters between tumor cells (Figure 7D, yellow arrows).