Mouse monoclonal β-catenin and N-cadherin antibodies were purchased from Transduction Laboratories (Mississauga, ON, Canada), a monoclonal E-cadherin antibody (anti-uvomorulin clone Decma-1) was purchased from Sigma (Oakville, ON, Canada), and monoclonal anti-ZO-1 and anti-vimentin antibodies were obtained from Chemicon International (Temecula, CA, USA) and ICN Biochemicals Inc. (Irvine, CA, USA), respectively. Polyclonal ErbB-1, ErbB-2, ErbB-3, and ErbB-4 antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA), a monoclonal ErbB-2/Neu antibody was obtained from NeoMarkers (Fremont, CA, USA), and a monoclonal anti-phospho-tyrosine antibody (clone 4G10) was purchased from Upstate Biotechnology (Charlottesville, VA, USA). Horseradish peroxidase conjugated secondary antibodies were obtained from Jackson Immuno Research Laboratories Inc (West Grove, PA, USA). Texas-red conjugated anti-rat and anti-mouse antibodies, Texas-red conjugated fungal toxin phalloidin, and Alexa-488 goat-anti-rat conjugates were purchased from Molecular Probes (Eugene, OR, USA). The FITC-conjugated anti-goat antibody was obtained from Santa Cruz Biotechnology and the rat anti-HA antibody was purchased from Boehringer-Ingelheim (Burlington, ON, Canada). Recombinant human epidermal growth factor (EGF) was obtained from Austral Biologicals (San Ramon, CA, USA), heregulin (HRG)-β₁ from NeoMarkers, and recombinant human TGF-β₁ and TGF-β₂ were purchased from R&D Systems (Minneapolis, MN, USA).

Mink lung cells (Mv1Lu) were obtained from the American Tissue Culture Collection (ATCC; Mannasas, VA, USA) and routinely cultured in Dulbecco's modified Eagles medium (DMEM; Wisent Inc., St.-Bruno, QC, Canada) supplemented with 10% fetal bovine serum (FBS; HyClone Laboratories Inc., Logan, UT, USA). Murine myeloma NS0 cells (a kind gift from C Guilbault) were maintained in DMEM/NCTC 109 medium (4:1; vol/vol) containing 10% FBS. All cell lines were cultured at 37°C in a humidified 5% carbon dioxide atmosphere.

FVB control and FVB MMTV/activated Neu transgenic mice were purchased from Charles River Laboratories (Wilmington, MA, USA). C3H×C57Bl/6 mice expressing MMTV/TβRII-AS [25] were backcrossed eight times with FVB mice and then interbred to generate mice expressing the MMTV/TβRII-AS on a congenic FVB background. These mice were mated with FVB MMTV/activated Neu transgenic animals to generate FVB MMTV/activated Neu + TβRII-AS bigenic animals. All offspring were screened for transgene expression by PCR using genomic mouse tail DNA. Neu (5'-GAAGGCACTGCCTCTCCGC-3') and TβRII-AS (5'-TTGTGGTTGATGTTGTTGG-3') specific primers were both used in combination with a polyadenylation signal specific primer (5'-CTCTGTAGGTAGTTTGTCC-3'), applying the following conditions: 94°C for 1 min; 34 cycles of 94°C for 30 s, 48°C (TβRII-AS) or 55°C (Neu) for 30 s, and 72°C for 1 min; followed by 72°C for 2 min.

Immediately after weaning, virgin MMTV/activated Neu and MMTV/activated Neu + TβRII-AS expressing female offspring were palpated twice a week for tumor formation. Mammary tumors were harvested, washed twice in ice-cold phosphate-buffered saline (PBS; Wisent Inc), cut into small pieces and incubated in trypsin/EDTA (Wisent Inc.) at 4°C for 30 min, while rocking. Cell suspensions were washed twice with ice-cold PBS, and resuspended in a mixture of DMEM and Ham's F12 (1:1) supplemented with 5% FBS and antibiotics/antimycotics (Wisent Inc.). The cells were further propagated in this medium, which is subsequently referred to as DF/5% unless otherwise stated.

The BRI-JM01 and BRI-JM04 cell lines were isolated by harvesting nonadherent cells 3 hours after plating the tumor suspension. These nonadherent cells were then reseeded and cultured for 3 days. Clusters of epithelial cells were then isolated and transferred to a 24-well plate for further propagation. The BRI-JM05 cell line was established by seeding the tumor cell suspension in growth factor reduced Matrigel without phenol red (Beckton Dickinson, San Jose, CA, USA). Cells were maintained in Matrigel until 'tree-like' structures appeared, which were then recovered using MatriSperse (Beckton Dickinson), rinsed twice with ice-cold PBS, resuspended in DF/5%, and further propagated on tissue-culture plastic.

PCR and RT-PCR analysis were used, respectively, to screen the harvested mammary tumors and generated cell lines for the presence and expression of the activated Neu and TβRII-AS transgenes. Total RNA was isolated from frozen tumors or 2.5 × 10⁶ BRI-JM01, BRI-JM04, and BRI-JM05 cells using Trizol (Invitrogen-Life Technologies, Burlington, ON, Canada). PolyA⁺ RNA was obtained using Oligotex resin (Qiagen, Mississauga, ON, USA), further purified using DNAse I (NEN-Life Technologies, Boston, MA, USA), and quantified using Ribogreen (Molecular Probes). First strand cDNA was synthesized in a final volume of 20 μl using 1 μg polyA⁺ RNA, an 18-base poly(dT) primer, and 200 units Superscript II (Life Technologies). Of this first strand reaction, 2 μl was used in a PCR reaction using Neu or TβRII-AS specific primers under the conditions described for screening of the transgenic mice. Screening of the BRI-JM01 cell line for the presence of TβRIII mRNA transcripts was carried out using two sets of primers (set 1: 5'-CCT-CCT-CCA-CAG-ATT-TT-CCA-3' and 5'-TGA-GTG-CTC-CCT-ATG-CTG-TG-3'; set 2: 5'-ACG-ATC-CAT-GAC-AGT-GAC-CA-3' and 5'-GAA-GCT-CAG-GAG-AAT-GGT-G-3'). For the RT-PCR reactions we used glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as control (5'-ACC-ACA-GTC-CAT-GCC-ATC-AC-3' and 5'-TCC-ACC-ACC-CTG-TTG-CTG-TA-3'). The conditions used for both GAPDH and TβRIII PCR reactions are similar to those described for the Neu amplification.

Plating efficiency of the three cell lines was determined by seeding 40, 75, 150, 300, 600, and 1250 cells/well using six-well plates. After 7 days the wells were rinsed with PBS, the adherent colonies were stained with crystal violet (0.2% weight/vol in 100% ethanol) for 5 min and rinsed with water, and colonies visible were counted. The plating efficiency was calculated as follows: (number of colonies/number of cells seeded) × 100.

Population doubling time of the three cell lines was determined by seeding 1 × 10⁵ cells/well using six-well plates. Cells were trypsinized after 24, 48, 72 and 96 hours of culturing and counted using a hemacytometer.

The anchorage-independent growth of the three cell lines was studied by plating 10⁵ cells in 0.3% agarose on top of a layer of 0.6% agarose, both in DF/5%, using 100 mm dishes. Cells were fed every week and colony growth was evaluated by light microscopy after 7, 14, and 21 days of culturing.

The ability of the BRI-JM01, BRI-JM04, and BRI-JM05 cell lines to grow in three-dimensional culture was evaluated by seeding 5 × 10⁴ cells in growth factor reduced Matrigel (Becton Dickinson) using 12-well plates. In addition to these cells, we also seeded BRI-JM01 cells that had been exposed to TGF-β₁ for 48 hours and subsequently cultured for more than five passages in the absence of this growth factor. Cellular structures were evaluated 1, 2, and 3 weeks after seeding by light microscopy and documented using a Leitz Labovert (Leitz, Wetzlar, Germany) inverted microscope.

The cell cycle distribution of the three cell lines was determined by flow cytometry. Briefly, exponentially growing cells were trypsinized, washed and resuspended in ice-cold PBS, and fixed in 70% ethanol (4°C). Cells were washed the next day with ice-cold PBS, treated with 1 mg/ml RNAse in PBS (15 min at 37°C), and stained (30 min at room temperature) with propidium iodide (Molecular Probes) to a final concentration of 50 μg/ml in PBS before analysis. Cell populations were selected by forward and side scattering using a 488 nm dichroic long pass filter, and doublets were discriminated using the peak versus the integrated signal of propidium iodide red fluorescence (645 nm dichroic long pass and 620 nm band pass filter set). Data were analyzed using Multicycle Software (Phoenix Flow Systems, San Diego, CA, USA).

The epithelial character of the cell lines was evaluated by cell surface E-cadherin expression. Cells were harvested using nonenzymatic cell dissociation buffer (Sigma), then washed twice with 10% FBS containing ice-cold PBS (PBS/10%), and incubated (20 min at 4°C). Cells were pelleted, resuspended in PBS/10%, divided into three equal portions, and incubated with an E-cadherin antibody (1:1600), a control isotype equivalent antibody (anti-HA; 1:1600), or a secondary antibody Alexa-488 (1:400) in PBS/10% (1 hour at 4°C). The cells were rinsed twice in ice-cold PBS/10%, incubated with Alexa-488 secondary antibody in the dark (1 hour at 4°C), washed twice with ice-cold PBS/10%, and finally resuspended in 1 ml PBS. Viable cell populations were selected using forward and side scattering parameters with a 488 nm dichroic long pass filter. FITC or Alexa green fluorescent emission was detected using a 550 nm dichroic long pass and 525 nm band pass filter. Both cell cycle distribution and the cell surface E-cadherin expression of the three cell lines were analyzed using a Coulter EPICS™ XL-MCL flow cytometer (Beckman-Coulter, Fullerton, CA, USA) equipped with a 15 mW at 488 nm argon ion laser as the excitation source.

Mv1Lu and BRI-JM01 cells (2 × 10⁴ cells/well), and BRI-JM04 and BRI-JM05 cells (4 × 10⁴ cells/well) were seeded in 24-well plates. The next day, cells were incubated for 4 hours in either serum-free (recombinant human EGF and HRG-β₁) or medium containing 1% FBS (TGF-β₁), after which serial dilutions of recombinant human EGF, HRG-β₁ or TGF-β₁ were added. After 24 hours, cells were pulse-labeled for 3 hours by adding 0.5 μCi [³H]thymidine/well (Amersham, Piscataway, NJ, USA). Cells were then rinsed with PBS, trypsinized, and harvested using a cell harvester (Cambridge Technology, Cambridge, MA, USA). [³H]thymidine incorporation was determined by scintillation counting (Universol; ICN Biochemicals Inc.) using a β-counter (LKB-Wallac, Turku, Finland).

Mv1Lu and BRI-JM01 cells were seeded at a density of 1 × 10⁵cells/well, and BRI-JM04 and BRI-JM05 cells at a density of 2 × 10⁵cells/well, using 12-well plates. The next day, cells were cross-link affinity labeled with 100 pmol/l [¹²⁵I]TGF-β₁ (NEN-Life Science Products) or [¹²⁵I]TGF-β₂ (iodination according to Philip and O'Connor-McCourt [32]) in the absence or presence of 100× excess unlabeled TGF-β₁ in 300 μl binding buffer (D-PBS/Ca²⁺/Mg²⁺, 1 mg/ml BSA; pH7.4). Receptor bound ligand was cross-linked using Bis(sulfosuccinimidyl) suberate (Pierce Biotechnology, Rockford, IL, USA), as described previously [33]. Cells were solubilized (1% Triton X100, 10% glycerol, 1 mmol/l EDTA, 20 mmol/l Tris-HCl; pH 7.4) and equal amounts of total protein were resolved on a linear 3–12% gradient SDS-PAGE. Immunoprecipitations of the labeled receptor complex were performed using antisera directed against activin receptor-like kinase (ALK)-1, ALK-5 (Tβ RI) or ALK-7 (antisera kindly provided by Dr P ten Dijke [34]), TβRII and TβRIII, before SDS-PAGE. For treatment of the cells with glycosaminoglycan-cleaving enzymes, subconfluent monolayers were treated with 100 mU heparinitase I, II and III, and 10 mU chondroitinase ABC (Ibex Technologies, Montreal, PQ, USA) in 200 μl binding buffer (3 hours at 37°C) before cross-link affinity-labeling with 100 pmol/l [¹²⁵I]TGF-β₁ or [¹²⁵I]TGF-β₂. Gels were fixed (1 hour, 10% HAc/30% MeOH) at room temperature and dried, and [¹²⁵I]TGF-β cross-link affinity labeled proteins were imaged using a PhosphorImager 401E and analyzed using ImageQuant software (both purchased from Molecular Dynamics, Sunnyvale, CA, USA).

Cell motility was determined using two independent assays. For the wound closure assay, cells were seeded at a density of 2 × 10⁵ (BRI-JM01) and 4 × 10⁵ (BRI-JM04 and BRI-JM05) cells/well in 12-well plates in DF/5% and grown to confluency. Monolayers were then 'wounded' with a pipet tip and the ability of the cells to migrate into this 'wound area' was monitored in the absence or presence of 100 pmol/l TGF-β₁.

For the 'black ink assay', 24-well plates were coated with ink according to Al-Moustafa and coworkers [35], with modifications. Briefly, 24-well plates were coated with 0.1% gelatin (30 min at 37°C). Excess gelatin was aspired from the wells, and a black ink solution (Demco, St.-Lambert, PQ, Canada; 1:3 dilution in sterile ddH₂O) was added, covering the bottom of the wells. After 5 min, the excess ink was removed and the wells were rinsed vigorously with sterile ddH₂O and finally PBS. Cells were seeded immediately after the ink-coated wells were prepared at a density of 2 × 10⁴ cells/well in DF/5% in the absence or presence of 100 pmol/l TGF-β₁.

Cell motility in both assays was determined 24 hours after TGF-β addition using a Leitz Labovert inverted microscope. The ink particle-free tracks, produced by the migrating cells in the black ink assay, of five microscopic fields were quantified using ImageJ freeware . Differences in endogenous and TGF-β₁ induced motility between the three cell lines was expressed as the relative fold/cell per 24 hours using the least motile BRI-JM05 cell line as a standard.

BRI-JM01 cells were seeded in glass chamber slides (Lab-Tek, NalgeNunc International, Rochester, NY, USA) at a density of 1 × 10⁴ cells/0.8 cm² in DF/5%. A total of 100 pmol/l TGF-β₁ was added and the cells were incubated for an additional 24 hours. Cells were rinsed with PBS (repeated after each of the following steps), fixed in ice-cold 100% methanol (10 min), and permeabilized with 0.1% Triton X100 in PBS (5 min at room temperature). Nonspecific sites were blocked (30 min at room temperature) with 1.5% nonfat milk-powder in PBS (PBS/1.5% milk), followed by an incubation with antibodies raised against ZO-1 (1/500), E-cadherin (1/500), β-catenin (1/1000), N-cadherin (1/1000), and vimentin (1/50), all in PBS/1.5% milk. This was followed by incubation with a fluorescently conjugated secondary antibody (1/500 in PBS/1.5% milk) for 1 hour at room temperature. F-actin filaments were stained, after fixation of the cells in 4% paraformaldehyde, with Texas-red conjugated phalloidin (30 min at room temperature). Nuclei were counter stained on all slides (5 min at room temperature) using 0.4 μg/ml 4,6-diamidino-2-phenylindole (Sigma) in PBS. Slides were mounted using antifade (Molecular Probes) and fluorescent images were captured using a cooled Coolsnap CCD digital camera mounted on a Leitz upright microscope.

Semiconfluent monolayers of BRI-JM01, BRI-JM04, and BRI-JM05 cells were rinsed with PBS before lysis in CSK buffer (150 mmol/l NaCl, 10 mmol/l Pipes, 30 μmol/l MgCl₂, 300 mmol/l sucrose, 2% Triton X-100) supplemented with a protease inhibitor cocktail (Roche Diagnostics, Laval, PQ, USA). The total ErbB-1, ErbB-2, ErbB-3, and ErbB-4 content of the cell lines was determined by resolving 20 μg total protein by 7.5% SDS-PAGE followed by western blotting. The phosphorylation status of the ErbB-2/Neu receptor was determined by immunopreciptation of ErbB-2/Neu, followed by western blotting with a phosphotyrosine specific antibody. Briefly, semiconfluent monolayers of BRI-JM01, BRI-JM03, and BRI-JM04 cell lines were lysed in lysis buffer (150 mmol/l NaCl; 50 mmol/l Tris-HCl, pH 8.0; 0.55% NP-40; 50 mmol/l NaF). Total protein (500 μg) was precleared with protein G sepharose (4 hours at 4°C) and the ErbB-2/Neu receptor was precipitated (16 hours at 4°C) using 2 μg monoclonal ErbB-2/Neu antibodies coupled to protein G sepharose. The next day precipitates were rinsed five times with lysis buffer and resolved by 7.5% SDS-PAGE.

The E-cadherin, N-cadherin, and β-catenin content of the BRI-JM01 cells was determined by lysing BRI-JM01 cells in CSK buffer after 24, 72, and 120 hours of TGF-β₁ treatment (100 pmol/l final concentration) and resolving 20 μg total protein by 10% SDS-PAGE followed by western blotting.

All proteins were transferred to nitrocellulose, nonspecific sites were blocked using 5% non-fat milk powder in TBS (20 mmol/l Tris-HCl, pH 7.6; 137 mmol/l NaCl) containing 0.1% Tween 20 (TBS-T) for 1 hour at room temperature, and filters were then incubated with primary antibody (E-cadherin, ErbB-1, ErbB-2, ErbB-3, and ErbB-4, 1:500; N-cadherin and β-catenin, 1:1000; phospho-tyrosine, 1 μg/ml) in TBS-T/5% nonfat milk (16 hours at 4°C). Filters were washed with TBS-T (3 × 10 min), incubated with the appropriate secondary HRP-conjugated antibody (1:10,000, 1 hour at room temperature) in TBS-T/5% nonfat milk, and washed with TBS-T (3 × 10 min). Immunoreactive bands were visualized by chemiluminescence (Perkin Elmer, Boston, MA, USA).

To provide mammary epithelial cell lines that may exhibit a combination of reduced TGF-β signaling due to TβRII-AS expression, and activated oncogenic pathways due to Neu expression, we isolated two cell lines (named BRI-JM01 and BRI-JM05) from tumors from the bigenic MMTV/activated Neu + TβRII-AS mice that we have generated. As a potential control, we also isolated a cell line (termed BRI-JM04) from a tumor taken from an MMTV/activated Neu expressing animal. Our analysis of transgene expression at the level of the tumors using RT-PCR analysis of polyA⁺ RNA showed that all of the tumors from four bigenic MMTV/activated Neu + TβRII-AS mice expressed both the Neu and TβRII-AS transgenes (Fig. 1a, left panel). Additionally, as expected (Muller WJ, personal communication), RT-PCR analysis indicated that all the mammary tumors isolated from MMTV/activated Neu transgenic animals expressed the Neu transgene (Fig. 1a, right panel). In control experiments conducted in the absence of reverse transcriptase, no PCR products were generated for any of the transgenes (data not shown), indicating that the polyA⁺ RNA was free of genomic DNA contamination.

In contrast to the results obtained with the tumor samples, RT-PCR analysis of the BRI-JM01 and BRI-JM05 cell lines (which were established from MMTV/activated Neu + TβRII-AS bigenic tumors) unexpectedly indicated that the TβRII-AS transgene was not expressed (Fig. 1b, left panel). The BRI-JM04 cell line (from an MMTV/activated Neu mammary tumor) did not contain the TβRII-AS message, as expected. PCR analysis, using genomic DNA isolated from the three cell lines, showed that the lack of TβRII-AS expression in the BRI-JM01 cells was due to the loss of the TβRII-AS sequence from the genome (Fig. 1c, middle panel). In contrast, the PCR analysis showed that BRI-JM05 genomic DNA contained the TβRII-AS sequence, implying that some form of gene silencing was responsible for the loss of TβRII-AS expression in this cell line (Fig. 1c, middle panel). With respect to Neu transgene expression, RT-PCR and PCR analysis showed that all three cell lines carried the Neu oncogene sequence in their genomes (Fig. 1c, right panel); however, only the BRI-JM04 and BRI-JM05 cells expressed the Neu transgene (Fig. 1b, right panel). The results of these PCR and RT-PCR analyses are summarized in Table 1 and indicate that these cell lines will not serve as good models for studying the effect of TβRII-AS expression on TGF-β signaling because the expression of the TβRII-AS transgene was not maintained. However, the results do suggest that these lines may be useful for studying TGF-β signaling in the absence (BRI-JM01) versus presence (BRI-JM04 and BRI-JM05) of activated Neu oncogene signaling.

The procedures that were used to establish these new cell lines favor the isolation of epithelial cells. However, using light microscopy, we observed that only the BRI-JM01 cells exhibited the typical epithelial-like polygonal morphology (Fig. 2a). Flow cytometry (Fig. 2b), using an anti-E-cadherin antibody and nonpermeabilized cells, revealed that more than 95% of the cells in the three cell lines expressed cell surface E-cadherin, confirming their epithelial character.

Next, we determined the doubling time and the plating efficiency of the three mammary tumor cell lines. The BRI-JM01 cell line exhibited the shortest doubling time (14 hours) and a relatively high plating efficiency (78%). BRI-JM04 and BRI-JM05 cells were shown to have much longer doubling times (22 and 25 hours, respectively) and lower plating efficiencies (55% and 41%, respectively). Using flow cytometry, we determined the cell cycle distribution of exponentially growing BRI-JM01, BRI-JM04, and BRI-JM05 cells. The S+G₂/M ratio confirmed that the BRI-JM01 cell line contained the most rapidly growing cell population, whereas the BRI-JM04 and BRI-JM05 populations multiplied more slowly (Table 2), which is consistent with the doubling time results. It is interesting that the one cell line that did not maintain expression of the activated Neu oncogene exhibited the fastest growth rate.

Culturing the three cell lines for several passages in phenol red-free modified Eagle's medium containing 10% charcoal-stripped/dextran-treated bovine serum did not alter their growth characteristics, indicating that these cells are able to grow in the absence of steroid hormones.

Anchorage independent growth of the cells in soft agar, and growth of the cells in three-dimensional culture (Matrigel) was assessed. Only the BRI-JM05 cells were capable of forming colonies in soft agar; however, the clonogenic capacity of this cell line was under 5% (data not shown). Both the BRI-JM04 and BRI-JM05 cell lines proliferated relatively rapidly in Matrigel and formed structures approximately 50–100 μm in diameter within 2 weeks after seeding the cells (Fig. 2c). The BRI-JM01 cell line also formed structures but at a slower rate, with structures of approximately 25–50 μm in diameter being detected after 2 weeks of culturing.

We demonstrated that the BRI-JM04 and BRI-JM05, but not the BRI-JM01, cells express the message for the activated Neu oncogene (Fig. 1b). Next, we wished to determine the growth response of the three cell lines to EGF and HRG-β₁ by measuring [³H]thymidine incorporation, and to correlate this response with the levels of expression of the ErbB-1, ErbB-3 and ErbB-4 receptors, in addition to that of the activated Neu receptor. The BRI-JM01 and BRI-JM04, but not the BRI-JM05 cells, were growth stimulated by recombinant human EGF (Fig. 3a), whereas only the BRI-JM01 cells were significantly stimulated by HRG-β₁ (Fig. 3b). Western blot analysis (Fig. 3c) showed that the BRI-JM01 and BRI-JM04 cells expressed significant levels of ErbB-1, whereas the BRI-JM05 cells expressed much lower levels of ErbB-1. Western blot analysis also showed that the BRI-JM04 and BRI-JM05 cells expressed high levels of the activated Neu protein (Fig. 3c), confirming the RT-PCR data (Fig. 1b). In addition, immunoprecipitation of the ErbB-2/Neu receptor followed by western blot analysis using an anti-phosphotyrosine antibody showed that the receptor was highly phosphorylated in the activated Neu expressing BRI-JM04 and BRI-JM05 cell lines (Fig. 3c). Finally, all three cell lines expressed relatively high levels of ErbB-3 and ErbB-4 receptors. The correlation between ErbB-1, ErbB-2, ErbB-3, and ErbB-4 receptor expression and recombinant human EGF and HRG-β₁ responsiveness is discussed in terms of receptor expression and heterodimer formation below.

Treatment with TGF-β₁ in the presence of serum resulted in growth inhibition of all three cell lines (Fig. 3d), as might be expected of cells of epithelial origin. Cross-link affinity labeling of cell surface TGF-β receptors using [¹²⁵I]TGF-β₁ revealed the presence of two receptors with molecular weights of 53 kDa and 70 kDa on the three cell lines (Fig. 3e, upper panel), whereas a third receptor (molecular weight approximately 250 kDa) could be detected on only the BRI-JM04 and BRI-JM05 cells. These molecular weights are characteristic of type I, II, and III TGF-β receptors. The identity of the type II and III TGF-β receptors was confirmed by immunoprecipitation of the [¹²⁵I]TGF-β₁ cross-link affinity labeled receptors, using specific antibodies (data not shown). In the case of type I TGF-β receptors, it is known that TGF-β most often interacts with and activates the Alk-5 type I receptor. However, it is also capable of interacting with and activating another type I receptor, termed Alk-1 [36]. In all three cell lines, using specific antibodies, we determined that the affinity labeled 53 kDa receptor corresponded to Alk-5 and not Alk-1.

The type III receptor is known to contain heparin and chondroitin sulfate glycosaminoglycan chains, which results in it exhibiting a diffuse migration pattern on SDS-PAGE. In order to collapse the type III receptor into a tighter band and thereby enhance our ability to detect it at the cell surface, we treated the three cell lines with heparinitase I, II and III, and chondroitinase ABC before cross-link affinity labeling with [¹²⁵I]TGF-β₁ or [¹²⁵I]TGF-β₂ (Fig. 3e, bottom panels). Although, as expected, this deglycosylation procedure collapsed the TβRIII receptor band into a labeled core protein of approximately 115 kDa on the BRI-JM04 and BRI-JM05 cells, it did not reveal the presence of the TβRIII receptor on the BRI-JM01 cells. However, RT-PCR analysis (using two independent sets of TβRIII specific primers) indicated that these cells express the TβRIII message (Fig. 3f). These results indicate that either the TβRIII message is not translated efficiently or that trafficking of the TβRIII to the cell surface is impaired in these cells.

Table 3 summarizes the ErbB receptor and TGF-β receptor expression patterns in the three cell lines, and Table 4 gives an overview of their responsiveness to TGF-β₁, recombinant human EGF and HRG-β₁.

TGF-β is known to stimulate motility in several epithelial cell lines, including the NMuMG and EpH4 mammary epithelial cell lines. We therefore assessed the endogenous, and TGF-β₁-induced, motility of the three cell lines that we have isolated using two independent assays. In the first assay, confluent monolayers of cells were 'wounded' with a pipet tip, and the movement of the cells into the wound area was monitored in the absence or presence of 100 pmol/l TGF-β₁. Figure 4a shows that BRI-JM01 cells exhibited a TGF-β₁-induced increase in motility. This 'wound healing' effect was not due to an increase in cell proliferation because we demonstrated that these cells were clearly growth inhibited by 100 pmol/l TGF-β₁ (Fig. 3d). In contrast, TGF-β₁ did not affect the motility of the BRI-JM04 and BRI-JM05 cell lines in the wound assay.

The second assay is a 'black ink assay' in which the cells were sparsely seeded on an ink-coated surface and then monitored for cell movement, which appears as ink particle cleared tracks on the plastic. This motility assay confirmed the results of the wound assay (i.e. the presence of TGF-β₁ significantly increased the motility of the BRI-JM01 cells, but not of the BRI-JM04 and BRI-JM05 cells; Fig. 4b). Quantification of these particle-free tracks using ImageJ software showed that the basal motility of the BRI-JM01 and BRI-JM04 cells was approximately twice that of the BRI-JM05 cells, and that TGF-β₁ stimulated the motility of the BRI-JM01 cells approximately 1.7-fold (Fig. 4c).

Using phase contrast microscopy, we also observed that, in addition to having their growth inhibited and their motility stimulated by TGF-β₁, BRI-JM01 cells exhibited a striking morphologic change in response to this growth factor, either in the presence (Fig. 5a) or in the absence (data not shown) of serum. The appearance of the cells changed from a homogeneous epithelial-like cuboidal morphology to a spindle-shaped morphology (Fig. 5a, phase contrast). We therefore investigated whether these cells were undergoing an EMT similar to that described for the NMuMG and Ras-transformed EpH4 cells [22,23] by evaluating the expression and/or relocalization of several epithelial and mesenchymal markers.

Using immunofluorescence microscopy, we observed that the BRI-JM01 cells were indeed undergoing an EMT-like process, which was characterized by the following: relocalization of the cell adhesion proteins E-cadherin, N-cadherin, and β-catenin; loss of the tight junctional protein ZO-1; acquisition of actin stress fibers; and reorganization of the cytoskeletal intermediate filament protein vimentin (Fig. 5a). Flow cytometric analysis, using an anti-E-cadherin antibody and nonpermeabilized cells, confirmed that TGF-β₁ exposure caused a relocalization of E-cadherin. Cell surface E-cadherin expression decreased approximately 50% after 48 hours, and almost 100% after 120 hours of TGF-β₁ treatment (Fig. 5b). Immunoblot analysis of whole cell lysates showed that the effect on E-cadherin was primarily one of relocalization because the total E-cadherin content of BRI-JM01 cells was only slightly diminished over the TGF-β₁ time course (Fig. 5c). A similar phenomenon was observed for N-cadherin and β-catenin (i.e. both proteins translocate from the cell surface to cytosolic compartments upon TGF-β₁ exposure [Fig. 5a], with no significant decrease in total protein [Fig. 5c]).

There is an ongoing discussion in the literature as to what criteria may be used to define an EMT in vitro and to distinguish the EMT phenotype from a 'scattering' phenotype [23]. Both phenotypes exhibit disruption of cell junctions, fibroblastoid morphology, and enhanced motility. However, it has been suggested that they can be distinguished by reversibility studies with the EMT phenotype, but not the scattering phenotype, persisting after withdrawal of the stimulus [23]. We investigated the reversibility of the EMT phenotype in the BRI-JM01 cells by treating cells with TGF-β₁ for 48 hours and then culturing them for at least five passages in the absence of TGF-β₁. In two-dimensional culture on plastic, these cells maintained their mesenchymal morphology (data not shown). When grown for 3 weeks in three-dimensional culture in Matrigel, the TGF-β₁ 'primed' cells formed larger structures than did the nontreated BRI-JM01 cells (Fig. 5d). Together, these results indicate that, in the BRI-JM01 cells, EMT is maintained as a metastable phenotype that is independent of TGF-β₁ stimulation.