PMC42-LA cells were grown at 37°C in RPMI-1640 medium with 10% (vol/vol) foetal bovine serum (FBS; Thermo Trace, Melbourne, Australia). Primary human NMFs were isolated from human breast tissue with appropriate consent from women undergoing reduction mammoplasty, in accordance with the standards of St. Vincent's Hospital Human Ethics Committee. Primary mammary CAFs were obtained from minced tumour tissue of a mammary tumour biopsy. Tissue was minced and digested with 200 ml collagenase (Sigma-Aldrich, St. Louis, MO, USA) and 100 ml hyaluronidase (Sigma-Aldrich) in Dulbecco's modified Eagle medium (DMEM)/Ham's F-12 (1:1) with 10% (vol/vol) FBS containing penicillin, streptomycin and amphoporicin B (Thermo Electron Corporation, Melbourne, Australia) overnight at 37°C and 5% carbon dioxide with gentle rocking. Digested tissue was centrifuged at 600 g for 10 min, and the cell pellet was resuspended in medium and filtered through a 40 μm cell strainer. The cells were allowed to attach overnight, cultured in growth medium as above, and used within 10 passages. SV40 immortalized human skin fibroblasts (GM847), obtained for diagnostic purposes, were grown and maintained in basal medium (Eagle's) with 10% (vol/vol) FBS (Sigma-Aldrich) for use as a control. Cells were passaged in 0.05% trypsin/EDTA (Sigma-Aldrich) when confluent; primary cells were passaged a maximum of 11 passages.

To induce organoid formation of PMC42-LA cells, 150 μl undiluted EHS (Sigma-Aldrich) was spread with a cell scraper on chilled PET track-etched/porous membrane cell culture inserts (0.4 μm pore size; Becton Dickinson Labware, Franklin Lakes, NJ, USA). The filters were incubated at 37°C for 30–45 min to allow the EHS to set. Once set, 150 μl diluted (4% vol/vol in chilled distilled water) EHS was placed on the surface of the set EHS, followed by 2 ml of RPMI 10% FBS (ThermoTrace) containing 10⁶ PMC42-LA cells. Then, 3 ml RPMI 10% FBS was placed beneath the filter using a pipette. Cells were incubated for 5 days at 37°C with 5% carbon dioxide before processing. For EHS matrix with fibroblasts beneath the filter, the same procedure was followed except that 0.5 × 10⁵ previously trypsinised fibroblasts (NMFs or CAFs) were placed in the well beneath the filter insert with 1 ml DMEM/Ham's F-12 (10% FBS; Thermo Electron Corporation) and 2 ml RPMI (Thermo Trace).

For conditioned medium experiments, attached confluent fibroblasts were rinsed in phosphate-buffered saline (PBS), and fresh DMEM:Ham's F-12 (10% FBS) media added to flasks. Flasks were returned to 37°C with 5% carbon dioxide, and after 48 hours the conditioned medium was collected, double filter sterilized using an eccentric tip syringe (Terumo Corporation, Tokyo, Japan) and sterile nonpyrogenic 0.2 μm filter (Schleicher & Schuell BioScience, Keene, NH, USA), and used at a 1:1 ratio in addition to fresh RPMI on top or beneath the filter.

Either 10 ml (1×), 5 ml (2×), 2.5 ml (4×), or 1 ml (10×) fresh DMEM:Ham's F-12 (10% FBS) was added to confluent NMFs. After 24 hours, medium was collected and used to culture PMC42-LA cells on glass cover slips before scratch wound assays.

Cell counting was conducted using a standard haemocytometer, and cell viability tests were conducted using dye exclusion with trypan blue solution (0.4%; Sigma-Aldrich).

Nitric acid-treated glass cover slips were used to culture PMC42-LA cells in six-well plates (Nunc, Roskilde, Denmark) for scratch tests. Using a sterilized plastic cafeteria fork, a series of scratches were made on confluent cover slips and these were processed for indirect immunocytochemistry as described below. For each cover slip, four scratches were made, averaging between 367 μm (minimum) and 392 μm (maximum) width, at time 0. Each scratch was measured at several time points and averaged, and the extent of wound closure was measured microscopically, compared with the original width of scratches and averaged. Data were processed and standard deviation calculated using Excel (Microsoft Corp., Redmond, WA, USA).

Mouse anti-human β-actin, anti-human catenin-β and goat anti-human vimentin antibodies were purchased from Sigma-Aldrich. Mouse anti-human vimentin and mouse anti-human E-cadherin antibodies were purchased from Zymed laboratories, Inc. (San Francisco, CA, USA) and mouse anti-human α-SMA antibodies were purchased from Dako (Denmark, Europe). Goat anti-human E-cadherin and rabbit anti-active caspase-3 antibodies were purchased from Chemicon International Inc. (Temecula, CA, USA) and rabbit anti-human α-SMA antibodies from Abcam plc. (Cambridge, UK); anti-human FAP antibodies (muF19) from the biological production facility at Austin Hospital were kindly donated by the Ludwig Institute of Cancer Research (Heidelberg, Australia). The Alexa Fluor secondary antibodies used for immunocytochemistry were purchased from Molecular Probes Inc. (Eurgene, OR, USA) and included goat anti-mouse IgG labelled with Alexa 488, donkey anti-sheep IgG labeled with Alexa 594 and donkey anti-rabbit IgG labelled with Alexa 594. Anti-mouse horse-radish peroxidase conjugated antibody from Sigma-Aldrich was used as a secondary antibody in Western blot analysis.

Confluent PMC42-LA cells cultured in flasks were treated with 0.05% (vol/vol) trypsin/EDTA solution (Sigma-Aldrich), cell suspensions were centrifuged and the pellet re-suspended in 5 ml PBS. Following two washes in PBS and re-centrifugation, the pellets were stored at -80°C until needed, or prepared for quantification as described.

PMC42-LA cells cultured in EHS to form organoids were washed twice in PBS, following which 2 ml of 5 mg/ml dispase (Sigma-Aldrich; 20 mg/ml stock concentration) in RPMI/10% FBS was added to the filters. The filters were incubated for 30 to 90 min, or until cells were detached. Cells were centrifuged at 3,000 rpm for 5 min, washed twice with PBS, and the pellet was dried and then stored at -80°C.

Cell pellets were re-suspended in 500 μl of 1% (weight/vol) SDS in 10 mmol/l Tris-HCl (pH 7.5) on ice and homogenized by passaging through a 21-guage needle 15 times on ice, after which cells were sonicated (40% power output, 30% duty cycle) twice using a Microson Ultrasonic cell disrupter (Misonix Incorporated, Farmingdale, NY, USA). Samples were centrifuged and the supernatant quantified using the Pierce BCA Protein Assay Reagent Kit (Perbio, Rockford, IL, USA), in accordance with the manufacturer's instructions. The protein concentrations were determined by measuring the absorbance at 595 nm.

Proteins were separated using SDS-PAGE and transferred to a nitrocellulose membrane at 10 V for 40 min. The membranes were blocked in 1% (weight/vol) casein/Tris-buffered saline (TBS) overnight at 4°C in a sealed bag. Following this, the membrane was exposed to the appropriate antibody diluted in 1% casein/TBS at varying concentrations overnight at 4°C, and then washed three times for 10 min each in 1% (weight/vol) casein/TBS. The membrane was then exposed to anti-mouse horse radish peroxidase conjugated antibody, diluted at 1:2,000 in 1% (weight/vol) casein/TBS, for 2 hours at room temperature. The membrane was then washed twice in 1% (weight/vol) casein/TBS for 10 min and, finally, in TBS Tween 20 solution twice for 15 min each. A Roche BM chemiluminescence Blotting substrate POD kit (Roche, Indianapolis, IN, USA) was used, in accordance with the manufacturer's instructions, and placed on membranes for 2 min, after which the membranes were placed inside a LAS-3000 Fuji Film intelligent dark box. The illuminated bands were detected and the image captured using Image reader LAS-3000 software. For each result given, Western blot analysis was performed on three separate occasions with new cell lysates.

To re-probe membranes and analyze protein loading, membranes were stripped with a re-blot solution (Re-blot plus strong; Chemicon International Inc.) at a 1:10 dilution, in accordance with the manufacturer's protocol. The membrane was then blocked with 1% (weight/vol) casein/TBS for 1 hour at room temperature and re-probed with anti-β-actin (1:5,000 in 1% casein/TBS) as a control for protein loading and processed, as described above, as a control for protein loading.

To determine the density of the bands present in Western blots, a Multi-Gauge V2.3 program (FUJIFILM Medical Systems Inc., Stamford, Connecticut, USA) was used to read band densities. Data were processed, standard deviations calculated, and t-tests performed using Microsoft Excel software.

Once confluent, cells cultured on glass cover slips were washed three times with PBS, fixed in 4% (weight/vol) paraformaldehyde (PFA/PBS; Sigma-Aldrich) for 10 min at room temperature, and washed again three times in PBS. To permeabilize the cells, 0.1% (vol/vol) Triton X-100/PBS (Amresco Inc., Solon, OH, USA) was applied to cells for 10 min at room temperature, after which the cells were washed three times in PBS and incubated with 1% (weight/vol) bovine serum albumin in PBS (BSA/PBS; Sigma-Aldrich) for 2 hours at room temperature to reduce nonspecific binding of primary antibodies. Following this, 30 μl of the appropriate primary antibody, diluted in 1% (weight/vol) bovine serum albumin/PBS was placed on each cover slip, and the cover slips incubated overnight at 4°C. Cover slips were then washed three times at 15 min each with PBS and were incubated with 30 μl of the secondary antibody (Alexa goat anti-mouse IgG) diluted 1:2000 with 1% (weight/vol) bovine serum albumin/PBS, for 2 hours at room temperature. Cover slips were then washed three times for 5 min each with PBS, and stained with ethidium bromide (diluted 1:10,000 in PBS; Sigma-Aldrich) for 3 to 4 min to visualize nuclei. A final series of three PBS washes of 5 min each was then performed, after which the cover slips were dried and mounted on 1.5 μl of Bio-Rad FluoroGuard™ Anti-fade Reagent (Bio-Rad, Hercules, CA, USA) and cover slip edges sealed with nail polish.

For cells cultured on EHS gel, the filter was inverted and the membrane cut into six to eight segments using a razor. Segments were placed in 35 mm sterile Petri dishes and gently washed three times for 5 min each with PBS. The same procedure was followed for indirect immunocytochemistry, as previously described.

Cells were viewed using a Leica TCS SP2 AOBS laser scanning confocal microscope (Leica, New South Wales, Australia), using oil immersion and a 40× or 100× objective. Images were captured by the photomultiplier tube using the Leica TCS SP2 laser, and viewed on a workstation using Leica microsystems TCS SP2 software. Images were presented using Microsoft PowerPoint software.

The phenotypes of human primary NMFs, obtained from reduction mammoplasty tissue, and human primary mammary CAFs, obtained from minced mammary tumour biopsy tissue, were determined by analysis of protein markers associated with the myofibroblast phenotype.

Western blot analysis revealed that α-SMA (Figure 1a) was expressed by both NMFs and CAFs; however, densitometry revealed 2.7-fold higher expression of α-SMA in CAFs as compared with NMFs. A control sample derived from human immortalized skin fibroblasts (GM847) was negative for α-SMA expression (Figure 1a). Western blot analysis indicated that FAP was expressed by both NMFs and CAFs, but not by the GM847 control fibroblasts (Figure 1b). An increase in band density by twofold was seen for CAFs as compared with NMFs (Figure 1b).

PMC42-LA cells cultured on ECM gel coated filters to form organoids with fibroblasts beneath the filter or fibroblast-conditioned medium were analyzed for myoepithelial-specific and luminal epithelial-specific protein markers.

Compared with control, expression of the luminal epithelial marker E-cadherin, as measured by Western blot analysis, was not significantly upregulated by PMC42-LA organoids cultured with either NMFs (fold increase: 1.8 ± 1.17) or CAFs (fold increase: 1.8 ± 1.31) beneath the filter (Figure 2a, lanes 1 and 2, respectively). Similar values were seen when fibroblast-conditioned medium was used in place of the fibroblasts (fold increase compared with control: 1.5 ± 2.13 with NMF-conditioned medium and 2.4 ± 1.10 with CAF-conditioned medium; Figure 2a, lanes 3 and 4). Immunocytochemistry confirmed that E-cadherin was expressed by most cells in all conditions tested, with no obvious changes in the organization of E-cadherin expressing cells within organoid structures (Figure 2a, lanes 1 to 4).

In the absence of fibroblasts or their conditioned medium, α-SMA was not expressed by PMC42-LA organoids (Figure 2b). When cultured with NMFs beneath the filter, PMC42-LA organoids exhibited a significant 2.5 (± 1.16)-fold increase in α-SMA (P < 0.01), similar to that seen in PMC42-LA cultures with CAFs beneath the filter (fold increase: 2.7 ± 0.42, P < 0.01; Figure 2b, lanes 1 and 2, respectively). Compared with control, conditioned medium from NMFs caused a significant 6.7 (± 2.76)-fold increase in α-SMA expression (P < 0.01) in PMC42-LA cells, which was similar to the significant 7.4 (± 1.20)-fold increase seen in PMC42-LA cells (P < 0.001) cultured in CAF-conditioned medium (Figure 2b, lanes 3 and 4, respectively). Immunocytochemistry revealed that α-SMA-expressing cells were organized similarly within organoids in all fibroblast treatment conditions (Figure 2b, lanes 1 to 4).

Cytokeratin 14 was not expressed by PMC42-LA organoids in the absence of fibroblasts/conditioned medium (Figure 2c). However, expression of cytokeratin 14 was induced significantly by the presence of either NMFs (fold increase: 11.5 ± 9.1, P < 0.01) or CAFs (fold increase: 12.3 ± 9.3, P < 0.01) beneath the filter (Figure 2c, lanes 1 and 2, respectively). Fibroblast-conditioned medium induced expression of cytokeratin 14 similarly, with a significant 10.9 (± 8.8)-fold increase by PMC42-LA in NMF-conditioned medium (P < 0.01), and a 14.4 (± 10.2)-fold increase by cells in CAF-conditioned medium (P < 0.01; Figure 2c, lanes 3 and 4, respectively). Immunocytochemistry confirmed no obvious changes in the organization of cytokeratin 14-expressing cells within organoids in all fibroblast treatment conditions (Figure 2c, lanes 1 to 4).

Compared with control, vimentin expression was significantly increased in cells cultured with NMFs beneath the filter (fold increase: 3.9 ± 2.50, P < 0.001) and in cells with CAFs beneath the filter (fold increase: 4.4 ± 2.70, P < 0.001; Figure 2d, lanes 1 and 2, respectively). In the presence of NMF-conditioned medium, vimentin expression was significantly upregulated by 3.2 (± 1.30)-fold (P < 0.001), which was similar to the 3.3 (± 1.80)-fold (P < 0.001) increase by cells in CAF-conditioned medium (Figure 2d, lanes 3 and 4, respectively) compared with control. In control PMC42-LA organoids, vimentin-positive cells were distributed throughout the organoids with no apparent pattern in organization (Figure 2d, control). However, when cells were cultured with NMFs beneath the filter, vimentin-positive cells became more organized, and were only visible on the outer layer of organoids (Figure 2d, lane 1). This was also the case when cells were cultured with CAFs beneath the filter (Figure 2d, lane 2) and in NMF-conditioned medium (Figure 2d, lane 3). Surprisingly, PMC42-LA cells cultured in CAF-conditioned medium appeared to form organoids with the same organization as control cells, with vimentin-positive cells being dispersed throughout organoids (Figure 2d, lane 4).

In the absence of fibroblast influences, E-cadherin appeared to have a junctional localization in PMC42-LA cells (Figure 3a part A, left and right panels). The presence of either NMFs or CAFs beneath the filter, or their respective conditioned media, induced a change in E-cadherin localization from cell-cell junctions (Figure 3a part A, left and right panels) to cytoplasm, where it adopted a granular distribution (Figure 3a parts B to E). To better visualize this change in localization within individual cells and obtain sections through cells, PMC42-LA monolayers cultured in fibroblast conditioned medium were also immunostained for E-cadherin; confocal sections through these cells clearly demonstrate some junctional and granular cytoplasmic localization in cells cultured in NMF conditioned medium (Figure 3a part B, right panel). Significantly, the E-cadherin localization appeared more cytoplasmic in PMC42-LA cells in the presence of CAF-conditioned medium (Figure 3a part C, right panel).

In the absence of fibroblast-conditioned media, E-cadherin and β-catenin colocalized to the cell junctions in PMC42-LA monolayer cells (yellow in Figure 3b part A), with sparse areas of independent localization (red/green). When PMC42-LA cells were cultured in NMF-conditioned medium, this colocalization was still apparent (yellow in Figure 3b part B), with some areas of non-colocalization (green/red). However, in the presence of CAF-conditioned medium, β-catenin exhibited a predominantly cytoplasmic label, with some nuclear label (red in Figure 3b part C), with E-cadherin also localised to the cytoplasm (green/yellow in Figure 3b part C). This was performed on PMC42-LA monolayer cultures to allow one to visualize clearly the localization of these two proteins within individual cells.

Using total cell counts at 24 hours, 3 days and 5 days, some differences in rates of proliferation of PMC42-LA cells in relation to fibroblast-conditioned medium were seen (data not shown). No significant difference was detected at 24 hours, but at 3 days numbers of PMC42-LA cells in CAF-conditioned medium were considerably greater than numbers of PMC42-LA cells cultured in NMF-conditioned medium. At 5 days, this difference was still evident although not significant. This pattern was also seen in three-dimensional culture, with no difference in PMC42-LA proliferation in NMF-conditioned media or CAF-conditioned media at 24 hours; considerably higher cell numbers in CAF-conditioned media at 3 days; and slightly higher, but not significant, proliferation of PMC42-LA in CAF-conditioned medium at 5 days compared with NMF-conditioned medium (data not shown).

No differences were seen in viability of PMC42-LA cell cultures in the presence of NMFs or CAFs at 24 hours, 3 days and 5 days (Figure 4a), as determined by trypan blue dye exclusion (viable cells calculated as a percentage of total cells counted). Western blot analysis for active caspase-3, a marker of apoptosis, revealed no differences in expression levels for PMC42-LA cells cultured with NMFs beneath the filter as compared with CAFs beneath the filter (densitometry ratio 1:1.2), or in cells cultured in NMF-conditioned medium as compared with those in CAF-conditioned medium (densitometry ratio 1:1.4; Figure 4b).

PMC42-LA cells were cultured on glass to enable the scratch removal of cells and the measurement of wound closure over a period of 48 hours. In the presence of NMF-conditioned medium, PMC42-LA cell wound closure was similar to that in control medium after 48 hours (Figure 5 panels d to f versus panels a to c). Cells cultured in CAF-conditioned medium exhibited rapid wound closure, with most wounds being completely closed by 48 hours (Figure 5 panels g, h and i part ii), visible only by the areas of intense vimentin label among areas of less label (Figure 5i part ii). When the widths of the wounds were measured and averaged, control cells had wounds of approximately 367 ± 29 μm at 0 hours, 188 ± 21 μm at 24 hours and 81 ± 29 μm at 48 hours. Cells in NMF-conditioned medium had wounds of approximately 383 ± 38 μm at 0 hours, 213 ± 33 μm at 24 hours and 52 ± 16 μm at 48 hours, and cells cultured in CAF-conditioned medium had wounds of approximately 392 ± 52 μm at 0 hours, 58 ± 52 μm at 24 hours and 17 ± 29 μm at 48 hours (Figure 5j). Vimentin expression in control cells did not appear changed at 24 hours and was not increased until 48 hours after wound infliction (Figure 5c), whereas cells cultured in the presence of NMF-conditioned medium exhibited a slight increase in vimentin expression by 24 hours (Figure 5e). Cells cultured in CAF-conditioned medium appeared to express more vimentin at 24 and 48 hours (Figure 5h and 5i part i), as compared with control cells and cells cultured in NMF-conditioned medium. In addition to this, some cells in CAF-conditioned medium cultures exhibited an elongated/spindle shape, with some detached cells, neither of which were seen in control or NMF cultures (Figure 5h and 5i part ii).

PMC42-LA cells were cultured on glass to enable the scratch removal of cells and the measurement of wound closure 24 hours after scratch. In the presence of NMF-conditioned medium at a 1× concentration/control, PMC42-LA cell wounds at 24 hours averaged 307 ± 24.29 μm (Figure 6a). Cells cultured in a 2× NMF-conditioned medium had an average wound width of 319 ± 23.15 μm (Figure 6b) at 24 hours. Cells cultured in a 4× NMF-conditioned medium had wounds of 232 ± 49.65 μm (Figure 6c) at 24 hours and cells cultured in a 10× NMF-conditioned medium had wounds of 160 ± 62.43 μm (Figure 6d) 24 hours post-scratch. Vimentin localization did not appear changed in any of the concentrated media compared with control, but PMC42-LA cells cultured in 10× NMF-conditioned medium cultures exhibited an elongated/spindle shape, with some detached cells, neither of which were seen in the less concentrated NMF-conditioned media cultures (Figure 6d versus a to c). Average wound widths of these cultures are displayed graphically (Figure 6e).

When cultured in control three dimensional-culture, PMC42-LA cells formed round organoids, with few or no single cells present in the culture (Figure 7a,b). When cultured in three-dimensional culture with CAF-conditioned medium, PMC42-LA organoids remained predominantly spherical, although the appearance of some budding edges on organoids and the presence of some single cells in the cultures were noted (Figure 7c,d). Furthermore, when PMC42-LA cells were grown in three dimensional culture with CAF-conditioned medium placed as a chemoattractant beneath the filter insert, organoids once again exhibitied budding edges, in addition to many single cells and clusters of single cells present among the organoids (Figure 7e,f).