Fibroblast Hepatocyte Growth Factor Promotes Invasion of Human Mammary Ductal Carcinoma In situ

An altered stroma surrounding a preneoplastic lesion may play a critical role in neoplastic progression and may also affect treatment strategies as has been shown for breast cancer (1, 2). Fibroblasts accumulate during progression of human breast carcinomas, including in preinvasive ductal carcinoma in situ (DCIS) lesions (3). In several tumor types, infiltrating fibroblasts have been shown to undergo alterations in secretion of growth factors, for example, transforming growth factor-β1 and hepatocyte growth factor (HGF; ref. 4). The HGF/c-Met pathway has effects on proliferation, motility, invasion and angiogenesis, linking it to cancer progression (5). Together, HGF and c-Met form a simple paracrine signaling loop linking HGF in stromal cells and c-Met in epithelial cells (6). In breast cancer, staining for HGF and c-Met increases with progression from normal breast/benign hyperplasia to DCIS to invasive carcinoma. Higher levels of HGF and c-Met staining in DCIS are associated with other aggressive tumor markers including comedo histology, high nuclear grade, p53 positivity, and bcl-2 negativity (7). Elevated expression of c-Met at the advancing margins (8, 9) suggests a strong association between c-Met signaling and transition to invasive carcinoma (10, 11). Thus, targeting the HGF/c-Met pathway with small-molecule inhibitors, decoy receptors, receptor antagonists, and humanized antibodies is an area of active investigation (12, 13).

HGF/c-Met signaling regulates downstream proteolytic pathways involved in tumor growth, progression, and extracellular matrix degradation, including urokinase-type plasminogen activator (uPA) and its receptor (uPAR; refs. 14, 15). The role of the plasminogen cascade in invasion and metastasis is well established (for review, see ref. 16); however, uPA and uPAR also support cell migration and invasion by plasmin-independent mechanisms, including interactions between uPA, uPAR, extracellular matrix proteins, integrins, endocytosis receptors, and growth factors such as HGF (17).

Elegant studies by the Bissell and Brugge laboratories have established that growing breast epithelial cells in three-dimensional reconstituted basement membrane (rBM) overlay cultures distinguishes between normal and malignant epithelial cells and identifies pathways that mediate morphogenesis and oncogenesis of normal epithelial cells (for review, see refs. 18, 19). We have shown that incorporation of fibroblasts into three-dimensional rBM cultures can recapitulate effects of the tumor microenvironment on cell growth and progression, including increases in proteolysis (20). Here, we used three-dimensional rBM overlay cultures to mimic DCIS growth and progression in vivo. Fibroblasts expressing HGF and recombinant HGF (rHGF) were able to induce the development of invasive outgrowths from three-dimensional DCIS cultures and to increase their degradation of collagen IV and their secretion of uPA and uPAR. In vivo xenograft studies recapitulated the increase in invasiveness observed in vitro.

MCF-10A human mammary epithelial cells were maintained in DMEM/F-12 (Sigma) supplemented with 5% horse serum (Invitrogen), 5 ng/mL epidermal growth factor (EGF), and 100 μg/mL insulin (Sigma). MCF10.DCIS (21) and SUM102 (22) human mammary DCIS cell lines were maintained in DMEM/F-12 (Sigma) supplemented with 5% horse serum and Ham's F-12 (Sigma) supplemented with 10% fetal bovine serum, respectively. Normal mammary fibroblasts (MF; ref. 23) and normal mammary fibroblasts engineered to secrete HGF (MF:HGF; ref. 24) were maintained in DMEM (Sigma) supplemented with 10% fetal bovine serum.

For homotypic cultures, 6-well plates were coated with 10 mg/mL Cultrex without phenol red (Trevigen) and allowed to solidify for 20 min at 37°C. MCF10.DCIS or SUM102 cells (3.0 × 10⁵) were seeded as single cells onto the solidified rBM. For heterotypic cocultures, 1.5 × 10⁵ fibroblasts were mixed with the rBM before coating the wells. Cultures were grown in M171 Mammary Epithelial Medium (Invitrogen) supplemented with Mammary Epithelial Growth Supplement (Invitrogen) and 2% rBM. Where indicated, concentrated, conditioned medium from fibroblasts, 100 ng/mL rHGF (R&D Systems), 2 μmol/L SU11274 (Calbiochem), or rHGF plus SU11274 were added to culture medium. Medium was replenished every 3 days. The HGF concentration was based on preliminary studies establishing an increase in expression of uPA/uPAR, a response known to occur in parallel with HGF-induced tubulogenesis of MDCK cells (19).

Three-dimensional structures were harvested from rBM by incubation with cold PBS supplemented with 5 mmol/L EDTA, 1 mmol/L NaVO₄, and 1.5 mmol/L NaF on ice for 45 min. Conditioned medium was collected and concentrated 10 times in Ultrafree-0.5 PBGC Centrifugal Filter Units with 10-kDa MWCO Biomax Membranes (Millipore). Cell lysates were prepared in lysis buffer [250 mmol/L sucrose, 25 mmol/L 2-(N-morpholino)ethanesulfonic acid, 1 mmol/L EDTA, 0.1% Triton X-100 (pH 6.5)] and cleared by centrifugation at 14,000 × g for 10 min at 4°C and the supernatant was used for all subsequent procedures. Supernatants and conditioned medium were separated by SDS-PAGE, transferred to nitrocellulose membranes, and immunoblotted with polyclonal goat anti-HGF (R&D Systems), rabbit anti-uPA (Abcam), or rabbit anti-uPAR (Abcam) or monoclonal mouse anti-β-actin (Sigma) or mouse anti–glyceraldehyde-3-phosphate dehydrogenase (Millipore).

Equal numbers of MF or MF:HGF were cultured in M171 medium for 4 days. Conditioned medium was centrifuged at 100 × g to pellet any floating cells and centrifuged again at 800 × g to remove debris. Cleared medium was concentrated to 1/10th of their original volume with a Centriprep Ultracell YM-10 (Millipore), stored at −20°C, and diluted 1:2 with fresh medium before use.

Day 9 cultures were imaged in triplicate for development of invasive outgrowths by differential interference contrast (DIC) imaging on an Ultraview ERS (Perkin-Elmer) using a 10× objective. Invasive outgrowths were defined as consisting of two or more cells migrating away from their structure of origin. A minimum of 10 images was captured and analyzed for each experimental condition.

Assays for quantification of degradation of DQ-collagen IV substrate (Invitrogen) by live cells were done as described previously (25). Where indicated, conditioned medium, 100 ng/mL rHGF, or 1 μmol/L aprotinin (Calbiochem) were added to culture medium. For the overlaid DQ-collagen IV degradation assay, MCF10.DCIS cells were grown in three-dimensional rBM overlay culture on round glass coverslips for 4 days. Coverslips containing preformed structures were then incubated with CellTracker Orange (Invitrogen) and inverted onto rBM containing 25 μg/mL DQ-collagen IV, which had been diluted to 10 mg/mL with fibroblast conditioned medium. These cultures were imaged live on a Leica TCS SP5 confocal microscope with a 20× PL APO N.A. 0.7 objective.

Transwell invasion assays were conducted over 5 days to allow formation of three-dimensional structures. Briefly, 8 μm BioCoat control inserts (BD Biosciences) were coated with 20 μL of 5 mg/mL rBM and the rBM was allowed to polymerize. MCF10.DCIS cells (1.0 × 10⁴) in M171 medium were seeded on rBM-coated inserts and cultured in a 24-well plate in the absence or presence of 1.0 × 10⁴ MF or MF:HGF. Noninvasive cells were removed with a cotton swab, invasive cells were fixed with 3.7% formaldehyde (Polysciences), and nuclei were stained with 4′,6-diamidino-2-phenylindole (Invitrogen). Images obtained with a Zeiss LSM510 META NLO confocal microscope were scored for the number of cells (nuclei) that had invaded by two independent observers.

Three-dimensional rBM overlay cultures on coverslips were fixed with 3.7% formaldehyde, permeabilized with PBS containing 0.1% Triton X-100, blocked with PBS containing 2% bovine serum albumin, and then incubated with PBS containing 0.1% Triton X-100 and 4 units/mL Alexa Fluor 546 phalloidin (Invitrogen). Cultures were imaged on a Zeiss LSM 510 META using an Acroplan 10×/0.3 water immersion objective and LSM AIM software. Z-stack data were imported into Volocity software (Improvision) to generate three-dimensional reconstructions and movies.

Xenografts were generated by injecting 1.0 × 10⁶ MCF10.DCIS cells ± 5.0 × 10⁵ MF or MF:HGF in 0.1 mL Cultrex s.c. at the base of the nipple of gland 5 of female ICRSC-M mice (Taconic Farms). Mice were maintained under aseptic conditions according to the Institutional Animal Care and Use Committee guidelines. Four weeks post-injection, mice were sacrificed and xenografts were removed, fixed with Z-FIX (Anatech), embedded in paraffin, sectioned, and stained with H&E.

To mimic increases in stromal fibroblasts (3) and stromal-derived HGF levels (7), we introduced MF or MF:HGF into three-dimensional MCF10.DCIS rBM overlay cultures. We have established previously by immunoblot analysis of conditioned medium from MCF10.DCIS, MF, and MF:HGF cells that only the MF:HGF cells secreted HGF (105-210 ng/mL; n = 3; data not shown). Over a period of 9 days in either homotypic or heterotypic three-dimensional rBM overlay cultures with MF cells, MCF-10A cells were observed to form acinar structures (Fig. 1A) and MCF10.DCIS cells were observed to form irregular aggregates (Fig. 1A; Supplementary Movies 1 and 2). In heterotypic cultures with MF:HGF cells, MCF-10A acini were unchanged (Fig. 1A), but many of the MCF10.DCIS structures exhibited “invasive outgrowths” consisting of two or more cells migrating into the surrounding extracellular matrix (Fig. 1A; Supplementary Movie 3). To assess the reproducibility of this effect, we scored cultures of MCF10.DCIS ± fibroblasts labeled with cytoplasmic dyes to facilitate discrimination of invasive outgrowths (Fig. 1A). In cocultures of MF:HGF and MCF10.DCIS cells, a significantly higher percentage of three-dimensional structures were found to display invasive outgrowths than did MCF10.DCIS cells alone or MCF10.DCIS cells in coculture with MF cells (Fig. 1B). Thus, the preinvasive DCIS cells, but not the immortalized MCF-10A cells, exhibited a paracrine response to MF:HGF.

To determine whether the increase in invasiveness of MCF10.DCIS cells required direct contact with the fibroblasts, we performed an invasion assay in which the two cell types were separated by a Transwell filter coated with a thick layer of rBM that allowed formation of three-dimensional structures by the MCF10.DCIS cells. The presence of MF:HGF cells, but not MF cells, in the lower chamber resulted in a 2-fold increase in invasion of MCF10.DCIS cells (Fig. 1C), a finding consistent with soluble factors secreted by the MF:HGF cells inducing invasion of MCF10.DCIS cells.

The invasion assays suggested that a soluble factor, potentially pro-HGF secreted by the MF:HGF cells, stimulated MCF10.DCIS invasion. Fibroblast conditioned medium was therefore added to three-dimensional rBM cultures of MCF10.DCIS cells and growth and development of invasive outgrowths was monitored. At 24 h after plating, we observed that fibroblast conditioned medium had stimulated migration of MCF10.DCIS cells; as a result, three-dimensional structures in these cultures originated from clusters of cells rather than from single cells as in the control cultures. Consequently after 9 days, there were fewer three-dimensional structures in cultures incubated with MF or MF:HGF conditioned medium than in control cultures (270 ± 40 and 231 ± 22, respectively, compared with 325 ± 17; mean ± SE). Conditioned medium from the MF:HGF cells induced the formation of large invasive outgrowths from MCF10.DCIS three-dimensional structures (Fig. 2A) and significantly increased (∼3-fold) the percentage of structures that developed invasive outgrowths (Fig. 2B). The observed comparable stimulation of invasive outgrowths by coculture with MF:HGF cells and by addition of MF:HGF conditioned medium was consistent with soluble HGF secreted from the MF:HGF cells being the causative factor.

To examine whether the increased invasion was associated with increased degradation of the basement membrane, we used a live-cell confocal microscopy-based assay to image and quantify degradation of a dye-quenched form of the basement membrane protein collagen type IV (25). Degradation of DQ-collagen IV by MCF10.DCIS cells was significantly increased by MF:HGF conditioned medium (∼2.5-fold; Fig. 2C), an increase similar to that in the percentage of structures with invasive outgrowths (Fig. 2B). In addition, we examined the effect of fibroblast-secreted HGF on DQ-collagen IV degradation by preformed MCF10.DCIS structures. Live imaging revealed a dramatic increase in degradation products surrounding the MCF10.DCIS structures that had been incubated with MF:HGF conditioned medium (Fig. 2D). Thus, HGF-enhanced collagen IV degradation was associated with invasion of preformed structures and not solely the result of proteolysis by growing and migrating DCIS cells.

Binding of HGF, the only known ligand for the c-Met receptor, results in receptor activation and phosphorylation of tyrosine residues within the activation domain (26). Small-molecule inhibitors that compete for ATP binding, such as SU11274, inhibit c-Met phosphorylation (27). To determine if HGF secreted by MF:HGF fibroblasts bound to and activated c-Met on DCIS cells, we collected cell lysates from three-dimensional DCIS cultures treated with MF:HGF conditioned medium ± SU11274. To detect activation of c-Met, we immunoblotted cell lysates using antibodies specific for the double phosphorylation at phosphoepitopes Y1234/Y1235 of c-Met. Following MF:HGF conditioned medium treatment, we observed a sustained phosphorylation of c-Met in both MCF10.DCIS and SUM102 DCIS lines and SU11274 abrogated the phosphorylation (Fig. 3A). SU11274 also significantly decreased the development of invasive outgrowths induced when MCF10.DCIS cells were cocultured with HGF-expressing fibroblasts (Supplementary Fig. S1), consistent with HGF activating c-Met in these cocultures.

HGF treatment in two-dimensional monolayer cultures has been shown to increase expression and secretion of proteases capable of degrading extracellular matrices, including the uPA/uPAR system (15, 28, 29). We therefore determined whether treatment with MF:HGF conditioned medium (shown above to stimulate c-Met phosphorylation) increased the expression and secretion of uPA and uPAR in three-dimensional rBM cultures of MCF10.DCIS and SUM102 cells. Cultures treated with MF:HGF conditioned medium exhibited increased levels of uPA and uPAR in cell lysates and conditioned medium compared with controls and SU11274 abrogated those increases in the MCF10.DCIS cells and partially in the SUM102 cells (Fig. 3B). As uPA and uPAR were not detected in concentrated MF:HGF conditioned medium itself (data not shown), the changes seen in the MF:HGF conditioned medium–treated cultures reflected changes in expression and secretion by MCF10.DCIS and SUM102 cells.

MF:HGF conditioned medium stimulated invasive outgrowths, invasion, and degradation of DQ-collagen IV by MCF10.DCIS cells, consistent with an involvement of HGF. To confirm that HGF acting through c-Met activation was responsible for the increase in invasive outgrowths, we incubated the DCIS cell lines with rHGF and the c-Met inhibitor SU11274. Both MCF10.DCIS and SUM102 cell lines formed dysplastic three-dimensional structures with few structures exhibiting invasive outgrowths in the presence of the DMSO vehicle control or SU11274. Invasive outgrowths were observed in DCIS cell lines treated with rHGF but not treated with both rHGF and the c-Met inhibitor (Fig. 4A). The percentage of structures with invasive outgrowths was significantly increased by rHGF (Fig. 4B). Stimulation of c-Met appeared to be responsible for the increase of invasive outgrowths because rHGF did not increase the percentage of structures with invasive outgrowths in the presence of SU11274 (Fig. 4B). Together, these experiments suggest that MF:HGF within the tumor microenvironment have the potential to increase invasiveness of DCIS lesions and thus their transition to invasive carcinomas.

Because expression of uPA and uPAR was increased in three-dimensional rBM cultures treated with MF:HGF conditioned medium (Fig. 3), we analyzed whether rHGF could elicit similar results. Total cell lysates and concentrated conditioned medium from cultures such as those shown in Fig. 4A were separated by SDS-PAGE and immunoblotted for uPA and uPAR. rHGF increased secretion of uPA and uPAR from DCIS cell lines (Fig. 4C). Immunoblotting of cell lysates with a c-Met phosphorylation-specific antibody confirmed that c-Met was activated by rHGF and inhibited by SU11274 in this series of experiments (data not shown). Our results suggest that HGF-stimulated invasiveness was associated with increases in plasmin-generating potential through an imbalanced uPA/uPAR system (16).

HGF/c-Met–stimulated development of invasive outgrowths may be associated with increased proteolysis initiated by increased uPA secretion. Therefore, we used a live-cell proteolysis assay to image and quantify the degradation of DQ-collagen IV by MCF10.DCIS or SUM102 cells grown for 18 h in the presence of 100 ng/mL rHGF ± 1 μmol/L aprotinin or untreated. We observed degradation products bordering the periphery of DCIS structures as well as intracellularly (Fig. 5A). rHGF increased the intensity of peripheral degradation products associated with MCF10.DCIS and SUM102 three-dimensional cultures (Fig. 5A and B; rHGF). We confirmed that the increase was significant by quantifying degradation products in the entire volume of the three-dimensional cultures and normalizing to the number of cells contributing to degradation in that volume (Fig. 5C; data are average integrated intensity per cell). Aprotinin moderately decreased HGF-induced fluorescence surrounding DCIS structures (Fig. 5A and B; rHGF + A), but this reduction was not significant (Fig. 5C). Our results thus suggest an involvement of more than one family of proteases in the increases in degradation of type IV collagen associated with HGF/c-Met–induced invasive outgrowths in DCIS structures.

To determine whether HGF also affects invasiveness in vivo, we compared xenografts of MCF10.DCIS cells alone or coinjected with MF or MF:HGF cells. Tumor take was similar among the three groups of mice (n = 8 per group), yet the median wet weight of the tumors was greatest in mice coinjected with DCIS cells and MF:HGF (680 mg; range, 70-869 mg) compared with DCIS cells alone (99 mg; range, 66-124 mg) and those coinjected with DCIS cells and MF cells (171 mg; range, 90-502 mg). Strikingly, coinjection of DCIS cells and MF:HGF cells enhanced progression to invasive ductal carcinomas (Fig. 6). These results confirmed our in vitro findings that paracrine HGF/c-Met signaling between fibroblasts and preinvasive DCIS enhances invasiveness.

In this study, we established that DCIS cell lines grown in three-dimensional rBM overlay cultures replicate characteristics of preneoplastic to neoplastic phenotypes that accompany the transition from preinvasive to invasive. We observed that HGF/c-Met signaling increased the invasive phenotype of DCIS cells, including their ability to migrate and to degrade collagen type IV and their expression and secretion of uPA and uPAR. This appears to be generalizable as we obtained comparable results in two unrelated human DCIS cell lines [MCF10.DCIS (21) and SUM102 (22)]. Similar results were obtained in vivo in that MF:HGF enhanced the invasive phenotype of MCF10.DCIS xenografts.

We have shown previously that, when grown in three-dimensional rBM overlay culture, isogenic MCF-10A epithelial variants form an in vitro progression series for analysis of preneoplastic and neoplastic phenotypes (30). In three-dimensional rBM overlay cultures, MCF10.DCIS cells grow into large dysplastic structures (30), a finding recapitulated here for another DCIS cell line: SUM102. Interestingly, the dysplastic structures formed by both DCIS cell lines resemble in morphology and size those depicted for MCF-10A cells that have been transfected with ErbB2 (31). In MCF10.DCIS cells, mitogen-activated protein kinase/extracellular signal-regulated kinase activity is elevated (30), whereas, in SUM102 cells, EGF receptor (EGFR) is elevated and EGFR ligands are secreted (22). ErbB2, mitogen-activated protein kinase/extracellular signal-regulated kinase, and EGFR all have been linked to kidney tubulogenesis (32). Our results are thus consistent with the preneoplastic invasive outgrowths from dysplastic DCIS structures having the same molecular underpinnings as tubular outgrowths formed during normal developmental processes.

Fibroblasts induce formation of capillary-like three-dimensional networks by endothelial cells (33) and branching morphogenesis of mammary epithelial cells (34). Growth and angiogenesis of s.c. MDA-MB-231 human breast carcinoma tumors is increased by coimplantation of fibroblasts, an effect reduced by downregulating HGF in the fibroblasts or c-Met in the carcinoma cells (35). These latter in vivo results are consistent with those reported here for DCIS-fibroblast three-dimensional rBM cocultures and xenografts. We suggest that there may be interaction between c-Met signaling and aberrant ErbB2 signaling in the DCIS cell lines. Such interactions have been shown for MCF-10A cells that express an inducible ErbB2 receptor and are infected with a retrovirus expressing HGF cDNA. These cells develop into large dysplastic structures with enhanced protrusive behavior (36), resembling the ones formed by the two DCIS lines in response to HGF. Our results would thus be consistent with ErbB2-related alterations in MCF10.DCIS and SUM102 cells interacting with HGF/c-Met signaling to promote development of large dysplastic structures and invasive outgrowths. Interestingly, HGF/c-Met signaling also may contribute to the intrinsic resistance to EGFR tyrosine kinase inhibitors seen in breast cancers. In cooperation with c-Src, HGF/c-Met signaling has been shown to transactivate EGFR even in the presence of EGFR tyrosine kinase inhibitors (37).

The accumulation of fibroblasts and increased HGF found in DCIS lesions in vivo in parallel with losses in basement membrane suggest a relationship for fibroblasts and HGF in progression of DCIS from preinvasive to invasive. The three-dimensional rBM overlay culture and coculture models for DCIS replicated the morphology and progression of human DCIS in vivo and allowed us to show that paracrine HGF/c-Met signaling stimulated invasiveness, upregulated expression and secretion of uPA/uPAR, and increased degradation of collagen IV. We suggest that the three-dimensional DCIS rBM overlay cultures/cocultures will be useful for analyzing the contributions of stromal-derived factors to DCIS progression and for screening therapeutics that target this progression, for example, agents that target HGF/c-Met signaling. We are presently unable to distinguish DCIS lesions that will rapidly progress to invasive carcinomas from those that will not, thus subjecting many women to unnecessary biopsies and others who should be treated to extended periods of monitoring during which their disease may progress. This is of clinical relevance as ∼30% of newly diagnosed breast cancers are found at the DCIS stage (21).

No potential conflicts of interest were disclosed.

Grant support: R01 CA56586 and DOD BC013005 (B.F. Sloane) and DOD BC051230 (C. Jedeszko). Imaging was done at the Microscopy and Imaging Resources Laboratory supported in part by NIH grants P30 CA22453 and P30 ES 06639 and by U54 RR02084330.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Drs. Fred Miller and Stephen Ethier (Karmanos Cancer Institute), Kornelia Polyak (Dana-Farber Cancer Institute), and Charlotte Kuperwasser (Tufts University) for the kind gifts of cell lines and Bruce Linebaugh and Mackenzie Herroon for assistance with xenograft studies.