Hepatocyte growth factor stimulates extensive development of ...

413

Journal of Cell Science 108, 413-430 (1995) Printed in Great Britain © The Company of Biologists Limited 1995

Hepatocyte growth factor stimulates extensive development of branching duct-like structures by cloned mammary gland epithelial cells J. V. Soriano1, M. S. Pepper1, T. Nakamura2, L. Orci1 and R. Montesano1,* 1Department of Morphology, University of Geneva Medical School, 1 rue Michel-Servet, CH-1211 Geneva 4, Switzerland 2Division of Biochemistry, Biomedical Research Center, Osaka University Medical School, Suita, Osaka 565, Japan

*Author for correspondence

SUMMARY Although epithelial-mesenchymal (stromal) interactions are thought to play an important role in embryonic and postnatal development of the mammary gland, the underlying mechanisms are still poorly understood. To address this issue, we assessed the effect of fibroblast-derived diffusible factors on the growth and morphogenetic properties of a clonally derived subpopulation (clone TAC-2) of normal murine mammary gland (NMuMG) epithelial cells embedded in collagen gels. Under control conditions, TAC2 mammary gland epithelial cells suspended within collagen gels formed either irregularly shaped cell aggregates or short branching cord-like structures. Addition of conditioned medium from Swiss 3T3 or MRC-5 fibroblasts dramatically stimulated cord formation by TAC-2 cells, resulting in the development of an extensive, highly arborized system of duct-like structures, which in appropriate sections were seen to contain a central lumen. The effect of fibroblast conditioned medium was completely abrogated by antibodies against hepatocyte growth factor (also known as scatter factor), a fibroblast-derived polypeptide that we have previously shown induces tubu-

logenesis by Madin-Darby canine kidney epithelial cells. Addition of exogenous recombinant human hepatocyte growth factor to collagen gel cultures of TAC-2 cells mimicked the tubulogenic activity of fibroblast conditioned medium by stimulating formation of branching duct-like structures in a dose-dependent manner, with a maximal 77fold increase in cord length at 20 ng/ml. The effect of either fibroblast conditioned medium or hepatocyte growth factor was markedly potentiated by the simultaneous addition of hydrocortisone (1 µg/ml), which also enhanced lumen formation. These results demonstrate that hepatocyte growth factor promotes the formation of branching ductlike structures by mammary gland epithelial cells in vitro, and suggest that it may act as a mediator of the inducing effect of mesenchyme (or stroma) on mammary gland development.

INTRODUCTION

postnatal life) on mammary gland development (Kratochwil, 1969; Donjacour and Cunha, 1990; Sakakura, 1991) is likely to be mediated in part by extracellular matrix components, as suggested by the ability of reconstituted matrices to promote alveolar morphogenesis and differentiation in vitro (Barcellos Hoff et al., 1989; Aggeler et al., 1991). However, the finding that culture of mammary gland epithelial cells in collagen gels allows only limited growth and branching of duct-like structures (Yang et al., 1980; Bennett, 1980; Ormerod and Rudland, 1982; Danielson et al., 1984; Reichmann et al., 1989) raises the possibility that additional signals from living stromal cells are required to support formation of an extensive ductal tree in vitro. The observation that conditioned medium from mammary fibroblasts enhances the proliferation of mammary epithelial cells in monolayer culture (Enami et al., 1983) also supports a role for paracrine stromal-epithelial interactions in mammary gland development. To date, however, soluble stromal factors that stimulate duct formation and branching have not been identified.

Embryonic and postnatal development of the rodent mammary gland involves a precise sequence of morphogenetic events that result in the formation of an arborized system of epithelial ducts embedded in a connective tissue stroma. During pregnancy, ductal elongation and branching resume, and clusters of alveoli subsequently bud off from the growing ducts (reviewed by Daniel and Silberstein, 1987). These processes are thought to be driven by a complex interplay of both systemic and local regulatory factors, including circulating hormones and epithelial-mesenchymal (stromal) interactions (Daniel and Silberstein, 1987; Borellini and Oka, 1989; Donjacour and Cunha, 1990; Sakakura, 1991; Haslam, 1991). While the multihormonal requirements for mammary gland development have been defined in considerable detail (Topper and Freeman, 1980; Imagawa et al., 1990), the role of the stroma is less well understood. The inducing effect of mesenchyme (during embryonic development) or stroma (during

Key words: scatter factor, morphogenesis, extracellular matrix, corticosteroid hormone, growth factor, epithelial-mesenchymal interaction, c-met protooncogene

414

J. V. Soriano and others

Hepatocyte growth factor (HGF) (Nakamura et al., 1989) is a pleiotropic cytokine, which was originally identified in the serum of partially hepatectomized rats as a potent mitogen for cultured hepatocytes (Nakamura et al., 1984; Michalopoulos et al., 1984), and later shown to promote the growth of a broad spectrum of epithelial cells and other cell types (Rubin et al., 1991; Kan et al., 1991; Matsumoto and Nakamura, 1993). It is identical to scatter factor, an independently isolated and characterized fibroblast-derived protein which induces dispersion and migration of epithelial cells (Stoker et al., 1987; Gherardi et al., 1989; Weidner et al., 1991; Naldini et al., 1991a; Furlong et al., 1991; Konishi et al., 1991; Bhargava et al., 1992). HGF is thought to elicit its various biological activities by binding to a membrane-spanning tyrosine kinase receptor encoded by the c-met protooncogene (Bottaro et al., 1991; Naldini et al., 1991b; Higuchi et al., 1992; Giordano et al., 1993; Weidner et al., 1993) expressed by epithelial cells (Chan et al., 1988; Iyer et al., 1990; Prat et al., 1991; Di Renzo et al., 1991; Tsarfaty et al., 1992; Sonnenberg et al., 1993). We have recently reported that Madin-Darby canine kidney (MDCK) epithelial cells form branching tubules when grown in collagen gels in the presence of fibroblasts or fibroblast conditioned medium (CM) (Montesano et al., 1991a), and that the fibroblast-derived factor responsible for epithelial tubulogenesis is HGF (Montesano et al., 1991b). On the basis of these findings, we proposed that HGF may act as a paracrine mediator of morphogenetic epithelial-mesenchymal interactions during the development of parenchymal organs (Montesano et al., 1991b), a hypothesis supported by recent in situ hybridization experiments (Sonnenberg et al., 1993). The present study was undertaken to determine whether diffusible factors released by fibroblasts could promote formation of duct-like structures by mammary gland epithelial cells embedded in collagen gels and, if so, to ascertain whether this effect could be mediated by HGF. To address this question, we have utilized the established normal murine mammary gland (NMuMG) epithelial cell line (Owens et al., 1974), which has previously been reported to form cavitary structures when sandwiched between two collagen layers (Hall et al., 1982) and to undergo gland-like morphogenesis in vivo (David et al., 1981). We demonstrate here that fibroblast CM induces the formation of a very extensive system of branching tubules by a clonally derived subpopulation of NMuMG epithelial cells, and that this effect is mediated by HGF. These results suggest that HGF is an important stromal mediator of mammary gland development. MATERIALS AND METHODS Reagents Recombinant human HGF (rhHGF) and recombinant rat HGF (rrHGF) were purified as described (Nakamura et al., 1989) from culture medium of CHO cells transfected with a plasmid containing a human or rat HGF cDNA. Natural mouse epidermal growth factor (EGF) was purchased from Collaborative Research Inc. (Bedford, MA). Human recombinant epidermal growth factor (rhEGF) and porcine platelet-derived growth factor (PDGF) were from Boehringer Mannheim (Rotkreuz, Switzerland), transforming growth factor-β1 (TGF-β1) from R&D Systems Europe (Oxon, UK), and transforming growth factor-α (TGF-α) from Bachem (Bubendorf, Switzerland). Human recombinant basic fibroblast growth factor (bFGF), rat

insulin-like growth factor-II (IGF-II), nerve growth factor (NGF), and human recombinant keratinocyte growth factor (KGF) were kindly provided by Dr P. Sarmientos (Farmitalia Carlo Erba, Milan, Italy), Dr N. Yanaihara (Laboratory of Bioorganic Chemistry, Shizuoka-shi, Japan), Dr L. Aloe (Consiglio Nazionale Ricerche, Rome, Italy) and Drs J. Rubin and S. Aaronson (Laboratory of Cellular and Molecular Biology, NIH, Bethesda, MD), respectively. Rabbit polyclonal antiserum against rhHGF and anti-rhHGF IgGs, produced as previously described (Montesano et al., 1991b), were generously provided by Dr K. Matsumoto (Biomedical Research Center, Osaka, Japan). Rabbit polyclonal antiserum directed against rat proalbumin decapeptide was kindly provided by Drs K. Davidson and T. Peters (Cooperstown, NY). Rabbit polyclonal antiserum against secretin was a gift from Dr R. S. Yalow (V.A. Medical Center, Bronx, NY). Rabbit IgGs against mouse IgG fraction were from Cappel Laboratories (Philadelphia). Type I collagenase (from Clostridium histolyticum) was purchased from Worthington Biochemical Corporation (Freehold, NJ). Laminin and type IV collagen were from Bethesda Research Laboratories (Gaithersburg, MD), and human plasma fibronectin from Collaborative Research. All hormones used in this study, as well as cis-hydroxy-proline and L-proline were purchased from Sigma Chemical Co. (St Louis, MO). Cells NMuMG cells (CRL 1636) (Owens et al., 1974) were purchased from the American Type Culture Collection (ATCC, Rockville, MD) and routinely grown in tissue culture flasks (Falcon, Becton-Dickinson and Co., San José, CA) in high glucose Dulbecco’s modified Eagle’s medium (DMEM, GIBCO, Basel, Switzerland) supplemented with 10% fetal calf serum (FCS) (Flow Laboratories, Baar, Switzerland). The NMuMG TAC-2 clone was established as described below and cultured in DMEM supplemented with 10% FCS in collagen-coated flasks. Swiss 3T3 (ATCC, CCL 92) mouse embryo fibroblasts and MRC-5 (ATCC, CCL 171) human embryonic lung fibroblasts were cultured according to the instructions provided in the ATCC Catalogue of Cell Lines and Hybridomas (7th edition, Rockville, MD, 1992). All culture media were supplemented with penicillin (500 i.u./ml) and streptomycin (100 µg/ml). Collagen gel cultures Parental NMuMG or clonal TAC-2 cells (see below) were harvested using trypsin-EDTA, centrifuged, and embedded in three-dimensional collagen gels as described (Montesano et al., 1983, 1991a). In brief, 8 volumes of rat tail tendon collagen stock solution (approximately 1.5 mg/ml) were mixed with 1 volume of 10× concentrated minimal essential medium (Gibco) and 1 volume of sodium bicarbonate (11.76 mg/ml) in a sterile flask kept on ice to prevent premature collagen gelation. Cells were resuspended in the cold mixture, and either 400 or 1500 ml aliquots of cell suspension were dispensed into 16-mm wells or 35-mm dishes (Nunc, Kampstrup, Roskilde, Denmark), respectively. After the collagen solution had gelled, 1.5 ml of complete medium (DMEM + 10% FCS) was added to each dish or well. Media were changed every 2-3 days, and the cultures incubated at 37°C for the times indicated. Cocultures with fibroblasts were prepared as described (Montesano et al., 1991a) by suspending NMuMG or TAC-2 cells within a collagen gel cast on top of a preformed gel layer containing Swiss 3T3 cells. To prevent contact between the upper and lower cell populations, a cell-free gel layer was interposed between the two cellcontaining collagen gels. Establishment of the TAC-2 clone Clone TAC-2 was established from a single colony of epithelial cells formed in a collagen gel coculture of NMuMG cells and Swiss 3T3 fibroblasts, as follows. Swiss 3T3 cells were seeded into 35-mm plastic dishes at 2×105 cells/dish. After overnight attachment and spreading, the fibroblasts were treated with 10 µg/ml mitomycin C

HGF and mammary gland morphogenesis (Sigma) for 4 hours, washed three times with PBS, and overlaid with a cell-free collagen gel (1 ml). A second collagen layer (700 µl) containing a suspension of NMuMG cells (100 cells/ml) was then layered on top of the cell-free gel. After 10 days of coculture, a small piece of collagen gel containing an individual NMuMG colony with a tubulo-alveolar organization (see Results) was manually removed under sterile conditions with the aid of fine needles, and subsequently digested by incubation with collagenase (4 mg/ml) at 37°C for 10 minutes. The released epithelial colony was then dissociated into single cells with trypsin-EDTA. To assess the morphogenetic properties of the cells thus recovered, the procedure described above was repeated by suspending the isolated epithelial cells in a collagen gel cast on top of mitomycin C-treated Swiss 3T3 cells. After 10 days of coculture, this resulted in the development of numerous colonies having a similar tubulo-alveolar organization. The collagen gel containing the tubulo-alveolar colonies was then digested with collagenase, the released colonies dissociated with trypsin-EDTA and the resulting cell suspension plated in a collagen-coated (see below) 16mm well. After 5 days, when the culture had attained confluence, the cells were trypsinized, expanded by successive passages in collagencoated dishes, and subjected to a second cycle of cloning by the same coculture procedure described above. One of the clones thus obtained was subcultured in collagen-coated flasks and used between passages 7 and 23. This clone will be referred to as TAC-2 (for ‘tubuloalveolar’ colony-2). For collagen coating, the rat tail tendon collagen solution described above was diluted to 300 µg/ml in cold sterile bidistilled water and poured into plastic culture flasks (approximately 100 µl/cm2). After 5-10 minutes incubation at 37°C, the collagen solution was aspirated and the flasks washed with PBS. Quantification of cord length and branching TAC-2 cells were suspended at 5×103 cells/ml in collagen gels (1.5 ml) cast into 35-mm dishes or 22-mm wells of 12-well plates (tissue culture Cluster, Costar, Cambridge, MA) and incubated with either control medium, fibroblast conditioned medium (CM) (prepared as described by Montesano et al., 1991a), or the indicated agents. After 7 or 9 days, the cultures were fixed as described below, and at least 30 randomly selected colonies per experimental condition in each of at least three separate experiments were photographed under transmitted light in a Nikon Diaphot TMD inverted photomicroscope, by focusing at the level of the major axis of each colony. The total length of all epithelial cords in each colony was measured on a graphic tablet (Tektronix, 4953 MOD AA) connected to a XT IBM computer programmed for the semiautomatic evaluation of individual mean parameters. Cord length was considered as 0 in: (a) colonies showing a cyst-like shape; and (b) structures in which the length to diameter ratio was less than 2. Quantification of branching was performed by counting all identifiable branch points in each colony on positive prints. Values of cord length and branching obtained from the largest colonies are underestimates, since in these colonies a considerable proportion of cords were out of focus and therefore could not be measured. The mean values for each experimental condition were compared with controls using Student’s unpaired t-test. Values of cord length in cultures treated with MRC-5 or Swiss 3T3 CM were also compared with controls by applying the Kolmogorov-Smirnov twosample test to the cumulative frequency distribution of each sample. Proliferation assays To measure proliferation in collagen-coated dishes, TAC-2 cells were seeded at 50 or 100 cells/cm2 in 16-mm wells and allowed to attach at 37°C for 2 hours, at which time medium was aspirated and replaced with fresh medium with or without the indicated treatment. Culture media and treatments were renewed every 2-3 days. After 9 days, cells were harvested with trypsin-EDTA and counted with a hemocytometer. To measure proliferation within three-dimensional collagen gels, TAC-2 cells were suspended at 5×103 cells/ml in a gelling collagen

415

solution (400 µl) cast into 16-mm wells. After the collagen had gelled, 500 µl of complete culture medium (DMEM + 10% FCS) was added to each well, with or without the indicated growth factors. Medium and growth factors were changed every 2-3 days. After 9 days, the collagen gels were removed from the wells and digested by incubation with collagenase (4 mg/ml) at 37°C for 15 minutes. The released cell clumps were recovered by centrifugation and dissociated by incubation with trypsin-EDTA. The isolated cells thus obtained were counted with a hemocytometer. To measure proliferation on three-dimensional collagen gels, TAC2 cells were seeded at 100 cells/cm2 on the surface of collagen gels in 16-mm wells and allowed to attach at 37°C for 2 hours, at which time medium was aspirated and replaced with fresh medium with or without the indicated treatment. Culture media and treatments were renewed every 2-3 days. After 9 days, cells were harvested with trypsin-EDTA and counted with a hemocytometer. All data are expressed as mean ± s.e.m. and are compared by Student’s unpaired t-test. Collagen binding assay Collagen-coated dishes were prepared by incubating 60-mm bacteriological plastic dishes (Falcon, cat. no. 1016) with 3 ml of a solution of rat tail tendon collagen (prepared as described above and diluted to approximately 50 µg/ml in bidistilled water) for 20 minutes at 37°C, at which time the collagen solution was aspirated and the dishes washed three times with PBS. The dishes were subsequently incubated with 0.5% BSA in PBS for 2 hours at 37°C to saturate nonspecific cell binding sites, and washed again three times with PBS. In control dishes, the collagen coating step was omitted. TAC-2 cells were harvested with trypsin-EDTA from confluent stock cultures following a 48 hour pre-incubation in the presence or the absence of 10 ng/ml rhHGF, centrifuged and resuspended in serum-free DMEM supplemented with 0.5% BSA (DMEM-BSA). The cells were then seeded into triplicate collagen-coated bacteriological dishes prepared as described above at a concentration of 2×104 cells/ml (4 ml of cell suspension per dish). After a 30 minute incubation at 37°C, the medium was aspirated, the dishes gently washed two times with DMEM-BSA, and the adherent cells fixed in 2.5% glutaraldehyde in PBS. Ten randomly selected fields were photographed in each dish using a Nikon Diaphot TMD inverted photomicroscope and a 4× objective, and the number of attached cells per field was counted. Values are expressed as mean number of cells/cm2 ± s.e.m. and are compared by Student’s unpaired t-test. Processing for light and electron microscopy Collagen gel cultures were fixed in situ overnight with 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4). After extensive rinsing in the same buffer, the collagen gels were cut into 3 ×3 mm fragments. These were postfixed in 1% osmium tetroxide in Veronal-acetate buffer for 45 minutes and processed as described (Montesano et al., 1991a). Semi-thin (1 µm) sections were cut with an LKB ultramicrotome, stained with 1% methylene blue, and photographed under transmitted light using an Axiophot photomicroscope (Zeiss, Germany). Thin sections were stained with uranyl acetate and lead citrate, and examined with a Philips EM 300 electron microscope. Immunofluorescence microscopy Collagen gel cultures were removed from the culture dishes and fixed overnight with 4% paraformaldehyde in PBS. After extensive rinsing in the same buffer, the collagen gels were incubated in 15% sucrose in PBS for at least 15 hours and frozen in liquid nitrogen-cooled methylbutane. 4-6 µm-thick sections were cut with a Jung Cryocut 3000 cryostat (Leica Cambridge Ltd., Cambridge, England) and stained by indirect immunofluorescence. Briefly, sections were incubated for 2 hours at room temperature with either sheep antiserum to mouse type IV collagen (1:200 dilution), rabbit antiserum to mouse

416

J. V. Soriano and others

laminin (1:50 dilution) (both gifts from Drs H. Kleinman and G. Martin, Bethesda) or rabbit antiserum to mouse entactin (1:50 dilution) (Carlin et al., 1981; gift from Dr A. Chung, Pittsburgh). The sections were then washed in PBS, further incubated for 1 hour with either a 1:200 dilution of FITC-conjugated rabbit IgGs anti-sheep IgGs (Byosis, Compiègne, France) (for anti-type IV collagen) or a 1:400 dilution of FITC-conjugated goat IgGs anti-rabbit IgGs (Byosis) (for anti-entactin and anti-laminin) before counterstaining with 0.03% Evans Blue. After extensive washing in PBS, the sections were mounted with a glass coverslip in PBS-glycerol (1:2, v/v) containing 0.02% paraphenylenediamine and photographed in a MRC600 laser-scanning confocal imaging system (Bio-Rad Laboratories, Richmond, CA) connected to a Zeiss Axiophot photomicroscope. The percentage of colony perimeter occupied by immunoreactive material was determined on at least 15 colonies per experimental condition using a graphic tablet (Tektronix) connected to a XT IBM computer. Controls included exposure of sections during the first incubation to one of the following reagents: (1) non-immune goat, sheep or lamb serum; (2) preimmune rabbit serum; (3) an unrelated (anti-insulin) antiserum; and (4) the FITC-conjugated antibodies normally used during the second incubation step. None of these control incubations resulted in specific staining of peritubular collagen matrix. RNA extraction and northern blot hybridization Total cellular RNA was extracted from subclonfluent monolayers of TAC-2 cells according to a modification of the acid guanidine-phenolchloroform method of Chirgwin et al. (1979) as described by Chomczynski and Sacchi (1987). RNA was denatured with glyoxal, electrophoresed in a 1% agarose gel (20 µg RNA per lane), and transferred overnight onto nylon membranes (Hybond, Amersham) as described by Thomas (1980). RNAs were crosslinked by exposure of filters to UV light (302 nm), and stained with methylene blue to assess 18 S and 28 S ribosomal RNA integrity. Filters were prehybridized for 2 hours at 65°C and hybridized at 65°C with 2×106cpm/ml of 32Plabelled mouse c-met, α1(IV) collagen, laminin A, B1 or B2 chain cRNA probes for 18 hours as described (Busso et al., 1986). The filters were washed twice at 65°C with 3× SSC (1× SSC = 0.15 M NaCl, 0.015 M Na citrate, pH 7.0), 2× Denhardt’s solution (Maniatis et al., 1982), and three times at 70°C with 0.2× SSC, 0.1% SDS and 0.1% sodium pyrophosphate. As an internal control for determination of the amount of RNA loaded, the filters were hybridized simultaneously with a 32P-labelled chicken glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cRNA probe. Filters were exposed to Kodak XAR-5 films at room temperature or at −80ºC between intensifying screens. Autoradiographs were scanned with a GenoScan laser scanner (Genofit, Geneva, Switzerland); mRNA levels were normalized relative to GAPDH mRNA in the same samples, and expressed relative to control untreated cultures. The median values for each condition were compared with controls using the median test, and a significant value was taken as P≤0.01. Ribonuclease protection assay RNase protection assays were performed using an RPAII kit from Ambion Inc. (Austin, TX). A 20 µg sample of total cellular RNA prepared from rat tissues as described above was hybridized with 32Plabelled pSP64rHGF5′ and pRcMET#7 cRNA probes (see below) for 15 hours at 45°C. Hybridization, RNase digestion and detection of protected fragments by polyacrylamide gel electrophoresis were performed according to the manufacturers’ instructions. Dried gels were exposed to Kodak XAR-5 films at −80°C between intensifying screens. Plasmid construction and in vitro transcription pSP65mmet was constructed by subcloning fragment B, a 2.1 kilobase pair mouse c-met cDNA derived from NIH3T3 cells (Chan et al., 1988) into the EcoRI site of pSP65 (Melton et al., 1984). pSP64cGAPDH was constructed by subcloning a 1.1 kilobase pair

chicken muscle GAPDH cDNA (Dugaiczyk et al., 1983), into the PstI site of pSP64 (Melton et al., 1984). pRcMET#7 was constructed by subcloning a 462 base pair rat c-met cDNA into pBluescript KS−. The rat c-met cDNA was amplified by PCR from IEC-6 (ATCC, CRL 1592) rat intestinal epithelial cell total cellular mRNA with degenerate primers designed from conserved regions in human and mouse cmet cDNA sequences (M.S. Pepper et al., unpublished data). pSP64rHGF5′ was constructed by subcloning a 336 base pair HincIIEcoRI fragment of clone pRBC1, a 1.4 kilobase pair rat HGF cDNA derived from rat liver exposed to CCl4 (Tashiro et al., 1990), into the HincII-EcoRI sites of pSP64 (Melton et al., 1984). pmLB1 was constructed by subcloning a 920 base pair BglII-EcoRI fragment isolated from pPE49, a plasmid containing a 1.1 kb mouse laminin B1 chain cDNA (Barlow et al., 1984) between the EcoRI and BamHI sites of pSP65 (Melton et al., 1984). pmLB2 was constructed by subcloning a 675 base pair EcoRI-PstI fragment isolated from pPE49, a plasmid containing a 675 base pair mouse laminin B2 chain cDNA (Barlow et al., 1984) into pSP64 (Melton et al., 1984). p1238 was constructed by subcloning a 1.15 kb EcoRI-SSTI fragment of mouse laminin A chain cDNA (Sasaki et al., 1988) into pGEM2. pmα1(IV) was constructed by subcloning a 800 base pair BamHI-HindIII fragment isolated from pFAC, a plasmid containing a 2.1 kb mouse α1(IV) collagen cDNA (Oberbäumer et al., 1985) into pSP64 (Melton et al., 1984).

RESULTS Morphogenetic properties of NMuMG epithelial cells embedded in collagen gels The objective of our initial experiments was twofold: firstly, to investigate the morphogenetic properties of NMuMG epithelial cells suspended within three-dimensional collagen gels; and secondly, to assess whether these properties might be influenced by coculture with fibroblasts. Culture of NMuMG epithelial cells in collagen gels resulted, after 7-10 days, in the formation of different types of colonies. These included essentially: (a) thin branching cords apparently devoid of lumen; (b) thick, stubby epithelial cords, also devoid of a visible lumen; (c) small, irregularly shaped cystic structures, and (d) bundles of elongated cells with a fibroblastoid morphology (not shown). These observations, together with the heterogeneous morphology of confluent monolayer cultures (Fig. 1A; see also Hall et al., 1982), strongly suggested that the NMuMG strain is composed of different cell populations. To examine the effect of coculture with fibroblasts, NMuMG cells were suspended within a collagen gel cast on top of a gel layer containing Swiss 3T3 cells, as described in Materials and Methods. Under these conditions, the various types of colonies described above developed more rapidly and, over the same time period, reached a greater size than colonies formed in control cultures (i.e. in the absence of fibroblasts). The most striking finding in cocultures, however, was the development of an additional, peculiar type of colony showing a high degree of structural organization. These colonies, which were not identified in control cultures, consisted of a central alveolar-like cavity extending into radially disposed tubular structures, and will therefore be referred to as tubulo-alveolar colonies. Both the alveolar and tubular portions of these cavitary structures were delimited by a uniformly thick epithelial wall (Fig. 1B). The occurrence of this type of colony in cocultures suggested that the NMuMG strain contained a subpopulation of cells endowed with the ability to form highly

HGF and mammary gland morphogenesis

417

organized duct-like and/or alveolar-like structures in response to fibroblast-derived soluble factors. Since initial attempts to isolate this subpopulation of epithelial cells by classical cloning procedures (i.e. limiting dilution or ring cloning) were unsuccessful, we followed the strategy of manually removing single tubulo-alveolar colonies from collagen gel cocultures of NMuMG and Swiss 3T3 cells. Cells obtained by enzymatic dissociation of the isolated colonies attached poorly and failed to spread when seeded into conventional plastic tissue culture dishes or plastic dishes which had been previously coated with either gelatin, fibronectin, laminin or type IV collagen; however, they attached and spread on dishes coated with type I collagen. A clonally derived cell population (clone TAC-2) established from a single tubulo-alveolar colony as described in Materials and Methods was subcultured in type I collagen coated flasks and used in subsequent experiments. TAC-2 cells exhibited an apparently homogeneous polygonal morphology in confluent monolayer cultures (Fig. 1C).

Fig. 1. Establishment of the TAC-2 clone. (A) A confluent monolayer culture of parental NMuMG cells shows a heterogeneous morphology (phase-contrast microscopy). (B) Tubulo-alveolar colony formed by parental NMuMG cells grown for 7 days in a collagen gel cast on top of a Swiss 3T3 fibroblast-containing collagen gel layer (bright field illumination). The colony consists of a central alveolar-like cavity extending into radially disposed tubular structures (arrows). Notice that both the alveolar and tubular portions of the colony are delimited by a uniformly thick epithelial wall. (C) Morphology of the TAC-2 clone established from a single tubulo-alveolar colony similar to that shown in B (phase-contrast microscopy). The monolayer culture is composed of apparently homogeneous polygonal cells. Bars, 100 µm.

Fibroblast conditioned medium stimulates formation of branching duct-like structures by TAC-2 cells When suspended in collagen gels under control conditions (i.e. in the absence of fibroblasts), TAC-2 cells gave rise, within 710 days, to small slowly growing colonies with a morphology ranging from irregularly shaped cell aggregates to poorly branched structures (see Fig. 2A). In contrast, in cocultures with Swiss 3T3 fibroblasts, TAC-2 cells formed thick branching cords which underwent progressive multifocal cavitation. After 9-10 days, coalescence of the focal lumina resulted in the development of tubulo-alveolar structures delimited by a thick epithelial wall (data not shown). To assess whether the effect of coculture was mediated by fibroblast-derived diffusible factors, TAC-2 cells were suspended in collagen gels and incubated with conditioned medium (CM) from either Swiss 3T3 or MRC-5 cells. This experimental condition partially mimicked the effect of coculture by markedly stimulating the growth of epithelial colonies in collagen gels. Contrary to cocultures, however, fibroblast CM did not induce formation of thick-walled tubuloalveolar structures with patent lumina, but promoted instead the development of an extensive network of branching cords (Fig. 2B). A quantitative analysis revealed that Swiss 3T3 and MRC-5 CM (50%, v/v) induced a 276-fold or 210-fold increase, respectively, in mean total additive cord length per colony (Table 1). A significant (P