ABSTRACT
The mammary gland is a renewing tissue in which morphogenetic processes and differentiation occur cyclically during the menstrual cycle, pregnancy and lactation. These events have been shown to be dependent upon epithelial-mesenchymal interactions. Studies of the effects of individual factors, their cellular source and their target cell populations in the different developmental stages of the mammary gland are greatly facilitated by the accessibility of this organ and the application of new techniques that allow purification of the major epithelial and stromal components of this tissue.
Here we demonstrate that HGF/SF and its cellular receptor, c-met, are expressed and regulated temporally during mouse mammary development and differentiation. We show that human and mouse mammary fibroblasts produce HGF/SF and that HGF/SF is not only mitogenic but morphogenic and motogenic for both human and mouse mammary epithelial cells. We have found that human luminal and myoepithelial cells express c-met differentially and that HGF/SF has different effects on these two mammary epithelial cell populations. HGF/SF is mitogenic for luminal cells but not myoepithelial cells, and morphogenic to myoepithelial cells but not luminal cells. This is discussed in the context of the proliferative compartments in the normal mammary gland and the potential role of the myoepithelial cells to act as the skeleton for ductal development.
INTRODUCTION
The mammary gland is a unique organ since growth, morphogenesis, cell diversification and full phenotypic differentiation occurs in several stages during postnatal and adult life (Sakakura, 1987). In both rodent and human the rudimentary ducts elongate extensively and through dichotomous branching invade the mammary fat pad. In rodents the result of this intensive growth and branching morphogenesis is a tree-like ductal network which fills the adipose-rich stroma (Williams and Daniel, 1983). Although less is known about the human gland, similar stages of morphogenetic change appear to occur (Anbazhagan et al., 1991; Atherton et al., 1994a).
Mammary ductal elongation is the direct result of proliferative activity in the mammary stroma and a highly specialised region of the duct called the ‘end bud’ (Berger and Daniel, 1983; Bresciani, 1965; Russo and Russo, 1978, 1980; Anbazhagan et al., 1991; Atherton et al., 1994a). The advancing edge of the end buds appear to be specialised for penetration of the surrounding fatty stroma (Silberstein and Daniel, 1982) while the posterior region of the end bud provides a supply of differentiating luminal and myoepithelial cells for elongation and subtending ducts.
During early mouse pregnancy, lateral (‘alveolar’) buds appear. From these lateral buds, alveoli develop through a process of rapid growth and morphogenesis. The alveoli organise into lobular structures in which the epithelia assumes a secretory function which will be continued during lactation (Cole, 1933; Russo and Russo, 1978, 1980). Similar changes are thought to occur in the human breast but there is very little data available. In the adult human mammary gland, cyclic patterns of proliferation, morphogenesis, differentiation and regression occur during the estrous cycle (Anderson et al., 1982). In both human and murine mammary gland, pregnancy and lactation results in a dramatic cyclic tissue remodelling.
Inductive interactions between epithelium and mesenchyme are essential for growth and differentiation of the mammary gland during embryogenesis, and continue to play a critical role in the growth and differentiation of this organ in the postnatal animal and throughout adult life (Sakakura et al., 1979; Berger and Daniel, 1983; Daniel and Silberstein, 1987). The mammary gland is therefore a suitable system for identification and characterisation of the mediators of these processes. During normal human fetal breast development, there is an intimate relationship between the condensed specialised stroma and the epithelial buds (Nathan et al., 1994). At birth and in the first two years after birth, there is a specialisation of the breast stroma into the functionally distinct components, the interlobular stroma around the glandular epithelium and the intralobular stroma, which is found within the lobules (Atherton et al., 1992). These two populations of fibroblasts, which are defined by their expression of the ectopeptidase dipeptidyl peptidase IV (DPP IV), are characteristic of the adult breast. Specific benign and malignant tumours arise from the specialised intralobular stroma (Atherton et al., 1992) further exploring their distinct identity.
A complex range of molecular signals have already been shown to be involved in the regulation of growth and branching morphogenesis of the mammary gland (Vonderhaar, 1987; Topper and Freeman, 1980; Coleman et al., 1988; Robinson et al., 1991). In an attempt to identify signalling molecules that mediate mesenchymal-epithelial interactions, we sought to examine the role of HGF/SF in normal mammary gland growth and differentiation. Hepatocyte Growth Factor/Scatter Factor, HGF/SF, is produced by mesenchymal cells and exerts its effect on epithelial cells. It is therefore a paracrine mediator of epithelial interactions. HGF/SF has been shown to be mitogenic for a broad spectrum of epithelial cells, to induce the formation of branching tubules and to induce invasiveness and motility in a variety of epithelial cells (Gherardi and Stoker, 1991). Further, recent studies suggest that HGF/SF may play a significant role in mouse embryogenesis (Sonnenberg et al., 1993; Schmidt et al., 1995; Uehara et al., 1995) and chick morphogenesis (Streit et al., 1995: Stern et al., 1990; Myokai et al., 1995). Although expression of HGF/SF transcript and that of its cellular receptor the c-met proto-oncogene have been demonstrated in a variety of tissues, the physiological role of HGF/SF and the molecular mechanism by which it appears to elicit the four distinct functions outlined above, via a single receptor, remain to be elucidated (Gherardi and Stoker, 1991).
In this study, using both mouse and human tissues, we have evaluated the role of HGF/SF and c-met in normal breast growth and development. Using novel techniques developed in our laboratories, we have isolated and highly enriched the fibroblast (separated into interlobular and intralobular) and epithelial components (separated into luminal and myoepithelial) that constitute this tissue (Atherton et al., 1994b; Clarke et al., 1994). Here we identify the mammary fibroblasts as the source of HGF/SF and demonstrate for the first time the response of primary mouse and human mammary epithelial cells to the spectrum of HGF/SF’s activities. The work presented here is unique in that it represents the first organ in which the contribution of HGF/SF to growth and morphogenesis of individual cellular components has been functionally demonstrated.
METHODS AND MATERIALS
Cell lines
MRC 5, human embryonic lung fibroblasts (Jacobs et al., 1970) and MDCK, Madin canine kidney epithelial cells (Madin and Darby, 1958) were purchased from the European Collection. SK23, a human melanoma cell line was a gift from Dr I Hart (ICRF, London). MRC 5 fibroblasts and SK23 were maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% foetal calf serum (FCS) while MDCK cells were maintained in DMEM plus 5% FCS. All cells were grown at 37 °C in a humidified 12% CO2 atmosphere.
Preparation of primary human mammary epithelial cells
Human mammary luminal and myoepithelial cells were prepared from normal human breast material, obtained from reduction mammoplasties of individuals aged between 18 and 33 years. Primary mammary epithelial cells were prepared as described by O’Hare et al., 1991. Briefly, mammary tissue was cut into small pieces and processed by progressive collagenase digestion, followed by sedimentation and filtration to produce organoids (ductal and lobulo-alveolar fragments). The organoids were then placed in culture for a period of 7-10 days. This allows the organoids to attach, spread and grow into semi-confluent cultures, where myoepithelial cells grow out to form a basal layer on top of which colonies of luminal cells grow. At the end of this period, myoepithelial and luminal cell populations were separated and highly enriched using an immunomagnetic separation technique described by Clarke et al. (1994). In this process, human luminal and myoepithelial cells are separated by virtue of their exclusive expression of Epithelial Membrane Antigen (EMA), and Common Acute Lymphoblastic Leukemia Antigen (CALLA/CD10), respectively (O’Hare et al., 1991).
ICR2, a rat monoclonal antibody against EMA (Imrie et al., 1990) and MAS 231P, a mouse monoclonal antibody against CALLA/CD10 (SeraLab) were used in conjunction with appropriately coated MACS microbeads (Becton Dickinson, Cowley, Oxford, UK) to purify human luminal and myoepithelial cells, respectively. The purity of the enriched populations of mammary epithelial cells generated by this procedure was checked in each preparation by immunocytochemistry using cytokeratins 18 and 19 as markers for luminal cells and cytokeratin 14 as the marker for myoepithelial cells (Taylor-Papadimitriou and Lane, 1987). This was done using LLOO2, (IgG3) a mouse monoclonal antibody against cytokeratin 14, LE61, (IgG1) a mouse monoclonal antibody against cytokeratin 18, and LP2K, (IgG2b) a mouse monoclonal antibody against cytokeratin, 19 (Taylor-Papadimitriou and Lane, 1987). The above antibodies (generous gifts from Dr B. Lane) were used in pairs (LE61/LLOO2 and LE61/LP2K) and simultaneously visualised using appropriate subclass-specific tetramethyl rhodamine isothiocynate (RITC) and fluorescein isothiocynate (FITC)-conjugated second antibodies (Southern Biotechnology, via Europath, Bude, UK).
In our hands, this technique is extremely efficient at generating large quantities of highly enriched populations of human luminal and myoepithelial cells, at 95-99% purity as checked by immunocytochemistry. Unless otherwise stated human luminal and myoepithelial cells were grown in RPMI 1640 supplemented with 10% FCS, insulin, I, (5 μg/ml), hydrocortisone, HC, (5 μg/ml) and cholera toxin, CT, (100 ng/ml).
Preparation of human mammary fibroblasts
Heterogeneous populations of human mammary fibroblasts were prepared by centrifugation (90 g, 5 minutes) of the filtrate (through 53 μm sterile nylon filter) from the partially digested mammary tissue (see above). The cells were maintained in DMEM plus 10% FCS.
Preparation of human interlobular mammary fibroblasts
The supernatant from partially digested mammary tissue derived as described above, was centrifuged (400 g for 5 minutes) and the resulting pellet washed and filtered sequentially through 140 μm, 53 μm and 35 μm sterile nylon filters. The filtrate containing single cells was collected and viable stromal cells plated at a density of 8000 cells per cm2 in Ham’s F12:DMEM (1:1) containing 20% FCS. The cells were subcultured at a split ratio of 1:3 on reaching confluence. After three passages, the cultures were routinely fed with DMEM plus 10% FCS (Atherton et al., 1994b).
Preparation of human intralobular mammary fibroblasts
This was achieved as described by Atherton et al. (1994b). Fragments of undigested breast tissue, obtained by limited enzymatic digestion (see above), were digested further for 3 hours at 37 °C in collagenase (2 mg/ml in Leibovitz L-15 medium). The resulting preparation, composed principally of epithelial organoids, were pelleted, washed and filtered through a 140 μm sterile nylon filter in order to retain the larger fragments, which are to be discarded. The filtrate was then passed through a 53 μm filter and the small/medium-sized organoids that were retained were retrieved, resuspended in L-15 and filtered through a 35 μm filter to remove all single cells.
Individual organoids that had visible intralobular stroma still attached were plated in separate wells of a 24-well culture plate containing Ham’s F12:DMEM plus 20% FCS and left to mobilise and grow for 7-14 days. After this period proliferating intralobular fibro-blasts could be seen at the periphery of some explants. Any wells that contained stromal cells not directly associated with an epithelial outgrowth were neglected.
Fibroblasts were isolated from satisfactory preparations by brief (1-2 minute) trypsinisation and resuspended in Ham’s F12:DMEM plus 20% FCS. The harvested cells were filtered through a 35 μm filter to remove the larger undigested epithelial fragments. The resulting intralobular fibroblasts from several original wells were pooled and grown in Ham’s F12:DMEM plus 20% FCS. After three passages the cells were routinely maintained in DMEM plus 10% FCS.
Separated cultures of interlobular and intralobular fibroblasts were then further characterised by virtue of their exclusive expression of dipeptidyl peptidase IV (DPP IV) and differential expression of Neutral endopeptidase (NEP) at early passages, respectively, in order to assess the purity of the populations (Atherton et al., 1994b). Indirect immunofluorescence microscopy was employed to examine DPP IV and NEP expression using a mouse monoclonal antibody to NEP (SS2/3b, Dako, 1:200) and a rabbit antisera to DPP IV (a generous gift from Dr A. J. Kenny, University of Leeds, Leeds, UK; 1:500).
Preparation of primary mouse mammary cells
Primary mouse mammary epithelial cells and fibroblasts were prepared as described by Imagawa et al. (1982) and Emerman and Bissel (1988). Briefly, mammary gland (number four) from 10-12 weeks old Parks mice were excised and the attached lymph node removed. Finely minced mouse mammary tissue was partially digested, with stirring for 1 hour at 37°C, using a digestion cocktail composed of 1.5 mg/ml trypsin (Gibco BRL) and 3 mg/ml collagenase A (Boehringer Mannheim) dissolved in Ham’s F12 plus 10% FCS. For each gram of tissue, 4 ml of digestion cocktail was used. After centrifugation (3 g for 30 seconds) to allow undigested fragments to settle, the fat was discarded while the supernatant and pellet were separated for further processing to yield mammary fibro-blasts and epithelial cells respectively.
Preparation of mouse mammary fibroblasts
The supernatant of the partially digested mammary tissue (obtained as described above) was centrifuged (190 g for 10 minutes) and the pellet predominantly composed of single stromal cells washed and seeded in DMEM plus 10% FCS.
Preparation of primary mouse mammary epithelial cells
Fragments of undigested mouse mammary tissue obtained by limited enzymatic digestion (see above) were further digested with stirring at 37 °C for 30 minutes, using 2 ml digestion cocktail per gram of the original tissue. The resulting epithelial organoids (ductal and lobulo-alveolar fragments) were then pelleted (55 g for 3 minutes) and washed in DMEM plus 10% FCS. The organoids were then seeded into Ham’s F12:DMEM (1:1) supplemented with 10% FCS, I (10 μg/ml), CT (10 ng/ml), transferin, T, (10 μg/ml) and Epidermal growth factor, EGF, (10 ng/ml) and, left for 2 hours. This allows any contaminating fibroblasts to attach preferentially, effectively enriching the non-attached population for epithelial cells. The non-attached cells were then removed and reseeded in the above media and left to grow for a period of 3-4 days after which the cells could be trypsinised into single epithelial cells and used for further study. Mouse mammary epithelial cells were routinely maintained in the above media referred to as complete media.
This procedure is efficient in generating large quantities of highly enriched populations of mouse mammary epithelial cells free of contaminating fibroblasts. However, further purification of epithelial and fibroblast cells into discrete representative subpopulations is currently not possible with the mouse mammary cells since specific markers that distinguish mouse mammary fibroblast subpopulations and mouse mammary epithelial subpopulations have not as yet been identified. Hence the mouse mammary epithelial cells used in this study are a mixture of myoepithelial and luminal cells, while the mouse mammary fibroblasts used represent a population of interlobular and intralobular fibroblasts.
HGF/SF
Mouse recombinant HGF/SF (mrHGF/SF) and human recombinant HGF/SF (hrHGF/SF) were generous gifts from Dr E. Gherardi.
Mitogenicity assay
2×104 cells were seeded in 24-well dishes in basal media and left overnight to settle. The media were then changed to test media containing recombinant HGF/SF. Throughout this study 50 ng/ml mrHGF/SF was used for mouse epithelial cells while 50 ng/ml hrHGF/SF was used for human epithelial cells. At indicated time intervals cells were trypsinised and cell numbers estimated using a Coulter counter (Coulter electronics). In all growth curves, the media was changed every three days.
Morhogenecity assay
Cells and organoids were embedded in collagen gels (Rat tail type I, Becton Dickinson) by mixing 250 μl collagen solution with 40 μl 7.5% sodium hydrogen carbonate and 210 μl media (appropriate to the cells) containing 2×104 cells or approximately 15 organoids. The gels were then poured into 24-well dishes already containing 500 μl collagen gel base per well, allowed to set, covered with media and left overnight. The following day, recombinant HGF/SF was added to the gels at 50 ng/ml and left for 5-7 days with daily feeding of HGF/SF.
Motogenicity assay
2×104 cells were seeded into 24-well dishes containing the appropriate media and left overnight to settle. Recombinant HGF/SF was then added at 50 ng/ml and the cells left for 24-48 hours after which time the cells were photographed.
Co-cultures
Co-cultures were initiated by seeding a well-mixed single cell suspension containing 4×105 epithelial cells together with 4×105 MRC 5 in 2 ml DMEM containing 10% FCS in a 3.5 cm dish. The controls were composed of 4×105 MRC 5 cells seeded in 2 ml DMEM plus 10% FCS in a 3.5 cm dish. The co-cultures and their controls were left to condition their media for a period of 3 days. The conditioned media were centrifuged at 90 g for 5 minutes to remove cell debris and stored at 4 °C until use.
Scatter factor activity assay
Using 96-well plates, twofold serial dilutions of test samples were incubated with 3000 MDCK cells at 37 °C for 24 hours. The plates were then fixed with 4% formaldehyde in PBS, stained with 0.2% Coomassie blue (15 minutes) followed by 1% crystal violet (60 minutes) and each well assessed for scattering. The highest dilution at which scattering could be observed was recorded as the end point of the assay. Since the assay is performed in serial dilutions of 2, the results are presented as log2, ie. SF activity at a dilution of 1/32 is presented as (log2 1/32 =) −5.
Northern analysis of HGF/SF and c-met transcripts
Cultured cell RNA extraction, northern blotting and hybridization procedures were performed as described previously (Boehm et al., 1988). For mouse developmental studies, total RNA was extracted by the method of Chomczynski and Sacchi (1987) from the fourth abdominal mammary glands of female Parks mice (MRC; out-bred) as previously described (Weber-Hall et al., 1994). Northern blot analysis was carried out on poly(A)+ RNA isolated from this material. HGF/SF and c-met probes were as follows; (i) a 2.6 kb cDNA fragment containing the complete mouse HGF coding region (M. Sharpe unpublished data), (ii) a 0.8 kb cDNA fragment encoding part of the C-terminal region of the mouse c-met gene (Chan et al., 1988), (iii) a human HGF/SF, 0.7 kb cDNA fragment covering the β chain region of the h-HGF mRNA (Nakamura et al., 1989), (IV) a 886 bp cDNA fragment encoding parts of the cytoplasmic domain of the human c-met gene (Chan et al., 1987) and, (V) a 400 bp cDNA fragment derived from the 5′ end of the mouse β casein (a kind gift from R. Humphries). Following hybridisation filters were washed to a stringency of 2.5× SSC, 0.1% SDS at 65 °C and autoradiographed for 3 days at −70 °C, using X-OMAT XAR-5 film. The relative quality and quantity of RNA samples used in this study was determined by probing for GAPDH mRNA using a 1.3 kb cDNA clone (Weber-Hall et al., 1994).
Analysis of HGF/SF and c-met gene expression by RT-PCR
A 271 bp region of the human HGF/SF mRNA, encompassing exon 1 and exon 2 sequences (Seki et al., 1991), and a 378 bp region of the human c-met mRNA, encompassing exons n-1, n and n+1 which encode the transmembrane region of this receptor (Lee and Yamada, 1994), were amplified using the following primers:
HGF/ SF-F: 5′-TTCTTTCACCCAGGCATCTC-3′,
HGF/ SF-R: 5′-ATTAGCACATTGGTCTGCAG-3′
and
c-met-F: 5′-CCTGCTGAAATTGAACAGCGAG-3′,
c-met-R: 5′-TGCACTTGTCGGCATGAACC-3′.
Control PCR reactions were carried out against human glucose-6 phosphate dehydrogenase (G6 PD) mRNA using primers described by Boehm et al. (1991) and human β-actin primers as described by Luqmani et al. (1992). The concentration of each RNA preparation to be analysed by RT-PCR was adjusted to 1 μg/ml and the integrity checked by electrophoresis on 1.5% agarose mini gels. Random primed cDNA was reverse transcribed from 2 μg of total RNA in a total reaction volume of 20 μl. PCR reactions (100 μl) were carried out using 2 μl of cDNA template. Amplification consisted of 30 cycles, each composed of a DNA denaturing step of 95 °C for 1 minute, a primer annealing step of 55 °C for 1 minute and a DNA synthesis step of 72°C for 1 minute. A final synthesis step of 72 °C for 5 minutes was carried out. PCR products were analysed on 2% NuSieve / 1% HGT.
RESULTS
HGF/SF expression in mouse mammary tissue
HGF/SF and c-met gene expression was followed during mouse mammary development and differentiation. Northern blots of poly(A)+ RNA to the various stages of development and differentiation were sequentially probed for mouse HGF/SF, mouse c-met and GAPDH transcripts (Fig. 1). The results show that both HGF/SF and c-met are coordinately expressed during the development and differentiation of mouse mammary tissue. The cleared fat pad represents the mesenchymal component of this organ, devoid of mammary epithelial cells, after removal of the rudimentary mammary gland by cauterisation (DeOme et al., 1959). Therefore, the expression of HGF/SF detected in the mammary fat pad indicates that the mesenchymal cells of the mammary gland are likely to be the source of HGF/SF expression in this organ. These data are in agreement with the existing body of evidence demonstrating that HGF/SF expression is restricted to mesenchymal cells (Stoker et al., 1987). The expression of c-met detectable in the cleared fat pad is likely to be due to the endothelial cells (Bussolino et al., 1992).
Expression of HGF/SF and c-met during mouse mammary gland development. A northern blot of poly(A)+ RNA from different stages development and differentiation was sequentially probed for HGF/SF and c-met (3 day exposure). Hybridisation of a control GAPDH probe shows equal RNA loading in all lanes apart from 2 and 10 day lactation (1 hour exposure). In these tracks, β casein demonstrates the presence of comparable levels of RNA in comparison to other tracks (8 minute exposure).
Expression of HGF/SF and c-met during mouse mammary gland development. A northern blot of poly(A)+ RNA from different stages development and differentiation was sequentially probed for HGF/SF and c-met (3 day exposure). Hybridisation of a control GAPDH probe shows equal RNA loading in all lanes apart from 2 and 10 day lactation (1 hour exposure). In these tracks, β casein demonstrates the presence of comparable levels of RNA in comparison to other tracks (8 minute exposure).
Expression of HGF/SF and c-met was detected in virgin mice as early as 6 weeks of age. By 12 weeks, the expression of both genes was elevated and remained at this level for at least the first 12.5 days of pregnancy. However, HGF/SF and c-met appeared to be expressed at undetectable levels at around 17 days of pregnancy and throughout lactation. The presence of undegraded polyadenalated mRNA in these tracks is demonstrated by the detection of full-length β casein and GAPDH mRNA (Gavin and McMahon, 1992; Barker et al., 1995) (Fig. 1). It is now well established that milk protein mRNAs constitute up to 95% of the mRNA population in a lactating rat and mouse mammary gland (casein and whey acidic protein, respectively, represent 80% and 15% of the total poly(A)+ RNA in the mammary gland (Richards et al., 1981, Hobbs et al., 1982)]. Therefore, it is likely that the apparent down regulation of expression of HGF/SF and c-met is a consequence of the feature of lactating mammary gland to express milk proteins preferentially over and above proteins, the expression of which are not crucial for specific requirements of the gland at this point in time. The relative abundance of these transcripts is further reflected in the exposure times required for the detection of HGF/SF, c-met, GAPDH and β casein transcripts (Fig. 1). Interestingly, levels of both HGF/SF and c-met remained high during the early stages of involution, a time at which this tissue undergoes extensive remodelling.
Mammary fibroblasts express HGF/SF
In order to identify the source of HGF/SF in the mammary tissue, populations of interlobular and intralobular fibroblasts (derived from the breast stroma) as well as luminal and myoepithelial (isolated from from organoids) were derived as described in methods. Significant HGF/SF activity was detected in the culture supernatant of the mouse mammary fibroblasts (Fig. 2A) but not in the culture supernatant of the human mammary fibroblasts. However, HGF/SF expression was readily detectable in the human mammary fibroblasts by PCR. Human luminal and myoepithelial cells do not express HGF/SF (Fig. 2B). Interestingly, HGF/SF appears to be more highly expressed in the human intralobular fibroblast population. HGF/SF activity was not detected in the culture supernatant of mouse (data not shown) or human mammary epithelial cells (Fig. 9). Together, these results demonstrate that mammary fibroblasts are the source of HGF/SF in the breast.
HGF/SF activity in mammary fibroblasts. (A) HGF/SF activity in supernatant of mouse and human mammary fibroblasts. Cells were seeded in basal media and left to condition their media for 4 days. The media was then removed and assayed for HGF/SF activity using the MDCK assay. The error bars represent an average of 4 independent values. (B) HGF/SF gene expression in purified human breast cell populations. HGF/SF gene expression was determined by RT-PCR (30 cycles) on RNA made from purified luminal cells (2), Myoepithelial cells (3), heterogeneous fibroblast population derived from primary organoids (4), inter lobular fibroblasts from three separate individuals (5, 7, 9) and intra lobular fibroblasts from the former three individuals (6, 8, 10). MRC 5 RNA was used as a positive control for HGF/SF expression (1). Efficiency of cDNA synthesis from RNA isolated from cell populations was monitored by co-amplification of G6PD sequences (150bP).
HGF/SF activity in mammary fibroblasts. (A) HGF/SF activity in supernatant of mouse and human mammary fibroblasts. Cells were seeded in basal media and left to condition their media for 4 days. The media was then removed and assayed for HGF/SF activity using the MDCK assay. The error bars represent an average of 4 independent values. (B) HGF/SF gene expression in purified human breast cell populations. HGF/SF gene expression was determined by RT-PCR (30 cycles) on RNA made from purified luminal cells (2), Myoepithelial cells (3), heterogeneous fibroblast population derived from primary organoids (4), inter lobular fibroblasts from three separate individuals (5, 7, 9) and intra lobular fibroblasts from the former three individuals (6, 8, 10). MRC 5 RNA was used as a positive control for HGF/SF expression (1). Efficiency of cDNA synthesis from RNA isolated from cell populations was monitored by co-amplification of G6PD sequences (150bP).
Morphogenic effects of HGF/SF on human organoids
Morphogenic effects of HGF/SF on the human breast were examined by treating organoids embedded in collagen gels with HGF/SF at physiological concentrations. After 5 days in culture, in the absence of HGF/SF, the organoids appeared somewhat spread with a conspicuous absence of morphogenic changes (Fig. 3). In the presence of HGF/SF however, the organoids exhibited a striking display of extensive branching tubules. Although structures resembling end buds were not seen in these cultures, it is possible that these tubules were derived by processes similar to those involved in the ductal elongation phase of mammary development at puberty and hence may encompass growth together with morphogenesis (Fig. 3).
HGF/SF induction of branching morphogenesis in human organoids. Human organoids embedded in collagen gel were treated with hrHGF/SF at 50 ng/ml for a period of 5 days. (A) Untreated control cultures; (B) HGF/SF treated cultures. Scale bar = 200 μm.
Morphogenic effects of HGF/SF on mammary epithelial cells
In order to examine the morphogenic effects of HGF/SF on mammary epithelial cells, human luminal and myoepithelial were embedded in collagen gels. Mouse mammary epithelial cells were used as a pooled epithelial population, free from fibroblasts. Mouse mammary epithelial cells when grown embedded in collagen gels and treated with physiological levels of HGF/SF produced extensive tubules with well-formed hollow lumina (Fig. 4).
HGF/SF induction of branching morphogenesis of mouse mammary epithelial cells. Dissociated and highly enriched mouse mammary epithelial cells were embedded in collagen gels and treated with mrHGF/SF (50 ng/ml) for 5 days. (A) Control untreated cultures. (B) HGF/SF treated cultures. (C) Cross section of HGF/SF treated culture demonstrating a well formed lumen. Scale bar in A, B = 200μm. Scale bar in C=10 μm.
HGF/SF induction of branching morphogenesis of mouse mammary epithelial cells. Dissociated and highly enriched mouse mammary epithelial cells were embedded in collagen gels and treated with mrHGF/SF (50 ng/ml) for 5 days. (A) Control untreated cultures. (B) HGF/SF treated cultures. (C) Cross section of HGF/SF treated culture demonstrating a well formed lumen. Scale bar in A, B = 200μm. Scale bar in C=10 μm.
Interestingly, human mammary luminal and myoepithelial cells appeared to respond differently to HGF/SF in that myoepithelial cells form extended branching tubules (Fig. 5A,B) whilst luminal cells were unresponsive to the morphogenic effects of HGF/SF (Fig. 5C,D). Further, luminal cells did not appear to proliferate when embedded in collagen gels in the presence of HGF/SF.
Induction of branching morphogenesis in human mammary epithelial cells. (A) Control untreated cultures of human mammary myoepithelial cells embedded in collagen gel. (B) Human mammary myoepithelial cells embedded in collagen gel and treated with hrHGF/SF (50 ng/ml) for 5 days, demonstrating branching morphogenesis. (C) Control untreated cultures of human mammary luminal cells embedded in collagen gel. (D) Human mammary luminal cells embedded in collagen gel and treated with hrHGF/SF (50 ng/ml) media for 5 days. Scale bar for A, B = 200 μm. Scale bar for C, D = 100 μm.
Induction of branching morphogenesis in human mammary epithelial cells. (A) Control untreated cultures of human mammary myoepithelial cells embedded in collagen gel. (B) Human mammary myoepithelial cells embedded in collagen gel and treated with hrHGF/SF (50 ng/ml) for 5 days, demonstrating branching morphogenesis. (C) Control untreated cultures of human mammary luminal cells embedded in collagen gel. (D) Human mammary luminal cells embedded in collagen gel and treated with hrHGF/SF (50 ng/ml) media for 5 days. Scale bar for A, B = 200 μm. Scale bar for C, D = 100 μm.
Motogenic effects of HGF/SF on mammary epithelia
Treatment of human luminal and myoepithelial cells with HGF/SF when grown on plastic produced classical scattering of both human epithelial cell populations accompanied by a change in morphology of the epithelial cells to a more fibroblastic or polar appearance (Fig. 6). Hence both myoepithelial and luminal cells are responsive to motility effects of HGF/SF.
Motogenic effect of HGF/SF on human mammary epithelial cells. (A) Control untreated cultures of human mammary myoepithelial cells. (B) Human mammary myoepithelial cells treated with hrHGF/SF (50 ng/ml) for 48 hours. (C) Control untreated cultures of human mammary luminal cells. (D) Human mammary luminal cells treated with hrHGF/SF (50 ng/ml) for 48 hours. Scale bar for A, B = 200 μm. Scale bar for C, D = 200 μm.
Motogenic effect of HGF/SF on human mammary epithelial cells. (A) Control untreated cultures of human mammary myoepithelial cells. (B) Human mammary myoepithelial cells treated with hrHGF/SF (50 ng/ml) for 48 hours. (C) Control untreated cultures of human mammary luminal cells. (D) Human mammary luminal cells treated with hrHGF/SF (50 ng/ml) for 48 hours. Scale bar for A, B = 200 μm. Scale bar for C, D = 200 μm.
Mitogenic effect of HGF/SF on mammary epithelial cells
In order to examine the mitogenic effect of HGF/SF on mammary epithelial cells, human mammary epithelial cells (luminal and myoepithelial cells) and mixed mouse mammary epithelial cells were treated with recombinant HGF/SF and their response examined using growth curves.
The results illustrate that HGF/SF produce a fourfold increase in growth of mouse mammary epithelial cells, when grown in basal media [DMEM + 10% FCS] (Fig. 7A). HGF/SF produced a further twofold increase in cell numbers in the presence of a combination of other established growth promoting factors (I, EGF, CT and T). This media (DMEM + 10% FCS, I, EGF, CT and T) referred to as complete media, was used for basic maintenance of these cells. It is interesting to note that the addition of just HGF/SF to basal media, in the absence of other growth-promoting agents, was sufficient to achieve the same growth rate as that achieved in complete media which contained four different additives (Fig. 7A). Moreover, HGF/SF in conjunction with the above additives improved the growth of mouse mammary epithelial cells tenfold.
Mitogenic effects of HGF/SF on mammary epithelial cells. (A) Mouse mammary epithelial cells grown in DMEM plus 10% FCS (○); DMEM plus 10% FCS and mrHGF/SF (•); DMEM:F12 plus 10% FCS, I, CT, T and EGF (▿); DMEM:F12 plus 10% FCS, I, CT, T, EGF and mrHGF/SF (▾). (B) Human mammary luminal cells grown in RPMI plus 1% FCS, I, CT and HC (▴); RPMI plus 1% FCS, I, CT, HC and hrHGF/SF (▵); RPMI plus 10% FCS, I, CT and HC (○); RPMI plus 10% FCS, I, CT, HC and hrHGF/SF (•). (C) Human mammary myoepithelial cells grown in RPMI plus 1% FCS, I, CT and HC (○); RPMI plus 1% FCS, I, CT, HC and hrHGF/SF (•); RPMI plus 10% FCS, I, CT and HC (▫); RPMI plus 10% FCS, I, CT, HC and hrHGF/SF (▪).
Mitogenic effects of HGF/SF on mammary epithelial cells. (A) Mouse mammary epithelial cells grown in DMEM plus 10% FCS (○); DMEM plus 10% FCS and mrHGF/SF (•); DMEM:F12 plus 10% FCS, I, CT, T and EGF (▿); DMEM:F12 plus 10% FCS, I, CT, T, EGF and mrHGF/SF (▾). (B) Human mammary luminal cells grown in RPMI plus 1% FCS, I, CT and HC (▴); RPMI plus 1% FCS, I, CT, HC and hrHGF/SF (▵); RPMI plus 10% FCS, I, CT and HC (○); RPMI plus 10% FCS, I, CT, HC and hrHGF/SF (•). (C) Human mammary myoepithelial cells grown in RPMI plus 1% FCS, I, CT and HC (○); RPMI plus 1% FCS, I, CT, HC and hrHGF/SF (•); RPMI plus 10% FCS, I, CT and HC (▫); RPMI plus 10% FCS, I, CT, HC and hrHGF/SF (▪).
HGF/SF is also a potent mitogen for human luminal cells, improving the growth rate of these cells fivefold over cells grown in RPMI plus 10% FCS supplemented with I, HC and CT (Fig. 7B). HGF/SF improves the growth of luminal cells by ninefold when grown in RPMI plus 1% FCS supplemented with I, HC and CT (Fig. 7B). However, HGF/SF appears to have no growth stimulatory effects on myoepithelial cells (Fig. 7C). It should be noted that, in RPMI supplemented with I, HC and CT, myoepithelial cells grow significantly better in the presence of 1% FCS rather than 10% FCS. We further examined the response of human myoepithelial cells to HGF/SF in a variety of growth media and so far have not observed any mitogenic effects of HGF/SF upon myoepithelial cells (data not shown).
We have demonstrated that both human myoepithelial and luminal cells express c-met. Human luminal cells appear to express relatively higher levels of c-met than myoepithelial cells (Fig. 8A). Further, c-met expression in human mammary fibroblasts was undetectable by RT-PCR analysis (Fig. 8C).
c-met gene expression in purified human luminal and myoepithelial cells. 25 mg total RNA from purified luminal and myoepithelial cells was probed by northern blot for c-met gene expression (A). The position of full-length c-met transcript is labelled. This figure shows that c-met gene expression is higher in luminal cells than myoepithelial cells (4 day exposure). Equivalence of RNA in each track is demonstrated by hybridisation to human GAPDH probe (B) (1 day exposure). RT-PCR analysis of human mammary fibroblast and epithelial cells is shown in C. Primers that flank sequences encoding the transmembrane region of the human c-met receptor were used to examine c-met expression in human luminal (lane 2), myoepithelial (lane 3) and fibroblast (lane 4) cells. A single c-met-specific PCR product (378 bp) is seen in luminal and myoepithelial cells as well as the control melanoma cell line SK23 (lane 1). No PCR product could be seen with mammary fibroblast cDNA (lane 4). Control b-actin PCR reactions for the same samples are shown in lanes 5-8.
c-met gene expression in purified human luminal and myoepithelial cells. 25 mg total RNA from purified luminal and myoepithelial cells was probed by northern blot for c-met gene expression (A). The position of full-length c-met transcript is labelled. This figure shows that c-met gene expression is higher in luminal cells than myoepithelial cells (4 day exposure). Equivalence of RNA in each track is demonstrated by hybridisation to human GAPDH probe (B) (1 day exposure). RT-PCR analysis of human mammary fibroblast and epithelial cells is shown in C. Primers that flank sequences encoding the transmembrane region of the human c-met receptor were used to examine c-met expression in human luminal (lane 2), myoepithelial (lane 3) and fibroblast (lane 4) cells. A single c-met-specific PCR product (378 bp) is seen in luminal and myoepithelial cells as well as the control melanoma cell line SK23 (lane 1). No PCR product could be seen with mammary fibroblast cDNA (lane 4). Control b-actin PCR reactions for the same samples are shown in lanes 5-8.
Inhibition of HGF/SF activity released by MRC 5 cells in coculture with human mammary epithelial cells. HGF/SF activity in co-cultures of MRC 5 cells with human mammary myoepithelial and luminal cells ◼ using MDCK as a positive control (□) in comparison to MRC 5 control cultures (■). The error bars represent the standard deviation of six readings.
Inhibition of HGF/SF activity released by MRC 5 cells in coculture with human mammary epithelial cells. HGF/SF activity in co-cultures of MRC 5 cells with human mammary myoepithelial and luminal cells ◼ using MDCK as a positive control (□) in comparison to MRC 5 control cultures (■). The error bars represent the standard deviation of six readings.
Co-culture of mammary epithelial cells with MRC 5 fibroblasts
In our earlier studies of regulation of HGF/SF expression, we have shown that epithelial cells can inhibit HGF/SF expression in HGF/SF-producing fibroblasts (i.e. MRC 5) when the two cell types are co-cultured (Kamalati et al., 1992). The capacity to inhibit HGF/SF expression in MRC 5 fibroblasts is a feature exclusive to epithelial cells, since cells of mesenchymal origin fail to inhibit HGF/SF activity in co-culture with MRC 5 cells (Kamalati et al, 1992). In order to examine the capacity of human myoepithelial and luminal cells to inhibit HGF/SF expression by MRC 5 cells, we co-cultured these cells with MRC 5 fibroblasts. The results (Fig. 9) show that luminal and myoepithelial cells do not express HGF/SF activity, in agreement with the lack of HGF/SF transcript in these cells shown above (Fig. 2). However, in co-culture with MRC 5 fibroblasts, human luminal cells appear to be extremely efficient at inhibiting HGF/SF expression while myoepithelial cells failed to inhibit HGF/SF expression by MRC 5 cells.
DISCUSSION
The importance of cellular interactions in the control of cell motility and morphogenic events during development is now well established (Grobstein, 1954; Kratochwill, 1972, 1983). Interactions between epithelial and mesenchymal cells influence epithelial proliferation, differentiation and morphogenesis (Grobstein, 1967; Saxen et al., 1976; Saxen, 1977). Mesenchymal and epithelial cells use soluble growth factors to mediate local effects through binding to their cellular receptors. Since HGF/SF is a multifunctional paracrine effector of epithelial cells, we aimed to evaluate the role of HGF/SF and its cellular receptor, c-met, in the mammary gland. Using both mouse and human models, we have demonstrated that HGF/SF may have an important role in mammary growth and differentiation since it is a potent mitogen for mammary epithelial cells and can induce motility and branching morphogenesis in these cells.
HGF/SF and c-met expression in mouse mammary tissue
Our data demonstrate that HGF/SF and its receptor are expressed during mouse mammary gland development and differentiation. This expression appears to be coordinately regulated through development, pregnancy, lactation and involution suggesting a role for HGF/SF in growth, morphogenesis, differentiation and tissue remodelling. This proposed role is supported by the in vitro findings, which show that HGF/SF is mitogenic, motogenic and morphogenic for mammary epithelial cells. Hence, HGF/SF is likely to play an important role in the mammary gland at times of cellular interactions and tissue remodelling.
Our data show that mammary fibroblasts are the source of HGF/SF in human and mouse mammary gland. Specifically, in the human mammary gland, the intralobular fibroblasts rather than interlobular fibroblasts appear to be the subpopulation responsible for HGF/SF expression. In contrast, c-met is expressed by both mouse and human mammary epithelial cell populations. Interestingly, the c-met transcript appears more abundant in human luminal cells relative to myoepithelial cells.
Mesenchymal-epithelial interactions in the mammary gland
The retention of inductive activity in adult mammary stroma and the maintenance of the ability of adult epithelium to respond to inductive mesenchymal influences provide suggestive evidence that stromal-epithelial interactions are active in the postnatal gland (Hoshino, 1964; Daniel et al., 1968; Sakakura et al., 1979; Cunha et al., 1992). Direct evidence for this is provided by studies of postnatal mouse mammary ductal elongation, which occur at puberty (Daniel and Silberstein, 1987). In this context, our data on the morphogenic effects of HGF/SF upon human organoids suggest that the branching morphogenesis observed in our studies may be obtained by processes similar to those that are involved in the ductal elongation phase of mammary development; i.e. rapid growth accompanied by cellular movement and morphogenesis of the epithelial component of the gland. Hence, HGF/SF may be a candidate molecule for a naturally occurring factor that triggers signal transduction pathways leading to the activation of these processes.
Role of HGF/SF in mammary gland growth and development
Our studies on the effects of HGF/SF upon mammary epithelial cells show that both human and mouse mammary epithelial cells are highly responsive to the full spectrum of HGF/SF activities. We have found HGF/SF to be a potent mitogen for mouse mammary epithelial cells causing up to tenfold increase in cell growth as determined by increased cell numbers. Previous studies of the mitogenic role of HGF/SF have been limited to the measurements of DNA synthesis and not absolute changes in cell number. HGF/SF is also a potent morphogen for these cells, inducing extensive tubular branching and well-formed lumina.
Interestingly, HGF/SF appears to effect human mammary epithelial cells differentially, depending on the epithelial cell type. Although HGF/SF was found to be a potent mitogen for human luminal cells, as determined by a ninefold increase in their cell numbers, it appeared to have no measurable mitogenic activity on myoepithelial cells. Further, we demonstrate that HGF/SF elicits a morphological response in the human myoepithelial cells, inducing the formation of extensive branching tubules. Under identical experimental conditions, human luminal cells do not appear to respond to the morphogenic capacity of HGF/SF. Finally, HGF/SF exerts a motogenic effect on both human luminal and myoepithelial cells, causing classical scattering of these cells accompanied by a change in the morphological appearance of the cells. The mitogenic response of the luminal cells to HGF/SF compared with myoepithelial cells is consistent with our earlier in vivo observations. Luminal cells have been shown to have a higher [3H]thymidine labelling index in vivo whilst myoepithelial cells have very low labelling (Joshi et al., 1986). In relation to the contrasting morphogenic responses in these two populations and their motility on addition of HGF/SF, it could be postulated that the myoepithelial cells form the ductal skeletal network down which the luminal cells migrate. In the rodent, the cap cells of the end buds give rise to the outer sheath of myoepithelial cells and it is thus hypothesized that HGF/SF may be responsible for inducing the ductal elongation, morphogenic and motility responses mediated through c-met expression on the differentiated myoepithelial cell populations; whilst the effect on the luminal cells is mitogenic and motogenic, requirements to populate the lumen. It is in this context that the current evidance on the clonal analysis of luminal and myoepithelial cells indicates that they are independent self-renewing populations (O’Hare et al, 1991; Gusterson et al., 1995)
HGF/SF has been previously shown to stimulate DNA synthesis in transformed human mammary epithelial cells (Rubin et al., 1991). Also, fibroblasts derived from a human breast carcinoma have been shown to express HGF/SF transcript but not protein (Seslar et al., 1993). The data presented here are the first reporting the effects of HGF/SF on primary normal mammary epithelial cells and expression of HGF/SF in primary normal human mammary fibroblasts.
The differential response of human luminal and myoepithelial cells to HGF/SF as a mitogen and morphogen, respectively, is particularly interesting since these two epithelial cell populations express c-met differentially. This hence begs the question whether the response to different capacities of HGF/SF by these two epithelial cell populations is governed by the receptor density or receptor subtype (given that c-met generates a number of transcripts) displayed on the cell surface. Identification of the number of functional receptors expressed by these cells may help elucidate this observation, since it is now well documented that receptor density can effect the extent to which downstream signalling pathways are engaged or activated (Hempstead et al., 1992; Heasley and Johnson, 1992; Traverse et al., 1994; Dikic et al., 1994).
The apparent differential response displayed by these cells could be used as a model system to investigate the intracellular signalling pathways that are involved in executing mitogenic, motogenic and morphogenic responses elicited by one ligand, apparently via one receptor. There now exists a compelling body of evidence demonstrating that individual factors are expressed differentially throughout mammary gland development and distributed within the gland in unique patterns, suggesting distinct intrinsic roles for the individual factors based on their temporal and spatial localisation as well as their receptor density on target cells (Robinson et al., 1991; Coleman et al., 1988).
Mammary luminal and myoepithelial cells are derived from the basal layer of the foetal periderm (Gusterson et al., 1994). However, myoepithelial cells express smooth muscle α-actin (O’Hare et al., 1991), a feature of smooth muscle cells. In this context, the inability of myoepithelial cells to inhibit HGF/SF expression in MRC 5 fibroblasts is yet another feature distinguishing myoepithelial cells from luminal cells and further demonstrates some aspects of the non-epithelial characteristics of these cells.
Myoepithelial cells form the outermost monolayer of mammary ducts, with processes extending laterally along ducts (Daams et al., 1987). The functional significance of myoepithelial cells in secretory mammary tissue, where they cause milk ejection, is clear. However, in virgin animals, this contractile tissue does not appear to have an immediately obvious function. In this context, the differential response to HGF/SF demonstrated by luminal and myoepithelial cells is interesting, since it suggests that myoepithelial cells may have an important role in branching morphogenesis of the gland and hence functionally contribute to the continuously renewing architecture of the gland.
Evidently, other factors must be involved in the regulation of proliferation of myoepithelial cells (Coleman et al., 1988). Indeed it is rapidly becoming clear that the explosion of growth and morphogenetic events at puberty are the result of highly regulated and precisely timed interactions between a variety of systemic hormones and peptide growth factors (Coleman et al.1988; Williams and Daniel, 1983; Silberstein et al., 1990). Recently, the HGF/SF gene promoter has been shown to contain two estrogen-responsive elements (Liu et al., 1994). Using RT-PCR we could not detect estrogen receptors in the cultured human mammary fibroblast and epithelial cells. Further, using 17-β estradiol, we were unable to induce HGF/SF expression in human mammary epithelial cells or to enhance the expression of HGF/SF in human mammary fibro-blasts (data not shown).
Here we demonstrate that HGF/SF is produced by mammary fibroblasts and can elicit mitogenic, motogenic and morphogenic responses in breast epithelial cells. HGF/SF must therefore be considered a strong candidate for a naturally occurring mutlifunctional mammary tissue cytokine.
ACKNOWLEDGEMENTS
The authors are indebted to Dr E. Gherardi for generous gifts of recombinant mouse and human HGF/SF. The authors wish to thank Ms C. Clarke and Dr M. O’ Hare for their advise on the separation of mammary epithelial cells, Dr S. Ali for his help and advise on estrogen receptor studies, Dr M.Crompton for his continued support and the Cancer Research Campaign for supporting this work. L. B. would also like to thank the MRC for additional support.