Stem cells in mammary gland differentiation and cancer

The mammary glands of both adult rats and humans consist of a system of branching ducts terminating in alveoli and embedded in a fatty stroma (Raynaud, 1961). The mammary ducts are composed of one or more layers of cuboidal, epithelial cells, some of which border a lumen that is continuous throughout the ductal system. The epithelial cells are surrounded by a layer of elongated, myoepithelial cells (Hollman, 1974; Vorherr, 1974). These two fully differentiated cell types have been distinguished in the past by their characteristic ultrastructural morphologies. The ductal epithelial cells possess apical microvilli and specialized junctional complexes Key words: mammary gland, stem cells in vivo and in vitro, differentiation and cancer. with associated desmosomes, whereas the myoepithelial cells possess smooth musclelike myofilaments with pinocytotic vesicles and basement membranes on their basal surfaces. A third functionally differentiated cell type, the secretory cell, is found in the mammary alveoli. This cell type is characterized by its ultrastructure and, during lactation, by the synthesis and secretion of milk products (Ozzello, 1971; Radnor, 1971).

More recently, immunocytochemical stains have been used to distinguish between the different cell types (Table 1). In the rat, the epithelial cells are stainable by antisera to milk fat globule membrane (MFGM) (Warburton et al. 1982ɑ) and keratin monoclonal antibody (MAB) LE6l (Taylor-Papadimitriou et al. 1983). The myoepithelial cells can be stained by antisera to vimentin, actin, myosin (Dulbecco, 1982; Warburton et al. 1982ɑ), keratin MAB LP34 (Taylor-Papadimitriou et al. 1983; Warburton et al. 1987), and by the lectins Griffonia simplicifoliaΛ (GS-1) and pokeweed mitogen (Hughes, 1988). The basement membrane, which is probably synthesized at least in part by the myoepithelial cells, stains with antisera to laminin, type IV collagen and Thy-1 antigen (Dulbecco, 1982; Rudland et al. 1982; Warburton et al. 1982a; Monaghan et al. 1983). In the human gland, antisera (Heyderman et al. 1979) and MABs (Foster et al. 1982; Taylor-Papadimitriou et al. 1983) to MFGM are primarily against a single glycoprotein termed epithelial membrane antigen (EMA) (Ormerod et al. 1984; Mcllhinneyet al. 1985), whilst the two lectins (Hughes, 1988) and the antiserum to human keratin (Gusterson et al. 1982) fail to show any discriminatory staining. MAB LICR-LON-23.IO which recognizes basal cells of the skin and blood vessels (Gusterson et al. 1985) and MABs to the common acute lymphoblastic leukaemia antigen (CALLA) (Gusterson et al. 1986) preferentially recognize the human myoepithelial cells. The secretory alveolar cells from both species are characterized by being stainable by peanut lectin (Newman et al. 1979) and, during lactation, by antisera to their respective caseins (Rudland et al. 1983ɑ; Earl & Mcllhinney, 1985) (Table 1).

The development of the mammary parenchyma takes place predominantly after birth, but prior to puberty (Myers, 1919; Dawson, 1934), by the lengthening and branching of primative ducts within the mammary fat pad. During this period of growth, the ducts terminate in globular structures called terminal end buds in rats (TEBs), which contain most of the dividing parenchymal cells (Dawson, 1934; Russo et al. 1982). The number of globular structures reaches a maximum in rats of about 20 days old (Russo et al. 1982), and in humans of about 13 years old (Dawson, 1934) (Fig. 1A). Thereafter the number rapidly declines as the globular structures differentiate to terminal ducts and alveolar buds in rats (Russo & Russo, 1978) or to terminal ductal alveolar units (TDLUs) in humans (Dawson, 1934) (Fig. IB). The alveolar buds and TDLUs are the direct precursors of the secretory alveoli. In rats the TEBs and to a lesser extent the alveolar buds consist of a heterogeneous collection of cells which show a gradation in ultrastructural and immunocytochemical-staining characteristics towards the epithelial cells on the one hand, and to the myoepithelial cells of the subtending duct on the other hand (Williams & Daniel, 1983; Ormerod & Rudland, 1984). Some of the terminal structures in humans also show some evidence for similar morphological gradations (Stirling & Chandler, 1976; Smith et al. 1984; Rudland& Hughes, 1989). Thus some evidence exists in vivo relating epithelial cells to myoepithelial cells on the one hand and to alveolar cells on the other hand, although this evidence is stronger in the rat than in the human at present.

The susceptibility of the rat mammary gland to chemical carcinogenesis correlates with the presence in the gland of TEBs and terminal ducts (Russo et al. 1977). The tumours induced by dimethyl-benzanthracene (DMBA) (Huggins et al. 1961) or nitrosomethyl urea (NMU) (Gullino et al. 1975) are predominantly cytologically benign in the authors’ experience (Williams et al. 1981). These relatively benign tumours contain areas of epithelial and elongated, myoepithelial-like cells in duct-like arrangements (Murad & von Haam, 1972). However, many of the elongated, myoepithelial-like cells possess a more undifferentiated appearance than the myoepithelial cells of mature mammary ducts (Dunnington et al. 1984a). Hormonal stimulation of the host leads to production of a small proportion of alveolar-like cells which can synthesize casein (Supowit & Rosen, 1982). However, the amount of casein and casein mRNA produced by these cells is only 1-5 % of that produced by the alveolar cells of the lactating mammary glands in normal rats, when animals bearing the tumours are subsequently mated (Herbert et al. 1978; Supowit & Rosen, 1982; Rudland et al. 1983a). Chemical induction in partially immune-deficient rats that are then subjected to nonspecific immunostimulation can produce metastatic tumours (Table 2) which disseminate widely, some like the human disease (Kim, 1979). However, no cells with any myoepithelial characteristics are seen in such malignant carcinomas (Dunnington et al. 1984a).

The primary carcinogens which induce mammary tumours in humans are completely unknown, apart from two exceptions where women had been exposed to high doses of radiation after atomic bomb explosions at Hiroshima and Nagasaki (McGregor et al. 1977). As in rats, the most susceptible developmental stage for these radiation-induced breast cancers is probably in prepubertal/adolescent females (McGregor et al. 1977). The specific phenotypic feature which best correlates with increased risk of neoplastic disease in humans is the presence of atypical epithelial cell proliferations in terminal ductal structures (Wellings et al. 1975). These atypical structures probably represent a spectrum from benign lesions to carcinoma-in-situ (Wellings & Yang, 1983), the direct precursor to mammary carcinoma, although there are contrary views (Azzopardi, 1979). Ultrastructural (Ahmed, 1978; Azzopardi, 1979; Macartney et al. 1979; Gould et al. 1980) and immunocytochemical techniques (Albrechstein et al. 1981; Barsky et al. 1982; Bussolati etal. 1980; Macartney et al. 1979; Gusterson et al. 1982; 1985; 1986) have shown that some myoepithelial cells are always present in the major categories of benign breast disease (epitheliosis, adenosis and fibroadenoma), but they are almost entirely lost in infiltrating ductal carcinomas. Similarly, in the few cases examined with well-characterized reagents, pregnant/lactating women bearing benign tumours can produce neoplastic cells that secrete casein, whereas none were seen in malignant carcinomas (Earl & Mcllhinney, 1985; Earl, 1987). Thus the broad pattern of malignant cell types in human breast neoplasms is similar, to some extent, to that found in the corresponding rat mammary tumours; both the myoepithelial cell and the putative alveolar cells are lost in malignant compared with nonmalignant tumours.

To determine whether one cell type can give rise to another cell type directly, it is frequently necessary to obtain immortalized cell lines cloned from a single cell, and to observe the different cell types that such a system will generate.

In the rat, limited digestion of mammary glands or carcinogen-induced mammary tumours yields ‘organoids’ that can subsequently adhere to the surface of a tissue culture vessel and produce growing cultures of epithelial cells (Hallowes et al. 1977b; Rudland etal. 1977). After a few passages most of these cells die out, but the occasional spontaneously transformed, immortalized epithelial cell is generated which can eventually be cloned (Bennett et al. 1978). In this way single-cell-cloned epithelial cell lines have been obtained from the normal mammary glands of 7-day-old inbred, Furth-Wistar rats (Ormerod & Rudland, 1985), from DMBA-induced benign tumours of out-bred, Sprague-Dawley rats (Bennett et al. 1978), or inbred, Furth-Wistar rats (Dunnington et al. 1983) and from an NMU-induced rat mammary tumour (Dulbecco et al. 1981) (summarized in Table 2). Cultures of mammoplasty specimens from otherwise normal human breasts that have been obtained in virtually the same way as those from the rat mammary glands, however, fail to undergo spontaneous transformation events, and eventually die out after several passages in culture (Hallowes et al. l977a; Stampfer et al. 1980; Easty et al. 1980; Rudland et al. l989b). To obviate this problem, human epithelial cells have been immortalized by transforming them with simian virus 40 (SV4O) (Fig. 2) (Chang et al. 1983; Rudland et al. 19896). All the rat and human epithelial cell lines discussed above behave in a similar manner (Table 2), and thus the results of one, rat mammary 25 (Rama 25) from a benign rat mammary tumour, is described in detail below.

The epithelial cell lines, both rat and human, are conveniently cultured on a plastic substratum where they grow with a cuboidal morphology (Fig. 2A). When such cultures become densely packed, small, dark, polygonal cells are formed (Fig. 2B) which can contain vacuoles or ‘droplets’ at their peripheries (droplet cells) (Fig. 2F) (Bennett et al. 1978). These droplet cells form hemispherical blisters or domes (Fig. 2F) that arise from the unidirectional pumping action of the ouabain-sensitive sodium/potassium ATPase (Paterson et al. 1985a). The overall process can be accelerated with the erythroleukaemic differentiating agent, dimethyl sulphoxide (Friend et al. 1971; Bennett et al. 1978), or retinoic acid (Rudland et al. 19836) in the presence of the mammotrophic hormones, prolactin, estrogen, hydrocortisone and insulin. Such cultures of Rama 25 produce authenticated rat /3-casein, although the small amounts (5O-IOO× less than in lactating rat mammary glands) may reflect the neoplastic origins of this particular cell line (Warburton et al. 1983). The discrete morphological stages observed in the formation of the casein-secreting, doming cultures in vitro (Paterson et al. 19856) are paralleled by changes in a small number of specific polypeptides (Paterson & Rudland, 1985A). The final casein-secreting stage resembles alveolar-like cells, particularly those found in the benign rat mammary tumours (Rudland et al. 1983). In addition, the human mammary epithelial cell lines can also keratinize in culture (Fig. 2E), whereas this is observed infrequently in the corresponding rodent cell lines (Rudland et al. 19896).

Although the epithelial cell lines have been single-cell cloned at least once, confluent cultures at high passage-number yield ridges of elongated cells and subconfluent cultures yield from 0·1 % (human cell lines: Rudland et al. 19896) up to 3 % (rat cell lines: Warburton et al. 19826; Ormerod & Rudland, 1985) of clones of cells with an elongated morphology (Fig. 2C). Similar morphological forms occur in epithelial cell lines of mouse mammary tumours (Sanford et al. 1961; Dexter et al. 1978; Hager et al. 1981). From a comparison of the ultrastructure and immunocytochemical-staining characteristics of histological sections of rat (Ormerod & Rudland, 1982; Warburton et al. 19826) and human mammary glands (Gusterson et al. 1982, 1985, 1986) and of their primary cultures (Warburton et al. 1985; Rudland et al. 1989a,b), the elongated cells derived from such cultures are thought to be related to myoepithelial cells rather than to fibroblasts (Fig. 2D). However, the final phenotype of these elongated cells can vary. In general, cells of a more mature myoepithelial phenotype have been derived from the epithelial cells of normal mammary glands, e.g. the rat cell lines Rama 7O4E (Ormerod & Rudland, 1985) and Rama 401 (Warburton et al. 19816), than from the epithelial cells of mammary tumours, e.g. the rat cell lines Rama 29 (Bennett et al. 1978) and Rama 37E5 (Dunnington et al. 1983) (Table 2). This result is consistent with the finding in the previous section that the better differentiated myoepithelial cells occur in normal rat mammary glands rather than in their tumours. Thus, based on the above results, the majority of the more-elongated cells in vitro are classified as myoepithelial-like rather than as mature myoepithelial cells (Rudland et al. 1980, 19896).

The epithelial cell lines can thus give rise to both alveolar-like cells and myoepithelial-like cells (Fig. 3). They are therefore possible candidates for stem cells for the mammary gland, since they can undergo in vitro the morphological transitions observed in terminal ductal structures in vivo (pp. 97-98) and in primary cultures in vitro (Rudland, 1987). Moreover, when grown on floating collagen gels which mimic the stromal matrix of the mammary gland, such epithelial cell lines form branched, duct-like structures reminiscent of the immature ducts found in neonatal mammary glands (Bennett, 1980; Ormerod & Rudland, 1982, 1985, 1988), further confirming their possible stem-cell properties.

Cloned cell lines that are intermediate (I) in morphology and known marker content between Rama 25 epithelial cells and elongated, myoepithelial-like cells have been isolated, and they form a series in the order: Rama 25 cuboidal cells, Rama 25-12, Rama 25-11, Rama 25-14, and elongated cells, e.g. Rama 29 (Table 2; Fig. 4: Rudland et al. 1986). When grown on floating collagen gels, the intermediate cell line Rama 25-12 forms more-mature, duct-like structures than the parental Rama 25, and ultrastructural and immunocytochemical analysis suggests that Rama 25-11 and Rama 25-14 resemble the intermediate cells of the terminal ductal structures in vivo (Rudland et al. 1986). The intermediate cell lines in vitro thus may represent the heterogeneous cells observed earlier (pp. 97-98) in budded structures in vivo. The above cellular order is maintained for increasing abundance of 7 polypeptides which are characteristic of elongated, myoepithelial-like cells and decreasing abundance of 4 polypeptides which are characteristic of cuboidal epithelial cells (Fig. 4) (Paterson & Rudland, 1985è; Rudland et al. 1986). The majority of these new proteins do not correspond to the known proteinaceous markers of the myoepithelial cells since the two dimensional-gel systems used to identify them cannot easily detect the lower levels of most of the known marker proteins of the myoepithelial cell.

One of the earliest detectable increases in a protein along the myoepithelial-like differentiation pathway is that due to polypeptide 13 (Fig. 4) (Rudland et al. 1986), a novel protein of 9000 mol.wt. termed p9Ka by Barraclough et al. (1982, 1984a). The nucleotide sequence of its gene (Fig. 5) suggests that it may be related to a class of small, regulatory, calcium-binding proteins (Fig. 6) (Barraclough et al. 1987ò). The p9Ka protein may therefore serve to trigger the changes in the cytoskeleton (Paterson & Rudland, 1985ò) which have been observed along this pathway to myoepithelial-like cells (Warburton et al. 198lɑ,6). Antibodies to purified p9Ka and to a synthetic peptide corresponding to a short stretch of its deduced amino acid sequence (Barraclough et al. 19876) bind only to myoepithelial cells in the rat mammary parenchyma, confirming the myoepithelial origins of p9Ka (Haynes, 1988). This protein and its messenger RNA coordinately increase initially in the intermediate Rama 25-11 cells (Barraclough et al. 1984a,b Rudland et al. 1986; B. R. Barraclough, unpublished results). Similarly Thy-1 protein and mRNA coordinately increase, but this increase occurs mainly in the Rama 25-14 and elongated cells (Rudland et al. 1986; Barraclough etal. 1987A). These results suggest that asynchronous or stepwise regulation of the production of marker proteins for the myoepithelial-like cell is controlled mainly at the level of transcription of the DNA. This is not always the case. Thus the increase in laminin and type IV collagen which occurs mainly in the Rama 25-11 and Rama 25-14 cells (Rudland et al. 1986) is not due to major changes in their levels of mRNA (Warburton et al. 1986; Barraclough et al. l987a) but, in the case of type IV collagen, to decreases in its rate of intracellular degradation (Warburton et al. 1986).

In the case of p9Ka, preliminary evidence suggests that at least part of the increase in its accumulation in the myoepithelial-like cells in vitro arises from an increased rate of transcription of its mRNA (B.R. Barraclough, unpublished observation). In many genes, regions of DNA which are important in controlling the synthesis of their mRNAs are often located immediately adjacent to those sequences that correspond to the 5’ end of their mRNAs. However, these regions, such as the TATA (Breathnach & Chambon, 1981) and CAAT (Benoist et al. 1980) consensus sequences are not found close to the p9Ka gene corresponding to p9Ka mRNA. In at least two other such cases, the murine Thy-1 gene (Giguere et al. 1985; Ingraham & Evans, 1986) and the 3-hydroxy-3-methylglutaryl coenzyme A reductase gene (Reynolds et al. 1984), multiple initiation sites for the transcription of the mRNAs have been reported, and this is also the case for the mRNA for p9Ka (Barraclough et al. 1987a). Thus it is possible that the genes for p9Ka and Thy-1 may contain a common or closely related novel promoter which regulates their expression between epithelial and myoepithelial-like cells.

In the rat, a variety of epithelial cell lines have been obtained from different transplantable tumours of the mammary gland, and these show similar metastatic potentials to those of their parental tumours (Table 2). These rat cell lines have been obtained with difficulty and grow as loosely-adherent colonies (Fig. 7A) much more slowly than cells from normal glands or benign tumours (Rudland, 1987). In contrast to cultured cells from normal and benign rat mammary glands, those from our metastasizing rat mammary carcinomas, as well as any cell lines developed from them, yield no myoepithelial cells nor any casein-producing, alveolar-like cells under the requisite hormonal conditions (Dunnington et al. 1984a; Williams et al. 1985; Rudland et al. unpublished). The weakly metastasizing Rama 600 cell line (Table 2) does contain a more elongated cellular component, but if this represents differentiation to a myoepithelial cell it is of a partial and incomplete nature, and probably reflects only a vestige of the complete pathway (Williams et al. 1985). Similarly, Rama 600 cells also appear to retain only a vestige of the differentiation pathway to alveolar cells (Rudland, 1988). Perhaps neoplastic transformation of the epithelia-1/intermediate cells results in a truncation of both differentiation pathways, and this truncation occurs earlier with increasing metastatic potential (Rudland, 1987).

Like most of the metastasizing rat mammary tumours, the culture of human mammary carcinomas has been extremely difficult (Hallowes et al.1977aKirkland et al. 1979). Routine digestion of over 100 primary infiltrating ductal carcinomas with collagenase, by slight modifications of the methods used for the benign rat mammary tumours (pp. 99-102), yields loosely adherent, malignant-looking cell clusters (Fig. 7B) and fast-adherent, less malignant-looking epithelium on collagen gels (Hallowes et al. 1983; Rudland et al. 1985). Metastases in axillary lymph nodes and pleural effusions yield only the loosely adherent clusters, whilst normal mammary glands and benign fibroadenomas yield only fast-adherent colonies (Hallowes et al. 1983; Rudland etal. 1985). These results suggest that the fastgrowing adherent sheets of epithelium from primary ductal carcinomas (Smith et al. 1981) do not usually represent the most-metastasizing cell populations but, as in the rat above, the latter are best represented by the slow-growing, loosely adherent aggregates (Rudland, 1987). Continued passage of one preparation of loosely adherent cell clusters has yielded a continuously growing cell strain, Ca2-83 (Table 2; Fig. 7B), which has not yet undergone a period of crisis (Rudland et al. 1985), unlike most other cell lines established from malignant breast cancer cells (Semen et al. 1976; Lasfargues et al. 1978; Engel et al. 1978). Since the fastadherent sheets of epithelium from cultures of different human mammary tissues always contain elongated, myoepithelial-like cells, but the loosely adherent clusters do not, myoepithelial-like cells are usually found in cultures of fibroadenomas and uninvolved peritumoral tissue adjacent to carcinoma (Rudland et al. 1985). However, they are almost invariably missing from cultures of metastases, from the cultures of the malignant cell strain Ca2-83, and from cultures of the loosely adherent aggregates of malignant cells of ductal carcinomas (Rudland et al. 1985). Moreover, Ca2-83 cells fail to produce casein and alveolar-like cells under the requisite hormonal conditions (Rudland, 1987).

The retention of the differentiating ability of the benign neoplastic cells from human breasts and its loss in human carcinoma cells in culture are facts which are consistent with both the pathology of neoplastic breast disease in humans (pp. 98-99) and the above findings from culturing the equivalent rat mammary tumours. The presence of abnormal organoidal structures of epithelial and myoepithelial-like cells in some of the primary ductal carcinomas and their absence in metastatic tumours (Rudland et al. 1985) probably reflects progression of the primary tumour from a less-malignant to a more-malignant phase. As in the rat, these findings in humans are also more likely to be consistent with a mutational event occurring in an epithelial stem cell with gradual truncation of its differentiation pathways during the progressive phase of the disease (Rudland, 1987) than with mutational events occurring simultaneously in the epithelial stem cell and an adjacent nondifferentiating epithelial cell that ultimately gives rise to the malignancy (Taylor-Papadimitriou et al. 1983). The loss of differentiating ability of epithelial stem cells to myoepithelial cells in the normal breast seems to be one of the few consistent changes wrought in the malignant breast cancer cell (Rudland, 1987). Thus the novel regulatory elements postulated in the previous section for transcriptional control of some of the events in the process of differentiation to myoepithelial cells may also be relatively unique in their capacity to become inactivated in the malignant breast cancer cell.

We thank our former colleagues of the now defunct Sutton Branch of the Ludwig Institute for Cancer Research for assistance and helpful discussions, Christine Hughes for expert histological and immunocytochemical assistance, and the North Western Cancer Research and the Cancer and Polio Research Funds for support in Liverpool.