Cytodifferentiation of mouse mammary epithelial cells cultured on a reconstituted basement membrane reveals striking similarities to development in vivo

The development and function of the mammary gland have been subjects of intense interest for many years. Pioneering studies of its morphological differentiation during pregnancy and lactation were among the earliest electron-microscopic investigations to trace the subcellular route of newly synthesized secretory proteins from the rough endoplasmic reticulum (RER) to the trans Golgi face, where secretory vesicles are formed (Wellings et al. 1960; Bargmann, 1962; Kurosumi et al. 1968; Wellings, 1969). During the past decade, many studies have been focussed on both molecular and cellular aspects of milk protein gene expression. A variety of milk protein genes have now been cloned, and their regulatory sequences are beginning to be dissected (Bisbee and Rosen, 1987). At the same time the cellular mechanisms underlying the expression of milk proteins have been studied in some detail, especially through the development of mammary epithelial cell culture systems (Bissell and Hall, 1987). These cell culture models confer many advantages for investigating the mechanisms underlying differentiation and tissue-specific gene expression, including the ability to isolate a single cell type, to manipulate hormonal stimuli and other growth conditions, and to control the microenvironment, including cell-cell and cell-matrix interactions. In the case of the mammary epithelium, such studies have made it clear that the underlying basal lamina and its components exert important control over milk protein expression (Li et al. 1987; Aggeler et al. 1988; Streuli and Bissell, 1990).

Our previous work has indicated that mammary epithelial cells derived from mid-pregnant animals and cultured on a reconstituted basement membrane re-form alveolar-like structures and synthesize and secrete milk proteins into sealed lumina (Barcellos-Hoff et al. 1989). We now report that alveolar re-formation is accompanied by differentiation of these cultured mammary epithelial cells from a mid-pregnant phenotype to one closely resembling lactation. In addition, expression of the tissue-specific milk protein, β-casein, appears to be localized to highly differentiated luminal epithelial cells.

Primary mouse mammary epithelial cells were isolated from midpregnant (13- to 15-day) CD-I mice (Charles Rivers Laboratories, Wilmington, MA), as previously described (Lee et al. 1984; Bissell et al. 1987; Barcellos-Hoff et al. 1989). After mincing and digestion with collagenase and trypsin, epithelial cells were separated by differential low-speed centrifugation, resuspended in F-12 medium containing 10 % fetal bovine serum (both from Gibco, Grand Island, NY), and plated at 2 x10^s to 5 ×10^s cells cm^-2 in culture dishes precoated with a reconstituted basement membrane extract isolated from the Englebreth-Holm-Swarm (EHS) tumor (Kleinman et al. 1986). Serum was removed after 2 days of culture and cells were maintained in serum-free medium for the remainder of the experiment. Fresh medium was added every day. All cultures were carried out in the presence of the lactogenic hormones, ovine prolactin (3 μg ml^-1; National Hormone and Pituitary Program, National Institutes of Health, Bethesda, MD), hydrocortisone (2 μg ml^-1) and insulin (5 μg ml^-1) (both from Sigma Chemical Co., St Louis, MO), which were added daily. The isolated cell preparations contained >90 % epithelial cells (Seely and Aggeler, 1991), as judged by staining for cytokeratins (Lane, 1982).

Pregnant and lactating mammary gland and cultured mammary epithelial cells were prepared for immunofluorescence microscopy or in situ hybridization by fixation with 2% paraformaldehyde, followed by freezing and sectioning with a Leitz cryotome, as previously described (Streuli and Bissell, 1990). Immunofluorescence microscopy was carried out using a β-casein-specific mouse monoclonal antibody (kindly supplied by C. Kaetzel, Case Western Reserve University, Cleveland, OH), followed by biotinylated rabbit anti-mouse IgG and streptavidin-Texas Red (both from Amersham Corp., Arlington Heights, IL). Slides were counterstained with 4’,6’-diamidino-2-phenyl-indole (DAPI) (Sigma) to localize cell nuclei. The distribution of /J-casein mRNA was determined by in situ hybridization, according to the method of Cox et al. (1984). A total of 540 bases of mouse /5-casein cDNA coding sequence (originally isolated by Dr Jeffrey Rosen, Baylor College of Medicine, Houston, TX) were recloned into transcription vectors (PGEM-1, Promega Corp., Madison, WI; and pT₇/T₃a-19, Bethesda Research Labs, Gaithersburg, MD) and high specific activity S-labeled riboprobes (sense and antisense, respectively) were prepared by transcription using bacteriophage T7 RNA polymerase. Frozen sections were applied to organosialinized microscope slides (Turtellotte et al. 1987), treated with proteinase K (0.1 pg ml^-1), postfixed with 2 % paraformaldehyde, acetylated and dehydrated through a series of graded ethanols (Martins-Green and Bissell, 1990). Slides were typically hybridized with 10⁶ctsmin^-1 of labeled probe overnight at 55°C, washed, treated with RNAase A (20 μg ml^-1) in 2xSSC for 30 min, washed again, coated with Kodak NBT-2 photographic emulsion, and exposed at —20 °C for 2–5 days before development and staining with Diff-Quick.

Purification of mRNA and Northern blot analysis were carried out as previously described (Streuli and Bissell, 1990). Gels were run with 2 pg/lane total RNA and loading was judged to be equal, on the basis of ethidium bromide staining of the 18 S and 28 S RNA bands after electrophoresis.

Cells grown in 16-mm diameter wells coated with EHS matrix were fixed with 2% glutaraldehyde in Hanks’ balanced salt solution containing 20mM Hepes (pH6.8) for Ih at room temperature, followed by postfixation with 1 % OsO₄ for 1 h at 4 °C. Fixed cells were dehydrated and embedded in Epon/Araldite in the culture dish, then re-embedded and sectioned perpendicular to the dish, as previously described (Li et al. 1987). Ultrathin sections were stained with lead citrate and examined in a Philips 400 electron microscope at an accelerating voltage of 80 kV. In one experiment, cells were postfixed with 1 % tannic acid for 1 h at room temperature before osmication.

Epithelial cells are isolated from mid-pregnant mouse mammary gland as a mixture of single cells and small clusters of 10—40 cells, which attach to the EHS matrix and appear to pull it up around them, so that they come to be completely surrounded by this basement membrane material (Barcellos-Hoff et al. 1989; Seely and Aggeler, 1991). For the first few days in culture, lumina are either collapsed or absent (Figs 1, 2). By day 4, apical/basal polarity has been re-established and lumina are found in cell aggregates throughout the cultures. Synthesis and vectorial secretion of milk proteins are observed at this time, as seen in sections of cultures stained with a monoclonal antibody against β-casein (Fig. 2B,C). Accumulation of newly synthesized milk proteins in lumina correlates with a twofold increase in lumen size between days 4 and 8 in culture (Fig. 1).

Cells of the mid-pregnant mammary gland are characterized by a relatively high nuclear to cytoplasmic ratio, isolated ER cistemae, well-developed Golgi apparatus and limited ability to synthesize casein, as evidenced by some casein micelle formation (Fig. 3A). After parturition, the synthetic capacity of mammary epithelium is increased markedly, and cells become filled with endoplasmic reticulum, swollen Golgi profiles and secretory vesicles (Fig. 3C). At the same time, milk fat globule secretion is observed. Cells from mid-pregnant mice cultured on the EHS matrix undergo a similar transition from a phenotype characteristic of the pregnant gland on day 4 (Fig. 3B) to one reminiscent of lactating gland from day 6 on (Fig. 3D).

The abundant synthesis and secretion of milk proteins (Fig. 2) suggest that mammary epithelial cells maintained on EHS matrix differentiate from a pregnant to a lactating phenotype in culture. Such differentiation is also indicated by the dramatic increase in the amount of RER during culture (Fig. 4A). In well-differentiated cultures, the hypertrophied RER is present as prominent arrays of ribosome-studded membrane stacks found throughout the cytoplasm (Fig. 4B). In a small proportion of cells (<5 %), the RER appears swollen, and active formation of casein-containing secretory vesicles is not evident (Fig. 4B). Hurley et al. (1989) recently reported that similar swollen RER cistemae in mammary epithelial cells cultured on floating type I collagen gels contain j8-casein, and suggested that the swelling was caused by a block in the secretory protein sorting pathway. That this pathway is intact and functioning normally in EHS cultures is confirmed by the observation of abundant casein micelle formation in the trans Golgi network of these actively synthesizing cells (Fig. 4D; also see Fig. 5C). As shown in Fig. 4A, the differentiation that we observe from the pregnant phenotype to the actively secretory lactating one is not accompanied by an increase in Golgi profiles in these cells. Rather, as shown in Fig. 4C, Golgi stacks in early, non-synthetic cultures appear well-organized, but simply lack any indication of the casein micelle formation characteristic of the fully differentiated later cultures. This is also consistent with the low levels of casein synthesis and secretion observed during the first 2–3 days in culture (Fig. 2; Barcellos-Hoff et al. 1989).

The caseins are a group of phosphoproteins that are secreted in crystalline form as micelles during lactation. One of the most striking characteristics of mouse mammary epithelial cells cultured on EHS matrix is the formation and secretion of abundant casein micelles (Fig. 5A), a functional marker infrequently observed in cultured cells. After 4 days in culture secretory vesicles containing micelles are present in virtually all luminal cells in the apical cytoplasm (Fig. 5A), associated with intracellular lipid droplets (Fig. 5B), and at the trans face of the Golgi apparatus (Fig. 5C). In some synthetically active cells, multiple micelles are observed within a single secretory vesicle (Fig. 5D), as is also observed in lactating gland (Fig. 3C). As has been previously reported (Franke et al. 1976), small areas of clathrin coat were observed on some casein-containing secretory vesicles in a culture postfixed with tannic acid (Fig. 5E). Micelles secreted into the lumen in a day 4 culture had a modal diameter of approximately 80 nm and a size range of 15 to >300 nm (Fig. 6), similar to the values reported for micelles in lactating gland (Wellings et al. 1960; Hollmann, 1974). A smaller sample of intracellular micelles showed a similar distribution, but lacked the very large micelles observed in the lumen, which presumably represent aggregates that form after secretion.

In addition to milk proteins, another important component of milk is fat, secreted as characteristic milk fat globules, which are extruded from mammary epithelial cells surrounded by a sheath of apical plasma membrane. Although fat synthesis and formation of intracellular lipid droplets are observed in mammary epithelial cells during pregnancy, extrusion of well-formed milk fat globules does not occur (Mather, 1987). In contrast, mammary epithelial cells cultured on EHS matrix for 6–8 days show evidence of milk fat globule formation (Fig. 7A) and extrusion into the lumen (Fig. 7B). Between days 2 and 8, total lipid in these cultures increases approximately 2.6-fold, while lipid droplets secreted into lumina increase by > 5-fold (Table 1). Interestingly, intracellular lipid remains relatively abundant during the first 3 days in culture and then decreases dramatically on day 4, a pattern similar to that observed in vivo at parturition (Hollman, 1969,1974). As is observed in the lactating mammary gland (Fig. 7C), milk fat globules formed in culture are extruded from the cell surrounded by a membrane derived from the apical plasma membrane (Fig. 7D).

We have shown previously that mammary epithelial cells cultured on EHS matrix for a week or more contain high levels of mRNA for the milk proteins β-casein (Li et al. 1987) and transferrin (Chen and Bissell, 1987), compared to cells grown on plastic. When we measured mRNA for these proteins during early times in culture on EHS matrix, a striking decrease to almost undetectable levels was observed (Fig. 8, lane 1). By day 3, mRNAs for both proteins were again abundant, demonstrating re-expression of these tissue-specific genes. Using in situ hybridization with an antisense probe for β-casein mRNA, milk protein expression was clearly localized within epithelial cells lining the lumina in EHS cultures and was absent from non-luminal cells (Fig. 9A).

The mammary gland is a rich model system for studying the mechanisms that control tissue-specific gene expression, both because of the number of gene products available and because of the hormone-regulated cyclical nature of this expression. Nevertheless, as with many other complex biological processes, sorting out the relative contributions of a variety of regulatory mechanisms to the induction and modulation of milk protein synthesis has been difficult to accomplish in vivo. For this reason, considerable effort has been expended to develop cell culture system(s) in which these putative regulatory molecules can be tested directly. The present results highlight the power of basement membrane matrices not only to support milk protein gene expression in cultured mammary epithelial cells, but also to permit developmental progression that mimics events in vivo. We now have a cell culture model in which the signals regulating differentiation from a pregnant to a lactating phenotype can be studied.

The intracellular route followed by newly synthesized secretory proteins is well-established, and actively secretory cells, are usually characterized by abundant RER cistemae and well-developed Golgi stacks. Early morphometric studies by Hollman (1969, 1974) indicated that the RER develops during the first half of pregnancy and then remains essentially unchanged during the last week of gestation (approximately days 14 to 21 in the mouse). Beginning at parturition, there is a dramatic hypertrophy of the RER, which increases by threefold during the first 5 days of lactation. This pattern is closely paralleled by the extent of RER development that we observe in mammary epithelial cells cultured on EHS matrix (approximately fourfold increase between days 4 and 6), although the timing of these changes is somewhat accelerated compared to the situation in vivo (3–4 days in culture compared to a week in vivo). Nevertheless, the pattern and extent of RER development support the conclusion that epithelial cells cultured on EHS matrix differentiate from a pregnant to a lactating phenotype.

Alveolar epithelial cells of the lactating mammary gland synthesize a family of calcium-binding phosphoproteins, the caseins, which form characteristic crystalline aggregates or micelles within secretory vesicles at the trans Golgi face of actively secreting cells (Wellings et al. 1960; Hollmann, 1974). These micelles have a fairly broad size distribution, ranging from 40 to 300 nm in diameter in mouse mammary gland (Wellings et al. 1960) to 30 to 600 nm in diameter in bovine milk (McGann et al. 1980). Both the range (15 to 600 nm) and the distribution (modal diameter 80 nm) of casein micelles synthesized by cells cultured on EHS matrix are essentially identical to those found in lactating gland. The size of individual micelles is apparently determined by the proportion of k-casein incorporated, with smaller micelles having more of this casein species than larger ones (McGann et al. 1979,1980). This relationship suggests that, in addition to α- and β-casein (Barcellos-Hoff et al. 1989), k-casein is also being synthesized in abundance by cells cultured on EHS matrix. Casein micelles have occasionally been seen in mammary epithelial cells in organ (Mills and Topper, 1970) or cell culture (Hamamoto et al. 1988; Hahm et al. 1990), but not in the numbers that we observe in cells cultured on EHS matrix. Although some micelles are formed as early as day 8 of pregnancy (Aggeler et al. 1990), significant secretion is not observed, and those micelles that do form during pregnancy may be degraded intracellularly, a form of milk stasis (Hollmann, 1974). Only after parturition do micelles accumulate in large numbers in the alveolar lumina, a condition that is paralleled by cells cultured on EHS matrix for four or more days. The presence of multiple micelles within individual swollen secretory vesicles reflects amounts of casein synthesis similar to those observed in lactating gland, and the marked accumulation of fluid in these vesicles may indicate increased lactose synthesis, as well.

Milk contains fat as a readily available energy source. During pregnancy, some fat synthesis takes place and lipid droplets, sometimes quite large, are commonly found free in the cytoplasm of secretory epithelial cells. This lipid is not usually secreted. Towards the end of gestation, characteristic changes in lipid metabolism take place (Mather, 1987), and abundant lipid synthesis and secretion commences after parturition. Thus, fat secretion is characteristic of lactation. A unique feature of cells cultured on EHS matrix is the secretion of milk fat globules and their accumulation within the lumina of differentiated cultures. This has not previously been documented in cultured mammary epithelial cells and is another indication that cells cultured on EHS matrix differentiate to a lactating phenotype. Milk fat globules are secreted from mammary epithelial cells by extrusion within an envelope of apical plasma membrane (Wellings et al. 1960; Bargmann, 1962; Wellings, 1969; Keenan and Dylewski, 1985; Mather, 1987), and this is also observed in cells cultured on EHS matrix. As has been described in lactating gland in vivo, milk fat globule formation appears to occur when intracytoplasmic lipid droplets approach the apical plasma membrane and become closely apposed to it. Evidence in vivo has suggested that cytoplasmic membranes may also contribute to the milk fat globule membrane (Wooding, 1977), and our observation of casein micelle-containing secretory vesicles surrounding cytoplasmic lipid droplets is consistent with this model. Previous studies have indicated that secretion of two milk fat globule membrane markers, butyrophilin and xanthine oxidase, is increased within cells seeded onto floating collagen gels (Lee et al. 1984) and in the lumina of cultures on EHS matrix (Barcellos-Hoff et al. 1989). Taken together, these results suggest that this experimental model should be useful for studying the unusual secretory mechanism(s) underlying milk fat globule formation.

We have previously reported that cultures of primary mouse mammary epithelial cells plated on a reconstituted basement membrane contain high levels of mRNA for the milk protein genes, /Lcasein (Li et al. 1987), transferrin (Chen and Bissell, 1987) and whey acidic protein (Chen and Bissell, 1989), and that these cultures synthesize and secrete significant amounts of these proteins (Barcellos-Hoff et al. 1989; Chen and Bissell, 1989). Induction of tissue-specific gene expression in cells cultured on this extracellular matrix occurs in parallel with their reorganization into sealed spheres that resemble secretory alveoli in the mammary gland (Barcellos-Hoff et al. 1989). Although it was apparent from these earlier studies that the cells in alveolar structures in culture were typical of a secretory epithelium (apical/basal polarity, tight junctions, well-developed RER and Golgi apparatus), these characteristics are also seen in the pregnant gland; thus, a more detailed investigation of this cytodifferentiation has been necessary. In fact, we find that these highly differentiated epithelial cells are more similar morphologically to those of lactating mammary gland than they are to those of the mid-pregnant glands from which they were isolated. Most interestingly, in situ hybridization studies have revealed that it is precisely the differentiated secretory cells lining the lumina within the alveolar structures in our cultures that contain abundant mRNA for the specific milk protein, -casein. The striking reorganization and functional segregation of secretory cells induced by the extracellular matrix provides further evidence that histiotypic architecture and concomitant cytoplasmic differentiation are important determinants of cell function. Thus, these studies confirm at the level of individual cells our hypothesis that cell structure is an important determinant of tissue-specific gene expression (Bissell et al. 1982; Aggeler et al. 1984; Bissell and Hall, 1987; Aggeler et al. 1988).

The authors thank Dr Dallas Hyde, University of California, Davis, CA, for his help in designing and analyzing the morphometric studies and Dr Gordon Parry, Lawrence Berkeley Labs, Berkeley, CA, for helpful discussion of this work. The electron micrographs in this study were taken at the Electron Microscopy Laboratory of the School of Medicine, University of California, Davis, CA. This work was supported by grants to J.A. from the National Dairy Council and the California Milk Producers Advisory Board through the California Dairy Food Research Center and from the National Science Foundation (no. DCB-87 19874), by a grant to M.J.B. from the Health Effects Research Division, Office of Health and Environmental Research, U.S. Department of Energy, contract no. DE-AC03–76SF00098, and by a gift for research from the Monsanto Company. C.H.S. was supported by fellowships from the European Molecular Biology Organization and the Fogarty International Center.