EXPERIMENTAL
CELL
RESEARCH
201, 225-234 (19%)
Mammary Epithelial Reorganization on Extracellular Matrix Is Mediated by Cell Surface Galactosyltransferase M. Building
74-159A, Cell and Molecular
Biology Division,
H. BARCELLOS-HOFF Lawrence
Berkeley Laboratory,
thelial When plated at appropriate densities in serum-free media, the COMMA-D mammary epithelial cell line rapidly reorganizes into multicellular spheres on the basement membrane matrix derived from EngelbrethHolm-Swarm murine tumor. Using time-lapse videoof reorganization were four stages microscopy, discerned during the first 24 h of culture. In the first few hours, cells attached to the matrix, elongated, migrated, and formed chains. In the next 6 h, chains of cells linked together in anastomosing networks. In the period between 8 and 18 h postplating, the networks contracted, resulting in dense cords radiating from central aggregates. During the final 6 h, the cords were drawn into the aggregates, which condensed further into spheres. The events occurring during mammary epithelial cell reorganization on the matrix were shown to be mediated by cell surface B-1,4-galactosyltransferase (GalTase), a receptor that binds iV-acetylglucosamine residues on glycosylated proteins. GalTase activity was evident at the surface of cells cultured on reconstituted matrix for 3 h but was absent from cells on glass. The protein a-lactalbumin (a-LA) inhibits the association of GalTase with N-acetylglucosamine. a-LA present from the beginning of culture on reconstituted matrix had no effect on cell attachment but caused concentration-dependent inhibition of the first two steps of reorganization, i.e., cell elongation and network formation, which then interfered with subsequent events. These observations were replicated using polyclonal antibodies to GalTase. Reorganization was impaired when a-LA was added during the first two stages but no effect was observed when it was added during the last two stages. Cells cultured on plastic, which lack surface GalTase activity, were unperturbed by incubation with a-LA. Thus certain events (cell elongation and network elaboration) during mammary epithelial cell reorganization on reconstituted matrix are GalTase dependent, while others (attachment, network contraction, and compaction) are not. The functional and temporal specificity of GalTase involvement indicates that GalTase mediates cell-matrix, but not cell-cell, interactions during epi-
Press,
University
morphogenetic
of California,
events
in
Berkeley,
culture.
California
0
1992
94720
Academic
Inc.
INTRODUCTION
Morphogenesis of epithelial tissues is dependent upon, and directed in part by, the extracellular matrix (ECM).’ Adhesion, migration, and condensation are distinct aspects of this multifaceted process, which most likely involves multiple receptors binding to distinct regions of various extracellular matrix proteins. The relative importance of each component in terms of its function during discrete phases of morphogenesis has yet to be defined, but it is thought that a balance between ECM adherence and cell-cell adhesion is critical. The concept that ECM composition may be informative during morphogenesis has been supported by studies in culture. Although epithelial cell culture on plastic or glass generally leads to a spread cell morphology and decreased tissue-specific function, culturing cells on ECM proteins has profound effects on both morphology and function [ 11. In particular, a number of studies have demonstrated that cell culture on a complex basement membrane-like matrix derived from the EngelbrethHolm-Swarm (EHS) murine sarcoma promotes histiotypic function and morphology for a variety of primary epithelial cells. Rat hepatocytes form cords that maintain albumin secretion [a], human endothelial cells form capillary-like structures with lumina and express factor VIII-related antigens [3], and mouse mammary epithelial cells reorganize into hollow, alveolar-like structures into which significant quantities of milk proteins are secreted vectorially [4]. In each of these examples, polarized cytoplasmic organization and membrane differentiation at the level of individual cells accompany ’ Abbreviations used: a-LA, n-lactalbumin; ECM, extracellular matrix; EHS, Engelbreth-Holm-Swarm tumor; GalTase, P-1,4-galactosyltransferase; GlcNAc, N-acetylglucosamine.
225 All
Copyright 0 1992 rights of reproduction
0014.4827/92 $5.00 by Academic Press, Inc. in any form reserved.
226
M. H. BARCELLOS-HOFF
multicellular reorganization into tissue-specific architecture. Studies using primary mammary epithelial cells indicate that multicellular architecture has a role in regulating gene expression of some tissue-specific proteins [5], and certainly some functions of epithelial cells, such as compartmentalization, can only be accomplished in three dimensions [4]. Electron microscopic studies of the reconstitution of a luminal compartment by mammary epithelial cells indicate that reorganization is effected via migration, matrix remodeling, and cell-cell interactions that culminate in the expression of many characteristics associated with the lactating phenotype [6]. However, it is difficult to analyze specific events leading to reorganization into alveolar-like spheres by primary mammary epithelial cultures because small clusters, rather than single cells, are usually plated. Thus, the mechanisms by which epithelial cells interact with the ECM and each other to establish such complex spatial organization are not well described or understood. The mammary epithelial COMMA-D cells, which form a monolayer of epithelial cells when cultured on plastic, undergo reorganization similar to primary mammary epithelial cells and demonstrate increased tissue-specific gene expression when cultured on reconstituted EHS matrix [7]. The COMMA-D cell line also retains the ability to undergo normal mammary gland morphogenesis in a cleared mammary fat pad in uiuo [8]. In this paper, I analyzed the early events (~24 h) leading to COMMA-D sphere formation on reconstituted basement membrane matrix in serum-free media. Using phase contrast time-lapse videomicroscopy, four stages of reorganization following attachment were observed: (i) cell elongation; (ii) formation of networks of chained cells, (iii) contraction of the networks; and (iv) compaction of the resulting aggregates into spheres. The rapid reorganization makes this system amenable for dissection of the molecular mechanisms by which individual stages are guided. The role of cell surface P-1,4-galactosyltransferase (GalTase) during epithelial morphogenetic reorganization was interesting to examine for two reasons: first, it does not appear to be involved in cell attachment [lo]. Second, GalTase participates in both cell-matrix [g-11] and cell-cell [ 12, 131 interactions. Cell surface GalTase is a transmembrane protein that recognizes N-acetylglucosamine (GlcNAc) residues on other cells [12] and in the ECM [lo], for example on laminin fragments [ 111. GalTase is a member of a large family of Golgi enzymes that add oligosaccharide chains to complex glycoconjugates. Golgi GalTase transfers galactose from UDP-galactose to GlcNAc residues [ 141. Cell surface GalTase is enzymatically inactive in the absence of the nucleotide sugar, which is not found extracellularly.
It has been suggested that a distinct pool of GalTase, which is regulated differently from the Golgi pool, is available for cell surface localization [ 151. I examined whether COMMA-D mammary cell reorganization on reconstituted EHS matrix is GalTase dependent by using agents to inhibit the interaction between GalTase and its ligand, GlcNAc carbohydrate residues. a-Lactalbumin (a-LA) binds GalTase and alters its substrate affinity for GlcNAc residues in favor of glucose [lo, 141, which inhibits cell surface GalTasemediated events in a variety of systems [13, 16, 171. Anti-GalTase antibodies were also used to block the interaction between GalTase and a-LA [12,18,19]. Both agents blocked cell elongation on EHS matrix in a temporally specific and concentration-dependent manner. These experiments indicated that GalTase is actively involved in mammary epithelial cell reorganization on reconstituted matrix. Certain events, cell elongation and network elaboration, were GalTase dependent, while others, attachment, network contraction, and compaction, were not. Taken together, these data suggest that GalTase specifically modulates cell-matrix interactions but does not appear to be involved in cell-cell interactions in this model of epithelial morphogenesis. MATERIALS
AND
METHODS
Reagents. Reconstituted EHS basement membrane matrix (gift of Dr. M. J. Bissell, Lawrence Berkeley Laboratory, Berkeley, CA) was prepared according to published methods [20] from an EHS tumor that was maintained intramuscularly in lathrytic C57/bl male mice. Affinity-purified rabbit anti-GalTase IgG (gift of Dr. B. D. Shur, M. D. Anderson Cancer Center, Houston, TX) was raised against affinity-purified bovine milk GalTase [21]. All other reagents were obtained from Sigma Chemical Co. (St. Louis, MO) unless otherwise noted. o-LA was dialyzed twice against distilled water and the stock solution (25 mg/ml in phosphate-buffered saline, pH 7.4) was stored at 4°C. Cell culture. COMMA-D mouse mammary epithelial cells frozen at passage 11 were obtained from Dr. M. J. Bissell (Lawrence Berkeley Laboratory). Cells were grown routinely in 1:l F12:DMEM supplemented with 5% fetal calf serum, insulin, transferrin, and selenium on tissue culture plastic in a 7.5% CO1 humidified incubator at 37°C and passaged twice weekly. Cells were used before passage 18. Because serum contains abundant soluble GalTase, GalTase perturbation experiments were carried out in serum-free medium; plating efficiency and initial growth on reconstituted matrix with or without serum were similar. The absence of serum from the culture media had no apparent effect on the rate or sequence of events described below. EHS reconstituted matrix-coated dishes were prepared by spreading 12.5 pi/cm’ of cold reconstituted matrix (5-7 mg/ml) onto cold dishes. The matrix was allowed to gel at 37’C for about 1 h before cells were plated at a density of 5 X 104/cm2. GalTuse actiuity. In some experiments the reconstituted EHS matrix was galactosylated according the method of Begovac and Shur [19]. EHS matrix (12.5 pi/cm’) was incubated with bovine GalTase (20 pg/ml) in F12:DMEM containing 2 mM UDP-galactose and 10 mM MnCl, for 24 h at 37°C with three changes of reagents. Control dishes were treated in parallel without GalTase or UDP-galactose. The relative amount of GalTase binding sites in native reconstituted
GALTASE
MEDIATES
EPITHELIAL
matrix following this protocol was assessed by substituting a buffer (127 mM NaCl, 5.3 n&f KCl, 18.2 r&4 Hepes) for medium and UDP[3H]galactose (10 PCi; 20 CilmM, 35 mCi/mg; Amersham) for unlabeled UDP-galactose during a final 2-h incubation. The substrata were washed four times and the radioactivity was solubilized in 1 N NaOH followed by liquid scintillation counting. To determine cell surface GalTase activity, lo6 cells in 2 ml of serum-free medium with either (U-LA or bovine serum albumin (5 mg/ml) were plated into 2-well chamber slides that had been partially coated with reconstituted matrix. After 3 h of incubation, 1 ml of medium was withdrawn and 25 ~1 containing 2.5 &i UDP[3H]galactose and 250 FM UDP-galactose was added. The cultures were incubated for 1 h, rinsed twice with phosphate-buffered saline, and fixed for 5 min with 4% paraformaldehyde in phosphate-buffered saline. Following two rinses with phosphate-buffered saline containing 10 mM glycine the cultures were dehydrated, air-dried, and processed for autoradiography. Cell reorganization and elongation assay. At various times after plating, dishes were examined and photographed on a Nikon inverted phase contrast microscope. Only during the first phase of reorganization (t3 h) were individual cells sufficiently distinct to assess their shape. Deviations from round, phase bright refractive morphology were quantitated by scoring photographs of cells at l-2 h postplating. Elongated, linked cells were included in these scores as well as single cells; clusters of more than six cells were excluded. Data are expressed as the ratio of the control to the experimental percentage of cells deviating from round morphology. Cell attachment efficiency was assessed in some cultures by removing an aliquot of the medium after gentle mixing and counting the cells in the sample using a Coulter counter. Cells examined using videomicroTime-lapse uideomicroscopy. scopy were allowed to equilibrate for l-2 h in medium containing 10 mM Hepes in the standard incubator. The dishes were then sealed with parafilm and placed on a phase contrast inverted microscope stage warmed to 37°C by continuous forced air during recording. These images confirmed the impressions gained from periodic static phase contrast photomicrographs and allowed the continuous observation of one field of cells. -
RESULTS
Morphogenetic Matrix
Reorganization
on EHS-Reconstituted
COMMA-D cells are routinely cultured on tissue culture plastic where they assume a typical monolayer appearance early in culture. When plated at a density of 5 X lo4 cells/cm2 on matrix-coated dishes, the majority of the cells reorganize during the first 24 h into small multicellular spheres (Fig. 1). When this process was observed using videomicroscopy (Fig. 2), four stages of reorganization could be discerned. Attachment occurred rapidly (~30 min) and during the first few hours (2-3 h) cells began to elongate and form chains of cells aligned end to end (Fig. 2A). During the next 6 h on reconstituted matrix, the cells underwent further chain formation and began forming anastomosing networks (Fig. 2B). These networks were formed initially by migration of individual cells, which then forged linkages between chains (Fig. 2C). The majority of cells were drawn into these networks, so that few cells were isolated and vir-
REORGANIZATION
227
tually none spread. The next 6- to 8-h period was characterized by the active contraction of networks, so that chains of cells became denser and cord-like, and thickened centers were formed of aggregated cells (Figs. 2D and 2E). In comparison, cells on plastic continued to spread and form a monolayer (not shown). In the final 6 h of observations, the aggregates compacted and the remaining cords were drawn into the cell mass, resulting in roughly spherical multicellular aggregates (Figs. 2F and 1D). This architecture is stable for over a week in culture, although the aggregates enlarge. This is probably due to cell proliferation since the growth rate of cells cultured on EHS matrix is similar to that of cells on plastic, exhibiting a population doubling time of approximately 24 h (data not shown). Cell Surface GalTase Enzymatic Activity Is Substratum Dependent The rapidity with which cells reorganized from single cells to spheres on reconstituted matrix, and the distinct stages of reorganization elucidated using videomicroscopy, provides an opportunity to determine what molecular mechanisms are involved in specific events in this sequence. I tested whether GalTase was involved by asking whether it was expressed on the cell surfaces of COMMA-D cells under a variety of conditions. To localize extracellular GalTase activity, COMMA-D cells were cultured on EHS matrix or plastic for 3 h, then incubated with radiolabeled UDP-galactose, its conventional sugar donor, which is usually absent extracellularly. Cultures were then processed for autoradiography. Grains were observed overlying the surface of many cells cultured on EHS matrix (Figs. 3A and 3B) but were absent from cells on plastic substrata (Fig. 3D). Furthermore, the addition of a-LA to the cells cultured on matrix prevented expression of galactosylation (Fig. 3C), which is consistent with its modifing GalTase substrate specificity away from GlcNAc residues. The appearance of silver grains over cells cultured on EHS matrix, but absence in the presence of a-LA or when cells were cultured on glass, indicates that GalTase expression at the cell surface is substratum dependent. However, this method does not indicate where or what the galactosylated ligand is. The presence of GalTase ligand, i.e., GlcNAc carbohydrate residues, in reconstituted EHS matrix was confirmed by galactosylating the matrix using soluble GalTase and radiolabeled UDPgalactose (2.4 X lo5 cpm/well vs 828 cpm/well in the absence of GalTase), as previously shown [9]. GalTase Substrate Modifier Inhibits Specific Stages of Reorganization To examine whether cell surface GalTase plays a functional role in mammary epithelial cell interactions during reorganization on matrix, a-LA was added to serum-free medium during cell culture. (U-LA forms a
228
M. H. BARCELLOS-HOFF
FIG. 1. Cells cultured on reconstituted matrix undergo dramatic reorganization during the first 24 h of culture. Four stages of reorganization are evident from phase contrast photomicrographs of COMMA-D mammary epithelial cells cultured on reconstituted matrix. At 2 h (A), cells have elongated and chained, 4 h (B), networks have formed; 7.5 h (C), networks have begun to contract; 23 h (D), the networks have condensed into spheres.
complex with GalTase that has decreased affinity for GlcNAc residues and inhibits its binding to extracellular matrices [9,19]. Addition of &-LA to the medium at the time of plating dramatically altered the ability of the cells to undergo the sequence of events described above. Videomicroscopy revealed that assumption of the elongated cell shape was inhibited and that migration was affected. Rather than elongating and moving along one axis, the cells remained rounded, yet adherent (Figs. 4A-4D). Cells that were close together were able to make short chains, but were not able to recruit more distant cells. However, further observation of such cultures revealed that, although the contraction and compaction phases of morphogenesis was attenuated due to the
short chain formation, the later stages of reorganization were able to proceed in areas of high cell density (Figs. 4E-4H). Thus, although a-LA blocked cell elongation and network formation, if the cells happened to be very close to each other, and thus need not elongate to make contact, then they were able to aggregate, albeit in a curtailed fashion. Additionally, if a-LA was added at 2 or 4 h postplating, during the chaining and networking phases, then the cultures were inhibited at that point in the reorganization, but addition of (U-LA after 6 h, i.e., when networks were established, had no apparent effect on reorganization. Matrices galactosylated prior to cell culture were examined in order to test whether altering the GalTase recognition site (i.e., GlcNAc residues) would have the
GALTASE
MEDIATES
EPITHELIAL
229
REORGANIZATION
FIG. 2. Time-lapse videomicroscopy reveals the sequence by which mammary epithelial cells reorganize into spheres during the first 24 h of culture on reconstituted matrix. Photomicrographs of video images are shown of COMMA-D reorganization on reconstituted matrix. Cells were plated at 10:00 AM. (A) 3 h, (B) 7 h, (Cl 9 h, (D) 11.5 h, (El 13 h, and (F) 23 h postplating.
same effect as inhibiting ceil surface GalTase [lo]. However, this experiment was not informative since the control matrices, which received the extensive incubations and washing without any GalTase present, failed to elicit the typical reorganization (not shown).
The Effects of Inhibitors of GalTase Activity Concentration Dependent
Are
Cell attachment and subsequent elongation were monitored to quantitate the concentration response ef-
230
M. H. BARCELLOS-HOFF
FIG. 3. Cell surface GalTase enzymatic activity is detectable only on cells cultured on reconstituted matrix. Nomarski optics photomicrographs of cells cultured on reconstituted matrix (A and B), or on reconstituted matrix in the presence of a-LA, 1 mg/ml (C), or on glass (D) and labeled with UDP-[3H]galactose. Silver grains were deposited heavily over many cells cultured on reconstituted matrix (A, B). No grains are deposited over cells on glass (D) or rounded cells on reconstituted matrix cultured in the presence of o-LA (C).
fects of GalTase inhibitors on cell functions. a-LA or polyclonal GalTase antibodies added at the time of plating inhibited cell elongation in a concentration-dependent manner (Fig. 5). Cell elongation was greatly curtailed by various concentrations of a-LA or antibodies to GalTase when used at concentrations shown by others to inhibit B16-FlO cell spreading [lo]. Cells remained rounded, although firmly attached, in the presence of a-LA. The number of cells attached to the substratum was the same in the controls as when incubated with a-LA for 2 or 24 h. Addition of up to 10 mg/ml of (Y-LA at any time to cells cultured on plastic had no effect on cell spreading or monolayer morphology. Although the effects were less pronounced, antibodies to GalTase inhibited events subsequent to elongation in a
manner similar to that described for a-LA. Thus altering GalTase substrate specificity by a-LA or masking cell surface GalTase by antibody binding had qualitative and quantitatively similar effects. DISCUSSION
The mechanisms by which epithelia organize during development and reorganize at specific times postnatally are a fundamental puzzle in biology. Specific interactions between cells or between cells and their ECM occur both sequentially and concordantly during morphogenesis to bring the appropriate cells into contact to form tissue-specific structures. Cell culture systems have been employed in an attempt to dissect some of
GALTASE
MEDIATES
EPITHELIAL
REORGANIZATION
231
FIG. 4. Perturbation of GalTase specifically inhibits the first two stages of reorganization, which impedes subsequent events. Photomicrographs of video images of cells plated on reconstituted matrix at lo:30 AM in the presence of 1 mg/ml (U-LA. One field of observation at (A) 2 h, (B) 3 h, (C) 5 h, or (D) 6 h postplating. Note that cells remain rounded for the most part. A second field in the same dish demonstrating that some networks form if cells are adjacent to one another and that such networks can contract in the presence of a-LA: (E) 7 h, (F) 9 h, (G) 13 h, (H) 17 h postplating.
232
M. H. BARCELLOS-HOFF
0.2 -
1 ,.,: 0
1
Concentration,
2
3
mglml
FIG. 5. Modifiers of GalTase quantitatively inhibit cell elongation. Concentration-dependent effects of (Y-LA (squares; control was bovine serum albumin; n = 97) and anti-GalTase IgG (circles; control was normal rabbit serum; n = 194) are expressed as the percentage of control of nonround cells under the experimental condition. The symbols represent the mean of two determinations.
these events by analyzing adherence of cells to each other or to ECM proteins. The observation that culture on EHS-reconstituted matrix elicits histiotypic reorganization by primary cells from adult epithelia [2,3, 221, and even cell lines as reported here and elsewhere [23], is intriguing. This process of reorganization provides a more complex, but potentially more informative, model of cell-ECM interactions than cells cultured as monolayers. EHS-reconstituted matrix is composed of laminin (~SO%), collagen IV, heparan sulfate proteoglycan, and entactin [20]. Laminin has been widely implicated in cell adhesion and migration [17, 241, but it has been noted in several studies that individual matrix components do not effectively substitute for the complex matrix derived from the EHS tumor in terms of cell reorganization [2, 3, 221. It has been suggested that both protein concentration [3,5] and matrix flexibility [4,5] are important factors for eliciting morphogenic reorganization. Using videomicroscopy it is possible to distinguish four stages of reorganization following COMMA-D mammary epithelial cells attachment to reconstituted matrix: (i) cell elongation and chaining; (ii) network formation; (iii) network contraction; and (iv) compaction of aggregated networks. None of these events occur when COMMA-D cells are cultured on plastic, on which cells rapidly spread and form monolayers, or on colla-
gen gels [26] or PHFR-9 cell-secreted ECM [27]. Similar sequences, although different time frames, have been described for a cell line from small intestinal epithelium [23] and for fetal type 2 pneumocytes [25]. Each stage involves several distinct events. The first stage requires cell shape changes, single cell migration, and establishment of cell-cell contacts. These processes are also important in the second phase, network formation, which is characterized by chained cells moving together and establishing a network of cell-cell contacts. A similar course of events has been described for formation of capillary-like tubules by endothelial cells cultured on reconstituted matrix, which elicits rapid (
CELL
RESEARCH
201, 225-234 (19%)
Mammary Epithelial Reorganization on Extracellular Matrix Is Mediated by Cell Surface Galactosyltransferase M. Building
74-159A, Cell and Molecular
Biology Division,
H. BARCELLOS-HOFF Lawrence
Berkeley Laboratory,
thelial When plated at appropriate densities in serum-free media, the COMMA-D mammary epithelial cell line rapidly reorganizes into multicellular spheres on the basement membrane matrix derived from EngelbrethHolm-Swarm murine tumor. Using time-lapse videoof reorganization were four stages microscopy, discerned during the first 24 h of culture. In the first few hours, cells attached to the matrix, elongated, migrated, and formed chains. In the next 6 h, chains of cells linked together in anastomosing networks. In the period between 8 and 18 h postplating, the networks contracted, resulting in dense cords radiating from central aggregates. During the final 6 h, the cords were drawn into the aggregates, which condensed further into spheres. The events occurring during mammary epithelial cell reorganization on the matrix were shown to be mediated by cell surface B-1,4-galactosyltransferase (GalTase), a receptor that binds iV-acetylglucosamine residues on glycosylated proteins. GalTase activity was evident at the surface of cells cultured on reconstituted matrix for 3 h but was absent from cells on glass. The protein a-lactalbumin (a-LA) inhibits the association of GalTase with N-acetylglucosamine. a-LA present from the beginning of culture on reconstituted matrix had no effect on cell attachment but caused concentration-dependent inhibition of the first two steps of reorganization, i.e., cell elongation and network formation, which then interfered with subsequent events. These observations were replicated using polyclonal antibodies to GalTase. Reorganization was impaired when a-LA was added during the first two stages but no effect was observed when it was added during the last two stages. Cells cultured on plastic, which lack surface GalTase activity, were unperturbed by incubation with a-LA. Thus certain events (cell elongation and network elaboration) during mammary epithelial cell reorganization on reconstituted matrix are GalTase dependent, while others (attachment, network contraction, and compaction) are not. The functional and temporal specificity of GalTase involvement indicates that GalTase mediates cell-matrix, but not cell-cell, interactions during epi-
Press,
University
morphogenetic
of California,
events
in
Berkeley,
culture.
California
0
1992
94720
Academic
Inc.
INTRODUCTION
Morphogenesis of epithelial tissues is dependent upon, and directed in part by, the extracellular matrix (ECM).’ Adhesion, migration, and condensation are distinct aspects of this multifaceted process, which most likely involves multiple receptors binding to distinct regions of various extracellular matrix proteins. The relative importance of each component in terms of its function during discrete phases of morphogenesis has yet to be defined, but it is thought that a balance between ECM adherence and cell-cell adhesion is critical. The concept that ECM composition may be informative during morphogenesis has been supported by studies in culture. Although epithelial cell culture on plastic or glass generally leads to a spread cell morphology and decreased tissue-specific function, culturing cells on ECM proteins has profound effects on both morphology and function [ 11. In particular, a number of studies have demonstrated that cell culture on a complex basement membrane-like matrix derived from the EngelbrethHolm-Swarm (EHS) murine sarcoma promotes histiotypic function and morphology for a variety of primary epithelial cells. Rat hepatocytes form cords that maintain albumin secretion [a], human endothelial cells form capillary-like structures with lumina and express factor VIII-related antigens [3], and mouse mammary epithelial cells reorganize into hollow, alveolar-like structures into which significant quantities of milk proteins are secreted vectorially [4]. In each of these examples, polarized cytoplasmic organization and membrane differentiation at the level of individual cells accompany ’ Abbreviations used: a-LA, n-lactalbumin; ECM, extracellular matrix; EHS, Engelbreth-Holm-Swarm tumor; GalTase, P-1,4-galactosyltransferase; GlcNAc, N-acetylglucosamine.
225 All
Copyright 0 1992 rights of reproduction
0014.4827/92 $5.00 by Academic Press, Inc. in any form reserved.
226
M. H. BARCELLOS-HOFF
multicellular reorganization into tissue-specific architecture. Studies using primary mammary epithelial cells indicate that multicellular architecture has a role in regulating gene expression of some tissue-specific proteins [5], and certainly some functions of epithelial cells, such as compartmentalization, can only be accomplished in three dimensions [4]. Electron microscopic studies of the reconstitution of a luminal compartment by mammary epithelial cells indicate that reorganization is effected via migration, matrix remodeling, and cell-cell interactions that culminate in the expression of many characteristics associated with the lactating phenotype [6]. However, it is difficult to analyze specific events leading to reorganization into alveolar-like spheres by primary mammary epithelial cultures because small clusters, rather than single cells, are usually plated. Thus, the mechanisms by which epithelial cells interact with the ECM and each other to establish such complex spatial organization are not well described or understood. The mammary epithelial COMMA-D cells, which form a monolayer of epithelial cells when cultured on plastic, undergo reorganization similar to primary mammary epithelial cells and demonstrate increased tissue-specific gene expression when cultured on reconstituted EHS matrix [7]. The COMMA-D cell line also retains the ability to undergo normal mammary gland morphogenesis in a cleared mammary fat pad in uiuo [8]. In this paper, I analyzed the early events (~24 h) leading to COMMA-D sphere formation on reconstituted basement membrane matrix in serum-free media. Using phase contrast time-lapse videomicroscopy, four stages of reorganization following attachment were observed: (i) cell elongation; (ii) formation of networks of chained cells, (iii) contraction of the networks; and (iv) compaction of the resulting aggregates into spheres. The rapid reorganization makes this system amenable for dissection of the molecular mechanisms by which individual stages are guided. The role of cell surface P-1,4-galactosyltransferase (GalTase) during epithelial morphogenetic reorganization was interesting to examine for two reasons: first, it does not appear to be involved in cell attachment [lo]. Second, GalTase participates in both cell-matrix [g-11] and cell-cell [ 12, 131 interactions. Cell surface GalTase is a transmembrane protein that recognizes N-acetylglucosamine (GlcNAc) residues on other cells [12] and in the ECM [lo], for example on laminin fragments [ 111. GalTase is a member of a large family of Golgi enzymes that add oligosaccharide chains to complex glycoconjugates. Golgi GalTase transfers galactose from UDP-galactose to GlcNAc residues [ 141. Cell surface GalTase is enzymatically inactive in the absence of the nucleotide sugar, which is not found extracellularly.
It has been suggested that a distinct pool of GalTase, which is regulated differently from the Golgi pool, is available for cell surface localization [ 151. I examined whether COMMA-D mammary cell reorganization on reconstituted EHS matrix is GalTase dependent by using agents to inhibit the interaction between GalTase and its ligand, GlcNAc carbohydrate residues. a-Lactalbumin (a-LA) binds GalTase and alters its substrate affinity for GlcNAc residues in favor of glucose [lo, 141, which inhibits cell surface GalTasemediated events in a variety of systems [13, 16, 171. Anti-GalTase antibodies were also used to block the interaction between GalTase and a-LA [12,18,19]. Both agents blocked cell elongation on EHS matrix in a temporally specific and concentration-dependent manner. These experiments indicated that GalTase is actively involved in mammary epithelial cell reorganization on reconstituted matrix. Certain events, cell elongation and network elaboration, were GalTase dependent, while others, attachment, network contraction, and compaction, were not. Taken together, these data suggest that GalTase specifically modulates cell-matrix interactions but does not appear to be involved in cell-cell interactions in this model of epithelial morphogenesis. MATERIALS
AND
METHODS
Reagents. Reconstituted EHS basement membrane matrix (gift of Dr. M. J. Bissell, Lawrence Berkeley Laboratory, Berkeley, CA) was prepared according to published methods [20] from an EHS tumor that was maintained intramuscularly in lathrytic C57/bl male mice. Affinity-purified rabbit anti-GalTase IgG (gift of Dr. B. D. Shur, M. D. Anderson Cancer Center, Houston, TX) was raised against affinity-purified bovine milk GalTase [21]. All other reagents were obtained from Sigma Chemical Co. (St. Louis, MO) unless otherwise noted. o-LA was dialyzed twice against distilled water and the stock solution (25 mg/ml in phosphate-buffered saline, pH 7.4) was stored at 4°C. Cell culture. COMMA-D mouse mammary epithelial cells frozen at passage 11 were obtained from Dr. M. J. Bissell (Lawrence Berkeley Laboratory). Cells were grown routinely in 1:l F12:DMEM supplemented with 5% fetal calf serum, insulin, transferrin, and selenium on tissue culture plastic in a 7.5% CO1 humidified incubator at 37°C and passaged twice weekly. Cells were used before passage 18. Because serum contains abundant soluble GalTase, GalTase perturbation experiments were carried out in serum-free medium; plating efficiency and initial growth on reconstituted matrix with or without serum were similar. The absence of serum from the culture media had no apparent effect on the rate or sequence of events described below. EHS reconstituted matrix-coated dishes were prepared by spreading 12.5 pi/cm’ of cold reconstituted matrix (5-7 mg/ml) onto cold dishes. The matrix was allowed to gel at 37’C for about 1 h before cells were plated at a density of 5 X 104/cm2. GalTuse actiuity. In some experiments the reconstituted EHS matrix was galactosylated according the method of Begovac and Shur [19]. EHS matrix (12.5 pi/cm’) was incubated with bovine GalTase (20 pg/ml) in F12:DMEM containing 2 mM UDP-galactose and 10 mM MnCl, for 24 h at 37°C with three changes of reagents. Control dishes were treated in parallel without GalTase or UDP-galactose. The relative amount of GalTase binding sites in native reconstituted
GALTASE
MEDIATES
EPITHELIAL
matrix following this protocol was assessed by substituting a buffer (127 mM NaCl, 5.3 n&f KCl, 18.2 r&4 Hepes) for medium and UDP[3H]galactose (10 PCi; 20 CilmM, 35 mCi/mg; Amersham) for unlabeled UDP-galactose during a final 2-h incubation. The substrata were washed four times and the radioactivity was solubilized in 1 N NaOH followed by liquid scintillation counting. To determine cell surface GalTase activity, lo6 cells in 2 ml of serum-free medium with either (U-LA or bovine serum albumin (5 mg/ml) were plated into 2-well chamber slides that had been partially coated with reconstituted matrix. After 3 h of incubation, 1 ml of medium was withdrawn and 25 ~1 containing 2.5 &i UDP[3H]galactose and 250 FM UDP-galactose was added. The cultures were incubated for 1 h, rinsed twice with phosphate-buffered saline, and fixed for 5 min with 4% paraformaldehyde in phosphate-buffered saline. Following two rinses with phosphate-buffered saline containing 10 mM glycine the cultures were dehydrated, air-dried, and processed for autoradiography. Cell reorganization and elongation assay. At various times after plating, dishes were examined and photographed on a Nikon inverted phase contrast microscope. Only during the first phase of reorganization (t3 h) were individual cells sufficiently distinct to assess their shape. Deviations from round, phase bright refractive morphology were quantitated by scoring photographs of cells at l-2 h postplating. Elongated, linked cells were included in these scores as well as single cells; clusters of more than six cells were excluded. Data are expressed as the ratio of the control to the experimental percentage of cells deviating from round morphology. Cell attachment efficiency was assessed in some cultures by removing an aliquot of the medium after gentle mixing and counting the cells in the sample using a Coulter counter. Cells examined using videomicroTime-lapse uideomicroscopy. scopy were allowed to equilibrate for l-2 h in medium containing 10 mM Hepes in the standard incubator. The dishes were then sealed with parafilm and placed on a phase contrast inverted microscope stage warmed to 37°C by continuous forced air during recording. These images confirmed the impressions gained from periodic static phase contrast photomicrographs and allowed the continuous observation of one field of cells. -
RESULTS
Morphogenetic Matrix
Reorganization
on EHS-Reconstituted
COMMA-D cells are routinely cultured on tissue culture plastic where they assume a typical monolayer appearance early in culture. When plated at a density of 5 X lo4 cells/cm2 on matrix-coated dishes, the majority of the cells reorganize during the first 24 h into small multicellular spheres (Fig. 1). When this process was observed using videomicroscopy (Fig. 2), four stages of reorganization could be discerned. Attachment occurred rapidly (~30 min) and during the first few hours (2-3 h) cells began to elongate and form chains of cells aligned end to end (Fig. 2A). During the next 6 h on reconstituted matrix, the cells underwent further chain formation and began forming anastomosing networks (Fig. 2B). These networks were formed initially by migration of individual cells, which then forged linkages between chains (Fig. 2C). The majority of cells were drawn into these networks, so that few cells were isolated and vir-
REORGANIZATION
227
tually none spread. The next 6- to 8-h period was characterized by the active contraction of networks, so that chains of cells became denser and cord-like, and thickened centers were formed of aggregated cells (Figs. 2D and 2E). In comparison, cells on plastic continued to spread and form a monolayer (not shown). In the final 6 h of observations, the aggregates compacted and the remaining cords were drawn into the cell mass, resulting in roughly spherical multicellular aggregates (Figs. 2F and 1D). This architecture is stable for over a week in culture, although the aggregates enlarge. This is probably due to cell proliferation since the growth rate of cells cultured on EHS matrix is similar to that of cells on plastic, exhibiting a population doubling time of approximately 24 h (data not shown). Cell Surface GalTase Enzymatic Activity Is Substratum Dependent The rapidity with which cells reorganized from single cells to spheres on reconstituted matrix, and the distinct stages of reorganization elucidated using videomicroscopy, provides an opportunity to determine what molecular mechanisms are involved in specific events in this sequence. I tested whether GalTase was involved by asking whether it was expressed on the cell surfaces of COMMA-D cells under a variety of conditions. To localize extracellular GalTase activity, COMMA-D cells were cultured on EHS matrix or plastic for 3 h, then incubated with radiolabeled UDP-galactose, its conventional sugar donor, which is usually absent extracellularly. Cultures were then processed for autoradiography. Grains were observed overlying the surface of many cells cultured on EHS matrix (Figs. 3A and 3B) but were absent from cells on plastic substrata (Fig. 3D). Furthermore, the addition of a-LA to the cells cultured on matrix prevented expression of galactosylation (Fig. 3C), which is consistent with its modifing GalTase substrate specificity away from GlcNAc residues. The appearance of silver grains over cells cultured on EHS matrix, but absence in the presence of a-LA or when cells were cultured on glass, indicates that GalTase expression at the cell surface is substratum dependent. However, this method does not indicate where or what the galactosylated ligand is. The presence of GalTase ligand, i.e., GlcNAc carbohydrate residues, in reconstituted EHS matrix was confirmed by galactosylating the matrix using soluble GalTase and radiolabeled UDPgalactose (2.4 X lo5 cpm/well vs 828 cpm/well in the absence of GalTase), as previously shown [9]. GalTase Substrate Modifier Inhibits Specific Stages of Reorganization To examine whether cell surface GalTase plays a functional role in mammary epithelial cell interactions during reorganization on matrix, a-LA was added to serum-free medium during cell culture. (U-LA forms a
228
M. H. BARCELLOS-HOFF
FIG. 1. Cells cultured on reconstituted matrix undergo dramatic reorganization during the first 24 h of culture. Four stages of reorganization are evident from phase contrast photomicrographs of COMMA-D mammary epithelial cells cultured on reconstituted matrix. At 2 h (A), cells have elongated and chained, 4 h (B), networks have formed; 7.5 h (C), networks have begun to contract; 23 h (D), the networks have condensed into spheres.
complex with GalTase that has decreased affinity for GlcNAc residues and inhibits its binding to extracellular matrices [9,19]. Addition of &-LA to the medium at the time of plating dramatically altered the ability of the cells to undergo the sequence of events described above. Videomicroscopy revealed that assumption of the elongated cell shape was inhibited and that migration was affected. Rather than elongating and moving along one axis, the cells remained rounded, yet adherent (Figs. 4A-4D). Cells that were close together were able to make short chains, but were not able to recruit more distant cells. However, further observation of such cultures revealed that, although the contraction and compaction phases of morphogenesis was attenuated due to the
short chain formation, the later stages of reorganization were able to proceed in areas of high cell density (Figs. 4E-4H). Thus, although a-LA blocked cell elongation and network formation, if the cells happened to be very close to each other, and thus need not elongate to make contact, then they were able to aggregate, albeit in a curtailed fashion. Additionally, if a-LA was added at 2 or 4 h postplating, during the chaining and networking phases, then the cultures were inhibited at that point in the reorganization, but addition of (U-LA after 6 h, i.e., when networks were established, had no apparent effect on reorganization. Matrices galactosylated prior to cell culture were examined in order to test whether altering the GalTase recognition site (i.e., GlcNAc residues) would have the
GALTASE
MEDIATES
EPITHELIAL
229
REORGANIZATION
FIG. 2. Time-lapse videomicroscopy reveals the sequence by which mammary epithelial cells reorganize into spheres during the first 24 h of culture on reconstituted matrix. Photomicrographs of video images are shown of COMMA-D reorganization on reconstituted matrix. Cells were plated at 10:00 AM. (A) 3 h, (B) 7 h, (Cl 9 h, (D) 11.5 h, (El 13 h, and (F) 23 h postplating.
same effect as inhibiting ceil surface GalTase [lo]. However, this experiment was not informative since the control matrices, which received the extensive incubations and washing without any GalTase present, failed to elicit the typical reorganization (not shown).
The Effects of Inhibitors of GalTase Activity Concentration Dependent
Are
Cell attachment and subsequent elongation were monitored to quantitate the concentration response ef-
230
M. H. BARCELLOS-HOFF
FIG. 3. Cell surface GalTase enzymatic activity is detectable only on cells cultured on reconstituted matrix. Nomarski optics photomicrographs of cells cultured on reconstituted matrix (A and B), or on reconstituted matrix in the presence of a-LA, 1 mg/ml (C), or on glass (D) and labeled with UDP-[3H]galactose. Silver grains were deposited heavily over many cells cultured on reconstituted matrix (A, B). No grains are deposited over cells on glass (D) or rounded cells on reconstituted matrix cultured in the presence of o-LA (C).
fects of GalTase inhibitors on cell functions. a-LA or polyclonal GalTase antibodies added at the time of plating inhibited cell elongation in a concentration-dependent manner (Fig. 5). Cell elongation was greatly curtailed by various concentrations of a-LA or antibodies to GalTase when used at concentrations shown by others to inhibit B16-FlO cell spreading [lo]. Cells remained rounded, although firmly attached, in the presence of a-LA. The number of cells attached to the substratum was the same in the controls as when incubated with a-LA for 2 or 24 h. Addition of up to 10 mg/ml of (Y-LA at any time to cells cultured on plastic had no effect on cell spreading or monolayer morphology. Although the effects were less pronounced, antibodies to GalTase inhibited events subsequent to elongation in a
manner similar to that described for a-LA. Thus altering GalTase substrate specificity by a-LA or masking cell surface GalTase by antibody binding had qualitative and quantitatively similar effects. DISCUSSION
The mechanisms by which epithelia organize during development and reorganize at specific times postnatally are a fundamental puzzle in biology. Specific interactions between cells or between cells and their ECM occur both sequentially and concordantly during morphogenesis to bring the appropriate cells into contact to form tissue-specific structures. Cell culture systems have been employed in an attempt to dissect some of
GALTASE
MEDIATES
EPITHELIAL
REORGANIZATION
231
FIG. 4. Perturbation of GalTase specifically inhibits the first two stages of reorganization, which impedes subsequent events. Photomicrographs of video images of cells plated on reconstituted matrix at lo:30 AM in the presence of 1 mg/ml (U-LA. One field of observation at (A) 2 h, (B) 3 h, (C) 5 h, or (D) 6 h postplating. Note that cells remain rounded for the most part. A second field in the same dish demonstrating that some networks form if cells are adjacent to one another and that such networks can contract in the presence of a-LA: (E) 7 h, (F) 9 h, (G) 13 h, (H) 17 h postplating.
232
M. H. BARCELLOS-HOFF
0.2 -
1 ,.,: 0
1
Concentration,
2
3
mglml
FIG. 5. Modifiers of GalTase quantitatively inhibit cell elongation. Concentration-dependent effects of (Y-LA (squares; control was bovine serum albumin; n = 97) and anti-GalTase IgG (circles; control was normal rabbit serum; n = 194) are expressed as the percentage of control of nonround cells under the experimental condition. The symbols represent the mean of two determinations.
these events by analyzing adherence of cells to each other or to ECM proteins. The observation that culture on EHS-reconstituted matrix elicits histiotypic reorganization by primary cells from adult epithelia [2,3, 221, and even cell lines as reported here and elsewhere [23], is intriguing. This process of reorganization provides a more complex, but potentially more informative, model of cell-ECM interactions than cells cultured as monolayers. EHS-reconstituted matrix is composed of laminin (~SO%), collagen IV, heparan sulfate proteoglycan, and entactin [20]. Laminin has been widely implicated in cell adhesion and migration [17, 241, but it has been noted in several studies that individual matrix components do not effectively substitute for the complex matrix derived from the EHS tumor in terms of cell reorganization [2, 3, 221. It has been suggested that both protein concentration [3,5] and matrix flexibility [4,5] are important factors for eliciting morphogenic reorganization. Using videomicroscopy it is possible to distinguish four stages of reorganization following COMMA-D mammary epithelial cells attachment to reconstituted matrix: (i) cell elongation and chaining; (ii) network formation; (iii) network contraction; and (iv) compaction of aggregated networks. None of these events occur when COMMA-D cells are cultured on plastic, on which cells rapidly spread and form monolayers, or on colla-
gen gels [26] or PHFR-9 cell-secreted ECM [27]. Similar sequences, although different time frames, have been described for a cell line from small intestinal epithelium [23] and for fetal type 2 pneumocytes [25]. Each stage involves several distinct events. The first stage requires cell shape changes, single cell migration, and establishment of cell-cell contacts. These processes are also important in the second phase, network formation, which is characterized by chained cells moving together and establishing a network of cell-cell contacts. A similar course of events has been described for formation of capillary-like tubules by endothelial cells cultured on reconstituted matrix, which elicits rapid (
