Determination of apical membrane polarity in mammary epithelial cell cultures: The role of cell-cell, cell-substratum, and membrane-cytoskeleton interactions

EXPERIMENTALCELL

RESEARCH

188.302-311

(1990)

Determination of Apical Membrane Polarity in Mammary Epithelial Cell Cultures: The Role of Cell-Cell, Cell-Substratum, and Membrane-Cytoskeleton Interactions GORDON PARRY,*,~ JAMES C. BECK,* LENNY MOSS,* JACK BARTLEY,*

AND GEORGE K. OJAKIAN~

*Cell and Molecular Biology Division, Lawrence Berkeley Laboratory, University of California, Berkeley, California 94720 and TDepartment Anatomy and Cell Biology, State University of New York, Health Science Center, Brooklyn, New York 11203

that borders the epithelial lumen and the basolateral domain facing the serosal compartment. These plasma membrane domains are both structurally and biochemically different and are specialized to carry out the specific polarized functions characteristic of epithelia [l-5]. The existence of these domains leads to the consideration of how proteins are confined to one domain or the other. Three factors have been considered to be important: First, tight junctions form the boundary between apical and basolateral membranes and several studies have suggested that these act as barriers to the passive lateral movement of glycoproteins in the plane of the membrane [6-111. These studies have been carried out using both intact epithelial sheets, such as those of toad bladder [7], as well as cultured monolayers of the Madin-Darby canine kidney (MDCK) epithelial cell line [9, lo]. Second, epithelial cell polarity is established relative to the attachment substratum and a number of studies have considered cell-substratum interactions to be important in the generation and maintenance of membrane domains [12-151. In this respect, it has been found that MDCK cells exhibit polarized budding of lipid envelope viruses in low density cultures that have formed cell-substratum contacts but not cell-cell junctions [12]. In addition, biophysical studies of antigen mobility on cultured endothelial cells have demonstrated that the restriction of antigens to the apical domain may be dependent upon the composition of the extracellular matrix beneath the cells [14, 151. Finally, very recent studies of glycoproteins located exclusively in the apical or basolateral membrane domains have indicated that submembraneous cytoskeletal elements appear to be tightly associated with certain polarized membrane proteins of both MDCK cells [16-201 and mammary epithelial cells [21]. From these studies, it has been hypothesized that cytoskeletal-membrane protein interactions may restrict these antigens to the correct membrane domain. PAS-O is a major differentiation antigen that is expressed on these membranes and also on the surface of a number of human carcinoma cell lines [22-251. In this report, we present studies of the apical polarization of

The membrane glycoprotein, PAS-O, is a major differentiation antigen on mammary epithelial cells and is located exclusively in the apical domain of the plasma membrane. We have used 734B cultured human mammary carcinoma cells as a model system to study the role of tight junctions, cell-substratum contacts, and submembraneous cytoskeletal elements in restricting PAS-O to the apical membrane. Immunofluorescence and immunoelectronmicroscopy experiments demonstrated that while tight junctions demarcate PAS-O distribution in confluent cultures, apical polarity could be established at low culture densities when cells could not form tight junctions with neighboring cells. In such cultures the boundary between apical and basal domains was observed at the point of cell contact with the substratum. Immunocytochemical analysis of these cell-substratum contacts revealed the absence of a characteristic basement membrane containing laminin, collagen (IV), and heparan sulfate proteoglycan. However, serum-derived vitronectin was associated with the basal cell surface and the cells were shown to express the vitronectin receptor on their basolateral membranes. Additionally, treatment of cultures with antibodies against the vitronectin receptor caused cell detachment. We suggest, then, that interactions between vitronectin and its receptor, are responsible for establishment of membrane domains in the absence of tight junctions. The role of cytoskeletal elements in restricting PAS-O distribution was examined by treating cultures with cytochalasin D, colchicine, or acrylamide. Cytochalasin D led to a redistribution of PAS-O while colchicine and acrylamide did not. We hypothesize that PAS-O is restricted to the apical membrane by interactions with a microfilament network and that the cytoskeletal organization is dependent upon cell-cell and cell-substratum interactions. Q leso Academic PWSS, IW.

INTRODUCTION In simple epithelia, plasma membrane proteins are segregated into two distinct domains: the apical domain 1To

whom

reprint requestsshould be addressed.

0014-4827/90 $3.00 Copyright 0 1990 by Academic Press, All rights of reproduction in any form

302 Inc. reserved.

of

APICAL

MEMBRANE

POLARITY

PAS-O in cultured human mammary carcinoma cells and assess the relative roles of cell-cell, cell-substratum, and membrane-cytoskeleton interactions in determining PAS-O cell surface distribution. Our results provide evidence that interactions between PAS-O molecules and the submembraneous actin microfilament network are important in restricting the antigen to the apical plasma membrane and that the organization and stability of such a network may be dependent upon cell-cell and cell-substratum interactions. MATERIALS

AND

METHODS

Cell culture. 734B cultures were obtained from Dr. Adeline Hackett (Peralta Cancer Research Institute Oakland, CA) and cultured in medium 199 supplemented with 5% fetal calf serum and insulin (10 rm/ml). Penicillin and streptomycin were added routinely to the medium. Cultures were transferred using trypsin-EDTA and used for experiments 48 h or longer, after trypsinization. Antibodies. Antibodies reactive with PAS-O were generated in rabbits (polyclonal) and mice (monoclonal), using purified human milk fat globule membranes [22,24]. The specificity of the antibodies was determined by Western blotting procedures [24]. The monoclonal antibodies used in these experiments, LBL-1 and LBL-6, both recognized oligosaccharide determinants on this heavily glycosylated mutin-like glycoprotein [24]. Zmmuno.!ocalization procedures. Cells on coverslips and lactating mammary gland were fixed in phosphate-buffered saline (PBS) containing formaldehyde (3.5%) and glutaraldehyde (0.05%). The fixed monolayers were then incubated (15 min) in PBS containing 50 mM glycine or ammonium chloride, then washed twice (15 min each wash) with PBS containing 0.1% bovine serum albumin (PBS/BSA). Monolayers were then incubated in primary antibody (undiluted hybridoma culture supernatant) at room temperature for 1 h, washed 3 X 15 min in PBS/BSA and incubated for 1 h in conjugated goat anti-mouse or goat anti-rabbit immunoglobulins (Tago, H + L chain specific). For immunofluorescence microscopy, the second antibodies were either fluoresceinor rhodamine-conjugated while peroxidase-conjugated antibodies or Protein A coupled to 10 nm colloidal gold (PA/gold; Sigma Chemical Co.) were used for immunoelectron microscopy. Fluorescent samples were viewed under a Zeiss inverted photomicroscope equipped with a (X63) objective. Photographs were taken using Kodak Tri-X film rated at 800 ASA. The peroxidase-conjugated antibodZmmunoelectron microscopy. ies were visualized by use of diaminobenzidine (1 mg/ml) and H202 (0.01%). The monolayers were then fixed successively in 2% glutaraldehyde in 0.1 M phosphate buffer (pH 7.2) and 1% osmium tetroxide in 0.1 M sodium cacodylate buffer (pH 7.2) for 1 h at room temperature. After dehydration through ethanol, monolayers were embedded in Epon 812, sectioned perpendicular to the growth plane of the monolayer, and photographed in a JEOL 1OOC electron microscope. For immunofluorescence localization of extracellular matrix components and receptors, the fixed cells were permeabilized with acetone for 1 min at -2O’C prior to antibody staining. The antibodies against vitronectin and the vitronectin receptor were purchased from Telios Pharmaceuticals (San Diego, CA) and do not cross-react with either fibronectin or the fibronectin receptor (Telios Pharmacentical’s data sheet, and personal communication). Antibodies against other matrix components have been described by us previously [26].

RESULTS

The antigen recognized by the monoclonal antibodies LBL-1 and LBL-6, was identified in 734B cells by West-

IN MAMMARY

EPITHELIAL

CELLS

303

ern blotting [24]. This high-molecular-weight mucinlike glycoprotein (designated PAS-O) migrated as a broad band (M, > 406 kDa). PAS-O cell surface distribution was observed on cryostat sections of lactating breast tissue by immunofluorescence microscopy and demonstrated to have an exclusive apical distribution (Fig. 1). This result confirms that previously reported by Mather and colleagues using antibodies to guinea pig PAS-O [23,25]. Polarity of PAS-O in Cultured CelLs: Density Dependence In order to assess the relative importance of cell-cell and cell-substratum interactions in establishing membrane polarity, 734B cultures were plated at different densities and PAS-O distribution determined by immunofluorescence microscopy. Since these experiments were performed at least 48 h after plating, cell-substratum contacts were established for all cells, regardless of plating density. In contrast, the extent to which cell-cell contacts were established was a function of cell density. Islands of three or four cells had cell contacts only on a portion of their lateral surfaces while the remainder of the lateral membrane remained noncontacted (Fig. 2). However, the majority of cells in confluent sheets and large islands were associated with adjacent cells on all lateral surfaces and had extensive cell-cell contacts. In the small islands, cell surface staining of PAS-O was observed as a punctate pattern that probably represented apical microvilli (Fig. 2C), but no staining of lateral regions of the plasma membrane was apparent (Fig. 2C). This finding indicated that even at low cell densities apical membrane polarity was probably established. At higher cell densities (Figs. 2A and 2B), a similar staining pattern was seen and only apical microvillar staining was discernable. Control experiments were carried out to ensure that access of antibodies to the basolateral membranes of 734B cultures at low and intermediate densities was not a problem for monitoring polarity. Monoclonal antibody El which recognizes a 23-kDa glycoprotein (gp23) on the basolateral membrane of MDCK cells [27, 281 was also shown to stain the basolateral membrane of both low and intermediate density cultures (Fig. 3), but not of high density cultures. Treatment of high density cultures with the detergent NP40 to permeabilize cells prior to antibody incubation did not change the PAS-O staining pattern but did allow demonstration of gp23 staining on the basolateral membrane (data not shown). It should be noted that the heterogeneity with respect to cell staining with anti PAS-O antibodies has been well documented (Ref. [24] and references therein) and reflects cell to cell heterogeneity with respect to PAS-O glycosylation. It does not influence the interpretation of the data presented here.

304

PARRY

ET AL.

FIG. 1. Localization of PAS-O in the lactating mammary gland. Corresponding immunofluorescence (A) and phase (B) micrographs of 5brn frozen sections from lactating mouse mammary gland stained with monoclonal antibody LBLB. The antibody binds exclusively to the apical membrane. L, lumen. Bar, 8 pm.

Observations of PAS-O Distribution on the Edge of Islands of Cells A striking feature of intermediate-sized islands was that cells on the edge of the islands had a very distinctive PAS-O cell surface distribution. While all the cells in the interior regions of the islands exhibited a punctate apical staining pattern, many cells on the edge stained intensely on both the apical surface and the lateral surface that was not bordered by adjacent cells (Fig. 4). PAS-O distribution on these “edge cells” was examined at higher resolution using immunoperoxidase electron microscopy. It was found that PAS-O was excluded from the basal membrane (Fig. 5) and from the lateral membranes in regions of cell-cell contact (not shown). Examination of a number of thin sections revealed that the boundary between the stained and unstained membrane domains was always in close proximity to the point of cell-substratum contact. These ultrastructural results are consistent with the immunofluorescence observations and imply that interactions at the points of contact of the cell and the substratum can influence PAS-O distribution on the plasma membrane.

Synthesis of Basement Membrane Components We considered the possibility that the development of apical polarity was correlated with the deposition of a basement membrane synthesized by the 734B cells. However, immunofluorescence and Western blot analysis using antibodies to type IV collagen, laminin, and both heparan and chondroitin sulfate proteoglycans failed to detect the presence of any of these matrix components (data not shown), implying that none of these was required for maintenance of polarity. Electron mi-

croscopy studies, however, had clearly demonstrated the presence of cell contacts with the substratum. We therefore investigated whether serum-derived attachment proteins, including vitronectin and fibronectin, might be mediating cell attachment and be promoting membrane polarity at low cell density. Affinity-purified antibodies against vitronectin, fibronectin, and the vitronectin receptor were used to determine if these components were involved in cell-substratum adhesion. By focusing at the plane of the coverglass, immunofluorescence demonstrated that vitronectin was localized under each cell (Fig. 6B) and that the vitronectin receptor could be detected on both basal and lateral membranes (Fig. 6D). To test whether the interactions between vitronectin and its integrin receptor were indeed responsible for 734B cell attachment, cultures were plated in the presence of vitronectin receptor antibody. The antibody blocked cell attachment and, additionally, was able to promote detachment of cells maintained in culture for 24 h prior to antibody addition (Fig. 7B). Similar results were obtained with the peptide RGDS, which blocks vitronectin interactions with its receptor (data not shown). Other peptides, SDGR and YIGSR, did not influence cell attachment [29]. These results, then, are consistent with the idea that vitronectin is the principal matrix component at the cell-substratum contacts of 734B cells. It is possible that fibronectin may also contribute to this contact, in that immunofluorescence analysis did detect some fibronectin beneath the basal cell surface (data not shown), and, indeed, the peptide RGDS can disrupt both vitronectin and fibronectin interactions with their receptors. However, the vitronectin receptor antibody used does not react with the fibronec-

APICAL

MEMBRANE

POLARITY

IN MAMMARY

EPITHELIAL

CELLS

305

FIG. 2. Cell surface distribution of PAS-O on cultured 734B cells. Immunofluorescence (A, B, C) and corresponding phase (A’, B’, C’) micrographs of monolayers grown to different densities on coverglasses. In confluent cultures (A, A’), PAS-O was found exclusively on the apical surface. In subconfluent cultures (B, B’), where the cells are clustered into large islands, the antigen was also polarized. In cultures at very low density, staining was usually not detectable in lateral membranes where adjacent cells contacted each other (C, c’; see arrowheads) and the antigen appeared restricted to the apical membrane. Bar, 12 pm.

tin receptor, thus demonstrating nectin in cell contact.

a specific role for vitro-

The Role of Cytoskeletal Elements in Stabilizing Membrane Domains The involvement of microfilaments, microtubules, and keratin filaments in stabilizing the apical domain was investigated by including the anti-cytoskeletal drugs cytochalasin D (CD), colchicine, or acrylamide in the culture medium for 2 h prior to fixation. CD treatment (1 pg/ml) produced a reorganization of the apical cell surface with a notable loss of microvilli and a redistribution of PAS-O into large aggregates on long membrane extensions (Fig. 8B). Although colchicine caused cell shape changes, redistribution of PAS-O was not observed. Acrylamide, a known disruptor of keratin fila-

ments [30] did not have an effect on either cell morphology or PAS-O distribution. The aggregation of PAS-O into patches could be accentuated by including a polyclonal antibody to PAS-O in the presence of cytochalasin (data not shown). Strikingly the antibody did not have any effect in the absence of cytochalasin. This set of data then is consistent with there being either direct or indirect interactions between PAS-O and the actin containing cytoskeleton. Further data consistent with PAS-O association with cytoskeletal components were obtained by preparing detergent-insoluble cytoskeleton fractions from 734B cells using Triton X-100 and EDTA. Western blot analysis of PAS-O presence in the cytoskeleton revealed that approximately 30% of the cellular PAS-O was associated with a cytoskeleton fraction that was collected by lowspeed centrifugation (2000 rpm X 5 min).

306

PARRY

ET AL.

FIG. 3. Subconfluent 734B cultures stained with monoclonal antibody El, which reacts with a 23-kDa basolateral fluorescence staining of the basolateral membranes (arrowheads) clearly demonstrates that penetration of antibodies was not impeded at this cell density. A, B are both representative fields of monolayers stained with El. Bar, 12 pm.

DISCUSSION In this study, several factors that could influence apical membrane polarity in cultured human mammary epithelial cells were investigated. We have considered the importance of tight junctions, cell-extracellular matrix contacts, and membrane-cytoskeletal interactions in determining the cell surface distribution of the high-molecular-weight mucin-like glycoprotein PAS-O. Our principle findings were: (a) apical glycoprotein polarity is established on the majority of 734B cells independent of cell density and includes those cells that do not have cell-cell junctions on all aspects of their lateral membranes (“edge cells”); (b) polarity in “edge cells” may require a tight-junction/zonula adherens complex on one aspect of the cell and cell-substratum contacts at the free aspect of the cell surface; (d) microfilaments appear to play an important role in restricting the PAS-O glycoprotein to the apical surface. These observations support a model in which a submembraneous, actin-containing cytoskeletal network interacts with an apical membrane glycoprotein. This network appears to interact with the zonula adherens/ tight junction complex in confluent monolayers of cells to restrict the lateral movement of proteins in the plane of the membrane. In subconfluent monolayers which lack extensive cell-cell contacts, we propose that cellsubstratum interactions may stabilize the cytoskeleton. Our model places considerable emphasis on adhesiontype junctions in actively establishing an apical macromolecular cytoskeletal network that serves as a restricting force upon membrane glycoproteins.

glycoprotein. Immunobeneath the monolayer

While our studies using cytochalasin treatment imply that PAS-O is restricted to the apical domain by microfilaments, the molecular nature of interactions between PAS-O in the membrane and the cytoskeleton remain to be determined. However, the findings of the Carraways [31-351 on actin-membrane interactions in microvilli of rat mammary ascites tumor cells may prove to be significant. These studies have identified a cell surface protein (CAG) that binds directly to both actin on the cytoplasmic face of the plasma membrane and to membrane glycoproteins on the cell surface. In these rat ascites cells, a 120kDa glycoprotein ASGP-2 binds directly to both CAG and to ASGP-1, a mucin-like glycoprotein which has many similarities to PAS-O. Even though these cells are not polarized and the quantitites of ASGP-1 and ASGP-2 appear to be in excess of the available microfilament binding sites, the information obtained from these studies may prove to be important for understanding microfilament control of apical polarity in polarized epithelial cell cultures. Specific data regarding the involvement of cytoskeletal elements in cell polarity has been obtained in other experimental systems. Recently, Nelson and co-workers have obtained solid evidence that a submembraneous cytoskeleton containing ankyrin and fodrin is responsible for restricting (Na+, Kf) ATPase to the basolateral domain of MDCK cells [16-U]. In cultured mouse mammary epithelial cells, Rapraeger et al. [21] have found that a transmembrane proteoglycan becomes associated with the actin-rich cytoskeleton and sequestered to the basolateral membrane during polarity development. Furthermore, there is also good evidence that 135 and

APICAL

MEMBRANE

POLARITY

IN MAMMARY

EPITHELIAL

CELLS

307

this is a continuous process requiring several days in culture [9,13]. Our results on human mammary epithelial cells are consistent with those obtained for MDCK cells in which the differential modulation of apical and basolateral protein distribution has been found [9,13,19,20]. The finding that PAS-O is associated with actin microfilaments appears to be inconsistent with the observations that cytochalasin D failed to affect the polarity of viral budding or delivery of influenza virus hemagglutinin to the MDCK apical surface [36]. It is conceivable that viral proteins do not behave in the same manner as native epithelial proteins. However, it is also possible that the subcellular and molecular mechanisms generating polarity (such as those operating during viral bud-

FIG. 4. PAS-O distribution on an island of 734B cells (A) observed in two focal planes. B, an immunofluorescence image focused on the basal region demonstrates intense staining of the lateral surface at the edge of the island and absence of staining on lateral membranes of contacted cells. C, the same island of cells observed in another focal plane demonstrating PAS-O on the apical surface. Bar, 10 pm.

184-kDa apical glycoproteins [13,19,20] are associated with the apical cytoskeleton of MDCK cells [19, 201. These results, together with those reported here, are consistent with the idea that polarized membrane glycoproteins of epithelia can be restricted to domains of the plasma membrane by specific interaction with cytoskeleton components. Our data also imply that the mechanisms maintaining apical polarity in culture are distinct from those maintaining basolateral membrane polarity. Previous work has demonstrated that basolateral membrane of MDCK cells become polarized _ -glycoproteins .-. only

after

tight

junctions

have assembled

[9] and that

FIG. 5. Ultrastructural localization and cell surface distribution of the PAS-O glycoprotein. 734B cells grown into intermediate-sized islands were fixed and PAS-O was localized by immunoelectron microscopy using horseradish peroxidase (HRP)-labeled antibodies. At the edge of these islands (A, B), the electron-dense HRP maction product localizing PAS-O was observed only on the apical cell surface (Ap) and not beyond a clearly defined transition region (large arrowheads). PAS-O was not detected on the basolateral membrane (small arrowheads). In some cells (B), tight cell-substratum interactions or focal contacts (fc; the boundaries are indicated by small arrowheads) containing numerous microfilaments (mf) were observed. Bar, 0.75 pm (A) or 0.50 pLm(B).

FIG. 6. Immunofluorescence localization of vitronectin and the vitronectin receptor was performed on subconfluent cells either 48 h El or 18 h after plating (F-J). B and G show vitronectin distribution; D and I show vitronectin receptor distribution; E and J demonstrate absence of staining in controls. In 18-h cultures, vitronectin was distributed in patches over the cell surface, but was diffusely distributed at basal surface in 48-h cultures. The vitronectin receptor was localized fairly evenly over the surface on the rounded 18-h cultures and on basal and lateral membranes of the 48-h cultures. Bar, 12 pm. 308

(Athe the the

APICAL

MEMBRANE

POLARITY

IN MAMMARY

EPITHELIAL

CELLS

309

FIG. 7. The role of the vitronectin receptor in cell spreading and attachment. After trypsinization, the cells were allowed to attach to coverglasses in medium either lacking (A) or containing (B) antibodies against the vitronectin receptor (1:lOO dilution) for 24 h. Control nonimmune serum did not cause detachment. The effect of the vitronectin receptor was reversible in that cells reattached after removal of the antiserum. Bar, 8 pm.

FIG. 8. Immunogold localization of PAS-O on CD-treated (1 fig/ml) cells. PAS-O localized by PA/gold appeared to have a uniform distribution on the apical surface of untreated cells (A) and was found on both microvilli (arrowheads) and intermicrovillar spaces. PA/gold staining was not observed on the basolateral membrane (B). After treatment with CD (C), PAS-O was localized primarily in aggregates (arrowheads) on large apical cell surface extensions. Bar, 0.22 pm.

310

PARRY

ding) are distinct from those involved in maintaining and stabilizing distinct plasma membrane domains. The full extent and significance of membrane cytoskeleton interactions in determining cell polarity will clearly depend on an analysis of the behavior of several different glycoproteins in a variety of cell types. 734B cells plated on glass coverslips were able to establish a polarized distribution of PAS-O and immunofluorescence microscopy demonstrated that vitronectin, fibronectin, and the vitronectin receptor were found associated with the basal surface. These results suggest the interesting possibility that integrin interactions with extracellular matrix components may be responsible for organizing a polarized cytoskeleton that could establish apical cell surface polarity in the absence of tight junctions. This interpretation is supported by previous observations demonstrating that the distribution of an apical membrane protein in endothelial cells was modulated by substratum-associated extracellular matrix components [ 14, Xi] and that the A chain of laminin is critical for the establishment of cell polarity in developing kidney [37]. The fact that polarity could be established in the absence of any of the major basal lamina components, laminin, heparan sulfate proteoglycan, and IV collagen, was surprising. These components are, of course, the principal constituents of the basal lamina in Go. In contrast, vitronectin is not usually a basal lamina component in uivo, but clearly functions in cell attachment in culture. This anomalous behavior is most likely a consequence of the tumorigenic state of the cells. While the matrix components used by the 734B cells in culture are distinct from those used by normal epithelia, the simplicity of cell-matrix interactions in 734B cell cultures is attractive from an experimental viewpoint. The ability to specifically modify cell-substratum interactions using antisera against the vitronectin receptor offers an excellent opportunity to study the specific mechanisms by which interactions of matrix molecules with their integrin receptors lead to membrane polarization. Finally, while the data presented here relate specifically to the mechanisms involved in membrane domain formation, our conclusions may be relevant to the broader question of how cell-cell and cell-substratum interactions regulate mammary epithelial cell functions [38-401. In this respect it is possible that the cytoskeleton network may mediate the effects of cell-cell and cellsubstratum on tissue specific gene expression and other aspects of cellular differentiation.

We thank Lucinda Olney and Dorothy Sprague for typing the manuscript and Alex Fulop and Agnes Ortega for assistance with some of the experiments. This research was supported by National Institutes of Health Grants CA44398 (G. P.) andDK-30537 (G. K. O.), and by the Office of Health and Environmental Research, Office of Energy Research, Department of Energy under Contract DEAC03-76SF

ET AL.

REFERENCES 1.

Rodriguez-Boulan, E., and Nelson, W. J. (1989) Morphogenesis of the polarized epithelial cell phenotype. Science 245, 718-725.

2.

Simons,K., and Fuller, S. D. (1985) Cell-Surface

polarity

in epi-

thelia. Annu. Rev. Cell Biol. 1,243-288. 3.

Misfeldt, D. S., Hammamoto, S. T., and Pitelka, D. R. (1976) Transepithelial transport in cell culture. Proc. Natl. Acad. Sci. USA 73,1212-1215.

4.

Cereijdo, M., Ehrenfeld, J., Meza, I., and Martinez-Palomo, A. (1980) Structural and functional membrane polarity in cultured monolayers of MDCK cells. J. Membr. Biol. 62,147-159.

5.

Bisbee, C. A., Machen, T. E., and Bern, H. A. (1979) Mouse mammary epithelial cells on floating collagen gels: Transepithelial ion transport and effects of prolactin. Proc. Natl. Acad. Sci. USA 76, 536-540.

6.

Galli, P., Brenna, A., DeCamilli, P., and Meldolesi, J. (1976) Extracellular calcium and the organization of tight junctions in pancreatic acinar cells. Enp. Cell Res. 99,178-183.

7.

Pisam, M., and Ripoche, P. (1976) Redistribution of surface macromolecules in dissociated epithelial cells. J. Cell Biol. 71,907-

8.

Hoi-Sang, V., Saier, M. M., and Ellisman, M. (1979) Tight junction formation is closely linked to the polar redistribution of intramembranous particles in aggregating MDCK epithelia. Exp. CellRes. 122,384-392.

9.

Herzlinger, D. A., and Ojakian, G. K. (1984) Studies on the development and maintenance of epithelial cell surface polarity with monoclonal antibodies. J. Cell Biol. S&1777-1787.

10.

Gonalez-Mariscal, L., Chavez de Ramirez, B., and Cereijido, M. (1985) Tight junction formation in cultured epithelial cells (MDCK). J. Membr. Biol. 86.113-125.

11.

Van Merr, G., and Simons, K. (1986) The function of tight junctions in maintaining differences in lipid composition between the apical and the basolateral cell surface domains of MDCK cells. EMBO J. 5,1455-1464.

12.

Rodriguez-Boulan, A., Paskiet, K., and Sabatini, D. D. (1983) Assembly of enveloped viruses in Madin-Darby canine kidney cells: Polarized budding from single attached cells and from clusters of cells in suspension. J. Cell Biol. 96,866-874.

13.

Vega-Salas, D. E., Salas, P. J. I., Gundersen, D., and RodriguezBoulan, E. (1987) Formation of the apical pole of epithelia (Madin-Darby canine kidney) cells: Polarity of an apical protein is independent of tight junctions while segregation of a basolatera1 marker requires cell-cell interactions. J. Cell Biol. 104,905916.

14.

Nakache, M., Schreiber, A. B., Gaub, H., and McConnell, H. M. (1985) Heterogeneity of membrane phospholipid mobility in endothelial cells depends on cell substrate. Nature (London) 317, 75-77.

15.

Nakache, M., Gaub, H. E., Schreiber, A. B., and McConnell, H. M. (1986) Topological and modulated distribution of surface markers on endothelial cells. Proc. Natl. Acad. Sci. USA 83, 2874-2878.

16.

Nelson, W. J., and Veshnock, P. J. (1986) Dynamics of membrane-skeleton (fodrin) organization during development of polarity in Madin-Darby canine kidney epithelial cells. J. Cell Biol.

17.

Nelson, W. J., and Veshnock, P. J. (1987) Modulation of fodrin (membrane skeleton) stability by cell-cell contact in MadinDarby canine kidney epithelial cells. J. Cell Biol. 104, 15271537. Nelson, W. J., and Veshnock, P. J. (1987) Ankyrin binding to (Na+ + K+) ATPase and implications for the organization of

920.

103,1751-1766.

18

APICAL membrane

domains

in polarized

MEMBRANE cells. Nature

POLARITY (London)

328,

533-536. 19.

Salas, P. J. I., Vega-Salas, D. E., Hochman, J., Rodriguez-BouIan, E., and Edidin, M. (1988) Selective anchoring in the specific membrane domain: A role in epithelial cell polarity. J. Cell Biol.

107,2363-2376. 20. Ojakian, G. K., and Schwimmer,

R. (1988) The polarized distribution of an apical cell surface glycoprotein is maintained by interactions with the cytoskeleton of Madin-Darby canine kidney cells. J. CellBiol. 107,2377-2387. 21. Rapraeger, A., Jalkanen, M., and Bernfield, M. (1986) Cell surface proteoglycan associates with the cytoskeleton at the basolatera1 cell surface of mouse mammary epithelial cells. J. Cell Biol.

103,2683-2696. 22. Shimizu, M., and Yamauchi,

K. (1982) Isolation and characterization of mucin-like glycoprotein in human milk fat globule membrane. J. Biochem. (Tokyo) 91,515-524. 23. Johnson, V. G., and Mather, I. H. (1985) Monoclonal antibodies prepared against PAS-l, butyrophilin and GP-55 from guineapig milk fat globule membrane bind specifically to the apical pole of secretory epithelial cells in lactating mammary tissue. Exp. Cell Res. 156, 144-158. 24. Moss, L., Greenwalt, D., Cullen, B., Dinh, N., Ranken, R., and Parry, G. (1988) Cell to cell heterogeneity in the expression of carbohydrate based epitopes of mucin-type glycoprotein on the surface of human mammary epithelial cells. J. Cell. Physiol. 137,

310-320. 25. Mather, I. H. (1987) in The Mammary

Gland (Neville, M. C., and Daniel, C. W., Eds.), pp. 217-267. Plenum, New York. 26. Parry, G., Cullen, B., Kaetzel, C. S., Kramer, R., and Moss, L. (1987) Regulation of differentiation and polarized secretion in mammary epithelial cells in culture: Extracellular matrix and membrane polarity influences. J. Cell Biol. 106,2043-2051. Ojakian, G. K., Romain, R. E., and Herz, R. E. (1987) A distal nephron glycoprotein that has different cell surface distributions on MDCK cell sublines. Amer. J. Physiol. 253,433-443. 28. Herz, R. E., and Ojakian, G. K. (1989) Differential targeting of an apical cell surface glycoprotein in polarized epithelial cell lines. J. Biol. Chem. 264,4605-4612. Received September 15,1989 Revised version received January

9,199O

IN MAMMARY 29.

EPITHELIAL

311

CELLS

Ruoslathi, E., and Pierschbacher, M. D. (1987) New perspectives in cell adhesion: RGD and integrins. Science 238,491-497.

30. Eckert, B. S. (1985) Alteration tion in PtKl cells by acrylamide.

of intermediate filament distribuEur. J. Cell Bi01. 37,169.

31. Sherblom, A. P., and Carraway,

K. L. (1980) A complex of two cell surface glycoproteins from ascites mammary adenocarcinoma cells. J. Biol. Ckem. 255,12051-12059.

32. Helm, R. M., and Carraway,

K. L. (1981) Evidence for the association of two cell surface glycoproteins of 13762 mammary ascites tumor cells. Exp. Cell Res. 135,418-424.

33. Carraway,

C. C., Jung, G., Creaik, J. R., Rubin, R. W., and Carraway, K. L. (1983) Identification of a cytoskeleton-associated glycoprotein from isolated microvilli of a mammary ascites tumor. Exp. CellRes. 143,303-308.

34. Carraway,

C. C., Jung, G., Hunkley, R. E., and Carraway, K. L. (1985) Isolation of microvillar microfilaments and associated transmembrane complex from ascites tumor cell microvilli. Exp. CellRes. 157, 71-82.

35. Carraway,

K. L., and Spielman, J. (1986) Structural and functional aspects of tumor cell sialomucins. Mol. Cell. Biochem. 72, 109-120.

36. Salas, P. J. I., Misek, D. E., Vega-Salas, D., Gunderson, D., Cerejido, M., and Rodriguez-Boulan, E. (1986) Microtubules and actin filaments are not critically involved in the biogenesis of epithelial cell surface polarity. J. Cell Biol. 102,1853-1867.

37. Klein, G., Langegger, M., Timpl, R., and Ekblom, P. (1988) Role of laminin A chain in the development Cell 55,331-341.

of epithelial

cell polarity.

38. Bissell, M. J., Hall, H. G., and Parry, G. (1982) How does the extracellular 31-68.

matrix

direct gene expression?

J. Tkeor. Biol. 99,

39. Lee, E. Y-H., Parry, G., and Bissell, M. J. (1984) Modulation of secreted proteins of mouse mammary epithelial collagenous substrata. J. Cell Biol. 98, 146-155.

cells by the

40. Bissell, M. J., and Hall, H. G. (1987) in The Mammary (Neville, M. C., and Daniel, New York.

Gland C. W., Eds.), pp. 97-146. Plenum,