17-β-Estradiol induced alterations of cell-matrix and intercellular adhesions in a human mammary carcinoma cell line

Estrogen is closely associated with promotion and control of human breast cancer tumor growth. Evidence of estrogen- dependent breast cancer is found in studies demonstrating remission after ovariectomy (Beatson, 1896; Pike et al., 1981), remission after treatment with anti-estrogens (King et al., 1982; Clark and McGuire, 1989), and the nearly exclusive post- puberty onset of breast cancer in women (Henderson et al., 1988). Animal studies have also shown that the natural estrogen, 17-β-estradiol (E₂), increases the incidence of mammary tumors in mice and rats (IARC, 1979). In vitro E₂treatment of hormone-sensitive MCF-7 human breast cancer cells elicits a number of responses that are also characteristic of breast tumors in vivo (Dickson, 1992; Lippman et al., 1976). MCF-7 cells introduced into ovariectomized, athymic mice will only form tumors if supplemented with estrogen (Soule and McGrath, 1980; Shafie and Liotta, 1980), indicating that E₂promotes metastasis as well as primary tumor growth. The estrogen-dependency of breast tumor growth and metastasis are strongly indicated by these studies, yet the role of E₂in the loss or dysfunction of stable cellular adhesions, which is necessary throughout these processes (Nicholson, 1989; Blood and Zetter, 1990), is not well understood.

The MCF-7 breast carcinoma cell line, established from the pleural effusion of a patient with breast adenocarcinoma (Soule et al., 1973), frequently serves as a model for estrogen- dependent control of tumor cell function in vitro and in vivo. E₂treatment of MCF-7 cells induces expression of a number of genes closely associated with control of cell proliferation, including transforming growth factor-α (Dickson et al., 1986a), insulin-like growth factor-I (Huff et al., 1986) and -II (Yee et al., 1988), platelet-derived growth factor (Bronzert et al., 1987), fibroblast growth factor-like and epidermal growth factors (Dickson et al., 1986b), and the epidermal growth factor receptor (Berthois et al., 1989). The induction of these genes and subsequent hyperproliferation of cells is a potential basis for E₂-dependent breast tumor growth and metastasis (Preston-Martin et al., 1990). However, hyperproliferation of epithelial cells, while obviously critical to the tumorigenic and metastatic processes, alone cannot fully account for the invasive nature of tumor growth and metastasis. The develop- ment of carcinomas and metastatic spread requires tumor cells to: (1) destabilize or weaken adhesion to underlying basement membrane and adjacent cells; (2) traverse the basement membrane; (3) invade the interstitial tissue; (4) invade blood- vessel basement membrane; (5) enter the bloodstream to be transported to distant sites; (6) adhere to the luminal surface of the endothelial lining; (7) invade the endothelial lining by migration through intercellular junctions; and (8) attach to the subendothelial matrix, where the process may begin again (Nicholson, 1989; Blood and Zetter, 1990). The tumor cell must have the capacity to form and dissolve adhesions with each of the different surfaces encountered during this process. Functional adhesion receptors must also be expressed to support malignant cell movement, while factors governing the formation of stable, immobilizing, cellular adhesions are sup- pressed. The hormonal dependency of human breast cancer and of MCF-7 cell metastasis in mice suggests a fundamental role for E₂in governing tumor-cell adhesion and migration at some or all of these steps.

The integrity and organization of epithelial cell layers, in vivo and in vitro, is maintained by the presence of several types of intercellular adhesions including desmosomes, tight junctions and intermediate junctions. Intermediate junctions, or adherens junctions, consist of two types: cell-to-cell and cell- to-matrix adherens junctions (CCAJ and CMAJ, respectively). Actin filaments insert into an adhesion plaque located on the cytoplasmic face of the plasma membrane of both CMAJ and CCAJ. The adhesion plaques of both types of junction contain the cytoskeletal proteins vinculin (Geiger, 1979), α-actinin (Lazarides and Burridge, 1975), tenuin (Tsukita et al., 1989) and zyxin (Crawford and Beckerle, 1991). However, CCAJ and CMAJ differ in the further composition of the plaques and the classes of adhesion receptors present in the membrane. Talin and paxillin are present in the plaque of the CMAJ but not the CCAJ (Geiger et al., 1985; Turner et al., 1990). Plako- globin, a major component of the desmosomal plaque, and α- and β-catenin are also present in the plaque of the CCAJ, but not the CMAJ (Cowin et al., 1986; Herrenknecht et al., 1991; Peifer et al., 1992; Knudsen and Wheelock, 1992). The CCAJ contain cadherins, adhesion receptor proteins that belong to a family of cell-cell adhesion molecules (Geiger and Ayalon, 1992). Cadherins mediate cell-cell adhesion directly through homotypic protein interactions between adjacent cells (Takeichi, 1988). The extracellular matrix receptor proteins, integrins, are present in CMAJ (Hynes, 1992). Integrins and cadherins are both transmembrane proteins, and the cytoplas- mic domains provide a basis for a link between the cytoskele- ton of one cell and the adjacent cells or underlying matrix (Burridge et al., 1988; Geiger and Ginsburg, 1991; Tsukita et al., 1992).

E₂treatment in vitro induces a morphological differentiation in post-confluent cultures of MCF-7 cells. The in vitro differ- entiation is characterized by the development of solid, multi- cellular structures or nodules (Pourreau-Schneider et al., 1984; Gierthy et al., 1991), which can extend up to 80 μm above the surrounding monolayer. The marked transition from a typical flattened epithelial monolayer to disordered nodules potentially reflects overt changes in either the structure or the composition of CMAJ and CCAJ, which may precede or parallel nodule development. Therefore, the present study was undertaken to evaluate E₂-dependent changes in MCF-7 cell-matrix and cell- cell adhesions. We examined E₂-induced MCF-7 cells for changes in the distribution of talin and vinculin at CMAJ, and plakoglobin and vinculin at CCAJ. Talin- and vinculin-rich CMAJ adhesion plaques were markedly reduced as a result of E₂-induction. Vinculin staining was also reduced at CCAJ of monolayers and absent from nodule cell-cell borders. Destabi- lization of CMAJ and CCAJ resulting from the loss of adhesion plaque proteins could provide a basis for E₂-dependent tumor- cell detachment from the substratum and adjacent cells and increased tumor cell motility.

The strain of MCF-7 human metastatic mammary carcinoma cell line used in these experiments was obtained from the laboratory of Dr Alberto C. Baldi (Institute of Experimental Biology and Medicine, Buenos Aires, Argentina). Cells were maintained in Phenol Red-free Dulbecco’s modified Eagle’s medium (DMEM) supplemented with penicillin (100 units/ml), streptomycin (100 μg/ml), 10 ng/ml insulin (Collaborative Research, Inc.), 2 mM L-glutamine, 0.1 mM nonessen- tial amino acids (Gibco), and 5% bovine calf serum (BCS, Hyclone). For experiments cells were seeded in E₂-deprived medium, which consisted of the above components except that 5% charcoal-dextran- stripped BCS was substituted for normal BCS. Cells were seeded on glass coverslips at a density of 10⁵cells/35 mm well. Twenty-four hours after seeding, the medium was removed and replaced with E₂- deprived medium supplemented with either: (a) 10^-9M E₂(Sigma); (b) 0.01% DMSO (Sigma); or (c) no E₂or DMSO. Estrogen stocks were in 100% DMSO; final concentration of DMSO was always 0.01%. Cultures were refed with the appropriate medium every 3 days for a total 12 days of treatment. Cultures attained confluence at 6 days and nodules appeared after 7 days, being well developed and numerous at 12 days.

Cells were fixed using 3.7% paraformaldehyde in a PIPES-buffered salt solution, pH 7.2 (Small, 1981). This buffer was used in all pro- cessing steps. The initial fixation was at 37°C and all remaining steps carried out at room temperature. Fixation was followed by permeabi- lization with 0.5% Triton-X 100 for 20 minutes, treatment with 0.1 M glycine for 10 minutes, and a 0.1% Tween-20 wash. Cells were then incubated with primary antibody for 30 minutes at room tem- perature. Antibodies used in this study were mouse monoclonals to human vinculin (clone hVIN-1, Sigma) diluted 1:800, plakoglobin (clone PG5.1, Pierce; Cowin et al., 1986) at 5 μg/ml, and cytokeratins (clone 8.13, Sigma) diluted 1:40. A rabbit antiserum to human talin was diluted 1:100 (kindly provided by Dr K. Burridge, UNC, Chapel Hill). TRITC- and FITC-conjugated affinity-purified goat IgG directed against either mouse or rabbit IgG (Pierce) diluted 1:100 were used as secondary antibodies. Incubation times were the same as with primary antibody. All antibody dilutions were in buffer con- taining 0.1% Tween-20. Controls consisted of either mouse IgG or normal rabbit serum substituted for primary antibody, which resulted in little or no nonspecific background staining. Bodipy-phallacidin (Molecular Probes) was used at a concentration of 250 nM to label F- actin. Specimens were then mounted in buffer containing 1 mg/ml paraphenylenediamine (Sigma) and viewed with a Nikon microscope equipped for conventional epifluorescence. Fluorescence micrographs were taken on T-MAX 400 film (Kodak) with an exposure index of 3200. Micrographs of antibody distribution were printed relative to the low level of nonspecific staining obtained with control preparations. Exceptions to this are Fig. 3Fand H(see Results). Confocal laser scanning microscopy was carried out using a Bio-Rad MRC 600 mounted on an Olympus IMT-2 microscope equipped with a ×60 objective lens (NA 1.4; planapochromatic), argon ion laser, and single label rhodamine block. Emitted signal was digitized by Accumulative filter collection with each section scanned two (rhodamine-phalloidin) or four (rhodamine-conjugated antibodies) times. Z-axis sections were taken at 1 μm increments from substratum to nodule apex. Monitor images were recorded on Kodak Tech Pan film.

Cells cultured in plastic Petri dishes were fixed in situ for 1 hour in a phosphate buffer (pH 7.2) containing 4% formaldehyde and 1% glu- taraldehyde (McDowell and Trump, 1976). After washing with 0.1 M sodium cacodylate (pH 7.2), cells were post-fixed for 1 hour in 1% osmium tetroxide (prepared in 0.1 M sodium cacodylate), washed with 0.1 M sodium cacodylate, dehydrated with a graded ethanol series and embedded in an Epon/Araldite mixture. After polymeriza- tion at 60°C for 36-48 hours, the embedded cells were removed from the culture dishes by applying gentle pressure to the plastic, viewed by light microscopy, and areas of interest were cut from the embedded material with a scalpel. These selected pieces were oriented either parallel or perpendicular to the cutting surface of flat embedding molds, which were then filled with fresh Epon/Araldite and cured at 60°C for an additional 48 hours. Sections of the resulting material were retrieved on Formvar-coated slot grids, stained with uranyl and lead salts, and viewed in the electron microscope.

MCF-7 cells grown in estrogen-depleted medium are flat and bounded by prominent intercellular spaces (Fig. 1A,C,E). Sub-confluent cultures (Fig. 1A) display typical epithelial characteristics, growing as clusters that can assume a polarized motile morphology with a leading edge and trailing end (DiPasquale, 1975; Kolega et al., 1982). Confluent and post- confluent cultures remain flattened with an overall rough cob- blestone appearance and retain the large intercellular spaces (Fig. 1C,E). The intercellular spaces are not present in sub- confluent cultures treated with E₂but otherwise the cells retain the epithelial morphology characteristic of untreated cultures (Fig. 1B). Cells of estradiol-treated confluent (6 to 7 days) and postconfluent (>7 days) cultures are densely packed (Fig. 1D,F). The development of nodules is restricted to postcon- fluent cultures (Gierthy et al., 1991) and the nodules are not detected in earlier cultures (Fig. 1D). The development of multicellular nodules might depend upon the disassembly of cell-matrix and intercellular adhesions, thereby allowing cells to migrate and grow beyond the boundaries of a typical epithe- lial layer. Therefore we examined CMAJ and CCAJ in untreated and treated MCF-7 cells to determine any changes in cytoskeleton components of these adhesions that might be influenced by estradiol.

Cultures seeded at low density and allowed to attach for 24 hours were used to examine individual and small clusters of cells. Approximately 5% of the cells had assumed an apparent motile morphology with an overall canoe shape and small but well-developed lamellipodia, which contained F-actin-rich microspikes (not shown). This morphology is essentially the same as reported for several epithelial cell types (Bereiter- Hahn et al., 1981; Kolega et al., 1982; Mittal and Bereiter- Hahn, 1985). The remainder of the cells were spread over a greater individual area, being flattened and round in circum- ference, or as small clusters of these. Individual MCF-7 cells have prominent vinculin- (Fig. 2) and talin-rich adhesion plaques, which are associated with the termini of F-actin-rich stress fibers. Interference reflection microscopy (IRM) showed that the adhesion plaques corresponded to focal contacts (not shown). Cells at the periphery of clusters had an appearance similar to that of individual cells. No subpopulation of indi- vidual MCF-7 cells or clusters was found that lacked adhesion plaques or focal contacts. Arrangement of adhesion plaques was typical of epithelial cells: located mostly at the periphery of apparently static cells (Fig. 2A,B) or concentrated just behind lamellipodia of apparently migrating individual cells (not shown). However, clear-cut boundaries at the ventral surface of adjacent cells within a cluster did not exist. Lamel- lipodia, and sometimes larger portions of cells, were found to underlap neighboring cells, as determined by staining for F- actin. Adhesion plaques were thus present beneath the entire cluster. Treatment with E₂for 24 hours had no overt effect on the arrangement of these structures.

In order to assess changes in cellular adhesions that longer-term E₂treatment could effect while leading to nodule devel- opment, cultures were examined at 3, 6, 9 and 12 days. Cell cultures are confluent at 6 days and the resulting F-actin and plaque protein fluorescence images are difficult to compare precisely. The F-actin staining pattern with respect to the ventral surface of induced cells is quite complex. E₂treatment results in the appearance of numerous slender, linear F-actin- rich structures that apparently originate from basolateral and ventral surfaces (not shown). Structures similar to, but dis- tinctly different from, lamellipodia arise from and cover most of the ventral region of induced cells (not shown). Further- more, underlapping lamellipodia from adjacent cells are also apparent. Characterization of E₂-mediated alteration of actin filaments is currently being undertaken. Therefore changes in adhesion plaque distribution were assessed by labelling with either anti-vinculin or anti-talin and selecting random fields for evaluation. For consistency in comparison with later cultures, micrographs presented here from the subconfluent 3-day time point were taken so that a multicellular cluster occupied the whole field of view. Vinculin or talin adhesion plaques are situated throughout the ventral surface of cell clusters or mono- layers, consistent with the extensive underlapping of adjacent cells. E₂treatment had no overt effect on either vinculin or talin adhesion plaque staining until 6 days, when adhesion plaque- free zones were observed sporadically distributed throughout a monolayer culture (Fig. 3D). Otherwise, the remainder of the culture was indistinguishable from untreated cultures (Fig. 3C). After 9 days, talin- and vinculin-rich adhesion plaques were reduced in size and number per field, while the plaques of untreated cells were slightly smaller but still numerous (Fig. 3E,Fand G,H). E₂induced an overall reduction in the size and number of adhesion plaques over a 12-day time course of treatment. The hormone-induced reduction of vinculin- and talin-rich adhesion plaques was not absolute. Regions of treated monolayers retained a typical plaque distribution although the extent of these unaffected areas never approached that of widespread plaque-depleted areas. The retention of plaques likely reflects the presence of E₂-insensitive subpopu- lations of MCF-7 cells or residual adhesion plaques.

During the period of E₂treatment a diffuse intracellular staining for either talin or vinculin was detected as early as after one day and increased significantly throughout the 12-day period. The relatively low intensity of this intracellular staining posed no problem in comparison with adhesion plaques from shorter duration cultures (<6 days). However, the high intensity of this staining in 9- and 12-day cultures raised the concern that details of underlying adhesion plaques might be obscured and thus account for the observed reduction in plaque distribution. Resolution of this matter was obtained by optical sectioning with confocal microscopy. Examination of ventral regions underlying and surrounding nodules indicated that anti-vinculin-labelled adhesion plaques were sparse in the monolayer (see Fig. 5A). Anti-talin-labelled plaques were virtually absent, with only small, thin plaque-like or dot staining visualized (see Fig. 5C). These results agreed with the overall loss of plaque staining obtained with conventional flu- orescence microscopy described above.

MCF-7 cells maintained in E₂-deprived medium form CCAJ at apical regions of adjacent cells. The typical staining pattern obtained from double labelling for either vinculin or plako- globin showed the colocalization of the two plaque proteins with F-actin (not shown). Talin is specifically associated with CMAJ (Geiger et al., 1985; Burridge et al., 1988) and, consis- tent with this, anti-talin staining was never detected at MCF-7 CCAJ (data not shown). E₂-treated cultures showed a marked reduction in the extent of vinculin staining in this region, indi- cating a loss of association with the cytoplasmic adhesion plaque of CCAJ (Fig. 4). The change in vinculin distribution in CCAJ occurs earlier in response to E₂than that observed for CMAJ. Discrete cell-cell vinculin staining is reduced after 3 days of cell growth in the presence of E₂compared with controls (Fig. 4A,B). This reduction in the extent of discrete staining proceeds through 6 days (Fig. 4D); after 9 days (Fig. 4F) discrete cell-cell vinculin staining is detected only in occa- sional areas of treated monolayers. Vinculin (Fig. 4C,E), F- actin and plakoglobin staining remained unchanged in control cultures (latter two proteins not shown). Staining for F-actin and plakoglobin indicated that the two proteins remained colo- calized despite the presence of E₂(not shown).

Discrete accumulations of vinculin staining could not be detected at borders of adjoining cells within the nodule mass when optically sectioned by laser scanning confocal microscopy (Fig. 5B). Diffuse vinculin staining was observed in nodule cells and, considering the absence of any staining by nonimmune IgG, could represent a soluble pool of the plaque protein. Plakoglobin and F-actin were both concentrated at adjoining cell borders (Fig. 5Eand F, respectively). Plakoglo- bin staining was absent in areas of nonadjoining cells, which in postconfluent cultures are only found at the apex of nodules (not shown), consistent with its known accumulation only at sites of cell-cell contact (Cowin et al., 1986). Plakoglobin is also present in desmosomes as well as CCAJ (Cowin et al., 1986), raising the possibility that staining of nodule cell-cell borders reflects an extensive distribution of desmosomes. Although some of the plakoglobin staining within nodules and in the surrounding monolayer can be attributed to the presence of desmosomes, the bulk of it cannot. Serial optical sections in the Z-axis show that each internal nodule cell is entirely bounded by plakoglobin. If all of this staining is associated with desmosomes, then ultrastructural analysis of nodules would indicate the presence of that adhesion type all along adjoining cell membranes. Examination of thin sections cut perpendicular to the vertical axis of nodules demonstrated that desmosomes are in low abundance along adjacent cell membranes (see Fig. 8). Therefore desmosomes alone cannot account for the intensely stained plakoglobin accumulations detected along adjacent cells by confocal analysis. Contacts between nodule cells consisted mainly of microvilli-like cyto- plasmic strands (see Fig. 8D). The similar staining pattern of F-actin and plakoglobin observed with confocal microscopy in parallel preparations of nodules suggests that the intercellular contacts of nodule cells are derived from CCAJ.

That the absence of vinculin in CCAJ of nodule cells is not due to inaccessibility of antibody to these sites is supported by the following. First, residual anti-vinculin- or anti-talin- labelled adhesion plaques are detected beneath the nodules (Fig. 5A). Second, vinculin-rich plaques are detected simulta- neously beneath the monolayer and at CCAJ in E₂-untreated and early E₂-treated cultures (Figs 3, 4, and 5). Finally, plakoglobin antibody is distributed as a discrete, intensely staining pattern at cell borders throughout the nodular mass. All three of these points demonstrate the accessibility of antibodies under the fixation-permeabilization conditions employed here. It should be borne in mind, however, that, despite the apparent absence of vinculin from nodule cell-cell borders, the level of structural resolution achieved in this study does not preclude the presence of vinculin within a finer structure distributed more uniformly over the cytoplasmic face of the membrane. This question can best be resolved by immunolabelling at the ultrastructural level. Nonetheless, the absence of discrete vinculin accumulations from F-actin and plakoglobin-rich intercellular contacts suggests that these adhesions may be less stable than the typical CCAJ counterpart seen in untreated cultures.

Talin antiserum was used to determine whether CMAJ are present at nodule cell-cell junctions. Staining for talin was never localized as discrete accumulations at adjoining cells but was present as a diffuse cytoplasmic pattern (Fig. 5D). Fur- thermore, immunofluorescence using antibodies to human fibronectin and vitronectin that cross-react with the bovine counterpart detected only weak intensity cytoplasmic staining but not discrete fibrils or junctional accumulations (not shown). In contrast, plakoglobin, a marker for CCAJ-type adhesion plaques, was always present at cell-cell borders of nodules. Thus the intense and similar patterns of staining for F-actin and plakoglobin, combined with the absence of discrete talin or extensive ECM protein staining, and the general absence of desmosomes, suggest that nodule intercellular adhesions are derived from CCAJ.

Subconfluent, confluent and postconfluent MCF-7 cells that have been maintained in low-E₂or E₂-deprived medium display prominent intercellular spaces as detected with phase- contrast optics (Fig. 1). Rhodamine-phalloidin staining demon- strated that the apparently structureless regions are actually rich in F-actin. A distinctive vesicular subapical pattern of F- actin staining, believed to arise from the overlapping of inter- cellular cytoplasmic strands, was found to correspond to the phase-contrast spaces (Fig. 6A,B). The pattern was distinct from that of CCAJ (Fig. 6C,D). F-actin staining in the baso- lateral region extended continuously from the ventral region of adjacent cells upwards to the apical region, where the pattern appears to converge with typical CCAJ (not shown). Consis- tent with the loss of the phase-contrast spaces after E₂treatment, F-actin staining was rearranged into a more regular or linear pattern at basolateral regions of adjoining cells (Fig. 7compare A and C). Ultrastructural studies showed that, in untreated cultures, thin cytoplasmic strands are present between, and often contact, adjacent cells (Fig. 8A,B). Typical CCAJ are present at the apical region of adjoining cells. Desmosomes were never found in high density and were never situated along the entire length of opposing membranes of untreated cells. In contrast, the membranes of adjacent E₂- treated cells are closely apposed, with microvillous structures, some CCAJ-like plaque structures, and more sparse desmo- somes (Fig. 8).

In an effort to determine whether these overlapping strands form adhesions, cells were stained with antibodies to the different adhesion plaque proteins. Plakoglobin colocalized with F-actin in a nearly identical vesicular pattern (Fig. 7), extending from the ventral region continuously upwards to and converging at CCAJ. When cells were treated with E₂the vesicular plakoglobin staining pattern was rearranged into a more linear, but punctate, pattern, just as was observed for F- actin (Fig. 7, compare A,B with C,D). Vinculin staining was rarely, and talin staining never, detected along the actin strands (data not shown). Also, antibodies to keratin and cytokeratins failed to label this region, despite fibrillar or diffuse cell-body staining, in agreement with results of ultrastructural studies indicating a lack of desmosomes in this region. The colocal- ization of F-actin and plakoglobin along basolateral surfaces suggests the presence of potentially functional CCAJ-derived structures that are spatially removed from the typical, apically situated CCAJ.

MCF-7 carcinoma cells grown in low E₂or E₂-deprived medium display a morphology characteristic of normal epithe- lial cells. Despite the similarity to normal epithelial cells, MCF-7 cells acquire more tumor cell-like characteristics after E₂treatment. Particularly notable in this respect is the E₂- dependent development of nodules in postconfluent cultures, an event suggesting that E₂promotes radical changes in cellular adhesions. We find that, in vitro, E₂treatment of MCF- 7 cells results in a gradual reduction in staining for vinculin and talin, suggesting a loss of cell matrix adhesion. Parallel to the loss of CMAJ/adhesion plaques vinculin staining was diminished at CCAJ of monolayer cells, while staining for F- actin and plakoglobin was retained at these sites. Subsequently vinculin could not be detected at nodule cell-cell junctions, in contrast to the intense plakoglobin and F-actin staining observed along cell-cell borders. The E₂-induced changes in adhesion plaque composition observed in this study may reflect a loss of stable CMAJ and CCAJ.

The cytoplasmic adhesion plaque of cell-matrix adhesions consists of several cytoskeletal proteins (Burridge et al., 1988; Geiger and Ginsberg, 1991), which potentially link F-actin to the cytoplasmic domain of the integrin β₁subunit. In vitro, talin has been shown to bind the integrin β₁subunit (Horwitz et al., 1986) and vinculin (Burridge and Mangeat, 1984). Vinculin has also been shown to bind the actin-bundling protein, α-actinin (Wachsstock et al., 1987). The sequential interaction of these proteins in vivo may anchor cells to the extracellular matrix by physically linking the internal micro- filament system to external substratum components. The absence of discrete vinculin- and talin-rich adhesion plaques on E₂-treated MCF-7 cells implies that the loss of cell-matrix adhesions is due to the disruption of the actin-integrin linkage. The loss of talin would remove appropriate sites for binding of either the β₁subunit or vinculin, while α-actinin would no longer be anchored to talin in the absence of vinculin. Moreover, other actin-membrane linkages or plaque-stabiliz- ing interactions, such as that between talin and F-actin (Muguruma et al., 1990), vinculin binding to paxillin (Turner et al., 1990), vinculin self-association (Milam, 1985; Molony and Burridge, 1985), α-actinin binding to both F-actin (Meyer and Aebi, 1990) and vinculin (Wachsstock et al., 1987), and zyxin binding to α-actinin (Crawford et al., 1992), may all be abolished in E₂-induced MCF-7 cells. α-Actinin alone may also mediate the linkage of F-actin to the β₁subunit (Otey et al., 1990). The subsequent dissolution of adhesion plaques would reduce substratum adhesion and increase the potential for individual cell or cell cluster motility. However, it remains to be directly tested whether all of these adhesion plaque-asso- ciated proteins are similarly affected by E₂as are vinculin and talin.

The actin filament system is also believed to be physically linked to cadherin cytoplasmic domains at CCAJ, mediated by a system of proteins analogous to those functioning in CMAJ plaques (Kemler and Ozawa, 1989; Geiger and Ginsberg, 1991; Tsukita et al., 1992). A hypothetical linkage of F-actin to the cytoplasmic domain has been proposed (Tsukita et al., 1992), where vinculin links α-actinin to α-catenin, which in turn is bound to cadherin. A physical link between the cadherin cytoplasmic domain and plakoglobin in vivo is suggested by co-aggregation of the two proteins into patches induced by cadherin antibodies (Knudsen and Wheelock, 1992). E₂- induced alterations in expression or binding activity of CCAJ plaque proteins could lead to the dissolution of stable CCAJ. Consequently E-cadherin, expressed by MCF-7 cells (Damsky et al., 1983; Wheelock et al., 1987), may be redistributed throughout the cell surface while retaining a limited capacity to mediate cell-cell adhesion. The inability of PC9 lung cancer cells to form tight cell-cell adhesions despite intense surface expression of E-cadherin is attributed to the absence of α- catenin in this cell line (Shimoyama et al., 1992). The loss of vinculin from MCF-7 CCAJ could have a similar effect by pre- venting the stabilizing linkage of α-actinin-F-actin to α- catenin, resulting in a uniform cell-surface redistribution of E- cadherin. The uniform distribution of plakoglobin at basolateral regions of monolayer cells and at the periphery of nodule cells suggests that E-cadherin is also present all along these opposing membranes, despite the absence of discrete CCAJ.

Downregulation of E-cadherin expression is generally cor- related to increased tumor cell invasiveness in several carcinoma types in vitro and in vivo (Frixen et al., 1991; Shiozaki et al., 1991; Bussemakers et al., 1992; Umbas et al., 1992; Mayer et al., 1993; Oka et al., 1993). Examination of antibody staining patterns in breast carcinoma tissue in some of these studies reveals that, despite decreased spatial expression, E-cadherin is often found at the entire periphery of ductal carcinoma cells and metastatic breast-cancer cell nests situated in distant lymph nodes (Oka et al., 1993). The spatial distribution of E-cadherin along the periphery of tumor cells strongly suggests that invasive tumor cells can and do form unstable cell-cell adhesions via available surface E-cadherin. The loose aggregates formed by PC9 cells suggests that E- cadherin, uncoupled from the cytoskeleton, is capable of mediating weak cell-cell adhesions. A more recent finding shows that E-cadherin transfection of invasive breast carcinoma cells deficient in E-cadherin, β-catenin and plako- globin, but expressing α-catenin, results in uniform surface expression of E-cadherin and aggregation with L-cells express- ing the homologous cadherin (Sommers et al., 1993). Our pre- liminary work also indicates that E-cadherin staining at MCF- 7 cell-cell borders is diminished but not eliminated by E₂. All of these findings indicate that weak tumor-cell adhesion can occur in the absence of stabilizing plaque-protein interactions. Although the absence of vinculin may destabilize adhesions, unstable cell-cell adhesions may still be retained in nodules and monolayers. α-Catenin, if retained at these sites, or plakoglo- bin, may contribute to weak cadherin-mediated nodule and monolayer cell-cell adhesions. Small aggregates of α-catenin complexed with E-cadherin may be capable of mediating weak, labile cell-cell adhesions while the absence of vinculin prevents the assembly of the small aggregates into larger stable complexes, that is, CCAJ. Thus E₂-induced MCF-7 monolay- ers and nodules may retain a degree of cohesiveness due to weak or unstable cell-cell adhesions. Simultaneously, individ- ual MCF-7 cells within the monolayer or nodules may become increasingly motile. The formation of transient cell-cell adhesions would potentially allow motile tumor cells to use surrounding monolayer or nodule cells as a substratum upon which to migrate.

MCF-7 cells formed prominent intercellular spaces under low-E₂culture conditions employed in this study. The spaces or gaps are actually actin and plakoglobin-rich cytoplasmic strands that overlap or contact strands of neighboring cells. The codistribution of F-actin and plakoglobin at subapical/basolat- eral regions of adjacent cells suggests that CCAJ-derived adhesions extend all along the membranes of this region, par- ticularly where intercellular strands contact. The sporadic presence of vinculin accumulations indicates that small, possibly transient, CCAJ do form in the basolateral area while the general absence of this plaque protein indicates that the adhesions are not stable structures. Consistent with the formation of transient CCAJ, preliminary long-term time-lapse video microscopy indicates that the intercellular spaces are in a constant state of flux, appearing and disappearing. It is possible then that adjacent cells are constantly making and breaking contacts in the basolateral region, in effect trying to translocate, but are constrained by the presence of CCAJ and CMAJ. The constraints may be removed by E₂-induction. It is not yet clear whether the basolateral contacts are related to intercellular contacts of nodule cells. The general absence of vinculin, presence of plakoglobin, and abundance of F-actin- rich microvillous cytoplasmic structures between adjacent cells, suggest a common origin. An intriguing possibility is that once CCAJ and CMAJ are weakened in response to E₂, indi- vidual cells may begin to crawl over each other, leading to mul- tilayering and, combined with proliferation, even to nodule development.

Cell-matrix and cell-cell adherens junctions maintain the integrity and organization of epithelial cell layers in vivo and in vitro. The development of MCF-7 cell nodules in response to E₂is a dramatic departure from this natural order, arising from destabilization of these adhesions, and may reflect events related to tumor growth in vivo. Recently, E-cadherin and vinculin were shown to be retained at CCAJ during cell division of normal epithelium and cultured epithelial cells (Baker and Garrod, 1993). The authors conclude that the loss of CCAJ in neoplastic cells, required for invasion and metas- tasis, most likely does not result solely from over-proliferation of these cells. Thus E₂-induced proliferation of MCF-7 cells, alone, probably does not initiate CCAJ disassembly. E₂may promote predisposed MCF-7 cells to escape the constraints of stable cellular adhesions through the loss or inactivation of specific cytoskeletal adhesion-plaque proteins, such as vinculin. Therefore, the MCF-7 cell multilayering and nodule development may be the combined result of E₂-induced hyper- proliferation and incapacity to form stable, immobilizing CCAJ.

MCF-7 cells depend on E₂for tumor formation in immuno-compromised mice (Soule and McGrath, 1980; Shafie and Liotta, 1980) and nodule development in postconfluent cell cultures (Gierthy et al., 1991). Nodule development may therefore reflect E₂-dependent tumorigenic events that occur in vivo. Particularly notable in this respect is the loss of vinculin staining at CMAJ and CCAJ, which parallels cell multilayer- ing and nodule formation. The uncoupling of vinculin from the adhesion-plaque complex may destabilize the adhesion and allow cellular overgrowth. Consistent with this argument, SV40-transformed 3T3 and BSp73ASML cell lines expressing low or undetectable levels of vinculin also form tumors in mice (Rodriguez Fernandez et al., 1992). Restoration of normal levels of vinculin expression by transfection suppresses the tumor-forming capability of these cells, implying that vinculin is a necessary factor for tumor suppression. Furthermore, the absence of focal contacts and reduced cell-cell adhesion in mutant F9 cells has also been directly attributed to reduced vinculin expression (Samuels et al., 1993). Therefore a potential mechanism underlying MCF-7 nodule growth is an E₂-promoted reduction in expression or post-translational modification of vinculin, either of which could prevent its incorporation into and stabilization of adhesion plaques. The similarity in E₂-dependence between MCF-7 tumor growth in vivo and nodule development in vitro suggests that inactiva- tion of the vinculin tumor suppression function may be a fun- damental event for E₂-dependent human breast tumor growth.

J.D. extends his gratitude to Dr James Turner and Don Szarowski for their guidance with confocal microscopy. The confocal micro- scope is part of the WCLR Biomedical Imaging and Reconstruction Resource (BMIRR). This work was supported by EHS grant no. ESO3561.