ECM-induced gap junctional communication enhances mammary epithelial cell differentiation

Gap junctional intercellular communication (GJIC) is critical in diverse cell and tissue functions (White and Paul,1999). Gap junctions are perceived as 'modulators of cellular differentiation' in several systems (Pitts et al., 1988; Paul et al.,1995; Bruzzone et al.,1996; Kumar and Gilula,1996) and recent studies have revealed a role for connexins in stratification and differentiation of human epidermal cells(Wiszniewski et al., 2000) and lung alveolar epithelial cells (Alford and Rannels, 2001), and in bone homeostasis, promoting osteoblast differentiation (Gramsch et al.,2001; Romanello et al.,2001; Schiller et al.,2001). The basic structural component of the gap junction is connexin (Cx). Connexins constitute a family of more than 20 homologous proteins that are temporally and spatially distributed throughout the body(reviewed by Goodenough et al.,1996; Kumar and Gilula,1996; Kidder and Mhawi,2002).

Besides humoral mediators (Hynes et al., 1997; Hennighausen et al., 1997), the extracellular matrix (ECM) has been regarded as the dominant regulator of mammary differentiation(Weaver et al., 1997; Boudreau and Bissell, 1998; Schmeichel et al., 1998; Smalley et al., 1999; Klinowska and Streuli, 2000; Hansen and Bissell, 2000). The few studies that addressed the role of cell-cell interaction in mammary differentiation (Streuli et al.,1991; Desprez et al.,1993; Alford and Taylor-Papadimitriou, 1996; Hansen and Bissell, 2000)undervalued its role as compared with the effect exerted by the ECM.

Studies have suggested that gap junctions play a critical role in the coordinated changes through development, differentiation, maintenance and involution of the mammary gland (Monaghan et al., 1994; Monaghan et al.,1996; Pozzi et al.,1995; Yamanaka et al.,1997; Locke et al.,2000; Yamanaka et al.,2001). However, no studies have established a clear correlation between functional GJIC and mammary epithelial differentiation, either in vivo(Perez-Armendariz et al.,1995; Pozzi et al.,1995; Monaghan and Moss,1996; Locke et al.,2000) or in vitro (Lee et al.,1991; Lee et al.,1992; Tomasetto et al.,1993; Hirschi et al.,1996; Sia et al.,1999).

The CID-9 mouse mammary cell culture system, responsive to both lactogenic hormones and substrata, consists of a heterogeneous cell strain of epithelial,myoepithelial and fibroblastic cells and is a widely accepted model that mimics in vivo differentiation of mammary cells(Schmidhauser et al., 1992; Talhouk et al., 2001). To determine the role of GJIC in modulating the differentiation phenotype of mammary cells, gap junction proteins of mammary CID-9 cells were characterized and their regulation by ECM assessed. The cause-and-effect relationship between GJIC and mammary epithelial differentiation was also investigated. We demonstrate that mammary CID-9 cells express Cx26, Cx32 and Cx43 proteins,which are modulated by ECM, and that proper cell-ECM interaction favours GJIC. Our studies suggest CID-9 cells are capable of differentiating and expressingβ-casein in the absence of an exogenous basement membrane in aβ ₁-integrin-independent pathway, provided the cells are coupled via functional gap junctions.

Highest grade materials were used: 18α glycyrrhetinic acid (18αGA), 8-Br-cAMP, bovine serum albumin (BSA), insulin, Lucifer Yellow (LY),ovine hydrocortisone, ovine prolactin, polyHEMA [poly-2-hydroxyethyl methacrylate], propidium iodide and Trypsin-EDTA were obtained from Sigma (St Louis, MO). Protease inhibitors (Complete™, Boehringer Mannheim,Germany). Hybond-N membrane, Rediprime kit and α-³²P dCTP were from Amersham Pharmacia Biotech (Uppsala, Sweden). Affinity-purified polyclonal rabbit anti-Cx 26 (Cat# 71-0500), 32 (Cat# 71-0600) and 43 (Cat#71-0700) antibodies raised against peptide portions of the cytoplasmic domains of the respective connexins were from Zymed Laboratories (San Francisco, CA). Function-blocking integrin antibody against the β₁ (Ha2/5)integrin subunit was purchased as azide- and endotoxin-free from PharMingen(San Diego, CA). Enhanced chemiluminescence (ECL) and horseradish peroxidase(HRP)conjugated anti-rabbit IgG were from Santa Cruz Biotechnology (Santa Cruz, CA), FITC-conjugated secondary goat anti-rabbit IgG (H+L) and Prolong antifade were from Molecular Probes (Eugene, OR). Cell culture media and reagents were purchased from Gibco BRL Life Technologies (Parsley, UK). MatTek glass bottom tissue culture plates and Bio-Rad protein assay were from MatTek(Ashland, MA) and Bio-Rad (Hercules, CA), respectively. EHS-matrix growth-factor-reduced Matrigel was purchased from Collaborative Biomedical Products (Bedford, MA). CID-9 mammary cell strain, polyclonal rabbit anti-mouse milk antiserum and β-casein c-DNA inserts were provided by Mina Bissell (Lawrence Berkeley National Laboratory, Berkeley, CA, USA).

A low passage number (17 to 21) of the CID-9 mouse mammary cell strain was used throughout. Cells were grown in 'growth medium' consisting of Dulbecco's Modified Eagle's Medium Nutrient Mixture F12 Ham (DMEM/F12) with 5% FBS,insulin (5 μg/ml) and gentamycin (50 μg/ml) in a humidified incubator(95% air 5% CO₂) at 37°C. Cells were propagated by trypsinization and plated either on tissue culture plastic petri dishes or on petri dishes coated with different substrata.

CID-9 cells were seeded at 3.0×10⁶ or at 5.0×10⁶ cells/75cm² dish on culture dishes or dishes coated with the reconstituted basement membrane, growth-factor-reduced Matrigel, respectively. Alternatively, diluted Matrigel (1.5% vol/vol) in HBSS was dripped onto cells 24 hours after plating(Streuli et al., 1995a). Cells cultured on EHS-matrix were directly plated in differentiation or non-differentiation media consisting of DMEM/F12 containing insulin (5μg/ml), hydrocortisone (1 μg/ml) and either supplemented with or lacking ovine prolactin (3 μg/ml), respectively. Cells cultured on plastic or dripped with EHS-matrix were first plated in growth medium for initial cell attachment and spreading. Twenty-four hours after plating, cells were washed three times with HBSS, and the growth medium was replaced with either differentiation or non-differentiation media. Media were changed on a daily basis.

PolyHEMA, a non-adhesive substratum, was prepared using a solution of 6 mg/ml in 95% ethanol and was added to culture plates at 5.0×10^-2 ml/cm² and allowed to evaporate to dryness at 37°C. The plates were then washed twice with HBSS and CID-9 cells were plated at a concentration of 5.0×10⁵ cells/ml and in differentiating media. Since the cells were grown in suspension, cAMP diluted in differentiating media was added to the existing media on days 1, 3 and 5 of culture. On day 6 of culture, the cells were harvested for analysis.

Total RNA was extracted from cells at day 6 after plating as described elsewhere (Chomczynski and Sacchi,1987). For northern analysis, 5 μg of total RNA were electrophoresis through 1% agarose/formaldehyde gel, blotted overnight onto Amersham Hybond-N membrane in 10× SSC and UV crosslinked for subsequent hybridization. β-Casein c-DNA inserts were ³²PdCTP-labelled using Rediprime kit and hybridization was performed overnight at 42°C in a shaker-incubator. The blots were then washed at high stringency (0.1% SSC,65°C) and signals were detected by fluorography.

Proteins were extracted by scraping the cells into lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P40, 0.5% sodium deoxycholate). The scraped cells were then sheared by passing them several times through a 21-gauge needle. Protease inhibitors were added at a concentration of 40 μl per 1 ml of lysis buffer, and the cell extracts were centrifuged. The protein content of the supernatants was determined by Bio-Rad assay and equal amounts of protein were resolved by gel electrophoresis. To detect milk proteins, the membranes were blocked overnight in a wash buffer (100 mM Tris-HCl buffer, pH 7.5, 150 mM NaCl. 0.3% Tween 20) with 2% fatty acid-free BSA. The membranes were then incubated for 1 hour in polyclonal rabbit anti-mouse milk antiserum at room temperature and washed three times, for 20 minutes each, to remove unbound antiserum.

For Cx26, Cx32 and Cx43 proteins, membranes were blocked for 1 hour in a wash buffer [Dulbecco's phosphate buffered saline (PBS), 0.1% Tween 20] with 3% skim milk. They were then incubated for two hours in a humid chamber in the corresponding polyclonal rabbit anti-Cx antibody at a concentration of 0.5μg per ml of blocking buffer. Bound antibody was detected by enhanced ECL for casein, Cx32 and Cx43 immunoblots. Addition of HRP-conjugated anti-rabbit IgG followed by tetramethyl benzidine (TMB) was used for detection of Cx26 immunoblots.

Cultured CID-9 cells were washed three times with warm HBSS and fixed in ice-cold (-20°C) 70% ethanol overnight. Fixed cells were first rinsed twice with PBS and then incubated for 1 hour at room temperature with 3%normal goat serum. After blocking, cells were labelled for 2 hours at room temperature with rabbit anti-connexin 26, 32 and 43. This was followed by labelling with a FITC-conjugated secondary goat anti-rabbit IgG (H+L) that was incubated for 1 hour with the fixed cells. Concentrations of the primary and secondary antibodies were used as recommended by the supplier. Nuclei were then counter-stained by incubation for 3 minutes with propidium iodide at 5μg/ml. Washing with PBS was performed twice between incubations. Finally,cells were mounted on slides and staining was preserved by addition of antifade to the stained cells, which were kept at 4°C. Cells were then observed under fluorescence microscopy (LSM 410, Zeiss, Germany).

For microinjection assay, CID-9 cells were cultured on MatTek glass bottom tissue culture plates. Cells were microinjected with 5% LY CH in 150 mM LiCl. The solution of LY was injected by pressure injection into cells through microelectrodes. The spread of the dye fluorescence to neighbouring cells was recorded photographically using fluorescence microscopy.

For scrape-loading assay, CID-9 cells were cultured on plastic or on EHS-drip, into 4-chamber polystyrene vessel tissue culture-treated glass slides. After 24 hours, medium was supplemented with 10 μM 18α GA, or 50 μM 8-Br-cAMP.

The scrape-loading method was performed as described elsewhere(El-Fouly et al., 1987). Cells plated on 4-chamber vessel slides were washed three times with warm HBSS before addition of LY at 0.1% dilution in PBS. Using a scalpel, cuts were made throughout the monolayer, followed by incubation for 10 minutes at 37°C. The cells were then washed with warm HBSS and fixed with 4% formaldehyde. Slides were preserved by mounting in antifade and stored at 4°C. Observation of dye spread was recorded photographically as described earlier.

CID-9 cells were seeded on EHS-matrix and treated with 10 μM 18αGA on day 1 of culture. Medium supplemented with 18α GA was changed on a daily basis up to day 10 in culture. Alternatively, 18α GA was supplemented to the medium for the first 5 days in vitro and later removed from the medium for days 6-10. Trypan Blue staining was used to determine cell viability as affected by 18α GA for the duration of the treatment. Samples were counted in triplicate wells. Proteins for western blot analysis(normalized to equal cell counts) were extracted on day 5 and day 10 of culture. Control cells were not treated with 18α GA.

CID-9 cells treated with 8-Br-cAMP were seeded in 100 mm petri dishes on tissue culture plastic. Twenty-four hours later, growth medium was replaced with differentiation medium and 8-Br-cAMP was added at a concentration of 50μM. Medium supplemented with 8-Br-cAMP was changed on a daily basis up to day 5 of culture, when proteins for western blot analysis were extracted. Control cells were not treated with 8-Br-cAMP.

CID-9 cells were seeded in 6-well plates on plastic substratum or EHS-drip. Twenty-four hours after plating, cells were washed three times with HBSS and growth medium was replaced with differentiation medium containing 100 μg/ml of the function-blocking β₁ integrin antibody. The antibody was supplemented daily with the media and the cells were harvested at day 4 after plating.

β-Casein, connexin expression and functional GJIC using LY scrape-load assays were quantified from different experiments using NIH Image 1.62 software. Quantification of β-casein and connexin expression was normalized with respect to β-actin. For LY scrape-load assays,quantification was based on measuring the integrated fluorescence intensity,at the scrape site, over an equivalent area in both control and experimental conditions. The degree of significance of variations between control and experimental values was assessed by ANOVA uni-variant test using the Graph Pad Prism software version 3.00. Where applicable, the quantification was from three different experiments.

CID-9 cells exhibited different morphologies in vitro depending on the substratum upon which they were plated. On tissue culture plastic, CID-9 cells spread to form a monolayer by day 6 in culture(Fig. 1Aa), whereas when cells were dripped with EHS-matrix (EHS-drip), CID-9 cells grew in multi-layers and aggregated in sheet-like structures (Fig. 1Ab). However, when plated on EHS-matrix, CID-9 cells grew in clusters (Fig. 1Ac). The expression of β-casein mRNA was only noted in cells grown in the presence of prolactin and EHS-matrix, regardless of whether the cells were dripped or plated on EHS-matrix. Cx43 mRNA was equally expressed on plastic and EHS-drip and was drastically downregulated on EHS-matrix, whereas Cx26 mRNA was non-significantly affected by EHS-matrix(Fig. 1B).

To determine the effect of substratum on expression of connexin proteins by CID-9 cells and their subcellular distribution, western blot analysis and immunolocalization studies were performed. Whereas no significant change was noted in Cx32 expression, slight but significant (P<0.05)upregulation of Cx26 expression in CID-9 cells was noted on EHS-drip as compared with cells on plastic. All Cx43 isoforms (NP: non-phosphorylated inactive form; P1 and P2: phosphorylated, active forms) were expressed on plastic, EHS-drip and on EHS-matrix on day 6 of culture. However, on EHS-matrix, the P2 form of Cx43 protein became predominant and was significantly (P<0.001) elevated compared with the levels detected on plastic or EHS-drip (Fig. 2A).

To assess the relationship, if any, between connexin intercellular distribution and cell differentiation, immunohistochemistry was employed to grade for connexin cytosolic or plasma membrane distribution on either plastic or EHS-drip after 6 days of culture. Cx26, Cx32 and Cx43 distribution was mainly cytosolic when cells were grown on plastic(Fig. 2Ba, Bc, Be). In certain limited areas of the culture, Cx26, Cx32 or Cx43 localized to the plasma membrane of cells on plastic (Fig. 2Ba, Bc,Be; inserts). Connexin distribution became predominantly membranous when cells were cultured on EHS-drip(Fig. 2Bb, Bd, Bf). Note that certain cells, as evident by immunostaining, did not express Cx26 and Cx43 under any of the culture conditions studied. Whether these cells are fibroblastic, myoepithelia or epithelial was not addressed in our studies.

The change in connexin distribution from intercellular on plastic to plasma membrane for cells grown on EHS-drip compared with those on plastic strongly suggests that communication via gap junctions is enhanced by cell-ECM interaction. To test for gap junction functionality, cells were injected with LY or scraped, and transfer of fluorescent dye into neighbouring cells was observed. In confluent cultures of CID-9 cells plated on plastic, dye transfer was limited to a few cells neighbouring the dye-injected cell(Fig. 3A); similarly, dye transfer was restricted to 1-2 layers deep beyond the scrape site(Fig. 3B). However, on EHS-drip, LY was transferred to several cells immediately neighbouring the microinjected cell (Fig. 3C)and beyond 4-5 cell-layers deep away from the scrape site(Fig. 3D). Quantitative analysis, as described in the Materials and Methods, of the integrated fluorescence intensity at the scrape site from three different experiments showed an increase in fluorescence from 36±10 for cells on plastic to 132±36 for cells on EHS-drip. This demonstrated that cells on EHS-drip communicate via gap junctions more readily than cells on plastic.

To determine if gap junctional communication is necessary for β-casein expression, two approaches were undertaken. In the first, 18α GA was utilized to inhibit gap junctional communication of cells cultured on EHS-matrix. In the other, cAMP was utilized to enhance gap junctional communication by cells on plastic. In contrast to control cells, those cells treated with 10 μM 18α GA and cultured on EHS-drip showed disrupted GJIC, and downregulated their Cx43 and β-casein expression without drastically affecting cell viability as determined by Trypan Blue staining(Fig. 4A). Disrupting GJIC also affected the clustering behaviour of CID-9 cells grown on EHS-matrix, yielding smaller size clusters that lacked well-defined membrane-like boundaries(Fig. 4Bb compared with Fig. 4Ba). β-Casein protein expression on day 6 of 18α GA-treated culture was dramatically decreased compared with the control (upto 95%) non-treated cells. The effect of the gap junction inhibitor 18α GA was reversible. Adding media devoid of 18α GA beyond day 6 showed that partial recovery of clustering morphology was noted (Fig. 4Bc). β-Casein expression was also partially (approximately 50%) recovered 4 days after the depletion of 18α GA from the medium(Fig. 4C).

cAMP-treated CID-9 cells plated on plastic showed a threefold enhanced LY transfer beyond 2-3 cell-layers deep away from the scrape site. cAMP also induced aggregation of CID-9 cells, resulting in enhanced clustering(Fig. 5A). Northern and western blot analyses showed that cAMP-treated CID-9 cells on plastic expressedβ-casein mRNA and proteins (Fig. 5B,C) as well as higher levels of Cx26 and Cx43 proteins on day 6 of culture compared with the non-treated control cells grown in parallel(Fig. 5C).

Although cAMP enhanced Cx43 and Cx26 expression and upregulated gap junctional communication as well as β-casein expression, it was not clear whether casein expression was owing to enhanced gap junctional communication or other cAMP-mediated effects. To assess this, cells were treated with both cAMP and 18α GA. This treatment induced downregulation of β-casein(Fig. 6A).

To find out whether the cAMP effect on CID-9 cells may also require cell-ECM interaction, cAMP-treated cells were incubated with function-blockingβ ₁-integrin antibody. Whereas, in cAMP-treated cultures,β-casein expression was partially downregulated by β1-integrin function-blocking antibody, this treatment markedly (P<0.01)decreased β-casein production in cells cultured on EHS-drip(Fig. 6B). This suggested that,in the presence of cAMP, treated cells, with enhanced GJIC, can expressβ-casein in a β₁-integrin-independent manner.

cAMP also induced differentiation of CID-9 cells in suspension independently of a cell-ECM interaction. CID-9 cells were plated on the non-adhesive polyHEMA substratum and treated with cAMP. Cells on PolyHEMA grew in small clusters not exceeding 5-6 cells in each cluster and did not expressβ-casein. By contrast, cells on PolyHEMA treated with cAMP aggregated into larger clusters, and upregulated their Cx43 expression. These cells also expressed β-casein (Fig. 7).

The CID-9 mammary cell strain is a suitable model to study connexin expression and regulation of GJIC in relation to ECM-induced mammary cell differentiation (Talhouk et al.,2001) because these cells differentiate, in culture, in an ECM-dependent manner (Schmidhauser et al.,1990; Schmidhauser et al.,1992; Desprez et al.,1993).

Evidence from the literature has shown that the main role of cell-cell interactions is to enhance mammary cell differentiation, but in a matrix-dependent manner (Emerman et al., 1977; Lee et al., 1985; Streuli and Bissell, 1990; Petersen et al., 1992; Desprez et al., 1993; Talhouk et al., 1993; Slade et al., 1999). A study by Streuli et al. emphasized the importance of cell-cell interaction in mammary cell differentiation when they demonstrated that casein production was synergistically elevated upon cell-cell interaction(Streuli et al., 1991). A consequence of cell-cell interaction could be intercellular communication through gap junctions. Although the importance of gap junctions is well established in the differentiation of many cell types(Pitts et al., 1988; Paul et al., 1995; Bruzzone et al., 1996; Kumar and Gilula, 1996; Wiszniewski et al., 2000; Alford and Rannels, 2001; Gramsch et al., 2001; Romanello et al., 2001; Schiller et al., 2001), few studies have characterized gap junctions in the mammary gland. The major connexins described in the mammary gland are Cx26, Cx32 and Cx43(Lee et al., 1991; Lee et al., 1992; Tomasetto et al., 1993; Perez-Armendariz et al., 1995; Pozzi et al., 1995; Monaghan and Moss, 1996; Yamanaka et al., 1997; Sia et al., 1999; Yamanaka et al., 2001), but no studies have established a clear correlation between connexin expression,functional GJIC and mammary epithelial differentiation. In this study, we provided evidence for the importance of intercellular interaction for targeting and functionality of connexins in mammary cells, and the differentiated phenotype resulting thereof.

The use of northern and western blot analysis coupled with immunolocalization studies established that CID-9 cells expressed and modulated all major mammary gland connexins and provided better insight to the locale and, hence, the role of connexin in the differentiation of mammary epithelial cells. The regulation of connexin expression by the ECM has been reported by Guo et al. (Guo et al.,2001), whereby lung alveolar type II epithelial cells modulated the expression and distribution of Cx43 and Cx26 depending on the substrata the cells were plated on. The downregulation of Cx43 mRNA but not Cx26 mRNA in differentiated CID-9 cells cultured on EHS-matrix was reminiscent of earlier studies (Rosenberg et al.,1996), whereby Cx26 and Cx43 transcripts declined whereas those of Cx32 increased as hepatic cells differentiated. In this study, the P2 phosphorylated form of Cx43 was upregulated in cells on EHS-matrix, suggesting that Cx43 regulation during cellular differentiation occurred at the post-translational level through phosphorylation. This correlated with in vivo studies (Yamanaka et al.,1997), and was supported by data from our laboratory demonstrating that, during lactation, Cx43 mRNA in the mammary gland is downregulated whereas the P2 form of Cx43 protein is upregulated (R.S.T. et al.,unpublished).

On day 6 of culture on EHS-drip, Cx26, Cx32 and Cx43 localized mostly to the plasma membrane. By contrast, connexins localized mostly to the cytoplasm in cells cultured on plastic. Studies performed on human luminal and basal mammary cells in culture revealed mostly cytoplasmic, rarely membranous, Cx26 staining associated with luminal epithelial cells(Monaghan et al., 1996). Cx43 localized mostly to membranes of cells with large nuclei of the CID-9 cell population that stain positive for smooth muscle actin. These represent the myoepithelial-like subpopulation of CID-9 cells(Desprez et al., 1993). Other studies (Monaghan et al.,1996; Yamanaka et al.,1997; and Locke,1998) reported Cx43 localization to myoepithelial cells. Cx26 and Cx32 immunostaining had no distinct cell-type distribution. These data supported by gap junction functionality assays, suggested that at least Cx26,Cx32 and Cx43 redistributed and assembled, by day 6 in culture, into communicating junctions due to matrix-dependent post-translational modifications. Similar data were reported for redistribution of intracellular stores of Cx43 in a quiescent MAC-T bovine mammary epithelial cell line(Sia et al., 1999) and in alveolar epithelial cells as demonstrated by Guo et al.(Guo et al., 2001). It is difficult to rule out the potential involvement of other connexins that we did not assay for.

Thus, differentiation is coupled to enhanced connexin membrane localization and GJIC. However, whether gap junctions actually mediate the differentiation process and/or whether the differentiation process depends solely on ECM-transduced signals is not clear. The differentiation re-acquired by CID-9 cells on EHS-matrix upon withdrawal of 18α GA suggested that cell-ECM interaction was not sufficient to maintain an optimal differentiated phenotype, whereas proper cell-cell and cell-ECM interaction were crucial for optimal differentiation. Many studies reported that 18α GA blocked GJIC in different cell types (Martin et al.,1991; Taylor et al.,1998; Venance et al.,1998) and others have shown that 18α GA rapidly and reversibly blocked GJIC and that extended exposure to 18α GA inhibited both Cx43 mRNA and protein expression in a time- and dose-dependant manner(Guo et al., 1999). In the present study, we showed that 18α GA downregulated Cx43 expression. The mechanism of action of 18α GA on connexin expression requires further elucidation.

Regulation of gap junctions by cAMP has been demonstrated at the transcriptional, translational and post-translational levels. The latter was mediated by phosphorylation of connexin proteins, by redistribution of gap junction plaques to the cell membrane, or even by increase in the rate of trafficking of individual connexins to the membrane(Atkinson et al., 1995; Paulson et al., 2000; Lampe et al., 2000; Lampe and Lau, 2000). In this study, we showed that cAMP-enhanced GJIC was coupled to increased Cx43 expression and its phosphorylation as well as increased Cx26 expression by cells on plastic. This was in agreement with other studies(Furger et al., 1996) where cAMP treatment of human granulosa cells enhanced GJIC by increasing the levels of Cx43 phosphorylation. Most importantly, enhanced GJIC due to cAMP was implicated in CID-9 differentiation even in the apparent absence of an exogenous basement membrane. In addition to the increased β-casein production, cAMP led to a decrease in the production of laminin and collagen IV (Sfeir et al., 2001). Pervious studies (Streuli and Bissell,1990) showed that expression of laminin and type IV collagen were lowest in culture conditions that favoured β-casein expression. However,this correlated with deposition of a continuous basement membrane. The increased β-casein expression in our studies could have thus occurred in a gap-junction-independent pathway and as a result of the endogenously deposited basement membrane as suggested by previously(Petersen et al., 1992). However, the loss of differentiation as well as Cx43 expression by cAMP-treated cells together with a gap junction inhibitor, 18 α GA,suggested that the effect of cAMP on β-casein expression and mammary differentiation was achieved through a gap junctional communication manner. The concentration of 18α GA and the duration of its use were not cytotoxic to cells (Davidson et al.,1986), and showed no drastic effect on cell viability; hence, the downregulation of β-casein expression in CID-9 cells was not due to nonspecific effects. Moreover, we suggest that heterocellular gap junction communication is a critical element in mediating mammary epithelial differentiation, since 18α GA inhibited the β-casein (R.S.T. et al., unpublished) that is typically expressed in co-cultures of SCp2 and SCg6 cells on plastic in the absence of exogenously provided basement membrane(Desprez et al., 1993).

We confirmed that β-casein expression by CID-9 cells on EHS-drip was disrupted, in a dose-dependent manner (data not shown), by function-blocking anti-β₁-integrin antibody at concentrations previously shown to markedly affect β-casein expression(Streuli et al., 1995b; Muschler et al., 1999). cAMP-treated cells cultured on plastic and incubated with function-blocking anti-β₁-integrin antibody showed only a marginal decrease inβ-casein production, suggesting that enhanced GJIC can induce partial differentiation in the absence of cell-ECM interaction. Studies(Roskelley et al., 1994) have indicated the presence of two classes of signals that are generated by the ECM. The physical signals that involve cell rounding and clustering and are mimicked by plating the cells on PolyHEMA, and theβ ₁-integrin-mediated biochemical signals. CID-9 cells expressed β-casein on PolyHEMA only when supplemented with cAMP,suggesting that enhanced GJIC lead to β-casein expression on a non-adhesive substratum by substituting for theβ ₁-integrin-mediated biochemical pathway.

In conclusion, these studies demonstrated that cell-cell communication via gap junctions is essential for growth and differentiation of mammary epithelial cells in vitro. Moreover, for the first time, it is reported that cell-cell communication via gap junctions was able to initiate a partial differentiation process on a non-adhesive substratum and in the absence of an exogenous basement membrane. Finally, cell-ECM interaction alone was not sufficient to induce an optimal differentiation phenotype.

The authors are grateful to M. J. Bissell and F. Homeidan for their critical reading of the manuscript. R. Bajani, S. Bazzi, A. Ismail and T. Zeitouni are acknowledged for assisting in the preparation of the manuscript. This work is supported by the University Research Board (R.S.T. and M.E.S.);Medical Practice Plan, Diana Tamari Sabbagh Research Fund, Terry Fox Cancer Research Fund and Lebanese National Council for Scientific Research (M.E.S.);Third World Academy of Science and Lebanese National Council for Scientific Research (R.S.T.); and Contract DE-AC03-76F00098 from the United States Department of Energy, Office of Biological and Environmental Research, and by NIH Grant CA57621 (to M. J. Bissell).