Extracellular Matrix Selectively Modulates the Response of Mammary Epithelial Cells to Different Soluble Signaling Ligands*

In adherent cells, cell-substratum interactions are essential for the propagation of some growth factor signaling events. However, it has not been resolved to what extent different types of extracellular matrix regulate the signals elicited by different soluble ligands. Our previous work has shown that prolactin signaling in mammary epithelium requires a specific cell interaction with the basement membrane and does not occur in cells plated on collagen I. We have now investigated whether the proximal signaling pathways triggered by insulin, epidermal growth factor (EGF), and interferon-γ are differentially regulated in primary mammary epithelial cell cultures established on basement membrane and collagen I. Two distinct signaling pathways triggered by insulin exhibited a differential requirement for cell-matrix interactions. Activation of insulin receptor substrate (IRS) and phosphatidylinositol 3-kinase was restricted to cells contacting basement membrane, whereas the phosphorylation of Erk occurred equally in cells on both substrata. The amplitude and duration of insulin-triggered IRS-1 phosphorylation and its association with phosphatidylinositol 3-kinase were strongly enhanced by cell-basement membrane interactions. The mechanism for inhibition of IRS-1 phosphorylation in cells cultured on collagen I may in part be mediated by protein-tyrosine phosphatase activity since vanadate treatment somewhat alleviated this effect. In contrast to the results with insulin, cell adhesion to collagen I conferred greater response to EGF, leading to higher levels of tyrosine phosphorylation of the EGF receptor and Erk. The mechanism for increased EGF signaling in cells adhering to collagen I was partly through an increase in EGF receptor expression. The interferon-γ-activated tyrosine phosphorylation of Jak2 and Stat3 was independent of the extracellular matrix. It is well recognized that the cellular environment determines cell phenotype. We now suggest that this may occur through a selective modulation of growth factor signal transduction resulting from different cell-matrix interactions.

evidence shows that a collaboration between ECM and growth factors promotes cell proliferation, migration, survival, and differentiation. The mechanisms by which these different arrays of signals are integrated have not been completely delineated. Interactions with ECM can regulate a cell's ability to transduce signals from specific growth factors. For example, in comparison with cells kept in suspension, adhesion to an ECM promotes the activation of signaling proteins in response to soluble ligands such as EGF, platelet-derived growth factor, and insulin (2)(3)(4)(5)(6). In both developmental and pathological processes, including developmentally regulated migration events, wound healing, and neoplasia, cells can temporally encounter different types of ECM. This suggests that within different cellular microenvironments, they may become differentially responsive to signals from the same growth factors. However, few studies have addressed the question of to what extent the adhesion to different types of ECM might affect growth factor signaling.
In mammary gland, the secretory epithelium resides on basement membrane (BM), which is required for milk protein gene expression and cell survival (7). Mammary epithelial cells cultured on a reconstituted BM matrix isolated from Engelbreth-Holm-Swarm tumor assume a spherical structure similar to alveoli in vivo. These cells survive and acquire differentiated mammary functions in the presence of lactogenic hormones (8,9). In contrast, neither three-dimensional tissue architecture nor functional differentiation can be recapitulated when cells are in contact with thin collagen I matrices. Furthermore, these cells undergo apoptosis even in the presence of soluble survival factors. The mammary gland therefore provides a good model to study the biochemical mechanisms underlying the cross-talk between ECM and soluble factors.
Both laminin and ␤ 1 integrin have previously been shown to be required for lactogenic hormone-stimulated milk protein expression (10,11). One mechanism for this control of gene expression is through ECM-mediated modulation of prolactin signaling (12). Mammary cells cultured for several days on BM are responsive to prolactin, which leads to tyrosine phosphorylation of the prolactin receptor, Jak2, and Stat5, whereas cells plated on collagen I are refractory to prolactin (13). All the components for prolactin signaling are present in cells on collagen I, and we have therefore argued that one function for integrin-mediated interactions with BM is to organize the prolactin signaling components into a functionally active complex. Since mammary epithelial cells contact BM in vivo, but are not normally associated with collagen I, it is possible that mammary cell adhesion to BM is a prerequisite for the orchestration of all growth factor signaling events. Alternatively, BM might provide a degree of specificity for some growth factor signaling pathways, but not for others.
To discriminate between these two possibilities, we examined the proximal signaling events triggered by insulin, EGF, and interferon-␥ (IFN-␥) in mammary epithelial cells plated on either collagen I or BM. In contrast to other studies that assessed growth factor signaling in cells plated on ECM for very short periods of time (3,4,6,14), cell signaling in long-term cultures was investigated. We examined growth factor signaling in cells that had established stable cell-matrix interactions over several days, thus mimicking the type of interactions they would experience in vivo. In addition, we investigated the signaling responses of wild-type primary cell cultures, rather than those of reconstituted cell lines transfected with vectors overexpressing signaling components, as has largely been performed previously. Our results show that the cellular microenvironment selectively modulates the response of mammary cells to different soluble signaling ligands and that this occurs through different mechanisms.

EXPERIMENTAL PROCEDURES
Reagents-Murine EGF was from Harlan Sera-Lab (Leicestershire, United Kingdom). Bovine insulin and hydrocortisone were purchased from Sigma (Poole, UK). Murine IFN-␥ was the generous gift of Dr. A. Groves (Ludwig Institute for Cancer Research, London, UK). Monoclonal anti-phosphotyrosine antibody 4G10 and polyclonal antibodies to IRS-1, IRS-2, Jak2, and PI3K were obtained from Upstate Biotechnology, Inc. (Lake Placid, NY). Monoclonal antibody to EGFR and polyclonal antibody to IR were from Transduction Laboratories (Lexington, KY). Polyclonal antibodies to Erk (extracellular signal-regulated kinase), IR, and Stat3 were from Santa Cruz Biotechnology (Santa Cruz, CA). Polyclonal antibody to phospho-Erk (Thr 202 -Tyr 204 ) was obtained from New England Biolabs Inc.(Beverly, MA). Polyclonal antibody to ErbB2 was the generous gift of Professor W. G. Gullick (Imperial Cancer Research Fund, London). Calf intestine alkaline phosphatase was purchased from Roche Molecular Biochemicals.
Substrata and Cell Culture-Collagen I-coated dishes were prepared by incubating plates overnight at 4°C with rat tail collagen at 8 g/cm 2 . The plates were washed extensively with phosphate-buffered saline before use. In some experiments, collagen I-precoated dishes (Falcon, Oxford, UK) were used. Reconstituted basement membrane matrix was prepared from the Engelbreth-Holm-Swarm tumor (15) and coated onto dishes at 14 mg/ml as described previously (8,15).
All experiments were performed with first passage mammary epithelial cells derived from mid-pregnant ICR mice. Primary epithelial cultures were prepared from isolated mammary alveoli (15) and plated on different substrata in nutrient mixture F-12 (Sigma) containing 10% heat-inactivated fetal calf serum (Advanced Protein Products, Brierley Hill, UK), 1 mg/ml fetuin (Sigma), 5 ng/ml EGF, 1 g/ml hydrocortisone, and 5 g/ml insulin. Insulin was omitted in experiments in which insulin-stimulated signaling was examined. After 72 h, cells were serum-starved overnight in Dulbecco's modified Eagle's medium/nutrient mixture F-12 (Sigma) containing hydrocortisone and then stimulated with EGF (10 ng/ml), IFN-␥ (200 units/ml), or insulin (5 g/ml).
To determine the phosphorylation status of IRS-1, the IRS-1 immunoprecipitates were incubated with 50 mM Tris (pH 7.5) and 1 mM MgCl 2 for 10 min at 30°C, followed by addition of calf intestine alkaline phosphatase (40 units) for another 20 min. The enzyme reaction was stopped by addition of an equal volume of 2ϫ SDS sample buffer.

RESULTS
Growth Factor-activated Tyrosine Phosphorylation Is Differentially Regulated by ECM-We have previously shown that prolactin-activated signal transduction in mammary epithelial cells is dependent on cell-BM interactions, but does not occur when cells are plated on tissue culture plastic or dishes coated with collagen I (13). To examine whether this reflected a general response of mammary epithelium to growth factors, cells cultured on collagen I and BM were stimulated by various growth factors, and protein tyrosine phosphorylation was monitored by Western blot analysis. All the experiments described here were performed with primary cultures of mammary epithelial cells established on either BM or collagen for 3-4 days.
Numerous proteins were tyrosine-phosphorylated in the absence of growth factors, but the gross extent of tyrosine phosphorylation was comparable in cells cultured on both substrata ( Fig. 1, lanes 1 and 4). However, we noted a dramatic difference in the tyrosine phosphorylation protein profile between cells cultured on BM and collagen I in response to 15-min treatments with insulin or EGF. Insulin induced the tyrosine phosphorylation of an ϳ170-kDa protein in cells cultured on BM, but to a lesser degree in cells cultured on collagen I (Fig. 1, compare lanes 5 and 2). However, tyrosine phosphorylation of an ϳ95-kDa protein in response to insulin was evident in cells on both substrata. Similar results were obtained for insulinlike growth factor I and II treatment (data not shown). In contrast, EGF triggered substantial tyrosine phosphorylation of an ϳ185-kDa protein in cells contacting collagen I, but very low levels of phosphorylation of this protein were observed in cells cultured on BM (Fig. 1, compare lanes 3 and 6).
Tyrosine-phosphorylated proteins are often the regulatory proteins involved in membrane-proximal signaling events. Our results reveal that cell-ECM interactions selectively modulate growth factor-activated tyrosine phosphorylation. To understand in more detail how ECM regulates growth factor signaling, we examined its effect on insulin-, EGF-, and IFN-␥-triggered signaling pathways.
Inhibition of Insulin Signaling by Collagen I Occurs Down-FIG. 1. Insulin-and EGF-stimulated tyrosine phosphorylation is differentially regulated by ECM. In all experiments described here, primary mouse mammary epithelial cells were isolated from midpregnant mice and cultured on collagen I (CI) or BM substratum. Cells were incubated with medium alone (Ϫ), 5 g/ml insulin, or 10 ng/ml EGF for 15 min. Total cell lysates were separated by 6% SDS-PAGE, and protein tyrosine phosphorylation was detected by immunoblotting with anti-phosphotyrosine antibody (PY). Note that insulin triggered the tyrosine phosphorylation of a 170-kDa protein in cells cultured on BM (*), whereas a 95-kDa phosphoprotein was detected in cells cultured on both collagen I and BM (E). In contrast, EGF triggered the tyrosine phosphorylation of a 185-kDa protein exclusively in cells cultured on the collagen I substratum (**).

stream of the Insulin Receptor through a Vanadate-sensitive
Mechanism-Insulin is a pleiotropic growth factor with multiple functions in mammary gland development, including mammogenesis and lactogenesis. Moreover, in cultured mammary epithelial cells, insulin along with BM is essential for milk protein gene expression and also provides survival signals to prevent default apoptosis (16). 2 Here we addressed the possibility that ECM influenced insulin signaling. Insulin binding to its receptor triggers tyrosine phosphorylation of the receptor, thereby promoting the association and tyrosine phosphorylation of a number of proteins, including IRS (17). Phosphorylated IRS provides docking sites for various signaling molecules, such as PI3K, SHP-2, Nck, Grb2, and Fyn (17). In this study, we examined some of the proximal insulin signaling events involving activation of IR, two isoforms of IRS (i.e. IRS-1 and IRS-2), and also the interaction of PI3K with IRS.
Incubation with insulin for 30 min resulted in tyrosine phosphorylation of IR, IRS-1, and IRS-2 in cells cultured on BM (Fig. 2), the first two corresponding to the ϳ95and ϳ170-kDa proteins, respectively, in Fig. 1 (lane 5). Furthermore, insulin induced the association of PI3K with IRS proteins in these cells (Fig. 2, B and C, lane 4). The phosphorylation of IRS-1 and IRS-2 and their association with PI3K in response to insulin were impaired in cells cultured on collagen I, whereas the phosphorylation of IR was unaffected (Fig. 2). These results show that insulin is able to trigger tyrosine phosphorylation of its receptor independently of cell-ECM interactions, but activation of signaling pathways downstream of IR is greater when cells are in contact with BM.
Although fairly low levels of tyrosine phosphorylation of IRS-1 occurred in cells cultured on collagen I, we noticed that insulin stimulation led to a decrease in the mobility of IRS-1 in cells on both substrata (Fig. 2B, lanes 1 and 3 versus lanes 2 and 4). To determine whether this change in mobility was due to serine/threonine phosphorylation, calf intestine alkaline phosphatase was added to IRS-1 immunoprecipitates, and the mobility of the protein was monitored by immunoblotting with anti-IRS-1 antibody (Fig. 3). Alkaline phosphatase treatment increased the mobility of IRS-1 isolated from cells on both substrata, suggesting that IRS-1 underwent serine/threonine phosphorylation upon insulin stimulation. Thus, insulin-induced serine/threonine phosphorylation of IRS-1 is not influenced by cell-ECM interactions, whereas efficient tyrosine phosphorylation of IRS-1 requires cell adhesion to BM.
We have previously demonstrated that prolactin-induced tyrosine phosphorylation is inhibited in cells cultured on collagen I by a mechanism involving protein-tyrosine phosphatase (PTP) (13). To determine whether PTP also plays a role in blocking insulin-activated tyrosine phosphorylation of IRS-1, mammary cells were pretreated with vanadate to inhibit PTP prior to addition of insulin. Combined treatment with insulin and vanadate partially restored tyrosine phosphorylation of IRS-1 in cells cultured on collagen I, whereas vanadate alone had no effect (Fig. 4, lanes 1-4). In contrast, insulin triggered a greater extent of tyrosine phosphorylation of IRS-1 in cells adhering to BM, and this was not further increased by addition of 100 M vanadate (Fig. 4, lanes 7 and 8). Thus, following the treatment of cells cultured on collagen I with vanadate, insulin is able to trigger tyrosine phosphorylation of IRS-1.
Taken together, our results demonstrate that the tyrosine phosphorylation of IRS-1 and IRS-2 and the association of IRS with PI3K only occur efficiently in mammary cells cultured on BM. Inhibition of insulin signaling by culture on collagen I occurs downstream of IR, and PTP may partially contribute to the suppression of signaling.
Insulin-activated Signaling Is More Sustained in Response to BM than Collagen I-Since insulin signaling is obligatory for milk protein gene expression and cell survival, and both of these events last for a long period of time in vivo, we investi- In cells on both substrata, prolonged insulin treatment resulted in a slight decrease in tyrosine phosphorylation of IR in comparison with that observed after 1 h (Fig. 5A, compare lanes 3, 4, 7, and 8 with lanes 2 and 6). This was due in part to lower levels of IR being present after 1 or 2 days of insulin treatment. The levels of tyrosine phosphorylation of IR declined after 1 day of insulin treatment, but in neither case, on collagen I or on BM, were they completely diminished. In contrast to IR, tyrosine phosphorylation of IRS-1 occurred predominantly in cells contacting BM, regardless of the time that the cells were treated with insulin, and was barely visible in cells on collagen I (Fig. 5B). As the exposure to insulin continued, tyrosine phosphorylation of IRS-1 in the BM cultures was only reduced very slightly. Similar results were obtained for the association of PI3K with IRS-1 (Fig. 5B). These results are consistent with our observation for short-term (i.e. 30 min) insulin stimulation (Fig. 2), confirming that insulin-triggered tyrosine phosphorylation of IRS-1 and its association with PI3K occur efficiently only when mammary cells are in contact with BM. Our data also demonstrate that insulin signaling is sustained over long periods of time. Interestingly, this is in contrast with some other signaling pathways where, for example, the prolactin-activated phosphorylation of Jak2/Stat5 diminishes to near background levels after 1 h of stimulation (13).
To determine whether sustained activation of insulin signaling in cells cultured on BM requires the continuous presence of insulin, a "pulse-chase" experiment was performed. Primary mammary epithelial cells isolated from mid-pregnant mice were plated on either collagen I or BM in the presence of insulin for 2.5 days. The insulin-containing medium was then removed and replaced with insulin-free medium. Cells were harvested at varying time periods after the removal of insulin, and the phosphorylation of IRS-1 and its association with PI3K were assessed. A dramatic difference in the tyrosine phosphorylation of IRS-1 and the amount of PI3K associated with IRS-1 was observed after incubation with insulin for 2 days, with a much greater extent of activation in cells adhering to BM (Fig. 6, lanes 1 and 5). Following insulin removal, the low levels of phosphorylated IRS-1 and its associated PI3K in cells cultured on collagen I were further attenuated (Fig. 6, lanes 1-4). However, the signal remained high in cells contacting BM even after removal of the ligand (Fig. 6, lanes 5-8). The mechanisms underlying the sustained phosphorylation of IRS-1 in cells cultured on BM are not clear. Whether this is due to an autocrine effect through induction of insulin-like growth factor or to sequestration of insulin within BM requires further investiga-tion. Together, these experiments show that ECM affects not only the amplitude, but also the duration of insulin signaling in mammary epithelial cells.
Insulin-triggered Erk Phosphorylation Occurs in Cells Cultured on both Collagen I and BM-Although tyrosine phosphorylation of IRS-1 by insulin was inhibited in cells cultured on collagen I, IR remained active ( Fig. 2A). In addition to the IRS protein family, Shc and Grb10 are also recruited to active IR (17,18). We hypothesized that cells on collagen I might be competent for induction of some other signaling pathways in Insulin-containing medium was then removed and replaced with fresh medium without insulin. Cells were harvested at the indicated times (days (d)) after the removal of insulin, and total cell lysates were immunoprecipitated (IP) with anti-IRS-1 antibody, followed by immunoblotting with anti-phosphotyrosine antibody (PY). The blot was stripped and reprobed with anti-IRS-1 and then anti-PI3K antibody. response to insulin, and we therefore examined the phosphorylation of Erk.
In contrast to our results with PI3K, insulin triggered the phosphorylation of both Erk1 and Erk2 to a similar extent in mammary cells adhering to either collagen I or BM (Fig. 7). Thus, although insulin does not activate IRS-1 and the downstream effector, PI3K, in cells on collagen I, they are not completely devoid of insulin signaling.
Taken together, our results indicate that different insulintriggered signaling pathways exhibit a differential requirement of cell-ECM interactions. The phosphorylation of Erk is independent of the ECM on which cells reside, whereas the activation of IRS/PI3K is restricted to cells contacting BM.

Epidermal Growth Factor Receptor Expression and Erk Phosphorylation Are Enhanced in Cells Cultured on Collagen
I-EGF is a potent mitogen for mammary gland, promoting ductal growth and lobulo-alveolar development. However, culture studies have shown that EGF exhibits a detrimental effect on the functional differentiation of mammary cells by antagonizing prolactin-driven milk protein gene expression (19). We therefore investigated whether EGF signaling in mammary epithelial cells was dependent on ECM, starting with an examination of tyrosine phosphorylation of the receptor. EGF induced a greater extent of tyrosine phosphorylation of EGFR in cells on collagen I than in cells on BM (Fig. 8A). This confirmed our previous result (Fig. 1, lanes 3 and 6) and extended it by showing that the 185-kDa protein was probably EGFR. However, in this case, we found that the increase in EGFR tyrosine phosphorylation was due to higher levels of EGFR expression in cells on collagen I (Fig. 8A). This contrasts with the ECM control of insulin signaling, where the levels of all its components remain comparable on both substrata. EGF signaling is complicated by the presence of four ErbB family receptors, EGFR (ErbB1), ErbB2, ErbB3, and ErbB4 (20). EGF induces the formation of EGFR/EGFR homodimers and EGFR/ErbB2 heterodimers (20). Since ErbB2 is also a receptor for EGF, its phosphorylation in response to EGF was also assessed. In the absence of ligand, very low levels of ErbB2 tyrosine phosphorylation were detected in cells on both substrata, but they were somewhat more pronounced when cells contacted BM (Fig. 8B, lanes 1 and 3). In contrast to the results for EGFR, exposure to EGF enhanced tyrosine phosphorylation of ErbB2 in mammary cells cultured on both substrata; indeed, it was more pronounced in cells on BM (Fig. 8B, lanes 2 and 4). However, ErbB2 expression levels were not significantly altered by cell-ECM interactions (Fig. 8B).
To determine whether ECM affected subsequent events in the EGF signaling pathway, the phosphorylation status of Erk1 and Erk2, downstream effectors of EGFR, was examined. The EGF-induced tyrosine phosphorylation of Erk1 and Erk2 was more predominant in cells cultured on collagen I, reflecting strongly phosphorylated EGFR (Fig. 8C). However, we also noted that Erk1 and Erk2 were phosphorylated, but to a lesser degree, in cells adhering to BM, most likely reflecting a signal derived from ErbB2 phosphorylation. Our experiments suggest that cell adhesion to collagen I enhances EGFR expression, leading to a corresponding increase in tyrosine phosphorylation of EGFR and the subsequent triggering of Erk phosphorylation.
Thus, our study demonstrates that under the influence of ECM, mammary cells interpret signals from EGF differently than those from insulin. For EGF, this occurs at the level of receptor control. In contrast, insulin-induced IRS phosphorylation and its association with PI3K are controlled partially by a mechanism involving PTP.
IFN-␥-stimulated Tyrosine Phosphorylation of Jak2 and Stat3 Is Not Affected by ECM-IFN-␥ is not known to be involved in normal mammary gland development, but it may have some impact on the tissue in pathological states through its immunomodulatory, anti-proliferative, and pro-apoptotic effects (21,22). We wished to determine whether or not the signaling pathway triggered by a cytokine not normally involved with mammary gland physiology was also affected by cell-ECM interactions. IFN-␥ activates tyrosine phosphorylation of Jak2 and Stat3 in an analogous way to the activation of Jak2 and Stat5 by prolactin. Since we had previously demonstrated an ECM dependence of prolactin signaling, we compared the effect of cell-ECM interactions on IFN-␥ signaling with that on prolactin signaling. In agreement with our previous work, prolactin activated the tyrosine phosphorylation of Jak2 exclusively in cells cultured on BM (Fig. 9A, lanes 2 and 5). However, mammary cells on either collagen I or BM were equally competent for induction of Jak2 tyrosine phosphorylation in response to IFN-␥ (Fig. 9A,  lanes 3 and 6). Similar results were obtained for Stat3 (Fig.  9B), where comparable levels of tyrosine phosphorylation of Stat3 were induced by IFN-␥ in cells cultured on both substrata.
These results show that although tyrosine phosphorylation of Jak2 by prolactin is dependent on cell-BM interactions, ECM has no effect on the phosphorylation of Jak2 and Stat3 triggered by IFN-␥. Thus, regulation of the Jak-Stat pathway by ECM is growth factor-specific. It is possible that the signaling components of these two similar cytokine pathways are compartmentalized differently in response to the cell's ECM environment. DISCUSSION Our study demonstrates that in the primary culture of untransfected adherent mammary epithelial cells, cell-ECM interactions determine the outcome of growth factor signaling. No longer can it be considered that the binding of a particular ligand to its receptor will necessarily trigger a specific intracellular signaling pathway. Here we extend our previous work demonstrating that prolactin signaling depends on cell-BM interactions (12,13) and show that mammary cell adhesion to ECM also influences insulin and EGF signaling and that this occurs by distinct mechanisms. We also show that ECM does not regulate all growth factor signaling pathways since IFN-␥ does not discriminate between BM and collagen I for activation of Jak2 and Stat3.
Insulin Signaling and Cell-Matrix Interactions-The crosstalk between integrin-and insulin-triggered signaling has previously been documented. Insulin regulates the tyrosine phos-phorylation of focal adhesion kinase (pp125 FAK ) as well as cell adhesion (6,23). Conversely, activation of pp125 FAK results in tyrosine phosphorylation of IRS-1 even in the absence of insulin (24). Several lines of evidence from other cell systems have also demonstrated the importance of cell adhesion for insulin signaling. First, insulin stimulates the association of IR and IRS-1 with ␣ v ␤ 3 integrin in Rat-1 and NIH 3T3 cell lines (2,5). Second, adhesion to fibronectin potentiates insulin-stimulated tyrosine phosphorylation of IR and IRS-1, the association of PI3K with IRS-1, and protein kinase B activation in adipocytes and Chinese hamster ovary cells overexpressing IR (6,25). Moreover, in our study, a reconstituted BM matrix conferred a greater response to insulin signaling in primary mouse mammary epithelial cells. All these findings underscore the importance of cell-ECM interactions in insulin signaling. We have yet to determine the role of integrins in the response of mammary cells to insulin. However, the major integrins expressed in mammary cells are ␣ 2 ␤ 1 , ␣ 3 ␤ 1 , and ␣ 6 ␤ 1 (26). The ␤ 3 integrin subunit is barely detectable, correlating with the poor adhesion of primary mammary epithelial cells to vitronectin. Thus, although a specific association of IR with ␤ 3 integrin has been demonstrated in some cell types (2,5), ␤ 1 integrin is more likely to be involved in the cross-talk with insulin signaling in mammary cells.
Regulation of growth factor signaling by cell adhesion has been shown to occur at different levels of signal propagation (1). Although other reports show that that signaling initiated by cell-ECM interactions intervenes with the insulin signaling pathway at the level of IR (2,5,6,25), our data suggest that these two signaling pathways converge at the level of IRS-1 since no discernible differences in activation of IR were detected in cells cultured on collagen I and BM. These contradictory data may reflect the different cell types used in these studies. Alternatively, they may indicate that intracellular signaling pathways are regulated differently in cells established on ECM for several days to those in cells plated on ECM for short times. We suggest that in mammary epithelial cells, there is a matrix-specific restriction point in signal transduction at the level of IRS tyrosine phosphorylation leading to a control on its subsequent downstream signaling.
Models for the Collagen I-dependent Inhibition of Insulin Signaling-There are several possible explanations for the ECMdependent inhibition of IRS signaling. One is that the interaction between IR and IRS-1 may be hindered, either by inhibitory molecules or by incorrect subcellular compartmentalization. Alternatively, IRS-1 may be subject to dephosphorylation by PTP. Both situations would result in high levels of tyrosine phosphorylation of IR, but low levels of IRS activation.
IRS protein contains an amino-terminal pleckstrin homology domain followed by a phosphotyrosine-binding domain. The phosphotyrosine-binding domain binds to the phosphorylated NPEY motif in the juxtamembrane region of IR (17). The pleckstrin homology domain of IRS is essential for its tyrosine phosphorylation after insulin treatment. Thus, both pleckstrin homology and phosphotyrosine-binding domains are critical for coupling IRS-1 with IR, leading to proper signal relay. Several proteins have been shown to bind to these sequences and to interrupt insulin signaling. 14-3-3 protein and Rho-associated protein kinase-␣ bind to the phosphotyrosine-binding domain of IRS-1 (27,28), whereas nucleolin binds to the pleckstrin homology domain (29). Interestingly, overexpression of nucleolin results in a reduction in tyrosine phosphorylation of IRS-1, but exerts no effects on IR and Shc (29). These observations led to a model whereby accurate interactions between IR and IRS are required for the propagation of insulin signaling, and disruption of this association causes abortion of signaling down- stream of IR. We are therefore exploring whether a similar mechanism is involved in the differential response to insulin when mammary cells adhere to different ECMs.
An important aspect in the regulation of cell signaling is compartmentalization of signaling molecules. Cell fractionation experiments show that IR mainly localizes to the plasma membrane, whereas IRS-1 is restricted to an intracellular membrane compartment that may associate with the cytoskeleton (30,31). Compartmentalization is likely to be critical for insulin signaling since okadaic acid treatment results in the movement of IRS-1 from the intracellular membrane compartment to the cytosol, a location where IRS-1 may be in close proximity to PTP or inaccessible to tyrosine kinases (30). Such a relocalization of IRS-1 may explain the observation that okadaic acid inhibits tyrosine phosphorylation of IRS-1, but has no effect on IR (32). Thus, the accurate subcellular localization of IRS-1 facilitates the propagation of signals downstream of IR and may therefore be an important determinant of ECM regulation.
PTPs are alternative candidates for the inhibition of signaling downstream of IR in mammary cells cultured on collagen I. This possibility is supported by our observation that vanadate partially eliminates the suppression of insulin-stimulated IRS-1 tyrosine phosphorylation. Several PTPs have been shown to down-regulate insulin signaling, including PTP1B, LAR, SHP-1, and RPTP␣ (33,34). Although these phosphatases associate with IR and inhibit its tyrosine phosphorylation and subsequent signaling, they may not be involved in the inhibition of insulin signaling seen in our study since IR phosphorylation was not affected by culture on collagen I. To our knowledge, the only PTP that targets to IRS is SHP-2. SHP-2 has previously been considered to have a positive role in insulin-stimulated Erk activation and DNA synthesis. However, interaction of SHP-2 with IRS-1 has recently been shown to reduce tyrosine phosphorylation of IRS-1 and its associated PI3K activity (35). We are therefore examining whether SHP-2 has a role in the ECM control of insulin signaling.
Of particular interest is the possibility that one of these PTPs or a novel PTP is regulated by cell-ECM interactions. Although several PTPs have been shown to regulate cell adhesion (36 -40), the effect of cell adhesion on the expression, activation, or translocation of PTP is not well understood. LAR is a receptor for laminin, but it has not yet been determined if its phosphatase activity changes upon cell adhesion (41). A number of examples of cell adhesion-induced PTP translocation between different cellular compartments have been reported (14,42,43). One interesting observation is that SHP-2 targets to the transmembrane glycoproteins SHPS-1/SIRP when cells are plated onto fibronectin or laminin (44). The association of SHP-1 and SHP-2 with SHPS-1/SIRP also occurs when cells are stimulated with various growth factors, which in turn affects growth factor signaling (45,46). Thus, translocation of PTP in response to cell adhesion has a critical role in controlling growth factor signaling and may contribute to the mechanism for altered insulin signaling in mammary cells adherent to collagen I.
We have previously shown a similar control of prolactin signaling, where ligand-dependent phosphorylation of the prolactin receptor occurs in mammary cells on BM, but is induced in collagen I cultures only after vanadate treatment (13). Thus, PTP may represent an important effector for the control of signal transduction in the long-term response of mammary cells to ECM. However, the level of intervention may vary according to ligand, affecting either receptor or post-receptor signaling events.
Biological Consequences of ECM Control of Insulin Signaling in Mammary Epithelium-In this study, we provide a link between insulin signaling and the subsequent biological responses under the influence of cell-ECM interactions. Insulin is essential for milk protein gene expression and cell survival in the mammary gland. One possible mediator for insulin action is PI3K since inhibition of PI3K by LY294002 or wortmannin results in both the abolition of ␤-casein expression and cell death (16). 2 Our data show that the activation of IRS/PI3K in response to insulin occurs efficiently only when cells are in contact with BM. Both the amplitude and duration of the response are particularly marked in long-term cultures. Thus, one implication of our findings is that cells adhering to BM may be endowed with the ability to respond to differentiation and survival signals from insulin. In the absence of this type of cell-ECM interaction, for example, when mammary cells are in contact with collagen I, signal transduction is not efficient and results in repressed differentiation and survival. EGF Signaling and Cell-Matrix Interactions-Cell-ECM interactions also have a great impact on EGF signaling. The interaction of fibroblasts with fibronectin-coated beads results in localization of EGFR to focal contact regions, and this is accompanied by elevated tyrosine phosphorylation of the receptor (47). In vascular smooth muscle cells, the engagement of ␣ v ␤ 3 integrin by tenascin C promotes EGF signaling (48). Glomerular epithelial cells adherent to collagen I exhibit higher levels of ligand-induced EGFR phosphorylation than those plated on plastic (49). However, other studies have shown that cell adhesion also affects EGF signaling, but that this occurs downstream of its receptor (3,4). We document here that primary mouse mammary epithelial cells cultured on collagen I show higher levels of EGFR and Erk phosphorylation than those plated on BM, thus agreeing with previous models in which EGF signaling is regulated by cell-ECM interactions. However, we find that this is mainly due to up-regulation of EGFR expression in cells on collagen I, which is distinct from other reports in that EGFR expression levels do not vary with ECM (47)(48)(49). We have not yet elucidated the mechanism for altered EGFR levels. One possibility is that as the cells are plated for several days before being stimulated with EGF, cell interactions with other collagen receptors, for example, discoidin domain receptors, might regulate EGFR expression (50,51).
Since mammary epithelial cells do not directly contact stromal matrix in vivo, but instead interact with a BM, the cultures on collagen I may be considered to be similar to wound healing. Wound healing is a complex process and includes cell spreading, migration, and proliferation. EGF is a potent mitogen and also promotes cell migration. Application of EGF to wounded skin has been shown to stimulate the re-epithelialization of wounds, leading to resurfacing of skin (52). Indeed, elevation of EGFR was detected during the process of epidermal wound repair, and this preceded epidermal hypertrophy (53). In mammary cells cultured on collagen I, EGFR up-regulation may therefore reflect other more generic changes that occur on wound healing (54). We (26,55,56) and others (57) have shown previously that expression of transforming growth factors ␣ and ␤, integrins, and ECM proteins is augmented in mammary cells cultured on collagen I or plastic. Thus, the expression of a repertoire of growth factors and ECM proteins and their receptors becomes elevated in response to culture on stromal matrix, equivalent to that encountered in a wound healing situation.
An important corollary of these observations is that elevated EGFR expression in cells exposed to collagen I may increase their sensitivity to proliferation signals. This may have significant implications for understanding the increased epithelial proliferation observed in neoplasia since the cells of late-stage breast carcinomas no longer interact with BM, but rather are embedded within the stromal matrix of the breast. The possibility that increased expression of proliferation receptors in breast cancer may occur through epigenetic mechanisms such as an altered ECM environment should therefore be seriously considered.
It is well known that dynamic interplay between cells and their microenvironment determines cell behavior. Our research has now brought insight into some of the mechanisms whereby ECM modulates the response of cells to different signaling ligands. In considering how growth factors regulate cell behavior, we now emphasize the importance of appreciating that cell-matrix interactions influence the outcome of specific ligand-receptor interactions and thereby profoundly affect cellular phenotype.