R-Ras Signals through Specific Integrin α Cytoplasmic Domains to Promote Migration and Invasion of Breast Epithelial Cells

T47D cells (American Type Culture Collection) were transfected with constitutively active constructs for R-Ras(87L) or (38V), TC21(72L), N-Ras(12D), or K-Ras(12V), or with dominant negative constructs for R-Ras(41A) or (43N), or TC21(26A) that were expressed in pZIP. Cells were selected in G418 as described (Keely et al., 1995) and expanded as pools of stably transfected cells. Control cells were transfected with pZIP alone and selected in parallel with the other cells. Expression of constructs was determined by immunoblotting. Because these different molecules could not be probed by the same antibody on the same blot, it was impossible to determine absolute expression levels of R-Ras compared with TC21, N-Ras, or K-Ras. However, a comparison of each molecule to endogenous levels was made: R-Ras(87L) was 8.8-fold above endogenous R-Ras (or 8.8×), R-Ras(43N) = 10.7×, R-Ras(41A) = 1.8×, TC21(72L) = 4.0×, TC21(26A) = 1.0×, N-Ras(12D) = 0.9×, and K-Ras(12V) = 2.1×. For double transfectants to create X4 chimeric cell lines, cells were cotransfected with X4 chimeras expressed in pSFneo and pCGN (20:1), and selected in hygromycin-containing growth medium as described (Keely et al., 1997). Expression of α4 integrin on the surface of the cell, indicating expression of the X4 chimeras, was determined by flow cytometry (see below). X4 chimeras were a gift of Dr. Martin Hemler (Dana-Farber Cancer Institute).

Migration and invasion across transwells were performed as described previously (Santoro et al., 1994; Keely et al., 1995, 1997). Transwells were coated from the underside with 6 μg/ml collagen I or 25 μg/ml fibronectin (Collaborative Biomedical Products). For studies with X4 chimeras, transwells were coated with specific α4 ligands: 6 μg/ml of the 40-kD chymotryptic heparin binding fragment of fibronectin (GIBCO BRL) (McCarthy et al., 1986) or 50 μg/ml of the CS-1 peptide (EILDVPST; Peninsula Labs Inc.) (Humphries et al., 1987). CS-1 peptide was coupled to ovalbumin as described (Haugen et al., 1990) and the best coating concentration for cell migration was determined. Cell migration was performed in the absence of serum or growth factors for 16 h. For inhibitor assays, cells were pretreated for 15 min with DMSO (control), Wortmannin (30 nM), LY294002 (25 μM), bisindolylmalemide (1 μM) all from Calbiochem-Novabiochem Corp., or PD98059 (25 μM; Alexis Biochemicals), and allowed to migrate across collagen in the continued presence of the inhibitor for 5 h.

For antibody inhibition assays, cells were preincubated with control IgG or anti–α2 integrin antibody, P1E6 (GIBCO BRL) at a final concentration of 1:1,000. Migration was performed in the continued presence of the antibody for 16 h. Invasion was assessed on serum starved cells that were seeded onto the top of 2-mm collagen gels and allowed to migrate in response to 10% serum placed in the lower chamber. Invasion assays were allowed to proceed for 26–36 h. Unless otherwise noted, values presented are the mean ± SEM of at least three experiments normalized to control values. Data were analyzed by performing ANOVA using Microsoft Excel software. Morphogenesis in three-dimensional collagen gels was assessed after 7 and 16 d as described (Santoro et al., 1994; Keely et al., 1995).

Adhesion to Immobilon plates (Dynatech) coated with various concentrations of collagen I or fibronectin was performed and adherent cells quantitated by hexosaminidase activity as described (Haugen et al., 1990; Keely et al., 1995). Values given represent the average of two separate experiments, each of triplicate determinations, ± SEM.

Cells were detached in versene, washed in PBS containing 5 mg/ml BSA, and incubated with primary antibody on ice. To determine cell surface integrin expression, cells were incubated with the following mAbs: P4C2 against the α4, P1D5 against the α5, P1E6 against the α2, or P1B5 against the α3 subunits (all from GIBCO BRL), or TS2-7 against the α1, or 4F10 against the α6 subunits (Serotec). Primary antibody was diluted 1:1000. After a 20-min incubation, cells were washed once in PBS/BSA, and incubated with FITC-conjugated donkey anti–mouse secondary antibody at 1:100 (Jackson Labs) for 20 min on ice. Cells were washed again and analyzed on a Beckman FACScan^®. For conformation studies, cells were either treated with EDTA or with manganese at 2 mM before incubation with 12G10 antibody (Serotec), which detects conformational changes in the β1 integrin subunit. After this primary antibody incubation, analysis proceeded as above.

Cells were pretreated with DMSO or bisindolylmaleimide (1 μM) for 20 min, lysed in buffer containing 0.05% Triton-X 100, and the in vitro activity of PKC was determined on crude lysates using the PKC assay system (SignaTECT; Promega Corp.), according to manufacturer's directions. In brief, a single purification step over a DEAE cellulose column was used to generate a crude PKC-containing fraction. Background activity (nonspecific) was determined in the absence of phosphatidylserine and diacylglycerol, and compared with PKC activity determined in the presence of phosphatidylserine and diacylglycerol.

Cells expressing R-Ras(87L), R-Ras(38V), TC21(72L), K-Ras(12V), N-Ras(12D), or control vectors were cultured overnight either in the presence or absence of serum, treated for 20 min with DMSO (control) or PD98059 (25 μM), and lysed. Lysates were electrophoresed, immunoblotted with anti-p42/44 ERK antibody (Santa Cruz Biotech), and the activation of MAPK was determined by gel-shift (Cox et al., 1994). Determination of the mean image density was made using NIH image software.

Well differentiated human breast carcinoma T47D cells maintain the ability to polarize into tubule and ductlike structures when cultured in three-dimensional collagen gels (Keely et al., 1995). This differentiation is dependent on the α2β1 integrin (Keely et al., 1995; Zutter et al., 1995). One of the characteristics of T47D differentiation into tubules is the multicellular nature of the structure, such that individual cell–cell boundaries become more difficult to distinguish, as can be noted in Fig. 1 A. Stable expression of activated R-Ras(87L) in T47D cells disrupted the ability of these cells to differentiate and polarize in collagen gel culture, instead the cells grew as disorganized clumps and sheets of individual, nonpolarized cells (Fig. 1 B). Similar results were obtained with cells stably expressing another activated R-Ras isoform, R-Ras(38V) (not shown). Thus, activation of R-Ras results in a less differentiated, more transformed phenotype. In contrast, stable expression of dominant negative R-Ras(43N) enhanced differentiation; tubule structures developed earlier and were more extensive (Fig. 1 D). This suggests that endogenous R-Ras may play a negative role in regulating the rate or extent of normal epithelial differentiation. Like R-Ras, expression of constitutively active TC21(72L) disrupted differentiation in collagen gels (Fig. 1 C), whereas expression of dominant negative TC21(26A) enhanced differentiation in a qualitative manner (Fig. 1 E). This suggested that TC21, like R-Ras, is a negative regulator of breast cell differentiation.

Breast carcinoma progression is accompanied by the loss of polarization and differentiation, and the acquisition of a migratory invasive phenotype. Migration was determined using a Boyden chamber haptotactic assay in which transwells were coated on the underside with collagen. Expression of activated R-Ras isoforms, R-Ras(87L) or R-Ras(38V), significantly enhanced cell migration relative to control cells (Fig. 2 a), indicating that R-Ras activation promotes a more migratory phenotype. Dominant negative R-Ras(41A) inhibited migration by ∼45–50% (Fig. 2 a), suggesting a contribution of R-Ras to basal migratory mechanisms. Activated TC21(72L) also enhanced cell migration, albeit to a somewhat lesser extent than R-Ras (Fig. 2 a). Expression of activated isoforms of other Ras family members, N-Ras(12D) and K-Ras(12V), also enhanced cell migration across collagen, but to a lesser extent than R-Ras (Fig. 2 a).

The migration induced by R-Ras(87L) across collagen-coated filters was dependent on the α2β1 integrin, since it could be completely inhibited by anti-α2β1 blocking antibodies (Fig. 2 b). Similarly, TC21(72L)-induced cell migration was also α2β1 integrin-dependent (Fig. 2 b). These results are consistent with the finding that the α2β1 integrin is a major collagen receptor in breast epithelial cells that mediates many of their responses to collagenous matrices (Keely et al., 1995; Zutter et al., 1995).

To examine whether the effect of R-Ras was specific for collagen, cell migration across fibronectin was determined. Surprisingly, expression of R-Ras(87L) or R-Ras(38V) did not enhance and, in fact, partially inhibited migration across fibronectin by ∼40% (Fig. 2 c). R-Ras(87L) did not enhance or inhibit migration of cells across laminin (not shown). Similar results were obtained for TC21(72L), which also inhibited migration across fibronectin by 25%, although this difference was not statistically significant (Fig. 2 c). The difference between migration across collagen and fibronectin is not due to dramatically different levels of migration, since similar numbers of control cells migrate across each substratum, as shown by a representative experiment in Table I. In contrast to the results with R-Ras and TC21, expression of either N-Ras(12D) or K-Ras(12V) enhanced migration across fibronectin (Fig. 2 c). Thus, R-Ras and TC21 differ from N- and K-Ras in the specificity of substrata on which they enhance cell migration.

To metastasize, cells must not only migrate but must also invade through extracellular matrices. Therefore, we examined the ability of transfected cells to invade through a 2-mm-thick collagen gel in response to a serum gradient. Consistent with their more migratory phenotype, cells expressing activated R-Ras(87L) or R-Ras(38V) were significantly more invasive through collagen than control cells (Fig. 3). Similarly, activated TC21(72L), N-Ras(12D), and K-Ras(12V) induced cell invasion (Fig. 3). Because basal invasion levels of T47D cells are already so low, we did not further investigate the effect of dominant negative R-Ras or TC21 on invasion. Therefore, we find that expression of activated R-Ras or TC21 converts cells from a polarized, differentiated phenotype into a migratory and invasive phenotype.

Next, we examined the effect of R-Ras on adhesion of cells to collagen and fibronectin since R-Ras activation specifically enhanced cell migration across collagen, but decreased migration across fibronectin. This was of particular interest since it has been shown that expression of activated R-Ras(38V) increases the adhesion of myeloid cells to fibronectin and vitronectin through activation of integrin subunits (Zhang et al., 1996). Similar to these results, we found that expression of activated R-Ras(87L) enhanced adhesion to collagen by a modest but consistent amount (Fig. 4 a), demonstrating an effect of R-Ras activation on the avidity of cells for collagen. Dominant negative R-Ras(41A) decreased adhesion to collagen (Fig. 4 a) by a similar amount, demonstrating a contribution of endogenous R-Ras to cell adhesion. The other dominant negative isoform of R-Ras, R-Ras(43N), did not affect cell adhesion (not shown). Expression of TC21 isoforms had no consistent or significant effect on adhesion of cells to collagen (not shown). In contrast to the effect on cell adhesion to collagen, expression of R-Ras(87L) or R-Ras- (41A) did not affect cell adhesion to fibronectin (Fig. 4 b). This is in contrast to the results of Zhang et al. (1996) who found increased adhesion of myeloid cells to fibronectin substrata. This suggests that there are cell-type specific differences in the effects of R-Ras on different integrin subunits.

To determine whether changes in cell migration and adhesion could be explained by altered expression of endogenous integrin subunits, integrin α subunit expression was examined by flow cytometry. Cells expressing activated R-Ras(87L) or dominant negative R-Ras(41A) or R-Ras(43N) showed no change in the expression of integrin subunits α1, α2, α3, α5, or α6 (not shown). Additionally, there were no changes noted in expression of α4 (not shown), an integrin subunit not normally expressed in these cells. Therefore, R-Ras–induced changes in cell migration and adhesion are not due to changes in integrin subunit expression.

To examine if an apparent change in integrin avidity was reflected in a detectable change in integrin conformation, we used a conformationally sensitive anti–β1 integrin antibody, 12G10. The 12G10 epitope is known to be expressed on integrins that are activated by manganese. We found no increase in 12G10 binding, as determined by flow cytometry, in cells expressing R-Ras(87L) compared with control cells (not shown). This result suggests that changes in apparent integrin avidity because of R-Ras activation either do not correlate to conformational changes, or that R-Ras changes integrin conformation in a manner not detected by this antibody.

The R-Ras–enhanced migration and adhesion of cells to collagen via α2β1, but not fibronectin via α5β1, suggests that, in these cells, R-Ras affects specific integrin subunits. Since the β subunits are identical, we speculated that R-Ras specifically signals to the α2 cytoplasmic domain. To test this hypothesis, the effects of R-Ras on a chimeric integrin containing the α2 cytoplasmic domain were compared with its effects on a similar chimera containing the α5 cytoplasmic domain. Chimeric integrins with the extracellular domain of the α4 subunit and the cytoplasmic domain of either the α2 subunit (X4C2) or the α5 subunit (X4C5) were transfected into control and R-Ras(87L)– expressing cells. Some cells were instead transfected with the complete α4 subunit (X4C4). We selected α4 integrin chimeras since T47D cells normally do not express the α4 subunit (Fig. 5 A, mock profile) which allowed us to monitor the expression of the chimeras in our control and R-Ras–transfected cells. Each of the α4 chimeras was expressed at similar levels on the surface of control and R-Ras–expressing cells (Fig. 5, A–F). Since individual integrin α subunits are not expressed on the cell surface, this indicates that α4β1 heterodimers were formed. Importantly, expression of α4 chimeras did not affect the expression of endogenous α2 (Fig. 5, G and H) or α5 subunits (not shown).

The ligand for the α4β1 integrin is the CS-1 sequence found in the COOH-terminal 40-kD heparin-binding fragment of fibronectin (McCarthy et al., 1986; Humphries et al., 1987). This 40-kD fragment does not contain the RGD sequence that is the ligand for the α5β1 integrin in fibronectin. Therefore, to specifically assess the effects of R-Ras expression on the X4 chimeras, cell migration across membranes coated with either the CS-1 peptide or the 40-kD fragment was determined, since untransfected T47D cells do not migrate across these substrata (not shown). We found that expression of the X4 chimeras in T47D cells conferred the ability to migrate on the 40-kD fragment and the CS-1 peptide (Fig. 6 a). Interestingly, R-Ras(87L) enhanced migration of cells expressing X4C2 and decreased migration of cells expressing X4C5 (Fig. 6 a). This is consistent with our finding that R-Ras enhances migration across collagen and decreases migration across fibronectin (Fig. 2). R-Ras(87L) also enhanced migration of X4C4-expressing cells, suggesting that R-Ras not only signals to the α2, but also the α4 cytoplasmic domains in breast epithelial cells (Fig. 6 a). Cells expressing an α4 chimera containing no α cytoplasmic domain, X4C0, were unable to migrate across the 40-kD and CS-1 substrata (not shown), indicating the importance of an intact α cytoplasmic domain for cell migration. These results suggest that R-Ras activation enhances α2β1 integrin function and diminishes α5β1 integrin function, at least with regard to cell migration.

We also determined whether cells expressing α4 chimeras had altered migration across collagen. Since the α4 chimeras do not bind collagen, any effect will be due to pleiotropic effects on endogenous integrin subunits. Expression of X4C2, but not X4C4 or X4C0, blocked the increase in migration of cells across collagen that is induced by activated R-Ras(87L) (Fig. 6 b), suggesting that there is some form of competition between the X4C2 subunit and the endogenous α2 subunit. Thus, the effects of R-Ras on endogenous α2 subunits can be antagonized specifically by de novo expression of α2 cytoplasmic domains. Such dominant negative effects of integrin cytoplasmic domains have been noted previously with expression of exogenous β cytoplasmic domains (Y.P. Chen et al., 1994; LaFlamme et al., 1994; Lukashev et al., 1994). However, this is the first demonstration of transdominant inhibition by a specific α integrin subunit. This suggests that there may be competition for a limited supply of α cytoplasmic domain binding factors. Conversely, X4C5 expression slightly enhanced the R-Ras(87L)–induced migration across collagen (Fig. 6 b), although this enhancement is not statistically significant when compared to the results obtained with cells expressing X4C4 or X4C0. These results provide additional evidence that R-Ras signaling involves specific integrin subunits.

We previously found that migration induced by the small GTPases, Rac and Cdc42, was dependent on PI3K (Keely et al., 1997). Like Rac and Cdc42, R-Ras also has been shown to activate PI3K (Marte et al., 1996). Therefore, we determined whether pharmacological inhibitors of PI3K, Wortmannin, and LY294002, could inhibit the migration induced by R-Ras. Wortmannin (30 nM) and LY294002 (25 μM) significantly inhibited the migration induced by activated R-Ras(87L) or R-Ras(38V) (Fig. 7 a). However, unlike the total inhibition previously noted with Rac and Cdc42-expressing cells (Keely et al., 1997), the effect of PI3K inhibition on R-Ras–induced migration was only partial; migration was inhibited only by ∼50%. Wortmannin and LY294002 also only partially inhibited the migration induced by activated TC21(72L) (Fig. 7 a). Thus, PI3K activity contributes to R-Ras–induced cell migration, but other signaling pathways are probably also involved.

PKC has been implicated in the migration of some cell types (Vuori and Ruoslahti, 1993; Derman et al., 1997; Batlle et al., 1998). Pretreatment of cells with 1 μM bisindolylmalemide, an inhibitor of most isoforms of PKC, significantly inhibited the migration of R-Ras(87L) or R-Ras(38V)–expressing cells (Fig. 7 b). Once again, inhibiting PKC produced only partial inhibition of R-Ras– induced cell migration. Bisindolylmalemide also partially inhibited the migration of cells expressing TC21(72L) and K-Ras(12V), albeit to an even lesser extent than the effect on R-Ras cells (not shown). The concentration of bisindolylmalemide used was adequate, since pretreatment of cells with this concentration inhibited PKC activity to background levels in an in vitro PKC kinase assay (data not shown). Since certain isoforms of PKC (δ, ε, and η) can be activated downstream of PI3K (Toker et al., 1994; Derman et al., 1997), we wished to determine if PKC and PI3K might be part of the same signaling pathway leading to cell migration, or separate pathways that both contribute to migration. Treatment of cells with Wortmannin and bisindolylmalemide had an additive effect and completely inhibited the increased migration induced by R-Ras expression (Fig. 7 b), suggesting that PI3K and PKC each contribute to R-Ras–induced cell migration through separate pathways.

Other pharmacological inhibitors of signaling pathways potentially downstream of R-Ras were also tested. The MEK inhibitor, PD98059 (Dudley et al., 1995), had no effect on migration induced by activated R-Ras or TC21 (Fig. 7 c), despite inhibiting MAPK activation 50–60% in cells expressing R-Ras(38V), R-Ras(87L), or TC21(72L) (data not shown). In sharp contrast, PD98059 significantly inhibited K-Ras–induced cell migration (Fig. 7 c), demonstrating important differences in the mechanism by which R-Ras and K-Ras induce migration. Rapamycin, an inhibitor of p70 S6 kinase, also did not inhibit cell migration (not shown), which is consistent with previous results with Rac and Cdc42 (Keely et al., 1997).

In this study, we show that activated R-Ras disrupts mammary epithelial polarization in three-dimensional collagen gels and induces migration and invasion across collagenous matrices. These results implicate a role for R-Ras in the transformation of mammary epithelial cells, a role that has not been described previously. We find similar results for a molecule closely related to R-Ras, TC21, which extends previous results by Clark et al. (1996) demonstrating that TC21 can alter breast epithelial morphology. We also find that R-Ras enhances cell adhesion to collagen, which is consistent with its previously described role in promoting increased integrin avidity (Zhang et al., 1996). Importantly, R-Ras enhances cell migration and adhesion specifically on collagen, but not fibronectin. This substratum specificity reflects a difference in the effect of R-Ras on the α2 versus the α5 integrin cytoplasmic domains. Such specificity has not been described previously for the effects of R-Ras on integrin subunits, and could have important implications for inside-out integrin signaling events that lead to migration and invasion on collagenous matrices. Finally, migration induced by R-Ras is partially blocked by inhibitors of PI3K and PKC, but not by a MEK inhibitor, suggesting that unique combinations of signaling pathways are activated by R-Ras compared with other members of the Ras superfamily.

Currently, it is not understood how different integrin subunits specifically link to different inside-out or outside-in signaling pathways. Here, we demonstrate that R-Ras can stimulate cellular migration that is dependent on the α2β1 or α4β1 integrin, but not the α5β1 integrin. In fact, R-Ras activation seems to have a slight inhibitory effect on α5β1-mediated cell migration. The effect of R-Ras is specifically on the α cytoplasmic domain since only this domain differs between the X4C2 and X4C5 chimeras. Our results are consistent with other observations that α subunits appear to regulate the specificity and appropriateness of integrin response. For example, swapping integrin α cytoplasmic domains on the α2 integrin changes the response of cells to collagen (Chan et al., 1992), whereas deletion of the α2 cytoplasmic domain causes unregulated recruitment of integrins into focal complexes, even when the cells are attached to fibronectin (Kawaguchi et al., 1994). Here we are able to link a specific signaling pathway involving R-Ras to the α2β1 integrin, which may help explain these previous observations regarding specific α2 subunit effects. Activation of MAPK by integrins also depends on specific integrin subunits (Wary et al., 1996; Wei et al., 1998). Specifically, α5β1 integrin, but not α2β1 integrin, associates with Shc and activates the MAPK pathway in osteosarcoma cells (Wary et al., 1996). The failure of α2β1 to couple to the MAPK pathway may partially explain why R-Ras–induced migration through α2β1 was unaffected by MEK inhibition. Our results imply that cell migration will differ on different matrices, depending in part on the signaling pathway that has been activated.

Our finding that R-Ras enhances cell adhesion to collagen suggests that R-Ras enhances α2β1 integrin avidity for its ligand. An increase in integrin avidity is consistent with the findings of Zhang et al. (1996). However, we find that R-Ras enhances α2β1- but not α5β1-mediated adhesion, whereas Zhang et al. noted that R-Ras enhances adhesion to fibronectin through α4β1, α5β1, and αvβ3 integrins (Zhang et al., 1996). This difference could relate to the difference between the myeloid cells used in those studies and the epithelial cells used here. The R-Ras– induced increase in avidity determined by cell adhesion could not be correlated to conformational changes of the integrin as detected by the 12G10 anti-β1 antibody. Other investigators have also noted a lack of correlation between ligand binding capability and β1 conformation as detected by various conformationally sensitive antibodies (Bazzoni and Hemler, 1998; Crommie and Hemler, 1998). Our result could mean that R-Ras alters integrin conformation in a manner not detected by this antibody. Alternatively, changes in integrin avidity after R-Ras activation may not be due to obvious conformational changes but to alterations in integrin clustering, adhesion strengthening, or cytoskeletal attachments.

It is not clear at this point how R-Ras stimulation differentiates between the α2 and α5 integrin subunits. An attractive model invokes proteins or factors that bind specifically to either the α2 or α5 integrin subunits. Stimulation of R-Ras may alter the binding properties of such molecules, perhaps through phosphorylation/dephosphorylation events or membrane targeting. These molecules could bind to specific integrin subunits and activate their ligand binding properties, or could alter integrin association with the cytoskeleton. Alternatively, these factors could be inhibitory, such that R-Ras activation releases the integrin subunit from inhibition by these factors. Our finding that the X4C2 chimera blocks the R-Ras–induced migration of cells on collagen, which is not a ligand for X4C2, suggests that X4C2 acts as a dominant negative molecule in this case, presumably through competition for these putative integrin binding molecules. Such transdominant inhibition has been suggested to explain how one integrin β subunit can inhibit the function of another (Balzac et al., 1994; Diaz-Gonzalez et al., 1996; Fenczik et al., 1997) or how chimeras expressing the cytoplasmic tail of the β subunit suppress integrin function (Y.P. Chen et al., 1994; LaFlamme et al., 1994; Lukashev et al., 1994). Interestingly, transdominant β1 integrin suppression can be overcome by expression of CD98 (Fenczik et al., 1997), lending support to the notion that accessory binding molecules regulate integrin function. To our knowledge, ours is the first demonstration of transdominant inhibition on α integrin subunits and implies that factors bind not only to the β, but also the α cytoplasmic domains. Future studies should identify factors specific for the α2 cytoplasmic domain that may play a role in R-Ras–enhanced integrin adhesion and migration.

The fact that R-Ras specifically enhances α2β1 integrin– mediated migration and disrupts tubulogenesis is interesting in light of previous work specifically implicating the α2β1 integrin in these events in mammary epithelial cells (Keely et al., 1995; Zutter et al., 1995). Thus affecting α2β1 function, either through changes in expression levels or by inside-out signaling events, has important consequences for the phenotype of these cells. It is intriguing that apparent increased function of the α2β1 integrin by R-Ras results in the same outcome of increased migration and decreased cell polarization that is noted when α2β1 levels are decreased (Keely et al., 1995). This could indicate that R-Ras activation fundamentally changes the signaling pathways used by the α2β1 integrin, disrupting the normal signaling events that cause the cell to differentiate and polarize in response to the extracellular matrix. These same normal signaling events could be disrupted or diminished when α2β1 levels are decreased, as is noted in breast carcinomas (Zutter et al., 1990). It will be interesting in future studies to determine exactly how cross-talk between the α2β1 integrin and R-Ras affects cellular differentiation, migration, and invasion.

We find a new role for endogenous R-Ras and TC21 in the differentiation and migration of breast epithelial cells. Dominant negative R-Ras inhibits basal cell migration, suggesting that endogenous R-Ras plays a role in the migration of breast epithelial cells. Additionally, we find that inhibitors of PI3K and PKC partially inhibit basal migration, consistent with our model that these molecules are part of the mechanism by which R-Ras contributes to cell migration, whether basal or induced. This further suggests that the signaling pathways under investigation here are likely relevant to understanding the migration of spontaneously occurring breast carcinomas. Dominant negative isoforms of R-Ras or TC21 also enhance tubulogenesis in three-dimensional collagen culture, indicating that antagonizing activated R-Ras or TC21 promotes differentiation. This suggests that activated R-Ras and TC21 are negative regulators of breast cell differentiation, consistent with the finding that expression of activated R-Ras or TC21 isoforms disrupt tubulogenesis. These results could suggest that activation of R-Ras and TC21 turns on signaling pathways that are incompatible with the decision of a cell to slow proliferation and differentiate. Thus, inappropriate activation of R-Ras or TC21 could result in a cell with a more migratory and less differentiated phenotype, which is consistent with a role in transformation.

Our results add to the growing body of evidence pointing to an important role for PI3K in cellular migration events (Kundra et al., 1994; Keely et al., 1997; Shaw et al., 1997; Khwaja et al., 1998). The effect of R-Ras on PI3K possibly is direct since it has been shown that R-Ras binds to PI3K in vitro and activates PI3K in Cos-7 cells (Marte et al., 1996). Since R-Ras has direct effects on integrin avidity, it is interesting that a role for PI3K in affecting integrin activation has been shown in lymphocytes (Shimizu et al., 1995) and platelets (Kovacsovics et al., 1995). Other Ras superfamily members, including H-Ras, Cdc42, and Rac, also bind and activate PI3K (Rodriguez-Viciana et al., 1994; Zheng et al., 1994; Tolias et al., 1995), which in the case of H-Ras plays an important role in cellular transformation (Hu et al., 1995; Rodriguez-Viciana et al., 1997). We previously found a role for PI3K in migration induced by Rac and Cdc42 (Keely et al., 1997) as well as by N- and K-Ras (Keely, P., unpublished observation). Consistent with these observations, expression of activated PI3K promotes migration and invasion (Keely et al., 1997; Shaw et al., 1997).

Although R-Ras is similar to Ras, Rac, and Cdc42 in its use of a PI3K signaling pathway, it differs in its use of other signaling pathways. Migration induced by R-Ras or TC21 was only partially blocked by PI3K inhibitors; this effect was much less dramatic than for migration induced by Rac and Cdc42, which could be completely blocked by the same concentrations of these PI3K inhibitors (Keely et al., 1997). This suggests that PI3K-independent signaling pathways also contribute to R-Ras– stimulated migration in these cells. Although MEK and MAPK activation have been implicated in cell migration in various cells (Klemke et al., 1997), we found that MEK inhibition did not affect migration induced by R-Ras or TC21. In contrast, MEK inhibition completely abolished migration induced by K- and N-Ras. These results are consistent with observations that R-Ras activates PI3K but not Raf (Herrmann et al., 1996; Marte et al., 1996; Huff et al., 1997), unlike H-, N-, and K-Ras, which activate both PI3K and the Raf-MAPK pathway (Vojtek et al., 1993; Rodriguez-Viciana et al., 1997). Since a recent report suggests that the Raf-MAPK pathway might be involved in transformation by TC21 (Rosario et al., 1999), our results suggest that the mechanisms by which TC21 induces migration and transformation differ. Moreover, it appears that different small GTPases stimulate cell migration by activating different combinations of downstream signaling pathways. The use of different signaling pathways may explain why R-Ras differs from N- and K-Ras in stimulating migration across fibronectin.

We find a role for PKC, in addition to PI3K, in migration induced by R-Ras. Although certain isoforms of PKC (δ, ε, and η) can be activated downstream of PI3K and contribute to cell migration (Toker et al., 1994; Derman et al., 1997), the additive effect of PI3K and PKC inhibitors in our assays suggests that PKC is on a separate pathway from PI3K in these cells. Whether PKC is downstream of R-Ras, or part of an independent obligate pathway remains to be determined. To this end, we did not observe an increase in in vitro PKC activity in cells expressing R-Ras in preliminary experiments (our unpublished observation), although translocation of PKC to the membrane and subsequent activation in R-Ras–expressing cells cannot be ruled out. Dominant negative R-Ras was unable to block cell migration induced by PMA (our unpublished results), suggesting PKC is not upstream of R-Ras in our system. In addition to R-Ras, we find a role for PKC in migration induced by K-Ras. Others also have noted synergy between PKC and small GTPases. The Rac exchange factor, Tiam-1, can be activated by PKC (Fleming et al., 1997), which would place PKC upstream of Rac activation. Similarly, PKC is upstream of Ras activation in platelets (Shock et al., 1997). Additionally, PKC activation synergizes with Rho to induce focal adhesion kinase phosphorylation, cell spreading, and actin assembly (Lewis et al., 1996; Defilippi et al., 1997). Interestingly, activation of the α2β1 integrin in monocytes by ligand binding to α5β1 requires PKC (Pacifici et al., 1994), which would be consistent with a model in which PKC is part of the pathway by which R-Ras enhances avidity of the α2β1 integrin. Moreover, PKC activation is involved in activation of the αIIbβ3 integrin by a number of agonists in platelets (Karniguian et al., 1990). Our results are also consistent with other observations that PKC activity contributes to cell migration (Vuori and Ruoslahti, 1993; Derman et al., 1997; Batlle et al., 1998).

In summary, we find that activation of R-Ras stimulates unique combinations of downstream signaling pathways compared with other GTPases of the Ras superfamily. These unique signaling combinations lead to specific effects on certain integrin α subunits to alter cellular responses to collagen such as polarization, migration, and invasion. Such specificity will decide how a cell might respond to different extracellular environments, depending on which Ras family member is activated, and could ultimately determine if a given carcinoma is metastatic or not. These differences have important implications for targeting signaling pathways in antimetastatic therapies since affecting certain signaling pathways may not have the desired effects on all neoplastic cells.

The authors wish to thank Dr. Martin Hemler for contributing the X4 integrin chimeras, Dr. Geoffrey Clark (National Cancer Institute) for helpful discussion and MCF10 cell lines, and Dr. Shayne Huff (University of North Carolina) for helpful discussion.

This work was supported by grants from the Elsa U. Pardee Foundation and the National Institutes of Health grants 2-P01-HL45100-06 and 1-R01HL58919-01 (to L.V. Parise), 1-R29-CA76537-01 (to P.J. Keely), and CA61951, CA67771, and CA76092 (to A.D. Cox).

MAPK

mitogen-activated protein kinase

MEK

MAPK kinase

PI3K

phosphatidylinositol 3-kinase

PI4K

phosphatidylinositol 4-kinase

PKC

protein kinase C