Distinct Roles for Rho Versus Rac/Cdc42 GTPases Downstream of Vav2 in Regulating Mammary Epithelial Acinar Architecture

Abstract

Non-malignant mammary epithelial cells (MECs) undergo acinar morphogenesis in three-dimensional Matrigel culture, a trait that is lost upon oncogenic transformation. Rho GTPases are thought to play important roles in regulating epithelial cell-cell junctions, but their contributions to acinar morphogenesis remain unclear. Here we report that the activity of Rho GTPases is down-regulated in non-malignant MECs in three-dimensional culture with particular suppression of Rac1 and Cdc42. Inducible expression of a constitutively active form of Vav2, a Rho GTPase guanine nucleotide exchange factor activated by receptor tyrosine kinases, in three-dimensional MEC culture activated Rac1 and Cdc42; Vav2 induction from early stages of culture impaired acinar morphogenesis, and induction in preformed acini disrupted the pre-established acinar architecture and led to cellular outgrowths. Knockdown studies demonstrated that Rac1 and Cdc42 mediate the constitutively active Vav2 phenotype, whereas in contrast, RhoA knockdown intensified the Vav2-induced disruption of acini, leading to more aggressive cell outgrowth and branching morphogenesis. These results indicate that RhoA plays an antagonistic role to Rac1/Cdc42 in the control of mammary epithelial acinar morphogenesis.

Introduction

Differentiated epithelia display a polarized architecture that is essential for their functional role as protective barriers and secretory or absorptive surfaces. The polarized epithelial cells associate with each other through lateral cell-cell junctions, which functionally and biochemically segregate the apical surface from the extracellular matrix-contacting basal surface (1, 2). The cell-cell junctions and cell-extracellular matrix interactions stabilize the epithelial structure and ensure appropriate signaling (1, 2). Loss of apical and basolateral polarity is an invariant feature of tumors arising from epithelial cells, also known as carcinomas, which account for most human cancers (3).

In vitro polarity and morphogenesis of epithelia are typically studied using model cell lines, such as Madin-Darby canine kidney (MDCK)⁸ cells as monolayers or in three-dimensional extracellular matrix gels, where cells form a hollow cyst with apicobasal polarity (4). However, linkage of polarity and morphogenesis to oncogenicity has increasingly led to the use of immortalized, non-tumorigenic human epithelial cells. For example, immortalized, non-tumorigenic human mammary epithelial cells (MECs) form basolaterally polarized acinar structures in three-dimensional culture on reconstituted matrices, such as Matrigel (5, 6). These acini consist of a monolayer of cells surrounding a hollow lumen, which is formed during morphogenesis through the elimination of central cells (6, 7). MECs in mature acini exhibit basolateral polarity with an integrin-enriched basal surface contacting the extracellular matrix, basolateral E-cadherin-enriched adherens junctions (AJs), and an apical surface enriched in proteins, such as GM130 or Muc1 (7–9). Although the available immortalized and non-tumorigenic MEC lines, such as MCF10A, do not exhibit clear tight junctions, the ease of visualizing MEC architecture in three-dimensional culture has led to their extensive use in analyzing mechanisms of MEC morphogenesis and alterations of these processes during oncogenic transformation. When grown on Matrigel, non-tumorigenic MECs usually cease to proliferate by approximately day 14 to form quiescent, regular acinar structures (10, 11). In contrast, both oncogenically transformed MECs and breast cancer cells fail to form monolayer structures in Matrigel but proliferate continuously to form larger, irregular structures without hollow lumina (5, 12). The transition from acinar to irregular structures provides a relatively easy means of visualizing perturbations in polarity and morphogenesis as a result of alterations in specific biochemical pathways (6, 13, 14).

Receptor tyrosine kinases (RTKs) of the epidermal growth factor receptor (EGFR) family play critical roles in breast cancer tumorigenesis. EGFR overexpression is found in a significant proportion of breast cancers and correlates with increased aggressiveness and poor prognosis (15–17). When overexpressed in immortalized MECs, EGFR causes disruption of acinar structures (18), implying that EGFR levels need to be tightly controlled to maintain MEC homeostasis. Notably, EGFR levels are down-regulated during MEC acinar morphogenesis (19). Another EGFR family receptor, ErbB2, also induces irregular acinar structures when overexpressed in MECs (10).

Rho, Rac1, and Cdc42 are small GTPases that cycle between the GTP-bound active form and the GDP-bound inactive form, which are regulated by guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins, respectively (20). These GTPases control epithelial cell polarity, as demonstrated in both two- and three-dimensional cell culture systems (1, 21, 22). Previous work has shown that RhoA, Rac1, and Cdc42 are required for the establishment of AJs and participate in tight junction formation in model epithelial cells (1, 23). In three-dimensional culture, Rac1 and Cdc42 play essential roles in the establishment of apicobasal polarity of MDCK cells (24, 25). Paradoxically, these GTPases also disrupt cell-cell junctions and cell polarity when their constitutively active forms are expressed (26–29). Activation of these GTPases by RTKs is also known to regulate cell-cell junctions (30–33). For example, hepatocyte growth factor stimulation of the c-Met receptor activates Rho and Rac1, which in turn are critical for hepatocyte growth factor-induced loss of cell-cell adhesion and disruption of polarity in MDCK cells (34–37). In addition, transforming growth factor-β can induce loss of epithelial cell polarity through the ubiquitination and degradation of RhoA (38), and overexpression of constitutively active Rac1 can disrupt AJs (26). Notably, Rho, Rac1, and Cdc42 have been found to be overexpressed in breast cancer tissues, and their overexpression correlates with breast cancer progression (39). Understanding how Rho GTPases are regulated and how they function in controlling mammary epithelial architecture and morphogenesis, especially downstream of RTKs, is of considerable biological importance.

Rac1 and Cdc42 activation is associated with the disruption of epithelial polarity downstream of PI3K (27, 28). RTKs, such as EGFR, activate PI3K in MECs and are often overexpressed in breast cancer cells (40–42). Taken together, these results imply a role for Rac1- and Cdc42-directed GEFs in the loss of cell polarity in epithelial cells. Among the known Rac1 and Cdc42 GEFs, the Vav family of proteins (Vav1 to -3) are unique in that they directly couple to EGFR and undergo activation through phosphorylation of tyrosine residues located in their N-terminal acidic region (43). Vav2 is a ubiquitously expressed member of the Vav family. Studies in transfected cells have shown that Vav2 interacts with phosphorylated tyrosine residues on EGFR and is subsequently activated by tyrosine phosphorylation and interaction with PIP3, which is generated by PI3K (44). Although Vav2 was initially characterized as a GEF for Rho, Rac1, and Cdc42 (45, 46), recent findings suggest that it may predominantly activate Rac1 (47). Therefore, Vav2 is a likely candidate to be the Rac1-directed GEF downstream of EGFR and PI3K and a potential participant in the disruption of cell polarity in MECs and breast cancer cells.

Although stimulation of EGFR in model cells activates Vav2 and Rho GTPases (46, 48), our previous studies show that MEC lines, such as 16A5 and MCF10A (18), when grown in three-dimensional Matrigel cultures, retain their polarity despite the use of EGF-containing media. Because of this, we considered the possibility that alterations in Rho GTPase signaling machinery during three-dimensional culture may preserve MEC polarity. Because our preliminary data showed that Vav2 is the only Vav family member detected in a number of MECs, we investigated how Vav2 functions within MECs during three-dimensional culture. We found that growing 16A5 MECs in three-dimensional culture results in EGFR down-regulation. The three-dimensional cultured cells showed reduced Vav2 phosphorylation and a decrease in RhoA, Rac1, and Cdc42 activation in response to EGF stimulation. To test whether or not Vav2 and its downstream Rho GTPases affect MEC morphogenesis and architecture, we created a tetracycline-inducible (Tet-On) 16A5 MEC line in which the expression of a constitutively active Vav2 mutant (Y172F) is under the control of doxycycline (DOX). By inducing active Vav2 expression in combination with shRNA-mediated knockdown of RhoA, Rac1, or Cdc42 at different stages of acinar morphogenesis, we show that RhoA and Rac1/Cdc42 play distinct and seemingly opposite roles in the regulation of MEC polarity and morphogenesis.

EXPERIMENTAL PROCEDURES

Antibodies and Other Reagents

Rabbit anti-Vav2 peptide sera were generated against Vav2 amino acids 208–222 (QETEAKYYRTLEDIE) through a commercial vendor (Animal Pharma Inc.). The monoclonal anti-EGFR (528, ATCC) and anti-E-cadherin (clone E4.6) (49) (provided by Drs. Michael Brenner and Jonathan Higgins, Brigham and Women's Hospital, Boston, MA) antibodies were purified from mouse hybridoma supernatants. Purified anti-phosphotyrosine antibody 4G10 (50) was provided by Dr. Brian Druker (Oregon Health Sciences University, Portland, OR). The following antibodies were commercially obtained: monoclonal anti-RhoA, anti-Rac1, and anti-GM130 (BD Biosciences); anti-Rac1 (Cytoskeleton); anti-β-actin (Sigma); anti-α6 integrin (Chemicon Inc.); and anti-Cdc42 (Cell Signaling). Alexa Fluor 594-conjugated phalloidin used to stain polymerized actin was from Invitrogen.

shRNA Constructs and cDNAs

The shRNA sequences specific for the genes of interest were identified using the online S-fold software, subjected to a BLAST search against the NCBI data base to minimize off-target possibilities, and cloned into the pSuper.retro vector (OligoEngine Inc.). The scrambled control nucleotide sequence is AAGAGCATCTCCACCTCTA. The shRNA sequences are as follows: for RhoA, GCAGGTAGAGTTGGCTTTG (sequence 1) and CGACAGCCCTGATAGTTTA (sequence 2); for Rac1, GACACGATCGAGAAACTGA (sequence 1) and GTGAAGAAGAGGAAGAGAA (sequence 2); and for Cdc42, GATAACTCACCACTGTCCA (sequence 3) and ACACAGAAAGGCCTAAAGA (sequence 5). The C-terminally YFP-tagged murine Vav2 cDNA was obtained from the Signaling Consortium (The Signaling Gateway) and subcloned into the pRevTRE vector (Invitrogen) through an engineered SalI site.

Human MEC Lines and Cell Culture

The 16A5 cell line is an HPV E6/E7-immortalized derivative of the primary MEC line 76N obtained from a normal human mammoplasty specimen (51). These cells were routinely maintained in two-dimensional culture as described (51). Standard procedures were used to introduce pSuper.retro-based retroviruses into 16A5 cells to generate stable shRNA-expressing lines (18). The transductants were selected and maintained in DFCI-1 medium (51) supplemented with puromycin (0.5 μg/ml) or G418 (500 μg/ml) and used as polyclonal cell lines. The 16A5-Tet-On cell line is a clone of the 16A5 cell line infected with pRev-Tet-On vector (Invitrogen) and selected in G418 for maximal induction of transfected genes.

For three-dimensional Matrigel culture, 2.5 × 10³ cells in 0.4 ml of 2% reduced growth factor Matrigel (BD Biosciences) were added to DFCI-1 medium containing 3 ng/ml EGF and plated in a 60-mm plate onto a polymerized layer of 100% Matrigel, as described previously (11). Fresh medium was added to the cells every 2 days. Phase-contrast images were obtained at the indicated time points.

EGF Stimulation and GST Pull-down of Activated RhoA, Rac1, and Cdc42

For EGF stimulation, cells were grown for 3 days in DFCI-1 medium without EGF; EGF was then added at 100 ng/ml at the indicated time points before cell lysis. Bacterially expressed and purified GST-RBD (Rho-binding domain of rhotekin; interacts with activated GTP-bound RhoA) or GST-PBD (p21-binding domain of Pak1; interacts with activated GTP-bound Rac1 and Cdc42) was used to pull-down GTP-bound forms of RhoA or Rac1 and Cdc42, respectively, as described previously (52, 53). Two-dimensional cultured 16A5 MECs were washed with ice-cold Tris-buffered saline and lysed in radioimmune precipitation buffer (50 mm Tris, pH 7.2, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 10 μg each of leupeptin and aprotinin, and 0.1 mm phenylmethylsulfonyl fluoride). Three-dimensional cultured cells were lysed in the same buffer by scraping the Matrigel-cell mixture from the plates. Cell lysates were rocked at 4 °C for 10 min and cleared by centrifugation at 16,000 × g at 4 °C for 10 min to remove the insoluble fraction and Matrigel. Equal aliquots of lysates were incubated with 20–30 μg of purified GST-RBD or GST-PBD immobilized on glutathione-Sepharose beads at 4 °C for 45 min. The beads were washed four times with wash buffer (50 mm Tris (pH 7.5), 1% Triton X-100, 150 mm NaCl, 10 mm MgCl₂, and 1 mm phenylmethylsulfonyl fluoride). Bound RhoA, Rac1, or Cdc42 protein was detected by Western blotting for RhoA (catalog number 26C4, Santa Cruz Biotechnology, Inc. (Santa Cruz, CA)), Rac1 (catalog number ARC03, Cytoskeleton), or Cdc42 (catalog number 2462, Cell Signaling).

Confocal Immunofluorescence Microscopy

For immunofluorescence analysis, the three-dimensional cultures were prepared in 8-well chamber slides (BD Biosciences). The acinar structures were fixed in 4% formaldehyde, phosphate-buffered saline on the indicated days, permeabilized with 0.5% Triton X-100 for 5 min, and stained with anti-E-cadherin, anti-GM130, or anti-α6 integrin primary antibodies. This was followed either with Alexa Fluor 488- or Alexa Fluor 594-conjugated secondary antibodies and DAPI or Topro-3 (Molecular Probes) or with Alexa Fluor 594-conjugated phalloidin. The slides were mounted with Vectashield mounting medium (Vector Laboratories). Images were acquired with a Nikon C1 or a Zeiss LSM 510 confocal microscope under ×400 or ×600 magnification. All images represent the central plane of the acini using 0.5-μm-thick optical sections.

Statistical Analysis

One-way analysis of variance and Student's t test were used to determine the statistical significance of differences in loss of polarity between 16A5-Tet-On-172F cells expressing control shRNA versus RhoA, Rac1, or Cdc42 shRNAs.

RESULTS

Down-regulation of EGFR, Reduction in Vav2 Phosphorylation, and Activation of Rho GTPases in Mature Mammary Epithelial Acini

We have previously shown that the immortalized, non-malignant 16A5 MEC line resembles the MCF10A MEC line in forming basolaterally polarized acini in three-dimensional Matrigel culture (18) with relatively nondescript tight junctions, as observed with zona occludens-1 staining (data not shown). However, these structures are sensitive to the levels of EGFR; overexpression of EGFR induced abnormal structures in a significant proportion of acini (18). We therefore hypothesized that controlled EGFR signaling via downstream activation of Rho GTPases may be critical for the maintenance of acinar integrity. EGF stimulation indeed activated Rac1, Cdc42, and RhoA in two-dimensional culture, but surprisingly we found that EGF stimulation did not result in Rac1 and Cdc42 activation, whereas RhoA was weakly activated in three-dimensional as compared with two-dimensional cultures (Fig. 1, A–D). This result was confirmed in three independent experiments.

FIGURE 1.

Down-regulation of EGFR levels and phosphorylation, Vav2 phosphorylation, and activities of RhoA, Rac1, and Cdc42 in mammary epithelial cells grown in three-dimensional culture together with selective blunting of EGF-stimulated Rac1 and Cdc42 but not RhoA activation. A, 16A5 cells were grown either in two- or three-dimensional culture (2D or 3D, respectively) for 10 days and then EGF-starved for 3 days. The cells were either left unstimulated or stimulated with EGF (100 ng/ml) for 10 min prior to lysis. Whole-cell lysates were subjected to GST-RBD or GST-PBD pull-down and immunoblotted for RhoA, Rac1, or Cdc42 (representative image of three independent experiments is shown). B–D, the GTP-bound forms of Rac1, Cdc42, and RhoA were normalized to the loading control, and the average relative density in resting versus EGF stimulated cells and two-dimensional versus three-dimensional culture-grown cells were calculated (error bars represent S.D.). E, whole-cell lysates were immunoblotted for EGFR, phosphotyrosine (pY), or Vav2 and reprobed for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (loading control). Anti-Vav2 immunoprecipitates (IP) were immunoblotted for phosphotyrosine and reprobed for Vav2.

We reasoned that differential EGFR levels in two-dimensional versus three-dimensional may account for this difference. To test if EGFR was indeed down-regulated in three-dimensional cultures of 16A5 cells, we compared EGFR protein levels in 16A5 MECs grown in two-dimensional versus three-dimensional culture. Western blotting (Fig. 1E, top) and fluorescence-activated cell sorting analysis (25–30% lower mean fluorescence intensity; data not shown) showed that EGFR was indeed substantially down-regulated in three-dimensional culture (the latter was assessed when acini were fully formed; day 14). Furthermore, EGF-induced phosphorylation of EGFR was markedly decreased in three-dimensional compared with two-dimensional cultured cells (Fig. 1E, middle). These results reveal that the EGFR-signaling pathway is differentially regulated in two-dimensional versus three-dimensional cultures (19).

Because EGFR is known to activate Rac1 and Cdc42 through Vav2 (44, 46), which is the only detectable Vav family member in several MEC and breast cancer cell lines (data not shown), and Rac1 activation can disrupt breast cancer cell polarity (27), we examined Vav2 phosphorylation in response to EGF stimulation in two-dimensional versus three-dimensional cultured cells. Consistent with reduced EGFR activation, Vav2 phosphorylation in response to EGF was significantly lower in three-dimensional versus two-dimensional cultured cells despite similar protein levels (Fig. 1E). Thus, attainment of polarity in three-dimensional MEC culture is associated with EGFR down-regulation and reduced Vav2 activation, which is accompanied by a relatively selective blunting of EGF-induced Rac/Cdc42 but not RhoA activation.

Inducible Expression of Vav2-Y172F Mutant Activates Rac1 and Cdc42 but Not RhoA in MEC Acini and Induces the Disruption of Preformed Acinar Structure and Formation of Abnormal Structures in Three-dimensional Matrigel Culture

The correlative findings above suggested that Vav2, whose phosphorylation was reduced in three-dimensional culture in conjunction with reduced Rac/Cdc42 activity, could serve as an important driver of Rac1 and Cdc42 activation in MECs. To assess if this was indeed the case, we generated a Tet-On 16A5 cell line (16A5-Tet-On-172F) in which the expression of YFP-tagged, constitutively active Vav2 mutant (Vav2-Y172F) is DOX-inducible. DOX-induced expression of Vav2-Y172F in 16A5-Tet-On-172F cells grown in two-dimensional culture for 3 days led to increased levels of active Rac1 and Cdc42 but not of RhoA (Figs. 2, A and B, and 10C). Similarly, inducible expression of Vav2-Y172F in preformed 16A5 MEC acini (cultured in Matrigel for 14 days prior to induction) for 3 days activated Rac1 and Cdc42 but not RhoA (Fig. 2, C and D). In fact, expression of Vav2-Y172F in acini decreased the GTP-bound form of RhoA (Fig. 2, C and D). These results suggest that Vav2 is primarily a Rac1 and Cdc42 GEF in 16A5 MECs, a result consistent with recent results where Vav2 activated Rac1 but not RhoA in pancreatic cancer cells (47).

FIGURE 2.

Induced overexpression of constitutively active Vav2-Y172F mutant activates Rac1 (and Cdc42) but not RhoA in two-dimensional (2D) as well as three-dimensional (3D) culture. 16A5-Tet-On-Vav2-Y172F cells were grown in two-dimensional culture in DFCI medium without or with DOX induction (2 μg/ml) for 3 days (A and B). Alternatively, cells were grown in three-dimensional culture for 14 days and then treated with vehicle or DOX (2 μg/ml) for 3 days (C and D). Rac1, Cdc42, and RhoA activation was assessed using pull-down assays (A and C) and quantified (C and D) as in Fig. 1 (representative image of three independent experiments is shown).

FIGURE 10.

Overexpression of Vav2 induces remodeling of cell-cell junctional actin structures; role of Rac1 and RhoA. 16A5-Tet-On-Vav2-Y172F cells were grown in two-dimensional culture and EGF-starved together with treatment of vehicle (−DOX) or DOX (2 μg/ml, +DOX) for 3 days (expression of YFP-Vav2-Y172F (Y) in induced cells is seen in green). The cells were immunostained for E-cadherin (E, blue) and F-actin (phalloidin (P), red), and confocal images were acquired at the subapical plane. The top panels show images of whole-cell colonies, and selected areas are shown at higher magnification in the lower panels (middle panels, F-actin; bottom panels, E-cadherin).

Given the recently identified role for Rac1 in the disruption of mammary epithelial cell polarity downstream of PI3K (54), we surmised that activation of Rac1/Cdc42 by active Vav2 could disrupt MEC architecture and morphogenesis in three-dimensional culture. To test this hypothesis, 16A5-Tet-On-172F cells were grown in Matrigel for 14 days, and constitutively active Vav2 was then induced with DOX. Phase-contrast microscopy demonstrated that, similar to parental cells, 16A5-Tet-On-172F cells formed regular acini by day 14 when grown on Matrigel. Analysis after 6 days of DOX induction, using a phase-contrast microscope with both direct and fluorescence light, indicated that a substantial proportion of preformed acini showed disruptions of acinar architecture (Fig. 3A). Although parts of the original acinar structure were still discernible (Fig. 3A), other regions showed abnormal architecture, which we categorized as irregular (regional irregular outlining; Fig. 3A, open arrowhead and enlarged in the lower panel), multilobular (two or more irregular cellular masses; Fig. 3A, black arrowhead and enlarged in the lower panel), and branching (Fig. 3A, open arrow and enlarged in the lower panel).

FIGURE 3.

Inducible expression of constitutively active Vav2-Y172F in preformed mammary epithelial cell acini induces abnormal structures. A, 16A5-Tet-On-YFP or 16A5-Tet-On-Vav2-Y172F cells were grown in Matrigel for 14 days to generate acini and then induced with DOX (2 μg/ml) for 6 days. Cells were analyzed using a phase-contrast microscope under transmitted light for acinar morphology (Acini) and under fluorescence light for YFP-tagged Vav2 proteins. B, whole-cell lysates of acini prepared as in A were immunoblotted with an anti-green fluorescent protein antibody to detect YFP-tagged Vav2 protein (*, a postlysis cleavage product of YFP-Vav2-Y172F). C, irregular (see A, open arrowhead, enlarged in the lower panel), multilobular (see A, black arrow, enlarged in the lower panel), or branched (see A, open arrow, enlarged in the lower panel) acini were counted in three replicates of each cell line, and the mean percentage of abnormal acini are presented with S.D. as error bars. The p values of the difference between DOX-induced 16A5-Tet-On-YFP and 16A5-Tet-On-Vav2-Y172F cell lines (two-tailed t test) are indicated above the error bars. GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

In contrast, vehicle-treated 16A5- Tet-On-172F cells displayed structures comparable with those of uninduced cells (Fig. 3A) or DOX-induced 16A5-Tet-On-YFP control cells (Fig. 3, A–C). These results indicate that activation of Vav2 in mature 16A5 MEC acini disrupts the acinar architecture and results in regionally unregulated cell proliferation. Notably, overexpression of the Vav2-Y172F mutant did not affect EGFR degradation (supplemental Fig. 1E), excluding the possibility that this phenotype is a result of altered EGFR down-regulation.

Given our previous findings that EGFR overexpression in 16A5 cells leads to disruption of cell polarity (18), we asked whether Vav2 participates in the EGFR-mediated disruption of acinar architecture. 16A5-Tet-On-YFP, YFP-WT Vav2, and YFP-Vav2-Y172F cells grown in Matrigel for 14 days were induced with DOX in the presence or absence of EGF for 7 days. The abnormal acinar structures (as defined above) were quantified at day 21 of Matrigel culture. Interestingly, an increase in abnormal acinar structure was observed upon induced expression of WT Vav2 only in the presence of EGF (supplemental Fig. 1). On the other hand, Vav2-Y172F expression was associated with a significant increase in abnormal acinar structures in the absence of EGF, with a further increase in the presence of EGF (supplemental Fig. 1). These results suggest that Vav2 plays an important part in EGFR-dependent disruption of three-dimensional MEC architecture.

RhoA and Rac1/Cdc42 Play Opposite Roles in Regulating MEC Acinar Architecture and Morphogenesis

The differential down-regulation of EGF-induced Rac1/Cdc42, compared with RhoA activation, in three-dimensional MEC culture, together with the correlation of activated Vav2-induced disruption of MEC acini with Rac1/Cdc42 activation (without concomitant RhoA activation), suggests that RhoA activation is compatible with retention of normal acinar architecture, whereas Rac/Cdc42 activation is not. When combined, these data point to potentially distinct roles for these Rho family GTPases.

To address this possibility, we evaluated the effects of Rho GTPase knockdown on activated Vav2-induced disruption of MEC acinar architecture and morphogenesis. Western blot analyses demonstrate the selective knockdown of RhoA (lanes 3–6 in panel 4), Rac1 (lanes 7–10 in panel 2), and Cdc42 (lanes 11–14 in panel 3) when cognate shRNAs were stably expressed in 16A5-Tet-On-Vav2-Y172F cells (Fig. 4A).

FIGURE 4.

Opposite effects of RhoA versus Cdc42/Rac1 knockdown on Vav2-Y172F-induced abnormalities of mammary epithelial acinar structures. A, 16A5-Tet-On-Vav2-Y172F cells with stable expression of control shRNA, RhoA shRNAs, Rac1 shRNAs, or Cdc42 shRNAs were grown in Matrigel for 14 days to generate acini and then induced with DOX for 6 days. The cells were lysed, and whole-cell lysates were immunoblotted for Vav2, Rac1, Cdc42, and RhoA. B, the percentage of abnormal acini with characteristics of irregular, multilobular, or branching structures were counted in four replicates and are presented as mean percentages of abnormal acini with S.D. values shown as error bars. The significance of differences between the control and DOX-induced experimental groups was determined using one-way analysis of variance followed by the two-tailed t test. The following sets show a statistically significant difference in the mean percentage of abnormal acini between control and specific shRNA: RhoA shRNA 1, p = 0.0009; RhoA shRNA 2, p = 0.0016; Rac1 shRNA 1, p = 0.0078; Rac1 shRNA 2, p = 0.0143; Cdc42 shRNA 3, p = 0.0057; Cdc42 shRNA 5, p = 0.0086. C, the percentages of irregular, multilobular, and branching structures (based on data shown in B) are shown with S.D. values shown as error bars. Left panel, −Dox, right panel, +Dox.

The control shRNA or individual Rac1/Cdc42/RhoA shRNA-expressing 16A5-Tet-On-Vav2-Y172F cell lines were plated in Matrigel. Vav2-Y172F expression was induced starting either on day 3 (before the control cells attained polarity; see Fig. 6) or on Day 14 (after the control cells exhibited polarity; Figs. 4 and 5). Cells induced from day 3 onward were harvested at day 14, whereas those induced from day 14 onward were harvested on day 21. The proportion of cells with abnormal structures was quantified, and the cells were subsequently immunostained with polarity markers and visualized with a confocal microscope at the central plane of the acinar structures. When grown without Vav2 induction, cells with Rac1 or Cdc42 knockdown formed regular sized acini by day 20, whereas a relatively small proportion of acini in RhoA knockdown cells were irregular (Fig. 4B). When 16A5-Tet-On-Vav2-Y172F cells with control shRNA were induced from day 14 to day 20, they showed regional disruption of acinar architecture and formation of irregular, multilobular, or branching structures, as anticipated from previous results (Figs. 4–8).

FIGURE 6.

Confocal image analysis of the role for RhoA, Cdc42, and Rac1 in Vav2-Y172F-induced alterations in established MEC acini; analysis using the apical polarity marker GM130. A, the acini were prepared as in Fig. 5 and immunostained with anti-GM130 (G, red) antibody together with DAPI (D, blue) to visualize nuclei, and confocal images were acquired at the central plane of the acinar structures (scale bar, 50 μm). Green fluorescence (Y) represents induced Vav2 proteins. Black and white images are presented in the lower panels (white, GM-130; gray, nuclei). B, the number of cells (from a total of 10 acini for each condition) exhibiting apical orientation of GM130 upon visual inspection were counted and presented as a percentage of total cells; average values from three experiments are presented with S.D. shown as error bars.

FIGURE 5.

Confocal image analysis of the role for RhoA, Cdc42, and Rac1 in Vav2-Y172F-induced alterations in established MEC acini; analysis using the basal polarity marker α6 integrin. The 16A5-Tet-On-Vav2-Y172F cells with stable expression of control shRNA, RhoA shRNA, Rac1 shRNA, or Cdc42 shRNA were grown in Matrigel for 14 days to generate acini and then induced with DOX (2 μg/ml) for 6 days. The cells were immunostained with an anti- α6 integrin (I, red) antibody together with DAPI (D, blue) to visualize nuclei, and confocal images were acquired at the central plane of the acinar structures (scale bar, 50 μm). Yellow fluorescence (Y) represents induced Vav2 proteins. Black and white images are presented in the lower panels (white, α6 integrin; gray, nuclei).

FIGURE 7.

Confocal image analysis of the role for RhoA, Cdc42 and Rac1 in Vav2-Y172F-induced alterations in established MEC acini; analysis using the basolateral polarity marker E-cadherin. The acini were prepared as in Fig. 5 and immunostained with an anti-E-cadherin (E, red) antibody together with DAPI (D, blue) to visualize nuclei, and confocal images were acquired at the central plane of the acinar structures (scale bar, 50 μm). Green fluorescence (Y) represents induced Vav2 proteins. Black and white images are presented in the lower panels (white, E-cadherin; gray, nuclei).

FIGURE 8.

Confocal image analysis of the role for RhoA, Cdc42, and Rac1 in Vav2-Y172F-induced alterations in established MEC acini; analysis of apicolateral actin staining with phalloidin. The acini were prepared as in Fig. 5 and stained with phalloidin (P, red) together with DAPI (D, blue) to visualize nuclei, and confocal images were acquired at the central plane of the acinar structures (scale bar, 50 μm). Green fluorescence (Y) represents induced Vav2 proteins. Black and white images are presented in the lower panels (white, actin; gray, nuclei).

Immunostaining showed that α6 integrin (basal surface marker; Fig. 5) was confined to the basal surface in all of the cell lines. However, the α6 integrin staining showed an intermittent pattern (Fig. 5, box) in Rac1 knockdown cells as opposed to continuous staining in all other cell lines (Fig. 5). Quantification of cells with different staining patterns in three independent experiments indicated that more than 90% of the control shRNA-, Cdc42 shRNA-, or RhoA shRNA-expressing acini showed a continuous integrin staining pattern, whereas an average of 67% of Rac1 shRNA-expressing acini exhibited an intermittent integrin staining pattern (data not shown).

Staining for GM130 (apical marker) revealed that it was oriented to the apical side of the nuclei in a majority of the control shRNA-expressing cells in the absence of DOX (Fig. 6A, white arrows indicate apical orientation, and open arrows indicate irregular orientation). However, the apical orientation of GM130 appeared to be diminished in RhoA, Cdc42, and Rac1 shRNA-expressing acini (Fig. 6A). Visual inspection and quantification showed a remarkable reduction in the proportion of cells with apical orientation of GM130 in acini of RhoA, Cdc42, or Rac1 shRNA-expressing cells (Fig. 6B). Upon DOX induction, the polarized GM130 orientation was lost in cells that formed abnormal structures, and the proportion of cells with apical orientation of GM130 was significantly reduced (Fig. 6B).

Immunostaining for E-cadherin or F-actin revealed the basolateral distribution of E-cadherin (Fig. 7) and apicolateral distribution of F-actin (Fig. 8) in all of the cell lines. Notably, in cells that formed multilayered abnormal structures upon DOX induction, there was a diminution of staining for AJs (Fig. 7, white arrowhead) and junctional actin cables (Fig. 8, white arrowheads).

The formation of abnormal structures upon induction of Vav2-Y172F expression was significantly enhanced in RhoA knockdown cells, whereas it was reduced in Rac1 or Cdc42 knockdown cells (Fig. 4B). Further quantification of the abnormal acini as irregular, multilobular, and branching structures revealed that Vav2-Y172F expression primarily led to irregular acini in control shRNA-expressing cells (Fig. 4C, right). However, Vav2-Y172F expression in combination with RhoA knockdown led to a more pronounced abnormality, as shown by predominantly multilobular and branching acini (Fig. 4C, right). In contrast, the Vav2-Y172F-induced branching phenotype was not seen in Rac1 or Cdc42 knockdown cells (Fig. 4C, right).

Induction of Vav2-Y172F expression from day 3 onward in control shRNA-expressing cells resulted in primarily multilobular acini lacking lumina (Fig. 9). Vav2-Y172F expression in RhoA knockdown cells resulted in tubule-like branching in addition to multilobular acini (Fig. 9, arrowheads). On the other hand, Rac1 knockdown cells largely formed regular acinar structures even when Vav2-Y172F expression was induced (Fig. 9).

FIGURE 9.

Overexpression of Vav2-Y172F during the early phase of three-dimensional Matrigel culture blocks acinar morphogenesis; role of Rac1 and RhoA. 16A5-Tet-On-Vav2-Y172F cells with stable expression of control shRNA, RhoA shRNAs, or Rac1 shRNAs were seeded in Matrigel, and Vav2-Y172F expression was induced with DOX starting on day 3. On day 14 (when control cultures show regular acini), the cells were immunostained with an anti-E-cadherin antibody (red) together with Topro-3 (blue) to visualize nuclei, and confocal images were acquired at the central plane of the acinar structures (scale bar, 50 μm). Green fluorescence represents induced Vav2 proteins. The arrowheads indicate branching structures; these were observed in a majority of acini formed by RhoA knockdown cells.

Reorganization of Cell-Cell Junctional Actin Cytoskeleton and Disruption of AJs upon Inducible Expression of Activated Vav2; the Role of Rac1 and RhoA

Since Rac1 activation has been shown to disrupt AJs in epithelial cells (26) and Vav2 regulates the actin cytoskeleton in model cell systems (46), our findings of reduced E-cadherin and junctional actin staining in activated Vav2-induced abnormal acini prompted us to assess whether Vav2 activation is able to reorganize the junctional actin cytoskeleton and disrupt AJs in MECs.

Because the visualization of AJs and associated junctional actin structures in epithelial cells is optimal for cells grown as monolayers, 16A5-Tet-On-172F cells were grown in two-dimensional cultures in EGF-deprived DFCI medium until cell-cell adhesions formed. The cells were then treated with vehicle or DOX for 3 days to induce Vav2-Y172F expression; confocal imaging confirmed the YFP-Vav2-Y172F expression in DOX-treated but not in untreated cells (Fig. 10, green, top). Compared with control cells, Vav2-Y172F-expressing cells spread out more and became flatter as they increased in size. The induced expression of Vav2-Y172F was accompanied by a loss of circumferential actin cables and the formation of thin perijunctional actin bundles. In addition, E-cadherin staining at the cell-cell junctions was discontinuous, with more diffuse cytoplasmic staining in Vav2-Y172F-expressing cells (Fig. 10).

Similar to three-dimensional cultures, the phenotype induced by the overexpression of Vav2-Y172F was significantly blocked by Rac1 knockdown but not by RhoA knockdown (Fig. 10). Instead, RhoA knockdown exacerbated the Vav2-Y172F-induced disruption of AJs, with reduced E-cadherin staining at cell-cell interfaces. In addition, E-cadherin colocalized with the reorganized actin cytoskeleton, and more cytoplasmic E-cadherin staining was observed (Fig. 10). These results suggest that Rac1 but not RhoA is required for Vav2-mediated reorganization of the junctional actin cytoskeleton and disruption of AJs.

Knockdown of RhoA did not affect the Vav2-induced Rac1 activation (supplemental Fig. 3, A and B), thereby excluding the possibility that functional antagonism, with RhoA inhibiting the activation of Rac1 by Vav2, is involved. Furthermore, knockdown of Rac1 did not affect the Vav2-induced activation of Cdc42, and knockdown of Cdc42 did not change Rac1 activation by Vav2 (supplemental Fig. 3, C and D). These results suggest that Vav2 activates Rac1 and Cdc42 simultaneously and that the combined activity of Rac1 and Cdc42 action probably mediates the disruption of AJs and acinar architecture downstream of Vav2.

To test whether RhoA regulates AJs downstream of EGFR, WT RhoA was stably overexpressed in the telomerase (hTERT)-immortalized MEC line 81N-Tert (supplemental Fig. 2A) because RhoA could not be stably overexpressed in 16A5 cells. EGF-induced activation of RhoA was enhanced in RhoA-overexpressing 81N-Tert cells (supplemental Fig. 2A). Compared with control cells, RhoA-overexpressing 81N-Tert cells showed very little reorganization of the junctional actin cytoskeleton and AJs upon EGF stimulation. In contrast, circumferential actin cables were more intense in RhoA-overexpressing cells upon EGF stimulation (supplemental Fig. 2B). Altogether, these results strongly suggest a functional role for RhoA in maintaining epithelial AJs in the presence of activated EGFR and Vav2.

DISCUSSION

Rho GTPases have been implicated in mammary tumorigenesis based on their overexpression in breast cancer cells (39) and the functional roles they play in controlling cell proliferation, survival, migration, and polarity. An invariant feature of oncogenic transformation of mammary and other epithelial cells is a loss of polarity and an inability to undergo acinar morphogenesis in three-dimensional culture (13, 14). Thus, understanding the relative importance of distinct Rho GTPases in regulating mammary epithelial acinar morphogenesis and polarity is of substantial physiological as well as cancer-related importance.

Here, we show that three-dimensional acinar morphogenesis is associated with the down-regulation of Rac1/Cdc42 activation, whereas RhoA activation remains intact. By utilizing a DOX-inducible activated Vav2 expression system together with shRNA knockdown of Rac1, Cdc42, and RhoA, we show that different Rho GTPases play functionally opposite roles in the regulation of mammary epithelial cell polarity and acinar morphogenesis; the activation of Rac1/Cdc42 promotes disruption of polarity and abnormal acinar morphogenesis, whereas Rho activity appears to preserve polarity and acinar structure.

In the context of the regulation of polarity, Rac1 is required for the appropriate orientation of apical and basal polarity of MDCK cells (25) and is aberrantly activated downstream of PI3K within invasive breast cancer cells and perturbs their polarity (27). Cdc42 is also essential for the establishment of apical polarity in MDCK cells (24). However, less is known about the role of RhoA activation in acinar morphogenesis. Interestingly, RhoA activity was reported to be up-regulated in response to changes in extracellular matrix rigidity within three-dimensional culture, and this increase in RhoA activation induced the distortion of acinar structures (29). In our studies, EGFR stimulation is associated with RhoA activation both in two-dimensional and three-dimensional MEC culture. However, because Rac1/Cdc42 activation was not observed in three-dimensional culture, RhoA activity did not promote the disruption of acinar architecture.

Activation of Rho GTPases is associated with the stimulation of a number of cell surface receptors. In the context of epithelial cells, stimulation through RTKs, such as EGFR, is a potent means of controlling Rho GTPases (55). Consistent with studies in model cell lines, the Vav family of proteins represent some of the key Rho GTPase-directed GEFs (43). We found that several MEC lines, including the 16A5 cells utilized in this study, only express one member of this protein family, Vav2. This allowed us to drive Rho GTPase activation in MECs to assess the impact on cell polarity and acinar morphogenesis. Furthermore, down-regulation of EGFR, which is associated with acinar morphogenesis in three-dimensional culture, was observed in conjunction with reduced Vav2 phosphorylation and reduced Rac/Cdc42 activity, providing further support for the decision to use activated Vav2 to assess the role of downstream Rho GTPases.

DOX-inducible expression of a constitutively active form of Vav2 (Vav2-Y172F) in three-dimensional culture of 16A5 MECs led to the activation of Rac1 and Cdc42 but not RhoA. More importantly, controlled expression of activated Vav2 at different stages of MEC acinar morphogenesis demonstrated not only that Vav2 could block the completion of acinar morphogenesis but also that it could disrupt established acinar structures and lead to the formation of abnormal structures reminiscent of cancerous cellular growths. These alterations in acini were associated with a local loss of polarity as assessed with markers of apicobasal polarity.

The phenotype induced by Vav2-Y172F is largely dependent on the activation of Rac1 and Cdc42 because knockdown of Rac1 or Cdc42 significantly blocked the phenotypic characteristics associated with Vav2 induction. In contrast, knockdown of RhoA led to a partial distortion of acinar architecture by itself and further promoted the phenotype induced by activated Vav2. Moreover, compared with mainly irregular structures induced by Vav2-Y172F alone (a milder phenotype), overexpression of Vav2-Y172F in combination with RhoA knockdown resulted in the formation of predominantly multilobular and branching structures, a more severe phenotype associated with characteristics of cancerous cells (6). Thus, Rac1/Cdc42 and RhoA appear to play opposite roles in regulating three-dimensional acinar structure, with RhoA functioning to stabilize acinar structures, while Rac1/Cdc42 acts to promote acinar disruption.

Interestingly, knockdown of Rac1 resulted in an intermittent α6 integrin staining pattern at the basal surface and reduced the apical orientation of GM130 without remarkably affecting basolateral E-cadherin staining. Rac1 is required for the orientation of apical polarity in MDCK cells through the mediation of laminin assembly under the basal surface (25). Under our experimental conditions, the presence of laminin in Matrigel is seemingly sufficient for the proper orientation of α6 integrin to the basal surface. However, the distribution of integrin was different from the control cells, implying that Rac1 knockdown affected basal polarity, which may account for the diminished apical orientation of GM130. Cdc42 knockdown significantly reduced the apical orientation of GM130 in the cells, consistent with the finding that Cdc42 is essential for the establishment of apical polarity in MDCK cells (24).

Knockdown of RhoA also impaired the apical orientation of GM130. In MDCK cells, RhoA is essential for the formation of tight junctions (30, 56), which segregates apical and lateral membrane domains (57). Although 16A5 cells did not form tight junctions (based on ZO1 staining; data not shown) and less than 75% of cells had an apical orientation for GM130 (Fig. 6B), the significant reduction in the apical orientation of GM130 in RhoA knockdown cells suggests that RhoA plays a role in the apical orientation of these cells. The lack of intact junctional actin rings and AJs in RhoA knockdown cells, in conjunction with the maintenance of junctional actin rings and AJs by RhoA overexpression in response to EGF stimulation in two-dimensional culture, implies that RhoA, by promoting the formation of cell-cell junctions, is critical for the orientation of apical polarity. The inability to maintain cell-cell junctions is likely to contribute to the exaggerated phenotype induced by Vav2-Y172F expression in RhoA knockdown cells. More detailed analyses of how these Rho GTPases and their effectors regulate apicobasolateral polarity in the cell system described here should help reveal the coordinated mechanisms that control MEC morphogenesis and tumorigenic phenotypes.

Notably, knockdown of Rac1 or Cdc42 antagonized the effects of Vav2-Y172F expression but did not affect the activity of the other GTPase in Vav2-Y172F-expressing MECs (Fig. 10C), indicating that each one of these proteins plays a role in Vav2-induced disruption of acinar morphogenesis. Combined knockdown studies will be required to assess if they play an additive/synergistic role or if they influence each other's function in the context of acinar morphogenesis.

It is instructive to view the present observations of the down-modulation for the EGFR and Rac1/Cdc42 signaling axis during acinar morphogenesis of MECs in the context of 1) our previous results, where EGFR overexpression in 16A5 and MCF10A MECs, especially when combined with clinically observed c-Src overexpression, induced a loss of polarity and acinar architecture as well as invasive behavior in three-dimensional cultures (18); 2) the findings that EGFR is often overexpressed in breast cancer cells and associated with invasion (41, 42); and 3) the observation that abnormal Rac1 and Cdc42 activation contributes to polarity disruption in epithelial cells and breast cancer cells (28, 54). Collectively, it appears reasonable to suggest that activation of Rac1/Cdc42 plays an important role in the disruption of MEC architecture and morphogenesis and that aberration in this signaling axis may contribute to oncogenesis during the development of breast cancer. Interestingly, overexpression of the Rac-GEF Tiam-1 has been reported in breast cancer cell lines and patient samples (58). We have detected Vav2 protein expression in a number of breast cancer cell lines that express EGFR or overexpress ErbB2 and, in some cases, express constitutively active PI3K mutants (data not shown). Because ErbB receptors and PI3K are known to activate Vav2 (48), it is likely that Vav2 may play a role downstream of ErbB receptors, in collaboration with PI3K mutations, to activate Rac/Cdc42 signaling. Indeed, induced overexpression of WT Vav2 resulted in EGF-dependent disruption of acinar architecture in 16A5 cells. Future studies should help elucidate a role for Vav2 in the regulation of cellular architecture and oncogenesis in cancer cells.

Our findings that Vav2 activates Rac1 and Cdc42 but not RhoA in MECs are supported by similar recent observations in pancreatic cancer cells (47). In conjunction with the differential down-regulation of Rac1/Cdc42/RhoA activation during MEC acinar morphogenesis, these results suggest that Vav2 may play a more important role in activating Rac1/Cdc42 in MECs, whereas RhoA activation downstream of EGFR probably involves other GEFs. However, Vav2 has been previously characterized as a GEF for RhoA, Rac1, and Cdc42; therefore, other factors, such as induction of Rho-family member-specific inhibitors, may also need to be considered in future studies.

Consistent with our results, a recent study of mammary branching morphogenesis using primary, organotypic three-dimensional culture of mouse mammary epithelia has found that Rac mediates duct initiation, whereas activation of Rock, a RhoA effector, restores epithelial architecture (59). Moreover, c-MET-induced scattering of MDCK cells was blocked by RhoA activation (35, 37), whereas spatially restricted degradation of RhoA was shown to facilitate transforming growth factor-β-mediated disruption of tight junctions in MECs (38), further supporting a general role for RhoA in the maintenance of normal epithelial cell architecture. In the 16A5 MEC system used here, RhoA does not directly affect Vav2 activity because Vav2-induced activation of Rac1 is unaffected by knockdown of RhoA, suggesting that RhoA signaling antagonizes Rac1/Cdc42 at a step after their initial activation. One potential mechanism by which RhoA maintains epithelial cell polarity may be through the inactivation of cofilin, an actin-severing protein (60), which could possibly occur through the activation of both Rho kinase and LIM-kinase, the latter an inactivator of cofilin (61). Future studies will be necessary to explore the role of cofilin and other factors that could act as molecular switches between acinar and branching morphogenesis orchestrated by RhoA and Rac1/Cdc42.

When analyzing the DOX-inducible MEC system, we noted that the YFP-Vav2-Y172F protein accumulated and displayed higher fluorescence intensity near the apical side and within the lumina of a large proportion of acini after induction for 7 days. Because the lumen had already formed within acini when Vav2 expression was induced at day 14 (supplemental Fig. 1D), it is likely that Vav2-Y172F expression caused inward cell growth and partial refilling of the lumen. Staining of the acini for activated caspase-3 showed that although the vehicle-treated cells had positive caspase-3 staining in the lumen, the filled lumen in acini with YFP-Vav2-Y172F induction did not (supplemental Fig. 1C), implying that Vav2 may mediate antiapoptotic signaling. Some of the acini accumulated green YFP-Vav2 proteins in the lumen without positive nuclear staining. Whether or not Vav2 is toxic to the cells and caused non-apoptotic cell death with nuclear disappearance and exudation of Vav2 proteins to the lumen requires further investigation. Future studies will also focus on which antiapoptotic signaling pathways are regulated by Vav2 and whether such signaling pathways are an integral part of the Rac1/Cdc42 signaling axis in acinar morphogenesis.

In conclusion, our studies, utilizing a three-dimensional culture system of mammary epithelial acini together with inducible gene expression and knockdown strategies, reveal that the RhoA and Rac1/Cdc42 signaling pathways play disparate and apparently antagonistic roles in the remodeling of epithelial cell architecture and morphogenesis. These results suggest a potential role for the Vav2-Rac1/Cdc42 signaling pathway in RTK-mediated disruption of MEC architecture and breast tumorigenesis. Further biochemical and cell biological analyses of MECs and other epithelial cell systems should facilitate a better molecular understanding of the biological roles of Vav2 and the counterbalancing roles of the Rho family GTPases reported here.