Myosin II isoform switching mediates invasiveness after TGF-β–induced epithelial–mesenchymal transition

Abstract

Despite functional significance of nonmuscle myosin II in cell migration and invasion, its role in epithelial–mesenchymal transition (EMT) or TGF-β signaling is unknown. Analysis of normal mammary gland expression revealed that myosin IIC is expressed in luminal cells, whereas myosin IIB expression is up-regulated in myoepithelial cells that have more mesenchymal characteristics. Furthermore, TGF-β induction of EMT in nontransformed murine mammary gland epithelial cells results in an isoform switch from myosin IIC to myosin IIB and increased phosphorylation of myosin heavy chain (MHC) IIA on target sites known to regulate filament dynamics (S1916, S1943). These expression and phosphorylation changes are downstream of heterogeneous nuclear ribonucleoprotein-E1 (E1), an effector of TGF-β signaling. E1 knockdown drives cells into a migratory, invasive mesenchymal state and concomitantly up-regulates MHC IIB expression and MHC IIA phosphorylation. Abrogation of myosin IIB expression in the E1 knockdown cells has no effect on 2D migration but significantly reduced transmigration and macrophage-stimulated collagen invasion. These studies indicate that transition between myosin IIC/myosin IIB expression is a critical feature of EMT that contributes to increases in invasive behavior.

Normal mammary tissue consists of a branched multilayer ductal network residing in an expanse of adipocytes. The inner luminal epithelial layer is a cuboidal epithelium that is surrounded by an outer myoepithelial or basal cell layer that displays mesenchymal-like features, including a spindle-shaped morphology and expression of markers such as α-smooth muscle actin (SMA) and vimentin. During transition to a tumor state, breast epithelial cells characteristically exhibit features of either luminal or basal cell types. Thus, basal-like breast carcinomas are defined by having a gene expression profile similar to basal, or myoepithelial cells (1). In addition, basal-derived tumors are typically more invasive and metastatic than luminal-derived carcinomas (2).

Tumor cell metastasis is a process that includes migration to and intravasation of the vasculature, followed by extravasation and migration into the distant tissue to form a secondary tumor. One of the initial steps in metastasis is thought to be the process of epithelial–mesenchymal transition (EMT). During EMT, a polarized epithelial cell breaks down E-cadherin–based cell–cell contacts and acquires migratory and invasive properties, together with changes in gene and protein expression patterns (3). TGF-β is a known inducer of EMT (4) that signals through both Smad (5) and non-Smad pathways, including PI3K/Akt (6). Recent work has shown heterogeneous nuclear ribonucleoprotein-E1 (hnRNP-E1; hereafter referred to as E1) to be a downstream effector of the TGF-β–Akt2 pathway (7). E1 regulates translation of a number of critical EMT transcripts, including Dab2 and ILEI (7). Attenuation of E1 expression in epithelial cells induces EMT and increases metastatic capability (8).

Although the metastatic process can be separated into individual steps, one accepted theme is that tissue invasion requires cytoskeletal force generation. How cytoskeletal forces drive the mechanical process of invasion is not understood. Multiple migratory modes have been suggested, including amoeboid and mesenchymal, and some studies have suggested that cells can switch between migration modes depending on the extracellular environment (9). Recent work indicates that nuclear translocation can be a rate-limiting step during amoeboid 3D migration (10, 11). Others studies have shown that contraction of the cell rear is absolutely necessary for cancer cell invasion (12). Nonmuscle myosin II has been suggested to be involved in both of these processes.

The myosin II holoenzyme consists of two myosin heavy chains (MHC), two essential light chains, and two regulatory light chains. In mammals, three different genes encode nonmuscle MHC II proteins, which are named MHC IIA (Myh9), IIB (Myh10), and IIC (Myh14). These isoforms have both unique and redundant roles in vivo (13).

MHC phosphorylation has clear roles in regulating filament assembly in the model organism Dictyostelium discoideum (14). However, the contribution of heavy chain phosphorylation to mammalian myosin filament assembly remains less well understood. MHC IIA is phosphorylated on S1916, a putative PKC target (15), and S1943, a putative casein kinase II target (16). In vitro studies with recombinant MHC tail domains show that heavy chain phosphorylation shifts the monomer/filament equilibrium into the monomeric, disassembled state, suggesting a potential inhibitory role for heavy chain phosphorylation (17). However, recent studies in live cells suggest a model in which heavy chain phosphorylation is required for myosin IIA recycling from distal to anterior regions of the lamellum (18), where it can contribute to focal adhesion stability and maturation (19). Other studies have suggested that myosin IIA heavy chain phosphorylation increases breast cancer cell migration rates (20). Despite myosin II having roles in migration and invasion, and TGF-β treatment clearly leading to a more migratory and invasive phenotype in the context of EMT, the regulation of myosin II expression or phosphorylation by TGF-β signaling has not been examined.

In this study we show that myosin IIB expression and myosin IIA heavy chain phosphorylation are dramatically increased after TGF-β–induced EMT in mammary epithelial cells. Inhibition of myosin IIB expression in post-EMT mesenchymal cells reduces transmigration and invasion. These data indicate that shifts in myosin II isoform expression and possibly MHC IIA phosphorylation are essential for mediating mammary epithelial cell migration and invasion.

Results

Myosin Heavy Chain Isoforms Are Differentially Expressed in Normal Mouse Mammary Gland and Breast Epithelial Cell Lines.

To investigate the expression pattern of the myosin II isoforms in native mammary gland we immunostained sections of mouse mammary gland with markers for the luminal layer (cytokeratin 8, or K8) or myoepithelial layer (SMA). Myosin IIA and myosin IIC expression was primarily restricted to the luminal cell layer (Fig. 1A, Top and Bottom, respectively), whereas myosin IIB showed minimal expression in luminal cells but was prominently expressed in the myoepithelial layer (Fig. 1A, Middle).

Fig. 1.

MHC isoforms are differentially expressed in normal mouse mammary gland and breast epithelial cell lines. (A) Normal mouse mammary tissue was fixed, sectioned, and immunostained for MHC IIA (Top, red), MHC IIB (Middle, green), or MHC IIC (Bottom, red). The sections were coimmunostained with smooth muscle actin to mark the basal layer (Top and Bottom, green) or keratin-8 to mark the luminal layer (Middle, red). (Scale bar, 100 μm.) (B) Whole-cell lysates of the indicated cell lines were subjected to Western blot analysis with the indicated antibodies.

Given the observed differences in isoform distribution in the normal mammary gland, we sought to determine whether this correlated with differential isoform expression in breast cancer and breast epithelial cell lines that display luminal vs. basal characteristics as defined by transcriptome profiling (21). Analysis of a set of seven breast epithelial cell lines revealed elevated MHC IIB expression in three lines that have been previously defined as “basal” and very low levels of in MHC IIB expression in four lines defined in the same studies as “luminal” (Fig. 1B). MHC IIC was expressed in all four luminal cell lines but was absent in two of the three basal lines (Fig. 1B). In addition, the two cell lines that lack MHC IIC are “basal B” lines (MDA-MB-231 and MCF-10A), which are more consistently mesenchymal in their transcript profiles than the “basal A” line (MBA-MB-468) that has mixed basal/luminal properties (Fig. 1B) (22). Overall, these expression patterns generally match the distribution profile observed in the native tissue, suggesting that MHC IIC expression may be a marker for the luminal layer of ductal tissue and that MHC IIB expression is a part of the basal or myoepithelial gene expression program.

MHC IIA Phosphorylation Patterns in Breast Epithelial Cell Lines.

MHC IIA expression did not seem to correlate strongly with either luminal or basal breast cancer cell lines. To determine whether MHC IIA phosphorylation is differentially regulated in luminal vs. basal cell lines, we took advantage of our recently developed phospho-specific antibodies directed against two phosphorylated residues in the MHC IIA tail region, S1916 and S1943 (23). Western blot analysis revealed strongly elevated MHC phosphorylation at both target sites in all three basal-like cell lines, relative to the luminal lines (Fig. 1B). Given earlier analyses demonstrating that S1943 phosphorylation regulates myosin IIA filament dynamics at the leading edge of cells during polarized migration (18, 20), we suggest that elevated phosphorylation of S1943 and S1916 may have a role in enhancing the motile behavior of basal-like cells.

Myosin Heavy Chain Isoform Switching During EMT.

Given the preferential expression of myosin IIB and the elevated MHC IIA phosphorylation in basal-like lines, we asked whether luminal-like mammary epithelial cells, induced to undergo EMT by TGF-β treatment, would drive similar changes in myosin II expression. To test this possibility, we used the normal murine mammary gland (NMuMG) cell EMT model (24), which under normal tissue culture conditions forms a classic epithelial sheet with a “cobblestone” morphology (Fig. S1A). Upon treatment with TGF-β for 24 h, the cells convert to a spindle-shaped mesenchymal morphology (Fig. S1A). This morphological switch is accompanied by extensive transcriptional and phenotypic changes (25), including the induction of a number of classic mesenchymal cell markers, such as vimentin (Fig. 2A). Whereas MHC IIA showed a small but significant twofold increase at the mRNA and protein level during EMT (Fig. 2 A and D), MHC IIB protein expression was up-regulated dramatically at the 24-h time point (Fig. 2A). Quantitative PCR analysis showed that this is due to an ≈3,000-fold increase in mRNA level (Fig. 2D). Accompanying the dramatic increase in MHC IIB expression, an opposing decrease in mRNA abundance (approximately fivefold) and protein expression of MHC IIC was observed (Fig. 2 A and D).

Fig. 2.

MHC regulation during TGF-β and E1-shRNA treatment of NMuMG cells. (A) NMuMG, (B) E1-shRNA, and (C) E1⁺⁺ cells were untreated or treated with TGF-β for the indicated time. Whole-cell lysates were collected, normalized, and probed via Western blot with the indicated antibodies. Vimentin represents a positive control for EMT, and actin and GAPDH are shown as loading controls. (D) NMuMG cells were treated with TGF-β for the indicated time, after which RNA was collected and quantitative PCR was performed for the indicated myosin isoform. Data are the mean ± SEM from four independent experiments. *P < 0.05, Student’s t test, relative to the untreated. (E) RNA was collected from NMuMG, E1-shRNA, and E1⁺⁺ cells and quantitative PCR was performed for the indicated myosin isoform. For each cell line, relative isoform transcript level is expressed as a percentage of the total transcript abundance of all three isoforms. Data are the mean ± SEM. n = 3.

Multiple efforts have been made to identify targets of TGF-β treatment using expression microarrays. In agreement with our data, MHC IIA and MHC IIC reside within a list of genes that are up-regulated and down-regulated, respectively, during EMT (26–29). MHC IIB was previously included in a list of genes up-regulated in a lung epithelial cell EMT model (28). The present work shows that MHC IIB expression is also regulated in mammary epithelial cell EMT driven by TGF-β. We also confirm these expression changes at the protein level.

MHC IIA Phosphorylation Is Elevated During EMT.

To further investigate myosin regulation during EMT, we analyzed the same TGF-β–treated NMuMG time course for MHC IIA phosphorylation. Untreated NMuMG cells maintain low levels of MHC IIA phosphorylation at S1916 and undetectable phosphorylation at S1943. However, upon treatment with TGF-β there is a dramatic increase in MHC IIA phosphorylation, especially at S1943 (Fig. 2A). Interestingly, no increase in MHC IIA phosphorylation was observed until 12 h after the initiation of TGF-β treatment, with a more robust increase occurring after 24 h. The kinetics of these phosphorylation changes suggest regulation through TGF-β–mediated transcriptional or translational events, as opposed to immediate signaling downstream of TGF-β. The kinetics also correlate with previously reported times for increased migratory and invasive behavior by the post-EMT mesenchymal cells (30). Taken together these data further support the concept that expression of myosin IIB, and possibly elevated MHC IIA tail phosphorylation, are consistent features of a more motile phenotype, possibly relevant to the invasive behavior of cells that have undergone EMT.

E1 Modulation Alters Myosin Expression and Phosphorylation in NMuMG Cells.

Recent work has revealed E1 as a critical downstream effector of TGF-β treatment in NMuMG cells (7). E1 is a 3′UTR mRNA binding protein that represses translation of its targets. shRNA-mediated knockdown of E1 induces a mesenchymal state, whereas overexpression of E1 locks cells in an epithelial state that is resistant to TGF-β–stimulated EMT (7) [Fig. 2 B and C (see vimentin) and Fig. S1A]. E1 thus seems to be a master regulator of many downstream EMT events.

To determine whether changes in myosin expression and phosphorylation observed in TGF-β–treated NMuMG cells are downstream of E1, we analyzed cells with abrogated E1 expression (E1-shRNA) and cells overexpressing E1 (E1⁺⁺) during a TGF-β treatment time course. E1-shRNA cells showed the same profile of myosin expression and phosphorylation as NMuMG cells treated with TGF-β, namely strikingly increased MHC IIB expression, increased MHC IIA phosphorylation, and decreased MHC IIC expression (Fig. 2B). The effect of E1 shRNA on MHC IIB expression was confirmed using two additional E1 shRNA target sequences (Fig. S1B). This myosin profile was not altered upon treatment with TGF-β, supporting previous data demonstrating that E1 mediates the EMT-inducing effects of TGF-β. In contrast, E1⁺⁺ cells displayed a similar myosin isoform expression profile to the parental NMuMG cells, and this profile was not affected by TGF-β treatment (Fig. 2C). These data demonstrate that the myosin isoform switching and phosphorylation changes seen in TGF-β–treated cells are downstream of E1 and that manipulation of E1 expression has profound effects on myosin isoform and phosphorylation patterns independent of TGF-β addition. Given the large increase in MHC IIB transcript level observed upon TGF-β treatment (Fig. 2D) or in E1-shRNA cells (Fig. 2E), we suggest that changes in MHC IIB expression are mediated by transcriptional events downstream of E1, rather than via direct silencing of MHC IIB mRNA by E1. This contrasts with the translational repression seen in previously reported direct targets of E1 (7).

To quantify MHC II isoform switching downstream of E1, we performed mass spectrometric analysis to determine the relative molar abundance of MHC IIA, IIB, and IIC in E1⁺⁺ and E1-shRNA cell lines. In E1⁺⁺ cells, ≈99% of the MHC II detected was the IIA isoform, and ≈1% was the IIC isoform. No MHC IIB was detected. In contrast, in E1-shRNA cells, MHC IIC peptides were not detected by mass spectrometry, and the remaining MHC II peptides detected revealed a ratio of ≈90% MHC IIA and ≈10% MHC IIB. These data are consistent with mRNA levels in the cells (Fig. 2E). On the basis of the relatively low levels of MHC IIC present in the epithelial state and the relatively greater fold induction of MHC IIB expression upon EMT, we focused our analysis on the roles of MHC IIB in post-EMT mesenchymal phenotypes.

MHC IIB shRNA Inhibits Transmigration and Invasion.

In 2D migration assays, E1-shRNA cells are ≈50% more migratory compared with E1⁺⁺ cells (Fig. S2 A and B). Additionally, whereas E1⁺⁺ cells migrate as epithelial sheets, the E1-shRNA cells display properties more characteristic of mesenchymal cell migration, with little or no apparent cell–cell contacts and individual cells often emerging at the leading edge. In contrast to the 2D migration assay in which the E1-shRNA cells are slightly more migratory than the E1⁺⁺ cells, transwell assays revealed that E1-shRNA cells have a nearly 40-fold higher propensity for transmigration (Fig. S2C). Thus, E1-shRNA treatment only slightly increases the rate at which epithelial cells migrate in 2D but dramatically increases their ability to squeeze through small pores, a surrogate for invasive behavior. Taken together these data suggests that E1-shRNA cells recapitulate a mesenchymal-like phenotype and can be used as a model to further our understanding of the mechanics of mesenchymal cell invasive behavior.

To determine whether the increased expression of MHC IIB is important for the increased migration and invasion that occurs after E1 silencing, we used lentiviral-mediated shRNA to suppress MHC IIB expression in E1-shRNA cells (Fig. S3). Knockdown of MHC IIB did not affect mesenchymal morphology, significantly alter E1 shRNA cell migration in 2D, or alter expression of other EMT markers (Fig. 3A and Fig. S3). However, MHC IIB shRNA did significantly reduce the cells ability to migrate through 8-μm pores (Fig. 3B). Similar results were obtained when MHC IIB expression was suppressed in MDA-MB-231 cells using an shRNA construct with a different target sequence (Fig. S4). MHC IIB abrogation also inhibited E1-shRNA cells’ ability to invade through Matrigel-coated transwell chambers (Fig. S5). To complement these results, we examined the requirement for MHC IIB in a macrophage-induced invasion assay. As previously reported (31), MDA-MB-231 invasion into a collagen matrix can be potently enhanced via a macrophage-dependent CSF–EGF paracrine loop (Fig. 3C). Coculture of E1-shRNA cells with macrophages increased the number of invasive tumor cells by ≈3.5 fold, whereas knockdown of MHC IIB completely abolished macrophage-stimulated invasion (Fig. 3C). These results demonstrate a critical role for myosin IIB in 3D invasion in multiple settings.

Fig. 3.

MHC IIB shRNA inhibits transmigration. (A and B) 2D migration and transwell assays were performed on uninfected, scramble shRNA cells or MHC IIB shRNA cells. Data were normalized to the mean of scramble shRNA cells. Data points are the mean ± SEM. For B, n ≥ 28 fields of view from three independent experiments. For C, n ≥ 11 fields of view from three independent experiments. *P < 0.001 relative to scramble shRNA cells. (C) Scramble shRNA and IIB shRNA cells were tested for their ability to invade a collagen gel in the presence (black bars) or absence (gray bars) of BAC macrophages. MDA-MB-231 cells were used as a positive control. For D, n ≥ 12 regions analyzed from five experiments performed on three independent dates. *P < 0.001 relative to invasion in the absence of BACs. Data points are the mean ± SEM.

To dissect the mechanistic roles of myosin IIB in 3D migration, we immunostained migrating E1-shRNA cells for MHC IIA and MHC IIB in the 2D setting (Fig. 4A) and in the 3D-like setting of transwell migration (Fig. 4 B and C). Whereas myosin IIA appears preferentially localized to the anterior lamellar regions in 2D (Fig. 4A, yellow arrowheads), myosin IIB is enriched in the posterior retraction zones (Fig. 4A, blue arrows). This staining pattern is in agreement with previous studies in other cell types reporting a tendency for myosin IIB to reside in the posterior region of polarized cells relative to myosin IIA during 2D migration (32). However, the forces required for transmigration and 3D migration are quite different from the forces required for 2D migration (10, 33) and may result in altered myosin isoform localization that is not seen in 2D migration. To determine the localization of myosin IIA and myosin IIB in transmigrating cells, we fixed, immunostained, and imaged E1-shRNA cells as they were migrating through a transwell membrane. Confocal z-stacks were collected from individual cells and imaged from each side of the transwell membrane. 3D reconstruction of cells actively undergoing transmigration shows that myosin IIB is highly enriched in the rear of the cell (Fig. 4 B and C, blue arrows, and Fig. S6), especially around the nucleus that is just beginning to enter the pore. We hypothesize, on the basis of this posterior and perinuclear localization, that myosin IIB may specifically facilitate squeezing the cytoplasm and nucleus through tight spaces, an essential step in both intravasation and extravasation. In contrast, although myosin IIA appears more assembled in the anterior regions of the cell that has already passed through the pore (Fig. 4B, yellow arrowheads), quantitative assessment of myosin IIA localization suggests that it is not polarized in this assay (Fig. S6). These data show that myosin IIB, unlike myosin IIA, remains polarized during transmigration, with enrichment specifically in perinuclear regions.

Fig. 4.

MHC IIB localizes to the rear and perinuclear regions of migrating and transmigrating E1-shRNA cells. (A) E1-shRNA cells migrating in 2D were fixed and immunostained for MHC IIA (Top, green) and MHC IIB (Middle, red). The white arrow in the lower image indicates the direction of migration. (B) E1-shRNA cells actively moving through a transwell pore were fixed and immunostained for MHC IIA (green) and MHC IIB (red) and counterstained with DAPI (blue). Confocal z-stacks were collected. Images are orthogonal maximum intensity projections. Bottom: Merge 2 is a merge of DAPI, MHC IIA, and MHC IIB. Merge 1 excludes DAPI. The dotted line in merge 2 represents the position of the transwell membrane. (C) Same dataset used in Fig. 4B was separated into images above the membrane (Upper) and below the membrane (Lower). Each separated dataset was then averaged into a single image. (Scale bars, 20 μm.)

Discussion

The studies presented here report myosin II isoform expression patterns in normal mammary gland tissue, and demonstrate that isoform expression patterns in breast epithelial cell lines parallel the expression patterns in native luminal vs. basal tissue layers (e.g., expression of myosin IIC in luminal tissues and cell lines and expression of myosin IIB in basal tissue and cell lines). Considering the biophysical properties of myosin II isoforms, it is intriguing that the myosin IIB isoform is preferentially expressed in the myoepithelial cell layer, which also expresses many proteins involved in smooth muscle cell function. Of the three nonmuscle myosin II isoforms, myosin IIB has the highest duty ratio (34) and the highest affinity for ADP (35), suggesting that it may be tuned for prolonged force production and minimal energy consumption. Previous work has shown that myosin IIB contributes to force maintenance in aortic smooth muscle (36). The myoepithelial layer of mammary gland tissue is responsible for the contractile events during milk secretion (37) and has been suggested to be critical for proper branching architecture (38), an essential component of mammary gland development. We suggest that myosin IIB may be selectively expressed in the myoepithelial layer to contribute to the prolonged contractile events that are involved in milk secretion and branching morphogenesis.

We suggest that up-regulation of MHC IIA phosphorylation during EMT may have a role in the enhanced motility and invasiveness of mesenchymal cells. We favor a model in which heavy chain phosphorylation permits redistribution of myosin II from posterior to anterior regions, where it can facilitate adhesion stabilization and maturation, a concept supported by our recent studies using site-directed mutants of these target sites (18). Given the potential role for integrin-based adhesions in transmigration, it is logical that modulation of MHC IIA phosphorylation could alter cell invasion capability. Future studies are needed to explore this possibility. Additionally, MHC IIB phosphorylation has established roles in modulation of filament assembly in prostate cancer cells (39). Further studies directed at understanding MHC IIB phosphorylation events in the context of EMT and invasive behavior will be of great interest.

In our system, the up-regulation of MHC IIB mRNA levels after TGF-β treatment or E1-shRNA treatment suggests that there are transcription factors downstream of E1 that directly regulate MHC IIB expression. To date, few studies have analyzed the promoter region of MHC IIB in great detail. Those studies that have addressed this subject suggest possible regulation through the transcription factors KLF5 (40) and Hex (41), both of which have been shown to be downstream effectors of TGF-β family signaling (42, 43). Therefore, although there are interesting candidates for regulation of MHC IIB expression in this system, the precise molecular mechanism by which TGF-β and E1 regulate MHC IIB mRNA requires further investigation.

TGF-β–stimulated EMT involves induction of a large number of genes (26), many of which contribute to the acquisition of invasive behavior. Our work demonstrates induction of MHC IIB as a critical component of this transition. Interestingly, depletion of myosin IIB in the mesenchymal E1-shRNA cells has no effect on 2D migration, a moderate effect on transmigration through 8-μm pores, and a very strong negative impact on migration through dense (5.8 mg/mL) collagen gels in the macrophage-stimulated migration assay. We suggest that induction of myosin IIB when cells undergo EMT enhances their force-generating capacity for invasive migration and that this effect becomes more apparent the greater the resistance of the external environment is.

This study localized myosin IIA and IIB isoforms imaged in a setting where cells are migrating in a 3D-like setting in which the cell is experiencing significant mechanical resistance from its environment. We suggest that the posterior localization of myosin IIB in this setting may be critical for helping squeeze the nucleus through the transwell pore and that migration through dense collagen gels represents an even more demanding migration setting, such that depletion of myosin IIB by shRNA completely abrogates invasiveness. This model is consistent with earlier studies implicating myosin II as contributing to nuclear translocation when cells are migrating through narrow spaces (33, 11, 10). If this model proves valid with further studies, myosin IIB may prove to be an outstanding therapeutic target for intervention in invasive metastatic behavior of basal-like carcinomas.

Taken together, our data demonstrate a clear role for myosin IIB as contributing to invasive potential in mammary epithelial cells that have undergone EMT. Our data also demonstrate a consistent induction of myosin IIA heavy chain phosphorylation in post-EMT NMuMG cells and elevated phosphorylation of the same sites in basal-like breast epithelial lines. We hypothesize that MHC IIA phosphorylation and myosin IIB expression may both be critical components in the set of global changes that induce cellular invasion during EMT. Further studies are needed to test the roles of MHC IIA phosphorylation in invasive migration. We suggest that fluxes in myosin II expression and phosphorylation may also have roles in breast epithelial cell metastasis and possibly in developmental settings where EMT is essential for proper organogenesis.

Methods

Cell Culture and Treatments.

NMuMG, E1-shRNA, and E1⁺⁺ cells were maintained in DMEM with 10% FBS, 10 μg mL⁻¹ insulin, and antibiotic/antimycotic and, where indicated, were treated with 5 ng mL⁻¹ of TGF-β2. MDA-MB-231 cells were maintained in RPMI with 10% FBS and antibiotic/antimycotic. A list of shRNA target sequences is provided in Table S1.

Mammary Gland Immunohistochemistry.

Mammary glands were isolated from female wild-type C57BL/6 mice at ≈5 wk of age, fixed, embedded in paraffin, sectioned, and immunostained with the indicated antibodies.

2D Migration Assays.

Migration assays were performed as described previously (18).

Transwell Migration Assays.

Cells were plated in 12-well 8-μm-pore PET transwell membranes (Falcon) or CytoSelect 96-well Cell Invasion Assay Kit (Cell Biolabs), allowed to migrate for 18 h or 24 h, respectively, fixed and stained with DAPI. Fields of view were randomly imaged, and the number of cells present was quantified using ImageJ software. Additional details for imaging cells during transmigration can be found in SI Methods.

Collagen Invasion Assay.

The collagen invasion assay was performed as described previously (31).

Additional methods details can be found in SI Methods.