Research Article

Regulation of the Polarity Protein Par6 by TGFß Receptors Controls Epithelial Cell Plasticity

See allHide authors and affiliations

Science  11 Mar 2005:
Vol. 307, Issue 5715, pp. 1603-1609
DOI: 10.1126/science.1105718

Abstract

The transition of cells from an epithelial to a mesenchymal phenotype is a critical event during morphogenesis in multicellular organisms and underlies the pathology of many diseases, including the invasive phenotype associated with metastatic carcinomas. Transforming growth factor β (TGFβ) is a key regulator of epithelial-to-mesenchymal transition (EMT). However, the molecular mechanisms that control the dissolution of tight junctions, an early event in EMT, remain elusive. We demonstrate that Par6, a regulator of epithelial cell polarity and tight-junction assembly, interacts with TGFβ receptors and is a substrate of the type II receptor, TβRII. Phosphorylation of Par6 is required for TGFβ-dependent EMT in mammary gland epithelial cells and controls the interaction of Par6 with the E3 ubiquitin ligase Smurf1. Smurf1, in turn, targets the guanosine triphosphatase RhoA for degradation, thereby leading to a loss of tight junctions. These studies define how an extracellular cue signals to the polarity machinery to control epithelial cell morphology.

Epithelial-to-mesenchymal transition (EMT) is a striking example of cellular plasticity that involves the dissolution of epithelial tight junctions, modulation of adherens junctions, reorganization of the actin cytoskeleton, loss of apical-basal polarity, and induction of a mesenchymal gene-expression program (1, 2). The loss of epithelial homeostasis and the resultant acquisition of a migratory, mesenchymal phenotype are also essential for tumor invasion. Transforming growth factor β (TGFβ) has emerged as a key regulator of EMT in late-stage carcinomas, where it promotes invasion and metastasis (3, 4). TGFβ signals through type I and type II transmembrane receptor serine-threonine kinases (TβRI and TβRII, respectively). In the classically described Smad pathway (58), constitutively active TβRII phosphorylates and activates TβRI, which in turn phosphorylates receptor-regulated Smads. These then bind to Smad4 and accumulate in the nucleus, where they regulate gene transcription by interacting with DNA binding partners. Although both Smad-dependent and Smad-independent mechanisms have pivotal roles in regulating TGFβ-dependent acquisition of a mesenchymal phenotype, the contribution of nontranscriptional signaling conduits to the regulation of epithelial cell plasticity remains less well characterized.

Par6 interacts with the TGFb receptor. Using LUMIER (9), a luminescence-based assay for protein-protein interactions in mammalian cells, we performed a screen to identify interaction partners for TβRI. Briefly, in a 96-well format, human embryonic kidney (HEK293T) cells were transfected with a panel of FLAG-tagged prey proteins in conjunction with TβRI fused to Renilla luciferase (TβRI-Rluc). Lysates were subjected to immunoprecipitation with an antibody to FLAG (anti-FLAG) and assayed for luciferase activity to detect putative interactions with the receptor. In a preliminary screen, known interacting partners of TβRI, such as FKBP12 and TβRI itself as well as serine-threonine kinase receptor–associated protein (STRAP), were confirmed as interactors (Fig. 1A). This screen also identified partitioning-defective protein 6C (Par6C). Par6 is a component of a metazoan polarity complex that was originally characterized in Caenorhabditis elegans as a determinant of asymmetric cell cleavage (10). More recently, Par6 has been proposed to function as a scaffold for the assembly of a protein complex that regulates cell polarity and includes the Rho guanosine triphosphatase (GTPase) Cdc42, atypical protein kinase C ζ (PKCζ), and the PDZ domain–containing adapter Par3 (11, 12). Moreover, Par6 associates with components of additional polarity complexes, including Stardust-Crumbs-PATJ (1315) and DLG-SCRIB-LGL (16, 17). Par6 associates with proteins in these complexes through interaction domains that include a PB1 domain, a pseudo-CRIB (Cdc42/Rac interactive binding) motif, and a PDZ domain. Although Par6 regulates the assembly of tight junctions and the apical-basal polarity of mammalian epithelial cells (18), the mechanism for this action is unknown.

Fig. 1.

Binding of Par6 to TβRI. (A) LUMIER screen (9) of Renilla luciferase–tagged TβRI against a panel of FLAG-tagged proteins. Yellow represents the relative interaction of TβRI against the indicated protein (9). (B) Interaction of endogenous Par6 and TβRI. Lysates from Mv1Lu incubated with (+) or without (–) TGF β were subjected to immunoprecipitation with antibody to TβRI (α-TβRI) or control antibody (Co) followed by immunoblotting with antibody to Par6 (α-Par6). (C) Interaction of Par6 with TβRI but not with TβRII. Lysates from HEK293T cells expressing FLAG-Par6 (F-Par6), TβRI-HA, or TβRII-HA, as indicated, were subjected to anti-FLAG immunoprecipitation (IP) followed by immunoblotting (IB) with anti-FLAG or antibody to HA. Protein expression was confirmed by immunoblotting total lysate. (D) Interaction of Par6 with TGFβ receptor complexes at the cell surface. HEK293T cells transfected with Par6 and either wild-type (WT) or KR versions of TβRI and TβRII were affinity labeled, and Par6 bound to receptors was detected by autoradiography, as described (49). Protein amounts were determined as in (B). (E) A schematic of the strategy for detecting cell-surface receptors. (F) Differential localization of cell-surface TβRI and TβRII in NMuMG cells. HA-TβRII (red) or MYC-TβRI (blue) were detected as described (49), and tight junctions were visualized by ZO-1 staining (green). (G) Localization of cell-surface TβRI (green), Par6 (red), and ZO-1 (blue) was determined as described (49). Z-stacks are shown to highlight the tight junctions. (H) Localization of cell-surface TβRI (green) and TβRII (red) in tight junctions in NMuMG cells untreated or treated with TGFβ for 30 min, as indicated. (I) Z-stacks demonstrate TGFβ-dependent redistribution of TβRII to tight junctions.

TGFβ induces the loss of tight junctions during EMT (19). Therefore, we sought to investigate whether Par6 functions in this pathway. Immunoprecipitation and immunoblotting revealed association of endogenous Par6 with endogenous TβRI in mink lung epithelial (Mv1Lu) cells and demonstrated that Par6 remained bound to TβRI in both the absence and the presence of TGFβ (Fig. 1B). This interaction was specific for TβRI, because we were unable to detect any interaction between wild-type TβRII and Par6 when TβRII was expressed alone (Fig. 1C). We used affinity labeling with [125I]TGFβ to show association of TGFβ receptor complexes from the cell surface with immunoprecipitated Par6, as well as interactions with receptor complexes containing either wild-type or kinase-deficient (KR) TβRII or TβRI (Fig. 1D) (20). Thus, Par6 can associate with the TGFβ receptor complex by interaction with TβRI, independently of receptor kinase activity.

To map the region of Par6 that interacts with the TGFβ receptor, we generated a series of Par6 truncation mutants. The N-terminal 104 amino acids of Par6 were necessary for binding to TβRI (fig. S1A). This region of Par6 contains a PB1 domain, which mediates interaction with the PB1 domain of its partner protein, PKCζ (11, 12). Thus, we explored the relation between Par6 binding to TβRI and PKCζ. A mutant of Par6, Par6(K19A), which interferes with the PKCζ binding surface of the PB1 domain (21), bound to TGFβ receptors in amounts comparable to those of wild-type Par6 (fig. S1B). However, PKCζ association with the receptor complexes was lost in the presence of Par6(K19A). Thus, the N-terminal region of Par6 is necessary for association with the receptors and for PKCζ recruitment to the receptor complex.

TβRI colocalizes with Par6 at tight junctions. Par6 is localized to tight junctions in epithelial cells (18, 22). To investigate the distribution of cell-surface TβRI, we generated a MYC epitope–tagged version of TβRI (MYC-TβRI) in which the tag is placed in the extracellular domain downstream of the signal sequence. We confirmed the functionality of MYC-TβRI by demonstrating that it induced Smad2 phosphorylation, increased Smad transcriptional activity in a TGFβ-regulated manner, and bound [I125]TGFβ in a TβRII-dependent manner (23). We examined the cell-surface distribution of MYC-TβRI and hemagglutinin (HA)–TβRII transiently expressed together in confluent normal murine mammary gland (NMuMG) monolayers by immunofluorescence confocal microscopy (Fig. 1E) (24). These cells establish apical-basal polarity in culture, form tight junctions, and, like many other epithelial cell types, undergo EMT in response to TGFβ (19). Although TβRII was predominantly localized to puncta uniformly distributed over the apical aspect of the cell (Fig. 1F) (24, 25), the surface pool of TβRI was restricted to a discreet band around the apical periphery that colocalized with Zona Occludens-1 (ZO-1) (Fig. 1F), a structural component of tight junctions. Moreover, reconstructed confocal sections along the vertical axis (Z-stack) revealed that cell-surface TβRI localized with ZO-1 on the apical aspect of the cell together with Par6 (Fig. 1G) and was situated apically to endogenous E-cadherin, a marker of adherens junctions (fig. S1C). After stimulation of cells with TGFβ, cell-surface TβRII was redistributed to the tight junctions, where it localized with both TβRI and ZO-1 (Fig. 1, H and I, and movies S1 and S2). TβRI also constitutively interacts with occludin, a structural component of tight junctions, and, upon stimulation of cells with TGFβ, TβRII is recruited to interact with occludin (11). Thus, TβRI is localized to tight junctions where Par6 is also found, and TGFβ stimulation induces redistribution of cell-surface TβRII into tight junctions, likely due to ligand-dependent binding of TβRII to TβRI.

Par6 is a substrate of the TGFb receptor complex. The interaction of Par6 with the TGFβ receptor complex and the localization of cell-surface TβRI to tight junctions suggested that Par6 might serve as a substrate of the receptor complex and be involved in TGFβ-dependent EMT. We analyzed phosphorylation of Par6 bound to endogenous TβRI in [32P]phosphate-labeled NMuMG cells. This revealed TGFβ-dependent phosphorylation of TβRI-bound Par6 in vivo (Fig. 2A). To determine whether Par6 phosphorylation required receptor kinase activity, we analyzed FLAG-Par6 phosphorylation in HEK293T cells that expressed either wild-type or KR receptor complexes. When cells are transfected with both receptors, TβRII and TβRI associate because of intrinsic affinity (26) and signal in a kinase activity–dependent manner even in the absence of ligand. In the absence of transfected receptors, phosphorylated Par6 migrated as a single species, whereas in the presence of wild-type receptors we observed a slower migrating form of Par6, which we call Par6* (Fig. 2B). Phosphoamino acid analysis showed that Par6 is phosphorylated on serine residues in the presence of wild-type TGFβ receptors (23). We also analyzed the Par6 species separately by two-dimensional (2D) tryptic phosphopeptide mapping. Par6* contained a prominent neutrally charged phosphopeptide that was dependent on the kinase activity of TβRII, as well as several other peptides of weaker intensity (Fig. 2C). Par6* is a distinct phosphoisoform, as it completely lacked the tryptic phosphopeptide present in the faster migrating form.

Fig. 2.

Par6 as a substrate of the TGFβ receptor complex. (A) TGFβ-dependent phosphorylation of Par6. FLAG-tagged Par6-expressing NMuMG cells were [32P]phosphate labeled and treated with (+) or without (–) TGF β for 1 hour before isolating TGFβ receptor–bound Par6 (49). Expression of Par6 and interaction with TβRI was verified by immunoprecipitation and immunoblotting in nonlabeled cells as indicated. (B) Dependence of Par6 phosphorylation on TβRII kinase activity. HEK293T cells expressing F-Par6 and the indicated combinations of WT or KR TβRII and TβRI were [32P]phosphate labeled and Par6 isolated by anti-FLAG immunoprecipitation. F-Par6 expression was verified by anti-FLAG immunoblotting of a portion of the immunoprecipitate, and TGFβ receptor expression was verified in nonlabeled cells transfected in parallel. (C) 2D tryptic phosphopeptide mapping of Par6. F-Par6 and F-Par6* were isolated from [32P]phosphate-labeled cells expressing TβRI and TβRII as in (B) and were subjected to 2D tryptic phosphopeptide mapping (49). The TGFβ receptor–dependent phosphopeptide is indicated by a red arrow. (D) Mapping the TGFβ receptor–dependent Par6 phosphorylation site. The indicated mutants of F-Par6 were subjected to tryptic phosphopeptide mapping, as in (C). (E) In vitro phosphorylation of Par6 by isolated receptor complexes. HEK293T cells were transiently transfected with the indicated combinations of TβRI and TβRII. Receptor complexes were isolated by nickel purification and used in an in vitro kinase assay with bacterially expressed GST-Par6(258–346) or GST-Smad2 MH2 domain as substrate. (F) 2D tryptic phosphopeptide mapping of GST-Par6(258-346). Phosphorylated GST-Par6(258–346) was subjected to phosphopeptide mapping as in (C).

To map the site of phosphorylation, we analyzed Par6 mutants. Because truncation of the C terminus in Par6(1–338) abolished the neutrally charged phosphopeptide (Fig. 2D), we generated serine-to-alanine point mutations at the two serine residues in this peptide and assessed TGFβ receptor–dependent phosphorylation. Whereas both wild-type Par6 and Par6(S342A) were phosphorylated, Par6(S345A) and the Par6(S342, 345A) double mutant were not. We analyzed the phosphorylation of a bacterially expressed C-terminal fragment of Par6 by TGFβ receptors in vitro. TGFβ receptors phosphorylated glutathione S-transferase (GST)–Par6(258–346) in vitro to a similar degree as they did the MH2 domain of Smad2, a known substrate of the receptor complex (27) (Fig. 2E). Furthermore, 2D tryptic phosphopeptide mapping revealed a phosphopeptide that migrated similarly to one from Par6 phosphorylated by TGFβ receptors in vivo (Fig. 2F). These data suggest that TβRII phosphorylates Par6 at its penultimate residue, Ser345. At the C terminus of Par6 from various metazoans, Ser345 either is conserved or is a threonine residue in vertebrates, but it is not found in Par6 from Drosophila or C. elegans (fig. S2A). The latter two species lack tight junctions (28), which suggests that C-terminal phosphorylation of Par6 may be associated with tight-junction homeostasis in vertebrates. Altogether, our results indicate that in unstimulated cells, TβRI and Par6 exist in a complex at tight junctions. TGFβ stimulation then induces assembly of the TβRI-TβRII heteromer, thus bringing the type II receptor to TβRI-bound Par6, where its constitutively active kinase phosphorylates Par6 on Ser345.

Par6 mutants block TGFb-induced tight-junction dissolution. Par6 has been proposed to negatively regulate tight junctions (18). Therefore, we sought to determine whether Ser345 phosphorylation of Par6 was important for TGFβ-dependent dissolution of tight junctions during EMT. We generated multiple NMuMG clonal lines stably expressing comparable amounts of FLAG-tagged wild-type Par6 or Par6(S345A) (fig. S2B) that were correctly localized to tight junctions (fig. S2C). However, although expression of wild-type Par6 did not affect TGFβ-induced EMT, Par6(S345A)-expressing clones displayed stable tight junctions even after prolonged stimulation with TGFβ (Fig. 3A). These clones were also resistant to the TGFβ-induced actin cytoskeletal rearrangement that is associated with the acquisition of a mesenchymal phenotype (Fig. 3A) (19), and they retained cortical actin staining. Similarly, expression of Par6(S345A) blocked the dissolution of adherens junctions in response to TGFβ, as marked by E-Cadherin staining (fig. S2D), as well as the loss of adherens junction–associated structures, as marked by β-Catenin staining (23). Thus, phosphorylation of Par6 by the TGFβ receptor is important for TGFβ-dependent dissolution of tight junctions and rearrangement of the actin cytoskeleton during EMT.

Fig. 3.

Phosphorylation of Par6 on Ser345 is required for TGFβ-dependent EMT in NMuMG cells. (A) TGFβ-induced tight-junction dissolution in NMuMG cells. Control NMuMG or clones expressing either Par6 or Par6(S345A), as indicated, were grown to confluent monolayers and then treated with (+) or without (–) TGF β for 48 hours before staining with Cy3-conjugated phalloidin to detect filamentous actin (red) and immunostaining for ZO-1 (green) (49). Overlay of the two images is shown (merge), with colocalization appearing as yellow. (B) Expression of Par6(S345A) does not block induction of vimentin expression. Cells as in (A) were incubated with or without TGFβ for 60 hours before staining with antibody to vimentin (red) and antibody to ZO-1 (green). (C) Smad-dependent transcription in NMuMG cells is not impaired by Par6. NMuMG cells were transiently transfected with either control plasmid Par6 or Par6(S345A) in conjunction with 3TP luciferase and were treated with 100 pM TGFβ (closed bars) or left untreated (open bars) for 24 hours before lysis and measurement of luciferase activity. (D) Nuclear accumulation of endogenous Smad2 in response to TGFβ stimulation. Control, Par6-, and Par6(S345A)-expressing NMuMG cells were grown to confluence and incubated with or without TGFβ for 48 hours. Localization of endogenous Smad2 was visualized by immunostaining with mouse monoclonal antibody to Smad2 (red) and ZO-1 (green), as described (49).

Par6(S345A) does not block Smad activation. To further characterize the EMT block, we examined the transcriptional response of the gene that encodes vimentin, an intermediate filament protein of mesenchymal cells (29). TGFβ stimulated expression of vimentin in Par6(S345A)-expressing cells despite the complete retention of tight junctions (Fig. 3B). Thus, abrogating TGFβ-dependent Par6 phosphorylation interferes with the dissolution of tight junctions but does not impair the induction of vimentin gene expression that is associated with EMT.

Smad-dependent transcriptional regulation is an important mediator of TGFβ-induced EMT, likely participating in the induction of the mesenchymal gene-expression program (30, 31). Therefore, we sought to determine whether Par6(S345A) hindered Smad signaling. We examined whether transfection of either Par6 or Par6(S345A) in NMuMG cells interfered with expression of the 3TP-luciferase TGFβ reporter gene (20), which provides a transcriptional readout of Smad activation. TGFβ induced a strong expression of 3TP-luciferase that was not inhibited by expression of wild-type Par6 or Par6(S345A) (Fig. 3C). Immunofluorescent staining of endogenous Smad2 revealed that, although Par6(S345A) mutant clones retained their tight junctions in response to TGFβ stimulation, nuclear accumulation of Smad2 was comparable to that in control cells at either 4 hours, 24 hours, or 48 hours after stimulation with TGFβ (Fig. 3D) (23). Similarly, Smad2 phosphorylation in response to TGFβ was comparable in control, Par6, and Par6(S345A) clones (fig. S2E).

TGFb regulation of the Par6-Smurf1-RhoA pathway is required for EMT. Smurf1, an E3 ubiquitin ligase, functions as an effector of the polarity complex by mediating localized ubiquitination and degradation of RhoA in cellular protrusions (32). RhoA is the prototypical member of the Rho GTPase family, which regulates many cellular processes, including cellular adhesion, motility, and polarity (33), and is an important modulator of cell junction formation and stability (3436). Therefore, we hypothesized that in polarized epithelial cells, phosphorylation of Par6 by the TGFβ receptor might regulate tight-junction dissolution by controlling association of Par6 with Smurf1 and the localized degradation of RhoA. We analyzed the interaction of endogenous Smurf1 with either FLAG-Par6 or FLAG-Par6(S345A) in NMuMG cells. In the absence of TGFβ signaling, we detected that some Smurf1 precipitated with wild-type Par6, and the amount increased after treatment of cells with TGFβ (Fig. 4A). In contrast, when we examined Par6(S345A) mutants, there was little if any detectable association with Smurf1. This suggests that Smurf1 preferentially binds Par6 that is phosphorylated by TβRII on Ser345. To confirm this, we compared the phosphorylation pattern of total Par6 to that of Par6 bound to Smurf1 in the presence of TGFβ signaling. Par6 isolated from the total cellular pool displayed tryptic phosphopeptides corresponding to both the faster and slower migrating forms of Par6 (compare Fig. 4B to Fig. 2C). However, when we examined Par6 specifically bound to Smurf1 in the presence of TGFβ signaling, we observed only the phosphorylation pattern that corresponded to the Par6* phosphoisomer, which contains the Ser345 phosphopeptide that is a target of TβRII. Hence, TβRII-dependent phosphorylation of Par6 at Ser345 appears to stimulate association with Smurf1.

Fig. 4.

Interaction of Smurf1 with Par6 phosphorylated on Ser345 and regulation of EMT. (A) Assembly of Par6-Smurf1 complexes. Cells expressing FLAG-tagged WT Par6 or Par6(S345A) were treated with (+) or without (–) TGF β for 1 hour, and lysates were subjected to anti-FLAG immunoprecipitation followed by immunoblotting for Smurf1, as indicated. Protein amounts were confirmed by immunoblotting lysates. (B) Par6 bound to Smurf1 is phosphorylated at position Ser345. Total (left map) or FLAG-Smurf1–bound (right map) Par6 was analyzed by phosphopeptide mapping, as in Fig. 2. The position of phosphorylated Ser345 is indicated. (C) TGFβ-dependent redistribution of endogenous Smurf1. The indicated NMuMG cells were treated without (–) or with (+) TGF β for 6 hours before imaging endogenous Smurf1 (green) or ZO-1 (red), as described (49). (D) Reconstructed Z-stacks derived from confocal optical slices from cells treated as in (C). Colocalization appears yellow, and tight junctions are marked by white arrowheads. (E) Smurf1(ΔC2) blocks TGFβ-induced EMT. Confluent NMuMG cells were transiently transfected as indicated and incubated in the absence (–) or presence (+) of TGF β for 48 hours. Smurf1 was detected by FLAG immunostaining (red) before counterstaining for ZO-1 (green). The percentage of Smurf1-expressing cells with tight junctions in the absence (white) or presence (black) of TGFβ is plotted in the right panel (mean ± SD of 30 cells counted from three experiments). (F) Smurf1 siRNA blocks TGFβ-induced EMT. Confluent NMuMG cells were transiently transfected with either FITC-labeled luciferase control siRNA or FITC-labeled Smurf1 siRNA (green) and incubated in the absence (–) or presence (+) of TGF β for 24 hours before counterstaining for ZO-1 (red). The percentage of Smurf1-siRNA containing cells that retained apical tight junctions was plotted (right panel) (mean ± SD of 30 cells counted from three experiments).

To investigate the role of Smurf1 in the Par6 pathway, we examined the distribution of endogenous Smurf1 in NMuMG cells. In the absence of signaling, Smurf1 was not present in ZO-1 labeled regions (Fig. 4, C and D). However, after a 6-hour treatment with TGFβ, a point during EMT when tight junctions have not yet been lost, endogenous Smurf1 redistributed to cell-cell junctions. This TGFβ-dependent redistribution in control cells did not occur in NMuMG cells stably expressing Par6(S345A) (Fig. 4, C and D). To examine the role of Smurf1, we overexpressed either wild-type Smurf1 or Smurf1(ΔC2), which is a mislocalizing mutant that interferes with Smurf1 function (37). Unlike green fluorescent protein (GFP)–expressing control cells, more than 50% of Smurf1-expressing cells displayed impaired tight-junction stability or were located on the basolateral surface of the epithelial monolayer and exhibited a mesenchymal morphology even in the absence of TGFβ stimulation (Fig. 4E). This morphology is consistent with Smurf1 acting as a factor that promotes mesenchymal transition. In contrast, Smurf1(ΔC2)-expressing cells failed to undergo constitutive EMT and were also resistant to TGFβ-induced EMT (Fig. 4E). Next, we examined whether small interfering RNA (siRNA) against Smurf1 (35) affected TGFβ-dependent EMT. In cells transfected with control fluorescein isothiocyanate (FITC)–conjugated luciferase siRNA, tight-junction dissolution was unaffected, whereas in cells transfected with FITC-conjugated Smurf1 siRNA, tight junctions were retained in more than 50% of the cells after TGFβ treatment (Fig. 4F). When Smurf1 siRNA was transfected along with a siRNA-resistant version of FLAG-tagged Smurf1 (fig. S3A), cells adopted a mesenchymal phenotype consistent with a rescue of Smurf1 function (fig. S3B).

Localized degradation of RhoA is important for Smurf1-dependent regulation of protrusive activity (35). In control cells, TGFβ caused a reduction of 30% in the amount of RhoA, and treatment of cells with the proteasome inhibitor MG132 reversed the decrease (Fig. 5A and fig. S4A). This TGFβ-dependent decrease in RhoA was not observed in Par6(S345A)-expressing cells (Fig. 5B and fig. S4B), which suggests that TGFβ regulates RhoA levels through Par6. Using mutational analysis, we identified lysines 6 and 7 as the RhoA acceptors for ubiquitin transferred by Smurf1 (fig. S5). We confirmed in a RhoA activation assay that the corresponding RhoA(K6,7R) mutant bound GST-Rhotekin, a RhoA effector that associates specifically with active RhoA, to a similar degree as did wild-type RhoA, thereby demonstrating that the mutation does not affect GTP binding (23). Therefore, we examined whether Smurf1-resistant RhoA affected dissolution of tight junctions in response to TGFβ. Although wild-type RhoA had no effect on EMT, RhoA(K6,7R) inhibited TGFβ-dependent EMT (Fig. 5C). Furthermore, the proteasome inhibitor MG132 also impaired TGFβ-dependent dissolution of tight junctions (23), whereas transfection of FITC-conjugated RhoA siRNA resulted in the disruption of tight junctions in more than 60% of transfected cells (23). Thus, Smurf1-dependent RhoA turnover regulates EMT in response to TGFβ.

Fig. 5.

Requirement of RhoA turnover for TGFβ-dependent EMT. (A) TGFβ-dependent decrease in RhoA steady-state levels is blocked by proteasome inhibition. Cells were untreated (white bars) or TGFβ treated (black bars) for 6 hours, either in the absence (–) or presence (+) of the proteasome inhibitor MG132, as indicated. Endogenous RhoA was analyzed by immunoblotting and quantitated on a Fluor-S MultiImager (Bio-Rad, Hercules, CA) and is plotted relative to the amount of RhoA in untreated cells (mean ± SEM of four experiments). (B) Par6 (S345A) blocks TGFβ-dependent reduction in RhoA steady-state levels. RhoA steady-state levels in the indicated cell lines untreated (–) or treated (+) with TGF β were analyzed as in (A). (C) Expression of RhoA(K6,7R) blocks TGFβ-dependent EMT. Cells transiently expressing T7 epitope–tagged WT RhoA (T7-RhoA) or mutant RhoA [T7-RhoA(K6,7R)] were analyzed for TGFβ-dependent EMT, as described in Fig. 4D. The percentage of RhoA-expressing cells with tight junctions, in the absence (white bars) or presence (black bars) of TGFβ, was quantitated relative to GFP-expressing control cells (right). (D) A model of the TGFβ-Par6 pathway.

Altogether, our results demonstrate a direct link between TGFβ receptors, the polarity complex, and the regulation of tight-junction dissolution during EMT. In this model (Fig. 5D), cell-surface TβRI is localized to tight junctions in polarized epithelial cell sheets (9). TGFβ, which induces association of TβRI with TβRII, leads to the accumulation of receptor complexes in tight junctions, thereby bringing the TβRII receptor kinase to the Par6-TβRI complex. This leads to phosphorylation of Par6, which in turn stimulates binding of Par6 to Smurf1. The Par6-Smurf1 complex then mediates the localized ubiquitination of RhoA to enable the TGFβ-dependent dissolution of tight junctions during EMT.

The transition of epithelia from a highly organized sheet of cells exhibiting apical-basal polarity to a mesenchymal phenotype allows for independent motile behavior that is unrestricted by the 2D layer of the epithelium. In multicellular organisms, this may permit the evolution of specialized form and function during development (2), but in cancer, it is a critical pathological event during progression to a metastatic phenotype (1). EMT is a complex process that involves rearrangements of cell-cell contacts and cell-matrix contacts, as well as changes in gene expression. The actin cytoskeleton is extensively remodeled from its cortical distribution in epithelia to form contractile stress fibers typical of motile fibroblasts. However, it has been unclear how these events are orchestrated by extracellular cues such as TGFβ. In epithelial cells, polarity protein complexes have been proposed to serve as important modulators of tight-junction homeostasis and apical-basal polarity (13). Par6 is a key component of this machinery and functions as a scaffold to regulate the recruitment of effectors. Here, we link TGFβ signaling directly to the Par6-Smurf1 polarity complex. We previously described a PKCζ-dependent role for Smurf1 and localized RhoA turnover in the formation of cellular protrusions (32). It is unclear whether the mechanisms that regulate TGFβ-dependent assembly of the Smurf1-Par6 complex in cell junctions differ from those that control protrusion dynamics. However, integrin signaling can regulate the Par6 complex to control directed cell motility (38), and receptor tyrosine kinases are required in Par6 modulation of axonal polarity (39). Although the molecular mechanisms underlying these pathways are unclear, they suggest that Par6 may be a focal point for the control of cell polarity by multiple extracellular cues not only during development and homeostasis but also in the progression of cancer.

Reorganization of the cytoskeleton, and in particular the formation of cortical actin filaments, is critical for the establishment of tight and adherens junctions and the regulation of epithelial cell apical-basal polarity. The concerted activity of multiple Rho GTPases, which are key regulators of the actin cytoskeleton, coordinates these events (35). Notably, RhoA functions to maintain apical-basal polarity and cell junctions in colonic epithelia and keratinocytes (34, 36), and several reports have suggested that RhoA can stabilize tight junctions and increase transepithelial resistance (4042). These observations are consistent with our findings, as we propose that the localized degradation of RhoA is important in TGFβ-mediated EMT. However, other evidence suggests that RhoA activity promotes TGFβ-induced EMT (43, 44) and stress-fiber formation (33). These divergent outputs may be explained in part by the activity of downstream effectors of RhoA. RhoA-induced Rho-kinase (ROCK) activity can lead to the dissolution of cell-cell contacts and the formation of stress fibers, whereas mDia, another downstream effector of RhoA, promotes junction stabilization (36). Consequently, how RhoA regulates junction dynamics may be distinct from the mechanism underlying formation of stress fibers. Inhibition of ROCK blocks TGFβ-dependent formation of stress fibers but not dissolution of tight junctions (45); therefore, the requirement for RhoA activity in TGFβ-induced EMT may reflect its function in the latter pathway. Thus, RhoA may fulfill multiple spatiotemporal roles during TGFβ-induced EMT.

In late-stage cancers, autocrine production of TGFβ induces EMT and synergizes with activated Ras to promote metastasis (46). A number of other signaling conduits, including Smads, the mitogen-activated protein kinase p38, and phosphoinositide-3 kinase, have also been implicated in this process (3, 31, 47, 48). The specific events that are regulated by the latter two pathways are unclear; however, Smads are critical for the transcriptional responses accompanying EMT. Thus, our demonstration that both Smad activation and the induction of mesenchymal gene expression can be dissociated from a pathway controlling loss of tight junctions reveals a bifurcated signaling network at the level of the TGFβ receptor. This dynamic interplay of multiple pathways likely allows the TGFβ receptor system to orchestrate EMT by coordinating dissolution of tight junctions with gene-expression programs. Moreover, our high-throughput systematic analysis of protein-protein interactions in mammalian cells revealed that the TGFβ-Par6 signaling pathway is embedded in a larger network within which TGFβ components extensively connect to the p21-activated kinase signaling network (9). It will be interesting to determine the biological contexts in which this larger network functions to interpret TGFβ-family signaling.

Supporting Online Material

www.sciencemag.org/cgi/content/full/307/5715/1603/DC1

Materials and Methods

Figs. S1 to S5

Movies S1 and S2

References and Notes

View Abstract

Navigate This Article