The role of internalization in transforming growth factor beta1-induced Smad2 association with Smad anchor for receptor activation (SARA) and Smad2-dependent signaling in human mesangial cells.

Recent data investigating the role of the Smad anchor for receptor activation (SARA) in TGF-beta signaling have suggested that it has a crucial function in both aiding the recruitment of Smad to the TGF-beta receptor, and ensuring appropriate subcellular localization of the activated receptor-bound complex. The FYVE domain in SARA directs its localization to early endosomal compartments where it can interact with both the TGF-beta receptors and Smads. However, the necessity of endocytosis in the TGF-beta response remains controversial. We sought to examine the role of internalization in TGF-beta/Smad signaling in human kidney mesangial cells. Using co-immunoprecipitation studies, we show that endogenous Smad2 interacts with SARA after TGF-beta1 stimulation. Inhibition of clathrin-mediated internalization only slightly affects TGF-beta1-stimulated association between SARA and Smad2, Smad2 phosphorylation, or Smad2 interaction with Smad4. However, endocytosis inhibition decreases TGF-beta1-induced Smad2 nuclear translocation and thus abrogates Smad2-dependent transcriptional responses. The TGF-beta1-stimulated association between SARA and Smad2 peaks at 30 min followed by separation of the complex components. However, under conditions of inhibited endocytosis, Smad2 remains bound to SARA for at least 6 h without a significant decline in associated levels. This lack of complex dissociation correlates with a lack of Smad2 nuclear accumulation and reduction of Smad2-dependent ARE-Luc reporter activity. Our data therefore suggest that endocytosis plays a critical role in TGF-beta signaling in mesangial cells, and that internalization enhances the dissociation of Smad2 from the TGF-beta receptor-SARA complex, allowing Smad2 to accumulate in the nucleus and modulate target gene transcription.

Transforming growth factor (TGF) 1 -␤ is a ubiquitously expressed cytokine that has varied roles, affecting cellular processes including proliferation, differentiation, apoptosis, fibro-sis, and tumorigenesis (1). This diversity is likely caused by differential involvement of TGF-␤ downstream signaling mediators, and therefore understanding precise interactions among TGF-␤ signaling components is of critical importance.
TGF-␤ signaling is initiated when ligand-bound TGF-␤ type II receptor (T␤RII) binds to, and phosphorylates, the TGF-␤ type I receptor (T␤RI) (2)(3)(4). This phosphorylation, in the T␤RI cytoplasmic GS region, leads to its activation and its ability to activate the receptor-regulated Smads (R-Smads), Smad2, and Smad3, by C-terminal serine phosphorylation. Once phosphorylated, the R-Smads form a heteromultimeric complex with the common mediator (Co)-Smad (Smad4) and accumulate in the nucleus to regulate transcriptional responses (2)(3)(4). However, the seemingly simple initial delineation of the Smad signaling pathway has gained considerable complexity from the recent discovery of a number of additional factors that interact with the Smads and can regulate the signaling outcome. Among these interacting factors are those with anchoring or chaperone activity that aid in recruitment of R-Smads to the TGF-␤ receptor complex. Thus far, the best characterized of this type of Smad cofactor is SARA (Smad anchor for receptor activation). SARA contains both a Smad-binding domain (SBD), which has been shown to interact with Smad2 and Smad3 (5), and a C-terminal, TGF-␤ receptor complex-interacting region (6). SARA is therefore proposed to play a role in presenting R-Smads to the receptor for phosphorylation.
Additionally, SARA, by virtue of its FYVE domain binding to phosphatidylinositol 3-phosphate (PI3P), can interact with EEA1-positive, Rab 5-containing early endosomal compartments within the cell (6 -10). Disruption of the SARA endocytic localization through either expression of a mutant SARA lacking the FYVE domain (SARA/D1-664) (6) or inhibition of PI3P generation by wortmannin treatment (7) causes a redistribution of SARA from punctate endocytic structures to the cytosol, and can prevent TGF-␤-mediated transcriptional responses. However, while it has been clearly demonstrated that activated T␤RI and T␤RII are internalized into SARA-containing EEA1 endosomes, controversy remains as to the requirement for SARA binding and endocytosis in transducing the TGF-␤ signal. Although some reports indicate that TGF-␤-mediated, Smad2-dependent transcription is reduced by inhibition of endocytosis (9 -12), others find no effect of either inhibiting vesicle formation or interrupting SARA association on Smad2 (13) or Smad3 (14) responses at either the phosphorylation or transcriptional level. Additionally, a recent report investigating the role of endocytosis in the signaling mediated by the TGF-␤ family member, activin, through the Alk4 receptor, suggests that internalization is not required for Smad2 phosphorylation or transcriptional activity of the TGF-␤1/activin-responsive p3TP-Lux transcriptional activation (15). This inconsistency is likely due, at least in part, to the fact that almost every study of the Smad2-SARA interaction has been performed using transformed cell lines and overexpressed Smad, SARA, or TGF-␤ family receptors. Because high levels of SARA have been shown to lead to the formation of large nonvesicular aggregates (7) and to result in altered Smad localization (6), we sought to examine the specificity, localization and necessity of the interaction between endogenous SARA and Smad for TGF-␤ responsiveness.
We report here that endogenous SARA and Smad2 interact in response to TGF-␤1 treatment. There is a requirement for endocytosis in Smad signaling in that inhibition of endocytic vesicle formation abrogates the TGF-␤ transcriptional response. Our data suggest that an internalization step, although dispensable for phosphorylation of Smad2 and its interaction with Smad4, may be required to allow dissociation of Smad2 from SARA followed by its nuclear accumulation and transcriptional activity.
Cell Culture-Human mesangial cells were isolated by differential sieving of minced normal human renal cortex obtained from anonymous surgery or autopsy specimens. The cells were grown in Dulbecco's modified Eagle's medium/Ham's F12 medium, supplemented with 20% heat-inactivated new-born calf serum (NBCS), glutamine, penicillin/ streptomycin, sodium pyruvate, Hepes buffer, and 8 g/ml insulin (Invitrogen Life Technologies) as described previously (18), and were used between passages 5 and 8. Mouse mammary gland epithelial cells (NMuMG) were purchased from ATCC. They were passaged in Dulbecco's modified Eagle's medium with 4.5 g/liter glucose, 10 g/ml insulin, and 10% NBCS.
Potassium Depletion-Cells were switched to 1% NBCS-containing media for 18 -24 h prior to treatment. Media were then switched either to fresh 1% NBCS-containing media or to hypotonic media (50% Dulbecco's modified Eagle's medium, 50% H 2 0). After 10 min in hypotonic buffer, the cells were either switched to isotonic media without potassium (10 mM Tris-HCl, pH 7.5, 150 mM NaCl) or isotonic media containing 10 mM KCl. Cells were depleted for 30 min prior to TGF-␤1 treatment, and then lysed in RIPA buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS) containing protease and phosphatase inhibitors (Sigma). After clarification by centrifugation at 18,000 ϫ g, the protein content was determined by Bradford protein assay (Bio-Rad) and prepared for immunoprecipitation or Western blotting. Alternatively, the cells were fractionated as described below.
Cell Fractionation-Cells were scraped into a detergent-free buffer (20 mM Tris-HCl, pH 7.5, 0.5 mM EDTA, 0.5 mM EGTA, 10 mM ␤-mercaptoethanol) containing protease and phosphatase inhibitors. The cells were then disrupted by 15 strokes of a Dounce homogenizer, and centrifuged at 1,000 ϫ g for 5 min to remove nuclei. The supernatant was removed and further centrifuged at 100,000 ϫ g for 45 min. After centrifugation, the supernatant was saved as the cytosolic fraction. The pellet was resuspended in detergent-free buffer supplemented with 1% (final concentration) Triton X-100, sheared with a syringe, incubated for 30 min on ice, and centrifuged to remove insoluble material. This supernatant represents the soluble membrane fraction. The nuclei were resuspended in RIPA buffer, syringe-sheared, incubated for 30 min on ice, and centrifuged to remove insoluble material. This fraction represents the nuclear fraction. Protein concentration for each fraction was determined by Bradford protein assay.
Immunoprecipitation and Western Blot-300 -600 g of protein from total cell lysates were diluted in the same final volume of RIPA buffer containing protease and phosphatase inhibitors. Anti-SARA (N-20) or anti-Smad4 (H-552) antibody was added at a final concentration of 2 g. Immunoprecipitations and Western blotting were performed as described previously (19).
Transfection-Cells were plated, in triplicate wells per condition, at 1.4 ϫ 10 5 cells per well in 6-well plates. The following day, medium was changed to 1 ml per well 1% NBCS-containing medium and transfected as described previously (20)  and assayed for luciferase activity and also for ␤-galactosidase activity to correct for transfection efficiency.
Immunocytochemistry-Cells were grown to 70% confluence on gelatin-coated glass coverslips in 6-well dishes. For determination of Smad nuclear translocation, cells were transfected with wild type-or K44Adynamin using FuGENE 6 reagent (as described above). To confirm dynamin effectiveness cells were co-transfected with HA-tagged T␤RI along with either wild type or K44A-dynamin. To determine SARA subcellular localization and to confirm inhibition of endocytosis, cells were serum deprived and potassium depleted as described above. Cells were then either treated with TGF-␤1 for various time points, or labeled for 30 min with Alexa transferrin 594 (50 g/ml). Coverslips were prepared for immunocytochemistry by fixing in 3.7% formaldehyde for 15 min, followed by permeabilization with ice-cold 0.5% Triton X-100 for 5 min, and blocking in 3% bovine serum albumin-phosphate-buffered saline for 20 min. Antibody solutions were layered onto coverslips at a final concentration of 2 g/ml for Smad (H-2) or 4 g/ml for HA (Y-11), SARA (H-300), EEA1 (N-19), or T␤RI (V-22) and incubated in a humidified chamber for 2 h at 37°C, followed by three 5-min washes in phosphate-buffered saline. The coverslips were then incubated in secondary antibody for 30 -60 min with 2.5 g/ml either chicken anti-goatrhodamine, chicken anti-mouse-FITC, chicken anti-rabbit-FITC, or Al-exaFluor594 goat anti-rabbit secondary antibodies. To rule out background fluorescence, negative controls were incubated with secondary antibodies only. After three 5-min washes, cover slips were mounted onto glass slides, and images were scanned using a Zeiss LSM510 laser scanning confocal microscope.

TGF-␤1 Stimulates Smad2 and Smad3
Association with SARA-There are relatively few examples of SARA interaction with R-Smads (5,6,9,10,21,22), only one of which showed interaction of endogenous SARA with R-Smads (6), and none in human mesangial cells. We therefore first sought to examine the Smad2-SARA interaction in serum-deprived mesangial cells treated with 1 ng/ml (40 pM) TGF-␤1 for various time periods. After immunoprecipitation with anti-SARA antibody, the immunoprecipitated complexes were analyzed by Western blot with an antibody that recognizes both Smad2 and Smad3. As shown in Fig. 1A, there is little Smad2 or Smad3 detected in association with SARA in the absence of treatment. However, with 30 min of TGF-␤1 stimulation, there is a marked increase in co-immunoprecipitated Smad2. This interaction is apparent within 5 min (data not shown), peaks at 30 min, and continues to be somewhat increased over basal levels 24 h after treatment. Increased association is not caused by changes in total cellular SARA (1A) or Smad2 (1B) expression levels, which do not vary throughout 24 h of treatment with TGF-␤1. Interestingly, while Smad3 is strongly expressed in mesangial cells (Fig. 1B, lower panel), there is only a slight band revealed for Smad3-associated SARA (Fig. 1A).
TGF-␤1 activates Smad2 through receptor-mediated C-terminal phosphorylation, and the pattern of TGF-␤1-induced Smad2-SARA association parallels that of Smad2 phosphorylation as detected using an antibody specific for phosphoserine 465/467 (Fig. 1B). PDGF, which does not induce the phosphorylation of Smad2 ( Fig. 1B and Ref. 20), does not stimulate its association with SARA (Fig. 1A). The TGF-␤1-induced association between SARA and Smad2 does not appear to be restricted to mesangial cells, as we detect an increased association after 30 min TGF-␤1 treatment of an unrelated mouse mammary gland epithelial cell line (NMuMg) (Fig. 1C).
Clathrin-mediated Endocytosis Is Required for Smad2-dependent Transcriptional Activity-Because SARA has been shown to associate with EEA1-positive early endosomal compartments (7), and its mislocalization has been suggested to disrupt Smad2 signaling, we sought to determine whether human mesangial cells require endocytosis to signal through Smad2 after TGF-␤1 stimulation. We performed transient transfection of mesangial cells with the Smad2-specific ARE-Luc reporter construct along with a FAST-2 expression construct and either wild-type or K44A mutant dynamin. This K44A mutant of dynamin does not allow excision of vesicles from the cell membrane and thereby blocks clathrin-mediated endocytosis (16). Whereas wild-type dynamin does not affect the TGF-␤1 induction of ARE-Luc, this activation was significantly inhibited in the presence of the dynamin mutant ( Fig. 2A), confirming a role for internalization in Smad2 transcriptional activity. Because blocking the function of dynamin would also block caveolar excision from the membrane (23), and because lipid raft/caveolar internalization may mediate the ubiquitin-dependent degradation of the T␤R (24), we sought to rule out the possibility of caveolar compartmentalization by using nystatin to disrupt cellular cholesterol and thereby interfere with caveolae formation (25). Nystatin treatment has only a minimal inhibitory effect in comparison to that of the K44Adynamin mutant (Fig. 2B). Therefore the internalization requirement in TGF-␤1-mediated Smad2 signaling is likely specific to a non-caveolar, clathrin-mediated endocytic event.
Inhibition of Endocytosis Abrogates TGF-␤1-stimulated Nuclear Translocation of Smad2-Having established that endocytosis is required for Smad2 signaling, we sought to further investigate the specific role of internalization after TGF-␤ receptor stimulation. To determine whether endocytosis is required for nuclear Smad2 accumulation, we performed cell fractionation experiments under conditions of cellular potassium depletion to inhibit clathrin-coated pit formation (26). In agreement with other reports (9,11), under conditions of inhibited endocytosis, the level of nuclear Smad2 accumulation in response to TGF-␤1 is decreased compared with that of potassium control conditions (Fig. 3A). This is confirmed by immunocytochemical analysis for intracellular localization of Smad2. Whereas there is a striking increase in detection of Smad in the nucleus of TGF-␤1 treated cells under potassium control conditions (Fig. 3B compare panels b to a), potassium depletion completely abrogates this response (Fig. 3B, panels c and d). To ensure that the depletion of intracellular potassium had, in fact, reduced vesicle formation and thus endosomal association of SARA, we performed immunocytochemical analysis to detect SARA localization under potassium-depleted versus potassiumcontaining conditions. In control conditions, SARA shows the distinct punctuate staining pattern associated with endosomal localization (Fig. 4A, left panel). As expected (6 -9), these punctuate structures also contain the early endosomal marker EEA1 (27). However, in potassium-depleted cells SARA immunostaining becomes faint and diffuse throughout the cell and no longer shows extensive co-localization with EEA1 (Fig. 4A,  right panel). To confirm that the levels of SARA expression were not altered by potassium depletion, we performed Western blot analysis of SARA from cells treated with TGF-␤1 for various duration in either potassium-control or potassium-depleted conditions. As shown in Fig. 4B, throughout 6 h of treatment, there are no significant changes in the levels of SARA total protein expression. To further confirm that potassium depletion is inhibiting endosomal internalization, we treated the cells with fluorescently tagged transferrin (AF-tf594). Transferrin is known to internalize via clathrin-mediated endocytosis (28). As shown in Fig. 4C, top panel, cells incubated with AF-tf594 for 30 min show a punctate fluorescence. In contrast, in potassium-depleted conditions (4C, bottom panel) there is a lack of internalized AF-tf594. These results suggest that our potassium depletion conditions are effectively blocking endocytic internalization. However, there remains the possibility that the relatively nonspecific conditions of potassium depletion may be introducing alternate effects on the cells that disrupt the ability of Smad2 to translocate to the nucleus. Therefore, we used the K44A-dynamin mutant as an alternative means to block endocytic internalization. Cells were transfected with either HA-tagged wt-dynamin or K44A-dynamin and stained for Smad and HA. As shown in Fig. 5A, K44A-dynamin decreases TGF-␤1-induced nuclear accumulation of Smad2, whereas wild-type dynamin-expressing cells respond with nuclear accumulation similar to untransfected neighboring cells. This inhibitory effect of K44A-dynamin on translocation is seen in every transfected cell at both an early time point (Fig. 5A, panels a-d), corresponding to peak Smad2 phosphorylation (Fig. 1B) as well as at a longer duration of treatment of 6 h (Fig. 5A, e-h). This confirms the effect seen with potassium depletion in that Smad nuclear translo- cation requires endocytic internalization. To ensure K44A-dynamin was inhibiting internalization of T␤R in human mesangial cells as previously reported in other cell types (9, 13, 24), we co-transfected a T␤RI expression construct along with either wt-or K44A-dynamin constructs and performed immunocytochemical analysis of intercellular localization of the T␤RI and EEA1. As shown in Fig. 5B, there is T␤RI internalization and co-localization with EEA1 in cells transfected with wt-dynamin (Fig. 5B, left panel), but expression of T␤RI is predominantly detected at the cell periphery in cells expressing K44A-dynanim (Fig. 5B, right panel).

Phosphorylation of Smad2 Does Not Require Internalization and Can Occur While Smad2 Is in Association with SARA-
Interfering with endocytosis via potassium depletion could inhibit Smad2 signaling because mislocalization of SARA may not allow the SARA-Smad2 complex to be efficiently formed. To examine this possibility, we performed co-immunoprecipitations to determine the effect of potassium depletion on the levels of TGF-␤1-stimulated SARA-Smad2 association. As demonstrated in Fig. 6, the level of association is only slightly reduced by inhibition of internalization. Therefore the complex forms in response to TGF-␤1 even though SARA localization is altered (Fig. 4A).
As one of the potential functions of the endosome may be to act as an intracellular compartment to enhance the duration and level of membrane-associated signaling complexes, the requirement for endocytosis may be to enhance Smad2 receptormediated phosphorylation (9,11). To investigate this possibility, we analyzed Smad2 phosphorylation under control or potassium-depleted conditions. Total Smad2 phosphorylation was determined by immunoprecipitation of Smad2 followed by phosphoserine Western blotting. Additionally we examined the TGF-␤ receptor-mediated C-terminal phosphorylation using an antibody specific for Smad2-phosphoserine 465/467. Although there may be a slight reduction in overall phosphorylated Smad2 levels, we found that neither Smad2 total serine phosphorylation nor Smad2 C-terminal phosphorylation were con-siderably reduced when endocytosis was inhibited (Fig. 7). If both the TGF-␤1-induced Smad2-SARA complex formation and Smad2 phosphorylation are only minimally affected by potassium depletion, it would imply that the inhibition of endocytosis still allows for a membrane-associated complex of SARA and phosphorylated Smad2. To examine this possibility, SARA was immunoprecipitated from cells under control or potassium-depleted conditions, and the complexes were analyzed by immunoblotting with anti-phospho-Smad2. Phosphorylated Smad2 coimmunoprecipitates with SARA after treatment with TGF-␤1 (Fig. 8). Moreover, phosphorylation of Smad2 in association with SARA is not significantly affected by inhibition of endocytosis. Therefore, our data suggest that endocytosis is not critical for Smad2/SARA complex formation or for Smad2 phosphorylation in response to TGF-␤1.
Smad4 Associates with the SARA-Smad2 Complex-Because Smad2 forms a complex with Smad4 in response to TGF-␤1, and because it has been reported previously that Smad2 forms mutually exclusive complexes with either SARA or Smad4 (6), one possible explanation of the requirement for endocytic internalization is to allow Smad2-Smad4 complex formation. However, in contrast to that previous study (6), Smad4 is detected in SARA immunoprecipitates (Fig. 9A). The association correlates temporally with the enhanced Smad2-SARA association as can be seen by comparing the Smad2 detection (Fig. 9A, top panel) from SARA immunoprecipitation to the same blot unstripped and reprobed for Smad4 (Fig. 9A, middle panel). This effect of TGF-␤1 on SARA-Smad4 association is not caused by enhanced availability of Smad4, because total cellular Smad4 levels do not change (Fig. 9A, bottom panel). To further confirm that Smad4 interacts with the SARA-Smad2 complex, we performed immunoprecipitation of Smad4 and blotted for associated Smad2 and for SARA. As shown in Fig. 9B, Smad4 immunoprecipitates contain bands of associated Smad2 as well as SARA, and both of these interactions are enhanced by TGF-␤1 treatment. Because it has been reported that Smad4 does not bind SARA directly (5,6,22), we would anticipate that Smad4 is associated with SARA via its interaction with Smad2. However, we have not demonstrated a lack of endogenous Smad4-SARA association in our system, since a co-immunoprecipitation of SARA may pull-down both Smad2-SARA and Smad4-SARA complexes. Therefore the possibility exists that Smad2 may form mutually exclusive complexes with SARA or Smad4 as previously reported. To determine whether the three molecules can form a single complex, we performed successive immunoprecipitation reactions. First we immunoprecipitated Smad2, followed by a second round of immunoprecipitation of Smad4 from the supernatant of the Smad2 immunoprecipitation. As shown in Fig. 9C, immunoprecipitated Smad2 forms a complex with SARA as well as with Smad4 (Fig. 9C, left  panels). Both of these complexes are induced by TGF-␤1 treatment. However, Smad4 that is immunoprecipitated from the FIG. 6. Smad2 associates with SARA under conditions of inhibited endocytosis. Cells were either serum-starved only (C) or serumstarved and then incubated in potassium containing (K(ϩ)) or potassium depleted (K(Ϫ)) conditions, followed by 30 min of treatment with either vehicle (Ϫ) or 1 ng/ml TGF-␤1 (ϩ). Cells were immunoprecipitated with SARA and then blotted first for Smad2/3 (top) followed by blotting for SARA (bottom). Representative blots from one of three separate experiments are shown.

FIG. 7. Smad2 can be phosphorylated when endocytosis is in-
hibited. Cells were treated as described in Fig. 4 and either immunoprecipitated with Smad2/3 (N-19) and developed for phosphoserine (top) followed by re-probing for Smad2/3 (middle), or whole cell lysates were blotted for C-terminal Smad2 phosphorylation (bottom).
FIG. 8. Phosphorylated Smad2 can co-precipitate with SARA. SARA was immunoprecipitated from cells treated with TGF-␤1 for 30 min under control or potassium-deprived conditions. The immunoprecipitate was probed first for C-terminal phospho-Smad2 (top panel), followed by stripping and re-probing for Smad2 (middle), and then for SARA (bottom).
Smad2 supernatant does not show association with SARA (Fig.  9C, middle panels). Therefore, it appears that the Smad4 that is not associated with Smad2 is also not associated with SARA. Further, SARA immunoprecipitated from the lysates after Smad2 was removed no longer shows association with Smad4 (Fig. 9C, right panels). Thus it appears that endogenous Smad2 forms a complex with both SARA and Smad4, and that these complexes are not mutually exclusive.
Smad2 Does Not Separate from SARA under Conditions of Inhibited Endocytosis-The fact that Smad2 can associate with SARA, can be phosphorylated, and can at least be in proximity with Smad4 without requiring internalization suggests that the requirement for endocytosis lies somewhere in between these initial events and Smad2 translocation to the nucleus. This led us to speculate that the potential role for internalization in the Smad2 signaling pathway is to allow dissociation of the complex components, thereby allowing separation of the Smad2-Smad4 complex from the T␤R-associated SARA as required for Smad2 nuclear accumulation. As shown in Fig. 10A, co-immunoprecipitation to examine the Smad2-SARA complex over longer durations of TGF-␤1 treatment, indicate that, in control conditions, Smad2 dissociates from SARA between 0.5 and 4 h of treatment (left panel). In contrast, when endocytosis is inhibited, Smad2 is induced by TGF-␤1 to associate with SARA, and remains bound for at least 6 h (right panel). This trend, illustrated in Fig. 10B, demonstrates that without endocytosis Smad2 remains bound to SARA, and suggests that the separation of this complex occurs subsequent to internalization. Because potassium depletion for 6 h may be a relatively harsh condition, it is possible that the lack of SARA-Smad dissociation is because of a generally unhealthy state of the cells. Therefore to rule out a nonspecific toxicity effect, we transfected the cells with a CMV-␤-galactosidase reporter con-struct and analyzed the ␤-galactosidase activity of cells under potassium control or depletion conditions. As shown in Fig.  10C, cells that are incubated under potassium-depleted conditions do not vary in ␤-galactosidase activity compared with control cells, nor does TGF-␤1 treatment affect the transcriptional response. Additionally, if cells are incubated under potassium-depleted and control conditions for 1 h and then switched back to 1% NBCS-containing media for another 18 h, they have ␤-galactosidase activities identical to cells that have not undergone any media change (Fig. 10C). This suggests that depletion of intracellular potassium is not causing the cells to become apoptotic, and that the lack of Smad2 responsiveness is due specifically to the inhibition of endocytosis. Therefore our data suggest that the inhibition of internalization alters the dissociation of the Smad2-SARA complex, which is required for nuclear accumulation of Smad2 (Figs. 3 and 5) and thus for Smad2-mediated signaling (Fig. 2). DISCUSSION In this report we show that endogenous SARA and Smad2 associate in response to TGF-␤1 stimulation in human mesangial cells. Interestingly, under conditions of inhibited endocytic vesicle formation, we see both phosphorylation of Smad2 and its association with Smad4, but not Smad2 nuclear translocation or transcriptional activity. This suggests that neither phosphorylation of endogenous Smad2 nor its heteromerization with Smad4 is sufficient to fully propagate TGF-␤-stimulated Smad signaling. Our data confirm a critical role for endocytosis in TGF-␤ signaling through Smad2.
Based on our data and the data of others, SARA and Smad would associate with the T␤R at the plasma membrane where Smad2 is phosphorylated which allows for its interaction with Smad4. In our proposed model, endocytic internalization would FIG. 9. Smad4 can be associated with the Smad2-SARA complex. A, serum-deprived mesangial cells were incubated with 1 ng/ml TGF-␤1 for the indicated times, followed by lysis and immunoprecipitation of SARA. The blot was developed first for Smad2 (top) and then, without stripping, for Smad4 (middle). Bottom panel shows Smad4 detection from the whole cell lysates (WL). B, Smad4 was immunoprecipitated from serum-deprived mesangial cells treated with control vehicle (Ϫ) or 1 ng/ml TGF-␤1 (ϩ) for 30 min. Immunoprecipitated complexes (IP:Sd4) and whole cell lysates (WL) were analyzed with the indicated antibodies. C, left panels are Western blots of either lysates that were immunoprecipitated with anti-Smad 2/3 (N-19) (Sd2:IP) or WL from cells treated with either vehicle (Ϫ) or with 1 ng/ml TGF-␤1 for 30 min (ϩ). Middle panels represent Western blots of either Smad4 immunoprecipitation from Smad 2/3 supernatant (Sd4:ReIP) or WL. Bf represents immunoprecipitation of RIPA buffer alone as negative control. Right panels represent Western blots of WL or of SARA immunoprecipitated from Smad 2/3 supernatant (SARA:ReIP). In all cases blots were sequentially developed for SARA (top panels), Smad2 (middle panels), and Smad4 (bottom panels). Blots are representative of three separate experiments. allow the Smad2-Smad4 complexes to separate from SARA and enter the nucleus to modulate Smad2-dependent transcription. The robust TGF-␤1-stimulated interaction between SARA and Smad2 in our studies appears to differ from those of the initial report characterizing SARA (6) in which it was suggested that SARA preferentially interacts with unphosphorylated Smads. In that report, Tsukazaki et al. (6) proposed a TGF-␤-induced dissociation of SARA and Smad due to Smad phosphorylation. More recent work (22) supports this model, showing that phosphorylated Smad2 has a reduced affinity for SARA. While we detect the phosphorylated form of Smad2 in association with endogenous SARA, we also demonstrate a separation of Smad2 from SARA after the peak time of Smad2 phosphorylation. However, one novel finding of the present study is that this separation only occurs under conditions, which allow endocytosis. Therefore, our results do not contradict the idea that Smad2 phosphorylation induces dissociation from SARA, but rather suggest that this dissociation is not immediate upon stimulation of the TGF-␤ receptor and requires an intermediate internalization step.
The mechanism by which subsequent internalization of the membrane-localized receptor complex propagates the TGF-␤ signal remains unclear. While receptor-mediated endocytosis can be a means to blunt signals through the direction of receptors away from surface ligand interactions, endocytic vesicles may perform other functions such as the creation of specialized signaling com-partments to enhance signal intensity (11). A further role for endocytosis was recently described by Di Guglielmo et al. (24) who showed that internalization of the TGF-␤ receptor into EEA1 endosomal vesicles led to SARA-associated enhanced signal activation, but that internalization to caveolar compartments led to Smurf2-and Smad7-mediated receptor degradation. If either route of internalization was blocked, there was an increase in distribution to the other vesicular compartment, which suggests that one function of endocytosis may be to enhance signal propagation by sequestering the receptor from its degradationassociated caveolar pathway. However, a recent report by Mitchell et al. (27) suggests that clathrin-mediated endocytosis is the crucial event in both T␤R, rab11-dependent, recycling, and degradation, with raft/caveolae-dependent endocytosis being of minor importance.
Evidence suggests that SARA functions as a cytosolic retention factor (29). SARA binds a site referred to as the "hydrophobic corridor" in the MH2 domain of Smad2. Smad2 also contains a nucleoporin-interacting site that allows Smad2 to interact directly with nucleoporins CAN/Nup214 and Nup153, and this site overlaps the Smad2 SARA binding site. Thus the nuclear transport of Smad2 can be regulated through competitive binding, where TGF-␤-induced Smad2 phosphorylation reduces its affinity for SARA and allows interaction with the nuclear pore complex. The proposed competitive interaction between Smad2-SARA and Smad2-CAN/Nup214 involves di-FIG. 10. Without endocytic internalization, the Smad2-SARA complex does not dissociate. After incubation in hypotonic media, cells were placed in either control or potassium-depleted conditions. A, cells were treated with 1 ng/ml TGF-␤1 for the indicated times leading up to simultaneous harvest. SARA-immunoprecipitated complexes were analyzed by immunoblotting with anti-Smad2 (top) followed by reprobing with anti-SARA (bottom). B, graph representing densitometric analysis of three independent experiments for which the levels of Smad2 co-immunoprecipitated with SARA are corrected for the total immunoprecipitated SARA. Each point is expressed as fold induction over cells treated with vehicle in potassium-control conditions. C, cells were transfected with a CMV-␤-galactosidase reporter construct for 18 h followed by incubation under potassium control (K(ϩ)) or depleted (K(Ϫ)) conditions with TGF-␤1 (black bars) or vehicle (white bars) treatment for 6 h. Alternatively, ␤-galactosidase reporter-transfected cells are incubated under control (K ϩ wo) or depleted (K-wo) conditions, with or without TGF-␤1 for 1 h followed by wash off of media and replacement with 1% NBCScontaining media for an additional 18 h. At the end of each treatment time, cells were lysed and assayed for ␤-galactosidase activity. Bars represent a mean of at least three separate measurements. All measurements are standardized to cells that were transfected but left untreated in 1% NBCS-containing medium for the full time (1%). rect interaction of Smad2 with the nucleoporins. Because we do not observe Smad2 separation from SARA without internalization, a further possible role of the endocytic vesicle in TGF-␤1 signaling might be to aid in the transport of Smad2 to the nucleus. This type of requirement for endocytic vesicular transport in nucleocytoplasmic flux has been described previously for the transcription factor Stat3 in response to EGF stimulation (30).
Another unique finding of our studies is the presence of Smad4 in the SARA immunoprecipitates upon TGF-␤1 treatment. The difference in our data from results described previously showing that Smad2-SARA and Smad2-Smad4 complexes are mutually exclusive, is likely due to the fact that we have investigated only endogenous protein interaction. Nonetheless, cell-type differences cannot be ruled out. Because TGF-␤ treatment generates a variety of responses dependent on cell type and conditions, it is possible that differences in downstream signaling include alternative patterns of SARA-Smad2 or Smad2-Smad4 association.
Finally, because we detected such low levels of Smad3 in association with SARA, we chose to focus our studies on the Smad2-SARA interaction. However, as it is becoming clearer that Smad2 and Smad3 generate different downstream signals upon TGF-␤ stimulation, it would be interesting to examine potential differences in the requirement for SARA association or endocytosis in Smad3 signaling. One report of SARA-Smad3 association shows SARA preferentially binds monomeric Smad3 with phosphorylation-induced dissociation and activity followed by Smad3 trimerization leading to Ski-mediated down-regulation of Smad3 (31). Additionally, a recent article by Lin et al. (10) examined the role of the promyelocytic leukemia protein (pml) as a potential bringing protein between Smad2/3 and SARA. This report showed that add-back of wild-type pml, but not a Smad3-binding deficient mutant pml, restored TGF-␤1 responsiveness in pml-null fibroblasts. In contrast, another group has suggested that TGF-␤-induced Smad3 signaling may not require its association with SARA (14).
In summary, we have shown that endogenous Smad2 associates with SARA following TGF-␤1 treatment in human mesangial cells. Endocytosis, while not critical for Smad2 phosphorylation or for Smad4 association with the receptor-bound Smad2-SARA complex, is required downstream in order to facilitate the separation of the Smad complex from SARA and allow its nuclear accumulation and transcriptional activity.