Transforming Growth Factor β Activates Smad2 in the Absence of Receptor Endocytosis*

Like many other cell surface receptors, transforming growth factor β (TGF-β) receptors are internalized upon ligand stimulation. Given that the signaling-facilitating molecules Smad anchor for receptor activation (SARA) and Hrs are mainly localized in early endosomes, it was unclear whether receptor internalization is required for Smad2 activation. Using reversible biotin labeling, we directly monitored internalization of the TGF-β type I receptor. Our data indicate that TGF-β type I receptor is endocytosed via a clathrin-dependent mechanism and is effectively blocked by depletion of intracellular potassium or by expression of a mutant dynamin (K44A). However, blockage of receptor endocytosis by these two means has no effect on TGF-β-mediated Smad2 activation. Furthermore, TGF-β-induced Smad2 activation was unaffected by inhibition of hVPS34 activity with wortmannin or inhibitory anti-hVPS34 antibodies. Finally, we demonstrated that Smad2 interacted with cell surface receptors and that a SARA binding-deficient Smad2 mutant was phosphorylated by the receptors. Thus, our findings suggest that receptor endocytosis is dispersible for TGF-β-mediated activation of Smad2 and that this activation can be mediated by both SARA-dependent and -independent mechanisms.

Receptor endocytosis has long been regarded as an attenuation mechanism to switch off receptor signaling. However, accumulating evidence suggests that endocytosis may facilitate signaling by targeting signaling complexes to specific subcellular localization, either to increase access of activated receptor kinases to their substrates or to compartmentalize signaling complexes (1)(2)(3). Consistent with this idea, blocking of clathrinmediated endocytosis was found to attenuate the activation of the extracellular signal-regulated kinases by receptor tyrosine kinases and G-protein-coupled receptors (4 -7). Upon the activation of G-protein-coupled proteinase-activated receptor 2, a multiprotein signaling complex that contains ␤-arrestin 1, Raf-1, and extracellular signal-regulated kinases is formed on endocytic vesicles as well (8).
TGF-␤ 1 binds to its cell surface receptors, resulting in the formation of type I and type II receptor complexes. In the complex, the TGF-␤ type II receptor (T␤RII) phosphorylates and activates the TGF-␤ type I receptor (T␤RI), which in turn phosphorylates the C-terminal serine residues of Smad proteins Smad2 and Smad3. As a result, Smad proteins accumulate in the nucleus, bind to DNA, and regulate transcription. The ligand-stimulated receptor complexes undergo endocytosis and are eventually degraded in an ubiquitin/lysosome-dependent pathway (9,10).
The FYVE domain-containing proteins Smad anchor for receptor activation (SARA) and Hrs (the hepatocyte growth factor-regulated tyrosine kinase substrate) have been suggested to facilitate Smad2 phosphorylation by bringing Smad2 to TGF-␤ or activin receptors (11,12). Interestingly, immunofluorescence studies revealed that both SARA and Hrs are predominantly localized in early endosomes (11,13). This finding raises an important question: does Smad2 phosphorylation occur at the plasma membrane, where the receptors are exposed to TGF-␤, or in early endosomes, where the signal-facilitating molecules SARA and Hrs are mainly localized? To investigate the role of TGF-␤ receptor endocytosis in Smad2 activation, we directly followed endocytosis of T␤RI using a cell surface biotinylation protocol. Our data reveal that T␤RI is rapidly internalized via clathrin-mediated endocytosis but that inhibition of receptor internalization does not block Smad2 activation.

EXPERIMENTAL PROCEDURES
Cell Culture and Materials-L17-T␤RI cells and HeLa cells stably expressing wild-type and K44A dynamin were maintained as described previously (14,15). TGF-␤1 was purchased from R&D Systems, wortmannin was purchased from Sigma, anti-phospho-Smad2 antibody was purchased from Upstate Biotechnology, anti-phospho-Thr308-AKT and anti-AKT were purchased from Cell Signaling Technology, and other antibodies were purchased from Santa Cruz Biotechnology. Anti-VPS34 antibodies have been described previously (22). Unless indicated, all chemicals were from Fisher/ICN. SARA(dFYVE) and SARA(665-end) were generated by PCR-based deletion, and the sequences were confirmed by DNA sequencing.
Transfection, reporter assay, immunoprecipitation, Western analysis, and immunofluorescence microscopy were performed as described previously (16). Receptor affinity labeling was carried out as described previously (17) 1 The abbreviations used are: TGF-␤, transforming growth factor ␤; T␤RI, TGF-␤ type I receptor; T␤RII, TGF-␤ type II receptor; SARA, Smad anchor for receptor activation; PI(3)P, phosphatidylinositol 3-phosphate; HA, hemagglutinin. formed as described by Larkin et al. (18). Cells were treated with hypotonic medium (50% Dulbecco's modified Eagle's medium plus 50% H 2 O) for 10 min and with isotonic buffer (10 mM Tris, pH 7.5, 150 mM NaCl) with or without 10 mM KCl for 30 min. TGF-␤1 was then added to a final concentration of 100 pM, and the cells were incubated for another 30 min. Cell lysates were harvested for Western analysis.
Biotinylation and Receptor Endocytosis-L17-T␤RI cells were maintained in growth medium in the absence of tetracycline for 16 -24 h before biotinylation. Biotinylation was performed as described previously (19). After labeling with 0.5 mg/ml NHS-SS-biotin (Pierce), the cells were shifted to 37°C for the indicated time in the presence or absence of 100 pM TGF-␤1. Endocytosis was then stopped by transferring cells back to 4°C. After treatment with reducing solution (15.5 mg/ml glutathione, 75 mM NaCl, 75 mM NaOH, and 10% fetal bovine serum) and 5 mg/ml iodoacetamide (Sigma) in phosphate-buffered saline containing 0.8 mM MgCl 2 , 1.0 mM CaCl 2 plus 1% bovine serum albumin, the cells were lysed in TNE (10 mM Tris, pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40, and 1 mM EDTA). Equal amounts of cell lysates were used for precipitation of biotinylated proteins with streptavidin beads (Pierce). Precipitated proteins were eluted and analyzed by immunoblotting.

RESULTS
Rapid Clathrin-dependent Internalization of T␤RI-To study the relationship between receptor signaling and endocytosis, we chose to directly monitor T␤RI internalization in the stable cell line L17-T␤RI that is derived from the T␤RI-deficient mink lung epithelial L17 cells and expresses HA-tagged T␤RI under the control of tetracycline (15). Cell surface proteins were labeled at 4°C using a reducible biotinylation method (19). After additional incubation of labeled cells for various times at 37°C, biotin remaining at the cell surface was removed by incubation with impermeable reducing agents at 4°C. Internalized receptors, which are protected from reduction, were then precipitated with streptavidin beads and analyzed by immunoblotting.
First, we examined the ligand dependence of T␤RI internalization. The cell surface T␤RI was well labeled, and treatment of the cells with reducing agents efficiently removed biotin from the cell surface ( Fig. 1A, lanes 1 and 2). After the cells were shifted to 37°C, T␤RI was rapidly internalized, and its internalization was further increased by TGF-␤ treatment ( Fig 1A,  lanes 3-8; Fig. 1B).
To examine whether T␤RI endocytosis is a clathrin-mediated process, we used two protocols that have been previously demonstrated to disrupt clathrin-mediated vesicle formation. Depletion of intracellular potassium decreases clathrin-coated pit formation and thus inhibits clathrin-dependent endocytosis (18). Clathrin-mediated endocytosis is also regulated by the GTPase dynamin, and expression of dynamin mutants defective in GTP binding or hydrolysis blocks the endocytosis of transferrin and epidermal growth factor receptors (20).
L17-T␤RI cells were treated with hypotonic medium followed by isotonic potassium-free buffer to deplete intracellular potassium. Potassium depletion significantly inhibited T␤RI endocytosis ( Fig. 2A, lane 4). However, ligand-mediated T␤RI endocytosis was restored when 10 mM KCl was included in the buffer ( Fig. 2A, lane 5). Therefore, depletion of intracellular potassium potently blocked T␤RI internalization.
To confirm the clathrin dependence of T␤RI internalization, we employed stable HeLa cell lines that express wild-type dynamin or the K44A mutant under the control of tetracycline FIG. 1. Rapid internalization of T␤RI. A, L17-T␤RI cells were biotin-labeled at 4°C and then incubated at 37°C with or without 100 pM TGF-␤1 for the indicated time. The cells were transferred back to 4°C and treated with a reducing buffer. Biotinylated T␤RI was isolated with streptavidin beads and analyzed by anti-HA immunoblotting. One representative experiment is shown here. Lane 1 (TL), total labeled cell surface T␤RI; lane 2 (BG), reducing control; lanes 3-5 (ϪTGFb), internalized T␤RI in the absence of TGF-␤1; lanes 6 -8 (ϩTGFb), internalized T␤RI in the presence of TGF-␤. B, quantitation of T␤RI internalization. The bands were scanned and quantitated using NIH Image 1.6. The internalized T␤RI is expressed as a percentage of total labeled T␤RI (total labeled ϭ 100%) after subtraction of the background in the reducing control. The data were obtained from three independent experiments.

FIG. 2. T␤RI is internalized via a clathrin-mediated pathway.
A, potassium depletion blocks T␤RI endocytosis. L17-T␤RI cells were  (14); dynamin expression in these cells is induced by the withdrawal of tetracycline. The cells were transfected with HAtagged T␤RI and maintained in the growth medium in the absence of tetracycline, and T␤RI endocytosis was measured. As shown in Fig. 2B, T␤RI is efficiently endocytosed in HeLa cells expressing wild-type dynamin (lanes 3 and 4); a high level of ligand-independent T␤RI internalization was observed, which could be due to the high level of receptor expression. Nonetheless, expression of K44A dynamin completely blocked T␤RI internalization (Fig. 2B, lanes 7 and 8).
Smad2 Activation Is Unaffected by Inhibition of T␤RI Endocytosis-To test whether internalization of TGF-␤ receptors is required for Smad activation, L17-T␤RI cells were subjected to depletion of intracellular potassium. The cells were then stimulated with TGF-␤, and Smad2 phosphorylation was detected by immunoblotting with anti-phospho-Smad2 antibodies (Fig.  3A). Although potassium depletion potently inhibited T␤RI internalization, it had no effect on TGF-␤-mediated Smad2 phosphorylation (compare lanes 4 and 6). Similar results were obtained with TGF-␤1 at a concentration as low as 10 pM and were also observed in HeLa and Hep3B cells (data not shown).
To further study whether clathrin-mediated endocytosis is essential for Smad2 activation by TGF-␤, we examined Smad2 phosphorylation in dynamin-expressing HeLa cells. TGF-␤ stimulated the phosphorylation of Smad2 in parental HeLa cells (Fig. 3B, lanes 1 and 2), and this phosphorylation was not inhibited by the expression of wild-type or K44A dynamin (Fig.  3B, lanes 3-10). We also examined the effect of wild-type and mutant dynamin on TGF-␤-induced nuclear accumulation of Smad2. Smad2 is localized in the cytoplasm at the basal state, and TGF-␤ treatment resulted in its translocation into the nucleus (Fig. 3C). Expression of either wild-type or K44A dynamin had no effect on TGF-␤-induced nuclear accumulation of Smad2 (Fig. 3C).
To examine whether T␤RI endocytosis is necessary for the transcriptional activity of TGF-␤, we examined the expression of two TGF-␤-responsive reporters (3TP-luciferase and AREluciferase), whose expression is Smad-mediated. The 3TP-lu- FIG. 4. Inhibition of hVPS34 activity has no effect on Smad2 activation. A, wortmannin does not interfere with TGF-␤-induced Smad2 phosphorylation. L17-T␤RI cells were treated with 100 nM wortmannin (Wort) or solvent DMSO (Ϫ) for 30 min, followed by stimulation with 100 pM TGF-␤1 (Tb) or 100 nM insulin (Ins) for another 30 min. Smad2 phosphorylation was examined by anti-phospho-Smad2 immunoblotting (left panels), and AKT phosphorylation was examined with anti-phospho-Thr308-AKT (right panels). Protein expression is shown in the bottom panels. B, TGF-␤-mediated nuclear accumulation of Smad2 is unaffected by anti-hVPS34 inhibitory antibodies. HepG2 cells were microinjected with anti-hVPS34 antibodies (d) or control IgG (c) and treated with TGF-␤1 for 30 min (bϪd). The cells were stained with anti-Smad2 antibodies. Asterisks indicate injected cells. The data are representative of four separate experiments.

FIG. 5. L17-T␤RI cells were transfected with 3TP-luciferase construct with or without various forms of SARA constructs.
After TGF-␤1 treatment for 20 h, luciferase activity was determined. Relative luciferase activities are expressed as the mean Ϯ S.D. from triplicates. Similar results were obtained from three different experiments.
ciferase construct was transiently transfected into parental or dynamin-expressing HeLa cells, and reporter expression was measured after TGF-␤ treatment. TGF-␤-induced expression of this reporter was not inhibited by expression of either wild-type dynamin or K44A mutant (Fig. 3D, left panel). We further confirmed this result in Hep3B cells using the ARE-luciferase reporter. Expression of wild-type dynamin enhanced ARE-luciferase expression, which is consistent with slightly higher Smad2 phosphorylation in cells expressing wild-type dynamin (Fig. 3B, lane 6). However, expression of the T65A and K142A dynamin mutants, which have been shown to block transferrin uptake (21), did not inhibit TGF-␤ induction of the ARE reporter (Fig. 3D, right panel). Taken together, these data strongly suggest that T␤RI endocytosis is not required for Smad2 activation or Smad-mediated reporter expression in TGF-␤-stimulated cells.
Inhibition of hVPS34p Activity Has No Effect on Smad2 Activation-The FYVE domain of SARA binds to PI(3)P, and deletion of the FYVE domain-containing N terminus of SARA antagonizes TGF-␤-stimulated expression of ARE-luciferase (11). These data suggest that production of PI(3)P, presumably in the early endosome, is important for T␤RI signaling. To test this hypothesis, we treated L17-T␤RI cells with 100 nM wortmannin, which inhibits all phosphatidylinositol 3-kinases except the class II phosphatidylinositol 3-kinase C2␣ isoform. As shown in Fig. 4A, wortmannin had no effect on TGF-␤-stimulated Smad2 phosphorylation (lane 3), although it inhibited insulin-mediated Akt/protein kinase B phosphorylation at threonine 308 (lane 10). Similar results were obtained in Hep3B cells (data not shown). In addition, treatment of Hela cells with wortmannin did not affect TGF-␤-induced Smad2 nuclear localization (data not shown).
To specifically examine the role of PI(3)P in TGF-␤ signaling, we microinjected HepG2 cells with specific inhibitory antibodies against the class III phosphatidylinositol 3-kinase, hVPS34. These antibodies disrupt the trafficking of internalized platelet-derived growth factor receptors and the endosomal localization of EEA1 (22). Moreover, they completely disrupt the localization of the intracellular PI(3)P marker 2X-FYVE-GFP fusion protein in Chinese hamster ovary (23) and HepG2 cells (data not shown). However, microinjection of inhibitory anti-VPS34 antibodies had no effect on TGF-␤-stimulated nuclear accumulation of Smad2 in HepG2 cells (Fig. 4B).
To directly address whether the FYVE domain is required for SARA function, we examined TGF-␤-induced expression of 3TP-luciferase in the presence of SARA(dFYVE), a mutant that lacks the FYVE domain. This mutant was shown to lose a punctate subcellular localization but still interact with Smad2 (11). We reasoned that this mutant should have a dominant negative effect on TGF-␤ signaling by sequestering Smad2 and Smad3 away from the membrane. However, like the wild-type SARA, SARA(dFYVE) has no effect on TGF-␤ activity in mediating the expression of 3TP-luciferase. This is in contrast to SARA(665-end), a mutant lacking both the N terminus and the FYVE domain, which does inhibit TGF-␤ function (Fig. 5) (11). These results suggest that interactions between SARA and intracellular PI(3)P are not required for Smad2 activation.
Smad2 Is Recruited to the Receptor Complexes on the Plasma Membrane-To address whether Smad2 is activated on the plasma membrane, the receptor-Smad2 complexes were examined when receptor endocytosis was blocked. COS1 cells transfected with FLAG-tagged Smad1 or Smad2, T␤RI, and T␤RII were subjected to potassium depletion and then incubated with 125 I-TGF-␤1 to label cell surface receptors. After treatment with a cross-linking reagent, receptor-Smad complexes were examined by anti-FLAG immunoprecipitation and visualized by SDS-PAGE and autoradiography. Because the interaction between T␤RI and Smad2 is transient, Smad2 stably associated with the receptors only when the kinase-deficient T␤RI mutant was expressed (Fig. 6A, lane 3). Importantly, Smad2 still interacted with TGF-␤ receptor complex even when receptor endocytosis was inhibited by potassium depletion (lane 5), strongly indicating that Smad2 is recruited into receptor complexes on the plasma membrane.
To further investigate the role of SARA in Smad2 activation, we examined phosphorylation of Smad2(N381S), a mutant deficient in SARA-binding (24), in dynamin(K44A)-expressing HeLa cells. FLAG-tagged Smad2(N381S) was transfected into dynamin(K44A))-expressing HeLa cells together with T␤RI and T␤RII constructs. After induction of dynamin K44A expression for 16 h, the cells were treated with TGF-␤1, and Smad2(N381S) phosphorylation was revealed by anti-phospho-Smad2 immunoblotting after anti-FLAG immunoprecipitation. As shown in Fig. 6B, Smad2(N381S) was phosphorylated when coexpressed with TGF-␤ receptors, and this phosphorylation was further stimulated by TGF-␤. Phosphorylation of Smad2(N381S) in the absence of TGF-␤ could be due to protein overexpression. Nonetheless, these results demonstrate that SARA binding activity is not essential for receptor-mediated phosphorylation of Smad2. This phosphorylation is presumably mediated by the cell surface receptors because it is not abolished by inhibition of receptor endocytosis. DISCUSSION Our finding that T␤RI is internalized via clathrin-mediated endocytosis is in agreement with the earlier studies, which utilized granulocyte macrophage colony-stimulating factor-TGF-␤ chimeric receptors (25). The internalization of these chimeric receptors was abolished by potassium depletion. Furthermore, recent studies also suggest that internalization of T␤RII is clathrin-mediated (26). However, uptake of 125 I-TGF-␤1 was not blocked by chloroquine or transient expression of K44A dynamin (27). The mechanism of TGF-␤ uptake may differ from that of T␤RI internalization because TGF-␤ uptake can be mediated by different binding proteins, such as ␤-glycan. Thus, TGF-␤ uptake reflects the sum of internalized TGF-␤-bound proteins, some of which may utilize clathrin-independent internalization mechanisms.
SARA has been shown to play a role in Smad-mediated signaling. Given the endosomal localization of this protein, a model has emerged in which the internalization of T␤RI receptors into early endosomes facilitates interactions with SARA and enhances TGF-␤ signaling. In this study, we have tested several aspects of this model. We demonstrate that T␤RI internalizes via coated vesicles. However, inhibition of this process using two different methods has no effect on Smad2 phosphorylation and nuclear translocation and has no effect on TGF-␤-stimulated transcription. Furthermore, we find that disruption of the interaction between PI(3)P and the SARA FYVE domain, either by inhibiting cellular phosphatidylinositol 3-kinases with wortmannin, by specifically inhibiting hVPS34 with antibodies, or by deleting the FYVE domain, has no effect on TGF-␤-mediated signaling. In addition, we provide evidence that Smad2 can be recruited directly to cell surface TGF-␤ receptors. Therefore, we conclude that T␤RI endocytosis is qualitatively dispersible for Smad2 activation and that interactions between the FYVE domain of SARA and PI(3)P are not necessary for Smad2-mediated TGF-␤ signaling. We further demonstrate that the SARA binding-deficient mutant Smad2(N381S) is still phosphorylated by activated TGF-␤ receptors. This is consistent with our previous results showing that both wild-type Smad2 and Smad2(N381S) can stimulate ARE-luciferase expression, but only the activity of wild-type Smad2 is enhanced by SARA (24). Thus, receptor-mediated Smad2 activation is mediated by both SARA-dependent and -independent mechanisms.
Our results do not rule out the possibility that receptor endocytosis may contribute to the maximal activation of Smad2. Although not required for TGF-␤ signaling, interactions with endosomal SARA could modulate the intensity or lifetime of TGF-␤ signaling in the presence of submaximal levels of ligand. This is suggested by recent experiments in which a dominant negative Rab5 increases TGF-␤ signaling presumably by influencing early endosomal dynamics (28). Indeed, we observed that overexpression of wild-type dynamin slightly enhanced Smad2 phosphorylation. This suggests that overexpression of wild-type dynamin may accelerate receptor endocytosis and thereby facilitate Smad2 activation. Alternatively, enhanced Smad2 phosphorylation may be due to endocytosis-independent effects of dynamin on other signaling path-ways, as has been suggested for mitogenic and apoptotic signal transduction (29,30).
Receptor endocytosis has been shown to be required for certain signaling pathways but not for others. For instance, isoproterenol-stimulated extracellular signal-regulated kinase activation is inhibited by K44A dynamin, whereas activation of adenylate cyclase and phospholipase C via trimeric G-proteins is unaffected (7). Similarly, blockage of insulin-like growth factor receptor internalization interferes with the Shc/mitogenactivated protein kinase pathway, but not with the insulin receptor substrate-1 pathway (6). In addition to the Smad pathway, TGF-␤ has also been implicated to signal through other pathways, such as mitogen-activated protein kinases (reviewed in Ref. 31). It remains to be determined whether these signaling pathways are dependent on TGF-␤ receptor endocytosis.