Smad7 Protein Interacts with Receptor-regulated Smads (R-Smads) to Inhibit Transforming Growth Factor-β (TGF-β)/Smad Signaling*

TGF-β is a pleiotropic cytokine that regulates a wide range of cellular actions and pathophysiological processes. TGF-β signaling is spatiotemporally fine-tuned. As a key negative regulator of TGF-β signaling, Smad7 exerts its inhibitory effects by blocking receptor activity, inducing receptor degradation or interfering with Smad-DNA binding. However, the functions and the molecular mechanisms underlying the actions of Smad7 in TGF-β signaling are still not fully understood. In this study we report a novel mechanism whereby Smad7 antagonizes TGF-β signaling at the Smad level. Smad7 oligomerized with R-Smad proteins upon TGF-β signaling and directly inhibited R-Smad activity, as assessed by Gal4-luciferase reporter assays. Mechanistically, Smad7 competes with Smad4 to associate with R-Smads and recruits the E3 ubiquitin ligase NEDD4L to activated R-Smads, leading to their polyubiquitination and proteasomal degradation. Similar to the R-Smad-Smad4 oligomerization, the interaction between R-Smads and Smad7 is mediated by their mad homology 2 (MH2) domains. A positive-charged basic region including the L3/β8 loop-strand module and adjacent amino acids in the MH2 domain of Smad7 is essential for the interaction. These results shed new light on the regulation of TGF-β signaling by Smad7.

The TGF-␤ signaling pathway has been well documented (4, 9 -12). TGF-␤ family cytokines initiate signal transduction by binding to the receptor complex, wherein the type II receptor T␤RII activates the type I receptor T␤RI via phosphorylation. Then T␤RI in turn phosphorylates the C-terminal Ser-Xaa-Ser motif of receptor-regulated Smads (R-Smads, Smad2/3 for TGF-␤, and Smad1/5/8 for BMPs), leading to their oligomerization with the common Smad (Co-Smad, Smad4). The oligomeric Smad complex is then accumulated in the nucleus and controls target gene expression, in collaboration with other transcription factors or co-factors. Structurally, both R-Smads and Co-Smad consist of a conserved MH1 domain that mediates DNA binding (except Smad2) and a MH2 domain that mediates Smad oligomerization, Smad-receptor interaction, etc. The two domains are bridged by a proline-rich linker region that is divergent in length and amino acid sequence (9,11). Besides this canonical Smad pathway, other signaling molecules are reported to transduce signals from TGF-␤ and its receptors, such as mitogen-activated protein kinases (MAPKs, ERK, JNK, and p38), PI3K/Akt, RhoA GTPase, PAK2 and others in a context-dependent manner (13,14).
Although the function of Smad7 in TGF-␤ signaling has been extensively studied, the underlying molecular mechanisms by which Smad7 exerts its regulatory roles are not fully understood. In this study we uncover a novel mechanism whereby Smad7 directly inhibits R-Smad activity. Upon TGF-␤ treatment, Smad7 forms a heteromeric complex with R-Smads through the MH2 domain and hence interferes with R-Smad-Smad4 oligomerization in a competitive manner. In addition, Smad7 recruits the E3 ubiquitin ligase NEDD4L to the activated R-Smads, resulting in their polyubiquitination and degradation. Together, these results advance our understanding of the molecular functions of Smad7 in regulating TGF-␤ signaling.
Cell Culture and Transfection-Human embryonic kidney epithelial HEK293FT cells were maintained in Dulbecco's minimum essential medium (DMEM) (Corning) supplemented with 10% of fetal bovine serum (Gibco) at 37°C in a humidified, 5% CO 2 incubator. Human hepatocellular carcinoma Hep3B and T␤RI-deficient mink lung epithelial R1B/L17 cells were maintained in minimum essential medium (Corning), and MDA-MB-231 cells were maintained in RPMI 1640 (Corning) by the addition of 2 mM L-glutamine. Cell transfection was conducted with VigoFect (Vigorous Biotechnology, Beijing) or Lipofectamine 2000 (Invitrogen). The construct encoding  FLAG-tagged Smad7 under the control of CMV promoter and  carrying the puromycin-resistant gene was transfected into  MDA-MB-231 cells, paralleled by transfection of the control  empty vector. After puromycin selection, the drug-resistant  cells were pooled as stable cells. Luciferase Reporter Assay, Total RNA Extraction, Reverse Transcription (RT), and Quantitative RT-PCR-Cells were plated in 24-well plates one night before transfection. Transfection was performed as described above, and empty vectors were used to equalize the total amounts of plasmids in each sample. Luciferase activity was measured at 40 h post-transfection by using the dual luciferase reporter assay system (Promega, Madison, WI) following the manufacturer's protocol. The experiments were repeated in triplicate, and the data are presented as the means Ϯ S.D. after normalization to Renilla activity. Total cell RNA extraction, reverse transcription, and quantitative RT-PCR were described previously (30). The primers used were as follows: for human GAPDH (glyceraldehyde-3-phosphate dehydrogenase), 5Ј-ACCACAGTCCAT GCCATCAC-3Ј and 5Ј-TCCACCACCCTGTTGCTGTA-3Ј; for human Smad7, 5Ј-CCAACTGCAGACTGTCCAGA-3Ј and 5Ј-TTCTCCTCC-CAGTATGCCAC-3Ј.

Stable Cell Line Establishment-
Immunoprecipitation and Immunoblotting-Cells for immunoprecipitation (IP) were lysed on ice with lysis solution (50 mM Tris-Cl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, 10 mM NaF, 10 mM sodium pyrophosphate, 1 mM sodium orthovanadate dodecahydrate, and protease inhibitors) and rotated for more than 10 min at 4°C. After an aliquot was taken for protein expression analyses, the left cell lysates containing equivalent amounts of total proteins were precleared for 2 h with protein A-Sepharose (GE Healthcare) at 4°C. Immunoprecipitation was carried out by the addition of appropriate antibodies and protein A-Sepharose followed by incubation at 4°C overnight with gentle rotation. Then the immune complex was isolated by centrifugation and repeated washes with lysis buffer, analyzed by SDS-PAGE and immunoblotting, and detected with the enhanced chemiluminescent substrate (Pierce) according to the manufacturer's instructions.
Protein Turnover Analysis-HEK293FT cells were transfected with indicated plasmids in 12-well plates, and the amounts of plasmids in each sample were equalized by the addition of empty vectors. The cells were treated with 50 g/ml cycloheximide for the indicated time periods before harvest. Then cells were lysed for protein level analyses by immunoblotting.
In Vivo Ubiquitination Assay-HEK293FT cells were transfected with His-Myc-ubiquitin plasmid and other constructs as indicated. At 40 h post-transfection, cells were treated with 10 M MG132 for 4 h then collected and lysed in 1% SDS followed by 5 min of boiling. The lysates were immediately diluted 10-fold with lysis buffer (20 mM Tris-HCl (pH 7.4), 2 mM EDTA, 10 mM NaF, 10 mM sodium pyrophosphate, 1 mM sodium orthovanadate dodecahydrate, 1% Triton X-100) plus protease inhibitors and rotated for 30 min at 4°C. After centrifugation at 13,000 rpm for 10 min, an aliquot of the supernatants was taken for protein expression analyses, and the remaining lysates were subjected to immunoprecipitation and immunoblotting detection.
Homology Modeling of Smad7 MH2 Domain Structure-The structural model of Smad7 MH2 domain was constructed based on the three-dimensional structure of C-terminally phosphorylated Smad3 MH2 domain (RCSB Protein Data Bank (PDB) code 1U7F) using SWISS-MODEL.
Statistic Analysis-All the experiments were repeated at least three times. The values were presented as the mean Ϯ S.D., and the significance between the means was calculated using Student's t test. A p value Ͻ0.05 was considered as statistically significant.

Smad7 Inhibits the Signaling Activity of R-Smads-Although
Smad7 has been established as a pivotal negative regulator of TGF-␤ signaling, the underlying molecular mechanisms still need a better understanding (8,15). To clarify at what levels Smad7 acts in TGF-␤ signaling, we first carried out a reporter assay in HEK293FT cells using the Smad-responsive CAGAluciferase reporter (36). As shown in Fig. 1A, treatment with TGF-␤1, overexpression of the constitutively active (ca-) type I TGF-␤ receptor T␤RI, Smad3 alone, or Smad3 together with Smad4 stimulated reporter expression. Co-expression of Smad7 blocked all of these effects, indicating that Smad7 can function at both the receptor level and the Smad level. In addition, Smad7 also inhibited the expression of CAGA-luciferase induced by ectopic expression of Smad3 with or without Smad4 in the T␤RI-deficient mink lung epithelial R1B/L17 cells (Fig.  1B), further indicating that Smad7 is capable of inhibiting TGF-␤ signaling at the Smad level.
Smad7 has been demonstrated to inhibit Smad-driven transcription by competing with the R-Smad-Smad4 complex for DNA binding (21). To address whether Smad7 hampers the binding of R-Smad-Smad4 to DNA or directly inhibits R-Smad activity, we employed the Gal4-luciferase reporter in which luciferase expression is driven by a promoter that includes repeated Gal4 binding elements, and fusion proteins of Gal4 DBD with Smad3 or other transcription activators are able to induce the reporter expression ( Fig. 1C) (31). The Gal4-luciferase reporter was activated by Gal4-Smad3 overexpression, and TGF-␤1 further promoted the activation (Fig. 1D). However, Smad7 attenuated the Smad3-induced Gal4-luciferase expression even in the presence of the T␤RI kinase inhibitor SB431542, strongly suggesting that the effect of Smad7 is independent of receptor activity. Notably, SB431542 attenuated the activity of Smad3, indicating that the stimulating effect of Smad3 is partially due to autocrine TGF-␤ activity. A similar Gal4-luciferase reporter assay in R1B/L17 cells showed that . Smad7 knockdown efficiency was tested by transfection of NS and Smad7-specific shRNAs into HEK293FT cells followed by total mRNA extraction and RT-quantitative RT-PCR analysis (right). Smad7 mRNA expression level was normalized to that of GAPDH. **, p Ͻ 0.01. In all the reporter assays, empty vectors were used to equalize the total amounts of plasmids in each sample. Each experiment was performed in triplicate, and the data are presented as the mean Ϯ S.D. after normalization to Renilla activity.
both the Gal4-Smad2 and Gal4-Smad3 fusion proteins activated the reporter expression, and the activations were reinforced by Smad4 but inhibited by Smad7 (Fig. 1E). Furthermore, Smad7 inhibited Smad3 activity in a dose-dependent manner (Fig. 1F). Moreover, Smad7 was able to inhibit Smad1 activity in a similar manner (Fig. 1G). Although Smad6 was less effective in attenuating Smad3 activity, it efficiently blocked Smad1 activity (Fig. 1, F and G). Finally, to confirm the physiological role of endogenous Smad7 in regulating Smad3 activity, Smad7 was knocked down with a specific shRNA (Fig. 1H, right) (33). Consistently, silencing of Smad7 expression enhanced the Gal4-Smad3-induced expression of the Gal4 reporter (Fig. 1H, left). Together, these results indicate that Smad7 can inhibit the signaling activity of R-Smads downstream of the receptors, in addition to interfering with the binding of R-Smad-Smad4 complex to DNA.
Smad7 Interacts with R-Smads in Response to TGF-␤ Signaling-To explore how Smad7 inhibits Smad2/3 activity, we first examined whether Smad7 could oligomerize with R-Smad proteins. Hep3B cells treated with or without TGF-␤1 ligands were subjected to co-IP. Indeed, TGF-␤ induced the association of Smad2/3 not only with Smad4 but also with Smad7 ( Fig. 2A). To further characterize their interactions, we expressed Smad7, R-Smad proteins (Smad2/3 and Smad1/8), Smad4, and the other inhibitory Smad Smad6 in HEK293FT cells and performed co-IP experiments. Ectopic expression of Smad7 interacted with R-Smads involved in both TGF-␤ and BMP pathways at the basal level, and these interactions were greatly reinforced by co-expression of ca-T␤RI (Fig. 2B) or ca-ALK6/BMPRIB (Fig. 2C). Furthermore, we also noticed the existence of Smad7-Smad6 and Smad7-Smad4 associations (Fig. 2, B-C), which is consistent with a previous study reporting that Smad7 facilitates Smad4 degradation (37). To determine whether TGF-␤-induced carboxyl phosphorylation of R-Smads is required for their interaction with Smad7, we compared the binding of Smad7 to wild-type (WT) Smad3, Smad3(2D) mutant in which the last two serine residues at the C-terminal tail are mutated to aspartic acid to mimic phosphorylation (38), or Smad3(3A) mutant that contains three serineto-alanine mutations at the C-terminal tail and cannot be activated by TGF-␤. As shown in Fig. 2D, Smad7 interacted strongly with ca-T␤RI-phosphorylated wild-type Smad3 and 2D mutant but weakly with non-activated wild-type Smad3 or 3A mutant. Accordantly, wild-type Smad3 was able to induce the CAGA-luciferase reporter expression, and TGF-␤1 enhanced the effect of wild-type Smad3, comparable with that of 2D mutant (Fig. 2E). Furthermore, Smad7 attenuated the expression of the reporter induced by Smad3 (Fig. 2E).
Then we further confirmed the Smad7-Smad3 interaction using bimolecular fluorescence complementation approach. As demonstrated previously (32) and illustrated in Fig. 2F, if protein A and B could associate with one another in vivo, the protein A-fused N-terminal part of EYPF (EYFP(N)) and protein B-fused C-terminal part of Venus (Venus(C)) would be brought together and reconstitute functional fluorescent proteins, emitting green fluorescence. Indeed, co-expression of EYFP(N)-Smad3 fusion protein with Venus(C)-Smad4 gave rise to green fluorescence in the nucleus upon TGF-␤ treatment (Fig. 2G, upper panels), whereas co-expression of EYFP(N) with Venus(C)-Smad4 had no such effect (data not shown), indicative of in vivo specific interaction between Smad3 and Smad4. In contrast, TGF-␤-induced complex of EYFP(N)-Smad3 and Venus(C)-Smad7 were mainly localized in the cytoplasm, although a weak signal could also be detected in the nucleus (Fig. 2G, lower panels).
As TGF-␤ and BMP induce both the R-Smad-Smad4 association that transduces signals and the R-Smad-Smad7 interaction that interferes with signal transduction, the two events seem to contradict each other. To elucidate the physiological consequence of these interactions, we examined the kinetics of the R-Smad-Smad7 interaction in MDA-MB-231 cells that stably express FLAG-Smad7. The cells were treated with TGF-␤1, harvested at different time points, and subjected to anti-Smad2/3 immunoprecipitation and anti-Smad4 or anit-Smad7 immunoblotting. The Smad2/3-Smad4 association appeared quickly after ligand treatment, peaked at 1 h, and then decreased slowly (Fig. 2H). The kinetics was very similar to the one of phospho-Smad2. Similarly, Smad7 also associated with Smad2/3 at 0.5 h and lasted 1 h, then their associations quickly weakened. Based on these results, we reasoned that Smad4 and Smad7 might compete in interacting with the phosphorylated form of R-Smads, and the balance between Smad4-binding and Smad7-binding of R-Smads would determine the intensity or duration of TGF-␤ signaling.
The MH2 Domains of Smad7 and Smad3 Mediate Their Interaction-To test our hypothesis that Smad7 competes with Smad4 in binding to activated R-Smads, we first characterized the interaction between Smad7 and Smad3 by mapping the interacting domains. As shown in Fig. 3A, both the full-length (FL) Smad3 and its truncations containing the MH2 domain were able to interact with Smad7, suggesting that the MH2 domain of Smad3 mediates its interaction with Smad7. Similarly, the MH2 domain of Smad7 was also involved in Smad3 interaction (Fig. 3, B and E). Functionally, the Smad7 truncations containing the MH2 domain (including 90 -426 amino acids, 180 -426 amino acids, and the MH2 domain (260 -426 amino acids)) were capable of inhibiting the ca-T␤RI-induced CAGA-luciferase reporter expression and the Gal4-Smad3mediated Gal4-luciferase reporter activation (Fig. 3, C-E).
These results together demonstrate that the MH2 domains of Smad3 and Smad7 not only physically but also functionally mediate their interaction.
To consolidate the above conclusion, we then examined the interaction of Smad3 with various Smad7 mutants, including a nuclear localization signal (NLS)-tagged Smad7 that guides Smad7 to stay in the nucleus (21), a PY motif deletion mutant (⌬PY) that loses the ability to associate with WW-HECT type E3 ubiquitin ligases like NEDD4L, Smurf1/2, or WWP1 (39), and the 4E mutant that bears the mutations (K312E, K316E, K410E, and R409E) in the MH2 domain and affects the binding of Smad7 to TGF-␤ receptors (40). Smad7 could inhibit Smad3 activation by blocking receptor activity, thereby indirectly affecting Smad7-Smad3 interaction. To exclude this possibility, we co-expressed ca-T␤RI and HA-Smad3 in HEK293FT cells, whereas various FLAG-Smad7 constructs were separately expressed in different set of cells. After lysing the cells the lysates were mixed as indicated and subjected to co-IP assay. Interestingly, only Smad7 4E mutant lost the ability to associate with Smad3, whereas the other two mutants still retained the Smad3 binding activity (Fig. 3, E and F). Consistently, the Gal4luciferase reporter assays showed that the 4E mutant was unable to inhibit Smad3 or Smad1 activity, whereas the ⌬PY mutant exhibited an attenuated inhibitory effect, in comparison to those of the wild-type or NLS-tagged Smad7 (Fig. 3, E, G,  and H). These results indicate that the four amino acids Lys-312, Lys-316, Lys-410, and Arg-409 are important for Smad7 to interact with both T␤RI and R-Smads. The data also suggest that the Smad7-Smad3 interaction could occur in the nucleus and that the binding of WW-HECT type E3 ubiquitin ligases could play a role in Smad7-mediated inhibition of R-Smad activity.
Smad7 Interferes with R-Smad-Smad4 Complex Formation-As homo-oligomerization of R-Smads or Smad4 and heterooligomerization of R-Smads with Smad4 or Smad7 are all mediated by their MH2 domains, we examined whether Smad7 could interfere with the complex formation between R-Smads and Smad4. As shown in Fig. 4A, TGF-␤ stimulated robust Smad2/3-Smad4 interaction in the control MDA-MB-231 cells,

Smad7 Interacts with R-Smads
but the interaction was dramatically reduced in the FLAG-Smad7-expressing stable cells. To dissect whether Smad7 could interfere with the R-Smad-Smad4 interaction directly or indirectly due to the impaired receptor activity, co-IP experiment was carried out in T␤RI-deficient R1B/L17 cells. Ectopic inter-action between Smad3 and Smad4 was detected, and the interaction was attenuated by co-expression of Smad7 (Fig. 4B). To further confirm the direct interference of Smad7 on the Smad2/ 3-Smad4 complex formation, we performed a co-IP experiment similarly as done in Fig. 3F. Smad3 and Smad4 co-expressed WCL, whole cell lysate. B, plasmids were transfected as indicated into HEK293FT cells for co-IP analysis. C and D, CAGA-luciferase or Gal4-TK-luciferse reporter assays were carried out as in Fig. 1, A and B, respectively. E, diagrammatic representation of Smad7 mutants and summary of Smad7 domain-mapping results. F, HA-Smad3 and ca-T␤RI plasmids were transfected into HEK293FT cells as indicated, whereas FLAG-Smad7 proteins (WT, NLS-, ⌬PY, and 4E) and control empty vector were independently expressed. After cell lysis, the cell lysates were mixed together as indicated and subjected to co-IP experiments. G and H, pFR-luciferase reporter assays were performed in HEK293FT cells as described in Fig. 1D. Cells were treated with 100 pM TGF-␤1 (G) or 25 ng/ml BMP2 (H) for 20 h before harvested. *, p Ͻ 0.05.

Smad7 Interacts with R-Smads
with or without ca-T␤RI in HEK293FT cells were mixed with separately expressed Myc-Smad7, and then protein interactions were examined by co-IP. Again, the Smad3-Smad4 interaction was attenuated by Smad7 in a dose-dependent manner (Fig. 4C). In accordance, Smad4 was able to overcome the inhibitory effect of Smad7 on Smad3-induced Gal4-luciferase reporter expression (Fig. 4D).
Finally, in line with the above observation that the 4E Smad7 mutant no longer interacts with Smad3, this mutant was unable to interfere with Smad3-Smad4 association in co-IP assay (Fig.  4E). However, the PY motif-deleted (⌬PY) and the NLS-tagged Smad7 that retained the ability to associate with Smad3 decreased the Smad3-Smad4 interaction (Fig. 4E). These data together strongly support the notion that Smad7 can directly interfere with R-Smad-Smad4 hetero-oligomerization.
Although deletion of the PY motif (⌬PY) has little effect on the ability of Smad7 to associate with Smad3 (Fig. 3, E and F) or to interfere with the R-Smad-Smad4 complex formation (Fig.  4E), inhibition of the ⌬PY mutant on Smad3 or Smad1 signaling activity was less effective than wild-type Smad7 (Figs. 3, E, G,  and H, and 5, A and B). Together these results suggest that the recruitment of WW-HECT type E3 ubiquitin ligases may contribute to Smad7-mediated inhibition of R-Smad activity.
As Smad7 preferentially interacts with activated R-Smads, we tested whether Smad7 could modulate NEDD4L-elicited degradation of phospho-Smad3. As shown in Fig. 5C, NEDD4L was able to decrease TGF-␤1-induced C-terminally phosphorylated Smad3 (p-Smad3) level, and this effect was augmented by co-expression of Smad7. Moreover, the p-Smad3 level was restored in the presence of the proteasome inhibitor MG132, suggesting the contribution of Smad7 to NEDD4L-induced proteasomal degradation of p-Smad3. Next, we examined whether Smad7 could promote NEDD4L-induced turnover of Smad3(2D), the active Smad3 mimic. Indeed, expression of NEDD4L alone accelerated the turnover of Smad3(2D), and Smad7 remarkably enhanced the effect (Fig. 5, D and E). Functionally, wild-type NEDD4L, but not its ligase-deficient C821A derivative (35), inhibited Smad3(2D) activity in a dose-reliant manner, and the inhibition was greatly facilitated by Smad7 (Fig. 5F). Accordantly, silencing of NEDD4L or Smad7 gene expression enhanced Smad3(2D) activity, and silencing of either of the two genes was capable of attenuating the inhibitory effect of the other (Fig. 5G), again demonstrating that Smad7 and NEDD4L act in concert to inhibit R-Smad activity.

Smad7 Interacts with R-Smads
409) have been reported to mediate binding of Smad7 to T␤RI (40), and simultaneous mutation of the four residues to glutamic acids (4E mutant) disabled the Smad7-Smad3 association (Fig. 3, D-G). In fact, the double mutation of Lys-312/Lys-316 -2E, Lys-401/Arg-409 -2E, and Lys-401/Arg-409 -2A or either of the K359E, K401E, and R409E single mutations was able to disable Smad7 in inhibiting the activity of T␤RI or Smad3, whereas the K312E and K316E single mutation exhibited attenuated inhibitory effects (Fig. 6, D and E). In accordance with above data, the 4E, Lys-312/Lys-316 -2E and Lys-401/Arg-409 -2E as well as the K359E, K401E, R409E mutants were unable to associate with Smad3, and the binding ability of K312E and K316E single mutants were also partially impaired (Fig. 6, F and G). Together these results demonstrated that the basic region including the five key residues in the MH2 domain is essential for Smad7 to associate with Smad3 and inhibit its activity.

Discussion
TGF-␤ signaling is finely controlled both in strength and in duration in normal contexts and is deregulated in a variety of human diseases (1,3,5,7,8,13). Smad7, an inhibitory Smad, has been found to play a vital role in restricting TGF-␤ signaling (15,24,25,29). It has been reported to exert inhibitory effects at the receptor level or the transcription level (1,8,15). Here we report a novel mechanism whereby Smad7 inhibits TGF-␤/ Smad signaling by forming hetero-complexes with and inhibiting the activity of R-Smads directly (Fig. 6H).
Smad7 is able to form a stable complex with the activated TGF-␤ receptors, thereby inhibiting their activity or protein FIGURE 6. Identification of amino acids mediating the Smad7-Smad3 interaction. A, a ribbon drawing (left) and surface electrostatic potential (right) of C-terminally phosphorylated Smad3 MH2 domain (PDB code 1U7F). The positively charged regions and amino acids are shown in blue, and negatively charged regions are in red. B, co-IP experiment was carried out in HEK293FT cells as in Fig. 2B. IB, immunoblot; WCL, whole cell lysate. C, Smad7 MH2 domain modeling based on that of Smad3. D and E, effects of wild-type and mutant Smad7 on the activity of ca-T␤RI or Smad3. CAGA-luciferase (D) and pFR-luciferase (E) reporter assays were performed in HEK293FT cells. F, Myc-Smad3 and ca-T␤RI-HA were expressed as shown in HEK293FT cells, whereas wild-type or mutant Smad7 were separately expressed. The lysates were mixed as indicated before co-IP analysis. G, summary of the functions of Smad7 mutants. H, a schematic diagram showing regulation of R-Smad activity and TGF-␤/Smad signaling by Smad7.

Smad7 Interacts with R-Smads
stability (1,8,17,18). In the nucleus, Smad7 binds to DNA directly via its MH2 domain and impedes bindings of the functional R-Smad-Smad4 complex to DNA (21). Here we found that Smad7 was capable of inhibiting the expression of Gal4luciferase reporter induced by Gal4-Smad3 and Gal4-Smad2 independent of TGF-␤ receptor activity, as assessed by the addition of the T␤RI inhibitor SB431542 or in T␤RI-deficient R1B/L17 cells. Because R-Smads were brought to the Gal4 reporter by the Gal4 DNA binding domain and activated the reporter, our results demonstrate that Smad7 executes its inhibitory effect at another layer by directly inhibiting R-Smad activity.
The MH2 domains of Smad proteins mediate homo-oligomerization and hetero-oligomerization of R-Smads or Co-Smad and interaction of R-Smads or Smad7 to the receptors (9,11). Similarly, Smad7 also associates with R-Smads through the MH2 domain in response to TGF-␤ or BMP signaling, and mutation of four basic amino acids (Lys-312, Lys-316, Lys-401, and Arg-409) in the Smad7 MH2 domain impairs the Smad3-Smad7 interaction. These four amino acids have been reported to mediate the binding of Smad7 to activated T␤RI (40), suggesting that the R-Smad-Smad7 interaction shares a structural similarity as the T␤RI-Smad interaction or the R-Smad-Smad4 interaction. Indeed, structure modeling of the Smad7 MH2 domain revealed the existence of a highly positively charged region including the L3 loop and the ␤8 strand, similar to that of the Smad3 MH2 (9,11,44). Intriguingly, different from those of R-Smads or Smad4, an ␣ helix containing Lys-312 and Lys-316 also contributes to the basic region in Smad7 MH2.
In addition to interacting with R-Smads, Smad7 was also shown to associate with Smad4 upon TGF-␤ or BMP signaling. This is consistent with a previous study, which showed that the Smad4-Smad7 interaction depends on ligand-activated R-Smads (37). It is possible that Smad7 interacts with either Smad3 or Smad4 and thus prevents the R-Smad-Smad4 complex formation. However, we cannot rule out other possibilities, as Smad7 can form a homodimer (data not shown). Moreover, Smad6, the other inhibitory Smad that is more specifically involved in BMP pathways, was also able to oligomerize with Smad7 in the presence or absence of BMP signaling, with unknown biological significance. Together these results implicate that oligomerization could be an intrinsic and general nature of Smad family proteins.
In line with the importance of MH2 domains in mediating the interactions of both Smad7 and Smad4 with active R-Smads, Smad7 competed with Smad4 for R-Smad binding in a dose-dependent manner. Interestingly, as observed in the Smad7-expressing stable cells, TGF-␤ induced a rapid association of Smad2/3 with Smad4 or with Smad7, and then both interactionsdecreasedgraduallyalongwiththedecreasingphosphorylated R-Smad level. As Smad7 is a target gene of TGF-␤/ BMP signaling, it is possible that endogenous R-Smad-Smad7 interaction would increase along with signaling activation and duration to serve as a negative feedback loop. Taken into account that Smad6 also completes with Smad4 to interact with BMPs-activated Smad1 (45), it could be a general mechanism for the two inhibitory Smads to oligomerize with R-Smads, leading to inhibition of their signaling activity.
Although deletion of the PY motif in Smad7 had little effect on the interaction of Smad7 with Smad3, it alleviated Smad7mediated inhibition of Smad3 activity substantially. The PY motif is important for recruitment of WW-HECT type E3 ubiquitin ligases such as Smurf1/2, NEDD4L (NEDD 4 -2) and WWP1 (1,8,15,42). NEDD4L has been shown to associate with and degrade both Smad2 and Smad3 upon TGF-␤ signaling (41,43). In this study we found that Smad7 enhanced the interaction of Smad2/3 with NEDD4L and facilitated NEDD4L-induced polyubiquitination and proteasomal degradation of phospho-Smad3 and the turnover of Smad3(2D), a phosphorylation mimicking-mutant of Smad3. Therefore, Smad7 could act as an adaptor to recruit NEDD4L not only to the receptors but also to active R-Smads, and NEDD4L degrades R-Smads either directly or indirectly via Smad7. Moreover, Smurf2, which has been demonstrated to associate with Smad2/3 upon TGF-␤ signaling and target them for ubiquitination (46,47), could also cooperate with Smad7 to inhibit Smad2/3 activity in Gal4-luciferase reporter assays (data not shown), suggesting that NEDD4L, Smurf2, and other potential WW-HECT type E3 ubiquitin ligases could be engaged in Smad7-mediated inhibition of R-Smad activity via a similar mechanism.
Author Contributions-Y.-G. C. and X. Y. designed the study and wrote the manuscript. X. Y. carried out most experiments and analyzed the data. H. L. modeled the structure of Smad7 MH2 domain. M. C. generated some of the Smad3 mutant constructs. X. S. prepared plasmids for the bimolecular fluorescence complementation experiment. X.-H. F. and X. L. prepared plasmids for the Gal4-luciferase reporter system, established the MDA-MB-231 stable cells, and revised the manuscript.