Smad7 Inhibits Transforming Growth Factor-β Family Type I Receptors through Two Distinct Modes of Interaction*

The inhibitory Smads (I-Smads), i.e. Smad6 and Smad7, are negative regulators of transforming growth factor-β (TGF-β) family signaling. I-Smads inhibit TGF-β family signaling principally through physical interaction with type I receptors (activin receptor-like kinases), so as to compete with receptor-regulated Smads (R-Smads) for activation. However, how I-Smads interact with type I receptors is not well understood. In the present study, we found that Smad7 has two modes of interaction with type I receptors. One is through a three-finger-like structure in the MH2 domain, consisting of residues 331–361, 379–387, and the L3 loop. The other is through a basic groove in the MH2 domain (Mochizuki, T., Miyazaki, H., Hara, T., Furuya, T., Imamura, T., Watabe, T., and Miyazono, K. (2004) J. Biol. Chem. 279, 31568–31574). We also found that Smad6 principally utilizes a basic groove in the MH2 domain for interaction with type I receptors. Smad7 thus has an additional mode of interaction with TGF-β family type I receptors not possessed by Smad6, which may play roles in mediating the inhibitory effects unique to Smad7.

mutations of basic residues in the L3 loop (Lys-401 and Arg-409) of the Smad7 MH2 domain attenuate its association with both ALK-5 and ALK-6, whereas mutations in the ␣-helix1 (Lys-312 and Lys316) attenuate its association with ALK-5, but not with ALK-6 (26). These findings indicate that the mode of interaction with I-Smads and sensitivity to I-Smads vary among the subgroups of type I receptors. However, the molecular basis of the interaction between I-Smads and ALKs has not been clearly elucidated.
In the present study, utilizing the differential effects of Smad6 and Smad7 on ALK-2, we identified the important regions of Smad7 for interaction with type I receptors. The mode of interaction between type I receptors and I-Smads differed from that between R-Smads and type I receptors. We also found that Smad7 has two surfaces for interaction with type I receptors, whereas Smad6 has one. Usage of these two surfaces of Smad7 may affect its function as a negative regulator.
Homology Modeling-A structural model of Smad7 was constructed based on the three-dimensional structure of Smad2 (Protein Data Bank code 1DEV) using Biopackage (Molsoft).

MH2 Domains Are Responsible for the Differential Inhibitory
Effects of I-Smads on Signaling from ALK-2-Smad7 efficiently inhibits BMP signaling from both ALK-2 and ALK-3. In contrast, Smad6 efficiently inhibits BMP signaling from ALK-3 but only weakly inhibits that from ALK-2 (24). By utilizing these differential effects of Smad6 and Smad7 on ALK-2 signaling, we investigated the molecular basis of the Smad7-ALK-2 interaction.
We first examined which parts of I-Smads are responsible for the differential inhibition of signaling from ALK-2, with the use of truncated mutants of Smad6 and Smad7. Smad6 and Smad7 were divided into N-terminal and C-terminal fragments (1-330 and 315-496 for Smad6, 1-259 and 247-427 for Smad7, denoted Smad6N, Smad6C, Smad7N, and Smad7C), and their inhibitory effects on signaling from ALK-2 and ALK-3 were determined using luciferase reporter assay (Fig. 1). The isolated N domains of I-Smads (Smad6N and Smad7N) failed to repress signaling from both ALK-2 and ALK-3. Smad7C inhibited ALK-2 signaling as effectively as wild-type Smad7, whereas Smad6C had less inhibitory effect on ALK-2 signaling, similar to the full-length Smad6. Consistent with these results, Smad7C interacted with ALK-2 more stably than Smad6C, whereas the N domains of both I-Smads failed to interact with ALK-2 and ALK-3 (supplemental Fig. S1A).
The MH2 domains of I-Smads were thus found to be responsible for the differential inhibitory effects on signaling from ALK-2. In contrast, it is known that the N domain of Smad7 is required for inhibition by Smad7 of signaling from ALK-5 (23).
L3 Loops of I-Smads Do Not Determine the Specificity of Interaction with Type I Receptors-The L3 loops in the vicinity of the C terminus of R-Smads are crucial for the specific interaction of R-Smads with type I receptors (9). Amino acid sequences of the L3 loop are invariant among TGF-␤/activin-specific R-Smads FIGURE 1. MH2 domains are responsible for the differential inhibitory effects of I-Smads on signaling from ALK-2. Smad6 and Smad7 were divided into N-terminal (residues 1-330 and 1-259, respectively) and C-terminal (residues 315-496 and 247-427, respectively) moieties (schematically shown left), and inhibitory effects on ALK-2 and ALK-3 signaling were determined by luciferase reporter assay using BRE-Luc in C2C12 cells (right). Expression of each construct was confirmed by immunoblotting. Results are given as percentages of control. Error bars represent S.D.
(Smad2 and 3) and among BMP-specific R-Smads (Smad1, 5, and 8), respectively ( Fig. 2A). Swapping of the L3 loops of Smad1 and 2 results in switching of specificity of their C-terminal phosphorylation by type I receptors (9). We therefore examined whether the L3 loops of I-Smads are responsible for the specificity of interaction with type I receptors. We constructed mutants of Smad6 and 7 in which the regions spanning the L3 loop to the C terminus had been swapped (Fig. 2, A and B, left) and determined their inhibitory effects on ALK-2 and -3 signaling. We found that swapping of the region did not affect inhibition of ALK-2 and ALK-3 signaling (Fig. 2B, right). Binding experiments between I-Smad mutants with ALK-2 and ALK-3 also gave consistent results (supplemental Fig. S1B). We therefore concluded that the L3 loops of I-Smads do not determine the specificity of interaction with type I receptors.
L3 Loops of I-Smads Are Involved in Interaction with Type I Receptors-Next, to investigate whether the L3 loops of I-Smads are required for their inhibition of BMP signaling, we constructed chimeras of Smad6 or Smad7 with Smad4 because Smad4 is unable to interact with type I receptors. S6-S4L3 and S7-S4L3 are mutants based on Smad6 and 7, respectively, in which the regions spanning the L3 loop to the C terminus are replaced by the corresponding region of Smad4 (residues 505-552 for Smad4). S6-S4L3s and S7-S4L3s are short versions of S6-S4L3 and S7-S4L3, in which helix-5 (H5) and the C terminus of Smad4 (residues 505-531) are truncated (Fig. 2C, left). S6-S4L3, S7-S4L3, S6-S4L3s, and S7-S4L3s all lacked interaction with ALK-2/3 (supplemental Fig. S1B) and failed to suppress signaling from these type I receptors (Fig. 2C, right). Replacement of the L3 loop of S7-S4L3s by that of Smad7 resulted in recovery of inhibition ( Fig. 2D) as well as interaction with ALK-2 and ALK-3 (supplemental Fig. S1B). These findings indicate that the L3 loops of I-Smads are required for their inhibitory effects on ALK-2 and ALK-3 through physical interaction. The L3 loops of I-Smads are thus required for but not determinative of the differential effects of Smad6 and Smad7 on ALK-2 signaling, suggesting that other region(s) in I-Smads play crucial roles in determining the specificity of interaction with BMP type I receptors.
Identification of the Regions in Smad7 Responsible for Specific Inhibition of BMP Signaling from ALK-2-To identify which region(s) of Smad7 are determinants of the inhibitory effect on spanning the L3 loop to the C terminus of Smad6 (residues 467-496) and Smad7 (residues 399 -427) were swapped. C, regions spanning the L3 loop to the C terminus of Smad6 (residues 467-496) and Smad7 (residues 399 -427) were replaced by the corresponding region of Smad4 (residues 505-552 for S6-S4L3 and S7-S4L3) or the region in which the helix-5 (H5) and the C terminus of Smad4 were truncated (residues 505-531 for S6-S4L3s and S7-S4L3s). D, C-terminal region of Smad7 was progressively swapped with the corresponding region of Smad4.
We next attempted to specify the regions crucial for inhibition of ALK-2 signaling. First, we constructed four more Smad6/Smad7 chimeras that were intermediates between S6(384/314)S7 and S6(441/371)S7 (supplemental Fig. S2, left). The inhibitory effects on ALK-2 signaling became less potent with gradual substitution of residues 331-361 of Smad7 with corresponding residues of Smad6 (supplemental Fig. S2, right). Thus, the structure of the entire region comprising residues 331-361 of Smad7 appears to be important for the inhibition of ALK-2.
We also gradually replaced Smad7 sequences of S6(384/ 314)S7 with corresponding regions of Smad6 from the C terminus (Fig. 3C, left). Replacement of residues 379 -387 of Smad7 with the corresponding region of Smad6 altered the inhibitory effect on BMP signaling from Smad7-type to Smad6-type (Fig. 3C, right). These findings indicate that residues 379 -387 of Smad7 are involved in the inhibition of ALK-2. We further examined physical interaction of all of the chimeric mutants used in Fig. 3, A-C, with ALK-2 and ALK-3, which was well consistent with their inhibitory effects (supplemental Fig. S1, C and D).
Finally, we constructed a Smad6 mutant that contained residues 331-361 and 379 -387 of Smad7 (denoted S6-mt) and confirmed that these regions were sufficient for inhibition of signaling from ALK-2, as determined by luciferase reporter assay, physical interaction with ALK-2, and R-Smad phosphorylation (Fig. 4, A-C).
Smad7 Interacts with TGF-␤ Family Type I Receptors through Two Distinct Regions-We next constructed a Smad7 mutant, in which residues involved in the interaction with ALK-2 were replaced by corresponding Smad6 residues (denoted S7-mt) (Fig. 5A, left). Contrary to our expectation, S7-mt inhibited ALK-2 signaling as well as wild-type Smad7 and interacted with ALK-2 (Fig. 5, A and B), indicating that there are other site(s) involved in recognition of BMP type I receptors by Smad7. It has been reported that a basic groove composed of four basic residues (Lys-312, Lys-316, Lys-401, and Arg-409) in the MH2 domain of Smad7 plays critical roles in the interaction of Smad7 with ALK-5 (26). We therefore examined whether the basic groove is also involved in the interaction with ALK-2 and ALK-3. Although mutating one of the four basic residues, Lys-316, of wild-type Smad7 to a glutamic acid did not affect its inhibitory activity, the same mutation in S7-mt (denoted S7-K316E/mt) resulted in attenuation of the inhibitory activity and the interaction with ALK-2/3 (Fig. 5, A and B). These findings indicate that Smad7 interacts with BMP type I receptors through the basic groove, independent of the region identified above. In contrast, introduction of the corresponding mutation into Smad6 (K387E) resulted in loss of inhibition by Smad6 of signaling by ALK-3 (Fig. 5A) and interaction with ALK-3 (Fig. 5B).
We next examined inhibitory effects of the same set of Smad7 mutants on signaling from ALK-4 and ALK-5 (Fig. 5C) and their interaction with these receptors (Fig. 5D). We obtained essentially the same results for ALK-4. However, we obtained inconsistent results for ALK-5. S7-K316E failed to interact with ALK-5, indicating that Smad7 interacts with ALK-5 principally thorough the basic groove as previously reported (26). Unexpectedly, S7-K316E still retained an inhibitory effect on signaling from ALK-5, which may be due to regulation of ALK-5 signaling by Smad7 at other levels (30). We also examined effects of Smad6 and its mutant Smad6-K387E on signaling from ALK-4 and ALK-5 and their interaction with these receptors. Smad6 interacted with ALK-4 and ALK-5 through the basic groove ( Fig. 5D) but inhibited signaling only marginally (ALK-4) or moderately (ALK-5) (Fig. 5C). The underlying mechanism for these inert interactions remains to be elucidated.
Mapping of the Important Regions for Interaction with ALKs on a Three-dimensional Structural Model of Smad7-As described above, we have shown that multiple regions including the L3 loop and residues 331-361 and 379 -387 of Smad7 are responsible for interaction with ALKs. We aligned the amino acid sequences of Smad6 and Smad7 containing these regions (Fig. 6A). The amino acid sequences in residues 331-361 of Smad7 are conserved in Smad6 to the same extent as in other regions (53.3%). Residues 379 -387 of Smad7 are not conserved in Smad6. We next mapped the regions required for efficient interaction with ALKs on a structural model of the Smad7 MH2 domain, which was constructed by homology modeling based on the MH2 domain of Smad2 (Protein Data Bank Code 1DEV) (Fig. 6B). The L3 loop, residues 379 -387, and certain amino acids among residues 331-361 are present on a common surface of the three-dimensional structure of Smad7, comprising a three-finger-like structure, whereas the basic groove is located on a different side of the molecule. These findings suggest that, when Smad7 interacts with BMP type I receptors, two distinct modes of interaction exist, one through the three-finger-like structure and the other through the basic groove.
Two Molecular Surfaces Have Similar Binding Affinities for Type I Receptors-We next compared the relative affinity of the two binding surfaces on Smad7 for ALK-2. As shown in Fig. 6C, increasing amount of S7-K316E and S7-mt competed with each other in binding to ALK-2, indicating that the two molecular surfaces have similar binding affinities for ALK-2.

DISCUSSION
Ligands of the TGF-␤ family signal through type I receptor-mediated activation of R-Smads. Interaction between type I receptors and R-Smads is coordinated by robust molecular recognition. I-Smads are members of the Smad protein family which act as antagonists of Smad signaling principally through physical interaction with type I receptors to interrupt activation of R-Smads (13)(14)(15)(16). However, the mode of interaction between I-Smads and type I receptors has remained unclear. In the present study, we found that Smad7 inhibits TGF-␤ family type I receptors through two distinct modes of interaction.
Recognition of R-Smads by activated type I receptors appears to involve two matching components (Fig. 7A) (11). The interaction between the L3 loop of R-Smads and the L45 loop of the type I receptor determines substrate specificity (7)(8)(9)(10). The interaction between the basic groove and the phosphorylated GS loop appears to ensure recognition of R-Smads by activated receptors (11). The basic groove comprises two residues in the L3 loop and two residues in ␣-helix1 (26).
In the interaction of I-Smads with type I receptors, the L3 loops were   OCTOBER 1, 2010 • VOLUME 285 • NUMBER 40 found to be required for but not determinative of the specificity of interaction with these receptors. Smad6 and Smad7 also have a basic groove consisting of the L3 loop and the ␣-helix1. Mutation of a basic residue within the ␣-helix1 of Smad6, but not Smad7, decreased the magnitude of inhibition of ALK-3 signaling and association with ALK-3. These findings indicate that Smad6 principally utilizes the basic groove in its functional interaction with ALK-3. In ALK-3, the region along the nucleotide-binding loop and the ␣C helix, which is located near the L45 loop but apart from the GS loop, is a determinant of interaction with Smad6 (24) (Fig. 7B). The mode of interaction of Smad6 with ALK-3 thus differs from that of R-Smads with type I receptors.

Smad7 Interaction with Type I Receptors
Smad7 utilizes two alternative surfaces for interaction with type I receptors (Fig. 7, C and D). One is the basic groove (Fig.  7C) and the other the three-finger-like structure (Fig. 7D), both of which include the L3 loop as an essential constituent. The three-finger-like structure includes residues 331-361 and 379 -387 in addition to the L3 loop. Residues 379 -387 of Smad7 form a loop-like structure on the side opposite the L3 loop on an identical surface (Fig. 6B). On sequence alignment of Smad6 and Smad7, no amino acid residue in the 379 -387 region of Smad7 coincided with that of Smad6. Moreover, the amino acid sequences of this region of Smad6 are divergent, and only 38% amino acids are conserved even between human and mouse (supplemental Fig. S3); this explains why this region is not utilized in Smad6 for interaction with type I receptors. In the case of residues 331-361, inhibition of ALK-2 signaling became less potent as the length of the amino acid sequence of Smad7 substituted with that of Smad6 increased. Within residues 331-361, there is a protruding loop-like structure on the same surface as residues 379 -387 and the L3 loop. Smad7 thus possibly grabs type I receptors via this three finger-like loop structure.
For inhibition of ALK-5 signaling, Smad7 may have an alternative mode of regulation of ALK-5 signaling, which does not require its interaction with ALK-5 (12,30). S7-K316E failed to interact with ALK-5, but still inhibited signaling from ALK-5. Remarkably, the three-finger-like structure of Smad7 is also indispensable for this mode of ALK-5 inhibition because S7-K316E/mt lost its inhibitory effect on ALK-5.
At present, it is unclear why Smad7 utilizes two alternative surfaces in its interaction with type I receptors. The interaction of Smad7 with type I receptors through the three-finger-like structure appears to be involved in Smad7-specific modulation of type I receptor function, which remains to be elucidated. It is also unclear how utilization of two surfaces of Smad7 in interaction with type I receptors is controlled. One possibility is that some posttranslational modifications of Smad7 or type I receptors affect the mode of interaction between them.
In mammals, novel types of I-Smads, i.e. Smad6 and Smad7, have emerged via gene duplication events. Although Dad does not recruit dSmurf to type I receptors (32), mammalian I-Smads have acquired the ability to cooperate with Smurf family ubiquitin ligases (20 -24). In addition, the two I-Smads have distinct target specificity on type I receptors. Smad6 exhibits differential inhibitory effects on two subfamilies of BMP type I receptors. On the other hand, Smad7 has acquired the ability to inhibit ALK-4/5/7 in addition to ALK1/2 and ALK3/6, together with utilization of a distinct binding surface in its interaction with type I receptors. Control of TGF-␤ family type I receptor function by I-Smads via various mechanisms through multiple

Smad7 Interaction with Type I Receptors
surfaces results in the generation of spatiotemporally fine tuned signaling activity.