We have reported on some protein binding defects and gene expression changes caused by the mutation of specific amino acids on the surface of Smad3. As these amino acids are conserved in Smad2, the corresponding mutations in Smad2 might also affect Smad2 function, but Smad2 was not examined in these studies. The protein ligands used to query the binding properties of the Smad3 variants were a small subset of the identified Smad3-interacting proteins. They were used to reveal whether a mutation in a putative hot-spot on Smad3 reduced the binding to one or more proteins and, importantly, retained the binding to other Smad-binding partners, i.e., that each Smad3 mutation only affected binding to a portion of this test set of protein and peptide ligands. Although the test set was meant to represent the larger (>50) group of known Smad3-interacting proteins and was large enough to characterize selective protein-binding deficits in the variants, the test set of proteins and peptides cannot reveal the full extent of the protein-binding defects or which other Smad3 protein-protein interactions may be reduced by a specific mutation. As such, the effects of the mutations on the transcriptional functions of Smad3 are very likely mediated by other Smad3-interacting proteins that were not in the test panel. Therefore, one should not conclude that any of the specific protein ligands used in the reported binding studies are directly implicated in mediating the reported gene expression changes. One exception to this caveat may be the mutations that reduce binding to Smad4. Given that the canonical function of phosphorylated Smad3 is to form a heterotrimer with Smad4 that accumulates in the nucleus, reduced Smad4 binding is likely to be in part responsible for the reduced activity of these mutations in reconstituting Smad3-dependent reporter gene activity in JEG-3 cells.
The protein binding properties of the Smad3 variants were quantified in a modified version of the LUMIER assay 
. Proteins were co-expressed in HEK-293 cells with an activated Alk5 receptor to constitutively phosphorylate Smad3. All of the previously described Smad3 mutations exhibited the same defects in protein binding in this assay as had been initially reported using other assay methods. By using a larger panel of Smad3-binding proteins and peptides, we identified additional binding interactions that were impaired by each mutation and also demonstrated that other binding interactions were not affected under the conditions of the assay. In the cases where both a Smad binding domain and the corresponding full length protein were tested, the full length protein always provided more binding to the Smad3 variant. The increased binding by the full length protein may indicate that the Smad binding domain is in a more optimal conformation in the native protein than on the Trx scaffold or that the full length protein has other motifs that contribute to the binding in addition to the motif displayed on the Trx scaffold. In either case, full-length Smad-binding proteins will be a better indicator in future studies of the consequences of a hot-spot mutation in cells.
The binding assays revealed several examples where the amount of binding between a Smad3 variant and a Smad-binding protein was greater than with the wild-type Smad3 protein. For example, more Renilla-Smad3 E267A was pulled-down with Smad4, Smad3 and the Smad-binding domains of Smurf2 and Ski. Greater amounts of Renilla luciferase Smad3 V224A were pulled-down with SkiW274E and the Smad-binding domains from Smurf2, Ski and SIP1, but not by Smad3 or Smad4. The Smurf2 Smad binding domain pulled down more of nine different Renilla-Smad3 variants in . Previous studies have reported a propensity for mutant Smad proteins to undergo degradation, which would be consistent with increased interactions with an E3 ubiquitin ligase such as Smurf2 
. The molecular mechanism(s) for the increased binding in the present studies are not known. The mutations may enhance the binding affinity between Smad3 and a Smad-binding protein either by improving the binding interface directly or by altering the conformation of Smad3 to improve the binding interface. An alternative explanation is that these mutations reduce the amount of some Smad3-binding proteins interacting with Smad3, thereby making more Smad3 available to interact with other Smad3-binding proteins. Independent of the mechanism, the results indicate that the functional consequences of perturbing a protein-binding hot-spot on a hub protein by mutation or with a small molecule ligand or drug may be the result of both decreased interactions with some protein ligands and also increased interactions with other protein ligands. For example, the Smad3 E267A protein exhibited increased binding to several protein ligands and mediated increased basal levels of gene expression and wild-type (or greater than wild-type) TGF-β-induced levels of gene expression on several of the genes examined in C2C12 cells. This high basal activity resulted in a lower fold-induction by TGF-β, as indicated by the lower TGF-β inducibility in and . The increase in basal gene expression by Smad3 E267A was not due to it being expressed at higher levels of protein expression than the exogenous wild-type Smad3 or the other variants (). In future studies on the functional deficits of Smad3 variants it will be important to examine both the loss of binding partners and the increase of other Smad3-binding proteins in the regulatory complex relative to the components of the wild-type Smad3 regulatory complex.
Although only a small panel of Smad3-reponsive genes was examined, there was some evidence for different hot-spot mutations causing distinct effects on gene expression. For example, the Y297A mutation, decreased binding to SARA and Ski () and MEF2C and the Smad binding domain from FoxO3 (Table S1
) but provided 100% reconstitution of the Smad3-dependent reporter gene activation in JEG-3 cells (). However, Smad3 Y297A also had the unique property among the mutations examined of enhancing the TGF-β induciblity of five of the endogenous genes tested in C2C12 cells. Overexpression of wild-type Smad3 is well-known to induce gene expression changes in the absence of TGF-β, but the Smad3 Y297A variant induced lower levels of expression in the absence of TGF-β than did wild-type Smad3, resulting in a greater fold induction when TGF-β was added.
Another example for a selective effect was observed with Smad3 K341A. The K341A mutation decreased binding to SARA and Ski () and the Smad binding peptide from CBP (Table S1
) and provided less than 50% reconstitution of the Smad3-dependent reporter gene activation in JEG-3 cells (), but it mediated TGF-β inducibility similar to wild-type Smad3 for five of six endogenous genes tested. Interestingly, it was compromised by almost 50% for mediating TGF-β activation of Olfm2. Examination of a larger panel of Smad3 responsive genes and expression of the Smad3 K341A protein in an otherwise Smad3-deficient cell line may identify genes with an even greater dependence than Olfm2 on K341 in Smad3.
Several genes were identified from a gene array analysis of mRNAs from the TGF-β-induced C2C12 cells over-expressing wild-type Smad3 versus TGF-β-induced C2C12 cells infected with the control retrovirus that did not encode Smad3 (empty vector, EV). RT-PCR analysis confirmed the increased level of expression for thirteen genes and a reduced level of expression for three genes after 24 hours of TGF-β treatment; in most cases the responses were amplified by the overexpression of wild-type Smad3. Although we did not select for any particular functionality, most of the genes with increased expression levels encode proteins with roles in cell signaling, cell adhesion or extracellular matrix modifications. Several of the genes including Pai1, Mmp9 and Il11 have been identified as TGF-β-responsive genes in many other studies and cell-types. For example, Pai1 is induced by TGF-β in muscle satellite cells 
and MMP9 is an important Smad3 target in the process of cancer cell invasion 
. Regulation of Il11 by TGF-β signaling is important for metastasis to bone of breast and melanoma cancer cells 
. Pai1 and Cst6 were identified as direct targets for Smad3 regulation in a genome wide chromatin immunoprecipitation study 
. Cst6 encodes cystatin M/E, an inhibitor of the cathepsin L cysteine protease, which has been implicated in epidermal differentiation and tumor suppression 
. Methylation or loss of cystatin M/E expression has been reported in breast 
and prostate cancers 
. To the best of our knowledge Cst6 expression has not previously been reported in myoblasts.
The two other genes whose expression levels were reduced in the C2C12 cells were Myl1 and Sarcoglycan, Sgcg. Myl1 has been associated with muscle cell differentiation and knock-down of its expression enhanced myoblast proliferation 
. Sgcg mutations cause limb-girdle muscular dystrophy and Smad3 has been implicated in preventing inappropriate activation of Sgcg gene expression during myogenic differentiation 
. The reduced levels of Myl1 and Sgcg are consistent with the inhibition of muscle differentiation by TGF-β and Smad3. One of the earliest cellular responses reported for TGF-β was inhibition of myoblast differentiation in culture 
. Smad3 binds directly to the MyoD bHLH domain to block MyoD/E protein dimerization and DNA binding 
and to the myogenic transcription factor MEF2 to prevent muscle-specific gene expression 
. Persistent TGF-β expression contributes to the muscle fibrosis and loss of muscle function in neuromuscular diseases such as Duchenne muscular dystrophy 
Fermt1 encodes Kindlin-1, a focal adhesion protein involved in cytoskeletal interactions and implicated in integrin beta-1 activation 
. Loss-of-function of Fermt1 is associated with Kindler syndrome, an autosomal recessive disorder characterized by skin atrophy and blistering 
. Kindlin-1 is a member of a family of proteins that are key regulators of cell-matrix interactions 
. Tnfaip6 or TGS-6 (tumor necrosis factor, alpha-induced protein 6) encodes a hyaladherin, a hyaluronan-binding protein, and is involved in extracellular matrix stability and cell migration 
. TGS-6 expression is induced by TGF-β in dermal fibroblasts and is a required mediator of the TGF-β induced fibroblast-myofibroblast transition in wound healing 
. Olfm2 or olfactomedin 2 is a secreted glycoprotein belonging to a family of related olfactomedins 
. We are not aware of previous reports of a role for olfactomedins in muscle or myoblasts. A related family member, Olfm1, has been reported as a modulator of Wnt signaling in zebrafish 
. The Wnt family member that was induced in the studies reported here, Wnt11, has been implicated in fostering cell adhesion of cardiomyocytes essential for proper heart development 
Understanding the multiple functions of hub proteins requires new approaches. Knockdown approaches or knock-out mutations do not allow analysis of the distinct roles of a hub protein. Mutations in protein-binding hot-spots have the potential to only perturb a subset of the total binding interactions, providing an opportunity to learn the biological significance of those interactions. This approach is analogous to the generation of an allelic series of mutations in classical genetics for the purpose of revealing the different developmental roles of a specific gene. Future studies will benefit from a more comprehensive analysis of the gene expression changes caused by each hot-spot mutation, from analysis of the biological consequences on cellular phenotypes such as proliferation, differentiation, or pathological functions and from characterization of the changes in the composition of the protein complexes formed on the hub protein in the same cell-type and cell context being studied for phenotypic changes.