Dephosphorylation of the Linker Regions of Smad1 and Smad2/3 by Small C-terminal Domain Phosphatases Has Distinct Outcomes for Bone Morphogenetic Protein and Transforming Growth Factor-β Pathways*

Smad proteins transduce bone morphogenetic protein (BMP) and transforming growth factor-β (TGFβ) signals upon phosphorylation of their C-terminal SXS motif by receptor kinases. The activity of Smad1 in the BMP pathway and Smad2/3 in the TGFβ pathway is restricted by pathway cross-talk and feedback through protein kinases, including MAPK, CDK2/4, p38MAPK, JNK, and others. These kinases phosphorylate Smads 1-3 at the region that links the N-terminal DNA-binding domain and the C-terminal transcriptional domain. Phosphatases that dephosphorylate the linker region are therefore likely to play an integral part in the regulation of Smad activity. We reported previously that small C-terminal domain phosphatases 1, 2, and 3 (SCP1-3) dephosphorylate Smad1 C-terminal tail, thereby attenuating BMP signaling. Here we provide evidence that SCP1-3 also dephosphorylate the linker regions of Smad1 and Smad2/3 in vitro, in mammalian cells and in Xenopus embryos. Overexpression of SCP 1, 2, or 3 decreased linker phosphorylation of Smads 1, 2 and 3. Moreover, RNA interference-mediated knockdown of SCP1/2 increased the BMP-dependent phosphorylation of the Smad1 linker region as well as the C terminus. In contrast, SCP1/2 knockdown increased the TGFβ-dependent linker phosphorylation of Smad2/3 but not the C-terminal phosphorylation. Consequently, SCP1/2 knockdown inhibited TGFβ transcriptional responses, but it enhanced BMP transcriptional responses. Thus, by dephosphorylating Smad2/3 at the linker (inhibitory) but not the C-terminal (activating) site, the SCPs enhance TGFβ signaling, and by dephosphorylating Smad1 at both sites, the SCPs reset Smad1 to the basal unphosphorylated state.

The transforming growth factor ␤ (TGF␤) 3 family of cytokines transmit signals by activating their cognate membrane receptor serine/threonine kinases, which in turn phosphorylate receptor-regulated Smad proteins (R-Smads) at two C-terminal serine residues in a C-terminal SXS motif (1). Upon phosphorylation, the R-Smads translocate into the nucleus and regulate the transcription of many target genes (1). In general, among the R-Smads, Smads 2 and 3 respond to the TGF␤ branch of the family, which also includes activin and nodal factors, whereas Smads 1, 5, and 8 mediate responses to the BMP branch (2). Collectively these pathways control a plethora of cellular processes, including cell proliferation, recognition, differentiation, apoptosis, and cell fate, during embryogenesis as well as in mature tissues (1). Given the importance of TGF␤ signaling in metazoan biology, the contribution of different regulatory inputs is likely to be equally important.
R-Smads consist of two highly conserved globular domains, namely the MH1 and MH2 domains, that are connected by a more divergent linker region (1). The MH1 domain binds DNA, whereas the MH2 domain binds to receptors for activating phosphorylation, nucleoporins for nuclear translocation, and various partner proteins for transcriptional regulation in the nucleus (2). The diversity of the linker region, which consists of several serine and threonine residues, allows for regulation of R-Smads by multiple signaling inputs. The linker region of Smad1 consists of four MAPK phosphorylation sites (Ser-187, Ser-195, Ser-206, and Ser-214), whereas Smad2/3 consist of four SP/TP sites for proline-directed kinases (Fig. 1A). In response to mitogens, Erk MAPK mediates the phosphorylation of these sites in vivo (3,4). CDK2 and -4 have also been reported to mediate the phosphorylation of some of the linker residues in Smad2/3 in addition to residues at the N terminus of Smad2/3 (5). p38 MAPK and JNK also phosphorylate the linker region of Smad2/3 and regulate their transcriptional activity (6,7). The MAPK-mediated phosphorylation of the linker region * This research was supported by a grant from The Rockefeller University "Women & Science" fellowship program (to M. K.) and by National Institutes of Health Grants HD32105 (to A. H. B.) and CA34610 (to J. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  generally results in inhibition of Smad1 activity (3,8) and attenuation of nuclear accumulation of Smad1 (3). Similarly, MAPKmediated attenuation of Smad2 activity has been attributed to Smad2 linker phosphorylation (4,9). In Xenopus embryogenesis, linker phosphorylation of Smad1 through MAPK plays an important role in inhibiting BMP signaling, which results in neural induction (8). Linker phosphorylation of Smad2/3 during Xenopus embryogenesis results in cytosolic retention of Smad2/3 and inhibition of TGF␤ signaling (9). Recent studies have shed light on the identification and the roles of phosphatases in regulating R-Smad activity by dephosphorylation of the C-terminal SXS motifs of Smad1 and Smad2/3. Because the majority of phospho-Smad1 upon BMP stimulation translocates to the nucleus, nuclear phosphatase(s) may play dominant roles in dephosphorylating phospho-Smad1. Two distinct families of nuclear phosphatases have recently been reported to mediate the dephosphorylation of Smad1 tail to inhibit BMP signaling: the small polymerase II C-terminal phosphatases (SCPs) (10) and PPM1A phosphatases (11). The SCPs do not dephosphorylate the Smad2/3 C-terminal sites (10), but PPM1A acts on both the Smad1 and Smad2/3 C-terminal sites (11,12). With the use of a functional RNAi-based screening in Drosophila S2 cells, pyruvate dehydrogenase phosphatase, a predominantly mitochondrial metabolic enzyme, has also been identified as a phosphatase for MAD, the Drosophila homolog of Smad1 (13).
The SCPs consist of three closely related class C serine/ threonine phosphatases. SCP1-3 are related to the catalytic subunit of FCP1, which is a highly conserved, essential enzyme that dephosphorylates the C-terminal domain (CTD) of RNA polymerase II (pol II) and is required for pol II recycling and global DNA transcription (14). The SCPs can dephosphorylate the CTD of pol II in vitro (14), but a role for them as general regulators of transcription has not been shown. SCPs may play specialized roles in transcriptional control. They mediate silencing of neuron-specific gene expression in a mouse pluripotent cell line induced to undergo neuronal differentiation as well as in Drosophila S2 cells (15). Interestingly, SCP2 is amplified in a subset of osteosarcomas (16) and restricts BMP signaling in a SCPoverexpressing osteosarcoma cell line (10).
In the course of our work we noticed that SCPs interacted with both Smad1 and Smad2 independently of the phosphorylation state of the C-terminal sites and despite the fact that they display phosphatase activity against the C-terminal sites in Smad1 but not Smad2. We therefore investigated whether SCPs targeted additional phosphorylation sites within R-Smads. In the present study we have demonstrated that SCPs target the linker region of Smad1 as well as Smad2, affecting both branches of the TGF␤ pathway. Our observations complement and extend a recent report from others showing that SCPs act as Smad2/3 linker phosphatases leading to enhancement of TGF␤ signaling (17). We show that SCPs act as linker phosphatases for Smad1 as well as Smad2/3, that they act on BMP-or TGF␤-dependent linker phosphorylated Smads, and that their action has distinct consequences on the BMP and TGF␤ path-ways, i.e. the SCPs reset Smad1 to the base-line state, whereas they enhance Smad2/3 signaling.

EXPERIMENTAL PROCEDURES
Plasmid Constructions-Human SCP1, SCP2, and SCP3 (10) constructs were subcloned into EcoRI/HindIII sites in pCMV5 vectors containing N-terminal HA or FLAG epitope tags. Catalytically inactive mutants of SCP1(D97E), SCP2(D107E), and SCP3(D112E) were each generated by mutating a conserved aspartate residue to glutamate at the active site within the phosphatase domain. Human SCP1, SCP2, and SCP3 were also subcloned into the same sites in pGex6P-1 vector (Amersham Biosciences) in order to express GST-tagged SCPs in Escherichia coli. The construction of pCs2-XSCP2 and the mutant XSCP2 has been described previously (10). pCMV5 vectors encoding N-terminal FLAG-or HA epitope-tagged human Smad1 and Smad2 have been describe previously (3,4). All sequences were verified by DNA sequencing.
Cell Culture, Transfections, and Stimulations-HaCaT keratinocytes, HEK293 cells, and OsA-CL osteosarcoma cells were cultured as described (10). The transfection of siRNA oligonucleotides into HaCaT and OsA-CL cells was performed using Lipofectamine 2000 reagent (Invitrogen) as described previously (10). 300 pmol of each siRNA oligo was used per 10 cm dish containing 1 ϫ 10 7 cells. Cells were serum-starved for 12 h prior to treatment with BMP (25 ng/ml; 1 h), TGF␤ (100 pg/ml; 1 h), or EGF (100 ng/ml; 30 min), respectively. Cells were subsequently lysed and analyzed by immunoblotting as described previously (10). Transfection of pCMV5-HA or FLAG-tagged SCP1-3, Smad1, or Smad2 constructs (2 g/10cm-diameter dishes unless stated otherwise) into HEK293 cells (60% confluent at the time of transfection) was performed using Lipofectamine 2000 reagent as described above. FLAG immunoprecipitations were performed with 0.5 mg of protein lysates using FLAG-agarose beads, washed twice in 1 ml of TE buffer (Tris-HCl, pH 7.5, 0.1 mM EGTA, and 0.1% 2-mercaptoethanol) containing 0.5 M NaCl, twice in TE buffer and reconstituted in TE buffer. For phosphatase assays, 10% of these FLAG immunoprecipitates from Smad1-or Smad2transfected lysates were used for each reaction for the SCP phosphatase assays. Immunoblot analysis was performed as described previously (10).
Xenopus Embryo Manipulations-Xenopus embryo injections and manipulations were performed as described (18). For RNA synthesis, plasmids were linearized with AscI, and sense strand capped mRNAs were synthesized in vitro using the mMESSAGE mMACHINE system (Ambion) with SP6 RNA polymerase.
Isolation of GST-SCP1-3 and Phosphatase Assays-pGex6P-1 vectors encoding SCP1-3 were transformed into E. coli BL21 cells, which were cultured at 37°C until the A 600 was 0.6. Expression of GST-tagged SCP1-3 was induced by isopropyl-␤-D-galactosidase (250 M) at room temperature for 16 h. Cells were sonicated in ice-cold TE buffer containing 0.5 M NaCl, and GST-SCP1-3 was isolated by incubating cleared lysates in glutathione-Sepharose beads (Amersham Biosciences) following the manufacturer's protocol. Proteins were eluted using 20 mM glutathione and dialyzed in TE buffer containing 0.27 M sucrose prior to snap-freezing in liquid nitrogen and storage at Ϫ80°C. 10 pmol each of GST-SCP1-3 was incubated in a 50-l phosphatase assay mix containing 50 mM Tris-HCl, pH 7.5, 1 mM dithiothreitol, 20 mM MgCl 2 , and FLAG immunoprecipitates of linker-phosphorylated Smad1 or Smad2 (5 l of packed beads of FLAG-agarose per assay) as described above. The assays were performed at 30°C for 1 h with constant rocking and were stopped by boiling the assays in SDS sample buffer for 5 min. 10% of the reaction mix was immunoblotted with Smad1-LP or Smad2-LP as well as anti-FLAG and anti-GST antibodies.
Luciferase Assays and Quantification of mRNA by Real-time PCR Analysis-Luciferase assays in OsA-CL cells were performed with a mammalian TGF␤-inducible luciferase reporter construct (SBE-4X) as described previously (19). Quantification of mRNA by real-time PCR analysis was performed as described previously (10). The primers used for real-time PCR (Smad7, p15Ink4b, p21CIP1, and JAGGED1) have been described previously (20).

SCPs Dephosphorylate Smad1 and Smad2 at the Linker Region-
MAPKs and cyclin-dependent kinases (CDKs) are known to phosphorylate various serine and threonine residues in SP and TP sequences within the linker region of Smad1 (Ser-187, Ser-195, Ser-206, Ser-214) and Smad2 (Thr-220, Ser-245, Ser-250, Ser-255) (Fig. 1A), restricting Smad activity (3,4). Using the human keratinocyte cell line HaCaT, we recapitulated these observations by using antibodies that recognize Smad1 phosphorylated at Ser-206 (Smad1-LP) or Smad2 phosphorylated at Ser-245, Ser-250, and Ser-255 (Smad2-LP; Fig. 1B). Additionally, EGF induces phosphorylation of the Smad1 linker at Ser-187, Ser-195, and Ser-214 (3). Interestingly, BMP also induced phosphorylation of the Smad1 linker region at Ser-206 in addition to phosphorylation of the C-terminal SXS motif (Fig. 1B); this was not mediated through Erk MAPK, as BMP did not induce phosphorylation and activation of Erk MAPK (Fig. 1B). 4 Furthermore, TGF␤ also induced phosphorylation of the Smad1 linker at Ser-206 as well as the Smad2 linker region, albeit not to the same extent as phosphorylation induced by EGF (Fig. 1B). Smad1 phosphorylated at the linker region in response to EGF fractionated with the cytosolic fraction, where EGF-activated ERK was also detected (Fig. 1, C and D). In contrast, Smad1, which was phosphorylated at the linker region in response to BMP, fractionated mainly with the nuclear fraction (Fig. 1C) as did Smad2, phosphorylated at the linker region in response to TGF␤ (Fig. 1D). These results suggested that BMP-and TGF␤induced linker phosphorylation may occur after the nuclear translocation of Smads. In HEK293 cells transfected with FLAG-Smad1, we observed that the linker region of Smad1 4 G. Sapkota, C. Alarcon, and J. Massagué , unpublished results. , the phosphorylated C-terminal tails (Smad1-TP and Smad2-TP), and total Smad1 and Smad2/3. Phospho-ERK was used as control for EGF stimulation. C, HaCaT cells were treated with EGF or BMP as described in B. Cytosolic and nuclear fractions were analyzed for Smad1-LP. Histone 1B and ␣-tubulin Western immunoblotting was used as controls for nuclear and cytosolic fractions, respectively. D, same as in C, except cells were treated with or without TGF␤, and nuclear (N), cytosolic (C), or whole cell lysates were immunoblotted with the indicated antibodies. E, HEK293 cells expressing FLAG-tagged Smad1 or Smad2 were treated with or without EGF, and lysates were immunoblotted with the indicated antibodies.
was spontaneously phosphorylated, and treatment of cells with EGF did not yield a noticeable increase in linker phosphorylation of Smad1 (Fig. 1E). FLAG-Smad2 overexpressed in HEK293 cells is also phosphorylated at the linker region, and EGF moderately enhances the linker phosphorylation of FLAG-Smad2 (Fig. 1E).
We used these conditions to investigate in mammalian cells the possible role of SCPs as phosphatases acting on agonistinduced linker phosphorylation of endogenous Smads and overexpression-induced linker phosphorylation of ectopically expressed Smads. The co-expression of human SCP1, SCP2, or SCP3 with FLAG-Smad1 led to an almost complete loss of linker phosphorylation in overexpressed Smad1 ( Fig. 2A). When co-expressed with the catalytically inactive mutants of SCP1, SCP2, or SCP3, the linker phosphorylation of FLAG-Smad1 was unaffected, indicating that the phosphatase activity of the SCPs is required for the loss of linker phosphorylation. SCPs have been shown previously to act as Smad1 C-terminal phosphatases on endogenous Smad1 protein in mammalian cells and Xenopus embryos (10). However, in the present conditions of Smad1 protein overexpression, the level of BMP-induced C-terminal phosphorylation of Smad1 (at the SXS motif) was left intact or only slightly decreased by co-expression of FLAG-Smad1 with SCP1, -2, or -3 ( Fig. 2A). The dephosphorylation of the FLAG-Smad1 linker region by SCP1, -2, and -3 was dependent on the dose of these phosphatases (Fig. 2B). Surprisingly, the level of BMP-induced Smad1 C-terminal phosphorylation was significantly increased when FLAG-Smad1 was co-expressed with the catalytically inactive mutants of SCP1, -2, or -3 ( Fig. 2A), suggesting that these mutants may act as dominant negative forms over endogenous C-terminal phosphatases. Collectively, these results raised the possibility that the SCPs are Smad1 linker phosphatases. Furthermore, these results also suggested that the role of SCPs as Smad1 C-terminal phosphatases may depend on rate-limiting factors that are insufficient for the activity of overexpressed SCPs and become sequestered by overexpression of inactive forms of SCPs.
Xenopus SCP2 Dephosphorylates Smad1 and Smad2 at the Linker Regions-We have previously shown that dephosphorylation of the C-terminal SXS sites by Xenopus laevis SCP2 (XSCP2) is selective for Smad1 compared with Smad2/3 (10). We next assessed whether linker phosphorylation of Smad1 and Smad2 in Xenopus embryos is affected by XSCP2. We observed linker phosphorylation of endogenous Smad1 and Smad2 in the unfertilized eggs, with a subsequent drop after fertilization and early embryogenesis stages (Fig. 3A). Analysis of stage 6 (early blastula), 8 (midblastula), and 11 (midgastrula) embryos indicated that linker-phosphorylated Smad1 was absent at stage 6 and reappeared at stages 8 and 11, whereas linker-phosphorylated Smad2 was present at stage 6 but nearly absent at stages 8 and 11 (Fig. 3A). The expression of total Smad1 and Smad2 (Fig. 3A), as well as the status of the tail phosphorylation of Smad1 and Smad2 (data not shown) during these stages, was consistent with our previous report (10).
To test whether XSCP2 was able to dephosphorylate the linker region of Smad1 and Smad2 in Xenopus embryos, we injected synthetic RNAs encoding XSCP2 or catalytically inactive mutant of XSCP2 into both blastomeres of two-cell stage embryos. Embryos were harvested at mid-gastrulation (stage 11) and examined for the overall level of Smad1 phosphorylation. The expression of XSCP2 caused not only a loss of C-terminal phosphorylation, as reported previously (10), but also a loss of linker phosphorylation (Fig. 3B). Catalytically inactive XSCP2 did not induce a loss of linker or C-terminal Smad1 phosphorylation (Fig. 3B). For the analysis of Smad2 linker  DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 phosphorylation, XSCP2 or the inactive XSCP2 mutant were injected into two-cell stage embryos, and the embryos were harvested at stages 6.5 and 8, which is when Smad2-linker phosphorylation was observed (refer to Fig. 3A). As with Smad1, wild type XSCP2, but not the catalytically inactive mutant of XSCP2, caused a marked loss of Smad2 linker phosphorylation (Fig. 3C).

Dephosphorylation of R-Smad Linker Region by SCPs
Recombinant SCPs Dephosphorylate Smad1 and Smad2 at the Linker Sites in Vitro-We investigated the ability of recombinant GST-SCP1, -2, or -3 to dephosphorylate the linker sites of Smad1 and Smad2 in an in vitro phosphatase assay. As substrates, we used FLAG immunoprecipitates from HEK293 cells expressing FLAG-Smad1 or FLAG-Smad2, treated with EGF to induce maximal levels of linker phosphorylation. Incubation of these proteins with human recombinant GST-SCP1, GST-SCP2, or GST-SCP3 caused a nearly complete dephosphorylation of the linker sites of FLAG-Smad1 (Fig. 4A) and FLAG-Smad2 (Fig. 4B).
SCP1/2 Depletion Enhances Linker Phosphorylation of Smad1 and Smad2-To establish the effect of endogenous SCPs on the phosphorylation of the linker region of endogenous Smad1 and Smad2, we performed a loss-of-function analysis by introducing previously described siRNA oligonucleo-tides targeting SCP1 (iSCP1) and SCP2 (iSCP2) in human HaCaT keratinocytes (10). Expression of iSCP1 and iSCP2 resulted in the reduction of the respective mRNA levels by more than 80% (Fig. 5A). As a negative control we used iFoxO4, which effectively reduces the expression of FOXO4, without affecting the expression of two functionally redundant genes, FOXO1 and FOXO3 (20), or the expression of SCP1 or SCP2 (Fig. 5A) (10). Transfection of iSCP1/2 resulted in increased accumulation of phospho-linker Smad1 and phospholinker Smad2 in response to BMP and TGF␤, respectively (Fig. 5B), compared with iFoxO4. As reported previously (10), iSCP1/2 expression resulted in an increase in phospho-Smad1 tail in response to BMP, whereas the levels of TGF␤-induced phospho-Smad2 tail were unchanged in cells expressing iSCP1/2 (Fig. 5B). Similar results on the effects of iSCP1/2 were obtained in OsA-CL osteosarcoma cells, in which SCP2 is naturally amplified (data not shown).
Distinct Effects of SCP1/2 Depletion on TGF␤ and BMP Gene Responses-We next investigated whether the transcriptional ability of Smad1 and Smad2 was affected by knockdown of SCP1/2. Transfection with iSCP1/2 decreased the ability of TGF␤ to activate a luciferase reporter construct driven by an artificial Smad2/3-responsive promoter (SBE4X-luciferase) (Fig. 6A). This effect was observed in HaCaT cells and was even more pronounced in OsA-CL osteosarcoma cells, which express high levels of endogenous SCP2 naturally (16). We therefore used the OsA-CL cells to analyze the effects of SCP1/2 depletion on TGF␤ and compared these effects with those that we had described previously with BMP gene responses (10). In agreement with our previous results, the ability of SCP1/2 to mediate C-terminal dephosphorylation of Smad1 is coupled to an ability of iSCP1/2 to increase the BMPdependent induction of a known target gene, ID1 (Fig. 6B). This positive effect of iSCP1/2 on the BMP responsiveness of these cells may reflect a dominant effect of the increase in C-terminal phosphorylation over the increase in linker phosphorylation on the activity of Smad1 under these conditions (refer to Fig. 5B). In contrast to this gain in BMP responsiveness, the transfection of iSCP1/2 cause a marked decrease in the ability of TGF␤ to induce the expression of typical target genes, including SMAD7, p15INK4B, p21CIP1, and JAGGED1 (Fig. 6C). Smad7 is an inhibitory Smad that negatively regulates TGF␤ signaling by competing with R-Smads for receptors (21,22) and by targeting receptors for degradation (23,24). p15Ink4b and p21Cip1 are CDK inhibitors that mediate the cytostatic effect of TGF␤ (25,26), and Jagged1 is a Notch-signaling ligand implicated in epithelial-to-mesenchymal transitions induced by TGF␤ (27).

DISCUSSION
In this study we present evidence that human SCP1-3 as well as Xenopus SCP2 act as phosphatases for the regulatory phosphorylations in the linker region of Smad1 and Smad2. Enhanced expression of SCP2 in mammalian cells and in Xenopus embryos leads to an extensive loss of linker phosphorylation of Smad1 and Smad2. Human SCP1-3 efficiently dephosphorylate the linker region of Smad1 and Smad2 in vitro. Furthermore, RNAi-mediated knockdown of SCP1/2 leads to an increase in the levels of BMP-induced Smad1 linker phosphorylation as well as TGF␤-induced Smad2 linker phosphorylation. In cells expressing SCP1/2 siRNA, the TGF␤-induced luciferase reporter activity and the expression of TGF␤ target genes are significantly reduced compared with cells expressing control siRNA and in contrast to an increase in BMP responsiveness. Taken together, these results imply that SCPs regulate the BMP and the TGF␤ pathways by dephosphorylating the linker regions of Smad1 and Smad2 but do so with different outcomes depending on the pathway.
While our manuscript was in preparation, Wrighton et al. (17) reported that human SCP1-3 act as linker phosphatases for Smad2/3. To the extent that the findings of the two studies overlap, our findings are in complete agreement with this conclusion. Using in part different mammalian cell systems, both studies show that SCP1-3 act as Smad2/3 linker phosphatases in vitro as well as in vivo, that RNAi-mediated depletion of SCPs leads to enhancement of Smad2 linker phosphorylation, and that overexpression of active SCPs leads to enhancement of TGF␤ responses. However, our work additionally highlights several important points. We provide evidence that SCPs act as linker phosphatases for Smad1 as well as Smad2. We show that SCPs act as linker phosphatases under conditions in which linker phosphorylation is induced by BMP and TGF␤ themselves, in the context of feedback loops triggered by these two agonists. And, by demonstrating the role of XSCP2 as a Smad1 and Smad2 linker phosphatase in Xenopus embryos, we have established the generality of this role beyond mammalian cell systems. Importantly, together with these findings, our comparative analysis highlights the distinct outcomes of the effect of SCPs as Smad linker phosphatases in the BMP and TGF␤ pathways. Although the role of SCPs on Smad2/3 appears to be limited to de-inhibition of the TGF␤ pathway by removal of inhibitory linker phosphorylation, their effect on Smad1 is to reset it to the base-line state by dephosphorylating both the linker region and the C-terminal tail. These proposed roles are summarized schematically in Fig. 7.
We provide evidence that SCPs can induce the dephosphorylation of linker SP sites in Smad1 and Smad2/3 regardless of whether the phosphorylation occurred endogenously in Xenopus embryos or in response to BMP, TGF␤, or EGF in mammalian cells. Therefore, SCPs may play a broad role as antagonists of diverse feedback and cross-talk inputs into the linker region of the R-Smads.
The ability of SCPs to also dephosphorylate Smad1, but not Smad2/3, at the C-terminal sites has been shown by overexpression of the SCP proteins in Xenopus embryos, by knockdown of their expression in mammalian cells, and by in vitro phosphatase assays using recombinant SCPs (10) (see also Figs. 3 and 5). The SCPs differ in this selectivity from PPM1A, a nuclear phosphatase that mediates C-terminal dephosphorylation of both Smad1 and Smad2/3 (11,12). In different studies, the knockdown of SCPs or of PPM1A in mammalian cells consistently caused only a partial decrease in Smad1 C-terminal dephosphorylation (10 -12), suggesting that C-terminal dephosphorylation may be mediated by combinations of these different nuclear phosphatases and probably by non-nuclear phosphatases such as pyruvate dehydrogenase phosphatase (13). The ability of the SCPs to mediate the C-terminal dephosphorylation of Smad1 but not Smad2/3 gives rise to concordant consequences, as shown by loss-of-function experiments. The knockdown of SCP1/2 caused a marked decrease in typical TGF␤ gene responses (present results and Ref. 17), as would correspond to the loss of phosphatases that remove inhibitory phosphorylation from the linker region of Smad2/3. In contrast, the knockdown of SCP1/2 causes an increase in BMP responses (present results and Ref. 10), as would correspond to the loss of phosphatases that remove activating C-terminal phosphorylation in addition to removing inhibitory linker phosphorylation (Fig. 7).
The dual role of the SCPs as Smad1-3 linker phosphatases and Smad1 C-terminal phosphatases is surprising, given the differences in the amino acid sequences of these two types of  DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 JOURNAL OF BIOLOGICAL CHEMISTRY 40417 phosphorylation sites. However, several observations raise the possibility that the SCPs act on Smad1 linker and C-terminal sites with the involvement of distinct cofactors. Support for this idea comes from our observation that when SCP1-3 were co-expressed with Smad1 in mammalian cells, the Smad1 linker phosphorylation was markedly diminished, whereas the BMP-induced tail phosphorylation was largely unaffected. Furthermore, the co-expression of catalytically inactive SCPs mutants resulted in a clear enhancement of BMP-induced C-terminal phosphorylation of Smad1, suggesting that the inactive SCPs may sequester endogenous factors required for C-terminal dephosphorylation of Smad1 by endogenous SCPs. Of note, no evidence of a limiting cofactor was manifest in our experiments with Xenopus embryos, as SCP overexpression in this system led to extensive dephosphorylation at both the linker sites and the C-terminal sites in Smad1.

Dephosphorylation of R-Smad Linker Region by SCPs
It has been a decade since it was first reported that the Smad1 and 2 proteins are activated by C-terminal phosphorylation (28,29) and inhibited by phosphorylation at the linker region (3,4). In contrast to the clearly activating role of the C-terminal phosphorylation events, the regulation of Smad activity by phosphorylation and dephosphorylation of the linker region is complex. It is becoming increasingly clear that linker phosphorylation of Smads plays key roles in regulating Smad activity in mature cells and during embryogenesis, and different proline-directed kinases appear to be involved depending on the cell environment and the nature of the activating signals (3)(4)(5)7). The present results, together with recent results from us and others (10 -13, 17), suggest that a complex set of protein phosphatases is acting in opposition to these multiple proteins kinases in the control of Smad activity.