Protein Serine/Threonine Phosphatase PPM1A Dephosphorylates Smad1 in the Bone Morphogenetic Protein Signaling Pathway*>

Bone morphogenetic proteins (BMPs) are secreted polypeptides belonging to the transforming growth factor-β (TGF-β) superfamily that activates a broad range of biological responses in the metazoan organism. The BMP-initiated signaling pathway is under tight control by processes including regulation of the ligands, the receptors, and the key downstream intracellular effector Smads. A critical point of control in BMP signaling is the phosphorylation of Smad1, Smad5, and Smad8 in their C-terminal SXS motif. Although such phosphorylation, which is mediated by the type I BMP receptor kinases in response to BMP stimulation, is well characterized, biochemical mechanisms underlying Smad dephosphorylation remain to be elucidated. In this study, we have found that PPM1A, a metal ion-dependent protein serine/threonine phosphatase, physically interacts with and dephosphorylates Smad1 both in vitro and in vivo. Functionally, overexpression of PPM1A abolishes BMP-induced transcriptional responses, whereas RNA interference-mediated knockdown of PPM1A enhances BMP signaling. Collectively, our study suggests that PPM1A plays an important role in controlling BMP signaling through catalyzing Smad dephosphorylation.

Bone morphogenetic proteins (BMPs), 3 originally identified by their ability to cause bone differentiation (1), are signaling molecules that belong to the transforming growth factor-␤ (TGF-␤) superfamily. Presently, the biological functions of BMPs have been greatly expanded. BMPs regulate skeletal development as well as many non-osteogenic developmental processes, such as mesoderm patterning, left-right symmetry, neuronal patterning, and hematopoiesis (2)(3)(4)(5). Accumulating evidence indicates that BMPs play an important role in the regulation of stem cell properties (3, 6 -8). Signals from BMP ligands are transduced through binding to type I and II receptors on the cell surface, where type II receptors activate type I receptors, which in turn phosphorylate the downstream Smad1, Smad5, and Smad8. The phospho-R-Smads (P-Smads) form a complex with Smad4 and translocate into the nucleus, where they bind to the Smad binding sites and cooperate with other transcription factors to regulate BMP-induced gene expression (5,9,10).
Despite substantial effort devoted to understanding the actions of BMP/TGF-␤ and Smads, the precise regulation of Smads remains enigmatic. Regulation of Smads can be accomplished via various post-translational mechanisms, including phosphorylation and ubiquitin-dependent modifications (11,12). Among these, BMP-induced phosphorylation of Smad1/ 5/8, which is carried out by the BMP type I receptor (BMPRIA or BMPRIB) and occurs at the C-terminal SXS motif of Smad1/ 5/8, represents the most critical step in Smad signaling. The SXS phosphorylation triggers a cascade of intracellular events from Smad complex assembly in the cytoplasm to transcriptional control in the nucleus.
The reversible phosphorylation and dephosphorylation represents a fundamental strategy used by eukaryotic organisms to regulate a battery of biological functions. The cellular protein phosphorylation state is modulated by protein kinases and phosphatases. Until recently, how Smad functions are fine-tuned by Smad dephosphorylation remains poorly understood. R-Smads undergo continuous nucleocytoplasmic shutting, and their export from the nucleus requires dephosphorylation (13,14). A recent work identifies pyruvate dehydrogenase phosphatase (PDP) as a Smad phosphatase in the decapentaplegic pathway in Drosophila (15). On the other hand, we have found that PPM1A (but not PDPs) acts as a phosphatase for TGF-␤-activated Smad2 and Smad3 (16).
In this study, we have explored whether Smad1 is dephosphorylated by PPM1A. We have demonstrated that PPM1A physically interacts with and dephosphorylates Smad1. Furthermore, overexpression of PPM1A attenuates or abolishes BMP-induced transcription, and conversely RNA interference-mediated knockdown of PPM1A enhances BMP signaling. Thus, PPM1A-mediated dephosphorylation provides a novel mechanism in terminating BMP signaling.

EXPERIMENTAL PROCEDURES
Plasmids-Expression plasmids for epitope-tagged Smads have been previously described (17,18). PPM1A was obtained * This research was supported by National Institutes of Health Grants R01DK073932 and R21CA11293 (to X. L.) and R01GM63773 and R01CA108454 (to X.-H. F.). 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. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. S1 and S2. 1  were also obtained by PCR and similarly cloned into pRK5F. GCCGlux (19), Id1-luc (20), and Xvent-luc (21) were kindly provided by Kohei Miyazono, Peter ten Dijke, and Christof Niehrs, respectively. Cell Culture, Cell Transfection, Immunoprecipitation, and Western Blotting-HEK293T, HaCaT, Mv1Lu-tTA (22), HepG2, and C2C12 cells were cultured and transfected using Lipofectamine (Invitrogen) as previously described (16,17). Stable HaCaT cells with PPM1A knockdown and stable Mv1Lu cells with tet-off expression of PPM1A have been described elsewhere (16).
Immunoprecipitations were carried out as described previously (16,17). 293T cells were transiently transfected with cDNAs for HA-Smad1 and FLAG-PPM1A (for Smad1-PPM1A interaction) or FLAG-Smad1 and Myc-Smad4 (for Smad1-Smad4 interaction). Anti-HA or anti-FLAG antibodies were used to immunoprecipitate Smad1 from transfected cell lysates. To detect Smad1-bound PPM1A, the immunoprecipitated proteins were separated by SDS-PAGE, transferred onto nitrocellulose membrane, and immunostained with primary antibodies and finally detected by horseradish peroxidase-conjugated secondary antibodies and visualized by chemiluminescence (Pierce).
Antibodies against Smad1 (Cell Signaling) and PPM1A (Abcam) were used at 1:1000 dilutions in Western blotting assays to detect levels of endogenous proteins. P-Smad1 (Cell Signaling) was used to measure the level of P-Smad1.
Glutathione S-Transferase (GST) in Vitro Binding Assay-Production and purification of the GST-Smad1 fusion protein was done following the manufacturer's instructions (Amersham Biosciences). In vitro translation of PPM1A and in vitro binding assays were essentially carried out using the TNT kit (Promega) as previously described (16,17).
Real-time Quantitative RT (qRT)-PCR-Total RNAs were prepared using TRIzol reagent (Invitrogen) from HaCaT To stimulate Smad1 phosphorylation, either BMP2 (25 ng/ml) was added to cells for 1 (lanes 2 and 3) or constitutively active BMP type I receptor HA-ALK3QD was co-transfected (lanes 5 and 6). Phospho-Smad1 (P-Smad1), total Smad1, and PPM1A levels were determined by Western blotting. B, PPM1A dephosphorylates Smad5. Experiments were carried out as in Fig. 1A. C, PPM1A dephosphorylates Smad8. Experiments were carried out as described for A. D, PPM1A point mutation (D239N or R174G) abolishes its catalytic activity toward P-Smad1. Experiments were carried out as described for A. E, N-terminal deletion of PPM1A (Met-29, Met-36, or Met-112) abolishes its activity toward P-Smad1. Experiments were carried out as described for A. F, PPM1A directly dephosphorylates P-Smad1 in vitro. PPM1A was purified by anti-FLAG immunoprecipitation from cell lysates transfected with His-PPM1A, whereas P-Smad1 was purified by anti-His immunoprecipitation from cell lysates transfected with F-Smad1 and ALK3(Q233D). Purified PPM1A and P-Smad1 were mixed in Eppendorf tubes, and the P-Smad1 level was determined by anti-P-Smad1 Western blotting. G, 26 S proteasome inhibitor MG-132 has no effect on Smad1 dephosphorylation by PPM1A. P-Smad1 level was determined as described for A. MG-132 treatment at 20 M was done for 4 h. H, endogenous Smad1 phosphorylation is abolished by induced expression of PPM1A. Mv1Lu-stable cells expressing PPM1A under the control of the tet-off promoter were used. Withdrawal of doxycycline (ϪDox)-induced F-PPM1A expression. Exogenous (F-PPM1A) and endogenous PPM1A are indicated. treated with BMP2 (25 ng/ml) for 0, 4, 8, and 24 h. qRT-PCR was carried out using the Assay-on-demand kit (Applied Biosystems). mRNA levels of p21 and Id1 were normalized against 18 S RNA. Each target was measured in triplicates. Data were analyzed using Excel.
Reporter Assays-HepG2 cells at 25-30% confluency were co-transfected with expression plasmids for PPM1A (indicated in the text) and reporter plasmids. Cells were transfected with reporter plasmids (e.g. GCCG-lux, Id1-luc, Xvent-luc, or FRluc) and expression plasmids for PPM1A and then treated for 12 h with or without 25 ng/ml BMP2. BMP2-induced transcrip-tion was analyzed by measuring luciferase activity using a Luciferase Assay kit (Promega).

RESULTS
PPM1A Dephosphorylates Smad1-In a search to study Smad2/3 dephosphorylation in the phospho-SXS motif, we recently identified PPM1A as the phosphatase for Smad2 and Smad3 that is responsible for termination of TGF-␤ signals (Lin et al., (16)). Considering the highly conserved nature of the SXS motif in all R-Smads, we reasoned that PPM1A might also recognize the SXS motif in the BMP-activated Smad1. To test this, BMP-induced Smad1 phosphorylation was analyzed in HeLa cells that were transfected with HA-Smad1 and His-PPM1A. The phosphorylation level of Smad1 (P-Smad1) was determined by a phospho-SXS motif-specific antibody. In the absence of PPM1A, the level of P-Smad1 increased upon 1 h of BMP2 stimulation (Fig. 1A, lane 2). In contrast, co-transfection of His-PPM1A abolished BMP-induced Smad1 phosphorylation (Fig. 1A, lane 3), suggesting that PPM1A either prevents Smad1 phosphorylation or directly dephosphorylates Smad1. To rule out the possibility that PPM1A dephosphorylates BMP receptors to prevent Smad1 phosphorylation and activation, we used a mutant of BMP receptor ALK3(Q233D) that constitutively activates Smad1. Clearly, P-Smad1 level was increased by ALK3(Q233D) (Fig. 1A, lane 5), and this increase was abolished by the co-expression of PPM1A (Fig. 1A, lane 6).
Smad1, Smad5, and Smad8 are three R-Smads transducing BMP signals in vertebrates. After having established that PPM1A dephosphorylates Smad1, we then carried out an experiment to test whether PPM1A also dephosphorylates Smad5 and Smad8. The results of analysis showed that PPM1A dephosphorylated P-Smad5 (Fig. 1B) and P-Smad8 (Fig. 1C), which were recognized by the same P-Smad1 antibody (␣PS1).
We next determined whether the phosphatase activity of PPM1A is essential in reducing Smad1 phosphorylation. PPM1A point mutants at its catalytic domain (D239N and R174G) were generated (23). As shown in Fig. 1D, both D239N and R174G mutants were unable to eliminate ALK3(Q233D)induced P-Smad1. In addition, three N-terminal deletion mutants (Met-29, Met-36, and Met-112) of PPM1A were tested for their Smad1-dephosphorylating activity in 293T cells. We found that all three deletion mutants lost their activity toward Smad1 dephosphorylation (Fig. 1E). These results suggest that PPM1A reduces the P-Smad1 level through its phosphatase activity.
To further determine whether the effect of PPM1A on P-Smad1 accumulation is due to direct dephosphorylation of P-Smad1, an in vitro phosphatase assay was performed using purified PPM1A and P-Smad1 (Fig. 1F). P-Smad1 and PPM1A (wild type or mutants) were separately purified by immunoprecipitation from 293T cells, which were transfected with either F-Smad1 or His-PPM1A. Results in Fig. 1F showed clearly that the P-Smad1 level was reduced by wild-type PPM1A but not by either the D239N or R174G mutant, suggesting PPM1A directly dephosphorylates Smad1.
To exclude the possibility that PPM1A causes reduction in Smad1 phosphorylation dependent on the 26 S proteasome, the proteasome inhibitor MG-132 was included. Inhibition of the FIGURE 2. PPM1A physically interacts with Smad1. A, HA-Smad1 and FLAG-PPM1A were transfected in 293T cells. Smad1 was immunoprecipitated (IP) with anti-HA antibodies and then subjected to SDS-PAGE and Western blots (IB) with anti-FLAG antibodies to detect the Smad1-bound PPM1A. WCL, whole cell lysate. B, direct interaction between PPM1A and Smad1. In vitro translated, 35 S-labeled His-PPM1A or D239N and R174G mutants were incubated with purified glutathione bead-bound GST protein or GST-Smad1 fusion protein. The precipitated complex was subjected to SDS-PAGE followed by autoradiography. The quality of GST fusions is shown in supplemental Fig. S1. C, PPM1A interacts with both the N and C termini of Smad1. In vitro translated, 35 S-labeled His-PPM1A was incubated with an equal amount of GST proteins as follows: GST alone, GST-Smad1 full-length, N-terminal region (amino acids 1-144), linker region (amino acids 144 -268), or C-terminal region (amino acids 268 -465) as described in B. The quality of the GST fusions is shown in supplemental Fig. S1. PPM1A on the P-Smad1 level (Fig. 1G). This suggests that proteasome-dependent degradation does not contribute to PPM1A-induced loss of Smad1 phosphorylation.
Finally, we examined the effect of PPM1A on the dephosphorylation of endogenous Smad1. Mv1Lu cell lines that stably express PPM1A under the control of the tet-off promoter were established. Withdrawal of doxycycline (ϪDox), a tetracycline derivative, induced PPM1A expression (Fig. 1H, lanes 3 and 4). As a control, BMP2 treatment (1 h) resulted in a strong increase in the endogenous P-Smad1 level (lane 2). However, BMP2induced Smad1 phosphorylation was eliminated when PPM1A expression was induced (ϪDox) (Fig. 1H, lane 4).
PPM1A Physically Interacts with Smad1-To determine whether PPM1A interacts with Smad1, we carried out a coimmunoprecipitation experiment. FLAG-PPM1A was co-transfected with HA-Smad1 in 293T cells. Results from coimmunoprecipitation showed that PPM1A could interact with Smad1 ( Fig. 2A, lane 3). To determine whether there is a direct interaction between PPM1A and Smad1, an in vitro GST binding assay was performed. In vitro translated, 35 Slabeled wild type (WT), D239N, or R174G mutant of PPM1A was incubated with purified GST or GST-Smad1 fusion protein immobilized to glutathione-Sepharose beads. Wild-type PPM1A and its mutants clearly interacted with GST-Smad1 fusion protein but not with GST protein alone (Fig. 2B). This result indicates a direct interaction between PPM1A and Smad1 protein. To map the PPM1A binding domain on Smad1, recombinant GST fusions of the Smad1 N terminus, linker region, and C terminus were tested for their binding to in vitro translated, 35 S-labeled His-PPM1A. Interestingly, PPM1A interacts with both the N and C termini of Smad1; no interaction was found on the linker region (Fig. 2C).
Ectopic Expression of PPM1A Attenuates BMP Signaling-Upon BMP ligand binding, Smad1 is phosphorylated by the type I receptor and then forms a complex with Smad4, which translocates into the nucleus, where they exert their transcriptional regulation. After having determined that PPM1A dephosphorylates Smad1, we sought to investigate whether PPM1A affects accumulation of the Smad complex. Presence of the complex between FLAG-Smad1 and Myc-Smad4 was examined in 293T cells co-transfected with either WT or the D239N mutant of PPM1A. As a control, BMP2 induced an interaction between Smad1 and Smad4 (Fig. 3A, compare lanes  1 and 2). This interaction was significantly reduced in the presence of wild-type PPM1A (Fig. 3A, lane 3), whereas PPM1A D239N mutant had no effect on the Smad1-Smad4 interaction (Fig. 3A, lane 4), suggesting PPM1A antagonizes accumulation of the Smad1-Smad4 complex. In the light of its nuclear localization of PPM1A (data not shown), PPM1A may cause dissociation of the nuclear Smad1-Smad4 complex.
Because PPM1A dephosphorylates Smad1, we reasoned that PPM1A might inhibit the transactivation activity of Smad1. Using a Gal4 reporter system, where Smad1 was fused with the Gal4 DNA binding domain (Gal4-Smad1), we assessed the transcriptional activity of Smad1 on FR-luc, a heterologous Gal4 binding reporter gene. As expected, transcriptional activity of Smad1, as reflected by FR-Luc luciferase activity, was induced by BMP2 treatment (Fig. 3B). Co-transfection of wild- Transcription activity from the reporter plasmid was measured. Assays were performed in the presence (ϩ) or absence (Ϫ) of BMP2. Assays were done in triplicates. C, PPM1A inhibits BMP2-induced GCCG-lux reporter activity. HepG2 cells were transfected with PPM1A together with GCCG-lux plasmid. BMP2 treatment and luciferase assay were as described under "Experimental Procedures." D, PPM1A inhibits BMP2-induced Id1-luc reporter activity. HepG2 cells were transfected with WT PPM1A or PPM1A mutants (D239N and R174G) together with Id1-luc plasmid. BMP2 treatment and luciferase assay were performed as described in C. E, PPM1A inhibits BMP2-induced Xvent-luc reporter activity. DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48 type PPM1A (but not R174G mutant) abolished BMP2-induced transcriptional activity of Smad1.

Dephosphorylation of Activated Smad1 by PP2C
We next determined the effect of PPM1A on BMP-regulated Smad-dependent gene transcription. To this end, we used a synthetic reporter construct GCCG-lux, which contains multiple repeats of the GCCG motif and specifically responds to BMP stimulation but not to TGF-␤ or activin (19). As shown in Fig. 3C, BMP2 treatment caused an increased GCCG-lux reporter activity in HepG2 cells. In contrast, the presence of PPM1A completely blocked BMP2-dependent expression of GCCG-lux.
Several BMP early target genes have been reported. Among them, Id-1 encodes a protein that acts as a positive regulator of cell proliferation and a negative regulator of cell differentiation (24). Xvent1 is another BMP direct target gene, which is a member of the vox/vent homeobox gene family and is involved in dorsoventral axis determination (25). We studied the effect of PPM1A on the transcriptional activity of these target genes. As shown in Fig. 3D, BMP2 treatment caused an increased promoter activity of Id-1 in HepG2 cells. Co-expression of PPM1A completely eliminated this induction. Moreover, PPM1A D239N and R174G mutants did not affect Id1 promoter activity. Similarly, expression of PPM1A profoundly inhibited the transcription of Xvent-luc reporter gene (Fig. 3E).
Knockdown of PPM1A Enhances BMP Signaling-Having established the inhibitory effect of PPM1A overexpression on BMPinduced Smad1 phosphorylation and Smad1-dependent transcriptional responses, we took a loss-offunction approach to knock down PPM1A expression by RNA interference. Expression of small interfering RNA against human PPM1A (shPPM1A494) could efficiently knock down the exogenous expression of human PPM1A but not zebrafish PPM1A (Fig. 4A). As a result, shPPM1A blocked Smad1 dephosphorylation induced by hPPM1A (lane 3). The effect of PPM1A knockdown could be rescued by the expression of zPPM1A (lane 5). zPPM1A, which was resistant to shPPM1A-mediated knockdown, led to Smad1 dephosphorylation in the presence of shPPM1A494 (lane 6).
We also assessed the effect of PPM1A knockdown on BMP2induced promoter activity. When the shPPM1A plasmid was transfected in HepG2 cells, Id1 promoter activity was markedly increased compared with control cells (Fig. 4B). Furthermore, expression of zPPM1A abolished the increased induction of Id promoter activity by shPPM1A (Fig. 4B). We also noticed that HepG2 cells exhibited a weak p21 response to BMP2. Notably, shPPM1A could sensitize the cells to respond to BMP2, and conversely, zPPM1A completely shuts off BMP2-induced p21 response (Fig. 4C). We have stably knocked down PPM1A expression in HaCaT cells (Fig. 4D). In PPM1A-depleted HaCaT-stable cells, called shPPM1A-stable cells, the level of P-Smad1 induced by BMP2 was profoundly higher than that in control cells (Fig. 4D, lanes  2 and 4). To characterize the effect of PPM1A knockdown on BMP transcriptional responses, we examined the levels of p21 and Id1 mRNAs in HaCaT stable cells. p21 and Id1 are two representative Smad-dependent BMP target genes (20,26). In control cells, BMP2 induced a gradual increase in the p21 mRNA level over a period of 24 h, at which time it reached 2.5-fold higher than the basal level. Notably, knockdown of PPM1A in shPPM1A cells rendered cells more sensitive to BMP2, which induced a higher level of p21 mRNA (Fig. 4E). Similar results were obtained on BMP2-induced Id1 mRNA as higher and better induction of Id1 mRNA was observed in shPPM1A cells (Fig. 4F). Taken together, these results suggest that knockdown of PPM1A enhances Smad-dependent transcription responses.

DISCUSSION
It is generally accepted that phosphorylation in the SXS motif of R-Smads is the most critical intracellular event in the BMP/ TGF-␤ signal transduction pathways. Conversely, dephosphorylation of the SXS motif should provide a counter-mechanism terminating or antagonizing the functions of activated Smads. This study describes the identification and characterization of PPM1A as a phosphatase for BMP-activated Smad1. Our initial study on PPM1A-mediated dephosphorylation of Smad2/3 (16) led us to investigate whether PPM1A also acts as a Smad1 phosphatase. Considering the highly conserved sequences upstream of the distal SXS motif among all R-Smads, it is not surprising that all R-Smads are dephosphorylated by the same phosphatase. PPM1A physically interacts with Smad1 in both co-immunoprecipitation and GST pulldown assays. PPM1A dephosphorylates Smad1 in a number of cell lines tested, including 293T, HepG2, HaCaT, Mv1Lu, and C2C12 cells. Because of its activity toward the dephosphorylation of Smad1, overexpression of PPM1A inhibits BMP-induced transcriptional responses, whereas knockdown of its expression promotes BMP responses. In addition to our finding that Smad1 is dephosphorylated by PPM1A, we have found that PPM1A effectively dephosphorylates the phospho-SXS motif in Smad5 and Smad8.
Our finding that PPM1A is a phosphatase for R-Smads does not preclude the existence of other Smad phosphatases. We expect there is redundancy on Smad dephosphorylation. Recently, Chen et al. (15) has reported that PDP is a Mad phosphatase in Drosophila (15). Interestingly, mammalian PDPs (e.g. PDP1/PPM2C and PDP2) have no effect on Smad2 dephosphorylation (15,16), whereas PPM1A dephosphorylates all R-Smads (Ref. 16, this study, and data not shown). Unexpectedly, we failed to observe any effects of PDP1 or PDP2 overexpression on dephosphorylation of P-Smad1 (supplemental Fig. S2). Although our manuscript is in revision, Knockaert et al. (27) report dephosphorylation of Smad1 by small C-terminal domain phosphatases (SCPs) (27). Similar to the PDPs, overexpression of SCP1-3 has no effect on Smad1 dephosphorylation in mammalian cells, 4 although SCP1-3 clearly dephosphorylates the linker region of Smad2/3 (28). These discrepancies may be explained by the possibility that PDPs and/or SCPs may require additional cofactors or downstream effectors (which is limiting in mammalian cells) to have their full phosphatase activity toward dephosphorylation of the phospho-SXS motif of Smad1. Thus, in the absence of increased expression of these cofactors, increased expression of PDPs or SCPs does not suffice to achieve their effects. Further experiments are needed to clarify these issues, to identify new Smad phosphatases, and to characterize how these phosphatases specifically impact Smad-dependent responses.
Identification of PPM1A as a Smad1 phosphatase will aid in the understanding of the mechanisms underlying shut-off of BMP signaling. We anticipate that PPM1A might be involved in the regulation of many developmental processes such as skeletal and cardiovascular development as well as human diseases, partly through its role as a Smad phosphatase.