Retinoic acid regulates bone morphogenic protein signal duration by promoting the degradation of phosphorylated Smad1 Nengyin Shenga, Zhihui Xiea, Chen Wanga, Ge Baia, Kejing Zhanga, Qingqing Zhua, Jianguo Songa, Francois Guillemotb, Ye-Guang Chenc, Anning Lina,d, and Naihe Jinga,1 a Laboratory of Molecular Cell Biology, Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai 200031, China; bDivision of Molecular Neurobiology, National Institute for Medical Research, London NW7 1AA, United Kingdom; cState Key Laboratory of Biomembrane and Membrane Biotechnology, Department of Biological Sciences and Biotechnology, Tsinghua University, Beijing 100084, China; and dBen May Department for Cancer Research, University of Chicago, Chicago, IL 60637 Edited* by Melanie H. Cobb, University of Texas Southwestern Medical Center at Dallas, Dallas, TX, and approved September 20, 2010 (received for review June 28, 2010) The proper function of the bone morphogenic protein (BMP) pathway during embryonic development and organ maintenance requires its communication with other signaling pathways. Unlike the well-documented regulation of the BMP pathway by FGF/ MAPK and Wnt/GSK3 signals, cross-talk between BMP/Smad and retinoic acid (RA)/RA receptor (RAR) pathways is poorly understood. Here, we show that RA represses BMP signal duration by reducing the level of phosphorylated Smad1 (pSmad1). Through its nuclear receptor-mediated transcription, RA enhances the interaction between pSmad1 and its ubiquitin E3 ligases, thereby promoting pSmad1 ubiquitination and proteasomal degradation. This regulation depends on the RA-increased Gadd45 expression and MAPK activation. During the neural development in chicken embryo, the RA/RAR pathway also suppresses BMP signaling to antagonize BMP-regulated proliferation and differentiation of neural progenitor cells. Furthermore, this cross-talk between RA and BMP pathways is involved in the proper patterning of dorsal neural tube of chicken embryo. Our results reveal a mechanism by which RA suppresses BMP signaling through regulation of pSmad1 stability. B one morphogenic proteins (BMPs), a subfamily of cytokines of the TGF-β superfamily, play key roles in regulating a wide range of biological responses during embryonic development and adult tissue homeostasis (1). Deregulation of BMP signaling has been associated with developmental defects, carcinogenesis, and other diseases (2). BMP signal transduction involves the ligand binding to a type II receptor, which phosphorylates the type I receptor in a heterotetrameric receptor complex at the plasma membrane. This leads to the recruitment and phosphorylation of the cytoplasmic effectors Smad1/5/8, which are referred to as receptor-regulated Smads (R-Smads). Then, phosphorylated Smad1/5/8 together with Smad4, the common Smad protein (CoSmad), translocate to the nucleus where they regulate the expression of hundreds of target genes by interacting with transcriptional coactivators or corepressors (3–5). The pleiotropic regulatory functions of BMP signaling are finetuned extracellularly by antagonists (i.e., Chordin and Noggin) or intracellularly by Smad6 and Smad7, the inhibitory Smads (ISmads) (3, 6). In addition, the activity of the BMP pathway can be modulated by other signaling pathways either synergistically or antagonistically, depending on the biological context (7). Communications of the BMP signaling pathway with FGFs and other activators of the MAPK pathway rely on the phosphorylation of the R-Smads linker region (8, 9). The MAPK-mediated phosphorylation also primes the Smad1 linker region for GSK3 to phosphorylate, which further attenuates the Smad1 activity, whereas Wnt conversely prolongs the duration of BMP activity (10). BMP, FGF, and Wnt are important morphogens that set positional information in embryonic patterning. During early stages of Xenopus development, the dorsal-ventral (D-V) axis is established by a BMP gradient, and the anterior-posterior (A-P) morphogenetic gradient is dominated by Wnt signaling. The cross-talk between these two pathways suggests an interesting model where the graded signals 18886–18891 | PNAS | November 2, 2010 | vol. 107 | no. 44 that specify the D-V and A-P axes are integrated at the level of Smad1/5/8 to coordinate early embryonic patterning (11). However, it is less clear whether the BMP/Smad pathway is regulated by other morphogens during later phases of embryonic patterning. Retinoic acid (RA) is a universal morphogen that regulates multiple biological processes in embryonic development and adult tissue remodeling (12). When transported to the nucleus, RA binds to a transcription complex comprising heterodimers of RA receptor (RAR) and retinoic X receptor (RXR). These heterodimeric pairs then recruit a range of coactivators or corepressors to regulate gene transcription (13, 14). Moreover, RA signaling regulates many developmental processes through repression of protein growth factor signaling pathways (12), such as inhibition of FGF signaling during body axis extension (15). Along the D-V axis of the developing neural tube, BMP activity is restricted to the dorsal region, whereas RA signaling mainly resides in the intermediate region (13, 16). However, it is currently unknown whether and how RA pathway cross-talks with BMP signaling. In this study, we show that, through increasing Gadd45 expression and MAPK activation, RA enhances the interaction between pSmad1 and its ubiquitin E3 ligases. Consequently, RA promotes the ubiquitination and proteasomal degradation of pSmad1, thereby repressing BMP signaling activity. The crosstalk between the BMP and RA pathways is also involved in the patterning of the dorsal neural tube in chicken embryo. Results RA Reduces Smad1 Phosphorylation to Antagonize BMP/Smad Signaling. To address the possibility of RA regulating BMP signaling activity, we performed luciferase assays in mouse embryonic carcinoma P19 cells and mouse ES cells, in which both BMP and RA pathways are responsive. We used a four-repeat BMP responsive element (BRE), derived from the promoter of the human Id1 gene that is a canonical target of BMP signaling (17), to drive reporter expression. The BRE-luc response was significantly increased by BMP4 in both cell types, whereas RA treatment strongly inhibited BMP4-induced as well as basal BRE-luc activities (Fig. 1A, i and ii). To further confirm this negative regulation of BMP signaling, we used the natural promoter of human Id1 gene to drive luciferase expression and found that RA indeed prevented BMP4-induced activation of Id1-luc in P19 cells (Fig. 1A, iii). Consistently, real-time PCR analysis showed that basal and BMP4-induced expression of the downstream targets, Id1, Dlx5, and Msx2, was repressed by RA (Fig. 1B). In contrast, the expression of Smad1 Author contributions: N.S. and N.J. designed research; N.S. and Z.X. performed research; C.W., G.B., K.Z., Q.Z., and J.S. contributed new reagents/analytic tools; N.S., Y.-G.C., A.L., and N.J. analyzed data; and N.S., F.G., A.L., and N.J. wrote the paper. The authors declare no conflict of interest. *This Direct Submission article had a prearranged editor. 1 To whom correspondence should be addressed. E-mail: njing@sibs.ac.cn. This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1009244107/-/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.1009244107 RA Promotes the Ubiquitination and Proteasomal Degradation of pSmad1. In living cells, the level of pSmad1 is determined mainly by three factors: phosphorylation induced by the ligand and receptor complex, dephosphorylation by phosphatases, and degradation by the ubiquitin–proteasome system (18, 19). To determine which is responsible for the RA-induced decrease of Smad1 phosphorylation, serum-starved P19 cells were pretreated with RA for 4 or 5 h and then stimulated with BMP4 for another 1 or 2 h. We found that RA pretreatment did not repress BMP4-stimulated Smad1 phosphorylation (Fig. S1D), indicating that RA does not interfere with the kinase activity of the BMP ligand and receptor complex. To test whether RA activates phosphatases that dephosphorylate pSmad1 (20, 21), starved P19 cells were stimulated with BMP4 for 1 h, and after removal of BMP4, the cells were treated with or without RA for another 1 or 2 h. We found that pSmad1 levels decreased gradually after BMP4 was removed (Fig. S1E, lanes 2, 3, and 5), most likely owing to the endogenous phosphatases. However, RA did not facilitate the dephosphorylation of Smad1 (Fig. S1E, lanes 3–6), suggesting that the reduction of pSmad1 by RA does not involve the activation of R-Smad phosphatases. R-Smad stability is controlled by the ubiquitin–proteasomal machinery, which plays a critical role in the regulation of BMP and TGF-β pathways (19, 22). To examine the possibility that RAinduced pSmad1 reduction occurs through promotion of its degradation, we assessed whether the specific proteasome inhibitor MG-132 could restore pSmad1 levels after RA treatment. Immunoblotting analysis showed that the RA-induced pSmad1 reduction was indeed counteracted by MG-132, whereas total Smad1 levels were not affected (Fig. 1E, lanes 5 and 6). These data suggest that only a fraction of Smad1 is phosphorylated at the physiological BMP levels in P19 cells, in agreement with previous observations (10, 23), and that RA promotes pSmad1 degradation to reduce its level. Consistently, MG-132 also restored the expression of BMP target genes, Id1 and Dlx5, which was repressed by RA (Fig. S1F). Although okadaic acid, an inhibitor of type 1 and 2A protein phosphatases (24), blocked RA suppression of TGF-β–stimulated Smad2/3 phosphorylation in HL-60 cells (25), it was unable to rescue the RA-reduced pSmad1 level in P19 cells (Fig. 1E, lanes 8 and 9). Together, these results indicate that RA reduces Smad1 phosphorylation by promoting pSmad1 degradation. To determine whether the ubiquitination of pSmad1 is enhanced by RA, N2A cells were cotransfected with Flag–Smad1 and HA–ubiquitin. The ubiquitination of immunoprecipitated Smad1 was analyzed using anti-HA antibody. We found that BMP4 alone caused Smad1 ubiquitination (Fig. S1G, lanes 2 and 3) as previously Fig. 1. RA promotes pSmad1 degradation to antagonize BMP signaling. (A) Reporter activities of BRE-luc (i) and Id1-luc (iii) in P19 cells and BRE-luc (ii) in ES cells were measured after treatment with BMP4 and/or RA. The results are presented as mean ± SD (*P < 0.05). (B) Quantitative real-time PCR analyses of relative expression of Id1, Dlx5, and Msx2 to GAPDH in P19 cells that were stimulated with BMP4 and/or RA for 12 h before harvesting. The results are presented as mean ± SD (*P < 0.05). (C) ChIP assay of the pSmad1 binding on Id1 promoter in P19 cells incubated with the indicated additions for 12 h. Antibody against pSmad1/5/8 was used, and the immunoprecipitates were analyzed by real-time PCR for enrichment of the BRE1 and 2 and the region 1 kb upstream (Control). The data were normalized to the intact inputs respectively and represented as one of three independent experiments. The results are presented as mean ± SD (*P < 0.05). (D) P19 cells were treated as in B, except that Western blot was performed to examine the pSmad1 and total Smad1 levels in whole-cell lysates. (E) BMP4- and/or RA-treated P19 cells were simultaneously added with mock, MG132, or okadaic acid for 6 h and then harvested for immunoblotting of indicated antibodies. (F) P19 cells were treated with indicated additions for 6 h, lysed with SDS denaturing buffer, and boiled for 1 min. The immunoprecipitates with anti-Smad1 antibody and the whole-cell lysates were subjected to Western blot analyses with indicated antibodies. Sheng et al. PNAS | November 2, 2010 | vol. 107 | no. 44 | 18887 CELL BIOLOGY and Smad4, two mediators but not target genes of BMP signaling, remained constant in control and RA-treated cells (Fig. S1A). These results suggest that the RA suppression of BMP-regulated transcription is specific and not the result of cytotoxicity or general transcriptional repression. Because pSmad1 is the critical effector that controls transcriptional responses of BMP signaling, we reasoned that RA might inhibit BMP transcriptional activity by recruiting regulatory repressor(s) or preventing the pSmad1 transcriptional complex from binding to its target promoters. To distinguish between these two possibilities, a ChIP assay was carried out to examine whether RA interferes with BMP4-induced pSmad1 binding to the BRE region of Id1 promoter. Real-time PCR analysis indicated that BMP4 stimulation increased the enrichment of this BRE fragment in anti-pSmad1 immunoprecipitates and that RA treatment blocked this increase. In the control, no significant change was observed for the enrichment of a region upstream of the BRE after BMP4 and/ or RA treatment (Fig. 1C). These data suggest that RA interferes with the binding of pSmad1 to its cognate sequences. We then examined pSmad1 levels in starved P19 cells treated with BMP4 and/or RA. We found that RA reduced BMP4-activated Smad1 phosphorylation but did not affect the overall amount of Smad1 in whole-cell lysates (Fig. 1D, lanes 2 and 4). Cell fractionation assays showed that treatment with RA resulted in decrease of pSmad1 levels in both cytoplasm and nuclear fractions (Fig. S1B, lanes 2, 4, 6, and 8). To determine the kinetics of RA reduction of Smad1 phoshorylation, we treated P19 cells with BMP4 alone or BMP4 plus RA for various periods of time. There was no detectable difference in pSmad1 levels between the two treatments at 2 h poststimulation (Fig. S1C, lanes 2 and 3). However, there was a reduction of Smad1 phosphorylation when cells were treated with BMP4 plus RA for 4 h (Fig. S1C, lanes 4 and 5), and this reduction was more pronounced for 6-h treatment (Fig. S1C, lanes 6 and 7). These data suggest that RA decreases pSmad1 level to repress BMP signaling, and this repression requires long-term RA exposure. reported (26), and this effect was further facilitated by RA (Fig. S1G, lanes 3 and 4). When the immunoprecipitates were blotted by anti-Flag antibody, it was still found that Smad1 was significantly ubiquitinated by the cotreatment of BMP4 and RA (Fig. S1H). Moreover, the ubiquitination of endogenous Smad1 in P19 cells was increased by RA only in the presence of BMP4 (Fig. 1F). Taken together, these results suggest that RA reduces pSmad1 level by promoting its ubiquitination and proteasomal degradation. RA Enhances the Interaction Between pSmad1 and E3 Ubiquitin Ligases to Promote pSmad1 Ubiquitination and Degradation. The specification of recognition and ubiquitination of target proteins is dependent on E3 ubiquitin ligases, and the known Smad1-specific E3 ligases are Smurf1, Smurf2, and ChIP (22). In P19 cells, the expression of these E3 ubiquitin ligases was not increased by RA in the presence or absence of BMP4 (Fig. S2A), indicating that RA promotion of pSmad1 degradation is not through up-regulating these ligases. Next, we investigated whether depletion of these ligases could release the RA inhibition of BMP signaling. Because the expression of Smurf2 and CHIP is much higher than that of Smurf1 in P19 and N2A cells (Fig. S2B), we generated RNAi constructs to knockdown Smurf2 and CHIP expression simultaneously (Fig. S2 C and D). We found that their depletion blocked RA-reduced Smad1 phosphorylation (Fig. 2A, lanes 5 and 6). In accordance, knockdown of Smurf2 and CHIP decreased pSmad1 ubiquitination in N2A cells (Fig. 2B, lanes 1 and 3), and the RAinduced ubiquitination of pSmad1 was also inhibited (Fig. 2B, lanes 3 and 4). Furthermore, the dominant negative mutants of Smurf2 and CHIP also inhibited RA-enhanced pSmad1 ubiquitination (Fig. S2E). These results indicate that the R-Smad–specific E3 ubiquitin ligases are necessary for RA-induced ubiquitination and proteasomal degradation of pSmad1. The above results support the notion that RA interferes with BMP signaling through regulation of pSmad1 ubiquitination and stability. Next, we sought to examine whether RA enhances the interaction between the E3 ligases and pSmad1. Coimmunoprecipitation (Co-IP) assays in BMP4- and/or RA-treated N2A cells were used to analyze the association of Smad1 with mutated Smurf2(C716A) (27), which lacks ligase activity but retains its ability to interact with Smad1. We found that Smurf2(C716A) was present in the Smad1 immunocomplex (Fig. 2C, lane 3), and its amount was slightly increased on BMP4 treatment (Fig. 2C, lane 5), consistent with the observation that Smad1 ubiquitination was enhanced by BMP4 (Fig. S1G). Notably, in the presence of BMP4, RA strongly promoted the binding of Smurf2(C716A) to Smad1 (Fig. 2C, lanes 5 and 6), whereas RA treatment alone had no apparent effect (Fig. 2C, lanes 3 and 4). Moreover, proline to alanine mutations of the PPAY motif [Flag–Smad1(AAAY)], which abrogates the binding of Smad1 to E3 ubiquitin ligases (8, 28), prevented the RA-increased association of Smurf2(C716A) with Smad1 (Fig. 2D), suggesting that this RA regulation is indeed dependent on the recognition of Smad1 by its E3 ubiquitin ligases. To verify that RA preferentially increases Smurf2 interaction with pSmad1, we precipitated the Smurf2(C716A) complex. It was found that the pSmad1 only existed in the immunocomplex on BMP4 stimulation and that RA increased this pSmad1 binding significantly (Fig. S2F). Together, these data indicate that RA increases the interaction between pSmad1 and its E3 ubiquitin ligases to promote pSmad1 ubiquitination and degradation. RA Increases Gadd45 Expression and MAPK Activation to Promote pSmad1 Degradation. To determine whether RA nuclear receptors are involved in the RA regulation of BMP signaling, we first examined their expression in P19 and N2A cells. Because RARα and RARγ were the main RA receptors expressed in these cells (Fig. S3A), we used RNAi plasmids against RARα and RARγ to knockdown expression of both receptors (Fig. S3 B and D). Immunoblotting analysis showed that RA could not reduce pSmad1 levels after RARα and RARγ were depleted (Fig. 3A, lanes 5 and 6), suggesting that RA receptors are required for RA-promoted pSmad1 degradation. Consistently, Co-IP assays revealed that the reduction of RA receptors expression also blocked the 18888 | www.pnas.org/cgi/doi/10.1073/pnas.1009244107 Fig. 2. RA enhances the interaction between pSmad1 and E3 ubiquitin ligases to promote pSmad1 ubiquitination and degradation. (A) P19 cells were transfected with psuper vector or RNAi plasmids of Smurf2 and CHIP, and Western blot analyses of pSmad1 and total Smad1 in nuclear were performed after the cells were treated with indicated additions. (B) The Flag-Smad1– and HA-Ub–transfected N2A cells were coexpressed with psuper vector or RNAi plasmids of Smurf2 and CHIP and then treated with indicated additions. The Flag immunoprecipitates and whole-cell lysates were subjected to Western blot analyses with indicated antibodies. (C) N2A cells were transfected with expression plasmids of Flag-Smurf2(C716A) and HA-Smad1 and then treated with BMP and/or RA as indicated. Western blot was performed to analyze the levels of Smad1 and Smurf2 in the HA immunoprecipitates (IP) and whole-cell lysate (Input) with indicated antibodies. (D) Cells were transfected as in C, except that expression plasmids encoding Flag-Smad1(wild type) or FlagSmad1(AAAY) and Myc-Smurf2(C716A) were used for transfection as indicated and anti-Flag antibody was applied for immunoprecipitation. RA-enhanced interaction between pSmad1 and Smurf2(C716A) (Fig. 3B). Because RA usually forms a complex with RARs for transcriptional regulation (14) and our previous observation (Fig. S1C) indicated that RA-mediated reduction of Smad1 phosphorylation needed a 4- to 6-h treatment, we proposed that RARmediated de novo protein synthesis might be required for RA inhibition of BMP signaling. To confirm this notion, we used cycloheximide and actinomycin D to inhibit translation and transcription, respectively, and found that both inhibitors could rescue RA-reduced pSmad1 levels (Fig. 3C). More interestingly, we found that the MEK inhibitor PD184352 (Fig. 3D, lanes 5 and 6) but not the FGF receptor inhibitor PD173074 (Fig. 3D, lanes 8 and 9) could fully restore the RA reduction of Smad1 phosphorylation, suggesting that MAPK activation is required for RA-promoted pSmad1 degradation. Thus, it is possible that RA may regulate the expression of some MAPK regulator(s) to exert its inhibitory function. To test this possibility, P19 cells were treated with BMP4 or BMP4 plus RA for 6 h, and a microarray analysis was performed to search for differentially expressed genes, among which Gadd45β and γ were upregulated by RA treatment. Gadd45 members are MAPK cascade activators through regulating MEKK4 activity (29). Real-time PCR analysis confirmed that RA significantly increased Gadd45α, β, and γ expression, both in the absence and presence of BMP4 (Fig. 3E). To determine the function of Gadd45 in RA-regulated Smad1 phosphorylation, we generated RNAi constructs to knockdown Gadd45β and γ simultaneously (Fig. S3 C and E) and found that the depletion of Gadd45β and γ rescued the RA-repressed pSmad1 level (Fig. 3F, lanes 5 and 6) and disrupted the RAenhanced association of pSmad1 and Smurf2(C716A) (Fig. 3G, lanes 3 and 4). Together, these results indicate that RA increases Gadd45 expression and MAPK activation to enhance the interaction of pSmad1 with its E3 ubiquitin ligases, thereby promoting pSmad1 degradation. Sheng et al. portant morphogens, and their signaling activities must be finely regulated for proper development of the CNS (16, 30). To determine whether RA regulates BMP signaling during CNS development, we electroporated the reporter plasmid BRE-TK-RFP, in which the expression of RFP was controlled by four repeats of the BRE, into the neural tube of chicken embryos at Hamburger and Hamilton (HH) stages 10 and 11 (31). RFP expression was analyzed at HH stages 16 and 17 to visualize BMP signaling activity. RFP-positive cells were mainly located in the dorsal region of the developing neural tube, with a gradual decrease along the D-V axis (Fig. 4A, a), which is consistent with the endogenous gradient of BMP activity. Coelectroporation of RARγ strongly reduced the number and intensity of RFP-positive cells (Fig. 4 A, c and B). The reporter activity of BRE-luc was repressed by overexpressed RARγ but increased by RAR403, a dominant negative form of RAR (32) (Fig. 4C). Real-time PCR analysis also showed that RARγ inhibited the endogenous expression of the BMP target genes cId2, cId3, cMsx1, and cMsx2 (Fig. 4D). Meanwhile, expression of chick Gadd45α and β was increased by RARγ overexpression (Fig. S3F). Taken together, these results indicate that the RA/RAR pathway also negatively regulates BMP signaling activity in the neural tube of chicken embryo, and the underlying mechanism may be similar to that of in vitro models. During CNS development, activation of BMP signaling through the BMP receptor BMPRIA promotes the proliferation of neural progenitor cells and prevents precocious neurogenesis (33). To determine whether this BMP signaling activity is modulated by the RA/RAR pathway, we electroporated a construct of constitutively active BMPRIA (caBMPRIA), either alone or together with RARγ, into chick neural tubes at HH stages 10 and 11. Then, an immunostaining analysis of Tuj1, a neuron-specific class III βtubulin, was used to measure the number of mature neurons at HH stages 16 and 17. Overexpression of caBMPRIA significantly decreased the ratio of Tuj1+ cells between the transfected and control sides (Fig. 4 E, b and F), whereas control electroporation of a GFP plasmid had no effect (Fig. 4E, a). Notably, coelectroporation of RARγ with caBMPRIA could partially restore the number of Tuj1+ cells (Fig. 4 E, c and F). Progenitor cells in the embryonic neural tube are actively proliferating and can be labeled with BrdU. Thus, we asked whether the RA/RAR pathway could also interfere with the BMP regulation of neural progenitor proliferation. To this end, we injected the electroporated embryos with BrdU and incubated them for 1 h before harvesting at HH stages 22 and 23. Immunostaining analysis showed that caBMPRIA electroporation significantly increased the ratio of BrdU+ cells between the transfected and control sides (Fig. 4 E, e and F), whereas RARγ coelectroporation inhibited this effect (Fig. 4 E, f and F). Together, these data suggest that RA/RAR pathway can antagonize the BMP-regulated differentiation and proliferation of neural progenitor cells in chick neural tube. Integration of BMP/Smad and RA/RAR Pathways During Patterning of the Dorsal Neural Tube in Chicken Embryo. The specification of the distinct domains in the dorsal spinal cord is initiated by the proneural genes between embryonic day (E) 9.5 and E12.5 in the mouse embryo. The dorsal-most precursor cells adjacent to the roof plate (dI1) express the proneural gene Math1, whereas neurogenin 1 (Ngn1) and Ngn2 are expressed by the next ventral band of precursor cells (dI2) and Mash1 is expressed by the precursors that will become dI3–dI5 (34). BMPs secreted by the roof plate are required for the patterning of these dorsal domains (35). To examine whether the patterning activity of BMP signaling is regulated by the RA/RAR pathway, we electroporated caBMPRIA, either alone or together with RARγ, into chick neural tubes at HH stages 12–14 and examined the expression patterns of the proneural genes cNgn1, cNgn2, and Cash1 (the chick ortholog of Mash1) at HH stages 22 and 23. In the dorsal neural tube of caBMPRIA-electroporated embryos, the expression of Ngn1 (6/9), Ngn2 (6/9), and Cash1 (3/9) was clearly ventralized on the electroporated side compared with the control side (Fig. 5 Center), whereas electroporation of a control GFP construct had no effect (7/7) (Fig. 5 Left). This ventral migration of the dI2 and dI3 domains is consistent with the dorsal spinal cord phenotype of the BMPRIA and BMPRIB double-knockout mice (36). Interestingly, coelectroporation of caBMPRIA and RARγ rescued the ventralization phenotype of caBMPRIA, resulting in a localization of cNgn1 (8/9), cNgn2 (8/9), and Cash1 (9/9) transcripts similar to that of the nonelectroporated side of the neural tube (Fig. 5 Right). These results suggest that the RA/RAR pathway regulates BMP patterning activity in the dorsal neural tube. Collectively, these results indicate that the BMP/Smad and Fig. 3. RA increases Gadd45 expression and consequent MAPK activation to promote pSmad1 degradation. (A) Transfection was performed as in Fig. 2A, except that P19 cells were transfected with RARα and γ RNAi plasmids. (B) The HASmad1– and Flag-Smurf2(C716A)–transfected N2A cells were coexpressed with psuper vector or RNAi plasmids of RARα and γ and treated with BMP and/or RA as indicated. The coimmunoprecipitation and Western blot were performed as in Fig. 2C. (C ) BMP4- and/or RA-treated P19 cells were simultaneously added with mock, cyclohexmide, or actinomycin D for 6 h and then harvested for immunoblotting with indicated antibodies. (D) BMP4- and/or RA-treated P19 cells were simultaneously added with mock, PD184352, or PD173074 for 6 h and then harvested for immunoblotting with indicated antibodies. (E ) P19 cells were treated as in Fig. 1B, and the relative expression levels of Gadd45α, β, and γ were analyzed by real-time PCR. (F) Transfection was performed as in Fig. 2A, except that P19 cells were transfected with Gadd45β and γ RNAi plasmids. (G) Transfection was performed as in B, except that N2A cells were cotransfected with Gadd45β and γ RNAi plasmids and immunoprecipitation was used with the anti-HA antibody. Sheng et al. PNAS | November 2, 2010 | vol. 107 | no. 44 | 18889 CELL BIOLOGY RA/RAR Pathway Represses BMP Activity and Antagonizes BMP Regulation of the Proliferation and Differentiation of Neural Progenitor Cells in Chick Neural Tube. Both BMP and RA are im- that, by interfering with the stability of pSmad1, the RA/RAR pathway functions as a biological regulator to balance the duration of BMP/Smad signal. Smad1 is the main effector of BMP signaling and is regarded as the central component for integrated communication between BMP and other pathways (11). Previous studies have shown that the FGF/MAPK and Wnt/GSK3 pathways cross-talk with BMP/ Smad signaling by regulating the stability of Smad1. The dual inhibitory phosphorylations in the Smad1 linker region by MAPK and GSK3 are initiated by BMP receptor-activated phosphorylation, and these sequential modifications are important for balancing the duration of BMP pathway activity (8, 10). We found that the regulation of BMP signaling by the RA/RAR pathway also converges on Smad1, and RA specifically enhances the interaction between pSmad1 and E3 ubiquitin ligases, thereby promoting the ubiquitination and proteasomal degradation of pSmad1. RA treatment does not interfere with total Smad1 levels and does not increase the association of Smad1 and E3 ligases in the absence of BMP4. Therefore, RA-regulated Smad1 modification and stability depend on the activation of the BMP pathway and the phosphorylation of the Smad1 C terminal as triggered by BMP receptors primes Smad1 for RA modulation. This suggests that the RA/RAR pathway also regulates the duration of the BMP/Smad signal. The cross-talk between the BMP/Smad and RA/RAR pathways may provide an additional mechanism for regulating the intensity and duration of BMP/Smad signaling and the differential expression of target genes. The ubiquitin–proteasomal system plays a major role in regulating the transduction of the BMP pathway, and this regulation occurs at multiple levels, including the BMP receptors and Fig. 4. RA/RAR pathway represses BMP activity and antagonizes the BMP regulation of differentiation and proliferation of neural progenitor cells in chick neural tube. (A) At HH stages 10 and 11, the BRE-TK-RFP reporter was coelectroporated with GFP (a and b) or RARγ-IRES-GFP (c and d) into chick neural tubes, and transverse sections of HH stages 16 and 17 were analyzed about the expression of GFP and RFP. (Scale bar: 100 μm.) (B) Percentage of RFP+GFP+ cells among GFP+ cells in GFP (n = 8) and RARγ (n = 9) electroporated neural tubes. Data are presented as the mean ± SD (*P < 0.05). (C) Experiment was performed as in A, except that the expression plasmids were coelectroporated with BRE-luc, and the electroporated part of GFP- (n = 9), RARγ-IRES-GFP– (n = 9), or RAR403-IRES-GFP– (n = 6) expressed neural tubes was dissected for luciferase assay. (D) Experiment was performed as in A, except that the expression plasmids were electroporated alone, and the electroporated side of GFP (n = 8) or RARγ-IRES-GFP (n = 10) neural tubes was dissected for mRNA harvest. Real-time PCR was applied to analyze the relative expression levels of cId2, cId3, cMsx1, and cMsx2 to S17. (E) The right side of chick neural tubes was electroporated with indicated plasmids. Transverse sections of HH stages 16 and 17 were immunostained with antiTuj1 antibody (a–c; GFP: n = 9; caBMPRIA: n = 7; caBMPRIA plus RARγ: n = 8). For BrdU labeling, electroporated embryos were injected with BrdU and incubated for 1 h; then, they were harvested at HH stages 22 and 23. Transverse sections were immunostained with anti-BrdU antibody (d–f; GFP: n = 10; caBMPRIA: n = 11; caBMPRIA plus RARγ: n = 12). (Scale bar: 100 μm.) (F) The ratio of Tuj1+ and BrdU+ cells between the transfected side (indicated by +) and the control side (indicated by −) of the electroporated neural tubes in E. Data are presented as the mean ± SD (*P < 0.05). RA/RAR pathways are integrated for proper specification of neural progenitor cells in the developing spinal cord. Discussion Because deregulation of BMP signaling leads to many developmental disorders and diseases, the stringent control of its activation and termination is critical for normal development and tissue maintenance (1, 2). In this study, we show that RA represses the BMP signaling during proliferation and differentiation of neural progenitor cells and patterning of the dorsal neural tube in chicken embryo. Through increasing Gadd45 expression and MAPK activation, RA enhances the association of pSmad1 with its E3 ubiquitin ligases to promote pSmad1 ubiquitination and proteasomal degradation, thereby antagonizing BMP signaling activity. Our findings support the notion 18890 | www.pnas.org/cgi/doi/10.1073/pnas.1009244107 Fig. 5. Integration of BMP/Smad and RA/RAR pathways during patterning of chick dorsal spinal cord. Experiments were performed as in Fig. 4E, except that the electroporation was at HH stages 12–14, and embryos were harvested at HH stages 22 and 23. In situ hybridization was performed with probes of chick Ngn1 (a, d, and g), Ngn2 (b, e, and h), and Cash1 (c, f, and i). (Scale bar: 100 μm.) Sheng et al. progenitor domains of the dorsal neural tube. For proper D-V patterning of the neural tube, BMP signaling activity must be graded in the dorsal region, and the duration of the signal must also be precisely regulated (16, 36). Cross-talk between the BMP/ Smad and RA/RAR pathways plays an important role in this fine tuning, thus suggesting that positional information along the D-V and A-P axes is integrated to ensure correct neural tube patterning during CNS development. R-Smads (19). The ubiquitination of Smad1 is regulated by specific E3 ubiquitin ligases including Smurf1, Smurf2, and ChIP in both ligand-dependent and -independent manners (22). It was previously reported that MAPK and GSK3 could phosphorylate the Smad1 linker region, enabling Smad1 binding to Smurf1 and thus, leading to Smad1 ubiquitination and attenuation of BMP signaling (8, 11). Here, we show that, through enhancing the interaction of pSmad1 with Smurf2, RA/RAR pathway reduces pSmad1 stability to repress BMP signal activity, suggesting that different E3 ligases can mediate the cross-talk of BMP signaling with different pathways. We further find that the RA-induced pSmad1 degradation requires MAPK activation and RA-increased expression of the MAPK activator Gadd45. Therefore, different from the FGF/ MAPK and Wnt/GSK3 pathways, RA/RAR regulates BMP/Smad signaling through a mechanism that requires up-regulation of Gadd45 transcription to activate the MAPK cascade. BMP and RA are important morphogens, and both are involved in many developmental processes, including CNS development (16, 30), cardiogenesis, and limb induction (1, 12). When these two pathways participate in the same process, it is reasonable to speculate that they may regulate each other. Neural patterning is a crucial process that shapes the functional organization of the CNS during development. RA organizes the A-P patterning of the neural tube through its posteriorizing activity (30), and BMP activity plays an important role in the D-V patterning (16, 34). Additionally, BMP activity is restricted to the dorsal part of the neural tube along the D-V axis, whereas RA signaling mainly resides in the intermediate region (13, 16). Therefore, these signaling activities must be precisely regulated for proper development of the neural tube. We found that both BMP signaling activity and its inhibitory effect on neuronal differentiation are inhibited by the RA/RAR pathway in the developing spinal cord. Moreover, RA/ RAR signaling also regulates the BMP signal to specify the neural Immunoprecipitation and Western Blot. Coimmunoprecipitation experiments were performed as described previously (37). Western blot assays were performed as described previously (38) with slightly modification. The detailed procedures and antibodies used are included in SI Materials and Methods. 1. Wu MY, Hill CS (2009) Tgf-beta superfamily signaling in embryonic development and homeostasis. Dev Cell 16:329–343. 2. Massagué J (2008) TGFbeta in cancer. Cell 134:215–230. 3. Massagué J, Seoane J, Wotton D (2005) Smad transcription factors. Genes Dev 19: 2783–2810. 4. Feng XH, Derynck R (2005) Specificity and versatility in tgf-beta signaling through Smads. Annu Rev Cell Dev Biol 21:659–693. 5. Shi Y, Massagué J (2003) Mechanisms of TGF-beta signaling from cell membrane to the nucleus. Cell 113:685–700. 6. Massagué J, Chen YG (2000) Controlling TGF-beta signaling. Genes Dev 14:627–644. 7. Guo X, Wang XF (2009) Signaling cross-talk between TGF-beta/BMP and other pathways. Cell Res 19:71–88. 8. Sapkota G, Alarcón C, Spagnoli FM, Brivanlou AH, Massagué J (2007) Balancing BMP signaling through integrated inputs into the Smad1 linker. Mol Cell 25:441–454. 9. Kretzschmar M, Doody J, Massagué J (1997) Opposing BMP and EGF signalling pathways converge on the TGF-beta family mediator Smad1. Nature 389:618–622. 10. Fuentealba LC, et al. (2007) Integrating patterning signals: Wnt/GSK3 regulates the duration of the BMP/Smad1 signal. Cell 131:980–993. 11. Eivers E, Fuentealba LC, De Robertis EM (2008) Integrating positional information at the level of Smad1/5/8. Curr Opin Genet Dev 18:304–310. 12. Duester G (2008) Retinoic acid synthesis and signaling during early organogenesis. Cell 134:921–931. 13. Maden M (2007) Retinoic acid in the development, regeneration and maintenance of the nervous system. Nat Rev Neurosci 8:755–765. 14. Wei LN (2003) Retinoid receptors and their coregulators. Annu Rev Pharmacol Toxicol 43:47–72. 15. Diez del Corral R, et al. (2003) Opposing FGF and retinoid pathways control ventral neural pattern, neuronal differentiation, and segmentation during body axis extension. Neuron 40:65–79. 16. Liu A, Niswander LA (2005) Bone morphogenetic protein signalling and vertebrate nervous system development. Nat Rev Neurosci 6:945–954. 17. Katagiri T, et al. (2002) Identification of a BMP-responsive element in Id1, the gene for inhibition of myogenesis. Genes Cells 7:949–960. 18. Wrighton KH, Lin X, Feng XH (2009) Phospho-control of TGF-beta superfamily signaling. Cell Res 19:8–20. 19. Lönn P, Morén A, Raja E, Dahl M, Moustakas A (2009) Regulating the stability of TGFbeta receptors and Smads. Cell Res 19:21–35. 20. Knockaert M, Sapkota G, Alarcón C, Massagué J, Brivanlou AH (2006) Unique players in the BMP pathway: Small C-terminal domain phosphatases dephosphorylate Smad1 to attenuate BMP signaling. Proc Natl Acad Sci USA 103:11940–11945. 21. Duan X, Liang YY, Feng XH, Lin X (2006) Protein serine/threonine phosphatase PPM1A dephosphorylates Smad1 in the bone morphogenetic protein signaling pathway. J Biol Chem 281:36526–36532. 22. Inoue Y, Imamura T (2008) Regulation of TGF-beta family signaling by E3 ubiquitin ligases. Cancer Sci 99:2107–2112. 23. Alarcón C, et al. (2009) Nuclear CDKs drive Smad transcriptional activation and turnover in BMP and TGF-beta pathways. Cell 139:757–769. 24. Holmes CF, Luu HA, Carrier F, Schmitz FJ (1990) Inhibition of protein phosphatases-1 and -2A with acanthifolicin. Comparison with diarrhetic shellfish toxins and identification of a region on okadaic acid important for phosphatase inhibition. FEBS Lett 270:216–218. 25. Cao Z, et al. (2003) Levels of phospho-Smad2/3 are sensors of the interplay between effects of TGF-beta and retinoic acid on monocytic and granulocytic differentiation of HL-60 cells. Blood 101:498–507. 26. Gruendler C, Lin Y, Farley J, Wang T (2001) Proteasomal degradation of Smad1 induced by bone morphogenetic proteins. J Biol Chem 276:46533–46543. 27. Lin X, Liang M, Feng XH (2000) Smurf2 is a ubiquitin E3 ligase mediating proteasomedependent degradation of Smad2 in transforming growth factor-beta signaling. J Biol Chem 275:36818–36822. 28. Ingham RJ, Gish G, Pawson T (2004) The Nedd4 family of E3 ubiquitin ligases: Functional diversity within a common modular architecture. Oncogene 23:1972–1984. 29. Takekawa M, Saito H (1998) A family of stress-inducible GADD45-like proteins mediate activation of the stress-responsive MTK1/MEKK4 MAPKKK. Cell 95:521–530. 30. Maden M (2002) Retinoid signalling in the development of the central nervous system. Nat Rev Neurosci 3:843–853. 31. Hamburger V, Hamilton HL (1992) A series of normal stages in the development of the chick embryo. 1951. Dev Dyn 195:231–272. 32. Novitch BG, Wichterle H, Jessell TM, Sockanathan S (2003) A requirement for retinoic acid-mediated transcriptional activation in ventral neural patterning and motor neuron specification. Neuron 40:81–95. 33. Panchision DM, et al. (2001) Sequential actions of BMP receptors control neural precursor cell production and fate. Genes Dev 15:2094–2110. 34. Caspary T, Anderson KV (2003) Patterning cell types in the dorsal spinal cord: What the mouse mutants say. Nat Rev Neurosci 4:289–297. 35. Zhuang B, Sockanathan S (2006) Dorsal-ventral patterning: A view from the top. Curr Opin Neurobiol 16:20–24. 36. Wine-Lee L, et al. (2004) Signaling through BMP type 1 receptors is required for development of interneuron cell types in the dorsal spinal cord. Development 131:5393–5403. 37. Bai G, et al. (2007) Id sustains Hes1 expression to inhibit precocious neurogenesis by releasing negative autoregulation of Hes1. Dev Cell 13:283–297. 38. Shen C, et al. (2008) Hydrogen peroxide promotes Abeta production through JNKdependent activation of gamma-secretase. J Biol Chem 283:17721–17730. Sheng et al. Materials and Methods Cell Culture and Plasmids Construction. The methods of cell culture and details for constructs in this study are described in SI Materials and Methods. PNAS | November 2, 2010 | vol. 107 | no. 44 | 18891 CELL BIOLOGY ACKNOWLEDGMENTS. We thank Drs. Xinhua Feng (Baylor College of Medicine, Houston, TX), Zhijie Chang (Tsinghua University, Beijing, China), Pierre Chambon (Institute for Genetics and Cellular and Molecular Biology, Illkirch, France), Shanthini Sockanathan (Johns Hopkins School of Medicine, Baltimore, MD), Takenobu Katagiri (Saitama Medical School, Saitama, Japan), and Yuelei Chen (Institute of Biochemistry and Cell Biology, Shanghai, China) for providing crucial plasmids. We thank the members of the N.J. laboratory for discussion of the manuscript. This work was supported in part by the National Natural Science Foundation of China (Grants 30623003, 30721065, and 30830034), the National Key Basic Research and Development Program of China (Grants 2005CB522704, 2006CB943902, 2007CB947101, 2008KR0695, and 2009CB941100), the Shanghai Key Project of Basic Science Research (Grants 06DJ14001, 06DZ22032, and 08DJ1400501), and the Council of Shanghai Municipal Government for Science and Technology (Grant 088014199).