Smad1 and Smad4 Are Components of the Bone Morphogenetic Protein-4 (BMP-4)-induced Transcription Complex of the Xvent-2B Promoter*

To study the mechanism of transcriptional activation of the Xenopus homeobox geneXvent-2B, we have delineated the bone morphogenetic protein-4 (BMP-4)-responsive region between −275/−152 in the proximal promoter. Consistent with the BMP-4 inductive nature of this region, this element exhibits transcriptional activation upon ectopic expression of Smad1 and Smad4. Electrophoretic mobility shift assays with total cellular extracts demonstrated that a DNA fragment encompassing this region was competent in the formation of a BMP-4-induced protein-DNA complex containing Smad1. Two different Smad binding regions were localized, a distal binding region for Smad1 containing two GCAT motifs and proximal AGNC binding sites for Smad4, the latter being conserved in other transforming growth factor-β-responsive elements. Mutation of the Smad4 binding motif completely abolished transcriptional activation, whereas mutation or deletion of the Smad1 recognition sequence inhibited Smad1/Smad4 responsiveness. These results provide a functional characterization and identification of a vertebrate Smad1/Smad4 DNA response element induced by BMP-4 signaling and offers insight into the transcriptional regulation of a component essential for dorsoventral patterning inXenopus embryos.

To study the mechanism of transcriptional activation of the Xenopus homeobox gene Xvent-2B, we have delineated the bone morphogenetic protein-4 (BMP-4)-responsive region between ؊275/؊152 in the proximal promoter. Consistent with the BMP-4 inductive nature of this region, this element exhibits transcriptional activation upon ectopic expression of Smad1 and Smad4. Electrophoretic mobility shift assays with total cellular extracts demonstrated that a DNA fragment encompassing this region was competent in the formation of a BMP-4induced protein-DNA complex containing Smad1. Two different Smad binding regions were localized, a distal binding region for Smad1 containing two GCAT motifs and proximal AGNC binding sites for Smad4, the latter being conserved in other transforming growth factor-␤responsive elements. Mutation of the Smad4 binding motif completely abolished transcriptional activation, whereas mutation or deletion of the Smad1 recognition sequence inhibited Smad1/Smad4 responsiveness. These results provide a functional characterization and identification of a vertebrate Smad1/Smad4 DNA response element induced by BMP-4 signaling and offers insight into the transcriptional regulation of a component essential for dorsoventral patterning in Xenopus embryos.
The transforming growth factor-␤ (TGF-␤) 1 superfamily is a conserved class of cell-cell-signaling molecules implicated in the regulation of cellular proliferation and differentiation. Throughout induction and patterning of the amphibian mesoderm, multiple TGF-␤-signaling molecules are present, including bone morphogenetic proteins (BMP), which have been shown to promote the formation and patterning of ventral mesoderm (1).
Intracellularly, BMP-4 signals are transduced to the nucleus from heteromeric-coupled serine-threonine kinase receptors by an evolutionary-conserved family of Smad proteins (2,3). Most members can be classified into three groups based on both sequence homology and biological activity: receptor specific Smads, common mediator Smads, and inhibitory Smads. All members consist of two highly conserved domains in the amino and carboxyl terminus, denoted the MH1 and MH2 domains, respectively, separated by a proline-rich region of variable sequence and length.
Ligand binding to TGF-␤ receptors stimulates the association and phosphorylation of the receptor-specific Smad on carboxyl-terminal serine residues in the carboxyl-terminal SSXS motif. Smad1, Smad5, and most likely Smad8 are substrates for BMP receptors, whereas Smad2 and Smad3 are mediators of TGF-␤ and activin/Vg1 signaling. Functional differences between the receptor-specific Smads are best observed in Xenopus embryos, where ectopic expression of Smad1 and Smad5 mimic BMP-4 induction of ventral mesoderm (4,5), and Smad2 induces dorsalization of mesoderm and formation of a secondary axis in a fashion similar to activin A (6,7). Following phosphorylation, the receptor-specific Smads hetero-oligomerize with Smad4 and translocate to the nucleus, where they regulate the expression of specific genes.
Accumulating evidence suggests that the Smads directly participate in transcriptional complexes that elicit the effects of their corresponding ligands. First, the Smad proteins have an intrinsic transcriptional activation domain located in their carboxyl termini that is revealed upon fusion of this domain to a heterologous DNA binding motif (8,9). Second, the Smads associate and cooperate functionally with several well characterized transcription factors such as Sp1 (10), AP-1 (11)(12)(13), and the transcriptional coactivator CBP/p300 (14 -19). In addition, receptor-specific Smads and Smad4 are capable of direct DNA binding (20 -22), with target motifs present in the promoters of several TGF-␤ response genes including human type VII collagen (23), human plasminogen activator inhibitor-1 (24 -25), JunB (26), mouse Smad7 (27), and immunoglobulin germline C␣ (28). Additionally, the MH1 domain of Mad, the Drosophila homologue of Smad1, was found to bind Dpp-responsive elements in the vestigial (vg) (29), Ultrabithorax (Ubx) (30), and tinman (tin) (31) promoters. Moreover, Caenorhabditis elegans DAF-3 was reported to associate directly to elements within the C subelement of myo-2 (32). In contrast to the other receptor-specific Smads, Smad2 does not directly bind to DNA and has been shown to require the participation of FAST-1 as an adapter DNA-binding protein (33).
The framework of the molecular mechanism by which Smads activate their target genes has been elucidated through the use of a limited set of TGF-␤-, activin-, and Dpp-responsive genes. Further investigation of the regulation of other TGF-␤ early response genes should provide additional insight into the cellular response of these ligands. Of particular interest are BMP targets, which are poorly characterized. The Xvent homeobox genes have been shown to mediate BMP-4 signaling including dorsoventral patterning of the Xenopus mesoderm. This family * This work was supported by the Deutsche Forschungsgemeinschaft (Kn 200/4 -6; SFB 497/A1) and by Fonds der Chemischen Industrie. 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.
In this study, we have identified a BMP-4 response region in the Xvent-2B promoter between Ϫ275 and Ϫ152. This element is sufficient to confer BMP-4 responsiveness of a reporter as demonstrated by the strong transcriptional activation upon ectopic expression of BMP-4 or co-expression of Smad1 together with Smad4. Furthermore, this region was competent in the formation of a BMP-4-induced protein-DNA complex, which contained Smad1. In vitro binding studies demonstrated that Smad1 and Smad4 are capable of direct binding to this target sequence in two discrete regions. Modification of the Smad4 binding site by mutation was sufficient to abolish transcriptional activation of the Xvent-2B promoter. We also identify for the first time a Smad1 target sequence in a BMP-4 response element. Binding by Smad1 was found to require a GCAT sequence flanked by AT-rich regions. Mutation or deletion of these Smad1 binding sites also disrupted transcriptional activation, further underscoring the role of the identified elements and Smad1/Smad4 in Xvent-2B regulation.

EXPERIMENTAL PROCEDURES
Construction of Plasmids-Human Smad3 and Smad4 cDNAs were obtained from R. Derynck. Smad3 was excised from pRK5 with BamHI/ SalI and subcloned into the BamHI/XhoI sites of pCS2 (41). Smad4 was excised from pRK5 with EcoRI/SalI and subcloned into the EcoRI/XhoI sites of pCS2. His-Smad fusions were generated by subcloning the Smad cDNA into pRSET vectors (InVitrogen) for expression in bacteria.
Microinjections and Luciferase Assays-Xenopus embryos were obtained by in vitro fertilization, and dejellied in 2% cysteine hydrochloride in 0.1ϫ modified Barth's solution (MBS) and staged according to Nieuwkoop and Faber (42). Before injection, embryos were placed into 1ϫ MBS containing 4% Ficoll. Deletion mutants were injected dorsally or ventrally as indicated (20 pg/blastomere) into four-cell stage embryos. Injected embryos were collected at stage 11 and deep-frozen in liquid nitrogen. Luciferase assays were performed as described (43). An internal control was performed using the CMV-pRL vector (cytomegalovirus promoter in front of the Renilla luciferase, Promega) to confirm that ventralization of the embryos by BMP-4 or Smad1/Smad4 coinjections did not affect the transcription of the DNA.
Preparation of Embryonic Extracts-Embryos were injected dorsally at the 4-cell stage with 500 pg of BMP-4 mRNA, collected at stage 10, and frozen in liquid nitrogen. The embryos were homogenized on ice in lysate buffer containing 50 mM Tris-HCl, pH 8.0, 50 mM KCl, 1 mM dithiothreitol, 25% glycerol, 50 mM NaF, 10 mM sodium vanadate, and Complete protease inhibitor mixture (Roche Molecular Biochemicals) and cleared by centrifugation 3 times for 10 min at 10,000 rpm. Total protein concentration was measured by the Bradford assay.
Protein Preparation-Hexa-His-tagged proteins were expressed in Escherichia coli BL21(DE3) and purified under native conditions using Ni-NTA-agarose (Qiagen) according to the manufacturer's protocol. The purified proteins were quantified by the Bradford method.
Electrophoretic Mobility Shift Assays-The Ϫ275/Ϫ75 promoter fragment was amplified by polymerase chain reaction, restricted with BamHI, and labeled with [␣- 32  Binding reactions were performed in a total volume of 50 l containing 25 mM Tris-HCl, pH 8.0, 50 mM KCl, 5 MgCl 2 , 1 mM dithiothreitol, 10% glycerol, and 1 g of poly(dI-dC). Reactions contained 1 g of cellular extracts or His-Smad fusion proteins as indicated. The reactions were allowed to proceed for 30 min at 4°C and analyzed on a 8% native polyacrylamide gel containing 0.5ϫ Tris-borate-EDTA (TBE) (run at 250 V at 4°C). Smad1 (T-20) and Smad2 (S-20) antibodies used in the supershift experiments were purchased from Santa Cruz Biotechnology.
DNase I Footprinting-Binding reactions (total of 20 l) were performed as described for electrophoretic mobility shift assays (EMSAs). After incubation of the reactions, DNase I (0.1 unit) was added and incubated for an additional 2 min at room temperature. The reactions were quenched by the addition of 10 l of loading buffer (10 M urea, 1.5 mM EDTA) and analyzed on a 6% denaturing polyacrylamide gel. Chemical sequencing reactions were performed according to standard protocols.

Identification of the BMP-4-responsive Region in the Xvent-2B Promoter-We have previously demonstrated that
Xvent-2B is an immediate target for BMP-4 signaling and determined the 5Ј boundary responsible for activation and spatial distribution between Ϫ275 and Ϫ174 upstream from the transcription initiation site (36). As shown in Fig. 1, A and B, ventral injection of the Ϫ275/ϩ34 fragment fused to the luciferase reporter supports both basal activity on the ventral side, which maintains an active BMP-4 signal, as well as BMP responsiveness when dorsally co-injected with BMP-4 mRNA. This is in contrast to the Ϫ174/ϩ34 mutant, which exhibited a 3.5-fold reduction in ventral transcriptional activity. Although the Ϫ174/ϩ34 mutant afforded a significantly higher level of ventral activity than the Ϫ32/ϩ34 mutant, it lacked elements that mediate BMP-4 responsiveness on the dorsal side. Be-cause Smads are intracellular transducers of TGF-␤ signaling, we investigated if the identified BMP-4 regulatory region also exhibited Smad1 and Smad4 responsiveness. Consistent with Smad1 as a mediator of BMP-4 signaling, co-injection of Smad1 and Smad4 mRNA with the Ϫ275/ϩ34 construct resulted in nearly a 2-fold activation (Fig. 1B). The lack of transcriptional activation by Smad1/Smad4 on the Ϫ174/ϩ34 mutant parallels the results obtained by co-injection with BMP-4.
To further delineate the region responsible for regulation by BMP-4, a series of internally deleted mutants were constructed by fusing the indicated upstream sequences to the Xvent-2B minimal promoter (Ϫ32 to ϩ34) in front of the luciferase reporter. Internal deletions containing the Ϫ275 distal border injected into ventral blastomeres of four-cell stage Xenopus embryos displayed a progressive loss of transcriptional activity (Fig. 1C). Dorsal co-injection of the internal deletion mutants with Smad1/Smad4 mRNA demonstrated that constructs containing the region from Ϫ152 or longer were sufficient for BMP-4 responsiveness (Fig. 1D). The Ϫ275/Ϫ170 fragment, which showed an intermediate level of basal transcriptional activity on the ventral side, also supported a slight up-regulation of reporter activity when co-injected with Smad1/Smad4 mRNA. However, the level of induction of this mutant by Smad1/Smad4 was lower than the activation obtained with the longer constructs, suggesting elements required for optimal BMP-4 responsiveness are absent.
Taken together, the above results support the conclusion that the minimum BMP-4 response region of Xvent-2B promoter is located between Ϫ275 and Ϫ152. Accordingly, coinjection of the Ϫ275/Ϫ152 reporter construct with BMP-4 mRNA or constitutively active type I BMP receptor (CABR) resulted in a strong stimulation of transcriptional activity, whereas the dominant negative BMP receptor (DNBR) afforded a 5-fold reduction (Fig. 1E). The specificity for the BMP-4 pathway of this element is demonstrated by the lack of influence on transcriptional activity of the Ϫ275/Ϫ152 mutant upon co-injection of Smad4 in combination with Smad3 mRNA, which propagates activin and TGF-␤ signaling (Fig. 1E).
Smad1 Participates in the Nucleoprotein Complex-To further demonstrate that the identified region of the Xvent-2B promoter is responsible for regulation by BMP-4, EMSAs were performed with a 200-base pair fragment spanning the region from Ϫ275 to Ϫ75 containing the entire BMP-4-responsive element (BRE). Cell extracts were prepared from stage 10 embryos injected into the two dorsal blastomeres of 4-cell stage embryos with or without BMP-4 mRNA and incubated with the end-labeled Ϫ275/Ϫ75 DNA fragment. As shown in Fig. 2A, extracts prepared from BMP-4-injected embryos resulted in the induction of a DNA-protein complex, which was also present at a low level in the reaction from control extracts (compare lanes 2 and 3). To determine if Smad1 was a component of this complex, the binding reactions were incubated with an antibody against the linker domain of Smad1. This resulted in a significant decrease in the presence of the BMP-4 induced complex ( Fig. 2A, lane 4). That a disappearance of the complex is observed rather than a supershift in the presence of the antibody may be attributed to the masking of the supershifted complex by the strong BMP-4-independent complex that migrates slightly slower than the BMP-4-induced band. Additionally, the presence of the antibody may preclude DNA binding of Smad1. In any case, the specificity for Smad1 is demonstrated by the lack of reduction of this band in the presence of the Smad2 antibody (compare lanes 4 and 5, Fig. 2A).
Smad1 and Smad4 Directly Bind the BMP-responsive Re- gion-To determine if Smad1 and Smad4 were capable of directly binding to the BRE, bacterially expressed full-length Smad1, Smad4, and Smad3 were incubated with the 32 P-end labeled Ϫ275/Ϫ75 fragment. As shown in Fig. 2B, both Smad1 and Smad4 were capable of binding this fragment in a concentration dependent manner. That the binding is specific for the BMP-4 pathway is suggested by the lack of a DNA shift induced by Smad3. The identified BMP response region (Ϫ275/Ϫ152) of the Xvent-2B promoter contains five putative Smad binding elements (SBE I-V) that conform to the identified minimal Smad recognition sequence, GNCT (Fig. 2C) (21).
Smad4 Binds to a Conserved Smad Binding Element-Previous studies have also identified multiple SBEs in the promoters of TGF-␤-responsive genes that participate in transcriptional activation in a cooperative manner. The major TGF-␤responsive region of PAI-1 was found to contain two specific Smad binding sites, with the distal site conforming to the optimal Smad consensus sequence GTCT and the other GNCT site embedded in an AP1-like motif (45). These same elements are also found in the adjacent SBE III and IV of the Xvent-2B promoter, with a single A to G transition from the canonical AP-1 recognition sequence (AP-1: TGAg/cTCA; SBE III: TGGCTCA).
To confirm that the two putative Smad binding sites (SBE III and IV) in the Xvent-2B promoter are recognized by the Smads, a 29-mer duplex spanning from Ϫ204 to Ϫ176 was prepared and used as a substrate for EMSA. As shown in Fig. 3, Smad4 bound this sequence significantly stronger than Smad1. Compared with Smad4, Smad3 exhibited greater affinity toward this duplex. The strong binding of Smad3 to the SBE III-and IV-derived duplex is consistent with the similarity of this sequence to the previously described TGF-␤-responsive element. Interestingly, the binding affinity of Smad1 and Smad4, as compared with Smad3, to the Ϫ275/Ϫ75 fragment of DNA was significantly enhanced. These findings suggest that additional elements are present in the BRE that contribute to overall binding and selectivity for Smad1 and Smad4 (Fig. 2B). Mutation of both GNCT sites decreased Smad binding and abolished TGF-␤-induced transcriptional activation of the PAI-1 major response element (45). The same mutations introduced into the Xvent-2B promoter also disrupted binding by the Smads (Fig. 3).
A Single Smad4 Binding Site in the Xvent-2B Promoter Is Essential for Smad1/Smad4-induced Transcriptional Activation-The decrease in Smad binding to the mutated duplex suggested that mutation of SBE III and IV might also influence the biological activity of the BMP-4 response element. To study the biological relevance of Smad binding to these elements, the corresponding mutations were created in the BRE of the Xvent-2B promoter (Ϫ275/Ϫ152) and fused to the minimal promoter. As shown in Fig. 4, mutation of these putative Smad sites was sufficient to render the Xvent-2B promoter unresponsive to co-injection of Smad1/Smad4 mRNA. That this element is essential for BMP-4 induction was further demonstrated by the inhibition of Smad1/Smad4 transcriptional activation upon deletion of this sequence from the Ϫ275/ϩ34 mutant containing the downstream sequences (data not shown). To further evaluate the role SBE III and IV contribute to Smad1/Smad4mediated transcriptional activation, Ϫ275/Ϫ152 reporter constructs were prepared with each site separately mutated. Mutation of SBE IV alone inhibited transcriptional activation to levels comparable with that obtained when both SBE III and IV were altered. In contrast, when SBE III was disrupted, Smad1/ Smad4-induced activation was not impaired, but the basal transcriptional activity was severely reduced. Thus, the absolute activity compared with the wild-type reporter was significantly lower.
The other putative SBEs were also mutated to study their role in transcriptional activation of Xvent-2B (Fig. 4). Mutation of SBE II did not inhibit activation but actually led to an enhancement of activity compared with the wild-type sequence. Moreover, mutation of SBE V decreased transcriptional activation, suggesting this region may also contribute to Smad1/ Smad4 responsiveness. Deletion of SBE I had no influence on Smad1/Smad4 transcriptional activation (data not shown).
Localization of Smad1 and Smad4 Binding Sites by DNase I Footprinting Analysis-To determine the exact location of Smad1 and Smad4 binding to the Xvent-2B promoter, DNase I footprinting experiments were performed with the Ϫ275/Ϫ75 DNA fragment. As shown in Fig. 5A, Smad4 protected the region between Ϫ196 and Ϫ176, which contains SBE III and IV. Downstream sequences dispensable for transcriptional activation were not protected. In contrast to Smad4, Smad1 did not protect the proximal region of the BRE but bound to the upstream region in two adjacent sites between Ϫ253 and Ϫ210 (Fig. 5B). As receptor-specific Smads lacking the MH2 domain have been shown to exhibit stronger DNA binding activity than the corresponding full-length protein (29,45), DNA footprint analysis was also performed with Smad1⌬MH2. Protection between Ϫ255 and Ϫ210 by Smad1⌬MH2 was also observed, confirming our results obtained with full-length Smad1 (data not shown). Further evidence that the proximal element is responsible for Smad1 recognition is demonstrated by gel shift analysis with 32 P-end-labeled Ϫ267/Ϫ214 DNA fragment. As shown in Fig. 6A, Smad1 and Smad4 bound to this DNA fragment. However, Smad4 appears to have a stronger affinity to the downstream region compared with the upstream sequence, as this upstream region was not protected by Smad4 in the DNase I protection assay. In competition experiments, unlabeled duplex derived from the activin response element (ARE) of the Mix2 promoter (54) did not effectively compete for Smad1 binding compared with the unlabeled Ϫ267/Ϫ214 probe (compare lanes 3 and 4 with lanes 5 and 6, Fig. 6B) demonstrating the specificity of binding. These results correlate with the requirement of this region for optimal BMP-4 responsiveness. Interestingly, this region lacks characterized Smad binding sites.
Smad1 Binding Is Mediated by GCAT Sequences-To elucidate the recognition elements that facilitate Smad binding, the two regions of Smad1 protection were investigated for conserved motifs. Identified in close proximity to both Smad1protected regions were GCAT and GTAAA sequences. To gain further insight into the elements responsible for Smad1 binding, mutations were made in both the GCAT or GTAAA motifs (Fig. 7B, M2 and M5, respectively). Smad1 binding to the Ϫ267/Ϫ214 DNA fragment was not affected upon mutation of both GTAAA sequences, suggesting this motif does not contribute to Smad1 recognition (Fig. 7A, compare WT2 with M5). However, mutation of both GCAT motifs (M2) was sufficient to abolish Smad1 binding. To more precisely define the contribution of this element to Smad1 binding, the GCAT sequences were individually mutated. Alteration of the distal GCAT sequence (M3) was slightly more effective at disrupting Smad1 binding as compared with the more proximal site (M4) (Fig.  7A). To investigate if the GCAT motifs contribute to BMPinduced transcriptional activation of the Xvent-2B promoter, the sites were mutated in the BRE (Ϫ275/Ϫ152), and co-injection experiments were performed in dorsal blastomeres. In the biological assay, mutation of both GCAT sites consistently afforded a 2-fold reduction in transcriptional activation by Smad1/Smad4 compared with the wild-type BRE construct (Fig. 7C).
To determine if the GCAT sequence was sufficient to mediate Smad1 binding, this motif was embedded in a duplex spanning from Ϫ287 to Ϫ256 of the Xvent-2B promoter (Fig. 8). As shown in Fig. 8, Smad1 did not bind with greater affinity to the M7 duplex compared with the M6 probe, which lacks the GCAT motif (Fig. 8, compare lanes 2 and 4). However, insertion of the distal GCAT site as well as the flanking AT-rich regions (M8)  6 and 8, Fig. 8). Although the proximal GCAT site, like the upstream site, is also embedded in AT-rich regions, the exact nucleotide sequences are not conserved between the two sites. The different behavior of Smad1 toward the individually mutated GCAT sites located in the Ϫ267/Ϫ214 duplex is consistent with the influence of sequences outside the GCAT motif on Smad1 binding.
Taken together, the biological and the in vitro binding experiments indicate that elements between Ϫ275 and Ϫ152 of the Xvent-2B promoter are required for BMP-4-induced transcriptional activation mediated by Smad1/Smad4. The identified BRE contains a proximal site that is essential for activity and constitutes the conserved Smad4 binding element, whereas the upstream region mediates Smad1 association via GCAT motifs flanked by AT-rich sequences. The Smad4 site was found to be absolutely necessary for transcriptional activation mediated by Smad1/Smad4. To facilitate full-activation of the Xvent-2B BRE, the Smad1 binding sites were also required. DISCUSSION BMP-4 plays an essential role in vertebrate embryogenesis, where it promotes the patterning of ventral mesoderm. Of its downstream targets in Xenopus laevis, the Xvent-2 family is of high interest because it has been shown to be an immediate response gene of BMP-4 signaling (36,37). Ectopic expression of Xvent-2 members affords the same ventralized phenotype as observed with BMP-4. Furthermore, Xvent-2 can rescue the majority of effects induced by the dominant negative type I BMP receptor, suggesting it is a central mediator of BMP signaling (36 -40, 46). The requirement for Xvent-2 function in dorsoventral patterning was further demonstrated through the use of dominant negative constructs (46). Investigation of the molecular mechanism by which BMP-4 activates Xvent-2 not only provides insight into the patterning in embryogenesis but can also be used to further deduce the manner in which TGF-␤ family members exert their discrete cellular responses.
Although promoter/enhancer elements that respond directly to TGF-␤, activin, and Dpp have been characterized for several genes, studies of an element that exhibits an exclusive BMP-4-induced transcriptional activation are limited. A recent study of the Xvent-2 promoter using cell transfection experiments identified a 53-base pair region in the Xvent-2 promoter that responds to BMP (47). Presently, through the use of reporter assays in Xenopus embryos, we have delineated the BMP-4responsive region in the Xvent-2B promoter between Ϫ275 and Ϫ152. The Xvent-2B BRE contains regulatory elements similar to the closely related Xvent-2 BRE (47). Both elements were found to contain a CAGAC motif at its core and require flanking sequences downstream from the Smad4 site. In the present report, we demonstrate through the use of gel mobility shift assays and DNase I protection experiments with bacterially expressed protein that Smad4 directly binds this CAGAC sequence. Moreover, the identified BRE in the Xvent-2B promoter requires an additional upstream region that is necessary for full transcriptional activation of Xvent-2B by Smad1 and Smad4. Through the use of in vitro binding studies, Smad1 binding to the upstream region was identified and characterized. The present study provides a more detailed analysis of the Smad1 and Smad4 recognition elements of the Xvent-2B promoter, which mediate activation by BMP-4.
The crystal structure of the Smad3 MH1 domain bound to a palindromic GTCTAGAC octamer has been determined, revealing insight into the specificity of DNA binding by the Smads (21). In the crystal structure, sequence-specific DNA recognition was facilitated by interaction of amino acids with only three bases of the Smad binding element. As these residues are invariant among the Smads, the GNCT motif or its palindrome AGNC is expected to be recognized by other Smads with the exception of Smad2. Smad binding elements within promoters of TGF-␤ and activin-responsive genes have been shown to contain multiple motifs containing this core sequence. However, the regulation of TGF-␤ signaling specificity remains unclear and must be controlled by additional parameters, as some of the identified Smad recognition sites were found to be dispensable for biological activity (20).
The BMP-responsive region of Xvent-2B contains five putative AGNC SBEs differing in their contribution toward Smadinduced transcriptional activation. Through deletion and point mutations it was demonstrated that loss or modification of SBE I and III did not disrupt Smad1/Smad4 responsiveness, but mutants exhibited significantly lower basal levels of activity. Thus, the absolute level of activation was reduced. Mutation of SBE II enhanced Smad1/Smad4-mediated transcriptional activation as compared with the corresponding wild-type sequence, suggesting that this site may participate in the negative regulation of Xvent-2B. Alternatively, the SBE II may have been converted to an element that is recognized by a transcription factor that cooperates with the Smads. In contrast to the upstream SBEs, deletion or mutation of the proximal SBE V significantly disrupted Smad1/Smad4 responsiveness. SBE IV was the only SBE that was found to be essential for Smad1/ Smad4 responsiveness, consistent with the recent analysis of the Xvent-2 promoter (47).
A palindromic GTCTAGAC sequence was defined by polymerase chain reaction selection studies as the optimal binding site for Smad3 and Smad4. However, studies have not been performed that demonstrate that this is the optimal Smad1 target sequence (22). Although the amino acids that directly participate in base-specific DNA recognition are conserved among the Smads, several residues immediately preceding the DNA binding ␤-hairpin domain are conserved between receptor specific Smads of the same class that may confer subtle variations in Smad binding specificity (21). Consistent with this observation, Smad1, Smad4, and Smad3 were capable of binding this sequence; however, slightly different binding affinities were observed (21).
Differences in Smad DNA binding specificity were also observed on the Xvent-2B promoter where Smad1 and Smad4 preferentially protected distinct sites of the BRE in DNase I protection assays. Smad4 bound to the sequences between Ϫ196 and Ϫ176, which contain SBE III and IV, the latter of which was shown to be essential for Smad1/Smad4-mediated transcriptional activation. These results suggest that additional sites outside the core Smad site confer specificity to TGF-␤ signaling. This has already been suggested by the predominance of the sequence AGc/aCAGACA in multiple TGF-␤responsive promoters, which contains at its core a GNCT sequence in reverse orientation (24). Interestingly, the essential SBE IV of the Xvent-2B promoter has flanking sequences very similar to this AGAC Smad3/Smad4 target site found within several TGF-␤-responsive regions and adjacent to FAST binding sites in the activin-responsive element (Table I) (24,26,28,32,48,49). This motif is not limited to vertebrates, as Daf-3 from C. elegans, which is most closely related to Smad4, was also reported to mediate a transcriptional response through this motif (32). An exception may be the Drosophila Smad4 homologue, Medea, which was shown to recognize GC-rich tar-get sites (31,50). As these elements are responsive to multiple TGF-␤ pathways, this site most likely represents the Smad4 binding site, which is shared among the various pathways. However, demonstrated here (Fig. 3A) and in previous reports, this sequence also represents a Smad3 recognition site. Thus, for full-transcriptional activation by TGF-␤, at least two copies of this sequence would be required to accommodate Smad3 and Smad4. This is supported by the requirement of multimerized copies of this element from the PAI-1 promoter for TGF-␤ responsiveness (22,24,51). A recent study has demonstrated that a region of the PAI-1 promoter containing two SBEs, when present in a single copy, was sufficient for TGF-␤ transcriptional activation (52). Additionally, when the JunB TGF-␤responsive SBE was multimerized four times, it was also induced by activin and BMP-2. The activation of this artificial reporter was most likely due to binding via Smad4, because Smad1 and Smad2 were not able to bind this element (26). Further support that this conserved element is a Smad4 binding site is demonstrated by activin A-responsive elements that harbor a copy of this sequence. The SBE is bound by Smad4, whereas Smad2 DNA association is achieved indirectly through FAST binding (48). The Xvent-2B BRE contains a single copy of this conserved Smad4 recognition sequence with a different sequence motif being recognized by Smad1.
Whereas the oligomeric state of receptor-specific Smads and Smad4 in the transcriptional complex is unclear, Kawabata et al. (55) suggest that Smad hetero-oligomers may be trimers but do not exclude that other oligomeric states of the Smads exist. The identified Smad recognition sites in the Xvent-2B promoter would accommodate binding for a trimer composed of one molecule of Smad4 and two Smad1 molecules. The 53-base pair  DNase I footprinting experiments demonstrated that Smad1 preferentially protects the DNA between Ϫ253 and Ϫ210 in two distinct regions, which are required in addition to the Smad4 binding site to mediate optimal Smad1/Smad4 transcriptional activation. An element that is sufficient to mediate Dpp responsiveness of tinman expression was also found to contain two essential Smad binding sites, one that bound Mad exclusively and another that was recognized by both Medea and Mad (31). Interestingly, the site of Smad1 binding does not resemble that of the Smad4/Smad3 target site, the GCCGNCGc recognition sequence of Mad, or the clusters of GNCN repeats that are often found in TGF-␤-responsive regions (29,51). Both of the regions protected by Smad1 in the Xvent-2B promoter contained GCAT sequences, which through mutational analysis of the promoter were shown to mediate Smad1 binding. This motif contains the first and last positions of the previously identified Smad binding element, which together participate in four of the five base-specific hydrogen bonds between Smad3 and DNA (21). An intact GCAT sequence is required for Smad1 binding; however, it was not sufficient, as Smad1 binding was dependent on the flanking AT-rich regions (Fig. 8).
Although mutation of the conserved Smad4 site was sufficient to abolish Smad1/Smad4-induced transcriptional activation of the BRE, mutation of both Smad1 binding sites only led to a reduction in Smad1/Smad4 transcriptional responsiveness. The contribution of the GCAT sequence toward Smad1/Smad4mediated transcriptional activation is further demonstrated by the reduction of activity of the Ϫ245/Ϫ152 mutant, which does not contain the distal GCAT site, compared with the Ϫ251/ Ϫ152 mutant upon co-injection of Smad1/Smad4 mRNA (data not shown). It has been previously suggested that BMP signaling can occur without Smad1 binding (26). The lack of requirement of the upstream Smad1 binding sites for transcriptional activation of Xvent-2B resembles TGF-␤ and activin-responsive promoters that utilize FAST binding in addition to Smad4. In the case of the activin-inducible element of Mix.2, the FAST binding site was demonstrated to be absolutely essential for transcriptional activation (48). When the Smad binding site was mutated, stimulation by activin was observed, but activation was impaired compared with the wild-type response. The FAST binding site in the gsc element was also shown to be sufficient to mediate activin-induced activation; however, the Smad4 site was required for optimal activation (53).
Smad binding sequences of TGF-␤-responsive genes are often located adjacent to binding motifs for other transcriptional activators, with the type of factor being dependent on the nature of the responsive element. It is therefore thought that transcriptional activation by the Smads is obtained through the use of double DNA-specific promoters. One element confers specificity to the Smads, and the other site is bound by transcription factors that synergize with the Smads. The results presented here demonstrate that Smads exhibit a double specific nature in binding to the Xvent-2B promoter, with a distal Smad1 region and a proximal Smad4 binding site. Detailed analyses of BMP-4-responsive elements from additional genes are required to reveal, if the identified elements in the Xvent-2B promoter represent a conserved mechanism for BMP-4 signaling. Moreover, the cooperation of these identified sites with other transcription factors may be required to confer additional DNA binding specificity and full activation by BMP-4, as has been demonstrated for other target genes of TGF-␤ signaling. Such a transcriptional co-activator has just recently been reported (47). The zinc finger protein OAZ was shown to bind directly to Smad1 and activate the Xvent-2 promoter. A dominant-negative OAZ was also able to decrease the levels of Xvent-2 transcripts and partially inhibit the effects of BMP in Xenopus explant assays. However, although expression of OAZ in the early gastrula stage is ubiquitous and could thus be a candidate for Xvent-2 activation, at later stages the spatial expression patterns of Xvent-2 and OAZ do not overlap. Therefore, the search for additional factors that are expressed at the correct location is under present investigation.