Identification and Functional Characterization of Distinct Critically Important Bone Morphogenetic Protein-specific Response Elements in the Id1 Promoter*

Transforming growth factor-β (TGF-β) family members, which include bone morphogenetic proteins (BMPs) and TGF-βs, elicit their cellular effects by activating specific Smad proteins, which control the transcription of target genes. BMPs and TGF-βs have overlapping as well as specific effects on mesenchymal cell differentiation for which the mechanisms are incompletely understood. Here we report that Id1, a dominant negative inhibitor of basic helix-loop-helix proteins, is a direct target gene for BMP. BMP, but not TGF-β, strongly activates the Id1 promoter in an Smad-dependent manner. We identified two BMP-responsive regions in the mouse Id1 promoter, which contain three distinct sequence elements; one region contains two Smad binding elements (SBEs), and the other region contains a GGCGCC palindromic sequence flanked by two CAGC and two CGCC motifs. Whereas SBEs and GGCGCC sequence are critically important, the CAGC and CGCC motifs are needed for efficient BMP-induced Id1 promoter activation. Smads are part of nuclear transcription factor complexes that specifically bind to SBEs and GGCGCC sequence in response to BMP but not TGF-β. Multimerization of the all three distinct sequence motifs is needed to generate a highly sensitive and BMP/Smad-dependent specific enhancer. Our results provide important new insights into how the BMP/Smad pathway can specifically activate target genes.

. Therefore, the mechanism on how the BMP/Smad pathway specifically activates target genes is still incompletely understood.
The affinity of Smads for DNA is weak (18). Smads need thus to co-operate with other DNA binding factors to bind efficiently to promoters of target genes. The 30-zinc finger nuclear protein OAZ associates with BMP R-Smads in response to BMP (24). Expression of OAZ is cell type-specific and cannot be detected in mesenchymal cells. A member of the core binding factor (CBF) family of transcription factors, Cbfa1, also termed Runtrelated gene 2 (RUNX2) was shown to interact directly with Smad1 and Smad5 (25,26). RUNX2 precedes the appearance of osteoblasts, and mice deficient in RUNX2 lack osteoblasts, and bone ossification is completely blocked (27,28). RUNX2 and Smads have been shown to co-operate in transcriptional responses and BMP-induced osteoblast differentiation (29,30). Ectopic expression of RUNX2 in C2C12 cells was found to induce many of the extracellular matrix proteins induced by BMPs (29). However, RUNX2 alone is not sufficient to induce the whole onset of osteoblast differentiation without co-operation with Smad5 (29). Thus, other BMP targets genes with critical roles in BMP-induced osteogenesis remain to be identified.
Inhibitors of differentiation (Id) proteins act as dominant negative inhibitors of basic helix-loop-helix transcription factors, including members of the MyoD family of myogenic transcription factors, and also pocket proteins, including the retinoblastoma protein (31). Id and basic helix-loop-helix proteins dictate in an opposite manner cellular programs of differentiation and proliferation in various cell types. Because BMPs can induce Id expression (32,33) and ectopic expression of Id proteins can mimic certain BMP-induced responses, e.g. inhibiting myoblast differentiation (34,35), 2 Id proteins may serve an important effector function for BMPs. In the present paper we show that BMP specifically transcriptionally activates Id1 expression in C2C12 myoblasts and map two distinct BMP-specific responsive regions containing three pivotal distinct sequence motifs in the Id1 promoter.

EXPERIMENTAL PROCEDURES
Expression and Reporter Plasmids-Expression plasmids for Smads and type I receptors have been previously described (36). Mouse Id1 promoter fragment (Ϫ1575/ϩ88) (37) and human Id1 promoter have been described before (38). Id1, Id2, and Id3 cDNAs were provided by the Dr. H. Weintraub laboratory.
Mouse Id1 promoter 5Ј deletion constructs were made by PCR approach using mouse Id1-(Ϫ1575/ϩ88) fragment as a template. For precise mapping of the BMP-responsive element, Id1 promoter fragments were PCR-amplified using Id1-(Ϫ1231/ϩ88) fragment as a template and subcloned between KpnI and XmaI sites upstream of minimal adenoviral major late promoter (MLP) reporter construct (17). Site directed mutagenesis was performed using QuikChange site-directed mutagenesis kit (Stratagene) protocol. The wild-type GTCT sites were replaced by the mutated GagT, the wild-type CAGC sites were changed into the mutated Ctca sequence, the wild-type GCCG elements were replaced by GaCG-mutated sequence, and the wild type GGCGCC palindrome, GCЈ-1, and GCЈ-2 sequence elements were replaced with GaattC, tGCa, tcaC, respectively. All constructs were DNA sequence-verified.
Transient Transfections and Transcriptional Reporter Assays-Cells were seeded at a density of 1.5 ϫ 10 4 cells/cm 2 in 12-well (C2C12 cells) or 6-well (MDA-MB468 cells) plates. The next day, cells were tran-siently transfected with the reporters in the absence or presence of expression plasmids using pcDNA4 plasmid to keep total amount of transfected DNA constant. Transfection was carried out using the calcium phosphate co-precipitation method (4 g of total plasmid DNA/ well) in the case of MDA-MB468 and HepG2 cells or FuGENE 6 (Roche Molecular Biochemicals) transfection reagent (500 ng of total plasmid DNA/well) in the case of C2C12 cells, following the manufacturer's protocol. ␤-Galactosidase co-transfection was used as an internal control for normalizing transfection efficiency. Luciferase and ␤-galactosidase activity were quantified using the luciferase assay (Promega) with Victor luminometer (Wallac) as described previously (39).
RNA Isolation and Northern Blot Analysis-RNA was isolated with using TriZol reagent (Invitrogen) in accordance with manufacturer protocol. 20 g of total RNA/lane was loaded on denaturating (6% formaldehyde) 1% agarose gels. Blotting and hybridization using 32 Plabeled (Amersham Biosciences, Inc.) probes were performed as described previously (36).
Electrophoretic Mobility Shift Assays-Nuclear extracts from untreated C2C12 cells and C2C12 cells stimulated with BMP-6 (100 ng/ ml) or TGF-␤3 (5 ng/ml) were prepared according to Schreiber et al. (41). GST-Smad fusion proteins were prepared as described before (17). The binding reaction containing 20 pM 32 P-end-labeled oligonucleotides and appropriate nuclear extracts (15 g of protein/sample) or GST-Smad proteins (500 g of purified protein/sample) were performed as described in Dennler et al. (17). A ϫ100, ϫ250, or ϫ500 molar excess of unlabeled oligonucleotides was used as a competitor where indicated. One g of purified anti-Smad5 (42) and anti-Smad2 and three g of anti-Smad4 (43) rabbit polyclonal antibodies were used for supershift experiments.

Id1 Is a Direct BMP Target
Gene-Previous studies show that BMPs, but not TGF-␤, potently induce the expression of Id proteins (32,33,44), which are inhibitors of basic helix-loophelix proteins (31). Because we wanted to obtain new insights into the molecular mechanism of how certain genes can be selectively induced by BMP, but not by TGF-␤, the Id genes were selected for further study. We found that BMP4 and BMP7 potently induce mRNA levels of Id1, Id2, and Id3 in C2C12 cells (Fig. 1A). Induction of Id expression by BMPs peaks around 1-2 h. Among the three genes analyzed, Id3 was found to be most stably induced within the 1-24-h time period, and Id1 was found to be most rapidly induced upon BMP challenge (Fig. 1A); Id1, likely to be an immediate early BMP response gene, was therefore chosen for further analysis. Induction of Id1 by BMP6 was indeed observed in the presence of cycloheximide (CHI), indicating that de novo protein synthesis is not required for this response (Fig. 1B). Id1 is thus a direct BMP target gene. Id1 mRNA was actually more induced in the presence of BMP6 and CHI (Fig. 1B), probably as a result of an increase in mRNA stability or loss of transcriptional repressors by CHI. The addition of actinomycin D (ActD), an inhibitor of transcription, inhibited the BMP-induced effect on Id1 expression (Fig. 1B), indicating that the up-regulation of the Id1 protein is transcriptionally dependent. BMP6 stimulation also resulted in an up-regulation of the Id1 protein in C2C12 cells. This effect was blocked by ActD (and CHI) (Fig. 1C), indicating that in these cells BMP does not increase the Id1 expression by increasing the stability of the Id1 protein.
BMP/Smad-induced Activation of the Id1 Promoter-When we transfected the mouse or human Id1-Luc reporter constructs in C2C12 cells and stimulated the cells with BMP or TGF-␤, we found that only BMPs activated these Id1 reporters ( Fig. 2, A and B). On another transcriptional reporter, TGF-␤ activated a transcriptional response, indicating that C2C12 cells are TGF-␤-responsive, as previously reported (5). Consistent with these findings, BMP-induced activation of reporter is potently inhibited by dominant negative BMP type I receptors and stimulated by constitutively active (ca) BMP type I receptors, i.e. caALK2, caALK3, and caALK6 but not caALK4 and caALK5 (which are type I receptors for TGF-␤ and activin, respectively) ( Fig. 2C and data not shown). To examine whether Smad proteins are involved in the BMP6-induced transcriptional activation of Id1-Luc reporter, we co-transfected C2C12 cells with the mouse Id1-Luc reporter and expression constructs for R-Smads, i.e. Smad1 or Smad5 or inhibitory Smads, i.e. Smad6 or Smad7. We found that Smad1 or Smad5 strongly promoted both basal and BMP-induced activation of this reporter (Fig. 2D) and that both I-Smads strongly inhibited both basal and BMP-induced luciferase levels (Fig. 2E). These findings suggest an involvement of Smad proteins in BMPinduced activation of Id1-Luc reporter. The observation that basal levels are reduced by I-Smads suggests that there is FIG. 1. Id genes are induced by BMPs in C2C12 cells. A, Northern blot analysis of Id1, Id2, and Id3 mRNA expression in C2C12 cells stimulated with BMP4 (100 ng/ml) or BMP7 (300 ng/ml) for the indicated times. B, effect of CHI (5 g/ml) or ActD (5 g/ml) on BMP6-induced Id1 expression in C2C12 cells. CHI and ActD were added 1 h before BMP6. Equal loading of RNA samples loaded (20 g/lane) in A and B was checked by the ethidium bromide staining pattern of gels before Northern blotting. Ribosomal markers 18 S and 28 S are indicated. C, effect of BMP6 on Id1 protein expression was measured in the absence or presence of CHI and ActD. Expression of Id1 was detected by Western blotting using anti-Id1 antibody. As negative and positive controls for Id1 recognition by anti-Id1 antibody, cell lysates from mock-transfected COS cells and COS cells transfected with Id1 expression plasmid were analyzed. Cell lysates were also analyzed for phosphorylated Smad1 to show response of C2C12 cells to BMP in the absence or presence of CHI and ActD. As positive controls for phosphorylated Smad1, a cell lysate from COS cells transfected with caALK2 and Smad1 was used.
autocrine BMP signaling in C2C12 cells.
To show that Smad4 is required for the BMP6-induced activation of Id1-Luc reporter, we used MDA-MB468 cells. MDA-MB468 cells are human epithelial cells deficient in Smad4. In these cells, BMP6 has no effect on Id1-Luc reporter activity (Fig. 2F). However, co-transfection of an expression construct encoding Smad4 restored the BMP6 activation of Id1-Luc reporter, demonstrating that Smad4 is necessary for the BMP6 transcriptional effect. This BMP6-induced activation of Id1-Luc reporter can be further enhanced by co-expressing Smad4 with Smad1 or Smad5. These results suggest that the Co-Smad4 and R-Smads (i.e. Smad1 and Smad5) co-operate in BMP6-induced activation of Id1-Luc reporter.
Identification of Two Distinct BMP-response Elements in Id1 Promoter-Interestingly, when we compared the mouse Id1 with the human Id1 promoter sequences, which are both strongly activated by BMPs (Fig. 2, A and B), we noted that these two promoters are most similar in the identified BMPresponsive region (Fig. 4). Importantly, we noticed within the Id1-(Ϫ1133/Ϫ1025) fragment the presence of five GTCT motifs, also termed SBEs (17,19), three GCCG sequences, which have previously shown to be important for activation of BMP-responsive promoters (22,23), five conserved CGCC, and four conserved CAGC motifs. The latter two sequence motifs have not been previously implicated as BMP response elements.
To examine the importance of SBE, GCCG, CAGC, and CGCC sequence elements in BMP-induced activation of mouse Id1 promoter, we performed site-directed mutagenesis on several of these motifs and analyzed the effect on BMP-induced transcriptional activation of the Id1-Luc reporter constructs. Comparing the effect of BMP on Id1-(Ϫ1133/Ϫ996) (construct 1; Fig. 5) with Id1-(Ϫ1133/Ϫ1025) (construct 9; Fig. 5A), we found that both fragments were potently activated by BMP6, but the Id1-(Ϫ1133/Ϫ996) was consistently more so than the Id1-(Ϫ1133/Ϫ1025) (Fig. 5). This indicates that SBE-4, SBE-5, CAGC-3, and CAGC-4 ( Fig. 5A) may contribute to response but are not critically important. When we mutated the GCCG elements (GC-1 and GC-3; constructs 8 and 12), we found little change on the fold of BMP activation and even an increase in the level of BMP-induced activation compared with wild-type constructs (constructs 1 and 9) (Fig. 5). Mutation of CAGC-1 and CAGC-2 showed that both sites, but in particular CAGC-1, are important in mediating BMP response to this promoter. The maximal level of BMP-induced activation is severely affected by mutation of CAGC-1, and BMP fold of activation is greatly diminished by mutation of CAGC-2. Mutation of SBE-2 and SBE-3 led to a drastic reduction in BMP6 induction compared with the wild-type Id1 promoter, indicating that these sites are critically important. Upon mutation of both sites, the reduction was even greater, and this mutated reporter construct only minimally responded to BMP6 (Fig. 5). Also, the basal activity of the reporter construct was affected by mutation of SBE-2 and SBE-3. The reduction in basal transcriptional activity is likely caused by interference with autocrine BMP signaling (Fig. 2E).
Upon mutating the GCЈ-3,-4 elements (construct 13), which are arranged as a palindrome GGCGCC, we observed a pronounced inhibitory effect on both basal and the fold activation by BMP6. The inhibitory effect upon mutating the palindrome was intermediate to the effect we found upon mutating both CAGC or both SBE elements (Fig. 5). When we mutated GCЈ-1 and GCЈ-2 (construct 14), we affected BMP-induced activation but much less as mutating the GCЈ-3,4 sequence motifs. GCЈ-1 and GCЈ-2 are needed for most efficient activation of the Id1 promoter by BMP. Because GCЈ-1,-2 sequence elements overlap with putative binding sites for Egr-1, Sp-1, and YY1, it is possible that these transcription factors participate in BMPinduced activation of Id-1 promoter. However, we were unable to promote BMP-induced Id1 promoter activation by ectopic expression of YY1 and Egr-1 (data not shown).
Binding of Smad5 and Smad4 to BMP Response Elements-To test whether transcription factor complexes can bind to the BMP-responsive Id1 promoter fragments, we performed EMSA using nuclear extracts from C2C12 cells that were not treated or treated with BMP6 or TGF-␤. We could identify binding complexes to the Id1 promoter fragment (Ϫ1070/Ϫ1025) containing SBE-2 and SBE-3 with nuclear extracts from cells treated with BMP6 but not with TGF-␤. The BMP6-induced binding was specific since it could be competed with excess SBE containing oligonucleotides (Fig. 6A). In addition, no BMP6-induced binding was observed to Id1-(Ϫ1070/ Ϫ1025), in which two SBEs were mutated (data not shown). Based upon previous studies (17), Smad4 is likely to mediate the direct interaction of the transcription factor complex with SBE-2-and SBE-3-containing Id1-(Ϫ1070/Ϫ1025) fragment. Indeed, we observed direct and specific binding Escherichia coli-expressed GST-Smad4, but not GST-Smad5⌬MH2, to the Id1-(Ϫ1070/Ϫ1025) fragment (Fig. 6B).
To our surprise, we also found specific BMP-inducible (but not TGF-␤-inducible) binding of transcription factor complexes on the Id1-(Ϫ1133/Ϫ1071) fragment containing GCЈ-1 to GCЈ-4 flanked by two CAGC sites using nuclear extracts from C2C12 cells (Fig. 6E). The BMP6-induced binding to Id1-(Ϫ1133/Ϫ1071) was competed with excess unlabeled wild-type fragment Id1-(Ϫ1105/Ϫ1080) containing GGCGCC palindrome flanked by two CAGC sites or the same unlabeled fragment in which the two CAGC motifs were mutated (Fig.  6C). However, BMP6-induced binding to Id(Ϫ1133/Ϫ1071) was not competed with the same unlabeled fragment in which the palindrome was mutated or with excess SBE oligonucleotides (Fig. 6C). These results indicate that GCЈ motifs, but not CAGC elements, contribute mainly to the BMP6-induced binding. To determine which GCЈ motifs are important, we compared the BMP6-induced binding to wild-type Id1-(Ϫ1133/Ϫ1071) or to the same fragment in which either GCЈ-1,-2 or GGCGCC palindrome were mutated. Whereas mutation of palindrome completely abrogated binding, BMP6induced binding to the fragment with mutation of GCЈ-1,-2 was slightly less efficient than to the wild type fragment (Fig.  6D). Taken together, these results indicate that GGCGCC palindrome is the main DNA binding determinant in Id1-(Ϫ1133/Ϫ1071). Furthermore, we could supershift the BMPinduced binding of the transcription factor complex(es) to the Id1-(Ϫ1133/Ϫ1071) fragment with a specific antibody to Smad5 and Smad4 but not with a antibody specific to Smad2 (used as a negative control in the experiment) (Fig. 6E). Thus, the BMP6-induced transcription factor complex that binds to Id1-(Ϫ1133/Ϫ1071) contains both Smad5 and Smad4. To determine whether the binding of Smads to Id1-(Ϫ1133/Ϫ1071) is direct, we used E. coli-expressed GST-Smad5⌬MH2 and GST-Smad4 proteins in EMSA. The Smad5⌬MH2 domain and Smad4 were found to specifically interact with Id1-(Ϫ1133/Ϫ1071) (Fig. 6E). The Smad DNA binding results are in full agreement with the activation of the Id1-luc reporter by overexpression of Smad4 and Smad5 (Fig. 2).  Fig. 3B, the two Id1 fragments (Ϫ1070/Ϫ1025 and Ϫ1133/Ϫ1071) containing SBEs and CGCC sequence motifs flanked by CAGC motifs, respectively, are both necessary but only together are sufficient to confer strong BMP enhancer function to a minimal promoter reporter. These two Id1 fragments contain (putative) transcription factor binding sites for YY1, cAMP-response element-binding protein (CREB)/ATF2, Egr-1, and Sp1 and may thus respond to other stimuli (37). In an attempt to generate a BMP-specific transcriptional reporter assay, we fused the Id1-(Ϫ1052/Ϫ1032) containing only SBE-2, SBE-3, and GCЈ-5 with Id1-(Ϫ1105/Ϫ1080), which just contains GGCGCC palindrome flanked by CAGC-3 and CAGC-4 (both fragments lack the putative transcription factor binding sites distinct from Smads) and cloned it upstream of a minimal promoter reporter. Although this reporter construct was insufficient to confer BMP inducibility (data not shown), two repeats of these two fused Id1 fragments provided a strong BMPspecific enhancer function (Fig. 7A). This construct was termed the BRE-Luc reporter. Two repeats of (i) the fragment containing GGCGCC palindrome flanked by CAGC motifs, (ii) the fragment containing SBE-2,3 and GCЈ-5, or (iii) the fragment containing SBE-2,3, GCЈ-5, and CAGC-1,2 were insufficient to confer BMP responsiveness to the minimal promoter reporter construct (Fig. 7A). A weak BMP-induced activation was observed with two repeats of the Id1 fragment containing SBE-2,-3, GC-5Ј, and GGCGCC palindrome (Fig. 7A). These results are fully consistent with functional inactivation studies by site-directed mutagenesis in Fig. 5; SBE-2,3 and GGCGCC palindrome are most important for BMP-induced Id1 promoter activation, and CAGC motifs are needed for efficient activation.

Generation of a Specific BMP/Smad Transcriptional Reporter-As shown in
A strict dose-dependent response is observed when cells transfected with BRE-Luc reporter are stimulated with increasing concentrations of BMP6 (Fig. 7C). The BRE-Luc reporter is also specifically activated by BMP6 but not TGF-␤ or activin in other BMP-responsive cell types, including HepG2 cells (Fig. 7D and data not shown). Whereas wild-type Id1 promoter can be activated by serum (37), we found that the BRE-Luc reporter is not activated by serum (Fig. 7D). DISCUSSION In the present investigation we show that BMPs, but not TGF-␤, rapidly activates the mouse Id1 promoter via two regions, one of which contains CGCC sequence elements flanked by CAGC motifs and the other of which harbors two SBEs. Although SBE-2 and SBE-3 and GGCGCC palindrome are most important for DNA binding of Smads and conferring BMP responsiveness to the Id1 promoter, CAGC-1, -2, and GCЈ-1,-2 motifs are needed for efficient BMP-induced mouse Id1 promoter activation. Like the mouse Id1 promoter, the human Id1 promoter is also robustly activated by BMP. Interestingly, all sequence elements that were identified to be important for BMP6-induced activation of mouse Id1 promoter are completely conserved in the human Id1 promoter (Fig. 4). Mutations in GTCT and CAGC boxes or GC-rich sequence elements were generated by site-directed mutagenesis in Id1-(Ϫ1133/Ϫ996) and Id1-(Ϫ1133/Ϫ1025). Both fragments were cloned upstream of the MLP luciferase reporter. The wild-type GTCT sites were replaced by the mutated GagT, the wildtype CAGC sites were changed into the mutated Ctca sequence, the wild-type GCCG elements were replaced by GaCG mutated sequence, and the wild type GGCGCC palindrome, GCЈ-1, and GCЈ-2 sequence elements were replaced with GaattC, tGCa, tcaC, respectively. Wildtype and mutated Id1 promoter luciferase constructs were transfected into C2C12 cells and subsequently treated with or without BMP6 (100 ng/ml). Fold-induction values in the presence of BMP6 compared with basal levels are indicated.
We observed that BMPs induce the Id1 expression by directly activating its promoter (Fig. 2). The effect of BMP on Id1 mRNA expression was found to be transcriptionally dependent and occurs in the absence of de novo protein synthesis (Fig. 1). Whereas a first report on BMP-enhanced Id1 expression had shown that BMP-induced effect on Id1 expression was (contrary to our findings) not blocked by ActD (33), our results are in agreement with subsequent studies in which BMP exerts effects on Id1 expression at the transcriptional level (32,45). The level of activation of the mouse and human Id1 promoters FIG. 6. BMP-inducible binding of nuclear transcription factors and direct binding of Smads to Id1 promoter fragments. A, an EMSA was performed with nuclear extracts from C2C12 cells non-treated or treated with TGF-␤3 (5 ng/ml) and BMP6 (100 ng/ml) using Id1-(Ϫ1070/Ϫ1025) containing SBE-2 and SBE-3 as a 32 P-labeled probe. Comp., competitor. B, An EMSA of E. coli-expressed GST-Smad4 or GST-Smad5⌬MH2 using Id1-(Ϫ1070/Ϫ1025) as a 32 P-labeled probe. Unlabeled (SBE) 4 was used as competitor in A and B. C, an EMSA was performed with nuclear extracts from C2C12 cells from non-treated or treated with BMP6 (100 ng/ml) and Id1-(Ϫ1133/Ϫ1071) containing GCЈ1-4 repeats flanked by CAGC motifs as a 32 P-labeled probe. Wild-type Id1-(Ϫ1105/Ϫ1080) (designated Pwt/CAGCwt), Id1-(Ϫ1105/Ϫ1080) with either mutated GGCGCC palindrome (designated Pmut/CAGCwt) or mutated CAGC motif (designated Pwt/CAGCmut), or Id1-(Ϫ1052/Ϫ1032) containing SBE-2 and SBE-3 (designated (SBE) 2 wt), was used as competitors. D, an EMSA was performed with nuclear extracts from C2C12 cells from non-treated or treated with BMP6 (100 ng/ml) using wild-type Id1-(Ϫ1133/Ϫ1071) (designated Wt) or the same fragment with GCЈ-1,-2 mutated (designated GCЈ-1,2mut) or GCЈ-3,-4 mutated (designated P-mut) as a 32 P-labeled probe. E, an EMSA was performed with nuclear extracts from C2C12 cells from non-treated of treated with BMP6 (100 ng/ml) or TGF-␤3 (5 ng/ml) and Id1-(Ϫ1133/Ϫ1071) as a labeled probe. Specific anti-Smad5, anti-Smad2, and anti-Smad4 antibodies (Ab) were added where indicated. F, an EMSA of E. coli-expressed GST-Smad5⌬MH2 or GST-Smad4 using Id1-(Ϫ1133/1071) as a 32 P-labeled probe. Id1-(Ϫ1052/Ϫ1032) (designated (SBE) 2 ) or Id(Ϫ1105/Ϫ1080) (designated P/CAGC) were used as competitors. Where indicated, a 100-fold molar excess (in A and B) or a 250-fold molar excess (in C and F) of non-radiolabeled oligonucleotide was added as competitor. The retarded probe-protein complexes are indicated with an arrow or arrowheads, in the case of supershifted complexes. that is achieved by BMP6 is high (17-and 15-fold for mouse and human promoters, respectively) compared with other promoters that are activated by specific extracellular stimuli. For example, we could only observe 5.5-fold induction by TGF-␤ of the PAI-1 promoter, a gene that is strongly and directly induced by TGF-␤ (17).
The observation that BMPs, but not TGF-␤, strongly activate the Id1 promoter can be explained by the presence of Smad4 and Smad5 binding elements that are present in multiple copies in the Id1 promoter (Fig. 5). Smad5 and Smad4 are pivotal components in BMP pathway; their ectopic expression is sufficient to activate the Id1 promoter, and blocking of their activation or removing Smad4 abrogates the BMP-induced Id1-Luc activation (Fig. 2). The Id1 promoter is not activated by TGF-␤, which is consistent with the finding that TGF-␤ (unlike BMP) did not induce Smad binding to SBE or GGCGCC/CAGC-containing sequences in Id1 promoter fragments (Fig. 6). The flanking sequences of the SBEs previously were shown to be an important determinant for Smad binding (46). Probably, flanking sequences of the SBEs in Id1 promoter may not favor the in vivo binding of the Smad3⅐Smad4 complex. Of note, among the putative Drosophila Mad binding sites in vestigial quadrant enhancer and Ubx midgut enhancer, GC-rich sequences flanked by a CAGC motif were found to bind Drosophila Mad (which is highly similar to mammalian BMP R-Smads) with highest affinity (47). We are currently analyzing whether clusters of SBE and GGCGCC palindrome with flanking CAGC sequences in the promoters of other direct BMP target genes such as JunB and connective tissue growth factor are conferring BMP responsiveness to these genes. SBE, CGCC, and to a lesser extent CAGC sequence motifs were identified as most important for BMP-induced Smad binding and activation of the Id1 promoter. SBE, but not CGCC or CAGC sequence elements, have previously been implicated in BMP R-Smad binding. The three sequence motifs are distinct but also have similarity to each other. The 5Ј-CAGACA-3Ј sequence is often found as SBE (17) in which the nucleotides indicated in bold are involved in making direct contact with Smad (18). Change of these three nucleotides indicated in bold to other nucleotides were also found to have a profound effect on Smad4 binding, whereas other nucleotides had less effect (19). The CAGC sequence motif is often followed by an extra C in the Id1 promoter and is thus similar to CAGACA sequence and may bind Smads. The CGCC element differs from the SBE sequence and has higher similarity to CAGC motif. If we presume that Smad5 and Smad4 contact GGCGCC palindrome via their ␤-hairpin loops in an analogous fashion as Smad3 with SBE, then the GC-rich sequences are not expected to bind FIG. 7. Generation of a specific BMP/Smad transcriptional reporter. A, C2C12 cells were transfected with BRE-Luc reporter, i.e. two copies of the Id1-(Ϫ1105/1080) fragment fused to two copies of the Id1-(Ϫ1052/Ϫ1032) fragment cloned upstream of the MLP promoter or selected parts of these two Id1 fragments cloned upstream of the MLP promoter. B, C2C12 cells transfected with BRE-Luc reporter were stimulated with TGF-␤3 (5 ng/ml), activin (20 ng/ml), or BMP6 (100 ng/ml). C, dose-dependent response of BMP-6 on BRE-Luc in C2C12 cells. D, HepG2 cells were transfected with BRE-Luc reporter and not treated or treated with 10% fetal calf serum (FCS) in the absence or presence of BMP-6, TGF-␤3, or activin A. A-D, after 20 h, luciferase activity was measured, and values were corrected for transfection efficiency. Representative results are shown. Smad5 or Smad4 with high affinity (18). However, binding of BMP6 activated the Smad5⅐Smad4 complex to GGCGCC palindrome or SBE2 and SBE3 containing Id1 fragments appears equally strong. It is possible, therefore, that Smad5 and Smad4 make contact with GGCGCC palindrome via a domain distinct from a ␤-hairpin loop or that another cellular factor that is part of a common complex with Smad5⅐Smad4 facilitates the interaction of Smad5⅐Smad4 with GGCGCC palindrome.
BMPs, but not TGF-␤, activin, or serum activate the BRE-Luc reporter in C2C12 cells in a strictly dose-dependent manner (Fig. 7, B-D). This reporter may therefore be used to measure BMP-like biological activity in biological samples. We found that the BRE-Luc reporter is functional and specific in cells other than C2C12 cells, including HepG2 cells, although the fold of activation by BMP is cell type-dependent. The BRE-Luc reporter may be used in any cell that has a functional BMP receptor/Smad pathway. This is unlike the specific BMP reporter containing the Xvent promoter, which works only in selected cells, because of a requirement for a cell type-specific expressed Smad partner (24). In addition to the activated BMP type I receptors, activated ALK1, which signals via the same Smad proteins as BMP type I receptors, also stimulates the BRE-Luc reporter (data not shown). The BRE-Luc reporter complements our previous developed reporters, (SBE) 4 -luc (39) and (CAGA) 12 -Luc (17). Whereas (SBE) 4 -luc can be activated by all TGF-␤ family members, (CAGA) 12 -luc is only activated by TGF-␤ and activin (17,39).
BMP inhibits myogenic differentiation (6). Id1 is strongly induced by BMP and is likely to be important mediator of the inhibitory effect of BMP on myogenic differentiation. Like BMPs, TGF-␤ also inhibits myogenic differentiation (9). However, TGF-␤ has been reported to have no effect on Id1 expression (44), and we found that TGF-␤ does not activate Id1 promoter (Fig. 2). Consistent with this notion, TGF-␤ has been shown to inhibit the activity of myogenin through a mechanism that is independent of interference with myogenin DNA binding (44) and is, thus, independent of Id proteins. Therefore, the mechanism by which BMP and TGF-␤ inhibit myogenic differentiation of mesenchymal precursor cells is likely to be different.