Identification of a BMP-responsive element in Id1, the gene for inhibition of myogenesis

Authors


  • Communicated by: Kohei Miyazono

* Correspondence: E-mail: katagiri@dent.showa-u.ac.jp
aPresent address: Department of Oral and Maxillofacial Surgery, Tokyo Medical University, Tokyo, Japan.

Abstract

Background: Bone morphogenetic protein-2 (BMP-2) stimulates osteoblast differentiation, but inhibits myogenic differentiation in C2C12 myoblasts. BMP-2 induces transcription of Id1, an inhibitor for myogenesis, within 1 h in the cells. To examine the molecular mechanism of the action of BMP-2, we analysed a BMP-2-responsive element (BRE) in the 5′ flanking region of the human Id1 gene.

Results: A GC-rich region between −985 bp and −957 bp of the human Id1 gene was identified as a BRE. The BRE containing promoter activity was stimulated by BMP-2 or by constitutively active BMP receptors (BMPR-IA and BMPR-IB). The stimulation was blocked by co-transfecting with dominant negative BMPR-IA or Smad7. A unique DNA–protein complex was induced in response to BMP-2 on the BRE. The complex induced by BMP-2 contained Smad1 and Smad4, possibly as a complex of both Smads. BMP-2 failed to stimulate the expression of Id1 mRNA in Smad4-deficient cells. Over-expression of Smad4, but not Smad1, stimulated the Id1 reporter activity and the expression of endogenous Id1 mRNA in Smad4-deficient cells.

Conclusion: Signalling of BMP-2 to stimulate the expression of Id1 would be transduced by BMPR-IA and mediated by Smad1 and Smad4, both of which form a complex on the 29 bp GC-rich element.

Introduction

Bone morphogenetic proteins (BMPs), members of the transforming growth factor-β (TGF-β) superfamily, control the proliferation, differentiation and apoptosis of various types of cells (for reviews, see Kingsley 1994; Hogan 1996; Wozney & Rosen 1998; Reddi 2001). BMPs were originally identified as an activity that induces ectopic bone formation when it is implanted into muscular tissue (Urist 1965). BMPs are the active proteins responsible for ectopic bone formation (Wozney et al. 1988; Luyten et al. 1989; Celeste et al. 1990; Sampath et al. 1990). Skeletal abnormalities are found in animals and patients with mutations in BMP genes, confirming the physiological importance of BMPs in skeletal development in vertebrates (Kingsley et al. 1992; Storm et al. 1994; Thomas et al. 1996, 1997). Decapentaplegic (Dpp), a BMP homologue in Drosophila, also functions as a key morphogen in fly development (for review, see Raftery & Sutherland 1999). Several recombinant BMP proteins have been shown to induce not only ectopic bone formation in vivo but also osteoblast differentiation of mesenchymal cells in vitro (Wozney et al. 1988; Katagiri et al. 1990; Sampath et al. 1990; Yamaguchi et al. 1991; Sampath et al. 1992; Asahina et al. 1993; Katagiri et al. 1994).

Signalling by TGF-β superfamily members including BMPs is initiated following their binding to their respective membrane receptors. Two types of serine/threonine kinase receptors, types I and II, are required for the signal transduction (for reviews, see Sakou 1998; Kawabata et al. 1998; Miyazono 1999; Massague 2000; Massague & Chen 2000; Wrana 2000). Among them, BMPR-IA, BMPR-IB and ALK2 are classified as the type I receptors, and BMPR-II, ActR-II and ActR-IIB are the type II receptors. The kinase activity of the ligand-bound type II receptor phosphorylates the type I receptor to activate its kinase activity (Wrana et al. 1994). Over-expression of the kinase domain-truncated BMP type I receptors blocks the signal transduction induced by BMPs (Maeno et al. 1994; Suzuki et al. 1994; Namiki et al. 1997). In contrast, substitution mutations of the BMP type I receptors at the conserved residue in the GS domain activate the downstream signal transduction without ligand-binding or phosphorylation by the type II receptors (Wieser et al. 1995; Akiyama et al. 1997; Fujii et al. 1999). The activated type I receptors phosphorylate Smad proteins as substrates in the cytoplasm. The type I receptors also activate a signalling pathway through mitogen activated protein (MAP) kinases (Yamaguchi et al. 1995).

In the intracellular signalling activated by the type I receptors of BMPs, Smad proteins have been found to play critical roles in osteoblast differentiation induced by BMPs (Yamamoto et al. 1997; Nishimura et al. 1998; Fujii et al. 1999). Eight Smad proteins have been identified in mammals. These Smads are classified into three subgroups: R-Smads, Co-Smads and I-Smads, according to their structure and function (for reviews, see Kawabata et al. 1998; Sakou 1998; Miyazono 1999; Massague 2000; Massague & Chen 2000; Wrana 2000; Shi 2001). The R-Smads consist of Smad1, Smad2, Smad3, Smad5 and Smad8. They are directly phosphorylated at the carboxy terminal by the type I receptors. BMP receptors phosphorylate Smad1, Smad5 and Smad8, while the TGF-β and activin receptors phosphorylate Smad2 and Smad3. Thus, the differences in biological activities between BMPs and TGF-β/activin appear to be due to the different usage of the specific R-Smads by the type I receptors. The phosphorylated R-Smads form complexes with Co-Smad, Smad4, then move into the nucleus, bind to the regulatory regions of the target genes, and control their expression. In contrast, I-Smads such as Smad6 and Smad7 inhibit signal transduction of the TGF-β superfamily members by stably interacting with the type I receptors to block phosphorylation of R-Smads (Hayashi et al. 1997; Imamura et al. 1997; Nakao et al. 1997).

To examine the molecular mechanism of ectopic bone formation induced by BMPs, we established an in vitro model system using C2C12 myoblasts (Katagiri et al. 1994). BMPs inhibit the myogenic differentiation of C2C12 cells, and convert their differentiation pathway into that of osteoblast lineage cells (Katagiri et al. 1994). TGF-β1 also inhibits myogenic differentiation, but fails to induce the osteoblast differentiation of C2C12 cells (Katagiri et al. 1994). C2C12 cells express BMPR-IA, but no detectable BMPR-IB (Akiyama et al. 1997; Namiki et al. 1997). Over-expression of the constitutively active (ca)BMPR-IA or caBMPR-IB in C2C12 cells mimicked the effects of the treatment with BMPs (Akiyama et al. 1997; Fujii et al. 1999). In contrast, over-expression of the kinase domain-truncated BMPR-IA, but not BMPR-IB, blocked the BMP-2 response in a dominant negative fashion in C2C12 cells (Namiki et al. 1997). Other studies also reported that the intracellular signalling of BMPs is mainly mediated by the BMP-specific R-Smads such as Smad1 and Smad5, since the over-expression of Smad1 or Smad5 in C2C12 cells induced osteoblast differentiation and inhibited myogenic differentiation (Yamamoto et al. 1997; Nishimura et al. 1998; Fujii et al. 1999). However, the direct target genes of Smad proteins, which are involved in the conversion of the differentiation pathway of C2C12 cells by BMPs, are still not known.

Myogenic differentiation is controlled by the MyoD family transcription factors that have a conserved basic helix-loop-helix (bHLH) structure (for reviews, see Tapscott & Weintraub 1990; Ludolph & Konieczny 1995; Molkentin & Olson 1996; Arnold & Winter 1998; Perry & Rudnicki 2000; Buckingham 2001). The Id family proteins are capable of forming heterodimers with other bHLH proteins, such as MyoD family proteins, through their HLH domains (Benezra et al. 1990). However, the Id proteins lack the DNA-binding activity, because they do not have the basic region, an essential domain for the DNA binding (Benezra et al. 1990). Thus it is reasonable that the HLH heterodimers formed with the Id proteins inhibited the transcriptional activity in a dominant-negative mode. Indeed, myogenic differentiation was suppressed when Id1 was over-expressed in C2C12 cells (Benezra et al. 1990). We previously reported that the expression of Id1 was stimulated in C2C12 cells within 1 h after treatment with BMP-2 (Katagiri et al. 1994). Recently, Id1 was identified as a typical early responsive gene for BMP treatment in various types of cells in mice and humans (Hollnagel et al. 1999; Clement et al. 2000). These findings suggest that Id1 could be a direct target gene in the intracellular signalling pathway of BMPs.

In the present study we characterized and identified a 29 bp GC-rich element as a BMP responsive element (BRE) in the 5′ region of the human Id1 gene. Electrophoresis mobility shift assay (EMSA) showed that this element was recognized by Smad1 and Smad4 in response to BMP-2. BMP-2 failed to stimulate Id1 expression in Smad4-deficient cells, but the transient over-expression of Smad4 rescued the BMP-2 response in these cells. These results suggest that BMP-2 induced Id1 expression is mediated by the binding of Smad1 and Smad4 to the BRE in the Id1 gene.

Results

Identification of a 29 bp region in the Id1 gene as a BMP-responsive element

To identify a BMP-responsive element (BRE) in the human Id1 gene, we constructed a series of luc reporter plasmids from Id2.1-luc, which carries a 2.1 kb fragment of the 5′ flanking region of the human Id-1 gene (Fig. 1A). Although Id2.1-luc, Id1.3-luc, Id1.2-luc and Id1.0-luc responded similarly to BMP-2 in C2C12 cells, Id0.8-luc completely lost the ability to respond to BMP-2 (Fig. 1A). The region between −1.0 kb to −0.8 kb contained several potential regulatory elements which may be recognized by transcription factors such as Egr-1, CREB and Smads (Fig. 1B). We further generated an additional series of deletion plasmids and examined their responsiveness to BMP-2. Although Id985-luc responded to BMP-2 similarly to Id1.0-luc, neither Id956-luc, Id916-luc, nor Id889-luc responded to BMP-2 (Fig. 1B). The 29 bp region between −985 and −957 of the human Id1 gene showed 100% homology with the corresponding region of the mouse Id1 gene (Fig. 1C). This region contains a consensus sequence of Egr-1 which has been identified as a responsive element for the expression of Id1 in response to serum (Tournay & Benezra 1996). To clarify whether Egr-1 is involved in the Id1 expression induced by BMP-2 as an enhancer, we constructed Id985mutA-luc and Id985mutB-luc, in which substitutional mutations had been introduced at the inside and the outside of the Egr-1 consensus, respectively (Fig. 1C,D). Id985mutA-luc partially responded to BMP-2, but Id985mutB-luc almost completely lost the ability to respond to BMP-2 (Fig. 1D). To further characterize the ability of the 29 bp region to respond to BMP-2, we put four copies of the 29 bp fragment with the wild-type or mutB sequence in tandem in the forward (IdWT4F-luc and IdmutB4F-luc) or reverse orientation (IdWT4R-luc) in front of the SV40 promoter (Fig. 1E). Both IdWT4F-luc and IdWT4R-luc were activated in response to BMP-2, but IdmutB4F-luc was not (Fig. 1E).

Figure 1.

Identification of a BRE in the human Id1 gene. (A) Deletion analysis of the 5′ region of the human Id1 gene in luciferase reporter constructs. Deletion constructs were generated from Id2.1-luc. C2C12 cells were transfected with one of the reporter constructs and treated overnight with or without BMP-2. (B) Identification of BRE between −985 bp and −957 bp of the human Id1 gene. Constructs generated by PCR using Id2.1-luc as a template were transfected in C2C12 cells, and their response to BMP-2 was then examined. Consensus elements for Egr-1 (Egr-1), CREB (CRE) and Smads (CAGA) are indicated. (C) Comparison of the human sequence around the BRE, with the corresponding region of the mouse gene. The BRE identified in the human Id1 gene is underlined. Identical nucleotides between the human and mouse Id1 genes are indicated by asterisks. Positions of substitutional mutations used in this study are shown as MutA and MutB. The consensus sequence of Egr-1 is boxed. (D) MutB sufficiently blocked the response to BMP-2 in Id985-luc reporter. Mutations mutA and mutB were introduced in Id985-luc to generate Id985mutA-luc and Id985mutB-luc, respectively. The luciferase activities were determined in C2C12 cells in the presence and absence of BMP-2. (E) Characterization of the GC-rich 29 bp between −985 bp and −957 bp of the human Id1 gene as a BMP-responsive enhancer. Reporter constructs carrying four copies of the 29 bp element with wild-type or mutB sequence in forward (IdWT4F-luc and IdmutB4F-luc) or reverse orientation (IdWT4R-luc) were transfected in C2C12 cells, and luciferase activity was determined in the presence or absence of BMP-2. Data are means ± s.d. (n = 3).

BMPR-IA transduces intracellular signals to activate Id1 gene expression in C2C12 cells

Using Id1.0-luc, we examined signalling pathways to activate Id1 expression in response to BMP-2 in C2C12 cells. In agreement with our previous findings (Katagiri et al. 1994), TGF-β1 slightly suppressed the Id1.0-luc in C2C12 cells, whereas BMP-2 greatly stimulated it in them (Fig. 2A). Similar results were obtained in C3H10T1/2 cells (data not shown). When Id1.0-luc was co-transfected with the constitutively active BMP type I receptors, caBMPR-IA or caBMPR-IB, both receptors significantly stimulated Id1.0-luc, even in the absence of BMP-2 in C2C12 cells (Fig. 2B). The stimulatory activity of caBMPR-IB appeared slightly higher than that of caBMPR-IA. In contrast, the activation of Id1.0-luc by BMP-2 was blocked by co-transfecting with the dominant negative BMP receptor, dnBMPR-IA, but not with dnBMPR-IB (Fig. 2B). Co-transfection with Smad7, which blocks Smad-dependent signalling pathways, also blocked the activation of Id1.0-luc by BMP-2 (Fig. 2C). To visualize the induction of Id1 in living cells, we constructed Id985-EGFPd2 by replacing the coding sequence of luciferase in Id985-luc with an EGFPd2 cDNA. We have reported that the responsiveness to BMP-2 in C2C12 cells was blocked by over-expressing dnBMPR-IA, but not dnBMPR-IB (Namiki et al. 1997). The subclonal cell line C2C12ΔIA12, was established from the C2C12 cultures stably transfected with the dnBMPR-IA (Namiki et al. 1997). When C2C12 and C2C12ΔΙΑ12 cells were transfected with Id985-EGFPd2 and treated with BMP-2, an increased number of EGFP-positive cells appeared in the C2C12 cell cultures within 4 h, but not in the C2C12ΔΙΑ12 cell cultures (Fig. 2D). BMP-2 failed to increase the number of the EGFP-positive cells in C2C12 cultures transfected with Id985mutB-EGFPd2 (data not shown).

Figure 2.

Characterization of signalling pathways to activate Id1 transcription. (A) BMP-2 specifically activates Id1.0-luc. C2C12 cells were transfected with Id1.0-luc, and then treated with BMP-2 or TGF-β1. Luciferase activity was determined after overnight incubation. Data are means ± s.d. (n = 3). (B) BMPR-IA mediates BMP-2 signals in C2C12 cells. C2C12 cells were co-transfected with Id1.0-luc and one of the constitutive active (c) or the kinase domain-truncated (dn) BMPR-IA (IA) or BMPR-IB (IB), and then incubated with or without BMP-2. After an overnight incubation, luciferase activity was determined. Data are means ± s.d. (n = 3). (C) Smad7 blocks the BMP-2-induced response of Id1.0-luc. C2C12 cells were co-transfected with Id1.0-luc and a Smad7 expression construct. After overnight treatment with BMP-2, the luciferase activity was determined. Data are means ± s.d. (n = 3). (D) Id985-EGFPd2 is activated by BMP-2 in C2C12 cells but not in C2C12 cells stably expressing a kinase domain-truncated BMPR-IA (C2C12ΔIA12). Id985-EGFPd2 was transfected in C2C12 and C2C12ΔIA12 cells, and then the cells were incubated for 4 h with and without BMP-2. Cells were monitored using a fluorescence microscope.

Smad1 and Smad4 bind to the 29 bp BRE of the Id1 gene

Next, we characterized transcription factors bound to the 29 bp BRE of the Id1 gene in EMSA. BMP-2 induced DNA-binding protein(s) in nuclear extracts of C2C12 cells within 1 h, which formed Complex A with the 29 bp BRE probe (Fig. 3A). In contrast, PMA, a potent stimulator of Egr-1, induced Complex C, a slower migrating DNA-protein, but not Complex A (Fig. 3A). In competition experiments, a 20-fold excess amount of cold oligo DNA with the wild-type sequence partially competed with Complex A induced by BMP-2, and a 200-fold excess almost completely competed with it (Fig. 3B). MutA oligo also partially competed with Complex A at a 200-fold excess, but neither mutB, Egr-1, nor Sp1 consensus oligo DNA competed with Complex A (Fig. 3B). Complex B was a sequence with nonspecific binding, since it was competed by all of the cold oligo DNAs examined. The addition of anti-Egr-1 antibody to the reaction mixtures specifically abolished Complex C induced by PMA, but not Complex A induced by BMP-2 (Fig. 3C). In contrast, antibodies against Egr-2, Egr-3, Sp1 or p300 affected neither the amounts nor mobility of Complexes A and C in EMSA (Fig. 3C and data not shown). In contrast, addition of anti-Smad1 or anti-Smad4 antibody to the BMP-2-treated nuclear extracts reduced the amount of Complex A and induced supershifted bands (Fig. 3D). Moreover, when both the anti-Smad1 and anti-Smad4 antibodies were added together, a slower migrating double supershifted band appeared (Fig. 3D). In this condition, no other supershifted bands appeared. The addition of anti-Smad6/7 or anti-Smad3 antibodies did not induce a supershift of Complex A (Fig. 3D).

Figure 3.

Characterization of transcription factors bound to the GC-rich 29 bp BRE of the Id1 gene in EMSA. (A) BMP-2 specifically induces Complex A. Nuclear extracts were prepared from C2C12 cells treated with and without BMP-2 or PMA. EMSA was performed using the BRE as a probe. (B) Competition analysis of Complex A induced by BMP-2. Various cold oligo nucleotides were added together with the BRE probe to the mixtures. (C) Complex C but not Complex A contains Egr-1. Antibodies against Egr-1, Egr-2, Egr-3 and Id1 (unrelated control) were added to the nuclear extracts prepared from C2C12 cells treated with BMP-2. (D) Complex A contains both Smad1 and Smad4. Antibodies against Smad1, Smad3, Smad4 or Smad6/7 were added to the nuclear extracts prepared from C2C12 cells treated with BMP-2. Smad1+4; Antibodies for Smad1 and Smad4 were added simultaneously to the mixture.

A Smad4-dependent signalling controls the expression of Id1 induced by BMP-2

We further examined the role of Smads in the expression of Id1 induced by BMP-2 using human MDA-MB468 breast cancer cells, in which the Smad4 locus has been deleted (Schutte et al. 1996). BMP-2 failed to activate IdWT4F-luc even in MDA-MB468 cells after co-transfecting with Smad1 or Smad3 (Fig. 4A). However, co-transfection with Smad4 stimulated both basal and BMP-2-induced luc activities in these cells (Fig. 4A). Moreover, simultaneous co-transfection with Smad1 and Smad4 markedly enhanced the luc activity of IdWT4F-luc, even in the absence of BMP-2 in MDA-MB468 cells (Fig. 4A). Similar results were obtained when Id1.0-luc was used in place of IdWT4F-luc (data not shown).

Figure 4.

Smad4 is required for the expression of Id1 in response to BMP-2 in MDA-MB468 cells. (A) Transient transfection with Smad4, but not Smad1 or Smad3, stimulates the expression of Id1 in response to BMP-2 in MDA-MB468 cells. Co-transfection with Smad1 and Smad4 stimulates the expression of Id1, even in the absence of BMP-2 in MDA-MB468 cells. Cells were transfected with a fixed amount of DNA with and without Smad1, Smad3 or Smad4, and then incubated in the presence or absence of BMP-2. Their luciferase activities were determined after overnight incubation. Data are means + s.d. (n = 3). (B) Transient over-expression of Smad4 in MDA-MB468 cells induces the expression of endogenous Id1 mRNA in response to BMP-2. Cells were transfected with Smad4 and then treated with BMP-2 for 1 h. Expression of Id1 and GAPDH mRNA was determined by Northern blotting. Expression levels of Id1 were normalized to the levels of GAPDH.

Furthermore, we examined the expression of the endogenous Id1 mRNA induced by BMP-2 in MDA-MB468 cells. BMP-2 failed to induce the expression of Id1 mRNA in MDA-MB468 cells (Fig. 4B). However, transient over-expression of Smad4 in these cells induced the expression of endogenous Id1 mRNA in response to BMP-2 (Fig. 4B).

Effects of specific inhibitors for MAPK, PI3K and de novo protein synthesis on the Id1 expression induced by BMP-2

Recently, it was reported that BMP-2 activated not only Smad proteins but also mitogen-activated protein kinase (MAPK) and phosphatidylinositol 3-kinase (PI3K) in some types of cells (Iwasaki et al. 1999; Vinals et al. 2002). The contribution of these signalling pathways to the Id1 expression induced by BMP-2 was examined using specific inhibitors for the kinases; SB203580 for p38 MAPK, PD98059 for MEK, and wortmannin and LY294002 for PI3K. These inhibitors affected neither the both basal nor the BMP-2-induced levels of the Id1 mRNA in Northern blot analysis (Fig. 5A). In contrast, when C2C12 cells were pre-treated with cycloheximide to block the de novo protein synthesis, the basal level of Id1 mRNA was increased to a similar level induced by BMP-2, but no further stimulation of Id1 mRNA was observed, even in the presence of BMP-2 (Fig. 5B).

Figure 5.

Effects of specific inhibitors for MAPK, PI3K and de novo protein synthesis on the levels of Id1 mRNA induced by BMP-2. C2C12 cells were pre-treated for 1 h with and without 10 µm SB203580, 50 µm PD98059, 0.1 µm wortmannin, or 20 µm LY294002 (A), and 10 µg/mL of cycloheximide (CHX) (B). The cells were then cultured for an additional 1 h with fresh media containing each inhibitor in the presence or absence of 300 ng/mL of BMP-2. Expression of Id1 and GAPDH mRNAs was determined by Northern blotting using 20 µg of total RNA.

Discussion

In the present study, a responsive element for BMP-2 was identified in the human Id1 gene. We previously reported that BMP-2 induced the expression of Id1 mRNA within 1 h after BMP-2 treatment in C2C12 cells (Katagiri et al. 1994). Other studies also identified Id1 as an early responsive gene induced by BMP-2 in various types of mouse and human cells (Hollnagel et al. 1999; Clement et al. 2000). To elucidate the molecular mechanisms underlying BMP-2-induced Id1 expression, we cloned and analysed the 5′ region of the human Id1 gene. We found that a GC-rich region between −985 bp and −957 bp of the human Id1 gene was the responsive element to BMP-2. This 29 bp element showed 100% homology between human and mouse Id1 genes, suggesting that this element has a critical role in the expression of Id1 in mammals. We previously reported that the over-expression of either BMPR-IA or BMPR-IB with the constitutively active mutation induced the osteoblast differentiation in C2C12 cells (Akiyama et al. 1997). In contrast, the only BMPR-IA lacking a kinase domain inhibited the responsiveness to BMP-2 in C2C12 cells (Namiki et al. 1997). In agreement with our previous findings, the stimulation of the Id1 reporter activity by BMP-2 was blocked by dnBMPR-IA, but not by dnBMPR-IB. These results suggest that BMPR-IA has a much higher affinity for BMP-2 than BMPR-IB when these receptors are expressed on the surface of C2C12 cells. This possibility is supported by the findings of Ebisawa et al. (1999), who demonstrated, using affinity cross-linking followed by the immunoprecipitation, that BMP-2 only binds to BMPR-IA on the surface of C2C12 cells.

In EMSA, using the 29 bp region in the Id1 gene as the BRE probe, we identified both Smad1 and Smad4 in a DNA-protein complex induced by BMP-2. Interestingly, when antibodies against Smad1 and Smad4 were added simultaneously to the complex, a double supershifted band was observed. In this condition, however, we could not see any other supershifted bands corresponding to that observed by adding each individual antibody against Smad1 or Smad4. These results suggest that Smad1 and Smad4 bind to the element together, possibly as a complex, rather than individually. Transient over-expression of Smad4 in Smad4-deficient MDA-MB468 cells recovered the expression of not only the reporter gene but also endogenous Id1 mRNA in the presence of BMP-2. Taken together, these results suggest that a complex of Smad1 and Smad4 mediates the transcription of Id1 in response to BMP-2 by binding to the GC-rich 29 bp element. The Id1 protein induced by Smad1/Smad4 would be involved, at least in part, in the BMP-2-dependent conversion of the differentiation pathway of C2C12 myoblasts into that of osteoblast lineage cells. This possibility is under investigation in our laboratories.

It has been reported that BMP-2 activates not only Smad proteins but also MAPK and PI3K to induce the respective intracellular signals (Iwasaki et al. 1999; Vinals et al. 2002). However, none of the specific inhibitors for MAPK or PI3K used in this study showed significant effects on the Id1 mRNA expression in C2C12 cells in the presence or absence of BMP-2. In addition, these inhibitors showed only small effects on the differentiation of C2C12 cells into mature muscle cells and osteoblasts in the absence and presence of BMP-2, respectively (data not shown). Although we cannot rule out the possibility that MAPK and/or PI3K are involved in the effects of BMP-2, the contribution of these signalling pathways would be minor at least in the expression of Id1 in C2C12 cells.

Smad proteins are capable of binding to several different DNA sequences in vitro. It has been reported that Mad, a Drosophila homologue of mammalian Smad1, binds to a GC-rich sequence of the target genes in Drosophila (Kim et al. 1997; Xu et al. 1998). Recently, a similar GC-rich sequence was identified in a regulatory region of the mouse Smad6 gene, and this element could be recognized by the BMP-specific Smads such as Smad1 and Smad5 (Ishida et al. 2000). It was reported that the MH1 domains of Smads also bind to the CAGA sequence (Dennler et al. 1998; Jonk et al. 1998; Shi et al. 1998). However, this sequence appears to have a lower ligand specificity, since it was recognized by both BMP- and TGF-β/activin-specific Smads, together with Smad4. Indeed, there were three copies of the CAGA sequence within the BMP-responsive region between −1.0 kb and −0.8 kb of the human Id1 gene, but these sequences were not necessary for the BMP-2-induced Id1 expression. Moreover, the treatment of C2C12 cells with TGF-β1 or the transient over-expression of Smad3 did not activate expression of Id1. In contrast, the GC-rich element between −985 bp and −957 bp of the human Id1 gene was critical and sufficient for the response to BMP-2. This element was recognized by Smad1 and Smad4, but not Smad3. Taken together, it can be concluded that this GC-rich element, rather than the CAGA sequence, behaves as the BRE in the Id1 gene. Further studies are necessary to determine the exact binding sequence of Smad1 and Smad4 to form a DNA-protein complex.

It was reported that the Id1 expression induced by serum is mediated by the binding of Egr-1 to a consensus sequence within the GC-rich region of the 29 bp BRE (Tournay & Benezra 1996). In contrast to serum, BMP-2 induced Complex A rather than Complex C in the EMSA containing Smad1/Smad4 and Egr-1, respectively. Mutational analysis also showed that a 3′ portion of the Egr-1 consensus sequence in the 29 bp element is more critical than the consensus itself in the expression of Id1 induced by BMP-2. Moreover, both competition and supershifting experiments in EMSA showed that the DNA-protein complex induced by BMP-2 contained no detectable Egr-1 protein. In our preliminary experiments, PMA, a potent inducer of Egr-1 in various types of cells, induced Egr-1, but failed to stimulate the Id1 mRNA expression in C2C12 cells. These results suggest that Egr-1 is not involved in the expression of Id1 in response to BMP-2.

Pre-treatment of C2C12 cells with cycloheximide increased the basal levels of Id1 mRNA, but no further stimulation was observed, even in the presence of BMP-2. In addition, DNA-protein complexes were not detected in EMSA, when bacterially expressed full length Smad1 and Smad4 were incubated with the 29 bp BRE as a probe (data not shown). These results suggest that, although a Smad protein itself binds to a specific DNA sequence with a low affinity, it forms a transcriptional complex with other DNA-binding protein(s) to enhance the ability to regulate the target gene expression. The DNA-binding proteins may have a short half life and be continuously degraded through the de novo protein synthesis pathway as a rate-limiting transcriptional factor in the expression of Id1 mRNA. It has also been suggested that the role of Smad proteins in Id1 expression is to form a complex with such an unstable transcriptional factor to stabilize them on the promoter. A number of transcription factors have been identified as partners for Smads. However, transcription factors, which are associated with Smad1 and Smad4 to enhance the expression of Id1 in response to BMP-2, are not known at present. Although P300/CBP was reported to stimulate the transcriptional activity of Smads (Feng et al. 1998; Janknecht et al. 1998; Nishihara et al. 1998; Pouponnot et al. 1998; Topper et al. 1998), we failed to detect them in a supershift experiment in EMSA. OAZ, a zinc finger protein, was reported as a DNA-binding factor that is associated with Smad1 and Smad4 in response to BMP-2 (Hata et al. 2000). The transient over-expression of OAZ stimulated the Xvent-2 reporter expression in mammalian cells including C2C12 and C3H10T1/2 cell lines (Hata et al. 2000). However, OAZ does not appear to be involved in the expression of Id1 in response to BMP-2, since OAZ is undetectable in C2C12 cells in their basal condition (Hata et al. 2000). Further studies are necessary to characterize the transcription factors in the DNA-protein complex induced by BMP-2.

While we were preparing this manuscript, Lopez-Rovira et al. (2002) and Korchynski & ten Dijke (2002) independently reported a BMP-2 responsive element and a BMP-6 responsive element, respectively, in the Id1 gene. In agreement with our findings, they showed the importance of the GC-rich element overlapping with the BRE—which was identified in the present study—for the BMP-dependent stimulation of the Id1 gene through Smad1/Smad5 and Smad4. Although the CAGA sequences were not essential for the BMP response in our study, they reported that these elements were also required for the BMP response (Korchynski & ten Dijke 2002; Lopez-Rovira et al. 2002). This discrepancy has to be clarified by future experiments. Competitive displacement experiments in EMSA also suggested that Smad1 and Smad4 bind to the GC-rich region (Korchynski & ten Dijke 2002; Lopez-Rovira et al. 2002). We were able to localize the binding region of the Smad1 and Smad4 complex into a 29-bp GC-rich region in EMSA. Furthermore, the present report shows for the first time that the transient over-expression of Smad4 can rescue endogenous Id1 expression in response to BMP-2 in Smad4-deficient cells.

In conclusion, we have identified a 29-bp BRE from the human Id1 gene. This GC-rich element was recognized by Smad1 and Smad4, and it was sufficient for the response to BMP-2. We also successfully generated an EGFP reporter construct driven under the control of the BRE to visualize the BMP-dependent signalling in living cells. The BRE sequence, as well as the reporter constructs characterized in this study, would be useful for an understanding of the molecular mechanism of BMP-dependent gene expression.

Experimental procedures

Construction of reporter plasmids

To construct a luciferase reporter plasmid, a 2.1-kb fragment of the 5′ region of the human Id1 gene was cloned by a standard PCR method using Platinum Pfx DNA polymerase (Invitrogen, Groningen, the Netherlands) with human genomic DNA as a template. It was then subcloned between the XhoI and HindIII (Takara Shuzo Co., Shiga, Japan) recognition sites of pGL3-Basic (Promega, Madison, WI, USA). The primer sequences were as follows; 5′-ttctcgagTGAAAAAGATACAGAAGTTGA-3′ (hId1–5Xho) and 5′-ttaagcttTCTTGGCGACTGGCTGAAACA-3′ (hId1–3Hd) (lower cases indicate flanking sequences). Id1.3-luc, Id1.2-luc, Id1.0-luc and Id0.8-luc were made by digesting Id2.1-luc at the EcoRI, StuI, PstI and PvuII sites, respectively. Id985-luc, Id956-luc, Id916-luc, Id889-luc, Id985mutA-luc and Id985mutB-luc were generated by PCR using each specific upper primer and the hId1–3Hd primer with Id2.1-luc as a template. The primer sequences were as follows; 5′-ttctcgagCATGGCGACCGCCCGCGCGG-3′ (Id985), 5′-ttctcgagGACAGTCCGTCCGGGTTTTA-3′ (Id956), 5′-ttctcgagCCTGGCGTCTAACGGTCTGA-3′ (Id916), 5′-ttctcgagGTTCAGACGCTGACACAGAC-3′ (Id889), 5′-ttctcgagCATGGCGACCGTTTGCGCGG-3′ (Id985mutA), and 5′-ttctcgagCATGGCGACCGCCCGCGCTTTGCCA-3′ (Id985mutB) (lower case and underline indicate flanking sequences and mutations, respectively). To generate IdWT4F-luc, IdWT4R-luc and IdmutB4F-luc, the specific oligo DNAs of the 29 bp BRE with the wild-type or the mutB sequence flanked by six nucleotides were annealed, digested with XhoI and SalI, purified from agarose gels, and subcloned into pGL3-Promoter (Promega). The sequences of the oligo DNAs were as follows; 5′-tcgagCATGGCGACCGCCCGCGCGGCGCCAGCCTg-3′ (IdWT-S), 5′-tcgacAGGCTGGCGCCGCGCGGGCGGTCGCCATGc-3′ (IdWT-AS), 5′-tcgagCATGGCGACCGCCCGCGCTTTGCCAGCCTg-3′ (IdmutB-S) and 5′-tcgacAGGCTGGCAAAGCGCGGGCGGTCGCCATGc-3′ (IdmutB-AS) (lower cases and underlines indicate flanking sequences and mutations, respectively). PCR products were confirmed by DNA sequencing using a GeneRapid DNA sequencer (Amersham Pharmacia Biotech, Buckinghamshire, UK). Id985-EGFPd2 and Id985mutB-EGFPd2 were generated by replacing the coding regions of luciferase with a destabilized EGFP cDNA of pd2EGFP-C1 (Clontech Laboratories, Palo Alto, CA, USA) in Id985-luc and Id985mutB-luc, respectively.

Construction of Smad expression vectors

Complementary DNAs for the mouse Smad1, Smad3, Smad4 and Smad7 coding regions, except for the first Met residues, were obtained by an RT-PCR technique using Platinum Pfx DNA polymerase (Invitrogen) as previously described (Yanai et al. 2001). Total RNAs prepared from C2C12 myoblasts or mouse calvaria of the ddY strain were used for the RT-PCR. The primer sequences were as follows; 5′-ttaagcttAATGTGACCAGCTTGTTTTC-3′ (mS1–5Hd), 5′-tttctagaGACGGAAGCCACAGGTCTTT-3′ (mS1–3Xb), 5′-ttaagcttTCGTCCATCCTGCCCTTCAC-3′ (mS3–5Hd) and 5′-tttctagaCCCGCTCCCTTTACTCCTA-3′ (mS3–3Xb), 5′-aaaagcttGACAATATGTCTATAACAAA-3′ (mS4–5Hd), 5′-aatctagaAATGGTTAGGGCGTCCGTGG-3′ (mS4–3Xb), 5′-ttaagcttTTCAGGACCAAACGATCT-3′ (mS7–5Hd) and 5′-tttctagaTGTCCTCTTCTCCCCACC-3′ (mS7–3Xb) (lower cases indicate flanking sequences). PCR products were digested with HindIII and XbaI (Takara Shuzo Co.), purified from agarose gels, and subcloned into pFLAG-CMV-2 (Sigma-Aldrich, St Louis, MO, USA). Each cDNA was confirmed by DNA sequencing.

Cell culture, transfection and luciferase assay

Mouse C2C12 myoblasts and human MDA-MB468 breast cancer cells were obtained from the American Type Culture Collection, and mouse C3H10T1/2 (10T1/2) fibroblasts from the RIKEN Cell Bank (Tsukuba Science City, Ibaraki, Japan). C2C12 cells and its subclonal cell line, C2C12ΔIA12 cells, which are stably expressing a kinase domain-truncated BMPR-IA (Namiki et al. 1997), and 10T1/2 cells were maintained in Dulbecco's Modified Eagle's Medium (DMEM) containing 15% foetal bovine serum (FBS) (Sigma Chemical Co.). MDA-MB468 cells were maintained in Leibovitz's L-15 medium containing 10% FBS. Cells were inoculated at 1.5 × 104 cells/well in 96-well plates with growth media 1 day before transfection by Lipofectamine or Lipofectamine 2000 (Invitrogen). For normalization, pRL-SV40 plasmid (Promega) was co-transfected with luc reporter plasmids. The amount of each plasmid used for the transfection was fixed at 200 ng/well by appropriate empty plasmids. After an overnight incubation, the culture media were replaced with fresh media containing 2.5% FBS in the presence or absence of 300 ng/mL of BMP-2 (provided by Yamanouchi Pharmaceuticals, Tokyo, Japan), and then cultured for an additional 24 h. Luciferase activities in the cell extracts were determined using a dual luciferase assay kit (Toyo Ink, Tokyo, Japan) using a luminometer.

Both Id985-EGFPd2 and Id985mutB-EGFPd2 were transfected in C2C12 and C2C12ΔIA12 cells as described above, in place of the luciferase reporter plasmids. One day after transfection, the cells were incubated with fresh media containing 2.5% FBS in the presence or absence of 300 ng/mL of BMP-2. Expression levels of EGFP in the cultures were monitored under a fluorescence microscope (Olympus, Tokyo, Japan).

Electrophoresis mobility shift assay

EMSA was performed essentially as previously described (Katagiri et al. 1997). C2C12 cells were incubated for 1 h, with or without 300 ng/mL of BMP-2 or 1 µm PMA, and then nuclear extracts were prepared from the cells by the method of Schreiber et al. 1989) with a protease inhibitor cocktail (Roche Diagnostics, Mannheim, Germany). The double strand BRE probe was labelled with T4 polynucleotide kinase and [γ32P]ATP (New England Nuclear, Boston, MA) using a labelling kit (Takara Shuzo Co.). For the competition experiments, the cold double strand oligos were added to the mixtures with the labelled probe. The sequences of oligo DNAs we used were as follows; 5′-CATGGCGACCGCCCGCGCGGCGCCAGCCT-3′ (wild-type BRE), 5′-CATGGCGACCGTTTGCGCGGCGCCAGCCT-3′ (BRE mutA), 5′-CATGGCGACCGCCCGCGCTTTGCCAGCCT-3′ (BRE mutB), 5′-GGATCCAGGGGGGGCGAGCGGGGGCGA-3′ (Egr-1; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and 5′-ATTCGATCGGGGCGGGGGCAGC-3′ (Sp1; Santa Cruz Biotechnology). Six micrograms of the nuclear extracts were incubated with a labelled BRE probe with the wild-type sequence. For the supershift experiments, antibodies were added prior to the addition of the probe, and incubated for 15 min at room temperature. Antibodies against Smad1 (T-20), Smad3 (I-20), Smad4 (B-8), Smad6/Smad7 (N-19), Id1 (C-20), Egr-1 (C-19), Egr-2 (C-14), Egr-3 (C-24) and p300 (N-15) were obtained from Santa Cruz Biotechnology. The reaction mixture was loaded on to 5% polyacrylamide gel in 0.5 × TBE (44.5 mm Tris base, 44.5 mm boric acid and 1 mm EDTA) and resolved by electrophoresis.

Northern blot analysis

Northern blot analysis was performed as previously described (Katagiri et al. 1994). In brief, 20 µg of total RNA prepared with Trizol (Invitrogen) was resolved by electrophoresis in a 1.2% agarose-formaldehyde gel, and transferred on to a Hybond-N membrane (Amersham International, Amersham, UK). The membrane was sequentially hybridized with [α32P]-labelled cDNA probes for mouse Id1 and GAPDH. All of the inhibitors for MAPK (SB303580 and PD98059) and PI3K (wortmannin and LY294002) were purchased from Calbiochem, Darmstadt, Germany.

Acknowledgements

We thank Yamanouchi Pharmaceuticals, Japan for supplying the BMP-2, Dr E. Hara for generously providing Id1 cDNA, and M. Irie for technical help. This work was supported in part by a grant-in-aid from the Ministry of Education, Culture, Sports, Science and Technology of Japan, a special research grant for the development of characteristic education from the Promotion and Mutual Aid Corporation for Private Schools of Japan, a grand-in-aid from the Japan Space Forum, and a grant-in-aid for young researchers from the Kitasato University Alumni Association (to T.K.).

Ancillary