Transcriptional Activation of β-Tropomyosin Mediated by Serum Response Factor and a Novel Barx Homologue, Barx1b, in Smooth Muscle Cells*

Tropomyosin (TM) is a regulatory protein of actomyosin system. Muscle type-specific expression of TM isoforms is generated from different genes and by alternative splicing. β-TM isoforms in chicken skeletal and smooth muscles are encoded by a single gene and transcribed from the same promoter. We previously reported a smooth muscle cell (SMC) phenotype-dependent change in β-TM expression (Kashiwada, K., Nishida, W., Hayashi, K., Ozawa, K., Yamanaka, Y., Saga, H., Yamashita, T., Tohyama, M., Shimada, S., Sato, K., and Sobue, K. (1997) J. Biol. Chem. 272, 15396–15404), and identified β-TM as an SMC-differentiation marker. Here, we characterized the transcriptional machinery of the β-TM gene in SMCs. Promoter and gel mobility shift analyses revealed an obligatory role for serum response factor and its interaction with the CArG box sequence in the SMC-specific transcription of the β-TM gene in differentiated SMCs. We further isolated a novel homologue of the Barx homeoprotein family, Barx1b, from chicken gizzard. Barx1b was exclusively localized to SMCs of the upper digestive organs and their attached arteries and to craniofacial structures. Serum response factor and Barx1b bound each other directly, coordinately transactivated the β-TM gene in differentiated SMCs and heterologous cells, and formed a ternary complex with a CArG probe. Taken together, these results suggest that SRF and Barx1b are coordinately involved in the SMC-specific transcription of the β-TM gene in the upper digestive organs and their attached arteries.

The vertebrate muscular system consists of skeletal, cardiac, and smooth muscles, all of which are derived from the mesoderm. The molecular basis for muscle contraction is the actomyosin system, which converts the chemical energy of ATP into mechanochemical force (1). However, the actomyosin system is differentially regulated depending on the presence of different contractile proteins (2)(3)(4), whose expression patterns are muscle-type dependent. Therefore, these contractile proteins are considered to be muscle-specific molecular markers. Despite intensive study in this area, only one general concept concerning the transcription of muscle-specific marker gene has been agreed upon. Serum response factor (SRF) 1 and its DNA binding sequence, CArG box (CC(A/T) 6 GG), are thought to comprise one of the core machineries for the muscle-specific transcription of the skeletal ␣-actin (5), caldesmon (6), cardiac ␣-actin (7), ␣ 1 -integrin (8), SM22␣ (9), telokin (10), smooth muscle myosin heavy chain (11), smooth muscle ␣-actin (12), calponin (13), and desmin (14) genes. SRF was originally identified as a transcription factor of the c-fos gene (15). It is expressed at high levels in all three muscles (16), and at a far lower level in non-muscle cells (15). A recent study using SRF-deficient mice revealed the misformation of mesoderm leading to fetal death at around embryonic day 12.5; no skeletal, cardiac, and smooth muscle ␣-actin transcripts were detected in these mice (17). These findings suggest the possible involvement of co-activators in the SRF and CArG interaction that confer the muscle specificity on the transactivation of muscle-specific marker genes. One set of candidates for such co-activators includes members of the homeoprotein family. Grueneberg et al. (18) first reported a paired class homeoprotein, Phox1, which enhances the DNA binding ability of SRF (18). In addition, two NK family homeoproteins have been reported as co-activators of SRF. Nkx-2.5/Csx and SRF activate the cardiac ␣-actin transcription (7) and Carson et al. (19) reported that Nkx-3.1 potentiates the transcription of the smooth muscle ␥-actin gene with SRF.
Tropomyosin (TM), which is a regulatory protein of the actomyosin system, has several isoforms that are generated from different genes and by alternative splicing in a muscle-specific context (reviewed in Ref. 20). Three distinct chicken TM genes, ␣-TM, ␤-TM, and cardiac-TM, have been identified (20). In chicken smooth muscle cells (SMCs), TM forms an ␣/␤ heterodimer. We reported previously that ␣-TM exhibits a change in its isoform from a smooth muscle-type (␣-TMsm) to a fibroblast one (␣-TM-F1 and ␣-TM-F2), and that the expression of smooth muscle-type ␤-TM is down-regulated at the mRNA and protein levels during the de-differentiation of vascular and visceral SMCs. Consequently, ␣and ␤-TMs are also regarded as SMC-differentiation molecular markers (21). The chicken ␤-TM gene contains muscle and non-muscle promoters, and both smooth (␤-TMsm) and skeletal (␤-TMsk) muscle isoforms are transcribed from the same muscle promoter (22). Furthermore, ␤-TMsm and ␤-TMsk are generated by alternative splicing and contain exons 6a/9d and exons 6b/9a, respectively (20). Toutant et al. (23) reported the E, C, and CArG boxes as putative cis-elements of the ␤-TM muscle promoter in skeletal myoblasts. However, there are no other reports concerning the tissue-specific transcription of the ␤-TM gene.
Here, we investigated the SMC-specific transcription of the ␤-TM gene and demonstrated that CArG box, but not other cis-elements, is critically involved in the SMC-specific transcription. We further cloned a novel homologue of Barx1 homeoprotein family, Barx1b, as a partner of SRF. Barx1b is predominantly expressed in the SMC layer of the upper digestive organs, their attached arteries, and in craniofacial structures. In visceral SMCs and heterologous cells, SRF and Barx1b coordinately transactivated the ␤-TM gene. Gel mobility shift analysis showed a ternary complex between the SRF/ CArG probe and Barx1b, and immunoprecipitation analysis revealed a DNA-independent interaction between SRF with Barx1b in vivo.
Antibodies-An anti-Barx1b antibody was purified from anti-GST-Barx1b-C (amino acids 195-247) rabbit serum by the sequential application of GST and GST-Barx1b-C affinity columns. Anti-human SRF polyclonal, anti-GST polyclonal, and anti-Myc monoclonal antibodies were purchased from Santa Cruz Biotechnology. An anti-FLAG M2 antibody was obtained from Sigma.
Cell Culture and Transfection-Differentiated chicken gizzard SMCs were cultured as described previously (26). Mouse C3H10T1/2 fibroblasts grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum were transfected for 48 h with 1 g of ␤-galactosidase control plasmid and 1.5 g of luciferase reporter plasmid with or without SRF/Barx1b expression vectors using TransIT LT1 lipofection reagent (Pan Vera Corp.). The cell lysates were measured for luciferase and ␤-galactosidase activities. The transfection efficiencies were normalized to the ␤-galactosidase activity.
Purification of GST-SRF and Barx1b-GST-SRF or GST-Barx1b derivatives were transformed into Escherichia coli BL21. Protein expression was induced by 1 mM isopropyl-␤-D-thiogalactopyranoside. The cells were resuspended with lysis buffer (50 mM Tris-HCl, pH 8.3, 500 mM NaCl, 1 mM EDTA, 2 mM dithiothreitol, and 50 g/ml phenylmethylsulfonyl fluoride) and lysed with a French pressure cell press (SIM-AMINCO). The recombinant proteins were purified with glutathione-Sepharose 4B gels. The purity and quantity of the recombinant proteins were determined by SDS-PAGE after Coomassie blue staining.
Pull-down Assay-A series of GST-SRF proteins were purified and coupled to glutathione-Sepharose 4B gels. Barx1b derivatives were translated in vitro using TNT-quick coupled transcription and translation system kit with [ 35 S]methionine (Promega). The binding reaction was carried out with 2 g of GST-SRF and 10 l of in vitro translated Barx1b in 500 l of binding buffer (20 mM Tris-HCl, pH 7.5, 80 mM KCl, 10% glycerol, 0.05% Nonidet P-40, 1 mM EDTA, and 5 mM MgCl 2 ) at 4°C for 2 h. GST-SRF bound Sepharose was washed three times, treated with 2% SDS sample buffer, and analyzed by SDS-PAGE. Labeled proteins were visualized by autoradiography.
Gel Mobility Shift Assay-Oligonucleotide probes were subcloned into pBluescript SK(ϩ) (Stratagene) and excised by BamHI and HindIII digestion. The probes were labeled with [ 32 P]dCTP and purified with QIA quick nucleotide removal kit (Qiagen). Nuclear extracts from chicken gizzards at embryonic day 15 (E15) were prepared by the method of de Jonq et al. (27). Chicken gizzard nuclear extracts and the labeled ␤-TM CArG probe were incubated in 10 l of binding buffer (20 mM Tris-HCl, pH 8.0, 75 mM KCl, 0.5 mM EDTA, 0.05% Nonidet P-40, 50 g/ml bovine serum albumin, 5 mM dithiothreitol, 5% glycerol, and 25 g/ml herring sperm DNA). Purified fusion protein(s) and the labeled ␤-TM CArG probe were incubated in 10 l of binding buffer (20 mM HEPES-KOH, pH 7.9, 75 mM KCl, 0.05% Nonidet P-40, 50 g/ml bovine serum albumin, 5 mM dithiothreitol, 5% glycerol, and 1 g/ml herring sperm DNA) at room temperature for 20 min, then the mixtures were loaded onto a 4% polyacrylamide gel. Electrophoresis was carried out under constant current of 10 mA in 0.5 ϫ TBE buffer. The gel was dried and visualized by autoradiography.
In Situ Hybridization-Whole chicken embryos or gizzards at E15 were embedded with Tissue-Tek (SAKURA) and cut into 12-15-m thick by cryostat (Bright, Huntingdon, United Kingdom). The sections were fixed and hybridized with cRNA probes as described previously (8). The fluorescent signal was detected using the TSA Plus DNP AP system (PerkinElmer Life Sciences Inc.) and HNPP/Fast Red TR (Roche Molecular Biochemicals) according to the manufacturer's recommendation. A ␤-TM cRNA probe encoding exons 1a and 2, which are common to both the smooth and skeletal muscle isoforms, was used.
Immunoprecipitation-10T1/2 cells were seeded at a density of 9 ϫ 10 5 cells per 10-cm plate, and transfected with 2 g of pcDNA3.1-wt-SRF-FLAG and 2 g of pCS2ϩMT-wt-Barx1b. Transfected cells were washed three times with phosphate-buffered saline and lysed with 600 l of IP buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 100 mM KCl, 0.5 mM EDTA, 1 mM dithiothreitol, 0.5% Nonidet P-40, 50 g/ml phenylmethylsulfonyl fluoride, 2 g/ml pepstatin, and 2 g/ml aprotinin). Cell lysates were incubated with 200 g/ml DNase I and 10 g/ml RNase A for 30 min at 26°C, and centrifuged at 100,000 ϫ g for 15 min. The supernatant (250 l) was rocked with 2.5 g of anti-Barx1b polyclonal or 2 g of anti-SRF polyclonal antibody for 12 h at 4°C, then protein A-Sepharose gels (Amersham Pharmacia Biotech) were added. After 6 h incubation, the gels were washed three times and treated with 2% SDS sample buffer. Samples were separated by SDS-PAGE and subjected to Western blotting with mouse anti-FLAG M2 or anti-Myc monoclonal antibody. Target proteins were detected with peroxidase linked anti-mouse IgG (Amersham Pharmacia Biotech) and Supersignal West reagent (Pierce).

SRF-CArG-dependent Transcription of the ␤-TM Gene in
Chicken Gizzard SMCs-To elucidate the essential cis-elements of the ␤-TM gene, we cloned the chicken ␤-TM promoter region, and analyzed its promoter activity in differentiated gizzard SMCs (Fig. 1A). A series of deleted reporter genes (BTM1272-BTM117) showed only a slight reduction in their promoter activities, even when the E and C boxes were deleted (BTM117). In contrast, loss of the CArG box (BTM81) resulted in a marked decrease in the promoter activity. Mutation of the CArG box (CArG box MUT), but not the E and C boxes (E box MUT and C box MUT), in BTM230 also markedly decreased the promoter activity. Gel mobility shit analysis was then performed using nuclear extracts prepared from differentiated gizzard SMCs and 32 P-labeled ␤-TM CArG probe (Fig. 1B). The labeled CArG probe and nuclear protein complex (Fig. 1B, lane 1), which was displaced by an excess amount of unlabeled probe (Fig. 1B, lane 2), was supershifted with anti-SRF antibody (Fig.  1B, lane 4). These results suggested that CArG box is the critical cis-element of the ␤-TM gene in visceral SMCs, and that SRF serves as a trans-factor of the CArG box.
Chicken Barx1b, A Novel Homologue of Barx Homeoprotein Family-To elucidate the mechanism of SMC-specific transcription, we screened the homeobox genes from a gizzard SMC cDNA library and cloned the gax, hox cluster (hoxA5 and hoxB3-B5), and a novel clone of Barx family genes. This clone was 1,170 base pairs in length and encoded a single open reading frame of 247 amino acids ( Fig. 2A). Barx1 cDNAs have been recently isolated from mouse (28), human (29), and chicken (30). The novel clone shows 82-84% total identity with known Barx1 homologues. The homeodomain of this clone is completely identical with that of other Barx1 family members, including unique Thr and Tyr residues in the third helix. It also shares four leucine repeats and a Barx-specific domain at the COOH terminus ( Fig. 2A). On the other hand, the novel clone has a long NH 2 -terminal sequence and shows a substitution of 13 amino acids in the NH 2 terminus (amino acids 102-130) and 10 amino acids in the COOH terminus (amino acids 228 -247) compared with other Barx1 family members. Based on these findings, we called this novel homologue chicken barx1b. In addition, we noticed that there is a sequence similarity between the Barx1b and Nkx-2.5 families (see "Discussion"). They share the consensus sequence of the FIL domain (31) at their NH 2 terminus, and contain basic residue-rich sequence around the NH 2 -terminal boundary of the homeodomain (Fig.  2, B and C).
Expression of Barx1b in Chicken Embryos-The expression patterns of the cloned homeobox genes, including barx1b in E15 chicken embryos, were compared with those of ␤-TM and SRF by in situ hybridization. The Barx1b mRNA was exclusively expressed in upper digestive tissues such as the esophagus, crop, and gizzard, and in craniofacial structures (Fig. 3C). gax was expressed in all muscle lineages, and hox genes were ubiquitously expressed (data not shown). The ␤-TM mRNA was seen in all smooth and skeletal muscle tissues, but was faint in cardiac muscle (Fig. 3A). The SRF mRNA was ubiquitous but prominent in all three muscle types (Fig. 3B). The ␤-TM, SRF, and Barx1b mRNAs in gizzard were then precisely localized by fluorescent in situ hybridization. The ␤-TM and Barx1b mRNAs were only seen in the SMC layer of the gizzard (Fig. 3, D and F), but the SRF mRNA was detected from the glandular to the SMC layers (Fig. 3E). Signals of all three transcripts were also seen in the adjacent arteries (Fig. 3, D-F). As a control, we performed the same analysis using sense probes, and found no signals in the E15 whole embryo (data not shown). Northern blotting revealed that the expressions of Barx1b and SRF mRNAs in gizzards preceded that of ␤-TM mRNA and were most prominent at E10 -13. In addition, the expression patterns of Barx1b, SRF, and ␤-TM were sustained after hatching. The ␤-TM mRNA was up-regulated during development of the gizzard (data not shown).
Coordinated Activation of the ␤-TM Promoter with SRF and Barx1b-We next investigated the possible involvement of Barx1b in the SRF-dependent activation of the ␤-TM promoter. The ␤-TM reporter gene (BTM117), which contains a CArG box, was co-transfected with SRF and/or Barx1b expression vectors in differentiated gizzard SMCs (Fig. 4). When SRF or Barx1b was transfected alone, the activation rates of the reporter gene were doubled. However, co-transfection with Barx1b and SRF showed an approximately 5-fold increase. The co-transfection of other homeoproteins, Gax or MHox, with SRF, which was done as a control, failed to show the coordinated activation of the reporter gene. When a mutation was introduced into the CArG box of the reporter gene (CArG MUT), no significant activation was observed even in the presence of SRF and Barx1b (Fig. 4, right four columns). Taken together, these results indicate that SRF and Barx1b coordinately transactivate the ␤-TM gene through the CArG box in differentiated SMCs.
Interaction between SRF and Barx1b-The interacting domains of SRF and Barx1b were mapped by a pull-down assay. 35 S-Labeled Barx1b derivatives were incubated with GST-SRF immobilized on glutathione-Sepharose gels (Fig. 5A). Wild-type Barx1b (wt) and its COOH terminus deletion (⌬C) bound to SRF, but its NH 2 terminus deletion (⌬N) did not. Interestingly, deletion mutants composed 31 or 11 residues of the NH 2 -ter- minal arm plus the homeodomain (31ϩHD or 11ϩHD) retained their binding activity with SRF, but the homeodomain (HD) alone or the homeodomain deletion (⌬HD) lacked the activity. These results suggest the critical involvement of 11 residues of the NH 2 -terminal arm plus the homeodomain of Barx1b in its interaction with SRF. An essential domain of SRF required for Barx1b binding was also determined (Fig. 5B). SRF derivatives fused with GST were incubated with the 35 S-labeled minimum binding domain of Barx1b (11ϩHD). Wild-type (wt) and the MADS domain (MADS) of SRF interacted with Barx1b, but the NH 2 (N) and COOH (C) termini of SRF did not. Thus, the domains of Barx1b and SRF essential for their physical interaction were mapped to 11 residues of the NH 2 -terminal arm plus the homeodomain of Barx1b and the MADS domain of SRF. To confirm the direct interaction between SRF and Barx1b in vivo, FLAG-tagged SRF and Myc-tagged Barx1b were co-expressed in 10T1/2 cells and immunoprecipitated (Fig. 5C). The proteins precipitated with anti-SRF or anti-Barx1b antibodies were analyzed by Western blotting using anti-FLAG (Fig. 5C, lanes 1-4) or anti-Myc (Fig. 5C, lanes 5-8) antibodies. FLAG-SRF and Myc-Barx1b were successfully coimmunoprecipitated with anti-Barx1b and anti-SRF antibodies, respectively (Fig. 5C, lanes 4 and 7).
We then examined whether the minimum interaction domain of SRF and Barx1b are actually involved in ␤-TM transcription in 10T1/2 fibroblasts (Fig. 5D). Expression of wt-Barx1b or SRF alone showed a 2-or 5-fold increase in the promoter activity, respectively. Co-expression of SRF and wt-Barx1b, however, showed a striking increase in activity about 17-fold. Consistent with the results of pull-down assays (Fig. 5,  A and B), SRF and Barx1b-⌬C or Barx1b-11ϩHD, which directly binds to SRF, coordinately activated the ␤-TM promoter by 16-or 13-fold, respectively. Barx1b-⌬N, Barx1b-HD, or Barx1b-⌬HD, which lack SRF binding activity, failed to potentiate the promoter activity, even when SRF was present. This coordinated activation was not seen with an SRF mutant (SR-Fpm) that lacks DNA binding ability.
Ternary Complex Formation between SRF, the CArG Box, and Barx1b-To study the interaction of SRF and Barx1b with the ␤-TM cis-element, gel mobility shift analysis was performed using 32 P-labeled putative cis-elements. In this experiment, we used Barx1b-⌬C, which fully retained the SRF binding ability and the ␤-TM transcription activity with SRF. To reveal the binding site(s) in the ␤-TM promoter region for Barx1b, we examined the interactions of putative homeoprotein binding AT-rich sequences (Ϫ1272 to Ϫ1242, Ϫ1270 to Ϫ1040, Ϫ431 to Ϫ401, and Ϫ437 to Ϫ417) with Barx1b (24). None of these 32 P-labeled AT-rich probes, however, bound to GST-Barx1b-⌬C. The same result was obtained using GST-Barx1b-wt (data not shown). This finding was consistent with promoter analyses, which showed that no deletions of AT-rich sequences (BTM1013, BTM785, and BTM230) affected the ␤-TM promoter activity (Fig. 1A). Furthermore, mutation of the CArG box efficiently abolished the transactivation of the ␤-TM gene with SRF and/or Barx1b (Fig. 4). These results suggest that Barx1b acts as a DNA-independent co-activator of SRF, and that it requires the SRF and CArG interaction as a base foothold. Finally, we examined the interaction between SRF, the 32 P-labeled CArG probe, and Barx1b (Fig. 6). SRF bound to the labeled CArG probe in a dose-dependent manner (Fig. 6A,  lanes 6 and 7, arrowhead) and the SRF-DNA complex was supershifted with anti-SRF antibody (Fig. 6A, lane 9, two asterisks). In contrast, GST-Barx1b-⌬C alone did not bind to the labeled CArG probe (Fig. 6A, lanes 2-5). The same result was obtained using GST-Barx1b-wt (data not shown). When both SRF and GST-Barx1b-⌬C were incubated with the labeled CArG probe, a band shift was observed (Fig. 6B, lanes 3 and 4,  arrow). Because this slowly migrating band was supershifted with anti-GST or anti-SRF antibodies (Fig. 6B, lanes 5 and 6,  two asterisks), the band was identified as a ternary complex between SRF, CArG probe, and Barx1b. In addition, as GST-Barx1b-⌬C amounts were increased, free probes were gradually reduced (Fig. 6B, lanes 3 and 4, asterisk). No ternary complex was observed, when GST was incubated with SRF and CArG probe (data not shown). DISCUSSION Here, we demonstrated the SMC-specific transcription of the ␤-TM gene. Promoter analyses using a series of deleted and mutated ␤-TM reporter genes and gel mobility shift assays revealed an obligatory role of the SRF and CArG interaction for ␤-TM expression in SMCs (Fig. 1). We also isolated a novel homologue of the Barx family, Barx1b, and found it to act as a co-activator of SRF (Figs. 2 and 4). Barx1b was exclusively localized to the SMC layer of the upper digestive organs and its attached arteries and in craniofacial structures (Fig. 3). We further demonstrated that SRF and Barx1b, which directly bound each other, coordinately transactivated the ␤-TM gene (Figs. 4 and 5), and formed a ternary complex with a CArG probe (Fig. 6).
Although the transcription of muscle-specific genes has been extensively studied, no general mechanism explaining musclespecific transcription has been revealed except for the SRF and CArG box interaction. SRF is dominantly expressed in all three muscle types, which are derived from mesoderm ( Fig. 3B and Ref. 16) and is involved in muscle-specific gene transcription. However, SRF also plays a vital role in the expression of immediate-early genes (32). These diverse roles of SRF might be regulated by different co-activators for different genes. In fact, ternary complex factor recruits SRF and potentiates the transactivation of the c-fos gene (32). Therefore, it is likely that other co-activators of SRF confer muscle-specific transcription. Currently, candidates for such co-activators include members of the homeoprotein family. Grueneberg et al. (18) cloned PHOX1 from a human glioblastoma cDNA library, and found it to enhance the affinity of SRF for DNA. Mhox, a murine homologue of PHOX1, is widely distributed in mesoderm-derived cells, including those of skeletal and smooth muscles and the heart (33). Hautmann et al. (34) reported that the angiotensin-II-stimulated smooth muscle ␣-actin transcription is mediated by SRF and MHox. However, whether there is a physical interaction between SRF and MHox and the details of their involvement in coordinated transactivation remains unclear. Nkx-2.5/Csx, a member of the NK2 family, was originally cloned from a mouse heart cDNA library as a mammalian homologue of Drosophila tinman, which is a critical transcrip- tion factor for cardiogenesis in the fly (35)(36)(37). Chen et al. (7) reported a physical interaction between Nkx-2.5/Csx and SRF and their coordinated transactivation of the cardiac ␣-actin gene. Recent genetic analysis showed that NKX-2.5/CSX is also involved in development of the septum and the atrioventricular conduction in humans (38). However, no expression of Nkx-2.5/Csx is observed in a smooth muscle lineage. In Drosophila, bagpipe controls the midgut formation under the control of tinman (35). bagpipe belongs to the NK3 family, and two mammalian homologues, Nkx-3.1 and Nkx-3.2, have been recently reported (39,40). Carson et al. (19) reported the coordinated activation of the smooth muscle ␥-actin gene by SRF and Nkx-3.1. However, Nkx-3.1 is not expressed in the muscular layer of smooth muscle tissues (39). Thus, other homeoproteins are likely to be involved in the SMC-specific transcription.
Several studies have reported the isolation of homeobox genes from smooth muscle tissues. Two independent groups cloned the Hox cluster genes from rat and human aorta cDNA libraries (41,42), but the possible involvement of such Hox cluster genes in the transactivation of SMC marker genes remains unknown. Walsh et al. (43) isolated Gax from cultured rat aortic vascular smooth muscle cells. Gax is a homeobox gene expressed in all three muscle lineages, and its expression is rapidly down-regulated during the G 0 to G 1 transition in vascular smooth muscle cells (43). In this study, we screened a chicken gizzard cDNA library using degenerative probes targeted against a most conserved third helix region of the homeodomain and cloned six homeobox genes; barx1b, gax, and hox cluster genes (hox A5, B3, B4, and B5). Of these, Barx1b mRNA exclusively localized to the SMC layer of the upper digestive organs, their attached arteries, and to craniofacial structures (Fig. 3, C and F). The distribution of Barx1b completely overlapped that of ␤-TMsm mRNA in the SMC layer of gizzards and their attached arteries (Fig. 3, D and F). At present, barx1b is the only known homeobox gene whose expression is restricted to SMCs.
The Bar family was first identified in the Drosophila Bar locus (BarH1 and BarH2) (44). A vertebrate homologue of the Drosophila Bar family, Barx, is composed of Barx1 and Barx2 (28,45). The homeodomain of Barx1 is completely conserved among species (Fig. 2A) and is 87% identical to those of the Barx2 family (45). The expression pattern of Barx1b is similar to that of other Barx1 family members, and includes the craniofacial structures, stomach, and limb bud (28,30). Barx1b and  (25 ng) and SRF or SRFpm (12.5 ng) were co-transfected with BTM117 reporter plasmid (1.5 g) and pSV-␤-galactosidase control vector (1.0 g) in mouse C3H10T1/2 fibroblasts. Raw luciferase activities were normalized to the ␤-galactosidase activity and represented as fold activation relative to BTM117 alone. SRF wild type, wt; and mutant, pm.
chicken Barx1, which was cloned from a chicken head cDNA library, show completely different sequences at their NH 2 and COOH termini ( Fig. 2A). Barx1b has a long NH 2 terminus compared with other Barx1 family members, therefore, it is a novel homologue of the Barx1 family. Saito et al. (31) defined a characteristic domain consisting of Phe, Iso, and Leu at the NH 2 terminus of the Bar family, and named this sequence a "FIL" domain. Chicken Barx1b also contains a FIL domain at its NH 2 terminus (Fig. 2B). It has been reported that NK2 family members share a "TN" domain at their NH 2 terminus (reviewed in Ref. 46). Ranganayakulu et al. (47) reported that a cardiogenic domain of tinaman is mapped to NH 2 -terminal amino acids 1-52, which include the TN domain. Interestingly, we noticed that the FIL consensus sequence is also present in the TN domain of the NK2 family proteins. Furthermore, the Barx1b homeodomain shows about 70% similarity with that of the NK2 family. In addition to these sequence similarities, both Barx1b and Nkx-2.5/Csx show unique tissue-restricted expression patterns, respectively. Therefore, there might be some cross-relationships between the Barx and NK2 families.
We investigated the direct interaction between Barx1b and SRF using pull-down (Fig. 5, A and B), immunoprecipitation (Fig. 5C), and gel mobility shift (Fig. 6) analyses. The results demonstrated that Barx1b and SRF bind to each other, and their binding involves 11 residues of the NH 2 -terminal arm plus the homeodomain of Barx1b and the MADS domain of SRF, and is independent of DNA (Fig. 5, A and B). A region around the NH 2 -terminal boundary of the Barx1b homeodomain contains an Arg/Lys-rich sequence and is conserved among all Barx1 family members ( Fig. 2A). Interestingly, NK2 family also contains these basic residues around the NH 2 boundary of the homeodomain. Chen et al. (7) reported that Nkx-2.5/Csx requires the NH 2 -terminal arm plus the homeodomain for SRF binding. Thus, the basic residues around the NH 2 terminus of the homeodomain of Barx1b (and possibly all Barx1 family members) and of Nkx-2.5/Csx might be critical for SRF binding. Our gel mobility shift assay revealed the involvement of Barx1b in a ternary complex with SRF and CArG box (Fig. 6). Although Barx1b alone did not directly bind to ␤-TM CArG probe, it increased the DNA binding ability of SRF, resulting in formation of a ternary complex. The same DNA independence is observed with Nkx-2.5/Csx. Nkx-2.5/Csx does not require its own DNA binding ability for the transactivation of the cardiac ␣-actin promoter when it is co-expressed with SRF (7). Thus, Barx1b and Nkx-2.5/Csx seem to function similarly, but they are also different in that the COOH terminus of Nkx-2.5/Csx serves as an inhibitory domain for the cardiac gene transcription (48), but Barx1b-wt or Barx1b-⌬C alone did not significantly change the transactivation of the ␤-TM gene (Fig. 5D). Lastly, we identified the essential domain of Barx1b (11ϩHD) for SRF binding that was sufficient for the transactivation of the ␤-TM gene (Fig. 5D). Barx1b markedly increased the ␤-TM promoter activity only in the presence of SRF (Figs. 4 and 5D). In addition, the coordinated transactivation was abolished by the overexpression of an SRF mutant that lacks DNA binding ability (Fig. 5D). We also examined the effect of other homeoproteins, Gax and MHox, on the ␤-TM promoter. None of them, however, showed a similar cumulative transactivation with SRF (Fig. 4). Thus, the coordinated transactivation of the ␤-TM promoter with SRF is a characteristic function of Barx1b, and other mesodermally restricted homeoproteins lack this activity.
Recently, GATA family transcription factors have been shown to take part in muscle gene regulation, especially in the heart. At present, there are six known GATA proteins (GATA-1 to -6), which are divided into two subgroups. One includes GATA-1, -2, and -3, which are important for the differentiation of hematopoietic cells (49), and the other consist of GATA-4, -5, and -6, which are expressed in variety of tissues, including the heart, gut, lung, and urogenital tracts (50). Among the latter group, GATA-4 is restrictively expressed in the heart and potentiates cardiac muscle genes with Nkx-2.5/Csx (48) or SRF (51). GATA-6 is also expressed in the heart, but cannot substitute for GATA-4 in interacting with Nkx-2.5/Csx (52). Among the GATA family members, only GATA-6 is known to be expressed in vascular SMCs (53), but no one has reported its involvement in the transcription of SMC genes. In a preliminary experiment, we found that co-expression of GATA-6 with SRF and Barx1b potentiated the ␤-TM promoter activity even more than SRF and Barx1b alone (data not shown). However, GATA-6 is expressed not in the SMC layer, but in the glandular layer of the gizzard (data not shown). These findings suggest that there may be another novel GATA family protein in visceral SMCs.
In this study, we showed the SMC-specific transcription of the ␤-TM gene by a ubiquitous transcription factor, SRF, and an SMC-restricted homeoprotein, Barx1b. Barx1b is highly expressed in the SMC layer of the upper digestive organs and their adjacent arteries, but in neither lower digestive organs nor the aorta (Fig. 3C). It is well known that there are at least two distinct SMC populations in the aorta (54). SMCs in the aortic arch and the upper thoracic aorta are derived from neural crest cells, and thus are of ectodermal origin. In contrast, SMCs in the abdominal aorta are derived from lateral plate mesoderm. The fact that Barx1b is expressed in the SMC layer of the upper digestive organs and their adjacent arteries, but is absent in the aorta, indicates that homeoproteins might play a role in tissue-specific gene regulation in a homeotic fashion. It is possible that other unknown homeoproteins are expressed in SMCs of the aorta, lower digestive organs, and visceral organs, and act as co-activators of SRF. So far, the muscle-specific transcription had been considered in a lineagedependent context, but there would also be genetic hierarchies in other tissues. In fact, MyoD family had been thought to be a master transcription factor determining the skeletal muscle lineage. Tajbakhsh et al. (55) reported the agenesis of body wall muscles in Pax-3/Myf-5-deficient mice. The skeletal muscles in the head were, however, not affected in these mice. Their findings indicate that MyoD expression is definitely controlled by the homeoprotein Pax-3 in the trunk, but that there is another Pax-3-independent pathway in the head. Our findings also suggest the involvement of homeoproteins in tissue-dependent gene transcription within the SMC lineage.
Addendum-While this article was under review, a report by Herring et al. (56) appeared. The authors reported that Barx2b physically interacts with SRF and enhances DNA binding affinity of SRF. In addition, Barx2b localized in smooth muscle, skeletal muscle, and brain, but did not transactivate the telokin promoter with SRF.