Smooth Muscle Cell Phenotype-dependent Transcriptional Regulation of the α1 Integrin Gene*

The expressional regulation of chicken α1 integrin in smooth muscle cells was studied. The α1 integrin mRNA was expressed developmentally and was distributed dominantly in vascular and visceral smooth muscles in chick embryos. In a primary culture of smooth muscle cells, α1 integrin expression was dramatically down-regulated during serum-induced dedifferentiation. Promoter analyses revealed that the 5′-upstream region (−516 to +281) was sufficient for transcriptional activation in differentiated smooth muscle cells but not in dedifferentiated smooth muscle cells or chick embryo fibroblasts. Like other α integrin promoters, the promoter region of the α1 integrin gene lacks TATA and CCAAT boxes and contains binding sites for AP1 and AP2. The essential difference from other α integrin promoters is the presence of a CArG box-like motif. Deletion and site-directed mutation analyses revealed that the CArG box-like motif was an essential cis-element for transcriptional activation in differentiated smooth muscle cells, whereas the binding sites for AP1 and AP2 were not. Using specific antibodies, a nuclear protein factor specifically bound to the CArG box-like motif was identified as serum response factor. These results indicate that α1 integrin expression in smooth muscle cells is regulated transcriptionally in a phenotype-dependent manner and that serum response factor binding plays a crucial role in this regulation.

Integrins, heterodimeric transmembrane proteins, consist of an ␣ and a ␤ subunit and play a role in cell-to-cell and cell-toextracellular matrix adhesion as well as intracellular signal transduction (1). Cell adhesion mediated by integrins is important for cell differentiation, proliferation, and migration. There are at least 15 ␣ and 8 ␤ subunits, and their combinations generate various integrins that are distributed widely in a variety of cells. It has been generally accepted that the ␣ subunit dictates ligand binding specificity and that the cytoplasmic domain of the ␤ subunit, which interacts directly with the cytoskeleton, is involved in signal transduction from the extracellular matrix to the cytoplasm. Among variations of the integrin family, ␣1␤1 integrin is a receptor for both laminin (2) and collagens (3). During quail embryogenesis, the expression of ␣1 integrin is curious; ␣1 integrin is expressed transiently in nervous tissues and skeletal and cardiac muscle cells of early embryos, whereas its expression is increased dramatically in smooth muscle tissues as development proceeds (4). At a late stage of embryogenesis and afterward, ␣1 integrin expression comes to be restricted in visceral and vascular smooth muscle cells (SMCs) 1 and microvascular endothelium, whereas it is absent from most epithelial tissues. The down-regulation of ␣1 integrin expression has been reported in cultured SMCs under serum-stimulated conditions (5,6) and in some leiomyosarcomas (7); therefore, ␣1 integrin expression is closely associated with a phenotype of SMC.
Transcriptional machineries of integrin genes such as ␣2 (23), ␣4 (24 -26), CD11a (␣L) (27), and CD11c (␣X) (28) have been well studied. In general, the integrin promoters contain multiple Sp1, AP1, and/or AP2 binding sites, but not TATA and CAAT boxes, as cis-elements. In the case of ␣1 integrin, its expression depends upon a phenotype of SMC as described above. It is therefore worthwhile to investigate an SMC-specific transcriptional regulation of the ␣1 integrin promoter. In this study, we attempted to understand how ␣1 integrin expression is regulated in SMCs. As a first step, we cloned full-length cDNA encoding chicken ␣1 integrin and found that ␣1 integrin expression during phenotypic modulation of SMC is regulated at the mRNA level. We further characterized the 5Ј-upstream region of the ␣1 integrin gene and identified a CArG box-like motif as a cis-element involved in the SMC phenotype-dependent transcriptional regulation.

MATERIALS AND METHODS
cDNA and Genomic DNA Cloning-A cDNA library of gizzard SMCs from 15-day-old chick embryos was constructed in gt11 and was screened by plaque hybridization with 32 P-labeled probes derived from a partial cDNA fragment encoding the I-domain of chicken ␣1 integrin (GenBank U1041). To confirm the sequences of 5Ј-and 3Ј-portions, rapid amplification of cDNA end (RACE) was introduced. Genomic clones carrying the 5Ј-upstream regions of the chicken ␣1 integrin gene were isolated from a chicken genomic DNA library constructed in DASHII (29) using a 32 P-labeled cDNA subclone, C3b (Fig. 1A), as a probe. The sequences of cDNA and genomic clones were determined.
Expression of the Cloned Full-length ␣1 Integrin cDNA in Cultured SMCs-The full coding region of the ␣1 integrin cDNA was amplified by polymerase chain reaction (PCR) using a cDNA clone, C8-1 (approximately 5 kbp in length) (Fig. 1A), as a template, and the amplified cDNA was inserted into the downstream of the SR␣ promoter in pc-DLSR␣296 (30). The plasmid thus obtained was expressed transiently in dedifferentiated gizzard SMCs for 24 h as described below. The cells were lysed with Laemmli sample buffer under nonreducing conditions and were analyzed by Western blotting using anti-chicken ␣1 integrin polyclonal antibodies (31).
Genomic Southern Blotting-Genomic Southern blotting was performed as described previously (29). Briefly, genomic DNA was isolated from chick embryo gizzards and was digested with the indicated restriction enzymes. The DNA fragments were separated on 0.7% agarose gels and transferred to a nylon membrane (Hybond-N ϩ , Amersham) according to the method of Southern. A 32 P-labeled DNA fragment, expanding from ϩ103 to ϩ306 in the 5Ј-upstream sequence (see Fig. 4A, probe E1), was used as a probe. Hybridization and washing were carried out under the same conditions as described under "Northern Blotting." Northern Blotting-Total RNAs were extracted from 15-day-old chick embryo gizzard, aortic medial muscle layers, and primarily cultured SMCs and chick embryo fibroblasts (CEFs) using an ISOGEN RNA extraction kit (Nippon Gene, Japan). ␣1 integrin cDNA fragments, expanding from 575 to 1975 (Fig. 1A, probe N1) and expanding from Ϫ177 to 27 (Fig. 1A, probe N2), were amplified by PCR, and they were 32 P labeled on the antisense strands and used as probes. CaD cDNA (GenBank M28417) fragments (expanding from 286 to 810 (for the probe common to h-and l-CaD) and expanding from 811 to 1438 (for the probe specific to h-CaD)), a calponin cDNA (GenBank M63559) fragment (expanding from 1 to 867), an SM22␣ cDNA (GenBank M83105) fragment (expanding from 99 to 524), and a serum response factor (SRF) cDNA (GenBank U50596) fragment (expanding from 30 to 803) were labeled in the same way. 2 g of total RNAs were separated on 1.0% agarose-formaldehyde denaturing gels and then transferred to nylon membranes. The membranes were hybridized with the probe at 42°C for 16 h in 50% formamide, 6 ϫ SSC, 5 ϫ Denhardt's solution (1 ϫ Denhardt's solution is 0.02% polyvinylpyrrolidone, 0.02% bovine serum albumin), 0.5% SDS, and 0.1 mg/ml denatured herring sperm DNA. The blots were washed in decreasing SSC concentrations with a final concentration of 0.1 ϫ SSC containing 0.1% SDS at 52°C. To quantify the applied RNAs, ribosomal RNAs were stained with 0.02% methylene blue.
Primer Extension and S1 Mapping-Primer extension was performed using a 32 P end-labeled antisense oligonucleotide, corresponding to the 5Ј-upstream sequence from ϩ57 to ϩ77 (see Fig. 5A, HO0129), as a primer. The 32 P-labeled primer (1 ϫ 10 6 cpm) was annealed to 20 g of total RNAs from 15-day-old chick embryo gizzards and was extended with Rous-associated virus 2 reverse transcriptase (Takara Shuzo, Japan) for 1 h at 42°C. The extended products were analyzed on 8% polyacrylamide-urea denaturing gels.
As a probe for S1 mapping, a 140-bp genomic DNA fragment was amplified by PCR using a 32 P end-labeled antisense primer, HO0129 (see Fig. 5A). Purified 32 P-labeled probe (1 ϫ 10 6 cpm) was hybridized to 25 g of total RNAs from 15-day-old chick embryo gizzards for 16 h at 37°C in hybridization buffer containing 40 mM PIPES, pH 6.5, 0.4 M NaCl, 1 mM EDTA, and 80% formamide. The hybridized mixture was digested with S1 nuclease (Takara Shuzo, 200 units/ml) in a buffer containing 50 mM sodium acetate, pH 4.5, 250 mM NaCl, 4.5 mM ZnCl 2 , and 1 mg/ml denatured herring sperm DNA for 30 min at 37°C. The protected products were analyzed as described for primer extension analysis.
Construction of Chloramphenicol Acetyltransferase (CAT) Reporter Plasmids-Varied lengths of the 5Ј-upstreams of ␣1 integrin gene were amplified by PCR, and each of the amplified DNA fragments was inserted in pUC0CAT (promoterless CAT plasmid vector) (32). Sitedirected mutations were introduced in ␣1g05ϩCAT (see Fig. 6A) by in vitro mutagenesis using PCR. As a result, the sequences of AP1A, CArG box-like motif, AP1B, and AP2 (see Fig. 5A) were replaced by GGATCCA, GAGCTCAAGG, CGAGCTC, and ATCCGGGC, respectively.
Cell Culture, Transfection, and CAT Assay-Gizzard SMCs were isolated from 15-day-old chick embryos and were cultured in Dulbecco's modified Eagle's medium supplemented with 0.2% bovine serum albumin on laminin-coated dishes (10). Under such culture conditions, a differentiated phenotype of gizzard SMC was maintained for several days (see "Results"). The isolated SMCs were also cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum for more than 1 week to promote dedifferentiation (see "Results"). We used gizzard SMCs cultured on laminin without serum and under serum-stimulated conditions for promoter analysis in differentiated and dedifferentiated SMCs, respectively. Vascular SMCs were isolated from chick aortic media by explant methods (33) and were cultured under serum-stimulated conditions. During this process, a phenotype of vascular SMCs was converted from differentiated to dedifferentiated state. CEFs were cultured as described elsewhere (34). Passaged vascular SMCs and CEFs were used for promoter analyses. CAT assays were carried out as described previously (19,29) with some modifications. Briefly, cells were plated on six-well culture dishes at a density of approximately 1.1 ϫ 10 5 cells/cm 2 for differentiated gizzard SMCs or approximately 5.4 ϫ 10 4 cells/cm 2 for the others to make an equal adhered cell density and were all transiently transfected using Tran-sIT TM -LT1 polyamine transfection reagents (Pan Vera Corporation) according to the manufacturer's recommended procedure. TransIT TM -LT1 reagent-DNA complex mixtures containing 2 g of CAT construct and 1 g of control plasmid carrying the luciferase gene under Rous sarcoma virus promoter (Rous sarcoma virus-luciferase) were added to the cells in Opti-MEM (Life Technologies, Inc.). After a further 4-h incubation, the medium was replaced with fresh culture medium, and the transfected cells were harvested 48 h later. Standardization of transfection efficiency was carried out using luciferase activity as described elsewhere (19,34). Cell extracts containing equal amounts of luciferase activity were used for CAT assay. The transfection experiments were repeated at least three times on duplicate cultures with two or three different plasmid preparations. The CAT activity was quantified by Scanning Imager (Molecular Dynamics).
Analysis of DNA-Protein Interaction by Gel Shift Assay-Probes used for these analyses are shown in Fig. 7A. Nuclear extracts from differentiated and dedifferentiated gizzard SMCs were prepared as described elsewhere (35). Whole cell extracts from chicken aortic tissue and cultured vascular SMCs by explant methods were prepared according to the method described elsewhere (36). For characterization of DNAprotein interaction, samples of nuclear extracts (4 g) were mixed with 0.1-0.2 pmol of 32 P-labeled probe (see Fig. 7A) and 2 g of denatured herring sperm DNA in the presence or absence of nonradiolabeled competitor in 20 l containing 5 mM HEPES, pH 7.8, 5 mM 2-mercaptoethanol, 1 mM EDTA, 60 mM NaCl, 5 mM spermidine, and 10% glycerol at room temperature for 20 min. Samples were analyzed on 5% polyacrylamide gels in 0.5 ϫ Tris borate-EDTA (TBE) buffer. Anti-SRF polyclonal antibodies were purchased from Santa Cruz Biotechnology. We confirmed the cross-reactivity of these antibodies to chicken SRF (data not shown).
In Situ Hybridization of Chick Embryos-Chick embryos at various stages were frozen and sectioned 12-15 m thick by cryostat (Bright, U. K.). The sections were mounted on polylysine-coated glass slides and fixed with 4% formaldehyde in 0.1 M phosphate buffer for 20 min followed by treatment with 10 g/ml proteinase K and acetylation with 0.25% acetic acid anhydrous and 0.1 M triethanolamine. The slides were washed in 0.1 M phosphate buffer and dehydrated in ethanol. A subfragment of chicken ␣1 integrin cDNA, expanding from 1137 to 1835 (Fig. 1A, probe I) and exon 3b of the chicken CaD gene were subcloned into pGEM-4Z vector (Promega). Radiolabeled cRNA probes were prepared using T7 RNA polymerase (Promega) and [␣-35 S]UTP (NEN Life Science Products) and purified with a CHROMASPIN-100 column (CLONTECH). Hybridization was performed at 55°C in 20 mM Tris-HCl, pH 8.0, and 10 mM phosphate buffer containing 1 mM EDTA, 0.3 M NaCl, 50% formamide, 1 ϫ Denhardt's solution, 0.2% Sarcosyl, 500 g/ml yeast tRNA, 200 g/ml herring sperm DNA, and the labeled probe (2-3 ϫ 10 5 cpm). Next, a high stringency wash was performed in 2 ϫ SSC, 5% 2-mercaptoethanol, and 50% formamide for 30 min at 65°C. The remaining probes were digested with RNase A treatment, and an additional high stringency wash was carried out. Finally, the slides were sequentially dehydrated in ethanol and exposed to Fuji PhosphorImaging plate. Data were collected and analyzed with BAS-5000 PhosphorImager (Fujifilm, Japan).

RESULTS
cDNA Cloning of Chicken ␣1 Integrin-To isolate full-length cDNA, a partial cDNA fragment of chicken ␣1 integrin encoding the I-domain was amplified by PCR and was used as a probe for screening. Six overlapping cDNA clones (Fig. 1A) were isolated. Sequence analyses revealed an open reading frame of full-length chicken ␣1 integrin cDNA. The 5Ј-and 3Ј-untranslated cDNA sequences were also cloned by RACE as described under "Materials and Methods." The cDNA fragment obtained by the 3Ј-RACE was terminated at the d(A) 30 sequence in the 3Ј-untranslated region, but a polyadenylation signal is absent within the preceding 50 bp (Fig. 1A). As shown in Fig. 1A, both overlapping sequences cloned by plaque hybridization and RACE were identical. A sequence surrounding the ATG triplet for initiating methionine agreed well with the consensus sequence for functional translation initiating codons in eukaryotic mRNAs (37). Deduced amino acid sequence of chicken ␣1 integrin showed about 65% identity with that of rat (38) or human (39) ␣1 integrin (Fig. 1B). However, the sequence of predicted signal peptides was not homologous. Using a cDNA probe carrying the 5Ј-untranslated region and the signal peptide coding region (probe N2 in Fig. 1A), the hybridization pattern observed was the same as that using the cDNA probe belonging to the homologous region (probe N1 in Fig. 1A) by Northern blotting (see Fig. 3B and data not shown). We further constructed an expression plasmid carrying the full-length open reading frame in eukaryotic expression vector and transiently expressed in dedifferentiated gizzard SMCs in which the endogenous ␣1 integrin was scarcely detectable (Fig. 1C). Western blotting analysis with anti-chicken ␣1 integrin polyclonal antibodies revealed that the transfected cells expressed a 175-kDa protein, and its migrating position in SDS gels was identical to that of endogenous ␣1 integrin in differentiated gizzard SMCs (Fig. 1C). These results indicate that cloned cDNAs certainly encode chicken ␣1 integrin.
Distribution of ␣1 Integrin mRNA in Developing Chick Embryos-It has been reported by immunohistological study that ␣1 integrin is expressed widely in early embryos and comes to be restricted in smooth muscles and endothelium thereafter (4). However, there is no assessment of mRNA expression of ␣1 integrin. We examined the distribution of ␣1 integrin mRNA in developing chick embryos (8-day-old to posthatched) by in situ hybridization. The distribution of h-CaD mRNA was also examined as a control of SMC-specific molecular marker (8,40). The expressional pattern of ␣1 integrin mRNA is very similar to that of h-CaD mRNA (Fig. 2); both mRNAs were strongly expressed in aortic media, gizzard, small intestine, and the digestive tract. Additionally, ␣1 integrin mRNA was significantly detected in the kidney. Consequently, our present results demonstrated that ␣1 integrin is expressed predominantly in smooth muscles at late stages of embryogenesis.
Phenotype-dependent Expression of ␣1 Integrin mRNA in Cultured SMCs-We found that primarily cultured SMCs on laminin-coated plates without serum were able to maintain differentiated phenotype-specific gene expression for several days (10). Under such culture conditions, the expressions of h-CaD, calponin, and SM22␣ were maintained (Fig. 3A, lanes 1 and 2). In our separate experiment, we have found that differentiated phenotype-specific isoforms of tropomyosin, ␣-TM-SM and ␤-TM-SM, were expressed dominantly in such primary culture systems of differentiated SMCs (10). On the other hand, under serum-stimulated conditions, a phenotype of SMC is converted from the differentiated to the dedifferentiated state (41). During this transition, h-CaD was completely converted to l-CaD, and both calponin and SM22␣ were down-regulated (Fig. 3A, lane 3). It has been reported that the expression of ␣1 integrin protein is also down-regulated during dedifferentiation of SMCs (5, 6). Northern blotting was performed to investigate the expressional change of ␣1 integrin mRNA during phenotypic modulation of SMCs. The blots of total RNAs from 15-day-old embryo gizzard, aorta, and cultured SMCs under the indicated conditions were hybridized with a 32 P-labeled probe N1 (Fig. 1A). We also used total RNAs from CEFs as a control for non-muscle cells. Two distinct mRNAs were detected in 15-day-old embryo gizzard and aorta: a major 9.0-kb and minor 4.5-kb mRNA (Fig. 3B, lanes 1 and 4). Differentiated gizzard SMCs cultured on laminin without serum also maintained a high level of ␣1 integrin expression (Fig. 3B, lane 2). In contrast, the expression of ␣1 integrin mRNAs decreased dramatically during serum-induced dedifferentiation of gizzard and vascular SMCs (Fig. 3B, lanes 3 and 5) and was faint in CEFs (Fig. 3B, lane 6). We analyzed the growth of SMCs cultured on laminin under nonstimulated and serum-stimulated conditions and investigated whether the expressional pattern of such molecular markers was affected by the growth state of cells or cell density. A dramatic growth of SMCs was observed under serum-stimulated conditions, whereas significant cell growth was not found under nonstimulated conditions. However, once a phenotype of SMCs was converted to the dedifferentiated state, the cells were not able to recover the expression of differentiation-specific markers even though they were placed in a quiescent state by serum starvation (data not shown) or contact inhibition in the postconfluent culture (10). Thus, this phenotypic conversion is irreversible. It was the same with ␣1 integrin (data not shown). These results suggest that the expression of ␣1 integrin would be transcriptionally regulated in an SMC phenotype-dependent manner.
Identification of Promoter Region of ␣1 Integrin Gene-To isolate the 5Ј-upstream region of the ␣1 integrin gene, we screened a chicken genomic library using a cloned cDNA fragment. As a result of positive clone mapping, a 3.4-kbp EcoRI fragment was hybridized with an oligonucleotide probe G1 (corresponding from ϩ161 to ϩ178 in the 5Ј-upstream sequence (see Fig. 5A)). Sequencing analysis revealed that this fragment contained approximately 2.5 kbp of 5Ј-flanking region, exon 1, and part of intron 1 (Fig. 4A). Genomic Southern blotting using probe E1 (Fig. 4A) resulted in a single hybridized band (Fig.  4B), suggesting that the ␣1 integrin is a single copy gene. A part of the 5Ј-upstream sequence is shown in Fig. 5A. We then carried out primer extension and S1 nuclease mapping analyses to identify the transcriptional starting site. Reverse transcriptase-mediated extension from an antisense oligonucleotide HO0129 (Fig. 5A) yielded two major bands positioned at 277 and 276 bp upstream, respectively, from the translational initiation codon (data not shown). On the other hand, S1 nuclease mapping using an end-labeled antisense DNA (expanding from Ϫ62 to ϩ77 in the 5Ј-upstream sequence (Fig. 5A)) yielded two major protected fragments positioned at 280 and 284 bp upstream, respectively, from the translation initiating codon (Fig. 5B). There was a slight discrepancy between the transcriptional starting sites mapped by primer extension and S1 mapping analyses. Such a discrepancy might be attributed to an arrest of reverse transcription by a secondary structure of mRNA during extension. Accordingly, we designated the more prominent site under S1 nuclease mapping as a major transcriptional starting site (ϩ1) (Fig. 5B). Like other integrin promoters characterized previously (23-28), the upstream region from the transcriptional starting site lacks TATA and CAAT boxes. Several canonical binding sites for common transcription factors such as AP1 and AP2 are found in the upstream region from the translational initiation codon. In addition, a CArG box-like motif is present between Ϫ156 to Ϫ147, just adjacent to an E box (Fig. 5A).
Functional Analysis of the ␣1 Integrin Promoter-To determine whether the 5Ј-upstream region of the ␣1 integrin gene functions to direct transcription in SMCs, we constructed a CAT plasmid, ␣1g05ϩCAT, carrying the 5Ј-upstream sequence of the ␣1 integrin gene from Ϫ516 to ϩ281 (Fig. 6A) and transfected it into primarily cultured SMCs. This construct exhibited sufficient transcriptional activity in differentiated gizzard SMCs (Fig. 6B). We constructed a series of 5Ј-and/or 3Ј-deletion mutants from ␣1g05ϩCAT (Fig. 6A) and tested it in transfection assays to define the potential cis-acting regulatory elements that might be essential for tissue-or phenotypespecific as well as basal expression of the ␣1 integrin gene. In differentiated gizzard SMCs, the deletion of nucleotides from Ϫ516 to Ϫ106 (␣1g05ϩCAT and CArGdel in Fig. 6) led to a dramatic decrease in the transcriptional activity, suggesting the presence of positive regulatory elements within the region. Furthermore, the removal of nucleotides from ϩ173 to ϩ77 (del5Ј-3Ј-2 and del5Ј-3Ј-1 in Fig. 6) resulted in drastic loss of the activity, indicating that the basal promoter activity of the ␣1 integrin gene was attributed to this region. Accordingly, we designated the CAT activity produced by the construct del5Ј-3Ј-2 as the basal promoter activity and compared the activities of the other CAT constructs relative to it among different types of cells. The CAT activity of ␣1g05ϩCAT in differentiated gizzard SMCs was 7-10-fold higher than that in dedifferentiated SMCs or CEFs (Fig. 6B). These results are consistent with Northern blottings as shown in Fig. 3B. We further characterized differentiated SMC-specific transcription based on ␣1g05ϩCAT. In this region expanding from Ϫ516 to ϩ281, a single binding site for AP2, a couple of binding sites for AP1, transcription by S1 mapping. The sequences of cis-elements indicated with bold letters are underlined. Two binding sites for AP1 are designated AP1A and AP1B, respectively. Bold ATG triplets are the codon for initiating methionine of the ␣1 integrin protein. A 140-bp DNA fragment used as an S1 mapping probe is underlined. The corresponding sequences of antisense primers, HO0129, and probe G1, are shown by long horizontal arrows. Panel B, the 5Ј-end of antisense strand of the 140-bp DNA fragment indicated in panel A was specifically 32 P-end labeled and was used for S1 mapping as described under "Materials and Methods." Lane 1, protected products; and lane 2, undigested probe. The positions of major and minor protected products (lane 1) are indicated in the left margin. For size standard, the 5Ј-upstream region subcloned in bacterial plasmid was sequenced using HO0129 as a primer.  S1 mapping (panel B). Panel A, major and minor transcription starting sites determined by S1 mapping are marked by large and small bent arrows, respectively. Dotted characters are the transcription starting sites determined by primer extension using the end-labeled antisense primer, HO0129. Bases are numbered (indicated in the left margin) with respect to the major starting site of and a CArG box-like motif are present (Fig. 5A). To evaluate the contribution of these cis-elements to the transcriptional activity, we introduced mutations in each element and assessed the transcriptional activity of the respective mutated constructs in differentiated gizzard SMCs (Fig. 6, A and B). In comparison with the wild type construct (␣1g05ϩCAT), CArG mutant (CArGMUT) exhibited a more than 10-fold decreased activity, whereas the transcriptional activity was not affected by the mutation of each AP1 site (AP1AMUT and AP1BMUT) or AP2 site (AP2MUT) (Fig. 6B). The effect of mutation in the CArG box-like motif was also examined in dedifferentiated gizzard SMCs, vascular SMCs derived from aortic media explants, and CEFs. No significant change in the transcriptional activities was observed in these cells (Fig. 6B). These findings suggest that the CArG box-like motif would play an essential role in differentiated SMC-specific transcription of the ␣1 integrin gene.
SRF Is a Core Factor Interacting with the CArG Box-like Motif-To investigate interaction between the CArG box-like motif and nuclear proteins, we performed gel shift assays using indicated duplex DNA probes (Fig. 7A) and nuclear extracts from SMCs. A 32 P-labeled ␣1CArG (22) formed a single prominent DNA-protein complex in nuclear extracts derived from differentiated gizzard SMCs (Fig. 7B, lane 1), whereas the complex formed in nuclear extracts derived from dedifferentiated gizzard SMCs was faint (Fig. 7B, lane 4). The DNA-protein FIG. 6. Functional analysis of the ␣1 integrin promoter and identification of a cis-element for SMC phenotype-specific transcription. Panel A, schematic structures of the 5Ј-upstream region of the ␣1 integrin gene and its deleted and mutated CAT constructs. The construct ␣1g05ϩCAT carries the 5Ј-upstream sequence (Ϫ516 to ϩ218) of ␣1 integrin gene upstream from the CAT structural gene. Constructs CArGdel, del5Ј-5, del5Ј-3Ј-2, and del5Ј-3Ј-1 are deletion mutants derived from ␣1g05ϩCAT. Constructs AP1AMUT, CArGMUT, AP1BMUT, and AP2MUT are site-directed mutants to the AP1A site, the CArG box-like motif, the AP1B site, and the AP2 site, respectively, as described under "Materials and Methods." Panel B, relative CAT activities of the constructs in each type of cells. Differentiated or dedifferentiated gizzard SMCs (GzSMC), dedifferentiated vascular SMCs (VsSMC), and CEFs were transfected with the constructs indicated to the left of the panel. Cotransfection with Rous sarcoma virus-luciferase was used to control for transfection efficiency. After a 48-h incubation, the cells were harvested, and CAT activities of cell extracts were assayed. Relative transcriptional activities to the basal promoter (the construct del5Ј-3Ј-2) are shown in each type of cells. Each value represents the average Ϯ S.D. of at least three independent experiments. The promoterless control CAT plasmid, pUC0CAT, did not exhibit a detectable CAT activity in any type of cell (data not shown).

FIG. 7. A nuclear protein factor binds to the CArG box-like motif.
A, the top column shows the CArG box-like motif (enclosed in a rectangle) and its flanking DNA sequence in the ␣1 integrin promoter region. The middle panel shows a 22-bp double-stranded DNA sequence containing the CArG box-like motif. It was designated ␣1CArG (22) and used as a probe or a competitor in gel shift assays. The bottom panel shows a mutant DNA sequence derived from ␣1CArG (22) . It was designated ␣1CArGMUT and used as a competitor in gel shift assays. The replaced nucleotides in ␣1CArGMUT are indicated by dotted letters. B, gel shift assays revealed a specific complex between the CArG box-like motif and nuclear protein factors from SMCs. 4 g of nuclear extract from gizzard SMCs from 15-day-old chick embryo (differentiated gizzard SMCs; lanes 1-3), three time-passaged gizzard SMCs under serum-stimulated conditions (dedifferentiated gizzard SMCs; lanes 4 -6), adult aortic tissue (differentiated vascular SMCs; lanes 7-9), or three time-passaged vascular SMCs under serum-stimulated conditions (dedifferentiated vascular SMCs; lanes 10 -12) were reacted with 32 P endlabeled ␣1CArG (22) without a competitor (lanes 1, 4, 7, and 10) or with a 100-fold excess of unlabeled ␣1CArG (22) (lanes 2, 5, 8, and 11) or with a 100-fold excess of unlabeled ␣1CArGMUT (lanes 3, 6, 9, and 12). Reaction mixtures were separated on 5% polyacrylamide gels. The shifted complexes are indicated by an arrowhead. interaction was specific for the CArG box-like motif because the radiolabeled complex was efficiently replaced by an excess amount (100-fold) of unlabeled ␣1CArG (22) , but not unlabeled ␣1CArGMUT in which the CArG box-like motif was replaced (Fig. 7, A and lanes 2 and 3 in B). The specific DNA-protein interactions observed in these assays were consistent with the transcriptional activities of ␣1 integrin promoter in differentiated and dedifferentiated SMCs. Similar results were also obtained using nuclear extracts from vascular SMCs (Fig. 7B,  lanes 7-12). The intensity of radiolabeled ␣1CArG (22) -protein complex in nuclear extracts from aortic tissue was stronger than that in nuclear extracts from dedifferentiated vascular SMCs. These results suggest that the interaction between ␣1CArG (22) and a protein factor depends on phenotypes of both visceral and vascular SMCs. It is well known that the serum response element (SRE) in c-fos promoter contains a CArG box motif, and SRF binds to this motif (42). The radiolabeled ␣1CArG (22) -protein complex competed efficiently with c-fos SRE (data not shown), suggesting the involvement of SRF in such a complex. To confirm the binding of SRF to the CArG box-like motif, we carried out supershift assay using anti-SRF antibody. As shown in Fig. 8, anti-SRF antibodies supershifted the ␣1CArG (22) -protein complex in a dose-dependent manner, but nonimmune IgG did not. This protein factor, in both nuclear extracts from gizzard and vascular SMCs, was recognized by the anti-SRF antibodies. This finding suggests that, at least in part, SRF is a core factor to interact with the CArG box-like motif and is a key factor for transcription of the ␣1 integrin gene in both SMCs.
To investigate whether the expression of SRF is down-regulated during dedifferentiation of SMCs, we performed Northern blotting. As shown in Fig. 3A, the SRF mRNA did not decrease as drastically as the SRF binding activity in the gel shift assay did. This result suggests the presence of additional factor(s) that regulate the SRF binding in an SMC phenotype-dependent manner.

DISCUSSION
Phenotypic modulation of SMC is associated with atherosclerosis and SMC-derived tumor progression. The SMCs in primary culture under serum-stimulated conditions display phenotypic modulation from the differentiated to dedifferentiated state. During this process the expression of some cytoskeletal proteins is altered dramatically, and these alterations are regulated at mRNA levels. Such expressional changes of cytoskeletal proteins are used for favorable molecular markers indicating the SMC phenotype. Expression of ␣1 integrin in SMCs has been reported to be up-regulated during embryonic development (4). In this study, we focused on the expression of ␣1 integrin in association with an SMC phenotype and analyzed the transcriptional machinery of this gene using both differentiated and dedifferentiated phenotypes of SMCs.
We first cloned the full-length cDNA encoding chicken ␣1 integrin and the genomic DNA carrying the 5Ј-upstream region of the ␣1 integrin gene. The NH 2 -terminal half (extracellular domain) corresponding to the I-domain and seven repeated domains, the transmembrane domain, and the COOH-terminal cytoplasmic domain of chicken ␣1 integrin are highly homologous with those of rat (38) and human (39) ␣1 integrins, although homology is not evident among the sequences corresponding to signal peptides (Fig. 1B). These results suggest that the domains involved in the function of ␣1 integrin are highly conserved among different species. The reliability of our cDNA clone was confirmed by immunoreactivity of the protein produced by forced expression of the cloned cDNA against the anti-chicken ␣1 integrin antibodies (Fig. 1C).
In situ hybridization revealed that the ␣1 integrin mRNA was expressed predominantly in vascular and visceral smooth muscles in late stages of chick embryo. It was clearly detected in 8-day-old embryo and thereafter increased gradually in proportion to the expression of h-CaD mRNA (Fig. 2). Such an expressional profile of the ␣1 integrin mRNA almost completely coincided with its protein level in developing embryos (4). In contrast, the ␣1 integrin was down-regulated dramatically in cultured SMCs during serum-stimulated dedifferentiation (Fig. 3B). These results indicate that the SMC phenotypedependent expression of the ␣1 integrin gene is regulated transcriptionally and that ␣1 integrin is available as a molecular marker of SMC differentiation. In addition to SMCs, a low but a significant amount of ␣1 integrin mRNA was detected in the kidney (Fig. 2). Korhonen [5][6][7][8] were reacted with 32 P end-labeled ␣1CArG (22) with increasing amounts of anti-SRF antibodies (0, 150, 300, or 600 ng of antibodies was added to each reaction). Nonimmune IgG was added to each reaction to adjust the total amount of IgG to 600 ng. Reaction mixtures were separated on 5% polyacrylamide gels. The supershifted complexes are indicated by an asterisk.
The 5Ј-upstream sequences of the ␣1 integrin gene share structural similarities with other ␣ integrin genes (Fig. 5A); both TATA and CAAT boxes are absent, and consensus binding sites for AP1 and AP2 are present (23)(24)(25)(26)(27)(28). It is still difficult to perform promoter analysis using differentiated vascular SMCs because no one has succeeded in preparing cultured vascular SMCs in a differentiated phenotype. We have established a primary culture system maintaining a differentiated phenotype of gizzard SMCs. Using our culture system, we analyzed the phenotype-dependent transcriptional regulation of the ␣1 integrin gene. The comparative analysis of a series of deleted ␣1 integrin promoter-CAT constructs revealed that the basal promoter activity of the gene was attributed to the region from ϩ77 to ϩ173 and that the 5Ј-upstream sequence (Ϫ516 to ϩ281) produced sufficient transcriptional activity (more than 10-fold activity compared with the basal activity) in differentiated gizzard SMCs (Fig. 6B). However, the transcriptional activities of the same sequence in dedifferentiated gizzard and vascular SMCs were much lower than that in differentiated gizzard SMCs (Fig. 6B). Site-directed mutagenesis revealed that the CArG box-like motif was important for the transcriptional activation in differentiated SMCs, whereas binding sites for AP1 or AP2 were not (Fig. 6B). Moreover, mutagenesis on the CArG box-like motif did not affect the transcriptional activities in both dedifferentiated SMCs and CEFs. These findings indicate that the CArG box-like motif serves as a positive element only in a differentiated phenotype of SMCs (Fig. 6B). The CArG box is known to be involved in quite different facets of the transcriptional regulation. One is the serum response of immediate early genes, mediated by SRE that has been initially identified in the c-fos proto-oncogene promoter (42). Another is the transcriptional activation of muscle-specific genes such as the skeletal and cardiac ␣-actin genes (44,45) and the dystrophin gene (46). The involvement of the CArG box in gene expression has been emerging not only in striated muscle cells but also in SMCs. Promoter analyses of the smooth muscle ␣-actin (17,18), smooth muscle myosin heavy chain (20), and CaD (19) genes have proven that the CArG box plays an essential role in SMC-specific transcription. The promoter region of the SM22␣ gene which directs tissue-specific expression in arterial SMCs in transgenic mice contains two CArG boxes (47,48). It has been reported that SRF binds to the CArG boxes in the skeletal, cardiac, and smooth muscle ␣-actin promoters (18). Recently, Soulez et al. (49) directly showed that SRF is necessary for the CArG box-dependent transcriptional activation of muscle-specific genes. We demonstrated by supershift assays using anti-SRF antibodies that SRF is involved in the specific complex formation between the ␣1 integrin CArG boxlike motif and SMC nuclear factors (Fig. 8). This complex formation was intensive when nuclear extracts from differentiated gizzard or vascular SMCs were used, whereas only faint complex formation was observed using nuclear extracts from both dedifferentiated SMCs (Fig. 7). Taken together, we first clarified that the CArG box-like motif in the ␣1 integrin promoter is a core element for SRF binding and that its binding is essential for the phenotype-dependent transcriptional regulation in SMCs. There was little difference in SRF expression between differentiated and dedifferentiated gizzard SMCs, whereas the binding activity of SRF to the CArG box-like motif in differentiated SMCs was much higher than that in dedifferentiated SMCs. This suggests the possible involvement of additional factor(s) that modulates the SRF binding to the CArG box-like motif. The CArG boxes have been shown to be responsible for SMC-specific gene expression as described above. Indeed, the CArG box-like motif in the ␣1 integrin gene may be replaceable with the CArG box sequence from another gene because gel shift assays revealed that the SRF binding to it competed efficiently not only with c-fos SRE but also with the CArG box in the CaD gene (data not shown). Despite the consensus core sequences of the CArG boxes and the CArG box-like motifs, there was no sequential homology surrounding these core sequences among SMC-specific genes, suggesting that the surrounding sequences are not essential for SMCspecific expression. SRF is distributed widely in all types of muscle tissues and other non-muscle cells (50). Therefore, SMC-specific transcriptional activation of the ␣1 integrin and other genes cannot be explained simply by SRF transactivation through its binding to the CArG box-like motif or the CArG box. Homeodomain proteins such as Phox/Mhox (51) or Nkx-2.5 (52) have been known to enhance the binding activity of SRF to CArG box and the transcriptional activity mediated by its binding. Considering these evidences, SMC-specific homeodomain proteins might be a potent candidate responsible for SMC phenotype-dependent and/or SMC-specific transcriptional regulations.
In summary, the present report is a first analysis of the ␣1 integrin promoter and has demonstrated that the interaction between the CArG box-like motif and SRF is essential for SMC phenotype-dependent transcriptional regulation. Further studies regarding a possible involvement of SMC-specific homeodomain proteins in the SMC-specific gene expression are necessary for elucidating the molecular mechanism of phenotypic modulation of SMCs.