Organization and Myogenic Restricted Expression of the Murine Serum Response Factor Gene

Serum response factor (SRF), a member of an ancient family of DNA-binding proteins, is generally assumed to be a ubiquitous transcription factor involved in regulating growth factor-responsive genes. However, avian SRF was recently shown (Croissant, J. D., Kim, J.-H., Eichele, G., Goering, L., Lough, J., Prywes, R., and Schwartz, R. J. (1996) Dev. Biol. 177, 250–264) to be preferentially expressed in myogenic lineages and is required for regulating post-replicative muscle gene expression. Given the central importance of SRF for the muscle tissue-restricted expression of the striated α-actin gene family, we wanted to determine how SRF might contribute to this muscle-restricted expression. Here we have characterized the murine SRF genomic locus, which has seven exons interrupted by six introns, with the entire locus spanning 11 kilobases. Murine SRF transcripts were processed to two 3′-untranslated region polyadenylation signals, yielding 4.5- and 2.5-kilobase mRNA species. Murine SRF mRNA levels were the highest in adult skeletal and cardiac muscle, but barely detected in liver, lung, and spleen tissues. During early mouse development,in situ hybridization analysis revealed enrichment of SRF transcripts in the myotomal portion of somites, the myocardium of the heart, and the smooth muscle media of vessels of mouse embryos. Likewise, murine SRF promoter activity was tissue-restricted, being 80-fold greater in primary skeletal myoblasts than in liver-derived HepG2 cells. In addition, SRF promoter activity increased 6-fold when myoblasts withdrew from the cell cycle and fused into differentiated myotubes. A 310-base pair promoter fragment depended upon multiple intact serum response elements in combination with Sp1 sites for maximal myogenic restricted activity. Furthermore, cotransfected SRF expression vector stimulated SRF promoter transcription, whereas dominant-negative SRF mutants blocked SRF promoter activity, demonstrating a positive role for an SRF-dependent autoregulatory loop.

Serum response factor (SRF), 1 a member of an ancient family of DNA-binding proteins, contains a highly conserved DNAbinding/dimerization domain of 90 amino acids, termed the MADS box (1). The structure of the MADS box domain, recently elucidated by Pellegrini et al. (2), was assembled before the divergence of plants and animals. The identical MADS box structures were present in yeast transcription factors MCM1 and ARG80, a large number of homeotic like plant proteins, and invertebrate and vertebrate SRFs (reviewed in Ref. 3). All of these transcription factors, through their common MADS boxes, virtually bind to the same DNA sequences (1) and interact with similar kinds of co-accessory regulatory factors (reviewed in Refs. 4 and 5). Molecular dissection of human SRF revealed phosphorylation sites in the N-terminal domain that influence DNA binding, whereas sequences downstream of the MADS box contain the C-terminal transcription activation domain (6,7). These extra MADS box sequences are well conserved in vertebrate SRF species, but have diverged from lower animal and plant species.
Earlier studies (8 -10) have demonstrated that SRF-binding sites, termed serum response elements (SRE; GG(A/T) 6 CC), play a primary role in regulating early response genes such as c-fos and egr-1 (reviewed in Ref. 11). Growth factor signaling through the c-fos SRE appeared to be mediated through the formation of ternary complexes with an accessory factor, p62 TCF , and through protein-DNA interactions with a purinerich sequence at the 5Ј-end of the c-fos SRE (12,13). The ETS domain proteins Elk-1 (14) and SAP-1 (15) possess biochemical activities that are characteristic of p62 TCF (16) and interact with the C-terminal portion of the MADS box (17). Furthermore, another MADS box accessory factor, a paired-like homeodomain protein, Phox1 (18), has been shown to facilitate the DNA binding activity of SRF on the c-fos SRE. In addition, Phox1 and Elk-1 potentiate the ability of SRF to transcriptionally activate the c-fos promoter in response to growth-mediated events (18).
SRF plays an obligatory role in regulating post-replicative muscle gene expression. The multiple SREs in the promoters of vertebrate striated ␣-actin genes are required for myogenic expression (19 -22). Site-directed mutagenesis of the SREs of the avian striated ␣-actin promoters revealed that these SREs acted in combination with each other and were necessary for transcription (22,23). We (24) and others (7,25) have shown that SREs are not equivalent in function due to the contextural sequences embedding each SRE. For example, the skeletal ␣-actin proximal SRE allowed a minimal promoter to be activated during muscle differentiation (26). Indeed, SRF could activate the avian skeletal ␣-actin promoter in transient transfection experiments by competing against a negative-acting YY1 factor for binding on the proximal SRE (7,25). Avian SRF mRNA and protein dramatically increase as primary myoblasts withdraw from the cell cycle and fuse. In addition, SRF becomes localized to the nucleus in differentiated myotubes (27).
Thus, the increase in SRF mRNA appeared prior to the upregulation of ␣-actin gene activity during myogenesis (28). Vandromme et al. (29) demonstrated that microinjection of SRF antibodies prevented the progression of myogenic differentiation, implying an early dependence on SRF. In addition, Croissant et al. (27) demonstrated that a dominant-negative SRF mutant, defective in DNA binding but capable of heterodimerizing with other SRF monomers, inhibited the transcriptional activity of the skeletal ␣-actin gene promoter in myogenic cultures and also blocked terminal differentiation. Thus, SRF has a requisite role in ␣-actin gene transcription during terminal skeletal muscle differentiation.
Recently, Spencer and Misra (30) demonstrated that a 322-bp promoter region of the murine SRF gene was responsive to serum stimulation in NIH3T3 fibroblasts and that the SREs and Sp1-binding sites present within this promoter region were responsible. Since it is generally assumed that SRF is a ubiquitous transcription factor, no attempt has so far been made to understand the basis for the muscle tissue enriched expression of SRF. Given the central importance of SRF for the muscle tissue-restricted expression of the sarcomeric actin gene family, we wanted to determine how SRF might contribute to this muscle-restricted expression. Here we have characterized the SRF genomic locus. Murine SRF gene activity was reminiscent of the expression pattern of another MADS box-containing factor, MEF-2 (31), being primarily restricted to cell types derived from embryonic mesoderm such as skeletal, cardiac, and smooth muscles and, to a lesser extent, to cell types of neuroectodermal origins. Also, SRF was virtually absent in endoderm-derived tissues such as the liver, lung, and spleen. To understand the mechanisms responsible for the tissue-regulated expression of SRF, we also analyzed the cis-acting elements in the SRF promoter region. Our results indicated the 310 bp of the SRF promoter region upstream of the cap site as the 5Ј-regulatory boundary required for muscle-restricted expression. Furthermore, dominant-negative SRF mutants blocked SRF promoter activity in muscle cells. SRF gene activity appeared to be under an autoregulatory loop, in which two high affinity SREs in the core promoter were required for the SRF expression in skeletal muscle cells.

MATERIALS AND METHODS
Cloning of Mouse SRF cDNA-A fragment of mouse SRF cDNA corresponding to nucleotides 834 -1520 of human SRF (a gift from Dr. E. Olson) was used to screen a mouse heart cDNA library (Stratagene). One million plaques were initially screened with the above-mentioned mouse SRF cDNA fragment as the probe, and five clones were isolated. Restriction mapping and limited sequencing indicated that four of these clones were identical. The PstI fragment from the 3Ј-end of one of these clones corresponding to nucleotides 1747-1985 was used to rescreen the same library, and seven more clones were isolated. The remaining 3Ј-untranslated region sequences were isolated by screening 5 ϫ 10 5 plaques with the NheI-BglII fragment of the mouse SRF genomic DNA, which corresponded to nucleotides 3354 -4201 of human SRF cDNA. The mSRF cDNA restriction fragments were subcloned and sequenced using the Sequenase Version 2.0 sequencing system (U. S. Bioscience, Inc.).
Isolation of SRF Genomic Clones-The mouse strain 129 genomic library constructed in the EMBL2 -vector was a gift from Dr. Philip Soriano. Three clones, each containing ϳ14 kbp of insert DNA, were isolated by screening one million plaques with the mSRF cDNA fragment corresponding to nucleotides 834 -1520 of the human SRF cDNA. Restriction mapping and Southern analysis of these clones with various probes derived from the murine SRF cDNA indicated that clone -5 contained the complete SRF gene. Appropriate fragments from this clone were subcloned and sequenced. The exon/intron borders were assigned by aligning the cDNA and genomic DNA sequences.
RNA Isolation and Analysis-Total RNA was isolated from adult mouse tissues according to Chomczynski and Sacchi (32). Thirty g of total RNA was resolved on formaldehyde-containing 1% agarose gel and then blotted onto GeneScreen membrane (NEN Life Science Products).
The prehybridization, hybridization, and washings were according to the manufacturer's recommendations. The coding region (nucleotides 939 -1747) and the post I poly(A) region probes were hybridized overnight in 50% formamide at 42°C and washed at 65°C. The 23-mer primer used for primer extension analysis was complementary to nucleotides 82-104 of mSRF cDNA. The primer was end-labeled with T4 polynucleotide kinase and [␥-32 P]dATP, and 100,000 cpm of the labeled primer was hybridized overnight at 30°C with 50 g of skeletal muscle total RNA in 80% formamide, 100 mM sodium citrate, pH 6.4, 300 mM sodium acetate, pH 6.4, and 1 mM EDTA. The primer-RNA hybrid was precipitated and extended with 200 units of SuperScript II reverse transcriptase (Life Technologies, Inc.) in 50 mM Tris-HCl, pH 8.3, 75 mM KCl, 3 mM MgCl 2 , 0.01 M dithiothreitol, and 0.5 mM dNTPs for 1 h at 45°C. The reaction was treated with 1 g of RNase A; extracted once with phenol/chloroform; and analyzed on 8 M urea, 6% denaturing polyacrylamide gel. A dideoxy sequencing ladder generated using the same primer was used as size markers.
In Situ Hybridization of SRF in Mouse Embryos-In situ hybridization was performed on 7-mm sections of day 11.5 mouse embryos as described by Croissant et al. (27). A murine SRF probe, corresponding to the 3Ј-untranslated region subcloned into pBluescript, was linearized with the appropriate restriction enzymes to produce antisense 35 Slabeled copy riboprobes. Sections were hybridized overnight at 58°C and washed at 64°C. Sections were processed for emulsion autoradiography, post-stained with Hoechst 33258, and visualized by epifluorescence and dark-field microscopy.
Recombinant DNA Constructions-The wild-type pϪ456 plasmid was constructed by ligating the PCR-amplified 479-bp fragment containing 456-bp 5Ј-and 23-bp 3Ј-sequences relative to the transcriptional start site of the SRF gene and inserted between the NheI and XbaI sites of the luciferase reporter plasmid pGL2. To construct plasmids pϪ310, pϪ292, and pϪ227, pϪ456 was digested with KpnI and briefly digested with Bal-31 exonuclease. After inactivation of Bal-31 exonuclease, the 5Ј-overhangs were removed by T4 DNA polymerase digestion. The remaining DNA was trimmed with EcoRI, which cuts once within the luciferase vector. The SRF promoter region and part of the luciferase vector were gel-purified and ligated into the SmaI/EcoRI-digested fragment of pGL2. Plasmids pϪ136, pϪ57, and pϪ37 were derived from intermediate constructs containing a BglII site introduced by sitedirected mutagenesis into plasmid pϪ456. The smaller BglII fragments from these intermediate plasmids containing 136, 57, and 37 bp of the SRF promoter were cloned at the BglII site of pGL2-basic to construct pϪ136, pϪ57, and pϪ37, respectively.
Site-directed Mutagenesis of SRF Promoter Elements-Site-directed mutations were introduced into the Ϫ456 background by PCR. Polymerase chain reactions were performed in a 50-l volume of Pfu polymerase buffer (100 pmol each of the primer pairs, 0.2 mM dNTPs, and 5 units of Pfu polymerase (Stratagene)). The upstream and downstream wild-type primers were designed with NheI and XbaI sites, respectively, at their 5Ј-ends. Mutagenic primers contained either BglII (for SRE mutation) or EcoRI (for Sp1 site mutation). The conditions for PCR were as follows: initial denaturation at 95°C for 10 min and an additional 30 cycles of denaturation at 95°C for 1 min, annealing at 55°C for 1 min, and extension at 72°C for 2 min. The final extension was at 72°C for 20 min. The products of the secondary PCR containing the desired mutations were digested with NheI and XbaI and ligated at the NheI site of pGL2. All mutations were confirmed by sequencing.
Tissue Culture and Plasmid DNA Transfections-Chicken embryo primary skeletal myoblasts were isolated from day 11 embryonic breast muscle tissue as described previously (33). After transfection, cells were placed in minimum Eagle's medium containing 10% horse serum and 2% chicken embryo extract. CV1 and HepG2 cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum both before and after transfection. Cells were transfected with 1 g of the indicated plasmid DNA along with 0.2 g of pCMV-␤-galactosidase by LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's recommendations. Transactivation assays were performed with SRF promoter-reporter constructs (1 g cotransfected with 150 ng of SRF expression vector or empty vector). SRF promoter inhibition assays were performed with 150 ng of a SRF dominant-negative mutant (pCGNSRFpm1 or pCGNSRF⌬C). Cells were harvested 48 h post-transfection, and luciferase activity was measured according to standard methods in a luminometer. Luciferase activity was normalized for transfection efficiency using the ␤-galactosidase values. For the promoter transactivation experiments, luciferase activity was normalized to the total protein.
Preparation of Nuclear Extracts and DNA Binding Assays-Nuclear extracts were prepared according to Bohinski et al. (34). The protein concentration of extracts was estimated by the Bio-Rad protein assay reagent. Electrophoretic mobility shift assays used 5-10 g of the nuclear extract prepared from myotubes. The nuclear extract was first incubated for 10 min at room temperature with 1 g of poly(dG-dC) in 1 ϫ binding buffer (50 mM NaCl, 20 mM HEPES-KOH, pH 7.5, 0.1 mM EDTA, 0.5 mM dithiothreitol, and 10% glycerol). Specific and nonspecific double-stranded oligonucleotides and antibodies were included in the reaction for competition and supershift assays, respectively. Subsequently, 0.01 pmol of the indicated end-labeled probe was added and incubated for a further 10 min. DNA-protein complexes were resolved on a 5% polyacrylamide gel cast and run in 0.5 ϫ Tris/glycine buffer. The gel was dried and autoradiographed.
DNase I Footprinting of the SRF Gene Promoter-The SRF promoter fragment from Ϫ253 to ϩ23, relative to the transcriptional start site, was end-labeled at ϩ23 and used for DNase I footprinting. One g of poly(dI-dC) was incubated at room temperature for 5 min without or with increasing amounts of GST-SRF in 20 l of 1 ϫ binding buffer. The reaction was incubated at room temperature for an additional 10 min after adding 5 ϫ 10 5 cpm of probe. The MgCl 2 and CaCl 2 concentrations were adjusted to 5 and 1 mM, respectively, and the probe was digested with 0.004 units of pancreatic DNase I at room temperature for 90 s. The reaction was stopped by the addition of 100 l of stop buffer (100 mM NaCl, 100 mM Tris-HCl, pH 7.6, 15 mM EDTA, 0.375% SDS, 50 mg/ml sonicated salmon sperm DNA, and 100 mg/ml proteinase K) and incubated at 37°C for 20 min. After phenol/chloroform extraction and ethanol precipitation, the samples were dissolved in loading buffer. Approximately 20,000 cpm of the denatured samples was resolved on urea-6% polyacrylamide gel, dried, and autoradiographed.

Organization and Sequence Conservation of the Murine SRF
Gene-A mouse 129 genomic library was screened with the human SRF cDNA (35) corresponding to nucleotides 834 -1520, overlapping the conserved MADS box region. Three overlapping clones, each containing ϳ14 kbp of mouse genomic DNA, were isolated, and one of these clones (SRF 5) contained the complete SRF locus. The SRF structural gene extends over 10 kbp and consists of seven exons (Fig. 1A), encoding the SRF protein of 504 amino acids (Figs. 1B and 2). The sequence at the exon/intron borders conformed to the GT-AG consensus sequence as shown in Table I (36). The size of the exons ranged from 77 to 848 bp. However, based on which of the two polyadenylation signal sequences are used, the last exon was either 833 or 2337 bp in size. The methionine start codon (ATG) was located in the first exon, 347 bp downstream of the major cap site. The first exon contained 347 bp of GC-rich 5Ј-untranslated region and 501 bp of the coding region. The conserved MADS box region was split by the first intron and bordered by the second intron. The transcription activation domain was spread over exons 4 -7. The stop codon (TGA) was located in the seventh and last exon.
Comparison of the mouse SRF amino acid sequence with the human, chicken, and Xenopus sequences revealed a very high degree of sequence conservation during evolution (Fig. 2). The 90-amino acid MADS box region (Fig. 2, boldface) was identical in all three vertebrate SRF species. Mouse SRF is more closely related to human SRF than to Xenopus SRF. The coding region of mSRF is 94 and 95% similar to the nucleic acid and amino acid sequences of human SRF, respectively. The length of the coding region was also conserved for these two mammalian species. However, in comparison with more ancient relatives, such as the Drosophila SRF pruned, homologous sequences were limited only to the MADS box. Examination of the Nterminal domain revealed the presence of a conserved 36-amino acid insert in the amino-terminal region in human and mouse SRFs, which was absent in the Xenopus SRF. A higher degree of sequence divergence between mammalian and amphibian SRF species was observed in the N-terminal domain as compared with the carboxyl-terminal domain, which is involved in transcriptional activation.
Characterization of the Murine SRF Promoter-The 5Ј-cap  site was identified by primer extension analysis. A 23-nucleotide-long antisense olignucleotide from the 5Ј-end of the murine cDNA, which corresponded to nucleotides ϩ82 to ϩ104, was end-labeled, hybridized to total mouse muscle RNA, and extended with reverse transcriptase. Results of primer extension analysis indicated that the majority of the SRF transcripts were initiated from the guanosine located 347 bp 5Ј of the methionine start codon (Fig. 3A), which we previously reported (37). The major cap site for mSRF transcripts was 4 nucleotides downstream of that for human SRF (35). Several minor transcriptional start sites were also detected downstream of the major initiation site. Inspection of the genomic sequences immediately upstream of the cap site revealed a TATA box at positions Ϫ16 to Ϫ22 (Fig. 3B). The TATA box sequence also resembled the consensus binding element for YY1 and a tissue-restricted transcription factor, MEF-2. There are two consensus SREs located at Ϫ42 and Ϫ62. In addition, there are two divergent SREs located at Ϫ142 and Ϫ222. Potential binding sites for other immediate-early gene products (AP-1, Egr-1, and Ets-1) are identified. Overlapping the Egr-1 sites are Sp1-binding sites. Potential binding sites for NF-Y, GATA factors, and TEF-1 are also present within the SRF promoter (Fig. 3B).
Two SRF mRNA Species Are Enriched in Cardiac and Skeletal Muscle Tissues-We asked, how does SRF fulfill its role in regulating striated ␣-actin genes, and how might SRF contrib-ute to tissue-restricted expression? To investigate the expression pattern of SRF, total RNA from various tissues was analyzed by Northern blot analysis. As shown in Fig. 4A, a coding region probe downstream of the MADS box detected SRF transcripts of 2.5 and 4.5 kilobases, which were abundantly expressed in mesoderm-derived tissues such as skeletal and cardiac muscle and, to a lesser extent, in neuroectoderm-derived brain tissue. Liver, spleen, lung, and kidney tissues, which are derived from the endoderm, barely expressed SRF mRNA (Fig.  4). Close examination of the mSRF genomic sequence revealed the presence of two polyadenylation signal sequences, each separated from the other by 1.5 kbp. Differential utilization of these two polyadenylation signals for mRNA 3Ј-end formation could have contributed to the size differences observed for the two SRF RNA species. To examine this possibility, we used a probe that overlapped the second polyadenylation region, which detected only the 4.5-kilobase species, as shown in Fig.  4B. Thus, the two mSRF RNAs arose from post-transcriptional processing of the two polyadenylation sites. As observed for blots probed with the SRF coding region, SRF mRNA was detected primarily in skeletal and cardiac muscle and brain, but not in endoderm-derived tissues, reinforcing our conclusion that the expression of SRF is tissue-restricted (33).

SRF Transcripts Appear Enriched in Embryonic Skeletal, Cardiac, and Vascular Smooth
Muscle-During vertebrate embryogenesis, expression of striated ␣-actin transcripts serves as an early marker for differentiation of cardiac, skeletal, and vascular smooth muscle cell lineages. We asked if SRF mRNA expression patterns were also locally restricted to early embryonic cardiac, skeletal, and smooth muscle cell types. In sec- tioned day 11.5 mouse embryos, SRF transcripts were seen in the neuroectoderm of the brain and the neural tube, but were absent in the underlying notochord (Fig. 5A). Transverse sections revealed high levels of SRF expression in the bulbus cordis and the right atrial portions of the myocardium, as shown in Fig. 5B. SRF was also detected at high levels in the myotomal portion of somites and in the emerging smooth muscle cells surrounding the second branchial arch artery (Fig. 5A). Lower levels of SRF was also detected in the sympathetic trunk (Fig. 5B) and in the cardinal vein. SRF was barely detected in the lung bud and liver (Fig. 5B). Sense probes in all cases showed background levels of hybridization in all tissues (data not shown). These in situ hybridization experiments demonstrated that SRF gene expression was developmentally regulated and largely restricted to the cardiac and skeletal muscle cell lineages, consistent with the early specific expression of the ␣-actin genes in the embryo.
Tissue-regulated Expression of SRF-To investigate the mo-lecular basis for the muscle tissue enriched expression and to map the essential cis-acting elements, we transiently transfected chicken primary myotubes and the human liver cancer cell line HepG with SRF promoter constructs. The pϪ456 plasmid was 80-fold more active in myotubes than in HepG2 cells (Fig. 6). A further deletion to Ϫ310 uncovered the presence of a negative-acting element(s). This deletion of 146 bp resulted in a nearly 2.2-fold increase in the promoter activity in myotubes. The pϪ310 construct was 96-fold more active in myotubes than in HepG2 cells. A deletion of 18 bp to Ϫ292 decreased the promoter activity only slightly in myotubes and by 56% in HepG2 cells. A further deletion to Ϫ227 decreased the pro- FIG. 5. Serum response factor expressed in murine embryonic neuroectoderm-and cardiac, somitic, and smooth muscle mesoderm-derived tissues. In situ hybridization localization (A and B) of SRF transcripts in sectioned day 11.5 mouse embryos was performed with a 35 S-labeled cRNA probe as described under "Experimental Procedures." Sections were processed for emulsion autoradiography, poststained with Hoechst 33258, and visualized by epifluorescence, which labels cell nuclei as white spots against a blue background, and by dark-field microscopy, which shows autoradiographic grains as red spots. A shows an enlargement of the labeled neural tube. The notochord and the first and second branchial arches did not display SRF transcripts, except around the second branchial arch artery. B shows SRF transcripts in the outside wall of the bulbus cordis and right atria of the heart and the somitic myotomes. The sympathetic trunk and lung bud were lightly labeled, whereas the liver did not display in situ SRF labeling. The following abbreviations mark embryonic tissues as shown: moter activity by ϳ60% in myotubes, but not in HepG2 cells. Additional muscle-specific positive-acting element(s) were revealed by deletion to Ϫ136, which resulted in a 4-fold decrease in the promoter activity in myotubes and a slight decrease in HepG2 cells. Two non-consensus SREs are present within this deleted 91-bp region. A further deletion of 79 bp to Ϫ57, which removed one of the two consensus SREs (SRE2) and overlapping Egr-1-and Sp1-binding sites, reduced the promoter activity by 12-fold in myotubes. Plasmid pϪ57, which contains a single consensus SRE (SRE1) and the TATA box region, was 25-fold more active than the promoterless control plasmid pGL2 in myotubes. The pϪ57 construct was not active in HepG2 cells, which express only low levels of endogenous SRF.

FIG. 8. The SRF promoter requires two intact SREs and an Sp1 site for transcriptional activity in muscle cells. A and B
show binding of SRF and Sp1 from myotube nuclear extracts to cognate probes. End-labeled double-stranded SRE1 and SRE2 (A) and distal Sp1 (B) oligonucleotides were incubated with 5 g of myotube nuclear extract in 1 ϫ binding buffer. Where indicated, unlabeled competitor DNA was added at a 50-fold molar excess before adding the probe. Polyclonal antibodies for SRF, YY1, Sp1, and Egr-1 were used in supershift assays. Oligonucleotides to SRE1 (Ϫ53 to Ϫ30 bp, GCCTCGC-CATAAAAGGAAACATTG), the SRE1 mutant (Ϫ55 to Ϫ25 bp, CGGC-CTCGCCATAgAtctAAACATTGTATC), SRE2 (Ϫ74 to Ϫ51 bp, GGGCTCGCCATATAAGGAGCGGCC), the SRE2 mutant (Ϫ75 to Ϫ46 bp, GGGGCTCGCCATAgAtctAGCGGCCTCGCC), Sp1.1 (Ϫ94 to Ϫ71 bp, CCAATGGGGCGGGGGCGCTGGGGC), the Sp1.1 mutant (Ϫ97 to Ϫ66 bp, GGACCAATGGGGCGaattCGCTGGGGCTCGCC), Sp1.2 (Ϫ263 to Ϫ244 bp, CAACCCAGGGGGGCGGAACT), and the Sp1.2 mutant (Ϫ266 to Ϫ236 bp, CTCCAACCCAaGaattCGGAACTGGT-TCGGC) were used as duplexed DNA probes in band shift assays. Complexes were resolved by electrophoresis on 5% nondenaturing polyacrylamide gel cast and run in 0.5 ϫ Tris borate/EDTA. SRF and YY1 complexes are identified in A and B. SRF promoter activity was regulated by both SRF and Sp1. Site-specific mutations at SRE1 and SRE2 (C) and proximal and distal Sp1 sites (D) were introduced into pϪ456 by PCR-directed mutagenesis using the mutated sequences shown above. The wild-type and mutant plasmids were transfected into primary embryonic myoblasts. Cells were cultured for 2 days in differentiation medium before harvesting. Luciferase activity was normalized to cotransfected ␤-galactosidase activity. The promoter activity of mutated SRE and Sp1 site constructs was compared with wild-type pϪ456 activity, which was set at 100%. Data are presented as the mean Ϯ S.E. of three experiments done in duplicate.

FIG. 9. SRF promoter activity elevated during myogenesis is stimulated by SRF in fibroblasts and blocked by SRF dominantnegative mutants.
A, the pϪ456 SRF promoter-luciferase reporter gene was cotransfected with pCMV-␤-galactosidase into primary chicken embryo myoblast cultures. Assays of luciferase light units normalized to ␤-galactosidase activity were carried out at 24-h intervals during myogenesis in culture. B, the pϪ456 SRF promoter-luciferase reporter gene or the SV40 promoter-enhancer-driven reporter was cotransfected in primary myoblasts with pCGNSRFpm1, pCGNSRF⌬C, or the pCGN empty vector. Cells were harvested after 2 days of incubation, and luciferase activity was normalized to total protein. Normalized luciferase activity of pSV2Luc and Ϫ456Luc cotransfected in primary myoblasts with the empty vector pCGN was set at 100. C, SRF promoter-luciferase reporter gene was cotransfected with pCGNSRF or the pCGN control vector in CV1 cells. Data are presented as the mean Ϯ S.E. from three experiments done in duplicate.
A deletion of 20 bp to Ϫ37, which eliminated SRE1, reduced the promoter activity to background levels in myotubes, suggesting that the SREs were required for SRF promoter activity.
Multiple SRF-binding Sites in the SRF Promoter-Results of promoter deletion analysis suggested that the SREs present within the SRF promoter were required for transcriptional activity. DNase I protection assays performed with bacterially expressed purified SRF were used to ascertain if the two consensus SREs present in the SRF promoter bind SRF. Both SRE1 and SRE2 were well protected at the lowest levels of GST-SRF. In addition, the TATA box region was also protected, but at considerably higher inputs of GST-SRF (Fig. 7). No protection was observed over the two potential, but non-consensus SREs located at positions Ϫ142 and Ϫ222.
SREs may also serve as binding targets for YY1, NFIL-6, SRE-ZBP, SRE-BP, and several other uncharacterized factors. Many of these factors are also expressed in skeletal muscle tissue (Refs. 7 and 38; reviewed in Ref. 11). Having demonstrated specific binding of bacterially expressed SRF, we investigated the interaction of proteins from myotube nuclear extracts with SREs from the SRF promoter. Double-stranded oligonucleotide probes corresponding to SRE1 and SRE2 were incubated with the nuclear extract prepared from chicken embryo myotubes. A doublet of slowly migrating complexes (complexes I and II) and a fast migrating complex (complex III) were observed with the SRE2 probe (Fig. 8A). Nuclear extracts from the fibroblast cell line NIH3T3 also gave rise to a similar doublet of SRF complexes when a longer probe was used for gel shift assays (30). Complexes I and II were competed by a 50-fold molar excess of cardiac SRE1 and wild-type SRE1 and SRE2 oligonucleotides from the SRF promoter, but not by nonspecific Sp1/Egr-1 and mutant SRE1 and SRE2 oligonucleotides ( Fig. 8A) (data not shown). Furthermore, complexes I and II were either supershifted or abolished by SRF antiserum, but not by YY1 antiserum, indicating that these two complexes contain SRF (Fig. 8A) (data not shown). Complex III was identified as a YY1-containing complex based on several criteria. First, this complex was abundant in extracts from replicating myoblasts and decreased in myotube extracts as myogenesis progressed (data not shown). Second, this complex was competed by SRE2 and the YY1-binding site-containing skeletal actin SRE1 oligonucleotide, but not by SRE1 and cardiac actin SRE1 oligonucleotides, which do not bind YY1 (Fig. 8A) (data not shown). Third, complex III was abolished by YY1 antiserum and unaffected by SRF antiserum. SRE1 was similar to SRE2 with respect to SRF binding, except that it did not bind YY1 (Fig. 8A). We have also examined the binding of SRF to noncanonical SREs (SRE3 and SRE4) by competition gel shift assays in which SRE1 was the probe. Both SRE3 and SRE4 oligonucleotides competed for SRF binding to the SRE1 probe, albeit poorly (data not shown).
Mutations at Both (but Not Individual) SREs Block Promoter Activity-To investigate the role of SRE1 and SRE2 in the muscle enriched expression of SRF, site-directed mutational analysis of these SREs was performed in the context of the pϪ456 promoter. The BglII restriction site, which prevented the binding of SRF, was substituted for these SREs (Fig. 8A). Mutation of SRE1 did not affect the activity of the promoter, suggesting that the SREs may have partially redundant functions as SRE2 could functionally substitute for mutated SRE1 (Fig. 8C). In contrast to the SRE1 mutation, the SRE2 mutation increased the activity of the promoter by 2.2-fold (Fig. 8C). However, mutation of both SRE1 and SRE2 resulted in ϳ75% reduction in the promoter activity, suggesting that either of the two SREs is required for the complete promoter activity in myotubes.

Sp1
Binding Activity Is Required for SRF Promoter Activity-In the SRF promoter, there were two potential Sp1 sites located at Ϫ86 bp (proximal Sp1) and Ϫ251 bp (distal Sp1). Overlapping these Sp1 sites were potential binding sites for the zinc finger-containing transcription factor Egr-1. Sp1 sites from several other promoters contain similarly overlapping Egr-1 sites and bind both Sp1 and Egr-1. We tested if the Sp1 sites from the SRF promoter could bind both Sp1 and Egr-1 present in myotube nuclear extracts. Three closely migrating complexes were seen with the distal Sp1 site probe (Fig. 8B). These complexes were specifically competed by a 50-fold molar excess of unlabeled consensus Sp1 site and proximal and distal wild-type Sp1 sites, but not mutated Sp1 sites. The identity of these complexes was confirmed by an antibody supershifting experiment. The complexes were supershifted by the addition of Sp1 antibody, but not by Egr-1 antibody, indicating that the complexes contain Sp1 and not Egr-1. Even though the proximal Sp1 site contains an overlapping consensus Egr-1 site, binding of only Sp1 was evident under our electrophoretic mobility shift assay conditions (data not shown).
Spencer and Misra (30) showed by mutational analysis that the two Sp1 sites present in the murine SRF promoter were essential for serum induction of the promoter in 3T3 fibroblasts, but mutated Sp1 sites did not affect the basal promoter activity. The role of these Sp1 sites in the muscle enriched expression of the SRF promoter was also examined by mutagenesis. Sp1 site-directed mutations abolished the binding of Sp1 from myotube nuclear extracts (Fig. 8B). Mutagenesis of the proximal Sp1 site actually increased the promoter activity by 2-fold (Fig. 8D), whereas a mutation over the distal Sp1 site reduced the promoter activity by 60%.
Myogenic Up-regulation of SRF Promoter Activity-Recently, avian SRF RNA and protein and SRF DNA binding activity were shown to increase when cultured primary myoblasts were allowed to withdraw from the cell cycle and fuse to form multinucleated myotubes (7,27). We wanted to determine if the up-regulation of SRF gene activity was under transcriptional control by examining the activity of transfected SRF promoter-reporter constructs in primary chicken embryo myogenic cultures. The SRF promoter fragment from Ϫ456 to ϩ23 linked to a luciferase reporter gene displayed low activity in replicating pre-fusion myoblasts, but was up-regulated ϳ6-fold in late-stage myotubes (Fig. 9A). Thus, cis-acting sequences required for the up-regulation of SRF gene activity are contained within the 456-bp promoter region.
The up-regulation of SRF during myogenesis, the presence of multiple SREs in the SRF promoter, and the high affinity binding of SRF to these SREs suggested the possibility that SRF autoregulates itself. Thus, we compared luciferase reporter activities of the pϪ456 SRF promoter construct cotransfected with or without SRF expression vectors in non-myogenic CV1 fibroblasts. Coexpression of an exogenous SRF resulted in up to a 5-fold increase in SRF promoter activity (Fig. 9C). We then asked if dominant-negative SRF mutants would block SRF promoter function in transfected chicken embryo myotubes. The SRFpm1 mutant (39) dimerizes with other SRF monomers and interferes with wild-type SRF by forming DNA binding-defective heterodimers. In addition, another dominant-negative SRF mutant, SRF⌬C, in which the C-terminal transcription activation domain (amino acids 266 -504) was deleted, acts as a de facto repressor by occupying SREs through specific DNA binding, but is incapable of activating SRE-dependent transcription. Cotransfection of either SRFpm1 or SRF⌬C with pϪ456 resulted in 40 and 95% decreases in SRF promoter activity, respectively, thus suggesting that the myogenic up-regulation of SRF promoter activity was mediated by SRF (Fig. 9B). Inhibition by SRFpm1 and SRF⌬C was specific to the SRF promoter because SV40 promoter activity was not significantly affected by these dominant-negative SRF mutants. These results indicate that SRF autoregulates its own promoter and that this autoregulation is primarily mediated through SRE1 and SRE2. DISCUSSION The hypothesis that introns and RNA splicing facilitated the evolution of ancient genes in the progenote organism was recently reviewed (40). The function of introns in the evolution of genes can be explained by the proposal that either introns appeared late in evolution and could not participate in the construction of primordial genes or that RNA splicing and introns existed in the earliest organisms, but were lost during the evolution of the modern prokaryotes. Blake (41) suggested that evidence for intron-facilitated evolution of a gene might be found by comparing the borders of functional protein domains with the placement of introns. The recent elucidation of the x-ray crystal structure of the SRF MADS box demonstrated a novel DNA-binding motif, a coiled-coil, and a stratified structural subdomain involved in dimerization (2). We showed here ( Fig. 1 and Table I) that the murine SRF gene consists of seven exons interrupted by six introns. Exon/intron borders are well conserved between the Xenopus SRF (42) and murine SRF genes. The first intron was found to sever the N-terminal extension, which makes specific base contacts within the minor grove of an SRE half-site, from the dimerization subdomain encoded in exon II. In comparison with the genomic organization of MEF-2B (43), which was conserved in Drosophila dmef-2, other MEF-2 relatives, and the plant AGL3 gene (44), the first intron also bisected the unstructured N-terminal extension, whereas the second intron was close to the C-terminal border of MADS boxes found in animal and plant SRFs and the MADS/MEF-2 boxes in all MEF-2 genes. Thus, in all MADS box-containing genes yet examined, introns closely circumscribed the dimerization subdomain. Based on conservation of primary sequence of the MADS box region and gene organization analysis, introns might have participated in the construction of the earliest MADS box-containing genes, prior to the diversion of plants and animals that occurred at least one billion years ago.
Despite these similarities, MADS box proteins also have evolved to perform diverse functions such as specification of mating type in yeast, homeotic activities in plants, pulmonary system development in Drosophila, and elaboration of mesodermal structures in vertebrates. Interestingly, the overall structural divergence of SRF proteins among evolutionarily distant species of animals appears to be related to differences in the spatial expression pattern. For example, of the variety of animal species examined, pruned, the SRF homolog of Drosophila SRF (45), was the most divergent, in which sequence conservation was limited only to the MADS box domain. The localization of Drosophila SRF expression was different from that of Xenopus, avian, and murine SRFs, which are more unified in structure and tissue expression. Drosophila SRF was localized to the insect tracheal system (46), whereas vertebrate SRFs, like several of their MEF-2 counterparts (reviewed in Ref. 5), were localized to avian and murine skeletal and cardiac muscle and neuroectoderm-derived tissues (Figs. 4 and 5) (27).
How does SRF play a central role in regulating musclespecific genes that are expressed under cell differentiationpromoting conditions? We have shown that mSRF has a distinct striated muscle tissue enriched expression pattern and further identified the SRE as a mediator of SRF promoter regulation. Although it is generally assumed that SRF serves a role as a constitutive factor during its association with acces-sory factors, we have shown that SRF binding activity actually increased dramatically following the ending of cell replication primarily due to change in the cellular content of SRF in primary myoblasts (7). Surveys of early avian (27) and murine (Fig. 5) embryos also indicated tissue-restricted expression of SRF transcripts, which substantially increased the cellular mass of SRF in the myotomal portion of somites, cardiac myocytes, and vascular smooth muscle cells. The expression of SRF continues in these tissues at high levels through adulthood (Fig. 4). During primary myogenesis in culture, SRF promoter activity, RNA, and protein mass increase significantly preceding fusion of myoblasts and the appearance of muscle-specific structural genes (33). We showed that mutations of both SRE1 and SRE2 were required to down-regulate SRF promoter activity, suggesting that SRE1 and SRE2 were functionally redundant (Fig. 8C). Overexpression of SRF from a plasmid vector substantially increased SRF promoter activity in transfected fibroblasts (Fig. 9C). In comparison, the dominantnegative mutants of SRF, SRFpm1 and SRF⌬C, inhibited the muscle enriched expression of SRF and other muscle-specific genes. 2 Thus, SRF has a primary role in directing muscle differentiation.
Tissue-restricted expression of SRF was also evident from comparison of its promoter activity, in which the SRF promoter activity was at least 2 orders of magnitude greater in primary cultured myotubes than in liver HepG2 cells. Low SRF promoter activity cannot be attributed to the presence of strong liver tissue-specific silencer elements in the SRF promoter because serial deletions in the promoter did not activate the promoter. Although less likely, it is possible that a silencer element located within the 37-bp cap upstream region strongly repressed the SRF activity in liver cells. Another possibility for the lack of SRF promoter activity in liver cells could be the absence of SRF promoter-specific trans-acting factors from this cell type. Endoderm-derived liver tissue did not express endogenous SRF (Fig. 4). Furthermore, a minimal SRF promoter containing a single SRE, which was otherwise active in myotubes, was not active in HepG2 cells, indicating that the SRF promoter activity was SRF-dependent. The SREs were required for the higher basal level of expression in myotubes (Fig.  8C).
The basal activity of the SRF promoter in NIH3T3 cells was shown to be dependent on the more ubiquitous CAAT boxbinding factor, whereas the SREs and Sp1 sites were dispensable (30). In contrast, SRE1 and Sp1 sites were required for both serum-induced promoter activity (30) and muscle tissuerestricted activity. SRF promoter deletion analysis demonstrated that SRF was necessary but not totally sufficient for driving the SRF promoter. Furthermore, autoregulation of the SRF promoter by SRF alone cannot account for the muscle tissue enriched expression of the SRF gene. The tissue enriched expression of SRF might be accomplished by the concerted action of SREs with other cis-acting elements. Accordingly, deletion of sequences 5Ј to SRE1 and SRE2 decreased the promoter activity, suggesting that the interaction of these deleted sequences with SRE1 and SRE2 may be required for the complete activity of the promoter. Of the two Sp1-binding sites present in the SRF promoter, only the distal Sp1 site appeared to be important for SRF expression. Sartorelli et al. (47) have shown that a functional interaction among SRF, the ubiquitous transcription factor Sp1, and the cell type-restricted myogenic factor MyoD is required for human cardiac ␣-actin gene expression.
One mechanism by which SRF might promote muscle-spe-cific gene expression would be by interfering with the activity of negative regulators of muscle differentiation. YY1, a ubiquitously expressed C 2 H 2 zinc-finger protein (48,49) that binds the consensus sequence AANATGGNG, has been shown to bind several SREs (7,50). We observed that in proliferating myoblasts, the skeletal ␣-actin gene was repressed by mutually exclusive binding of YY1 over SRE1 (7). Interestingly, gel mobility shift assays with SRE2 from the SRF promoter also uncovered an overlapping YY1 site (Fig. 8A). Furthermore, comparison of nuclear extracts prepared from proliferating myoblasts with myotubes revealed mutually exclusive binding of SRF and YY1 over SRE2 (data not shown). Consistent with the repressor role of YY1, mutation of SRE2 resulted in a 2.2-fold increase in SRF promoter activity (Fig. 8C), indicating that the SRF promoter activity might also be modulated by YY1. Displacement of YY1 by increased SRF binding activity during myogenesis may facilitate the SRF autoregulatory loop. The combinatorial interaction of SRF with tissue-specific accessory factors might be another way to confer tissue specificity to SRF (reviewed in Refs. 51 and 52). A cardiac-specific homeodomain protein, Nkx-2.5, was shown to be expressed with a spatio-temporal pattern similar to that observed for avian SRF. Physical interaction of SRF with Nkx-2.5 resulted in enhanced expression of both endogenous and transfected chicken cardiac ␣-actin genes (52). Building up of SRF⅐Nkx-2.5 complexes might compete off negative-acting factors such as YY1 and allow for saturation of the multiple SREs with positive-acting SRF complexes. Conversely, SRF⅐Nkx-2.5 complexes in cardiac myocytes might serve to repress the c-fos promoter through forming nonproductive complexes via its SRE. Therefore, SRF might be able to mediate accessory factor interactions with certain homeodomain factors that either activate or repress transcription.
Likewise, interactions of SRF with skeletal muscle-restricted basic helix-loop-helix proteins of the MyoD family may also confer skeletal muscle specificity to SRF (53). Although consensus E box-binding sites for basic helix-loop-helix proteins are absent in the SRF promoter, it is still conceivable that the SRF promoter could be activated by myogenic regulatory factors in an E box-independent manner, as has been demonstrated for the chicken myoD promoter (54). The mechanism of imparting muscle tissue specificity to SRF is not limited to physical interaction between SRF and a tissue-restricted transfactor. Functional interaction of SRF and MyoD or myogenin bound to different sites on the promoter can confer muscle tissue specificity to SRF (47). A similar interaction of SRF on the interleukin-2 receptor ␣-chain gene promoter with the Rel homology protein NF-B confers T-cell specificity to SRF (55,56). Thus, additional interactions of SRF with different cell type-restricted coactivators may also determine the response of different tissues to SRF.