Identification of Direct Serum-response Factor Gene Targets during Me2SO-induced P19 Cardiac Cell Differentiation*

Serum-response factor (SRF) is an obligatory transcription factor, required for the formation of vertebrate mesoderm leading to the origin of the cardiovascular system. Protein A-TEV-tagged chromatin immunoprecipitation technology was used to collect direct SRF-bound gene targets from pluripotent P19 cells, induced by Me2SO treatment into an enriched cardiac cell population. From 242 sequenced DNA fragments, we identified 188 genomic DNA fragments as potential direct SRF targets that contain CArG boxes and CArG-like boxes. Of the 92 contiguous genes that were identified, a subgroup of 43 SRF targets was then further validated by co-transfection assays with SRF. Expression patterns of representative candidate genes were compared with the LacZ reporter expression activity of the endogenous SRF gene. According to the Unigene data base, 84% of the SRF target candidates were expressed, at least, in the heart. In SRF null embryonic stem cells, 81% of these SRF target candidates were greatly affected by the absence of SRF. Among these SRF-regulated genes, Raf1, Map4k4, and Bicc1 have essential roles in mesoderm formation. The 12 regulated SRF target genes, Mapk10 (JNK3), Txnl2, Azi2, Tera, Sema3a, Lrp4, Actc1, Myl3, Hspg2, Pgm2, Hif3a, and Asb5, have been implicated in cardiovascular formation, and the Ski and Hes6 genes have roles in muscle differentiation. SRF target genes related to cell mitosis and cycle, E2f5, Npm1, Cenpb, Rbbp6, and Scyl1, expressed in the heart tissue were differentially regulated in SRF null ES cells.

Serum-response factor (SRF), 1 a DNA-binding protein, is composed of two monomers that homodimerize and symmetrically contact DNA-binding sites with a consensus sequence CC(A/T) 6 GG, named a CArG box (1,2), an SRE (3), or a CBAR (4 -6). SRF is a member of an ancient DNA-binding protein family, which shares a highly conserved DNA binding/dimerization domain of 90 amino acids, termed the MADS box (7). SRF from human through yeast transcription factors MCM1 and ARG80 and several plant proteins, such as Deficiens, all have an evolutionarily conserved MADS box with similar DNA sequence binding specificity (8). In addition, SRF-related proteins (RSRF/MEF-2) constitute a subfamily of the MADS box family of transcription factors (9,10). MEF-2 factors contain the MADS box and an adjacent MEF-2 box (10). MEF2 proteins bind to MEF-2 sites, CTA(A/T) 4 TAG, which can be found in the regulatory regions of both non-muscle and muscle-specific genes (10,11). Despite their similarities, MADS box proteins have evolved to perform important biological functions such as specification of mating type in yeast (12), homeotic activities in plants (13), pulmonary development in Drosophila (14), and elaboration of mesoderm structures in vertebrates (15).
SRF is a key regulator of immediate early gene expression, which frequently results in mitogenesis, and also of terminal muscle differentiation (3,16,17). SRF acts as a platform to interact with other regulatory proteins and ultimately the regulation of specific gene programs. For example, at the c-fos promoter, SRF recruits proteins having an Ets domain and forms ternary complexes through the B box region of the Ets protein (18). The contiguous Ets DNA-binding site adjacent to the SRE and mitogen-activated protein kinase phosphorylation of the B box stimulates SRF-dependent c-fos activity in replicating cells (19). For the most part, Ets factor-dependent ternary complexes do not play a role in regulating SRF-dependent myogenic gene targets. Instead, SRF co-accessory factor association with homeodomain factors like Nkx2-5 (20,21), zinc finger proteins like GATA4 (22), LIM factors like CRP2 (23), and myocardin (24) and its close relatives MRTFs and MAL (25,26) deliver powerful SRF-dependent muscle gene activity.
In contrast to the c-fos gene, which contains a single high affinity binding site for SRF in its promoter, many musclespecific genes, including skeletal and cardiac ␣-actins and smooth muscle expressed genes, are dependent upon multiple CArG boxes as mixtures of both strong and weak binding sites (16,17,20,21). In addition, high levels of SRF expression and increased SRF mass appeared to coincide with the expression of cardiac, skeletal, and smooth muscle ␣-actins (19) noted as early markers for terminal striated and smooth muscle differentiation. The homologous recombinant knock-out of the murine SRF gene demonstrated a severe block for mesoderm formation during mouse embryogenesis (15). These very early lethal SRF-deficient embryos, which appear to have normal cell replication, also have a severe gastrulation defect and do not develop to term or express SRF-dependent genes.
Currently, there are a limited number of direct SRF gene * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ Supported by National Institutes of Health Grant P01 HL49953. ‡ ‡ To whom correspondence should be addressed: Institute of Biosciences and Technology, the Texas A & M University System Health Science Center, 2121 W. Holcombe Blvd., Houston, TX 77030. Tel.: 713-677-7710; E-mail: rschwartz@ibt.tamhsc.edu. 1 The abbreviations used are: SRF, serum-response factor; PATChIP, protein A-TEV-tagged chromatin immunoprecipitation; ChIP chromatin immunoprecipitation assay; ES, embryonic stem; SRE, serum-response element; TEV, tobacco etch virus; ProtA, protein A; PBS, phosphate-buffered saline; RT, reverse transcription; DMEM, Dulbecco's modified Eagle's medium; VEGF, vascular endothelial growth factor; E(n), embryonic day n. targets, which are regulated during the formation of the early cardiovascular mesoderm. The chromatin immunoprecipitation assay (ChIP) was harnessed to explore regulatory networks in the developing cardiovascular mesoderm. ChIP assays have begun to reveal co-regulated groups of genes that share cis-regulatory elements that bind the same transfactors. The great advantage of combining global transcription factor binding analysis with expression profiling is that the direct targets of transcription factors can be distinguished from indirect downstream effects. ChIP is a powerful tool for elucidating in vivo DNA-protein interactions. Recently, ChIP has been used to identify the target genes of a variety of DNA-binding proteins from yeast to human, such as the Gal4 gene in yeast (27), AGL15 of plants (28), Engrailed of Drosophila (29), and E2F genes in humans (30,31). We developed a systematic nonbiased search for downstream sequence genes that may be directed through the SRF to capture genetic targets. A generic protocol was devised to purify the DNA fragments bound to SRF in vivo from the whole genome by Protein A-TEV-tagged Chromatin Immunoprecipitation (PATChIP). We used PATChIP technology to purify the DNA fragments bound in vivo to serumresponse factors, and we identified 188 SRF-binding fragments and 92 contiguous genes as SRF targets. Here, we focused on 43 of these SRF-binding fragments that appear to play important roles in mesoderm formation and cardiovascular and muscle differentiation.

Construction of the Protein A-tagged-TEV-SRF Expression Vectors-
A DNA fragment containing a Kozak sequence, the two IgG binding units of protein A (ProtA), and TEV protease recognition sequences was amplified by PCR from pNOPPATAIL (32) and subcloned between the NheI and XbaI sites of a pcDNA3.1(Ϫ) myc-his (Invitrogen) to build a pcDNA-ProtA-TEV construct (N-terminal ProtA construct). The C-terminal ProtA construct, pcDNA-TEV-ProtA, was obtained by a ligation of three fragments. Fragment A (1.4 kb) is an ScaI-BamHI fragment of the pcDNA3.1(Ϫ) myc-his vector containing a partial ampicillin resistance gene and a cytomegalovirus promoter. Fragment B (4.4 kb) is a KpnI-ScaI fragment from pcDNA-ProtA-TEV. It consists of the two IgG binding units of ProtA, a poly(A), neomycin resistance gene, and a partial ampicillin resistance gene from the pcDNA-ProtA-TEV construct. Fragment C, containing BamHI, HpaI, and HindIII sites, the TEV protease recognition site, and a KpnI site, was obtained by annealing the following two synthesized oligonucleotides: 5Ј-GATC-CACGTTAACAAAGCTTGAAAACCTGTATTTTCAGGGCTCTG-3Ј and 5Ј-CAGAGCCCTGAAATACAGGTTTTCAAGCTTTGTTAACGTG-3Ј. An XbaI-XbaI fragment containing the SRF gene from a pCGN-SRF construct was cloned into the XbaI site of the pcDNA-ProtA-TEV construct to obtain pN-ProtA-TEV-SRF. A similar cloning strategy was used to make a pC-SRF-TEV-ProtA construct. An XbaI-BamHI PCR product without stop codon from the pCG-SRF construct was ligated into XbaI and BamHI sites of pcDNA-TEV-ProtA.
SRF-binding Fragment, Hsp68 Luciferase Construct-An 871-bp fragment from the HindIII to NcoI sites of the Hsp68-LacZ construct was subcloned into the HindIII and NcoI sites of pGL3-basic vector (Promega). The SRF-binding fragments were cut with EcoRI and XhoI and then subcloned into EcoRI and XhoI sites of the Hsp68 luciferase construct.
Cell Culture, Transfection, Stable Line Selections, Differentiation, and Luciferase Assays-The P19 mouse EC cell line from the American Type Culture Collection was cultured in growth medium (Dulbecco's modified Eagle's medium (DMEM), Invitrogen; growth medium is DMEM with 5% heat-inactivated fetal bovine serum, 5% normal calf serum, 2 mM L-glutamine, 50 units/ml penicillin, 50 g/ml streptomycin, and 10 mM non-essential amino acids according to Skerjanc (33)). P19 cells were transfected with pN-ProtA-TEV-SRF, pC-SRF-TEV-ProtA, pcDNA-ProtA-TEV, or pcDNA-TEV-ProtA. Stable transfectants were selected with G418 and pooled in P19 cells. Several stable lines were obtained, including S18 (expressed the fusion protein, SRF N-tagged ProtA) and SM1 (expressed the SRF C-tagged ProtA). Another P19-ProtA stable line was also selected as control, in which the ProtA protein only was expressed. For differentiation, P19 cells, or S18, and SM1 stable lines were seeded in bacterial Petri dishes at 1 ϫ 10 6 cells/dish; the cells grew and aggregated with differentiation medium (growth medium1% Me 2 SO) for 3-4 days. These aggregates were placed onto standard tissue culture plates with growth medium for another 3-4 days. For early differentiation, the cells were cross-linked with 1% formaldehyde after 4 days in differentiation medium. For luciferase assays, the SRF-binding fragment Hsp68-luciferase constructs were co-transfected with the pCGN-SRF construct alone into CV-1 cells. After 48 h, the transfectants were harvested and lysed, and 20 l of cell lysates was analyzed for luciferase activity as described previously.
Electrophoretic Mobility Shift Assays-Nuclear extracts were prepared according to Chen and Schwartz (20). 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 transfected SRF null ES cells with expression vectors driving the Prot A-tagged SRF species. The nuclear extract was first incubated for 10 min at room temperature with 1 g of poly(dG-dC) in binding buffer (50 mM NaCl; 20 mM HEPES-KOH, pH 7.5; 0.1 mM EDTA; 0.5 mM dithiothreitol; 10% glycerol). Specific c-fos SRE and nonspecific E box double-stranded oligonucleotides and anti-SRF antibodies were included in the reaction for competition and supershift assays, respectively. Subsequently, 0.01 pmol of the indicated endlabeled probe was added and incubated for a further 10 min. DNAprotein complexes were resolved on a 4% polyacrylamide gel cast and autoradiographed.
Purification and Cloning of SRF-binding Fragments-The following solutions for PATChIP were used: Cell lysis buffer (10 mM Tris-HCl, pH 7.5; 10 mM NaCl; 3 mM MgCl 2 ; 0.1% Nonidet P-40); Nuclei lysis buffer PATChIP, Strategy and Protocol-Protein-DNA complexes from SM1 or P19-proteinA cells were cross-linked by 1% formaldehyde at room temperature for 20 -30 min, and then glycines were added to stop the reaction at a final concentration of 0.125 M for 5 min. Cells were then washed two times with 1ϫ PBS, harvested into a 15-ml tubes, and centrifuged at 700 ϫ g at 4°C for 5 min. After that the cell pellet was suspended in 1 ml of Cell lysis buffer on ice for 10 min and centrifuged at 2000 rpm for 10 -15 min at 4°C to pellet the nuclei. Isolated nuclei were suspended in 1 ml of Nuclei lysis buffer, incubated on ice for 10 min, and sonicated to fragment DNA with an average length of 200 -400 bp. After centrifugation at 20,000 ϫ g, the soluble chromatin was transferred to a 15-ml column and 12 ml of IPP-150 buffer, 300 l of pre-activated IgG-Sepharose (Amersham Biosciences) was added for incubation at 4°C for 2 h. The Sepharose beads in the column were washed with 45 ml of ice-cold IPP-150 and then 10 ml of TEV buffer. To cleave the SRF-DNA complexes, Sepharose beads were incubated with 1 ml of TEV buffer with 100 units of TEV protease (Invitrogen) and were rocked gently at room temperature for 3 h or at 4°C overnight. Then the SRF-DNA complexes were eluted into a new 1.5-ml microtube by gravity, and the Sepharose beads were washed with 1 ml of IP buffer. 100 l of SRF antibody (Santa Cruz Biotechnology) and pre-coupled protein G beads (Upstate Biotechnology, Inc.) were added into the tube that contained the SRF-DNA complex eluates and were incubated overnight at 4°C. Beads were washed twice each with IP buffer, once with high salt solution, LiCl solution, and TE at room temperature. 500 l of elution buffer A was incubated with the beads, rotating at room temperature for 15-30 min. Harvested SRF-DNA complexes were carefully transferred into a new 1.5-ml tube and treated with 1 l of 0.5 M EDTA and 5 l of proteinase K (20 mg/ml) for 2 h of incubation at 55°C. 20 l of 5 M NaCl was then added into the 500-l eluate and incubated at 65°C for overnight for reverse cross-links. SRF-binding fragments were purified by the Qiagen column (from Qiagen PCR purification kit), and 100 l of 10 mM Tris-HCl, pH 8.5, was added to recover the DNA. To check the quality of DNA from PATChIP, 2 l of recovered DNA was used as the template to amplify the mouse cardiac actin and c-fos fragment by PCR with the specific primers for the cardiac actin and c-fos genes.
Construction of SRF-binding Fragment Library, BLAST Searches of Genomic Sequences, and Determination of Putative Binding Motif for DNA-binding Proteins-To make the SRF-DNA binding fragment library, DNA was blunted by T4 DNA polymerase, phosphorylated by T4 kinase, and then cloned into the EcoRV site of Bluescript KSII vector. Individual clones were purified and sequenced. The genomic locus of each sequence was mapped by BLAST searches of Ensembl, UCSC, and Celera mouse genome server. MatInspector (www.genomatix.de) and TESS (www.cbil.upenn.edu) were used for searching the putative binding motif for DNA-binding proteins within each SRF-binding fragment.
RNA Preparation, Probe Labeling and Microarray, and RT-PCR-Srf Ϫ/Ϫ , Srf ϩ/Ϫ , and Srf ϩ/ϩ (wild-type) ES cells (murine embryonic stem cells) were kept without feeder cells on gelatin-coated dishes in complete medium (DMEM) with 2 mM L-glutamine,100 units/ml penicillin, 100 g/ml steptomycin, 0.1 mM ␤-mercaptoethanol, 15% fetal bovine serum, and 1000 units/ml leukemia inhibitory factor. To initiate differentiation, ES cells were seeded and grown in monolayer fashion on gelatin-coated dishes and kept overnight in complete medium. The 1st day after seeding was counted as day 0 of differentiation. The medium was then replaced with complete medium without LIF. The medium was replaced every other day, and the cells were harvested at days 2, 4, 8, and 14. Total RNA was then isolated by Trizol (Invitrogen), and the RNA pellets were redissolved in diethyl pyrocarbonate distilled H 2 O. According to the protocol of Affymetrix, two-cycle amplification by in vitro transcription, and antisense RNA was labeled with biotin. After fragmentation and hybridization to MOE430A chips, data were analyzed using Microarray Suite-5 and d-Chip software (34). Reverse transcription reaction was carried out with Superscript II and oligo(dT), according to the manufacturer's instructions (Invitrogen). The following PCR primers are available upon request: RNA Probe Construction, Whole Mount in Situ Hybridization, and ␤-Galactosidase Staining-The RNA probe constructs were obtained by cloning the RT-PCR fragment of gene-specific primers into the EcoRV site of Bluescript KSII. Whole-mount in situ hybridization was performed by the Robotic machine InsituPro (Intavis AG), and our detailed protocols are available upon request. Generation of Srf ϩ/Ϫ LacZ mouse lines will be described elsewhere. Staged embryos from E 6.0 to E 10.5 were stained for ␤-galactosidase after fixation in 2% formaldehyde, 0.2% glutaraldehyde at room temperature for 60 min, were washed three times with 1ϫ PBS containing 0.1% sodium deoxycholate and 0.2% Nonidet P-40, and then stained in the reaction solution (10 mM ferrocyanide, 5 mM ferricyanide, 2 mM MgCl 2 , and 1 mg/ml of 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-Gal) in dimethylformamide) at 37°C overnight.

RESULTS
Generation of the PATChIP System in P19 Cells-We developed a protein A tag-based technology to recover transfactor-DNA complexes, as diagrammed in Fig. 1A. This flow chart  2-6). S18 and S15 are stable lines selected for expressing SRF-N-terminal-tagged ProtA (SRF-N); SM1 is a stable line selected for expressing SRF-Cterminal-tagged ProtA (SRF-C), and S10 and S12 are lines expressing only ProtA. C, SRF-N and SRF-C, in electrophoretic mobility shift assay. Electrophoretic mobility shift assays were performed as described by Chen and Schwartz (20) with minor modifications. Electrophoretic mobility shift assays used 5-10 g of the nuclear extract prepared from transfected SRF null ES cells with expression vectors driving Prot A-tagged SRF species. Specific c-fos SRE and nonspecific E box double-stranded oligonucleotides and rabbit IgG were included in the reaction for supershift assays, respectively. The DNAprotein complexes were fractionated on 5% polyacrylamide gels that were processed for autoradiography x-ray film. D, the cleavage of TEV to the SRF-ProtA complexes by Western blot. The cleavage of SRF-protein complexes in the 1st lane was treated with TEV protease and in the 2nd to the 5th lanes was treated with PBS. The uncompleted cleavage of SRF-ProtA shows two bands in Western blotting. E, the differentiation of P19 and SM1 (a SRF-C-terminal-tagged ProtA expressing stable line) by Me 2 SO. 0, Me 2 SO treatment Day 0; 6, Me 2 SO treatment Day 6. H, mouse adult heart tissue used as positive control for RT-PCR; CT, negative control. Group 1 marked by red letters contains 43 DNA fragments and their contiguous genes. Group 2 contains the remaining 150 cloned fragments, some of which have CArG boxes contiguous to ESTs. Group 3 contains the remaining fragments that cannot be tied to genes due to incompleteness in assembly of the whole mouse genome. These groups were organized using mouse genome assembly as of May 2004. The abbreviations used are as follows: Cons, conservation; 5UTR, 5Ј-untranslated region; 3UTR: 3Ј-untranslated region. Conserved represented cross-species conservation of the fragment by C Ͼ75% conservation, P ϭ 30 -75% conservation, and U Ͻ 30% conservation; species conservation, R, rat; h, human; d, dog; C, chick; CArG box, CArG box and CArG-like box; CArG&L, total number of CArG box and CArG-like box; CArG consv., CArG or CArG-like box conservation.
shows the following strategy: (i) starts with formaldehyde chromatin cross-linking within stable transfected cells; (ii) chromatin sonication; (iii) retrieval of ProtA-tagged SRF chromatin fragments; (iv) reversal of formaldehyde cross-links; (v) subsequent formation of a library, and sequencing of potential targets; and (vi) the identification of potential SRF gene targets. First, P19 cells were chosen as a tractable biological system for retrieval of SRF gene targets, because P19 cells can differentiate into several cell types, including cardiac, skeletal, and smooth muscle (33). P19 cells also naturally express endogenous SRF (Fig. 1B) and, when grown on a large scale, provide sufficient chromatin DNA for cloning after multiple steps of PATChIP purification. Second, the ProtA tag has been widely used for protein complex purification (35,36) and modified to place a cleavable TEV site between the SRF coding region and the dual ProtA moiety. ProtA-TEV-SRF (N-tagged) and SRF-TEV-ProtA (C-tagged) constructs were stably transfected into P19 cells (Fig. 1B). Antibody blots of P19 cellular proteins on PAGE showed expression of heavier NЈ-tagged (cell line S18 and S15) and CЈ-tagged SRF species (cell line SM1) in comparison with wild type SRF expressed in control P19 cell lines (S10 and S12) only expressing endogenous SRF in the presence of ProtA. We did not observe any differences in forming DNAbinding complexes as shown by electrophoretic mobility shift assay (Fig. 1C) or their ability to be cleaved by TEV (Fig. 1D). Because a tag placed on the CЈ terminus of a protein may still be preferable because of less interference with protein folding than with NЈ-terminal tags, studies were continued with the SRF-TEV-ProtA (C-tagged) P19 stable lines. P19 and stably transfected C-tagged SRF cell lines were then treated with 1% dimethyl sulfoxide (Me 2 SO), aggregated, and then differentiated into cardiogenic cell types as shown by expression of specific cardiac marker genes, Nkx2-5 and ␣-MHC (Fig. 1E). Third, after formaldehyde cross-linking and sonication of the chromatin, SRF-DNA complexes were captured on IgG-Sepharose by high affinity binding of their ProtA tags (Fig. 1A, IIIa). Gentle elution of enriched SRF-DNA complexes, although leaving behind nonspecific bound complexes, was achieved by the incubation with TEV protease (Invitrogen), which cut the unique TEV site Glu-Asn-Leu-Tyr-Phe-Gln-Gly (Fig. 1A, IIIc). To obtain greater enrichment of SRF-binding fragments, an additional step was used to immunoprecipitate the SRF-DNA complexes with SRF-specific antibody (Fig. 1A, IIId and IIIe). Finally, a pool of SRF-binding fragments were then filled in, blunt-ended, and ligated into Bluescript plasmid vectors to build an SRF-DNA target library. Samples were taken for quality control assays and were found to be enriched in cardiac ␣-actin promoter sequences over nonfractionated genomic DNA (data not shown).
We sequenced 242 individual clones from the SRF-DNA target library, and we compared their sequences against the published mouse genome (that was assembled by December, 2002) by BLAST search engines of from the University of California, Santa Cruz (genome.ucsc.edu), Ensembl (Sanger Institute, www.ensembl.ori), and Celera (www.celera.com). A cloned fragment was validated as a potential SRF-DNA binding target, by the following three criteria. 1) The fragment must have 95% or greater identity to the sequence of C57/black mouse genome, because P19 cells were derived from a CH3 mouse. 2) The fragment must be located within 100 kb upstream or downstream of a gene, based on previous reports that some enhancers may reside 85 kb upstream of its promoter (37). 3) The fragment must contain at least one SRF-binding site as a CArG consensus sequence or a CArG-like box (CArGL), being off the consensus sequence by no more than 1 base (16). Fig. 2 shows group assignment of the PATChIP cloned fragments, as SRF gene targets. Of the 242 cloned fragments, we found that 188 DNA clones (78%) contained at least a CArG or CArGL box, thus validating PATChIP technology as a highly efficient tool for retrieving SRF DNA-binding sites, as listed in Table I. In addition, 54 fragments that did not contain CArG boxes were also listed in Table I, in the event that they might also serve as indirect SRF targets. Table I also revealed a high level of sequence conservation of the fragments and their CArG boxes across mammalian species. We observed that of the 188 DNA fragments, 62% have conserved CArG boxes shared between two mammalian species and cross-homology could have even been greater, because sequencing of the rat genome is incomplete. Furthermore, of the 242 cloned fragments, 95% of them were found as hits in the whole mouse genome, whereas  only 92 fragments could be connected to their contiguous genes (assigned in Group 1 and Group 2 of Table I). Unfortunately, the remaining fragments cannot be tied to genes (Group 3) because of incompleteness in assembly of the whole mouse genome and because the map was published at the end of 2002. Table I displayed the updated search results of all 242 fragments based on the mouse genome assembled by NCBI, as of May 2004. Further characterization of 92 potential gene targets revealed that about half of these fragments were 150 -400 bp in length, and 70% of these fragments were located within the promoter, or within an intron, and no further than 10 kb from the transcription initiation site gene (Table II). As shown by the flow chart in Fig. 2, we further focused on 43 well defined SRF gene targets genes (not the 28 ESTs) out of 71 CArG box-containing genes.

SRF-dependent Transactivation of DNA Targets Correlated Well with the Quality and Quantity of SRF-binding Sites-Do
the CArG boxes within these SRF-binding fragments contribute to transcription activity? SRF-binding fragments were cloned into plasmid vectors to form DNA target fragment-Hsp68-luciferase reporter gene constructions. The Hsp68 promoter served as a neutral basal promoter. SRF-binding fragment-Hsp68-luciferase constructs co-transfected with the SRF expression vectors into CV-1 cells showed that most of the SRF-DNA binding fragments were activated by 3-20-fold in the presence of SRF expression. Activation of the SRF-binding fragments correlated well with the quality of CArG boxes, in that higher levels of SRF-induced transactivation trended with consensus CArG sequences rather than imperfect CArGL sequences (Table III and Fig. 3). Reporter gene activity was also well influenced by multimeric CArG and CArGL boxes, indicating dependence of transfected gene activity on SRF. SRF DNA binding efficiency was shown to be strongly influenced by co-accessory factor interactions (22,23). An enriched assortment of consensus binding sites for transcription factors known to associate with SRF, such as GATA, NKE, HNF, Smad, STAT and NF-B, were detected in the sequenced DNA fragments by MatInspector (www.genomatix.de) and TESS (www.cbil.upenn.edu) (Table III). Frequent YY1 sites were also detected. Previously, YY1 was shown to bind to a subset of SREs and competed with SRF for binding to the c-fos and skeletal and cardiac ␣-actin promoters (21,24,38). DNA factorbinding sites in Table IV appeared at a frequency much greater than chance in the genome, suggesting that these DNA fragments may be serving primary roles such as enhancers, silencers, and promoters in gene regulation.
SRF and Its Putative Target Genes Were Co-expressed during Embryogenesis-Another validation step was to ascertain the embryonic expression patterns of representative SRF target genes and to determine whether they overlay SRF protein in mouse embryos. Mouse lines were generated by homologous recombination in which the bacterial ␤-galactosidase gene (lacZ) replaced the first coding exon of the SRF gene locus (77). At E 7.0, the SRF-LacZ protein was strongly expressed in mesoderm and endoderm, primitive streak, amnion, and allantois (Fig. 4A). Later, the SRF-LacZ was highly expressed in the linear and looping heart tube and in pre-somite of the caudal region at E 8.0 -8.5 (Fig. 4, B-D). After embryo "turning," the SRF-LacZ was highly expressed in developing heart and in the tail tip with a gradient fashion (at E 9.0, Fig. 4E), as well as in the first and second branchial arches, otic vesicles, limbs, and the some regions of the neural tube of the embryo (at E 9.5-10.0, Fig. 4F). Similar to the SRF-LacZ expression pattern, transcripts of representative gene targets Hif3␣, Myl3, Actc1, Spnb3, Azi2, and Itg␣9 were detected in the developing heart (Fig. 4, G and M) and were also expressed in other tissues, such as neural tissue, limbs, and somites. Myl3 and Pgm2 genes (Fig. 4, G and H) were also highly expressed in the developing heart similar to those of Actc1 (cardiac ␣-actin (39)). Hes6 and Txnl2 were highly expressed in the somites and limbs, whereas Itg␣9 was mainly expressed in the vasculature of the embryos. Most interestingly, Hif3␣, Myl3, Pgm2, and Hes6 were also expressed in the otic vesicles of mouse embryos. Thus, several novel SRF target genes displayed expression patterns that either partially and or completely overlapped with embryonic SRF gene activity.
Activity of SRF Target Genes Were Greatly Affected by SRF Null ES Cells-Previously, c-fos and ␣-actin genes, two known SRF target genes, were shown to be dramatically down-regulated in Srf Ϫ/Ϫ mouse embryos and in the SRF (Srf Ϫ/Ϫ ) null ES cells (15,40). As a final validation step, we asked if SRF target gene activity was compromised by the absence of SRF in ES cells. RNA isolated from the wild type and Srf Ϫ/Ϫ ES cells on days 2, 4, 8, and 14 after withdrawal of LIF from media was converted to labeled RNA following two rounds of cDNA amplification and was hybridized on Affymetrix MEO430A chips. We defined a "regulated" SRF target gene if the gene activity responded as either up-or down-regulated by greater than 2-fold when comparing Srf Ϫ/Ϫ ES cells and wild type ES cells. Of 43 regulated SRF target genes (Table V) genes may be SRF direct targets. Mapk10 (JNK3), Txnl2 (thioredoxin-like2 PICOT), Tera (a Wnt antagonist), Map3k14 (NF-B-inducible kinase), and Azi2 (5-azacytidine induced gene 2) identified as SRF putative target genes are involved in the Wnt/JNK cascade that plays an important role in heart formation during embryogenesis (46,47). However, only Mapk10 (JNK3), Txnl2, and Azi2 (in Table V) were downregulated, whereas Tera was up-regulated in the Srf Ϫ/Ϫ ES cell TABLE III Enrichment of SRF and co-accessory transfactor DNA-binding sites from sequenced SRF-PATChIP fragments The abbreviation of DNA-binding motifs found by MatInspector program are as follows: CArGL, CArG-like box; CArG, CArG box; NKE, Nkx3-1; and Nkx3-2 (homeobox gene); GATA, all GATA factors (zinc finger protein); mTATA, muscle TATA box; MyoDs, MyoD/b-HLH family (MyoD, Myf5, and myogenin); HNF1/4, hepatic nuclear factor 1, 4, and 6; STATs, Smad, Smad3, Smad4, and FAST-1; Comp, cooperates with myogenic proteins in multicomponent complex; STAT3, -5 and -6; mTEF, muscle TEF; Ets, Ets family; NF-B, YY1. differentiation, suggesting SRF may up-regulate these targets during heart formation.
Microarray informatics data were further corroborated by surveying the expression of 15 SRF target genes, which included transcription factors, signaling molecules, and structural or metabolic proteins by RT-PCR in SRF null ES cells. Among these 15 genes, only the Npm1 gene was unchanged in the RT-PCR assay. Fig. 5 showed a subset of RT-PCR data in which Actc1 (cardiac actin) and Myl3 (myosin light chain ventricular) gene activities were blocked on days 2, 4, and 8 in SRF Ϫ/Ϫ ES cell culture. Ski (Sloan-Kettering viral oncogene), E2f5, Raf1, and Map4k4 genes were down-regulated, whereas Bicc1 (Bicaudal-C), Hes6 (Hair and enhancer split 6), and Snx2 (sortin nexin 2) gene were up-regulated in the Srf null ES cells,  Table III. Within these sequences we predicted the random appearance of the transcription DNA-binding sites shown above by MatInspector, observed to expectation indicated enrichment for most of the co-accessory transcription-binding sites. An exception was GATA-binding sites that are so numerous to begin with as compared with other transfactors.  in comparison to wild type ES cells, thus confirming their expression activity as determined by microarray analysis.

DISCUSSION
There is an apparent inconsistency between the activity of gene expression in SRF Ϫ/Ϫ ES cells and the cotransfection of DNA targets with plasmid-based SRF expression. Bicc, Hes6, Snx2, Tera, Cenbp, and Scyl are prime examples of genes that increased transcription in the absence of SRF, yet they were activated by SRF in transfection assays as shown in Fig. 3. One explanation for this apparent conundrum is that SRF may become a potent de facto repressor under conditions where its 3Ј activation domain is deleted either by alternative splicing (68) and or by caspase 3 cleavage, as found in human heart failure (69). Thus, the physical state of SRF may vary depending upon cell type, developmental context, and or disease to act as either an activator and or repressor of targets. In addition, deletion of the activation domain still allows SRF to interact with other cofactors that can further modify SRF-dependent gene activity. For example, ATF6 (68) and myocardin (24,40,70) can supply transcription activity in the absence of the SRF activation domain. Furthermore, the role of cofactors that com-pete for SRF binding such as YY1 or homeodomain factors that act as SRF co-repressors such as HOP (71,72) and bring histone deacetylases (73) to SRF adds to the great complexity of SRF gene regulation.
We noted that a high percentage of the fragments contained multiple CArG boxes and were located close or within the promoters. Fig. 3 showed reporter gene activity influenced by the increased numbers of CArG and CArGL boxes. In contrast to the c-fos gene promoter, which contains a single high affinity SRE, the promoters of many muscle-specific genes, including the skeletal, cardiac, and smooth muscle ␣-actins, contain combinations of three or more CArGs that bind SRF in a highly cooperative manner (74). Distance and phasing between multiple CArG boxes also contribute greatly to promoter strength, where close proximity of CArG boxes strengthens the cooperative binding of SRF to the weaker CArGL (75). The cardiomyogenic gene CArG boxes recruit collateral factors such as Nkx2.5 and GATA4, and these in turn strongly enhance SRF-DNA binding affinity, permitting the formation of higher order DNA binding complexes at relatively low SRF levels. Similarly, CRP2 LIM protein, which bridges SRF and GATA factors (23) and myocardin, greatly enhances SRF DNA binding (24) and thus increases the apparent concentration, or mass, of SRF.
SRF Transcriptional Network-Many of the SRF target genes that have been identified over the last decade have been related to immediate early genes (16,18,48,49). More recently, SRF appears to function as a master regulatory platform that allows for switching of genetic programs depending upon specific factor-factor associations with its MADS box. Combinatorial interactions shared between SRF with Nkx2-5, GATA4, myocardin, and LIM factors may be obligatory for the progression of mesoderm precursor cells to committed cardiac mesoderm. This idea is supported by the demonstrated combinatorial interactions shared between SRF with these factors that appear to require specific physical association through the MADS box. Here we uncovered over 70 novel SRF gene targets by PATChIP technology, many of which have been shown to play important roles in mesoderm formation and cardiovascular development as shown in a schematic diagram outlining SRF networking (Fig. 6).
SRF Target Genes, Raf1, Map4k4, and Bicc1, Play Roles in Mesoderm Formation-SRF may activate Raf1 and Mak4k4 and other genes to promote or permit mesoderm formation and repress Bicc1 to prevent overgrowth of endoderm tissue. Raf1, a signaling molecule, is an important component in the Ras/ Raf/MAP signaling cascade in which SRF is a mediator (12,50).
Cripto activates the Ras/Raf/MAP pathway in embryogenesis, and the lack of Cripto results in defective precardiac mesoderm unable to differentiate into functional cardiomyocytes (51). Cripto was strongly up-regulated and Raf1 was severely downregulated in SRF-deficient null ES cells, 2 indicating that the Raf1 gene down-regulated results from SRF deficiency rather than the repression of Cripto. In mammalian cells, there are three highly conserved Raf genes, Araf, Braf, and Craf (Raf1). Two out three Raf genes are indispensable for life because disruption of either of them invariably results in embryonic perinatal lethality. The double Braf Ϫ/Ϫ /Craf Ϫ/Ϫ mutants (disruption of all four copies of Braf, and Craf) were present in the maternal decidua as small clumps of cells. Loss of three copies of four copies of Braf, and Craf (Braf Ϫ/ϩ /Craf Ϫ/Ϫ or Braf Ϫ/Ϫ / Craf Ϫ/ϩ ) resulted in underdeveloped brain structure, heart, and limbs (52). Most severely affected embryos were arrested at the early stages of gastrulation. Therefore, the Raf1 gene down-regulated in these SRF Ϫ/Ϫ ES cells may be one of the primary causes in forming defective mesoderms. Most interestingly, the SRF-binding fragment located within intron-6 of the Raf1 gene also resided at a region 6 kb downstream of a zinc The right column is the fold changes of the gene expression in the Srf Ϫ/Ϫ ES cell differentiation. The indicates down-regulated and the OE indicates up-regulated by more than 2-fold. The » indicates no change in activity. *, Unigene was used to determine whether the target GENE was expressed in the heart cDNA library and does not construe tissue specificity. finger protein-encoding gene, Makorin-2, which runs antisense in the orientation of the Raf-1 gene. In humans, both genes were co-expressed in the same tissues and cell lines, suggesting activated makorin may affect the expression of Raf-1 (53).
Map4k4 was down-regulated in SRF Ϫ/Ϫ ES cells during differentiation. It has been reported that Mapk4k4 Ϫ/Ϫ embryos die during postgastrulation between embryonic E 9.5-10.5 with the severe impairment of mesodermal and endodermal cell migration (42).
Bicaudal-C (Bicc1), an RNA-binding protein, a homologue of Drosophila gene Bicaudal C, is believed to function at the post-translational level. It has been shown that Bicaudal-C is a strong inducer for endoderm formation in Xenopus. Overexpression of xBicc1 leads to impaired normal development and blocks the expression of mesoderm and neuronal markers, forming excessive endoderm tissue (44). Mouse Bicc1 mRNA is found in the growing oocyte, Hensen's node, and the developing kidney (45). In Srf Ϫ/Ϫ ES cells differentiation, we found that Bicc1 was highly up-regulated, along with several endoderm markers in microarray analysis (data not shown). Therefore, along with reduction of expression of Raf1 and Map4k4 genes, overexpression of Bicc1 possibly resulted in excessive endoderm tissue and impairment in mesoderm formation in Srf Ϫ/Ϫ ES cells.
SRF Target Genes Play an Inductive Role in Cardiovascular Development-Both Wnt/␤-catenin and Wnt/JNK cascade play important roles in cardiovascular development (40,46,47,54). Among the SRF-regulated target genes, Mapk10 (JNK3), Txnl2, Tera, and Azi2 were involved in the Wnt/JNK cascade. Mapk10(JNK3) was expressed in the heart, the brain, and the testis, whereas JNK1 and JNK2 are widely expressed. Txnl2 (thioredoxin-like 2) plays a role in protecting the heart tissue from various types of stress in JNK cascade (55). Tera is a Wnt antagonist identified recently and is expressed in neural tubes and somites (56). We observed that Txnl2 and Mapk10 (JNK3) were down-regulated, whereas Tera was up-regulated in Srf Ϫ/Ϫ ES cell differentiation, indicating that SRF may regulate these genes via cross-talk to other components of Wnt/JNK cascade to play a role in heart formation.
Among the regulated SRF target genes, nine genes are Actc1 (cardiac actin), Myl3 (myosin light chain, slow), Lrp4 (low density lipoprotein-related protein 4), Hspg2 (perlecan), Pgm2 (phosphoglucomutase), Hif3␣ (hypoxia-induced factor 3␣), VEGFR3 (Flt4), Itg␣9 (integrin ␣9), and Asb5 (ankyrin repeated and Soc box-containing protein 5). Actc1, the cardiac actin gene, severely down-regulated in Srf Ϫ/Ϫ ES cells is a key contractile protein component of the sarcomere. Sema3a Ϫ/Ϫ and Hspg2 Ϫ/Ϫ mutant embryos displayed cardiac defects during mouse development (57,58). Lrp4 (corin) is a type II transmembrane serine protease abundantly expressed in the heart, and Lrp4 converts pro-atrial natriuretic factor to the active form of atrial natriuretic factor by protein modification in the heart. The function of Pgm2 in heart development is not known, although it was expressed in the embryonic heart. Hif3␣ and Asb5 were reported to play roles in cardiovascular development (59 -61) but also to regulate VEGF receptors and angiogenesis (62). Asb5 has been reported to be expressed in cardiac and skeletal muscles and in endothelial cells of blood vessels in rabbit, suggesting that it may be a component for myogenesis and arteriogenesis (63). Among the regulated SRF target genes, Ski and Hes6 are important transcription factors for myogenesis (64,65). Ski was down-regulated and Hes6 was up-regulated in Srf Ϫ/Ϫ ES cell differentiation, indicating that SRF plays a role in myogenesis through the regulation of its target genes Ski and Hes6.
Recently, Selvaraj and Prywes (76) have identified a subset of genes that are MKL (also known as MAL, MRTF-A and -B, and BSAC)-dependent by expression profiling using a cell line that expressed a dominant negative MKL1. This approach identified SRF target genes whose activation was MKL-dependent. Twenty eight of 150 serum-inducible genes were found to be MKL-dependent, but unfortunately none of them coincided with the SRF targets discovered in our study.
ChIP assays have begun to reveal co-regulated groups of genes that share cis-regulatory elements and bind the same transfactors (reviewed in Ref. 66). Young and co-workers (67) have applied this approach to microarray expression profiling data in yeast at different stages in the cell cycle. They discovered not only known protein-binding motifs that regulate genes in the cell cycle but also completely new cis-acting elements that could be assigned to other groups of genes with related functions. The great advantage of combining global transcription factor binding analysis with expression profiling will be that the direct targets of transcription factors can be distinguished from indirect downstream effects, all of which are observed if, for example, expression alone is analyzed (47). We developed a systematic nonbiased search for downstream genes that may be directed through SRF ChIP analysis. The PAT-ChIP will be harnessed to explore regulatory network of other transfactors in the developing cardiovascular system.