Combinatorial expression of GATA4, Nkx2-5, and serum response factor directs early cardiac gene activity.

Herein, the restricted expression of serum response factors (SRF) closely overlapped with Nkx2-5 and GATA4 transcripts in early chick embryos coinciding with the earliest appearance of cardiac alpha-actin (alphaCA) transcripts and nascent myocardial cells. The combinatorial expression of SRF, a MADS box factor Nkx2-5 (a NK4 homeodomain), and/or GATA4, a dual C4 zinc finger protein, in heterologous CV1 fibroblasts and Schneider 2 insect cells demonstrated synergistic induction of alphaCA promoter activity. These three factors induced endogenous alphaCA mRNA over a 100-fold in murine embryonic stem cells. In addition, the DNA-binding defective mutant Nkx2-5pm efficiently coactivated the alphaCA promoter in the presence of SRF and GATA4 in the presence of all four SREs and was substantially weakened when individual SREs were mutated and or serially deleted. In contrast, the introduction of SRFpm, a SRF DNA-binding mutant, blocked the activation with all of the alphaCA promoter constructions. These assays indicated a dependence upon cooperative SRF binding for facilitating the recruitment of Nkx2-5 and GATA4 to the alphaCA promoter. Furthermore, the recruitment of Nkx2-5 and GATA4 by SRF was observed to strongly enhance SRF DNA binding affinity. This mechanism allowed for the formation of higher ordered alphaCA promoter DNA binding complexes, led to a model of SRF physical association with Nkx2-5 and GATA4.

A theme is emerging in which the appearance and diversification of nascent embryonic cardiac and smooth muscle cells requires the combinatorial interactions of serum response factor (SRF) 1 with other transcription factors enriched in early progenitor cells. A feature of a large number of cardiac and virtually all of the smooth muscle-expressed genes to date is their dependence upon a consensus sequence CC(A/T) 6 GG, a high affinity binding site for SRF. As a member of an ancient DNA-binding protein family, SRF shares a highly conserved DNA-binding/dimerization domain of 90 amino acids termed the MADS box. SRF serves as a versatile protein that binds to its cognate sites in a multitude of promoters to integrate intracellular signals and assists as a docking surface for the binding of accessory factors that may confer the regulation of specific gene programs (1)(2)(3)(4)(5)(6). SRF is a key regulator of immediate early gene expression, which frequently results in mitogenesis, and also a key regulator of myogenic terminal differentiation. Most SRF accessory proteins identified as coregulators of c-fos induction also act as endpoints of growth factor-induced signal transduction cascades. In contrast to the c-fos gene, which contains a single high affinity SRE, many muscle-specific genes including skeletal, cardiac, and smooth muscle ␣-actins contain combinations of at least three or more strong and weak affinity SREs that bind SRF in a highly cooperative manner and do not contain an adjacent ETS sequence (6 -8). Collateral accessory factors may then play essential roles in either facilitating or impeding SRF binding on multi-SRE muscle gene promoters, thus stimulating or repressing the transcription of SRF-dependent gene targets.
The recent homologous recombinant knock-out of the murine SRF gene locus supports the observation that SRF is absolutely required for the appearance of cardiac mesoderm during mouse embryogenesis (9). One of the SRF cofactors is Nkx2-5, a homeobox vertebrate homologue of Drosophila tinman (10,11) that is one of the earliest markers of vertebrate heart development and is important for the regulation of cardiac-restricted gene activity required for cardiac morphogenesis. We determined that the Nkx2-5 target sequences resemble the AT-rich central core of the serum response element and subsequently demonstrated that Nkx2-5 binds weakly to the consensus SREs of the ␣CA promoter. Nkx2-5 is also recruited by SRF to SREs, resulting in modest activation of endogenous ␣CA gene transcription (2,3,12,13). In addition, Nkx2-5 can associate with the dual C4 zinc finger transcription factor GATA4 to activate a variety of cardiac specified genes (14 -16). Likewise, SRF recruits GATA factors to activate SRE-containing myogenic and non-myogenic promoters (1,17,18). Recently, Nishida et al. (19) showed that the triad of SRF, Nkx-3.2, and GATA-6 was coexpressed in the medial smooth muscle layer of arteries. These factors transactivated the promoters of smooth muscle genes including ␤ 1 integrin, SM22, and caldesmon genes. This triad of factors provides transcriptional potency similar to the recently identified myocardin, a SAP factor, enriched during cardiogenesis, which also serves as a potent SRF coaccessory factor (20). Herein, we also provide strong evidence for SRF in playing a leading role in the commitment of cardiac progenitors by virtue of its obligatory requirement for acting as a myogenic restricted platform to interact with other early cardiac en-riched transcription factors, Nkx2-5 and GATA4. However, in this study, we show that Nkx2-5 and GATA4 facilitate potent activation of the cardiac ␣-actin promoter via the recruitment of these factors to multiple serum response elements, coincident with the earliest activation of the cardiac ␣-actin gene in promyocardial and embryonic stem cells.

MATERIALS AND METHODS
Recombinant Plasmids-The luciferase reporter vectors described previously (2,3,14) consist of the following plasmids: ␣CA-luciferase containing the SmaI-BstEII genomic fragment of the avian ␣CA promoter from Ϫ315 to ϩ15 relative to the transcription start site linked to the luciferase reporter gene; the deletion derivatives Ϫ200 del, Ϫ150 del, and Ϫ100 del; and the site-directed mutations over SRE1 (SRE1M), SRE2 (SRE2M), SRE3 (SRE3M), SRE4 (SRE4M), and Ϫ100 del ϩ SRE1M (Ϫ100 delϩ1M). The expression of Nkx2-5, SRF, and GATA4 was controlled by CMV promoter expression vectors described previously (14). Insect cell expression vector pAcX-SRF was kindly provided by Dr. Michael Gilman (Ariad Pharmaceuticals) (for review see Ref. 21). pAcX-GATA4 was constructed by excising a 726-bp fragment from pCG-GATA4 with XbaI and XcmI and a 600-bp band from pBS-GATA4-(1-126) with XcmI and BamH1. The two fragments were ligated into the pAcX vector after the removal of SRF from pAcX SRF with XbaI and BamH1. pAcX-Nkx2-5 was constructed by excising Nkx2-5 from pCGN-Nkx2-5 with XbaI and EcoRI, blunting the ends, and cloning into the pAcX vector after the removal of SRF from pAcX SRF with XbaI and BamH1 and blunting. pAct-Sp1 (22) (24). The synthesis of digoxigenin-labeled RNA probes corresponding to SRF, Nkx2-5, GATA4, and ␣CA and whole mount in situ hybridization were performed as described previously (24). All of the probes were alkalinehydrolyzed to the size of ϳ300 bp to enhance probe penetration. Cross sections of 10 m were cut from paraffin-embedded whole mounts.
Cell Culture, Transfection, and Luciferase Assays-CV1 fibroblasts were maintained in Dulbecco's modified Eagle's medium containing 10% newborn calf serum. Cells were grown to 50 -60% confluence in 6-cm diameter dishes and transfected as described previously (14). Drosophila Schneider 2 cells (S2) were grown in serum-free Sf-900 II serum-free medium (Invitrogen) at 26°C and passaged every week by diluting the culture 20-fold with the same medium. Before transfection, cells were placed into 12-well plates (ϳ5 ϫ 10 4 cells/well) and cultured overnight or until ϳ50% confluent. For each well, a total of 400 ng of plasmid DNA (200 ng of expression vectors and 200 ng of reporter vector) were mixed with 0.6 l of FuGENE 6 (Roche Molecular Biochemicals) in 100 l of Sf-900 medium and incubated for 10 -20 min at room temperature. This mixture was then added to the cells. After 48 h, the plates were centrifuged at 1000 ϫ g for 5 min, the supernatant was discarded, and the cells were lysed in 100 l of reporter lysis buffer (Promega). The processing of extracts and luciferase assays was done as described previously (14).
Nuclear Extract Preparation-Nuclear extracts of transfected cells were prepared by using a mini-extract procedure (25). The protein concentration was determined by the method of Bradford using a Bio-Rad kit.
Electrophoretic Mobility Shift Assay (EMSA)-EMSAs were performed with 20-l reaction mixtures at room temperature as previously described by Chen and Schwartz (2) in which 0.5 g of poly(dG⅐dC) was used as a nonspecific competitor. For EMSA antibody interference assays, proteins were incubated with cold antiserum for 5 min before the addition of the probe.
Affinity Purification of Protein Complexes with Protein A Fusion Proteins-Protein A pull-down assays were performed as described previously (14). Whole cell extracts from CV1 cells transfected with pCGN-Nkx2-5 and pCG-GATA4 together with plasmids expressing either protein A or protein A-SRF were lysed in EBC buffer, incubated with IgG-Sepharose beads (Amersham Biosciences) for 15 min at 4°C, and washed four times with 500 ul of EBC buffer. Proteins were solubilized in Laemmli's SDS-loading buffer, separated by SDS-PAGE, and visualized by immunoblotting with anti-HA and anti-GATA4 antibodies (Santa Cruz Biotechnology, Santa Cruz, CA).
DNA Affinity Chromatography Assay-Biotinylated ␣CA promoter DNA was reacted with translated 35 S-labeled SRF and GATA4. Casein kinase II was used to label Nkx2-5pm with [␣-32 P]ATP. DNA binding solution (50 mM NaCl, 15 mM HEPES, pH 7.6, 1 mM EDTA, 0.5 mM dithiothreitol, 5% glycerol) contained 5 mg/ml bovine serum albumin, 5 g of poly(dI⅐dC) as detailed by Cheng and Schwartz (2) with the addition of 5% fat-free milk. Avidin magnetic beads (Promega) at 100 g/assay point were concentrated with a magnetic rack (Promega) and repeatedly washed in binding buffer. After the complete removal of binding buffer, the beads were resuspended in SDS-loading buffer, split into two equal fractions, and analyzed by SDS-10% PAGE.

RESULTS
Overlapping SRF, GATA4, Nkx2-5 Expression Patterns Coincided with ␣CA Gene Activity in Early Heart Development-SRF, GATA4, and Nkx2-5 gene expression patterns were mapped to determine whether they coincided with that of cardiac restricted ␣CA gene activity, a marker for committed cardiac myocytes. Whole mount-sectioned in situ hybridized chick embryos ( Fig. 1) revealed that by HH stage 8 (ϳ4 somites), cNkx2-5, cSRF, cGATA4, and ␣CA have overlapping patterns of expression with the anterior intestinal portal (AIP) marking the apex of expression (Fig. 1, A-D). However, SRF and ␣CA are not expressed in the medial AIP ( Fig. 1, B and D, d and j). At the level of the AIP, cNkx2-5 was expressed throughout the mesoderm and endoderm (Fig. 1E, a), whereas GATA4 was expressed only in the endoderm (Fig. 1G, g). In the paired heart tubes, both genes were expressed in the mesoderm (Fig. 1C, h) The first somite marks the posterior border of cSRF and cGATA4 expression (Fig. 1, F and G), but cNkx2-5 expression barely extends to this level (Fig. 1E, c). Thus, SRF expression appears to mark the anterior border of ␣CA expression, whereas cNkx2-5 expression appears to coincide with the posterior border. The approximate overlap of SRF mRNA with Nkx2-5 and GATA4 transcripts was coincident with ␣CA gene activity and the appearance of the myocardial cells (as shown in Fig. 1b, e, h, and k). At HH stage 10 (ϳ10 somites), the heart tube is fused at the midline, and cNkx2-5, cSRF, and ␣CA were expressed throughout the straight heart tube ( Fig. 1, E, F, and H), whereas cGATA4 expression is primarily in the sinus venosus. By HH stage 10, all four genes were expressed in the myocardium.
SRF Cooperated with GATA4 and Nkx2-5 to Activate the ␣CA Promoter-The reporter plasmid ␣CA-luciferase contains the avian ␣CA promoter from Ϫ315 to ϩ15 relative to the transcription start site. When cotransfected with SRF alone, ␣CAluciferase activity varied between 6-and 10-fold, whereas GATA4 or Nkx2-5 alone resulted in only 1-4-fold transcriptional activation in CV1 fibroblasts. The pairings of SRF and GATA4 resulted in the coactivation of ϳ20 -30-fold above the basal level. The transfection of all of the three factors resulted in robust synergistic activation of the ␣CA promoter to levels up to 160-fold ( Fig. 2A). In contrast, no coactivation was seen with the SV40 promoter (data not shown). Fig. 2B shows robust expression of plasmid DNA expression vectors driving wild-type gene activity of transfected CV1 fibroblasts as detected by antibody protein blots. These results suggest that most of the synergistic coactivation requires the efficient binding of SRF to the SRE. Consistent with this idea, the cotransfection of SRFpm1, a triple point mutant of SRF that weakens DNA binding with GATA4 and Nkx2-5, resulted in a 75% reduction in coactivation (Fig. 2, A and C). To investigate whether triad factor coactivation requires the binding of Nkx2-5 to the ␣CA promoter, the non-DNA-binding mutant Nkx2-5pm, a point mutation (Asn to Gln) at position 10 of the DNA recognition helix, was substituted for wild-type Nkx2-5 in the cotransfection assay with SRF and GATA4 (Fig. 2, A and C). No reduction in activity was noted. In fact, a small increase of ϳ15% reporter activity was seen. These findings suggest that as shown for the interaction of Nkx2-5 with SRF (2), the binding of Nkx2-5 to the ␣CA promoter is not required and may actually decrease the efficiency of the coactivation with SRF and GATA4.
It is conceivable that the interaction of Nkx2-5 with SRF and GATA4 allows the activation domain of Nkx2-5 to contribute significantly to the synergistic effect as demonstrated in cotransfection assays of Nkx2-5 plus GATA4 (14,15,26). To investigate whether other members of the Nk2 family are also able to cooperate with SRF and GATA4 in a triple interaction, cotransfections were performed as described above, substituting Nkx2-5 for Nkx2-1 (TTF1), Nkx2-8, and D. tinman (Fig.  2C). The results showed that although Nkx2-1 and D. tinman were able to replace Nkx2-5 albeit at a lower efficiency, Nkx2-8 was unable to cooperate with SRF and GATA4 in activating the ␣CA promoter, indicating nonequivalent roles for the Nkx2 factors.
Drosophila S2 cells, which are devoid of significant SRF expression (21), allowed us to study the activation of the ␣CA promoter in the absence of endogenous SRF. The transfection of SRF into S2 cells resulted in minimal (2-fold) activation of the ␣CA promoter (Fig. 2D). Transfection of VP16-SRF resulted in 190-fold activation of the same promoter (results not shown), indicating that this promoter binds SRF and is functional in S2 cells, whereas the natural activation domain of SRF is relatively weak in S2 cells. Similarly, GATA4 and Nkx2-5 by themselves or in combination did not significantly activate the ␣CA promoter in these cells. A double transfection of SRF with GATA4 or with Nkx2-5 resulted in activation below the additive level (Ͻ3 times). However, triple cotransfection of SRF, GATA4, and Nkx2-5 resulted in synergistic (13-fold) stimulation of the ␣CA promoter. We also observed that the removal of the strong inhibitory C-terminal domain in the Nkx2-5 mutant Nkx2-5-(1-203) was by itself a weak activator of ␣CA (3 times) but a much stronger coactivator with SRF (18-fold) and with GATA4 and SRF (40-fold). These results together with the cotransfection experiments in CV1 fibroblasts (Fig. 2, A and C) (for review see Ref. 14) suggest that GATA4 activated Nkx2-5 by the removal of the repressive effect of the C-terminal domain of Nkx2-5. Moreover, the interaction of Nkx2-5 and GATA4 on the ␣CA promoter was mediated by exogenous SRF in insect cells, because the level of activation in the absence of SRF is minimal.
Sp1 has been reported to physically interact with GATA1 (27-29), Nkx2-1 (30), and MEF2C (31) and to functionally interact with SRF to regulate the ␣CA promoter (32)(33)(34). Because S2 cells have been shown to lack Sp1 (22), we then asked whether Sp1 interacts with these factors. When Sp1 was cotransfected into S2 cells with a minimal promoter containing multimerized Sp1 sites, luciferase activity was 160-fold compared with the same promoter without Sp1 sites (results not shown). However, Sp1 alone or with SRF had minimal effect on ␣CA-luciferase (Fig. 2D). Sp1 had only an additive effect on the  f, i, and l). a, Nkx2-5 is expressed in the pharyngeal endoderm and promyocardium at the level of AIP. b, Nkx2-5 expression detected in pharyngeal endoderm and promyocardium and weaker expression detection in the underlying endoderm tube at the level of the heart-forming region. c, at the level of the first somite, no detected Nkx2-5 expression. d, SRF detected in the neural tube at the AIP level. e, SRF expression detected in the promyocardium and neural tube at the heart-forming region. f, SRF expression at the first somite level in the splanchic mesoderm and neural tube. g, at the level of AIP, GATA4 expressed in the endoderm underlying the promyocardium but not in the promyocardium. h, in the heart-forming region, GATA4 was expression in the promyocardium. i, at the first somite level, GATA4 expression was detected in the splanchnic mesoderm. ␣CA expression was not detected at the level of the AIP (j) or at the level of the first somite shown by the arrows (l). k, at the level of the heart-forming region, ␣CA was detected only in the promyocardium and nowhere else. aip, anterior intestinal portal; ec, ectoderm; en, endoderm; h, heart; m, mesoderm; nt, neural tube; pe, pharyngeal endoderm; pm, promyocardium; ps, primitive streak; so, somite; tel, telenchephalon; vv, viteline vein. coactivation of the ␣CA promoter by SRF plus GATA4 and reduced coactivation by SRF plus GATA4 and Nkx2-5 (Fig. 2D). These results indicate that Sp1 does not stimulate ␣CA transcription via interactions through this triad of factors.
SRF, Nkx2-5, and GATA4 Activated ␣CA Gene Activity in Murine ES Cells-Is the coexpression of Nkx2-5, GATA4, and SRF capable of activating the endogenous ␣CA gene? The expression vectors encoding these transcription factors were transiently transfected into pluripotent murine AB2-129 embryonic stem cells. ␣CA mRNA was not detected in the parental 6-day post-aggregated ES cells transfected with the empty pCGN plasmid as shown in Fig. 3. Increased expression of any one of these factors in ES cells reproducibly elicited significant transcription of the ␣CA gene. SRF and GATA4 alone were sufficient to increase ␣CA transcripts by 40-fold over control ES cells. The paired groupings of SRF with GATA4 or with Nkx2-5 stimulated ␣CA activity ϳ30% more than the combination of Nkx2-5 and GATA4 (Fig. 3). The combination of SRF, Nkx2-5, and GATA4 induced ␣CA mRNA levels over 100-fold versus control levels, indicating a central role for SRF in directing embryonic cardiac gene expression.
Nkx2-5 and GATA4 Facilitated Formation of Multimeric SRF DNA Binding Complexes-How does Nkx2-5, GATA4, and SRF drive ␣CA activity? SRF binds to a variety of SREs including the four avian ␣CA SREs and to the greatest extent to the first (SRE1) and fourth (SRE4) SREs. Non-consensus second (SRE2) and third (SRE3) SREs contain a single GC substitution in the six AT-rich SRE core, and all of these SREs are highly conserved across the evolution of Aves and mammals (35,36). SRE2 and SRE3 are weak SRF binding sites that resemble preferred Nkx2-5 binding elements (2) and also contain overlapping inhibitory YY1 binding sites (3). Several ␣CA promoter mutants (Fig. 4) linked to the luciferase reporter gene were cotransfected with the three cofactors in CV1 fibroblasts to assay their transcriptional activities (Fig. 4). The wild-type promoter activity in these fibroblasts was set at one and activated ϳ58-fold with the SRF, Nkx2-5, and SRF. The mutations over individual SRF binding sites such as SRE1 (57%), SRE2 (35%), and SRE3 (48%) caused substantial inhibition of ␣CA promoter activity. Although the mutation of the distal SRE4 FIG. 2. SRF cooperated with GATA4 and Nkx2-5 and activated the ␣CA promoter in CV1 fibroblasts and Drosophila S2 cells. SRF, GATA4, and Nkx2-5 expression vectors were cotransfected with 1 g of the indicated luciferase reporter construct. pCGN-SRF and pCGN-SRFpm (150 ng), pCGN-Nkx2-5 and pCGN-Nkx2-5pm (400 ng), or pCG-GATA4 (400 ng) were added in various combinations with an appropriate amount of pCG empty vector to make the total mass of pCG-derived vectors equal to 1 g. CV1 cells were transfected at ϳ50% confluence, and extracts collected after 48 h culture in Dulbecco's modified Eagle's medium were transfected with 3% horse serum and 15 g/ml of insulin. The luciferase activity of the empty vector alone was set as 1. Error bars represent the mean Ϯ S.D. of at least two separate assays. B, Western blot of transfected CV1 fibroblasts with 2 g of pCGN-SRF, pCGN-Nkx2-5, or pCG-GATA4. After 72 h, extracts were prepared in EBC buffer, and proteins were separated by SDS-PAGE. Immunoblotting was done with anti-HA (lanes 1-5) or GATA4 (lanes 6 -10). HA antibody nonspecifically labels three cellular proteins (asterisk) that indicate equal loading of the gel and specifically show SRF and Nkx2-5 expression. The indicated expression plasmids were cotransfected with ␣CA-SRE1 into CV1 cells and assayed as in A. The full-length wild-type sequences are represented as wt, whereas pm represents point mutants that are unable to bind DNA. C, other members of the Nk2 family cotransfected with SRF and GATA4 in a triple interaction were performed as above, substituting Nkx2-5 for Nkx2-1 (TTF1), Nkx2-8, and D. tinman. D, Drosophila Schneider 2 embryonic cells were transfected in 12-well plates with 200 ng of ␣CA-luciferase and 50 ng of each of the indicated insect cell expression vectors or empty vector to a total of 200 ng of plasmid DNA. The full-length wild-type sequences are represented. 1-203 represents Nkx2-5 with amino acids C-terminal to residue 203 removed. After 48 h post-transfection, luciferase activity was assayed and compared with the activity obtained by transfection with the empty expression vector alone. Error bars represent the mean Ϯ S.D. of at least two assays. did not significantly alter transcription, the removal of the promoter sequence (Ϫ310 to Ϫ200), which included the distal SRE4, decreased promoter activity to 56% of the wild-type value. A further removal of additional 50-bp promoter sequence, construct del-150 that consists of the weak SRE2 and the strong SRE1, reduced the promoter activity to the lowest level approximately equal to Ϫ100 del with a mutated SRE1 (Ϫ100ϩSRE1M). The deletion of SRE2 to Ϫ100 del tripled the promoter activity. These results suggest that a negative regulatory element between Ϫ150 and Ϫ100 is probably SRE2, and that negative-acting YY1 binding to this element is dominant over the positive regulators SRF Nkx-2.5 and GATA4. In addition, we found that the mutant Nkx2-5pm efficiently coactivated in the presence of all four SREs but was substantially weakened when individual SREs were mutated and or deleted. In contrast, the introduction of SRFpm blocked the activation with all of the ␣CA promoter constructions. Thus, these assays indicate the importance for cooperative SRF binding in facilitating the recruitment of nonbinding Nkx2-5pm mutant and GATA4.
In Fig. 5, EMSA was performed with a suboptimal dose of SRF resulting in a barely detected single-shifted complex. Increased amounts of Nkx2-5 homeodomain added to a constant amount of SRF spurred SRF binding and induced the appearance of two additional slowly migrating complexes. Even though GATA4 was unable to bind to the ␣CA promoter, it was also able to facilitate SRF binding as a single complex. However, Nkx2-5 and GATA4 in combination with low dosage of SRF strongly increased SRF binding, which resulted in fourwell resolved shifted complexes, indicating complete occupancy of strong and weak SREs. Anti-SRF antibodies supershifted SRF and weakened the appearance of the multimeric complexes. In comparison, anti-GATA4 weakened overall binding but failed to supershift the primary and the secondary complexes. Together, anti-GATA4 and anti-SRF blocked the formation of both primary and secondary SRF multifactorial shifted complexes, thus demonstrating a dependence of Nkx2-5 and GATA4 in facilitating SRF binding.
SRF Recruited GATA4 and Nkx2.5 to the Cardiac ␣-Actin Promoter-Previous studies demonstrated that a fusion protein of GATA4 and protein A was able to interact with SRF and with Nkx2-5 (1,14). Similar experiments were performed with SRF fused to protein A. CV1 fibroblasts were cotransfected with protein A-SRF (or protein A alone as a control), GATA4, and Nkx2-5. Protein complexes interacting with protein A proteins were purified from whole cell extracts on IgG-Sepharose beads. Immunoblots from unprocessed extracts show similar levels of expression of protein A, protein A-SRF, GATA4, and Nkx2-5 (Fig. 5B, lanes 1-3 and 7-8). Unexpectedly however, protein A-SRF was able to specifically interact with GATA4 only in the presence of Nkx2-5 (Fig. 3B, lanes 11 and 10). Perhaps the N-terminal protein A moiety interferes with the ability of SRF to interact directly with GATA4. This result suggests that GATA4 is recruited to the complex by Nkx2-5 bound to protein A-SRF, lending support to the hypothesis that a physical complex of SRF, GATA4, and Nkx2-5 occurs in vivo. Fig. 5C shows an experiment in which SRF attracted Nkx2-5 and/or GATA4 to biotin-linked ␣CA promoter DNA pulled out of the binding reaction with avidin magnetic beads. Weakly binding SRF mutant SRFpm interfered with the recruitment of Nkx2-5 and GATA4 to biotinylated ␣CA DNA. Also, the DNAbinding defective mutant Nkx2-5pm was pulled out only by SRF wild type, not by mutant SRFpm. Thus, Nkx2-5 and GATA4 brought about robust synergy with SRF by strongly enhancing the formation of promoter DNA binding complexes. DISCUSSION In vertebrates, SRF expression is restricted to tissues of mesoderm and neuroectoderm origins (3,37). During chicken embryogenesis and the progression of gastrulation, strongly localized SRF mRNA expression was observed in the lateral plate mesoderm, the precardiac splanchic mesoderm, the myo- FIG. 4. Multiple SREs bestowed cooperative dependent and maximal ␣CA promoter activity. A, schematic diagram of the cardiac ␣-actin promoter SRE mutations and serial deletion constructs were previously described in Chen and Schwartz (2,12). SREs were indicated by closed boxes. B, transfection analysis of various ␣CA promoter mutant reporter constructs in CV1 fibroblasts is shown. The luciferase activity of full-length ␣CA promoter in these cells was set at 1 and designated as control (C). Shown is cotransfection analysis of reporter constructs with SRF, Nkx-2.5, and GATA4 expression vectors designated as condition 1, SRF plus defective DNA-binding mutant Nkx2-5pm plus GATA4 designated as condition 2, and defective SRF DNA-binding mutant SRFpm plus Nkx2-5 plus GATA4 as designated condition 3. cardium, and the myotomal portion of the somites (37). We have shown that SRF is expressed at high levels in a symmetrically split crescent capping the extreme anterior and lateral parts of the embryo appearing like the split cardiac progenitor cell populations described from fate mapping experiments (37). The initial expression pattern resolves into a complete crescent and undergoes changes consistent with morphogenesis of the linear and S-shaped heart tube. In the mouse, we showed that the highest SRF mRNA levels were seen in adult skeletal and cardiac muscle. During mouse embryonic development, SRF transcripts were found to be enriched in smooth muscle media of the vessels, the myocardium of the heart, and myotome portion of somites (1). Recently, the homologous recombinant knock-out of the murine SRF gene demonstrated a severe block for mesoderm formation during mouse embryogenesis (9). 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. They consist of misfolded ectoderm and endoderm cell layers that do not form primitive streak or any detectable mesoderm and fail to express the very early developmental marker genes Bra(T), Bmp2/4, and Shh (9).
We observed that the overlap of SRF transcripts with Nkx2-5 and GATA4 transcripts coincided with ␣CA gene activity and the appearance of the early committed myocardial cells of the heart (Fig. 1). The consequence of expression of any one of these transcription factors in ES cells was adequate to stimulate significant endogenous ␣CA gene activity, whereas the combination of SRF, Nkx2-5, and GATA4 induced ␣CA RNA levels over 100-fold versus control levels (Fig. 3). The coactivation was dependent on intact SRF as shown by the dominant inhibition of ␣CA expression with mutant SRFpm (Figs. 2 and 4). In addition, Nishida et al. (19) showed that the cardiac musclespecific triad of Nkx2-5, serum response factor, and GATA-4 transactivated the cardiac atrial natriuretic factor gene, which contains three separate elements, a CArG-like box, a GATA binding box, and a NK binding element. These observations demonstrate that transcripts emanating from the endogenous ␣CA and ANF genes can be increased by elevated levels of SRF, GATA4, and Nkx2-5 and lend further support to the notion that the combination of these three factors are among the earliest endogenous activators of cardiac gene transcription.
A Model of GATA4 and Nkx2-5 Interaction with SRF-SRF serves as a versatile protein that binds to its cognate sites in a   FIG. 5. Nkx2-5 and GATA4 recruited by SRF facilitates multimeric SRF binding. A, electrophoretic mobility shift assays were performed as described by Chen and Schwartz (13) with minor modifications. The DNA-protein complexes were fractionated on 5% polyacrylamide gels that were processed for autoradiography x-ray film. Pb, probe; Sab, SRF antibody; Gab, GATA4 antibody; S, SRF; N, Nkx2-5 GATA4; G, GATA4; trp, triple factors. B, CV1 fibroblasts were transfected with vectors coding for either protein A or protein A-SRF together with pCG-GATA4 and with or without pCGN-Nkx2-5. Equivalent levels of expressed proteins are shown on Western blots of whole cell extracts (input lanes 1-3 and 7-9). Complexes interacting with protein A or protein A-SRF were purified on IgG-Sepharose, washed with EBC buffer, and visualized by immunoblotting with the indicated antibodies (lanes 4 -6 and lanes 10 -12). Protein A and protein A-SRF can be visualized by their ability to nonspecifically bind the primary and secondary antibodies used through the Fc receptors. Input lanes are overexposed to show protein A-SRF bands. GATA4, protein A, protein A-SRF (pA-SRF) are indicated with arrows. Other bands not indicated with an arrow represent nonspecific binding of the antibodies in input lanes and IgG light chains on lanes 4 -6. Proteins were separated on SDS-10% PAGE and processed for autoradiography on x-ray film. C, DNA recruitment assay was carried out with translated [ 35 S]methionine-labeled SRF and GATA4 proteins. Casein kinase II [ 32 P]ATP-labeled Nkx2.5pm is shown. DNA binding assays were performed as detailed by Chen and Schwartz (2) with the addition of 5% fat-free milk. Prewashed magnetic beads linked with avidin were incubated at room temperature with DNA binding solution containing the binding reactions that were washed, split into two equal fractions, analyzed by SDS-10% PAGE, and exposed to two layers of autoradiographic film (Kodak X-Omat AR). Proximal film layer recorded 35 S signal of SRF and GATA4 within 24 h and distal film layer recorded 32 P signal of  multitude of promoters and serves as a docking surface for the binding of many different accessory proteins that confer specific functional abilities to the promoter (2,4,38). The elucidation of the x-ray crystal structure of the SRF core bound to DNA (39,40) provides an explanation for mutually inclusive binding of coaccessory factors to a single SRE. As shown schematically in Fig. 6, the coiled coil formed by the MADS box ␣I helices (amino acids 153-179) lies parallel and on top of a narrow DNA major groove making contacts with the phosphate backbone on a SRE half-site. In addition, the unstructured N-terminal extension from the ␣I helix (amino acids 132-152) makes critical base contacts in the minor groove. The dimerization of the MADS box occurs above the ␣I helix by a structure composed of two ␤-sheets in the monomer that interact with the same unit in its partner. A second ␣II helix in the C-terminal portion of the MADS box stacked above these ␤-sheets completes this stratified structure. Touching the MADS box at either the first or second ␣-coils may play important roles in determining whether gene activity is directed toward cell differentiation or cell growth.
The recruitment of Nkx2-5 to a SRE is dependent upon SRF DNA binding activity. The dominant negative SRFpm1 mutant, which dimerizes with wild-type SRF, blocked the recruitment of Nkx2-5. Even though Nkx2-5 can bind weakly to some SREs, we found that the activation of a minimal promoter consisting of a single SRF binding site was dependent upon increasing the cellular levels of SRF (3,37). When Nkx2-5 binding activity was blocked by a point mutation in the third helix of the homeodomain, SRF was still capable of recruiting mutated Nkx2-5 to the ␣CA promoter (2). An investigation of protein-protein interactions demonstrated that Nkx2-5 could bind to SRF in the absence of DNA as soluble protein complexes isolated from cardiac myocyte nuclei. In addition, Nkx2-5 and SRF complexes could be detected as coassociated binding complexes on the proximal SRE1 (2). The recruitment of Nkx2-5 to a SRE was dependent upon SRF DNA binding activity and could be blocked by the dominant negative SRFpm mutant, which dimerizes with wild-type SRF monomers but cannot itself bind to DNA. In addition, Nkx2-5 protein and SRF interact directly in the absence of the SRE. A short 30-amino acid peptide (amino acids 142-171), which encompasses the basic region of SRF in the ␣1 coil of the MADS box, is sufficient to mediate protein-protein contacts with the Nkx2-5. The N terminus/helix I and helix II regions of the Nkx2-5 homeodomain interacts with the MADS box as modeled in Fig. 6.
Nkx2-5 can also cooperate with GATA4 to activate the ␣CA (14), ANF (15,16,26), and minimal A20 (14) promoters containing multimerized Nkx2-5 DNA binding sites. Transcriptional activity requires the N-terminal activation domain of Nkx2-5 and binding activity through its homeodomain, but it does not require the activation domain of GATA4. Minimal interactive regions were mapped to the homeodomain of Nkx2-5 and the second zinc finger of GATA4. The removal of Nkx2-5 C-terminal inhibitory domain stimulates robust transcriptional activity compared with the effects of GATA4 on wild-type Nkx2-5, which in part facilitates Nkx2-5 DNA binding activity. We postulated a simple model that GATA4 induces a conformational change in Nkx2-5 that displaces the C-terminal inhibitory domain; thus, eliciting transcriptional activation of promoters containing Nkx2-5 DNA binding targets. Also, coexpression of both GATA4 and SRF in fibroblasts resulted in robust activation of both muscle-restricted and ubiquitous SREdependent promoters. GATA4 second zinc finger binds avidly to the ␣I coil of the MADS box. Interestingly, deletion of the N-terminal activation domain and the first zinc finger of GATA4 increased the ability of GATA4 to synergize with SRF, suggesting that the domain surrounding the first zinc finger may interfere with interaction of SRF and GATA4. This notion is supported by recent reports describing interaction the multizinc finger coaccessory proteins such as FOG-1 and FOG-2 which inhibit the transcriptional activity of GATA-1 and GATA4 by interacting with the first zinc finger (41)(42)(43) which in Drosophila is also similar to the interaction of pannier with U-shaped (44).
The coexpression of all three proteins resulted in robust synergistic activation of the ␣CA promoter two orders of magnitude above base line. The combination of Nkx2-5 homeodomain and GATA4 enhanced the formation of SRF dependent DNA binding complexes. Possibly, conformational changes in SRF structure facilitated by Nkx2-5 and GATA4 made SRF a more efficient DNA binding factor; allowing it to bind to weaker nonconsensus SREs and, stimulating ␣CA activity under limiting amounts of SRF. This level of activation required intact SRF-SRE binding, but was even higher when Nkx2-5pm was  (40) who described the x-ray crystallographic structure of structure of serum response factor core bound to DNA. The paired monomers form a coiled coil by the MADS box ␣I helices (amino acids 153-179) and lies parallel and on top of a narrow DNA major groove making contacts with the phosphate backbone on a SRE halfsite, whereas an unstructured N-terminal extension from the ␣I helix (amino acids 132-152) makes critical base contacts on paired Gs in the minor groove. Dimerization of the MADS box occurs above the ␣I helix by a structure composed of two ␤-sheets in the monomer that interact with the same unit in its partner. A second ␣II helix in the C-terminal portion of the MADS box stacked above these ␤-sheets completes this stratified structure. B, the second zinc finger of GATA4 binding to the N-terminal extension and ␣-coil I of the MADS box. Removal of the first zinc finger of GATA4 actually stimulated GATA4 synergy with SRF by the removal of interfering multi-zinc finger coaccessory inhibitory proteins such as FOG-1 and FOG-2. C, the binding of Nkx2-5 homeodomain through its helix I/II to the other monomer of SRF and the interaction with GATA4 through helix III. Note that helix III may be positioned to bind the exposed major groove at the center of the SRE. Nkx2-5 and SRF interact directly in the absence of the SRE. A short 30-amino acid peptide (amino acids 142-171), which encompasses the basic region of SRF in the ␣1 coil of the MADS box is sufficient to mediate proteinprotein contacts with the Nkx2-5. The N terminus/helix I and helix II regions of the Nkx2-5 homeodomain interact with the MADS box. The fact that Phox-1⅐SRF (45,46) and Nkx2-5⅐SRF (2) interactions required N-terminal arm/helix 1/helix 2 region of the homeodomain lead us to propose an earlier model that helix1/2 is responsible for homeodomain-SRF interactions (2). As shown above in C, this model shows GATA factors such as GATA4 binding to a monomer of SRF and Nkx2-5 binding to the other SRF monomeric subunit. Cross-linking assays from our laboratory indicated that a monomer of Nkx2-5 was bound to a SRF dimer (C. Y. Chen and R. J. Schwartz, unpublished observation). Thus, this model represents our data presented here and elsewhere that demonstrated Nkx2-5 and GATA4 coassociation with SRF in solution and on SRE DNA binding targets. substituted for wild-type Nkx2-5. The lack of requirement for Nkx2-5 binding to DNA is also observed with Nkx2-5/SRF coactivation of the ␣CA-SRE1 promoter (2,3). This may indicate that Nkx2-5 acts by providing a strong transcriptional activation domain, whereas the role of SRF is to attract Nkx2-5 to the ␣CA promoter and facilitate the recruitment of GATA4. We did not observe significant synergistic activity of SRF and Nkx2-5 in insect cells. However, the deletion of the C-terminal repressor domain of Nkx2-5 resulted in increased cooperative activation with SRF similar to the levels obtained with coexpressed GATA4, SRF, and Nkx2-5. These results suggest that the interaction with SRF is not sufficient to remove the repressive effect of the C-terminal domain and that interaction with GATA4 (recruited by SRF or Nkx2-5) is required for coactivation. This effect of GATA4 on the repressor domain of Nkx2-5 may thus explain why triple coexpression resulted in much greater activation of the ␣CA promoter.
Recently, Wang et al. (20) identified myocardin, a novel and highly potent transcription factor that is expressed in cardiac and smooth muscle cells. Myocardin belongs to the SAP domain family of nuclear proteins and activates cardiac muscle promoters by recruitment to SRF. By virtue of the ability of myocardin to restructure chromatin, it is probable that myocardin could further stimulate the ability of this triad to transactivate target genes. In conclusion, the reciprocal recruitment among Nkx2-5, GATA4, and SRF may expand the spectrum of genes regulated by either one of these factors while conferring additional level of specificity. Our study underscores the ability of these proteins to interact in a combinatorial manner to drive cardiogenic gene expression programs.