Myocardin-dependent Activation of the CArG Box-rich Smooth Muscle γ-Actin Gene

Serum response factor (SRF) is a ubiquitously expressed transcription factor that binds a 10-bp element known as the CArG box, located in the proximal regulatory region of hundreds of target genes. SRF activates target genes in a cell- and context-dependent manner by assembling unique combinations of cofactors over CArG elements. One particularly strong SRF cofactor, myocardin (MYOCD), acts as a component of a molecular switch for smooth muscle cell (SMC) differentiation by activating cytoskeletal and contractile genes harboring SRF-binding CArG elements. Here we report that the human ACTG2 promoter, containing four conserved CArG elements, displays SMC-specific basal activity and is highly induced in the presence of MYOCD. Stable transfection of a non-SMC cell type with Myocd elicits elevations in endogenous Actg2 mRNA. Gel shift and luciferase assays reveal a strong bias for MYOCD-dependent transactivation through CArG2 of the human ACTG2 promoter. Substitution of CArG2 with other CArGs, including a consensus CArG element, fails to reconstitute full MYOCD-dependent ACTG2 promoter stimulation. Mutation of an adjacent binding site for NKX3.1 reduces MYOCD-dependent transactivation of the ACTG2 promoter. Co-immunoprecipitation, glutathione S-transferase pulldown, and luciferase assays show a physical and functional association between MYOCD and NKX3.1; no such functional relationship is evident with the related NKX2.5 transcription factor despite its interaction with MYOCD. These results demonstrate the ability of MYOCD to discriminate among several juxtaposed CArG elements, presumably through its novel partnership with NKX3.1, to optimally transactivate the human ACTG2 promoter.

Smooth muscle cell (SMC) 6 differentiation requires synchronized expression of general and cell-restricted cytoskeletal and contractile genes encoding proteins conferring the unique contractile activity of this cell type (1). Many SMC differentiation genes contain proximal cis regulatory elements, called CArG boxes, which bind the broadly expressed transcription factor, serum response factor (SRF) (2). SRF is a weak activator of gene expression, requiring physical interactions among over 60 cofactors that themselves recruit additional proteins to fully direct unique patterns of gene expression. The discovery of one such SRF cofactor, called myocardin (MYOCD), has revolutionized our understanding of the basic transcriptional mechanisms underlying the specification of a differentiated SMC phenotype (3). Myocd mRNA is restricted mainly to cardiac and SMC and its encoded protein is one of nature's most powerful transactivators of gene expression (3), although the level of SMC gene activation appears to be both cell and promoter context-dependent (4). MYOCD elicits structural and functional attributes of SMC suggesting this cofactor has characteristics of a master regulator of the differentiated SMC phenotype (5). Consistent with this idea, conventional inactivation of Myocd results in midgestational arrest at embryonic day 10.5 of the mouse because dorsal aortic SMC are poorly differentiated (6), a finding that has recently been extended to neural crest-derived SMC of the great arteries (7).
In theory, SRF binds up to 1,216 permutations of the CArG box (8) with the consensus forms (CCW 6 GG) binding SRF more avidly than non-consensus forms (9). Many SMC contractile genes contain two or more CArG elements in either their 5Ј promoter or intronic regions (2,10). Olson and colleagues (11) proposed a model wherein MYOCD forms a molecular bridge over two or more CArG boxes by undergoing homo-oligomerization through its leucine zipper-like domain thus maximizing transactivation of SMC contractile genes harboring more than one CArG element. According to this model, SMC cytoskeletal/contractile genes with one CArG box will only be activated moderately by MYOCD. On the other hand, several multi-CArG containing genes are refractory to high level MYOCD-dependent transactivation, which likely is attributable to sequences immediately flanking CArG elements (3,(12)(13)(14)(15)(16). Thus, the number of CArG boxes, the sequence character therein, and flanking sequences appear to be key determinants of MYOCD-mediated transcriptional activity.
SMC are often defined structurally and physiologically as vascular or visceral SMC. Although expressing comparable levels of many cytoskeletal/contractile genes, microarray data indicate unique gene expression signatures between both SMC types (17). For example, the SRF-and MYOCD-dependent telokin gene is abundantly expressed in visceral SMC with lowlevel expression in vascular SMC (18). However, a chimeric promoter of telokin harboring a fragment of the Tagln promoter confers vascular activity in transgenic mice suggesting that SMC promoters harbor important sequence content for vascular versus visceral SMC expression (19). Two smoothelin isoforms exhibit SMC type-restricted expression in adults with smoothelin A expressed predominantly in visceral SMC and smoothelin B in vascular SMC (20). Smoothelin A contains conserved CArG elements that are mildly responsive to MYOCD, whereas smoothelin B, without conserved CArG boxes, is refractory to MYOCD transactivation in vitro (15).
The murine smooth muscle ␥-actin gene (Actg2) is expressed abundantly in visceral SMC with lower levels in vascular SMC (21). Transgenic mice carrying 13.7 kb of 5Ј promoter and 3Ј intronic mouse Actg2 sequences similarly exhibit visceral SMC activity (22). In this report, we have revisited expression of Actg2 in vascular SMC and evaluated the role of MYOCD in regulating human ACTG2 promoter activity in vitro. Results demonstrate Actg2 mRNA and protein expression as well as specific promoter activity in vascular SMC. Despite the presence of four closely juxtaposed CArG boxes in the 5Ј proximal region of ACTG2, MYOCD appears to preferentially utilize a single CArG element (CArG2) that confers high-level promoter activity through elevations in SRF-MYOCD binding. Importantly, an adjacent binding site for NKX3.1 is necessary for full transactivation, and co-transfection reporter assays of MYOCD show NKX3.1 potentiates MYOCD-dependent stimulation of the human ACTG2 promoter. Co-immunoprecipitation, GST pulldown, and luciferase reporter assays suggest this activity occurs through physical contacts between the C-terminal domain of MYOCD and NKX3.1. These results reveal a CArG bias for MYOCD transactivation of Actg2 gene transcription and suggest this discrimination stems from the novel interaction of MYOCD with the NKX3.1 homeodomain protein.

EXPERIMENTAL PROCEDURES
Cell Culture-Primary-derived rat aortic smooth muscle cells and the PAC-1 and A7r5 SMC lines, as well as NIH 3T3, CV-1, COS-7, HeLa, BC 3 H1, L6 myoblasts, and a human uterine SMC line (SKLMS, ATCC) were maintained in Dulbecco's modified Eagle's medium containing high glucose, supplemented with 10% fetal bovine serum and 200 M L-glutamine. Cells were cultured in the absence of antibiotics and antimycotics and used for transfections or RNA/protein analysis when 70 -90% confluent. The same medium conditions were used for prostate stromal and epithelial cells as well as LNCaP and PC3 prostate cell lines. In some experiments the androgen, methyltrienolone (R1881, PerkinElmer Life Sciences), was used at 1 or 10 nM (see below). For L6 stable cell lines carrying myocardin (4), cells were cultured in the presence of 900 g/ml G418.
Transient Transfection and Luciferase Assay-The calcium phosphate co-precipitation method was used for transient transfections as described (25). Cells were dispersed in 24-well plates at 50,000 cells/well and grown to subconfluence. Except where indicated, cells were co-transfected with 0.5 g of ACTG2 promoter linked to firefly luciferase and either 0.5 g of empty vector (pcDNA3.1 or pCGN) or an equal molar quantity of the same vector carrying various forms of MYOCD, NKX3.1, NKX2.5, or SRF-VP16. Fifty nanograms of the internal Renilla reporter gene (Promega) were used in all samples to normalize luciferase activity. The following day (12-16 h post-transfection), cells were washed with phosphate-buffered saline and refed fresh growth medium for an additional 24 h. Cells were then lysed with 200 l of 1ϫ Passive lysis buffer (Promega) and 20 l of cell lysate was processed for Dual Luciferase Assay according to the manufacturer's manual (Promega). Bioluminescence was measured over 10 s with an AutoLumat LB 953 luminometer (EG&G Berthold, Gaithersburg, MD). All transfections were performed in quadruplicate in at least two independent experiments and expressed as the normalized (luciferase to Renilla) fold-change over baseline (Ϫ55 ACTG2, see Fig.  2B, and Ϫ55 ACTG2 absent MYOCD in see Fig. 3), percent change under Ϫ285 ACTG2 (see Fig. 5), or fold-increase over pcDNA3.1 or pcGN control vectors (see Fig. 6).
Immunohistochemistry-Sections (5 m) of rat heart and aorta were processed for immunohistochemistry with an ACTG2-selective antibody as described (26). Under these conditions (Ͼ1:500 antibody dilution) we and others have shown that the monoclonal antibody clone B4 used in this study is highly selective for the smooth muscle ␥-actin isoform, ACTG2 (48,49).
Co-immunoprecipitation and Western Blotting-Interactions between NKX3.1 and MYOCD were examined by co-immunoprecipitation following co-transfection with FLAGtagged MYOCD and HA-tagged NKX3.1. Transfected cells were lysed in immunoprecipitation lysis buffer containing 50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 5 mM MgCl 2 , 0.5% Nonidet P-40, 1 mM dithiothreitol, 10 mM sodium fluoride, 1 mM sodium orthovanadate, 200 M phenylmethylsulfonyl fluoride and protease inhibitor mixture (Sigma). Lysates were cleared by centrifugation (5 min at 14,000 rpm), followed by incubation with anti-FLAG antibodies conjugated to agarose beads (M2, Sigma) overnight at 4°C. The beads were washed three times with lysis buffer, the immune complexes were eluted by boiling in Laemmli buffer for 5 min and subjected to polyacrylamide gel electrophoresis followed by electrotransfer to Immobilon-P membrane (Millipore). The proteins on the membrane were then analyzed by Western blotting with anti-HA or anti-FLAG antibodies followed by horseradish peroxidase-conjugated secondary antibodies (Calbiochem), and developed by enhanced chemiluminescence reagent (Pierce).
In Vitro Binding Assay-The [ 35 S]methionine-labeled in vitro translated myocardin probe was generated in rabbit reticulocyte lysate (Promega, Madison, WI), following the manufacturer's protocol. The [ 35 S]MYOCD probe was incubated with purified GST-NKX3.1 or GST alone (bound to glutathione-Sepharose beads) for 3 h at 4°C in a buffer containing 20 mM HEPES, 75 mM KCl, 1 mM EDTA, 2 mM MgCl 2 , 0.1% Nonidet P-40, 2 mM dithiothreitol. The resin was subsequently washed four times with 1 ml of the binding buffer, and then boiled in Laemmli sample buffer. Proteins were resolved by SDS-PAGE and transferred to a polyvinylidene difluoride membrane. The proteins on the membrane were stained with Amido Black, and the [ 35 S]MYOCD was detected by autoradiography.
Electrophoretic Mobility Shift Assay (EMSA)-To examine SRF-MYOCD binding to CArG boxes in the human ACTG2 promoter, the following oligonucleotides (CArG and NKE sites underlined, mutant bases in lowercase) and their respective complement (not shown) were heated at 65°C for 10 min and annealed prior to end labeling with [␥-32 P]dATP as described (27) The wild type ACTG2 Ϫ205 promoter fragment containing CArG1, NKE, and CArG2 was end-labeled using T4 polynucleotide kinase (New England Biolabs). SRF and MYOCD were in vitro translated by the TNT Coupled Reticulocyte Lysate system (Promega). Endogenously expressing MYOCD was from lysates of L6 cell lines stably transfected with MYOCD (4). Samples were resolved on a 5% polyacrylamide gel using indicated competitors and antibodies essentially as described (27).
Computational Biology-Comparative sequence analysis of the human ACTG2 genetic locus with orthologs in chimp (Pan troglodytes, Ptr), dog (Canis familiaris, Cfa), mouse (Mus musculus, Mmu), and chicken (Gallus gallus, Gga) was done with the VISTA Gene Alignment tool (28). Sequence logos representing position weight matrices of orthologous CArG and NKE elements of the ACTG2 promoter from the same species as those used in the VISTA analysis were generated with WebLogo (29). Simple protocols for performing these analyses are available upon request.

ACTG2 Expression in Vascular Smooth Muscle and Prostate-
Actg2 mRNA is abundantly expressed in aorta, bladder, and cultured SMC with less expression in other tissues and cell lines (Fig. 1A). In contrast, Acta2 mRNA is more broadly expressed. Interestingly, we observe differential splicing of Srf mRNA in aorta and bladder (primarily full-length) versus other tissues and cultured SMC (full-length plus ␦5 splice variant, Fig. 1A); the molecular basis for this differential splicing phenomenon is presently unknown. As expected, Nkx2.5 mRNA is restricted to heart but detectable transcripts are also seen in the A7r5 SMC line (Fig. 1A). Expression of the related Nkx3.1 mRNA is only seen in L6 myoblasts and A7r5 SMC (the latter cells only show detectable Nkx3.1 transcripts after two sequential rounds of PCR). Actg2 and Nkx3.1 mRNA are present in prostate cell lines (39). To see whether Myocd is similarly co-expressed with these transcripts, we measured mRNA levels in several distinct human prostate cell lines in the absence or presence of androgen. Results reveal detectable transcripts to Myocd in each cell line (Fig. 1B). Moreover, Myocd is co-expressed with Actg2 and Nkx3.1 mRNA in mouse prostate tissue (Fig. 1C). To determine whether expression of the ACTG2 protein coincides with its mRNA expression in vascular SMC, we performed immunohistochemistry with an ACTG2-selective antibody on sections of rat heart and aorta. Levels of ACTG2 protein are present in coronary artery, aortic and microvascular SMC (Fig. 1D, d, e, and f, respectively), but not cardiomyocytes (Fig. 1D, d) or perivascular tissue (Fig. 1D, f). Together, the expression data establish mRNA and protein expression of SM ␥-actin (Actg2) in distinct vascular SMC types and the co-expression of Myocd, Nkx3.1, and Actg2 mRNA in prostate tissue/cell lines.
SMC-specific Basal and MYOCD-dependent Activation of Actg2-Comparative genomics between human (Hsa) and chimp (Ptr), dog (Cfa), mouse (Mmu), and chicken (Gga) reveals high sequence conservation in the immediate 5Ј promoter region of ACTG2 where 4 conserved CArG elements and a binding element for NK factors (NKE) reside over ϳ200 bp of genomic DNA (Fig. 2, A and B). The presence of CArG1 and NKE elements mediates mild basal promoter activity in several SMC lines from human and rat (Ϫ95 in Fig. 2C). Inclusion of CArG2 dramatically elevates basal human ACTG2 promoter activity in a SMC-specific manner (Ϫ205 in Fig. 2C). The addition of CArG3 and CArG4 has no further stimulatory effect on ACTG2 promoter activity in SMC (Ϫ285 in Fig. 2C). To determine the relative effects of MYOCD on each ACTG2 promoter construct, transient co-transfections were performed in COS7 and the rat pulmonary artery PAC1 SMC line using the 807amino acid form of MYOCD (3) (Fig. 3, A and B). In both COS7 and PAC1 SMC, MYOCD is unable to stimulate increases in the ACTG2 Ϫ55 promoter construct above baseline levels and only mildly stimulates Ϫ95 ACTG2. However, addition of CArG2 (Ϫ205) results in 100-fold increases in ACTG2 promoter activity in COS7 cells (Fig. 3A). Less apparent induction is seen with Ϫ205 ACTG2 in PAC1 SMC due to the higher basal activity of the Ϫ205 reporter (Fig. 3B). Similar to data in Fig. 2C, the presence of CArG3 and CArG4 (Ϫ285) confers no further MYOCDdependent transactivation of ACTG2 promoter activity (Fig. 3,  A and B). Consistent with previous data using other CArG-dependent promoters (3), the basic and C-terminal transactivation domain of MYOCD are essential in mediating ACTG2 promoter stimulation (data not shown). To determine whether the potent stimulation of MYOCD on the ACTG2 promoter translates into activation of the endogenous rat Actg2 gene, we evaluated mRNA expression levels of Actg2 in L6 myoblasts stably transfected with Myocd. In three independent stable cell lines, the presence of Myocd transcripts correlated with induction of the Actg2 mRNA (Fig. 3C). Moreover, RNA knockdown of SRF in BC 3 H1 cells elicits corresponding decreases in endogenous Actg2 transcripts (Fig. 3D). Together, these results firmly establish Actg2 as an SRF-MYOCD-dependent SMC target gene, consistent with previous expression studies (11).
SRF-Myocardin Preferentially Binds CArG2 of the Human ACTG2 Promoter-A previous study using the chicken Actg2 promoter showed differential binding of SRF to each of the 4 conserved CArG elements (26). We used EMSA to assess the association of MYOCD with SRF bound to CArG1 or CArG2 of the human ACTG2 Ϫ205 promoter because luciferase data reveal high level activity within this region (Fig. 3, A and B). A nucleoprotein ternary complex of in vitro translated SRF and MYOCD can readily be seen and validated with appropriate antibodies to SRF and FLAG-tagged MYOCD (Fig. 4, lanes  5-7). Competitive EMSA experiments with cold Ϫ205 probe or CArG2 oligonucleotide disrupt CArG-SRF-MYOCD ternary complex formation, whereas CArG1 or NKE have minimal effect (Fig. 4, lanes 8 -11). These results suggest that SRF-MYOCD associate preferentially with CArG2 of the human ACTG2 Ϫ205 promoter.
We next evaluated effects of point mutations in each of the 4 CArG elements or NKE on the ability of MYOCD to transactivate the ACTG2 promoter. Mutations in CArG1, -3, and -4, leaving CArG2 intact, mildly reduce MYOCD transactivation of the ACTG2 Ϫ285 promoter (Fig. 5A). On the other hand, mutating CArG2 in context of wild type CArG1, -3, and -4 completely abolishes MYOCD-dependent transactivation of the ACTG2 promoter suggesting that CArG2 is the preferred CArG element MYOCD utilizes for ACTG2 transcription (Fig.  5A). Surprisingly, mutation of NKE (without CArG mutations) also reduces MYOCD-dependent transactivation suggesting that NKE and one of its binding proteins may influence the ability of MYOCD to activate through CArG2 (see below). Similar results are seen in other cell types (L6 and HeLa) and with the ACTG2 Ϫ205 promoter (data not shown).
We next performed EMSA with the ACTG2 Ϫ285 promoter (with 4 CArG boxes) as a probe; chromatin immunoprecipitations cannot adequately discriminate SRF-MYOCD association between such closely juxtaposed CArG elements. We find that mutation of all 4 CArGs is incompatible with SRF-MYOCD ternary complex formation (Fig. 5B, lane 2 versus lane 1). Consistent with the luciferase activity above, the same probe with only CArG2 intact supports full binding activity (lane 3 versus lane 1), but mutation of CArG2 in the context of WT CArG1/

3/4 leads to near complete loss in SRF-MYOCD ternary complex formation (lane 4).
Competitive EMSA studies further substantiate the pivotal role of CArG2 in SRF-MYOCD binding to the ACTG2 Ϫ285 promoter (lanes 6 -11). We next determined the effects of substituting CArG2 with other CArG elements on ACTG2 promoter activity and SRF-MYOCD complex formation. Substantial loss in both promoter activity (Fig.  5C) and SRF-MYOCD ternary complex formation (Fig. 5B,  lanes 12-15) is seen when CArG2 is replaced with any of the other CArG elements of the ACTG2 promoter. Taken together, results support the concept of MYOCD preferentially utilizing a single CArG element (CArG2), among many, to effect ACTG2 gene transcription.
NXK3.1 Potentiates ACTG2 Promoter Activity and Physically Associates with MYOCD-Because mutation of the NKE site adjacent to CArG1 reduces MYOCD-dependent transactivation of the ACTG2 promoter (Fig. 5A), we hypothesized that NKX3.1 would augment MYOCD-dependent ACTG2 promoter activity. Initial co-transfection studies with NKX3.1 and MYOCD failed to show any additive effect of NKX3.1 on ACTG2 promoter activity, probably because the level of MYOCD expression was high thus masking any positive effect of NKX3.1 (data not shown). However, when we reduce the concentration of MYOCD Ͼ10-fold, a strong stimulatory effect of NKX3.1 is observed in COS7 cells (Fig. 6A). This effect requires MYOCD and is specific for NKX3.1 as the related NKX2.5 factor is without effect (Fig. 6A) despite its apparently higher expression (Fig. 7A, lower panel). The effect of NKX3.1 on MYOCD-dependent transactivation is dose-dependent (Fig. 6B) and specific for the ACTG2 promoter as no such activity is seen with other SMC promoters or the ACTG2 Ϫ95 promoter (data not shown). Moreover, similar findings are seen in HeLa and L6 cells indicating the effects are not restricted to only one cell type (data not shown). Similar to data in Fig. 5A, the synergistic stimulation for NKX3.1 of the ACTG2 promoter activity is severely attenuated when either CArG2 or the NKE site is mutated (Fig. 6C). Collectively, these results suggest MYOCD functionally associates with NKX3.1 to effect ACTG2 transcription.
To ascertain whether the functional association between NKX3.1 and MYOCD extends to a physical interaction, we performed co-immunoprecipitations in COS7 cells co-transfected with Myocd and Nkx3.1 or Nkx2.5. Both NKX factors interact with MYOCD (Fig. 7A, middle panel) and the interaction of NKX3.1 was lost when the C-terminal transactivation domain of MYOCD was deleted (Fig. 7B, middle panel). The direct interaction between MYOCD and NKX3.1 was confirmed by the in vitro binding assay of [ 35 S]MYOCD with GST-NKX3.1 (Fig. 7C). These results strongly suggest that NKX3.1 mediates its MYOCD-dependent transactivation of ACTG2 transcription via direct physical contact through the C-terminal transactivation domain of MYOCD.

DISCUSSION
The results of this report demonstrate expression of SM ␥-actin (ACTG2) in vascular SMC in vitro and in vivo, thereby validating this traditionally viewed visceral SMC marker (26), a vascular SMC marker as well. The human ACTG2 promoter, containing four conserved CArG elements, exhibits SMCspecific promoter activity in vitro and is highly responsive to MYOCD-dependent transactivation. Several lines of evidence support MYOCD preferentially utilizing a single CArG element . Blue vertical peaks indicate exons, whereas pink peaks indicate at least 75% sequence homology over 100 bp of non-coding sequence. Note the progressive loss in non-coding sequence homology from chimp to chicken. B, sequence logos for the 5 major cis-acting elements in the immediate 5Ј promoter region of ACTG2. These logos represent position weight matrices derived from human, chimp, dog, mouse, and chicken orthologous sequence elements (position of human elements relative to transcription start site indicated at the left of each logo). Note that CArG4 and CArG2 are 100% homologous across all species analyzed. Where sequences diverge in other elements, the height of the letters diminishes with inclusion of variant base sequences. C, firefly luciferase activities of various human ACTG2 promoter constructs in 3 distinct SMC lines (PAC1 pulmonary artery, human SKLMS uterine SMC, and A7r5 aortic SMC) versus two non-SMC types (COS7 and L6). Luciferase activity is presented here as the fold-activation of normalized (ratio of luciferase to Renilla control) luciferase to the minimal Ϫ55 promoter, set to 1. Each bar represents the average (and standard deviation) of 4 replicates in each of the indicated cells lines. These experiments were performed multiple times with similar trends in relative fold-activation. Note the y axis here and in Fig. 3 is based on a logarithmic scale.
for transactivation of ACTG2. First, we show through mutagenesis studies that MYOCD mediates ACTG2 transcription primarily through CArG2. Second, the CArG2 sequence is shown to confer the highest level of SRF-MYOCD binding. Finally, replacing CArG2 with another consensus CArG (CArG1) only weakly activates the ACTG2 promoter demonstrating that the sequence character of CArG2 is critical for MYOCD-dependent transactivation. Mutation of a binding site for NKX3.1 (24), adjacent to CArG2, reduces MYOCD-dependent transactivation of ACTG2, suggesting a functional association between NKX3.1 and MYOCD. Indeed, we show that NKX3.1 potentiates the activity of MYOCD in a CArG2-and NKE-dependent manner, and binding assays reveal direct association of NKX3.1 to the C-terminal domain of MYOCD. Taken in aggregate, the results suggest a novel mechanism for the ability of MYOCD to discriminate among multiple CArG boxes through the physical and functional association of MYOCD with the NKX3.1 homeodomain protein.
In mammals, there are six major actin isoforms encoded by separate genes that are transcribed as six independent transcripts encoding proteins with Ͼ90% amino acid identity (30). All six actin genes contain at least one SRF-binding CArG element that is important for gene transcription (8), yet each gene displays unique expression patterns during development and in postnatal tissues (21,31). The manner in which the ubiquitous SRF transcription factor mediates such disparate patterns of gene expression relates to its ability to recruit and bind contextor cell-restricted transcription factors, some of which associate with neighboring cis elements. For example, the cardiac ␣-actin (Actc1) promoter has CArG elements that bind SRF and NKX2.5, both of which counter the repressive effects of another CArG-binding factor, YY1 (32). SRF and NKX2.5 physically   associate and the latter is proposed to direct expression of Actc1 independent of its ability to bind DNA (33,34). Thus, vertebrate hearts express the cardiac isoform of actin largely through the coordinate activities of SRF and heart-restricted factors such as NKX2.5. The notable lack of NKX3.1 expression in heart (35)(36)(37) could explain, in part, the weak expression of Actg2 in the myocardium (36) (and Fig. 1) despite the presence of MYOCD. These examples highlight a broader complex regulation of SRF-dependent muscle-restricted genes through the coordinate activities of over 60 SRF cofactors (8).
Previously we showed that a proximal CArG box of the chicken Actg2 promoter binds SRF avidly through an interaction with NKX3.1 that binds an NKE adjacent to the CArG box (24). In this report, we further advance this concept by showing a functional and physical interaction between NKX3.1 and MYOCD. The functional association of NKX3.1 with MYOCD appears to be unique because a related homeodomain protein (NKX2.5) binds MYOCD but shows no activity over the ACTG2 promoter. It is possible that interactions between MYCOD and NKX2.5 potentiate cardiac-restricted genes.
Currently available assays could not reveal the presence of a complex between SRF, MYOCD, and NKX3.1 (not shown); however, we postulate that a multiprotein complex encompassing these and perhaps other factors over CArG2 and NKE exists to optimally transactivate the ACTG2 promoter in a context-dependent fashion. One such context may be the prostate epithelium where levels of Nkx3.1 (35,(37)(38)(39), Actg2 (40), and possibly Myocd (this report) are expressed. Loss in NKX3.1 results in defects of the prostate, including hyperplasia of the ductal epithelium with no obvious defects in vascular tissues (38,39). Whether levels of ACTG2 are reduced in the prostate or vascular tissues from Nkx3.1 null mice is currently unknown. Interestingly, ACTG2 and NKX3.1 are considered tumor suppressors with the former part of a molecular signature for human metastatic disease (41). MYOCD is also known to reduce growth (4) and itself has been implicated as a tumor suppressor (42). Thus, it may be important to maintain normal levels of NKX3.1, ACTG2, and MYOCD for normal prostate function.
Although there are four closely spaced CArG boxes within the ACTG2 promoter, the two proximal sequences, CArG1 and CArG2, are the most avid players with regard to MYOCD-mediated transcriptional responses. This would be consistent with the concept advanced by Olson and colleagues (11) that proposes MYOCD acting as a molecular bridge between adjacent CArG boxes in CArG-rich promoters to mediate strong, cellspecific transcription (43). However, there are exceptions to the FIGURE 6. NKX3.1 potentiates MYOCD-dependent activation of the ACTG2 promoter. A, COS7 cells were co-transfected with the ACTG2 Ϫ285 promoter plus the indicated plasmids, and luciferase activity (per "Experimental Procedures") was measured 24 h later. The amount of MYOCD input was 50 ng, whereas NK factors were used at 500 ng. Results are displayed as the mean Ϯ S.D. and are representative of two independent studies. B, similar study as in panel A only the amount of NKX3.1 input was varied and balanced by the empty pcGN vector control. C, the wild type ACTG2 Ϫ285 promoter or the indicated mutants were co-transfected with 50 ng of MYOCD and 500 ng of NKX3.1 and luciferase activity measured (per "Experimental Procedures"). Results are displayed as those in panels A and B. multi-CArG, MYOCD responsive target gene concept including Egr1 and Akap12␣ promoters, which contain multiple CArG elements but are nevertheless refractory to MYOCD-dependent transactivation (12,14) (data not shown). Additionally, there are single CArG-containing gene promoters, such as telokin and Cald1, that are strongly activated by MYOCD (16,44). Although the bulk of evidence supports the concept of MYOCD-dependent gene activation requiring multiple CArG sequences, the growing exceptions to this idea may indicate a role for accessory proteins to potentiate or modify cell-specific MYOCD-mediated responses. In the activation of Egr1 transcription, there appears to be a competitive inhibition of MYOCD response by association of ELK1 with SRF (13). As indicated in our studies, MYOCD-mediated transcription of ACTG2 is potentiated by its interaction with the homeodomain, NKX3.1. The functional consequences of this interaction are dependent upon DNA context as shown by the inability of NKX2.5-MYOCD complexes to activate ACTG2 promoters. Thus our results support the concept that MYOCD-mediated transcriptional responses require CArG context within the promoter/activator segment as well as the appropriate association of multiple proteins within the transcriptional complex.
Two of the four CArG elements in the ACTG2 promoter (CArG3 and CArG4) are non-consensus CArG boxes that bind SRF weakly (26) and confer no further MYOCD-dependent transactivation. Therefore, the ACTG2 promoter presents a model to gain an understanding of MYOCD-modulated, cellspecific transcriptional activation in promoters containing active and inactive CArG elements. Similar to ACTG2, the TAGLN promoter contains two similarly spaced CArG boxes within its 5Ј proximal promoter. Although this gene is highly MYOCD inducible, studies in transgenic mice have shown that a single CArG box is responsible for cell-specific promoter activity (45,46), indicating that multiple CArGs may participate in gene activation through several mechanisms. Thus, it is possible that the ACTG2 non-consensus CArG boxes play an important context-dependent role in appropriate MYOCD/ SRF transcriptional responses. Of note is our previous studies demonstrating a time-dependent binding of SRF to CArG3 and CArG4 in SMC with elevated SRF expression (26). This would indicate that promoter context plays a significant role in appropriate cell-specific transcription. That is, these distal elements may potentiate ACTG2 transcription that was initiated by MYOCD-dependent activation at the proximal CArG elements resulting in appropriate SMC ACTG2 expression. Alternatively, there may be little functionality of these sequences and their conservation may simply reflect their juxtaposition to a block of sequence resistant to genetic drift. It will be important to address these questions in transgenic mice where each of the ACTG2 CArG elements is mutated.
In summary, we provide new insight into the transcriptional regulation of a CArG-rich human SMC-restricted promoter, ACTG2. The finding that SRF and MYOCD utilize a distinct CArG element, even among a cluster of similar elements over a short genomic landscape, supports a model wherein the sequence content of the CArG box, per se, imparts structural information that dictates how SRF-associated factors such as MYOCD interact. Molecular structure studies of different SRF-CArG nucleoproteins in the absence or presence of MYOCD will likely offer novel insight into how the CArG sequence character confers MYOCD-dependent transactivation.