CARP, a cardiac ankyrin repeat protein, is downstream in the Nkx2-5 homeobox gene pathway

Cardiac muscle progenitor cells originate from the lateral splanchnic mesoderm in vertebrates, migrate anterolaterally, fuse to form a single cardiac tube at the ventral midline and then undergo an elaborate series of steps of differentiation and morphogenesis to form a functional heart (Garcia-Martinez and Schoenwolf, 1993; Rosenquist and DeHaan, 1996; Sater and Jacobson, 1990; Wilens, 1995). Despite the extensive knowledge at the embryological level, the molecular mechanisms of the complex process of cardiogenesis is relatively unclear. In this regard, the specification of distinct atrial and ventricular chambers is one of the critical steps during vertebrate cardiogenesis. To identify molecular pathways that guide this process, we have utilized the ventricular specific myosin light chain-2 (MLC-2v) gene as a model system, which is the earliest ventricular chamber-specific gene expressed during mammalian cardiogenesis (O’Brien et al., 1993). Using a combination of transient assays and transgenic approaches, we have recently identified a 28-bp minimal cis-element (HF1), which contains two adjacent (HF-1a and MEF-2) sites, which confer cardiac and ventricular chamber-specific expression in an anterior-posterior gradient (Zhu et al., 1991; Navankasattusas et al., 1992, 1994; Ross et al., 1996). The ubiquitous transcription factor YB-1 has been shown to bind to the HF-1a site and can regulate MLC-2v promoter activity in co-transfection assays (Zou and Chien, 1995). Biochemical studies have demonstrated that YB-1 has a cofactor and it is this complex that occupies the HF-1a site (Zou and Chien, 1995). As such, the identification of YB-1-associated proteins becomes critical in elucidating the pathways of ventricular specification.

In the current study, we report the isolation and initial characterization of a YB-1 associated protein that is cardiac restricted early during murine cardiogenesis and which contains ankyrin repeats. CARP is a nuclear factor and has negative transcriptional activity. In situ analysis reveals that carp is expressed abundantly and specifically in the early embryonic heart, supporting an important role during cardiogenesis. Using both in situ and RNase protection analyses, carp expression was found to be dramatically and selectively reduced in embryos that carry a targeted interruption of the homeobox gene Nkx2-5, a tinman homologue required for ventricular chamber specification and morphogenesis. Co-transfection assays using an Nkx2-5 dominant-negative cDNA construct suggest that Nkx2-5 can regulate the carp gene activity at the transcriptional level. A carp promoter-lacZ transgene, which recapitulates the cardiac-specific expression of the endogenous gene, is significantly reduced in the Nkx2-5^−/− background, providing additional evidence that Nkx2-5 either directly or indirectly regulates the carp promoter during in vivo cardiogenesis. Thus, carp is downstream in the Nkx25 homeobox gene regulatory network and may act as a negative regulator of HF-1-dependent pathways for ventricular muscle gene expression. As such, these studies suggest that Nkx2-5 may simultaneously activate positive and negative regulators to titrate the level of expression of specific subsets of cardiac muscle genes.

To isolate YB-1 associated proteins, we utilized the two-hybrid interaction trap system developed by the laboratory of Dr Roger Brent (Gyuris et al., 1993). Briefly, poly(A) RNA was isolated from a total RNA preparation from rat neonatal ventricular chamber and the first strand cDNA was synthesized using the Moloney virus reverse transcriptase with an oligo(dT) primer including an XhoI site. After second strand cDNA synthesizes, the 5′ ends were ligated to an EcoRI adapter and the cDNA inserts were ligated into the vector pJG4-5, yielding 1.3 million independent clones. 92% of the cDNA clones from the library contained inserts with an average of 1.0 kb and ranged from 500 base pair to several kbs. 1.3 million clones were screened using a YB-1-LexA fusion protein as the bait construct, following previously described procedures (Gyuris et al., 1993). 400 positive clones were obtained during the leucine minus selection and 26 positive clones displayed blue color during the lacZ selection. Four of these positive clones were identical and encoded a c-terminal region of carp. To assess tissue distribution of CARP expression, northern blot analysis was performed (Clontech) and exposed for 4 days with double intensifying screens and at − 80°C. The full-length rat carp cDNA was cloned by screening a rat neonatal heart λ gt11 cDNA library with the c-terminal region of carp as a probe.

The carp cDNA was subcloned into the pGEX vector (Pharmacia) and expressed in the BL21 strain. After IPTG induction, the bacterial cells were harvested and lysed by several freeze-thaw cycles, followed by sonication. The GST-CARP fusion protein was purified from the bacterial lysate by glutathione beads and injected into rabbits to generate polyclonal antiserum. Bleeds were tested at various dilution by western blotting. Immunoprecipitation was performed as previously described (Zou and Chien, 1995). 10 μg of either the GST protein alone or the GST-CARP protein were added to 1 ml of the cardiac myocyte whole-cell extracts and incubated at 4°C for 2 hours with gentle shaking. Glutathione-sepharose beads (Pharmacia) were added to the tube and incubated for 1 hour. The beads were washed three times with lysis buffer and the GST or GST-CARP fusion proteins were eluted with 10 mM glutathione under native conditions, followed by SDS-PAGE and western blotting.

The nucleotide sequence of the mouse and rat carp cDNA was 97% identical, including the 5′ and 3′ noncoding region. The 1700 base pair rat carp fragment was subcloned into pRC/CMV via a NotI site in the sense and antisense orientations downstream of the T7 promoter. The template constructs were linearized with XbaI. Digoxigenin labeling was performed according to Boehringer Mannheim. Alkaline hydrolysis was carried out to reduce the probe size to an average of 250 base pairs. Analogous studies were performed with the MLC-2a probes, prepared as previously described (Kubalak et al., 1994). A fragment of rat YB-1 cDNA (500 bp at the C-terminal cDNA coding region) was amplified by PCR subcloned into pRc/CMV via NotI and sequence verified. The rat and mouse YB-1 cDNA display 99% sequence identity.

To generate the ³⁵S-labeled riboprobe, in vitro transcription assays were carried out using ³⁵S-α UTP, instead of the Dig-UTP, with 10 mM DTT. cRNA probes were alkaline hydrolysed in a 20 mM NaHCO₃/Na₂CO₃ (2:3) buffer with 10 mM DTT for 60 minutes at 60°C. Free nucleotides were removed by running through the neutralized hydrolysis reaction through a Sephadex G-50 spin column.

Timed pregnant mice were purchased from Harlan Sprague Dawley, Inc. Mouse embryos were isolated under a dissecting microscope and submerged in 1× PBS. The whole-mount in situ hybridization closely followed previously published methods (Wilkinson, 1992). For in situs with the embryonic sections, the timed mouse embryo paraffin-embedded sections were purchased from Novagen. For the ³⁵S-labeled cRNA probe, we followed the protocol kindly provided by Dr Gary Lyons (Lyons et al., 1990, 1991a,b). For in situs with the Dig-labeled probe, we followed previously published methods (Becker et al., 1992; O’Donnell et al., 1989).

A 260 bp cDNA fragment of the rat carp was used for a RNase protection probe. In this region, the rat and mouse carp have 99% sequence identity. The E9.0 heart tube was dissected out and subjected to RNase protection assays using ‘Direct Protect’ kit from Ambion. After lysis of heart tube samples, they were divided equally to two tubes. One was hybridized to a carp probe and an EF1α probe, which serves as internal control; the other was hybridized to MLC-2a probe and EF1α probe as internal control. The MLC-2a and EF1α probes were prepared according to previously published information (Kubalak et al, 1994).

Cardiac myocytes were prepared by a percoll gradient technique and plated in gelatin-coated 6 cm plates at 10⁶ cell/plate in plating medium (Sheng et al., 1996). After overnight incubation, cardiac myocytes attached to the plates and the medium was switched to 4% horse serum for 3-5 hours prior to transfection. Co-transfection in cardiac myocytes was performed by calcium-phosphate precipitation, as initially described (Chen and Okayama, 1987) and subsequently modified (Zou and Chien, 1995). The luciferase activity was normalized to the β -gal activity to account for variations in the efficiency of transfection, as previously described (Zou and Chien, 1995). For the GAL4-CARP fusion construct, the reporter was a GAL4 DNA-binding site and CAT reporter (Clontech). CAT assays were carried out according to the method described by Promega, using the liquid scintillation counting. The rat Nkx2-5 cDNA was kindly provided by Dr YuChang Fu. Its homeodomain was fused C-terminal to a nuclear localization signal and the inhibiting domain of drosophila transcriptional repressor engrailed (Han and Manley, 1993; Jaynes and O’Farrell, 1991; John et al., 1995; Badiani et al., 1994; and Conlon et al., 1996). This construct has been shown to work very efficiently in co-transfection assays with a reporter construct harboring the Nkx2-5 responsive site TTF-1 (Evans et al., unpublished results).

Primary cardiac myocytes were isolated from neonatal rat heart ventricular chambers and plated on the chamber slides precoated with 1% gelatin and laminin (Sigma). After culturing for 24 hours, the cells were rinsed in 1× PBS for 15 minutes, fixed in 4% paraformaldehyde for 10 minutes and neutralized in 50 mM NH₄Cl (in PBS) for 5 minutes each. Cells were then washed twice with PBS, permeabilized and blocked in PBS, 1% BSA, 0.3% Triton X-100 for 15 minutes followed by a 60 minutes incubation with the primary antibody in PBS, 1% BSA, 2% newborn goat serum, 0.1% Tween-20 with 1:100 dilution of the antibody. The primary antibody was then washed away with PBS for 5 minutes. The secondary antibody was incubated in PBS, 1% BSA, 2% newborn goat serum, 0.1% Tween-20 for 60 minutes and washed away 4× with PBS for 5 minutes each. The slides were briefly washed for 5 minutes, dried and mounted with Gelvatol plus 2.5% DABCO. The Gelvatol was 2.4 g Gelvatol 20-30 (Monsanto Chemicals) to 6 g of glycerol, stirred and mixed, and 6 ml of water was added. After several hours at room temperature, 12 ml of 0.2 M Tris (pH8.5) was added and heated to 50°C for 10 minutes with occasional mixing. After the Gelvatol was dissolved, the solution was clarified by centrifugation at 5000 g for 15 minutes. DABCO was 1,4diazobicyclo[2.2.2]-octane. The slides were subsequently subjected to fluorescent microscopy.

A mouse carp genomic DNA clone was isolated using the rat carp cDNA as the probe and various lengths of the promoter were tested for cardiac-specific activity (Chen, Kuo, Zou and Chien, unpublished data). A 2.5 kb fragment can confer cardiac-specific expression in cotransfection assays. Transgenic lines were established using a 2.5 kb carp promoter fragment fused to a lacZ cDNA, which resulted in cardiac-specific expression in early cardiac development (Chen, Kuo, Zou and Chien, unpublished data). The 2.5 kb transgene was crossed into Nkx2-5 knock-out background. Embryos at day 9.5 were analyzed for β -gal staining as described by Ross et al. (1996).

Using YB-1 as the bait in the yeast two-hybrid system, 1.3 million clones from the neonatal rat heart ventricular cDNA library were screened, resulting in the isolation of a partial cDNA encoding a cardiac ankyrin repeat protein, which we designated as CARP. As displayed in Fig. 1A, the deduced amino acid sequence of the CARP cDNA revealed the presence of four conserved ankyrin-like repeats with an additional half less conserved ankyrin repeat c-terminal to the four repeats. These ankyrin repeats were most similar to those in the transcription factor GA-binding protein (LaMarco et al., 1991). CARP belongs to a distinct class of ankyrin repeat proteins, since it does not contain the ETS domain that is characteristic of the GA-binding proteins (LaMarco et al., 1991). During the course of this work, isolation of a human ankyrin repeat protein (c139) from cytokine stimulated human endothelial cells by a subtraction approach was reported (Chu et al., 1995), which displays a high similarity to CARP at the amino acid sequence level (Fig. 1B). In addition, the deduced amino acid sequence predicts several putative phosphorylation sites, a PEST-like sequence (Chu et al., 1995), a nuclear localization signal and an amino terminus that contains a high content of positively and negatively charged amino acids. Comparison with a Xenopus homologue (XCARP) isolated from a Xenopus adult heart library cDNA by low-stringency screening, indicated a 67% overall amino acid identity and 82% identity in the ankyrin repeats (Zou, Fu, Evans and Chien, unpublished results). Analysis of the tissue distribution of carp mRNA by northern blotting revealed that it is highly enriched in the adult heart (Fig. 1C), with a trace amount detected in skeletal muscle. As a control, the same northern blot was hybridized with the human β -actin probe, which detects an equal amount of signal in all tissues assayed (Fig. 1D). The expression of carp in murine lung tissue reflects the presence of atrial tissue, which invaginates the pulmonary vein in the adult mouse (Jones et al., 1994). Further evidence for the lack of expression in lung tissue has been revealed by analysis of carp knockout mice which harbor a knock-in of lacZ into the endogenous carp locus (Chen and Chien, unpublished results).

To independently confirm that CARP is a YB-1-binding protein in native cardiac myocytes, we employed biochemical analyses using immunoprecipitation. As assessed by coimmunoprecipitation and GST-CARP pulldown assays, YB-1 and CARP formed a stable complex in rat cardiac myocytes, suggesting that CARP is an endogenous YB-1-binding protein. Using an anti-YB-1 antibody (Fig. 2, lane 2), we detected a 40× 10³M_r band (the size of CARP by in vitro translation, data not shown), in addition to the YB-1 band, in the co-immunoprecipitation complexes from [³⁵S]methionine-labeled cardiac myocyte extracts. In the preimmune control, the 40× 10³M_r band was not detected, documenting the specificity of the antibody (lane 1). Using the anti-CARP antibody, CARP protein was detected in the co-immunoprecipitation complex by the anti-YB-1 antibody (lane 5). As shown in Fig. 2 (lane 7), a band corresponding to the expected size of YB-1 was observed, in addition to the CARP band, in the co-immunoprecipitation complex from the [³⁵S]methionine-labeled cardiac myocyte extracts, using the anti-CARP antibody. A strong band with a relative molecular mass of 65-68× 10³M_r was observed in the co-immunoprecipitation complex (lane 7). The identification of this protein is currently in progress. Since the YB-1 protein has a similar migration rate to the IgG, the purified GST-CARP fusion protein was added to the cardiac myocyte extracts to pull down interacting proteins. The antiYB-1 antibody detected the YB-1 band in the interacting complex by western blotting (lane 10). It should be noted that the conditions to assess YB-1-CARP protein-protein interactions were performed at near physiological conditions to represent the endogenous cell context. Thus, these data indicate that CARP forms a stable physical complex with the YB-1 protein within cardiac myocytes. Yeast two hybrid assays revealed that the ankyrin repeat domain of CARP was sufficient to mediate the interaction with YB-1, thereby providing further support for the specificity of the YB-1/CARP interaction (data not shown).

As noted above, the deduced amino acid sequence of CARP contained a consensus nuclear localization sequence. To determine the subcellular localization of the CARP protein, we performed immunolocalization studies in cardiac myocytes. The anti-CARP antibody revealed that the endogenous CARP protein was localized in the nuclei of cardiac myocytes (Fig. 3A). A nuclear signal was not detected in a noncardiac cell line COS1 (Fig. 3D) and after staining of cardiac myocytes with preimmune serum control (data not shown). In addition, a strong signal enriched in the nucleus was not detected in cardiac myocytes in serum-free medium (Fig. 3B). The CARP protein levels in cells cultured in serum-free medium or serum-containing medium did not show any significant difference when equal amount of whole-cell extracts were analyzed by western analysis (Zou and Chien, unpublished data). In contrast, the YB-1 protein was always found in the nucleus in serum, and in serum-free and phenylephrine-containing media (Fig. 3E,G,H). Fig. 3F displays a typical control, using myomesin antibody to document that these cells were indeed cardiac myocytes. These data suggest that the CARP protein might be actively transported into the nucleus and that this transportation may be promoted by some components in serum. In serum-free medium, CARP could passively diffuse throughout the cell including the nucleus. Currently, we are identifying the extracellular signals that induce the nuclear localization of CARP. Stimulation of myocyte growth/hyper trophy with the adrenergic agonist phenylephrine did not induce nuclear localization (Fig. 3C) (Zou and Chien, unpublished results).

In situ hybridizations were carried out to determine the expression pattern of carp during murine cardiogenesis. carp expression was detected at least as early as at E8.5, with [³⁵S]UTP-labeled riboprobe (see Fig. 4A,B). Expression of carp is uniform throughout all cardiac muscle compartments, as assessed by whole-mount in situ hybridizations with mouse embryos of E9.0 (Fig. 4C). At E9.0, carp was specifically and evenly expressed at a high level in the looping heart tube throughout the common atrial chamber, the common ventricular chamber and the outflow tract. A sense control probe yielded no detectable signal in the embryo (data not shown). To verify this uniform distribution of CARP expression, in situ hybridization with sections of timed E9.0 embryos were performed, yielding identical results as the whole-mount in situ analyses (Fig. 4D). E10.0-10.5 embryos show that carp expression started to decrease in the ventricle and remained strong in the outflow tract and atrium (Fig. 4E). The endocardium and cushion tissues do not express carp at E9.0 (not shown) and E10.0-10.5 (Fig. 4H) hearts. The expression of carp was reduced and detectable at relatively low levels in the E11.0 heart by in situ hybridization of a number of heart sections, but the carp signal became more readily detectable at E12.0 and E13.0 (not shown). In summary, carp expression starts at least as early as E8.5, specifically in the myocardium, and its expression is temporally and spatially regulated during cardiogenesis. In neonates and adults, carp was expressed at a high level in cardiac muscle. No skeletal muscle expression was detected at any of these embryonic stages within the sensitivity of in situ hybridization.

To test whether YB-1 was also expressed in the heart at early stages, we performed in situ hybridization with a YB-1 probe. Fig. 4F shows that YB-1 was highly expressed in the myocardium of the cardiac tube at E8.0 with a much lower level in pericardium and the neural tube (except the floorplate). At E10.0, YB-1 was also strongly expressed in the trabeculated myocardium. In fact, heart is one of the tissues that displays the highest level of YB-1 expression at E10 (Fig. 4H). YB-1 was also expressed highly in the E9.0 heart and, interestingly, shows a strong signal in the dermamyotome and a much weaker signal in the sclerotome suggesting that YB-1 might be important to the skeletal muscle development as well (Fig. 4G). Thus, YB-1 is expressed at a high level in embryonic heart, thereby documenting co-localization of CARP and YB-1 during cardiogenesis.

In Drosophila, cardiac mesoderm forms a simple tube containing myocytes within the dorsal vessel (Bodmer, 1995). The homeobox gene tinman is expressed in the dorsal vessel, and a tinman null mutation results in the loss of the dorsal vessel and its progenitors (Bodmer, 1993). Vertebrate homologues of tinman have been identified in mouse (Lints et al., 1993; Komuro and Izumo, 1993) Xenopus (Tonissen et al., 1993; Evans et al., 1995) zebrafish (Lee et al., 1996) and chick (Schultheiss et al., 1995), all of which display highly restricted expression in early cardiac mesoderm. A knockout of the mouse homologue Nkx2-5 reveals a specific loss of the ven-tricular myosin light chain-2 gene expression and results in an arrest of cardiac tube looping morphogenesis, with embryonic lethality around E9.5-E10.0 (Lyons et al., 1995). In contrast to MLC2v, expression of several other cardiac muscle genes was found to be unaffected: α myosin heavy chain, β myosin heavy chain, myosin light chain-2a, cardiac actin, myosin light chain 1a, myosin light chain-1v and troponin I, documenting the specific relationship with MLC-2v (Lyons et al., 1995). Since CARP interacts with a component of the HF-1a/MEF-2 pathway of ventricular specificity and displays an early cardiac-restricted pattern of expression during early stages of cardiogenesis, it became of interest to determine if CARP expression would require Nkx2-5. As displayed in Fig. 5A, carp expression was dramatically reduced in the Nkx2-5^{− /−} embryos, as indicated by whole-mount in situ hybridization. As a control, whole-mount in situ hybridization with the riboprobe of myosin light chain-2a was performed in parallel. As previously reported, (Lyons et al., 1995), the level of MLC 2a expression was unchanged (Fig. 5B), indicating that the reduction of carp signal is not simply due to a generalized loss of cardiac tissue in the Nkx2-5^{− /−} background. RNase protection assays using a carp probe confirmed the specific decrease of expression using RNA derived from either the wild-type or the Nkx2-5^{− /−} heart tubes at E9.0 (Fig. 6). It is noted that carp expression in the anterior portion of the Nkx2-5^{− /−} heart tube, probably corresponding to the conotruncal portion of the outflow tract, displayed the lowest level of carp signal (below the level of detection) compared to the more posterior portions of the heart tube (Fig. 5C,D), which also display dramatically reduced carp expression compared to the wild-type heart tube. This suggests that carp expression at the rostral end of heart tube was particularly dependent on Nkx2-5. This alteration in the uniform expression of carp at E9.0 supported the notion that the loss of carp expression at the anterior region of the heart tube might contribute to the defect of cardiac looping morphogenesis.

To confirm that carp is downstream of Nkx2-5 and to test whether Nkx2-5 can regulate the carp gene at the transcriptional level, we utilized a CARP promoter construct in cotransfection assays employing a dominant negative-Nkx25 mutant construct. A 10 kb fragment of carp genomic DNA contained sequences which can confer cardiac-specific expression to a luciferase reporter in cultured cells (Chen, Kuo, Zou and Chien, unpublished results). Deletion studies also demonstrate that a 2.5 kb fragment still retains cardiac specificity in cultured cells, although the promoter activity decreases with larger deletions (Chen, Kuo, Zou and Chien, unpublished results). A fusion protein of Nkx2-5 with the inhibitory domain of Engrailed can inhibit the 10 kb and the 2.5 kb promoter activities over 80-85% in cardiac myocytes (Fig. 7A,B). The dominant negative Nkx2-5 mutant can functionally block the endogenous Nkx2-5 thus mimicking the loss of Nkx2-5 function in cardiac myocytes. The engrailed inhibitory domain alone displayed moderate inhibitory activity on the carp promoter constructs, although to a significantly less extent than the specific inhibition caused by the Nkx2-5-Engrailed fusion protein. Both the EngrailedNkx2-5 and the Engrailed only constructs contained the nuclear localization signal. All transfections were normalized by co-transfecting an unrelated SV40-LacZ control reporter. Taken together, these studies support the concept that carp is downstream of Nkx2-5 and that the regulation is at the transcriptional level. Defining the precise Nkx2-5-responsive element in the carp promoter and defining the mechanism by which Nkx2-5 regulates the CARP promoter is currently in progress.

To independently test whether Nkx2-5 regulates the carp promoter at the transcription level during in vivo cardiogenesis, we employed a transgenic approach by generating transgenic lines with carp promoter-lacZ reporter genes. Transgenic mice lines harboring the 2.5 kb CARP promoter driving lacZ displayed cardiac-specific β -gal staining of the reporter, thereby mimicking the pattern of endogenous carp expression (Chen, Kuo, Zou and Chien, unpublished results). We subsequently crossed these transgenic lines into the Nkx2-5^{− /−} background. As displayed in Fig. 8A, the carp promoter transgenes recapitulated the specific reduction of CARP message in the Nkx2-5^{− /−} background, thereby providing further direct support for the conclusion that carp is downstream of Nkx2-5 in the heart regulatory network, as shown in the in vitro co-transfection assays. Moreover, while a residual amount of carp transgene expression can be detected in Nkx2-5 mutant embryos, the conotruncus region of the mutant heart was not expressing the transgene (Fig. 8B,C), which again recapitulates the specific loss of carp mRNA in the conotruncus region as assessed by whole-mount in situ (Fig.5C,D). The β-gal-staining was clearly negative in the endocardium, either in wild-type or mutant embryos, which also agrees with the results that were obtained by in situ hybridization studies (Fig. 4). Fig. 8D-F displays the sectioning of the β-gal-stained embryos.

To test whether CARP has a positive or negative role in the regulation of HF-1-dependent pathways of transcriptional activation, we employed co-transfection assays with a CARP expression vector and a luciferase reporter gene driven by an HF-1 element fused to a minimal TK promoter, identical to the HF-1 element utilized in our previous studies that documented ventricular chamber-specific expression in transgenic embryos (Ross et al, 1996). As shown in a titration experiment (Fig. 9A), increasing the amounts of CARP expression vector resulted in a reduction of the HF-1-TK promoter activity in cardiac myocytes, as assessed by the luciferase reporter levels. In these studies, the activity has been normalized to β -galactosidase activity driven by an SV40 promoter as internal control. In a similar titration assay employing a CARP expression vector, CARP did not affect the minimal TK promoter itself, supporting the notion that the negative regulation by CARP was dependent on the HF-1 element (Fig. 9B).

To assess whether CARP contains an inhibitory domain, a construct encoding a GAL4 DNA-binding domain GAL4(DBD)-CARP fusion protein was expressed in both cardiac myocytes and noncardiac cells, COS1 with a Gal4 DNA-binding site-CAT reporter construct. In titration experiments (Fig. 9C,D), in both cell types, GAL4DBD-CARP fusion protein inhibited the Gal 4 reporter activity down to 2025% of the initial reporter activity, suggesting that CARP contains a transcriptional inhibitory domain. These results are consistent with the negative function manifested by co-transfection assays with the HF-1 reporter noted previously. Taken together, these results suggest that CARP may serve as a negative regulator of HF-1-dependent pathways for ventricular muscle gene expression.

The formation of distinct atrial and ventricular chambers is a critical step during cardiogenesis. Currently, relatively little is known regarding the molecular and positional cues that dictate individual steps of cardiac chamber specification, maturation and morphogenesis. Studies employing ventricular muscle cell-fibroblast heterokaryons suggest that a combinatorial pathway may exist for activation of ventricular muscle genes (Evans et al., 1994). To address this question, we have utilized a variety of transgenic and gene targeting approaches in mouse model systems to identify molecular pathways that control these critical steps during cardiogenesis. To define the molecular cues that guide early stages of ventricular chamber specification, our laboratory has employed the myosin light chain 2 ventricular gene (MLC-2v) as a model system (Chien et al., 1993). MLC-2v is the earliest known ventricular-specific marker in vertebrate cardiogenesis, displaying bilateral symmetrical expression in the restricted zone of the cardiogenic crescent prior to fusion of the progenitors in the midline at day E7.5-8.0 and restricted expression to the ventricular segment of the linear heart tube (O’Brien et al., 1993; Ross et al., 1996). Previous studies have established that a 250 bp MLC-2v promoter fragment can confer ventricular specificity in transgenic mice (Lee et al., 1992, 1994). Recently, we have provided evidence that a 28 bp element, termed HF-1, which is composed of adjacent HF-1a and MEF-2 sites (Zhu et al., 1991; Navankasattusas et al., 1992), is sufficient to confer ventricular-specific expression of a lacZ reporter gene during early stages of murine cardiogenesis (Ross et al., 1996). In addition, this combinatorial element confers an anterior/posterior gradient of transgene expression during ventricular chamber morphogenesis, providing a model system in which to investigate the cues that dictate anterior/posterior (right ventricle/left ventricle) gradients during mammalian heart development.

The ubiquitous transcription factor YB-1 binds to the HF-1a site in the MLC-2v promoter in conjunction with a co-factor and positively regulates the MLC-2v promoter (Zou et al., 1995). In a search for interacting co-factors, we now report the isolation and characterization of a nuclear ankyrin-like repeat protein (CARP) (cardiac ankyrin repeat protein) isolated from a rat neonatal heart cDNA library via two-hybrid screening, using YB-1 as bait. Northern analysis revealed that CARP mRNA is highly enriched in the adult heart with only trace levels in skeletal muscle.

A number of independent approaches, including coimmunoprecipitation and GST CARP pulldown studies, have revealed that CARP forms a physical complex with YB-1 in cardiac myocytes. To our knowledge, CARP is the first described tissue-specific co-factor for YB-1 and raises the question as to the potential role of this tissue-restricted cofactor in the regulation of YB-1 function. In situ analyses in the current study demonstrate that, although YB-1 is widely expressed, it is preferentially expressed in cardiac muscle during early heart development and comprises one of the major tissues that expresses YB-1 at high levels. In addition, YB-1 also appears to be expressed preferentially in the dermamyotome compartment versus the sclerotome compartment in somites during somitogenesis, implicating a role for YB-1 in the control of the skeletal muscle gene program. The interaction of YB-1 with tissue-specific partners, such as CARP, might identify YB-1 as playing a more specialized role in the control of tissue-specific gene programs, which has been suggested by transient transfection assays in a wide variety of other cell types (Grant and Deeley, 1993; Ruiz-Lozano et al., 1994; Wolffe et al., 1992).

Although CARP interacts with YB-1, a number of distinct criteria suggest that CARP is not a component of the major endogenous HF-1a-binding activity. Previous studies have identified the size of the major interacting co-factor to be 30× 10³M_r, while the size of CARP is approximately 40× 10³M_r. In addition, antibodies directed against CARP do not supershift the major endogenous HF-1a-binding activity (Zou and Chien, unpublished data) and CARP has a negative transcriptional regulatory effect on an HF-1-dependent promoter activity, as opposed to the positive regulatory effect identified for the endogenous HF-1a activity. Finally, GAL4 assays indicate that CARP contains domains that can act as negative transcriptional regulators, supporting a potential role of CARP in the negative regulation of a subset of ventricular muscle genes. In this manner, one might propose that CARP may act to antagonize the actions of p30 via interaction with YB-1, serving as a negative regulator of HF-1a-dependent genes by sequestration of YB-1. The analysis of CARP-deficient murine embryos should be valuable in identifying the precise role of CARP in the control of ventricular chamber specification, maturation and adaptation. Studies in the initial two hybrid screening identified other positive clones that interact with YB1, which are currently the subject of analyses to determine their relationship to the p30 YB-1 co-factor (Zou et al., 1995).

Recent studies have underscored that combinatorial pathways may be responsible for the activation of the cardiac muscle gene program (Evans et al., 1994; Ross et al., 1996). In addition, an increasing body of evidence suggests that discrete programs may exist for various regions of the heart: right ventricular chamber, left ventricular chamber, atrial and outflow tract (Ross et al., 1996). A number of nuclear factors have been implicated in the control of the cardiac muscle gene program and are expressed early during cardiogenesis, including MEF2 (Edmondson et al., 1994; Lilly et al., 1995), GATA-4 (Grepin et al., 1995; Jiang and Evans, 1996) and Nkx2-5 (Lints et al., 1993), dHAND and eHAND (Cserjesi et al., 1995; Olson and Srivastava, 1996). Each of these transcription factors are markedly enriched in heart, but are also expressed in a variety of other tissues. In fact, none of these genes appear to be expressed in an exclusively muscle-restricted fashion. The nuclear regulatory factors that might influence the cardiac muscle gene program and that would themselves be more restricted in their pattern of expression might serve as downstream target genes or cofactors that work to restrict the subsequent program of gene expression in the myocardium during the course of cardiogenesis. In this regard, CARP expression occurs specifically in myocardium at a relatively early stage during cardiogenesis (at least E8.5). In addition, recent studies analyzing embryos that harbor a knock-in of the lacZ reporter gene into the endogenous CARP locus also document early restricted expression of CARP in the primitive heart tube before the completion of heart tube fusion at early E8.0 (Chen, Kuo, Zou and Chien, unpublished results). At later points in time (after E9.5), trace detection of CARP promoter activity can be observed in skeletal muscle shown by the knock-in of lacZ. However, throughout embryogenesis, CARP expression is found primarily in heart. Trace levels in skeletal muscle, which can be revealed by lacZ, are not detected by in situ hybridizations. These studies suggest that CARP may be an excellent marker for early stages of the process of cardiac specification and elucidating the mechanisms that mediate the cardiac-restricted expression of CARP could be instrumental in identifying hierarchical signaling pathways that activate expression of the muscle gene program. In addition, it could become of interest to determine whether the expression of CARP in various regions of the heart tube, i.e., outflow tract, right ventricle, left ventricle and atrial, is under the control of separate genetic mechanisms. As noted in the current study, a 2.5 kb promoter fragment of the CARP promoter can direct a high level of cardiac-restricted expression of a lacZ reporter gene early during cardiogenesis. Thus, the CARP promoter may have utility for directing the early, high level cardiac-specific expression of genes for rescue strategies in various embryos harboring the loss of function of ubiquitously expressed genes that produce embryonic lethality due to cardiac defects, such as the RXRa (Sucov et al., 1994; Dyson et al., 1995; Gruber et al, 1996), β ARK-1 (Jaber et al., 1996), erb B2 (Lee et al., 1995), erb B4 (Gassmann et al., 1995) and neuregulin (Meyer and Birchmeier, 1995)-deficient embryos.

The absence of the MyoD family of myogenic factors in cardiac myocytes has led to the view of cardiac muscle as an exciting model system to study the genetic mechanisms of how these two closely related, but distinct mesodermal cell lineages achieve the differentiated myogenic phenotype. To date, the only candidate that may be capable of specifying the cardiac muscle cell lineage is the Drosophila homeobox gene tinman. tinman is expressed in the two major cell types of the Drosophila heart, the cardial and pericardial cells. In tinman mutants, no visceral mesoderm of the midgut or cardiac mesoderm forms, causing the loss of the dorsal vessel and its progenitors. tinman has been shown to regulate several genes that are important for visceral mesoderm development, including a related NK-homeodomain gene, bagpipe/NK-3 (Azpiazu and Frasch, 1993). Vertebrate homologues of tinman are highly conserved across species and have been identified in mouse (Komuro and Izumo, 1993; Lints et al., 1993), Xenopus (Tonissen et al., 1994; Evans et al., 1995), zebrafish and chick (Schultheiss et al., 1995). The mammalian homolog of the Drosophila tinman homeobox gene, Nkx2-5, is specifically required for ventricular chamber-specific myosin light chain-2 (MLC-2v) expression and looping morphogenesis during mammalian heart development.

While it is apparent that the Nkx2-5 gene is required for normal cardiogenesis, it is less clear how this homeobox genes triggers or influences downstream cardiac type-specific differentiated gene programs. Nkx2-5, as with many other homeobox genes, is expressed in a number of tissues (heart, pharyngeal endoderm, thyroid, tongue, spleen) and must interact with other regulatory circuits that control tissue-specific transcription. The current study suggests that one of the mechanisms by which Nkx2-5 may establish cell-specific gene programs is via the transcriptional activation of tissue-restricted nuclear cofactors, such as CARP, which in turn may participate in the regulation of differentiated gene programs.

In Nkx2-5^{− /−} embryos, carp expression was found to be significantly and selectively reduced by in situ hybridization and RNase protection studies. In addition, transient co-transfection assays have provided evidence that CARP promoter activity can be regulated by Nkx2.5 in the cardiac muscle cell. Finally, transgenic mice that harbor a 2.5 kb carp promoter-lacZ reporter gene display cardiac-specific expression of the transgene early during cardiogenesis. Breeding these carp transgenic mice into the Nkx2-5^{− /−} background has resulted in the significant reduction of the Carp-lacZ reporter gene in the Nkx2-5^{− /−} embryos, providing further direct evidence that the carp gene lies downstream of Nkx2-5. Interestingly, both the endogenous carp gene and the 2.5 kb carp-lacZ transgene display a preferential loss of expression in the conotruncal region of the heart tube. The carp expression in the conotruncus region is more sensitive to Nkx2-5 mutation. This study provided an Nkx2-5 downstream regulatory gene in the cardiac regulatory network that is highly restricted to heart. It should be noted that previously our laboratory has reported that a 250 bp MLC-2v-lacZ transgene, which displays high levels of expression in the conotruncus and bulbus cordis region, was not affected by the Nkx2-5 mutation (Ross et al., 1996). While it is clear that Nkx2-5 regulates carp at the transcriptional level, it is currently unclear whether Nkx2-5 has a direct effect or works through regulating other signals in a hierarchical fashion. This is currently under investigation.

Previous studies indicate that Nkx2-5 is required for the endogenous MLC2v gene expression (Lyons et al, 1995). However, the 250 base pair promoter of the MLC2v gene was not affected by the Nkx2-5 mutation (Ross et al, 1996). Therefore, it is probable that the positive Nkx2-5-dependent element lies upstream of the 250 base pair proximal promoter. CARP displays a negative effect on the HF1 element of the 250 base pair MLC2v promoter. Its negative function is in consistency with the reduction of its expression and the persistence of the 250 base pair MLC2v promoter activity in the Nkx2-5 mutant background. We propose that Nkx2-5 either directly regulates MLC2v or activates another factor(s) which in turn regulates MLC2v positively via an upstream element. In the meantime, Nkx2-5 also activates, directly or indirectly, the carp gene, which negatively modulates the MLC2v expression via the HF1 element so that the MLC2v expression can be kept at a desired level. Since Nkx2-5 is an early cardiogenic factor, it might activate a broader panel of genes, including positive and negative regulatory genes. These downstream genes then might play a role to shape the cardiac muscle-specific gene expression in response to other signals either developmental or, later on, physiological.

Many thanks to Wei Yan for the co-transfection experiments with Nkx2-5 constructs and the carp promoter. Dr Yu Chang Fu provided the rat Nkx2-5 cDNA. We thank Dr Roger Brent for providing the vectors, constructs and procedures for the interaction trap system (Gyuris et al., 1993) and Dr Rae Wu and Dr Gordon Gill for their valuable technical support at the initial stages of two-hybrid screening. Mahmoud Itani helped in sectioning the lacZ stained embryos. Dr Susumu Minamisawa helped in sequencing the GAL4CARP construct. Dr Steve Kubalak was an instrumental help in the maintenance of the Nkx2-5 knockout line and in embryonic dissection techniques. Y. Z. was supported by an institutional pre- and postdoctoral NRSA fellowship from NIH and K.R.C. was supported by grants from the NIH/NHLBI and the American Heart Association.

Note: During the course of this work, Dr Larry Kedes’ laboratory has independently cloned an identical rat carp gene via a differential display approach and has documented cardiac restricted expression during murine embryogenesis.