Cardiovascular basic helix loop helix factor 1, a novel transcriptional repressor expressed preferentially in the developing and adult cardiovascular system.

We have cloned a cardiovascular-restricted basic helix-loop-helix factor that interacts with arylhydrocarbon receptor nuclear translocator (ARNT) in a yeast two-hybrid screen. Cardiovascular helix-loop-helix factor 1 (CHF1) is distantly related to the hairy family of transcriptional repressors. We analyzed its expression pattern during mouse embryo development. At day 8.5, the expression of CHF1 is first detected in the primitive ventricle of the primordial heart tube and persists throughout gestation. In rat hearts, this expression is down-regulated after birth, concurrent with terminal differentiation of cardiomyocytes. In the developing vasculature, CHF1 first appears in the dorsal aorta at day 9.0, which precedes the reported expression of smooth muscle cell markers, and persists into adulthood. In an in vitro system of smooth muscle cell differentiation, CHF1 mRNA was barely detectable in undifferentiated cells but was induced highly in differentiated smooth muscle cells. To determine whether CHF1 might affect the function of ARNT, we performed transfection studies. Co-transfection of CHF1 inhibited ARNT/EPAS1-dependent transcription by 85%, and this inhibition is dose-dependent. In electrophoretic mobility studies, CHF1 inhibited the binding of the ARNT/EPAS1 heterodimer to its target site. Our data suggest that CHF1 functions as a transcriptional repressor and may play an important role in cardiovascular development.

Transcriptional control of cardiovascular development in vertebrates is a complex process requiring the coordinated expression of a variety of factors in a temporally and spatially defined manner (for review, see Refs. 1 and 2)). Gene knockout experiments in mice have identified factors such as Nkx 2.5, GATA-4, MEF2C, dHAND, and eHAND as regulators of various aspects of cardiac development (3)(4)(5)(6)(7)(8). However, the present picture is incomplete, particularly with regard to transcriptional control of smooth muscle differentiation in the developing vasculature.
Basic helix-loop-helix (bHLH) 1 proteins have been implicated in control of lineage commitment and differentiation in several cell types such as neurons, B cells, and skeletal muscle cells (9 -12). In skeletal muscle cells, the best studied proteins are the muscle regulatory factors such as MyoD and myogenin, which serve as nodal regulators for myogenesis by turning on the transcription of a cascade of skeletal muscle-specific genes (reviewed in Ref. 13). To date, no proteins that function analogously in cardiomyocytes or vascular smooth muscle cells (VSMC) have been identified. Although most bHLH proteins function as transcriptional activators, some bHLH proteins function as transcriptional repressors, including MyoR (14) and the hairy family of bHLH factors. Hairy-related factors have been implicated in the control of neuronal differentiation in developing mice (15,16).
Members of the hairy family of transcriptional repressors have three functional domains. They contain a proline in the basic region of the bHLH that affects the DNA-binding specificity, an Orange domain, and a WRPW motif at or adjacent to the carboxyl terminus (reviewed in Ref. 17). The bHLH region of the hairy proteins binds to class B E-boxes or N-boxes (18 -20). The Orange domain mediates interaction with specific transcriptional activators (21). The WRPW motif is critical for recruitment of members of the groucho family of corepressors (22).
Although initially identified in Drosophila, members of the hairy family are present in many species including humans. All hairy family members identified to date are expressed primarily in the nervous system where they are thought to inhibit differentiation. In mammals, the evidence for such a role comes primarily from studies of HES-1, the first mammalian homologue of hairy to be identified (18). Overexpression of HES-1 in the developing brain of mice leads to delayed neuronal differentiation (16). Targeted disruption of the HES-1 gene in mice leads to up-regulation of the proneural gene MASH-1 and premature neuronal differentiation, resulting in severe neural tube defects (15). HES-1 has also been implicated as a downstream effector of Notch in the lateral inhibition pathway (23). In PC12 pheochromocytoma cells that overexpress Wnt-1, HES-1 is up-regulated and the cells fail to differentiate in response to nerve growth factor (24).
To identify novel helix-loop-helix factors present in the cardiovascular system, we performed a yeast two-hybrid screen using the bHLH/PAS domain of ARNT as a bait to probe a human heart cDNA library. ARNT has previously been shown to have an important role in gene regulation by heterodimerization with several other factors, including the tissue-specific factor EPAS1 (25)(26)(27)(28). We identified two novel cDNAs that interact with ARNT in yeast. At the sequence level, these proteins are closely related to each other and distantly related to the hairy family of transcriptional repressors. Their patterns of expression are predominantly cardiovascular during development, consequently we named them "cardiovascular helixloop-helix Factor 1 and 2" (CHF1 and CHF2). Because CHF1 is expressed at high levels in the adult aorta we focused our initial efforts on CHF1. We found that CHF1 was highly expressed in developing cardiomyocytes and vascular smooth muscle cells but was down-regulated in adult heart when cardiac myocytes became terminally differentiated. CHF1 inhibited ARNT/EPAS1-dependent transcription in a dose-dependent manner. Thus, CHF1, like other hairy family members, can function as a repressor of transcription. In electrophoretic mobility shift assays, CHF1 inhibited the binding of the ARNT/ EPAS1 heterodimer to its target DNA-binding site. Taken together, these data suggest that CHF1 is likely to play an important role in cardiovascular development.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screening-A human heart cDNA library cloned into the yeast vector pACT2 was obtained from CLONTECH (Palo Alto, CA) and prepared according to the manufacturer's instructions. The yeast strain YRG-2 (Stratagene, La Jolla, CA) was made competent by the lithium acetate method (29). Approximately 10 10 yeast cells were transformed with 100 g of the ARNT bait plasmid pBDGAL4-ARNT and 100 g of library plasmid followed by plating on SD medium lacking tryptophan, leucine, and histidine. Surviving clones were then replated, transferred to paper filters, and screened for ␤-galactosidase activity as described (30). Clones that remained positive were then grown in liquid medium and yeast miniprep DNA containing the cDNAs were prepared as described (30). Yeast miniprep DNA was used to transform ultracompetent XL-1 blue cells (Stratagene, La Jolla, CA). The cDNA inserts were then sequenced by the dideoxy chain termination method with Thermo Sequenase (Amersham Pharmacia Biotech) or with an automated sequencer according to the manufacturer's instructions (Licor, Lincoln, NE). The interactions were verified by a quantitative ␤-galactosidase assay (31).
Cloning of Mouse Homologues by Low Stringency Hybridization-The mouse homologues of CHF1 and CHF2 were cloned by low stringency screening of a mouse 11-day embryo library (CLONTECH, Palo Alto, CA) with radiolabeled probes derived from the human cDNAs. A 0.3-kb EcoRI to BamHI fragment of the hCHF1 cDNA encoding the bHLH region was labeled with a random priming kit according to the manufacturer's instructions (Stratagene, La Jolla, CA). Approximately 1.6 million phage were plated and screened. Filters were washed with 0.5 ϫ SSC (75 mM sodium chloride, 7.5 mM sodium citrate) and 0.1% SDS at 42°C. To obtain the entire mouse CHF1 cDNA sequence, the 5Ј and 3Ј ends of the mouse cDNA were obtained by 5Ј-and 3Ј-rapid amplification of cDNA ends according to the manufacturer's protocol (Life Technologies, Inc., Gaithersburg, MD). To clone mouse CHF2, the entire 2.2-kb hCHF2 cDNA was used as a probe. The screening strategy was identical to that for mCHF1 except that the low stringency wash consisted of 0.2 ϫ SSC, 0.1% SDS at 42°C.
Construction of Plasmids-The human ARNT cDNA (25) was a gift from Oliver Hankinson (Los Angeles, CA). The ARNT bait plasmid pBDGAL4-ARNT was constructed by amplification of cDNA nucleotides 262 to 1581 encoding amino acids 70 -508 of the human ARNT protein by the polymerase chain reaction using the oligonucleotide primers 5Ј-ATAGTCGACACGATAAGGAGCGGT-3Ј and 5Ј-ATAGTCGACGA-CAAACGACAACGACGACG-3Ј. The amplified fragment was digested with SalI and cloned into the SalI site of the commercially available yeast bait vector pBDGAL4 Cam (Stratagene). The hCHF1 expression plasmid pcDNA-CHF1 was created by cloning a 0.6-kb BamHI-Acc65I fragment from the original yeast library clone pACT-CHF1 and a 0.8-kb Acc65I-XhoI fragment into the BamHI and XhoI sites of pcDNA3 (Invitrogen, Carlsbad, CA). The mammalian expression vector for E12 was made by polymerase chain reaction amplification of full-length E12 and cloning into pCR3 (Invitrogen, Carlsbad, CA). The yeast bait construct was made by taking an E12 insert encoding amino acids 477-654, which contains the bHLH region, from a previously described plasmid, pEG202-E12 (32) and cloning it into the EcoRI site of pBDGAL4. The mammalian expression vectors for human ARNT, endothelial PAS domain protein 1 (EPAS1), and the human vascular endothelial cell growth factor (VEGF) promoter/luciferase construct are described elsewhere (28). The cytomegalovirus promoter/MyoD expression plasmid pCS2MyoD6MT was obtained from Andrew Lassar (Boston, MA). The yeast GAL4 activation domain vector, pGAD424 was obtained commercially (CLONTECH, Palo Alto, CA).
The deletion mutants of hCHF1 were generated by cloning restriction fragments of the hCHF1 cDNA into the expression vector pFLAG-CMV-2 (Sigma). This plasmid contains the human cytomegalovirus promoter and an NH 2 -terminal methionine-FLAG epitope to allow expression of DNA fragments that do not contain an initiation codon. For amino acids 38 -337, a 1.1-kb EcoRI to XhoI cDNA fragment was cloned into the EcoRI and SalI sites of the vector. For amino acids 147-337, a 0.7-kb BamHI to XhoI fragment was cloned into the BglII and SalI sites of the vector. For amino acids 38 -146, a 0.3-kb EcoRI to BamHI fragment was cloned into the EcoRI and BamHI sites of the vector.
RNA Preparation, Northern Blot Analysis, and in Situ Hybridization-Total RNA was prepared from mouse organ tissue and cultured cells as described (33). Northern blot analysis was performed as described (33). In situ hybridization was performed as described (33). The mCHF1 riboprobe was derived from a 0.5-kb EcoRI to Acc65I fragment encoding the bHLH domain cloned into pBluescript II SK. 33 P-Labeled RNA probes were synthesized with T7 or T3 polymerase according to the kit manufacturer's instructions (Ambion, Austin, TX). Embryo sections were purchased from Novagen (Madison, WI) and deparaffinized prior to use according to the manufacturer's instructions. The SM22␣ and cyclin A probes for Northern analysis have been described previously (34,35).
Cell Culture, Transient Transfection, Luciferase, and ␤-Galactosidase Assays-Monc-1 cells were cultured and induced to differentiate into smooth muscle cells as described (34). NIH 3T3 cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. Cells were plated onto 6-well trays 1 day prior to transfection and were transfected with the indicated plasmids mixed with FuGENE 6 transfection reagent according to the manufacturer's protocol (Roche Molecular Biochemicals). Cells were harvested with passive lysis buffer and luciferase assays were performed according to the manufacturer's instructions (Promega, Madison, WI). ␤-Galactosidase assays were performed as described (36).
FIG. 1. CHF1 and CHF2 interact specifically with ARNT but not E12 in yeast. The yeast strain YRG-2 was transformed with plasmids encoding the indicated proteins in-frame with either the GAL4 DNA-binding domain or the GAL4 activation domain. Yeast were harvested in mid-log phase and ␤-galactosidase assays were performed as described under "Experimental Procedures." Electrophoretic Mobility Shift Assays-Electrophoretic mobility shift assays with in vitro transcribed and translated ARNT and EPAS1 binding to the HIF1-binding site were performed as described (28). In vitro transcribed and translated hCHF1 and E12 were generated using the plasmids pcDNA-hCHF1 and pCR3-E12, respectively (see above). All in vitro transcription and translation reactions were verified by SDS-PAGE.

RESULTS
Yeast Two-hybrid Screening-Out of approximately 2 ϫ 10 6 independent transformants, 230 yeast colonies initially grew on selective media. These colonies were then tested for ␤-galactosidase activity by a filter assay. Of the 22 ␤-galactosidase positive clones, one represented the entire open reading frame The amino acid sequences for hCHF1 and hCHF2 were deduced from the human cDNA sequences, while the hHES-1 amino acid sequence was obtained from GenBank. Alignment was performed using the Clustal V method included in the MegAlign module of the Lasergene software package (DNASTAR, Inc., Madison, WI). B, phylogenetic analysis of hairy family proteins. The amino acid sequences of hCHF1, mCHF1, hCHF2, and mCHF2 were deduced from their cDNA sequences. All other amino acid sequences were obtained from GenBank. Phylogenetic analysis was performed using the MegAlign module of the Lasergene software package (DNASTAR, Inc.). Bar denotes bHLH region. Dashed bar denotes Orange domain. * denotes critical glycine residue in basic region. **** denotes YRPW motif. of hCHF1 and two represented full-length hCHF2.
Isolation of Full-length Mouse cDNA Clones-After screening of 1.6 million plaques from the mouse 11-day embryo library with the human CHF1 probe, a total of 6 positive plaques were isolated, 3 were sequenced and one clone containing the entire open reading frame was characterized further. After sequencing this clone, 5Ј-and 3Ј-rapid amplification of cDNA ends were performed on 11-day mouse embryo RNA. Both 5Јand 3Ј-specific bands were isolated that overlapped with the original phage clone and gave a total length of 2,541 nucleotides, consistent with the predicted mRNA size on Northern blots. The same library was screened with the human CHF2 probe. A total of 9 clones were isolated, 2 were sequenced, and 1 represented a 2.2-kb full-length mCHF2 cDNA.
CHF1 and CHF2 Interact Specifically with ARNT-To verify that CHF1 and CHF2 interact with ARNT, quantitative yeast ␤-galactosidase assays were performed. The yeast strain YRG-2 was transformed with plasmids pBDGAL4-ARNT and pACT-CHF1, pACT-CHF2 or pGAD424 (a GAL4 activation domain vector). As a negative control, YRG-2 was also transformed with pBDGAL4-E12, containing the human E12 cDNA and either pACT-CHF1 or pACT-CHF2. Binding between SV40 T antigen and p53 was used as a positive control. As shown in Fig. 1, the combination of either CHF1 or CHF2 with ARNT but not E12 were sufficient to activate the endogenous ␤-galactosidase gene, indicating specific binding of CHF1 and CHF2 to ARNT in yeast.
Homology and Phylogenetic Analysis of CHF1 and CHF2-Computer data base searching initially revealed weak homology between hCHF1 and members of the hairy family of transcription factors at the protein level. A search with hCHF2 gave similar results. In order to determine the homology of CHF1 and CHF2 to other members of the hairy family, protein alignment and phylogenetic analysis were performed. In Fig.  2A, we aligned the open reading frame of hCHF1, hCHF2, and hHES-1, the first mammalian member of the hairy family to be identified (18). CHF1 and CHF2 are most homologous to hHES-1 in the bHLH region. CHF1 and CHF2 are distinct from hHES-1 and other hairy family members by the substitution of glycine for proline in the basic region and by the presence of a YRPW motif in place of the canonical WRPW at the COOH terminus. Similarly, the Orange domains (implicated in protein-protein interaction (21)) of CHF1 and CHF2 are highly related to each other and divergent from that of HES-1. CHF1 and CHF2 share an overall amino acid similarity of 51.3% while their similarities to hHES-1 are 22.9 and 23.9%, respectively. Because of these divergent features, which suggest that CHF1 and CHF2 are distinct from other members of the hairy family, we performed a phylogenetic analysis. As shown in Fig.  2B, CHF1 and CHF2 are most closely related to each other, and diverge early from other members of the hairy family.
Expression Pattern of CHF1 in Adult Tissues-Previously described hairy-related genes are expressed primarily in the nervous system. To determine the expression pattern of CHF1 in adult mice, Northern analysis was performed using RNA prepared from different mouse organs and a mouse CHF1 probe derived from the coding region (Fig. 3). This probe hybridized to an approximately 2.6-kb band expressed most abundantly in the adult aorta, with lower expression detected in the heart, brain, and lung. Similar results were obtained with a probe derived from the 3Ј-untranslated region (data not shown). CHF2 showed much lower expression in the adult heart and aorta (data not shown), so our initial efforts focused primarily on CHF1.
Expression Pattern of CHF1 in Developing Mouse Embryos-Given the importance of bHLH proteins in cell type specifica-tion and the established role of hairy-related genes in embryonic development, we examined the timing and tissue distribution of CHF1 expression in developing mouse embryos. As shown in Fig. 4, a very low level of CHF1 message was detected in whole embryos by Northern blot analysis as early as embryonic day 7.5. This low level expression of CHF1 persisted until day 11.5, when CHF1 mRNA was markedly increased. At this time point, there is a marked reorganization in branchial arch arteries, and septation begins within the heart and outflow tract. To identify the tissues expressing CHF1 in the developing mouse embryo, we performed in situ hybridization. CHF1 expression was detected in the developing ventricle as early as day 8.5, but was not detected in the neural tube and the bulbus cordis (Fig. 5). Expression of CHF1 at this time is concurrent with the reported expression of MLC2v, the earliest known marker of ventricular specification (37). At day 9.0, CHF1 mRNA was also detected in the developing paired dorsal aorta prior to fusion. CHF1 expression was observed primarily in the cells surrounding the primitive aortic endothelial tube, presumably in cells committed to the VSMC lineage. Its expression in this location occurred prior to the reported appearance of the smooth muscle markers SM22␣ and smooth muscle ␣-actin at day 9.5 (38), CRP2/SmLIM at day 10 (39), and smooth muscle myosin heavy chain at day 10.5 (40). At later stages of development (embryonic day 11.5-16.5), expression of CHF1 persisted in the ventricle and aorta (Fig. 5, and data not shown). At embryonic day 11.5, abundant CHF1 message was detected in mesenteric arteries, indicating that the expression of CHF1 in developing vessels is not limited to the aorta. At these time points, ARNT expression overlaps that of CHF1, suggesting that their interaction is biologically relevant (41).
Regulation of CHF1 Expression in Monc-1 Cells-To deter- mine whether CHF1 is induced during VSMC differentiation, we used an in vitro VSMC differentiation system. We have previously reported that the Monc-1 neural crest-derived cell line can be induced to differentiate with high efficiency into smooth muscle cells (34). As shown in Fig. 6, CHF1 is upregulated within 24 h of induction of smooth muscle differentiation and the CHF1 mRNA abundance persists at day 6, when Monc-1-derived VSMC are highly differentiated. This pattern of induction during the onset of differentiation in vitro is similar to the expression pattern in vivo (Fig. 5) and suggests an important role for CHF1 in VSMC differentiation.
Regulation of CHF1 Expression in Developing Rat Hearts-To explore the regulation of CHF1 expression in hearts during the transition from a proliferative to a terminally differentiated phenotype, we performed Northern blot analysis of RNA derived from rat hearts harvested at various time points before and after birth. As shown in Fig. 7, CHF1 RNA is abundant in fetal rat hearts and is markedly down-regulated 2 days after birth. It is noteworthy that down-regulation of CHF1 correlates with down-regulation of cyclin A, cell cycle withdrawal, and terminal differentiation of cardiomyocytes (35).
Regulation of ARNT-dependent Transcription-CHF1 interacts directly with ARNT, but the functional consequences of this binding are unknown. Others have shown that ARNT heterodimerizes with a variety of partners, such as the arylhydrocarbon receptor (25), hypoxia-inducible factor 1␣ (26), and EPAS-1 (27,28) to activate transcription of target genes. We have previously shown that ARNT and EPAS1 together heterodimerize and activate the VEGF promoter by binding to the HIF1-binding site (28). To assess the effect of CHF1 on ARNT/ EPAS1-dependent transcription, we transfected a plasmid containing the VEGF promoter and luciferase reporter gene with different combinations of expression plasmids encoding ARNT, EPAS1, and CHF1 into NIH3T3 cells, which do not express endogenous CHF1. As shown in Fig. 8A, ARNT and EPAS1 increase the VEGF promoter activity in NIH3T3 cells by 5-fold. In the presence of CHF1, induction of VEGF promoter activity is decreased by 85%. This inhibition is specific to CHF1 because E12 and MyoD expression plasmids did not block induction of the VEGF promoter (Fig. 8A). In contrast, these E12 and MyoD expression plasmids are able to increase the activity of the myogenin promoter (data not shown). The effect of CHF1 on ARNT-dependent transcription is also dose-dependent, as shown in Fig. 8B.
CHF1 Prevents Binding of ARNT/EPAS1 to the HIF1-binding Site-The HIF1 site in the VEGF promoter is known to confer inducibility by facilitating the binding of ARNT/EPAS1 or ARNT/HIF1␣ heterodimers (28). In order to determine the mechanism by which CHF1 inhibits VEGF promoter activation, we performed EMSAs to determine the effect of CHF1 on the binding of the ARNT/EPAS1 heterodimer to the HIF1binding site. As shown in Fig. 9, in vitro transcribed and translated ARNT ϩ EPAS1 form a protein-DNA complex with the HIF1 oligonucleotides and CHF1 abolishes this complex. To demonstrate that CHF1 inhibition is specific, a parallel reaction was performed with E12, which did not inhibit complex formation. To verify that this band is a specific complex, unlabeled HIF1 oligonucleotides (specific competitor) were also added to the reaction and inhibited formation of the ARNT/ EPAS1-labeled complex. In contrast, a nonspecific competitor oligonucleotide did not inhibit formation of the observed complex.
Mutation of CHF1 Abolishes Transcriptional Repression-As a first step in mapping the domain of CHF1 required for transcriptional repression, we generated a series of deletion mutants, as shown in Fig. 10. In transient transfection analysis, full-length CHF1 abolished ARNT/EPAS1-dependent transcription. Deletion of amino acids 1-37 had no effect. In contrast, deletion of amino acids 1-146, removing the bHLH do- FIG. 5. In situ analysis of CHF1 expression in developing mouse embryos. Mouse embryo sections were hybridized with 33 P-labeled riboprobes as described under "Experimental Procedures." All sections were probed with both antisense and sense probes (not shown). Sense probes gave no appreciable background. A, bright field microscopy of mouse embryo sections from indicated developmental time points. B, dark field microscopy of the same sections probed with CHF1 antisense riboprobe. Abbreviations: A, atrium; AA, ascending aorta; DA, dorsal aorta; V, ventricle; MA, mesenteric artery.

FIG. 6. CHF1 is induced during smooth muscle differentiation in vitro.
Monc-1 cells were induced to differentiate into smooth muscle and harvested for RNA at the indicated time points. Total RNA (10 g) from each time point was analyzed on MOPS-formaldehyde 1.3% agarose gels and transferred to nitrocellulose. The blot was hybridized with a mouse CHF1 probe as described in the legend to Fig. 3. The blot was also probed for SM22␣ expression and with an oligonucleotide complementary to 28 S RNA to control for variability in loading.
FIG. 7. CHF1 is down-regulated in developing rat hearts. Rat hearts were harvested for RNA at the indicated time points. Fetal RNA was from day 18 of gestation. Total RNA from each time point (10 g) was separated, blotted, and probed as described in the legend to Fig. 3. The blot was also probed for cyclin A expression to verify withdrawal of cardiomyocytes from the cell cycle. main and part of the Orange domain, abolished repression. Deletion of amino acids 1-37 and 147-337, leaving primarily the bHLH domain, also abolished repression. Taken together, these results suggest that the bHLH domain is necessary but not sufficient for repression and an intact Orange domain or other domains in the COOH terminus are also required for repression of ARNT-dependent transcription. DISCUSSION We have isolated two novel cDNAs from a human heart library encoding developmentally regulated bHLH proteins that interact with the bHLH-PAS motif of ARNT in yeast ( Fig.  1). At the amino acid level, these proteins are most related to the hairy family of transcriptional repressors, which have been implicated in the control of neuronal differentiation. CHF1 and CHF2 diverge from other members by the presence of a glycine in place of proline within the basic region, and by the presence of a YRPW motif in place of the WRPW motif at or near the carboxyl terminus ( Fig. 2A). The proline residue appears to be critical for DNA-binding specificity, and is thought to explain the divergent binding specificity of hairy-related proteins from other bHLH proteins. In general, members of the hairy family bind to class B E-box motif of the consensus sequence CA(C/ T)GTG as well as the N-box sequence CACNAG. Interestingly, recombinant CHF1 and CHF2 do not bind to E-boxes and N-boxes. 2 Thus, they may regulate a different set of genes via an as yet unidentified binding site.
In all hairy family members, the WRPW motif is critical for interaction with members of the groucho family of transcriptional co-repressors (22). It is unknown whether the YRPW motif in CHF1 and CHF2 can also interact with groucho-related proteins. Of note, the related sequence WRPY in the runt family of transcription factors can also interact with groucho, and the interaction may be regulated (42).
In Drosophila hairy, the Orange domain, in conjunction with the bHLH domain, is thought to mediate repression of specific transcriptional activators, such as Scute (21). Our analysis of the interaction between CHF1 and ARNT/EPAS1 is consistent with these findings. Because the bHLH and Orange domain are not highly conserved between CHF1 and other hairy family members, CHF1 may regulate a specific class of transcriptional activators that are distinct from those regulated by other hairy family members.
We have also cloned the mouse homologue of CHF1 and determined its expression in mouse embryos and adults. CHF1 is expressed early in the developing ventricle and vasculature and this pattern of expression persists into adulthood (Figs. 3 and 5). Expression of CHF1 is correlated with the onset of smooth muscle differentiation in developing vasculature and in an in vitro model of smooth muscle differentiation (Figs. 5 and 6). In the developing ventricle, CHF1 expression correlates with the reported expression of early ventricular markers such as MLC2v (37). Interestingly, CHF1 expression is down-regulated in isolated rat hearts as a function of age (Fig. 7). The time course of down-regulation correlates with down-regulation of cyclin A and withdrawal from the cell cycle (35). In the adult, CHF1 is down-regulated in the heart but continues to be expressed highly in the aorta. Since cardiomyocytes are terminally differentiated and have lost the capacity to divide, while VSMCs in the aorta retain the ability to re-enter the cell cycle, it is tempting to speculate that CHF1 may function as a repressor to prevent terminal differentiation. Total DNA was kept constant by the addition of the appropriate amounts of pcDNA3. For all transfections, an SV40 promoter-driven ␤-galactosidase reporter plasmid, pSV␤gal (0.25 g, Promega) was included to correct for differences in transfection efficiency. B, CHF1 repression is dose-dependent. NIH3T3 cells were transfected with the reporter plasmid pGL2hVEGF (0.25 g), the ARNT expression plasmid phARNT (0.5 g), the EPAS1 expression plasmid phEP-1 (0.5 g), and varying amounts of pcDNA-CHF1 (0, 0.125, 0.25, and 0.5 g). Total DNA was kept constant by the addition of appropriate amounts of pcDNA3 vector. All transfections were performed in triplicate at least three times. Relative VEGF promoter activity is presented in arbitrary units (mean Ϯ S.E.). * denotes statistical significance when compared with ARNT/EPAS-1 induced activity by one-way analysis of variance (ANOVA) with p value Ͻ0.05.
FIG. 9. CHF1 inhibits binding of the ARNT/EPAS1 heterodimer to the HIF1-binding site. The indicated proteins were expressed by in vitro transcription and translation and incubated with radiolabeled HIF1 oligonucleotides as described under "Experimental Procedures." Protein-DNA complexes were separated on 4% low ionic strength, nondenaturing gels, dried, and exposed. Abbreviations: SC, specific competitor oligonucleotide; NSC, nonspecific competitor oligonucleotide.
We have found that CHF1 and CHF2 interact with ARNT in a yeast two-hybrid screen (Fig. 1). This interaction with ARNT is specific, as CHF1 and CHF2 do not interact with the bHLH protein E12. In transient transfection assays, CHF1 functions as a transcriptional repressor by interfering with ARNT-dependent activation of the VEGF promoter, while E12 and MyoD do not (Fig. 8). The mechanism by which this occurs is through inhibition of ARNT/EPAS1 binding to the HIF1 site (Fig. 9). Targeted disruption of the ARNT gene in mice has been shown to cause defects in yolk sac angiogenesis (43). In the yolk sac vessels of knockout mice, the smooth muscle layer is deficient. It is believed that the vascular phenotype in ARNT-null mice can be attributed, at least partially, to reduction of VEGF, which is critical for normal blood vessel development (44,45). Our findings that CHF1 is expressed in developing vasculature, regulates ARNT/EPAS1-dependent activation of the VEGF promoter (and other as yet unidentified promoters) suggest a potential mechanistic link between this protein and vascular development. For example, one potential role for CHF1 may be to control the timing of smooth muscle development in the medial layer, in a manner analogous to the function of HES-1 in the developing nervous system.
To our knowledge, CHF1 is the first tissue-restricted bHLH protein to be expressed in vascular smooth muscle. In the future, we hope to determine the contribution of CHF1 and CHF2 to organismal development by targeted disruption of their genes in mice. We also hope to elucidate the role of CHF1 in the adult vasculature by identification of its downstream target genes.