The Homeodomain Transcription Factor NK-4 Acts as either a Transcriptional Activator or Repressor and Interacts with the p300 Coactivator and the Groucho Corepressor*

NK-4 (tinman) encodes an NK-2 class homeodomain transcription factor that is required for development of the Drosophila dorsal mesoderm, including heart. Genetic evidence suggests its important role in mesoderm subdivision, yet the properties of NK-4 as a transcriptional regulator and the mechanism of gene transcription by NK-4 are not completely understood. Here, we describe its properties as a transcription factor and its interaction with the p300 coactivator and the Groucho corepressor. We demonstrate that NK-4 can activate or repress target genes in cultured cells, depending on functional domains that are conserved between Drosophila melanogaster andDrosophila virilis NK-4 genes. Using GAL4-NK-4 fusion constructs, we have mapped a transcriptional activation domain (amino acids 1–110) and repression domains (amino acids 111–188 and the homeodomain) and found an inhibitory function for the homeodomain in transactivation by NK-4. Furthermore, we demonstrate that NK-4-dependent transactivation is augmented by the p300 coactivator and show that NK-4 physically interacts with p300 via the activation domain. In addition, cotransfection experiments indicate that the repressor activity of NK-4 is strongly enhanced by the Groucho corepressor. Using immunoprecipitation and in vitropull-down assays, we show that NK-4 directly interacts with the Groucho corepressor, for which the homeodomain is required. Together, our results indicate that NK-4 can act as either a transcriptional activator or repressor and provide the first evidence of NK-4 interactions with the p300 coactivator and the Groucho corepressor.

Homeobox genes encode sequence-specific DNA-binding proteins that play important roles during development by regulating transcription in organisms as diverse as insects, nematodes, mammals, and plants (1,2). These homeodomain transcription factors can act as transcriptional activators or repressors that require separate but sometimes overlapping functional domains including a DNA-binding domain for their transcriptional activities (3)(4)(5). In addition, it was recently demonstrated that homeoproteins interact with other transcription factors that lead to their functional specificity in vivo, utilizing a specific but conserved domain of the homeoproteins for these interactions (5)(6)(7)(8)(9)(10)(11). In general, sequence-specific DNA-binding proteins recruit a coactivator or corepressor for their transcriptional activation or repression of target genes (12,13), yet the mechanism of activation or repression by homeoproteins is still poorly understood.
The NK-4 homeobox gene tinman (14 -17), belongs to the NK-2 class that includes a large number of vertebrate homeobox genes (18 -20). Analysis of tinman mutant embryos revealed that tinman function is required for development of the dorsal mesoderm including the heart during Drosophila embryogenesis (16,17). Additionally, tinman activity is required for the formation of glia-like dorsal-median cells (21), gonadal mesoderm (22)(23)(24), and a subset of somatic body wall muscles (16). In mammals, targeted disruptions of the mouse Nkx-2 genes (Nkx-2.1/TTF1, Nkx-2.2, and Nkx-2.5) revealed their functions in organogenesis including heart development (25)(26)(27). Likewise, it has recently been revealed that human NKX2-5 is involved in nonsyndromic, human congenital heart disease (28). Also, cross-phylum rescues of the NK-2 class of homeobox genes have recently been demonstrated (29,30). For example, zebrafish nkx 2.5 efficiently rescues a ceh-22 mutant when expressed in the pharyngeal muscle of Caenorhabditis elegans (29), and mouse Nkx2-5 rescued only visceral mesoderm when tested for its ability to rescue the tinman mutant phenotype of Drosophila (30). Thus, genetic evidence clearly indicates that NK-4 plays an important role during development and suggests that its function is well conserved in the animal species.
NK-4 is initially expressed in the presumptive mesoderm at the cellular blastoderm stage, continues to be expressed in all mesodermal cells during germ band elongation, and eventually becomes restricted to dorsal mesodermal cells including precursor cells of the heart (15). Recently, NK-4 has been shown to be regulated by the myogenic factor Twist during embryogenesis (31,32). In addition, previous work has shown that inductive signals from dorsal ectodermal cells such as Dpp can induce tinman expression in the dorsal mesoderm (33)(34)(35). Similarly, chicken Nkx-2.5, a potential vertebrate homologue of the Drosophila NK-4 gene, was shown to be induced by BMP-2 and BMP-4 (36), suggesting that regulatory pathways for its induction are also conserved. We have previously demonstrated that NK-4 binds to a target sequence (5Ј-TCAAGTG-3Ј) (31,37) that is related to the binding sites of other NK-2 class homeodomain proteins (20, 31, 38 -41) and autoregulates its own gene in cultured cells (31). Additionally, it has recently been shown that NK-4 directly activates the D-mef2 gene in cardial cells (37) and ventral muscle founder cells (42), making D-mef2 the first known in vivo target of NK-4 to be described. Cofactors of NK-4 are totally unknown. However, physical interaction of Nkx-2.5 with GATA-4 has been reported, suggesting that they may act as mutual cofactors (43)(44)(45).
Despite the growing body of evidence that supports an important role for NK-4 during embryogenesis, the properties of NK-4 as a transcriptional regulator and the mechanism of gene transcription by NK-4 are still poorly understood. In this study, we characterize the functional domains of NK-4 that are responsible for its transcriptional activity. Our analysis reveals that NK-4 acts as either a transcriptional activator or repressor that utilizes separate functional domains for its transcriptional activity. Furthermore, we provide the first evidence of interaction of NK-4 with the p300 coactivator and the Groucho corepressor.
Site-directed Mutagenesis-A PCR-based method was used to generate mutations within either the TN domain or the homeodomain. For the NK4A6TDm construct that contains a mutation (Phe 3 Ala at aa 38, TTT 3 GCT) within the TN domain, DNA fragments (TDm, aa 1-43) were amplified by PCR with specific primers (primer 20 and primer NK4FA (5Ј-GTTCAGGATATCCTTGACCGAAGCTGGAGTGG-TGTTCAGGGC-3Ј), and amplified DNA fragments were codigested with EcoRI and EcoRV. The corresponding region of the NK4A6 construct was replaced by the TDm DNA fragments to generate the NK4A6TDm expression vector. For the A6-NK4HD(NQ) construct that contains a mutation (Asn 3 Gln at aa 351, AAT 3 CAG) at the helix III region of the homeodomain, two separate DNA fragments (fragment A, aa 188 -356; fragment B, aa 345-416) were amplified by PCR with specific primers (fragment A, primer 422 and primer 9540 (5Ј-CGATT-TGTAGCGCCGCTGCTGGAACCAAATCTTCACTTG-3Ј); fragment B, primer 9541 (5Ј-CAAGTGAAGATTTGGTTCCAGCAGCGGCGCTACA-AATCG-3Ј and primer 502). Mixed DNA fragments (A and B) were denatured and reannealed. DNAs (39-nucleotide overlap) were gapfilled with Vent DNA polymerase, and gap-filled DNAs were subjected to PCR with specific primers (primer 422 and primer 502). Amplified DNAs were gel-eluted and subcloned into the EcoRI/XbaI sites of the pSG424 expression vector to generate the NK4HD(NQ) plasmid. For the A6NK4HD(SA) construct that contains a mutation (Ser 3 Ala at position 309, TCC 3 GCC) at the helix I region of the homeodomain, two separate DNA fragments (fragment C, aa 267-309; fragment D, aa 309 -364) were amplified by PCR with specific primers (fragment C, primer 9539 (5Ј-CAAGAATTCGATAACAGCCAGGTGACC-3Ј and primer NK4SA-1 (5Ј-GACTGAAGACTGGGCAAAGAGCACGCGAGG-CTTTCG-3Ј); fragment D, primer NK4SA-2 (5Ј-GACTGAAGACTTTG-CCCAGGCACAGGTCCTGGAGCT-3Ј) and primer 9538). DNA fragments (C and D) were digested with BbsI and ligated together. The ligated DNAs were codigested with EcoRI and XbaI and then subcloned into the corresponding sites of the pSG424 vector to generate the NK4HD(SA) plasmid. Mutations and open reading frames of the clones were confirmed by sequencing.
Expression Vectors, Cell Transfections, and CAT Assays-Cell growth, transfections into cells with indicated reporters and expression vectors using the calcium phosphate precipitation method, normalization of transfection efficiency with ␤-galactosidase activity, and CAT assays with a CAT enzyme-linked immunosorbent assay kit (Roche Molecular Biochemicals) were performed as described (31). For the constructions of truncated NK-4 expression vectors (d8 and d6), DNA fragments (121 bp, from 5Ј-untranslated region to the first ATG codon) were amplified by PCR with specific primers (primer NK453 (5Ј-TAA-GTCGACGCGGCCGCGTACCCAGATTCCAATTC-3Ј) and primer LEE10 (5Ј-GATGAATTCCATCCTCGCTGTGCGAT-3Ј)). Amplified DNA was codigested with EcoRI and SalI and subcloned into the EcoRI/ SalI sites of pBluscript vector in order to generate the pKSII-NK4N plasmid. Subsequently, coding regions of NK-4 from GAL4-NK-4 chimeras (A18 and A2 in Fig. 2A, d8 and d6 in Fig. 1, respectively), which were obtained by codigestion with EcoRI and XbaI followed by gel elution, were subcloned into the EcoRI/XbaI sites of the pKSII-NK4N plasmid to generate the pKSII-NK4d8 and the pKSII-NK4d6 plasmids. These plasmids were digested with NotI, and DNA fragments were subcloned into NotI site of pRC-CMV expression vector (Invitrogen) to generate d8 or d6 expression vectors. Expression vector for full-length NK-4 (NK-4, CMV-NK-4) and reporter plasmids (P1, P1E2m, E2CAT, 1 The abbreviations used are: PCR, polymerase chain reaction; CAT, chloramphenicol acetyltransferase; GRO, Groucho; GST, glutathione S-transferase; bp, base pair; aa, amino acid(s); RD, repressor domain; HD, homeodomain; PAGE, polyacrylamide gel electrophoresis; TN, Tinman. and E2mCAT) were described previously (31). The G5EnXCAT reporter plasmid, which contains SV40 enhancer, was described previously (47). For the construction of the Groucho expression vector, PCR-amplified cDNAs were subcloned into the pKSII vector, and the resulting plasmid (pKS-GRO) was confirmed and used for the construction of the Groucho expression vector (pCI-GRO) by inserting cDNA into EcoRI/SalI sites of the pCI vector (Promega). For the construction of the Myc-tagged Groucho expression vector, the NcoI/XbaI DNA fragment from the pSPUTK-GRO was subcloned into the pCS3(ϩ)MT expression vector. For the construction of the GST-NK-4 fusion protein expression vectors, DNA fragments of the corresponding regions of NK-4 were excised from GAL4-NK-4 fusion constructs ( Fig. 2A) and subcloned into either the EcoRI site of the pGEX5X-1 plasmid or into the EcoRI/XbaI sites of the pGEX5X-1Xb plasmid containing an extra XbaI site. For the constructions of GST-TD and GST-TDm expression vectors, corresponding regions were amplified (primer 20 and primer 500) from NK4A6 and NK4A6TDm, respectively, and subcloned into pGEX5X-1. Expression and purification of the fusion proteins were performed as described previously (31). p300N-(1-670), p300M-(671-1194), and p300C-(1135-2414) plasmids were described previously (48). For the constructions of the truncated p300 expression plasmids, the NK-4, and the Groucho expression plasmids used for in vitro translation (Figs. 4 and 5), DNA fragments of the corresponding regions of p300 and Groucho were amplified by PCR with specific primers (in the case of construction of the NK-4 expression vector, the corresponding regions of the NK-4 were obtained from GAL4-NK-4 constructs) and subcloned into the EcoRI/ SalI sites of the pSPUTK vector. Primer sets used for amplifications are as follows: p300(300 -450), primer p300N1 (5Ј-GATGAATTCCCCAA-CATGGGTCAACAGCCA-3Ј) and primer p300C1 (5Ј-GATGTCGACG-GCAGACTGTTGACCCACCCC-3Ј; p300(451-670), primer p300N2 (5Ј-GATGAATTCCCCAACCTAAGCACTGTTAGT-3Ј) and primer p300C2 (5Ј-GATGTCGACATTCATGGAAACTGGAACCAT-3Ј); p300(620 -1891), primer p300N3 (5Ј-GATGAATTCCCCTGCGATCTGATGGATGGT-3Ј) and primer p300C3 (5Ј-GATGTCGACTTGAGTCCTGGGCAAGTAGGG- In Vitro Pull-down Assays and Immunoblot Analysis-For pull-down assays, p300, NK-4, and groucho cDNAs were subcloned into pSPUTK vector (Invitrogen) and subjected to in vitro translation using the TNT Coupled Reticulocyte Lysate System (Promega). Pull-down assays were performed by incubating equal amounts of GST or indicated GST-NK-4 and GST-Groucho fusion proteins, immobilized onto glutathione-Sepharose beads (Amersham Pharmacia Biotech), with in vitro translated, 35 S-labeled p300, Groucho, and NK-4 proteins. The mixture was placed on ice for 2 h and washed five times with buffer A (20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% Nonidet P-40), and bound proteins were eluted, separated by 12% SDS-polyacrylamide gel electrophoresis, and autoradiographed. In the case of interactions of Groucho with NK-4, precipitates were washed five times under high stringency conditions (20 mM Tris-HCl, pH 8.0, 350 mM NaCl, 1% Nonidet P-40). For immunoblot analysis, any pull-down proteins were separated on a standard 12% SDS-polyacrylamide gel and transferred to a nylon membrane (NEN Life Science Products). Filters were prehybridized in TBST (10 mM Tris-HCl, pH 8.0, 50 mM NaCl, 0.05% Tween 20) containing 5% nonfat dry milk followed by incubation with either an anti-p300 antibody (4 g/ml) (Upstate Biotechnology) or with an anti-Myc antibody (1 g/ml) (Invitrogen) overnight at 4°C. After three washes in TBST, secondary horseradish peroxidase-conjugated antibody (Life Technologies) was added in TBST containing 5% nonfat dry milk, incubated for 1 h, and then filters were washed three times in TBST. The signal was detected with chemiluminescence (Super Signal TM ; Pierce).

NK-4 Can Act as either a Transcriptional Activator or Repressor and Contains Multiple Functional Domains-Previ-
ously, we showed that NK-4 is regulated by Twist and autoregulates its own promoter (31). We further characterized the properties of NK-4 as a transcription factor using the same reporters (Fig. 1, P1 and P1E2m) and truncated forms of the NK-4 expression vectors (Fig. 1, d8 and d6). As described pre-viously (31), the P1 reporter contains NK-4-responsive elements (the E2 cluster) and is activated by NK-4. The P1E2m reporter contains mutated NK-4 binding sites but otherwise is exactly the same as the wild-type P1 reporter, which also contains several weak NK-4 binding sites. As shown in Fig. 1, we found that NK-4 can activate the P1 reporter (6-fold activation, lane 2). In contrast, NK-4 down-regulates the P1E2m reporter gene (3-fold repression, lane 6). In this case, NK-4 binds to weak binding sites and represses the P1E2m reporter gene. These results indicate that NK-4 can act either as a transcriptional activator or repressor, depending on the context of the reporters (P1 or P1E2m). Additionally, we demonstrate that this phenomenon is dependent on the functional domains of NK-4 ( Fig. 1). For example, deletion of the amino-terminal region of NK-4 ( Fig. 1, d8 construct) abrogates activation of the P1 reporter (Fig. 1, lane 3), indicating that the amino terminus (aa 1-110) of NK-4 is required for transcriptional activation. Indeed, this NK-4 mutant (d8) represses gene expression of both the P1 and the P1E2m reporter (Fig. 1, lanes 3 and 7). Further deletion of NK-4 ( Fig. 1, construct d6) relieved this repression, irrespective of the reporter gene used (Fig. 1, lanes  4 and 8). These results suggest that the region following the amino terminus of NK-4 (aa 111-188) is required for the repressor activity of NK-4. Taken together, these results indicate that, depending on the context of the target genes (for example P1 or P1E2m; see lanes 2 and 6), NK-4 can act as either a transcriptional activator or repressor and that these different transcriptional activities are dependent on functional domains of NK-4.
In order to define the domains responsible for transcriptional activation and repression in great detail, various GAL4-NK-4 chimeras were generated by inserting DNAs coding for different regions of NK-4 into the pSG424 expression vector that contains a heterologous GAL4 DNA binding domain ( Fig. 2A). In these experiments, the normalized CAT activity (CAT activities were corrected for transfection efficiencies by ␤-galactosidase assays), obtained from cotransfection with a test effector and reporter plasmid (lanes 1-4, P1; lanes 5-8, P1E2m), was divided by the corresponding value obtained with the indicated reporter and pRC-CMV (empty vector) and is shown as relative CAT activity.
CAT activities from cells cotransfected with the indicated plasmids and a G5BCAT reporter plasmid were measured and compared with activities from cells cotransfected with pSG424 as a control (Fig. 2B). From this analysis, we found that the amino-terminal region of the NK-4 protein (construct A6, aa 1-110) contains a transcriptional activation domain, which is consistent with results described above (Fig. 1). Thus, this activation domain functions both in intact NK-4 and in a fusion form when tethered to the heterologous GAL4 DNA binding domain. Within this activation domain, the Gln, Asp, and Glu residues are enriched (24%). Of note is the presence of the TN domain (aa 39 -47) that is well conserved in the NK-2 class homeoproteins (19) and the YYD tripeptide (aa 14 -16) that is similar to the FYD motif that was found in the activation domains of MyoD, myogenin, and MRF (49). In fact, deletion of the first 13 amino acid residues within the activation domain (Fig. 2B, construct A10) did not affect transcriptional activation. However, further deletion of both the YYD motif and the TN domain (Fig. 2B, construct A14) markedly reduced transcriptional activation.
To map the repressor domain (RD) of NK-4, we constructed the G5EnXCAT reporter plasmid by inserting the SV40 enhancer into the G5BCAT reporter ( Fig. 2A). The G5EnXCAT reporter alone showed very high CAT activity so that we could measure the repression activity by NK-4 more easily. The A1 construct showed strong repressor activity when cells were cotransfected with the G5EnXCAT reporter (Fig. 2C). Constructs A12 and A18, which contain a deletion of the activation domain, showed comparable repressor activities (Fig. 2C, lanes  A12 and A18, respectively). Further deletion of NK-4 domains (constructs A19, A2, A23, and A24) revealed that both the region following the activation domain (RD, aa 110 -188) and the homeodomain (HD) are important for repressor activity. Interestingly, either domain alone (constructs A31-A34) showed weak repressor activity. Strong repressor activity, however, was restored when both domains were combined (constructs A35 and A18). Thus, when fused to the heterologous GAL4 DNA binding domain, these domains (RD and HD) can act as a transcriptional repressor domain. These results are consistent with our observation that deletion of the RD (Fig. 1, were cotransfected with the G5BCAT reporter (1 g/transfection) into CV1 cells, and CAT activities were measured. In this series of experiments, the normalized CAT activity obtained from transfection with a test expression vector was divided by the corresponding value obtained with pSG424, and -fold activation values of GAL4-NK-4 chimeras are shown (averages of three sets of independent experiments). C, identification of repression domains. The G5EnXCAT reporter (1 g/transfection) was cotransfected with various GAL4-NK-4 chimeras (listed on the left) as described above, and CAT activities were measured. The normalized CAT activity obtained from transfection with G5EnXCAT and pSG424 was divided by the corresponding value obtained with a test expression vector, and -fold repression of the GAL4-NK-4 chimeras is shown (averages of three sets of independent experiments). construct d6) caused a loss of repressor activity of NK-4 and support the idea that the RD is necessary for the repressor activity of intact NK-4. Within this region (aa 111-188), strong homology (69% identity) between Drosophila virilis NK-4 (DvNK-4) and NK-4 was observed. 2 Taken together, these results indicate that NK-4 has both transcriptional activation and repressor domains.
The Homeodomain Is an Inhibitory Domain-As described above, although construct A1 contains a strong activation domain (aa 1-110) (Fig. 2B, lane A6), it still could not activate the G5BCAT reporter gene (Fig. 2B, lane A1). These results suggested that NK-4 contains inhibitory domains that suppress the transactivation function of NK-4. Since the homeodomain (in conjunction with the RD) showed repression of reporter gene expression, we suspected that the homeodomain might provide such an inhibitory function. In order to explore this possibility, various constructs were generated by the in-frame fusion of different domains of NK-4 to construct NK4A6 that contains the activation domain of NK-4 (aa 1-110) (Fig. 3A). As shown in Fig. 3, the carboxyl-terminal half of NK-4 including the homeodomain inhibited transactivation (Fig. 3B, lane 3), and the region containing the NK-4 homeodomain (aa 267-364) fully inhibited the transactivation function of the activation domain (Fig. 3B, lane 4). Indeed, we found that the NK-4 homeodomain itself is sufficient to show an inhibitory effect (Fig. 3B, lane 5). This result is consistent with the previous reports that the repressor activities of HoxA7 and Msx-1 are mediated by their homeodomains (50,51). This phenomenon seems to be a general characteristic of the homeodomain, since the same effect is seen with the NK-1 homeodomain (aa 531-628) (Fig. 3B, lane 6). Also, it has recently been demonstrated that the Dfd homeodomain masks the Dfd activation function  (52). The mutant NK-4 homeodomain (N 3 Q mutation in the helix 3 of the homeodomain) that is defective in binding DNA still showed this inhibitory effect (Fig. 3B, lane 7), suggesting that DNA binding is not involved in this inhibition.
Other transcription factors such as ATF-2 also showed inhibition of transactivation by their DNA binding domains (53,54). For example, the bZIP DNA-binding domain of ATF-2 suppresses the activation region by an intramolecular interaction, and the inhibition observed is activation domain-specific (53). We found that the amino terminus of NK-4 (aa 1-53) that contains the TN domain showed weak interaction with the NK-4 homeodomain (Fig. 3E, lane 3). Thus, one possible explanation for this inhibition is that the homeodomain can interact intramolecularly with the activation domain and could mask the transactivation function. To test this possibility, we generated the construct NK4A6TDm (Fig. 3A), which has a point mutation within the TN domain (Phe 3 Ala mutation). This mutation inhibits interaction between the TN domain and the homeodomain (Fig. 3E, lane 4). However, this construct still showed strong transactivation (Fig. 3C, lane 3), and the transactivation was inhibited by the homeodomain as efficiently as the wild type (Fig. 3C, lanes 4 -6). Moreover, we found that the homeodomain can also suppress the transactivation function of VP16 (Fig. 3D, lanes 6 -8). These results suggest that the inhibitory function of the homeodomain is not activation domain-specific and may not be caused by intra-molecular interaction. Rather, these results imply that the homeodomain itself might be a repressor domain that may directly serve as a protein-protein binding interface with a co-repressor complex or with the basal transcription machinery. Consistent with this notion, the homeodomains of both HoxA7 and Msx-1 showed repression activity, specifically through residues in the N-terminal arm of the homeodomain, and the Msx-1 homeodomain interacts with the TATA binding proteins (50,51). It is of interest to note that one hypomorphic allele of the even-skipped homeoprotein containing a mutation at a residue in the aminoterminal arm (Thr-residue at the ninth position of the homeodomain) was reported (55). Since this residue (either Thr or Ser) is highly conserved in the homeodomains of known homeodomain proteins (18), we mutated this residue in the NK-4 homeodomain and tested its effect on the inhibitory effect of the homeodomain (Fig. 3A). Interestingly, we observed that this point mutation (S 3 A mutation) partially relieved the inhibitory effect (Fig. 3D, lane 4). We do not know how this mutation alleviates the inhibitory effect of the homeodomain, but we speculate that it may affect protein-protein interaction of the homeodomain with unknown partner. Taken together, we defined the homeodomain as a transcriptional inhibitory domain, adding an another function of the homeodomain in addition to the known DNA and RNA binding activities (56,57).
p300 Augments NK-4-mediated Transactivation and Physically Interacts with NK-4 -It was recently shown that the p300/CBP coactivator enhances transcription mediated by various transcription factors including CREB, AP-1, MyoD, and nuclear hormone receptors (58). Hence, we reasoned that p300 may be required for the transactivation of homeoproteins as well. In order to explore this possibility, we tested whether p300 can enhance NK-4-dependent transactivation using transient expression assays. For testing NK-4-dependent reporter gene activation, two different reporters containing either the wild-type NK-4 binding sites (E2 elements, E2CAT reporter) that were previously shown to be involved in autoregulation of NK-4 (31) or mutated NK-4 binding sites (E2m elements, E2mCAT reporter) were used for transfection. Cells were cotransfected with NK-4 in the presence (Fig. 4A, lanes 4 and 8) or absence (Fig. 4A, lanes 2 and 6) of the p300 expression vector, and CAT activities were measured. As shown in Fig. 4A, co-expression of p300 considerably increased CAT activity (lane 4) compared with expression of NK-4 alone (lane 2). However, this coactivation by p300 was not observed with the E2mCAT reporter that lacks NK-4-responsive elements (Fig. 4A, lane 8), suggesting that p300 enhanced transcriptional activation mediated by NK-4.
Next, we tested whether NK-4 can physically interact with p300. To this end, HeLa nuclear extracts were mixed with various GST-NK-4 proteins bound to glutathione-Sepharose beads, and any co-precipitated p300 was separated on SDS-PAGE and detected by Western blot analysis with an anti-p300 antibody. As shown in Fig. 4B, the p300 protein can bind to the full-length NK-4 protein (lane 3) and binds more strongly to the amino-terminal half of NK-4 (lane 4), which includes an activation domain. p300 also interacted weakly with the carboxylterminal half of NK-4 that contains the homeodomain (Fig. 4B,  lane 5).
Functional Interaction of NK-4 with the Groucho Corepressor-As described above, NK-4 can act as an active transcriptional repressor (Fig. 1) for which the homeodomain is required (Fig. 2C and Fig. 3). As is the case for other active transcriptional repressors, NK-4 may recruit corepressors for its repression activity. Groucho proteins are well-known corepressors that are required for transcriptional repression by several distinct types of active transcriptional repressors (59). Among homeodomain proteins that act as transcriptional repressors, Engrailed was recently shown to interact with Groucho (60,61). Like NK-4 and NK-3, Engrailed also contains the eh-1 domain, which is similar to the TN domain. Hence, we tested whether Groucho can act as a corepressor for NK-4.
As shown in Fig. 5A, coexpression of Groucho greatly enhanced the repressor activity of GAL4-NK-4 ( lanes 6 -8), whereas the effect of Groucho on GAL4 alone was negligible (lanes 2-4), suggesting that Groucho may act as a transcriptional corepressor for NK-4. In order to test whether NK-4 can directly interact with Groucho, we transfected cells with the Myc-tagged Groucho expression vector. Nuclear extracts from transfected cells were mixed with a GST-NK-4 protein bound to glutathione-Sepharose beads, and any co-precipitated Groucho was separated on SDS-PAGE and detected by Western blot analysis with an anti-Myc antibody. As shown in Fig. 5B, Groucho protein can bind to the full-length NK-4 protein.
Using full-length NK-4 and various in vitro translated Groucho proteins, we mapped the NK-4-binding domain of Groucho (Fig. 5C). Indeed, NK-4 can bind efficiently to the full-length Groucho protein (Fig. 5C, lane 3). Also, we found that the amino-terminal region of the Groucho protein (residues 1-192) is required for the interaction with NK-4 (Fig. 5C, lane 9). The WD repeat region does not show any interaction with NK-4 under high stringency washing conditions (Fig. 5C, lanes 13-15). Next, we further delineated the Groucho-binding domains of NK-4 using in vitro pull-down assays with a variety of in vitro translated NK-4 and GST-Groucho(1-719). As expected, Groucho can bind strongly to full-length NK-4 (Fig. 5D, lane 3). Deletion of either the RD (construct 188 -416) or the homeodomain (construct 1-267) significantly weakened the binding of NK-4 to Groucho (Fig. 5D, lanes 9 and 12). Also, whereas the construct (110 -364) containing both the RD and the homeodomain showed a strong interaction with Groucho (Fig. 5D, lane  27), constructs containing either the homeodomain or the RD showed weak interaction with Groucho (Fig. 5D, lanes 24 and  21) Thus, for a strong interaction of NK-4 with Groucho, both the RD and the homeodomain are required, which is consistent with the result that both the RD and the homeodomain are required for the strong repressor activity of NK-4 when fused to the heterologous GAL4 DNA binding domain (Fig. 2C). Inter-estingly, deletion of the TN domain (construct 46 -416) does not affect NK-4 binding to Groucho (Fig. 5D, lane 6). Taken together, the data show that Groucho directly interacts with NK-4 and enhances NK-4-mediated repression, suggesting that Groucho can act as a corepressor for NK-4. DISCUSSION NK-4 plays a key role in cardiogenesis (19), yet the regulatory mechanisms of gene transcription by NK-4 are poorly understood. In the present study, we have characterized functional domains of NK-4 that are required for transcriptional activity of NK-4. Our results indicate that NK-4 can act either as a transcriptional activator or repressor that utilizes separate functional domains. Furthermore, we provide the first evidence of NK-4 interactions with the p300 coactivator and the Groucho corepressor.
Other classes of transcription factors recruit coactivators for their transcriptional activation (58), and coactivators have been shown to form a multimeric activation complex with  5) bound to Glutathione-Sepharose beads, and the co-precipitated p300 was separated on SDS-PAGE and detected by Western blot analysis with an anti-p300 antibody. Lane 1 (HeLaNE) contained 5% of the amounts used for binding. The major band is indicated by an arrow. C, mapping the NK-4 interaction domain of p300. In vitro pull-down assays were performed with different in vitro 35 S-labeled p300 proteins and glutathione-Sepharose-bound GST-NK-4. After washing, bound proteins were resolved by SDS-PAGE and autoradiographed. Input samples contained 5% of the amount used for binding. GST, negative control with glutathione-Sepharose-bound glutathione S-transferase protein; NK-4, pull-down assays with glutathione-Sepharose-bound GST-NK-4. Numbers in parentheses indicate amino acid residues of the coding region used for expression of the proteins. D, schematic diagram of p300 showing interaction domains with NK-4 and other transcription factors. Regions of p300 used for pull-down assays with NK-4 are indicated below the diagram, and NK-4 binding is shown. ϩ, weak interaction; ϩϩ, strong interaction; Ϫ, no interaction. Numbers indicate amino acid residues. Other classes of transcription factors that showed interaction with p300 previously are indicated above the diagram. Bromo, bromodomain; C/H3, C/H3 domain. P/CAF and p300/CBP (62,63). As a potential coactivator for the NK-4 homeoprotein, we have demonstrated that p300 augments NK-4-mediated transcriptional activation in cultured cells (Fig. 4A). Consistent with these data, expression of viral protein E1A alleviated this effect, 2 and p300 physically interacted with NK-4 both in vivo and in vitro (Fig. 4, B and C). These results support the notion that p300 can act as a coactivator for homeoproteins. The interaction of p300 with NK-4 is mediated by common domains of p300, such as C/H3, which was shown to act as binding sites for many other transcription factors, including viral proteins (Fig. 4D) (58). Thus, there may be a competition between NK-4 and other factors for binding to p300. It has recently been shown that the Drosophila CBP can act as a coactivator of Ci, a transcription factor homologous to the Gli family of proteins (64), and that Drosophila CBP is necessary for Dorsal-mediated activation of the twi promoter (65). These results indicate a common role for CBP as a coactivator of different transcription factors and suggest that the Drosophila CBP/p300 may also act as a coactivator for homeoproteins during Drosophila embryogenesis. Nevertheless, it is still conceivable that homeoproteins may also recruit specific coactivators for their transactivation function, because CBP/ p300 represents a common coactivator that is required in addition to distinct coactivators for the function of different classes of transcription factors in mammals (66). Thus, it will be interesting to see whether specific coactivators of homeodo- Groucho proteins can act as corepressors for specific active repressors such as Hairy-related proteins and Runt domain proteins (59,60). For the interactions between Groucho and these class repressors, the WD repeat region of Groucho and the four-amino acid WRPW (WRPY) motif of repressors are important (67). As shown in Fig. 5, we have demonstrated that Groucho enhances the repressor activity of NK-4 and directly interacts with NK-4. These results suggest that Groucho can act as corepressor for NK-4. Furthermore, we found that Groucho can act as a corepressor for NK-3. 3 In both cases, the homeodomain is required for the strong interaction with Groucho (Fig. 5C). The WD repeat domain of Groucho appears not to be important for the interaction with NK-4 under our experimental conditions (Fig. 5D). Instead, the amino-terminal region of Groucho, which was assigned to a transcriptional repressor domain, is required for the interaction (Fig. 5D). The Engrailed homeodomain protein also interacts with Groucho (61) via the eh1 domain, which is related to the TN domain of the NK-2 class. However, the deletion of the TN domain from the NK-3 and NK-4 proteins does not affect interactions with Groucho. Thus, NK class homeodomain proteins have a unique feature in this sense, suggesting a different regulatory role for the TN domain in protein-protein interaction. We have recently shown that the NK class of homeoproteins can interact with homeodomain-interacting protein kinases (HIPKs) that can act as co-repressors of homeoproteins (47). Indeed, we found that homeodomain-interacting protein kinases (HIPKs) also interact with the Groucho corepressor. 3 We have demonstrated that NK-4 can act either as a transcriptional activator or repressor by recruiting the p300 coactivator and the Groucho corepressor, respectively. It was recently shown that Dorsal, which plays an important role for the body axis formation of Drosophila, recruits Drosophila CBP and Groucho for its transcriptional activity (65,68). Thus, our results raise one interesting question as to how NK-4 can be switched from a transcriptional activator to repressor. As shown in Fig. 6, the context of the target gene promoter could be critical for determining whether activation or repression occur. For example, Dorsal can interact with Twist for the activation of the target gene (69), whereas DSP1, an HMG-like protein, can switch a Dorsal activator to a transcriptional repressor (70). Likewise, NK-4 may interact with other cofactors, thereby acting as either an activator or repressor. As was seen in the D-mef-2 activation in cardial cells, a GATA site binding factor collaborates with NK-4 (37,42). Also, we found that NK-4 can interact with other transcription factors such as Twist and NK-3 in cultured cells. 4 Thus, it is conceivable that, depending on gene promoters that contain different cis-acting regulatory elements, NK-4 can interact with other cofactors, thereby providing different protein-binding interfaces to recruit either coactivator or corepressor complexes. Since a corepressor complex often contains histone deacetylase activity (71), it will be of great interest to see whether the NK-4 homeodomain transcription factor, which interacts with the Groucho corepressor, recruits histone deacetylases.