|Home | About | Journals | Submit | Contact Us | Français|
The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the open access version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal and Oxford University Press are attributed as the original place of publication with the correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use permissions, please contact email@example.com.
At the molecular level, members of the NKx2.2 family of transcription factors establish neural compartment boundaries by repressing the expression of homeobox genes specific for adjacent domains [Muhr et al. (2001) Cell, 104, 861–873; Weiss et al. (1998) Genes Dev., 12, 3591–3602]. The Drosophila homologue, vnd, interacts genetically with the high-mobility group protein, Dichaete, in a manner suggesting co-operative activation [Zhao and Skeath (2002) Development, 129, 1165–1174]. However, evidence for direct interactions and transcriptional activation is lacking. Here, we present molecular evidence for the interaction of Vnd and Dichaete that leads to the activation of target gene expression. Two-hybrid interaction assays indicate that Dichaete binds the Vnd homeodomain, and additional Vnd sequences stabilize this interaction. In addition, Vnd has two activation domains that are typically masked in the intact protein. Whether vnd can activate or repress transcription is context-dependent. Full-length Vnd, when expressed as a Gal4 fusion protein, acts as a repressor containing multiple repression domains. A divergent domain in the N-terminus, not found in vertebrate Vnd-like proteins, causes the strongest repression. The co-repressor, Groucho, enhances Vnd repression, and these two proteins physically interact. The data presented indicate that the activation and repression domains of Vnd are complex, and whether Vnd functions as a transcriptional repressor or activator depends on both intra- and inter-molecular interactions.
Homeobox genes play essential instructional roles in many developmental processes, including patterning the early embryo and specifying embryonic cell lineages. The NK-type homeobox genes encode transcription factors with a divergent homeodomain and two additional highly conserved domains, the candidate Groucho interaction domain referred to as the NK decapeptide or Engrailed homology (EH) domain, and the NK-2 box or the NK-2 specific domain (1–3). The Drosophila NK-2 homeobox gene, vnd, specifies ventral neuroblast identity (4,5). This dorsal–ventral patterning gene represses the expression of the homeobox gene, intermediate neuroblast identity (ind), which is expressed in neuroectodermal cells adjacent to the bilateral columns of vnd expressing ventral cells. Three vnd binding sites in the ind enhancer mediate this repression (6).
The question of whether Vnd can directly activate transcription has not been addressed, despite the fact that various potential target genes are not activated in vnd mutants. These include the proneural genes, achaete and scute, the homeobox gene, Nk6 and the zinc finger transcription factor, Escargot. vnd also autoactivates its own expression (7–9). In addition, epistasis tests and double mutant analyses suggest that Vnd and the high-mobility group (HMG) proteins, Dichaete and Sox Neuro, co-operate to regulate patterning and cell fate decisions in the Drosophila neuroectoderm (10,11). NKx2.2, the closest vertebrate homologue of vnd, is also required for dorsal–ventral patterning of the ventral neural tube, as well as pancreatic specification (12–14). Within the neural tube, NKx2.2 interacts with the co-repressor, Groucho, to repress inappropriate expression of target genes (1). Groucho lacks a DNA-binding domain (DBD), but regulates transcription by interacting with DNA-binding transcription factors to form a repression complex [for review see (15)].
Vnd is a 723 amino acids protein (3,16) that is significantly larger than its vertebrate counterparts, such as NKx2.2. This Drosophila protein is likely to be complex, as it may perform functions that are executed by the multiple vertebrate NKx2 family members. In this report, we dissected the domains of the Vnd protein using a heterologous system in transient transfection assays. If Vnd regulatory domains can function in a heterologous context with a DBD from another transcription factor, that domain contains sufficient sequence information to execute a specific regulatory function. These domain swap experiments identified multiple repression domains in Vnd. We found that the conserved homeodomain and NK-2 box have significant repressor activity when expressed in conjunction with other Vnd sequences. We also uncovered two activation domains that are masked by the repressor activity of the intact Vnd protein in cell culture. In immunoprecipitation and mammalian two-hybrid interaction assays, we found that Vnd physically interacts with the high-mobility Sox activator, Dichaete, through the homeodomain. While Vnd had little effect on its own enhancer, co-expression with Dichaete enhances its activation capacity co-operatively in transient transfections. Likewise, under similar assay conditions Vnd requires the co-repressor, Groucho, to repress the ind enhancer, through which Vnd represses ind expression in embryos. These data are the first biochemical evidence for Vnd-dependent activation, and suggest that Vnd-mediated activation or repression depends to a large extent on the availability of co-factors.
A 2.44 kb fragment corresponding to the Vnd open reading frame (ORF), plus 10 bp 5′ and 257 bp 3′ of the ORF, was cloned into the EcoR1 site of the pCMV-flag vector (Sigma) for expression in mammalian cells. CMV-flag-groucho 4 (CMV-grg4) was kindly provided by Greg Dressler (Department of Pathology, University of Michigan, Ann Arbor, MI). A 2.5 kb EcoR1–BamH1 fragment corresponding to one of the vnd enhancers was isolated from a P1 library and subcloned into the Kpn1–BglII site of the pGL3 promoter vector (Promega), following the addition of linkers. A 600 bp fragment corresponding to a dimer of the ind enhancer element, to which Vnd binds (6), was also cloned into pGL3. The pFox Luc1-7xNK2 vector that contains seven NKx2.2 consensus-binding sites upstream of the luciferase gene was provided by Michael German (Department of Medicine, University of California at San Fransisco, CA). A 1 kb Pvu11–XbaI fragment, lacking sequences encoding the first 30 amino acids of Dichaete, was cloned into the Not1 site of pACT (Promega) and of pCMV-flag, following fill in with Klenow. John Nambu (Biology Department, University of Massachusetts at Amerherst Amerherst, MA) kindly gave us the Dichaete cDNA. The pACT vector encodes the VP16 activation domain.
For the generation of Gal4–Vnd fusion constructs, relevant vnd cDNA fragments were subcloned into pBind vector (Promega), which encodes the Gal4 DBD (amino acids 1–147) driven by the cytomegalovirus (CMV) promoter and enhancer, using appropriate restriction enzymes and Klenow fragments (see Table 1). pBind also encodes Renilla luciferase under the control of the SV40 early enhancer and promoter. To express the NK decapeptide and the NK-2 box as Gal4 chimeras, complimentary oligonucleotides were used. The sequence of the positive-strand oligonucleotide for the NK decapeptide is: 5′-GGCTTCCATATATCGGACATCTTGAATTTGGAGGGCTCTGAGCTGAAGAATGCAGCAGC-3′. The sequence of the NK-2 box oligonucleotide is: 5′-CATCGCCCCGTCGGGTAGCCGTTCCAGTTCTGGTGAGGAACGGAAAGCCCTGCTTGGGCGATAGTTCCAAACTGGGAGCCG-3′. Complimentary oligonucleotides were kinase-treated, denatured, annealed and cloned into pBind. Each construct was sequenced using vector-specific primers.
293 cells were plated at 500000 cells/ml into 6-well plates, cultured in DMEM + 10% fetal calf serum (FCS) containing 1% penicillin/streptomycin at 37°C, and transfected the following day with Fugene-6 (Roche), according to the manufacturer's instructions. The volume of Fugene-6 used was twice that of the DNA. pBind (Promega) encodes Renilla luciferase, which was used to monitor transfection efficiency. The pG5luc reporter (Promega) encodes firefly luciferase downstream of five Gal4-binding sites. For the chimeric one-hybrid assays, 1.5 μg of DNA was transfected and the molar ratio of the expression vector to the reporter vector was 2:1. For the co-transfection of pCMV-flag-groucho 4 andpCMV-flag-dichaete with the pBind-vnd and pCMV-vnd constructs, the ratio of pBind to pG5luc was 2:1. pGem-3Z was added to maintain the amount of DNA transfected constant. For the two-hybrid assay, the molar ratio of the pACT to pBind to pG5Luc constructs was 1:1:1, and 2.5 μg of DNA was used per transfection. The Checkmate Two Hybrid kit (Promega) was used according to the manufacturer's instructions.
Forty-eight hours after transfection, cells were washed with phosphate-buffered saline (PBS), scraped into passive lysis buffer (Promega) and rocked for 15 min gently at room temperature (RT). Debris was pelleted and 20 μl of lysate was assayed for Renilla and firefly luciferase activity using the Dual Luciferase Kit (Promega), according to the manufacturer's instructions. Note that the pBind vector encodes the Renilla luciferase, while the pG5luc vector encodes the firefly luciferase. Each transfection was performed at least three times. Drosophila S2 cells were maintained at RT in M3 media containing 10% FCS and 1% penicillin/streptomycin, and were otherwise handled similar to the 293 cells.
4.5 μg of pBind-vnd or empty pBind, 1.5 μg of pCMV-flag-grg4 or pCMV-flag-dichaete or empty vector, and 1.6 μg of pGL3-vnd or pGL3-ind were transfected into 100 mm dishes containing 293 cells. Forty-eight hours after transfection, cells were washed with PBS, scraped and the supernatant removed following centrifugation. Cells were resuspended in immunoprecipitation (IP) buffer [20 mM Tris–HCl, 100 mM NaCl, 10 mM NaF, 1 mM Na3VPO4, 1 mM phenylmethylsulfonyl fluoride (PMSF) and protease inhibitor cocktail (Roche) containing 0.5% Triton X-100] and briefly vortexed. Non-specific binding proteins were removed by pre-incubating the supernatant with protein A/G PLUS agarose (Santa Cruz Biotechnology) for 30 min at 4°C, and then removing them by centrifugation. Cell lysates were then incubated with M2 anti-Flag antibody (Sigma), and rotated overnight at 4°C, following incubation with protein A/G PLUS agarose, and rotation at 4°C for 2 h. Beads were precipitated by centrifugation, and washed three times with IP buffer containing 0.1% Triton X-100. Then, beads were resuspended in SDS–PAGE sample buffer. Immunoprecipitates were separated by SDS–PAGE electrophoresis, transferred onto Immobilon-P (Millipore) membrane, and western blotted. Duplicate blots were incubated with anti-Flag antibody (Sigma) to detect the Flag-tagged Groucho 4 and Dichaete proteins, or a Gal4-specific antibody (Santa Cruz Biotechnology) to detect Gal4–Vnd chimeric protein. Binding of peroxidase-conjugated secondary antibodies was detected by chemiluminescence, using the Lightning kit (Perkin Elmer).
Asp1–Bgl1, Asp1–EcoRV and Kpn1–Bgl1 vnd cDNA fragments (Table 1) were PCR-amplified using a 5′ primer with an EcoR1 site added, and a 3′ primer with a Sal1 site added, and cloned into the bait vector, pGBKT7 (Clontech). PCR products encoding the ORFs of Dichaete and fly groucho, containing synthetic 5′ EcoR1 and 3′ Xho1 sites, were also cloned into the pGADT7 prey vector (Clontech). Sequencing was used to confirm the identity of the PCR products, and that the individual fragments had been cloned in frame. Using the T7 promoter, located upstream of the myc tag in pGBKT7, and the hemagglutinin (HA) tag in pGADT7, individual prey and bait proteins were in vitro transcribed and translated using the TNT T7 Coupled Reticulocyte Lysate kit (Promega) with 35S-labeled methionine, according to the manufacturer's instructions. Immunoprecipitations were performed using the HA antibody from the Matchmaker Co-Immunoprecipatation kit (Clontech), according to the manufacturer's instructions. Following separation by SDS–PAGE, labeled proteins in the in vitro transcription–translation reactions and the co-immunoprecipitations were identified using autoradiography.
Transient expression of Vnd in Drosophila S2 or vertebrate 293 cells had minimal effects on reporter gene activation or repression, using three different candidate targets, despite strong Vnd expression in both cell types. The candidate target sequences tested were as follows: the vnd enhancer that directs vnd-specific expression from stage 6 onwards (17), the repression element previously identified in the ind enhancer (6) and a concatenate of seven binding sites corresponding to the consensus NKx2.2 recognition motif (18) (data not shown). Watada et al. (18) similarly found that intact Nkx2.2 had minimal effects on reporter expression driven by seven consensus Nkx2.2 binding sites. To facilitate the identification of the functionally important domains in Vnd, we used a one-hybrid assay in which Vnd was fused to the DBD of Gal4. Using this assay, we were able to delineate different Vnd domains and test their efficacy on reporter gene activation or repression using a Gal4-specific reporter construct (Figure 1).
Full-length Vnd fused to the Gal4 DBD represses firefly luciferase reporter expression 7- to 8-fold (construct A1, Figure 1C), consistent with its potential role as a repressor in the developing nervous system. To delineate the repression domain more precisely, we made C- and N-terminal deletions of Vnd and fused these to the Gal4-DBD. Deletion of 100 amino acids at the C-terminal of Vnd, including the conserved NK-2 box (construct A2, Figure 1C), leads to a slight, but reproducible, 10% increase in repression activity, relative to A1. Further deletions encompassing the homeodomain (construct A3, Figure 1C) cause a 75% decrease in repressor activity relative to full-length Vnd. However, repressor activity is restored to the level of full-length Vnd if additional sequences are deleted to amino acid 217 (constructs A4 and A5, Figure 1C). These results suggest that sequences spanning amino acids 217–406 likely have de-repression or activator function. The first 217 amino acids of Vnd (construct A5, Figure 1C) include the NK decapeptide or the EH domain (1,2). Surprisingly, when the NK decapeptide is removed (construct A6, Figure 1C), the repression activity of the chimera is not greatly affected. Thus, C-terminal deletions of Gal4–Vnd chimeras uncover multiple redundant repression domains.
To confirm and extend our observations, we also analyzed N-terminal deletions (Figure 1D). Deletion of amino acids 3–217 significantly reduces repressor activity (construct A7, Figure 1D), consistent with the activity of constructs A5 and A6. Further N-terminal deletion to amino acid 519 increases repressor activity (constructs A8 and A9, Figure 1D), although not quite to the full-length Vnd levels. Again, these results are consistent with the homeodomain having repressor function. The C-terminal 60 amino acids are able to activate reporter gene expression, consistent with the slight increase in repression observed upon deletion of this sequence in A2 (Figure 1C). Taken together, the C- and N-terminal deletions indicate that the sequences spanning the homeodomain and NK-2 box, and the sequences spanning 3–217, are able to independently repress activation. However, the region spanning 217–519 can function to de-repress these domains individually. In the presence of both repressor domains, the effect of the depression region is minimal.
To assess the potential of the candidate regulatory domains to function independently, we assayed the activity of constructs encoding specific internal Vnd regions that were delineated by the N- and C-terminal deletions. Surprisingly, we found that the capacity of the EH domain, the homeodomain and the NK-2 box to repress transcription is minimal, when expressed independent of other Vnd sequences (Figure 2, constructs A12, A16–A19, respectively). Assuming that steric hindrance does not account for these observations, these results suggest that these domains interact with one another, or other Vnd sequences, in the intact protein to repress transcription. More significantly, a strong activation domain was found that maps onto the amino acids 217–406 (construct A14, Figure 2). Again, this is consistent with the de-repression activity observed in the previous deletions. No difference in luciferase levels was seen when Vnd was expressed independent of the Gal4-DBD with the empty pBind vector, which indicates that the Vnd-dependent effects seen are not due to Vnd activity in a target DNA-independent manner. Thus, Vnd has two activation domains, spanning 217–406 and 657–717, which are typically masked in the intact protein, suggesting that Vnd can function as an activator under the appropriate circumstances. For each of the constructs generated in Figures 1 and and2,2, western blotting with an antibody against the Gal4-DBD generated the expected sized protein at levels consistent with the efficiency of transfection (data not shown). Furthermore, all of the Vnd chimeric proteins were found in the nucleus of transfected cells when the same antibody was used for immunostaining (data not shown).
For transcriptional activation, co-activators are typically required in addition to the general transcription factors needed for basal transcription. Recent genetic analyses suggest that the HMG proteins, Dichaete and Sox neuro interact with Vnd (10,11). Zhao and Skeath (12) found that the frequency of loss of marker expression in two ventral neuroblasts was increased in vnd, Dichaete double mutants relative to that seen in either vnd or Dichaete single mutants. Because HMG proteins bend DNA to align non-contiguous DBDs for interaction with other transcription regulators and physically interact with other regulatory proteins to activate transcription (19–23), we asked whether Vnd physically interacts with Dichaete by doing a two-hybrid interaction assay in mammalian cells using the Checkmate Kit (Promega). We introduced a chimeric plasmid, pACT-Dichaete, which encodes the VP16 activation domain in frame with the Dichaete ORF into the one-hybrid assay. We found that when increasing concentrations of VP16–Dichaete are co-expressed with Gal4–Vnd, the repression of baseline luciferase activity caused by the full-length Vnd–Gal4 chimera was significantly reduced. When Dichaete was expressed at a 4-fold molar excess relative to Vnd, repression was reduced ~6-fold (Figure 3, A1).
Figure 3 shows the results of mapping the region of Vnd that interact with Dichaete, using our series of Gal4–vnd deletion constructs. There was a 3- to 3.5-fold alleviation of repression when a Gal4–Vnd chimera encoding the N-terminal of Vnd, including the homeodomain (Figure 3, amino acids 3–613, A2), was co-expressed with VP16–Dichaete, relative to empty pBind plus pACT-Dichaete. Excluding the homeodomain in the N-terminal construct (Figure 3, amino acids 3–519, A3) leads to only very slight alleviation of repression, which suggests that the Vnd homeodomain is important for Dichaete interaction; whereas, expression of the C-terminal amino acids 519–717, including the homeodomain (Figure 3, A9), resulted in an 8- to 9-fold alleviation of repression. This indicates that Vnd sequences in this region must interact with Dichaete strongly. To further define the Dichaete interaction domain in the C-terminal region of vnd, we expressed the Vnd homeodomain and the downstream region, independent of each other, as Gal4 chimeras. When expressed in conjunction with VP16–Dichaete, neither peptide activated luciferase activity significantly, compared to when the domains were expressed together (Figure 3, A10 and A16). Although we cannot rule out steric hindrance accounting for the inability of the smaller Vnd sequences to interact with Dichaete, these results suggest that the homeodomain is necessary, but not sufficient, for binding of Vnd to Dichaete. The C-terminal sequences downstream of the homeodomain may facilitate Dichaete binding. In agreement with these observations, the homeodomains of Bicoid, and Oct-type homeodmain proteins has been implicated in the interaction of these transcription factors with other HMG proteins (24–26).
Dichaete interacts with the CNS-midline-specific proteins, Single-minded and Tango, to activate ventral midline Slit gene expression (23). A 2.5 kb region of the vnd enhancer directs ventral cord expression (17,27). We examined this region for the presence of consensus Dichaete-binding sites [AACAAT and AACAAAG (22)] and Vnd-binding sites [GTGAACT (4)]. Two non-contiguous consensus Dichaete-binding sites are present at positions 350 and 1350, independent of three candidate Vnd-binding sites that have six out of seven of the consensus nucleotides at positions 50, 700 and 1020, respectively. We hypothesized that if Dichaete interacts with Vnd physically, this could cause bending of the vnd enhancer, so that the non-adjacent binding sites for Vnd and Dichaete are juxtaposed. This modulation of target DNA availability, combined with the likely conformational changes in Vnd resulting from its interaction with Dichaete, could potentially lead to the activation of reporter expression.
When we addressed this possibility initially in S2 cells, we found that although Vnd is strongly expressed in these cells under the control of the actin promoter, the effects on reporter expression driven by the vnd enhancer are negligible. Dichaete, when expressed alone caused weak activation of reporter expression, whereas co-expression of Dichaete and Vnd lead to enhancer activation at levels greater than when Dichaete was expressed on its own (data not shown). We repeated the experiment in 293 cells, using the CMV promoter to drive protein expression (Figure 4). With increasing concentrations of pCMV-Dichaete, luciferase expression driven by the vnd enhancer was increased to a maximum level of 150% [when the baseline was set at 0 (Figure 4, B)], whereas increasing the concentration of pCMV-vnd lead to a slight reduction in luciferase expression (Figure 4, A). However, when Vnd and Dichaete were co-expressed, reporter expression was activated to a maximum level of 250% (Figure 4, C), relative to the baseline level of reporter expression. Because the sum of the effects of Vnd and Dichaete expressed together is significantly greater than when each transcription factor is expressed independently, our results suggest that these two regulators co-operate with the vnd enhancer to activate reporter expression. The capacity of Vnd to activate is target gene dependent. Dichaete did not affect Vnd activity when co-expressed in cells in the presence of a reporter driven by the ind enhancer, through which Vnd mediates repression in the embryo (6).
We also asked whether exogenous Groucho affects the capacity of intact Vnd to repress luciferase reporter expression, because Vnd has an NK decapeptide or EH domain. In addition, there are AT-rich domains in close proximity to the binding sites for Vnd in the ind enhancer (6). Groucho repression complexes typically bind these AT-rich domains when appropriately positioned (28). When we used the ind enhancer to direct luciferase reporter expression, luciferase activity was only very slightly repressed when Vnd was expressed alone. The repression was variable and the maximum level seen was a bare 15% decrease in luciferase activity (Figure 4, D). Likewise, when Grg4 was expressed alone, it had minimal effects on luciferase activity (Figure 4, E). However, when Vnd and Grg4 were co-expressed, repression of luciferase activity was significantly enhanced, and there was less variability (Figure 4, F). Again, Vnd's capacity to co-operate with Groucho in repressing gene expression was target-dependent, and required the ind enhancer. Co-expression of Vnd and Groucho in the presence of the vnd enhancer had no effect (data not shown). These combined results indicate that both co-factor availability and target DNA accessibility influence Vnd's capacity to activate and repress transcription.
Our data suggest that Vnd interacts directly with the co-activator, Dichaete, and the co-repressor, Groucho, and this interaction leads to either target activation or repression. To confirm this, we performed co-precipitation analyses. Because our Vnd antibody was relatively ineffective for western-blot analyses, we confirmed that the expression of Vnd fused to the Gal4-DBD with Flag-tagged Groucho or Dichaete resulted in enhanced repression and activation of vnd targets, similar to intact Vnd without an additional tag (see Figure 4; data not shown). When we immunoprecipitated Flag-tagged Dichaete or Grg4 from cell lysates using a Flag antibody, Gal4-tagged Vnd was pulled down with both co-regulators. These results indicate that Dichaete and Grg4 physically interact with Vnd (Figure 5).
We further examined the physical interactions of Groucho and Dichaete with Vnd by addressing whether these co-factors interact in a yeast two-hybrid assay using the Matchmaker Gal 4 Two Hybrid System 3 kit (Clontech). We cloned full-length vnd into the bait vector, pGBKT7, as well as two smaller vnd domains. Because of concerns about steric interference that arose in the Vnd–Gal4 chimeric assays, we selected the A3 and A8 Vnd constructs for the yeast assays. A3 lacks the homeodomain and its C-terminal sequences, but includes the EH domain, whereas A8 extends from amino acid 406, and includes the homeodomain and Nk-2 box. Full-length fly groucho and Dichaete were also cloned into the pGADT7 prey vector. AH109 cells were transformed with the various constructs. Because all the three bait proteins drove alpha gal expression without an interacting protein, we monitored prey–bait interactions by measuring growth on HIS− medium. We found that both Dichaete and Groucho interacted with full-length Vnd very poorly in this assay. When the co-transformants were plated on HIS−, LEU−, TRP− medium with concentrations of 3-AT ranging from 0 to 15 mM, it took 5 days at 30°C for small colonies to grow, relative to unrelated controls that grew well after 3 days. Transformants encoding Groucho and the A3 Vnd deletion mutant failed to activate HIS− expression, while transformants encoding the A8 Vnd deletion mutant and Dichaete grew well on HIS−, LEU−, TRY− medium containing 10 mM 3-AT after 3 days at 30°C (data not shown).
Since these observations question our finding that Vnd can co-precipitate with both Dichaete and Grg 4, we tested whether the bait and prey proteins interact in co-precipitations assays. Using the Matchmaker Co-IP kit (Clontech) and the TNT T7 Coupled Reticulocyte Lysate System (Promega), we made 35S-labeled A1, A3 and A8 Vnd from the bait plasmids in vitro, with N-terminal Myc-tags (but lack the Gal4-DBD; Figure 6, lanes 1–3). Similarly, we made 35S-labeled Groucho and Dichaete from the T7 prey plasmids, which have a HA tag (but lack the Gal4 activation domain; Figure 6, lanes 4 and 5). The HA antibody proved more effective than the Myc antibody in immunoprecipitations (data not shown). Figure 6 (lanes 6–9) shows the results of pull downs with the HA antibody. Confirming our results from the co-immunoprecipitations from tissue culture cells, we found that either HA-tagged Dichaete or Groucho can pull down full-length Vnd (Figure 6, lanes 6 and 8, respectively). Vnd lacking the first 408 amino acids (A8) interacted with Dichaete (Figure 6, lane 7), but the amount precipitated was less than when full-length Vnd was co-precipitated (Figure 6, lane 6). We were surprised to find that the Vnd deletion mutant, A3, which contains the EH domain but lacks N-terminal sequences including the homeodomain and the NK-2 box, failed to interact with Groucho (Figure 6, lane 9) in three independent experiments. These results indicate that sequences in the C-terminal region of Vnd are required for Groucho binding to the EH domain, and potentially explain why we were unable to assign repressor function to the EH domain in the Gal4–Vnd chimera dissection assays. In agreement with these findings, Winnier et al. (29) reported that in Caenorhabditis elegans mutant Unc-4 failed to interact with Unc-37, the worm Groucho, in two-hybrid interaction assays, and did not exhibit in vivo repressor activity when the EH1 domain was intact, but sequences N-terminal to the domain are missing.
Our dissection of the Vnd homeodomain protein in cell culture suggests that this transcription factor is a complex combination of repression and activation domains. Both intra- and inter-molecular interactions are apparently involved in the DNA binding, activation and repression activity of Vnd. Chimeric analyses of Vnd subdomains indicate that there are at least three repression domains and two activation domains that can contribute to the regulatory capacity of this protein. These activation and repression domains likely interact in the intact protein to generate different patterns of regulation. Our analyses indicate that the co-repressor, Groucho, promotes the repressional activity of Vnd, whereas interaction with the transcriptional activator and DNA bending protein, Dichaete, leads to target activation. Figure 7 summarizes our current model for Vnd regulation of target gene expression that is based on the data presented.
The Vnd homeodomain plays a complex role in the function of this protein. Although capable of binding its target in vitro (3,6,30), the homeodomain apparently does so very inefficiently within the context of the native protein in transient transfection assays. Watada et al. (18) showed that removal of the C-terminal of NKx2.2 downstream of the homeodomain leads to NKx2.2-activating reporter expression in tissue culture, although the intact protein was unable to do so. Co-factors are potentially required to stabilize the binding of the homeodomain to its target in the context of the complete protein. NMR analyses indicate that the conformation of the Vnd homeodomain is altered upon binding to target DNA (30). This potentially results in the protein surfaces available for contact with the transcriptional machinery, and with other regulatory proteins being altered. These intra-molecular interactions would then provide a mechanism for transducing information from the DBD to the other regulatory domains within the protein. Our deletion analyses of Vnd–Gal4 chimeras in transient transfection assays suggest that the homeodomain interacts with repression domains at the N-terminal of the protein. In keeping with this speculation, we found that Groucho did not interact with a Vnd deletion, lacking the N-terminal 200 amino acids including the homeodomain and the NK-2 box, despite the presence of the EH domain (Figure 6). Paradoxically, the homeodomain and sequences on its C-terminal side are also required for the interaction of Vnd with the transcription activator and DNA bending protein, Dichaete. Likewise, Bicoid and Oct 2A have been shown to interact with HMG proteins through their homeodomains (24–26). Taken together, these results suggest that conformational changes in the native protein, including local protein folding, play important roles in the capacity of Vnd to regulate gene expression.
Extending the findings of Xhao and Skeath (11), we show that Vnd's interaction with the high-mobility Sox protein, Dichaete, confers weak activation capacity on this homeodomain protein in transient transfections. Although Dichaete co-operated with Vnd to activate reporter expression driven by a vnd enhancer containing both Dichaete and Vnd binding sites, no effects were seen when both proteins were expressed in cells with a reporter driven by the ind enhancer (data not shown). Thus, Vnd's capacity to activate is dependent on both the availability of co-factors and the DNA target. In the context of the vnd enhancer, Vnd's interaction with Dichaete changes both the context of these proteins and also that of the target DNA. As an HMG protein, Dichaete's transcriptional co-activator ability is associated with its capacity to bend DNA to align non-contiguous sites for interaction with other transcription regulators [(20–22) and Figure 7]. The capacity of Dichaete to modulate target DNA architecture so that the contact DNA is altered potentially generates a new dimension to NKx2.2/Vnd regulation. To address the possibility that NKx2.2 regulates neural tube patterning by activating transcription, Muhr et al. (1) over-expressed the NKx2.2 homeodomain in frame with the VP16 activation domain in the chick neural tube. This substitution did not affect patterning of the neural tube, whereas over-expression of the NKx2.2 homeodomain fused to the Engrailed repression domain did. Consequently, Muhr et al. (1) concluded that NKx2.2's capacity to activate gene expression was not important in ventral neural tube patterning. However, these experiments modified the NKx2.2 protein but not its target DNA. Sox 1, 2 and 3, the vertebrate homologues of Dichaete, are expressed in overlapping patterns in the developing neural tube and Sox 2 function is necessary, but not sufficient, to direct cells to a neural fate (31,32). If Nkx2.2 interacts with one, or all, of these proteins in the neural tube, this interaction could potentially confer transcriptional activation capacity on NKx2.2, by both reorganizing the target DNA and directly interacting with this transcription factor. This possibility has not yet been addressed. Interestingly, Watada et al. (17) also identified an activation domain at the C-terminal end of NKx2.2 that could be used in the presence of the appropriate transcriptional co-activators to activate target gene expression.
The intrinsic activity of Vnd's activation domains is suppressed in transient transfection assays in the context of the full-length protein, or when full-length Vnd is fused to Gal4. Both activation domains that we identified are predicted to have subdomains that are helical and enriched in homopolymeric alanine residues interspersed with bulky amino acids. The structure of the activation domain is as important as the amino acid sequence to its function (33,34). Further work will be required to define the limits of these activation domains and the critical amino acids involved.
We were surprised to find that a domain with no homology to the published regulatory domains at the N-terminus of Vnd had the strongest repression activity in our domain swap experiments. This region is moderately conserved between Drosophila virilis, Drosophila pseudoobscura and Drosophila melanogaster (Z. Yu and D.M. Mellerick, unpublished data). However, the amino acid composition is not striking. One possibility is that the N-terminal extension of Vnd, lacking in vertebrate Vnd-type proteins, reflects greater functional complexity relative to the vertebrate counterparts. The alternative possibility that this domain is not functionally significant in the developing Drosophila embryo cannot be excluded at this time. Although, full-length Vnd co-precipitates with Groucho, a Vnd deletion lacking the N-terminal 200 amino acids, including the homeodomain and the Nk-2 box, failed to do so (Figures 5 and and6).6). Potentially, this explains why we were unable to assign repressor activity to the EH domain in the Gal4–Vnd chimera dissection analyses (Figures 1 and and22).
It should be kept in mind that how a transcription factor functions in a transient transfection assay does not always fully reflect its function in the context of the developing embryo, where the availability of co-activators, repressors and targets may be different. Tolkunova et al. (35) previously found that Engrailed uses two mechanisms to repress transcription—one that is predominant under normal transient transfection assay conditions, the other, which is predominant in an in vivo repression assay. The EH domain had only weak activity in transient transfection assays, despite the fact that this Engrailed domain mediates the in vivo repression activity and interacts specifically with Groucho, whereas two additional repression domains were more potent in transient transfections than in vivo. Koizumi et al. (36) recently found that over-expression of mutant Vnd lacking either the NK-2 box or the NK domain in transgenic embryos did not affect the early repression activity of the vnd transgene. Since both these domains are highly conserved, and associated with the repressor activity of the protein, these results were unexpected. One possible explanation for Koizumi's result is that deleting repression domains does not alter the activation capacity of Vnd. The mutant transgenes may be able to autoactivate expression of the endogenous vnd gene, resulting in its over-expression. Consequently, the effects of the mutations would be hidden by the over-expression of the endogenous gene. Determining whether the mutant vnd transgenes can rescue loss-of-function, vnd mutants will address this possibility. Thus, both tissue culture and transgenic vnd over-expression experiments have limitations in terms of their capacity to assign a specific function to an individual domain. Both these approaches highlight the context-dependence of Vnd activity.
We thank Heather Hand for making the pActPPA-vnd construct, Vicki Modica for making the pGL3-ind enhancer construct, Jay Uhler for making the pGL3-vnd enhancer construct and Huanqing Zhang for help with the immunoprecipitations. We appreciate the generosity of Greg Dressler, Michael German, John Nambu and Susan Parkhurst for providing us with their constructs. We are also grateful to Jay Uhler for her critical comments on this manuscript. This work was supported by RO1 HD0348403 to D.M.D. from NIH. Funding to pay the Open Access publication charges for this article was provided by RO1 HD0348403.