Groucho Oligomerization Is Required for Repression In Vivo

Abstract

Drosophila Groucho (Gro) is a member of a family of metazoan corepressors with widespread roles in development. Previous studies indicated that a conserved domain in Gro, termed the Q domain, was required for repression in cultured cells and mediated homotetramerization. Evidence presented here suggests that the Q domain contains two coiled-coil motifs required for oligomerization and repression in vivo. Mutagenesis of the putative hydrophobic faces of these motifs, but not of the hydrophilic faces, prevents the formation of both tetramers and higher order oligomers. Mutagenesis of the hydrophobic faces of both coiled-coil motifs in the context of a Gal4-Gro fusion protein prevents repression of a Gal4-responsive reporter in S2 cells, while mutagenesis of a single motif weakens repression. The finding that the repression directed by the single mutants depends on endogenous wild-type Gro further supports the idea that oligomerization plays a role in repression. Overexpression in the fly of forms of Gro able to oligomerize, but not of a form of Gro unable to oligomerize, results in developmental defects and ectopic repression of Gro target genes in the wing disk. Although the function of several corepressors is suspected to involve oligomerization, these studies represent one of the first direct links between corepressor oligomerization and repression in vivo.

Groucho family corepressors play multiple roles in metazoan development. The archetypal member of this family, Drosophila Groucho (Gro), is required for almost every aspect of embryonic and imaginal development, including sex determination, neurogenesis, segmentation (54), dorsoventral patterning (19), terminal patterning (25, 55), eye development (21), and wing development (3). The vertebrate members of the Groucho family, the transducin-like Enhancer of Split (TLE) proteins, play similarly diverse roles in development, including roles in embryonic patterning and neurogenesis (11, 22). In addition, these factors have been implicated in human disease. For example, Gro/TLE factors may serve to suppress colorectal cancers that result from defects in Wnt signaling (57). The widespread role of Groucho family proteins in development and disease is due to their requirement for transcriptional repression by a myriad of DNA-bound regulatory factors. Repressors that require Groucho family corepressors include Hairy family proteins (27, 54), Runt family proteins (5), Engrailed family proteins (33), Brinker (31, 67), Huckebein (25), Even skipped (35), TCF/Lef family factors (10, 39), and Dorsal (19, 24).

Groucho family proteins are characterized by a conserved N-terminal glutamine-rich region (the Q domain) and a conserved C-terminal WD-repeat domain (11). The WD-repeat domain is required for binding to a variety of proteins involved in repression (33, 64, 68). The Q domain has also been implicated in several protein-protein interactions (10, 15, 20, 45, 59), but, perhaps most significantly, this domain is required for the homotetramerization of Groucho (13, 56). The amino-terminal enhancer of split (AES) proteins, which constitute a Groucho subfamily, contain the Q domain but lack the WD-repeat domain, and a number of studies suggest a dominant negative role for AES proteins in Gro-mediated processes (59, 60). This dominant negative function may result from the formation of mixed oligomers between AES proteins and full-length members of the Gro/TLE family, thereby sequestering these factors in an inactive form. Thus, the dominant negative function of AES proteins is consistent with the idea that the Q domain mediates oligomerization in vivo.

Between the highly conserved Q and WD-repeat domains of Groucho family proteins are three less well conserved domains, the GP domain, the CcN domain, and the SP domain. The GP domain may be important for interactions with the histone deacetylase HDAC1 (10, 12, 15), while the entire N-terminal region of the protein, including everything other than the WD-repeat domain, may be important for binding to hypoacetylated forms of the core histone N-terminal tails (23, 53).

Most of the available evidence suggests that Gro mediates long-range repression (16, 22, 43). In other words, this corepressor is able to silence a locus regardless of where it is tethered to the template relative to DNA-bound activators or the core promoter. This contrasts with dCtBP, another Drosophila corepressor, which represses transcription only when bound within about 100 bp of an activator or the core promoter (43, 49, 51). Long-range repression by Gro could relate to its ability to both oligomerize and interact with core histones, thus allowing the corepressor to polymerize along the template and establish a large transcriptionally silent domain. Recruitment of HDAC1 by Gro could serve to reenforce the silent state since histone deacetylation increases Gro binding affinity. It is interesting that other regulatory factors that establish large silent domains, in particular those involved in heterochromatic silencing, have similarly been shown to both oligomerize and bind histones (16, 46).

Although corepressor oligomerization is implicated in many instances of repression, in no case has a definitive link been made between the ability of a factor to oligomerize and its ability to repress transcription in vivo. In the experiments presented here, we show that Gro oligomerization is very probably mediated by two coiled-coil motifs in the Q domain. Furthermore, we show that mutations that disrupt the hydrophobic faces of these coiled coils prevent both tetramerization and the formation of higher-order Gro oligomers. Finally, we show that the same point mutations that disrupt oligomerization prevent Gro-mediated repression in transfected cells and in vivo, strongly suggesting that oligomerization is a prerequisite for repression.

MATERIALS AND METHODS

In vitro translation and coimmobolization assays.

pET17b-Gro (13) was used to generate wild-type full-length Gro. Wild-type and point mutant pET3c-6HGro(2-194) (13) were used to generate His-tagged N-terminal Gro. For cotranslation, constructs encoding two forms of Gro were mixed before being added to the TNT T7 quick coupled transcription-translation system (Promega). The translation products were diluted into binding buffer (25 mM HEPES [pH 7.6], 0.45 M NaCl, 10 mM imidazole, 0.1% Tween 20, 1 mM dithiothreitol) and incubated with Ni-nitrilotriacetate (NTA) beads. The beads were then extensively washed with binding buffer. Bound proteins were eluted and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and autoradiography.

Protein preparation and gel filtration.

FLAG-tagged Gro mutant proteins were expressed in stably transfected S2 cells and purified on anti-FLAG antibody conjugated beads (Sigma). Briefly, a pRM expression construct (9) encoding FLAG-tagged Gro and a plasmid encoding the selectable hygromycin resistance gene (HPH) marker were cotransfected into S2 cells. After selection with hygromycin B, the cells were amplified and Gro expression was induced by the addition of 0.5 mM CuSO₄ to the cell culture medium. Gro mutant proteins were purified from nuclear extracts as described previously (14). A Superdex 200 gel filtration column connected to a fast protein liquid chromatography system was used to analyze the native size of Gro mutants. The gel filtration running buffer contained 25 mM HEPES (pH 7.6) and 250 mM NaCl.

Transient-transfection assays and RNAi.

Calcium phosphate-mediated cotransfection was performed with S2 cells as described previously (17). The luciferase reporter activity was analyzed by the dual-luciferase reporter assay system (Promega). For the RNA interference (RNAi) experiment, 1 μg of double-stranded RNA (dsRNA) directed against the Gro 3′ untranslated region (UTR) (matching the sequence from bp 95 to 683 downstream of the stop codon) or control dsRNA directed against the glutathione S-transferase (GST)-conjugated mRNA was cotransfected with the reporter.

Generation and analysis of flies overexpressing Gro mutants.

PCR fragments encoding Gro^40,89D and Gro^38,87D were inserted into the p131 transformation vector (1) so that the Gro coding region was in frame with the amino-terminal 6×myc epitope tag. The upstream activation sequence (UAS) constructs were then introduced into the germ line by standard procedures.

To determine lethality rates, Actin-Gal4/CyO flies were crossed with homozygous UAS-myc-Gro^40,89D or UAS-myc-Gro^38,87D lines. The percent lethality was defined as 100 − [(straight-winged flies/curly-winged flies) × 100]. About 300 curly winged flies were scored per cross.

Clonal overexpression of Gro mutants was achieved as described previously for overexpression of wild-type Gro (31). Briefly, hs-flipase; vgQ-lacZ; Actin>CD2>Gal4/TM6 females were crossed with homozygous UAS-myc-Gro^40,89D or UAS-myc-Gro^38,87D males. Progeny were heat shocked (40 min at 35°C) as first-instar or early-second-instar larvae. After 2 or 3 days, late-third-instar larvae lacking TM6 were selected for dissection and staining.

Immunocytochemistry and X-Gal staining.

Primary antibodies used were mouse anti-Gro monoclonal Ab (kindly provided by Christos Delidakis; 1:2,000), mouse anti-FLAG monoclonal Ab (Sigma; 1:4,000), mouse anti-myc monoclonal Ab (Santa Cruz; 1:1,000), mouse anti-CD2 monoclonal Ab (Serotec; 1:1,000), and rabbit anti-β-galactosidase polyclonal Ab (ICN/Cappel; 1:1,000). Secondary antibodies were from Sigma: fluorescein isothiocyanate-conjugated anti-mouse immunoglobulin G (IgG) (1:100), tetramethylrhodamine-5-isothiocyanate-conjugated anti-rabbit IgG (1:200), Cy3-conjugated anti-mouse IgG (1:800), and peroxidase-conjugated anti-mouse IgG (1:5000). For wing disk staining, the anterior halves of late-third-instar larvae were inverted in 1× phosphate-buffered saline fixed, and antibody staining or 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) staining performed as described previously (62).

Coimmunoprecipitation.

S2 cells transfected with wild-type or mutant Gal4-Gro were resuspended in lysis buffer (20 mM Tris-HCl [pH 7.6], 150 mM NaCl, 2 mM EDTA, 2 mM EGTA, 0.8% Triton X-100, 0.5 mM dithiothreitol, 1× complete protease inhibitor cocktail [Roche]) and rocked at 4°C for 15 min. Clarified lysates were incubated with rabbit anti-Gal4 DNA binding-domain polyclonal Ab (Santa Cruz) and protein A-Sepharose beads (Amersham). The beads were then washed with the lysis buffer, eluted with SDS sample buffer, and analyzed by Western blotting with anti-Gro Ab. To detect the interaction between exogenous and endogenous Gro in the Drosophila embryo, Actin-Gal4/CyO flies were crossed with w¹¹¹⁸, UAS-myc-Gro^40,89D, or UAS-myc-Gro^38,87D lines. Overnight embryos were collected and homogenized in lysis buffer. The lysates were subjected to immunoprecipitation with mouse anti-myc monoclonal Ab.

RESULTS

Oligomerization of Groucho is mediated by putative coiled-coil motifs.

Previous cross-linking, gel filtration chromatography, native gel electrophoresis, and velocity sedimentation data all confirmed that Gro forms a homotetramer and that formation of this oligomer was mediated by the N-terminal Q domain (13). Analysis of this domain with an algorithm designed to search for coiled-coil motifs (42) found two segments with a high likelihood of coiled-coil formation extending from residues 24 to 52 and 73 to 100 (Fig. 1A and B). We referred to these putative amphipathic α-helices as AH1 and AH2, respectively. When helix-breaking proline residues were introduced into AH1 and AH2, tetramerization did not occur (13).

FIG. 1.

Mutagenesis of AH1 and AH2 and predicted effect on coiled-coil probability. (A) Primary structure of Gro. (B) Helical-wheel projections of AH1 and AH2 motifs modeled as parallel coiled coils. Residues targeted for mutagenesis (L38, A40, L87, and T89) are boxed and numbered. The residues at positions a and d occupy the hydrophobic cores of the coiled-coil motifs. Potential favorable electrostatic contacts between the residues at positions e and g in the AH1 homodimer are indicated by dotted lines. (C) Coiled-coil probabilities for wild-type and mutant Q domains determined with the Coils algorithm (42). Wild-type Gro (solid line) exhibits two regions of high coiled-coil probability, which correspond to AH1 and AH2. Conversion of one amino acid at the putative hydrophobic core of each motif to aspartate (Gro^38,87D, dashed line) greatly reduces coiled-coil probability, while conversion of one amino acid on the hydrophilic face of each motif to aspartate (Gro^40,89D, dotted line) has little effect on coiled-coil probability.

To test the prediction that AH1 and AH2 represent coiled coils, we introduced less drastic mutations into these motifs. The residues chosen for mutagenesis were L38, A40, L87, and T89. In each mutant, one or two of these residues were changed to aspartate residues, which, unlike proline residues, are compatible with α-helix formation. L38 and L87 are located on the putative hydrophobic faces of AH1 and AH2, respectively, while A40 and T89 are located on the putative hydrophilic faces of AH1 and AH2, respectively (Fig. 1B). A coiled-coil prediction algorithm (42) suggests that the L38 and L87 mutations will disrupt the coiled-coils while the A40 and T89 mutations will have no effect on coiled-coil formation (Fig. 1C).

To test the effects of these mutations on Gro oligomerization, we carried out coimmobilization assays. These assays used a histidine-tagged N-terminal Gro fragment containing the Q and GP domains (His-GroN), as well as wild-type full-length untagged Gro. These two forms of Gro were expressed together or separately in an in vitro translation system and then incubated with Ni-NTA beads. When expressed separately, untagged Gro did not bind the beads (Fig. 2A, lane 1). However, untagged Gro was efficiently recruited to the beads when cotranslated with His-GroN, demonstrating the formation of mixed oligomers containing both the full-length and truncated proteins (lane 2). Under the conditions of these experiments, formation of the mixed oligomers required cotranslation of the two proteins. When full-length Gro and His-GroN were mixed posttranslationally, very little untagged Gro was recruited to the beads (data not shown), strongly suggesting that oligomer dissociation and/or reassociation is slow relative to the time of the coimmobilization assay.

FIG. 2.

Coimmobilization assays reveal altered oligomerization affinity of different Gro mutants. (A) Untagged wild-type full-length Gro was cotranslated in vitro with wild-type (wt) or point mutant forms of histidine-tagged Gro N-terminal region containing the Q and GP domains (His-GroN) (lanes 2 to 8) or translated alone (lane 1) in the presence of [³⁵S]methionine. The translation products were incubated with Ni-NTA beads, and after extensive washing, Ni-bound proteins were eluted and analyzed by SDS-PAGE and autoradiography. Upper panel shows input proteins, while lower panel shows Ni-NTA bound proteins. (B) Quantitative analysis of the data in panel A. The amount of bound full-length Gro in each lane was normalized to the amount of His-GroN in each lane.

Mutant forms of His-GroN containing aspartate substitutions not predicted to disrupt coiled-coil formation (His-GroN^40D, His-GroN^89D, and His-GroN^40,89D) bound untagged Gro as well as wild-type His-GroN (Fig. 2A, lanes 2 to 5). In contrast, mutant forms of His-GroN containing aspartate substitutions predicted to disrupt the coiled coils bound untagged Gro with reduced efficiency (lanes 6 to 8). The single mutations in these residues (His-GroN^38D and His-GroN^87D) reduced binding by ∼60 to 70%, while the double mutation (His-GroN^38,87D) abolished binding (Fig. 2B).

To further assess the effects of the mutations on Groucho oligomerization, we analyzed FLAG-tagged Gro^40,89D and Gro^38,87D expressed in transformed S2 cell lines. Both proteins localized primarily to the nucleus (Fig. 3A) and were purified from nuclear extracts by anti-FLAG immunoaffinity chromatography (Fig. 3B). On analytical gel filtration chromatography (Fig. 3C), Gro^40,89D, which oligomerizes as efficiently as the wild-type protein according to the coimmobilization assays, eluted primarily as a tetramer, a finding consistent with our previous analysis of wild-type Gro (13). In addition, some of the Gro^40,89D behaved as a higher-molecular-mass aggregate, eluting just after the void volume (the resin has a nominal exclusion limit of 1,300 kDa) but before the 669-kDa marker. The wild-type protein exhibits a similar high-molecular-mass peak on gel filtration chromatography in addition to the tetramer peak (data not shown). In contrast, Gro^38,87D, which failed to oligomerize in the coimmobilization assay, behaved exclusively as a monomer during gel filtration chromatography - we observed neither the tetramer nor any sign of the high-molecular-mass aggregate (Fig. 3C). These experiments support the idea that two coiled-coil motifs corresponding to AH1 and AH2 are required for Gro tetramerization, as well as for the formation of higher-order Gro oligomers.

FIG. 3.

Analysis of Gro oligomerization domain mutants by gel filtration chromatography. (A) S2 cells expressing FLAG-tagged Gro^38,87D or Gro^40,89D were fixed, incubated with anti-FLAG antibody and Cy3-conjugated secondary antibody, and examined by fluorescence microscopy. (B) Coomassie blue-stained SDS-polyacrylamide gel showing FLAG-tagged Gro^40,89D (lane 1) and Gro^38,87D (lane 2) purified from S2 cells. Lane M shows a 10-kDa ladder, with sizes given in kilodaltons (C) Superdex 200 gel filtration profiles for purified FLAG-Gro^40,89D (solid line) and FLAG-Gro^38,87D (dashed line). Arrows indicate the elution positions of protein markers as determined from a calibration run.

Gro oligomerization is required for repression in transfected cells.

To determine whether oligomerization is required for repression, we compared the ability of wild-type Gro and Gro^40,89D, both of which oligomerize, with the ability of Gro^38,87D, which does not oligomerize, to repress transcription. Repression was examined both in transient-transfection assays using S2 cells and in overexpression assays using transgenic flies.

In the transient-transfection assays, Gal4-Gro fusion proteins containing Q-domain mutations were introduced into S2 cells and tested for their ability to repress a cotransfected luciferase reporter. In addition to Gal4 sites, the reporter contained multiple sites for the activators Dorsal and Twist, and cotransfection of the reporter with vectors encoding Dorsal and Twist resulted in a ∼30-fold increase in reporter gene expression (Fig. 4A). When a vector encoding Gal4-Gro^wt was also included in the transfection, reporter activity was significantly reduced (1 and 5 μg of a vector encoding Gal4-Gro^wt resulted in ∼10- and 15-fold repression, respectively). The control mutant Gal4-Gro^40,89D also repressed transcription, although with slightly reduced efficiency relative to Gal4-Gro^wt (1 and 5 μg of a vector encoding Gal4-Gro^40,89D resulted in ∼4- and 12-fold repression, respectively). In contrast, Gal4-Gro^38,87D exhibited no ability to repress transcription (Fig. 4A). An anti-Gro immunoblot (Fig. 4B) demonstrated that the inability of Gal4-Gro^38,87D to repress is not due to poor expression, since Gal4-Gro^38,87D was expressed better than Gal4-Gro^wt and Gal4-Gro^40,89D.

FIG. 4.

Repression by Gal4-Gro in S2 cells: the role of the coiled-coil motifs and endogenous Gro. (A) The G₅DE₅tkLuc firefly luciferase reporter (diagrammed at the top) (5 μg) and a Renilla luciferase internal control reporter (0.1 μg) were introduced into S2 cells. The transfections also included vectors encoding Dorsal (D1, 90 ng), Twist (Twi, 30 ng), and a wild-type (WT) or mutant form of a Gal4-Gro (G4Gro) fusion protein (1 or 5 μg), as indicated. At 2 days posttransfection, firefly luciferase activities were measured and normalized to the control Renilla luciferase activities. The values plotted are the average of duplicate measurements; error bars indicate standard deviation. (B) Lysates of S2 cells transfected as in panel A (5 μg of the Gal4-Gro vectors) were subjected to SDS-PAGE and anti-Gro immunoblotting. (C) S2 cells were transfected and luciferase activites were analyzed as in panel A (2 μg of the Gal4-Gro vectors). Some of the transfections included dsRNA (1 μg) directed against the Gro 3′-UTR or GST. Statistically significant repression was observed on addition of wild-type Gal4-Gro (P < 0.01), Gal4-Gro^38P (P < 0.01), and Gal4-Gro^87P (P < 0.05). Statistically significant derepression due to the addition of Gro dsRNA was observed in the presence of Gal4-Gro^38P (P < 0.05) and Gal4-Gro^87P (P < 0.05). The derepression observed due to the addition of Gro dsRNA in the presence of wild-type Gal4-Gro was not statistically significant (P = 0.39). Lysates of S2 cells transfected with the indicated combinations of expression vectors (5 μg) and dsRNA (1 μg) were also subjected to SDS-PAGE and anti-Gro immunoblotting (bottom). (D) S2 cells transfected with 5 μg of wild-type or mutant Gal4-Gro were subjected to Western blotting (WB) with anti-Gro antibody (upper panel) or immunoprecipitation (IP) with anti-Gal4 DNA binding domain (DBD) antibody followed by Western blotting with anti-Gro antibody (lower panel).

S2 cells contain significant levels of endogenous Gro. Given the ability of Gro to oligomerize, it is possible that the repression directed by Gal4-Gro is enhanced by an interaction with endogenous Gro. To test this possibility, we carried out additional transient-transfection assays in which we used RNAi to reduce levels of endogenous Gro. We were able to target endogenous Gro for degradation without also targeting the recombinant Gal4-Gro by using dsRNA directed against the Gro 3′-UTR, which is absent from the Gal4-Gro expression vector.

In these experiments, we looked at the role of endogenous Gro in repression by the wild-type Gal4-Gro fusion protein and by Gal4-Gro fusion proteins containing mutations disrupting single coiled-coil motifs. The mutant fusion proteins contained single proline substitutions at either L38 (disrupting AH1) or L87 (disrupting AH2). Transient-transfections assays demonstrated that the single proline substitutions reduced but did not abolish repression (Fig. 4C, top). While the wild-type fusion protein repressed transcription by 13-fold, the single mutants repressed transcription by 2- to 3-fold. Although the RNAi treatment significantly reduced the levels of endogenous Gro (Fig. 4C, bottom), there was little or no effect on repression of the luciferase reporter by the wild-type Gal4-Gro fusion protein (Fig. 4C, top). On the other hand, repression by the single-mutant forms of Gal4-Gro was dependent on endogenous Gro, since it was largely abolished on treatment of cells with Gro dsRNA. This finding is consistent with the idea that repression by the single mutants requires oligomerization with endogenous wild-type Gro. Coimmunoprecipitation assays confirmed that both single-mutant forms of Gal4-Gro can oligomerize with endogenous Gro (Fig. 4D). Consistent with the results of the Ni-NTA pulldown assays (Fig. 2), the single-mutant form of Gal4-Gro bound endogenous Gro less avidly than did wild-type Gal4-Gro, while double-mutant Gal4-Gro completely failed to bind endogenous Gro.

In conclusion, the robust repression directed by wild-type Gal4-Gro is largely independent of endogenous Gro. However, when oligomerization is weakened by the disruption of a single coiled-coil motif, repression is enhanced by endogenous wild-type Gro. This further supports the idea that oligomerization contributes to repression, since the endogenous Gro lacks a Gal4 DNA binding domain and can presumably be recruited to the reporter only by oligomerization with Gal4-Gro.

Oligomerization is required for the lethality and developmental defects that result from Gro overexpression.

To determine the effects of mutations in the putative coiled-coil motifs on repression in the fly, we overexpressed mutant forms of Gro by using the binary Gal4-UAS overexpression system. For these experiments, we generated lines containing transgenes encoding myc-tagged Gro^38,87D and myc-tagged Gro^40,89D under control of the UAS. To determine the levels of myc-tagged Gro expression in various transgenic lines, the UAS-myc-Gro lines were crossed with a line containing an Actin-Gal4 driver, which directs ubiquitous expression, and overnight embryos were analyzed by anti-myc immunoblotting. We observed that while there was some line-to-line variability, the range of expression levels of the two forms of Gro was comparable (Fig. 5A). Consistent with the results of the Ni-NTA pulldown assay (Fig. 2), myc-tagged Gro^40,89D but not myc-tagged Gro^38,87D could oligomerize with endogenous Gro (Fig. 5B).

FIG. 5.

Oligomerization is required for Gro-mediated ectopic repression of Brinker target genes omb and vg in the wing disk. (A) Three lines carrying independent insertions of the UAS-myc-Gro^40,89D transgene and four lines carrying independent insertions of the UAS-myc-Gro^38,87D transgene were crossed to an Actin-Gal4 driver line. Lysates of the resulting embryos were subjected to SDS-PAGE and anti-myc immunoblotting. (B) Embryos expressing myc-tagged Gro^40,89D or myc-tagged Gro^38,87D and control w¹¹¹⁸ embryos were lysed and subjected to immunoprecipitation (IP) with anti-myc antibody followed by Western blotting (WB) with anti-Gro antibody. (C to E) omb-lacZ; C765-Gal4 flies were crossed with w¹¹¹⁸, UAS-GFP, UAS-Gro^wt, UAS-myc-Gro^40,89D, or UAS-myc-Gro^38,87D flies. The expression pattern of the C765-Gal4 driver is shown by green fluorescent protein (GFP) fluorescence of a third-instar larval wing disk carrying the UAS-GFP transgene (C, left panel). Wing disks were stained with X-Gal to reveal the pattern of lacZ expression. The omb-lacZ reporter is expressed in a broad stripe at the center of the disk in the flies from the control w¹¹¹⁸ cross (C, right panel). The stripe narrows on introduction of the UAS-Gro^wt transgene (D). Quantitative analysis of the relative width of omb-lacZ stripes in wing disks of each genotype is also shown (E). The width of the stripe was measured as a fraction of the total width of the disk. Forty disks of each genotype were measured. The values plotted represent the averages; error bars indicate standard deviation. Gro^wt and Gro^40,89D reduced the width of the stripe, while Gro^38,87D had no effect. (F to H) Third-instar larval imaginal wing disks from flies bearing the vgQ-lacZ reporter, stained for the CD2 marker (left) and for β-galactosidase (center); the merged image is also shown (right). vgQ-lacZ expression (center, red) is silenced in Gro^WT and myc-Gro^40,89D overexpression clones, marked by loss of the CD2 marker (left, green) (F and G). In contrast, overexpression of myc-Gro^38,87D did not result in repression of vgQ-lacZ (H).

When embryos from this cross were allowed to develop, overexpression of myc-Gro^40,89D resulted in partial lethality while the overexpression of myc-Gro^38,87D did not. Among the six UAS-myc-Gro^40,89D lines tested, rates of lethality ranged from 70 to 97%. The variation in lethality correlated with the variation in expression levels as revealed by the anti-myc immunoblot assay. In contrast, the progeny from crosses between UAS-Gro^38,87D lines and the Actin-Gal4 driver line were always fully viable.

Survivors containing the UAS-myc-Gro^40,89D transgene and the Actin-Gal4 driver always exhibited visible defects. These included wing blisters, abundant extra wing vein tissue, and missing pigment bands on the 6th and 7th abdominal tergites (data not shown). In contrast, flies expressing the UAS-myc-Gro^38,87D transgene never exhibited visible defects, consistent with the dramatically reduced ability of the protein to oligomerize.

Oligomerization is required for Gro-mediated repression of Brk targets in the wing disk.

To further assess the requirement for oligomerization in Gro function, we examined the ability of various mutant forms of Gro to mediate repression by Brinker in the wing disk. In the wild-type wing disk, a Dpp signal emanating from the center of the disk results in the graded activation of the transcription factor heterodimer Mad/Med (2, 47). The Mad/Med heterodimer then cooperates with the transcriptional repressor Schnurri to repress Brinker expression. As a result, Brinker concentrations are highest at the edges of the disk and gradually decrease toward the center. A number of promoters, including the optomotor blind (omb) and vestigal (vg) promoters, are activated by Mad/Med and repressed by Brinker. Antagonism between the opposing Dpp and Brinker gradients is critical in determining the anteroposterior boundaries of the expression domains of these genes. omb is expressed in a narrower anteroposterior domain than is vg, apparently reflecting the somewhat greater sensitivity of this promoter to repression by Brinker. A previous study showed that overexpression of Gro in the wing disk changed the expression pattern of reporters under the control of omb and vg regulatory modules by strengthening repression by Brinker (31), and, as described below, we have taken advantage of this observation to examine the activity of Gro mutations in the wing disk.

To examine the effects of mutant Gro overexpression on the activity of an omb-lacZ reporter, we employed the C765-Gal4 driver, which directs expression throughout the wing disk (Fig. 5C, left panel). In the absence of Gro overexpression, the reporter is expressed in a broad stripe at the center of the third-instar larval wing disk (right panel). Overexpression of wild-type Gro or Gro^40,89D, but not of Gro^38,87D, resulted in a narrowing of the stripe as well as a reduction in its intensity (Fig. 5D and data not shown). To quantitate this change, we measured the width of the stripe relative to the width of the disk in 40 disks of each genotype (Fig. 5E). The measurements reveal that overexpression of wild-type Gro or Gro^40,89D reduced the average width of the stripe by about one-third while overexpression of Gro^38,87D had no effect on the width or intensity of the stripe.

Using a strategy previously employed for wild-type Gro (31), we similarly overexpressed the Gro Q-domain mutants in marked clones in the wing disk. We then assessed the resulting effects on the expression patterns of vgQ-lacZ, a reporter containing the dpp-responsive vgQ enhancer from the vg gene. Both clones overexpressing wild-type Gro (Fig. 5F) and clones overexpressing myc-Gro^40,89D (Fig. 5G) exhibited cell-autonomous repression of vgQ-lacZ. In contrast, clones overexpressing myc-Gro^38,87D exhibited no repression of the vgQ-lacZ reporter (Fig. 5H).

Previous analysis showed that the Q domain is not required for binding to Brinker (31). However, this result was obtained with a Q-domain deletion mutant. To show that the Q-domain point mutations do not alter Gro conformation in a way that prevents binding to Brinker, we carried out coimmbolization assays using GST-Brk fusion proteins. As expected, full-length Gro bound to full-length GST-Brk (Fig. 6A, lane 3) as well as to GST-Brk(441-589) (lane 4), a GST fusion protein containing the central region of Brk, which includes an FKPY Gro-interaction motif (31, 67). In contrast, Gro(2-194), containing the Gro Q and GP domains, bound very poorly to GST-Brk (lane 7) and not at all to GST-Brk(441-589) (lane 8). When we introduced mutations to disrupt AH1 (Fig. 6B, lanes 4 to 6), AH2 (lanes 7 to 9), or both (lanes 10 to 12), we observed a modest reduction in the binding to both GST-Brk (Fig. 6B) and GST-Brk(441-589) (data not shown). This reduced binding can be accounted for by the diminished ability of the mutants to oligomerize, thus reducing the number of Gro protomers immobilized by each GST-Brk/Gro binding interaction. Similar results were observed for the binding of wild-type and mutant Gro to GST-Hairy and GST-Huckebein (data not shown). Thus, the Q-domain point mutants retain the ability to bind Brinker, Hairy, and Huckebein. This reinforces the conclusion that the inability of these mutants to repress transcription is probably the result of their inability to oligomerize.

FIG. 6.

Both wild-type and mutant Gro bind to Brk in vitro. [³⁵S]methionine-labeled Gro proteins were incubated with glutathione-agarose beads bearing GST or GST fusion proteins. After extensive washing, bound proteins were eluted and analyzed by SDS-PAGE and autoradiography. (A) Full-length Gro (lanes 1 to 4) or Gro(2-194) (lanes 5 to 8) were incubated with glutathione-agarose beads containing GST (lanes 2 and 6), GST-Brk (lanes 3 and 7), or GST-Brk(441-589), designated in the figure as GST-BrkΔ (lanes 4 and 8). (B) Full-length Gro^WT (lanes 1 to 3), Gro^38P (lanes 4 to 6), Gro^87P (lanes 7 to 9), or Gro^38,87P (lanes 10 to 12) were incubated with glutathione-agarose beads bearing GST (lanes 2, 5, 8, and 11) or GST-Brk (lanes 3, 6, 9, and 12). The percentage of input protein that bound GST-Brk beads is indicated at the bottom of the GST-Brk lanes.

DISCUSSION

Role of coiled-coil motifs in oligomerization.

The experiments presented here strongly suggest that AH1 and AH2 are coiled-coil motifs required for Gro oligomerization since mutations predicted not to interfere with coiled-coil formation had no effect on oligomerization while mutations predicted to disrupt the coiled coils reduced or eliminated oligomerization. The gel filtration analysis demonstrates that in addition to the previously observed tetramer, the coiled coils mediate the formation of higher-order oligomers. The breadth of this high-molecular-mass peak suggests that it is heterogeneous. For example, it may represent an open-ended polymer with variable numbers of subunits.

The ability of the same domain to mediate both tetramerization and the formation of higher-order oligomers could easily be explained by the idea of “domain swapping” (8, 40). This is a process whereby a relatively small closed oligomer can be converted into a larger open oligomer by an exchange of a structural element between subunits. In the case of Gro, the tetramer may represent a closed oligomer while the high-molecular-mass form may represent an open oligomer (Fig. 7).

FIG. 7.

Domain swapping can convert a closed oligomer into a polymer. The black and white objects represent two parts of an oligomerization domain that interact noncovalently and are connected by a flexible linker capable of assuming multiple conformations. In the case of Gro, the two coiled coils are hypothesized to contribute in some unspecified way to the interface.

The coimmobilization and transfection assays show that both AH1 and AH2 are required for high-affinity oligomerization and for efficient repression. Single-point mutations disrupting either coiled-coil motif allowed some residual oligomerization. When the single-point mutants were introduced into a Gal4-Gro fusion protein, weak residual repression of Gal4 binding-site-containing reporters was observed. However, this repression was dependent on endogenous Gro as shown by an experiment in which we specifically reduced the level of the endogenous protein by RNAi. Presumably, when the oligomerization domain of Gal4-Gro is crippled by mutagenesis of a single coiled-coil motif, the formation of mixed oligomers between the DNA-bound Gal4-Gro mutant and wild-type endogenous Gro is required for repression.

Long- versus short-range repression.

Repressors that employ Gro as a corepressor, such as Hairy and Dorsal, generally mediate long-range repression (7, 66). That is, they are often able to silence a locus completely, regardless of where they are bound relative to the core promoter or relative to the binding sites for activators. In contrast, a number of Gro-independent repressors, including Krüppel, Knirps, and Snail, function as short-range repressors (4, 26). Specifically, they block activation only by activators bound to the template within about 100 bp of the repressor. Short-range repressors are able to interfere with the action of one enhancer while still allowing another enhancer to function. This so-called enhancer autonomy is important at complex loci such as the pair rule loci, in which a single promoter is under the control of multiple enhancers that need to function independently of one another. Interestingly, the short-range repressors mentioned above all rely on the corepressor dCtBP (50). Thus, Gro and CtBP may function by distinct mechanisms that ensure long and short-range repression, respectively.

Studies of other types of long-range repression, particularly heterochromatic silencing, suggest that long-range repression involves the establishment of large, transcriptionally silent chromosomal domains with altered chromatin structure (28, 41). The establishment of a large silent domain often involves the recruitment of enzymes such as histone deacetylases and histone methyltransferases that change the posttranslational modification state of histones. For example, Gro recruits the histone deacetylase HDAC1 via its GP domain, a region just C-terminal to the Q domain (10, 12, 15). Paradoxically, however, the short-range corepressor dCtBP also appears to function, at least in part, by recruiting histone deacetylases to the template (18, 63). Therefore, the difference between long- and short-range repression is not determined by whether a corepressor serves to recruit histone-modifying enzymes. Indeed, studies with yeast have shown that histone-modifying enzymes such as histone deacetylases may normally act in a very local manner to modify only about a 100- to 200-bp region surrounding the site of recruitment (34, 61).

Our finding that Gro can form high-order oligomers and that oligomerization is essential for Gro-mediated repression suggests another way in which long-range repression might be distinguished from short-range repression. In particular, long-range repressors may generally possess the ability to polymerize, perhaps allowing them to spread along the chromatin template and establish a transcriptionally silent domain. In addition to Gro, other long-range corepressors that may have the ability to polymerize include Sir3/Sir4 (29), which is involved in heterochromatic silencing in budding yeast, and HP1/Swi6, which mediates heterochromatic silencing in fission yeast, metazoans, and plants (52). In addition to having the ability to oligomerize, these long-range corepressors bind to specifically modified forms of histones—both Sir3/Sir4 and Gro bind hypoacetylated histone tails (23, 29, 53), while HP1 binds very specifically to the lysine 9-methylated histone H3 tail (6, 36). This interaction between the template and the corepressor may facilitate spreading along the template from a nucleation site. In each case, the corepressor has the additional ability to recruit a histone-modifying enzyme that changes the modification state of the template in such as way as to increase the affinity of the corepressor for the template. Specifically, Gro recruits HDAC1 (a class I histone deacetylase) (10, 12, 15), Sir3/Sir4 recruits Sir2 (an NAD⁺-dependent histone deacetylase) (32, 37), and HP1 recruits Su(var)3-9 (a histone H3 lysine 9 methyltransferase) (30, 58). This may result in the establishment of a self-reinforcing transcriptionally silent state and, in the case of heterochromatic silencing, may be important for the epigenetic stability of the silenced state.

Overexpression of Gro makes Brinker a more efficient repressor.

Dpp is expressed at the anteroposterior midline of the wing disk, from which it is thought to diffuse, resulting in the formation of a Dpp concentration gradient (38, 48). Since the Dpp signal results in the activation of Mad/Med (a transcriptional activator) and the transcriptional silencing of brk (which encodes a Gro-dependent repressor) (44, 65), the gradient of Dpp activity results in a parallel gradient of Mad/Med activity but an opposing gradient of Brk activity. Competition between Mad/Med and Brk is then thought to determine the domains of transcription of dpp target genes such as omb (2, 47). Based on this model, one might expect the on/off state of Dpp target genes such as omb to depend only on the balance between Mad/Med and Brk. However, the findings presented here and elsewhere (31) suggest that the concentrations of coregulators can also be critical. Specifically, when functional Gro was overexpressed in the wing disk, the domain of omb expression was narrowed, indicating that limiting concentrations of Brk can be compensated for by increased levels of Gro. This finding is compatible with the idea that Gro-dependent repression depends on the cooperative assembly of a nucleoprotein complex termed a repressosome. In addition to DNA-bound repressors, this repressosome probably includes corepressors, nucleosomes, and architectural factors and may be analogous to the enhanceosomes that mediate transcriptional activation (16).