Patterning along developing body axes is regulated by gradients of transcription factors, which activate or repress different genes above distinct thresholds. Understanding differential threshold responses requires knowledge of how these factors regulate transcription. In the Drosophila wing, expression of genes such as omb and sal along the anteroposterior axis is restricted by lateral-to-medial gradients of the transcriptional repressor Brinker (Brk). omb is less sensitive to repression by Brk than sal and is consequently expressed more laterally. Contrary to previous suggestions, we show that Brk cannot repress simply by competing with activators, but requires specific repression domains along with its DNA-binding domain. Brk possesses at least three repression domains, but these are not equivalent; one, 3R, is sufficient to repress omb but not sal. Thus, although sal and omb show quantitative differences in their response to Brk, there are qualitative differences in the mechanisms that Brk uses to repress them.
Introduction
During animal development, spatial patterns of gene expression can be established along one body axis by a single transcription factor if the concentration or activity of this factor is graded along that axis, and if target genes are differentially sensitive to activation or repression by that factor. The best examples of this are found in the early Drosophilaembryo. For example, the transcription factor Hunchback is graded along the anteroposterior (AP) axis and establishes distinct domains of gap and pair-rule gene expression along that axis(Rivera-Pomar and Jackle,1996). Along the dorsoventral (DV) axis, there is a nuclear-cytoplasmic localization gradient of the Dorsal transcription factor,which is exclusively nuclear in the ventral region and cytoplasmic dorsally. The distinct expression domains of several genes along the DV axis can be explained by their sensitivity to either activation or repression by Dorsal(Stathopoulos and Levine,2002).
Our understanding of how a single transcription factor can activate or repress a gene at one concentration, but is required at higher levels to influence another, is still fairly rudimentary. Again, some of the best insights into how threshold responses to transcription factor gradients are deciphered at the level of cis-regulatory elements have come from studies of transcription factors acting during early Drosophila embryogenesis. The most obvious mechanisms involve modifying the number of binding sites or the affinities of those sites in enhancers to vary their sensitivity to a transcription factor, and these mechanisms appear to be used in enhancers regulated by Dorsal, Hunchback, Kruppel and Knirps(Clyde et al., 2003; Jiang et al., 1992; Langeland et al., 1994). However, sensitivity to activation or repression may be achieved in other ways; for example, sensitivity to repression by Giant can to be modified by the positioning of binding sites in relation to a promoter(Hewitt et al., 1999).
Gradients of transcription factors are used to generate spatial patterns of gene expression in many other systems, including in the Drosophilawing, where the transcriptional repressor Brinker is expressed in a lateral-to-medial gradient in the anterior and posterior halves(Campbell and Tomlinson, 1999; Jazwinska et al., 1999; Minami et al., 1999; Muller et al., 2003). This pattern of expression is established by the morphogen Decapentaplegic (Dpp, a TGF-β), which is expressed at the center of the AP axis and, following secretion, becomes distributed in medial-to-lateral gradients in the anterior and posterior compartments (Blackman et al., 1991; Entchev et al.,2000; Masucci et al.,1990; Teleman and Cohen,2000). Dpp, acting through the intracellular signal transducer Mad(a Smad) and Schnurri (a nuclear zinc finger protein), represses brkexpression, so that the graded expression of brk mirrors that of the Dpp protein (Marty et al.,2000; Muller et al.,2003).
Brk functions to repress the expression of genes in the wing that were originally classified as Dpp targets. In fact, in the absence of both Dpp and Brk, these targets are still expressed, indicating that Dpp regulates their expression largely indirectly, through the repression of Brk. These targets include spalt (sal), optomotor-blind (omb; bifid, bi – FlyBase) and the vestigial quadrant enhancer(vg-QE, an enhancer recapitulating a portion of the expression of the vestigial gene) (Campbell and Tomlinson, 1999; Jazwinska et al., 1999; Kim et al.,1996; Marty et al.,2000), and have nested expression domains centered on the dpp stripe, so that the vg-QE domain is wider than that of omb, which is wider than that of sal(Kim et al., 1997; Lecuit et al., 1996; Nellen et al., 1996). As all three genes appear to be targets of Brk, this pattern of expression can be explained by a differential sensitivity to Brk, with sal being repressed by very low levels of Brk, but vg-QE requiring higher levels.
Here, we investigate how Brk represses gene expression, to further our understanding of why some targets are more sensitive to Brk than others. Previous studies have suggested that Brk may repress different targets by different mechanisms. First, Brk-binding sites in the cis-regulatory regions of some embryonic targets, including zen and Ubx, overlap with those of an activator, namely Mad (which can function both to activate and repress at different loci), and in vitro studies indicate that Brk and Mad can compete for binding to the same region of DNA(Kirkpatrick et al., 2001; Rushlow et al., 2001; Saller and Bienz, 2001). Second, Brk possesses interaction motifs for the co-repressors Groucho (Gro)(Chen and Courey, 2000) and CtBP (Chinnadurai, 2002),indicating that it may use more active mechanisms to repress targets(Hasson et al., 2001; Saller and Bienz, 2001; Zhang et al., 2001). Loss of Gro or CtBP does result in derepression of some Brk targets, such as the vg-QE, but not others, such as omb, indicating that Brk may use different mechanisms to repress different genes(Hasson et al., 2001).
We show that Brk requires its DNA-binding domain (DBD) plus a repression domain to act as a repressor, the DBD alone is insufficient to repress targets, even those that have overlapping Brk- and Mad-binding sites. This poses the question of whether competition is a real phenomenon in vivo. Brk possesses at least three independent repression domains, the Gro and CtBP interaction motifs (GiM and CiM, respectively) and one other domain, defined here as 3R. However, these domains are not equivalent; 3R is sufficient for repression of omb but not sal, and this difference may be related to the spacing of Brk-binding sites in relation to activator sites. Thus, although sal and omb show quantitative differences in their response to Brk, this may actually be based more on qualitative differences in the mechanisms that Brk uses to repress them.
Materials and methods
Sequence comparisons
Sequences for brinker from Anopheles gambiae (accession number AY578799), Drosophila pseudoobscura (included in AADE01000981)and Bombyx mori (composite of AU004448 and AV402267) were identified through BLAST searches of NCBI databases.
Fly strains and mutational analysis
Flies carrying the following existing alleles or transgenes were used: brkF124, brkE427, brkF138, brkM68, CtBP87De-10, groE48, vg-QE(P{vg(806)-lacZ}), hs-GFP (Avic\GFPhs.T:Hsap\MYC), hs-flp(P{hsFLP}22), FRT18A (P{ry[+t7.2]=neoFRT}18A), FRT82B (P{neoFRT}82B), Ubi-GFP(P{Ubi-GFP(S65T)nls}3R), omb-lacZ (P{lacW}biPol-1), C765(Scer\GAL4C-765), en-Gal4 (P{en2.4-GAL4}e16E), UbxB (P{Bhz}) and 24B-Gal4 (P{GawB}how24B). Unless indicated otherwise in parentheses, all genotypes are as denoted in FlyBase(http://flybase.bio.indiana.edu),where more information on each can be found. To molecularly characterize mutants, the brk gene was amplified by PCR from genomic DNA from hemizygous embryos and was sequenced.
Generation of in vitro mutated/modified UAS-brk transgenes
The UAS-brkA438 line, which produces an unmodified, untagged wild-type Brk protein, was generated by cloning a brk cDNA into a modified pUAST vector, which had the white gene removed. A >y+> flp-out cassette was inserted between the UAS sequences and the cDNA. Transgenic flies were generated by standard procedure and the flp-out cassette was removed from transformants using hs-flp, to generate UAS-brkA438. All other transgenes were cloned into the standard pUAST vector and included a sequence to introduce two copies of the HA tag to the C terminus, the sequence of this was (GS/EF/RS)MAGNIYPYDVPDYAGYPYDVPDYAG (the HA sequence is underlined).
The exact locations of mutations and deletions in different transgenes are indicated in Fig. 2. C-terminal deletions were generated by PCR using an external primer, and internal primer flanked by KpnI and either EcoRI, BglII or BamHI sites, respectively. Mutations and internal deletions were generated by inverse PCR followed by religation using two internal primers flanked by restriction sites generating the following changes to sequence. CM mutation (CiM), PMDLSLG to AMAAALA (NotI); GM mutation (GiM), FKPY to FAAA (NotI; similar mutations were shown to result in loss of CtBP and Gro binding to Brk); NA deletion (3R), residues 148-200 to RS(BglII); ΔA17 deletion, residues 173-189 to RS(BglII). The NLS and NLSW proteins were generated by PCR using an internal primer flanked by sequences encoding the following, PPKKKRKV(matching the NLS sequence from SV40 T antigen(Kalderon et al., 1984) plus WRPW for NLSW. The EC construct was generated by amplifying residues 151-228 and 383-452, respectively, by PCR and cloning onto NLS. More details on each construct and methods of construction are available on request.
Clonal analysis and ectopic expression
Homozygous mutant clones were generated in imaginal discs by hs-flp/FRT-induced mitotic recombination. Clones were generated in the second or early third instar of larvae with the following genotypes:
y omb-lacZ brkF138 FRT18A/hsGFP FRT18A; hs-flp (and similarly for brkM68, brkE427 and brkF124);
y brkF138 FRT18A/hs-GFP FRT18A; vg-QE; hs-flp (and similarly for brkM68, brkE427 and brkF124); and
hs-flp; FRT82B CtBPl(3)87De–10groE48/FRT82B Ubiq-GFP (and similarly for single mutant clones).
Clones in discs were identified by the loss of GFP.
Ectopic expression of UAS-transgenes was achieved by independently crossing transformant lines to two Gal4-expressing lines: en-Gal4 (expressed in the posterior) and C765 (ubiquitous expression in the wing). For assigning activity level as observed in adult wings, the following criteria were used.
++++, wild-type level (no modified/mutated protein achieved this level). No adults were obtained with en-Gal4 even when reared at 17°C. With C765 at 20°C, there was an almost complete loss of wing blade.
+++, some adults were obtained with en-Gal4 at 20-25°C, but showed substantial loss of posterior wing tissue and veins. With C765 at 25°C,there was an almost complete loss of wing blade.
++, adult flies were obtained with en-Gal4 at 25-30°C. Their wings had a loss of tissue or a fusion of veins IV and V and loss of the posterior crossvein. With C765 at 30°C, the wings were slightly smaller and had vein defects, including extra crossveins.
+, adult flies obtained with en-Gal4 at 25-30°C, with loss of the posterior cross vein. With C765 at 30°C there was little or no effect on the wings.
–, no activity. No abnormal phenotype under 25°C; at 30°C there was often some disruption to wing venation, such as extra small veins around the posterior crossvein and vein V. This was distinct from the other phenotypes above and may be caused by weak dominant-negative activity.
At least three lines of each construct were tested apart from F124 (one line). Although there was some variability in the level of activity from line to line, in general most lines from any one construct fell into the same category of activity level. To be assigned to one of the above categories, at least two lines from a construct had to have a similar level of activity; in fact, for most constructs, at least three lines had similar levels of activity.
Generation of mutations in UAS-brkA438
A homozygous strain of UAS-brkA438 was mutated with ethylmethane sulphonate (EMS) using standard procedure(Grigliatti, 1986) and crossed to the ubiquitous Gal4 line C765. The progeny were raised at 25°C, which normally results in flies with almost no wing blade(Fig. 3B); flies with larger wings were selected. The UAS-brk transgene was amplified from potential mutants by PCR and sequenced.
Immunostaining, X-gal staining and analysis of adult wings
Dissection and staining of imaginal discs was carried out by standard techniques. omb expression was detected using a lacZenhancer trap. The following antibodies were used: anti-Sal (rabbit, 1:50)(Kuhnlein et al., 1994); anti-βgal (rabbit, 1:2000, Cappell; and chicken,1:2000, Abcam); anti-HA (mouse, 1:1000, Covance). Embryos carrying the UbxB reporter were stained with X-gal by the standard protocol(Lawrence and Johnston, 1986);all embryos were fixed and stained along side each other. Wings from adult flies were mounted in GMM.
Results
Comparison of Brk homologs in other species identifies the DBD and the GiM and CiM
As an initial approach to identifying the regions of the Brk protein required for it to repress gene expression, a comparison was made between Brk proteins from four different insect species: the Dipterans, Drosophila melanogaster, D. pseudoobscura and Anopheles gambiae, and the more distantly related Lepidopteran, Bombyx mori. This identified a single region of extended similarity consisting of 60 amino acids(Fig. 1) that coincides with the helix-turn-helix (HTH) motif predicted from the original sequence and the DBD identified by in vitro studies(Rushlow et al., 2001; Saller and Bienz, 2001; Sivasankaran et al., 2000; Zhang et al., 2001). The size of this domain and the positioning of the HTH are consistent with a previous suggestion (Jazwinska et al.,1999) that the Brk DBD has similarity to a homeodomain, which is the same size and is composed of three helices, the last two comprising a HTH. Outside of the DBD, a single shorter region with significant similarity was identified, centered on the GiM (Fig. 1). The CiM is also conserved amongst all four species(Fig. 1). However, outside of these motifs and the DBD, there are no significant regions of similarity shared by all four proteins.
Generation and analysis of mutated/modified Brk protein
To determine whether the DBD, GiM, CiM or other domains of Brk are required for repression of targets in the wing, three approaches were taken to identify or generate mutant/modified forms of Brk protein in which one or more of these domains/motifs were mutated or deleted. (1) The molecular lesions in three previously reported mutations in the endogenous gene, brkF124, brkF138 and brkE427 (Lammel et al., 2000; Lammel and Saumweber, 2000), were characterized(Fig. 2B). (2) Five new mutations were generated in a wild-type brk transgene carried in line UAS-brkA438 (Fig. 2C). This screen was performed by driving ubiquitous expression of this transgene in the wing using the Gal4 line C765: this results in an almost complete elimination of the wing blade at 25°C(Fig. 3B). Five point mutants in UAS-brkA438 were identified as flies with a significantly increased wing size in this assay (Fig. 3). (3) Modified UAS-brk transgenes were generated that produced HA-tagged proteins in which different regions were included, modified, mutated or fused to other domains, including nuclear localization sequences (NLS) and repression motifs (Fig. 2D).
The activity of these mutated/modified Brk proteins was compared with that of the wild-type protein by analyzing the expression of known Brk targets: sal, omb and the vg-QE in wing discs, and a Ubx reporter, UbxB, in embryos (Kirkpatrick et al.,2001; Saller and Bienz,2001). For endogenous mutants (brkF124, brkF138 and brkE427), expression of these targets was examined in marked homozygous mutant clones. For null mutants, such clones show misexpression of Brk targets in lateral regions(Campbell and Tomlinson, 1999; Jazwinska et al., 1999; Minami et al., 1999).
The UAS-brk transgenes were misexpressed in the posterior of wing discs using en-Gal4, and the expression of Brk targets in the posterior was compared with that in the anterior. Larvae were raised at different temperatures to vary the amount of protein ectopically expressed (Gal4 is cold sensitive). At 20°C, wild-type Brk3PF3 completely repressed sal and omb, and almost completely repressed the vg-QE(Fig. 2D, Fig. 5A,J). At 17°C, sal is still completely repressed, but some omb expression can be detected,indicating that wild-type Brk is more effective at repressing salthan omb (Fig. 5B). In embryos, UAS-brk transgenes were misexpressed in mesoderm using the 24B driver; at 25°C, Brk3PF3 completely repressed UbxB(Fig. 6B).
The activity of UAS-brk transgenes was also assessed in adult wings using both en-Gal4 and the weaker C765 driver. UAS-brk3PF3; en-Gal4 animals did not survive to adults, even at 17°C, whereas UAS-brk3PF3; C765 adults emerged at 20°C with an almost complete loss of the wing blade (Fig. 7B). The modified/mutated Brk proteins showed a wide range of activity levels based on the wing phenotype they induced(Fig. 7C-F). Activity level was classified into five categories (`–', no repressor activity to `++++',full, wild-type activity; see Materials and methods for criteria).
The results of these studies are summarized in Fig. 2 and are described in detail below.
Mutations in the DBD reduce or completely eliminate Brk activity
Mutations in the DBD were identified in one of the endogenous mutants, brkF124, and four of the UAS-brkA438 mutants,D44, S4, C and F2 (Fig. 2). brkF124 has an amino acid substitution in the recognition helix R82W (Fig. 2B), and resulted in a protein with little or no activity: brkF124mutant clones in wing discs were indistinguishable from those of null alleles,showing autonomous misexpression of sal, omb and the vg-QE(Fig. 4A,B). The mutations in UAS-brkA438 were located in different regions of the DBD(Fig. 2C) and had varying effects on the activity of the resulting Brk protein, as assessed by the size of the wing produced with C765 (Fig. 3), although none appeared to completely eliminate activity. We also generated UAS-brkF124, which generated a full-length,tagged protein with the same mutation that is found in brkF124. This also had no repressor activity, as assessed by sal and omb expression, although when misexpressed at high levels it did result in the formation of ectopic veins in adult wings,which may be due to an abnormal activity of this protein; for example, it may have a modified DNA-binding specificity(Fig. 2D, Fig. 7F).
The DBD of Brk alone has no repressor activity
If Brk represses simply by binding-site competition, a protein consisting primarily of the DBD should be capable of doing this. Residues 44-99 have previously been shown to be sufficient to bind DNA and compete with activators in vitro (Saller and Bienz,2001). Two mutants were identified that should produce proteins consisting primarily of the DBD, brkE427 and UAS-brkA438-53 (truncated at residues 102 and 115, respectively; Fig. 2). However, both mutants had no detectable activity in vivo (Fig. 2B,C; Fig. 3G). One possible reason for their inactivity was that the mutant proteins lack an NLS and were predominantly cytoplasmic. However, their localization could not be characterized with available antibodies, so an equivalent tagged protein,Brk53 (Fig. 2D), was expressed in wing discs and was shown to be primarily cytoplasmic (data not shown). Consequently, we attached a NLS to produce the protein BrkNLS. This was localized in the nucleus(Fig. 5E) but still had no detectable activity (Fig. 2D),having no effect on sal or omb expression(Fig. 5C,D), or on the size of adult wings (Fig. 7F).
The in vitro studies suggesting that Brk may use binding site competition as a mechanism for repression were carried out with genes expressed in the embryo (Kirkpatrick et al.,2001; Rushlow et al.,2001; Saller and Bienz,2001), and it is possible that Brk does not use this mechanism in the wing. Consequently, we tested one of these embryonic targets, UbxB(Thuringer et al., 1993),which is expressed in the midgut mesoderm, but found that it too was not repressed by BrkNLS (Fig. 6C).
Addition of a minimal repressor motif to the DBD restores activity
One possible reason for the inactivity of the BrkNLS protein is that it cannot actually bind DNA in vivo. To test this, we designed an identical construct, UAS-brkNLSW, with the addition of a minimal repressor motif consisting of just four amino acids, WRPW(Fig. 2). WRPW functions as a repression motif by recruiting Gro (Aronson et al., 1997; Fisher et al.,1996). BrkNLSW protein has considerable activity; it significantly reduced the size of adult wings (but is not as effective as wild-type protein), and it could repress sal completely and omb almost completely (Fig. 5F). In this regard it behaves similarly to wild-type protein,i.e. it represses sal more effectively than omb. It had no effect, however, on vg-QE expression (Fig. 5K), but could substantially repress UbxB in the embryo(Fig. 6D).
The CiM and GiM are required for Brk to repress some, but not all targets
If the DBD alone is unable to repress gene expression, other regions of the protein must be required, the obvious candidates being the CiM and GiM. However, analysis of the brkF138 mutant, which lacks both motifs, revealed that Brk could repress at least one target, omb, in their absence. This mutant is predicted to produce a protein truncated at residue 332, i.e. before these two motifs(Fig. 2B). In contrast to null clones, omb was not derepressed in brkF138 clones(Fig. 4C,D), although we cannot rule out the possibility that there may be a very minor expansion of expression of one or two cell diameters (the edge of the endogenous domain is difficult to resolve precisely, because it is not completely straight). However, sal and the vg-QE were ectopically expressed in some brkF138 clones (Fig. 4C,D,E). For sal, this ectopic expression was restricted to the omb domain, i.e. expansion of the sal domain did not extend to the edge of the wing pouch, as it would in brk null mutant clones.
A recent study showed that sal expression is dependent upon Omb(del Alamo Rodriguez et al.,2004), suggesting that the lack of ectopic sal expression in lateral brkF138 clones (i.e. outside of the omb domain) could be explained by the absence of ectopic ombin these clones, rather than by BrkF138 directly repressing sal. In support of this, sal is not misexpressed in omb brk null double-mutant clones (data not shown). del Alamo Rodriguez et al. also suggested that vg-QE expression is dependent upon Omb(del Alamo Rodriguez et al.,2004). However, vg-QE expression actually shows expansion into lateral brkF138 clones(Fig. 4E), i.e. in the absence of any ectopic omb expression. Possible explanations for these conflicting results are that vg-QE expression may not be dependent upon Omb in all situations, or that there is some ectopic omb expression in these clones but it cannot be detected with the omb enhancer trap used.
CtBP or Gro is required for repression of sal
BrkF138 protein appears to be able to repress ombwithout recruiting either CtBP or Gro, but either one or both may be required for repression of sal and the vg-QE. This is consistent with previous studies showing that neither CtBP nor Gro is required for the repression of omb, and that Gro is required for repression of the vg-QE(Hasson et al., 2001). We have extended these studies, and have found that, in contrast to omb, Gro is required for the full repression of sal, whereas CtBP can partially substitute for the loss of Gro [this is in contrast to previous claims that neither Gro nor CtBP is required for repression of sal(Hasson et al., 2001)]. CtBP/gro double mutant clones in the wing had the same phenotype with respect to sal expression as brkF138 clones, i.e. ectopic expression, but only within the omb domain (Fig. 4F). By contrast, sal expression is normal in CtBP single mutant clones (Fig. 4G), demonstrating that the ectopic expression of salfound in CtBP/gro double mutant clones can be rescued completely by Gro alone. sal expression in gro single mutant clones, however, is occasionally expanded(Fig. 4H), but not as extensively as in the CtBP/gro double mutant clones,indicating that CtBP can provide some, but not full, repressor activity to narrow the lateral limit of sal expression.
CtBP/Gro-independent repression by Brk is dependent upon a domain,3R, located just C-terminal to the DBD
The inactivity of BrkNLS in comparison to BrkF138suggested that the region between the DBD and the CiM contains an additional repression domain. This was tested by misexpression of additional modified Brk proteins. BrkStop1 is similar to BrkF138, being truncated immediately before the CiM (Fig. 2), and was tested initially to rule out the possibility that any Brk activity in the brkF138 mutant was due to read through of the stop codon. Misexpression of BrkStop1 in the posterior compartment at 25°C resulted in complete repression of omb and sal, and of the vg-QE (Fig. 5H,L). At 20°C (with concomitant lower expression of the transgene), omb was still almost completely repressed, although there was still weak residual expression remaining in the posterior compartment(Fig. 5I). By contrast, at this temperature sal was still strongly expressed in the posterior(although the size of its domain was significantly reduced), demonstrating that BrkStop1 is much more effective at repressing ombthan sal, and, thus, that it differs from the wild-type protein.
Although we cannot rule out the possibility that BrkStop1represses sal directly 25°C, it is likely that it does this indirectly: sal requires Omb for expression(del Alamo Rodriguez et al.,2004) and omb is completely repressed at this temperature(Fig. 5H). It should be pointed out, however, that at 20°C, omb expression cannot be detected in some cells that are expressing sal(Fig. 5I). Because previous studies (del Alamo Rodriguez et al.,2004), and our own (not shown), indicate that salexpression is absolutely dependent upon Omb, the likeliest explanation for this observation is that omb is expressed in these cells at levels sufficient for expression of sal, but that is just too low to be detected with the omb-lacZ line.
BrkStop1 can also repress the reporter UbxB and the vg-QE. It can, in fact, repress UbxB even more effectively than BrkNLSW does(Fig. 6E). It is unclear why BrkStop1 can repress the vg-QE when BrkF138 appears to be compromised in this respect (Fig. 4E), and when vg-QE expression is upregulated in gromutant clones (Hasson et al.,2001). It is possible that Gro is required for repression of the vg-QE when Brk is present at physiological levels, but not at the higher levels that can be achieved with the UAS/Gal4 system.
Additional truncations identified a minimal protein, BrkA,truncated at residue 206 (Fig. 2D), with similar activity to BrkStop1. A series of additional constructs, which either had repressor activity, BrkECand BrkStop1Δ17, or did not, BrkStop1NA and BrkA2, identified the region 151-206 as being the minimum region sufficient to confer repressive activity. This region has been termed 3R, for the third repression domain, in addition to the CiM and GiM(Fig. 2A). 3R consists of a histidine-rich region, a stretch of poly-alanine and some unique sequence at the C terminus. Further studies are required to narrow down essential sequences in this region; for example, the fact that BrkStop1ΔA17 still has considerable activity suggests that the stretch of poly-alanine is probably not essential.
Brk protein possessing only the CiM has significant activity
Experiments described above show that a Brk protein possessing only a GiM,or 3R, has significant activity (Fig. 2). We further demonstrated that a protein,BrkStop1NAC, possessing only a CiM (deleting the 3R domain and terminating immediately after the CiM; Fig. 2D) repressed omb and sal, and significantly reduced the size of adult wings (Fig. 2D, Fig. 5G). It behaved similarly to wild-type Brk, and was more effective at repressing sal than omb.
Inactivation of a single repression domain/motif reduces the activity of Brk
To test whether the loss of an individual repression domain/motif resulted in reduced Brk activity, we tested the activity of Brk proteins in which either the CiM (BrkCM), GiM (BrkGM) or 3R(BrkNA) was mutated or deleted. In contrast to wild-type Brk3PF3 (Fig. 7B),adult flies did survive following the misexpression of each of these single mutant transgenes with en-Gal4 at 20°C(Fig. 7C) indicating that they were less active than the wild-type protein. In terms of gene expression, each could repress sal, omb and the vg-QE, but were essentially too similar in their activity level to each other, and to the wild-type Brk3PF3 protein, to make any clear conclusions.
Brk must possess an additional repression domain(s)
If Brk requires a DBD plus a repression domain to function, and 3R, CiM and GiM are the only repression domains/motifs, then mutation or deletion of all three should render the Brk protein inactive. However, in Brk3M, 3R is deleted, and the CiM and GiM are mutated, but this protein still had significant activity (Fig. 2D, Fig. 7D), indicating that there must be at least one more repression domain/motif, which is probably located between the CiM and the C terminus. Preliminary studies indicate a fourth repression domain may lie between the CiM and GiM, but further analysis is required to confirm this finding (not shown).
Discussion
A key step in Dpp acting as a morphogen in the Drosophila wing is its generation of lateral-to-medial gradients of Brk expression, a pattern that mirrors the distribution of Dpp itself. Brk represses `Dpp targets', such as sal and omb, and their lateral limits of expression are defined by their sensitivity to Brk(Campbell and Tomlinson, 1999; Entchev et al., 2000; Jazwinska et al., 1999; Minami et al., 1999; Muller et al., 2003; Teleman and Cohen, 2000). Consequently, the less sensitive target omb is expressed in a wider domain than the more sensitive target sal. The question we wished to address was: why are some genes, such as sal, repressed by lower levels of Brk than others, such as omb? Here, we have investigated the mechanisms that Brk uses to repress gene expression, to determine whether these are the same for different genes. If so, any difference in sensitivity between genes would be simply quantitative: Brk would operate identically, but just more effectively on more sensitive targets. However, if Brk does not always use the same mechanism, then any perceived difference in sensitivity may result from relative differences in the ability of Brk to use alternative mechanisms. We show that Brk can use different mechanisms to repress sal or omb.
Brk does not appear to repress by simple competition
The simplest method of transcriptional repression involves competition with an activator, and can operate at the level of DNA if the activator and the repressor have the same, or overlapping, binding sites in an enhancer. In theory, assuming a transcription factor is nuclear, it should only require a DNA-binding domain to act in this fashion. Brk has been shown to possess an N-terminal sequence-specific DNA-binding domain (DBD; Fig. 1)(Kirkpatrick et al., 2001; Rushlow et al., 2001; Saller et al., 2002; Sivasankaran et al., 2000; Zhang et al., 2001), and here we have identified several mutations in this domain that either completely inactivate or reduce the activity of the protein (Figs 2, 3, Fig. 4A), indicating that this region is essential for Brk activity.
Previous studies suggested that Brk could function by competition, more specifically, by competing with Mad for overlapping binding sites in vitro(Kirkpatrick et al., 2001; Rushlow et al., 2001; Saller and Bienz, 2001). However, a nuclear localized Brk protein consisting primarily of the DBD,BrkNLS, cannot repress any Brk target in vivo(Fig. 2D, Fig. 5C, Fig. 6C), including the embryonic UbxB reporter, which has been shown to possess overlapping Brk and Mad binding sites that Brk and Mad can compete for in vitro(Kirkpatrick et al., 2001; Saller and Bienz, 2001). It is possible that BrkNLS cannot bind to DNA in vivo. However, a modified protein, BrkNLSW, which is identical to BrkNLSapart from the addition of the four amino acids WRPW(Fig. 2D) that recruit the co-repressor Gro (Aronson et al.,1997; Fisher et al.,1996), can repress targets(Fig. 5F, Fig. 6D), indicating that BrkNLS should also be capable of binding to these targets in vivo.
Competition has been proposed as a mechanism for many transcriptional repressors. However, direct in vivo support for or against such proposals is rare, at least of the sort presented here, i.e. testing, in vivo, the ability of a protein consisting largely of a functional DBD, which has access to the nucleus, to repress a target for which there is in vitro evidence for overlapping binding sites with an activator. There is some in vivo evidence that the Drosophila embryonic repressor Kruppel can repress a synthetic enhancer containing overlapping binding sites with the activators Dorsal and Bicoid (Nibu et al.,2003). However, although this repression was CtBP-independent, and further studies are required to rule out additional domains outside of the DBD being required (Licht et al.,1990) in a similar fashion to the 3R domain in Brk. The paucity of good examples of binding-site competition in vivo in eukaryotes is in stark contrast to in prokaryotes (Ptashne and Gann, 2001), and raises the question of how common this phenomenon really is in eukaryotes.
Brk possesses at least three independent repression domains/motifs
If Brk cannot repress by competition it must possess repression domains/motifs, and previous studies identified interaction motifs for the co-repressors CtBP and Gro (CiM and GiM)(Hasson et al., 2001; Saller and Bienz, 2001; Zhang et al., 2001). However,repression of at least one Brk target, omb, was previously shown not to require CtBP or Gro (Hasson et al.,2001). This is consistent with our demonstration that the protein produced by the endogenous mutant brkF138, which is truncated before the CiM and GiM, can still repress omb(Fig. 4D). Truncated proteins that lack the CiM and GiM, BrkStop1, BrkEC and BrkA (Fig. 2D), can also repress omb (Fig. 5H), but only if they contain a specific region between the DBD and CiM that has been classified as a third repression domain, 3R(Fig. 2A). Further studies are required to determine if 3R is a true autonomous repression domain, i.e. if it can function outside of Brk, or if it is more specific (for example,antagonizing activators such as Mad), and to determine what its specific properties are (for example, how close do Brk sites have to be to activator sites for 3R to be effective?).
Differential activity of the repression domains
The three repression domains/motifs of Brk are not equivalent(Fig. 2D). Wild-type Brk and proteins containing only a GiM, BrkNLSW, or only a CiM,BrkStop1NAC, can repress both sal and omb, and they are more effective at repressing sal than omb(Fig. 5A,B,F,G). Analysis of gro and CtBP single and double mutant clones revealed that Gro is required for normal repression of sal in wing discs, and that CtBP can provide some, but not always complete, activity for the repression of sal in the absence of Gro (Fig. 4F-H). By contrast, Gro and CtBP are not required for repression of omb (Hasson et al.,2001).
The 3R domain is sufficient for Brk to repress omb(Fig. 4D, Fig. 5H) and the UbxB enhancer in embryos (Fig. 6E), but is deficient for the repression of sal(Fig. 4D). Furthermore,misexpression of proteins possessing only the 3R domain (plus the DBD) are much more effective at repressing omb than sal, i.e. the converse of wild-type Brk or Brk possessing only a GiM or a CiM(Fig. 5I). Although some results suggested that 3R may confer a limited ability to repress sal(Fig. 5H), this is probably indirect, because a previous study (del Alamo Rodriguez et al., 2004), and our own (not shown),demonstrated that sal requires Omb to be expressed, and if omb is repressed directly, sal will be lost also. However,we cannot completely rule out the possibility that high levels of proteins possessing only the 3R domain can repress sal directly.
Contradictory results were obtained regarding the ability of 3R to repress the vg-QE. Expression of the vg-QE did show expansion in some brkF138 clones (Fig. 4E), indicating that the truncated protein produced in this mutant(which only has the 3R repression domain) cannot efficiently repress this enhancer. However, similar in vitro truncated proteins, such as BrkStop1, could efficiently repress vg-QE expression when misexpressed using the UAS/Gal4 system(Fig. 5L). Such a difference could simply be a reflection of the high levels of expression achieved with the Gal4/UAS system, and that, at physiological levels, the 3R domain is not sufficient for complete repression of the vg-QE.
Whether a single repression domain is sufficient for Brk to repress a particular target may depend upon the positioning of Brk sites in relation to activator sites (or possibly the promoter) at that target. The UbxB reporter has overlapping Brk and activator (Mad) sites(Kirkpatrick et al., 2001; Saller and Bienz, 2001). Analysis of an omb enhancer revealed that an important Brk site may also overlap with an activator(Sivasankaran et al., 2000). Conversely, analysis of the cis-regulatory elements of the sal gene indicate that activator and Brk sites are separated(Barrio and de Celis, 2004). Proteins possessing only 3R can repress UbxB and omb, but not sal, suggesting that 3R may only be sufficient for the repression of genes in which the Brk sites are situated very close to activator sites.
Multiple repression domains
Why does Brk possess at least three, probably four, independent repression domains/motifs? There are two obvious answers: qualitative, different repression domains/motifs are required for the repression of different targets; and quantitative, more domains/motifs provide greater repressor activity. Other transcription factors have multiple repression domains and there is evidence that they have these for either qualitative or quantitative reasons, and, in some cases, both. For example, in the Drosophilaembryo, the pair-rule protein Runt requires Gro for the repression of one stripe of the pair-rule genes, even skipped (eve) and hairy, but not for the repression of engrailed(Aronson et al., 1997). The gap protein Knirps represses different stripes of eve; for stripes 4 and 6 it requires CtBP, but for stripes 3 and 7, it does not. However, this appears to be a quantitative difference, because increasing the levels of Knirps allows it to repress stripes 4 and 6 even in the absence of CtBP(Struffi et al., 2004). Similarly, Gro appears to increase the repressor activity of the Eve protein(Kobayashi et al., 2001).
As discussed above, there is some difference in the ability of the three domains/motifs in Brk to repress different targets. For example, the 3R domain is sufficient for the normal repression of omb but not sal. However, either the CiM or GiM appear to be sufficient for the repression of both sal and omb (Fig. 2D, Fig. 5F,G), so why does Brk need the 3R domain? In the absence of Gro and CtBP, the Brk protein appears fully active in its ability to repress omb, and recruiting Gro and CtBP does not seem to increase its activity towards omb; otherwise, the width of the omb domain would be expected to shift in brkF138 mutant cells, which have no CiM or GiM, or in CtBP gro double mutant cells, but it does not(Hasson et al., 2001). It is possible that, in regard to omb, the 3R domain is more efficient than either of the other two and provides Brk with sufficient activity to establish the omb domain in the correct position.
Brk needs to recruit either CtBP or Gro for the repression of some targets,including sal (Fig. 3E-G) and brk itself(Hasson et al., 2001), or just Gro for some others, including the vg-QE(Hasson et al., 2001). Consequently, why does Brk need to recruit CtBP? Mutation of the CiM alone, in common with mutation or deletion of just the GiM and 3R, does reduce activity of Brk, as judged by its effect when misexpressed(Fig. 2D). However, there is no evidence that CtBP is required specifically for the repression of any Brk target in the wing, because CtBP mutant clones have no effect on the expression of any known Brk target in the wing(Fig. 4G)(Hasson et al., 2001). The CtBP and Gro motifs in Brk have been conserved over millions of years(Fig. 1), and thus, recruiting CtBP is presumably important for Brk activity. It is possible that CtBP is required outside of the wing – for example in the embryo(Hasson et al., 2001) –or for some other, as yet, uncharacterized targets in the wing.
Recruiting both CtBP and Gro does appear to be a little illogical from what is known about their basic properties, CtBP acting only over a short range,while Gro acts over much longer ranges. It might be assumed that different transcription factors would use either Gro or CtBP(Zhang and Levine, 1999),because the primary advantage of recruiting CtBP is that it would allow a transcription factor to repress one enhancer without disrupting the activity of one nearby, which would be repressed if Gro was recruited, although this simple model does not always hold (Nibu et al., 2001). Consequently, most transcription factors do recruit only one of these co-repressors. However, there are two other exceptions,Hairy and Hairless. In Hairy it appears that CtBP may actually be functioning to antagonize Gro activity and not in its standard role as a co-repressor(Phippen et al., 2000; Zhang and Levine, 1999). There is no evidence that it does this in Brk, where it can provide repressor activity. For Hairless, there is genetic evidence that both CtBP and Gro provide repressor activity to the protein(Barolo and Posakony, 2002),although it is not clear if CtBP is required to increase the general activity of Hairless, or for repression of specific targets that cannot be repressed adequately by Gro.
With the exception of the brkF138 mutant, our analysis has been limited to analyzing the effects of misexpressing modified Brk proteins in positions where the endogenous protein is not found. Consequently,further insights into the precise roles of individual repression domains will require replacing the endogenous gene with one in which only one or two repressions domain have been mutated or deleted.
Thresholds
To conclude, it is often assumed that the sensitivity of one enhancer to a transcription factor compared with that of another enhancer is based largely upon the number or the affinity of the binding sites for that transcription factor in each enhancer. However, other factors are also important; for example, the ability of the Giant transcription factor to repress a promoter is related to how closely it binds (Hewitt et al., 1999). Here, we have shown that the two best characterized outputs of the Dpp morphogen gradient, sal and omb, appear to be regulated differently by Brk. Consequently, simply counting binding sites and measuring their affinity will not reveal why one is more sensitive to Brk than the other, and we need to factor in what specific repressive mechanisms are being used, and the relative efficiencies of each.
Acknowledgements
We thank J. Cochran for preliminary studies on this project. We thank K. Arnold, K. Price, A. Bodnar and E. Hitchcock for technical assistance. We thank the following people for materials used in this study: S. Barolo, K. Basler, M. Bienz, E. Bier, S. Cohen, O. Cook, P. Hasson, U. Lammel, Kathy Matthews, Kevin Cook and the Bloomington Stock Center, B. Mollereau, B. Muller, Z. Paroush, J. Posakony, R. Schuh, G. Struhl and A. Tomlinson. We thank D. Chapman, V. Twombly and B. Stronach for comments on the manuscript. This work was supported by a National Institutes of Health grant, GM60368, to G.C.