Engrailed cooperates with extradenticle and homothoraxto repress target genes in Drosophila

The engrailed (en) gene is a homeobox-containing gene crucial for the establishment and maintenance of compartment boundaries duringDrosophila development (Akam,1987; Blair, 1995). Its homologs in mammals are important in brain and limb development, where they might provide a similar function(Augustine et al., 1995;Joyner, 1996;Wurst et al., 1994). The En gene product regulates several key target genes. It helps to maintain the activities of the hedgehog gene and of its own gene(Heemskerk et al., 1991), and it represses the genes wingless (wg)(Heemskerk et al., 1991) andcubitus interruptus (ci)(Eaton and Kornberg, 1990),all of which are involved in the functions of compartment boundaries and the signals they provide to developing tissues(DiNardo et al., 1994).

En carries out these functions, at least in part, by repressing downstream target genes (Smith and Jaynes,1996; Tolkunova et al.,1998). It has two well-characterized repression domains, one of which binds the Groucho co-repressor complex(Han and Manley, 1993;Jaynes and O'Farrell, 1988;Jaynes and O'Farrell, 1991;Jimenez et al., 1997;Tolkunova et al., 1998). Like many eukaryotic DNA-binding proteins, the specificity of DNA binding by En alone appears to be lower than that required for highly selective interaction with specific target genes. This suggests that, like many other DNA binding proteins, En interacts with cofactors that increase its binding specificity. En has been shown to be capable in vitro of binding cooperatively to DNA with the Extradenticle protein (Peltenburg and Murre, 1996), although a role for this interaction in vivo has not been demonstrated.

The extradenticle (exd) gene is also a homeobox-containing gene that was initially characterized as a mutation causing multiple homeotic transformations in Drosophila, without affecting the expression patterns of the homeotic genes. These observations suggested that it might serve as a cofactor for Hox gene products(Peifer and Wieschaus, 1990). Homeotic transformations are seen in zygotic mutants but, when the maternal contribution to exd function is also removed, embryos show alterations that suggest a loss of en function, including a loss ofen gene expression, at later embryonic stages(Peifer and Wieschaus, 1990;Rieckhof et al., 1997). Molecular studies have confirmed the role of Exd as a cofactor for Hox proteins, including a function in altering their specificity of binding to DNA(reviewed in Mann and Chan,1996). This function appears to be conserved in the mammalian homologs of Exd, the Pbx proteins(Peltenburg and Murre, 1997;Phelan et al., 1995).

A third homeobox gene, Meis1, was discovered as an ecotropic retroviral insertion site in mice (Moskow et al., 1995), where its overexpression in conjunction with a subset of Hox proteins leads to leukemic transformation of hematopoietic stem cells (Nakamura et al., 1996).Meis1 is a mammalian homolog of the Drosophila homothoraxgene (hth), which has been shown to regulate the nuclear localization of Exd (Rieckhof et al., 1997)and to form complexes with Hox proteins, including co-complexes with Exd(Ryoo et al., 1999). As with the cofactor functions of Exd, the interactions and functions of the Meis1 family appear to be largely conserved between Drosophila and mammals(Jacobs et al., 1999;Liu et al., 2001;Saleh et al., 2000;Shanmugam et al., 1999;Shen et al., 1999). In mammals, a Hox-class (Antp-class) protein that is part of an endodermally expressed ParaHox cluster (Brooke et al.,1998; Leonard et al.,1993) has also been shown to interact functionally with mammalian homologs of Exd and Hth (Swift et al.,1998).

In this work, we show that En interacts specifically with Exd and Hth in vitro and in cultured cells, that Exd and En mediate co-operative repression in cultured cells, and that the genetic functions of both exd andhth are crucial to the repression activity of En in embryos. We identify sloppy paired (slp) as a direct target gene of En,and show that exd and hth are required for En effectively to repress both slp and a second key target gene, wg, in developing embryos. The involvement of exd and hth function in direct repression by En, in conjunction with the observed molecular interactions, suggest that they are likely to participate in En repression complexes in vivo.

Yeast two-hybrid assays were performed as previously described(Tolkunova et al., 1998),using a lacZ reporter gene. Briefly, strain Y190 was co-transformed with a pAS2 and a pACT2 derivative, and lacZ activity assays were performed on colonies streaked and grown on filters. The experiments shown inFig. 5A used pACT2-Exd, in which the complete Exd open reading frame (ORF) was inserted,pACT2-Hth, containing the complete Hth ORF, pACT2-Meis1b and pAS2-Meis1b (Steelman et al.,1997), and pAS2-En (Tolkunova et al., 1998). The p53 controls used pAS2-p53, which contains the mouse p53 ORF. Details of plasmid construction are available on request.

Glutathione-S-transferase (GST) pull-down assays were performed as described previously (Tolkunova et al.,1998). GST-En contains En amino acids 1-544 in pGEX-5X-1;full-length Exd was transcribed from pET15b-Exd, which contains the complete Exd ORF. Full-length Meis1b was expressed from pET28b-Meis1b (Steelman et al.,1997).

Transfections of Schneider line 2 (S2) cells were performed as described previously (Jaynes and O'Farrell,1991). Plasmids used for Fig. 5C were pRM-HA3-hth (encoding full-length Hth, 3 μg per 60 mm plate), pAc-en (Jaynes and O'Farrell,1988) (8 μg per plate), pP_Ac-His₆Exd(encoding full-length Exd tagged at its N-terminus, driven by theactin5C promoter, 8 μg per plate) and pCaSpeR-hs as filler (to 19μg total DNA per plate). Nuclear extracts for co-immunoprecipitation (coIP)experiments were performed as described by Han and Manley(1993). For coIPs, antibodies were immobilized on either Protein-A/agarose or Protein-G/sepharose in Buffer A (20 mM HEPES pH 7.9, 150 mM NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 1 mM DTT, 0.5 mM PMSF, 10% glycerol), blocked with non-transfected nuclear extract for 4 hours, incubated with transfected nuclear extract overnight, washed extensively at 23°C with Buffer A plus 0.2% Triton X-100, eluted in Laemmli buffer, and analyzed by western blotting essentially as described by Han and Manley (Han and Manley,1993), except that specific proteins were visualized using the ECL detection system (Amersham). All incubations were at 4°C unless otherwise indicated.

Plasmids used for Fig. 6were: (1) pT₃CM₆-CAT, a chloramphenicol acetyltransferase (CAT)-expressing plasmid that can be activated by glucocorticoid receptor (GR) and is derived from pT3N6D-33CAT(Jaynes and O'Farrell, 1991)by replacing the N6 homeodomain binding sites with a tandem array of six composite En-Exd binding sites (Peltenburg and Murre, 1996) or by mutated versions of this site (as indicated in the text and in the legend for Fig. 6); (2) pTAT₃-CAT(Jaynes and O'Farrell, 1991),the equivalent plasmid without En-Exd sites; (3) pRK232(Heemskerk et al., 1991), an En expression construct; (4) pRM-HA3-hth, a metallothionein-promoter-driven Hth expression construct; and (5) pLac82SU(Dorsett et al., 1989) as a reference. The metallothionein promoter was induced by adding CuCl₂to 0.7 mM 24 h after transfection; the GR was activated by adding triamcinolone acetonide to 10^-7 M 48 hours after transfection.Fig. 6A used 1 μg each of pRK232 and either pTAT₃-CAT or pT₃CM₆-CAT, as indicated (without or with En-Exd sites, or with mutated sites), 0.5 μg pRM-HA3-hth, and 0.5 ng pLac82SU per 60 mm culture dish.Fig. 6B used 1 μg each of pT₃CM₆-CAT and pRK232, 0.04 μg pP_Ac-GR(Jaynes and O'Farrell, 1991),and 0.5 ng pLac82SU per 60mm culture dish, and either 0.5 (`low') or 1 μg(`hi') of pRM-HA3-hth where indicated in the figure. Total DNA was normalized to 6 μg per dish in both Fig. 6A and Fig. 6B.Fig. 6C used 0.36 μg pT₃CM₆-CAT, 0.5 μg pRM-HA3-hth, 0.014 μg pP_Ac-GR and 0.18 ng pLac82SU per 35 mm culture dish, and either 0.108 (`low') or 0.36 μg (`high') pRK232, as indicated in the figure. Total DNA was normalized to 2.16 μg per dish. Three hours prior to transfection,each dish was treated as described by Clemens et al.(Clemens et al., 2000) with 24μg of double-stranded RNA (dsRNA) encoding either Rad9 or Exd.

In situ hybridization was performed as described(Tautz and Pfeifle, 1989). Antibody staining with anti-lacZ, anti-En, anti-Exd(Rieckhof et al., 1997) and female-specific anti-Sex-lethal antibodies(Bopp et al., 1991) was performed as described previously(Kobayashi et al., 2001).

To generate hth mutants with hs-En(Heemskerk et al., 1991),P[hs-En E3D]; hth^P2/TM3 Sb ftz-lacZ strains were self-crossed; as controls, P[hs-En E3D]; TM3 Sb ftz-lacZ/+strains were self-crossed.

To generate exd maternal mutant embryos carrying hs-En,exd¹ FRT 18D/FM7-B¹ females were crossed withOVO^D2 FRT 18D; hs-Flp1 males to generateexd¹ FRT 18D/OVO^D2 FRT 18D; hs-Flp1/+females (non-B), which were heat shocked during the wandering larval stage (1 hour at 37°C every 12 hours) to induce mitotic recombination. These females with homozygous mutant germ-line clones were then crossed with either P[hs-En E5] (on the third chromosome) or wild-type (control)males. The exd¹ allele (a.k.a.exd^XP11) is amorphic(Rauskolb et al., 1995).

To generate exd maternal mutant embryos carrying UAS-en(Tabata et al., 1995),exd¹ FRT 18D/FM7-B¹;; UAS-en females were crossed with OVO^D2 FRT 18D; hs-Flp1;UAS-en males to generate exd¹ FRT 18D/OVO^D2FRT 18D; hs-Flp1/+; UAS-en females (non-B),which were heat shocked as above to induce mitotic recombination. These females were then crossed with prd-Gal4/TM3 Sb hb-lacZ males.

Cuticle preparations were performed as described previously(Fujioka et al., 1999). ForFig. 4, embryos were collected and aged for 22 hours at 25°C. Cuticles were prepared and categorized by severity of defects as follows. The number of abdominal denticle bands fused was determined in at least 120 embryos (for each data point) of the genotypes indicated in Fig. 4 and its legend. The categories, from least to most severe, had the following number of denticle bands fused (two bands completely fused together were counted as two bands fused, partial fusions were counted as one band fused): 1-3, 4, 5, 6,7-8.

In the absence of exd or hth function, expression of the endogenous en gene in embryos is unstable and is progressively lost during the middle and later stages of development(Peifer and Wieschaus, 1990;Rieckhof et al., 1997). Therefore, although there is loss of en activity in the absence ofexd, this might be secondary to the loss of en expression. To circumvent this ambiguity and to test whether exd and hthare required for the function of En in vivo, we expressed En from transgenes that are independent of exd function. We assayed en activity by examining expression of its target genes, and by examining the consequences for the pattern of embryonic cuticular structures, which is a sensitive function of en activity.

First, we expressed En ubiquitously using a heat shock promoter-driven transgene. This allowed us to focus on a direct target gene of En, based on the immediate response to ectopic En expression. We examined the responses ofslp and wg, which are repressed by en, and of theen gene, which is ectopically activated by ubiquitously expressed En(Heemskerk et al., 1991). The gene that responded most rapidly to En was slp, which was noticeably repressed relative to control heat-shocked embryos within 7 minutes of a 5 minute heat pulse (data not shown). Such a rapid response makes it very likely that the response is direct (Manoukian and Krause, 1992; Saulier-Le Drean et al., 1998), and strongly suggests that slp is a direct target of repression by En.

To determine whether repression of slp by En was affected by a loss of hth or exd function, we examined the effect of the strong hth allele hth^P2(Kurant et al., 1998) in combination with a hs-En transgene. As shown inFig. 1A-E, the loss ofhth function clearly reduced the repression of slp by En. This repression of slp expression persisted to later stages of embryogenesis (data not shown), although the degree of persistence varied from embryo to embryo because of the relatively mild induction of ectopic En used here. This mild induction resulted in a level of expression comparable to or less than that in the endogenous En stripes. We also tested whether loss ofhth function affected the expression of En from the hs-Entransgene by staining embryos with En-specific antiserum. We detected no difference in the strength or persistence of the ectopic En signal between wild-type and hth mutant embryos (data not shown), showing that the repression activity of En is specifically affected in the hthmutants.

Although repression of slp was strongly reduced in hthmutants, it was not abolished. With longer inductions of hs-Enexpression, slp was more effectively repressed, again consistent with a residual activity in hth mutants (data not shown). A possible explanation for this residual activity is that the hth^P2allele is not a complete null. Alternatively, En might retain some repression activity even in the absence of Hth.

We used the same approach to test whether En activity in embryos requires Exd, using a null allele of exd. As shown inFig. 1F-H, complete removal of both maternal and zygotic exd function caused a dramatic drop in En repression activity on the slp gene. However, as with hth,there appeared to be residual activity even in the absence of exdfunction. Thus, although En retains a residual repression activity in vivo without Hth and Exd, its effectiveness is severely reduced in the absence of either.

To examine further the requirements for exd and hth by En, we expressed En ectopically using the Gal4-UAS system(Brand and Perrimon, 1993), in a pattern that partially overlaps with the expression of two En target genes,wg and slp. This patterned ectopic expression provides an internal control within individual embryos of unrepressed target gene expression, as well as avoiding the potential complicating effects of heat shock. In this system, a transgene (the driver) expresses the Gal4 activator protein in a pattern, and this drives the expression of a responding transgene, in this case UAS-en. Using the prd-Gal4 driver[made by L. Fasano and C. Desplan (seeYoffe et al., 1995)],UAS-en is activated in a striped pattern(Fig. 2B) that covers every other stripe of both wg and slp in the segmented portion of the embryo (wg is expressed within the slp domain, just anterior to each normal en stripe). The effects of this ectopic En expression were assayed in wild-type and exd mutant embryos. The effects on wg expression are shown inFig. 2. Alternate wgstripes (in even-numbered parasegments) were repressed completely in wild-type embryos (Fig. 2D), whereas repression of these stripes was less complete in embryos that lacked a maternal supply of Exd protein and also had a reduced zygotic gene dose, and appeared to be delayed relative to the wild-type controls(Fig. 2F, and data not shown;notice that some expression remains, particularly in parasegments 0, 2 and 4). In embryos that completely lack exd function (both maternally and zygotically), the situation is complicated by the fact that continuedwg expression requires exd. The weakening of wgstripes in exd null embryos might be secondary to the loss ofen expression, because en is necessary to maintainhedgehog and wg transcription(Heemskerk et al., 1991;Ingham, 1993). Thus, even without ectopic En expression, wg expression is weaker, particularly in odd-numbered parasegments such as 3 and 5(Fig. 2G,I). Nevertheless,against this `background' of weakened odd-numbered wg stripes, the ability of prd-Gal4-driven En (expressed only in even-numbered parasegments) to repress wg was significantly reduced. Although this effect was also seen in more-posterior parasegments, the effect was noticeably stronger anterior to the abdomen (particularly in parasegments 0, 2 and 4),where there appeared to be very little repression of wg compared withexd mutant embryos in which En was not ectopically expressed(Fig. 2G-J).

In similar collections of embryos, we examined the consequences of loss ofexd function on the ability of En to repress slp. As withwg, alternate slp stripes were strongly repressed byprd-Gal4-driven En (Fig. 3A,B). Simultaneously eliminating the maternal contribution and reducing the zygotic dose of exd⁺ caused a delay in repression and an ultimately less-complete reduction of slpexpression (Fig. 3E,F). In embryos completely lacking both maternal and zygotic exd function,repression activity was strongly reduced(Fig. 3G,H), consistent with the results involving ubiquitously expressed En(Fig. 1). As with repression ofwg, the repression of slp by En was more dependent onexd function in the gnathal-thoracic region than in the abdomen(Fig. 3H, parasegments 0, 2 and 4 vs 6, 8, 10 and 12). However, even in the abdomen, repression by En was significantly reduced in the absence of exd(Fig. 3G,H vs A,B).

To test the significance of the changes in the specific En target genes described above for the overall ability of En to regulate its target genes, we examined the extent to which reduced hth and exd function affected the ability of ectopically expressed En to alter the pattern of structures produced at the end of embryogenesis in the external larval cuticle. First, in hth mutant embryos, we examined the severity of cuticle defects produced by hs-En. AsTable 1 shows, the ability ofhs-En to cause severe cuticle defects was reduced in hthheterozygotes and homozygotes, suggesting that the observed reductions in En repression activity on the specific target genes wg and slpaccurately reflects its overall ability to regulate the target genes responsible for cuticle pattern.

We also quantified the severity of cuticle defects caused by ectopically patterned En expression and compared the results between theexd^+/+ and exd^+/- populations, which are both wild type in the absence of hs-En. The analysis(Fig. 4) showed that, in the embryo, the reduction in exd activity caused by simultaneously removing maternal exd function and reducing the zygotic gene dose had a significant effect on the ability of prd-Gal4-driven En expression to disrupt the proper development of abdominal cuticular structures. This confirms the developmental impact of the changes in target gene expression characterized above. These results also confirm that En has a significant requirement for exd in the abdomen, as well as in more anterior regions.

As a first step in determining the mechanisms whereby exd andhth contribute to repression by En in vivo, we examined the possibility of direct interaction. Previous studies had shown that En can bind co-operatively with Exd in vitro to artificial DNA sites(Peltenburg and Murre, 1996). We tested whether a direct En-Exd interaction could also occur in other contexts, using yeast two-hybrid and in vitro assays. We also tested whether En could interact similarly with Hth. Fig. 5A shows the results of two-hybrid assays in yeast. En appeared to interact robustly with Exd in this system, because the signal strength observed with both isolated colonies (data not shown) and colony streaks(Fig. 5A) was consistently higher than that seen with some of our positive controls, including the functionally important interaction between En and Groucho (data not shown)(Tolkunova et al., 1998). This signal was also comparable to that seen with Exd and the mouse homolog of Hth,Meis1 (Fig. 5A; Hth fused with the Gal4 DNA-binding domain gave a high background signal, so that parallel results using Hth with Exd were uninformative). En also gave a somewhat weaker, but apparently specific, signal in combination with either Hth or Meis1 (Fig. 5A).

In vitro, En also interacts specifically with both Exd and Meis1(Fig. 5B). Here, En fused with GST, but not GST alone, effectively pulls down either Exd or Meis1. Meis1 was used in these studies because of the high level of non-specific interaction observed with in-vitro-translated Hth, perhaps owing to the heterologous nature of the translation system. In this system, it is unlikely that the interactions are due to co-operative binding to DNA, and we interpret these results to mean that these interactions can occur in solution. Furthermore,Meis1 appears to interact more strongly with En in the presence of Exd,suggesting that the three proteins form a co-complex.

We tested whether these molecular interactions can also occur in culturedDrosophila cells (S2 cells). We transfected these cells with a Hth expression plasmid, which induces nuclear localization of endogenously expressed Exd (Rieckhof et al.,1997), either alone or with expression plasmids for En and for a tagged form of Exd (His₆Exd, see Materials and Methods). In nuclear extracts from these cultures, anti-Hth antiserum specifically precipitated both En and endogenous Exd, as well as the tagged form of Exd(Fig. 5C). Conversely,anti-His₆ antibodies specifically precipitated both En and Hth(Fig. 5C). This ability of either Hth or His₆Exd to mediate precipitation of En occurs at a salt concentration of 150 mM and with 0.2% Triton (see Materials and Methods),suggesting that the complexes involved are reasonably stable. However, the interaction of Hth with Exd, which is known to be functionally important,might be somewhat more stable, because the background seen under conditions needed to generate a readily detectable signal is somewhat lower between Hth and endogenous Exd, and between His₆Exd and transfected Hth, than that seen with En (notice the slight background visible in the `—antibody' lane at the positions of En and tag-Exd inFig. 5C, and the lack of any background at the positions of endogenous Exd and Hth). Although the interactions seen with En might be weaker than those between Exd and Hth, they are nonetheless strong enough to suggest the direct involvement of Exd and Hth in repression by En in vivo.

As a first step in assessing the influence of the physical interactions identified above on the transcriptional activity of En, we used a previously constructed co-operative binding site for En and Exd(Peltenburg and Murre, 1996)as a target site in transfection assays. Repression by En has been extensively studied in Drosophila S2 cells, which have been shown to express Exd,but not Hth, constitutively (Rieckhof et al., 1997). Co-operative binding sites were inserted into a reporter vector previously shown to be unresponsive to En in the absence of inserted sites (Jaynes and O'Farrell,1991) and their responses to En and to En in combination with Hth were measured. Consistent with previous results, En was able to repress this artificial target gene only when it contained the inserted sites and, without these sites, neither Hth alone nor En plus Hth caused any repression(Fig. 6A). Importantly, with the binding-site-containing reporter, the presence of Hth significantly increased the effectiveness of repression by En, either on the basal expression level of the target gene (Fig. 6A) or, in an established assay for active repression, when this gene was activated from separate activator binding sites(Fig. 6B). Because Hth has been shown to induce the nuclear localization of endogenously expressed Exd in these cells (Abu-Shaar et al.,1999), this might be due to an increased nuclear concentration of En-Exd complexes, to a Hth-En interaction or to a combination of the two.

In order to test whether the increase in repression by En caused by expression of Hth was mediated by cooperative binding between En and Exd, we tested mutated versions of the cooperative binding site. A two-nucleotide change in an Exd core consensus binding sequence (TGAT) within the 30-nucleotide oligonucleotide has previously been shown to abolish co-operative binding by En and Exd (van Dijk, 1994). As shown inFig. 6, this Exd-site-mutated target gene still responded to En, but the increased repression caused by Hth was largely absent. Although there was no apparent increase in repression of basal transcription caused by Hth with this mutated target gene, some residual increase in repression of activated transcription appeared to persist. This was true even when the other core Exd consensus binding sequences (ATCA)within the oligonucleotide were also mutated (data not shown). Thus,eliminating co-operative binding between Exd and En significantly reduced the cooperative repression caused by Hth expression, suggesting that this effect of Hth is mediated at least in part by co-operative binding between Exd and En, whereas a secondary effect of Hth expression on repression of activated transcription might be independent of such co-operative binding. Consistent with previous results (Jaynes and O'Farrell, 1991), when the En core consensus sequence ATTA was mutated to AGGA, repression by En, or by En plus Hth, was eliminated (data not shown).

To test the influence of Exd more directly, we turned to RNA interference(Clemens et al., 2000) to reduce endogenous Exd levels in S2 cells. As shown inFig. 6C, treatment of S2 cells with Exd dsRNA reduced En-mediated repression from the co-operative binding sites significantly, relative to non-specific dsRNA. This effect was not seen at higher concentrations of En, possibly owing to saturating levels of En. However, the increase in repression by Hth was clearly reduced by Exd dsRNA at both high and low En concentrations, suggesting that Hth and Exd cooperate to increase the repression activity of En.

Although these results might be accounted for by a combination of Hth-induced nuclear localization of Exd and co-operative binding by Exd and En to the target gene, the observed ability of En to form complexes with both Hth and Exd suggest the simple model that repression by En alone is less effective than repression by a trimolecular complex of En, Exd and Hth.

We show here that En, a non-Hox homeodomain protein with multiple roles in development, is dependent on the functions of the Hox cofactors exdand hth to repress target genes in vivo. Molecular studies demonstrate the ability of En to form complexes with Hth and Exd in vitro, in yeast and in Drosophila cells, leading to the model that complexes among these genetic partners carry out repression in vivo. This model is supported by the ability of Exd and Hth to facilitate repression by En from co-operative binding sites in co-transfection assays. Previously, Hth and Exd were known to act as cofactors only for Hox proteins. This work therefore sheds new light on how En might achieve target specificity in vivo, and expands the diversity of transcriptional regulators that require hthand exd function. Below, we discuss the evidence supporting this model, as well as its implications.

En can form complexes with both Exd and Hth (as well as its mammalian homolog Meis1) in vitro, in yeast and in cultured cells(Fig. 5). Interestingly, rather than the two-way interactions being mutually exclusive, En appears to be able to interact simultaneously with Exd and Meis1 in vitro(Fig. 5B). In extracts from transfected cultures, En is specifically precipitated by antisera against both Exd and Hth (Fig. 5C). Therefore, En might interact with these cofactors similarly to Hox proteins,which have been shown to form such three-way complexes(Ferretti et al., 2000;Jacobs et al., 1999;Ryoo et al., 1999).

In cultured Drosophila cells, Exd and Hth cooperate with En to repress transcription. Using a co-operative binding site for Exd and En(Peltenburg and Murre, 1996)to construct an En-responsive target gene, we found that both Exd and Hth are required for full repression activity (Fig. 6). When a mutation was introduced into an Exd consensus binding sequence that eliminates co-operative binding, co-operative repression was largely eliminated, whereas mutating the En consensus binding sequence eliminated repression. This, along with the fact that RNA interference directed against Exd mRNA also largely eliminated co-operative repression(Fig. 6C), suggests that a complex containing Exd and En is responsible for the co-operative repression caused by coexpression of Hth and En (Exd is constitutively expressed in these cells). Because Hth regulates the nuclear localization of Exd, it can allow Exd-En repression complexes to form in the nucleus. In addition, the observed molecular interactions suggest that the fully active repression complex might include all three proteins.

Loss of exd function has been shown to result in a loss ofen expression at later embryonic stages(Rieckhof et al., 1997). Because en function is required to maintain its own expression(Heemskerk et al., 1991), the loss of en expression could be a downstream effect of a loss ofen function, or it could be due to some other consequence of the lack of exd. This ambiguity concerning the role of exd inen function led us to investigate whether the activities of ectopically expressed En (Heemskerk et al., 1991) are dependent on exd function. We expressed En ectopically in two ways: from a heat-shock promoter and using a patterned Gal4`driver' transgene. An advantage of the former approach is that one can often distinguish between immediate and secondary downstream effects based on how rapidly they occur following heat induction. Advantages of the second approach include having normal and altered expression in parts of the same embryo,providing a rigorous internal control. Both of these approaches led to similar conclusions (Figs1,2,3,4),that exd function is important for the repression by En of its direct target gene slp, that wg also shows a strong dependence onexd function for its repression by En and that the ability of En to alter the pattern of embryonic cuticles is sensitive to the gene dosage ofexd. Further, in each set of experiments, the observed dependence of repression on exd was accompanied by a residual repression activity when exd function was removed both maternally and zygotically. This residual exd-independent repression activity might be due to the ability of En to bind to target sites independently of exd but with a reduced affinity, or it could be accounted for by the existence of two classes of binding sites, one exd dependent and the other exdindependent. This possibility is paralleled by the relationship of Exd with Ubx, which has been shown to function either co-operatively with Exd or alone on multiple binding sites in target genes(Galant et al., 2002). Alternatively, exd might be exerting an indirect effect on repression by En. However, because Exd forms complexes with En in yeast and in vitro, and because it appears to facilitate repression by En directly in cultured cells,it seems likely that the dependence of En on exd function in vivo is due at least in part to the direct action of En-Exd complexes. Confirmation of this model will require the analysis of specific regulatory sites, which have not yet been identified, in target genes such as slp. If this model is correct then our results suggest that the repression activity of Exd-En complexes might come exclusively from En repression domains(Han and Manley, 1993;Jaynes and O'Farrell, 1991;Jimenez et al., 1997;Smith and Jaynes, 1996;Tolkunova et al., 1998),because Exd has been shown to act as a cofactor in the activation of target genes in vivo in conjunction with Hox proteins (reviewed inMann and Chan, 1996;Grieder et al., 1997;Inbal et al., 2001;Liu et al., 2001;Pinsonneault et al., 1997;Ryoo et al., 1999).

The effects of eliminating exd function on repression by En appear to be different in the abdomen and the more-anterior regions (Figs2,3), in that En is less dependent on exd in the abdomen (parasegments 6-12). Similar results have recently been obtained by Alexandre and Vincent(Alexandre and Vincent, 2003),as described in their accompanying paper. One possible explanation is thathth can provide the observed exd-independent activity. However, in exd mutants, Hth levels are reduced, probably because Hth protein is less stable without Exd(Abu-Shaar et al., 1999). Nevertheless, our data are consistent with the possibility that, on their own,either Hth or Exd might provide partial cofactor activity, whereas both together might be required for full activity. The latter possibility is suggested by our observation that maximal repression activity in S2 cells requires all three gene products.

An additional possibility to account for the residual exd- andhth-independent repression activity of En in the abdomen is that other cofactors assist En in binding to its target genes in the abdomen. If there are other cofactors at work, it is likely that their activity (or expression) is dependent, either directly or indirectly, on the Hox genesUbx and abd-A, because these genes are responsible for all known aspects of differential segment identity in this region of the embryo. This expectation has been directly confirmed by Alexandre and Vincent(Alexandre and Vincent,2003).

It is noteworthy that the difference in the dependence of En onexd in the abdomen versus the thorax is seen only after stage 9 (for example, it is not seen in Fig. 1), when the levels of Hth, and the consequent nuclear concentration of Exd, have declined in the abdomen(Rieckhof et al., 1997). Thus,the dependence of En on exd parallels the nuclear concentration of Exd, and might reflect an evolutionary adaptation to the changing levels of Exd in different regions of the embryo.

Hth has been shown to act in part through its facilitation of the nuclear localization of Exd, and strong hth and exd mutants have very similar phenotypes (Rieckhof et al.,1997). Although Hth can also interact with En independently of Exd(Fig. 5), transfection assays in cultured cells suggest that Hth might depend entirely on Exd for its ability to increase repression by En, at least from artificial En-Exd co-operative binding sites (Fig. 6C). Because Hth forms complexes with En in these cells(Fig. 5C), in addition to increasing its repression activity (Fig. 6A,B), a simple model is that maximal repression activity is due to complexes containing En, Exd and Hth. However, we cannot rule out the possibility that Hth acts solely by making Exd available to interact with En on target sites, through its ability to bring Exd into the nucleus.

We tested whether the repression activity of ectopically expressed En in vivo is dependent on hth function, using assays similar to those used for exd (Fig. 1,Table 1). In each case, we observed a close similarity to results with exd mutants. En activity showed a strong dependence on hth function, although residual activity remained in hth mutants. In addition, En activity showed a sensitivity to the hth gene dose. All of these results are consistent with the effects of Hth being exerted through its effect on Exd nuclear localization, provided that the nuclear targeting of Exd is necessary for its ability to function with En. However, as noted above, Hth might also increase the effectiveness of En repression directly, by forming complexes with En and/or as part of En-Exd complexes. A detailed analysis of a number of in vivo target sites will be necessary to distinguish among these possibilities.

Exd and Hth are essential to the correct regulation of target genes by the homeodomain proteins of the Hox clusters (reviewed inMann and Affolter, 1998). However, their functional interactions have not previously been shown to extend beyond the highly restricted subset of homeodomain proteins that are found within the Hox clusters (the Antp, Abd-B and Labial classes). The identification of functional interactions with En suggests that exdand hth might provide functional specificity in conjunction with other non-Hox-class homeodomain proteins.

The identification of slp as a direct target gene of En has implications for the mechanism by which En helps to maintain the activity of its own and other genes, including hedgehog, within its domains of expression, the posterior compartments. The slp locus produces two closely related, coordinately regulated gene products (Slp1 and Slp2), which have essentially indistinguishable functions(Cadigan et al., 1994;Grossniklaus et al., 1992). They are forkhead-domain transcription factors(Grossniklaus et al., 1992)that repress en expression(Cadigan et al., 1994), and both contain a conserved motif (homology region II)(Grossniklaus et al., 1992)that is similar to the Groucho-binding domain of En(Tolkunova et al., 1998). Slp1 has also been shown to bind the Groucho co-repressor in vitro(Kobayashi et al., 2001),suggesting that it is a repressor and therefore that its action on theen gene is likely to be direct. Thus, the mechanism of enautoregulation, as well as the ability of En to activate other target genes,is likely to be due, at least in part, to an indirect effect of repression ofslp expression. Similar conclusions have been reached by Alexandre and Vincent (Alexandre and Vincent,2003), as described in the accompanying paper. In addition, En might activate target genes indirectly by repressing other repressors that are also normally excluded from its expression domain, such as Odd-skipped(Mullen and DiNardo, 1995;Saulier-Le Drean et al., 1998)and the repressor form of Cubitus interruptus(Eaton and Kornberg,1990).

Although there have been previous suggestions that Exd and Hth might participate in active repression as well as activation complexes(Abu-Shaar and Mann, 1998;Abzhanov et al., 2001;Manak et al., 1994;White et al., 2000), most of the well-characterized direct Exd-Hth-Hox target genes are activated in anexd- or hth-dependent fashion(Manak et al., 1994;Pinsonneault et al., 1997;Ryoo and Mann, 1999;Ryoo et al., 1999). In fact,these observations raised the question of whether Exd and Hth might be dedicated to gene activation (Pinsonneault et al., 1997). Recently, Hth and Exd have been shown to act directly with Ubx to repress the Hox target gene Distalless in theDrosophila abdomen (Gebelein et al., 2002). The partnership with En in repression further argues that these cofactors can increase the target site discrimination of homeodomain proteins without restricting the resulting transcriptional activity to activation alone. Based on these results, we suggest that Hth and Exd increase the target-site discrimination of several classes of homeodomain proteins and that they do so without defining the transcriptional activity of the resulting protein complex.

We thank J. Montgomery and A. M. Buchberg for Meis1 plasmids, P. Schedl for anti-Sex Lethal antibodies, C. Girdham and P. O'Farrell for anti-En antibodies, G. L. Yusibova, Jian Zhou, and Y. Emi-Sarker for excellent technical assistance, and A. M. Buchberg and A. Mazo for helpful comments on the manuscript. This work was supported by NIH (GM50231) and NSF (0110856)awards to J.B.J. and by NIH and Leukemia and Lymphoma Society of America awards to R.S.M.