Anterior repression of a Drosophila stripe enhancer requires three position-specific mechanisms

In Drosophila, a hierarchy of genetic interactions divides the embryo into fourteen segments along the anterior posterior axis (reviewed byPankratz and Jackle, 1993;Small, 1997). A crucial component of this hierarchy is the pair-rule gene even skipped(eve), which is expressed in a pattern of seven stripes in early blastoderm embryos (Frasch et al.,1987; Macdonald et al.,1986). This pattern is established by five modular enhancers, each of which controls the expression of a single stripe or a pair of stripes(Fig. 1)(Frasch et al., 1987;Fujioka et al., 1999;Goto et al., 1989;Harding et al., 1989;Macdonald et al., 1986). In general, eve stripes are activated by broadly distributed proteins,and stripe borders are set by localized repressors. Each enhancer contains a different combination of activator and repressor binding sites, and consequently a unique patterning activity.

Although all five enhancers are active at the same time in development,they function independently, based in large part on their modularity(Small et al., 1993). Stripe borders are formed by short-range repression mechanisms, which permit different enhancers to be in different states of transcriptional activity within the same nucleus. The best-characterized enhancer directs the expression of eve stripe 2 (eve 2). This stripe is first activated in a broad anterior domain by the anterior morphogens Bicoid (Bcd)and Hunchback (Hb), and then refined to a stripe by repressive mechanisms involving the gap proteins Giant (Gt) and Kruppel (Kr), which form the anterior and posterior borders of the stripe, respectively(Arnosti et al., 1996;Small et al., 1992;Small et al., 1991;Stanojevic et al., 1991). All four proteins bind in vitro to multiple sites within a 480 bp minimal stripe element (MSE; Fig. 1), which is sufficient for stripe expression in vivo. Mutations in individual sites cause changes in the level of expression or the shape of the stripe, suggesting that the enhancer acts as a switch element that directly measures concentrations of activator and repressor proteins.

The model for eve 2 regulation is the result of intensive study,but is incomplete in at least two respects. First, the combined activities of Bcd and Hb do not seem to be sufficient for activation at the position ofeve stripe 2. For example, reporter genes containing three high affinity Bcd sites and three Hb sites cannot respond to the low levels of these proteins at the position of eve stripe 2(Simpson-Brose et al., 1994). In addition, several attempts to construct artificial stripe 2 enhancers using up to ten Bcd and/or Hb sites have been unsuccessful (S. S. and M. Levine,unpublished). These results suggest that cis-regulatory sequences other than the Bcd and Hb sites are important for activation. Such sequences may contain low-affinity Bcd- and Hb-binding sites, or sites for other activator proteins. Alternatively, they may simply function to provide the correct spacing between known activator or repressor sites.

Second, the mechanism(s) that control repression of eve 2 in anterior regions are not well understood. As the activators (Bcd and Hb) are distributed throughout the anterior half of the embryo, repressive mechanisms must exist that prevent activation in all nuclei anterior to the position of the stripe. The gap protein Gt has been suggested as a critical anterior repressor, but there is only a modest expansion of the stripe in embryos lacking gt activity (Small et al., 1992). Furthermore, ectopic expression of Gt does not efficiently repress eve 2 (Nibu and Levine, 2001; Wu et al.,1998). Thus, Gt activity is required for setting the anterior stripe border, but it probably acts by potentiating the activity of another anterior repressor (X). Deleting the Gt-binding sites causes a more extensive expansion than that seen in loss-of-function gt mutants(Small et al., 1992),consistent with the idea that these sites are bound by both Gt and X, which then cooperate in eve 2 repression. As Gt contains a b-ZIP dimerization domain (Capovilla et al.,1992), the simplest model is that Gt and X form a heterodimer that mediates effective repression through these sites. However, the expansion ofeve 2 caused by deleting these sites still does not extend all the way to the anterior tip. This suggests that other unknown mechanism(s) prevent activation in this region.

To identify other activities that regulate eve 2, we tested a series of enhancer deletions in the context of a lacZ reporter gene. These experiments led to the identification of a binding site(GTTT)₄, which is crucial for repression in a more anterior subregion. We show that the forkhead domain (FD) protein Slp1, which is expressed in a broad anterior domain, binds to this site and is sufficient for repression of stripe 2 of the endogenous eve gene. Further genetic experiments identify a separate repression activity near the anterior pole that is dependent on the terminal patterning gene torso. Thus, three position-specific activities are required for anterior repression ofeve 2. Two other eve stripes (eve 1 andeve 3) are also repressed by ectopic Slp1 expression. As the enhancers that control these stripes contain sites similar to(GTTT)₄, it is possible that repression of all three enhancers in this region is controlled by a similar mechanism.

For the initial enhancer analysis, four deletions (D1-D4) were generated by oligo-mediated mutagenesis (Muta-Gene, BioRad Laboratories, Richmond, CA) of a pBS-SK+ clone containing the eve 2 MSE fused upstream of aneve basal promoter fragment. An 800 bp fragment composed of the 500 bp eve 3+7 MSE and a 300bp spacer sequence was inserted upstream of the eve 2 MSE to control for changes in expression levels(Small et al., 1993). Deletion 1 (D1) removes 78 bp between the Gt 1 and Bcd 2 sites. D2 removes 62 bp between the Bcd 2 and Kr 4 sites. D3 removes 41 bp between the Bcd 3 and Gt 2 sites. D4 removes 38 bp between the Gt 2 and the Gt 3 sites. A fifth deletion(D5) removes 87 bp that lies 5′ of the Bcd 5 site (S. S. and D. Arnosti,unpublished). Precise deletion end points were confirmed by sequence analysis.

The deletion and base pair substitution of the (GTTT)₄ site was made in the context of a pBS-SK+ clone containing only the enhancer and theeve basal promoter. The following oligos were used in these experiments:

• eve2delGTTT, 5′ GGCTAATCCCAGCATGCGCCGCAGAAGGATG 3′; and

• eve2mutGTTT, 5′GCTAATCCCAGCAGGTACCTGCGTGCACGCCGCAGAAGG 3′.

Restriction sites used to screen for mutant clones are indicated by underlined sequences. All mutations were confirmed by DNA sequencing, andEcoR1-XhoI fragments containing the mutant enhancers were cloned along with a 4.5 kb XhoI-XbaI eve lacZ fragment into the CaSpeR vector (Pirrotta,1988), which was cut with EcoR1 and XbaI. Constructs were introduced into the Drosophila germline via standard microinjection procedures (Small,2000; Spradling,1986). At least three independent lines were generated for each construct, and analyzed by in situ hybridization using an antisenselacZ RNA probe as previously described(Jiang et al., 1991b).

The mutant alleles used in this study were tor^PM51,gt^YA82, croc⁵⁹, fkh³³³¹,hb^14F21 and slp^Δ34BCyO/ 1(2), which is a small deficiency that makes no slp1 or slp2 mRNA(Grossniklaus et al., 1992). The slp deficiency is on the CyO balancer chromosome, which facilitated identification of flies carrying this mutation. For analysis of transgenes in zygotic mutants, embryos were collected from inter se crosses among flies carrying a single copy of the transgene and single copies of the mutations to be tested. slp, gt double mutant embryos were unambiguously identified by triple staining with anti-sense slp1, eveand lacZ mRNA probes.

Four tandem copies of a 30 bp sequence containing the(GTTT)₄-binding site were inserted into the pBluescript SK+ vector using standard cloning procedures. As a negative control plasmid, four copies of a mutated version of the same sequence were inserted into a parallel construct. The sequences of the oligos used for these constructs are as follows:

• (GTTT)₄, 5′ GATCGCGGCGTTTGTTTGTTTGTTTGCTGG 3′;and

• (Gmut)₄, 5′ GATCGCGGCGTGCACGCAGGTACCTGCTGG 3′.

The integrity of each repeat was assayed by sequence analysis, and fragments containing the intact tetramers were excised asSmaI-XbaI fragments, blunt ended and cloned into the blunt-ended XbaI site of the pHisi-1 vector and into the blunt-endedXhoI site of the placZi vector. These vectors were provided in the Matchmaker One-Hybrid kit from Clontech (Palo Alto, CA).

A yeast strain (YM4271) carrying both the lacZ and His3 reporter constructs was generated using conditions recommended by the manufacturer(Clontech), and transfected with a λACT library of cDNAs made from 0-to 10-hour-old embryos (Yu et al.,1999).

An expression clone containing the full length Slp1 coding region,pET3a/slp1c (Cadigan et al.,1994), was transformed into E. coli BL21 cells and expression was induced with lmM IPTG for 3 hours at 37°C. Cells were harvested by centrifuging at 4200 g for 10 minutes. The bacterial pellet was then resuspended in 2 ml phosphate buffer (20 mM NaPO₄,500 mM NaC1, 1 mM EDTA, 0.1% Tween20, 1×Complete™ Protease Inhibitor Cocktail) followed by sonication three times for 15 seconds on ice. Crude lysates were fractionated by centrifugation at 16,000 g for 30 minutes. The supernatant was mixed 1:1 with 100% glycerol, flash frozen in liquid nitrogen and stored at -80°C. Fraction content was monitored using SDS-PAGE.

Double stranded probes were created by annealing single stranded oligos containing the (GTTT)₄ site and the mutant version of the site. The sequences of the oligos used are as follows:

• GTTT top, 5′ GATCGCGGCGTTTGTTTGTTTGTTTGCTGG 3′;

• GTTT bottom, 5′ GATCCCAGCAAACAAACAAACAAACGCCGC 3′;

• MutGTTT top, 5′ GATCGCGGCGTGCACGCAGGTACCTGCTGG 3′; and

• MutGTTT bottom, 5′ GATCCCAGCAGGTACCTGCGTGCACGCCGC 3′.

Annealed oligos were labeled using the Klenow fragment of E. coliDNA polymerase I and ³²P-dGTP, and purified on a 15% polyacrylamide gel. Gel retardation assays were performed by incubating nuclear extracts from 0-12 hours wild-type embryos (Han et al.,1993) or bacterial extracts with (GTTT)₄ or a mutant version of the probe in 20 μl binding cocktail (50% glycerol, 12.5 mM Hepes pH 7.5, 62.5 mM NaCl, 7 mM β-mercaptoethanol, 3.125 mM MgCl₂,1.25 mg/ml bovine serum albumin and 0.1 mg/ml poly dIdC) at 25°C for 30 minutes and resolving DNA-Protein complexes on a 4% acrylamide, 2.5% glycerol gel containing 1×TBE. Labeled complexes were visualized by autoradiography.

A general transformation vector for ventral misexpression was constructed using a 2.8 kb NsiI-Asp718 snail (sna)promoter fragment (Ip et al.,1992), which includes the native basal promoter and 100 bp of the 5′ UTR. This was fused upstream of a 2.5 kb Asp718 fragment containing an FRT-STOP-FRT cassette, and a double-strandedKpnI-NotI linker that contains a PmeI site for the insertion of any cDNA sequence. These fragments were ligated into the CaSpeR CGB3 vector `PBKN' (a gift from Miki Fujioka), which was cut withPstI and NotI. The slp1 cDNA used here is a 1.6 kbEcoRV-XhoI fragment isolated from the one-hybrid clone. TheEcoRV site is a native site that lies 17 bp upstream of the translation initiation site (Grossniklaus et al., 1992). This fragment includes the entire slp1 ORF and 370 bp of the 3′ UTR. The gt cDNA used here is a 1.6 kb fragment that extends from an artificial NdeI site at the translation initiation codon to an EcoRI site that lies ∼400 bp downstream of the termination codon (Kraut and Levine,1991). These fragments were blunt-ended, and cloned into thePmeI site of the ventral expression vector described above. Transgenic flies carrying these constructs were then generated using standard microinjection techniques (Small,2000; Spradling,1986).

For these experiments, three independent sna-slp1 and fivesna-gt transgenic lines were analyzed. To activate mis-expression, we generated males containing a given misexpression transgene along with aβ2-tubulin-FLP transgene, which activates expression of the yeast FLP recombinase and recombination between the FRTs during spermatogenesis (Struhl et al.,1993). These males were then mated with w¹¹¹⁸females or with females carrying various lacZ reporter genes. The reporter genes used here are eve2-lacZ(Small et al., 1992),eve3+7-lacZ (Small et al., 1996), -7.8eve-lacZ (D. Kosman and S. S.,unpublished) and eve1+5-lacZ(Fujioka et al., 1999). All reporter genes contain the proximal eve basal promoter and 5′untranslated region. Embryos were collected and analyzed by in situ hybridization (Kosman and Small,1997). All misexpression lines gave similar results in the assays described here.

Previous DNA-binding experiments identified twelve sites for the genetically defined regulators within the eve 2 MSE(Small et al., 1991;Stanojevic et al., 1989). To identify other sequences important for eve 2 regulation, we constructed a series of mutant enhancers that contain deletions (D1-D5) of the regions between the known binding sites(Fig. 1). Each deletion was tested separately in the context of an eve 2 lacZ fusion gene that also contains the eve 3+7 MSE as an internal control for levels of expression. Several independent lines for each construct were obtained by P-element mediated transformation. Embryos were collected from these lines, and examined by in situ hybridization for expression oflacZ mRNA.

All five deletions disrupt the normal function of the eve 2 enhancer. Four (D1, D3, D4, and D5) cause a significant reduction in the level of stripe activation (Fig. 2;data not shown). It is not clear whether these reductions are caused by removing discrete activator sites or by changing the spacing between the known sites. By contrast, the D2 deletion, which removes a 62 bp sequence between the Bcd 2 and the Kr 4 sites (Fig. 1), causes an apparent strengthening of stripe 2 activation, and ectopic expression in more anterior regions(Fig. 2E,F). The derepression is quite broad early in nuclear cycle 14(Fig. 2E), but refines later to an ectopic anterior stripe (Fig. 2F). These results suggest that the D2 region contains sequences required for repression in the region of the ectopic stripe.

Because important functional binding sites are likely to be evolutionarily conserved, we compared the eve 2 sequence from D. melanogaster with published sequences from four other Drosophilaspecies (Fig. 3A)(Ludwig et al., 1998;Sackerson, 1995). The Kr4 and Bcd2 sites are well-conserved among all five species, and thus represent excellent anchors for the careful analysis of the intervening region. The best-conserved sequence block in this region is a 16 bp sequence that consists of four repeats of the sequence GTTT. Although this type of repeat is unusual for a functional binding site, we tested it by deletion or mutagenesis in the context of an eve 2-lacZ reporter gene. Both disruptions cause severe anterior derepressions (Fig. 4C-F). By contrast, a deletion that removes the rest of the D2 sequence (46 bp), but leaves the (GTTT)₄ sequence intact, does not cause any detectable change in enhancer activity (data not shown). Thus, the(GTTT)₄ sequence is the major binding site for a repressive activity that prevents expression of the eve 2 enhancer in a specific anterior region.

Previous experiments suggested that the gap gene gt and the Gt-binding sites are required for the correct positioning of the anterioreve 2 border (Small et al.,1992). To test the relationship between the(GTTT)₄-binding activity and Gt-mediated repression, theeve2Δ(GTTT)₄-lacZ construct was crossed into a gt mutant background. If the two repression mechanisms are independent, we expect an additive effect from combining these two perturbations. If, however, Gt-mediated repression is partially redundant with the (GTTT)₄-binding activity, removing both might cause a more severe derepression. The result of this cross is shown inFig. 5D. There is an anterior shift and slight expansion of stripe 2 that is similar to the effects on the wild-type eve 2 transgene in gt mutants(Fig. 5C). No new effect is detected on the band of derepression created by deleting the(GTTT)₄ site, and a small repressed area is still maintained between the two parts of the pattern (Fig. 5D). This result is consistent with an additive effect, and suggests that the (GTTT)₄-binding activity functions independently of Gt-mediated repression. The failure to derepress in the region between the two parts of the pattern probably reflects the activity of the unknown protein X, which normally participates with Gt in repression.

The eve2Δ(GTTT)₄-lacZ transgene is also repressed at the anterior tip (Fig. 5B), even in gt mutants(Fig. 5D), suggesting that yet another mechanism prevents activation in this region. This mechanism could work through another localized repressor activity, or by modifying Bcd, the major activator of eve 2. It has been previously shown that Bcd-dependent activation of hb and orthodenticle(otd) is downregulated by the Tor phosphorylation cascade at the anterior tip (Ronchi et al.,1993), consistent with the latter possibility. To test whethertor controls the ability of Bcd to activate eve 2, theeve2Δ(GTTT)₄-lacZ transgene was crossed into embryos lacking tor activity. This causes a significant derepression at the anterior tip (Fig. 5F),suggesting that tor-mediated modification of bcd activity is important for preventing activation in this region. A similar derepression is not detected with the wild-type eve2-lacZ transgene in tormutants (Fig. 5E), suggesting that Tor-mediated repression is dependent on the (GTTT)₄-binding activity (see Discussion). In summary, these results suggest that multiple activities are required for anterior repression of eve 2, and that three different mechanisms prevent activation in different anterior regions.

Because of the peculiar sequence of the (GTTT)₄ site, we performed gel shift experiments using nuclear extracts from 0- to 12-hour-old wild-type embryos (Fig. 6A). These experiments showed the formation of several specific protein-DNA complexes. To identify specific proteins that bind the (GTTT)₄sequence, we conducted a yeast one-hybrid assay with constructs containing four tandem copies of the intact site. From an initial screen of ∼500,000 clones, we obtained 66 true positives (based on survival on his^-medium and increased lacZ production). These clones were then transformed into a yeast strain containing identical reporters except for base pair substitutions in the (GTTT)₄ sequence. Forty-nine clones also activated one of the mutant constructs, leaving only 16 that activated the(GTTT)₄ constructs, but not the negative controls(Table 1). Among these 16 were two clones that encode histone H1 and a single clone that encodes the forkhead domain (FD) protein Slp1. slp was originally classified as a pair-rule mutation (Nusslein-Volhard et al., 1985), but the slp locus contains two tightly linked genes, slp1 and slp2(Grossniklaus et al., 1992). These genes are related in their primary structure, and their expression patterns overlap significantly. However, slp1 is expressed much earlier in an anterior `gap gene-like' domain, which first appears as an anterior cap, and then evolves into a broad stripe at approximately 80% egg length (Grossniklaus et al.,1992). Double staining experiments with gt show that both genes are expressed at the same time, and that slp1 expression overlaps the anterior part of the gt expression domain(Fig. 7A). Thus, the temporal and spatial expression patterns of slp1 are consistent with a role in anterior repression of eve 2.

The (GTTT)₄ sequence bears little resemblance to the only previously defined Slp1-binding site (TCTTCGATGTCAACACACC)(Yu et al., 1999). Thus, we tested whether bacterially expressed Slp1 can bind directly to the(GTTT)₄ sequence in vitro (Fig. 6B). These experiments show that Slp1 binds specifically to this sequence, suggesting that it may directly interact with this sequence in vivo. Similar results were obtained using a fragment of the Slp1 protein that contains only the forkhead domain (data not shown).

If Slp1 acts as an anterior repressor of eve 2, genetically removing it might cause an anterior derepression of theeve2-lacZ expression pattern. To test this, we crossed the reporter into a slp deletion mutant that completely removes theslp1-coding region and disrupts slp2(Grossniklaus et al., 1992). No anterior expansion was detected in this experiment (data not shown). We also analyzed endogenous eve expression in this mutant background,and did detect slight anterior shifts of stripes 1 and 2 (data not shown), but no significant derepression in anterior regions. To test whether Slp1-mediated repression requires gt, we examined eve and theeve2-lacZ reporter gene in gt; slp double mutant embryos. The double mutant shows no increase in the anterior derepression over that caused by removal of gt alone (data not shown). These results argue against a role for Slp1 in anterior repression of eve, but do not rule out the possibility that Slp1 is one of several redundant proteins that repress through the (GTTT)₄ site. Two other FD proteins, Fkh and Crocodile (Croc), are expressed in anterior regions of the embryo(Hacker et al., 1995;Weigel et al., 1989). However,the expression domains of both proteins are located very near the anterior pole, making it unlikely that either gene is involved in this repression mechanism. To make sure, we examined eve and eve2-lacZ expression in each mutant; neither shows an anterior derepression. Thus these two genes are unlikely to play important roles in this repression mechanism.

To further test the roles of Slp1 and Gt in eve patterning, we used a fragment of the snail (sna) promoter(Ip et al., 1992) and the yeast FLP-FRT system (Struhl et al.,1993) to drive ectopic domains of each gene along the ventral surface of the embryo (Fig. 7B,C). This method is an efficient way to test whether any gene is sufficient for repression of individual stripes because the ventral expression domain intersects all seven eve stripes. In this assay, Slp1 expression alone distorts the expression of eve 1 in ventral regions by shifting it posteriorly, and causes a strong repression of eve 2 and a weaker repression of eve 3(Fig. 7E,H). By contrast, there is no detectable effect on the posterior stripes. Thus, Slp1 activity is sufficient for repression of specific anterior stripes including eve2. By contrast, ventrally misexpressed Gt causes only a weak repression ofeve 1 and 2, but strongly affects eve 5, a repression target of the posterior gt expression domain(Fig. 7F)(Nibu and Levine, 2001). The minor effect of Gt on eve 2 is transient, and the stripe recovers and expands posteriorly later in cycle 14 (Fig. 7I). This expansion is probably caused by repression ofKr, which forms the posterior border of eve 2(Wu et al., 1998). These results confirm that Gt is not sufficient for effective repression ofeve 2, and that its effect is much weaker than Slp1-mediated repression. We also generated embryos that contain ventral expression domains of both Slp1 and Gt. While effects of both genes were detected within the same embryos, there was no evidence of synergistic repression activity in these experiments (data not shown). This is consistent with the demonstration that the (GTTT)₄ site is independent of Gt-mediated repression(Fig. 5).

The repressive effects of ectopic Slp1 on the three anterior evestripes suggest a common mechanism for repression in anterior regions of the embryo. To test this, we searched the eve locus for binding sites similar to the (GTTT)₄ site in the eve 2 enhancer(Adams et al., 2000). Interestingly, there are only two other such sites in the eve locus,which are located within the boundaries of the stripe 1 and stripe 3+7 enhancers (Fig. 3B). We next tested whether repression by ectopic Slp1 is mediated through these enhancers and the eve 2 MSE. In these experiments, ectopic Slp1 caused a ventral repression of stripe 1 in the context of aneve1+5-lacZ transgene(Fig. 8B). A similar repression of stripe 3 was observed in the context of an eve3+7-lacZtransgene (Fig. 8D,F). These results are consistent with the idea of a common mechanism. By contrast, no ventral repression of the eve2-lacZ transgene was detected (data not shown), which is surprising in light of the fact that eve 2 is the most strongly affected stripe in the context of the endogenous gene.

The reason for the discrepancy between the reporter and the endogenous gene is not clear. It is possible that Slp1-mediated repression of eve 2 requires eve sequences outside the minimal enhancer. For example, the late element (LE) (Fig. 1)mediates the refinement of all seven eve stripes after they are initially positioned (Fujioka et al.,1996; Jiang et al.,1991a). Perhaps interactions between the LE and/or othercis sequences are required for effective repression by Slp1. To test this, we used a larger reporter gene (-7.8 eve-lacZ) that contains all native sequences from the 5′ border of the locus to the transcription start site. This transgene contains the LE, the eve 2 and eve 3+7 enhancers, and all native sequences that lie between these elements, and drives expression of stripes 2, 3 and 7, and a single line of nuclei located within the normal position of stripe 1(Fig. 8G). Expression from this reporter is effectively repressed at the position of stripes 1 and 3 in embryos containing ventrally expressed Slp1, but there is still no effect on the stripe 2 response (Fig. 8H). Thus, the addition of these extra sequences does not restore the sensitivity of eve 2 to Slp1-mediated repression. This suggests that undefined properties of the endogenous eve locus are required for Slp1-mediated repression of eve 2.

The results presented here indicate that three distinct mechanisms are required for anterior repression of eve 2, with each activity functioning within a specific subregion(Fig. 9A).

In subregion III, the Gt-binding sites(Fig. 1) are crucial for repression — deletion of these sites leads to an anterior expansion of the stripe. However, it is clear that Gt does not act alone, and that at least one other factor (X) must be involved in repressing through these sites(Small et al., 1992). The identity of X is not clear, but genetic studies have localized a Gt-like patterning activity to the left arm of chromosome II(Vavra and Carroll, 1989). Segmental aneuploids that remove this arm show an expansion similar to that seen in gt mutants.

In subregion II, repression of eve 2 is mediated by the(GTTT)₄ site described in this paper. We have further identified a candidate protein, Slp1, which is expressed at the right time and place for the repression activity and binds specifically to this site in the yeast 1-hybrid experiment and in vitro. The (GTTT)₄ site shows little similarity to the other known Slp1-binding site(Yu et al., 1999), but is quite similar to sites bound by other members of the FD protein family. For example, a 115 amino acid FD fragment of Fkh binds specifically to the site CTTTGTAAA (Kaufmann et al.,1994), which bears some resemblance to the (GTTT)₄site. Also, the hepatocyte FD protein HNF-3 binds to a site (TGTTTGTTTTAGTT)that contains two perfect GTTT repeats(Pani et al., 1992). Ventrally expressed Slp1 specifically represses eve 2, strongly supporting a role in regulation of the endogenous eve gene. However, there is no effect on eve 2 in slp mutants, suggesting that Slp1 is redundant with at least one other protein (Y), which also mediates repression through the (GTTT)₄ site. The existence of multiple complexes in gel shifts with embryo extracts is consistent with this, but the identity of Y is still unknown.

In subregion I, eve 2 repression is controlled by Tor, which may act by downregulating Bcd-dependent activation. This is consistent with the previous demonstration that Tor interferes with Bcd-dependent activation ofhb, otd and slp1(Grossniklaus et al., 1994;Ronchi et al., 1993).

In summary, at least five different protein activities are involved in three distinct mechanisms that repress eve 2 in anterior regions. Interestingly, it seems that all aspects of eve 2 regulation are controlled, directly or indirectly, by the Bcd morphogen gradient(Fig. 9B). The eve 2 enhancer is directly activated by Bcd, but activation is prevented near the anterior pole by Tor. The anterior expression patterns of the defined repressors of eve 2 (Slp1 and Gt) are also activated by Bcd. We propose that the relative positions of these domains, and the ultimate position of eve 2, are controlled by differential sensitivity to the Bcd concentration gradient. Future experiments on the cis-elements that regulate slp1 and gt transcription will be required to test this.

The study of eve regulation is an excellent paradigm for how complex promoters integrate the activities of multiple enhancers. Previous studies suggest that all five enhancers function independently in the segmented part of the embryo. The autonomy of each enhancer depends on short-range repression mechanisms and sufficient linear spacing between enhancers along the DNA sequence (Gray et al., 1994; Small et al.,1993). These factors are crucial for creating the pattern of seveneve stripes because they permit different enhancers to be in different transcriptional states within the same nuclei. For example, in nuclei at the position of stripe 1, the eve 1 enhancer will activate transcription even though the eve 2 enhancer is in a repressed state.

A different scenario exists in regions anterior to the striped pattern,where none of the eve enhancers are activated. The genetic removal of various repression activities suggests that at least two eveenhancers (eve 2 and eve 3+7) can be activated in this region. The mechanism of eve 1 activation is still unknown, but it is reasonable to suggest that its activators are also distributed in this region. Thus, mechanisms must be in place to prevent activation by each enhancer. Alternatively, repression could occur by an anterior repression activity that directly contacts the basal transcription complex. We have shown that the(GTTT)₄ site in the eve 2 enhancer mediates anterior repression, and that there are similar binding sites in the eve 1 andeve 3+7 enhancers. Furthermore, ectopic Slp1 expression represses all three stripes. These results argue against the mechanism of direct contact with the basal machinery, and suggest that these three enhancers share a common mechanism for repression in a specific anterior region of the embryo. Genetic experiments also show that the Tor phosphorylation cascade participates in polar repression of the eve 2 and eve 3+7 enhancers, which suggests another common repression mechanism that is shared by at least two of these enhancers.

By contrast, the eve 5 and eve 4+6 enhancers do not appear to contain sites similar to the (GTTT)₄ sites(Adams et al., 2000), and they are immune to repression by ectopic Slp1. These enhancers are expressed in posterior regions of the embryo, and thus may be activated by factors localized there, making anterior repression unnecessary for their function.

slp was originally classified as a pair-rule gene based on its cuticular phenotype, and has been shown to function at both the level of the pair-rule genes and the segment polarity genes(Grossniklaus et al., 1992). Specific patterning functions for the early anterior Slp1 expression domain,however, have remained unclear, although strong alleles of slp1exhibit severe defects in the mandibular lobe. Our results suggest that Slp1 acts at the level of the gap genes by repressing enhancer elements that control the initial eve stripes. Thus, Slp1 function is required at three different levels of the segmentation hierarchy.

The mechanism involved in Slp1-mediated repression of eve is unknown, but may involve an interaction with the co-repressor Groucho (Gro)(Paroush et al., 1994;Paroush et al., 1997). The Slp1 protein sequence contains a motif (FSIDAIL)(Grossniklaus et al., 1992),which is very similar to the EH1 Gro-binding consensus (FSIDNIL)(Jimenez et al., 1997) and Slp1 has been shown to bind Gro in vitro(Kobayashi et al., 2001).

It has been proposed that Gro mediates repression by creating a direct physical link between DNA-bound proteins and components of the basal transcription machinery (Courey and Jia,2001; Nibu et al.,2001). As such, repressors that act through Gro can function over very long distances, and have been classified as long-range repressors. The ability of Slp1 to interact with Gro suggests such a long-range mechanism, but several considerations are not consistent with this model. For example,ventral expression of a long-range repressor that `locks' the basal transcription machinery should repress all seven eve stripes, not just the anterior three. Also, the three (GTTT)₄ sites described here are all located within minimal enhancer elements that control specific stripes. In a long-range mechanism, these sites could be located anywhere in the promoter, and need not be associated with specific enhancers.

One of the most intriguing findings of this study is that ectopic Slp1 represses eve 2 in the endogenous gene, but not in the context of several lacZ reporter genes. Despite much effort, we have not resolved this discrepancy, but it is informative to compare the structural differences between the endogenous gene and the tested transgenes. One obvious difference between the lacZ reporter genes tested in these experiments and the endogenous eve gene is copy number. Perhaps Slp1-mediated repression requires two copies of the enhancer in a homozygous situation. A pairing-sensitive element (PSE) that reduces marker gene expression in homozygotes has been identified in the far 3′ region of the eve gene (Fujioka et al.,1999), which is consistent with this hypothesis. Two experiments argue against this hypothesis. First, ectopic Slp1 still represses endogenouseve in Df (eve)/+ heterozygotes (data not shown). Also, Slp1 fails to repress eve 2-containing transgenes when they are homozygosed (data not shown).

Another difference between the endogenous gene and the reporters is that the endogenous gene contains genomic regions outside those tested in the reporter genes, and is located in a different genomic position. Perhaps control sequences in the 3′ region of the gene, or further 5′ are required for this repression mechanism. As mentioned above, the eve 1 enhancer, which is located in the 3′ region, contains a(GTTT)₄-binding site. Perhaps effective repression of eve2 requires all three sites contained in the three different enhancers. This will be tested in future experiments.

Finally, it is possible that the native eve locus is organized in a specific chromatin conformation that permits repression by Slp1, and this configuration is not maintained when eve 2 transgenes are inserted into ectopic genomic locations. The fact that Slp1 protein contains an FD DNA-binding domain is interesting in this regard. Structural studies suggest that FD domains from a `winged-helix' very similar to the globular DNA-binding domain of the linker histone H1 (Clark et al., 1993). Furthermore, it has been shown that the mammalian FD protein hepatic nuclear factor 3 (HNF3) competes with H1 for binding to specific sites, and that this competition is critical for the in vivo regulation of the albumin liver-specific enhancer(Cirillo et al., 1998). Such a mechanism may be involved in Slp1-mediated repression of eve 2. Consistent with this, we isolated two clones that encode histone H1 in the one-hybrid experiment with the (GTTT)₄ site(Table 1). This suggests that both proteins can bind to this site, and supports the idea that regulation of chromatin structure may be an important part of Slp1-mediated repression ofeve 2. More experiments will be required to test this hypothesis.

The eve 2 enhancer is one of the best-characterized patterning elements in Drosophila development. Proteins involved in activation and repression have been identified, and a simple model has emerged that explains the basic activity of the enhancer(Fig. 1). We have shown that anterior repression of this element requires at least three position-specific mechanisms, which significantly extends our understanding of this aspect of enhancer function. Our results also suggest that the current model for activation of this enhancer is also incomplete. The deletion analysis(Fig. 2) identified four regions that are required for efficient activation of the enhancer. These effects of these deletions may be caused by changing the spacing between known activator and/or repressor sites within the enhancer. However, it possible that these regions contain specific binding sites required for activation. Consistent with this, there are several well-conserved sequence blocks that might represent specific sites required for activation. Base-pair substitutions that disrupt the conserved sequences without changing site spacing will be used to initially test this.

We thank Gary Struhl for the FRT cassette and the β-tub-FLP flies;Miki Fujioka for the modified CaSpeR vector with glass-binding sites and flies containing the eve 1+5 reporter gene; Leslie Pick for the Gal4AD library used in the one-hybrid experiment, the pET-Slp1 expression vector and the embryonic extracts used for gel shift analysis; Tony Ip for thesna promoter fragment; Alain Vincent for the croc flies; and Manfred Frasch for the slp^Δ34B flies. We thank members of Eric Chang's laboratory for advice on the yeast one-hybrid experiment, and members of the Small and Desplan laboratory for support. We thank Claude Desplan, Dot Clyde and Mary Anne Pultz for comments on the manuscript. L. P. A. was supported by a fellowship from the FAPESP organization in Brazil. This work was funded by NIH grant GM 51946, NSF grant 9982535 and the Margaret and Herman Sokol Foundation.