Hox repression of a target gene: extradenticle-independent, additive action through multiple monomer binding sites

Many animal body plans are constructed of serially repeated structures that develop from initially equivalent fields of cells during embryogenesis. Serially homologous structures often attain different morphologies and functions within a single animal and changes in their number, form and function mark many important evolutionary differences between taxa (Carroll et al., 2001). Understanding this basic modular aspect of animal design and evolution requires detailed knowledge of the genetic and developmental mechanisms that regulate the formation and identity of body parts. In arthropods and vertebrates, selector genes play central roles in these processes and encode transcription factors that possess the distinct ability to direct the formation of specific tissues, such as the eye or heart and the differentiation of homologous body parts, such as segments, vertebrae and appendages. Selector proteins are proposed to regulate the expression of numerous target genes within regulatory networks that control the development of these tissues (Guss et al., 2001; Halder et al., 1998; Mann and Morata, 2000; Weatherbee et al., 1998).

One class of selector genes, the Hox genes, encode homeodomain-containing proteins that regulate regional identity along the anteroposterior axis in animals and differentiate the identities of serially homologous structures such as vertebrae (Burke et al., 1995; Cohn and Tickle, 1999) and segments and appendages in arthropods (Abzhanov and Kaufman, 2000; Averof and Akam, 1995; Averof and Patel, 1997; Carroll et al., 1995; Grenier et al., 1997; Kaufman et al., 1990; Lewis, 1978; Panganiban et al., 1995; Rogers et al., 1997; Warren et al., 1994). Elucidating the mechanisms of Hox protein function is therefore critical to understanding the development and diversification of serially homologous structures. However, many facets of Hox target gene regulation are not well understood. Only a few direct genetic targets of Hox regulation have been identified, including cis-regulatory elements involved in Hox gene auto- and cross-regulation (Appel and Sakonju, 1993; Beachy et al., 1993; Beachy et al., 1988; Bergson and McGinnis, 1990; Chan et al., 1994; Dessain et al., 1992; Ferretti et al., 2000; Frasch et al., 1995; Gould et al., 1997; Grieder et al., 1997; Haerry and Gehring, 1996; Jacobs et al., 1999; Li and McGinnis, 1999; Maconochie et al., 1997; Malicki et al., 1992; Pinsonneault et al., 1997; Popperl et al., 1995; Regulski et al., 1991; Thuringer et al., 1993; Zeng et al., 1994) and the regulation of the Drosophila genes Distal-less (Dll) (Vachon et al., 1992), decapentaplegic (dpp) (Capovilla and Botas, 1998; Chan et al., 1994; Manak et al., 1994), teashirt (McCormick et al., 1995), forkhead (Ryoo and Mann, 1999) and apterous (Capovilla et al., 2001). A key issue arising from studies of these targets and of DNA-binding by Hox proteins is the relatively low DNA-binding specificities of Hox proteins (Ekker et al., 1994). Hox proteins generally bind a six base pair DNA sequence containing a TAAT core (Ekker et al., 1991) that occurs with high frequency throughout animal genomes (about once every kilobase pair), including the cis-regulatory elements of many genes, suggesting that there are many potential targets for Hox proteins.

How is Hox target selectivity achieved? In other words, how does a particular Hox protein recognize and regulate a subset of target genes among a much larger number of potential targets within a genome? One ‘widespread binding’ model proposes that the binding of Hox proteins to multiple monomer sites increases their occupancy of cis-regulatory elements, perhaps co-operatively (Biggin and McGinnis, 1997). In this scenario, numerous Hox binding sites within cis-regulatory elements would be required for targets to be regulated by Hox proteins. Alternatively, a ‘co-selective binding’ model proposes that Hox proteins could regulate cis-elements through co-operative interactions with protein cofactors that increase Hox protein DNA-binding affinities for larger compound binding sites (Biggin and McGinnis, 1997).

Many studies have demonstrated that complexes of Hox proteins and Extradenticle (Exd), a DNA-binding cofactor of the PBC family that also includes the vertebrate Pbx and nematode Ceh-20 proteins, are required to directly regulate several target genes (reviewed by Mann and Affolter 1998; Mann and Morata, 2000). Complexed with PBC cofactors, Hox proteins bind to a compound DNA sequence with greater DNA-binding affinity and specificity and hence, increased selectivity (Chan et al., 1994; Chan et al., 1997; Chang et al., 1995; Mann and Affolter, 1998; Ryoo and Mann, 1999; van Dijk and Murre, 1994). MEIS family homeodomain proteins (Hth, Meis and Prep) promote the nuclear localization of PBC proteins and increase the DNA-binding specificities and/or affinities of Hox/PBC complexes (Abu-Shaar et al., 1999; Berthelsen, 1999; Berthelsen et al., 1998; Ferretti et al., 2000; Jacobs, 1990; Mann and Affolter, 1998; Mann and Morata, 2000; Mercader et al., 1999; Rieckhof et al., 1997; Ryoo et al., 1999; Shanmugam et al., 1999; Vlachakis et al., 2001). Thus, there exists a great deal of support for the ‘co-selective binding’ model for Hox protein selectivity.

However, the fact that most target genes that have been analyzed require PBC proteins for their regulation has led, in our view, to perhaps an overemphasis on the primacy of Hox/PBC interactions in determining the selectivity of Hox proteins. Some studies have suggested that Hox proteins may act through widespread binding to monomer sites (Appel and Sakonju, 1993), while other studies have shown that some Hox-regulated targets are controlled independently of Exd (Pederson et al., 2000; Pinsonneault et al., 1997). Moreover, many Hox-regulated structures in Drosophila do not require the activity of either Exd or Homothorax (Hth) for their proper development, most notably the distal appendages (Abu-Shaar and Mann, 1998; Casares and Mann, 1998; Gonzalez-Crespo et al., 1998; Gonzalez-Crespo and Morata, 1996; Mann and Morata, 2000; Rauskolb et al., 1995; Wu and Cohen, 1999), including the haltere (Azpiazu and Morata, 1998; Azpiazu and Morata, 2000; Casares and Mann, 2000; González-Crespo and Morata, 1995). Furthermore, distal appendage development in other arthropods (Gonzalez-Crespo and Morata, 1996) and in vertebrates (Capdevila et al., 1999; Gonzalez-Crespo et al., 1998; Mercader et al., 1999; Mercader et al., 2000; Selleri et al., 2001) requires neither PBC nor MEIS function. Because Hox regulation of gene networks is critical to distal appendage development and does not involve PBC cofactors, the question of whether or how Hox proteins act selectively in the absence of PBC cofactors has particular importance in understanding the genetic mechanisms underlying the development and diversification of distal appendage morphologies and functions in arthropods and tetrapods.

Here, we focus on the regulation of haltere identity by Ultrabithorax in Drosophila as a model to understand how Hox proteins selectively regulate specific subsets of target genes and the differentiation of serial homologs in the absence of PBC/MEIS cofactors. We analyze the regulation of several genes that are repressed in the haltere by Ubx and show that the spalt (sal) gene is cell-autonomously repressed by Ubx in flight appendages. We demonstrate that the repression of a flight appendage-specific cis-regulatory element of sal in the haltere is directly regulated through multiple Ubx binding sites. In addition, we show that individual Ubx binding sites within this element contribute to, but are not sufficient for its complete repression in the haltere. These findings suggest that Hox proteins can act through multiple monomer binding sites in the absence of known cofactors. We propose that the evolution of some Hox target genes involved the gradual accumulation of multiple Hox monomer binding sites within cis-regulatory elements.

Clones expressing UbxIa protein were generated by heat induction of FLP recombinase (Xu et al., 1993) at 34°C for 20 minutes at 75-99 hours after egg laying (AEL) in flies of the genotype hsFLP122; Act5C promoter-FRT-CD2-FRT-GAL4/vgQ; UAS-UbxIa/+ (Castelli-Gair et al., 1994; Pignoni and Zipursky, 1997; Struhl and Basler, 1993). Third instar imaginal discs were later dissected, fixed and immunostained at 116-140 hours AEL using methods previously described (Halder et al., 1998). Ubx protein was detected using a mouse monoclonal anti-Ubx antibodies FP3.83 or FP6.87 (Kelsh et al., 1994) (gifts from R. White), the anti-Kn and anti-Sal antibodies were generously provided by Michèle Crozatier (Crozatier and Vincent, 1999) and Rosa Barrio (de Celis et al., 1996), respectively and the anti-β-galactosidase antibodies are commercially available (Molecular Probes and Promega). Embryos were stained as described by Galant and Carroll (Galant and Carroll, 2002).

DNaseI footprinting analysis of the sal 1.1 element was performed with the Ubx homeodomain (a generous gift from Phil Beachy) using methods previously described (Halder et al., 1998). Sites that were protected by Ubx in the sal 1.1 element and their mutant counterparts (listed in Fig. 2B) were further analyzed by electromobility shift assays on 20 base pair double-stranded oligonucleotides containing individual sites centered within its native flanking DNA sequences. These oligo probes were radioactively labeled by end-filling in two T overhangs on the 5′ and 3′ ends with [α-³²P]dATP using the Klenow fragment. The labeled probes were incubated for 30 minutes at room temperature (RT) with 0, 1.1, 3.3, 10, 30, or 90 ng of the Ubx homeodomain in 20 mM Hepes pH 7.5, 200 mM KCl, 0.25 mM EDTA, 1 mM DTT, 100 μg/ml BSA and 8% glycerol. They were then run on a 5% polyacrylamide gel (19:1 bis:acrylamide) at 150 V in 0.5× TBE for 2 hours. The gels were dried and exposed to XOMAT-AR X-ray film (Kodak). After developing the film, the concentration of Ubx at which half-maximal binding occurred to each wild-type probe was compared to its mutant counterpart to verify that the mutations decreased Ubx affinity for the probe to the level of non-specific DNA sequences without a TAAT core sequence (at least a ten-fold decrease).

Mutant variants of the sal 1.1 and 328 mutant elements were each created by using primers carrying appropriate base pair changes (listed in Fig. 2B) in two rounds of PCR, one round with mutant primers to amplify overlapping mutant DNA fragments individually and a second with flanking primers to amplify across all the fragments and generate the entire element. The mutant elements were cloned into the hsp-lacZ-CaSpeR reporter plasmid (Nelson and Laughon, 1993), their sequences verified by sequence analysis and they were subsequently used to produce transgenic fly lines. All primer sequences and further details are available upon request. The sal 1.1 element sequence is submitted in GenBank (accession no. AF46408712). β-galactosidase activity driven by the various wild-type and mutant elements was detected with X-gal (Halder et al., 1998) and observed to be consistent among at least three independent transgenic lines.

The construct used to drive Ubx YAAA mutant protein expression was created by amplifying two overlapping DNA fragments from a UAS-UbxIa cDNA (Castelli-Gair et al., 1994) using primers carrying mutations that change the YPWM motif in Ubx to YAAA in one round of PCR (primer sequences available upon request), amplifying across the two overlapping mutant fragments with flanking primers in a second round of PCR and cloning the final PCR product into pUAST (Brand and Perrimon, 1993). The Ubx YAAA mutant cDNA was confirmed by sequence analysis. Ectopic expression of the Ubx and Ubx YAAA proteins was produced by crossing female flies carrying the arm¹¹-Gal4 driver (available from the Bloomington fly stock facility) and the Dll304 embryonic limb enhancer β-galactosidase reporter gene (Vachon et al., 1992) to males carrying a UAS-UbxIa or UAS-Ubx YAAA transgene. The progeny from these crosses were either collected as embryos at 0-20 hours AEL and stained with antibodies (Panganiban et al., 1995) or their cuticles prepared 24-48 hours AEL (Roberts, 1998).

Gel shift analyses were performed with a radiolabeled oligonucleotide probe of the sequence 5′ TTAGCGATGATTAATTGCCTCCTT 3′ with in vitro transcribed and translated full-length Exd, Ubx and Ubx YPWM-YAAA proteins as previously published (Galant and Carroll, 2002). When indicated, 3 μl of Ubx protein and/or 1 μl of Exd were used in gel shift reactions.

We have previously shown that many cis-regulatory elements and genes, including the vestigial quadrant enhancer (vgQ) and the vein patterning genes Drosophila Serum Response Factor (DSRF)/blistered and sal, exhibit wing-specific expression patterns, but a complete absence of expression within the developing haltere field because they are repressed by Ubx (Weatherbee et al., 1998). Because both cell-autonomous and non cell-autonomous regulation of some target genes by Ubx has been observed (Shashidhara et al., 1999; Weatherbee et al., 1998), this repression could be mediated directly by Ubx or indirectly through its repression of upstream activators of these target genes, including important signaling molecules (Weatherbee et al., 1998). In order to elucidate the mechanism of Ubx action in the developing haltere, we performed additional genetic experiments to identify Ubx-regulated genes that would be most amenable to molecular analysis.

Previous studies analyzed the responses of potential Ubx-regulated target genes to broad, ectopic expression of Ubx in wings and in large, loss-of-function Ubx mutant clones in halteres (Shashidhara et al., 1999; Weatherbee et al., 1998). This can make it difficult to distinguish between cell-autonomous (potentially direct) and non cell-autonomous (indirect) effects. We reasoned that the expression of Ubx in small, discrete groups of cells would minimize the effects of non cell-autonomous regulation of target genes by Ubx. We monitored the autonomy of the effects of Ubx in small clones of cells that ectopically express Ubx on the expression of several downstream targets in third instar wing imaginal disks. Reporter gene expression driven by the vgQ enhancer was cell-autonomously repressed in Ubx-expressing clones close to the dorsal-ventral (DV) boundary in the wing (Fig. 1A). This result suggests that the vgQ enhancer may be a direct target of Ubx repression and indicates that Ubx regulates the vgQ enhancer at two levels, because Ubx also represses an unidentified signal emanating from the DV boundary in the haltere that affects vgQ enhancer activation (Shashidhara et al., 1999). We also found that the DSRF gene is regulated by Ubx at more than one level. knot (kn), a Hedgehog signaling-responsive gene that encodes a COE family transcription factor is required for the activation of DSRF along the anteroposterior (AP) boundary of the developing wing (Vervoort et al., 1999). Neither DSRF nor Kn is expressed in the developing haltere pouch (Weatherbee et al., 1998) (data not shown). In small clones of ectopic, Ubx-expressing cells located on the AP boundary, the repression of Kn by Ubx was cell-autonomous (Fig. 1B). It follows then that the repression of DSRF along the AP boundary of the wing disk by Ubx is achieved at least in part through Kn (i.e. is indirect). However, because DSRF is expressed in a broader domain than that of Kn (Montagne et al., 1996; Vervoort et al., 1999), DSRF expression may also be directly repressed in the haltere by Ubx.

In contrast to vg and DSRF, sal appears to be regulated by Ubx solely in a direct manner. The two known upstream activators of sal, the Dpp signaling pathway (Lecuit et al., 1996; Nellen et al., 1996) and the Vestigial (Vg)/Scalloped (Sd) flight appendage selector protein complex (Guss et al., 2001), are both active in the wing and haltere (Weatherbee et al., 1998). Previous work has shown while other Dpp-regulated genes, such as optomotor-blind/bifid, are expressed in the haltere, the sal gene is repressed and thus selectively regulated by Ubx. Clones of cells that ectopically expressed Ubx and were located within the Sal expression domain in the wing exhibited cell-autonomous repression of Sal (Fig. 1C). Because both the vgQ enhancer and DSRF are regulated by Ubx at multiple levels, the analysis of their direct regulation by Ubx in vivo is complicated by these indirect effects. Therefore, we focused our molecular analyses on the regulation of the sal gene by Ubx.

sal expression in the wing is regulated by a discrete cis-regulatory element that is directly activated by the Vg/Sd selector protein complex and is also bound by Mothers-against-dpp, a transcriptional effector of Dpp signaling (Guss et al., 2001). Both this 1.1 kb element (sal 1.1) and a smaller sub-element contained within it (sal 328) drove reporter gene expression in third instar wing discs, but no significant expression in haltere discs (see Fig. 3A,B,E,F,). Hence, these sal regulatory elements, like Sal, are repressed in the haltere and thus may be directly regulated by Ubx.

To test whether sal could be a direct target of Ubx repression, we searched for potential Ubx binding sites within the sal 1.1 element by DNaseI footprinting analysis with purified Ubx homeodomain protein. We identified seven regions protected by Ubx (a representative set of footprints is shown in Fig. 2A). Five of these regions contained one consensus TAAT core site (Fig. 2B, sites 1,2,5-7), one region contained two core sites (Fig. 2B, site 4) and one contained three core sites (Fig. 2B, site 3). The nucleotide sequences flanking the TAAT core site in all seven regions of Ubx binding in the sal 1.1 element showed characteristics of sites bound by Ubx in vitro (Ekker et al., 1991; Ekker et al., 1994). The sites protected by Ubx were scattered throughout the sal 1.1 element and three of these were located within the sal 328 element (Fig. 2C). The presence of these Ubx binding sites in the sal elements supports the possibility that sal may be directly repressed in vivo by Ubx.

If repression of the sal 1.1 and sal 328 elements in the haltere is directly mediated by Ubx binding to these sites, then abolishing the binding sites should result in derepression of reporter gene expression driven by the mutant elements in the haltere. To test this, we altered all seven of the Ubx binding sites within the sal 1.1 element and the three sites in the sal 328 element (shown in Fig. 2B) by introducing mutations that abolished the specific binding of Ubx to them (Ekker et al., 1994). We performed electromobility gel shift analyses on short oligonucleotide probes containing individual wild-type sites footprinted by Ubx or their mutant variants to verify that the affinity of Ubx protein for each mutant site was reduced to that of Ubx for non-specific DNA sequences (lacking a TAAT core sequence), at least ten-fold lower for each site than its wild-type counterpart (a representative example is shown in Fig. 2D). We then reintroduced these elements with mutated Ubx binding sites into a lacZ reporter gene vector and assayed β-galactosidase activity in developing wing and haltere imaginal disks from transgenic Drosophila. We observed that the mutant sal 1.1 element lacking all seven Ubx binding sites (sal 1.1 u1-7) now strongly drove β-galactosidase expression in the haltere along the anteroposterior boundary in a pattern similar to that in the wing (i.e. it was derepressed) (compare Fig. 3C,D). Similarly, the mutant sal 328 element in which three Ubx binding sites were altered (sal 328 u5-7) drove strong reporter gene expression in the haltere (Fig. 3G,H). We conclude that repression of the sal 1.1 and sal 328 elements in the haltere requires some or all of the Ubx binding sites we identified in vitro and that these elements are direct targets of Ubx repression.

Cis-regulatory elements in Drosophila display a wide range of both the number and composition of Hox binding sites that contribute to their regulation, from an individual Hox/PBC compound binding site that can mediate a detectable level of target gene regulation (Vachon et al., 1992; White et al., 2000), to a very large number of apparently monomer Hox binding sites (Appel and Sakonju, 1993), to a combination of both compound Hox/Exd and Hox monomer binding sites (Capovilla and Botas, 1998; Capovilla et al., 1994; Chan et al., 1994; Manak et al., 1994; Sun et al., 1995). The diversity of Hox binding sites within cis-regulatory elements raises the question of whether Hox-regulation is mediated by a net effect through multiple binding sites or primarily through a single monomer or compound site? For most Hox-regulated elements the number and the nature of binding sites sufficient for regulation by Hox proteins is unknown and this is of primary importance in understanding the regulation of Hox target genes and their evolution. The properties of Ubx-regulation of the sal 328 cis-regulatory element presented an opportunity to investigate the scope of the ability of individual Ubx binding sites to contribute to repression.

We examined whether each of the individual Ubx binding sites contribute to repression of the sal 328 element in the haltere in two ways. First, we examined the necessity of each of the three binding sites for repression by mutating each one singly and examining β-galactosidase activity driven by lacZ reporters for each of these elements in developing wings and halteres. Abolishing specific Ubx binding to sites 5 (Fig. 4A, right) and 6 (Fig. 4B, right) in the sal 328 element resulted in the ability of these single site mutant sal 328 elements to drive significant reporter activity above background levels in the haltere (i.e. partial derepression) while abolishing site 7 did not (Fig. 4C, right). Therefore, both Ubx binding sites 5 and 6 are necessary for complete repression of the sal 328 element, while Ubx binding site 7 is not.

Our second approach was to examine the degree to which these single sites were sufficient to impart repression of the sal 328 element in the haltere by mutating pairwise combinations of two Ubx binding sites at a time, leaving one site intact within each mutant element. We then assayed β-galactosidase activity driven by each of these elements. All three elements containing a single Ubx binding site drove reporter gene expression in haltere disks (Fig. 4D,E,F), indicating that no single Ubx binding site within the sal 328 element is sufficient to impart full repression by Ubx. However, we also observed that none of the single Ubx binding site elements was expressed at the same level in the haltere as was the triple mutant construct (compare Fig. 4D,E,F with Fig. 3H). These necessity and sufficiency tests demonstrate that each individual Ubx binding site contributes to the repression of the sal 328 element and that complete repression is a net effect of all three sites. Among the three Ubx binding sites, site 5 appeared to mediate the strongest level of repression of the sal 328 element (Fig. 4D). Because Drosophila haltere development does not require any activity of the Hox cofactor Exd and single Ubx binding sites appear to contribute to, but are insufficient for complete repression of, the sal 328 element by Ubx, we conclude that Ubx regulates this target cis-regulatory element in the absence of Exd via multiple monomer binding sites.

The ability of Ubx to regulate genes in the haltere independently of Exd raised the question of whether Ubx can do so elsewhere during development, especially in tissues where Exd is present. Co-crystal structures of partial Hox/PBC complexes demonstrate that interactions between these two proteins are mediated via contacts between a YPWM peptide that is located upstream of Hox protein homeodomains and very highly conserved among Hox proteins and a hydrophobic pocket in the PBC protein (Passner et al., 1999; Piper et al., 1999). Many studies have shown that the YPWM motif is essential for Hox/PBC interactions (Chang et al., 1995; Johnson et al., 1995; Knoepfler and Kamps, 1995; Neuteboom et al., 1995; Phelan et al., 1995) and both X-ray crystallographic and other biochemical experiments indicate that the tryptophan residue is critical for complex formation (Neuteboom et al., 1995; Passner et al., 1999; Phelan et al., 1995; Piper et al., 1999). To create a Ubx protein that should not interact with Exd, we mutated the YPWM motif in the Ubx protein to the sequence YAAA, a sequence that incorporates several mutations that have been shown to abolish the interaction between several Hox proteins and Exd (Chang et al., 1995; Knoepfler and Kamps, 1995; Neuteboom et al., 1995; Phelan et al., 1995; Shanmugam et al., 1997). Interestingly however, this protein still displayed significant (but slightly reduced) cooperative binding with Exd to a compound Ubx/Exd site in vitro (Fig. 5A). Since this result was not expected based on the literature, we also tested another mutant version of the YPWM motif (AAAM). This protein also bound cooperatively with Exd to a compound site (data not shown). These results suggest that mutation of the YPWM alone is not sufficient to eliminate Ubx/Exd interactions in vitro in the presence of a DNA binding site. We note that the original report of Ubx/Exd interactions detected this interaction with a form of Ubx protein that lacked the YPWM motif (Chan et al., 1994), suggesting that other residues in the Ubx protein must also interact with Exd. Furthermore, it should be pointed out that co-crystallographic study of Ubx/Exd was performed with a form of Ubx that lacked most of the N terminus and all of the C terminus of the protein (Passner et al., 1999), so additional contacts that might involve these residues would not have been observable. Lastly, we note that yeast two-hybrid experiments that indicated a requirement for the YPWM motif were performed in the absence of target DNA sequences (Johnson et al., 1995). Based upon our data, it appears that Ubx and Exd monomers, when brought into proximity by the presence of a compound binding site, may still interact in vitro.

We tested the activity of the YAAA mutant Ubx protein by ectopically expressing it during Drosophila embryogenesis. Ectopic expression of wild-type Ubx during embryogenesis transforms third thoracic (T3) segmental identity to that of the first abdominal (A1) segment (compare Fig. 5B with 5C) (Mann and Hogness, 1990). However, ectopic expression of the Ubx YPWMrYAAA mutant protein transformed T3 segmental identity to that of A2 (Fig. 5C). Importantly, in an exd zygotic mutant, Ubx specifies A2 segmental identity and when ectopically expressed in an exd mutant background, Ubx induces transformations of thoracic segmental identities to that of A2 (Peifer and Wieschaus, 1990). Because Ubx specifies A1 identity in the presence of Exd, but A2 segmental identity in its absence, our results suggest that the A2 segmental identity transformation induced by the Ubx YAAA protein may be the result of Ubx action that is independent of Exd. This raises the possibility that the YAAA mutation affects the activity of the Ubx/Exd complex more so than in vitro binding to compound sites.

In order to ascertain whether specific target genes are regulated by this mutant Ubx protein, we examined reporter gene expression driven by a Dll embryonic limb enhancer. In the abdomen, this enhancer is directly repressed by Ubx (Vachon et al., 1992) and it has recently been proposed that this is mediated by a Ubx/Exd compound binding site (White et al., 2000). Ectopic expression of wild-type Ubx protein in the embryonic ectoderm repressed the Dll enhancer (Fig. 5E,F) (Chan and Mann, 1993; White et al., 2000). We also observed that ectopic expression of Ubx YAAA strongly repressed Dll (Fig. 5G). Previous work has shown that abolishing the YPWM motif in the Hox protein Labial caused both an increase in in vitro DNA-binding affinity and hyperactivity in the regulation of a target cis-regulatory element in vivo (Chan et al., 1996). We tested this possibility for our Ubx YAAA mutant, but observed no difference between the binding affinities of the Ubx and Ubx YAAA proteins for an oligonucleotide bearing the compound Ubx/Exd binding site from the Dll enhancer in electromobility gel shift assays (data not shown). Together, these results suggest that a Ubx protein with an altered YPWM motif may act independently of Exd to regulate target genes but the in vitro data indicates that caution is necessary concerning whether this mutation affects Ubx binding to targets and/or Ubx activity in vivo.

We have shown that three targets of Ubx in the haltere, the vgQ enhancer, kn and sal, are cell-autonomously regulated by Ubx and that the sal gene is directly repressed by Ubx. This repression is mediated by multiple Ubx binding sites in a flight appendage-specific sal cis-regulatory element in which each individual Ubx binding site contributes to the overall complete repression of gene expression. Because the development of the haltere does not require PBC or MEIS protein activity, we suggest that Ubx is acting through simple monomer sites. We further provide evidence that Ubx may regulate other target genes independently of the Hox DNA-binding cofactor Exd in tissues other than the haltere. These results show that Hox target gene selectivity can be achieved through several Hox monomer binding sites without requiring either PBC or MEIS protein activity. This has important implications regarding the mechanisms underlying the regulation and evolution of Hox target genes and the development and diversification of serially homologous structures.

The low DNA-binding specificities of Hox proteins in vitro have been a long standing problem in determining how various Hox proteins selectively regulate target genes in vivo. This challenge has been further compounded because relatively few direct Hox targets have been identified. Molecular studies of Hox target genes have been limited to several auto- and cross-regulated Hox gene cis-regulatory elements (Appel and Sakonju, 1993; Beachy et al., 1993; Beachy et al., 1988; Bergson and McGinnis, 1990; Chan et al., 1997; Dessain et al., 1992; Ferretti et al., 2000; Frasch et al., 1995; Gould et al., 1997; Grieder et al., 1997; Haerry and Gehring, 1996; Jacobs et al., 1999; Li and McGinnis, 1999; Maconochie et al., 1997; Malicki et al., 1992; Pinsonneault et al., 1997; Popperl et al., 1995; Regulski et al., 1991; Thuringer et al., 1993; Zeng et al., 1994) and to only six other target gene cis-regulatory elements known in Drosophila (Capovilla and Botas, 1998; Capovilla et al., 2001; Chan et al., 1994; Manak et al., 1994; McCormick et al., 1995; Pederson et al., 2000; Ryoo and Mann, 1999; Vachon et al., 1992). The regulation of all but two of these elements has been shown to involve Hox/PBC complexes that, compared to Hox proteins alone, exhibit an increase both in DNA-binding affinity and the size of the binding sites that they occupy in cis-regulatory elements. Target genes regulated in vivo by Hox/PBC protein complexes generally require a single ten base pair compound binding site consisting of an Exd binding site neighboring a Hox binding site for their regulation (Chan et al., 1996; Chan et al., 1997; Grieder et al., 1997; Jacobs et al., 1999; Popperl et al., 1995; Ryoo and Mann, 1999; Ryoo et al., 1999; White et al., 2000). In addition, several of these elements also require a MEIS binding site for their activation (Berthelsen et al., 1998; Jacobs et al., 1999; Ryoo et al., 1999). Altogether, a Hox/PBC compound binding site and a MEIS binding site span at least sixteen base pairs and will occur rarely at random (see below). Hence, Hox/PBC/MEIS interactions bestow upon Hox proteins a much greater capacity to select among many potential target genes within animal genomes.

Recent studies of Hox-regulated target genes have focused largely upon Hox/PBC/MEIS protein complexes and their binding sites. The question we address here is how Hox proteins selectively regulate the expression of target genes independently of these co-factors. Specifically, we examine the potential contribution of single Hox monomer binding sites to the repression of Hox target genes, which we suggest has been largely underestimated. We have shown that several Ubx binding sites individually mediate partial repression of the sal 328 element in the haltere (Fig. 4D-F) and that the additive contribution of three sites completely represses sal expression (Fig. 3B,F). Because these sites are apparently simple monomer sites and the activities of Exd and Hth are not genetically required for Ubx function in the development of the haltere (Azpiazu and Morata, 1998; Azpiazu and Morata, 2000; Casares and Mann, 2000; González-Crespo and Morata, 1995), Ubx repression of the sal element is clearly independent of Exd or Hth. Our results suggest that Hox proteins do regulate the activity of cis-elements through single monomer binding sites without requiring either of the currently identified DNA-binding Hox protein cofactors and these provide support for the ‘widespread binding’ model for Hox protein selectivity (Biggin and McGinnis, 1997).

We have not ruled out the possible existence of unidentified DNA-binding cofactors that may be required for Ubx to repress target genes. One would predict that such a DNA-binding cofactor of Ubx would bind to similar flanking sequences that are shared among Ubx binding sites within a cis-regulatory element, as is the case for Ubx/Exd composite sites. However, we do not observe any shared motifs outside of the core Ubx binding site sequences among sites 5, 6 and 7 (Fig. 2B). Rather, we believe that the evidence points toward the ability of Hox proteins to mediate Hox repression through monomer binding sites without requiring other DNA-binding cofactors. Three previous studies of Hox-regulated cis-elements have demonstrated or implicated the ability of apparent individual Hox monomer binding sites to contribute to target gene regulation. In a remarkable example, the Antp gene contains as many as 41 Ubx binding sites in its P2 cis-regulatory element, a large number of which are required for its repression (Appel and Sakonju, 1993). In another case, in addition to two Exd binding sites and a Hox/Exd compound binding site that is required to regulate the dpp midgut enhancer, seven Ubx/Abdominal-A (Abd-A) monomer binding sites also contribute to target gene regulation (Capovilla and Botas, 1998; Capovilla et al., 1994). Furthermore, Pederson et al. (Pederson et al., 2000) have shown that Deformed may regulate a target independently of Exd. Here, we have provided evidence that individual Ubx monomer binding sites contribute to and that just three sites are sufficient for, the complete repression of the sal 328 element (Fig. 4D-F). Therefore, Hox repression of cis-regulatory elements may be realized through the net, additive effect of Hox proteins binding to multiple monomer sites that individually mediate weak regulation.

Hox regulation of the activity of cis-regulatory elements through multiple monomer binding sites may operate through a number of different modes. For instance, the presence of multiple sites may increase Hox-binding site occupancy in cis-elements either through cooperative interactions between Hox proteins (Beachy et al., 1993) or by increasing the probability that a site is bound by a Hox protein by virtue of a large number of sites (i.e. the more Hox binding sites, the greater the probability that any one site is bound by a Hox protein at a given time). Because Hox/PBC/MEIS compound sites also serve to increase binding site occupancy by Hox proteins, these mechanisms may serve essentially equivalent functional roles in regulating cis-elements. Hox proteins may also regulate some of their targets without requiring PBC or MEIS cofactors by binding to Hox monomer sites that are positioned in specific sequence contexts, such as their proximity to other activator or repressor binding sites within cis-regulatory elements. For instance, among the three Ubx binding sites within the sal 328 element, Ubx binding site 5 is the closest to the Sd binding sites that are required for the activation of this cis-regulatory element (Fig. 2C) and it appears to mediate the strongest repression (Fig. 4D).

Repression by Ubx could be mediated by a number of mechanisms, including steric hindrance of Sd or other transcriptional activators from binding to the sal 328 element or by affecting local chromatin structure. We have recently localized a motif at the C terminus of the Ubx protein that is involved in repression activity (Galant and Carroll, 2002). This motif may potentiate repression activity through interactions with components of the basal or activated transcriptional machinery, or by recruiting corepressors. Such interactions may be sufficient to account for the ability of single monomer binding sites to affect gene regulation.

The evolution of Hox target genes has accompanied a major evolutionary trend of diversification of animal body plans and morphologies (reviewed by Carroll et al., 2001). Several studies have demonstrated that changes in Hox target gene regulation are correlated with morphological differences among homologous body regions (Belting et al., 1998; Kopp et al., 2000; Palopoli and Patel, 1998; Warren et al., 1994; Weatherbee et al., 1999). For example, it has been postulated that the serially homologous fore- and hindwings of the four-winged ancestor of Dipterans had largely identical morphologies and presumably largely identical gene expression patterns and that during the evolution of the Dipteran haltere from this ancestral state, a subset of genes involved in flight appendage development became repressed by Ubx (Weatherbee et al., 1999). Elucidating the molecular mechanisms by which Hox proteins selectively regulate cis-elements is therefore key to understanding how Hox target genes may have evolved and contributed to such events.

Our findings that individual Hox monomer binding sites can contribute to Hox target gene regulation and that as few as three sites are sufficient to mediate complete repression suggest a simple, but potentially very important mechanism that may underlie Hox target gene evolution. We propose that the evolution of the Hox-repression of a target gene involves the stepwise accumulation of Hox monomer binding sites within cis-regulatory elements and progresses from initially quantitative differences in the activity of a cis-element to a qualitative difference (i.e. full repression) (Fig. 6). Our scenario is as follows: (i) point mutations can easily give rise to a single monomer Hox binding site that mediates partial repression of the activity of a cis-regulatory element in a tissue that expresses a given Hox protein (Fig. 6, red shading); (ii) if the Hox binding site becomes fixed under selection for a decreased level of gene activity in this tissue, then (iii) directional selection on the activity of a cis-regulatory element can fix additional sites that further repress it; and (iv) the evolution of several Hox monomer binding sites in the cis-element can eventually lead to its full repression and thus a qualitative difference in gene expression.

This scenario suggests that Hox target genes can evolve gradually through multiple intermediate steps. We favor this view because of the greater probability of functional sites arising as monomers and because smaller, gradual changes in gene regulation are likely to be less deleterious to development than larger ones. The rate at which target gene cis-elements can evolve regulation by Hox proteins is determined by the number and composition of the binding sites required for Hox proteins to regulate the activity of cis-regulatory elements. Hox monomer binding sites (TAAT[g/t][g/a]) should occur approximately once every 500 base pairs, on average. However, high affinity Hox/PBC binding sites (TGATNNAT[g/t][g/a]) should occur around 16 times less frequently (≈1/8,200 base pairs). If composite Hox/PBC and MEIS binding sites were minimally required to mediate Hox-regulation, then no fewer than 16 specific base pairs would need to evolve within a cis-regulatory element and the occurrence of combinations of a high affinity Hox/PBC binding site and a high affinity MEIS binding site (CTGTCA) located within 20 base pairs of each other is more than 800 times less probable (≈1/420,000 base pairs) than the occurrence of Hox monomer binding sites. These probabilities indicate that a Hox monomer binding site is much more likely to arise de novo than either a Hox/PBC compound site or a Hox/PBC/MEIS trimeric binding site. In this manner, the low DNA-binding specificities of Hox proteins may actually facilitate the evolution of new binding sites. The yeast alpha2 homeodomain repressor protein has a similarly low DNA-binding specificity and this results in its ability to repress target genes in vivo through many different DNA sequences (Smith and Johnson, 1994). Because Hox monomer sites appear to be functional, selection can favor the evolution of higher affinity Hox/PBC composite and Hox/PBC/MEIS trimeric sites through the sequential modification of monomer sites.

Hox proteins have played a central role in the diversification of serially homologous structures. Subsets of serially homologous structures express different Hox proteins (e.g. regions of vertebrate hindbrains, vertebrae, insect flight appendages and arthropod body segments), thereby uncoupling the development of serial homologs and enabling them to follow independent evolutionary trajectories. This regulatory logic facilitated the evolution of morphological differences between thoracic and cervical vertebral identities among vertebrates (Burke et al., 1995; Cohn and Tickle, 1999), between wings and halteres in Dipterans (Warren et al., 1994; Weatherbee et al., 1998) and the vast diversity of appendage number, shape and function in arthropods (reviewed by Gellon and McGinnis, 1998; Carroll et al., 2001). We suggest that the accumulation of simple Hox protein binding sites within cis-regulatory elements that direct gene expression in serial homologs has played an important role in the evolution of morphological diversity. This is certainly the case in the distal appendages of arthropods and vertebrates where the activity of neither PBC nor MEIS proteins is required and is perhaps a general mechanism governing the development and evolution of Hox-regulated characters.

We thank M. Crozatier, R. Barrio and R. White for antibodies, P. Beachy for purified Ubx homeodomain, D. Nellen and K. Basler for the sal elements, R. Mann for the Exd plasmid, K. Guss for reagents and fly stocks, K. Vaccaro for technical assistance, C. Nelson and A. Kopp for critical readings of the manuscript and J. Carroll for its preparation. R. G. is supported by an NIH predoctoral training grant provided to the Genetics Department and S. B. C. is an investigator of the Howard Hughes Medical Institute.