DWnt-4, a novel Drosophila Wnt gene acts downstream of homeotic complex genes in the visceral mesoderm

In response to the positional information delivered by maternal morphogens and segmentation gene products, the HOM-C selector genes are expressed in defined body parts of the embryo (St Johnston and Nüsslein-Volhard, 1992; Ingham and Martinez Arias, 1992; McGinnis and Krumlauf, 1992). Their functions are to assign identities and specify the diversified morphogenesis of metameric units (Lewis, 1978). Homeoproteins, acting as sequence-specific transcription factors (Desplan et al., 1988), are thought to achieve their effect by regulating a battery of subordinate target genes (Garcia-Bellido, 1975; Andrew and Scott, 1992). Hence, each metameric unit will enter a specific morphogenetic pathway, depending on the effector target genes regulated by the set of HOM-C genes which are expressed in it. While it is mostly from Drosophila that the molecules participating in this genetic cascade have been isolated, growing evidence suggests that the molecular processes involved have their counterparts in vertebrates (Duboule, 1992; McGinnis and Krumlauf, 1992; Bachiller et al., 1994).

Most of the few putative targets of HOM-C genes identified so far encode regulatory molecules, thus demonstrating further complexity in the genetic cascade upstream of the ‘realizator’ genes whose products act directly in morphogenesis and differentiation processes (Morata, 1993). Besides genes that themselves encode transcriptional regulators, several are involved in cell-cell communication, and presumably therefore in signal transduction pathways downstream of which transcription factors function (Botas, 1993). The products of two candidate targets identified using immunopurification procedures, connectin (Gould et al., 1992) and scabrous (Graba et al., 1992), function as cell signalling molecules. connectin encodes a cell surface protein which acts as a homophilic cell-cell adhesion protein and probably has a role in the formation of neuromuscular connections (Nose et al., 1992). scabrous protein is thought to be involved in lateral inhibition during the development of the nervous system (Baker et al., 1990). Two more genes, decapentaplegic (dpp) and wingless (wg), encode secreted glycoproteins whose expression is regulated in the visceral mesoderm (VM) by the action of homeotic proteins (Immerglück et al., 1990; Reuter et al., 1990; Capovilla et al., 1994). dpp is a member of the TGF-β family (Padgett et al., 1987) and wg is a segment polarity gene, which belongs to the Wnt family (Rijsewijk et al., 1987).

A variety of Wnt genes has now been discovered in various species (Nusse and Varmus, 1992; McMahon, 1992; Sidow, 1992). Every Wnt deficiency or ectopic expression provokes dramatic developmental defects, in the morphogenesis of the central nervous system (McMahon and Bradley, 1990; Thomas and Capecchi, 1990), cell proliferation (Roelink et al., 1990; Skaer and Martinez Arias, 1992), embryonic axis specification and mesodermal patterning (McMahon and Moon, 1989; Takada et al., 1994), embryonic segmentation (Nüsslein-Volhard and Wieschaus, 1980; Noordermeer et al., 1992) or limb patterning (Couso et al., 1993; Struhl and Basler, 1993). Wnt proteins are thus recognised as cell signalling mediators that play important regulatory functions in a number of basic developmental processes.

Orthologs of the murine Wnt-5 and Wnt-7, have been recently identified in Drosophila (Eisenberg et al., 1992; Russel et al., 1992). In this paper, we report the characterisation of a novel Drosophila Wnt gene, cloned from a library of genomic DNA fragments associated in vivo to Ubx proteins (Graba et al., 1992). This gene locates cytogenetically to the same chromosomal position as wg, the 5′ ends of their translated regions being separated by 30 kb of DNA. It shows a segment polarity-like pattern of expression in the ectoderm and is controlled by HOM-C genes in the VM.

Oregon R was used as the wild-type standard. The mutant alleles used were Scr^W17 (Wakimoto and Kaufman, 1981), Antp^W10 (Wakimoto and Kaufman, 1981), Ubx⁹²² (Kerridge and Morata, 1982), abd-A^MX1 (Sanchez-Herrero et al., 1985), Df(3R)109 (Lewis, 1978), dpp^{shv4 (}Segal and Gelbart, 1985) and dpp^{shv6 (}Segal and Gelbart, 1985). The Antp-Ubx chromosome was generated by recombination using the Antp^W10 and Ubx⁹²² alleles.

Flies were grown in standard culture conditions at 25°C. Egg collections were made from transheterozygotes over a wild-type chromosome, dechorionated in 50% bleach for 2 minutes, prefixed in a mix (v/v) of n-heptane and PBS buffer containing 50 mM EGTA, 10% formaldehyde, devitelinised in a −70°C prechilled mix of n-heptane/methanol (v/v) containing 50 mM EGTA, and stored in ethanol at −20°C.

In situ hybridisation to whole embryos using digoxigenin-labelled probes (Boehringer) was performed according to the method of Tautz and Pfeifle (1989). To detect DWnt-4 RNA, we used either a 4.2 kb EcoRI genomic fragment, which overlaps 1.4 kb of the DWnt-4 transcribed region, or the 3.2 kb DWnt-4 cDNA. To detect wg RNA, we used a 1 kb fragment lying in the 5′ coding sequence of the gene. After phosphatase alkaline detection, embryos were mounted in 90% glycerol, 100 mM Tris, pH 7.5 and observed under an AxioPhot Zeiss microscope using Nomarski optics. In situ double labelling experiments were performed according to the method of Röder et al. (1992), using the monoclonal antibody 4D9 (Patel et al., 1989) to visualise the En stripes. In situ hybridisation to a polytene chromosome was performed using biotinylated probes and the Vectastain peroxydase detection kit (Vector Laboratories).

Phage (Maniatis et al., 1978) and cosmid (Hoheisel et al., 1991) Drosophila genomic libraries were screened with clone CU83 and DNA fragments 5′ to wg. Overlapping clones (4 phages and one cosmid) spanning 64 kb were recovered. DWnt-4 RNA was identified on northern blots probed with the 4.2 kb EcoRI fragment located at the left end of the λ83.1 phage (Fig. 1A). We failed to detect any other transcript within the intercalating DNA between wg and DWnt-4. Eight overlapping cDNA clones were recovered by probing a 4 to 8 hour embryonic cDNA library (Brown and Kafatos, 1988) with the 4.2 kb EcoRI fragment.

The longest cDNA clone (3.2 kb) was subcloned in both orientations into the Bluescript vector. Nested deletions of the insert fragments were generated by digestion with ExoIII and the nucleotide sequence (on both strands) was determined using Sequenase 2.0 kit (US Bio-chemicals). A fragment of 277 bp, which corresponds to the 3′ end of DWnt-4 coding sequence, was amplified from genomic DNA using the primers: 5′-CGCAATACACCACGCTGTAC and 5′-AAGTCT-GCGAGAGATGGCAC. The amplified fragment was subcloned in the PCRII plasmid (invitrogen) and sequenced.

The alignment of amino acid sequences of Drosophila Wnt proteins was constructed with the Geneworks Alignment Program (Intelligenetics Inc.) and refined by hand in order to minimise the ‘cost’ given by an algorithm similar to FASTA.

The phylogenetic tree was determined without taking into account the sequences upstream of the first conserved cysteine and obvious inserted sequences. This corresponds in Wg to the deletion of amino acids 1-57, 167-178, 205-208, 293-373. Homothetic deletions were made in the other Wnt sequences. These truncated sequences were then aligned as described above, and a phylogenetic tree constructed using the unweighted pair group method with an arithmetic mean (UPGMA-Tree Program, Geneworks, Intelligenetics Inc.).

The Ubx fusion protein construct in the T7 polymerase expression vector pET3a was kindly provided by R. A. H. White (University of Cambridge). Heparine-agarose purified Ubx protein extracts were separated by 12.5% SDS-PAGE, the fusion protein cut out from the gel and electroeluted. The purified Ubx protein was dialysed for 4 hours against a denaturing buffer (25 mM Hepes, 100 mM KCl, 0.1% NP40, 6 M Urea) and overnight in a renaturing buffer (25 mM Hepes, 100 mM KCl, 1 mM PMSF, 1 mM EDTA, 1 mM DTT and 20% glycerol, pH 7.9), and stored in small aliquots at −80°C.

DNA-protein interactions were performed in 10 mM Hepes, 0.1 mM EDTA, 30 mM KCl, 1 mM MgCl2, 0.1 mg/ml BSA, pH 8.0. Ubx protein was preincubated in a 20 μl reaction volume for 30 minutes on ice. 10⁴ cpm of ³²P-labelled CU83 fragment was added and the incubation continued for 30 minutes on ice. The formation of DNA protein complexes was detected by autoradiography after 5% PAGE (20 mA) performed under non-denaturing conditions (TBE). In super shift assays, FP3.38 monoclonal antibody (50 ng) was added to Ubx fusion protein during the preincubation step of 30 minutes, before the addition of CU83-labelled fragment.

We previously reported a library of genomic DNA fragments associated in vivo with Ubx homeoproteins (Graba et al., 1992). In a further screen of this library, we selected a clone, CU83, which is particularly abundant, representing about 0.8% of the total clones. The nucleotide sequence of CU83 (127 bp) revealed the presence of a motif (underlined in Fig. 1A) containing the ATTA core motif found in most homeodomain binding sites (Desplan et al., 1988). This motif, TTAATTAGC, resembles the consensus TTAAT(G/T)(G/A)CC defined by Ekker et al. (1992) as the optimal DNA sequence recognised by the Ubx homeodomain in vitro and hence suggests that the gene associated with the fragment represents a bona fide target of homeoproteins. Gel shift assays confirm this (Fig. 2) and demonstrate that Ubx protein efficiently binds clone CU83, in vitro.

Genomic DNA from the locus was recovered using CU83 to screen genomic libraries (Fig. 1). Probing northern blots of 5 to 15 hour embryonic mRNA with all the EcoRI fragments of phage λ83.1, identified a single 3.2 kb transcript, 7 kb away from clone CU83 (not shown). Furthermore, by probing an embryonic cDNA library constructed by Brown and Kafatos (1988), we isolated 8 overlapping cDNA clones of the corresponding gene, which we also named CU83.

CU83 was localised, by in situ hybridisation, at 28A on the left arm of chromosome 2 (not shown), i.e. at the same chromosomal position as wg. We mapped it more precisely within a chromosomal walk initiated from probes 5′ to wg and CU83. The translation starts in wg and CU83 are separated by 30 kb of DNA. On the basis of restriction analysis and polymerase chain reaction (PCR) amplification data, two introns of 18 kb and 300 bp were identified and localised in the coding of the gene (Fig. 1B).

The complete nucleotide sequence of the longest cDNA is available from data banks. Comparison of its length (3201 bp) with the size of the mRNA detected by northern blot (3.2 kb) suggests that all of the sequence of the CU83 transcript has been recovered. The longest open reading frame (ORF) extends 1167 bp. The sequence surrounding the first methionine codon of the ORF at nucleotide 1492, 5′-CAACATGG-3′, matches the consensus, 5′-(A/C)AA(A/C)-ATGG-3′, for translation start in Drosophila (Cavener, 1987). This ATG is the first methionine codon preceded in frame by several stop codons in the cDNA sequence (at position 946, 1018 and 1039). An AATAAA poly(A) addition signal is present at position 3147, near the 3′ end of the untranslated region.

The deduced protein sequence, shown together with those of Wg, DWnt-2 and DWnt-3 in Fig. 3A, is highly homologous to the members of the Wnt gene family. Accordingly, the CU83 gene will be now referred to as DWnt-4. A signal sequence (Von Heijne, 1985), from residue 1 to 27, as well as putative sites for N-linked glycosylation predict a secreted glycoprotein. The most significant feature of Wnt proteins is an abundance of strongly conserved cysteine residues (Nusse and Varmus, 1992; Sidow, 1992). Besides the 22 cysteines found in all Wnt sequences, DWnt-4 contains two residues, at position 90 and 103, which exist in all proteins other than Wnt-1 orthologs. It lacks the cysteine unique to Wnt-1 (position 124 in Wg). Another rule among Wnt proteins is that the spacing between the conserved cysteines is largely invariant. In three places in the carboxy-terminal part of DWnt-4 this rule is not satisfied. Relative to the other family members, DWnt-4 has 2 extra amino acids between cysteines 19 and 20 and two gaps of 8 and 4 amino acids between cysteines 13/14 and cysteines 14/15, respectively. We confirmed the primary structure of the C terminus of DWnt-4 by sequencing a genomic DNA fragment spanning the 3′ coding region (from nucleotides 2403 to 2680).

On the whole, the DWnt-4 sequence is 35% identical to the Wg sequence, 32% to DWnt-2, and 33% to DWnt-3, not taking into account the non-conserved amino termini and the inserts found in each protein. Comparison with Wnt proteins from other species gave identity scores also ranging from 30% to 35%. The dendogram of Fig. 3B, which provides a phylogenetic relationship between mouse and Drosophila Wnt proteins, puts DWnt-4 in a branch distinct from the others. Thus, DWnt-4 is not orthologous to a known vertebrate Wnt gene nor paralogous to a Drosophila Wnt gene. Interestingly, out of the 59 predicted C-terminal regions of Wnt proteins (from cysteines 11 to 19) identified by Sidow (1992), Wnt-9 sequences from hagfish (Eptatretus stouti) and shark (Alopius vulpinus) are the only ones that show an unusual structure with gaps at similar positions to those in DWnt-4, between cysteines 13/14 and 14/15 (Fig. 3C). It must be noted that Wnt-9 conservation between species separated by more than 500 millions years (hagfish and shark), means that the genes are functional. While the comparison has so far only been made on partial sequences, DWnt-4 could yet constitute the Drosophila ortholog of vertebrate Wnt-9.

Transcripts are first detected in the ectoderm of stage 8 embryos at the anterior tip and in the trunk, where a segmentally repeated striped pattern becomes visible ventrally (Fig. 4B). Shortly afterwards, expression in the head region is seen in the foregut primordium and dorsal patches that presumably correspond to cells of labral and ocular segment primordia (Fig. 4C). By completion of germ band extension, the segment polarity-like staining in the trunk intensifies and 15 stripes are clearly seen in parasegments (PS) 0 to 14 (Fig. 4C). Double staining for Engrailed (En)/Invected (Inv) proteins as well as wg or DWnt-4 transcripts indicate that stripes are one cell wide and overlap the ventralmost wg cells (Fig. 5).

Before germ band retraction, the DWnt-4 pattern in the trunk ectoderm evolves. Transcripts become visible in the dorsolateral ectoderm of each PS, in a short row of 3-4 cells which parallels the En stripe but which is located at the middle of the PS (arrow in Fig. 5A). At the posterior tip of the germ band, a complex expression is initiated with one additional stripe posterior to the En stripe, which marks the anterior border of PS14, and a strong staining of the telson (Figs. 4F, 5C).

During germ band shortening, the striped staining in the ectoderm slowly diminishes in intensity and fades at the end of the process. By this time, transcripts become detectable in the central nervous system and developing gut. Expression in the nerve cord is not uniform, but instead increases progressively from the anterior to posterior regions of the embryo (Fig. 4G, I). In the gut, transcripts are seen in foregut, hindgut and at two locations in the VM of the midgut (Fig. 4H). The anteriormost domain, which is separated from the foregut by 3-4 unlabelled cells, presumably corresponds to PS4. The second domain was localised through double-label experiments for DWnt-4 transcripts and Ubx or En proteins (not shown). The landmarks provided by Ubx in PS7 of the VM (Bienz and Tremml, 1988) and the En stripes indicated that the posterior-most domain corresponds to PS8. The two domains of DWnt-4 transcription in the VM coincide with primordia of the gastric caeca and the second midgut constriction. Later, the gene is clearly expressed at, and adjacent to, the sites of these differentiative events (Fig. 4J).

Four HOM-C genes, expressed in exclusive parasegmental domains of the VM, control gut morphogenesis: Sex combs reduced (Scr), Antennapedia (Antp), Ubx and abdominal-A (abd-A) in posterior PS3 to PS4, PS5 to PS6, PS7, and from PS8 to PS12, respectively (Tremml and Bienz, 1990; Reuter and Scott, 1990). Since DWnt-4 was originally identified as a putative target of Ubx, we investigated whether its transcription in the VM might depend on Ubx and other HOM-C gene activities.

In embryos deficient for Ubx or abd-A, expression in PS8 is absent (Fig. 6B,C). Thus, Ubx in PS7 and abd-A in PS8 are necessary for DWnt-4 activation in PS8. This is reminiscent of the regulation of wg by HOM-C genes in the VM, and suggests a positive and direct control by abd-A and a positive but indirect control by Ubx. We examined whether the secreted Dpp protein, which mediates the action of Ubx on wg in PS8 of the VM (Immerglück et al., 1990), might also control DWnt-4 expression. In the shortvein class of mutations (dpp^shv), dpp expression, which normally occurs in PS4 and PS7 of the VM, is disrupted either in one or both domains (Hursh et al., 1993). In dpp^shv6 embryos, which lack dpp activity in PS7, we fail to detect DWnt-4 transcripts in PS8 (Fig. 6D). In the dpp^shv4 embryos, which lacks dpp expression in PS4 and in PS7, transcripts are no longer detected in the VM, either in PS4 or in PS8 (Fig. 6E). dpp is therefore required for DWnt-4 activation not only in PS8, as it is for that of wg, but also in the anterior part of the midgut (PS4). Scr is dispensable for DWnt-4 regulation since in the null mutant Scr^W17, the gene is still expressed in PS4 (not shown).

In Antp deficient animals, DWnt-4 becomes ectopically expressed in PS5 and PS6 (Fig. 6F), indicating a negative regulation by Antp protein in this portion of the VM. Regulatory interactions between HOM-C genes occur in the VM as well as in the epidermis (Bienz and Tremml, 1988; Reuter and Scott, 1990); for instance, Ubx sets the posterior border of Antp expression such that, in Ubx mutants, Antp is derepressed posteriorly in PS7. We asked whether the ectopic expression of Antp could silence DWnt-4 in PS7 of Ubx homozygotes. If so, Antp would replace Ubx as a repressor of DWnt-4 in this PS; in other words, the derepression of DWnt-4 in PS7 resulting from the absence of Ubx activity would be masked by the ectopic accumulation of Antp. One would therefore expect DWnt-4 to be expressed not only in PS5-6 but also in PS7 of embryos deficient for both Antp and Ubx functions. Indeed, transcription in these embryos extends posteriorly to include PS7 (Fig. 6G). Note that expression is clearly absent in PS8, which is likely to be due to the lack of dpp activation by Ubx. Thus, ectopic Antp is able to repress DWnt-4 in PS7 of embryos lacking Ubx function, a strong indication that the down-regulation, in PS7 of wild-type embryos, is achieved by Ubx. Comparison of DWnt-4 patterns in PS7 and PS8 of Antp, Ubx and Antp-Ubx mutants, illustrates the dual regulatory function of Ubx which in the wild type turns DWnt-4 off in PS7 and on in PS8.

In abd-A mutants, Ubx expression extends to the posterior end of the midgut (Tremml and Bienz, 1989). To test whether this ectopic expression could have an effect, we analysed the DWnt-4 pattern in embryos deficient for both Ubx and abd-A. In such embryos, as the ectopic expression of Antp does not extend back to PS7 (Bienz and Tremml, 1988), none of the HOM-C genes are expressed from PS9 to PS12 of the VM. We found that DWnt-4 became ectopically expressed from PS9 to PS12 (Fig. 6H). The simplest explanation is that abd-A silences DWnt-4 in this portion of the VM of wild-type embryos and that, in the absence of abd-A function, repression is achieved by Ubx. These data clearly illustrate that abd-A also achieves a dual regulation of DWnt-4 in the VM, positive in PS8 and negative from PS9 to PS12.

The strong and complex expression of DWnt-4 seen in the ectoderm at the posterior tip of the germ band (Fig. 4F) as well as in the posterior region of the nerve cord during germ band retraction (Fig. 4I), seems to be related to that of Abdominal-B (Abd-B). Analysing these patterns in several mutant backgrounds lacking the Abd-B function (Abd-B⁻, abd-A⁻-Abd-B⁻, Antp⁻-Ubx⁻-abd-A⁻-Abd-B⁻), we do not observed any change relative to the wild type. It appears therefore that firstly, Abd-B does not play on DWnt-4 in the caudal region of the embryo, and secondly, regulation by homeotics takes place in one germ layer only, the VM.

The regulatory network composed of dpp and HOM-C genes that controls DWnt-4 expression in the VM is schematised in Fig. 7.

The identification of genetic functions acting downstream of HOM-C genes is of central importance in the understanding of the homeotic control of regional identity and morphogenesis. The immunoselection of DNA fragments bound to homeoproteins in vivo constitutes an approach specially designed to clone direct effectors of homeotic genes. DWnt-4, isolated from a further screen of a library of Ubx-associated DNA fragments (Graba et al., 1992), presents several features shared by most of the putative targets identified so far (Botas, 1993). First, the gene is regulated by several HOM-C genes; its pattern of expression in the VM results from regulatory interactions between the homeotic genes active in this germ layer. Second, in keeping with the expression pattern, DWnt-4 presumably is involved in developmental processes other than downstream of HOM-C genes. Third, the identification of DWnt-4 as a putative target of homeotic proteins is a further indication that effector genes encode molecules involved in growth and cell fate decisions rather than, so far, housekeeping functions.

DWnt-4 encodes a protein with features characteristic of Wnt gene products. First, the amino-terminal signal sequence and prospective sites for N-linked glycosylation predict a glycoprotein destined for secretion. Second, the 24 cysteines in the DWnt-4 sequence correspond to conserved residues found in all Wnt proteins other than Wnt-1. Based on the conservation of cysteines as a criterion for the relatedness of the Wnt genes, DWnt-4 is less related to Wnt-1, Wg in Drosophila, than to the other Wnt proteins. In terms of overall conservation of amino acid residues, homology scores with the other family members, from Drosophila as well as from widely divergent species, range around 30-35%. It appears therefore that neither a paralog nor an ortholog of this gene exists within the current repertoire of full-length Wnt sequences. Comparison with 59 Wnt partial sequences revealed that the C terminus of vertebrate Wnt-9 and DWnt-4 show the same unusual structure, with gaps at similar positions between conserved cysteines. This suggests that the corresponding genes could be orthologous. The connection with vertebrate Wnt-9 indicates that DWnt-4 presumably does not arise from a gene duplication in some ancestral arthropod, but much earlier, before the divergence of deuterostome and arthropod lineages.

Several loci consisting of related genes have been described. Generally, expression patterns are virtually identical or largely overlapping and the predicted proteins show considerable homology, at least in the functional domains; in some cases, it has been demonstrated that the genes are functionally related (e.g. Rothe et al., 1992; Grossniklaus et al., 1992; Zhang et al., 1994). wg and DWnt-4 are not paralogous Wnt genes and have distinct intron-exon structures, consistent with the previous proposal that the genes diverged early in evolution. Their expression profiles, while related, differ in several respects. Both genes are transcribed in the same ventral cells of the posteriormost ectoderm of each PS, in PS8 of the VM, in the head region and in foregut and hindgut primordia; but they are also found in specific domains: wg and DWnt-4 in distinct cell rows of the dorsolateral ectoderm, wg in Malphigian tubules, and DWnt-4 in the nerve cord and PS4 of the VM. The close physical linkage, coexpression and coregulation suggest that the two Wnt genes share cis-control elements and belong to the same gene networks in developmental processes. Supporting this view, we isolated several EMS induced regulatory mutations that affect expression of the two genes; moreover, while the absence of an amorphous DWnt-4 mutation prohibits us from conclusively demonstrating a function in segmental patterning and gut morphogenesis, we identified a hypomorphic DWnt-4 allele which has a segment polarity phenotype, indicating a function in embryonic epidermis (not shown).

The homeotic proteins Antp, Ubx and Abd-A control, directly or indirectly, DWnt-4 expression in several portions of the VM. This regulatory network, which also includes Dpp, leads to DWnt-4 expression at two locations in the VM where major morphogenetic events occur: the formation of the gastric caeca and the second midgut constriction.

Evidence for a direct relationship exists only for Ubx and clone CU83: the clone has been selected from the library of Ubx targets, it presents an ATTA nucleotide motif required for homeodomain association and actually binds Ubx in vitro. Accordingly, the interaction between Ubx and the 127 bp element could be responsible for the DWnt-4 silencing observed in vivo in PS7 of the VM. The gene is also downregulated by Antp and abd-A in PS5-6 and PS9-12, respectively. Moreover, ectopic Antp in PS7 and Ubx in PS9-12 repress DWnt-4. Comparable situations, where HOM-C proteins can be replaced by one another to achieve repression, have been described: Ubx can substitute for Abd-A, and Abd-A for Abd-B in the down-regulation of Dll (Vachon et al., 1992) and scabrous (Graba et al., 1992) in abdominal segments. Analysing the repression of the Antp P2 promoter by Ubx and Abd-A, Appel and Sakonju (1993) proposed a model where inactivation results from a steric blockade due to the fixation of HOM-C gene products on homeodomain binding sites, rather than from specific interactions with transactivating factor(s). The model provides a simple and attractive explanation as to the interchangeability of HOM-C proteins in the repression of a common target.

Ubx and Abd-A are required for DWnt-4 expression in PS8. Ubx acts indirectly; it activates dpp in PS7 (Capovilla et al., 1994), which in turn stimulates DWnt-4 in PS8. Two scenarios are possible regarding the role of abd-A. In the first, transcriptional activation would only require the dpp signal received from PS7, and Abd-A protein would have no direct and positive effect on DWnt-4. In abd-A mutants, the ectopic accumulation of Ubx would be sufficient to account for the silencing seen in PS8. Thus, the function of abd-A would serve only to prevent Ubx expression posterior to PS7. Alternatively, DWnt-4 transcription in PS8 depends on the coordinate action of Dpp and Abd-A, as does wg. In that case, the apparent paradox that Abd-A acts as both an activator (PS8) and a repressor (PS9-12) is not resolved, which could suggest that a co-factor determines the specificity of action of Abd-A. For example, interaction of Abd-A with a PS8-specific protein could be required for the activation of DWnt-4. The absence of the co-factor in more posterior PS would then lead Abd-A to act as a repressor. Growing evidence for the existence of co-factors for homeodomain proteins is currently emerging (Manak and Scott, 1993).

Ubx and abd-A control gut morphogenesis by driving the expression of secreted signalling factors in two adjacent domains at the boundary of which the second midgut constriction will form. The functions of Dpp and Wg in this morphogenetic event have been demonstrated (Bienz, 1994). The data reported in this paper indicate that a third signalling molecule, DWnt-4, whose expression in PS7-8 depends on Ubx and abd-A, could also be involved in this process. The isolation of mutants deficient for DWnt-4 in the VM is a requisite for addressing this point.

We thank C. Goridis, B. Jacq, S. Kerridge, A. Martinez Arias, R. Rosset and J. Singer for critical reading of the manuscript and for helpful discussions and suggestions. We are especially grateful to A. Sidow for his comments about vertebrate Wnt-9 and DWnt-4 and to G. Chen, J. Manak, A. Martinez Arias and R. Nusse for DNA probes from the wg locus. We acknowledge J. D. Hoheisel for providing high-density filters of cosmid libraries and clone JHD4 41 C 6. This work was supported by the CNRS and grants from ‘l’Association pour la Recherche sur le Cancer’ and ‘Ministère de la Recherche et de la Technologie’ to J. Pradel. K. Gieseler is fellow of the Boehringer Ingelheim Fonds.

The databank accession number for the DWnt-4 cDNA sequence is L25316.