The combgap locus encodes a zinc-finger protein that regulates cubitus interruptus during limb development in Drosophila melanogaster

Signaling between distinct cell populations often establishes organizing centres that pattern embryonic fields during animal development (Lawrence and Struhl, 1996). One of the best examples of such an embryonic field is the anteroposterior (A/P) axis of fly limbs (reviewed in Blair, 1995; Brook et al., 1996), where axial patterning is controlled by a series of cell interactions, beginning with signaling between cells of the anterior and posterior lineage compartments. Hedgehog protein (HH) is secreted by posterior cells and establishes a specialized group of anterior cells that act as an organizing centre to direct growth and patterns of cell differentiation throughout the A/P axis (Basler and Struhl, 1994; Tabata et al., 1995; Zecca et al., 1995).

The compartment specific expression of cubitus interruptus (ci) and engrailed/invected (en/inv) is critical for the regulation of both hh and HH-responsive genes in the elaboration of A/P pattern. The expression of the en/inv homeobox genes in posterior cells results in both the adoption of posterior fate and the repression of ci transcription (Dominguez et al., 1996; Tabata et al., 1995). In turn, the resulting anterior-specific expression of ci, a zinc-finger transcription factor related to the vertebrate Gli proteins, represses the transcription of the hedgehog gene (hh), resulting in posterior only expression of the HH secreted signaling molecule (Aza-Blanc et al., 1997; Dominguez et al., 1996; Hepker et al., 1997). The spatially restricted expression of HH-responsive genes that leads to the formation of the A/P organizer is the result of the combined action of EN/INV and the activating and repressing forms of the CI transcription factor, CI-155 and CI-75.

HH signaling controls the post-transcriptional modification of the full-length 155 kDa form of CI. In anterior cells outside the range of the HH signal, CI-155 is cleaved to a 75 kDa form, CI-75, which acts as a transcriptional repressor of the morphogen encoded by decapentaplegic (dpp) (Aza-Blanc et al., 1997). In the stripe of anterior cells adjacent to the A/P boundary that receive the HH signal, the 155 kDa form is not cleaved and is activated by HH signaling (Methot and Basler, 1999; Ohlmeyer and Kalderon, 1998). The activated form (CI-155*) induces the expression of the HH target genes including dpp, wingless (wg), patched (ptc), knot (kn) and vein (Aza-Blanc et al., 1997; Methot and Basler, 1999; Mohler et al., 2000; Ohlmeyer and Kalderon, 1998; Vervoort et al., 1999; Wessells et al., 1999). EN/INV repress the expression of dpp in the posterior lineage compartment (Sanicola et al., 1995; Tabata et al., 1995). In leg imaginal discs, CI-155, CI-75 and EN/INV regulate dpp in dorsal cells. In ventral cells, wg is regulated through similar mechanisms (Hepker et al., 1997; Sanicola et al., 1995). The local induction of organizer activity in anterior cells near the A/P compartment boundary is the result of three processes: (1) repression of morphogen genes in anterior cells outside the range of the HH signal by CI-75; (2) activation of the same genes in cells of the organizer by CI-155; and (3) repression of morphogen genes by EN/INV in posterior cells (see Fig. 1).

The segmentation-gene hierarchy establishes the initial pattern of en/inv and ci gene expression in the limb primordia (Eaton and Kornberg, 1990; Orenic et al., 1990). However, little is known about how the compartment-specific expression of ci and en/inv is maintained throughout limb development. EN has been shown to be a direct repressor of ci transcription in the posterior compartment (Schwartz et al., 1995), and some members of the Polycomb-Group (Pc-G) of transcriptional repressors have been shown to maintain the spatially restricted patterns of ci and en/inv (Busturia and Morata, 1988; Maschat et al., 1998). Other factors are likely to be required to maintain the A/P dichotomy of ci and en/inv and one candidate is combgap (cg). Past studies have described loss or duplication of anterior or posterior structures in wings and legs caused by the cg¹ mutation (Bridges cited in Lindsley and Zimm, 1992). Some of these defects are similar to those caused by mutations in ci or en/inv. Moreover, cg interacts genetically with mutations in both genes (House cited in Lindsley and Zimm, 1992), suggesting an important role for cg in the genetic control of the A/P axis. To study these novel aspects of mechanisms underlying the anterior versus posterior distinction in limb development, we undertook the characterization of the cg locus. We report here that cg encodes a zinc-finger chromosomal protein required for the correct regulation of ci in limb development.

Flies carrying the cg¹ allele (Lindsley and Zimm, 1992) and cg² allele (also known as l(2)07569) a PZ insertion mapped to 50E1,2) were obtained from the Bloomington Stock Center and the Berkeley Drosophila Genome Project (BDGP). The cg³ mutation was a second-site mutation generated in a P-excision screen on a PZ{lacZ;ry⁺}H15, b cn chromosome (W. J. B., unpublished). The lacZ reporter is intact in the strain but the ry⁺ is mutated resulting in a PZ{lacZ;ry^-}H15, b cn cg³ chromosome. The mutant was mapped to meiotic position 2-71, based on 17 recombinants between purple (pr) (2-54.5) and curved (c) (2-75.5). The cg³ mutant failed to complement cg¹ and cg². The cg²ptc-gal4, the cg¹ptc-gal4 and the PZ{lacZ; ry⁺}dpp-lacZ cn cg² chromosomes were generated for this study.

Genomic DNA flanking the PZ insertion in cg² was cloned by plasmid rescue (Mlodzik et al., 1990) of the bacterial plasmid in the PZ transposon. A 7.9 kb fragment flanking the insertion site was cloned and used to screen a lambda Dash II library (Stratagene) (Ng et al., 1995) to isolate several genomic phages representing DNA on both sides of the insertion. Genomic fragments were used to screen embryonic and imaginal disc cDNA libraries. One cDNA truncated at the 5′ end represented a C₂H₂ zinc-finger protein; database searches identified a cDNA from the Drosophila EST collection (LD05357) that represents a full-length cDNA corresponding to the zinc-finger protein.

Full phage inserts as well as a 10 kb subcloned fragment were cloned into the Casper4 P-element transformation vector (Thummel and Pirrotta, 1991) and transformed into w¹¹¹⁸ flies. Several independent transformants of each construct were crossed into the cg¹ background and tested for their ability to rescue the cg phenotype. All four lines were able to rescue the cg phenotype. The rescuing fragments P{103} (−8 kb to +7.8 kb), P{104} (−9.5 kb to +4 kb), P{105} (−14.5 kb to+1.2 kb), and P{111} (−8.5 kb to +1.2 kb), define a 9.2 kb region sufficient for wild-type combgap function (−8 kb to +1.2 kb relative to the P-insertion point of cg²). For the ci rescue experiment, larvae of the genotypes w; dpp-lacZ cg²/ptc-gal4 cg¹;+/+ and w; dpp-lacZ cg²/ptc-gal4 cg¹;UAS-ci/+ were generated either as siblings from the same cross or in parallel crosses.

pGEXN61 contains a 495 bp EcoRI/NsiI fragment from cDNA clone L05357 cloned downstream and in-frame with the GST coding sequences. The EcoRI/NsiI fragment is predicted to encode the first 164 amino acids of the combgap protein. The same fragment of the combgap cDNA was cloned in-frame downstream of the His-tag coding region of the pET28a vector generating QEN1 (cloning details available upon request). New Zealand White rabbits were immunized with gel purified GST:CG fusion protein (from pGEX4T1). Antibodies to CG were purified from polyclonal sera using the purified His-tag:CG fusion protein (from QEN1) coupled to CNBr-activated Sepharose 4B. Crude and purified sera produced the same patterns in imaginal disc and chromatin staining. Unpurified polyclonal sera were used for most immunofluorescence and immunohistochemistry protocols.

Immunohistochemistry and detection of β-galactosidase activity in imaginal discs were carried out as described previously (Brook et al., 1996). Rabbit anti-CG primary was used at a dilution of 1:200, mouse anti-WG (4D4) was used at 1:10, mouse anti-EN (4D9) was used at 1:50, mouse anti-PTC (2E9) was used at 1:100, rabbit anti-KN was used at 1:200. Biotinylated goat anti-mouse and goat anti-rabbit secondary antibodies were used at 1:500, followed by histochemical staining using the Vectastain ABC kit (Vector Laboratories). Stocks of cg mutants balanced over T(2;3)SM6-TM6, Cy Hu Tb were crossed to allow identification of homozygous or hetero-allelic cg mutant combinations as Tb⁺ larvae.

Salivary gland chromosomes were prepared essentially as described by Decamillis et al. (1992). Purified anti-CG primary antibodies were used at 1:40 dilution. Secondary antibodies (biotinylated goat anti-rabbit, Jackson Labs) were used at 1:200 dilution and Texas Red conjugated streptavidin (Jackson Labs) was used at 1:1000 in PBS to visualize positive bands. Chromosomes were counterstained with 4′,6-diamidino-2-phenylindole (DAPI, 1 μg/ml in PBS).

We undertook the molecular characterization of cg because the patterning defects and genetic interactions displayed with ci and en/inv suggested a role for cg in A/P limb development. We identified a recessive lethal mutation that displayed prominent leg defects in lethal pharate adults. Genetic mapping and complementation tests showed that it was allelic to both the original cg¹ allele and l(2)07659 (hereafter called cg²), a P-element insertion from the BDGP collection (see Materials and Methods). DNA flanking the cg² insertion was cloned into the CaSpeR4 P-element transformation vector (Thummel and Pirrotta, 1991), transformed into w¹¹¹⁸ flies and tested for their ability to rescue cg. All four constructs rescued the cg¹/cg¹ adult wing, leg, antenna and fertility phenotypes (Fig. 2A). The genomic fragments rescuing the cg phenotype define a 9.2 kb region, from 8 kb to the left of the cg² P-insertion extending 1.2 kb to the right, necessary for rescue (Fig. 2B). Southern blot analysis of cg¹ and cg³ genomic DNA using probes from the rescuing region indicated that the cg¹ allele was associated with a >6 kb insertion and cg³ was associated with a >4.3 kb insertion in the rescuing region (Fig. 2B).

Sequence analysis of the genomic rescuing region (AF276663) with the gene-finder software Genie (BDGP, www.fruitfly.org) predicted that a putative zinc-finger protein was the only coding gene within the region. Database searches identified a Drosophila EST (LD05357) (AF276664) representing a cDNA corresponding to the zinc-finger protein that included a 5′ UTR and an open reading frame of 671 codons. The predicted open reading frame of LD05357 contains 11 canonical C₂H₂ zinc-finger sequences (C-X_1-2-C-X₃-F-X₅-L-X₂-H-X_3-4-H) typical of many DNA-and chromatin-binding proteins (Fig. 2C) (Coleman, 1992). A glutamine-rich region (19/24 residues) spans residues 569 to 592. We find no significant matches to other proteins outside the zinc-finger and glutamine-rich motifs. GST fusion proteins of the N-terminal 164 amino acids of the predicted CG protein were expressed in Escherichia coli and injected into rabbits to produce anti-sera. The affinity-purified anti-sera recognized a single band at approximately 75 kDa in western blots of imaginal discs extracts (data not shown). Immunolocalization of the protein using three independent anti-sera showed uniform CG expression in all imaginal discs and in many larval tissues (Fig. 3), as well as uniform expression in embryos (data not shown). As was most clearly seen in the peripodial membrane of imaginal discs (Fig. 3B) and in larval tissues, such as the fat body (Fig. 3D), the protein was enriched in the cell nuclei in all tissues.

To confirm that the protein recognized by the anti-sera corresponded to CG, we examined the staining in cg mutants. Animals homozygous for cg² or cg³ are lethal with most cg³/cg³ animals surviving to the pharate adult stage and very few cg²/cg² animals surviving to third larval instar stage. Approximately 10% of cg¹/cg², and 50% of cg¹/cg³ trans-heterozygous animals eclose to adult stages with most of the remaining animals dying as pharate adults. Flies carrying cg²/cg³ survive to pharate adult stage but never eclose. Thus cg² appears to have the least activity while cg¹ has the most. The protein was found to be slightly reduced in cg¹/cg¹ tissues (not shown), and was reduced to background levels in cg³/cg³ (Fig. 3F) and cg²/cg³ tissues (Fig. 3C), supporting the conclusion that the predicted C₂H₂ zinc finger is indeed the product of the cg locus.

In order to understand how the ubiquitously expressed CG protein acts in A/P limb patterning, we examined the effect of cg mutations on both morphological and molecular markers of limb development. Wings from the three cg genotypes producing viable adults (cg¹/cg¹, cg¹/cg² and cg¹/cg³) had defects in some combination of the posterior compartment longitudinal veins 4 and 5 (L4 and L5) and the anterior compartment longitudinal vein 2 (L2). Defects in vein 4 (L4), ranging from a slight thinning of the vein to gaps of increasing size (Fig. 4B), were seen in 100% of wings from all hetero-allelic combinations. There was a higher penetrance of posterior compartment (L4 and L5) defects in cg¹/cg¹ flies, which is predicted to be the weakest mutant combination, based on both viability and leg phenotypes (described below) (see Fig. 4). Wings from the stronger cg¹/cg² mutants had a higher incidence of anterior compartment defects (see Fig. 4). More rarely (<1%), all three genotypes displayed outgrowths in the costal region of the anterior wing margin (Fig. 4B). These outgrowths were reminiscent of the pattern triplications caused by mutations in ci and other genes affecting HH signaling (Dominguez et al., 1996; Lepage et al., 1995; Phillips et al., 1990; Simpson and Grau, 1987).

The L4 gap phenotype is also seen in loss-of-function en alleles and in gain-of-function ci regulatory alleles, both of which result in ectopic posterior ci expression (Eaton and Kornberg, 1990; Slusarski et al., 1995). Vein L4 gaps have also been reported in flies heterozygous for both cg¹ and either en or ci alleles, as well as compound heterozygotes of en and ci (House cited in Lindsley and Zimm, 1992). We found that compound heterozygotes of all three cg alleles with Df(2R)en¹¹– a small deficiency removing engrailed and invected (Sanicola et al., 1995) – had the L4 gap phenotype (Fig. 4D and data not shown). As noted by House for cg¹ (cited in Lindsley and Zimm, 1992), flies heterozygous for cg² or cg³ and ci^W had an enhancement of the vein L4 defects observed in ci^W/+ (Fig. 4F and data not shown). Many of these lack vein L4 completely (Fig. 4F) and thus resemble ci^W/ci^W homozygous wings (Slusarski et al., 1995). The dominant interactions between ci, en and cg suggest that the ectopic ci expression may be due to the disruption of EN-mediated repression of ci in posterior cells, and that cg plays a role in this regulation of ci by EN.

To clarify the relationship between cg, ci and en further, we examined the expression of ci and en in cg mutant wing imaginal discs. Using the 4D9 antibody that recognizes both EN and INV (Patel et al., 1989), we found that the expression of EN/INV was normal in cg mutant discs (data not shown). To assess the effects of cg on ci expression, we crossed ci-lacZ reporter constructs into cg mutant backgrounds. We found weak ectopic expression of ci-lacZ (Eaton and Kornberg, 1990) in the posterior compartment and reduced anterior ci-lacZ compared with wild-type (compare Fig. 5A and B). To confirm these results and exclude the possibility of an artefact that was due to the reporter construct, we stained cg imaginal discs using antibodies to CI proteins. We were able to distinguish between the different forms of CI using either an N-terminal-directed antibody that recognizes both CI-155 and CI-75 (Schwartz et al., 1995), or an antibody directed against the C terminus that is specific for CI-155 (Aza-Blanc et al., 1997; Motzny and Holmgren, 1995). The levels of CI-155 were reduced in the anterior compartments of cg mutant discs and CI-155 was ectopically expressed in the posterior (Fig. 5C,D). This is also true of overall CI levels detected with the N-terminal-directed antibody. The effects were more pronounced in stronger allelic combinations. For example, in cg¹/cg¹ (Fig. 5F) homozygotes, there was less of a reduction of CI in the anterior compartment but more ectopic expression of CI in the posterior compartment, when compared with the strongest allelic combination, cg²/cg³ (Fig. 5G). Thus, effects of cg mutants on the expression of ci are very similar to those described for gain-of-function ci alleles that are disrupted in a regulatory element 5′ of the ci gene (Fig. 5H) (Dominguez et al., 1996; Slusarski et al., 1995). These mutations also result in the ectopic expression of ci in the posterior compartment and diminished expression of ci in cells of the anterior wing compartment. Given these results and the similarity of the phenotypes of the cg and ci gain-of-function alleles, it seems most likely that the effects of cg on wing patterning are due to the misregulation of ci. The wild-type function of cg in wing development appears to be different in the two compartments. CG is required for activation of ci in anterior cells, and, probably in conjunction with EN, is required to repress ci in the posterior compartment.

How does the misexpression of CI in the cg mutants affect the regulation of hh and HH-responsive genes ptc, dpp, and kn? In even the strongest cg mutant backgrounds, hh-lacZ (Fig. 6A,B) (Lee et al., 1992) and dpp-lacZ (Fig. 6D,E) (Jiang and Struhl, 1995) were expressed in the wild-type pattern. The transcription of both hh and dpp are normally repressed in the anterior compartment by CI-75 (Aza-Blanc et al., 1997; Methot and Basler, 1999) and thus sufficient levels of CI-75 must be present in cg mutants for the repression of both hh and dpp. The occasional anterior bifurcations seen in cg mutant wings may be due to rare (<1%) ectopic expression of either dpp or hh and may thus have been missed in our sampling.

Unlike dpp and hh, two HH-responsive genes activated by CI-155 in the intervein region between veins L3 and L4 were altered in cg mutants. The ptc gene is normally expressed weakly throughout the anterior compartment of wing discs, except for an elevated level of expression in the stripe of anterior organizer cells adjacent to the A/P compartment boundary (Fig. 6G) (Phillips et al., 1990). In cg mutants, we found that PTC was ectopically expressed in the posterior cells (Fig. 6H), consistent with observed ectopic CI expression in cg mutants. The kn locus encodes a transcription factor required for the suppression of vein formation in the 3/4 inter-vein (Mohler et al., 2000; Vervoort et al., 1999). kn expression is usually restricted to the stripe of anterior cells adjacent to the A/P boundary in the wing pouch (Fig. 6C). Using an anti-KN antibody (Crozatier et al., 1999), we found that, like ptc, this protein was expressed ectopically in posterior cells of cg mutant imaginal wing discs (Fig. 6F,I). In general, we found that the levels of both ectopic and endogenous KN decreased with increasing strength of the cg allelic combination used, which in turn probably reflects altered ci expression. In cg¹/cg¹ homozygotes, the levels of staining in the endogenous domain of KN expression was similar to wild type, with weaker, ectopic staining observed in posterior cells (Fig. 6F). In stronger allelic combinations such as cg³/cg³ and cg²/cg³, we observed lower levels of both endogenous and ectopic KN (Fig. 6I). This may reflect the lower levels of both endogenous and ectopic CI detected in stronger cg allelic combinations. The relatively high levels of CI and KN expression in the posterior compartments of wings from the weaker cg¹/cg¹ mutants may also explain the higher incidence of posterior compartment vein defects in this background (Fig. 4).

We examined the leg phenotypes of all allelic combinations (except for cg² homozygotes, which die prior to leg differentiation). All combinations were found to exhibit an expansion of anterior dorsal and anterior ventral pattern elements, most notably in the tibia and in the tarsae. For example, two ventral anterior markers, the sex combs and the transverse rows, were abnormal in the male prothoracic legs in all cg mutants. The sex comb abnormalities range from a few extra bristles in cg¹/cg¹ through to highly disorganized structures consisting of as much as a threefold increase in the number of sex comb teeth in cg²/cg³ legs (Fig. 7B,C). The transverse rows were similarly affected. Duplications of dorsal structures such as the pre-apical bristle and occasional leg bifurcations were also observed (Fig. 7C,D). These overgrowths of ventral and dorsal anterior structures and pattern duplications are consistent with the changes in dorsal and ventral molecular marker gene expression observed in cg mutants (described below). Owing to variability of the observed phenotypes, it was difficult to order the severity of the phenotype; however, cg¹/cg¹ clearly had the mildest defects, followed by cg¹/cg² and cg¹/cg³, which had intermediate defects, followed by cg³/cg³ and cg²/cg³, which were the most severe. This is the same order of allele strength as predicted by viability.

The effects of cg on wing development are very similar to the effects of gain-of-function ci alleles. However, even the mildest leg defects seen in cg mutants are more severe than the ci^W gain-of-function mutants, which have normal legs (Schwartz et al., 1995; Slusarski et al., 1995). Using the N-terminal-directed antibody that recognizes CI-155 and CI-75, we observed a decrease in overall CI levels in leg imaginal discs in the weakest cg mutant combinations and substantial decreases in the stronger cg allelic combinations (Fig. 8B and data not shown). The overall reduction of anterior CI staining in cg²/cg³ legs was less severe than that observed in cg²/cg³ wings (compare Fig. 5G with Fig. 8B) and, in contrast to wing discs, we have not observed ectopic CI staining in posterior legs.

Can the changes in CI levels also explain the observed defects in A/P patterning? We examined the expression of EN/INV as well as the ci targets hh, ptc, wg and dpp in leg imaginal discs. No ectopic expression of posterior markers was observed using anti-EN/INV, hh-lacZ or anti-HH sera (Tabata and Kornberg, 1994; data not shown). Leg imaginal discs from cg mutants were expanded in the A/P axis, and staining of hh-lacZ (Fig. 8C,D) or EN/INV (not shown) in cg²/cg³ leg discs show that the expansion was largely due to a substantial increase in the relative size of the anterior compartment. This overgrowth is associated with the ectopic expression of ptc, dpp and wg. There is increased expression of PTC (Fig. 8G,H) and ectopic expression of dpp-lacZ (Fig. 8E,F; see also Fig. 9A) throughout the anterior compartment of cg²/cg³ mutant leg discs. Using a monoclonal antibody to WG, weak ectopic expression was observed in the anterior ventral quadrant of cg²/cg³ mutant leg discs (Fig. 8I,J). To confirm that there was an expansion of WG signaling, we examined the expression of H15-lacZ, a target of WG regulation in the ventral leg that is normally expressed in a ventral wedge centred around the WG domain (Fig. 8I) (Brook and Cohen, 1996). We found that the expression of H15-lacZ expanded throughout the anterior ventral quadrant of the cg³/cg³ leg imaginal discs (Fig. 8K,L).

The regulation of HH-responsive genes is very similar in leg and antennal development (Brook and Cohen, 1996; Diaz-Benjumea et al., 1994). Antennae in cg mutants have an overgrown phenotype reminiscent of the defects seen in cg mutant legs and similar ectopic expression of ptc, dpp, wg and H-15 (data not shown).

A likely explanation for the observed effects on the spatial distribution of dpp and wg in combgap mutant legs and antenna is that there is insufficient CI-75 to repress the two genes ectopically expressed in anterior cells. This has been reported for loss-of-function mutations in ci (Methot and Basler, 1999). We tested this directly by seeing if additional ci expression in cg mutants using a transgene could rescue cg leg phenotypes (Brand and Perrimon, 1993). As PTC expression throughout the anterior compartments of cg mutant legs was still present, and indeed somewhat elevated (Fig. 8H), we used ptc-gal4 (Speicher et al., 1994) as a tool to further increase ci expression in its normal domain using UAS-ci (Methot and Basler, 1999). The expression of dpp-lacZ in discs of the genotype ptc-gal4 cg¹/dpp-lacZ cg²; UAS-ci/+ was compared with ptc-gal4 cg¹/dpp-lacZ cg² controls. Control flies had increased expression of dpp-lacZ in the anterior compartment and exhibited the previously observed overgrowth of the disc in the A/P axis (Fig. 9A), while flies carrying the UAS-ci transgene showed no upregulation of the dpp-lacZ reporter and had normal disc morphology (Fig. 9B). Control flies lacking the UAS-ci transgene had the leg overgrowth typical of cg¹/cg² (Fig. 9C), while flies from the same cross carrying the UAS-ci transgene were similar to the milder phenotype seen in cg¹/cg¹ flies, the weakest cg mutant combination (Fig. 9D). A similar rescue was observed of the cg antennal phenotype and WG and dpp-lacZ expression in antenna (data not shown).

The preceding results show that the effects of cg mutations on limb development can be largely if not completely explained through misregulation of ci. Does the zinc-finger protein encoded by cg directly regulate ci expression? Recombinant CG protein produced in E. coli did not show specific binding to the 5′ region of the ci gene using gel mobility-shift assays (S. D. G. M., M. K. and W. J. B., unpublished). However, we examined the distribution of CG in the polytene chromosomes of salivary glands and found dozens of strong, discrete binding sites on each of the five major chromosome arms, and many more weak sites of CG binding (Fig. 10A,B). This staining was not observed in the chromosomes of cg²/cg³ mutants (data not shown). Of particular interest, a band of CG binding was found on chromosome 4 at approximately 101F, the cytological location of the ci gene (Fig. 10D,E). This suggests that the CG protein may be a direct regulator of ci transcription. Attempts to show that CG localizes to ci regulatory elements in transgenes inserted at ectopic chromosomal locations have been confounded by the large number of endogenous CG-positive bands (P. C. S. and W. J. B., unpublished).

Transcriptional control of the ci gene and post-translational regulation of the CI protein are essential for the placement and function of the A/P organizer in limb development. The action of the HH-signaling pathway results in a relatively high ratio of CI-155 to CI-75 at the A/P boundary, which ensures the localized expression of HH-responsive genes necessary for the formation of the A/P organizer. HH signaling influences the post-translational modification of the CI protein at multiple levels including proteolytic cleavage, phosphorylation, subcellular location and nuclear import (reviewed in Aza-Blanc and Kornberg, 1999). In contrast to the details of post-translational modification of CI-155 by elements of the HH-signaling pathway, relatively little is known about the transcriptional control of ci. We observed decreased levels of both CI protein and expression from ci-lacZ reporter constructs in the anterior compartments of cg mutant imaginal discs, as well as ectopic expression in posterior cells of wings. Thus, CG affects both the activation and repression of ci transcription.

The dominant interaction between cg and en/inv mutations that gives rise to the gap in vein L4 strongly suggests that CG and EN/INV act together to repress posterior ci transcription. Posterior expression of EN represses the transcription of ci resulting in anterior specific expression (Dominguez et al., 1996; Guillen et al., 1995; Schwartz et al., 1995; Simmonds et al., 1995; Tabata et al., 1995; Zecca et al., 1995). EN has been shown to interact directly with the ci regulatory elements (Schwartz et al., 1995). In cg mutant wing imaginal discs, we found weak ectopic expression of ci-lacZ reporter constructs in posterior cells (Fig. 5B), thus CG may act in concert with EN to repress posterior ci. Hypomorphic mutants in either cg or en/inv can give rise to the reduction in vein L4 that is characteristic of ectopic ci expression (Fig. 4; Eberlein and Russell, 1983; Kornberg, 1981).

Many proteins with multiple C₂H₂ zinc finger motifs like those found in cg have been shown to be transcription factors, DNA-binding proteins or chromatin proteins (Coleman, 1992). The location of CG on salivary gland chromosomes is consistent with all of these activities. While our data have not yet established direct action of CG on the ci regulatory elements, the binding of CG to the ci region of polytene chromosomes suggests that it could be a direct regulator of ci transcription. We have been unable to demonstrate direct binding of CG produced in E. coli to DNA from the ci regulatory region (S. D. G. M., M. K. and W. J. B., unpublished). However, given that the transcriptional regulation of ci is likely to be complex, CG may not act at the level of direct DNA binding. The involvement of the Pc-group genes in the repression of ci (Maschat et al., 1998) suggests that intricate regulatory modes are necessary to maintain the correct levels and spatial patterns of ci transcription during imaginal disc development. Furthermore, the ci-regulatory regions have been shown to be subject to transvection effects, indicating that interchromosomal interactions also govern ci regulation (Locke and Tartof, 1994). Thus CG may act at any level, from generally influencing the chromosome pairing through to direct binding of ci enhancer elements. Finally, the positive and negative effects of cg mutants on ci transcription and the genetic interaction with en/inv suggest that CG may be required in conjunction with other transcription factors for the function of ci enhancers and that CG may not specify activation or repression itself.

The changes in the A/P pattern observed in cg mutant limbs are caused by the mis-regulation of HH-responsive genes regulated by the CI-155 and CI-75 transcription factors. In cg mutant wing imaginal discs, CI-155 is ectopically expressed in the posterior compartment and is associated with posterior compartment defects and posterior misexpression of genes such as ptc and kn. Ectopic expression of kn is sufficient to suppress vein fate (Mohler et al., 2000). Thus, the misexpression in the posterior compartment of kn and other genes regulated by high levels of CI-155 probably leads to the vein defects described here. The occurrence of both higher levels of posterior CI and KN expression, and higher frequency of posterior compartment defects in cg¹/cg¹ mutant wings supports this explanation. Stronger allelic combinations of cg have lower levels of ectopic CI and KN, and lower incidence of posterior compartment wing defects but they result in a greater reduction in CI in the anterior compartment and more anterior vein defects. The posterior and anterior vein defects, as well as occasional anterior wing margin bifurcations, resemble the effects of regulatory mutants of ci that cause the ectopic expression of ci in posterior cells and the reduction of ci expression in the anterior compartment (Dominguez et al., 1996; Schwartz et al., 1995; Slusarski et al., 1995).

In legs and antennae, overall CI levels are decreased in the anterior compartment, resulting in circumferential overgrowth of the anterior compartment and ectopic anterior expression of the morphogens wg and dpp. Similar effects on leg morphology have been reported previously when wg and dpp are ectopically activated in anterior cells (Basler and Struhl, 1994; Diaz-Benjumea et al., 1994; Jiang and Struhl, 1995; Jiang and Struhl, 1998; Lepage et al., 1995; Li et al., 1995; Pan and Rubin, 1995; Theodosiou et al., 1998). The rescue of cg mutant leg defects by additional expression of ci in the anterior compartment using the Gal4/UAS system indicates that the phenotypes result from a reduction of Ci-75 leading to the derepression of wg and dpp. Thus, we conclude that the CG protein is critical for the proper levels and spatial patterns of CI and that the A/P limb patterning defects in cg mutants are due largely, if not completely, to mis-regulation of ci.

The effects of cg mutants on ci expression are seen only in the anterior compartment of legs but in both anterior and posterior compartments in wings. What is the basis for this difference? While anterior CI is reduced in both limbs, ectopic posterior CI is only seen in wings. One possibility is that the alleles we are studying may have different effects on cg expression in anterior versus posterior compartments and/or legs versus wings. However, we saw little CG imaginal disc staining in cg²/cg³, suggesting little or no CG protein is produced, and so the phenotype may be near null (there are no deficiencies uncovering the cg locus, so we could not test this genetically). CG may not be required for repression of ci in the posterior compartment of leg discs, or alternatively, there may be a much lower threshold for CG function in legs.

The different effects on ci expression in cg mutant leg and wing imaginal discs suggest that while the broad framework is similar, there may be unique aspects to A/P patterning in dorsal versus ventral limbs. The predominance of wing phenotypes in ci^W and similar ci regulatory mutations also suggests a difference in the way ci is regulated in wings versus legs (Schwartz et al., 1995; Slusarski et al., 1995). Another difference was seen in the effects of reduced CI levels on the expression of dpp. A greater reduction of CI staining was seen in the anterior compartments of wings compared with leg imaginal discs; paradoxically, ectopic expression of dpp was seen in all cg mutant leg imaginal discs but none was seen in our sample of cg²/cg³ wing imaginal discs. Although there are limits to how accurately real levels of CI may be inferred from histochemical staining in different tissues, the simple conclusion is that dpp responds to different thresholds of CI in legs and wings.

We have been able to demonstrate that the role of cg in the A/P patterning of limb development involves regulation of ci expression. However, the broad tissue distribution of CG and the numerous binding sites on polytene chromosomes make it likely that cg has functions other than those in limb development described here. The adult viable cg mutants are sterile and also exhibit defects such as eye roughening and ectopic thoracic bristles (data not shown). Furthermore, several deficiencies from the Bloomington Deficiency Collection had dominant phenotypes (i.e., wing notching, ectopic veins) when in compound heterozygotes with cg² (E. Ciechanska and W. J. B., unpublished). While some of these phenotypes may be due to effects on ci regulation, others are not easily ascribed to changes in ci expression. Thus, CG may be required for the regulation of many genes but ci may be particularly sensitive to reduction in cg function.

We thank Stephen Cohen in whose lab this project was initiated and Ann-Marie Voie for assistance with transformants. We thank Kirsten Guss, Suzanne Eaton, Georg Halder, Todd Laverty, Teresa Orenic, Carol Schwartz, Anna Wild, Anita Taylor, Phil Ingham, Cairine Logan, Michel Crozatier, Alain Vincent, the Berkeley Drosophila Genome Project, and the Bloomington and Umea Drosophila stock centres, for fly stocks, cDNAs or anti-sera. We thank D. Daniel, T. Heslip, P. Mains and J. McGhee for many helpful suggestions. This work was supported by the Alberta Heritage Foundation for Medical Research and the Medical Research Council of Canada.