A cap ‘n’ collar protein isoform contains a selective Hox repressor function

The homeotic selector (Hox) genes are important components in a developmental pathway that diversifies morphology on the anterior-posterior body axis of many animals (Kenyon, 1994; Lawrence and Morata, 1994; Manak and Scott, 1994). Loss-of-function mutations in Hox genes, or their ectopic expression, can cause homeotic transformations in which body structures are duplicated in abnormal positions on the anterior-posterior axis (McGinnis and Krumlauf, 1992; Krumlauf, 1994). The protein products of Hox genes are DNA-binding transcriptional regulators that apparently accomplish their developmental functions by activating and repressing many target genes (Laughon, 1991; Botas, 1993; Graba et al., 1997). The similar in vitro DNA-binding properties of Hox proteins (Ekker et al., 1994) and the relative paucity of Hox genes compared to the diverse morphologies under their control, suggest that other genetic functions that act in parallel to Hox proteins play a crucial role in diversifying segmental morphology. Recently, a few Drosophila genes (extradenticle, teashirt, homothorax, cap ‘n’ collar) have been discovered that can mutate to give homeotic transformations that are largely independent of their effects on Hox patterns of transcription (Peifer and Weischaus, 1990; Roder and Kerridge, 1992; Mohler et al., 1995; Rieckhof et al., 1997). These genes appear to modulate the functional activity and specificity of the Hox proteins, thus we refer to them as the Hox modulator class. The understanding of Hox modulator functions is beginning to provide insight into the mechanisms of how Hox proteins achieve functional specificity, as well as broadening our knowledge of the developmental genetic circuitry that controls anterior-posterior body axis patterning. The modulator proteins so far characterized play a variety of biochemical roles, including enhancing the DNA-binding affinity of Hox proteins (Mann and Chan, 1996), regulating their transcriptional activities (Pinsonneault et al., 1997) and regulating the nuclear entry of Hox cofactors (Rieckhof et al., 1997).

An interesting example of the Hox modulator class is encoded in the cap ‘n’ collar (cnc) locus. cnc was identified and named (Mohler et al., 1991) based on its striking expression pattern in the anteriormost labral segment (cap) and the mandibular segment (collar) of embryos. Deletion mutants of cnc coding sequences indicate that cnc functions are required for the normal development of both labral and mandibular structures (Mohler et al., 1995). In place of the missing mandibular structures, some maxillary structures – mouth hooks and cirri – are ectopically produced (Harding et al., 1995; Mohler et al., 1995). The genetic function of the homeotic gene Deformed (Dfd) is required in the cnc mutant background to produce ectopic mouth hooks, and Mohler et al. (1995) have proposed that Dfd and cnc function in combination to specify mandibular identity. Previous research on the molecular genetics of cnc (Mohler et al., 1991) showed that the locus encodes a protein that is one of the founding members of the CNC/bZIP class of transcription factors, which includes NF-E2 (Andrews et al., 1993a), LCR-F1 (Farmer et al., 1997), Skn-1 (Bowerman et al., 1992) and Nrf-1 (Chan et al., 1993).

We recently isolated EMS-induced mutations in the cnc gene in a genetic screen for modifiers of Dfd (Harding et al., 1995) and are interested in the mechanism by which cnc alters the morphogenetic function of Dfd. To that end, we have further explored the molecular stucture of the cnc locus and found that three different protein isoforms are encoded by cnc. One of these isoforms, the CncB protein, exhibits spatially localized expression that is limited to the mandibular and labral segments. In mandibular cells, CncB antagonizes the ability of Dfd protein to transcriptionally activate response elements of downstream genes. Ectopic expression of CncB in embryos results in an ablation of normal maxillary structures, while having mild or no effects on the functions of homeotic proteins of the trunk. We propose that the CncB protein has properties that allow it to selectively repress many regulatory elements that are normally activated by Dfd.

Collections were made from cages containing flies heterozygous for mutant chromosomes over wild-type chromosomes to avoid balancer chromosome effects. Embryos were collected on apple juice/agar plates for 4 hours, aged at 25°C for 24 hours, then harvested and dechorionated in 100% bleach. They were devitellinized, cleared, mounted and analyzed as described in Harding et al. (1995).

Polyadenylated RNA was isolated from 0-2, 2-8, 8-12 and 12-24 hour embryos and 4 μg of each staged aliquot was loaded onto a pre-run 0.8% agarose formaldehyde gel. The RNA was blotted onto nitrocellulose filters and probed as in Sambrook et al. (1989). Radioactively labeled probes were made by the method of nick translation from the following DNA fragments. (i) cncA probe, a 900 bp XhoI-EcoRI fragment from a genomic clone that includes exon A1. (ii) cncB probe, a 1.1 kb BstXI-EcoRV fragment from the cDNA clone pNBcnc27, which was isolated from a 8-12 hour embryonic cDNA library (Brown and Kafatos, 1988). The cncB probe includes part of exon B1, all of exons B2 and B3, and most of exon B4. (iii) cncB-specific probe, a 2.2 kb EcoRI genomic fragment in clone pBstB1spec, which includes exon B1 sequences and no other B exon sequences. (iv) cncC probe, a 2.3 kb BamHI-EcoRI fragment from the cDNA clone pNBcnc23, which was isolated from a 8-12 hour embryonic cDNA library (Brown and Kafatos, 1988). This probe contains part of exon C1, all of exons C2 and C3 and most of exon C4. (v) The cnc common probe is a 310 bp BamHI fragment entirely contained within exon A2, which was isolated from the cDNA clone cnc1A10 (Mohler et al., 1991). See Fig. 2 for the location of the probes on the map of the cnc locus.

Several D. melanogaster libraries were screened for both cDNA and genomic clones. The 4-8 hour, 8-12 hour and 12-24 hour embryonic cDNA libraries of Brown and Kafatos (1988) were screened using the 310 bp BamHI fragment from exon A2. The 8-12 hour library was also screened with an 83 bp PCR product made solely of exon A1 and the 4-8 hour and 12-24 hour libraries were additionally screened with a PCR fragment covering the 5′ end of exon B2 to the middle of exon B4. An iso-1 genomic library consisting of Sau3A partial digests ligated into λEMBL3 (Tamkun et al., 1992) was screened with a 2.3 kb BamHI-EcoRI fragment from pNBcnc23, a cncC cDNA clone. A randomly primed λgt10 cDNA library (Clontech, Cat # IL1010a) was screened with a 4 kb EcoRV-HindIII fragment from pNBcnc27, which includes sequences from within exon B2 to beyond the translation stop. All potential positives obtained from the primary screens were picked, replated at lower density and rescreened according to standard protocols to ensure plaque/colony purity. PCR screening was employed as a rapid method to detect and characterize the structure of cDNA clones containing different exons and/or to assay the extent of 5′ exon sequence.

For two mutant alleles, cnc^2E16 and cnc^C7, and the parental chromosome Ki Dfd^rV8p^p, DNA sequence was obtained for chromosomal regions corresponding to known exons, as well as all exon/intron boundaries including 40-50 bp beyond each boundary into the introns. All mutant and parental chromosomes were heterozygous and balanced over TM3 Sb 35UZ (Irvine et al., 1991). DNA was extracted from 1 g of flies according to standard protocols and isolated on a CsCl gradient. Regions of the genome were PCR amplified using nested primers chosen from cDNA and genomic sequences. Primers were made approximately 100 bp from intron/exon boundaries so that all boundaries could be sequenced from both directions. The PCR products were then either sequenced using ³⁵S and the Sequenase PCR Product Sequencing Kit from Amersham (Catalogue no. US70170) or they were agarose gel purified and sequenced according to kit protocols without the use of Exonuclease I and Shrimp Alkaline Phosphatase. At least one strand was sequenced for 100% of the regions, both strands were sequenced for the majority of the open reading frames and 7-deaza dGTP reactions were used to uncover any ambiguities due to compression. No nucleotide substitutions were present in any cnc open reading frames or splice junctions of the cnc^2E16 and cnc^C7 mutant chromosomes when compared to the parental chromosome sequence.

Polyclonal antiserum was raised against the common Cnc protein domain produced as a glutathione S-transferase (GST)-fusion protein.A cDNA clone (pNBcnc10) containing a full-length CncA open reading frame was digested with BstEII and SacI and blunt ended. The resulting 1721 bp fragment was ligated into the SmaI site of pGEX-4T-3 (Pharmacia Biotech) to produce an in-frame fusion with GST. The GST-Cnc common region polypeptide was purified on glutathione-Sepharose columns according to the manufacturer’s protocol, and the antiserum raised in rabbits at the Pocono Rabbit Farm and Laboratory, Inc.

For affinity purification of antibodies, pNBcnc10 was digested with HincII and SacI and the resulting 1827 bp fragment that included the entire CncA open reading frame was ligated into the pQE-32 vector (Qiagen) which had been digested with BamHI, blunt-ended and then cut with SacI. The fusion protein containing a 6xHis-tag at the N terminus was purified on the Qiagen Ni-NTA Agarose column under denaturing conditions and refolded. The protein was coupled to Actigel ALD beads and antibodies directed at the common Cnc protein domain were purified using the Quickpure system (Sterogene Bioseparations, Inc.). Antibody staining for Dfd and Cnc proteins were done as described in Zeng et al. (1994). For fluorescence microscopy, FITC-conjugated anti-rabbit secondary antibodies were used to detect Cnc protein and Cy-5-conjugated anti-guinea pig antibodies were used to detect Dfd protein (Jackson ImmunoResearch Laboratories, Inc.). Optical sections of fixed and stained embryos were taken every 0.2 μm using a DeltaVision microscope system (Applied Precision, Inc.) with a computer-controlled stage. An Olympus 60×/1.40 objective was used. Following image acquisition, out-of-focus blur was removed using constrained iterative deconvolution (Agard et al., 1989).

For RNA in situ staining, sense and antisense digoxigenin-labeled RNA probes were produced from subclones of cDNA or genomic fragments based on Tautz and Pfeifle (1989). The lacZ reporter and other gene expression patterns were detected by staining of whole-mount embryos as described in Bergson and McGinnis (1990).

A cncA cDNA clone, 1A10 (Mohler et al., 1991) was digested with EcoRI and cloned into the EcoRI site of pHSBJ-CaSpeR (Jones and McGinnis, 1993). The resulting pHSBJ-CaSpeR-cncA contained 100 bp of 5′ UTR and 800 bp of 3′ UTR from exon A3.

pNBcnc27 cDNA was digested with EcoRV and ligated with EcoRV-digested pHSBJ to produce pHST17. The heat-shock cassette was cut out with NotI and inserted into the NotI-digested pCaSpeR4 vector to produce pHSBJ-CaSpeR4-cncB. This construct contained 390 bp of 5′ UTR and 1230 bp of 3′ UTR from exon A3. hs-cncC pNBcnc23 cDNA was digested with HindIII and NcoI, as was the pHST17 plasmid. Corresponding DNA fragments were ligated to produce pHSBJ-cncC. The resulting cassette was cut out with NotI and ligated with NotI-digested pCaSpeR4 to create pHSBJ-CaSpeR4-cncC. The first ATG codon in this cncC expression construct is 37 bp from the HindIII site and it encodes the methionine residue found at position 21 in the conceptual translation of the cncC open reading frame shown in Fig. 3. This construct also contains 1230 bp of the normal cnc 3′ UTR.

We recovered three EMS-induced mutant alleles of cnc (cnc^2E16, cnc^C7 and cnc^C14) in a screen for mutations that interact with the Hox gene Dfd (Harding et al., 1995).

Embryos homozygous for these EMS-induced alleles have ectopic duplications of maxillary mouth hooks and cirri, but retain normal labral structures and some normal mandibular structures, e.g. the lateralgräten and median tooth (Fig. 1). This contrasts with the phenotype of deletion mutants of cnc, which lack all mandibular and labral derivatives (Mohler et al., 1995). The difference between the phenotypes of the EMS-induced alleles when compared to the deletion alleles prompted us to consider the possibility that multiple functions are encoded in the cnc locus.

Previous studies detected one transcript isoform at cnc (Mohler et al., 1991), but our molecular analyses of the locus indicate that three transcript and protein isoforms are produced from the cnc gene. As shown in Fig. 2C, a probe homologous to the region that encodes the b-ZIP region of cnc detects three different sizes of polyadenylated RNAs on embryonic northern blots. We will refer to these as the cncA, cncB and cncC transcripts. No other embryonic transcripts were detected is present in 0-2 hour embryos, is barely detected in 2-12 hour embryos and is detected at relatively higher levels in 12-24 hour embryos (Fig. 2C).

To identify cDNAs corresponding to the cncA, cncB and cncC transcripts, 212 cDNA clones from libraries covering all stages of Drosophila embryonic development were isolated and characterized. The first class of cDNAs corresponded to the cncA transcript. This is the same class characterized by Mohler et al. (1991), and is distinguished by the incorporation of exon A1. Exons A2 and A3, which encode the CNC and b-ZIP domains, are present in cncA and the other two isoforms of cnc. A probe containing exon A1 sequences specifically hybridizes the 3.3 kb cncA transcript on northern blots (Fig. 2C). The cncA open reading frame begins with an ATG codon near the 5′ end of exon A2 and is predicted to encode a 533 amino acid protein (Mohler et al., 1991).

A second class of cDNAs from the locus corresponded to cncB transcripts. Such cDNAs lacked sequences from exon A1, but did contain five additional exons (B1-B5) spliced onto the 5′ end of exon A2. A probe containing the B1-B4 exons detects the 5.4 kb cncB transcript and the 6.6 kb cncC transcript on northern blots (Fig. 2C). The total extent of the cncB transcription unit is approximately 17 kb (Fig. 2B). Since exon A2 sequences contain no stop codons upstream of the initiating ATG for the CncA codons, the open reading frame in cncB transcripts includes the entirety of the CncA protein, as well as an additional 272 codons from exons B3, B4, B5 and A2 (Fig. with genomic probes that spanned the region from −35 to +5 kb shown in Fig. 2A. The 3.3 kb cncA transcript is present in 0-2 hour embryos, presumably from maternal stores and is also abundantly expressed in 12-24 hour embryos. The 5.4 kb cncB transcript is absent from 0-2 hour embryos, but present at all other embryonic stages. The 6.6 kb cncC transcript 2B). The predicted 805 amino acid CncB protein thus is distinguished from CncA by a 272 amino acid region that includes His-Pro repeats, Ala-repeats, a Pro-repeat and Val-Gly repeats, but exhibits no extended sequence similarity to other proteins in database searches besides the CNC/b-ZIP domain that it shares with CncA (Fig. 3).

The third class of cDNAs from the locus corresponded to cncC transcripts. These cDNAs have identical sequence to the cncB cDNAs, except that exon B1 is absent, and five additional exons (C1-C5) are spliced onto the 5′ end of exon B2. A probe containing the C1-C4 exons detects the 6.6 kb cncC transcript on northern blots (Fig. 2C). Since exon B2 and the 5′ end of exon B3 contain no stop codons upstream of the initiating ATG for the CncB codons, the ATG-initiated open reading frame in cncC transcripts includes the entirety of the CncB protein, as well as an additional 491 codons that derive from the C3, C4, C5, B2 and B3 exons. The extent of the entire cncC transcription unit is approximately 39 kb. The 491 amino acid CncC-specific domain at the N terminus of the predicted 1296 residue CncC protein includes regions that are rich in Ser and Thr residues, other regions with abundant concentrations of Glu and Asp residues, but exhibits no extended sequence similarity to other proteins in database searches (Fig. 3). Interestingly, the fuzzy onions gene, which encodes a testis protein required for mitochondrial fusion in Drosophila spermatids (Hales and Fuller, 1997), is encoded in the sequence interval between the C5 and B1 exons (Fig. 2).

We next were interested in defining which of the cnc isoforms were involved in modulating Dfd function. As a first step, the expression patterns of the three transcript isoforms were analyzed using cncA, cncB, or cncC exon-specific probes both on wild-type and our EMS-induced cnc mutants. In wild-type embryos, a cncB probe detects cytoplasmic transcripts limited to the mandibular and labral segments from cellular blastoderm to the end of embryogenesis (Fig. 4A,B). The cncB transcripts are expressed throughout both anterior and posterior regions of the mandibular lobes. In contrast, a cncA-specific probe detects a ubiquitous distribution of presumably maternal RNA at syncytial and early cellular blastoderm stages (Fig. 4F,G). After cellular blastoderm, cncA transcripts are not detectable until stage 14, when the level of ubiquitous cytoplasmic transcript increases in abundance and remains high for the remainder of embryogenesis. cncC-specific probes also detect a ubiquitous distribution of RNA in syncytial stage embryos and a low level ubiquitous expression pattern in embryos after stage 14 (data not shown).

Based on the above results, the labral and mandibular stripes of transcription that were detected by Mohler et al. (1991) using a probe including the cnc common exons (A2 and A3) correspond primarily to cncB transcripts. Since cncB is the transcript isoform that is expressed throughout the entire mandibular segment during mid-embyronic stages, our working hypothesis is that cncB encodes the principal function that modulates Dfd function in the mandibular segment. To further test this hypothesis, we assayed whether cncB transcript or protein abundance was altered in embryos homozygous for the cnc^2E16 and cnc^C7 mutations. We found that the pattern of zygotic RNA expression detected with a cncB probe is unaltered in the EMS-induced cnc mutant embryos (Fig. 4C). The signal due to cncA and cncC transcripts was also unchanged in these mutants (data not shown).

However, the use of polyclonal antiserum raised against the common domain of the cnc isoforms (anti-Cnc) indicates that CncB protein expression is strikingly reduced in both cnc^2E16 and cnc^C7 mutant embryos. In wild-type embryos, the anti-Cnc antiserum exhibits a low-level ubiquitous staining in syncytial embryos, presumably due to maternally deposited CncA andC CncC isoforms. From cellular blastoderm (stage 5) until stage 14, the staining detected by the anti-Cnc antiserum is localized in the nuclei of mandibular and labral cells (Fig. 4D). Although the anti-Cnc antiserum used in these experiments cross-reacts with all three Cnc proteins, only cncB RNA expression is localized in mandibular and hypopharyngeal regions from stages 6 through 14. From stage 14 to the end of embryogenesis, the antiserum detects a low level global staining, upon which is superimposed much stronger levels of staining in mandibular and labral cells. As can be seen in Fig. 4E, cnc^2E16 mutants (and cnc^C7 mutants, not shown) accumulate much lower levels of Cnc antigen in both mandibular and labral cells of stage 11 embryos. These results provide further evidence that the cnc^2E16 and cnc^C7 mutations result in a loss of cncB function, and is consistent with the idea that CncB protein is required to prevent the maxillary-promoting function of Dfd from being active in mandibular cells.

We determined the sequence of all of the coding exons and exon/intron boundaries for all isoforms on the cnc^2E16 and cnc^C7 mutant chromosomes (see Materials and Methods) in an attempt to find the molecular lesion responsible for the decreased amount of CncB protein in the mutant embryos. However, no nucleotide substitutions were detected when the coding and splice site sequences were compared with parental chromosome sequence. Though we do not yet know the location of the mutations that alter CncB protein expression, they could plausibly reside in translational regulatory sequences for cncB.

In another test of the functions of the Cnc protein isoforms, we placed each of the cncA, cncB and cncC open reading frames under the control of the heat-shock promoter in P-element vectors and generated transgenic fly strains carrying these constructs. Using the Cnc common-region antiserum to stain heat-shocked embryos, it appears that all three isoforms are produced at similar levels, localized in nuclei and possess similar stabilities after ectopic expression (Fig. 5C,E,G). However, their morphogenetic and regulatory effects are very dissimilar. Heat-shock-induced ectopic expression of CncA during embryogenesis has no effect on embryonic morphology. Nearly all of the hs-cncA embryos hatch and proceed through larval development, and many eclose as viable adults.

In contrast, ectopic expression of CncB at mid-stages (4-10 hours) of embryonic development is lethal. When ectopic expression is induced at 6 to 8 hours after egg lay, a defective embryonic head phenotype, which resembles the mutant phenotype of strong Dfd hypomorphs is produced (Fig. 5I, compare with Fig. 1B). These hs-cncB embryos develop with rudimentary mouth hooks, H-piece and cirri. In addition, the anterior portion of the lateralgräten are truncated. All of these structures are components of the head skeleton that are absent or abnormal in Dfd mutant embryos (Merrill et al., 1987; McGinnis et al., 1990). The head defects seen in the hs-cncB embryos also include an absent or abnormal dorsal bridge, a structure that is usually unaffected in Dfd mutant embryos. Many other head structures that develop in a Dfd-independent manner, such as the antennal sense organ, vertical plates and T-ribs (Jurgens et al., 1986) develop normally in the hs-cncB embryos. The hs-cncB head defects are produced at high penetrance (>95%) by heat shocks in mid-embryogenesis (4-10 hours). In 10-70% of these embryos, depending on the stage of heat shock, abdominal denticles near the ventral midline are replaced with naked cuticle (Fig. 5L). We have not observed the ectopic formation of mandibular or hypopharyngeal structures such as lateralgräten, dorsolateral papillae or T-ribs (Jurgens et al., 1986) in other head or trunk segments in hs-cncB embryos.

Ectopic induction of hs-cncC at 6-8 hours of development also results in highly penetrant defects in head development that include the loss of maxillary mouth hooks and cirri (Fig. 5J) as well as head involution defects that are more profound than those induced by hs-cncB. In addition to the morphological defects described for CncB, ectopic CncC induces the formation of an abnormal head sclerite that develops as an extension of the normal lateralgräten. The position and appearance of this extra fragment of head skeleton suggests that it might correspond to ectopic production of lateralgräten or longitudinal arms of the H-piece.

Since CncB encodes a function that is required and sufficient to antagonize the maxillary-promoting effects of the Hox gene Dfd, we next were interested in whether CncB protein acts upstream to repress Dfd transcription, or in parallel to inhibit Dfd protein function. It is possible for CncB to do both, since Dfd protein function is required to establish an autoactivation circuit that provides persistent Dfd transcription in maxillary and mandibular cells (Zeng et al., 1994). In wild-type embryos at stage 9, both Dfd and CncB proteins are expressed throughout the entire mandibular segment (Fig. 6A). By stage 11, Dfd protein is present at lower levels in the anterior when compared to posterior mandibular nuclei, while CncB protein persists at relatively high levels throughout the segment (Fig. 6B). Finally, at stage 13, Dfd protein expression is no longer detected in anterior mandibular nuclei, although it is still abundant in posterior nuclei (Fig. 6C). cnc is required for this progressive repression of Dfd expression in the anterior mandibular segment, since cnc null mutants as well as the EMS-induced mutants show inappropriate persistence of Dfd transcripts and protein after stage 11 in anterior mandibular cells (Fig. 6E). All of this data suggests that CncB is not capable of repressing Dfd expression before stage 11. But after that stage CncB represses the maintenance phase of Dfd transcription in mandibular cells, perhaps by repressing the autoactivation circuit that is normally established during stages 9 and 10 (Zeng et al., 1994).

We also found that CncB is sufficient to repress Dfd transcription outside the mandibular segment. When CncB is ectopically expressed in embryos, Dfd transcript levels in the maxillary segment are reduced, especially in the anterior region of the segment (Fig. 6G). Note that these transcript expression assays were done at a time after heat shock when Dfd protein is still present at wild-type levels (Fig. 7A,B). Only the CncB isoform is capable of repressing Dfd transcription. Neither the ectopic expression of CncA nor CncC have an effect on the abundance or pattern of Dfd transcripts in the maxillary epidermis (Fig. 6F,H). Since the phenotypic effect of hs-cncC in epidermal cells strongly resembles that of hs-cncB, this indicates that the effect of Cnc gene products on maxillary epidermal development may not require repression of Dfd transcription per se.

The CncB repressive effect on Dfd expression might be mediated, at least in part, through autoactivation elements. To test this, we assayed the activity of three subregions from the Dfd-epidermal autoactivation element (Dfd-EAE) in hs-cncB embryos, using hs-cncA and wild-type embryos as controls. At 30 minutes after heat-shock induction of CncB expression, when no change in Dfd protein abundance is detectable (Fig. 7A,B), there is a decrease in the activity of various modules of the Dfd-EAE. One of these, module F, consists of a 471 bp fragment at the 3′ end of the Dfd EAE (Zeng et al., 1994). This element activates Dfd-dependent reporter expression in posterior maxillary cells of wild-type embryos and is equally active in hs-cncA controls (Fig. 7C). Ectopic expression of CncB nearly abolishes the activity of this element (Fig. 7D). Another Dfd-EAE fragment, 570 bp module C, activates Dfd-dependent reporter transcription in most maxillary epidermal cells (Zeng et al., 1994; Fig. 7E). When compared with hs-cncA embryos, module C exhibits lower activity in hs-cncB embryos (Fig. 7F). The smallest known module of the Dfd-EAE with a significant amount of autonomous activity is the 120 bp module E, which is directly targeted by both Dfd and Exd proteins (Zeng et al., 1994; Pinsonneault et al., 1997). In hs-cncB embryos, expression levels from module E are lower than those observed in wild-type or hs-cncA embryos (Fig. 7G,H). Conversely, in cnc^2E16, cnc^C7 or cnc^VL110 embryos (shown), module E is ectopically expressed in mandibular cells (Fig. 7K,L). From these experiments, we conclude that cnc function is required to repress a variety of Dfd response elements in mandibular cells and that ectopic CncB is sufficient to reduce the activity of all of those elements in maxillary cells. The hs-cncB-induced repression of these elements occurs at a time after heat shock when Dfd protein levels in the maxillary segment are unchanged, evidence that the CncB effect on these elements is not indirectly caused by CncB repression McGinnis, 1988) and hs-cncA develop ectopic maxillary cirri and mouth hooks in 69% (n=94) of the first thoracic segments. However, when a hs-cncB is substituted for hs-cncA in the same genetic background, only 25% of the first thoracic segments bear thoracic cirri. Thus, even when Dfd expression of Dfd protein levels produced from the endogenous Dfd locus.

Although the cuticular phenotype conferred by ectopic expression of hs-cncB suggested it does not generally antagonize the function of Hox proteins that specify trunk regional identities, we wished to test whether cis-regulatory elements that are activated by other Hox proteins exhibited any response to ectopic CncB. To address this question, we tested the activity of the Ubx-activated dpp674 element (Capovilla et al., 1994; Sun et al., 1995), the Abd-B-activated ems-1.2 kb filzkörper element (Jones and McGinnis, 1993) and an element that is activated by the Hox protein Labial (Popperl et al., 1995; Chan et al., 1996). All of these elements exhibited patterns and amounts of reporter expression in hs-cncB embryos that were indistinguishable from controls (Fig. 7I,J and data not shown).

Another experiment to address whether CncB represses the maxillary-promoting function of Dfd by antagonizing its function on downstream target elements involved testing the effects of CncB on Dfd-dependent structures when Dfd expression is driven by an exogenous promoter. Heat-shocked embryos that are heterozygous for both hs-Dfd (Kuziora and is driven in ectopic positions by a heat-shock promoter, the maxillary-promoting function of Dfd protein is reduced in the presence of CncB. This could either be due to CncB-mediated repression of the Dfd autoactivation circuit in ectopic positions or to CncB repression of downstream target elements of Dfd protein, or to both of these effects.

One of the downstream genes that is activated by Dfd in maxillary cells is Distal-less (Dll). Dll is required for the formation of the larval appendage primordia and the distal regions of adult appendages (Cohen et al., 1989). In the maxillary segment, Dll is expressed in two patches of cells: a dorsal patch that gives rise to the maxillary sense organ and a ventral patch that consists of the primordia for the maxillary cirri (Fig. 8A). The dorsal maxillary domain of Dll expression is largely independent of Dfd function, while the ventral maxillary patch of Dll is activated by Dfd through a 3′ enhancer (the ETD6 element) (O’Hara et al., 1993). In cnc^2E16 mutant embryos, Dll is ectopically expressed in ventral mandibular cells, suggesting that cncB represses Dll transcription in mandibular cells (Fig. 8B). In hs-cncB embryos 30 minutes after heat shock, when Dfd protein abundance is normal (Fig. 7A,B), Dll expression is repressed in the ventral maxillary segment (Fig. 8D) but other domains of Dll expression in the head and thorax are relatively unaffected, indicating that CncB selectively represses the Dfd-dependent portion of the Dll expression pattern. In hs-cncC and hs-cncA embryos, the ventral maxillary expression of Dll is not selectively repressed (Fig. 8C). Reporter gene expression from the Dll ETD6 enhancer follows the expression of Dll as the enhancer is ectopically activated in the ventral mandibular region in cnc mutants and repressed in hs-cncB embryos (data not shown).

The cnc gene has been proposed to encode a spatial determinant required for proper segmental identity in the posterior head (Mohler et al., 1995). Here we find that a crucial part of this segmental identity function is provided by the localized expression of the CncB protein isoform. In loss-of-function mutants in which CncB function is reduced or absent, mandibular epidermal cells assume the identity of maxillary epidermis. This is in part because Dfd function is no longer antagonized in mandibular cells, which our current evidence suggests is mechanistically accomplished by CncB exerting a repressive effect on many Dfd response elements. These elements include ‘downstream’ autoactivation elements at the Dfd locus and Dfd response elements at downstream genes such as Dll and 1.28 (Mohler et al., 1995). When CncB function is reduced or lost, these downstream target genes and Dfd response elements are inappropriately activated in mandibular cells. CncB is one of the growing list of genetic functions, other examples being Exd/Pbx class proteins, the products of the teashirt and homothorax genes and other Hox genes, that modulate the functional activity and specificity of the Hox system, and thereby diversify A/P body axis morphology (Peifer and Weischaus, 1990; Gonzalez-Reyes and Morata, 1990; Roder and Kerridge, 1992; Jurgens and Hartenstein, 1993; Mohler et al., 1995; Rieckhof et al., 1997; Kurant et al., 1998; Pai et al., 1998). Unlike Exd/Pbx, there is as yet no evidence that the modulatory effect is mediated by direct binding of CncB to Dfd protein.

CncB is not only required to repress Dfd target genes and Dfd response elements in most mandibular cells, but also is sufficient to partially repress such Dfd targets in the maxillary segment. When CncB is ectopically expressed during mid-embryogenesis, many of the normal Dfd-dependent cuticular structures produced by maxillary cells are reduced or absent. This is associated with the ability of CncB to reduce the levels of Dfd transcription in maxillary cells. This is due, at least in part, to CncB action on Dfd epidermal autoactivation elements, since the activity of some of these elements is rapidly repressed by ectopic CncB. No ectopic mandibular structures (e.g. lateralgräten, dorsolateral papillae) are detected in the hs-cncB cuticular preparations, therefore it appears that the heat-shock-induced expression of CncB, coupled with Dfd function, is not sufficient to specify mandibular segmental identity in place of maxillary identity.

Among the many known and suspected targets of Hox proteins, the CncB repression function appears to be highly selective for Dfd targets. In normal embryos, this is partly due to the highly restricted expression pattern of CncB, which overlaps the expression pattern of Dfd, but not the expression of other Hox genes except for proboscipedia (Pultz et al., 1988). However, even when hs-cncB is ectopically expressed, the epidermal phenotype indicates that the function of Hox genes that act in the trunk (e.g. Scr, Antp, Ubx, abd-A and Abd-B) are largely unaffected. This selective activity might be due to the presence of CncB-binding sites in many Dfd response elements and the absence of such sites from downstream elements that are activated by other Hox proteins. All of the Drosophila Cnc protein isoforms are closely related in their CNC and b-ZIP domains to the well-studied p45 subunit of the mammalian NF-E2 transcription factor. Based on this, it is expected that one mode of CncB action will involve heterodimerization with small Maf-class proteins on sequences with the consensus GCTGANTCAT (Andrews et al., 1993a,b). The underlined nucleotides designate the palindrome found in many sites bound by homodimers or heterodimers of b-ZIP proteins. None of the elements (modules C, E or F of the Dfd epidermal autoregulatory enhancer) that show ectopic activity in cnc mutants and are repressed by ectopic expression of CncB, have sequences that match the TGANTCA core of the b-ZIP-binding site. Module E has been previously subjected to systematic mutageneses that involved clustered substitutions throughout its length (Zeng et al., 1994; Gross and McGinnis, 1995; Pinsonneault et al., 1997). None of the mutant versions of module E exhibited ectopic activity in the mandibular segment, although discrete subregions were required for the activity of this Dfd response element in the maxillary segment. The sum of this evidence suggests that the sequences in module E that transduce the cnc repressive effect overlap with the sequences required for module E activation. At present, the mechanism by which CncB function is directed to Dfd target elements is unknown. Since the ectopic expression of CncA protein has no detectable effect on morphology or viability, even though it possesses the same DNA-binding domain as CncB, it seems unlikely that a Cnc DNA-binding function alone is sufficient to repress the activity of Dfd response elements.

One region of Dfd expression that appears to be immune to the repressive effects of CncB is in posterior mandibular epidermis. Throughout most of embryogenesis, these cells maintain abundant levels of CncB protein and levels of Dfd protein that are only somewhat lower than the levels detected in maxillary cells (Fig. 6). One possible reason for this is that some of the autoactivation enhancers that supply Dfd expression in this region are not completely repressed by CncB. It is only a subset of such elements that are immune to CncB, since other autoregulatory enhancers that contribute to Dfd expression in this region are strongly activated when CncB function is absent (Fig. 7K,L). We believe that the most-likely possibility is that some of the persistent Dfd transcription in the posterior regions of the maxillary and mandibular segments is supplied by elements that require Dfd protein only to achieve the normal levels of expression and have independent sources of regulatory input that determine their spatial limits of expression. There is evidence for the existence of such elements, which act in the posterior region of the maxillary segment (Zeng et al, 1994), and this also applies to the posterior mandibular segment at early stages of embryogenesis (G. Gellon and W. McG., unpublished results).

It is possible that protein-protein interactions play a role in the selectivity of CncB action. The difference between the CncA protein and the CncB protein resides in the 272 amino acids found at the N terminus of CncB. This 272 amino acid domain, required for the repressive effect on Dfd response elements, has no significant similarity to non-repetitive amino acid sequence of proteins that are currently listed in public databases. It does, however, share a number of repetitive amino acid motifs that are found in a variety of other transcription factors. These include His-Pro repeats, Ala repeats, Val-Gly repeats and Pro repeats. Alanine-rich regions have been associated in a few instances with transcriptional repression functions (Licht et al., 1990; Han and Manley, 1993,Hanna-Rose and Hansen, 1996). Thus, it is possible that the 272 amino acid domain has a relatively generic repression function that accounts for its ability to interfere with Dfd protein on target elements. This 272 amino acid domain might also contain sequences that allow it to specifically repress Dfd response elements, while having little effect on other Hox response elements.

The predicted CncC protein also possesses this 272 amino acid domain, but in addition has an additional 491 amino acids at its N terminus that are unique to this isoform. At present, there are no known mutant alleles that remove only the cncC function and leave the function of the other isoforms intact, but the hs-cncC-induced defects allow us to speculate about the role of cncC in normal development. Heat-shock-induced ectopic expression of CncC protein has a profound influence on head development, removing all Dfd-dependent maxillary epidermal structures and affecting the normal morphogenesis of many other cuticular structures from a variety of head segments. Interestingly, it ablates the maxillary-promoting function of Dfd without any detectable affect on Dfd transcription levels. The cncC transcript isoform is apparently maternally deposited, and expressed in all or virtually all cells during embryogenesis at levels that vary at different stages. Using staining intensity as a guide with probes of similar sizes, the cncC transcript levels appear to be much lower than those of cncB in the mandibular and hypopharyngeal regions. One model consistent with all these results is that the CncC function sets a threshold level of repressor that must be overcome in order for Dfd and other head-patterning functions to activate transcription from some or many of their downstream target genes. Conceivably, the threshold level of this repressor might even change the segment identity function of Dfd protein by regulating its ability to activate different numbers of target genes. Consistent with this model, when the level is dramatically raised by heat-shock-induced expression of CncC, the ability of Dfd and other head-patterning genes to promote the development of head structures is lost. Further experiments with cncC-specific null mutations, in combination with markers for target genes of Dfd and other head-patterning factors, will be required to explore this and other potential explanations for the cncC gain-of-function and loss-of-function phenotypes.

CncB and CncC may act in the mandibular segment in a manner that resembles how posterior Hox proteins act to influence the function of anterior Hox genes during the process of phenotypic suppression (Gonzalez-Reyes and Morata, 1990; Macias and Morata, 1996). For example, the Hox protein Ultrabithorax (Ubx) can suppress the thoracic-promoting function of Antennapedia (Antp) in a manner that is independent of Ubx regulatory effects on Antp transcription. One mechanism that has been proposed to explain phenotypic suppression is competition for common binding sites by the different members of the Hox family, although other mechanisms are equally plausible (Duboule and Morata, 1994). Binding-site competition seems highly unlikely to be sufficient for the CncB suppressive effect on Dfd response elements, since CncA also possesses the same DNA binding and dimerization motif as CncB, and has no detectable influence on Dfd expression or function. Though binding site competition in itself seems to be an unlikely mechanism, it is intriguing that the half site that is recognized by CNC class b-Zip proteins, ^G/_ATCAT, resembles the preferred half-site (ATCA) for proteins of the PBC class (e.g Drosophila Exd, mammalian Pbx). The Exd/Pbx proteins bind cooperatively to DNA with many Hox proteins (Mann and Chan, 1996) and the exd function in Drosophila appears to be required for many of the Hox proteins to activate downstream target genes, but not for their repression effects on targets (Pinsonneault et al., 1997). Perhaps one way in which CncB acts is by antagonizing the function of this known Hox coactivator on certain Dfd response elements.

In many interesting ways, the interactions of CncB with Dfd resemble those of the teashirt (tsh) gene with Hox genes of the trunk, particularly Sex combs reduced (Scr). Scr is normally expressed both in the labial and 1st thoracic segments, while tsh expression overlaps only the 1st thoracic (T1) portion of the Scr domain (Fasano et al., 1991). Both tsh and Scr are required for T1 identity, and in order to achieve the normal morphology of this segment, tsh represses Scr transcription and function in T1 (Fasano et al., 1991; Roder and Kerridge, 1992). Teashirt is also capable of repressing some of the morphogenetic functions of Scr in other segments when Scr expression is driven by heterologous promoters (Andrew et al., 1994). The mechanism that integrates the functions of tsh and Scr is not yet known, but presumably occurs on common downstream target elements, since the tsh gene encodes a zinc-finger protein with a sequence-specific DNA-binding function (Alexandre et al., 1996).

CncB is one of three isoforms produced from the cnc locus. The other two, CncA and CncC, are expressed maternally, as their transcripts are present in early syncytial embryos. Thus either or both of these isoforms may play a role in oocyte development or in early stages of embryogenesis. Analysis of a P-element insertional allele of cnc indicates that a function at the locus is required for germ cell viability or early oogenesis (Perrimon et al., 1996). Both the CncA and cncC isoforms are also expressed at later stages in most or all embryonic cells. Studies of their roles in development, and how these roles are integrated with the role of CncB, await mutations that selectively eliminate their functions.

We thank Jym Mohler, Xuelin Li and Juan Botas for sharing clones, flies and regulatory elements, Brian Florence and Elizabeth Wiellette for critical readings of the manuscript and Raffi Aroian for instruction on the confocal microscope. The Bloomington Drosophila stock center and Flybase supplied invaluable support. This research was supported by an NIH grant (W. M.), a Fulbright fellowship (E. R.) and by an HHMI pre-doctoral fellowship (A. V.).