The Drosophila rotund gene is required in the wings, antenna, haltere, proboscis and legs. A member of the Rac family of GTPases, denoted the rotund racGAP gene, was previously identified in the rotund region. However, previous studies indicated that rotund racGAP was not responsible for the rotund phenotypes and that the rotund gene had yet to be identified. We have isolated the rotund gene and show that it is a member of the Krüppel family of zinc finger genes. The adjacent roughened eye locus specifically affects the eye and is genetically separable from rotund. However, roughened eye and rotund are tightly linked, and we have therefore also isolated the roughened eye transcript. Intriguingly, we show that roughened eye is part of the rotund gene but is represented by a different transcript. The rotund and roughened eye transcripts result from the utilization of two different promoters that direct expression in non-overlapping domains in the larval imaginal discs. The predicted Rotund and Roughened Eye proteins share the same C-terminal region, including the zinc finger domain, but differ in their N-terminal regions. Each cDNA can rescue only the corresponding mutation and show negative effects when expressed in each others domain of expression. These results indicate that in addition to the differential expression of rotund and roughened eye, their proteins have distinct activities. rotund and roughened eye act downstream of early patterning genes such as dachshund and appear to be involved in Notch signaling by regulating Delta, scabrous and Serrate.
INTRODUCTION
The Drosophila rotund (rn) locus is recessive viable causing male and female sterility as well as defects in adult body structures (Cavener et al., 1986). These include defects in the antennae, wing, haltere and proboscis as well as fusion of all five leg tarsi into one fused tarsal-like segment. Analysis of third instar larvae imaginal discs revealed localized cell death in the regions giving rise to the affected adult structures (Kerridge and Thomas-Cavallin, 1988). The rn locus has previously been molecularly analyzed (Agnel et al., 1989) and a cDNA encoding a member of the Rac family of GTPase-activating proteins (GAP) was isolated from this genomic region (Agnel et al., 1992b). Since this gene was located in the rn genomic region it was denoted the rotund racGAP (rnracGAP), but molecular analysis of multiple rn alleles indicated that the rnracGAP is not responsible for the rn phenotypes (Agnel et al., 1992a). In fact, all studies to date instead point to an uncharacterized larger transcript as the likely candidate for the rn gene (Agnel et al., 1992a; Hoemann et al., 1996).
The closely linked roughened eye (roe) locus affects a late step in the development of the eye, and roe mutants display rough eye morphology and reduction of photoreceptors (Renfranz and Benzer, 1989). The roe gene is genetically separable from rn, but the two genes show complex complementation (Brand and Campos-Ortega, 1990; Kerridge and Thomas-Cavallin, 1988; Ma et al., 1996). This previously led to the suggestion that rn and roe may be ‘two classes of mutation of the same gene, each of them disrupting a subfunction’ (Ma et al., 1996). To address the tight link between these two adjacent loci we have isolated the rn and roe genes. Intriguingly, our results show that roe is part of the rn gene but is represented by a different transcript. These two transcripts encode predicted proteins with an identical C-terminal region, containing a Krüppel-type zinc finger domain, but with different N-terminal regions. rn and roe are expressed in non-overlapping domains in the larval imaginal discs. Each cDNA can rescue only the corresponding mutation and when misexpressed in each others domain of expression has negative effects. Our results indicate that these two loci are genetically separable not only because of their differential expression but also because of distinct activities of the Rn and Roe proteins. By analyzing the expression of a number of markers in the developing imaginal discs, we further show that rn and roe act downstream of early patterning genes such as dachshund, but may act to modulate Notch signaling by regulating expression of Delta, Scabrous and Serrate.
MATERIALS AND METHODS
Fly stocks
w1118, roe3, UAS-lacZ, and pp,cu (Bloomington Stock Center); rn89 (Couso and Bishop, 1998) identified as P089 in Flyview stock collection (http//pbio07.uni-muenster.de); rn16, rn19, rn20 (Agnel et al., 1989); sev-GAL4 (A. Bailey and G. M. Rubin); GMR-GAL4 (Hay et al., 1997); UAS-rn#1, UAS-rn#32, UAS-roe#18, UAS-roe#88 and rnGAL4#5 (this study). Mutations were maintained over standard balancers with lacZ or GFP markers.
Isolation of rn and roe cDNAs
Using genomic fragment D (Agnel et al., 1989) (provided by R. Griffin-Shea) as a probe, three Drosophila cDNA libraries were screened for a total of 11 million plaques and colonies. A larval λgt11 cDNA library (Clontech) yielded a 1.3 kilobase pair (kb) positive clone (4H). Comparison of the 4H sequence with Drosophila genomic sequence revealed that the 4H cDNA was truncated on both ends owing to internal EcoRI sites. To obtain the remainder of the cDNA we used PCR to amplify a 700 bp fragment downstream of the 3′ EcoRI site and used this PCR fragment to screen the same larval library. From 4 million plaques a 2.3 kb clone (22-4) was isolated and sequenced. The compiled cDNA sequence (4H/22-4) contained a long open reading frame (ORF) encoding a putative protein of 945 amino acids (aa; GenBank AF395905). There are several putative start codons at the beginning of the ORF, one of which closely matches the Drosophila consensus (Cavener and Ray, 1991). Owing to internal EcoRI sites at the 5′ of clone 4H and the 3′ of clone 22-4, the precise extent of the rn gene was not determined. Immediately 3′ of clone 22-4 the genome sequence reveals a number of polyadenylation sites that likely are used as termination signals.
We used a 3′ fragment from rn clone 22-4 (bp 2714-3658 of GenBank AF395905) as a probe to screen the larval cDNA library used for isolation of the rn cDNA. This yielded 2 positive clones out of 5 million plaques. Both clones contained truncated roe cDNAs, corresponding to bp 332-2856 and 621-2856 (GenBank AF395904). Both inserts crossed the junction between exon 1 and exon 2 of the predicted roe gene, extending past the end of the Roe ORF. Since we did not obtain a full-length roe cDNA, we verified the structure of the roe transcript by amplifying part of it using RT-PCR. For this, RNA from w1118 embryos was isolated and purified using RNAsol (Tel-Test, Inc.) and Qiagen Oligotex (Qiagen). We designed a primer in the predicted first exon, 5′ to stop codons in all three reading frames and followed by the predicted Roe start methionine (TAAAATTGTGCTTGGACCAGTGAA), and 2 primers in exon 2 (ATGCGAGAGCTGCGTGAACTT and TGCGACAGATACGACGAGTTGG). Using these primers, nested PCR was performed and a product of the predicted size was generated. Sequencing of this fragment was in agreement with our prediction for the intron/exon structure of roe (GenBank AF395904).
Generation of UAS-rn and UAS-roe
rn sequences corresponding to position 0-3373 (GenBank AF395905) of rn cDNA, and roe sequences corresponding to 0-2160 of roe cDNA (GenBank AF395904) and 86 bp of upstream genomic DNA, were cloned into the pUAST vector (Brand and Perrimon, 1993). Three independent UAS-rn and eight independent UAS-roe transgenic lines were generated using P-element transformation (Spradling and Rubin, 1982). These lines were tested for expression using GMR-GAL4 and all gave strong phenotypes indicating similar levels of expression.
P-element analysis
The insertion of the rn89 enhancer trap, a P[lArB] insert, was determined using standard plasmid rescue methods. This revealed that P[lArB] is inserted at position –440 bp upstream of the rn cDNA (GenBank AF395905).
Conversion of P[lArB] in rn89 to P[GawB] was carried out as previously described (Sepp and Auld, 1999) with some modifications. Briefly, males of the genotype w1118, elavC155P[GawB];;rn89/D2-3,Sb were crossed to w1118 females and their progeny screened for red-eyed males (indicating that the P[GawB] had mobilized onto the autosomes). These males were crossed singly to UAS-GFP/TM3,Sb and their progeny screened for the rn expression pattern in larvae. From 30 lines screened, 3 independent insertions (rnGAL4#5, rnGAL4#13, rnGAL4#14) expressed GFP in the rn pattern and subsequently failed to complement rn. The site of insertion and the orientation of P[GawB] was determined by PCR amplification and sequencing. In all three cases P[GawB] was inserted in the exact same position as rn89 P[lArB]. For the rescue experiments rnGAL4#5 was used. The three rnGAL4 lines enhance the wing phenotype of Ser1, common to many third chromosome balancer lines (not shown).
To verify that the rn89 and rnGAL4#5 mutant phenotypes were due to the insertion of the P elements, we excised them by standard methods. For rn89, six independent revertants were isolated using their complementation of rn. Two independent revertant lines (rn#1–5 and rn#2–1) were homozygous viable and showed no rn phenotype. They were further analyzed by PCR and sequencing to determine the structure at the P-element insertion site. In both cases the P element had imprecisely excised but left a 30 bp (rn#1–5) and 37 bp (rn#2–1) ‘footprint’ containing the expected direct duplication of the 8 bp P-element target sequence and additional sequences from both ends of the P element. These ‘footprints’ are outside the identified rn exons thus explaining why they reverse the rn phenotype. Additionally, four stronger independent alleles were identified, one of which, rnΔ2–2 was analyzed in more detail. Southern blot analysis using multiple probes, revealed that rnΔ2–2 retained P[lArB] but is deleted for 3′ flanking genomic DNA removing the first and part of the second rn exon (Fig. 1A). For the reversion of rnGAL4#5 a similar strategy was used and we obtained 5 independent revertant lines that complemented multiple rn alleles, and in addition had lost the white marker and GAL4 expression.
Analysis of roe3
To identify the EMS-induced mutation in roe3, we amplified a 1.5 kb genomic region covering the first exon of roe (primers were ATGCGAGAGCTGCGTGAACTT and CCAAATGGAAGGCCGTCTCA). Three independent PCR fragments using genomic DNA from w1118, roe3/rn20 and ppcu1 were sub-cloned and three clones from each were sequenced (ppcu1 was used as a second control since the roe3 parental chromosome could not be obtained). We found several conservative changes between roe3 and each of the other two lines, but only one non-conservative change between roe3 and both w1118 and ppcu1. This was a nonsense C→T mutation resulting in a glutamine to amber stop codon change at aa position 191 (bp 629 in GenBank AF395904) in the Roe ORF (Fig. 1A). This mutation would truncate the predicted Roe protein and the mutant protein would lack the entire C-terminal region including the ZF domain.
In situ hybridization and immunohistochemistry
Standard in situ protocols were used to examine expression of rn and roe (Tautz and Pfeile, 1989). We used three probes, 4H, containing rn-only sequences (0-1331of GenBank AF395905), roe, containing the first exon of roe (0-785 of GenBank AF395904) and ZF, containing common 3′ sequences including the ZF domain (2016-3373 of GenBank AF395905). Sense probes showed no signal in embryos or larvae. For the roe rescue experiments, adult eyes were cryo-sectioned and immunostained for Elav, a marker for photoreceptors (O’Neill et al., 1994). More than 14 ommatidia from more than four flies per genotype were analyzed and the total number of R1-7 photoreceptors determined. For epistatic analysis, third instar imaginal discs were immunostained using the following primary antibodies: anti-Elav (1:10), anti-Dac (1:25), anti-Boss (1:2000), anti-Sca (1:10), anti-Ser (1:1000), anti-Bab (1:2000) and anti-Dl (1:20).
RESULTS
Isolation of rotund and roughened eye
To isolate the rn cDNA we used genomic fragment D (Fig. 1A), shown to hybridize to the putative rn transcript (Agnel et al., 1989). The cDNA sequence indicates that rn encodes a Krüppel-type zinc finger (ZF) protein and contains six C2H2 ZFs. The predicted Rn protein has a high degree of homology to the predicted protein of Drosophila gene CG5557 (Adams et al., 2000), and to C.elegans Lin-29 (Rougvie and Ambros, 1995). Over the ∼150 aa ZF domain these two proteins display 84-90% identity to Rn (Fig. 1C). Among mammalian proteins, a recently identified rat cDNA, Cas-Interacting Zinc finger (CIZ) (Nakamoto et al., 2000), displays the highest homology (59% in the ZF) to Rn. Rn and CG5557 also share a short C-terminal domain of high homology not found in the other proteins (Fig. 1C). In line with the complex genetics of this area, the alignment of the rn cDNA with the genomic sequence reveals that rn spans ∼50 kb and extends on both sides of the rnracGAP (Fig. 1A).
The roe gene shows complex complementation with rn and a number of roe alleles are also rn (Agnel et al., 1989; Brand and Campos-Ortega, 1990; Kerridge and Thomas-Cavallin, 1988; Ma et al., 1996). The rn gene structure together with previous molecular work on rn alleles gave us some initial insight into the identity of roe. Particularly informative were the rnΔ2–2 and rn19 alleles. The rnΔ2–2 P-element excision allele (materials and methods) contains a deletion in the rn 5′ region removing the first and part of the second exon of rn (Fig. 1A). Complementation analysis of rnΔ2–2 shows that it is a null allele of rn but does not cause roe phenotypes (see below). Furthermore, the rn19 allele, shown to contain a larger deletion in the rn 5′ region (Agnel et al., 1989), acts as a rn null allele and, although it removes at least one other lethal complementation group, does not cause roe phenotypes. These results indicated the existence of roe-specific functions encoded in the genomic region proximal to the breakpoint of rn19 (Fig. 1A). One model could be the existence of roe specific exon(s) that are spliced and utilized specifically in the eye. However, the fact that rn19 extends further distally, uncovering other complementation group(s), but does not produce roe phenotypes argues against eye-specific splicing of a long transcript originating from a promoter in the rn region. Instead, a more likely scenario would be the existence of an eye-specific promoter and exon(s). This notion was further supported by analysis of P-element insertions in the rn 5′ area that result in the rn phenotype and matching expression but not in the roe phenotype or eye disc expression (see below). These results prompted us to look for additional exons that could explain the molecular nature of the roe gene. By screening a larval cDNA library with a rn 3′ probe and by subsequent PCR analysis we isolated the roe cDNA. The roe gene utilizes the same two 3′ exons as rn but contains a different 5′ exon (Fig. 1A). As a result the predicted Roe protein shares the C-terminal region, including the ZF domain except the first finger, with Rn but differs in the N-terminal region (Fig. 1B). It is interesting to note that the rn genomic structure was not revealed by the analysis of the sequences carried out by the Drosophila Genome Project (Adams et al., 2000). Although parts of the rn coding regions were identified (CG14600, CG14601, CG14603 and CG10040), the rn transcript was not predicted, probably because rn has several small exons spread over 50 kb. In contrast, the roe transcript was accurately predicted, short of one aa error in the splice junction between exons 1 and 2 (CG10040). At the submission of this study, the rn and roe cDNAs had not been isolated in the BDGP or RIKEN expressed sequence tag (EST) projects.
Molecular analysis of rotund and roughened eye mutations
The genomic structure of the rn locus that we propose fits well both with previous studies as well as with our molecular analysis of rn and roe alleles. First, rn16 and rn20 are deletions that show both rn and roe phenotypes, while the rn19 deletion only shows rn phenotypes (Agnel et al., 1989). In agreement, rn16 deletes both the common ZF coding exons and roe-specific exons, rn20 deletes the whole region, and rn19 removes most of the rn-specific exons (Fig. 1A). Second, we sequenced roe3, a strong roe-specific allele, and show that it is the result of a nonsense mutation in the roe-specific exon. This mutation does not affect the common 3′ exons and explains why roe3 acts as a roe null allele but does not show rn phenotypes. Third, rn89, a lacZ-containing P-element transposon allele (Couso and Bishop, 1998) was shown to be inserted within the 5′ region of the rn gene. This explains why it only displays rn and not roe phenotypes. In addition, imprecise excision of rn89 yielded rnΔ2–2, which contains a deletion of the first and part of the second rn exon (Fig. 1A). As expected, rnΔ2–2 displays a rn null phenotype (Fig. 3C,I) but no eye phenotype. In agreement with this, in situ hybridization failed to detect any rn transcript in rnΔ2–2 mutant discs (not shown). We further generated rnGAL4#5 by P-element conversion of rn89. rnGAL4#5 displays a stronger leg phenotype than rn89, possibly due to differences in the structure of the P element, but again no aberrant eye phenotype (not shown). Wild-type revertants of rn89 and rnGAL4#5 were generated that complement other rn alleles, verifying that in both cases the rn phenotype was caused by the P-element insertion.
Expression of rotund and roughened eye
We detect expression of rn and roe in developing imaginal discs, as well as in the embryonic and larval CNS. Here we will focus on the expression in the imaginal discs. Expression of rn commences during the early third larval instar in the leg, wing, haltere and antennal part of the eye-antennal imaginal disc (Fig. 2E-H). Expression of rn is observed as a ring in the leg and antenna discs and in the presumptive wing pouch and capitellum of wing and haltere discs respectively. In late third instar, expression of rn in the leg disc is no longer evident, but is maintained in the other discs (Fig. 2G). We also studied the expression of lacZ in both rn89 and in rnGAL4#5/UAS-lacZ larvae to determine rn expression. In both genotypes, expression of lacZ is in agreement with the rn in situ hybridization, except for the persistence of tarsal expression (Fig. 2A-D), but in neither line do we detect expression in the eye disc. Expression of roe commences in the third instar and is confined to the eye part of the eye-antennal imaginal disc in a band of 4-6 cells at the morphogenetic furrow (Fig. 2I,K). We find no evidence of roe expression in other imaginal discs.
The expression of rn and roe is in agreement with the observed phenotypes. For instance, rn mutants have defects in wings and halteres, and correspondingly rn is expressed in the appropriate presumptive regions in wing and haltere imaginal discs. In the leg, rn mutants display fusion of all 5 leg tarsi into one fused tarsal-like segment. In agreement with this, rn is expressed in a sub-distal ring that represents the presumptive tarsus, as revealed by the persistent tarsal expression of rn-driven lacZ in late third instar discs. Similarly, roe specifically affects the eye, and mutants have rough eyes and reduced numbers of photoreceptors (Ma et al., 1996). Accordingly, we observe expression of roe in the eye part of the eye-antennal imaginal disc but not in other imaginal discs. The mutually exclusive patterns of expression of rn and roe raised the issue of whether they may in fact negatively regulate each other. To determine this, we analyzed the expression of roe in rn mutant imaginal discs and conversely the expression of rn in roe mutant imaginal discs. These studies revealed no apparent changes in the expression of rn and roe when compared to wild type, indicating that there is no cross-regulation between rn and roe (not shown).
Rescue of rotund
Owing to the complexity of the rn locus we wanted to further verify the authenticity of our rn and roe cDNAs by rescue experiments. For the rn rescue we focused on the leg phenotype and used the rnGAL4#5 line that shows strong leg phenotypes over rn20 (Fig. 3A,D,I). By providing rn function with UAS-rn, we observe rescue of the rnGAL4#5/rn20 leg phenotypes, often to a level indistinguishable from the wild-type leg (Fig. 3F,I, P<0.001). We do not observe any dominant effect in the leg of UAS-rn in a heterozygous background (Fig. 3E,I).
The structure of the rn genomic region and the differential expression in imaginal discs explains why rn and roe can be genetically separated and affect different tissues. However, the rn and roe gene products are also different, and the first ZF is truncated in the Roe protein (Fig. 1B), intriguing given that the first finger of Krüppel-type ZF proteins has been shown to be involved in DNA-binding (Avram et al., 1999; Hamilton et al., 1998). Rn and Roe further differ in the N-terminal regions where they contain stretches of glutamine/serine (Roe) or alanine (Rn), often found in transcriptional activator and repressor domains respectively (Gerber et al., 1994; Lanz et al., 1995; Licht et al., 1994; Madden et al., 1993; Nowling et al., 2000). This raised the possibility that these two proteins may have different activities and may not be interchangeable. To address this issue we misexpressed roe in the leg disc and also attempted to rescue rn with roe. When roe is misexpressed in the developing leg disc using rnGAL4#5, we noticed a negative effect with reduced number of tarsi, similar to rn mutants (Fig. 3G,I). Furthermore, in a rn mutant background (rnGAL4#5/rn20) we observe no evidence of rescue by UAS-roe (Fig. 3H,I).
Rescue of roughened eye
We also wanted to rescue roe mutants using the GAL4/UAS system. The roe rescue was complicated by the fact that we did not have a GAL4 insertion in the roe gene. This is especially relevant given the dynamic pattern of roe expression in the eye disc, with transient expression in a band of approx. 4-6 cells at the morphogenetic furrow (Fig. 2I,K). We were unable to identify a GAL4 line that would express precisely in the roe pattern and instead attempted to rescue roe using GAL4 drivers that would drive in photoreceptors. To this end, we tested several eye disc GAL4 driver lines for ectopic effects. Not surprisingly, strong pan-eye drivers such as GMR-GAL4 lead to dramatic phenotypes with loss of pigment and bristle cells (Fig. 4D). A novel sevenless-GAL4 (sev-GAL4) line that expresses GAL4 in the photoreceptors, cone and mystery cells (Fig. 4A,B) showed little if any sign of rough eye morphology when crossed to UAS-roe (not shown). Using sev-GAL4 crossed to UAS-roe in a roe null mutant background (rn16/rn20) we observe partial rescue of the eye phenotypes with increased eye size and reduced roughness (Fig. 5A-C). To quantify the roe rescue we counted the number of adult R1-7 photoreceptors in wild-type, mutant and rescued flies. These results confirm previous studies (Ma et al., 1996) and show that roe mutants have a reduced number of photoreceptors compared to wild type (Fig. 5E). In line with the apparent morphological rescue we find significantly increased numbers of photoreceptors in rescued flies when compared to mutants (P<0.04, Fig. 5E). Given that we were unable to use a GAL4 driver line that perfectly matched the dynamic expression of roe in eye discs, we believe that this partial rescue supports the proposed identity of the roe gene. As in the rn rescue experiments, we wanted to address whether rn is interchangeable with roe and could provide rescue activity in the eye. First we tested the activity of UAS-rn in the eye by misexpressing it using GMR-GAL4 and sev-GAL4. This leads to severe rough eye phenotypes with GMR-GAL4 (Fig. 4C) and little if any sign of rough eye morphology with sev-GAL4 (not shown). In a roe null mutant background (rn16/rn20) we find no evidence of rescue by adding UAS-rn (Fig. 5B-E).
Molecular context for rotund and roughened eye activity
Previous studies suggested that rn and roe act late during development of their respective tissues, perhaps during terminal differentiation (Godt et al., 1993; Renfranz and Benzer, 1989). To further explore the function of rn and roe during leg and eye development, we have examined the expression of genes that play key roles during development of these tissues. We first studied the leg disc and analyzed genes whose expression abuts or overlaps that of rn. Dachshund (Dac), a nuclear factor required for normal leg development, is expressed at early stages of leg development in a ring pattern that abuts the early rn-expressing ring (M. I. G., S. A. Bishop and J. P. C., unpublished). Bric a brac (Bab), a BTB-domain containing transcription factor, has been suggested to be active late in limb development and is expressed in a similar pattern to rn in the leg (Godt et al., 1993). Furthermore, bab mutants show similar (though not identical) phenotypes to rn mutants in the tarsal segments of the leg (Godt et al., 1993). Interestingly, neither Dac nor Bab appears to be regulated by rn as revealed by staining of third instar leg discs (Fig. 6A,B; not shown). These results suggest that rn might act in parallel to, or downstream of, dac and bab to specify tarsal segment identity. Ser, a ligand for the Notch (N) receptor, is expressed in presumptive joint areas in larvae and pupa leg discs and controls the development of the leg joints (Bishop et al., 1999). In wild-type mid-third instar leg discs, Ser is expressed in the first tarsal fold, which coincides with the rn-expressing ring. In rn, Ser is down-regulated in the tarsal ring but not outside it (Fig. 6C,D). In pupal leg discs, Ser expression, normally present in four stripes within the presumptive tarsal area (Fig. 6E), is present in fewer and less defined stripes in rn (Fig. 6F).
The roe rough eye phenotype is reflected in reduced numbers of photoreceptors present in adult ommatidia (Brand and Campos-Ortega, 1990) (this study). To determine whether roe mutants show early patterning defects in the eye-antennal disc, we analyzed expression of Dac, which plays an early role in the eye disc and is expressed in a broad domain spanning both sides of the morphogenetic furrow (MF) (Mardon et al., 1994). Since dac mutants have a more severe eye phenotype than roe we anticipated that Dac would not be regulated by roe, and as expected we observe no change in the pattern of Dac staining in roe when compared to wild type (Fig. 7A,B). Next we analyzed third instar eye-antennal discs with antibodies to Elav and to Bride of Sevenless (Boss), a marker of R8 photoreceptors (Hart et al., 1990). In wild-type eye discs, Elav and Boss are expressed in a stereotyped pattern immediately posterior to the MF (Fig. 7C,E). In roe mutants, expression of Elav and Boss reveals abnormal photoreceptor differentiation with apparent gaps in the expression of both markers posterior to the MF (Fig. 7D,F). Elav expression also indicates that photoreceptor clusters frequently have fewer photoreceptors than normal (Fig. 7E,F). Expression of Elav and Boss further reveals an apparent failure of the MF to progress in a straight line from dorsal to ventral. The MF appears to progress more slowly in some areas, creating a wave-like appearance of developing photoreceptor clusters near the MF (Fig. 7C-F). These results indicate that roe function is centered around the MF, a notion that fits well with the strong but transient roe expression seen at the MF (Fig. 2I,K). We therefore analyzed markers expressed at the MF, and since roe has been shown to interact genetically with the NSpl mutation (Brand and Campos-Ortega, 1990), we examined expression of Delta (Dl), a N ligand (Vassin et al., 1987), and Scabrous (Sca), a secreted glycoprotein implicated in N signaling (Baker et al., 1990). In wild type, Dl and Sca are expressed in clusters of cells at the MF, and expression is maintained posterior to the MF in subsets of cells (Fig. 7G,I). In roe mutants, the punctate expression of Dl and Sca is lost at the MF and replaced by a diffuse band of expression. Posterior to the MF, expression is punctate but appears disorganized (Fig. 7H,J).
DISCUSSION
The rn and roe loci are tightly linked and this study reveals the underlying molecular basis for this linkage. Intriguingly, our work shows that roe is part of the rn gene and is represented by a related but distinct transcript. The rescue and misexpression experiments support the notion that rn and roe play different roles during imaginal disc development not only because of their differential expression but also because of distinct activities of the Rn and Roe proteins. These activities could involve different target DNA sequences and/or different transcriptional effects, perhaps based on their different ZF and glutamine/alanine/serine stretches.
Regarding the function of the rnracGAP, both our work and previous studies argue against any involvement of rnracGAP in the rn or roe phenotypes (Agnel et al., 1989; Agnel et al., 1992a; Hoemann et al., 1996). In situ studies indicate that rnracGAP is only expressed at low levels in the imaginal discs during pupal stages (Agnel et al., 1989; Agnel et al., 1992a; Hoemann et al., 1996). In addition, there is no obvious difference in the severity of rn and roe phenotypes whether or not the rnracGAP is simultaneously removed. For instance, we have found no significant difference in the severity of rn leg phenotypes in rn20/rn20 (that removes rn, roe and rnracGAP) compared to rn19/rn20 (rn19 does not remove rnracGAP). Similarly, roe3/rn20 (roe3 has a premature stop codon in the roe-specific exon) displays as severe of an eye phenotype as rn20/rn20 (not shown). Furthermore, we can rescue rn and roe mutants with the rn and roe cDNAs. Recent studies may indicate an involvement of rnracGAP specifically in male fertility, and high levels of rnracGAP expression have been observed in the adult testis (Agnel et al., 1989; Agnel et al., 1992a; Hoemann et al., 1996). The rn89 and rnGAL4#5 P-element insertions described here may provide useful starting materials for the generation of mutations specifically affecting the rnracGAP by local P-element mobilization.
Little is known about the genetic cascades within which roe and rn are acting. The results from eye-antennal imaginal discs indicate that roe acts at the morphogenetic furrow, as evident both from its expression and from the effects on Dl and Sca expression in roe mutants. Both Dl and sca play roles in spacing the array of ommatidial preclusters in the morphogenetic furrow (Baker et al., 1990; Baker and Zitron, 1995; Ellis et al., 1994), and it is interesting to note that the expression of roe at the furrow is not evenly distributed and appears stronger in clusters of cells (Fig. 2I). Genetic screens for modifiers of the Nspl mutation identified roe as an enhancer, and sca and Dl as suppressors of the Nspl eye phenotype (Brand and Campos-Ortega, 1990). Given the dynamics of N signaling, these results support models where Roe acts to either positively or negatively regulate Dl and Sca. A genetic interaction screen for enhancers of glass also identified roe (Ma et al., 1996), an interesting finding given that ectopic expression of roe using GMR-GAL4 leads to a glass-like phenotype with a loss of bristles and pigment cells (Fig. 4E,F).
In the leg, rn expression is the earliest marker known for tarsal development (Couso and Bishop, 1998). rn is required for the development of this region and for its subsequent patterning, as observed by the loss of Ser expression. Thus, the transient expression of rn in the leg might reveal that the intercalation of the presumptive tarsal region between the distal tip and medial leg regions occurs during early third instar.
It is increasingly common, even in invertebrates, to find genes that utilize two or more promoters (Gower et al., 2001; Krishnan et al., 1995; Li et al., 1999; Mevel-Ninio et al., 1995). Although this may lead to the generation of different proteins, it is often unclear whether the proteins have distinct activities. In fact, this issue is not easily resolved by traditional forward genetics and subsequent molecular analysis, since even if a locus can be genetically dissected into different subfunctions, this does not identify whether the different proteins have distinct activities. Perhaps the best way to test whether the variant proteins are interchangeable in vivo, is by cross-rescue in each others domain of expression. The rn gene is a clear example of a locus that utilizes both tissue-specific promoters and functionally distinct proteins to achieve functional diversity, a scenario likely to be observed more and more frequently in the post-genomic era.
Acknowledgements
We thank J. B. Thomas and P. H. Taghert for advice. We thank A. Bailey and G. M. Rubin for the sev-GAL4 lines, and R. Griffin-Shea for sharing fly lines and DNAs. G. Gloor provided advice on P-element conversion. We thank L. Zipursky, K. Irvine and E. Knust for antibodies. We are grateful to The Bloomington Stock Center for providing fly lines. We thank N. Perrimon, A. Michelson and A. Simcox for critically reading the manuscript. B. Borsari provided excellent technical assistance. This work was supported by grants from NIH (RO1 NS39875-01, T32MH20017) to S. T. and S. E. S. and by The Wellcome Trust (Senior Research Fellowship and supplementary Grant) to J. P. C. and M. I. G.