Drosophila imaginal discs undergo extensive pattern formation during larval development, resulting in each cell acquiring a speciﬁc adult fate. The ﬁnal manifestation of this pattern into adult structures is dependent on pulses of the steroid hormone ecdysone during metamorphosis, which trigger disc eversion, elongation and differentiation. We have deﬁned genetic criteria that allow us to screen for ecdysone-inducible regulatory genes that are required for this transformation from patterned disc to adult structure. We describe here the ﬁrst genetic locus isolated using these criteria: crooked legs (crol). crol mutants die during pupal development with defects in adult head eversion and leg morphogenesis. The crol gene is induced by ecdysone during the onset of metamorphosis and encodes at least three protein isoforms that contain 12-18 C2H2 zinc ﬁngers. Consistent with this sequence motif, crol mutations have stage-speciﬁc effects on ecdysone-regulated gene expression. The EcR ecdysone receptor, and the BR-C, E74 and E75 early regulatory genes, are submaximally induced in crol mutants in response to the prepupal ecdysone pulse. These changes in gene activity are consistent with the crol lethal phenotypes and provide a basis for understanding the molecular mechanisms of crol action. The genetic criteria described here provide a new direction for identifying regulators of adult tissue development during insect metamorphosis.
The spectacular biological transformations associated with insect and frog metamorphosis provide an ideal context for understanding the hormonal regulation of key developmental pathways, including tissue morphogenesis, remodeling and programmed cell death (Gilbert et al., 1996). We are studying the metamorphosis of the fruit fly, Drosophila melanogaster, as a model system for understanding how steroid hormones direct stage- and tissue-speciﬁc biological responses during development. Pulses of the steroid hormone 20hydroxyecdysone (referred to here as ecdysone) signal the destruction of larval tissues and their replacement by adult tissues and structures during Drosophila metamorphosis (Riddiford, 1993). The external structures of the adult head and thorax develop from imaginal discs that proliferate and undergo pattern formation during larval development (Cohen, 1993). Pulses of ecdysone then signal the imaginal discs to evert and differentiate into their appropriate adult form (Fristrom and Fristrom, 1993; von Kalm et al., 1995). Although a great deal is known about how imaginal discs are spatially patterned, much less is understood about how these patterns are elaborated into speciﬁc adult structures in response to ecdysone during metamorphosis. In this paper, we deﬁne genetic criteria that allow us to identify genes that regulate this hormone-driven transformation, from patterned disc to differentiated adult structure.
Two high titer pulses of ecdysone direct the animal through the early stages of metamorphosis (Riddiford, 1993). A high titer ecdysone pulse at the end of larval development triggers puparium formation initiating the prepupal stage of development. This phase is terminated by another ecdysone pulse, approximately 10 hours after puparium formation, that triggers adult head eversion and signals the onset of pupal development (Sliter and Gilbert, 1992). Initial insights into the mechanisms by which ecdysone directs these developmental responses arose from studies of the puffing patterns of the larval salivary gland polytene chromosomes (Becker, 1959; Ashburner et al., 1974). Ecdysone directly induces a small set of early puffs in the polytene chromosomes. Some of these puff genes encode transcription factors that both cross-regulate their own activity and induce large sets of secondary-response late puff genes (Russell and Ashburner, 1996; Thummel, 1996). The late genes, in turn, appear to direct the appropriate stage-and tissue-speciﬁc developmental responses to ecdysone during metamorphosis.
Two early puff genes have been characterized in detail: the Broad Complex (BR-C) (DiBello et al., 1991) and E74 (Burtis et al., 1990). The BR-C and E74 genes are complex and contain multiple nested ecdysone-inducible promoters that direct the synthesis of families of related protein isoforms. The BR-C encodes multiple protein isoforms that can be distinguished by their use of one of four possible pairs of C2H2 zinc ﬁngers, while the E74A and E74B proteins share an identical ETS DNA-binding domain. The BR-C and E74 are widely expressed and are required for the proper development of both larval and imaginal tissues during metamorphosis (Kiss et al., 1988; Thummel et al., 1990; Boyd et al., 1991; Emery et al., 1994; Fletcher et al., 1995).
Interestingly, aspects of the BR-C and E74 mutant phenotypes bear resemblance to one another. For example, some E74B mutant pupae display cryptocephalic phenotypes and short, twisted legs (Fletcher et al., 1995). This latter phenotype, referred to as the malformed leg phenotype, is a characteristic of some heteroallelic combinations of BR-C mutant alleles and is associated with incomplete disc eversion during prepupal development (Kiss et al., 1988). BR-C and E74 mutants also show similar effects on the transcription of some ecdysone-regulated target genes, providing a molecular framework for interpreting the similarity in their developmental phenotypes (Guay and Guild, 1991; Karim et al., 1993; Fletcher and Thummel, 1995).
These related mutant phenotypes suggest that a number of ecdysone-inducible regulatory genes may circuit into these complex developmental pathways. The identiﬁcation of these genes based on their puffing pattern in the larval salivary gland polytene chromosomes, however, has strict limitations. First, not all genes that are expressed in the salivary glands form puffs in the polytene chromosomes. Second, at least some genes that affect adult tissue development are not expressed in larval tissues (Appel et al., 1993). Third, mutant phenotypes can only be discerned after the difficult and time-consuming effort of reverse genetics. Indeed, the absence of genetic screens for mutants that affect metamorphosis has been a signiﬁcant hindrance to progress in this area of research.
In an effort to circumvent these difficulties, we have established a set of genetic criteria that allow us to identify new ecdysone-inducible genes that play a role in adult tissue development during metamorphosis. We describe here the ﬁrst results of this effort, the isolation of a new pupal lethal mutant called crooked legs (crol). crol mutants die during pupal development with two phenotypes reminiscent of mutations in the BR-C and E74: an inability to properly evert their head and defects in leg morphogenesis. Furthermore, crol mutations speciﬁcally affect the transcription of a number of ecdysone-induced genes during the prepupal-pupal transition, including the EcR ecdysone receptor gene (Koelle et al., 1991), as well as the BR-C, E74 and E75 early puff genes. The crol locus encodes at least three related protein isoforms that contain 1218 clustered C2H2 zinc ﬁngers, suggesting that crol manifests its effects on metamorphosis by directly regulating gene expression. Finally, crol transcription is induced by ecdysone during late larval and prepupal development, and appears to be expressed in a number of ecdysone target tissues including imaginal discs, salivary glands and the central nervous system. The crol gene thus appears to be a new regulator of the prepupal and early pupal genetic responses to ecdysone. The identiﬁcation of crol provides a means of extending our studies of metamorphosis beyond the conﬁnes of the puffing response in the larval salivary glands, directly into the morphogenesis of adult tissues.
MATERIALS AND METHODS
Fly stocks and phenotypic characterization
The crol alleles are available through the Berkeley Drosophila Genome Project as l(2)04418, l(2)06470 and l(2)k08217 (Spradling et al., 1995). The crol4418 allele was kindly provided by S. Shaver and A. Hilliker as part of a collection of prepupal and pupal lethal mutants (S. A. Shaver, M. B. Sokolowski, and A. J. Hilliker, personal communication). The crol6470 and crolk08217 alleles were provided by T. Laverty and G. M. Rubin. l(2)04418 and l(2)06470 are derived from the mutagenesis described in Karpen and Spradling (1992) while the l(2)k08217 allele is from Torok et al. (1993). The y; cn1; ry506 parental stock was provided by A. Spradling. Df(2L)esc10 was obtained from the Bloomington stock center. Other stocks are described in Lindsley and Zimm (1992). Flies were raised on a standard cornmeal/molasses/yeast medium at 25° C.
For lethal phase analysis, 300 eggs were collected from y w; crol4418cn/CyO y+ flies after an 8 hour egg laying period, and transferred to a Petri dish with hard agar and fresh yeast paste. Hatched ﬁrst instar larvae were collected after 30 hours and crol homozygous mutant larvae were distinguished from their crol/CyO y+ siblings by the yellow phenotype of their mouth hooks and denticle belts. CyO y+ homozygotes die during embryogenesis. The number of crol homozygous larvae was 33% of the total, indicating no embryonic lethality. To examine possible larval lethality, 80 crol4418 homozygous and crol4418/CyO y+ heterozygous ﬁrst instar larvae were transferred to different vials and allowed to develop for about 3 days, after which both mutant and control third instar larvae were counted. To analyze prepupal or pupal lethality, 60 third instar larvae of both genotypes were collected and allowed to develop at 25° C. The vials were checked every day for 7 days in order to score for a delay in pupariation as well as progression through metamorphosis and adult eclosion.
Lethal phases were also determined in animals that carried crol4418 in combination with Df(2L)esc10 as well as controls that carried y; cn/Df(2L)esc10. 100 eggs were collected from y w; Df(2L)esc10/CyO y+ animals mated with y w; crol4418cn/CyO y+ animals, and 32% of the hatched ﬁrst instar larvae were yellow mutants, indicating complete survival of the crol4418/Df(2L)esc10 embryos. Similarly, 100 embryos collected from y w; Df(2L)esc10/CyO y+ animals mated with y; cn; ry506 animals yielded 48% yellow mutant ﬁrst instar larvae, indicating that there are no early haploinsufficient phenotypes associated with the crol locus. Lethal phase analysis of later stages was performed using either 80 ﬁrst instar larvae, or 70 third instar larvae of the same genotypes: y w; crol4418cn/Df(2L)esc10 and y; cn/ Df(2L)esc10 as a control.
For phenotypic characterization, crol4418cn homozygous or crol4418cn/Df(2L)esc10 hemizygous wandering third instar larvae were collected and allowed to develop at 25° C. Animals were followed for 7 days and scored for morphological markers representative of successive pupal stages (Bainbridge and Bownes, 1981). To document the phenotypes, each pupa was embedded in a small drop of water (10 µ l) to dissolve the glue and gently detached from the wall using a wet paintbrush. About 15-20 pupae were transferred to a 1.5 ml Eppendorf tube containing 1 ml of boiling water, and left for 2 minutes at 95°C. The pupal case was then gently dissected in water for analysis of pharate adults. All animals were viewed by dark-ﬁeld microscopy and photographed with Kodak Ektar 25 color print ﬁlm. Legs were dissected in PBS, cleared by incubation overnight in 10% KOH at room temperature, dehydrated in ethanol and mounted in Euparal.
Staged third instar larvae and prepupae were dissected in PBS and ﬁxed in 2.5% glutaraldehyde, 50 mM sodium cacodylate, pH 7.0, in PBS for 10-15 minutes. Organs were then rinsed three times with PBS and incubated in a staining solution containing 0.4% X-Gal for 16-20 hours at 25° C (Ashburner, 1989). After two rinses in PBS, the stained organs were mounted in 50% glycerol in PBS and viewed and photographed using a Zeiss Axiophot microscope.
Northern blot hybridization and organ culture
Third instar larvae were maintained on food containing 0.1% bromophenol blue and staged essentially as described (Andres and Thummel, 1994). Staged prepupae were synchronized at the white prepupal stage (0 hour prepupae) and allowed to age at 25°C for the appropriate time. For the northern blot hybridization shown in Fig. 6B, crol alleles were balanced over a CyO y+ chromosome in a y w background. Homozygous animals were identiﬁed by the yellow phenotype of their mouth hooks and denticle belts. For the northern blots shown in Figs 5 and 8A, yellow animals were collected from the following crosses: for crol™ mutants, y w; Df(2l)esc10/CyO y+ virgin females were crossed to yw; crol4418cn/CyO y+ males, and for control animals, y w; Df(2l)esc10/CyO y+ virgin females were crossed to y; cn1; ry506 males. In all cases, 10 µg of total RNA was used in each lane. Hybridization with rp49 was used as an internal control for loading, transfer and hybridization conditions.
RNA was extracted, fractionated by formaldehyde agarose gel electrophoresis and blotted as described (D’Avino et al., 1995). Filters were hybridized, washed and stripped as described (Karim and Thummel, 1991). DNA probes are described in Andres et al. (1993), except the Brg-P9 cDNA probe, described in Emery (1995), and the Sb probe, which was synthesized from the EcoRI-SalI fragment at the 5′ end of the cDNA described by Appel et al. (1993). This fragment does not contain the coding region for the Sb protease domain. Probes were labeled by random priming (Prime-It kit, Stratagene) of gel-puriﬁed fragments.
The genomic regions flanking the P-element insertions in crol4418 and crol6470 were isolated by plasmid rescue as described by Hamilton and Zinn (1994). The region flanking the insertion in crolk08217 was isolated by inverse PCR as described (Yeo et al., 1995). The rescued fragment from the crol4418 allele was used to map this insertion on the walk of Frei et al. (1985) by Southern blot 90 hybridization and restriction mapping. A 2.3 kb HindIII fragment that gave the best signal in northern blot hybridizations (‘probe’ in Fig. 6A) was used to screen seven cDNA libraries. These included five oligo(dT)-primed and random-primed cDNA libraries constructed from RNA isolated from staged prepupae (provided by P. Hurban), one library prepared from RNA isolated from third instar larval organs cultured in the presence of 20-hydroxyecdysone and cycloheximide (Hurban and Thummel, 1993) and one library prepared from 8-12 hour embryos (Brown and Kafatos, 1988). The longest cDNAs obtained were sequenced on both strands using oligonucleotide primers and the Prism Dye Primer kit on an ABI 373A automated sequencer. A combination of DNA sequencing, Southern blot hybridization and PCR was used for intron/exon mapping. PCR ampliﬁcations were performed in an Idaho Technology air thermocycler essentially as described in Fisk and Thummel (1995). The 2.3 kb EcoRI-HindIII genomic fragment containing the third exon was completely sequenced on one strand. RT-PCR was performed using a Perkin-Elmer 480 DNA thermal cycler with the GenAmp RNA PCR kit (Perkin-Elmer) according to the manufacturer’s instructions. To identify the crolα isoform, an oligonucleotide located in the sixth exon of crolγ (E22: 5′ TGTGGTTTCTCACCCGAATG 3′) was used to reverse transcribe (30 minutes at 42° C) RNA isolated from Canton S white prepupae. This cDNA was then ampliﬁed by PCR using a second primer (7A14: 5′ TACGCGCAAGGAACACTATG 3′) located in third exon downstream from the crolγ splice site. An ampliﬁed fragment was obtained of the appropriate size for the crolα isoform and its identity was conﬁrmed by Southern blot analysis and a second round of PCR ampliﬁcation using nested primers.
Identiﬁcation and initial characterization of crol mutants
For our initial screen, we restricted our efforts to collections of prepupal/pupal lethal mutants, the times during development when BR-C and E74 mutations lead to lethality. We further restricted our analyses to collections of single P-elementinduced mutations in order to facilitate subsequent gene isolation and to avoid being misled by phenotypes due to multiple mutations. We eliminated mutations that cause a delay in puparium formation in order to avoid genes that might be involved in ecdysone metabolism. Subsequently, we selected those mutants that speciﬁcally affect the morphogenesis and/or differentiation of the imaginal discs during prepupal or pupal development, but neither their size nor morphology in late third instar larvae. Finally, we prepared northern blots from RNA isolated from mutant animals and screened for defects in ecdysone-regulated gene transcription during the onset of metamorphosis. In this way, we directed our efforts toward the isolation of genes that affect imaginal disc differentiation through the mis-regulation of target genes in the ecdysone regulatory hierarchies.
Using these criteria, we identiﬁed three P-element-induced pupal lethal mutations that deﬁne a single lethal complementation group: l(2)04418, l(2)06470 and l(2)k08217. These mutations all behave as recessive loss-offunction alleles. Based on the mutant phenotype, we have named the corresponding locus crol, for crooked legs. As expected from the criteria used in our screen, crol mutants show no delay in puparium formation and their imaginal discs are normal in size and morphology. Furthermore, we see normal patterns of engrailed and distalless expression in crol4418 mutant leg discs dissected from late third instar larvae, indicating that there are no apparent defects in disc patterning (data not shown).
Lethal phase analysis revealed that homozygous crol4418 mutants survive through embryogenesis and hatch to form ﬁrst instar larvae (see Materials and methods). Furthermore, collections of crol4418 ﬁrst and third instar larvae progress normally through prepupal development, but die before adult eclosion (Fig. 1). An identical lethal phase can be seen in animals that carry crol4418 in combination with Df(2L)esc10 (Fig. 1), a 380 kb deﬁciency that removes the crol locus (from 33A1-33B2, Frei et al., 1985). This observation suggests that crol4418 represents a null allele for this locus, a conclusion that is consistent with the absence of crol transcription in this mutant (Fig. 6B). Genetic and molecular analysis of crol6470 and crolk08217 demonstrated that these are also null alleles with a lethal phase essentially identical to that of crol4418 (data not shown).
crol mutants die during two stages of pupal development. The ﬁrst lethal phase occurs at the beginning of pupal development, 14-16 hours after puparium formation, corresponding to stage P5 of Bainbridge and Bownes (1981). These mutants display severe defects in head eversion and leg elongation (Fig. 2B). The remaining 40-50% of the mutants die at the end of pupal development, stage P14, with a microcephalic phenotype and malformed legs (Fig. 2D). While the penetrance of the microcephalic phenotype is variable, the ‘crooked legs’ phenotype is highly penetrant and is similar in all animals examined. This can be easily seen in the third pair of legs, which are smaller than wild type and distorted (Fig. 3B). We also see a kink near the middle of the femur (Fig. 3B) and a few missing bristles (Fig. 3C). The other two crol alleles display lethal phenotypes that are indistinguishable from those seen in crol4418 mutants (data not shown). Finally, excision of the P element in crol4418 results in a complete rescue of lethality, indicating that these lethal phenotypes are due to the P-element insertion (data not shown).
Accordingly, we examined the morphology of crol4418 mutant leg discs 6 hours after puparium formation, when they should have completed their eversion and elongation (Fristrom and Fristrom, 1993). Legs dissected from control 6 hour prepupae showed a similar size and shape, indicative of complete eversion and elongation (Fig. 3D). In contrast, a range of defects are evident in legs dissected from crol4418 prepupae (Fig. 3E,F). Of nine mutant animals examined, three contained legs with less severe defects. The 2nd-5th tarsal segments of these legs are slightly more expanded than wild type and the leg is not fully elongated (Fig. 3E). The remaining six crol4418 mutants had legs that were severely distorted and reduced in length (Fig. 3F). It is possible that these legs arise from the class of crol mutants that die early in pupal development with more severe leg defects (Fig. 2B). These observations indicate that at least part of the crooked legs phenotype is due to defects in leg disc elongation.
lacZ is widely expressed from the crol4418 P-element insertion and is induced at puparium formation
The crol4418 allele was derived from a P-lacZ enhancer trap mutagenesis, and thus carries a lacZ reporter gene that should provide an indication of the temporal and spatial patterns of crol expression (Karpen and Spradling, 1992). As shown in Fig. 5, lacZ is induced in leg imaginal discs, salivary glands and the central nervous system (CNS) isolated from crol4418/+ late third instar larvae and prepupae. This induction of lacZ expression is coincident with the high titer ecdysone pulse that triggers puparium formation suggesting that crol expression is regulated by ecdysone.
Expression of lacZ can be detected initially in leg imaginal discs isolated from late third instar larvae (™ 4 hours) and is restricted to the precursors of the tarsal segments. This expression expands in early prepupal leg discs to the precursors of the femur and tibia. Other imaginal discs in crol4418 animals do not express lacZ (data not shown).
Expression of lacZ in the salivary glands is induced at puparium formation, slightly later than lacZ induction in leg discs (Fig. 4). An identical pattern of expression is present in both fat bodies and trachea (data not shown). In contrast, β -galactosidase can be detected in the ventral ganglion and presumptive optic lobes of the CNS in mid-third instar larvae. Expression in these cell types increases noticeably at puparium formation, in apparent synchrony with lacZ expression in the salivary gland. Interestingly, lacZ is also expressed speciﬁcally in the corpus allatum of the ring gland at all stages examined. The corpus allatum is the endocrine organ responsible for releasing juvenile hormone (Riddiford, cell type is, however, difficult to predict since the role of juvenile hormone is not well understood during pre-adult Drosophila development.
crol is required for proper ecdysone-regulated gene expression during prepupal development
The BR-C and E74 both manifest their effects on metamorphosis through the activation of genetic regulatory hierarchies. In order to determine if crol also functions in these hierarchies, we examined the temporal patterns of transcription for a number of ecdysone primary- and secondary-response genes in crol mutant animals. RNA was isolated from staged third instar larvae and prepupae hemizygous for the crol4418 allele (y w; crol4418cn / Df(2L)esc10) or for the crol4418 parental chromosome (y; cn/Df(2L)esc10) and analyzed by northern blot hybridization (Fig. 5). Identical results were obtained using RNA samples isolated from an independent time course, as well as from staged crol6470 homozygotes. Only one difference was noticed between the two alleles, affecting the levels of β FTZ-F1 transcription (see below). The transcription patterns observed in the parental control stock (cn/Df) are essentially 1993). A possible function for crol in this identical to those reported in wild-type Canton S animals (Andres et al., 1993).
We examined the expression patterns of several genes that encode key transcription factors in the ecdysone cascades. These include the EcR ecdysone receptor gene as well as the BR-C, E74A, E74B, E75A, E75B, DHR3 and β FTZ-F1. E75A and E75B are two isoforms of the E75 early puff gene that encode orphan members of the nuclear receptor superfamily (Segraves and Hogness, 1990). DHR3 and β FTZ-F1 encode distinct orphan receptors, with DHR3 functioning as an inducer of β FTZ-F1 expression in mid-prepupae (Lavorgna et al., 1993; Lam et al., 1997; White et al., 1997). E75B inhibits this DHR3 activation function through direct heterodimerization (White et al., 1997). β FTZ-F1, in turn, appears to function as a competence factor that facilitates the reinduction of the early genes by ecdysone in late prepupae (Woodard et al., 1994).
DHR3 is speciﬁcally expressed in early prepupae and is unaffected by crol mutations (data not shown). In contrast, the other genes are all expressed at later stages and, interestingly, their transcription is selectively reduced in mid- and late crol4418 mutant prepupae. EcR and E74B are both submaximally transcribed in crol4418 mid-prepupae (Fig. 5). Similarly, the peak of BR-C, E74A, E75A and E75B transcription in response to the prepupal ecdysone pulse is signiﬁcantly reduced, while the earlier induction of these genes in response to the late larval ecdysone pulse is unaffected (Fig. 5). Consistent with the stage-speciﬁcity of this mutant phenotype, we also see a signiﬁcant reduction in the transcription of the stage-speciﬁc early gene E93 (Baehrecke and Thummel, 1995). The timing of these transcriptional responses conﬁrms that crol mutations have no effect on the duration of larval and prepupal development, but rather indicates that crol is required for the proper magnitude of ecdysone-induced gene expression in prepupae. The level of β FTZ-F1 mRNA is also reduced in crol4418/Df mutants (Fig. 5). However, crol6470 homozygotes show only an approximate two-fold reduction in β FTZ-F1 mRNA levels, yet the reduction in early gene transcription in these mutants is indistinguishable from that seen in crol4418 mutants (data not shown). This observation suggests that crol works independently of β FTZF1 to regulate the prepupal genetic response to ecdysone.
We also examined several ecdysone-inducible genes that are expressed speciﬁcally in the imaginal discs and salivary glands. The expression of four salivary gland-speciﬁc secondary-response genes is unaffected by crol mutations: the Sgs-3 and Sgs-5 glue genes (Meyerowitz and Hogness, 1982; Guild and Shore, 1984), and the L71-1 and L71-6 late puff genes (Wright et al., 1996) (data not shown). This is consistent with the wild-type expression of known regulators of these secondary-response genes, the BR-C and E74A, in crol mutant late third instar larvae (Fig. 5). Similarly, expression of the fat body speciﬁc gene Fbp-1 (Lepesant et al., 1978) is unaffected by crol mutations (data not shown). Because of the leg phenotypes in crol mutants, we also examined four genes that appear to play a role in imaginal disc morphogenesis. These included the Stubble gene (Sb), which encodes an apparent integral membrane serine protease required for disc eversion (Appel et al., 1993), the Brg-P9 secondary-response gene, which encodes an apparent protease inhibitor (Emery, 1995), IMP-E1, which is induced directly by ecdysone and may play a role in epithelial movements during disc eversion (Natzle et al., 1988), and the EDG-84A pupal cuticle gene (Fechtel et al., 1988). None of these genes are signiﬁcantly affected in crol mutants, with the exception of Brg-P9, which appears to be expressed for a longer time and at higher levels in crol mutant prepupae (Fig. 5).
The crol gene encodes at least three related C2H2 zinc ﬁnger proteins
The P element in crol4418 maps to a region of the genome that is covered by the chromosomal walk of Frei et al. (1985). Southern blot hybridization using genomic probes rescued from the P-element insertions in crol4418, and crol6470 demonstrated that these P elements lie within several kilobase pairs of one another at coordinate ™ 90 on the genomic walk. This corresponds to position 33A6-7 in the polytene chromosomes and positions the P-element insertions approximately 15-20 kb from the esc gene at 33B1-2 (Fig. 6A). Two transcripts, 5.3 and 6.0 kb in length, were detected by northern blot hybridization using probes from a 7 kb region surrounding the crol4418 P-element insertion site (Fig. 6B). These RNAs were not detectable in the three crol mutant backgrounds, strongly suggesting that these transcripts correspond to the crol gene (Fig. 6B).
84 cDNA clones were isolated from screens of seven cDNA libraries, and the six longest overlapping clones were sequenced on both strands. Introns were mapped by genomic DNA sequencing, Southern blot hybridization and PCR ampliﬁcation. These studies deﬁned two mRNA isoforms that we have called crolβ and crolγ . Two independent cDNA clones contain the complete crolβ coding region, and one cDNA contains a complete crolγ coding region. The 5′ sequences of the crolβ and crolγ cDNAs overlap an independent clone that contains the 5′ non-coding exons 1 and 2, leading to the proposed gene structure shown in Fig. 6A. The crolß and crolγ isoforms are generated by differential splicing such that crolβ lacks the sixth exon of crolγ, and a splice donor site within the third exon of crolβ is used to generate the crolγ mRNA. The existence of these mRNA isoforms was conﬁrmed by RT-PCR analysis of RNA isolated from white prepupae (data not shown). A third mRNA isoform was also identiﬁed by RTPCR, designated crolα, which utilizes the distal exon 3 splice site of crolβ and carries the sixth exon of crolγ . The crol gene thus encodes at least three different mRNA isoforms: crolα is 6263 nucleotides in length, crolβ is 6050 nucleotides in length, and crolγ is 5645 nucleotides in length. It is unclear, however, how these isoforms relate to the 6.0 and 5.3 kb RNAs detected by northern blot hybridization (Fig. 6B). Only the 6.0 kb RNA corresponds in length to a crol mRNA isoform – crolβ . Furthermore, probes derived from exons 1 and 2, the 3′ end of exon 7, or 3′ sequences of exon 3, which are speciﬁc to crolα and crolβ, all detect both 6.0 and 5.3 kb RNA size classes on northern blots. This observation suggests that the crol mRNA isoforms may utilize more than one promoter and/or 3′ polyadenylation signal. It seems likely that additional crol mRNA isoforms will be identiﬁed by further molecular characterization of this locus.
The locations of the P elements in crol4418crol6470 are consistent with a complete elimination of crol transcription in these mutant animals (Fig. 6B). The crol6470 mutation arises from a P-element insertion in the ﬁrst intron of the crol gene, and crol4418 and crolk08217 are due to P-element insertions in the second intron (Fig. 6A). Furthermore, the promoters for all crol mRNA isoforms must lie upstream from, or near, the crol6470 insertion site, since no crol transcripts can be detected in this mutant background (Fig. 6B).
Sequence analysis revealed that the crol mRNAs contain an unusually long 5′ leader sequence of at least 1418 nucleotides. Furthermore, this 5′ leader contains 13 AUG triplets upstream from the long open reading frame that encodes the crol protein products. Most of these 5′ open reading frames are short (1-22 codons in length) and the AUG triplets are in a poor context for optimal translational initiation (Kozak, 1986; Cavener and Ray, 1991). There are two exceptions to this rule: the tenth open reading frame, six codons in length, has only a one nucleotide mismatch from the optimal AUG sequence context. More interestingly, the ﬁrst open reading frame in the crol mRNAs shows a similarly favorable sequence context for translational initiation, and it is the longest open reading frame of the 5′ leader – encoding a 37 amino acid arginine- and serine-rich polypeptide. It thus seems likely that this short open reading frame is normally translated in vivo, although its possible function remains unknown. Long 5′ leader sequences seem to be a characteristic of only a few Drosophila regulatory genes, including Antennapedia and the ecdysone-inducible E74A early gene (Burtis et al., 1990; Oh et al., 1992). Both of these genes use internal ribosome entry to allow translation of their long open reading frame, suggesting that crol mRNAs may also contain sequences that facilitate internal ribosome entry (Oh et al., 1992; Boyd, 1993). These observations also raise the interesting possibility that crol expression may be regulated at the translational level.
The three crol mRNA isoforms have identical 5′ and 3′ sequences flanking a variable internal region that is generated by differential splicing (Fig. 6A). This sequence arrangement leads to the synthesis of three distinct CROL protein isoforms that have identical amino- and carboxy-terminal regions flanking a variable number of internal C2H2 zinc ﬁngers. Translation of the three CROL protein isoforms initiates with the same AUG triplet, which lies in a near ideal context for initiation. Conceptual translation of the CROL protein products is depicted in Fig. 7A. The longest isoform, CROLα, is 962 amino acids in length and contains 18 zinc ﬁngers while the other two isoforms, CROLβ and CROLγ, are 891 and 756 amino acids in length and contain 16 and 12 zinc ﬁngers, respectively. CROLβ is missing one and a half of the C- terminal zinc ﬁngers in CROLα, while CROLγ is missing the ﬁrst six zinc ﬁngers of CROLα (Fig. 7B).
crol transcription is induced by ecdysone during the onset of metamorphosis
The expression of β -galactosidase in crol4418 animals suggested that the gene marked by this enhancer trap insertion might be regulated by the steroid hormone ecdysone (Fig. 4). As an initial test of this hypothesis, we examined the temporal proﬁle of crol transcription in staged late third instar larvae and prepupae. The developmental northern blots shown in Fig. 5 were hybridized with a radiolabelled probe derived from exon 3 of the crol gene, sequences that are shared by all three mRNA isoforms. As expected, hybridization of the northern blot prepared from crol4418 mutants showed no detectable crol transcription (data not shown). In contrast, both the 6.0 and 5.3 kb crol mRNAs were detected in control animals, and their levels fluctuated in parallel with the changes in ecdysone titer (Fig. 8A). crol mRNA can be detected in mid-third instar larvae, consistent with the expression of β -galactosidase in the CNS of crol4418 animals at this stage in development (Fig. 4). The level of crol mRNA then increases in late third instar larvae, in parallel with the high titer ecdysone pulse that triggers puparium formation (Fig. 8A). The levels of crol transcription decrease to low levels in mid-prepupae and then rise signiﬁcantly in 12 hour prepupae, following the ecdysone pulse that triggers head eversion. This correspondence between the rises in ecdysone titer and the induction of crol transcription are consistent with crol being an ecdysone-inducible gene. To test this hypothesis more directly, salivary glands were dissected from mid-third instar larvae and cultured for 4 hours in the absence or presence of ecdysone. RNA was then isolated and crol transcription was analyzed by northern blot hybridization (Fig. 8B). This study revealed that crol mRNA levels are induced approximately two-fold by ecdysone, similar to the level of induction seen in vivo in late third instar larvae. A similar induction of crol transcription was seen in cultures of mixed larval organs treated with ecdysone (data not shown). These observations support the hypothesis that crol transcription is inducible by ecdysone, but the relatively low level of induction suggests that other factors may contribute to this regulation.
Drosophila metamorphosis represents a spectacular biological transformation in both form and function. Most organs and tissues that are necessary for larval growth and viability are destroyed during metamorphosis as new tissues and structures are assembled into the adult fly (Robertson, 1936; Bodenstein, 1965). An integral aspect of this remodeling is the external structures of the adult head and thorax that develop from imaginal discs. The discs are determined during embryogenesis and proliferate during larval development as they undergo pattern formation (Cohen, 1993). The patterned imaginal disc, however, is itself a static structure. Its growth, morphogenesis and terminal differentiation are dependent on pulses of ecdysone during metamorphosis (Fristrom and Fristrom, 1993; von Kalm et al., 1995). We would like to understand how ecdysone directs the imaginal discs to assume their appropriate morphological and functional properties during these ﬁnal stages of their development. To achieve this goal, we need a means of identifying ecdysone inducible regulatory genes that are required for appropriate disc morphogenesis and differentiation. The isolation and characterization of the crooked legs gene provides a ﬁrst step toward this goal. Below, we describe the crol mutant phenotypes and present a model for crol function during the early stages of metamorphosis.
crol is required for adult head eversion and leg morphogenesis during metamorphosis
Mutations in the crol locus lead to lethality during pupal development with defects in adult head eversion and leg development. Although no earlier lethality is evident in these mutants, crol mRNA is maternally provided and thus crol may have essential functions during early stages of development (data not shown). Experiments to examine crol mutant embryos derived from germline clones are currently underway.
Although at least part of the leg phenotype associated with crol mutations can be attributed to defects in leg elongation during early prepupal development, crol also appears to exert functions in the leg during later stages of development. One manifestation of this is a kink near the middle of the femur (Fig. 3B). The femur and tibia are initially fused at 18 hours after puparium formation and are divided into distinct segments between 21 and 36 hours after puparium formation (Fristrom and Fristrom, 1993). It is possible that the kink in crol mutant femurs is due to a defect in this morphogenetic process. In addition, some bristles are occasionally missing from crol mutant legs, indicating defects in the ﬁnal stages of leg differentiation (Fig. 3C).
One goal of our screen was to identify new regulatory genes that contribute to the complex developmental pathways controlled by ecdysone during metamorphosis. Consistent with this goal, the crol mutant phenotypes resemble those associated with some heteroallelic combinations of BR-C alleles and with a null E74B mutation. Legs dissected from br1/npr13BR-C mutant pupae are malformed and signiﬁcantly shorter than those of wild-type animals (Kiss et al., 1988), resembling those seen in crol mutant pupae. Furthermore, this phenotype can be attributed, at least in part, to defects in leg elongation during prepupal development (Fristrom et al., 1987).
The E74B lethal phenotypes also resemble those seen in crol mutant pupae. E74B mutants die during prepupal and early pupal development with defects in head eversion and leg elongation (Fletcher et al., 1995). These E74B phenotypes, however, are distinct from those seen in crol mutants. The E74B cryptocephalic phenotype is associated with a failure in gas bubble translocation from the middle of the prepupa to its anterior end. In wild-type animals, this process creates a space at the anterior end of the prepupa into which the adult head can later evert (Chadﬁeld and Sparrow, 1985). The gas bubble in crol mutant prepupae, however, appears to translocate properly. Similarly, the leg defects of E74B mutants are distinct from those associated with crol mutations. E74B mutant legs are smaller than wild type and are missing distal segments of the tarsus (Fletcher et al., 1995). These similar, but nonetheless distinguishable mutant phenotypes suggest that crol and E74B have unique and perhaps overlapping functions in adult head and leg development. These observations also support the hypothesis that these complex developmental responses involve interactions among a number of ecdysone-inducible regulatory genes and identify crol as a critical member of these developmental pathways.
crol encodes at least three related zinc ﬁnger proteins
Structural characterization of the crol locus revealed that this is a complex gene encoding at least three related protein isoforms. These proteins have identical N-terminal and C-terminal sequences flanking a variable central region that encodes tandem repeats of 12-18 C2H2 zinc ﬁngers. The presence of zinc ﬁngers within the CROL proteins suggests a molecular mechanism for crol function, through the direct regulation of gene expression. Furthermore, the N- and C-terminal sequences shared by the three CROL protein isoforms are rich in proline, alanine, glutamine, serine and threonine residues. Similar stretches of homopolymeric amino acids are associated with a number of transcription factors, further supporting a role for crol in gene regulation (Mitchell and Tjian, 1989). Consistent with this possibility, crol mutations result in defects in ecdysone-regulated gene expression during prepupal and early pupal development (Fig. 5). It remains unclear, however, at what level CROL exerts its regulatory functions since, unlike most transcription factors, at least some zinc ﬁnger proteins are also capable of binding RNA (El-Baradi and Pieler, 1991).
crol transcription is induced by ecdysone at the onset of metamorphosis
The pattern of lacZ staining in crol4418/+ larvae and prepupae combined with the temporal pattern of crol transcription during the onset of metamorphosis provide evidence in support of a role for ecdysone in inducing crol expression. Furthermore, crol mRNA levels increase approximately two-fold in salivary glands cultured in the presence of ecdysone, similar to the level of crol induction seen in vivo in late third instar larvae (Fig. 8B). Although these observations indicate that crol transcription is induced by ecdysone, they do not allow us to determine if this is a primary- or secondary-response to the hormone. An attempt to address this issue by using cycloheximide in cultured salivary glands was complicated by the ability of this drug to stabilize crol mRNA levels (data not shown).
The pattern of crol transcription in early pupae is consistent with crol induction as an ‘early-late’ response to ecdysone, similar to the response of the DHR3 and E78B orphan receptor genes (Koelle et al., 1992; Stone and Thummel, 1993). The levels of crol mRNA increase 12 hours after puparium formation, 2 hours after the initial induction of E74A and E75A transcription and thus 2 hours after the peak in ecdysone titer (Fig. 5). This delay in crol induction is similar to the delay seen in DHR3 and E78B induction relative to the late larval ecdysone pulse. This regulation has been attributed to a dual requirement for both the ecdysone-receptor complex and the synthesis of ecdysone-induced proteins (Ashburner and Richards, 1976; Stone and Thummel, 1993; Horner et al., 1995). Further studies of crol regulation should provide insights into its mode of regulation by ecdysone.
Models for crol function during prepupal and early pupal development
The changes in ecdysone-regulated gene expression seen in crol mutant prepupae provide a framework for understanding the molecular basis of the crol mutant phenotypes (Fig. 9). Four genes were examined that are expressed in imaginal discs and that appear to play a role in imaginal disc development: IMP-E1, Sb, Brg-P9 and EDG-84A. Of these genes, only BrgP9 was affected in crol mutants – Brg-P9 is expressed at higher levels, and for a longer duration, in crol mutant prepupae, suggesting that crol may normally repress this gene (Fig. 5). Interestingly, Brg-P9 encodes a protein with sequence similarity to the kunitz class of serine protease inhibitors (Emery, 1995). Several studies predict an important function for proteases in imaginal disc morphogenesis during prepupal development (Poodry and Schneiderman, 1971; Fekete et al., 1975; Pino-Heiss and Schubiger, 1989). In addition, mutations in the Stubble-stubbloid gene (Sb), which encodes an apparent transmembrane serine protease, interact with the BR-C to regulate appendage elongation (Beaton et al., 1988; Appel et al., 1993). Some heteroallelic combinations of Sb alleles lead to defects in leg elongation that are similar to those seen in crol mutants and proper Sb leg elongation can be restored by simply culturing the mutant leg discs in the presence of trypsin (Appel et al., 1993). It is possible that increased levels of Brg-P9 expression, and perhaps other serine protease inhibitors, could block the activity of Sb and other serine proteases in the leg discs, and thus prevent proper leg elongation during prepupal development (Fig. 9A). It is also likely that crol regulates other, as yet unidentiﬁed, target genes that function during leg elongation. The identiﬁcation of other secondary-response genes that are regulated by crol and expressed in leg imaginal discs should provide a better understanding of crol function in this tissue.
crol mutations lead to stage-speciﬁc effects on ecdysone-induced regulatory gene expression during the onset of metamorphosis. The levels of EcR and E74B transcription are reduced in 6-10 hour crol mutant prepupae, and the BR-C, E74A, E75A, E75B and E93 early genes are submaximally induced in response to the ecdysone pulse in 10 hour prepupae (Fig. 5). These effects on gene expression provide a molecular basis for understanding the defects in adult head eversion seen in crol mutants (Fig. 9B). As mentioned above, E74B has been shown to be required for head eversion, although the mechanism(s) by which E74B regulates this response remain unknown (Fletcher et al., 1995). The reduced expression of E74B in crol mutant prepupae thus provides one means of interpreting the effect of crol mutations on adult head development (Fig. 9B). Alternatively, reduced levels of EcR expression in crol mutants could attenuate early gene induction by ecdysone and thereby indirectly affect head eversion (Fig. 9B). Finally, it is possible that crol directly regulates early gene expression in prepupae (Fig. 9B). In this regard, it is interesting to note that preliminary studies have shown that crol is expressed normally in BR-C and E74 mutants, conﬁrming that crol functions either upstream from, or in parallel with, these regulatory genes (data not shown). Further molecular and genetic studies of the crol locus should provide insights into how this gene can exert its multiple effects on the ecdysone regulatory hierarchies during metamorphosis.
A genetic screen for regulators of adult tissue development during metamorphosis
The screening strategy used to identify the crol locus provides a new avenue for studies of ecdysone function, by providing a genetic approach to the identiﬁcation of regulatory genes that function during insect metamorphosis. All known regulatory genes in the ecdysone hierarchies have been isolated based on their cytogenetic location or membership in the nuclear receptor superfamily (Russell and Ashburner, 1996; Thummel, 1996). crol represents the ﬁrst gene isolated independently of these criteria, based solely on its mutant phenotypes. In fact, crol would not have been identiﬁed based on its cytogenetic location since it does not map to an ecdysone-inducible puff in the salivary gland polytene chromosomes (Ashburner, 1972). Given the successful identiﬁcation of this gene, we have expanded our screen to a collection of 1300 single P-elementinduced lethal mutations on the second chromosome. From this screen, we have identiﬁed a dozen genes that appear to be regulated by ecdysone and play an essential role in imaginal disc development during metamorphosis (J. Gates and C. S. T., unpublished results). It seems likely that this approach will provide new opportunities to unravel the molecular mechanisms of ecdysone action during Drosophila development. In addition, molecular and genetic characterization of these loci should provide new insights into the mechanisms by which ecdysone triggers the remarkable transformation from a patterned imaginal disc to a differentiated adult structure.
We are grateful to S. Shaver, A. Hilliker, J. Fristrom, G. Guild, A. Hammonds, T. Laverty, K. Matthews, M. Noll, G. M. Rubin, and A. Spradling for fly stocks and reagents. We also thank P. Reid for excellent assistance with DNA sequencing, and J. Broadus, A. Letsou, and J. Schmidt for comments on the manuscript. We are grateful to members of the Thummel laboratory, past and present, for valuable discussions during the course of this work. P. P. D. is a Postdoctoral Research Associate and C. S. T. an Associate Investigator with the Howard Hughes Medical Institute.