The Drosophila disembodied gene controls late embryonic morphogenesis and codes for a cytochrome P450 enzyme that regulates embryonic ecdysone levels

The steroid hormone 20-hydroxyecdysone (20E) coordinates many aspects of Drosophila development (reviewed by Riddiford, 1993). Some of the best-characterized responses to ecdysteroids are those that occur at each major postembryonic developmental transition. Molting for example, is triggered by pulses of ecdysone released from the prothoracic gland in response to a neuropeptide signal from the brain. Other responses influenced by ecdysone during the larval-prepupal and pupal transitions include eversion of imaginal discs and deposition of pupal cuticle (reviewed by Fristrom and Fristrom, 1993), movement of the morphogenetic furrow in the eye (Brennan et al., 1998), proliferation of histoblasts and cells of the imaginal rings (Bender et al., 1997), remodeling of the central nervous sytem (CNS) and apoptosis of most larval cells (Bender et al., 1997; Truman, 1996; Schubiger et al., 1998). Ecdysteroids have also been implicated in regulating oogenesis (Buszczak et al., 1999) and in patterning of the embryonic cuticle (Bender et al., 1997).

Much is now known about the mechanisms by which ecdysteroids control developmental events. In Drosophila, response to these hormones is mediated through the products of the EcR and usp genes (reviewed by Thummel, 1995). The EcR gene produces three isoforms, which are members of the nuclear receptor family (Talbot et al., 1993). Each of these isoforms can form heterodimers with Usp, a Drosophila RXR receptor, and mediate specific developmental responses by binding to DNA elements located in the regulatory sequences of target genes (Yao et al., 1992; Horner et al., 1995). As first proposed by Ashburner et al. (1974), the coordination of these developmental responses is controlled by a hierarchical system in which many of the primary response genes, such as E74, E75 and products of the Broad Complex, code for transcription factors, which in turn, implement the appropriate genetic programs of each target tissue (reviewed by Thummel, 1995).

Although the genetic hierarchy that controls responses to 20E has received considerable attention, little is presently known about the genes involved in ecdysteroid biosynthesis and possible regulatory mechanisms that control hormone production. In arthropods, ecdysteroids are generally synthesized from dietary cholesterol or phytosteroids via a series of hydroxylation steps (reviewed by Grieneisen, 1994). Most of these hydroxylations have been reported to occur in the insect prothoracic gland or its crustacean equivalent, the Y organ. In addition, the ovaries are a major source of ecdysteroid synthesis in several insects (Hoffmann et al., 1986). Almost all of the enzymes involved in these hydroxylations exhibit biochemical properties consistent with classical P450 cytochrome enzymes and are associated with either microsomes or mitochondria (Grieneisen et al., 1993).

Although several Drosophila mutations including ecdysoneless (l(3)ecd1), l(3)DTS3, dre4 and giant have been shown to affect ecdysone titers, the mechanisms responsible are not entirely clear (Garen et al., 1977; Walker et al., 1987; Sliter and Gilbert, 1992; Schwartz et al., 1984). The giant locus encodes a b-ZIP transcription factor and so may indirectly regulate production of ecdysteroids at certain developmental stages or in certain tissues through transcriptional control. In the case of l(3)ecd1, recent biochemical tracer data suggest that this mutation does not involve a block in any of the enzymes involved in ecdysone production. Instead, the defect appears to be in a gene that governs translocation of sterol intermediate(s) between subcellular compartments (Warren et al., 1996). Neither molecular nor biochemical data are presently available concerning the functions of the l(3)DTS3 and dre4 genes.

The only locus identified to date that is likely to affect ecdysteroid biosynthesis directly is the recently described Drosophila dare gene (Freeman et al., 1999). dare encodes a homolog of human adrenodoxin reductase (AR). In mammals, AR catalyzes electron transport from NADPH to adrenodoxin, which in turn can donate a pair of electrons to all known mitochondrial P450s (Miller, 1988). Mutations in dare cause abnormal behavioral responses to olfactory stimuli, blocks in molting and oogenesis, and degeneration of the adult nervous system (Freeman et al., 1999; Buszczak et al., 1999). Since the molting defects could be partially rescued by feeding mutant larvae 20-hydroxyecdysone, at least some of the phenotypes associated with this mutation likely result from reduced synthesis of ecdysteroids.

One phenotype predicted for mutations that disrupt function of the biosynthetic pathway for embryonic ecdysteroids would be inability to produce a cuticle during the second half of embryogenesis. disembodied (dib) is one of a set of mutations that result in a poorly differentiated embryonic cuticle (Jürgens et al., 1984; Nüsslein-Volhard et al., 1984; Wieschaus et al., 1984). In this report, we describe the cloning, genetic characterization and physiological defects associated with mutations in the dib gene, which encodes a cytochrome P450-like enzyme. Mutations in dib prevent dorsal closure, and interfere with embryonic cuticle deposition, midgut morphogenesis and head involution. Consistent with a potential role in ecdysone biosynthesis, RNA in situ hybridization reveals that two major sites of dib expression are the ring gland and the adult ovaries. Using enzyme immunoassay, we show that the titers of ecdysone and 20E are greatly reduced in dib homozygous mutant animals compared to sibling controls. Consistent with this observation, we also find that induction of IMP-E1 and L1, two ecdysteroid inducible genes, is blocked in certain tissues of dib mutant embryos. Finally, we investigate ecdysteroid levels in several other mutants that exhibit poorly differentiated cuticle and show that one of these, spook (spo), exhibits a range of phenotypic defects similar to dib, shows low ecdysteroid titers and inhibited IMP-E1 induction. These studies suggest several new roles for ecdysteroids during Drosophila embryonic development and identify at least two of the genes involved in the biosynthesis of these hormones.

Drosophila strains were cultured on standard cornmeal/yeast extract/dextrose medium. Wild type is considered y,w or cn; ry. dib, phm, sro, hau, gho mmy and spo mutants were obtained from the Bloomington stock center and are described in Wieschaus et al. (1984), Nüsslein-Volhard et al. (1984) and Jurgens et al. (1984). dib alleles where obtained from Steve Harrison and are described by Harrison et al. (1995). The zip mutant was obtained from Jim Fristrom (UC Berkeley).

For genomic rescue constructs, lambda phage inserts were excised from the lambda II fix vector (Stratagene) by digestion with NotI. The insert was then ligated into the NotI site of pCaSpeRhs70. The rescue construct pMBO1901 was constructed by ligating an approx. 7.0 kb SalI-HindIII (filled in) fragment from lambda SE20 (Fig. 2) into XhoI-XbaI (filled in) of pCasPeR4. All germline transformants were obtained using standard protocols. Injection was into a y,w stock. For rescue experiments, strains carrying a transgene on the second chromosome were used. Males and virgin females that had a single copy of the transgene, together with any of the dib^F8/TM6B, dib^13A/TM6B or dib^P3/TM6B dib genotype, were crossed. Progeny from these crosses were scored for the survival of dib/dib homozygous adults at 25°C. Observed survival was 60-112% of expected, depending on the transgenic line used, versus 0% survival in the absence of the transgene.

A 1.7 kb dib cDNA clone was isolated by screening an embryonic cDNA library (a gift from C. Thummel) with a 1.2 kb BamHI genomic fragment from phage SE20 (Fig. 2). Both strands of the cDNA were sequenced by primer walking. To identify the changes in dib mutants, genomic DNA of heterozygous mutants was isolated from adult flies and the dib gene of the different alleles, together with the balancer allele, were amplified by PCR. Genomic DNA from each DNA preparation was then sequenced using the Thermosequenase cycle sequencing kit (USB Inc.) according to the manufacturer’s instructions. Mutations were identified by the appearance of two bases (one from the mutant and one from the balancer) at a given position in the sequence.

Embryos were fixed for 20 minutes in a 1:2 mixture of PBS:heptane that contained 4% formaldehyde. Devitellinizing was accomplished by washing the embryos several times in methanol. Imaginal disks and ovaries were dissected into phosphate-buffered saline (PBS), fixed in 4% formaldehyde and stored in ethanol at −20°C until use. Sense and antisense digoxigenin-labeled RNA probes (Boehringer Mannheim) were generated by transcription of pBluescript dib, IMP-E1 or L1 subclones using T3, T7 or the Sp6 promoter. Embryos collected from heterozygotes containing a balancer chromosome marker with either Ubx-lacZ (for third chromosome mutants), or wg-lacZ (for second chromosome mutants), were double stained with anti-β-gal antibody and RNA probe in order to distinguish the homozygous mutants from those that carried the lacZ balancer chromosome. RNA hybridization and detection was carried out using standard methods. The rabbit anti-Spectrin antibody (a gift from T. Hays) was used at 1/100 dilution and the mouse β-gal monoclonal antibody (Promega) was used at 1/1000 dilution. Staining for both antibodies was visualized using an HRP-coupled secondary antibody (Vector laboratories) and diaminobenzidine as substrate. Embryos, disks and ovaries were mounted for photography in 10% PBS and 90% glycerol.

7- to 12-hour (Table 1) or 7- to 16-hour (Table 2) embryos were collected, dechorionated and frozen at −20°C until used. Embryos were homogenized in 0.5 ml of methanol and incubated at 4°C for 2 hours. The supernatant was saved and the embryos were again incubated in 0.5 ml methanol overnight. Both supernatants were pooled and dried under N^2. Ecdysteroids were analyzed with an enzyme immunoassay modified from that of Porcheron et al. (1989), using a 20-hydroxyecdysone-peroxidase conjugate as a tracer and either DBL2-polyclonal anti-ecdysone antiserum or EC19 monoclonal anti-20-hydroxyecdysone-specific antibodies (for more details see Aribi et al., 1997; Pascual et al., 1995). Briefly, dried samples were solubilized in a phosphate buffer (0.1 M, pH 7.4), placed on microplate wells (previously coated with secondary anti-IgG antibodies), then a known amount of tracer was added along with the primary anti-ecdysteroid antibody. After a 3-hour incubation at room temperature, bound peroxidase activity was revealed using tetramethylbenzidine as substrate and microplates were analyzed using a SpectraMax 340 microplate reader (Molecular Devices). Results were compared with data obtained from reference concentrations of ecdysone or 20-hydroxyecdysone and were normalized to the weight of the samples.

Homozygous mutant germline clones were generated in the background of a dominant female-sterile mutant ovo^D. Virgin females w, hs-Flp/w: FRT79, dib^F8/TM3 were crossed to males FRT79, ovo^Dw⁺/TM3. The first instar larvae of this cross were heat shocked at 37°C for 90 minutes and the non TM3 females were crossed to males dib^P3/TM3, ftz-lacZ. Eggs were stained with anti-β-gal antibody in order to distinguish the homozygous dib germline clone embryos from the heterozygous embryos.

The disembodied mutation was first identified in screens for lethal genes that affect embryonic cuticular patterning (Jürgens et al., 1984). As shown in Fig. 1, mutations in dib prevent full differentiation of embryonic cuticle (Fig. 1A compare with E and I). Specifically, no denticle belts, dorsal hairs or differentiated mouth parts are evident in dib homozygotes. Only a thin cuticular remnant containing both dorsal and anterior holes is secreted by most mutant embryos.

In an effort to determine when and where dib mutants first show developmental defects, we examined the morphology of dib mutant embryos at different developmental stages. Homozygous dib mutant embryos derived from dib^F8/TM2, Ubx-lacZ × dib^P3/TM2, Ubx-lacZ parents were unambiguously identified by the absence of staining for Ubx-lacZ. As shown below, these dib alleles are the result of stop codons and therefore likely represent null alleles. We find that dib embryos appear to develop normally to approximately stage 14 (Fig. 1B compare with F). At this point, many morphogenetic movements including dorsal closure, head involution, midgut morphogenesis and hindgut looping become abnormal and staging of the embryos becomes difficult. Midgut cells form the bilateral vessel but often fail to migrate to enclose the gut and the normal constrictions are either not formed or form abnormally (Fig. 1J). The aberration is not due to a failure to express homeotic genes such as Ubx and Antp in the developing visceral mesoderm (data not shown). At later times, the failure of the midgut to enclose the yolk leads to yolk granules being distributed throughout the embryo and they often extrude through the dorsal opening. During this time, the epidermis also begins to contract along the anteroposterior axis leaving the thoracic and abdominal segments in a highly compacted state possibly because head segments fail to involute properly (Fig. 1C and D compare with G and H). The hindgut also becomes much more extended than normal or may show unusual coiling (Fig. 1G). Many embryos never complete dorsal closure (Fig. 1G) and, when development terminates, the embryos are highly contracted with non-involuted head segments protruding anteriorly and gut and yolk protruding posteriorly (Fig. 1L). We also stained embryos with Acridine orange, DAPI and reaper to examine whether there was an increase in apoptosis during these late stages. We did not observe any unusual DAPI staining, or an increase in AO-positive cells or reaper-expressing cells suggesting that abnormal numbers of cells were not dying during this time.

To determine if the phenotype was altered by removal of maternal components, we also generated germline clones. We found that embryos derived from mutant germlines exhibited exactly the same range and timing of defects (Fig. 1I,K) as did simple zygotic mutants, suggesting that dib is not required in the germline during oogenesis or during early embryogenesis before stage 14.

In an effort to understand the molecular basis for the dib mutant phenotype, we cloned the locus and characterized its product. The dib locus was previously localized to 64A3 on the left arm of the third chromosome by deficiency mapping (Harrison et al., 1995). During a low-stringency screen for new BMP-type II receptors (G. M. and M. B. O’C., unpublished data), a new receptor was found to be localized to the 64A3 interval and an approx. 30 kb phage walk was established in the region. Several cDNAs were isolated and positioned within the walk by Southern hybridization analysis and sequencing (Fig. 2). To begin to associate particular transcripts with mutant genes, we used P-element-mediated transformation to generate transgenic lines carrying various overlapping genomic DNA fragments. We found that P elements carrying DNA from either phage SE9 or SE20 (Fig. 2) were able to rescue the l(3)64Ak complementation group. This complementation group has been previously shown to be allelic to dib (Harrison et al., 1995). A single transcription unit defined by a 1.7 kb cDNA was localized to the DNA contained on both these phages. To further test whether this transcription unit might correspond to dib, we generated a smaller genomic rescue construct (H-S fragment in Fig. 2) containing this transcription unit together with approximately 4 kb of 5′ sequences. This transgene was found to rescue different allelic combinations of dib mutants to 100% viability and fertility. Additional proof that this transcript codes for the Dib product was obtained by sequencing several mutant alleles (see below).

As a first step in determining the possible biochemical function of Dib, we sequenced the 1.7 kb cDNA. As shown in Fig. 3, conceptual translation of the Dib product revealed that it is a new member of the cytochrome P450 superfamily. Based upon consultation with the cytochrome P450 Nomenclature Committee (http://drnelson.utmem.edu/cytochromeP450.html), this sequence has been assigned the name CYP302A1 and is the first member of a new subfamily since it is <40% identical to any of the other cytochrome P450 gene presently in the database. The most closely related genes are the Drosophila acanthoptera CYP12B1, which shows 28% identity, and members of the CYP24 family (23% identity), which encode 1,25-dihydroxyvitamin D₃ 24-hydroxylase, a mitochondrial enzyme involved in vitamin D₃ inactivation (Chen et al., 1993; see also http://drnelson.utmem.edu/nelsonhomepage.html for dendograms showing relationship of Dib to other P450s). All eukaryotic cytochrome P450 enzymes are membrane-associated proteins that are found in either the endoplasmic reticulum or the mitochondria. Overall, the Dib sequence is most similar to the mitochondrial P450 class.

As further confirmation that this enzyme is in fact the product of the dib gene, we sequenced genomic DNA obtained from lines carrying three different dib alleles. As shown in Fig. 3, dib^13A, dib^P3 and dib^F8 all resulted from the introduction of stop codons. Both dib^F8 and dib^P3produce truncated proteins that lack the heme binding site suggesting that they are likely null alleles. The dib^13A allele is a stop codon that removes only the last five amino acids. Not surprisingly, this mutation was found to be hypomorphic, and when crossed to either dib^F8 or dib^P3, produced a low level (<1%) of escaper flies that appear normal and are fertile. Embryos produced by crossing flies containing this allele to either of the other two mutant lines also showed less-severe embryonic defects (data not shown).

The lack of cuticle deposition in dib mutants and the interference with many morphogenetic movements after stage 14 suggested a strong requirement for the dib product during late stages of embryogenesis. To learn more about which tissues might require dib function, we determined the temporal and spatial expression of dib throughout development by in situ hybridization. In the zygote, dib expression is first seen at the early syncytial blastoderm stage where it is enriched in the anterior half of the embryo in a hunchback-like pattern (Fig. 4A). As cellularization proceeds, additional posterior stripes rapidly appear and by the time the germband is extended, most of the epidermal cells express dib (Fig. 4B-D) The expression then fades during further germband elongation and, by about stage 12, only a small number of unidentified cells scattered in each segment express dib (Fig. 4F-H). During germband retraction, dib expression fades from these cells and, during dorsal closure, precursors of the ring gland begin to express dib in the prothoracic component (Fig. 4I,J). Expression remains strong and exclusively in the prothoracic component of ring gland until cuticle deposition at stage 17, at which time probe accessibility becomes limited. We also examined expression in wandering third instar larvae and white prepupae. In late third instar larvae, expression is again found exclusively in the prothoracic component of the ring gland (Fig. 4K). Its expression then appears to recede in white prepupae that were stained for an equivalent length of time (Fig. 4L). In adult females, dib is expressed in the follicle cells of the ovary at about stage 8 of oogenesis (Fig. 4M). Little if any expression is seen in the nurse cells and, by stage 11, expression in the follicle cells is greatly diminished (data not shown).

The finding that Dib is most closely related to vertebrate steroidogenic P450 enzymes suggested that it might function in either the synthesis or breakdown of ecdysteroids, the primary hormones that signal and control major developmental transitions in insects. The observation that dib is expressed in both the ring gland and the ovaries, two tissues that are known to be major sites of ecdysteroid biosynthesis (reviewed by Riddiford 1993), is also consistent with this view. As a first test of whether dib mutants negatively influence ecdysteroid titers, we examined the expression of the 20E-inducible genes IMP-E1 and L1 in dib mutant embryos. IMP-E1 and L1 were isolated in a screen for imaginal disc, membrane-bound polysomal RNAs that show upregulation in response to 20E (Natzle et al., 1986). These genes encode novel secreted proteins that respond either directly (IMP-E1) (i.e. inducible in the presence of cycloheximide), or indirectly (IMP-L1), to 20E (Natzle et al., 1988, 1992) in culture. Recently both genes have also been found to be expressed in complex patterns in the embryo (J. E. N., unpublished data) including the epidermis (Fig. 5A,B,D). Since the timing of the epidermal expression coincides with the rise in embryonic ecdysteroid titers, we examined whether these patterns are altered in a dib mutant background. As shown in Fig. 5C,F, the epidermal expression of both IMP-E1 and IMP-L1 is greatly reduced or absent consistent with the notion that dib mutants may be defective in ecdysteroid production.

Previous determinations of ecdysteroid titers during Drosophila embryogenesis have shown that there is a peak in the production of these hormones between 7 and 12 hours of development (Maróy et al., 1988), which is at about the time that dib mutants first begin to show developmental defects. To determine directly whether dib mutant embryos show reduced levels in ecdysone and 20-HE titers, we quantified the levels of these two ecdysteroids by EIA. 7- to 16-hour old embryos were collected from parents heterozygous for dib^F8 and a TM3 balancer marked with a arm-GFP reporter. Embryos were sorted by fluorescence into GFP-negative pools containing the dib homozygous mutant embryos and a GFP-positive pool corresponding to dib/ TM2, arm-GFP heterozygotes and TM2, arm-GFP homozygotes. Two antibodies, DBL-2 and EC 19, which specifically recognize E and 20E, respectively (Aribi et al., 1997), were used to quantify the levels of these two ecdysteroids. As shown in Table 1, homozygous dib mutants contained very low concentrations of both E and 20E, compared to control embryos. The residual low levels of these products seen in the homozygous mutants could correspond to either remaining maternal products, to other metabolites weakly recognized by the antibodies, or to low-level synthesis of each product by another less-efficient pathway. These results strongly implicate dib in catalyzing a step in the production of ecdysteroids.

Superficially, the phenotype of dib mutants resembles that produced by mutations in several other genes isolated in embryonic lethal screens (Jürgens et al., 1984; Nüsslein-Volhard et al., 1984). These include spook, haunted, mummy, shroud, ghost and phantom. We examined the IMP-E1 expression pattern in embryos mutant for each of these genes in order to determine whether these mutants, like dib, might have an imbalance in ecdysone metabolism. We found that hau and mum both showed relatively normal levels of IMP-E1 expression while gho was somewhat reduced (Fig. 6B,C and data not shown). This observation suggests that the defects in cuticle differentiation in these mutant backgrounds likely lie downstream of ecdysone and 20E biosynthesis and do not involve the general ecdysone response transcription machinery. In contrast, no or very little staining was found in phm, spo and sro mutants (Fig. 6D-F), consistent with the possibility that these genes may also affect ecdysteroid titers. To examine this directly, we also determined the ecdysone and 20E titers in five of these mutant backgrounds. As shown in Table 2, spo homozygous mutants have very low levels of these two ecdysteroids suggesting that, like dib, this mutation also affects ecdysteroid biosynthesis. Interestingly, shroud homozygous mutants have relatively normal levels of ecdysteroids despite the fact that mutants show no epidermal IMP-E1 expression. Like dib, all three of these mutants (sro, phm and spo) also fail to make cuticle and show defects during late embryogenesis in morphogenetic movements that lead to a failure of head involution (Fig. 6E) and defective midgut formation (Fig. 6G,H).

At present, all species of arthropods examined appear to be able to synthesize ecdysteroids from smaller organic precursors. The primary precursor in Drosophila is probably cholesterol, which is obtained through dietary uptake. A general scheme for the production of 20E from cholesterol in the ovaries and ring glands consisting of a series of oxidation, reduction and hydroxylation steps has been proposed (reviewed by Grieneisen 1994; Warren et al., 1996). In a number of different insects, it appears that at least five of these steps are catalyzed by enzymes with properties unique to P450-type proteins (Grieneisen, 1994). In this report, we have shown that the Drosophila Dib product is a new member of the P450 cytochrome gene family. Since Dib shows <40% identity to other known P450s, it has been assigned to a new class. Nonetheless, it shows a high degree of similarity to vitamin D3 24-hydroxylase, a vertebrate P450 involved in steroid metabolism. This finding, together with the fact that ecdysone and 20E levels are extremely low in dib homozygous mutant embryos and the observation that dib message is highly expressed in the prothoracic gland, a known site of ecdysone biosynthesis, leads us to conclude that the Dib protein is involved in biosynthesis of ecdysteroids.

With the data presently available, we are not able to prove the exact step at which Dib functions. However, based on a number of observations, we can offer some arguments that limit the possibilities. First, we believe that Dib is not likely to catalyze the last step involving 20-hydroxylation of ecdysone, since both 20E and ecdysone are low in dib mutants. In addition, another novel cytochrome P450 isolated from Locusta migratoria has been proposed to be the 20-hydroxylase (Winter et al., 1999), and this enzyme shows only 20% identity at the amino acid level to Dib.

In addition to the 20-hydroxylase, at least four other steps in the production of ecdysone in locusts and hornworms appear to utilize P450-type enzymes. These include an early 7,8-dehydrogenating enzyme as well as the enzymes involved in the later terminal hydroxylations at C-25, C-22 and C-2 (Kappler et al., 1986, 1988). It is likely that Dib could catalyze one of these steps in Drosophila. Interestingly, all of these enzymes are also found in the prothoracic glands in these insects; however, they show differences in their subcellular localization. The 7,8-dehydrogenating enzyme and the C-25 hydroxylase are both found in the microsomal fraction, while the C-22 and C-2 hydroxylases are located in mitochondria (Kappler et al., 1988). In vertebrates, microsomal P450s all have an N-terminal hydrophobic segment that serves as both a signal and a stop insertion sequence (Omura and Ito, 1991). The result is that these enzymes are inserted with their N terminus within the luminal side of the ER membrane and their C-terminal catalytic domain facing the cytoplasm. Examination of the Dib N-terminal region does not reveal a typical hydrophobic signal sequence. In contrast to microsomal enzymes, most mitochondrial P450s contain an amphipathic N terminus of 25 to 39 residues that aids in localization of these proteins to the inner mitochondrial membrane (Omura and Ito, 1991). Examination of the Dib N-terminal sequence for this type of trait shows that it contains five basic and five hydrophilic residues within the first 30 amino acids, which is more consistent with it being a mitochondria-type P450 rather than a microsomal type. This view is also consistent with its overall sequence similarity to the vitamin D3 24-hydroxylase, which is a mitochondrial enzyme. If Dib is indeed a mitochondrial enzyme then this location would be consistent with it catalyzing either the C-22 or C-2 hydroxylations.

Another observation relevant to identifying the possible enzymatic step at which Dib might function is the finding that, in vitro, incubation of Drosophila ovaries with radiolabelled sterol substrates results in termination of ecdysteroid biosynthesis at the 2dE step prior to C2 function (Warren et al., 1996). Since Dib is expressed in the follicle cells of ovaries, this may indicate that Dib functions prior to this step. Taken together with Dib’s potential mitochondrial location, these observations are most consistent with Dib being the C-22 hydroxylation enzyme.

In Drosophila, ecdysteroid pulses are known to play key roles in regulating larval molts, the transition between larval and the prepupal stage, and differentiation of adult tissues during pupation (Riddiford, 1993). What role, if any, ecdysteroid levels play during embryonic development has not been clear, in part due to a lack of mutations that directly affect ecdysteroid synthesis during these early stages. Our results demonstrate that dib and spo mutants greatly reduce ecdysone and 20E levels during embryogenesis, and strongly suggest that these steroids are critical for allowing proper embryonic development to proceed beyond stage 14 at about 9 hours after fertilization. The timing of this requirement fits well with the known peak of ecdysteroid production in embryos, which occurs between 7 and 12 hours postfertilization (Richards, 1981; Maróy et al., 1988).

Since P450s catalyze a wide variety of oxidative, peroxidative and reductive steps during the metabolism of a diverse range of compounds including fatty acids, it remains possible that some other steroid or non-steroid metabolite produced by Dib might contribute to the observed phenotypes. Nonetheless, we believe that the phenotypes that we observe are completely consistent with the previously described roles of ecdysteroids in modulating morphogenesis and cellular differentiation during later larval and pupal stages of development. For example, one of the major processes regulated by ecdysteroids during larval stages is the formation and subsequent sclerotization of new cuticle at the larval molts (Kaznowski et al., 1985; Fristrom and Liebrich, 1986). Since the embryonic ecdysteroid peak just precedes the onset of first instar cuticle production at about 12 hours (Hillman and Lesnick, 1990), it is not surprising to find that elimination of this peak by mutations in dib and spo abolishes formation of this cuticle. New cuticle synthesis during larval and pupal stages actually requires a drop in ecdysteroid levels (Fristrom et al., 1982; Doctor et al., 1985). Thus, the larval pulses appear to condition cells such that they become permissive for expression of cuticle-producing genes during the downturn in ecdysteroid levels. The timing of the events in the embryo appears to be consistent with a similar conditioning phase. The conditioning phase would presumably set in motion a genetic program that leads to terminal differentiation of epidermal cells and subsequent production of the cuticle biosynthetic machinery. This process may require a 20E-dependent transcription factor activation cascade similar to what has been described for puffing of salivary gland chromosomes and ecdysteroid-stimulated transitions during pupal development (Ashburner et al., 1974; Thummel, 1995).

It is interesting to speculate that the shroud gene might encode an early transcription factor that implements the proper hormonal response at this stage of development. While shroud does not appear to affect ecdysteroid levels directly, it does affect inducibilty of downstream targets, such as IMP-E1, and results in phenotypes very similar to dib and spo. Alternatively, shroud could also code for the C2 enzyme as the DBL-2 and EC19 antibodies used in the E1A are not well suited to detect changes in ecdysteroid titers resulting from a defect in C2 hydroxylation. It should also be pointed out that, while spo mutants do show low levels of ecdysone and 20E, the gene need not encode a catalytic enzyme involved directly in ecdysteriod biosynthesis. Instead, as has been suggested for the ecd locus (Warren et al., 1996), spo might encode a transporter that is involved in shuttling earlier intermediates between the two subcellular compartments in which the biosynthetic machinery resides.

Another issue to consider with regard to embryonic ecdysone biosynthesis is the location and identity of the cells involved in producing the pulse of embryonic ecdysteroids. During larval stages, the main source of ecdysteroids is the prothoracic gland. Indeed, we see high level expression of dib in both the larval prothoracic gland as well as the embryonic prothoracic gland. It seems unlikely, however, that the embryonic prothoracic gland is involved in producing the embryonic pulse of ecdysone since dib expression is not observed in this gland until stage 16-17 at about 16 hours of development. This is well after the embryonic peak in ecdysone levels is reached and is also well after the onset of the morphological defects in dib mutants. If we use dib expression as a guide, it is possible that the epidermal cells, or the unidentified segmentally repeated cells that express high levels of dib during stages 11-12 (Fig. 4), may be the source of embryonic ecdysteroid synthesis. The identity of the segmentally repeated cells remains unresolved and it will be interesting to see if other genes involved in embryonic ecdysteroid production, possibly members of the spo locus, are also expressed in these cells. The expression of dib at the blastoderm stage is also noteworthy since it shows a distinct gap-gene-like pattern and mutations in at least one gap gene, giant, have been reported to show low levels of ecdysteroids (Schwartz et al., 1984). Additional characterization of the transcriptional control regions of dib and spo may provide insight into the important question of what initiates the production of the embryonic ecdysone pulse.

Work from a number of laboratories has raised the question of the role of maternally deposited ecdysteroids during embryonic development. In many insects, maternally synthesized ecdysteroids are stored in the oocyte in the form of polar (e.g. phosphorylated derivatives) or apolar (e.g. ecdysteroid fatty acid esters) conjugates of ecdysone, 20E and 2dE (Bownes et al., 1988; Lagueux et al., 1981; Dinan, 1997; Tawfik et al., 1999). In Locusta (Laguex et al., 1981; Dinan, 1997) and Drosophila (Bowes et al., 1988), these conjugates are apparently bound to the yolk protein vitellin. However, it is not clear whether these conjugates serve as a storage reservoir for release of ecdysteroids during embryogenesis or as an inactivation mechanism for maternal ecdysteroids from the ovary. Since most maternal conjugates are formed from ecdysteroids after the point in the biosynthetic pathway where Dib is postulated to act, it is not likely that Dib function would be necessary for later conversion of the conjugates into active hormone. De novo synthesis of ecdysteroids in the early embryo could supplement these potential maternal stores. However, the clear effect of zygotic dib mutations on embryonic ecdysteroid levels suggests that a pathway that utilizes the maternal ecdysteroids cannot substitute for the absence of the zygotic dib product.

One aspect of the phenotype that also deserves comment is the contrast between the dib and spo zygotic loss-of-function phenotypes and that seen with mutations in other genes that have been implicated in either 20E production or in mediating its response. Mutations in the recently described dare locus (Freeman et al., 1999), for example, should eliminate embryonic 20E production since AR is the only known source of reducing equivalents for mitochondrial P450s. Yet these mutants arrest during larval stages and show no embryonic defects. In contrast, null mutations in the ecdysteroid receptor components EcR and Usp and the 20E inducible orphan nuclear receptor DHR3 are embryonic lethal. However, the phenotypes of these mutants are milder than those seen in dib or spo mutants. EcR and usp mutants, for example, show cuticle defects but these consist primarily of a reduction in denticle number and size, along with occasional aberrations in mouth-part formation in the case of EcR mutants (Bender et al., 1997), and minor abnormalities in abdominal segments in some usp alleles (Oro et al., 1992), rather than a total lack of cuticle as found in dib and spo mutants. Depending on the timing of the temperature shift, the temperature-sensitive mutants ecd1 and DTS-3 have also been shown to affect egg viability, but the phenotypic effects on embryogenesis have not been described (Audit-Lamour and Busson, 1981; Walker et al., 1987).

One likely explanation for some of the phenotypic differences observed between these various mutants is that the encoded products may be either differentially loaded into the egg during oogenesis or exhibit different stabilities in the zygote. Consistent with the former view is the recent demonstration that EcR and dare transcripts are both loaded into the egg at stage 10 (Buszczak et al., 1999) while we find that dib is not very highly transcribed in the nurse cells and may be only weakly, if at all, provided maternally to the egg. It is also possible that some ecdysteroid responses may not need certain components at specific stages or in certain tissues. Usp, for instance, is not required for certain ecdysteroid-mediated responses during mid-third instar development (Hall and Thummel, 1998). Such differential requirements could also account for some of the phenotype variation observed between mutants in the different ecdysteroid synthesizing and utilization components.

Perhaps the most novel aspect of the phenotype associated with dib and spo mutants is the dramatic defects seen in morphogenetic movements after stage 12. These include failures of head involution and dorsal closure as well as defects in gut morphogenesis. With respect to the last observation, it is interesting to note that it has been previously reported that deficiencies that remove E75, a 20E-regulated gene that encodes an orphan nuclear receptor, also result in alterations in the spatial positioning of midgut constrictions as do dib and spo mutants (Bilder and Scott, 1995). As in the case of the cuticle defects described above, these effects on embryonic morphogenesis are also consistent with the types of defects that have been associated with ecdysone deficiencies at later stages of development. For example, recently it has been shown that an ecdysone signal is required for progress of the morphogenetic furrow across the eye (Brennan et al., 1998). Likewise, remodeling of larval neuronal connectivity during pupal stages requires the ecdysteroid receptor (Schubiger et al., 1998) and a hormonal pulse is known to be required for evagination of imaginal discs at puparium formation (Fristrom et al., 1982). One common aspect of all these processes is that they require cell shape changes and/or cell movement. Thus, one likely general molecular function for 20E is to stimulate tissue remodeling by coordinating changes in either the actin-myosin cytoskeleton, the adhesive properties of cells and/or the composition of the extracellular matrix. This conjecture is consistent with the demonstration that several new disc cell surface proteins are induced by 20E (Rickoll and Fristrom, 1983; Sater et al., 1984; Paine-Saunders et al., 1990) including IMP-E2 and IMP-E1 (Natzle et al., 1988, 1992; Natzle, 1993). In addition, 20E also stimulates the appearance of extracellular proteases and the cleavage of certain surface glycoproteins (Pino-Heiss and Schubiger, 1989; Birr et al., 1990). Thus, it seems likely that the interference with several late embryonic morphogenetic processes also reflects a failure to induce genes necessary for tissue movement and remodeling.

Our genetic analysis of dib mutants revealed no germline requirement for dib function. This observation is surprising given that the ovaries are known to be a major site for the synthesis of ecdysteroids in many insects and several studies have suggested that ecdysteroids are required for oocyte maturation. The most compelling of these reports is the recent demonstration that germline clones, carrying mutations in the dare gene, arrest during oogenesis at stage 8-9 (Buszczak et al., 1999). Since AR, the product of the dare locus, is essential for transport of electrons to all known mitochondrial steroidogenic P450s, the oogenesis defect seen in dare mutant clones is highly suggestive that ecdysone, or some other precursor or degradation product, is essential for oogenesis. It remains possible however, that precisely because AR function is essential for activity of all mitochondrial P450s, dare mutants may have pleiotropic defects that are not directly associated with disruption of steroid biosynthesis. Nevertheless, a role for ecdysteroids in the germline seems likely since germline mutations of EcR as well as E75, a transcriptional factor that is know to mediate ecdysteroid responses, both arrest oogenesis at a similar stage as found for dare mutant clones (Buszczak et al., 1999). If ecdysteroids are essential for oogenesis, then why do dib mutant clones survive? One likely explanation is that the primary source of ecdysteroids in the ovary might be the follicle cells and not the nurse cells. In locusts and cockroaches, the follicle cells seem to be the primary site of ecdysteroid synthesis (Lagueux et al., 1977; Zho et al., 1983) and this observation is consistent with our findings that dib is primarily expressed in the follicle cells. This presents a dilemma however, since dare has been reported not to be expressed in the follicle cells and it would presumably need to be in this tissue if Dib is a mitochondrial enzyme. If such a scenario is correct, then it would require shuttling of intermediates not only between subcellular compartments, but also between the nurse cells and the follicle cells to produce the final ecdysone end product. Another explanation is that there is some expression of either dib in nurse cells or dare in follicle cells that is not readily detectable by in situ hybridization. The fact that very low levels of transgene expression were sufficient to rescue the dare mutant phenotype (Freeman et al., 1999) would be consistent with follicle cell expression levels that are functional but generally undetectable.

We thank Steve Harrison for generously sending us his collection of mutants from the 64A region and Katy Sullivan (Rubin lab, UC Berkeley) for providing the armadillo-GFP flies. We also thank the Drosophila stock center at Bloomington for numerous stocks and Tom Hays for the spectrin antibody. We are grateful to Carl Thummel for his continued interest in this project and comments on the manuscript. V. M. Chávez was supported in part by PHS grant GM47462 to M. BO. J. B. was supported by a predoctoral training grant HD07029 and G. M. was a NATO fellow. Work in the lab of J. E. N. was supported by Califonia Agricultural Experiment Station grant CA-D-NCB-5373. M. B.O. is an Associate Investigator for the Howard Hughes Medical Institute.