The Drosophila retinoid X receptor homolog ultraspiracle functions in both female reproduction and eye morphogenesis

The actions of the steroid, thyroid and retinoid hormones are predominantly mediated by the members of the steroid receptor superfamily. These receptors are ligand-dependent transcription factors that alter the expression of a large number of genes in response to hormones, thus controlling important developmental and physiological pathways (Evans, 1988). Much progress has been made toward the understanding of the biochemical and molecular nature of steroid receptor action, including the definitions of functional domains interacting with ligands, target genes and accessory transcription factors (for reviews, see Beato, 1989; Green and Chambon, 1988; Evans, 1988). However, little is known about the cellular and physiological roles of the nuclear receptors in development. Perhaps the best studied receptor/ligand system for the control of morphogenesis involves the vitamin A derivative, retinoic acid (RA). Classical endocrine studies demonstrate that RA can function in multiple tissues and induce differentiation of a wide variety of vertebrate cell lines from different cell lineages (Sporn and Roberts, 1983; Breitman et al., 1980; Strickland and Mahdavi, 1978). Further, RA alters positional information in developing avian and amphibian limbs (Brockes, 1989) by mimicking the actions of the zone of polarizing activity (ZPA) (Tickle et al., 1975). RA most likely acts indirectly by controlling the production of a morphogenic signalling molecule (Noji et al., 1991; Wanek et al., 1991; Tabin, 1991). Two subfamilies of the nuclear receptor superfamily mediate the transcriptional activities of RA, the retinoic acid receptors, which bind with high affinity to RA, and the retinoid X receptors, which bind to the 9-cis form of RA (Mangelsdorf and Evans, 1991; Heyman et al., 1992). The molecular mechanisms through which these retinoid-responsive receptors control morphogenesis are unknown.

Three isoforms of the RXR have been identified and have partially overlapping expression patterns during development (Mangelsdorfet al., 1992). Recent in vitro experiments demonstrate that many of the receptors physically interact with other receptors and other transcription factors (Forman and Samuels, 1990). RXR appears to play a central role in these interactions as it has been shown to multimerize in vitro with the COUP-TF, vitamin D, thyroid hormone, and retinoic acid receptors (Kliewer et al., 1992a,b). However, the biological significance of RXR and its interactions with other transcription factors during development has not been addressed.

As part of a genetic analysis of the hormonal control of development, we have examined the Drosophila genome for homologs of the retinoid receptors (Oro et al., 1988). While no homolog of the retinoic acid receptor has been isolated, a receptor gene sharing significant homology with the retinoid X receptor was identified. This cognate is encoded by the Drosophila ultraspiracle (usp) gene (Oro et al., 1990). This gene was independently identified based on binding of the usp protein to an essential promoter element of the chorion S15 gene (Shea et al., 1990) and by screening libraries of genomic DNA using receptor DNA-binding domain consensus oligonucleotides (Henrich et al., 1990). usp is expressed at high levels in adult females, as well as at lower levels throughout development, suggesting requirements for usp throughout development and in the adult (Oro et al., 1990; Shea et al., 1990). Mutations in the Drosophila RXR cognate make it possible to analyze in vivo interactions between an RXR family member and other receptors. The initial step in such an analysis is a characterization of the extent of the usp phenotype.

Here we analyze usp functions in development. In order to examine usp requirements after the initial lethal phase, we have generated conditional lethal usp genotypes and used them to reveal novel maternal and zygotic usp phenotypes. Through the analysis of usp mutants and usp imaginal disc mosaics, we show that usp is required for female reproduction as well as for normal eye development. These requirements involve both cell autonomous and non-cell autonomous processes and suggest that usp controls the production of factors important for cell-cell communication as well as for cellular differentiation.

Flies were raised at 25°C on standard medium containing cornmeal, sugar, yeast and agar. Balancer chromosomes used are described in Lindsley and Zimm (1990). Canton-S is the wild-type strain in our studies. Basic usp stocks were gifts of Norbert Perrimon and were crossed to a yellow (y) white (w) forked (f) stock to create y w f-marked usp chromosomes.

hs-usp, hsneo was constructed by placing the 2C8 cDNA from the usp locus into the EcoRI site of hshsneo vector (Boggs et al., 1987). P element injection and transformation procedures were performed essentially as described by Boggs et al. (1987). G1 offspring were selected on 250 μ g/ml G418 then balanced and grown on regular food. The hs-usp Ki p^p chromosome was constructed by recombination between the original hs-usp hsneo chromosome and a se Ki p^p chromosome. While six independent hs-usp lines were generated, only one of the six was able to complement usp mutants in the heat-shock assays.

The phenotypic analysis of usp mutants was performed using three usp alleles, usp^VE653, usp^VE849, and usp™²¹. usp^VE653and usp^vE849 are cytologically normal alleles generated by chemical mutagens and usp^KA21 is a rearrangement allele. Because of a lack of deficiencies for the 2C5-9 region of the genome, a complete genetic analysis of the usp alleles could not be performed. However, usp^KA21 is a transposition mutation, which contains a breakpoint in the middle of the usp transcription unit, interrupting the region encoding the DNA-binding domain, and therefore should be a strong or null allele (Oro et al., 1990). Moreover, the other two alleles also appear to be strong alleles: they have an identical zygotic lethal phase and exhibit the same phenotypes in germline clones as usp^KA21. usp^KA21 is complemented by the heat-inducible usp chromosome (see below) with a relative viability that is fivefold less than usp^VE653 or usp^VE849 and is sterile when the zygotic usp requirements are complemented by a P element containing the 8 kb EcoRI usp genomic region. We ascribe these differences to hyperploidy of the third chromosome as the usp^KA21 transposition contains a large segment of the left arm. Finally, usp/+ organisms are phenotypically wild type in all aspects tested, supporting the assertion that the usp alleles behave as true recessive alleles.

Morphology of embryos and larvae was examined by mounting in Hoyers mountant as previously described (van der Meer, 1977). Embryos were dechorionated in bleach and fixed in glycerol acetic acid (4:1) for 1 hour at 58°C then mounted in Hoyers. Larvae were digested for 4 hours. The chorion was examined by briefly washing embryos in PBS then mounting embryos directly in Hoyers. Morphology was examined under a dissecting microscope or a compound scope with Nomarski optics. Staging of the pupae was performed as described previously (Bainbridge and Bownes, 1981).

In situ hybridization to usp embryonic RNA was performed as described previously (Tautz and Pfielfle, 1989). Ovaries were removed from wild-type females after 2 days on well-yeasted food. In situ hybridization to ovaries was performed as previously described. This modification included addition of DMSO in the incubation buffer and appeared to reduce nonspecific background (McKearin and Spradling, 1990). Specimens were mounted and examined in 80% glycerol-PBS. Staging was performed as described previously (King, 1970). We used three different usp probes and varying proteinase K digestion times to optimize the usp signal. For studies of both ovaries and embryos, low levels of specific usp staining were observed that were not seen with control probes of vector and heterologous DNA.

The production of usp RNA in response to heat treatment was analysed using adults of the genotype +; hs-usp. Females of the indicated genotypes were placed in a 37°C bath in water up to the cotton plug for 60 minutes and subsequently returned to 25°C for the required period of time. Of the times tested, this length of heat shock gave the highest induction of usp RNA. Total RNA was isolated from flies as previously described (Thummel et al., 1990).

Primer extension of usp transcripts was performed using uniformly labeled, single-stranded DNA primers. A 800 bp PstI-SalI genomic fragment containing the usp 5’ region was cloned into M13 and used to generate single-stranded DNA template. A 30mer starting at nucleotide position 199 of the usp sequence (Oro et al., 1990) was hybridized to the template in 10 mM Tris pH 7.5 and 10 mM MgC12 and then the hybrid was extended for 45 minutes in 10 mM Tris pH 7.5,10 mM MgCl₂, 50 mM NaCl, 100 μ g/ml BSA, 1 mM dATP, dGTP, dTTP, 100 μCi [³²P]dCTP, and 2 U Klenow and chased for 15 minutes by the addition of 100 μ M dCTP and 1 U Klenow. The extended product was cleaved by digestion with MboII for one hour and isolated by electrophoresis on a denaturing polyacrylamide gel. Primer extension was performed as previously described (Ausubel, 1987). Labelled probe (300,000 cts/minute) was incubated overnight with 30 μ g total RNA in a 30 μl volume of hybridization buffer at 30°C and extended with Mu-LV Reverse Transcriptase (BRL) for 90 minutes at 43°C. The samples were digested with RNAase, precipitated and analyzed on a 6% sequencing gel.

The effect of ectopic usp expression was examined by crossing the +; hs-usp line to heterozygous usp mutant females and examining progeny classes in the presence or absence of heat shock. For the first cross, y usp^VE849/FM7c females were crossed with y usp^VE849/Y; hs-usp/+ males (see below). Offspring of this cross were collected and heat shocked starting from 12 hours after egg laying until eclosion by placing bottles containing them into a 37°C bath with the water to the level of the cotton plug for 60 minutes. This regimen was performed every 12 hours. During the intervening period, the flies were housed at 25°C.

Examination of additional usp lethal phases was performed by growing organisms at room temperature, giving them two 1 hour heat shocks separated by 12 hours during the first instar stage and then returning them to room temperature. The organisms of the genotype usp^VE653w f/Y; hs-usp Ki p^p/+ were selected by analysis of the Malpighian tubule phenotype and followed for their terminal usp phenotype.

Examination of the adult usp phenotype was performed by the generation of y usp^VE849/y usp^VE849-, hs-usp/ + females and y usp^VE849/Y’, hs-usp/+ males by crossing rescued usp/Y;, hs-usp/ + males to usp/FM7 females. 1 hour heat shocks were given every 12 hours through development until approximately one day before eclosion. Subsequently the flies were grown at room temperature for all phenotypic analyses. Adult fertility was ascertained by crossing usp males or females to wild-type females or males, respectively. Hatching assays were performed on apple juice plates every 24 hours. Unhatched eggs were examined for cuticular defects as described above. As controls for the effects of heat shock and gain of function usp effects, usp/FM7; hs-usp/+ and usp/FM7; +/+ which had undergone identical heat-shock regimens were compared. The hatching rate remained constant over the 7 days examined after eclosion, indicating that the remaining fertility was not due to perdurance of usp activity from developmental heat shocks. Moreover, the sterility is not attributable to ectopic maternal usp expression from the heat-shock promoter, as the usp⁺ control class of mothers had wild-type levels of hatching in our assay.

Analysis of large patches of usp tissue was attempted using the second chromosome paternal loss (pal) mutation (Baker, 1975). pal causes the loss of paternally derived chromosomes at early cleavage stage divisions resulting in XO/XX and XO/XY mosaics in both imaginal and non-imaginal tissues. y/y+Y’, pal’, spa^po1 males were crossed to usp^VE653w f/FM7 females and adult offspring analysed for presence of mosaic tissue. Cuticles were examined under a dissecting microscope for either white, forked or yellow patches. Note that in this analysis all cuticular tissues could be scored for hemizygosity of the usp chromosome. The control chromosome, FM7c, could only be scored in the eye, notum and sexually dimorphic regions of the genitalia and forelegs. This difference, coupled with the demonstrably lower viability of FM7c relative to the usp w f chromosome (as scored in males containing a single copy of a usp⁺ P-element on the third chromosome), indicates that the number of control class FM7c/0//FM7c/y gynandromorphs underestimates the number of usp w f/O/Jusp w f/y animals that were produced and failed to survive.

Analysis of smaller usp clones in imaginal tissues used gamma-ray-induced somatic recombination. This technique generates smaller usp clones by recombination after irradiation and subsequent cell division (Morata and Ripoll, 1975). Flies of genotype usp^VE653w f/Y-, hs-usp Ki pf/+ generated as described above were crossed to M(l) o^sp/FM6 females. Larvae were collected 36 – 48 hours after egg laying and irradiated with 1000 R from a ⁶⁰Co source. Subsequently, adults of the genotype usp^VE653w f/M(1) o^sp genotype were examined for the presence of forked bristle patches or white eye patches. As positive controls, f as described above were crossed to y w f/Y males were crossed to M(1)o^sp/FM6, irradiated and examined for yellow, white, or forked bristle patches as described above. usp⁺ and usp − clones were generated in approximately the same frequency indicating usp is not a general cell lethal. The mosaic retinas were fixed and sectioned as previously described (Cagan and Ready, 1989b). 1.5 μm sections from plastic-embedded retinas, stained with toluidine blue, were examined under a light microscope. Our method of tissue fixation washed pigment granules from photoreceptor cells, preventing unambiguous assignment of photoreceptor genotypes. The sunken eye phenotype was examined by embedding the mosaic heads in Tissue-tek, sectioning on a cryostat with 12 μ m sections, staining with Hoechst 33258 stain and visualizing with a UV filter.

Analysis of usp clones in the female germline was performed using the dominant female sterile mutation Fs(1)K1237 (Perrimon and Gans, 1983). This technique generated clones in the female germline simultaneously with removing a dominant female sterile mutation. Flies of the genotypes y usp^VE849/FM.7, usp^VE653w f/FM.7, or usp^KA21’ f/FM7 were crossed to Fs(1)K1237 v/Y males. Larvae were collected 36 – 48 hours after eclosion and irradiated with 1000 R with a ⁶⁰Co source. usp/Fs(l)K1237 virgins were collected and crossed to wild-type males. Fertility, hatching and morphology of the eggs from these mothers were examined as described above. One explanation for the appearance of the chorion phenotype in germline clones is that in the generation of the clones, both the germ cells and the mesoderm precursors of the follicle cells, which are not normally clonally related, were rendered mutant. Analysis of eggs from individual females bearing germline usp−clones shows that each mother lays eggs showing the chorion phenotype, with none of the clones giving rise only to eggs with wild-type appearance. This result is not compatible with the two clone hypothesis for mutant eggshell production.

The role of RXR homologues in Drosophila development was examined by exploring the consequences of loss of usp function. The earliest requirement for usp occurs during embryonic development (Perrimon et al., 1985). Females with usp − germ cells give rise to fully viable and fertile usp/+ daughters but non-viable usp/Y sons that die at or about the period of hatching from the egg. A detailed analysis of the embryonic phenotype (see below and Fig. 7A) indicates that most usp/Y progeny of usp − germline mothers die just prior to hatching, with the remainder dying shortly thereafter. The unhatched embryos die with cuticular scarring in the region posterior to the ninth abdominal segment. In contrast, usp/Y embryos derived from usp⁺ germ cells hatch and become first instar larvae. Most die between the first and second instar periods with no gross cuticular defects (data not shown). A small but significant proportion die as second instar larvae, as judged by spiracles and mouthhooks, with the remnants of the posterior part of the first instar cuticle loosely attached. This leads to the appearance of both first and second instar spiracles on the dead larvae, resulting in the name ultraspiracle (Perrimon et al., 1985).

In the absence of gross defects in these mutants, it is difficult to assess whether usp has widespread or restricted function. We have examined these alternatives by first characterizing the levels and distribution of usp transcripts during oogenesis and early development followed by an examination of the effects of a mosaic loss of usp function. In situ hybridization to whole mounts of ovaries from wild-type mothers reveals high levels of usp message accumulation in the nurse cells beginning at approximately stage 8 of oogenesis (Fig. 1A). The message accumulation appears equivalent among the various nurse cells and increases to a maximum in stage 10 egg chambers (Fig. 1B). By stage 12, the usp RNA is confined to the degenerating nurse cells (Fig. 1C). Because of the crosslinking of the chorion around the oocyte, usp message accumulation could not be examined in stages 13 and 14. The usp RNA synthesized in nurse cells and injected into the oocyte is uniformly distributed in the embryo and remains so through cellular blastoderm and germ band elongation (Fig. 1D-E). As the germ band retracts, a more distinct pattern of usp expression emerges with higher levels of staining occurring in the ventral nervous system and developing midgut (Fig. IF). The message accumulation is equivalent across the anterior/posterior axis in the stained regions. There is no staining in regions such as the amnioserosa. The observed patterns suggest that the generalized usp expression is a result of the maternal contribution, while the later more localized contribution is a result of activation of the zygotic genome. The widely distributed expression pattern suggests the existence of widespread zygotic requirements for usp.

To characterize specific usp requirements better, we have created mosaic animals that harbor usp⁺ and usp − tissues. These genetic mosaics were produced using the second chromosome mutation paternal loss (pal) (Baker, 1975). When homozygous in males, pal causes the loss of paternally derived chromosomes in the early nuclear divisions of their progeny, resulting in the production of gynandromorphs with large regions of usp − tissue. Regions of the body not requiring usp⁺ function could sustain usp mutant patches identifiable by markers linked to usp. Approximately 3500 usp/y progeny were examined for usp mutant patches and none were found, while a comparable number of control sibs yielded 58 gynandromorphs (Table 1). This absence of usp patches supports the notion that usp is required in multiple tissues for normal development.

The early lethal usp phenotype precludes simple examination of late usp functions. To circumvent this problem, a conditional expression system was constructed using the heat-inducible hsp70 promoter driving a usp cDNA (referred to as hs-usp, see Fig. 2A). Transgenic animals containing hs-usp were initially established in the absence of heat shock. Transcripts from the hs-usp gene are approximately 150 bp larger than those from the endogenous usp locus and are easily distinguishable from the endogenous product (Fig. 2B). Heat shocking hs-usp females causes the hs-usp RNA levels to increase to approximately fifty times the unheat-shocked level (compare w, hs-usp, heat – with w, hs-usp, heat +), while heat-shocking control females does not alter the level of endogenous usp transcripts (compare w, heat –with w, heat +). Upon returning the flies to room temperature (“hours”), the hs-usp RNA level decfined with an apparent half-life of 4 hours, although substantial levels of hs-usp RNA were still present 10 hours after heat shock. Expression from the endogenous gene was unaffected in these experiments, indicating that usp protein is not a positive or negative regulator of its own promoter.

In a usp mutant fly, basal expression of the hs-usp allele delays the lethal phase such that the majority of usp/Y; hs-usp individuals die in the second instar rather than dying at the boundary between the first and second instar larval stages. To see if higher levels of usp expression could rescue usp − individuals through the rest of the life cycle, usp − hs-usp animals were given heat pulses every 12 hours throughout development, starting at 12 hours after egg laying. If usp functions were regulated, such as through the local production of a hormone or cell-specific factor, high levels of usp would be expected to rescue usp mutants without any effects of ectopic expression. Indeed, continual high levels of usp expressed uniformly were not detrimental (Fig. 2C, compare the ratios of FM7/usp in hs-usp⁺, heat to hs-usp⁺, control and FWl/usp in hs-usp⁻, heat to hs-usp ⁻, control). Moreover, elevating usp levels by giving periodic heat pulses complemented the usp mutants, allowing survival up to adulthood (Fig. 2C). These higher levels of usp⁺ activity rescue both males and females equally (Fig. 2C, usp/Y and usp/usp columns, hs-usp⁺) to a relative viability of approximately 50% compared to non-usp, heat-shocked controls. The rescued adults had normal morphology and lifespan. The usp/Y; hs-usp/+ males, kept at room temperature after eclosion, were fertile while the usp mutant females were partially sterile (see below). These results suggest that tissue-specific expression is not the rate-limiting step for usp⁺ function and support the notion that an additional cell-specific factor or hormone controls usp activity.

In order to uncover additional usp functions after the initial lethal phase, first instar usp − hs-usp mutant larvae were collected and given a single 1 hour heat pulse, returned to room temperature, and followed during development. Most progressed into the late third instar and early pupal stages LI and L2 before dying while 50% died during the pupal stage P4. Because none of the animals examined in this experiment developed past pupal stage P4, examination of the early developing eye and putative later pupal phenotypes required a separate approach (see below). In contrast, usp − hs-usp larvae that were heat shocked in the first instar and additionally heat shocked during the mid-third instar period survived to adulthood. These results support the idea of a continuing requirement beyond the second instar which is partially rescued by perdurance of product expressed at an earlier stage. Moreover, the pupal usp phenotype suggests that usp plays an additional role during metamorphosis as usp mutants fail to progress through pupariation.

The requirement of usp⁺ function to complete metamorphosis suggests that it may contribute to the formation of adult structures. To examine the role that usp plays in particular cell lineages in the developing imaginal discs, usp mosaic organisms were generated by gamma-ray-induced somatic recombination. The usp gene was marked with the white (w) and forked (f) mutations; mutant cells could be identified by the presence of white patches in the retina or forked bristles on body cuticle, usp clones in the abdomen or the notum were generated at a wild-type frequency and had normal cuticle, bristle spacing and number (data not shown). In contrast, usp clones in the head revealed two distinct morphogenic abnormalities. First, usp retinas contained enlarged, split rhabdomeres in the mutant regions (Fig. 3). These rhabdomeres appear to wrap around the cell body, and are wider and more disorganized than wild-type rhabdomeres. Moreover, although wild-type rhabdomeres extend the entire depth of the retina from the distal surface to the more proximal fenestrated membrane, usp mutant rhabdomeres extend only the outside 20% of the depth of the retina with cell bodies continuing inward (compare Fig. 3A with 3B). The spacing of ommatidia clusters and photoreceptor cell number per ommatidium appear grossly normal, although exceptions are seen. In general, the photoreceptors in each ommatidium displayed either an abnormal phenotype with apical rhabdomeres or a wild-type phenotype, with few ommatidia displaying a mosaic phenotype.

A large fraction of usp − clones in the head revealed an additional phenotype in which the ventral third of the retina looked atrophic and significantly thinner than the dorsal portion of the retina. This sunken eye phenotype was often associated with a slightly irregular shaped eye, but with normal cuticle and bristles surrounding the retina (Fig. 4A). The sunken phenotype was always most severe at the ventral aspect of the eye and gradually lessened dorsally. The precise boundaries of the affected portion of the retina varied and often extended more dorsally at the periphery of the retina. This phenotype perhaps suggested involvement of other visual system structures. Examination of the head in frozen sections stained with the nuclear stain Hoechst 33258 revealed an atrophic retina, with cells in the retina being shorter than wild type, and displaying a decreased distance between the Rl-6 nuclei and the R8 nuclei (Fig. 4B-C). The lamina, medulla and brain in these flies appeared to be intact, although individual connections were not tested. External eye morphology indicated that gross ommatidial morphology was intact. This phenotype always correlated with the presence of a usp clone in the eye-antennal imaginal disc. Remarkably, the sunken phenotype did not correlate with the presence of a usp mutant retina as judged by the presence of white patches in the retina or forked patches in the cuticle. 36% of the head clones with this phenotype contained genotypically wild-type retinas. In the clones with a sunken eye and a mutant retina, the mutant clone could extend beyond the boundaries of the atrophic portion of the retina, while the atrophic portion of the retina continued beyond the boundaries of the clone. These data suggest that usp is required in a non-retinal cell type for a developmental process that ultimately manifests itself in mutant retinas.

The sunken eye usp phenotype is a non-cell autonomous phenotype and suggests that another cell in the eye-antennal imaginal disc requires usp⁺ activity for normal eye morphology. In principle, the location of the usp⁺-requiring cells should be identifiable by correlating the locations of the clonal patch in animals with sunken eyes. By the use of bristle and eye markers linked to usp on the X chromosome, 25 heads with usp − patches and the sunken eye phenotype were examined and the positions of usp − clones on the fate map of the imaginal disc scored relative to one another (Bryant, 1978). Clones that had boundaries that included portions of the eye (present 64%), bristles on the second antennal segment (present 64%), the orbital bristles (present 32%), and the vibrissae (present 32%) most often correlated with the presence of a the sunken eye phenotype (Fig. 5). A clone resulting in a sunken eye whose boundary included the eye always occurred on the medial border of the retina. The mere presence of a clone involving markers for these tissues did not indicate that the adjacent retina would be mutant. Clones were identified whose boundaries included these markers but that contained a wild-type retina, indicating that the cells in the imaginal disc that require usp⁺ activity are not the bristle or retinal cells themselves, but rather associated with the cuticle that lies between them. This analysis supports the unexpected conclusion that cells that lie between eye and antennal anlage are required for normal retinal morphology.

The ability to rescue female usp −, hs-usp individuals to adulthood with the conditional expression system allowed us to examine maternal usp contributions to female reproduction. Although the rescued mutant females with only basal levels of hs-usp being expressed were morphologically normal, they exhibited partial sterility when crossed to wild-type males. 60% of the eggs laid by usp −, hs-usp females were unfertilized (Fig. 6A). These eggs had wild-type polarity and displayed a morphologically normal micropyle. 20% of the eggs from these mothers hatch and develop into fertile adult females of the genotype usp/+, indicating that the eggs received a paternal wild-type X chromosome. The remaining 20% of the eggs were fertilized, but did not hatch and died in late embryogenesis (Fig. 7B). Some of these embryos had cuticular defects in the region posterior to the ninth abdominal segment, portions that include the anal pad. The anal pad structure was always present, but often had defective cuticle covering the structure. The fact that all of the embryos that survived to adulthood received a paternal X chromosome shows that the late embryonic maternal defect can be complemented by paternally supplied usp+ activity and is equivalent to the late embryonic defect observed in the male progeny of usp − germ cells as mentioned above. Both the fertilization defect and the late embryonic lethal phenotype can be partially rescued by heat shocks of adult usp; hs-usp females (data not shown). These results indicate that two distinct phenotypic processes contribute to the sterility seen in usp maternal mutants.

The eggs laid by usp; hs-usp mothers also exhibited a partially penetrant chorion abnormality. The chorion in 30 –50% of the eggs laid by usp/usp-, hs-usp/ + mothers exhibited patches of abnormally thin chorion, clear and shiny in contrast to its normal white opaque coloring (Fig. 7D). The anterior portion of the egg was most often affected, while the entire anterior-posterior axis occasionally displayed the chorion defect. These chorion abnormalities were not accompanied by embryonic pattern defects in that the anterior-posterior and dorsal-ventral polarity of the eggs appeared normal. Moreover, the chorion phenotype appeared unrelated to fertility as the presence of a clear chorion did not correlate with the inability to hatch in our assay. Fertilization and hatching frequencies were the same for eggs displaying the defective chorion phenotype as it was for those that did not show the phenotype.

To understand the role of usp in female fertility better, we have examined both usp expression and function in the germline and soma of the ovary. As shown above, ovarian usp expression occurs predominantly in the nurse cells and not in follicle cells up to stage 12 (Fig. 1B-C), arguing against a function for usp in follicle cells. Moreover, the functional contributions of the germline and somatic lineages in the ovary to the various usp phenotypes were examined by the analysis of mosaics generated through somatic recombination in the female ovary (Perrimon and Gans, 1983). These mothers have a wild-type soma and a usp mutant germline. Mothers mutant in their germline lay few unfertilized eggs, indicating that the fertilization defects seen in usp mutant mothers can be attributed to the lack of usp in the maternally derived soma (Fig. 6B). Of the fertilized eggs, 50% hatch and give rise to normal fertile females, while the usp/Y eggs die as embryos with a cuticular scar resembling that seen in usp/Y eggs derived from maternal usp mutants (Fig. 7A). This confirms previous work (Perrimon et al., 1985) indicating that the maternal germline and not the soma is used for normal embryogenesis and demonstrates a novel maternal somatic usp requirement.

The origins of the chorion defect were also examined through the analysis of germline mosaic animals. Since the chorion is secreted by the follicle cells surrounding the oocyte, mothers with a mutant germline and a wildtype soma (including follicle cells) would be expected to lay eggs with wild-type chorions. However, germline clones from three different usp alleles show a chorion phenotype similar to that seen in eggs from usp/usp; hs-usp/+ mutant mothers (Fig. 7C). As with the defect observed in eggs from usp/usp; hs-usp/+ mutant mothers, there is no adverse affect on egg viability associated with the chorion defect in these eggs. Thus we conclude that usp⁺ activity is required in the germ cells to produce a phenotype involving the follicle cells in the ovary. The only visible chorion phenotype associated with loss of usp function can be attributed to loss of function in the germ cells and not to loss of function in the cells that produce the eggshell.

We have analyzed usp mutants to begin to understand how members of the RXR subfamily may function in development. As expected, usp is required in multiple tissues and at multiple times in development. To gain additional information about usp requirements, we have developed a conditional usp expression system to examine the range of usp⁺ functions and have shown novel maternal and zygotic usp phenotypes. Through mosaic analysis, we have shown specific requirements for usp in the developing eye-antennal disc and during oogenesis. In both cases, usp is required for cell autonomous and non cell-autonomous interactions.

The usp phenotype is pleiotropic in that the gene is required at multiple stages of development for multiple functions which cannot be localized by mosaic analysis to a particular tissue. One of the first usp requirements is a mid-embryonic function required for cuticle production in the posterior abdominal region. Using a heat-inducible usp expression system to supply usp⁺ activity, we have identified usp lethal phases at each of the larval instars and up until stage P4 of pupal development. Continual high levels of usp activity rescue usp mutants to adulthood. While long heat shocks during the first instar stage allow usp mutants to continue to develop through the larval instars to stage P4 of pupariation, shorter heat shocks only delay the lethal phase until the late third instar stage. These results suggests that there is a threshold of wild-type usp activity required for normal development.

The fact that high levels of ectopic usp expression did not result in additional lethality or abnormality suggests that tissue- or temporal-specific synthesis of usp product is not normally the critical regulatory step in the processes with which usp is involved. This contrasts with other Drosophila nuclear receptors such as tailless (Steingrimsson et al., 1991) and other transcription factors, such as fushi tarazu (Struhl, 1985), for which ectopic expression results in distinctive alterations in development. One possible explanation for the benign behavior of ectopic usp expression is that the non-liganded receptor is inactive and its activity is dependent upon a locally produced hormone or cell-specific factor. We favor this model for two reasons. The first is that usp shares significant homology with the vertebrate nuclear receptor RXR, which transcriptionally regulates target genes in cultured Drosophila and mammalian cells only in the presence of RA (Mangelsdorf et al., 1990). Second, studies of usp in these systems reveal that usp does not compete efficiently with hormone-activated RXR, even though it can bind to the same target gene sequences (Oro, 1991). The simplest explanation is that usp, like RXR, requires a hormone for transactivation, but we cannot exclude the possibility of a requirement for an additional cell-specific factor or covalent modification.

The creation of a usp conditional expression system coupled with mosaic analysis allowed us to examine the role for usp in the development and differentiation of the eye and ovary, usp function is essential in both the germline and the soma of the adult female. Maternal somatic usp function is required for fertilization of the egg, a process that requires a wide range of soma-derived ovarian and extra-ovarian factors. Ovarian-specific processes include synthesis by the border cells of a sperm canal for sperm entry or the synthesis and uptake by the follicle cells of yolk proteins (Parks and Spradling, 1987). That ovarian usp expression appears low in the stage 12 centripetally migrating follicle cells suggests the defect is not the result of failure of usp action in the follicle cells.

In contrast, the germline is involved in supplying usp RNA to the oocyte. Maternal germline-derived usp in a zygotic usp background can complement the embryonic functions and allow development until the lethal phase at the end of the first instar. This maternal RNA is required for a late embryonic function involving the production of cuticle covering the ventral portions posterior to the eighth abdominal segment. While phenotypic analysis reveals that usp is not involved in specification of the eighth and ninth abdominal segments (data not shown), it may play a role in the terminal tenth and eleventh abdominal segments (Jurgens, 1987).

A second ovarian phenotype revealed a signalling pathway between the maternal germline and the follicle cells. Eggs derived from mothers with a usp mutant germline display a defective chorion phenotype similar to that observed in eggs from usp/usp; hs-usp/+ mothers. This indicates that the mutant chorion phenotype can be fully explained by the lack of usp in the germ cells without postulating a role for usp in the follicle cells. Consistent with this hypothesis, the follicle cells up to stage 12 of oogenesis accumulate little or no usp message. This is in apparent contrast to previous studies which identified a usp cDNA from a follicle-cell-enriched expression library on the basis of its ability to bind to a region of the chorion S15 promoter absolutely required for expression (Shea et al., 1990). Since our somatic usp − individuals were expressing the usp product at the level of basal heat shock our data cannot absolutely rule out a role for usp in follicle cells. However, the simplest model for the observed effects of usp on choriogenesis is one in which normal chorio-genesis requires usp-dependent local signalling between the germline and soma. Interactions between the germ and soma have been shown previously in the process of dorsal-ventral patterning and egg shell synthesis for genes such as gurken and fs(l)K10 (Schupbach, 1987). These two genes are required in the germline for the production of a normal chorion pattern and are thought to be involved in the correct production or distribution of spatial information from the germ cells to the follicle cells.

Using imaginal disc mosaic analysis, two distinct developmental eye abnormalities have been identified. In the first, the photoreceptor rhabdomeres lose their normal cylindrical morphology and become apically displaced, although the gross spacing and apparent cell number per ommatidium is normal. This is a novel phenotype in that the total cell number appears normal, but the morphology of each cell is abnormal. This suggests that at least part of the usp function in eye development occurs after the assembly of the ommatidia during retinal morphogenesis. Rhabdomeres are known to form, shortly after the photoreceptors differentiate during stage P7 of pupal development, by sequential stacking of the membranes until there is a nearly crystalline lattice of villi (Cagan and Ready, 1987b). The altered rhabdomere morphology suggests that usp is involved in a process that disrupts photoreceptor cell membrane stacking. This process could directly involve membrane stacking or be a consequence of an alteration in another developing cellular structure.

usp is not the only member of the steroid receptor superfamily that functions in the developing retina. Phenotypic analysis indicates that the Drosophila nuclear receptor sevenup (svp) is required for eye morphogenesis (Mlodzik et al., 1990). svp is the Drosophila cognate of COUP-TF, a vertebrate transcription factor that regulates the chicken ovalbumin promoter (Wang et al., 1989) and is structurally no more related to usp than to other members of the steroid receptor superfamily. Analysis of the function of RXR and COUP indicates that they may functionally interact through the recognition of similar target sequences and the ability to form heterodimers in vitro (Kliewer et al., 1992a). This suggests that usp and svp may also interact and regulate similar developmental processes. While usp and svp both function in the eye, the respective rhabdomere phenotypes appear distinct. svp is required for the specification of photoreceptor cell fates in the developing ommatidia, but is not apparently required for normal rhabdomere morphology. This supports the idea that usp and svp are not part of a linear pathway of rhabdomere morphogenesis. The interaction of the two genes in other processes not examined in this study, such as the determination of photoreceptor cell fate, may uncover novel biological interactions.

The sunken eye phenotype appears to be caused by the lack of usp function in cells of the imaginal disc between the photoreceptor and antennal anlage. The phenotype appears to involve primarily the thickness of the retinal layer and the length of the photoreceptor and accessory cells, although connectivity of the photoreceptors to layers of the visual brain was not examined. The affected portion of the eye always began ventrally in each clone and included varying degrees of the retina. Because multiple clones exhibited the phenotype and each animal that displayed the phenotype had a clone in the eye-antennal disc, we conclude that the origin of the usp⁺-requiring cells is from the eye-antennal imaginal disc alone and not from embryologically distinct structures like the optic lobes. The correlation of usp − cells with the sunken eye phenotype narrows the location of the usp-requiring cells to the region of the disc between precursors of the antennal and eye structures (See Fig. 5).

Genetic analysis of the developing fly has identified a growing class of genes that function in the development of multiple tissues. One such gene is Notch, which plays important roles in cell fate decisions in the embryonic neuroectoderm as well as developing sensory organs (Hartenstein and Posakony, 1990), ommatidia (Cagan and Ready, 1989b) and egg chambers (Ruohola et al., 1991). This observation suggests that some neurogenic genes, as well as other regulators, may be re-used in key signalling pathways to trigger different effects at different times in development. In this case usp would function to regulate a set of distinct target genes in different tissues in response to hormonal signals. Thus, usp in conjunction with its ligand may constitute a self-contained or modular regulatory unit controlling such diverse systems as female reproduction and eye morphogenesis.

The non-cell autonomous interactions occurring in the developing ovary and eye-antennal disc are occurring at a local level. Similar local interactions have been known to play important roles in photoreceptor cell fate determination (Banerjee and Zipursky, 1990) and egg morphogenesis (Manseau and Schupbach, 1989). From our analysis, the exact distance over which these signals must cross are not known. For the sunken eye phenotype, the large distances between the proposed usp⁺-requiring cells and the affected cells appear to rule out simple models of cell-cell interactions, unless cells derived from the usp⁺-requiring region somehow intercalate within the developing retina, an unlikely possibility given previous studies of eye morphogenesis (Cagan and Ready, 1987a). An alternative and more likely model is the existence of secreted factors produced by the cells between the anlage for the eye and antenna in response to the action of usp.

The usp signalling pathways in the retina and ovary have interesting parallels to the mechanisms of steroid hormone and retinoid action in developing vertebrate tissues. RA acts in the developing chick limb bud by inducing the ZPA to produce an apparently diffusible morphogenic signalling molecule (Noji et al., 1991; Wanek et al., 1991; Tabin, 1991). RA presumably acts through its receptors to control the production of the proposed morphogen, just as a putative usp hormone would act through usp in eye-antennal disc or germ cells to produce a factor important in eye or eggshell morphogenesis. Similar processes occur in the testis during spermatogenesis. Peritubular cells of the testis respond to androgens produced by the Leydig cells to secrete the paracrine factor P-MOD-S across the bloodtestis barrier. P-MOD-S is an important modulator of Sertoli cell function, crucial for spermatogenesis (Skinner, 1991). While the precise molecules that mediate the morphogenesis may or may not be identical, the study of usp action may be a useful model system for understanding more complex developmental processes in both vertebrates and invertebrates.

The authors would like to thank Ross Cagan and Larry Zipursky for sections of the usp mutant eyes and Ramakris-nan Rangarathan and Charles Zuker for whole head sections. The authors would like to also thank Tony Manly for technical assistance, and Bill Segraves and Tso-Pang Yao for helpful discussions. This work was supported by Medical Scientist Training Program (A.E.O.) and grants from HHMI (R.M.E.), Pew Scholars Program in Biomedical Sciences (M.McK.), NIH (M.McK.) and a Cancer Center Core Grant (CA-14195) to the Salk Institute.