The Sf1-related nuclear hormone receptor Hr39 regulates Drosophila female reproductive tract development and function

Some molecular pathways of sex determination evolve rapidly whereas others are relatively conserved (reviewed by Marin and Baker, 1998). The roles played by steroid hormones in vertebrates and invertebrates appear to be among the most divergent. Multiple steroid hormones serve non-autonomously as master regulators of male or female sexual development in mammals, whereas the Drosophila molting hormone ecdysone acts sex-specifically only during adult oogenesis (Buszczak et al.,1999; Carney and Bender,2000; Hackney et al.,2007; Li et al.,2000). Steroid-producing (`steroidogenic') tissues arise early in mammalian embryos under the control of the nuclear receptor steroidogenic factor 1 (SF1) (reviewed by Val et al.,2003) and function throughout life. The closely related protein LRH1 (NR5A2) also plays an important role, especially in ovarian function(reviewed by Fayard et al.,2004). The Drosophila genome encodes two proteins that are closely related to SF1 and LRH1, FTZ-F1 and HR39, which bind to similar target sequences (Ohno et al.,1994), but neither has been implicated in sex determination(reviewed by King-Jones and Thummel,2005). Steroid production in the ovary, the only known site of adult steroidogenesis, is not known to depend on either gene.

Despite these differences, in both mammals and Drosophila the gonads and reproductive tract develop in a generally similar manner. During mammalian embryogenesis, SF1 is required to produce androgens and Müllerian-inhibiting substance, a TGFβ family member that causes the oviduct precursors to degenerate. Drosophila reproductive tract precursors also develop in a sex-specific manner within the bipotential genital disc (Keisman et al.,2001), but a corresponding genetic pathway has not been found. In Drosophila (reviewed by Bloch Qazi et al., 2003), the spermathecae and seminal receptacle, which carry out long-term and short-term sperm storage, branch from the oviduct,along with the glandular parovaria (Fig. 1A). In mammals, long-term sperm storage takes place within the male epididymis, whereas the oviducts receive glandular secretions and can maintain sperm briefly (see Suarez and Pacey, 2006).

Gametes also undergo a complex maturation process in both mammals and invertebrates. Following production in the testis, mammalian sperm are immotile and incapable of fertilization. Only after passing through two other steroid-regulated tissues, the male epididymis, where they encounter extracellular proteases, antioxidants and anti-bacterial proteins (reviewed by Cooper and Yeung, 2006), and the female reproductive tract, where they contact mucins and membrane glycoproteins (reviewed by Suarez and Pacey, 2006) are sperm fully capacitated for fertilization. At the time of mating, Drosophila sperm are mixed with bioactive peptides and other proteins from the male accessory gland. Following transfer to the female, sperm have been proposed to interact with proteins synthesized by the female reproductive tract prior to storage in the seminal receptacle and spermathecae (Bloch Qazi et al.,2003; Lawniczak and Begun,2007). However, the identity, function, origin and regulation of female sperm-interacting proteins remain poorly known.

We find that Hr39 functions in a manner reminiscent of SF1. Hr39 is required for the normal development and function of spermathecae and parovaria. Thus, as in mammals, a Drosophila SF1-related gene mediates the sex-specific development of an essential region of the reproductive tract. Moreover, our results show that spermathecae and parovaria secrete proteins that function in sperm maturation, as well as in storage. Conserved steps in sperm maturation may take place at the sites of sperm storage, i.e. the epididymis in mammals or the female sperm storage organs in Drosophila. Our work reveals closer connections between Dipteran and mammalian reproductive biology than previously believed, and raises the possibility that novel steroid hormones regulate aspects of Drosophila reproduction.

Fly stocks were maintained at 20-25°C on standard cornmeal-agar-yeast food. The yw strain was used as a control in all experiments and to generate transgenic flies. The Hr39 alleles were isolated in three different single P-element mutagenesis screens as described in Table S1 (see supplementary material). The Hr39^k13215 line was obtained from Carl Thummel and is described in Horner and Thummel(Horner and Thummel,1997).

The fertility and spermathecae number of wild-type, heterozygous and homozygous mutant female flies was determined in the following manner: a single female was placed with two yw males in a vial for 5 days at 25°C. On the fifth day, the flies were removed and the female dissected to determine the number of spermathecae. The vial was then allowed to develop for 20 additional days at 25°C before the progeny in each vial was counted and recorded.

Full-length Hr39 cDNA, LD45021, was cloned into the Gateway entry vector and then swapped into the pUASt vector to make a P-element construct in which protein expression is under control of the yeast upstream activating sequence (UAS). P-element transformation was performed by standard procedures. 26 lines were generated, of which 11 were homozygous viable. The chromosomal locations of the P-elements were determined through standard crosses and appropriate transgenic animals were crossed into each mutant Hr39line in order to obtain the homozygous mutant Hr39, transgene and heat-shock driver in one fly. These flies were maintained at 20°C to minimize leakiness of the transgene and then heat shocked during larval development as third instar larvae for 30 minutes at 37°C. Fertility assays of transgenic animals were performed as described above.

RNA was isolated using either TriZOL reagent (Invitrogen) or Qiagen RNeasy kit following the manufacturer's protocol. The RNA was treated with 2 U/μl DNase overnight (Ambion) according to manufacturer's instructions. One-Step RT-PCR (Qiagen) was then performed using 0.5 μg of the isolated RNA and primers designed to span an intron within Hr39 or RpS17 as a control. Primer sequences are available upon request. The PCR machine was an MJ Research PTC-100 Programmable Thermal Controller and the program used: 30 minutes at 50°C, 15 minutes at 95°C, 29 cycles of 30 seconds at 94°C, 30 seconds at 55°C, 1 minute at 72°C, followed by 10 minutes at 72°C. PCR products were resolved on 1% agarose LE gels (Roche) in 0.5× TBE buffer with 0.25 μg/μl ethidium bromide. Gel images were acquired by using the BioRad Gel Doc XR scanner and Quantity One software(V4.5.2). All experiments were carried out at least in triplicate and a representative data set is shown.

RNA was isolated and Qiagen's One-Step RT-PCR kit used as described above under RT-PCR. Primers were designed to span introns for each gene tested(Hr39, RpS17, Cyp4d21, takeout, AttC and Pbprp1) and sequences are available upon request. Quantitative RT-PCR reactions were carried out on an Opticon Monitor 2 (MJ Research) using a 25 μl reaction comprising 0.25 μg total RNA and 0.25 μl of a 7.5× SYBR Green stock (Molecular Probes). The program used was 30 minutes at 50°C, 15 minutes at 95°C, followed by 40 cycles of 30 seconds at 94°C, 30 seconds at 55°C and 1 minute at 72°C. Finally the melting curve of each sample was determined. Results were analyzed using the Opticon Monitor software. Transcripts were expressed relative to the transcript of the control RpS17 gene and normalized to the control female for each gene tested.

Whole-mount samples were fixed with 4% paraformaldehyde for 15 minutes and processed using standard procedures (Cox and Spradling, 2003). The following antisera were used: rabbit anti-β-gal (pre-absorbed against lower reproductive tracts or ovaries,1:1000) (Cappel), mouse-anti-Fas2 (1:2) (1D4, Developmental Studies Hybridoma Bank), mouse-anti-Dac (1:200) (Abdac2-3, Developmental Studies Hybridoma Bank), and mouse-anti-Wg (1:50) (4D4, Developmental Studies Hybridoma Bank). Secondary antibodies were used at 1:500 and are as follows: goat anti-rabbit conjugated to Alexa 488 and goat anti-mouse conjugated to Cy3 (Molecular Probes). For DNA labeling, DAPI was added 1 μg/ml for 5 minutes.

Whole-mount in situ hybridization was performed by generating sense and antisense RNA probes by in vitro transcription from PCR products. Methods were previously described by Liu et al. (Liu et al., 2006) and Lécuyer et al.(Lécuyer et al.,2007).

Confocal images were taken with a 20× (NA 0.70) or a 40× (NA 1.25) Plan Apo objective on laser-scanning confocal microscopes (NT or SP2;Leica). Images were taken with the laser intensity and photomultiplier gain adjusted so that pixels in the region of interest were not saturated(`glow-over' display). Contrast and relative intensities of the green (Alexa 488), red (Cy3) and blue (DAPI) images were adjusted with Photoshop (Adobe). All confocal images are projected z-stacks.

Electron microscopy was carried out essentially as described(Cox and Spradling, 2003).

RNA from either lower reproductive tract (minus spermathecae) and spermathecae was made by dissecting young wild-type or Hr39⁰⁴⁴⁴³ females 3 days after mating as described above. Tissue samples were quick frozen in liquid nitrogen and kept at -80°C until enough sample was isolated (∼2000 spermathecae and ∼1000 lower reproductive tracts). The microarray was performed by the Johns Hopkins Microarray Core Facility on Drosophila version 2.0 Affymetrix chips using either 10 μg lower reproductive tract RNA or 2 μg spermathecae RNA. The microarray was performed twice using two different sets of RNA. The absolute difference between the replicate measurements for all genes called as present averaged less than 26% of their mean. The changes in expression observed for selected genes were fully verified by quantitative RT-PCR on RT samples (see below).

To investigate the role of Hr39 in female gametogenesis, we analyzed five mutations (see Table S1 in the supplementary material) isolated as causing female sterility in three different single P-element screens that contain insertions within the differentially spliced Hr39transcription unit (Fig. 1B). Complementation tests showed that these mutations are allelic (not shown), and that they also fail to complement a specific deficiency, Df(2L)Exel6048 (see Fig. S2 in the supplementary material), that deletes all Hr39-coding sequences but not upstream sequences. Homozygous females from each line produce apparently normal mature oocytes, but females bearing four of the alleles are usually sterile (Hr39^ly92,Hr39^neo8, Hr39⁰³⁵⁰⁸, Hr39⁰⁷¹⁵⁴)whereas Hr39⁰⁴⁴⁴³ and a sixth allele Hr39^k13215 (Horner and Thummel, 1997) are almost as fertile as wild type. Further examination revealed that the reproductive tracts of all the mutants frequently lack the two spermathecae and two parovaria characteristic of wild type (Fig. 1A,C). Mutant adults bearing the four strongest alleles usually lack parovaria and spermatheca or contain just one spermatheca (Fig. 1D,H). Flies bearing weaker alleles with one or two spermathecae also usually have one or two parovaria. Hr39⁰⁴⁴⁴³ females are unique because they frequently contain an extra spermatheca(Fig. 1G,H). Interestingly,although these 3-spermathecae-containing Hr39⁰⁴⁴⁴³ females still contain just two parovaria, their parovaria are distinctly larger than wild type (compare Fig. 1E with 1F).

The observed defects in spermathecae and parovaria might be responsible for the sterility of mutant females bearing strong Hr39 alleles, or there might be unapparent defects in the ovary or some other tissue. To distinguish these alternatives, we investigated whether spermathecal and/or parovarial content correlated with fertility at the level of individual female flies. Such a relationship was suggested by our observation that most Hr39mutant females were completely sterile, and what appeared to vary between alleles was the frequency of rare females with significantly greater fecundity(data not shown). Consequently, we scored the fertility of hundreds of individual Hr39 mutant females and subsequently determined the number of spermathecae and parovaria they contained(Table 1, see also Table S2 in the supplementary material).

Our results reveal an extremely strong correlation between spermathecae and fertility, such that possession of even 1 spermatheca is associated with a many-fold increase in fertility and fecundity, while possession of two spermathecae engenders near wild-type fertility. Whether parovaria alone can support fertility could not be determined from these data because we never recovered females that possess a parovarium but no spermatheca. Some females contained spermathecae that were smaller and morphologically abnormal; such organs were included in the counts. Despite this, we suspect that the very low level of residual fertility in females `lacking' both spermathecae and parovaria (8% fertility, with a fecundity of only four progeny), is due to rare females that contain tiny defective spermathecae or parovaria that escaped detection, but that retain a low level of function. Combining the information in Fig. 1H and Table 1 explains why the Hr39 insertions were originally isolated as female sterile mutations.

A correlation between spermathecae and fertility has been shown previously in studies of lozenge mutations that also produce females with a variable number of morphologically normal or defective spermathecae, but no parovaria (Anderson, 1945). It has been proposed that spermathecae produce a product required for sperm storage. However, when we examined the seminal receptacles of 3- to 6-day-old wild type and strong Hr39 mutant flies mated on day 1 that lacked spermathecae and parovaria, both DAPI staining (see Fig. S1 in the supplementary material) and electron microscopy (not shown) revealed normal numbers of sperm in the mutants. We also noted the presence of sperm in seminal receptacles of much older Hr39 mutants. Thus, infertile Hr39 mutant females lacking spermathecae and parovaria, still transfer normal amounts of sperm at mating (data not shown) and maintain normal amounts of sperm in their seminal receptacles. Consequently, our data suggest that spermathecae (possibly including their small associated segment of fat body) produce a product that is required for sperm to function, despite their ability to be stored.

To analyze how the insertion mutations affect spermathecal and parovarial development, we attempted to analyze Hr39 expression in the genital disc, the anterior region of which is known to give rise to both structures from tiny primordia shortly after the onset of prepupal development(Keisman et al., 2001). However, we were unable to generate specific anti-Hr39 antibodies or to carry out whole-mount in situ hybridization on larval genital discs with either Hr39-specific or control probes. Furthermore, no defects in the structure or gene expression of these discs was apparent in late stage larvae, as we could detect no changes in the expression patterns of Engrailed,Wingless, Dachshund or Abdominal-B using specific antibodies (see Fig. S3 in the supplementary material; data not shown).

Consequently, to investigate the tissue-specificity of Hr39expression and to verify that the insertions in the sterile alleles alter Hr39 expression, we analyzed Hr39 RNA in adults. A significant level of Hr39 transcripts was detected in RNA from adult females (and lower levels in adult males) when analyzed using RT-PCR targeting a common region within the known transcript isoforms(Fig. 2A,B). Females bearing each of the mutations contained significantly reduced Hr39 transcript levels, as expected. The Hr39⁰⁴⁴⁴³ allele behaved genetically like a hypermorph with respect to spermathecal and parovarial production (see Fig. S2 in the supplementary material). Hence, an increase in Hr39 expression is expected in the genital disc or some other larval or early pupal tissue; however, this fact does not contradict the observed reduction in adult expression levels. The expression and independent regulation of Hr39 in adults suggests that this gene functions in adults as well as during development.

Individual tissues expressing Hr39 were identified using whole-mount in situ hybridization (Fig. 2). Except as noted, the presence of RNA detectable by in situ hybridization corresponded closely to the enhancer trap expression patterns of the Hr39⁰³⁵⁰⁸ and Hr39⁰⁴⁴⁴³ alleles. Thus, Hr39 RNA was detected directly and by enhancer trap staining in the lateral (ecdysone-producing) cells of the larval ring gland(Fig. 2C), the larval ovary(Fig. 2E), the spermathecae and parovaria (Fig. 2F,G), the seminal receptacle (Fig. 2H),the spermathecae-associated fat body (Fig. 2I), the spermathecal capsule cells(Fig. 2J-L), the adult ovariole(Fig. 2M), and the adult testis(Fig. 2N). In addition, a low uniform level of enhancer trap staining was observed throughout the entire larval genital disc (Fig. 2D). All these tissues contribute directly or indirectly to the development and function of reproductive tissue. Hr39 expression in the gland cells of the spermathecae was mosaic when assayed by whole-mount in situ hybridization, while enhancer trap staining was more uniform, presumably owing to the longer perdurance of the β-galactosidase protein(Fig. 2J-L). Possible cyclic activity of these cells has been noted previously(Filosi and Perotti,1975).

We carried out rescue experiments to verify that the defects in fertility and reproductive tract development within the Hr39 insertion strains are caused by alterations in this gene. Both the spermathecal, parovarial and fertility defects of all four strong Hr39 mutants were rescued in the presence of an hsp70GAL4 driver and a full-length Hr39-RA cDNA under control of a UAS promoter (Fig. 3; see Table S3 in the supplementary material). Sometimes spermathecal development was incompletely rescued as only one spermatheca(Fig. 3B) or two smaller glands(Fig. 3C) were restored. Elevated Hr39 expression affected spermathecal number, as even wild-type files bearing these constructs frequently contained extra spermathecae (Fig. 3A). Furthermore, applying heat shocks during larval development to Hr39mutants that also contained hsp70GAL4::UAS-Hr39 frequently caused the number of spermathecae (but not of parovaria) to increase to three(Fig. 3D,E) or four(Fig. 3F). Although Hr39-RA overexpression affects spermathecal number in a similar manner as Hr39⁰⁴⁴⁴³ mutation, it differs in that the parovaria are not enlarged. Differences in the cellular locations and particular Hr39 isoforms overproduced under these two conditions probably account for this discrepancy.

Our previous observations suggested that spermathecae, and possibly parovaria, produce a secreted product(s) required for sperm function. We further investigated the nature of this product and the effects of Hr39 mutations using electron microscopy of wild-type and mutant spermathecae and parovaria. Wild-type spermathecae(Fig. 4A) contain multiple gland cells (outlined), connected via end apparati (EA) and ducts to the lumen(L) of the capsule. The lumen (Fig. 4B) is filled head first with a highly ordered collection of sperm(S) surrounded by lightly staining material (M). In spermathecae from 3- to 5-day-old wild-type mated females, most of the capsule cells appear to be actively secreting material into the lumen, because their end apparati are swollen with a poorly staining material that partially obscures the villi(Fig. 4C, arrow). By contrast,a few cells appear to lack secretion. Their end apparati are smaller and contain more readily visible microvilli(Fig. 4D). These apparent differences in capsule cell secretion may reflect the cyclic activity described previously.

Studies on wild-type parovaria revealed a surprising resemblance in cellular organization to spermathecae, but with a much thinner cuticle. Parovaria are largely made up of gland cells with a similar morphology to those of the spermathecal capsule (Fig. 4E). Each is connected to an end apparatus that appears to contain only a relatively small amount of product. Neither sperm nor the lightly staining material seen in spermathecae is ever present in the lumen.

Our studies of Hr39 mutant females that manage to acquire a spermatheca show that they are much more fertile than their siblings that lack these organs, but that a majority remain sterile(Table 1). Consistent with this observation, rare spermathecae produced by strong Hr39 mutant females(Fig. 1H) show a range of structures when examined under the electron microscope. Some are small,abnormal and contain many necrotic cells but still store sperm (not shown). These probably correspond to the 57% of females with one spermatheca that are still sterile (Table 1). Others, however, are generally normal in structure, and contain secretory material in their end apparati (Fig. 4F, red arrow), although the amount of secretory material is usually less than in wild-type spermathecae. These probably correspond to the fertile females that showed reduced fecundity.

Once again, the behavior of the Hr39⁰⁴⁴⁴³ mutant females differed from females bearing any of the other alleles. Hr39⁰⁴⁴⁴³ females contain two and frequently three spermathecae, and their general structure and sperm content was normal(Fig. 5A,B), except for a possible increase in the frequency of dying cells (asterisk). However,detailed examination of the secretory cells suggested that these glands produce little or no secretion. No secretory product was present in their end apparati (Fig. 5C), causing all the cells to resemble inactive normal cells(Fig. 4D). Despite the great reduction in spermathecal secretion, sperm surrounded by normal lightly staining material are present in the lumen of these glands(Fig. 5B), which support nearly wild-type levels of fertility (Table 1).

A likely explanation for this paradox was found when we examined Hr39⁰⁴⁴⁴³ parovaria, which we noted previously are significantly enlarged (Fig. 1E,F). Hr39⁰⁴⁴⁴³ parovaria and their end apparati are highly swollen with secretion(Fig. 5D, arrow). These observations suggest that Hr39⁰⁴⁴⁴³ mutation either directly stimulates increased parovarial secretion, or by blocking spermathecal secretion indirectly induces parovaria to produce a compensating product. All previous data are also consistent with the idea that parovaria can produce a functionally equivalent secretion to that of the spermathecae. The loss of parovaria alone has not been correlated with any defects in female reproduction, whereas in sterile mutations that lack spermathecae, parovaria are also always absent.

To further characterize the biology of spermathecae and the role of Hr39, we carried out gene expression studies on young females 3 days after mating to wild-type males. RNA was isolated from more than 2000 hand dissected wild-type spermathecae, and from 1000 derived from Hr39⁰⁴⁴⁴³mutant females. Unfortunately, it was not feasible to carry out a similar isolation of the rare spermathecae from other Hr39 mutants or from parovaria. However, for comparison we prepared RNA from the remainder of the female reproductive tract (RT), i.e. oviducts,uteri and seminal receptacles from each genotype. Parovaria were also present in this material, but are expected to contribute a very small fraction of the RNA. The genes expressed within these unamplified RNA populations were analyzed using Affymettrix gene chip 2 arrays and the highly reproducible results were analyzed without any further mathematical manipulation (see Materials and methods).

Frequently, the most highly expressed mRNAs within a secretory tissue encode its secretion products. Consistent with this expectation, wild-type spermathecae express a small number of genes at higher levels even than most ribosomal protein mRNAs (Table 2). Eight of the genes (CG17239, CG32834, CG31681, CG32277,CG17012, CG18125, CG9897 and CG17234) encode serine-type peptidases, a class of protein whose regulated activity is known to be important for sperm maturation and fertility(Friedlander et al., 2001; Bloch Qazi et al., 2003). These genes contain candidate signal sequences and their expression in most cases has not been observed outside the female reproductive tract, consistent with the idea that their products are part of a tissue-specific secretion. Three reside in a cluster of five consecutive serine protease genes at 22D5(Table 2), and include the three best currently known examples of genes that are induced by mating(McGraw et al., 2004; Lawniczak and Begun, 2004; Lawniczak and Begun, 2007). RNA in situ hybridization verifies that CG18125 is expressed in spermathecae (Lawniczak and Begun,2007), whereas CG17012 is expressed specifically in spermathecae and parovaria (Arbeitman et al., 2004). CG32834 and CG9897 define a new cluster at 59C1, while CG32277 is the only spermathecae-expressed member of a third serine protease cluster at 63B1. We did not observe spermathecal expression of one previously reported serine protease in the 22D5 cluster, CG17240 (McGraw et al.,2004). None of these genes was altered in expression within Hr39⁰⁴⁴⁴³ spermathecae.

Our microarray study greatly expands the number of genes known to be expressed in spermathecae. Several other abundant gene transcripts encode members of protein classes that have also been implicated in sperm maintenance, including anti-microbial proteins, antioxidants and serpins(Table 2). Surprisingly, all three yolk protein genes (YP1-YP3) are also highly expressed in spermathecae(Table 2) and reproductive tract (not shown), suggesting that these tissues, like fat body(Barnett et al., 1980) and follicle cells (Brennen et al., 1982), contribute to yolk production. Again,the expression of all these abundant gene products with the possible exception of the serpin CG18525 was unaltered in Hr39⁰⁴⁴⁴³spermathecae.

As Hr39 mutation did not affect the most highly expressed genes within the spermathecae, we looked for genes whose levels were significantly reduced in the mutant spermathecae (Table 3) to try and understand why secretory products are strongly reduced in the end apparati of the mutant cells. Two genes, GlcAT-Pand PAPS appear to be particularly strong candidates. Expression of GlcAT-P, encoding a putative N-acetyllactosamineβ-1,3-glucuronosyltransferase (Kim et al., 2003) was reduced to one thirtieth of its original levels. This gene has been implicated in glycoprotein, glycosphingolipid and proteoglycan biosynthesis. Expression of PAPS synthetase, an essential step in sulfur metabolism, was reduced to one twenty-fifth of its original levels. PAPS is required for the production of sulfated proteins, proteoglycans and lipids. Paps mutations abolish mucus production in the embryonic salivary gland(Zhu et al., 2005), indicating the importance of the gene in this secretory tissue. In addition, expression of sytIV gene, a gene that functions in synaptic vesicle exocytosis,is entirely dependent on Hr39. Vesicle exocytosis may be needed for secretion from spermathecal cells. The expression of several other genes that may be involved in carbohydrate or lipid metabolism were also reduced to less than one-tenth of their original levels(Table 3).

A major mechanism of SF1 action is the modulation of levels of steroid-modifying cytochrome P450 genes, which leads to changes in the identity and levels of steroid hormones(Val et al., 2003). Our data shows that Hr39 controls the expression of six cytochrome P450 genes in the spermathecae (Tables 3, 4). Cyp4p2 and Cyp6a14 are each downregulated more than 10-fold. Multiple closely related cytochrome P450 genes exist in Drosophila and humans, and each may be able to act on a wide variety of substrates; hence, it is not possible to predict the biochemical consequences of these changes. The most closely related human gene to Cyp6a14, for example, is CYP3A4, a gene expressed in liver and prostate that can oxidize a variety of small molecules, including steroids. Interestingly, Cyp4g1and Cyp6a17, genes that are closely related to Cyp4p2 and Cyp6a14, are strongly increased in expression by Hr39mutation (Table 4). Another potential P450 pair is Cyp312a1, whose expression is decreased 2.6-fold, and Cyp305a1, whose expression is increased 3.4-fold by Hr39 mutation. Thus, the effect of Hr39 mutation is to modulate the expression levels of specific members of the closely related cytochrome P450 gene families in the spermathecae.

When the genes that are highly regulated in reproductive tract(Table 4) were compared with those in the spermathecae (Table 3), additional evidence of an SF1-like mode of action was observed. Cyp4p2 is also downregulated in reproductive tract(14-fold), but in this tissue, Cyp4d21, rather than Cyp4g1,is strongly upregulated (27-fold) to become the Cyp gene with the highest level of expression in this tissue. Also known as sex-specific enzyme 1 (sex1), Cyp4d21 is normally expressed selectively in the fat cells within male but not female heads(Fujii and Amrein, 2002). The expression of another such gene, Opb99b, is also induced by Hr39 mutation (Table 4). The largest gene expression increase we observed (84-fold) was of takeout (to). Like Cyp4d21 and Obp99b,to is normally expressed in male heads, where it plays a role downstream from the somatic sex-determination genes doublesex and fruitless in integrating nutritional status and circadian cycle with courtship behavior (Dauwalder et al.,2002; Kadener et al.,2006). The requirement for to is autonomous to fat body,and To protein, part of a family of juvenile hormone-binding proteins, is secreted and circulates in the hemolymph of males but not females(Lazareva et al., 2007). These changes in gene expression were verified by quantitative RT-PCR and similar changes were observed in other Hr39 alleles(Table 5). Our observations show that Hr39 not only activates spermathecal secretion, a female characteristic, but represses production of male-specific secretory proteins in the female reproductive tract.

These studies show that the nuclear receptor encoded by Hr39 is not a redundant gene, but is essential for the development of spermathecae and parovaria. Previously, a genetic requirement for this gene was not detected through studies of the Hr39^k13215 allele(Horner and Thummel, 1997). Although, the Hr39^k13215 mutation reduces Hr39expression in adults, its effects on spermathecal and parovarial development were the weakest of any studied Hr39 allele. Differences between the alleles, which probably result from the insertions blocking promoter access to multiple enhancers located in the first two introns and from disrupting splicing, were useful in practice. Although no single allele was completely null for Hr39 function, Hr39^ly92 appeared close to null for spermathecal development and Hr39⁰⁷¹⁵⁴ was close to null for adult function. Additional insight into Hr39function will probably require analyzing double mutants with ftz-f1. The closely related FTZ-F1 protein may be expressed in tissues where loss of HR39 did not cause a detectable phenotype, such as in developing ovarian follicles.

Clearly, the most sensitive tissue requiring Hr39 function is the anterior genital disc at the time of metamorphosis. Previous studies have localized the primordial of both spermathecae and parovaria in this region and documented the rapid growth, migration, eversion and differentiation of spermathecal and parovarial cells during the first 18 hours after the prepupal molt (Anderson, 1945; Keisman et al., 2001). Either autonomously or non-autonomously, our studies show that these events depend in a dose-sensitive manner on Hr39 gene action. All of the phenotypic effects we observed could be explained if the amount of an Hr39-dependent product influenced the number (and/or behavior) of progenitor cells in a spermathecal field that arises during early pupal development, with excess cells leading to additional spermathecae and cell deficits leading to smaller abnormal glands. The regulation, as well as the timing, of spermathecal and parovarial development appear to be closely connected, as evidenced by their common expression and requirement for the lozenge transcription factor(Anderson, 1945). Among all the female genital disc derivatives, parovaria are unique in arising from the otherwise male-specific A9 segment(Keisman et al., 2001) and this may somehow result in the special Hr39 requirement for the development of both tissues. Fortunately, these developmental issues did not detract from the usefulness of the Hr39 alleles in studying the roles played by spermathecae, parovaria and Hr39 in female reproduction.

The data reported here strongly argue that spermathecae and parovaria are redundantly required for female fertility owing to their production of a secretory product that acts throughout the female reproductive tract. Fertility correlates strongly with the number of spermathecae(Table 1), arguing that it is the presence of this tissue rather than some other defect in the Hr39mutants that is responsible for their reduced fertility and fecundity. Moreover, our demonstration that the spermathecae that do form in mutant animals are frequently still defective in secretion, and that Hr39⁰⁴⁴⁴³ mutant spermathecae lack secretion entirely and have parovaria with increased secretory activity, all support this conclusion. The observation that at least one major serine protease, CG17012, is expressed in both tissues (Arbeitman et al., 2004) provides one example of this redundancy.

Many steps are required before the gametes produced by the ovary and testis can undergo successful fertilization. After mating, sperm are introduced into the female reproductive tract along with dozens of proteins (Acps) that mediate sperm storage and behavior, and can even reduce female lifespan(reviewed by Bloch Qazi et al.,2003). Multiple Acps undergo proteolytic processing within the female reproductive tract, and seminal fluid contains serine proteases and protease inhibitors (serpins) that may interact with female-produced factors to regulate this process. At least seven Acps, including four serpins, enter the sperm storage organs after mating (see Lawniczak and Begun, 2007). For example, the male-produced Acp36DE, which is required for efficient sperm storage (Tram and Wolfner,1999), can be found in the spermathecae and is proteolytically processed after transfer to the female(Neubaum and Wolfner, 1999). The serpin encoded by Acp62F, which is required for fertility, enters the spermathecae (Lung et al.,2002). The many spermathecal secretory proteins we identified,including at least eight serine proteases and a serpin, are candidates for the female factor in these interactions. Consistent with this idea, some spermathecal serine protease genes are induced by mating(McGraw et al., 2004; Lawniczak and Begun, 2004; Lawniczak and Begun, 2007) and undergo rapid selective evolution(Lawniczak and Begun,2007).

Our experiments show that the spermathecal and parovarial secretion acts after sperm have been transferred to the female reproductive tract and successfully stored. Hr39 mutant females lacking spermathecae still mated successfully and stored normal amounts of sperm in their seminal receptacles, yet they were sterile in the absence of a spermatheca. This implies that the secretion normally mixes with sperm in the reproductive tract and acts to make them fertilization competent regardless of their eventual storage site. It is unclear why these results differed from studies based on lozenge mutations that suggested a spermathecal requirement for efficient sperm storage (Anderson,1945; Boulétreau-Merle,1977). It is possible that, in the absence of spermathecae and parovaria, the processing of Acps and of sperm is altered or slowed. These defects must not prevent storage, but the resulting sperm may remain incapable of fertilization.

These studies suggest new parallels between Drosophila and mammalian reproductive biology. Following completion of their development within the testis, mammalian sperm move through the lumen of the epididymis,where they undergo a complex process of maturation. Epididymal cells secrete proteases, protease inhibitors, antioxidants, anti-bacterial proteins and other molecules into the epididymal fluid, and they also take up and modify or degrade materials shed by sperm (reviewed by Cooper and Yeung, 2006). Drosophila sperm are exposed to similar classes of molecules after transfer to the female and storage in the spermathecae or seminal receptacle. Thus, the spermathecae and parovaria may play a similar role to that carried out by the caudal epididymis, where under the influence of products secreted by epididymal cells, sperm become motile, fertilization competent and can be stored for long periods. It is possible that the final steps of maturation can be accomplished in the reproductive tracts of either sex, but that some advantage exists in carrying them out at the storage site.

Several studies have been carried out on the genes expressed in the epididymis (Jelinsky et al.,2007; Johnston et al.,2007). These include antioxidant glutathione peroxidases, which are thought to protect against the peroxidation of polyunsaturated fatty acids within sperm plasma membranes (reviewed by Drevet, 2006). Drosophila spermathecae express the similar genes (Prx6005, PHGPx, GstS1 and CG1633). Two genes comprising the `polyol' pathway are found to be associated with membranous vesicles in the epididymal fluid known as`epididymosomes' aldose reductase and sorbitol dehydrogenase(Frenette et al., 2006). Sorbitol dehydrogenase 2 is expressed in spermathecae and its transcript level falls 19 times to undetectable levels in Hr39 mutants. Whether any of these genes carries out an important function in the spermathecae remains to be tested genetically.

Mammalian sperm are motile, but still not fully fertilization competent when they leave the epididymis. In the female they continue to interact with maternal products, such as the mucins that line the reproductive tract and retard movement, as well as other products secreted by the reproductive tract epithelia and the specialized glands it contains, such as Bartholin's gland(reviewed by Suarez and Pacey,2006). In addition to secreting molecules that assist in sperm maturation and preservation, our studies show that spermathecae expressed genes involved in carbohydrate and lipid metabolism, including two genes, GlcAT-P and PAPS, that are strongly associated with glycoprotein, sulfoprotein and lipoprotein secretion. Products dependent on these genes may enter the reproductive tract, especially at the anterior uterus where the spermathecae and parovaria connect to the reproductive tract. This is the region that sperm must traverse en route to the micropyle of the egg and fertilization. How this process occurs remains almost completely unknown. However, the presence of specific glycoproteins, glycolipids,sphingolipids and sulfated molecules might facilitate this final step, and ensure that sperm arriving at the micropyle are fully capacitated for fertilization, which, in Drosophila, must be very highly efficient. Thus, the fertility-essential functions of the spermathecae lie in its secretion rather than in sperm storage, a view consistent with the presence of a separate sperm-storage organ and the independent evolution of spermathecae from these structures (Pitnick et al.,1999).

The similarities between Hr39 expression and function in Drosophila and SF1 in mammals suggest that these genes play roles that at least in part have been conserved during evolution. The expression of Hr39 in reproductive and steroid-producing tissues, in gonadal duct progenitors that develop differentially between the sexes, and in regulating cytochrome P450 genes are all strikingly similar to Sf1 or Lrh1. HR39 function, however, appears to be confined to female development. Male Hr39 mutants were viable, fertile and apparently normal. Indeed, the major function of the gene is in the development of spermathecae and parovaria. Hr39 is also likely to control gene expression within spermathecae in adults, based on the specific gene expression defects observed in Hr39⁰⁴⁴⁴³ spermathecae. This is analogous to SF1 and steroid hormone-dependent production of numerous products throughout multiple mammalian reproductive tissues.

Further evidence that Hr39 has not simply evolved a new role in controlling the spermathecae was our observation that Hr39 mutant females turn on male courtship genes. Expression of Cyp4d21, takeoutand Obp99b are normally undetectable in the reproductive tracts of wild type females, but all three are expressed in the fat body of male heads. Our studies suggest that Cyp4d21 expression might control production of a male specific steroid in the fat body that is responsible for inducing the other genes. This pathway might have been retained from a time when Hr39 played a wider role in controlling reproduction in both sexes. Perhaps a wider role for a conserved regulatory pathway will be uncovered by examining the effects of removing both ftz-f1 and Hr39 at various times during development. However, even if the role of Sf1-like genes is much more limited in Drosophila than in mammals, the finding of any conservation has important implications for our understanding of the evolution of sex-determination mechanisms(Marin and Baker, 1998).

Our observations that Hr39, like Sf1, controls the expression of a small set of cytochrome P450 genes, raises the issue of whether it might act by mediating the production of steroids other than ecdysone and 20-OH ecdysone. Many other steroids have been found in Drosophila and other insects, but none has been clearly implicated in sex-specific reproductive functions (reviewed by De Loof et al., 1998). By defining specific biological functions and specific target Cyp genes, it will now be easier to further investigate the mechanism of Hr39 action,and to determine whether it involves the production of new steroid derivatives. Such studies have the potential to significantly deepen our understanding of how reproduction is regulated and how this regulation evolved.

The authors thank Carl Thummel for providing the Hr39^k13215 stock, Mike Sepanski for assistance with electron microscopy, and members of the Spradling laboratory for comments on the manuscript and helpful discussions. A.C.S. is an Investigator of the Howard Hughes Medical Institute.