Drosophila Pxt: a cyclooxygenase-like facilitator of follicle maturation

Prostaglandins (PGs) are locally synthesized and locally acting hormones that mediate a wide variety of biological activities, ranging from inflammation and pain to reproduction and cancer(Funk, 2001). Cells generate PGs from arachidonic acid via the action of cyclooxygenase (COX) enzymes,which, by two distinct catalytic activities, cyclooxygenation and peroxidation, generate the PG intermediate, PGH₂. The two subclasses of vertebrate COX enzymes, Cox1 and Cox2, differ in their structure, susceptibility to inhibitors and biological function. Specific PG synthases act on PGH₂ to synthesize additional PGs, including PGD₂, PGE₂ and PGF_2a. PGs are ligands for G-protein-coupled receptors (GPCRs) and possibly for nuclear hormone receptors, thereby regulating critical signaling cascades(Bos et al., 2004; Gilmour and Mitchell, 2001). However, the specific downstream events mediated by PG signaling remain unclear. Many pharmaceuticals have been designed to block COX enzyme catalytic activity and, therefore, PG synthesis(Funk, 2001; Taketo, 1998).

In a variety of organisms, prostaglandins regulate female reproduction,including follicle maturation, ovulation, fertilization, maintenance of pregnancy, and induction of labor(Langenbach et al., 1999; Loftin et al., 2002). Cox2 knockout mice are defective in all aspects of female reproduction(Lim et al., 1997), while loss of Cox1 blocks labor. PGs and PG synthetic activity has also been found in many invertebrates (Stanley,2006; Stanley-Samuelson and Pedibhotla, 1996), including Drosophila(Pages et al., 1986). PG levels in these animals frequently rise upon mating, while exogenous PGs stimulate both egg development and laying(Stanley, 2006; Stanley-Samuelson and Pedibhotla,1996). However, the molecular mechanisms of PG signaling during oogenesis have not been established.

The Drosophila ovary is composed of ovarioles containing sequentially maturing follicles termed egg chambers(Spradling, 1993). New follicles are continuously produced from stem cells in the anterior region or germarium. Groups of 16 interconnected sibling germline cells differentiate into 15 nurse cells and an oocyte before being surrounded by epithelial cells to form follicles. Subsequently, follicles mature through 14 morphological stages (S1-14), during which they increase greatly in size (S1-10), take up yolk proteins from the hemolymph (S8-10), and dump the nurse cell contents into the oocyte (S11). Follicles subsequently undergo a complex process of maturation that involves the programmed migration of specific subpopulations of follicle cells, including the border cells, centripetal cells and dorsal appendage cells that secrete and shape distinct parts of the protective eggshell.

We have used pharmacology and genetics to explore the roles of PGs during Drosophila oogenesis. By using an in vitro egg maturation assay, in which S10B egg chambers mature to S14s in culture, we find that Drosophila egg maturation requires a Cox1-like activity. Genetic studies of the COX-like peroxidase Pxt indicate that it functions upstream of PGs to regulate actin during nurse cell dumping. These studies provide the first evidence that PGs mediate Drosophila oogenesis, making this a valuable system with which to elucidate the molecular mechanisms by which prostaglandin signaling acts during reproduction.

Fly stocks were maintained at 20-25°C on standard cornmeal-agar-yeast food. Wet yeast, dry active yeast reconstituted to a thin paste with water,was provided for all ovary analyses including IVEM, immunofluorescence and in situ hybridization. y¹ w¹ was used as the control in all experiments. EY03052 (Bellen et al., 2004) was obtained from the Bloomington Stock center and f01000 (Thibault et al., 2004)was obtained from the Harvard Exelixis collection.

S10B follicles were hand dissected in room temperature freshly made IVEM media composed of Grace's medium (BioWhittaker, Cambrex) with heat-inactivated 10% fetal bovine serum (Invitrogen) and 1× penicillin/streptomycin glutamine (100×, Gibco). Dissected S10Bs were transferred to fresh IVEM media while dissection continued and were then transferred into 24-well tissue culture plates with 1 ml of IVEM media plus COX inhibitors and PGs as indicated and allowed to develop. For each experiment, controls were treated with the same volume of ethanol and/or DMSO used for the experimental conditions. S10Bs were allowed to develop for 10 or more hours and then stages were scored. Those chambers not undergoing nurse cell dumping (s10-11) were scored as stalled, while those mostly through completely dumped were scored as developed. Each experiment was performed at least in duplicate and standard deviations are shown, except that standard error is presented for experiments with three or more replicates. COX inhibitors were obtained from Cayman Chemical, except indomethacin (Sigma). U44069, the stabilized form of PGH₂, Fluprostenol and PGF_2a were from Cayman Chemical. All other prostaglandins were from Sigma. Live time-lapse movies of S10Bs were generated using a Zeiss Stereolumar, images were taken every 10 minutes for 10 hours.

Fertility was determined by mating a known number of females with y¹w¹ males, in the case of the pxtalleles, and heterozygous sibling males for the rescue experiments. The flies were allowed to mate for 3-4 days and then sequentially transferred to new vials over a period of 2-3 weeks. All transfers were kept and the progeny were counted 20 days after mating.

Full-length pxt cDNA, LD43174, and mouse Cox1 cDNA, B0811H12, were cloned into the Gateway entry vector (Invitrogen), and swapped into a pUASp vector (a kind gift from Terence Murphy), thereby placing the cDNA expression under the control of the yeast upstream activating sequence (UAS) modified to enhance germline expression. Mouse Cox1 was also cloned by the same means without the starting ATG and placed into the pUASp Venus vector (kind gift from Terence Murphy), generating an N-terminal Venus fusion. P-element transformation was performed by Genetic Services (Cambridge, MA). Multiple insertion lines were generated for each element. Chromosomal locations of the P-elements were determined by standard crosses. Insertions on the 2nd chromosome were crossed to pxt^f01000 to obtain a stock of UASp/balancer;f01000/balancer. These flies were crossed to hsGAL4 or c587;Sco/CyO;f01000/balancer to generate GAL4;UASp;f01000/f01000. hsGAL4 flies were incubated at 37°C (air) for 1 hour, on three consecutive days. A known number of females were then allowed to lay eggs in a new vial over the next 4 days. Fertility/female was assessed by counting the number of adults eclosing by day 19.

Total RNA was isolated from fattened ovaries (15-20 pairs of ovaries) or whole flies (15-20 females) using either TriZOL reagent (Invitrogen) or Qiagen RNeasy kit following the manufacturers' directions. RNA was treated overnight at room temperature with 2U RNase-free DNase (Ambion). Concentration was quantified using UV spectrometry. RNA (1 μg) was used to generate cDNA via the Reverse Transcription System (Promega) using random primers and following manufacturer's directions. PCR was then performed using 2 μl or 1/10th of the RT reaction with primers to pxt and rps17, as the control. The following pxt primers were used: to generate a ∼750 bp fragment starting from the ATG, 5′pxt rt ATGAGTCGCATTTTATTT and 3′ pxt rt TGTGGTTCTGGATTATTG; to generate a ∼550 bp fragment,5′ pxt 901 GATCGTCCTCATCCTAAGT and 3′ pxt 1390/1450 3′TGGTTTGTTCGGCCATCT; and to generate the control ∼250 bp fragment of rps17, 5′ rps17 CGAACCAAGACGGTGAAGAAG and 3′ rps17 CCTGCAACTTGATGGAGATACC. An MJ Research PTC-100 Programmable Thermal Controller was used to amplify 50 μl reactions with GoTaq Flexi DNA polymerase and 2.5 mM MgCl₂ with the following program: 2 minutes at 95°C, 30 cycles of 30 seconds at 95°C, 30 seconds at 42°C, 1 minute at 72°C, followed by 5 minutes at 72°C. Products were resolved on 0.8%agarose LE gels (Roche) in 1×TAE buffer with 0.25 μg/μl ethidium bromide and imaged with BioRad Gel Doc XR Scanner and Quantity One software(V4.5.2). Experiments were carried out at least in triplicate and representative data are shown. Real time quantitative RT-PCR reactions were performed on ovary RT reactions on an Opticon Monitor 2 (MJ Research) using a 50 μl reaction comprising 2 μl RT and 0.5 μl of a 7.5× SYBR Green stock (Molecular Probes). The program used was 2 minutes at 95°C, 35 cycles of 30 seconds at 95°C, 30 seconds at 42°C, 1 minute at 72°C, followed by 5 minutes at 72°, and a melting curve of each sample was determined. Results were analyzed using the Opticon Monitor software. Transcripts were expressed relative to the level of the control rps17gene and normalized to the control ovaries. Each reaction was carried out in triplicate on at least two separate RNA samples.

Whole-mount in situ hybridization was performed as previously described(Lecuyer et al., 2008) using an antisense RNA probe to full-length pxt cDNA generated by in vitro T7 transcription from HindIII-digested LD43174.

Whole-mount samples were fixed with 4% paraformaldehyde for 15 minutes and processed using standard procedures (Cox and Spradling, 2003). The following antisera from the Developmental Studies Hybridoma Bank were used: mouse anti-Hts (1B1, 1:20) and mouse anti-Fas3 (7G10, 1:50). Goat anti-mouse conjugated to Alexa 488(Molecular Probes) was used at 1:2000. Rhodamine phalloidin (1:200;Invitrogen) was added to both the primary and secondary antibody incubations. To visualize DNA, DAPI (1 μg/ml) was added to the final wash before mounting the samples in Vectashield (Vector Labs).

DIC imaging was performed on a Zeiss Axiophot microscope with a Qimaging RETIGA 1300 camera and QCapture software. Time-lapse microscopy was performed on Zeiss Stereolumar microscope and acquired with AxioVision software. Static confocal images were taken with either a 20×(NA 0.7) or a 63×(NA 1.32) PlanApo lens on laser-scanning confocal microscope (SP2 or SP5; Leica). All confocal images are projected z-stacks.

Drosophila S10B egg chambers can reliably complete development in vitro (Fig. 1A-C), termed in vitro egg maturation (IVEM, Fig. 1D, Movie 1), a property they share with mammalian follicles(Sirois et al., 1992). We took advantage of this capability to investigate whether a DrosophilaCOX-like protein functions during late oogenesis by asking whether COX inhibitors interfere with in vitro follicle maturation. The effect of each inhibitor on follicle development was tested at multiple concentrations and the IC₅₀, the concentration at which half of the egg chambers do not mature in vitro, was determined (Table 1 and supplementary material Fig. S1). We define not maturing or`stalled' as incomplete nurse cell dumping and breakdown, morphologically resembling stages 10-11. Nurse cell dumping is a prerequisite for normal maturation, and, when disrupted, it leads to smaller oocytes and abnormal dorsal appendages. Several Cox1-specific inhibitors block egg chamber maturation in a dose-dependent manner(Table 1 and Fig. 1D-I), but most Cox2 inhibitors have little or no affect on maturation(Table 1). NS-398, a preferential Cox2 inhibitor, is an exception, as it blocks maturation but at a higher concentration known to inhibit Cox1 function(Table 1, Fig. 1J). The IC₅₀concentrations we observed are slightly higher than what is seen in cell culture experiments, most probably owing to lower drug penetration into a multicellular tissue. Thus, pharmacological inhibition of Cox1 activity disrupts egg maturation (Fig. 1I,J; Movie 2).

If the COX inhibitor-induced block in egg maturation is due to the loss of PG synthesis then addition of exogenous PGs should restore development. A complicating factor is that PGs are locally acting, anionic lipids and therefore do not survive long in an aqueous environment. To circumvent this problem, stabilized analogs of PGs were tested. Addition of a stabilized form of the PG intermediate, PGH₂, which in mammals can serve as a substrate for endogenous PG synthases, partially rescues the aspirin- or NS-398-mediated block to egg maturation(Fig. 1K,L). The incomplete development of PGH₂-rescued egg chambers may be due to the uniform application of the PG, as endogenous PGs are likely to be produced in a spatially and temporally restricted manner.

PGH₂ can be converted by specific PG synthases into the individual classes of PGs that activate distinct signaling cascades. To determine which specific PG(s) mediate egg chamber development we examined the exogenous effects of the individual PGs. Most exogenous PGs, except PGH₂ (the precursor PG) and PGF_2α, inhibit egg chamber development after stage 12 (see Fig. S2 in the supplementary material). This suggests that during S10-14 egg chamber development PGH₂ is processed into PGF_2α, and not the other PGs, to facilitate maturation. Consequently, we tested whether exogenous PGF_2α can rescue COX inhibition. Both PGF_2α(not shown) and its stabilized analog Fluprostenol (Flu) significantly rescue the aspirin or NS-398-induced block in maturation(Fig. 1L). We conclude that COX inhibitors block cultured Drosophila egg chamber development at expected concentrations and this inhibition is significantly overcome by supplying exogenous PGs. These observations strongly suggest that a COX-like enzyme and PGs play a physiological role during late follicle development in Drosophila.

We used BLAST analysis to identify the pxt gene (AY119616) as a candidate Drosophila COX. Mammalian COX enzymes are thought to have evolved from heme-dependent peroxidases such as myeloperoxidase by acquiring a cyclooxygenase activity center (Garavito and Mulichak, 2003; Picot et al., 1994). Drosophila Pxt is such a heme peroxidase,with the conserved catalytic residues needed for heme coordination. Within the 362 amino acids of the homology region, Pxt is 38% similar and 24% identical to vertebrate Cox1 enzymes, the closest such match among Drosophilaproteins. Like mammalian COX enzymes (Otto et al., 1993), Pxt is predicted to be membrane bound, as the first 18 amino acids encode a signal peptide, and to be glycosylated (PROSITE analysis) (Hulo et al.,2006).

Mutations of pxt have been recovered in large insertional mutagenesis screens that allow the function of this gene to be probed. We studied two such alleles: a P-element insertion EY03052 and a Piggybac insertion f01000 (Fig. 2A). The pxt^EY03052 insertion is located 17 bp upstream of the predicted 5′ UTR, while pxt^f01000 contains an insertion within the 5′ UTR, 38 bp upstream of the starting methionine. Reverse transcriptase-PCR (RT-PCR and qRT-PCR) indicates that both insertions substantially reduce the level of pxt RNA in the ovary(Fig. 2B) and whole fly (data not shown). However, whole-mount in situ hybridization (see Fig. S3 in the supplementary material) indicates that pxt^EY03052 exhibits some mislocalized and misregulated pxt expression, while pxt^f01000 exhibits almost a complete loss of expression.

Like genetic loss of mammalian COX enzymes(Lim et al., 1997; Stanley, 2006; Stanley-Samuelson and Pedibhotla,1996), loss of pxt affects female fertility. Homozygous pxt^f01000 females are sterile, while the fertility of pxt^EY03052 females drops with age more rapidly than normal(Fig. 2C). pxt mutant follicles frequently show defects in nurse cell dumping(Fig. 2D) that resemble wild-type chambers developed in vitro in the presence of COX inhibitors(Fig. 1F). In 4-day-old pxt^f01000females, 93% of S14 follicles are short in length and failed to undergo complete nurse cell dumping (n=131). pxt^EY03052 and pxt^EY03052/pxt^f01000 mutant follicles display these same phenotypes but with lower penetrance and less severity(data not shown), in agreement with the molecular data suggesting that pxt^EY03052 is a weaker allele. Thus, both genetic loss of pxt and COX inhibition cause a block in nurse cell dumping.

Whole-mount in situ hybridization was used to determine the spatial and temporal patterns of pxt expression, and therefore the likely sites of PG synthesis, during oogenesis. pxt RNA is present at low levels in the germarium and early follicles (Fig. 2E). Expression begins to increase in S6 in both somatic and germline cells, becomes progressively stronger during vitellogenesis(Fig. 2E-G, see Fig. S3A in the supplementary material), and is highly expressed in germ cells(Fig. 2F,G) and somatic cells(outlined in Fig. 2G) from S10 throughout the remainder of oogenesis (Fig. 2H,I). A subset of follicle cells, termed border cells (BC),migrates from the anterior of the egg chamber to the nurse cell/oocyte boundary during S9 and exhibit a high level of pxt expression(Fig. 2F,F′, arrow). The high levels of pxt expression in late-stage follicles are consistent with Pxt being the enzyme susceptible to inhibition by mammalian Cox1 inhibitors during in vitro follicle development.

To verify that the sterility of pxt^f01000 females and their egg chamber defects are caused by disruption of pxt, we performed rescue experiments. Expression of a full-length pxt cDNA under the control of the UASp promoter (which allows expression in the germline) using heat-shock GAL4 (HS-GAL4) beginning in adulthood, restored fertility at an average level of 18 progeny/female over the 4 days following heat shock treatment (Fig. 3A). For comparison, pxt heterozygous females produced 26 progeny/female in a similar period, while unrescued mutants produced none. Such rescue proves that the pxt^f01000 phenotype is due to the loss of pxt function, and that the RB isoform alone can provide substantial gene function without being developmentally regulated. When pxt cDNA was expressed using the c587 somatic cell driver, fertility was also restored but to a lesser extent, 5.4 progeny per female in the same interval. Thus,somatic expression of pxt is sufficient to rescue fertility but expression in both germ cells and somatic cells driven by HS-GAL4 is more efficacious. Fertility is also restored by precise excisions of either element(data not shown).

If Pxt acts as a Cox1 enzyme, then the effects of pxt^f01000 mutation on follicle development might largely be due to reduced PG production. To test this possibility, we first examined the ability of pxt mutant S10B follicles to develop in vitro. Approximately 75% of pxt^f01000 stage 10B follicles fail to mature in culture, compared with ∼7% of wild-type controls(Fig. 3B,C). (This differs slightly from our previous measurement of 93% with arrested dumping in vivo because only completely normal-looking S10B follicles were selected for in vitro development.) Addition of either 10 μM PGH₂ or 10 nM Fluprostenol (Flu), the stabilized analog of PGF_2α, greatly improves development (Fig. 3B). In the presence of these PGs, ∼70% of the follicles undergo nurse cell dumping (Fig. 3D). Therefore,Pxt functions in a process required for egg chamber maturation that can be rescued by exogenous PGs, strongly supporting the idea that Pxt acts as a Drosophila COX enzyme to synthesize PGs.

If Pxt functions as a Cox1 enzyme to generate PGs, then it might be possible to rescue pxt function using a mammalian COX protein. Consequently, we also generated transgenic lines expressing mouse Cox1 full-length cDNA (MCOX1, NM008969) under the control of the UASp promoter. Inducing MCOX1 expression using heat-shock GAL4 in adult pxt^f01000 females restores fertility to a level of 19 progeny/female (Fig. 3E), which is as effective as rescue with Drosophila Pxt under the same conditions (Fig. 3A). The S14 defects observed in pxt mutants are suppressed by MCOX1 expression,as nurse cell dumping is completed (Fig. 3G, compare with 3F). Thus, Pxt is homologous to and can be functionally replaced by mouse Cox1; pxt mutants and COX inhibitors block egg maturation, and development can be rescued by exogenous PGs. These observations strongly argue that Drosophila Pxt functions as a Drosophila COX during follicle maturation and acts by generating PGs.

We examined the actin cytoskeleton of S10B pxt egg chambers to investigate how loss of pxt and PG production interferes with nurse cell dumping. At the onset of nurse cell dumping, the regulated formation of non-contractile actin bundles in S10 nurse cells tethers the nurse cell nuclei and prevents them from plugging up the oocyte ring canals(Robinson and Cooley, 1997). Moreover, non-muscle myosin II-mediated contraction of nurse cell subcortical actin provides the contractile force behind dumping and egg chamber elongation(Wheatley et al., 1995; Edwards and Kiehart, 1996). pxt^f01000 mutant follicles contained significantly fewer actin bundles in the nurse cell cytoplasm during S10 compared with wild type(Fig. 4B,B′, compare with 4A,A′). We did not observe nurse cell nuclei in the ring canals of mutant follicles, suggesting that pxt mutation disrupts the actin-based contraction that generates the force for dumping. In support of this, subcortical actin levels are strongly decreased in the mutant follicles(Fig. 4E,E′, compare with 4D,D′). The loss of both subcortical actin and cytoplasmic actin bundles in pxt mutants is likely to be the underlying cause of the nurse cell dumping defects in the mutant.

The actin defects observed in pxt^f01000 females are almost completely suppressed by MCOX1 expression(Fig. 4A,C). Robust actin bundles (Fig. 4C,C′) and subcortical actin (Fig,4F,F′) form in S10B nurse cells in the rescued mutants. The ability of mouse Cox1 to efficiently substitute for Pxt during oogenesis strongly argues that production of PGs or other Cox1 products are essential for actin-based contraction and actin bundle formation during nurse cell dumping. However, the mechanism(s) by which, either directly or indirectly,COX activity is required remains to be determined.

To verify that prostaglandins play a role in modulating actin during nurse cell dumping, we cultured S10B follicles with or without COX inhibitors,allowed them to mature to stage 12, and examined actin structures by phalloidin staining (Fig. 4G-I). (Prolonged incubation in vitro interferes with normal fixation of the large S10B cytoplasmic actin fibers, so we focused on subcortical actin.) COX inhibition, either with aspirin or NS-398, results in a loss of nurse cell subcortical actin and causes large actin aggregates to form (Fig. 4H, compare with G,and data not shown). Addition of exogenous PGH₂ results in a more normal pattern of actin organization (Fig. 4I). These data argue that PGs rather than some other product of the Cox1-like activity of Pxt are needed for normal actin structures during s10-11.

We also used pxt mutants to look for possible roles of PGs during the earlier stages of oogenesis. Drosophila ovarioles normally contain 6-7 egg chambers that regularly increase in developmental age(Fig. 5A). By contrast,ovarioles from pxt mutant females typically contain only 3-5 egg chambers (Fig. 5B), and are deficient in maturing follicles (pxt^f01000 contain 80%fewer S10s/ovariole; see Fig. S4 in the supplementary material). This is probably due to follicle instability because 25% of pxt^f01000 ovarioles contain a fusion of two adjacent egg chambers and in 58% of ovarioles there is at least one degenerating follicle(see Fig. S5 in the supplementary material). pxt^f01000nurse cell plasma membranes are unstable and frequently break down to produce multinucleate nurse cells (Fig. 5C, nuclei are outlined in white, compare with S8 in Fig. 5A), in which an actin-rich aggregate containing ring canal remnants remains (arrows in Fig. 5C). The multinucleate cells are probably caused by membrane breakdown and not cytokinesis defects,because this phenotype is not seen in the early stages and begins to arise around S8. The membrane breakdown, fusion and degeneration defects increase with age, because in 25-day-old pxt females, the ovarioles are reduced to a single fused structure (Fig. 5D).

Another striking defect was observed during border cell migration, a process requiring dynamic cytoskeletal and adhesion changes(Montell, 2003). In both homozygotes and transheterozygotes of pxt^EY03052 and pxt^f01000, more than 50% of S9-10 border cell clusters to not complete migration normally. Some mutant cells reach the oocyte, whereas others form a trail of lagging cells along the migration path (outlined in Fig. 5E, compare with Fig. 5A, outlined and labeled BC).

Intriguingly, pxt mutant phenotypes within an egg chamber usually decrease in severity along the anteroposterior (AP) axis. Membrane breakdown preferentially occur in anterior nurse cells (outlined in yellow in Fig. 5F). Conversely, posterior S10B nurse cells tend to retain more normal actin structures, including membrane and filaments (outlined in white in Fig. 5F). Interestingly, these observations suggest the possibility that a posterior to anterior gradient of PGs, or a downstream target of PGs, influences factors that regulate nurse cell dumping, membrane stability, and cell migration within the ovarian follicle.

PGs play important and diverse roles in insects and other invertebrates,such as regulating immune responses and reproduction, including oogenesis(Stanley, 2006; Stanley-Samuelson and Pedibhotla,1996). Biochemical studies suggest that invertebrate PGs are produced in a similar manner to those in vertebrates(Stanley, 2000), and mammalian COX inhibitors have been shown to block insect PG synthetic activity in tissue extracts (Buyukguzel et al.,2002; Stanley-Samuelson and Ogg, 1994). However, the genes encoding the proteins that mediate invertebrate PG production have not been previously identified. This has slowed attempts to understand the detailed functions of PGs in Drosophila and precluded the use of genetics to address many important issues about PG signaling. We find that Drosophila S10B-14 follicle development requires a Cox1-like activity that can be satisfied by exogenous PGs, including PGF_2α. Thus, follicle maturation provides a phenotypic focus for further genetics studies of PG action in Drosophila.

Using pharmacology, sequence homology and developmental genetics, we have identified Pxt as a candidate COX enzyme. Pxt is a heme peroxidase, and like mammalian COX enzymes, is predicted to be glycosylated and membrane bound. Despite only moderate homology to known COX enzymes, genetic loss of pxt resembles the effects of COX inhibitors on follicle development in vitro. Exogenous PGs can overcome the block in development caused by COX inhibitors and by pxt mutation. Therefore, PGs act downstream or in parallel to Pxt, supporting the idea that Pxt is responsible for the synthesis of PGs. Furthermore, we find that expression of mouse Cox1 in pxt^f01000 mutant adults fully rescues female fertility and follicle development. The fact that mouse Cox1 can functionally replace Pxt during Drosophila oogenesis strongly argues that Pxt is a Drosophila COX enzyme.

Although Drosophila COX and mammalian COX enzymes appear to have diverged over time, they are likely to have the same basic three-dimensional structure within the conserved domain(Garavito and Mulichak, 2003; Picot et al., 1994). The critical residues (Gln203, His207, His388) for heme coordination necessary for the peroxidase activity are conserved in Pxt (Gln396, His402, His590). The conservation is too weak to predict the existence of the hydrophobic substrate-binding channel within the cyclooxygenase active site and the substrate binding residues Arg120 and Tyr355 are not clearly conserved. However, Pxt does contain a candidate Tyr385 (Pxt Tyr564), the residue at the active site that acquires an activating electron from the peroxidase part of the enzyme and is required for the cyclooxygenase activity. This suggests that there may be substantial differences in the structure of Pxt and Cox1, despite the ability of Cox1 to substitute functionally for Pxt.

The limited previous studies of Pxt structure and expression have suggested alternative functions for the protein. Distant homologies exist with peroxinectin (17-31% identity), a crayfish cell-adhesion peroxidase(Altschul et al., 1997; Vazquez et al., 2002). Pxt has also been suggested, by homology to a putative Aedes aegypti chorion peroxidase, to be the peroxidase responsible for eggshell hardening, a process caused by the formation of dityrosine crosslinks(Li et al., 2004; Petri et al., 1976). Although our more detailed expression data and the phenotypes of pxt mutants indicate that Pxt functions as a COX enzyme and is required during oogenesis prior to eggshell formation, we cannot rule out additional functions for the protein. Studies of purified Pxt will be required to address its full range of biochemical and enzymatic properties.

Both pharmacological and genetic studies show that PGs are required for mammalian follicles to mature normally and undergo ovulation. Cox2 knockout mice are defective in both follicle maturation and ovulation(Loftin et al., 2002). In rats, inhibition of PG synthesis results in mistargeted and incomplete follicle rupture. The small percentage of follicles that do ovulate fail to be fertilized, most probably owing to impaired oocyte maturation(Lim et al., 1997). COX inhibitors, such as NSAIDs, cause reversible female infertility and may act in a similar manner (Akil et al.,1996; Pall et al.,2001; Smith et al.,1996). Like many mammals, Drosophila females also ovulate only one mature oocyte at a time, most probably by inducing contraction of the muscles surrounding just one of the more than 30 ovarioles that make up the two ovaries. In addition to their defects in follicle maturation, pxtmutants ovulate only rarely (see Fig. S4 in the supplementary material; T.L.T. and A.C.S., unpublished). Thus, the roles of PGs during follicle development and ovulation may be conserved between Drosophila and mammals.

Each cycle, the single mammalian follicle that will be ovulated, the dominant follicle, accumulates high levels of smooth muscle myosin and actin which contributes to the contractile force needed for ovulation(Gougeon, 1996). These changes in actin organization are hypothesized to be downstream of PG signaling. PGF_2α, the PG that had the most effect on DrosophilaIVEM, is intimately associated with muscle contractions in mammals(Funk, 2001; Langenbach et al., 1999). Consequently, our finding that Pxt and PGs affect some oogenic processes that are modulated by actin and myosin, including border cell migration and nurse cell dumping, represents a potential parallel between the role of PGs in Drosophila and mammalian oogenesis.

Even the induction of PG-dependent maturation may occur in a similar manner. In mammals, PG synthesis is hormonally upregulated during the 10 hours preceding ovulation (Duffy and Stouffer,2001; Murdoch et al.,1993; Sirois et al.,1992), and thereafter mediates the terminal differentiation of follicles (El-Nefiawy et al.,2005). Similarly, we find that pxt levels are upregulated during stages 9-10, and that COX activity, Pxt and PGs are subsequently required to complete the last 10-15 hours of Drosophila follicular development. Interestingly, the steroid hormone ecdysone is known to regulate the S8 oogenic checkpoint, which commits egg chambers to finish developing into mature follicles (Buszczak et al.,1999). Drosophila pxt expression and PG synthesis may be upregulated by ecdysone at the onset of S8 as part of the maturation program.

Many PGs serve as short-range hormones that act as ligands for G protein-coupled receptors (GPCRs). In mammals there are eight receptors with distinct specificities for the active PGs, and each receptor favors the initiation of a specific signaling cascade(Bos et al., 2004). Such GPCR signaling can secondarily modulate additional signaling pathways. We have found that PGF_2α mediates Drosophila egg maturation in vitro. Mammalian PGF_2α acts as a ligand not only for the F receptor (FP), but also for two E receptors (EP₁ and EP₃), thereby increasing the possible signaling outcomes. One known downstream target is protein kinase A (PKA), which can then activate multiple MAPKs (Bos et al., 2004). It should be possible in the future to discern which effects of Pxt are exerted autonomously, and which are downstream of intercellular signals.

Despite a possible plethora of mechanisms, PGs have been frequently found ultimately to affect muscle contraction and ovulation, or to modulate the actin cytoskeleton in mammals. Within many cultured mammalian cell types, PGs cause changes in the organization, stability and polymerization of actin that influence membrane permeability and cell motility(Banan et al., 2000; Sheng et al., 2001). PG action appears to be rapid, suggesting that sophisticated feedback circuits may be involved. For example, reduction in the level of actin microfilaments has been reported to stimulate PG synthesis and release(Wang and Hatton, 2006). Thus,PGs may facilitate fine-tuning of the actin cytoskeleton in a rapid time frame.

Our studies show that Pxt and PGs can affect the actin cytoskeleton in Drosophila. Actin filaments are greatly reduced in pxtmutant stage 10B follicles, which phenotypically resemble follicles lacking major components of the actin cytoskeleton(Robinson and Cooley, 1997). PG-mediated regulation of actin dynamics is currently not well understood,particularly when intercellular interactions are involved. Drosophilaoogenesis provides an attractive system to elucidate how PGs modulate actin in the context of developing tissues.

Our data suggests that disruption of PG signaling has a graded effect within the developing follicle. The actin-related defects in pxtmutants are more prevalent at the anterior of the egg chamber, including nurse cell membrane loss and accumulation of actin puncta. Conversely, actin bundles are more likely to form in a normal fashion in the nurse cells closest to the oocyte. A PG signaling gradient might be relevant to many incompletely understood biological differences that exist along the AP follicle axis, such as the gradient of nurse cell ploidy, the temporal gradient of vitelline membrane formation within the oocyte, the directional migration signal for border cells and the programmed reorganization of the oocyte cytoskeleton in mid-oogenesis that underlies patterning. Ultimately, by using Drosophila genetics, it may be possible to better understand how PGs act, providing new insights and therapeutic targets for the widely conserved biological processes they influence.

The authors thank the members of the Spradling, Farber and Yanowitz laboratories for critiques of the manuscript, the Spradling and Farber laboratories for helpful discussions, Vanessa Damm for assistance with FISH,Sheryl Southard and Julio Castaneda for their rotation work using IVEM, Mahmud Siddiqi for assistance with time-lapse imaging, and Rafael Villagaray for assistance with Quicktime movie generation. A.C.S. is an Investigator of the Howard Hughes Medical Institute. T.L.T. is supported by NIH 5F32CA119676.