The Drosophila G-protein-coupled receptor kinase homologue Gprk2 is required for egg morphogenesis

Patterning a Drosophila oocyte requires the highly integrated activities of 15 interconnected nurse cells and an overlying layer of epithelial follicle cells (reviewed by Spradling, 1993; Ray and Schüpbach, 1996). Signaling events between follicle and germline cells precede the initial formation of new egg chambers (stage 1) and continue throughout development of the mature egg (stage 14). Using maternal information, the anteroposterior axis is specified very early, prior to egg chamber formation. hedgehog (hh) production by small groups of somatic cells at the tip of the ovary plays a role in establishing this axis, and in controlling egg chamber assembly (Forbes et al., 1996a). A likely Hedgehog receptor, encoded by the gene smoothened (smo), is a protein homologous to mammalian G protein-coupled serpentine receptors (Alcedo et al., 1996; van den Heuvel and Ingham, 1996). The dorsal-ventral axis is laid down much later, following the relocation of the oocyte nucleus to the future dorsal side. The TGF-α family member, gurken, is expressed locally around the oocyte nucleus and binds to the Drosophila EGF-receptor homologue, Torpedo, on the surface of the overlying dorsal follicle cells. This initiates a signaling pathway that defines embryonic dorsal-ventral polarity (Chasan and Anderson, 1993; Neuman-Silberberg and Schüpbach, 1993).

The Gurken signal not only establishes the dorsal-ventral axis, but it also participates in patterning the anterior end of the egg. Eggs contain specialized structures in this region, including a micropyle (to allow sperm entry), two dorsal appendages (to allow embryonic respiration), and an operculum (to allow the mature embryo to escape from the egg shell). These structures are shaped by special subsets of migrating follicle cells. The ‘centripetally migrating’ cells produce the operculum and collaborate with the ‘border’ cells to make the micropyle. A distinct subset of ‘dorsal appendage’ follicle cells migrate anteriorly where they lay down the tubular dorsal appendages. At least one other major signaling pathway, involving the TGF-β family member decapentaplegic (dpp), also influences anterior eggshell production by these cells (Twombly et al., 1996). dpp is expressed by a set of 40-50 follicle cells that overlie the nurse cells and by the centripetally migrating follicle cells. Mutations reducing the activity of dpp or of the Dpp receptors encoded by saxophone (sax) and thick veins (tkv) cause eggs to be produced that have shortened dorsal appendages and other anterior defects. Overexpression of dpp causes an expansion of these anterior structures (Twombly et al., 1996).

We have identified a new mutant, gprk2⁶⁹³⁶, that produces anterior eggshell defects that resemble those seen in flies with reduced levels of Dpp signaling. The gene disrupted in this mutant encodes a putative member of the family of kinases that phosphorylate G protein-coupled receptors. Activation of G protein-coupled receptors leads to the dissociation of G proteins into α and βγ subunits, which in turn activate effector molecules including adenylyl cyclase and protein kinase A, phospholipases, ion channels, and the Ras/MAP kinase pathway (Gilman, 1987; Logothetis et al., 1987; Birnbaumer, 1992; Luttrell et al., 1995; van Biesen et al., 1996). G protein-coupled receptor kinases (GRKs) rapidly downregulate these pathways by phosphorylating activated receptors at multiple sites which in turn allows the binding of proteins of the arrestin family. Arrestin binding uncouples the receptor from its associated G protein (Wilden et al., 1986) and contributes to resensitization of receptors through sequestration and recycling (YMénard et al., 1996; Yu et al., 1993)

The GRK family contains six members from mammals (reviewed in Premont et al., 1995), two from Drosophila (Cassill et al., 1991), and one whose existence is suggested from sequencing of the C. elegans genome (accession no. U22833). The biochemistry and cell biology of GRKs have been determined primarily through analysis of the original members of the family: rhodopsin kinase and β-adrenergic receptor kinase (βARK; Benovic et al., 1986; Palczewski et al., 1988). All of the mammalian family members phosphorylate agonist-bound G protein-coupled receptors, but which receptors they activate and the role of desensitization in vivo is not known with certainty. The Drosophila GRKs (GPRK1 and GPRK2) were identified by homology to βARK1 (Cassill et al., 1991) but no mutations had been previously identified. In addition to smoothened, several other Drosophila genes have been identified that encode putative G protein-coupled receptors (Shapiro et al., 1989; Saudou et al., 1992; Arakawa et al., 1990; Li et al., 1991, 1992; Monnier et al., 1992; Gotzes et al., 1994; Colas et al., 1995; Feng et al., 1996; Han et al., 1996), including a putative Wingless receptor Frizzled-2 (Bhanot et al., 1996). Our studies of the gprk2⁶⁹³⁶ mutant suggest that a signaling pathway involving a G proteincoupled receptor functions during egg chamber maturation.

Flies were maintained on yeasted cornmeal/molasses media under standard culture conditions. ry⁵⁰⁶ stocks were used as controls in all experiments (Lindsley and Zimm, 1992). The gprk2⁶⁹³⁶ mutant, fs(3)06936, and a lethal dpp allele, l(3)10638, were recovered from a P element mutagenesis screen (Karpen and Spradling, 1992). The dpp allele is described by Twombly et al. (1996). Balancer chromosomes are as described by Lindsley and Zimm (1992). The BS3.0, dpp-lacZ enhancer line, which was kindly provided by Dr. W. Gelbart, gives the complete pattern of dpp staining in the blade of the imaginal wing disc (Blackman et al., 1991).

To test for an interaction between gprk2⁶⁹³⁶ and dpp two sets of crosses were carried out. To examine wing imaginal discs, the homozygous BS3.0 dpp-lacZ enhancer line was crossed to gprk2⁶⁹³⁶/TM3, Sb. Sb⁺ male progeny were crossed to gprk2⁶⁹³⁶/TM6, Tb and Tb⁺, third-instar larvae were dissected open and stained for lacZ activity. Tissues were fixed and stained as described by Margolis and Spradling (1995). Homozygous gprk2⁶⁹³⁶ larvae were identified by the strength of β-gal staining in the central nervous system.

To examine the effects of the gprk2⁶⁹³⁶ mutation on dpp expression in the ovaries, a gprk2⁶⁹³⁶ chromosome that harbors the recessive markers thread (th), scarlet (st) and curved (cu) was crossed to the dpp allele, l(3)10638. l(3)10638/+; gprk2⁶⁹³⁶, th st cu/+ males were crossed to gprk2⁶⁹³⁶/, th st cu/TM3, Sb females and gprk2⁶⁹³⁶ homozygotes were identified by the presence of th, st and cu. Ovaries were dissected, fixed and stained for β-gal activity.

Recovery of genomic DNA flanking the P element insertion was carried out as described by Cooley et al. (1988). A 2.0 kb fragment flanking the fs(3)06936 insertion was used to screen a Charon 4A library of Canton-S genomic DNA (Maniatis et al., 1978) to obtain overlapping clones for the 5′ region of the gene. The same library was probed with the B6936 cDNA (see below) to obtain clones for the 3′ region of the gene. These hybridizations were carried out in 6× SSC + 0.5% dry milk, as described by Sambrook et al. (1989). All probes were generated by random hexamer priming.

A 3.1 kb genomic fragment from the 5′ region of the gene was used to screen a Drosophila ovary cDNA library (Stroumbakis et al., 1994). Hybridizations were carried out as above. Three identical clones of approximately 2.5 kb were isolated (named T6936). A 1.0 kb 3′ T6936 fragment was used to probe a 0-4 hour, cDNA embryo library (Brown and Kafatos, 1988) and two clones of approximately 3.5 kb were isolated. These clones were judged to be identical based on restriction mapping and sequencing of 5′ and 3′ ends, and only one of these clones was further characterized (named B6936). P1 clones from the 5′ and 3′ ends of the gene were obtained from the Berkeley Drosophila Genome Project (BDGP). P1 DNA was isolated according to the BDGP protocol.

DNA sequencing was carried out manually with Sequenase (United States Biochemical) or with Taq DyeDeoxy terminator Cycle AutoSequencing (Applied Biosystems Model 373A). Genomic and P1 fragments and cDNA clones B6936 and T6936 were subcloned into Bluescript-(KS−). Sequencing was carried out using T3, T7 and Gprk2 primers on subcloned DNA, as well as on templates generated by nested deletions. cDNA B6936 was sequenced entirely on both strands whereas the T6936 3′ fragment and genomic fragments were sequenced on one strand and compared to the B6936 sequence.

To obtain RNA from Drosophila ovaries, females were dissected in Drosophila Ringers and their ovaries were frozen immediately on dry ice. Tissues were thawed and homogenized in extraction buffer containing 50 mM Tris, pH 8.0, 1 mM EDTA, 100 mM NaCl, 0.5 % SDS, and an equal volume of phenol/chloroform/isoamyl alcohol (25:24:1 ratio). Tissues were repeatedly extracted until the protein interface was minimal and the RNA was precipitated by the addition of sodium acetate and ethanol.

To obtain RNA from staged embryos and third instar larvae, tissues were frozen on dry ice and homogenized in 8 M guanidine isothio-cyanate buffer using a polytron homogenizer (as described by Sambrook et al., 1989). Poly(A)⁺ RNA was purified on oligo(dT)-cellulose columns (Collaborative Research), size-fractionated on formaldehyde agarose gels, and transferred to GeneScreen nylon membranes in 10× SSC. Hybridization was carried out using the buffers described by Church and Gilbert (1984). Relative amounts of RNA on northern blots were determined by hybridization with β-tubulin probes.

The ovary northern blot (Fig. 6A) contained 4 μg of poly(A)⁺ RNA per lane. The blot was probed with fragments of B6936 cDNA or the entire T6936 cDNA. The developmental northern blot (Fig. 6B) contained 1 μg of poly(A)⁺ RNA per lane. The blot was probed first with a T6936-specific probe and exposed for 10 days. Next the blot was probed (without stripping) with the entire B6936 cDNA and exposed to film overnight. Finally, the blot was probed (again without stripping) with β-tubulin and exposed to film overnight. The T6936-specific probe was generated by digesting the T6936 cDNA insert with HinfI, and isolating a 453 bp internal fragment by electrophoresis. This fragment has no sequences in common with B6936.

Purified antibodies to GPRK2 were generously provided by Charles Zuker and are described by Cassill et al. (1991). The antigen used to generate the antibody was an expressed protein containing amino acids 429-714 (according to the numbering in this article).

For western blot analysis, tissues were dissected in Drosophila Ringers and immediately frozen on dry ice. Tissues were thawed and homogenized in sample buffer (10 mM sodium phosphate buffer, 6 M urea, 2% SDS, 1% β-mercaptoethanol, 10 mM dithiothreitol, and 20 μg/ml PMSF). Samples were boiled for 10 minutes, electrophoresed on 7% SDS-PAGE gels, and transferred to nitrocellulose (Schleicer and Schuell) using the Bio-Rad MiniProtean system. Complete transfer was assessed by prestained molecular weight standards (Bio-Rad) and by staining of transferred proteins with amido black.

Nitrocellulose filters were preincubated in blocking solution (3% dry milk and 3% BSA in PBS) for 2 hours at room temperature or overnight at 4°C. Primary antibody (1:100 of anti-GPRK2) was added to the blocking solution and allowed to incubate overnight at room temperature. Washes were carried out in PBS + 0.1% Tween 20 at room temperature. Incubations in horseradish peroxidase-coupled, secondary antibody (Amersham; diluted 1:5000 in blocking solution) were carried out for 1-2 hours at room temperature. Antibody binding was visualized using the Amersham ECL kit.

For immunohistochemistry, tissues were dissected in Drosophila ringers and fixed in 4% paraformaldehyde (e.m. grade, Ted Pella) diluted in PBS. Optimal times of fixation were derived for each tissue: 10 minutes for ovaries, 20 minutes for imaginal discs, and 40 minutes for central nervous systems. Fixed tissues were washed in PBS + 0.3% Triton X-100 and blocked for 4 hours at room temperature or overnight at 4°C in PBS, 0.3% Triton X-100, 3% normal goat serum (NGS). Primary antibody was added to the blocking solution at a dilution of 1:50 or 1:100 and allowed to incubate 1-2 nights at 4°C. Tissues were washed 3 times for 5 minutes and then 6 times for 20 minutes in PBS + 0.3% Triton X-100. Secondary antibody coupled to horseradish peroxidase (Jackson Labs) was diluted 1:300 in PBS, 0.3% Triton X-100, 3% NGS and incubated with the tissue overnight at 4°C. Tissues were washed as described above and antibody binding was visualized by reaction with diaminobenzidene (DAB; Sigma Chemical). In some cases the anti-GPRK2 antibody was pre-absorbed against gprk26936 ovaries; this treatment reduced cytoplasmic staining in the nurse cells but did not affect membrane-associated staining in any tissue. Due to the weakness of GPRK2 staining, tissues were viewed with bright-field rather than interference-contrast microscopy.

Double staining in wing discs was carried out using the GPRK2 antibody and an Engrailed (En) monoclonal antibody that was kindly provided by N. Patel (4F11, Patel et al., 1989). Expression of dpp was assessed using the enhancer trap staining of l(3)10638 (Twombly et al., 1996) or the dpp-lacZ enhancer line, BS3.0 (Blackman et al., 1991) visualized by a β-galactosidase antibody. The En antibody was used at a concentration of 1:2 and the β-gal antibody was used at a concentration of 1:5000. Staining with these antibodies was carried out as described above for anti-GPRK2 and according to Patel (1994) for anti-En and anti-β-gal. To achieve two-color staining, the first antibody (anti-GPRK2) was visualized using DAB intensified with nickel chloride and the second antibody (anti-En or anti-β-gal) using DAB with no intensification.

Whole-mount in situ hybridization was carried out as described previously (Yue and Spradling, 1992).

To collect staged embryos, females were allowed to lay eggs on agar plates for 4 hours. After 24 hours, unhatched eggs were dechorionated, mounted in Hoyer’s medium, and heated to 50°C for several hours. Cuticles were visualized and images were captured under brightfield optics. Pseudo-darkfield images were generated by inverting the image colors using Adobe Photoshop.

Adult viable mutations of decapentaplegic (dpp), or its putative receptor saxophone (sax), cause homozygous females to produce short rounded eggs with abnormal anterior structures (Twombly et al., 1996). We identified a new female sterile mutation, fs(3)06936, during a single-P element mutagenesis screen (Karpen and Spradling, 1992) that caused similar effects on egg shape. Mature oocytes produced by homozygous fs(3)06936 females are slightly shorter and more rounded than wild type (compare Fig. 1A with Fig. 1B-D). Although positioned normally, dorsal appendages in fs(3)06936 eggs are generally shorter and broader than wild type (Fig 1C,D), and the two appendages on a single egg chamber frequently differ in length. The operculum is oriented more vertically than in wild type, giving the eggs a ‘square-ended’ appearance (compare Fig. 1A with Fig. 1C,D), but the chorion within the operculum retains its distinctive appearance (Fig. 1E) and the micropyle forms (Fig. 1F). Nurse cells often fail to completely transfer their contents into the oocyte, leaving residual material that may interfere with anterior end formation (Fig. 1B,F). Thus fs(3)06936 appears to affect specific aspects of egg formation without grossly altering the major pattern axes of the egg.

Two additional ovarian defects suggested that fs(3)06936 also functions at earlier stages of oogenesis. First, homozygous fs(3)06936 egg chambers degenerate during vitellogenic stages (stages 8-10A) much more frequently than expected. 26.8% of ovarioles (n=594) from 4-day-old fs(3)06936 females contained a degenerating vitellogenic chamber compared to only 0.7% of wild-type ovarioles (n=276). Second, egg chamber formation slows or ceases entirely within a significant number of fs(3)06936 ovarioles. 5.2% of the mutant ovarioles (n=594) contained only 0-2 egg chambers instead of the 6-7 that are present in wild type (Fig. 1C). Germaria in such ovarioles were often smaller and thinner than in wild type, like the germaria of agametic ovarioles (see Margolis and Spradling, 1995). Although cyst production normally declines in old females, 4-day-old wild-type females contained no similar ovarioles (n=276).

The fs(3)06936 mutant exhibited additional defects that indicate roles for this gene outside of the ovaries. Homozygous fs(3)06936 females lay a small number of eggs, but those that are laid display a maternal effect that is partially rescued by zygotic fs(3)06936⁺ expression. 23.7% of embryos (n=300 eggs laid) produced by homozygous females hatch when crossed to wild-type males compared to only 10.3% (n=390) following crosses to homozygous males. The unhatched eggs displayed a wide variety of defects including twisted gastrulation, fused adjacent segments, and perforated dorsal and ventral cuticle (Fig. 1G,H). These defects were more severe when the embryos lacked both maternal and zygotic fs(3)06936⁺ function (as seen in Fig. 1H).

To further analyze the fs(3)06936 mutation, we cloned the gene disrupted in this strain. The single P element insertion in the fs(3)06936 mutant was localized to cytological position 100C by in situ hybridization (data not shown). Seventy-two of eighty-six lines in which the P element was excised reverted to fertility, confirming that the insertion was responsible for the oogenic defect. We first cloned genomic DNA flanking the insertion by plasmid rescue (Cooley et al., 1988), and used it to isolate overlapping clones from a lambda genomic library (Fig. 2). One of these genomic clones, λMgB, contains the P element insertion site and hybridized in situ at position 100C, confirming that the genomic DNA derived from the appropriate chromosomal region (not shown).

We used the genomic map to identify candidate transcripts encoded by the fs(3)06936 gene. A 3.1 kb fragment containing the insertion site was used to select a clone, T6936, from an ovary cDNA library (Stroumbakis et al., 1994) that hybridizes to transcripts of 4.0, 5.0 and 5.5 kb in length in ovarian poly(A)⁺ RNA. The sizes of the T6936-complementary RNAs were drastically altered in fs(3)06936 mutant ovaries (see below), indicating that they were good candidates for fs(3)06936 transcripts. A 3′ segment of T6936 was used to probe a cDNA library prepared from 0-4 hour embryos (Brown and Kafatos, 1988) and a longer 3.5 kb cDNA, B6936, was recovered. The 3547 nucleotide sequence of B6936 (Fig. 3) predicts a protein of 714 amino acids preceded by a long 5′ untranslated region (1118 bp). Nucleotides 1719 through 3532 are identical to a previously described cDNA sequence, the G protein-coupled receptor kinase 2 gene (Gprk2, Cassill et al., 1991). However, the B6936 cDNA extends further in the 5′ direction than the previously described clone, and predicts a larger protein due to an extended amino-terminal region. The P element insertion was located between nucleotides 142 and 143 of B6936, within the 5′ untranslated region of the gene. These observations imply that the gene disrupted in fs(3)06936 corresponds to Gprk2; and we will subsequently refer to fs(3)06936 as gprk2⁶⁹³⁶.

The Gprk2 gene was originally isolated based on its sequence similarity to the human βARK1gene (Cassill et al., 1991). The predicted protein from Gprk2 is 47% identical to βARK1 protein in the kinase domain. Outside of this region there is no significant similarity. In contrast, the GPRK2 protein predicted by the B6936 sequence shares a high level of sequence identity with two other mammalian members of the GRK family, GRK5 and GRK6 (Fig. 4). This similarity extends over the entire protein, except for a 121 amino acid region that is unique to GPRK2. The physiological functions of GRK5 and GRK6 have not yet been established, although both proteins are able to phosphorylate and desensitize activated G protein-coupled receptors.

Comparing genomic DNA clones with cDNA B6936 revealed that only the first exon (where the P element had inserted) derives from the genomic region that we had analyzed. To obtain genomic clones containing the rest of the gene, we used a 3′ fragment of B6936 as a probe to re-screen the genomic phage library. In addition, P1 clones of this region were obtained from the Berkeley Drosophila Genome Project. Mapping of the genomic and P1 clones to B6936 showed that Gprk2 is encoded by at least 13 exons (Fig. 2). Phage clones that mapped 3′ to exon 1 and 5′ to exon 3 did not overlap suggesting that these two exons are at least 15 kb apart. According to the cDNA sequence, there are 48 nucleotides between exon 1 and 3 that are encoded in this uncloned region. This is the only portion of the cDNA B6936 for which the genomic counterpart was not sequenced. We have putatively designated this 48 nucleotide sequence as exon 2, although it could be composed of more than one mini-exon. Sequencing of the 3′ end of cDNA T6936 demonstrated that it ends in a polyadenylation site that is 750 nucleotides 3′ to the B6936 polyadenylation site. The 5′ end of T6936 was not fully characterized, but hybridization analysis demonstrated that it contains sequences from the 5′ end of B6936. This suggests that the middle portion of the gene is not present in T6936, either due to alternative exon processing or to deletions that were generated during construction of the cDNA library.

To determine how the P element disrupts expression of the Gprk2 gene, we analyzed RNA expression using probes from both ends of cDNA B6936 and from the 3′ end of cDNA T6936 (Fig. 5A). All three probes hybridized to transcripts of 4.0, 5.0, and 5.5 kb in length in poly(A)⁺ RNA from wild-type ovaries. In mutant ovaries, the full-length transcripts were no longer detectable. However, with the 5′-most probe, new transcripts of 1.3 and 0.8 kb were recognized. This suggests that transcription begins normally but is terminated by the P element (Horowitz and Berg, 1995). A developmental analysis of expression (Fig. 5B) showed that the 4.0 and 5.0 kb transcripts are abundantly expressed in 0-3 hour embryos, as a result of maternal expression. The 5.0 kb transcript is no longer detectable after 0-3 hours while the 4.0 kb transcript continues to be expressed at a low level. The 5.5 kb transcript is expressed at a low level in early embryogenesis, with an increase in expression from 12 to 24 hours. Thus the three transcripts are differentially expressed during development.

A probe derived from the 3′ region that is specific to T6936 hybridized to only the two largest transcripts (data not shown). This suggests that B6936 corresponds to the smallest (4.0 kb) transcript, which is consistent with the fact that this cDNA was isolated from a 0-4 hour library. The cDNA sequence reported by Cassill et al. (1991) did not terminate in a poly(A) tail, however it extended further 3′ than either B6936 or T6936 suggesting that the Gprk2 gene utilizes at least three polyadenylation sites (Fig. 5C). Whether use of these three sites gives rise to the three transcript sizes will require further confirmation.

The expression of GPRK2 protein was analyzed using a polyclonal antibody that was kindly provided by Charles Zuker. This antibody was generated against a bacterially expressed protein which contained the final 145 amino acids of the kinase domain and the entire 139 amino acid carboxyl region (Cassill et al., 1991). DNA probes from the corresponding genomic region hybridize to all three transcripts (not shown). In immunoblot analysis of adult tissues, we detected a single band of 80 kDa in ovaries and whole females, in close agreement to the 80.7 kDa protein predicted by the B6936 cDNA. This band was absent in ovaries from gprk2⁶⁹³⁶ homozygotes whereas expression in whole females remained (Fig. 6). Thus, as expected, GPRK2 is expressed in the ovary and its expression is disrupted in the gprk2⁶⁹³⁶ mutant. In larval tissues, we again observed an 80 kDa band in both the central nervous system (CNS) and in carcasses (the entire animal minus the CNS). In homozygous gprk2⁶⁹³⁶ larvae, expression was no longer detectable in CNS tissue but appeared to be unaffected in carcasses. These results demonstrate that gprk2⁶⁹³⁶ selectively disrupts Gprk2 expression in some but not all tissues.

The tissue distribution of the protein was studied by staining whole-mount tissues with the same antibody. The anti-Gprk2 antibody labeled developing wild-type egg chambers (Fig. 7), and most of the staining was eliminated in gprk2⁶⁹³⁶ mutant ovaries. Expression was first seen in region 2B of the germarium, the stage where germline cysts are being enveloped by follicle cells (Fig. 7B, gm). In early egg chambers, membranes between adjacent nurse cells and between nurse cells and the oocyte were labeled (Fig. 7A-C, m). In addition, staining around the germinal vesicle was sometimes observed (Fig. 7C, gv). Membrane-associated staining persisted through stage 6 but decreased in intensity at stage 7, except between the nurse cells and the oocyte (Fig. 7A, S7). In these early stages we observed no staining in follicle cells or in the region of the nurse cells that lies adjacent to the follicle cells. During stages 8-11 membrane-associated staining decreased in the nurse cells and appeared around the entire circumference of the oocyte (Fig. 7A and D, m). We also observed weak cytoplas-mic staining in nurse cells and follicle cells of these stages; however, the cytoplasmic (and germinal vesicle) staining sometimes persisted in gprk2⁶⁹³⁶ ovaries (Fig. 7E). Thus GPRK2 protein appears to be preferentially associated with nurse cell and oocyte plasma membranes during much of oogenesis. Although in situ hybridization studies also detected gprk2 RNA only in the nurse cells (not shown), a lower level of expression in somatic cells can not be ruled out by these experiments.

Specific staining using anti-GPRK2 was also detected in non-ovarian tissues, consistent with a role for this gene outside of the ovaries. The central nervous system (CNS) of both larvae and adults stained intensely. In larvae, staining was present in axon fascicles, especially large ones, including nerves projecting to the optic lobes, the longitudinal connectives, and portions of the mushroom bodies (ol, lc, and mb in Fig. 8A). We detected little staining in cell bodies within the CNS. However, we consistently observed strong staining in the cell bodies and nerves of the corpus allatum of the ring gland (rg in Fig. 8A). In the adult CNS, staining was restricted to two major structures within the brain. First, the nerves within and projecting to the ellipsoid body of the central complex were consistently stained (eb in Fig. 8C). The ellipsoid body contributes to higher brain function in flies, such as locomotion (Strauss and Heisenberg, 1993). Second, we also detected strong staining in the mushroom bodies that included the Kenyon cells and nerve processes within the peduncles and the alpha lobes (not shown). The mushroom bodies have been implicated in memory and learning. Interestingly, the learning mutants dunce, rutabaga, and DCO, which all disrupt genes involved in G protein signaling, are predominately expressed in the mushroom bodies (reviewed in Davis et al., 1995). In the mutant, GPRK2 staining was abolished from both the larval and adult CNS, and from the larval ring gland (Fig. 8B,D).

Most of the other tissues in the larva and adult were only stained weakly and it was not possible to determine if staining was affected in the mutant. One exception to this was in the wing imaginal disc where we observed staining in the notum and in a stripe that paralleled the anterior/posterior boundary of the wing blade (Fig. 8E, A/P, and 9A). Within this stripe, staining was always weakest in the region corresponding to the tip of the wing. This staining was eliminated in gprk2⁶⁹³⁶ wing discs (Fig. 8F).

The disc staining was particularly interesting because dpp is also expressed along the anterior/posterior boundary of the wing disc. To determine how GPRK2 staining corresponds to the anterior/posterior boundary, we compared expression of GPRK2 with Engrailed (En) and dpp. En protein, which is expressed in the posterior compartment of the wing disc, was visualized with a monoclonal antibody (4F11, Patel et al., 1989). dpp expression was visualized by β-gal immunoreactivity in flies bearing a dpp-lacZ fusion gene (BS3.0; Blackman et al., 1991). GPRK2 and dpp are co-expressed along the anterior/posterior boundary with two exceptions (Fig. 9C). First, the stripe of GPRK2 staining is thinner, about one cell in thickness as opposed to a 2-3 cell thickness of dpp expression. The stripe of GPRK2 expression coincided with the dpp-expressing cells closest to the boundary. Second, near the wing tip, dpp is not expressed in the cells immediately adjacent to the boundary. In this region GPRK2 staining was present in cells that do not express dpp.

At the tip of the wing En staining crosses the anterior/posterior boundary, as previously described (Blair, 1992; Raftery et al., 1991). Therefore, in this region GPRK2 and En were co-expressed whereas outside of this region the patterns of the two proteins abutted (Fig. 9B). These results (summarized in Fig. 9D) confirm that GPRK2 is expressed along the anterior side of the anterior/posterior border in a pattern that overlaps dpp.

One result of signaling through G protein-coupled receptors is the induction of cAMP which activates cAMP-dependent protein kinase (PKA). PKA has been shown to repress dpp expression in imaginal discs (Jiang and Struhl, 1995; Lepage et al., 1995; Li et al., 1995; Pan and Rubin, 1995; Strutt et al., 1995). If PKA plays a similar role in egg chambers, then Gprk2 may control the level of dpp expression in discs and ovaries by influencing PKA activity. To test this hypothesis we followed the expression of dpp-lacZ enhancer lines in gprk2⁶⁹³⁶ homozygotes. l(3)10638 and the BS3.0, dpp-lacZ enhancer line were used to assess dpp expression in the ovaries and wing imaginal discs, respectively. We observed no differences in dpp expression in either ovaries or wing imaginal discs in these studies (not shown).

We characterized a Gprk2 mutation, gprk2⁶⁹³⁶, and showed that Gprk2 activity is required during oogenesis and embryo-genesis. Gprk2 encodes a putative member of the G protein-coupled receptor kinase (GRK) family. The hallmark of GRKs is that they phosphorylate G protein-coupled receptors when they are in the active (agonist-bound) state (Premont et al., 1995). Although we have not demonstrated biochemical activity of GPRK2, several characteristics suggest that it is a true member of the GRK family. First, the sequence similarity in the kinase domain between GPRK2 and other family members with known activity is high, ranging from 47% with βARK1 to 86% with GRK5. Second, the GPRK2 kinase domain contains amino acids that are conserved in all other members of the GRK family as well as residues that are required for the activity of all serine/threonine kinases (Fig. 3; Hanks and Quinn, 1991; Taylor et al., 1993). Finally, GPRK2 and its putative human homologue, GRK5, share extensive sequence similarity outside of the kinase domain including a serine/threonine pair that is autophosphorylated in GRK5 (Kunapuli et al., 1994). One difference between GPRK2 and the other GRKs is the length of the amino-terminal region. Most GRKs have an amino terminus of 183-188 amino acids. GPRK2 has a 121 amino acid insertion with respect to the GRK5 sequence, generating an amino terminus of 309 amino acids. This region of the protein is thought to play a role in substrate specificity.

Despite the presence of multiple transcripts, our studies identified only a single GPRK2 protein of 80 kDa. Previously, Cassill et al. (1991) predicted a GPRK2 protein of 50 kDa by assuming that translation initiates at a methionine just prior to the kinase domain. In our sequence this methionine lies at amino acid 288, and the predicted size of the protein is 80.7 kDa. Because only the 80 kDa protein was detected using the antibodies reported in Cassill et al. (1991), and because this band was specifically eliminated in gprk2⁶⁹³⁶ mutant ovaries and CNS, we believe that it corresponds to the Gprk2 gene product. Comparing the GPRK2 sequence predicted from the B6936 cDNA with GRK5 provided further support that we had analyzed the full-length protein. The first 114 amino acids share 53% identity with GRK5 including the first 19 amino acids, 18 of which are identical between GPRK2 and GRK5.

If the effects of the gprk2⁶⁹³⁶ mutation on oogenesis are caused by the disruption of the second messenger, cAMP, then mutations in genes that lie downstream of G protein-coupled receptors might also affect female fertility. In fact, three candidate genes have been identified: dunce (dnc), rutabaga (rut), and DCO. All three genes act in the G protein-coupled receptor pathway that is mediated by the second messenger, cAMP. dnc encodes a cAMP-dependent phosphodiesterase, and in dnc mutants the level of cAMP is elevated in comparison to wild type. dnc homozygous females are sterile although, like gprk2⁶⁹³⁶ females, they produce mature oocytes with defects in dorsal appendage formation and nurse cell dumping (Bellen et al., 1987; L. E. S., unpublished observations). The female sterility of dnc is suppressed by mutations in the rut gene, which encodes Ca⁺/calmodulin-dependent adenylyl cyclase (Bellen et al., 1987). This suppression occurs by decreasing the levels of cAMP. These phenotypes are consistent with a potential role for Gprk2 in this classic G protein pathway. If GPRK2 down-regulates a G protein-coupled receptor, then the gprk2⁶⁹³⁶ mutant should cause prolonged receptor activity resulting in increased cAMP levels, as seen in dnc mutants.

Mutations at the DCO locus disrupt the major catalytic component of cAMP-dependent protein kinase (PKA). In ovaries of DCO females, late stage nurse cells are often binucleate, although younger egg chambers have no such defects (Lane and Kalderon, 1993). This suggests that the breakdown of nurse cell membranes, which normally begins in stage 11, takes place prematurely in the mutant. In some respects, gprk2⁶⁹³⁶ mutant egg chambers have the opposite phenotype since nurse cells do not break down completely. A reduction in GRK activity and a reduction in PKA activity would be expected to exert opposite effects on signaling. Moreover, GPRK2 protein was detected most strongly on nurse cell membranes throughout much of oogenesis and PKA activity is known to be required in the germline.

Our results suggest that GPRK2 may be part of a G protein signaling pathway in the nervous system as well. GPRK2 was strongly expressed in the mushroom bodies, a neural structure that has been implicated in learning and memory. dnc, rut and DCO exhibit defects in olfactory learning (reviewed in Davis et al., 1995), and all three are abundantly expressed in the mushroom bodies (de Belle and Heisenberg, 1994; Erber and Menzel, 1980; Heisenberg et al., 1985). Gprk2 may act in a common pathway with dnc, rut and/or DCO in both ovaries and mushroom bodies. Since the gprk2⁶⁹³⁶ mutation eliminated gprk2 expression in the CNS, it would be worthwhile testing whether the mutant adults exhibit defects in learning and memory.

A number of seven-transmembrane receptors that are likely to couple to G proteins have been identified in Drosophila. However, most of the known Drosophila G protein-coupled receptors, including those for dopamine, serotonin, octopamine, tyramine, tachykinin, and the muscarinic acetyl-choline receptor, have been cloned by homology and no mutations are known (Shapiro et al., 1989; Saudou et al., 1992; Arakawa et al., 1990; Li et al., 1991, 1992; Monnier et al., 1992; Gotzes et al., 1994; Colas et al., 1995; Feng et al., 1996; Han et al., 1996). Therefore, it remains uncertain whether these receptors function during development.

Two putative G protein-coupled receptors have been identified recently that play major roles during development. smoothened (smo), encodes a putative Hedgehog receptor (Alcedo et al., 1996; van den Heuvel and Ingham, 1996) while frizzled2 (Dfz2), specifies a putative Wingless receptor (Bhanot et al., 1996). Both predicted proteins contain seven transmembrane domains although they have little sequence similarity to other G protein-coupled receptors. There are currently three distinct superfamilies of receptors that couple to G proteins: those in the βAR family, the metabotropic glutamate receptors, and members of the GI hormone family (Premont et al., 1995). It is possible that Smo and Dfz2 represent additional families of G protein-coupled receptors.

Are the Smo and Dfz2 receptors likely to be good substrates for GPRK2? GRK5 phosphorylates βAR at 6 sites within the last 40 amino acids of the receptor; no phosphorylation occurs on cytoplasmic loops (Fredericks et al., 1996). The same regions of the Smo and Dfz2 carboxy termini contain 6 or 8 potential phosphorylation sites, respectively. However, Smo may act as the better substrate for GPRK2 because GRK5, like βARK, shows a preference for target residues located within an acidic environment (Fredericks et al., 1996). Dfz2 has no acidic amino acids in the last 40 amino acids of the protein. In contrast, Smo has several acidic residues in the 40 amino acid domain including an acidic pair upstream of the most aminoterminal phosphorylatable residue in the 40 amino acid fragment. This same pattern occurs in receptors that are phosphorylated by GRK5 and βARK. Thus, from a structural stand-point Smo is an excellent candidate to mediate the developmental effects of GPRK2.

The role played by hh and wg during oogenesis has been analyzed recently (Forbes et al., 1996a,b). Both genes are expressed in small groups of cells at the anterior tip of each ovariole. The hh-expressing cells influence the proliferation of somatic follicle cell precursors and the specification of polar follicle cells. However, effects on oogenesis similar to those caused by gprk2⁶⁹³⁶ were not observed following changes in hh or patched activity. While the ovarian hh pathway resembles the pathway in early embryos and imaginal discs (reviewed by Perrimon, 1994; Ingham, 1995) in its requirement for patched, ovarian hh signaling is not mediated by wg or dpp. A separate role for wg in oogenesis is suggested by the observation that overexpression of wg causes multiple developmental defects in the ovary (Forbes et al., 1996b). However these defects do not mimic the effects of the gprk2⁶⁹³⁶ mutation. Consequently, changes in inhibitory phosphorylation of Smo and Dfz2 are not likely to be responsible for the defects we observed in gprk2⁶⁹³⁶ females.

The similar effects on oogenesis of disrupting Gprk2 and dpp suggest that these two genes may be involved in a common pathway. Dpp receptors differ significantly from serpentine receptors and are unlikely to serve as GPRK2 substrates. While it is possible that GPRK2 can act directly on targets other than G protein-coupled receptors, mammalian GRK family members are thought to phosphorylate serpentine receptors exclusively. It is more likely that Gprk2 influences either the expression of dpp itself or the effects of Dpp signaling downstream of the receptors. We were unable to detect any change in a dpp enhancer line in homozygous gprk2⁶⁹³⁶ ovaries, suggesting that gprk2⁶⁹³⁶ does not affect dpp transcription. It remains a possibility that gprk2⁶⁹³⁶ could disrupt Dpp signaling by altering the posttranslational processing of dpp. Gprk2⁶⁹³⁶ could also affect Dpp signaling by acting downstream of Dpp receptors, a relationship that has been demonstrated between tyrosine kinase receptor and G protein-coupled receptor pathways (Luttrell et al., 1995; Van Biesen et al., 1995; Hafner et al., 1994). This second possibility is favored by our expression studies. dpp is expressed most strongly in the anterior follicle cells and its activity is required in these soma. In contrast, GPRK2 is expressed predominantly in the nurse cells although immunoreactivity was also detected in stage 9-10 follicle cells.

The co-expression of GPRK2 and dpp-lacZ in the wing imaginal disc suggested that these two molecules interact in the patterning of the wing. However, we have not been able to detect any wing defects or changes in dpp-lacZ expression in gprk2⁶⁹³⁶ homozygotes. The effects of the gprk2⁶⁹³⁶ mutation on dpp expression and wing patterning may be too subtle to be detected with the tools that we used. Alternatively, gprk2⁶⁹³⁶ may play no role in wing patterning and its expression at the anterior-posterior boundary of the wing disc may play another role altogether.

The finding that a putative G-protein coupled receptor kinase functions during several developmental processes raises the question of the developmental role of receptor desensitization. The primary effect of gprk2⁶⁹³⁶ during oogenesis was to disrupt anterior chorion formation, a process mediated by rapidly migrating subpopulations of follicle cells. G protein signaling pathways are known to be important for cell migration and/or shape changes in other developing systems. Examples include the aggregation of Dictyostelium discoideum to form a migrating slug and fruiting body (Devreotes, 1994), gastrulation in Drosophila embryos which requires a specific Gα subunit encoded by the concertina locus (Parks and Wieschaus, 1991), FGF-stimulated endothelial cell migration which is pertussis toxin-sensitive (Sa and Fox, 1994), and Rho-dependent neurite retraction in mammals (Postma et al., 1996).

Because GRKs phosphorylate activated receptors, desensitization provides an ideal mechanism for regulating the responsiveness of receptors based on their previous signaling activity. Based on this idea, we propose three potential developmental roles for desensitization by GRKs. First, desensitization could allow a cell to generate a discrete response to a signal whose presence in the environment is prolonged. This could play an important role for a cell that is using a given signal as a migratory cue as it passes through an area where that signal is constant. Second, expression of a particular GRK could allow cells to respond differentially to particular ligands. In Chinese hamster ovary cells, the neurokinin1 and neurokinin2 receptors display different desensitization responses to substance P and neurokinin A (Garland et al., 1996). Third, desensitization could be the first step in receptor recycling to allow the cell to respond rapidly to changes in the level of an ongoing signal. All of these mechanisms would contribute to the ability of a cell to navigate through a complex and changing environment. Further study of Gprk2 should help elucidate the detailed role GPRKs play during development.