Signaling from germ cells mediated by the rhomboid homolog stet organizes encapsulation by somatic support cells

Germ cells normally differentiate while in intimate contact with somatic support cells. In mammals, differentiating male germ cells are enclosed in somatically derived Sertoli cells (Desjardins and Ewing, 1993) and oocytes are surrounded by somatic granulosa cells (Erickson, 1986; Hsuesh and Schomberg, 1993). In both cases, interactions between germ cells and surrounding somatic cells play important roles in gametogenesis (Marziali et al., 1993; Bitgood et al., 1996; Pesce et al., 1997; Ojeda et al., 2000; Matzug, 2000). Similarly, in Caenorhabditis elegans, early germ cells are closely associated with the somatic distal tip cell, which provides crucial signals that govern germ cell proliferation versus differentiation (Kimble and White, 1981; Berry et al., 1997). At subsequent stages, C. elegans germ cells interact with somatic sheath and spermathecal cells (Church et al., 1995; McCarter et al., 1997; Hall et al., 1999). In insects as well, germ cells are closely associated with somatic cells (King, 1970; Hardy et al., 1979; Bünning, 1994), which play key regulatory roles in germ cell fate (reviewed by Kiger and Fuller, 2001; Xie and Spradling, 2001).

In Drosophila males, germline stem cells lie at the apical tip of the testis, in intimate contact with somatic hub and cyst progenitor cells. Upon stem cell division, the daughter cell displaced away from the hub becomes encapsulated by two somatic cyst cells and initiates differentiation (Hardy et al., 1979). The surrounding somatic cyst cells play an important role in the initiation of germ cell differentiation (Kiger et al., 2000; Tran et al., 2000), and later in the transition from mitosis to meiosis (Gönczy et al., 1997; Matunis et al., 1997). In Drosophila females, somatic cells at the apical tip of the germarium form a specialized niche in which germline stem cells are maintained through signaling from the soma (Xie and Spradling, 1998; King and Lin, 1999; Xie and Spradling, 2000). After mitotic amplification, clusters of 16 interconnected female germ cells become surrounded by follicle cells, which form an epithelial sheath around each developing egg chamber. Interactions between germ cells and follicle cells regulate such critical events as egg chamber formation and determination of the polarity of the developing oocyte (reviewed by Ray and Schüpbach, 1996; Morgan and Mahowald, 1996).

Signaling via the epidermal growth factor receptor (Egfr) mediates many cell-cell interactions where one cell influences the proliferation or differentiation of a closely apposed partner (Schweitzer and Shilo, 1997; Freeman, 1998). Despite the exquisitely localized and temporally specific requirements for Egfr activation in normal development documented in Drosophila, the Egfr and its major ligand spitz (spi) are widely expressed (Lev et al., 1985; Kammermayer and Wadsworth, 1987; Rutledge et al., 1992). Spatial and temporal control of Egfr pathway activation appear to be achieved at the level of ligand activation. spi is synthesized as a transmembrane protein. Proteolytic cleavage of spi by the transmembrane protein rhomboid (rho) within the Golgi apparatus of the signal sending cell produces a potent diffusible ligand (Rutledge et al., 1992; Schweitzer et al., 1995; Golembo et al., 1996; Lee et al., 2001; Urban et al., 2001). Expression of rho is spatially and temporally controlled, providing developmental specificity to activation of the Egfr pathway (Bier et al., 1990)

In Drosophila oogenesis, germ cells signal via the germline Egfr ligand gurken (grk) to specify the correct behavior of follicle cells in encapsulating each individual cluster of 16 germ cells (Goode et al., 1992), and later to pattern the follicle cell layer (Schüpbach, 1987; Gonzales-Reyes et al., 1995). So far it has been unclear how Egfr is activated during oogenesis. Germline clones mutant for rho produced wild-type eggs, suggesting that rho is not required in germ cells. Instead, rho is expressed in follicle cells depending on Egfr activation (Ruohola-Baker et al., 1993), most likely to spread and amplify the initial signaling event (Wasserman and Freeman, 1998).

We show that stet, a homolog of rho, plays a crucial role in signaling from germ cells to somatic cells. Wild-type function of stet is required for encapsulation of germline stem cells and their progeny by somatic support cells and germ cell differentiation in both Drosophila males and females. Clonal analysis and rescue experiments in testes have demonstrated that stet function is required in germ cells. The conserved protease motif in the Stet protein (Urban et al., 2001) and its subcellular localization (Ghiglione et al., 2002) suggest that stet functions through the same biochemical mechanism as rho. In support of this, expression of rho in germ cells rescued the stet mutant testes phenotype. We propose that stet activates signaling from germ cells to the Egfr on somatic support cells to set up the crucial associations between germ cells and soma that are required for normal gamete differentiation.

Flies were raised on standard cornmeal molasses agar medium at 25°C. The original allele stet⁸⁷¹ was identified in an EMS screen for male sterility by J. Hackstein. Seven stet alleles (stet^{1, 2, 3, 8A, 8B, 8F, 9}) were isolated on a red, ebony chromosome in a screen in our laboratory for EMS induced mutations that failed to complement a chromosome carrying the stet⁸⁷¹ allele. Six additional stet alleles (stet^{Z3-0369, z3-4806, z3-3671, z3-0919, z3-3835, z3-2244}) were identified as male sterile mutations by B. Wakimoto and D. Lindsley in a collection of 12000 EMS-induced viable lines generated in the laboratory of C. Zucker. We identified these Zucker lines as carrying stet alleles by failure to complement the stet⁸⁷¹ allele. All other Drosophila mutants and balancer chromosomes are as described elsewhere (Lindsley and Zimm, 1992).

The stet⁸⁷¹ mutation was mapped by recombination between roughoid (ru) and hairy to 1.4 map units proximal to ru. stet was localized to polytene interval 62A1 with the following deficiencies generated in our laboratory: stet was uncovered by Df(3L)29b (61C;62A5) and Df(3L)PX62 (62A1), but not by Df(3L)PX49-15 (62A1;A8). The generation of deficiencies and their breakpoints are described elsewhere (Schulz et al., 2002). The stet mutant phenotype was analyzed in flies trans-heterozygous for loss-of-function alleles stet⁸⁷¹, stet² and stet^z3–3671 over Df(3L)PX62. stet⁸⁷¹, stet² and stet^z3-3671 displayed the same phenotype trans-heterozygous to each other as over DF(3L)PX62. However, stet⁸⁷¹ and stet^z3–3671 carried additional mutations on their chromosomes, giving rise to stronger phenotypes in ovaries when homozygous. Unless otherwise stated, images of stet mutant gonads shown were from stet⁸⁷¹/Df(3L)PX62 animals.

Ovaries and larval testes were stained for immunofluorescence, histochemistry, or β-galactosidase activity following standard protocols (Ashburner, 1989). Testes used for anti-Map-kinase immunohistochemistry were dissected in testes buffer with phosphatase inhibitors (10 mM Tris-HCl, pH 6.8, 180 mM KCl, 50 mM NaF, 10 mM NaVO₄ and 10 mM β-glycerophosphate) before the staining procedure. Immunofluorescence experiments on squashed testes were performed as described previously (Hime et al., 1996). The hybridoma/monoclonal antibodies mouse anti-α-spectrin (1:5) (developed by D. Branton and R. Dubreuil), mouse anti-fasciclin III (1:10) (developed by C. Goodman) and mouse anti-Sxl (1:200) (developed by P. Schedl) were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA 52242. The monoclonal mouse anti-Map-kinase antibody (Sigma) was used at 1:200, polyclonal rabbit anti-phosphorylated Histone-H3 antibody (Upstate Biotechnology, NY) was used at 1:100. Secondary antibodies (Jackson Immuno Research Laboratories) were used at 1:200. DAPI (Sigma) was used at 1 μg/ml. Counts of cell types were performed by scoring 20 to 50 testes or ovarioles of wild-type and stet mutant animals. Expression of UAS-constructs under control of the GAL-4 activator proteins (Phelps and Brand, 1998) were temporally controlled by growing flies at 18°C and shifting them to 29°C as larvae, pupae or adults.

Whole-mount in situ hybridization was performed as described by Tautz and Pfeifle (Tautz and Pfeifle, 1989), with modifications for RNA probes described by Klingler and Gergen (Klingler and Gergen, 1993). Ribonucleotide probes were generated from linearized plasmid using the Roche Molecular Biochemicals (Indianapolis) RNA-labeling kit.

The alleles stet⁸⁷¹ and stet^Z3-3671 were recombined onto FRT-80 chromosomes (FRT-80-stet). Males carrying the FRT-80-stet chromosome and control animals carrying the FRT-80 chromosome were crossed to females carrying the FLP recombinase gene under control of a heat-shock promoter and a FTR-80 chromosome marked with a nuclear targeted GFP (FRT-80-GFP). Progeny were heat-shocked during pupal stages or as adults for 2 hours in a 37°C waterbath. Testes from adult males were dissected 7 to 10 days after heat-shock. On average, one out of seven testes showed GFP-negative clones. Under these conditions, control animals often contained several clusters of GFP negative cells, while males carrying the FRT-80-stet chromosome usually contained only one cluster of GFP negative clones (nine out of 10 testes with clones scored).

All molecular techniques were performed using standard protocols (Sambrook et al., 1989). The genomic walk across cytological region 62A is described elsewhere (Schulz et al., 2002). The left molecular breakpoint of Df(3L)PX62 mapped in cosmid 116G11, in the 5′ non-translated region of the rho gene. The right breakpoint of Df(3L)PX62 mapped in cosmid 6814, 250 bp 5′ of the Drosophila puromycin-sensitive aminopeptidase gene translational start. 11 of the 12 potential transcription units in the 62A1 area were determined not to be stet by several approaches. Some of the transcription units were excluded because known mutant alleles complemented stet mutants. Others were excluded because genomic rescue constructs potentially containing the whole transcription units did not rescue the stet mutant phenotype when introduced into flies by P-element-mediated transformation (Spradling, 1986). Finally, for those of the 11 transcription units expressed in testes, we did not detect lesions in the coding regions when sequencing several strong stet alleles. Molecular information about the 62A area, cDNA clones of transcripts in this interval, as well as information about their expression pattern, and several genomic rescue constructs are available from the authors on request.

Wild-type function of the stet locus is required for male germ cells to proceed through early stages of differentiation. Loss-of-function stet mutant males were viable but sterile. Adult stet mutant males had tiny testes filled with small cells (Fig. 1B, arrowheads) resembling cells normally found only at the tip of wild-type testis (Fig. 1A, arrowhead). Early male germ cells failed to differentiate and instead accumulated in third instar larval testes from loss-of-function stet animals, based on appearance in phase contrast and DIC microscopy, nuclear size in DAPI-stained preparations (data not shown) and expression of cell-type specific markers. In wild type, early germ cells (stem cells, gonialblasts and spermatogonia) are located at the apical tip of the testis (Fig. 2A) and express the lacZ enhancer trap marker S3-46 (Fig. 1C, arrowhead). Spermatocytes are located more distally, fill most of the larval testis and do not express the S3-46 enhancer trap marker. Larval testes from stet mutant males were filled with cells expressing β-galactosidase from the S3-46 marker, suggesting that they were early germ cells (Fig. 1D).

In wild-type testes, mitotically active early germ cells were observed exclusively at the apical tip upon staining with anti-phosphorylated Histone-H3 antibody. Germline stem cells and gonialblasts divide as single cells (Fig. 1E, arrowheads), while spermatogonia divide in groups of two, four or eight cells (Fig. 1E, arrow). In stet mutants, many phosphorylated Histone-H3-positive cells were scattered throughout the testes, suggesting that the early germ cells accumulating in stet mutant testes remained mitotically active. Many anti-phosphorylated Histon-H3-positive cells were detected as single cells throughout stet mutant testes (Fig. 1F, arrowheads), indicating that cells with stem cell or gonialblast identity had been displaced away from the tip.

stet mutant testes appeared to contain a mixture of germ cells with stem cell, gonialblast and spermatogonial identities. In wild-type testes, α-spectrin is localized to a ball-shaped spectrosome in germline stem cells and gonialblasts and to the branched fusome structure passing through the intercellular bridges between spermatogonia (Fig. 1G). In wild-type testes, 10 to 20 cells with a spectrosome dot could be detected at the apical tip. In stet mutant testes, we detected 20 to 40 cells with spectrosome dots (Fig. 1H, arrowheads) at the apical tip, and many cells with a spectrosome dot displaced away from the tip, suggesting an increased number of stem cells and/or gonialblasts. However, most of the small germ cells accumulating in stet mutant testes were interconnected by short, branched fusomes, suggesting spermatogonial identity (Fig. 1H, arrows).

Staining for escargot (esg) mRNA also suggested an increased number of cells with stem cell characteristics. In wild-type, esg mRNA was detected in the somatic hub cells and in the five to nine germline stem cells around the hub (Fig. 1I, arrow), but not in gonialblasts and spermatogonia. stet mutant testes had in average 40 esg-positive cells, ranging from the normal five to more than 100, with some at the apical tip and some displaced away from the apical tip (Fig. 1J, arrows).

Staining for somatic hub cell lacZ enhancer trap markers (254, S2-11) or the hub cell surface marker Fasciclin III (FasIII) revealed that somatic hub cells were present at the apical tips of stet third instar larval testes (Fig. 1F,H). However, the hub often appeared slightly enlarged and less tightly packed than in wild type (Fig. 1E,G), much as in agametic testes from sons of oskar mutant mothers (Gönczy et al., 1992).

In wild-type testes, somatic cyst progenitor cells act as stem cells for the cyst cell lineage (Gönczy and DiNardo, 1996) and lie next to the hub adjacent to germ line stem cells (Fig. 2A). The cyst progenitor cells produce somatic cyst cells, two of which encapsulate each gonialblast and its progeny throughout all subsequent stages of male germ cell differentiation. Somatic cyst cells and cyst progenitor cells were present in loss-of-function stet mutant testes based on the presence of traffic jam (tj) protein (Fig. 2C), a transcription factor detected in nuclei of cyst progenitors and early somatic cyst cells in wild-type (Fig. 2B). The number of Tj-positive somatic cyst cell nuclei detected in stet mutant testes varied, ranging from 20 to 90, compared with 70 to 80 Tj-positive somatic cyst cell nuclei detected in wild-type testes. Somatic cyst cells were also detected in stet mutant testes by several nuclear targeted lacZ enhancer-trap markers (11, 600, 473, data not shown).

Despite the presence of somatic cyst cell nuclei, cyst cells did not appear to envelop germ cells in stet mutant testes. In wild type, somatic cyst progenitors and cyst cells surround early germ cells in a net-like pattern that can be visualized using cytoplasmic cyst cell markers. In wild-type testes, β-galactosidase activity encoded by the 17-en-40 insert (wingless-lacZ enhancer trap marker) was detectable throughout the cell body and cytoplasmic extensions of somatic cyst cells as they surround the developing germ cells (Fig. 2D, arrows). In stet mutant testes stained for the same cytoplasmic cyst cell marker, cyst cells appeared round (Fig. 2E, arrows), with a small percent (10-30%) having detectable short cytoplasmic extensions (Fig. 2E, arrowheads). Similar results were obtained by expressing a cytoplasmic UAS-GFP under the control of the patched-GAL4 (ptc-GAL4) transcriptional activator. In wild-type testes, the GFP-positive cytoplasm of somatic cyst cells formed a net-like pattern surrounding the germ cells (Fig. 2F, arrows). In contrast, in stet mutant testes cyst cells were mostly detected as round GFP-positive structures (Fig. 2G, arrows), and only a few GFP-positive cytoplasmic extensions were observed (Fig. 2G, arrowheads). The number of somatic cyst cells in stet mutant testes detected with cytoplasmic markers (ranging from seven to 46) was lower than the number of cyst cells (20 to 90) detected with the nuclear marker Tj.

Analysis of male germline clones indicated that stet function is required in germ cells. Clones of cells that lack the stet gene were generated in stet/+ animals using a FRT/FLP recombination system (Xu and Rubin, 1993) and identified by lack of expression of a nuclear targeted GFP (see Materials and Methods). GFP is expressed under control of the ubiquitin promotor, allowing for detection of both, the round nuclei of germ cells (Fig. 3A, arrowheads) and the triangular shaped nuclei of somatic cells (Fig. 3A, arrows). Control clones wild-type for stet produced clusters of GFP-negative germ cells (Fig. 3B, circles) that developed normally into spermatocytes, based on appearance by phase contrast microscopy (data not shown) and nuclear size when stained with DAPI (Fig. 3C). In contrast, stet/stet mutant germ cells did not differentiate into spermatocytes. Instead, germline clones produced large clusters of GFP-negative cells (Fig. 3D, circle, Fig. 3F) resembling early germ cells, based on appearance by phase contrast microscopy (data not shown) and by their small, bright nuclei when stained with DAPI (Fig. 3E). The cells in stet mutant clones expressed piwi mRNA (Fig. 3G, arrowheads) and other early germ cell markers normally restricted to the anterior tip of the testis. Staining with esg mRNA and α-spectrin (data not shown) revealed that the cells within individual stet mutant clones were a mixed population resembling stem cells, gonialblasts and spermatogonia, much as the germ cells accumulating in testes from stet homozygous mutant males.

Loss of stet function in the germline affected association with the stet/+ heterozygous somatic cyst cells. Based on GFP-expression, 60% (100 clones tested) of stet mutant germ line clones were not associated with triangular GFP-positive nuclei (Fig. 3F). Triangular GFP-positive nuclei were detected on top, under or at the side of the clusters of stet mutant germ cells in the remaining 40% of the clones. In 30% we detected 1 GFP-positive nucleus. In 10% we detected two GFP-positive nuclei (data not shown). The triangular GFP-positive somatic cyst cell nuclei could have been associated with the stet mutant germ cell clone or with a neighboring stet/+ germ cell cluster. The presence of somatic cyst cells associated with stet mutant germ cell clones was further examined by anti-Tj staining. The nuclear early cyst cell nuclei marker Tj was detected in somatic cyst cell nuclei at the apical tip of wild-type testes, but not in later stage cyst cells. In testes containing stet mutant germ cells clones, many Tj-positive cyst cells were detected at the apical testes tip (Fig. 3H, arrows). However, nuclei expressing the early cyst cell marker Tj were usually not found associated with the cluster of stet mutant germ cells. For 21 out of 25 stet mutant germ cell clones carefully examined in all planes of focus, we did not detect any Tj-positive nuclei in, on top of, under or next to the clone (Fig. 3I). For three of the 25 stet mutant germ cell clones, we observed one associated Tj-positive nucleus (Fig. 3J, arrow). For one out of the 25 stet mutant germ cell clones, we observed two associated Tj-positive nuclei (Fig. 3K, arrows). None of the 25 stet mutant germ cell clones was associated with more than two Tj-positive nuclei.

To test whether somatic cyst cells associated with stet mutant germ cell clones instead expressed late cyst cell markers, we stained for the late cyst cell marker eyes absent (eya). Eya is normally expressed in nuclei of cyst cells associated with spermatogonia. For 50 stet mutant clones examined, we did not detect any Eya-positive cyst cell nuclei on top, within or under the clone. Eya-positive cyst cell nuclei were detected next to the stet mutant germ cell clones (Fig. 3L, arrows). However, these cyst cells could have been associated either with the stet mutant clone or with a neighboring cluster of stet/+ spermatocytes.

Together our analysis indicates that most stet mutant germ cell clones were not associated with the two accompanying somatic cyst cells.

Females mutant for loss-of-function stet alleles that cause severe defects in male germ cell differentiation produced few progeny (one to three adult progeny/female) and showed a variety of defects in oogenesis. In young stet mutant females, 60% of the ovarioles contained egg chambers at several different stages of differentiation (Fig. 4B). DAPI staining (Fig. 4C) and phase contrast microscopy (data not shown) revealed that egg chambers from stet mutant females often contained abnormal numbers and arrangements of germ cells. In older stet mutant females, 90% of the ovarioles usually had only a few egg chambers, which commonly showed signs of degeneration (Fig. 4D). By 2 weeks after hatching, all ovarioles from stet mutant females were mostly empty, except for the germaria, which contained increased numbers of early germ cells as described below. stet mutant females became completely sterile with increasing age.

Early germ cells appeared to accumulate at the apical tip of the germarium in both young and old stet mutant females. In wild-type germaria, germline stem cells lie at the tip, followed by their immediate daughters, the cystoblasts, and then the interconnected cystocytes (Fig. 5A). In wild type, germline stem cells and cystoblasts can be distinguished from later stage germ cells by several subcellular markers. Sex-lethal (Sxl) protein accumulates in the cytoplasm of stem cells and cystoblasts to a much higher level than in later stage germ cells (Fig. 5B, arrow) (Bopp et al., 1993). In addition, α-spectrin is localized to the ball-shaped spectrosome in wild-type stem cells and cystoblasts (Fig. 5D, arrows) but localizes to the branched fusome in cystocytes (Fig. 5D, arrowhead). Germaria from stet mutant females had an elevated number (ranging from the normal six to 75 cells) of early germ cells with cytoplasmic Sxl (Fig. 5C, arrows) and a spectrosome (Fig. 5E, arrows) compared with wild-type germaria (four to six cells). The apparent accumulation of cells resembling stem cells and/or cystoblasts in stet mutant germaria suggests that wild-type function of stet in females plays a role in allowing differentiation of early germ cells.

Wild-type function of stet appears to facilitate the contacts between female germ cells and a population of somatic cells in region 1 and 2A of the germarium. In wild-type and stet mutant germaria, 11 to 12 inner sheath cells were detected in region 1 and 2A of the germarium based on the nuclear targeted lacZ enhancer trap marker I-72 for inner sheath cells (data not shown). These inner sheath cells form cytoplasmic extensions between stem cells, cystoblasts and clusters of cystocytes in region 1 and 2A of the germarium, that can be seen at the ultrastructural level (Fig. 5H, arrows). The cytoplasmic extensions can also be seen upon expression of cytoplasmic GFP (UAS-GFP) under control of either an engrailed-GAL4 (en-GAL4, Fig. 5F) or a ptc-GAL4 transcriptional activator (data not shown). In wild-type germaria, we detected nine to 12 GFP-positive extensions from inner sheath cells between germ cells in region 1 and 2A of the germarium. In germaria from stet mutant females, six to 12 GFP-positive inner sheath cells were present (Fig. 5G, arrows). However, they did not form normal numbers of cytoplasmic extensions. In 50% of the germaria from newly enclosed females, two to eight GFP-positive cytoplasmic extensions from inner sheath cells were detected around or between germ cells. By 1 week after hatching, no cytoplasmic extensions from inner sheath cells were detected in germaria from stet mutant females.

To explore the mechanism of action of stet, we identified the stet gene product by positional cloning. We localized the stet gene to cytological region 62A1 by recombination analysis and deficiency complementation. The stet mutation was uncovered by Df(3L)PX62, which removes ∼60 kb of genomic DNA in 62A1. Analysis of the genome sequence in this region revealed 12 potential transcription units (data not shown, see Materials and Methods).

Sequence analysis of genomic DNA from several EMS induced stet alleles identified stet as a predicted seven transmembrane protein (Table 1, Fig. 6A). The predicted stet protein showed high sequence similarity to the Drosophila rho protein (data not shown) and had been published under the names CT5484 (FlyBase, 1999), rhomboid-2 (rho-2) (Wasserman et al., 2000) and brother of rhomboid (brho) (Guichard et al., 2000). In the following, we will refer to the stet gene product as stet based on naming genes by the mutant phenotype. Two strong stet alleles introduced stop codons in the stet protein-coding region, truncating the predicted protein (Fig. 6A). Another strong stet allele introduced a splice site change resulting in a frame shift that led to a premature stop codon in the predicted stet protein (Fig. 6B). Several other EMS alleles altered conserved amino acids in the predicted stet transmembrane domains (Fig. 6A); stet ^8A, stet ³ and stet ^z3–0369 had amino acid replacements in the conserved protease motif. Comparison of the genomic sequence with several independent cDNAs isolated from a testes cDNA library revealed that the stet testis transcript contained four exons (Fig. 6B). The predicted protein from the stet testes cDNA had stop codons in all three reading frames upstream of the predicted initial methionine, located in exon 2.

We detected stet transcript on northern blots from adult testes, adult ovaries and 0-4 hour embryos. Transcript was not detected on similarly loaded northern blots of mRNA from 4-24 hour embryos (Fig. 7A). Although the stet mutant phenotype clearly demonstrates a requirement for stet function in male germ cells, we did not detected stet mRNA by in situ hybridization or stet protein by immunofluorescence staining of whole testes. The high load required to detect stet transcript on northern blot and the failure to detect stet RNA or protein in whole mount testes suggest that stet is expressed at extremely low level. Similarly, although stet function is clearly required in region 1 and 2A of the germarium, neither stet mRNA nor stet protein accumulated at detectable levels at the tip of the germarium. stet mRNA was detected in germ cells in region 2B and 3 of the germarium (Fig. 7B, arrowhead). Signal from the stet mRNA was extremely low in germ cells of stage 1 and 2 egg chambers, but increased in germ cells of stage 3 to stage 8 egg chambers. In stage 1 to 8 egg chambers, stet mRNA appeared to accumulate in the posterior region of the egg chambers, in the position of the developing oocyte (Fig. 7B, arrows).

The identity of stet was confirmed by rescue experiments. Consistent with stet function in germ cells, expression of a UAS-stet-cDNA in germ cells of stet mutant testes under control of the germ cell-specific driver nanos-GAL4-VP16 restored spermatogenesis (Fig. 8B). As expression with the UAS-GAL4 system is temperature dependent, rescue was not always complete and occasionally clusters of stet mutant germ cells were detected in the testes (Fig. 8B, arrowhead). In contrast, expression of UAS-stet under control of the somatic cell driver ptc-GAL4 did not rescue the stet mutant phenotype (data not shown), suggesting that stet function in germ cells is both required and sufficient to allow germ cell differentiation.

Expression of rho in germ cells of stet mutant testes also restored spermatogenesis (Fig. 8C), indicating that stet and rho may function through the same biochemical mechanism. To explore how stet function in early male germ cells might relate to the Egfr signal transduction pathway, we tested the expression and effects of other components of the pathway on early male germ cell differentiation. Expression of secreted forms of the Egfr ligands spi and grk in male germ cells under the control of the nos-GAL4 activator did not modify the stet mutant phenotype (data not shown), raising the possibility that another ligand may play a role in male germ cells. rho normally acts synergistically with the transmembrane protein Star within the signaling cell to activate spi (Rutledge et al., 1992; Kolodkin et al., 1994; Pickup and Banerjee, 1999; Guichard et al., 1999). In situ hybridization with a Star mRNA probe to wild-type testes revealed high levels of Star expression at the apical tip (Fig. 8D, arrow).

Consistent with a potential role for the Egfr in somatic cells, activated MAP-kinase is detectable in somatic cyst cells of wild-type testes (Fig. 8F, arrowheads) (Kiger et al., 2000). In stet mutant testes, MAP-kinase expression was restricted to the somatic hub cells (Fig. 8G, arrow) and a few (two to five) cells next to the somatic hub in the position corresponding to that of cyst progenitor cells (Fig. 8F, arrowhead). Although cyst cells were present in stet mutant testes (Fig. 2C), we did not detect activated MAP-kinase in cyst cells displaced away from the hub. Likewise, we detected activated MAP-kinase in the cytoplasmic extensions of inner sheath cells of wild-type germaria, but only in a few inner sheath cells in germaria from stet mutant females (data not shown).

The stet gene plays a crucial role in germ cell differentiation in both males and females. In animals that lack stet function, somatic support cells failed to surround germ cells properly and germ cells accumulated at early stages of differentiation. Mosaic analysis in testes suggested that stet function is required in germ cells for normal association between early germ cells and somatic cyst cells. This, along with the molecular identity of the stet gene, suggests that stet activates signaling from germ cells to the soma to allow normal interactions between germ cells and somatic support cells.

The stet gene encodes a homolog of rho, which plays an essential role in Egfr signaling (Sturtevant et al., 1993; Sapir et al., 1998; Wassermann and Freeman, 1998). Rho has recently been shown to localize to the Golgi apparatus, where it acts as a protease to cleave the Egfr ligand spi (Lee et al., 2001; Urban et al., 2001). The stet protein also localized to the Golgi apparatus in cell culture experiments (Ghiglione et al., 2002), and contains the protease motif described for Rho (Urban et al., 2001). Consistent with the idea that stet may encode a protease, three strong stet alleles alter residues in the conserved protease domain.

Mosaic analysis and rescue experiments showed that stet function is required in male germ cells for normal germ cell differentiation. We cannot exclude the possibility that this is a cell-autonomous function of stet within the germ cells. However, the rho family of proteins have been shown to play roles in the production of signals sent to neighboring cells, and do not seem to be directly required for differentiation of the ligand producing cell itself (Golembo et al., 1996; Wasserman et al., 2000). We therefore favor the idea that stet functions primarily by activating signaling from germ cells to somatic cells. We propose that once proper contacts between germ cells and somatic cells are established, signals from somatic cells then regulate germ cell differentiation. We hypothesize that germ cells in stet mutants fail to differentiate because they lack the somatic micro-environment that provides essential cues regulating germ cell differentiation.

Experiments in wing discs demonstrate that stet is able to collaborate with Star to promote signaling through the Egfr/MAP-kinase pathway. stet can activate the Egfr ligands spi and grk when ectopically expressed in wing discs or follicle cells (Guichard et al., 2000; Ghiglione et al., 2002). In this study, we showed that expression of rho in germ cells can substitute for stet function, and that Star is expressed at the tip of wild-type testes. In addition, in stet mutant testes most somatic cyst cells failed to express activated MAP-kinase, the downstream indicator for Egfr signaling. Based on these observations, we propose that the stet gene functions as an activator of signaling from early germ cells to the Egfr presented on the surface of somatic cells and that activation of the Egfr in somatic cells is required to establish normal interactions between germ cells and somatic cells.

Testes from animals carrying a temperature-sensitive allele of the Egfr showed accumulation of germ cells that appeared to be stem cells, gonialblasts and spermatogonia (Kiger et al., 2000). This similarity to the stet mutant phenotype in testes is consistent with stet and the Egfr functioning in the same pathway. However, the Egfr^ts mutant phenotype in testes did not exactly resemble the stet mutant phenotype. In stet mutant testes, somatic cyst cells did not envelope clusters of early germ cells properly. Testes from Egfr^ts mutant animals displayed many defects in the association of somatic cyst cells and early germ cells, including some cases in which germ cell clusters were associated with multiple somatic cyst cells (Kiger et al., 2000). As analysis of the Egfr^ts phenotype was performed after a temperature shift, we hypothesize that testes from Egfr^ts animals may have had sufficient residual Egfr activity to allow some and possibly abnormal association of early germ cells and somatic cyst cells. In addition, the Egfr^ts mutant may not be null for Egfr function at 29°C, the temperature assayed. Consistent with this likely possibility, Kiger et al. (Kiger et al., 2000) have reported a failure to recover cyst cell clones mutant for Egfr null alleles, even though they did detect somatic cyst cells in the Egfr^ts allele. In contrast, in our study we report the phenotype of animals null mutant for stet throughout development.

stet may activate a yet unidentified Egfr ligand to recruit somatic cells for germ cell encapsulation. Even though stet can activate spi and grk when ectopically co-expressed in wing discs (Guichard et al., 2000), expression of secreted forms of spi or grk in male germ cells did not rescue the stet mutant phenotype in our study. Loss-of-function alleles of grk that cause severe defects in female gametogenesis did not show an early germ cell over-proliferation phenotype in testes (C. S., unpublished), suggesting that stet does not function through grk activation. In females, eggs laid by stet mutant mothers did not display the grk or spi mutant phenotypes, but instead either showed a variety of defects or developed into phenotypically normal adults. Further investigation of the Egfr signal transduction pathway remains to be undertaken to identify additional components of the pathway and test their potential role in interactions between early germ cells and surrounding somatic cells.

The early stages of gametogenesis in Drosophila are strikingly similar in males and females in that, in both sexes, germ cells are in constant contact with encapsulating somatic cells. Based on ultrastructural studies by electron microscopy and light microscopy analysis using several markers (Margolis and Spradling, 1995) (this study), the inner sheath cells in region 1 and 2A of the germarium appear to form cytoplasmic extensions that contact female germ line stem cells, cystoblasts and cystocytes. Our study provides the first insight into the function of the inner sheath cells. In stet mutant females, the cytoplasmic extensions from the inner sheath cells failed to surround the germ cells and the germ cells failed to progress to the cystocyte stages. We propose that the inner sheath cells at the tip of the germarium may play a role similar to the somatic cyst cells surrounding germ cells in testes, providing a microenvironment that directs female germ cell differentiation.

Our data predict a new function for the Egfr signaling pathway in the female gonad. Egfr signaling, activated by stet, may be required to set up the normal interactions of early female germ cells and somatic inner sheath cells in region 1 and 2A of the germarium. We did not observe accumulation of early germ cells with cytoplasmic Sxl protein and spectrosomes at the tip of the germarium after shifting animals carrying the Egfr^ts allele to the restrictive temperature as adults (data not shown). However, many ovarioles from females carrying Egfr^ts alleles also did not display defects at later stages of oogenesis in these experiments, again indicating that the Egfr^ts allele had residual Egfr activity and may not reflect the Egfr loss-of-function situation.

We cannot rule out the possibility that stet activates other signaling pathways to set up proper physical interactions between germ cells and somatic support cells. In females, normal encapsulation of germ cells by somatic follicle cells requires the neurogenic genes brainiac (brn) and egghead (egh). brn encodes a secreted protein (Goode et al., 1996a) and egh encodes a secreted or transmembrane protein (Goode et al., 1996b). Double mutant combinations of grk and brn displayed much stronger defects in encapsulation of germ cells than either grk or brn mutants alone, suggesting that the brn and egh pathway and the Egfr pathway function partially redundantly in formation of the prefollicular epithelium (Goode et al., 1992). This opens the possibility that encapsulation of early female germ cells by inner sheath cells and encapsulation of male germ cells by somatic cyst cells depend on another signaling pathway instead of, or in addition to the Egfr signal transduction pathway.

We thank J. Hackstein, B. Wakimoto and D. Lindsley; E. Bier, S. DiNardo, M. Steinmann-Zwicky, T. Schüpbach and the Drosophila Bloomington stock center for Drosophila strains. Anti-Tj-D1 antibody was a generous gift from R. Avancini and D. Godt; anti-Eya antibody was from S. DiNardo; esg DNA was obtained from S. Hayashi; and piwi DNA was obtained from H. Lin. stet cDNAs were isolated from a testes library constructed by the Berkeley Drosophila Genome Project from testes mRNA isolated by members of the Fuller laboratory. The authors especially thank A. T. Carpenter for the electron microscope image and M. Freeman for sharing results prior to publication. We thank S. Alexander, B. Bolival, A. A. Kiger and T. Maa for technical assistance. The authors are grateful to D. Godt, P. J. Langer, T. Schüpbach and members of the Fuller laboratory for helpful discussions and comments on the manuscript. This work was supported by an EMBO postdoctoral fellowship to C. S. (ALTF700-1995), a Lilly Fellowship of the Life Sciences Research Foundation and a Stanford Reproductive Biology Training Program fellowship (HD07493) to D. L. J., funding from the NIH Reproductive Scientist Developmental Program and the Burroughe Wellcome Fund to S. I. T., and by NIH grant DK53074 (subproject #4) to M. T. F.