The sex determination master switch, Sex-lethal, responds to Hedgehog signaling in the Drosophila germline

Sex-lethal (Sxl) is the sex determination master switch and it controls somatic sexual development. It is a splicing and translational regulator. Sxl is activated in females where the X chromosome to autosome (X:A) ratio is 1. It remains ‘off’ in males where the X:A ratio is 0.5. Once set, these two modes of Sxl expression are maintained through the rest of the life cycle (Sanchez and Nothiger, 1983; Cline, 1984). In females, a positive autoregulatory feedback loop that functions through alternative splicing maintains the on state (Bell et al., 1991). The male mode of splicing is maintained by default.

Sxl directs somatic sexual development by controlling the dosage compensation system (Lucchesi, 1978) and the somatic sexual differentiation pathway (reviewed by Cline and Meyer, 1996). The dose compensation system is turned off by Sxl, through both splicing regulation and translational repression (Bashaw and Baker, 1997; Kelley et al., 1997). In somatic sexual development, Sxl promotes female differentiation by controlling the female specific splicing of the transformer gene (Boggs et al., 1987; McKeown et al., 1987).

The final requirement of Sxl is in oogenesis. Pole cell transplantation experiments have demonstrated that female germ cells require Sxl for regulating mitosis. Germ cells that lack Sxl develop as tumorous cysts of many small undifferentiated cells (Schupbach, 1985), a phenotype that is shared by the female sterile alleles of Sxl. Additionally, Sxl regulates the splicing of its own transcripts (Hager and Cline, 1997), as well as the female-specific process of meiotic recombination (Schütt et al., 1998; Bopp et al., 1999). The germline targets of Sxl are not known, but they are not the somatic ones (Marsh and Wieschaus, 1978; Schupbach, 1985).

Drosophila ovaries are made up of 14-16 ovarioles with the germarium as the anteriormost structure. Germline stem cells reside in the anteriormost region of the germarium, immediately under the terminal filament (Fig. 1A). Each stem cell divides to form a cystoblast and another stem cell; the cystoblast undergoes four synchronized divisions with incomplete cytokinesis to form a cyst of 16 interconnected cells. Cysts move down the germarium and cells of somatic origin, the follicle cells, surround them pinching off an egg chamber at the posterior end of the germarium.

The five to nine germ cells (stem cells and early cystoblasts) immediately under the terminal filament, show high levels of Sxl protein primarily in the cytoplasm (Fig. 1B). As these cells mature, there is a marked drop in the level of cytoplasmic Sxl. After this downregulation, Sxl is detected in mid-region 1 of the germarium as bright nuclear foci (Bopp et al., 1993). The intensity of Sxl in the nucleus and cytoplasm increases as the germ cells progress down the germarium (Figs 1B, 2A). The remarkable change in both localization and levels of Sxl in early germ cells suggests that the process is of functional significance. We demonstrate that some of these changes are controlled by the Hedgehog (Hh) signaling pathway.

Hh specifies cell fate and patterning in the development of several different tissues. It is a secreted protein that binds to its receptor, Patched (Ptc). This relieves Smoothened (Smo) from inhibition by Ptc, enabling Smo to activate Cubitus interruptus (Ci), a transcription factor. Cells that have not been exposed to Hh, have the predominant form of Ci as a 75 kDa isoform (Aza-Blanc, et al., 1997), which acts as a transcriptional repressor. Phosphorylation of Ci by protein kinase A (PKA) promotes processing of Ci to the 75 kDa isoform. In the presence of Hh, phosphorylation of Ci is reduced to generate the full-length 155 kDa isoform (Chen, et al., 1999a) that activates transcription of wingless, decapentaplegic and ptc (reviewed by Ingham, 1998).

The processing and nuclear import of Ci is regulated via a complex of Ci with the cytoplasmic members of the Hh signaling pathway, Costal-2 (Cos2; Cos – FlyBase), Fused (Fu) and Suppressor of Fused (Su(fu); Robbins et al., 1997; Sisson et al., 1997; Stegman et al., 2000). Fu appears to be a serine threonine kinase (Therond et al., 1996), Cos2 has sequence similarity to the motor domain of kinesin, while Su(fu) shows no homology to any known protein (Preat et al., 1993). The Ci-containing complex is thought to be tethered to microtubules by Cos2. On Hh signaling, the complex is released from microtubules and full-length Ci enters the nucleus (Robbins et al., 1997; Sisson et al., 1997).

We report here that Ptc and Fu in the germarium undergo changes in expression that are coincident with Sxl. Immunoprecipitations from ovarian extracts show that Sxl is in a complex with Fu and Cos2, along with β- and γ-tubulin. We propose that γ-tubulin together with Cos2, tether Sxl to microtubules maintaining Sxl in the cytoplasm until Hh, or an effector of the Hh signal, releases Sxl from the complex. Proteasome inhibitor studies and germline clones of hh pathway components suggest that the levels of Sxl in early germ cells are decreased in response to the Hh signal, most likely limiting the amount of the protein that enters the nucleus.

hh^ts2 is a temperature-sensitive allele, hh^AC is a strong loss-of-function allele (kind gifts from P. Beachy). hs-hhM11 (synonym hh^hs.PI) is a construct of the full-length coding sequence of hh under the hsp70 promoter (a gift from P. Ingham; Ingham, 1993). Sxl^f4, Sxl^fP7BO, Sxl^M4, cos2^W1 (synonym cos¹; kind gifts from K. Ho and M. Scott), Su(fu)^LP (a gift from D. Kalderon) and fu³³ are described in FlyBase (http://flybase.bio.indiana.edu). Stocks used to generate clones were hsp70flp, FRT40A DCO^H2/CyO, FRT40A smo²/CyO and FRT40A armlacZ (generously provided by D. Kalderon). R. Steward provided FRTG13 armlacZ/CyO. OreR was used as the wild-type control.

Transheterozygous hh flies were produced by crossing hh^AC/TM3 to hh^ts2/TM3 flies at 17°C. hh^AC/hh^ts2 progeny were shifted to 29°C for more than 7 days. fu³³ flies were raised at 17°C until eclosion and shifted to 29°C for 7 days. Homozygous hs-hhM11 flies were heat shocked in a 37°C water bath for 1 hour every 12 hours for 3 days, and ovaries dissected within 2 hours of the final heat shock.

Germline clones were generated by the FLP-FRT mitotic recombination system (Chou and Perrimon, 1992). Recombination was induced in first and second instar larvae with a 1 hour heat shock at 37°C to induce hs-flp. Clones were created for smo² (cold-sensitive amorph), DCO^H2 (Pka-C1 catalytic subunit null) and cos2^W1, and marked by armlacZ. Clones were generated in the genotypes PKA (hsp70-flp; FRT40A DCO^H2/FRT40AarmlacZ); smo (hsp70-flp; FRT40A smo²/FRT40A armlacZ); and cos2 (hsp70-flp;FRTG13 cos2^W1/FRTG13 armlacZ)

Ovaries were fixed and stained (Bopp et al., 1993). Fifty or more ovarioles from a minimum of two separate stainings were counted from at least six different pairs of ovaries. Anti-Fu (D. Robbins; Robbins et al., 1997), anti-Cos2 (K. Ho and M. Scott; Sisson et al., 1997), anti-Vasa (P. Lasko) and anti-Sxl (ascites, P. Schedl) antibodies were used at 1:500, anti-α-Spectrin was used at 1:50 (Developmental Studies Hybridoma Bank) (Byers et al., 1987), and anti-Orb at 1:20 (P. Schedl). Anti-Patched antibody (I. Guerrero) was used at 1:200; anti-γ-tubulin antibody at 1:1000 (Sigma). This antibody does not distinguish the isoforms of γ-tubulin, but only γ-tub23 is in the germarium (Wilson et al., 1997). Biotinylated goat anti-rabbit and sheep anti-rat antibodies (Amersham Life Sciences) were used at 1:200. Biotinylated donkey anti-mouse antibody (Jackson ImmunoResearch Laboratories) was used at 1:1000. Cy™3-conjugated (Jackson ImmunoResearch Laboratories) Alexa™594 (Molecular Probes) and Oregon Green™-conjugated streptavidin (Molecular Probes) were used at 1:1500. Yo-Pro™ (Molecular Probes) nucleic acid stain was used at 1:8000 for 5 minutes. Propidium Iodide (Molecular Probes) was used at 1 μg/ml for 15 minutes. Before nucleic acid staining, ovaries were treated with RNaseA for 15 minutes in phosphate-buffered saline. The ovaries were mounted in aquaPolyMount (Polysciences).

RNA extraction, RT, PCR and Southern blot analysis were performed (Bopp et al., 1993) but a mixture of Pfu DNA polymerase (Stratagene) and Klentaq1 (Ab Peptides) was used (Barnes, 1994).

Approximately 75 pairs of Sxl^f4 or OreR ovaries, or 75 Sxl^fP7BO (Sxl null) males were homogenized in IP buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 0.05% NP40, 1 μg/ml leupeptin, 1 μg/ml pepstatin, 10 μg/ml Aprotinin, and 10 mg/ml PMSF). Colchicine was supplemented to 1 μg/ml when used. The homogenate was on ice for 10-15 minutes and cellular debris then removed by centrifugation. Extracts were incubated for 4 hours with protein A beads (Pharmacia Biotech) crosslinked (by DMP (Sigma)) to either anti-Sxl, anti-Bic-D or anti-γ-tubulin antibodies. The beads were pelleted and supernatant saved as post-IP lysate. The beads were washed with supplemented IP buffer (300 mM NaCl). Immunoprecipitates and 10% of the post-IP lysates were run on a 10% SDS-polyacrylamide gel and transferred onto nitrocellulose membrane (Life Technologies). Membranes were blocked in 5% nonfat dry milk (Carnation). Anti-Fu, anti-Sxl, anti-phosphoglucomutase (gift of D. Bedwell) and anti-γ-tubulin antibodies were used at 1:500. Anti-Cos2 was used at 1:200. Anti-Bic-D antibodies (4C2 and 1B11, R. Steward) were each used at 1:20. Horseradish peroxidase-conjugated goat anti-rabbit, goat anti-mouse and goat anti-rat antibodies (Southern Biotechnology Associates) were used at 1:10,000. The western blot signals were detected using ECL (Amersham Life Sciences). Quantitation of the protein in the immunoprecipitates relative to the protein in the lysate was determined using Image Quant™ from Molecular Dynamics.

Ovaries from OreR (4 day old) or heat-shocked hs-hhM11 females (as above) were rocked at room temperature for 3 hours in 250 μM MG132 (Calbiochem), 250 μM ALLN (N-Acetyl-Leu-Leu-Norleucinal, Sigma), 0.1 mg/ml colchicine or DMSO (1:200) in DS2 medium (Mediatech).

Tumorous cysts in ovaries of females transheterozygous for a fu deficiency (Df(1)fu) and fu¹, show Sxl localized primarily in the cytoplasm. Homozygotes of the temperature-sensitive (ts) fu allele (fu³³) kept at the non-permissive temperature, also show an increase in the number of germ cells with cytoplasmic Sxl (Bopp et al., 1993); see Fig. 2G). These data suggest that Fu is necessary for reducing high levels of cytoplasmic Sxl and that the Hh pathway may be involved in regulating Sxl distribution.

As an initial test of whether there was a link between the Hh signaling pathway and Sxl localization, we examined the expression pattern of several genes in the pathway. Fig. 2 shows the expression of Fu, Ptc and Cos2 in a wild-type germarium. Fu and Ptc proteins appear to mimic the changes in Sxl expression. With the exception of the nuclear foci, Fu shows a striking resemblance to Sxl (Fig. 2A-C). The temporal distribution of Ptc also resembles Sxl but, as previously reported (Forbes et al., 1996b), the staining is in aggregates (Fig. 2D). Like Fu, Ptc is not detected in the nucleus and shows a decrease beyond the stem cells and early cystoblasts. This downregulation of Fu and Ptc is followed by an increase of the proteins in the cells that show an increase in Sxl. This pattern of expression is not seen for all proteins, for example, Snf, a general splicing factor, is expressed uniformly throughout the germarium (see Bopp et al., 1993). Cos2 is also uniform in the cytoplasm of all germ cells with no downregulation in the germarium (Fig. 2E).

The cytoplasmic expression of γ-tubulin (isoform at 23C, γ-tub23) also strongly resembles the first phase of Sxl expression (see Fig. 2F and Wilson et al., 1997). γ-tub23 is located on the centrosomes of all follicle cells and the mitotic germ cells, but cytoplasmic γ-tub23 is restricted to the early germ cells. Double staining of wild-type germaria with anti-Sxl and anti-γ-tub23 antibodies confirms that the clearing of cytoplasmic γ-tub23 occurs in the same cells that show a decrease in cytoplasmic Sxl (not shown). The centrosomal expression of γ-tub23 shows no correlation with Sxl expression and persists until the four mitotic divisions are complete (Fig. 2F; Wilson et al., 1997).

To determine whether there is a connection between the expression of Sxl and fu, we examined the distributions of the two proteins under conditions where each gene was altered. fu activity was reduced by maintaining homozygous fu³³ females at the non-permissive temperature (29°C) for 7 days and the ovaries stained for Fu and Sxl. We find that both proteins are not downregulated normally in mid-region 1 of the germarium (Fig. 2G-I) and Sxl cytoplasmic expression persists beyond the region that contains the early germ cells. Not all of the germ cells in the fu^ts germarium show the altered localization of the two proteins, most likely a reflection of residual Fu activity. It is worth noting, however, that expression of Sxl is altered in all germ cells where the expression of Fu is altered.

This phenotype of persistent high levels of cytoplasmic Sxl resembles that of the Sxl female sterile mutations, Sxl^f4 and Sxl^f5. The Sxl female steriles are normal for somatic function, but the germ cells appear arrested at an early stage in oogenesis with high levels of cytoplasmic Sxl (Fig. 2K), as seen in wild-type germline stem cells and early cystoblasts (Bopp et al., 1993). The mutation in Sxl^f4 changes a proline to a serine in a stretch of Sxl that is proline rich, while Sxl^f5 alters a proline to a leucine in the same region (Bopp et al., 1993).

The assumption that the germ cells in Sxl^f4 are arrested at an early stage is supported by their staining pattern with anti-α-Spectrin antibodies (Fig. 2J). Spectrin antibodies stain the fusome, a germ cell-specific organelle that is spherical in stem cells and early cystoblasts (spectrosome) and branches as the 16-cell cyst develops (reviewed by McKearin, 1997; Fig. 1C). Sxl^f4 germ cells show mainly the spectrosome (Fig. 2J) and Fu expression is affected as all the germ cells show a constant level of cytoplasmic Fu (Fig. 2L).

Fu and Sxl show the same localization in a background that is mutant for either gene. While this co-dependent expression does not demonstrate a functional link, the developmental arrest caused by either gene would appear to impact the process(es), which regulates their distributions.

If the Hh signaling pathway regulates Sxl, then changes in Hh expression should affect the distribution of Sxl. We tested this by either reducing Hh (using a ts allele of the gene) or by overexpressing Hh (using a transgene of the hh cDNA under heat-shock control; hs-hhM11) (Ingham, 1993). As hh positively regulates fu, and fu mutants give rise to persistent cytoplasmic Sxl, overexpression of Hh is predicted to decrease the amount of cytoplasmic Sxl in early germ cells. Conversely, removing Hh should resemble the loss of fu and result in an increase in germ cells with cytoplasmic Sxl.

hh null over ts allele (hh^AC/hh^ts2) females were shifted to 29^oC after eclosion, a condition that reduces the number of follicle cells and produces egg chambers with greater than 16 germ cells (Forbes et al., 1996a; Zhang and Kalderon, 2000, Fig. 3A-E). These hh^AC/hh^ts2 ovaries contain germaria and early egg chambers with persistent high levels of cytoplasmic Sxl (Fig. 3A), and no visible nuclear foci, even though the germ cells appear to have differentiated past the stage where Sxl is normally strongly cytoplasmic. Ptc expression, like Sxl, is not decreased when Hh activity is reduced (Fig. 3B).

That the germ cells have differentiated, albeit abnormally, is suggested by the presence of branched fusomes in germ cells close to the terminal filament (Fig. 3C). In wild-type ovaries, the branched form of the fusome is seen further down the germarium (Fig. 1C). Differentiation is also suggested by the presence of polyploid nurse cell nuclei and the expression of Orb protein. Wild-type Orb expression begins at the very end of germarial region 1. It localizes to the future oocyte in region 2 and is associated with the oocyte throughout ovarian development. Orb expression in hh^ts2/hh^AC females occurs immediately after the stem cell stage (compare Fig. 1D with Fig. 3D-E). The accumulation of Orb is abnormal, showing more than one germ cell with high levels of the protein. This is consistent with the fusome expression, suggesting that the germ cells may be more advanced in development than normal.

Later chambers of hh^ts2/hh^AC ovaries show little or no Sxl in their germ cells (Fig. 3F). As Sxl expression requires the productive splicing of its transcripts via an autoregulatory feedback loop (Bell et al., 1991; Hager and Cline, 1997), retaining Sxl in the cytoplasm should affect the splicing efficiency of its transcripts. Without nuclear Sxl, male-spliced mRNAs are produced, resulting in the absence of the protein. Male-spliced Sxl mRNAs are detected in ovaries of certain female steriles, such as Sxl^f4, where Sxl is mainly cytoplasmic (Fig. 3H) (Bopp et al., 1993), but not in wild-type ovaries. Fig. 3H shows that in ovaries with reduced Hh, significant amounts of male Sxl mRNA are detected. Female Sxl mRNA is also detected, presumably from the somatic cells and the older chambers present before the heat shock.

These data suggest that without the Hh signal, Sxl is retained in the cytoplasm at a level that compromises the splicing feedback loop. A constitutive allele of Sxl might be predicted to restore Sxl expression in these late stage egg chambers. When Sxl^M4, the strongest of the Sxl constitutive alleles (Bernstein et al., 1995), is introduced into the hh^ts2/hh^AC background, Sxl expression is maintained and the protein remains cytoplasmic (Fig. 3G).

Overexpressing Hh has been shown to produce an effect opposite to the hh^ts condition. The follicle cells overproliferate, forming extended stalks between egg chambers and some egg chambers have fewer than sixteen germ cells because the increased number of follicle cells encapsulates the egg chambers prematurely (Forbes, 1996a). When Hh is overexpressed, the downregulation of Sxl, Fu (Fig. 4A-B) and γ-tub23 (not shown) appears to occur earlier than normal. Only one or two cells immediately under the terminal filament show high levels of cytoplasmic Sxl. Additionally, we detect more Sxl in the nuclei of these cells. We also find Orb expression comes on earlier in the germarium – late region 1 – and that the protein fails to show a distinct accumulation in the presumptive oocyte (Fig. 4C).

Beyond the cells immediately under the terminal filament, the germ cells show very little Sxl protein and the majority of the staining comes from the surrounding follicle cells. As might be expected when little Sxl protein is available for splicing regulation, these ovaries show Sxl mRNA spliced in the male mode (Fig. 3H). Flies with a single copy of the hs-hhM11 and the constitutive allele of Sxl, Sxl^M4, also show reduced levels of Sxl protein in early germ cells (less dramatically than the wild type with 2 copies of the hs-hhM11). The low expression of Sxl is also seen in late egg chambers (not shown), suggesting that the constitutive expression of Sxl cannot significantly increase Sxl levels when Hh is overexpressed.

As overexpression of Hh appears to decrease the levels of Sxl, we introduced the hs-hhM11 transgene into a Sxl^f4 background to see if increasing the levels of Hh would rescue the Sxl^f4 phenotype. The Sxl^f4 protein is refractory to Hh as the hs-hhM11 transgene has no effect. This result places Sxl downstream of hh.

Ovaries that overexpress Hh also show strongly reduced Ptc in the germarium and the bulk of the signal observed is in the somatic cells (Fig. 4D-E). The signal is in aggregates and late stage chambers appear unaffected. Overexpressing Hh also reduces the levels of Spectrin, suggesting that the fusome is compromised (Fig. 4F). Previously, Ptc was described as accumulating in the fusome (Forbes et al., 1996b). The lack of Ptc and Spectrin when Hh is overexpressed suggests that assembly or maintenance of the fusome may be dependent on Hh signaling.

As hh and fu affect Sxl expression in the germline, we wanted to test whether other components of the Hh signal transduction pathway would affect Sxl. The Hh pathway has been shown to regulate the development of the follicle cells. To demonstrate a direct involvement of the pathway in germ cells, we generated mutant germ cell clones. The FLP/ FRT mitotic recombination system (Chou and Perrimon, 1992) was used, because, with the exception of Su(fu), which has no overt phenotype, all strong loss-of-function mutations in the hh signaling pathway are lethal.

To ensure that germline stem cells would be affected, clones of cos2, smo and PKA were generated by heat shocking first or second instar larvae. Clones were marked by the absence of armadillo-lacZ (see Materials and Methods), which is expressed in both somatic and germ cells. The clones described were mutant for germ cells and wild type for surrounding follicle cells. Changes in Sxl localization should thus reflect the removal of a component of the hh pathway in only the germ cells.

Early germ cells that were mutant for cos2W1 (Sisson et al., 1997) showed severely reduced levels of Sxl (Fig. 5A-C). In the section shown, there is only one early germ cell that is not mutant for cos2 (β-galactosidase positive). This cell expresses the normal high levels of cytoplasmic Sxl. The other early germ cells are cos2 negative, and they express very low levels of Sxl. Further down the germarium, mutant germ cell clusters appear to have normal nuclear foci of Sxl protein. These clusters are surrounded by follicle cells and are pinched off normally. They continue to develop relatively normally to about stage 6 or 7 and then become necrotic.

We next examined the effect of smo mutations. Because Smo transduces the Hh signal in somatic tissues, we anticipated that loss of Smo would result in retention of cytoplasmic Sxl, as seen in the germ cells of hhts ovaries. The allele of smo we used (smo2) is cold sensitive (FlyBase); at 25°C it has reduced activity and is considered amorphic at 18°C. Early germ cell clones of smo2 show the normal levels of Sxl and downregulation of Sxl when the flies are raised at 25°C or 18°C (not shown). No significant developmental defects are seen in later chambers either. This suggests that a protein other than Smo might transduce the Hh signal in early germ cells. Alternatively, the Hh signal may be indirectly transduced to the germ cells via a downstream somatic effector.

Mitotic clones of two alleles of PKA (PKAH2, a null, and PKAH3, a hypomorph) were examined. PKA is not part of the cytoplasmic complex of the hh pathway, although PKA phosphorylation of Ci results in the partial degradation of Ci to generate the 75kD repressor form. PKAH2 clones in early germ cells result in an increase in nuclear localization of Sxl, as judged by the size and intensity of the nuclear foci (Fig. 5D-F). In the section shown, two of the early germ cells are PKA negative (negative for β-galactiosidase) and contain very visible nuclear foci of Sxl, with slightly reduced levels of cytoplasmic Sxl. The adjacent cell, which is not a clone, has no clear nuclear focus of Sxl. PKA mutant germ cells further down the germarium also appear to have brighter nuclear foci. Clones of the weaker allele, PKAH3 are similar to PKAH2, but the increase in intensity of the nuclear foci is not as obvious (not shown). This result suggests that PKA functions to either inhibit Sxl nuclear entry or maintain Sxl in the cytoplasm. The mutant germ cell clusters appear to pinch from the germarium and develop normally.

Mutations in Su(fu), even with the amorphic allele Su(fu)LP, are not lethal and show no patterning defects unless in the presence of a cos2 or fu mutation (Préat et al., 1993). Although many egg chambers of homozygous Su(fu)LP females are normal, there are also a number of degenerating egg chambers (Fig. 5G). To test for an interaction between Sxl and Su(fu), we examined females homozygous for both Sxlf4 and Su(fu)LP. These females remain sterile, and most of the egg chambers contain degenerating germ cells (Fig. 5H). This germ cell death is much more prevalent than in the homozygous Su(fu)LP ovaries. The germ cells appear to detach from the follicle cells and degenerate. This effect could simply arise from an exacerbation of the system caused by two molecules important to germ cell viability, or may reflect an interaction between Su(fu) and Sxl.

The similar temporal patterns of expression of Sxl, Ptc and Fu in the germarium, and the effects of the germline clones strongly suggested that Sxl might directly interact with some of the cytoplasmic components of the hh signaling pathway. Such an interaction, particularly with Cos2, which appears necessary for the high levels of cytoplasmic Sxl in early germ cells, would provide the mechanism for maintaining Sxl in the cytoplasm.

To test this hypothesis, we determined whether the various cytoplasmic Hh signaling components were present in immunoprecipitates of Sxl. Fig. 6A shows that immunoprecipitation of Sxl from hand dissected wild-type and Sxlf4 extracts co-immunoprecipitates both Fu and Cos2. As a control for specificity, the immunoprecipitations were carried out using extracts from males with a null allele of Sxl (SxlfP7BO, Salz et al., 1987). Neither Fu nor Cos2 (not shown) was detected in these immunoprecipitates. As additional controls, we tested whether the Sxl immunoprecipitates from wild-type ovaries contained Bicaudal-D (Bic-D) or phosphoglucomutase. Bic-D was tested because it has sequence similarity to the tail domain of myosin heavy chain and the microtubule motor kinesin, making it similar to Cos2 (Suter et al., 1989; Wharton and Struhl, 1989). Phosphoglucomutase was tested because it is a highly expressed cytoplasmic enzyme. Neither Bic-D (Fig. 6A) nor phosphoglucomutase (not shown) were detected in the Sxl immunoprecipitates. Additionally, when the immunoprecipitation is carried out with Bic-D antibodies, neither Fu nor γ-tubulin (discussed below) was detected, indicating that their presence in the Sxl immunoprecipitates is specific (Fig. 6B).

As the timing for the downregulation of cytoplasmic γ-tubulin is similar to that of Sxl and Fu, we examined whether γ-tubulin was also in the Sxl-Fu-Cos2 complex. Immunoprecipitates of Sxl from wild-type and Sxlf4 ovaries contain γ-tubulin (Fig. 6A). Surprisingly, wild-type immunoprecipitates also show γ-tubulin. Compared with the total germ cell contribution, the proportion of the stem cells and early cystoblasts is very small in wild-type ovaries. This result is readily explained, however, if we assume that the isoform of γ-tubulin we immunoprecipitate is the γ-tub37 isoform, which is expressed from stage 5 onwards in ovaries. Taken with the presence of Fu and Cos2 in the Sxl immunoprecipitates from wild-type ovaries, the data suggest that Fu, Cos2, Sxl and γ-tubulin are in a complex not only in early germ cells but also in most of the germ cells of the ovary.

To test the specificity of the γ-tubulin immunoprecipitation results, we did the immunoprecipitation in reverse, using an anti-γ-tubulin antibody. Sxl is in the immunoprecipitate, indicating that γ-tubulin does indeed bind to a complex containing Sxl (Fig. 6B).

The Sxl immunoprecipitates also contain β-tubulin (not shown). Interestingly, when the microtubule destabilizer, colchicine, was added to the extracts before immunoprecipitation, γ-tubulin and Fu continue to immunoprecipitate with Sxl but β-tubulin does not (not shown). It appears that the γ-tubulin interaction in the Sxl-Cos2,-Fu complex is different from the β-tubulin interaction. The colchicine sensitivity of β-tubulin presumably reflects the tubulin-tubulin interaction between γ and β. This interaction could serve to tether the Sxl-Fu-Cos2-γ-tubulin complex to the microtubule network.

In somatic cells, the Fu-Cos2 complex contains Ci protein (Robbins et al., 1997; Sisson et al., 1997). Ci is not detected in germ cells but is present in the surrounding somatic follicle cells. Ci protein is not detected in the Sxl immunoprecipitates (not shown). This may be because the contribution of the follicle cells in the extracts is relatively small compared with the germ cells. If, however, Sxl precludes Ci from being in the same complex with Cos2 and Fu, then Ci should not be present in the Sxl immunoprecipitates, as observed.

To determine whether the downregulation of Sxl in the germarium involves the proteasome, wild-type ovaries were treated with proteasome inhibitors (MG132 and ALLN; Rock et al., 1994; Palombella et al., 1994). Treatment with both inhibitors shows an increase in Sxl in the cytoplasm and nucleus of early germ cells of wild-type germaria, but does not affect the strong downregulation of Sxl (Fig. 7B). (Note that in other samples we see an increase in Sxl in the somatic cells at the distal end of the germarium, see Fig. 7E.) As a control, the ubiquitous germ cell-specific protein, Vasa, was examined. The same treatment does not affect Vasa in the germarium (not shown).

As the Hh pathway regulates the partial proteolysis of Ci through the proteasome (Chen et al., 1999b), Hh might also be involved in the turnover of Sxl. Hh is expressed in the terminal filament and cap cells and its overexpression decreases the number of early germ cells with high cytoplasmic Sxl. If Hh regulates the proteolysis of Sxl, then treatment of ovaries with the MG132 inhibitor should counteract the effects of Hh overexpression. MG132 sustains high levels of cytoplasmic Sxl in early germ cells when Hh has been ectopically induced (compare Fig. 7D with 7E). It would appear that the major outcome of the Hh signal is to decrease Sxl in the early germ cells. This effect includes the germ cells immediately under the terminal filament, as the levels of Sxl in these cells are increased in wild-type germaria by the addition of the proteasome inhibitors (Fig. 7B).

Cos2 is thought to sequester the Ci-containing complex in the cytoplasm by binding to microtubules. To test if Sxl is also affected by the microtubules, we treated wild-type ovaries with colchicine, a microtubule destabilizer. This treatment produces the same effect as the proteasome inhibitors, increasing the levels of Sxl in the early germ cells (Fig. 7C). The effect of colchicine is not observed in all germ cells but appears restricted to the germ cells near the terminal filament. The signal intensity of Sxl is so great that the scan shows little definition of the later germ cells. When wild-type ovaries are treated with taxol, a microtubule stabilizer, no significant effect is observed (not shown). As colchicine produces essentially the same effect as the proteasome inhibitors, we tested whether colchicine could counteract the effects of Hh overexpression on Sxl. This appears to be the case; colchicine re-establishes wild-type levels of Sxl in the early germ cells (Fig. 7F). These data suggest that the decrease of Sxl in early germ cells is dependent on microtubules, and is triggered by Hh.

As the effects of Hh overexpression are rapidly reversed (3 hours of incubation with proteasome inhibitors or colchicine), the change in number of cells with high levels of cytoplasmic Sxl must be occurring in situ. This timescale precludes significant cell division (and subsequent differentiation), as stem cells divide about every 48 hours (Wieschaus and Szabad, 1979) but the cystoblast completes its four mitotic divisions in 24 hours (Spradling et al., 1997). We also find that the effects on the fusome are reversed <5 hours after overexpressing Hh (C. V., unpublished). A change in gene expression, rather than an advancement in differentiation of the early germ cells, must be occurring when Hh is overexpressed.

Sxl is essential to oogenesis, but unlike the soma, the targets of Sxl in the germline have remained elusive. Sxl functions in early germ cells as a regulator of mitosis, and is necessary later for the progression of the meiotic cycle (Schupbach, 1985; Bopp et al., 1999). As early germ cells mature, a distinct change in both localization and levels of Sxl takes place. As these changes occur during the early mitotic divisions of the cystoblasts, it seems reasonable to suppose that the process is regulated and is related to Sxl function. That the Sxl female sterile alleles have ovaries with cysts of many small, undifferentiated cells with high levels of cytoplasmic Sxl, is consistent with this view.

In this study, we examined the effect of the hh signaling pathway on the distribution and levels of Sxl in early germ cells. Various genetic and biochemical observations indicate a connection between Hh and the Hh pathway downstream components with Sxl expression.

One of the first suggestions that the Hh pathway may be linked to the regulation of Sxl, was the similarity in expression of Fu and Ptc to that of Sxl in early germ cells. Although their internal cellular distributions differ, particularly that of the membrane protein Ptc, both mimic the changes Sxl shows as the germ cells mature. Sxl, Fu and Ptc are downregulated, then their levels are increased in the same cells.

We found that the cytoplasmic expression of γ-tub23 in early germ cells also disappears in the same cells that dramatically reduce cytoplasmic Sxl. Unlike Sxl, Fu and Ptc, the cytoplasmic expression of γ-tub23 does not reinitiate further down the germarium. Despite this difference in late expression, other data suggest that γ-tubulin may be part of the cytoplasmic components of the Hh pathway.

Our immunoprecipitation analyses show both Fu and Cos2 are complexed with Sxl. As the extracts were from hand-dissected ovaries, we could not generate sufficient material to prove (e.g. by fractionation) that the proteins were all in a single complex. We suppose that, as in patterning, the proteins are in one complex and that Sxl replaces Ci, which is not present in germ cells.

We also found γ-tubulin associated with Sxl in Sxlf4 ovaries, which express only the γ-tub23 isoform, and in wild-type ovaries, which express the γ-tub23 isoform early and the γ-tub37 isoform from stage 5 onwards. We propose that in germ cells, the Sxl-Cos2-Fu-γ-tubulin complex is bound to microtubules via γ-tubulin and Cos2 (Fig. 8); γ-tubulin bridges the complex to microtubules in a role that is analogous to its bridging of the centrosome to microtubules (Joshi et al., 1992). In keeping with such a hypothesis, we found that ovarian extracts pre-treated with colchicine continued to co-immunoprecipitate γ-tubulin with Sxl but not β-tubulin. This result presumably reflects the colchicine sensitivity of the interaction between the tubulins that does not appear to apply to the interaction of γ-tubulin to the complex.

At least three functions of the Sxl-Cos2-Fu-γ-tubulin complex are apparent. First, as suggested by the cos-2 negative germline clones and the fu ts allele, the complex stabilizes and anchors Sxl in the cytoplasm. Cytoplasmic Sxl might function to repress the translation of a specific message(s), consistent with the somatic role of Sxl in dosage compensation (Bashaw and Baker, 1997; Kelley et al., 1997). Second, the complex may facilitate and regulate the nuclear entry of Sxl, as Sxl lacks a canonical nuclear localization signal. The recent demonstration that both Cos2 and Fu shuttle into the nucleus (Méthot and Basler, 2000) would be consistent with such a role. We found that reducing Hh levels compromises the nuclear function of Sxl splicing autoregulation. Young egg chambers of hhts ovaries show most of the Sxl protein in the cytoplasm (Fig. 3A), whereas older chambers have a severe reduction of the protein (Fig. 3F). Under these conditions, the constitutive Sxl allele (SxlM4), which bypasses the requirement for Sxl in autoregulation, was able to restore Sxl expression (Fig. 3G). This suggests that it is the lack of nuclear Sxl that leads to its eventual loss in older hhts egg chambers. Finally, the effects of the proteasome inhibitors, even under conditions of Hh overexpression, suggest that the strongest effect of the Hh signal is the proteolysis of Sxl. This effect appears to depend on polymerized microtubules (Fig. 7) and may be balanced against nuclear entry. When Hh is globally overexpressed, we find Sxl expression is lost in all germ cells, other than a few cells immediately under the terminal filament (Fig. 4A). The presence of the SxlM4 allele was not able to overcome this effect, consistent with the idea that when Hh is overexpressed, most of the Sxl is being degraded. Although the data do not demonstrate that Sxl is a direct target of the proteasome, the effects of the Hh signal on Sxl are linked with this complex, as they can be overcome by proteasome inhibitors.

The proteolysis triggered by Hh decreases the overall amounts of Sxl in early germ cells (Fig. 7B,E), and may limit the amounts of Sxl that can enter the nucleus. Several observations suggest that the amounts of nuclear Sxl may be regulated in early germ cells. First, wild-type stem cells and early cystoblasts show only low levels of nuclear Sxl. Second, overexpression of Hh, which should lead to the disassembly of the cytoplasmic hh components (and the bound Sxl), shows an increase in detectable nuclear protein in the cells that still express high levels of cytoplasmic Sxl (Fig. 4A, inset). Third, using the proteasome inhibitors in both wild-type ovaries and ovaries where Hh had been overexpressed, Sxl could readily be detected in the nucleus (Fig. 7B (inset),E). Finally, removal of PKA, which normally functions negatively in response to the Hh signal, also results in increased nuclear Sxl. This increase is not as strong as the effects of the proteasome inhibitors, indicating that PKA is not the key effector of Hh effected Sxl proteolysis. PKA may regulate the rate of Sxl nuclear entry or exit; phosphorylation of the nuclear transport machinery has been shown to downregulate nuclear import in vitro (Kehlenbach and Gerace, 2000).

Keeping the level of nuclear Sxl low would seem incompatible with the requirement of Sxl to positively autoregulate the splicing of its transcripts. However, early germ cells may rely on other gene products to generate the Sxl female splice. Hager and Cline demonstrated that the female mode of Sxl splicing could be activated in male germ cells, but not somatic cells, by doubling the dose of the Sxl splicing co-factor snf (Hager and Cline, 1997). The need to keep Sxl nuclear levels low in early germ cells may have caused this reliance on products such as Snf to generate the Sxl female splice. This reliance also changes as germ cells mature. virilizer is not needed to productively splice Sxl RNA in early germ cells but is necessary late in the germarium and onwards (Schütt et al., 1998).

Previous studies on Hh signaling in the germline have focused primarily on the follicle cells (Forbes et al., 1996a; Forbes et al., 1996b; Zhang and Kalderon, 2000). We have focused on the effects of the hh pathway on Sxl in early germ cells, but our analyses also indicate that the generally universal response of Hh to upregulate Ptc expression does not apply to these germ cells. In the follicle cells, as seen in patterning, overexpression of Hh induces Ptc, whereas loss of Hh results in the loss of Ptc expression (Forbes et al., 1996a; Forbes et al., 1996b). In the germ cells of the germarium, the responses of Ptc to both over- and under-expression of Hh are the opposite of those in somatic cells and suggests that Ptc is under a different mode of regulation. This difference may arise from the fact that Ci, which normally upregulates Ptc in somatic tissues (Ingham, 1998), is not present in germ cells (Forbes et al., 1996b).

We also find that when Hh is over- or under-expressed, germ cells show an apparent shift towards premature differentiation. Conditions that reduce Hh activity show the branched form of the fusome in germ cells close to the terminal filament, which is much earlier than normal. Orb expression, which normally does not begin until the very end of region 1, appears to come on immediately after the stem cell stage. On the other hand, germaria and early egg chambers have persistent high levels of cytoplasmic Sxl (Fig. 3A) and Ptc (Fig. 3B), a phenotype more consistent with an early arrest, as these two proteins have not undergone the reduction that germ cells usually show as they mature.

When Hh is overexpressed, we observe greatly reduced numbers of early germ cells with high levels of cytoplasmic Sxl, Fu and cytoplasmic γ-tub23, as though the germ cells have differentiated early and have prematurely downregulated these proteins. The expression of Orb would support this conclusion, as Orb is detected in cells higher up in the germarium than normal (Fig. 4C). Overexpression of Hh also leads to loss of the Ptc and Spectrin signals (Fig. 4D-F) suggesting that the fusome has been compromised. In somatic cells, Hh leads to the internalization and eventual degradation of Ptc (Denef et al., 2000). If this were also true in germ cells, then overexpression of Hh would result in a general reduction of Ptc. Without Ci in germ cells to upregulate ptc transcription, little Ptc would remain. It is not clear how Hh overexpression affects Spectrin, but as Ptc has been described as accumulating in the core region of the fusome (Forbes et al., 1996b), compromising one component of the fusome could lead to a similar effect on other components.

These apparently contradictory differentiation responses to opposite levels of Hh can be reconciled if the phenotypes are considered similar but not identical. Hh is normally released from a localized source so a decrease in Hh presumably promotes germ cell differentiation as seen in follicle cells (Zhang and Kalderon, 2000). Overexpression of Hh, however, may also drive differentiation, but indirectly, by compromising the assembly or maintenance of the fusome. The fusome has been implicated in germ cell maturation (see Spradling et al., 1997) so its premature breakdown could lead to premature differentiation. Although the expression of Orb is earlier than normal when Hh is overexpressed, there is poor accumulation of the protein in the presumptive oocyte (compare Fig. 1D with 4C). This indicates a failure in the mechanism that transports determinants within the 16-cell cluster and may reflect a defective fusome, as this organelle has been postulated to be key to this intercellular transport process.

Unlike the follicle cells (Forbes et al., 1996b; Zhang and Kalderon, 2000), germ cells do not appear to require either smo or ptc. Using a strong smo allele under conditions that render it amorphic, we found that loss of smo had little effect on germline development, or on the expression of Sxl. It has also been reported that Ptc is not required in germ cells (Forbes et al., 1996b). In this case the allele used, ptcS2, encodes the protein but is null for Hh signal transduction (Chen and Struhl, 1996). This latter result raises the interesting possibility that the ptcS2 protein may still retain another function besides its capacity to bind Hh. As discussed above, overexpression of Hh severely reduces the levels of Ptc in germ cells and appears to compromise the fusome (Fig. 4D,E). If Ptc is functional in the fusome, then the lack of an effect with the S2 allele may indicate that this protein retains a structural activity necessary for the fusome that is separable from Hh signal transduction.

The existence of a Hh-dependent and Ptc-independent activity has been described in the phosphorylation of Fu in the posterior compartment cells of the wing imaginal disc (Ramirez-Weber et al., 2000). It is also possible that germ cells use homologs of Ptc and Smo to transduce the Hh signal. Mammals have at least two ptc genes (Ptch and Ptch2) and their expression patterns do not fully overlap: Ptch is expressed throughout the mouse embryo but Ptch2 is found at high levels in spermatocytes and skin. Ptch2 has been proposed to be the germline receptor for the mammalian Hh homolog, Desert Hedgehog, which is expressed specifically in the testis and is required for germ cell development (Carpenter et al., 1998). The Drosophila genome encodes at least two ptc homologs (FlyBase; Haines and van den Heuvel, 2000), so the effects of Hh on Sxl and the fusome could occur through either or both of these. A homolog of Smo has not yet been identified, so germ cells may use a molecule other than Smo to effect the presence of Hh. Alternatively, Smo in somatic cells may activate the release of a second signal from the somatic cells to the germ cells.

In summary, our studies suggest a direct involvement of the hh downstream components in the regulation of Sxl. These same components are involved in the regulation of Ci, but the outcome of Hh is frequently the opposite between Sxl and Ci. Loss of Cos2 severely reduces the levels of cytoplasmic Sxl, whereas for Ci, loss of Cos2 (using the same allele) increases the levels of the protein and localizes it to the cytoplasm (Sisson et al., 1997). The degradation of Sxl by Hh, is also opposite to Ci. Hh signaling inhibits the proteolysis of Ci. Sxl, however, is degraded in response to Hh. In addition, different from the normal Hh signaling process, germ cells either do not require Smo and Ptc, or they use a homolog of these proteins to transduce the Hh signal. Despite these differences, it is striking that in sharing the cytoplasmic components, both Ci and Sxl use the pathway for their intracellular distribution and involve the proteasome in this regulation.

We thank Drs R. Holmgren, K. Ho, M. Scott, D. Robbins, K. Nybakken, R. Steward, D. Bedwell, P. Lasko, P, Schedl, P. Ingham, P. Beachy, D. Kalderon and Isabel Guerrero for reagents; P. Schedl, V. A. Bankaitis and R. Johnson for helpful discussions and critical reading of the manuscript; and Albert Tousson, Shawn Williams and Pam Kontzen from the UAB imaging facility. The Spectrin antibody developed by D. Branton and R. Dubreuil was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa. This research was supported by grants to J. I. H. by the NIH and ACS.