CSN5/Jab1 mutations affect axis formation in theDrosophila oocyte by activating a meiotic checkpoint

Polarization of the anteroposterior (AP) axis of the Drosophilaoocyte occurs early in oogenesis, while the presumptive oocyte is still in the germarium (Gonzalez-Reyes and St Johnston,1998). The dorsoventral (DV) axis is set up much later and relies on transfer of the AP axis polarity from the oocyte to the somatic follicle cells at the posterior end of the oocyte. During stages 4-6 in wild-type egg chambers, grk RNA that is localized next to the nucleus at the posterior end of the oocyte is translated and signals through the EGFR pathway to establish the adjacent follicle cells as posterior(Gonzalez-Reyes et al., 1995). After these posterior follicle cells signal back to the oocyte, microtubule orientation in the oocyte is reversed, and the oocyte nucleus migrates along the microtubules to an anterior corner of the oocyte. During stages 8-9 this anterior corner is defined as dorsal by translation of grk RNA and activation of EGFR signaling in the overlying follicle cells(Gonzalez-Reyes et al., 1995;Neuman-Silberberg and Schupbach,1994; Roth et al.,1995). Thus, grk signaling is required for elaboration of the AP axis and establishment of the DV axis.

Recent results have shown that establishment of both AP and DV axes also depends on the successful repair of DNA double strand breaks (DSBs) that are formed during meiotic recombination(Ghabrial et al., 1998;Ghabrial and Schupbach, 1999). Meiotic prophase begins in early region 2a of the germarium, and both recombination and repair are probably completed before oocyte determination occurs in region 2b (Huynh and St Johnston, 2000).

Meiosis and axis establishment are related to each other because the accumulation of Grk protein in the oocyte cytoplasm depends on the successful completion of meiotic recombination(Ghabrial et al., 1998;Ghabrial and Schupbach, 1999). Mutations in the spindle-class genes, spindle-B(spn-B), spindle-C (spn-C) and okra(okr), cause a delay in oocyte determination and a failure to accumulate Grk protein, leading to defects in AP and DV patterning in late oogenesis (Gonzalez-Reyes et al.,1997; Ghabrial and Schupbach,1999). spn-B and okr encode Drosophilahomologs of the RAD51 and RAD54 genes from yeast that are required for DSB repair (Ghabrial et al.,1998; Kooistra et al.,1997). Their effects on Grk appear to be mediated by a DNA damage checkpoint governed by Mei-41, a Drosophila member of the ATM/ATR family of kinases that are required for DNA damage and recombination checkpoints in yeast, worms and humans, as well as flies (reviewed by Melo and Toczysky, 2002; Weinert, 1998;Murakami and Nurse, 2000). Because they eliminate the checkpoint, mei-41 mutations suppress the effects of spn or okr mutations. The spn andokr mutations can also be suppressed by mutations inmei-W68, which encodes the Drosophila homolog of yeast geneSPO11, a gene required for the induction of DSBs during recombination(Ghabrial et al., 1998;Roeder, 1997) These results indicate that the spn or okr patterning defects result from activation of a meiotic checkpoint in response to the presence of unrepaired DSBs.

We show that, like the spindle-class genes, CSN5 is required for the repair of recombination-induced DSBs duringDrosophila oogenesis. The CSN5 protein (also known as Jab1), is a subunit of the eight protein COP9 signalosome complex (CSN) originally identified in plants and conserved from plant to mammalian cells (for reviews,see Seeger et al., 2001;Schwechheimer and Deng, 2001;Bech-Otschir et al., 2002). As the genes for the CSN subunits were identified, a striking similarity was noticed between them and the eight subunits of the regulatory lid of the proteasome, suggesting a common ancestry and related function(Glickman et al., 1998;Seeger, 1998; Wei et al.,1998). This similarity was intriguing because examinations of CSN function have shown that it regulates protein stability in pathways leading to ubiquitination and degradation by the proteasome (reviewed bySeeger et al., 2001;Schwechheimer and Deng, 2001;Kim et al., 2001).

The CSN has been implicated in many regulatory and signaling functions including activation of the Jun transcription factor, stabilization of nuclear hormone receptors and interactions with integrins. Most relevant here, the CSN or its subunits have been shown to regulate multiple steps in the mitotic cell cycle. For example, the CSN regulates the ubiquitination and degradation of the CDK inhibitor, p27^kip1, and either a small, CSN5-containing subcomplex or CSN5 alone promotes p27^kip1 nuclear export(Yang et al., 2002;Tomoda et al., 1999). In addition, a CSN-associated kinase activity promotes degradation of p53,thereby allowing cell cycle progression(Bech-Otschir et al.,2001).

In Drosophila CSN5 is essential for development(Freilich et al., 1999) and was recently shown to be required in photoreceptor cells to induce glial cell migration (Suh et al., 2002). We report the first example of a CSN5 effect on meiosis and on axis determination. We find that homozygous CSN5-mutant clones disrupt both the DV and AP axes of the oocyte as a result of decreased Grk protein. These effects on axis determination appear to be caused by activation of the meiotic recombination checkpoint.

Canton S and w¹¹¹⁸ were used as standard strains.placW insertion line l(3)L4032 (referred to here asCSN5^L4032) was obtained from the BerkeleyDrosophila Genome Project(Spradling et al., 1999).grk^2B, grk^HK36, grk^HF,mei-41^D3, mei-41^D1 and mei-41^RTlines were a gift from T. Schupbach (described byNeuman-Silberberg and Schupbach,1993; Ghabrial and Schupbach,1999). kek^15A6 was obtained from N. Perrimon and described in Musacchio and Perrimon(Musacchio and Perrimon,1996). Enhancer trap line PZ6256(Liu and Montell, 1999) was a gift from D. Montell. Strains mei-W68¹,mei-W68^k05603, EGFR^f2, EGFR^E1,EGFR^t1, slbo⁰¹³¹⁰ (slbo¹) were obtained from the Bloomington Drosophila Stock Center. Standard conditions were used for raising flies. Crosses were performed at 25°C except as described in the text. Embryos were collected on molasses/agar plates. Flies carrying a GFP balancer were used to determine the lethal stage of development.

CSN5 homozygous-mutant germline clones were produced by using the dominant-female-sterile, FLP/FRT technique(Chou and Perrimon, 1992;Chou and Perrimon, 1996). Females of genotype w; CSN5^L4032 P{neoFRT}82B/TM3B, Sb orw; CSN5^35ex P{neoFRT}82B/TM3B, Sb were mated with males of genotype w hsFLP; ovo^D1 FRT82B/TM3B, Sb. Their progeny were heat shocked as third instar larvae or early pupae for two hours at 37°C for 2 consecutive days to induce FLP expression. Follicle cell mosaic clones were induced as described by Duffy et al.(Duffy et al., 1998): flies carrying w; P{en2.4-GAL4}e22c P{UAS-FLP1.D}JD1/CyO; P{neoFRT}82B P{Ubi-GFP(S65T)nls}3R were mated with w; P{neoFRT}82B CSN5^*/TM3B, Sb flies. Eggs were collected and examined for several days after eclosion. Females were dissected to confirm the presence of homozygous-mutant follicle cells marked by the absence of GFP.

The original CSN5 P element insertion l(3)L4032 was mobilized by introducing the P[ry⁺(Δ2-3)]99B transposase source (Engels et al., 1987). Derivatives that had lost the w⁺ marker carried by the original insert were crossed back to CSN5^L4032 to identify imprecise excisions. The majority of new excision lines appear to be precise excisions of the original P-element insertion. They were fully viable and had no ovarian defects when homozygous or when heterozygous withCSN5^L4032. Several weak alleles of CSN5 were also identified. They had poor viability and weak ovarian defects when heterozygous with CSN5^L4032. Finally, several lines failed to complement the lethality of CSN5^L4032, and produced, as germline clones, similar ovarian defects as didCSN5^L4032.

The fixation and visualization of egg chambers and embryos was performed as described (Cant et al., 1994;Verheyen and Cooley, 1994). For immunostaining, the following antibodies were used: mouse anti-Grk (1:20),rat anti-Grk (1:500), rabbit anti-sperm-tail (1:500), rabbit anti-Vasa(1:1000) (gifts from T. Schupbach, R. Cohen, T. Karr and P. Lasko). To monitorlacZ expression of the P-lacZ insertion mutations, ovaries were treated according to Verheyen and Cooley(Verheyen and Cooley, 1994). For actin visualization, ovaries were stained with rhodamine-conjugated phalloidin (Molecular Probes). To visualize nuclei, tissues were stained with DAPI. High magnification fluorescent images were collected on a Zeiss 510 confocal microscope.

In situ hybridization using digoxigenin-labeled antisense RNA probes was carried out as described (Tautz and Pfeifle, 1989) with modifications(Harland, 1991). Hybridization signals were visualized by histochemical staining with alkaline phosphatase. Embryos and ovaries were mounted in 70% glycerol and viewed and photographed with Nomarski optics on a Leica DMRB microscope.

Protein extracts for western blot analysis were prepared as described by Sambrook et al. (Sambrook et al.,1989). Drosophila CSN5/JAB1 protein was detected using a mouse polyclonal, and three independent mouse monoclonal, anti-mouse Jab1 antibodies (GeneTex), or a rabbit polyclonal, anti-mouse Jab1 antibody (Santa Cruz Biotechnology). On a western blot, all of these antibodies recognized the same 37-38 kDa band, consistent with the predicted size of DrosophilaCSN5. No other specific bands were detected. This band is strongly reduced in extracts from CSN5^L4032 germline clone ovaries and is reduced to different extents by the hypomorphic alleles derived fromCSN5^L4032. Monoclonal antibody MS-JAB11-PXS (GeneTex) was used for the western blots in this paper. We used rat polyclonal and mouse monoclonal anti-Grk antibodies (gifts from T. Schupbach and R. Cohen), rabbit anti-Vasa (a gift from P. Lasko), or monoclonal anti-Actin (ICN). Secondary antibodies for signal detection were a goat anti-rat or anti-mouse and a protein-A horseradish peroxidase conjugate (Molecular Probes; Santa Cruz Biotechnology). Proteins were visualized using chemiluminescent detection (NEN Life Science Products).

Total or polyA⁺ RNA was isolated from ovaries as described(Sambrook et al., 1989). RNA was resolved on formaldehyde-agarose gels, transferred to nylon membranes,crosslinked and hybridized by standard procedures.

Since most CSN5 homozygotes die during larval or pupal development(this paper) (Freilich et al.,1999), it seemed likely that embryos receive a maternal contribution of CSN5 RNA or protein. In situ hybridization confirmed this expectation, showing that CSN5 RNA accumulates in the nurse cells beginning in the germarium and continuing through most of oogenesis(Fig. 1A-D). During stage 10,CSN5 RNA is transferred to the oocyte along with the bulk of the nurse cell cytoplasm.

In embryos, uniformly distributed maternal RNA is evident until gastrulation begins. The earliest zygotic expression is in an anterior stripe during cellular blastoderm. During gastrulation, zygotic expression becomes evident in the ventral furrow, the cephalic furrow, and both the anterior and posterior midgut invaginations (Fig. 1E-H).

To enable an analysis of early embryonic requirements for CSN5, we induced homozygous, CSN5-mutant germline clones(Chou and Perrimon, 1992). These clones revealed requirements for CSN5 during oogenesis as well as embryogenesis. In ovarian germline clones the level of CSN5 RNA is dramatically reduced, but still detectable, indicating that the P-element-induced allele, CSN5^L4032, is hypomorphic(Fig. 1C,D). Depending on the paternal allele, embryos derived from the germline clones showed either a reduced amount of CSN5 RNA in the zygotic pattern(Fig. 1F) or no detectableCSN5 RNA (not shown).

Flies carrying CSN5 germline clones laid eggs with a range of abnormal phenotypes that were affected by temperature(Fig. 2,Table 1). Flies grown at 25°C laid eggs with phenotypes closest to normal. The most frequent defects at 18°C were different from those at 29°C. At 18°C many of the defective eggs had fused dorsal appendages(Fig. 2B). At 29°C there was an increasing frequency of properly separated but short dorsal appendages(Fig. 2E). These results suggest that aberrations in patterning the follicular epithelium predominate at 18°C, while defects in follicle cell migration predominate at 29°C.

Because the eggshell phenotypes were only partially penetrant, it was possible that they were caused by somatic, rather than germline, CSN5clones. To test this possibility, we induced somatic clones in the ovary by using the follicle cell driver E22c-GAL4 to induce expression ofUAS-FLP (Duffy et al.,1998). Under these conditions, there were no eggshell defects at any temperature, indicating that this requirement for CSN5 function is limited to the germline.

In addition to the eggshell defects, the viability of CSN5 mutants also depends on temperature. At 29°C the original P-element mutation is lethal during early development with fewer than 1% of the mutant larvae becoming prepupae. By contrast, at 18°C 90% of the mutant larvae pupariate and 1-2% escape as adults. Mobilization of the original P-element insertion confirmed that it was responsible, not only for lethality, but also for the eggshell defects; precise excisions were viable and had normal dorsal appendages.

Some mutations that disrupt the DV patterning of the eggshell also affect the patterning of the embryo. To look for effects on the embryonic DV fate map, we used as markers the expression of three zygotic genes:decapentaplegic (dpp), rhomboid (rho) andtwist (twi) (Fig. 3). dpp is expressed on the dorsal side of the embryo as well as its anterior and posterior ends(St Johnston and Gelbart,1987). rho is expressed in two, eight-cell-wide ventrolateral domains and later also in a narrow stripe on the dorsal side of the embryo (Bier et al., 1990).twi, a marker for the mesoderm, is expressed ventrally in the embryo(Thisse et al., 1988).

For all three of the markers, many of the CSN5-mutant embryos appeared to be ventralized (Fig. 3B,E,H). In these embryos dpp expression on the dorsal side was reduced or absent. The dorsal rho stripe was reduced and the lateral stripes were moved dorsally. twi expression appeared to expand dorsally about halfway around the embryo. Some embryos showed stronger ventralization at their anterior or posterior ends (data not shown). There were also infrequent embryos that appeared to be dorsalized(Fig. 3C,F,I).

To characterize CSN5 mutants further, we examined the spatial localization of the RNAs for two determinants of AP polarity, bicoid(bcd) and oskar (osk). The localization ofbcd RNA to the anterior pole of the oocyte is crucial in the establishment of AP polarity(Nusslein-Volhard et al.,1987; Berleth et al.,1988; St Johnston et al.,1989). In CSN5 mutant oocytes and embryos, bcdmRNA was abnormally expressed in 10-15% of oocytes(Fig. 4). In these abnormal oocytes, the bcd mRNA is diffusely distributed and sometimes accumulated near the center of the oocyte(Fig. 4B). In mutant embryos,the bcd RNA often shifted toward the dorsal side of the embryo(Fig. 4F).

The posterior pole of the egg chamber is defined by the tight, posterior localization of osk RNA (Ephrussi et al., 1991; Kim-Ha et al.,1991). Although most CSN5-mutant oocytes and embryos were nearly normal, osk RNA in 10-15% of mutant oocytes and embryos was reduced or mislocalized (Fig. 4). In the abnormal oocytes, the osk RNA was typically diffuse or concentrated in the center of the oocyte(Fig. 4D). Only small amounts were localized at the posterior pole. In the abnormal embryos only a small amount of osk RNA at the posterior pole remained. In these embryos the osk RNA appeared to be shifted slightly dorsally from its normal position at the extreme posterior end (Fig. 4H).

Since the localization of osk and bcd RNAs depends on polarization of the microtubule lattice, we used a reporter for the motor protein kinesin to examine microtubule organization in CSN5 germline clones (Clark et al., 1994). Kinesin moves toward the plus ends of microtubules, and in stage 8-9 wild-type egg chambers kinesin-β-gal localizes to the posterior of the oocyte. However, in some CSN5-mutant oocytes kinesin-β-gal staining was diffuse or mislocalized (not shown).

In addition to its role in determining the AP axis, CSN5 may have a distinct role in pole cell development. In normal embryos, the pole cells form as a tight, contiguous cluster at the posterior end of the embryo(Fig. 4I). As gastrulation and germ band extension begin, somatic epithelial cells at the posterior end of the embryo form a shallow cup that will eventually become the posterior midgut invagination. The pole cells adhere to this cup and remain tightly clustered on its surface as they are conveyed over the dorsal side of the embryo and then into its interior. In CSN5-mutant embryos the number of pole cells is often reduced, as might be expected because of the inefficient localization of oskar RNA (Fig. 4J). In addition, the pole cells are occasionally found in a loose, non-contiguous group near, but not tightly associated with, the posterior end of the embryo (Fig. 4K). This is an unusual phenotype, not seen in other mutants that impair the formation of pole plasm. Thus, in addition to its role inoskar RNA localization, CSN5 may have a separate role in organizing the pole cell cluster.

Since CSN5 germline clones caused defects in both the AP and DV axes, it seemed possible that grk signaling was compromised(Gonzalez-Reyes et al., 1995;Roth et al., 1995). As described in the introduction, grk is unusual among axis-determining genes in being required for both axes.

To assess the role of CSN5 in grk signaling, we used reporters for either the posterior or the dorsal Grk signal. In the absence of the posterior Grk signal, the posterior follicle cells appear to adopt the anterior follicle cell fate and express markers that are characteristic of the border cells (Gonzalez-Reyes et al.,1995; Roth et al.,1995). We used two such markers, an enhancer trap called PZ6356(Fig. 5A) and aslbo-lacZ enhancer trap (Fig. 5C) (Montell et al.,1992; Tinker et al.,1998; Liu and Montell,1999), to monitor whether CSN5 is required for the early Grk signal. For both markers, loss of CSN5 from the germline causedlacZ expression in the posterior follicle cells of many egg chambers,suggesting a reduction in Grk signaling(Fig. 5B,D). To monitorEGFR signaling to the dorsal follicle cells at stages 9 and 10, we used a kekkon (kek)-lacZ reporter construct(Fig. 5E). Because thekek gene acts downstream of the EGFR pathway in the follicle cells, it can serve as a sensitive indicator of grk activity coming from the oocyte (Musacchio and Perrimon,1996; Sapir et al.,1998). We found that at 18°C kek expression is abnormal in about a third of CSN5-mutant egg chambers at stage 10(but only 3-4% at 25°C). In most of these egg chambers, expression in the dorsal anterior follicle cells over the oocyte was reduced or, rarely, absent(Fig. 5F,G). A small number of egg chambers show broader expression of kek in the follicle cells,probably reflecting the small number of dorsalized embryos arising from these mutant egg chambers (Fig. 5H). We conclude that in most egg chambers both posterior and dorsal Grk signaling are impaired in CSN5-mutant germline clones.

Further evidence that CSN5 affects Grk signaling came from testing for genetic interactions between CSN5 and either grk orEGFR. Females heterozygous for strong grk alleles lay eggs with fused or partially fused dorsal appendages(Table 2). This dominant phenotype provides a sensitive background for detecting interactions. With the exception of a precise P-element excision, all CSN5 alleles showed strong enhancement of the dominant grk phenotype(Table 2). In addition,CSN5^L4032 weakly enhanced the dominant eggshell phenotype of a loss of function EGFR allele, EGFR^f2.

These results suggested that production of grk RNA or protein might be affected in CSN5 germline clones. In situ hybridization using a grk probe showed normal or nearly normal localization ofgrk RNA in most CSN5-mutant stage 10 oocytes(Fig. 6B). In some of these mutant oocytes the messenger was improperly localized, probably because the oocyte nucleus was no longer located at the dorsal corner of the oocyte(Fig. 6C). Interestingly, in these oocytes the `dorsal' follicle cells were often columnar as though the nucleus had been properly localized at an earlier stage(Fig. 6C). A northern blot showed nearly normal amounts of grk mRNA in ovaries carryingCSN5-mutant germline clones, consistent with the strong signals seen by in situ hybridization in most oocytes(Fig. 6G).

Immunostaining of egg chambers using anti-Grk antibodies showed a more extreme effect. Grk protein was strongly reduced in CSN5 mutants compared with controls, although the residual protein usually appeared to be properly localized (Fig. 6D-F). This reduction was confirmed by western blot analysis(Fig. 6H). There were also a few cases of Grk protein mislocalization, sometimes being present all along the anterior end of the oocyte (data not shown).

The reduction in Grk protein appeared to be most extreme at early stages in oogenesis. Grk expression begins in region 2a in wild-type germaria. The signal appears in several cells per cyst in regions 2a and 2b and then becomes concentrated in the oocyte cytoplasm by region 3(Fig. 7A). In viable,hypomorphic combinations such asCSN5^ex21/CSN5^L4032, we could not detect Grk expression in the germarium (Fig. 7B). With this combination Grk does become detectable from stage 2-3 onwards (data not shown), suggesting that a reduction in CSN5 causes a delay in the beginning of Grk accumulation (see Discussion). Taken together these results show that the major effect of CSN5 mutations appears to be on grk RNA translation or on stability of the protein.

Because of the similarity between the CSN5 andspindle-class phenotypes, we tested for a connection betweenCSN5 and the meiotic checkpoint mediated by mei-41. As mentioned above, the viable hypomorphic combinationCSN5^ex21/CSN5^L4032 caused a reduction in Grk protein level, especially during the early stages of oogenesis(Fig. 7B). Five to fifteen percent of eggs laid by these transheterozygotes had fused dorsal appendages,indicating a partial reduction of Grk. WhenCSN5^ex21/CSN5^L4032 flies were also homozygous-mutant for mei-41, however, the normal Grk protein level was restored (Fig. 7C), and the eggshell phenotype was rescued (not shown).

Interestingly, checkpoint activation leads to modification of the Vasa protein, as shown by a slightly reduced mobility during SDS polyacrylamide gel electrophoresis (Ghabrial and Schupbach,1999). This result is relevant to the spindle-class andCSN5 phenotypes because Vasa regulates translation of Gurken and, as a consequence, axial patterning (Styhler et al., 1998; Tomancak et al.,1998). This Vasa modification is checkpoint dependent since it is present in spn-B mutants but absent in mei-41 spn-B double mutants (Ghabrial and Schupbach,1999).

We detected a similar reduced mobility of Vasa protein in CSN5mutants (Fig. 7D). For viableCSN5 mutants there were two Vasa bands: one corresponding to Vasa from wild-type ovaries and a second with lower mobility. In stronger mutant combinations, most of the Vasa protein was modified, while in weaker combinations most Vasa had normal mobility. The shift in Vasa mobility was suppressed by mei-41 mutations. Interestingly, removal of one dose ofmei-41 completely restored normal Vasa mobility for a weakCSN5 combination. For stronger CSN5 mutants, full restoration of Vasa mobility required removal of both mei-41 genes(Fig. 7E).

The gene mei-W68 is required for the initiation of meiotic recombination in Drosophila ovaries and is likely to induce DNA double strand breaks (DSBs) as recombination begins(McKim and Hayashi-Hagihara,1998). Mutations in mei-W68 were shown to rescuespindle-class defects, including Grk protein accumulation, eggshell morphology and Vasa modification (Ghabrial and Schupbach, 1999). These results suggested that since DSBs were not formed in the absence of mei-W68, DNA repair by thespindle-class genes was not required. A similar interaction was seen between mei-W68 and CSN5. Hetrerozygosity formei-W68 was sufficient to suppress the phenotypes of both strong and weak CSN5 allelic combinations(Fig. 7E).

These data demonstrate that absence of CSN5 function during meiosis activates a DNA-damage checkpoint that is mediated by Mei-41. Because the reduction in DSBs in mei-W68 heterozygotes removes the requirement for CSN5, it is likely that CSN5 promotes DNA repair, as do thespindle-class genes.

Establishment of both AP and DV polarity requires expression of the TGF-α homolog Gurken in the oocyte and activation of the EGF receptor and its downstream effectors in the adjacent follicle cells. Our results show that CSN5 is required in the germline for these critical signaling events. Several results tie CSN5 to Grk-EGFR signaling. First, CSN5mutations affected both axes as shown by DV defects in the eggshell,mislocalization of bcd and osk RNAs in both the oocyte and embryo, and mislocalization of dpp, rho and twi expression in the embryo. Second, CSN5 germline clones affected the expression of follicle cell reporters for Grk-EGFR signaling: slbo and the PZ6356 enhancer trap in the posterior follicle cells, kek expression in the dorsal anterior follicle cells. Third, CSN5 alleles show strong genetic interactions with grk alleles. Finally, Grk protein is reduced in CSN5 germline clones, starting in region 2a of the germarium but still evident in stage 10 egg chambers or in ovary extracts.

Previous studies have shown that the accumulation of Grk protein can be affected by activation of a meiotic checkpoint in response to the persistence of DNA double-strand breaks (Ghabrial and Schupbach, 1999). Mutations in several genes that play a role in DNA repair (okra, spn-B, spn-C and spn-D) activate this meiotic checkpoint and disrupt axial patterning in the oocyte. There is a remarkable similarity between the CSN5-mutant phenotype and defects caused by mutations in these spindle-class genes (described byGonzalez-Reyes et al., 1997;Ghabrial et al., 1998;Ghabrial and Schupbach, 1999). In both cases mutant females produced eggs with a variety of partially penetrant eggshell defects: mild or strongly ventralized, dorsalized, or small eggs or eggs with multiple dorsal appendages. Embryonic patterning was also disrupted, and both axes were affected. As had been seen inspindle-class mutants, the oocyte of some CSN5-mutant egg chambers was positioned laterally or at the anterior end, and some had defects in karyosome morphology (data not shown). There was also a similar, strong reduction in Grk protein, with one intriguing difference. At early stages of oogenesis in CSN5 mutants, the level of Grk protein was always strongly reduced, both in germline clones of the strongCSN5^L4032 allele and in hypomorphic combinations ofCSN5^L4032 with viable excision mutants(Fig. 7). Although Grk was also strongly reduced in CSN5^L4032 germline clones at later stages (Fig. 6), it often appeared to be present at higher levels than in the germarium. With the hypomorphic combinations, it was often difficult to detect any reduction in Grk protein at later stages. By contrast, in spn-B and spn-Dmutants, Grk accumulates normally in early oogenesis but then declines and is often undetectable by stage 9-10 (Ghabrial et al., 1998). In okr mutants, the amount of Grk protein varies from one egg chamber to the next in a single ovariole, but a bias towards lower levels at early stages was not reported(Ghabrial et al., 1998). Thus,there seem to be three different patterns of Grk accumulation in these mutants. CSN5 mutants appear to cause a more immediate response of Grk to DNA damage than do spn-B and spn-D mutants.

Because of the similarities between the phenotypes and because at least two of the spindle-class genes, okr and spn-B, encode components of the RAD52 DNA repair pathway, it seems likely that CSN5directly or indirectly regulates DSB repair. The fact that mei-41 andmei-W68 mutations can suppress the CSN5 phenotypes reinforces this conclusion. Kinases in the ATM/ATR subfamily that includes Mei-41 play a central role in checkpoint-mediated responses to DNA damage (for reviews, see Melo and Toczysky, 2002;Weinert, 1998;Murakami and Nurse, 2000). These checkpoint kinases are thought to act as sensors of DNA damage, becoming activated on binding damaged DNA. Phosphorylation of several downstream effectors, including the Chk1 and Chk2 kinases and p53, then restrains cell cycle progression until the DNA damage is repaired and the checkpoint kinases dissociate from the DNA. In Drosophila mei-41 mutants, the checkpoint cannot be activated, and oocytes with damaged DNA, such as those mutant forspindle-class genes, can proceed through oogenesis. Suppression ofCSN5 phenotypes by mei-41 mutations demonstrates that theCSN5-mutant lesion acts upstream of the DNA damage checkpoint and suggests that DSBs arising during meiotic recombination cannot be efficiently repaired in CSN5-mutant cells(Fig. 7).

Suppression by mei-W68 restricts the possible role of CSN5 further. mei-W68 encodes a topoisomerase II-like protein homologous to S. cerevisiae Spo11 and has been proposed to create the DSBs needed to initiate meiotic recombination(McKim and Hayashi-Hagihara,1998). In flies mutant for mei-W68, DSBs are absent and meiotic recombination is eliminated. In double mutants of mei-W68with either okr, spn-B or spn-C, Grk protein accumulation and eggshell patterning are normal and other spindle-class defects are suppressed (Ghabrial and Schupbach,1999). We found that heterozygosity for mei-W68 was sufficient to suppress hypomorphic CSN5-mutant phenotypes(Fig. 7). Combination of this result with the mei-41 suppression result indicates that CSN5 acts in the recombination pathway to regulate the formation of DSBs or their successful repair (Fig. 8).

vasa mutants show similar effects on axis determination and Grk protein accumulation as do spindle mutants and CSN5 GLCs(Styhler et al., 1998). However, the vasa phenotypes are not suppressed by mei-41 ormei-W68 mutations, indicating that Vasa acts downstream of the meiotic checkpoint (Ghabrial and Schupbach, 1999). Indeed, Vasa is one of the targets of Mei-41 activity as Vasa electrophoretic mobility is changed in spn-B mutants but restored in mei-41 spn-B double mutants(Ghabrial and Schupbach,1999). Since Vasa protein binds to grk mRNA and is required for both its localization in the oocyte and its translation, it seems likely that the checkpoint effects on Grk accumulation are directly mediated by Vasa, although other Mei-41 targets cannot be excluded(Fig. 8). Our results that show effects of CSN5 mutants on Vasa mobility are entirely consistent with the previous spn-B results, as would be expected if both types of mutants activate the same checkpoint.

We propose that in CSN5-mutant oocytes DSBs created by Mei-W68 during meiotic recombination are repaired more slowly than in wild type. Accumulation of unrepaired DNA breaks would then activate themei-41-dependent checkpoint leading to a block in the progression of meiotic prophase (Fig. 8). Since activated Mei-41 is an ATR-related kinase, it might modify Vasa directly or through downstream kinases such as Chk1 or Chk2. Modified Vasa would then prevent efficient Grk translation. Because CSN5 mutants are likely to affect the stability rather than the presence or absence of repair proteins,the DNA DSBs might be slowly repaired during the checkpoint-induced delay,thereby allowing cell cycle progression to resume. Delayed repair might explain why the early CSN5 effects on Grk expression are stronger than at later times. It might also explain why CSN5-mutant phenotypes are weaker and less penetrant than in okra and spn-Bmutants, in which repair proteins are absent and DNA probably remains unrepaired.

How might CSN5 regulate DNA repair? Two mechanisms of CSN activity have been reported, and either might affect the activity or stability of proteins involved in DNA repair. In addition, since there is an excess of CSN5 relative to other CSN subunits in many cells (Yang et al., 2002), CSN5 might regulate DNA repair independent of the large CSN complex.

The best-documented mechanism for CSN activity works through regulation of the SCF (Skp1/cullin-1/F-box) ubiquitin ligases(Lyapina et al., 2001;Yang et al., 2002). This pathway is attractive here because SCF-dependent ubiquitination mediates the degradation of many cell-cycle regulators, including not only p27^kip1, but also cyclins E, A and B, CDK inhibitor p21, E2F1,β-catenin and IκBα(Michel and Xiong, 1998;Russell et al., 1999;Yu et al., 1998;Carrano et al., 1999;Marti et al., 1999;Kitagawa et al., 1999;Hatakeyama et al., 1999). Recently, a connection has been made in C. elegans between the SCF complex and the regulation of meiosis. Members of the Skp1-related(skr) gene family in C. elegans are required for the restraint of cell proliferation, progression through the pachytene stage of meiosis, and formation of bivalent chromosomes at diakinesis(Nayak et al., 2002).

The CSN regulates SCF activity by removing the ubiquitinlike protein Nedd8 from the cullin subunit of SCF (Lyapina et al., 2001). Nedd8/Rub1 is covalently attached to target proteins through an enzymatic cascade analogous to ubiquitination(Lammer et al., 1998;Liakopoulos et al., 1998;Osaka et al., 2000). It is ligated to all cullin family proteins, and so far cullins are the only known targets for neddylation (Hori et al.,1999; Read et al.,2000). Nedd8 modification enhances the ubiquitinating activity of the SCF complex in vitro and is required in vivo for embryogenesis in both mice and nematodes (Kawakami et al.,2001; Tateishi et al.,2001; Jones et al.,2002). As the CSN mediates cleavage of the Nedd8 conjugate, it can antagonize SCF-dependent protein degradation. For example CSN inhibits ubiquitination and degradation of p27^kip1 in vitro and injection of the purified complex inhibited the G1-S transition in cultured cells(Yang et al., 2002).

Although this deneddylation activity of the CSN would explain our results,the kinase activity associated with the CSN might also be important. This kinase activity co-purifies with the CSN complex though it is uncertain whether it is intrinsic to one of the CSN subunits(Bech-Otschir et al., 2002). It phosphorylates and stabilizes the Jun transcription factor against proteasomal degradation (Musti et al.,1997). Conversely, it sensitizes p53 degradation by the SCF-ubiquitin pathway (Bech-Otschir et al.,2001).

Although the DNA repair-related targets of CSN5 or the CSN remain unclear, proteins encoded by the spindle-class genes or bymei-W68 are strong candidates (seeFig. 8). The deneddylation activity of the CSN might protect a DNA repair protein from SCF-dependent degradation. Alternatively, the kinase activity might promote Mei-W68 turnover, thereby limiting the production of DSBs. Further investigation may help to distinguish among these and other hypotheses and find direct targets for CSN5 in oogenesis.

We are grateful to Erica Roulier, who initiated the study of CSN5in oogenesis. We are indebted to Robert Cohen, Paul Lasko, Tim Karr, Denise Montell, Norbert Perrimon, Hannele Ruohola-Baker and especially Trudi Schupbach for sharing reagents and fly stocks, and for advice. We are also grateful to the Berkeley Drosophila Genome Project and Bloomington Drosophila Stock Center. In addition, we thank Mark Stern and Vidya Chandrasekaran for critically reading the manuscript. Finally, we thank an anonymous reviewer for suggestions that significantly clarified the manuscript.