The RNA-binding protein Squid is required for the establishment of anteroposterior polarity in the Drosophila oocyte

Many cells have an obvious polarity that is essential to their function. One important mechanism for establishing polarity involves the asymmetric localization of mRNAs, which is then translated into asymmetric protein localization and function. In many cases, RNAs are localized by directed transport along microtubules (MTs)(Lipshitz and Smibert, 2000). This is true of the Drosophila oocyte, in which the MT-dependent localization of three primary mRNA determinants polarizes the cell and establishes the body plan of the future embryo. Localization of bicoid (bcd) and oskar (osk) mRNAs to the anterior and posterior termini of the oocyte, respectively, establishes the anteroposterior (AP) axis, and localization of gurken (grk)mRNA to the presumptive dorsal anterior corner determines the dorsoventral(DV) axis (van Eeden and St Johnston,1999). The large size of the Drosophila oocyte, along with its relatively slow development, has allowed substantial dissection of the processes involved in the establishment of polarity using molecular genetics, thus affording insights into fundamental mechanisms that organize MT polarity and localize RNAs (Johnstone and Lasko, 2001; Riechmann and Ephrussi, 2001).

The Drosophila oocyte develops within an egg chamber, which consists of 15 nurse cells and an oocyte surrounded by an epithelial monolayer of somatic follicle cells (Spradling,1993). The oocyte and nurse cells are produced from a single germline cystoblast, which arises when a germline stem cell divides. The cystoblast undergoes four synchronous mitotic divisions to produce a germline cyst of 16 cystocytes; each division ends with incomplete cytokinesis, leaving the cystocytes interconnected by cytoplasmic bridges called ring canals. As the cyst forms, MTs are organized by a specialized cytoskeletal structure called the fusome, which spans the entire cyst. During the cystocyte divisions, the fusome attaches to one pole of every mitotic spindle, ensuring the stereotypical arrangement of cystocytes(Grieder et al., 2000). Subsequently, the fusome directs the formation of an MT organizing center(MTOC) within the oocyte, which enables cytoplasmic determinants to accumulate within the ooplasm and maintain oocyte fate(Huynh and St Johnston, 2004). The cyst then rearranges such that the oocyte is positioned posteriorly, and somatic follicle cells encapsulate the cyst. At this stage, defined as stage 1, the MTOC is positioned at the anterior of the oocyte, and the MT plus-ends extend through the ring canals into the nurse cells(Theurkauf et al., 1992). At stage 2, the egg chamber buds off from the germarium, where it is formed, and enters the vitellarium, where it will develop into a mature egg. The MTOC and the oocyte nucleus move to the posterior of the oocyte(Theurkauf et al., 1992), and the polarized MT network directs the localization of several cytoplasmic factors that are synthesized in the nurse cells to the oocyte(Koch and Spitzer, 1983; Pokrywka and Stephenson,1995). In this manner grk RNA becomes concentrated within the oocyte and accumulates at its posterior cortex, where it is translated during stages 2-6 (Neuman-Silberberg and Schupbach, 1993; Van Buskirk and Schupbach, 1999). The posteriorly localized Grk protein, a TGF-α homolog, signals via the Epidermal growth factor receptor (EGFR)to the follicle cells overlying the oocyte at the posterior of the egg chamber, causing posterior cells to express different markers and behave differently from the other follicle cells(Gonzalez-Reyes et al., 1995; Gonzalez-Reyes and St Johnston,1998; Roth et al.,1995). At stage 7-8, there is a rearrangement of the germline MT cytoskeleton such that the MT network emanating from the posterior MTOC is replaced by MTs nucleated from multiple sites at the anterior cortex of the oocyte and extending toward the posterior(Theurkauf et al., 1992). The newly polarized MT cytoskeleton allows bcd RNA to be directed toward the anterior of the oocyte in association with a minus-end directed motor(Arn et al., 2003; Schnorrer et al., 2000), and osk RNA to be transported to the posterior by the plus-end directed motor, kinesin (Brendza et al.,2000). MT rearrangement also allows the oocyte nucleus to move from its early position near the posterior to the presumptive dorsal anterior corner (Koch and Spitzer,1983; Swan et al.,1999). During stages 8-11, grk RNA accumulates in a cap between the nucleus and the plasma membrane, and the resulting Grk protein signals via the EGFR to specify dorsal cell fate in the overlying follicular epithelium (Neuman-Silberberg and Schupbach, 1993; Neuman-Silberberg and Schupbach,1996).

Mutations that alter Grk signaling during stages 8-11 disrupt DV patterning of the follicle cells, leading to altered DV polarity in the eggshell and embryo. For instance, loss of Squid (Sqd) activity in the germline leads to the delocalization of grk RNA and Grk protein along the entire anterior circumference of the oocyte, resulting in loss of ventral follicle cell fates (Neuman-Silberberg and Schupbach, 1993; Neuman-Silberberg and Schupbach,1996). Mutations that prevent Grk signaling to the posterior follicle cells during stages 2-6 disrupt AP patterning of the oocyte. For example, mutations in grk itself lead to an aberrant MT rearrangement at stages 7-8, altered localization of bcd and osk RNAs, and failure of the oocyte nucleus to migrate to the anterior(Gonzalez-Reyes et al., 1995; Roth et al., 1995). This,along with the discovery of similar AP phenotypes when specific gene activities (e.g. Notch and Laminin A) are lost in the follicle cells alone(Deng and Ruohola-Baker, 2000; Ruohola et al., 1991), led to the proposal that correctly specified posterior follicle cells must signal to the oocyte at stages 6-7 to allow normal MT rearrangement. However, this hypothetical signal has not been defined.

Protein kinase A (PKA; PKA-C1 - FlyBase) was identified as a potential transducer of the posterior follicle cell signal because it is required in the germline for normal MT polarity and RNA localization in stage 8-9 oocytes(Lane and Kalderon, 1994). In this study, we characterized the role of PKA further and identified a mutation in sqd as a weak modifier of the PKA mutant phenotype. Sqd is a member of the heterogeneous nuclear ribonucleoprotein (hnRNP) A/B family of RNA-binding proteins, and its role in establishing DV polarity has been well characterized (Kelley,1993; Matunis et al.,1994; Neuman-Silberberg and Schupbach, 1993; Neuman-Silberberg and Schupbach,1996). We found that loss of Sqd function produces a highly penetrant defect in MT organization along the AP axis at stages 8-9, as well as aberrant localization of osk RNA and other molecules that are normally localized posteriorly. A less penetrant defect in MT polarity, which affects Grk localization, is also evident before stage 6 in sqdmutants. Despite this earlier defect, sqd mutants display normal specification of posterior follicle cells in response to the early Grk signal. Thus, Sqd is essential for the establishment of MT polarity in both early and mid-oogenesis.

Mutant clones were generated by mitotic recombination using the FLP/FRT system and an X-chromosomal hs-flp(Xu and Rubin, 1993). Germline clones were made using the DFS technique(Chou et al., 1993; Chou and Perrimon, 1996) and the following alleles: PKA-C1^H2(Lane and Kalderon, 1993),P-element lethal mutations from BDGP(Spradling et al., 1999), sqd^ix50 and sqd^ix77 (gifts from T. Schupbach). Other alleles used for genetic analyses were: PKA-C1^A13 and Df(2L)Tw2(Lane and Kalderon, 1993), sqd¹ and Df(3R)urd (gifts from A. Norvell), hrp48^k16203 (gift from A. Ephrussi), hrp48^K02814, hrp48⁰²⁶⁴⁷, Df(3R)126C,mago¹ and Df(2R)F36 (from Bloomington Stock Center). Somatic clones were marked using FRT82B Ubi-GFP (from Bloomington) or FRT82B hsCD2, y⁺ (gift from G. Struhl). Rescue constructs used were: mC^* (Li et al.,1995), GFP-LKB1^S535E(Martin and St Johnston, 2003)and SqdS (Norvell et al.,1999). Marker lines used were: mirror^6D1(Ruohola-Baker et al., 1993), Orb-Pz (Lantz and Schedl,1994), pnt^998/12, BB127, L53b and 5A7(Gonzalez-Reyes and St Johnston,1998), kinesin-lacZ, pnt^rm254(Lee and Montell, 1997), GFP-Stau (gift from D. St Johnston) and m1⁴¹⁴lacwt (Gunkel et al., 1998).

Ovaries were hand-dissected from ten females fattened on yeast paste. Ovaries were teased open with forceps and fixed for 9 minutes in a biphasic 1:6 mixture of 6% formaldehyde:heptane. Ovaries were washed in PBS + 0.1%Tween-20 (PBT), permeabilized for 1 hour in PBT + 1% Triton-X 100, dehydrated into methanol, and stored overnight at -20°C in 100% methanol. After rehydration, ovaries were treated for 10 minutes with 50 μg/ml proteinase K and for 2 minutes with 2 mg/ml glycine. Ovaries were post-fixed for 15 minutes in 4% formaldehyde and washed 5×5 minutes in PBT. Subsequent steps were according to Tautz and Pfeifle with slight adjustments. Digoxigenin-labeled riboprobes were hybridized at 1:50 overnight at 55°C. Hybridization buffer was adjusted to pH 5. Alkaline-phosphatase-conjugated anti-digoxigenin antibody was used at 1:2000 (Roche). Ovaries were mounted in 60% glycerol in PBS.

Riboprobes were synthesized from cDNAs subcloned into a Bluescript vector(bcd) or a pNB40 vector (osk) using Ampliscribe transcription kit (Epicentre) and DIG RNA Labeling Mix (Roche). Following the transcription reaction, probes were hydrolyzed in carbonate buffer (2×:120 mmol/l Na₂CO₃, 80 mmol/l NaHCO₃, pH 10.2)for 20 minutes at 65°C. Hydrolysis was stopped with one volume 0.2 mol/l NaOAc (pH 6), and the probes were precipitated at -20°C with 0.1 volume 4 mol/l LiCl, 20 μg/ml tRNA, and 2.5 volumes chilled 100% ethanol. Probes were stored at -20°C in 200 μl hybridization buffer.

Ovaries were dissected and fixed as for RNA in-situ hybridization. After washing in PBT, ovaries were blocked for 1 hour in PBT + 1% Triton X-100, 1%BSA. Primary antibodies were diluted in PBT + 0.1% BSA and incubated overnight at 4°C. Anti-Grk (mouse, 1:30), anti-Orb (mouse, 1:30) and anti-N^intra (mouse, 1:1000) were from the Developmental Studies Hybridoma Bank. Anti-β-galactosidase antibodies were from Promega (mouse,1:1000) and Cappel (rabbit, 1:4000). Anti-PKCζ was from Santa Cruz Biotech (rabbit, 1:1000). Anti-Dhc (mouse, 1:100), anti-Baz (rabbit, 1:1000),anti-Sqd (mouse, 1:10) and anti-Osk (rabbit, 1:3000) were gifts from T. Hays,A. Wodarz, T. Schupbach and A. Ephrussi respectively.

Ovaries were washed 3×10 minutes in PBT + 0.1% BSA and incubated for 1 hour at room temperature with Alexa fluor-conjugated secondary antibodies at 1:500 (Molecular Probes). After 3×10-minute washes in PBT, ovaries were mounted in Aquapolymount (Polyscience) or Fluoromount G (Southern Biotech). Before mounting, ovaries were sometimes incubated for 15 minutes in 1 mg/ml Rnase A at 37°C and stained for 20 minutes with 10 μg/ml propidium iodide (Molecular Probes) in PBT. Alternatively, ovaries were stained for 20 minutes with rhodamine-conjugated phalloidin (Molecular probes) at 1:40.β-Galactosidase activity staining was performed as described previously.

Ovaries were dissected as for RNA in-situ hybridization and fixed for 10 minutes in 100% methanol at -20°C. After rehydration into PBT, ovaries were blocked as for antibody staining. FITC-conjugated anti-α-tubulin antibody (Sigma) was incubated at 1:250 in PBT + 0.1% BSA overnight at 4°C. Ovaries were washed 3×10 minutes in PBT + 0.1% BSA, incubated for 1 hour at room temperature with Alexa 488-conjugated secondary antibodies at 1:500 (Molecular Probes), and mounted in Aquapolymount or Fluoromount G. For partial MT depolymerization, ovaries were rocked at room temperature in 20μg/ml colchicine + 0.1% DMSO in modified Robb's medium(Theurkauf et al., 1992) for 10 minutes and rinsed in PBT before methanol fixation.

When the activity of the Drosophila PKA catalytic subunit, PKA-C1(also known as DC0), is reduced in oocytes, the MT cytoskeleton fails to reorganize properly at mid-oogenesis, leading to mislocalization of bcd and osk mRNAs (Lane and Kalderon, 1994). Movement of the oocyte nucleus from the posterior to the anterior cortex was not impaired, however(Fig. 1A), despite the fact that nuclear movement is dependent on MTs and is disrupted by several other mutations that affect oocyte MT polarity(Gonzalez-Reyes et al., 1995; Koch and Spitzer, 1983; Roth et al., 1995; Swan et al., 1999). Consequently, dorsal follicle cell fate, as visualized with the lacZenhancer trap mirror^6D1, is specified correctly in PKA mutants (Fig. 1B).

Germline clones of a PKA-C1 null allele, H2, showed over 90% penetrance of mid-oogenesis AP polarity defects, as detected by a kinesin-β-galactosidase (kin-β-gal) protein that marks the plus-ends of MTs and a GFP-Staufen (GFP-Stau) (St Johnston et al., 1991) protein that generally co-localizes with osk RNA (Fig. 1E,G). Such polarity defects can be a consequence of failed posterior follicle cell specification, as in grk and cornichon (cni)mutants, for example (Gonzalez-Reyes et al., 1995; Roth et al.,1995). However, the expression of the posterior follicle cell marker pointed^998/12 (pnt) was normal in 89% of egg chambers containing PKA-C1^H2 germline clones(n=187; Fig. 1C). pnt^998/12 expression was absent from posteriorly positioned follicle cells only when those cells also lacked PKA activity,consistent with previously noted cell-autonomous effects. Similarly,expression of anterior markers in posterior follicle cells surrounding PKA mutant germline clones was never observed for the centripetal cell marker BB127 (Fig. 1D) and only very rarely for the general anterior marker L53b and the border cell marker 5A7 (5% each, data not shown). The rare ectopic expression of L53b and 5A7 at the posterior is probably due to loss of PKA in the follicle cells and cannot plausibly be responsible for the >90% incidence of AP polarity phenotypes in PKA null germline clones. Thus, loss of PKA in the germline disrupted oocyte polarity without substantially affecting established markers of posterior follicle cell identity. These results support the hypothesis that PKA affects the oocyte MT network in mid-oogenesis in response to a normal posterior follicle cell signal.

As PKA activity is essential for normal oocyte polarity, it has been suggested that PKA might actively transduce the polarizing signal from the posterior follicle cells to the oocyte(Lane and Kalderon, 1994). Therefore, we tested whether normal regulation of PKA activity is required in this setting. The PKA holoenzyme (R₂C₂) is activated by binding of cAMP to its regulatory (R) subunits, thereby releasing active monomeric catalytic (C) subunits. A specific altered mouse PKA catalytic subunit, mC^*, binds poorly to regulatory subunits and is therefore constitutively active and largely unaffected by cAMP concentration or other inputs mediated by the regulatory subunits(Li et al., 1995). Expression of mC^* from an actin5C gene promoter (by flp-mediated conversion of the conditional actp>y⁺>mC^*transgene to actp>mC^*) rescued both the kin-β-gal and GFP-Stau mislocalization defects of PKA-C1^H2 germline clones from >90% to approximately 10% penetrance(Fig. 1E-H). The fusion of nurse cell membranes, another characteristic of PKA-C1 germline clones, was also fully rescued. Thus, PKA activity need not be regulated by cAMP or the regulatory subunits in order to promote normal oocyte polarity.

Even though mC^* efficiently rescued oocyte polarity and nurse cell fusion defects in PKA-C1^H2 germline clones, those eggs did not develop normally. Fewer than 4% hatched even if zygotically rescued by a wild-type paternal PKA-C1 allele, and most were either unfertilized or arrested development very early (see Table S1A in the supplementary material). Of the embryos that hatched, roughly half developed to adulthood if zygotically rescued, but all died before third instar in the absence of zygotic rescue. mC^* was also unable to rescue animals zygotically null for PKA-C1 (but derived from heterozygous parents)beyond second instar (see Table S1B in the supplementary material). Thus,there are PKA functions in oogenesis or early embryogenesis, as well as during larval development, that are not adequately substituted by mC^*.

Drosophila homologs of the Caenorhabditis elegans partition defective (par) genes are involved in organizing the germline MT cytoskeleton at mid-oogenesis (Pellettieri and Seydoux, 2002). Germline mutation of the Drosophila par-4 homolog, lkb1, can produce a mid-oogenesis phenotype similar to that of PKA mutants. As PKA can phosphorylate LKB1 at a specific site in vitro, and as rescue activity of an LKB1 transgene is reduced if the PKA site is substituted with an alanine (GFP-LKB1^S535A) and enhanced if substituted with a glutamate residue (GFP-LKB1^S535E),it has been suggested that LKB1 is a key target of PKA in oocyte polarity development (Martin and St Johnston,2003). If the sole role of PKA in AP polarity were to phosphorylate LKB1, then expressing the LKB1 transgene carrying the phosphorylation-mimicking mutation in the PKA site should bypass the need for PKA. However, PKA-C1^H2 germline clones that overexpress GFP-LKB1^S535E displayed 100% kin-β-gal mislocalization(n=19; Fig. 1J),whereas wild-type oocytes overexpressing the transgene did not exhibit any AP polarity defects (Fig. 1I). Thus, LKB1 does not seem to be a major target for PKA function in oocyte polarity.

In an effort to identify factors involved in transducing the mid-oogenesis signal with PKA, we performed a dominant enhancer screen in a PKAhypomorphic background. We found that the P-element allele l(3)j4B4dominantly enhanced the kin-β-gal mislocalization phenotype of the PKA hypomorph from less than 10% penetrance to 29%. In the absence of any PKA mutation, l(3)j4B4 homozygous germline clones made from either of two independent recombinant chromosomes showed over 90%kin-β-gal mislocalization in stage 9 (n=58; Fig. 2A) and close to 100%GFP-Stau mislocalization in stages 9 and 10 (n=106; Fig. 2C). Like PKAgermline clones, l(3)j4B4 germline clones showed no defect in movement of the oocyte nucleus to the anterior cortex (data not shown). To verify that the P-element insertion was the cause of the phenotype, we generated a revertant line in which the P-element had been mobilized and excised completely out of the genome. The revertant line was homozygous viable and showed 93% normal GFP-Stau localization (n=149; data not shown),a penetrance that is well within the range of GFP-Stau localization observed for various controls. The l(3)j4B4 P-element is inserted into the first intron of the sqd gene(http://flybase.org/). Thus, a P-element insertion in sqd is responsible for a strong defect in AP patterning in the oocyte.

To test whether l(3)j4B4 behaves like other alleles of sqd, we examined l(3)j4B4 germline clones for classic sqd phenotypes. DV defects, including eggshells with dorsal appendage material around their entire anterior circumference, specification of dorsal fate in all the follicle cells surrounding the anterior of the oocyte, and Grk protein accumulation along the entire anterior cortex of stage 9-10 oocytes,were observed for l(3)j4B4 germline clones (see Fig. S1A-D in the supplementary material). Furthermore, in both l(3)j4B4 germline clones and sqd¹ mutant ovaries, the nurse cell chromosomes failed to disperse completely beyond stage 6 (see Fig. S1F in the supplementary material) (Goodrich et al.,2004), in contrast to wild-type oocytes, in which the nurse cell chromosomes remain polytene only for the early stages of oogenesis (see Fig. S1E in the supplementary material, arrow). Finally, no Sqd protein was detected in the nurse cells or oocyte in l(3)j4B4 germline clones(see Fig. S1H in the supplementary material). Thus, l(3)j4B4 mutants clearly are deficient for Sqd activity, displaying all the previously described oogenesis phenotypes characteristic of sqd mutants. We conclude that l(3)j4B4 is indeed an allele of sqd, and we now refer to it as sqd^j4B4.

Because sqd^j4B4 shows a highly penetrant defect in AP patterning during oogenesis, we examined whether other sqd alleles also exhibit a similar phenotype. We found GFP-Stau to be substantially mislocalized in stage 9-10 oocytes of escaper females from all strong sqd allelic combinations tested, including sqd^j4B4/sqd^ix50,sqd^j4B4/sqd^ix77, sqd^j4B4/Df,sqd^ix77/Df, as well as in females containing sqd^ix77 germline clones(Table 1A). For all these genotypes, we often observed a small amount of GFP-Stau in the correct location at the posterior of the oocyte, in addition to the ectopic GFP-Stau in the center of the oocyte, and the incidence of oocytes with normal localization was consistently higher in later stages. We also observed GFP-Stau mislocalization in homozygotes of the weak, viable allele sqd¹ and in sqd^j4B4/sqd¹trans-heterozygotes (Table 1A). In these genotypes, GFP-Stau rarely accumulated as a `cloud' in the center of the oocyte as it does in sqd<sup>j4B4</sup> germline clones; rather it seemed to aggregate in `flecks' in the ooplasm or in ectopic locations along the lateral cortex (Fig. 2D). Additionally, all sqd alleles showed aberrant GFP-Stau aggregation surrounding the nuclei of the nurse cells(Fig. 2J). A SqdS transgene that expresses one of the three described Sqd splice variants(Norvell et al., 1999) rescued the partial lethality of sqd^j4B4 in combination with other strong sqd alleles and also restored normal GFP-Stau localization in those females (Table 1A,B). Both the lethality complementation tests and the GFP-Stau mislocalization data suggest that sqd^j4B4 is a stronger allele than the previously described allele sqd^ix77, which is the result of a small deletion in the 5′ UTR(Kelley, 1993)(Table 1). Germline clones of the probable null allele, sqd^ix50(Kelley, 1993), did not produce vitellogenic egg chambers. We conclude from these data that all sqd alleles disrupt AP polarity during oogenesis to a degree commensurate with their allelic strength.

In order to characterize further the effect of Sqd on AP patterning, we examined key molecules that normally localize to the posterior of oocytes at stage 9 and later. osk RNA was mislocalized in 95% of sqd^j4B4 germline clones (n=61), usually to the center of the oocyte (Fig. 2E). A less penetrant transient mislocalization of osk RNA at the anterior of stage 9 sqd¹ oocytes was recently reported(Norvell et al., 2005). We observed some anterior osk RNA in a small percentage of sqd^j4B4 germline clones but not in sqd^j4B4/sqd¹ oocytes. In sqd^j4B4/sqd¹ oocytes, osk RNA was mislocalized at high penetrance (74%, n=163), typically appearing as`flecks' in the ooplasm or along the lateral cortex in a pattern similar to that of mislocalized GFP-Stau in oocytes of this genotype(Fig. 2F).

Dynein heavy chain (Dhc) is also localized to the posterior pole of wild-type stage 9 oocytes (Fig. 2G) (Li et al.,1994), but it never localized to the posterior pole in sqd^j4B4 germline clones(Fig. 2H). Thus, in sqd mutant oocytes at stages 9-10, localization of posterior factors,including the MT polarity marker kin-β-gal(Fig. 2A), is universally disrupted, suggesting an underlying defect in MT organization.

We therefore examined MT distribution directly with α-tubulin antibody. In wild-type stage 8 and 9 oocytes, α-tubulin staining is higher at the anterior than at the posterior(Fig. 3A), because MTs are nucleated primarily at the anterior cortex(Theurkauf et al., 1992) and are excluded from the posterior pole (Cha et al., 2002). In sqd^j4B4 germline clones, we detected MTs at uniform density all around the oocyte cortex, indicating that in these mutants the asymmetrical mid-oogenesis MT array does not form correctly (Fig. 3B).

We did not find any combination of sqd alleles that shows a significantly greater penetrance of mislocalization for osk RNA than for kin-β-gal. In the viable trans-heterozygote, sqd^j4B4/sqd¹, kin-β-gal was improperly localized at 65% penetrance, either in the center of the oocyte or diffusing away from the posterior pole (n=165; Fig. 2B). This penetrance is comparable to the 65% GFP-Stau mislocalization and 74% oskmislocalization observed in oocytes of this genotype(Table 1A and Fig. 2D,F), implying that the mislocalization of osk RNA in this and other sqd allelic combinations results principally from aberrant MT organization. This is in contrast to, but not mutually exclusive with, the suggestion that Sqd guides osk RNA to the posterior by a direct interaction(Norvell et al., 2005). Our observation that GFP-Stau particles cluster aberrantly around the periphery of nurse cell nuclei in sqd mutants supports the idea that Sqd affects the behavior of complexes likely to include osk RNA, despite the fact that we did not observe any specific effect of Sqd on osk RNA localization independent of the MTs.

Cytoplasmic streaming normally occurs at stages 10b-12 and depends on reorganization of the MT cytoskeleton after stage 10a(Theurkauf, 1994; Theurkauf et al., 1992). In cappuccino (capu) mutants, mislocalization of oskRNA is the result of premature cytoplasmic streaming during stages 8-10a(Manseau et al., 1996). In sqd^j4B4 germline clones, we observed normal streaming at stage 10b but not earlier (data not shown), indicating that the MT arrangements that govern streaming do not depend on Sqd function.

In a subset of polarity mutants, including PKA, bcd RNA localizes ectopically to the oocyte posterior at stages 8-10; it has been suggested that this reflects localization to an MTOC aberrantly persisting at the oocyte posterior after stage 7 (Lane and Kalderon, 1994). We did not detect ectopic bcd at the posterior of sqd^j4B4 mutant oocytes (n=40, Fig. 3H), even though it was readily observed in PKA-C1^H2 germline clones (31%, n=26, data not shown), as well as in another polarity mutant, mago nashi (mago; Fig. 3G).

To further investigate the nature of the MT defects in sqd and PKA mutants, we examined MTs with α-tubulin antibody following partial depolymerization with colchicine. In wild-type stage 8 and 9 oocytes,MT stubs emanate mostly from the anterior cortex following partial depolymerization (Fig. 3C),reflecting the presence of an anterior MTOC after stage 7(Theurkauf et al., 1992). By contrast, partial MT depolymerization in sqd^j4B4 germline clones revealed MTs emanating from the entire oocyte cortex at stages 8 and 9(Fig. 3D,F). Likewise, in PKA-C1^H2 germline clones, MT stubs could be detected all around the oocyte cortex following partial depolymerization at stages 8-10. Additionally, however, a strong focus of MTs sometimes could be detected at the posterior pole of PKA-C1^H2 germline clones(Fig. 3E), whereas such staining was never seen in sqd^j4B4 germline clones. This suggests that the whole cortex, including the posterior pole, nucleates MTs in both sqd and PKA mutants but that only PKA mutants retain a discrete posterior MTOC. The cortical MTs that persist at the posterior pole in sqd mutants presumably do not represent a true organizing center, as they are not capable of supporting ectopic posterior bcd RNA, in contrast to those of the PKA mutant.

Sqd has been shown to affect grk RNA localization and translation in mid- to late oogenesis(Neuman-Silberberg and Schupbach,1993; Neuman-Silberberg and Schupbach, 1996; Norvell et al., 1999). If grk were also affected in early oogenesis and unable to signal to the posterior follicle cells, this could account for the mid-oogenesis AP polarity defect(Gonzalez-Reyes et al., 1995; Roth et al., 1995). In wild-type oocytes, Grk forms a cap at the posterior of stage 2-6 oocytes(Neuman-Silberberg and Schupbach,1993) (Fig. 4A). We observed normal Grk localization in only 48% of sqd^j4B4germline clones at those stages, with Grk localized diffusely throughout the ooplasm in the remaining cases (n=86; Fig. 4B).

To test whether early Grk signaling suffices to specify posterior follicle cells in sqd^j4B4 germline clones, we examined two established markers of follicle cell fate. Expression of the posterior follicle cell marker, pnt^rm254, was normal in egg chambers where the germline was mutant for sqd(Fig. 6A). Additionally, we did not see ectopic expression of the anterior centripetal cell marker, BB127, in posterior follicle cells surrounding sqd mutant oocytes (Fig. 6B). Thus, the early grk signal, which is required for pnt^rm254expression, is transmitted appropriately from sqd mutant oocytes, and we conclude that the mid-oogenesis phenotype of sqd oocytes is not the result of an earlier defect in grk signaling or in posterior follicle cell specification.

Although the anterior and posterior follicle cells are specified correctly in egg chambers containing sqd^j4B4 germline clones, the migration of the anterior border and centripetal cells in those egg chambers was defective at high penetrance (see Fig. S2F in the supplementary material).

The mislocalization of Grk protein in early sqd mutant oocytes led us to question whether early sqd oocytes display other polarity defects. Like Grk, Orb protein and osk RNA localize in a cap at the posterior of wild-type oocytes in stages 2-6(Fig. 4C and data not shown). As was found for Grk, Orb was diffusely localized in the ooplasm of early sqd mutants, at variable penetrance ranging from 12-52%(n=339; Fig. 4D); the penetrance increased with the age of the females, as was the case for various other defects (see below). osk RNA was also mislocalized in a manner similar to Orb and Grk in a subset of early oocytes (data not shown). Likewise, stage 2-6 sqd^j4B4 oocytes often had an even distribution of MTs throughout their cytoplasm(Fig. 4F), in contrast to wild-type oocytes, which have a strong focus of tubulin staining at the posterior, indicative of the posterior MTOC(Fig. 4E). Thus, sqdmutants display abnormal distributions of germline MTs in both early and mid-oogenesis.

Ovarioles containing sqd^j4B4 germline clones exhibited a number of additional phenotypes likely to originate in the germarium. These increased in penetrance over time, presumably reflecting progressive depletion of perduring Sqd gene products below functional thresholds in the germarium. Thus, for animals with germline clones induced during the third larval instar,the penetrance of these phenotypes collectively increased from 17% of early egg chambers in 1-3-day-old females (n=127; 6-8 days after clone induction) to 89% in 10-12-day-old females (n=280; 15-17 days after clone induction).

The most prominent of these phenotypes was the absence of stalks separating adjacent egg chambers (see Fig. S2B in the supplementary material). Also, many egg chambers contained aberrant numbers of germ cells, from as few as two (see Fig. S2C in the supplementary material) to well in excess of 16. Furthermore,the number of germ cells in adjacent egg chambers generally did not add up to 16, nor were excess germ cells present in multiples of 16. Cysts with 16 nurse cells and no oocyte were also observed (data not shown), although more commonly, cysts with no oocyte also contained abnormal numbers of germ cells(see Fig. S2B in the supplementary material). Additionally, in egg chambers containing the proper complement of nurse cells and an oocyte, the oocyte was sometimes mispositioned within the cyst, failing to reside at the posterior(see Fig. S2D in the supplementary material). In 10-12-day-old females, a number of egg chambers beyond stage 6 were seen to degenerate. The scoring of defects in AP polarity was always performed using young females and only included egg chambers with 16 germ cells and a normally positioned oocyte.

Normally, osk RNA is translationally repressed until it becomes localized to its final destination at the posterior pole. Consequently, Osk protein can be detected only at the posterior of stage 9-10 oocytes(Kim-Ha et al., 1995)(Fig. 5A). In sqd^j4B4 germline clones, Osk protein was seen in the center of stage 9 and 10 oocytes at 29% penetrance(Fig. 5B); no Osk protein was visible in 59% of mutant oocytes, and in the remaining cases (12%) Osk protein was detected only at the posterior (n=259). We never observed ectopic osk translation in sqd oocytes before stage 9 or in the nurse cells at any stage. An osk translational reporter, m1⁴¹⁴lacwt, which carries the osk 5′ and 3′ UTRs flanking the lacZ gene(Gunkel et al., 1998),confirms the observations made with Osk antibody. sqd^j4B4germline clones showed 34% ectopic β-galactosidase localization in stages 9-10 (n=58) but no ectopic expression in earlier stages(Fig. 5D). Recently, it was reported that excessive PKA activity can result in osk translation before the RNA is localized, whereas translation of normally localized osk can be reduced by decreasing PKA activity(Yoshida et al., 2004). This led to the suggestion that PKA is required for osk translation. We therefore looked at Osk protein in germline clones that were doubly mutant for both PKA-C1^H2 and sqd^j4B4. Mislocalized Osk was still seen in 27% of PKA-C1, sqd double mutants(n=15, Fig. 5F),compared with 18% of sqd single mutant sibling controls in this experiment (n=130). Furthermore, we also found ectopic Osk away from the posterior in 9% of PKA-C1^H2 mutant oocytes at stages 9 and 10, even in the absence of any sqd mutation(Fig. 5E). Most PKA-C1^H2 mutant oocytes had no detectable Osk, although 24% had Osk solely at the posterior (n=164). Thus, the absence of PKA in the germline does not prevent Osk translation, either at the posterior or in the center of stage 9-10 oocytes.

To determine whether Sqd affects polarity of somatic cells in addition to its role in polarizing the oocyte, we examined the localization of several markers of apical polarity in sqd^j4B4 follicle cell clones. We did not detect any effect on the localization of the apical proteins Notch (N), Bazooka (baz), or Atypical protein kinase C (aPKC) in sqd mutant follicle cell clones, even as long as 15 days after clone induction (Fig. 6C). Additionally, we did not detect any defects in planar cell polarity in the trichome bristles of the wing blade, either in sqd^j4B4clones or in the wings of sqd^j4B4/Df escaper flies (Fig. 6D), and no planar polarity or other defects were apparent in sqd^j4B4homozygous eye clones (data not shown). Thus, Sqd acts specifically in the female germline to organize MTs and establish polarity.

It is well established that the hnRNP Sqd participates in Drosophila DV axis formation(Kelley, 1993; Neuman-Silberberg and Schupbach,1993; Neuman-Silberberg and Schupbach, 1996; Norvell et al., 1999). Following a screen to identify factors that interact with PKA in mid-oogenesis, we discovered that Sqd is essential also for AP axis formation. The localization of posterior factors, including oskRNA, GFP-Stau, kin-β-gal, and Dhc, was disrupted at stages 9-10 in all sqd allelic combinations tested, was highly penetrant in strong sqd alleles, and could be rescued by expression of a Sqd cDNA transgene. These defects can be attributed to the failure of sqdmutants to establish a normally polarized MT array at mid-oogenesis. We also observed defects in MT organization and in the localization of posterior factors, including Grk protein, in sqd mutant oocytes at stages 2-6. Despite the imperfect localization of Grk before stage 6, posterior fate appears to be specified normally in the follicle cells overlying sqdmutant oocytes. Thus, sqd mutations affect germline-specific processes required for the polarization of MTs in both early and mid-oogenesis.

Loss of PKA in the germline did not affect MT polarity in early oogenesis,as judged by Orb localization in stages 2-6 (data not shown), but, similarly to sqd mutants, it disrupted MT polarity in mid-oogenesis without discernibly altering follicle cell fate. However, despite several similarities, we observed a significant difference in the MT organization of sqd and PKA mutants at stages 8-10, as discussed in greater detail below.

In grk and cni mutants, where the posterior follicle cells do not differentiate properly, both the subsequent RNA localization defects and the failure of the oocyte nucleus to migrate from the posterior to the anterior have been attributed to defects in MT reorganization resulting from loss of a posterior follicle cell signal(Gonzalez-Reyes et al., 1995; Roth et al., 1995). In PKA and sqd mutant oocytes, the anterior migration of the oocyte nucleus, which depends on MT function(Koch and Spitzer, 1983), was unaffected, despite the accompanying MT defects and the highly penetrant mislocalization of osk RNA and other posterior factors. Thus, it appears that either discrete aspects of the MT organization, which direct nucleus migration, are spared in PKA and sqd mutants or the overall disruption of MT organization by loss of PKA or Sqd is simply less severe than that caused by loss of posterior follicle cell fate.

The normal organization of MTs in stage 8-10 oocytes is not entirely clear. In addition to MTs nucleated at the anterior cortex, MTs have been proposed to emanate from all cortical positions, with the exception of the posterior pole(Cha et al., 2001; Cha et al., 2002). This assertion is based on the observations that components associated with MT minus-ends, such as γ-tubulin and the centrosome component Centrosomin(Cnn) (Cha et al., 2002), can be seen along the entire oocyte cortex, and that injected bcd RNA localizes to the lateral cortices as well as the anterior, but not to the posterior pole (Cha et al.,2001). Hence, normal posterior localization of osk RNA may require the clearing of MTs nucleated both from a discrete posterior MTOC established before stage 6 and from dispersed cortical sites established after stage 7.

Staining with α-tubulin antibody following partial MT depolymerization revealed MT stubs emanating mostly from the anterior in wild-type oocytes, whereas PKA and sqd mutant oocytes retained short MTs around the entire oocyte cortex, including the posterior pole. Some PKA mutant oocytes also showed an elevated posterior concentration of MTs not seen in sqd mutant oocytes. Thus, it appears that the primary MT defect in sqd mutants is the failure to eliminate cortical sites of MT nucleation beyond stage 7, whereas PKA mutants additionally retain a posterior MTOC beyond stage 6. This hypothesis can explain why ectopic bcd RNA localizes at the posterior of PKA mutant oocytes but not sqd mutant oocytes. It should,however, be noted that since classical MTOC components, such asγ-tubulin, are present along the entire oocyte cortex at stages 9-10 even in wild-type oocytes (Cha et al.,2002) (J, Steinhauer, unpublished data), the inference of a discrete posterior MTOC from partial MT depolymerization experiments cannot be confirmed directly.

In a proportion of sqd mutant stage 2-6 oocytes, Grk, Orb, osk RNA and MTs were distributed evenly throughout the ooplasm rather than localizing in a cap at the oocyte posterior. Although these defects did not appear to cause the subsequent AP defects by preventing posterior follicle cell specification, we cannot rule out the possibility that the early and late polarity phenotypes are causally related in some other way. For instance, it is possible that a molecule(s) required at the posterior of the oocyte for the MT reorganization at stages 7-8 is improperly localized by stage 6 in sqd mutants, as are Grk, Orb and osk RNA. If MT rearrangements were very sensitive to the localized concentration of this hypothetical regulator, an early polarity defect of apparently low penetrance could be translated into a much more penetrant polarity phenotype at mid-oogenesis.

sqd is not the only mutant to cause polarity defects in both early and mid-oogenesis. For example, defects in early polarity are caused by mutations in Armitage (Armi), a component of the RNA silencing machinery, and these defects were proposed to be the cause of a mid-oogenesis AP polarity phenotype (Cook et al., 2004). However, we found that pnt^998/12 expression was not disrupted in armi¹ homozygotes (J. Steinhauer,unpublished). Weak par-1 alleles also affect mid-oogenesis polarity without affecting posterior follicle cell fate(Shulman et al., 2000),whereas strong alleles disrupt early polarity severely, causing oocyte identity to be lost (Cox et al.,2001; Huynh et al.,2001). Thus, for several mutations, including sqd, armiand par-1, it is unclear whether MT organization is disrupted independently at two distinct phases of development or whether there is a causal connection between the early and later polarity phenotypes that is not evident as a failure in posterior follicle cell specification. In either case,a single molecular target might account for both the early and mid-oogenesis phenotypes.

Several additional phenotypes became prevalent in older sqd^j4B4 germline clones, rising to very high penetrances after 2 weeks. Among the varied late onset sqd phenotypes, the oocyte sometimes was mispositioned within the egg chamber, even in those egg chambers containing the normal complement of nurse cells to oocyte. This phenotype can arise in several ways, including as a result of delayed oocyte specification(Gonzalez-Reyes et al., 1997). A role for Sqd in oocyte specification is supported by the presence of cysts with 16 nurse cells and no oocyte in these older ovarioles. In other cases,cysts with fewer than 16 germ cells were observed, implicating Sqd in cystocyte mitosis. Both oocyte specification and the normal cystocyte divisions depend on specific arrangements of the MT cytoskeleton in the germarium (Huynh and St Johnston,2004). Thus, it is likely that some of these late onset sqd phenotypes, like the polarity phenotypes, are caused by a defect in regulating MT dynamics.

In sqd^j4B4 germline clones, we noticed the accumulation of Osk protein in the cytoplasm of stage 9-10 oocytes. A similar observation was reported for mutations in another hnRNP A/B family member, Hrp48 (Hrb27C -FlyBase) (Yano et al., 2004). Normally, Osk protein accrues only at the posterior cortex, and translation of osk RNA is presumably repressed elsewhere(Kim-Ha et al., 1995). Therefore, loss of sqd may cause de-repression of osktranslation. However, we did not see ectopic osk translation in stage 6-8 sqd mutant oocytes, detected either with Osk antibody or with an osk translation reporter, in contrast to the premature expression observed with the osk translation reporter in hrp48 mutants(Yano et al., 2004) or with similar reporters lacking specific repressor elements(Gunkel et al., 1998; Kim-Ha et al., 1995). Thus,although we do not discount the idea that sqd is directly involved in translational regulation of osk, we propose an alternative hypothesis for the ectopic Osk protein accumulation in sqd mutants. As most of the posterior components that we examined were mislocalized to the center of sqd^j4B4 oocytes, we believe that the primary AP defect in sqd mutants is that the MT plus-ends are focused incorrectly at the center of the oocyte. Hence, all the necessary components for osktranslation may be localized together, and we hypothesize that the osk translation machinery is assembled and activated in the middle of the sqd mutants as it normally is at the posterior of wild-type oocytes at stage 9.

We also detected a low penetrance of ectopic Osk protein in PKAmutants. The scenario outlined above could be true for PKA mutants as well. Regardless of the mechanism, it is clear from this result that PKA is not absolutely required for Osk translation, although it may enhance osk translation, as previously suggested(Yoshida et al., 2004).

Although sqd was identified in a screen for modifiers of PKA in oocyte polarity, retesting with various alleles indicated that there is not a strong genetic interaction between the two loci (data not shown). Both Sqd and PKA act in mid-oogenesis to reorganize the oocyte MTs in response to a normal posterior follicle cell signal, but specific MT defects differ between the two mutants, as discussed above. Thus, they probably have different targets and mechanisms in this complex process.

The hnRNP Sqd is an RNA-binding protein. Another hnRNP of the same family,Hrp48, is also required for MT reorganization at mid-oogenesis(Yano et al., 2004). Sqd and Hrp48 bind each other in vitro, cooperate in grk RNA localization(Goodrich et al., 2004) and have similar localization patterns throughout oogenesis(Matunis et al., 1994; Yano et al., 2004) (see Fig. S1G in the supplementary material). Thus, one might expect these two proteins to act together in MT reorganization. Although we were unable to detect a strong genetic interaction between sqd and hrp48 in AP polarity (data not shown), we speculate that they are collectively necessary for the localization and translation of one or a small number of specific RNA molecules required for MT repolarization at mid-oogenesis.

hnRNPs normally participate in the processing of many RNAs(Dreyfuss et al., 2002), but their generic functions may be partially redundant, so that, for example, in sqd mutants, continued cell viability is not impaired despite the presence of a strong AP polarity defect. Our ability to induce large,persistent somatic cell clones for sqd^j4B4 without causing any polarity or other phenotypes supports this idea. Follicle cell polarity was also normal in PKA mutant clones (data not shown). Thus, the disruptions in MT polarity that we observe for both PKA and sqd mutants represent specialized functions of these proteins in germline cells.

We thank: Yan Zhang for data on rescue of PKA phenotypes by mC^*; Andrew Tomlinson for data on sqd^j4B4 eye clones; Yan Zhang, Kelli Chung, Rahul Jhaveri, and Helen Qiujiajing for conducting the PKA modifier screen; Bloomington Stock Center, Trudi Schupbach, Amanda Norvell, Anne Ephrussi, Daniel St. Johnston, Kathleen Smith,Gary Struhl, Developmental Studies Hybridoma Bank, Tom Hays, and Andreas Wodarz for reagents; and Tulle Hazelrigg for invaluable assistance with confocal microscopy. This work was supported by a National Science Foundation Graduate Research Fellowship to J.S. and NIH grant GM062938 to D.K.