Oskar is targeted for degradation by the sequential action of Par-1, GSK-3, and the SCF⁻Slimb ubiquitin ligase.

Translation of oskar messenger RNA (mRNA) is activated at the posterior of the Drosophila oocyte, producing Long Oskar, which anchors the RNA, and Short Oskar, which nucleates the pole plasm, containing the posterior and germline determinants. Here, we show that Oskar is phosphorylated by Par-1 and GSK-3/Shaggy to create a phosphodegron that recruits the SCF(-Slimb) ubiquitin ligase, which targets Short Oskar for degradation. Phosphorylation site mutations cause Oskar overaccumulation, leading to an increase in pole cell number and embryonic patterning defects. Furthermore, the nonphosphorylatable mutant produces bicaudal embryos when oskar mRNA is mislocalized. Thus, the Par-1/GSK-3/Slimb pathway plays important roles in limiting the amount of pole plasm posteriorly and in degrading any mislocalized Oskar that results from leaky translational repression. These results reveal that Par-1 controls the timing of pole plasm assembly by promoting the localization of oskar mRNA but inhibiting the accumulation of Short Oskar protein.

Introduction

Asymmetric accumulation of proteins within single cells is critical to establish polarity, segregate cell fate determinants, and to specify the body axis of multicellular organisms. In Drosophila, Oskar (Osk) protein accumulates at the posterior of the oocyte to direct the formation of a specialized cytoplasm, the pole plasm, which determines posterior patterning and germ cell formation after fertilization (Breitwieser et al., 1996; Markussen et al., 1995; Vanzo and Ephrussi, 2002). The levels of Osk protein directly correlate with the number of germ cells and the embryonic pattern (Ephrussi and Lehmann, 1992; Smith et al., 1992). Furthermore, ectopic expression of Osk at the anterior induces a duplication of posterior structures and specification of anterior pole cells (Ephrussi and Lehmann, 1992; Smith et al., 1992). The localized accumulation of Osk is therefore strictly controlled at several levels, including the localized transport and translation of its messenger RNA (mRNA) and the anchoring of osk mRNA and protein at the posterior (for review, see Kugler and Lasko, 2009). There are two Osk protein isoforms produced by selective use of two alternative in-frame start codons in the Osk coding sequence. Long Osk is required to anchor osk mRNA and Short Osk protein to the posterior cortex but does not function as a determinant, whereas Short Osk initiates the assembly of the pole plasm (Breitwieser et al., 1996; Markussen et al., 1995; Vanzo and Ephrussi, 2002).

Par-1 kinase is a major regulator of cell polarity in multiple biological contexts (St Johnston and Ahringer, 2010). Par-1 localizes at the oocyte posterior and controls the formation of a polarized microtubule cytoskeleton by excluding microtubule minus ends from the oocyte posterior (Doerflinger et al., 2010; Shulman et al., 2000; Tomancak et al., 2000). This polarized microtubule network then directs the localization of osk mRNA to the posterior pole of the oocyte (Zimyanin et al., 2008). It has also been reported that Par-1 directly phosphorylates Osk protein to prevent its degradation (Riechmann et al., 2002). Although this work did not decipher the molecular mechanism involved in Osk destabilization, it suggests that the regulation of Osk protein stability might confer another layer of control over the level of this crucial posterior determinant.

Here, we show that Osk stability is regulated by Slimb (β-TRCP), an F-box protein that is the substrate recognition subunit of the SKP1/Cullin/F-box (SCF) protein E3 ubiquitin ligase complex that targets proteins for proteosomal-dependent degradation (Cardozo and Pagano, 2004). Slimb recognizes the doubly phosphorylated DpSGXXpS destruction motif (Degron) present in Armadillo/β-catenin and IκB (Maniatis, 1999). Sgg/GSK-3 kinase is required for Slimb-mediated degradation of Armadillo/β-catenin and Cubitus interruptus (Ikeda et al., 1998; Smelkinson and Kalderon, 2006). Importantly, Sgg/GSK-3 phosphorylation requires a priming phosphorylation by another kinase (Dajani et al., 2001; Fiol et al., 1987; Jia et al., 2002; Price and Kalderon, 2002). We show that Par-1 primes Osk for GSK-3 phosphorylation to form a phosphodegron, thereby targeting Osk for degradation, rather than stabilizing it as previously proposed.

Results

Slimb Is Recruited to the Oocyte Posterior by Short Osk

We recently identified slimb, the ortholog of human β-TRCP, in a germline clone screen for mutants that disrupt the polarity of the oocyte (E.M. and D.StJ., unpublished data; Martin et al., 2003). During the course of this work, we discovered that Slimb protein accumulates at the oocyte posterior from stage 9 onward, i.e., after the oocyte has been polarized, suggesting an additional role during Drosophila oogenesis. We therefore investigated how Slimb is targeted to the posterior cortex using a GFP-tagged Slimb transgene that faithfully reproduces the localization of the endogenous protein. GFP-Slimb is first recruited to the posterior at stage 9 of oogenesis and remains there throughout most of mid/late oogenesis, but it largely disappears by stage 14 when the oocytes have matured (Figure 1A).

Figure 1

Slimb Accumulates at the Posterior of the Oocyte in an Osk-Dependent Manner

(A) GFP-Slimb localizes to the oocyte posterior from stage 9 onward, but this localization is reduced in mature oocytes.

(B) GFP-Slimb localization in an oocyte from a female treated with colcemid to depolymerize microtubules. Mislocalization of the oocyte nucleus confirms the efficiency of drug treatment.

(C–H) GFP-Slimb localization in mutants that disrupt oocyte polarity or pole plasm assembly: grk/grk (C), par-1/par-1 (D), stau (E), and osk/Df(3R)p (F) abolish the posterior localization of Slimb, indicating that its posterior recruitment depends on the localization of Osk protein, whereas Slimb localizes at the posterior in vas/Df(2L)A267 (G) and vas/Df(2L)A267 (H), which block the assembly of the pole plasm downstream of Osk. (A–H) UAS-GFP-Slimb was expressed in the germline using a nanos-GAL4:VP16 driver.

To determine how Slimb is recruited to the posterior, we analyzed its localization under various conditions that disrupt the polarization of the oocyte and the formation of the pole plasm. Treatment with the microtubule-depolymerizing drug, colcemid, blocks the posterior enrichment of Slimb, suggesting that this depends on a polarized microtubule cytoskeleton (Figure 1B). Consistent with this, Slimb is not localized in gurken and par-1 mutants, which disrupt the polarity of the oocyte upstream of microtubule organization (González-Reyes and St Johnston, 1994; Roth et al., 1995; Shulman et al., 2000; Tomancak et al., 2000) (Figures 1C and 1D). A key function of the polarized microtubule network in the oocyte is to direct the transport of osk mRNA to the posterior cortex (Clark et al., 1994; Zimyanin et al., 2008). osk mRNA localization and anchoring at the posterior require the RNA-binding protein, Staufen (Stau), which colocalizes with the mRNA at the posterior pole (Ephrussi et al., 1991; Kim-Ha et al., 1991; St Johnston et al., 1991). stau mutants also abolish Slimb localization, indicating that Slimb recruitment to the posterior is downstream of the localization of Stau and osk mRNA (Figure 1E).

We next asked whether Osk protein is required for Slimb localization using a nonsense allele, osk, in which osk mRNA localizes normally at stage 9, but produces no protein (Ephrussi et al., 1991; Kim-Ha et al., 1991; St Johnston et al., 1991). Slimb is not localized posteriorly in osk/Df oocytes, demonstrating that Slimb recruitment depends on Osk protein or a downstream component of the pole plasm (Figure 1F). To distinguish between these possibilities, we analyzed Slimb localization in a hypomorphic allele of vasa, vas, which lacks catalytic activity, preventing the subsequent steps of pole plasm assembly (Liang et al., 1994). GFP-Slimb still accumulates posteriorly in vas but shows a reduced localization in the stronger vasa mutant allele, vas (Figures 1G and 1H). This latter allele abolishes the posterior localization of Vasa and reduces the amount of posterior Osk protein, because Vasa participates in a positive feedback loop to enhance Short Osk accumulation (Breitwieser et al., 1996; Markussen et al., 1995; Rongo et al., 1995). Thus, Slimb seems likely to be directly recruited to the posterior by Osk, because its localization does not depend on Vasa or any downstream steps in pole plasm assembly and correlates with the amount of localized Osk protein.

To determine whether the posterior recruitment of Slimb depends on the Long or Short isoform of Osk, we examined the localization of endogenous Slimb in osk RNA null oocytes expressing either an osk transgene that can only produce Short Osk or an osk transgene that produces only Long Osk (Markussen et al., 1995). Long Osk is necessary and sufficient for the anchoring of Short Osk and Stau protein at the posterior (Vanzo and Ephrussi, 2002). Consistent with this, when Short Osk is expressed alone, Stau and Osk protein spread away from the posterior cortex, and Slimb localizes in these detached particles (Figures 2A and 2B). By contrast, Slimb is absent from the posterior of oocytes expressing only Long Osk, despite the normal localization of Stau (Figure 2C). These results imply that the posterior recruitment of Slimb depends specifically on Short Osk. To confirm this hypothesis, we targeted either Long or Short Osk to the anterior of the oocyte using transgenes in which the osk 5′UTR and coding region are fused to bcd 3′UTR (Ephrussi and Lehmann, 1992; Tanaka and Nakamura, 2008). Expression of Short Osk (osk-bcd3′UTR), but not Long Osk (osk-bcd3′UTR), recruits GFP-Slimb to the anterior (Figures 2D and 2E). Furthermore, ectopic localization of Short Osk recruits Slimb during earlier stages to sites in which bcd mRNA is normally located (Figure 2F). Thus, Short Osk is necessary and sufficient for Slimb localization at either the anterior or posterior of the oocyte, suggesting that it recruits Slimb directly.

Figure 2

Slimb Localization Depends on Its Substrate Interaction Domain and Short Osk

(A–F) Slimb is recruited by Short Osk, but not Long Osk. (A–C) Egg chambers of the indicated genotypes were stained for Stau (Green) and Slimb (red). (A) Endogenous Slimb forms a posterior crescent in wild-type oocytes. (B) Slimb still colocalizes with Stau at the posterior of UAS-osk;osk/Df(3R)p oocytes, which only express Short Osk, although Stau and osk mRNA are not anchored at the posterior. (C) Slimb is not localized to the posterior of UAS-osk; osk/Df(3R)p oocytes that express only Long Osk. (D) GFP-Slimb (green) is not recruited to the anterior by Long Osk expressed from UAS-oskM139L-bcd3′UTR but still localizes posteriorly with endogenous Osk. (E and F) GFP-Slimb is recruited to the anterior by Short Osk expressed from UAS-osk-bcd3′UTR in stage 10A (E) and stage 7 (F) oocytes.

(G and H) Slimb’s substrate interaction domain is necessary and sufficient for posterior localization. (G) GFP-Slimb-Δ-WD40 does not localize to the posterior. (H) GFP-Slimb-Δ-F-Box, which contains the substrate interaction domain, but lacks the F-Box domain, localizes to the oocyte posterior. A schematic representation of the domains is shown. (A–H) GFP-Slimb and Osk transgenes were expressed using nanos-GAL4:VP16.

See also Figure S1.

In order to better understand the mechanism of Slimb recruitment to the posterior of the oocyte, we produced transgenes expressing GFP-Slimb lacking the F-Box (Slimb-Δ-F-Box) or the WD40 repeats (Slimb-Δ-WD40). As the F-Box domain is required for Slimb’s interaction with the rest of the SCF complex and the WD40 repeats bind to SCF-Slimb substrates, the localization of these truncated GFP-Slimb transgenes should indicate if Osk recruits Slimb directly because it is a substrate or indirectly by recruiting the rest of the SCF complex. GFP-Slimb-Δ-F-Box localizes to the posterior as efficiently as full-length Slimb, whereas GFP-Slimb-Δ-WD40 does not accumulate posteriorly (Figures 2G and 2H). Thus, Slimb localization at the posterior depends on its substrate interaction domain but is independent on its interaction with the rest of SCF complex.

To address if Slimb is a stable component of the pole plasm or continually turns over at the posterior, we performed fluorescence recovery after photobleaching (FRAP) experiments with GFP-Slimb-Δ-F-box (Figure S1 available online). The GFP-Slimb-Δ-F-box signal returns to around 60% of its prebleaching level in 10 min, which is comparable to other dynamically localized components within the oocyte, such as grk and bcd mRNAs (Jaramillo et al., 2008; Weil et al., 2006). Thus, Slimb is constantly recruited to the posterior, consistent with the hypothesis that it recognizes a substrate that is constantly produced at the posterior.

Slimb Regulates Excessive Osk Protein Levels via Protein Ubiquitylation

As the posterior localization of Slimb requires its substrate-interacting domain and Short Osk, Osk is a good candidate to be a substrate for polyubiquitylation by the SCF complex. We therefore analyzed Osk protein levels in ovary extracts from combinations of slimb hypomorphic mutants rescued to adulthood by larval expression of hs-Slimb (Figure 3A). Short Osk is strongly overexpressed in slimb/slimb and slimb/slimb mutant ovaries, and this is largely rescued by germline expression of GFP-Slimb. In contrast to Osk, the levels of Vasa do not change under any of these conditions. Thus, Slimb appears to specifically target Short Osk for degradation.

Figure 3

Slimb Regulates Osk Protein Levels

(A) Osk is stabilized in slimb and sgg mutants. Western blots of ovary extracts from the indicated genotypes were probed with α-Osk and α-tubulin/α-Vasa as loading controls. The levels of Short Osk are significantly increased in sgg germline clones and in slimb/slimb transheterozygotes that have been rescued to adulthood by hs-Slimb expression in the larva and dissected 3 to 4 days after the last period of Slimb induction. GFP-Slimb expression reduces Osk levels in slimb mutants (lanes 3–4).

(B–D) Osk-GFP expression in stage 9 oocytes from heterozygous control (B), slimb/slimb (C), and sgg germline clones (D).

(E) Osk localization in an oocyte in which osk mRNA has been overexpressed using the mat-α4-GAL4:VP16 to drive UAS-osk. osk mRNA overexpression disrupts oocyte polarity, leading to the ectopic Osk protein in the oocyte center, but does not strongly increase Osk protein levels.

(F and F’) Osk localization in a UAS-osk/mat-α4-GAL4:VP16;slimb/slimb oocyte. In the absence of Slimb, osk mRNA overexpression leads to the accumulation of Osk protein all over the oocyte cortex. (F’) is a z stack projection from the oocyte shown in (F) to show the uniform cortical localization of Osk protein.

See also Figure S2.

The observation that Osk levels are increased in slimb mutants raises the question of whether Slimb is involved in the spatial control of Osk accumulation or whether it simply acts to limit Osk levels. We therefore examined the distribution of Osk in slimb mutant oocytes, focusing on transheterozygous combinations of hypomorphic mutants that rarely disrupt oocyte polarity. The most frequent defect in these oocytes is the enlargement of the Osk, Stau, and Vasa staining at the posterior, coupled with a failure to form a tight posterior crescent (Figures 3B and 3C; Figures S2A, S2B, S2D, and S2E). Although this phenotype can result from an anchoring defect caused by a lack of Long Osk protein, this is not the case for slimb mutants as Long Osk is present at normal or even slightly elevated levels. Thus, this phenotype may reflect a failure to degrade Osk that is not at the posterior cortex or the production of so much Short Osk that it can no longer be efficiently anchored.

As an alternative approach to investigate where in the oocyte Slimb targets Osk for degradation, we analyzed the localization of GFP-Slimb-Δ-F-Box, which cannot rescue Slimb function but interacts with Slimb substrates and therefore reveals their distribution. GFP-Slimb-Δ-F-Box reproduces Osk localization in slimb mutants, forming a broad posterior crescent (Figure S2G). Furthermore, both Osk protein and GFP-Slimb-Δ-F-Box are found in the center of the slimb mutant oocytes that display polarity defects (Figures S2C and S2H). Osk translation is thought to be repressed until osk mRNA reaches the oocyte posterior, but this phenotype indicates that some mRNA escapes this repression and that the resulting mislocalized protein is normally targeted for degradation by Slimb. To further analyze the potential role of Slimb in removing excess Osk protein, we compared the effects of overexpressing Osk in wild-type and slimb oocytes. When Osk is overexpressed in wild-type oocytes, Osk forms an enlarged posterior crescent and is often found in an additional ectopic dot but is excluded from most of the oocyte cortex (Figure 3E; Zimyanin et al., 2007). By contrast, overexpression of Osk in slimb mutants results in its accumulation all over the oocyte cortex (Figures 3F and 3F’). Slimb therefore limits the levels of Osk protein at the posterior and also degrades mislocalized Osk, indicating that it targets Osk for degradation wherever it is made in the oocyte. We do not, however, detect any accumulation of Osk in the nurse cells in Osk-overexpressing egg chambers, suggesting that osk mRNA translation is fully repressed in the nurse cells or that any Osk protein is removed by an alternative degradation pathway.

The SCF-Slimb ubiquitin ligase catalyzes the linkage of a polyubiquitin chain to its substrates, thereby targeting them for degradation by the 26S proteasome. Poly- and monoubiquitylated proteins can be detected by staining with the FK2 antibody, which does not bind free ubiquitin (Fujimuro et al., 1994). FK2 staining produces a specific signal at the posterior of most wild-type oocytes, which is particularly clear at stage 10 of oogenesis (Figure S2K). This signal is strongly reduced in slimb mutant oocytes and completely absent in osk/Df mutants, indicating that Slimb and Osk protein are required for the accumulation of ubiquitylated proteins at the posterior (Figures S2L and S2M). Furthermore, mild Osk overexpression driven by nanos-GAL4:VP16 increases the posterior staining of ubiquitylated species, whereas strong overexpression driven by mat-α4-GAL4:VP16 induces FK2 staining in a dot in the middle of the oocyte that overlaps with ectopically localized Osk (Figures S2N and S2O). Finally, the FK2 signal at the posterior of the oocyte is not due to the accumulation of ubiquitylated pole plasm components, because it persists in the vas mutant, which abolishes pole plasm formation downstream of Osk (Figure S2P). These results are consistent with the idea that SCF-Slimb polyubiquitylates Osk protein to target it for degradation.

Shaggy Kinase Localizes to the Oocyte Posterior and Regulates Osk Protein Stability

Slimb recruits the SCF complex to its substrates by recognizing a phosphorylated destruction motif. In a number of cases, such as Armadillo/β-catenin and Cubitus interruptus, the recruitment of Slimb requires phosphorylation of this motif by Shaggy (Sgg), the Drosophila ortholog of Glycogen synthase kinase-3 (GSK-3) (Maniatis, 1999). We therefore tested whether Sgg is also required to regulate Osk protein levels. Western blots of ovary extracts from sgg germline clones show elevated levels of Osk protein, and Osk-GFP accumulates to higher levels at the posterior of sgg mutant oocytes (Figures 3A and 3D). Furthermore, a GFP protein trap insertion in sgg that reports the localization of endogenous Sgg accumulates in the oocyte during early oogenesis and localizes in a crescent at the posterior of the oocyte at stages 9–11 (Figures 4A–4C). As this localization mirrors that of Slimb, we next examined whether it has the same requirements. The posterior localization of Sgg is disrupted in stau and osk mutants but is maintained in a strong hypomorphic vasa allele, vas (Figures 4D–4F). Thus, like Slimb, Sgg localization requires Osk protein but is upstream of the formation of the pole plasm. This suggests that Sgg may phosphorylate Osk to target it for Slimb-mediated degradation.

Figure 4

Shaggy Localizes to the Posterior of the Oocyte

Localization of Sgg (GSK-3) is visualized using the sgg GFP protein trap line. Sgg accumulates in the oocyte from early oogenesis onward (A) and forms a crescent at the oocyte posterior from late stage 9 to stage 11 (B and C) but is absent in oocytes mutant for stau (D) or osk (E). Sgg localizes normally in vasa mutants (F).

Par-1 Phosphorylates Osk at Ser248, Priming It for Sgg/GSK-3 Phosphorylation on Ser244

To determine whether Sgg/GSK-3 can phosphorylate Osk directly, we performed in vitro kinase assays using recombinant rabbit GSK-3 and Osk but observed no phosphorylation of Osk. Given that Par-1 has been previously shown to phosphorylate Osk (Riechmann et al., 2002), we used Par-1 as a positive control to analyze Osk phosphorylation. We confirmed that Osk is phosphorylated by both recombinant MBP-Par-1 and GFP-Par-1 immunoprecipitated from ovaries (Figures 5A and 5B). Furthermore, no Osk phosphorylation was observed when we immunoprecipitated a kinase-dead version of GFP-Par-1, ruling out the possibility of phosphorylation by a copurifying kinase (Figure 5B). Surprisingly, Par-1 did not phosphorylate the expected region of Osk (292–606) (Riechmann et al., 2002) and instead phosphorylated a more N-terminal fragment spanning amino acids (aa) 217–294 (Figures 5A and 5C). We mapped this phosphorylation replacing each serine or threonine in this region with alanine by site-directed mutagenesis, except for Ser244 and Ser248, where the alanine substitutions were unstable in Escherichia coli and serine was replaced with glutamate instead. This analysis revealed that mutation of Ser248 alone was sufficient to abolish phosphorylation of this fragment, whereas none of the other mutations had any detectable effect (Figures 5D and 5E).

Figure 5

Par-1 Phosphorylates Osk on Serine 248

(A) In vitro kinase assay using purified recombinant MBP-Par-1 (amino acids 245–521) incubated with Osk protein fragments spanning the entire protein, expressed as MBP fusions.

(B) In vitro kinase assay using GFP-Par-1 (IP-Par-1) or GFP-Par-1 kinase dead (IP-KD) purified from ovarian extracts by immunoprecipitation with α-GFP and MBP-Osk217-294 as a substrate.

(C) In vitro kinase assays using purified MBP-Par-1 or immunoprecipitated GFP-Par-1 incubated with the MBP-Osk 217-294 and MBP-Osk292-606. Only Osk217-294 is phosphorylated, indicating that there are no Par-1 phosphorylation sites in the Osk C-terminal region.

(D) In vitro kinase assay with recombinant MBP-Par-1 and Osk217-294 containing point mutations in putative phosphorylation sites. All fragments show wild-type levels of phosphorylation, except the S444E, S248E mutant, which is not phosphorylated.

(E) In vitro kinase assays using MBP-Par-1 or IP-Par-1 incubated with Osk217-292 S244E and Osk217-292 S248E. The S248E mutation abolishes phosphorylation, identifying S248 as the Par-1 phosphorylation site. The lower panels in (A)–(E) show gels stained with Instant Blue before autoradiography to verify that each reaction contained equal amounts of substrate.

(F) Wild-type or par-1/par-1 stage 10 egg chambers stained with an antibody that recognizes Osk phosphorylated on S248 (p-Osk, green) and anti-Osk (red). The wild-type egg chambers received no treatment (−), were treated with λ phosphatase to remove phosphate groups (λ), or were treated with λ phosphatase + Na3VO4, which blocks phosphatase activity (λ+O). Note that Osk is found in the middle of par-1 mutant oocytes (arrow), but it is not detected by p-Osk (arrowhead).

(G) Stage 10 egg chambers from osk/Df(3R)p females carrying one copy of a wild-type genomic osk construct (osk) or a construct in which S248 is mutated to alanine (oskS248A) stained with α-OskS248-P or α-Osk. The phosphospecific antibody does not stain OskS248A protein, although it is highly expressed at the posterior of the oocyte, confirming the specificity of the antibody for Osk phosphorylated on S248.

To test whether Par-1 phosphorylates Osk on S248 in vivo, we generated a phosphospecific antibody against OskS248-P. This antibody detects a crescent at the posterior of the oocyte at stage 9 that corresponds to the site of Osk localization and appears to specifically recognize phosphorylated Osk, as treating the oocytes with λ phosphatase abolishes this localized signal (Figure 5F). To confirm the specificity of the antibody, we generated oocytes expressing a nonphosphorylatable form of Osk (OskS248A; see below) in an osk protein null mutant background. Although OskS248A protein localizes normally to the posterior of the oocyte, it is not detected by the anti- OskS248-P antibody, verifying that the antibody specifically recognizes Osk that is phosphorylated on S248 (Figure 5G). No staining for OskS248-P is observed at the site of Osk localization in the middle of par-1 mutant oocytes, indicating that this phosphorylation depends on Par-1 kinase (Figure 5F). Taken together, these results demonstrate that Par-1 phosphorylates Osk in vivo on S248.

The sequence surrounding the Par-1 phosphorylation site (SSGAPS) resembles the consensus site for phosphorylation by GSK-3 (T/SXXXSp/Tp), which requires a priming phosphorylation in the +4 position (Dajani et al., 2001; Fiol et al., 1987). We therefore performed in vitro kinase assays to ask whether GSK-3 could phosphorylate Osk after it had been phosphorylated by Par-1. Recombinant GSK-3 is unable to phosphorylate OskWT alone (Figure 6A, lane 2). However, when both Par-1 and GSK-3 are added to the reaction, OskWT is much more phosphorylated than by Par-1 alone (lane 3, compared to lane 1). This increase in Osk phosphorylation on adding GSK-3 is not observed with OskS244E, which is mutant for the predicted GSK-3 phosphorylation site, and this protein is phosphorylated by Par-1 and GSK-3 to the same extent as by Par-1 alone (lane 6, compared to lanes 1 and 4). In addition, OskS248E is not phosphorylated in any of the conditions tested (lanes 7–9), demonstrating that phosphorylation of S248 by Par-1 is required for the subsequent phosphorylation of S244 by GSK-3.

Figure 6

Par-1 Primes Osk for GSK-3 Phosphorylation and Slimb Recruitment

(A) In vitro kinase assays with MBP-Par-1 and GSK-3 in reactions containing wild-type Osk217-294 (OskWT), OskS244E, or OskS248E as substrates. MBP-Par-1 was added 25 min before the addition of GSK-3. OskWT is more highly phosphorylated when Par-1 and GSK-3 are added sequentially, but the S244 mutation blocks the enhancement upon GSK-3 addition, and the S248 mutation blocks all phosphorylation. This indicates that GSK-3 phosphorylates S244, but only if S248 is already phosphorylated.

(B) Egg chambers expressing wild-type or nonphosphorylatable forms of Osk in an Osk protein null mutant background were stained for Slimb (green) and Vasa (red) or Osk (red). Slimb is recruited to the posterior by OskWT and OskK261A, but not by Osk that cannot be phosphorylated by GSK-3 (OskS244A) or Par-1 (OskS248A), or the double mutant (OskS244A,S248A).

See also Movie S1.

Slimb Is Recruited to Doubly Phosphorylated Osk

The region of Osk that is phosphorylated by Par-1 and GSK-3 closely matches the consensus for the Slimb degron (DSpGφXSp - φ; hydrophobic, X; any amino acid), suggesting that sequential phosphorylation by these two kinases recruits Slimb to target Osk for degradation. To test this hypothesis, we mutated each or both phosphorylation sites to alanine within a full-length osk genomic fragment to produce nonphosphorylatable versions of Osk that are expressed from the endogenous promoter. These transgenes were integrated into the same genomic attB site using PhiC31 integrase-mediated recombination and crossed into an osk/Df background. Each transgene is therefore expressed at the same level and provides the only source of Osk in the oocyte. OskWT protein recruits Slimb to the oocyte posterior as expected, whereas OskS244A, OskS248A, and OskS244A,S248A do not (Figure 6B). Thus, both serines are required for Slimb recruitment, consistent with the idea that the sequential phosphorylation of Osk by Par-1 and GSK-3 generates a Slimb binding site. Osk and its downstream effector Vasa are still present at the posterior in osk, osk, and osk oocytes (Figure 6B; data not shown). However, three-dimensional reconstructions of stage 10 oocytes indicate that unphosphorylatable OskS248A is not as tightly localized as OskWT and spreads away from the posterior cortex (Movie S1).

Because the nonphosphorylatable osk mutants fail to recruit Slimb, one would expect Osk proteins to be stabilized in these mutants, and this is indeed the case: OskS244A (GSK-3 site mutant), OskS248A (Par-1 site mutant), and OskS244A,S248A accumulate to higher levels than OskWT in ovary extracts (Figure 7A). This is consistent with a model in which the recruitment of SCF-Slimb complex catalyzes the polyubiquitylation of Osk to target it for degradation by the proteosome. SCF-Slimb attaches polyubiquitin chains to lysine residues, which are usually located near the phosphodegron, and Osk contains lysines in appropriate positions (K236, K261). Mutation of either K236 or K261 had, however, no effect on Osk stability, suggesting that other lysines can be used as Ubiquitin acceptors (Figure 7A; data not shown).

Figure 7

Nonphosphorylatable Osk Is Stabilized, Leading to Increased Numbers of Pole Cells and Patterning Defects

(A) Western blot of ovary extracts from females expressing osk transgenes with point mutations in the GSK-3 (OskS244A), Par-1 (OskS248A), or both phosphorylation sites (OskS244A,S248A) in an osk/Df(3R)p background, probed for Osk protein. The Osk proteins with mutant phosphorylation sites accumulate to much higher levels than the wild-type protein (OskWT) expressed from the same insertion site.

(B) Expression of nonphosphorylatable Osk leads to an increase in pole cell number. The left panel shows the average number of pole cells ± SD per embryo in 12 embryos from osk/Df(3R)p, osk mothers and osk/Df(3R)p, osk mothers (p < 0.005, Student’s t test with a two-tailed distribution). The right panel shows projections of stage 11 embryos stained with Vasa to mark the pole cells.

(C) Nonphosphorylatable Osk accumulates in mago mutants. Western blot of ovary extracts from mago/Df(2R)F36 females carrying no transgene, with one copy of OskWT or with one copy of OskS244A,S248A probed with α-Osk. The membrane was reprobed with α-tubulin as a loading control.

(D) Quantification of the phenotypes at 25°C of embryos derived from osk/Df(3R)p, osk/Df(3R)p, wild-type, and mago/Df(2R)F36 females carrying the indicated copies of OskWT or OskS244A,S248A transgenes.

The nonphosphorylatable osk mutants provide a way of testing the in vivo significance of the regulation of Osk stability by Par-1, GSK-3, and SCF-Slimb that is not possible with mutations in any of these trans-acting factors, which all disrupt other processes in oogenesis (Muzzopappa and Wappner, 2005; Shulman et al., 2000; Tomancak et al., 2000). Both single phosphorylation site mutant transgenes (osk and osk) and the double mutant transgene (osk) gave similar phenotypes, consistent with the fact that they all block Slimb recruitment. Because Osk nucleates the pole plasm, which contains the posterior determinant, nanos mRNA, and the pole cell determinants, the number of pole cells provides a sensitive readout for the level of Osk activity (Ephrussi and Lehmann, 1992; Smith et al., 1992). Embryos from mothers expressing one copy of nonphosphorylatable Osk contain an average of 28 pole cells, compared to 24 in embryos from mothers expressing one copy of OskWT (p < 0.005) (Figure 7B). This small but significant increase is less than one might expect from the observed increase in Osk protein levels but may reflect the fact that much of the additional Osk in the nonphosphorylatable mutants is not tightly localized to the posterior.

We also assayed the ability of the transgenes to rescue the abdominal defects of the osk null mutant. One copy of osk rescued the arrest in oogenesis of osk mRNA null (osk/Df) flies. However, 10% of the resulting embryos were missing abdominal segments, displaying a weak posterior group phenotype (Figure 7D). By contrast, one copy of nonphosphorylatable Osk rescued abdomen formation completely but also produced a significant frequency of embryos with head defects, which are characteristic of Osk overexpression. Furthermore, 20% of the embryos laid by females with two copies of the osk transgene in an osk protein null background developed head defects, and 10% were symmetric bicaudals, whereas two copies of the wild-type transgene in the same background produced a much lower frequency of anterior defects. The effect of mutating the two phosphorylated serines in Osk is also apparent when the transgenes are expressed in the presence of wild-type osk. Nonphosphorylatable Osk produces three times as many embryos with head defects as OskWT, with 19 % bicaudal embryos. Thus, the nonphosphorylatable mutants frequently disrupt anterior development in both osk mutant and wild-type backgrounds, demonstrating that Slimb-dependent degradation of Osk is important for normal anterior-posterior axis formation.

The head defects produced by nonphosphorylatable Osk expression can be caused by too much Osk at the posterior but are most commonly due to mislocalized anterior Osk, which recruits ectopic Nanos to repress the translation of bicoid mRNA (Ephrussi and Lehmann, 1992; Wharton and Struhl, 1991). This phenotype and the recruitment of Slimb to the anterior of oocytes expressing osk-bcd3′UTR raise the possibility that degradation plays a key role in removing mislocalized Osk protein. To test this hypothesis directly, we crossed one copy of osk or osk into mago nashi (mago) mutant females, in which most osk mRNA is mislocalized to the anterior of the oocyte (Boswell et al., 1991). The addition of one extra copy of osk results in a small increase in the levels of Osk protein and partially rescues the abdominal phenotype of the resulting embryos (Figures 7C and 7D). The levels of Osk are, however, much higher in the presence of one copy of osk, and almost none of the embryos develop abdominal defects. Instead, nearly half the embryos have head defects, with 19% showing the more severe, symmetric bicaudal phenotype (Figures 7C and 7D). Taken together, these results indicate that Par-1/GSK-3/Slimb-dependent degradation plays a key role in repressing Osk accumulation at the anterior of the oocyte to prevent ectopic Nanos from disrupting head development.

Discussion

The results presented here reveal a mode of Osk regulation that appears to play two important roles in development. First, by reducing the amount of Osk protein that accumulates at the correct site at the posterior of the oocyte, SCF-Slimb-dependent degradation limits the amount of pole plasm produced and thereby restricts the number of pole cells (the primordial germ cells) that form in the early embryo. This activity therefore regulates the balance between the germline and the soma, a key decision point in development that is under strong selective pressure (Strome and Lehmann, 2007). Perhaps more importantly, Osk degradation also serves to remove Osk from inappropriate regions of the oocyte. osk mRNA is translationally repressed until it is localized, and this is thought to be the main mechanism that restricts Osk protein to the oocyte posterior (Kim-Ha et al., 1995; Markussen et al., 1995; Rongo et al., 1995). However, even unlocalized osk mRNA sediments in the polysomal fraction in sucrose gradients, suggesting that this mRNA is translated (Braat et al., 2004; Chekulaeva et al., 2006). Because even small amounts of ectopic Osk protein are sufficient to disrupt head development, the translation of unlocalized osk mRNA even at low levels means that there must be a mechanism to get rid of this ectopic protein (Braat et al., 2004). Our results demonstrate that unlocalized osk mRNA is indeed translated, as nonphosphorylatable Osk protein accumulates in mago oocytes, in which osk mRNA transport to the posterior is impaired. Furthermore, SCF-Slimb-dependent degradation provides the mechanism that removes this mislocalized Osk protein, as the translation of OskWT from the mRNA at the anterior of mago mutants only rarely disrupts head development, whereas Osk that cannot recruit Slimb causes a significant frequency of bicaudal embryos. Thus, the translational repression of unlocalized osk mRNA is leaky and is not sufficient on its own to restrict Osk protein to the posterior, and Slimb-dependent degradation provides an essential backup mechanism to dispose of small amounts of mislocalized Osk.

Slimb is recruited to Osk by a typical phosphodegron that must undergo two sequential phosphorylations to be recognized, much like other well-characterized Slimb/β-TRCP substrates, such as Armadillo/β-catenin, Cubitus Interruptus, and IκB (Fuchs et al., 2004; Smelkinson et al., 2007). Osk is targeted for degradation by Par-1 phosphorylation, which primes it for a second phosphorylation four amino acids downstream by GSK-3, to generate a strong Slimb binding site that recruits the SCF-Slimb complex. Our data show that only the pole plasm inducing isoform Short Osk is able to recruit Slimb. As Long Osk also contains the Slimb binding site, the folding of its additional N-terminal extension may block phosphorylation or the interaction with Slimb.

The role of Par-1 in Osk degradation is surprising, given a previous report that Par-1 phosphorylation stabilizes Osk (Riechmann et al., 2002). The reasons for this discrepancy are unclear, but there are several differences between the two analyses that may account for this contradiction. First, the in vitro phosphorylation experiments performed by Riechmann et al. (2002) used a truncated version of Osk (aa 292–606) that lacks serine 248, which is the only site that was strongly phosphorylated by Par-1 in our analysis of the entire protein. Second, they assayed the stability of truncated in vitro translated Osk in Drosophila protein extracts and Xenopus oocytes, which may not reflect the regulation that occurs in the normal context of the Drosophila oocyte. Although we cannot rule out the possibility that Par-1 phosphorylates other, lower efficiency sites in Osk, our data strongly suggest that the major effect of Par-1 phosphorylation on Osk protein is to target it for polyubiquitylation and degradation by the proteasome. The phosphospecific antibody against OskS248-P confirms that serine 248 is phosphorylated in vivo, and mutation of this site to alanine stabilizes the protein. Furthermore, when osk mRNA is overexpressed in par-1 mutants, the levels of Osk protein are hugely increased compared to osk mRNA overexpression in a wild-type background, and ectopic protein accumulates all over the oocyte (Zimyanin et al., 2007).

Par-1 plays at least two positive roles in Osk regulation. First, the recruitment of Par-1 to the posterior cortex of the oocyte in response to the polarizing signal from the posterior follicle cells induces the repolarization of the oocyte microtubule cytoskeleton that directs osk mRNA localization to the posterior pole (Doerflinger et al., 2006, 2010; Shulman et al., 2000; Tomancak et al., 2000). Because osk mRNA is released from translational repression at the posterior, its localization enhances the synthesis of Osk protein. Second, Osk protein recruits a distinct population of Par-1 to its site of accumulation, and this reinforces the polarity of the microtubule cytoskeleton to increase the localization of osk mRNA (Zimyanin et al., 2007). The results presented here show that Par-1 also negatively regulates Osk by targeting it for degradation. Because Osk protein is degraded by the Par-1/GSK-3/Slimb pathway wherever it is made within the oocyte and not just at the posterior pole, Osk is presumably phosphorylated by the second population of Par-1 that is recruited by Osk protein itself.

It seems paradoxical that Par-1 participates in both positive and negative feedback loops to regulate Osk levels, but one reason why this might be important is suggested by the fact that Par-1 positively regulates osk mRNA localization but negatively regulates the accumulation of Osk protein. osk mRNA can only be transported by kinesin to the posterior of the oocyte during stages 9–10A of oogenesis, when the microtubule cytoskeleton is polarized with the plus ends extending toward the posterior pole, as the microtubules are rearranged by the fast cytoplasmic streaming that begins at stage 10B (Dahlgaard et al., 2007; Manseau et al., 1996; Parton et al., 2011; Zimyanin et al., 2008). By contrast, Osk protein only accumulates to high levels during stages 13–14 of oogenesis, coincident with the decline of Slimb localization to the posterior (Snee et al., 2007). Thus, Par-1 seems to induce a temporal delay between the localization of osk mRNA and the accumulation of Short Osk protein. During stages 9–10A, Par-1 acts to polarize the microtubule cytoskeleton to promote the efficient localization of osk mRNA, while preventing the excessive accumulation of Short Osk. The mRNA is then stably anchored at the posterior by Long Osk, which does not recruit Slimb. At the end of oogenesis, the Par-1/GSK-3/Slimb pathway appears to be less active, and high levels of Short Osk accumulate so that the pole plasm can assemble just before it is needed to induce pole cell formation in the fertilized egg. These results highlight the central role that Par-1 plays in defining the posterior and regulating the assembly of the pole plasm, both in space and time.

Experimental Procedures

Drosophila Strains and Genetics

Drosophila strains and genetics are described in detail in the Supplemental Experimental Procedures.

sgg germline clones (GLC) were produced using the dominant female sterile technique (Chou and Perrimon, 1996). Although slimb mutations are recessive lethal, slimb mutant adults can be obtained by inducing Slimb expression under the control of a heat shock promoter in first-instar larvae until the end of pupariation (Grima et al., 2002). Transheterozygous slimb/slimb mutants were analyzed 3 to 4 days after the last period of hs-Slimb induction.

In Vitro Kinase Assays

In vitro kinase assays were performed as previously described (Benton et al., 2002). Recombinant rabbit GSK-3 was purchased from New England BioLabs and used at 1 μg per reaction. The full protocol is described in the Supplemental Experimental Procedures.

Immunological Methods

An antibody against phosphoserine 248 of Osk was raised in rabbits by injection of the phosphorylated peptide SDYSSGAP[pS]LENIPRAP and subsequent immunodepletion with the unmodified peptide SDYSSGAPSLENIPRAP and affinity purification with the phosphorylated peptide (21st Century Biochemicals). Immunofluorescence (IF) and western blotting were performed in accordance with standard protocols using antibodies described in the Supplemental Experimental Procedures. Actin was visualized with rhodamine-conjugated phalloidin (Invitrogen).