The actin-binding protein Lasp promotes Oskar accumulation at the posterior pole of the Drosophila embryo

In Drosophila melanogaster, maternally provided mRNAs and proteins are transported from the nurse cells into the growing oocyte and stored until they are required at later stages of development. oskar mRNA, which encodes the posterior determinant of the fly, is transported to the posterior pole of the oocyte, where it is localized from stage 8 of oogenesis onward(Ephrussi et al., 1991; Kim-Ha et al., 1991). oskar mRNA is exclusively translated after its localization(Kim-Ha et al., 1995; Markussen et al., 1995; Rongo et al., 1995), and thus locally induces the assembly of the pole-plasm, the germ-plasm of Drosophila, causing abdomen and germline formation in the embryo. Oskar protein is both necessary and sufficient for abdomen induction and germline formation (Ephrussi and Lehmann,1992). Mislocalization of oskar mRNA to the anterior, as occurs in BicD mutants (Ephrussi et al., 1991), or ectopic expression of Oskar at the anterior pole from an oskar-bicoid (osk-bcd) transgene(Ephrussi and Lehmann, 1992),causes strong embryonic patterning defects and, in the latter case, the formation of primordial germ cells at the anterior of the embryo. It is therefore essential that Oskar activity remains restricted to the posterior pole of the oocyte and embryo.

oskar mRNA and protein remain tightly associated with the posterior pole during oogenesis and early embryogenesis, in spite of the strong cytoplasmic streaming that occurs after stage 10B of oogenesis(Gutzeit and Koppa, 1982), and experiments with the actin-severing drugs Cytochalasin D and Latrunculin A suggest that oskar mRNA and protein are attached to the actin cytoskeleton (Lantz et al.,1999). In addition, and in contrast to Oskar, Bicoid protein diffuses along the anteroposterior axis after its synthesis and is detected down to 30% of the egg length, although its mRNA is restricted to the anterior pole in the embryo [100-80% egg length(Driever and Nusslein-Volhard,1988)], which suggests that mRNA localization is not sufficient for protein restriction. Posterior anchoring of Oskar requires the Oskar protein itself (Ephrussi et al.,1991; Kim-Ha et al.,1991), and is a dynamic process involving both actin-dependent and-independent processes (Babu et al.,2004; Tanaka and Nakamura,2008; Vanzo et al.,2007). We performed an extensive yeast two-hybrid screen with Short-Oskar (the isoform of Oskar that is required for pole cell formation) as bait, to search for potential anchoring proteins, and found that the Drosophila homolog of mammalian Lasp1(Grunewald and Butt, 2008)binds to Oskar and helps to anchor Oskar to actin at the posterior pole. Furthermore, we found that efficient pole cell formation in vivo requires a Lasp SH3 domain that is capable of interacting with Oskar.

A detailed description of constructs and transgenes is available on request.

The yeast two-hybrid screen was performed using a modified mating strategy(Fromont-Racine et al., 1997). In brief, frozen aliquots of the oligo-dT primed ovarian cDNA library(Grosshans et al., 1999)transformed into the yeast strain EGY48 were mated to RSY473 strains previously transformed with both Short-Oskar fused to NLSLexA at the C terminus and JK103 for 4 hours on YP-Gal plates to simultaneously induce prey protein expression. Cells were washed with medium lacking leucine, tryptophan,uracil and histidine, and plated onto S-Gal/Raf plates lacking leucine,tryptophan, uracil and histidine. Mating efficiencies varied between 2% and 10%. Diploid cells (3.8×10⁷) were screened, corresponding to 5.7×10⁶ primary yeast transformants, and 480 positive clones were analyzed. Table S1 shows a compilation of the screening results (see supplementary material). Eighteen colonies of type 3.30 were further analyzed for this report [3.30 corresponds to bases 2243 to 3240 of the sequence submitted to GenBank (AJ294538)]. Isolated prey plasmids (in pJG4-5) were retransformed into EGY48 and the specificity of the interaction tested by mating to independent bait proteins.

The insert of clone 3.30 was used as a probe to screen an oligo-dT primed ovarian cDNA library in lambda ZAPII kindly provided by A. Spradling and the Berkeley Drosophila Genome Project. Several of the longest cDNA clones were characterized by restriction mapping and end-sequencing. Two clones were chosen for sequencing. Templates were generated using the GPS-1 transposon system (NEB, Madison, WI, USA) and sequenced with fluorescent-labeled primers on an Arakis sequencer (EMBL).

S-Lasp was amplified from EST AT23571 with Oligo3_BglII(CGGCAGATCTATGAATAAAACCT) and Oligo2a_Not (GCGGCCGCTTATATAACC), and cloned into pCRIITopo (Invitrogen, Carlsbad, CA, USA). RNA extraction from ovaries and embryos was done as described previously(Hachet and Ephrussi, 2004). Expression of S-Lasp was confirmed by One-step RT-PCR (Invitrogen) using primers LASP859U (GAGAACACCAAAATCCAGTCCAAC) and LASP1552L(CTGTGGTGGCGGCATCTG). The accession numbers of the assembled cDNA sequences are AJ294538 (L-Lasp) and AT23571 (S-Lasp).

The thermoscript RT-PCR System (Invitrogen) was used to synthesize cDNA for quantitative RT-PCR for oskar (see Fig. S3 in the supplementary material). Q-PCR reactions were run on an ABI 7000 Real-time PCR system with primers HJ46F (AACAAATCTTGCACCGCTGGGC) and HJ47R (GACTTGGCGTGGTGAGGCCTGA). RpS49 was used for normalization (rp49F, GCTAAGCTGTCGCACAAA; rp49R,TCCGGTGGGCAGCATGTG).

The expression constructs for Gst-chicken Src, Gst-mouse NSrc and pSGT_HA_SAM68 were kindly provided by G. Superti-Furga (Ce-M-M, Vienna,Austria) (Lock et al., 1996). Gst-fusion proteins were expressed and purified as described, except that protein expression was induced at 30°C(Breitwieser et al., 1996). HA-tagged Short-Oskar and SAM68 were translated in vitro from pβTHShortHA and pSGT_HA_SAM68 using the TNT-coupled transcription translation kit,according to the manufacturer's instructions (Promega, Madison, WI, USA).

For Gst pull-down experiments, 10 μg of the appropriate Gst-fusion protein were incubated with 15 μl pre-equilibrated glutathione-Sepharose beads (Amersham-Pharmacia, Piscataway, NJ, USA) in 500 μl BP [50 mM Tris/HCl (pH 7.5), 150 mM KCl, 1% NP40, 5 mM EGTA, 5 mM EDTA and 1×EDTA free protease inhibitor cocktail (Roche, Basel, Switzerland)] for 30 minutes at 4°C on a rotating arm. The beads were then washed three times with 1 ml BP and incubated with 5 μl of an in vitro translation reaction in a total volume of 500 μl BP for 1 hour at 4°C. The beads were then washed three times with 1 ml BP and resuspended in 20 μl 2×concentrated SDS sample buffer. Samples were then analyzed on a 12% SDS polyacrylamide gel, dried,exposed to Biomax film (Kodak) and quantified on a Molecular Dynamics phosphorimager.

For the Gst pull-down experiments shown in Fig. S1 (see supplementary material), 10 μg of the appropriate Gst-fusion protein were incubated for 50 minutes with ovary extracts (50 ovary pairs per sample) at 4°C and processed as above. Samples were analyzed on a 10% SDS polyacrylamide gel and processed for western blotting.

The GN protein was expressed as described above, except that 1 mM DTT was added to the dialysis buffer. Actin co-sedimentation assays were performed following the protocol supplied by the manufacturer (Cytoskeleton, Denver, CO)with several modifications: To pre-clear, the protein solutions were centrifuged for 1 hour at 436,000 g in a TLA 100 rotor in a Beckmann TL100 ultracentrifuge. G-actin (36 μg) was polymerized in the presence of 30 μg GN or control proteins for 30 minutes at room temperature. The samples were then centrifuged at 279,000 gunder the same conditions as above, for 30 minutes. Ten percent of the resulting pellet was then loaded onto a 10% SDS polyacrylamide gel. All assays were run in duplicate.

To generate an anti-Lasp antibody, the insert of two-hybrid clone 3.30 was cloned as a EcoRI(blunt)/XhoI fragment into the SmaI and SalI sites of pQE31 (Qiagen, Valencia, CA, USA) and expressed as a His₆-tagged fusion protein in E. colistrain M15[pRep4] (Qiagen). Protein production was induced with IPTG at 1 mM for 4 hours at 37°C and the protein purified under denaturing conditions as indicated by the manufacturer, except that the protein was step-eluted in buffer C containing 250 mM imidazole. Rabbit (#6528 and #6529) injections and bleedings were carried out by Eurogentec (Belgium). Analogously, the N-terminal BglII/PstI fragment of pSP72_L_Lasp was subcloned, expressed and used for immunization (#7271). To generate the anti-Oskar antibody, the EcoRV/SphI fragment of the Blue-osk cDNA (Ephrussi et al., 1991)was cloned into the BamHI(blunt)/SphI site of pQE32(Qiagen). The His₆-tagged fusion protein was expressed and purified as above, except that the culture was grown at 30°C. The antibody was produced in a rabbit (Anix) by EMBL Laboratory Animal Resources.

Whole-mount antibody staining and RNA in situ hybridizations were performed as described previously (Riechmann and Ephrussi, 2004; Riechmann et al., 2002; Tomancak et al.,2000). Anti-Lasp terminal-bleed serum (#6529), rat anti-Vasa and pre-absorbed rat 56 anti-Oskar (Ephrussi et al., 1991) antibodies were used at dilutions of 1:500-1:2000,1:1000 and 1:2000, respectively. Fluoresceine-conjugated goat anti-rabbit and rhodamine-conjugated goat anti-rat were used at a dilution of 1:500 (in PBT,Amersham). Samples were analyzed on a Zeiss Axiovert 200 or a Leica TCS SP2 confocal microscope and quantified using ImageJ.

The preparation of ovarian extracts and western blotting were performed as described previously (Markussen et al.,1995). Rabbit anti-Oskar and anti-Lasp antibodies (#6528 and#7271) were used at dilutions of 1:2000 and 1:4000, respectively. An antiα-tubulin monoclonal antibody DM 1A (1:2000, Sigma, St Louis, MO, USA)was used for standardization.

Directional hobo-mediated deletions were generated by remobilizing the Hobo element of the P{wHy}Lasp^DG14505 insertion (kindly provided by W. Gelbart, Harvard University, USA) as described(Huet et al., 2002). After selecting for y^- or w^- chromosomes, 44 deletion events were analyzed for loss of the appropriate P-element flanking sequence, by single fly genomic PCR using primers for pendout(CGACGGGACCACCTTATGTT) and the insertion flanking lasp (5′primer, ATAGCGAGTCGTACCATTACCATC; and 3′ primer,GGAAGACCACAAAGCCAATTTATA). The extent of the Hobo-mediated deletion was then mapped by inverse PCR as described (Huet et al., 2002). Two y^-/w⁺deletions were further characterized. lasp^y45 and lasp^y41 delete 2859 bp and 3514 bp with respect to the original insertion point of P{wHy}Lasp^DG14505 (nt 16675541 3L; GB accession number AE014296, release 5.7). RT-PCR (not shown) and western analysis demonstrated that neither lasp mRNA is present, and that neither the C terminus (Fig. 4D; anti-Lasp Cterm antibody #6528) nor the N terminus (anti-Lasp Nterm #7271; not shown) of Lasp is expressed. lasp^y45 and lasp^y41 are thus molecular null alleles.

Transgenic fly lines were generated by P element-mediated transformation of a w¹¹¹⁸ stock and then balanced(Rubin and Spradling, 1982). Several independent lines were assayed for each construct and results with typical lines are shown in this study.

In a two-hybrid screen using Short-Oskar as bait, we identified CG3849(Lasp), the Drosophila homolog of vertebrate Lasp1 (LIM and SH3 protein 1; see Fig. S1 in the supplementary material; Fig. 4A)(Chew et al., 1998; Grunewald and Butt, 2008; Schreiber et al., 1998), as being a strong interactor. Lasp consists of an N-terminal LIM domain, followed by two nebulin-like repeats (NEB; Fig. 4A), a spacer region and a C-terminal SH3 domain (for a review,see Grunewald and Butt, 2008). NEB repeats were first identified in human nebulin, an actin-binding,structural muscle protein (Labeit and Kolmerer, 1995). SH3 domains are evolutionarily conserved protein-protein interaction domains of about 60 amino acids(Musacchio et al., 1992a),which are frequently present in cytoskeletal proteins and proteins involved in signal transduction. Within the LIM, NEB and SH3 domains, DrosophilaLasp is highly similar to vertebrate and Caenorhabditis elegans Lasp,whereas the spacer between the NEB repeats and the SH3 domain is not conserved, and its length varies greatly between different species (see Fig. S1E in the supplementary material).

More detailed analysis showed that both the Short- and Long-Oskar isoforms interact with Lasp and that the interaction is specific for the SH3 domain of Lasp (see Fig. S1A-C in the supplementary material). Furthermore, GST fused to the original two-hybrid fragment of Lasp (3.30) or to the SH3 domain only were able to precipitate [³⁵S]-labeled, in vitro-translated Short-Oskar(Fig. 1A). However, as quantified by phosphorimager analysis, a mutant Lasp SH3 domain, in which a tryptophan that is strictly required for physiological SH3 interactions(Musacchio et al., 1992b; Musacchio et al., 1994) was changed to alanine (Gst_W→A), was at least 15-fold less efficient in precipitating Short-Oskar, confirming that the Lasp-Oskar interaction behaves as expected for a canonical SH3 domain interaction(Fig. 1A). Neither equal amounts (Fig. 1B) of Gst alone,nor Gst fused to the SH3 domain of Src (Gst_Src) or to the neuronal variant of the Src SH3 domain (Gst_NSrc) was capable of binding Short-Oskar(Fig. 1B). GST-pull downs of Oskar from ovary extracts showed similar results (see Fig. S1D in the supplementary material). Thus, independent assay systems show that Oskar and Lasp interact specifically.

For the interaction of Lasp with Oskar to be physiologically relevant, the distribution of the two proteins should overlap in vivo. Consistent with in situ hybridization data (not shown), antibodies against the initial two-hybrid fragment of Lasp strongly stain region 2 and two or so cells at the anterior tip of the germarium (Fig. 2A). Double-labeling experiments showed that these cells do not express Hts, which suggests that they are cap cells rather than germline stem cells (data not shown) (Lin et al., 1994). During mid-oogenesis, Lasp is strongly expressed in nurse cells and is highly enriched in the oocyte (Fig. 2B). Lasp appears to be mildly enriched at the posterior of the oocyte and is also expressed in the pair of anterior and posterior polar follicle cells; it is also present on ring canals(Fig. 2B, data not shown). At stage 10 of oogenesis, Lasp is detected at the nurse cell borders, ring canals and around the cortex of the oocyte (Fig. 2C), and it tightly colocalizes with Oskar at the posterior pole of the oocyte [Fig. 2D-F; Fig. 5A; the inset in Fig. 2F shows a fluorescence profile of Oskar (red) and Lasp (green) peaking at similar positions across the oocyte cortex]. Similarly, co-detection of Lasp and Oskar in embryos at the preblastoderm stage shows that the proteins colocalize at the posterior pole (Fig. 3B-D). The antibody staining is specific, as no staining is detected in lasp mutants(Fig. 2A′, Fig. 3, Fig. 5A; see below for mutant generation). Thus, Oskar and Lasp are closely associated in both oocytes and embryos.

The sequence of Drosophila Lasp contains two NEB repeats, which are predicted to bind to actin (Chen et al., 1993). To test whether Drosophila Lasp binds F-actin in vitro, we expressed the N terminus of Lasp, including the LIM domain and both NEB repeats, as a Gst-fusion protein (GN) and performed co-sedimentation assays. After polymerization of purified G-actin in the presence of the GN fusion protein, GN co-sedimented with F-actin fibers(Fig. 1C). Almost no GN protein was found in the pellet in the absence of F-actin(Fig. 1C). None of BSA, Gst alone or Gst fused to the C terminus of Lasp (lacking the NEB repeats)co-sedimented with F-actin, whereas the actin-binding protein α-actinin co-sedimented quantitatively under the same conditions (data not shown). Consistent with the antibody stainings, functional (data not shown) GFP-tagged Lasp expressed under control of the germline specific nanos-Gal4VP16 driver (Rorth, 1998; Van Doren et al., 1998) showed a distribution very similar to that of F-actin(Fig. 1D-F).

In the early embryo, F-actin is highly concentrated in the cortical cytoplasm as punctate aggregates and in filamentous structures(von Dassow and Schubiger,1994). Similarly, Lasp was detected as a network throughout the embryo cytoplasm and around the cortical dense layer at the preblastoderm stage (Fig. 3A). At the syncytial blastoderm stage, Lasp accumulated apically along the invaginating plasma membrane (Fig. 3E,F), in a distribution resembling that of F-actin(Fig. 3G). At the blastoderm stage, both Lasp (Fig. 3H,I)and F-actin (Fig. 3J)accumulated apically and laterally in the epithelial cells. No Lasp staining was detected at either stage in lasp mutants(Fig. 3A′-D′,H′,I′). Taken together, the colocalization and binding data suggest that Lasp anchors Oskar to the actin cytoskeleton. Consistent with this hypothesis, both the cortical Lasp (see Fig. S2 in the supplementary material) and posterior Oskar localization(Babu et al., 2004) are lost upon treatment of the oocytes with the actin polymerization inhibitor Latrunculin A.

FlyBase predicts a long Lasp (L-Lasp) isoform of 660 amino acids (aa) and a short isoform of 504 aa (S-Lasp), encoded by alternatively spliced transcripts(Fig. 4A,B). The short isoform lacks a part of the spacer region encoded by the fifth exon. RT-PCR revealed that, although transcripts encoding each isoform are expressed in the ovary,only the long isoform-encoding transcript is detected in 0- to 2-hour embryos(Fig. 4C). Consistent with this, both Lasp isoforms are detected in the ovary, whereas only L-Lasp is detected in 0- to 2-hour embryos by western blotting(Fig. 4D).

To determine Lasp function during oogenesis and early embryogenesis, we generated lasp null alleles by remobilizing a hobotransposable element (Huet et al.,2002) inserted in the second intron of lasp(P{wHy}DG14505, kindly provided by the Gelbart laboratory, Harvard University, USA; Fig. 4B). In two of the mutants we generated, lasp^y41 and lasp^y45, a large region including exons 3, 4 and 5 of the lasp-coding region was deleted(Fig. 4B) and no mRNA could be detected in ovaries by RT-PCR (data not shown). Western blotting confirmed that neither S-Lasp nor L-Lasp was present in either the ovaries of lasp^y41 or lasp^y45 homozygous females,or the embryos laid by them (lasp mutant embryos; Fig. 4D).

lasp^y41 and lasp^y45 are semi-lethal alleles, and 40-50% of lasp mutant eggs fail to develop, mostly because they are unfertilized (as assessed by DAPI staining; data not shown). Hatching is restored by expressing transgenic wild-type L-Lasp, but not by expressing S-Lasp or a version of L-Lasp bearing the W→A mutation in its SH3 domain (e.g. driven by maternal tubulin-Gal4 or pCOG; data not shown), revealing that both the SH3 domain and the spacer region (fifth exon)are crucial for this Lasp function. The other 50-60% of embryos hatch and develop normally.

Homozygous mutant lasp females all contain normally developed ovaries. Oskar protein (Fig. 5A), oskar mRNA and Staufen protein, which is required for oskar mRNA localization and serves as an oskar mRNA reporter (data not shown) (St Johnston et al., 1991), are localized normally, and oskar mRNA levels appear to be normal, as evaluated by quantitative RT-PCR (data not shown). However, western blotting revealed a mild reduction in Oskar protein levels in the ovary and a strong reduction in 0- to 2-hour lasp mutant embryos compared with in wild-type controls (Fig. 5B). Accordingly, in situ hybridization and antibody staining of embryos derived from lasp mutant mothers showed reduced amounts of oskar RNA and protein at the posterior pole(Fig. 5C; quantification of fluorescent signals in Fig. 5D). By contrast, bicoid mRNA detected at the anterior pole was unchanged (Fig. 5C,D),demonstrating that the reduction was specific to oskar. As quantitative RT-PCR (see Fig. S3 in the supplementary material) revealed similar total levels of oskar mRNA in lasp mutant and wild-type embryos (0-2 hours), the reduced amount of localized oskarmRNA observed at the posterior pole of lasp mutant embryos did not result from mRNA degradation, but rather might result from its failure to remain localized at the posterior pole, a process that requires the Oskar protein itself (Ephrussi et al.,1991). In conclusion, lasp appears to be required specifically for the stable accumulation of wild-type levels of oskarmRNA and Oskar protein at the posterior pole of the embryo.

Oskar activity at the posterior pole is required for abdomen and germline formation in the embryo (Ephrussi et al.,1991; Lehmann and Nusslein-Volhard, 1986), with germline formation being more sensitive to reductions in Oskar dosage than is abdominal patterning and thus an ideal read-out for threshold Oskar levels(Lehmann and Nusslein-Volhard,1986). The progeny of heterozygous osk⁵⁴/+females and lasp^y45, osk⁵⁴/+ double heterozygotes develop no germline (grandchildless phenotype) at very low frequency (e.g. osk⁵⁴/+: 1% no ovary, 8% only one ovary; Fig. 6A). By contrast, oskar heterozygous females lacking both copies of lasp show a penetrant grandchildless phenotype, with 20% of their offspring displaying a complete absence of ovaries, and another 25% having only a single ovary(Fig. 6A). Western blot analysis showed that the amount of Oskar protein in lasp^y45,osk⁵⁴/lasp^y45 ovaries was reduced by 40-50% relative to osk⁵⁴/+ ovaries (normalized to Tubulin and Kinesin Heavy Chain, respectively; see Fig. S4 in the supplementary material), which is consistent with the increased penetrance of the grandchildless phenotype.

We further quantified the effect of the absence of lasp in germline formation by counting the number of germ cells formed in lasp wild-type and mutant embryos(Fig. 6B,C). Whereas wild-type embryos displayed an average of 33 germ cells, an average of only 26 cells was observed in lasp mutant embryos, reflecting the reduction in Oskar levels at the posterior pole. Again, we detected a strong genetic interaction between oskar and lasp: whereas osk⁵⁴/+embryos displayed an average of 23 germ cells, lasp^y45,osk⁵⁴/lasp^y45 embryos formed only 12.

We conclude that oskar and lasp cooperate to ensure that Oskar protein accumulates at the posterior pole of the embryo at levels that are sufficient for germline formation (see Discussion).

To test whether the SH3 domain of Lasp is required for Oskar accumulation at the posterior pole of the embryo, we performed rescue experiments using either a wild-type Lasp transgene or one bearing the W→A mutation that disrupts the Lasp-Oskar interaction in vitro(Fig. 1; see also Fig. S1 in the supplementary material). The grandchildless phenotype of lasp^y45, osk⁵⁴/lasp^y41females was rescued almost fully by a lasp transgene(Fig. 7A), thereby also confirming that absence of lasp was responsible for the phenotype. By contrast, no rescue of the grandchildless phenotype was observed with the L-Lasp (W→A) transgene, indicating that the SH3 domain that is crucial for the Lasp-Osk interaction in vitro is also crucial in vivo. The S-Lasp transgene also failed to rescue the phenotype. The number of germ cells detected in the progeny of lasp^y45,osk⁵⁴/lasp^y41 females expressing the different lasp transgenes (Fig. 7B,C) paralleled the penetrance of the grandchildless phenotype(Fig. 7A). These results indicate that the SH3 domain is essential for Lasp function during germline formation and that the Lasp-Oskar interaction is relevant in vivo.

It has been shown that wild-type Oskar protein produced from an oskar-K103′UTR transgene accumulates around the entire oocyte cortex (Riechmann et al.,2002; Vanzo and Ephrussi,2002). Consequently, the embryos develop anterior patterning defects or even a bicaudal phenotype, in which anterior structures (head and thorax) are replaced by a duplicated abdomen of reversed polarity(Fig. 8A). To test whether Lasp is required for this ectopic Oskar-induced phenotype, we expressed the osk-K103′UTR transgene in laspmutant females. Indeed, whereas wild-type females expressing the osk-K103′UTR transgene produced embryos that displayed the full range of Oskar-related anterior patterning phenotypes with high penetrance, embryos of lasp mutant osk-K103′UTR females showed anterior patterning defects with both significantly reduced penetrance and severity(Fig. 8A,B).

Similarly, an oskar-bicoid (osk-bcd) 3′UTR transgene (Ephrussi and Lehmann,1992) caused the formation of ectopic pole cells at the anterior in 43% of embryos, whereas in a lasp mutant background, pole cells formed at the anterior in only 10% of embryos(Fig. 8C). Furthermore, only 20% of the embryos of lasp^y41/+ females developed pole cells at the anterior pole, revealing an effect of lasp gene dosage on Oskar accumulation. All fertilized eggs produced by osk-bcd-expressing lasp^y41 females developed into embryos displaying the bicaudal phenotype, confirming that abdominal patterning is less sensitive than germ cell formation to the reduction of Oskar levels in the embryo. Taken together, these results demonstrate that Lasp is necessary to achieve the high levels of Oskar activity required for pole cell formation in the embryo.

We have shown that the Drosophila homolog of mammalian Lasp1 interacts via its SH3 domain with Oskar, binds directly to F-actin in vitro,and colocalizes with F-actin in vivo. Western analysis of lasp mutant oocytes shows a slight reduction in Oskar levels that is not readily visualized by in situ hybridization. Posterior oskar mRNA maintenance in the ovary, which depends on Oskar protein, also appears to be relatively normal. However, in the embryos laid by lasp mutant mothers(lasp mutant embryos), oskar protein and mRNA at the posterior pole are strongly reduced. The overlap of Oskar, Lasp and actin distributions at the posterior pole during oogenesis and early embryogenesis(Figs 2, 3; Fig. 5A), as well as the actin-and Oskar-binding properties of Lasp (Fig. 1), suggest that Lasp connects Oskar to actin. This is further supported by the genetic interaction between oskar and lasp,and the dependence of ectopic Oskar activity on Lasp. We therefore propose that Lasp links Oskar to the actin cytoskeleton and helps to restrict Oskar activity to the posterior pole after its localized translation. Even though we cannot exclude that Lasp also protects Oskar from degradation, the reduced amounts of oskar mRNA and Oskar protein observed in laspmutants are probably due to the interdependence of oskar mRNA localization and Oskar protein. Less well-anchored Oskar cannot keep its mRNA restricted to the posterior pole, which in turn leads to less mRNA available for localized translation.

Lasp concentrates in actin-rich subcellular regions, including focal adhesions and lamellipodia in migrating cells(Butt et al., 2003; Chew et al., 2002; Schreiber et al., 1998; Terasaki et al., 2004), and motile growth cones of cultured neurons(Phillips et al., 2004). In our study, we have shown that the NEB repeats of Drosophila Lasp bind actin in vitro and that Lasp colocalizes with actin in vivo. laspmutant egg-chambers display no obvious actin defects, consistent with the observation that rabbit Lasp1 has no detectable effect on actin polymerization and that cells depleted of Lasp1 can form focal adhesions(Chew et al., 2002; Chew et al., 2000). Lasp1 has also been shown to be a substrate of protein kinase A (PKA)(Chew et al., 1998; Schreiber et al., 1998). After PKA stimulation of parietal cells in isolated gastric glands, Lasp1 redistributes from cortical regions to the actin-rich intracellular canalicular region (Chew et al.,2000). In Drosophila, egg chambers lacking PKA catalytic subunit activity contain multinucleate nurse cells and abnormal ring canals. In addition, pka mutant oocytes fail to respond to a signal from posterior follicle cells and consequently fail to reorganize the microtubule cytoskeleton (Lane and Kalderon,1993; Lane and Kalderon,1994). Oocytes lacking Lasp function show neither of these phenotypes. By contrast, excess PKA activity, as in Pka-RI mutant oocytes, causes the accumulation of high levels of ectopic Oskar protein and embryos displaying the bicaudal phenotype(Yoshida et al., 2004). This ectopic accumulation could be due to delocalized oskar mRNA translation, or to the stabilization of Oskar protein produced ectopically that would normally be targeted for degradation. Given the reduced Oskar levels we observed in lasp mutant embryos and the dependence of ectopic Oskar activity on Lasp (Figs 7, 8), it would be interesting to determine whether PKA phosphorylation of Lasp contributes either to the accumulation of Oskar or to its anchoring in the embryo.

Oskar anchoring becomes crucial at stage 10B of oogenesis, when cytoplasmic streaming starts, a vigorous process that ensures the mixing and even distribution of cytoplasmic mRNAs and proteins in the egg(Gutzeit and Koppa, 1982). Treatment of embryos with the actin-severing drugs Cytochalasin D and Latrunculin A disrupts the localization of oskar mRNA and protein,but not of bicoid mRNA (Lantz et al., 1999), which is similar to what we observed in laspmutant embryos and consistent with a role of Lasp in actin-dependent Oskar anchoring. Several mechanisms contribute to oskar RNA and protein anchoring. Bifocal (Bif) and Homer (Hom) redundantly promote Oskar anchoring at the oocyte posterior pole via an actin-dependent and an unknown,actin-independent process, respectively(Babu et al., 2004). Remarkably, the Oskar detachment defects of Latrunculin A-treated homegg chambers are stronger than in hom/bif double mutants(Babu et al., 2004), suggesting there are additional actin-dependent attachment mechanisms. However, we were unable to detect genetic interactions between lasp and hom/bif, as neither hom/lasp nor bif/lasp double mutants showed an Oskar localization defect in ovaries (data not shown).

The ubiquitous expression of Oskar in the oocyte leads to its enrichment all around the cortex, indicating that Oskar protein can be recruited all around the F-actin-rich oocyte cortex (data not shown)(Riechmann et al., 2002). As Lasp is uniformly distributed along the actin-rich cortex, it could participate in Oskar anchoring at ectopic cortical locations. Consistent with this, the deleterious effects of ectopic anterior (osk-bcd) or cortical (osk-K10) Oskar on embryonic patterning are suppressed in lasp mutant embryos. In further support of our anchoring model, only Lasp with a functional SH3 domain (i.e. one capable of interacting with Oskar in vitro and in the two-hybrid assay; see Fig. 1 and below; see also Fig. S1 in the supplementary material), is able to rescue the grandchildless phenotype and the reduction in pole cell number observed in osk^-, lasp^-/lasp^-embryos.

The interaction of Oskar with Lasp depends on its SH3 domain, because a single W→A point mutation on the binding surface of the SH3 domain abolishes its binding to Oskar (Fig. 1; see also Fig. S1 in the supplementary material). SH3 domains usually interact with short, proline-rich sequences with the consensus RxxPxxP(class I binding site) or PxxPxR (class II binding site)(Kay et al., 2000). The region of Oskar that is sufficient to interact with the SH3 domain of Lasp (aa 290-369; data not shown) does not encode a perfect class I or class II binding site, nor does it contain any of the less frequently occurring SH3-binding motifs (Kang et al., 2000; Mongiovi et al., 1999). Three PxxP motifs are present within the SH3-binding region of Oskar, but all of them lack the neighboring basic amino acid. Nevertheless, Oskar is highly specific for the SH3 domain of Lasp, as it does not interact with six other SH3 domains tested, and the interaction capability is conserved in Oskar of D. virilis (not shown).

Interestingly, the Lasp SH3 domain shares a high degree of similarity with several proteins involved in clathrin-mediated endocytosis: the actin-binding proteins Cortactin and Abp1, as well as Syndapin and Amphiphysin, both of which bind Dynamin and contain an F-actin- or an Arp2/3-interacting domain(Qualmann and Kessels, 2002). It has been reported that both Lasp and Dynamin II localize to the apical membrane of parietal cells, and that they interact in vitro(Okamoto et al., 2002). Thus,it is possible that Lasp is not only an actin binding-protein, but is also involved in vesicle trafficking. Endocytic trafficking and actin-based mechanisms also contribute to Oskar anchoring, and it is thus possible that Lasp, together with other functionally related proteins, represents a link between these processes. Furthermore, one of the two Oskar isoforms -Long-Oskar - plays a crucial role in the anchoring of both Oskar isoforms and oskar mRNA at the oocyte cortex at stage 10 of oogenesis(Vanzo and Ephrussi, 2002). Oocytes lacking Oskar have a dramatically reduced endocytic compartment and lack the thick actin bundles that are normally observed at the posterior pole. Thus Oskar anchoring in the oocyte appears to be a dynamic process, with Oskar-stimulated recycling endocytosis and filamentous actin outgrowths playing an important role in the maintenance of Oskar at the cell cortex(Tanaka and Nakamura, 2008; Vanzo et al., 2007). It is thus conceivable that Oskar, Lasp, and other possibly redundant actin-binding proteins act in feedback loops with components of the endocytic pathway to maintain Oskar at the posterior pole.