Par-1 regulates bicoid mRNA localisation by phosphorylating Exuperantia

The Ser/Thr kinase Par-1 is involved in several polarisation events in the Drosophila female germline. During early oogenesis, one of the 16 germline cells is selected to become the oocyte and the other 15 cells develop as nurse cells, providing the oocyte with mRNAs and proteins essential for its growth and patterning (for a review, see Riechmann and Ephrussi, 2001). The maintenance of oocyte cell fate is the earliest known function of Par-1 during oogenesis. In germline clones of a par-1 null allele, the oocyte fails to polarise during early oogenesis and reverts to a nurse cell fate, resulting in egg chambers with 16 nurse cells and no oocyte(Cox et al., 2001; Huynh et al., 2001). Analysis of hypomorphic par-1 alleles that allow oocyte development past early oogenesis revealed that par-1 is required for repolarisation of the oocyte during mid-oogenesis (Shulman et al., 2000; Tomancak et al.,2000). At this stage par-1 is required for polarisation of the oocyte microtubule network, and for localisation of oskarmRNA, which encodes the posterior determinant of the fly. In wild-type oocytes, microtubules nucleate from the anterior and lateral cortex, their plus ends directed towards the posterior pole of the oocyte(Cha et al., 2001), and oskar mRNA is localised in a kinesin-dependent process to the posterior pole (Brendza et al.,2000). By contrast, in par-1 mutants microtubules nucleate from the entire oocyte cortex, including the posterior pole,directing their plus ends towards the centre of the oocyte. In these mutants, oskar mRNA is mislocalised to the centre of the oocyte(Shulman et al., 2000; Tomancak et al., 2000). As oskar mRNA localisation is kinesin dependent, the mislocalisation of oskar mRNA is most likely the consequence of the aberrantly polarised microtubule network (Shulman et al.,2000). Finally, during the late stages of oogenesis, Par-1 accumulates at the posterior pole of the oocyte. Here, Par-1 guarantees the maintenance of posterior polarity by phosphorylating, and thus stabilising,Oskar protein (Riechmann et al.,2002).

Recently, it has been reported that par-1 is also involved in localisation of the anterior determinant bicoid mRNA(Benton et al., 2002). bicoid mRNA localisation occurs in two phases(St Johnston et al., 1989; Schnorrer et al., 2002): in the early phase, the mRNA is transported to the anterior margin of the oocyte,resulting in its distribution in a ring-shaped pattern. Anterior localisation of bicoid mRNA is microtubule dependent(Pokrywka and Stephenson,1991), and is most probably mediated by plus end-directed motors. It is unclear, however, how the bicoid mRNA localisation complex distinguishes between microtubules nucleating from the lateral cortex and those nucleating from the anterior cortex of the oocyte. It has been speculated that anterior-nucleating microtubules might be qualitatively different from lateral microtubules and that this difference might be perceived by the bicoid mRNA-transport complex(Cha et al., 2001). In the late phase, bicoid mRNA redistributes along the entire anterior cortex,resulting in a disc-shaped distribution of the mRNA. This localisation is dependent on the function of γTub37C and Dgrip75(Grip75 – FlyBase), suggesting that it requires the formation of an anterior microtubule organising centre (MTOC)(Schnorrer et al., 2002). In hypomorphic par-1 mutants, the early phase of bicoid mRNA localisation is disrupted, such that the mRNA is not restricted to the anterior of the oocyte, but rather is distributed down the lateral cortex towards the posterior pole (Benton et al.,2002).

Another gene essential for bicoid mRNA localisation is exu. In exu mutants, bicoid mRNA localisation is affected beginning from the early phase, when the mRNA fails to localise specifically at the anterior and is dispersed throughout the oocyte cytoplasm(Berleth et al., 1988; St Johnston et al., 1989). Exu protein contains no known domains, with the exception of a region with weak homology to an RNA-binding motif(Macdonald et al., 1991; Marcey et al., 1991). However,this domain is dispensable for the bicoid mRNA localisation function of Exu (Wang and Hazelrigg,1994). By electron microscopy, Exu is detected in large electron-dense structures in the nurse cells, the sponge bodies, where Exu is thought to associate with bicoid mRNA(Wilsch-Bräuninger et al.,1997). A series of RNA injection experiments has led to the hypothesis that, in the nurse cells, Exu promotes the recruitment of anterior-targeting factors to bicoid mRNA. This assembly of the bicoid mRNA localisation complex would render the mRNA competent for transport to the anterior margin of the oocyte during the early phase localisation (Cha et al.,2001).

We have identified Exu in a proteomic screen for direct targets of Par-1 kinase. We show that Exu is a phosphoprotein whose phosphorylation is dependent on Par-1. In the nurse cells, Exu and Par-1 colocalise in patches that also contain bicoid mRNA. We have determined Par-1 phosphorylation sites in Exu, and show that mutation of these sites abolishes the ability of Exu to mediate bicoid mRNA localisation during mid-oogenesis. These mutants, which specifically affect phosphorylation of Exu protein, allow us differentiate between two phases of exu-dependent bicoid mRNA localisation: an early phase during mid-oogenesis, which is strictly dependent on Exu phosphorylation; and a late phase at the end of oogenesis, in which the requirement for Exu phosphorylation is less stringent.

To identify Par-1 substrates, we screened the first release (DGC-1) of the Drosophila Unigene collection of Expressed Sequence Tags (EST)produced by the Berkeley Drosophila Genome Project (BDGP). Plasmid DNAs of the ESTs were prepared from bacterial glycerol stocks using a miniprep robot. Pools of eight ESTs were expressed in vitro and ³⁵S-methionine labelled, using the coupled transcription-translation system (Promega TNT system). Protein pools were split and incubated either with purified Par-1 kinase and kinase buffer [40 mM HEPES (pH 7.2), 3 mM MgCl₂, 5 mM EGTA, 1 mM ATP], or with buffer alone, for 2 hours at 37°C. Proteins were separated on polyacrylamide gels, in which SDS is only provided in the running buffer but not in the gel itself (Anderson gels). We used 15 ml 30% Acrylamide, 2.58 ml 1%bis-Acrylamide 7.5 ml 1.5 M Tris (pH 8.8), 4.74 ml H₂O, 150 μl 10% Ammonium Persulfate and 15 μl TEMED to prepare a 30 ml resolving gel;2.5 ml 30% Acrylamide, 2 ml 1% bis-Acrylamide 1.875 ml 1.5 M Tris (pH 6.8),8.625 ml H₂O, 150 μl 10% Ammonium Persulfate and 15 μl TEMED for a 15 ml-stacking gel; and 60 g Tris-base, 288 g Glycine, 10 g SDS to prepare 2 litres of 5 ×running buffer. After SDS-PAGE and autoradiography, pools containing Par-1 substrates were identified by scoring for proteins with altered electrophoretic mobility. Identification of the cDNAs of interest within a positive pool was achieved by translating each of the cDNAs separately, and running the eight translation products in parallel with the original pools. We confirmed the identified cDNAs as Par-1 substrates by performing the kinase assay with each cDNA individually. Among the 5849 screened cDNAs, 133 (2.2%) encode proteins that showed altered mobility after incubation with Par-1. Those were analysed using data provided by the BDGP and FlyBase. Based on the presence of certain protein domains, homologies to proteins from other species and functional data, seven groups of substrates were classified: Cytoskeletal proteins (11), proteins involved in different aspects of signal transduction (35), DNA-associated proteins (34),RNA-associated proteins (11), enzymes (9), novel proteins (25) and others(9).

Antibody staining of ovaries and embryos was performed as described(Tomancak et al., 2000). In situ hybridisation was performed as described by Wilkie et al.(Wilkie et al., 1999), and an RNA probe corresponding to full-length bicoid mRNA was generated by in vitro transcription using an Ambion Megascript kit. Double staining for bicoid mRNA and Exu GFP was performed as described by Vanzo and Ephrussi (Vanzo and Ephrussi,2002) using a mouse anti-GFP antibody (Roche) in a 1:200 dilution. For double staining for Exu-GFP and Par-1, the same antibody was used in combination with a rabbit anti-Par-1 antibody(Tomancak et al., 2000) at a 1:40 dilution. Anti-Bicoid (Kosman et al.,1998) antibody was used at a 1:600 dilution.

Preparation of ovarian extracts and western blotting was performed as described (Riechmann et al.,2002). An anti-Exu antibody from rabbit was used at a 1:10.000 dilution. For the blot shown in Fig. 1B, Exu was detected with Enhanced Chemiluminescence using an anti-rabbit antibody coupled to horseradish peroxidase. For the blot in Fig. 4E, a fluorochrome-coupled anti-rabbit antibody was used, and signal was detected using an Odyssey scanner (LI-CORE).

exu mutants were generated by mutating the original EST from the Unigene collection using the Quick change mutagenesis kit from Stratagene according to the manufacturer's protocol. Mutant cDNAs were confirmed by sequencing, and in vitro translated for kinase assays with Par-1. exutransgenes are based on a previously described P-element transformation vector designed for the expression of Exu-GFP(Wang and Hazelrigg, 1994). The vector contains the exu upstream regulatory sequences and the exu open reading frame (ORF) supplemented at its 3′ end with a sequence encoding GFP. To remove the GFP sequence and to introduce the mutations, we excised a NotI-BamHI fragment that includes the 3′ end of the ORF and the GFP sequence and replaced this fragment with a NotI-BamHI fragment bearing the mutation.

Following fly stocks were used: exu^XL and exu^VL (protein null alleles), par-1^W3(a null allele), and par-1^9A and par-1⁵⁷⁴ (hypomorphic alleles). All females used for staining were grown at 18°C.

To elucidate the pathways by which Par-1 establishes polarity during oogenesis, we modified an assay developed by Stukenberg et al.(Stukenberg et al., 1997) to identify proteins phosphorylated by Par-1 kinase. This assay is based on the fact that phosphorylation usually alters the mobility of proteins on SDS polyacrylamide gels. We screened the Drosophila genome for Par-1 substrates, by testing in vitro-translated and radiolabelled proteins for mobility shifts after incubation with purified Par-1 (see Materials and methods). Exu, an essential mediator of bicoid mRNA localisation(Berleth et al., 1988; St Johnston et al., 1989; Macdonald et al., 1991), was among the substrates identified in the screen. Migration of in vitro translated Exu protein changed dramatically after incubation with Par-1,suggesting phosphorylation of the protein(Fig. 1A, lanes 1,2). We confirmed by phosphatase treatment that the altered mobility was due to phosphorylation (Fig. 1A, lane 3). To test if Exu phosphorylation is also dependent on Par-1 in vivo, we compared the migration of Exu in ovarian extracts of females possessing different levels of Par-1 activity. Remarkably, overexpression of Par-1 causes hyperphosphorylation of Exu (Fig. 1B, lane 4), while the majority of Exu protein is unphosphorylated in par-1 mutants. In addition, the degree of Exu phosphorylation is dependent on the strength of the par-1 allele, as strong mutants show a stronger reduction in Exu phosphorylation than weak mutants(Fig. 1B, lanes 2 and 3). These data show that the phosphorylation of Exu is dependent on Par-1, suggesting that Par-1 phosphorylates Exu during oogenesis.

In wild type oocytes, two phases of bicoid mRNA localisation can be distinguished by in situ hybridisation(St Johnston et al., 1989; Schnorrer et al., 2002). First, bicoid mRNA forms a ring at the anterior margin of stage 9-10a oocytes (Fig. 2A,D). Next, at stage 10b, the ring of bicoid mRNA evolves into a disk, such that the mRNA is present along the entire anterior surface of the oocyte(Fig. 2G). As a first step towards the analysis of the role of Exu phosphorylation by Par-1, we re-examined the distribution of bicoid mRNA in exu and par-1 mutants. Consistent with previous reports(Berleth et al., 1988; St Johnston et al., 1989), we found that in exu null mutants(exu^XL/exu^VL)(Marcey et al., 1991) bicoid mRNA localisation is severely affected and the anterior ring does not form (Fig. 2B). Interestingly however, bicoid mRNA localisation is not completely abolished in exu mutants, as the mRNA still concentrates in the anterior region of stage 9 oocytes (Fig. 2B,E). In addition, residual amounts of bicoid mRNA are detected at the anterior cortex of disk stage oocytes(Fig. 2H). Residual anterior localisation of bicoid mRNA was also observed in to other exu alleles (exu^QR and exu^PJ), confirming that exu does not completely abolish bicoid mRNA localisation to the anterior of the oocyte.

As par-1 null mutants arrest oogenesis before bicoid mRNA is localised, we analysed bicoid mRNA distribution in a combination of par-1 alleles that retains some Par-1 activity(par-1^W3/par-1^9A)(Cox et al., 2001; Huynh et al., 2001). In these mutants, formation of the anterior bicoid ring is severely affected,as revealed by spreading of the mRNA along the lateral cortex of the oocyte(Benton et al., 2002)(Fig. 2C,F). However, this defect recovers during stage 10 and only trace amounts bicoid mRNA are mislocalised (Fig. 2I). In conclusion, both exu and par-1 mutants show strong bicoid mRNA localisation defects during the early phase of bicoid mRNA localisation. These defects, together with the finding that Par-1 phosphorylates Exu, raises the possibility that Par-1 acts via Exu phosphorylation in bicoid mRNA localisation.

To further investigate the connection between Exu and Par-1, we compared their distributions by detecting Exu as a GFP fusion protein and Par-1 with an antibody. Both proteins have previously been detected in the cytoplasm of the nurse cells (Wang and Hazelrigg,1994; Theurkauf and Hazelrigg,1998; Shulman et al.,2000; Tomancak et al.,2000). Strikingly, the proteins show a near-identical localisation in patches in the apical regions of the nurse cells(Fig. 3A). It has previously been shown that Exu-GFP colocalises with bicoid mRNA in the nurse cells (Cha et al., 2001). We therefore tested whether bicoid mRNA is also present in these patches in the nurse cells. bicoid mRNA is enriched in the apical patches,although a substantial amount of the mRNA is also detected outside of the patches (Fig. 3B). The extensive colocalisation of Par-1 and Exu-GFP in these patches suggests phosphorylation of Exu by Par-1 may occur at these sites. The presence of bicoid mRNA in the apical patches supports our hypothesis that Exu phosphorylation is involved in the localisation of bicoid mRNA.

To test the hypothesis that Par-1 regulates bicoid mRNA localisation by phosphorylating Exu, we searched for potential Par-1 phosphorylation sites in Exu protein. Within the C-terminal part of Exu, we identified two sites resembling 14-3-3 binding motifs, which are known targets of Par-1 phosphorylation (Benton et al.,2002) (Fig. 4A). We mutated all relevant serines within these sites, produced the mutant proteins by in vitro translation and analysed their phosphorylation state by SDS-PAGE analysis after incubation with Par-1 kinase. Remarkably, mutations in both sites cause an increased mobility of Exu protein, indicating reduced phosphorylation of the mutant proteins. Although mutations in site A cause only a modest increase in Exu mobility(Fig. 4B, A6) mutations in site B lead to a dramatic increase (Fig. 4B, B3). An even stronger reduction in Exu phosphorylation,resembling the situation after phosphatase treatment, is observed when all Serines within sites A and B are mutated simultaneously (compare B3 and A6B3 in Fig. 4D). By mutating single Serine residues in site A and B, we mapped two serines in site A (S438 and S440, see Fig. 4C, A3 and A4)and one serine in site B (S457, see Fig. 4D, B2), the mutation of which causes a significant reduction of Exu phosphorylation. A triple mutant, in which these three serines were simultaneously mutated shows a slightly greater degree of phosphorylation than a mutant in which all serines were mutated, indicating that the other serine residues are also subject to phosphorylation(Fig. 4D, A5B2 and A6B3).

After identifying Par-1 target sites in Exu in vitro, we generated transgenes to express the mutant Exu proteins in ovaries under the control of the exu promoter (Wang and Hazelrigg, 1994). The transgenes were crossed into an exuprotein null background to assess the phosphorylation state of the mutant Exu proteins by western blot analysis. As in the in vitro assay, a knockout of site A causes only a slight decrease in Exu phosphorylation(Fig. 4E, A6), while a knockout of site B leads to a dramatic reduction(Fig. 4E, B3). We observed the same strong decrease in phosphorylation when only S457 of site B was mutated as when both relevant serines in B were mutated(Fig. 4E, B2 and B3) confirming that this Serine is especially important for Exu phosphorylation. Most importantly, upon mutation of all serines in A and B(Exu^ΔA+B), the protein migrates as a single band, indicating absence of any phosphorylation (Fig. 4E, A6B3). Hence, the motifs mediating phosphorylation of Exu by Par-1 in vitro also mediate Exu phosphorylation in vivo. Together with the finding that Exu phosphorylation is dependent on Par-1 during oogenesis, these data strongly suggest that Par-1 phosphorylates Exu on sites A and B.

Different aspects of Exu function have been revealed by analysis of an Exu-GFP fusion protein, which, when expressed under the control of the endogenous promoter, rescues the exu phenotype(Wang and Hazelrigg, 1994). Exu-GFP can be detected at the anterior of the oocyte and accumulates at the posterior of stage 9-10 oocytes (Fig. 3D). Within the nurse cells, Exu-GFP forms particles that move through the ring canals into the oocyte(Theurkauf and Hazelrigg,1998). To examine whether one of these properties of Exu is phosphorylation dependent, we expressed a mutant Exu-GFP, in which all serines in site A and B are mutated in ovaries of exu null mutant females. First, we confirmed by western analysis that the mutant Exu-GFP is not phosphorylated (Fig. 4E, A6 B3-GFP). Next, we analysed the distribution of mutant Exu-GFP protein in egg chambers and found that neither its subcellular localisation in the oocyte nor in the nurse cells is affected (Fig. 3C,E). Consistent with the finding that the overall distribution of the protein is not altered, we could not detect any differences in the mobility of unphosphorylated and phosphorylated Exu-GFP particles (data not shown). Finally, we tested if Exu phosphorylation affects the ability of the protein to colocalise with bicoid mRNA in the nurse cells. Notably, bicoid mRNA is still enriched in the Exu-GFP containing apical patches in the nurse cells (Fig. 3C). In conclusion, our Exu-GFP analysis suggests that phosphorylation of Exu is neither required for its localisation, nor its mobility, nor its ability to colocalise with bicoid mRNA.

A Bicoid protein gradient controls the hierarchy of segmentation genes at beginning of embryogenesis (Driever,1993). The mislocalisation of bicoid mRNA in embryos from exu females disturbs this gradient, resulting in head defects(Schüpbach and Wieschaus,1986; Frohnhöfer and Nüsslein-Volhard, 1987). To test whether phosphorylation of Exu is required for anterior patterning of the embryo, we first attempted to rescue the head defects of exu null mutant embryos using the exu^ΔA+B transgene, which expresses unphosphorylated Exu. More than one-third of the embryos produced by exu^VL/exu^XL;exu^ΔA+B females develop head defects typical of exu mutants and fail to hatch (data not shown), indicating that the Bicoid protein gradient is affected when Exu is unphosphorylated. Next, we directly analysed the Bicoid gradient in embryos produced by exu^ΔA+B mothers, using an anti-Bicoid antibody. We could detect no Bicoid gradient in approximately one-third of the embryos (Fig. 5E). In the rest of the embryos, a gradient of Bicoid protein was detected, however, its extent appears reduced compared with the wild type(Fig. 5E). These results suggest that the head defects observed in one third of the embryos produced by exu^ΔA+B mothers are caused by the absence of a Bicoid gradient in the embryo at the beginning of embryogenesis. The fact that two third of the embryos hatch indicates that, in those embryos, the Bicoid gradient is sufficiently strong to support anterior patterning.

To investigate more precisely the role of Exu phosphorylation in formation of the anterior-posterior axis of the embryo, we made use of our allelic series of exu transgenes encoding Exu proteins with different phosphorylation levels. For this analysis, we chose those transgenic lines that express equivalent levels of Exu protein as revealed by western blot analysis (Fig. 4E). We first examined the segmentation of the embryos produced by exu^VL/exu^XL mothers expressing Exu proteins mutant in the different Par-1 phosphorylation sites, staining for the protein product of the pair-rule gene even skipped (eve). Mutations in either site A or B, as well as the triple mutation of serines 438, 440 and 457, cause a weak anterior shift of embryonic segments typical of embryos produced by females in which bicoid activity is slightly reduced(Frohnhöfer and Nüsslein-Volhard,1987) (Fig. 6B′-D′). We observed two types of patterning defects in embryos from exu^ΔA+B females:(1) embryos with a stronger anterior shift of Eve stripes than that observed in mutants in which Par-1 phosphorylation is merely reduced and (2) embryos that, in addition to an anterior extension of the first Eve stripe also show an enlargement of the first two stripes and a compression of the posterior stripes (Fig. 6E′). To determine whether there was a correlation between the reduction of Exu phosphorylation and the segmentation defects, we compared the extent of anterior shift of the first Eve stripe in the different mutants. In Table 1, the average extension of the first Eve stripe in the different phosphorylation defective backgrounds is indicated, and in Fig. 6those embryos with the maximal extension of the first Eve stripe are shown. Strikingly, the more Exu phosphorylation is reduced, the more the first Eve stripe extends anteriorly. Thus, there is a direct correlation between the degree of reduction of Exu phosphorylation and the position of the first Eve stripe. In summary, we observed two classes of phenotypes in the embryos of the phosphorylation defective mutants. First, in one third of the exuΔA+B mutants we found a severe phenotype in which the Bicoid protein gradient is abolished, resulting in strong segmentation defects and lethality. Second, in two thirds of the exuΔA+B mutants and in the mutants that affect only a subset of the Exu phosphorylation sites we observed a milder phenotype reflected by a reduction of the Bicoid gradient and an extension of the first Eve stripe.

To test if the embryonic patterning defects we observed could be explained by defects in bicoid mRNA localisation during oogenesis, we analysed bicoid mRNA distribution in egg chambers by in situ hybridisation. A transgene encoding wild-type Exu rescues all aspects of bicoid mRNA localisation in exu^XL/exu^VL oocytes(Fig. 6A). Strikingly, the transgene expressing Exu protein with a total knockout of phosphorylation,entirely fails to rescue the bicoid localisation defects of the exu mutant oocytes from stage 9 through stage 10b(Fig. 6E). These bicoid mRNA localisation defects are indistinguishable from exu-null mutants (Fig. 2E,H) and have been observed in all analysed oocytes(n=31). Exu proteins with intermediate levels of phosphorylation caused by mutations in site A or B are unable to rescue bicoid mRNA localisation at stage 9, but possess sufficient residual activity to rescue of bicoid mRNA localisation at stage 10(Fig. 6B,C). Thus, the capacity of Exu protein to localise bicoid mRNA correlates with the phosphorylation state of the protein. Therefore, we conclude that Exu phosphorylation is required for localisation of bicoid mRNA during mid-oogenesis. Furthermore, these results strongly suggest that the embryonic pattering defects in the phosphorylation defective mutants are the consequence of bicoid mRNA mislocalisation during oogenesis.

Although exu-null and the phosphorylation defective exu^ΔA+B mutants abolish bicoid mRNA localisation during mid-oogenesis, the two mutants differ in the strength of their embryonic patterning defects. Although we never observed Bicoid protein at the anterior of the embryos from exu null mutants, 73% of the progeny of the exu^ΔA+B mutants form a Bicoid protein gradient (Fig. 5). Though the extension of this gradient is reduced compared with the wild type,its existence in a large proportion of the embryos reveals that unphosphorylated Exu retains some activity. The fact that exu null and exu^ΔA+B mutants show identical phenotypes until stage 10b suggests that the residual activity of unphosphorylated Exu protein is required only during the later stages. We therefore compared bicoid mRNA localisation in the two mutants after mid-oogenesis. During late oogenesis, exu^ΔA+B oocytes start to accumulate bicoid mRNA at the anterior cortex(Fig. 7C), while in exu null mutants only traces of bicoid mRNA are anteriorly localised, and most of the mRNA is either uniformly distributed in the ooplasm or cortically localised (Fig. 7B). The partial recovery in the exu^ΔA+B mutants shows that bicoid mRNA localisation is less phosphorylation dependent during late oogenesis. We conclude that during mid-oogenesis, bicoid mRNA is localised by a mechanism that is strictly dependent on Exu phosphorylation,and that during late oogenesis the mode of bicoid mRNA localisation changes to a mechanism that is still dependent on Exu but less dependent on its phosphorylation by Par-1. The fact that the amount of bicoid mRNA that is localised during late oogenesis in the exu^ΔA+B mutants is sufficient for the formation of a Bicoid protein gradient in 73% of the embryos shows that mutants, in which localisation is abolished during the early stages of bicoid mRNA localisation can recover in late oogenesis. Therefore,the two mechanisms that mediate bicoid mRNA localisation during mid-and late oogenesis are redundant, and defects that occur during mid-oogenesis can be compensated in late oogenesis.

We have shown that during mid-oogenesis, Exu phosphorylation is required for anterior localisation of bicoid mRNA, and have identified Par-1 as the kinase that phosphorylates Exu. Thus, Par-1 regulates bicoidmRNA localisation by phosphorylating Exu. However, par-1 and exu null mutants differ in their bicoid mRNA localisation phenotypes: bicoid mRNA is spread along the lateral cortex of par-1 mutant oocytes (Fig. 2C,F), while very little bicoid mRNA is detected at the cortex of exu mutant oocytes (Fig. 2B,E). The interpretation of the bicoid mRNA localisation phenotype of the par-1 mutants is complicated by the fact that par-1 is not only required for localisation of bicoid mRNA,but is also involved in other processes that establish polarity within the oocyte. In particular, par-1 mutants show defects in polarisation of the microtubule network (Shulman et al.,2000) and, as bicoid mRNA localisation is microtubule dependent (Pokrywka and Stephenson,1991), this defect might well contribute to the localisation defect. We therefore investigated if the bicoid mRNA mislocalisation in par-1 mutants is exclusively due to reduced Exu phosphorylation,or if an Exu-independent process also contributes to the bicoid mRNA localisation defect. To this end, we generated an `activated' version of Exu independent of phosphorylation by Par-1, by replacing (phospho-)serine 457 by the negatively charged residue glutamic acid. Although replacement of serine 457 by alanine results in severe bicoid mRNA localisation defects at the ring stage (Fig. 6C), its replacement by glutamic acid fully rescues bicoid mRNA localisation in exu-null mutants (Fig. 8B and Fig. 2D). If reduced Exu phosphorylation were solely responsible for the bicoidmRNA localisation defect in par-1 mutant oocytes, then the activated Exu protein should rescue this defect in par-1 mutants. However, the Exu<sup>Ser457Glu</sup> transgene does not rescue the par-1 phenotype completely, and bicoid mRNA is aberrantly distributed along the lateral cortex of the oocyte and in the middle of the anterior cortex at stage 9 (the `ring' stage) (Fig. 8D). This indicates that Par-1 has Exu-independent functions which, when they fail in par-1 mutants, contribute to the mislocalisation of bicoid mRNA in these oocytes. The predominant cortical localisation of bicoid mRNA in par-1 oocytes expressing the activated Exu protein suggests that the cortical mislocalisation defect occurs upstream of exu function.

To test the genetic hierarchy of par-1 and exu, we recombined the two mutations and compared the distribution of bicoidmRNA in the double-mutant with that in the single exu and par-1 mutants. Strikingly, bicoid mRNA in par-1 exudouble-mutant oocytes is mainly cortical and resembles the distribution of bicoid mRNA in par-1 oocytes, confirming that the cortical localisation defect of par-1 occurs upstream of exu function(Fig. 8G). We conclude that par-1 has two functions in bicoid mRNA localisation: a first Exu-independent function, which – when it fails – causes cortical accumulation of bicoid mRNA; and a second function, which is to phosphorylate Exu and render the protein competent for anterior targeting of bicoid mRNA.

We have shown that Par-1 has two distinct functions in bicoid mRNA localisation. First, Par-1 is necessary for the release of bicoidmRNA from the oocyte cortex. This role of Par-1 is epistatic to its second function in regulating anterior localisation of bicoid mRNA by phosphorylating Exu. By generating mutants that abolish Exu phosphorylation,we could further distinguish two phases of Exu dependent bicoid mRNA localisation. An early phase, in which bicoid mRNA localisation is abolished when Exu is unphosphorylated and a late phase, in which the requirement for Exu phosphorylation is less stringent. Thus, our results show that bicoid mRNA localisation is a multi-step process, and that redundant mechanisms are used to ensure the anterior accumulation of bicoid mRNA.

Exu protein is an essential mediator of bicoid mRNA localisation. In this study we have shown that Par-1 kinase phosphorylates Exu, and that this phosphorylation is necessary for anterior localisation of bicoidmRNA during mid-oogenesis. We have also shown that Exu phosphorylation does not affect Exu localisation, its ability to form mobile particles, or its colocalisation with bicoid mRNA. How then might Par-1 phosphorylation enable Exu to mediate bicoid mRNA localisation? Experiments in which fluorescently labelled bicoid mRNA was microinjected into living egg chambers have revealed that Exu is required in the nurse cells for anterior localisation of bicoid mRNA within the oocyte. These experiments have led to a model whereby Exu associates in the nurse cells with bicoidmRNA and mediates the recruitment of additional nurse cell factors required for targeting of bicoid mRNA to the anterior of the oocyte(Cha et al., 2001). Our finding that mutation of Exu phosphorylation sites results in a phenotype that is,during mid-oogenesis, indistinguishable from that of exu-null mutants suggest that Exu phosphorylation is involved in the recruitment of these anterior-targeting factors in the nurse cells. Phosphorylation might increase the binding affinity of Exu for these nurse cell factors, promoting their association with bicoid mRNA. The colocalisation of Exu-GFP, Par-1 and bicoid mRNA in patches in the nurse cells suggests that this is where the bicoid RNP complexes assemble.

The consequences of exu and par-1 mutations on bicoid mRNA localisation are distinct. Although loss of exufunction results in diffuse bicoid mRNA distribution in the ooplasm,a reduction in par-1 function causes cortical localisation of the mRNA. We have generated an Exu protein that localises bicoid mRNA independent of phosphorylation by Par-1 and rescues exu mutants, but that is unable to rescue bicoid mRNA localisation in par-1mutants. Therefore, the cortical mislocalisation of bicoid mRNA in par-1 mutant oocytes is independent of Exu function. What might be the other function of Par-1 in localisation of bicoid mRNA? The fact that bicoid localisation requires the microtubule cytoskeleton(Pokrywka and Stephenson,1991), together with the report that oocyte microtubules are improperly polarised in par-1 mutants(Shulman et al., 2000),suggests that cortical localisation in the mutants is caused by a microtubule defect. It has been proposed that microtubules of different qualities may nucleate from different regions of the oocyte cortex(Cha et al., 2001). A simple explanation for the aberrant localisation of bicoid mRNA in par-1 oocytes would be that the subset of microtubules nucleating from the anterior corners of the oocyte and serving as tracks for anterior transport of bicoid mRNA are not restricted to the anterior corners,but spread along the cortex, resulting in the lateral cortical localisation of bicoid mRNA. However, this model is not supported by our genetic epistasis experiments, which indicate that the exu independent function of par-1 acts at a step upstream of exu in bicoid mRNA localisation. Therefore, we favour a different model, in which in wild-type oocytes bicoid mRNA first localises cortically preceding its targeted transport along microtubules. In this model, most of the bicoid mRNA entering the oocyte moves in a nonpolar fashion,either passively or by active transport, to the oocyte cortex. Only after this cortical localisation does the targeted transport of bicoid mRNA to the anterior corners of the oocyte commence. In par-1 mutants, the improperly organised microtubule cytoskeleton prevents release of the mRNA from the cortex to the (anterior-targeting) microtubules and the mRNA remains cortically localised. In exu mutants, the polarity of the microtubules is normal and bicoid mRNA is released from the cortex. However, its targeted transport to the anterior is impaired and the mRNA is diffusely distributed in the ooplasm.

The requirement for Exu phosphorylation in bicoid mRNA localisation decreases during the later stages of oogenesis. This is revealed by the partial recovery of bicoid mRNA localisation in exumutants that abolish phosphorylation. These mutants are indistinguishable from exu-null mutants through stage 10b of oogenesis, but during early embryogenesis two-thirds of the mutants localise enough bicoid mRNA at the anterior to support formation of a Bicoid protein. This indicates that the mechanism of bicoid mRNA localisation changes after stage 10b of oogenesis, from an early phase that is strictly dependent on Exu phosphorylation, to a late phase that is less dependent on phosphorylation. Stage 10b is the stage at which ooplasmic streaming commences, providing a possible mechanism for localisation of bicoid mRNA in mutants in which Exu phosphorylation cannot occur. Before stage 10b, anterior targeting of bicoid mRNA could be mediated solely by directed transport of bicoid mRNA complexes along microtubules, a process that is strictly dependent on Exu phosphorylation. After stage 10b, this directed transport might be complemented or replaced by a passive trapping mechanism, which has also been postulated for the localisation of oskar and nanosmRNAs during late oogenesis (Glotzer et al., 1997; Forrest and Gavis,2003). This mechanism relies on the movements generated by ooplasmic streaming, which could bring bicoid mRNA complexes into contact with the anterior cortex of the oocyte, where the mRNA could be trapped by localised anchoring molecules. This change in the mechanism of bicoid mRNA localisation would occur at the time of assembly of the anterior MTOC that is essential in the late phase of bicoid mRNA localisation (Schnorrer et al.,2002), suggesting that the MTOC might be involved in the trapping mechanism. Such a trapping mechanism would be differentially affected in Exu-null mutants and in mutants that specifically abolish Exu phosphorylation. It is possible that Exu provides bicoid mRNA not only with factors required for anterior targeting, but also with factors required for anchoring of bicoid mRNA. Unphosphorylated Exu might be inactive in recruiting the factors for anterior targeting, but be competent for binding of factors required for anchoring.

In summary, our data supplement the models for bicoid mRNA localisation previously presented by Cha et al.(Cha et al., 2001) and Schnorrer et al. (Schnorrer et al.,2002) in the following way: in the first phase of bicoidmRNA localisation, the mRNA is transported to the anterior corners of the oocyte, resulting in a ring-like distribution. This targeted transport requires the formation of RNP complexes that contain bicoid mRNA and specific anterior-targeting factors that allow the RNPs to identify those microtubules that nucleate from the anterior corners of the oocyte. Assembly of this complex takes place in the nurse cells and requires the phosphorylation of Exu by Par-1. Upon entry of the complex into the oocyte, a specific proportion of the RNP complexes encounter the microtubules that nucleate from the anterior corners, and these complexes are directly transported to their final destination. However, a large proportion of the complexes does not find these microtubules directly, and moves first to the oocyte cortex. The transfer of these cortically localised complexes to microtubules nucleating from the anterior corners solely requires a properly polarised microtubule network. Only at this stage can the nurse cell factors assembled on the mRNA act to transport the cortically localised complexes to the anterior corners of the oocyte. During the second phase of bicoidmRNA localisation, the ring-shaped distribution changes to a disc-shaped distribution and a MTOC forms at the anterior of the oocyte. The third phase of bicoid mRNA localisation begins after the onset of ooplasmic streaming. In this late phase, the mechanism of bicoid mRNA localisation changes from targeted transport to passive trapping, mediated by ooplasmic streaming, and the mRNA is anchored at the anterior margin. The generation of exu mutants that abolish phosphorylation allows us to distinguish between the early and the late mechanisms of bicoid mRNA localisation, as the two mechanisms differ in their sensitivity to Exu phosphorylation.

We have previously shown that Par-1 controls posterior patterning by phosphorylating Oskar (Riechmann et al.,2002). Here, we have shown that Par-1 regulates anterior patterning by phosphorylating Exu. Although Oskar is an intrinsically unstable protein whose stability is increased by Par-1 phosphorylation, Par-1 phosphorylation does not affect Exu stability(Fig. 1B, Fig. 3C,E; data not shown) but does affect its ability to mediate bicoid mRNA localisation. Thus,Par-1 uses at least two different mechanisms to generate polarity within the same cell. Interestingly, these two Par-1 substrates, Oskar and Exu, are unique to Diptera, showing that during evolution Par-1 gained fly-specific mediators of cell polarisation as substrates. Par-1 is therefore flexible in the mechanisms and in the targets by which it mediates cell polarisation. This is in striking contrast to the PDZ-containing proteins Par-3 and Par-6, which appear to establish polarity by the assembly of a conserved protein complex.

We thank T. Hazelrigg for her gift of the Exu-GFP transformation vector; P. Macdonald for the anti-Exu antibody; R. Rivera Pomar; C. Alonso and J. Reinitz for ant-Bicoid antibodies; S. Yoshida for the bicoid probe; S. Schmidt and V. Benes for preparing miniprep DNA; the Tübingen stock-keepers for fly stocks; and S. Roth for helpful comments on the manuscript. We are grateful to A.-M. Voie for generation of transgenic flies and P. Tomancak for the original idea for the in vitro screen. The project was supported by the DFG Schwerpunkt `Zellpolarität' (EP 37/1-1).