The identification of novel genes required for Drosophilaanteroposterior axis formation in a germline clone screen using GFP-Staufen

Drosophila oogenesis provides an excellent model system in which to study cell polarity and mRNA localisation, because the anteroposterior (AP)axis of the embryo is determined by the localisation of two different mRNAs to the anterior and posterior pole of the oocyte(van Eeden and St Johnston,1999; Riechmann and Ephrussi,2001). bicoid mRNA is localised at the anterior of the oocyte, and is translated after fertilisation to generate a morphogen gradient of a homeodomain transcription factor that patterns the head and the thorax of the embryo (Driever, 1993). oskar mRNA localises to the posterior of the oocyte, where Oskar protein directs the formation of the pole plasm that contains the pole cell and abdominal determinants (Ephrussi et al., 1991; Kim-Ha et al.,1991; Ephrussi and Lehmann,1992).

The localisation of bicoid and oskar mRNAs depends on reciprocal signalling between the oocyte and the overlying somatic follicle cells. In early stages of oogenesis, an MTOC is present at the posterior of the oocyte and nucleates microtubules that extend through the nurse cells(Theurkauf et al., 1992). During stages 5-6, Gurken protein signals from the oocyte to induce the adjacent somatic follicle cells to adopt a posterior fate(Gonzalez-Reyes et al., 1995; Roth et al., 1995). These posterior cells are then thought to signal back to the oocyte to induce the disassembly of the posterior MTOC, which leads to the formation of an AP gradient of microtubules, in which most minus ends lie at the anterior of the oocyte, with the plus ends extending towards the posterior pole(Theurkauf et al., 1992; Clark et al., 1994; Clark et al., 1997). This polarised microtubule cytoskeleton defines the destinations of bicoidand oskar mRNAs, and also directs the migration of the oocyte nucleus and gurken mRNA from the posterior of the oocyte to the anterior margin, where Gurken signals a second time to define the dorsal-ventral axis(Neuman-Silberberg and Schüpbach,1993).

As the localisation of bicoid and oskar mRNAs depends on the polarised microtubule cytoskeleton, a simple model is that these mRNAs are transported to the anterior and posterior by minus- and plus-end-directed microtubule motors, respectively (Clark et al., 1994; Pokrywka and Stephenson, 1995). In support of this model, oskar mRNA localisation requires the plus-end-directed motor, Kinesin 1(Brendza et al., 2000). However, it is unclear whether Kinesin directly transports oskar mRNA to the posterior, and if so, how it is coupled to its mRNA cargo(Glotzer et al., 1997; Cha et al., 2002; Palacios and St Johnston,2002). The actin cytoskeleton may also be important in this process, as some alleles of the actin-binding protein Tropomyosin II (TmII;Tp2 - FlyBase) disrupt oskar mRNA localization(Erdelyi et al., 1995). How bicoid mRNA is directed to the anterior of the oocyte is even less well understood, because the minus ends of microtubules are found along the lateral cortex as well as the anterior (Cha et al., 2001). Indeed, bicoid mRNA localises in a microtubule-dependent manner to both the anterior and lateral cortex when it is injected into the oocyte, and only localises specifically to the anterior if it has been exposed to nurse cell cytoplasm. This has led to the proposal that bicoid RNA is transported to the anterior along a specific subpopulation of microtubules, and that nurse cell factors render it competent to distinguish these microtubules from those nucleated laterally.

Very little is known about how the oocyte microtubule cytoskeleton becomes polarised, but three classes of mutants disrupt this organisation.

1. In mutants that abolish the follicle cell signal, such as gurkenmutants, the posterior MTOC persists after stage 7, leading to the formation of a symmetric microtubule cytoskeleton, which directs the localisation of bicoid RNA to both poles of the oocyte and oskar mRNA to the centre (Gonzalez-Reyes et al.,1995; Roth et al.,1995). Germline clones of PKA, mago nashi and Y14 (tsunagi) mutants cause similar phenotypes in the oocyte without affecting the follicle cells, suggesting that they are required for the transduction of the polarising signal in the oocyte(Lane and Kalderon, 1994; Micklem et al., 1997; Newmark et al., 1997; Hachet and Ephrussi, 2001; Mohr et al., 2001) However,Mago and Y14 are components of the exon-exon junction complex that marks where introns have been removed from mRNAs, and they presumably act on microtubules indirectly (Le Hir et al.,2001).

2. In par-1 and 14-3-3ϵ mutants, the posterior MTOC disassembles normally, but the microtubule plus ends and oskar mRNA become focussed in the centre of the oocyte(Shulman et al., 2000; Tomancak et al., 2000; Benton et al., 2002). These proteins may therefore play a role in recruiting plus ends to the posterior.

3. Mutants in the actin-binding proteins, Cappuccino, Spire and Profilin(Chickadee) disrupt oskar mRNA localisation by causing the formation of large cortical bundles of microtubules that drive premature cytoplasmic streaming (Theurkauf, 1994b; Manseau et al., 1996). This suggests that the actin cytoskeleton plays an important role in regulating the formation of the polarised microtubule array.

Genetic screens have identified a number of genes that play a role in the localisation of bicoid and/or oskar mRNAs once the microtubule network has been organised. Amongst these, staufen is unique, as it is required for both bicoid and oskar mRNA localisation. Staufen plays a central role in the regulation of oskarmRNA, as it is required not only for its transport from the anterior to the posterior of the oocyte, but also for the anchoring and translation of the RNA once it has reached the posterior pole(Ephrussi et al., 1991; Kim-Ha et al., 1991; St Johnston et al., 1991; Rongo et al., 1995; Micklem et al., 2000). Furthermore, Staufen co-localises with oskar mRNA throughout oogenesis, in both wild-type and all mutant conditions examined so far, and this localisation requires oskar mRNA(St Johnston et al., 1991; Ferrandon et al., 1994). Because Staufen is a dsRNA-binding protein, these results indicate that it binds directly and stably to oskar mRNA(St Johnston et al., 1992; Ramos et al., 2000). In freshly laid eggs, Staufen is also localised at the anterior, where it is required to anchor bicoid mRNA(St Johnston et al., 1989; Ferrandon et al., 1994). The anterior localisations of Staufen and bicoid mRNA are mutually-dependent, again suggesting that Staufen interacts directly with the RNA. However, it is unclear when Staufen first associates with bicoidmRNA, as it is not enriched at the anterior of the oocyte at stage 10a, and the later stages of oogenesis cannot be examined by antibody staining because of the impermeable vitelline membrane that is deposited around the oocyte.

Several other genes are necessary for the anterior localisation of bicoid mRNA at earlier stages. exuperantia is required for the anterior accumulation of bicoid RNA from stage 7 onwards, and appears to function in the nurse cells to render the RNA competent to localise in the oocyte (Berleth et al.,1988; Cha et al.,2001). In swallow mutants, the initial localisation of bicoid mRNA is normal, but it is not anchored at the anterior of the oocyte from stage 10b (Berleth et al.,1988; St Johnston et al.,1989). Swallow protein physically interacts with the Dynein light chain, and localises to the anterior of the oocyte, but it is unclear whether it plays a direct role in the anchoring of bicoid mRNA or functions indirectly in the organisation of the microtubule cytoskeleton(Schnorrer et al., 2000). In support of the latter view, Swallow interacts with γTub37C and Grip75,which are components of the γTubulin ring complex that nucleates microtubules; mutants in both of these genes also disrupt bicoid mRNA anchoring (Schnorrer et al.,2002).

The localisation of oskar mRNA is also affected by mutations in a number of other known genes. In barentsz mutants and hypomorphic alleles of mago nashi and Y14, oskar mRNA fails to localise to the posterior of the oocyte and accumulates along the anterior margin instead (Newmark and Boswell,1994; Micklem et al.,1997; Hachet and Ephrussi,2001). All three proteins transiently localise with oskarmRNA at the posterior of stage 9 oocytes, suggesting that they are directly involved in its transport (Newmark et al.,1997; Hachet and Ephrussi,2001; Mohr et al.,2001). Once oskar mRNA reaches the posterior of the oocyte, it needs to be securely anchored. The actin cytoskeleton may be important in this process as two actin-binding proteins, TmII and Dmoesin(Moesin-like - FlyBase), are necessary for the anchoring of oskarmRNA to the posterior cortex (Tetzlaff et al., 1996; Jankovics et al.,2002; Polesello et al.,2002). Oskar protein also plays an important role in anchoring, as oskar mRNA is not maintained at the posterior in oskarprotein null mutants (Ephrussi et al.,1991; Kim-Ha et al.,1991). Mutations in genes necessary for oskartranslation, such as aubergine, therefore result in similar defects in oskar mRNA anchoring (Harris and Macdonald, 2001). In addition, the homologue of the Xenopus cytoplasmic polyadenylation element-binding (CPEB) protein,ORB, binds oskar mRNA, and is required for its localisation and translation (Lantz et al.,1992; Christerson and McKearin, 1994; Hake and Richter, 1994; Chang et al.,1999).

The majority of the mutants described above were identified in genetic screens for maternal-effect lethal and female-sterile mutations(Nüsslein-Volhard et al.,1987; Schüpbach and Wieschaus, 1989; Schüpbach and Wieschaus,1991). Although these screens were very successful at identifying germline- specific factors required for oocyte polarisation and mRNA localisation, they could only recover homozygous viable mutations, and therefore missed many of the essential genes that also play a role in polarity and mRNA localisation in other cell types. This problem can be overcome by using the FLP/FRT/DFS system to perform screens in germline clones(Xu and Rubin, 1993; Chou and Perrimon, 1996). This system allows the recovery of lethal mutations in essential genes, because it selects for germline clones, while most somatic cells remain heterozygous. Perrimon et al. successfully used this technique to screen a sample of 500 lethal P-elements recombined onto FRT autosomes, and identified maternal-effect lethal factors involved in proper egg shell formation and patterning of the cuticle (Perrimon et al.,1996).

A limitation of all of these screens is their use of the embryonic cuticle as a readout for AP patterning, as this precludes the identification of any mutants that affect AP axis formation in the oocyte but block development before cuticle formation. To circumvent this problem, we have designed a novel genetic screen in germline clones for mutations that alter the distribution of GFP-Staufen in living oocytes. Here, we report the results of this screen on chromosome arm 3R, and the characterisation of 23 new complementation groups that play a role in AP axis formation.

The following fly stocks were used: w; P[ry⁺; hsp70:neo;FRT]82B, ry⁵⁰⁶ (Xu and Rubin,1993); w; P[ry⁺; hsp70:neo; FRT]82B, ry⁵⁰⁶P[w⁺; ovo^D1-18]3R1, P[w⁺;ovo^D1-18]3R2/TM3, Sb (Chou and Perrimon, 1996); y, w, P[w⁺; mat-tub-alpha4:GFP Staufen], [ry⁺; hs:FLP]; Pr, Dr/TM6B, Tb – this line contains a GFP-Staufen transgene driven from the α4 tubulin promoter and a FLPase transgene under control of the heat-shock promoter(Golic and Lindquist, 1989; Schuldt et al., 1998).

The following alleles were used for complementation tests: osk⁵⁴ (Lehmann and Nüsslein-Volhard, 1991), cnc⁰³⁹²¹(Guichet et al., 2001), spnE^hlsΔ157(Gillespie and Berg, 1995), orb^F343 (Lantz et al.,1994), orb^mel(Christerson and McKearin,1994), Ets97D^tne-4(Schulz et al., 1993), tmII^gs1 (Erdelyi et al., 1995), btz²(van Eeden et al., 2001). The yps^JM2 allele and the yps^JM2 orbrecombinant chromosomes were obtained from Dr Hazelrigg(Mansfield et al., 2002).

w; FRT82B males were starved for 6 hours before they were treated with 25 mM EMS (Sigma) in 1% sucrose for 18-24 hours to induce an average of one lethal hit per chromosome arm. The number of lethal hits was estimated by monitoring the number of X-linked lethals. Mutagenised males were mated with w; Pr, Dr/TM3 virgin females. Single w; FRT82B, */TM3 virgin females (where the asterisk indicates the mutagenised chromosome) were mated with y, w, hs-FLP, GFP-Staufen; FRT82B, ovo^D1/TM6B males. The progeny were heat-shocked three times for 2 hours in a 37°C incubator during the third larval instar and pupal stages. Ovaries from 3-5 females of the genotype y, w,hs-FLP, GFP-Staufen/w; FRT82B, */FRT82B, ovo^D1 were dissected and screened for defects in GFP-Staufen localisation, using an inverted fluorescence microscope (see Fig. 2). If a phenotype was observed, w; FRT82B, */TM6B males were mated with w; Pr, Dr/TM3 virgin females and balanced stocks were established. Mutants identified in the primary screen were re-screened under the confocal microscope.

Induction of the FLPase in females containing a non-mutagenised FRT chromosome produced germline clones in over 90% of the females. As we dissected 3-5 females for each mutant chromosome, the likelihood of getting clones in at least one female was higher than 99.9%. The mutagenised chromosomes for which no clones were recovered (about 25%) carried either a cell lethal mutation or a mutation arresting oogenesis at early stages, and were indistinguishable from the ovo^D1 egg chambers. These mutants were discarded.

Two lines were considered allelic if no trans-heterozygous progeny were recovered, or if the trans-heterozygous females were sterile or maternal-effect lethal. We did not routinely test for weaker maternal- effect defects, such as a grandchildless phenotype.

The primary screen was performed on a Leica inverted fluorescence microscope, using a broadband filter (Leica 13 513828) in which the GFP fluorescence appears green, and yolk autofluorescence yellow. For live observation, the ovaries were dissected in Voltalef 3S (screen) or 10S(movies) oil (Elf Atochem) on a coverslip. The time-lapse movies were obtained by collecting z series of five sections at 1 μm intervals every 30 seconds for 20 minutes on a BioRad confocal MRC1024 microscope. The moving particles were visualised by excitation with 568 nm light, and collection of the emission through an OG515 filter (Palacios and St Johnston, 2002). The Kalman images in Fig. 7E-H were obtained using the Kalman averaging function of the BioRad software to merge eight consecutive scans taken at 3 second intervals. Static particles form dots in these images, whereas moving particles appear as lines that represent the direction and speed of movement. Actin was visualised by fixing ovaries in 4%paraformaldehyde and staining with rhodamine-phalloidin (1:500; Molecular Probes). In situ hybridisations were carried out using RNA probes labelled with Digoxigenin-UTP (Roche). Immunohistochemical detection was performed with alkaline phosphatase-conjugated anti-DIG (1:5000; Roche). For microtubule staining, samples were fixed for 10 minutes in 8% paraformaldehyde and stained with a FITC-conjugated monoclonal anti-α-tubulin antibody (1:400; Sigma)(Theurkauf, 1994a).

To develop a marker for mRNA localisation in living oocytes, we analysed a fusion between green fluorescent protein (GFP) and Staufen that is specifically expressed in the female germline(Schuldt et al., 1998). The GFP-Staufen fusion is fully functional, as it rescues both the oskarand bicoid mRNA localisation defects of a staufen null mutant (data not shown). Like endogenous Staufen, GFP-Staufen accumulates in the oocyte during stages 2-8 (Fig. 1A), and co-localises with oskar mRNA at the posterior from stage 9 onwards (Fig. 1B,C). Furthermore, GFP-Staufen localisation parallels that of oskar mRNA in mutants in which oskar mRNA is mislocalised:it accumulates along the anterior margin in stage 9 barentsz mutant oocytes, and concentrates in the middle of the oocyte in grk mutants(Fig. 1D-F)(Gonzalez-Reyes et al., 1995; Roth et al., 1995; van Eeden et al., 2001). Because the fusion protein does not need to be stained, it can also be visualised during the later stages of oogenesis, when the oocyte is impermeable to antibodies. At stage 10b, GFP-Staufen accumulates at the anterior cortex of the oocyte, where it remains for the rest of oogenesis(Fig. 1C,G). This anterior localisation is abolished in exu or swa mutants, which disrupt the localisation of bicoid mRNA at an earlier stage, and is increased in females that overexpress bicoid mRNA(Fig. 1H and data not shown)(Berleth et al., 1988). Thus,Staufen associates with bicoid mRNA at the anterior of the oocyte from stage 10b, where it functions to anchor the RNA during the final stages of oogenesis.

Because GFP-Staufen co-localises with oskar and bicoidmRNAs at opposite poles of the oocyte, and at different stages of oogenesis,it provides an in vivo marker for the localisation of both transcripts, and hence the polarity of the AP axis. To identify new genes required for axis formation, we performed a genetic screen in living oocytes for mutants that affect the localisation of GFP-Staufen. We mutagenised flies carrying the FRT82B chromosome with ethyl methyl sulphonate (EMS), and used the FRT/FLP-ovo^D1 system to generate germline clones of chromosome arm 3R, using the crossing scheme outlined in Fig. 2. From the primary screen of 5023 independent mutagenised lines, we recovered 400 lines with defects in GFP-Staufen localisation. These were then re- screened in fixed samples, which were counterstained with rhodamine-phalloidin to reveal the morphology of the egg chambers. This identified 141 lines that showed reproducible and penetrant phenotypes. We performed in situ hybridisations for bicoid or oskar mRNAs on most of these lines to assay their localisation directly, and this correlated with the localisation of GFP-Staufen in all cases.

We first crossed the mutants to alleles of all of the known genes on 3R that affect oskar or bicoid mRNA localisation, and identified two new alleles of oskar, four of orb, six of spn-E, one of D-elg (Ets97D) and three of cnc, but no alleles of tmII or barentsz were identified (Ephrussi et al.,1991; Kim-Ha et al.,1991; Christerson and McKearin, 1994; Erdelyi et al., 1995; Gajewski and Schulz, 1995; Gillespie and Berg, 1995; Tetzlaff et al.,1996; Guichet et al.,2001; van Eeden et al.,2001). As expected, all of these mutants affect oskarmRNA localisation, and the three new alleles of cnc also affect the localisation of bicoid mRNA. This indicates that the screen was efficient, and achieved a reasonable degree of saturation.

We classified the remaining mutants on the basis of their phenotypes and used this as a guideline in the complementation tests(Table 1). As expected, we recovered mutants that affect only oskar mRNA localisation (74), only bicoid mRNA localisation (18), or the localisation of both mRNAs(40). To identify complementation groups, mutants within each phenotypic class were crossed to each other, and the progeny assayed for lethality,female-sterility and maternal-effect lethal phenotypes. In addition, mutants that affect both bicoid and oskar mRNAs were crossed to mutants that affect the localisation of only one mRNA. With this approach, we identified 23 novel complementation groups(Table 2).

The majority of the mutants affect the posterior localisation of GFP-Staufen and oskar mRNA, but have little or no effect on the anterior localisation of GFP-Staufen and bicoid mRNA(Fig. 3). These mutants display three distinct classes of phenotypes that indicate defects in the three processes required for oskar mRNA localisation: microtubule organisation, mRNA transport and mRNA anchoring. The first two classes are represented by 55 mutants, in which GFP-Staufen and oskar mRNA accumulate in the oocyte but fail to localise to the posterior in a significant proportion of oocytes, and either remain diffuse throughout the cell or localise to its centre. These mutants comprise five spn-Ealleles, the new alleles of orb and D-elg, and nine novel complementation groups (see Table 2).

Mutations in errance, sverdrup, poulpe, strindberg, fraenkel and lkb1 display phenotypes indicative of a defect in oocyte polarisation. The oocyte nucleus is very occasionally mislocalised in errance, sverdrup, poulpe and strindberg mutants. In addition, GFP-Staufen and oskar mRNA often localise in an ectopic central dot in poulpe and fraenkel mutants, as they do in par-1 and 14-3-3ϵ mutants(Fig. 3E,F)(Shulman et al., 2000; Tomancak et al., 2000; Benton et al., 2002). Examination of the microtubule cytoskeleton with Tau-GFP reveals that poulpe mutants display a disorganised microtubule network (data not shown), as do mutants of lkb1, which encodes the Drosophilahomologue of C. elegans par-4(Martin and St Johnston,2003). Thus, these six complementation groups probably affect mRNA localisation indirectly, by regulating microtubule organisation.

vagabond is the only complementation group in this class that gives the characteristic phenotype of mutants that disrupt oskar mRNA transport, in which the mRNA remains at the anterior of oocyte(Fig. 3C,D)(van Eeden et al., 2001). vagabond mutant oocytes display a range of additional phenotypes,including a mislocalised nucleus and aberrant organisation of the follicle cells, suggesting that the gene has other functions in addition to its role in oskar mRNA transport. In this respect, it may be similar to mago nashi and Y14 (Micklem et al., 1997; Newmark et al.,1997; Hachet and Ephrussi,2001; Mohr et al.,2001).

The third class of mutants consists of those in which oskar mRNA localises to the posterior, but fails to adopt a wild-type pattern (19 mutants, of which nine fall into four groups). Mutants in this group are unlikely to affect the transport of oskar mRNA or the polarisation of the oocyte, as oskar mRNA can reach the posterior. Instead, they probably disrupt the anchoring of oskar mRNA to the posterior cortex. In most cases, oskar mRNA and GFP-Staufen initially localise to the posterior normally (stage 9; Fig. 3G,H) but fail to be maintained there in later stages(Fig. 3I,J). In glissade (Fig. 3K,L)and sedov, GFP-Staufen and oskar mRNA localise to the posterior part of the oocyte but fail to restrict to a cortical crescent, a phenotype very similar to that of the oskar protein- null allele osk⁵⁴, which suggests that mutants in this group could also affect oskar translation(Ephrussi et al., 1991; Kim-Ha et al., 1991). In agreement with this, we found two novel oskar alleles in this class. Western blot analysis failed to detect Oskar protein in these mutants,suggesting that they are protein nulls (data not shown). Thus, most mutants in this group probably affect the translation or anchoring of oskarmRNA.

Several mutants disrupt the anterior localisation of GFP-Staufen but have little or no effect on posterior localisation, suggesting that they specifically affect the localisation of bicoid mRNA(Fig. 4). These mutants can be separated into two distinct classes based on when the defect in GFP-Staufen localisation is first observed. In the first class (five single alleles),GFP-Staufen never concentrates at the anterior of late stage oocytes and bicoid mRNA fails to accumulate at the anterior of the oocyte at any stage (Fig. 4C,D), indicating that these mutants affect an early step in the localisation of bicoidmRNA. In the second class (13 mutants, six of which form three complementation groups), bicoid mRNA localises to the anterior of the oocyte in stage 8-10 egg chambers, but GFP-Staufen fails to accumulate at the anterior of late stage oocytes (Fig. 4E,F). These mutants therefore affect a later event in the anterior restriction of bicoid mRNA. In agreement with this, bicoid mRNA is not concentrated at the anterior of these mutants in late stage oocytes or freshly laid eggs (data not shown; U. Irion, personal communication).

Forty mutants affect the localisation of both oskar and bicoid mRNAs (Fig. 5). Amongst these, we recovered four mutants (including one new spn-Eallele) in which the oocyte nucleus could be mispositioned and bicoidmRNA was occasionally localised at the posterior of the oocyte. These included one allele of the lethal complementation group boussole(Fig. 5C,D). oskarmRNA localises to the centre of the oocyte in this mutant, and bicoidmRNA can occasionally be seen at the posterior of the oocyte, along with the nucleus. This phenotype is reminiscent to that of mago or grk and suggests that this gene may be required in the oocyte to receive the polarising signal back from the posterior follicle cells(Gonzalez-Reyes et al., 1995; Roth et al., 1995; Micklem et al., 1997). However, this defect often correlates with morphological changes in the posterior follicle cells, and may reflect the presence of mutant follicle cell clones.

Twenty-one mutants affect all markers of polarity within the oocyte: oskar mRNA usually fails to localise to a posterior crescent, the oocyte nucleus is mislocalised, and bicoid mRNA associates strongly with the cortex above the nucleus. This class contains the 3 novel alleles of cnc. We also recovered 3 other complementation groups that give this phenotype: abruzzi, nansen and weyprecht. Of these, abruzzi displays the most striking phenotype(Fig. 5E,F): bicoidmRNA localises in a cortical ring at the level of the nucleus and is absent from the anterior cortex. In nansen and weyprecht, bicoidmRNA localises next to the mislocalised nucleus but is also present along the anterior cortex of the oocyte. To investigate whether this bicoidmRNA localisation phenotype is due to a defect in the organisation of the microtubule cytoskeleton, we stained germline clones of a subset of mutants from this class with an anti-α-tubulin antibody. All the mutants studied showed defects in the organisation of microtubules between stage 8 and 10(data not shown).

Two further complementation groups, mertz(Fig. 5G,H) and ellsworth, affect the localisation of both oskar and bicoid mRNAs. In these mutants, the mRNAs localise to their respective poles in much reduced amounts, or are diffusely distributed throughout the oocyte. However, the oocyte nucleus is always positioned normally, suggesting that the oocyte has been polarised correctly. These mutants may therefore affect factors, such as staufen, that are necessary for the transport of both oskar and bicoid mRNAs(Ferrandon et al., 1994). In agreement with this, oskar mRNA persists at the anterior of the oocyte in ellsworth mutants, as it does in other mutants that specifically disrupt oskar mRNA transport.

A number of mutants recovered for their defect in GFP-Staufen localisation also display unexpected additional phenotypes. sorcière(Fig. 6A) and gerlacheare dumpless mutants in which the bulk of GFP-Staufen remains in the nurse cells, in or just anterior to the ring canals. Although this phenotype suggests a defect in ring canal growth, the morphology and size of the ring canals is indistinguishable from wild type, as assayed by rhodamine-phalloidin staining. In trou mutants, the posterior of the oocyte detaches from the follicle cells (Fig. 6B). This detachment may interfere with signalling from the posterior follicle cells, which could account for the variable defects in oskar and bicoid mRNA localisation in these mutants. In sedov,vagabond and wellman mutant egg chambers, excess follicle cells are sometimes found between the oocyte and the nurse cells, indicating a possible defect in the control of follicle cell migration or proliferation(Fig. 6C). We also found mutants that affect the growth of the oocyte (seven mutants, of which two form the complementation group nain). As GFP-Staufen still accumulates in one cell in these mutants, the oocyte appears to be correctly determined, but fails to grow larger than the nurse cells. Finally, we found two mutants that affect the number of germ cell divisions, as the egg chambers contain more than 16 cells, and an oocyte with five or six ring canals(Fig. 6E).

As mentioned above, we recovered six new alleles of spn-E and four of orb: two genes that encode RNA-binding proteins required for the polarisation of both AP and dorsoventral (DV) axes(Lantz et al., 1992; Christerson and McKearin,1994; Lantz et al.,1994; Gillespie and Berg,1995). SPN-E is an RNA helicase required at multiple steps during oocyte development; notably, spn-E mutants disrupt oskar and gurken mRNA localisation, and have an abnormal microtubule cytoskeleton at stage 9 (Gillespie and Berg, 1995; Gonzalez-Reyes et al., 1997). Interestingly, one of our spn-E alleles displays a novel phenotype in which bicoid mRNA is occasionally localised to the posterior of the oocyte at stage 8-9, suggesting that this may be a stronger allele than those previously studied.

We examined the organisation of the microtubule network in three of the new spn-E alleles (spn-E^2A9-14, spn-E^4E2-14 and spn-E^8D4-11) and found that all display thick microtubule bundles at stage 9(Fig. 7A,B). This microtubule organisation resembles that observed in capu, spir or chicmutants, which cause similar defects in oskar and gurkenmRNA localisation (Neuman-Silberberg and Schüpbach, 1993; Theurkauf, 1994b; Manseau et al., 1996). As these mutants also cause premature cytoplasmic streaming, we studied the movement of autofluorescent vesicles in spn-E mutant oocytes. Whereas wild- type oocytes show slow and chaotic cytoplasmic movements at stage 9,40-60% of spn-E mutant oocytes show a very fast circular movement(Fig. 7E,F; see Movie 1). Thus, spn-E is required for the formation of the polarised microtubule network at stage 9 and for the regulation of cytoplasmic streaming.

The role of ORB during mid-oogenesis has previously been studied using a single hypomorphic allele, orb^mel, in which oskarmRNA fails to localise to the posterior and grk mRNA fails to localise to the dorsoanterior corner of the oocyte(Christerson and McKearin,1994). As ORB binds oskar mRNA, these phenotypes have been interpreted as a function of ORB in RNA transport and translation(Christerson and McKearin,1994; Chang et al.,1999). ORB is also necessary for the determination of the oocyte,as strong (orb^F303) and null (orb^F343)alleles arrest oogenesis at early stages(Lantz et al., 1994). We ordered the new orb alleles into an allelic series by analysing the strength of their phenotype in trans-heterozygous combinations over orb^F343 (Fig. 7I). Whereas our stronger allele arrests at early stages of oogenesis (orb^9D6-3), the weaker alleles show similar defects to orb^mel: trans-heterozygous females lay ventralised eggs with fused or no dorsal appendages, and the rare eggs that are fertilised have a reduced number of abdominal denticle belts, indicating a defect in pole plasm assembly.

Because orb mutants cause a similar defect in both AP and DV patterning to spn-E mutants, we studied the organisation of the microtubule cytoskeleton. Both anti-α-tubulin antibodies and Tau-GFP reveal large cortical bundles of microtubules in over 90% of stage 9 orb mutant oocytes (Fig. 7A,C; data not shown). This phenotype was observed in germline clones of all four new alleles, as well as in trans-heterozygous combinations of these alleles and orb^mel, over orb^F343. In addition, orb mutant oocytes display a very fast circular cytoplasmic movement at stage 9 that is characteristic of premature cytoplasmic streaming (Fig. 7E,G; see Movie 2). Thus, orb hypomorphs have the same effect on the microtubules and cytoplasmic streaming as capu, spir, chic or spn-E mutants,suggesting that this is the primary cause of the defect in oskar mRNA localisation. Interestingly, ORB protein levels are dramatically reduced in spn-E mutants, suggesting that SPN-E may be involved in the post-transcriptional regulation of orb mRNA(Fig. 7J).

Ypsilon Schachtel (YPS) is another RNA-binding protein that co-localises with oskar mRNA at the posterior, and has been shown to act antagonistically to ORB in oskar mRNA localisation and translation(Mansfield et al., 2002). Whereas oskar mRNA fails to localise to the posterior in over 90% of orb^mel/orb^F303 oocytes, about 50% of yps^JM2 orb^mel/yps^JM2orb^F303 have wild-type amounts of oskar mRNA at the posterior (Mansfield et al.,2002) (data not shown). This has been interpreted as a direct antagonism between YPS and ORB in the transport and translation of oskar mRNA (Mansfield et al.,2002). However, we do not detect significant changes in the level of OSK protein in orb yps versus orb mutants, indicating that YPS does not counteract the translational activation of oskar by ORB (Fig. 7J). By contrast,half of yps orb double- mutant oocytes have a wild-type microtubule organisation and a normal pattern of cytoplasmic streaming, which indicates that YPS acts antagonistically to ORB in the organisation of the microtubule cytoskeleton (Fig. 7D,H; see movies). This observation suggests that the presence of oskar mRNA at the posterior of yps orb double mutants reflects the restoration of a wild-type microtubule network, rather than a direct effect of YPS on oskar mRNA. YPS does not appear to exert its effect by altering the levels of ORB protein, as ORB protein levels are not increased in ypsmutants (Fig. 7J).

We describe here a genetic screen for mutants affecting the AP polarisation of the Drosophila oocyte that has two major advantages over screens performed previously. First, by using GFP-Staufen as an in vivo marker for the localisation of bicoid and oskar mRNAs, we screened directly for defects in oocyte polarity and did not rely on the resulting maternal-effect phenotypes in the embryonic cuticle. To compare these two approaches, we examined whether the embryos derived from germline clones of 51 of our mutants showed a cuticle phenotype. Only 3 produced a clear AP defect in the embryonic cuticle (data not shown), whereas the majority of mutants failed to produce eggs, or laid eggs that were unfertilised. These mutants would therefore not have been identified in any of the previous screens.

Second, we performed the screen in germline clones, which allowed us to recover lethal mutations. Indeed, 17 of the 23 new complementation groups are essential for viability and would have been missed in female-sterile or maternal-effect lethal screens. This suggests that about 70% of the genes necessary for AP axis formation have essential functions in other cell types. In the one case that we have examined in detail, we found that lkb1is also required for epithelial polarity(Martin and St Johnston,2003). Thus, a proportion of the mutants identified in this screen may be general factors regulating polarity in multiple cell types.

Surprisingly, four mutants that display clear defects in the localisation of GFP-Staufen produce eggs that hatch and develop into normal larvae. Two of these are only partially penetrant, but the others produce similar phenotypes to barentsz mutants, in which the posterior localisation of GFP-Staufen and oskar mRNA is strongly reduced. It has been suggested that the defect in oskar mRNA localisation in barentszmutants is partially rescued by localised translation, resulting in a weak grandchildless phenotype, and this may also be the case for these alleles(van Eeden et al., 2001). This illustrates another advantage of a direct visual screen that allows the identification of mutations with subtle phenotypes that do not cause lethality. The large size of the oocyte makes it a particularly good system in which to detect defects in cellular organisation, and we recovered mutations with a variety of polarity phenotypes that would have been very difficult to identify in smaller cells.

One disadvantage of germline clone screens is that they are laborious,which limits the number of lines that can be screened. We screened over 5000 mutagenised chromosomes and recovered new alleles for five of the seven known genes on chromosome 3R required for oskar mRNA localisation(oskar, spn-E, Ets97D, orb and cnc). However, we failed to recover new alleles of barentsz and tmII. One reason for this may be that these genes are not easily mutable by EMS, because all of the existing alleles that disrupt oskar mRNA localisation are either P-element insertions or deletions (Erdelyi et al., 1995; Tetzlaff et al.,1996; van Eeden et al.,2001). Alternatively, hypomorphic alleles of these genes may exist in our collection of mutants but have been overlooked in the complementation tests, because they only cause a nonpenetrant grandchildless phenotype, as is the case for btz¹ (van Eeden et al., 2001). Thus, the screen was efficient in finding mutants affecting the posterior localisation of GFP-Staufen but did not find mutants in all genes that can be mutated to give this phenotype.

Another indication that the screen did not reach saturation is that 66 mutants (47%) could not be attributed to any complementation group. However,the degree of saturation varies considerably between the different phenotypic classes. It is likely that we missed many mutants that only affect the anterior localisation of GFP-Staufen, as this localisation is more difficult to score in living oocytes than the posterior localisation because of the strong yolk autofluorescence at late stages of oogenesis. In addition,GFP-Staufen only localises to the anterior in late oocytes and is therefore visible in fewer egg chambers per ovary. Another class of genes for which we probably did not reach saturation are those required at multiple stages of oogenesis. orb, par-1 and 14-3-3ϵ, for example, not only function to polarise the oocyte during mid-oogenesis, but are also necessary for the determination of the oocyte in the germarium(Lantz and Schedl, 1994; Cox et al., 2001; Huynh et al., 2001; Benton et al., 2002). As germline clones of null mutations in these genes produce no late egg chambers,one can only recover hypomorphic mutations in a screen of this type. It is therefore likely that many of the single alleles correspond to hypomorphic mutants in such loci. However, it is now relatively straightforward to clone genes for which only one mutant allele is available by meiotic mapping with single nucleotide polymorphisms (Berger et al., 2001; Martin et al.,2001).

Although we may have missed a number of genes for these reasons, the screen was very successful in identifying new mutants that affect the posterior localisation of GFP-Staufen and oskar mRNA. For example, over 70% of the mutants with phenotypes similar to cnc belong to complementation groups, which suggests that we are approaching saturation for this phenotype. oskar mRNA localisation depends on the polarisation of the oocyte microtubule cytoskeleton, the transport of the mRNA to the posterior and its anchoring at the posterior cortex, and we identified new complementation groups that are required for each of these steps. The cloning of these genes should therefore provide valuable insights into the molecular mechanisms that underlie these processes.

One of the surprising results of the screen was the large number of mutants that appear to affect the polarity of the oocyte. These produce a range of different phenotypes that can be classified into at least four classes.

1. We recovered only a few mutants in which bicoid mRNA localises to both poles of the oocyte, which is the phenotype expected for a defect in the transduction of the polarising signal(Lane and Kalderon, 1994; Micklem et al., 1997; Newmark et al., 1997). All of these are either single alleles, or are the only mutants in a complementation group that give this phenotype, suggesting that at least some are not simple loss-of-function mutations. Thus, there do not appear to be many genes that are specifically required for the transduction of the polarising signal.

2. Most alleles of spn-E and orb, and capu, spirand chic mutants, cause the formation of large cortical bundles of microtubules that correlate with premature cytoplasmic streaming, and this disrupts the localisation of oskar but not bicoid mRNA(Theurkauf, 1994b; Manseau et al., 1996).

3. poulpe and fraenkel mutants give a similar phenotype to par-1 and 14-3-3ϵ, in which Staufen and oskarmRNA localise to a dot in the centre of the oocyte(Shulman et al., 2000; Tomancak et al., 2000; Benton et al., 2002). These genes may therefore be required to recruit the plus ends of microtubules to the posterior. A related phenotype is produced by lkb1, strindbergand sverdrup mutations, but in these mutants Staufen and oskar mRNA are usually found in a diffuse cloud in the centre of the oocyte instead of in a tight `dot'. A more detailed analysis of lkb1indicates that it also disrupts the recruitment of plus ends to the posterior(Martin and St Johnston,2003). The difference between the central dots and clouds is unclear, but may reflect the degree to which microtubule plus ends are focussed in the centre of the oocyte. As PAR-1 and LKB1 are protein kinases that function in the same signalling pathway, the other genes in this group may encode novel components of this pathway(Shulman et al., 2000; Martin and St Johnston,2003).

4. abruzzi, nansen, weyprecht and cnc mutants disrupt the localisation of all three markers of oocyte polarity: the oocyte nucleus, bicoid mRNA and oskar mRNA(Guichet et al., 2001). It is particularly striking that some or all bicoid mRNA localises around the cortex adjacent to the misplaced nuclei in these mutants. As both the nucleus and bicoid mRNA are believed to localise to microtubule minus-ends, this phenotype may reflect a defect in the nucleation of microtubules at the anterior of the oocyte. Nothing is known about how the new diffuse anterior MTOC forms at stage 7, but this occurs normally in grk mutants and is therefore independent of the posterior follicle cell signal. The mutants in this group may therefore provide an insight into how the microtubules are nucleated from the anterior cortex.

The diversity of phenotypes described above indicates that the polarisation of the microtubule cytoskeleton is a complex process that can be disrupted in several different ways. It is therefore likely that the signal from the posterior follicle cells is not transduced in a simple linear manner, but impinges on several parallel pathways that act together to generate the polarised microtubule array.

ORB protein has been shown to bind oskar mRNA, and is required for its localisation and translation at the posterior of the oocyte(Christerson and McKearin,1994; Lantz et al.,1994; Chang et al.,1999). Although this suggests that ORB plays a direct role in both processes, our results indicate that its effects on mRNA localisation are indirect. All hypomorphic orb alleles disrupt the organisation of the microtubule cytoskeleton and cause premature cytoplasmic streaming, and this can account for the dramatic reduction in the amount of localised oskar mRNA. Indeed, other mutants that cause premature cytoplasmic streaming, such as capu and spir, cause an identical oskar mRNA localisation phenotype(Theurkauf, 1994b; Manseau et al., 1996). Thus,ORB presumably controls the organisation of the microtubules by regulating the translation of some other mRNA(s), and the principal function of its interaction with oskar mRNA is to regulate its translation by promoting the formation of a long polyA tail(Chang et al., 1999; Castagnetti and Ephrussi,2003). This function is very similar to that of its Xenopus homologue, CPEB, which binds cytoplasmic polyadenylation elements (CPE) in the 3′UTRs of several maternal mRNAs to stimulate their polyadenylation and translation(Lantz et al., 1992; Hake and Richter, 1994).

The RNA-binding protein YPS has also been proposed to play a direct role in the localisation and translation of oskar mRNA, because it co-localises with the mRNA to the posterior of the oocyte and antagonises the effects of ORB (Mansfield et al.,2002). Our results again suggest that this effect is indirect. yps mutants partially rescue the microtubule and premature streaming phenotypes of orb hypomorphs but have little or no effect on the levels of Oskar protein. Thus, YPS presumably counteracts the effects of ORB on the translation of RNAs that regulate the microtubule organisation. This is consistent with the biochemical characterisation of YPS as a component of a multi-protein complex, containing EXU and ME31B, that has been proposed to repress the translation of many oocyte mRNAs(Wilhelm et al., 2000; Nakamura et al., 2001).

Mutants in spn-E produce very similar microtubule and premature streaming phenotypes to orb mutants, and have strongly reduced ORB protein levels. SPN-E is an RNA-helicase, raising the possibility that it exerts its effect by regulating the processing of orb mRNA(Gillespie, 1995). Alternatively, as ORB is known to regulate its own translation, ORB and SPN-E may function together to regulate the translation of common target RNAs, including that of orb itself(Tan et al., 2001). These results reveal a novel regulation of the microtubule network at the level of RNA translation, and it will be important in future to identify the RNA substrates of ORB, YPS and SPN-E.

We wish to thank Anne Ephrussi, Antoine Guichet, Tulle Hazelrigg and the Bloomington stock centre for fly stocks, and Alex Sossick for help with confocal microscopy. This work was supported by a Wellcome Trust Principal Fellowship to D.StJ. S.G.M. was supported by a studentship from the Swiss National Science Foundation. V.L. was a European Community Marie Curie post-doctoral fellow.