Patronin/Shot Cortical Foci Assemble the Noncentrosomal Microtubule Array that Specifies the Drosophila Anterior-Posterior Axis

1Many differentiated animal cells and all plant cells lack functional centrosomes, yet form highly organized microtubule (MT) arrays that play essential roles in cell polarity, organization, and function. For example, both Drosophila and rodent hippocampal neurons develop normally without active centrosomes, with the latter extending and even regenerating axons independently of centrosomal MT nucleation. Most Drosophila tissues lack functional centrosomes or microtubule organizing centers (MTOCs) in interphase.

2Anterior-posterior axis formation in the Drosophila oocyte provides a well-studied example of the role of noncentrosomal MTs. Although the oocyte contains centrosomes, which cluster near the nucleus, oogenesis proceeds normally in their absence. Instead, the majority of MTs grow from the anterior/lateral cortex, but not from the posterior, where the plus ends concentrate. This noncentrosomal MT array directs the localization of bicoid and oskar mRNAs to the anterior and posterior poles of the oocyte, respectively, to define the main body axis of the embryo. 3D modeling of the oocyte MT cytoskeleton has shown that restricting MT minus ends to the anterior/lateral cortex is sufficient to generate an MT network that can direct the transport of oskar mRNA to the oocyte posterior by kinesin.

3The formation of this polarized MT array is under the control of the PAR proteins, which localize in mutually antagonistic anterior and posterior cortical domains. The posterior crescent of the Par-1 kinase transmits this cortical polarity to the MT cytoskeleton by excluding minus ends from the oocyte posterior. It is not known, however, how PAR-1 activity is transduced into the asymmetric organization of MT minus ends, nor how the minus ends associate with the anterior/lateral cortex.

4The recent discovery of the Patronin family of MT minus-end-binding proteins, consisting of Patronin in Drosophila, CAMSAP1, 2, and 3 in mammals, and PTRN-1 in worms, has begun to reveal how the minus ends of noncentrosomal MTs are organized and maintained. The Patronins recognize and stabilize free MT minus ends by protecting them from depolymerization. Patronins appear to play a particularly important role in organizing MTs in differentiated cells. CAMSAP3 localizes to the apical domain in epithelial cells, where it is required for the formation of the apical-basal array of MTs. CAMSAP2 stabilizes neuronal MTs in axon and dendrites, and its knockdown leads to defects in axon specification and dendritic branch formation. Similarly, Caenorhabditis elegans PTRN-1 is required for normal neurite morphology and axon regeneration. The function of Drosophila Patronin has only been examined in cultured S2 cells, where its depletion leads to a decrease in MT number and an increase in free moving MTs.

5Although it is now clear that the Patronins play an important role in organizing noncentrosomal MTs in differentiated cells, little is known about the regulation of the distribution and activity of the Patronins themselves. Here we show that Patronin is recruited to the anterior/lateral cortex of the Drosophila oocyte by the spectraplakin, Shot, under the control of Par-1. These Shot/Patronin complexes form the cortical noncentrosomal MTOCs that organize the polarized MT network in the oocyte, which specifies the anterior-posterior axis.

6We previously isolated 11 new alleles of short stop (shot) in a screen for dominant suppressors of the bicaudal phenotype caused by mislocalizing oskar mRNA to the oocyte anterior. Shot is the single Drosophila spectraplakin, a giant cytoskeletal linker protein with an N-terminal actin-binding domain and two C-terminal domains that bind MT, the GAS2 domain, which binds to the MT lattice, and a more C-terminal domain that associates with MT plus ends through the +TIP, EB1. Null alleles of shot block the specification of the oocyte, and this is also the case for 10 out of 11 of the new alleles. Some germline clones of shot2A2 are not blocked in oogenesis, however, and develop to later stages, occasionally laying fertilized eggs that develop into larvae that lack an abdomen. Since this is a typical posterior group phenotype, we examined whether the posterior determinant, oskar mRNA, is correctly localized in shot2A2 mutants. Both oskar RNA and Staufen-GFP (an RNA-binding protein associated with oskar) fail to localize to the oocyte posterior in shot2A2 germline clones (Figures 1A and 1B). To determine whether Shot is specifically required for oskar mRNA localization or plays a more general role in kinesin-dependent transport to the posterior, we also examined the localization of Dynein and the dynactin subunit, Glued, which are transported to the oocyte posterior by kinesin independently of oskar mRNA. Neither Dynein nor Glued are localized in shot2A2 oocytes, indicating that either kinesin activity is inhibited or the MT plus ends are not concentrated at the posterior pole (Figure 1C).

7We next examined the overall organization of the MTs in fixed and living oocytes. Staining of fixed oocytes with anti-tubulin and in vivo labeling of MTs in living oocytes with Jupiter-GFP reveals the anterior-posterior gradient of MTs in wild-type with the highest concentration of MTs at the anterior (Figures 1D and 1E; Movie S1). This anterior enrichment is lost in shot2A2 and the MT organization becomes somewhat variable, with a much more even distribution throughout the oocyte cytoplasm (Figures 1D and 1E, right panels; Movie S2).

8Shot localizes to the anterior and lateral cortex of the oocyte, but is absent from the posterior, following the predicted distribution of MT minus ends. Shot is also strongly enriched at the apical side of the epithelial follicle cells that surround the developing egg chamber (Figure 2A, left). YFP-tagged Shot expressed from a transgenic bacterial artificial chromosome (BAC) rescuing construct shows an identical distribution in both the follicle cells and oocyte. We therefore examined whether the interaction of Shot with the oocyte cortex is under the control of the cortical Par proteins that control the polarity of the MT cytoskeleton. In par-1 mutant oocytes, MTs grow from the posterior cortex as well as the anterior/lateral cortex, and the MT cytoskeleton loses its asymmetry, whereas Par-1T786A, which has a uniform cortical distribution, abolishes all MT growth from the cortex. Shot responds to Par-1 activity in the same way as MTs: it extends around the posterior in the absence of Par-1, and is lost from the cortex in oocytes overexpressing Par-1T786A (Figures 2A and 2B). Thus, Shot is downstream of Par-1, consistent with it playing a role in MT minus-end localization.

9Sequencing of shot2A2 reveals that it is a point mutation in the first calponin homology domain of the N-terminal actin-binding domain (ABD) of Shot, changing Val224 (isoform PE) to Asp (Figure 2C). Val224 is well conserved among ABD-containing proteins. Structural analysis of the interaction of fimbrin with F-actin showed that the equivalent to Val224 (Val212 in fimbrin) directly contacts F-actin. In agreement with this, Shot loses its association with the actin-rich cortex in shot2A2 and is mainly cytoplasmic (Figure 2D). Like full-length Shot, the Shot ABD is enriched at the anterior-lateral cortex (Figure 2E, left). Introducing the Val224 to Asp mutation into the Shot ABD disrupts its cortical localization, although the protein still shows an enrichment at the ring canals, which is not observed with the full-length protein (Figure 2E, right). Thus, Shot is recruited to the cortex through its ABD, presumably by direct binding to cortical F-actin, and this interaction is inhibited at the posterior by Par-1.

10We took advantage of the recent identification of Patronin/CAMSAP as an MT minus-end-binding protein to analyze the relationship between cortical Shot and the distribution of MT minus ends in the oocyte. Live imaging of both transgenic and endogenously tagged Patronin reveals that it localizes to anterior/lateral cortex in the expected distribution of MT minus ends (Figures 3A and 3B; Movie S3, left panel). Importantly, Patronin co-localizes with Shot in distinct cortical foci (Figure 3C). Patronin localization is Shot dependent, as it becomes largely cytoplasmic in shot2A2 (Figure 3D and Movie S3, right panel). Furthermore, the cortical Patronin foci extend around the posterior cortex in par-1 mutant oocytes, as Shot does, consistent with the two proteins being in the same complex (Figure 3E). In agreement with this, Patronin co-immunoprecipitates with Shot-YFP from ovary extracts (Figure 3F). The fact that both Patronin and Shot are no longer cortical in shot2A2 indicates that Shot anchors Patronin to the cortex, providing an explanation of how the asymmetric localization of Shot controls the polarized distribution of MT minus ends in the oocyte.

11It has previously been shown, using Tau-GFP to label MTs and EB1-GFP to label the growing MT plus ends, that oocyte MTs grow out from noncentrosomal foci that can be visualized using an MT regrowth assay. Upon colcemid treatment, both proteins accumulate in cortical foci. Local inactivation of the colcemid with a pulse of UV light allows MTs to regrow from the cortex (Figure 4A). We therefore examined whether the MTs grow from the Patronin foci. Both EB1-GFP and Tau-GFP accumulate in the cortical Patronin foci upon colcemid treatment, indicating that these contain stable MT minus ends (Figures 4B and S1B). Furthermore, after colcemid inactivation with UV light, EB1-GFP and Tau-GFP label growing MTs that emerge from the Patronin foci (Figures 4C, S1C, and S1D; Movies S4 and S5). The Patronin foci also act as a source of growing MTs under steady-state conditions in the absence of colcemid (Figure 4D and Movie S6). After colcemid inactivation, each Patronin focus produces an average of 11.5 new MTs per minute (n = 15; SEM = 0.75), providing a source of MTs that grow in multiple directions (Figure 4E). Moreover, these foci are the only visible source of growing MTs at the oocyte cortex, strongly suggesting that they represent the noncentrosomal, cortical MT organizing centers (ncMTOCs) from which MTs grow to form the polarized cytoskeleton in the oocyte.

12In shot2A2 mutant oocytes, many of the foci fail to be retained at the oocyte cortex and redistribute throughout the oocyte cytoplasm, consistent with the loss of most Shot and Patronin from the cortex in this mutant (Figure 4F and Movie S7). These cytoplasmic foci remain active, however, producing growing MTs after colcemid inactivation, explaining why the overall polarity of the MT network is disrupted (Movie S7).

13A patronin null mutant blocks oogenesis at an early stage. To test whether Patronin is required for the activity of the cortical ncMTOCs in the oocyte, we therefore used a hypomorphic allele, patronin05252, which strongly reduces Patronin levels. patronin05252 homozygous oocytes contain 90% fewer cortical EB1-GFP foci after colcemid treatment than wild-type, and the remaining foci also generally recruit less EB1-GFP (Figures 5A and 5B). Nevertheless, the Patronin foci that form are still active, acting as a source of growing MTs after colcemid inactivation (Figure 5C and Movie S8). The density of MTs is also significantly reduced in patronin05252 clones, as expected from the reduced number of cortical ncMTOCs (Figure 5D). Despite the dramatic reduction in MT number, there are still sufficient MTs to direct the localization of Staufen/oskar mRNA complexes to the oocyte posterior, although the levels of localization are reduced by >40% (Figures 5E and 5F).

14To further investigate the nature of the Shot/Patronin noncentrosomal MTOCs, we asked whether they contain γ-tubulin as the source of new MTs. Antibody staining of oocytes for γ-tubulin label only the centrosomes adjacent to the oocyte nucleus, but overexpressed γ-tubulin 37C-GFP is also seen in weak foci along the anterior/lateral cortex. We therefore co-expressed γ-tubulin-GFP and Cherry-Patronin to determine whether the two proteins co-localize (Figures 6A and 6A′). Patronin labels some of the nuclear-associated, γ-tubulin foci, which probably correspond to the active centrosomes. The cortical Patronin foci do not co-localize with the γ-tubulin-GFP foci, however, and Shot/Patronin ncMTOCs contain no detectable γ-tubulin. Since MTs start to grow out from Patronin foci within 1 s of colcemid inactivation, and these foci are the only visible source of cortical MTs, it seems most likely that the MTs are seeded from Patronin-stabilized MT minus-end stumps and not from de novo nucleation by the γ-tubulin ring complex.

15Overexpression of the centriolar duplication factors dSAS6, dSas4, Sak/PLK4, and Ana2/STIL can promote the formation of acentriolar MTOCs in the oocyte. Moreover, expression of membrane-tethered Cep152/Asl and PLK4 is sufficient to induce formation of ectopic acentriolar MTOCs in mouse oocytes. To test whether any of these acentriolar MTOC components are involved in the formation of the Shot/Patronin ncMTOCs, we co-expressed Cherry-Patronin with Asl-GFP (Figure 6B), Ana2-GFP (Figure 6C), dSas6-GFP, dSas4-GFP, and Sak-GFP (data not shown). None of these proteins co-localize with the Patronin foci, however, indicating that they are not components of the ncMTOCs (Figures 6B and 6C, and data not shown).

16An alternative mechanism that can contribute to the formation of new MTs is the severing of existing MTs to generate minus ends that act as seeds for new microtubule growth. The mammalian Patronin orthologs, CAMSAP2 and CAMSAP3, associate with the microtubule severing protein, Katanin. This association is conserved in Drosophila, as a protein trap insertion that labels endogenous Katanin 80 co-localizes with Patronin in the cortical foci in the oocyte and at the apical side of the follicle cells (Figure S2). Furthermore, Katanin 80-YFP co-immunoprecipitates with Patronin from ovary extracts, confirming that it is a component of the cortical Patronin complex (Figure 3F). Thus, MT severing by Katanin may contribute to the generation of new MTs in the Patronin ncMTOCs.

17In epithelial cells, noncentrosomal MTs form apical-basal arrays with their MT minus ends concentrated at the apical cortex. The mammalian Patronin homolog, CAMSAP3, localizes to the apical cortex of mouse intestinal cells and human Caco2 cells, and mutation of camsap3 leads to a random orientation of MTs. To test whether Patronin ncMTOCs play a similar role in the formation of the apical-basal array of MTs in Drosophila epithelia, we analyzed the localization of Patronin in the follicle cells, larval salivary glands, and male ejaculatory duct (Figures 3B, 7A, and S3). Patronin localizes apically in all three epithelia, forming multiple apical foci in the follicle cells, but is excluded from the adherens junctions (Figure 7B). Live imaging of EB1-GFP and Jupiter-GFP reveals that most MTs grow from the region of apical Patronin foci (Figure 7C and Movie S9). Although capsap3 null cells contain relatively normal numbers of MTs, patronin05252 mutant cells have very few MTs (Figure 7D), presumably because it is the only copy of this gene in Drosophila. In addition, larger patronin05252 mutant clones often lead to tissue disorganization and multi-layering (Figures 7G and S4A). This suggests that Patronin apical foci act as ncMTOCs in epithelial cells and that they are crucial for tissue integrity.

18Shot also localizes apically in the follicle cells and the embryonic salivary gland epithelium, and has been proposed to link acentrosomal MT minus ends to medial actomyosin, although this does not appear to require its ABD. This suggests that Shot may have similar role as an anchor of Patronin ncMTOCs in epithelial cells. In agreement with previous studies, we observed that Shot is strongly enriched at the apical side of the follicle cells, where it co-localizes with Patronin (Figures 2A and 7A). In homozygous clones of the ABD mutant, shot2A2, Shot protein at the apical cortex is slightly reduced and the protein is found throughout the cytoplasm, indicating that the ABD contributes to efficient apical recruitment (Figure 7E).

19To examine the role of Shot in MT organization, we generated clones of shot3, a null mutation. Mutant clones lose the pronounced apical enrichment of MTs seen in wild-type cells and have fewer MTs than normal, with the remaining MTs mainly along the lateral cortex (Figures 7D and S4B). shot3 mutant cells contain more MTs than patronin mutant cells, however, and the absence of Shot does not disrupt the apical localization of Patronin (Figure 7F).

20It has previously been shown that Patronin functions during spindle elongation in the embryo and in interphase S2 cells to protect MT minus ends from the depolymerizing kinesin, Klp10A (kinesin-13), as simultaneous knockdown of Klp10A and Patronin rescues the MT phenotype of Patronin knockdown alone. To ask whether Patronin also antagonizes KLP10a in epithelial cells, we examined the MT phenotype of klp10a patronin double-mutant clones. Loss of KLP10a partially rescues MT abundance in patronin mutant cells, but does not rescue the apical enrichment of MTs, resulting in an MT phenotype that is similar to that seen in shot3 (Figure 7G). By contrast, klp10a has no effect on MT density or organization in shot3 cells (Figure 7H). Thus, Patronin is required both to position MT minus ends apically and to protect them from depolymerization by Klp10A. Shot is not required for Patronin's activity in protecting MT minus ends, but the fact that shot and klp10a patronin mutants produce very similar defects in MT organization suggests that Shot and Patronin act in the same pathway to anchor MTs apically. We also tested whether Patronin functions in the oocyte to protect MT minus ends from depolymerization by Klp10A. However, klp10a patronin double-mutant germline clones show the same reduction in MT density as the patronin single mutant, suggesting that Klp10A plays little role in the germline (Figure S5).

21The polarized arrangement of the MTs in the Drosophila oocyte depends on the posterior crescent of the Par-1 kinase, which excludes MT minus ends from the posterior cortex. Here we show that Par-1 acts by preventing the association of Shot with the posterior actin cortex, thereby restricting the formation of noncentrosomal MTOCs to the anterior and lateral cortex. Computer modeling has shown that this asymmetric localization of MT minus ends is sufficient to explain the formation of the weakly polarized MT network that directs the transport of oskar mRNA to the posterior pole. Thus, the regulation of the interaction of Shot with the cortex by Par-1 transmits cortical PAR polarity into the polarization of the MT cytoskeleton that localizes the axis determinants (Figure 7I).

22The mechanism by which Par-1 excludes Shot is unknown. The interaction of Shot with the cortex depends on its N-terminal calponin homology domains, which bind to F-actin. Thus, Par-1 could phosphorylate Shot to inhibit its binding to the cortex. If this is the case, Par-1 would have to modify the activity or accessibility of the N-terminal ABD of Shot, as this domain recapitulates the posterior exclusion and cortical recruitment of the full-length protein. We have not detected any phosphorylation of the ABD by Par-1 in vitro, however, and it seems more likely that Par-1 acts by modifying the cortex to block the binding of Shot.

23Shot and its vertebrate ortholog, MACF1, have previously been shown to interact with the MT plus-end tracking protein EB1 through their C-terminal SxIP motifs and with the MT lattice through their Gas2 and C-terminal domains. Our results indicate that in addition to binding to MT plus ends and to the MT lattice, Shot also interacts with MT minus ends through its association with the Patronin/Katanin complex. The exact nature of the interaction between Shot and the Patronin complex is unclear, but Shot was found to interact with Katanin 60 in a high-throughput yeast two-hybrid screen. Thus, one possibility is that Katanin acts as a link between Shot and Patronin. Since Shot is exclusively cortical in the oocyte, the protein does not appear to bind to MT plus ends or along the body of MTs in this system. It will therefore be interesting to investigate whether the different modes of MT binding by Shot are mutually exclusive and how this is regulated.

24Several models have been proposed to explain the formation of noncentrosomal MTs. Upon centrosome inactivation in postmitotic Drosophila tracheal cells and C. elegans intestinal cells, γ-TuRC complexes and other pericentriolar material (PCM) components are released from the centrosome and transported toward the apical membrane, where they nucleate MT. Whole MTs released from the centrosome can also be delivered and anchored to the apical domain or cell junctions by Ninein. Alternatively, new MTs can grow from MT ends generated by severing enzymes, a mechanism that is thought to be important in plant cells and neurons. Here, we present evidence that this latter mechanism is responsible for the formation of the MT array that directs Drosophila axis formation. Firstly, Shot/Patronin ncMTOCs contain stable minus ends even after treatment with the MT-depolymerizing drug, colcemid, as shown by the persistent recruitment of Tau-GFP and EB1-GFP to these foci. This is consistent with the ability of Patronin and CAMPSAPs to capture and stabilize minus ends of single MTs in vitro and in cells. Secondly, MTs start to grow out in all directions from the Shot/Patronin foci immediately after colcemid inactivation. Indeed all visible growing MTs emanate from Patronin foci, indicating that they are the principal source of MTs in the oocyte. Thirdly, the foci contain no detectable γ-tubulin and do not co-localize with PCM proteins. This is consistent with observations in Caco-2 cells, which showed that CAMSAP2 and CAMSAP3 do not co-localize with γ-tubulin and in the C. elegans epidermis, where PTRN-1 and γ-tubulin function in parallel pathways to assemble circumferential MTs.

25Taken together, these results suggest a model in which the Shot/Patronin foci act as ncMTOCs by capturing and stabilizing MT minus-end stumps that then act as templates for new MT growth. One attractive feature of this model is that it uncouples MT organization from MT nucleation in both space and time. The Shot/Patronin complex bypasses the need to continually nucleate new MTs by preventing existing microtubules from completely depolymerizing. Thus, once a cell has nucleated sufficient MTs, it can maintain and reorganize its MT cytoskeleton by stabilizing MT minus-end stumps in appropriate locations and using these, rather than the γ-tubulin ring complex, to provide the seeds from which new MTs grow. The number of MTs can even increase in the absence of new MT nucleation if MT-severing proteins chop up existing MTs to produce new minus ends that can then be captured and stabilized. The presence of the severing protein, Katanin, in the Shot/Patronin foci is intriguing in this context, as it raises the possibility that it severs existing MTs to provide a local source of minus ends for Patronin to capture.

26Shot and Patronin also co-localize at the apical cortex of the epithelial follicle cells, where they are required for apical-basal MT organization. This consistent with the recent observation that CAMSAP3 is required for the recruitment of MT minus ends to the apical cortex of mammalian intestinal epithelial cells. Thus, this function of Patronin has been evolutionarily conserved. Furthermore, the similarities between roles of Shot and Patronin in the oocyte and the follicle cells suggest that the complex may provide a general mechanism for organizing noncentrosomal MTs. The relationship between Shot and Patronin is different in the follicle cells compared with the oocyte, however, as Shot is not required for the apical recruitment of Patronin. Nevertheless, loss of either protein produces a very similar loss of apical MT and a reduction in overall MT density. Although we cannot rule out the possibility that they act in parallel pathways, this observation suggests that they collaborate to anchor MTs to the apical cortex. The combination of Patronin binding to the MT minus ends and Shot binding to the MT lattice may therefore provide a robust anchor to retain MTs at the apical cortex.

27The protocol was modified from. Flies were starved for 3 hr and then fed colcemid (Sigma) in yeast paste (66 μg/ml) for 2–3 hr. Ovaries were dissected and imaged as described below. Colcemid was inactivated with a brief UV pulse (3–5 s).

28For live imaging, ovaries were dissected and imaged in Voltalef oil 10S (VWR International) on an Olympus IX81 inverted microscope with a Yokogawa CSU22 spinning disk confocal imaging system (40× 1.35 NA Oil UPlanSApo, 60× 1.35 NA Oil UPlanSApo, and 100× 1.3 NA Oil UPlanSApo). Fixed preparations were imaged using Olympus IX81 (40× 1.35 NA Oil UPlanSApo, 60× 1.35 NA Oil UPlanSApo) and Zeiss LSM510 (40× NA 1.3 Oil Plan-NeoFluor) confocal microscopes. Images were collected with Olympus Fluoview, LSM510 AIM software, or MetaMorph software and processed using ImageJ. The oocyte cortex was imaged by collecting 10–15 z sections spaced 0.5 μm apart and then merging them.

29Ovaries were fixed for 10 min in 10% paraformaldehyde and 2% Tween in PBS. Ovaries were then blocked with 10% BSA in PBS for 1 hr at room temperature. Ovaries were incubated with the primary antibody for 16 hr in PBS with 0.2% Tween and for 4 hr with the secondary antibody. In situ hybridizations were performed as previously described. We used the following primary antibodies: mouse anti-α-tubulin fluorescein isothiocyanate at 1:250 (Sigma); mouse anti-Dynein heavy chain at 1:50 (DSHB); rabbit anti-Glued antibody raised against amino acid residues 1–400 of Glued and used at 1:100; mouse anti-DIG Cy3 at 1:200 (Jackson Immunoresearch), rabbit anti-Patronin at 1:300 (gift from R. Vale, HHMI and UCSF, USA); mouse anti-Armadillo at 1:100 (DSHB); and guinea pig anti-Shot antibody raised against amino acid residues 2,602–3,640 (isoform PE) and used at 1:500. Conjugated secondary antibodies (Jackson Immunoresearch) were used at 1:100.

30To generate a rescuing genomic shot transgene with C-terminal YFP tag, we used the PACMAN CH321-44M3 BAC clone covering the entire shot locus. The BAC was modified using the galK positive/counter-selection cassette and recombineering. Transgenic flies were created by Genetivision.

31The Patronin C-terminal YFP knockin was made by injecting nos>Cas9 embryos with a single guide RNA targeting the region of the stop codon in patronin (5′-GGCGCTTGTAATCTAAGCGG-3′, the stop codon is in bold) and a donor plasmid with 4-kb homology arms surrounding the Venus sequence.

32pUASP-mKate-ABD was constructed by amplifying Shot ABD and mKate2 with the following primers: 5′-ATGTAGCGGCCGCCCGCGATGCCATTCAGAAGA-3′ and 5′-ATGTATCTAGATCAAATGTACGTGATGAGGGACT-3′; 5′ACGTGGTACCATGGTGAGCGAGCTGATT-3′ and 5′ATGTAGCGGCCGCGGAAGAGGAAGATCTGTGCCCCAGTTTGCT-3′. The amplified fragments were cloned into the pUASP vector. The mutated Shot ABD was amplified with 5′-GATCAAACTGGACAACATACG-3′ and 5′-CGTATGTTGTCCAGTTTGATC-3′. Shot RE cDNA was obtained from A. Prokop (University of Manchester, UK).

33For generation of pUASP-mCherry-Patronin and pUMAT-mCherry-Patronin, patronin RI and mCherry were amplified with 5′-ATGTAGGTACCATGGTGAGCAAGGGCGAGGAGGATAACA-3′ and 5′-GCATTCTAGATTAGATTACAAGCGCCATGTCTTTT-3′ from the pMT-mCherry-Patronin plasmid (Addgene) and cloned into the pUASP vector and the pUMAT vector.

34For generation of pUMAT-YFP-Patronin, patronin RI and YFP were amplified with 5′-ATGGACGAGCTGTACAAGCACCGGTATACAAGT-3′ and 5′-GCATTCTAGATTAGATTACAAGCGCCATGTCTTTT-3′, and 5′-TAGTAGGTACCCATGAGCAAGGGCGAGG-3′ and 5′-ACTTGTATACCGGTGCTTGTACAGCTCGTCCAT-3′, respectively and cloned into the pUMAT vector.

35shot2A2 genomic DNA was isolated from homozygous embryos and larvae using the Gentra Puregene Cell Kit (Qiagen), and exonic regions were amplified by PCR and sequenced. Primer sequences are available on request.

36D.N. performed most of the experiments and data analysis. A.R.F. performed immunoprecipitations. D.N. and D.St J. planned the experiments. D.N. and D.St J. conceived the project and wrote the manuscript.