Souffle/Spastizin Controls Secretory Vesicle Maturation during Zebrafish Oogenesis

During oogenesis, the egg prepares for fertilization and early embryogenesis. As a consequence, vesicle transport is very active during vitellogenesis, and oocytes are an outstanding system to study regulators of membrane trafficking. Here, we combine zebrafish genetics and the oocyte model to identify the molecular lesion underlying the zebrafish souffle (suf) mutation. We demonstrate that suf encodes the homolog of the Hereditary Spastic Paraplegia (HSP) gene SPASTIZIN (SPG15). We show that in zebrafish oocytes suf mutants accumulate Rab11b-positive vesicles, but trafficking of recycling endosomes is not affected. Instead, we detect Suf/Spastizin on cortical granules, which undergo regulated secretion. We demonstrate genetically that Suf is essential for granule maturation into secretion competent dense-core vesicles describing a novel role for Suf in vesicle maturation. Interestingly, in suf mutants immature, secretory precursors accumulate, because they fail to pinch-off Clathrin-coated buds. Moreover, pharmacological inhibition of the abscission regulator Dynamin leads to an accumulation of immature secretory granules and mimics the suf phenotype. Our results identify a novel regulator of secretory vesicle formation in the zebrafish oocyte. In addition, we describe an uncharacterized cellular mechanism for Suf/Spastizin activity during secretion, which raises the possibility of novel therapeutic avenues for HSP research.


Introduction
Oogenesis prepares the egg to start the development of a new organism. During their development, oocytes actively import and secrete proteins using the basic cellular mechanism of vesicle transport [1][2][3]. As a consequence, oocytes contributed substantially to our understanding of vesicle trafficking e.g. Clathrin-coated vesicles were first described in mosquito oocytes [4]. Moreover, a plethora of novel regulators were discovered by exploiting the genetics of the Caenorhabditis elegans oocyte [5][6][7][8][9]. In zebrafish, oogenesis starts with a burst of secretory vesicle formation, which are called cortical granules [10,11]. Shortly afterwards, oocytes endocytose enormous amounts of the yolk precursor protein Vitellogenin (Vtg) by receptor-mediated endocytosis increasing its volume about 3000-fold within ten days [12,reviewed in 13]. Hence, the zebrafish oocyte provides the opportunity to integrate vertebrate genetics to visualize active trafficking of abundant and large vesicles in one big cell.
In humans, defects in vesicle trafficking lead to neurodegenerative diseases [14][15][16]. The disorder Hereditary Spastic Paraplegia (HSP) is characterized by progressive loss of lower limb motility [17][18][19]. At the cellular level, this spasticity is caused by the axonal degeneration of neurons in the corticospinal tracts, which are considered the longest axons in the human body. However, the cellular analysis of the dying neurons is hampered by the adult onset of the disease and the complexity of the nervous system. Therefore, the precise cellular mechanism for most HSP genes is currently controversial.
One of the HSP genes with multiple cellular roles is SPG15, which encodes SPASTIZIN aka ZFYVE26 or FYVE-CENT [20,21]. In a cell culture RNAi screen, SPASTIZIN was shown to be necessary for cytokinesis, but this defect is not observed in human and murine mutants [21,22]. This cell culture study found that the endogenous protein localizes to centrosomes and the midbody during cytokinesis of human fibroblasts. In neuronal tissue culture, SPASTIZIN was shown to localize to microtubules, ER, mitochondria and endosomal vesicles [22,23]. In additional tissue culture reports, SPASTIZIN interacted with the UVRAG complex during DNA repair [24] and autophagosome maturation [25]. Furthermore, SPASTIZIN binds to the recently discovered adaptor protein complex 5 (AP5) regulating multivesicular body (MVB) formation [26,27]. Although these localization data in different tissue culture cells are not mutually exclusive, it is difficult to reconcile the described cellular functions of SPASTIZIN in fibroblasts into an underlying cellular defect causing neurodegeneration in neuronal cells.
Mutants provide an essential tool to determine the endogenous role of a gene. Previously, we used a mutagenesis screen in zebrafish to discover vertebrate regulators of egg development and early embryogenesis, which isolated the souffle (suf) mutation named after its defect during oogenesis [28]. Here, we positionally cloned suf and show that it encodes the homolog of SPASTIZIN.
In suf mutants, we show that oocytes expand a Rab11b-positive compartment, but correctly transport the recycling cargo Transferrin. Moreover, Suf colocalizes with Rab11b on secretory vesicles called cortical granules in the oocyte. We demonstrate genetically that Suf/Spastizin is essential for the formation of cortical granules. Importantly, our subcellular analysis indicates that loss of Suf/Spastizin inhibits vesicle maturation, probably during sorting, which is necessary to complete the formation of Clathrin-coated buds and eventually to pinch-off vesicles. Finally, blocking vesicle scission with the pharmacological Dynamininhibitor Dynasore mimics the mutant phenotype supporting the hypothesis that Suf is required for vesicle fission, which is critical for the maturation of secretory granules in the egg. Collectively, these results identify Suf/Spastizin as a novel key gene controlling the maturation of cortical granules in zebrafish oocytes, which may also bring us closer to understand the cellular etiology of HSP.

Souffle encodes the zebrafish homolog of SPASTIZIN
Contrary to transparent eggs of wild-type (wt) mothers, maternal suf mutants spawn opaque eggs, which fail to proteolytically cleave yolk proteins [28]. In addition, mutant stage V eggs did not elevate their chorion after activation as observed in wt ( Figure 1A, B). With this early defect the suf mutation rather causes a female-sterile than a maternal-effect phenotype. To characterize the molecular mechanism of the defect, we positionally cloned the disrupted gene in suf mutants. We previously located the suf mutation on chromosome 13 of the zebrafish genome. Genotyping 1183 females identified 5 recombinants with the SSLP marker z25580/G47633 and another 5 recombinants with z21403/ G41743 restricting the critical interval with the suf mutation to 1.22 Mb ( Figure 1C). Generating novel SSLP markers AL13-10 and AL13-13, we reduced the interval to 270 kb. Within this genomic region, we sequenced the cDNAs of arginase, vti1b (t-snare), rdh12 (retinol dehydrogenase 12), zfyve26 (spastizin), galectin and pleckstrin2 ( Figure 1D). The zfyve26 gene consists of 41 exons and encodes a predicted mRNA of 7798 bp ( Figure 1E). Comparing the zfyve26 cDNA-sequence between wild type and the p96re allele of suf mutants showed a 25 bp deletion at the 39-end of exon 35 ( Figure 1G). However, the genomic sequence revealed a single point mutation in a splice donor, which probably results in cryptic splice donor selection 25 bp upstream of the wt splice site ( Figure 1F). The deletion of 25 bp results in a frameshift, which creates a termination codon after six aberrant amino acids ( Figure 1G). The premature STOP deletes 282 of the 2552 amino acids in the predicted Zfyve26 protein resulting in a shortened protein of 2270 amino acids (Figures 1H). Searching for protein motifs with the Prosite database [29] and MyHits [30] revealed a bipartite nuclear localization signal (amino acid 714-730), a serine-rich domain (aa 1251-1342) and a zinc finger FYVE domain (aa 1807-1865). FYVE domains bind to the lipid phospatidylinositol-3-phosphate (PI3P) predominantly present on endosomes [31][32][33][34]. Phylogenetic analysis detected one homolog in most metazoan genomes and a similar gene in Drosophila melanogaster (CG5270), albeit no homolog in the C. elegans genome ( Figure 1I). Remarkably, although teleosts underwent an additional genome duplication compared to tetrapods [35], all ten teleost genomes available at the Ensembl-database (http://www. ensembl.org) contain a single paralog of Suf/Spastizin, which we confirmed for zebrafish by BLAST searches with full-length Suf (data not shown). Alignment of the vertebrate protein sequences discovered a highly conserved C-terminus, which we termed Sufdomain. Scanning the zebrafish genome with the Suf-domain did not detect other proteins with a similar amino acid motif. This motif is also conserved in plants, but the protein does not contain a FYVE-domain (data not shown). In human tissue culture cells, this domain in SPASTIZIN interacts with BECLIN1, KIF13A and TTC19 [21,36] and is predicted to form alpha-helical solenoids as found in Clathrin heavy chain [27]. Since the Suf domain is deleted in the p96re allele, it is essential for Suf function during zebrafish oogenesis. Together, these data identified souffle as the zebrafish homolog of SPASTIZIN [20].
Souffle/spastizin is maternally expressed HSP patients require SPASTIZIN in the nervous system [20] and similarly, a recent mouse k.o. shows an adult onset neurodegeneration but not an embryological defect [22]. In zebrafish, inhibiting Spastizin by injection of morpholino-oligonucleotides (MO) leads to embryogenesis defects such as a twisted tails [37]. A fraction of these embryos without morphological changes show motor axon outgrowth failure leading to reduced motility, but are not paralyzed. Unexpectedly, we did not detect a neurological phenotype in zygotic zebrafish suf 2/2 embryos such as a touch response defect or a failure in motor axon outgrowth (data not shown). However, upon MO-injection, we confirmed the previously reported severe phenotypes ( Figure S1).
Hence, we investigated whether the suf/spastizin gene is expressed at the appropriate time to control oogenesis in zebrafish. Real-time PCR analysis of isolated follicles at selected stages showed that suf/spastizin mRNA is expressed during oogenesis with

Author Summary
Oocytes of egg laying animals frequently represent the biggest cell type of a species. The size of the egg is a consequence of active transport processes, e.g. the import of yolk proteins, which results in the massive storage of vesicles. In addition, secretory vesicles termed cortical granules are stored in the oocyte to be discharged right after fertilization during cortical reaction, which also occurs in mammals. Their secretion leads to chorion expansion, which prevents the lethal entry of additional sperm and protects the developing embryo against physical damage. Mutants with a defect in membrane transport are successful tools to discover genes regulating vesicle formation. We molecularly identify the disrupted gene in the recessive maternal-effect mutation souffle, which encodes a homolog of human SPASTIZIN. SPASTIZIN was previously implicated in endocytosis, but our cellular analysis of mutant oocytes connects this gene also with the regulation of cortical granule exocytosis. More precisely, we show that Suf/Spastizin is crucial for the maturation of cortical granules into secretion competent vesicles describing a novel role for this protein. Since SPASITIZN causes the disease Hereditary Spastic Paraplegia in humans, our results will help to decipher the pathogenesis of this neurodegenerative disorder.
a slight increase at the onset of vitellogenesis (stage II, for staging see Figure 1A) and another increase after ovulation (stage V) ( Figure 2A). This expression profile is consistent with a role of suf/ spastizin during zebrafish oogenesis.
During embryogenesis, suf/spastizin mRNA decreased after 4 hpf (hours post fertilization) similar to other maternal genes ( Figure 2B). At 30 hpf suf/spastizin showed a small peak of expression. Later during larval stages, the expression steadily increases consistent with microarray data in the Espresso database (http://zf-espresso.tuebingen.mpg.de; Unigene ID: Dr.21642). To analyze sex-specific expression of suf/spastizin, we compared mRNA levels in whole females, females without ovaries and males. Contrary to its specific phenotype in the oocyte, suf/spastizin was strongly expressed in males and even outside the female germline suggesting that it also acts in somatic cells ( Figure 2C).
The higher expression of Suf in males can be explained by two non-exclusive reasons. A simple explanation might be technical, i.e. the genes gapdh and odc1, which we used to normalize mRNA levels, may be differentially expressed between males and females as reported for ef1a [38]. Alternatively, Suf is indeed higher expressed in males, possibly in one of the non-reproductive organs with sex-specific gene expression such as the brain or the liver [39]. However, the p96re allele clearly demonstrates that Suf is required in the oocyte, but does not exclude that it has a critical role in other organs, which we did not notice.
Since we did not observe a mutant phenotype outside the germline, we addressed whether the p96re mutation causes a complete loss-of-function null-allele or forms a hypomorph. Comparison of suf/spastizin mRNA levels between +/+, +/2 and 2/2 ovaries showed a strong reduction after loss of one suf/ spastizin copy in heterozygous adults, but no phenotype ( Figure 2D). By contrast, 2/2 mutant females showed only a minor reduction of mRNA compared to +/2 heterozygotes, but eggs displayed the mutant phenotype. To examine whether an alternatively spliced suf/spastizin mRNA hides a potential zygotic mutant phenotype in other tissues or in males, we analyzed the expression of exon 35 carrying the mutation during zebrafish embryogenesis and oogenesis. We did not observe a shorter transcript lacking exon 35, which would generate a 413 bp product ( Figures 2E and S2A). However, in hetero-or homozygotes we detected the predicted 25 bp shorter transcript consistent with the mutant splice donor. Since suf/spastizin expression at the mRNA level was not completely eliminated, the residual protein might have sufficient activity to compensate for Suf/Spastizin requirement in somatic cells. To analyze Suf/Spastizin protein expression, we generated an antibody against the zebrafish protein ( Figure S2B) and compared mutant and wt oocytes in whole-mount immunostainings. Although mRNA levels were strongly reduced in mutant ovaries, Suf/Spastizin protein still sufficiently accumulated during oogenesis and localized to the membrane and cytoplasmic vesicles in wt as well as mutant oocytes ( Figure 2F). Taken together, Suf/ Spastizin was expressed during zebrafish oogenesis, but was also present outside of the germline.

souffle/spastizin mutants accumulate Rab11b-positive vesicles
In tissue culture cells, the FYVE domain of Suf/Spastizin interacts with the endosomal lipid PI3P indicating a role in endosomal trafficking [21]. Moreover, in human and mouse cells Spastizin binds to the novel AP5 complex regulating endosomal transport [22,26,27,40]. To examine genetically in zebrafish oocytes, whether Suf/Spastizin is involved in endocytosis during oogenesis, we compared endosomal compartments between wt and mutant. The gross morphology of oocyte vesicles showed no difference in early endosomes (Rab5) or late endosomes (Rab7) ( Figure 3A) [41,42]. In contrast, Rab11b-positive recycling endosomes showed a remarkable transformation of their tubular shape in wt to patches accumulating below the nuclei of the surrounding follicle cells in mutant oocytes ( Figure 3A) [43].
To quantify the defect in suf/spastizin mutants, we counted Rab positive foci in optical sections in deeper layers of the oocyte cytoplasm about 50 mm below the cortex, where single, small vesicles are easier to discriminate than the large compartments at the cortex ( Figure 3B confirming the initial observation that Suf/Spastizin controls trafficking of recycling endosomes in zebrafish oocytes. Consistent with previous reports in cell culture and mouse mutants [22,25], we also discovered in zebrafish that smaller lysosomes accumulated in suf/spastizin oocytes, possibly preventing yolk proteolysis in the egg and causing the opaque cytoplasm phenotype ( Figure 3B).
To functionally analyze during oogenesis whether the observed accumulation of the compartment-specific Rab proteins reflects a defect in the corresponding transport route as described for tissue culture cells, we established cargo trafficking assays in the zebrafish oocyte. LDL follows the degradative transport route to lysosomes and the yolk-receptor belongs to the LDL-receptor superfamily  [44,45]. Adding fluorescent LDL to the culture medium labeled yolk globules in zebrafish oocytes, which correspond to lysosomes of somatic cells [46]. However, we observed no difference in the LDL transport to wt or mutant lysosomes, which were also identified by a characteristic black halo after fixation ( Figure 3D, also visible in Figure 3B) suggesting that transport along the degradative pathway is not disrupted.
Transferrin follows the recycling route [47][48][49]. The zebrafish Transferrin-receptor is maternally expressed and the Transferrin recycling assay was previously also applied to Xenopus oocytes  [50,51]. Notably, contrary to the accumulation of Rab11 vesicles, fluorescent Transferrin did not accumulate in suf/spastizin oocytes ( Figure 3D). More interestingly, we found rare overlap of Transferrin with Rab11 suggesting that the Rab11b antibody and Transferrin label different compartments in zebrafish oocytes.
To analyze whether Suf/Spastizin directly controls the trafficking of endosomes or whether the accumulation of Rab11b-positive vesicles is a secondary defect, we investigated Suf/Spastizin and Rab colocalization in the oocyte. Although early endosomes did not show a defect in suf/spastizin mutants, we also discovered colocalization of Suf/Spastizin with Rab5 as shown in the mouse [22], but not with Rab7 ( Figures 3E, S4A and S4B). Although Rab11b and Suf/Spastizin mostly overlapped in their localization, some Rab11b vesicles were negative for Suf/Spastizin ( Figures 3E  and S4C). This finding suggests that Suf/Spastizin might not be involved in all processes regulated by Rab11b in the zebrafish oocyte. In summary, the loss of Suf leads to an accumulation of Rab11b positive vesicles, which in zebrafish oocytes are not involved in Transferrin recycling.
In eukaryotes, recycling vesicles and Rab11 are involved in cytokinesis during mitosis [reviewed in 57, [58][59][60]. Consistently, after RNAi depletion of the Suf-homolog FYVE-Cent in HeLa cells, a cytokinesis defect was observed [21]. To analyze cytokinesis in embryonic cells with defective maternal Suf protein, we examined embryos from suf mutant mothers. We labeled the cell cortex of 32-cell embryos with Phalloidin and their nuclei with DAPI ( Figure 4F). Whereas embryos from wt mothers showed one nucleus per cell, age-matched embryos from suf mutants exhibited multinucleated cells. This result suggests that the maternally controlled cell cycles require Suf/Spastizin similar to the cytokinesis defect discovered in HeLa cells. However, these results did not conclusively provide evidence for a role of Suf in endosomal recycling of zebrafish oocytes.
In oocytes of many organisms including humans, secretory vesicles are also designated cortical granules [reviewed in 69,70]. They are most similar to large, dense-core vesicles found in secretory cells in humans, e.g. neurons or pancreatic b-cells [71][72][73]. The most prominent cargoes of cortical granules are carbohydrates, which after secretion increase their volume by hydration leading to chorion elevation and thereby create the perivitelline space between oocyte and chorion. As in mammals, cortical granules are secreted after fertilization during a process termed ''cortical reaction'' and the induced chorion elevation is important to inhibit lethal polyspermy and mechanical damage to the embryo [reviewed in 69,70].
To determine in fish oocytes, whether Rab11b marks secretory granules in fish oocytes, we double-labeled them with the cortical granule marker MPA-lectin [74]. Rab11b colocalized with MPAlectin identifying the Rab11 vesicles of the zebrafish oocyte as secretory, cortical granules ( Figure 5A, A9). Moreover, Suf/ Spastizin protein also colocalized with MPA on cortical granules ( Figure 5B, B9). Since not all MPA-vesicles were positive for Rab11b or Suf, we hypothesized that colabeling is only observed at the vesicle surface. In contrast, if the optical section is more central in the vesicle, the green Rab11b or Suf-signal was encompassing the luminal MPA-positive cargo. Since Suf was highly enriched in the cortex and hence, colocalization might be caused by protein abundance, we analyzed optical sections in deeper layers of the oocyte with less Suf signal ( Figure 5C-E). Indeed, Rab11b and Suf mostly localized outside the granulelumen ( Figure 5C, D). Higher magnifications showed that Rab11 in general stains the entire surface of the vesicle ( Figure 5E, upper row), whereas Suf was also found to be restricted to extraluminal microdomains ( Figure 5E, lower row). Taken together, the colocalization of Suf with Rab11b and MPA-lectin supports a role of Suf/Spastizin in the formation of cortical granules during zebrafish oogenesis.

Souffle/Spastizin controls cortical granule maturation during oogenesis
To analyze the role of Suf/Spastizin in cortical granule formation during oogenesis, we compared wt and mutant oocytes using electron microscopy. More cortical granules were visible in suf/spastizin oocytes ( Figures 6A and S5A). Interestingly, their electron dense core was not visible, which is a remarkably strong phenotype compared to other factors involved in dense-core vesicle maturation [75]. We also noted more and smaller lysosomes consistent with the results of the lysotracker staining. To verify an increase in cortical granules, we stained oocytes with MPA-lectin [74]. Wild-type zebrafish oocytes showed a few, large cortical granules in the cytoplasm and numerous, smaller vesicles at the cortex ( Figures 6B and S5B). Conversely, mutant oocytes appeared filled with MPA-lectin indicating that Suf/Spastizin is involved in cortical granule formation during oogenesis.
The observed additional granules in zebrafish mutants could be generated by two alternative mechanisms. Either Suf/Spastizin could suppress the formation of granules, which in the mutant leads to an increase of mature, fusogenic vesicles. Alternatively, Suf/Spastizin could regulate the sorting of the MPA epitope, which in the mutant oocyte leads to an increased number of MPApositive vesicles. At egg activation, cortical granules released their contents during exocytosis leading to chorion elevation ( Figure 6C) [76]. Therefore, the first alternative would lead to a stronger chorion-elevation after exocytosis. However, after triggering exocytosis in wt and suf/spastizin stage V eggs by H 2 O exposure, we observed no chorion elevation 10 min after egg activation in mutants ( Figure 1B). Furthermore, in 30 mpf (min post fertilization) embryos from mutant mothers the chorion was weakly elevated (Figure 6E), although the regular exocytosis process is completed within six minutes after sperm entry [77]. We quantified chorion and embryo diameters to exclude growth defects during oogenesis as a cause for the size differences.
Embryo size in wt and mutants were similar, while chorion diameters were reduced in embryos from mutant mothers, which was also apparent in eggs activated without sperm, supporting the idea that chorion elevation is impaired ( Figure  S5C-G).  These results indicate that Suf/Spastizin controls the sorting of MPA-positive cargo, whose failure leads to the accumulation of immature vesicles in the mutant egg. This hypothesis predicts that immature cortical granules in the mutant egg do not secrete their cargo. Indeed, after activation, mutant eggs still contained numerous cortical granules, whereas wt lost their vesicles close to the cortex with a few left in the inner cytoplasm (compare Figures 6B and 6D).
To directly visualize the fusion of the vesicles, we analyzed the kinetics of exocytosis by labeling cortical actin [74]: 60 s after activation fusing cortical granules generate negatively stained crypts in the wt actin meshwork, whereas suf mutants show no vesicle exocytosis ( Figure 6F); and 180 s after activation the collapsing crypts form scars of accumulating f-actin in wt oocytes, while the cortex of suf mutants did not change. These experiments demonstrate that cortical granules of suf mutants are not fusogenic and thus, support the hypothesis that their maturation is controlled by Suf.
In cell culture, immature secretory granules remove specific SNARE proteins from their cytoplasmic surface to acquire the competence to fuse with the plasma membrane [78][79][80]. For instance, depletion of GGA3 in neuroendocrine cells inhibits sorting of VAMP4 away from neuroendocrine secretory granules, which needs to be removed to permit vesicle fusion with the membrane [81]. When we analyzed VAMP4 in zebrafish oocytes, suf/spastizin mutants accumulated this SNARE protein compared to wt ( Figure 6G). These results show that in zebrafish oocytes Suf/Spastizin is essential for the maturation of fusion-competent cortical granules.
Since chorion elevation provided a sensitive read-out for Suf functionality in zebrafish oocytes, we determined whether wt suf rescues the mutant phenotype. However, the length of the suf gene with almost 8 kb made it problematic to obtain sufficient in vitro transcribed RNA for injection and hence, we injected plasmid DNA into oocytes [82][83][84]. Furthermore, we used stage III oocytes, since they can be incubated for longer periods to allow for protein expression than matured stage V eggs. Moreover, zebrafish stage III oocytes also show spontaneous chorion expansion [82]. After 12 h, the majority of wt oocytes elevated their chorion (84.3%66.9), whereas suf mutant oocytes rarely showed a perivitelline space (7.3%62.9) (Figures 6H and S6A) also after injection of control plasmid ( Figure S6B). Wt Suf plasmid partially rescued chorion expansion in mutant oocytes (58.5%619.0), but the chorion was not elevated to the same extend as in wt (Figures 6I and S6A). Interestingly, when we overexpressed the p96re allele, we also observed partial rescue, but at a lower rate (40.0%613.4) (Figures 6I and S6A). To confirm the rescue with molecular markers, we stained injected oocytes with MPA-lectin and VAMP4. Both markers were reduced in mutant oocytes after injection of plasmid encoding wt Suf or Suf p96re , but not after injection of control plasmid ( Figure S6B). This result confirms our previous hypothesis that the zebrafish p96re allele encodes a hypomorph with reduced activity. Taken together these data demonstrate that Suf controls secretory vesicle maturation probably via sorting during zebrafish oogenesis.

Suf/Spastizin regulates vesicle fission in zebrafish oocytes
In cell culture experiments, VAMP4 is removed from immature secretory granules after it is sorted into Clathrincoated buds, which finally pinch off [78,85,86]. The localization of Suf to compartmental microdomains also supports a role for Suf in sorting. Consistent with our results, tissue culture experiments previously proposed a role for Suf in sorting [27], which is also a prerequisite for vesicle abscission. To analyze in zebrafish oocytes whether suf/spastizin mutants show a defect in vesicle abscission, we returned to the EM analysis and investigated mutants at higher magnification. Interestingly, close to the cortex of mutant oocytes we discovered compartments covered with Clathrin-coated buds, which appeared to not complete budding and abscission (Figures 7A (white arrowheads) and S7A). In wt oocytes these prominent compartments with Clathrin-buds were never observed suggesting that Suf/Spastizin controls a step such as sorting to initiate vesicle budding and fission. To confirm that Clathrin-coated buds accumulate, we compared the localization of Clathrin in wt and mutant oocytes. Clathrin accumulated in mutants corroborating the EM data ( Figures 7B, D and S7B, D). The compartments accumulating Clathrin in mutant oocytes looked similar to cisternae formed in the temperature-sensitive Dynamin mutant shibire in Drosophila [87]. Dynamin exerts the ultimate step after sorting and budding during the fission process [88][89][90]. Indeed, we observed an accumulation of Dynamin on cortical granules in zebrafish suf/spastizin mutants suggesting that Suf/Spastizin controls a molecular step, which permits Dynamin mediated fission in oocytes ( Figures 7C, D and S7C, D). If fission by Dynamin is essential for cortical granule maturation, treatment of zebrafish oocytes with the Dynamin-specific inhibitor Dynasore should mimic the mutant phenotype as long as Suf is required genetically upstream of vesicle fission [91]. Interestingly, wt oocytes accumulated the cortical granule marker MPA similar to mutants after Dynasore treatment (Figures 7E, F and S8A-F). However, we also noted differences such as mature cortical granules in Dynasore-treated wt oocytes, which probably formed during oogenesis before the drug treatment ( Figure  S8F). Although Dynamin is involved in additional processes as revealed by the Dynasore inhibition, these data confirm that Dynamin mediated fission is required during maturation of secretory granules in the zebrafish egg.
Cortical granule formation was previously considered to be an ongoing process during oogenesis, but whether it continued after ovulation remained unclear [70]. To analyze whether cortical granule maturation still continues in ovulated, fully matured stage V eggs, we treated them for 5, 15 and 30 min with Dynasore and examined chorion elevation. Indeed, the level of chorion elevation corresponded well with the duration of Dynasore treatment (Figures 7G and S8G). Remarkably, the 30 min treatment with Dynasore inhibited chorion elevation completely. Most importantly, Dynasore also reverted the transparency of the egg cytoplasm back to opaqueness similar to suf mutant eggs, which we confirmed by MPA-Lectin and VAMP4 staining ( Figures 7G  and S8H). The transparency of the egg cytoplasm is caused by whereas the cytoplasm stays opaque in suf/spastizin oocytes. Scale bar: 100 mm (D) suf/spastizin oocytes form a polar body. In vitro matured oocytes from wt and mutants stained with DAPI (blue) to highlight the nuclei and f-actin (green) to mark the polar body. Scale bar: 25 mm. (E) suf/spastizin oocytes arrest in metaphase II. Confocal images of b-tubulin (green) highlight the spindle in wt and mutant oocytes, which is located in the lower left corner, whereas the polar body is in the upper right corner. Scale bar: 25 mm. (F) Suf/Spastizin is required for cytokinesis. Embryos at 32-cell stage from heterozygous wt (top panel) or homozygous (bottom panel) suf/spastizin mothers with labeled nuclei (DAPI; blue) and plasma membranes (Phalloidin; green). In contrast to wt, few eggs from mutant mothers initiate cell division after fertilization, but then show cells with multiple nuclei (arrowhead). doi:10.1371/journal.pgen.1004449.g004 changes in the crystal structure of yolk globules, which are considered dormant lysosomes [92][93][94]. Theses intriguing results suggest that cortical granule maturation is critical for the function of lysosomal yolk globules during all stages of oogenesis. In summary, our results show that Suf/Spastizin is a key regulator of secretory vesicle maturation during zebrafish oogenesis probably before the fission of Clathrin-coated buds through Dynamin ( Figure 7H).

Discussion
Here we take advantage of zebrafish genetics and its oocytes with their high vesicle trafficking activity to describe a novel role for the SPASTIZIN homolog Souffle. We show that during zebrafish oogenesis Suf/Spastizin is essential for the maturation of secretory granules. Suf/Spastizin colocalizes with Rab11b to an intermediate compartment during the formation of dense-core vesicles. During this process, Suf/Spastizin is necessary for fission from immature secretory granules in zebrafish oocytes.
A striking phenotype in suf oocytes is the accumulation of Rab11b positive vesicles. However, in fibroblasts it is unusual that Transferrin does not colocalize with Rab11b and consistently, none of the other phenotypes besides the defect in cytokinesis supports a role for Suf/Spastizin in endosomal recycling. Interestingly, in certain human cell types, secretory cargoes also pass through Rab11-positive endosomes [reviewed in 61]. Moreover, polarized tissue culture cells spatially and functionally separate Rab11a/Transferrin and Rab11b [68,95]. This remarkable separation of Rab11a and 211b is mostly observed in polarized epithelial cells, which form the Rab11-positive subapical compartment (SAC) or common endosome as a central sorting hub for recycling [reviewed in 96]. Although the zebrafish oocyte is highly polarized along the animal-vegetal axis, polarity at the level of vesicle transport has not been previously described. Therefore, the Rab11b positive compartment in the zebrafish oocyte requires further analysis to confirm its homology to the subapical compartment of human epithelial cells.
Another study performed in rat neuroendocrine PC12 cells implicated Rab11b in the formation of dense-core vesicles [64]. This report showed that Rab11b, but not Rab11a and Rab25 of the Rab11 protein family, are involved in secretion. Moreover, Rab11b was suggested to control the sorting of secretory cargo in neuroendocrine cells. This result in rat tissue culture also provides a molecular mechanism for Suf/Spastizin function most consistent with our results in the zebrafish oocyte. Interestingly, when they repeated their experiments in unpolarized human fibroblasts, e.g. HeLa cells, Rab11b showed different effects [64]. This cell-type dependent role of Rab11b in mammals might also explain why the Spastizin homolog was not implicated in regulated secretion before our study, since it was mostly analyzed in unpolarized human fibroblasts [21,26,27]. It remains to be determined to which degree our results in zebrafish oocytes are comparable to those in mammalian tissue culture experiments.
The data from human fibroblasts showed that Suf/Spastizin interacts with the novel AP5 complex during the formation of multivesicular-bodies (MVB) and that knock-down of AP5 generates empty MVBs [26,27]. Consistently, the mouse k.o. demonstrates genetically that Spastizin is critical in vivo during this process, which leads to the formation of lysosomes. In zebrafish, we also detected fragmented lysosomes in suf/spastizin mutant oocytes. Moreover, mutations in AP5 and SPASTIZIN cause HSP in humans confirming genetically that both proteins are involved in the same process [20,24]. In addition, Suf/Spastizin also binds to Kif13a [21], which was implicated by tissue culture overexpression experiments to regulate secretion [97]. Since all these studies are carried out in different organisms and different tissues, it is difficult to extract the precise function of Suf/Spastizin.
A major question arising from our study is how the various observed phenotypes could be reconciled. Two alternative scenarios integrate all results: Suf/Spastizin primarily acts in MVB/lysosomes and the maturation defect of secretory vesicles is secondary e.g. through retrograde transport to the Trans-Golgi-Network [reviewed in 98] or alternatively, Suf/Spastizin primarily acts in immature secretory granules and the lysosomal defect is secondary e.g. through sorting of the mannose-6-phosphate receptors transporting hydrolytic enzymes into lysosomes [reviewed in 71,99]. Currently, we favor the second model, which is supported by proteins such as AP-3, which play a role in dense core vesicle formation [100] and lysosomal maturation [reviewed in 101]. By contrast, defects in retrograde transport were hitherto not reported to affect the formation of the dense core in secretory granules as in the suf mutant, but rather lead to missorting of lysosomal enzymes [reviewed in 98]. This model would also predict that cargo sorted away in immature secretory granules is necessary for homotypic lysosome fusion ( Figure 7H) as described in other cells [102]. These open questions make clear that additional studies are necessary to determine the precise role of Suf/Spastizin and whether a defect in lysosome formation or secretion can lead to neuronal degeneration in HSP patients.
In humans, Suf/Spastizin encodes one of the more than 50 loci involved in the neurodegenerative disorder HSP controlling diverse cellular processes [18]. However, regulated secretion was so far not considered. Interestingly, defects in regulated secretion as observed in the zebrafish oocyte may explain some of the HSP symptoms. Cortical granules in the zebrafish oocyte appear identical to large dense-core vesicles (DCV) in neurons [71][72][73]103], which store neurotrophic factors at the synapse [104]. Neurotrophic factors are responsible for the dynamics and maintenance of synaptic connections during long-term potentiation [105,106]. In contrast to neurotransmitter vesicles, DCVs need to be transported from the cell body to the synapse. This transport model would resolve why in HSP preferentially the longest axons degenerate from their synapse. In addition, a defect in long-term potentiation would explain why HSP neurons are not maintained and the symptoms become apparent in juveniles and adults, but not in embryos. Investigating the molecular network regulated by Suf/Spastizin in zebrafish oocytes as well as transferring these results to the nervous system as described for Atlastin (SPG3A) in zebrafish [42] and Drosophila [107] could provide novel insights into the biochemical etiology of HSP and bring us closer to a therapeutic treatment for patients.

Ethics statement
Fish were maintained as described [108] in accordance with regulations of the Georg-August University Goettingen, Germany.

Oocyte and embryo methods
Oocytes were dissected and cultured as previously described [83] and their stage indicated with Roman numerals as published [11]. For in vitro maturation and plasmid injection, stage III oocytes were manually dissected and incubated for 1 hr in 60% L-15 medium. After removing damaged oocytes, meiosis resumption was induced by a 90 min-exposure to 1 mg/ml DHP or EtOH as carrier control.
For the trafficking assay, stage III oocytes were incubated with 125 mg/ml Transferrin-Alexa594 (Molecular Probes) in OR2 buffer [108] for 10 and 25 min followed by 30 min of chasing in Transferrin-free OR2 buffer (recycling) or with 10 mg/ml of LDL Dil (Molecular Probes) in OR2 buffer for 10 min (degradation). Then, oocytes were fixed with MEMFA (1 M MOPS pH 7.4, 20 mM EGTA, 10 mM MgSO 4 , 3.7% formaldehyde) after washing twice with OR2 and twice with PBT and stained with antibodies or fluorescent dyes.
For the chorion elevation assay, ovulated stage V eggs were squeezed from gravid females. Eggs were activated by adding E3medium [108] and imaged at 30 min [108] after activation. Images were used to measure chorion elevation using Fiji software [109].
For Dynasore treatment, live oocytes were collected in OR2 buffer and incubated with 500 mg/ml of Dynasore (Sigma) or the carrier DMSO as control for 1 hr at room temperature. Then, oocytes were washed thrice and processed for immunofluorescence staining.

Quantitative Real-Time PCR
qRT-PCR on selected stages of oogenesis and embryogenesis was performed as previously described [83].

Meiotic mapping and genotyping
The described mapping position of the suf mutation on chromosome 13 [28] was used for fine mapping as previously published for the bucky ball locus [83].

Phylogenetic analysis
Alignments were performed with CulstalW and the similarity was quantified with vector NTI (Invitrogen). The phylogenic tree was constructed as described [83] with 1000 iterations.

Immunofluorescence
Oocytes were dissected from gravid female and fixed with MEMFA buffer for 1 h at room temperature after proteinase K (50 mg/ml) treatment for 3 min. Oocytes were washed thrice with PBT and blocked with PBT containing 2% BSA and 2% Horse serum for 2 h at room temperature. Subsequently, oocytes were incubated with primary antibody: Rab5 For double labeling, oocytes were incubated with Suf-antibody directly labeled with ATTO590 (SySy, Goettingen, Germany) for 2 h at room temperature. The Suf antibody was generated by immunizing two rabbits each with two peptides TEQVKVPAKDRNRE (aa 187-200) and LNKTSTNKGMSKTD (aa 1007-1020) and then purified by affinity-chromatography (Biogenes, Berlin Germany). DNA was stained with 1 mM of Hoechst 33342. Cortical granules were stained with 50 mg/ml MPA Lectin Texas Red (EY Labs Inc.) and lysosomes with 70 nM Lysotracker DND-99 Red (Molecular Probes). Phalloidin staining was performed as described [74].

Confocal microscopy
Images were captured at room temperature using a LSM780 confocal microscope (Carl Zeiss) with a Plan Apochromat 636/ 1.4 NA and 256/0.8 NA oil-immersion and a digital microscope camera (M27; Carl Zeiss). After washing the oocytes with PBT, yolk was cleared with Murray's solution (Benzylbenzoate (66%)/ Benzylalcohol (33%)) during imaging. A multiple wavelength laser was used to visualize red (561) fluorescence, green (488, 405) and blue (405) and images were acquired and processed using ZEN 2011 software (Carl Zeiss).

Electron microscopy
For high-pressure freezing EM, living oocytes were placed in aluminum platelets of 150 mm depth containing 1-hexadecen [110]. The platelets were frozen using a Leica Em HPM100 highpressure freezer (Leica Mikrosysteme Vertrieb GmbH, Wetzlar, Germany). The frozen oocytes were transferred to an automatic Freeze Substitution Unit Leica EM AFS2. The samples were substituted at 290uC in a solution containing anhydrous acetone, 0.1% tannic acid for 24 h and in anhydrous acetone, 2% OsO 4 , 0.5% anhydrous glutaraldehyde (EMS Electron Microscopical Science, Ft. Washington, USA) for additional 8 h. After a further incubation over 20 h at 220uC, samples were warmed up to 4uC and washed with anhydrous acetone subsequently. The samples were embedded at room temperature in Agar 100 (Epon 812 equivalent) at 60uC over 24 h. Images were taken in a Philips CM120 electron microscope (Philips Inc.) using a TemCam 224A slow scan CCD camera (TVIPS, Gauting, Germany).