Mutations in Cog7 affect Golgi structure, meiotic cytokinesis and sperm development during Drosophila spermatogenesis

The conserved oligomeric Golgi (COG) Complex is required for Golgi integrity, vesicle trafficking and glycosylation in yeast and mammalian cells (Smith and Lupashin, 2008; Ungar et al., 2006). Based on yeast genetics and electron microscopy, the COG complex was proposed to have a bi-lobed structure, with eight subunits (Cog1-8) arranged in two distinct subcomplexes: Lobe A (Cog1-4) and Lobe B (Cog5-8) (Loh and Hong, 2004; Ram et al., 2002; Ungar et al., 2002; Walter et al., 1998; Whyte and Munro, 2001). In yeast, subunits of Lobe A are essential components of the complex (VanRheenen et al., 1998; Whyte and Munro, 2001; Wuestehube et al., 1996), whereas Lobe B subunits are not substantially required for cell growth or internal membrane organisation (Ram et al., 2002; Whyte and Munro, 2001). Mutations in the genes encoding human COG1, COG4-COG8 have been associated with congenital disorders of glycosylation (CDG) (Foulquier et al., 2006 Foulquier et al., 2007; Kranz et al., 2007; Lübbehusen et al., 2010; Ng et al., 2007; Paesold-Burda et al., 2009; Reynders et al., 2009; Spaapen et al., 2005; Steet and Kornfeld, 2006; Wu et al., 2004) indicating a role for COG in the transport and/or stability of Golgi glycosylation enzymes. Indeed studies in both yeast and mammalian cells have suggested that COG complex might function as a vesicle-tethering factor in intra-Golgi retrograde COPI transport (Ungar et al., 2002), thus regulating the distribution and the stability of Golgi resident proteins (Oka et al., 2004; Shestakova et al., 2006; Suvorova et al., 2001; Suvorova et al., 2002; Walter et al., 1998). A set of Golgi proteins called GEARs including giantin matrix proteins and glycosyltransferases/glycosidases were shown to mislocalize and to be abnormally degraded in Chinese hamster ovary (CHO) cells mutant for either Cog1 or Cog2 subunits (Oka et al., 2004). Depletion of Cog3 in HeLa cells resulted in the accumulation of COG complex-dependent (CCD) vesicles carrying Golgi v-SNARE proteins but no anterograde cargo molecules (Zolov and Lupashin, 2005). Consistent with a role in Golgi trafficking, COG subunits interact with intra-Golgi SNAREs, the COPI coat, small GTPases and other tethering proteins (Laufman et al., 2011; Sohda et al., 2007; Sohda et al., 2010; Suvorova et al., 2002).

Drosophila spermatogenesis provides a well suited model to study the requirement of Golgi trafficking in vivo, in a multicellular organism. Mutations in vesicle trafficking cause defects in various aspects of sperm development. For example, Drosophila spermatocytes are quite large cells (more than 20 µm in diameter), that complete two meiotic divisions in less than two hours (Fuller, 1993; Giansanti et al., 2012). Spermatocyte cytokinesis proved to be sensitive to mutations affecting vesicle traffic components (Brill et al., 2000; Farkas et al., 2003; Dyer et al., 2007; Gatt and Glover, 2006; Giansanti et al., 2006; Giansanti et al., 2007; Polevoy et al., 2009; Robinett et al., 2009; Xu et al., 2002a; Zhou et al., 2011). During Drosophila spermiogenesis, round spermatids, ∼12 µm in diameter, undergo drastic morphological changes in cell surface before reaching the impressive 1.8 mm in length of sperm tails (Fuller, 1993; Tates, 1971; Zhou et al., 2011). The 100-fold increase in cell length that characterises spermatid elongation is particularly dependent on membrane biosinthesis and membrane remodelling (Fabian et al., 2010; Farkas et al., 2003, Wei et al., 2008; Zhou et al., 2011). Among the COG subunits, the Drosophila Cog5 homologue Four way stop (Fws) has been implicated in both male meiotic cytokinesis and spermatid elongation (Farkas et al., 2003).

Here we have characterised the Drosophila orthologue of human Cog7. Loss of human COG7 is responsible for a rare, congenital disorder of glycosylation (CDG), namely CDG IIe (Morava et al., 2007; Ng et al., 2007; Wu et al., 2004). We show that mutants carrying null alleles of Drosophila Cog7 are viable but male sterile. Cog7 function is essential to maintain Golgi structure and to regulate axoneme architecture during sperm development. Similar to Fws, Cog7 also controls furrow ingression during cytokinesis. Importantly, Cog7 is required to localise the small GTPase Rab11 and the phosphatidylinositol transfer protein (PITP) Giotto (Gio) to the cleavage site of spermatocytes. In addition Gio coimmunoprecipitates with both Cog7 and Rab11 in Drosophila testes suggesting that these proteins may interact in male germ cells.

Three mutant alleles of Drosophila Cog7 were isolated by a cytological screen of the Zuker collection of male sterile lines (see Materials and Methods). Males homozygous for each allele and transheterozygotes carrying the allelic combinations Cog7^z-2805/Cog7^z-4495, Cog7^z-2805/Cog7^z-5797, Cog7^z-5797/Cog7^z-4495 are viable but male sterile. Mutant testes displayed frequent multinucleate spermatids suggesting failures in meiotic cytokinesis (Fig. 1A,B). Testes from Cog7 males also exhibited a failure in spermatid elongation, a phenotypic trait previously observed in mutants in the Drosophila Cog5 homologue Fws (Fig. 1C) (Farkas et al., 2003).

All the mutations were uncovered by both Df(3R)Exel6213 and Df(3R)BSC861. DNA sequencing revealed that each of the three mutant alleles carried nonsense mutations in the CG31040 gene (Fig. 1D) encoding a 742 amino acid polypeptide that is homologous to human Cog7 (Ungar et al., 2002).

To determine the localisation of Cog7 in testes we generated transgenic flies expressing a GFP-Cog7 fusion protein. The GFP-Cog7 transgene rescued both male sterility and the phenotypic defects associated with Cog7 mutations indicating that the encoded protein is fully functional. In prophase I, GFP-Cog7 was enriched at multiple round structures that colocalized with the Golgi marker Lava Lamp (Lva, Sisson et al., 2000) indicating that Cog7 localises to the Golgi stacks (Fig. 2A). During metaphase and anaphase of meiosis I, GFP-Cog7 was associated with Golgi organelles in the polar regions of the cell (Fig. 2B,C). In telophase cells, Cog7 signals appeared excluded from the cell equator just like Lva-Golgi organelles (Fig. 2C,D; supplementary material Fig. S1). In onion-stage spermatids, Cog7 localised to Golgi derived acroblasts (Fig. 2E).

The subcellular localisation of Cog7 in spermatocytes is in accordance with the localisation of the Fws/Cog5 protein (Farkas et al., 2003) (Fig. 2F). To determine the functional dependence between Fws/Cog5 and Cog7, Cog7 mutants expressing GFP-Fws and fws mutants expressing GFP-Cog7 were examined. Cog7 mutations completely abolished the localisation of Fws at the Golgi and Cog7 was diffuse in the cytoplasm in the absence of Fws (Fig. 2F).

Since Cog7 and Fws/Cog5 are believed to reside in the same COG subcomplex (Ungar et al., 2002), we assessed whether the stability of either protein was affected in the reciprocal mutant (Fig. 2G,H). Western blots showed that GFP-Cog5 protein was decreased by 56% in testes from Cog7 mutants relative to wild type (Fig. 2G) and GFP-Cog7 protein was decreased by 20% in testes from fws mutants relative to wild type (Fig. 2H).

In order to verify whether other Drosophila putative COG proteins were also essential for male meiotic cytokinesis we used RNA interference (RNAi) induced by double stranded RNA (Dietzl et al., 2007). Cog3 depletion resulted in frequent multinucleate spermatids suggesting that other COG components are required for spermatocyte cytokinesis (supplementary material Fig. S2).

Because in mammalian cells, defects in COG subunits disturb Golgi architecture (Oka et al., 2005; Reynders et al., 2009; Ungar et al., 2002), we asked whether Cog7 is essential for maintaining Golgi integrity in Drosophila spermatocytes. Mature wild-type spermatocytes stained for Lva displayed multiple Golgi stacks of similar size (Fig. 3A,B). Conversely, in spermatocytes from Cog7 males, most Lva signals appeared as small puncta (Fig. 3A) and the spherical structures were less numerous than in control cells (Fig. 3B). Similarly, in wild-type spermatocytes the Golgi Mannosidase II (GMII, Rabouille et al., 1999) concentrated in the Golgi stacks, whereas in Cog7 the majority of this protein appeared associated with small puncta (Fig. 3C).

In wild-type mature spermatocytes analysed by transmission electron microscopy (TEM), the Golgi complex consisted of distinct clusters of membrane-bound compartments (Fig. 3D; n = 42 cells analysed). Each Golgi stack was made of three to five flat disc-shaped cisternae, which were associated to abundant tubular-reticular networks and vesicles of various size. The edges of the disc-shaped cisternae bulged out in small vesicles and tubules. In Cog7 spermatocytes the morphology of Golgi stacks was severely disrupted and Golgi consisted of clusters of irregularly sized and swollen cisternae (Fig. 3E,F; n = 57 cells analysed).

Mutations in fws caused defects in F-actin ring constriction of telophase spermatocytes (Farkas et al., 2003). Staining of Cog7 spermatocytes for tubulin and F-actin revealed a cytokinetic phenotype similar to fws (Fig. 4A). In 80% of mid-late telophases I from Cog7^z3-4495/Df(3R)BSC861 (Cog7) mutants, the F-actin rings failed to constrict to completion and the central spindles appeared less dense than in wild type (n = 25 for wild type; n = 30 for Cog7; Fig. 4A). In order to further substantiate the phenotype of Cog7, we examined the localisation of Myosin II. Loss of function of Cog7 also resulted in failure to constrict the cortical ring of Myosin II complex proteins in telophase spermatocytes visualised by the Drosophila Myosin regulatory light chain Spaghetti Squash tagged with GFP (Sqh-GFP, Royou et al., 2002), or by anti Myosin II antibodies (Myo II, Royou et al., 2002). 74% of late telophases from Cog7 males displayed defective Sqh-GFP rings (Fig. 4B; n = 36 for wild type, n = 34 for Cog7). Likewise, 79% of Myo II rings were large or broken during late telophase (Fig. 4C; n = 60 for wild type, n = 68 for Cog7).

COG subunit deficiencies are associated with defects in processing carbohydrate chains of both glycoproteins and glycolipids resulting in altered cell surface glycoconjugates (Kingsley et al., 1986; Oka et al., 2005; Peanne et al., 2011; Reddy and Krieger, 1989; Ungar et al., 2006; Wu et al., 2004). We thus asked whether mutations in Cog7 affected the pattern of glycoconjugates at the cleavage furrow.

Previous studies in sea urchin eggs revealed a unique equatorial membrane domain enriched in monosialotetrahexosylganglioside (G_M1), cholesterol and activated tyrosine kinases forms during late anaphase (Ng et al., 2005). By staining with fluorescent cholera toxin subunit B (CTB), we found that G_M1 accumulates in the furrow region of wild-type spermatocytes in late anaphase and telophase (Fig. 5A, n = 98). Staining of Cog7 spermatocytes failed to detect an accumulation of G_M1 in 87% of late anaphase/telophase cells (Fig. 5A; n = 114).

Glycoproteins with terminal sialic or N-acetylglucosamine residues concentrate in the cleavage furrow and can be visualised by staining with fluorescent wheat germ agglutinin (WGA) (Yoshigaki, 1997; Ng et al., 2005). Fluorescent WGA also stains the equatorial plasma membrane of Drosophila wild-type spermatocytes during late anaphase/telophase indicating an enrichment of sialic or N-acetylglucosamine glycoconjugates (Fig. 5B) (Giansanti et al., 2007) but failed to detect an accumulation of WGA receptors in all the Cog7 cells examined (Fig. 5B, n = 80 for wild type, n = 84 for Cog7).

We examined whether loss of Cog7 affected the localisation of membrane traffic components involved in cytokinesis. Previous data showed that the phosphatidylinositol transfer protein (PITP) Giotto/Vib (Gio/Vib) and the small GTPase Rab11 accumulate at the cleavage site and are both required for spermatocyte cytokinesis (Gatt and Glover, 2006; Giansanti et al., 2006; Giansanti et al., 2007; Polevoy et al., 2009). It was suggested that Rab11, Gio and the Phosphatidylinositol 4-kinase β (PI4Kβ) Fwd function in the same pathway during cytokinesis, with Gio and Fwd acting upstream of Rab11 (Giansanti et al., 2007; Polevoy et al., 2009). Our immunofluorescence analysis revealed that loss of Cog7 impaired the recruitment of Gio at the cell equator in 89% of telophase spermatocytes (Fig. 6A; n = 32 for wild type; n = 36 for Cog7). Cog7 mutations also disrupted the localisation of Rab11 in dividing spermatocytes. In both wild-type and Cog7 prophase spermatocytes expressing Rab11-GFP, Rab11 was enriched at round structures corresponding to Lva-enriched Golgi stacks (supplementary material Fig. S3). Rab11-GFP colocalized with Lva in both wild-type and Cog7 prophase cells. However, Cog7 mutant spermatocytes displayed few Golgi organelles relative to wild-type control cells. In addition, Rab11 did not accumulate at the cell midzone in 90% of mutant telophases (Fig. 6B,C; n = 26 for wild type; n = 30 for Cog7). In contrast, the analysis of GFP-Cog7 distribution in gio and Rab11 mutant spermatocytes revealed a normal localisation of this protein in both prophase and telophase (supplementary material Fig. S4). Since localisation of Rab11 to the cleavage site is dependent on both Gio (Giansanti et al., 2007) and Cog7 (this study) we performed Co-IP experiments using Drosophila testis protein extracts aimed at verifying interactions among these proteins. Protein extracts from testes expressing either GFP-Cog7 or Rab11-GFP were immunoprecipitated using the GFP-Trap (see Materials and Methods). We found that Gio coimmunoprecipitated with both Cog7 and Rab11 in Drosophila testes (Fig. 6D). Gio/Cog7 and Gio/Rab11 interactions were confirmed by a second experiment in which extracts from testes expressing Gio-RFP and either Rab11-GFP or GFP-Cog7 were immunoprecipitated using the RFP-Trap (Fig. 6E,F).

Because Gio coimmunoprecipitated with Cog7 in testis extracts, we asked whether the two proteins colocalized in dividing spermatocytes. To test this, we imaged spermatocytes expressing both GFP-Cog7 and Gio-RFP. Our analysis revealed that Gio partially colocalized with Cog7 at the Golgi membranes (Fig. 6G). Gio also localised to parafusorial and astral membranes where Cog7 was not detectable. In fixed dividing spermatocytes coexpressing Gio-RFP with Rab11-GFP, Gio colocalized with Rab11, consistent with previous results with anti-Gio and anti-Rab11 antibodies (supplementary material Fig. S5) (Giansanti et al., 2007). During telophase Gio and Rab11 concentrated at the cleavage furrow (Fig. 6A,B; supplementary material Fig. S5), while Cog7 was excluded from the cell equator (Fig. 2; supplementary material Fig. S1). In live preparations, Gio-RFP could be clearly detected on vesicles at the polar regions and in proximity of the cell equator (supplementary material Fig. S5).

In wild-type onion-stage spermatids stained for Lva, the acroblast is a cone-shaped structure adjacent to the nucleus (Fig. 7A, n = 120). In Cog7 mutant spermatids, anti-Lva labelled numerous puncta but failed to detect a cone-shaped acroblast (Fig. 7A, n = 96). Moreover immunostaining against the Golgi SNARE Syntaxin16 (Syx16, Xu et al., 2002b) revealed that this protein is enriched in wild-type acroblasts (n = 60) but was entirely diffuse in mutant cells (Fig. 7B; n = 54). TEM analysis confirmed failure to form acroblasts (Fig. 7C,E). In wild-type onion-stage spermatids, the acroblast was found in a juxtanuclear position (Fig. 7C; n = 18 spermatids). Cog7 mutant spermatids displayed numerous vesicles, scattered throughout the cytoplasm and lacked acroblasts (Fig. 7E; n = 17 spermatids). TEM analysis revealed that loss of Cog7 also affects flagellar biogenesis. In wild-type onion-stage spermatids, the basal body was found attached to the surface of the nuclear envelope and an electron-dense adjunct formed a collar around this structure (Fig. 7D; n = 37). In Cog7 spermatids the structure of the basal body and the centriolar adjunct was not affected (Fig. 7F). However, 87% of basal bodies and centriolar adjuncts appeared displaced from the nuclear envelope, suggesting defects in basal body docking to the nuclear envelope (Fig. 7F; n = 45). Cross sections of wild-type axonemes (Fig. 7G) revealed the typical organisation of axonemal microtubules into an arrangement composed of nine outer doublets and a central pair (9+2). Mutant spermatids were often defective in axoneme architecture. In our TEM studies, 25% of mutant axonemes appeared irregular in cross section and lacked peripheral doublets and/or central microtubules (Fig. 7H; n = 97 mutant axonemes; n = 131 control axonemes). These phenotypic defects were not specific for Cog7 mutants; similar, although less frequent aberrations were also observed in gio and Rab11 mutant males (supplementary material Fig. S6). Axoneme architecture was disrupted in 18% of spermatids from gio males (n = 273) and 15% of spermatids from Rab11 males (n = 213) (supplementary material Fig. S6).

Because Cog7 null mutants are viable but male sterile we wondered whether Cog7 is required for Golgi integrity in cell types other than spermatocytes. In salivary glands and central nervous system (CNS) of larvae expressing GFP-Cog7, Cog7 localised on Golgi bodies (not shown). In wild-type interphase neuroblasts Lva concentrated in multiple Golgi bodies of similar size. In Cog7 Lva-enriched structures were less numerous and appeared smaller than in wild type (Fig. 8). Golgi was also defective in mutant salivary gland cells (supplementary material Fig. S7). Given the defects in Golgi structure of Cog7 mutants, we asked whether Cog7 protein is essential during mitotic divisions of larval neuroblasts. Analysis of larval brain squashes revealed a low frequency of tetraploidy: 3% of metaphases were tetraploid in Cog7 (n = 495 metaphases) versus 0% in wild type (n = 510). Immunostaining of larval brains for Anillin also indicated mild defects during cytokinesis; 15% of mid-telophases from Cog7 displayed poorly constricted or broken contractile rings (Fig. 8D–F). Loss of Cog7 did not appear to affect other mitotic parameters; the mitotic index and the frequency of prometaphases/metaphases relative to all mitotic figures were not significantly increased in Cog7 larval brain preparations (Fig. 8F).

Wild-type function of the Drosophila homologue of Cog7, a subunit of the conserved octameric oligomeric complex (COG), involved in membrane trafficking and glycoconjugate synthesis (Smith and Lupashin, 2008; Ungar et al., 2002; Ungar et al., 2006), is required for maintaining Golgi structure in both spermatocytes and spermatids. Studies in yeast and mammalian cells suggest a primary role for the COG complex in tethering retrograde vesicles that traffic within the Golgi (Bruinsma et al., 2004; Oka et al., 2004; Sohda et al., 2010; Suvorova et al., 2002; Zolov and Lupashin, 2005) and between endosomal compartments and the Golgi (Laufman et al., 2011; Smith et al., 2009; VanRheenen et al., 1999), thus contributing to the normal structure and function of the Golgi apparatus (Smith and Lupashin, 2008; Ungar et al., 2006). The dependence of other trafficking steps, such as endoplasmic reticulum (ER) export and ER-to-Golgi transport (VanRheenen et al., 1998; VanRheenen et al., 1999; Wu et al., 2004; Wuestehube et al., 1996) on Cog proteins may be secondary to a primary role in maintaining Golgi integrity and function.

Mutations in the genes encoding human COG1, COG4-COG8 have been associated with congenital disorders of glycosylation (CDG) (Foulquier et al., 2006; Foulquier et al., 2007; Kranz et al., 2007; Lübbehusen et al., 2010; Ng et al., 2007; Paesold-Burda et al., 2009; Reynders et al., 2009; Spaapen et al., 2005; Steet and Kornfeld, 2006; Wu et al., 2004) suggesting a role for COG in the retention and/or retrieval of Golgi glycosylation enzymes. Indeed, loss of COG subunits impaired the localisation and the stability of Golgi resident proteins, which are known to recycle within the Golgi stacks (Oka et al., 2004; Zolov and Lupashin, 2005). In mammalian cells, deficiencies in COG subunits affect Golgi integrity causing several types of structural alterations. Depletion of Cog3 in HeLa cells caused fragmentation of the Golgi ribbon (Zolov and Lupashin, 2005), Cog1 and Cog2 CHO null mutants and Cog5-deficient HeLa cells displayed cisternal dilation (Oka et al., 2005; Ungar et al., 2002), fibroblasts from Cog7 and Cog8 patients exhibited fragmented and/or vesiculated stacks, undulated appearance of stacks and swollen cisternae (Reynders et al., 2009).

It was surprising that flies homozygous for null alleles of Cog7 were viable, although males were sterile and displayed various defects during spermatogenesis, as mutations in the gene encoding human Cog7 are associated with a lethal congenital disorder of glycosylation (CDG), causing multisystemic deficiencies including neurological, metabolic and anatomical abnormalities (Morava et al., 2007; Ng et al., 2007; Wu et al., 2004).

Drosophila Cog7, along with the Cog5 homologue Fws, localised to Golgi throughout spermatogenesis. In addition Cog7 and Fws were interdependent for localisation to Golgi membranes. The organisation of Golgi into many stacks scattered throughout the cytoplasm in most Drosophila cells (Kondylis and Rabouille, 2009) allowed identification of subtle structural alterations in the Cog mutants that might have been missed in analysis of mutant mammalian cells, where Golgi stacks are interconnected into a pericentriolar ribbon. Loss of function of Drosophila Cog7 affected the integrity and the number of Golgi stacks in primary spermatocytes, as well as the structure of Golgi cisternae, consistent with previous EM studies in COG7 patient cells (Reynders et al., 2009) and in CHO cells carrying either Cog1 or Cog2 null mutations (Ungar et al., 2002; Vasile et al., 2006 ). Loss of function of Cog7, like loss of function of fws, also disrupted assembly or stability of the Golgi-based acroblasts that form after meiosis II from Golgi-derived vesicles in onion-stage spermatids (Farkas et al., 2003; Giansanti et al., 2006; Giansanti et al., 2007).

Rather than fusing to form a compact juxtanuclear acroblast, Cog7 mutant Lva-containing vesicles were scattered and the Golgi SNARE Syntaxin16 (Syx16, Xu et al., 2002b) was entirely diffuse in mutant cells. Based on these data, the defects in acroblast formation might be a consequence of a failure to recruit Syntaxin proteins to Golgi organelles at the end of meiosis, thus impairing Golgi membrane fusion. Consistent with a role for the Cog proteins in recruiting Syntaxin proteins, it has been recently demonstrated that loss of function of Cog6 in HeLA cells abrogated the steady state level of Syntaxin 6, markedly reduced localisation of Syntaxin 16 to Golgi membranes and impaired the assembly of the Syntaxin 6-Syntaxin 16-Vti1a-VAMP4 SNARE complex (Laufman et al., 2011).

Cog7 mutant males displayed defects in cell elongation in differentiating spermatids, similar to fws and syntaxin 5 mutants. Spermatid elongation, which involves a 100-fold increase in cell surface area, may necessitate the addition of new membrane derived from Golgi vesicles and so may be especially sensitive to defects in Golgi structure and/or function (Fabian et al., 2010; Farkas et al., 2003; Fuller, 1993; Tates, 1971; Wei et al., 2008; Zhou et al., 2011). In addition, loss of Cog7 affected architecture of the developing flagellar axonemes and resulted in dissociation of the basal body from the nuclear envelope, suggesting defects in the docking to the nuclear envelope. These defects were not specific for Cog7; similar although less frequent alterations were also observed in gio and Rab11 males.

Recent studies demonstrated that depletion of the phosphoinositide phosphatidylinositol 4,5 biphosphate (PIP2) results in defects in both axoneme architecture and in basal docking to the nuclear envelope (Wei et al., 2008). However, axonemal of PIP2-depleted flies also exhibited alterations not observed in Cog7 flies, such as triplets as well as doublets among outer microtubules, suggesting a role for PIP2 or a downstream second messenger in controlling the transition from basal body to axonemal microtubule arrays (Wei et al., 2008). The defects in axoneme architecture are not merely the consequence of cytokinesis failure, as other mutants defective in cytokinesis form regular axonemes (Wei et al., 2008). Further investigations will be necessary to clarify whether Cog7 defects in flagellar biogenesis are the consequence of Golgi malfunction.

Function of Cog7, like the Cog5 homologue Fws (Farkas et al., 2003), is essential for successful cytokinesis in Drosophila spermatocytes. Cytokinesis involves a rapid and massive expansion in plasma membrane surface area that strongly depends on endocytic and/or secretory membrane trafficking pathways in several cell systems (Albertson et al., 2005; Boucrot and Kirchhausen, 2007; McKay and Burgess, 2011; Neto et al., 2011; Prekeris and Gould, 2008). Previous studies have demonstrated that spermatocyte cytokinesis is sensitive to Brefeldin A (Robinett et al., 2009) and requires the wild-type function of the Golgi proteins Syntaxin 5 (Xu et al., 2002a), Fws (Farkas et al., 2003), Fwd (Polevoy et al., 2009) and Brunelleschi (Robinett et al., 2009), indicating a requirement for effective Golgi trafficking. The defective morphology of Golgi stacks in mutant spermatocytes strongly suggests that the failure of cytokinesis in Cog7 mutant spermatocytes might be a consequence of defects in Golgi trafficking and/or function.

Previous studies showed that the phosphatidylinositol transfer protein (PITP) Giotto/Vib (Gio/Vib) and the small GTPase Rab11 concentrate to the cleavage furrow and are both required for spermatocyte cytokinesis (Gatt and Glover, 2006; Giansanti et al., 2006; Giansanti et al., 2007; Polevoy et al., 2009). Rab11, Gio and the Phosphatidylinositol 4-kinase β (PI4Kβ) Fwd might function in the same pathway controlling membrane addition to the spermatocyte cleavage site, with Gio and Fwd acting upstream of Rab11 (Giansanti et al., 2007). Fwd is required for the synthesis of PI4P on Golgi membranes and for the formation of Rab11-and PI4P containing organelles at the cell equator (Polevoy et al., 2009). Because PITPs can stimulate vesicle budding from the trans-Golgi network and also provide vesiculating activity for scission of coatomer-coated vesicles in vitro (Jones et al., 1998; Ohashi et al., 1995; Simon et al., 1998) Gio might be involved in vesicle formation. We found that localisation of Gio and Rab11 to the cleavage furrow in dividing spermatocytes requires the wild-type function of Cog7, indicating that Cog7 might be an upstream component in a gio-Rab11 pathway during cytokinesis (Fig. 6I). Gio coimmunoprecipitates with both Rab11 and Cog7 in testes, suggesting that these proteins may form a complex in male germ cells. Based on our results, Cog7 might be implicated in the Gio-mediated formation of Rab11 associated secretory organelles which become enriched in the cleavage site during spermatocytes (Fig. 6H,I). However, we cannot exclude a role for Cog7 and the Cog complex in the regulation of a retrograde/anterograde traffic between Golgi membranes and ER during cytokinesis (Fig. 6I).

In addition to defects in recruitment of membrane trafficking proteins, Cog7 mutant spermatocytes also exhibited incomplete actomyosin ring constriction, as also observed in fws (Farkas et al., 2003), gio (Giansanti et al., 2004; Giansanti et al., 2006), fwd (Giansanti et al., 2004) and Rab11 (Giansanti et al., 2007) mutant spermatocytes. Although regular actomyosin rings formed during early telophase, they failed to contract during cytokinesis. Several studies have indicated that actomyosin ring assembly and remodelling during cytokinesis is strictly dependent on membrane trafficking during furrowing. For example in Dictyostelium, clathrin-minus cells fail to assemble myosin II into a functional contractile ring and mutations in syntaxin1 disrupt actin organisation during furrow formation of Drosophila embryos (Niswonger and O'Halloran, 1997; Burgess et al., 1997). Interestingly, live analysis of fluorescent tagged vesicles and F-actin movement during cytokinesis of postcellurized Drosophila embryos has suggested a model in which actin and membrane are delivered as a unit to the invaginating furrows (Albertson et al., 2008).

A possible explanation for how loss of Cog7 could affect contractile ring constriction is suggested by result that fluorescent WGA receptors concentrate at the equatorial site of telophase cells in wild type but not in Cog7 mutants, indicating a defect in the accumulation of glycoproteins with terminal sialic or N-acetylglucosamine residues. We speculate that particular glycoproteins must be enriched at the equatorial site to control the contraction of the actomyosin ring during cytokinesis. Mutations in Cog7 might affect glycosylation in the Golgi and so impair localisation of these glycoproteins at the cell surface, resulting in failure of ring constriction. CHO cells deficient for either Cog1 or Cog2 exhibit defects in Golgi-associated processing reactions for the synthesis of N-linked, O-linked carbohydrate chains of glycoproteins and also glycolipids (Kingsley et al., 1986). Consistent with a requirement for the COG complex in processing lipid-linked carbohydrate chains, an enrichment of the ganglioside G_M1 could be visualised in telophase spermatocytes stained with fluorescent CTB in wild type but not in Cog7 mutants.

Previous work in sea urchin showed that a membrane equatorial domain, enriched in G_M1 cholesterol and signalling molecules, forms during late anaphase with the contractile ring and suggested the existence of signalling platforms that control cytokinesis (Ng et al., 2005). Thus, failure to localise G_M1 might disrupt the structure of a special lipid raft domain in dividing spermatocytes and affect several aspects of cytokinesis such as the activation of signalling pathways that regulate furrowing, actin remodelling and membrane traffic.

Despite its fundamental role in spermatogenesis and spermatocyte cytokinesis, Cog7 was not essential for normal development in Drosophila. Karyotyping of mitotic chromosomes in larval brain cells from Cog7 revealed a low frequency of tetraploidy and immunostaining indicated regular furrow ingression in 85% of telophases. Analysis of Golgi in mutant neuroblasts indicated a requirement for Cog7 in Golgi integrity also in these cells. However, loss of Cog7 and the consequent Golgi defects did not substantially affect cytokinesis of neuroblasts. Other mutations in membrane traffic components such as fwd, fws, bru, gio and Arf6 disrupt spermatocyte cytokinesis causing little or no effects on cytokinesis of larval neuroblasts or S2 cells (Brill et al., 2000; Eggert et al., 2004; Farkas et al., 2003; Giansanti et al., 2004; Giansanti et al., 2006; Dyer et al., 2007; Robinett et al., 2009). The large size of spermatocytes and the rapid succession of two meiotic divisions might explain the particular requirement of male meiotic cytokinesis for vesicle trafficking pathways.

Cog^z-4495, Cog^z-5797 and Cog^z-2805 alleles of Drosophila Cog7 were obtained by screening the C. Zuker collection of male sterile lines (Wakimoto et al., 2004; Giansanti et al., 2004). Chromosomal deficiencies Df(3R)BSC861 and Df(3R)D1-BX12 were obtained from the Bloomington Drosophila Stock Center at Indiana University. Flies carrying the four way stops (fws), giotto (gio) and Rab11 mutations fws^Z0161, fws^Z1201, gio^RM1-r7, gio^z3-3934, gio^EP513, Rab11^e(To)3 and Rab11^93Bi and flies expressing GFP-Fws have been described previously (Farkas et al., 2003; Giansanti et al., 2006). UAS::DCog3-RNAi flies were from the Vienna Drosophila RNAi Collection (VDRC). Flies expressing GAL4 (used to inactivate Cog3 or to drive the expression of Gio-RFP from the UAS-gio-RFP transgene) were as follows: Bam-GAL4 (Chen and McKearin, 2003), gift of J. Wakefield (University of Exeter, UK); P{tubP-GAL4}LL7 (tub-GAL4) obtained from the Bloomington Stock Center, P{tubP-GAL4}LL7 gio^RM1-r7 (described in Giansanti et al., 2006). Flies expressing Rab11-GFP (Dollar et al., 2002) were a gift from R. S. Cohen (University of Kansas); flies expressing Sqh-GFP (Royou et al., 2002) were a gift from R. E. Karess (CNRS, University of Paris Diderot). To construct GFP-Cog7; gio^z3934/Df(3R)D1-BX12 and GFP-Cog7; Rab11^e(To)11/Rab11^93Bi mutant strains, GFP-Cog7 (see below) was crossed into the gio and Rab11 mutant background.

To generate the GFP-Cog7 construct, the EGFP CDS was fused in frame to the N-terminus of the full length CG31040 cDNA and cloned into the transformation vector pJZ4 (provided by G. D. Raffa; Raffa et al., 2010) under the control of a tubulin promoter. The Gateway strategy (Invitrogen) was used to generate the UAS-gio-RFP construct; full length gio cDNA (Giansanti et al., 2006) was cloned into the transformation vector pPWR (DGRC, Indiana University) that contains the Gateway cassette, the UASp promoter and the C-terminal monomeric red fluorescent protein (mRFP) sequence. Germline transformation was performed by Bestgene, Inc. (Chino Hills, CA). GFP-Cog7 was crossed into the Cog7 mutant background to test for phenotypic rescue of male sterility and meiotic cytokinesis failures. To verify whether the UAS-gio-RFP transgene could rescue gio mutations, females of genotype w; UAS-gio-RFP; P{tubP- GAL4}LL7 gio^RM1-r7/Tm6B were crossed to w/Y; Df(3R)D1-BX12/Tm6B males and the progeny was scored for viability and fertility.

Cytological preparations were made with testes, brains or salivary glands from third instar larvae. Cog7^z-4495/Df(3R)BSC861 mutants were used in all the experiments involving Cog7 mutations except the ones described in Fig. 1. To visualise α tubulin with either Rab11-GFP or Sqh-GFP or to stain F-actin with Rhodamine-phalloidin (Molecular Probes) larval testes were fixed in 4% formaldehyde as described in Giansanti et al. (Giansanti et al., 2006). For other immunostaining with testes or larval brains, preparations were fixed using 3.7% formaldehyde in PBS and then squashed in 60% acetic acid according to Giansanti et al. (Giansanti et al., 1999). To visualise mitotic chromosomes, larval brains were dissected in NaCl 0.7%, treated with hypotonic solution for 7′, fixed in 45% acetic acid, processed as per Giansanti et al. (Giansanti et. al., 1999) and mounted in Vectashield medium with DAPI (Vector Laboratories, Burlingame, CA). Salivary glands from third-instar larvae were dissected in PBS and fixed in 4% formaldehyde for 10′. Monoclonal antibodies were used to stain α tubulin (T6199, Sigma-Aldrich, diluted 1∶200). Polyclonal antibodies were as follows: anti-Myosin II (Royou et al., 2002), gift from R. E. Karess, diluted 1∶400; rabbit anti Anillin (Field and Alberts; 1995), gift from C. Field (Harvard Medical School), diluted 1∶300; rabbit anti-Lava Lamp (anti-Lva, Sisson et al., 2000), gift from O. Papoulas (University of Texas), diluted 1∶500; rabbit anti Mannosidase II (anti-dGMII, Rabouille et al., 1999), gift from D. B. Roberts, diluted 1∶200; rabbit anti-Gio (Giansanti et al., 2006) diluted 1∶3000; rabbit anti-dSyntaxin16 (antidSyx16, Xu et al., 2002b), gift from W. Trimble (University of Toronto), diluted 1∶20. For staining with fluorescein-labelled cholera toxin B subunit (Sigma-Aldrich, C1655, diluted 1∶200) or with FITC-conjugated wheat germ agglutinin (WGA, Molecular Probes) testes preparations were fixed according to Giansanti et al. (Giansanti et. al., 1999). Secondary antibodies Alexa 555-conjugated anti rabbit IgG (Molecular Probes) and rhodamine/FITC-conjugated anti-mouse IgG (Jackson Immunoresearch, West Grove, PA), were used at 1∶250 and 1∶20 respectively. In all cases slides were mounted in Vectashield medium with DAPI. Images were captured with a charged-coupled device (CCD camera, Photometrics Coolsnap HQ), connected to a Zeiss Axioplan, epifluorescence microscope, equipped with an HBO 100-W mercury lamp, and 40× or 100× objectives as described in Giansanti et al. (Giansanti et al., 2006).

Areas of Golgi bodies were measured using image J (NIH; http://rsbweb.nih.gov/ij/).

Testes from both mutant and control larvae or pupae were dissected in PBS and fixed overnight at 4°C in 2.5% glutaraldehyde in PBS. After two rinses (2×20 minutes) in PBS, specimens were post-fixed in 1% osmium tetroxide in PBS for 2 hours. Preparations were then washed for 30 minutes in PBS and for 30 minutes in distilled water, dehydrated through a series of graded ethyl alcohols and then embedded in an Epon-Araldite mixture and polymerized at 60°C for 48 hours. Ultrathin sections, obtained with a LKB ultratome Nova, were collected on Formvar copper coated grids and stained with uranyl acetate and lead citrate. Samples were then observed with a Philips CM 10 operating at 80 kV. Full serial 400 nm sections through the nuclear region were collected on copper Gilder slot grids, 2×1 mm, coated with Formar. Only selected sections were shown in the figures.

Larval testes were prepared for time lapse as per Inoue et al. (Inoue et al., 2004) and imaged as described in Giansanti et al. (Giansanti et al., 2006). Spermatocytes were examined with a Zeiss Axiovert 20 microscope equipped with a 100×, 1. 25 NA and a 63×, 1.4 NA objectives and a filter wheel combination (Chroma Technology Corp.). Images were collected at one-minute intervals with a CoolSnap HQ camera (Photometrics) controlled through a Metamorph software (Universal imaging); eleven fluorescence optical sections were captured at 1-µm z steps. Movies were created using the Metamorph software and each frame shows the maximum-intensity projection of all the sections.

For immunoblotting analysis of GFP-Cog7 and GFP-Fws, 40 adult testes from males of each genotype, were homogenised in 100 µl of Lysis buffer (10 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5 mM EDTA, 0.5% NP40, 1 mM PMSF, 1× Protease inhibitor Cocktail) at 4°C. Immunoprecipitation experiments from adult testes expressing GFP- and/or RFP-tagged proteins, were performed using the GFP/RFP trap-A kits and control beads purchased from ChromoTek (Planegg-Martinsried). At least 200 adult testes were homogenised in 500 µl of Lysis buffer at 4°C. Lysates were cleared by centrifugation and protein concentration was quantified using the Bradford protein assay (Bio-Rad, Hercules, CA). 4% of each sample was retained as the ‘input’, the remainder was incubated with 20 µl of either GFP trap-A/RFP trap-A or control beads for two hours at 4°C. In all cases beads were washed three times and bound proteins were eluted by boiling in SDS sample buffer. For immunoblots, samples, were separated on 4–20% SDS polyacrylammide precast gels (Bio-Rad) and blotted to PVDF membranes. Blocking and antibody incubation were in Tris-buffered saline (Sigma-Aldrich) with 0,05% Tween-20 (TBST) containing 4% nonfat dry milk (Bio-Rad; Blotting GradeBlocker). Primary antibodies were as follows: anti-Giotto (Giansanti et al., 2006) diluted 1∶8000; rat monoclonal anti-RFP (Chromotek, 5F8) diluted 1∶1000. GFP was detected using a peroxidase labelled anti-GFP (Vector Laboratories). HRP-linked secondary antibodies (GE Healthcare) were used at 1∶5000. After incubation with the antibodies, blots were washed in TBST and imaged using ECL detection kit (GE Healthcare).

We thank Rebecca Farkas for the isolation of z-2805 and z-5797 mutant lines; J. A. Brill and R. Piergentili for critical reading of the manuscript; G. D. Raffa for technical advices; J. Wakefield, R. Karess and S. Cohen for fly stocks; O. Papoulas, R. Karess, D. B. Roberts, C. Field and W. Trimble for antibodies.