Identification of GOLPH3 Partners in Drosophila Unveils Potential Novel Roles in Tumorigenesis and Neural Disorders

Golgi phosphoprotein 3 (GOLPH3) is a highly conserved peripheral membrane protein localized to the Golgi apparatus and the cytosol. GOLPH3 binding to Golgi membranes depends on phosphatidylinositol 4-phosphate [PI(4)P] and regulates Golgi architecture and vesicle trafficking. GOLPH3 overexpression has been correlated with poor prognosis in several cancers, but the molecular mechanisms that link GOLPH3 to malignant transformation are poorly understood. We recently showed that PI(4)P-GOLPH3 couples membrane trafficking with contractile ring assembly during cytokinesis in dividing Drosophila spermatocytes. Here, we use affinity purification coupled with mass spectrometry (AP-MS) to identify the protein-protein interaction network (interactome) of Drosophila GOLPH3 in testes. Analysis of the GOLPH3 interactome revealed enrichment for proteins involved in vesicle-mediated trafficking, cell proliferation and cytoskeleton dynamics. In particular, we found that dGOLPH3 interacts with the Drosophila orthologs of Fragile X mental retardation protein and Ataxin-2, suggesting a potential role in the pathophysiology of disorders of the nervous system. Our findings suggest novel molecular targets associated with GOLPH3 that might be relevant for therapeutic intervention in cancers and other human diseases.


Molecular Cloning
dGOLPH3-mRFP was generated by cloning full-length Drosophila GOLPH3 (dGOLPH3, sauron, sau) cDNA into a pCasper4-tubulin [4] in frame with C-terminal mRFP. dGOLPH3-mRFP was crossed into the dGOLPH3 (sau z2217 /Df(2L)Exel7010) mutant background to test for phenotypic rescue of male sterility and meiotic cytokinesis failure. To generate the RFP-βCOP fusion construct, the cDNA of βCOP was cloned into a pCasper4-tubulin in frame with N-terminal mRFP. The GFP cDNA was cloned into a pCasper4-tubulin to generate the GFP transgenic flies. Transgenic flies were generated by P-element mediated germline transformation, performed by Bestgene Inc. (Chino Hills, CA, USA).

Co-Immunoprecipitation Experiments
Co-immunoprecipitation (Co-IP) experiments were performed as described in [27]. For the experiments shown in Figures 3 and 4, 400 adult testes of each genotype were homogenized in 500 µL of lysis buffer [25 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1mM EDTA, 1% NP-40] with protease inhibitors (#11697498001, Roche, Basel, Switzerland) on ice using a Dounce homogenizer. For the experiments of AP-MS, 3000 adult testes from either dGOLPH3-RFP or RFP males, were homogenized on ice in 1ml of lysis buffer with protease inhibitors using a Dounce homogenizer. Lysates were clarified by centrifugation and protein concentration was quantified using a NanoDrop 2000c Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). 4% of each lysate was retained as the "input". The remainder was precleared with control agarose beads (bab-20, ChromoTek, Planegg, Germany). Co-IP experiments from lysates expressing GFP or RFP-tagged proteins, were performed using GFP/RFP trap-A purchased from ChromoTek (#gta-100, #rta-20), following the protocol previously described [33]. The beads were rinsed once with ice-cold IP Lysis buffer and washed extensively (4 × 5 min) on the wheel at 4 • C. After the final wash, the beads were resuspended in 30 µL of SDS sample buffer [20% glycerol, 4% SDS, 0.2% BBF, 100 mM Tris-HCl (pH 6.8), 200 mM DTT] and boiled for 10 min.
To immunoprecipitate dGOLPH3, the testis extract from 400 adult testes expressing dAtx2-HA was precleared with Protein A-Agarose (SC-2001, Santa Cruz, Biotechnology, Dallas, TX, USA) and divided into two. Fractions were incubated with either 4 µg of rabbit anti-GOLPH3 antibody L11047/G49139/77 [4] or 4 µg of rabbit pre-immune serum L11047/G49139 [4] from the same animal before the immunization. After antibody incubation, Co-IP was carried out using the Protein A-Agarose (#SC-2001, Santa Cruz) following the manufacturer's instructions. Co-IP experiments were performed in triplicate with identical results.

Glutathione S-Transferase (GST) Pull-Down Assays
GST and GST-dGOLPH3 proteins were expressed in bacteria and purified using glutathione-Sepharose 4B beads (#17-0756-01, GE Healthcare, Arlington Heights, IL, USA) following the manufacturer's instructions, as described previously [27,37]. GST pull-down experiments were performed with testis lysates using the procedure described in [38]. Testis lysates were incubated with either GST or GST-dGOLPH3 (at the appropriate concentration), bound to glutathione-Sepharose 4B beads, with gentle rotation at 4 • C for 2 h. After rinsing in wash buffer (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP-40, 1 mM EDTA, Protease and phosphatase inhibitors) three times, the beads were boiled in SDS sample buffer and separated by SDS-PAGE. Bound proteins were analyzed by Western Blotting. Before immunoblotting, PVDF membranes were stained with Ponceau (#P3504, Sigma-Aldrich). GST pull-down experiments were performed in triplicate with identical results.

Proteomics and Data Analysis
Visualization of protein bands was obtained using a colloidal Coomassie staining. From each SDS-PAGE lane, ten slices were excised and submitted to a trypsin proteolysis [39]. Peptide mixtures were then extracted from the gel matrix and submitted to a desalting step by solid phase extraction before mass spectrometric analyses [40]. Nano-liquid chromatography tandem mass spectrometry (nanoLC-MS/MS) analyses were performed using an Ultimate3000 system (Thermo Fisher Scientific) equipped with a splitting cartridge for nanoflows and connected on-line via a nanoelectrospray ion source (Thermo-Fisher Scientific) to an LTQ-Orbitrap XL mass spectrometer (Thermo-Fisher Scientific). Each sample was automatically loaded from the autosampler module of the Ultimate 3000 system at a flow rate of 20 µL/min onto a trap column (AcclaimPepMap µ-Precolumn, 300 µm × 1 mm, Thermo Fisher Scientific) in 4% ACN containing 0.1% FA. After 4 min, peptides were eluted at 300 nL/min onto a 15 cm column (360 µm OD × 75 µm ID, 15 µm Tip ID; PicoFrit, New Objective, Littleton, MA, USA) and custom packed by reverse phase (C18.5 µm particle size, 200 Å pore size; Magic C18AQ, Michrom Bioresources, Auburn, CA, USA) using a two-step gradient of solvent B (from 5% to 40% in 120 min and from 40% to 85% in 15 min). Data-dependent tandem mass spectroscopy (MS/MS) was performed using full precursor ion scans (MS1) collected at 30,000 resolution, with an automatic gain control (AGC) of 1 × 106 ions and a maximal injection time of 1000 ms. The 5 most intense (>200 counts) ions with charge states of at least +2 were selected for collision-induced dissociation (CID). Dynamic exclusion was active, with 90 ms exclusion for ions selected twice within a 30 ms window. For MS/MS scanning, the minimum MS signal was set to 500, activation time to 30 ms, target value to 10,000 ions and injection time to 100 ms. All MS/MS spectra were collected using a normalized collision energy of 35% and an isolation window of 2 Th. All MS/MS samples were analyzed using the software package MaxQuant (version 1.3.0.5, Max Planck Institute of Biochemistry, Martinsried, Germany). Peptides sequences were searched against the Drosophila melanogaster Uniprot proteome database and common contaminant proteins.
We set oxidation (methionine) and phosphorylation (serine, tyrosine, threonine) as variable modifications, carbamidomethylation (cysteine) as a fixed modification, mass tolerance of 20 ppm for the precursor ion (MS) and of 0.5 Da for the fragment ions (MS/MS). High-confidence peptide-spectral matches were filtered at <1% false discovery rate. Proteins recognized as having a low confidence level [i.e., (i) number of unique peptides ≤ 0, (ii) identified only by a modified peptide, (iii) less than 3 MS/MS spectra] were filtered out. Individual MS/MS spectra were manually inspected for proteins represented by a single tryptic peptide.

Computational Analysis of the dGOLPH3 Interactome
Protein classes and GO over-representation analyses were performed using the PAN-THER database [41], while GO enrichment analysis was performed using the GOrilla tool [42]. Prism 9 (GraphPad Software, San Diego, CA, USA) and Excel (Microsoft Corporation, Redmond, WA, USA) software were used for statistical analyses and to prepare graphs.

Proximity Ligation Assay
Larval testes were dissected in PBS and fixed using 4% methanol-free formaldehyde in PBS. Samples were blocked with the blocking solution contained in the kit (Duolink In Situ PLA Probes, #DUO92001/DUO92005, Sigma-Aldrich), following the instructions provided by the manufacturer. After blocking, samples were incubated with primary antibodies diluted in Duolink In Situ Antibody Diluent included in the kit (Duolink In Situ PLA Probes, Sigma-Aldrich) overnight in a humid chamber at 4 • C. Monoclonal antibodies were used to stain dFmr1 (1:600, clone 6A15, #F4554, Sigma-Aldrich). Polyclonal antibodies were: anti-HA (1:600, clone C29F4, # 3724, Cell Signaling) mouse anti-Rab1 (1:750, antibody S12085a [27]) and rabbit anti-dGOLPH3 (1:1500, [4]). The PLA probe incubation and the detection protocol were performed in accordance with the procedures described in the Duolink In Situ-Fluorescence User Guide, using the Duolink In Situ PLA Probes and Duolink In Situ Detection Reagents (#DUO92013/DUO92014, Sigma-Aldrich). Following the detection steps, specimens were mounted in Vectashield Vibrance Antifade Mounting Medium containing DAPI (#H-1800, Vector Laboratories). Images were captured with a charged-coupled device (Axiocam 503, mono CCD camera) connected to a Zeiss Cell Observer Z1 microscope (Carl Zeiss AG, Oberkochen, Germany) equipped with an HXP 120 V inclusive built-in power supply, lamp module and a 63X/1.4 objective. Images were acquired using the ZEN2 software along the z-axis. Projections were created using the Extended Depth of Focus function of the ZEN2 software and processed in Photoshop. Quantification of the number of PLA signals per cell was obtained using the Analyze Particles tools of the ImageJ software. Number of PLA signals, compared to background signals in the control, was examined for statistical significance using the nonparametric Mann-Whitney test.

Identification of the dGOLPH3 Interactome in Drosophila melanogaster
To identify the in vivo interactome of dGOLPH3, we performed affinity purification from testis extract of flies expressing either dGOLPH3-RFP or RFP alone as control. We first assessed whether dGOLPH3-RFP localized to the Golgi in testes co-expressing dGOLPH3-RFP and GFP-Cog7 by time-lapse imaging. dGOLPH3-RFP co-localized with GFP-Cog7 [33] to the Golgi organelles of primary spermatocytes and to the ribbon-like acroblast of spermatids ( Figure 1). dGOLPH3-RFP and its associated partners were characterized by RFP affinity purification, coupled with mass spectrometry (AP-MS), and selected interactors were subsequently validated by co-immunoprecipitation (Co-IP) or gluthatione S-transferase (GST) pull-down.
Using the MaxQuant searching platform, we identified a list of dGOLPH3-interacting proteins that met a cut-off criterion of confidence greater than 95% and were absent in controls (Supplementary Table S1). The proteins identified by AP-MS were categorized into broad functional classes using PANTHER [41] (Figure 2a and Supplementary  Table S2). Gene ontology (GO) enrichment and over-representation profiles of the dGOLPH3 interactome were analyzed using GOrilla [42] and PANTHER [41], respectively, and revealed a significant enrichment of proteins involved in cytokinesis and vesicle-mediated trafficking (Figure 2b; Supplementary Tables S3 and S4).  Table S2 for details. (b) Heat maps showing the GO annotation enrichment profiles of the dGOLPH3 interactome. GO enrichment profiles were analyzed using GOrilla tool [42] under the category "process" and PANTHER database [41] under the category "GO-slim biological process". Over-represented/enriched GO terms are shown in different color shades according to their fold enrichment as indicated in the color scale bar at the bottom; actual fold enrichment values are shown within the heat map (see Supplementary Tables S3 and S4 for p-values). For simplicity and to improve visual representation, for the PANTHER over-representation analysis only the headings for each GO-slim biological process are shown in the graph, while the full results are reported in the Supplementary Table S4. Only results for Bonferroni-corrected analysis (p < 0.05) were considered.

Profiling dGOLPH3 Interactors Reveals an Enrichment of Membrane Trafficking Proteins
dGOLPH3-RFP pulled down well-established interactors of GOLPH3/Vps74p, which include coat protein complex I (COPI) subunits and the phosphatidylinositol 4-phosphatase Sac1 ( [7,12,43,44]; Table 1, Supplementary Table S1). Golgi vesicle transport-GO: 0048193 Rab protein signal transduction-GO: 0032482  Golgi organization-GO: 0007030 Cell cycle-GO: 0007049 Signaling-GO: 0023052 Cilium assembly-GO: 0060271 Nervous system process-GO: 0050877 1 UniProtKD entry; unique and stable entry identifier. 2 Drosophila annotation symbol; Current FlyBase annotation identifier of the gene. CG, prefix for protein-coding genes. 3 Drosophila symbol; Approved Drosophila gene/protein symbol. 4 PEP; Posterior Error Probability of the identification. This value essentially operates as a p-value and represents the probability that the observed peptide spectrum match (PSM) is incorrect (a smaller value is more significant).
Co-IP and GST pull-down experiments further validated the association of dGOLPH3 with COPI subunits (Figure 3a-d). protein coprecipitated with RFP-βCOP but not with γCOP-RFP and RFP. Protein extracts from testes expressing RFP-βCOP, γCOP-RFP (a) and RFP (RFP IP CTRL) (b), were immunoprecipitated with RFP-trap beads (α-RFP) and blotted for either RFP or dGOLPH3. 4% of the total lysates and one third of the IP were loaded and probed with the indicated antibodies. (c-e) GST pull-down to test dGOLPH3 interaction with αCOP (c), δCOP (d) and Sec31 (e) proteins. (c) Bacterially expressed GST and GST-dGOLPH3 were purified by Gluthatione-Sepharose beads, incubated with testis protein extracts from Oregon-R males and blotted for αCOP protein. (d) Bacterially expressed GST and GST-GOLPH3 were purified by Gluthatione-Sepharose beads, incubated with testis protein extracts from males expressing δCOP-HA. (e) Bacterially expressed GST-dGOLPH3 and GST purified by Gluthatione-Sepharose beads were incubated with testis protein extracts from Oregon-R males and blotted for Sec31. Ponceau staining in (c-e) is shown as a loading control. 2% of the input and 25% of the pull-downs were loaded and probed with the indicated antibody. Molecular masses in (a-e), expressed in kilodaltons.
Consistent with the known requirement of GOLPH3 proteins for Golgi architecture maintenance and vesicular trafficking [1][2][3][4][5]17], the dGOLPH3 interactome comprises vesicle coat proteins, Rab GTPases, proteins of the tethering and fusion machinery and endocytic trafficking regulators (Table 1). Listed among the vesicle transport proteins are not only the COPI subunits but also the COPII proteins Sec31, Sec23, and Sec24CD, suggesting the role of dGOLPH3 in controlling Endoplasmic Reticulum (ER) to Golgi trafficking ( Table  1, Supplementary Table S1). The binding of dGOLPH3 to Sec31 was further validated by using GST pull-down (Figure 3e). The Rab family GTPases that bound to dGOLPH3 have been implicated in multiple steps of intracellular trafficking and tethering (Table 1, Supplementary Table S1, [45]). Among these Rab GTPases, Rab1, Rab11, and Rab5 had already been identified as molecular partners of Drosophila and/or human GOLPH3 in previous studies [4,27,46,47]. Rab1 controls ER to Golgi and intra-Golgi trafficking whereas Rab5 and Rab11 regulate endocytic trafficking [4,27,37]. Rab8, Rab10 and Rab14 regulate post-Golgi trafficking from the trans-Golgi network to the plasma membrane [48]. Rab32 has been associated with vesicle trafficking through lysosome and is required for autophagy and lipid storage [49,50]. Co-IP experiments further validated binding of GOLPH3 to Rab8 and Rab10 in Drosophila testes (Figure 4a). We also identified the Drosophila ortholog of human COG7 (namely Cog7), a subunit of the Conserved Oligomeric Golgi (COG) complex, which plays a pivotal role in tethering retrograde vesicles that traffic within the Golgi and between the endosomes and the Golgi. Further evidence of COG-dGOLPH3 association was provided in our previous studies [27,38]. Of interest is also the presence in the dGOLPH3 interactome of Sec22 and Slh, which both regulate membrane fusion events. Sec22, a v-SNARE (vesicle soluble N-ethylmaleimide-sensitive factor attachment protein receptor), is necessary in promoting efficient membrane fusion in the cis-Golgi and in the contact sites between the ER and the plasma membrane [51,52]. Slh is the Drosophila homolog of human Sly1, a Sec1/mammalian Unc-18 (SM) protein which functions at the ER-Golgi to regulate SNARE complex assembly and membrane fusion [53][54][55][56]. Further validation of Sec22-dGOLPH3 interaction was obtained by Co-IP experiments (Figure 4b).
The candidate dGOLPH3 partners regulating protein transport include Sec63 and Srp54k, which are components of the signal recognition particle (SRP), the multimeric ribonucleoprotein machine that, along with its conjugate SRP receptor, controls targeting of secretory proteins to the rough ER [57]. Among additional partners, we also found Drosophila Gilgamesh (Gish), a plasma membrane-associated casein kinase that has been involved in the maintenance of germline stem cell sperm individualization in testes [58,59]. Gish also regulates polarized Rab11-vesicle trafficking during trichome formation [60]. In accordance with our previous data [4], we found evidence of the association between dGOLPH3 and the clathrin heavy chain subunit (Chc) ( Table 1, Supplementary Table S1). Besides Chc, other dGOLPH3-partners involved in clathrin-mediated endocytic trafficking are Shibire and SH3PX1, that are, respectively, the Drosophila orthologs of dynamin and Sorting nexin 9 (Snx9) ( Table 1, Supplementary Table S1). GST pull-down experiments further provided experimental validation of Shibire and SH3PX1 as dGOLPH3 protein partners (Figure 5a,b).

The dGOLPH3 Interactome Reveals Functions in Several Glycosylation Pathways
Experimental data from both yeast and human cultured cells revealed that GOLPH3 controls COPI-mediated Golgi trafficking of several Golgi glycosyltransferases required for N-and O-glycosylation [7,8,12]. Our results suggest that dGOLPH3 might bind to glycosyltransferase enzymes that control multiple glycosylation pathways such as Nand O-linked glycan synthesis and glycosylphosphatidylinositol (GPI) anchor processing (Table 1, Supplementary Table S1). Four proteins in the list of dGOLPH3 interactome, CG6790, CG5342, CG4907 and PIG-T, are predicted to be involved in GPI-anchor biosynthesis [61]. In the context of N-glycosylation, the dGOLPH3-interactome indicates an association with proteins involved in the early steps of N-glycosylation. Ost∆ is a subunit of the oligosaccharyl transferase (OST) complex that catalyzes the initial transfer of Glc3Man9GlcNAc2 from dolichol-pyrophosphate to an asparagine residue within an Asn-X-Ser/Thr consensus motif in nascent polypeptides [62]. Uridine diphosphate (UDP)glucose:glycoprotein glucosyltransferase plays a crucial role in glycoprotein quality control in the ER [63,64]. The other two proteins involved in N-glycosylation are the Drosophila orthologs of human alpha-1,2-mannosyltransferase (ALG11) and phosphomannomutase type 2 (PMM2) [65][66][67][68][69][70]. ALG11 controls the addition of the first alpha-1,2-linked mannose residues to growing linked-oligosaccharide [66]. Human PMM2 catalyzes the second step in the mannose pathway, which converts mannose-6-phosphate to mannose-1-phosphate, the precursor of GDP-mannose [67][68][69][70].
Our results implicate dGOLPH3 in the synthesis of mucin-type O-glycans (initiated by GalNAc-Ser/Thr) and glycosaminoglycan (GAG) chains. Consistent with the previous report from Chang and coauthors [16], we found that dGOLPH3 interacted with the exostosin Brother of tout-velu (Botv), a glucuronyl-galactosyl-proteoglycan 4alpha-N-acetylglucosaminyltransferase required for heparan sulfate proteoglycan synthesis. Additional molecular partners include Pgant5 and Pgant7, which display Nacetylgalactosaminyltransferase activity required to initiate mucin-type O-glycosylation [71].

dGOLPH3 Partners Control Lipid Homeostasis and Golgi Architecture
Along with the phosphatidylinositol 4-phosphatase Sac1, other proteins that bound GOLPH3 have a role in lipid homeostasis (Table 1, Supplementary Table S1). The protein Small wing (Sl) is a phosphatidylinositol-specific phospholipase type C [72] that catalyzes the hydrolysis of phosphatidylinositol (4,5) bisphosphate into two second messengers, inositol 1,4,5-trisphosphate (IP3) and diacylglycerol (DAG). Multi-substrate lipid kinase (Mulk) is the Drosophila ortholog of human Acylglycerol kinase (AGK) ceramide kinase which reports to phosphorylates ceramide and acts in Wnt-mediated migration of primordial germ cells [73]. PAPLA1 enzymes cleave the ester bond at the Sn-1 position of Phosphatidic acid to produce lysophosphatidic acid, a bioactive phospholipid that mediates several signaling functions [74,75]. The genome of Drosophila encodes a unique PAPLA1 enzyme, while the mammalian PAPLA1 family consists of three members: DDHD1, DDHD2 (or KIAA0725p), and the SEC23 interacting protein (SEC23IP) [76][77][78]. Drosophila PAPLA1, by interacting with the COPII proteins Sec23 and Sec31, regulates ER to Golgi transport and glycosylation of Rhodopsin 1, an N-glycosylated G-protein coupled receptor [77].
The Drosophila Bond protein is a member of the Elovl family of enzymes that is required for the elongation of very-long-fatty-acids, commonly found in sphingolipids and essential for sphingolipid function [79,80]. Similarly to dGOLPH3, Bond has been involved in spermatocyte cytokinesis [80]. Moreover, both PAPLA1 and Bond proteins are required for sperm individualization during Drosophila spermatogenesis [77,81]. Among the proteins involved in lipid metabolism we found the very-long-chain enoyl-CoA reductase Sc2, required for very long-chain fatty acid biosynthetic process [82] and the Sphingosine-1-phosphate lyase Sply, which catalyzes the conversion of sphingosine-1-phosphate to ethanolamine phosphate and a fatty aldehyde [83]. The interactome of dGOLPH3 comprises several proteins required for Golgi structure and function. The ubiquitin-selective AAA-ATPase valosin-containing protein (VCP) (also known as Transitional endoplasmic reticulum 94, TER94) controls the cell-cycle-dependent Golgi fragmentation/assembly as well as the ubiquitin-proteasome system [84]. Ergic53 is the Drosophila ortholog of mammalian ER-Golgi intermediate compartment-53 (ERGIC-53), also known as p58 and lectin mannose binding 1 (LMAN1), a type I transmembrane protein containing a mannose binding domain that has been established as a marker of the ER-Golgi intermediate compartment (ERGIC or IC) [85,86]. Mammalian P58/ERGIC-53/LMAN1, by interacting with COPI and COPII coats, operates as a cargo receptor and a recycling protein at the ER-Golgi interface [87,88] and is essential for maintaining the architecture of ERGIC and Golgi [86].
Drosophila Alphabet is a metal-dependent serine/threonine phosphatase of the PP2C family, closely related to mammalian PP2Cα/β isoforms, that acts as a negative regulator of RAS/MAPK signaling [101,102].
The Mitogen-activated protein kinase kinase (MAPKK) 4 is predicted to have a Jun kinase kinase activity [103]. In accordance with our previous work [22], we provide evidence for the interaction between dGOLPH3 and the regulatory light chain Spaghetti Squash (Sqh) of non-muscle Myosin II, a structural component of the actomyosin contractile ring during cytokinesis. The Drosophila GTP binding proteins, Rac and Cdc42, are also key regulators of actin cytoskeleton organization and have been involved in vesicle trafficking, cell polarization and JNK kinase activation [104][105][106][107][108] Moreover, studies in mammalian cell culture systems have shown that coatomer-bound Cdc42 regulates actin assembly, dynein recruitment and bidirectional transport at the Golgi [109][110][111]. Co-IP experiments from testes expressing GFP-tagged Rac1 and Rac2 further validated binding of dGOLPH3 to Rac2 GTPase (Figure 4b).

dGOLPH3 Interactors Suggest an Involvement in the Assembly or Organization of Ciliary and Flagellar Axonemes
Many of the candidate partners of dGOLPH3 have been implicated in motile cilia assembly and/or function, suggesting that GOLPH3 might have a role in this process (  [121]. Remarkably, the human orthologs of these proteins have been involved in primary ciliary dyskinesia and Kartagener syndrome, rare autosomal recessive genetic disorders affecting motile cilia function and characterized by chronic respiratory infections and defects in male fertility [122]. Cut up (Ctp) and Cytoplasmic dynein light chain 2 (Cdlc2) belong to the dynein light chain family and exhibit dynein intermediate chain binding activity [123][124][125]. Defective transmitter release (Dtr) protein is the Drosophila ortholog of human dynein axonemal assembly factor 1 [121], a protein that is required for the stability of the ciliary structure and involved in cytoplasmic preassembly of dynein arms [126]. The kinesins Kinesin-like protein at 10A (Klp10A) and Kinesin-like protein at 59D (Klp59D) and Centrosomal protein of 97 kDa (CEP97) are also required for sperm ciliogenesis [127][128][129][130].
Immunostaining of primary spermatocytes for dAtx2-HA and dFmr1 showed that these proteins co-localize in the cytoplasm of premeiotic spermatocytes as expected from their described roles in translational control (Figure 7a). Figure 7. dFmr1 localization in premeiotic and dividing spermatocytes. (a) dFmr1 co-localizes with dAtx2 in the cytoplasm of primary spermatocytes. Testes expressing dAtx2-HA were stained for dFmr1(green), HA (dAtx2, red) and DNA (DAPI, blue). n = 40 spermatocytes randomly selected from images taken in five independent experiments. (b) Premeiotic and dividing spermatocytes were stained for dFmr1 (green) dGOLPH3 (red) and DNA (DAPI, blue). dFmr1 localizes to the cytoplasm of premeiotic spermatocytes and concentrates at the midzone (arrows) and at the astral membranes (arrowheads) at each pole of dividing spermatocytes (telophase). dGOLPH3 protein is visible in the cytoplasm and enriched in the Golgi stacks of premeiotic spermatocytes (yellow arrowheads). During telophase dGOLPH3 protein localizes to the midzone (arrows) and is enriched in puncta at the astral membranes of dividing spermatocytes (arrowheads). n = 40 premeiotic spermatocytes and n = 25 telophase spermatocytes, randomly selected from images taken in five independent experiments. Bars, 10 µm.
Immunostaining of testes with anti-dFmr1 and anti-dGOLPH3 antibodies showed that dFmr1 is enriched at the midzone and overlaps with dGOLPH3-enriched organelles at the astral membranes of dividing spermatocytes (Figure 7b). Recent data have shown that dAtx2 interacts and functions with dFmr1 in neuronal translational control to mediate long-term olfactory habituation [133]. We further validated the interaction between dGOLPH3/dFmr1 and dFmr1/dAtx2 in fixed spermatocytes using a proximity ligation assay (PLA, Figure 8). Taken together, our results indicate that dFmr1 interacts with both dGOLPH3 and dAtx2 in the cytoplasm of male meiotic cells.

The dGOLPH3 Interactome Reveals an Enrichment in Vesicle-Mediated Trafficking and Cytokinesis Proteins
In this paper, we report the first comprehensive analysis of the interactome of GOLPH3 protein. We have exploited the advantages of Drosophila spermatogenesis, which offers a well-suited model system for dissecting membrane trafficking pathways and their role in cytokinesis and cell differentiation [23][24][25][26][27]. Consistent with our previous findings that dGOLPH3 controls membrane trafficking during cytokinesis [4,21,22,27], the dGOLPH3 interactome revealed an enrichment of proteins involved in cytokinesis and vesicle-mediated trafficking. Importantly, we have identified well-established molecular partners of GOLPH3/Vps74 such as COPI subunits, Sac1 and Rab1 [7,12,43,44,46,147]. Among the novel molecular interactors of dGOLPH3, we found vesicle coats, small GT-Pases of the Rab family, and tethering and fusion factors, indicating roles in both secretory and endocytic trafficking pathways. We showed that dGOLPH3 bound Rab8 and Rab10 proteins that regulate post-Golgi trafficking from the trans-Golgi network to the plasma membrane [48]. Rab8 functions with Rab10 and Rab14 in GLUT4 cycling [148]. Moreover, both Rab8 and Rab10 contribute to ciliogenesis [148,149]. dGOLPH3 also bound other Rab GTPases involved in controlling endocytic trafficking. Importantly, distinct Rab proteins localize at specific membrane-bound compartments and act in concert with different phosphoinositides to regulate all the vesicular trafficking pathways [150]. Although GOLPH3 localization to the Golgi membranes depends on PI(4)P, both human and Drosophila GOLPH3 proteins were able to bind PI(3)P and PI(4,5)P2 in lipid-binding assays [4,13]. Moreover, by using surface plasmon resonance, Wood and co-authors [13] demonstrated that the human GOLPH3 binds PI(3)P with a mere threefold affinity compared with PI(4)P. Thus, GOLPH3 proteins might associate with either PI(4)P-vesicles or PI(3)P-enriched endosomes and regulate secretory and endosomal membrane dynamics in concert with specific Rab GTPases, during interphase and cytokinesis.
In the context of the GOLPH3-controlled vesicle trafficking in cytokinesis, a vast amount of literature has discussed the role of endocytosis and endocytic recycling pathways during furrow ingression and the final steps of cytokinesis in model organisms and mammalian cultured cells [151][152][153]. At least two endocytic Rab GTPases, namely Rab11 and Rab35, control distinct endocytic recycling pathways required for completion of cytokinesis in mammalian cells [154][155][156][157][158][159]. Consistent with our previous studies on fly spermatocytes [20,27], our AP-MS experiments identified Rab11 (but not Rab35) as a molecular partner of dGOLPH3. In Drosophila melanogaster Rab11 is essential for cytokinesis of S2 cells and spermatocytes [159][160][161] as well as for furrow ingression during embryonic cellularization [162,163]. In dividing spermatocytes, Rab11 concentrates to the cleavage furrow, together with its effector Nuclear fallout [160,161], providing an essential function for contractile ring constriction and furrow ingression [160]. Data from the Brill group showed that the Drosophila type III PI 4-kinase four wheel drive (Fwd), localizes to the Golgi of male meiotic cells, recruits Rab11 to the Golgi complex and is required for the accumulation of PI(4)P-vesicles co-localizing with Rab11 at the cell equator of dividing spermatocytes [161]. However, because Fwd does not localize to the cleavage furrow, targeting of PI(4)P and Rab11 vesicles to the equatorial site depends on dGOLPH3 function [4]. Thus, data in this paper together with our previous analyses [4,21,22] suggest a model whereby the PI(4)P effector dGOLPH3 forms a complex with Rab11 and myosin II and coordinates contractile ring assembly with phosphoinositide signaling and vesicle trafficking during cytokinesis.
Further studies on mammalian cells should clarify whether GOLPH3 functions with the Rab11 pathway to promote actin clearance and ESCRT-III dependent abscission in late cytokinesis [158].
Importantly, human SNX9 subfamily proteins are required for accumulation of active myosin II at the cleavage site and normal furrow ingression during cytokinesis of HeLa cells [179]. Further work will clarify whether SH3PX1/SNX9 protein cooperates with GOLPH3 to regulate membrane remodeling and actomyosin dynamics during cytokinesis.
Although most studies on membrane trafficking during cytokinesis have revealed the essential role of the Golgi and endocytic pathways proteins, it has been suggested that the ER might provide an important membrane storage within the dividing cells [190]. Consistent with a possible role of the ER in cytokinesis, a proteomic analysis of purified midbodies isolated from Chinese hamster ovary cells led to the identification of both ERresident proteins and proteins involved in ER to Golgi traffic including Sec13, endoplasmin, Sec23, Sec31 and COPI [191]. Moreover, time-lapse fluorescence analysis of the ER disulfide isomerase GFP chimera protein revealed pronounced reorganization during cytokinesis and a redistribution of this protein to the spindle poles and the spindle equator of Drosophila dividing cells [192,193]. The interactome of dGOLPH3 comprises the COPII proteins Sec31, Sec23 as well as ERGIC-53 and Sec22 proteins, which are known to operate at the level of the ERGIC/cis-Golgi. Additionally, consistent with our previous findings [27], data in this paper confirm the interaction between dGOLPH3 and Rab1, which controls ER to Golgi and intra-Golgi trafficking. Future studies will be required to explore the functional dependence between dGOLPH3 and ER/ERGIC/cis-Golgi proteins and the possible implications of these interactions during cytokinesis.

dGOLPH3 Interacts with Proteins Required for Protein Glycosylation and Lipid Homeostasis
It has been amply demonstrated that GOLPH3 proteins control COPI-mediated Golgi trafficking of specific Golgi glycosyltransferases [7,8,10,12]. Our findings suggest that dGOLPH3 is required for N-and O-linked glycan synthesis and glycosylphosphatidylinositol (GPI) anchor processing. Importantly our results indicate a role of dGOLPH3 in the biosynthesis of glycosaminoglycan chains, that was also reported by Chang and coauthors [16]. Moreover, our results implicate dGOLPH3 in the synthesis of mucin-type O-glycans. It has been proposed that the oncogenic properties of human GOLPH3 might be correlated with defects in protein glycosylation [5]. Indeed, altered glycoprotein glycosylation represents a hallmark of cancer and in particular aberrant mucin-type O-glycans have an important role in cancer pathogenesis as they affect the adhesive properties of the neoplastic cells and promote cell invasion and tumor metastasis [194].
The interactome of dGOLPH3 indicates an important role in lipid metabolism and signaling that correlates with tumorigenesis and a variety of human genetic diseases [195]. In this context, several molecular interactors of dGOLPH3 are required for the synthesis of sphingolipids and other signaling lipids including lysophosphatidic acid and DAG. These data support the model whereby PI(4)P-GOLPH3 exerts a key function to coordinate lipid homeostasis with vesicle trafficking and glycosylation at the Golgi.

The dGOLPH3 Interactome Indicates Molecular Targets that Might Be Relevant for Therapeutic Intervention in Cancer and Other Neurological Diseases
It has been suggested that human GOLPH3 and Golgi alterations might have a potential role in the pathophysiology of neurological disease [196]. Remarkably, Golgi fragmentation of specific groups of neurons is an early preclinical event in many neurodegenerative diseases, preceding pathological symptoms [197]. Our findings that dGOLPH3 interacts with dFmr1 and dAtx2 suggest a link with polyglutamine diseases. Studies on the members of Ataxin-2 from human cells and model organisms indicated a conserved role of ATX2 proteins in regulating of mRNA stability and translation [133,134,[198][199][200][201][202]. ATX2 has been also involved in endocytic trafficking [203,204] and has been localized to ER membranes and Golgi [205,206]. Recent studies provided evidence for a conserved role of Atx2 in ER dynamics and structure in C. elegans as well as in Drosophila embryos and cultured neurons suggesting a possible mechanism that involves vesicle trafficking in SCA2 disease [207]. Mammalian FMRP and dFmr1 are mainly localized in the cytoplasm, where they bind specific mRNAs acting as translation regulators [133,136,[208][209][210]. FMRP proteins also interact with components of the microRNA and Piwi-interacting RNA pathways [211][212][213][214][215]. Recent data have shown that dAtx2 interacts and functions with dFmr1 in neuronal translational control to mediate long-term olfactory habituation [133].
Our results indicate that dFmr1 forms a complex with both dAtx2 and dGOLPH3 in male meiotic cells. In addition, dFmr1 protein is enriched at the poles and in the midzone of the dividing spermatocytes, suggesting a potential role in cytokinesis. We speculate that dFmr1 protein might control localization and transport of specific mRNAs to midzone microtubules where they are locally activated during cytokinesis. A similar regulatory process occurs in neurons where mRNAs, together with the machinery for RNA translation, are transported from the cell body to synapses where they are locally translated [216]. Moreover, cytokinesis of early C. elegans embryos requires ATX2 function, which controls a molecular mechanism required to target and maintain the kinesin ZEN-4 to the spindle midzone through the posttranscriptional regulation of PAR-5 [217]. Consistent with a role of dFmr1 in cytokinesis, several mRNA/protein targets of dFmr1, so far identified, are involved in actin cytoskeleton remodeling. dFmr1 binds Cytoplasmic FMRP Interacting Protein (CYFIP), the fly ortholog of vertebrate FMRP interactor CYFIP1, that is part of the WAVE regulatory complex that regulates actin polymerization [218,219]. In addition, the function of dFmr1 in dendritic development depends on the small GTPase Rac1 and the Rac1-encoding mRNA is present in the Fmr1-messenger ribonucleoprotein complexes [209]. Finally, dFmr1 controls actin cytoskeleton dynamics in Drosophila neurons by binding the mRNA of the Drosophila profilin homolog chickadee and regulating Profilin protein expression [210]. Although the relationship between the dGOLPH3 and the dFmr1-mediated mRNA transport remains to be determined, proteins of the COPI vesicle complex interact with specific mRNAs and disruption of the COPI complex results in mis-localization of RNAs in human neurons [220,221]. Moreover, COPI α binds a specific set of mRNAs that overlaps with FMRP-associated mRNAs which encode proteins that are known to localize to the plasma membrane and cytoskeleton [222]. Uncovering the molecular mechanisms that involve dGOLPH3, and dFmr1 in the dynamics of RNAs during cell division will further our understanding of diseases of the nervous system. Moreover, recent studies indicate the involvement of human FMRP in different cancer types including breast cancer and melanoma [223,224]. Thus, investigating the functional dependence between GOLPH3 and FMRP will be also important in the light of a therapeutic strategy in human cancer.
The Co-IP of dGOLPH3 is also enriched with proteins controlling cell cycle progression and cell proliferation including protein kinases and phosphatases. A significant finding in our work is the interaction of dGOLPH3 with proteins that are known to play a role in the TOR kinase signaling pathway. Importantly, in the context of cancer pathogenesis, human GOLPH3 function has been associated with enhanced AKT/mTOR signaling, although the precise biochemical basis for its activity remains to be determined [17]. Importantly, findings in this paper suggest that GOLPH3 might form a complex with proteins that have been involved in the TOR signaling pathway: Tctp, 14-3-3 ζ [112,113] and the conserved TOR-binding protein LST8 [118][119][120]. Further work will clarify whether the association between GOLPH3 and these proteins can impact on the TORC1 and TORC2 complexes and consequently on cell growth and proliferation.