Detailed Analysis of the Interaction of Yeast COG Complex

. The Golgi apparatus is a central station for protein trafficking in eukaryotic cells. A widely accepted model of protein transport within the Golgi apparatus is cisternal maturation. Each cisterna has specific resident proteins, which are thought to be maintained by COPI-mediated transport. However, the mechanisms underlying specific sorting of these Golgi-resident proteins remain elusive. To obtain a clue to understand the selective sorting of vesicles between the Golgi cisterenae, we investigated the molecular arrangements of the conserved oligomeric Golgi (COG) subunits in yeast cells. Mutations in COG subunits cause defects in Golgi trafficking and glycosylation of proteins and are causative of Congenital Disorders of Glycosylation (CDG) in humans. Interactions among COG subunits in cytosolic and membrane fractions were investigated by co-immunoprecipitation. Cytosolic COG subunits existed as octamers, whereas membrane-associated COG subunits formed a variety of subcomplexes. Relocation of individual COG subunits to mitochondria resulted in recruitment of only a limited number of other COG subunits to mitochondria. These results indicate that COG proteins function in the forms of a variety of subcomplexes and suggest that the COG complex does not comprise stable tethering without other interactors.


Introduction
It is essential for cellular organelles to maintain characteristic molecular compositions. Protein transport between single-membrane-bounded organelles is mostly mediated by membrane carriers, such as vesicles and tubules that bud from a donor membrane and fuse with a target membrane. The transport of cargo molecules is regulated by sets of proteins such as SNAREs, small GTPases of Sar/Arf and Rab families, vesicle coat and tethering proteins (Bonifacino and Glick, 2004;Cai et al., 2007). The Golgi apparatus functions as a central station of the membrane trafficking system in eukaryotic cells. It consists of flattened membrane-enclosed compartments, called cisternae, which are orderly differentiated in their functions and structures from cis to trans. Each cisterna has specific resident deletion of lobe A subunits causes severe growth defects, whereas deletion of lobe B subunits does not affect cell growth (Lees et al., 2010). COG subunits directly interact with multiple Golgi SNARE proteins, COPI coat and Rab GTPases (Suvorova et al., 2002;Willett et al., 2013). Thus mutations in yeast COG subunits exhibit a variety of defects such as Golgi morphology, intra-Golgi and endosomal transport and Golgi glycosylation enzyme activities (Whyte and Munro, 2001;Suvorova et al., 2002;Bruinsma et al., 2004;Fotso et al., 2005). In studies of mammalian cells, COG subunits are present as independent subcomplexes on membranes (Willett et al., 2016) and are shown to recruit different SNARE proteins and work as specific landmarks of vesicle tethering on the Golgi membrane (Willett et al., 2013). Every subunit has been shown essential for the COG complex function in Golgi trafficking, though to varying extents . Relocation of a lobe A subunit, COG4, to mitochondria recruited membranes bearing Golgi SNARE protein STX5 (Sed5 in yeast). A lobe B subunit, COG8, recruited membranes that carry TGN SNARE protein, STX16 (Tlg2 in yeast). The interaction between individual COG proteins and other Golgi transport machinery has also been investigated by pull-down and yeast two-hybrid methods (Suvorova et al., 2002;Shestakova et al., 2007). However, the details of in vivo interactions among COG subunits are not fully understood in yeast.
To understand the molecular arrangements of the COG subunits in vivo, we analyzed in detail the interactions among COG subunits in cytosolic and membrane fractions by co-immunoprecipitation in the present study. Both the lobe A and the lobe B subunits interacted with all other subunits in the cytosolic fraction, suggesting that cytosolic COG subunits exist as octamers. By contrast, membraneassociated COG subunits form a variety of subcomplexes. The results indicated complex features of inter-subunit interactions in yeast cells. Further investigating interactions of COG proteins, one of each COG subunits were ectopically targeted to mitochondria membrane. Mitochondrial COG relocation experiments showed that exogenously expressed mitochondria-associated COG proteins were able to recruit other subunits, but could not outcompete the endogenous Golgi-associated COG complex.

Microscopy
Fluorescence microscopy was performed by super-resolution confocal microscopy (SCLIM) with a UPlanSApo 100 X NA 1.4 oil objective lens (Olympus, Tokyo, Japan), a high-speed spinningdisk confocal scanner (Yokogawa Electric, Tokyo, Japan), a custom-made spectroscopic unit, image intensifiers (Hamamatsu Photonics, Hamamatsu, Japan) with a custom-made cooling system, and two EM-CCD cameras (Hamamatsu Photonics) (Kurokawa et al., 2013). For microscopic observation, all strains were grown in selective medium (0.67% yeast nitrogen base without amino acids and 2% glucose) with appropriate supplements. For imaging of COG-mCherry-Fis1 cells, 42 optical slices spaces 0.1 μm apart were collected. Maximum intensity projection was conducted by z project of Fiji plugin Max Intensity (Schindelin et al., 2012).
Cells were resuspended 1 mL of IP buffer [150 mM NaCl/50 mM Tris (pH 7.4)/7 μL/mL Halt Protease Inhibitor Cocktail EDTA free (Thermo Fisher)/1 mM PMSF] and homogenized by dounce homogenizer. Total cell lysates were centrifuged at 2,000 × g for 5 minutes at 4°C to remove debris. Supernatants were separated into supernatant (cytosol) and pellet (membrane) fraction by ultracentrifugation at 100,000 × g for 1 h at 4°C in a Beckman Optima MAX-XP Ultracentrifuge (TLA 55 rotor). Pellet was resuspended in 500 μL of IP buffer containing 1% Triton X-100, incubated on ice for 30 minutes and centrifuged at 10,000 × g for 5 minutes at 4°C. Supernatant was used as membrane fraction (P100) for the following immunoprecipitation analysis. 500 μL of supernatant (S100, cytosol fraction) was solubilized with Triton X-100 (final concentration 1%) on ice for 30 minutes. 1 μL of anti-GFP antibody (mouse monoclonal, 3E6, Thermo Fisher) was added to both membrane and cytosol fractions and incubated on ice in a cold room overnight. 60 μL of 30% Protein G beads (Protein G-Agarose, Roche Diagnostics, Mannheim, Germany) was added to each tube and incubated with gentle mixing in a cold room for 2 h. Beads were then washed four times with PBS containing 0.05% Triton. Samples were eluted in 2x Laemmli sample buffer (Bio-Rad, Hercules, California, USA) and heated for 5 minutes at 95°C to elute the bound proteins.

mCherry beads Immunoprecipitation
Cells were grown at 30°C in selective medium with 2% galactose overnight. 200 OD 600 unit cells were harvested, incubated at 30°C for 10 minutes in 0.1 M Tris-HCl (pH 9.4)/10 mM DTT, harvested, resuspend in 50 mM HEPES pH 7.2/0.8 M mannitol/20 μL zymolyase, and spheroplasted for 1 h at 30°C. Spheroplasts were harvested by centrifugation at 2,000 × g for 5 minutes. Spheroplasts were resuspend 1 mL of IP buffer and homogenized by dounce homogenizer. Total cell lysates were centrifuged at 2,000 × g for 5 minutes at 4°C to remove debris. Triton X-100 (final concentration 1%) was added to supernatant and incubated on ice to solubilize. 60 μL of 50% mCherry-Nanobody beads (LaM4.2-His beads) (Fridy et al., 2014) was added and incubated with gentle mixing at room temperature for 1.5 h. Beads were then washed four times with PBS containing 0.05% Triton. Samples were eluted in 2x Laemmli sample buffer (Bio-Rad) and heated for 5 minutes at 95°C to elute the bound proteins.

COG-GFP immunoprecipitation
To investigate the in vivo interaction among yeast COG subunits in the cytosol and on the membrane, cells expressing GFP-tagged COG proteins were subjected to immunoprecipitation by the anti-GFP antibody. The immunoprecipitates were analyzed for the interaction with other subunits by Western blotting using antibodies against the individual COG subunits. GFP-tagged COG cells grew as wild type cells, indicating GFP-fusion COG proteins are functional. Cell lysates were separated into cytosolic (S100) and membrane (P100) fractions by centrifugation at 100,000 × g. COG proteins were found in both cytosolic and membrane fractions as described previously (Suvorova et al., 2002). In the cytosolic fraction, all the eight subunits were co-immunoprecipitated with Cog1, Cog3, Cog5 and Cog6-GFP, confirming that all eight subunits were stably associated in a COG1-8 octameric complex in the cytosol fraction (Fig. 1). On the other hand, COG subunits showed different interaction patterns in the membrane fraction. Cog1 co-immunoprecipitated both the lobe A (Cog1-4) and the lobe B (Cog5-8) subunits. Cog3 co-immunoprecipitated mainly the lobe A (Cog1-4) subunits. Cog5 did not show strong interactions with other subunits, whereas Cog6 interacted strongly with Cog8 (Fig. 1). These co-immunoprecipitation data suggest that membrane-associated COG subunits exist in the forms of subcomplexes; lobe A and lobe B. These results are consistent with the observation of HeLa cells, which showed that complete COG1-8 octamers were formed in the cytosolic fraction but a combination of octamers and subcomplexes were present in the membrane fraction (Willett et al., 2016). Probably COG subunits strongly interact with each other in the cytosolic fraction but weakly in the membrane fraction. In other words, the COG complex may tend to dissociate into subcomplexes when they bind to the membrane or to membrane-associated protein partners.

Ectopic targeting of COG subunits to mitochondria
To determine the role of COG subunits in vesicle targeting, we developed constructs in a way that one of the COG subunits would be ectopically tethered to the mitochondrial outer membrane (Willett et al., 2013). The mitochondrial protein Fis1 integrates its C-terminal transmembrane domain into the mitochondria outer membrane and exposes its N-terminus to cytosol (Mozdy et al., 2000). The Cterminus of each of the eight individual COG subunits was fused to mCherry-Fis1 (COG-mCherry-Fis1 collectively, Fig. 2A). Expression of COG-mCherry-Fis1 fusion proteins was placed under control of the GAL1 promoter (strongly induced in a galactose-containing medium and repressed in a glucose-containing medium).
First, the growth of COG-mCherry-Fis1 or COG-Fis1 cells was examined on galactose-containing plates. As shown in Fig. 2, neither the overexpression of untagged COG subunits by 2μ multicopy plasmids nor the COG-mCherry-Fis1 expression by chromosomal integration plasmids showed growth defects. By contrast, the overexpression of COG-mCherry-Fis1 by 2μ multicopy plasmids caused marked growth defects for the cases of Cog3, 4, 5 and 6. This indicates that overexpression of some specific COG subunits on the outer mitochondrial membrane is deleterious to cells. Interestingly, Western blotting with anti-mCherry or anti-Cog4 antibodies did not detect a band corresponding to Cog4-mCherry-Fis1 either for integration or multicopy plasmid constructs (Fig. S1), indicating that the expression of the full-length Cog4-mCherry-Fis1 protein was toxic and the cells kept the amount of proteins below the detection limit (see Fig. 2). COG-mCherry-Fis1 protein was confirmed to colocalize with a GFP-tagged mitochondria marker protein, Cit1 (Fig. S2).

COG subunits mislocalized to mitochondria recruit other COG subunits
Using mitochondria-localizing COG-mCherry-Fis1, we examined the recruitment of other COG subunits. For the subsequent analysis, we used the cells that expressed COG-mCherry-Fis1 by the integration plasmid, because they showed no growth defect. First the interaction among COG subunits was analyzed by co-immunoprecipitation. COG-mCherry-Fis1 was immunoprecipitated using mCherry-Nanobody (LaM4.2)-conjugated beads (Fridy et al., 2014) and analyzed for interacting proteins by Western blotting with anti-COG antibodies.
Mitochondrially-mislocalized subunits of the lobe A, Cog1 and Cog3, co-immunoprecipited other subunits of both the lobe A and the lobe B (Fig. 3). The mislocalized lobe B subunit, Cog5-mCherry-Fis1, recruited other lobe B subunits and Cog4. The mislocalized Cog6-mCherry-Fis1 recruited other subunits of both the lobe A and the lobe B. The mislocalized Cog8-mCherry-Fis1 did not show appreciable interaction with other subunits, indicating that Cterminal tag may interfere with efficient incorporation of COG8 in the COG complex. In a previous study, Cog1 was shown to be important for recruiting the lobe B subunits (Fotso et al., 2005). Here we showed that not only Cog1 but also Cog3 and Cog6 have the ability to recruit other subunits of both the lobe A and the lobe B.

The COG complex alone is insufficient to recruit Golgi derived vesicles
It has been shown that COG proteins interact with Golgi SNARE proteins, Sed5, Ykt6, Gos1 and Sed22 in yeast cells (Suvorova et al., 2002). In mammalian cells, mislocalized COG4 and COG8 proteins are reported to recruit vesi- Fig. 1. Interactions of COG subunits. Lysates from cells expressing Cog1-GFP (A), Cog3-GFP (B), Cog5-GFP (C) or Cog6-GFP (D) were separated into S100 (cytosol) and P100 (membrane) fractions by centrifugation at 100,000 × g. S100 and P100 fractions were immunoprecipitated (IP) with anti-GFP. IP and 10% immunodepleted/flow-through (FT) fractions were separated by SDS-PAGE and blotted with antibodies against Cog1-8. 10% input of solubilized proteins was run as a control. The bottom bar graph shows IP efficiency of upper blots. IP efficiency values were calculated by dividing IP by input.

Fig. 3. COG-mCherry-Fis1
recruited other subunits to the mitochondria. COG-mCherry-Fis1s were expressed under GAL1 promoter by plasmids integrated into chromosome. Yeast cell lysates were immunoprecipitated (IP) with mCherry-Nanobody beads. IP and 5% immunodepleted/flow-through (FT) fractions were separated by SDS-PAGE and blotted with antibodies against mCherry, Cog1, 2, 4, 5, 6 and 8. 5% input of yeast total lysate was run as a control. The bottom bar graph shows IP efficiency of upper blots. IP efficiency values were calculated by dividing IP by input. cles including Golgi and TGN SNARE proteins (Willett et al., 2013). We examined localization of COG and SNARE proteins in cells expressing chromosomally-integrated COG-mCherry-Fis1. Cells were observed at 3 h after induction, under which condition mitochondria did not show abnormal morphology. As shown in Fig. 4, Golgi SNAREs, Sed5 and Gos1 and TGN SNAREs, Tlg1 and Tlg2 were not recruited to Cog3 or Cog6. This result suggests that the tested mislocalized COG subunits are unable to fulfill full tethering activities of COG subunits. Long time induction (more than 6 h) of COG-mCherry-Fis1 often caused aggregation of mitochondria, however we could not observe SNARE mislocalizations even under this condition. This suggests that COG-mCherry-Fis1 could not stably recruit the Golgi derived vesicles to mitochondria. Alternatively, this level of overexpression was not sufficient to change fractionation profiles of the bulk of Golgi proteins. From these results, unfortunately, it remains elusive whether the individual COG subunit has the ability to recruit or keep tethering vesicles to the membrane. If not, other interactors, in addition to COG subunits themselves, might be needed for this function.

Discussion
The COG complex has been characterized as the peripheral protein complex that functions in the tethering of Golgi derived vesicles. Studies of yeast and mammalian cells have shown that the COG complex directly interacts with the Golgi SNARE complexes (Suvorova et al., 2002;Shestakova et al., 2007). In mammalian cells, COG4 and COG8 are reported to recruit different SNARE proteins and form different vesicle tethering platforms (Willett et al., 2013). We hypothesized that the COG subunits of yeast cells also play roles as specific landmarks for different vesicles, and mediate specific intra-Golgi vesicle transport. In this study, we first analyzed interactions among the COG subunits in cytosol and on the membrane. COG subunits form a stable octameric complex in cytosol, whereas they form not only octamers but also various subcomplexes in the membrane fraction. Considering its function as a tethering complex, the cytosolic COG complex may not be fully active and the COG subunits that associate with each other in a variety of forms on the membrane should work as tethers. It is consistent with the result on mammalian cells that most of cytosolic COG subunits exist as octomers and membrane-associated COG proteins exist as smaller subcomplexes (Willett et al., 2013(Willett et al., , 2016. Whereas the quadruple lobe B mutant yeast cells (Δcog5 Δcog6 Δcog7 Δcog8) grow normally like wild-type cells in the laboratory condition (Lees et al., 2010), the mutations in lobe B subunits in human cause congenital disorders of glycosylation (CDG) diseases (Freeze and Ng, 2011). The roles of lobe B proteins might be different between species. Our result that COG proteins form multiple forms of subcomplexes on the membrane suggests that octamers and smaller subcomplexes play different roles on the membrane. Deciphering the different roles of the octameric complex and subcomplexes will need further investigations. Cog6-mCherry-Fis1 were expressed under GAL1 promoter by plasmids integrated into chromosome. GFP tagged SNARE proteins, Sed5, Gos1, Tlg1 and Tlg2 were constitutively expressed by low copy plasmids. Cells were grown in synthetic medium containing 1% raffinose overnight, then in synthetic medium containing 2% galactose for 3 h at 30°C and observed by SCLIM. Maximum intensity projections were shown. Scale bar, 5 μm.
To further understand the roles of COG subunits, mitochondrial targeting assays have been employed (Sengupta et al., 2009;Willett et al., 2013;Luo et al., 2014;Willett et al., 2016). We used this method because direct vesicle sorting from the Golgi apparatus to the mitochondria is not known and the mislocalization of secretory machineries to mitochondria would separate them from the normal secretory pathway. This type of assay worked well in mammalian cells, in which mislocalized COG subunits specifically recruit vesicles containing SNARE proteins (Willett et al., 2013) and in yeast cells, mislocalization of the CATCHR family tethering complex, exocyst subunit Sec3, recruited secretory vesicles (Luo et al., 2014). We expected that ectopic targeting of specific COG subunits to mitochondria would give us some information about the specific vesicle tethering mechanism of the COG complex. It was unclear at this point whether the growth defects were due to relocation of the COG complex or they recruited unknown essential proteins to mitochondria. Importantly, overexpression of mitochondrial-targeted COG4 was very toxic in yeast cell, suggesting that potential recruitment of Sed5containing membranes to mitochondria was interfering with normal yeast cell growth. Cog1, 3 and 6 recruited both lobe A and lobe B subunits, indicating that they are able to form a platform to assemble the octameric COG complex. If mitochondria-localized COGs were indeed unable to recruit vesicles, it would suggest that the COG complex interacts with other components for a short period of time and COG itself cannot maintain stable interactions with the vesicles. Recently, Anderson et al. (2017) reported that the mammalian CASP homolog, Coy1, interact with both COG proteins and Golgi SNARE proteins (Anderson et al., 2017). Further studies on interactors of the COG complex would provide better understanding of intra-Golgi trafficking.