The vesicle-tethering protein p115 functions in endoplasmic reticulum–Golgi trafficking. We explored the function of homologous region 2 (HR2) of the p115 head domain that is highly homologous with the yeast counterpart, Uso1p. By expression of p115 mutants in p115 knockdown (KD) cells, we found that deletion of HR2 caused an irregular assembly of the Golgi, which consisted of a cluster of mini-stacked Golgi fragments, and gathered around microtubule-organizing center in a microtubule-dependent manner. Protein interaction analyses revealed that p115 HR2 interacted with Cog2, a subunit of the conserved oligomeric Golgi (COG) complex that is known another putative cis-Golgi vesicle-tethering factor. The interaction between p115 and Cog2 was found to be essential for Golgi ribbon reformation after the disruption of the ribbon by p115 KD or brefeldin A treatment and recovery by re-expression of p115 or drug wash out, respectively. The interaction occurred only in interphase cells and not in mitotic cells. These results strongly suggested that p115 plays an important role in the biogenesis and maintenance of the Golgi by interacting with the COG complex on the cis-Golgi in vesicular trafficking.
To maintain an ordered flow through the secretory pathway, a pair of membranes must specifically recognize and subsequently fuse with each other. Thus, organelles must possess docking and fusion machinery that allows specific recognition of incoming membranes. Biochemical and genetic studies have identified several proteins that support the fidelity of membrane fusion, such as soluble N–ethylmaleimide sensitive factor attachment protein receptors (SNAREs) (1), small guanosine triphosphatases (GTPases) (2) and tethering proteins (3,4). Tethering is thought to involve the formation of physical links, often over considerable distances, between two membranes before trans-SNARE complex formation. In the membrane trafficking pathway, long coiled-coil proteins and the multi-subunit complex are proposed in parallel as candidate tethering factors at each transport step. Both sets of proteins can be recruited to the membrane by small GTPases and both have been suggested to interact with SNAREs to support targeting specificity in the membrane traffic (5–7). It is, however, not clear whether two kinds of tethering complexes share a common role and/or a cooperative role in the tethering process.
p115 is a well-characterized and highly conserved vesicle-tethering protein, which tethers coat protein I (COPI) vesicles to Golgi membranes (8). Although originally identified by using the Rothman intra-Golgi transport assay (9), p115 also functions in endoplasmic reticulum (ER)–Golgi transport (5,6,10,11) and in stacking Golgi cisternae at the M/G1 transition (12). p115 is a peripheral membrane protein mainly localized in the vesicular tubular clusters of ER to Golgi intermediates/ER-Golgi intermediate compartment (VTCs/ERGIC) or cis Golgi network (CGN), and the cis Golgi (9,13). A reduction in p115 below a detectable level by antibodies or small interfering RNA (siRNA) led to Golgi fragmentation and an apparent accumulation of Golgi-derived vesicles (10,11,14–16). p115 is known to interact with several proteins, such as GM130 (8,17), Rab1 (5,18), SNAREs (19), Golgi brefeldin A (BFA) resistance factor 1 (GBF1) (20) and phospholipase Cγ (21), which are involved in vesicular transport or biogenesis of the Golgi apparatus. However, the precise role of p115 in vesicular transport and Golgi biogenesis remains largely unclear.
In this report, we analyzed the role of the p115 HR2 region in the Golgi structure and protein transport using p115 knockdown (KD) cells with or without expression of p115 mutants. The results revealed a novel interaction of p115 with Cog2, a subunit of the conserved oligomeric Golgi (COG) complex, which is required for the Golgi ribbon structure.
The importance of HR2 domain for the ER to Golgi transport
Figure 1A shows a schematic representation of p115, consisting of an NH2-terminal globular head domain (head), a coiled-coil tail that is divided into four parts (CC1–CC4) and a short acidic domain (6,22). The p115 head domain is similar in size to the yeast counterpart, Uso1p and possesses two highly homologous regions (HR1 and HR2, with 62 and 60% identity over 34 and 48 residues, respectively). To study the function of HR2 region, we compared the effect of overexpression of the FLAG-tagged wild type (WT) and the deletion mutant of HR2 region (ΔHR2) in HeLa cells after 20-h incubation of transfection. As shown in Figure 1B, WT and ΔHR2 p115 showed Golgi localization, and no significant change was observed in the Golgi structure, suggesting that the HR2 region plays a role other than targeting the Golgi membrane. An extreme overexpression of p115 caused Golgi fragmentation in HeLa cells (data not shown). When cells expressing WT and ΔHR2 mutant were comparatively examined, cells with Golgi structural defects were found in about 10 and 1.3% of cells expressing WT and ΔHR2, respectively. The less negative effect by the overexpression of ΔHR2 implies an important function of HR2 region.
We then examined the effect of overexpression of WT or ΔHR2 p115 on the secretion of secreted form of dipeptidyl peptidase IV (sDPPIV). Interestingly, in cells overexpressing WT p115, the secretion of sDPPIV was significantly retarded and remained at a half level compared with that of control cells at 80-min chase (Figure 1C). The initial delay of secretion in WT p115–expressing cells resembles the secretion kinetics in cells treated with brefeldin A (23), suggesting that an early step of intracellular transport was affected by overexpressing WT p115. In co-transfected cells, similar amounts of WT and ΔHR2 p115 were expressed (Figure 1D). In addition, we confirmed that there was no significant difference in the secretion kinetics of sDPPIV with overexpressed p115 with or without N-terminal FLAG tag in both WT and ΔHR2. To see the effect of the ER to Golgi transport, oligosaccharide processing of a membrane form of DPPIV (mDPPIV) was analyzed. Cells were transfected with p115 and mDPPIV construct, and were pulse labeled and chased. Immunoprecipitates of 0- and 200-min chase cells with anti-DPPIV antibody was treated with or without endoglycosidase H (endoH) (Figure 1E). At 200-min chase, there was less amount of endoH-resistant form of mDPPIV in WT p115–expressing cells as compared with that of the cells without exogenous p115 expression or with ΔHR2 mutant. These results suggest that the overexpression of WT p115 inhibited the anterograde transport probably at the ER to Golgi transport level. Strikingly, overexpression of ΔHR2 mutant only showed marginal retardation of sDPPIV secretion and no reduction of oligosaccharide processing (Figure 1C,E).
These results clearly indicated that HR2 region has an important role in the inhibition of the ER to Golgi transport seen by p115 WT overexpression. The excess p115 in the HR2 region probably interacted with (an) intrinsic partner protein(s) in non-functional manner resulting in the reduction of the ER to Golgi transport.
Maintenance of the Golgi ribbon requires the HR2 domain of p115
To examine the precise role of p115 HR2 region, we microinjected the WT- or indicated-mutant complementary DNAs (cDNAs) of p115 into the nuclei of p115 KD cells and observed Golgi reassembly after 5 h of incubation. When p115 was depleted, the Golgi apparatus was fragmented and dispersed (Figure 2A, mock). The amount of p115 was reduced in almost all the cells after 96 h of siRNA treatment. The average level of the remaining p115 in used KD cells was 1.9 ± 0.6% by immunostaining and 4.9 ± 0.8% by immunoblotting. The dispersed Golgi fragments consisted of short-stacked cisternae and vesicular structures (16). When WT p115 was expressed in p115 KD cells, an intact ribbon structure of the Golgi apparatus was restored (intact; Figure 2A, WT, Figure S1, available online at http://www.blackwell-synergy.com). The fragmented Golgi remained dispersed (dispersed) when mutant p115 lacking the SNARE-binding domain (ΔCC1) was expressed (Figure 1A), as reported previously (11). In contrast, the expression of ΔHR2 p115 in p115 KD cells resulted in the formation of an irregular Golgi structure consisting of converged tubular and vesicular structures (irregular) at the perinuclear area (Figures 2A,B, ΔHR2, S1). The irregularly assembled Golgi that looked like clustered Golgi fragments was clearly different from dispersed Golgi in p115 KD cells (Figures 2B giantin staining, S1B).
As found in WT-injected cells, the trans (p230) and cis (GM130) Golgi were clearly segregated from each other in ΔHR2-injected cells (Figure 2B, see p230 and GM130 staining), suggesting that the polarity of the Golgi stack may be maintained. However, the continuity of the Golgi ribbon was clearly lost in the irregularly assembled Golgi. The distribution of cells with intact, dispersed or irregular Golgi (viewed by giantin) as in Figure 2A is summarized (Figure 2C). The exogenous expression of WT p115 in p115 KD cells restored the intact Golgi morphology almost completely, while the Golgi remained dispersed in ΔCC1-expressing cells. In contrast, the expression of ΔHR2 resulted in the occurrence of irregular Golgi (57%) in addition to dispersed (41%) and intact Golgi (1%). The defect of ΔHR2 mutant to restore intact Golgi ribbon structure may be caused by the delay of the Golgi assembly process. To examine this possibility, we carried out longer incubation of ΔHR2-injected cells. When ΔHR2 cells were incubated for longer (up to 7.5 h), the ratio of cells with irregular Golgi increased to 65% and dispersed Golgi decreased to 34%. However, there was no significant increase in cells with intact Golgi (1%). It is unlikely that the irregular Golgi is an intermediate form that finally recovers to an intact Golgi. The result shows that the irregular Golgi in ΔHR2-injected p115 KD cells is due to the arrest rater than the delay of the Golgi-formation process.
Irregularly assembled Golgi is a cluster of short fragmented Golgi cisternae
Typical Golgi structure with long cisternae was observed by the re-expression of WT p115 in p115 KD cells (Figure 3A, top panel). In contrast, the irregularly assembled Golgi in ΔHR2-injected p115 KD cells consisted of fragmented Golgi with short tubular and vesicular structures gathered around the microtubule-organizing center (MTOC) (Figure 3A, bottom panel). The irregularly assembled Golgi structure appeared to be a gathering of the scattered mini-stacked Golgi fragments found in p115 KD cells (16). Quantitative analysis confirmed that the average length of Golgi cisternae in ΔHR2-expressing cells was similarly reduced compared with that of p115 KD cells (Figure 3B, compare KD and ΔHR2 to control). In contrast, Golgi cisternae in WT-expressing cells grew longer similar to the level that was found in control cells (Figure 3B, compare control and WT to KD). This observation suggested that the HR2 domain has an essential role in the elongation of Golgi cisternae probably promoted by the fusion of fragmented Golgi cisternae.
The irregularly assembled Golgi localized near the MTOC stained with γ-tubulin as observed in the immunoelectron microscopy, and did not assemble at the perinuclear region upon nocodazole treatment, showing scattered structures at multiple peripheral sites (Figure 4, +noc). The results indicate that the irregular Golgi assembly in ΔHR2-expressing in p115 KD cells shows microtubule dependency, as observed for the normal Golgi. It is probable that CC1 domain is involved in the migration of transport containers to the MTOC, while HR2 is the essential domain of p115 for intact Golgi assembly. The irregularly assembled Golgi was the cluster of the fragmented Golgi that was somehow reformed by the expression of the ΔHR2 p115 mutant from the once fragmented and scattered Golgi fragments by the depletion of p115 but could not be fully assembled in a ribbon structure.
Localization of SNAREs in ΔHR2-expressing cells
It has been reported that p115 directly interacts with SNAREs (syntaxin5, GS27/membrin and GS28) to promote membrane fusion (5,6). We thus examined the localization of ER–Golgi SNAREs in irregularly assembled Golgi. Including p115-binding SNAREs (GS27 and GS28), another SNARE, Bet1, a member of the GS27 and GS28 SNARE complexes, was co-localized with giantin on irregularly assembled Golgi (Figure 4A). The results suggest that the irregularly assembled Golgi was not due to the loss of SNAREs on the Golgi membrane.
Next, we examined whether there is any disturbance in SNARE distribution in p115 KD cells and ΔHR2-expressing cells with irregularly assembled Golgi. In control cells, almost all GS27-positive structures were distinguished from the ER exit site marker hSec13-positive components, whereas in p115 KD cells a portion of GS27 was co-localized with hSec13 (Figure 4B). In ΔHR2-expressing cells, GS27 was gathered to the irregularly assembled Golgi and was distinguished from the peripheral staining of hSec13 more or less similar to the control cells. The result suggests that the trafficking of GS27, a SNARE recycling between the ER and the Golgi, was not perturbed in ΔHR2 mutant–expressing cells. Therefore, the function of p115 in ER to Golgi translocation, that is, its recruitment on COPII vesicle via the Rab1-binding domain CC1 (6,18) and/or the migration of p115-containing ER-Golgi transport carriers to the MTOC was suggested to be normal in ΔHR2 mutant–expressing cells.
HR2 domain is required for full transport efficiency
Next, we examined the ability of the irregularly assembled Golgi in anterograde transport. The secretion of pulse-labeled sDPPIV was measured during the 200-min chase in control and p115 KD cells with or without expression of exogenous p115 constructs. It was found that the amount of protein secreted from p115 KD cells was decreased by half compared with that of control cells (Figure 5A, compare control and KD) (16). This result suggested that the anterograde transport was delayed in p115 KD cells. In ΔHR2-expressed p115 KD cells, the secretion of sDPPIV was significantly retarded and remained at a low level, suggesting the slower transport process (Figure 5A, 40-min chase). Although WT-expressed p115 KD cells restored the secretion to 75% of total sDPPIV, ΔHR2-expressed p115 KD cells showed the same level (47% secretion) as that of p115 KD cells (Figure 5A). Despite such a difference in the Golgi structure and secretion rate, oligosaccharides of sDPPIV secreted from ΔHR2-expressed p115 KD cells were fully processed to the complex type with sialic acid, as observed in WT-expressing p115 KD cells (Figure 5B). The results suggest that the irregularly assembled Golgi exerts less efficient protein trafficking than the intact Golgi while remaining almost normal in the oligosaccharide-processing function.
p115 interacts with Cog2 through HR2
To investigate the role of the HR2 region, we tried to identify a protein that interacts with the HR2 region using yeast two-hybrid screening with the p115 cDNA fragment containing the HR2 sequence (residues 21–247) as bait (Figure 6A). A 2.1-kbp cDNA clone (named H19) was isolated as a candidate p115-binding partner. The H19 cDNA was found to encode part of Cog2/ldlC (residues 554–738) (24,25). The Cog2-binding region on p115 was mapped to residues 200–247, covering HR2 (Figure 6A). To confirm the direct binding of Cog2 fragment to HR2 region, a binding assay using purified recombinant fragments were performed (Figure 6B). A glutathione S-transferase (GST)-tagged p115 fragment containing both HR1 and HR2 region (GST-HR1–2), similar fragments in which HR1 or HR2 region was deleted (GST-ΔHR1, GST-ΔHR2), a fragment only containing HR2 region (GST-HR2) or GST alone were purified and incubated with a His-tagged H19 fragment (His-H19) or a control (His-GCP60N; His-tagged N-terminal fragment of GCP60) (26). GST-fragments were then isolated with glutathione beads and the co-isolated Cog2 fragment (His-H19) was detected by anti-Cog2 antibody. The H19 fragment was only co-isolated with the fragments containing HR2 region and not with the fragments lacking HR2 (Figure 6B, blot). The control His-GCP60N was not co-isolated with any of the fragments. These results strongly suggested that Cog2 binds HR2 region of p115 directly.
It was reported that p115 is predominantly associated with VTCs/ERGIC or CGN adjacent to the Golgi stack (10), while the COG complex is localized on CGN and the Golgi cisternae (25,27). Double staining with antibodies for p115 and Cog2 showed that p115 was partially co-localized with Cog2, even after the cells were treated with nocodazole (Figure 6C). This finding is consistent with a previous observation that p115 was only partially co-localized with subunits of the COG complex, such as GTC90/Cog5 (28) and exogenously introduced hSec34/Cog3 (29). Cog2 was associated with Golgi membranes in both p115 KD cells and ΔHR2-injected p115 KD cells (Figure 6D), suggesting that Cog2 localization does not depend on p115 interaction. These results suggest that the interaction of p115 and Cog2 is dispensable for the proper targeting of Cog2 to the Golgi apparatus.
p115 interacts with the COG complex and golgins
To examine the interaction between p115 and Cog2 in vivo, a Triton-X-100 (TX100) extract of rat Golgi membranes was immunoprecipitated with anti-Cog2 antibody, and examined by blotting with anti-p115 antibody. Figure 7A shows that p115 in the detergent extract was co-precipitated with Cog2. Unfortunately, we could not co-precipitate the COG complex with anti-p115, which was raised to the rod region of p115 (30). To circumvent this problem, we expressed FLAG-WT p115 and subjected to immunoprecipitation with anti-FLAG antibody. As shown in Figure 7B, Cog2 but not the control early endosome antigen 1 (EEA1) was co-precipitated with FLAG-p115 confirming the specific interaction of p115 and Cog2. The COG complex consists of two distinct lobes, each of which comprises four Cog proteins: one lobe consists of Cog1, Cog2, Cog3 and Cog4 and the other of Cog5, Cog6, Cog7 and Cog8 (25). p115 and Cog2 were also co-precipitated with anti-Cog3 and anti-Cog7 antibodies, suggesting that p115 is associated with the entire COG complex through Cog2 (Figure 7C). Efficiency of Cog2 recovery with antibody for Cog2, Cog3 or Cog7 was 7.7 ± 1.3, 3.9 ± 1.1 or 2.8 ± 0.8% in rat Golgi membrane, respectively. The recovery of p115 with anti-Cog7 antibody was 53 ± 8% of that with anti-Cog3 antibody (normalized with recovered Cog2). One possible interpretation is that the antibodies to individual Cog subunits may have different reactivity to the p115-interacting COG complex.
The recovery of p115 in the immunocomplex with anti-Cog antibodies (anti-Cog2 or anti-Cog3) was in the range of 0.8–0.6% and 0.14–0.12% of the total protein in rat Golgi membrane and in the postnuclear supernatant (PNS) of HeLa cells, respectively. The recovery of Cog2 in immunocomplex with anti-Cog3 antibody in HeLa cells PNS was about one-third of that in the rat Golgi membrane (1.3 ± 0.6%). Thus, there is not so much difference in the recovery of p115 in the immunocomplex between the Golgi membrane and the HeLa PNS. The low recovery of p115 may reflect the nature of the p115-Cog complex that is formed by transient and dynamic interaction.
To confirm the interaction of p115 with Cog2 through HR2 region in vivo, HeLa cells were transfected with WT or ΔHR2 of FLAG-p115 and analyzed by co-immunoprecipitation (Figure 7D). In WT-transfected cells, the 0.5 ± 0.2% of FLAG-p115 was recovered in the immunocomplex with anti-Cog3 antibody. In contrast, no detectable level of the FLAG-ΔHR2 was co-precipitated (<0.001%). These results strongly supported the notion that the interaction between p115 and Cog2 is dependent on the HR2 region.
The immunoprecipitate with anti-Cog3 prepared from TX100 extract of rat Golgi membranes was further analyzed for interaction between the COG complex and other Golgi-associated proteins. Two golgins, giantin and GM130 were detected in the precipitate together with Cog2, Cog7 and p115 (Figure 7E). However, the trans-Golgi protein golgin-97 (4) and the medial Golgi protein α-mannosidase II (ManII) were not detected in the immunocomplex, substantiating the specificity of immunoprecipitation. Finally, p115 and Cog2 were also co-precipitated with anti-GM130 antibody confirming that p115, GM130 and the COG complex form a large complex (Figure 7F).
In mitotic phase cells, the Golgi apparatus is fragmented and dispersed throughout the cytoplasm, and the intracellular vesicular transport process is interrupted (19). To analyze the regulation of interaction between p115 and Cog2 during the cell cycle, synchronized mitotic HeLa cells were used for co-immunoprecipitation. As shown in Figure 7G, p115 co-immunoprecipitated with anti-Cog3 antibody remarkably decreased in mitotic phase cells. The result implies that the p115–Cog2 interaction is involved in the Golgi integrity in interphase cells. The results also reinforced the specificity of the p115–COG binding.
Overexpression of the HR2-binding region of Cog2 perturbs p115-mediated Golgi reassembly in p115 KD cells
To further characterize the interaction between p115 and Cog2, the p115-binding region of Cog2 was mapped using the yeast two-hybrid system. A 56-residue region of Cog2 (residues 613–669) was found to interact with the HR2 domain of p115 (Figure 8A).
We then tried to confirm the interaction of p115 and 613–669 region of Cog2. Cells were transfected with FLAG-p115 with or without hemagglutinin-tagged p115-binding region of Cog2 (HA-Cog2613–669), and immunoprecipitated with anti-FLAG antibody followed by immunoblotting with anti-Cog2. Cog2 was shown to be co-immunoprecipitated with FLAG-p115 (Figure 8B, left lane). In the cells co-expressing HA-Cog2613-669, the amount of Cog2 co-immunoprecipitated was about half of that in the cells expressing FLAG-p115 alone (Figure 8B, right lane). The level of another p115-binding protein, GM130 was not significantly changed with or without the HA-Cog2613–669 fragment. The results support the specificity of p115–Cog2 interaction through the Cog2613–669 region in vivo.
Next, we examined the interaction of p115 with Cog2 by injecting p115 constructs into p115 KD cells, in combination with overexpression of the HA-Cog2613–669, which prevents p115 binding to endogenous Cog2 (Figure 8C). The co-injection of WT p115 and HA-Cog2613–669 caused irregular Golgi reassembly, as found in cells injected with ΔHR2 alone, in contrast to the intact Golgi assembly found in WT-expressing cells. The overexpression of HA-Cog2613–669 alone showed no effect on the Golgi structure observed in p115 KD cells. As the coexpression of HA-Cog2613–669 and ΔHR2 showed an irregular Golgi structure, as observed in ΔHR2-injected cells, the overexpression of HA-Cog2613–669 might have little influence on the other functions of p115. Figure 8D summarizes the distribution of cells with intact, dispersed or irregular Golgi in which p115 and HA-Cog2613–669 were coexpressed in p115 KD cells. When HA-Cog2613–669 was coexpressed with WT p115, intact Golgi were observed in only 38% of the injected cells in contrast to 90% recovery of intact Golgi when WT p115 and other Cog2 fragment without the p115-binding region (HA-Cog2554–612) were coexpressed. In other cells, the Golgi structure remained ‘dispersed’ (25%) and ‘irregularly reassembled’ (37%). However, coexpression of ΔHR2 and HA-Cog2613–669 showed a similar distribution of cells with dispersed and irregular Golgi as cells injected with ΔHR2 alone (compare Figure 8D with Figure 2C). Taken together, these results strongly suggested that the binding of the COG complex to the HR2 region of p115 is essential for the reassembly of the Golgi structure.
Overexpression of the HR2-binding region of Cog2 perturbs p115-mediated Golgi reassembly in BFA wash out cells
As expected from the results of p115 ΔHR2 mutant expression in cells without siRNA treatment (Figure 1B), the overexpression of HA-Cog2613–669 showed no effect on the Golgi structure in HeLa cells without p115 KD (Figure 9A, -BFA). Previous observation demonstrated that COG complex is BFA sensitive (24) and p115 is distributed in VTCs/ERGIC by BFA treatment (10). We therefore examined whether HA-Cog2613–669 has some effect on the Golgi reassembly that is observed after forced redistribution of Golgi proteins by BFA treatment and wash out. The cells were expressed with or without HA-Cog2613–669, treated with BFA and washed out to start reassembly of the Golgi apparatus (Figure 9A). The morphology of the Golgi apparatus was categorized and the number of cells showing each category was counted as in Figure 2C (Figure 9B). After 2-h BFA wash out, irregularly assembled Golgi was observed in cells expressing HA-Cog2613–669. Expression of HA-Cog2554–612 did not affect the Golgi structure after BFA wash out. The perturbation of Golgi reassembly with HA-Cog2613-669 showed that the p115 and Cog2 interaction was essential for Golgi ribbon reformation.
To examine the effect of overexpressing HA-Cog2613-669 to the secretion kinetics and the oligosaccharide processing, we performed pulse-chase experiments of sDPPIV in BFA treatment and wash out cells transfected with or without the HA-Cog2613-669 (Figure 9C). Transfected cells were pulse labeled for the last 20 min of BFA treatment, and then chased up to 200 min without BFA. The kinetic analysis of secretion of sDPPIV showed that the level of secretion of cells with overexpression of HA-Cog2613-669 was a half of that in mock-transfected cells after BFA wash out. The result supported that the anterograde trafficking was impeded in cells overexpressing the p115-binding domain of Cog2, as observed in ΔHR2 p115–expressing p115 KD cells (Figure 5A, KD/ΔHR2). Again, we observed no significant alteration of the oligosaccharide processing in cells overexpressing HA-Cog2613-669 in comparison with mock-transfected cells, when analyzed at 200 min of chase (Figure 9D). This indicates that overexpression of the p115-binding region of Cog2 had no effect on the distribution of oligosaccharide-processing enzymes in the Golgi apparatus. Finally, we analyzed the interaction of p115 partner proteins, GM130 and giantin, with the COG complex in cells overexpressing HA-Cog2613-669 after BFA wash out. Recoveries of giantin and GM130 in the immunocomplex with anti-Cog3 antibody were reduced remarkably in cells overexpressing the p115-binding domain of Cog2 (Figure 9E). These results strongly suggested that the expressed Cog2613–669 region specifically interacted with p115 and competed the binding with endogenous Cog2. Furthermore, the recovery of giantin and GM130 is much higher as compared to p115 in the control experiment. The result implies the presence of specific p115-independent interactions between giantin, GM130 and the COG complex.
Previously, we have shown that in p115 KD cells, the Golgi apparatus was fragmented into short-stacked cisternae and vesicular structures, which retained the normal cis–trans Golgi organization and anterograde transport with reduced efficiency (16). In this paper, we demonstrated that two tethering complexes, p115 (golgin) and Cog2 (COG complex) interact. Cog2 binds HR2 region of the p115 head domain and the inhibition of this binding caused irregular assembly of the Golgi apparatus, which consisted of a mini-stacked Golgi with vesicles gathering around the MTOC without reforming a ribbon structure (Figures 2–4). The defect under the inhibition of p115-Cog2 binding was manifested at the reassembly process of the Golgi apparatus after the forced disassembly (p115 re-expression in p115 KD cells or BFA wash out), while the inhibition showed only marginal effect on preassembled Golgi ribbon (Figure 1, 2, 8 and 9). It was thought that p115-Cog2 binding was necessary for the tethering and/or fusion of mini-stacked Golgi apparatus to reform the interconnected ribbon-like structure. Then, what is the importance of the p115-Cog2 binding in the cell? The anterograde transport was only delayed and the oligosaccharide-processing function of the Golgi apparatus was not perturbed in the cell lacking p115–Cog2 interaction. Therefore, the trafficking through the Golgi apparatus is more or less normal even when the p115 cannot interact with Cog2. As it was thought that sDPPIV was mainly accumulating in the ER and/or pre-Golgi compartment in ΔHR2-expressing p115 KD cells (Figure 5), it is probable that the defect in p115–Cog2 binding leads to the reduced tethering and fusion of VTCs/ERGIC to the cis Golgi cisternae resulting in reduced ER–Golgi transport. This idea was supported with histochemical analyses that VTCs/ERGIC contain a significant amount of p115 for fusion with the Golgi stack (31). Furthermore, we have found that the co-precipitation of p115 with the COG complex was greatly reduced in mitotic cells. Therefore, it is possible that p115–Cog2 binding plays an essential role in the reassembly of the Golgi apparatus after the mitosis.
The effect of co-expression of the p115-binding region of Cog2 (HA-Cog2613-669) with WT p115 in p115 KD cells was quite similar on the morphology of the Golgi compared with those in ΔHR2-expressing cells (compare Figure 2 with Figure 8). Furthermore, Golgi apparatus was more or less normally transported around the MTOC in the ΔHR2-expressing cells strongly suggest that ΔHR2 functions normally in microtubule–dependent vesicle movement (Figure 4). Therefore, the irregularly assembled Golgi structure found in ΔHR2-expressing cells is a specific result of the impairment of the p115–Cog2 binding and not the result of the indirect effects caused by global defect on the protein folding.
It has been thought that the tethering complex plays a critical role in the vesicle-docking process to determine the specificity of membrane fusion. Munro and colleagues proposed that the vesicle-tethering process is divided into two steps. The first step is simply kinetic, in which the vesicle is tethered to vicinity of its destination, and the second step is thermodynamic, in which tethering factors actively promote SNARE-mediated fusion. In this aspect, the long coiled-coil proteins act as kinetic tethers and the multi-subunit complexes act as thermodynamic tethers (3,32). Our finding on the interaction between p115 and Cog2 supports their idea that there are cross talks between the two classes of tethering proteins, the long coiled-coil proteins and the multi-subunit complex. However, we do not have direct evidence that interaction of p115 with Cog2 causes SNARE pairing or membrane fusion. The candidate ER-Golgi SNAREs, GS27, GS28 and Bet1 (1), involved in anterograde transport were localized in the irregularly assembled Golgi in ΔHR2-injected cells. Therefore, it is likely that the irregularly assembled Golgi in ΔHR2-injected cells is not due to a loss of fusion machinery. Similar to the defect found in ΔHR2-expressed cells or in cells being perturbed by p115–Cog2 interaction, the acute Cog3 KD by Cog3 depletion with siRNA caused the fragmentation of Golgi structure and showed some kinetic delay in ER–Golgi anterograde transport (33). These observations support the idea that p115 and COG co-operate in vesicle tethering and/or in the fusion to form a fully assembled ribbon-like structure.
It is approved that the cis–Golgi localized COG complex acts as a tether for retrograde COPI-coated vesicles that originated from distal trans-Golgi/endosomal compartments in mammalian cells (33–35). Similarly, the yeast COG complex, Sec34/35 complex was proposed to act as a tether that connects cis-Golgi and COPI vesicles in intra-Golgi retrograde transport (7). However, the COG complex has been implicated in anterograde Golgi trafficking (28) and ER to Golgi protein delivery (36) in mammalian cells. Consistent with this, yeast genetic and biochemical analysis indicated that Sec35p, the yeast counterpart of Cog2, is required for the vesicle-docking stage catalyzed by Uso1p in ER–Golgi trafficking (37) and for vesicle tethering and the fusion process in ER–Golgi transport of glycosyl-phosphatidylinositol (GPI)-anchored protein (38). The COG complex genetically interacts not only with proteins involved in intra-Golgi retrograde transport, Sed5, Ykt6, and COPI coat (7), but also with proteins involved in ER–Golgi anterograde transport, Ypt1, Uso1, and Yos9 (37–39). Therefore, it is possible that the COG complex interacts with different protein sets for playing roles in both anterograde and retrograde transport at cis-cisternae of the Golgi apparatus.
The HR2 region of p115 is evolutionally conserved between yeast and mammal (22). The p115-binding region of Cog2 (residues 613–669) is conserved between a fly (Drosophila melanogaster, 75% similarity to humans), plants (Arabidopsis thaliana, 70% and Oryza sativa, 63%), fishes (Danio, 81% and Tetraodon nigroviridis, 80%) and mammals (mouse, 89%; rat, 91% and bovine, 93%). This domain, however, is not found in the yeast counterpart Sec35p (25). In addition, Sec35p (275 amino acid residues) is much smaller than human Cog2 (738 residues). This might suggest that p115 HR2 domain functions not only through binding to the COG complex but also by interacting with (a) factor(s) not yet identified. Therefore, we could not exclude the possibility that the direct interaction of p115–COG complex is not essential for major functions associated with p115 and the COG complex, while it is important for the maintenance of complete mammalian Golgi ribbon structure. However, there are emerging evidences showing that VTCs/ERGIG are developed in higher eukaryotes but not in yeast (40). Therefore, it seems plausible that mammalian Cog2 acquired some expanded roles to accommodate the needs for the developed architectures of the Golgi apparatus in mammalian cells.
Further analyses of interactions between the COG complex and other p115-binding proteins (SNAREs, Rab proteins and golgins) will help in understanding the complex mechanism of tethering process.
Materials and methods
The following polypeptides were injected into rabbits to raise antibodies: synthetic 19-residue peptides (corresponding COOH-terminal region of Cog2) (24); GST-fusion protein of Cog2 (amino acids 1–108); Cog3 (25) (amino acids 21–190); GM130 (amino acids 53–96, GenBank accession number D86426); and His-tagged Cog7 (25). Polyclonal anti-ManII antibody was a gift from Dr K.W. Moremen. Monoclonal antibodies for GS27, GS28 and bet1 were purchased from BD Bioscience (Franklin Lakes, NJ, USA). Monoclonal anti–γ tubulin was purchased from Sigma-Aldrich Japan Co. (Tokyo, Japan). Polyclonal antibody for protein disulfide-isomerase (PDI) was purchased from Nventa Biopharmaceuticals Corp. (San Diego, CA, USA). Other antibodies were as previously described (16,26,30).
RNA interference treatment
The RNA interference was performed on HeLa cells using Oligofectamine (Invitrogen Carlsbad, CA, USA) with duplex RNA oligos (Dharmacon Inc., Chicago, IL, USA) as previously described (16). Two days after treatment, the cells were retreated with duplex RNA oligos for an additional 2 days. The level of p115 was examined by immunofluorescence microscopy and immunoblotting with p115 antibodies.
Construction and transfection of expression plasmids
For microinjection, codons for Ile118, Lys119 and Gln120 of p115 were mutated by site-directed mutagenesis with synthetic oligonucleotides 5′-CAGAAATTTTCATCAAGCAGCAGGAAAATGTCAC-3′ and 5′-CAGAAATTTTCATCAAACAACAGGAAAATGTCACTCTTCTGT-3′ to generate RNA interference-resistant versions of the protein without changing the encoded amino acids. Deletion mutants of p115 were generated by site-directed mutagenesis with synthetic oligonucleotides 5′-CAGTGAAGAAGATAAACAGTATAATCTTCTTAAAATAC-3′ for ΔCC1 and 5′-CAAGAAGCAATGGTGCAATCAATTTTTTTAAAGAAGGCTCATATATTC-3′ for ΔHR2. Complementary DNAs encoding the WT and deletion mutants (ΔCC1 and ΔHR2) of human p115 were inserted into the EcoRI–XhoI site of the pSG5 vector. Deletion mutants of Cog2 were obtained by polymerase chain reaction (PCR) amplification of Cog2 cDNA with a set of 25-mer primers synthesized for the indicated positions shown in Figure 8. All the constructs prepared by site-directed mutagenesis and PCR-generated fragments were verified by sequencing. The cDNAs encoding the WT and ΔHR2 of p115 or Cog2 mutants were inserted into the EcoRI–XhoI site of the pSG5 vector downstream of the sequence encoding the Met-FLAG™ tag or Met-HA tag, respectively. For transfection experiments, each construct was transfected into HeLa cells with the Lipofectamine 2000 transfection reagent (Invitrogen).
Microinjection and immunocytochemistry
HeLa cells were cultured on a coverslip, and microinjection was performed using a semi-automated microinjection system (InjectMan; Eppendorf, Hamburg, Germany). For nuclear injections, plasmids at 100 μg/ml were used, and Cascade Blue-conjugated dextran (MW 10 000’ Invitrogen) was used as an injection marker. After the desired incubation times, the cells were fixed and immunostained as previously described (26). Stained cells were observed with an LSM PASCAL confocal microscope with ×63 objective (Carl Zeiss, Jena, Germany). Confocal images are presented as sections 0.42 μm in thickness. Where indicated, stained images were presented as projections of stacked image. For electron microscopic studies, cells were grown on a coverslip with the grid of 175 μm. The cells were injected with appropriate constructs and green fluorescent protein (GFP) expression plasmid (BD Bioscience Clontech, Palo Alta, CA, USA) as an injection marker. Uninjected cells that lacked co-injected GFP fluorescence were completely removed using an injection needle. After fixation, the cells were processed for immunoelectron microscopy as previously described (16).
Yeast two-hybrid screening and assay
The cDNA fragment of human p115 containing the HR1 and HR2 regions (amino acid positions 21–247) (30) was used for the bait construct. Screening was performed as previously described (26). Out of approximately 106 clones screened, seven independent clones were obtained. One positive clone was designated as H19, which was found to contain an incomplete open-reading frame of Cog2 cDNA (amino acids 554–738) (24). In the in vitro pull-down assay, H19 was recovered with the p115 fragment used for the bait as described in the results. We could not detect the interaction of the p115 fragment with other six clones in the in vitro pull-down assay (not shown). A cDNA clone (2.2 kbp) containing the complete open-reading frame of Cog2 was isolated by PCR using double-strand cDNA from HeLa cells as a template.
Cell fractionation and co-immunoprecipitation
HeLa cell membrane fractions (26) and rat liver Golgi fractions (17) were prepared by the methods previously described. The membrane fraction of HeLa cells (108 cells) or rat liver Golgi fraction was suspended in 10 mm Tris–HCl (pH 7.5), 60 mm KCl and 1% TX100 (TKT buffer) and incubated for 20 min on ice, followed by centrifugation at 105 000 × g for 20 min. When indicated, interphase and mitotic phase cells that had been prepared as described previously (30) were subjected to immunoprecipitation. The supernatant was subjected to immunoprecipitation with the antibodies indicated, in combination with protein A- or G-Sepharose. The immunoprecipitates were then analyzed by 7% SDS–PAGE and immunoblotting with the antibodies indicated. The immunoreactive proteins were visualized using an ECL kit (Amersham Bioscience, Piscataway, NJ, USA). Blots were scanned and quantified using NIH Image.
GST or GST-p115 constructs on GST-Sepharose was incubated with His-tagged H19 (Cog2 residues 554–738, 2 μg/ml) or His-GCP60N (residues 1–174 of GCP60, 2 μg/ml) in TKT buffer for 1 h at 4°C, and the GST-Sepharose was washed four times with TKT buffer. The amounts of recombinant proteins were standardized using Coomassie brilliant blue staining. The bound proteins were separated on 12.5% SDS–PAGE, and subjected to immunoblotting with anti-Cog2 C-terminal antibody or anti-GCP60 antibody (26).
Secreted or membrane form of DPPIV cDNA was transfected into control and p115 KD cells. After 20 h of transfection, cells were pulse-labeled with [35S]methionine for 20 min and chased. At the indicated times of chase, cell lysates and media were prepared and subjected to immunoprecipitation with polyclonal anti-DPPIV antibodies as described previously (16). The immunoprecipitates were treated with or without endoH or N-acetylneuraminidase, and analyzed by SDS–PAGE (7.5% gels) and fluorography. Fluorograms were scanned and quantified for labeled sDPPIV using NIH Image. In the case of p115 KD cells expressing p115 mutants, sDPPIV and p115 constructs were co-transfected into p115 KD cells, which were used after 20-h incubation.
We thank Dr K. W. Moremen (University of Georgia) for reagents, and Dr K. Nakayama (Kyoto University) for useful discussions and comments on the manuscript. This work was supported in part by grants from the Ministry of Education, Science, and Sports of Japan, Narushige Zoological Science Award, and the Central Research Institute of Fukuoka University.