p125 Is Localized in Endoplasmic Reticulum Exit Sites and Involved in Their Organization*

Transport vesicles coated with the COPII complex, which is assembled from Sar1p, Sec23p-Sec24p, and Sec13p-Sec31p, are involved in protein export from the endoplasmic reticulum (ER). We previously identified and characterized a novel Sec23p-interacting protein, p125, that is only expressed in mammals and exhibits sequence homology with phosphatidic acid-preferring phospholipase A1 (PA-PLA1). In this study, we examined the localization and function of p125 in detail. By using immunofluorescence and electron microscopy, we found that p125 is principally localized in ER exit sites where COPII-coated vesicles are produced. Analyses of chimeric proteins comprising p125 and two other members of the mammalian PA-PLA1 family (PA-PLA1 and KIAA0725p) showed that, for localization to ER exit sites, the p125-specific N-terminal region is critical, and the putative lipase domain is interchangeable with KIAA0725p but not with PA-PLA1. RNA interference-mediated depletion of p125 affected the organization of ER exit sites. The structure of the cis-Golgi compartment was also substantially disturbed, whereas the medial-Golgi was not. Protein export from the ER occurred without a significant delay in p125-depleted cells. Our study suggests that p125 is a mammalian-specific component of ER exit sites and participates in the organization of this compartment.

the selection of cargo proteins and membrane deformation, the latter of which leads to the generation of vesicular structures (5)(6)(7)(8)(9)(10). In mammalian cells (11) and yeast Pichia pastoris (12), COPII-coated vesicles are produced at a specialized ER subdomain known as ER exit sites or the transitional ER. Recent studies involving living cells revealed that ER exit sites are immobile yet dynamic (13) and are generated de novo (14,15). In mammalian cells, newly formed ER exit sites appear to move slowly from the cell periphery toward the perinuclear Golgiproximal region (15).
Biochemical and live cell imaging studies have suggested that the COPII coat is replaced with the COPI coat in close proximity to ER exit sites (16 -20). This replacement results in the formation of a membrane compartment named vesicular tubular clusters (VTCs) or the ER-Golgi intermediate compartment (ERGIC). These cargo-containing VTCs move in a COPIdependent manner along microtubule tracks toward the cis-Golgi (21). It has been suggested that ER-resident proteins are segregated in VTCs and transported back to the ER by COPIcoated vesicles via the retrograde pathway (22). Although the boundaries between the ER exit sites, VTCs, and cis-Golgi are not fully defined, several marker proteins have been reported for each compartment.
COPII consists of two heterotrimeric complexes, Sec23p-Sec24p and Sec13p-Sec31p, and a low molecular weight GTPbinding protein Sar1p (5). Activation of Sar1p on the ER membrane results in the sequential recruitment of the Sec23p-Sec24p complex and the Sec13p-Sec31p complex, which leads to vesicle budding (5,7). Concomitant with this budding process, cargo proteins are selectively incorporated into forming vesicles most likely through interaction with Sec24p (8 -10). Sec23p exhibits a GAP activity toward Sar1p, and hydrolysis of GTP on Sar1p, perhaps occurring after vesicle fission, results in the depolymerization of the COPII coat, thereby leading to uncoating (23).
We identified previously and characterized a mammalian Sec23p-interacting protein, p125, that appears to be only expressed in mammals and exhibits a wide tissue expression distribution (24,25). p125 contains an N-terminal proline-rich region responsible for the interaction with Sec23p (25) and central and C-terminal regions, which exhibit significant sequence homology to phospholipid-modifying proteins, especially PA-PLA 1 , which was identified by Higgs et al. (26,27). A data base search led to the discovery of another p125-like protein, KIAA0725p, that shows 52.6% aa identity with p125 over the entire sequence but lacks the N-terminal, proline-rich Sec23p-interacting region (28). It seems that PA-PLA 1 , p125, and KIAA0725p constitute a novel protein family, i.e. the PA-PLA 1 family. Although the physiological implication of this family is poorly understood, members of this family may in general be involved in membrane transport or organelle maintenance. Recent work on an Arabidopsis mutant has demonstrated that SGR2p, a member of this family, is involved in gravity sensing, perhaps by mediating vacuole and/or vesicle transport to the vacuoles in endodermal cells (29,30). Understanding of the function of this family may provide an insight into the relationships between lipid metabolism, signaling, and membrane transport.
Although our previous studies involving an ectopic expression system contributed to the understanding of the role of p125 in the early secretory pathway (24,25), more direct approaches are necessary to define the exact localization and function of p125. In this study, we produced antibodies that specifically recognize endogenous p125, and we examined the effect of depletion of p125 on the localization of proteins within the early secretory pathway. Our results demonstrated that p125 is mainly localized in ER exit sites, where COPII-coated vesicles are produced, and contributes to the organization of this compartment.

EXPERIMENTAL PROCEDURES
Antibodies-The cDNA fragment encoding aa 1-134 of p125 (p125N) was inserted into expression vector pGEX-4T-1 (Amersham Biosciences). The GST-p125N fusion protein was expressed in Escherichia coli cells, purified, and then injected into rabbits according to the standard method. The antisera produced were first depleted of anti-GST antibodies by chromatography on a GST column, and then the anti-p125 antibody was affinity-purified using antigen-coupled beads. For the production of mAb, the purified protein was injected into BALB/c mice. Hybridoma cells producing antibodies against p125 were obtained according to the standard protocol. The monoclonal antibodies were purified from ascites using a HiTrap TM protein G HP column (Amersham Biosciences).
The anti-FLAG mAb (M2) and the polyclonal anti-GST antibody were obtained from Sigma and Santa Cruz Biotechnology, respectively. The polyclonal antibodies against Sec23p and mannosidase II were purchased from Affinity BioReagents and Chemicon International, respectively. The polyclonal antibodies against VSV-G, ␤-COP, and Sec31p were prepared in this laboratory. The monoclonal antibody against VSV-G (8G5) was prepared from ascites. The antibody against ribophorin II was kindly provided by Dr. D. Meyer (UCLA).
Cell Culture-Cell culture was performed as described (24). To examine p125-depleted HeLa cells by immunofluorescence microscopy, cells were grown on poly-L-lysine-coated coverslips.
Immunofluorescence and Immunoelectron Microscopy-Immunofluorescence microscopy was performed as described previously (31). Briefly, cells grown on coverslips were fixed with 4% paraformaldehyde, followed by sequential incubation with primary antibodies and fluorescein isothiocyanate-conjugated and/or Texas Red-conjugated secondary antibodies. Immunoelectron microscopy was performed as described (32).
Plasmid Construction and Transfection-cDNA clone AK058137 (Toyobo), which encodes aa 360 -872 of human PA-PLA 1 , was cleaved with KpnI and XhoI, and the resultant fragment was inserted into pBluescript SKϩ. To obtain the upstream region of human PA-PLA 1 cDNA, PCR was carried out using a human cDNA library (MATCHM-AKER human leukocyte cDNA library, Clontech) as a template and synthetic oligonucleotides (5Ј-gcgggtaccgagaattcaagatctatgaattacccgggccgcgg and 5Ј-ggcggtaccactacttgatgctttag) as sense and antisense primers, respectively. The sense primer contained a KpnI site and nucleotide positions 1-20 of human PA-PLA 1 , the sequence of which was taken from the human genome sequence data base. The antisense primer contained nucleotides complementary to positions 69 -91 of the AK058137 sequence, which includes a KpnI site. The amplified DNA fragment, comprising the putative initiation codon to the KpnI site of human PA-PLA 1 , was treated with KpnI and then inserted into the pBluescript-derived vector to generate the full-length cDNA of human PA-PLA 1 .
Chimeric protein constructs were produced by PCR and verified by DNA sequencing. Chimera A was composed of the N-terminal aa 1-234 of p125 fused to a KIAA0725p fragment starting from Ser-2. Chimera B was a fusion protein comprising the N-terminal aa 1-259 of p125 and the entire PA-PLA 1 . For the expression of proteins, Vero or 293T cells plated on 35-mm dishes were transfected with 1-2 g of expression plasmids using Lipofectamine PLUS reagent (Invitrogen) according to the manufacturer's instructions.
Membrane Binding Assays-Recombinant Sar1 proteins were purified as described previously (33). Microsome membranes and rat liver cytosol were prepared as described previously (16,34). Membrane binding of p125 was analyzed by using microsome membranes (20 -40 g) supplemented with recombinant Sar1 proteins and rat liver cytosol, as described for the analysis of Sec23 binding (35,36).
Construction of a HeLa Cell Line Expressing VSV-045G-A pcDNA3-based expression plasmid for VSV-045G, which encodes aa 1-511 of the VSV-045G protein with the C-terminal 4 aa (RTAA) followed by a T7 tag (MASMTGGQQMG), was constructed. The cDNA of VSV-045G was derived from pVSVG-GFP, which was a kind gift from Dr. J. Lippincott-Schwartz (National Institutes of Health). HeLa cells were transfected with the expression plasmid for VSV-045G and then selected for stable expression with 400 g/ml G418 (Invitrogen). Individual clones were cultured in the presence of 400 g/ml G418 and screened for expression by immunofluorescence microscopy using the anti-VSVG mAb. We sought to obtain clones that stably express VSV-045G, but all clones obtained gradually lost the ability to express the protein while retaining resistance to the selection marker. In this study, we used cells moderately expressing VSV-045G.
RNAi-The RNA duplexes used for targeting of p125 (aagttgaccatttggtgtttg) and lamin A/C (aactggacttccagaagaacatc) were purchased from Japan Bioservice, Inc. Transfection of HeLa cells was performed using Oligofectamine (Invitrogen). The final concentration of an RNA duplex was 200 nM. At 48 or 72 h after transfection, the cells were processed for immunoblotting and immunofluorescence.
VSV-045G Transport Assays-HeLa cells stably expressing VSV-045G were transfected with or without a duplex RNA and then incubated for 48 -55 h at 37°C. The cells were further incubated at 40°C (a total of 72 h after transfection) and then shifted to 32°C to allow protein transport. After incubation for the indicated times, the cells were fixed and examined by immunofluorescence microscopy. Alternatively, the cells were solubilized with 0.5% SDS and 1% 2-mercaptoethanol (0.1 ml/35-mm dish) and heated at 100°C for 10 min. A portion of the lysate was digested with Endo H (New England Biolabs), according to the manufacturer's protocol, and then subjected to SDS-PAGE on 10% gels. VSV-045G was visualized by immunoblotting with the polyclonal anti-VSV-G antibody, and the intensities of the immunostained bands were quantitated with NIH image software.

RESULTS
p125 Is Localized in ER Exit Sites-In a previous study, we used an ectopic expression system to determine the localization of p125, because the antibody we produced against a p125 peptide was not suitable for immunofluorescence analysis (24). When FLAG-tagged p125 was transiently expressed at low levels in cultured cells, it was found to be colocalized with ERGIC53, a well known marker protein for VTCs (37), and ␤-COP, a subunit of COPI located in both VTCs and the cis-Golgi (38).
Until recently, ER exit sites, VTCs, and the cis-Golgi could not be fully discerned as distinct compartments because of the lack of marker proteins for ER exit sites. However, recent studies revealed that ER exit sites are distinct from VTCs and the cis-Golgi in that they are characterized by COPII components (13,15,(17)(18)(19). To determine whether p125 is located in ER exit sites, we prepared monoclonal antibodies against p125. As an antigen, we chose the N-terminal 134 aa of p125 because this region exhibits no significant sequence homology with KIAA0725p or PA-PLA 1  By using mAb 1D4, we investigated the localization of endogenous p125 in HeLa cells. As shown in Fig. 1B, a vesicular staining pattern was observed throughout the cells but was concentrated in the perinuclear region (a, d, and g). Essentially the same staining pattern was observed with a polyclonal antibody (see Fig. 3B, b) that specifically recognizes p125 (data not shown). Double labeling with mAb 1D4 and a polyclonal antibody against Sec31p, a COPII component, showed that the two proteins were well colocalized (Fig. 1B, a-f). A similar colocalization pattern was observed for p125 and another COPII component Sec23p (Fig. 1B, g-i). On the other hand, p125-positive structures showed limited overlapping with ␤-COP-positive ones. In the perinuclear region, p125 staining was observed in concentrated dot-like structures (Fig. 1C, d), whereas ␤-COP staining revealed more closely connected ones (e).
Next, we analyzed the localization of p125 at the ultrastructural level. As shown in Fig. 2, immunogold labeling of p125 was observed in regions between the ER and Golgi but not within the Golgi stacks including the cis-Golgi area. This immunolabeling pattern is similar to that observed for Sec31p (39).
The localization of p125 in ER exit sites was confirmed in two ways. First, cells were treated with BFA. This reagent is known to inactivate Arf1, thereby leading to the dissociation of COPI from the Golgi, which is followed by redistribution of most Golgi membrane proteins to the ER (40). When cells were treated with BFA, ␤-COP staining changed from a perinuclear pattern to a diffuse, cytosolic one (Fig. 3A, e), whereas p125 retained the dot-like pattern (a and d) similar to Sec31p (b). Clearly, p125 was still colocalized with Sec31p in BFA-treated cells. Next, we examined the effect of expression of constitutive active Sar1p on p125 localization. Because uncoating of COPIIcoated vesicles is regulated via the GTP/GDP cycle of Sar1p (23), expression of the constitutively active form (Sar1p[H79G]) prevents uncoating, which leads to the generation of vesicular tubular membrane clusters containing COPII components and cis-Golgi matrix proteins (19). As shown in Fig. 3B, p125 (b), similar to Sec31p (e), was observed in perinuclear clusters in FLAG-Sar1p[H79G]-expressing cells. These immunofluorescence and electron microscopic results taken together strongly suggest that p125 is primarily localized in ER exit sites rather than the cis-Golgi.
p125 Is Recruited to Microsome Membranes in an Active Sar1p-dependent Manner-Given that p125 is present in Arrows indicate the position of the ER. The Golgi apparatus (G) was slightly dilated due to brief fixation for immunoelectron microscopy. Immunogold labeling for p125 was found around the Golgi but not in the ER nor within the Golgi stacks in normal rat kidney cells. Bar, 0.5 m. A, a perinuclear region; B, a region rich in ER. Sar1p[H79G]-induced clusters containing a COPII component, we wondered whether p125 is recruited to membranes in an active Sar1p-dependent manner. To test this, we performed in vitro membrane binding assays (Fig. 4). Microsome membranes were incubated with rat liver cytosol and increasing concentrations of Sar1p[H79G] under conditions that support COPII assembly (35,36). At the end of the incubations, the membranes were washed and collected by centrifugation, and the binding of p125 to the membranes was assessed by immunoblotting with mAb 1D4. Activated Sar1p led to enhancement of p125 association with the membranes (Fig. 4A, compare lane 1  to lanes 2-5). A minor signal was observed when microsome membranes were omitted from the incubation, attesting to the specificity of membrane binding (Fig. 4A, lane 6). The enhancement of p125 interaction with microsome membranes was dependent on Sar1p activation. Incubation of microsome membranes with a Sar1p mutant that mimics the GDP-bound form (Sar1p[T39N]) did not support COPII recruitment (analyzed by the membrane binding of Sec23p), as reported previously (6), nor did it enhance p125 interaction with microsome membranes (Fig. 4B, compare lanes 5 and 6 to lanes 3 and 4). This biochemical result complements the morphological observation that p125 is redistributed to Sec31-enriched perinuclear clusters in cells which express constitutively active Sar1p.
Overexpression of p125 Causes Perturbation of the Structures of ER Exit Sites, VTCs, and the cis-Golgi-Our previous study showed that expression of p125 at high levels in cells causes dispersion of ␤-COP and ERGIC53, whereas cis-Golgi matrix proteins such as p115 (41) and GM130 (42) coexist with overexpressed p125 in clusters formed in the perinuclear region (24,25). We examined whether the distribution of COPII changes upon ectopic expression of p125 (Fig. 5A). When expressed, FLAG-p125 was colocalized with Sec31p (Fig. 5A, a-c) and Sec23p (d-f), demonstrating that expressed FLAG-p125 retains the targeting ability. It should be noted that in cells expressing FLAG-p125 at high levels, Sec31p and Sec23p were not dispersed but rather colocalized with overexpressed p125 in aggregated structures, as observed for p115 and GM130 (25). These results imply that overexpression of p125 perturbs the structure or organization of ER exit sites. Given that ER exit sites are located upstream of VTCs and the cis-Golgi in the secretory pathway, it is reasonable to speculate that disorganization of VTCs and the cis-Golgi is a consequence of perturbation of ER exit sites. The ability of p125 to disturb the structure of ER exit sites, VTCs, and the cis-Golgi explains why expressed p125, which is supposed to be principally targeted to ER exit sites, is colocalized with VTCs and/or ERGIC markers (ERGIC53 and ␤-COP) (24). When ectopically expressed at low levels, p125 may induce local coalescence of ER exit sites with VTCs and the cis-Golgi.
In mammals, the PA-PLA 1 family consists of three members, p125, KIAA0725p, and PA-PLA 1 . Our previous study involving an expression system showed that the localization of KIAA0725p may be similar but significantly different from that of p125 (28). The localization of PA-PLA 1 has not been investigated in detail so far, although it was originally purified from the cytosolic fraction (26). We examined the effect of overexpression of FLAG-KIAA0725p or FLAG-PA-PLA 1 on the organization of ER exit sites (Fig. 5B). Expressed FLAG-KIAA0725p, like expressed FLAG-p125, was observed in dotlike structures and/or aggregates (Fig. 5B, a-c), whereas FLAG-PA-PLA 1 was distributed exclusively in the cytosol regardless of its expression level (d). In contrast to the case of p125 overexpression, overexpression of KIAA0725p or PA-PLA 1 had no marked effect on the organization of ER exit sites. This confirms the specificity of the effect of p125 overexpression on ER exit sites and suggests that the function and localization of KIAA0725p and PA-PLA 1 are different from those of p125.
N-terminal Region of p125 Is Required for the Localization to ER Exit Sites-We hypothesized previously (25) that the putative phospholipase domain of p125 mediates membrane attachment, and the N-terminal region determines membrane specificity. Given that KIAA0725p associates with membrane compartments different from p125-localized ones, and that PA-PLA 1 does not bind to membranes, these two proteins appear to be ideal for the chimeric approach to test our hypothesis.
We constructed two chimeric proteins, chimeras A and B (Fig. 6A). As KIAA0725p exhibits similarity with the sequence starting from aa 235 of p125, the N-terminal region of p125 (aa 1-234) was fused to KIAA0725p to construct chimera A. Chimera B was a fusion protein comprising the N-terminal aa 1-259 of p125 and PA-PLA 1 . When expressed at moderate levels, FLAG-chimera A was observed in large aggregates in cells (Fig. 6B, a), similar to the case of KIAA0725p and p125. In these cells, Sec31p existed in the FLAG-chimera A-positive aggregates (Fig. 6B, b and c). On the other hand, FLAG-chimera B, when expressed, was localized in the cytosol (Fig. 6B,  d), as observed for PA-PLA 1 . These results imply that the N-terminal region is important for targeting to ER exit sites and that the putative lipase domain is interchangeable with KIAA0725p but not with PA-PLA 1 , thus confirming our previous hypothesis. In cells overexpressing chimera B, the number of Sec31p-positive dots decreased depending on the expression level (Fig. 6B, e). It is likely that cytosolic chimera B substantially induced the dissociation of COPII components from the membrane, perhaps by interacting with Sec23p.
Depletion of p125 Affects the Distribution of ER Exit Sites but Not the Transport of VSV-045G from the ER-To gain an insight into the function of p125, p125 mRNA was targeted for degradation by RNAi. A p125-specific RNA duplex was transfected into HeLa cells, and then the cells were analyzed for p125 protein expression by immunoblotting and immunofluorescence microscopy. As shown in Fig. 7A, lane 3, the amount of p125 protein decreased by about 10-fold in p125-targeted cells, whereas the expression levels of Sec23p and actin did not change. Concomitant with the reduction in the amount of p125 protein, the intensity of immunofluorescence substantially decreased (Fig. 7B, b), confirming that our antibody is able to specifically recognize p125 on immunofluorescence analysis.
In p125-depleted cells, the distribution of Sec23p was changed (Fig. 7B, a). In control cells, Sec23p was observed in numerous vesicular structures distributed throughout the cells but concentrated in the perinuclear region. It has been suggested that this concentration results from the slow movement of newly formed ER exit sites from the cell periphery to the central region of the cell (15). In p125-depleted cells, the intensity of Sec23p staining in the perinuclear region was markedly reduced, and Sec23p-positive dots were observed equally throughout cells. The same change in the distribution pattern was observed for Sec31p (data not shown). On the other hand, no marked change in the distribution of Sec31p or Sec23p was observed in mock-treated cells (Fig. 7B, c) or ones treated with a lamin A/C RNA duplex (data not shown). These results suggest that p125 plays a role in the organization of ER exit sites in cells. In p125-depleted cells, the staining pattern of ␤-COP was also changed (Fig. 7B, e). ␤-COP lost its compact perinuclear localization, showing a relatively diffuse pattern. In contrast, the staining pattern of mannnosidase II, a medial-Golgi protein, did not significantly change in cells depleted of p125 (Fig. 7B, f). We reasoned that the change in ␤-COP localization is a consequence of the perturbation of ER exit sites. To test this idea, we analyzed an early stage of p125 depletion (48 h after transfection of the p125 RNA duplex). As shown in Fig.  7C, the staining pattern of Sec23p had already changed after 48 h (Fig. 7C, a), whereas the staining pattern of ␤-COP had The positions of the consensus motif for lipases (Gly(Ser)-X-Ser-X-Gly; where X represents any aa) are indicated. B, Vero cells were transfected with the plasmid for FLAG-chimera A (a chimeric protein comprising p125 and KIAA0725p) or FLAG-chimera B (a chimeric protein comprising p125 and PA-PLA 1 ). At 24 h after transfection, the cells were fixed and double-stained with the antibodies against FLAG (a and d) and Sec31p (b and e). c and f present merged images. not (Fig. 7C, e). These results suggest that depletion of p125 principally affects the distribution of ER exit sites, which leads to a change in ␤-COP localization as a secondary effect.
To assess the effect of p125 depletion on protein transport, we measured the transport of VSV-045G from the ER in p125depleted cells (Fig. 8). First, we performed a morphological assay. HeLa cells stably expressing VSV-045G were transfected with the p125-specific RNA duplex and then incubated at the nonpermissive temperature. When kept at the nonpermissive temperature, VSV-045G was present in the ER (Fig.  8A, 0 min). At 30 -60 min after the temperature shift to the permissive one, VSV-045G was detected in the perinuclear Golgi area in mock-transfected cells (ERGIC/Golgi). In p125depleted cells, VSVG-045G was also observed in the perinuclear region (Fig. 8A, 30 min) and colocalized with ␤-COP (data not shown). At 2 h, VSV-045G had reached the plasma mem- brane in both p125-depleted and control cells. Semi-quantitative analysis (Fig. 8A, right panel) revealed that depletion of p125 does not markedly inhibit VSV-045G transport from the ER. This conclusion was supported by the result from a biochemical VSV-045G transport assay based on the acquisition of Endo H resistance, an event that occurs in the medial-Golgi. Delay in the acquisition of Endo H resistance of VSV-045G was little, if any, in p125-depleted cells (Fig. 8B). DISCUSSION p125 was identified by affinity chromatography on mammalian Sec23p-coupled resins (24). Our previous studies involving a mammalian expression system showed that p125 is probably localized in VTCs and/or the cis-Golgi compartment and that both the N-terminal proline-rich region, which is responsible for the interaction with Sec23p, and the putative phospholipase domain are required for its localization (24,25). In this study, we investigated the localization of p125 using monoclonal and polyclonal antibodies that specifically recognize endogenous p125, and we assessed its function by means of RNAi.
Several lines of evidence suggest that p125 is localized in ER exit sites rather than the cis-Golgi. First, immunofluorescence analysis showed the colocalization of p125 with ER exit site markers, i.e. COPII components (Sec31p and Sec23p). Second, immunoelectron microscopy revealed the presence of p125 in the vicinity of the Golgi, a region where the COPII components are located, but not within the Golgi stacks. Third, similar to that observed for COPII components (19), no redistribution of p125 took place upon treatment with BFA, a Golgi-disrupting reagent (40). Fourth, expression of constitutively active Sar1p caused p125 to be redistributed to the perinuclear clusters positive for Sec31p, and recombinant constitutively active Sar1p enhanced the association of cytosolic p125 with microsome membranes in vitro.
Based on the results obtained using truncation mutants, we proposed previously that the phospholipase domain of p125 is a primary determinant for membrane attachment and that the N-terminal Sec23p-intertacting region coordinates membrane specificity (25). The present study involving chimeric proteins comprising p125 and other members of the mammalian PA-PLA 1 family verified this hypothesis. The N-terminal region of p125 is able to direct membrane-associated KIAA0725p to ER exit sites but has no ability to direct PA-PLA 1 , a bona fide cytosolic protein, to ER exit sites or any other membrane compartment. Perhaps after membrane attachment, the N-terminal domain of p125 specifies the location by interacting with Sec23p. It is interesting that p125 and KIAA0725p bind to membranes, whereas PA-PLA 1 is present almost exclusively in the cytosol. A future study involving comparison of lipid binding specificity between the membrane-associated and cytosolic members of the PA-PLA 1 family will provide an insight into the mechanism underlying membrane association.
Overexpression or depletion studies suggested that p125 is involved in the architecture of ER exit sites. Overexpression of p125 caused coalescence of ER exit sites with VTCs and perhaps also with the cis-Golgi. This phenotype is similar to that observed when constitutively active Sar1p was expressed in cells. In contrast to the formation of the perinuclear COPII aggregates in p125-overexpressing cells, the perinuclear concentration of ER exit sites, which is observed in normal cells, was decreased in cells depleted of p125 by means of RNAi. Concomitantly, the structure of the cis-Golgi was moderately disturbed. The disturbance of the cis-Golgi structure probably results from the perturbation of ER exit sites, because the change in the cis-Golgi structure occurred much later after that in the distribution of ER exit sites.
It should be noted that depletion of p125 changed the distribution of ER exit sites but not their size or shape at least at the microscopic level. Previous work involving yellow fluorescence protein-tagged Sec23p in mammalian cells demonstrated that the number of ER exit sites increases greatly during interphase FIG. 8. Depletion of p125 does not markedly inhibit the intracellular transport of VSV-045G. A, HeLa cells stably expressing VSV-045G were transfected without (Mock) or with (KD) an RNA duplex specific for p125. At 72 h after transfection, transport of VSV-045G from the ER was monitored as described under "Experimental Procedures." At 0, 30, 60, and 120 min after a temperature shift to 32°C, the cells were fixed and stained with the monoclonal anti-VSV-G antibody. Quantitative data represent the average of two independent experiments. About 200 VSV-045G-expressing cells were examined for each time point. B, VSV-045G-expressing cells were treated as described above, except that the cells were lysed, subjected to Endo H treatment, and analyzed by immunoblotting with the polyclonal anti-VSV-G antibody. A representative experiment is shown. R and S denote Endo H-resistant and -sensitive forms, respectively. Asterisks indicate a nonspecific immunostained band. The amounts of the Endo H-resistant form were determined and plotted. Quantitative data represent the average of two independent experiments. and newly formed peripheral ER exit sites are slowly translocated to the juxtanuclear area of cells (15). This movement seems to be independent of cargo transport from the ER to the Golgi (15). Because the perinuclear concentration of ER exit sites is decreased in p125-depleted cells, it is possible that cells depleted of p125 have defects in this process.
To our surprise, transport of VSV-045G proceeded normally in cells depleted of p125. This implies that the change in the distribution of ER exit sites and the cis-Golgi does not have a significant effect on membrane transport from the ER. Although we could not exclude the possibility that a trace amount of p125 is sufficient to maintain membrane transport or that p125 is involved in the transport from the ER of certain cargo molecules other than VSV-G, our results at least suggest that p125 is not a structural component, such as Sec23p and Sar1p, of the general transport machinery. The present observations are reminiscent of the recent work by Kondylis and Rabouille (43), who reported that although depletion of p115 in Drosophila S2 cells leads to changes in the organization of the Golgi stack and ER exit sites, intracellular protein transport is only marginally impaired. Several studies have revealed that membrane transport occurs, albeit with reduced efficiency, in the absence of the integrity of the Golgi apparatus (44,45). It is possible that membrane transport from the ER does not strictly require the morphological integrity of ER exit sites and the Golgi. This is true for Saccharomyces cerevisiae. In S. cerevisiae, COPII vesicles are formed from the entire ER and not from a specialized ER subdomain, and the Golgi is scattered throughout the cells (12). Nevertheless, efficient membrane traffic is maintained in this organism.
Recently, PA-PLA 1 family members were discovered not only in mammals but also in plants and yeast. SGR2, a PA-PLA 1 family member that exists in a higher plant, Arabidopsis, was found to be localized in vacuoles and may function in shoot gravitropism by regulating membrane systems (29,30). In S. cerevisiae, the YOR022C gene encodes a protein similar to PA-PLA 1 . Global localization analysis has suggested that the YOR022c protein is localized in mitochondria (46). These findings combined with our results and the results of others (24,25,47, and this study) suggest that PA-PLA 1 family proteins appear to be localized in diverse subcellular compartments and may be involved in various biological processes.