Proteomics of Endoplasmic Reticulum-Golgi Intermediate Compartment (ERGIC) Membranes from Brefeldin A-treated HepG2 Cells Identifies ERGIC-32, a New Cycling Protein That Interacts with Human Erv46* □ S

Cycling proteins play important roles in the organiza-tion and function of the early secretory pathway by par-ticipating in membrane traffic and selective transport of cargo between the endoplasmic reticulum (ER), the intermediate compartment (ERGIC), and the Golgi. To identify new cycling proteins, we have developed a novel procedure for the purification of ERGIC membranes from HepG2 cells treated with brefeldin A, a drug known to accumulate cycling proteins in the ERGIC. Membranes enriched 110-fold over the homogenate for ERGIC-53 were obtained and analyzed by mass spectrometry. Major proteins corresponded to established and putative cargo receptors and components mediating protein maturation and membrane traffic. Among the uncharacterized proteins, a 32-kDa protein termed ERGIC-32 is a novel cycling membrane protein with sequence homology to Erv41p and Erv46p, two proteins enriched in COPII vesicles of yeast. ERGIC-32 localizes to the ERGIC and partially colocalizes with the human homologs of Erv41p and Erv46p, which mainly localize to the cis -Golgi. ERGIC-32 processed for immunofluorescence in 1% BSA containing PBS. Confocal laser-scanning microscopy pictures were acquired with a TCS NT microscope (Leika). To determine the topology of ERGIC-32, cells were fixed with 3% paraformaldehyde on ice, and the plasma membrane was permeabilized with 20 (cid:2) g/ml digitonin (Calbiochem-Novabiochem) for 5 min on ice and washed with PBS. The cells were then further permeabilized or not with 0.1% saponin and processed for immunofluores- cence microscopy. Immunofluorescence was recorded with a Zeiss Axio-vert 135M microscope coupled to a charge-coupled device camera. RNA Interference— ERGIC-32 gene expression was silenced using synthesized small interfering RNA duplexes (Eurogentech, Liege, Bel- gium). The targeted sequences of ERGIC-32 were AACGAGCTCTAT-GTCGATGAC and AATACGTCGCCTACAGCCACA for si1 and si2, respectively. Their specificity for the ERGIC-32 mRNA was confirmed by a BLASTn search in NCBInr database. Similarly the oligonucleotide AGCAGGGAATTTTCACATA was designed to interfere specifically with hErv41 gene expression. As a negative control, we used the oligonucleo- tide AATTCTCCGAACGTGTCACGT, which gives no significant match in a BLASTn search (human NCBInr database). Oligonucleotides were transfected into HeLa cells with RNAifect (Qiagen) according to the recommendations of the manufacturer. In typical experiments, 50–80% confluent cells grown in 6-well plates were transfected with 5 (cid:2) g of duplex siRNA and 20 (cid:2) l of RNAifect/well. 48 h later the cells were trypsinized and seeded into a fresh well before a new round of transfection. 48 h after the second transfection, the cells were subcultured for subsequent experiments.

drug has little effect on the ERGIC, which keeps its identity (32), although the ERGIC clusters become larger and more uniformly distributed in the cytoplasm of the cells. Several cycling proteins have been shown to accumulate in the ERGIC clusters upon BFA treatment, including ERGIC-53 (32,34), KDEL-R (11,35), and proteins of the p23/24 family (13,36). Based on these observations it may be speculated that the purification of the ERGIC from cells treated with BFA could lead to the identification of novel cycling proteins and thereby provide new insight into the functions of this compartment as well as ER to Golgi transport.
Here we describe a new procedure to highly purify ERGIC membranes from BFA-treated HepG2 cells and to identify major ERGIC proteins by mass spectrometry. Among novel cycling proteins we have characterized ERGIC-32, a new ERGIC protein that interacts with human Erv46, a protein previously characterized in yeast and functioning in ER to Golgi protein trafficking.
Cells-HepG2 and HeLa cells were grown in minimum Eagle's medium and Dulbecco's modified Eagle's medium, respectively, supplemented with 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, and 1 g/ml fungizone (Sigma). Cells were transfected with FuGENE TM 6 transfection reagent according to the supplier's recommendations (Roche Diagnostics). Stable cell lines were obtained by selection with 1 mg/ml Geneticin (Invitrogen). Expression of human Erv41 (hErv41)-Myc was induced by incubation for 16 h with 10 mM butyric acid (Sigma).
Purification of ERGIC Membranes-Subcellular fractionation ( Fig.  1A) was performed essentially according to Klumperman et al. (7) using Nycodenz gradients (Nycomed, Oslo, Norway). 5 days postconfluent HepG2 cells (one 150-mm diameter dish/gradient) were treated with 10 g/ml brefeldin A (Epicenter Technologies, Madison, WI) for 90 min and rinsed with 150 mM NaCl and homogenization buffer (buffer H: 120 mM NaCl, 5 mM KCl, 25 mM NaHCO 3 , pH 7.4). Cells were homogenized in buffer H supplemented with 40 g/ml phenylmethylsulfonyl fluoride (Sigma) using a ball-bearing homogenizer (45). A postnuclear supernatant was prepared by centrifugation at 100 ϫ g for 10 min. The postnuclear supernatant (2 mg of proteins) was loaded on top of a linear 13-29% Nycodenz gradient with a 35% cushion. Nycodenz solutions contained 1 mM EDTA and 10 mM triethanolamine, pH 7.4, with the exception of the 13% solution that contained additional 120 mM NaCl. After centrifugation at 80,000 ϫ g for 3 h, fractions of 1.2 ml were collected from the bottom of the tube. Fractions 9 -12, which are specifically enriched in ERGIC-53 and KDEL-R, were pooled and diluted five times with Nycodenz buffer containing 0.1% BSA (Sigma). Prewashed M-450 Dynabeads coupled to goat anti-mouse IgG (Dynal Biotech GmbH, Hamburg, Germany) were incubated overnight with the mAb against KDEL-R in PBS, 0.1% BSA and washed four times with this buffer and once with Nycodenz buffer, 0.1% BSA. 1 ml of diluted membranes was incubated with 0.2 ml of the anti-KDEL-R magnetic beads for 2 h at 4°C with gentle end-over-end mixing in the presence of protease inhibitors. Beads and membranes were washed four times with PBS, 0.1% BSA, 1 mM EDTA and twice with PBS, 1 mM EDTA. 30 separate immunopurification reactions were carried out in parallel. The washed beads were pooled and treated with 0.1 M NaHCO 3 , pH 11.5, for 30 min on ice. Membranes were recovered in the supernatant after this treatment and pelleted by centrifugation at 100,000 ϫ g for 1 h. The membranes were resuspended in Laemmli sample buffer containing 6.5 mM iodoacetamide (Sigma) and loaded on a 6 -12% SDS gel. Protein was determined with the BCA assay (Pierce). Markers of the different organelles were analyzed by Western blotting and quantified with ChemImager TM (Alpha Inotech Corp.).
Mass Spectrometry-Slices of silver-stained gels (46) were washed five times (1 min each) with 30 l of 40% n-propanol followed by five washes with 30 l of 0.2 M NH 4 HCO 3 containing 50% acetonitrile. The gel slices were dried in a SpeedVac concentrator (Savant, Farmingdale, NY) and reswelled in 0.5 g of trypsin (sequencing grade-modified trypsin, Promega, Madison, WI) in 10 l of 100 mM NH 4 HCO 3 . Enough digestion buffer (100 mM NH 4 HCO 3 ) was added to completely immerse the gel slices in the liquid. Digestion was at 37°C overnight. The supernatant was collected, and the gel slices were extracted with 15 l of 0.1% formic acid for 5 min followed by 15 l of acetonitrile for 1 min. Extraction was repeated twice, and all supernatants were pooled and dried in a SpeedVac concentrator. For mass spectral analysis, the peptides were redissolved in 10 l of 0.1% trifluoroacetic acid, and 5 l was used for identification by liquid chromatography MS/MS. The peptides were separated on fused silica capillaries (100 m inner diameter ϫ 280 m outer diameter, LC-Packings) packed with POROS R2 material with a linear gradient from 0 to 80% methanol in 0.02% acetic acid at a flow rate of 200 nl/min. Mass spectral data were acquired on a TSQ7000 triple quadrupole instrument (Finnigan, San José, CA), with data-controlled switching between precursor ions and daughter ions during single chromatographic runs. The daughter ion spectra were used to identify the proteins with the Mascot search software (www.matrixscience.com).
Recombinant DNA-The cDNA clone HF00712 (47) corresponding to the gene KIAA1181 and encoding ERGIC-32 was cloned by the HUGE (human unidentified gene encoded) project and kindly provided by the Kasuza DNA Research Institute (Chiba, Japan). This cDNA contains a 5Ј-untranslated region. No EST with an initiation codon upstream of the first one found in this cDNA clone was identified in the NCBI databases. hErv41 (NCBI accession number gi͉7706105͉), hErv46 (gi͉4929577͉), Surf-4 (gi͉12644052͉), and hYif1p (gi͉12654907͉) were cloned by RT-PCR using different sets of specific primers: 5Ј-GCG GTA CCC CTG AAG ATG AGG CGA CTG AAT C-3Ј and 5Ј-CGT CTA GAC  TTC AAT CGG GAG GTG TTA ATG TG-3Ј for hErv41, 5Ј-CGA ATT  CCC TTT CCG GCC GGT CCC CAT G-3Ј and 5Ј-CGT CTA GAC CGA  GGG TGA CTA CGT TGT CTT CC-3Ј for hErv46, 5Ј-CGA ATT CCA  CCA TGG CTT ATC ACT CGG GCT ACG GA-3Ј and 5Ј-GCT CTA GAT  GGG GCC AGG GGG TCA CC-3Ј for hYif1p, and 5Ј-CGA ATT CCA CCA  TGG GCC AGA ACG ACC TGA TG-3Ј and 5Ј-GCT CTA GAA CCA GTC  CTT GAC GGC CAC GG-3Ј for Surf-4. Total RNAs were prepared from HepG2 cells using the Peq-gold RNA pure kit (Peqlab Biotechnology GmbH, Herlangen, Germany). cDNAs were amplified with the Omniscript TM kit (Qiagen, Basel, Switzerland). RT-PCR-specific products were cloned and sequenced. HA-and Myc-tagged versions of ERGIC-32, hErv41, and hErv46 were obtained by PCR, cloned, and analyzed for correct sequence. All constructs were cloned into pcDNA3 vector except GST-ERGIC-32, which was cloned into pGEX-5X2.
Immunofluorescence Microscopy-Cells grown in 8-well multichamber glass slides (Miles Laboratories, Naperville, IL) were fixed with 3% paraformaldehyde, permeabilized with 0.1% saponin in PBS, and processed for double immunofluorescence microscopy (5). Alternatively, an acidified ethanol fixation/permeabilization procedure was used for anti-ERGIC-32. Cells were washed in cold PBS, dehydrated in 95% ethanol, 5% acetic acid at Ϫ20°C for 5 min, washed three times with PBS, and processed for immunofluorescence in 1% BSA containing PBS. Confocal laser-scanning microscopy pictures were acquired with a TCS NT microscope (Leika). To determine the topology of ERGIC-32, cells were fixed with 3% paraformaldehyde on ice, and the plasma membrane was permeabilized with 20 g/ml digitonin (Calbiochem-Novabiochem) for 5 min on ice and washed with PBS. The cells were then further permeabilized or not with 0.1% saponin and processed for immunofluorescence microscopy. Immunofluorescence was recorded with a Zeiss Axiovert 135M microscope coupled to a charge-coupled device camera.
RNA Interference-ERGIC-32 gene expression was silenced using synthesized small interfering RNA duplexes (Eurogentech, Liege, Belgium). The targeted sequences of ERGIC-32 were AACGAGCTCTAT-GTCGATGAC and AATACGTCGCCTACAGCCACA for si1 and si2, respectively. Their specificity for the ERGIC-32 mRNA was confirmed by a BLASTn search in NCBInr database. Similarly the oligonucleotide AGCAGGGAATTTTCACATA was designed to interfere specifically with hErv41 gene expression. As a negative control, we used the oligonucleotide AATTCTCCGAACGTGTCACGT, which gives no significant match in a BLASTn search (human NCBInr database). Oligonucleotides were transfected into HeLa cells with RNAifect (Qiagen) according to the recommendations of the manufacturer. In typical experiments, 50 -80% confluent cells grown in 6-well plates were transfected with 5 g of duplex siRNA and 20 l of RNAifect/well. 48 h later the cells were trypsinized and seeded into a fresh well before a new round of transfection. 48 h after the second transfection, the cells were subcultured for subsequent experiments.

Purification of ERGIC Membranes from BFA-treated Cells-
To identify new cycling proteins in the early secretory pathway, we developed a procedure for the purification of ERGIC membranes from HepG2 cells treated with BFA. The rationale of using BFA is 2-fold. BFA leads to an accumulation of rapidly cycling proteins in the ERGIC, as exemplified by ERGIC-53 and KDEL-R, and to the fusion of the Golgi apparatus with the ER. The latter effect facilitates the purification of ERGIC membranes which, in the absence of BFA, largely cofractionate with trans-Golgi membranes on density gradients (7). In a first step, a postnuclear supernatant of BFA-treated HepG2 cells was separated by Nycodenz gradient centrifugation. ER, Golgi, and plasma membrane distributed to the bottom fractions of the Nycodenz gradient, whereas ERGIC membranes, identified by ERGIC-53 and KDEL-R, floated to the top fractions ( Fig. 1A; also see supplementary Fig. 1). The ERGIC-containing top fractions were still contaminated by endosomes (transferrin receptor, not shown), lysosomes (LAMP-1), trans-Golgi network (TGN46), and some ER (BAP31) (Fig. 1B, lane T). In a second step, these contaminating membranes were removed by immunopurification. The pooled top fractions were incubated with magnetic beads to which an antibody against a cytoplasmically exposed epitope of the KDEL-R was coupled (Fig. 1). The resulting membrane fraction was devoid of detectable transferrin receptor, LAMP-1, TGN46, and BAP31 but contained some antibodies that had leaked from the immunobeads. The immunopurified membranes were then subjected to the classical carbonate, pH 11.5, treatment to remove the antibodies as well as peripheral proteins. The final ERGIC fraction was 110 Ϯ 19.5-fold (mean Ϯ S.E., n ϭ 3) enriched over the homogenate.
Proteomics of ERGIC Membranes-To determine the protein composition of the purified ERGIC membrane fraction we scaled-up the purification and separated the proteins by 6 -12% SDS-PAGE. One-dimensional instead of two-dimensional gels were used because of the unresolved difficulty of solubilizing membrane proteins by isoelectric focusing (50). 30 bands in a molecular mass range of 20 to more than 300 kDa were reproducibly detected on silver-stained gels (Fig. 2). Bands were excised, trypsin-digested in gel, and analyzed in parallel by MALDI-TOF MS and nano-electrospray tandem mass spectroscopy techniques. 19 proteins were unambiguously identified by screening NCBInr databases. Several proteins were identified with both techniques, whereas some could only be detected with one. A refinement of the gel analysis using SDS gels containing urea, to better solubilize membrane proteins and resolve low molecular weight proteins, led to the identification of five more proteins, bringing the total number to 24 identified proteins ( Table I).
The identified proteins can essentially be grouped into four classes: cargo receptors, proteins involved in membrane traffic, chaperones involved in protein maturation, and uncharacterized proteins (Table I). ERGIC-53 is apparently the most abundant protein of the purified ERGIC fraction. VIP36, another lectin that is related to ERGIC-53 (51,52), was also identified, as well as p24A, Gp25L2, Tmp21, and CGI-100, four members of the p24 protein family. In accord with these results, previous studies have shown that VIP36 cycles between the ER and Golgi (53), although it may also operate at a post-Golgi level in some cells (54). P24 family members are known to cycle in the early secretory pathway, both in higher eukaryotes and in yeast, and to operate as cargo receptors linking cargo to COPI and COPII coats (13,(55)(56)(57)(58).
Among the proteins having functions in membrane trafficking, the small GTPase Rab1 and the t-SNARE Sec22b are already known to be associated with the ERGIC (18,21,22). The largest protein in the ERGIC fraction was identified as giantin, a cis/medial Golgi protein implicated in intra-Golgi traffic (39,59,60). By immunofluorescence microscopy we confirmed that giantin indeed colocalizes in part with ERGIC elements in BFA-treated HepG2 cells (not shown) as previously shown for other cells (61).
Some luminal chaperones associated mainly with the ER were also identified, including BIP and protein disulfide isomerase. These proteins carry a COOH-terminal KDEL motif that allows them to cycle between ER and Golgi (62). Class 1A Man9-mannosidase is an enzyme involved in the maturation of glycoproteins (63), and CBP1/Hsp47 is a procollagen-binding chaperone involved in procollagen folding and transport (64). Finally, CALNUC (nucleobindin1), a cis-Golgi protein implicated in calcium homeostasis, and its homolog NEFA (nucleo-bindin2), were present in the fraction (65). The overall composition of the ERGIC fraction confirms its high purity, because it is devoid of marker membrane proteins for other organelles. The presence of some ER chaperones is unlikely to be due to cross-contamination with ER membranes, because major ER membrane proteins such as BAP31 were not detectable in this fraction by immunoblotting.
hYif1p and Surf-4 have not been described as cycling proteins before and are mainly uncharacterized. hYif1p is the human ortholog of yeast Yif1p a protein enriched in COPII vesicles (66). Surf-4 (surfeit locus 4) gene belongs to a cluster of highly conserved housekeeping genes (67), and the corresponding protein is the human ortholog of Erv29p of yeast. Erv29p is also enriched in COPII vesicles and operates as a transport receptor for specific cargo in yeast (66,68). Surf-4 encodes a 30-kDa protein localized in the early secretory pathway, but its function is unknown (69). KIAA1181 is a putative protein of 290 amino acids the cDNA of which was cloned by the Kasuza project. The protein has not been characterized yet (47).
To verify the proposed recycling of these newly identified proteins, KIAA1181, Surf-4, and Yif1p were HA-tagged and expressed in HepG2 cells. All three proteins localized in the early secretory pathway but displayed slightly different distributions, Surf-4 being more enriched in the ER compared with KIAA1181 and Yif1p (Fig. 3). Upon BFA treatment these proteins redistributed to membrane spots in the cytoplasm that co-labeled for KDEL-R (Fig. 3). Thus, the three proteins behave like established cycling proteins. These findings further vali-date our procedure regarding its suitability for identifying cycling proteins of the early secretory pathway (also see enrichment of ERGIC-32 along the purification procedure in supplementary Fig. 1).
ERGIC-32 Is Homologous to Erv46p and Erv41p-For the remainder of this report we will focus on the characterization of KIAA1181. The corresponding gene is localized on human chromosome 5 at locus 5q35.2 and comprises 10 exons. The cDNA was cloned by the Kasuza DNA Research Institute from human adult brain (clone hf00712a) and contains an open reading frame of 800 base pairs and 5Ј-and 3Ј-untranslated regions (47). Conceptual translation yields a protein of 290 amino acid residues and a predicted molecular mass of 32.5 kDa in agreement with the apparent molecular mass on SDS gels (Fig. 2). We designate the protein ERGIC-32 (Swiss-Prot accession number Q969X5) referring to its molecular weight and its enrichment in ERGIC membranes (see below). ESTs corresponding to ERGIC-32 were found in NCBI EST databases for most tissues, indicating that ERGIC-32 is ubiquitously expressed in agreement with the RT-PCR enzyme-linked immunosorbent assay data provided by the Kasuza human cDNA project (www.kazusa.or.jp/huge/). Data base searches also revealed orthologs of ERGIC-32 in mouse and Caenorhabditis elegans with sequence similarities of 98 and 55%, respectively (Fig. 4A). The function of these proteins is unknown. Moreover, two human proteins, CGI-54 (also named "putative 43.2-kDa protein") and CDA14, with sequence identity scores of 34 and 21% were identified (Fig. 4B). These proteins are the human orthologs of Erv46p and Erv41p, two proteins enriched in ERderived COPII vesicles (66). ERGIC-32 shares higher homologies with Erv46p (34 and 26% identity with human and yeast Erv46, respectively) than with Erv41p (21 and 18% identity) ( Fig. 4B). Obviously ERGIC-32, Erv46p and Erv41p belong to the same protein family (Fig. 4D).
Molecular Characterization of ERGIC-32-The hydrophobicity profile of ERGIC-32 suggests two hydrophobic domains at the NH 2 and COOH termini, each of sufficient length to span the membrane bilayer, but the predicted protein sequence lacks a typical NH 2 -terminal signal sequence (Fig. 4, A and C). According to TopPred2 predictions (70), the NH 2 -and COOHterminal ends of the protein are on the cytoplasmic side, whereas the greater part of the protein in between the two transmembrane domains is luminally exposed. The luminal domain carries a consensus site for N-glycosylation (Fig. 4A). To further characterize ERGIC-32, we raised a rabbit polyclonal antibody against a recombinant GST-ERGIC-32 fusion protein, which immunoprecipitated a protein with the expected M r from HepG2 cells metabolically labeled with [ 35 S]methionine (Fig. 5A, lane 1). The reaction could be quenched with purified recombinant GST-ERGIC-32 (Fig. 5A, lane 3), and as expected, a stronger signal with a slightly higher M r was obtained with cells expressing HA-tagged ERGIC-32, confirming the specificity of the antibody (Fig. 5A, lane 2). ERGIC-32 is N-glycosylated as probed with endoglycosidase H but does not appear to undergo complex glycosylation as revealed by pulsechase experiments (Fig. 5B). ERGIC-32 is a bona fide membrane protein because it remains membrane-associated after carbonate, pH 11.5, treatment (Fig. 5C) like the type 1 membrane protein ERGIC-53 (5) but unlike the soluble receptorassociated protein RAP (71).
The membrane topology of ERGIC-32 was probed in situ by antibody accessibility in cells permeabilized with digitonin alone or digitonin plus saponin. Digitonin permeabilizes the plasma membrane but leaves internal membranes intact, permitting the selective detection of cytoplasmic epitopes on transmembrane proteins of internal membranes (72). Fig. 5D shows that digitonin permeabilization allowed the detection of giantin with an antibody against its cytoplasmic domain (73), whereas an antibody against the luminal domain of GPP130 (74) gave no reaction (upper panels). In contrast, permeabilization with digitonin and saponin allowed the detection of both cytoplasmic and extracytoplasmic (luminal) epitopes (Fig. 5D, lower pan-els). HA epitopes attached to either the NH 2 or the COOH terminus of ERGIC-32 were detected with both conditions of permeabilization, indicating that both ends of the molecule are exposed to the cytoplasm. We conclude that ERGIC-32 has two transmembrane domains and that the greater part of the protein is luminally exposed, whereas the NH 2 and COOH termini are cytoplasmic. Consistent with this topology, the (used) consensus sequence for N-glycosylation is located between the two transmembrane domains.
Endogenous ERGIC-32 Localizes to the ERGIC and Overlaps with the Distribution of hErv41 and hErv46 -Next, we examined the localization of endogenous ERGIC-32 in HeLa and HepG2 cells by immunofluorescence microscopy using affinitypurified anti-ERGIC-32 (Fig. 6). In both cell lines ERGIC-32 localizes to dots scattered throughout the cytoplasm and in the Golgi area. The dots also co-stain for ERGIC-53. In BFAtreated cells ERGIC-32 localizes to the typical ERGIC-53-positive punctate elements (Fig. 6, bottom panels). Consistent with the morphological data, ERGIC-32 co-fractionated largely with ERGIC-53 on Nycodenz gradients (Fig 7B). Unlike ERGIC-53, ERGIC-32 was almost excluded from fraction 9, which is enriched in the cis-Golgi marker GPP-130. We conclude that ERGIC-32 is an ERGIC marker.
The homology of ERGIC-32 with Erv41p and Erv46p proteins prompted us to compare the three proteins. We first compared the localization of ERGIC-32 and hErv46 by confocal immunofluorescence microscopy. Mouse Erv46 localizes to the cis-Golgi and the ERGIC (41). Using the rabbit anti-mouse Erv46 antibody, which cross-reacts with the human antigen, we found prominent colocalization of C-terminally HA-tagged ERGIC-32 with hErv46 in the Golgi area. In contrast, most of the peripheral ERGIC clusters were positive for ERGIC-32 only (Fig. 7A, upper panels). This localization was confirmed by Nycodenz gradient centrifugation. ERGIC-32 was more enriched in ERGIC fractions than hErv46, which was enriched in cis-Golgi fractions (Fig. 7B). To visualize hErv41, we isolated the cDNA of human hErv41 (CDA14/hErv41) by RT-PCR using mRNA from HepG2 cells and stably expressed NH 2 -and COOH-terminally tagged hErv41 in HepG2 cells. Both hErv41 variants were found to colocalize with the cis-Golgi marker GPP-130 (data not shown). Double labeling experiments revealed strong colocalization of hErv41-Myc and hErv46 in the Golgi (Fig. 7A, lower panels). Collectively, these results indicate that ERGIC-32, hErv41, and hErv46 partially colocalize in the ERGIC/Golgi area, although ERGIC-32 is more localized to the ERGIC.
ERGIC-32 Interacts with hErv46 -In yeast, Erv41p and Erv46p form a complex and cycle between the ER and Golgi. Complex formation is required for the stability of the two proteins and their co-transport to the Golgi (75). We tested whether ERGIC-32 can interact with hErv46 in a similar way. To this end ERGIC-32HA was immunoprecipitated with anti-HA from lysed membranes of HepG2 cells that had been treated with the membrane-permeable thiol-cleavable crosslinker DSP and probed for the presence of hErv46. Fig. 8A (lane  3) shows that some hErv46 specifically co-precipitated with ERGIC-32HA but not in absence of the cross-linker (not shown). As a further control, hErv46 was not detectable in immunoprecipitates from nontransfected HepG2 cells (Fig. 8A,  lane 2).
To further analyze the interaction between ERGIC-32 and hErv46, we took advantage of the observation that transfected Myc-hErv46 is trapped in the ER and does not concentrate in ERGIC upon BFA treatment (Fig. 8B) but co-precipitates HAtagged ERGIC-32 to the same extent as in nontreated cells (not shown). This mislocalization is most likely a result of limited availability of the partner hErv41 and not to the Myc tag, because nontagged, overexpressed hErv46 also accumulates in the ER. Interestingly, coexpression of ERGIC-32HA changed the ER pattern of Myc-hErv46 to a Golgi-like pattern (Fig. 8C). To confirm the relocalization to the Golgi we performed triple labeling with the KDEL-R in cells cotransfected with Myc-hErv46 and ERGIC-32HA. In HepG2 cells the KDEL-R is localized in the cis-Golgi and to a more minor extent to the ERGIC. Confocal analysis showed that Myc-hErv46 indeed relocalizes to KDEL-R positive structures, confirming that overexpression of ERGIC-32HA pulls Myc-hErv46 out of the ER. Collectively these observations indicate that ERGIC-32 interacts with hErv46 in the ER and that this interaction leads to transport of the two proteins to ERGIC and cis-Golgi.
ERGIC-32 Silencing Affects hErv46 Stability-To study the function of ERGIC-32 we reduced its expression by RNA interference (76). Synthetic duplexes specific for ERGIC-32 were transfected twice into HeLa cells, and the expression of ER-GIC-32 was measured by Western blotting after 1 week of interference. As shown in Fig. 9A, ERGIC-32 was strongly reduced (by 80%) in cells transfected with the more efficient siRNA (si1) compared with cells transfected with control siRNA. Other organelle markers were not affected by the treatment, but the expression of hErv46 was slightly reduced (Fig.  9A). Since in yeast Erv41p and Erv46p stabilize each other by forming a complex (66), we tested whether ERGIC-32 silencing FIG. 4. ERGIC-32 is a putative transmembrane protein homologous to yeast Erv41p and Erv46p. A, protein sequence alignment of ERGIC-32 orthologs. Human (hERGIC-32, gi͉15215343͉), mouse (mERGIC-32, gi͉12835932͉), and C. elegans (ceERGIC-32, gi͉13775507͉) ERGIC-32-related sequences were aligned using ClustalW. Identical residues are shaded, putative transmembrane domains are boxed as deduced by the Toppred2 prediction program (84), and a consensus site for N-glycosylation is underlined. B, ERGIC-32 protein sequence identity with Erv41p and Erv46p proteins. Alignment of human and C. elegans ERGIC-32 with Erv46p (gi͉6319274͉) and Erv41p (gi͉6323573͉) of yeast and their human homologs hErv46 (CGI-54/putative protein of 43.2 kDa, gi͉4929577͉), hErv41 (CDA14, gi͉7706105͉) was obtained by ClustalW analysis. The numbers indicate percentages of identity (BLOSUM matrix analysis). C, ERGIC-32 hydrophobocity profile (Kyte and Doolittle). Two amino acids stretches of sufficient length and hydrophobicity are predicted to be transmembrane domains near the N and C termini according to the Toppred2 program. D, dendogram of the ERGIC-32/Erv41/Erv46 protein family. Multiple alignment (clustalW program) and tree drawing (PHYLIP program) were performed using the Biology Workbench interface (workbench.sdsc.edu). Mouse Erv46 (mErv46, gi͉12844094͉), Erv41 (mErv41, gi͉12841082͉), and C. elegans Erv46 (ceErv46, gi͉3878494͉), Erv41 (ceErv41, gi͉25809204͉) sequences were also included. modifies the half-life of hErv46. Fig. 9B shows that this is indeed the case. The half-life of hErv46 was reduced from 3 to 1 h in cells treated with ERGIC-32 siRNA (si1) but was not affected by control siRNA (Fig. 9C) 3 and 5). An unspecific background band in anti-HA immunoprecipitates is indicated by an asterisk. C, membrane association of ERGIC-32. HepG2 total membranes were treated with 0.1 M Na 2 CO 3 pH 11.5, on ice and then pelleted. Like ERGIC-53, ERGIC-32 remains associated with the pellet fraction as opposed to the soluble protein RAP that appears in the supernatant (SN). D, ERGIC-32 membrane topology. HepG2 cells expressing ERGIC-32 that was HA-tagged either at the N or the C terminus were fixed and permeabilized with digitonin alone (upper panels) or digitonin and saponin (lower panels). They were then co-labeled with anti-HA (right panels) and anti-giantin or anti-GPP130 (left panels). Like the cytoplasmic epitope of giantin, the HA epitopes at both ends of ERGIC-32 were accessible under both permeabilization conditions, whereas the luminal epitope of GPP130 was not accessible when cells are permeabilized with digitonin alone. Bar, 10 m.

FIG. 6. ERGIC-32 colocalizes with ERGIC-53 in HeLa and HepG2 cells.
Cells treated with or without BFA were fixed, processed for double-labeling immunofluorescence microscopy, and analyzed by confocal microscopy. ERGIC-32 (affinity-purified antibody, green in overlay) localizes to punctate structures in the cytoplasm and co-localize with ERGIC-53 (mAb, red in overlay). After treatment of HepG2 cells with BFA, ERGIC-32 and ERGIC-53 accumulate in ERGIC clusters. Bar, 20 m. transfected with control RNA. Silencing hErv41 by siRNA also reduced the protein levels of hErv46 (Fig. 9A), and by silencing hErv46, hErv41 was not detectable after transfection, presumably because it is rapidly degraded (not shown). In contrast, reduced levels of either hErv46 or hErv41 had no effect on ERGIC-32. The results indicate that hErv41 and hErv46 stabilize each other as in yeast but that the ⌭rv41-Erv46 complex is further stabilized by ERGIC-32.

DISCUSSION
In this study we present the first comprehensive characterization of ERGIC membranes from BFA-treated cells. The new procedure we have developed for the isolation of ERGIC membranes is superior to previous methods (4,77) and results in a 2-3-fold higher purity. The enrichment factor is comparable with that of highly enriched Golgi membrane fractions used for proteomic analysis (78). Unlike the present method, the previously published procedures for the purification of the ERGIC could not completely remove contaminating membranes from other organelles. The new procedure was specifically designed to identify new cycling proteins and takes advantage of two different effects of BFA. First, the in situ merging of Golgi with ER by BFA allows efficient separation of the ERGIC from Golgi membranes by density gradient centrifugation. Second, the BFA-induced accumulation of KDEL-R in ERGIC membranes facilitates their subsequent affinity isolation by magnetic immunobeads. No membrane marker is known to exclusively localize to the ERGIC, rendering it difficult to purify these membranes from untreated cells. In BFA-treated cells, however, KDEL-R and ERGIC-53 almost quantitatively redistribute to the ERGIC and can therefore serve as ERGIC-specific markers under these conditions. The low level of contaminating proteins and the identification of a number of known and new cycling proteins in the final fraction clearly document the value of our approach. It should be noted that the new BFA-based procedure does not provide quantitative information on the number of individual proteins present in the ERGIC of untreated cells. For instance, KDEL-R is mostly localized to the cis-Golgi in untreated HepG2 cells, whereas ERGIC-53 is enriched in the ERGIC. Upon BFA treatment, however, the two markers almost completely co-localized in ERGIC spots. Thus, although most Golgi proteins redistribute to the ER in response to BFA, the fraction we have isolated contains some Golgi proteins and is therefore occasionally designated ERGIC/Golgi remnant membranes. Moreover, we cannot exclude the possi- bility that the purified membranes also contain some ER exit site membranes, although ER exit sites do not merge with the ERGIC in BFA-treated cells (79). Currently, there is no accepted membrane marker for ER exit sites with which to test this issue.
By mass spectrometry we have identified 24 proteins enriched in the ERGIC membranes, many of which are already well characterized and their function known or inferred from proteins of the same family. The identified proteins represent rather abundant proteins, but the methodology can also be applied directly for a more refined analysis of minor proteins in view of the purity of the membrane fraction. The inclusion of a sodium carbonate, pH 11.5, treatment favors the identification  Equal amounts of total proteins were analyzed by SDS-PAGE and Western blotting using antibodies against BAP31, TfR, hErv46, and ERGIC-32. ERGIC-32 depletion reduces hErv46 expression. hErv46 is strongly reduced by silencing hErv41. Similar results were obtained with two other oligonucleotides targeting hErv41 mRNA. B, HeLa cells transfected with control and ERGIC-32 siRNA (si1) were pulsed and chased for the indicated times. hErv46 was immunoprecipitated and analyzed by SDS-PAGE (fluorogram). Relative amounts of hErv46 were quantified and expressed as the percentage of time 0 chase. C, HeLa cells treated with siRNA oligonucleotides were pulsed and chased for 3 h as in B. The amount of hErv46 was quantified and expressed as the percentage remaining after 3 h of chase compared with the amount present at the beginning of the chase (mean Ϯ S.D., n ϭ 3; Student's t test for control versus si1, p Ͻ 0.01). of integral membrane proteins at the expense of peripheral membrane proteins and soluble proteins, although some soluble chaperones were still present, the reason for which is not entirely clear. We do not know to what extent the substantial amount of these chaperones is due to a BFA effect or whether the ERGIC of nontreated cells also contains high levels of these chaperones in HepG2 cells. However, it is noticeable that the chaperone CBP1/HSP47 is known to be associated with procollagen in the ERGIC from where it is retrieved to the ER via its RDEL motif (64,80).
Interestingly, some proteins we have identified in the ERGIC fraction have orthologs or homologs in yeast where they are enriched in COPII vesicles (66). These human proteins are p24 family members, Surf-4, Yif1p, and ERGIC-32. Data base searches reveal the existence of at least eight human p24 proteins, four of which we have identified, namely p24A, gp25L2, TMP21, and CGI-100. The reason the other p24 family members have escaped our analysis may be the differential cycling properties of the individual p24 proteins (36), low expression levels, or difficulties in detecting them by mass spectrometry. Surf-4 has been described as an ER protein (69). The present study indicates that Surf-4 can cycle in the early secretory pathway and may therefore function as a cargo receptor similar to its yeast ortholog Erv29p. Yif1p binds the transport GTPase Yipt1p and is required for fusion competence of ERderived vesicles with the Golgi membranes in yeast (81,82). Its presence in ERGIC membranes together with Rab1 (the mammalian homolog of Yipt1p) may suggest that Yif1p controls vesicle fusion at the ERGIC and/or Golgi. Further studies are required to uncover the functions of Surf-4 and Yif1p.
In discussing potential functions of ERGIC-32, it may be informative to consider its relatives Erv41p and Erv46p previously studied in yeast. These proteins are selectively and efficiently packaged into COPII vesicles and cycle between the ER and Golgi (66). Their expression levels are interdependent, and deletion of the Erv41p-Erv46p complex influences the membrane fusion stage in a cell-free assay that reproduces transport between the ER and Golgi. However, the precise molecular mechanism by which these proteins function in membrane trafficking between the ER and Golgi is so far unknown. Although we did not find hErv41 and hErv46 by our proteomic approach, these proteins are also enriched in the ERGIC of BFA-treated cells (41) 2 and are bona fide cycling proteins in higher eukaryotes. Like ERGIC-32, hErv41 and hErv46 lack a cleavable signal sequence and are predicted to have large luminal and short NH 2 -and COOH-terminal cytosolic domains. Transport signals present in the COOH-terminal part of Erv41p and Erv46p proteins and acting in trans are responsible for the COPII-dependent exit of the complex from the ER and its cycling between the ER and Golgi in yeast (75). Two hydrophobic residues, Ile-349 and Leu-350 in positions Ϫ4 and Ϫ3 of the COOH terminus of Erv41p, constitute a transport motif required for the recruitment of the Erv41p-Erv46p complex into COPII vesicles. However, the Ile-Leu motif is not sufficient for ER export. The Erv46p tail contains a Phe-Tyr sequence in positions Ϫ13 and Ϫ14 from the COOH terminus cooperating with the Ile-Leu motif in Erv41p for ER export of the complex. The combined presence of both motifs in a specific orientation is a prerequisite for efficient packaging. These motifs appear to be conserved in mammalian Erv41 and Erv46. A di-leucine motif is present in positions Ϫ6 and Ϫ7 of mammalian hErv41 and an Ile-Tyr sequence is located at the same distance from the membrane in hErv46 as the Phe-Tyr sequence in yeast Erv46p. A di-lysine motif in positions Ϫ3 and Ϫ4 of the Erv46p may mediate the Golgi to ER retrograde trafficking of the Erv41p-Erv46p complex (66,75). Only the lysine in position Ϫ3 is conserved in mammalian Erv46. However, in some instances a lysine at position Ϫ3 is sufficient for retaining type 1 membrane proteins in the ER (83). Extrapolating from the similarities of transport motifs, the mechanisms of trafficking of Erv41 and Erv46 are most likely similar in yeast and mammals. Moreover, the interdependence and identical localization of hErv46 and hErv41 suggest that these two proteins presumably form a stable complex as described in yeast.
ERGIC-32 has many similarities to hErv46. Apart from its overall sequence similarity, its C-terminal tail is closely related and the transport motifs are conserved. This is the case for the two hydrophobic residues, Ile274 and Phe275, that may function in COPII packaging and the Lys288 in position Ϫ3 that might function in retrieval. However, there are notable differences between ERGIC-32 and hErv46. First, ERGIC-32 has a shorter luminal domain than hErv46. Second, ERGIC-32 localizes to the ERGIC, whereas hErv46 localizes more to the cis-Golgi as does hErv41. Third, overexpressed ERGIC-32 localizes correctly to the ERGIC as opposed to overexpressed hErv46 that is trapped in the ER unless ERGIC-32 is co-transfected. Fourth, silencing ERGIC-32 increases the turnover of hErv46 but not vice versa. Fifth, silencing hErv41 increases the turnover of hErv46 but not ERGIC-32. Sixth, ERGIC-32 is not present in yeast.
Obviously, ERGIC-32 does not require teaming up with either hErv46 or hErv41 for correct localization, nor does its stability depend on these two proteins. Together with the finding of different steady state localization it is unlikely that ERGIC-32 is permanently associated with the Erv41-Erv46 complex beyond the ER. Notably, correct localization of the Erv41-Erv46 complex does not require the presence of ERGIC-32. What then is the function of ERGIC-32? A likely scenario is that ERGIC-32 stabilizes monomeric hErv46 in the ER and thereby promotes optimal assembly of hErv46 with hErv41. In addition, or alternatively, ERGIC-32 may act as a transport chaperone facilitating ER to ERGIC transport of the Erv41-Erv46 complex. Such a function may not be required in yeast, where ERGIC-32 is absent, but in higher eukaryotes which have developed a more complex early secretory pathway.
The function of Erv41, Erv46 and ERGIC-32 in mammalian cells is so far unknown. Combined depletion of ERGIC-32 and hErv46 had no effect on protein or glycoprotein secretion in a pulse-chase approach (not shown). Because hErv46 and hErv41 are interdependent, this double knock down is in fact a triple knock down. Unless the function of these three proteins can be compensated by other proteins, the silencing data suggest that they are probably not essential for vesicular trafficking. Further studies are required to elucidate the function of these three proteins in mammalian cells.
The new procedure for purification of ERGIC membranes from BFA treated cells described in the present study should prove useful to ultimately identify all the proteins rapidly cycling early in the secretory pathway. Although the protein pattern of ERGIC membranes appears rather simple regarding the major proteins (Fig. 2), we anticipate many more proteins to cycle through the ERGIC.