Molecular Cloning and Expression of a 58-kDa cis- Golgi and Intermediate Compartment Protein*

An abundant 58-kDa (p58) homodimeric and hexa- meric microsomal membrane protein has been biochem-ically characterized and localized to tubulo-vesicular elements at the endoplasmic reticulum-Golgi interface and the cis- Golgi cisternae in pancreatic acinar cells J. Cell Sci. Here we report the purification of p58 by two-dimensional gel electrophoresis, and the cloning and sequencing of the rat and part of the Xenopus laevis cDNAs. The rat cDNA encodes a 517-amino acid protein having a putative signal sequence, a transmembrane domain close to the C terminus and a short cytoplasmic tail. The C-terminal tail contains a double-lysine motif (KKFF), known to mediate retrieval of proteins from the Golgi back to the endoplasmic reticulum. The rat p58 sequence was found to be 89% identical with those of ERGIC-53 and MR60, two previously identified human membrane proteins. Strong homology with the frog sequence was also observed indicating high evolutionary conservation. Overexpression of c-Myc-tagged p58 resulted in accumulation of the protein both in the endo- plasmic reticulum and in an apparently enlarged Golgi complex, as well as its leakage to the plasma membrane. Immunolocalization using antibodies raised against a lumenal

The endomembranes participating in the early stages of protein maturation and transport consist of the endoplasmic reticulum (ER), 1 the cis-Golgi cisternae and tubulo-vesicular elements at the interface between these organelles (Palade, 1975). After their translocation across the membrane of the rough ER, newly synthesized secretory and membrane proteins undergo folding and post-translational modification reactions, before becoming competent to be transported out of the ER (Helenius et al., 1992;Gething and Sambrook, 1992;Helenius, 1994). Transport along this pathway occurs via an intermediate compartment which could originally be visualized in Semliki Forest virus-infected cells using low temperature as a reversible transport arrest (Saraste and Kuismanen, 1984). Identification and localization of two marker proteins, rat p58 (Saraste et al., 1987) and human p53 (Schweizer et al. (1988); renamed ERGIC-53, Schindler et al. (1993)), has provided additional information of the structural complexity of this compartment. The tubulo-vesicular elements, to which these proteins concentrate (Schweitzer et al., 1991;Lahtinen et al., 1992), have been shown to be involved in the transport of viral glycoproteins to the Golgi complex (Schweitzer et al., 1990;Saraste and Svensson, 1991;Plutner et al., 1992).
The removal of membrane from the ER, that accompanies the formation of vesicular carriers, is balanced by the counterflow of lipids and retrieval of membrane proteins from the Golgi (Wieland et al., 1987). This retrograde pathway serves as a mechanism to return escaped, KDEL-containing, resident proteins back to the ER (Pelham, 1988;Dean and Pelham, 1990) and maintain the localization of either KKXX-or KXKXXcontaining ER membrane proteins by retrieval (Nilsson et al., 1989;Jackson et al., 1990;Jackson et al., 1993;Townsley and Pelham, 1994;Gaynor et al., 1994).
Studies using in vitro assays and yeast secretion mutants have identified several cytosolic components needed for vesicle budding, transport, and fusion (reviewed by Rothman, 1994). Coatomer is a complex of seven proteins (␣-, ␤-, ␤Ј-, ␥-, ␦-, ⑀-, and -COP), which assemble on membranes in an ADP-ribosylation factor-dependent manner forming a cytoplasmic coat (Serafini et al., 1991;Waters et al., 1992;Orci et al., 1993;Ostermann et al., 1993). Coatomers are involved both in intra-Golgi (Rothman and Orci, 1992) as well as ER to Golgi transport (Pepperkok et al., 1993;Peter et al., 1993). Recently, coatomer subunits and double-lysine motifs have been shown to interact in vitro . The functional significance of this interaction in membrane dynamics between the ER and Golgi was revealed by genetic studies in yeast  showing that several coatomer subunits were essential for protein retrieval.
As a first stage in trying to understand both the function and the molecular mechanism for trafficking of p58 we have cloned and expressed its cDNA and developed peptide antibodies against different domains of the protein. We report here that p58 is the rat homolog of the human ERGIC-53/MR60 protein (Schindler et al., 1993;Arar et al., 1995), which has been suggested to function as a membrane bound lectin in the early secretory pathway Arar et al., 1995). Furthermore, immunolocalization using the anti-tail peptide antibodies detected p58 only in the Golgi complex, suggesting masking of the tail in earlier compartments. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) U44129 (rat) and U44130 (X. laevis).

EXPERIMENTAL PROCEDURES
Materials-Methionine-and cysteine-free minimal essential medium were obtained from Sigma, all other cell culture media and reagents including Lipofectin® were from Life Technologies, Inc. Pro-Mix ([ 35 S]methionine and -cysteine) was from Amersham. Sequenase version 2.0 Sequencing kit from U. S. Biochemical Corp., rat pancreas ZAP II and Xenopus laevis liver UNI-ZAP XR cDNA libraries from Stratagene and AmpliFINDER RACE kit from Clontech. Trypsin (sequence grade) was from Promega and endoproteinase Lys-C (Achromobacter lyticys protease I) from Wako Chemicals GmbH. Rabbit reticulocyte lysate, canine pancreatic microsomal membranes, and Poly(A)Tract mRNA Isolation System were purchased from Promega. FITC-coupled goat anti-rabbit and anti-mouse IgG were from Biosys and TRITCcoupled anti-rabbit IgG from Sigma. SDPD, protein A-Sepharose CL-4B and NAP-25 columns were from Pharmacia. Quick Spin Sephadex G-50 columns were from Boehringer Mannheim and Gene Screen Plus membranes from DuPont NEN. 4-Vinylpyridine was from Fluka and all chemicals for two-dimensional gels from sources described previously (Lahtinen et al., 1992).
Antibodies-Antiserum raised against the catalytic domain of ␣-mannosidase II (Moremen et al., 1991) and affinity-purified monoclonal anti-c-Myc antibodies (9E10; Evan et al., 1985) were kindly provided by Dr. Marilyn Farquhar and Dr. Tommy Nilsson, respectively; monoclonal anti-HA antibodies (12CA5) were from Berkeley Antibody Co. The preparation and affinity-purification of polyclonal anti-p58 antibodies raised against SDS-gel-purified p58 has been described elsewhere (Saraste and Svensson, 1991).
Cell Culture, Transfection, and Metabolic Labeling-BHK-21 cells were grown in minimal essential medium containing 5% of both fetal calf serum and tryptose phosphate broth and supplemented with 2 mM L-glutamine, 100 IU/ml penicillin, and 100 g/ml streptomycin. Cells were transfected by lipofection using the Semliki forest virus-based expression system (Liljeström and Garoff, 1991). Subconfluent plates were first preincubated in Opti-MEM for 10 min and 9 g of Lipofectin was mixed with 2.3 g of RNA in 1 ml of Opti-MEM by vortexing. The cells were incubated with Lipofectin-RNA mixture for 2 h at 37°C and thereafter the transfection mixture was replaced with BHK growth medium.
For metabolic labeling, the cells were first preincubated with methionine-and cysteine-free medium containing 10% dialyzed fetal calf serum for 1 h and then labeled with Pro-Mix (500 Ci/ml).
Preparation and Affinity Purification of Peptide Antibodies-Two peptides corresponding to amino acids 158 -170 (FFDSFDNDGKKNN, lumenal) and 507-517 (TQQEAAAKKFF, tail) were synthesized with an additional cysteine at the N terminus. For immunizations, peptides were coupled via the cysteine residue to keyhole limpet hemocyanin using SPDP as a heterobifunctional cross-linker. 10 mg of keyhole limpet hemocyanin were incubated in the presence of a 20 times molar excess of SPDP in 100 mM sodium phosphate buffer (pH 7.5) containing 100 mM NaCl for 1 h at room temperature. The reaction mixture was passed through a NAP-25 column to separate uncoupled SPDP. Five times molar excess of peptide was added and the reaction was continued for 1 h at room temperature or overnight at 4°C depending on the solubility of the peptide.
Due to the relatively high background observed in immunofluorescence staining, ImmunoPure Ag/AB Immobilization Kit #2 (Pierce) was used for affinity purification of the anti-tail antibodies according to the manufacturer's instructions.
Preparation of Microsomes-Preparation of rat pancreas microsomes was carried out as described in detail elsewhere (Saraste et al., 1986). Microsomes were extracted with sodium carbonate (pH 11.3) according to Howell and Palade (1982).
SDS-PAGE and Two-dimensional Gel Electrophoresis-10% SDSpolyacrylamide gels were run to separate proteins in one dimension according to Laemmli (1970). To purify p58, alkaline-extracted microsomal proteins were separated first by isoelectric focusing (O'Farrell, 1975), followed by SDS-PAGE in the second dimension as described before (Lahtinen et al., 1992), except that proteins were alkylated in 40 mM 4-vinylpyridine before the second dimension. Proteins were transferred onto nitrocellulose filters, stained with Ponceau S, and the spots corresponding to p58 (Lahtinen et al., 1992) were excised.
Amino Acid Sequence Analysis-Membrane pieces from blots of about 50 two-dimensional gels were combined, and a total about 3 g of protein were digested with either trypsin or endoproteinase Lys-C following a protocol of Aebersold et al. (1987). Fragments were isolated by reverse phase liquid chromatography on a RPC C2/C18 SC 2.1/10 column (Pharmacia Biotech). The eluting acetonitrile gradient (0.5%/ min in 0.1% trifluoroacetic acid) was generated in a Beckman Gold System. Flow rate was 100 l/min, and fractions were collected manually. Sample peptides were sequence analyzed in an Applied Biosystems model 470A instrument following the manufacturer's instruction (Foster City).

Molecular Cloning of Rat and Frog p58 cDNAs-Two overlapping degenerate oligonucleotide mixtures 5Ј-TA(C/T)CA(A/G)GA(A/G)GA(A/ G)TT(C/T)GA(A/G)CA-3Ј and 5Ј-GA(A/G)CA(C/T)TT(C/T)CA(A/G)CA-(A/G)AT(C/T)C-3Ј
derived from a peptide sequence were use to screen a rat pancreas ZAP II cDNA library. Duplicate nylon filters were hybridized overnight at 37°C in the presence of 32 P-labeled probe in 6.6 ϫ SSC, 5 ϫ Denhardt's solution, 0.5% SDS, and 100 g/ml salmon sperm DNA. To isolate the 3Ј region of the p58 cDNA, the library was further screened with a 32 P-labeled 600-base pair restriction fragment derived from the obtained 3Ј region at 65°C in the same hybridization solution. The 5Ј region was isolated by using 5Ј AmpliFINDER RACE PCR method according to the manufacturers protocol (Clontech). Poly(A) ϩ RNA was isolated from rat liver total RNA using the Poly(A)Tract mRNA Isolation System (Promega). 5Ј-GTCACTTCCACTTCCCAGT-TCTCGAAGGC-3Ј was used as a gene-specific primer for cDNA synthesis and a nested primer 5Ј-CGAGCGGATCCCGGAACCTCTTTGGC-TCTTTAAGGACGG-3Ј, containing a BamHI site for PCR amplification and cloning. A full-length cDNA was constructed into pGEM-3Zf plasmid from overlapping clones and the PCR-amplified fragment.
To introduce either a c-Myc-or HA-epitope at the N terminus (alanine 42) an unique NarI-restriction site was used. Sense and antisense oligonucleotides encoding amino acids GEQKLISEEDLG (c-Myc, bold) or GYPYDVPDYAG (HA, bold) and containing NarI sites were synthesized, annealed, and ligated.
For transfections with both the wild type and c-Myc-or HA-tagged p58, the cDNAs were transferred into the pSFV1 vector (Liljeström and Garoff, 1991).
Nucleotide sequences were determined from both strands by using the Sequenase version 2.0 Sequencing Kit (U. S. Biochemical Corp., Amersham).
In Vitro Transcription and Translation-For each transcription reaction, 2 g of linearized plasmid and T7 or SP6 RNA polymerase was used. Transcriptions were performed according to the manufacturer's protocol (Promega), but to increase the yield of RNA the incubation time was increased. First, 2 l of 10 mM GTP was added and the incubations were continued for 15 min, thereafter 20 units of RNA polymerase was added, and incubations were continued for 45 min. RNA was phenolchloroform extracted and precipitated with ethanol. For transfection experiments the RNA was further purified on Quick Spin columns.
In vitro translations using rabbit reticulocyte lysate and canine pancreatic microsomal membranes were performed as described by the manufacturer (Promega). Proteinase K digestions were carried out as described earlier (Lahtinen et al., 1992).
Immunoprecipitation-Immunoprecipitations from lysates from metabolically labeled BHK cells were performed as described (Lahtinen et al., 1992). For each precipitation, 5 l of preimmune or immune serum and 100 l of 10% (v/v) preswollen protein A-Sepharose CL-4B was used.
Immunofluorescence Microscopy-BHK cells were fixed either with methanol for 5 min at Ϫ20°C (anti-tail p58 antibodies) or with 3% paraformaldehyde in 100 mM sodium phosphate buffer (pH 7.4) for 30 min (all other antibodies), quenched with 50 mM ammonium chloride in phosphate-buffered saline for 30 min, and permeabilized with 0.05% Triton X-100 in phosphate-buffered saline. The cells were incubated in phosphate-buffered saline containing 2% goat serum and 0.2% bovine serum albumin for 30 min to block unspecific binding of antibodies. Primary antibodies were visualized either with FITC-or TRITC-coupled goat anti-rabbit or anti-mouse antibodies. The coverslips were mounted in 80% glycerol in phosphate-buffered saline supplemented with 1 mg/ml phenylenediamine and were viewed and photographed with a Zeiss Axiophot photomicroscope.

RESULTS
Purification of p58 -Two-dimensional gel analysis of rat pancreatic microsomal proteins demonstrated that p58 was a major membrane protein in this subcellular fraction (see Fig. 3, B and D, in Lahtinen et al. (1992)). This observation allowed us to purify the protein to homogeneity. Total microsomes were isolated from rat pancreatic homogenates and membrane proteins were extracted with alkaline sodium carbonate, resulting in a considerable enrichment of p58. The proteins were separated by isoelectric focusing followed by SDS-PAGE, transferred onto nitrocellulose filters, and after staining with Ponceau S, the spots corresponding to p58 were excised. For protease digestions in situ with either trypsin or endoproteinase Lys-C, pooled p58 from 40 blots was used. Peptides were separated by reverse-phase high performance liquid chromatography and four major peptides were microsequenced.
Molecular Cloning of p58 -Two overlapping degenerate oligonucleotide probes were designed on the basis of the peptide sequence YQEEFEHFQQIL (observe that in the sequence deduced from the cDNA a glutamate is found instead of an isoleucine, Fig. 1) and about 2 ϫ 10 6 clones of a rat pancreas ZAP II cDNA library were screened. Several positive clones were isolated, one of which contained sequences encoding all of the four microsequenced peptides, but was lacking both the start and the stop codons. Further screening with a 600-base pair long C-terminal cDNA probe derived from the original clone lead to the isolation of two C-terminal clones which overlapped with the original clone and contained a potential stop codon. No poly(A) sequence was found in these clones. To isolate and confirm the 5Ј region, mRNA was purified from rat liver, cDNA was synthesized, and the RACE PCR method (see "Experimental Procedures") was used for amplification. With the primers used a 390-base pair fragment was obtained and sequenced. A full-length cDNA was constructed from overlapping clones and the PCR fragment, and cloned into the pGEM-3Zf plasmid.
The p58 cDNA has an open reading frame encoding 517 amino acid residues (Fig. 1). The N terminus is predicted to contain a 30-residue long and cleavable signal sequence (von Heijne, 1985), followed by a lumenal domain. Based on the hydropathy profile (Kyte and Doolittle, 1982; data not shown) and the position of the flanking charged residues, a 18-residue long hydrophobic stretch close to the C terminus is likely to form a transmembrane domain, followed by a 12-amino acid long cytoplasmic tail. Mature p58 is predicted to have molecular mass of 54,755 daltons and pI of 5.9. In agreement with previous biochemical data (Lahtinen et al., 1992), no consensus sites for N-glycosylation were found.
Conservation of p58 -Sequence comparison (Fig. 2) in the Swissprot and EMBL data bases shows 89% identity (94% similarity) with the human intermediate compartment ER-GIC-53 protein (Schindler et al., 1993) and the human mannose-specific lectin MR60 (Arar et al., 1995). The sequences of ERGIC-53 and MR60 are identical, differing only in 1 amino acid residue. The similarity in the rat and human proteins is evenly distributed along the sequences, but in the rat sequence an 8-amino acid insertion is found close to the N terminus of the mature protein. In the cytoplasmic tail the two lysines at positions Ϫ3 and Ϫ4 are conserved with only a serine at position Ϫ11 being replaced by a threonine.
To obtain further information of the degree on evolutionary conservation of p58, we screened a X. laevis UNI-ZAP XR cDNA library with a 1.3-kilobase rat p58 cDNA fragment. Two positive overlapping clones were isolated having a long open reading frame and a stop codon, but both lacking the start codon. The Xenopus protein was estimated to have the same size as the rat p58 since anti-p58 antibodies recognized a protein comigrating with the rat p58 in immunoblots of frog liver homogenates (data not shown). The C-terminal 420-amino acid sequence was 71% identical (85% similar) to the rat sequence (Fig. 2).
Membrane Insertion and Topology of in Vitro Translated p58 -Membrane insertion and signal sequence cleavage were analyzed by translating the in vitro transcribed p58 mRNA in the presence or absence of canine microsomal membranes followed by immunoprecipitations (Fig. 3). In the presence of microsomes p58 was inserted into the membrane and co-migrated with the endogenous protein, immunoprecipitated from metabolically labeled BHK cells (Fig. 3A, lanes 1 and 3). In the absence of microsomes the mobility of the protein was slightly slower (lane 2), indicating the presence of a functional and cleavable signal sequence. The protein could not be post-translationally inserted into the membranes (data not shown). Both the c-Myc-and HA-tagged forms of p58 behaved as the wild type (lanes 4 -7).
When membranes were subjected to extraction with sodium carbonate (pH 11.3) followed by centrifugation, the in vitro translated p58 was almost quantitatively recovered in the resulting pellet (Fig. 3B), confirming proper membrane integration. Previous experiments have shown that p58 remained largely protected in microsomes after protease treatments (Lahtinen et al., 1992). To probe the membrane topology of in vitro synthesized p58, anti-tail peptide antibodies (see below) were used.
Both anti-tail and polyclonal anti-p58 antibodies precipitated the intact protein (Fig. 3C, lanes 1 and 2). After digestion with proteinase K, anti-tail antibodies failed to precipitate the protein, indicating that the short tail was accessible to the enzyme FIG. 2. Comparison of the rat p58, human ERGIC-53, and frog p58 amino acid sequences. Rat and human sequences share 89% identity (94% similarity), and rat and frog sequences 71% identity (85% similarity) indicating high evolutionary conservation of the protein. Together these results show that p58 is a type I integral membrane protein, having the bulk of the protein on the lumenal side and a short Cterminal cytoplasmic tail in agreement with the topology deduced from the amino acid sequence (see above).
Peptide Antibodies Reveal Different Populations of p58 -Antisera were raised against two different peptides, one from the lumenal N-terminal region (corresponding to amino acids 158 -170; anti-lumenal) and the other corresponding to the C-terminal tail (amino acids 507-517; anti-tail). Both antibodies gave a strong signal in immunoblots of rat pancreatic microsomal proteins (Fig. 4, lanes 1 and 3), which could be competed out with the corresponding peptide (not shown). They also specifically precipitated a 58-kDa protein from metabolically labeled BHK cells, although precipitation with the antitail antibodies from cell lysates was considerably less efficient (lanes 2 and 4).
When tested by immunofluorescence a distinct difference in the staining patterns obtained with the two peptide antibodies was observed in BHK cells. The staining with the anti-lumenal antibodies was similar to the staining obtained with the previously described affinity-purified anti-p58 antibodies (Saraste and Svensson, 1991); the Golgi region as well as vesicular elements concentrated in the Golgi area, but also scattered throughout the cell, were seen (Fig. 5, panels A and B). In contrast, the anti-tail antibodies stained a perinuclear, often ring-shaped structure typical of the Golgi complex (panel C).
No scattered punctate elements could be visualized even in transfected cells having higher expression levels of p58 (data not shown). This suggests that in the vesicular elements the epitope(s) in the C-terminal tail may be hidden and therefore not accessible to the anti-tail antibodies.
Expression of c-Myc-tagged p58 -Due to the high, endogenous background of p58 staining in murine and other cell lines tested, a c-Myc-epitope tag (Evan et al., 1985) was introduced at the N terminus downstream from the predicted signal se-quence cleavage site. To ensure that the insertion of a foreign peptide sequence did not interfere with proper signal sequence cleavage or cellular localization of p58, we also tested an influenza virus HA-tagged (Field et al., 1988) version of p58. In vitro translation experiments with these constructs showed that signal sequence cleavage occurred normally in both cases (Fig.  3A, lanes 4 -7).
For in vivo expression, both the c-Myc-and HA-tagged p58 cDNAs were cloned into a Semliki Forest virus-based pSFV1vector (Liljeström and Garoff, 1991), the DNA was transcribed in vitro, and transfected into BHK cells by lipofection. Four hours post-transfection, cells were treated for 1 h with 50 g/ml cycloheximide and then fixed and processed for immunofluorescence. Staining with anti-c-Myc (Fig. 6, panel A) antibodies showed a punctate, Golgi-concentrated pattern resembling the distribution of endogenous p58. Double-staining with antimannosidase II antibodies (panel B) showed overlapping distributions of the two antigens in the Golgi area as previously observed both in BHK and normal rat kidney cells (Saraste and Svensson, 1991). In cells having a higher expression level, the protein could also be clearly detected in reticular ER membranes. This is in good agreement with previous immunolocalization studies and the quantitation of p58 in subcellular fractions (Saraste and Svensson, 1991;Lahtinen et al., 1992). The same distribution was observed with the HA-tagged p58 (data not shown). In addition, c-Myc-tagged p58 was shown to form FIG. 4. Specificity of the peptide antibodies. Anti-lumenal p58 antibodies and anti-tail p58 antibodies (lanes 1 and 3, respectively) specifically recognize a 58-kDa protein in immunoblots of rat pancreatic microsomes. Immunoprecipitates from cell lysates from 35 S-labeled BHK cells with anti-lumenal (lane 2) and anti-tail (lane 4) antibodies. The high molecular weight bands probably represent aggregated p58. both homodimers and oligomers as well as heterodimers and oligomers with the endogenous protein indicating that the oligomerization (Lahtinen et al., 1992) was not affected by the introduction of a foreign peptide sequence at the N terminus (not shown).
Overexpression of the c-Myc-tagged p58 resulted in the accumulation of the protein in the intracellular membranes. No increase in punctate elements could be detected while the ER network was heavily stained, and the Golgi complex, identified by double-staining with anti-mannosidase II antibodies (not shown), appeared enlarged (Fig. 6, panel C). When non-permeabilized cells were stained with anti-c-Myc antibodies, plasma membrane patches could be visualized (panel D), indicating that overexpression of the c-Myc-tagged p58 led to its mislocalization and leakage from the intracellular membranes to the plasma membrane.

DISCUSSION
During the last several years p58 (Saraste et al., 1987;Saraste and Svensson, 1991;Lahtinen et al., 1992) and p53 (Schweitzer et al., 1988(Schweitzer et al., , 1990renamed ERGIC-53, Schindler et al. (1993)), two membrane proteins identified in rat and human cells, respectively, have been widely used as markers for the intermediate compartment at the ER-Golgi interface. These proteins have very similar intracellular distributions and biochemical properties, but their structural and functional relationship has been unknown. The abundance of p58 in pancreatic microsomal membranes (Lahtinen et al., 1992) allowed us to purify the protein to homogeneity, obtain peptide sequences, and clone the cDNA. The sequence reveals that rat p58 and human ERGIC-53 (Schindler et al., 1993) are homologous proteins sharing 89% identity at the amino acid level. Furthermore, the observed 71% identity between the isolated partial X. laevis p58 sequence and the rat sequence indicates a high degree of evolutionary conservation of the protein among different species.
The cDNA sequence, together with biochemical data, suggest that p58 is synthesized as a 517-amino acid protein having a predicted and cleavable signal sequence and a single transmembrane domain close to the C terminus. This type I membrane topology was confirmed by using peptide antibodies generated against lumenal and cytoplasmic domains of p58.
Although p58 has been suggested to contain immature N-glycans (Hendricks et al., 1991), no consensus sites for N-glycosylation were found.
The expression of c-Myc-tagged p58 in BHK cells resulted in a distribution identical to the endogenous protein (Saraste and Svensson, 1991). A typical punctate, Golgi-concentrated pattern could be visualized by immunofluorescence microscopy. In cells overexpressing p58, the protein appeared to accumulate mostly in the Golgi complex and could also clearly be detected in the ER, but no increase in vesicular elements was observed. This is in contrast to the vesicular pattern and suggested enlargement of the intermediate compartment observed in Vero cells overexpressing ERGIC-53 (Schindler et al., 1993). In rat pancreas the staining of p58 in actively secreting acinar cells is more punctate and disperse, as compared to the reticular Golgi-staining pattern of ␤-cells (Lahtinen et al., 1992). It is thus possible that the distribution of p58 at the ER-Golgi interface correlates with the secretory status of the cell. In fibroblasts, where the level of secretion is relatively low, the amount of cargo proteins and/or cytosolic factors, e.g. coatomers, could be limiting in cells overexpressing p58, and therefore the production of anterograde and/or retrograde transport vesicles would not be enhanced. This could result in the preferential accumulation of the protein in donor or acceptor organelles, respectively.
In non-permeabilized cells patchy staining of the cell surface, sometimes concentrated close to the cell edges, could also be visualized. Whether these structures represent clathrin-coated pits or caveolae is presently unknown. The same surface staining pattern has also been described for ERGIC-53 in overexpressing COS cells (Kappeler et al., 1994). This is likely to be due to saturation of the intracellular components needed for the retention of p58, resulting in leakage of the protein to the plasma membrane. The cytoplasmic tails of p58 and ERGIC-53 contain a double-lysine motif (Nilsson et al., 1989;Jackson et al., 1990), which mediates retrieval of membrane proteins from the Golgi to the ER (Jackson et al., 1993;Townsley and Pelham, 1994;Gaynor et al., 1994). Overexpressed ERGIC-53 has been shown to be endocytosed from the plasma membrane (Kappeler et al., 1994). Surprisingly, the C-terminal amino acid sequence K-K/R-F/Y-F/Y needed for internalization was found to be re- FIG. 6. Expression of c-Myc-tagged p58. c-Myc-tagged p58 was expressed by using the Semliki Forest virus system. Immunofluorescence double-staining with anti-c-Myc antibodies (A) and anti-mannosidase II antibodies (B). Overexpression of c-Myc-tagged p58 and immunofluorescence staining of permeabilized (C) and nonpermeabilized (D) cells with anti-c-Myc antibodies. lated to the KKXX ER retrieval signal (Itin et al., 1995).
The immunofluorescence staining pattern obtained with the peptide antibodies made against the short cytoplasmic tail appeared to differ from the staining seen with antibodies made against a lumenal sequence. Anti-tail antibodies recognized the subpopulation of p58 residing in the Golgi cisternae, indicating that the epitope is masked or not accessible in the vesicular elements, and exposed only in the Golgi complex. Knowing the distribution of p58 in the Golgi complex (Saraste et al., 1987), this is predicted to be the cis cisternae of the Golgi. Whether anti-tail antibodies can be used as a specific cis-Golgi marker has to be resolved at the electron microscopic level. The differential exposure of the double-lysine ER retrieval signal might regulate the trafficking of p58 by at least two mechanisms. First, the accessibility of the double-lysine to the retrieval machinery might correlate with different homo-oligomeric forms of p58 (Lahtinen et al., 1992). For example during the oligomeric assembly of the IgE receptor the ␣-chain is retained in the ER. Steric masking of the two lysines when the assembly with the ␥-chain has been completed allows the receptor complex to be transported from the ER (Letourneur et al., 1995). Second, specific protein-protein interactions might be responsible for the transport to the Golgi complex, and recycling back to the ER. Some of the coatomer subunits interact directly with double-lysine containing peptides in vitro . Recently, coatomer subunits have been shown to be essential for the retrieval pathway .
The ectodomains of p58 and VIP36, an integral membrane protein isolated from exocytic carrier vesicles in epithelial Madin-Darby canine kidney cells , were found to be related, sharing 30% identity and 53% similarity. Both p58 and VIP36 are non-glycosylated type I integral membrane proteins but localize to different parts of the secretory pathway. VIP36 was recently shown to be related to leguminous plant lectins . Although the similarity at the primary amino acid level was quite low, the secondary structure and critical amino acids involved in sugar residue and metal binding were conserved. As a consequence of the lectin properties and the cellular localization a possible function in the sorting of glycolipids and GPI anchored proteins was proposed . Based on the sequence homology between VIP36 and ERGIC-53,  postulated a new family of animal lectins operating at different stages along the secretory pathway. It is worth mentioning that the critical three amino acids needed for lectin activity pointed out in the ERGIC-53 sequence are conserved both in the rat and frog proteins. The argument for a possible lectin function was strengthened, when MR60, a human mannose-binding lectin purified by affinity chromatography, was cloned and found to be identical to ERGIC-53 differing only in one amino acid residue (Pimpaneau et al., 1991;Arar et al., 1995). MR60 was also shown to have structural similarities with galectins, animal ␤-galactoside binding lectins. Lectin-like affinity for glycoproteins containing terminal mannoses might be needed for sorting and concentrating proteins into transport vesicles in the ER (Mizuno and Singer, 1993;Balch et al., 1994;Pelham, 1995). This would be followed by delivery of the ligand in the the Golgi complex, and the recycling of unoccupied receptor back to the ER. Alternatively, p58 could have a function for retrieval of escaped, not fully matured, or folded glycoproteins back to the ER.