Profile-based Data Base Scanning for Animal L-type Lectins and Characterization of VIPL, a Novel VIP36-like Endoplasmic Reticulum Protein*

Consensus profiles were established to screen data bases for novel animal L-type lectins. The profiles were generated from linear sequence motifs of the human L-type lectin-like membrane proteins ERGIC-53, ERGL, and VIP36 and by optimal alignment of the entire carbohydrate recognition domain of these proteins. The search revealed numerous orthologous and homologous L-type lectin-like proteins in animals, protozoans, and yeast, as well as the sequence of a novel family member related to VIP36, named VIPL for VIP36-like. Sequence analysis suggests that VIPL is a ubiquitously expressed protein and appeared earlier in evolution than VIP36. The cDNA of VIPL was cloned and expressed in cell culture. VIPL is a high-mannose type I membrane glycoprotein with similar domain organization as VIP36. Unlike VIP36 and ERGIC-53 that are predominantly associated with postendoplasmic reticulum (ER) membranes and cycle in the early secretory pathway, VIPL is a non-cycling resident protein of the ER. Mutagenesis experiments indicate that ER retention of VIPL involves a RKR di-arginine signal. Overexpression of VIPL redistributed ERGIC-53 to the ER without affecting the cycling of the KDEL-receptor and the overall morphology of the early secretory pathway. The results suggest that VIPL may function as a regulator of ERGIC-53. these primer proof-reading polymerase 5 (cid:2) -end primer an additional Bam HI site and the Kozak transcription initiation sequence preceding the start AUG codon of VIPL. 3 (cid:2) -end primer had a Xba I site the stop codon. The cDNA was cloned into pcDNA 3.1 (Invitrogen) via Bam HI and Xba I sites. A HA epitope (YPYDVPDYA) was introduced downstream of the signal sequence cleavage site between amino acids 44 and 45 of full-length VIPL. This construct was generated by PCR-based splicing (37). Selected amino acids were substituted by oligonucleotide-directed PCR mutagenesis or sequence-over- lap-extension PCR (38), and mutant fragments were recloned via Bam HI/ Xba I or Sac II/ Xba I sites into VIPL constructs. For creating chimeric GM constructs with the cytoplasmic tail of VIPL (GM-ViTa), a Bgl II restriction site was introduced adjacent to the transmembrane domain of GM by silent mutagenesis changing the codon of the arginine 499 in the tail to AGA. This site allows insertion of fragments via Bgl II and Xba I into GM constructs. A cDNA encoding the cytoplasmic tail of VIPL was prepared by annealing complementary oligonucleotides as described (10) and into the GM construct via Bgl II and Xba I sites. Additional mutations were introduced by PCR and recloned as Acc I/ Xba I fragments into GM-ViTa construct. All constructs were confirmed by sequencing using standard methods an ABI

Understanding the molecular basis of secretion requires knowledge of how secretory proteins are correctly folded and assembled in a process known as quality control, and how these itinerant proteins are sorted from resident proteins along the secretory pathway. Evidence is mounting that intracellular lectins play an important role in these processes (1,2). A majority of secretory proteins acquires N-glycans during translocation into the ER, 1 and these modifications are most often required for the folding of the proteins. The folding process is assisted by the membrane lectin calnexin and the related soluble lectin calreticulin. These two lectins recognize monoglucosylated trimming intermediates and, in a remarkable cycle of de-and reglucosylation, guarantee that only correctly folded glycoproteins leave the ER (2)(3)(4).
If folding is unsuccessful, newly synthesized glycoproteins are degraded by the proteasome after retrotranslocation to the cytosol. Degradation appears to involve calnexin and an ER ␣-mannosidase-like protein, termed EDEM in mammalian cells, that enhances degradation of misfolded glycoproteins carrying Man 8 (GlcNAc) 2 glycans (5-7). The strict functional dependence of EDEM on a Man 8 (GlcNAc) 2 structure suggests it may function as a lectin.
Correctly folded glycoproteins leave the ER in COPII-coated vesicles (8) and are transported to the Golgi via the ER-Golgi intermediate compartment (ERGIC,Ref. 9). The exit signals that direct proteins out of the ER are poorly understood with few exceptions (10). The mannose lectin ERGIC-53 (p58 in rat) mediates efficient ER-exit of some secretory glycoproteins by serving as a transport receptor (9,(11)(12)(13). ERGIC-53-assisted proteins include blood coagulation factors V and VIII (14), cathepsin C (15), and cathepsin Z (16). ERGIC-53 is a major membrane protein of the tubulovesicular clusters of the ERGIC (17) and efficiently cycles between ERGIC and ER (18 -20). A second lectin cycling in the early secretory pathway, VIP36 is related to ERGIC-53 (21). VIP36 is associated with ERGIC and Golgi (22) and to some extent with the plasma membrane (23). VIP36 binds mannose (24) and GalNAc (25) and may operate in glycoprotein transport to the apical plasma membrane of polarized cells (26).
ERGIC-53 and VIP36 are type I membrane proteins. In their luminal segment they carry a single domain of about 200 amino acids that exhibits sequence similarity to the carbohydrate recognition domain (CRD) of soluble lectins of leguminous plants (21). This domain is known as L-type CRD (27). The L-type lectin-like domain of animal lectins is designated LTLD in the present study. The family of plant L-type lectins includes numerous members whereas only three L-type lectins are known in animals: ERGIC-53, VIP36, and a recently discovered ERGIC-53-like protein termed ERGL (28). Unlike ERGIC-53 and VIP36 that are ubiquitously expressed proteins, the ERGL gene is expressed in a limited number of tissues only, with highest mRNA levels in normal and neoplastic prostate cells. The ERGL protein remains to be characterized.
We wondered if there are additional, yet unidentified members of the animal L-type lectin family. To identify such proteins we established profiles characteristic for animal L-type lectins. By scanning data bases with these profiles we identified orthologous and homologous L-type lectin-like proteins in animals, protozoans, and yeast. A novel member of this family displays molecular similarities with VIP36 and was therefore dubbed VIPL (for VIP36-like). The corresponding protein was cloned and expressed in cell culture. VIPL was found to localize to the ER and to affect the recycling of ERGIC-53.
Recombinant DNA-Standard molecular biology protocols were adapted from Refs. 32 or 33. Oligonucleotides were from Microsynth (Switzerland). ERGIC-53 constructs containing a Myc-tag and an artificial N-glycosylation site, termed GM, have been described (34,35). VIPL cDNA was obtained by RT-PCR. Total RNA of subconfluent HepG2 cells was isolated using peqGOLD RNApure reagent (peqLab, Biotech GmbH, Germany) with a protocol modified from Ref. 36. Reverse transcription was done with 2 g of total RNA using Omniscript RT enzyme (Qiagen, Switzerland) and oligo d (T) 14 V primer. Subsequent PCR was performed with VIPL-specific primer pairs and Tgo DNA proof-reading polymerase (Roche Molecular Biochemicals, Switzerland). The 5Ј-end VIPL primer contained an additional BamHI site and maintains the Kozak transcription initiation sequence preceding the start AUG codon of VIPL. The 3Ј-end primer had a XbaI site after the stop codon. The resulting cDNA was cloned into pcDNA 3.1 vector (Invitrogen) via BamHI and XbaI sites. A HA epitope (YPYDVPDYA) was introduced downstream of the signal sequence cleavage site between amino acids 44 and 45 of full-length VIPL. This construct was generated by PCR-based splicing (37). Selected amino acids were substituted by oligonucleotide-directed PCR mutagenesis or sequence-overlap-extension PCR (38), and mutant fragments were recloned via BamHI/XbaI or SacII/XbaI sites into VIPL constructs. For creating chimeric GM constructs with the cytoplasmic tail of VIPL (GM-ViTa), a BglII restriction site was introduced adjacent to the transmembrane domain of GM by silent mutagenesis changing the codon of the arginine 499 in the tail to AGA. This site allows insertion of fragments via BglII and XbaI into GM constructs. A cDNA encoding the cytoplasmic tail of VIPL was prepared by annealing complementary oligonucleotides as described (10) and cloned into the GM construct via BglII and XbaI sites. Additional mutations were introduced by PCR and recloned as AccI/XbaI fragments into GM-ViTa construct. All constructs were confirmed by sequencing using standard methods and an ABI Prism 310 Genetic Analyser (PE Applied Biosystems, Switzerland).
Cell Culture and Transfection-COS-1 cells were cultured and transfected (DEAE-dextran method) as described (35). HEK293 cells (kindly provided by T. Meier, Myocontract, Switzerland) were cultured in Dulbecco's minimal essential medium (4.5 g/liter glucose) supplemented with 10% fetal calf serum and 100 IU/ml penicillin, 100 g/ml streptomycin, and 1 g/ml fungizone. HEK293 cells were transfected by the calcium phosphate precipitation method (32). HepG2 cells were cultured as described (20) and transfected by use of FuGENE 6 transfection reagent (Roche Molecular Biochemicals, Switzerland). For immunofluorescence experiments cells were plated in poly-L-lysine-coated 8-well multichamber glass slides (Lab Tek, Nalgene-Nunc Intl.). All cultures were grown at 37°C with 5% CO 2 in humidified air, and transfection of cells was carried out 1 day after plating.
Metabolic Labeling and Immunoprecipitation-42 h after transfection the cells were washed twice with phosphate-buffered saline, starved in labeling medium (MEM without methionine, supplemented with 10% dialyzed fetal calf serum) and pulsed with 100 Ci/ml [ 35 S]methionine/cysteine (EasyTag™ EXPRE 35 S 35 S Protein Labeling, PerkinElmer Life Sciences). Cells were immediately processed or chased with complete Dulbecco's minimal essential medium containing 10 mM L-methionine. For immunoprecipitation the cells were washed twice with ice-cold phosphate-buffered saline and resuspended in lysis buffer (100 mM sodium phosphate, 1% Triton X-100, pH 8) supplemented with 0.2 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, and 0.5 g/ml pepstatin). In some experiments 20 mM iodoacetamide or N-methylmaleimide was included in the lysis buffer. Cells were lysed by passing them five times through a 25-gauge needle. After 1 h on ice, the lysate was cleared by centrifugation at 100,000 ϫ g for 1 h. The super-natant was added to protein A-Sepharose beads (Amersham Biosciences) to which antibodies had been prebound. After incubation for at least 1 h on a rotary shaker in the cold, the beads were washed four times with lysis buffer, once with 100 mM sodium phosphate (pH 8) and once with 10 mM sodium phosphate (pH 8).
Endoglycosidase Digestion-For digestion with endoglycosidase H (endo H) the immunoprecipitates were boiled for 3 min in 50 mM TrisCl, 1% SDS, and 0.1 M ␤-mercaptoethanol (pH 6.8). An equal volume of 0.15 M sodium citrate (pH 5.3) supplemented with protease inhibitors was added, and digestion with 10 milliunits of endo H (Roche Molecular Biochemicals) was carried out at 37°C overnight. For digestion with endoglycosidase F (PNGase F), the immunoisolates were boiled for 3 min in 100 mM sodium phosphate, 0.1% SDS, 0.1 mM ␤-mercaptoethanol, and 10 mM EDTA (pH 7.2). An equal volume of the same buffer containing 1% Triton X-100 instead of SDS was added and the sample was incubated with 400 milliunits of PNGase F (Roche Molecular Biochemicals, Switzerland) at 37°C overnight.
Sodium Carbonate Extraction-Membrane association of proteins was tested by the sodium carbonate procedure (39). After metabolic labeling, the cells were collected in ice-cold homogenization buffer (20 mM Hepes/KOH, 300 mM sucrose, 0.2 mM phenylmethylsulfonyl fluoride, 20 mM NEM, pH 7.4), resuspended in 0.1 M sodium carbonate (pH 11.5), passed 10 times through a G25 needle, and kept on ice for 30 min. The sample was then layered onto a small cushion of homogenization buffer, and membranes were pelleted by centrifugation at 100,000 ϫ g for 1 h and resuspended in lysis buffer (see immunoprecipitation). The pH of the supernatant was neutralized by adding one-fifth volume of 0.5 M potassium phosphate (pH 8), and Triton X-100 was added to 1% final concentration. All samples were kept on ice for 30 min before centrifugation at 100,000 ϫ g for 1 h. The supernatants were subjected to immunoprecipitation.
Subcellular Fractionation-42 h after transfection with VIPL-HA cDNA, HepG2 cells were incubated for 3 h with 5 g/ml BFA or with solvent only before subcellular fractionation using Nycodenz gradients (35). The distribution of organelle markers and VIPL-HA was determined by Western blotting.
Gel Electrophoresis and Western Blotting-Samples were separated by gradient SDS-PAGE. Radioactivity was visualized by fluorography using sodium salicylate and BioMax MR-1 films (Rochester). Fluorograms were quantified with a ChemImager TM and AlphaEase TM software (Alpha Inotech Corporation). For Western blotting, proteins were transferred to nitrocellulose B85 (45 m, Schleicher & Schuell, Germany) at 100 V for 1 h in the cold using transfer buffer containing 15.6 mM Tris, 120 mM glycine, and 20% (v/v) methanol (pH 8.4). The nitrocellulose was rinsed with 50% methanol and stained by Amido Black (Serva, Germany). All subsequent incubations were in PBS containing 5% nonfat dry milk and 0.05% Tween 20 (Serva, Germany): Blocking for 1 h, incubation with the first antibody for 60 min, rinsing three times 10 min, incubation with peroxidase-coupled secondary antibody for 60 min, and rinsing 10 min. After three final 10-min washes with phosphate-buffered saline containing 0.05% Tween 20, the nitrocellulose was processed using enhanced chemiluminescence (ChemiGlow ECL reagent, AlphaInotech Corp.) and exposed to BioMax MR-1 films or directly analyzed in a ChemImager TM .
Immunofluorescence Microscopy-The procedure has been described (10). Specimens were examined with a Polyvar microscope or a Leica confocal laser scanning microscope.
Biocomputing-The GCG programs (Madison, WI) were used for sequence analysis (40). Other software used is available at ExPASy server (www.expasy.org/) (41). SwissProt (release 40.7) and TrEMBL/ release 19.1) data bases were searched. MEME/MotifSearch (42) was performed with full-length sequences of human ERGIC-53, ERGL, and VIP36 with default settings. The algorithm automatically generated a profile, based on settings with 6 motifs of 8 amino acids in length each with one or zero occurrence in each sequence. A ProfileSearch (43) against SwissProt and TrEMBL data bases was done with a profile generated of the alignment of ERGIC-53 (amino acids 44 -292), VIP36 (amino acids 50 -286), and ERGL (amino acids 31-271). For secondary structure predictions we used the GCG program PeptideStructure and PredictProtein (44). Potential sites for domains and motifs were identified by MOTIF search in the Prosite library (45) and by PSORT (psort.ims.u-tokyo.ac.jp). Coiled coil regions were identified with CoilScan (46), and signal sequences with SPScan and signalP (47). For detection of putative transmembrane domains the programs HMMTOP (48), and TMpred were applied in combination with structure predictions. The phylogenetic tree was constructed using the GrowTree software of the GCG package, and distance correction was calculated according to Jukes-Cantor.

Consensus Profiles for the Identification of Animal L-type
Lectins-Profiles for the identification of lectins have been established in several data bases (27). For plant L-type lectins such profiles include PS00307 and PS00308 in PROSITE (45) and PF00138 and PF00139 in Pfam (49). We noticed, however, that these profiles do not identify animal L-type lectins accurately in the existing data bases. Therefore, we developed new consensus profiles specific for animal L-type lectins.
In a first approach we used software MEME/MotifSearch of GCG (42) to establish a consensus motif pattern from nonaligned full-length sequences of the known human L-type lectins ERGIC-53, ERGL, and VIP36. The algorithm generated a consensus motif pattern of 6 linear sequence motifs of 8 amino acids each with single occurrence in a defined order. All sequences were restricted to the LTLD. One motif encompassed part of loop A of the CRD of L-type lectins (50) that includes a conserved aspartate required for sugar and metal binding (Fig.   1B). Two motifs covered the two cysteines known to form an intracellular disulfide bond in native ERGIC-53 (Ref. 51 and our own observations). The cysteines are conserved in the LTLD of all three human lectins and in most of their orthologs assembled in Fig. 1. The other four motifs encompassed conserved regions of ␤-sheets 6 and 15 of ERGIC-53/p58 as well as the 3 10 helix turn that separates the ␤-sheets 1a and 1b of ERGIC-53/p58 (51).
A second consensus profile for proteins with a putative LTLD was generated by a ProfileSearch method included in the GCG package (43). This approach is based on a sequence matrix profile that was generated on the optimal alignment of the LTLDs of human ERGIC-53, ERGL, and VIP36. The similarity of LTLDs was found to be 58% for ERGIC-53 and ERGL, 46% for ERGIC-53 and VIP36, and 45% for VIP36 and ERGL. The algorithm of the profile scan calculates the score (quality) of the optimal alignment between the consensus profile and each sequence in the scanned data bases.  Table I. B, alignment of partial LTLDs of the human L-type lectins ER-GIC-53, VIP36, and VIPL with selected plant lectins (Favin, Lol I, LSL, Con A, and ECorL), as well as the putative lectins Uip5 of yeast (SwissProt P36137), Q9GBP0 of T. bruci and Q9GRK5 of L. leishmania. Shown are sequences from ␤-sheet 6 -15 referring to the resolved structure of rat ERGIC-53/p58 (51). The alignment performed with ClustalW (69) was edited based on defined or predicted secondary structures of the lectins and formatted by ESPript program (prodes.toulouse.inra.fr/ESPript/cgi-bin/ nph-ESPript_exe.cgi). The secondary structures are indicated above the sequences. Boxed arrows indicate ␤-sheets, cylinders indicate ␣-helices. Conserved cysteines are highlighted by $. Key residues for metal and sugar binding are marked by asterisks. Addition residues important for metal and sugar binding in plants are marked by circles. Black boxes represent identity and white boxes similarity.
The scored proteins of both scanning methods were then analyzed by alignment and comparison to animal L-type lectins, and the presence of residues functionally important for metal/sugar binding was investigated. Moreover, structural features of the LTLD were examined by hydrophilicity plots and secondary structure predictions. Finally we analyzed the identified proteins for possible protein motifs and domains.
Identification of Orthologs and New Family Members of Animal L-type Lectins by Profile Scanning-Both profile scans recovered all the known orthologs of ERGIC-53, ERGL, and of VIP36 from the data bases (Table I and Fig. 1A). The scores were higher for ERGIC-53 orthologs since profile characteristics were contributed mostly by ERGIC-53 and ERGL and to a lesser extent by VIP36. Two entries (SpTrEMBL accession numbers Q9H0V9 and Q9BQ14) also appeared with high scores. They stand for the same sequence that encodes a novel VIP36-like protein, we termed VIPL (VIP36-like, see below).
With lower scores the scans also identified lectin-like proteins in yeast. These proteins are the ERGIC-53 like Emp47p (52) and its close relative Emp46p of Saccharomyces cerevisiae (56% similarity to each other) and two entries of Schizosaccharomyces pombe (SpTrEMBL O42707 and O94401). Emp47p and Emp46p are type I membrane proteins with a putative LTLD fold and a stalk containing a coiled coil domain preceding the transmembrane domain. Emp47p contains a DXL(X) 5 N metal/ sugar binding motif in loop C of the LTLD that is found in the plant lectins UEA I and II as well as LAA I (50). By contrast, animal lectins have a metal/sugar binding site of the DXF/YXN type in loop C that is common to most plant lectins (Fig. 1B). The Emp46p protein does not contain any of the key residues for metal/sugar binding although a LTLD fold is predicted. A deletion of Emp47p results in intolerance to high Ca 2ϩ concen-tration in the growth medium, while no such phenotype was observed in an Emp46p deletion strain (53). Nevertheless, overexpression of Emp46p can suppress the Emp47p defect, suggesting that the two proteins functionally overlap.
Both lectin-like proteins of S. pombe contain a LTLD fold similar to animal L-type lectins. Examination of the similarity to animal lectins and of the domain organization suggests that entry O42707 is related to Emp47p and Emp46p whereas entry O94401 is presumably the ortholog of the novel VIPL (Figs. 1A and 2). While the O94401 protein has all the hallmarks of a functional LTLD, the O42707 protein lacks typical key residues required for metal and sugar binding in its LTLD fold.
The scans identified two putative lectins in the protozoans Trypanosoma cruzi (SpTrEMBL Q9GPB0) and Leishmania major (SpTrEMBL Q9GRK5). These proteins exhibit a LTLD fold similar to that of animal L-type lectins including the putative intramolecular disulfide bond (Fig. 1B). Both lectins contain predicted coiled-coil regions downstream of the LTLD similar to ERGIC-53, but apparently no transmembrane segment, suggesting that they are secretory proteins.
Rescanning the data bases with consensus profiles that include the newly identified VIPL sequence gave a result similar to that of the initial screen, with the exception that an additional protein of S. cerevisiae was uncovered, known as Uip5 (Ulp1-interacting protein 5, SwissProt P36137). The protein is predicted to be a type I membrane protein with a signal sequence, suggesting localization in the secretory pathway. The LTLD fold of Uip5 resembles that of animal L-type lectins. It comprises all key residues for metal/sugar binding of a functional LTLD (Fig. 1B) but does not display the conserved cysteine residues of the animal LTLD. Uip5 has been identified in a two-hybrid screen as interacting partner of the Smt3-specific protease Ulp1 that mediates deconjugation of Smt3 from septin components during G 2 /M phase transition of the cell cycle (54). A GFP-tagged version of Uip5 was localized to the nuclear membrane. No specific phenotype was found in the deletion strain. VIPL, a Novel Gene Related to VIP36 -Alignment of the SpTrEMBL entries Q9H0V9 (gene designation DKFZP564L2423) and Q9BQ14 revealed that they encode the same hypothetical protein although the latter lacks a defined N terminus. The predicted protein is 43% similar (35.2% identity) to human ERGIC-53 and 68% similar (57.8% identity) to human VIP36. Because of its high similarity to VIP36 we named the protein VIPL for VIP36-like. The LTLD fold of VIPL is similar to that of ERGIC-53 and VIP36. It contains all residues required for a functional LTLD (Fig. 1B) including the conserved cysteine residues required for intramolecular disulfide bond formation in ERGIC-53.
VIPL is predicted to be a type 1 membrane protein with a cytoplasmic tail of 12 amino acids. Unlike ERGIC-53, the LTLD of VIPL is not followed by a stalk, and no coiled coil is predicted. Thus, VIPL exhibits a domain organization very similar to VIP36. The VIPL protein is expected to be translocated into the ER by a 38-amino acid long N-terminal signal sequence. The mature protein (residues 39 -348) has a predicted mass of 35.6 kDa and a pI of 7.51. Moreover, the net charge is almost zero in a pH range of 6.8 -8.0.
Detailed examination by gap alignments and evolutionary analysis indicates that the previously found lectins of Drosoph-ila (SpTrEMBL Q9VCC2) and Caenorhabditis elegans (SpTrEMBL Q22170) are orthologs of VIPL rather than VIP36, suggesting that VIP36 appeared later in evolution than VIPL. An alignment of human VIPL and its orthologs in Drosophila (55.1% similarity), C. elegans (53.5% similarity), and the newly discovered S. pombe ortholog (O94401, 44.2% similarity) is shown in Fig. 2. The mouse VIPL (NCBI XP_129848) exhibits 92.5% similarity to the human sequence. Sequence conservation among VIPL orthologs is striking in the LTLD fold, particularly in the functional loops comprising the key residues for metal and sugar binding (Fig. 2). The cysteine residues at positions 200 and 237 of human VIPL are also conserved.
The human gene DKFZP564L2423 (EMBL AL136617) for VIPL is located on chromosome 2 (q11.2) and has a total length of 34.14 kb with 8 exons. For comparison, the gene of human VIP36 with a total length of 14.9 kb also contains 8 exons but is located on chromosome 5 (q35.5). The VIPL gene has an ATG start codon within a proper initiation context (55) and an inframe termination codon. The 5Ј-untranslated region showed no other open reading frame of significant length suggesting the defined start site is correct. The correctly spliced cDNA has a total length of 2416 bp with a coding region of 1046 bp. An expression profile obtained from cDNA and expressed sequence tag (EST) sources (UniGene cluster Hs. 18627 for VIPL at NCBI server, and SAGE expression profile of VIPL gene with Ensemble ID ENSG00000114988 at EBI server) suggests that VIPL is a ubiquitous protein expressed in many cell types. VIPL Is an N-Glycosylated Membrane Protein of the ER-To study the VIPL protein the corresponding cDNA was isolated from total RNA of HepG2 cells by RT-PCR. The cloned sequence was identical to the hypothetical cDNA of the gene DKFZP564L2423. VIPL was tagged with a HA tag and expressed in various cell lines. To test for membrane association, homogenates of transfected HEK293 cells metabolically la-beled with [ 35 S]methionine were subjected to the carbonate/pH 11 extraction procedure. Like the integral membrane protein ERGIC-53, VIPL-HA quantitatively distributed to the pellet fraction, while Sec31p, a component of the COPII coat that is peripherally associated with membranes was quantitatively recovered in the soluble fraction as expected (Fig. 3A). The results confirm the prediction that VIPL is an integral membrane protein.
VIPL carries a single consensus site for N-glycosylation at position 181 (Fig. 2). To test if this site is used, VIPL-HA was immunoprecipitated from cells and treated with glycosidases. Fig. 3B shows that VIPL-HA is sensitive to both endo H and PNGase F indicating that VIPL is a glycoprotein. The persistent sensitivity to endo H suggests that the N-glycan of VIPL is of the high-mannose type. This finding was confirmed by pulsechase experiments in which VIPL remained endo H-sensitive throughout a 3-h chase (Fig. 3C). Similarly, VIPL-HA did not acquire Endo D sensitivity when expressed in Lec-1, a cell line deficient in GlcNAc-transferase (data not shown).
We next studied the localization of transiently transfected VIPL-HA by immunofluorescence microscopy. In HepG2 cells VIPL-HA displayed a reticular pattern typical for ER (Fig. 4, E-H). Double immunofluorescence microscopy with the ER marker BAP31 confirmed this finding (Fig. 4A). A similar localization was found in COS-1, HEK293, Lec-1, and HeLa cells ( Fig. 7B and not shown).
To test if VIPL cycles in the early secretory pathway, VIPL-HA transfected cells were treated with BFA. BFA redistributes rapidly cycling proteins of ERGIC and Golgi, such as ERGIC-53 (18), VIP36 (22), and the KDEL receptor (56), to characteristic punctate structures in the cytoplasm that remain separate from the ER. While BFA treatment led to the expected punctuate pattern for the KDEL receptor (Fig. 6, E and G) it had no effect on the localization of VIPL-HA (Fig. 6,  D, H, L). These results suggest that VIPL-HA does not cycle between ER and post-ER compartments.
We confirmed the ER resident localization of VIPL-HA by subcellular fractionation (Fig. 5). A postnuclear supernatant of HepG2 cells transiently expressing VIPL-HA was separated by Nycodenz density gradient centrifugation, and the position of VIPL-HA and organelle markers was probed by SDS-PAGE followed by immunoblotting. Fig. 5A shows the expected distribution of BAP31 (ER), GPP130 (cis-Golgi), and ERGIC-53 (ER and ERGIC) (20). VIPL co-distributed with the ER. Upon BFAtreatment, GPP130 relocalized to ER fractions at the bottom of the gradient (Fig. 5B), while ERGIC-53 accumulated in the ERGIC near the top of the gradient typical for cycling proteins.
The distribution of VIPL-HA was largely unchanged by BFA treatment. The results confirm the IF data that VIPL is a non-cycling ER protein.
Overexpression of VIPL Affects the Localization of ERGIC-53-ERGIC-53 is localized in punctuate structures scattered in the cytoplasm with concentration near the Golgi (17) (Figs. 4B  and 6I). Surprisingly, overexpression of VIPL led to a striking redistribution of ERGIC-53 to the ER (Figs. 4B and 6I). This effect was observed in all the tested cell lines transiently transfected with VIPL. Interestingly, the relocalized ERGIC-53 was insensitive to BFA. It remained in the ER without accumulation in dots (Fig. 6K). To test if the redistribution of ERGIC-53 to the ER required an active LTLD we substituted the conserved asparagine 163 of VIPL by aspartate (N163D). This mutation can be expected to disrupt VIPL's lectin function. VIPL(N163D) still redistributed ERGIC-53 to the ER suggesting that an active LTLD was not required (not shown). Likewise, this redistribution does not involve an active LTLD of ERGIC-53 since the lectin-impaired N156A mutant of ER-GIC-53 was also redistributed (not shown) To test if overexpression of VIPL affects the morphology of the early secretory pathway we studied the localization of Sec31p (a component of COPII coats) and GPP130 by immunofluorescence microscopy. Figs. 4 and 6 show that the distribution of these markers was unchanged suggesting that the general morphology of the secretory pathway is unaffected by overexpressing VIPL. Moreover, the normal distribution of the cycling proteins KDEL-receptor (Fig. 6) and VIP36 (not shown) in VIPL-overexpressing cells indicates that VIPL does not just arrest any rapidly cycling protein in the ER.
ER Retention of VIPL Requires an RKR Motif in the Cytoplasmic Segment-The localization of VIPL to the ER is surprising since its cytoplasmic segment possesses a putative ERexit motif (i.e. an aromatic amino acid in position Ϫ2, Ref. 10) but no di-lysine ER localization signal typical for many type I ER membrane proteins (57). ER localization is also unexpected because VIP36, having a similar C-terminal tail sequence (Fig.  7A), is localized to Golgi and ERGIC (22), and the C-terminal four amino acids of VIP36 can mediate efficient ER export when appended to ERGIC-53 (58).
To test if ER retention of VIPL is caused by its luminal domain we generated a VIPL mutant lacking the transmembrane and cytosolic domains. This soluble mutant was secreted into the cell culture medium (not shown) suggesting that the ER localization signal is not contained in the luminal domain. We noted a conserved RKR motif in the cytoplasmic tail of all orthologs of VIPL that is not present in VIP36 (Fig. 7A). The motif is reminiscent of di-arginine motifs (RR or RXR) initially characterized as ER retention motifs near the N terminus of type II membrane proteins (57,59) and near the cytoplasmic C terminus of multispanning membrane proteins (60). To study the role of the RKR motif in ER retention we substituted it by three serines. Furthermore, to prevent interference by a pos-sible lectin activity of VIPL, we also mutated asparagine 163 to aspartate in the putative lectin domain. Like wild-type VIPL-HA, VIPL(N163D)-HA was localized to ER in transfected COS cells (Fig. 7B, panels a and b). By contrast, VIPL lacking the RKR motif was also found at the cell surface in addition to intracellular membranes (Fig. 7B, panel c). This became more obvious when the cells were not permeabilized. While hardly any staining of VIPL-HA was detected in cells transfected with wild-type or VIPL(N163D)-HA cDNA (Fig.  7B, panels d and e), surface staining of VIPL-HA lacking the RKR motif was clearly evident (Fig. 7B, panel f). Despite its transport to the cell surface the SSS mutant of VIPL-HA remained endo H-sensitive as revealed by pulse-chase experiments (Fig. 7C). In addition, no acquisition of Endo D sensitivity was observed when expressing the VIPL constructs in Lec-1 cells (data not shown). The results of this mutagenesis approach indicate that the RKR motif is required for the ER localization of VIPL.
VIPL Lacking the ER Retention Signal Does Not Pull ER-GIC-53 to the Cell Surface-Does overexpression of VIPL lacking the RKR retention motif result in surface localization of ERGIC-53? To test this, HepG2 cells were transiently transfected with VIPL-HA carrying a RKR to SSS substitution and probed surface expression of ERGIC-53 by IF. The transfected cells showed intracellular but no apparent cell surface staining of ERGIC-53 (Fig. 8A), while the mutant VIPL exhibited cell surface in addition to intracellular staining, as expected (Fig.  8B). The results were confirmed by using non-permeabilized cells. Mutant VIPL but not ERGIC-53 was detectable at the cell surface (Fig. 8, C and D). These results suggest that the interaction of overexpressed VIPL with ERGIC-53 does not persist beyond the ER.

RKR-mediated ER Retention Can Be Overridden by an ERexit Signal Presented in an Oligomeric
Form-We wondered if the RXR motif was capable of retaining ERGIC-53 in the ER. To this end the cytoplasmic tail of VIPL was appended to an ERGIC-53 variant carrying a Myc-tag and a N-glycosylation site (GM, Fig. 9, A and B; Ref. 35). Appending different cyto- plasmic tails to GM does not interfere with its folding (10). The KK retrieval signal in GM has been substituted to alanines to prevent cycling. GM is efficiently transported from ER to Golgi when its C-terminal di-phenylalanine motif is present (GMA 5 FF), but transport is inefficient when the motif is replaced by two alanines (GMA 7 ) (Refs. 10, 35, and 61 and Fig. 9, C and D). When the tail of VIPL was appended to the GM reporter (GM-ViTa, Fig. 9, A and B), a slight but statistically significant reduction in ER to Golgi transport was observed (Fig. 9, C and D). When the RKR motif was mutated to serines the GM-mtViTa chimera was transported as efficiently as the GM construct containing the di-phenylalanine ER-export signal. The inability of the RKR motif to fully retain GM may be due to the fact that the potential ER exit motif "FY" of VIPL is more active in an oligomerized protein such as ERGIC-53 (Ref. 10). 2 Indeed, when we replaced the putative ER exit motif FY in the tail of GM-ViTa (named GM-ViTa(AA) by two alanines, the construct hardly became endo H-resistant (Fig. 9). Immunofluorescence analysis confirmed that GM-ViTa(AA) was localized to the ER (not shown). We conclude that the RKR motif can mediate efficient retention of VIPL despite the presence of an ER exit motif. However, the RKR motif cannot override the ER exit motif in a homo-oligomeric reporter construct. Furthermore, it appears that the RKR motif does not mediate ER retrieval since the GM-ViTa construct failed to localize to characteristic dots after BFA treatment (not shown).

DISCUSSION
Our profile scanning revealed notable differences between animal and plant L-type lectins. The differences include two conserved cysteines, present in the LTLD of animal but not plant lectins, that form an intramolecular disulfide bond in ERGIC-53 and presumably other animal lectins (Fig. 1B). Another characteristics is the unique turn of 3 10 helix in animal LTLDs that separates the ␤-sheets 1a and 1b (51). Further differences became obvious when the (partial) LTLDs of plant and animal L-type lectins were aligned (Fig. 1B). Animal lectins contain a shorter B loop but an elongated loop between the two ␤-sheets after loop C. The elongated loop forms an ␣-helix in ERGIC-53 (51) and is predicted to be also present in ERGL, VIP36, and the newly identified protozoan lectins, as well as in the putative yeast lectin Uip5. Uip5 has an additional predicted ␣-helix in loop B not present in animal and plant lectins (Fig. 1B). Yet another difference between plant and animal LTLDs concerns the residues mediating metal/sugar binding in loop C. While plant lectins possess a glutamate and an aspartate that are conserved and responsible for Mn 2ϩ binding, the animal lectins lack these residues (Fig. 1B). Accordingly, lectin activity of ERGIC-53 and VIP36 does not require Mn 2ϩ (12,25). Whether or not the protozoan lectins and the putative lectin Uip5 of yeast require Mn 2ϩ for lectin activity is difficult to predict since they lack the glutamate although the aspartate is conserved. Finally, the general domain organization also differs between plant and animal L-type lectins (including the orthologs and homologs in lower organisms, such as yeast). Animal lectins are type I membrane proteins as opposed to plant lectins that are soluble proteins.
The legume lectin fold has evolved independently several times (62). It is not only present in L-type lectins but also in galectins and pentraxins. However, the position of the ligand binding site is not conserved among the lectin families. In the animal L-type-like lectin family, major features of the carbohydrate binding site of legume lectins are conserved. Thus, a divergent evolution from a common ancestor is likely. We propose that Uip5 represents a common ancestor of plant and animal L-type lectins.
With the discovery of VIPL the family of L-type lectins in humans appears to be complete although some uncertainty remains since the annotation of the human genome has not been completed. The animal L-type lectin family comprizes the 4 members ERGIC-53, ERGL, VIP36, and VIPL. The predicted overall structure and domain organization of VIPL most closely resembles that of VIP36. Different from ERGIC-53, secondary structure predictions could not define an ␣-helix between the ␤-sheets after loop C in the VIPL LTLD, although, similar to ERGIC-53, the corresponding region is longer than in the LTLD of plant lectins (Fig. 1B). Interestingly, orthologs of VIPL are found in mouse, fly, worm, and yeast (S. pombe) but not S. cerevisiae whereas VIP36 is restricted to higher organisms (Fig. 1A). It appears, therefore, that VIP36 has evolved from VIPL, presumably by gene duplication.
Unlike ERGIC-53 and VIP36 that cycle in the early secretory pathway, VIPL is a resident of the ER when expressed in cultured cells. The following considerations argue against the possibility that the ER localization is due to misfolding. First, we have no evidence for aggregation of transfected VIPL, both biochemically and by immunofluorescence microscopy. Second, ER retention was saturable by high expression. Third, ER retention was abolished by mutating the conserved RKR motif in the cytoplasmic domain of the molecule or by deleting both the transmembrane and cytoplasmic domain. These observa-  (34). The transmembrane domain is followed by the amino acid sequence of the cytoplasmic tails in single letter code. The amino acid sequence of the mutant cytoplasmic tails is shown in panel B. GMA 7 and GMA 5 FF have been described (10,35). In chimeric GM-ViTa constructs, the underlined residues in cytoplasmic tail are RKR in wild-type constructs and SSS in mutant (mt) constructs. The most extreme FY motif in GM-ViTa (shown in italic letters in panels A and B) has been additionally mutated to AA in GM-ViTa(AA). C, effect of RKR motif on GM transport. COS-1 cells were transfected with the indicated constructs (panel B) and subjected to pulse-chase/endo H analysis using tions suggest that endogenous VIPL is also localized to the ER.
ER targeting of VIPL is mediated by a conserved di-arginine motif, RKR. VIPL is the first type I membrane protein with such a determinant. Di-arginine ER targeting motifs (RR or RXR) have been found in the N-terminal tail of type II membrane proteins (59) and more recently in the C-terminal tail of multispanning membrane proteins (60,63). Because lysines cannot substitute for the retention function of the arginines (59,63) and the X in RXR can be any basic or neutral amino acid but not an acidic one (63,64), we conclude that only the two arginines in the RKR motif are responsible for ER targeting of VIPL.
VIPL also possesses a C-terminal ER-exit motif (10) that appears to be dominated by the di-arginine ER retention signal. However, RKR was unable to retain a chimeric ERGIC-53 reporter protein carrying the same ER-exit motif. In contrast, the RKR motif in the cytoplasmic tail of Kir6.2, that lacks an ER exit motif, fully retains a CD4 reporter in the ER (63). This discrepancy is most likely due to the fact that the ERGIC-53 reporter is homo-oligomeric, while VIPL, like CD4, is not. 2 Our studies on transport motifs in ERGIC-53 revealed that ER-exit motifs are more efficient when presented in an oligomeric protein (Ref. 10). 2 Accordingly, inactivation of the ER-exit motif in the ERGIC-53 reporter led to ER retention of the construct in the present study. The results indicate that the RKR motif can indeed operate as an ER targeting signal in type I proteins.
An unexpected finding was the arrest of endogenous ER-GIC-53 in the ER upon overexpression of VIPL. The effect was selective since other rapidly cycling proteins, such as KDEL receptor and VIP36, were not affected by the overexpression of VIPL, and the secretory pathway was apparently unaffected. Since ER arrest of ERGIC-53 was also observed in cells expressing low apparent levels of VIPL, we assume that endogenous VIPL must be expressed at rather low levels. The redistribution of ERGIC-53 by VIPL overexpression was still observed when the lectin function of either protein was abolished by mutagenesis. This suggests that VIPL and ERGIC-53 interact in a lectin-independent manner, either directly or indirectly via a third component. Since overexpression of both proteins gave the same phenotype, the latter possibility appears less likely to us, unless a third interacting component is present in large excess. However, we could not detect any interaction by co-immunoprecipitation or by chemical crosslinking. The mechanism by which VIPL and ERGIC-53 interact remains to be elucidated. A hydrophobic surface patch opposite to the carbohydrate binding site of the LTLD of ERGIC-53 (51) may mediate this interaction given the fact that VIPL is uncharged at physiological pH. We speculate that VIPL may control the function of ERGIC-53 by modulating its exit from the ER.
Is VIPL a lectin? The presence of key residues for metal and sugar binding in the LTLD of VIPL would suggest lectin activity. However, HA-tagged VIPL failed to bind to immobilized mannose or fluorescein isothiocyanate-labeled bovine serum albumin conjugated with glucose, mannose, or GlcNAc under conditions used to successfully establish the mannose specificity of ERGIC-53 (12). Moreover, binding of ERGIC-53 to immobilized mannose was not affected by the presence of VIPL. The metal/sugar binding site of VIPL differs slightly from that of ERGIC-53 and VIP36 in loop C (Fig. 1B). VIPL contains an elongation by one amino acid after the key binding site. A similar elongation is found in Con A but not other plant lectins (Fig. 1B). This variance in length correlates with different affinities of plant lectins for different mannose/glucose derivatives (50). The slightly elongated C loop in VIPL may explain why experiments that demonstrated lectin activity of ER-CIG-53 or VIP36 failed for VIPL. Nevertheless, we speculate that VIPL possesses binding preference for mannose based on findings by Sharma and Surolia (50) who reported that legume lectins with a short D loop, such as Favin, LOL I, LSL, or Con A, preferentially bind mannose and glucose while lectins with a long D loop, such as EcoRL, prefer GalNAc (see Fig. 1B). Like ERGIC-53 and VIP36, VIPL has a short D loop.
An ortholog of VIPL was recently identified in zebra fish (EMBL Aam29497) that exhibits 71% similarity to human VIPL and is required for early development (65). Embryos lacking VIPL are touch insensitive and insensitive to tapping on the dish, but they are capable of spontaneous movements. We speculate that this defect may be caused by inefficient processing or transport of some secretory molecules in the absence of VIPL. A case in point is combined deficiency of coagulation factors V and VIII in humans, an inherited disease in which mutations in ERGIC-53 lead to inefficient secretion of coagulation factors V and VIII (14). Interestingly, some patients suffering from this disease have entirely normal ER-GIC-53 (66 -68). It is conceivable that such patients synthesize excessive levels of VIPL that would render ERGIC-53 nonfunctional by retaining it in the ER. The findings of the present study provide a basis to test this notion.