Structures of the Carbohydrate Recognition Domain of Ca2+-independent Cargo Receptors Emp46p and Emp47p*

Emp46p and Emp47p are type I membrane proteins, which cycle between the endoplasmic reticulum (ER) and the Golgi apparatus by vesicles coated with coat protein complexes I and II (COPI and COPII). They are considered to function as cargo receptors for exporting N-linked glycoproteins from the ER. We have determined crystal structures of the carbohydrate recognition domains (CRDs) of Emp46p and Emp47p of Saccharomyces cerevisiae, in the absence and presence of metal ions. Both proteins fold as a β-sandwich, and resemble that of the mammalian ortholog, p58/ERGIC-53. However, the nature of metal binding is distinct from that of Ca2+-dependent p58/ERGIC-53. Interestingly, the CRD of Emp46p does not bind Ca2+ ion but instead binds K+ ion at the edge of a concave β-sheet whose position is distinct from the corresponding site of the Ca2+ ion in p58/ERGIC-53. Binding of K+ ion to Emp46p appears essential for transport of a subset of glycoproteins because the Y131F mutant of Emp46p, which cannot bind K+ ion fails to rescue the transport in disruptants of EMP46 and EMP47 genes. In contrast the CRD of Emp47p binds no metal ions at all. Furthermore, the CRD of Emp46p binds to glycoproteins carrying high mannosetype glycans and the is promoted by binding not the addition of Ca2+ or K+ ion in These results suggest that Emp46p can be regarded as a Ca2+-independent intracellular lectin at the ER exit sites.

In eukaryotic cells, post-translational modification of secreted proteins and intracellular protein transport between organelles are ubiquitous features. One of the most studied systems is the N-linked glycosylation pathway in the synthesis of secreted glycoproteins (1)(2)(3)(4). The N-linked glycans are subject to diverse modification and transport from the endoplasmic reticulum (ER) 2 to the Golgi apparatus in transport vesicles. Incorporation of the glycoproteins (cargo) into the transport vesicles is thought to be mediated by transmembrane cargo receptors, which have been identified as intracellular lectins (non-enzymatic carbohydrate-binding proteins) (5,6).
In mammalian cells, the canine vesicular integral membrane protein of 36 kDa (VIP36) and human ER-Golgi intermediate compartment (ERGIC)-53 have been identified as cargo receptors in the Golgi apparatus and ERGIC, respectively (5,6). The cargo receptors are type I membrane proteins, which have lumenal, transmembrane, and cytoplasmic domains. The N-terminal lumenal domains of VIP36 and ERGIC-53 share homology with L (leguminous)-type lectins and are thus called carbohydrate recognition domains (CRD). We previously reported that VIP36 has high avidity for high mannose-type glycans containing Man␣132Man␣132Man residues in Man 7-9 GlcNAc 2 ⅐Asnpeptides (7). Recently, Kamiya et al. (8) reported details of the carbohydrate recognition mechanism of VIP36 using NMR (8). These observations have implicated that VIP36 is a cargo receptor involved in the intracellular transport of glycoproteins carrying high mannose-type glycans. On the other hand, a chemical cross-link experiment has shown co-isolation of ERGIC-53 with soluble cathepsin-Z-related glycoproteins in a Ca 2ϩ -dependent manner (9). They form complexes in the ER, and dissociate in the ERGIC fraction. Although the physiological functions of the cathepsin-Z-related protein remain unclear, this finding strongly supports the presence of soluble glycoprotein cargo receptors at the ER exit sites. Genetic studies of ERGIC-53 have shown that its mutations lead to a bleeding disorder known as a combined deficiency of coagulation factors V and VIII in circulation (10,11). In addition, it has been found that a second protein involved in this disease, MCFD2, an EF-hand protein, is co-purified with ERGIC-53 in a Ca 2ϩ -dependent manner (12). The crystal structures of the CRD of p58, a rat ortholog of ERGIC-53, in the absence and presence of Ca 2ϩ ions were recently determined, confirming its structural similarity to the L-type lectins (13,14). It was shown that the putative ligandbinding site of p58/ERGIC-53 is similar to the mannose-binding site of the L-type lectins. However, no complex structures of ligand/receptor have been solved. Thus the structural basis for N-linked glycoprotein transport by cargo receptors still remains largely unknown.
Emp47p and Emp46p, yeast orthologs of ERGIC-53, have been proposed as cargo receptors between the ER and the Golgi apparatus in Saccharomyces cerevisiae (15,16). Emp47p is a receptor for Emp46p responsible for the selective transport of Emp46p from the ER to the Golgi apparatus by forming hetero-oligomerization between the two proteins (17). The lumenal domain of Emp47p consists of a CRD with homology with ERGIC-53 and VIP36, and a putative coiled-coil domain, which is responsible for the formation of a homo-oligomeric complex of itself and a hetero-oligomeric complex with Emp46p in the ER. As for Emp46p, the lumenal domain consists of a CRD and a putative coiled-coil domain, which is required for forming the complex with Emp47p. On the other hand, the C-terminal cytoplasmic regions of Emp46p and Emp47p both contain binding sites for coat protein complexes I and II (COPI and COPII), which are required for cycling between the ER and Golgi apparatus. Thus, Emp47p and Emp46p apparently play important roles in selective packaging of specific glycoproteins into ER-derived vesicles. Indeed, gene disruption of both EMP47 and EMP46 leads to a marked defect in the secretion of a subset of glycoproteins (16). Unfortunately, however, specific cargo proteins for Emp47p and Emp46p have not been identified so far. To investigate the structural basis for the glycoprotein transport by Emp46p and Emp47p, we determined crystal structures of their CRDs and studied their binding affinities against high mannose-type glycoproteins using surface plasmon resonance (SPR).

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The DNA fragments for residues 6 -229 of CRD of Emp46p, 7-227 of CRD of Emp47p (⌬1-Emp47p CRD), and 1-254 of CRD of Emp47p of S. cerevisiae were cloned into the BamHI and SmaI sites of the pGEX4T-1 plasmid (Amersham Biosciences). The native proteins of Emp46p, ⌬1-Emp47p, and Emp47p CRDs were expressed in Escherichia coli BL21(DE3) cells. Cells were harvested after induction with 0.1 mM isopropyl ␤-D-thiogalactoside (Wako) for 6 h at 20°C, and lysed by sonication in phosphate-buffered saline buffer. The cell lysate was loaded on a glutathione-Sepharose 4B column (Amersham Biosciences). The glutathione S-transferase (GST) fusion proteins were eluted by glutathione (Wako) and cleaved by thrombin protease (Amersham Biosciences). The cleaved proteins were passed through a glutathione-Sepharose 4B to remove GST protein, and further purified by benzamidine-Sepharose 4FF column (Amersham Biosciences) to remove thrombin protease. Emp46p and ⌬1-Emp47p CRDs were purified by Superdex 75 gel filtration column (Amersham Biosciences) chromatography. Emp47p CRD was purified by Mono Q column (Amersham Biosciences) chromatography. The selenomethionine (SeMet)-substituted proteins of Emp46p and Emp47p CRDs were expressed in the E. coli DL41 cell. The SeMet-substituted proteins were purified by the same procedure as native proteins. The purified proteins were dialyzed against 10 mM Tris-HCl (pH 7.5).
Crystallization and X-ray Data Collection-All of the crystallizations were carried out by the sitting-drop vapor diffusion method. Crystals of native and SeMet-substituted Emp46p CRD in its K ϩ -bound forms were obtained in a buffer containing 8 mg ml Ϫ1 protein, 20% (w/v) PEG3350, 0.3 M KF, 0.1 M HEPES-Na (pH 7.5), and 10% (v/v) ethylene glycol, with incubation at 289 K for 4 days. On the other hand, the crystal of the metal-free Emp46p CRD was obtained in a buffer containing 8 mg ml Ϫ1 protein, 22% (w/v) PEG1000, 0.1 M HEPES-Na (pH 7.5), and 10% (v/v) ethylene glycol, with incubation at 289 K for 4 days. These native and SeMet-substituted proteins in the absence and presence of the K ϩ ion were crystallized in space group P2 1 with two molecules per asymmetric unit. All data sets of Emp46p CRD were collected under cryogenic conditions with crystals soaked in the reservoir solution.
As for Emp47p, crystals of native and SeMet-substituted Emp47p CRD in its metal-free forms were obtained in a buffer containing 28 mg ml Ϫ1 protein, 1.2 M NaH 2 PO 4 , 0.8 M K 2 HPO 4 , 0.2 M Li 2 SO 4 , and 0.1 M CAPS (pH 10.5): with a final pH 6.1, with incubation at 277 K for 1 week. Data sets of native and SeMet-substituted Emp47p CRD were collected under cryogenic conditions with crystals soaked in the cryoprotectant buffer containing 20% (v/v) glycerol. Both native and SeMet-substituted proteins were crystallized in space group P4 3 2 1 2 with one molecule per asymmetric unit (form 1).
To obtain the metal-bound form of Emp47p CRD, we further constructed and crystallized a ⌬1-Emp47p CRD. This construct was designed from a structured (visible) region of the form 1 Emp47p CRD structure. Crystals of ⌬1-Emp47p CRD were obtained in the following conditions (forms 2-4). Nevertheless, the protein was crystallized as metal-free forms in these conditions. The form 2 crystal was obtained in a buffer containing 3.5 mg ml Ϫ1 protein, 20% (w/v) PEG3350, and 0.2 M NH 4 Cl, with incubation at 283 K for 3 days. The crystal belongs to space group C2 with one molecule per asymmetric unit. The form 3 crystal was obtained in a buffer containing 3.5 mg ml Ϫ1 protein, 20% (w/v) PEG3350, and 0.2 M CH 3 COOK, with incubation at 283 K for 3 days. The crystal belongs to space group P2 1 with two molecules per asymmetric unit. The form 4 crystal was obtained in a buffer containing 3.5 mg ml Ϫ1 protein, 20% (w/v) PEG4000, 0.2 M CH 3 COONH 4 , 50 mM sodium cacodylate (pH 7.0), and 10 mM CaCl 2 , with incubation at 283 K for 3 days. The crystal belongs to space group P2 1 2 1 2 1 with four molecules per asymmetric unit. All data sets of ⌬1-Emp47p CRD were collected under cryogenic conditions with crystals soaked in the cryoprotectant buffer containing 15% (v/v) ethylene glycol. The diffraction data were processed using HKL2000 (18). The native crystal parameters of Emp46p and (⌬1)-Emp47p CRDs are shown in Tables 1 and 2, respectively. The SeMet-substituted crystal parameters of Emp46p and (⌬1)-Emp47p CRDs are shown in supplementary Tables 1 and 2, respectively.
Structure Determination and Refinement-The crystal structures of the K ϩ -bound Emp46p CRD and form 1 Emp47p CRD were solved by the MAD method using crystals of the SeMet-substituted proteins. The initial phases were determined with SOLVE (19) and improved with RESOLVE (20). As the electron density maps with the experimental phases were very clear, most residues in the structures could be traced automatically using RESOLVE. The initial molecular models were built automatically using ARP/wARP (21). The crystal structures of metalfree Emp46p CRD and forms 2-4 of ⌬1-Emp47p CRD were solved by the molecular replacement method using the program MOLREP (22) with K ϩ -bound Emp46p CRD and form 1 Emp47p CRD as search models, respectively. The refinement procedures were carried out with CNS (23), REFMAC5 (24), and SHELX97 (25). Further model fitting to the electron density maps was performed manually using O (26) and Turbo-FRODO (27). Stereochemical qualities of the final models were assessed by PROCHECK (28). Phasing statistics of Emp46p CRD and Emp47p CRDs are summarized in supplementary Tables 1 and 2, respectively. Final refinement statistics of Emp46p CRD and (⌬1)-Emp47p CRDs are summarized in Tables 1 and 2, respectively. Figures were prepared using the GRASP (29) and PyMOL (30).
Mutational Experiments-The mutation (Y131F) was introduced to the Emp46 gene by the PCR method. The amplified DNA fragment with the mutation was cloned into the BamHI and BglII sites of wild-type Emp46p/pGEX4T-1 plasmid. The overexpressed mutant protein was purified and crystallized according to the method used for the wild type. The crystal parameters of Y131F-Emp46p are shown in Table 1. The plasmid construction for yeast overexpression of the mutant and the cell manipulation were carried out as previously described (16).

RESULTS
Structures of Metal-free and K ϩ -bound Emp46p CRD-Following the conventional notion of Ca 2ϩ dependence of its mammalian orthologs ERGIC-53 and VIP36, we first tried to determine crystal structures of Emp46p CRD in the presence or absence of Ca 2ϩ ion. Despite numerous trials to form Ca 2ϩ -bound Emp46p crystals, for example, crystallization in the presence of 10 mM CaCl 2 , Ca 2ϩ was never found in the crystals. Instead, to our surprise, we found that Emp46p CRD bind another metal ion K ϩ bound in the crystal structures. Initially, the potassium ion was thought originat from the crystallization condition (0.3 M KF), but, as will be described below, we later confirmed that Emp46p is in the K ϩ -bound form in the physiological conditions. The crystal structure of K ϩ -bound Emp46p CRD was determined using MAD phasing at 1.8-Å (supplementary Table 1). As for metal-free Emp46p, we first prepared a metal-free protein sample whose metal compositions were confirmed by atomic absorption spectroscopy (data not shown). We used this solution for crystallization in the buffer that does not contain any metal ions. The structure of metal-free Emp46p CRD was solved by the molecular replacement method using the K ϩ -bound Emp46p as a search model. The final model of K ϩ -bound Emp46p refined to 20.0 -1.52-Å resolution has an R-factor of 18.9% and an R free of 21.8%. On the other hand, the final model of metal-free Emp46p refined to 20.0 -1.75-Å resolution has an R-factor of 21.0% and an R free of 23.7%. Final refinement statistics of the K ϩ -bound and metal-free Emp46p are summarized in Table 1. In both cases, two molecules (A and B) of Emp46p were contained in an asymmetric unit. Emp46p, however, was eluted at positions corresponding to a monomer in gel filtration chromatography (data not shown), and the contact area of the two molecules is very small (319 Å 2 ). Thus we concluded that the crystallographic dimer of Emp46p is nonphysiological. In the K ϩ -bound Emp46p, two K ϩ ions were observed: one is located at the edge of a concave ␤-sheet of molecule A (the first site), and the other in the dimer interface (the second site). The K ϩ ion in the second site is considered non-physiological because the K ϩ ion was replaced by a water molecule (data not shown) in other crystallization conditions with lower K ϩ ion concentrations, 100 -150 mM, which correspond to physiological intracellular conditions (31, 32). Therefore, the structure of molecule A is described hereafter as the K ϩ -bound form.
Residues Cys 150 (strand ␤10) and Cys 184 (strand ␤13) form a disulfide bond. A structure-based sequence alignment of Emp46p CRD with other CRDs, rat p58/ERGIC-53, canine VIP36 and Emp47p, along with the secondary structure assignments shown in Fig. 2. K ϩ Ion Binding Site and Its Structural Changes in Emp46p CRD-The F o Ϫ F c electron density map of Emp46p CRD shows one prominent peak between two loops that are termed Loop 1 (between ␤8 and ␤9) and Loop 2 (between ␤9 and ␤10). Moreover, at ϭ 2.0 Å, a prominent anomalous signal (ϳ20 ) was observed at the corresponding site (Fig. 4A), indicating that it is identified as a metalbinding site. When the structure was refined with the metal atom as either Ca 2ϩ or K ϩ , the B values of the metal after refinement ( . Metal binding in this position has never been observed in L-type lectins (14,33,34). The metal-binding site of Emp46p was formed by a negatively charged pocket (Fig. 3, A and B), which is favorable for cation binding. The metal ion has a distorted octahedral coordination, with a side chain oxygen atom of Tyr 131 , main chain carbonyl oxygen atoms of Ile 141 , Glu 142 , and Ser 144 , and two water molecules (Fig. 4A). The distances between metal ion and the coordinating residues and waters (Tyr 131 , Ile 141 , Glu 142 , Ser 144 , Wat 1 , and Wat 2 ) are 2.9, 2.6, 2.9, 2.6, 2.8, and 2.9 Å, respectively. Distances for the coordination of Ca 2ϩ and K ϩ ions are known to be ϳ2.4 and 2.8 Å, respectively (35,36). We demonstrated that two prominent peaks were not observed in the F o Ϫ F c electron density map when the metal-free crystal was soaked with 1-10 mM Ca 2ϩ ion, whereas the peaks were observed when the metal-free crystal was soaked with 25-150 mM K ϩ ion. Taken together, we conclude that the metal ion found in the Emp46p crystal is K ϩ ion. Ribbon models of the CRD of Emp46p monomer are shown in A and B, which is rotated by 90°around a vertical axis. The overall structure of the metal-free form of Emp46p is quite similar to that of the K ϩ -bound form (r.m.s. deviation of 0.3 Å for the C␣ atoms). Ribbon models of the CRD of the Emp47p monomer are shown in C and D as in Emp46p. Positions of the N and C termini are indicated by red letters. The secondary structures are highlighted (␤-strands belonging to the concave ␤-sheets, yellow; ␤-strands belonging to convex ␤-sheets, blue; ␤-strands belonging to ␤-hairpin, cyan; helices, red) and the loops are colored green. The bound potassium ion is shown as a magenta sphere.
Upon K ϩ ion binding, small conformational changes take place around Loops 1 and 2 (Fig. 4B), and the electron density of the Loop 2 region was improved. Movements of the K ϩ ion coordinating atoms were observed for the side chain oxygen atom of Tyr 131 , and main chain oxygen atoms of Ile 141 , Glu 142 , and Ser 144 . The distances were 0.5, 0.7, 1.0, and 0.4 Å, respectively. In Loop 1, Asp 121 and Asp 122 shift by 0.6 and 1.1 Å for the C␣ atoms, respectively, suggesting that the side chain oxygen atoms are attracted by K ϩ ion. As a result of the conformational changes, the two loops shift toward each other. Similar but more pronounced structural changes of Loops 1 and 2 upon metal binding are also observed in concanavalin A (37) and p58/ERGIC-53 (14).
To investigate biological relevance of the K ϩ ion, we introduced a point mutation (Y131F) into a K ϩ -coordinating residue. First, we deter-mined the crystal structure of the Y131F mutant, and studied its K ϩ ion-binding site. Mutation of Tyr 131 to Phe prevents K ϩ from binding even at 300 mM K ϩ concentration, and leads to conformational changes around Loops 1 and 2 (Fig. 4C). Unexpectedly, Loops 1 and 2 of the Y131F mutant are very different from those of the metal-free form. The loss of the hydroxyl oxygen atom of Tyr 131 may perturb some hydrogen network in the Loop 1 region. In Loop 2, large movements were observed: the Ser 144 C␣ atom moves by 5.1 Å as compared with the K ϩ -bound form. In addition, the side chain of Glu 142 moves toward Loop 1 with a shift of 4.3 Å for the C␦ atom. As a consequence, the movement reverses the side chain of Asp 122 . A comparison of B-factor values of the three Emp46p structures reveals that K ϩ confers a rigid structure in the Loop 2 region of Emp46p (data not shown). From these results, we conclude that the function of the K ϩ ion is to stabilize Loops 1 and 2.
Next, we constructed a gene disrupted strain of emp46 and emp47 and observed its phenotype. As we previously reported, the disruptants of emp47⌬ and emp47⌬emp46⌬ showed a severe growth defect at a restrictive temperature (37°C), and the defect of both strains was rescued by an overexpression of Emp46p (16). In contrast, the rescue ability was drastically reduced by introduction of the Y131F mutation to Emp46p (Fig. 5A). It suggests that the K ϩ ion binding of overexpressed Emp46p is required for viability of the emp47⌬emp46⌬ cell at 37°C.
High-mannose Glycoprotein Binding of Emp46p CRD-SPR measurements were used to investigate whether Emp46p CRD binds glycoproteins with various characteristic sugar chain structures. We observed binding of Emp46p to porcine thyroglobulin (Fig. 6A), which has high mannose-type sugar chains containing mostly Man 7-9 GlcNAc 2 (38). There was weak interaction between Emp46p and ovalbumin, whereas the binding to ribonuclease B, transferrin, asialofetuin, and ␣1-acid glycoprotein was not detected (data not shown). Ribonuclease B and ovalbumin also contain one high mannose-type but a smaller sugar chain per molecule, Man 5 GlcNAc 2 and Man 6 -7 GlcNAc 2 , respectively (39, 40). On the other hand, transferrin, FIGURE 2. Alignment of amino acid sequences of Emp46p, Emp47p, p58/ERGIC-53, and VIP36. Red, green, and orange letters indicate conserved amino acid residues, K ϩ -coordinating residues, and Ca 2ϩ -coordinating residues, respectively. Purple boxes indicate p58/ERGIC-53 residues involved in mannose binding. Secondary structures of Emp46p and Emp47p are shown above and below the amino acid sequences, respectively, and are colored as described in the legend to Fig. 1. D). The surface models of Emp46p and Emp47p are shown in the same orientations as in Fig. 1 and colored according to the electrostatic surface potential (blue, positive; red, negative; scale from Ϫ10 to ϩ10 kT/e).

FIGURE 4. K ؉ ion binding site and its conformational changes upon K ؉ ion binding of Emp46p.
A, K ϩ ion binding site of Emp46p. Residues coordinating the K ϩ ion are shown in balland-stick models. The magenta sphere indicates the K ϩ ion. Water molecules are shown as W1 and W2. Pink spheres indicate Ca 2ϩ ions at the Ca1 and Ca2 sites in p58/ERGIC-53 (14). An anomalous Fourier map (blue mesh) is contoured at 15 . B, comparison between the metal-free and K ϩ -bound Emp46p structures. Residues coordinating the K ϩ and those with significant conformational changes are shown in ball-and-stick models. The metal-free and K ϩ -bound structures are colored in cyan and yellow, respectively. C, comparison between the metal-free and Y131F Emp46p structures. The metal-free and Y131F structures are colored in cyan and green, respectively. asialofetuin, and ␣1-acid glycoprotein have bi-to tetra-antennary complex-type glycans (41)(42)(43). In addition, the total amount of binding to the glycoproteins carrying high mannose-type glycans decreased upon removal of sugar chains by Endo H, indicating that Emp46p indeed recognizes sugar portions of the glycoproteins (Fig. 6A). The remaining binding to the Endo H-treated thyroglobulin might come from some residual sugar portions such as complex-type and/or high mannosetype glycans.
Interestingly, the thyroglobulin affinities of Emp46p were not affected by metal ions including Ca 2ϩ , Mg 2ϩ , Mn 2ϩ , and K ϩ ions (data not shown), despite the existence of the K ϩ ion-binding site. In addition, the Y131F mutation of Emp46p, which prevents K ϩ binding, did not show any effect with glycan binding affinities (Fig. 6B). These results suggest that the K ϩ ion was not involved in the binding ability for the glycoprotein.
Structure of Emp47p CRD-Crystals of the CRD of Emp47p were obtained in various crystallization conditions containing Ca 2ϩ or K ϩ ions (crystal forms 1-4). However, no electron density was observed for the metal ions in any of these crystal forms. Besides, a significant negatively charged pocket was not observed on the molecular surface (Fig. 3,  C and D). The crystal structure of form 1 Emp47p was determined using the MAD phasing at 2.0 Å (supplementary Table 2). The structures of forms 2-4 Emp47p were solved by the molecular replacement method using the structure of form 1 Emp47p as a search model. Final refinement statistics of crystal forms 1-4 Emp47p are summarized in Table 2. The final model of form 1 Emp47p contains residues 7-227 and 244 -249. Residues 228 -243 and a few N-and C-terminal residues were not visible in the electron density map. On the other hand, the final models of forms 2 and 3 Emp47p contain residues 7-227, namely no disordered residues in these crystal forms. As for form 4, there are some disordered regions at the N and C termini. The Emp47p crystals in forms 1 and 2 contain one molecule per asymmetric unit, whereas those in forms 3 and 4 contain two and four molecules, respectively. The dimer interfaces of forms 3 and 4 are different from each other, although the contact areas are comparable: 517 and 559 Å 2 , respectively. Furthermore, the Emp47p was eluted at positions corresponding to a monomer in gel filtration chromatography (data not shown). From the results of the gel filtration analysis, the disagreement of dimer interfaces, and the small contact areas, we concluded that the crystallographic dimer of Emp47p CRD is non-physiological. The crystal structures of forms 1-4 are quite similar to each other with r.m.s. deviation values of 0.17-0.57 Å for all C␣ atoms of Lys 11 -Val 227 residues, suggesting that there are no significant structural differences among them. Therefore, the structure of form 2 Emp47p, which has the highest resolution, the lowest R-factor, and no disordered region, is described hereafter.
The CRD of Emp47p has a globular shape, very similar to the CRD of Emp46p, with a ␤-sandwich of two antiparallel ␤-sheets, a small ␣ helix, and one turn of a 3 10 helix (Fig. 1, C and D). There are slight discrepancies compared with the Emp46p CRD: the crevice is formed by a 17-residue (as opposed to 18 in Emp46p) loop between ␤7 and ␤8, and a 15-residue (as opposed to 13) loop between ␤9 and ␤10. Like in the Emp46p CRD, Emp47p CRD forms a disulfide bond between residues Cys 151 (strand ␤10) and Cys 185 (strand ␤13).
Most L-type lectins (13,33,34) have a core structure composed of a ␤-sandwich with a seven-stranded concave ␤-sheet and a six-stranded convex ␤-sheet. In contrast, the convex ␤-sheets of Emp46p and Emp47p are composed of five ␤-strands. The N terminus of p58/ER-GIC-53 starts with two short ␤-strands (␤1a and ␤1b) separated by a 3 10 helix turn. On the other hand, the N termini of the Emp46p and Emp47p simply start with a 3 10 helix. The sequences of the N-terminal regions at the corresponding sites of ␤1a and ␤1b are not conserved. Likewise, the Emp46p and Emp47p have no peptide bonds observed in the cis-conformation, whereas p58/ERGIC-53 has three cis-peptide bonds: (i) between residues Gly 62 and Pro 63 at the end of ␤1b, (ii) between Ala 128 and Asp 129 at the entrance of ␤7, and (iii) between Asn 170 and Pro 171 before ␤9 (13,14). In addition, the proline residues are conserved in all sequences of the mammalian ERGIC-53-and VIP36-like families. Furthermore, there is an insertion loop (9 residues) between ␤12 and ␤13 in the p58/ERGIC-53.

DISCUSSION
Comparison of the Metal-binding Site and the Potential Ligand-binding Site of Emp46p with Those of Other L-type Lectins-It is known that leguminous lectins have Ca 2ϩ and Mn 2ϩ ions in their ␤-sandwich structures (34). The Ca 2ϩ and Mn 2ϩ ions of the leguminous lectins interact with oligosaccharide ligands indirectly; their divalent cations are used to stabilize the binding site and fix the positions of amino acids that interact with oligosaccharide ligands, whereas the Ca 2ϩ ion of C-type (Ca 2ϩdependent) lectins forms direct coordination with oligosaccharide ligands (44). Velloso et al. (14) showed that calcium-dependent p58/ ERGIC-53 contains two Ca 2ϩ ions termed Ca1 and Ca2 (Fig. 4A), and that Ca2 is exactly at the Ca 2ϩ site of the leguminous lectins, whereas Ca1 does not lie at the corresponding Mn 2ϩ site. However, the electron density maps of Emp46p and Emp47p show no peak that can be FIGURE 5. Phenotype of Y131F-Emp46p. A, isogenic wild-type (YPH500) and emp47⌬emp46⌬ (KSY008) cells transformed with a multicopy plasmid, a multicopy plasmid with EMP46, or a multicopy plasmid with EMP46-Y131F were grown at 37°C. B, the amount of the expressed protein in each cell was estimated by Western blotting. The expression amount of the Y131F mutant is almost identical to that of the wild type.
assigned as the Ca 2ϩ ion at the corresponding site, nor anywhere else. Thus Emp46p and Emp47p are the first examples of the L-type lectin family that has no Ca 2ϩ ion. Loop 1 of p58/ERGIC-53 is 4 residues longer than those of Emp46p and Emp47p. Likewise, the residues coordinating the Ca 2ϩ ion in p58/ERGIC-53 (Asp 160 , Phe 162 , Asp 163 , Asn 164 , Asp 165 , Asn 169 , Asn 170 , and Asp 189 ; shown in orange in Fig. 2) are poorly conserved in Emp46p and Emp47p. In addition, the position of the ␣ helix in Loop 2 of p58/ERGIC-53 is different from that of Emp47p. The ␣ helix in the p58/ERGIC-53 contains two key residues: His 186 , which is thought to be a pH/Ca 2ϩ sensor in the ligand binding (45), and Asp 189 , which is one of the Ca 2ϩ coordinating residues (14). As for Emp46p, there is no such helix in Loop 2, and hence no binding of Ca 2ϩ .
Strikingly, Emp46p contains a monovalent cation, K ϩ ion, between Loops 1 and 2 (Fig. 4, A and B). When coordinates of Emp46p and p58/ERGIC-53 are superimposed, the K ϩ ion of Emp46p is 7.5 and 8.2 Å away from the two Ca 2ϩ ions (Ca1 and Ca2) of the p58/ERGIC-53, respectively. The K ϩ ion is mainly coordinated by residues of Loop 2 of Emp46p, whereas the divalent cations mainly coordinate to those of Loop 1 in p58/ERGIC-53. The Loop 2 of Emp46p is 2-3 residues shorter than those of Emp47p and p58/ERGIC-53. Moreover, the sequences of K ϩ ion coordinating residues (Tyr 131 , Ile 141 , Glu 142 , and Ser 144 ; letters in green in Fig. 2) are not conserved in p58/ERGIC-53 and Emp47p. To the best of our knowledge, structures of lectins complexed with the K ϩ ion have never been reported. Thus the structure of the K ϩ -bound Emp46p is the first example in lectin families, although there is an established theory in sugar recognition; lectin affinity is modulated by divalent cat-ions but not monovalent cations (44). We therefore considered a possibility where the K ϩ ion functions as a stabilizing molecule of the ligandbinding site as in the case of divalent cations, which indirectly interacts with cognate oligosaccharides.
On the basis of structural similarities to other L-type lectins and SPR experiments, we have supposed that Emp46p as well as p58/ERGIC-53 (14) recognize Man 8 GlcNAc 2 moieties of glycoproteins. Mutagenesis studies have implicated four residues, Asp 129 , Asp 160 , Asn 164 , and His 186 , to be required for binding of p58/ERGIC-53 to mannose beads (6,45). In addition, Velloso et al. (13) speculated that Gly 259 and Gly 260 as well as Asp 129 , Asn 164 , and His 186 of p58/ERGIC-53 interact with sugar residues on the basis of the structural similarity to the isolectin-1⅐mannose complex (33). The peptide bond between Ala 128 and Asp 129 is in a cis-conformation in p58/ERGIC-53, as in the leguminous lectins. The cis-conformation is crucial for the correct geometry of the Ca 2ϩbinding site and for sugar binding in the leguminous lectins (33). However, these residues are poorly conserved in Emp46p and Emp47p (Fig.  2). Besides, the Emp46p and Emp47p have no peptide bonds observed in the cis-conformation. Combined with the SPR data, our results strongly suggest that the Ca 2ϩ -independent carbohydrate recognition by Emp46p are quite different from that of the Ca 2ϩ -dependent p58/ERGIC-53.
Ca 2ϩ -independent Glycoprotein Binding of Emp46p-SPR experiments revealed that CRDs of Emp46p bind some glycoproteins, which have high mannose-type oligosaccharide chains, and recognize largersized oligosaccharide chains (Man 7-9 GlcNAc 2 ). In the ER exit site, it has been thought that ERGIC-53 sorts glycoproteins containing Man 8 GlcNAc 2 , which are properly folded and processed through the ER-associated degradation system (2)(3)(4). Our results suggest that Emp46p also functions as sorting receptors for glycoproteins at the ER exit site. However, we have not been successful in determining precise carbohydrate binding specificities of Emp46p CRD because of the very weak binding affinities of the CRDs (Fig. 6). In addition, no interactions between Emp47p CRD and the glycoproteins were observed by our SPR measurements (data not shown). Recently, we have shown that Emp47p oligomerizes through the coiled-coil domain for the selective transport of Emp46p and its own exit from the ER (17). In this study, we used monomeric Emp46p and Emp47p CRDs for the SPR measurement because the lumenal domains of the homo-oligomeric Emp47p and hetero-oligomeric Emp46/47p complexes could not be purified so far because of degradation problems. Many lectins, such as mannose-binding protein and asialoglycoprotein receptor, achieve higher affinity and selectivity through oligomerization of their CRDs (44). Elucidation of the molecular details of the hetero-oligomeric Emp46/47p complex with high mannose-type glycoproteins and glycoconjugates will provide further insight into the role of these proteins as cargo receptors.
To investigate whether Ca 2ϩ and K ϩ ions are involved in carbohydrate recognition of Emp46p, we carried out SPR experiments in the presence and absence of Ca 2ϩ and K ϩ ions. Calnexin, calreticulin, and L-type lectins are known to be all dependent on Ca 2ϩ ion for their CRD functions (6,8,44,46). The SPR observation that the binding of Emp46p is not enhanced by the Ca 2ϩ ion is consistent with the crystallographic results on the Ca 2ϩ unbinding properties. Intracellular organelles have characteristic lumenal pH values suitable for their biochemical function. Organelles of the secretory and endocytic pathways encounter a gradient of decreasing pH; pH values of the ER and the Golgi apparatus are ϳ7.2 and 6.4, respectively, in typical mammalian cells (32). Recently, Appenzeller-Herzog et al. (45) proposed a glycoprotein traffic model between the ER and the ERGIC; a pH-induced loss of Ca 2ϩ in ERGIC-53 triggers glycoprotein cargo release. On the other hand, we have shown that hetero-oligomerization of Emp46p and Emp47p occurs in the ER, and their dissociation occurs in the Golgi apparatus (17). Therefore, in this case the dissociation of the complex between Emp46p and Emp47p, but not the Ca 2ϩ ion loss, might trigger glycoprotein cargo release in the Golgi apparatus. We showed that K ϩ ion is not required for lectin activity of Emp46p using the binding assay of the wild-type and Y131F-Emp46p in the presence and absence of the K ϩ ion (Fig. 4). In conclusion, we postulate Emp46p as a Ca 2ϩ -independent intracellular lectin.
To investigate whether the binding and release of the K ϩ ion from Emp46p occurs either in the ER or Golgi, we carried out soaking experiments using the metal-free Emp46p crystals with various pH (6.6, 7.0, and 7.4) and [K ϩ ] (25, 75, 100, and 150 mM conditions). Although the K ϩ ion concentration ([K ϩ ]) in the ER has not been reported yet, those of the cytoplasm and Golgi apparatus are reported to be 140 and 107 mM, respectively (31, 32). The results showed that there was little difference in the K ϩ -binding site structure regardless of the variation of pH and [K ϩ ] (data not shown), which suggests that Emp46p always binds the K ϩ ion in vivo. We have suggested that the K ϩ ion contributes to stabilization of Loop 1 and 2 regions that appear unrelated to sugar recognition (Figs. 4 and 6). It might be possible that the K ϩ ion is required for recognition of hitherto unidentified cargo proteins by the hetero-oligomeric complex of full-length Emp46p and Emp47p (Fig. 5).
In summary, we determined the first crystal structures of CRD of the Ca 2ϩ -independent K ϩ -bound Emp46p and metal-unbound Emp47p. Furthermore, we showed that the CRD of Emp46p binds to glycoproteins carrying high mannose-type glycans. Structure determination of the hetero-oligomeric complex of Emp46p and Emp47p with high mannose-type glycoproteins and/or glycoconjugates will provide new Ca 2ϩindependent carbohydrate recognition modes in N-linked glycoprotein traffic by intracellular lectins.