Structure of the Lectin Mannose 6-Phosphate Receptor Homology (MRH) Domain of Glucosidase II, an Enzyme That Regulates Glycoprotein Folding Quality Control in the Endoplasmic Reticulum*

Background: Glucosidase II is an endoplasmic reticulum enzyme involved in quality control of glycoprotein folding. Results: The structure of the lectin domain of GII was determined by NMR spectroscopy. Conclusion: GII lectin domain structure contains a unique Trp residue critical for GII activity. Significance: GIIβ MRH domain structure is the first determined of an MRH domain present in a protein with enzymatic activity. Here we report for the first time the three-dimensional structure of a mannose 6-phosphate receptor homology (MRH) domain present in a protein with enzymatic activity, glucosidase II (GII). GII is involved in glycoprotein folding in the endoplasmic reticulum. GII removes the two innermost glucose residues from the Glc3Man9GlcNAc2 transferred to nascent proteins and the glucose added by UDP-Glc:glycoprotein glucosyltransferase. GII is composed of a catalytic GIIα subunit and a regulatory GIIβ subunit. GIIβ participates in the endoplasmic reticulum localization of GIIα and mediates in vivo enhancement of N-glycan trimming by GII through its C-terminal MRH domain. We determined the structure of a functional GIIβ MRH domain by NMR spectroscopy. It adopts a β-barrel fold similar to that of other MRH domains, but its binding pocket is the most shallow known to date as it accommodates a single mannose residue. In addition, we identified a conserved residue outside the binding pocket (Trp-409) present in GIIβ but not in other MRHs that influences GII glucose trimming activity.

Mannose 6-phosphate receptor homology (MRH) 3 domains were first described in the two receptors responsible for delivering lysosomal enzymes from the trans Golgi network to endosomal compartments in mammalian cells. In these mannose 6-phosphate receptors (MPRs), the MRH domains require the presence of mannose-linked phosphate groups in N-glycans for binding, a structural feature not required by MRH domains present in other proteins of the secretory pathway (1). Of the known MRH domain-containing proteins, three are endoplasmic reticulum (ER)-resident proteins, OS-9 (Yos9p is the yeast homolog of mammalian OS-9), Erlectin (also called XTP3-B), and glucosidase II (GII) (2).
A quality control mechanism reliant on N-glycan structures exists in the ER and ensures the correct folding of glycoproteins within the secretory pathway of eukaryotic cells (see Fig. 1A) (3,4). OS-9 and Erlectin are apparently involved in driving misfolded glycoproteins to proteasomal degradation. The N-glycan structural requirements for OS-9/Yos9p binding have been studied in detail. Optimal recognition occurs in glycans lacking residues i and k (see Fig. 1B), thus exposing an ␣(1,6)-linked mannose unit (residue j) (5)(6)(7). ER mannose removal (first residue i followed by k) from the transferred glycan (Glc 3 Man 9 GlcNAc 2 ) is a relatively slow process compared with deglucosylation and thus occurs mainly in slow folding/misfolded glycoproteins. Thus, the N-glycan binding specificity of OS-9/Yos9p is particularly suited for the known function of this protein. The binding specificity of the C-terminal MRH domain of Erlectin for Man 9 has recently been reported, and it was proposed to serve as a negative regulator of ER-associated degradation (8). In contrast to these two proteins, GII is an early player in N-glycan processing as it removes the two inner glucoses from all transferred glycans (residues l and m; see Fig. 1B) irrespective of the folding status of the protein. Removal of the middle glucose (residue m) creates a monoglucosylated epitope that is recognized by two unconventional chaperones, the lectins calnexin and calreticulin. Their interactions with glycoproteins enhance the folding efficiency of nascent polypeptides and prevent the exit of folding intermediates and misfolded glycoproteins from the Golgi. Monoglucosylated N-glycans may be also formed by the action of UDP-Glc:glycoprotein glucosyltransferase (UGGT) on fully deglucosylated glycans linked to protein moieties not displaying their native structures (see Fig.  1A).
GII is a heterodimer composed of a catalytic GII␣ subunit and a regulatory GII␤ subunit that contains an MRH domain and in most species a KDEL-like ER retrieval sequence at its C terminus (9, 10) (see Fig. 2A). Contrary to OS-9/Yos9p, maximal N-glycan binding by GII␤ MRH domain requires the presence of mannose residue k of arm C (Fig. 1B). Furthermore, in vivo and cell free assays show that a decrease in N-glycan mannose content results in lower GII enzymatic activity (11)(12)(13)(14)(15). Not only does optimal GII deglucosylation activity require the presence of the full complement of mannose units on nascent glycoproteins but also the four conserved amino acids of the GII␤ subunit that are known to be essential for the ability of the MRH domains of MPRs to bind mannose 6-phosphate (Man-6-P)-containing lysosomal enzymes (15,16) (Fig. 2B).
Why does GII differ from most other glycosidases for which their maximal activity does not require the presence of a lectinlike domain? We have shown previously that a decrease in N-glycan mannose content sharply decreases in vivo GII activity without affecting the activity of UGGT, thereby prolonging the half-life of monoglucosylated glycans produced by UGGT (15). We speculate that ER mannosidase-mediated removal of mannose units would result in an increased interaction of slow folding/misfolded glycoproteins with the lectin chaperones. This in turn would increase the chances for proper folding of glycoproteins displaying an arduous folding process and would result in a more stringent prevention of a surreptitious exit of misfolded glycoproteins to the Golgi. Therefore, the presence of the MRH domain converts GII into a key regulator of the quality control mechanism of glycoprotein folding. As the structure of OS-9 MRH domain was recently determined (7), the significant difference between its N-glycan binding specificity and that of GII prompted us to determine the structure of GII␤ MRH domain. Here we report that GII␤ MRH domain folds into a flattened ␤-barrel similar to that of the MPRs and OS-9, but its binding pocket is the most shallow solved to date. Moreover, we identified a residue (Trp-409) important for GII activity and propose two models of how it could be influencing GII␤-mediated enhancement of GII activity.
Cloning and Expression of Recombinant Mature S. pombe GII␤-Total RNA from S. pombe was obtained from exponentially growing S. pombe wild-type cultures as described (19). GII␤-coding DNA lacking the signal peptide (first 23 amino acids) sequence was obtained by RT-PCR using the primers B2sNde1 (5Ј-GGAATTCCATATGGCAAATGACCTCCGT-GGTG-3Ј) and B2aNot1 (5Ј-ATAGTTTAGCGGCCGCCT-CATCGACAGATGATTC-3Ј) that include NdeI and NotI restriction sites (underlined) and a Superscript II reverse transcriptase kit (Invitrogen). The PCR product was cloned in NdeI/NotI sites of pET22b(ϩ) vector as a C-terminal His 6 fusion protein. The construction was transformed into BL21 3030 cells. After an overnight induction at 18°C with 1.5 mM isopropyl ␤-D-thiogalactoside, bacteria were lysed by sonification in binding buffer (20 mM Tris-HCl, pH 7.9, 0.5 mM NaCl, 5 mM imidazole) with 0.5% (v/v) Triton X-100, 1 mg/ml lysozyme, and protease inhibitors (100 M PMSF, 10 M leupeptin, 10 M pepstatin, and 10 M E64) and centrifuged for 20 min at 20,000 ϫ g. The protein was purified from the supernatant by immobilized metal ion affinity chromatography using a column loaded with chelating Sepharose (GE Healthcare) precharged with NiCl 2 . After extensive washes with binding buffer followed with wash buffer (binding buffer plus 60 mM imidazole), elution was performed with elution buffer (binding buffer plus 1 M imidazole). The protein was dialyzed against 50 mM phosphate buffer, pH 7.4, 100 mM NaCl. The purification and identity of the protein in the soluble fraction were verified by 10% SDS-polyacrylamide gel electrophoresis (PAGE) and Western blot, respectively, with a mouse polyclonal anti-GII␤ S. pombe antibody (1:1000; Ref. 14) and mouse anti-His (1:1000; Sigma).
S. pombe Microsomal Fraction Preparations-S. pombe microsomes were prepared from 250 ml of exponentially growing cultures at an A 600 of 2. Cells were harvested, washed with 5 mM NaN 3 , and broken by 10 repetitive cycles of 1-min vortexing on ice with glass beads in 0.25 M sucrose, 20 mM imidazole, 5 mM EDTA with protease inhibitors (100 M PMSF, 10 M L-1-tosylamido-2-phenylethyl chloromethyl ketone, 10 M N-p-tosyl-L-ly-sine chloromethyl ketone, 10 M leupeptin, 10 M pepstatin, and 10 M E64), and the microsomal fraction was obtained as described (10). Protein concentrations were determined using a Bio-Rad protein assay as described by the manufacturer.
Protein Expression and Purification of S. pombe GII␤ MRH Domain-MRH GII␤ cDNA encoding residues 357-450 was cloned into a pQE30 vector modified to express the small ubiq-FIGURE 1. Quality control of glycoprotein folding in the ER. A, schematic diagram of the glycan-processing steps during quality control of glycoprotein folding in the ER. During N-glycosylation, a glycan (Glc 3 Man 9 GlcNAc 2 ; B) is transferred to the Asn-X-Ser/Thr sequence of nascent proteins that are being translocated into the ER. The sequential action of glucosidase I (GI) and GII produces monoglucosylated glycans able to interact with the lectin chaperones calnexin (CNX) and/or calreticulin (CRT). This interaction facilitates both glycoprotein folding and interaction with the protein-disulfide isomerase ERp57, thus preventing exit from the ER of misfolded glycoproteins until proper folding is achieved (4,45). The formation of monoglucosylated glycoproteins occurs either by deglucosylation of the transferred N-glycan or by reglucosylation by UGGT. UGGT is a glycoprotein folding status sensor that transfers a glucose residue to the Man 9 GlcNAc 2 oligosaccharide of glycoproteins that have not yet acquired their native structure. Removal of mannose residues by ␣-mannosidase(s) generates a degradation signal on misfolded/slow folding glycoproteins. OS-9 recognizes Man␣1,6Man␣1,6Man on the trimmed C arm via its MRH domain (7) and facilitates entry of misfolded glycoproteins into the ER-associated degradation pathway where the misfolded glycoproteins exit the ER and are degraded by the proteasome in the cytosol. SEC61, translocon; OST, oligosaccharyltransferase; ribbon diagram representing the crystal structure of calnexin is shown (Protein Data Bank code 1JHN (53)); ␣-MAN, ␣-mannosidase; ribbon diagram of human OS-9 is shown (Protein Data Bank code 3AIH (7)). B, structure of the oligosaccharide transferred to proteins during N-glycosylation. Lettering (a-n) follows the order of monosaccharide addition during the synthesis of Glc 3 Man 9 GlcNAc 2 -P-P-dolichol. Glucosidase I (GI) removes residue n. GII removes residues l and m. UGGT adds residue l to residue g.
uitin-like modifier (SUMO) protein with an N-terminal hexa-His tag and transformed into BL21(pREP4) cells. Constructs containing the amino acid substitutions W409A and W409F were generated using the QuikChange Lightning site-directed mutagenesis kit (Stratagene). Cells were grown at 37°C until an OD of ϳ0.7 at which time protein expression was induced by the addition of 1 mM isopropyl ␤-D-thiogalactoside. Cells were harvested after 18 h. Protein was purified from insoluble inclusion bodies using metal affinity chromatography. Refolding was performed by diluting the protein into 100 mM Tris, pH 8.0, 10 mM cysteine, 0.5 mM cystine and stirring for 2 h at 22°C. Following concentration, the SUMO-specific peptidase ULP-1 was added to remove the SUMO domain. The protein was dialyzed and subjected to metal affinity chromatography to separate the cleaved SUMO peptide from the MRH domain. Protein for NMR analyses was then buffer-exchanged into a solution containing 10 mM deuterated imidazole (pH 7.1) and 150 mM NaCl. The protein was further purified by size exclusion chromatography using a Superdex 75 column. The concentration of the purified protein was determined using the Bradford assay (Bio-Rad) with bovine serum albumin (BSA) as the standard.
GII␤ Mutagenesis and Expression in S. pombe-The Gateway pDONR201 plasmid containing clone 26/D11 (S. pombe GII␤ subunit, which was obtained from RIKEN DNA Bank (20)) was used as the template for single amino acid PCR mutagenesis of the GII␤ MRH domain (W409A or W409F). The amplified mutant DNA containing both GII␤ and vector sequences was phosphorylated, religated, and electroporated into DH5␣ cells. The following primers were used (mutagenic codons are underlined): W409F forward mutagenic primer 5Ј-TTCAATGG-TCCTCATAGATCTGCC-3Ј or W409A forward mutagenic primer 5Ј-GCTAATGGTCCTCATAGATCTGCC-3Ј was used with W409 reverse primer 5Ј-GCAACTTTGACCATTT-TCATAC-3Ј. Wild-type and mutant GII␤ DNA clones were transferred to the pREP1-ccdb2 Gateway-compatible S. pombe destination expression vector (RIKEN DNA Bank) by the LR recombination reaction (Invitrogen). S. pombe competent ⌬GII␤ cells were electroporated with the pREP1-GII␤ episomal constructs. The MRH binding pocket mutant constructs used in this study (pREP1-GII␤ Y462F and pREP1-GII␤ E456Q) were described in Stigliano et al. (14). Please note that the numbering in that case starts with the initial methionine, although throughout this work the numbering corresponds to the mature protein sequence.
GII Activity Assays-GII activities using the small substrate analog pNPG and labeled N-glycans as substrates were assayed in S. pombe microsomal fractions as described (14). Briefly GII activity in ⌬GII␤ mutants transformed with wild-type or mutant regulatory subunits was assayed using 5 mM pNPG as the substrate in 0.1 M HEPES buffer, pH 7.2 for 20 min at 37°C, and absorbance at 405 nm was measured, or [ 14 C-Glc]Glc 1 Man 9 GlcNAc was used as the substrate in 40 mM sodium phosphate buffer, pH 7.2 for 15 min at 30°C. In the latter case, glucosidase activity was determined as the percentage of total glucose released. Complementation assays between GII␣ and endogenous GII␤ were performed using a mixture of 125 g of proteins from microsomes containing GII␣ but not GII␤ (⌬GII␣␤ cells transformed with pREP3x-GII␣VDEL as described (14)) with 125 g of proteins from microsomes containing GII␤ but not GII␣ (⌬GII␣ S. pombe cells expressing only endogenous GII␤). When assaying GII␤ enhancement of GII␣ activity toward N-glycans of recombinant GII␤ or MRH, the microsomal source of GII␤ subunit was replaced with 125 ng of pure recombinant GII␤ (this is in the linear range of activation; data not shown) or with the indicated MRH/GII␤ or BSA/GII␤ mass ratios. The mixtures were preincubated for 30 min at 4°C in the presence of 1% Triton X-100 and then assayed for GII activity with [ 14 C-Glc]Glc 1 Man 9 GlcNAc as described above.
Purification of Man 9 GlcNAc-Man 9 GlcNAc was obtained from endo-␤-N-acetylglucosaminidase H-digested denatured soybean agglutinin purified from untoasted soybean meal by affinity chromatography on acid-treated Sepharose 6B (23). Soybean agglutinin (10 mg) was denatured in 8 M urea for 5 h at 65°C; dialyzed against H 2 O; preincubated for 10 min at 95°C in 0.5% SDS, 0.1 M ␤-mercaptoethanol in 50 mM triethylamine acetate buffer, pH 5.5; and treated with 20 milliunits of endo-␤-N-acetylglucosaminidase H (Sigma) for 16 h at 37°C. Proteins were precipitated with 66% methanol. To separate Man 9 from remaining proteins, the supernatant was dried, resuspended in 10% ethanol, and resolved in a Superdex Peptide gel filtration column (GE Healthcare) in 10% ethanol at 1 ml/min. The glycan-containing fractions were detected using the phenol-sulfuric acid assay (24), and protein elution was monitored by absorbance at 280 nm.
Analysis of Glycans Synthesized in Vivo-To assess ER N-glycan composition, S. pombe cells in the exponential growth phase expressing GII␤ mutations were harvested, extensively washed with 1% yeast nitrogen base medium without glucose, and resuspended in 2 volumes (v/w) of the same medium. Cells (0.5 ml) were then preincubated for 5 min in 5 mM DTT and pulsed for 15 min with 5 mM glucose containing 150 Ci of [ 14 C]glucose (300 Ci/mol). Further details on the labeling procedure and the preparation of whole cell endo-␤-N-acetylglucosaminidase H-sensitive N-glycans have been described previously (25). Glycans were separated by paper chromatography using Whatman 1 papers and 1-propanol/nitromethane/H 2 O (5:2:4) as solvent, and the peaks were identified by standards run in parallel. To improve the resolution, the identified glycans were eluted from paper and resolved by HPLC using a TSKgel Amide-80 column (4.6-mm inner diameter ϫ 25 cm; Tosoh) with a mobile phase of H 2 O/CH 3 CN in a linear gradient from 35:65 to 55:45 over 65 min and a flow rate of 0.75 ml/min at room temperature. Because of slight variations in retention times among runs, the positions of the peaks in paper chromatography and not the retention times from HPLC were used to identify glycans.
Surface Plasmon Resonance (SPR) Analyses-Biosensor studies were carried out as described previously (26). All SPR measurements were performed at 25°C using a BIAcore 3000 instrument (GE Healthcare). Purified protein (acid ␣-glucosidase (GAA), GAA phosphomonoester, or GAA phosphodiester) was immobilized on a CM5 sensor chip (GE Healthcare) following activation of the surface using 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide and N-hydroxysuccinimide. The proteins were injected onto the activated dextran surface at a concentration of 10 -20 g/ml in 10 mM sodium acetate buffer, pH 5.0 (GAA phosphomonoester), pH 4.5 (GAA phosphodiester), or pH 5.0 (GAA) using immobilization buffer (10 mM MES, pH 6.5, 150 mM NaCl, 0.005% (v/v) P20) as the running buffer. After coupling, unreacted N-hydroxysuccinimide ester groups were blocked with ethanolamine. The reference surface was treated in the same way except that protein was omitted. To assess binding affinities, samples of purified proteins were prepared in running buffer (50 mM imidazole, pH 7.4, 150 mM NaCl) and injected in a volume of 80 l over the coupled and reference flow cells at a flow rate of 40 l/min. After 2 min, the solutions containing the purified proteins were replaced with buffer, and the complexes were allowed to dissociate for 3 min. The sensor chip surface was regenerated with a 10-l injection of 10 mM HCl at a flow rate of 10 l/min. The surface was allowed to re-equilibrate in running buffer for 1 min prior to subsequent injections. The response at equilibrium (R eq ) for each concentration of protein was determined by averaging the response over a 10-s period within the steady state region of the sensorgram using the BIAevaluation software package (version 4.0.1). The response at equilibrium (R eq ) was plotted versus the concentration of protein and fitted by nonlinear regression to a one-site saturation binding model using the equation y ϭ (R max ϫ [MPR])/(K d ϩ [MPR]) (SigmaPlot version 10.0, Systat Software, Inc.). All response data were double referenced (27) where controls for the contribution of the change in refractive index were performed in parallel with flow cells derivatized in the absence of protein and subtracted from all binding sensorgrams.
Data Collection and Structure Determination by Heteronuclear NMR Spectroscopy-NMR spectra were acquired at 25°C on a Bruker 600-MHz spectrometer equipped with a triple resonance CryoProbe TM . Spectral data were processed using NMRPipe (28) and visualized using XEASY (29). Complete 1 H, 15 N, and 13 C resonance assignments for GII␤ MRH domain were obtained in a semiautomated manner using peak lists from three-dimensional HNCO, HNCOCA, HNCOCACB, HNCA, HNCACB, HNCACO, CCONH, HBHACONH, HCCONH, and HCCH total correlation spectroscopy and 13 C (aromatic)-edited NOESY-HSQC spectra and the assignment program GARANT (30). Automated assignments were verified and completed by manual inspection. Structures of the GII␤ MRH domain were calculated using distance constraints derived from three-dimensional 15 N-edited NOESY-HSQC and 13 C-edited NOESY-HSQC spectra ( mix ϭ 80 ms). Backbone and dihedral angle constraints were generated using TALOSϩ (31) and the secondary shifts of the 1 H, 13 C␣, 13 C␤, 13 CЈ, and 15 N nuclei. Initial structures were generated in an automated manner using the NOEASSIGN module of the tor-sion angle dynamics program CYANA (32). Initial structures underwent iterative rounds of manual refinement in CYANA to eliminate constraint violations. The top 20 CYANA conformers with the lowest target function were further refined using an Xplor-NIH (33) molecular dynamics protocol in explicit solvent (34). The structural statistics are listed in Table 1.
Determination of Binding Affinities by Heteronuclear NMR Spectroscopy-NMR studies in which 15 N-labeled GII␤ MRH domain was titrated with increasing concentrations of ligand (Man-6-P (Sigma), Man␣1,2Man (Dextra Laboratories), or Man 9 GlcNAc purified from soybean agglutinin were carried out as described previously (35).

RESULTS AND DISCUSSION
Full Functionality of the E. coli-expressed S. pombe GII␤ MRH Domain-We have previously shown that, in the absence of GII␤, GII␣ folds into an active subunit that can hydrolyze the small substrate analog pNPG but is unable to efficiently hydrolyze glucose from N-glycans in vivo or in vitro (14). We developed a complementation assay and showed that a mixture of microsomes derived from mutant S. pombe cells devoid of GII␣ or GII␤ subunits could partially restore the level of GII deglucosylation of Glc 1 Man 9 GlcNAc observed in microsomes from wild-type cells (14). Using this assay, addition of recombinant GII␤ dramatically increased the Glc trimming from Glc 1 Man 9 GlcNAc of S. pombe microsomes lacking GII␤ (Fig.  3A). To determine whether the isolated MRH domain of S. pombe GII␤ (amino acids 357-450; Fig. 2) was functional when expressed in E. coli, we used the same assay. Addition of up to 50 times more molar amounts of the recombinant GII␤ MRH domain (compared with full-length GII␤) to microsomes containing only GII␣ subunits did not enhance GII enzymatic activity (Fig. 3A). This result was expected as the MRH domain lacks the portion of the GII␤ subunit responsible for the interaction between GII␣ and GII␤ subunits (GII␤ G2B domain; Fig.  2A) (15,16). However, addition of GII␤ MRH domain inhibited GII activity in mixtures of microsomes containing GII␣ plus GII␤ subunits in a dose-dependent manner (Fig. 3B), thus strongly suggesting that the recombinant GII␤ MRH domain was sequestering the high mannose glycan portion of the substrate. In contrast, BSA had no effect (Fig. 3B).
To directly assess the ability of GII␤ MRH domain to bind high mannose-type glycans, SPR experiments were conducted in which a glycoprotein, GAA, was immobilized and GII␤ MRH domain was flowed over the surface of the sensor chip. The immobilized recombinant GAA contains exclusively high mannose-type glycans with 8.35 mannose residues on average per glycan (36). In these studies, we used a robust E. coli expression system in which recombinant GII␤ MRH domain was purified from insoluble inclusion bodies and refolded. The results dem-onstrate that the MRH domain binds high mannose-type glycans with an estimated K d of 120 M (Fig. 4A). The affinity measured by SPR, although lower, is comparable with the affinities (K d ϭ ϳ20 -80 M) determined for a tetrameric form of the human GII␤ MRH domain using frontal affinity chromatography and 2-aminopyridine-labeled glycans (13). Differences may be due to differences between species, between substrates, in the analytical methods used for measurements, and between monomeric and tetrameric avidities of the protein preparations. Taken together, these results demonstrate that the refolded recombinant S. pombe GII␤ MRH domain is functional in binding high mannose N-glycans and thus is suitable for structural analyses (13).   presence of a carbohydrate ligand (35,37). These structures allowed us to identify four conserved residues (Gln, Arg, Glu, and Tyr), which interact with the 2-, 3-, and 4-hydroxyl groups of the mannose ring, that are essential for Man-6-P recognition by the MRH domains of MPRs. These four residues, which form the binding pocket, are also conserved in the GII␤ MRH domain (Fig. 2B). Mutagenesis of the analogous residues in S. pombe GII␤ MRH domain abolished the ability of GII␤ to enhance glycan trimming by GII in live cells (14). Previous studies showed that the GII␤ MRH domain recognizes mannose residues in the B and/or C arms of N-glycans as the absence of these mannose residues in the glycan prevents trimming of the glucose in arm A by GII both in vitro and in living cells (11-13, 15, 38). The differences in substrate specificity between this ER glycoprotein folding quality control enzyme and OS-9, a receptor involved in ER-associated degradation, prompted us to obtain the solution structure of the isolated GII␤ MRH domain from S. pombe using three-dimensional NMR techniques. Sam-ple conditions were optimized as described under "Experimental Procedures" to obtain high quality spectra of 15 N/ 13 C-labeled recombinant GII␤ MRH domain at pH 7.2 (Table 1).
S. pombe GII␤ is the shortest (89 residues; Fig. 2) of the MRH domain structures that have been determined to date (eight of the 15 MRH domains of CI-MPR (35, 39 -41), CD-MPR (42), and OS-9 (7)). GII␤ MRH domain folds into a flattened ninestranded ␤-barrel comprising two antiparallel ␤-sheets oriented orthogonally over one another (Fig. 5, A-C). This structure is similar to the structures reported for the MRH domains of the MPRs and OS-9 with the overall fold of GII␤ MRH domain being most similar to that of OS-9 (r.m.s.d. ϭ 1.32 Å; 85 C␣ atoms) (Fig. 5D). Both OS-9 and GII␤ adopt a compact ␤-barrel structure due to their lack of ϳ20 amino acids present at the N terminus of the CD-MPR and CI-MPR MRH domains that form either an ␣-helix or two to three short ␤-strands, respectively. However, S. pombe GII␤ MRH domain loops A and B as well as ␤-strands 4/5 and 6 are significantly shorter than those in OS-9 (Fig. 5D). Additionally, S. pombe GII␤ MRH domain contains only four cysteine residues in contrast to the six (CD-MPR, CI-MPR domain 5, and OS-9) or eight (domains 3 and 9 of the CI-MPR) cysteine residues found in the other glycan-binding MRH domains. The absence of the N-terminal disulfide bond in S. pombe GII␤ MRH domain that is found in other MRH domains as well as other species of GII␤ does not appear to affect the overall fold of the domain. The structural stability provided by the N-terminal disulfide bond may be . The response at equilibrium (R eq ) to the GAA surfaces is plotted versus the concentration of protein (0, 5, 10, 20, 40, 80, 120, and 160 M) and fitted as described previously (35). GAA contains exclusively high mannose-type glycans with 8.35 mannose residues on average per glycan and either 0.25 mol of Man-6-P/mol of glycan (phosphomonoester) or GlcNAc-modified glycan (phosphodiester) (36). The mean and S.D. from three independent analyses are as follows for the high mannose, phosphomonoester, and phosphodiester surfaces, respectively: wild type, 121    in the ligand-bound state. These results demonstrate that the structure of GII␤ MRH domain is relatively constrained with the exception of residues in loops C and D (Fig. 5, A, C, and E).
Binding Pocket and Specificity of GII␤ MRH Domain-To map the binding pocket of GII␤ MRH domain with carbohydrates, increasing amounts of the disaccharide Man␣1,2Man, the monosaccharide Man-6-P or glucose 6-phosphate (Glc-6-P), or the glycan Man 9 GlcNAc were added to 13 C/ 15 N-labeled sample, and changes in chemical shifts of amide groups located in either backbone or side chains were monitored. These ligand titrations were also used to (i) determine whether pure glycan and simple carbohydrate molecules bound specifically to the conserved MRH domain recognition site and (ii) test the hypothesis that the GII␤ MRH domain selectively recognizes mannose in contrast to glucose. An overlay of spectra acquired in the presence of increasing concentrations of Man␣1,2Man, Man-6-P, or Man 9 GlcNAc demonstrated progressive chemical shift changes for residues either within or adjacent to the proposed binding pocket (Fig. 6A). The weighted changes in combined 1 H/ 15 N chemical shift perturbations were used to determine relative binding affinities for Man␣1,2Man (K d ϭ 65 Ϯ 3 mM), Man-6-P (K d ϭ 77 Ϯ 7 mM), and Man 9 GlcNAc (K d ϭ 2.6 Ϯ 0.4 mM) (Figs. 7, A-C). These results show a significantly higher relative affinity (ϳ25-fold) for the larger glycan than for the disaccharide. The ability to recognize both mannose and Man-6-P suggests that substitution at the 6-position (i.e. hydroxyl or phosphate) does not constitute a major recognition determinant for GII␤ MRH domain. This hypothesis is further validated by SPR analyses show that K d values for GII␤ MRH domain interacting with an immobilized glycoprotein bearing high mannose (Fig. 4A), phosphomonoesters (Fig. 4B), or phosphodiesters with a GlcNAc on the 6Ј-OH group of the mannose (Fig. 4C) are similar. Thus, the GII␤ MRH domain differs from the P-type lectin MRH domains of the MPRs that exhibit a preference to bind either phosphate on the 6-hydroxyl of mannose (domain 3 of the CI-MPR and CD-MPR) or a phosphodiester (domain 5 of the CI-MPR) (36) and from human OS-9, which binds mannose but not Man-6-P (43).
To probe whether the GII␤ MRH domain can recognize glucose, 15 N-1 H HSQC collected in the presence of Glc-6-P, which differs from Man-6-P only in the positioning of the 2Ј-OH group, produced chemical shift perturbations localized to the same regions as Man-6-P although to a significantly lower extent (Figs. 6B and 7D). These weak interactions observed with Glc-6-P demonstrate that the axial 2Ј-OH present in mannose is a key recognition determinant for the GII␤ MRH domain. Although K d values of 65-77 mM for the monosaccharide Man-6-P and disaccharide Man␣1,2Man may not equate to the binding interactions occurring between GII and its heterogeneous population of N-glycan-containing substrates in the ER, the results demonstrate that GII␤ MRH domain interacts preferentially with mannose. Taken together, the observation that the affinity for the monosaccharide Man-6-P and the disaccharide Man␣1,2Man are similar along with the ability of Glc-6-P to cause specific but very low chemical shift perturbations supports the structural findings that GII␤ MRH domain has a shallow binding pocket (see below) that accommodates preferentially a single mannose residue. From a functional standpoint, the inability of GII␤ MRH domain to efficiently recognize glucose eliminates the potential for the ␤-subunit to compete with the catalytic ␣-subunit of GII for glucose present in arm A of the N-glycan.
Mannose Binding Pocket-GII␤ MRH domain contains the four residues (Gln-384, Arg-414, Glu-433, and Tyr-439) known to be essential for Man-6-P binding by the MPRs (Fig. 2B), and mutational studies support their involvement in GII activity (13,14). To evaluate the role these residues play in carbohydrate binding, measurements of weighted differences in com- Missing values correspond to proline residues or residues not observed in the 15 N-1 H HSQC spectra. Middle panels depict chemical shift mapping on the GII␤ MRH domain structure. The molecular surface highlighted in red corresponds to amino acids with the largest chemical shift perturbations (Ͼ1), whereas those colored in cyan represent moderate perturbations (0.5). Right panels, chemical shift differences for the eight selected residues (labeled in A) were calculated as ⌬␦ ϭ [(5 ϫ ⌬NH␦) 2 ϩ ⌬N␦] 1/2 . Representative curves for four of the eight residues monitored for the Man␣1,2Man and Man-6-P titrations and curves for the three residues monitored for the Man 9 GlcNAc titration are shown. Dissociation constants (K d ) were determined from monitoring changes in the chemical shift of the eight selected residues as described under "Experimental Procedures." bined 1 H/ 15 N chemical shift perturbations with increasing amounts of ligand were used to map the location of the Man␣1,2Man, Man-6-P, and Man 9 GlcNAc (Fig. 7, A-C) binding site to the surface of GII␤ MRH domain. The greatest chemical shift perturbations of the ligands map to an area (loops C and D) corresponding to the known carbohydrate binding sites of CD-MPR, of domains 3 and 5 of the CI-MPR, and of OS-9. Although chemical shift perturbations were observed for residues in the proposed binding pocket, two of the essential ligand binding residues, Gln-384 and Tyr-439, did not show significant changes in chemical shift in the presence of ligand (Fig. 7). We postulate from the structure that the amide protons are unable to shift due to extensive hydrogen bonding with neighboring atoms due to their location in ␤-strands that are part of the two ␤-sheets. Residues in loop A experienced slightly greater chemical shift differences in the presence of the larger ligands: a stronger chemical shift perturbation at Gly-370 and additional perturbations for Tyr-372 and Thr-373 were observed (Fig. 7, A-C). However, the striking similarity of the chemical shift mapping between the disaccharide and the monosaccharide coupled with their similar affinities supports the lack of an extended binding pocket. This notion is further supported by the observation that the branched Man 9 GlcNAc glycan, which exhibited ϳ25-fold higher affinity than Man-6-P or Man␣1,2Man, failed to produce additional measurable chemical shift perturbations at the highest concentration used (Fig.  7). Although these structural analyses failed to identify additional protein contacts to explain the increased affinity observed for the Man 9 GlcNAc glycan compared with the smaller ligands, the longer glycan may have constrained the terminal and penultimate mannose into an optimal conformation for binding. Furthermore, the observed ϳ20-fold higher affinity of GII␤ MRH domain for a glycoprotein bearing high mannose-type N-glycans compared with a glycan free in solution (K d ϭ ϳ120 M (Fig. 4A) versus K d ϭ 2.6 Ϯ 0.4 mM (Fig.  7C); SPR data were obtained with glycoproteins, whereas NMR data were obtained using free glycans) may be the result of protein-induced restrictions that increase the percentage of N-linked glycans in the optimal conformation for binding and/or of differences in the principles of the methods used to estimate affinities.
The lumen of the ER is a crowded environment with protein concentrations up to 200 -300 mg/ml (44). Given the high substrate concentration for the ER quality control machinery in which transient interactions occur, it is reasonable to expect that the affinities of ER MRH domains for glycoproteins in this environment would not be as high as those of the Golgi MRH domains in which the Man-6-P receptors must capture the very low abundance of lysosomal enzymes for transportation to late endosomes. The limited interactions with mannose and the failure of either Man 9 GlcNAc or a disaccharide to produce submillimolar binding affinities are likely a direct consequence of the shallowness of this binding pocket relative to the other MRH binding pockets solved to date (CD-MPR, domains 3 and 5 of CI-MPR, and OS-9) (Fig. 8). In S. pombe, this shallow mannose binding region is due to the dramatically shortened loop B relative to other species and the positioning of loop C more distally from the essential residues in the pocket, thereby allow-ing the binding site to accommodate mannose modified at the 6-hydroxyl (i.e. phosphate and GlcNAc) (Figs. 5 and 8). Notably, the four residues considered essential for mannose binding (Gln-384, Arg-414, Glu-433, and Tyr-439) overlay well with those in the other four MRH domains (Fig. 8). Because the ligand-bound crystal structures of domain 3 of the CI-MPR, OS-9, and the CD-MPR demonstrate directly that the common positioning of these four essential residues results in a nearly identical positioning of the mannose ring within the binding pocket, we predict that the mode of mannose binding by the GII␤ MRH domain to these four residues will be essentially the same.
Although the four mannose-interacting residues are conserved among these MRH domains, other structural features must exist to explain the different preferences displayed by these proteins for high mannose-type glycans (45). For example, OS-9 contains a double tryptophan motif that is proposed to determine the ␣1,6-linkage specificity of OS-9 (7). These residues form an extended surface to accommodate the penultimate and prepenultimate mannose residues of one arm of the glycan (see Fig. 8) (7). In contrast, GII␤ MRH domain contains a conserved tyrosine residue (Tyr-372; strand ␤2) as observed in the MPRs (Figs. 2B, cyan, and 8). This tyrosine has been proposed by Satoh et al. (7) to be the residue responsible for conveying ␣1,2-linkage specificity by the MPRs. However, we have crystallized domain 3 of the CI-MPR with a Man␣1,3Man in the binding pocket (39), suggesting that this tyrosine (Tyr-324) in domain 3 is not involved in specifying a linkage preference for ␣1,2-containing ligands. Furthermore, the side chain of Tyr-372 of the GII␤ MRH domain, unlike that of the MPRs, is oriented 90°away from the binding pocket such that the hydroxyl group is unable to interact with the glycosidic linkage of the ligand (Fig. 8). The significant number of NOE constraints for this residue along with the rest of the molecule (average of 21 NOE constraints per residue) further validates the positioning of the side chain of Tyr-372 in the GII␤ MRH domain structure determined in the absence of ligand. However, we cannot exclude the possibility that in the presence of ligand the side chain of Tyr-372 may reposition. Thus, additional studies are needed to evaluate the role of this conserved residue in the MPRs and GII␤.
Role of a Conserved Trp-409 among GII␤ MRH Domains-A comparative analysis of the residues within and surrounding the proposed binding pocket revealed another residue that is highly conserved and unique to the GII␤ MRH domain. Trp-409, which is located in loop C, is absolutely conserved among 77 of the 79 homologs analyzed (Fig. 8) and is not present in the other MRH domains (Fig. 2B). Our initial assessment of the role of the conserved Trp-409 was performed in the context of the isolated MRH domain. Mutants containing conservative (W409F) or non-conservative (W409A) substitution of Trp-409 were generated, and 15 N-labeled proteins were evaluated. 15 N-1 H HSQC experiments show that both constructs are folded, which is reflected in their similar 15 N-1 H HSQC spectra as the wild-type protein (Fig. 9A). Furthermore, addition of 50 mM Man␣1,2Man disaccharide to the W409A 15 N-labeled sample resulted in chemical shift perturbations on the order of those found for similar concentrations of the wild-type protein (Fig. 9B). In addition, SPR experiments show that W409A and W409F bound immobilized glycoproteins bearing high mannose (Fig. 4A), phosphomonoesters (Fig. 4B), or phosphodiesters bearing a GlcNAc on the 6Ј-OH group of the mannose (Fig. 4C) with a slight decrease in affinity (within 3.5-fold) compared with the wild-type protein. Thus, substitution of Trp-409 does not dramatically lower binding affinity toward glycans in the context of the isolated MRH domain.
To study the influence of Trp-409 of GII␤ MRH domain in GII activity, we mutated Trp-409 in the full-length GII␤ (W409A or W409F), expressed the wild-type or mutated GII␤ in S. pombe cells lacking endogenous GII␤ (⌬GII␤), and measured GII activity toward both the small substrate analog pNPG and [ 14 C-Glc]Glc 1 Man 9 GlcNAc in microsomal fractions obtained from the mutant cells. Results presented in Fig. 10A show that although the influence of Trp-409 in GII␤-mediated GII␣ activity enhancement was not as strong as mutation of the conserved residues (Gln, Arg, Glu, and Tyr) that interact with the 2-, 3-, and 4-hydroxyl groups of the mannose ring (Fig. 2B), mutation of Trp-409 precluded full GII activity toward N-glycans in vitro. We have reported previously that microsomal GII activity using pNPG reflects ER GII␣ content (14). Reduced GII activity observed in cells expressing Trp-409-mutated GII␤ is not due to a reduced ER GII␣ content as the activity toward pNPG was not reduced in mutant GII␤ W409A or GII␤ W409F (Fig. 10A). Moreover, Western blot analyses show normal GII␣ levels in microsomes (enriched in ER membranes) obtained from GII␤ Trp-409 mutants (Fig. 10B).
To determine whether substitution of Trp-409 in the GII␤ MRH domain affects GII glucose trimming activity in vivo, we analyzed the N-glycan pattern produced by ⌬GII␤ cells expressing Trp-409 GII␤ mutants (W409A and W409F) and compared it with the N-glycan patterns of cells expressing wildtype GII␤ or GII␤ mutated in amino acids predicted to interact with the mannose ring (Y439F and E433Q). Cells were incubated with [ 14 C]Glc for 15 min in the presence of 5 mM DTT, and whole cell N-glycans were isolated and analyzed as described under "Experimental Procedures." The total incorporation of the label into the N-glycans was linear during the incubation time, and the presence of DTT prevented the passage of glycoproteins to the Golgi, hindering further extension of the recently synthesized N-glycans (46). The N-glycan patterns obtained from S. pombe cells expressing wild-type GII␤ revealed that whereas deglucosylation of the transferred glycan Glc 3 Man 9 GlcNac 2 was so rapid that almost no glucose-containing glycans were detected (Fig. 10C) the amount of glucosylated glycans increased in cells expressing GII␤ Trp-409 mutants, although it was not as much of an increase as observed in the GII␤ Y439F mutant (Fig. 10, D-F) and as observed previously in ⌬GII␤ mutants (14). Taken together, the results demonstrate that substitution of Trp-409 in the context of the fulllength GII␤ has a moderate to high (60% inhibitory) effect on GII activity.
How Does GII␤ MRH Domain Interact with High Mannosetype Glycans?-Studies both in vitro and in vivo indicate that residue k in arm C plays a more significant role than residue i in arm B of the glycan in the enhancement of GII catalytic activity by GII␤ (12)(13)(14)(15). To envision how GII␤ enhances the catalytic activity of GII toward glycans, we propose two models that are consistent with the existing biochemical and current structural studies. In the first model, the terminal mannose (residue k) of arm C is captured by the shallow binding pocket of the MRH domain of GII␤ defined by the four conserved residues (Gln-384, Arg-414, Glu-433, and Tyr-439). Interaction of the terminal mannose (residue i) of arm B with Trp-409 would further constrain the N-glycan and facilitate optimal positioning of the glucose-containing arm A near the catalytic site of GII␣ (Fig.  11A). Visualization of this model was obtained by positioning the terminal mannose residue of arm C of a Man 9 glycan into the MRH binding pocket from the superimposition of the four mannose-binding essential residues (Gln-384, Arg-414, Glu-433, and Tyr-439) with that of domain 3 of CI-MPR containing bound Man-6-P (47). This overlay indicated that the terminal mannose of arm B of the glycan is near loop C and that Trp-409 in loop C could stack against the terminal mannose ring of arm B, thereby providing a bidentate mode of interaction between the MRH domain and two arms of the glycan. Although a previous molecular dynamics simulation of a Man 9 GlcNAc 2 glycan and NOE analyses indicate that significant flexibility of arms A, B, and C of the glycan (48) complicates such analyses, CH--mediated stacking interactions between aromatic residues and carbohydrates are commonly observed in lectins (49,50), including another ER-resident lectin, malectin, that recognizes the Glc␣1,3Glc moiety on Glc 2 -N-glycans (51). It is intriguing to speculate that Trp-409 is located in a relatively dynamic region of the GII␤ MRH domain (Fig. 5, A and C) to capture one of the many and related conformers of an N-glycan.
In the second model (Fig. 11B), Trp-409 interacts with other regions of the ␤-subunit. Although the GII␤ MRH domain alone is sufficient for binding glycans, its affinity for ligand may be augmented by the presence of the remaining portion of the ␤-subunit and/or the ␣-subunit. This enhancement in ligand binding affinity has been reported for both domain 3 (52) and domain 5 (26) of the CI-MPR. Trp-409 of GII␤ could be positioned to interact with residues on adjacent parts of the holoenzyme and act to either stabilize the position of loops C and/or D much like the linker region between domains 1 and 2 does for domain 3 of CI-MPR, resulting in increased affinity for Man-6-P (39, 47), or orient the MRH domain closer to GII␣, resulting   N-glycans. A, GII activity toward pNPG or Glc 1 Man 9 GlcNAc was measured in 125 g of microsomes from S. pombe cells lacking endogenous GII␤ (⌬GII␤) and expressing GII␤ or the same subunit displaying the indicated mutations. The activity of ⌬GII␤ cells transformed with wild-type GII␤ was taken as 100%. Error bars represent S.D. B, GII␣ ER content was evaluated. Microsomal proteins (250 g) of S. pombe ⌬GII␣, ⌬GII␤, or ⌬GII␤ expressing exogenous variants of GII␤ were resolved by 8% SDS-PAGE, transferred to a PVDF membrane, and blotted using mouse polyclonal anti-GII␣ (1:500) or anti-GII␤ (1:5000) subunit primary antibodies. Goat HRP anti-mouse IgG (1:5000) was used as the secondary antibody. Reactions were detected by chemiluminescence. Because endogenous GII␤ levels are significantly lower than exogenously expressed levels, the GII␤ band was detected in the GII␣ mutant (middle lane; ⌬GII␣) only upon overexposing the blot. C-F, glycan patterns synthesized by S. pombe ⌬GII␤ mutant cells expressing GII␤ (C) or mutated GII␤ versions (D and E). Endoplasmic reticulum N-glycans produced by ⌬GII␤ cells expressing GII␤ W409A were indistinguishable from those produced by cells expressing GII␤ W409F. The N-glycan pattern produced by ⌬GII␤ cells expressing GII␤ E433Q was indistinguishable from that of the same cells expressing GII␤ Y439F. Quantification of the relative amounts of the di-, mono-, and unglucosylated species in each mutant is shown in F. The label of Man 8 species was added to that in Man 9 species to account for unglucosylated glycans.
in arm A of the glycan becoming optimally positioned into the catalytic site. Both models indicate that the active site of GII␣ must be very close to that of the binding pocket of the MRH domain of GII␤. In comparison with the other MRH domains, the shorter loops between strands 2 and 3 and strands 4/5 and 6 along with shorter strands 4/5 and 6 ( Fig. 5D) may allow a close juxtapositioning of the catalytic site with the MRH domain.
Concluding Remarks-In the current study, we have determined the solution structure of the S. pombe GII␤ MRH domain, and its function was characterized alone and in the context of the full-length GII␤ subunit. This is the first time that the structure of an MRH domain of an enzyme has been determined. The structure reveals the conserved MRH fold observed in the MPRs (35, 39 -42) and OS-9 (7), but in contrast to these proteins, GII␤ MRH domain contains a shallow binding pocket that accommodates only a single mannose residue. In addition, we have identified a conserved residue, Trp-409, that plays an important role in GII glucose trimming activity in vivo. Additional structural studies are needed to fully understand how this heterodimeric enzyme interacts with N-glycans and thus regulates the quality control of glycoprotein folding in the ER.