Mechanism of Class 1 (Glycosylhydrolase Family 47) �-MannosidasesInvolved in N-Glycan Processing and Endoplasmic ReticulumQuality Control*□S

Received for publication, January 4, 2005, and in revised form, February 11, 2005Published, JBC Papers in Press, February 15, 2005, DOI 10.1074/jbc.M500119200

Khanita Karaveg‡§, Aloysius Siriwardena¶�, Wolfram Tempel§, Zhi-Jie Liu§, John Glushka‡,Bi-Cheng Wang§, and Kelley W. Moremen‡§**

From the ‡Complex Carbohydrate Research Center, University of Georgia, Athens, Georgia 30602, the §Department ofBiochemistry and Molecular Biology, University of Georgia, Athens, Georgia 30602, and the ¶Department of Chemistryand Biochemistry, University of Mississippi, University, Mississippi 38677

Quality control in the endoplasmic reticulum (ER) deter-mines the fate of newly synthesized glycoproteins towardeither correct folding or disposal by ER-associated degra-dation. Initiation of the disposal process involves selectivetrimming of N-glycans attached to misfolded glycoproteinsby ER �-mannosidase I and subsequent recognition by theER degradation-enhancing �-mannosidase-like proteinfamily of lectins, both members of glycosylhydrolase family47. The unusual inverting hydrolytic mechanism catalyzedby members of this family is investigated here by a combi-nation of kinetic and binding analyses of wild type andmutant forms of human ER �-mannosidase I as well as bystructural analysis of a co-complex with an uncleaved thio-disaccharide substrate analog. These data reveal the rolesof potential catalytic acid and base residues and the iden-tification of a novel 3S1 sugar conformation for the boundsubstrate analog. The co-crystal structure described here,in combination with the 1C4 conformation of a previouslyidentified co-complex with the glycone mimic, 1-deoxyman-nojirimycin, indicates that glycoside bond cleavage pro-ceeds through a least motion conformational twist of aproperly predisposed substrate in the �1 subsite. A novel3H4 conformation is proposed as the exploded transitionstate.

Newly synthesized membrane and secretory proteins arecommonly glycosylated as they enter the lumen of the endo-

plasmic reticulum (ER)1 (1). The protein-linked glycans areimmediately trimmed, and the resulting structures subse-quently serve as ligands for the luminal lectin chaperones,calnexin and calreticulin, which act in conjunction with otherchaperones and thiol disulfide oxidoreductases to help facili-tate protein folding, oligomerization, and disulfide bond forma-tion (2). Release from the chaperones and packaging into an-teriograde transport vesicles is generally accomplished onlyupon completion of the folding process. In contrast, proteinsthat fail to fold continually re-engage the chaperone machineryeither until folding is achieved or until the nascent glycopro-teins acquire a signal for disposal through the ER-associateddegradation (ERAD) pathway (3). Many normal proteins, aswell as many “defective” proteins in loss-of-function humangenetic diseases, have delayed rather than defective foldingkinetics that result in the premature disposal of the newlysynthesized polypeptides (2). Thus, the percentage of trans-ported functional protein is defined by a competition betweenthe kinetics of conformational maturation versus the rate ofacquiring the targeting signal for protein disposal.

The present model for ERAD invokes the specific cleavage ofa single mannose residue from the Man9GlcNAc2 glycan proc-essing intermediate by the slow acting enzyme, ER mannosi-dase I (ERManI), as a rate-limiting initiation step for targetingincompletely folded glycoproteins for disposal (4). The trimmedoligosaccharide, in the context of an unfolded polypeptide, issubsequently recognized by a collection of luminal ER lectins,termed EDEMs, which are nonhydrolytic homologs of ERManI(2, 5–8). The nature of the complex interactions between mis-folded glycoproteins, EDEMs, and other ER-associated compo-nents has still not been entirely resolved, but present modelsenvisage a “handing off” of the ERAD substrate to the Sec61translocon pore complex for retrotranslocation and subsequentproteasomal degradation (2).

ERManI, EDEMs, and a collection of Golgi processing hydro-lases, termed Golgi �-mannosidases IA, IB, and IC, are allmembers of glycosylhydrolase (GH) family 47 (9, 56). The hy-drolytic members of this family have unique branch specifici-

* Use of the Advanced Photon Source was supported by the UnitedStates Department of Energy, Office of Science, Office of Basic EnergySciences, under Contract W-31-109-ENG-38. Data were collected at theSE Regional Collaborative Access Team (SER-CAT) 22-ID beamline atthe Advanced Photon Source, Argonne National Laboratory (supportinginstitutions may be found on the World Wide Web at www.ser-cat.org/members.html). This work was supported by National Institutes ofHealth Research Grants GM47533 and RR05351 (to K. W. M.) and theGeorgia Research Alliance (to B. C. W). The costs of publication of thisarticle were defrayed in part by the payment of page charges. Thisarticle must therefore be hereby marked “advertisement” in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.

The atomic coordinates and structure factors (code 1X9D) have beendeposited in the Protein Data Bank, Research Collaboratory for Struc-tural Bioinformatics, Rutgers University, New Brunswick, NJ(http://www.rcsb.org/).

□S The on-line version of this article (available at http://www.jbc.org)contains an additional table and two figures.

� To whom correspondence may be addressed: Universite de PicardieJules Vernes, Faculte des Sciences, Laboratoire des Glucides, 33, ruesaint Leu, 80039 Amiens, France. Tel.: 33-322827527; Fax:33-322827562; E-mail: siriward@ccrc.uga.edu.

** To whom correspondence may be addressed: Complex Carbohy-drate Research Center, University of Georgia, Athens, GA 30602. Tel.:706-542-1705; Fax: 706-542-1759; E-mail: moremen@uga.edu.

1 The abbreviations used are: ER, endoplasmic reticulum; ERAD,endoplasmic reticulum-associated degradation; EDEM, ER degrada-tion-enhancing �-mannosidase-like protein; ERManI, ER �-mannosi-dase I; dMNJ, 1-deoxymannojirimycin; SPR, surface plasmon reso-nance; HPLC, high performance liquid chromatography; GH,glycosylhydrolase; Man8B, an isomer of Man8GlcNAc2 where the cen-tral branch mannose residue had been removed from the standardMan9GlcNAc2 structure (see Fig. 1 of Ref. 19); MES, 4-morpho-lineethanesulfonic acid; CHES, 2-(cyclohexylamino)ethanesulfonicacid; HEPBS, N-(2-hydroxyethyl)piperazine-N�-(4-butanesulfonic acid);RU, response unit.

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 280, No. 16, Issue of April 22, pp. 16197–16207, 2005© 2005 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

This paper is available on line at http://www.jbc.org 16197

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from

ties that are required for the complete trimming to the keyMan5GlcNAc2 intermediate structures necessary for matura-tion into complex type glycans on cell surface and secretedglycoproteins (9–11). Despite sequence similarities with family47 hydrolases, EDEM family members appear to lack hydro-lase activity, but they are proposed to accomplish their lectinfunction in ERAD by a mechanism analogous to glycan recog-nition during catalysis by the true hydrolases (2, 5–8). Thus,understanding how this family of enzymes and lectins accom-plish their functions will provide insights into the rate-limitingdecision between glycoprotein maturation and disposal in thesecretory pathway.

The two standard mechanisms for glycosidase-mediatedbond hydrolysis both involve the direct protonation of the gly-cosidic oxygen by an enzyme-associated catalytic acid and theattack of the anomeric carbon atom by a nucleophile (12).Glycosidases with retaining mechanisms generally employ anenzyme-associated nucleophile, usually a Glu or Asp carboxy-late side chain, resulting in the transient formation of an en-zyme-linked intermediate that is subsequently cleaved by anincoming general base-activated water molecule to result in areleased glycone that retains the configuration of the originalglycosidic linkage. In contrast, glycosidases with invertingmechanisms employ the direct nucleophilic attack by a generalbase-activated water molecule to result in inversion of config-uration of the released glycone. Both mechanisms require theformation of a transition state carrying a considerable positivecharge delocalized between the anomeric center (C-1) and thering oxygen (O-5) (12–14). The resulting partial double bondcharacter of the C-1–O-5 bond requires co-planarity of C-5, O-5,C-1, and C-2 at or near the transition state. Among the poten-tial pseudorotational conformational itineraries for sugar

pyranose ring interconversions, there are four potential confor-mations where the planarity of C-5, O-5, C-1, and C-2 is satis-fied (Fig. 1, 2,5B, B2,5, 4H3, and 3H4) (14). Recent data havesuggested that three of these potential transition state confor-mations (2,5B, B2,5, and 4H3) are employed among the knownGH families (14–18). Evidence favoring these conformationshas come from the analysis of x-ray structures of target en-zymes with trapped covalent intermediates using fluorinatedsubstrate analogs (for retaining glycosidases), with enzyme-bound nonhydrolyzable substrate mimics, or with natural sub-strates in combination with active site mutants (14–18). Of thefour possible coplanar transition states, only the 3H4 conforma-tion has not yet been identified as an intermediate in anyglycosidase mechanism.

Family 47 �-mannosidases from a variety of species have beenextensively studied in regard to enzyme kinetics, substrate spec-ificity, and structural analysis (9–11, 19–23). Mutagenesis stud-ies of potential residues involved in Ca2� binding and catalysishave also been carried out for S. cerevisiae ERManI (24), butthese studies predated the structural determination of either theyeast (21) or human (20) enzymes, and the roles of these residuesin catalysis and substrate binding have not been revisited. Twoclasses of �-mannosidase co-complexes have been characterizedby x-ray diffraction. Putative enzyme-glycan product co-com-plexes have been isolated for both ER and Golgi family 47 hydro-lases, revealing the structural basis for branch specificity bythese enzymes (19, 21). In addition, co-complexes between ER-ManI and the inhibitors, kifunensine and 1-deoxymannojirimy-cin (dMNJ), have revealed several unprecedented aspects of gly-cone binding and hydrolysis by this family of enzymes (20). Bothinhibitors were found to bind in the �1 subsite at the core of theenzyme (��)7 barrel catalytic domain (Fig. 2) in the equivalent of

FIG. 1. Pseudorotational itinerary for the interconversion of sugar ring conformations highlighting the proposed transitionstates employed by glycosidases. A partial map of the interconversions between pyranose ring sugar conformations (53) are shown with thenomenclature indicated for the reference plane of four ring atoms shaded blue as per IUPAC-IUB rules (54). The map highlights the likely routesfor interconversion between conformers from the low free energy 4C1 chair conformation through various half-chair (H), boat (B), and skew boat(S) conformations (14, 53). The ring oxygen is indicated by the O in the given pyranose ring structure, and the orientations of the sugars areapproximately the same for each monosaccharide conformer. Of the 21 conformers shown, only four have a structure in which C-5, O-5, C-1, andC-2 are co-planar (B2,5, 4H3, 2,5B, and 3H4) (14). The portions of the pyranose sugar rings with co-planarity at C-5–O-5–C-1–C-2 are indicated byred shading. Positions where there is an overlap of the nomenclature reference plane and the C-5–O-5–C-1–C-2 plane are indicated by brownshading. The blue boxed regions illustrate the proposed transition state conformations for B2,5, 4H3, and 2,5B bracketed by conformations that havebeen proposed based on the isolation of trapped intermediates in various co-crystal studies with inverting and retaining enzymes (14). The redboxed region illustrates the potential 3H4 transition state for glycosylhydrolase family 47 �-mannosidases bracketed by the 4C1 chair conformationfound for the ERManI�dMNJ co-complex and the 3S1 conformation of the ERManI-thiodisaccharide co-complex described here.

Class 1 (GH Family 47) �-1,2-Mannosidase Mechanism16198

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from

an unusual high free energy 1C4 conformation facilitated bothby a direct coordination of the inhibitor O-2� and O-3� hy-droxyls with an enzyme-associated Ca2� ion and by a matrixof hydrogen bonds and hydrophobic interactions. A novelinverting glycosidase mechanism was proposed where thewater nucleophile was one of the ligands directly coordinatedwith the enzyme-bound Ca2� ion, and the catalytic acid wasproposed to act indirectly in a through-water protonation ofthe glycosidic oxygen (20). The unusual nature of this pro-posed mechanism precluded an unambiguous identificationof the catalytic acid and base residues.

We describe here the kinetics and binding characteristics ofpotential catalytic mutants as well as a co-complex of ERManIwith a nonhydrolyzable thiodisaccharide pseudosubstrate.These data demonstrate the roles of various active site residuesas well as the novel conformational itinerary during glycosidebond hydrolysis.

MATERIALS AND METHODS

Synthesis of Methyl-2-S-(�-D-mannopyranosyl)-2-thio-�-D-mannopy-ranoside—The previous reported synthesis of the thiodisaccharide, meth-yl-2-S-(�-D-mannopyranosyl)-2-thio-�-D-mannopyranoside, was achievedusing the coupling reaction of methyl 4,6-O-benzylidene-2-O-triflu-oromethanesulfonyl-�-D-mannopyranoside with the thiolate anion of2,3,4,6-tetra-O-acetyl-1-thio-�-D-mannopyranose (25). We chose an alter-nate strategy using the alternative key coupling reaction of methyl 3-O-benzyl-4,6-O-benzylidene-2-S-thioacetyl-�-D-mannopyranoside with

2,3,4,6-tetra-O-acetyl-D-mannopyranosyl bromide and characterized thesynthetic product by NMR as described in the Supplementary Data.

Mutagenesis, Expression, and Purification of Human ERManI—Theexpression and purification of the human ERManI catalytic domain hasbeen described previously (20, 26). The cDNA encoding ERManI in thepPICZ�A vector (Invitrogen) was used to perform site-directed mu-tagenesis using the QuikChange™ mutagenesis kit from Stratagene(La Jolla, CA) based on the sequence of human ERManI (GenBankTM

accession number AF145732). The full coding region of each mutantwas sequenced to confirm that only the desired mutation was gener-ated. The plasmid constructs were then used to transform the Pichiapastoris strain X-33, and zeocin-resistant colonies were screened forERManI expression by performing Western blots using conditionedmedium from induced cultures as previously described (20). ERManIwas detected using a rabbit polyclonal anti-human ERManI antibodygenerated to the recombinant wild type enzyme, and the antibody wasdetected using a peroxidase-conjugated anti-rabbit IgG secondary an-tibody. Mutant enzymes were expressed in 1-liter shake flask culturesby induction in BMMY media, and the enzyme was purified from theconditioned media as previously described (20).

Man9GlcNAc2-PA and Glycopeptide Preparation—Crude soybean ag-glutinin was extracted from soy flour by acid and ammonium sulfateprecipitation (27) and was subsequently denatured, reduced, and carboxy-amidomethylated in 8 M guanidine HCl containing 50 mM dithiothreitoland 100 mM iodoacetamide (28) prior to proceeding to elastase digestion(29). The Man9GlcNAc2-glycopeptide was recovered by affinity chroma-tography using concanavalin A-Sepharose and further purified by HPLCon a Cosmosil C18 column (30) for SPR studies. For enzyme assays,Man9GlcNAc2 was liberated from the peptide by peptide:N-glycosidase Fdigestion (31) and derivatized with pyridylamine (32).

Enzyme Assays—The purified wild type and mutant enzymes wereassayed for �1,2-mannosidase activity using pyridylamine-taggedMan9GlcNAc2 (Man9GlcNAc2-PA) as substrate (32). The enzyme reac-tions (20 �l) were carried out in 96-well plates by adding 10 �l ofenzyme in 300 mM NaCl and 10 mM CaCl2 to a mixture of 5 �l of 4�universal buffer (80 mM succinic acid, 80 mM MES, 80 mM HEPBS, 80mM HEPES, and 80 mM CHES adjusted to pH with 5 M NaOH) and 5 �lof substrate. The reactions were performed at 37 °C for the indicatedtimes and stopped by the addition of 20 �l of 1.25 M Tris-HCl (pH 7.6)to the reaction mixture. The enzymatic products were resolved andquantitated using a Hypersil APS-2 NH2-HPLC column (32). Initialrates (v) for the enzymes were determined at various substrate concen-trations ranging from 10 to 300 �M. The catalytic coefficient (kcat) andMichaelis constant (Km) values were determined by fitting initial ratesto a Michaelis-Menten function by nonlinear regression analysis usingSigmaPlot (Jandel Scientific, San Rafael, CA)). kcat/Km values werederived from reciprocal plots of v and [S] where needed.

pH Rate Dependence Analysis—Values for kcat and Km were deter-mined from initial rates of enzyme reactions in the pH range from 4 to10 using the 1� universal buffer described above. Plots of log(kcat/Km)versus pH were used to estimate values of the macroscopic enzymeionization constants (pKE1 and pKE2) by linear extrapolation of theacidic and basic limbs of the curve (Fig. 3).

Binding Studies by Surface Plasmon Resonance (SPR)—SPR analy-ses were conducted using a Biacore 3000 apparatus (Biacore AB, Pis-cataway, NJ). Recombinant ERManI was immobilized on the SPR chipsurfaces at 25 °C by the amine coupling method (34). The flow cells wereactivated by injecting a mixture of 50 mM N-hydroxysuccinimide and200 mM 1-ethyl-3-(dimethylaminopropyl)carbodiimide over the CM5sensor chip surface for 7 min at 5 �l/min. The recombinant proteins (30�g/ml) prepared in 10 mM sodium succinic acid (pH 6.0) were passedthrough a 0.2-�m polyvinylidene difluoride filter (Millipore Corp.) anddiluted in the same buffer to obtain a concentration of 5 �g/ml prior toinjection onto the activated surface. The desired immobilization levelwas achieved by specific contact time. The remaining reactive groupswere blocked by injection of 1 M ethanolamine-HCl at pH 8.5 for 7 minat 5 �l/min. The immobilization efficiency for ERManI was about 2500RU/min at a flow rate of 5 �l/min in HPB-EP buffer (10 mM HEPES, pH7.4, 150 mM NaCl, 3.4 mM EDTA, 0.01% polysorbate P20) at 25 °C.Mock derivatized flow cells served as reference surfaces. The bindinganalyses were performed at 10 °C with a continuous flow (30 �l/min) ofrunning buffer (10 mM MES, pH 7.0, 300 mM NaCl, and 5 mM CaCl2).Analytes were prepared in running buffer by 2-fold serial dilution toobtain an appropriate concentration range. A concentration series ofMan9GlcNAc2-glycopeptide (0.4–400 �M) was analyzed over a low den-sity immobilization surface of recombinant protein (3000 RU), whereasdisaccharides (16–1000 �M) and dMNJ (2–1000 �M) were analyzed overa high density immobilization surface of recombinant protein (10,000

FIG. 2. Model for the structure and catalytic residues for hu-man ERManI used in mutagenesis studies described in thispaper. The structure of the co-complex between human ERManI anddMNJ (PDB 1FO2 (20)) was used to select potential residues involved incatalysis. The end (A) and side (B) views of the human ERManI ribbondiagram display the (��)7 barrel structure with the inhibitor (dMNJ;stick representation) bound in the core of the barrel directly coordinatedto the protein-bound Ca2� ion (blue space fill) as highlighted by the bluecircle. The residues examined in this study are shown in the stereodiagram (C), where the stick representation of dMNJ is shown in yellow,the Ca2� ion is shown as blue space fill, and the relevant residuesdescribed throughout are shown as white stick diagrams. The small redspace fill structures connected through cyan dotted lines to the Ca2� ionare water molecules that coordinate Ca2� ion. The small green space fillstructures connected through cyan dotted lines to the Ca2� ion representthe carbonyl oxygen and O-� of Thr688, the only protein residue directlycoordinated to the Ca2� ion. Proposed acid, base, and nucleophile tra-jectories as described throughout are illustrated with magenta dottedlines. Hydrogen bonds between Glu330 and Arg334 as well as Glu599 andHis524 are shown as green dotted lines.

Class 1 (GH Family 47) �-1,2-Mannosidase Mechanism 16199

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from

RU). Samples were injected for 1 min followed by at least 3 min ofdissociation phase. The base line returned to the original responselevel in 5 min for all analytes described here without a furtherregeneration of the chip surface. SPR data for each concentration ofanalyte were collected in duplicate and globally fit to a 1:1 Langmuirbinding algorithm model to calculate the on-rate (ka), the off-rate (kd),and the equilibrium dissociation constant (kd/ka � KD) using Biae-valuation version 3.1 software (35). Alternatively, the maximal equi-librium sensorgram values were used to plot a saturation bindingcurve and calculate values for the equilibrium dissociation constant(KD) directly.

Crystallization and X-ray Diffraction—In an effort to obtain higherresolution crystal structure data than was previously obtained for hu-man ERManI (20), a sparse matrix screen (36) was performed with theCrystal Screen HT™ from Hampton Research, carried out using anOryx 6 crystallization robot (Douglas Instruments) in Nunc HLA plates.The crystallization drop was prepared by mixing equal volumes (0.5 �l)of the protein and the screening solutions (final protein concentration10 mg/ml) and sealed with paraffin oil (15 �l). The plate was thenoverlaid with a mixture of silicone and paraffin oils (30:70) (37, 38) andkept at 18 °C. Crystal formation was monitored after 3 days. Additivescreening was carried out by mixing equal volumes (0.5 �l) of theprotein solution, the optimized crystallization solution (24% (w/v) pol-yethylene glycol 4000, 100 mM MES/NaOH, pH 6.0, and 50 mM ammo-nium sulfate), and 3� additive screen components (Hampton Research,catalog number HR2-428). The addition of 1,4-butanediol was found toproduce the best crystals and improved diffraction to beyond 1.35 Å forthe native enzyme (data not shown). Since we had a limited supply ofthe thiodisaccharide substrate analogue, direct co-crystallization wasattempted using the above conditions without further optimization.

For co-crystallization, the thiodisaccharide (100 mM in H2O) wasadded to the protein solution to yield 10 mg/ml recombinant protein, 20mM MES, pH 7.0, 150 mM NaCl, 5 mM CaCl2, 750 mM NDSB 201, and50 mM thiodisaccharide and incubated for 2 h at room temperatureprior to crystallization. A hanging drop was prepared over 1 ml ofmother liquor containing 24% (w/v) polyethylene glycol 4000, 100 mM

MES/NaOH, pH 6.0, 50 mM ammonium sulfate, and 10% (v/v) 1,4-butanediol. The crystallization drop contained 1.5 �l of protein/ligandsolution and 1 �l of mother liquor. Crystals formed within 4 h at 25 °Cand were mounted and flash-frozen 1 day later using 10% glycerol ascryoprotectant. Although the single co-crystals grown by this methodwere considerably smaller in size than that of the native crystals, one ofthe crystals produced diffraction data that could be scaled to 1.41 Å withgood statistics and completeness (Table III).

Crystals were examined using an ACTOR/Director automated crys-tal-screening system (Rigaku/MSC) consisting of an FR-D high inten-sity rotating anode, a Max-Flux confocal optics, an AFC-9 four-axisgoniometer, and a Saturn 92 CCD detector. Complete diffraction datawere collected at the Advanced Photon Source, beam line 22ID, using aMARCCD-225 detector and a crystal-to-detector distance of 130 mm. A

total of 360 1° oscillation images were recorded. Integration and scalingwere performed with the HKL software suite (39). The structure wassolved by molecular replacement using the program EPMR (40) andProtein Data Bank entry 1FMI (20) as a search model. Rebuilding (Xfit(41)) was repeatedly iterated with refinement (Refmac5 (42–44)) andvalidation (MOLPROBITY (45)). Diffraction analysis indicated thatboth native and co-crystals belong to space group P1 in contrast to theprior crystal form (P3121) for native and inhibitor-bound human ER-ManI (20). Matthews coefficients (46) and structure determinationindicated a single molecule per asymmetric unit. All of the unit cellconstants between the co-crystal and the native enzyme were within0.5 Å or 0.4° with the exception of angle �, where the co-crystal wascalculated to be significantly larger than that of the native crystal(89.5° for the co-crystal and 66.2° for the native crystal). Although thedifference in the unit cell constant suggests a significant impact of thepresence of thiodisaccharide on crystal packing, there were no signif-icant differences in overall protein structure between native crystal(data not shown), co-complex structures, or the prior crystal forms(20).

For the calculation of the Fo � Fc electron density map, inhibitoratoms were deleted from the model. Remaining model atoms weredisplaced randomly by an average 0.3 Å, and their temperature factorswere randomly altered by an average of 3 Å2 with the program Mole-man (47). Amplitudes and phases for the weighted difference map werecalculated after 10 cycles of REFMAC5 (43) refinement. Protein andglycan molecular structure figures were prepared using MacPymol(version 0.95; available on the World Wide Web at www.pymol.org) togenerate rasterized images.

RESULTS

Kinetic Analysis of ERManI Mutants—Mutations were gen-erated in three of the residues, Asp463, Glu330, and Glu599, thathave been previously implicated as the potential general acidand general base residues in the catalytic mechanism (Fig. 2)(20, 48). Two of these residues are adjacent to charged aminoacid side chains that may influence the ionization state of therespective catalytic residue (Arg334 with Glu330 and His524 withGlu599) (Fig. 2). Each of these residues was individually alteredby mutagenesis (D463N, E330Q, E599Q, R334A, and H524A)along with some of the combinations of double mutants andexpressed in P. pastoris as secreted catalytic domains (19, 20).Expression of R334A was not detected in the culture media andwas not characterized further.

All of the mutants had diminished kcat values relative to wildtype ERManI (Table I), but the mutations had a relativelyminor impact on Km (variation of �2.4-fold). The H524A mu-tant showed the least change in kcat (4.3-fold reduction),

FIG. 3. pH rate analysis for wild type and the E330Q mutant of ERManI. Kinetic analyses of wild type (open squares) and the E330Qmutant (open diamonds) of ERManI were performed using Man9GlcNAc2-PA as substrate and determining initial rates of hydrolysis byNH2-HPLC. Values for Km and kcat were determined as described under “Materials and Methods” at 0.5-pH unit intervals and plotted aslog(kcat/Km) versus pH (solid line). An estimation of the enzyme ionization constants (KE1 and KE2, upper left panel) was performed by graphicalextrapolation as indicated by the dotted lines in the plot in the right panel. The derived enzyme ionization constants are shown in the table at thebottom left.

Class 1 (GH Family 47) �-1,2-Mannosidase Mechanism16200

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from

whereas other mutants varied from a 44-fold reduction(E330Q) to a 105-fold reduction (E599Q) (Table I). Double mu-tants of E599Q with either E330Q or D463N showed significantreductions in kcat and slight increases in Km consistent with thedominant effect of the E599Q single mutant. pH rate analysisindicated that there was no significant change in pH optimum(�0.5 pH units) for any of the mutants except for E330Q, whichdisplayed a reduction in pH optimum of �1.8 units (Table I). Aplot of log(kcat/Km) versus pH confirmed that the E330Q mutanthad a major shift of �1.8 pH units for the peak value in the plot(Fig. 3). Linear approximations for the enzyme ionization con-stants (pKE values) for the acidic and basic limbs of the curveindicated that both values decreased in the E330Q mutant,with a 0.8-pH unit drop on the acidic limb and a 1.3-pH unitdrop on the basic limb. These data indicate that the mutationcauses a global change in the ionization state in the active site,influencing both the acidic and basic limbs of the pH curve. Aneffect on the basic limb of the curve would be consistent with aninfluence on the ionization of a potential general acid as aproton donor in the catalytic mechanism.

Glycan Binding Affinity Measurements to Human ER-ManI—In addition to kinetic analysis, the binding affinitiesbetween Man-�1,2-Man-containing glycans and recombinantwild type and mutant forms of ERManI were examined by SPR,including high mannose oligosaccharides, a nonhydrolyzablethiodisaccharide analog, and an equivalent O-disaccharide.The on-rates (ka) and off-rates (kd) for binding of Man9GlcNAc2-glycopeptide to the wild type and E599Q mutant were too fastfor accurate measurement by SPR, but the equilibrium disso-ciation constant (KD) could be derived from the equilibriumvalues of the binding sensorgrams (Fig. 4, inset plots). In con-trast, the E330Q mutation displayed significantly slowed on-and off-rates for all of the ligands, allowing the measurement ofka and kd and thus the equilibrium dissociation constant (KD �kd/ka) from both the kinetic and equilibrium binding data. Theresulting data (Table II) indicate that the E330Q mutant hadan �100-fold increase in binding affinity for the Man9GlcNAc2-glycopeptide ligand with a virtually unchanged affinity fordMNJ. These values, in combination with slowed on- and off-rates and the decreased kcat observed for the E330Q mutantindicate that the replacement of this carboxyl side chain withan uncharged isosteric amide causes a significantly reducedcatalytic turnover, slower binding kinetics, and higher bindingaffinity for the substrate without altering the affinity for theglycone residue. In contrast, the E599Q mutant had a 4-foldweaker binding affinity than the wild type enzyme forMan9GlcNAc2-glycopeptide and an �1000-fold reduction inbinding affinity to dMNJ (Fig. 4 and Table II), indicating thatthe mutation results in both a significant drop in kcat and anunfavorable binding of the glycone residue.

The D463N mutant revealed a surprising difference in bind-ing characteristics between the Man9GlcNAc2-glycopeptide

and dMNJ ligands. The binding affinity of dMNJ for the D463Nmutant was reduced �73-fold (Table II), but binding to theMan9GlcNAc2-glycopeptide (Fig. 4) as well as the disaccharideligands (data not shown) was virtually eliminated. These dataare consistent with a role for this latter residue in binding thesubstrate in the �1 subsite as supported by the ERManI-thiodisaccharide co-crystal structure described below.

Disaccharide Binding to Human ERManI—Methyl-2-S-(�-D-mannopyranosyl)-2-thio-�-D-mannopyranoside, a thiodisaccha-ride analog of Man-�1,2-Man-O-CH3 was synthesized by amodification of a previously published method (25), and itsstructure was confirmed by NMR spectroscopy (see Supple-mentary Data). Proton scalar couplings confirmed that bothpyranose rings were in the conventional 4C1 conformation insolution, and qualitative analysis of NOEs and carbon-protonscalar couplings showed that the glycosidic bond adopts anglessimilar to an equivalent O-glycoside (see Supplementary Data).The on- and off-rates of the thiodisaccharide and the equivalentO-glycoside for binding to wild type human ERManI (ka and kd

values in Table II) or the E330Q mutant (data not shown) weresignificantly slower than the Man9GlcNAc2-glycopeptide, andboth displayed lower equilibrium binding affinities. The on-and off-rates were significantly different for the two disaccha-rides, but the equilibrium dissociation constant (KD) was only�5-fold lower for the thiodisaccharide versus the O-glycoside.Similarly, the IC50 for inhibition of human ERManI by thethiodisaccharide (�200 �M) was lower than the Km for theO-glycoside substrate (�2 mM). These data, in combinationwith the NMR data above, indicate that the solution behaviorand binding of the thiodisaccharide to ERManI is likely to besimilar to those of the O-glycoside substrate.

Co-complex Structure of ERManI with the ThiodisaccharideAnalog—Co-crystallization of recombinant human ERManIwith the thiodisaccharide analog resulted in crystals that dif-fracted beyond 1.4 Å (Table III), and the structure of the co-complex was solved by molecular replacement using the struc-ture of human ERManI (Protein Data Bank code 1FMI (20)) asthe probe. The structure revealed the anticipated (��)7 barrelcatalytic domain (Fig. 2) (20) with an extra density correspond-ing to the thiodisaccharide in the proposed �1 and �1 subsites(Fig. 5A). As anticipated, the C–S bond lengths for the thiodi-saccharide were both �1.8 Å, significantly greater than thestandard �1.4 Å C–O bonds in a glycosidic linkage. Modeling ofMan-�-O-CH3 in the �1 subsite revealed an unambiguous 4C1

conformation, whereas the conformation of the sugar occupyingthe �1 subsite was clearly distorted (Fig. 5, A and B). Thehydroxyls associated with C-3 and C-4 as well as the C-5–C-6linkage were in axial orientations, and the O-2� hydroxyl wasequatorial, representing identical positions compared with the1C4 conformation adopted by dMNJ when bound in the �1subsite (20). However, the positions of C-1, O-5, and C-5 of the�1 residue were clearly distorted from a 1C4 chair into a 3S1

TABLE IKinetic constants for wild type and mutant human ERManI using Man9GlcNAc2 as substrate

pH optimuma Km kcat kcat/Km (kcat/Km)/(kcat/Km(wt))

pH �M s�1 � 103 s�1/M %

Wild type 7.1 110 � 8.0 3,700 � 110 34,000 � 2,600 100E330Qb 5.3 68 � 1 84 � 3 1,200 � 48 3.5

(7.1) (90 � 7) (18 � 3) (200 � 37) (0.60)D463N 6.8 110 � 1 4.6 � 0.2 42 � 2 0.1E599Q 6.7 160 � 22 0.026 � 0.003 0.16 � 0.03 0.0005E330Q/E599Q 6.6 150 � 10 0.28 � 0.01 1.9 � 0.1 0.006D463N/E599Q 6.6 260 � 28 0.027 � 0.003 0.1 � 0.02 0.0003H524A 6.9 210 � 21 860 � 45 4,100 � 460 12

a Assay data were fit to generate a bell-shaped curve to yield a pH optimum with an S.E. value of less than 0.1 pH unit.b The pH optimum of the E330Q mutant is 5.3, but the kinetic constants for the enzyme assayed at pH 7.1 (pH optimum for wild type enzyme)

are also shown in parentheses.

Class 1 (GH Family 47) �-1,2-Mannosidase Mechanism 16201

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from

skew boat conformation (Fig. 5B). The altered position of C-5resulted in a corresponding change in the position of C-6;however, a rotation of the C-5–C-6 bond resulted in a nearlyidentical position of the O-6� hydroxyl for hydrogen bonding toO-�1 of Glu599 and NH1 of Arg597. Thus, C-1, O-5, C-5, and C-6are the only atoms that were significantly different in positionfrom the 1C4 chair of the dMNJ co-complex structure (20), andnone of these atoms directly interact with the enzyme in the �1

subsite. Comparison of the 3S1 and 1C4 conformers reveals thattheir interconversion results from a twist at C-5, the O-5 ringoxygen, and C-1, and the 3H4 conformation is intermediatebetween the two conformers (Fig. 6). The movements for C-5,O-5, and C-1 would be �0.4, �1, and �0.3 Å, respectively,during this conformational switch.

Interactions between ERManI and the Thiodisaccharide Sub-strate Analog—Interactions between the nonreducing terminal

FIG. 4. SPR binding ofMan9GlcNAc2-glycopeptide or dMNJligands to wild type and mutant ER-ManI. Wild type (WT) or mutant ERManIwas immobilized on the SPR chip surfaceas described under “Materials and Meth-ods,” and various concentrations of eitherMan9GlcNAc2-glycopeptide or dMNJ li-gands were tested for binding. The datawere collected in duplicate, and repre-sentative SPR sensorgrams in the ligandconcentration series are shown. For thebinding studies with dMNJ, the on- andoff-rates (ka and kd, respectively) weresufficiently slow to allow curve fitting ofthe sensorgrams using the 1:1 Langmuirbinding algorithm model to determinevalues for the equilibrium dissociationconstants (KD � kd/ka). In binding studieswith wild type ERManI and the E599Qmutant, the kinetics of binding for theMan9GlcNAc2-glycopeptide ligand weretoo rapid for curve fitting, so the equilib-rium sensorgram values were used to plota saturation curve (insets in the upperand lower left plots) and calculate valuesfor KD. The E330Q mutant significantlyslowed the kinetics of binding of theMan9GlcNAc2-glycopeptide ligand, andthe data were fit with the 1:1 Langmuirbinding algorithm model in this instance.In contrast, the D463N mutant com-pletely abolished binding of theMan9GlcNAc2-glycopeptide ligand as in-dicated by the absence of a deflection inthe SPR sensorgram trace. The values forka, kd, and KD, based on the SPR studies,are shown in Table II.

TABLE IISummary of the SPR binding affinity data for wild type and mutant forms of human ERManI

Ligand Protein Fitting typea ka kd KD KD/KD(wt)

s�1M

�1 � 1000 s�1 � 0.001 �M -fold

dMNJ Wild type 1:1 2.94 � 0.73 4.74 � 1.47 1.61 � 0.64 1E330Q 1:1 1.06 � 0.56 2.62 � 1.63 2.47 � 2.02 1.5D463N 1:1 0.33 � 0.15 38.3 � 9.0 117 � 62 73E599Q 1:1 0.033 � 0.004 56.8 � 12.2 1730 � 440 1080

Man9GlcNAc2 glycopeptide Wild type SS 48.6 � 1.7 1E330Q 1:1 21.9 � 4.9 11.1 � 2.7 0.51 � 0.17 0.01D463N 1:1 NDb

E599Q SS 184 � 22 4O-Disaccharide Wild type 1:1 0.050 � 0.004 27.9 � 6.2 563 � 131Thiodisaccharide Wild type 1:1 0.71 � 0.03 78.3 � 6.2 110 � 10

a 1:1 refers to a 1:1 nonlinear Langmuir fit of the binding data to derive the ka, kd, and KD values from the kinetic SPR binding sensorgrams asdescribed under “Materials and Methods.” SS refers to a steady-state fit of the binding data from maximal equilibrium values of the SPRsensorgrams as described under “Materials and Methods.” For examples of each type of data fitting, see Fig. 4.

b ND, no detectable binding.

Class 1 (GH Family 47) �-1,2-Mannosidase Mechanism16202

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from

residue in the thiodisaccharide and the �1 subsite residues areidentical to those stabilizing the ERManI�dMNJ co-complex(20) and include direct coordination of the Ca2� ion to the O-2�and O-3� hydroxyls (Table IV and Fig. 7), hydrogen bonding tothe O-4� and O-6� hydroxyls, and a hydrophobic stacking ofPhe659 adjacent to the C-4–C-5–C-6 region of the �1 residue(Fig. 7A). The predominant interactions with the �1 residueare a pair of hydrogen bonds between the two carboxylateoxygens of Asp463 and the O-3� and O-4� hydroxyls of the �1sugar residue. Additional hydrogen bonds were found betweenthe enzyme and the O-4� and the O-6� hydroxyls of the �1residue.

Acid/Base Chemistry in the Active Site of GH Family 47Mannosidases—The positions of the amino acid side chains inboth the �1 and �1 subsites of ERManI are not significantlyaltered as a consequence of binding the thiodisaccharide. Theposition of the sulfur in the thiodisaccharide is �4.2 Å fromO-�2 of Glu330, indicating, as previously proposed (20, 48), thatthere is probably not a direct protonation of the glycosidicoxygen by this enzyme carboxyl side chain. Both Glu330 andArg334 are in positions similar to those in the ERManI�dMNJcomplex (20). A weak electron density of a proposed watermolecule, W8, is in proximity to the glycosidic sulfur (3.2 Å) andwithin hydrogen bonding distance from both carboxyl oxygensof Glu330 (2.6 and 3.0 Å, respectively) as well as NH2 of Arg334

(2.7 Å) (Table IV and Fig. 7). A direct interaction betweenGlu330 and Arg334 is also indicated by the proximity of O-�1 ofGlu330 to both NH2 (3.0 Å) and N-� (2.9 Å) of Arg334 (Fig. 7 andTable IV).

The positions of the proposed catalytic base, Glu599, and thewater nucleophile, W5, as well as the protein-associated Ca2�

ion are essentially unchanged from the free enzyme or theERManI�dMNJ co-complex (Table IV) (20). The distances fromGlu599 to W5 (2.7 Å) and from W5 to C-1 of the sugar in the �1site (3.2 Å) would indicate that the water nucleophile could beactivated by the Glu599 base for nucleophilic attack. The dis-tance from W5 to the C-1 of the glycone would be expected to beeven less at the 3H4 transition state conformation.

DISCUSSION

Class 1 �-mannosidases (family 47 glycosidases (56)) are keyenzymes in the maturation of N-glycans in the secretory path-way as well as playing key roles in defining the timing andrecognition for disposal of misfolded proteins by ER-associateddegradation (2, 4–8, 49, 50). Previous structural studies on the

co-complexes between human ERManI and the inhibitors,dMNJ and kifunensine, implicated three residues in proximityto the glycosidic bond, Asp463, Glu330, and Glu599, as potentialcatalytic residues (Fig. 2) (20). In addition, computational dock-ing studies of monosaccharides and disaccharides in the activesite of S. cerevisiae ERManI have suggested that the equiva-lents of Glu330 and Glu599 act as the catalytic acid and baseresidues, respectively, and that an E4 conformation is the likelytransition state (48). The mechanism proposed from thesestructural studies included several elements not common forinverting glycosidases, including the involvement of an un-usual high free energy intermediate adopting a 1C4 conforma-tion, a through-water protonation of the glycosidic oxygen, anda water nucleophile that was also directly coordinated to anenzyme-associated Ca2� ion (20). In an effort to understand themechanism for glycoside bond hydrolysis and possibly gaininsights into how the EDEM proteins are compromised in gly-cosidase activity, we generated mutations in the potential cat-alytic residues of human ERManI and examined their effectson substrate binding and catalysis. In parallel studies, we alsoexamined the structure of a co-complex between human ER-ManI and an uncleavable thiodisaccharide substrate analog.These studies have yielded insights into the roles of active siteresidues and the dynamics of sugar conformational changesduring glycoside bond hydrolysis.

Mutation of Glu330 to an isosteric amide (E330Q) resulted ina significant decrease in kcat and a slight drop in Km similar tothe effects seen in an equivalent mutant of S. cerevisiae ER-ManI (24). Further analysis revealed a �1.8-pH unit drop inthe pH optimum and a drop in the calculated pKE values forboth the acidic and basic limbs of the pH curve; however, theshifts in pKE values were more significant on the basic side ofthe curve. These data do not necessarily confirm a role of Glu330

as the general acid in the catalytic mechanism, since the re-moval of the general acid might have been predicted to result ina grater loss in kcat and predominately influence the pKE of thebasic limb of the pH curve by the loss of the ionizable residue.However, there are several nonstandard features of the generalacid function for the GH47 �-mannosidases. First, the distancebetween the Glu330 carboxyl side chain and the sulfur in thethiodisaccharide co-complex is far too large (�4.2 Å) to accountfor direct protonation of the glycosidic oxygen of the substrateby the carboxyl side chain as is standard in both inverting andretaining glycosidases. Second, the presence of a weak density

TABLE IIIData collection and refinement statistics

Parameter Valuea

Data resolution (Å) 48.80 to 1.41 (1.45 to 1.41)Unique observed reflections 84,600 (8,225)Completeness (%) 95.4 (92.6)I/(�I) 41.8 (13.0)Average redundancy 4.0 (3.5)Space group P1Unit cell lengths a, b, c (Å) 50.7, 53.9, 56.2Unit cell angles �, �, � (degrees) 89.5, 63.6, 62.6Matthews coefficient (Å3�Da�1) 2.4Solvent content (%, v/w) 47Mosaicity (degrees) 0.1–0.3No. of unique reflections used in refinement 76,092 (5,417)No. of reflection in free set 8,507 (631)Rcryst (%) 14.4 (15.8)Rfree (%) 16.2 (19.3)No. of nonhydrogen atoms refined 4,097Root mean square deviation (Å) bonds 0.014Root mean square deviation (degrees) angles 1.34Mean temperature factor (Å2) 8.4Wilson temperature factor (Å2) 9.3

a Values in parentheses refer to data from the high resolution shell.

Class 1 (GH Family 47) �-1,2-Mannosidase Mechanism 16203

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from

for an intervening water molecule, W8, as the only atom inproximity to the glycosidic oxygen would suggest that anygeneral acid function would act indirectly in a through-waterprotonation mechanism. Third, Arg334 is within ion-pairingdistance from Glu330, supporting an interaction that shouldlower the pKa of Glu330 rather than raise it to a value of �7.8as predicted from the pH rate studies (Fig. 3). Fourth, bothGlu330 and Arg334 are within hydrogen bonding distances of W8

(Fig. 7 and Table IV) and are the only ionizable residues inproximity to this intervening water molecule that can contrib-ute to the protonation of the glycosidic oxygen. A number ofstudies have suggested that Arg-carboxylate dyads can havesignificant deviations from predicted pKa values and play un-predicted roles in acid-base chemistry in catalysis (for a review,see Ref. 51). The lack of any other candidate residues withineither direct or indirect proximity of the sulfur in the thiodis-accharide co-complex structure would suggest that the Glu330-Arg334-W8 complex contributes to the general acid function inthe catalytic mechanism. An attempt to probe the role of Arg334

for its influence on catalysis by mutagenesis (R334A) was in-conclusive, but the high calculated pKa for the catalytic acid(pH �7.8) is likely to reflect the local environment of theGlu330-Arg334-W8 complex.

Mutation of Glu599 to the isosteric amide (E599Q) resulted ina much more profound drop in kcat (�105-fold) than the E330Qmutant and a �1000-fold loss in affinity for binding todMNJ. Surprisingly, the drop in affinity for binding of theMan9GlcNAc2-glycopeptide substrate by SPR was only 4-fold,suggesting that the loss in catalytic turnover due to the ab-sence of the catalytic base arose predominately from an unfa-vorable binding of the glycone residue in the �1 subsite ratherthan the docking of the ��1 residues in the glycan bindingcleft. Mutagenesis of the His residue adjacent to the Glu599

catalytic base (H524A) reduced the kcat value by �4-fold andslightly increased the Km, but the effects were relatively minorby comparison with those obtained for the conversion of thecatalytic base to the isosteric amide. These data would indicatethat Glu599 acts as the catalytic base and that there is not asignificant role of His524 in the catalytic mechanism ofERManI.

The role of Asp463 in the ERManI mechanism was tested byexamining the kinetics and substrate binding of an isostericamide mutant of the enzyme (D463N) and by examining thenature of the interactions between this residue and the sub-strate analog in the thiodisaccharide co-complex. The D463Nmutant resulted in an �800-fold drop in kcat, and binding of theMan9GlcNAc2-glycopeptide substrate to the enzyme was virtu-ally eliminated as shown by SPR (Fig. 4). In the co-crystal,Asp463 was clearly seen to coordinate the �1 residue by inter-acting with both the O-3� and O-4� hydroxyls of this residue(Fig. 7). We anticipated that the isosteric amide substitution atthis position (D463N) would allow hydrogen bonding capabilityto be retained while eliminating its potential role as a catalyticacid. Since this residue is clearly involved in stabilizing theinteraction with the �1 residue and does not act as a generalacid, we were surprised that the mutation would have such aprofound effect on substrate binding. Precedent for a signifi-cant decrease in bond length and increase in hydrogen bondstrength for carboxylate side chains over isosteric amides hasbeen previously noted (52). The critical role of the proposedhigher affinity interactions for the carboxylate over the amidewould be strongly supported by a significant drop in kcat andthe total loss of measurable binding affinity for high mannoseoligosaccharides in the D463N mutant.

Standard glycosidase mechanisms invoke a glycone interme-diate containing a partial positive charge delocalized betweenthe anomeric carbon and the ring oxygen at or near the tran-sition state. The partial double bond character requires co-planarity of the C-2–C-1–O-5–C-5 linkages (Fig. 1). Of the fourpotential sugar ring conformers that satisfy these criteria (Fig.1), only the 3H4 conformer had not yet been identified as apotential glycosidase transition state conformation (14). Theunusual 1C4 conformation adopted by dMNJ in theERManI�dMNJ co-complex has led to the suggestion that

FIG. 5. Normalized Fo � Fc electron density map for the thio-disaccharide in the active site of ERManI and comparison of thesugar ring conformations with the enzyme-bound conformationof dMNJ in the �1 subsite and the M7 mannose residue in the �1subsite. A, a stereographic representation of the difference electrondensity for the omitted inhibitor in the ERManI-thiodisaccharide co-complex. The inhibitor model is shown to aid in map interpretation. Thereducing terminal Man-�-O-CH3 is shown at the top, labeled as the �1subsite residue, and the nonreducing terminal Man residue in the 3S1conformation is labeled as the �1 subsite residue. Carbon and sulfuratoms in the structures are labeled as a reference. The electron densitymap was contoured at 3� for the gray mesh and 10� for the red mesh,demonstrating the significant electron density at the glycosidic sulfur,the O-3� and O-4� hydroxyls of the �1 residue, and the O-2�, O-3�, andO-4� hydroxyls of the �1 residue. B, the protein structure of the ER-ManI-thiodisaccharide co-complex was aligned with the correspondingprotein structures of the ERManI�dMNJ co-complex (20) and the co-complex of yeast ERManI containing a Man5GlcNAc2 glycan in theactive site (21) using Swiss-PdbViewer (version 3.7) (55). Displayed inthe figure are the structures of the thiodisaccharide (yellow stick figure),dMNJ (green stick figure), and the M7 residue of the Man5GlcNAc2glycan in the �1 subsite (white stick figure; see Ref. 19 for oligosaccha-ride residue nomenclature). Carbon and sulfur atoms in the structuresare labeled as a reference. The M7 residue is in an identical conforma-tion as the �1 residue of the thiodisaccharide and in a similar positionexcept for an offset of 0.5–0.7 Å resulting from the longer C–S bondlengths of the thiodisaccharide. The positioning of the �1 subsite res-idues (dMNJ versus the �1 residue of the thiodisaccharide) were vir-tually identical at the C-2, C-3, and C-4 positions. The main differencesbetween the two structures were found in the equivalent of the C-1, O-5,C-5, and C-6 positions.

Class 1 (GH Family 47) �-1,2-Mannosidase Mechanism16204

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from

the family 47 glycosidases may employ a 3H4 transition state(14). Evidence for this proposed transition state is providedhere with the structure determination of the enzyme-boundthiodisaccharide pseudosubstrate which has been unambigu-ously shown to adopt a 3S1 skew boat conformation for theglycone in the �1 subsite. This conformation is quite differentfrom that adopted in solution by unbound O-glycoside sub-strates or the thiodisaccharide analog. In the itinerary of sugar

ring conformational interconversions, the 1C4 chair and 3S1

skew boat conformations bracket the 3H4 half-chair conforma-tion (Figs. 1 and 6).

The co-complex structure described here demonstrates thatthe GH47 mannosidases bind to substrates constrained at the �1residue into axial configurations for the C-3� and C-4� hydroxyls,and the C-5–C-6 linkage (Fig. 6) reflecting a 3S1 conformation.Substrate binding in this high energy conformation is mediatedby direct interaction of the O-2� and O-3� hydroxyls of the �1residue with the enzyme-bound Ca2� ion in combination with amatrix of hydrogen bonding interactions with the O-4� and O-6�sugar hydroxyls and a hydrophobic stacking of Phe659 with theC-4–C-5–C-6 region of the �1 residue (Fig. 7A). Most notably, thepresence of Phe659 at the base of the �1 subsite precludes thebinding of a glycone residue containing equatorial configurationsat C-4 and C-5 by steric hindrance. Additional interactions be-tween the �1 residue and Asp463 play a critical role contributingto substrate binding affinity and in positioning the glycosidiclinkage in proximity to the active site residues. Constraining the�1 residue into its high free energy 3S1 conformation would bepredicted to reduce the activation energy needed for transforma-tion of the substrate into the required 3H4 transition state, whichcould be attained through a conformational twist at C-5, O-5, andC-1. Through-water proton transfer from a general acid involvingGlu330, Arg334, and W8 to the glycosidic oxygen would result inglycoside bond cleavage with a simultaneous attack at C-1 by thewater nucleophile (deprotonated by the general base, Glu599) andthe formation of the exploded transition state with a 3H4 confor-mation (Fig. 6). Collapse of the transition state would lead to afinal product adopting a 1C4 chair conformation and an invertedanomeric configuration. In summary, these data are consistentwith a mechanism of glycoside bond cleavage in which a leastmotion conformational twist of a properly predisposed substrate

FIG. 6. Proposed mechanism for hu-man ERManI-catalyzed hydrolysis ofMan-�1,2-Man linkages based on ki-netic analysis and the substrate ana-log co-crystal structure. Substratebinding to the �1 subsite results in dis-tortion of the glycone into a 3S1 skew boatconformation by binding interactions inboth the �1 and �1 subsites (left panel).Through-water protonation of the glyco-sidic oxygen from the Glu330-Arg334-W8complex presumably acts as the catalyticacid, leading to glycosidic bond cleavage,and the simultaneous attack of C-1 by thebase-activated (Glu599) water nucleophileleads to the 3H4 exploded transition stateconformation (center panel, in brackets).The nature of the proton source for thegeneral acid catalyst within the Glu330-Arg334-W8 complex has not been resolved,as indicated by the question marks. Col-lapse of the transition state results in theformation of an enzymatic product in a1C4 conformation with an inverted ano-meric configuration (right panel). Stickfigure representations of the enzyme-bound thiodisaccharide �1 residue anddMNJ are shown at the bottom left andright, respectively, reflecting the pro-posed conformations of the bound sub-strate and product.

TABLE IVRelevant distance measurements for atoms in the human

ERManI-thiodisaccharide co-complex relative to the equivalent atomsin the human ERManI�dMNJ co-complex

Atomsa Thiodisaccharideco-complex

dMNJco-complex (20)

Å Å

Ca2�-ligand O-2� 2.4 2.7Ca2�-ligand O-3� 2.5 2.6Ca2�-Thr688 O-� 2.5 2.7Ca2�-Thr688 O 2.4 2.6Ca2�-W1 2.5 2.7Ca2�-W4 2.5 2.7Ca2�-W5 2.4 2.8Ca2�-W6 2.4 2.6W5-Glu599 O-�1 2.7 2.6W5-ligand C-1 3.2 3.0W8-Glu330 O-�2 2.6 2.8W8-thiodisaccharide S 3.2Glu330 O-�1-Arg334 NH2 3.0 2.8Glu330 O-�1-Arg334 N-� 2.9 2.9

a Atom and residue numbering is based on equivalent residues inboth the thiodisaccharide co-complex described here and the dMNJco-complex with human ERManI (Protein Data Bank code 1FO2 (20)).The ligand atoms refer to the equivalents of the O-2� and O-3� hydroxyloxygens for the thiodisaccharide and dMNJ co-complexes in the respec-tive columns.

Class 1 (GH Family 47) �-1,2-Mannosidase Mechanism 16205

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from

at the �1 subsite accounts for the formation of the novel explodedtransition state.

Modeling of the sequence of the non-catalytic family 47�-mannosidase EDEM homologs upon a template of the humanERManI structure, surprisingly, did not result in a readilydiscernable difference in positions of any of the key catalyticresidues (data not shown). Future studies on the nonhydrolyticEDEM proteins similar to those described here, which combineSPR binding studies with structural analysis of protein-glycancomplexes, will hopefully reveal the structural basis of glycanrecognition and ERAD targeting of unfolded glycoproteins bythese lectins without catalyzing glycoside bond hydrolysis.

Acknowledgments—We thank Linghzi Sun for expert technical helpwith synthesis of the thiodisaccharide and Dr. Esther Arranz-Plaza forrunning and interpreting NMR spectra.

REFERENCES

1. Kornfeld, R., and Kornfeld, S. (1985) Annu. Rev. Biochem. 54, 631–6642. Helenius, A., and Aebi, M. (2004) Annu. Rev. Biochem. 73, 1019–10493. Cabral, C. M., Liu, Y., and Sifers, R. N. (2001) Trends Biochem. Sci. 26,

619–6244. Wu, Y., Swulius, M. T., Moremen, K. W., and Sifers, R. N. (2003) Proc. Natl.

Acad. Sci. U. S. A. 100, 8229–82345. Molinari, M., Calanca, V., Galli, C., Lucca, P., and Paganetti, P. (2003) Science

299, 1397–14006. Oda, Y., Hosokawa, N., Wada, I., and Nagata, K. (2003) Science 299,

1394–13977. Hosokawa, N., Wada, I., Hasegawa, K., Yorihuzi, T., Tremblay, L. O., Hersco-

vics, A., and Nagata, K. (2001) EMBO Rep. 2, 415–4228. Mast, S. W., Diekman, K., Davis, A. W., Karaveg, K., Sifers, R. N., and

Moremen, K. W. (2005) Glycobiology, in press9. Moremen, K. (2000) in Oligosaccharides in Chemistry and Biology: A Compre-

hensive Handbook (Ernst, B., Hart, G., and Sinay, P., eds) Vol. II, pp.81–117, John Wiley and Sons, Inc., New York

10. Herscovics, A. (1999) Biochim. Biophys. Acta. 1426, 275–28511. Herscovics, A. (1999) in Comprehensive Natural Products Chemistry (Pinto,

B. M., ed) pp. 13–35, Elsevier, New York12. Rye, C. S., and Withers, S. G. (2000) Curr. Opin. Chem. Biol. 4, 573–58013. Zechel, D. L., and Withers, S. G. (2001) Curr. Opin. Chem. Biol. 5,

643–64914. Davies, G. J., Ducros, V. M., Varrot, A., and Zechel, D. L. (2003) Biochem. Soc.

Trans. 31, 523–52715. Guimaraes, B. G., Souchon, H., Lytle, B. L., Wu, J. H. D., and Alzari, P. M.

(2002) J. Mol. Biol. 320, 587–59616. Golan, G., Shallom, D., Teplitsky, A., Zaide, G., Shulami, S., Baasov, T.,

Stojanoff, V., Thompson, A., Shoham, Y., and Shoham, G. (2004) J. Biol.

Chem. 279, 3014–302417. Guerin, D. M., Lascombe, M. B., Costabel, M., Souchon, H., Lamzin, V.,

Beguin, P., and Alzari, P. M. (2002) J. Mol. Biol. 316, 1061–106918. Miyake, H., Kurisu, G., Kusunoki, M., Nishimura, S., Kitamura, S., and Nitta,

Y. (2003) Biochemistry 42, 5574–558119. Tempel, W., Karaveg, K., Liu, Z. J., Rose, J., Wang, B. C., and Moremen, K. W.

(2004) J. Biol. Chem. 279, 29774–2978620. Vallee, F., Karaveg, K., Herscovics, A., Moremen, K. W., and Howell, P. L.

(2000) J. Biol. Chem. 275, 41287–4129821. Vallee, F., Lipari, F., Yip, P., Sleno, B., Herscovics, A., and Howell, P. L. (2000)

EMBO J. 19, 581–58822. Van Petegem, F., Contreras, H., Contreras, R., and Van Beeumen, J. (2001) J.

Mol. Biol. 312, 157–16523. Lobsanov, Y. D., Vallee, F., Imberty, A., Yoshida, T., Yip, P., Herscovics, A.,

and Howell, P. L. (2002) J. Biol. Chem. 277, 5620–563024. Lipari, F., and Herscovics, A. (1999) Biochemistry 38, 1111–111825. Johnston, B. D., and Pinto, B. M. (1998) Carbohydr. Res. 310, 17–2526. Gonzalez, D. S., Karaveg, K., Vandersall-Nairn, A. S., Lal, A., and Moremen,

K. W. (1999) J. Biol. Chem. 274, 21375–2138627. Evers, D. L., Hung, R. L., Thomas, V. H., and Rice, K. G. (1998) Anal. Biochem.

265, 313–31628. Rice, K. G., Rao, N. B., and Lee, Y. C. (1990) Anal. Biochem. 184, 249–25829. Deras, I. L., Sano, M., Kato, I., and Lee, Y. C. (2000) Anal. Biochem. 278,

213–22030. Deras, I. L., Kawasaki, N., and Lee, Y. C. (1998) Carbohydr. Res. 306, 469–47131. Sutton, C. W., and O’Neill, J. A. (1997) Methods Mol. Biol. 64, 73–7932. Lal, A., Pang, P., Kalelkar, S., Romero, P. A., Herscovics, A., and Moremen,

K. W. (1998) Glycobiology 8, 981–99533. Cleland, W. W. (1979) Methods Enzymol. 63, 103–13834. O’Shannessy, D. J., Brigham-Burke, M., and Peck, K. (1992) Anal. Biochem.

205, 132–13635. Myszka, D. G. (1999) J. Mol. Recognit. 12, 279–28436. Jancarik, J., and Kim, S.-H. (1991) J. Appl. Crystallogr. 24, 409–41137. Chayen, N. E., Shaw Stewart, P. D., Maeder, D. L., and Blow, D. M. (1990)

J. Appl. Crystallogr. 23, 297–30238. D’Arcy, A., Elmore, C., Stihle, M., and Johnston, J. E. (1996) J. Cryst. Growth

168, 175–18039. Otwinowski, Z., and Minor, W. (1997) Methods Enzymol. 276, 307–32640. Kissinger, C. R., Gehlhaar, D. K., and Fogel, D. B. (1999) Acta. Crystallogr.

Sect. D Biol. Crystallogr. 55, 484–49141. McRee, D. E. (1999) J. Struct. Biol. 125, 156–16542. Potterton, E., Briggs, P., Turkenburg, M., and Dodson, E. (2003) Acta. Crys-

tallogr. Sect. D Biol. Crystallogr. 59, 1131–113743. Murshudov, G. N., Vagin, A. A., and Dodson, E. J. (1997) Acta Crystallogr.

Sect. D Biol. Crystallogr. 53, 240–25544. Winn, M. D. (2003) J. Synchrotron. Radiat. 10, 23–2545. Davis, I. W., Murray, L. W., Richardson, J. S., and Richardson, D. C. (2004)

Nucleic Acids Res. 32, W615–61946. Matthews, B. W. (1968) J. Mol. Biol. 33, 491–49747. Kleywegt, G. J., Zou, J. Y., Kjeldgaard, M., and Jones, T. A. (2001) in Crystal-

lography of Biological Macromolecules (Rossmann, M. G., and Arnold, E.,eds) pp. 353–367, Kluwer Academic Publishers, Dordrecht, The

FIG. 7. Interactions between the thiodisaccharide and the �1 and �1 binding sites in human ERManI. Shown is a schematic diagram(left panel) of the interactions between the thiodisaccharide and ERManI in the �1 and �1 subsites, demonstrating hydrogen bonding interactions(green dotted lines), direct coordination of the enzyme-associated Ca2� ion (blue dotted lines), hydrophobic stacking of Phe659 with the C-4–C-5–C-6region of the �1 residue (black dotted lines), and proposed acid-catalyzed through-water (W8) protonation of the glycosidic oxygen (sulfur in thethiodisaccharide) and the base-catalyzed (Glu599) attack by the water nucleophile (W5) (red dotted lines). Residue numbering of amino acid sidechains in the respective subsites is indicated. The stereo view (center and right) illustrates a stick diagram of the interaction between thethiodisaccharide residues in the �1 and �1 subsites relative to the residues examined by mutagenesis here. Coordination to the Ca2� ion (bluedotted lines), and the proposed nucleophile trajectory and acid protonation of the glycosidic oxygen (red dotted lines) are indicated. The small redand green space fill structures representing the water molecules and carbonyl oxygen and O-� of Thr688 that coordinate the Ca2� ion are asdescribed in the legend to Fig. 2. The green dotted lines indicate hydrogen bonds between the respective residues.

Class 1 (GH Family 47) �-1,2-Mannosidase Mechanism16206

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from

Netherlands48. Mulakala, C., and Reilly, P. J. (2002) Proteins 49, 125–13449. Hosokawa, N., Tremblay, L. O., You, Z., Herscovics, A., Wada, I., and Nagata,

K. (2003) J. Biol. Chem. 278, 26287–2629450. Yoshida, H., Matsui, T., Hosokawa, N., Kaufman, R. J., Nagata, K., and Mori,

K. (2003) Dev. Cell 4, 265–27151. Guillen Schlippe, Y. V., and Hedstrom, L. (2005) Arch. Biochem. Biophys. 433,

266–27852. Taylor, R., and Kennard, O. (1984) Acc. Chem. Res. 17, 320–32653. Stoddard, J. F. (1971) Stereochemistry of Carbohydrates, Wiley-Interscience,

New York54. IUPAC-IUB Commission (1980) Eur. J. Biochem. 111, 295–29855. Guex, N., and Peitsch, M. C. (1997) Electrophoresis 18, 2714–272356. Henrissat, B., and Bairoch, A. (1996) Biochem. J. 316, 695–696

Class 1 (GH Family 47) �-1,2-Mannosidase Mechanism 16207

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from

Bi-Cheng Wang and Kelley W. MoremenKhanita Karaveg, Aloysius Siriwardena, Wolfram Tempel, Zhi-Jie Liu, John Glushka,

-Glycan Processing and Endoplasmic Reticulum Quality ControlN-Mannosidases Involved in αMechanism of Class 1 (Glycosylhydrolase Family 47)

doi: 10.1074/jbc.M500119200 originally published online February 15, 20052005, 280:16197-16207.J. Biol. Chem. 

  10.1074/jbc.M500119200Access the most updated version of this article at doi:

 Alerts:

  When a correction for this article is posted• 

When this article is cited• 

to choose from all of JBC's e-mail alertsClick here

Supplemental material:

  http://www.jbc.org/content/suppl/2005/02/24/M500119200.DC1

  http://www.jbc.org/content/280/16/16197.full.html#ref-list-1

This article cites 52 references, 12 of which can be accessed free at

by guest on March 7, 2020

http://ww

w.jbc.org/

Dow

nloaded from