Energetics of Substrate Binding and Catalysis by Class 1 (Glycosylhydrolase Family 47) α-Mannosidases Involved in N-Glycan Processing and Endoplasmic Reticulum Quality Control*[boxs]

Nascent glycoproteins are subject to quality control in the lumen of the endoplasmic reticulum (ER) where they can either be effectively folded with the aid of a collection of ER chaperones or they can be targeted for disposal in a process known as ER-associated degradation. Initiation of the ER disposal process involves selective trimming of N-glycans by ER α-mannosidase I and subsequent recognition by the ER degradation-enhancing α-mannosidase-like protein family of lectins, both members of glycosylhydrolase family 47. The kinetics and energetics of substrate binding and catalysis by members of this family were investigated here by the analysis of wild type and mutant forms of human ER α-mannosidase I. The contributions of several amino acid residues and an enzyme-associated Ca2+ ion to substrate binding and catalysis were demonstrated by a combination of surface plasmon resonance and enzyme kinetic analyses. One mutant, E330Q, shown previously to alter general acid function within the catalytic site, resulted in an enzyme that possessed increased glycan binding affinity but compromised glycan hydrolysis. This mutant protein was used in a series of glycan binding studies with a library of mannose-containing ligands to examine the energetics of Man9GlcNAc2 substrate interactions. These studies provide a framework for understanding the nature of the unusual substrate interactions within the family 47 mannosidases involved in glycan maturation and ER-associated glycoprotein degradation.

As polypeptide chains are extruded through the endoplasmic reticulum (ER) 1 membrane during co-translational translocation, they are commonly glycosylated on the amide side chains of Asn residues within the acceptor consensus sequon, Asn-X-(Ser/ Thr) (1). Trimming of terminal glucose residues results in the formation of glycan structures that can act as ligands for the luminal ER lectin chaperones, calnexin and calreticulin (2,3), which can aid in the folding of the nascent polypeptides in the lumen of the ER (2,4). Glycoproteins that have slow folding kinetics continually re-engage the lectin chaperones either until folding is complete (2,4,5) or until the nascent glycoproteins acquire a target signal for disposal (6 -8). For terminally misfolded glycoproteins, trimming of the oligosaccharide by the action of ER ␣-mannosidase I (ERManI) to generate a unique Man 8 GlcNAc 2 isomer product ( Fig. 1E) is the key rate-limiting initiation signal (9,10) that ultimately leads to retrotranslocation of the polypeptide back into the cytoplasm for degradation by the proteasome in a process known as ER-associated degradation (ERAD) (11). Inhibition of ERManI can cause the accumulation of misfolded model glycoproteins in the ER lumen (12)(13)(14)(15)(16)(17)(18)(19)(20), and ERManI overexpression has been shown to accelerate the "disposal clock," hastening the disposal of misfolded proteins and even early folding intermediates of wild type proteins (9). Thus, the efficiency of creating fully folded glycoproteins for transport from the ER is defined by a competition between the kinetics of conformational maturation versus the rate of acquiring the key glycan signal for glycoprotein disposal.
Many loss-of-function human genetic diseases result from delayed folding kinetics of potentially functional polypeptides, such as the ⌬Phe 508 mutant of cystic fibrosis transmembrane regulator (21), rather than generating terminally misfolded protein structures (22). Thus, treatment of many protein misfolding disorders could be achieved if pharmacological inhibition of the rate-determining steps for ERAD allowed sufficient time for completion of the protein folding process (7,23).
ERManI is a member of a larger family of proteins, termed Class 1 mannosidases (24) (CAZy family 47 glycosylhydrolases (25)(26)(27)(28)), involved in glycoprotein maturation and disposal. Two other subgroups within this family include a subfamily of hydrolases in the Golgi complex and a subfamily of lectins in the ER. The Golgi ␣-1,2-mannosidases (termed IA (29), IB (30), and IC (31)) are essential for trimming high mannose N-glycans to the Man 5 GlcNAc 2 -Asn intermediate necessary for maturation into complex type structures on cell surface and secreted glycoproteins. In the ER, the EDEM subgroups of proteins apparently have no hydrolase activity but act as lectins as a part of the ERAD disposal machinery (2,(32)(33)(34)(35)(36)(37)(38). The present models envisage recognition of the glycan structures by the EDEM proteins in a mode similar to substrate recognition during catalysis by the true hydrolases, followed by transfer to the Sec61 translocon pore, retrotranslocation into the cytosol, and proteasomal degradation (2). Thus, understanding how this family of enzymes and lectins accomplish their functions in recognition and catalysis will provide insights into the rate-FIG. 1. Model for the structure and catalytic residues of human ERManI used in the mutagenesis studies described in this paper. The end (A) and side (B) views of the human ERManI ribbon diagram (Protein Data Bank 1X9D (46)) display the (␣␣) 7 barrel structure with the N-glycan substrate (Man 5 GlcNAc 2 substrate, stick representation from Protein Data Bank 1DL2 (43) and the glycone residue in the Ϫ1 subsite limiting decisions between glycoprotein maturation and disposal in the secretory pathway.
Class 1 ␣-mannosidases have been studied extensively with regard to enzyme kinetics, substrate specificity, structure, and mechanism (24, 39 -46). The catalytic domains of enzymes from fungal and mammalian sources have similar (␣␣) 7 barrel structures (41)(42)(43)(44)(45) that are plugged at one end by a ␤-hairpin, whereas the opposite end is composed of a broad cleft leading to the catalytic residues in the barrel core (Fig. 1, A and B). In all of the enzymes, a Ca 2ϩ ion is bound at the apex of the ␤-hairpin in the core of the barrel where it is involved in direct interactions with two of the glycone hydroxyls during the catalytic cycle (Fig. 1, A-D) (42,46). Co-complex structures have been examined between ERManI and the glycone mimic inhibitors, 1-deoxymannojirimycin (dMNJ) and kifunensine (Kif) (Fig. 1F) (42), as well as a co-complex with an uncleaved thiodisaccharide pseudosubstrate (46), revealing a novel conformational itinerary during glycoside bond hydrolysis. Our recent studies suggest that the enzyme binds to the glycone residue in the Ϫ1 subsite in a high free energy 3 S 1 conformation (46), which allows the formation of a ring-flattened 3 H 4 transition state by a least motion conformational twist of the predisposed sugar ring, and produces an inverted enzymatic product in a 1 C 4 conformation. Novel general base (Glu 599 ) and general acid (Glu 330 and Arg 334 acting in a through-water protonation scheme) functions were identified through a combination of kinetic analyses and structure determination of the ERManIthiodisaccharide co-complex ( Fig. 1, C and D) (46). All of the known Class 1 mannosidase structures are essentially identical within the Ϫ1 and ϩ1 subsites, suggesting that the catalytic mechanism for bond hydrolysis is conserved among all of the true hydrolases.
Although all of the Class 1 mannosidases cleave Man-␣1,2-Man linkages, there are significant differences in branch specificities among the different family members (24). ERManI cleaves a single residue from the central branch of the Man 9 GlcNAc 2 substrate to produce a single Man 8 GlcNAc 2 isomer product (Fig. 1E) (39,40,47). In contrast, the Golgi subfamily of enzymes recognizes the other terminal Man-␣1,2-Man branches but instead cleaves the central branch with poor efficiency (48,49). Thus, these enzymes have a mutually exclusive but complementary specificity for the complete cleavage of Man-␣1,2-Man linkages on high mannose glycans (41). Putative glycan enzymatic product co-complexes have been isolated for members of both the ER (43) and Golgi (41) subclasses of enzymes, demonstrating that differences in the cleft structures leading from the catalytic core residues confer unique glycan branch specificities for the different subfamily members.
In the studies described here, we have complemented and extended our recent work on the characterization of the ER-ManI catalytic mechanism (46) by examining for the first time the energetics of substrate binding and catalysis by a class 1 ␣-mannosidase. Kinetic and binding analyses have allowed us to dissect the energetic contributions of individual amino acid residues and the protein-bound Ca 2ϩ ion during substrate binding and catalysis. Through the use of a general acid catalytic mutant that is compromised in hydrolysis, yet maintains substrate binding with similar characteristics to the wild type enzyme, we also mapped the contributions of individual residues in the Man 9 GlcNAc 2 substrate for their interactions with the extended enzyme glycan binding pocket. These studies revealed unanticipated roles for glycan interactions in the Ն ϩ1 subsites for facilitating catalysis and substrate specificity. The experimental strategy for the binding and kinetic analyses also provides a framework for further studies on the structural and energetic basis of substrate branch recognition and catalysis by other members of the class 1 (glycosylhydrolase family 47) ␣-mannosidases.

MATERIALS AND METHODS
Mutagenesis, Expression, and Purification of Human ERManI-The mutagenesis, expression, and purification of the human ERManI catalytic domain has been described previously (42,46). Briefly, the cDNA encoding the human ERManI catalytic domain in the pPICZ␣A vector (Invitrogen) was used to perform site-directed mutagenesis using the QuikChange TM mutagenesis kit from Stratagene (La Jolla, CA). Plasmid constructs were then used to transform the Pichia pastoris strain X-33, and Zeocin-resistant colonies were screened for ERManI expression by performing Western blots using conditioned medium from induced cultures as described previously (42,46). Mutant enzymes were expressed in 1-liter shake flask cultures by induction in BMMY media, and the enzyme was purified from the conditioned media as described previously (42,46).
Enzyme, Protein, and Carbohydrate Assays-The purified wild type and mutant enzymes were assayed for ␣1,2-mannosidase activity using Man 9 GlcNAc 2 -PA as substrate as described previously (46). Briefly, the enzyme reactions (20 l) were carried out in 96-well plates at 37°C for the indicated times, stopped by the addition of 20 l of 1.25 M Tris-HCl, pH 7.6, and resolved and quantitated using a Hypersil APS-2 NH 2 -HPLC column (48). One unit of enzyme activity is defined as the amount of enzyme that generates 1 mol of Man 8 GlcNAc 2 from Man 9 GlcNAc 2 in 1 min at 37°C. Protein concentration was determined using the BCA protein assay reagent (Pierce) as described by the manufacturer. Oligosaccharide concentrations were determined by phenol-sulfuric acid assays (51).
Kinetic Analysis-Initial rates (v) for the enzymes were determined at various substrate concentrations ranging from 10 to 300 M. The catalytic coefficient (k cat ) and Michaelis constant (K m ) values were determined by fitting initial rates to a Michaelis-Menten function by nonlinear regression analysis using SigmaPlot (Jandel Scientific, San from Protein Data Bank 1X9D (46)) bound in the core of the barrel highlighted by the dark blue circle. The glycone residue in the Ϫ1 subsite (light blue circle, B) is in direct association with the protein-bound Ca 2ϩ ion (blue spacefill). The residues examined in this study are shown in the stereo diagram (C) where the stick representation of the Man 6 GlcNAc 2 glycan substrate is shown in yellow, the Ca 2ϩ ion is shown as a blue spacefill, and the relevant residues described in the text are shown as stick diagrams. Residues mutated in this study are shown as green stick figures, whereas the Glu 330 residue studied previously (46) is shown as a light blue stick figure. Water molecules coordinating the Ca 2ϩ ion are indicated by small red spacefill structures with interactions with the Ca 2ϩ ion indicated by cyan dotted lines. Interactions between the carbonyl oxygen and O-␥ of Thr 688 and the Ca 2ϩ ion are also represented by cyan dotted lines. Proposed acid, base, and nucleophile trajectories as described previously (46) are illustrated with magenta dotted lines. Hydrogen bonds are shown as green dotted lines. A schematic diagram demonstrating the interactions between the N-glycan and the active site (D) employ a similar color scheme for hydrogen bonding, acid, base, and nucleophile trajectories and Ca 2ϩ coordination as C, with the exception of the hydrophobic stacking between Phe 659 and the C4-C5-C6 region of the Ϫ1 residue (black dotted lines). Residue numbering of amino acid side chains in the respective subsites is indicated in the figure. The residue nomenclature and linkages for the monosaccharides in the Man 9 GlcNAc 2 -Asn substrate are indicated (E), and the linkage cleaved by ERManI is also indicated. A similar monosaccharide nomenclature is used in C and D to label the respective residues in the glycan structure. Labeling of the enzyme subsites with negative (glycone in the Ϫ1 subsite) or positive (ϩ1 and ϩ2 subsite residues) numbers reflects their respective positions relative to the glycosidic bond being cleaved (64). Schematic structures of ␣-D-mannose and the corresponding inhibitors, dMNJ and Kif, are shown in F. Rafael, CA). k cat /K m values were derived from reciprocal plots of v and [S] where needed. In studies on the temperature dependence of catalysis, values for k cat were obtained between 5 and 40°C at 5°C intervals and were used to calculate activation energies (E a ) from the slopes (ϪE a /R) of Arrhenius plots (ln(k cat ) as a function of 1/T). The thermodynamic activation parameters were described by the Equations 1-3 (52,53): Ϫ lnk catͪ (Eq. 1) where R is the gas constant (8.314 J⅐mol Ϫ1 K Ϫ1 ); k B is the Boltzmann constant (1.3805 ϫ10 Ϫ23 J⅐K Ϫ1 ), and h is the Planck constant (6.6256 ϫ 10 Ϫ34 J⅐s). Calcium Equilibrium Analysis-Purified wild type ERManI or the T688A mutant was incubated with 200 mM EGTA for 2 h at 4°C prior to desalting over a Sepharose G-25 column (1 ϫ 40 cm), which was pretreated with 0.5 M EGTA, followed by pre-equilibration in calciumfree buffer (20 mM MES, pH 7.0, 150 mM NaCl). The calcium-free protein solution was concentrated to 1 mg/ml using an Amicon YM-10 membrane. The calcium content of the protein solution, confirmed by inductively coupled plasma mass spectrometry (ICP-MS), was below 50 ppb.
The total amount of calcium chloride required to generate defined concentrations of free Ca 2ϩ ion (0 -500 M) in an EGTA-containing buffer (20 mM MES, pH 7.0, 150 mM NaCl, 5 mM EGTA) was calculated using WEBMAXCLITE version 1.15 (54) (available online at www. stanford.edu/ϳcpatton/maxc.html). Solutions of defined Ca 2ϩ ion concentration were prepared, and the pH of the mixture was monitored and adjusted to pH 7.0 by addition of NaOH. All stock solutions were prepared in EDTA-treated plasticware. Total calcium ion concentrations in each solution were confirmed by atomic absorption and used to recalculate the free calcium ion concentration under the conditions of analysis (55).
The calcium-free enzyme solution was diluted in calcium-free buffer to obtain a stock solution of 80 g/ml for wild type ERManI and 140 g/ml for the T688A mutant. Aliquots of the enzyme solutions (5 l) were added to each of the buffers (10 l) containing defined Ca 2ϩ ion concentrations before addition of 5 l of 80 M Man 9 GlcNAc 2 -PA. Enzyme reactions were allowed to proceed at 37°C for 1 h, stopped, and analyzed by NH 2 -HPLC chromatography as described above. Plotting of mannosidase enzyme activity versus Ca 2ϩ ion concentration revealed a sigmoidal curve similar to data expected for a common equilibrium dialysis experiment, allowing the calculation of the Ca 2ϩ affinity constant (K Ca ) by nonlinear regression analysis using Equation 4 (55), where y is equal to the moles of Ca 2ϩ bound per mol of enzyme, measured as units of ␣-mannosidase enzyme activity resulting from Ca 2ϩ binding to the enzyme, and n is the apparent ERManI-specific activity.
Binding Studies by Surface Plasmon Resonance (SPR)-SPR analyses were conducted using a Biacore 3000 apparatus (Biacore AB) with recombinant ERManI immobilized on the SPR chip surfaces at 25°C by the amine-coupling method as described previously (46). Mock-derivatized flow cells served as reference surfaces. The binding analyses were routinely performed at 10°C with continuous flow (30 l/min) of running buffer except in the temperature-dependent interaction studies, which were performed between 5 and 35°C at 5°C intervals, consecutively, in an automated method. The running buffer was 10 mM MES, pH 7.0, 300 mM NaCl, and 5 mM CaCl 2 in all cases, except in calciumdependent binding studies. For the latter studies, the chip surface was subjected to overnight treatment with 20 mM MES/NaOH, pH 7.0, 300 mM NaCl, 5 mM EGTA, at a flow rate of 5 l/min, prior to binding analyses using 20 mM MES/NaOH, pH 7.0, 300 mM NaCl, and 200 M EGTA as running buffer. Analytes were prepared in the respective running buffers by 2-fold serial dilution to obtain an appropriate concentration range. The binding of Man 9 -5 GlcNAc 2 -PA and Man 9 GlcNAc 2 -glycopeptide glycans was analyzed in a concentration series (0.4 -400 M) over a low density immobilization surface (46) of recombinant protein (3000 response units), whereas dMNJ and Kif (2-1000 M) were analyzed over a high density immobilization surface (46) of recombinant protein (10,000 response units). The base line returned to the original response in 5 min for all analytes described here without a further regeneration procedure, except for analyses using Kif as the analyte, which did not dissociate from the chip surface even with extensive washing.
SPR data for each concentration of analyte were collected in duplicate and globally fit to a 1:1 Langmuir binding algorithm model to calculate the on-rate (k a ), the off-rate (k d ), and the equilibrium dissociation constant (k d /k a ϭ K D ) using the BIAevaluation 3.1 software (56). Alternatively, the maximal equilibrium sensorgram values were used to plot a saturation binding curve and calculate values for the equilibrium dissociation constant (K D ) directly.
K D values measured at different temperatures were used to calculate thermodynamic parameters (56, 57) of binding using van't Hoff equation (Equation 5), (58) and Gibbs free energy change for binding (⌬G) was also calculated (Equation 6). The van't Hoff equation allows the calculation of thermodynamic parameters using the linear relationship of y ϭ ln(K D ) versus x ϭ 1/T, which gives a slope of ⌬H/R and an intercept of Ϫ⌬S/R (58).
The effects of temperature on the association rates (k a ) and dissociation rates (k d ) were independently determined using the Eyring equation (Equation 7) (58).
Similar to the van't Hoff analysis, the Eyring equation allows thermodynamic parameters to be determined from measured k a and k d values at different temperatures by a linear relationship of y ϭ Rln(hk a /k B T) or y ϭ Rln(hk d /k B T) versus x ϭ 1/T, where the slope and the intercept of the Eyring plots are Ϫ⌬H † and ⌬S † , respectively (58). Isothermal Titration Calorimetry (ITC)-Calorimetry measurements were performed with a 4200-ITC calorimeter (Calorimetry Sciences Corp., Lindon, UT) as described (59,60). Protein solutions for ITC analysis were dialyzed overnight against buffer containing 20 mM MES, pH 7.0, 150 mM NaCl, 5 mM CaCl 2 and 0.75 M 3-(1-pyridino)-1-propane sulfonate (NDSB201; Calbiochem) at 4°C. The ligand solutions of Kif and dMNJ were prepared by diluting the compounds in the buffer used for protein dialysis. Aliquots (5-10 l) of the ligand solution (1-5 mM) were automatically delivered into 1.3 ml of protein solution (1-2 mg/ml) in the reaction cell. The calorimetry cell was allowed to return to equilibrium for 4 min prior to the next injection. The data analysis was performed using DataWork and BindWork software provided by manufacturer. Protein and glycan molecular structure figures were prepared using MacPymol (version 0.95) 2 to generate rasterized images.

RESULTS
Kinetic Analysis of ERManI Mutants-Mutations were generated previously in five residues that were hypothesized to be involved in catalysis by ERManI (46). These mutations tested the roles of putative general acid (E330Q and R334A) and general base (E599Q and H524A) functions, as well as a residue that proved to play a critical role in providing hydrogen bonding interactions with the mannose residue in the ϩ1 subsite (D463N) (Fig. 1, C and D). In the present study, we have mutagenized two residues at the base of the Ϫ1 subsite (Phe 659 and Thr 688 ), one residue involved in interactions with several Ն ϩ1 residues (Arg 461 ), and one residue that interacts with both the Ϫ1 and ϩ2 subsite residues (Arg 597 ) (Fig. 1, C and D), and characterized their roles in catalysis and substrate binding. Phe 659 was shown previously to provide van der Waals interactions to the C4-C5-C6 region of the glycone in the Ϫ1 subsite during catalysis (Fig. 1D) (42,46). Thr 688 is positioned at the apex of the ␤-hairpin in the core of the (␣␣) 7 barrel where it has been shown to be the sole protein residue that directly coordinates the bound Ca 2ϩ ion through both its O-␥ and carbonyl oxygens (Fig. 1D) (42,46). The other four points of coordination of the Ca 2ϩ ion to the enzyme are indirect through-water in-teractions with carboxylate side chains in the core of the barrel. Arg 461 interacts with several residues in the core of the Man 9 GlcNAc 2 substrate (Fig. 1, D and E, residues M7, M4, and M3) and has been proposed to contribute to branch specificity for Saccharomyces cerevisiae ERManI (61). A Leu residue is found at the equivalent position in the Golgi subclass of enzymes. Previous mutagenesis studies generating the equivalent of an R461L mutant for S. cerevisiae ERManI (61) resulted in an enzyme that had a hybrid activity between the specificity of ERManI that cleaves only the central branch mannose residue and the Golgi mannosidases that cleave the remaining ␣1,2-Man residues down to Man 5 GlcNAc 2 . Finally, Arg 597 appears to play a dual role in hydrogen bonding to the O-6Ј hydroxyl oxygen of the glycone in the Ϫ1 subsite via NH1 and the O-4Ј hydroxyl of the mannose in the ϩ2 subsite via NH2. Each of the mutants (T688A, F659A, R461A, R461L, and R597A) was expressed in P. pastoris as a secreted catalytic domain, and the detailed enzyme kinetic parameters for hydrolysis of Man 9 GlcNAc 2 -PA were determined and are summarized in Table I. The pH optima for all of the mutant enzymes were slightly below the value for wild type ERManI (pH 6.5-6.8 versus 7.0 for wild type; Table I). The catalytic rates (k cat ) and the catalytic efficiencies (k cat /K m ) of Man 9 GlcNAc 2 cleavage for all of the mutants were significantly decreased, resulting in a range of k cat /K m values that varied from 0.6 to 12% of wild type values. Surprisingly, the K m value for the T688A mutant was reduced 7.3-fold, whereas the Phe 659 mutant remained unaffected. In contrast, the K m values for the other mutants were all significantly increased by 3-5-fold.
For all of the mutants tested, the R461L mutant was unique in its ability to readily hydrolyze ␣1,2-mannoside residues from Man 9 GlcNAc 2 -PA to Man 8 -6 GlcNAc 2 -PA, as described previously for an equivalent mutant of S. cerevisiae ERManI (61) (data not shown). The 53-fold decrease in k cat and 3-fold increase in K m (Table I) indicated that although the amino acid substitution relaxed the specificity of the enzyme for glycan cleavage beyond Man 9 GlcNAc 2 , the enzyme lost significant catalytic efficiency as a result of the mutation. In contrast, the R461A or R597A mutants were unable to cleave beyond Man 8 GlcNAc 2 . However, the catalytic rate of the R461A mutant was intermediate between the wild type enzyme and the R461L mutant. These data suggest that the removal of the Arg 461 side chain in the R461A mutant moderately compromised catalysis while retaining ERManI substrate specificity. In contrast the R461L mutant was altered in substrate specificity, although its catalytic efficiency was severely compromised.
The catalytic rates (k cat ) of wild type ERManI and the E330Q and T688A mutants were also obtained from initial rates measured at different temperatures but under optimal pH conditions for the respective enzymes (pH 7.1, 5.3, and 6.5 for wild type, E330Q, and T688A, respectively). The enzyme activities increased with temperature, and activation energies (E a ) were calculated from the slopes of the Arrhenius plots (supplemental Fig. 1A and supplemental Table I). The wild type enzyme and T688A mutant appear to have similar trends, with significant enthalpy and entropy contributions to the activation energy, whereas the E330Q mutant had a slightly reduced entropic contribution (supplemental Fig. 1A and supplemental Table I).
Effect of the T688A Mutant of ERManI on Ca 2ϩ Ion Affinity and Enzyme Activity-The role of the protein-bound Ca 2ϩ ion in catalysis by ERManI was determined by depleting wild type ERManI or the T688A mutant of bound Ca 2ϩ and then performing enzyme assays at defined Ca 2ϩ concentrations controlled by the presence of the divalent cation chelator EGTA. In the absence of any added Ca 2ϩ , both proteins exhibited no detectable enzyme activity (supplemental Fig. 2). Addition of Ca 2ϩ resulted in a progressive appearance of enzyme activity, allowing the calculation of the Ca 2ϩ affinity for the enzyme, K Ca , by curve-fitting. The K Ca values for the wild type enzyme (0.24 Ϯ 0.02 M) and the T688A mutant (0.15 Ϯ 0.01 M) were quite similar, yet the specific activity of the T688A mutant was generally ϳ15-fold lower than the wild type enzyme at all Ca 2ϩ concentrations where activity could be detected. The observation that the T688A mutant is compromised in catalysis (reduced k cat ) but increased in substrate binding affinity (reduced K m ), while being essentially unaltered in Ca 2ϩ binding affinity, indicates that the mutation has a direct effect on catalysis rather than acting through a reduced affinity for binding and coordinating Ca 2ϩ . The K Ca value for wild type ERManI was similar to the affinity constants previously determined for yeast ERManI (62) and rabbit liver Golgi ManIA (63).
Glycan and Inhibitor Binding Affinity Measurements to Human ERManI-In addition to kinetic analysis, the binding affinities of inhibitors and high mannose oligosaccharides to wild type and mutant forms of ERManI were also examined by SPR. Prior SPR studies with wild type and mutant forms of ERManI revealed significant alterations in the on-rates (k a ) and off-rates (k d ) of binding to dMNJ and Man 9 GlcNAc 2 -glycopeptide ligands (46). Correlations were made with the positions of the mutations and their impacts on ligand binding affinity. In the present study, we performed similar types of SPR studies with wild type ERManI in the presence and absence of Ca 2ϩ , as well as testing the effects of the T688A, F659A, R597A, R461L, and R461A mutants on the binding of Man 9 GlcNAc 2 -glycopeptide, dMNJ, or Kif ligands ( Fig. 2 and Table II). As described previously (46), the equilibrium dissociation constants (K D ) could be measured from a combination of the on-rates (k a ) and off-rates (k d ) determined by curve-fitting of the SPR sensorgrams, where K D ϭ k d /k a . When the on-rates and off-rates were too fast for accurate measurement, plotting a Assay data were fit to generate a bell-shaped curve to yield a pH optimum with a standard error of Ͻ0.1 pH unit. b Kinetic constants for wild type and E330Q mutant of ERManI were as reported previously (46) and are shown as a reference. c The enzyme exhibits the ability to readily hydrolyze Man 9 GlcNAc 2 -PA to Man 6 GlcNAc 2 -PA.
of a saturation curve for the equilibrium values of the binding sensorgrams (Fig. 2, inset plots) allowed an alternative means of determining the K D values (46). Depletion of Ca 2ϩ from wild type ERManI resulted in a significantly slowed on-rate and off-rate for binding of the Man 9 GlcNAc 2 -glycopeptide ligand by SPR (Fig. 2), but resulted in only a 1.6-fold reduction in the equilibrium binding affinity for the glycan ligand (Table II). In contrast, the binding affinity of dMNJ was reduced 343-fold, largely as a result of a 127-fold reduction in the on-rate (Table II). These data indicate that the presence of Ca 2ϩ bound to the core of the (␣␣) 7 barrel influences the rate of glycone binding to the Ϫ1 subsite, but the overall equilibrium binding affinity of the larger Man 9 GlcNAc 2 substrate is not significantly influenced by the absence of the divalent cation.
Similar to the reduced on-and off-rates for Man 9 GlcNAc 2glycopeptide binding to the Ca 2ϩ -depleted enzyme, the T688A mutant also had significantly reduced on-and off-rates for high mannose glycan binding, but an ϳ50-fold increase in the equilibrium binding affinity (Table II) similar to the increase in glycan binding affinity previously observed for a E330Q general acid mutant (46). The increased glycan binding affinity for the T688A mutant, when combined with the significantly reduced k cat (Table I) and an unaltered Ca 2ϩ binding affinity (supplemental Fig. 2), indicates that the reduced catalytic turnover results in a substrate that is stabilized in an uncleaved form in the active site. This altered environment is less favorable for binding to dMNJ (5-fold decrease in binding affinity), which we have proposed previously (42,46) to resemble the conformation of the enzymatic product.
In contrast to the effects of the T688A mutant at the base of the Ϫ1 subsite, mutation of Phe 659 (F659A), which provides van der Waals interactions with the Ϫ1 subsite residue, had a minimal effect on Man 9 GlcNAc 2 binding (4-fold reduction in  D, F, and H) or Kif (I and J) ligands were tested for binding. The data were collected in duplicate, and representative SPR sensorgrams in the ligand concentration series are shown. If the on-and off-rates (k a and k d , respectively) were sufficiently slow, curve fitting of the sensorgrams was performed using the 1:1 Langmuir binding algorithm model to determine the values for the equilibrium dissociation constants (K D ϭ k d /k a ). In binding studies where the kinetics for binding of the Man 9 GlcNAc 2 ligand were too rapid for curve fitting, the equilibrium sensorgram values were used to plot a saturation curve (insets in the A and G plots) and calculate values for K D . In the F659A mutant, binding of dMNJ was completely abolished as indicated by the absence of a deflection in the SPR sensorgram trace. The values for k a , k d , and K D based on the SPR studies are shown in Table II. K D , see Table II), consistent with the lack of an effect of the mutation on the K m with the same glycan substrate. However, the binding of dMNJ was completely abolished for this mutant enzyme (K D Ͼ 10 mM; Fig. 2), consistent with 123-fold reduction in k cat (Table I). These data suggest that Phe 659 plays a key role in glycone binding within the Ϫ1 subsite to promote catalysis.
Finally, the binding of Kif to wild type ERManI was examined by SPR. The equilibrium binding affinity for this inhibitor could only be estimated at Ͻ30 nM, because there was no detectable dissociation of the compound from the enzyme, even after extensive washing (Fig. 2). Inhibitor binding studies by ITC confirmed the tight binding by the inhibitor, which was highly exothermic and had a strongly favorable enthalpy of binding (see supplemental Table II). The E330Q, D463N, and T688A mutants also formed stable nondissociable complexes with Kif (data not shown). To test the role of the van der Waals interactions between Phe 659 and the inhibitor at the base of the Ϫ1 subsite, we examined the binding of Kif to the F659A mutant. In contrast to the wild type enzyme, binding of Kif to the F659A mutant was reversible (Fig. 2), with measurable onand off-rates and a K D of 1.45 M (Table II).
Binding of Man 9 GlcNAc 2 to the R461A and R597A mutants exhibited reduced on-and off-rates, but the K D values were similar to the wild type enzyme (data not shown). In contrast, the R461L mutant exhibited no detectable binding (K D Ͼ 1 mM) to Man 9 GlcNAc 2 (data not shown), consistent with the poor catalytic efficiency of this mutant enzyme that has a hybrid activity between ERManI and the Golgi subclass of enzymes.
Temperature and pH Dependence of Glycan Binding-In an effort to examine the temperature dependence of glycan binding to the wild type and mutant forms of ERManI, we performed a series of SPR binding studies with Man 9 GlcNAc 2 as ligand at temperatures between 5 and 35°C. The sensorgram responses for wild type ERManI, E330Q, and T688A (supplemental Fig. 3) are representative of the effects of temperature on the respective enzymes. Because of the fast on-and off-rates, binding to the wild type enzyme could only be measured at the equilibrium plateau values for the sensorgrams, and below 15°C there was little effect of increasing temperature (supplemental Fig. 3). At higher temperatures there was a progressive decrease in the sensorgram amplitude, presumably as a result of a combination of binding and hydrolysis of the ligand and subsequent release of the enzymatic product. For the T688A mutant, the off-rate progressively increased with increasing temperature, whereas the equilibrium sensorgram values increased up to ϳ20°C and then progressively decreased (supplemental Fig. 3). We have interpreted these data to indicate that glycan binding was more prevalent than glycan hydrolysis at low temperature but that increased hydrolysis at higher temperature led to an increased off-rate and an inflection for the equilibrium sensorgram values. In contrast, the off-rate for the E330Q mutant was considerably less influenced by increasing temperature, whereas the on-rate increased with temperature, leading to an increase in equilibrium binding (supplemental Fig. 3). Similar results were revealed when the binding data were subjected to van't Hoff analysis, where the slopes of the plots indicated that the temperature dependence of the K D for the T688A mutant was greater than for the E330Q mutant (supplemental Fig. 1B). Eyring analysis demonstrated that the temperature dependence of the K D for the T688A mutant was mainly due to an increase in the dissociation rate with increasing temperature (supplemental Fig. 1, C and D). Calculation of the thermodynamic parameters for the glycan association and dissociation (supplemental Table I) indicated that there was a significantly greater entropic contribution to glycan dissociation for the T688A mutant by comparison to the E330Q mutant, as would be predicted for the greater temperature dependence of dissociation.
To confirm the temperature dependence of enzyme activity under the conditions used for SPR analysis (pH 7.0) we compared the specific activities of the various enzyme forms at 10 and 37°C at this pH. The wild type enzyme had a 30-fold increase in specific activity between 10 and 37°C, whereas the T688A mutant increased 10-fold and the E330Q mutant increased only 2-fold over the same temperature range (data not shown). Because the E330Q mutant has a reduced pH optimum for catalysis (pH 5.3) relative to the wild type enzyme (pH 7.1) (46), we also tested the pH dependence of glycan binding by the E330Q mutant. At pH values Ն7.0, the E330Q mutant exhibits high affinity binding (K D Ͻ 1 M) (Table III and supplemental Fig. 4), whereas at pH values below 7.0 both the on-and off-rates are altered to result in an enzyme with a 46-fold lower binding affinity at pH 5.0. We interpret the alterations in onand off-rates and decrease in glycan affinity to reflect an increase in glycan hydrolysis under conditions closer to the pH optimum of the mutant enzyme. These data indicate that the E330Q mutant is an effective model for binding analyses under our conditions for SPR (pH 7.0, 10°C) because, in contrast to wild type ERManI or the T688A mutant, there are minimal contributions of ligand hydrolysis. Energetic Contributions of Glycan Binding to the E330Q Mutant of ERManI-The E330Q mutant of ERManI was used as a model to investigate the contributions of various residues in a Man 9 GlcNAc 2 oligosaccharide for binding to the mutant enzyme. Individual Man 9 GlcNAc 2 , Man 8 GlcNAc 2 , Man 6 GlcNAc 2 , and Man 5 GlcNAc 2 glycans, as well as mannose disaccharides (Fig. 3A), were examined for their respective binding parameters (Table IV), and the resulting K D values were converted to their corresponding ⌬G values for binding. The contributions of individual residues or groups of residues were subsequently calculated by a combination of difference calculations (⌬⌬G values) using the rationale shown in Fig. 3B and Table V. These calculations indicate that binding of the glycone residue in the Ϫ1 subsite contributes only ϳ21% (Ϫ1.68 kcal/mol) to the overall binding energy of a Man 9 GlcNAc 2 ligand (Table V and Figs. 3C and 4), whereas the ϩ1 subsite residue (M7 in Figs. 1E and 4) contributes ϳ52% (4.2 kcal/mol) to the overall binding energy. Residues M9, M8, and M11 (Fig. 1E) together contribute ϳ19% (1.53 kcal/mol) to the glycan binding energy, whereas the remainder of the glycan contributes ϳ0.42 kcal/mol (ϳ6%), and the peptide backbone contributions are negligible (ϳ0.19 kcal/mol) ( Fig. 3C and Table V). DISCUSSION Class 1 (CAZy glycosylhydrolase family 47 (25-28)) ␣-mannosidases play key and diverse roles in glycan maturation and disposal of misfolded glycoproteins in the ER (2,9,24,(32)(33)(34)(35)(36)38). Prior studies have revealed that the overall protein fold for the catalytic domain of members of this family is an (␣␣) 7 barrel structure with a catalytic site in the core of the barrel (41)(42)(43)(44)(45). Co-complex structures between ERManI and dMNJ, Kif (42), or a mannobiose thiodisaccharide (46) revealed the conformational itinerary of the glycone during catalysis. Distortion of the glycone into a 3 S 1 conformation during substrate binding has been proposed to predispose the substrate for hydrolysis by a least motion conformational twist through a ringflattened 3 H 4 transition state producing an inverted enzymatic product in a 1 C 4 conformation (46). Several questions remain regarding the energetics of substrate binding during catalysis and the nature of the active site chemistry. Conservation in the positions of key residues in the Ϫ1 and ϩ1 enzyme subsites indicates that both the ER and Golgi subfamilies of hydrolases cleave Man-␣1,2-Man linkages by a similar mechanism. However, the ER and Golgi enzymes recognize distinctive terminal branches of Man 9 GlcNAc 2 substrates. Prior structural data on putative enzyme-product co-complexes for both the ER and Golgi enzymes support the proposal that the respective enzymes have distinctive geometries in the clefts leading from the barrel cores that engage the Ͼ ϩ1 residues of the respective substrates and confer the unique branch specificities for the different enzyme subfamilies (41). In an effort to understand the energetics of substrate binding and catalysis for the class 1 mannosidases, we have examined the interactions between the wild type and mutant forms of ERManI by using various substrates and inhibitors.
Two residues that were shown previously to play roles in FIG. 3. Strategy for determining the energetic contributions of respective Man 9 GlcNAc 2 glycan residues for binding to the E330Q mutant of ERManI. A series of SPR binding studies with the immobilized E330Q mutant enzyme was performed by using various ligand structures as indicated in A. In addition to Man 9 GlcNAc 2 -PA and Man 9 GlcNAc 2 -glycopeptide ligands, discrete isomers of Man 8 GlcNAc 2 -PA, Man 6 GlcNAc 2 -PA, and Man 5 GlcNAc 2 -PA, as well as Man␣1,2Man disaccharides, were tested. For each ligand, binding analyses were performed, and K D values were determined (Table IV). The resultant K D values were converted to values of ⌬G (Table IV) by using the relationship ⌬G ϭ ϪRTln(1/K D ), and energetic contributions to glycan binding were then calculated by a series of ⌬G difference calculations (⌬⌬G calculations) as shown in the example calculation (B). The calculations for combinations of residues are summarized in Table V, and the data are graphically summarized in C. The glycone residue was found to contribute only 20.9% to the overall glycan affinity, whereas the ϩ1 subsite residue contributed 52.4% to the glycan binding energy.

TABLE III
Summary of the pH dependency of Man 9 GlcNAc 2 binding to the E330Q mutant of ERManI SPR binding studies were performed with the immobilized E330Q mutant enzyme at 10°C at varied values of pH. At each pH the SPR sensorgram traces were obtained for the binding of Man 9 GlcNAc 2 (supplemental Fig. 4), and values for k a , k d , and K D were determined. Minimal changes in the sensorgram profiles were detected at pH 8 -6, whereas the on-rates and off-rates were both considerably altered at pH 5.0. Calculated values for K D indicated a progressive 40-fold lowering of glycan binding affinity between pH 8 and 5. binding to Ͼ ϩ1 subsite residues (46,61) were examined for their effects on catalysis and substrate binding. The R461L, R461A, and R597A mutants all resulted in reduced catalytic rates (varying 6 -34-fold) and slightly increased K m values (3-5-fold) (Fig. 4). Both the R461A and R597A had binding affinities for dMNJ and Man 9 GlcNAc 2 that were comparable with the wild type enzyme, whereas the R461L mutant had no detectable Man 9 GlcNAc 2 binding (K D Ͼ1 mM). Arg 461 has been proposed to play a major role in ERManI substrate binding and recognition (61), contrasting with a Leu at the equivalent position for the Golgi enzymes. Only the R461L mutant was found to readily catalyze the cleavage of Man 9 GlcNAc 2 -PA to Man 8 -6 GlcNAc 2 -PA, similar to data reported for S. cerevisiae ERManI (61). Our working hypothesis is that Arg 461 contributes to high substrate binding affinity and substrate specificity only within the context of the overall geometry of the ERManI glycan binding cleft. The R461L mutation would eliminate specific hydrogen bonding interactions that confer glycan affinity and specificity as well as creating space at the core of the glycan-binding site for more flexible interactions with alternative terminal glycan branches. However, the R461A mutant would be predicted to allow even greater flexibility for glycan binding, yet the latter mutant maintains the restricted substrate specificity of wild type ERManI. These data suggest that the Leu residue at this position provides a positive role in broadening the substrate specificity, but within the inappropriate context of the ERManI glycan-binding cleft steric constraints preclude high affinity glycan interactions. In contrast, within the context of the active site clefts of the Golgi subfamily of enzymes, the corresponding Leu residue would be expected to confer high affinities of substrate binding and appropriate branch specificities for the latter enzymes. Initial studies of inhibitor binding examined the energetics and kinetics of interactions by ITC and SPR. Both approaches indicated that binding of the inhibitors was highly exothermic, and the K D values from ITC were in close agreement with K i values from catalytic measurements and binding affinities by SPR (supplemental Table II). The major difference in structure between dMNJ and Kif is the fused five-membered ring in the latter compound (Fig. 1F) that causes Kif to be "pre-loaded" in the high free energy 1 C 4 conformation prior to binding to the enzyme (46). The restricted conformation and the additional interactions between the enzyme and the Kif five-membered ring were proposed to account for the higher binding affinity of Kif in comparison to dMNJ (42). Surprisingly, binding studies by SPR indicated that the on-rate for binding to wild type ERManI was comparable between dMNJ and Kif, but the offrate for Kif was considerably slower. Thus, the restricted conformation of the Kif six-membered ring did not accelerate binding of the inhibitor but significantly slowed dissociation. SPR binding studies between Kif and the F659A mutant, which should eliminate the van der Waals interactions with the C4-C5-C6 region of the ring-constrained inhibitor, significantly increased the off-rate and lowered the overall inhibitor binding affinity. Consistent with the lower binding affinity for Kif to the mutant enzyme, binding of dMNJ to the F659A mutant was not even detectable by SPR (K D Ͼ10 mM). These data indicate that Phe 659 plays a critical role in stabilizing the glycone in the active site and that the van der Waals interactions with the ring-constrained inhibitor contributes to its slow dissociation rate from the active site. By extension, the greater ring flexibility of dMNJ or the substrate/product glycone residue associated with the wild type enzyme would be predicted to contribute an entropic component favoring dissociation from the active site. Consistent with a role for Phe 659 in inhibitor binding, the F659A mutant also caused a 123-fold reduction in k cat , suggesting that interactions between Phe 659 and the glycone in the Ϫ1 subsite play a significant role in constraining the substrate into the 3 S 1 conformation required for catalysis.
At the base of the Ϫ1 subsite a protein-bound Ca 2ϩ ion interacts directly with the glycone 2Ј-and 3Ј-hydroxyl residues. The ion is coordinated directly with the O-␥ and carboxyl oxygens of Thr 688 and indirectly with four water molecules associated with Glu residues in the core of the Ϫ1 subsite. Two types of studies examined the roles of the Ca 2ϩ ion in glycan binding and catalysis. First, SPR binding studies on the Ca 2ϩdepleted enzyme indicated that the ion contributes to glycan on-and off-rates, but the enzyme had an almost identical equilibrium binding affinity for the Man 9 GlcNAc 2 ligand as the enzyme containing bound Ca 2ϩ . Binding of dMNJ was drastically reduced, with a predominant effect on reducing the onrate of the inhibitor. Second, altering the Thr 688 side chain to  subsite residue as the major contributor to glycan binding Free energy calculations were generated by a strategy shown in Fig.  3B using a color convention for the residues in the high mannose ligands as indicated in the same figure. ⌬⌬G calculations to define the energetic contributions of individual or combinations of residues were performed by calculating the differences in free energies of binding for various glycan ligands (from Table IV) to the E330Q mutant.
an Ala did not significantly alter Ca 2ϩ binding affinity for the enzyme, but it reduced k cat by 61-fold and increased Man 9 GlcNAc 2 binding affinity by 50-fold. These data suggest that alterations in the Ϫ1 subsite can lead to an increased binding affinity for an uncleaved substrate. The data also suggest that glycone interactions with the enzyme-bound Ca 2ϩ ion, in the context of the appropriate tethering by Thr 688 , directly facilitate catalysis.
One of the main goals of our studies on ERManI was to map the energetics of interaction between glycan substrates and the active sites of class 1 mannosidases and to identify which residues contribute to catalysis and substrate specificity for members of the different enzyme subfamilies. The strategy for these studies will be to examine the individual contributions of enzyme residues in their interactions with the glycan substrates and the individual contributions of glycan resides in their interactions with the enzyme. The wild type enzyme is not an effective model for SPR binding studies of this type, because the rapid on-and off-rates have contributions from both binding and ligand hydrolysis. A preferable model would be an enzyme form that is compromised in hydrolysis yet maintains substrate binding with similar characteristics to the wild type enzyme. For maximal utility, the mutant enzyme would also have high glycan binding affinities but relatively slower onand off-rates than the wild type enzyme so that subtle changes in association and dissociation rates could be readily measured.
The E330Q mutant appears to fulfill all of these criteria, because it has minimal catalysis under the conditions that we routinely use for SPR binding studies (10°C, pH 7.0), yet it retains high affinity glycan binding with reduced on-and off-rates. As a demonstration of the utility of this mutant, we examined the binding affinities of a collection of glycan ligands, and we used the relative binding energies to calculate the contributions of the respective glycan residues in a Man 9 GlcNAc 2 substrate for their interactions with the mutant enzyme. The method yielded reproducible results, because the binding energies calculated from different combinations of glycans yielded reasonably similar energetic contributions (Table  V and Fig. 3).
The results of the glycan binding studies revealed that the ϩ1 subsite residue (residue M7) contributes a majority (ϳ52%) of the binding energy for a Man 9 GlcNAc 2 ligand, whereas the glycone binding to the Ϫ1 subsite contributes only ϳ21% of the binding energy. Another ϳ19% of the binding energy is contributed from the other peripheral ␣1,2Man residues, and ϳ7.6% of the binding energy comes from core glycan residues. These data are consistent with binding data obtained from the respective ERManI mutants (Fig. 4). Mutations in the Ϫ1 subsite or conditions that significantly reduce dMNJ binding affinity or catalysis either have a minimal effect on Man 9 GlcNAc 2 glycan binding (F659A mutant or Ca 2ϩ depletion) or actually increase glycan binding affinity (T688A and E330Q mutants). Thus, the contributions of the Նϩ1 subsite residues to the overall binding affinity can compensate for compromised interactions with glycan substrates in the Ϫ1 subsite (Fig. 4).
In contrast, alterations in the ϩ1 subsite, such as the D463N or R461L mutants (Fig. 4), abolished glycan binding (K D values Ͼ1 mM). The former side chain anchors the interactions between the enzyme and the M7 residue in the ϩ1 subsite by hydrogen bonding with the sugar O-3Ј and O-4Ј hydroxyls (46). For the latter residue, Arg 461 has been shown to form a matrix of interactions with mannose residues M7, M4, and M3 (43, 61), yet substitution with an Ala residue (R461A) resulted in a near FIG. 4. Summary of the enzyme kinetics and binding studies for key protein and glycan residues in the ERManI active site. A schematic display of the catalytic and binding components in the ERManI active site is shown highlighting the residues interacting with the Ϫ1, ϩ1, and ϩ2 subsite glycan residues. Fold differences for kinetic parameters (values for k cat and K m ) and SPR binding parameters (K D values for binding to Man 9 GlcNAc 2 and dMNJ) relative to values for wild type ERManI for each mutant are indicated adjacent to each respective amino acid. In addition, the respective proposed function for each residue based on the binding and kinetic analyses is indicated adjacent to each amino acid in red text. Energetic contributions to binding within the Ϫ1 and ϩ1 subsites are also indicated by the dotted lines with shading. The "Discussion" summarizes the data from the kinetic and binding analyses for the wild type and mutant enzymes and describes their respective contributions to substrate binding and catalysis. wild type binding affinity for Man 9 GlcNAc 2 . In contrast, the R461L mutant has no detectable Man 9 GlcNAc 2 binding, confirming that there is likely a problem with steric hindrance in the latter mutant.
In conclusion, the combined studies on catalysis and glycan binding by wild type and mutant ERManI forms revealed an active site cleft that promotes substrate binding predominantly through interactions with the Ն ϩ1 subsite residues (Fig. 4). How do these data compare with our emerging model for catalysis by class 1 ␣-mannosidases (46)? The docking of the glycone residue in a high free energy 3 S 1 conformation, predisposed for glycoside bond hydrolysis, is partly facilitated by van der Waals interactions with Phe 659 and additional hydrogen bonding to Ϫ1 subsite residues. However, the predominant source for substrate binding energy is provided by interactions with the ϩ1 residue, dominated by the pair of hydrogen bonds from Asp 463 to the 3Ј-and 4Ј-hydroxyls of the ϩ1 subsite residue (Fig. 4). These latter interactions likely offset the entropic penalty for binding to a high free energy glycone conformation in the Ϫ1 subsite, creating favorable energetics for the conformational distortion required for glycoside bond hydrolysis. Additional binding energy and branch specificity are provided by the interactions with the Ͼ ϩ1 subsite residues. Interactions between the glycone residue and the enzyme-bound Ca 2ϩ ion influence the association and dissociation rates for glycan substrates, but do not significantly increase their respective equilibrium binding affinities. However, the Ca 2ϩ ion does promote catalysis, presumably through the assistance of Thr 688 , the sole residue involved in direct coordination with the divalent cation. In the absence of the Thr 688 side chain, a solvent water molecule likely replaces the lost point of Ca 2ϩ ion coordination. Ca 2ϩ binding affinity is not reduced in the T688A mutant, yet the catalytic rate is significantly reduced, and glycan binding affinity is surprisingly increased. These data strongly suggest that Thr 688 aids in Ca 2ϩ -mediated catalysis, either through appropriate positioning of the Ca 2ϩ ion in the active site or by providing appropriate electrostatics to the divalent cation. It is worth noting that the water nucleophile in the inverting catalytic mechanism is directly coordinated to the Ca 2ϩ ion (46), and an isosteric amide substitution of the adjoining general base residue (E599Q) reduced catalysis by 13,000-fold ( Fig. 4) (46). However, this mutation did not increase the binding affinity for the uncleaved substrate. Thus, altering the electrostatics of the general base function for activation of the water nucleophile does not account for the increased glycan binding affinity for the T688A mutant. An alternative role for the Thr 688 side chain could be a mechanical tethering of the Ca 2ϩ ion in a favorable position adjacent to the Ϫ1 subsite that is required for efficient catalysis. A similar effect of reduced catalysis and increased binding of an uncleaved substrate is found for the general acid mutant E330Q. Future studies on the structures of co-complexes between the T688A and E330Q mutants with uncleaved substrates or substrate analogs should provide insights into the roles of the glycone conformational changes and the enzyme-associated Ca 2ϩ ion in glycan hydrolysis.
The use of the E330Q mutant in SPR binding studies was also shown to be an effective tool in assessing the binding contributions of respective residues within the Man 9 GlcNAc 2 substrate. Applying a similar approach to map the contributions of oligosaccharide substrate residues for the other Class 1 mannosidases will reveal the molecular basis of substrate recognition and specificity for this diverse enzyme family. More importantly, the ability to measure detailed binding affinities and kinetics using the equivalent of the E330Q mutant as a model should provide critical information for the analysis of new selective inhibitors for class I mannosidases as potential targets for human protein misfolding disorders.