Analyses of Variant Acid β-Glucosidases

Acid β-glucosidase (GCase) is a 497-amino acid, membrane-associated lysosomal exo-β-glucosidase whose defective activity leads to the Gaucher disease phenotypes. To move toward a structure/function map for disease mutations, 52 selected single amino acid substitutions were introduced into GCase, expressed in an insect cell system, purified, and characterized for basic kinetic, stability, and activator response properties. The variant GCases from Gaucher disease patients and selected variant GCases from the mouse had decreased relative kcat and differential effects on active site binding and/or attachment of mechanism-based covalent (conduritol B epoxide) or reversible (deoxynojirimycin derivatives) inhibitors. A defect in negatively charged phospholipid activation was present in the majority of variant GCases but was increased in two, N370S and V394L. Deficits in saposin C enhancement of kcat were present in variant GCases involving residues 48-122, whereas ∼2-fold increases were obtained with the L264I GCase. About 50% of variant GCases each had wild-type or increased sensitivity to in vitro cathepsin D digestion. Mapping of these properties onto the crystal structures of GCase indicated wide dispersion of functional properties that can affect catalytic function and stability. Site-directed mutagenesis of cysteine residues showed that the disulfide bonds, Cys4-Cys16 and Cys18-Cys23, and a free Cys342 were essential for activity; the free Cys126 and Cys248 were not. Relative kcat was highly sensitive to a His substitution at Arg496 but not at Arg495. These studies and high phylogenetic conservation indicate localized and general structural effects of Gaucher disease mutations that were not obvious from the nature of the amino acid substitution, including those predicted to be nondisruptive (e.g. Val → Leu). These results provide initial studies for the engineering of variant GCases and, potentially, molecular chaperones for therapeutic use.

branched-chain amino acids). Furthermore, the relationships of genotype to phenotype in Gaucher disease should be based on known disruptive functional changes derived from mutations. For example, highly disruptive mutations could produce more significant effects on catalytic function and/or enzyme instability, leading to more severe or progressive disease manifestations. Also, knowledge of the distribution of the mutations and their functional effects on catalytic or stability properties of GCase could have significant implications for the design and/or development of new therapeutic agents for enzyme, chaperone, or genic therapeutic approaches. The vast majority of point mutations occur as compound heterozygotes with two different mutant alleles expressed in the same cells, obviating the direct characterization of the specific GCase properties. Separation of these alleles in recombinant systems allows for in vitro characterization of these properties.
Here, analyses of 42 point mutations identified in Gaucher disease patients, seven mutations of cysteine residues that are either involved in disulfide formation or are free, and three mutations of C-terminal histidines were conducted to gain insight into their effects on structure and function of GCase. These properties are correlated with the crystal structure of human GCase, along with comparisons of the properties of selected corresponding mutant mouse GCases to develop contextual insights into mutational effects.

Methods
Mutation Identification-The mutations were identified in genomic DNA from enzymatically confirmed Gaucher disease patients (i.e. those with defective acid GCase) from a variety of ethnic and demographic backgrounds. Two approaches were used with amplified genomic DNA: 1) exons alone were sequenced, or 2) the GCase genes were amplified by long range PCR and sequenced in their entireties. Sequences were analyzed by fluorescent labeling using the ABI capillary electrophoresis system. All sequence reads were verified by direct visualization following automatic base calls with Sequencher (GeneCodes; Ann Arbor, MI) or Mutation Surveyor (SoftGenetics, Inc., State College, PA). Many of the patients were submitted through the ICGG (International Collaborative Gaucher Group) program for GCase DNA sequence analysis. The majority of samples did not have readily identifiable clinical information and/or were heteroallelic for GCase mutations. Available phenotypic information is provided in Table 1. The objective of this work was to verify the pathogenic nature of the mutation (i.e. the predicted amino acid substitution) and its effects on in vitro enzymatic function. Additional mutagenesis was conducted to evaluate the effects of specific amino acids (e.g. the seven cysteines). Serine was chosen as an isosteric substitution for cysteines 4,16,18,23,126, and 248. The first four cysteines participate in disulfide bonds as follows: Cys 4 -Cys 16 and Cys 18 -Cys 23 . Cys 126 , Cys 248 , and Cys 342 are free (see "Results"). C342G and C16S are naturally occurring mutations from Gaucher disease patients (28,29). The mutant enzymes, R495H, R496H, and R495H/ R496H (a double mutant), were produced to evaluate the potential effects of the R495H substitution in the commercially available enzyme replacement product, Imiglucerase TM , and the mutation found in Gaucher disease patients, R496H, compared with the wild-type Arg 495 / Arg 496 enzyme, as determined by amino acid sequence analysis of the wild-type enzyme from human placenta (7).
Expression and Purification of GCases-Codons for amino acid substitution were produced individually in the cDNA of wild-type GCase and cloned into the plasmid pBluescript 4.5 (Invitrogen) following sitedirected mutagenesis (QuikChange; Stratagene). The GCase mRNAs from wild-type, N370S, V394L, D409H, and D409V homozygous mice (30) were isolated, reverse transcribed, sequenced, and cloned into the Bac-N-Blue vectors for recombination into baculovirus (see below). All cDNAs were completely resequenced to ensure that no spurious mutations were introduced. The wild-type and mutant cDNAs were inserted into the baculovirus by homologous recombination using the Bac-N-Blue system. The individual viruses were amplified in and purified from Sf9 cells as described (31,32). GCases expressed from these cDNAs were obtained from media of Sf21 cells that had been adapted to serum/ protein-free medium. Each GCase was partially purified by batch hydrophobic chromatography following n-butanol extraction as follows: 1) delipidation (n-butanol, 1:4, v/v), 2) batch binding to octyl-Sepharose (1:10, v/v), 3) step batch washing (2 bed volumes) with 50 mM phosphate, pH 5.8, containing 20 or 50% ethylene glycol; 4) GCase elution (100% ethylene glycol). All GCases had very similar purification properties requiring only minor modifications of the procedure. The only exception was the incorporation of protease inhibitor mixture III into the medium and butanol extraction steps for the ⌬36, F397S, D409H, and D409V mutants, since they were rapidly digested without added mixture III by proteases released from Sf21 cells into the SF900II medium. Following hydrophobic chromatography, all GCases could be stored for up to 3 months without evidence of proteolytic digestion. The wild-type, N370S, V394L, D409H, and D409V mouse GCases were expressed and purified as above to evaluate the potential effects of multiple amino acid substitutions on GCase properties from different species. The comparative properties for these specific GCase variants were from Grace et al. (31,32), although the properties of N370S were reverified here.
Additional GCase sequences from other mammalian species (chimpanzee, orangutan, porcine, fugu, partial rat, mouse, honey bee, and Caenorhabditis elegans) were obtained from public data bases and aligned using DNASIS 3.7 (Hitachi Software Engineering Co., San Francisco, CA). The canine GCase cDNA sequence was determined from a clone isolated from a constructed cDNA library using canine brain total RNA (Lambda AP kit).
Enzyme Purification and Crystallization-Wild-type GCase for crystallization was purified from Cerezyme™ to remove nonprotein components. During the conduct of these experiments, Sussman and co-workers (17,33) solved the GCase crystal structure at ϳ2.0 Å level. Our crystal structure was highly similar, except that the active site region had greater resolution than that of Dvir et al. (33). The deglycosylation pro-cedure developed here provides fully active GCase prior to crystallization. To achieve this, deglycosylated GCase was purified on octyl-Sepharose, and active fractions (70 -90% ethylene glycol) were pooled and exchanged into 10 mM sodium acetate, pH 5.2, immediately before the crystallization. This allowed for recoveries of 85-93% of initial enzyme activity. In comparison, the exchange into MES, pH 6.6 (33), was shown in preliminary experiments to irreversibly inactivate GCase. Final protein concentration was estimated by BCA protein assays to be 10 mg/ml. Activities for GCase to be crystallized were determined with 4MUG (4 mM) as substrate in the presence of brain phosphatidylserine and purified recombinant human saposin C in 25 mM citrate, 50 mM phosphate, pH 5.5. Recombinant human saposin C was expressed and purified as described (34). Crystals of N-glycanase-treated GCase were obtained by vapor diffusion at room temperature.
Relative Catalytic Activity-Comparisons of the catalytic activity, relative k cat , were determined by referencing to cross-reacting immunological material (CRIM) and expressed as CRIM specific activity (CRIM SA). This approach was needed, since the degree of purity of each GCase was not identical, but all activity toward the 4MUG substrate within each preparation was completely inhibitable by CBE, consistent with being a GCase. CRIM SA for each GCase variant was determined using immunoblots developed with anti-human GCase polyclonal antisera (27) and assuming equal reactivity of the GCase variants and wild-type GCase. Predetermined amounts of GCase activity were applied to SDS gels in amounts equivalent to standards of known masses of homogeneous human GCase (2.5, 5, and 10 ng). These amounts of GCase were known to provide readings in the linear range of the densitometer. The CRIM SA was determined from duplicates for each sample based on the standard curve densities.
Wild-type and Mutant GCase Characterization-Immunoblot analyses were as described using rabbit anti-human GCase polyclonal antiserum produced against purified human placental GCase (27). The pH optimum profiles were developed using citrate/phosphate buffers over the range of pH values from 4.4 to 7.2. pK a values were estimated from a diprotic model and varied by Ϯ0.3-0.6 pH units (n ϭ 3). Preliminary studies showed deoxynojirimycin and castanospermine to be rapidly reversible competitive inhibitors of all of the active enzymes used here, and K i values were determined from the equation, IC 50 ϭ K i (1 ϩ K m /[S]). Conduritol B epoxide (CBE) was shown by dilution experiments to be an irreversible inhibitor of all of the active enzymes used here. IC 50 values were determined under standard assay conditions. The standard GCase activity assay contained citrate/phosphate, pH 5.5, 0.25% sodium taurocholate, 0.25% Triton X-100, and 4 mM 4MUG. Assays in the presence of phosphatidylserine and/or saposin C did not contain either taurocholate or Triton X-100. For these assays, the GCases were preincubated at pH 5.5 with phosphatidylserine for 30 min at room temperature prior to the addition of saposin C.
Cathepsin D sensitivity for selected GCases was determined as follows. 2.9 pmol of each GCase variant were digested in the presence of cathepsin D (2 M; Calbiochem) in 20 l of 25 mM sodium acetate, pH 4.8, 50 mM NaCl, 1.25 mM EDTA, and 1.25 mM dithiothreitol at 37°C. After 48 h, the reaction mixtures were subjected to Western analyses using rabbit anti-human GCase polyclonal antiserum (27). The band densities of intact GCase (M r ϳ60,000) were quantified in digested and undigested aliquots using ImageQuant 5.2 (Amersham Biosciences). The results are expressed as the percentage of remaining intact GCase. The final enzyme preparations for the majority of GCases contained small M r contaminant proteins that varied between 10 and 30% of the total protein as estimated from stained SDS-PAGE. To ensure that the sensitivity toward cathepsin D was not altered by such contaminants, the pattern of fragments obtained on immunoelectroblotting was shown to be unaltered during the purification procedure and by doping of selected GCases from crude medium through final purification steps, with pure wild-type GCase in different concentrations. The pattern/ sensitivity of wild-type GCase was unaltered under these conditions, indicating that sufficient cathepsin D was used to avoid interference by nonspecific proteins. The amounts of variant GCases were estimated from densitometric quantification of immunoblots using pure wild-type GCase from imiglucerase as standard.

Mutations, Point Substitutions, and Characteristics of Mutant
Proteins-The variant GCase properties are shown in Tables 1 and 2 and in Figs. 1-6. Of the listed GCases, in Tables 1 and 2, the following have not been discovered in samples from Gaucher disease patients: the serine for cysteines at 4, 18, 23, 126, and 248 and R495H or R495H/ R496H. The majority of the natural variant substitutions occur in conjunction with another variant GCase allele (i.e. they are heteroallelic). All variants were expressed independently, and no interaction between GCase variants has been observed in double infections in the baculovirus systems (35).
The CRIM SA assessments using polyclonal antibodies provide a comparison of k cat relative to that from wild type. For the GCase variants found in Gaucher disease patients, three groups of CRIM specific activities were evident (Fig. 1). 1) Two GCases (⌬36 and E326K) had 35-50% wild-type CRIM SA. E326K was a stable enzyme with about 40 -50% of wild-type CRIM SA, although the mutation resulted in the addition of a bulky group with a complete charge change (i.e. Ϫ to ϩ). This variant has been found in phenotypically normal individuals as the heteroallele to L444P-or G202R-encoding alleles (36). The ⌬36 GCase, an in-frame codon deletion of threonine 36, could be partially purified in the presence of mixture III, and that enzyme had significant CRIM specific activity (ϳ35% of wild-type). 2) 15 variants with ϳ10% CRIM SA were found. 3) 25 variants had Յ5% CRIM SA. The M416V and C342G enzymes were stable in medium and could be produced in significant amounts but were essentially devoid of activity. The variants K198E, F397S, and R463P were rapidly degraded in medium, and the CRIM SAs were estimated directly from medium in the presence of protease inhibitors.
Several GCase variants were produced by "knock-in" approaches in mice (30). The corresponding mutant mRNAs were isolated from cultured skin fibroblasts of the respective homozygous mice, expressed in the baculovirus system, purified from media, and characterized (Table  1). These variants corresponded to the human Gaucher disease mutations encoding N370S, V394L, D409H, and D409V. The murine and human D409H and D409V GCases were unstable in medium surrounding Sf21 cells. For these murine GCases, mixture III was added to preserve significant GCase signals. The CRIM SA of the mutant murine GCases was ϳ11-18%. Those in human were comparable only for N370S and V394L (ϳ12-15%). The values for human D409H and D409V were very low (Ͻ0.2%) (32,37). The large S.D. values for estimates of mouse D409H or D409V derive from variations in the culture conditions, inhibition of proteases, and the preservation of these unstable mutant GCases. These results show that the specific mutations in mice and humans have highly similar enzymatic effects even with the 15% difference in amino acid context of the amino acid backbone (Fig. 8).
The results of active site interaction studies of the above variant GCases with CBE, deoxynojirimycin, and castanospermine are summarized in Table 1 and Figs. 2 and 3. CBE is a covalent inactivator of GCases with a two-step mechanism involving a reversible complex of CBE-GCase and a subsequent modification to the covalent complex. In comparison, deoxynojirimycin is a competitive inhibitor that forms a rapidly reversible inhibitor-GCase complex.
By comparing the effects of these inhibitors, one can dissect the binding and/or covalent complex formation components that could be significantly altered by the amino acid substitutions. The IC 50 or Castanospermine. c R495H GCase was purified and characterized in the baculovirus system, and also imiglucerase was purified from Cerezyme TM and contains R495H. The properties of these two GCases with R495H were identical. All other GCases were purified from medium of Sf21 cells infected with the corresponding baculovirus construct. PS fold activation was for equal molar amounts of intact enzyme, (E) t ϳ2 n M , in 8 M, or no PS at their optimal pH values. d ND, not done.      (Fig. 4). This deficit was about 5-fold (i.e. enhancements of 15-fold in wild type versus 2-5fold in the mutants). The exceptions included E326K, N370S, and V394L. The last two GCases, either human or mouse (Table 1), had excess activation compared with wild type (ϳ2-3-fold), as previously described (32,37), whereas E326K exhibited ϳ50% of wild-type activation. These results indicate that most of the Gaucher disease mutations lead to variant GCases that cannot develop an optimal phosphatidylserine-induced conformational change for catalysis.

TABLE 2 Properties of GCase cysteine and COOH-terminal histidine variants
In the presence of negatively charged phospholipids, the natural protein activator, saposin C, enhances wild-type GCase activity by an addi-tional ϳ2-fold (Fig. 5). The variant GCases R120Q, P122L, and L444P showed little effect of saposin C (150 nM), and variably deficient activations were noted over a narrow range of mutations that included R48Q, K79N, R120Q, P122L, and D127V. L264I showed about twice wild-type activation levels. These results imply additive activation effects of phosphatidylserine and saposin C that can function independently.
pH Optima for Variant GCase Activities-The optimum pH for activity of the mutant GCases and the estimated pK a values for this diprotic enzyme are listed in Tables 1 and 2. The majority of the active enzymes have wild-type values for these parameters, irrespective of the specific substitution. The D127V, E349K, G390R, and D399N have dis-  tinctly acidic shifts in pH optima with considerable acidic shifting of the entire curves, a mean 0.4 -0.6-pH unit leftward shift for both pK a values. These results suggest changes in the solvent environment near the active site residues and/or their interactions. Importantly, the variant GCases remain diprotic without loss of either arm of the relatively symmetric bell-shaped curves.
Sensitivity of Variant GCases to Cathepsin D-Cathepsin D sensitivity was used to evaluate the conformational effects of the specific Gaucher disease mutations (Fig. 6). 13 variants displayed extreme sensitivity to this treatment with little or no intact enzyme and/or enzyme activity remaining following the 48-h incubation period. Nine GCases had nearly wild-type or slightly improved resistance, including E326K and E349K. R48Q, L174F, L264I, and N382K had ϳ25-50% of the wild-type resistance.
Additional Mutations in GCase and Their Effect on GCase Properties-The disulfide bonds in wild-type GCase are between residues Cys 4 and Cys 16 and residues Cys 18 and Cys 23 . These are located in an isolated loop structure (domain 1) in the crystal (33) (Fig. 7). To evaluate the importance of these and the free cysteines at positions 126, 248, and 342, GCases containing singly mutated cysteines at each position were produced and characterized. Substitutions of serine for cysteine, an isosteric replacement, at 4, 16, 18, or 23 produced enzymes with severe decreases in catalytic activity and were easily detected on immunoblots (i.e. they were stable proteins). Similarly, C342G GCase had Ͻ0.2% CRIM SA. In comparison, the GCases C126S and C248S had ϳ80 and ϳ30% of wild-type GCase CRIM SA. These results indicated that any one of the cysteines involved in disulfide formation is essential to formation and/or preservation of an active enzyme. Cys 342 is near the active site nucleophile, Glu 340 (14), with potentially significant conformational and local effects. The C126S and C248S GCases had wild-type values for K i and K m ( Table 2). C126S had wild-type levels of activation by phosphatidylserine. The comparable value for C248S was ϳ35%. Both of these GCases had pH optimum curves that were indistinguishable from wild type ( Table 2).
The GCase variants with substitutions of histidine at positions 495 and/or 496 indicated that the carboxyl tail of the enzyme has important functions in catalytic activity. The histidine substitutions led to stable enzymes in the Sf21 medium. The R495H and R496H GCase variants had wild-type and severely diminished levels of CRIM SA, respectively. The R495H is present in the porcine GCase, but the R496H was discovered as a heteroallele in a patient with Gaucher disease (38). Curiously, the double mutant enzyme R495H/R496H showed recovery of enzyme activity to nearly normal levels with wild-type K i and K m values ( Table  2). R495H and the double mutant had 100% and ϳ60% of wild-type activation by phosphatidylserine, respectively, and comparable pH optima curves (Table 2).
Crystal Structure of Partially Deglycosylated GCase-The active site residues Glu 340 and Glu 235 are in the center of the domain 3 ␤-barrel. The region around these residues is the least ordered in the crystal structure. In particular, very weak electron density is observed for residues 313-319 and 342-346, both of which are in loops that form part of the opening to the active site. The closest approach between the side chains of active site residues Glu 340 and Glu 235 is 4.4 Å in molecule A and 3.9 Å in molecule B, the two asymmetric monomers in the GCase crystal. This is closer than the reported 5.2 Å (33). Also, Tyr 313 that was reported to form a hydrogen bond with the side chain carboxyl oxygen of Glu 235 is poorly ordered in molecule A of the present structure and is too far to form a strong hydrogen bond with Glu 235 in molecule B.

DISCUSSION
The overall objectives of this research were to confirm the pathogenicity of the various mutations detected in patients with Gaucher disease, to determine in vitro properties that were altered by the specific point mutations, and to gain insight into the structure/function correlations for GCase. The rationale for the in vitro studies was that most of the mutations occurring in Gaucher disease patients are heteroallelic, thus precluding determination of their unique properties. This is true for the kinetic properties and sensitivity of the mutant GCase to proteases and also is relevant to the misfolding properties induced by GCase mutations as assessed in cultured cells (39,40). Thus, the overall concept was to determine the degree of abnormality induced by the specific point mutation and to develop a correlation map with our crystal structure of GCase.
Essentially all of the mutant enzymes listed in Table 1 have significantly attenuated catalytic rate constants, k cat , as estimated by the CRIM specific activity. This approach was used because equal degrees of purity could not be obtained for all mutant GCases. For these analyses, polyclonal anti-human GCase antibody was used to quantify the amount of enzymatic protein present in each normal or mutant sample under the assumption that for the majority, if not all, of the mutant GCases, the polyclonal antibody would have equal avidity and interaction with the proteins. Thus, within these limitations, the catalytic rate constants of the majority of the variant GCases were decreased to below 10% of wild type; several had no activity. This shows that these mutations could be  Table 1 for specific substitutions. The substitutions indicated with an asterisk are those GCases with Ն10% of wild type. Those in boxes (N370S and V394L) have increased responsiveness to phosphatidylserine, and those in green font had wild-type resistance to in vitro cathepsin D digestion. FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7 pathogenic to the patients. This diminished catalytic capacity was coupled mostly to abnormalities in kinetics, inhibition, activation by phosphatidylserine or saposin C, or in vitro proteolytic sensitivity to cathepsin D. Using the cathepsin D assay as an assessment tool of the overall conformational changes of the mutant GCases, about half of the mutants led to an increased sensitivity of the protein to in vitro cathepsin D digestion. Some of the proteins (e.g. the ⌬36) were highly sensitive to endogenous insect cell proteases but, in the presence of protease inhibitors, could be purified and had substantial residual activity and high sensitivity to cathepsin D. Thus, the majority of the mutant GCases with point mutations from Gaucher patients led to major changes in proteolytic sensitivity in vitro and catalytic deficits.

Variant Human Acid ␤-Glucosidase
About 50% of the various point mutants had Ͻ5% of the wild-type GCase activity. These were scattered throughout our crystal structure, but more were interior in hydrophobic regions and around the active site. Since the mutations were selected from patients with Gaucher disease, it might be anticipated that low catalytic activity would be observed. It is interesting that the interior location produced significant conformational changes that led to severe catalytic deficits that were not detected with active site-directed inhibitors or substrates. Also, throughout phylogeny (Fig. 8), none of the identified Gaucher mutations occurred at a conserved residue region from humans to worms.
The E326K GCase had about 50% of the wild-type CRIM specific activity and a significantly diminished activation by phosphatidylserine. However, compound heterozygosity with G202R/E326K and L444P/ E326K genotypes and half-normal GCase activity levels indicated that normalcy for periods of 1-4 decades can exist in the presence of an E326K allele (36). Interestingly, the E349K GCase and E326K are approximately equally spaced from the active site in the crystal structure, but E349K had much greater effects on the CRIM specific activity

Variant Human Acid ␤-Glucosidase
and other enzymic properties than did E326K. Both enzymes, however, had equal and nearly wild-type in vitro sensitivity to Cathepsin D digestion. Thus, although the Glu 3 Lys change would be predicted to have major disruptive affects, such alterations were position-dependent.
Interestingly, the overall impact of most of the Gaucher disease mutations on phosphatidylserine activation was substantial. Only three of the point mutants had normal to increased responsiveness to phosphatidylserine, indicating that the vast majority of mutations affect the global conformation of the enzyme, leading to decreases in catalytic power. Interestingly, N370S, V394L, and I161N flank the active site and are Ͼ7 Å from the active center residues (Fig. 7). However, the binding of deoxynojirimycin and conduritol B epoxide was altered by 7-10-fold. Other Asn 370 mutations (i.e. Asp, Glu, or Gln substitutions for Asn) decrease catalytic capacity without concomitant changes in active site binding functions (15). This indicates that the disease mutation in humans, N370S, is residue-specific with local affects on active site function. Interestingly, an apparently nondisruptive substitution, L371V, alters the CRIM-specific activity but has no effect on active site interaction with the inhibitors, whereas the neighboring N370S substitution does. Asn 370 is not in proximity to the deeper aspects of the active site and has no obvious direct interaction at the active site based on its placement in the crystal structure (Fig. 7).
Five valine 3 leucine substitutions and one leucine 3 valine substitution have been discovered in patients with Gaucher disease (28,(41)(42)(43)(44)(45). The V394L GCase has major kinetic changes similar to those in N370S. V394L homozygotes have not been reported in humans, but compound heterozygotes with L444P or 84GG (a null) have neuronopathic Gaucher disease (46 -48), indicating a major disruption of the glucosylceramide pathway. Val 394 is located in domain 1 opposite the active site relative to Asn 370 and forms part of a ring of amino acids surrounding the entrance to the active site (Fig. 7) adjacent to Arg 395 that extends out from the surface. It is possible that the V394L substitution alters active site access. Indeed, during the final preparation of this paper, the crystal structure of GCase with bound CBE showed this to result from destabilization of an open loop conformation of residues Val 394 -Asp 399 (17). V15L also leads to very severe disease. Val 15 is in a hydrophobic pocket made up of Phe 9 , Leu 354 , and Tyr 412 , where replacement with the larger leucine side chain would be likely to result in steric clashes. Val 15 is also located in domain 1 near the first glycosylation site and could interfere with disulfide formation, similar to the lack of N-glycosylation at Asn 19 (5). Thus, replacement of the ␤-branched hydrophobic residue, Val, with the larger ␥-branched leucine leads to substantive and major alterations in enzyme function, suggesting a very limited target region for acceptable changes within the GCase molecule.
None of the mutations had a major impact on saposin C activation of the mutant GCases, except for those residues from 48 through 122. With these residue changes, there was some diminished responsiveness to saposin C with preserved interactions at the active site. Thus, no obvious binding site for saposin C is evident from such analyses.
GCase also appears to have somewhat malleable properties within the context of the mutations. For example, N370S, V394L, and D409H/V were produced in mice to mimic human mutations. These mice developed glucosylceramide storage and/or had diminished enzyme activity in their specific cell types (30). Also as shown here, the properties of the isolated mutant mouse enzymes are very similar to those in humans. For example, the N370S and V394L GCases have essentially identical kinetic properties and responses to phosphatidylserine and saposin C, although the background amino acid sequence differs by ϳ15% across the entire protein. Interestingly, the physiologic effects of the N370S mutation in the mouse are quite different from those in human (30). N370S homozygosity in humans is associated with less severe Gaucher disease, whereas in the mouse it is lethal. Since the properties of N370S are essentially identical in mice and humans, the lethality must relate to the differences in substrates that are present in mice and humans (49,50). The lethality of the N370S mutation in mice results from a defect in the skin permeability barrier, similar to that observed in the GCase null mouse (30,51). It is interesting that the composition of glucosylceramide in the skin of mice differs from that in a human by having longer average fatty acid acyl chain lengths (52). This suggests that the longer acyl chain glucosylceramides may be poor substrates for N370S enzymes with resultant effects on the skin permeability barrier, the cause of lethality in N370S/N370S mice (30). Indeed, in vitro hydrolysis of glucosylceramides with varying fatty acid acyl chains by partially purified GCases from unaffected and Ashkenazi Jewish Gaucher disease type 1 patients showed a progressively more efficient cleavage (ϳ2-fold) of C12-C18 glucosylceramides by wild-type GCase than by the N370S GCase (32).
The present structure and that of Dvir et al. (33) show disulfide bonds between cysteines 4 and 18 and between cysteines 16 and 23. The remaining cysteines at residues 126, 248, and 342 are free. Interestingly, the first four cysteines are conserved in all species, as is cysteine 342, which is near the catalytic nucleophile, Glu 340 . Cysteine 126 is conserved in all species except the honeybee, and cysteine 248 is conserved only in mammals (Fig. 8). Since the disulfide structures occur within a separate domain, domain 1, on the periphery of the main GCase structure and glycosylation occurs at residue Asn 19 , the folding of GCase has been proposed to be a vectoral process determined by proper formation of this first domain 1 (5). The lack of glycosylation at Asn 19 during synthesis leads to an inactive enzyme as does substitution of serine at any one of the first four cysteines (5). However, as shown here, nearly complete deglycosylation under appropriate conditions can produce an enzyme with wild-type activity. This implies that co-translational glycosylation plays an important role in disulfide formation. These data suggest that the glycosylation and the formation of the disulfides in this part of domain 1 provide a nidus for the directed folding of normal GCase.
If mutated, the cysteines at residues 126, 248, and 342 might be expected to have differential effects on GCase based on different conservation in phylogeny. Indeed, substitution of a glycine at Cys 342 leads to a catalytically inactive enzyme that has been found in Gaucher disease patients. The proximity of this residue to the catalytic nucleophile might have predicted significant effects on the active site function. C126S had nearly wild-type properties. C248S produced significant decreases in CRIM specific activity that might result from its proximity to the active site but not from phylogenetic comparisons (Fig. 8). Recent studies have suggested that the reduction state of the three free cysteines is critical to the preservation of catalytic activity and prevention of the enzyme aggregation. 4 Thus, these free cysteines have dual roles in the preservation of enzyme activity that might result from being nearly equidistant from and surrounding the active site.
Also of interest were the carboxyl-terminal histidines. Curiously, the R496H occurs as a pathogenic mutation in Gaucher disease, and its neighbor, R495H, is present in the enzyme replacement product, Imiglucerase TM . The R495H GCase has normal properties, whereas the R496H GCase mutation has little, if any, in vitro catalytic activity. More surprising is that the double mutant R495H/R496H leads to ϳ90% reconstitution of enzymatic activity. Arg 496 forms hydrogen bonds with Asp 474 , stabilizing the fold of domain 2, whereas the side chain at posi-tion 495 is on the surface and unlikely to participate in stabilizing interactions. In mammals, substitutions can occur at Arg 495 with Cys and/or His in the orangutan and pig, respectively. In comparison, Arg 496 is conserved in all mammals. These two arginine residues at the carboxyl terminus are not known to form salt bridges or participate in a lid of the active site as occurs in other lipases. Certainly, these are not evident from the crystal structure. Two loops involving Ser 345 -Glu 349 and Val 394 -Asp 399 apparently control active site access (17), but their roles in enzyme stability or in catalytic activity remain to be elucidated.
Whereas Gaucher-disease associated mutations are not restricted to a particular surface or the active site of the enzyme, some generalizations can be drawn in correlating the mutational analyses reported above and the three-dimensional structure of GCase. In particular, we discuss here three classes of residues: those on the enzyme surface, residues at domain interfaces, and residues in the hydrophobic core of the individual domains.
Of the residues reported in Table 2 Only M416V and R463P completely inactivate the enzyme. Interestingly, mutation of residues between the ␣ and ␤ portions of the domain 3 ␤-barrel reduce enzymatic activity to below 10%, as seen for the mutations M123V, I161N, L185F, G193E, F259L, L264I, R353W, Y363C, and D380H. Similarly, changes in the ␤-barrel core of domain 3 in the active site space (R120Q and C342G) and in the active site opening (D127V, V394L, and F397S) also substantially reduce activity. Both Arg 120 and Cys 342 are in close proximity to the active site and might be expected to have direct catalytic roles. The D127V mutation leads to modest changes in K i values for CBE, deoxynojirimycin, and castanospermine, consistent with Asp 127 participating in tight hydrogen bonding to the cyclohexitol of CBE bound to Glu 340 (17). The most modest reduction in enzymatic activity was associated with the surface mutation E326K (42% activity) in domain 3 and localized to a surface of the enzyme that has relatively few identified mutations in this group.
Our crystal structure provides some different information compared with that of the original structure proposed by Dvir et al. (33). Premkumar et al. (17) have redissolved such crystals and shown them to retain activity. However, in our hands, the published procedure leads to a completely inactivated GCase and, consequently, some direct effects on the structure in and around the active site. Indeed, the vast majority of differences in the crystal structure reported here and that previously reported by Dvir et al. (33) are near the active site. Clearly, more work on the structure of GCase remains, including the full elucidation of the active site structure and the docking of substrates to the enzyme. The available crystal structures do not provide insight into potential subsites for substrate or alkyl glycon binding to the active site as have been proposed from kinetic studies (31,53).
From the distribution of the mutations in Gaucher disease and the heterogeneity of the physical and kinetic properties of GCase resulting from those mutations, it is not clear that predictions can be made from GCase structure to function. There appear to be major sensitive areas leading to significant alterations in catalytic activity and, in particular, the minor changes of branched chain amino acids and other such changes that are not compatible with active forms of the enzyme.
Note Added in Proof-Wild-type GCase for crystallization was purified from Cerezyme TM ro remove non-protein components. The final enzyme was electrophoretically homogenous as evidenced by a tight single band on SDS-PAGE and the absence of extraneous, detectible cross-reacting bands by immunoblotting using rabbit anti-human GCase polyclonal antiserum. For crystallization, the pure GCase was deglycosylated as indicated below. Cerezyme TM (4.5 mg of GCase) was dissolved in H 2 O (1.6 ml, 4°C). The non-enzyme components of the Cerezyme TM preparation were removed by dilution and reconstitution using an ultrafiltration with YM10 Microcons. To this solution was added 0.5 M NaPO 4 , pH 7.5, to a total volume of 2 ml. Nonidet P-40 was added to a final concentration of 1% and N-glycosidase F (10,000 units) as added. The deglycosylation reaction was at room temperature for 24 h. The extent of deglycosylation was monitored by increasing M r on 7.5% SDS-PAGE. The final GCase M r was ϳ56,000. This migration was evidence of nearly complete removal of the oligosaccharide changes, since it was not different from N-glycanase-digested GCase obtained under denaturing conditions. The final solution was subjected to octyl-Sepharose column chromatography using ethylene glycol as the eluting agent. GCase activity containing fractions (70 -90% ethylene glycol) were pooled and exchanged into 10 mM sodium acetate, pH 5.2, immediately before crystallization procedures. Activities for GCase crystallization were determined with 4 mM 4MUG as substrate in the presence of brain phosphatidylserine and purified recombinant human saposin C in 25 mM citrate, 50 mM phosphate, pH 5.5. Crystals of Endo-F-treated GCase were obtained by vapor diffusion at room temperature. A 7.5 mg/ml solution of protein (in 10 mM sodium acetate, pH 5.2) was mixed with half the volume of 1.2-1.4 M ammonium sulfate, 96 mM sodium citrate, pH 6, and 10 mM MgCl 2 and equilibrated against a reservoir of 1.2-1.4 M ammonium sulfate, 96 mM sodium citrate, pH 6, 10 mM MgCl 2 . Crystals appeared between 2 and 7 days from setup. The crystals were stabilized in 1.4 M ammonium sulfate, 96 mM sodium citrate, 10 mM MgCl 2 , and 20% glycerol and then flash-frozen directly in a liquid nitrogen stream. The crystals diffracted beyond 2.5 Å using a CuK␣ source from a rotating anode generator (RU200) and belong to the space group C222 1 with unit cell dimensions of a ϭ 108.9 Å, b ϭ 284.8 Å, c ϭ 91.6 Å and two molecules in the unit cell. Diffraction data were recorded at 100 K on an RAXIS IV image plate detector. The images were indexed and the reflections integrated, scaled and post-refined with the HKL package (64). Since the space group and cell dimensions are similar to those reported by Dvir et al. (33), the published coordinates (1 Protein Data Bank code OGS) stripped of side chains and solvent molecules were used as a model in molecular replacement. The structure was refined by a combination of rigid body refinement followed by simulated annealing and positional refinement in CNS (65), using all data to a maximum resolution of 2.5 Å. Simulated annealing omit maps were systematically calculated and examined during the refinement. The final model included residues 1-497 and had R ϭ 19.6% and free R ϭ 25.6%. The coordinates have been deposited in the Protein Data Bank with accession code 2F61.