Role of the Cysteine Residues in the α1,2-Mannosidase Involved inN-Glycan Biosynthesis inSaccharomyces cerevisiae

The Saccharomyces cerevisiae α1,2-mannosidase, which removes one specific mannose residue from Man9GlcNAc2 to form Man8GlcNAc2, is a member of a family of α1,2-mannosidases with similar amino acid sequences. The yeast α1,2-mannosidase contains five cysteine residues, three of which are conserved. Recombinant yeast α1,2-mannosidase, produced as the soluble catalytic domain, was shown to contain two disulfide bonds and one free thiol group using 2-nitro-5-thiosulfobenzoate and 5,5′-dithiobis(2-nitrobenzoate), respectively. Cys485 contains the free thiol group, as demonstrated by sequencing of labeled peptides following modification with [3H]ICH2COOH and by high performance liquid chromatography/mass spectrometry tryptic peptide mapping. A Cys340-Cys385 disulfide was demonstrated by sequencing a purified peptide containing this disulfide and by tryptic peptide mapping. Cys468 and Cys471 were not labeled with [3H]ICH2COOH and a peptide containing these two residues was identified in the tryptic peptide map, showing that Cys468 and Cys471 form the second disulfide bond. The α1,2-mannosidase loses its activity in the presence of dithiothreitol with first order kinetics, suggesting that at least one disulfide bond is essential for activity. Mutagenesis of each cysteine residue to serine showed that Cys340 and Cys385 are essential for production of recombinant enzyme, whereas Cys468, Cys471, and Cys485 are not required for production and enzyme activity. These results indicate that the sensitivity to dithiothreitol is due to reduction of the Cys340-Cys385 disulfide. Since Cys340 and Cys385 are conserved residues, it is likely that this disulfide bond is important to maintain the correct structure in the other members of the α1,2-mannosidase family.

The processing ␣1,2-mannosidase present in the endoplasmic reticulum of Saccharomyces cerevisiae removes one specific mannose residue from Man 9 GlcNAc 2 to form Man 8 GlcNAc 2 during the formation of N-linked oligosaccharides (1)(2)(3)(4). Its gene (MNS1) encodes a type II membrane protein of 63 kDa with no significant cytoplasmic tail, an N-terminal transmembrane domain, and a large C-terminal catalytic domain (5). The yeast ␣1,2-mannosidase exhibits significant similarity in amino acid sequence and topology with ␣1,2-mannosidases cloned from rabbit, mouse, and human (71-73 kDa) that are essential for the formation of complex and hybrid N-linked oligosaccharides (6 -8). The yeast and mammalian proteins are about 35% identical in amino acid sequence in their C-terminal catalytic regions. Based on sequence homology and common properties (9), these enzymes were grouped as Class 1 ␣1,2mannosidases. They all contain an EF-hand Ca 2ϩ binding consensus sequence and require Ca 2ϩ for activity. In addition, they are inhibited by 1-deoxymannojirimycin and do not use -nitrophenyl-␣-D-mannopyranoside as substrate. The yeast ␣1,2mannosidase has a very high specificity for removal of a single mannose residue on Man 9 GlcNAc 2 , whereas the mammalian enzymes can remove up to four mannose residues from Man 9 GlcNAc 2 to form Man 5 GlcNAc 2 . The mammalian enzymes hydrolyze ␣-Man1,2␣-Man-OMe, whereas the yeast ␣1,2-mannosidase cannot hydrolyze this disaccharide. The smallest oligosaccharide substrate for the yeast ␣1,2-mannosidase is ␣-Man1,2␣-Man1,3␣-O(CH 2 ) 8 COOCH 3 , but it is a very poor substrate (K m ϭ 9 mM) (10). Recently, ␣1,2-mannosidases have also been cloned from Drosophila melanogaster (11), Penicilium citrinum (12), and Aspergillus saitoi (13), which have similar amino acid sequences to the yeast and mammalian enzymes. The Drosophila mas-1 gene encodes two ␣1,2-mannosidases (72.5 and 75 kDa) that differ in their N-terminal region and have the same topology as the yeast and mammalian ␣1,2mannosidases. The P. citrinum and A. saitoi ␣1,2-mannosidase genes encode secreted proteins of 56 -57 kDa with a cleavable signal peptide. Unlike the other members of this family, they do not contain an EF-hand Ca 2ϩ binding consensus sequence and do not require Ca 2ϩ for activity.
Little is known about the structure and mechanism of catalysis of any of the Class 1 ␣1, 2-mannosidases (recently named Family 47 in the classification of glycosyl hydrolases in Release 34.0 of the SWISS-PROT Protein Sequence Data Bank), and the three-dimensional structure is not known. Until recently, a major difficulty has been the purification of sufficient enzyme for study. However, we can now produce milligram quantities of the catalytic domain of the yeast ␣1,2-mannosidase (14). The yeast processing ␣1,2-mannosidase is the first member of this family that can be produced in sufficient quantity to study its structure and its mechanism of catalysis, and we have shown recently that it is a glycosidase of the inverting type (15).
In the present work we demonstrate that the yeast ␣1,2mannosidase has two disulfide bonds and one sulfhydryl group in its catalytic domain. Their location is documented by peptide analysis, and their respective roles in enzyme activity is established by site-directed mutagenesis. Only one of these two disulfide bonds is essential for catalytic activity, and the free sulfhydryl residue is not required. Taq polymerase, endoprotease Asp-N (sequencing  grade), and DTT 1 were purchased from Boehringer Mannheim. Oligonucleotides were synthesized at the Sheldon Biotechnology Centre (McGill University, Montréal, Canada). Restriction enzymes were from either Pharmacia Biotech Inc., New England Biolabs, or Boehringer Mannheim. The Pichia expression kit was from Invitrogen and includes the Pichia pastoris strains GS115 and KM71 and the vector pHIL-S1. CNBr was from Fluka. Urea (Ն99%) and iodoacetic acid were from ICN. DTNB, endoprotease Glu-C (sequencing grade) from Staphylococcus aureus V8, and TPCK-treated trypsin were from Sigma. Water for HPLC was obtained from a Barnstead nanopure water purification system or from Baxter. All other methods require water from a Milli-Q system with an Organex-Q cartridge. Acetonitrile suitable for chromatography was from BDH or Fisher. Trifluoroacetic acid (HPLC/Spectro Grade, Sequanal Quality) was from Pierce or trifluoroacetic acid (99% pure) was from Aldrich. Oligosaccharide substrates were obtained as described previously (14). All other chemicals were reagent grade.

Materials-
Plasmid Construction and Site-directed Mutagenesis-Escherichia coli DH5␣ and DH10 were used as the host for plasmid manipulations. The plasmid pBS9.5 contains the whole open reading frame of the MNS1 gene in pBluescript (16). The following oligonucleotides were used to isolate the DNA sequence encoding the catalytic domain (nucleotides 64 -1651 corresponding to amino acids 22-549) of the ␣1,2mannosidase from pBS9.5 by PCR: (i) a sense 5Ј oligonucleotide ATACTCGAGTGCCATGGTACGAACACTTTG containing an XhoI site and (ii) a 3Ј antisense oligonucleotide GGTGGATCCCTACAACGAC-CAACCTGTG containing a BamHI site. Preparative PCR was carried out as described previously (14). The specific PCR product was digested with XhoI and BamHI and ligated into the XhoI/BamHI sites of the pHIL-S1 vector to produce the plasmid YpHMNS1. Cysteine to serine mutations were accomplished by the unique site elimination procedure (17) (U.S.E. mutagenesis kit, Pharmacia) using the plasmid YpHMNS1 and the following oligonucleotides: GACCACCTCGTAAGCTTTATGGG for the C340S mutation, GGGATAACTGACACTAGTTATCAAATGTA-CAAGC for C385S, CTTTGAAAATACTAGTGTTGATTGTAATGACCC for C468S, CTGTGTTGATTCTAATGACCCAAAATTAAGG for C471S, and GGCGGTTCACTAGTTTAAGTGATTCTATCACGTTACCTAC for C485S. The plasmids YpHC340S, YpHC385S, YpHC468S, YpHC471S, and YpHC485S were thus constructed from the wild type YpHMNS1 plasmid. The entire coding region of all the constructs was sequenced by the dideoxy method (18).
Production of Recombinant ␣1,2-Mannosidase-The secreted recombinant ␣1,2-mannosidase lacking the transmembrane domain was produced using two different yeast expression systems, one in S. cerevisiae as already described (14) and the other in P. pastoris. BglII-digested plasmids were transformed by the spheroplast method or by electroporation into the P. pastoris strains GS115 (his4) or KM71 (his4, aox1) as described in the Pichia expression kit manual obtained from Invitrogen. Histidine-independent transformants were selected and subsequently screened for methanol utilization when the GS115 strain was used. Use of the KM71 strain does not require selection for methanol utilization. BMGY and BMMY media were prepared according to the Pichia expression kit manual. His ϩ , Mut Ϫ yeast cells were grown at 30°C (250 -300 rpm) for 48 h in BMGY medium (buffered with 100 mM potassium phosphate, pH 6.0), and then centrifuged and resuspended with BMMY medium (buffered with 100 mM potassium phosphate, pH 6.0) in 20 -40% of the original volume, and incubated for an additional 1-4 days at 30°C (250 -300 rpm). If the cells were induced for more than 2 days, methanol was added to a final concentration of 0.5% (v/v) after 2 days of induction. Cells were separated from the medium by centrifugation. In order to screen for clones expressing ␣1,2-mannosidase, 5 l of medium were assayed for ␣1,2-mannosidase activity. The assay mixture contained 8700 dpm [ 3 H]Glc 1 Man 9 GlcNAc, 1 mg/ml BSA, 0.1 M PIPES (pH 6.5), and 1 mM sodium azide in a total volume of 20 l and was incubated for 1 h at 37°C. In order to screen for intracellular expression of recombinant enzyme, the cells were vortexed 4 -5 min at 4°C in the presence of glass beads in 10 mM sodium phosphate buffer, pH 6.8, containing 1 mM sodium azide. Glass beads, buffer, and cells were added in a ratio of 1:1:0.5. The glass beads were allowed to settle, and aliquots of the supernatants were analyzed by Western blotting.
␣1,2-Mannosidase Assays-Unless otherwise indicated, enzyme samples were diluted in 10 mM PIPES (pH 6.5) containing 1 mg/ml BSA and 7.5 l were assayed under standard conditions with 0.2 mM Man 9 GlcNAc (17,400 dpm [ 3 H]Man 9 GlcNAc) as described previously (14). K m values for the mutant and wild type enzymes present in the medium were determined as described previously (14).
Quantitation of Disulfide Bonds and Free Sulfhydryl Groups-The disulfide bonds were quantitated using 2-nitro-5-thiosulfobenzoate, which was synthesized from DTNB (19). The reaction was carried out in a solution of pH 9.5 containing 0.2 M Tris-HCl, 0.1 M sodium sulfite, 3 mM EDTA, 3 M guanidine thiocyanate, and 0.5 mM NTSB. 100 l of the reagent solution were added to the protein solution (7-15 l containing 26 -147 g of ␣1,2-mannosidase). The reaction was monitored at 412 nm until a constant absorbance reading was achieved (45-90 min).
Carboxymethylation with 2-[ 3 H]ICH 2 COOH-Recombinant ␣1,2mannosidase (1.2 mg) in 2 mM sodium phosphate (pH 6.8), 1 mM sodium azide was lyophilized. Reaction with iodoacetic acid was carried out according to Carr et al. (21), with a few modifications. Iodoacetic acid (recrystallized in chloroform) was added to a final concentration of 0.1 M in 0.5 M Tris-HCl (pH 8.2), containing 2 mM EDTA and 6 M Gdn-HCl. The pH was adjusted to 8.2 with ammonium hydroxide and the solution was purged with nitrogen. 500 Ci of dried 2-[ 3 H]ICH 2 COOH (172 mCi/mmol, Amersham) were added to 2.4 ml of this solution, and this was then added to the lyophilized protein. The reaction was carried out on ice in the dark under a nitrogen atmosphere for 140 min. Three ml of 1% acetic acid were added to stop the reaction, and the sample was extensively dialyzed against 1% acetic acid (six changes of 1.0 liter) and lyophilized. Reaction with unlabeled 0.1 M iodoacetate was performed as above.
Cleavage of ␣1,2-Mannosidase (Scheme I)-Lyophilized ␣1,2-mannosidase was treated with a 100-or 500-fold molar excess (with respect to methionine content) of CNBr in 6 M Gdn-HCl/0.2 M HCl (22). The reaction was allowed to proceed for 24 h in the dark at room temperature, 1 ml of water was added, and the sample was lyophilized. One ml of water was added again and the sample was lyophilized and then stored at Ϫ80°C. CNBr-treated peptides were fractionated on a Sephadex G-50 column (1 ϫ 120 cm) using 0.1 M or 0.2 M acetic acid as solvent, monitoring the eluent at 230 nm. Five included peptide fractions, P 1 -P 5 , were collected in their order of elution. Aliquots from each of these fractions were subjected to HPLC before and after reduction. A set of peptides eluting between 44 and 48 min in fraction P 3 , and to a 1 The abbreviations used are: DTT, dithiothreitol; BMGY, buffered glycerol-complex; BMMY, buffered methanol-complex; BSA, bovine serum albumin; CM-Cys, S-carboxymethylcysteine; DTNB, 5,5Ј-dithiobis(2-nitrobenzoate); Gdn-HCl, guanidine hydrochloride; HPLC, high pressure liquid chromatography; MS, mass spectrometry; NTB, 2-nitro-5-thiobenzoate; NTSB, 2-nitro-5-thiosulfobenzoate; PAGE, polyacrylamide gel electrophoresis; PCR, polymerase chain reaction; PIPES, piperazine-N,NЈ-bis(2-ethanesulfonic acid); TPCK, L-1-tosylamido-2phenylethyl chloromethyl ketone. SCHEME I lesser extent in fraction P 2 , were observed to shift upon reduction. None of the other fractions (P 1 , P 4 , or P 5 ) contained peptides that changed elution position upon reduction.
For endoprotease Asp-N digestion, peptides were dried under a nitrogen stream and were then dissolved in 50 mM potassium phosphate (pH 6.0), 0.1 M Gdn-HCl at 1.25 g/l and treated with 0.8 g of endoprotease Asp-N/100 g of peptides for 16 h at 37°C. Gdn-HCl was added to help solubilize the peptides. Glacial acetic acid was added to a final concentration of 10% (v/v) to stop the reaction, and the sample was stored at Ϫ20°C.
For endoprotease Glu-C digestion, peptides were lyophilized and resuspended in 30 mM sodium phosphate (pH 7.8), containing 1.2 M urea and 2 mM EDTA. Urea was added to help solubilize the peptides. The peptides were digested at a concentration of 1.5 g/l with endoprotease Glu-C at a concentration of 13 g/100 g of peptides. The digestion was allowed to proceed for 24 h at 25°C and terminated by the addition of glacial acetic acid to 10% (v/v). The sample was stored at Ϫ20°C.
For trypsin digestion, recombinant ␣1,2-mannosidase was lyophilized and resuspended in 8 M urea, 0.1 M potassium phosphate (pH 6.5) at a concentration of 5 g/l. The protein was denatured by sonication for 1 min followed by incubation at 37°C for 5 min. This was repeated three times. The solution was then incubated at 37°C for 30 min. An equal volume of 0.1 M potassium phosphate buffer (pH 6.5) was then added before the addition of 1 l of trypsin. TPCK-treated trypsin was prepared in 0.1 mM HCl and was added to yield a final concentration of 5 g/100 g of ␣1,2-mannosidase. The digestion was allowed to proceed at 37°C for 5 h and stopped by freezing at Ϫ80°C. For the reduced sample, 4 times the volume of 0.5 M Tris-HCl (pH 8.5) containing 0.375 M DTT was added to an aliquot of the trypsin-digested ␣1,2-mannosidase and the mixture was incubated at 37°C for 10 min, then stored immediately at Ϫ80°C.
For reduction of CNBr-treated and endoprotease Asp-N-digested peptides, peptide solutions were dried in a vacuum concentrator (Savant), then reconstituted with 150 l of 50 mM DTT in 0.5 M Tris-HCl (pH 8.5) containing 6 M Gdn-HCl (23). The mixture was incubated at 37°C for 4 h. An identical sample lacking DTT was treated in the same way. The reaction was stopped by adding 15 l of glacial acetic acid and 40 l of 0.1% aqueous trifluoroacetic acid.
HPLC of Peptides-A Varian model 5020 HPLC system equipped with a reverse-phase C4 column (4.6 mm (inner diameter) ϫ 25 cm, 10 m, Vydac) was employed for peptide separations. Solvent A was 0.1% trifluoroacetic acid in water, and solvent B contained 0.1% trifluoroacetic acid, 95% acetonitrile, and 5% water. The sample was injected, and solvent A was passed through the column for 5 min. The peptides were eluted with a linear gradient of 0 -60% B over 60 min at a flow rate of 1 ml/min.
Microbore HPLC Electrospray Mass Spectrometry-5 g of trypsindigested ␣1,2-mannosidase (reduced or non-reduced) were fractionated on a microbore C18 column (1 mm (inner diameter) ϫ 25 cm, 5 m, Vydac), which was on-line with a Finnigan SSQ 7000 mass spectrometer equipped with an electrospray ionization source. A Hewlett Packard model 1090 HPLC system was used, and a discontinuous gradient was employed for elution: 5% B at 0 min, 33% B at 63 min, 60% B at 95 min, and 80% B at 105 min, using 0.05% trifluoroacetic acid/water as buffer A and 0.05% trifluoroacetic acid/acetonitrile as buffer B. The mass analysis was set at 5 s/scan with a range of 300-2500 m/z. The flow rate was 40 l/min for the non-reduced sample and 30 l/min for the reduced sample. For the non-reduced sample, mass data were collected 5 min after the gradient started. For the reduced sample, data collection started 10 min later.
Protein Analysis-The concentration of purified recombinant ␣1,2mannosidase was quantitated using the absorbance at 280 nm as described (24) or the Micro BCA reagent kit from Pierce. Peptide concentration was determined using a modified Lowry method (25). SDS-PAGE under reducing conditions was carried out according to Laemmli (26) using the Bio-Rad Mini-PROTEAN II apparatus. For Western blotting, proteins were transferred onto nitrocellulose membrane (Schleicher & Schuell) and visualized by the ECL Western blotting detection system (Amersham) using rabbit polyclonal antiserum raised against the purified soluble yeast ␣1,2-mannosidase (14) or against recombinant yeast ␣1,2-mannosidase (4).
Peptide Analysis-N-terminal sequencing was performed at Queen's University (Kingston, Canada) on an Applied Biosystems model 473A sequenator equipped with an on-line microgradient phenylthiodantoin analysis system or at the Sheldon Biotechnology Centre (McGill University, Montréal, Canada) on a Beckman integrated microsequencing system, model PI2090E, equipped with an on-line HP1090 HPLC.
For amino acid composition analysis, peptides were subjected to constant boiling hydrochloric acid (about 6 N) vapor hydrolysis for 16 -24 h at 110°C using a Pico-Tag workstation (Waters), and amino acid analysis was performed on a Beckman 6300 series autoanalyzer. The amino acids were separated by ion-exchange chromatography, and ninhydrin post-column detection/quantitation was used.

RESULTS
Expression of S. cerevisiae ␣1,2-Mannosidase in P. pastoris-In order to obtain sufficient protein for structural analysis, the catalytic domain of the yeast ␣1,2-mannosidase was cloned downstream of the PHO1 signal peptide and strong alcohol oxidase (AOX1) promoter in the pHIL-S1 vector. The construct was introduced into the P. pastoris genome by homologous recombination. Expression from P. pastoris was induced with methanol, and the recombinant ␣1,2-mannosidase was secreted into the medium. Different clones expressed different amounts of ␣1,2-mannosidase, but the highest yield obtained using this system was 30 mg/liter of purified recombinant enzyme, which is 50 times more than was produced in S. cerevisiae as described previously (0.6 mg/liter) (14). Recombinant ␣1,2-mannosidase produced from S. cerevisiae had been shown to have similar properties as the native ␣1,2-mannosidase (14), and the recombinant ␣1,2-mannosidase from P. pastoris has the same specific activity and K m as the enzyme produced in S. cerevisiae (data not shown).
Quantitation of Disulfides and Free Sulfhydryl Groups-The yeast ␣1,2-mannosidase contains five cysteine residues in its catalytic domain. NTSB in the presence of sodium sulfite was used to quantitate the number of disulfides plus free sulfhydryl groups. The protein was reduced with sodium sulfite releasing one thiol per disulfide bond, and NTSB reacts with free thiols to produce one NTB per free thiol. Reaction with ␣1,2-mannosidase yielded a value of 2.7 Ϯ 0.4 NTB molecules/protein molecule. DTNB was used to quantitate free sulfhydryl groups. A value of 1.1 Ϯ 0.1 sulfhydryl group/␣1,2-mannosidase molecule was obtained. In addition, carboxymethylation with iodoacetate under denaturing conditions resulted in 0.85 CM-Cys/ protein molecule as determined by amino acid composition. These results demonstrate that there is one free sulfhydryl group and two disulfide bonds in the yeast ␣1,2-mannosidase.
Effect of DTT on ␣1,2-Mannosidase Activity-The importance of the disulfide bonds and the free sulfhydryl group in the yeast ␣1,2-mannosidase was investigated. It was found that purified recombinant ␣1,2-mannosidase loses about 90% activity when incubated for 9 h with 10 mM DTT (Fig. 1). The rate of inactivation clearly shows first order kinetics, suggesting that reduction of one of the two disulfide bonds was likely to cause the loss of enzyme activity. No change in the migration of the protein on SDS-PAGE was observed during this treatment with DTT. These results demonstrate that at least one disulfide bond is essential to maintain ␣1,2-mannosidase activity. Treatment of the ␣1,2-mannosidase with the sulfhydryl specific reagents N-ethylmaleimide, -chloromercuribenzoate, iodoacetate, and iodoacetamide under native conditions did not affect enzyme activity (data not shown).
Labeling the Sulfhydryl Group with 2-[ 3 H]ICH 2 COOH-The protein was labeled with radioactive iodoacetate to identify the cysteine containing the free sulfhydryl group. Labeling was carried out by adding Gdn-HCl and iodoacetate simultaneously to the dried protein in order to minimize disulfide exchange. The CNBr-digested protein was fractionated by gel filtration (Fig. 2). There was one major radioactive peptide fraction (C 1 ) that contained about 43% of the radioactivity expected to be incorporated as CM-Cys. The radioactive peptides that were eluted between fractions 32 and 45 were larger than 10 kDa due to incomplete digestion with CNBr. The peptides in fractions 75 and 85 contained only about 10% of the recovered radioactivity.
Fraction C 1 was treated with endoprotease Glu-C, under conditions in which the enzyme would cleave after both aspartic and glutamic acid residues (27), in order to obtain peptides containing only one cysteine residue according to the amino acid sequence. HPLC of the undigested peptides showed that all of the radioactivity was eluted with a peak at about 48 min (Fig. 3). HPLC of an aliquot of the endoprotease Glu-C-digested peptides showed radioactive fractions eluting with peaks at about 31 and 40 min. Preparative HPLC was carried out on the remaining digest, and the fractions expected to contain radioactivity were collected manually. About 64% of the radioactivity was eluted as a doublet at about 31 min, about 30% was eluted at about 40 min, and 5% did not bind to the column. Fractions g 1 and g 2 were analyzed by N-terminal sequencing, collecting fractions for radioactivity measurement. The two fractions were found to contain the same radiolabeled peptide. The N-terminal sequence for the peptide in fraction g 1 is shown in Table I. The radioactivity elutes with the CM-Cys in the first cycle and the peptide was identified as: 485 CM-Cys-ITLPT-KKSNN-Hse 496 . The peptide contains homoserine at its C terminus and was eluted from the column as a doublet due to the equilibrium between homoserine and homoserine lactone (28).
Fraction g 3 was also analyzed by N-terminal sequencing (Table I) Fig. 2 (about 75 g of peptides) was treated with endoprotease Glu-C and fractionated by HPLC as indicated under "Experimental Procedures." The elution was monitored at 206 nm (A), and an aliquot from each fraction was used to quantitate total radioactivity per fraction (hatched bars in B). An aliquot of undigested C 1 was also applied to HPLC under the same conditions, and 0.5-ml fractions were collected. The total radioactivity in each fraction is plotted in B as white bars. Fractions subsequently analyzed by N-terminal sequencing are named g 1 , g 2 , and g 3 .
to the CM-Cys residue. These data show that fraction g 3 is a partially digested peptide also containing carboxymethylated Cys 485 .
Isolation and Characterization of Disulfide-bonded Peptides-In order to isolate peptides containing a disulfide bond, the ␣1,2-mannosidase was first treated with CNBr and the products were fractionated by Sephadex G-50 gel filtration chromatography as described under "Experimental Procedures." Disulfide exchange was prevented by CNBr treatment under acidic conditions. Five peptide fractions (P 1 -P 5 ) were pooled according to the pattern of absorbance at 230 nm. Aliquots from each peptide fraction were treated with or without DTT and analyzed by HPLC. Disulfide-containing peptides were identified by a change in their elution position upon reduction (data not shown). A set of peptides eluting between 44 and 48 min of HPLC were observed to shift upon reduction. These peptides were then digested with endoprotease Asp-N and were analyzed by HPLC before and after reduction with DTT (Fig. 4). Endoprotease Asp-N digestion was carried out at pH 6.0 in order to prevent disulfide rearrangement. Fraction a 1 disappears upon reduction and fraction a 2 is decreased by half. This residual absorbing material is a contaminant that is pres-ent in the buffer alone. Fractions a 1 and a 2 were collected manually and subjected to amino acid analysis. Fraction a 1 did not contain any cysteine and did not correspond to any possible disulfide-containing peptide in the yeast ␣1,2-mannosidase. From amino acid composition and N-terminal sequencing results (Table II), it was determined that fraction a 2 contains the following peptides, which are linked by a disulfide bond: 336 DHLVCF-Hse 342 and 383 DTCYQ-Hse 388 . These results demonstrate that there is a disulfide bond between Cys 340 and Cys 385 .
b The dashes represent the absence of phenylthiohydantoin-amino acids in the indicated cycle. In this case cysteine residues were degraded and arginine residues were poorly recovered.

TABLE II
Characterization of peptide a 2 Fraction a 2 from HPLC of endoprotease Asp-N-digested peptides (see Fig. 4) was analyzed for amino acid composition and N-terminal sequence. a Methionine is converted to homoserine by CNBr cleavage. Homoserine was present, but was not quantitated.
b Glutamic acid is obtained due to the deamidation of glutamine. c There was no standard for homoserine, but there was a peak where the homoserine is expected to elute. could not be identified as tryptic peptides. Additionally, peaks after scan 850 contained peptides above 4000 Da and contained mixtures of peptides that could not be resolved. However, 77% of the amino acid sequence was confirmed. Longer incubation with trypsin was not useful, since there were too many peptides that could not be assigned to expected tryptic peptides.
A peptide with mass 4053.3, corresponding to the peptide in which Cys 340 is bound to Cys 385 (t 16 ϩ t 19 ), was observed in the non-reduced sample (Fig. 5A and Table III). This peptide was not observed in the reduced sample, although the two peptides resulting from its reduction (2731.8, t 16 ; 1322.9, t 19 ) were only observed in the reduced sample ( Fig. 5B and Table III).
The tryptic peptide containing Cys 485 (t 23 ) was observed in both the non-reduced and reduced digests (Fig. 5 and Table III).
A peptide with a mass (5625.1) corresponding to residues 429 -475 (t 22 ) is observed in the non-reduced sample and in the reduced sample (5626.5) (Table III). This peptide elutes at scan number 1180 in both the reduced and non-reduced tryptic digests (data not shown). Trypsin does not cleave between Cys 468 and Cys 471 ; therefore, peptide t 22 contains both cysteine residues. There is an increase in mass (1.4 mass units) upon reduction, which is within experimental error of the expected value of 2 mass units due to the loss of two protons.
Mutagenesis of Cysteine Residues-In order to determine the role of each disulfide bond and of the free sulfhydryl residue in the yeast ␣1,2-mannosidase, each cysteine was individually mutated to serine. The plasmids were transformed into P. pastoris, and clones were screened for ␣1,2-mannosidase expression using Western blotting and ␣1,2-mannosidase assay. Although a significant number, 25, of transformants were screened for each mutant, recombinant ␣1,2-mannosidases containing C340S or C385S mutations were never detected in the medium. Western blotting of cellular extracts showed minimal or no ␣1,2-mannosidase in cells transformed with C340S or C385S mutant plasmids in contrast to cells transformed with the wild type plasmid (data not shown). The C468S, C471S, and C485S mutants were all expressed in the medium and were enzymatically active. However, the C468S and C471S mutations resulted in a reduction in the amount of ␣1,2-mannosidase found in the medium, whereas the C485S mutant was expressed at a similar level as the wild type protein. In order to compare the relative specific activity of the mutant ␣1,2-mannosidases to the wild type enzyme, aliquots of medium containing equal ␣1,2-mannosidase activity were subjected to Western blotting (Fig. 6). According to densitometric scanning, it was determined that the C485S mutation did not affect ␣1,2-mannosidase activity, whereas the C468S and C471S mutation reduced the specific activity of the enzyme. Several clones expressing the C468S and C471S mutant enzymes had about 50% of the specific activity present in the wild type enzyme. The K m values were 0.7, 0.7, and 0.4 mM for the C468S, C471S, and C485S mutants, respectively, compared to 0.3 mM for the wild type enzyme. The C471S and C485S mutants were incubated with labeled Man 9 GlcNAc, and the products were analyzed by HPLC as described previously (14). The results indicate that the mutant ␣1,2-mannosidases retained the same specificity as the wild type ␣1,2-mannosidase (data not shown).

DISCUSSION
In the present work, we have shown using several methods that the yeast ␣1,2-mannosidase contains two disulfide bonds and one free thiol group in its catalytic domain, and we have identified their position in the primary sequence of the enzyme. Cys 485 was found to contain the free thiol group by sequencing of labeled peptides following modification with radioactive iodoacetate and by HPLC/MS tryptic peptide mapping. The presence of a disulfide bond between Cys 340 and Cys 385 was demonstrated directly by sequencing a purified peptide containing this disulfide and by HPLC/MS mapping of tryptic peptides. The existence of the other disulfide bond between Cys 468 and Cys 471 was deduced from several observations. First, quantitation with DTNB and NTSB clearly demonstrated the presence of two disulfide bonds and one free thiol. Second, Cys 485 was the only residue labeled with iodoacetate and no CM-Cys was formed from Cys 468 and Cys 471 following carboxymethylation. Third, a tryptic peptide (peptide t 22 ) containing only Cys 468 and Cys 471 was identified by HPLC/MS peptide mapping. Because there is no cleavable tryptic site between these two residues, no large effect is observed upon reduction of this peptide. However, the fact that this peptide containing both Cys 468 and Cys 471 was identified by HPLC/MS, in conjunction with the assignment of the other cysteine residues, supports the conclusion that the second disulfide bond is present between Cys 468 and Cys 471 . Finally, a peptide containing Cys 468 or Cys 471 disulfide bonded to any of the other cysteine residues was never observed by isolating disulfide-containing peptides or by HPLC/MS peptide mapping.
Treatment of the yeast ␣1,2-mannosidase with DTT clearly shows that at least one of the two disulfide bonds is essential to maintain its activity. This conclusion is supported by the mutagenesis studies demonstrating that neither the disulfide bond between Cys 468 and Cys 471 nor the free thiol group on Cys 485 are essential for enzyme activity and that removal of the Cys 340 -Cys 385 disulfide bond by mutagenesis of Cys 340 or Cys 385 results in no secreted ␣1,2-mannosidase and little or no intracellular recombinant protein. The Cys 340 -Cys 385 disulfide bond is therefore essential for the protein to acquire its proper conformation. Although mutagenesis of Cys 468 or Cys 471 caused some decrease in specific activity and a small change in K m , these mutant enzymes were still catalytically active, consistent with the conclusion that the second disulfide bond is not essential for enzyme activity.
From alignment of the amino acid sequences of the known members of the ␣1,2-mannosidase family (Fig. 7), it is observed that Cys 340 and Cys 385 have been conserved through evolution. This observation indicates that this disulfide bond may also play an essential role for these enzymes, and it supports the conclusion that the yeast enzyme requires formation of the Cys 340 -Cys 385 disulfide bond to fold properly.
Although the Cys 468 -Cys 471 disulfide bond is not essential for enzyme activity, the binding affinity for the substrate and the specific activity of the C468S and C471S mutants decreased 2-fold compared to the wild type enzyme. It seems, therefore, that this disulfide bond stabilizes the enzyme. This idea is supported by difficulties encountered in purifying the C471S mutant, most likely due to traces of protease activity. The above observations, coupled with the fact that Cys 468 is conserved in all the members of the family (Fig. 7), suggest that the region corresponding to Cys 468 and Cys 471 in the yeast enzyme is likely to be important for proper structure and catalytic activity in other members of the family.
The present results demonstrate that mutation of Cys 485 does not affect activity, showing that the free sulfhydryl group is not required for activity of the yeast ␣1,2-mannosidase. This conclusion is only in apparent disagreement with a previous study showing inactivation of the rabbit ␣1,2-mannosidase FIG. 6. Expression of wild type and mutant ␣1,2-mannosidases in P. pastoris. The KM71 P. pastoris strain was transformed with either pHIL-S1, YpHMNS1, YpHC468S, YpHC471S, or YpHC485S. Transformed clones were grown in BMGY medium for 2 days, then induced in BMMY medium for 2 days. The medium was diluted 1/8 to 1/150 and assayed for ␣1,2-mannosidase activity. Different aliquots (0.   (7), Drosophila (11), A. saitoi (13), P. citrinum (12), and S. cerevisiae (5) ␣1,2-mannosidases. The amino acid numbers are given in the right margin, and the conserved amino acids are indicated with a black dot. The three conserved cysteine residues are highlighted with a box, and the two additional cysteine residues in the yeast ␣1,2-mannosidase are marked with an arrow.
with -chloromercuribenzoate (30). This inactivation was only observed in the presence of very low Ca 2ϩ concentration (0.01 mM), and normal enzyme activity was observed following -chloromercuribenzoate treatment in the presence of 2 mM Ca 2ϩ . Furthermore, the location of the free sulfhydryl group is not conserved between the yeast enzyme and other members of the family and cysteine residues have not been implicated in the catalytic mechanism of any glycosidases to date. Two acidic residues are usually directly involved in catalysis for both inverting and retaining enzymes (31).
This study combines results from protein chemistry and mutagenesis in order to elucidate the role of the cysteine residues in the yeast ␣1,2-mannosidase. It is the only member of the family for which this type of study is possible, since it is the only Family 47 ␣1,2-mannosidase that has been produced in sufficient quantity as a recombinant enzyme.