Characterization of a cDNA Encoding a Novel Human Golgi α1,2-Mannosidase (IC) Involved in N-Glycan Biosynthesis*

A human cDNA encoding a 70.9-kDa type II membrane protein with sequence similarity to class I α1,2-mannosidases was isolated. The enzymatic properties of the novel α1,2-mannosidase IC were studied by expressing its catalytic domain in Pichia pastoris as a secreted glycoprotein. α1,2-Mannosidase IC sequentially hydrolyzes the α1,2-linked mannose residues of [3H]mannose-labeled Man9GlcNAc to form [3H]Man6GlcNAc and a small amount of [3H]Man5GlcNAc. The enzyme requires calcium for activity and is inhibited by both 1-deoxymannojirimycin and kifunensine. The order of mannose removal was determined by separating oligosaccharide isomers formed from pyridylaminated Man9GlcNAc2 by high performance liquid chromatography. The terminal α1,2-linked mannose residue from the middle branch is the last mannose removed by the enzyme. This residue is the mannose cleaved from Man9GlcNAc2 by the endoplasmic reticulum α1,2-mannosidase I to form Man8GlcNAc2 isomer B. The order of mannose hydrolysis from either pyridylaminated Man9GlcNAc2 or Man8GlcNAc2 isomer B differs from that previously reported for mammalian Golgi α1,2-mannosidases IA and IB. The full-length α1,2-mannosidase IC was localized to the Golgi of MDBK and MDCK cells by indirect immunofluorescence. Northern blot analysis showed tissue-specific expression of a major transcript of 3.8 kilobase pairs. The expression pattern is different from that of human Golgi α1,2-mannosidases IA and IB. Therefore, the human genome contains at least three differentially regulated Golgi α1,2-mannosidase genes encoding enzymes with similar, but not identical specificities.

␣1,2-Mannosidases play an essential role in the maturation of N-glycans to hybrid and complex structures in mammalian cells (for reviews, see Refs. [1][2][3]. They remove the four ␣1,2linked mannose residues from Man 9 GlcNAc 2 , following cleavage of glucose from Glc 3 Man 9 GlcNAc 2 . Thus, ␣1,2-mannosidases provide the Man 5 GlcNAc 2 substrate required for GlcNAc transferase I that initiates formation of complex and hybrid N-glycans. They belong to class I ␣-mannosidases (family 47 of the glycosyl hydrolase classification (Ref. 4)) that have been conserved through eukaryotic evolution. The ␣1,2-mannosidases are type II transmembrane proteins with amino acid similarity throughout their large C-terminal catalytic domains. They are inverting calcium-dependent glycosyl hydrolases that are inhibited by 1-deoxymannojirimycin and kifunensine. However, they have different N-terminal regions and intracellular localizations. A class I ␣1,2-mannosidase localized to the ER 1 of mammalian cells has been cloned (5,6). It has the same properties as the yeast ER ␣1,2-mannosidase, the structure of which has recently been determined by x-ray crystallography (7). The ER ␣1,2-mannosidase removes a single specific mannose residue from Man 9 GlcNAc 2 to form Man 8 GlcNAc 2 isomer B that lacks the terminal ␣1,2-mannose from the middle branch of the oligosaccharide. Two class I Golgi ␣1,2-mannosidases, IA and IB, that are about 65% identical in amino acid sequence have also been cloned from mammalian cells (8 -11). These Golgi enzymes remove the four ␣1,2-linked mannose residues from Man 9 GlcNAc 2 to yield Man 5 GlcNAc 2 . Their specificity is complementary to that of the ER ␣1,2-mannosidase since the mannose residue cleaved by the ER enzyme is the last residue removed by the two Golgi ␣1,2-mannosidases (12). The major difference between Golgi ␣1,2-mannosidase IA and IB is their tissue-and cell-specific expression as shown by Northern blot analysis of human and murine tissues (9 -11), and by immunolocalization in cells of the rat testis (13). In addition, there is some difference in their specificity with Man 9 GlcNAc as substrate (12).
In the present work, the characterization of Golgi human ␣1,2-mannosidase IC, a novel member of the mammalian class I ␣1,2-mannosidases is reported. This enzyme displays a distinct pattern of tissue-specific expression and trims Man 9 GlcNAc 2 to Man 5 GlcNAc 2 , forming different high mannose oligosaccharide intermediates from those previously observed for mammalian Golgi ␣1,2-mannosidases IA and IB.  3Ј end of the transcript were aligned, and clones T54452, AA437353,  N30588, H67812, H68084, and W19722, spanning the length of the  consensus sequence (1.3 kb), were obtained from Genome Systems Inc. and sequenced. Primers within the 5Ј region of the consensus sequence (5Ј-ACCTGAACGTGAGCGGAGAAG-3Ј; 5 Ј-CCTGGTTGCCAGA GAGTTCAG-3Ј) were used to screen a fetal brain cDNA library by PCR and hybridization (Genome Systems). The isolated clone (2.2 kb) contained 0.9 kb of additional 5Ј sequence. In addition, a partially sequenced 2.9-kb EST clone (AA017666) was identified when the NCBI Human UniGene file was reorganized and listed clone sizes. The clone was obtained from Research Genetics and entirely sequenced.
␣1,2-Mannosidase IC Antibodies-Rabbits were immunized with 0.5 mg of the keyhole limpet hemocyanin-conjugated synthetic C-terminal peptide NHSDSSGRAWGRH emulsified in Freund's complete adjuvant and boosted 5 weeks later with the same peptide in Freund's incomplete adjuvant. Serum was collected 12 days later. Antipeptide antibodies were affinity-purified (17) using a column prepared by coupling the peptide to cyanogen bromide-activated Sepharose 4B according to the manufacturer's instructions (Amersham Pharmacia Biotech).
Expression of the Catalytic Domain in Pichia pastoris-The DNA sequence encoding the catalytic domain (amino acids 165-630) was amplified by PCR using a sense primer containing a KpnI site (5Ј-AAAGGTACCCAGGAGCCCCAGAGCCAAGTG-3Ј) and an antisense primer with a XbaI site following the stop codon (5Ј-AAATCTAGAT-CAGTGTCTGCCCCAGGCTCTG-3Ј). The amplicon was inserted into the KpnI/XbaI sites of pPICZ␣ A (Invitrogen) in frame with the ␣-factor signal sequence yielding the expression construct pZ␣AHMIC493. The expression construct (10 g) was linearized with PmeI and electroporated into P. pastoris strain GS115 (his4) (Invitrogen), and transformants were grown as described previously (11). Clones expressing recombinant ␣1,2-mannosidase were identified by assays with [ 3 H]Man 9 GlcNAc (18).

SDS-PAGE and Western
Blotting-Medium containing recombinant ␣1,2-mannosidase, with and without Endo H (New England Biolabs) treatment, was subjected to SDS-PAGE (19) using the Bio-Rad Mini-Protean II apparatus. The proteins were then transferred onto a nitrocellulose membrane (Schleicher & Schuell). Recombinant ␣1,2-mannosidase was detected using affinity-purified peptide antibodies visualized by the ECL detection system (Amersham Pharmacia Biotech).
To characterize the oligosaccharide isomers, 10 l of medium was incubated with either 50 pmol of either Man 9 GlcNAc 2 -PA or Man 8 GlcNAc 2 -PA in 50 mM MES, pH 5.9, containing 1 mg/ml BSA, 10 mM CaCl 2 , and 1 mM NaN 3. The assay mixtures were supplemented with fresh enzyme at 24 h. Samples (1/6) were collected at 0, 2, 4, 8, 24, and 48 h. The products were first fractionated according to size by HPLC on a TSK-Gel Amide 80 column (4.6 ϫ 250 mm, TosoHaas), eluted isocratically at 1 ml/min with a 1:1 v/v mix of acetonitrile, water, 500 mM acetic acid, pH 7.3 (75:15:10, v/v/v) and acetonitrile, water, 500 mM acetic acid, pH 7.3 (50:40:10, v/v/v). The pH of the 500 mM acetic acid was adjusted to 7.3 with triethylamine. Oligosaccharides were monitored with a Varian model 360 fluorescent detector at an excitation of 310 nm and an emission of 380 nm. The oligosaccharide products were collected manually and lyophilized. The isomers present in each fraction were then resolved by HPLC on a MicroPak-SP C18 column (4.6 ϫ 150 mm, Varian), eluted isocratically at 1 ml/min with 100 mM acetic acid containing 0.025% n-butyl alcohol adjusted to pH 4 with triethylamine. The isomers were monitored at an excitation of 320 nm and an emission of 400 nm. The identity of the products was determined by comparing their elution to that of standard Man 9 -5 GlcNAc 2 -PA.
Localization of ␣1,2-Mannosidase IC in MDBK and MDCK Cells by Immunofluorescence-The ORF sequence was amplified by PCR using a sense primer containing a HindIII site and Kozak sequence (5Ј-AAAAAAGCTTCCACCATGCTCATGAGGAAAGTG-3Ј). The antisense primer containing a NotI site was either 5Ј-AAAAAAAAGCGGCCGC-GAGTGTCTGCCCCAGGCTCTG-3Ј or 5Ј-AAAAAAAAGCGGCCGCT-CAGTGTCTGCCCCAGGCTCTG-3Ј (including a stop codon). The ORF amplicons were cloned into the HindIII/NotI sites of pMH (Roche Molecular Biochemicals) in frame with the C-terminal hemagglutinin tag, yielding the tagged (pMHHMICT) and untagged (pMHHMICS) constructs.
MDBK and MDCK cells were grown to 70% confluence in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM glutamine, and 5 g/ml gentamicin. Following trypsinization, 1.4 ϫ 10 6 cells in PBS containing 20 mM HEPES were transiently transfected by electroporation with 10 g of either pMHHMICT or pMHHMICS. The cells were grown overnight on coverslips in 35-mm dishes, and the following morning the medium was changed. At 24 and 48 h after electroporation, the cells were washed twice with PBS, fixed with 3% paraformaldehyde (prewarmed to 37°C) for 10 min, washed twice with PBS, permeabilized for 2 min with 0.2% Triton X-100 in PBS, and blocked with fetal calf serum for 1 h at room temperature.  21)). Following five washes with PBST, the primary antibodies were detected by incubation for 1 h with affinitypurified CY2 conjugated anti-mouse IgG (1:400) or goat anti-rabbit IgG conjugated with rhodamine (tetramethylrhodamine B isothiocyanate) (1:800) (Jackson ImmunoResearch). The cells were then washed five times with PBST and mounted onto slides in Immuno-Fluore mounting medium (ICN). They were viewed with a Nikon Eclipse 800 epifluorescence microscope and photographed with TMax P3200 film (Kodak).
DNA Sequencing and Alignments-DNA sequencing was done by the Sheldon Biotechnology Centre (McGill University, Montréal, Canada) using the ABI prism dye terminator sequencing kit and ABI 373A sequencer, or with the Thermo Sequenase fluorescent labeled primer cycle sequencing kit (Perkin Elmer) and ALFexpress sequencer. Sequencing was also done by Bio S&T Inc. (Montréal, Canada) using the SequiTherm EXCEL II kit (Epicenter Technologies) and a Long Readir 4200 sequencer. Sequences were assembled into contigs with the DNASTAR SeqMan program (Madison, WI), and deduced amino acid sequences were aligned using the PileUp and Gap programs (version 10.0) from the University of Wisconsin Genetics Computer Group (Madison, WI).  1 kb) as well as the 3Ј-untranslated region (1 kb) was identified by a PCR screen of a fetal brain cDNA library using primers within the 5Ј region of the consensus sequence. In addition, a partially sequenced EST clone identified in the UniGene data base was completely sequenced (2.9 kb) and shown to encode the entire ORF.

Isolation and Characterization of Human
The ␣1,2-mannosidase IC cDNA (2.9 kb) is predicted to encode a 70.9-kDa type II membrane protein with a short cytoplasmic tail of about 22 amino acid residues, a transmembrane domain of 22 residues and a large C-terminal domain (Fig. 1). The C-terminal domain contains a proline-rich "stem" region (amino acids 45-164) not required for enzyme activity, followed by the catalytic domain (amino acids 165-630). The latter encodes class I ␣1,2-mannosidase signature motifs (see the Carbohydrate-Active Enzymes server, available via the world wide web) and the nine invariant acidic amino acids and cysteine residues shown to be essential for the activity of the yeast class I ␣1,2-mannosidase (23,24). Three potential N-glycosylation sites are located within the catalytic domain. ␣1,2-Mannosidase IC is about 54% identical to the human (X74837, AF027156), murine (U04299, U03458), and porcine (Y12503) ␣1,2-mannosidases IA and IB, and 38% identical to the human ER ␣1,2-mannosidase (AF145732, AF148509) amino acid sequences (5, 6, 8 -11, 25).
␣1,2-Mannosidase Expression in Human Tissues-Northern blot analysis revealed variable expression of a major 3.8-kb transcript in most tissues with the exception of lung, muscle, and pancreas (Fig. 2). Remarkably high levels of the major transcript are expressed in the placenta. The ovary, liver, and placenta also expressed minor transcripts of about 2.4 kb, and several tissues expressed low levels of a 5.7-kb transcript. Both the transcript sizes and expression differ from those reported for human ␣1,2-mannosidase IA (3.8, 4.3 kb) and IB (7.5, 9.5 kb) (11) with the exception of the 5.7-kb transcript, but this same size transcript is different since it is expressed differentially in the tissues analyzed.
Expression of Recombinant ␣1,2-Mannosidase IC in P. pastoris-The catalytic domain starting at amino acid 165 was cloned in the P. pastoris expression vector pPICZ␣ A in frame with the ␣-factor signal sequence. ␣1,2-Mannosidase activity was detected in the medium 2 days following induction with methanol of yeast cells transformed with the resulting construct pZ␣AHMIC493. No activity was found in the medium of cells transformed with the empty vector pPICZ␣ A. The secreted recombinant ␣1,2-mannosidase consists of a 55-kDa and a heterogeneous 67-kDa form. Treatment with Endo H gives rise to a single band of the expected size of 52 kDa (Fig. 3).
These results indicate that one glycoform only acquires core N-glycans, whereas the other contains outer chains with an average of about 16 residues per core structure, assuming all three sites are equally glycosylated.
Properties of Recombinant ␣1,2-Mannosidase IC-The enzymatic properties of recombinant ␣1,2-mannosidase IC were analyzed using [ 3 H]mannose-labeled Man 9 GlcNAc as substrate. The enzyme has a pH optimum of about 5.9 and requires the addition of calcium for maximum activity. Inhibition of the enzyme by preincubating with 50 M EDTA is reversed by the addition of 10 mM Ca 2ϩ , but not by 10 mM Mg 2ϩ , Mn 2ϩ , Co 2ϩ , Zn 2ϩ , or Fe 2ϩ . The ␣1,2-mannosidase IC activity is inhibited by the class I ␣-mannosidase inhibitors 1-deoxymannojirimycin  (Table I). Therefore, ␣1,2-mannosidase IC has all the properties ascribed to class I ␣1,2-mannosidases.
Specificity of Human ␣1,2-Mannosidase IC-The enzyme was incubated with [ 3 H]mannose-labeled Man 9 GlcNAc, and the products obtained at different times were resolved by HPLC. The recombinant enzyme catalyzed the stepwise removal of mannose from Man 9 GlcNAc to form Man 6 GlcNAc and a small amount of Man 5 GlcNAc (Fig. 4). Man 9 GlcNAc incubated for 48 h with medium of yeast transformed with the vector alone was not hydrolyzed.
To determine the order of mannose removal, the enzyme was incubated with Man 9 GlcNAc 2 -PA. The Man 8 -6 GlcNAc 2 intermediates were first fractionated according to size by HPLC (data not shown). Each oligosaccharide fraction was then further fractionated into its component isomers by HPLC on a reverse phase column. Hydrolysis of Man 9 GlcNAc 2 yields primarily a single Man 8 GlcNAc 2 isomer (about 90%), equivalent amounts of two Man 7 GlcNAc 2 isomers, and a single Man 6 GlcNAc 2 isomer that were identified by comparison with elution of standard oligosaccharides-PA (Fig. 5, A-C). These results indicate that the terminal ␣1,2-linked mannose residue on the middle arm is the last to be removed. Since Man 8 GlcNAc 2 isomer B is formed by human ER ␣1,2-mannosidase I, Man 8 GlcNAc 2 -PA isomer B was also incubated with ␣1,2-mannosidase IC. In this case the enzyme first cleaves the terminal mannose on the ␣1,3-branch of the substrate, yielding essentially a single Man 7 GlcNAc 2 isomer (about 85%). Thereafter, equivalent amounts of two Man 6 GlcNAc 2 isomers (Fig. 5,  D and E) were formed by the hydrolysis of either of the two remaining ␣1,2-linked mannose residues. Thus, the order of mannose removal from Man 8 GlcNAc 2 -PA was identical to the order observed for the Man 9 GlcNAc 2 -PA.
Immunolocalization of ␣1,2-Mannosidase IC in Transfected MDBK and MDCK Cells-The full-length ␣1,2-mannosidase IC [␣-32 P]dATP-labeled ␣1,2mannosidase EST clone W19722 was hybridized to Northern blots containing 2 g of poly(A ϩ ) RNA isolated from human tissues. The blots were exposed to x-ray film for 5 days. Molecular size markers are indicated beside each blot.   was expressed in MDBK and MDCK cells to determine its subcellular localization by indirect immunofluorescence. Punctate perinuclear Golgi staining was detected in cells 24 -48 h after transfection with both the hemagglutinin tagged (pMH-HMICT) and native (pMHHMICS) ␣1,2-mannosidase IC (Fig.  6). The staining pattern shows that ␣1,2-mannosidase IC is in the Golgi since it co-localizes with endogenous Golgi ␤1,4galactosyltransferase. No immunofluorescence was observed with pre-immune serum, secondary antibodies alone, or cells transfected with the pMH vector.
Genomic Organization and Chromosomal Localization-The ␣1,2-mannosidase IC gene contains 12 exons encoded by Gen-Bank clones AL031280 and AL020996. These clones overlap by 2.3 kb within the intronic region between exons 2 and 3. The gene is localized on chromosome 1p35.1-36.13 and spans 167 kb of genomic sequence between the markers D1S2843 and D1S417 on Gene Map 98 (26). The intron and exon boundaries of the coding region are identical to those found in the human ␣1,2-mannosidase IA gene, which spans 188 kb on chromosome 6q22 (UniGene Hs.2750). The reported genomic organization of the ␣1,2-mannosidase IB (11) localized on human chromosome 1p13 differs from ␣1,2-mannosidase IA and IC at a few positions within the ORF (Fig. 7), particularly at the N terminus, which is encoded by two exons. DISCUSSION The present results demonstrate the existence of a previously unsuspected third mammalian Golgi ␣1,2-mannosidase derived from a distinct gene. This enzyme is capable of trimming high mannose oligosaccharides to Man 5 GlcNAc 2 during N-glycan biosynthesis. Human ␣1,2-mannosidase IC requires calcium for activity and is inhibited by 1-deoxymannojirimycin and kifunensine; thus, it possesses the characteristic properties of class I ␣1,2-mannosidases. The amino acid sequence of the catalytic domain is similar to previously described mammalian Golgi ␣1,2-mannosidases IA and IB, but the cytoplasmic tail and stem region sequence differ. Furthermore, ␣1,2mannosidase IC displays a distinct tissue-specific expression pattern and order of ␣1,2-linked mannose removal (Fig. 8).
Human Golgi ␣1,2-mannosidases IA, IB, and IC are encoded by independent genes on chromosomes 6q22, 1p13, and 1p35-36, respectively. Gene duplication occurring late in evolution probably gave rise to the mammalian Golgi ␣1,2-mannosidase gene family (27) since the positions of the intron and exon boundaries within the gene are very similar. However, in both humans and mice these genes are independently regulated, thus giving rise to distinct patterns of expression (9 -11, 13).
␣1,2-Mannosidase IC readily hydrolyzes three of the four ␣1,2-linked mannose residues of Man 9 GlcNAc 2 and slowly cleaves the remaining terminal ␣1,2-linked mannose residue on the middle branch (Fig. 8, upper section). The enzyme produces the same Man 8 GlcNAc 2 isomer as ␣1,2-mannosidase IA  4 -12). The intron and exon boundaries within the coding region are identical to those of the ␣1,2-mannosidase IC gene (AL031280 (exons 1 and 2) and AL020996 (exons 3-12)). The ␣1,2-mannosidase IB gene organization (11) differs at the indicated positions. Coding region exons are indicated by numbered boxes, introns are denoted by dotted lines, and solid lines represent the 5Ј-and 3Ј-untranslated regions. Numbers above the boxes correspond to the position of the 3Ј nucleotide in the exons relative to the first nucleotide of the ORF. (12) and then forms equivalent amounts of two Man 7 GlcNAc 2 isomers. One of the Man 7 GlcNAc 2 isomers is also formed by recombinant murine ␣1,2-mannosidases IA and IB (12), and purified rat Golgi ␣1,2-mannosidase (28), whereas the other isomer (not produced by IA or IB) is formed by recombinant insect (29) and fungal (30) ␣1,2-mannosidases, and is an inferred intermediate of purified porcine ␣1,2-mannosidase (31). The human Golgi ␣1,2-mannosidase activities are complementary to the human ER ␣1,2-mannosidase (5, 6) since they hydrolyze the terminal ␣1,2-linked mannose of the middle arm of Man 9 GlcNAc 2 last. Hydrolysis of Man 8 GlcNAc 2 isomer B by ␣1,2-mannosidase IC proceeds readily to Man 5 GlcNAc 2 (Fig. 8,  lower section). The mannose residues are removed in the same order observed for Man 9 GlcNAc 2 . However, this order differs from that observed with murine ␣1,2-mannosidases IA and IB (12). These results demonstrate that ␣1,2-mannosidase IC has a unique specificity that differs from that of mammalian Golgi ␣1,2-mannosidases IA and IB. Recent x-ray crystallographic studies of the yeast ER ␣1,2mannosidase indicate the active site of class I ␣1,2-mannosidases is located within an (␣␣) 7 barrel with many non-conserved amino acids interacting with different parts of the oligosaccharide substrate (7). Furthermore, mutation of one of these amino acids was demonstrated to change the specificity of the yeast ER ␣1,2-mannosidase (22). Therefore, it is likely that the variations in the order of mannose removal by the various class I ␣1,2-mannosidases is largely determined by the differences in non-conserved amino acids interacting with the oligosaccharide substrate within the barrel. The relative expression of mammalian Golgi ␣1,2-mannosidases with slightly different specificities can thus provide different high mannose oligosaccharide isomers with possible variation in recognition functions.