a-D-Mannosidases of Rat Liver Golgi Membranes MANNOSIDASE I1 IS THE GlcNAcMAN5-CLEAVING ENZYME IN GLYCOPROTEIN BIOSYNTHESIS AND MANNOSIDASES IA AND IB ARE THE ENZYMES CONVERTING MAN9 PRECURSORS TO MAN5 INTERMEDIATES*

Current evidence indicates that the trimming of man- nosy1 residues from intermediates in the biosynthesis of the N-linked oligosaccharides of glycoproteins oc- curs in the Golgi complex. We now present evidence that mannosidase 11 (Tulsiani, and (1977) J. Bwl Chem 252,3227-3233) is the Golgi enzyme that converts GlcNAcMans species to GlcNAcMans species in completing the mannosyl trimming process required in the biosynthesis of complex type glycoproteins. GlcNAc([3H]Man)6GlcNAc-mannos- idase andp-nitrophenyl a-D-mannosidase activities co-purify throughout the preparative procedure and show the same properties.

When it became evident that mannosyl trimming of biosynthetic intermediates occurred during glycoprotein biosynthesis, the rat liver Golgi a-D-mannosidase ("mannosidase 1I")l reported in 1973 by Dewald and Touster (7) became a candidate to be the cleavage enzyme. This enzyme, shown to be distinct from the lysosomal and cytosolic a-D-mannosidases (7), was purified (8) using p-nitrophenyl a-D-mannoside as substrate and was also shown to act on glycopeptides (9). Subsequently, Tabas and Kornfeld (10) prepared a second rat liver Golgi mannosidase ("mannosidase IA") active on a-1,2 linkages in high mannose oligosaccharides but possessing negligible activity toward PNP-Man.' Evidence that Golgi membranes also contain the mannosidase activity cleaving a-1,3and a-1,6-mannosyl residues from GlcNAcMan5-GlcNAcz to produce GlcNAcMansGlcNAcz has been presented (10)(11)(12).
The present paper reports the preparation from Golgi membrane of a second a-1,2-mannosidase ("mannosidase IB") and the demonstration that mannosidase I1 is the Golgi enzyme that converts GlcNAcMan5 species to GlcNAcMana species. A preliminary report of this work has been presented (5). EXPERIMENTAL PROCEDURES lan Industries, Inc. UDP, UDPGlcNAc, UDPGlc, and UDPGal were Materials-Male Wistar rats (125-150 g) were obtained from Harfrom Sigma, and Bio-Gel P-2 (200-400 mesh) and Bio-Gel P-4 (-400 mesh) were from Bio-Rad. Other chemicals were obtained from the sources described previously (8). The preparation of rat liver Golgi membranes was based on the method of Leelavathi et al. (13) as modified by Tulsiani et al. (8). These highly purified Golgi membranes were washed with 0.4 M NaCl to remove adsorbed proteins. Homogeneous preparations of Golgi mannosidase I1 were obtained from these membranes as described previously (a), except that the enzyme from the hydroxylapatite column was eluted with 0.115 M potassium phosphate buffer containing 0.1% Triton X-100. The purified, concentrated enzyme obtained in nearly 12% yield, has a specific activity of 7.0 units/mg of protein. The specific activity is 2-fold higher than previously reported (a), in part because twice as much substrate is used in the present assay procedure.
Uniformly The numbering of the Golgi a-D-mannosidases introduced in the present paper is in order of their action in the glycoprotein biosynthetic pathway and differs from the numbering in our previous reports (5,6).
The abbreviation used is PNP-Man, p-nitrophenyl a-D-mannoside.

GlcNAc
( D l (C) SCHEME 1 described elsewhere (2). This enzyme cleaves between the proximal GlcNAc residues of high mannose-type oligosaccharides (14,15). Jack bean a-mannosidase digestion of the [3H]mannose-labeled substrates yielded the expected ratios of radioactive peaks migrating as free mannose and Man-P-GlcNAc (2). Since Sindbis virus glycoproteins contain little if any MansGlcNAcz (2), ([3H]Man)sGlcNAc was prepared from glycopeptides isolated from cultured NIL-8 (hamster fibroblast) cells. In this case, 4.5 X IO5 cells were cultured in 60-mm plastic tissue culture dishes in 5 ml of Eagle's minimum essential medium (Gibco) containing 10% fetal calf serum and 300 pCi/ml D-[2-3H]mannose (New England Nuclear, 18.4 Ci/mmol). After 25-h incubation at 37 "C in a 5% COZ atmosphere, the cells from two dishes were harvested and the glycopeptides were prepared, treated with endo-P-N-acetylglucosaminidase H, and resolved on a column of Bio-Gel P-4 (1 X 215 cm, -400 mesh) as described previously (3), except that the elution buffer was 0.1 M pyridinium acetate, pH 5. The sugar composition of the ([3H]Man)~GlcNAc was confirmed as described for this oligosaccharide from chick embryo fibroblast cells (3). The ([3H]Man)9GlcNAc preparation contained a small amount (-10% of the radioactivity) of material behaving as GlclMansGlcNAc. It should be noted that although these substrates were not subjected to as extensive a structural determination as that applied to the substrates of Tabas and Kornfeld (lo), it is likely that the major route of processing high mannose derivatives is the same in the various cells used.
Uniformly labeled ([3H]Man)~GlcNAc was obtained in a similar manner from the cellular glycopeptides of clone 15B cells which had been incubated 24 h with [2-3H]mannose (300 pCi/ml). This mutant cell line, which was originally isolated by Gottlieb et al. (16), accumulates protein-linked Man5GlcNAc~ with the structure shown in Scheme 1, compound B (17) as a result of deficient UDP-G1cNAc:glycoprotein GlcNAc transferase activity (18). The clone 15B oligosaccharide and the Sindbis ([3H]Man)5GlcNAc behaved similarly in the enzymatic experiments reported in the present report.
Preparation of GlcNAc([3H/"n)~ClcNAc-([3H]Man)5GlcNAc from clone 15B cells (-250,000 cpm) was incubated at 37 "C in a total volume of 50 4 containing 100 mM sodium cacodylate buffer, pH 7.6, 10 mM MgC12, 10 mM UDPGlcNAc, and 5 p1 of rat liver Golgi membrane suspension (7.34 mg protein/ml of 0.3% Triton X-100). The reaction was terminated after 40 min by heating at 100 "C for 5-7 min. Since the N-acetylglucosaminylation was accompanied by some release of mannose by Golgi mannosidase, free [3H]mannose was removed by passing the reaction mixture through a column of Bio-Gel P-2 (0.6 X 90 cm) equilibrated with 0.1 M acetic acid. The flow rate was 4 ml/h, and 0.5-ml fractions were collected. Oligosaccharide appeared in fractions between 28 and 36, whereas free [3H] mannose appeared in fractions 46-50. The oligosaccharide pool was dried in a centrifugal Bio-Dryer (Virtis); the residue was suspended twice in 1 ml of glass distilled water and redried each time. Nearly 90-95% of the starting material was recovered in the oligosaccharide pool ("mixed oligosaccharide"). The remaining 5 to 10% of the radioactivity was found as free Man.
The mixed oligosaccharide produced from [3H]Man5GlcNAc was dissolved in 30 pl of Hz0 and fractionated on a Bio-Gel P-4 column (Fig. 1). The mixed oligosaccharide was thereby resolved into two major fractions (Peak Z and Peak ZI, Fig. 1B). Peak I, when reapplied to the same Bio-Gel P-4 column, exhibited a single peak with some trailing (Fig. IC). However, Peak I did not appear if UDPGlcNAc UDPGlcNAc and the 3H-labeled-oligosaccharide was separated on a column of Bio-Gel P-2 as described under "Experimental Procedures." Mixed oligosaccharide (20,000 cpm from mixture A and 210,000 cpm from mixture B ) was applied to a Bio-Gel P-4 column (0.6 X 122 cm) equilibrated with 0.1 M acetic acid. The column was washed with the same acetic acid at a flow rate of 1.25 ml/h. Fractions (0.5 ml) were collected and radioactivity was determined in aliquots as described was omitted from the original reaction mixture (Fig. lA), nor did any nucleotide other than UDPGlcNAc promote the appearance of Peak I. Consequently, it was concluded that Peak I in Fig. 1, B and C, largely consisted of N-acetylglucosaminylated material, namely G~cNAc([~H]M~~)sG~cNAc. The results of specific glycohydrolase digestions supported the tentative identification of Peak I as G~c N A c ( [~H ] M~~)~G~c N A c (Scheme 1, Structure C). Jack bean amannosidase digestion of Peak I released 46-53% of the radioactivity as free mannose (Fig. 2B). However, after treatment of Peak I with jack bean exo-P-N-acetylglucosaminidase (Sigma), which converted it to material migrating as MansGlcNAc (Fig. 2C), jack bean amannosidase released 80% of the radioactivity as free [3H]mannose (Fig. 20); the remaining 20% migrated as Man! GlcNAc (19). The same results were obtained with Peak I prepared from Man5[3H] GlcNAc, except that no radioactivity was released as free mannose Enzyme Assays-(i) PNP-a-D-mannosidase activity was assayed at pH 5.5 by measuring the hydrolysis of p-nitrophenyl-a-D-mannopyranoside as described previously (8), except that the substrate concentration was 4 mM instead of 2 mM. One unit of PNP-wDmannosidase activity is the amount of enzyme which catalyzes the release of 1 pmol ofp-nitrophenol/min. (ii) ['H]Man-a-~-rnannosidase activity was assayed by measuring the hydrolysis of [3H]Man-labeled oligosaccharide in a standard incubation mixture (0.05 ml) containing 2500-7500 cpm of radioactive substrate and 0.1 M sodium acetate buffer, pH 6.0. Incubation times (at 37 "C) were 10 min to 4 h as indicated in each experiment. The reaction was stopped by heating the samples at 100 "C for 5-7 min. Free ["HIMan was separated from the oligosaccharides by gel fdtration on a column of Bio-Gel P-2 (90 X 0.6 cm) equilibrated with 0.1 M acetic acid. The column was eluted at a flow rate of 4 ml/h. Fractions (0.5 ml) were collected and mixed with 10 ml of Aquasol and the radioactivity was measured by scintillation counting using ['HIMan as standard. ['HIMan-labeled oligosaccharides appeared in fractions 27-37, whereas free [3H]Man was present in fractions 46-50. One unit of [3H]Man-a-~-mannosidase is the amount of enzyme which catalyzes the release of lo00 cpm of ['HI Man/min. Under the standard assay conditions with either method, specific activity is the units of enzyme/mg of protein.
Protein was determined by the fluorometric assay of Anderson and Desnick (20). When samples contained Tris-HCl buffer, the appropriate volume of this buffer was included in the bovine serum albumin standard.
Enzymatic Digestion of Oligosaccharides-Jack bean a-mannosidase (Sigma) digestions were carried out as described (2). Jack bean P-N-acetylglucosaminidase (Sigma) digestions were performed as follows: to the radioactive oligosaccharide in 100 p1 of H 2 0 were added 15 pl of 0.3 M sodium citrate (pH 5.5) and 20 pl (2.5 units) 37 "C under toluene for 24 h.
of the enzyme suspension. The reaction mixture was incubated at Concentration of Enzyme Solutions-Enzyme solutions were concentrated in an Amicon concentrator equipped with a PM-30 membrane.
Unless otherwise indicated, Bio-Gel P-2 and Bio-Gel P-4 columns were run at room temperature, whereas enzyme purification steps were carried out at 0-4 "C.

Extraction of the a -D "annosiduses from Golgi Membranes
Sodium chloride-washed Golgi membranes (8) were extracted six times with Buffer A, and protein and a-D-mannosidase activity toward PNP-Man and ([3H]Man)9GlcNAc were assayed in the original Golgi suspension, the six extracts, and the final Golgi suspension. It is evident from Table I that the extraction patterns of a-D-mannosidase activity toward the two substrates differ, a result indicating the presence of more than one a-D-mannosidase activity in the Golgi membranes. Also shown in Table I (Experiment 2) are the similar patterns of extractability of activities toward PNP-Man and G~c N A c ( [~H ] M~~)~G~c N A c . T h e minor differences between our results and those of Tabas and Kornfeld (10) in the extractability of activities toward PNP-Man and GlcNAc(['H] Man)sGlcNAc may be explained by the fact that the 10 mM potassium phosphate buffer of pH 5.8 used in the extraction solution by these investigators is less effective than the 50 mM potassium phosphate buffer of pH 7.2 employed in the present study. Moreover, the activities toward PNP-Man and G~~N A C ( [~H ] M~~)~G~C N A C reported herein were assayed under identical conditions of buffer and pH. Tabas and Kornfeld used different buffers of different pH to assay the two activities (10).

Separation of Mannosidase IA from Mannosidase IB and II by Chromatography on Cellulose Phosphate
The six extracts ( Table 11, Experiment 1) were combined, the pH was adjusted to 5.8 with 1 M KH2P04, and the extracts were dialyzed against 40 volumes of Buffer B for 4-6 h with one change of buffer. The dialyzed enzyme was chromatographed on a cellulose phosphate column as shown in Fig. 3. Nearly 50% of the [3H]Man-ru-~-mannosidase activity applied

I Extraction of LY-D -mannosiduse activities and protein from Golgirich fraction
The Golgi membranes (8) were suspended in Buffer A (6.2 mg of protein/ml) by manual homogenization in a glass homogenizer equipped with a Teflon pestle. The suspension was kept on ice for 15 min and then centrifuged at 50,000 rpm for 30 min in the Spinco Ti-50 rotor. The supernatant solution was carefully removed by aspiration from the pellet and loosely packed membranes. The pellet and membranes were resuspended in the same volume of Buffer A, homogenized, and centrifuged as above. This process was repeated four more times. The pellet remaining after the sixth extraction was suspended in Buffer A. The two enzymatic assays and the determination of protein were performed as described under "Experimental Procedures."  " These values are approximate because of the small amounts of Extracts 4-6 were pooled and concentrated 10-fold before protein protein present in these fractions. and the two activities were assayed.
to the column was found in the Buffer C eluate, which contained negligible PNP-a-D-mannosidase activity. These results suggested that the [3H]Man-a-~-mannosidase activity eluted in Buffer C was mannosidase IA (10). Enzyme eluted from the cellulose phosphate column by NaCl gradient in Buffer C (Fig. 3) exhibited both [3H]Man-a-~-mannosidase and PNP-a-D-mannosidase activities. However, the ratio of the two activities decreased as the elution proceeded, and incubation at 50 "C for 3 h caused a greater loss of [3H]Mana-D-mannosidase activity than of PNP-a-D-mannosidase activity. Since these results suggested that the NaC1-eluted activity was due to more than one enzyme with a-D-mannosidase activity, fractions 55 to 80 were pooled, dialyzed, and further purified.

Separation of Mannosidase IB and II by Chromatography on a Second Cellulose Phosphate Column
The NaCl eluate was fractionated by a chromatographic

a-D "annosiduses
of Rat Liver Golgi Membranes step utilizing a second cellulose phosphate column in Buffer G (Fig. 4). More than 90% of the [3H]Man-a-~-mannosidase activity but less than 7% of the PNP-a-D-mannosidase activity applied to the column was present in the effluent and washes. The unadsorbed enzyme is mannosidase IB. Elution of the column with NaCl yielded fractions containing both mannosidase activities, the ratio of the activities in the fractions being essentially constant (0.32-0.43). The NaC1-eluted enzyme is mannosidase 11.
Purification of the Separated Mannosidases by Chromatography on Hydroxylapatite Columns Each of the three mannosidases was further purified on a column of hydroxylapatite (6 X 2 cm) equilibrated with Buffer D. Mannosidase IA was applied to the column and, after thorough washing with Buffer D, the column was eluted with Buffer E. Mannosidase I1 and mannosidase IB were similarly chromatographed on hydroxylapatite, except that the enzymes were eluted with Buffer F. The eluted enzymes were concentrated to small volumes (2.0-5.0 ml) as described under "Experimental Procedures." A summary of a typical experiment for the separation and purification of the three mannosidases is shown in Table 11. Mannosidases IA and IB were purified at least 25-fold over the Golgi-rich fraction. Mannosidase I1 was purified 20-fold (based on PNP-a-D-mannosidase activity) over the Golgi-rich fraction. Its extent of purification is not as high as previously reported (8) because the procedure used in the present report for the extraction of the Golgi membranes was designed to maximize the solubilization of all the mannosidases.
The separation and purification procedure described here resulted in recoveries of 22 and 17% for mannosidase IA and mannosidase IB, respectively. Much higher recovery of mannosidase IB was obtained by an alternate procedure in which the concentration of each fraction eluted from the first cellulose phosphate column was omitted. In brief, after the mannosidase IA was eluted from the first cellulose phosphate column with Buffer C, the column was eluted with 20 ml of Buffer G containing 0.5 M NaC1. The salt-eluted fraction containing mannosidases IB and I1 was dialyzed against Buffer G and applied to a second cellulose phosphate column equilibrated with Buffer G. The effluent and Buffer G column washes containing mannosidase IB in nearly 30% yield was further purified on a column of hydroxylapatite as described above.
Since, during the course of this investigation, the question of possible interconversions among the three mannosidases was addressed, a summary of certain additional experiments is presented at this point. Possible interrelationships between mannosidases IA and IB were examined as follows. (i) Equal amounts of purified mannosidase IA and mannosidase IB obtained from hydroxylapatite columns were mixed and applied to a cellulose phosphate column in Buffer B (pH 5.8). As expected from the results in Fig. 3, half of the recovered activity was eluted with Buffer C (pH 7.2). The remainder, mannosidase IB, was eluted by 0.15 M NaCl in Buffer C, a result suggesting that interconversion of the enzymes does not occur during cellulose phosphate chromatography. (ii) The ratio of the yields of the two enzymes was not affected by prior incubation of the Golgi membranes at 37 "C for 60 min or at 4 "C for 24 h before isolation of the enzymes. (iii) Prior storage of the cytoplasmic extract (supernatant solution) at 4 "C for 16 h gave a result indicating that mannosidase IB is not derived artifactually from mannosidase IA. Two distinct bands of crude membranes were observed above the 1.3 M sucrose instead of the one band obtained in the standard procedure without the prior storage of the extract (8). The Golgi membranes obtained from the upper band of pinkish membranes yielded mannosidases IA and I1 in amounts expected from Table 11. However, the recovery of mannosidase IB was lower than expected. These results cannot be explained by the assumption that mannosidase IB is a proteolytic product of mannosidase IA.

Kinetics
Under standard conditions, the rate of hydrolysis of PNP-Man by mannosidase I1 was directly proportional to enzyme concentration (0.6-2.5 pg of enzyme protein) and was linear for at least 30 min. Cleavage of ([3H]Man)sGlcNAc was also proportional to mannosidase I1 concentration. With 1 to 5 pg of enzyme, [3H]Man release was linear for 4 h. With mannosidase IA or IB, hydrolysis of ([3H]Man)9GlcNAc was proportional to enzyme concentration (70 to 300 ng of enzyme protein). With either enzyme, the reaction was linear for up to 2 h.

Effect ofpH on Enzyme Activity
The pH optimum of purified mannosidase I1 and PNP-Man as substrate was 5.5 in 0.1 M acetate buffer (or 0.1 M phosphatecitrate buffer), as reported previously @), but the pH optimum was 5.8 with ([3H]Man)8GlcNAc as substrate, as had been previously observed with a glycopeptide substrate (9). With ([3H]Man)8GlcNAc as substrate, both mannosidase IA and mannosidase IB showed maximum activity at pH 6.0.

a-~-Mannosidases of Rat Liver Golgi Membranes
Man)7GlcNAc. The reason for the difference in results is not known. With decreasing oligosaccharide size, there is a more rapid decrease of activity of mannosidase IB than of mannosidase IA. Moreover, unlike mannosidase IA, mannosidase IB shows a low level of activity toward PNP-Man which is not lost by rechromatography of the purifed enzyme. Mannosidase IA, on the other hand, is several times more active toward [3H]Man5GlcNAc than mannosidase IB. The detailed experiment reported in Table IV provides further evidence that mannosidase IA is considerably more effective in hydrolyzing ([3H]Man)5GlcNAc than are the other two Golgi mannosidases.

Stability
All three Golgi a-D-mannosidases were stable when stored at 2-4 "C in the presence of 10 m~ potassium phosphate buffer, pH 7.2, and 1% Triton X-100, the enzymes retaining nearly 80% of their activities after 4 weeks. They retained 50-

3665
70% of their activities when heated at 37 "C in 100 m M sodium acetate buffer, pH 5.8, for 24 h. However, they showed different stabilities at 50 "C ( Fig. 5), with mannosidase I1 being the most stable and mannosidase IB the least stable. (Table V) (also see below). Tris inhibits only mannosidases IA and IB, whereas p-chloromercuriphenyl sulfonic acid inhibits all three enzymes. The effects of Co2+, Tris-maleate, and EDTA on mannosidase IA are quite similar to those reported by Tabas and Kornfeld (10). Mannosidase I1 clearly behaves distinctively towards several reagents. Mannosidases IA and IB show small but reproducible differences towards some inhibitors, Le. Tris-C1, Tris-maleate, EDTA, andp-chloromercuriphenyl sulfonic acid.

TABLE IV Activities of the Golgi a-D-mannosidases toward ([3H]Man)sGlcNAc
Each incubation was performed as described for enzyme assays under "Experimental Procedures," using 0.01 unit of each mannosidase (based on MansGlcNAc substrate) except that the pH was 6.0 rather than 5.8. The reaction products for each mixture were characterized on a Bio-Gel P-4 column (115 X 1 cm).

Co-purification of PNP-mannosidase and G~cNAc([~H] Man)5GlcNAc-mannosidase-When PNP-a-D-mannosidase and GlcNAc([3H]
Man)5GlcNAc-mannosidase activities were assayed in the homogeneous preparations of mannosidase I1 (8), both these activities were present in the purified enzyme. Moreover, their patterns of purification from the crude membranes were essentially the same (Table VI). In addition, if the fiial purification on a hydroxylapatite column was effected by elution with a linear potassium phosphate gradient (0.04-0.5 M), the five fractions containing PNP-Man-cleaving activity (at approximately 0.09 M phosphate) also contained GlcNAc([3H]Man)5GlcNAc-cleaving activity. The two activities were present in the same ratio (+15%) in all five fractions.
Response to Cations-To determine whether the highly purified mannosidase 11 (8) still contained more than one mannosidase, the preparation was preincubated with cations which inhibit PNP-mannosidase activity. The hydrolyses of G~C N A C ( [~H ] M~~)~G~C N A C and ([3H]Man)8GlcNAc, as well as PNP-Man, were strongly inhibited by Cu2+ and Fe", with Fe3+ being a less potent inhibitor (Table VII). The magnitude TABLE V

Differential inhibition of rat liver Golgi a-D -mannosiduses
The reagent solution under study were added to the enzyme solution (50 p l ) in 0.1 M sodium acetate buffer, pH 5.8. After preincubation for 15 min at 0 "C, a-D-mannosidase was assayed by adding substrate (-5000 cpm of ([3H]Man)gGlcNAc) and incubating at 37 "C for 2 h. Free [3H]mannose was separated from the oligosaccharide on a column of Bio-Gel P-2 (0.6 X 90 cm). All reagent solutions were prepared and used within 1 h. The averages of duplicate or triplicate analyses are given, with the range of values obtained. Several values represent single determinations. It should be noted that these investigators assayed the two activities under different conditions. PNP-mannosidase activity was assayed in sodium acetate buffer, pH 5.5, whereas oligosaccharide hydrolysis was assayed in sodium phosphate buffer, pH 6.5 (10). We have examined the effect of Cu2+ in four different buffers (Table  VIII). The hydrolyses of PNP-Man, GlcNAc(r3H] Man)5GlcNAc, and ([3H]Man)sGlcNAc were inhibited similarly under all conditions employed (Table VIII). It is apparent that Cu2+ is a less effective inhibitor when tested in sodium phosphate buffer, pH 6.5, probably due to the precipitation of cupric phosphate.
Rates of Inactivation as a Function of pH-When highly purified mannosidase I1 was preincubated (37 "C) for 60 min at different pH values, and assayed for PNP-Man-,

G~C N A C ( [~H ] M~~)~G~C N A C -,
and ([3H]Man)8GlcNAc-cleaving activities, all three showed complete inactivation at pH 4.0, retention of about 70% of all activities at pH 6.0, and almost full retention at pH 8.0 (Table IX) Incubated for 15 min at 37 "C. Incubated for 2 h at 37 "C. Earlier experiments (7) had indicated that Fez+ does not inhibit PNP-Man activity of Golgi membranes. The fact that the mannosidase I1 in intact membranes or in partially purified preparations is not inhibited is apparently due to the oxidation of ferrous ion to ferric ions by these preparations.    terminal a-1,2-mannosyl residues (Fig. 6). It should be noted that nearly 20% of [3H]Man from G~C N A C ( [~H ] M~~)~G I C N A C (one mannosyl residue) was released within 7 min, with the release of the remaining [3H]Man requiring many hours. It is possible that the enzyme preferentially hydrolyzes either the 1,3-or 1,6-linked mannose. A tentative indication of the preferential cleavage of the 1,6 linkage was obtained by Harpaz and Schachter (12) in a study of the action of Golgi membranes on GlcNAcMan5GlcNAc.

DISCUSSION
The fact that six nonequivalent mannosyl residues in the Man9GlcNAcZ-containing precursors are cleaved during the biosynthesis of complex glycoproteins provides a basis for expecting that more than one a-mannosidase is involved in this biosynthetic process. Pulse-chase studies (3) suggest that these mannosidases might be located in the Golgi complex. The present report identifies three Golgi enzymes that can account for all of the required trimming of mannosyl residues from biosynthetic intermediates.
The cleavage of a-1,2-linked mannosyl residues by Golgi membranes is accounted for by two enzymes, namely, the a-1,2-mannosidase reported by Tabas and Kornfeld (10) (mannosidase IA) and by a second a-1,2-mannosidase (mannosidase IB) described in the present report. The existence of mannosidase IB was fiist indicated in chromatographic experiments with cellulose phosphate, when results were obtained indicating that a large part of the applied a-1,2-mannosidase activity remained adsorbed while the rest was eluted with Buffer C. Further purification steps led to the characterization of the adsorbed a-1,2-mannosidase and its designation as mannosidase IB.
Early purification studies not reported herein also disclosed that if the Golgi membrane extracts were subjected to DEAEcellulose (DE52) chromatography prior to cellulose phosphate chromatography, approximately half of PNP-Man-and (C3H] Man9GlcNAc-cleaving activities in the extracts passed directly through the column, whereas the adsorbed portions of each enzyme required elution with NaC1. Since both the unadsorbed and adsorbed fractions of the three enzymes gave very similar results in subsequent purification steps, the DE52 step eventually was omitted from the purification procedure. It had previously been observed that mannosidases I1 (8) and IA (10) exist in two forms differing in their adsorbability on DEAE-cellulose. It is evident that mannosidase IB does, too. This behavior could reflect differences between the two forms of each pair in acidic residues in the polypeptide chains or in the oligosaccharide moieties. Catalytic differences between the forms of a pair of mannosidases have not as yet been observed.
Although mannosidases IA and IB are quite similar in by guest on March 24, 2020 http://www.jbc.org/ Downloaded from substrate specificity, response to inhibitors, and certain other properties, they are clearly distinguishable on the basis of (i) their behavior on cellulose phosphate chromatography (Fig.  2), (ii) their thermolability (Fig. 4) and sensitivity to storage in liver cytoplasmic extract (see "Results"), and (iii) their activity toward ManSGlcNAc and PNP-Man. In repeated experiments with ManSGlcNAc, a relatively poor substrate mannosidase IA always showed about four times greater activity toward this substrate than did mannosidase IB (Table  111). Similarly, mannosidase IB which was free of contamination by mannosidase I1 always had a low level of activity toward PNP-Man, whereas mannosidase IA did not. The activity of mannosidase IB toward PNP-Man was not reduced by rechromatography of the enzyme on cellulose phosphate under conditions that readily separate mannosidase IB from I1 (i.e. Fig. 4). Attempts to obtain evidence that mannosidase IA and IB might be derived from each other gave negative results.
Further work will be required to shed light on possible biosynthetic or functional relationships between mannosidases IA and IB. Differentiation of their biological roles from each other may depend in part on studies of their action on glycopeptides or glycoproteins. It is possible that different processing routes occur within the Golgi apparatus, leading to different classes of glycoproteins and/or different subsequent routing. Interestingly, structural studies on human IgM suggest that the high mannose oligosaccharide attached to asparagine 563 is processed by a different sequence of mannose removal than occurs with the oligosaccharide at asparagine 402 (21, 22).
An a-l,2-mannosidase recently purified by Forsee and Schutzbach (23) from rabbit liver was stated to appear to differ from the a-1,2-mannosidase of Tabas and Kornfeld (mannosidase IA) because the former enzyme is strongly inhibited by copper ions whereas the latter is not. However, as the present paper indicates, mannosidase IA is in fact very sensitive to copper ions (Table VI) provided that the inhibition test is not carried out in phosphate buffer (Table IX). The rabbit liver enzyme shows a distinct difference from mannosidase IB in that the former enzyme possesses only a trace of activity toward PNP-Man (23).
The experiments reported herein establish that mannosidase I1 is the enzyme in the glycoprotein biosynthetic process that converts the pentamannosyl intermediate to the triman-nosy1 intermediate after the former has been N-acetylglucosaminylated (10)(11)(12). The activities of enzyme preparations toward GlcNAcMansGlcNAc and PNP-Man behave very similarly in all respects: (i) the two activities purify in a parallel manner, (ii) they respond very similarly toward effectors, and (iii) they are inactivated at the same rate.
The regulation of glycoprotein biosynthesis may in part involve the three Golgi mannosidases. As an aid to our understanding of these processes, it would be of great interest to establish the specific localization of these several enzymes within the Golgi complex.