Demonstration that Golgi endo-alpha-D-mannosidase provides a glucosidase-independent pathway for the formation of complex N-linked oligosaccharides of glycoproteins.

Studies on N-linked oligosaccharide processing were undertaken in HepG2 cells and calf thyroid slices to explore the possibility that the recently described Golgi endo-alpha-D-mannosidase (Lubas, W.A., and Spiro, R.G. (1987) J. Biol. Chem. 262, 3775-3781) is responsible for the frequently noted failure of glucosidase inhibitors to achieve complete cessation of complex carbohydrate unit synthesis. We have found that in the presence of the glucosidase inhibitors, castanospermine (CST) or 1-deoxynojirimycin, there is a substantial production of the glucosylated mannose saccharides (Glc3Man, Glc2Man, and Glc1Man) which are the characteristic products of endomannosidase action. Furthermore, in HepG2 cells, a secretion of these components into the medium could be demonstrated. Characterization of the N-linked polymannose oligosaccharides produced by HepG2 cells in the presence of CST (as well as 1-deoxymannojirimycin to prevent processing by alpha-mannosidase I) indicated the occurrence, in addition to the expected glucosylated species, of substantial amounts of Man8GlcNAc and Man7GlcNAc. Since Man9GlcNAc was almost completely absent and the Man8GlcNAc isomer was shown to be identical with that formed by the in vitro action of endomannosidase on glucosylated polymannose oligosaccharides, we concluded that this enzyme was actively functioning in the intact cells and could provide a pathway for circumventing the glucosidase blockade. Indeed, quantitative studies in HepG2 cells supported this contention as the continued formation of complex carbohydrate units (50% of control) during CST inhibition could be accounted for by the deglucosylation effected by endomannosidase.

Studies on N-linked oligosaccharide processing were undertaken in HepG2 cells and calf thyroid slices to explore the possibility that the recently described Golgi endo-cY-D-mannosidase (Lubas, W. A., and Spiro, R. G. (1987) J. Biol. Chem. 262, 3775-3781) is responsible for the frequently noted failure of glucosidase inhibitors to achieve complete cessation of complex carbohydrate unit synthesis.
We have found that in the presence of the glucosidase inhibitors, castanospermine (CST) or 1-deoxynojirimycin, there is a substantial production of the glucosylated mannose saccharides (GlcsMan, GlczMan, and GlclMan) which are the characteristic products of endomannosidase action. Furthermore, in HepG2 cells, a secretion of these components into the medium could be demonstrated. Characterization of the N-linked polymannose oligosaccharides produced by HepG2 cells in the presence of CST (as well as l-deoxymannojirimycin to prevent processing by cr-mannosidase I) indicated the occurrence, in addition to the expected glucosylated species, of substantial amounts of MansGlcNAc and Man,GlcNAc.
Since MansGlcNAc was almost completely absent and the MansGlcNAc isomer was shown to be identical with that formed by the in vitro action of endomannosidase on glucosylated polymannose oligosaccharides, we concluded that this enzyme was actively functioning in the intact cells and could provide a pathway for circumventing the glucosidase blockade. Indeed, quantitative studies in HepG2 cells supported this contention as the continued formation of complex carbohydrate units (50% of control) during CST inhibition could be accounted for by the deglucosylation effected by endomannosidase.
It is now recognized that in most eukaryotic cells N-glycosylation of proteins is accomplished by a co-translational transfer of a triglucosylated polymannose oligosaccharide (GicsMansGlcNAcz) from a lipid carrier followed by a complex series of processing reactions to yield mature asparaginelinked carbohydrate units (l-3). A prerequisite to the formation of the complex-type oligosaccharides is the removal of glucose residues, and this modification is generally believed to be accomplished early in the processing scheme through * This work was supported by Grant DK 17477 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore he hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom correspondence and reprint requests should be addressed: Elliott P. Joslin Research Laboratories, One Joslin Place, Boston, MA 02215. the sequential action of Lu-glucosidases I and II which are situated in the rough endoplasmic reticulum (3). However, despite this requirement for glucose excision, it has become apparent from studies in a variety of cells that glucosidase inhibitors fail to achieve complete cessation of complex carbohydrate unit synthesis (4-16), and, that furthermore, a glucosidase II-deficient cell line still has the capacity to form these oligosaccharides (17). These unexpected findings suggested to us that an alternate processing route which can circumvent a glucosidase blockade may exist and that this pathway may be provided by the Golgi endo-cu-D-mannosidase recently discovered in this laboratory (18,19). This enzyme which can achieve deglucosylation by cleaving the bond between the glucose-substituted mannose residue and the remainder of the polymannose unit is not inhibited by the commonly used agents which interfere with glucosidase action.
In the present investigation, we have explored this hypothesis by studying N-linked oligosaccharide processing in HepG2 cells and thyroid slices in the presence of glucosidase inhibitors.
In both cell types, an active endo-cY-D-mannosidase could be demonstrated by identification of the characteristic products of this enzyme. Furthermore, detailed quantitative studies of the HepG2 cells indicated that this alternate route can account for the continued formation of complex carbohydrate units during glucosidase blockade. 20-min labeling period with [2-3H]mannose (200 &i) was terminated by a wash with medium containing 2 mM mannose and 4 mM pyruvate, and this was followed by incubations in this unlabeled medium for various periods of time.
Incubation of Thyroid Slices--Slices (0.5 g) from fresh calf thyroids were incubated at 37 "C in I.7 ml of a previously defined medium (21) containing 5 mM pyruvate and 120 aCi of D-[U-"Clglucose in an atmosphere of 95% 02, 5% CO*. As with the cells, a 45-min incubation with inhibitors preceded the addition of the radioisotope.
Extractions of Cells and Slices-At the end of the He&2 cell incubations, the medium was carefully removed, and the plates were washed twice with 2 ml of phosphate-buffered saline; these fractions were pooled and stored at -20 "C for further analyses. Subsequently, 4 ml of an ice-cold mixture of methanol, 0.15 M Tris/HCl, pH 7.4, containing 4 mM MgCl, (2:l) was added to the rinsed cells which were then scraped from the plates and combined with a further 2-ml wash with this-reagent. Chloroform (6 ml) was then added to achieve a final mixture of chloroform/methanol/buffer (3:2:1) from which the upper phase containing the soluble oligosaccharides was removed for further study. The interphase material was, after a water wash, further extracted with chloroform/methanol/water (10:10:3) as previously described (21) to yield a delipidated protein pellet.
At the termination of the thyroid slice incubations, comparable upper phase oligosaccharide and interphase delipidated protein fractions were prepared after separating the tissue from the medium and disrupting the former with a Polytron homogenizer (Brinkmann Instruments) as previously reported (21).

Preparation of Free Oligosaccharides from Medium
and Cells-For the characterization of free oligosaccharides, organic solvents were evaporated in uucuo at 40 "C from the upper phase of the lipid extracts while the media, after addition of bovine serum albumin (1.25 mg), were deproteinized with a final concentration of 10% ice-cold trichloroacetic acid subsequent to which the latter was removed by repeated extractions with peroxide-free ether. After such treatment, the upper phase and medium samples, in aqueous solution, were passed through coupled columns of Dowex 50-X2, 200-400 mesh (H' form) and Dowex l-X2, 200-400 mesh (acetate form) to yield neutral salt-free fractions which in turn were applied to columns (3 X 10 mm) of charcoal-Celite (Darco G-BO-Celite 535, 1:l by weight) (22). After extensive washing with water to remove monosaccharides, these columns were eluted with 30% ethanol (9 ml) to yield an oligosaccharide fraction which was resolved by thin layer chromatography.

Preparation of Glvcopeptides
and Endo H Digestion-The delinidated protein pellets from the HepG2 cells or thyroid slices, as well as the ether-extracted trichloroacetic acid precipitates from the media of the cells, were digested at 37 "C with Pronase (Calbiochem) under the conditions previously specified (23). Upon termination of the digestion, the glycopeptides were absorbed on columns of Dowex l-X2 (acetate form) overlaid with Dowex 50-X2 (H+ form), eluted with 2 M pyridine acetate, pH 5.0, and lyophilized. Polymannose units were released by digestion of the glycopeptides with 4 milliunits of endo H (ICN Biochemicals Inc. or Genzyme) as previously described (24); after passage of the digests through columns of Dowex 50 over Dowex 1, the neutral oligosaccharides were collected in the effluent and water wash, whereas the endo H-resistant glycopeptides were eluted from the resins with 2 M pyridine acetate, pH 5.0.

Pectin Affinity
Chromatography-Glycopeptide samples  x lo5 dpm) were chromatographed at room temperature on columns (0.7 x 4 cm) of concanavalin A-Sepharose (Pharmacia LKB Biotechnology Inc.) equilibrated with 50 mM Tris/HCl buffer, pH 7.8, containing 100 mM NaCl, 1 mM CaC12, 1 mM MgC12, and 0.02% sodium azide. After allowing the sample to interact with the immobilized lectin for 30 min, elution was started with 10 ml of equilibration buffer (fraction 1) and continued with 10 ml of 10 mM methyl-or-Dglucoside (fraction 2) and 14 ml of 500 mM methyl-cY-D-mannoside (fraction 3) in the same buffer. Each fraction was concentrated by lyophilization, and the glycopeptides therein were separated from salt and glycoside by filtration on columns of Bio-Gel P-2 equilibrated with 0.1 M pyridine acetate, pH 5.0. Glycosidase Digestions-Jack bean cr-mannosidase (Sigma, Type III) digestions of oligosaccharides were carried out in 200 ~1 of 0.15 M sodium citrate buffer, pH 5.2, at 37 "C; a total of 7.5 units of enzyme were added in three equal portions over a period of 96 h. After terminating the reaction by heating at 100 'C for 3 min, the samples were desalted by passage through coupled columns of Dowex 50 (H' form) and Dowex 1 (acetate form) in preparation for examination by thin layer chromatography.
Treatment of oligosaccharides with amyloglucosidase (Sigma, from Rhizopus mold) was performed with 0.44 unit of enzyme in 200 ~1 of 0.1 M sodium acetate buffer, pH 4.8, at 27 "C for 24 h. At the end of the incubation, the samples were applied to a charcoal-Celite column, as already described, to separate released glucose (water wash) from the remaining oligosaccharides (30% ethanol eluate).
Digestions with rat liver endo-cu-D-mannosidase (100 /.rg of Golgi membrane protein) were carried out at 37 "C for 36 h in a loo-al volume under the conditions previously described (19) with the excention that 1 mM CST was substituted for the DNJ. The samples were deproteinized and desalted prior to thin layer chromatography by reported procedures (19).
All glycosidase incubations were performed in the presence of toluene.

Structural and Analytical
Procedures-Reduction of oligosaccharides was carried out with NaBHl as previously described (18). Acetolysis of reduced oligosaccharides was performed by the procedure of Varki and Kornfeld (25), and, after desalting (18), the products were identified by thin layer chromatography; radiolabeled standards for the characterization of acetolysis fragments were prepared as previously reported (18).
Quantitation of radiolabeled neutral sugars was carried out after acid hydrolysis (1 N HCl, 5 h at 100 "C under nitrogen) followed by passage of the samples through coupled columns of Dowex 50 (H' form) and Dowex 1 (acetate form) (22) and separation of the monosaccharides by thin layer chromatography in Solvent System C; radioactivity was determined by scintillation counting after elution of the components from the plate. The relative specific activities of glucose and mannose residues were determined from the ratio of the radioactivities observed in these two sugars after hvdrolvsis of purified GlcaMamGlcNAc.
Radiolabeled oligosaccharides released by endo H (0.5-3.0 X lo6 dpm) were coupled to 2-aminopyridine (Aldrich Chemical Co.) by reductive amination according to the procedure of Hase et al. (26). The reaction mixture was desalted on a column (1.5 x 34 cm) of Bio-Gel P-2 equilibrated with 0.1 N formic acid and further resolved by preparative thin layer chromatography on silica gel-coated plates in Solvent System B.

Chromatographic
Procedures-HPLC separation of pyridylaminoderivatized polymannose oligosaccharides was performed by a modification of the method of Hase et al. (27). Samples were applied to a Microsorb Cia column (4.6 x 250 mm, 5 pm, Rainin Instruments Co.) equilibrated with 20 mM ammonium acetate, pH 4.0, and eluted with the same solvent at a flow rate of 0.4 ml/min. The column was monitored with a Model 171 radioisotope detector (Beckman Instruments) employing a scintillation mixture in which Ready Flow III (Beckman Instruments) was mixed with the effluent in a 61 ratio. Data were recorded on a SP 4270 recording integrator interfaced with a Workstation integrator (Spectra-Physics Inc., San Jose, CA). Prior to HPLC analysis, each sample was mixed with a [2-3H]mannoselabeled GlciMam pyridylamine derivative to serve as an internal standard.
Thin layer chromatography of monosaccharides through pentasaccharides was carried out on plastic sheets precoated with cellulose (0.1 mm thickness, Merck) for 6 h in pyridine/ethyl acetate/water/ acetic acid, 5:5:3:1 (Solvent System A), while resolution of larger oligosaccharides (GlcaMangGlcNAcHz to Manr,GlcNAcHn) was carried out on plastic sheets precoated with Silica Gel 60 (0.2-mm thickness, Merck) for 26 h in 1-propanol/acetic acid/water 3:3:2 (Solvent System B); for the resolution of the pyridylamino derivatives of the oligosaccharides, chromatography was carried out for 36 h in this system. For the separation of hexoses from hexitols, chromatography was undertaken for 20 h on cellulose-coated plates in nitromethane/acetic acid/ethanol/water saturated with boric acid, 8:l:l:l (Solvent System C).. The chromatography in all systems was.carried out with a wick of Whatman No. 3MM paper clamued to the tou of the thin layer plates, and the components were detected by fluorography.
Quantitation was achieved by scraping the areas corresponding to the radioactive spots into vials to which 0.4 ml of water-was added followed by 4.6 ml of scintillation fluid (Ultrafluor, National Diaenostics, Inc.). For preparative purposes, resolved oligosaccharides were eluted from the plates with water, and the resultant eluates, after extraction with peroxide-free ether to remove scintillant, were passed through small coupled columns of Dowex 50 (H') and Dowex 1 (acetate).
For chromatographic standards, metabolically radiolabeled GIIz-~ MamGlcNAc and Mans.r,GlcNAc were prepared from thyroid oligosaccharide-lipids and glycoproteins, respectively, as previously de-scribed (19), and converted into their reduced forms by NaBHA treatment.
For the preparation of G&Man standards, "C-labeled Glc,.:iMangGlcNAcHz was treated with rat liver Golgi endo-cu-Dmannosidase as previously reported (19). Radioactiuity Measurements-Liquid scintillation counting was carried out in Ultrafluor with a Beckman LS 7500 instrument. Radioactive components on thin layer plates were detected using X-Omat AR film (Eastman) by fluorography at -70 "C after spraying with a scintillation mixture containing a-methylnaphthalene (28).

Identification of Endomannosidase-generated
Mono-, Di-, and Triglucosylated Mannose in HepG2 Cells and Thyroid Slices during Glucosidase Blockade-After incubation of HepG2 cells with [Ylglucose in the presence of glucosidase inhibitors, thin layer chromatographic examination of the free intracellular and medium oligosaccharides revealed the presence of radiolabeled components which co-migrated with the characteristic di-, tri-, and tetrasaccharide products (GlclMan,GlczMan, and GlcsMan) of in vitro endo-a-D-mannosidase action (Fig. 1). Although these glucosylated mannose oligosaccharides were not evident in control incubations, the presence of a glucosidase inhibitor, either DNJ or CST, resulted in their formation albeit in distinctively different ratios; the predominance of the tetrasaccharide in the presence of CST is consistent with its known effectiveness as a glucosidase I inhibitor (29,30).
While the components observed in the media were limited to the Glcl.aMan species, the cellular fraction contained a number of larger oligosaccharides (Fig. 1) which were identified as belonging to the previously reported (24) polymannose-G~cNAc~-~ series; in the presence of glucosidase inhibitors, they remained close to the origin due to their larger size. The components in the cellular fraction of control incubations (Fig. 1 GlcNAc-terminating oligosaccharides were not observed in the media (Fig. l), the presence of GlcleBMan components in this fraction cannot be attributed to a simple leakage from damaged cells. Indeed, pulse-chase studies suggested that the HepG2 cells actively secrete Glcl-sMan oligosaccharides into the medium. As illustrated for GlcsMan, this tetrasaccharide progressively accumulates in the medium after a period in which the concentration inside the cell reaches a plateau level (Fig. 2).
To confirm that the di-, tri-, and tetrasaccharide species which are produced by HepGP cells in the presence of the glucosidase inhibitors are in fact identical with the GlclMan, Glc*Man, and GlcaMan products of in vitro endomannosidase action (19), they were isolated by preparative thin layer chromatography. The purified components (Fig. 3) migrated to the same position on rechromatography as the in vitro products @Man = 0.76,0.61, and 0.28, respectively) and were found to be resistant to jack bean a-mannosidase treatment (data not shown). Furthermore, after acid hydrolysis of the native and NaBH4-reduced oligosaccharides, only glucose and mannose or glucose and mannitol, respectively, could be detected in ratios consistent with their proposed structures (Table I).
In order to determine if other cell types also display in vivo endomannosidase activity, the free oligosaccharides produced by calf thyroid slices during ['%]glucose labeling were examined by thin layer chromatography (Fig. 4). Although the chromatographic pattern from these cells was complicated by the presence of prominent radiolabeled components belonging to the maltooligosaccharide series (28), it was apparent that in the presence of DNJ or CST components migrating to the position of Glc,Man and GlcsMan, respectively, were produced (Fig. 4). Treatment of the thyroid oligosaccharide fractions with amyloglucosidase to degrade the maltose oligomers left the presumptive Glc,Man and GlcsMan components intact and presented a chromatographic picture comparable to that observed in the HepG2 cells (cf. Figs. 1 and 4); since the glucosylated mannose oligosaccharides were found to be resistant to amyloglucosidase action, the decrease of the spot at the GlcsMan position in the CST-inhibited sample after treatment with this enzyme (cf. lanes 6 and 7) is due to the removal of a maltooligosaccharide. In thyroid as in HepG2 cells, the glucosylated mannose oligosaccharides were not evident in The components were isolated by preparative thin layer chromatography from the medium after ["Clglucose labeling. The disaccharide and trisaccharide were obtained from incubations carried out in the presence of DNJ while the tetrasaccharide was isolated from CSTinhibited cells (see Fig. 1). Chromatography was performed for 6 h in Solvent System A on a cellulose-coated plate; 2000 dpm was applied to each lane, and the components were visualized by fluorography. The migration of standard mannose is indicated by the arrow. Following a 160-min incubation with 120 &i of radiolabeled glucose in the absence (CONTROL) or presence of glucosidase inhibitors (DNJ, 5 mM; CST, 2.2 mM), the oligosaccharides were isolated as described under "Experimental Procedures" and applied to a cellulose-coated plate with (+) or without (-) prior amyloglucosidase treatment. Chromatography was carried out in Solvent System A for 6 h, and the components were visualized by fluorography. The abbreviation-s for the glucosylated mannose standard are the same as in Fig. 1.
Mg W'g ' Isolated by preparative thin layer chromatography after labeling of HepG2 cells with ["C)glucose; the di-and trisaccharides were obtained from incubations performed in the presence of DNJ, while the tetrasaccharide was isolated from CST-inhibited incubations (see Fig. 1). * (+) indicates that the saccharide was reduced with NaBH4 prior to acid hydrolysis.
' Measured by scintillation counting after acid hydrolysis and thin layer chromatographic separation of the released hexoses and hexitols as described under "Experimental Procedures"; the values were corrected for the experimentally determined difference in specific activity of the glucose and mannose residues (Glc/Man = 1.2). No glucitol was detected in any of the reduced samples. After incubation of the cells with ['4C]glucose (100 PCi) in the presence of CST (2.2 mM) plus DMJ (2 mM) or absence of inhibitor (CONTROL), the reduced endo H-released oligosaccharides from cellular (C) and medium (M) glycoproteins were resolved by chromatography on a silica gel-coated plate in Solvent System B for 26 h. Aliquots representing 3% and 5%, respectively, of the cell and medium fractions were applied to the plate, and the components were detected by fluorography. The migration of radiolabeled standard oligosaccharides is indicated by the following abbre- Chromatographic examination of the radiolabeled oligosaccharides released from cell and medium glycoproteins by endo H revealed a very different pattern for control and glycosidase-inhibited incubations (Fig. 5). In the absence of the inhibitors, MansGlcNAc and MansGlcNAc were the predominant cellular species, while the saccharide units of the medium glycoproteins were largely endo H-resistant presumably because they had been processed to complex-type structures. In contrast, large amounts of endo H-susceptible oligosaccha-rides were present in cellular as well as medium glycoproteins formed in the presence of CST and DMJ and even a cursory examination indicated a substantial complement of slow moving glucosylated oligosaccharides as well as prominent ManaGlcNAc and ManGlcNAc components and yet very little of the MangGlcNAc (Fig. 5). After isolation by preparative thin layer chromatography, the oligosaccharides (a-f) from the inhibited cells were characterized by the products which they yielded after extensive jack bean a-mannosidase digestion.
The results of these studies indicated that components a,b,c, and d were predominantly GlcgManr,GlcNAc, GlcsMansGlcNAc, GlclMangGlcNAc, and GlclMan,GlcNAc, respectively, while components e and f consisted almost entirely of MansGlcNAc and ManGlcNAc, respectively. Treatment of components ae with rat liver Golgi endomannosidase confirmed these identifications indicating that indeed a and b were primarily triglucosylated, c and d monoglucosylated, and e unglucosylated (Fig. 6).
A quantitation of the various endo H-released oligosaccharides indicated that even in the glucosidase-inhibited cells a substantial portion was present in the nonglucosylated form as ManaGlcNAc and ManTGlcNAc (Table II). However, the almost complete absence of MangGlcNAc stood in marked contrast to the situation observed in control incubations and was consistent with an endomannosidase-mediated deglucosylation by the HepG2 cells during glycosidase inhibition (Table II).
The effectiveness of CST as an inhibitor of glucosidase II is evident from the virtual absence of MangGlcNAc; however, the presence of some monoglucosylated oligosaccharides suggest that the action of this agent on glucosidase I, although superior to DNJ (see Fig. l for the Glcl-sMan standards are the same as in Fig. 1. Occurring in HepG2 Cell Glycoproteins during Glycosidase Blockade-Analyses by reverse phase HPLC of the pyridylamino derivatives of ManRGlcNAc released by endo H from radiolabeled thyroid glycoproteins indicated that a clear separation of the three possible isomers of this oligosaccharide could be achieved by this procedure (Fig. 7), and these were observed to be present in ratios (21:70:9) similar to these reported by Byrd et al. (32) in their studies of bovine thyroglobulin.
Examination of the MansGlcNAc obtained from glycoproteins synthesized by HepG2 cells under control and exoglycosidase-inhibited (CST + DMJ) conditions indicated a distinct difference in the elution position of the predominant isomer (Fig. 7). In the uninhibited cells, MansGlcNAc coeluted with the major isomer from thyroid (component B) which is known to have the terminal mannose of the middle branch of the polymannose unit missing (18). The ManaGlcNAc from the cells incubated in the presence of CST and DMJ co-eluted almost exclusively with component A from thyroid (Fig. 7) and in a position identical to the product of in uitro action of rat liver Golgi endomannosidase on GlciMam,GlcNAc substrate (chromatograph not shown) in which the terminal mannose of the al,3-linked branch of the polymannose saccharide is absent (18). To further establish the identity of the ManeGlcNAc isomers, acetolysis studies were undertaken.
When this procedure, which selectively cleaves 1, 6- that from CST + DMJ-inhibited incubations yielded mannobiose, mannotriose, and MansGlcNAcHp as products (Fig.  8), thereby confirming the structures deduced from the HPLC separations. The acetolysis studies indicated, in agreement with the HPLC data, that the third possible Man,GlcNAc variant (isomer C), which would be expected to yield mannose, mannotriose, and ManlGlcNAcHz, was not present in significant amounts in either the control or inhibited HepGZ cells.
Since it has been suggested that the direct transfer of nonglucosylated Man,GlcNAc from lipid to protein might circumvent a glucosidase blockade (33), we undertook a characterization of the isomeric form of this oligosaccharide as present on the glycoproteins in control and exoglycosidaseinhibited HepG2 cells. Acetolysis of the reduced Man7GlcNAc produced fragments (mannobiose and Man,GlcNAcHz) consistent with the predominance of an isomer with structure II in the glycoproteins from both types of incubations (Fig. 9). Although the formation of small amounts of mannose and mannotriose suggested that a limited quantity of a Man7GlcNAc variant with structure I may also be present, the absence of Man,GlcNAcHz excluded the possibility that an isomer with structure III, which is characteristic of the Processing Pathway-Since analyses of the N-linked polymannose carbohydrate units produced in HepG2 cells indicated that the block in glucose removal could be circumvented by endomannosidase action, we undertook to determine if further processing into complex oligosaccharides could take place during glucosidase inhibition. When glycopeptides prepared after [2-3H]mannose labeling were fractionated by a concanavalin A chromatographic procedure (35), it became evident that indeed the synthesis of complex carbohydrate units continues in CST-inhibited cells, although in reduced amounts (Fig. 10). In the presence of the glucosidase inhibitor, an anticipated increase in the polymannose units of cellular glycoproteins was observed while there was about a 50% reduction in the complex multiantennary oligosaccharides; the concanavalin A fraction of the cellular glycoproteins in which the complex biantennary oligosaccharides are known to elute did not show a decrease presumably because of an enhanced formation of hybrid units (Fig. 10). While in the presence of CST, the radioactivity of glycopeptides from the medium was reduced to about one-half of the control levels, concanavalin A chromatography indicated that these secreted glycoproteins, unlike the cellular proteins, had the normal ratio of multiantennary, biantennary, and polymannose oligosaccharides (Fig. 10) In order to assess whether endomannosidase action could account for the persistent formation of complex carbohydrate units by HepG2 cells in the presence of CST, we determined the sum of the distinctive products released by this enzyme (GlcaMan, GlcpMan, and GlclMan) during an incubation with [2-3H]mannose and compared it, on a molar basis, to the endo H-resistant oligosaccharides formed during that time taking into account the presence of 3 mannose residues per unit and correcting for radiolabeling of fucose substituents. Such a calculation could be made with reasonable precision as in the presence of CST, without an additional mannosidase inhibitor, no N-linked unglucosylated polymannose oligosaccharides (Man5-pGlcNAc) could be detected (data not shown) presumably due to their rapid processing into complex carbohydrate units. The close correlation evident from these calculations between complex carbohydrate unit production and the appearance of mono-, di-, and triglucosylated mannose (Table III) indicated that endomannosidase does indeed provide the route by which the glucosidase blockage is circumvented.

DISCUSSION
It is apparent from the studies reported in this paper that the recently described 19) does function in uiuo and can initiate a processing route which permits the formation of complex N-linked carbohydrate units in the presence of a glucosidase blockade. Indeed, our findings indicate that the frequently noted failure of glucosidase inhibitors to stop complex oligosaccharide synthesis (4-16) in a variety of cells can be attributed to the alternate glucose-removal mechanism which the endomannosidase provides; furthermore, the continued formation of these oligosaccharides observed in glucosidase-deficient cells (17) could be explained on this basis.
The identification of the characteristic glucosylated man- nose products (GlcsMan, GlczMan, and GlciMan) of endomannosidase (19) in both HepG2 cells and thyroid slices in the presence of glucosidase inhibitors provided the major evidence for the in vivo action of this enzyme. These components, which were resistant to a-mannosidase and amyloglucosidase treatment, were found intracellularly as well as in the medium. Indeed, pulse-chase studies in HepG2 cells indicated that the GlclmsMan saccharides are actively secreted in contrast to the large assortment of polymannose-GlcNAcl_ 2 oligosaccharides (24) which remain associated with the cells. The mechanism by which the glucosylated mannose products of endomannosidase action reach the exterior of the cell is not known and is currently under investigation.
Characterization of the N-linked polymannose oligosaccharides formed in HepG2 cells during glucosidase blockade provided persuasive additional evidence that the endomannosidase is actively functioning. Incubation of the cells with CST, with the further addition of DMJ to prevent processing by mannosidase I of endomannosidase-generated saccharide units, resulted in the appearance of substantial amounts of N-linked ManBGlcNAcz which, in contrast to control incubations, occurred almost exclusively as the isomer in which the terminal mannose on the cul,3-linked branch is missing (isomer A). Also of importance was the observation that in the exoglycosidase-inhibited cells ManSGlcNAcz was almost completely absent although this component was present in substantial amounts in control cells and was, as anticipated (36), even more abundant when DMJ alone was used as a inhibitor (data not shown). The presence of substantial amounts of MansGlcNAcz in CST-inhibited cells despite the almost complete absence of MangGlcNAcz indicated that the endomannosidase-catalyzed conversion of Glcl-sMangGlc-NAcz to MansGlcNAcz must represent a major deglucosylation pathway which stands in contrast to the sequential removal of glucose residues by glucosidases I and II followed by mannose excision through the action of exo-a)-D-mannosidases (3).
The occurrence of N-linked ManTGlcNAcz and ManGGlc-NAcz oligosaccharides in the CST + DMJ-treated HepG2 cells is quite consistent with an endomannosidase-mediated deglucosylation route as in vitro studies have shown that this enzyme acts optimally on oligosaccharides with truncated mannose branches (19). Glucosylated MansGlcNAcn and Man?GlcNAcp, which have been noted as minor N-linked processing intermediates under physiological conditions (37)(38)(39), could, at the elevated levels occurring during glucosidase blockade, provide the substrate for the formation of substantial quantities of ManTGlcNAcp and MansGlcNAcZ, respectively, by endomannosidase action. Indeed, we have observed that Glcl-sMan is released in substantially greater amounts (1.6-fold) in CST-treated HepG2 cells than in those exposed to CST plus DMJ, and this is consistent with an enhanced in uivo action by endomannosidase on mannose-trimmed carbohydrate units.
Our studies with the HepGP cells indicated that, despite the demonstrably effective glucosidase inhibition brought about by CST, complex carbohydrate unit formation was inhibited by no more than 50%. This observation is in accord with the findings of numerous investigations (4-16) which indicated that only a partial blockade of complex oligosaccharide synthesis could be imposed by glucosidase inhibitors with an effectiveness ranging from 30% to 90%. Our measurements of radioactivity in the characteristic glucosylated mannose saccharides (Glci-BMan) produced by CST-inhibited HepG2 cells during [2-3H]mannose labeling clearly indicated that the endomannosidase activity was sufficient to account for all of the complex oligosaccharides formed during the glucosidase blockade. The possibility that the glucosidase inhibition could be circumvented by a direct transfer of ManTGlcNAcz from a dolichol intermediate to protein as has been suggested from studies with F9 teratocarcinoma cells (33) would not explain the large amount of MansGlcNAcp observed in the HepG2 cells and is furthermore made unlikely by the observation that endo H-released ManyGlcNAc in our system is distinct from the isomer which is associated with the lipid intermediate (34).
Although our studies with glucosidase-inhibited cells provide a clear indication of the potential capacity of the endomannosidase-initiated processing route, we are unable to determine at this time to what extent this alternate mechanism for deglucosylation operates under physiological conditions. Our inability to detect the characteristic Glcl_BMan saccharides in control HepG2 incubations does not necessarily indicate a lack of endomannosidase activity as these components might undergo degradation in the absence of a glucosidase inhibitor. While analyses of the N-linked polymannose oligosaccharides from uninhibited cells indicated that only relatively small amounts of the MansGlcNAcf isomer A was present in thyroid and essentially none in HepG2 cells, the interpretation of these findings in relation to endomannosidase activity has again to be qualified as this isomer once formed may undergo rapid further processing.
Judging from the substrate specificity of the endomannosidase (19), it is likely that, under normal circumstances, its primary in vivo action is on monoglucosylated N-linked polymannose oligosaccharides. Since it has been generally observed that excision of the outer 2 glucose residues occurs much more rapidly than that of the mannose-linked glucose (37,39,40), it may indeed be anticipated that oligosaccharides containing this latter substituent are the most likely to be carried into the Golgi complex in which the endomannosidase is known to be located (18). From our studies with glucosidaseinhibited cells, it is however apparent that even proteins with triglucosylated polymannose units can leave the rough endoplasmic reticulum and enter the Golgi cisternae to present themselves as substrate for the endomannosidase. The action of this enzyme apparently makes possible the secretion of proteins with a normal ratio of complex to polymannose Nlinked oligosaccharides as long as further processing by mannosidase I can proceed, however, if the latter enzyme is also inhibited, the HepG2 cells do export glycoproteins which contain almost exclusively the polymannose carbohydrate units.
It is apparent that the function of endomannosidase under physiological conditions requires further definition, and, as part of such a study, our laboratory is currently evaluating the selectivity of this enzyme for specific glycoproteins and N-glycosylation sites.