Cell Surface Glycoproteins Undergo Postbiosynthetic Modification of Their N-Glycans by Stepwise Demannosylation*

Primary rat hepatocytes and two hepatoma cell lines have been used to study whether high mannose-typeN-glycans of plasma membrane glycoproteins may be modified by the removal of mannose residues even after transport to the cell surface. To examine glycan remodeling of cell surface glycoproteins, high mannose-type glycoforms were generated by adding the reversible mannosidase I inhibitor deoxymannojirimycin during metabolic labeling with [3H]mannose, thereby preventing further processing of high mannose-type N-glycans to complex structures. Upon transport to the cell surface, glycoproteins were additionally labeled with sulfosuccinimidyl-2-(biotinamido)ethyl-1,3-dithiopropionate. This strategy allowed us to follow selectively the fate of cell surface glycoproteins. Postbiosynthetic demannosylation was monitored by determining the conversion of Man8–9GlcNAc2 to smaller structures during reculture of cells in the absence of deoxymannojirimycin. The results show that high mannose-typeN-glycans of selected cell surface glycoproteins are trimmed from Man8–9GlcNAc2 to Man5GlcNAc2 with Man7GlcNAc2 and Man6GlcNAc2 formed as intermediates. It could be clearly shown in MH 7777 as well as in HepG2 cells that demannosylation affects plasma membrane glycoproteins after they are routed to the cell surface. As was determined for total cell surface glycoproteins in HepG2 cells, this process occurs with a half-time of 6.7 h. By analyzing the size of high mannose-type glycans of glycoproteins isolated from the cell surface at the end of the reculture period, i.e. after trimming had occurred, we were able to demonstrate that glycoproteins carrying trimmed high mannose glycans become exposed at the cell surface. From these data we conclude that cell surface glycoproteins can be trimmed by mannosidases at sites peripheral to N-acetylglucosaminyltransferase I without further processing of their glycans to the complex form. This glycan remodeling may occur at the cell surface or during endocytosis and recycling back to the cell surface.

During their maturation, N-linked glycans of secretory and membrane glycoproteins undergo extensive processing by spe-cific glycosidases and glycosyltransferases in the ER, 1 the Golgi complex, and the TGN (for reviews, see Refs. [1][2][3][4]. The sequence of processing events includes trimming of the precursor oligosaccharide, Glc 3 Man 9 GlcNAc 2 , by glucosidases I and II and by distinct ␣1,2-mannosidases to form Man 5 GlcNAc 2 . Several of the processing mannosidases have been described (for reviews, see Refs. 4 and 5) such as ER mannosidases and Golgi mannosidase IA/IB. Following the action of N-acetylglucosaminyltransferase I and Golgi mannosidase II, the transfer of N-acetyl-D-glucosamine, D-galactose, L-fucose, and sialic acids by an array of glycosyltransferases generates the wide variety of oligosaccharide structures found on mature glycoproteins.
Several lines of evidence suggest that oligosaccharide processing of cell surface glycoproteins is not restricted to biosynthesis but may also occur after the initial passage through the compartments of the secretory pathway to the cell surface (for a review, see Ref. 6). First, measurements of the turnover rates of the different sugar residues of glycoproteins isolated from rat liver plasma membranes have shown that these turnover kinetics are distinctly influenced by the position of each sugar within the N-linked oligosaccharides (7)(8)(9)(10). The half-lives of the terminal or penultimate sugars, L-fucose, sialic acid, and D-galactose, are only 1 ⁄6 to 1 ⁄3 as long as that of the protein backbone. From these studies it has been proposed that terminal sugar residues may be removed from the nonreducing end of the N-glycans of plasma membrane glycoproteins. In distinct plasma membrane glycoproteins, even mannose residues were lost from the glycoproteins (11). Second, studies designed to examine the return of surface receptors to compartments of the secretory pathway have demonstrated that selected cell surface glycoproteins may also acquire terminal sugars, L-fucose, and sialic acid when recycling to fucosyl-and sialyltransferases in the medial/trans-Golgi and in the TGN (12)(13)(14)(15)(16)(17)(18)(19). In Chinese hamster ovary cells, the cation-independent mannose 6-phosphate/insulin-like growth factor-II receptor has been reported to recycle even to galactosyltransferases in the trans-Golgi region (19). It has been proposed that reglycosylation might serve as a repair mechanism for surface glycoproteins trimmed by glycosidases encountered on the cell surface or during endocytosis and recycling and that cell surface glycoproteins may pass several rounds of de-and reglycosylation (15,18,20). However, as compared with recycling to glycosyltransferases, far less is known about postbiosynthetic trimming of cell sur-face glycoproteins by glycosidases. In an important study, Snider and Rogers (21) demonstrated that TfR and glycoproteins from the total cellular protein pool may return to mannosidase I in the early Golgi region in K 562 cells. This was shown in that cells were metabolically labeled with [ 3 H]mannose in the presence of the reversible mannosidase I inhibitor 1-deoxymannojirimycin (dMM). Glycoproteins synthesized under these conditions retained immature oligomannosidic Nglycans during their initial transport through the Golgi complex. A return to early Golgi mannosidase I and a subsequent passage through peripheral Golgi elements was noticed by trimming of the immature oligomannosidic N-glycans and conversion to complex-type structures during reculture of cells in the absence of dMM. Employing this experimental strategy, a return to early Golgi mannosidase I was also shown for the cation-dependent and the cation-independent mannose 6-phosphate receptor in BW 5147 mouse lymphoma cells (12). It was calculated that glycoproteins recycle to mannosidase I at very low rates with half-times of 12 h for the total glycoprotein pool in K562 cells (19) and ϳ20 h for both mannose 6-phosphate receptors in BW 5147 cells (12). In none of these studies, however, was a return to Golgi mannosidase I examined by a sample of glycoproteins that had been covalently labeled on the cell surface beforehand. Hence, it could not be distinguished whether glycoproteins trimmed by early Golgi mannosidase I recycled from the cell surface or from other post-Golgi locations such as the TGN, secretory vesicles, endosomes, or lysosomes or even represented, in the case of the total glycoprotein pool, at least partly glycoproteins resident in the Golgi complex. Moreover, it remained unknown whether glycoproteins return to the cell surface after reentering the early Golgi. A recent study designed to examine the transport of TfR and DPPIV from the cell surface to compartments of the secretory pathway in HepG2 cells showed that oligomannosidic N-glycans of these two glycoproteins were not converted to complex structures during recycling (18). In accordance with this finding, Neefjes et al. (22) failed to detect conversion of oligomannosidic to complex-type glycans on recycling glycoproteins including TfR and HLA class II antigens in different cell lines, indicating that these proteins do not encounter mannosidase I in the early Golgi (i.e. at sites proximal to N-acetylglucosaminyltransferase I). This enzyme is localized in medial Golgi elements (23) and initiates the further processing of the oligomannosidic trimming intermediate Man 5 GlcNAc 2 to complex-type oligosaccharides. In a recent immunohistochemical study, however, mannosidase I was found to be less compartmentalized than previously assumed and was also detected in the medial and the trans-Golgi and, in some cell types, even in the TGN, in secretory vesicles, and in the plasma membrane (24). Hence, it became conceivable that cell surface glycoproteins could return to mannosidase I also at sites peripheral to N-acetylglucosaminyltransferase I, resulting in trimming of oligomannosidic glycans without further processing to complex N-glycans. In that case, membrane glycoproteins retaining trimmed oligomannosidic glycans might return to the plasma membrane and become exposed on the cell surface.
In the present paper, this assumption was examined in two hepatoma cell lines and primary rat hepatocytes by a sample of cell surface glycoproteins of well defined function, i.e. the TfR, the serine peptidase DPPIV, the cell adhesion molecule gp110/ cell-CAM105 (25) (a member of the Ig superfamily), and LIcadherin (26) (a member of the cadherin family of cell adhesion molecules).
Using a strategy based on the experimental design initially described by Snider and Rogers (21), in conjunction with covalent labeling of cell surface glycoproteins with NHS-SS-biotin, it was determined whether plasma membrane glycoproteins might be trimmed by mannosidases after transport to the cell surface. Moreover, to examine whether subsequently glycoproteins carrying trimmed oligomannosidic N-glycans are exposed on the cell surface, glycoproteins were allowed to encounter mannosidases and were, thereafter, isolated selectively from the cell surface. In this report, we present evidence that Nglycans of selected cell surface glycoproteins are postbiosynthetically trimmed from Man 8 -9 GlcNAc 2 to Man 5-7 GlcNAc 2 . Following demannosylation, glycoproteins carrying trimmed oligomannosidic structures become exposed on the cell surface, indicating that demannosylation occurs at sites peripheral to N-acetylglucosaminyltransferase I, either at the cell surface or during endocytosis and recycling back to the cell surface.
Metabolic Labeling of Cells-Confluent layers of cells were trypsinized and seeded on collagen I-coated dishes (50-mm diameter). Cells were allowed to adhere overnight. For labeling in the polypeptide moiety, the monolayers were washed and preincubated for 60 min in DMEM without L-methionine/L-cysteine. The cells were pulse-labeled for 4 h with L-[ 35 S]methionine/L-[ 35 S]cysteine (5.5 Mbq/3 ϫ 10 6 cells) and then chased for 3 h in DMEM with 1 mM unlabeled L-methionine/ L-cysteine. When used, 3 mM dMM was present during the preincubation, pulse, and chase periods. For labeling of glycoproteins in the oligosaccharide moiety, cells were washed, preincubated for 60 min in the presence of 3 mM dMM, and then labeled for 8 h with D-[2,6-3 H]mannose (14.8 Mbq/6 ϫ 10 6 cells) in glucose-free DMEM supplemented with 3 mM dMM, 5 mM D-galactose, and 10 mM pyruvate. Cells were then washed and chased for 3 h in the presence of 3 mM dMM.
Labeling of Cell Surface Proteins with NHS-SS-biotin-Cell surface proteins were labeled with NHS-SS-biotin essentially as described (18). After cooling on ice, cells were washed four times with ice-cold PBS/ Ca 2ϩ /Mg 2ϩ (PBS containing 0.9 mM CaCl 2 and 0.5 mM MgCl 2 ) and incubated with a freshly prepared solution of NHS-SS-biotin (1 mg/ml) in PBS/Ca 2ϩ /Mg 2ϩ for 20 min at 4°C. Cells were then washed twice with PBS/Ca 2ϩ /Mg 2ϩ containing 0.1% (w/v) bovine serum albumin and twice with PBS/Ca 2ϩ /Mg 2ϩ and were then either recultured or harvested for further analysis.
Isolation of Biotinylated Proteins-Cells were extracted as detailed above with lysis buffer A, which additionally contained 10 mM L-lysine (lysis buffer B). A cellular glycoprotein fraction was prepared from detergent extracts by binding to ConA-Sepharose. For this purpose, ConA-Sepharose (0.5 g of ConA-Sepharose/5 mg of protein) was added to the detergent extract and slowly shaken at 4°C for 12 h. ConA-Sepharose was then pelleted by centrifugation and washed four times in lysis buffer B. Bound glycoproteins were eluted with 200 mM ␣-meth-ylmannopyranoside in lysis buffer B. Biotinylated cell surface glycoproteins were then isolated from this cellular glycoprotein fraction by binding to streptavidin-agarose. Streptavidin-agarose (100 g/mg of protein) was added, and the mixture was shaken at 4°C for 4 h. After washing five times in lysis buffer A, biotinylated proteins were eluted by boiling for 3 min in 100 l of 0.4% SDS, 5% mercaptoethanol. Biotinylated DPPIV was isolated as described previously (18). Briefly, the protein was immunoadsorbed from detergent extracts and eluted from protein A-Sepharose with 3 M KSCN, 0.5% Nonidet P-40. Eluates were diluted 1:10 in dilution buffer (10 mM Tris/HCl, pH 8.0, 150 mM NaCl, 1 mM EDTA, 0.1% Nonidet P-40), and streptavidin-agarose was added. The suspension was shaken at 4°C for 4 h; washed four times in 50 mM Tris/HCl, pH 8.0, 500 mM NaCl, 1 mM EDTA, 1% Nonidet P-40, 0.3 M KSCN; washed twice in the same buffer without KSCN; and washed once with PBS. Biotinylated DPPIV bound to streptavidinagarose was eluted by boiling for 3 min in 100 l of 0.4% SDS, 5% mercaptoethanol.
Treatment of Glycoproteins with Endo H and PNGase F-Immunoabsorbed glycoproteins were eluted from protein A-Sepharose by boiling for 3 min in 0.4% SDS, 5% mercaptoethanol, and 10 mM EDTA. Biotinylated proteins bound to streptavidin-agarose were eluted with the same buffer (100 l per 100 l of streptavidin-agarose). Before the addition of the glycosidases, all samples were diluted 4-fold in the buffer recommended by the manufacturer. The diluted samples were cleared by centrifugation followed by the addition of a mixture of proteinase inhibitors (leupeptin, chymostatin, antipain, pepstatin) each at a final concentration of 20 g/ml. Treatment with 10 milliunits of Endo H from Streptomyces plicatus (EC 3.2.1.96) was performed in 50 mM sodium phosphate, pH 6.0, 0.1% Nonidet P-40 for 20 h at 37°C. Treatment with PNGase F (EC 3.2.2.18) from Flavobacterium meningosepticum was performed in 50 mM sodium phosphate, pH 8.0, containing 1% (v/v) decanoyl-N-methylglucamide for 20 h at 37°C.
Preparation of High Mannose-type Oligosaccharide Alditols-High mannose-type oligosaccharides were prepared from immunoadsorbed glycoproteins or from biotinylated proteins eluted from streptavidinagarose as described previously (36,37). Briefly, oligosaccharides were released from glycoproteins by incubation with Endo H and then separated from proteins by ultrafiltration through Centricon-10 microconcentrators. The filtrates were desalted by mixed bed ion exchange chromatography on a column (0.6 ϫ 10 cm) containing 500 l of Dowex ]cysteine for 4 h followed by a 3-h chase, either in the presence or absence of 3 mM dMM. Cells were detergent-extracted, and LI-cadherin, DPPIV, gp110/cell-CAM105, and TfR were immunoadsorbed as described under "Experimental Procedures." Immunoadsorbed glycoproteins from untreated cells were divided into three aliquots and were mock-incubated (lanes 1, 6, 11, and 16), treated with Endo H (lanes 2, 7, 12, and 17), or treated with PNGase F (lanes 3, 8, 13, and 18). Glycoproteins immunoadsorbed from dMM-treated cells were divided into two portions and were either mock-incubated (lanes 4, 9, 14, and 19) or treated with Endo H (lanes 5, 10, 15, and 20). After incubation, samples were separated by SDS-PAGE and visualized by radiofluorography.
AG50W-X12 and 500 l of Dowex AG3-X4. After washing the column with four bed volumes of water, the combined filtrates were dried by evaporation. The oligosaccharides were converted to their corresponding oligosaccharide alditols by reduction with sodium borohydride in 0.2 M sodium borate, pH 9.2, for 6 h at 30°C. The reaction was stopped by the dropwise addition of 1 M acetic acid. The solution was adjusted to pH 5.0 and passed through a cation exchange column containing 4 ml of AG 50W-X12. The column was washed with five bed volumes of water, and the combined filtrates were evaporated to dryness at 30°C. Boric acid was completely removed by 5-fold evaporation with 1 ml of methanol, and traces of acetic acid were removed by drying the oligosaccharides over sodium hydroxide in a desiccator.
HPLC Separation of Oligosaccharide Alditols-HPLC separation of the oligosaccharide alditols was performed as described previously (37) using a Bio-Rad model 700 chromatography workstation equipped with two Bischoff (Leonberg, Germany) model 2200 pumps, a Knauer (Berlin, Germany) dynamic mixing chamber, and a Shimadzu fluorescence HPLC monitor RF-535. Briefly, oligosaccharide alditols were separated on two Spherisorb-NH 2 columns (4.6 ϫ 250 mm, 5 m; Bischoff) equilibrated with a mixture containing 65% acetonitrile and 35% 15 mM sodium dihydrogenphosphate, pH 5.2, at a flow rate of 1.5 ml/min and were eluted by decreasing the proportion of acetonitrile to 45% within 100 min. Fractions of 1 ml were collected and assayed for radioactivity by liquid scintillation counting using a Tri-Carb 1900 CA liquid scintillation analyzer (Canberra Packard). A mixture of glucose oligomers (n ϭ 1-20), fluorescence-labeled by reductive amination with 8-amino-2-naphthol, was used as an internal standard. Columns were calibrated with authentic oligosaccharide alditols Man 5-9 GlcNAcOH, prepared from HA 2 subunits of influenza virus hemagglutinin by Endo H treatment and NaBH 4 reduction after metabolic labeling with D-[2-3 H]mannose (38).
High Performance Anion Exchange (HPAE) Separation of Oligosaccharides-In some experiments, oligosaccharides were released from glycoproteins with PNGase F according to Anumula and Taylor (39) with some modifications and separated by HPAE chromatography. Briefly, biotinylated proteins were eluted from streptavidin-agarose with 0.4% SDS, 5% mercaptoethanol by boiling for 3 min. Eluted proteins were concentrated in a Centricon-10 microconcentrator and suspended in 0.5% NH 4 HCO 3 , pH 8.0. Trypsin (2% by mass) was added to the samples from a fresh stock solution of 20 mg/ml in 0.5% NH 4 HCO 3 and incubated at 37°C for 16 h. Trypsin was inactivated by heating at 100°C for 5 min. After cooling, the pH was readjusted to 8.5, and PNGase F was added and incubated at 37°C for 16 h in a shaker. Oligosaccharides were purified by passing the samples through a column containing AG3-X4 in the bottom layer and AG50W-X12 in the top layer. After washing with two bed volumes of water, the filtrates were dried by evaporation. Oligosaccharides were separated using a Dionex (Sunnyvale, CA) DX-300 system and a CarboPac PA-100 (4 ϫ 250 mm) in series with a CarboPac PA-100 guard column as described (18). Columns were calibrated with authentic oligosaccharides Man 5-9 GlcNAc 2 .

RESULTS
To examine whether cell surface glycoproteins lose mannose residues from their oligomannosidic N-glycans during their life span, the following protocol was employed. Cells were metabolically labeled with either L-[ 35 S]methionine/L-[ 35 S]cysteine or D-[2,6-3 H]mannose in the presence of the Golgi-mannosidase I inhibitor dMM. As a consequence, the N-glycans of newly synthesized glycoproteins normally processed to the complex type retain high mannose-type structures. After a chase period sufficient to allow the passage of the newly synthesized glycoproteins through the Golgi region to the cell surface, the inhibition of mannosidase I was reversed by washout of the inhibitor. Cells were then recultured for different times, and the plasma membrane glycoproteins DPPIV, LI-cadherin, TfR, and gp110/ cell-CAM105 were isolated and analyzed for trimming of their oligomannosidic glycans.
Characterization of the N-Glycosylation of LI-cadherin, DP-PIV, gp110/cell-CAM105, and TfR-To characterize the Nglycans of LI-cadherin, DPPIV, gp110/cell-CAM105, and TfR generated either in the absence or in the presence of dMM, MH 7777 cells were pulse-labeled with L-[ 35 S]methionine/L-[ 35 S]cysteine and chased for 3 h. When used, dMM was present during the pulse and the chase period. DPPIV, LI-cadherin, TfR, and gp110/cell-CAM105 were then isolated by immunoadsorption and were analyzed by SDS-PAGE and radiofluorography. LI-cadherin and DPPIV immunoadsorbed from dMMtreated cells had a molecular mass of approximately 110 and 100 kDa, respectively, as assessed from migration on 7.5% SDS-polyacrylamide gels (Fig. 1, lanes 4 and 9). GP110/cell-CAM105 and TfR from these cells migrated as doublets with a molecular mass of 84/78 kDa (lane 14) and 92/88 kDa (lane 19), respectively. As has been shown previously, the doublet observed for gp110/cell-CAM105 represents two variants generated by alternative splicing that differ in the size of their C-terminal cytoplasmic domains (40). The doublet observed for the TfR has been reported for a wide variety of cell types and most likely represents the phosphorylated and the nonphosphorylated form of the receptor (41). Glycoproteins synthesized For the examination of the complex-type glycoforms of DPPIV and gp110/cell-CAM105, cells were labeled, chased, and recultured in the absence of dMM. DPPIV and gp110/cell-CAM105 were immunoadsorbed from the detergent extracts either immediately after the chase (0 h; lanes 1, 3, 5, 7, 9, 11, 13, and 15) or after the reculture period (70 h; lanes 2, 4, 6, 8, 10, 12, 14, and 16), divided into two portions, and either mock-incubated (lanes 1 and 2, 5 and 6, 9 and 10, and 13 and 14) or treated with Endo H (lanes 3 and 4, 7 and 8, 11 and 12, and 15 and 16). Samples were separated by SDS-PAGE and visualized by radiofluorography. as shown in Fig. 1. After washout of the inhibitor, cells were recultured in the absence of dMM. DPPIV, gp110/cell-CAM105, LI-cadherin, and TfR were immunoadsorbed at different times and were analyzed by SDS-PAGE and radiofluorography ( Fig.  2A). A distinct decrease in the molecular mass was noted for DPPIV and gp110/cell-CAM105. DPPIV obtained immediately after the chase (0 h) had a molecular mass of approximately 100 kDa, which shifted to 95 kDa during reculture. The decrease in the molecular mass of gp110/cell-CAM105 was more prominent, probably due to its high carbohydrate content of approximately 50% (25). No decrease in the molecular mass was detectable for the high mannose-type glycoforms of LIcadherin and TfR. Immunoprecipitates of DPPIV obtained during reculture contained an additional faint polypeptide band with a molecular mass of approximately 110 kDa that corresponded to the molecular mass of mature complex-type DPPIV and might represent some DPPIV processed to the complextype glycoform. For comparison, the behavior of the complextype glycoforms of the four glycoproteins, labeled and chased in the absence of dMM, was analyzed. In contrast to the high mannose-type glycoforms of DPPIV and gp110/cell-CAM105, none of the four glycoproteins with complex-type N-glycans exhibited a decrease in the molecular mass during reculture (shown for DPPIV and gp110/cell-CAM105 in Fig. 3, lanes 9  and 10). To exclude the possibility that the observed reduction in the molecular mass of the high mannose-type glycoforms of DPPIV and gp110/cell-CAM105 is a particular feature of transformed MH 7777 cells, the same experiments were performed with primary cultured rat hepatocytes. In accordance with the results obtained in MH 7777 cells, the molecular mass of the high mannose glycoforms of DPPIV and gp110/cell-CAM105 decreased after 70 h of reculture (Fig. 3, lanes 5 and 6), and no decrease in the molecular mass was observed for the complextype glycoforms of the two glycoproteins (Fig. 3, lanes 13 and  14). Whereas in MH 7777 cells the two isoforms of the oligomannosidic gp110/cell-CAM105 were labeled to a similar ex-tent, in hepatocytes the 78-kDa isoform prevailed (Fig. 3,  gp110, lanes 1 and 5). The 84-kDa isoform was only faintly detectable and could not be assessed. In accord with the results shown in Fig. 1 (lanes 11 and 12) the complex-type glycoforms of the two isoforms of gp110/cell-CAM105 generated in the absence of dMM migrated as a very broad band and could not be discriminated (Fig. 3, gp 110, lanes 9 -16). Taken together, these results indicate that the high mannose glycoforms of DPPIV and of gp110/cell-CAM105 undergo a trimming process most likely by loss of mannose residues from the nonreducing end of the oligomannosidic N-glycans.
The Decrease in the Molecular Mass of the High Mannose Glycoforms of DPPIV and gp110/cell-CAM105 Is Due to Glycan Trimming and Not to Limited Proteolysis-To rule out that the decrease in the molecular mass of the high mannose-type glycoforms of DPPIV and gp110/cell-CAM105 during reculture is caused by limited preoteolysis of the polypeptide backbone, the sizes of the deglycosylated polypeptides were compared immediately after the chase and after 70 h of reculture. MH 7777 cells and hepatocytes were radiolabeled with L-[ 35 S]methionine/L-[ 35 S]cysteine and chased, both in the presence of dMM, and were recultured after washout of the inhibitor. DPPIV and gp110/cell-CAM105 immunoadsorbed immediately after the chase (Fig. 3, lanes 1 and 5), and after 70 h of reculture (Fig. 3, lanes 2 and 6) they were digested with Endo H. Digestion with the endoglycosidase converted both forms of the glycoproteins (DPPIV: 0 h, 100 kDa; 70 h, 95 kDa; gp110/cell-CAM105: 0 h, 84/78 kDa; 70 h, 78/72 kDa) to polypeptides exhibiting the same molecular mass of approximately 88 kDa as for DPPIV and 58/48 kDa as for gp110/cell-CAM105 (Fig. 3, lanes 3 and 4  (MH 7777 cells) and lanes 7 and 8 (hepatocytes)). The additional polypeptide band of 110 kDa present in immunoprecipitates of DPPIV obtained from MH 7777 cells after reculture (Fig. 3, DPPIV, lanes 2 and 4) was resistant to Endo H and hence most likely reflects reprocessing of some DPPIV to the complex-type glycoform as mentioned above. These experi-

FIG. 4. High mannose-type N-glycans of DPPIV are postbiosynthetically trimmed from Man 8 -9 GlcNAc 2 to Man 5-7 GlcNAc 2 . MH 7777 cells were metabolically labeled with D-[2,6-3 H]mannose for 8 h followed by a 3-h chase, both
in the presence of 3 mM dMM. Cells were then recultured either in the absence or presence of 3 mM dMM. DPPIV and TfR were immunoadsorbed from the detergent extracts, and oligosaccharides were released from the glycoproteins by digestion with Endo H. Oligosaccharides were converted to their corresponding oligosaccharide alditols by reduction with NaBH 4 and were analyzed by HPLC as detailed under "Experimental Procedures." Separation of the high mannose oligosaccha- ments demonstrate that the decrease in the molecular mass does not result from limited proteolysis but reflects trimming of the oligomannosidic N-glycans.
High Mannose-type N-Glycans of Surface Glycoproteins Are Trimmed from Man 8 -9 GlcNAc 2 to Man 5-7 GlcNAc 2 after Transport to the Cell Surface-To characterize the mannose trimming of the high mannose N-glycans of DPPIV and gp110/cell-CAM105, in particular to determine the end product of the trimming, oligosaccharides were analyzed immediately after washout of dMM, i.e. after the chase period, and after reculture using the following protocol. MH 7777 cells were radiolabeled with D-[2,6-3 H]mannose and then chased, both in the presence of dMM, and recultured in the absence of dMM. From the glycoproteins exhibiting either a decrease in the molecular mass during reculture (DPPIV and gp110/cell-CAM105) or not (TfR and LI-cadherin), DPPIV and TfR were analyzed. The glycoproteins were immunoadsorbed immediately after the chase (0 h) and after 85 h (DPPIV) or 70 h (TfR) of reculture. The radiolabeled oligosaccharides were released from the glycoproteins by Endo H and then converted into their corresponding oligosaccharide alditols by reduction with NaBH 4 and separated by HPLC (Fig. 4). Columns were calibrated with authentic oligosaccharide alditols (Man 5-9 GlcNAcOH), and samples were mixed with fluorescence-labeled glucose oligomers as internal standards. From DPPIV and TfR isolated immediately after the chase (0 h), Man 9 GlcNAcOH and Man 8 GlcNAcOH were obtained as the major components (Fig.  4, A and D), in line with the inhibitory effect of dMM on processing mannosidases (for reviews, see Refs. 4 and 5). After reculture of cells, TfR and DPPIV exhibited a different pattern in the sizes of oligomannosidic N-glycans. In the case of DPPIV, Man 5 GlcNAcOH was the major component, while Man 6 -9GlcNAcOH were present as minor components (Fig. 4B). These results demonstrate that high mannose-type N-glycans of DPPIV are trimmed from Man 8 -9 GlcNAc 2 to Man 5 GlcNAc 2 during reculture with Man 6 -7 GlcNAc 2 , representing trimming intermediates. In contrast to DPPIV, no decrease in the size of the high mannose-type N-glycans was detectable for the TfR during reculture of cells (Fig. 4E). To examine whether DPPIV being trimmed from Man 8 -9 GlcNAc 2 to Man 5 GlcNAc 2 derives in fact from the plasma membrane, demannosylation of DPPIV was examined after covalent labeling with NHS-SS-biotin at the cell surface, as schematized in Fig. 5A. This approach allows us to unambiguously discriminate between cell surface proteins and proteins localized in intracellular compartments. Cells were radiolabeled and chased as above and then surfacelabeled at 4°C with NHS-SS-biotin prior to reculture. Biotinlabeled surface DPPIV was isolated by immunoadsorption in conjunction with affinity chromatography on streptavidin-agarose. HPAE-separation of the high mannose-type oligosaccharides released from biotinylated DPPIV by PNGase F revealed the same shift from Man 8 -9 GlcNAc 2 to Man 5-7 GlcNAc 2 structures (Fig. 6, A and B) that was observed for total cellular DPPIV (Fig. 4, A and B). This clearly demonstrates that trimming of high mannose-type N-glycans affects DPPIV molecules that have exited the secretory pathway and were exposed at the cell surface. Demannosylation was also observed for the bulk of cell surface glycoproteins in MH7777 cells (Fig. 6, C and D), showing that this process is not restricted to DPPIV. In an attempt to find out whether a class I or class II mannosidase is involved that is known to be inhibited by dMM or swainsonine, respectively (for reviews, see Refs. 4 and 5), reculture was performed in the presence of either one of the inhibitors. When cells were recultured in the presence of dMM, trimming of DPPIV was completely blocked (Fig. 4C), which is in agreement with previous reports (12,21). The inhibitory effect of dMM on demannosylation could also be demonstrated by the finding that the molecular mass of DPPIV and gp110/cell-CAM105 as analyzed by SDS-PAGE did not decrease during reculture in the presence of the inhibitor (Fig. 2B). In contrast, in the presence of swainsonine (3 g/ml) the high mannose-type Nglycans of DPPIV were trimmed to the same extent as in the absence of the inhibitor (not shown). In summary, these results show that high mannose-type glycans of cell surface glycopro- To determine whether the oligomannosidic glycans of glycoproteins are trimmed after their transport to the cell surface, glycoproteins were additionally labeled with NHS-SS-biotin at the cell surface prior to the reculture period (A). To examine whether trimmed glycoproteins are exposed at the cell surface at the end of the reculture period, i.e. after demannosylation occurred, glycoproteins were labeled with NHS-SS-biotin at the cell surface immediately after the reculture period (B). Biotinylated proteins were isolated selectively and analyzed for mannose trimming. teins are postbiosynthetically trimmed by a dMM-sensitive, swainsonine-resistant ␣-mannosidase.
In two previous studies on the recycling of cell surface glycoproteins in HepG2 cells, no reconversion of oligomannosidic to complex-type glycans could be detected for DPPIV, TfR, and HLA class I antigens (18,22), indicating that these proteins do not recycle to the cis-Golgi in this cell line. To find out whether, nevertheless, plasma membrane glycoproteins might be subject to mannose trimming, demannosylation was also examined in HepG2 cells. As is shown in Fig. 7, N-glycans of cell surface DPPIV were trimmed from Man 8 -9 GlcNAc 2 to Man 5-7 GlcNAc 2 essentially as found for surface DPPIV in MH 7777 cells. In conjunction with previous reports (18,22), these data show that high mannose-type glycans of DPPIV can be trimmed without further processing to complex structures.
Kinetics of Postbiosynthetic Mannose Trimming-A quantitative analysis of the kinetics of postbiosynthetic demannosylation was performed for total cell surface proteins in HepG2 cells. In these experiments, high mannose-type glycans of cell surface-labeled glycoproteins were analyzed after different times of reculture. Data from HPAE fractionations (shown in Fig. 8A) were quantitated by totaling the radioactivity in each peak and by correcting for the number of mannose residues. In Fig. 8B, the radioactivity of each oligosaccharide species, expressed as a percentage of the total oligosaccharides recovered, is plotted versus time of reculture. As can be seen from the hydrolysis curves, Man 9 GlcNAc 2 is converted to Man 5 GlcNAc 2 with Man 6 GlcNAc 2 , Man 7 GlcNAc 2 , and Man 8 GlcNAc 2 formed as trimming intermediates. The extent of demannosylation at each time point, defined as the conversion of Man 8 -9 GlcNAc 2 to Man 5-7 GlcNAc 2 , was calculated based on the amount of Man 8 -9GlcNAc 2 obtained immediately after the chase and plotted versus time. As can be seen in Fig. 8C, the extent of demannosylation increased during the first 24 h and then reached a plateau. With the assumption of first order kinetics, demannosylation followed the equation D t ϭ D t 3 ϱ ⅐(1 Ϫ e Ϫkt ), where D t is the extent of demannosylation at the time t ϭ x. The constant k can be determined from this equation as the negative slope of a plot of ln(D t 3 ϱ Ϫ D t ) over the time (Fig. 8D). For total cell surface glycoproteins of HepG2 cells, a half-time of 6.7 h was calculated for demannosylation according to the equation t1 ⁄2 ϭ ln 2/k.
Trimmed Glycoproteins Can Be Isolated from the Cell Surface after Postbiosynthetic Demannosylation-Trimming of plasma membrane glycoproteins after their transport to the cell surface could occur either at the cell surface or after endocytosis in intracellular compartments. In the latter case, it could not be distinguished on the basis of the previous experiments whether trimmed glycoproteins remain in intracellular compartments or return to the cell surface. To examine whether trimmed glycoproteins are exposed at the cell surface, mannose trimming was analyzed for glycoproteins isolated selectively from the cell surface at the end of the reculture period. To do so, MH 7777 cells were radiolabeled and recultured as in the experiments shown in Fig. 4. After reculture, proteins exposed at the cell surface were labeled with biotin immediately prior to harvesting and solubilization of cells and were then analyzed for mannose trimming, as schematized in Fig.  5B. HPLC separation of the high mannose-type glycans released from the biotinylated glycoproteins revealed that surface glycoproteins isolated after reculture carried mainly Man 5 GlcNAc 2 and minor amounts of Man 6 -9 GlcNAc 2 (Fig. 9B). In accordance with the results shown in Figs. 4 and 6, glycoproteins isolated immediately after the chase carried mainly Man 8 -9 GlcNAc 2 (Fig. 9A). These results demonstrate that after demannosylation had occurred, trimmed glycoproteins were exposed at the cell surface. This provides additional evidence that demannosylation can occur without further processing to complex structures. DISCUSSION Two major conclusions can be drawn from the results of the present study. First, after exit from the secretory pathway and transport to the cell surface, selected cell surface glycoproteins undergo trimming of their oligomannosidic N-glycans by ␣-mannosidase(s) sensitive to dMM. The observed demannosylation of cell surface glycoproteins results in the conversion of Man 8 -9 GlcNAc 2 species to Man 5-7 GlcNAc 2 with Man 5 GlcNAc 2 being the major product. Second, subsequent to demannosylation, trimmed glycoproteins are exposed at least partly at the cell surface. Hence, it is likely that demannosylation of glycoproteins occurs at sites peripheral to N-acetylglucosaminyltransferase I either at the cell surface or during endocytosis and recycling back to the cell surface. The data clearly demonstrate that modification of plasma membrane glycoproteins by trimming of their oligomannosidic N-glycans is not restricted to biosynthesis. These conclusions are based on the following evidence.
Analysis of the size of high mannose-type N-glycans of cell surface glycoproteins demonstrated that during reculture of MH 7777 cells and HepG2 cells Man 8 -9 GlcNAc 2 species were trimmed to Man 5 GlcNAc 2 with Man 6 GlcNAc 2 and Man 7 GlcNAc 2 formed as trimming intermediates. Mannose trimming has also been observed for the cation-dependent and the cation-independent mannose 6-phosphate receptors in BW 5147 mouse lymphoma cells (12) and for TfR and for the total cellular glycoprotein pool in K562 cells (21). However, in both of these studies, mannose trimming was not examined by a sample of glycoproteins after previous labeling on the cell surface. Therefore, it remained unknown whether glycoproteins modified by mannosidase I were derived from the plasma membrane or from intracellular compartments. In comparison, by analyzing membrane glycoproteins that had been labeled with biotin at the cell surface prior to reculture of cells (schematized in Fig.  5A), we were able to examine selectively the fate of cell surface glycoproteins. The results of these experiments unequivocally demonstrate that demannosylation of oligomannosidic N-glycans affects glycoproteins after they have been delivered to the cell surface. Moreover, previous studies did not address the question of whether trimmed glycoproteins become exposed at the cell surface. By analyzing glycoproteins that were allowed to encounter mannosidases and were thereafter isolated selectively from the cell surface (schematized in Fig. 5B), it is clearly shown in the present study that subsequent to demannosylation trimmed glycoproteins are present at the cell surface.
Which ␣-mannosidase is involved in the postbiosynthetic mannose trimming remains to be established. Based on the finding that postbiosynthetic demannosylation could be inhib- After surface labeling with NHS-SS-biotin, cells were recultured for the times indicated. Biotinylated cell surface glycoproteins were isolated after different times of reculture as detailed under "Experimental Procedures," and oligosaccharides were released and analyzed as described in the legend to Fig. 6. B, data from HPAE fractionations were quantitated by totaling the radioactivity in each peak and by correcting for the number of mannose residues. The radioactivity of each oligosaccharide species was expressed as a percentage of the total oligosaccharides recovered and plotted versus time. ϫ, Man 9 GlcNAc 2 ; ࡗ, Man 8 GlcNAc 2 ; OE, Man 7 GlcNAc 2 ; E, Man 6 GlcNAc 2 ; f, Man 5 GlcNAc 2 . C, the extent of demannosylation at each time point, defined as the conversion of Man 8 -9 GlcNAc 2 to Man 5-7 GlcNAc 2 , was calculated based on the amount of Man 8 -9 GlcNAc 2 obtained immediately after the chase. D, plot of ln(A t 3ϱ Ϫ A t versus time. A linear regression leads to k ϭ 0.1028 (R 2 ϭ 0.969).
ited by dMM, and since the oligomannosidic N-glycans were converted from Man 8 -9 GlcNAc 2 species to Man 5 GlcNAc 2 , it is likely that class I ␣-mannosidases of the ER and the Golgi complex are involved. These enzymes are known to cleave up to four mannose residues from Man 9 GlcNAc 2 to yield Man 5 GlcNAc 2 during the maturation of N-linked oligosaccharides. In contrast to class I ␣-mannosidases of the ER and the Golgi complex, the class II ␣-mannosidases (Golgi ␣-mannosidase II, ER/cytosolic ␣-mannosidase, and the lysosomal mannosidase (for review, see Refs. 4 and 5)) cannot account for the observed trimming reactions. These enzymes are dMM-resistant and inhibited to some extent by swainsonine, a mannose analogue that had no effect on demannosylation. In addition, class II ␣-mannosidases do not have the specificities required to account for the trimming from Man 9 GlcNAc 2 to Man 5 GlcNAc 2 and therefore cannot be involved in postbiosynthetic demannosylation. Several as yet unclassified ␣1,2/1,3/ 1,6-mannosidases have been described, purified from rat brain (42), rat sperm (43), and rat liver microsomes (44), that share many common characteristics. Swainsonine and dMM are only weakly inhibitory or not inhibitory to these enzymes. For example, rat liver ␣1,2/1,3/1,6-mannosidase is inhibited by dMM at concentrations more than 100-fold higher than that reported for purified Golgi mannosidase I (44). These ␣1,2/1,3/1,6-mannosidases cleave Man 4 -9 GlcNAc substrates to Man 3 GlcNAc, an oligosaccharide that could not be isolated from cell surface glycoproteins in the present study. Taken together, it is unlikely that these enzymes are involved in postbiosynthetic mannose trimming of cell surface glycoproteins. It should, however, be noted that the substrate specificities of several mannosidases were determined using Man X GlcNAc oligosaccha-rides as substrates and that the enzymes may yield other products with Man X GlcNAc 2 or glycopeptides as substrates as has been shown for the pig liver Man 9 -mannosidase (45) and the neutral ␣-mannosidase from Japanese quail oviduct (46). Apart from the known ␣-mannosidases sensitive to dMM the possibility cannot be excluded that another as yet unidentified dMM-sensitive ␣-mannosidase might be involved.
The subcellular site of postbiosynthetic mannose trimming is unknown. Demannosylation of glycoproteins could occur at the cell surface, during passage through endocytic compartments, or even after return to compartments of the secretory pathway. Trimming of high mannose-type N-glycans at the cell surface seems feasible, since in a recent immunohistochemical study Golgi ␣-mannosidase I has been detected at the cell surface of enterocytes, pancreatic acinar cells, and goblet cells (24). However, it is unknown whether the mannosidase is enzymatically active at this location. Moreover, a rat sperm ␣1,2/1,3/1,6mannosidase has been reported to be an intrinsic plasma membrane component that is enzymatically active when assayed in sperm plasma membranes and intact spermatozoa, respectively (43). As a second possibility, postbiosynthetic mannose trimming could occur in endocytic compartments during endocytosis and recycling of glycoproteins back to the cell surface. It has been shown that rat liver ␣1,2/1,3/1,6-mannosidase activity is enriched in endosomal fractions (47). Although, as discussed above, this enzyme is probably not involved with respect to its substrate specificity, other mannosidases might also be present in endosomes. Third, since several plasma membrane glycoproteins have been shown to return to the Golgi complex and the TGN (12-21), demannosylation could also occur after the return of cell surface glycoproteins to these compartments. From our data, the possibility cannot be excluded that some glycoproteins return to the processing mannosidases in the cis-Golgi, since a small fraction of the oligomannosidic glycoform of DPPIV was processed to the complex form during reculture in MH 7777 cells ( Fig. 2A), whereas no reconversion to the complex glycoform could be detected for surface DPPIV in HepG2 cells (18). From these observations, combined with the fact that surface DPPIV is trimmed from Man 8 -9 GlcNAc 2 species to Man 5-7 GlcNAc 2 in MH 7777 cells as well as in HepG2 cells, we conclude that most of the DPPIV is trimmed without further processing to the complex glycoform. This explanation is supported by the observation that trimmed glycoproteins can be found on the cell surface and is consistent with the recent finding of Velasco et al. (24) that Golgi mannosidase I previously assumed to reside specifically in the cis-Golgi (48) is less compartmentalized. In this study, the enzyme was primarily detected in the medial-and trans-Golgi cisternae, and in some cell types it was also localized in the TGN and even in secretory vesicles. Therefore, in case cell surface glycoproteins return to the Golgi at sites peripheral to N-acetylglucosaminyltransferase I, an enzyme that is localized in medial Golgi elements (23) and initiates the synthesis of hybrid and complex oligosaccharides, it seems feasible that high mannose-type N-glycans of recycling glycoproteins might be trimmed by Golgi mannosidase I without being further processed to complex structures.
Comparison of DPPIV, TfR, gp110/cell-CAM105, and LI-cadherin showed that postbiosynthetic trimming did not affect each of the four glycoproteins. This may be due to differences in the kinetics or routes of internalization and recycling or due to a different susceptibility of the oligomannosidic glycans to trimming mannosidases. Which of these different mechanisms is responsible remains to be established. Although distinct proteins escape demannosylation, this process seems to affect a large number of cell surface glycoproteins, since it could be demonstrated for the bulk of plasma membrane glycoproteins FIG. 9. Trimmed glycoproteins are exposed at the cell surface at the end of the reculture period. MH 7777 cells were radiolabeled, chased, and recultured as described in the legend to Fig. 4. Immediately after the chase (0 h) or after reculture (110 h), cells were labeled with NHS-SS-biotin at the cell surface. Cells were then detergent-extracted and a cellular glycoprotein fraction was prepared from the detergent extracts by affinity chromatography on ConA-Sepharose. Glycoproteins were eluted with 200 mM ␣-methylmannopyranoside. From this total glycoprotein fraction, biotinylated glycoproteins were isolated by affinity chromatography on streptavidin-agarose as described under "Experimental Procedures." Oligosaccharides were released from surface glycoproteins and converted to oligosaccharide alditols as described in the legend to Fig. 4. Separation of the high mannose oligosaccharide alditols of total glycoproteins, isolated from the cell surface immediately after the chase (0 h, A) or after reculture (110 h, B) is shown.
in MH 7777 cells as well as in HepG2 cells. A quantitative analysis of the time course of demannosylation revealed that this process obeyed first order kinetics with a calculated halftime of 6.7 h as determined for total cell surface glycoproteins in HepG2 cells. The process of demannosylation occurs distinctly faster than degradation of [ 35 S]methionine-labeled total membrane proteins in HepG2 cells (t 1/2 ϭ 65 h 2 ). This indicates that cell surface glycoproteins may encounter trimming mannosidase(s) several times during their life span. The physiological role of postbiosynthetic demannosylation is unknown. Trimming of oligomannosidic N-glycans could reflect the occasional removal of mannose residues from surface glycoproteins by a mannosidase present at the cell surface or encountered during endocytosis and recycling. Alternatively, postbiosynthetic processing could provide a means by which cells can modify N-glycans of cell surface glycoproteins. With respect to the role of cell surface glycoproteins in cell-substratum and cell-cell recognition processes, this remodeling of cell surface glycoproteins may be of relevance for cell surface functions.