Glucosidase I, a Transmembrane Endoplasmic Reticular Glycoprotein with a Luminal Catalytic Domain*

We have analyzed the functional domain structure of rat mammary glucosidase I, an enzyme involved in N-linked glycoprotein processing, using biochemical and immunological approaches. The enzyme contains a high mannose type sugar chain that can be cleaved by endo-8-N-acetyl-D-glucosaminidase H without signifi- cantly affecting the catalytic activity. Based on trypsin digestion pattern and the data on membrane topogra- phy, glucosidase I constitutes a single polypeptide chain of 86 kDa with two contiguous domains: a mem- brane-bound domain that anchors the protein to the endoplasmic reticulum and a luminal domain. A cata- lytically active 39-kDa domain could be released from membranes by limited proteolysis of saponin-perme- abilized membranes with trypsin. This domain ap- peared to contain the active site of the enzyme and had the ability to bind to glucosidase I-specific affinity gel. Phase partitioning with Triton X-114 indicated the amphiphilic nature of the native enzyme, consistent with its location as an integral membrane protein, whereas the 39-kDa fragment partitioned in the aqueous phase, a characteristic of soluble polypeptide. These results indicate that glucosidase I is a transmembrane protein with a luminally oriented catalytic do- main. Such an orientation of the catalytic domain may facilitate the sequential of enzyme activity, and the remaining portion was mixed with Laemmli's sample buffer for Western blotting. General Procedures-Protein determinations, peptide mapping, and silver staining were carried out as described elsewhere (7). After 10% SDS-PAGE and electrotransfer to nitrocellulose membrane, the immunoblotting was carried out by using anti-glucosidase I antibodies and either anti-rabbit IgG-alkaline phosphatase or anti-rabbit [lZ5I] IgG (21). Rainbow molecular mass markers kit containing 200-kDa myosin, 92-kDa phosphorylase b, 69-kDa bovine serum albumin, 46- kDa ovalbumin, 30-kDa carbonic anhydrase, 21.5-kDa trypsin inhibitor, and 14-kDa lysozyme was used for determining M, of polypep- tides.

one or more mannosyl residues may be removed and additional sugars may be added as the glycoprotein reaches its final destination, either within a membrane compartment of the cell or secretion into the extracellular environment (1). The processing may begin as soon as the glycosylation of the appropriate asparagine residue has occurred while the polypeptide backbone is still undergoing synthesis on the polysomes. These reactions continue as the newly assembled glycoprotein is being transported from the rough ER through the Golgi compartment during the terminal stages of its assembly (2).
The ensemble of glycosyltransferases and glycosidases involved in oligosaccharide assembly and processing act sequentially and in concert in which the product of one enzyme serves as the substrate for the next (3)(4)(5). Studies on membrane topography indicate that the biosynthesis of intermediates up to Man5GlcNAc2-P-P-dolichol occurs on the cytoplasmic face, whereas the transfer of additional mannose and glucosyl residues takes place in the lumen of the ER (6). The first enzyme of the dolichol cycle, uiz. UDP-G1cNAc:dolichol-P GlcNAc-1-P-transferase has been purified to homogeneity (7) and the membrane orientation of the enzyme, postulated from the cDNA sequence (8), is in agreement with the proposal that the catalytic domain of the enzyme is oriented toward the cytoplasmic face of ER membrane. Elongation of Man5GlcNAc2 -P-Pdolichol to the ultimate precursor, Glc3Man9GlcNAc2-P-P-dolichol, has been proposed to occur within the lumen of ER. Indeed, the catalytic domain of GDP-Man:dolichol-P mannosyltransferase, the enzyme that generates Man-P-dolichol for the elongation reactions, appears to be oriented toward the lumen (9, 10).
Among the processing-specific glycosidases located in the rough ER, glucosidases I and I1 cleave the a-1,2-and a-1,3linked glucosyl residues on the precursor (2, 11,12). The kinetic pattern of the removal of glucosyl residues indicated that the innermost glucosyl residue is removed at a later time point than the outer 2 residues, suggesting the location of glucosidases at different places along the secretory pathway (13). Glucosidase I1 has been localized in smooth and rough ER through immunoelectron microscopy (14, 15). Besides the glucosyl residues, one of the mannose residues in the oligosaccharide moiety is also removed by an a-mannosidase in rough ER (16).
Although the precise molecular basis for the localization of the above glycosidases has not been elucidated, current views favor the idea that the residence of a particular protein in an organelle requires a specific retention signal that retains the protein in that location. Such retention sequences recently have been identified in both soluble (17) and transmembrane (18,19) ER resident proteins. For example, ER residence of the adenoviral E3/19K protein, a type I transmembrane protein, requires the last six amino acids (DEKKMP) at the C-

16587
Membrane Topography of Glucosidase 1 terminal end of the short cytoplasmic tail (20). It is possible that the retention of processing glycosidases in ER could be determined by such sequences in their cytoplasmic tail.
Previously, we reported the purification of glucosidase I from rat mammary gland (ll), the enzyme activity was shown to be modulated as a function of gland ontogeny (21). It was also observed that the hormonal regulation of the enzyme activity in explant cultures was in good agreement with the glycosylation of a-lactalbumin, an N-linked glycoprotein and a characteristic marker of the lactating tissue (22). Given the positioning of glucosidase I in the multienzyme pathway of oligosaccharide assembly and processing, it is likely that this enzyme may play a critical role in the overall biosynthesis and regulation of protein N-glycosylation. In this investigation, we report the functional organization of glucosidase I in the ER membranes. Our results show that it is a transmembrane glycoprotein with a short cytoplasmic tail. The implications of distinct domains of the enzyme vis a vis its retention in ER also are discussed.

MATERIALS AND METHODS
All chemicals and reagents employed in the purification of glucosidase I and preparation of anti-rabbit antibodies against the enzyme were obtained from commercial sources (11). Sprague-Dawley lactating female rats were purchased from Hilltop Lab Animals Inc. (Scottdale, PA). 1-Deoxynojirimycin was a kind gift from Drs. D. Schmidt and Scangos, Bayer AG, Wuppertal, Federal Republic of Germany. Enzymes endo H, endo F, and N-glycnase were obtained from Genzyme (Cambridge, MA). Biotinylated lectins and alkaline-phosphatase substrate kit were purchased from Vector Laboratories (Burlingame, CA). Anti-rabbit IgG and modified trypsin were products of Promega Biotech (Madison, WI). Anti-rabbit ['251]IgG and Rainbow molecular weight markers were products of Amersham Corp. All other analytical grade fine chemicals were purchased from Sigma.
Purification of Glucosidase I and Preparation of Antibodies-The purification of glucosidase I was carried out essentially as described (11). The purified enzyme was stored in 200 mM potassium phosphate buffer, pH 6.8, containing 0.8% Lubrol PX (buffer A) at -70 "C.
Anti-glucosidase I antibodies were raised in rabbit and affinitypurified by the methods described (7,23).
Glucosidase I Assay-Glucosidase I assay was carried out by incubating the crude or pure enzyme preparation with 10,000 cpm of ["C] GlcaMan9GlcNAc2 in 50 mM potassium phosphate buffer containing 0.25% Lubrol PX in a final reaction volume of 100 rl. The reaction incubation was carried out for 30 min at 37 "C, and the reaction was stopped by boiling for 2 min. The release of [I4C]Glc was determined as described (11). Under these assay conditions Saponin and Triton X-114 in the range of 0.05-1% concentration did not affect the glucosidase I assay. One unit of enzyme activity is defined as the amount of enzyme that releases 100 cpm of [14C]Glc under the conditions of assay.
Peptide Mapping-5 pg of pure glucosidase I was subjected to 12.5% SDS-PAGE in Tricine buffer system (24) for 16 h. After staining and destaining, the gel portions corresponding to bands A and B were cut out separately with a razor blade. The peptide mapping of the polypeptide bands was carried out by chemical cleavage with 0.015 M N-chlorosuccinimide in urea:acetic acidwater (1 g:l ml:l ml), as described by Lischwe and Ochs (25). The peptide maps were visualized with silver staining.
Preparation of Subcellular Fractions-The rat mammary microsomes were prepared (26) and fractionated to obtain different subcellular fractions as described (7). Rough ER was purified further (27) and stored in 20 mM MOPS buffer, pH 6.8, containing 0.25 M sucrose (buffer B). The integrity of membrane vesicles was assessed by examining the latency of mannose-6-phosphatase essentially as described (28). Its latency was lost as a function of increasing concentration of the detergent, Lubrol PX, or saponin. The pronounced latency (93-97%) in mannose-6-phosphatase and glucosidase I activities, as a measure of membrane integrity (29), was monitored routinely in all experiments. The maximum activity of mannose-6phosphatase was 0.33 pg of Pi released/min/mg of protein.
Endo Enzyme Digestions-The digestion of pure glucosidase I with endo H, endo F, and N-glycnase was conducted at 37 "C for 2-16 h under toluene (12). A portion of the glucosidase I, following the digestion, was used immediately for determination of the residual enzyme activity, and the remaining portion was analyzed by either 12.5% SDS-PAGE in Tricine buffer system (24) followed by silver staining or by immunoblotting.
Lectin Binding Assays-The enzyme samples after digestion with endo enzymes were resolved on 10% SDS-PAGE and electrotransferred to nitrocellulose. Polypeptides were identified by using biotinylated Con A/avidin-alkaline phosphatase system following the manufacturer's instructions.
Inhibition by Treatment with NEM or Trypsin-For examining the inhibition of glucosidase I activity by NEM, intact or detergentpermeabilized membranes were preincubated with indicated concentrations of NEM for 10 min on ice, and the enzyme activity was determined.
Intact or detergent-permeabilized rough ER membranes were preincubated with sequencing grade modified trypsin at the indicated concentrations for 30 min on ice. Under these conditions, the trypsin concentration at 100 pg/ml did not affect the latency of mannose-6phosphatase significantly. Trypsin inhibitor was added (100 pg/ml), an aliquot of the enzyme was used immediately for the determination of its activity, and the remaining portion was mixed with Laemmli's sample buffer for electrophoresis (30) and Western blotting.
Release of Catalytic Domain-Freshly prepared intact or saponinpermeabilized ER membranes (8 mg/ml) in buffer B were treated either with or without 0.5% saponin in the presence or absence of modified trypsin (50 pg/ml) for 30 min at 20 "C . After the addition of trypsin inhibitor (100 pg/ml), the reaction mixture was centrifuged at 200,000 X g for 1 h in a Beckman refrigerated Airfuge. The final volume of pellets and supernatants was made to 200 p1 with buffer B. Glucosidase I activity in the pellet and the supernatant was determined. A portion of the samples also was analyzed by immunoblotting.
Phase Separation Using Triton X-114"ER membranes (8 mg of protein/ml) were homogenized in Triton X-114 buffer (4% Triton X-114,150 mM NaCl, 20 mM MOPS, pH 7.0) with a Dounce homogenizer and clarified by centrifugation at 200,000 X g for 60 min at 4 "C. Triton X-114-extracted membranes were incubated with or without trypsin (50 pg/ml) for 30 min at 20 "C as described above. The samples then were subjected to phase separation as described (31). Each sample (0.6 ml) was layered above a 0.4-ml sucrose cushion (6% sucrose, 20 mM MOPS, pH 7.0, 150 mM NaC1, buffer C) in a 1.5-ml Eppendorf tube, warmed to 37 "C for 5 min, and centrifuged in a Microfuge at 15,000 rpm for 3 min at 25 "C. The aqueous and detergent phases were diluted with buffer C to 0.8 ml. After concentration, a portion of both phases was used immediately for the determination of enzyme activity, and the remaining portion was mixed with Laemmli's sample buffer for Western blotting. General Procedures-Protein determinations, peptide mapping, and silver staining were carried out as described elsewhere (7). After 10% SDS-PAGE and electrotransfer to nitrocellulose membrane, the immunoblotting was carried out by using anti-glucosidase I antibodies and either anti-rabbit IgG-alkaline phosphatase or anti-rabbit [lZ5I] IgG (21). Rainbow molecular mass markers kit containing 200-kDa myosin, 92-kDa phosphorylase b, 69-kDa bovine serum albumin, 46-kDa ovalbumin, 30-kDa carbonic anhydrase, 21.5-kDa trypsin inhibitor, and 14-kDa lysozyme was used for determining M, of polypeptides.

RESULTS
Characterization of Glucosidase I-Although, the purified glucosidase I exhibited a band of 85 kDa on 10% SDS-PAGE under reducing conditions (21), it was resolved into two bands of 82 and 85 kDa on 12.5% SDS-PAGE in Tricine buffer system (24) (Fig. la). The purified glucosidase I from calf liver exhibited a doublet of about 85 kDa on 10% SDS-PAGE (32). Peptide mapping of the two bands by chemical cleavage with N-chlorosuccinimide revealed identical fragments indicating similarities in their primary structures (Fig. lb). Con A-Sepharose binding suggested that glucosidase I from rat mammary gland is an N-linked glycoprotein. To examine the nature of its sugar moiety, the purified enzyme was digested with endo H, endo F, and N-glycnase. In each case, the digestion resulted in an increased electrophoretic mobility both polypeptides (Fig. !?A), indicating the removal of one high mannose type N-linked oligosaccharide. Biotinylated Con A did not bind either of the polypeptide bands following digestion with endo enzymes, confirming that the sugar moiety was cleaved (Fig. 2B). Next, the purified enzyme was incubated at 37 "C with or without endo H for different time intervals, and its activity was determined. A complete deglycosylation of glucosidase I could be achieved within 2-4 h of incubation (Fig. 3); at this time point, the catalytic activity of the deglycosylated enzyme was comparable with that of the native enzyme. Further incubation under these conditions resulted in a rapid decline in the activity of the enzyme in both control as well as experimental sets, possibly due to enzyme inactivation a t 37 "C. Nevertheless, the catalytically active enzyme after digestion with endo H did not bind Con A-Sepharose (data not shown).  (Fig. &I). The accumulation of a 39-kDa fragment with concomitant decrease in the larger fragments indicates a precursor-product relationship between different fragments. A possibility existed that trypsin digestion of glucosidase I might have generated additional polypeptides that might not be immunoreactive. This was examined by staining the trypsin-generated peptides, after SDS-PAGE, with silver reagents (Fig. 4B). The results showed that the 69-, 55-, and 39-kDa polypeptides are the only major fragments of proteolytic digestion. Whereas the 69-kDa fragment was very faint on silver-stained gel, it was clearly visible on the immunoblot. It was interesting to note that even after 60 min of proteolysis, the activity of glucosidase I was found to be 61% of the control, and at this time of incubation the 39-kDa fragment was the major trypsin-generated peptide. These data also suggested that the 39-kDa fragment may contain the catalytic domain and therefore be enzymatically active. To explore this possibility, the purified enzyme was digested with trypsin under conditions that resulted in the complete disappearance of 85,69, and 55 kDa bands with a concomitant accumulation of the 39-kDa fragment. The reaction mixture was then incubated with N-(5-carboxypentyl)-l-deoxynojirimycin-Affi-Gel 102, a matrix that specifically binds glucosidase I (11). The enzyme activity bound to the affinity gel and could be eluted quantitatively with 100 mM l-deoxynojirimycin. The Western blot analysis revealed the presence of a 39 kDa band in the eluate (Fig. 5). These results confirm that the 39-kDa fragment contains the catalytic domain.
Membrane Orientation of Catalytic Domain-To address the orientation of the catalytic domain of the enzyme within the membranes, we examined its sensitivity to inhibitors and proteases in intact and detergent-permeabilized microsomal vesicles. Inhibition of glucosidase I with NEM was examined with sealed membrane vesicles in the presence and ahsence of detergents. There was no significant inhibition of enzyme activity when intact membranes were used. However, the  enzyme activity was inhibited severely by increasing concentrations of NEM when membranes were disrupted with Lubrol PX. This inhibition was abolished completely if an excess of dithiothreitol was included in the reaction milieu. A similar inhibition by NEM also was observed in the presence of saponin (Fig. 6), a detergent that permeabilizes the membranes without disrupting their integrity (31). These results indicate that NEM inhibited the enzyme by reacting with a site on the polypeptide that is protected by the membranes, but is accessible in the presence of the detergent. The inhibition of the enzyme activity in saponin-permeabilized membrane vesicles also suggested that the site affected by NEM may be luminally oriented.
Intact membranes were incubated with increasing concentrations of trypsin on ice in the presence and absence of detergent. The enzyme activity was not affected when the detergent was excluded; however, in its presence, up to 40% inhibition in activity was observed a t trypsin concentration as high as 100 pg/ml (Fig. 7). This would be possible if proteolytic products are also enzymatically active. The immunoblotting of these samples revealed that in the absence of detergent, the degradation of the enzyme was minimal and under these conditions, only a single fragment of 82 kDa was generated. On the other hand, in presence of the detergent, the 85-kDa polypeptide was degraded sequentially to a 39-kDa fragment, indicating that the tryptic cleavage a t these sites was protected by membranes. These results favor a Enzyme recovery  1. sample before tr-ypsin digestion; lane 2. after trypsin digestion; lnnp 3, supernatant after affinity hinding; and lnnr 4, the eluted fraction. luminal orientation of a proteolytically resistant catalytic domain. Several integral membrane enzymes contain luminally oriented catalytic domains and, in some cases, the catal-ytic domains could be released from saponin-permeabilized membranes by limited proteolysis (33,34). The sealed and saponinpermeabilized microsomal vesicles were subjected to limited proteolysis with trypsin at 20 "C to release the potential catalytic domain of glucosidase I. The digestion had no effect on the total enzyme activity in either intact or saponinpermeabilized membranes (Fig. 8); the activity was found exclusively in the pellet. These results suggest that while saponin allows access to the lumen of ER, the permeabilization does not solubilize the enzyme from the membranes. The digestion in presence of saponin, however, released only 18% of the enzyme activity from the membranes. Under these conditions, i.e. incubation with trypsin a t 20 "C in contrast to incubation a t 0 "C for the experiment in Fig. 7, trypsin concentrations higher than 50 pg/ml caused disruption of the integrity of membrane vesicles (28, 35) and were found unsuitable for examining the release of higher levels of glucosidase I. Western blot of the samples derived from this experiment revealed that the trypsin digestion of intact membrane vesicles resulted in the conversion of the 85-kDa polypeptide into a polypeptide of approximately 82 kDa that still was associated with the membranes. This confirmed that the cleavage had occurred on the cytoplasmically exposed portion of the enzyme. Digestion of saponin-permeabilized membranes with trypsin resulted in membrane-associated fragments of 69 and 55 kDa and a soluble fragment of 39 kDa, suggesting that the proteolysis occurred on sites facing the lumen of ER (Fig. 8).
The release of enzymatically active 39-kDa fragment from E R membranes could be attributed to the removal of a hydrophobic membrane-anchoring domain. This was confirmed by phase separation of proteolyzed enzyme with Triton X-114, a detergent whose solutions separate into a detergent phase and an aqueous phase a t temperatures above 20 "C (36). Upon phase separation, amphipathic integral membrane proteins partition into the detergent phase, while hydrophilic proteins partition into the aqueous phase (37). The native glucosidase I partitioned in both the detergent and aqueous phases, suggesting an amphipathic nature. Similarly, 69-and 55-kDa fragments partitioned into both layers. On the other hand, the 39-kDa fragment was highly enriched in the aqueous layer ( Fig. 9), favoring its luminal orientation. The native enzyme activity was distributed more or less equally in detergent and aqueous layers. However, after trypsin digestion, the activity in the aqueous layer was almost twice as much as in the detergent phase, consistent with the partitioning of the 39-kDa fragment into the aqueous layer.
The data presented here indicate that glucosidase I is a transmembrane, endoplasmic reticulum-localized, N-linked glycoprotein with distinct domains within its structure. DISCUSSION The biosynthesis of glucosidase I occurs in the endoplasmic reticulum. Similar to many other resident proteins of ER, it acquires asparagine-linked, high mannose type oligosaccharide. The results of this study show that the removal of carbohydrate moiety does not seem to affect its catalytic activity significantly. This raises questions regarding the significance of glycosylation of glucosidase I. The requirement for carbohydrate prosthetic groups in the biological activity of glycoproteins has been shown to be variable for different glycoproteins (1). For example, glycosylation of epidermal growth factor receptor has been demonstrated to be essential for its posttranslational activation and ligand-binding capacity (38). On the other hand, the deglycosylation of erythropoietin by glycosidases did not affect its in vitro biological activity in spite of a significant loss of the in vivo activity (39). It is possible that the glycosylation of glucosidase I may be required for its stability, conformation, or retention in the ER.
The major findings of the present study are that glucosidase I is a transmembrane, endoplasmic reticular N-linked glycoprotein; it contains distinct membrane anchoring and luminally oriented catalytic domains; and the enzyme has a cytoplasmically exposed tail.
To resolve definitively the question whether glucosidase I is a transmembrane protein, we examined the trypsin accessibility of different regions of this protein in sealed and saponin-permeabilized membranes. The digestions of sealed membranes generated a polypeptide band of approximately 8 2 kDa from the native enzyme demonstrating a short cytoplasmic tail. The presence of two bands (85 and 82 kDa) in purified enzyme, therefore, could be attributed to the generation of the 82 kDa band during the solubilization and purification of the enzyme by endogenous proteases even though a mixture of antiproteases was included in all buffers employed. The remaining tryptic cleavage sites appear to be located luminally, because the cleavage at those sites occurred only when membrane vesicles were permeabilized with saponin. The release of enzymatically active 39-kDa fragment from the membranes following trypsinization in the presence of saponin suggests that the catalytic domain is attached to the luminal face of the ER by means of a membrane-anchoring domain. This is consistent with the results of phase partitioning with Triton X-114 which revealed an enrichment of 69and 55-kDa fragments in the detergent phase and the 39-kDa fragment in the aqueous phase. It was rather surprising that a removal of more than 50% of the total enzyme mass was required for the change in the phase partitioning behavior. These results raise the possibility that, like glycosyltransferases in the ER (8, lo), glucosidase I also may contain more than one membrane-spanning region. The cDNA sequence of GDP-Man:dolichol-P mannosyltransferase suggests that the enzyme is anchored with both its N and C termini in the membrane, but the catalytic domain of the protein is oriented toward the lumen of ER membranes (10).
The presence of distinct domain structure in glucosidase I is similar to the domain structures of 3-hydroxy-3-methylglutaryl-coenzyme A (34), GDP-Man:dolichol-P mannosyltransferase (9, lo), mannosidase I1 (31), @-1,4-galactosyltransferase (40), @-galactoside a-2,6-sialyltransferase (18), and UDP-GlcNAc:a-3-~-mannoside @-1,2-N-acetylglucosaminyltransferase I (41). All these enzymes contain catalytic domains freely accessible to the substrate, while being anchored to the membrane. This type of membrane organization probably facilitates the regulation of the enzyme activity in vivo (42), as well as retention in the ER (43). When a hybrid gene, consisting of coding sequence for the membrane domain of 3hydroxy-3-methylglutaryl coenzyme A reductase linked to the coding sequence of the soluble enzyme @-galactosidase from Escherichia coli, was expressed in Chinese hamster ovary cells, the fusion protein was localized in ER. Importantly, like 3hydroxy-3-methylglutaryl coenzyme A reductase, the @-galactosidase activity of the fusion protein also exhibited the sterolregulated degradation (44). Thus, the membrane-anchoring domain may be important in determining both the correct intracellular localization and the in vivo regulation of the biological activity.
The membrane topography of glucosidase I has important functional implications. A luminal orientation for the enzyme is consistent with the in vivo availability of the substrate, Glc3Man9GlcNAc2-polypeptide (nascent), synthesized on the luminal side of the ER by the oligosaccharyl transferase complex (45). Further, glucosidase 11, the enzyme catalyzing the removal of the two glucosyl a-1,3-linked residues, also appears to be oriented luminally (14). This is in accord with the finding that the newly translated and glycosylated proteins were sequestered within the lumen of the ER (22).
Our current experiments are designed to investigate the detailed structural analysis of oligosaccharide attached to glucosidase I and the role, if any, it plays in the retention of the enzyme in ER.