Human acid beta-glucosidase. N-glycosylation site occupancy and the effect of glycosylation on enzymatic activity.

The five potential N-glycosylation sites (sequons) of human acid beta-glucosidase were individually mutated to determine site occupancy and the effect of site occupancy on selected catalytic and stability properties of this enzyme. Each N-glycosylation consensus sequence [Asn-Xaa-(Ser/Thr)] was obliterated by individually substituting glutamine (Q) for asparagine (N). By expression of the normal and mutated cDNAs in insect (Sf9) and COS-1 cells and subsequent immunoblotting with anti-human acid beta-glucosidase antibodies, the four sequons at Asn-19, Asn-59, Asn-146, and Asn-270 were shown to be glycosylated in either source. The sequon at Asn-462 was never occupied. The mutant enzymes N59Q, N146Q, and N270Q were catalytically active and had normal interactions with active site-directed inhibitors as well as with the activators, phosphatidylserine and saposin C. Of the occupied sequons, N-glycosylation of the first was critical to the synthesis of a catalytically active enzyme. Alteration of this sequon, Asn-19-Ala-20-Thr-21, by the substitutions N19Q, N19D, N19E, or T21G led to a lack of glycosylation at this site. Enzymes containing N19Q, N19E, or T21G had significant decreases (3- to 60-fold) in intrinsic enzyme activity. The N19D enzyme had nearly normal catalytic activity and had enhanced activation by phosphatidylserine. These results show that sequon occupancy as well as steric effects at residue 19 are important for the development of an active conformer of this enzyme. This is the first example of a lysosomal hydrolase that requires sequon occupancy for the synthesis of a catalytically active enzyme.

zyme requires the use of detergents, negatively charged lipids, and an activator protein, saposin C, for optimal hydrolysis of the natural or synthetic substrates (for review, see Ref. 3). Detailed kinetic studies of this enzyme indicate a complex enzyme with multiple subsites for interaction with phospholipids and saposin C. The active site also has subsites that accommodate the glycon, sphingosyl, and fatty acid acyl moieties of substrates and inhibitors (for review, see Ref. 3). Defects in the catalytic function, stability, and/or post-translational processing of acid p-glucosidase lead to the variants of Gaucher disease, the most prevalent lysosomal storage disease (4-9).
As is typical of lysosomal hydrolases (lo), the normal enzyme is synthesized as a precursor containing a 19-amino acid secretory signal sequence that is proteolytically clipped during transit through the membrane of the endoplasmic reticulum (11). No additional proteolytic processing occurs during transport through the Golgi apparatus to the lysosome (11,12). The mature, unglycosylated polypeptide has a calculated molecular weight of -56,000. Co-translational glycosylation occurs at four of the five (11) predicted N-glycosylation sites, Asn-Xaa-(Ser/Thr) (13), but site occupancy has not been determined. Complete deglycosylation with N - GlycanaseTM (14) or endoglycosidase F (11) indicated the presence of N-glycosylation. Complete characterization of the oligosaccharide moieties from the human placental enzyme demonstrated N-linked high mannosyl type and typical biand triantennary complex structures (15). Modification of the oligosaccharide composition has no effect on the stability or catalytic properties of the enzyme (16). However, prevention of glycosylation by expression of the human cDNA in bacterial systems or in tunicamycin-treated insect cells led to the synthesis of catalytically inactive enzyme forms (14). These results suggested that glycosylation at one or more sites may play a role in the formation of an active enzyme species (14).
To gain further insight into glycosylation site occupancy and the role of glycosylation in the development of active enzyme forms, site-directed mutagenesis was used to individually destroy each potential glycosylation consensus sequence or sequon. Expression of these cDNAs in insect and/or mammalian cells showed the oligosaccharide site occupancy and that occupancy of the first site is important for development of a catalytically active enzyme.  nitrocellulose filters (Schleicher & Schuell); @-cyanoethyl phosphoramidites, CPG synthesis columns, and reagent kits for model 380B DNA synthesizer (Applied Biosystems); 4-chloro-4-naphthol and sodium dodecyl sulfate (Bio-Rad); phosphatidylserine (Avanti, Alabaster, AL); Tricine (Sigma); BaculoGoldTM (PharMingen, San Diego, CA). All buffers for cell sonication contained 2 PM phenylmethylsulfonyl fluoride.
(20) was from Dr. Randall Kaufman (Genetic Institute, Cambridge, Site-directed Mutagenesis and Plasmid Construction-The cDNA encoding normal acid @-glucosidase cloned in M13mp19 was subjected to mutagenesis by oligonucleotide-directed site-specific reactions using the phosphorothioate DNA selection procedure (21, 22). Five individual cDNAs were prepared so that each was missing one Nglycosylation consensus sequence, Asn-Xaa-Ser/Thr, by creating codons for Gln instead of Asn. Additional cDNAs encoded the substitution of Asn to Asp or Asn to Glu at the sequon located at amino acid 19 (N19D, N19E), of Thr to Gly a t amino acid 21 (T21G), or of Asn to Asp a t amino acid 462 (N462D). The oligonucleotides for mutagenesis are shown in Table 1. One cDNA was prepared with all five of the sequons destroyed by substituting Asn with Gln. For this cDNA, one reaction mutated the first and the second glycosylation sites on the normal acid @-glucosidase cDNA, and an EcoRI-NcoI fragment, termed A, was isolated. The second reaction used the cDNA encoding N462Q, i.e. the fifth sequon, as a single-stranded template for mutagenesis of the sequons beginning at residues 146 and 270. An NcoI-Sac1 fragment, termed B, containing these sequons was isolated. Subsequently, A and B were ligated to form a full-length multiple mutant (-2 kilobases) EcoRI-Sac1 cDNA with all N-glycosylation sites mutated. The construct that had an intact N-glycosylation site only a t Asn-146 (G-3) was prepared by ligating the 446-base pair NcoI-ScaI fragment from the cDNA encoding N270Q into the 8804base pair ScaI-NcoI fragment of the multiple mutant. The mutant cDNAs were cloned and plaque-purified, and the coding regions of all 10 mutagenized cDNAs were sequenced in their entirety (23). All cDNAs were shown to be identical with the normal cDNA except at the specifically mutagenized amino acids. The mutant cDNAs were purified from ethidium bromide-stained 0.8% agarose gels following digestion with EcoRI and SacI restriction endonucleases and directionally cloned into the EcoRI and SacI sites of the plasmid pAc610. The EcoRI-Sac1 fragment of the acid-@-glucosidase cDNA contains the entire coding region and an additional 93 nucleotides 5' and 304 nucleotides 3' of the untranslated sequence. After ligation, transformation of Escherichia coli HB101, and selection of ampicillin-resistant colonies, the purified plasmids were obtained by standard methods for large scale preparation followed by cesium chloride gradient purification (24).
Construction and Purification of Recombinant Baculouirus-Recombinant baculovirus containing the normal or mutant cDNAs were produced by homologous recombination (5). Viral DNA (Baculo-GoldTM) was used for preparation of recombinant viruses which contained acid B-glucosidase cDNAs encoding the N19E, T21G, N462D, and G-3 enzymes. Infection of Sf9 cells with wild type or recombinant A . californica nuclear polyhedrosis virus, determination of viral titers, and calcium phosphate-mediated transfections were done as described (5). Recombinant viruses containing the acid @glucosidase cDNAs were purified by plaque hybridization (5,25). The levels of acid 8-glucosidase activity and protein expressed from independent recombinant viral isolates for each cDNA were monitored by enzyme assay and immunoblots, 3 days after infection. For these studies, titers were adjusted to achieve maximal specific activities (5).
Construction of Recombinant p91023(B) and COS-1 Tramfection-COS-1 cells were transfected by electroporation with p91023(B) plasmid containing a normal or a mutagenized cDNA. For cloning into the unique EcoRI site of p91023(B), EcoRI sites were introduced into the above cDNAs via the polymerase chain reaction using the following 5' and 3' primers containing EcoRI sites: 5"GGGAATT-CAGGGTAAGCATCATGGCTGGC-3' and 5'WGGAATTCT-CACTGGCAGAGCCACAGGTA-3'. PCR conditions were as follows: denaturation (1 min, 94 "C), annealing (30 s, 66 "C), extension (2 min, 72 "C). The resultant polymerase chain reaction product contained the coding sequence from the most 3' in-frame initiation ATG of the cDNA to the stop codon. Ligation of the 1.6-kilobase EcoRI fragment into the p91023(B) EcoRI site, transformation, and tetracycline selection were as described (20). Selection for p91023(B) with sense-oriented cDNA inserts was done by restriction fragment patterns on ethidium bromide-stained agarose gels. These plasmids were purified by cesium chloride gradients prior to use (24).
Zmrnunoelectroblotting-Immunoelectroblotting was conducted (17) using SDS or Tricine (26) PAGE. Briefly, Sf9 cells, infected with wild type (A. calijornica nuclear polyhedrosis virus) or pure recombinant virus as well as COS-1 cells mock-transfected or transfected with p91023(B) containing acid 8-glucosidase inserts, were harvested into 0.9% NaCl by vigorous striking of the flasks or by trypsin treatment, respectively. Cells were washed three times by centrifugation (525 X g, 10 min) and resuspension in 0.9% NaCl. Pellets were stored at -20 "C until use. Acid 8-glucosidase was solubilized at 4 "C by ultrasonic irradiation of the washed pellets in 0.1% Triton X-100 and 0.1% sodium taurocholate using a cup sonicator (Branson Cell Disruptor 200, 80 watts, pulse times of 30 s, 20 s, and 20 s) (8). The sonicates were clarified by centrifugation (875 X g, 20 min), and the resulting supernatants were used in the immunoelectroblotting studies (17). Immunoblots were conducted using the anti-human acid @glucosidase IgG monoclonal antibodies, MC-1 and MC-38, which recognize different epitopes on the enzyme or polyclonal rabbit antihuman acid 8-glucosidase antibodies (17). Extracts of Sj9 cells infected with wild-type virus or mock-transfected COS-1 cells gave no signals on immunoblots using the above antibodies.
N-Linked deglycosylation of normal (NL) and mutated human acid @-glucosidase expressed in Sj9 cells was achieved by treatment with N-Glycosidase F or N-GlycanaseTM under denaturing conditions, according to the manufacturer's instructions. With N-Glycosidase F, octyl 8-glucoside was used as the nonionic detergent to optimize electrophoretic separations on Tricine PAGE.
Enzyme Assays-Acid 8-glucosidase hydrolytic rates were determined fluorometrically using NBD-GC or 4MU-Glc substrates (27). Unless specified, the final reaction mixtures (0.2 ml) contained 0.04 M phosphate/citrate, pH 5.5, 4 mM 8-mercaptoethanol, and 1.25 mM EDTA as well as 0.25% Triton X-100 and 0.25% mM sodium taurocholate (Buffer A). Assays were done using crude sonicates prepared as described above for the immunoblotting studies. For the inhibition studies using DNM or Br-CBE, the inhibitor was added as a concentrated solution in water to the assay solution. The enzyme source was added last. ICw values, the concentration of inhibitor which results in a 50% decrease of initial enzyme activity, were determined as described (5,29). The amount of normal or mutant enzyme source, COS-1 or Sf9 cell sonicates, was adjusted to ensure that <5% of the substrate was hydrolyzed during the reaction. Reactions were terminated after 0.5-2 h at 37 "C (29).
Activation by phosphatidylserine was determined as described (8,27). Briefly, aliquots of crude sonicates in 0.06% Triton X-100 were incubated at room temperature for 15 min in 0.04 M citrate/phosphate buffer, pH 5.5, containing 0.06% Triton X-100 and varying concentrations of phosphatidylserine. Activity to 4 mM 4MU-Glc was assayed as described above with final phosphatidylserine concentrations ranging from 0 to 3 mM. The final concentration of Triton X-100 in the assay mixtures was 0.06%. Solutions containing phosphatidylserine and Triton X-100 were prepared from concentrated stock solutions in ch1oroform:methanol (2:l). The appropriate aliquots of each stock solution (5 mg/ml phosphatidylserine; Triton X-100) were transferred to a glass tube, dried under Nl, and then under vacuum for 2 h. The dried compounds were then dispersed into citrate/ phosphate buffer using a cup sonicator. Response to the protein activator, saposin C, was determined by assaying activity to 4MU-Glc in 0.2 M sodium acetate, pH 4.7, containing 20 pg/ml (0.025 mM) phosphatidylserine. Saposin C was prepared from the spleen of a Gaucher disease patient (18,28).
For heat inactivation studies, the above clarified supernatants from Sf9 cells were diluted 10-fold in Buffer A and incubated for 0 to 30 min at 50 "C. At each time point, the tubes were removed and placed immediately on ice. Activity toward the 4MU-Glc substrate was determined as above. The times of incubation at 50 "C that resulted in 50% loss of the initial activity (ts) were determined (5).

RESULTS
T o determine the N-glycosylation sequon occupancy for acid 8-glucosidase, the normal and mutant cDNAs were individually expressed in insect (Sf9) and COS-1 cells. The cells. The N19Q, N59Q, N146Q, N270Q, and N19D enzymes migrated faster than the enzyme expressed from the normal cDNA (NL). Similarly, the N19E and T21G enzymes migrated faster than the normal enzyme (Fig. 1B). These results indicated that the glycosylation consensus sequences at the first four positions are normally occupied in Sf9 cells. Essentially identical results were obtained in COS-1 cells (Fig. IC).
In either cell type, N462Q (Fig. IC) or N462D (Fig. 1B) migrated normally which indicated that this glycosylation site is not normally occupied. N146Q consistently had slower migration than the mutants N19Q, N59Q, N270Q, and N19D, but this altered migration was only slightly different from normal. To conclusively demonstrate occupancy a t Asn-146, the cDNAs encoding N146Q or with only the third glycosylation site (G-3) intact were expressed in Sf9 cells and deglycosylated with N-Glycosidase F. Faster migration of N146Q before deglycosylation again was obtained on Tricine PAGE. Treatment of the G-3 enzyme with N-Glycosidase F produced a shift in migration of the acid P-glucosidase so that the migration was identical with that for the mutant enzyme with all glycosylation sites obliterated (Fig. 2). The deglycosylated G-3 also migrated with the fastest band obtained by N -Glycosidase F treatment of the normal enzyme expressed in Sf9 cells (Fig. 2). This result shows that glycosylation occurs at this site. The cDNA containing mutations a t all glycosylation consensus sequences (multiple mutant) migrated identically with the normal enzyme which had been completely deglycosylated with N-GlycanaseTM (data not shown) or N -Glycosidase F (Fig. 2) or the unglycosylated human protein expressed in E. coli (data not shown).
The specific activities of the expressed normal and mutant enzymes were determined in Sf9 cells and COS-1 cells. Of the glutamine-substituted enzymes, detectable activities were obtained in Sf9 cells with N59Q, N146Q, and N270Q (Table 11). In contrast, the activities with the N19Q, N462Q, and N462D mutant enzymes were very low even when large amounts of immunologically reactive protein were expressed. These results demonstrate a very decreased catalytic rate constant for these mutant enzymes. The effects of different amino acid   ' Activation-fold is referenced to the values for phosphatidylserine (5-to 8-fold) or saposin C (4-to 6-fold) obtained for the normal human enzyme expressed in Sf9 cells. The 4MU-Glc substrate only was used for activation studies.
'Due to low activities, only the glucosylceramide substrate was used. Assays were conducted as described under "Experimental Procedures." ND, not determined.
substitutions within the first glycosylation site were evaluated by creating and expressing enzymes containing N19D, N19E, or T21G. In comparison to N19Q, these enzymes were catalytically active but had variably decreased specific activities based on the amount of cross-reacting immunologic material (CRIM) to anti-human acid P-glucosidase antibodies (5). These estimates indicated that the CRIM specific activities for the N19Q, N19E, and T21G enzymes were >60, -30, and -3 -fold decreased, respectively, compared to the normal human enzyme expressed in Sf9. The corresponding values for the N19D enzyme were nearly normal but depended on the assay conditions (see below). Essentially identical results were obtained in COS-1 cells (data not shown).
T o determine if any of the enzymes were secreted, media from either Sf9 or COS-1 cell cultures expressing the normal or active mutant enzymes were assayed for enzyme activity and the ratios were determined for the enzyme activity in the media compared with the total enzyme activity within cells and in the media. For cells infected or transfected with the normal cDNA and for Sf9 cells, under low lysis conditions, little enzyme was secreted into the media (<25% of total enzyme). This ratio was the same for all mutant enzymes tested. Similarly, little enzyme could be detected by immunoblotting of concentrated media from the respective COS-1 cells.
T o evaluate additional effects of glycosylation site occupancy on active site function or stability, inhibition (DNM and Br-CBE) and activation (phosphatidylserine and saposin C) studies were conducted (Table 111). The potent active sitedirected inhibitors (29), DNM and Br-CBE, had normal or nearly normal ICso values with all tested mutant enzymes. The exceeding low activity of the N462Q or N462D enzymes is due to its location near catalytically important regions of the enzyme.' Indeed, a mutation, R463C, has been identified as causing Gaucher disease (31,32). The interactions of acid @-glucosidase with its activators, phosphatidylserine and saposin c , were tested with the normal and selected active mutant enzymes. The hydrolysis of 4MU-Glc by mutant enzymes, except N19D, was enhanced normally, or nearly so, by these compounds. The N19D enzyme was hyper-responsive to phosphatidylserine. About 2.6-fold greater degrees of activation were achieved with this enzyme than with the normal M. E. Grace or other mutants. Heat inactivation studies demonstrate that only the destruction of the third glycosylation site (N146Q) had a significant effect (tH = 9.5 min uersus 15.5 for normal) on decreasing the thermostability of the enzyme.

DISCUSSION
The present studies show that the first four N-glycosylation consensus sequences, sequons, of human acid 0-glucosidase are normally occupied and that the catalytic activity of this enzyme depends on the occupancy and/or the nature of amino acid 19. The expression of the normal cDNA encoding acid 0-glucosidase in bacteria or in insect cells exposed to tunicamycin resulted in the synthesis of stable but catalytically inactive enzymes (14). In addition, deglycosylation of purified acid 0-glucosidase in the native state suggested that a single oligosaccharide moiety was easily removed and that its presence was needed to maintain the activity of the enzyme (14). In comparison, the trimming of glycosidic residues to the mannosyl core had little, if any, influence on catalytic activity (33). In fibroblasts, the type or degree of glycosylation of acid 0-glucosidase did not appear to influence either the stability or catalytic activity when inhibitors of glycosidic trimming enzymes were used to alter the type of glycosylation (16). However, site occupancy was not evaluated in these studies. The studies presented here show that the catalytic activity of a lysosomal hydrolase is dependent upon specific glycosylation site occupancy.
The activity of only a few other enzymes is modulated by glycosylation site occupancy, and this was due to either the presence or absence of oligosaccharide moieties at a particular consensus sequence (34)(35)(36). The influence of sequon occupancy of tissue plasminogen activator has been extensively and thoroughly analyzed in natural cells (34,37) and following heterologous expression (35). Tissue plasminogen activator has three glycosylation sequons (37). Occupancy of all three, but especially that at Asn-184, results in diminution of catalytic activity (34). The nature of the glycosylation at Asn-117 and Asn-448 also influenced the kinetic properties of this enzyme (34,35). Complete lack of glycosylation at all three sites leads to a higher activity enzyme which was more sensitive to enhancement of activity by a fibrinogen fragment (34).
In comparison to the normal enzyme, the acid /3-glucosidase mutant, N19D, which lacked glycosylation at the first sequon, had greater enhancement of activity by phosphatidylserine. Since N19Q was nearly inactive and N19E and T21G had decreased CRIM specific activities, these results indicate that the nature of the amino acids within the first sequon as well as occupancy of this sequon influence the formation of a catalytically active species. Steric influences were more important than functional effects of substituted amino acids for the synthesis of a fully active enzyme. Isofunctional substitutions at amino acid 19 led to major decreases in catalytic activity; i.e. the ratio of CRIM specific activities for the enzymes N19N (normal) and N19Q was -60, and the ratio for the enzymes N19D and N19E was -30. In comparison, isosteric substitutions had smaller effects on enzyme activity; i.e. the ratio of CRIM specific activities for the enzymes N19N and N19D was -1 and for the enzymes N19E and N19Q was "2 to 3. However, the normal thermostability and the enhanced effects of phosphatidylserine with N19D suggested that the conformation assumed by this mutant, while not grossly distorted, is not that of the normal enzyme. Computer calculations (38) showed that the presence of an aspartate or a glutamate at residue 19 significantly altered the predicted secondary structure in this region. In comparison, the pres-ence of a glutamine, substituted for an asparagine, did not (data not shown). Thus, the nature of the amino acid and, potentially, the degree or type of glycosylation at amino acid 19 may influence the maximal catalytic activity of the enzyme. By comparison, the lesser effect of the T21G substitution on catalytic activity suggests additional positional effects. Although not extensively investigated, tissue-specific (spleen uersus brain) differences in sensitivity of acid 0-glucosidase to exogenous negatively charged lipids (39) might be based in alternative glycosylation site compositions. Present results suggest that differences in the degree of glycosylation at a single site were present only at the sequon Asn-146 which may have smaller oligosaccharide moieties (Figs. 1 and 2). Prevention of glycosylation at sequon Asn-146 produced an enzyme with increased thermolability. These finding suggest that selection of the specific cDNA and/or cell type for expression of recombinant acid 0-glucosidase with altered glycosylation may have significant therapeutic implications (34,40).
By substitution of Ala or Gln for Asn-43, human lipoprotein lipase also was shown to require glycosylation of a single sequon for catalytic activity (36). The more conservative substitution of an Asp for Asn (41) was not investigated. In addition, glycosylation at residue 43 was needed for secretion of the lipoprotein lipase into the culture media (36). With acid 0-glucosidase, very little enzyme was secreted normally in either COS-1 or Sf9 cells, and none of the singly mutated enzymes was found in excess in the media. The enzyme protein with all glycosylation sites destroyed was stable within these cells and was not detected in the media. These results suggest that the signals for retention of acid &glucosidase in COS-1 and Sj9 cells were contained in the polypeptide sequence.
The mechanism by which the presence of an oligosaccharide moiety or the nature of the amino acid at the first sequon (residue 19) confers a catalytic conformer is unknown. However, it is interesting that all four occupied glycosylation sites are present in the amino-terminal 50% of the enzyme. In addition, the carboxyl-terminal 50% of the acid 0-glucosidase contains most of the hydrophobic amino acids and, probably, major components of the active site (30). Thus, sequon occupancy or the specific amino acid at residue 19 must affect catalytic activity by altering global conformation during the emergence of the nascent enzyme through the endoplasmic reticular membrane. It seems likely, that, under normal circumstances, glycosylation of Asn-19 provides a nidus for the vectorial folding of this enzyme during its synthesis.