Identification of Glu-540 as the catalytic nucleophile of human beta-glucuronidase using electrospray mass spectrometry.

Human beta-glucuronidase is a member of the Family 2 glycosylhydrolases that cleaves beta-D-glucuronic acid residues from the nonreducing termini of glycosaminoglycans. The enzyme is shown to catalyze glycoside bond hydrolysis with net retention of anomeric configuration, presumably via a mechanism involving a covalent glucuronyl-enzyme intermediate. Incubation of human beta-glucuronidase with 2-deoxy-2-fluoro-beta-D-glucuronyl fluoride resulted in time-dependent inactivation of the enzyme through the accumulation of a covalent 2-deoxy-2-fluoro-alpha-D-glucuronyl-enzyme, as observed by electrospray mass spectrometry. Regeneration of the free enzyme by hydrolysis or transglycosylation and removal of excess inactivator demonstrated that the covalent intermediate was kinetically competent. Peptic digestion of the 2-deoxy-2-fluoro-alpha-D-glucuronyl-enzyme intermediate and subsequent analysis by liquid chromatography coupled with electrospray ionization triple quadrupole mass spectrometry indicated the presence of a 2-deoxy-2-fluoro-alpha-D-glucuronyl peptide. Sequence determination of the labeled peptide by tandem mass spectrometry in the daughter ion scan mode permitted the identification of Glu-540 as the catalytic nucleophile within the sequence SEYGAET.

cludes ␤-glucuronidases, ␤-galactosidases, and ␤-mannosidases (6). Enzymes in this family hydrolyze their substrates with net retention of anomeric configuration, presumably via a double displacement mechanism involving the action of two active-site carboxylic acid residues (7). The first step involves nucleophilic attack by one of the carboxylates on the sugar anomeric center. General acid catalysis from the other carboxyl group aids in the departure of the aglycone to form an ␣-Dglycosyl-enzyme intermediate. In the second step, the intermediate is hydrolyzed by general base-catalyzed attack of water at the anomeric center, resulting in cleavage of the glycosidic bond with net retention of the anomeric configuration. Formation and hydrolysis of the glycosyl-enzyme intermediate proceed via transition states with substantial oxocarbenium ion character (Fig. 1).
The three-dimensional structure of HBG is one of two that have been solved for Family 2 ␤-glycosidases, the other being that of Escherichia coli lacZ ␤-galactosidase (8,9). The two enzymes have a similar multidomain structure, each possessing a jelly roll barrel, an immunoglobulin constant domain, and a TIM barrel. The active sites, however, are significantly different since E. coli ␤-galactosidase is a metalloenzyme, binding Mg 2ϩ at the active site. Removal of this metal ion reduces activity severely on most substrates. The role of this metal is unclear, one suggestion being that it acts as an electrophilic catalyst, interacting directly with the departing phenolate (7). Alternatively, it may function by correctly placing the acid/base catalyst. More important, HBG shows no metal ion dependence, and no metal ion is found in the x-ray crystal structure. In this case, it appears that Glu-451 functions as the acid/base catalyst. The equivalent residue in E. coli ␤-galactosidase (Glu-461) was previously suggested to play the role of the acid/base catalyst on the basis of affinity labeling studies and kinetic analyses of mutants (10 -13). Both residues are on the C-terminal side of a conserved asparagine that has been shown in other enzymes of Clan GH-A to hydrogen bond to the sugar 2-hydroxyl (14,28). The catalytic nucleophile of E. coli ␤-galactosidase was identified as Glu-537 by trapping of a glycosylenzyme intermediate and sequencing of the labeled peptide. Based on the above information and amino acid sequence homology, Glu-540 has been suggested to be the catalytic nucleophile of HBG. However, another residue (Asp-207) has been proposed as an alternative candidate based upon its position in the active site (8). Thus, confirmation of the identity of this important residue is necessary.
2-Deoxy-2-fluoro-␤-D-glycosyl fluorides have proven to be very useful reagents for the derivatization of active-site nucleophiles in a number of retaining ␤-glycosidases (15,16). These inactivators are mechanism-based as the C-2 fluorine destabilizes the transition states for both the glycosylation and deglycosylation steps, whereas the anomeric fluorine, a good leaving group, accelerates the first step, permitting the trapping of a 2-deoxy-2-fluoro-␣-D-glycosyl-enzyme intermediate. Proteolytic digestion of the labeled enzyme yields a mixture of peptides, one of which has the fluoro-sugar attached. Isolation of this peptide, followed by tandem mass spectrometric analysis, permits identification of the nucleophile. This study presents evidence that HBG is a retaining glycosidase and describes the synthesis of 2-deoxy-2-fluoro-␤-D-glucuronyl fluoride (2-FGlcUAF) and its use in the identification of the enzyme's catalytic nucleophile.
Synthesis of 2-Deoxy-2-fluoro-␤-D-glucopyranosyl Fluoride (4)-The fluoride (4) was prepared by dissolving 3 g of tetraacetate (2) (8.6 mmol) in 45% HBr/glacial acetic acid (10 ml) and stirring at room temperature for 2 h. The mixture was then diluted with water (200 ml) and extracted with CH 2 Cl 2 (2 ϫ 150 ml). The organic phase was washed with water and saturated sodium bicarbonate, dried over magnesium sulfate, and concentrated in vacuo. The resulting syrup was dissolved in 30 ml of HPLC-grade acetonitrile. To the solution was added silver fluoride (2.17 g, 17.1 mmol), and the suspension was allowed to stir overnight in the dark. The silver salts were then filtered off through a silica gel plug using ethyl acetate as the eluent, and the solvents were removed in vacuo. The syrup was dissolved in anhydrous methanol (30 ml), and then sodium methoxide (46 mg, 0.86 mmol) was added to the solution. After stirring at room temperature for 30 min, the reaction mixture was neutralized with Amberlite IR-120 (H ϩ ) resin and concentrated in vacuo. Silica gel chromatography (27:2:1 ethyl acetate/methanol/water) yielded 1.34 g (85%) of compound 4. The 1 H NMR spectrum was identical to that previously reported (19).
Synthesis of Phenacyl (2-Deoxy-2-fluoro-␤-D-glucopyranosyl Fluoride) Uronate (6)-To a stirred solution of the 2-fluoroglucosyl fluoride (4) (19.5 mg, 0.106 mmol) in water (0.5 ml) was added sodium bromide (3.3 mg, 0.032 mmol) and TEMPO (0.5 mg, 0.0032 mmol). After cooling to 0°C, 5.25% (v/v) bleach (3 ml, 2.1 mmol), already cooled to 0°C, was added slowly, and the reaction mixture was stirred at 0°C for 2 h. It was then acidified with 1 M HCl, and the water was evaporated in vacuo. The residue was suspended in N,N-dimethylformamide (1 ml). To this was added 2-bromoacetophenone (phenacyl bromide; 24.1 mg, 0.121 mmol), followed by triethylamine (16 l, 0.115 mmol). After stirring for 1 h at room temperature, the mixture was diluted with water (10 ml) and extracted with ethyl acetate (4 ϫ 10 ml). The organic phase was dried over magnesium sulfate and concentrated in vacuo. Enzyme Kinetics-Kinetic studies were performed at 37°C. All studies on human ␤-glucuronidase were performed in 100 mM sodium acetate buffer, pH 4.8 (buffer B). A continuous spectrophotometric assay based on the hydrolysis of pNPGlcUA was used to monitor enzyme activity by measurement of the rate of p-nitrophenolate release upon hydrolysis ( ϭ 360 nm, ⑀ ϭ 2.25 ϫ 10 3 M Ϫ1 cm Ϫ1 in buffer B) using a Unicam 8700 UV-visible spectrophotometer equipped with a circulating water bath. Michaelis-Menten parameters for the substrate, previously undetermined with this enzyme, were K m ϭ 1.56 mM and k cat ϭ 250 min Ϫ1 at 37°C in buffer B.

Catalytic Nucleophile of Human ␤-Glucuronidase
Inactivation Studies-The inactivation of human ␤-glucuronidase by 2-FGlcUAF was monitored by incubation of the enzyme (ϳ0.2 mg/ml) under the above conditions in the presence of various concentrations of 2-FGlcUAF at 37°C in a total volume of 100 l. Residual enzyme activity was determined at the appropriate time intervals by addition of a 10-l aliquot of the inactivation mixture to a solution of pNPGlcUA (1 mM, 700 l) in buffer A and measurement of p-nitrophenolate release over a period of 1 min. Pseudo first-order inactivation rate constants at each inactivator concentration (k obs ) were determined by fitting each curve to a first-order equation using the program GraFit (27). The value of k i /K i , assuming inactivation according to the kinetic model shown, was determined from the slope of a plot of k obs against inactivator concentration in the range . Binding of 2-FGlcUAF at the active site was proven by demonstrating protection against inactivation by a competitive inhibitor. Inactivation mixtures (100 l) containing 0.47 mg/ml enzyme and 150 M 2-FGlcUAF were incubated in buffer B at 37°C in the presence and absence of D-saccharic acid 1,4-lactone (0.68 M; K i ϭ 0.23 M). Because the lactone was found to bind to HBG slowly, preincubation of the enzyme with the inhibitor for a period of 1 h was necessary. At several time intervals, residual activity of the inactivation mixtures was determined. For samples containing no protecting ligand, aliquots (5 l) were removed and assayed immediately. In the presence of protecting ligand, aliquots (1 l) were removed, diluted into substrate, and assayed after a delay of 5 min (this delay allowed for the release of any protecting ligand still bound to HBG). The assays were performed as described above.
Turnover (Reactivation) of the 2-Fluoroglucuronyl-enzyme Intermediate-The 2-deoxy-2-fluoro-␣-D-glucuronyl-enzyme was generated by incubating HBG in the presence of 6 mM 2-FGlcUAF in buffer B at 37°C for 10 min. The trapped intermediate (100 l, 13.5 mg/ml) was freed of excess inactivator by concentration of the inactivated enzyme using 30-kDa nominal molecular mass cutoff centrifugal concentrators (Amicon Centricon-30), followed by dilution of the concentrated enzyme stock (100 -200 l) with buffer B to a final volume of 2000 l. 4-Fold repetition of the process ensured a Ͼ99% reduction in the free 2-FGl-cUAF concentration. Aliquots (30 l) of the inactivated enzyme, freed from 2-FGlcUAF, were then incubated at 37°C in buffer B alone, with 50 mM chitobiose, or with 50 mM N-acetyl-D-glucosamine. Reactivation was monitored by removing aliquots (1 l) at several time intervals and assaying for activity as described above. Approximately 50 -75% of the activity was recovered relative to a control of native enzyme treated in an identical manner.
Electrospray Mass Spectrometry-Mass spectra were recorded on a PE-Sciex API 300 triple quadrupole mass spectrometer (Sciex, Thornhill, Ontario, Canada) equipped with an Ionspray ion source. Peptides were separated by reverse-phase HPLC on an Ultrafast Microprotein analyzer (Michrom BioResources Inc., Pleasanton, CA) directly interfaced with the mass spectrometer. In each of the MS experiments, the proteolytic digest was loaded onto a C 18 column (Reliasil, 1 ϫ 150 mm) equilibrated with solvent A (solvent A: 0.05% trifluoroacetic acid and 2% acetonitrile in water). Elution of the peptides was accomplished using a gradient (0 -60%) of solvent B over 60 min, followed by 100% solvent B over 2 min (solvent B: 0.045% trifluoroacetic acid and 80% acetonitrile in water). Solvents were pumped at a constant flow rate of 50 l/min. Spectra were obtained in the single quadrupole scan mode (LC/MS) or the tandem MS product ion scan mode.
In the single quadrupole mode (LC/MS), the quadrupole mass analyzer was scanned over a m/z range of 400 -1800 Da with a step size of 0.5 Da and a dwell time of 1.5 ms/step. The ion source voltage was set at 5.5 kV, and the orifice energy was 45 V. In the tandem MS daughter ion scan mode, the spectrum was obtained by selectively introducing the parent ion (m/z ϭ 935) from the first quadrupole (Q1) into the collision cell (Q2) and observing the product ions in the third quadrupole (Q3). Thus, Q1 was locked on m/z ϭ 935; the Q3 scan range was 50 -1120 Da; the step size was 0.5 Da; the dwell time was 1 ms; the ion source voltage was 5 kV; the orifice energy was 45 V; Q0 ϭ Ϫ10 V; and IQ2 ϭ Ϫ48 V.

RESULTS AND DISCUSSION
Stereochemistry of HBG Hydrolysis-The use of proton NMR in the determination of the stereochemical course of enzymecatalyzed glycoside hydrolyses has been demonstrated previously (20,21). Chemical shifts and coupling constants of the anomeric protons of ␣and ␤-glycosides and the product hemiacetals are distinct and readily observed. When sufficient enzyme is used to complete the hydrolysis quickly (typically Ͻ2 min), the initially formed anomer is detected before mutarotation has occurred to any significant extent. Fig. 2 illustrates the experiment performed using HBG. Fig. 2A shows the anomeric proton region of pNPGlcUA in buffer A. The multiplet centered at ␦ 5.18 -5.28 ppm arises from the axial anomeric proton of the ␤-glycoside substrate, and the large resonance at ␦ 4.7 ppm is from HOD. Panels B-D were recorded at time intervals after addition of the enzyme. As shown in Fig. 2B, the enzymatic hydrolysis was essentially complete after 1.5 min since the resonance at ␦ 5.18 -5.28 ppm had almost disappeared. Simultaneously, a new resonance appeared at ␦ 4.58 ppm (J ϭ 7.9 Hz). The chemical shift and coupling constant identify this as the axial anomeric proton of ␤-D-glucuronic acid. After 4 min, a new doublet appeared at ␦ 5.18 ppm (J ϭ 3.6 Hz) (Fig. 2C). This corresponds to the equatorial anomeric proton of ␣-D-glucuronic acid resulting from mutarotation of the initial ␤-D-glucuronic acid product. After 20 min, the product D-glucuronic acid converted via mutarotation to the equilibrium ratio of anomers (34:66 ␣/␤) (Fig. 2D). The data unequivocally demonstrate that hydrolysis catalyzed by HBG proceeds with net retention of anomeric configuration, presumably via a double displacement mechanism.
Synthesis of Inactivator-The synthesis of 2-deoxy-2-fluoro-␤-D-glucuronyl fluoride is outlined in Fig. 3. The key steps involved the use of Selectfluor TM to introduce fluorine at C-2 and the use of TEMPO in the selective oxidation of the C-6 alcohol to the carboxylic acid. Selectfluor TM is an electrophilic fluorinating agent that is both stable and easy to use (22). It reacts with olefins in the presence of a weak nucleophile such as water or acetic acid to yield the addition product, and it has previously been shown to react with glycals to yield a variety of 2-deoxy-2-fluoroglycosides (23). In particular, tri-O-acetyl-Dglucal was shown to react with Selectfluor TM in N,N-dimethylformamide/water to give predominantly the manno-epimer. We found that by changing the solvent/nucleophile system to acetic anhydride/acetic acid the gluco-epimer was obtained in satisfactory quantities, albeit at lower yield relative to the mannoepimer (2:3 gluco/manno; 23 and 32%, respectively). Fortunately, separation of the gluco-and manno-products is readily achieved by silica gel chromatography, and the gluco-epimer was converted to 2-deoxy-2-fluoro-␤-D-glucopyranosyl fluoride by a series of steps involving bromination, fluoride displacement, and deprotection.
Glucopyranosyluronic acid was obtained via TEMPO-mediated oxidation of the C-6 alcohol of 2-deoxy-2-fluoro-␤-D-glucopyranosyl fluoride in water buffered at pH 10 with hypobromite (formed by the reaction of hypochlorite and bromide) as the regenerating oxidant (24 -26). This reaction is both mild and selective for the primary C-6 alcohol. To facilitate purification of the product away from the inorganic salts present, the uronic acid was first converted to its phenacyl ester and then purified chromatographically. The choice of the phenacyl ester was based upon the need for a protecting group that can be easily removed under mild conditions, i.e. non-acidic and non-basic conditions. This particular ester is easily cleaved via catalytic hydrogenolysis.
Inactivation Kinetics-Incubation of HBG with 2-FGlcUAF resulted in inactivation of the enzyme in a time-dependent man-ner according to pseudo first-order kinetics (Fig. 4A). However, no saturation was observed, even at the highest inactivator concentrations studied (Fig. 4B); yet higher concentrations could not be investigated due to the rapidity of inactivation, which precluded accurate sampling. Reliable values for the inactivation rate constant (k i ) or the reversible dissociation constant (K i ) therefore could not be determined. However, a reliable secondorder rate constant of k i /K i ϭ 286 min Ϫ1 M Ϫ1 was calculated for the slope of the plot of k obs versus .
Incubation of the enzyme with 2-FGlcUAF (150 M) in the presence of the competitive inhibitor D-saccharic acid 1,4-lactone (1) (0.68 M) resulted in a two-phase reaction (Fig. 4C). The initial inactivation rate constant was calculated to be 0.054 min Ϫ1 , nearly identical to the apparent inactivation rate constant of 0.057 min Ϫ1 obtained in the absence of the lactone. The final inactivation rate constant was calculated to be 0.019 min Ϫ1 . This observed protection from inactivation in the pres-  ence of the lactone is consistent with 2-FGlcUAF being active site-directed. These results suggest that inactivation is a consequence of accumulation of a stable covalent 2-deoxy-2-fluoro-␣-D-glucuronyl-enzyme intermediate, a conclusion that is supported by the mass spectral analysis of the inactivated enzyme. Catalytic Competence-Further evidence supporting the existence of a covalent 2-deoxy-2-fluoro-␣-D-glucuronyl-enzyme arises from demonstration of the catalytic competence of the trapped intermediate. Following removal of excess inactivator from the labeled enzyme, the sample was incubated at 37°C in the presence of buffer B alone, with 50 mM chitobiose, or with 50 mM N-acetylglucosamine, and the recovery of activity associated with the regeneration of the free enzyme was monitored. Reactivation kinetics of the 2-deoxy-2-fluoro-␣-D-glucuronylenzyme in buffer alone followed a first-order process with an apparent rate constant of k re ϭ 0.0040 h Ϫ1 (t1 ⁄2 ϭ 250 h) (Fig. 5). Rate constants for reactivation by transglycosylation (k trans ) were found to be 0.0033 h Ϫ1 (t1 ⁄2 ϭ 303 h) with N-acetylglucosamine and 0.0072 h Ϫ1 (t1 ⁄2 ϭ 139 h) with chitobiose. The higher enzyme reactivation rate observed in the presence of chitobiose (2-fold higher than the spontaneous reactivation rate) suggests that reactivation is accelerated by transglycosylation to an acceptor sugar. That chitobiose functions as a transglycosylation acceptor is not surprising since the natural substrates of HBG are the glycosaminoglycans, oligosaccharides composed of alternating glucuronic acid and N-acetylglycosamine residues, and chitobiose is a GlcNAc(␤1-4)GlcNAc disaccharide.
Identification of the Labeled Active-site Peptide by Electrospray MS-Peptic hydrolysis of native HBG or the 2-deoxy-2fluoro-␣-D-glucuronyl-enzyme resulted in a mixture of peptides, which were separated by reverse-phase HPLC using the electrospray MS as detector. When scanned in the normal LC/MS mode, the total ion chromatograms showed a large number of peaks, each corresponding to one or more peptides in the digest mixture (Fig. 6, A and C). The masses under each peak in the labeled sample were compared with the masses of the corresponding peptides in the native sample, searching for a peptide present only in the labeled sample that was 178 Da greater than a peptide present only in the unlabeled sample. Only one pair of peaks satisfied this requirement of difference in mass by that of the attached label, these being a peak at m/z 756 in the native digest that was not observed in the labeled digest (Fig.  6B) and a peak at m/z 934 in the labeled digest that was not observed in that of the native enzyme (Fig. 6D). The mass difference between these two peaks is 178 Da, which corresponds exactly to the mass of the 2-deoxy-2-fluoro-␣-D-glucuronyl label. The labeled parent ion (m/z 934) and the unlabeled intact peptide (m/z 756) thus appear as singly charged species.
Candidate peptides with a mass of 756 Ϯ 2 Da were then identified by inspection of the amino acid sequence of the enzyme and searching for all possible peptides with this mass. Twenty such peptides were identified, but of these, all but five were eliminated because their sequences did not contain either an aspartate or a glutamate residue. Precedent with all retaining glycosidases to date would predict that the nucleophile should be one of these two amino acids. The candidate peptides are 257 KLEVRL 262 , 60 EEQWY 64 , 318 DFYTLP 323 , 484 NSNYAAD 490 , and 539 SEYGAET 545 . The peptide was then unambiguously identified by peptide sequencing using tandem MS.
Peptide Sequencing-Information on the sequence was obtained by additional fragmentation of the peptide of interest (m/z 934) in the daughter ion scan mode (Fig. 7). The parent ion of interest (m/z 934) was selected in the first quadrupole and subjected to collision-induced fragmentation, and then the masses of the daughter ions were detected in the third quadrupole. Peaks resulting from YЈЈ ions correspond to fragments ET (m/z 249), AET (m/z 320), GAET (m/z 378), YGAET (m/z 540), and EYGAET (m/z 669). Peaks arising from B ions bearing the label include SE (m/z 395), SEY (m/z 558), SEYG (m/z 616), SEYGA (m/z 687), and SEYGAE (m/z 816). Because the B ions bearing the label include SE (m/z 395), we can infer that the label is linked to either Ser-539 or Glu-540. This information, in conjunction with the mass of the labeled peptide and the primary sequence of the enzyme, permits identification of the peptide containing the active-site nucleophile as 539 SEYG-AET 545 . Comparison of this sequence with those of the other Family 2 glycosidase sequences shows that the dipeptide EY is conserved throughout the known family members (Table I). However, the amino acids around this dipeptide are conserved only within a particular glycosidase category, i.e. galactosidase versus glucuronidase versus mannosidase. From this, we can assign Glu-540 as the catalytic nucleophile.
Doubts have been expressed previously concerning the identity of the catalytic nucleophile of HBG. In particular, on the basis of the three-dimensional crystal structure, it had been suggested that Asp-207 might well serve in this role (8). Furthermore, the fact that E. coli ␤-galactosidase is a metalloenzyme, requiring a Mg 2ϩ bound at the active site for catalytic activity, whereas HBG has no such requirement, causes concerns about parallels drawn between the two enzymes' active sites. These are further heightened by the fact that HBG cleaves a substrate containing a carboxylic acid, thus a different active-site composition may be required to accommodate this additional charge. However, the assignment of Glu-540 as the catalytic nucleophile of HBG removes any ambiguity concerning the identity of this residue and is completely consistent with expectations on the basis of sequence similarity (6).