Photoaffinity Labeling of Human Lysosomal β-Hexosaminidase B

The carbene precursor 3-azi-1-[([6-3H]-2-acetamido-2-deoxy-1-β-D-galactopyranosyl)thio]-butane (also designated [3H]-1-ATB-GalNAc) has been used as a photoaffinity label for human lysosomal β-hexosaminidase B (Hex B, EC 3.2.1.52) purified to apparent homogeneity from postmortal liver. [3H]-1-ATB-GalNAc behaved as an active site-directed inhibitor, which bound covalently to Hex B upon photolysis at 350 nm and resulted in 15% inactivation of enzyme activity. Up to 75% of the inactivation of Hex B was prevented by including the competitive inhibitor 2-acetamido-2-deoxy-D-glucono-1,5-lactone in the photoaffinity experiment. Incubation of [3H]-1-ATB-GalNAc with the enzyme followed by irradiation and subsequent separation of the three polypeptides composing the β-subunit led mainly to labeling of the βa-polypeptide. Subsequent proteolysis of βa with trypsin and separation of the resulting peptides by high pressure liquid chromatography yielded one prominently labeled peptide fraction. Edman degradation resulted in the sequence E339ISEVFPDQFIHLGGDEVEFK359. However, no modified amino acid was detected, indicating that the photoaffinity label was presumably bound to the peptide by a labile ester linkage. This was proven when the radiolabel was almost completely released from the peptide by treatment with aqueous ammonium hydroxide. Simultaneously, Glu-355 was converted into Gln-355, which is located within a region of Hex B that shows considerable homology with the α-subunit of human hexosaminidase A and other hexosaminidases from various species.

The major isoenzymes ␤-hexosaminidase A and ␤-hexosaminidase B (Hex B) 2 are composed of two noncovalently linked subunits of nearly equal molecular mass. ␤-Hexosaminidase A is a heterodimer (␣␤), whereas Hex B is a homodimer of two ␤-chains (␤␤) (3). The homodimer ␣␣, also called ␤-hexosaminidase S, has been detected in patients with Sandhoff's disease and seems to be unstable under normal conditions (4,5). The mature ␣-chain is composed of a major ␣-polypeptide (M r ϭ ϳ54,000) and a minor polypeptide ␣ p (M r ϭ ϳ6,000). The mature ␤-subunit originates from its precursor by two proteolytic cleavage events forming a small ␤ p -polypeptide (M r ϭ ϳ11,000), the ␤ b -chain (M r ϭ ϳ24,000), and the ␤ a -chain (M r ϭ ϳ28,000). In the mature ␤-subunit all 3 chain components are linked by disulfide bonds (6 -10). Dimerization of the subunits ␣ and ␤ or ␤ and ␤ is essential for catalytic activity of human ␤-hexosaminidase A and B, respectively (11). Each subunit possesses an active site that differs in its substrate specificity. The active site of the ␤-chain hydrolyzes uncharged substrates, whereas the ␣-subunit, in addition, slowly cleaves negatively charged substrates (12). However, only ␤-hexosaminidase A is able to hydrolyze the most important natural lipid-substrate, ganglioside G M2 , in presence of the G M2 activator protein at significant rates (13).
Mutations in any of the three genes encoding for the ␣or ␤-subunit or the G M2 activator protein can give rise to neuronal storage of ganglioside G M2 in one of the three variants of G M2 gangliosidoses. In variant B (Tay-Sachs disease), deficiency of the ␣-subunit caused by mutations of the HEXA gene brings about a lack of ␤-hexosaminidase A and ␤-hexosaminidase S activities, whereas the activity of Hex B is normal or increased. Variant 0 (Sandhoff's disease) is characterized by ␤-hexosaminidase A and B deficiency due to mutations of the HEXB gene encoding the ␤-subunit. In variant AB the G M2 activator protein is deficient due to mutations of the GM2A gene, whereas ␤-hexosaminidase A and B activity levels appear to be normal (for review see Ref. 14).
Although human hexosaminidases have been studied intensively, including their gene structures, peptide sequences, biosynthesis, and post-translational modifications, their active sites remain unknown. Moreover, although Hex B has been * This work was supported by Grant SFB 284 from the Deutsche Forschungsgemeinschaft. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This paper is dedicated to Prof. Heinz Egge on the occasion of his 65th birthday.
Therefore, we studied the binding site of Hex B using active site-directed labeling techniques employing [ 3 H]-1-ATB-Gal-NAc as a photoaffinity label for the purified enzyme. The chemical structure of this ligand is shown in Fig. 1. Because glycoside hydrolases usually show a high glycon-specificity and a relatively low aglycon-specificity, the photolabile group was introduced into the aglycon of the enzyme-resistant thioglycoside so that recognition of the compound as a substrate analogue was not impaired (16). After photolysis at 350 nm, the diazirine group located in the aglycon forms a highly reactive carbene, which can react covalently with amino acid side chains. [ 3 H]-1-ATB-GalNAc has been previously shown to specifically bind at the substrate binding site of ␤-hexosaminidase A (17). The K i -values of this compound were estimated to be 1.8 mM and 3.0 mM for the ␤-subunit of ␤-hexosaminidase A and B, respectively.
The present study includes specific photoincorporation of [ 3 H]-1-ATB-GalNAc into purified human lysosomal Hex B, tryptic digestion of the covalently labeled enzyme, and isolation and sequencing of the labeled peptide.
The photoaffinity label 1-ATB-GalNAc was synthesized and tritiated as previously reported (17). The resulting [ 3 H]-1-ATB-GalNAc had a specific radioactivity of 0.92 Ci/mmol. It was homogenous as judged by TLC, NMR spectroscopy, and fast atom bombardment-mass spectrometry. All other reagents were of the highest purity available.
Hexosaminidase and Protein Assays-Enzyme activity was measured with the fluorogenic substrates MUG and MUGS as described previously (13). One unit of enzyme activity is defined as the amount of enzyme that hydrolyzes 1 mol of MUG/minute. Protein concentrations were determined according to the procedure of Bradford (19) using bovine serum albumin as standard.
Purification of Lysosomal ␤-Hexosaminidase B-Isolation of Hex B from human liver was performed as described previously (20) and modified using fast protein liquid chromatography techniques. All buffers contained 0.01% NaN 3 (w/v) as preservative. Briefly, liver extracts were brought to 70% saturation with (NH 4 ) 2 SO 4 . The precipitated proteins were dissolved in phosphate buffer (50 mM, pH 7.0) and loaded onto a concanavalin A-Sepharose 4B column. Nonspecifically bound protein was washed away with the same buffer containing 200 mM NaCl. Bound glycoproteins were eluted using the same buffer containing 10% (w/v) methyl ␣-D-glucopyranoside. Fractions with hexosaminidase activity were dialyzed against phosphate buffer (10 mM, pH 6.0) and loaded onto a Q-Sepharose column, which was connected to a fast protein liquid chromatography system. Hex B, which elutes with the void volume, was subsequently loaded onto a S-Sepharose column and eluted with a linear gradient of 0 -0.5 M NaCl. Fractions with Hex B activity were concentrated and dialyzed against phosphate buffer (10 mM, pH 7.0, containing 150 mM (NH 4 ) 2 SO 4 ) and loaded onto a Superdex 200 gel filtration column (Q-Sepharose-HiLoad 16/10, S-Sepharose-HiLoad 16/10, Superdex 200-HiLoad 16/60, and the fast protein liquid chromatography system were from Pharmacia, LKB, Uppsala, Sweden). Enzyme activity was assayed during each step of purification. Protein was detected by continuous monitoring at 280 nm. Homogeneity of the enzyme was assessed by SDS-polyacrylamide gel electrophoresis.
Photoaffinity Labeling of Hex B and Determination of Inactivation-385 g of [ 3 H]-1-ATB-GalNAc (12 mM; specific radioactivity, 0.92 Ci/ mmol, 1.1 mCi) were dissolved in 100 l of citrate buffer (50 mM, pH 4.5) containing 600 g of Hex B (50 M, 150 units) in a custom-made tapered quartz vial under safelight conditions. The solution was sonified in an ice water bath for 5 s and deoxygenated by bubbling with argon for 2 min. After incubation at 37°C for 15 min, the solution was irradiated for 20 min with UV light ( max ϭ 350 nm) in a Rayonet RPR 100 reactor (Southern New England Ultraviolet Company, Middletown, CT) equipped with 16 lamps (20 W each, RPR 3500 A).
Inactivation was measured immediately after irradiation with 3 ϫ 2 l of the [ 3 H]-1-ATB-GalNAc-labeled enzyme solution. The aliquots were diluted 1:20,000 with water containing 0.02% bovine serum albumin (w/v) and assayed for enzyme activity using MUG as a substrate. Under these conditions the concentration of photolyzed [ 3 H]-1-ATB-GalNAc was lowered to approximately 1 M. Protein concentration of additional aliquots (4 ϫ 1 l) were measured to determine the specific activity of the labeled enzyme. As control, a photolabeling experiment under the same conditions but in the absence of [ 3 H]-1-ATB-GalNAc was performed and treated as above.
Photoaffinity Labeling of Hex B in Presence of ␦-Lactone and Determination of Inactivation-193 g of [ 3 H]-1-ATB-GalNAc (12 mM, 0.55 mCi) and 11 g of ␦-lactone (1 mM) were dissolved in 50 l of citrate buffer (50 mM, pH 4.5) containing 300 g of Hex B (50 M, 75 units) and treated as described above. Hex B activity after irradiation was determined as described above. Under these conditions the concentration of ␦-lactone was lowered to approximately 0.05 M.
Separation of ␤ a -and ␤ b -Polypeptides-The entire labeled mixture obtained from the photoaffinity experiment of Hex B with [ 3 H]-1-ATB-GalNAc was freeze dried and dissolved in 180 l of Tris-HCl buffer (200 mM, pH 8.0, containing 6.6 M guanidinium-HCl and 2.2 mM EDTA). The solution was incubated at 50°C for 30 min. Disulfide bonds were reduced by the addition of 10 l of dithiothreitol (1 M) and incubation at 50°C for 2 h. After cooling, 20 l of iodoacetamide (1.2 M) was added and allowed to react for 1 h at ambient temperature in the dark. The reaction was terminated by addition of 20 l of ␤-mercaptoethanol. Subsequently, the entire mixture was loaded onto a Superdex 75 gel filtration column to remove photolysis byproducts, ␤-mercaptoethanol and iodoacetamide. Gel filtration was performed under denaturing conditions using 6 M guanidinium-HCl in phosphate buffer (20 mM, pH 5.8) as solvent. Aliquots of protein containing fractions were subjected to SDS-PAGE. Fractions containing both the ␤ a -and ␤ b -polypeptide were pooled and subsequently separated by reverse phase HPLC as follows. HPLC grade water containing 0.1% trifluoroacetic acid was used as solvent A. Solvent B consisted of isopropanol containing 0.085% trifluoroacetic acid. Solvents were delivered using a LKB 2150 low pressure gradient system. Column temperature was maintained at 60°C throughout the separation. A LiChroSpher C 8 -column (particle size, 10 m; pore width, 100 Å; 4 ϫ 250 mm) was equilibrated with 10% solvent B. The prepurified photolabeling mixture obtained from gel filtration was injected through a 2-ml loop and eluted with a gradient of 10 -50% solvent B for 45 min and 50 -97% solvent B for 10 min at a flow rate of 1 ml/min. Following the absorption at 206 nm, fractions were cut manually in the range of 1.0 -2.0 ml. Aliquots were assayed for radioactivity and subjected to SDS-PAGE. Fractions containing the ␤ a -and ␤ b -polypeptide were pooled and freeze dried.
Tryptic Digest of ␤ a -and ␤ b -Polypeptides-The separated and lyophilized ␤ a -and ␤ b -polypeptides (ϳ5 nmol each) were dissolved in 100 l of Tris-HCl buffer (50 mM, pH 8.0, containing 6 M guanidinium-HCl) and diluted to 1 ml with the same buffer containing 1 mM CaCl 2 instead of guanidinium-HCl. Modified trypsin (20 g, final protease/protein ratio of ϳ1:6, w/w) was added, and the mixtures were incubated at 37°C for 8 h.

Isolation of [ 3 H]-1-ATB-GalNAc-labeled Peptides-
The peptides generated from ␤ a -and ␤ b -polypeptide digestion were recovered by reverse phase HPLC using 0.1% trifluoroacetic acid in HPLC grade water as solvent A and 0.085% trifluoroacetic acid in acetonitrile/isopropanol (2:1, v/v) as solvent B. Column temperature was maintained at 50°C throughout the separation. The entire digest of each polypeptide was loaded onto a Nucleosil C 8 -column (3 m, 120 Å, 4 ϫ 250 mm from Knauer, Berlin, Germany), which was equilibrated with 3% solvent B. Peptides were eluted at a flow rate of 1 ml/min with a linear gradient of 3-15% solvent B for 10 min, 15-45% solvent B for 45 min, and 45-97% for 10 min. Following the absorption at 206 nm, fractions representing individual peaks were cut manually in the range of 0.3-1.4 ml, and aliquots were assayed for radioactivity. In case of the ␤ b -digest, no peptide with a significant specific radioactivity was detected, low amounts of radioactivity being distributed over all the fractions. However, in the case of the ␤ a -polypeptide, a single prominently labeled peak was obtained. A sample of this peptide (ϳ50 pmol) was subjected directly to Edman sequencing.
Rechromatography of the Most Strongly Labeled Peptide Fraction-A sample (ϳ3 nmol) of the prominently labeled peak was further purified by narrow bore HPLC using a SMART system (Pharmacia) with a Nucleosil C 18 -column (3 m, 100 Å, 2 ϫ 100 mm). Separation was achieved with HPLC grade water, containing 0.2% hexafluoroacetone (adjusted with 25% aqueous NH 3 to pH 7.0) as solvent A and 84% acetonitrile in water containing 0.03% hexafluoroacetone as solvent B (21). The reverse phase column was equilibrated with 4% solvent B at ambient temperature. The peptide fraction was injected through a 100-l loop and eluted with a linear gradient of 4 -60% solvent B in 30 min at a flow rate of 100 l/min. UV absorbance was monitored at 214 nm, and the fractions were cut manually in the range of 50 -400 l and assayed for radioactivity. The rechromatography succeeded in the separation of three peptides, which were further analyzed by MALDI-TOF-MS and sequence analysis.
Base Treatment of the Radiolabeled Peptide-Radioactivity was released from an aliquot (ϳ100 pmol) of the purified labeled peptide by incubation with ammonium hydroxide (25%, v/v) for 8 h at 37°C as described previously (22). The resulting solution was rechromatographed on the SMART system under the same conditions as described above. The amino acid sequences of the resulting two peptides were determined. 3 H radioactivity was measured by liquid scintillation counting using Ultima Gold from Packard as scintillant. For SDS-PAGE, protein samples dissolved in 6 M guanidinium-HCl were dialyzed against Tris-HCl buffer (50 mM, pH 6.0, containing 4% SDS and 12% glycerol) before being applied to the gel. Denaturing SDS-PAGE was performed on 12% polyacrylamide gels according to Schä gger and von Jagow (23).
MALDI-TOF-MS-Matrix-assisted laser desorption ionization mass spectrometric analyses (24) of the peptides were conducted at Shimadzu-Europe (Duisburg, Germany) using a Kratos Maldi III laser desorption time-of-flight mass spectrometer equipped with a 70-cm flight tube and a 337 nm nitrogen laser with a 3-ns pulse width. The spectrometer was operated in reflector mode with an acceleration voltage of 20 kV. The matrix was a saturated solution of 3,5-dimethoxy-4hydroxycinnamic acid (sinapinic acid) in ethanol/water (1:1). Insulin was used as standard. Sample mixture for laser desorption mass spectral analyses contained 1 l of analyte obtained from HPLC and 1 l of matrix solution.
Peptide Sequencing-Isolated peptides were sequenced on an Applied Biosystems 477A Sequencer with an on-line phenylthiohydantoin analyzer (model 120A). For radiolabeled peptides, two-thirds of each PTH-derivative were injected into the PTH analyzer. One-third was routed to a fraction collector and used for scintillation counting in order to verify the location of the radiolabel within the sequence.
Amino Acid Alignment-The full amino acid sequence of the ␤-chain of human Hex B was compared with the SWISS-PROT protein sequence data base using the BLITZ automatic mail server (Biocomputing Research Unit, University of Edinburgh, Scotland) for the MPsrch program, which employs the Smith and Waterman best local similarity algorithm (25). Program parameters used for the alignment included a amino acid similarity matrix (PAM value) of 120 and a INDEL cost for inserting gaps of 13.

RESULTS
Enzyme Purification-Isolation of lysosomal Hex B from human liver resulted in a final preparation that was ϳ3000-fold enriched over the liver homogenate with an overall yield of ϳ10%. The specific activity was ϳ250 units/mg.

Covalent Binding of [ 3 H]-1-ATB-GalNAc to Hex B-Incubation of Hex B (50 M) with [ 3 H
]-1-ATB-GalNAc (12 mM; specific radioactivity, 0.92 Ci/mmol) and subsequent irradiation at 350 nm for 20 min resulted in a 15% reduction of the enzyme activity. The enzyme activity was not reduced in a control experiment in the absence of the photoaffinity label. However, in the simultaneous presence of the photoaffinity label (12 mM) and the ␦-lactone (1 mM) as a transition state analogue, enzyme activity was reduced by only 4% after irradiation indicating that the ␦-lactone prevented 75% of the covalent binding of the highly reactive carbene generated from [ 3 H]-1-ATB-GalNAc at the active sites of Hex B.
Subsequent size exclusion chromatography of the reduced and alkylated protein under denaturing conditions led to the isolation of two protein-containing peaks. The first major peak contained the main portion of protein-bound radioactivity (97%) and was found by SDS-gel electrophoresis to comprise both the ␤ a -and the ␤ b -polypeptide. The second minor peak contained only ϳ3% of protein-bound radioactivity and corresponded to the ␤ p -polypeptide, as shown by SDS-PAGE (Fig. 2). Additionally, a group of low molecular weight peaks was detected containing the reducing and alkylating reagents together with photolysis byproducts of [ 3 H]-1-ATB-GalNAc.
Separation of ␤ a -and ␤ b -Polypeptides-The ␤ a -and ␤ bpolypeptide mixture obtained from the photoaffinity experiment and subsequent size exclusion chromatography was directly loaded onto a C 8 -column. Separation was achieved by elution at 60°C using a gradient of 10 -97% isopropanol (0.085% trifluoroacetic acid), which led to two major radioactive protein peaks (Fig. 3A). SDS-PAGE revealed that the first peak contained the ␤ b -polypeptide and the second contained the ␤ a -polypeptide (Fig. 3B). Interestingly, the radioactivity bound to the ␤ a -chain (ϳ0.31 Ci/mmol) was three times higher than that of the ␤ b -polypeptide (ϳ0.1 Ci/mmol).
Lactone Reduces Labeling of ␤ a -In order to determine whether the ␤ a -or ␤ b -polypeptide participate in the active binding site of the enzyme, labeling was performed in the presence of a competitive inhibitor. Hex B was photoaffinity labeled with [ 3 H]-1-ATB-GalNAc (12 mM) in the presence of the powerful and competitive inhibitor ␦-lactone (1 mM), which is a transition state analogue having a K i value in the submicromolar range (26,27). Separation of the polypeptides resulted in labeled ␤ a -and ␤ b -chains of nearly identical specific radioactivity (Fig. 4). However, in the absence of the ␦-lactone specific radioactivity of the ␤ a -polypeptide increased 3-fold, whereas that of the ␤ b -polypeptide remained almost constant.

Identification of the Amino Acid Labeled by [ 3 H]-1-ATB-Gal-NAc-Labeled
␤ a -polypeptide (ϳ5 nmol) was digested with modified trypsin. Separation of the resulting peptides was then carried out by HPLC (Fig. 5A). Each peak was collected manually for determination of radioactivity by liquid scintillation analysis. Fig. 5B shows the distribution of radioactivity in the individual fractions. A prominently labeled single peak eluted at a retention time of 52 min. Inclusion of ␦-lactone in the photoaffinity experiment of Hex B specifically reduced the incorporation of radiolabel into this peptide 2.8-fold (Fig. 5C).
An aliquot (ϳ50 pmol) of this peptide fraction was directly subjected to automated sequence analysis. The effluent stream from the Edman degradation was split before it entered the phenylthiohydantoin analyzer, so that radioactivity of the effluent from each sequencing cycle could be correlated with analysis of PTH-derivatives. Automated Edman degradation of the labeled peptide yielded the sequence EISEVFPDQFIHLG-GDEVEFK (peptide I), corresponding to amino acid positions 339 -359, within the preprosequence of human lysosomal ␤-hexosaminidase B. The sequence assignment was unambiguous, no secondary sequences were observed. However, contrary to our expectations, no significant burst of radioactivity and no severely reduced PTH-derivative signal were observed in any of the cycles during sequence analysis. All of the radioactivity was recovered in the sequencer waste. This result strongly indicated the presence of a labile ester bond between an acidic side chain of either a glutamic or an aspartic acid and the photoaffinity label, which was then cleaved during sequencing under the conditions of Edman degradation.
For further analysis, the radiolabeled peptide (ϳ50 pmol) was rechromatographed by narrow bore HPLC (Fig. 6) at pH 2.0 using the standard solvent system (0.1% trifluoroacetic acid in water versus 0.085% trifluoroacetic acid and 84% acetonitrile in water). As shown in Fig. 6A, rechromatography yielded one single peak comprising all of the radioactivity. However, its calculated specific radioactivity (0.066 Ci/mmol) was rather low and suggested a contamination by an unlabeled peptide. There-fore, the peak was rechromatographed in the solvent system of Serwe et al. (21) (0.2% hexafluoroacetone in water, pH 7.0, versus 0.03% hexafluoroacetone and 84% acetonitrile in water). As shown in Fig. 6B, the apparently homogenous peak (Fig. 6A) separated into three peaks designated peaks 1, 2, and 3, which were further analyzed by MALDI-TOF-MS and sequence analysis. The majority of radioactivity was found in peak 3. The specific radioactivity calculated for this peptide of 0.91 Ci/mmol is close to the specific radioactivity of [ 3 H]-1-ATB-GalNAc indicating a 1:1 stoichiometry of labeling.
MALDI-TOF-MS gave a (MϩH) ϩ ion at m/z 2438.7 in the case of peak 1, which is close to the calculated mass of the peptide I, E 339 ISEVFPDQFIHLGGDEVEFK 359 with 2436.68. Peak 2 gave rise to a single molecular (MϩH) ϩ ion at m/z 3342.3, a value close to the calculated mass of an elongated peptide I (positions 339 -366; 3338.7), plus iodoacetamide. In this minor cleavage product, the potential tryptic site at lysine 359 has apparently been skipped. In case of peak 3, a (MϩH) ϩ ion at m/z 2729.9 was found matching the calculated mass for An aliquot (30 pmol) of peak 3 containing the photoaffinity labeled peptide I was subjected to sequence analysis. Again, no modified amino acid and no burst of radioactivity were detected. Obviously, the bond between the peptide and the ligand was cleaved again by the harsh conditions of Edman chemistry. Therefore, following the strategy of Gebler et al. (22), we decided to choose an indirect approach in order to identify the labeled amino acid. Assuming that an acid-labile isobutyl ester had been formed between the C 4 -aglycon of [ 3 H]-1-ATB-Gal-NAc and the carboxyl group of an amino acid with an acidic side chain, then this ester should be cleavable by base treatment. Aqueous ammonia should liberate tritiated butanol-thio-Gal-NAc from the peptide and simultaneously convert a significant proportion of the carboxyl group of the former attachment site to the corresponding amide.
Consequently, 100 pmol of labeled peptide from peak 3 (Fig.  6B) were reacted with 25% aqueous NH 3 at 37°C for 8 h and rechromatographed under the same separation conditions that had led to segregation of peaks 1, 2, and 3 (see Fig. 6B). All of the radioactivity now came off the column with the void volume, as had been expected for the highly polar GalNAc derivative. Within the time window where peaks 1 and 3 normally eluted, two peaks of equal intensity were observed. The retention time of the earlier eluting peak corresponded to the retention time of peak 1, the unlabeled peptide I. The second peak was intermediate in retention between peaks 1 and 3. None of the peaks bore any residual radioactivity.
Both peptides were subjected to sequence analysis. The earlier eluting peptide was found to be peptide I (positions 339 -359). The later eluting peptide had the same sequence, with the notable exception of cycle 17, where PTH-glutamine instead of PTH-glutamic acid was unambiguously identified. This result strongly suggests that [ 3 H]-1-ATB-GalNAc was covalently attached with its aglycon to Glu-355 via an ester bond, which could be cleaved by aminolysis reverting one-half of the derivatized Glu-355 back to glutamic acid and converting the other half to glutamine. DISCUSSION Our attempts to probe the binding sites of hexosaminidases were initially based on the pyrrolidine alkaloid 2-acetamido-1,4-imino-1,2,4-trideoxy-D-galactitol, which was shown to act as a competitive inhibitor of enzyme activity on both the ␣-(K i ϭ ϳ220 M) and the ␤-subunit (K i ϭ ϳ18 M) of ␤-hexosaminidase A (28). In order to achieve covalent photoincorporation of this compound into hexosaminidases, we introduced a photolabile diazirine group via a butyl spacer at the nitrogen ring atom. Unexpectedly, incubation of ␤-hexosaminidase A with this photolabel even in concentrations up to 1 mM and subsequent irradiation did not result in any reduction of enzyme activity, although the inhibitory potential (K i ϭ ϳ14 M, ␤-subunit of ␤-hexosaminidase A) of the galactitol derivative was not affected by attachment of the photoaffinity label (data not shown). The ineffectiveness of certain carbohydrate-linked diazirines to react covalently with their receptors has been reported by Lehmann and Petry (29). However, 1-ATB-GalNAc, although showing only a moderate K i value (ϳ3 mM) is able to reduce enzyme activity after binding and therefore was used as a photoaffinity label for purified Hex B.
Our results provide strong evidence that the amino acid sequence E 339 ISEVFPDQFIHLGGDEVEFK 359 is involved in the substrate binding site of human lysosomal ␤-hexosaminidase B. This conclusion is supported by the ability of ␦-lactone to compete effectively for the active sites of the enzyme and to reduce labeling of the ␤ a -polypeptide and peptide I with [ 3 H]-1-ATB-GalNAc. The fact that photoincorporation of the label was not totally prevented was somewhat surprising, because addition of 1 mM ␦-lactone into the labeling experiment should completely protect the active sites of Hex B from covalent reaction with [ 3 H]-1-ATB-GalNAc. On the other hand, it is known that the ␦-lactone is relatively labile and in equilibrium with 2-acetamido-2-deoxy-D-glucono-1,4-lactone and 2-acetamido-2-deoxy-D-gluconic acid (30). Therefore we compared the inhibitory potential of a freshly prepared aqueous solution of ␦-lactone (1 mM) with the activity of a ␦-lactone solution (1 mM) treated under the labeling conditions. It turned out that a freshly prepared aqueous solution of ␦-lactone showed a K i value (ϳ25 nM) that was about 20 times lower than the solution of ␦-lactone that ran through the whole labeling procedure (K i ϭ ϳ500 nM) (data not shown). Referring this result to the competition experiment, it is likely that the initially applied amount of ␦-lactone (1 mM) is severely reduced during the labeling experiment, so that the ␦-lactone concentration appears to be in the same order of magnitude as the enzyme concentration (50 M) or even below the active site concentration (100 M). Under these conditions, complete protection of labeling by the ␦-lactone cannot be expected anymore.
Laser desorption mass spectrometry of the predominantly labeled peptide identified peptide I with a single affinity ligand attached. During the isolation and purification procedures of peptide I, it was noticed that the label was partially lost if the peptide was treated with strongly acidic or basic solutions.

FIG. 5. Separation of the tryptic digest of the ␤ a -polypeptide after labeling with [ 3 H]-1-ATB-GalNAc by reverse phase HPLC.
A, HPLC separation of peptides derived from the labeled ␤ a -polypeptide. The arrow indicates the labeled peptide that was further purified by narrow bore HPLC and subjected to sequence analysis. Photoaffinity labeling of Hex B with [ 3 H]-1-ATB-GalNAc, separation of ␤ a -and ␤ bchains, digestion of labeled ␤ a -polypeptide with trypsin, and HPLC analysis were conducted as described under "Experimental Procedures." B, the corresponding radioactivity profile to A as obtained by liquid scintillation counting. C, control to B radioactivity profile obtained after labeling of Hex B in the presence of the competitive inhibitor ␦-lactone (1 mM) to demonstrate the specificity of the reaction by reduced label incorporation into the peak at 52 min. Experimental details are provided under "Experimental Procedures." Similarly, identification of the modified amino acid within the sequence of peptide I by Edman degradation failed, presumably because the protocols used for automated sequencing include treatment with 25% aqueous trifluoroacetic acid for 20 min at 64°C. These observations led to the assumption that the photolabel was bound to the peptide via an ester bond that would be hydrolyzed under strongly acidic or basic conditions (31). Therefore, we treated the labeled peptide I with 25% aqueous ammonium hydroxide (22), which resulted in quantitative cleavage of the label from the peptide and simultaneously, conversion of Glu-355 into Gln-355, as shown by Edman degradation. Exactly the same results were obtained when we performed the entire photoaffinity labeling experiment with ␤-hexosaminidase B partially purified from human placenta (specific activity 130 units/mg) instead of Hex B from human liver (data not shown).
The mechanisms of enzymatic glycoside hydrolysis and the approaches to identify active site residues by labeling techniques have been recently reviewed by Withers and co-workers (32,33). In case of retaining glycosidases, such as hexosaminidases (34), it is generally known that hydrolysis of a glycoside bond is accomplished by a pair of acidic side chains, which participate simultaneously in this reaction. The carboxyl group of one of these side chains functions as a general acid and base, whereas the other one acts as a nucleophile and a leaving group. In this context, it is interesting that in our experiments only one acidic amino acid is exclusively labeled by an active site-directed affinity ligand. The labeling is highly specific, because peptide I is the only peptide carrying significant amounts of radioactivity, and furthermore the labeling reaction can be suppressed to a significant extent by the addition of a competitive but nonreactive inhibitor. Moreover, the reaction happens to the side chain of Glu-355 exclusively and extends to neither Asp-354 nor Glu-357, making it unlikely that the carbene just picks out the best nucleophile around. On the other hand, the photoreactive site within the butyl residue is rather remote from the bound sugar and furthermore possesses an inherent flexibility, making it unlikely to assign labeled Glu-355 to one of the amino acids involved in catalysis. With the help of naturally occurring mutations of the B1 variant (reviewed in Ref. 35), Brown and Mahuran (36) reported that Arg-178 of the ␣-subunit (corresponding to Arg-211 on the ␤-subunit) could be implicated in the active site of human hexosaminidases. Although the data strongly suggest that these residues are important for the activity of hexosaminidase, the assignment that they participate in the cleavage of the glycosidic bond is not compelling. Moreover, as shown by molecular modeling, mutations of these basic amino acids result in a severe disruption of the three-dimensional conformation of the subunits (35) and therefore do not prove that ␣-subunit Arg-178 or ␤-subunit Arg-211 are involved catalytically in the active site. Although our results of photoaffinity labeling of the binding site of Hex B with 1-ATB-GalNAc focus attention on Glu-355, full understanding of the functional aspects of this amino acid will undoubtedly await elucidation of the tertiary structure of Hex B.
Because this is, to our knowledge, the first identification of a specific sequence at the binding site of the ␤-subunits from human hexosaminidase B, we were interested in the extent to which this sequence is conserved among hexosaminidases from other species and human ␣-chain in particular. Alignment of amino acid sequences for hexosaminidases from various species revealed that Glu-355 is highly conserved between the ␣-subunit of human ␤-hexosaminidase A, the ␣and ␤-subunit of mouse, and the ␣-subunit of Dictyostelium discoideum (Fig. 7). Moreover, there are several highly conserved amino acids in the vicinity of Glu-355, being potential targets for future studies using site-directed mutagenesis to probe the binding site in hexosaminidases. Interestingly, Glu-355 is located near Cys-360, which is bridged with Cys-309, connecting the ␤ a -and the ␤ b -polypeptides in Hex B. These cysteine residues and also the size of the loop they enclose are conserved in all known hexosaminidase subunit sequences (37).
It is clear from previous attempts to locate the active sites of A, the chromatogram obtained by rechromatography of an aliquot (50 pmol) of the prominently labeled peak and the corresponding radioactivity profile using 0.085% trifluoroacetic acid in 84% acetonitrile as solvent B. Tryptic digestion of the labeled ␤ a -chain, separation of the resulting peptides, and rechromatography of the prominently labeled peak on narrow bore HPLC were conducted as described under "Experimental Procedure." B, rechromatography of another sample (3 nmol) performed as described above but now using 0.03% hexafluoroacetone in 84% acetonitrile (pH 7.0) as solvent B. The apparently homogenous peak (see A) separated into three peaks, designated peaks 1, 2, and 3, which were further analyzed by MALDI-TOF-MS and sequence analysis.
hexosaminidases that there are several regions of homology among them. One was found using the predicted amino acid sequence of the bacterial ␤-N-acetylhexosaminidase gene from Vibrio vulnificus. Sequence alignments identified a region of 19 amino acids largely conserved around the central portion of the amino acid sequence among known hexosaminidases but not including Glu-355 (38). Although regions of homology that have been identified by sequence alignments may prove to be important in the structure of the active sites of hexosaminidases, our results with photoaffinity labeling and peptide sequencing provide the first chemical evidence identifying a specific amino acid at the binding site of Hex B.