Physiological Substrates for Human Lysosomal (cid:1) -Hexosaminidase S*

Human lysosomal (cid:1) -hexosaminidases remove terminal (cid:1) -glycosidically bound N -acetylhexosamine residues from a number of glycoconjugates. Three different isozymes composed of two noncovalently linked subunits (cid:2) and (cid:1) exist: Hex A ( (cid:2)(cid:1) ), Hex B ( (cid:1)(cid:1) ), and Hex S ( (cid:2)(cid:2) ). While the role of Hex A and B for the degradation of several anionic and neutral glycoconjugates has been well established, the physiological significance of labile Hex S has remained unclear. However, the striking accumulation of anionic oligosaccharides in double knockout mice totally deficient in hexosaminidase activity but not in mice expressing Hex S (Sango, K., McDonald, with changes in the ratio of n -hexane and water (55/44/1, 55/43/2, etc. until 55/37/8, v/v). Final purification of SM2 was achieved by repeated silica gel flash chromatography of the combined SM2-containing fractions using the solvent systems chloroform/methanol/water (62/30/1.8, v/v/v) and then chloroform/methanol/water (70/30/2, v/v/v). Final purification of SB2 was achieved by silica gel flash chromatography of the combined SB2-containing fractions using the solvent system chloroform/metha-nol/water (65/30/2.5, v/v/v).

and Hex B (␤␤) were believed to be the major functional isozymes, and Hex S (␣␣) a minor labile form without significant activity (2). Several attempts to isolate pure Hex S from human tissues or cell homogenates yielded preparations with poor enzymatic activity (3)(4)(5).
Each subunit possesses an active site characterized by its own substrate specificity (5). The active site of the ␤-subunit hydrolyzes uncharged substrates, whereas the ␣-subunit, in addition, cleaves negatively charged substrates. Only the ␣␤heterodimer Hex A is able to degrade ganglioside GM2 (Fig. 1B) at significant rates in the presence of the GM2 activator protein (GM2AP).
A group of severe neurodegenerative storage diseases, the GM2 gangliosidoses, results from mutations in any of the genes encoding the two hexosaminidase subunits and GM2AP. Tay-Sachs disease is caused by mutations in the HEXA gene resulting in a deficiency in Hex A and Hex S, whereas in Sandhoff disease the lack of Hex A and Hex B activity is observed due to mutations in the HEXB gene. The GM2 gangliosidoses are characterized by a massive accumulation of ganglioside GM2 and related glycolipids in neuronal lysosomes. Depending on the defect and its severity, other tissues may also be affected by lipid and oligosaccharide accumulation (1). The severity of the clinical phenotype is directly related to the amount of residual enzyme activity (6). In addition, the isozyme that is not affected by mutation is able to compensate in part for the loss of the affected hexosaminidase activity (1).
Mouse models have been generated for Tay-Sachs disease (deficient in the ␣-subunit but expressing Hex B), for Sandhoff disease (deficient in the ␤-subunit and expressing Hex S) as * This work was supported by Deutsche Forschungsgemeinschaft Grant SFB 284 and the Fonds der Chemischen Industrie. 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.
Surprisingly, the double knockout mice show a mucopolysaccharidosis phenotype with an accumulation of anionic oligosaccharides in addition to the pathological and biochemical features of GM2 gangliosidoses (8,9). In the Hexb Ϫ/Ϫ mice no increased accumulation of glycosaminoglycans (GAGs) was observed indicating that Hex S, which is still expressed in the mutant mice, is involved in GAG catabolism. GAGs contain ␤(1-4)-and ␤(1-3)-linked N-acetylhexosamine residues which become sensitive to hexosaminidases when exposed as terminal sugar residues during their degradation (10).
We have purified and characterized recombinant Hex S (rHexS) after expression in insect cells. To study its substrate specificity the storage glycans were isolated from the urine of the hexosaminidase double knockout mice. In addition, other water-soluble and lipid substrates were evaluated. The results demonstrate that Hex S is highly active on a wide range of substrates and support the contention that the enzyme is physiologically important in vivo.

Enzyme Preparation and Identification
Recombinant Hex S was expressed in insect cells using the baculovirus expression system. Sf21 cells were infected with a recombinant virus containing the cDNA encoding for the human hexosaminidase ␣-subunit. The baculovirus construct was prepared as described for the expression of recombinant Hex B (11). The expression was terminated 96 h after infection. After centrifugation of the serum-free culture medium of infected cells the supernatant was subjected to lectin affinity chromatography with concanavalin A. The column was eluted with 15% (w/v) methyl-␣-D-mannoside in equilibration buffer (25 mM sodium phosphate, pH 7.0, 200 mM sodium chloride) and the enzyme activity was monitored with MUG as substrate. The eluate of the concanavalin A column was concentrated and passed through a fast-flow cation exchange resin (POROS 20 HS) equilibrated in 50 mM sodium phosphate buffer, pH 6.5. rHexS was eluted without binding, in contrast to contaminating glycoproteins in the ConA eluate. Subsequently the fraction containing Hex S activity was applied to a weak anion-exchange material (POROS 20 HQ) equilibrated in 50 mM sodium phosphate buffer, pH 7.0. The column was eluted by increasing the concentration of sodium chloride in the equilibration buffer (by 10 mM/column volume), at a flow rate of 4 ml/min. Purification to 90% homogeneity was shown by SDS-PAGE. The purified preparation of rHexS was identified by Western blotting and N-terminal sequencing using Edman degradation and ESI-MS/MS sequencing of tryptic and chymotryptic peptides after narrow-bore reverse phase HPLC separation as described below. Western blotting was performed according to Towbin et al. (12) using an antiserum raised against the denatured hexosaminidase ␣-subunit (13). Hex A and Hex B were purified from human liver to apparent homogeneity according to Liessem et al. (14). Their purity and identity were ensured by SDS-PAGE and Western blotting (12,13).

Characterization of Post-translational Modifications of rHexS
After proteolytic cleavage and reverse phase separation of the resulting peptides the post-translational modifications of rHexS were analyzed by mass spectrometry: N-glycosylation of rHexS glycopeptides was demonstrated by treatment with N-glycosidase F (15) and by ESI-MS/MS fragment analysis as described below. Disulfide linkages were analyzed by comparing the masses of the proteolytic cleavage products of unmodified active rHexS with those of active rHexS alkylated by iodoacetamide, and those of reduced and alkylated rHexS as established in our laboratory for the analysis of the hexosaminidase ␤-subunit (15). In addition, chymotryptic cleavage of rHexS was used under the conditions described for tryptic degradation (15) and proteolytic digestion products were also analyzed by ESI-MS/MS as described below, in addition to MALDI-MS.

Activator Preparation
Recombinant GM2AP was expressed in insect cells using the baculovirus expression system and purified as described previously (16).

Determination of K m and V max
The K m and V max values for recombinant ␤-hexosaminidase S were determined by the method of Lineweaver and Burk (17). Enzyme activities were measured as described (5)

Determination of pH Optimum with the Artificial Substrates MUG and MUGS
Enzyme activities were measured as described (5)   Tay-Sachs disease (TSD) (7) Hexa Ϫ/Ϫ Hex B Limited GM2 storage in gray and white matter of the brain, average lifespan: Ͼ2 years (normal) Sandhoff disease (SD) (7) Hexb Ϫ/Ϫ Hex S Accumulation of GM2 and GA2 in gray and white matter, urinary excretion of N-glycan fragments, average lifespan: 4-5 months Combination of TSD and SD (8,9) Hexa Ϫ/Ϫ, Hexb Ϫ/Ϫ None Massive accumulation of GM2 and GA2 in gray and white matter, decrease in levels of cerebrosides and sulfatides (SM4)3 demyelination, urinary excretion of N-glycan fragments, mucopolysaccharidosis, average life span: 1-4 months 25 mM citrate buffer), respectively, and varying the pH of the solution in the range from 3.0 to 6.0.

Presentation of Data
The data obtained with the artificial substrates are means of at least duplicate determinations. Deviations did not exceed Ϯ 5% of the mean.

Animals
Generation of hexosaminidase-deficient mice through targeted gene disruption and cross-breeding of the knockout animals to obtain the double knockout mice has been described previously (7,8). The urine and kidneys of knockout mice and wild type littermates of 10 -20 weeks of age were analyzed.
Isolation of the Sulfoglycolipids SM2 and SB2 from Rat Kidney SM2 and SB2 were isolated according to Jennemann et al. (18) with some modifications. Kidneys were homogenized with an Ultra-Turrax, freeze-dried, and extracted twice with acetone using an ultrasonic bath. The residual pellet then was extracted for GSLs two times with chloroform/methanol/water (10/10/1, v/v), once with chloroform/methanol/ water (30/60/8, v/v) using an ultrasonic bath, and the combined chloroform/methanol/water extracts were dried in a rotary evaporator. To remove most phospholipids the chloroform/methanol/water extract was treated with 0.1 M methanolic KOH for 2 h at 37°C and neutralized with acetic acid. To remove salts the extract was dialyzed 5 times against water and subsequently freeze dried. Neutral and acidic GSLs were separated on a DEAE column and acidic GSLs were split into fractions by eluting with a stepwise gradient of 20, 80, 200, 500, and 1000 mM potassium acetate in methanol. Again salts were removed by dialyzing 5 times against water and subsequently freeze drying the micellar lipid solution. SM2 then was isolated from the 80 mM potassium acetate fraction and SB2 from the 500 mM potassium acetate fraction. Each fraction was further purified by silica gel flash chromatography using a stepwise gradient of propanol-2/n-hexane/water, with changes in the ratio of n-hexane and water (55/44/1, 55/43/2, etc. until 55/37/8, v/v). Final purification of SM2 was achieved by repeated silica gel flash chromatography of the combined SM2-containing fractions using the solvent systems chloroform/methanol/water (62/30/1.8, v/v/v) and then chloroform/methanol/water (70/30/2, v/v/v). Final purification of SB2 was achieved by silica gel flash chromatography of the combined SB2-containing fractions using the solvent system chloroform/methanol/water (65/30/2.5, v/v/v).

Preparation of a Trisaccharide from Chondroitin 6-Sulfate
The model substrate for Hex S was obtained by enzymatic degradation of chondroitin 6-sulfate according to Kresse et al. (20). Briefly, after exhaustive digestion of chondroitin 6-sulfate with testicular hyaluronidase (EC 3.2.1.35) in 100 mM NaAc buffer, pH 5.0, containing 150 mM NaCl, for 45 h at 37°C the ethanol-soluble degradation products were chromatographed on a column of Sephadex G-25 fine (24 ϫ 850 mm) equilibrated and eluted with 1.0 M NaCl, at a flow rate of 0.3 ml/min. The fractions were monitored by FACE profiling. Tetrasaccharide peak fractions were desalted using the same column equilibrated with water. The tetrasaccharide was then treated with ␤-glucuronidase (EC 3.2.1.31) in 100 mM sodium citrate buffer, pH 5.2, for 48 h at 37°C. The trisaccharide product C6S-3 ( Fig. 1C) was chromatographed and desalted as described for the tetrasaccharide.

Isolation of Urinary Glycans in the Hexosaminidase-deficient Mice
The urine was collected in a metabolism cage for mice (Scanbur A/S) in which the mice stayed 8 h per day. Neutral urinary oligosaccharides were prepared by solid phase extraction on graphitized carbon as described by Klein et al. (21). Briefly, the columns (0.5 ml) equilibrated with 3 ml of water were loaded with 150 -300 l of urine sample and washed with 2 ml of water again. Neutral oligosaccharides were eluted with 2 ml of 25% (v/v) acetonitrile in water. Anionic urinary oligosaccharides were prepared according to Lindahl et al. (22) with some modifications. A 2.6-ml DEAE-Sephacel column was equilibrated first with 50 mM Tris/HCl, pH 7.2, and then with washing buffer (50 mM sodium acetate buffer, pH 4.0) at a flow-rate of 1.2 ml/min. After loading of 150 -300 l of urine and 15 ml of washing buffer the column was eluted with either 150 mM lithium chloride in 50 mM sodium acetate buffer, pH 4.0, for the isolation of DS-5 (A) or 2.0 M lithium chloride in the same buffer for the preparation of GAG OS (B). The eluates A and B were separately desalted by size exclusion chromatography on a TSK-Gel HW40F column (16 ϫ 390 mm) with a flow rate of 1.0 ml/min.

Enzyme Assays
Hex S activity was assayed with the artificial substrates MUG or MUGS as described (5). Incubation mixtures using MUG as substrate contained the following components in a final volume of 200 l: sodium citrate buffer (50 mM, pH 4.4), 20-l aliquots of enzyme solution, 10 g of bovine serum albumin, and 1 mM MUG. When using MUGS the final substrate concentration was 0.5 mM and the buffer concentration was 25 mM, pH 4.0. The incubation was stopped after 30 min at 37°C by adding 1 ml of sodium carbonate/glycine buffer (200 mM each, pH 9.5). Then, the amount of 4-methylumbelliferone released was measured fluorimetrically. One unit of enzyme activity is defined as the amount of enzyme that catalyzes the hydrolysis of 1 mol of MUG/min at 37°C.
Oligosaccharide degradation by hexosaminidases was assayed as follows: 1-5 nmol of purified N-glycan fragment H 3 N 3 (Fig. 1D), 0.5-2 nmol/oligosaccharide DS-5 ( Fig. 1C) or GAG OS, and 4 nmol of C6S-3 ( Fig. 1C), respectively, were dried in a centrifugal vacuum evaporator (CVE, Savant). Standard assay mixtures contained 200 milliunits of (MUG) ␤-hexosaminidase activity and 4 g of bovine serum albumin in a final volume of 40 l. For subsequent mass spectrometric analysis sample solutions were buffered by ammonium acetate (5 mM, pH 4.0). By addition of 40 l of methanol the incubation was stopped after 5 h unless indicated otherwise. Concomitantly, samples lacking enzyme activity as well as samples devoid of substrate were incubated as control experiments under identical conditions.
The GAG OS mixture isolated from the urine of double knockout mice was incubated with chondroitinase AC (EC 4.2.2.5) and chondroitinase ABC (EC 4.2.2.4), respectively, under the same conditions as with the hexosaminidases except that 175 milliunits of enzyme activity and ammonium acetate buffer (5 mM, pH 8.0) were used; the samples were incubated for 9 h at 37°C and the reaction products were analyzed by FACE. Concomitantly, control assays lacking either enzymes or substrates were run under identical conditions, none of them indicating that any digestion took place. The degradation of glycolipid substrates by the hexosaminidase isozymes was tested in a micellar assay (A) and the hydrolysis of SM2 by Hex S was also examined in a liposomal assay (B).
Micellar Assay-Glycolipids were incubated with the A, S, and B isozymes under the same conditions as described in the oligosaccharide assay with the following alterations: 8 nmol of glycolipid per assay were used and sodium citrate buffer (5 mM, pH 4.3) replaced the ammonium acetate buffer in a final volume of 20 l. 1.5 g of GM2AP was added and incubation times ranged from 15 min to 17 h, as indicated in the legends to the figures. After the incubations were stopped with 20 l of methanol the assay mixtures were dried in a CVE and redissolved in 40 l of ammonium acetate buffer (0.3 M, pH 7.0). Prior to TLC analysis the samples were desalted by reverse phase chromatography. RP18 columns (0.5 ml) equilibrated with a solution of chloroform, methanol, 0.1 M KCl (3/48/47, v/v/v) were loaded with sample, washed with water, and eluted with 2 ml of chloroform/methanol 1/1 (v/v). The eluted lipids were applied to a TLC plate.
Liposomal Assay-LUVs containing SM2 as substrate were incubated with rHexS and GM2AP in a liposomal assay. The micellar assay described above was modified as follows. Standard incubation mixtures contained sodium citrate buffer (10 mM, pH 4.3), 4 g of bovine serum albumin, 1.0 g of GM2AP, and 30 milliunits of rHexS in a final volume of 40 l. A total volume of 20 l of protein mixture in citrate buffer was added to the same volume of unilamellar vesicles dissolved in Tris/Cl buffer (1 mM, pH 7.2). The liposomes (1 mM, amount of total liposomal lipids) were composed of 50 mol % PC, 20 mol % Chol, 20 mol % BMP and 10 mol % SM2 or 70 mol % PC, 20 mol % Chol, and 10 mol % SM2. After incubation stop with 40 l of methanol the mixtures were con-centrated to dryness by a stream of nitrogen and then subjected to alkaline methanolysis with 0.1 M NaOH in 200 l of methanol for 4.5 h at 37°C to remove the phospholipids. After neutralization with 1.2 l of glacial acetic acid, the samples were desalted by reverse phase chromatography on RP18 and applied to HPTLC plates.

Analysis of Glycan Samples by Fluorophore-assisted Carbohydrate Electrophoresis (FACE)
This method is based on the fluorescence labeling of glycans by reductive amination with 8-aminonaphthalene-1,3,6-tris-sulfonic acid (23). 5 l of a 0.15 M solution of the fluorophore in 15% (v/v) glacial acetic acid and 5 l of 1.0 M sodium cyanoborohydride in dimethyl sulfoxide were added to the glycan samples which had previously been dried in a CVE. After incubation for 15 h at 37°C the samples were dried in a CVE and dissolved in 8 l of HPLC water. 4 l of this solution were mixed with the same volume of 25% (v/v) glycerol in HPLC water and applied to a 32% polyacrylamide gel with a 4% stacking gel. The buffer solutions and gel components were the same as for SDS-PAGE according to Laemmli (24) except that SDS was omitted. The fluorescently labeled glycans resolved in the gel were visualized by illumina-  (Fig. 7). Digestion by chondroitinases ABC and AC indicated that the anionic storage oligosaccharides contained both iduronic and glucuronic acid residues. However, their position in the oligosaccharide chain has not been determined. Thus, the DS-5 structure shown here is one of four constitutional isomers. D, neutral oligosaccharide. The N-glycan fragment structure H 3 N 3 is the main component of the neutral storage glycans in the urine of hexosaminidase double knockout mice and Hexb Ϫ/Ϫ mice (Fig. 2). Physiological Substrates for Human Lysosomal ␤-Hex S tion with UV light of 366 nm wavelength. The gel image was photographed by a ccd camera. The migration distance depends on molecular weight and charge, i.e. the m/z values of the labeled glycans. An anionic oligosaccharide carrying intrinsic negative charges has a higher electrophoretic mobility than a neutral oligosaccharide of the same size. Thus, in the FACE gels it is only possible to compare migration distances of identically charged analytes. From our experiments using FACE we conclude for the GAG fragments with high negative net charge that the labeling efficiency decreases with increasing molecular weight of the saccharide. Therefore, band intensities are only comparable between analytes of the same molecular weight and charge. Despite these limitations the FACE display technique is one of the most powerful methods for glycan profiling currently at hand, combining high sensitivity down to the picomole range, high resolution, and relatively high tolerance of buffers and salt concentrations in the samples.

MALDI Mass Spectrometry
MALDI-MS analysis was performed on a TofSpec E (Micromass, Manchester, UK) mass spectrometer operating at an acceleration voltage of 20 kV with a 337-nm nitrogen laser. Mass spectra were externally calibrated using a maltooligosaccharide standard (4 to 10 glucose units) for glycan samples and commercially available peptides for peptide samples (Sigma). 1 l of matrix solution were mixed on-target with the same volume of glycan sample. Dihydroxybenzoic acid (10 mg/ml of acetonitrile/ water, 7/3, v/v, freshly prepared) was used as matrix for glycan analysis, ␣-hydroxycinnamic acid was used for small peptides (10 mg/ml acetonitrile, 0.1% trifluoroacetic acid in water, 6:4, v/v, freshly prepared) and sinapinic acid was used for high molecular weight peptides (10 mg/ml acetonitrile, 0.1% trifluoroacetic acid in water, 1/1, v/v).

ESI Mass Spectrometry
ESI mass spectra were acquired on a Q-Tof2 (Micromass, Manchester, UK) instrument operating at capillary voltages around 1000 V. It uses a quadrupole mass analyzer and an orthogonal acceleration time of flight mass spectrometer with a hexapole collision cell situated between the two mass analyzers. In MS/MS mode, this collision cell was flooded with argon for fragmentation analysis and collision energy was varied in the range from 20 to 60 eV. Peptides were dissolved in acetonitrile/ water, 1/1 (v/v) for positive mode analysis. According to Zaia and Costello (25) anionic glycans were dissolved in acetonitrile, 10 mM hydrochloric acid (1/1, v/v) and analyzed in negative mode. In addition, parent and daughter ion scans of DS-5 were performed on a Quattro II FIG. 2. Excretion of neutral glycans in the urine of the hexosaminidase deficient mice. The urine of mice deficient in Hex A and S (Hexa Ϫ/Ϫ) and deficient in Hex A and B (Hexb Ϫ/Ϫ) as well as the urine of mice totally deficient in hexosaminidase activity (double knockout) and of wild type littermates was collected for 8 h/day in a metabolism cage. The neutral oligosaccharides were prepared by solid phase extraction on Carbograph columns. The neutral glycans were eluted with 25% (v/v) acetonitrile in water. Aliquots of desalted glycan fractions corresponding to equal amounts of urine were evaporated to dryness and, together with a maltose oligosaccharide standard (4 to 10 glucose units), were profiled by FACE. This included fluorescence labeling with 8-aminonaphthalene-1,3,6-tris-sulfonic acid and subsequent electrophoresis on a 32% polyacrylamide gel.

Structural Analysis of Recombinant Hex S-Recombinant
human Hex S was partially purified from the serum-free culture medium of transfected insect cells by affinity chromatography on a concanavalin A-Sepharose column. This preparation contained precursor ␣-subunit and the processed ␣-subunit in a 1:1 ratio. By Edman sequencing the N terminus of the precursor ␣-subunit was identified as L 23 WPWPQNFQT and that of the polypeptide r␣ m in the processed ␣-subunit as G 85 KRHTLKNV which starts five amino acid residues upstream of the N terminus of the native ␣ m -chain (26,27). After further purification by cation and anion exchange chromatography the preparation contained Ͼ90% processed rHexS according to SDS-PAGE. Rechromatography on a POROS 20 HQ resin yielded small amounts of apparently homogeneous processed rHexS and precursor rHexS. This purified rHexS was 40 times more active against the neutral substrate MUG (52.9 units/mg of protein) than the purified pro-rHexS (1.3 units/mg). Both enzymes, precursor rHexS and rHexS, hydrolyzed the sulfated substrate MUGS at the same rate as the neutral substrate MUG. Since the complete separation of the precursor Physiological Substrates for Human Lysosomal ␤-Hex S and the processed rHexS by rechromatography caused considerable loss of material and activity we used the purified rHexS preparation still containing up to 10% precursor rHexS for the following studies.
The ␣-subunit of native Hex S carries three potential N-linked glycosylation sites, Asn-115, -157, and -295, each of which has been found to be modified by an oligosaccharide chain in a previous study (28). We analyzed the N-glycosylation of the rHexS using proteolytic cleavage and mass spectrometric analysis of the glycopeptides separated by reverse phase HPLC. N-Glycans were linked to all three N-glycosylation sites. Asn-115 carried an Nlinked GlcNAc 2 Man 3 oligosaccharide as found for the processed ␣-subunit in Hex A (29), or was not glycosylated; Asn-157 carried either GlcNAc or GlcNAc 2 Man 7-9 ; Asn-295 carried GlcNAc, GlcNAcFuc, or GlcNAc 2 Man 4 (data not shown).
To further characterize the structure of the rHexS the disulfide linkage pattern was examined using mass spectrometric analysis of proteolytic peptides as established before for the analysis of Hex B (15). We found that the postulated cystine linkage between the peptides ␣ p and r␣ m (27) is formed between Cys-58 and Cys-104 (data not shown). Another disulfide bridge of the ␣-subunit is formed by Cys-277 and Cys-328 (data not shown) as it was postulated after analysis of the native ␤-subunit (30). The third disulfide bond was detected between Cys-505 and Cys-522 (data not shown). This disulfide linkage pattern corresponds to that of the homologous ␤-subunit in Hex B (15). The ␣-subunit contains an additional cysteine residue in position 125. Both cystein residues, Cys-125 and Cys-458 of the unreduced and active enzyme protein could easily be alkylated by iodoacetamide, indicating that they are not involved in any disulfide bridge (data not shown). Taken together, the glycosyl-ation and disulfide pattern of the recombinant ␣-subunit in rHexS are in good agreement with data obtained for the hexosaminidase subunits in other studies (15,(27)(28)(29)(30).
The rHexS Cleaves the Synthetic Substrates MUG and MUGS at High Rates-The purified rHexS showed higher specific activities toward the artificial substrates MUG and MUGS (Fig. 1A) than described previously for partially purified Hex S preparations obtained from liver and an apparently homogeneous preparation from human fibroblasts (4,5). As given in Table II the corresponding V max values were of the same order of magnitude as for Hex A. Interestingly, rHexS catalyzed the hydrolysis of the anionic MUGS substrate even at a higher rate than the other isozymes (Table II). rHexS hydrolyzed the neutral substrate MUG optimally at pH 4.5 (data not shown) as did Hex S and Hex A prepared from human liver (3). The pH optimum of MUGS degradation by rHexS was at pH 3.8 (data not shown).
Neutral Oligosaccharides Are Excreted in the Urine of Hexosaminidase Knockout Mice-Uncharged glycans were found in the urine of both, the Hexb Ϫ/Ϫ mice and the double knockout mice (Hexa Ϫ/Ϫ; Hexb Ϫ/Ϫ) (Fig. 2). The most abundant neutral storage oligosaccharide was isolated and analyzed by ESI-MS/MS (data not shown) and enzymatic degradation by hexosaminidases (Fig. 3). Based on the data obtained from this and additional monosaccharide and methylation analysis (31) we assign the following structure to the major neutral storage glycan: Fig. 1D) which appears to be a product of partial N-glycan degradation (data not shown). Other neutral glycans were detected by MALDI-MS that contained 3 hexose (H) and 3-6 N-acetylhexosamine residues (N) and 4 hexose and 4 -7 N-acetylhexosamine residues. In the urine of Hexb Ϫ/Ϫ mice, ions that could be assigned to a saccharide exclusively composed of N-acetylhexosamine and hexose residues were observed up to m/z ϭ 2841.4 and corresponded to the sodium adduct of H 6 N 9 (calculated average mass: 2841.6 Da). In the urine of double knockout mice, H x N y compounds were detected up to a signal at m/z 3816.4 corresponding to H 7 N 13 (calculated average mass: 3816.5 Da; data not shown). In the urine of human patients deficient in Hex A and B activity, H 3 N 3 , two structural isomers of H 3 N 4 , and smaller fragments have been identified by NMR spectroscopy (32) carrying only N-acetylglucosamine residues at their non reducing ends. The structure of H 3 N 3 obtained from human urine is identical to the N-glycan fragment H 3 N 3 that we characterized in the knockout mice (Fig. 1D).
Neutral Storage Oligosaccharides Are Preferentially Degraded by Hex B-The N-glycan fragment H 3 N 3 (Fig. 1D) was isolated from the urine of Hex B-deficient mice. Long-term incubation with rHexS yielded only minor amounts of degradation products as detected in MALDI-MS (Fig. 3). The bottom panel in Fig. 3 shows that this substrate was most efficiently degraded by Hex B.
The Anionic Urinary Oligosaccharides Are Fragments of Dermatan Sulfate Degradation-The anionic storage glycans were prepared from the urine of the double knockout mice by anion exchange chromatography (Fig. 4). Elution with a buffer containing 150 mM LiCl (A) yielded a single, low molecular weight oligosaccharide termed DS-5 in fraction A4 (Fig. 4) whereas elution with 2.0 M LiCl (B) yielded a complex mixture of anionic oligosaccharides including DS-5 in fraction B4 (Fig. 4). This anionic oligosaccharide mixture was named GAG OS. Both, DS-5 and the GAG OS mixture, were analyzed structurally by ESI-MS/MS as well as by enzymatic degradation (Figs. 5-7). by 459 Da, by the expected mass of a disaccharide composed of a sulfated N-acetylhexosamine and a uronic acid residue as it occurs in galactosaminoglycans but not in keratan sulfate. The molecular mass of the free acid DS-5 was determined to be 1139.21 Da, which is identical to the molecular mass of a bis-sulfated GAG pentasaccharide composed of three N-acetylhexosamine and two uronic acid residues (1139.23 Da). Using a triple quadrupole ESI-MS/MS in the negative precursor ion mode, the pure DS-5 sample was scanned for molecular ions that should contain sulfate groups. Only molecular ions were scanned that upon fragmentation in the collision chamber gave rise to the fragment m/z 97, indicating the presence of sulfate or phosphate groups. By this, we detected the singly charged alkali metal ion adducts of DS-5 that corresponded by the loss of one Li ϩ to the doubly charged ions found by ESI-TOF-MS previously (Fig. 5B). To confirm that DS-5 contains sulfate but not phosphate we performed an ESI-MS/MS product ion scan of the molecular ion [M Ϫ 4H ϩ ϩ Li ϩ ϩ Na ϩ ] 2Ϫ (m/z 582.6, M ϭ DS-5) (Fig. 5C). By increasing the collision energy 4.7-fold (inserted spectrum) a new fragment of m/z 80 (due to [⅐SO 3 ] Ϫ ) but not m/z 79 (due to [PO 3 ] Ϫ ) appeared (33). Further fragments of this product ion scan confirmed the existence of a sulfated N-acetylhexosamine (HexNAc) residue [HexNAc-sulfate Ϫ H 3 O ϩ ] Ϫ at m/z 282 and a monosulfated trisaccharide containing one uronic acid (HexUA) and two HexNAc residues in complex with either a Li ϩ or Na ϩ (m/z 685 or 701: [sulfated (HexNAc) 2 HexUA Ϫ 2 H ϩ ϩ(Li ϩ or Na ϩ )] Ϫ and m/z 667 or 683 due to a further loss of H 2 O (Ϫ18 Da), respectively).
To determine whether DS-5 and the GAG OS are derived from either chondroitin or dermatan sulfate the GAG OS were incubated with the bacterial endolyases chondroitinase AC (EC 4.2.2.5) and chondroitinase ABC (EC 4.2.2.4). Whereas chondroitinase AC specifically hydrolyzes chondroitin sulfate and chondroitin sulfate-like units in polysaccharide chains, chondroitinase ABC additionally hydrolyzes dermatan sulfate (34). Both enzymes produce galactosaminoglycan disaccharides unsaturated between C4 and C5 of the hexuronic acid residues when incubated with appropriate substrates (34). By FACE analysis chondroitinase AC activity caused a shift of the complete ladder of oligosaccharides with the formation of low molecular weight products. However, degradation was not complete as it was after incubation with chondroitinase ABC (data not shown). The products of chondroitinase ABC digestion were analyzed by negative mode ESI-MS/MS. The main signal at m/z ϭ 458.07 was identified as a deprotonated monosulfated, unsaturated disaccharide (Fig. 6, calculated molecular mass 458.07 Da). Therefore, the origin of the GAG OS including the pentasaccharide DS-5 is dermatan sulfate.
The Repeating Disaccharide Unit in the Urinary GAG OS Is Sulfated in Position 4 of the GalNAc Residue-In ESI-MS/MS analysis the unsaturated disaccharide formed by chondroitinase ABC digestion of the GAG OS showed the same fragmentation pattern (Fig. 6D) as the unsaturated standard disaccharide ⌬Di-4-sulfate which is sulfated in position 4 of the GalNAc residue (Fig. 6B). We conclude that the dermatan sulfate fragments GAG OS and, thus DS of the hexosaminidase double knockout mice were degraded in vitro by rHexS and Hex A (Fig. 7A, lanes 2 and 3), but not by Hex B (Fig. 7A, lane 1). Incubation of the GAG OS or DS-5 with rHexS and Hex A, respectively, gave rise to two products (Fig.  7A, lanes 2 and 3, B, lanes 3 and 4), one of them comigrating with GalNAc. The second degradation product, derived from the pentasaccharide DS-5 by loss of one monosaccharide unit is termed DS-4. As indicated by the FACE gel in Fig. 7C, Nacetylgalactosamine (lane 3) but not -glucosamine (lane 2) or -galactosamine 6-sulfate (lane 4) was the monosaccharide released from DS-5 (lane 1). This indicates that an exo-sulfatase acted on the nonreducing sugar residue in vivo before DS-5 was excreted with the urine and confirms that the pentasaccharide DS-5 is derived from a galactosaminoglycan. Together with the structural information obtained from mass spectrometry and chondroitinase digestion this suggests the structure of the dermatan sulfate pentasaccharide DS-5 suggested in Fig. 1C.
rHexS and Hex A Degrade the Anionic Trisaccharide C6S-3-To study the degradation of chondroitin 6-sulfate by hex-  6 and 7) and devoid of enzyme (lane 8) were run under identical conditions. After incubation, the assay samples were subjected to FACE analysis. GalNAc-6-sulfate (lane 5) and the commercially available chondroitin 6-sulfate disaccharide bearing a glucuronic acid residue at the nonreducing terminus (lanes 1 and 9) served as standards for the digestion products. The assays were performed as described under "Experimental Procedures." After mild alkaline methanolysis the samples were desalted by reverse phase chromatography and applied to a HPTLC plate. The resolved bands were stained with cuprous sulfate reagent in phosphoric acid. The relative intensities of the product band, corresponding to SM3 and of the substrate band corresponding to SM2 were determined densitometrically. Blanks containing denatured enzyme were run for each assay under identical conditions and subtracted from the experimental values. A, time dependence. LUVs containing SM2 and BMP were incubated with rHexS in the presence of GM2AP in 10 mM citrate buffer (f) and in 50 mM citrate buffer (q). LUVs composed of 10 mol % SM2, 20 mol % Chol, and 70 mol % PC but devoid of BMP were incubated with rHexS in 50 mM citrate buffer in the presence of GM2AP (‚). LUVs containing both SM2 and BMP were incubated with rHexS in 50 mM citrate buffer in the absence of GM2AP (E). B, pH dependence. LUVs containing both SM2 and BMP were incubated for 30 min with rHexS in the presence (f) or absence (Ⅺ) of GM2AP. The pH ranged from 3.6 to 5.7.
osaminidases we prepared the bis-sulfated trisaccharide C6S-3 (Fig. 1C) from commercial chondroitin 6-sulfate using testicular hyaluronidase and ␤-glucuronidase. After incubation of C6S-3 with the hexosaminidase isozymes the reaction products were separated and visualized by FACE profiling as shown in Fig. 8. In the presence of either rHexS or Hex A two new product bands appeared (lanes 3 and 4) which comigrated with a GalNAc-6-sulfate standard (lane 5) and a commercially available saturated chondroitin 6-sulfate disaccharide standard (C6S-2, lanes 1 and 9), respectively. Hex B exhibited no detectable hydrolyzing activity on C6S-3 (Fig. 8, lane 2). rHexS and Hex A Degrade the Sulfated Glycolipid SM2 but Not SB2-We examined the enzymatic degradation of the monosulfated glycolipid SM2 and the bis-sulfated glycolipid SB2 by hexosaminidases. SM2 is an analogue of ganglioside GM2, and the glycolipid SB2 carries an additional sulfate ester bound to position 3 of the terminal GalNAc residue. The purified sulfated GSLs were incubated with the different hexosaminidase isozymes in a micellar assay (Fig. 9). rHexS was more active than Hex A and Hex B and catalyzed the degradation of SM2 to SM3 at significant rates even in the absence of GM2AP. Addition of GM2AP stimulated the SM2 hydrolysis by both, rHexS and Hex A. Hex B showed no significant activity in hydrolyzing SM2 in the presence or absence of GM2AP (Fig.  9A). None of the three hexosaminidases showed significant activity in degrading SB2 (Fig. 9B).
To reproduce the lysosomal conditions as closely as possible, additional degradation experiments were performed using a liposomal assay system. In this system, the lipid substrate was presented as a component of a unilamellar lipid bilayer to the water-soluble enzyme. Standard liposomes contained phosphatidylcholine and cholesterol as carrier lipids, and BMP, which is a characteristic component of lysosomal and intraendosomal membranes (35,36). Under these conditions rHexS degraded membrane-bound SM2. Significant degradation rates were achieved only in the presence of both, BMP and the GM2AP (Fig. 10A). The optimum for the reaction was pH 4.3 (Fig. 10B). DISCUSSION Hexosaminidase S (Hex S, ␣␣) is a labile lysosomal enzyme that occurs in much smaller quantities in normal human tissues than the major isoenzymes Hex A (␣␤) and Hex B (␤␤). Since rather low specific activities have been reported for purified Hex S (3-5) its physiological significance has been questioned (2). The striking accumulation of anionic oligosaccharides in the urine of mice deficient in all hexosaminidases compared with those still expressing Hex S (Hexb Ϫ/Ϫ) (8) prompted us to reinvestigate the substrate specificity of Hex S.
For these studies we expressed human Hex S cDNA in cultured insect cells and purified the recombinant Hex S from the secretions. All three N-glycosylation sites were utilized in the recombinant ␣-subunit consistent with previous studies on the enzyme (28). Its disulfide linkage pattern, as analyzed by mass spectrometry after proteolytic cleavage, corresponded to that of the homologous hexosaminidase ␤-subunit (15). rHexS hydrolyzed the synthetic anionic substrate MUGS at even higher rates than the other isozymes and was as active as Hex A on the neutral substrate MUG (Table II). Interestingly, the precursor form of rHexS showed only 2-3% of the specific molecular activity of processed, mature rHexS.
To identify physiological substrates of Hex S anionic and neutral glycans were isolated from the hexosaminidase double knockout mice (Hexa Ϫ/Ϫ, Hexb Ϫ/Ϫ) and characterized. Analysis by ESI-MS/MS combined with glycosidase digestion indicated that the isolated GAGs were derived from dermatan sulfate. In addition, elevated levels of neutral N-glycan fragments were found in the urine of the double knockout mice. This was expected from their occurrence in the urine of Hexb Ϫ/Ϫ mice. An anionic (DS-5) and a neutral oligosaccharide (H 3 N 3 ) were isolated from the urine of the double knockout mice, and used as physiological substrates for rHexS. As an additional putative substrate the sulfated sphingolipid SM2 was isolated from rat kidney. rHexS was as active as Hex A on the anionic bis-sulfated glycans, the trisaccharide C6S-3, and the pentasaccharide DS-5, and the sulfated glycolipid SM2. Each of these substrates was refractory to the action of Hex B (Table III). From these anionic substrates rHexS released N-acetylgalactosamine from the nonreducing end and, in the case of C6S-3, N-acetylgalactosamine 6-sulfate, thereby bypassing the exo-sulfatase step. The release of GalNAc-6-sulfate from keratan sulfate oligosaccharides by Hex A has been reported (37) and it was postulated that hexosaminidases are able to release sulfated hexosamines in vivo from the observation that 6-and 4-sulfated hexosamines are excreted with the urine of patients deficient in GAG specific sulfatase activity (38). However, if the terminal N-acetylgalactosamine residue is sulfated in the 3-position, as in the glycolipid SB2, it is resistant against the action of all three hexosaminidases A, B, and S. As with the degradation of other anionic and neutral sphingolipids (39 -43), the hydrolysis of the membrane-bound sulfated glycosphingolipid SM2 by rHexS is strongly dependent on both, an activator protein, in this case GM2AP, and the presence of an additional anionic phospholipid such as BMP in the substrate carrying membrane.
As observed by quantitative ESI-MS/MS analysis elevated levels of SM2 were found in the kidney tissues of both Hexa Ϫ/Ϫ mice and double knockout mice, both lacking Hex S and Hex A activities. 2 This suggests an important role of Hex S in the catabolism of SM2, a potentially important glycosphingolipid. High levels of SM2 have been observed in the human renal cancer cell line SMKT-R3 (44) as well as in a uterine endometrial adenocarcinoma (45). Moreover, SM2 and the bis-2 R. Sandhoff, manuscript in preparation. Physiological Substrates for Human Lysosomal ␤-Hex S sulfated SB2 have been identified as the most potent glycosphingolipid ligands for NKR-P1, a membrane protein on natural killer cells that contains an extracellular C type lectin domain, and their contribution to the activation of natural killer cells has been proposed (46). Comparing the hexosaminidase concentrations and incubation times used in vitro, anionic glycolipids were degraded much faster than anionic oligosaccharides. Rough calculations suggest that the sulfated GSL SM2 is degraded 500 times faster than the sulfated pentasaccharide DS-5 ( Fig. 7B and 9A). Slow degradation rates have also been described for digestion of oligosaccharides by other glycosidases, e.g. hyaluronidase and ␤-glucuronidase (34). Presumably, the suitable substrate conformation and the accessibility of the sensitive linkages for the enzyme is less often achieved in a freely movable, watersoluble oligosaccharide than in a glycolipid which may be conformatively fixed in an activator-lipid complex.
The results obtained with Hex A and rHexS expand the array of physiological glycoconjugate substrates acted upon by the lysosomal hexosaminidases. They also add support to the concept that Hex S catalyzes an important set of degradative reactions in vivo. Finally, the unique substrate specificity of Hex S make the enzyme an interesting and valuable research tool in glycobiology.