Association of N-Acetylgalactosamine-6-sulfate Sulfatase with the Multienzyme Lysosomal Complex of β-Galactosidase, Cathepsin A, and Neuraminidase POSSIBLE IMPLICATION FOR INTRALYSOSOMAL CATABOLISM OF KERATAN SULFATE

Abstract N-Acetylgalactosamine-6-sulfate sulfatase (GALNS) catalyzes the first step of intralysosomal keratan sulfate (KS) catabolism. In Morquio type A syndrome GALNS deficiency causes the accumulation of KS in tissues and results in generalized skeletal dysplasia in affected patients. We show that in normal cells GALNS is in a 1.27-MDa complex with three other lysosomal hydrolases: β-galactosidase, α-neuraminidase, and cathepsin A (protective protein). GALNS copurifies with the complex by different chromatography techniques: affinity chromatography on both cathepsin A-binding and β-galactosidase-binding columns, gel filtration, and chromatofocusing. Anti-human cathepsin A rabbit antiserum coprecipitates GALNS together with cathepsin A, β-galactosidase, and α-neuraminidase in both a purified preparation of the 1.27-MDa complex and crude glycoprotein fraction from human placenta extract. Gel filtration analysis of fibroblast extracts of patients deficient in either β-galactosidase (β-galactosidosis) or cathepsin A (galactosialidosis), which accumulate KS, demonstrates that the 1.27-MDa complex is disrupted and that GALNS is present only in free homodimeric form. The GALNS activity and cross-reacting material are reduced in the fibroblasts of patients affected with galactosialidosis, indicating that the complex with cathepsin A may protect GALNS in the lysosome. We suggest that the 1.27-MDa complex of lysosomal hydrolases is essential for KS catabolism and that the disruption of this complex may be responsible for the KS accumulation in β-galactosidosis and galactosialidosis patients.

Recent studies indicated the importance of the supramolecular association of enzymes catalyzing subsequent metabolic reactions (5,6), which may lead to an acceleration of the whole process resulting from the directed transport of intermediates. In lysosomes GAL and ␣-neuraminidase (NEUR, EC 3.2.1.18), the enzymes catalyzing the two first steps of ganglioside hydrolysis are associated with cathepsin A, also named protective protein (CathA, E.C. 3.4.16.1), into a lysosomal multienzyme complex necessary for both the activation of NEUR and stabilization of GAL within the lysosome (7)(8)(9)(10). Recently we proposed the molecular mechanism of GAL protection by CathA and demonstrated that the complex purified from human placenta has a molecular mass of about 1.27 MDa and contains multiple protein components (11,12). In the present work we demonstrate that GALNS is the component of a 1.27-MDa complex, that this association protects GALNS in lysosome, and we explore the hypothesis that the supramolecular organization of lysosomal hydrolases is important for the catabolism of KS.

MATERIALS AND METHODS
Purification of Lysosomal High Molecular Weight Complex-The 1.27-MDa lysosomal complex was purified from human placenta using affinity chromatography on a concanavalin A-Sepharose column followed by affinity chromatography on a p-aminophenyl ␤-D-thiogalactopyranoside (PATGAL)-agarose column, which binds GAL, or on a Phe-Leu-agarose column, which binds CathA (12,13). After affinity purification the preparations were dialyzed against 20 mM sodium acetate buffer, pH 5.2, containing 0.15 M NaCl and 0.02% (w/v) NaN 3 (buffer A), concentrated to ϳ3 mg of protein/ml, and applied to a FPLC Superose 6 column (Pharmacia Biotech Inc.), eluted with buffer A at a flow rate of 0.4 ml/min. 0.5-ml fractions were collected and analyzed for NEUR, GAL, CathA, and GALNS activities as well as by SDS-polyacrylamide gel electrophoresis (PAGE) as described below. Fractions from the peak, corresponding to a molecular mass of 1.27 MDa and containing NEUR, GAL, and CathA activities, were pooled, dialyzed against 25 mM bis-Tris (Sigma Chemical Co.) buffer, pH 6.3, concentrated to 1 ml, and applied to a FPLC Mono P column equilibrated with the same buffer. The column was eluted at a flow rate of 1 ml/min with Polybuffer PB 74 (Pharmacia), previously diluted 10-fold in water and adjusted to pH 4.0 with 1 M HCl in accordance with the manufacturer's protocol. Fractions containing NEUR, GAL, and CathA activities were pooled, dialyzed against buffer A, and concentrated to 0.5 ml. The preparation of purified 1.27-MDa complex was frozen and stored at Ϫ70°C until used.
Enzyme Assays-GAL, NEUR, and HEXA were assayed using the corresponding fluorogenic 4-methylumbelliferyl (Muf)-glycoside derivatives as substrates according to published methods (14 -16). CathA activity was measured using CBZ-Phe-Leu as a substrate (10). GALNS was assayed either with the tritiated substrate 6-sulfo-N-acetylgalactosamine-glucuronic acid-6-sulfo-N-acetyl-[1-3 H]galactosaminitol (Gal-NAc-6S, HSC Research and Development Ltd., Toronto), as described (17) or with the fluorogenic substrate 4-methylumbelliferyl-␤-D-galactopyranoside 6-sulfate (Muf-Gal-6S, Moscerdam, Amsterdam) using the following procedure. The reaction mixture containing 50 l of the cellular homogenate, 25 l of 1 M sodium acetate buffer, and 25 l of 1 mM Muf-Gal-6S in the same buffer was incubated overnight at 37°C. The GALNS activity was then inhibited by addition of 25 l of 2 M sodium phosphate buffer, pH 4.75, and the mixture was supplemented with 0.5 g (15-20 milliunits of GAL activity) of purified 680-kDa GAL⅐CathA lysosomal complex from human placenta (13). This reaction mixture was incubated further for 2 h at 37°C to convert all the Muf-Gal liberated during the first incubation into Muf (18). The reaction was terminated by the addition of 1.85 ml of 0.4 M glycine buffer, pH 10.5, and the concentration of Muf was assayed fluorometrically. One unit of enzyme activity is defined as the conversion of 1 mol of substrate/min. Proteins were assayed according to Bradford (19) with BSA (Sigma) as standard.
SDS-PAGE-The electrophoretic analysis of proteins was performed on SDS-polyacrylamide gel (11%, w/v) under reducing conditions according to the method of Laemmli (20). To increase the sensitivity, the proteins were first stained with Coomassie Blue R-250 (Sigma) and then using the Fast Silver Staining Kit (Bio-Rad). The ratio between the proteins in the complex was estimated by scanning of the stained gels with an Ultroscan XL laser densitometer (LKB). The linearity of the microdensitometer response was established up to 20 g of protein loaded on the gel. NH 2 -terminal Amino Acid Sequencing of Proteins-After electrophoretic separation, the proteins were electrotransferred to Immobilon-P membrane (Millipore) in a Trans-Blot Cell (Bio-Rad) for 2.5 h at 4°C in 10 mM CAPS buffer, pH 11, 10% (v/v) methanol at a current of 480 mA. After staining with Coomassie Blue R-250 the bands were cut off, and amino acid sequences were determined in an Applied Biosystem 470 A gas phase Sequencer and identified using the BLAST network service at the National Center of Biotechnology Information (Bethesda, MD).
Antibodies-Rabbit polyclonal antibodies against purified human CathA were prepared as described previously (12). Rabbit polyclonal antiserum against human HEXA was a generous gift of Dr. Roy A. Gravel (McGill University, Montreal).
The antibodies against human GALNS were prepared as follows. The preparation of 1.27-MDa lysosomal complex purified as described above was dialyzed against 10 mM Tris-HCl buffer, pH 7.5 (buffer B), and applied to a FPLC Mono Q column (Pharmacia). The column was eluted with linear 0 -0.4 M gradient of NaCl in buffer B at a flow rate of 0.5 ml/min. The fractions containing GALNS activity were pooled, concentrated to the volume of 0.2 ml, and applied to a Superose 12 column (Pharmacia), eluted with phosphate-buffered saline at a flow rate of 0.4 ml/min. Purified GALNS preparation (45 g), homogeneous by SDS-PAGE analysis, was emulsified with Freund's complete adjuvant and used to immunize a rabbit by subcutaneous injections. 10 and 20 days after the first immunization the rabbit was boosted with another 45 g of the same GALNS preparation in incomplete Freund's adjuvant. The antiserum was used for Western blotting in a dilution of 1:5,000.
Immunotitration-10 g of 1.27-MDa complex purified by FPLC chromatofocusing or 350 g of glycoprotein fraction purified from human placenta by concanavalin A-Sepharose chromatography (12) in 100 l of buffer A was added to 100 l of 0.15 M sodium phosphate buffer, pH 6.0, containing 1% (w/v) BSA. Increasing amounts (1-50 l) of antihuman CathA, anti-HEXA antiserum, or preimmune rabbit serum were added to this mixture. After 1 h of incubation at 4°C, 200 l of Pansorbin cells (Calbiochem) was added. The samples were incubated for 2 h at 4°C with constant shaking and then centrifuged at 5,000 ϫ g for 5 min; the supernatants were assayed for GAL, CathA, NEUR, GALNS, and HEXA activities as described.
Western Blotting of GALNS in Fibroblast Homogenates-Cultured skin fibroblast homogenates (50 g of protein) were subjected to SDS-PAGE and electrotransferred to Immobilon-P membranes as described above. To visualize protein bands, the membrane was stained with 0.1% (w/v) Ponceau S (Sigma) in 5% (v/v) acetic acid. The dye was then washed completely with 5% (v/v) acetic acid, and nonspecific binding sites were blocked by the incubation at 37°C for 2 h in phosphatebuffered saline, 0.05% (w/v) MgCl 2 , and 3% (w/v) BSA. The membrane was then incubated at 4°C for 1 h with the anti-GALNS antiserum (1:5,000 dilution) in phosphate-buffered saline, 0.05% (w/v) MgCl 2 , and 3% (w/v) BSA, washed with phosphate-buffered saline containing 0.2% (v/v) Tween 30, and incubated with an anti-rabbit IgG antibodies conjugated to horseradish peroxidase (Sigma). After washing as above, the GALNS-antibody complex was detected with TMB Blot and Slide reagent (Life Technologies, Inc.) using the manufacturer's protocol.
Cultured Skin Fibroblasts-The fibroblast lines from normal individuals and from patients affected with galactosialidosis, ␤-galactosidosis (both G M1 -gangliosidosis and Morquio type B), sialidosis, and Morquio type A were obtained from the NIGMS Human Genetic Mutant Cell Repository (lines GM02921, GM02685, GM05653, GM03251, GM01602, GM01259, GM00958, GM00593, GM00806, GM05076, and GM02438A) and from our cell bank (lines 89.40.11 and 83.29.34). The cells were cultured in Eagle's minimal essential medium (Mediatech, Washington, D. C.), supplemented with 20% (v/v) fetal calf serum and antibiotics in 150-cm 2 Falcon flasks up to confluence (2-4 mg of protein/flask). The cells from one flask were harvested by trypsinization, rinsed with saline, sonicated in 1-2 ml of distilled water (5 s at 60 watts, twice), and assayed for enzyme activities as described above.
Correction of GALNS Activity in Galactosialidosis Fibroblasts-Four days prior to the enzyme assays the purified placental 680-kDa GAL⅐CathA complex or CathA (12,13) was added to the culture medium in a final concentration of 50 g/ml. In another set of experiments, leupeptin was added to the medium at a final concentration of 20 M to inhibit intralysosomal proteolytic degradation. The cells were harvested, and the enzyme activities were determined as described above.

RESULTS AND DISCUSSION
We purified the 1.27-MDa GAL⅐CathA⅐NEUR complex from human placenta using two different affinity chromatography methods: a Phe-Leu-agarose column for CathA binding and a PATGAL-agarose column for GAL binding (12,13). The analysis of the affinity-purified preparations by FPLC gel filtration (Fig. 1, left panels) demonstrated that both contain two oligomeric forms: a 1.27-MDa GAL⅐CathA⅐NEUR complex (peak I in Fig. 1, a and b), which contains all NEUR and about 2% of total GAL and CathA activities in human placenta extract, and a 680-kDa GAL⅐CathA complex (peak II in Fig. 1, a and b), which contains most of GAL and 30% of total CathA. The remaining 70% of CathA in tissue is not associated with GAL and therefore is not retained on PATGAL-agarose affinity column. Free CathA is present in the Phe-Leu-agarose-purified preparation in the form of a 110-kDa dimer (peak III of the gel filtration profile in Fig. 1b).
The protein composition of the 1.27-MDa complex was studied by SDS-PAGE analysis (Fig. 1, a and b, right panels) and NH 2 -terminal amino acid sequencing. In both Phe-Leu-agarose and PATGAL-agarose-purified preparations of the 1.27-MDa lysosomal complex we found three new protein components of 46, 40, and 35 kDa, respectively, in addition to the previously described (7) 64-kDa subunit of GAL (GAL64 in Fig. 1), and two 32-and 20-kDa subunits of CathA (CathA32 and CathA20 in Fig. 1). A 46-kDa protein had the NH 2 -terminal sequence (QRMFEIDYSRD) of human GAL (21) and probably represents the recently reported 46-kDa product of GAL intralysosomal proteolytic digestion (22). The NH 2 -terminal residue of the 35-kDa-33-kDa doublet protein (P35 in Fig. 1) was probably blocked, since no amino acid sequence was obtained. Molecular mass and NH 2 -terminal amino acid sequence (APQPPNI-LLLLMDDM) of the 40-kDa protein were identical to those (23) of the heavy subunit of human GALNS. We estimated from SDS-PAGE analysis that GALNS represents about 7-10% of the total protein of the 1.27-MDa complex. The relative amount of GALNS present in the preparation did not depend on the purification method, which suggests that GALNS binds to the complex specifically. In agreement with our data it has been reported that a 40-kDa protein with an NH 2 -terminal sequence homologous to that of human GALNS copurifies with GAL⅐NEUR⅐CathA complex from bovine testis (24).
GALNS activity assay in FPLC gel filtration fractions demonstrated that 50% of total GALNS in human placenta was copurified with the 1.27-MDa complex (peak I in Fig. 1, a and  b). The other 50% of GALNS activity eluted as a second peak (peak III in Fig. 1, a and b), consistent with the reported 120-kDa homodimeric form of this enzyme (25).
Purified by gel filtration 1.27-MDa complex was subjected to an additional purification step, FPLC chromatofocusing (Fig.  2). GALNS, GAL, CathA, and NEUR coeluted from the column at a pH 4.8 that corresponds to the pI of the whole complex. Under these conditions, only 5-10% of GALNS and CathA are unbound as estimated from the assay of their activities in the fractions eluting at a pH 6.0 and 5.4, corresponding to the pI values of free GALNS and CathA, respectively.
The presence of GALNS within the 1.27-MDa complex was also confirmed by immunoprecipitation of the purified 1.27-MDa complex using antibodies against human CathA. The titration curves (Fig. 3a) for GALNS, CathA, GAL, and NEUR showed a proportional decrease of the activity of each of the four enzymes. Similar immunotitration curves were obtained with the crude glycoprotein fraction purified by concanavalin A-Sepharose affinity chromatography from human placenta extract (Fig. 3b). However, in this case about 40% of total GALNS activity in the glycoprotein fraction could not be precipitated by anti-CathA antibodies and probably represents the pool of free 120-kDa GALNS homodimers, which exists in equi-librium with the 1.27-MDa complex (Fig. 1). However, in vivo the ratio between the complex-bound and free GALNS could be sufficiently shifted toward the complex formation because of the extremely high protein concentration inside the lysosome, up to 100 mg of protein/ml (12). The activity of HEXA that is not part of the complex was unchanged by anti-CathA antibodies (Fig. 3b), and anti-human HEXA antibodies did not precipitate GALNS, CathA, GAL, and NEUR activities in both purified 1.27-MDa complex and crude glycoprotein fraction (Fig. 3,  c and d), suggesting that the immunoprecipitation of the 1.27-MDa complex by anti-CathA antibodies is specific.
In the lysosome GALNS and GAL catalyze the subsequent reactions of KS hydrolysis; therefore, transfer of the hydrolysis intermediates between the active sites of these enzymes in the 1.27-MDa complex may be an important step in the KS catabolic process. To explore the hypothesis that the association of GALNS with the complex is necessary for the proper intralysosomal KS catabolism we verified whether the complex exists in the cells of patients which accumulate KS. Two autosomal recessive diseases in which KS accumulates in abnormal amounts are known, both of which are associated with a characteristic skeletal dysplasia and corneal clouding in affected patients. The first, Morquio type A syndrome (MPS IV type A), is caused by the primary deficiency of GALNS (2)(3)(4). The second, Morquio type B syndrome (MPS IV type B) (26), clinically similar to MPS IV type A, results from GAL deficiency and is also named ␤-galactosidosis (27). Interestingly, a different set of mutations in GAL results in another clinical condition of ␤-galactosidosis, G M1 -gangliosidosis, in which lipid substrate of GAL, G M1 -ganglioside, accumulates in brain and viscera, resulting in mental retardation and hepatosplenomegaly. KS accumulation, which occurs in both clinical forms of ␤-galactosidosis (27), is presently explained by the primary deficiency of GAL. However, this hypothesis still does not explain the accumulation of both undersulfated and normally sulfated forms of KS, since normal or low normal levels of GALNS activity are detectable in cultured cells from patients (26,27).
Using FPLC gel filtration on Superose 6 column we analyzed fibroblast extracts of ␤-galactosidosis patients (both G M1 -gangliosidosis and MPS IV type B clinical forms) as well as extracts of control fibroblasts. About 30 -40% of the GALNS activity and all of the NEUR activity of normal fibroblasts (Fig.  4a) are associated with the 1.27-MDa complex (peak I). Under these conditions, the 1.27-MDa complex exists in equilibrium with a 680-kDa GAL⅐CathA binary complex (peak II), GAL tetramers (peak III), and CathA and GALNS dimers (peak IV). This distribution of oligomeric forms is similar to that reported previously for human placenta (12). In contrast, in fibroblasts from ␤-galactosidosis patients (MPS IV type B) the 1.27-MDa complex and 680-kDa GAL⅐CathA complex are not detectable, and GALNS is present only in the 120-kDa homodimeric form (Fig. 4b, peak IV). The elution profiles of the enzyme activities of fibroblast extracts from G M1 -gangliosidosis patients (not shown) were similar to those of MPS IV type B fibroblasts.
Using the same method, we also analyzed fibroblast extracts from CathA-deficient (galactosialidosis) and NEUR-deficient (sialidosis) patients because in both diseases a deficiency of a single component of the 1.27-MDa complex could result in the disruption of the complex. Gel filtration of galactosialidosis fibroblast extracts (Fig. 4c) demonstrated the absence of the 1.27-MDa complex (previously (8) also shown by the ultracentrifugation method) suggesting that patients with galactosialidosis may excrete higher than normal amounts of KS in urine, similar to those described for MPS IV type A and ␤-galactosidosis patients. Indeed, the KS urinary excretion of a 16-year-old female galactosialidosis patient was elevated at the level of 2.1 mg/24 h, similar to that of MPS IV type A patients (1.8 -2.4 mg/24 h; normal Ͻ 0.8 mg/24 h). 2 On the other hand, in fibroblasts of NEUR-deficient patients, which do not accumulate KS (28), the 1.27-MDa complex was present, and the amount of complex-bound GALNS corresponded to those in the control fibroblasts (Fig. 4d, peak I). These results also suggest that the presence of NEUR is not essential for the association of the other enzymes in the 1.27-MDa complex.
Interestingly, the disruption of the 1.27-MDa complex in galactosialidosis is also associated with a decreased level of GALNS activity (Fig. 4c). Further assay of GALNS activity in three other fibroblast lines of galactosialidosis patients (Table  FIG.  I) showed that the activity is reduced to 3-21% of normal. Of note, in MPS IV type A fibroblasts, used as a negative control, GALNS residual activity did not exceed 1-2% of normal (Table  I). We found that the activity of GALNS, as well as that of GAL and NEUR, could be restored in galactosialidosis fibroblasts (cell line 83.29.34) by the addition of purified CathA or GAL⅐CathA 680-kDa complex to the culture medium (Table II). GALNS and GAL activities were also increased 3-4-fold by leupeptin inhibition of lysosomal proteases (Table II). These data are consistent with the rapid proteolytic degradation of unbound GALNS within lysosomes similar to that demonstrated for GAL (7,8). Indeed, the reduced amount of GALNS cross-reacting material was detected by Western blotting in fibroblast homogenate of galactosialidosis patient (Fig. 5 On the other hand, GALNS activity is low normal (ϳ50% of control), and NEUR activity is normal in ␤-galactosidosis fibroblasts (Table I), in which the complex does not exist. Gel filtra-tion analysis of ␤-galactosidosis fibroblast extracts (Fig. 4b) demonstrated that NEUR is present as large (Ͼ3 MDa) aggregates with CathA which elute in the void volume (V 0 ) of the column (Fig. 4b, peak V). As was demonstrated previously (9), the association with CathA only is sufficient for the expression of NEUR activity. It is tempting to speculate that a similar protective mechanism may also apply to GALNS in ␤-galactosidosis cells, although a GALNS⅐CathA binary complex has not yet been directly demonstrated. The normal level of GALNS activity in sialidosis fibroblasts (Table I) is in agreement with the presence of the 1.27-MDa complex in these cells (Fig. 4d, peak I).
Altogether, our results demonstrate that the majority of GALNS in the lysosome is associated with the 1.27-MDa multienzyme complex, which also includes GAL, CathA, and NEUR. The complex is not present in the cells of ␤-galactosidosis and galactosialidosis patients which accumulate KS. We suggest that 1.27-MDa multienzyme complex is necessary for proper KS catabolism, which can reflect the importance of supramolecular structure for the hydrolytic function of the lysosome.