Purification and Characterization of Recombinant ... - Science Direct

7 downloads 0 Views 182KB Size Report
by Chinese Hamster Ovary Cells. Ke-Wei Zhao and Elizabeth F. Neufeld1 .... Fluorescence was measured with a Farrand Ratio-2 system filter flu- orometer at an ...
Protein Expression and Purification 19, 202–211 (2000) doi:10.1006/prep.2000.1230, available online at http://www.idealibrary.com on

Purification and Characterization of Recombinant Human ␣-N-Acetylglucosaminidase Secreted by Chinese Hamster Ovary Cells Ke-Wei Zhao and Elizabeth F. Neufeld 1 Department of Biological Chemistry, Brain Research Institute and Molecular Biology Institute, University of California, Los Angeles, California 90095-1737

Received February 4, 2000

␣-N-Acetylglucosaminidase (EC 3.2.1.50) is a lysosomal enzyme that is deficient in the genetic disorder Sanfilippo syndrome type B. To study the human enzyme, we expressed its cDNA in Lec1 mutant Chinese hamster ovary (CHO) cells, which do not synthesize complex oligosaccharides. The enzyme was purified to apparent homogeneity from culture medium by chromatography on concanavalin A–Sepharose, Poros 20 – heparin, and aminooctyl–agarose. The purified enzyme migrated as a single band of 83 kDa on SDS– PAGE and as two peaks corresponding to monomeric and dimeric forms on Sephacryl-300. It had an apparent K m of 0.22 mM toward 4-methylumbelliferyl-␣-Nacetylglucosaminide and was competitively inhibited by two potential transition analogs, 2-acetamido-1,2dideoxynojirimycin (K i ⴝ 0.45 ␮M) and 6-acetamido-6deoxycastanospermine (K i ⴝ 0.087 ␮M). Activity was also inhibited by mercurials but not by N-ethylmaleimide or iodoacetamide, suggesting the presence of essential sulfhydryl residues that are buried. The purified enzyme preparation corrected the abnormal [ 35S]glycosaminoglycan catabolism of Sanfilippo B fibroblasts in a mannose 6-phosphate-inhibitable manner, but its effectiveness was surprisingly low. Metabolic labeling experiments showed that the recombinant ␣-N-acetylglucosaminidase secreted by CHO cells had only a trace of mannose 6-phosphate, probably derived from contaminating endogenous CHO enzyme. This contrasts with the presence of mannose 6-phosphate on naturally occurring ␣-N-acetylglucosaminidase secreted by diploid human fibroblasts and on recombinant human ␣-L-iduronidase secreted by the same CHO cells. Thus contrary to cur1 To whom correspondence should be addressed at Department of Biological Chemistry, UCLA School of Medicine, 33-257 CHS, 10833 Le Conte Avenue, Los Angeles, CA 90095-1737. Fax: (310) 206-1929. Email: [email protected].

202

rent belief, overexpressing CHO cells do not necessarily secrete recombinant lysosomal enzyme with the mannose 6-phosphate-targeting signal; this finding has implications for the preparation of such enzymes for therapeutic purposes. © 2000 Academic Press

␣-N-Acetylglucosaminidase (EC 3.2.1.50) is a glycosidase required for the stepwise degradation of heparan sulfate within lysosomes. Deficiency of ␣-N-acetylglucosaminidase activity causes Sanfilippo syndrome type B (mucopolysaccharidosis III B), a disorder characterized biochemically by lysosomal storage of heparan sulfate and clinically by profound neurological deterioration with behavioral disturbances but relatively mild somatic manifestations (1–3). Patients generally survive only to their late teens, though some who have an attenuated form of the disease may live well into adulthood. The disease shows extensive molecular heterogeneity, with over 70 different mutations identified to date (4 –9). A mouse model of the Sanfilippo syndrome type B has been generated by homologous recombination (10). ␣-N-Acetylglucosaminidase has been purified from several human (11–14) and animal (4,15) sources, but its low abundance precluded extensive studies. The recent cloning of the human complementary and genomic DNA (4,14,16) has made possible the production of recombinant human enzyme. Chinese hamster ovary (CHO) 2 cells have been found very useful for the preparation of recombinant lysosomal enzymes. As first shown for ␣-galactosidase (17), stably transfected CHO cells selectively secrete the overexpressed lysosomal enzymes in substantial amounts, permitting easy purification from the me2 Abbreviations used: CHO, Chinese hamster ovary; endo-H, endo␤-N-acetylglucosaminidase-H; HE, heparin.

1046-5928/00 $35.00 Copyright © 2000 by Academic Press All rights of reproduction in any form reserved.

RECOMBINANT HUMAN ␣-N-ACETYLGLUCOSAMINIDASE

dium. The secreted recombinant lysosomal enzymes examined to date have all been shown to have the mannose 6-phosphate-targeting signal—␣-galactosidase (17), ␣-L-iduronidase (18 –20), iduronate sulfatase (21), N-acetylgalactosamine 6-sulfatase (22), cathepsin D and uteroferrin (23), heparan sulfamidase (24), and ␣-glucosidase (25,26). The mannose 6-phosphate signal facilitates endocytosis into a variety of cells and makes these enzymes candidates for enzyme replacement therapy of the respective deficiency diseases; for example, recombinant human ␣-L-iduronidase secreted by CHO cells has been shown effective in treatment of mucopolysaccharidosis I in dogs (18,27) and is currently in clinical trial (28). Macrophages represent an exception to the generality of the mannose 6-phosphate receptor, as these cells have receptors that recognize unphosphorylated mannose residues (29). We have used CHO cells for expression of human recombinant ␣-N-acetylglucosaminidase and present the purification of the enzyme and its characterization with respect to enzymatic and targeting properties. METHODS

Reagents, Supplies, and Cell Lines Concanavalin A–Sepharose 4B, aminooctyl–agarose, affinity chromatography medium (kit), protein standards for size-exclusion chromatography, phosphatedeficient and methionine/cystine-deficient DMEM, and other materials were from Sigma Chemical Company unless otherwise indicated. Poros 20 – heparin (HE) was from Perkin–Elmer/Perseptive Biosystems. Sephacryl S-300 was from Pharmacia. AP1 glass columns were from Waters. Centricon-100, Polycap 75 SPF prefilter, and cell culture plates were from Fisher Scientific. Fetal bovine serum was from Irvine Scientific. General cell culture and labeling media, supplements, Lipofectamine Plus, and BenchMark prestained protein standards for SDS–PAGE were from Gibco/Life Technologies, Inc. The Lec1 CHO cell line was from the American Type Culture Collection. The stably transfected CHO line used for synthesis of recombinant human ␣-L-iduronidase, 2.131, has been previously described (18). The normal human diploid fibroblast lines GM 4390A, GM 3440B, and IMR 90 were from the Human Genetic Mutant Cell Repository, Coriell Institute for Medical Research, and the fibroblast line from a Sanfilippo B patient, IT 424 (5), was from Dr. P. D. Natale. The radiolabeled compounds— 32 PO 4, H 2 35SO 4, and Expre 35s 35s 35S-protein labeling mix containing [ 35S]methionine and [ 35S]cystine—were from NEN Life Science Products. Endo-␤-N-acetylglucosaminidase (endo-H f) was from New England Biolabs. Pansorbin, G-418 (geneticin), and 4-methylumbelliferyl-␣-N-acetylglucosaminide were from Calbiochem. The plasmid pVGRXR and zeocin were from

203

InVitrogen. Methyl-␣-N-acetylglucosaminide, benzyl␣-N-acetylglucosaminide, and 2-acetamido-2-deoxyglucono-1,5-lactone were from Toronto Research Chemicals, Inc. 2-Acetamido-1, 2-dideoxynojirimycin, 6-acetamido-6-deoxycastanospermine, and Nagstatin were kindly provided by Drs. G. Legler, M. S. Kang, and T. Aoyagi, respectively. Determination of Enzyme Activity

␣-N-Acetylglucosaminidase was assayed by the method of Chow and Weissmann (30). In a standard assay, the sample (25 ␮l) was mixed with 25 ␮l 0.2 mM 4-methylumbelliferyl-␣-N-acetylglucosaminide in 0.1 M Na–acetate buffer, pH 4.3, containing 0.5 mg/ml bovine serum albumin. After incubation for up to 60 min at 37°C, the reaction was stopped by addition of 1 ml 0.5 M glycine–NaOH, pH 10.5. Fluorescence was measured with a Farrand Ratio-2 system filter fluorometer at an excitation wavelength of 360 nm and an emission wavelength of 450 nm. For kinetics studies, the reaction was performed in 96-well microplates (FluoroNunc) and 0.2 ml of 0.5 M glycine–NaOH, pH 10.5, was added to stop the reaction; fluorescence was measured with an HTS 7000 Plus BioAssay Reader (Perkin–Elmer). A unit of ␣-N-acetylglucosaminidase activity corresponds to release of one nmol of 4-methylumbelliferone per hour. Other lysosomal glycosidases were assayed in 96well plates under conditions similar to the ones used for ␣-N-acetylglucosaminidase. The 4-methylumbelliferyl glycoside substrates were used in all cases at the concentration of 0.10 mM (␤-glucuronide, ␣-N-acetylneuraminide) or 0.12 mM (␣-glucoside, ␤-galactoside, and ␤-N-acetylglucosaminide), in 0.1 M Na–acetate buffer, pH 4.3, for 1 h at 37°C; ␣-L-iduronidase was assayed with 0.025 mM 4-methylumbelliferyl-␣-L-iduronide in 0.2 M Na–formate buffer, pH 3.5, for 1 h at ambient temperature. Expression of Human ␣-N-Acetylglucosaminidase A human ␣-N-acetylglucosaminidase expression vector (pCMV-NAGLU) had been constructed earlier (5). It was electroporated into CHO Lec1 cells using 240 V and 960 ␮F for 10 s. Cell colonies were selected in ␣-MEM supplemented with 5% fetal bovine serum and 0.75 mg/ml G-418. The highest expressing clone, B12, was selected for further study and for production of enzyme. The cells were maintained in ␣-MEM with nucleosides, enriched with 5% fetal bovine serum, nonessential amino acids, pyruvate, penicillin, streptomycin, and, for vector-containing cells, 0.1 mg/ml G-418. For enzyme preparation, G-418 and nucleosides were omitted. Conditioned medium, collected every 2 to 3 days, was stored at 4°C. Methyl-␣-glucoside was added to a final concentration of 10 mM. The CHO line 2.131,

204

ZHAO AND NEUFELD

already stably transfected with IDUA and neo r (18), was transfected with pCMV-NAGLU and pVgRXR in a ratio of 10:1 using Lipofectamine and selected in the above medium containing 1 mg/ml each of G-418 and zeocin. The resulting clonal line, B29, which expressed large amounts of both ␣-L-iduronidase and ␣-N-acetylglucosaminidase, was maintained in 0.1 mg/ml G-418. The human embryonic kidney cell line 293 was transfected with pCMV-NAGLU using Lipofectamine and was maintained in 1 mg/ml G-418. Purification of ␣-N-Acetylglucosaminidase from Conditioned Medium A three-step purification scheme was established, with the order of the steps chosen to avoid concentration or dialysis of material applied to each column. All solutions were sterilized by passage through 0.2-␮m cellulose acetate filters, and sterile glassware and plasticware were used throughout (excluding chromatography columns). All purification procedures were performed at room temperature except for loading of conditioned medium in Step 1, which was performed at 4°C. Protein profiles were determined at A 280 either online or in fractions, and ␣-N-acetylglucosaminidase activity in fractions was assayed with a 10-min incubation time. Steps 2 and 3 were carried out on a BioCAD protein purification workstation (Perkin– Elmer/Perseptive Biosystems) which records online changes of pH and conductivity in addition to A 280 . Except for step 2 where a linear pH gradient had to be maintained, all buffers contained 10 mM phosphate in order to protect any mannose 6-phosphate groups that might be present on ␣-N-acetylglucosaminidase from phosphatase activity. Step 1: Lectin chromatography. Concanavalin A–Sepharose 4B, packed into a glass column (25 ⫻ 100 mm), was equilibrated with buffer A (10 mM Na–phosphate, pH 6.8, 50 mM NaCl, 10 mM methyl-␣-glucoside, 0.02% NaN 3, and 1 mM ␤-mercaptoethanol). Settled medium was loaded by gravity onto the column through a 1-␮m Polycap 75 SPF prefilter to remove any floating cells or debris. After it was loaded at 4°C, the column was moved to ambient temperature and extensively washed with buffer B (as buffer A but with addition of 10 mM methyl-␣-mannoside, 5 mM EDTA, and pH lowered to 5.8) until A 280 reached zero. Elution of the column was initiated by applying 500 mM methyl-␣-mannoside in buffer B at a rate of 0.5 ml per min, and fractions of 10 ml were collected in polyethylene tubes. After about one bed volume elution, the column was stopped, and elution was resumed after about 6 –12 h. Step 2: Poros 20 –HE perfusion chromatography. Poros 20 –HE resin packed into an AP1 glass column (10 ⫻ 90 mm) was equilibrated with 10 column vol of

buffer C (25 mM Hepes/Mes/acetate buffer, pH 4.5, containing 0.05 M NaCl). Active fractions from the concanavalin A column were pooled, adjusted to pH 4.5– 4.8 with 0.1 N HCl, and directly pump loaded onto the column. The flow rate was maintained at 10 ml/min throughout the process with pressure ranging from 800 to 1200 psi. A maximum pressure setting of 1500 psi was used to protect the column. Fractions of 5 ml were collected in polyethylene tubes containing 0.1 ml 0.5 M NaH 2PO 4. A peak of activity was eluted by a pH gradient from 4.5 to 6.5 in 0.05 M NaCl and in 50 column vol. The gradient was created by blending of 0.1 M Hepes/Mes/NaAc at pH 4.5 and 7.5, with 3 M NaCl and H 2O controlled by the BioCAD software. Proteins remaining on the column after elution with the pH gradient were eluted by a gradient of 0.05 to 2.0 M NaCl in buffer C at pH 7.5 for 5 column vol. Step 3: Hydrophobic interaction chromatography. Aminooctyl–agarose was selected for this step instead of the commonly used phenyl–Sepharose 4B, after preliminary tests using the Sigma affinity chromatography medium kit. Among the kit’s aminoacyl ligands ranging from 2 to 12 carbons, the aminooctyl ligand was found to be of optimal length to bind at high salt concentration and then release the enzyme when salt was withdrawn. Aminooctyl–agarose resin in an AP1 column (10 ⫻ 70 mm) was equilibrated with buffer D (10 mM Na–phosphate, pH 7.2, and 2.5 M NaCl). Active fractions from Poros 20 –HE column were pooled and adjusted to pH 7.2 with 0.1 N NaOH, mixed with an equal volume of 5 M NaCl/20 mM Na–phosphate, pH 7.2, and immediately pump loaded onto the column. The flow rate was maintained at 2 ml/min with pressure at about 500 psi, and fractions of 1.2 ml were collected in polyethylene tubes. After washing with buffer D until A 280 reached zero, the column was eluted by a gradient of 2.5 to 0 M NaCl in 10 mM Na–phosphate, pH 7.2, for 15 column vol. The broad peak of ␣-N-acetylglucosaminidase peak eluted from the aminooctyl–agarose at low salt concentration was pooled and concentrated to ⬃1 mg/ml in Centricon-100 and buffer exchanged to 10 mM Na–phosphate, pH 7.2, containing 150 mM NaCl. The column was regenerated for future use with 5 column vol of 50% acetonitrile in buffer D. Determination of Glycosaminoglycan Accumulation in Fibroblasts When cultured fibroblasts are labeled with H 2 35SO 4, incorporated 35SO 4 remaining after removal of the pericellular layer by trypsin represents newly made glycosaminoglycans stored in the lysosomes. This lysosomal pool was measured by a modification of an earlier method (31). Confluent fibroblasts in six-well plates were labeled for up to 44 h at 37°C, in 1.5 ml sulfate-

RECOMBINANT HUMAN ␣-N-ACETYLGLUCOSAMINIDASE

205

TABLE 1 Purification of ␣-N-Acetylglucosaminidase from Lec1-Conditioned Medium Step

Volume (ml)

Activity (units ⫻ 10 ⫺3)

Protein (mg)

Specific activity (units/mg)

Purification (fold)

Yield (%)

Medium Con A–Sepharose Poros 20–HE Aminooctyl–agarose Concentrate

7,710 830 45 40 0.8

227 256 86 34 23

11,600 147 6.3 1.1 0.66

20 1,740 13,700 31,000 35,000

1 87 680 1,500 1,800

100 113 38 15 10

deficient S-MEM supplemented with 0.02% CaCl 2, pyruvate, and nonessential amino acids. The medium also contained 8% dialyzed fetal bovine serum and 4 ␮Ci/ml H 2 35SO 4. Purified ␣-N-acetylglucosaminidase and potential inhibitors were diluted into the medium for correction and inhibition assays. After labeling, cells were washed three times with phosphate-buffered saline (2 ml) both before and after trypsinization and extracted by three cycles of freezing and thawing in 10 mM Na–phosphate, pH 5.8, containing 0.5% NP-40. After centrifugation at 10,000 rpm ⫻ 10 min, the supernatant fluids were used to determine radioactivity and protein concentration. Antibody Production and Use in Immunoprecipitation About 500 ␮g of the ␣-N-acetylglucosaminidase purified to apparent homogeneity was used to raise polyclonal antibodies in rabbits (Cocalico Biologicals, Inc). The rabbit antisera against ␣-N-acetylglucosaminidase and ␣-L-iduronidase were used essentially as described to precipitate the respective enzymes from metabolically labeled cells and medium (20). RESULTS

Overexpression of ␣-N-Acetylglucosaminidase in Stable Cell Line As described in Materials and Methods, CHO cells were stably transfected with a vector containing NAGLU cDNA under the control of the CMV promoter. The Lec1 mutant CHO cell line chosen for transfection lacks N-acetylglucosaminyltransferase I and therefore does not synthesize complex oligosaccharides (32). The highest expressing clone, B12, contained 940 units/mg protein of intracellular ␣-N-acetylglucosaminidase (corresponding to ca. 26 ␮g of enzyme, see below) and secreted ca. 350 units/24 h/mg cellular protein. The enzyme accumulated in the medium in linear fashion for 3 days, indicating that the enzyme was stable in culture medium. Overexpression of ␣-N-acetylglucosaminidase did not appear to affect intracellular or secreted levels of several other lysosomal enzymes. The ratio of enzyme

activities in homogenates of B12 cell versus the parental Lec1 cells were as follows: ␣-N-acetylglucosaminidase, 370; ␣-L-iduronidase, 0.81; ␣-glucosidase, 0.74; ␣-N-acetylneuraminidase, 0.92; ␤-galactosidase, 0.95; ␤-hexosaminidase, 1.0; and ␤-glucuronidase, 0.21. Secretion of those enzymes that could be reliably measured in the medium (␤-galactosidase, ␤-hexosaminidase, and ␤-glucuronidase) was not changed from the parental Lec1 cells, whereas ␣-N-acetylglucosaminidase secretion was increased 650-fold. Purification of Recombinant ␣-NAcetylglucosaminidase The purification of ␣-N-acetylglucosaminidase from medium conditioned by B12 cells is described under Materials and Methods and is summarized in Table 1. The use of the lec1 cell line, which makes glycoproteins with exposed mannose residues, allowed an easy separation from bovine serum proteins by chromatography on concanavalin A–Sepharose. The specific activity of the most purified ␣-N-acetylglucosaminidase was 35,000 units/mg using the assay of Chow and Weissmann (30). This corresponds to ca. 190,000 nmol/h/mg protein using the assay of Marsh and Fensom (33), which calls for 10 times as much substrate. The specific activity of the purified recombinant enzyme is six times higher than the specific activity (500 nmol/min/mg protein, Marsh and Fensom assay) reported for human placental enzyme that had been purified to apparent homogeneity (14). Characterization of Purified Recombinant Human ␣-N-Acetylglucosaminidase Size-exclusion chromatography using Sephacryl S-300 separated ␣-N-acetylglucosaminidase activity into two peaks of about 90 and 180 kDa corresponding to monomer and dimer, in a ratio of 1:2 (Fig. 1A). Upon analysis with SDS–PAGE, the purified ␣-N-acetylglucosaminidase migrated as a single band of ca. 83 kDa (Fig. 1B). In contrast to what had been observed for the enzyme purified from bovine testis (4), the dimeric form of the recombinant human ␣-N-acetylglu-

206

ZHAO AND NEUFELD

FIG. 1. Molecular mass and homogeneity of ␣-N-acetylglucosaminidase. (A) Purified ␣-N-acetylglucosaminidase was subjected to size-exclusion chromatography on a Sephacryl S-300 column (15 ⫻ 910 mm) equilibrated with 10 mM Tris–HCl, pH 7.4, at 4°C, and its elution was monitored by enzyme assay (closed triangles). The column was calibrated with blue dextran (V 0 ) and protein standards (open circles) consisting of apoferritin (443 kDa), ␤-amylase (200 kDa), alcohol dehydrogenase (150 kDa), transferrin (90 kDa), and bovine serum albumin (66 kDa). (B) ␣-N-Acetylglucosaminidase, ca. 4 ␮g, from the last purification step was subjected to PAGE under reducing and denaturing conditions, followed by silver staining. Its molecular size was estimated to be 83 kDa using BenchMark standard proteins.

cosaminidase was not seen in SDS–PAGE conducted without prior heating. Amino-terminal amino acid sequence analysis of the band isolated by SDS–PAGE showed the sequence of predicted from cDNA, 24 DEAREAAAVRALVARLLGP 42, confirming the authenticity of the purified protein. In addition, the sequence confirms earlier determinations of the amino terminus of ␣-N-acetylglucosaminidase purified from human placenta (14) and from human liver, cited in (4), implicating cleavage by signal peptidase upstream of Asp-24. The purified recombinant ␣-N-acetylglucosaminidase was found stable in a broad range of pH, using 20 mM buffers from pH 4.6 (acetate) to pH 8.1 (phosphate). The concentrated ␣-N-acetylglucosaminidase has been kept over 1 year at 4°C in 10 mM Na–phosphate buffer, pH 7.2/150 mM NaCl, without significant loss of activity. However, all activity was lost upon heating at 70°C for 4 min. The purified ␣-N-acetylglucosaminidase had an apparent K m of 0.22 mM toward 4-methylumbelliferyl-␣N-acetylglucosaminide. Inhibition by substrate analogs is shown in Table 2. Two imino sugars, 2-acetamido-1, 2-dideoxynojirimycin and 6-acetamido6-deoxycastanospermine, were found to be the most potent competitive inhibitors with K i of 0.45 and 0.087 ␮M, respectively. 2-Acetamido-2-deoxyglucono-1,5-lactone was much less inhibitory, with a K i of 0.12 mM; two structurally related compounds, deoxynojirimycin

and nagstatin, were not inhibitory at all at 0.25 and 4 mM, respectively. The K i for UDP-N-acetylglucosamine, a known substrate of ␣-N-acetylglucosaminidase (11), was 0.12 mM, and the K i for the product, N-acetylglucosamine, was 2.4 mM (not shown). N-Acetylglucosamine 6-sulfate, N-benzylglucosamine, methyl-␣-N-acetylglucosaminide, and benzyl-␣-N-acetylglucosaminide showed little if any inhibition at 5 mM (not shown). Mercuric ion (1.5 ␮M) and p-chloromercuribenzoate (5 ␮M) inhibited ␣-N-acetylglucosaminidase activity completely. Iodoacetate also inhibited completely at 2.5 mM. However, two other thiol reagents, N-ethylmaleimide and iodoacetamide, had no effect at concentrations up to 5 mM. Targeting Properties of Recombinant ␣-NAcetylglucosaminidase Addition of purified ␣-N-acetylglucosaminidase into the labeling medium reversed [ 35S]glycosaminoglycan accumulation in Sanfilippo B fibroblasts (Fig. 2A). The correction was dose-dependent, with half-maximal effect achieved at 2.5 units of activity per well or 0.6 nM ␣-N-acetylglucosaminidase. Inclusion of mannose 6-phosphate (5 mM) partially inhibited the corrective effect (Fig. 2B), to the extent predicted if the K i were ca. 130 ␮M; glucose 6-phosphate had much less effect, while glucose and mannose did not inhibit at all. These results indicate that some ␣-N-acetylglucosaminidase was taken up by the fibroblasts via the mannose 6-phosphate system. But the concentration required for correction was nearly a thousand times higher than expected on the basis of experience with ␣-L-iduronidase (18). Moreover, attempts to measure uptake of the enzyme directly were not successful with the concentration of enzyme used. This suggested that only a very small fraction of the purified enzyme had the mannose 6-phosphate signal. Biosynthetic Labeling of Newly Synthesized ␣-NAcetylglucosaminidase Because of the possibility that the enzyme had lost its phosphate groups during purification, the presence of mannose 6-phosphate in newly made ␣-N-acetylglucosaminidase was investigated by biosynthetic labeling. A simple diagnostic test for the presence of mannose 6-phosphate is incorporation of radioactive phosphate into immunoprecipitated enzyme and removal of the label by subsequent treatment with endo-H; the protein core, detected by labeling with [ 35S]methionine/cystine, becomes smaller upon treatment with endo-H but does not disappear. The amount of 32PO 4 and [ 35S]methionine/cystine used in the labeling medium was selected on the basis of previous experience with ␣-L-iduronidase (20), to allow compara-

RECOMBINANT HUMAN ␣-N-ACETYLGLUCOSAMINIDASE

207

TABLE 2 Inhibition of ␣-N-Acetylglucosaminidase Activity by Substrate Analogs Compounds

Structure

K i (␮M)

2-Acetamido-1,2-dideoxynojirimycin

0.45

6-Acetamido-6-deoxycastanospermine

0.087

2-Acetamido-2-deoxyglucono-1,5-lactone

120

Deoxynojirimycin

Not inhibitory

Nagstatin

Not inhibitory

ble intensity of the band of radioactive phosphate with that of the protein core in SDS–PAGE. NH 4Cl, 10 mM, was included in the medium during the chase in order to augment the amount of newly made enzyme that was secreted. As can be seen in Fig. 3A, the ␣-N-acetylglucosaminidase secreted by the B12 cell line contained only a trace of endo-H cleavable phosphate (Fig. 3A, lanes 3 and 4), an amount no greater than in the enzyme that was secreted by the parental Lec1 cell line (Fig. 3A, lanes 1 and 2). Likewise, the ␣-N-acetylglucosaminidase secreted by the cell line (B29) doubly transfected with NAGLU and IDUA contained only a trace of 32P (Fig. 3A, lanes 5 and 6). This should be contrasted with the secretion of recombinant ␣-L-iduronidase by a Lec1 cell line transfected with IDUA (Fig. 3B, lanes 7 and 8), the 2.131 cell line (Fig. 3B, lanes 9 and 10), and the doubly transfected B29 cell line (Fig. 3B, lanes 11 and 12). In each case, the ␣-L-iduronidase contained endo-H-cleavable 32P at a level approxi-

mately equal to 35S. Thus recombinant ␣-N-acetylglucosaminidase differed from recombinant ␣-L-iduronidase in having no mannose 6-phosphate, even when both enzymes were secreted by the same cells. In addition, the 32P found in intracellular recombinant ␣-Nacetylglucosaminidase was also much reduced in comparison to ␣-L-iduronidase (not shown). Because it had been reported that naturally occurring ␣-N-acetylglucosaminidase contained mannose 6-phosphate (34), enzyme secreted by normal human diploid skin fibroblasts was tested in the same manner. The ␣-N-acetylglucosaminidase made by the human fibroblasts was found to be heavily phosphorylated as judged by the similar intensity of 32P and 35S bands (Fig. 4A, lanes 1 and 3) and was comparable to ␣-Liduronidase (Fig. 4A, lanes 5 and 7). Removal of the 32P label with endo-H (Fig. 4, even-numbered lanes) confirmed that the label was on oligosaccharide chains. The normal human fibroblast line shown in Fig. 4A was GM4390A; similar results were obtained with the

208

ZHAO AND NEUFELD

rameric) forms that had also been observed in those preparations (11,13). It was competitively inhibited by two imino sugars, 2-acetamido-1,2-dideoxynojirimycin and 6-acetamido-9-deoxycastanospermine in the submicromolar range, with a K i/K m of 0.002 and 0.0004,

FIG. 2. Correction of [ 35S]glycosaminoglycan accumulation in deficient fibroblasts by ␣-N-acetylglucosaminidase. (A) The dose-dependent reduction of [ 35S]glycosaminoglycan (GAG) accumulation in Sanfilippo B fibroblasts by purified ␣-N-acetylglucosaminidase. The double reciprocal plot (inset) was used to calculate the concentration providing half-maximal correction. (B) Correction by 4.6 nM ␣-Nacetylglucosaminidase in a 22-h incubation with or without 5 mM mannose 6-phosphate (M6P), glucose 6-phosphate (G6P), mannose (Man), or N-acetylglucosamine (Gn) as indicated. The dotted line indicates the calculated maximal correction. A different specific activity of 35SO 4 was used in A and B.

normal fibroblast lines GM 3440B and IMR90 (not shown). To determine whether the difference between the ␣-N-acetylglucosaminidase secreted by human fibroblasts and CHO cells could be attributed to a speciesspecific difference in the posttranslational processing of that enzyme, the human embryonic kidney cell line 293 was transfected and tested in the same manner. It was found to secrete recombinant ␣-N-acetylglucosaminidase which had only a trace of phosphate, no more than the enzyme secreted by the untransfected 293 cells (Fig. 4B, compare lanes 15 and 11). DISCUSSION

Recombinant human ␣-N-acetylglucosaminidase is a stable enzyme, readily purified to apparent homogeneity in three steps starting from medium conditioned by overexpressing CHO cells. It exists in monomeric and dimeric forms, as seen for enzyme purified from natural sources, though not in the 300-kDa (tri- or tet-

FIG. 3. Comparison of phosphorylation of ␣-N-acetylglucosaminidase and ␣-L-iduronidase secreted by CHO cells. The cells in six-well plates were labeled with 50 ␮Ci/ml [ 35S]methionine/cystine or 32Pi, for a 1-h pulse followed by a 20-h chase. To induce secretion, 10 mM NH 4Cl was added for the chase. Medium was collected at the end of the chase and subjected to immunoprecipitation, SDS–PAGE, and fluorography (odd-numbered lanes). To confirm that 32P was derived from oligosaccharide groups, half of each immunoprecipitate was digested with 100 units of endo-H and incubated in 37°C for 1.5 h prior to electrophoresis (even-numbered lanes). A gutter between lanes indicates that the lanes were moved electronically from their original order on the gel. Exposure time was 3 days for the 32P set in A and overnight for the remainder. (A) Labeling of ␣-N-acetylglucosaminidase with 35S or 32P, as indicated. The cell lines used were: lanes 1 and 2, untransfected Lec1; lanes 3 and 4, B12 (Lec1 transfected with NAGLU); lanes 5 and 6, B29 (cell lines doubly transfected with NAGLU and IDUA). The 35S band with higher mobility in lanes 1 and 3 is a contaminating glycoprotein of endogenous origin in the Lec1cell line. (B) Labeling of ␣-L-iduronidase with 35S or 32P, as indicated. The cell lines used were: lanes 7 and 8, Lec1 transfected with IDUA; lanes 9 and 10, 2.131, the cell line stably transfected with IDUA (18); lanes 11 and 12, B29, the doubly transfected line.

RECOMBINANT HUMAN ␣-N-ACETYLGLUCOSAMINIDASE

FIG. 4. Phosphorylation of ␣-N-acetylglucosaminidase secreted by human cells. (A) Normal human skin fibroblasts, GM 4390A, were pulsed with 32Pi or [ 35S]methionine/cystine for 3 h, followed by chase for 40 h in the presence of 10 mM NH 4Cl. ␣-N-Acetylglucosaminidase was subjected to immunoprecipitation and SDS–PAGE without or with treatment with endo-H. (B) The 293 cell line, with or without transient transfection with pCMV-NAGLU, was labeled as described for A, except that the chase was for 20 h.

respectively. These compounds are known to inhibit at similar concentrations the phosphodiester-␣-N-acetylglucosaminidase (35) as well as ␤-N-acetylglucosaminidases from various sources, including lysosomal ␤-hexosaminidase (36,37). These very potent inhibitors may represent transition state analogs for both ␤- and ␣-N-acetylglucosaminidases. This is not the case for 2-acetamido-2-deoxyglucono-1,5-lactone, which was found three orders of magnitude less effective for ␣-N-acetylglucosaminidase than had been found for ␤-hexosaminidase (38). Strong inhibitors of a different class were organic and inorganic mercury, suggesting that cysteine residues are essential for ␣-N-acetylglucosaminidase activity. Because iodoacetamide and N-ethylmaleimide were not inhibitory, the sulfhydryl residues are presumed to be buried in hydrophobic regions of the enzyme. The inhibition by iodoacetate may be ascribed to the reactivity of that compound with histidine (39). Uptake of the recombinant ␣-N-acetylglucosaminidase into Sanfilippo B fibroblasts could not be demonstrated under conditions used. On the other hand, correction of the excessive accumulation of [ 35S]glycosaminoglycans was observed and was inhibited by mannose 6-phosphate. Since correction requires uptake,

209

there is an apparent discrepancy between the two observations, but it is readily explained by the difference in the sensitivity of the two tests. For recombinant ␣-L-iduronidase administered to deficient fibroblasts derived from Hurler patients, half-maximal correction had been observed at 0.7 pM, whereas half-maximal uptake was seen only at 0.7 nM (18). The half-maximal correction of the [ 35S]glycosaminoglycan accumulation of Sanfilippo B fibroblasts occurred at 0.6 nM, a concentration 800 times higher than required for halfmaximal correction of the accumulation in Hurler fibroblasts by ␣-L-iduronidase. The need for such a high concentration suggested that only a small fraction of the ␣-N-acetylglucosaminidase carried the mannose 6-phosphate signal. To verify that the absence of mannose 6-phosphate groups was not because these were blocked (i.e., in diester form) or had been lost during purification, we turned to biosynthetic experiments in which we could compare incorporation of radiolabeled phosphate into ␣-N-acetylglucosaminidase with incorporation into ␣-L-iduronidase by overexpressing cell lines. Recombinant human ␣-L-iduronidase has two of its six carbohydrate chains phosphorylated (20). If only one of the six putative carbohydrate chains of ␣-N-acetylglucosaminidase was phosphorylated, the resulting 32P signal should have been half as intense, relative to the protein, as that of ␣-L-iduronidase. Instead, there was never more than a barely detectable 32P signal, no stronger than when untransfected CHO cells were used under similar conditions. This led to the conclusion that any phosphorylated ␣-N-acetylglucosaminidase seen in conditioned medium was due to endogenous CHO enzyme. The level of ␣-N-acetylglucosaminidase in the secretions of untransfected Lec1 cells was 0.15% that in the stably transfected line; if that ratio had been maintained during purification, the concentration of endogenous CHO enzyme in the mixture giving half-maximal correction would be 0.9 pM, a value in the range of that found for recombinant ␣-Liduronidase. It is therefore reasonable to ascribe the corrective activity of the purified recombinant human enzyme to contamination by CHO enzyme. Rather remarkably, cell line B29, which overexpresses both ␣-L-iduronidase and ␣-N-acetylglucosaminidase, secreted the former in highly phosphorylated form and the latter with only a trace of phosphate. That is, ␣-N-acetylglucosaminidase did not compete with ␣-L-iduronidase as a substrate for UDPN-acetylglucosamine: lysosomal protein N-acetylglucosaminyl-1-phosphotransferase. This shows that the difference in the manner the two enzymes were phosphorylated must be ascribed to some difference in interactions between the newly made enzyme proteins and the processing machinery of the cell. Since it had been reported that ␣-N-acetylglu-

210

ZHAO AND NEUFELD

cosaminidase purified from human urine was efficiently taken up by the mannose 6-phosphate system (34) and that enzyme synthesized by human cells contained phosphorylated carbohydrate (40), we tested the phosphorylation of ␣-N-acetylglucosaminidase by diploid human fibroblasts. The cells were labeled in the presence of NH 4Cl, to induce secretion of all newly formed lysosomal enzymes. The secreted ␣-N-acetylglucosaminidase contained mannose 6-phosphate as predicted from the earlier findings. But the difference in acquisition of mannose 6-phosphate cannot be ascribed to a species-specific difference in posttranslational processing, since transfected 293 cells, also of human origin, did not incorporate phosphate into recombinant ␣-N-acetylglucosaminidase. Thus recombinant ␣-N-acetylglucosaminidase differs from the native enzyme as well as from recombinant ␣-Liduronidase (18,19). It also differs from other recombinant lysosomal enzymes secreted by CHO cells (17,21–26). The reason for this unexpected lack of the mannose 6-phosphate modification on the recombinant ␣-L-iduronidase is not clear. It is unlikely to be due to some error in the transfected NAGLU cDNA that would prevent recognition of the nascent recombinant enzyme by the processing enzyme, UDP-N-acetylglucosamine lysosomal protein N-acetylglucosaminyl-1-phosphotransferase, for two reasons: (a) the nucleotide sequence of the transfecting cDNA was identical to that of two cDNA sequences obtained independently (4,14), and (b) expression of mouse NAGLU in CHO cells also produced enzyme that was not taken up by deficient fibroblasts (not shown). It is also unlikely that all the newly made recombinant ␣-N-acetylglucosaminidase containing mannose 6-phosphate would have been retained in the cells, because the biosynthetic experiments were performed in the presence of NH 4, which induces secretion, and because the intracellular enzyme was not labeled with 32P (not shown). Perhaps the high local concentration of recombinant enzyme in early biosynthetic compartments leads to premature formation of dimers that would not be recognized by the N-acetylglucosaminyl phosphotransferase; such a hypothesis implies that the recombinant and endogenous enzymes do not interact in these compartments. Whatever the reason for the failure of recombinant ␣-N-acetylglucosaminidase to acquire mannose 6-phosphate residues, it underscores the need for quantitative measurements of uptake, to ensure that properties attributed to the recombinant enzyme are not in fact the result of trace contamination by endogenous CHO enzyme. ACKNOWLEDGMENTS We thank Drs. Gu¨nter Legler (University of Cologne, Cologne, Germany), Mohinder S. Kang (Merrell Dow Research Institute, Cin-

cinnati, OH), and Takaaki Aoyagi (Institute of Microbial Chemistry, Tokyo, Japan) for providing the inhibitors used in this study. This work was supported in part by grants from the NIH (Grant NS 22376) and the Children’s Medical Research Foundation.

REFERENCES 1. O’Brien, J. S. (1972) Sanfilippo syndrome: Profound deficiency of ␣-acetylglucosaminidase activity in organs and skin fibroblasts from type-B patients. Proc. Natl. Acad. Sci. USA 69, 1720 –1722. 2. Neufeld, E. F., and Muenzer, J. (1995) The mucopolysaccharidoses in “The Metabolic and Molecular Bases of Inherited Disease” (Scriver, C. R., Beaudet, A. L., Sly, W. S., and Valle, D., Eds.), pp. 2465–2494, McGraw–Hill, New York. 3. Cleary, M. A., and Wraith, J. E. (1993) Management of mucopolysaccharidosis type III. Arch. Dis. Child. 69, 403– 406. 4. Zhao, H. G., Li, H. H., Bach, G., Schmidtchen, A., and Neufeld, E. F. (1996) The molecular basis of Sanfilippo syndrome type B. Proc. Natl. Acad. Sci. USA 93, 6101– 6105. 5. Schmidtchen, A., Greenberg, D., Zhao, H. G., Li, H. H., Huang, Y., Tieu, P., Zhao, H. Z., Cheng, S., Zhao, Z., Whitley, C. B., Di Natale, P., and Neufeld, E. F. (1998) NAGLU mutations underlying Sanfilippo syndrome type B. Am. J. Hum. Genet. 62, 64 – 69. 6. Zhao, H. G., Aronovich, E. L., and Whitley, C. B. (1998) Genotype–phenotype correspondence in Sanfilippo syndrome type B. Am. J. Hum. Genet. 62, 53– 63. 7. Beesley, G. E., Young, E. P., Vellodi, A., and Winchester, B. G. (1998) Identification of 12 novel mutations in the ␣-N-acetylglucosaminidase gene in 14 patients with Sanfilippo syndrome type B (mucopolysaccharidosis type IIIB). J. Med. Genet. 35, 910 –914. 8. Bunge, S., Knigge, A., Steglich, C., Kleijer, W. J., van Diggelen, O. P., Beck, M., and Gal, A. (1999) Mucopolysaccharidosis type IIIB (Sanfilippo B): Identification of 18 novel ␣-N-acetylglucosaminidase gene mutations. J. Med. Genet. 36, 28 –31. 9. Weber, B., Guo, X. H., Kleijer, W. J., van de Kamp, J. J. P., Poorthuis, B., and Hopwood, J. J. (1999) Sanfilippo type B syndrome (mucopolysaccharidosis III B): Allelic heterogeneity corresponds to the wide spectrum of clinical phenotypes. Eur. J. Hum. Genet. 7, 34 – 44. 10. Li, H. H., Yu, W. H., Rozengurt, N., Anagnostaras, S., Fanselow, M., Gomez-Pinilla, F., Suzuki, K., Vanier, M., and Neufeld, E. F. (1999) A mouse model for Sanfilippo syndrome type B, a heritable neurological disorder. J. Neurochem. 73, S23. 11. von Figura, K. (1977) Human ␣-N-acetylglucosaminidase. 1. Purification and properties. Eur. J. Biochem. 80, 523–533. 12. Ro¨hrborn, W., and von Figura, K. (1978) Human placenta ␣-Nacetylglucosaminidase: Purification, characterization and demonstration of multiple recognition forms. Hoppe-Seylers Z. Physiol. Chem. 359, 1353–1362. 13. Sasaki, T., Sukegawa, K., Masue, M., Fukuda, S., Tomatsu, S., and Orii, T. (1991) Purification and partial characterization of ␣-N-acetylglucosaminidase from human liver. J. Biochem. (Tokyo) 110, 842– 846. 14. Weber, B., Blanch, L., Clements, P. R., Scott, H. S., and Hopwood, J. J. (1996) Cloning and expression of the gene involved in Sanfilippo B syndrome (mucopolysaccharidosis III B). Hum. Mol. Genet. 5, 771–777. 15. Mersmann, G., von Figura, K., and Buddecke, E. (1974) Physical properties and biological activities of two forms of ␣-N-acetylglucosaminidase from bovine spleen. Biochim. Biophys. Acta 364, 88 –96. 16. Zhao, Z., Yazdani, A., Shen, Y., Sun, Z., Bailey, J., Caskey, C. T., and Lee, C. C. (1996) Molecular dissection of a cosmid from a

RECOMBINANT HUMAN ␣-N-ACETYLGLUCOSAMINIDASE

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

gene-rich region in 17q21 and characterization of a candidate gene for ␣-N-acetylglucosaminidase with two cDNA isoforms. Mammal. Genome 7, 686 – 690. Ioannou, Y. A., Bishop, D. F., and Desnick, R. J. (1992) Overexpression of human ␣-galactosidase A results in its intracellular aggregation, crystallization in lysosomes, and selective secretion. J. Cell Biol. 119, 1137–1150. Kakkis, E. D., Matynia, A., Jonas, A. J., and Neufeld, E. F. (1994) Overexpression of the human lysosomal enzyme ␣-L-iduronidase in Chinese hamster ovary cells. Protein Express. Purif. 5, 225– 232. Unger, E. G., Durrant, J., Anson, D. S., and Hopwood, J. J. (1994) Recombinant ␣-L-iduronidase: Characterization of the purified enzyme and correction of mucopolysaccharidosis type I fibroblasts. Biochem. J. 304, 43– 49. Zhao, K. W., Faull, K. F., Kakkis, E. D., and Neufeld, E. F. (1997) Carbohydrate structures of recombinant human ␣-L-iduronidase secreted by Chinese hamster ovary cells. J. Biol. Chem. 272, 22758 –22765. Bielicki, J., Hopwood, J. J., Wilson, P. J., and Anson, D. S. (1993) Recombinant human iduronate-2-sulphatase: Correction of mucopolysaccharidosis-type II fibroblasts and characterization of the purified enzyme. Biochem. J. 289, 241–246. Bielicki, J., Fuller, M., Guo, X. H., Morris, C. P., Hopewood, J. J., and Anson, D. S. (1995) Expression, purification and characterization of recombinant human N-acetylgalactosamine-6-sulphatase. Biochem. J. 311, 333–339. Ling, P., and Roberts, R. M. (1993) Overexpression of uteroferrin, a lysosomal acid phosphatase found in porcine uterine secretions, results in its high rate of secretion from transfected fibroblasts. Biol. Reprod. 49, 1317–1327. Bielicki, J., Hopwood, J. J., Melville, E. L., and Anson, D. S. (1998) Recombinant human sulphamidase: Expression, amplification, purification and characterization. Biochem. J. 329, 145– 150. Fuller, M., Van der Ploeg, A., Reuser, A. J., Anson, D. S., and Hopwood, J. J. (1995) Isolation and characterisation of a recombinant, precursor form of lysosomal acid ␣-glucosidase. Eur. J. Biochem. 234, 903–909. Van Hove, J. L., Yang, H. W., Wu, J. Y., Brady, R. O., and Chen, Y. T. (1996) High-level production of recombinant human lysosomal acid ␣-glucosidase in Chinese hamster ovary cells which targets to heart muscle and corrects glycogen accumulation in fibroblasts from patients with Pompe disease. Proc. Natl. Acad. Sci. USA 93, 65–70. Kakkis, E. D., McEntee, M. F., Schmidtchen, A., Neufeld, E. F., Ward, D. A., Gompf, R. E., Kania, S., Bedolla, C., Chien, S. L., and Shull, R. M. (1996) Long-term and high-dose trials of enzyme replacement therapy in the canine model of mucopolysaccharidosis I. Biochem. Mol. Med. 58, 156 –167.

211

28. Kakkis, E., Muenzer, J., Tiller, G., Waber, L., Belmont, J., Passage, M., Izykowski, B., Phillips, J., Walot, I., Doroshow, R., Hoft, R., Yu, K. T., Okasaki, S., Lewis, D., Lachman, R., and Neufeld, E. F. (1998) Recombinant alpha-L-iduronidase replacement therapy in mucopolysaccharidosis I: results of a human clinical trial. Am. J. Hum. Genet. 63, A25. 29. Pontow, S. E., Kery, V., and Stahl, P. D. (1992) Mannose receptor. Int. Rev. Cytol. 137B, 221–244. 30. Chow, P., and Weissmann, B. (1981) 4-Methylumbelliferyl 2acetamido-2-deoxy-␣-D-glucopyranoside, a fluorogenic substrate for N-acetyl-␣-D-glucosaminidase. Carbohydr. Res. 96, 87–93. 31. Fratantoni, J. C., Hall, C. W., and Neufeld, E. F. (1968) The defect in Hurler’s and Hunter’s syndromes: Faulty degradation of mucopolysaccharide. Proc. Natl. Acad. Sci. USA 60, 699 –706. 32. Stanley, P., Narasimhan, S., Siminovitch, L., and Schachter, H. (1975) Chinese hamster ovary cells selected for resistance to the cytotoxicity of phytohemagglutinin are deficient in a UDP-Nacetylglucosamine: glycoprotein N-acetylglucosaminyltransferase activity. Proc. Natl. Acad. Sci. USA 72, 3323–3327. 33. Marsh, J., and Fensom, A. H. (1985) 4-Methylumbelliferyl ␣-Nacetylglucosaminidase activity for diagnosis of Sanfilippo B disease. Clin. Genet. 27, 258 –262. 34. Ullrich, K., Mersmann, G., Weber, E., and Von Figura, K. (1978) Evidence for lysosomal enzyme recognition by human fibroblasts via a phosphorylated carbohydrate moiety. Biochem. J. 170, 643– 650. 35. Zhao, K. W. (1994) “Lysosomal Enzyme Targeting: Purification and Characterization of the Two Enzymes That Catalyze the Formation of Mannose 6-Phosphate from Human Fibroblasts, Ph.D. dissertation, University of California, San Diego. 36. Kappes, e., and Legler, G. (1989) Synthesis and inhibitory properties of 2-aetamido-2-deoxynojirimycin (2-acetamido-5-imino2,5-dideoxy-D-glucopyranose, 1) and 2-acetamido-1,2- dideoxynojirimycin (2-acetamido-1,5-imino-1,2,5-trideoxy-D-glucitol, 2). J. Carbohydr. Chem. 8, 371–388. 37. Liu, P. S., Kang, M. S., and Sunkara, P. S. (1991) A potent inhibitor of ␤-N-acetylglucosaminidases: 6-Acetamido-6-deoxycastanospermine. Tetrahedron Lett. 32, 719 –720. 38. Horsch, M., Hoesch, L., Vasella, A., and Rast, D. M. (1991) N-Acetylglucosaminono-1,5-lactone oxime and the corresponding (phenylcarbamoyl)oxime: Novel and potent inhibitors of ␤-Nacetylglucosaminidase. Eur. J. Biochem. 197, 815– 818. 39. Crestfield, A. M., Stein, W. H., and Moore, S. J. (1963) Alkylation and identification of the histidine residues at the active site of ribonuclease. J. Biol. Chem. 238, 2413–2420. 40. von Figura, K., Hasilik, A., Steckel, F., and van de Kamp, J. (1984) Biosynthesis and maturation of ␣-N-acetylglucosaminidase in normal and Sanfilippo B-fibroblasts. Am. J. Hum. Genet. 36, 93–100.