Recognition of Arylsulfatase A and B by the UDP-N-acetylglucosamine:lysosomal Enzyme N-Acetylglucosamine-phosphotransferase*

The critical step for sorting of lysosomal enzymes is the recognition by a Golgi-located phosphotransferase. The topogenic structure common to all lysosomal enzymes essential for this recognition is still not well defined, except that lysine residues seem to play a critical role. Here we have substituted surface-located lysine residues of lysosomal arylsulfatases A and B. In lysosomal arylsulfatase A only substitution of lysine residue 457 caused a reduction of phosphorylation to 33% and increased secretion of the mutant enzyme. In contrast to critical lysines in various other lysosomal enzymes, lysine 457 is not located in an unstructured loop region but in a helix. It is not strictly conserved among six homologous lysosomal sulfatases. Based on three-dimensional structure comparison, lysines 497 and 507 in arylsulfatase B are in a similar position as lysine 457 of arylsulfatase A. Also, the position of oligosaccharide side chains phosphorylated in arylsulfatase A is similar in arylsulfatase B. Despite the high degree of structural homology between these two sulfatases substitution of lysines 497 and 507 in arylsulfatase B has no effect on the sorting and phosphorylation of this sulfatase. Thus, highly homologous lysosomal arylsulfatases A and B did not develop a single conserved phosphotransferase recognition signal, demonstrating the high variability of this signal even in evolutionary closely related enzymes.

The critical step for sorting of lysosomal enzymes is the recognition by a Golgi-located phosphotransferase. The topogenic structure common to all lysosomal enzymes essential for this recognition is still not well defined, except that lysine residues seem to play a critical role. Here we have substituted surface-located lysine residues of lysosomal arylsulfatases A and B. In lysosomal arylsulfatase A only substitution of lysine residue 457 caused a reduction of phosphorylation to 33% and increased secretion of the mutant enzyme. In contrast to critical lysines in various other lysosomal enzymes, lysine 457 is not located in an unstructured loop region but in a helix. It is not strictly conserved among six homologous lysosomal sulfatases. Based on three-dimensional structure comparison, lysines 497 and 507 in arylsulfatase B are in a similar position as lysine 457 of arylsulfatase A. Also, the position of oligosaccharide side chains phosphorylated in arylsulfatase A is similar in arylsulfatase B. Despite the high degree of structural homology between these two sulfatases substitution of lysines 497 and 507 in arylsulfatase B has no effect on the sorting and phosphorylation of this sulfatase. Thus, highly homologous lysosomal arylsulfatases A and B did not develop a single conserved phosphotransferase recognition signal, demonstrating the high variability of this signal even in evolutionary closely related enzymes.
Soluble lysosomal enzymes are synthesized at the rough endoplasmic reticulum (ER). 1 They are glycoproteins carrying N-linked oligosaccharide side chains. From the ER they are transported to the Golgi apparatus where they acquire mannose-6-phosphate (M6P) residues on their high mannose-type oligosaccharide side chains. In the trans-Golgi compartment the M6P-bearing enzymes bind to M6P receptors, which mediate the vesicular transport from the Golgi apparatus to the lysosomes (for review see Ref. 1).
The synthesis of M6P residues on the oligosaccharide side chains is a two-step process. The initial event is the recognition of soluble lysosomal enzymes by UDP-GlcNAc:lysosomal enzyme N-acetylglucosamine-1-phosphotransferase (briefly phosphotransferase). This enzyme transfers N-acetylglucosaminyl phosphate to mannose residues of the ␣1,6 and ␣1,3 branches of the N-linked oligosaccharide side chains, creating N-acetylglucosamine-1-phospho-6-mannose residues. Subsequently, the terminal GlcNAc residue is removed by N-acetylglucosamine-1-phosphodiester ␣-N-acetylglucosaminidase generating a M6P residue on the oligosaccharide side chains (1). Via these M6P residues the enzymes bind to M6P receptors, which mediate the further vesicular transport from the Golgi apparatus to the lysosomes. Thus, the critical step in the sorting of soluble lysosomal enzymes is the recognition by the phosphotransferase. This enzyme is able to distinguish lysosomal enzymes from secretory proteins. Various efforts have been made to define the structure of the topogenic signal of lysosomal enzymes, which is recognized by the phosphotransferase. No apparent sequence similarities where found when different lysosomal enzymes were compared, instead it was shown the topogenic signal is defined within the three-dimensional structure of soluble lysosomal enzymes (2). Initial experiments with the lysosomal protease cathepsin D identified lysine 203 and amino acids 265-292 in the C-terminal lobe of the enzyme as minimal lysosomal targetting elements (3), but other amino acids also contribute to the phosphotransferase recognition of cathepsin D (4). Overall, the amino acids involved cover a rather large region of about 1630 Å 2 on the surface of cathepsin D. Further experiments demonstrated an additional independent phosphotransferase recognition domain in the N-terminal lobe of the enzyme (5). Other studies, however, could not confirm a phosphotransferase recognition domain in the C-terminal lobe of cathepsin D (6). Thus, despite numerous efforts the topogenic determinants of lysosomal enzymes are still not satisfyingly defined.
Lysine residues, however, seem to play a critical role in the phosphotransferase recognition of various lysosomal enzymes, like cathepsin L (7), arylsulfatase A (ASA), arylsulfatase B (ASB) (9), cathepsin D (3,4), cathepsin L (8, 10), and aspartylglucosaminidase (11). Comparisons of three-dimensional structures of non-homologous cathepsins A, B, and D in combination with peptide inhibition experiments of in vitro phosphorylation indicated that a ␤-hairpin loop structure containing the relevant lysines may be an important common recognition determinant in the cathepsins and in ␤-glucuronidase (12,13). Comparable hairpin loops, however, where not found among those residues relevant for proper sorting of aspartylglucosaminidase (11). When the distance between critical lysine residues was determined, it was striking to find values of 34 -35 Å in various non-homologous * This work was supported by a grant from the Deutsche Forschungsgemeinschaft. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Arylsulfatase B (EC 3.1.6.12) (N-acetylgalactosaminidase-4sulfatase/ASB) removes the sulfate from position 4 of N-acetylgalactosamine sugar residues at the non-reducing terminus of dermatan and chondroitin-4-sulfate. It is synthesized as a 64-kDa precursor, which is proteolytically processed to yield the mature 47-kDa enzyme. ASB bears five N-linked oligosaccharide side chains (17), but their degree of phosphorylation has not yet been determined. Chemical modification of lysine residues on ASA and ASB abolish the in vitro phosphorylation of these enzymes by the phosphotransferase completely (9), and for ASA it was shown that large surface areas interact with the phosphotransferase (9,18).
ASA and ASB are highly homologous enzymes. Comparison of amino acid sequence reveals 29% identity of residues. Both enzymes also share a similar three-dimensional structure (19,20). Both consist of a central ␤-pleated sheet of ten strands, which is decorated by helices on both sides. They also share an intramolecular salt bridge and a more peripheral four-stranded ␤-pleated sheet. Therefore, they may also share a conserved phosphotransferase recognition domain.

EXPERIMENTAL PROCEDURES
Materials-Cell culture media and supplements were from Invitrogen. DNA digesting and modifying enzymes were from Invitrogen, New England BioLabs, and MBI Fermentas. [ 35 S]Methionine (specific activity Ͼ 39 TBq/mmol) and [ 32 P]orthophosphate (specific activity Ͼ 111 TBq/mmol) were from Amersham Biosciences. Isolation of plasmids was performed with a kit from Qiagen and sequencing with a kit from ABI Biosystems, according to the protocols provided by the manufacturers.
In Vitro Mutagenesis-The in vitro mutagenesis protocol was based on the pALTER-1 vector obtained from Promega. To facilitate the process of in vitro mutagenesis and expression of the mutants via transfection experiments, all elements necessary for expression of ASA cDNA were cloned into the polylinker of the pALTER plasmid. A fragment encompassing these elements was obtained from the previously described pBEH ASA expression plasmid (21,22). This fragment contains the SV40 early promoter, the ASA cDNA, and the SV40 polyadenylation signal. The pALTER-1/pBEH ASA fusion plasmid was prepared from bacteria grown in the presence of tetracycline, and single-stranded DNA was prepared using the f1 origin of replication present on the pALTER-1 plasmid and helper phage R408. Generation of singlestranded DNA and in vitro mutagenesis was performed following the protocol provided by Promega. Oligonucleotides used for the in vitro mutagenesis introduced the codons for the substitution of the respective lysines and concomitantly without a further change of the amino acids introduced unique restriction sites to facilitate identification of mutants via restriction enzyme digestion. All mutations were verified by DNA sequencing.
DNA Transfection and ASA and ASB Activity Determinations-Transfection of expression plasmids into BHK cells was performed with LipofectAMINE (Invitrogen). Twenty-four hours prior to transfection 3.5 ϫ 10 5 BHK cells were seeded on a 3-cm cell culture dish. For transfection, 5 l of LipofectAMINE was mixed with 1 g of DNA and 750 l of medium. After 30 min of incubation at room temperature DNA was added to cells in serum-free medium. Six hours later the DNA/ LipofectAMINE-containing medium was removed and replaced by Dulbecco's modified Eagle's medium containing 5% of fetal calf serum. Cells were harvested 48 h later.
To determine ASA activity 50 -100 l of cell lysate corresponding to about 50 -100 g of protein was incubated at 37°C for 30 -60 min with 200 l of substrate (10 mM para-nitrocatecholsulfate in 0.5 M sodium acetate, pH 5.0/10% w/v sodium chloride and 0.3% Triton X-100). The incubation was terminated by the addition of 200 l of 1 N sodium hydroxide. Absorption was read at 515 nm. ASB activity was measured similarly, however, ASB was immunoprecipitated prior to activity measurements, and activity was directly determined in the immunoprecipitates (17). Protein was determined with the DC assay protein determination kit from Bio-Rad following the protocol provided by the manufacturer.
Metabolic Labeling and Immunoprecipitation-The protocols for metabolic labeling with [ 35 S]methionine and [ 32 P]orthophosphate and the protocol for subsequent immunoprecipitation of ASA and ASB have been described in detail previously (23). To facilitate quantification of ASA-associated radioactivity with a Fuji Bioimager, in the experiments performed in this study, cells were labeled separately with either [ 35 S]methionine and in parallel under identical conditions with [ 32 P]orthophosphate. Immunoprecipitations were done from each of the labeled cells under identical conditions. After SDS-PAGE the radioactivity incorporated in ASA or ASB was quantified using a Fuji Bioimager. Pixels of the corresponding polypeptide band were integrated. After subtraction of background values the numbers obtained were taken as arbitrary values for 32 P and 35 S incorporation.
To calculate the extent of phosphorylation, the values for [ 32 P]orthophosphate and [ 35 S]methionine incorporation obtained from the phosphorimaging data were divided ([ 32 P]orthophosphate/[ 35 S]methionine). The quotient obtained for the wild-type enzyme was taken as 100%. ASA bears no phosphate in the polypeptide backbone.
Analysis of Phosphorylation of Individual Glycosylation Sites-For each sample four 28-cm 2 plates of transiently transfected BHK cells were labeled with 500 Ci of [ 32 P]orthophosphate each. After lysis of the cells, 40 g of unlabeled purified recombinant ASA was added. Subsequently, ASA was immunopurified (16,18) and subjected to reductive carboxymethylation, desalting, and digestion with trypsin. Tryptic peptides were separated by reversed-phase HPLC on a C8 column (0 -60% acetonitrile in aqueous 0.1% trifluoroacetic acid). Radioactivity in the chromatographic fractions was determined by liquid scintillation counting using a high energy channel (156 -1700 keV), which reduced background radioactivity to 2-3 cpm per sample. The glycopeptides P9 and P18 containing the relevant phosphorylated oligosaccharide side chains were identified by N-terminal sequencing and MALDI-TOF mass spectrometry. The entire procedure has been described in detail elsewhere (16,24,25).

RESULTS
In Vitro Mutagenesis of Lysine Residues-Human ASA has a total of eight lysine residues. Five of these residues (Lys-367, Lys-393, Lys-433, Lys-457, and Lys-463) are located on the surface of the enzyme, two are partially accessible (Lys-302 and Lys-395), and one is not accessible from the solvent (Lys-123). Lys-123 and Lys-302 have been shown to be part of the active center (19,24). In an initial in vitro mutagenesis experiment all lysines except for Lys-123 were replaced by alanine.
Expression plasmids containing the wild-type and lysine by alanine-substituted ASA cDNAs were transiently transfected into BHK cells in up to eight independent experiments. Fortyeight hours after transfection ASA activity was measured in the cells (see Fig. 1, bottom line). The K302A, K367A, K393A, and K433A substituted ASAs yielded 34 -102% of the activity of wild-type ASA-expressing cells. Little or no ASA activity could be expressed from ASA cDNAs coding for K395A, K457A, and K463A-substituted enzymes (data not shown). Because the amount of enzyme activity expressed by a mutated ASA cDNA can be considered as an indicator for structural integrity, additional amino acid substitutions were introduced at positions 395, 457, and 463 to obtain expression of higher enzyme activities. Because of the close proximity of residues 393 and 395, we introduced various amino acid substitutions at position 395 on the background of the cDNA coding for the K393A-substituted enzyme to obtain an enzymatically active 393/395 double-substituted enzyme. An ASA cDNA coding for a double-substituted K393A/K395G or K393A/K395H enzyme expressed about 35 and 30% of enzyme activity compared with wild-type ASA, respectively. Four different amino acid substitutions (Arg, Ser, His, and Gly) were introduced at position 457, of which only the K457R-and K457S-substituted enzymes yielded satisfying en-zyme activities of 45 and 25%, respectively. Thirteen different amino acid substitutions were introduced at position 463, of which only the K463R-and K463Q-substituted enzymes displayed residual enzyme activity of 39 and 14%, respectively.
To reveal whether the secretion of any of the enzymatically active lysine-substituted enzymes is enhanced, ASA activities in cells and in the media of transiently transfected BHK cells were compared (Fig. 1, A and B). Substitutions at positions 302, 367, 393, 395, and 433 had no effect on the distribution of enzyme activities; values were comparable to that of the wildtype enzyme of which an average of 35% of activity was found in the media. Secretion rates of K463Q-and K463R-substituted ASA were consistently lower (25%) than the wild-type enzyme.
In contrast, all amino acid substitutions introduced at position 457 lead to an increase of ASA activity in the media to about 60% of total activity. Fig. 1A shows the absolute values for the amounts of ASA in cells and media, whereas Fig. 1B shows the same data expressed as percentage of total ASA activity found in the media. The latter presentation makes it more obvious that even mutant enzymes with low residual enzyme activity (K457G) show enhanced secretion.
As a control we have used an ASA (N158Q/N350Q) in which the two attachments sites for phosphorylated N-linked oligosaccharide side chains at Asn-158 and Asn-350 were mutated. The remaining oligosaccharide side chain at Asn-184 is poorly phosphorylated (16), and the mutant is hypersecreted (23). In summary, the data demonstrate that only mutations at position 457 lead to enhanced accumulation of ASA activity in the medium.
Immunoprecipitation of Lysine-substituted ASA in Cells and Media of Metabolically Labeled Cells-To verify the results obtained by ASA activity determination, BHK cells which were transiently transfected with ASA cDNAs coding for the wildtype or the lysine-substituted ASA were metabolically labeled for 1.5 h with [ 35 S]methionine and chased for 6 h. ASA was immunoprecipitated from cells and media. Fig. 2A  experiments ( Fig. 2A) only the K457S and K457R amino acidsubstituted enzymes on average yielded an amount of ASA cross-reacting material in the media comparable to the nonphosphorylated, hypersecreted N158Q/N350Q ASA. Thus, these results are in accordance with enzyme activity determinations shown in Fig. 1.
Phosphorylation of Lysine-substituted ASA-The enhanced secretion of lysine 457-substituted ASA should correlate with a decreased amount of M6P residues on the N-linked oligosaccharide side chains. To determine the degree of mannose-phosphorylation, BHK cells transiently transfected with expression plasmids containing the wild-type and various lysine-substi-tuted ASAs were metabolically labeled with [ 35 S]methionine or [ 32 P]orthophosphate. ASA was immunoprecipitated from cells and analyzed by SDS-PAGE. Radioactivity incorporated into ASA polypeptides was determined by quantification with a Fuji bioimager. The values (total pixel) obtained for [ 32 P]orthophosphate and [ 35 S]methionine incorporation in ASA were divided to obtain an arbitrary value for ASA phosphorylation. The value obtained for the wild-type enzyme was taken as 100%. Fig. 3 shows an example of such an experiment, the table at the bottom of Fig. 3 summarizes results of several independent experiments. Whereas most lysine-substituted ASAs show a degree of phosphorylation that was comparable to the wild-type , and ASA or ASB were immunoprecipitated. The amount of 32 P and 35 S incorporated into ASA and ASB was quantified with a bioimager (Fuji). These arbitrary values for 32 P and 35 S incorporation were divided. The quotient for the wild-type was taken as 100%. The values for single experiments were obtained by comparison of the substituted enzymes to the respective wild-type controls and expressed as percent of wild-type incorporation for every experiment. The table at the bottom summarizes these values for a number of experiments performed with ASA. n.d., non-detectable. The N158Q/N350Q-substituted enzyme has been shown to bear no M6P (16) and served as a negative control. In contrast to ASA, ASB is synthesized as a 64-kDa precursor and processed intralysosomally to a mature 47-kDa form (20). Because it has been shown that in BHK cells mannose 6-phosphate residues are rapidly dephosphorylated upon arrival in the lysosome (30), the mature 47-kDa form is not visible in the panel showing the 32 P labeling. Phosphorylation rates for ASB were obtained by using the 35 S and 32 P incorporation of the 64-kDa ASB precursor only.
ASA, the K457R-substituted ASA showed only 33% of phosphorylation when compared with the wild-type enzyme. Because the replacement of lysine 457 by arginine conserves the positive charge, we also investigated the extent of phosphorylation of the K457S-and K457G-substituted ASAs. Compared with the wild-type enzyme these mutants showed residual phosphorylation of 50 and 31%, respectively.
Phosphorylation of Individual N-Linked Oligosaccharides-Wild-type ASA has three N-glycosylation sites. It was shown previously that only oligosaccharides at the first (Asn-158) and the third (Asn-350) glycosylation site are phosphorylated, whereas the second (Asn-184) is not (16). To examine how the K457R substitution affects the phosphorylation of the first and the third oligosaccharide side chain we analyzed the 32 P incorporation into individual oligosaccharide side chains. BHK cells were transiently transfected with the wild-type ASA and K457R-substituted cDNAs and labeled with [ 32 P]orthophosphate for 3 h. After cell lysis 40 g of unlabeled purified ASA were added, and ASA was immunopurified from the cell lysates. The purified enzyme was carboxymethylated and digested with trypsin. Tryptic peptides were separated by reversed-phase HPLC. The tryptic peptides P9 and P18 containing N-linked oligosaccharides were identified by N-terminal sequencing and MALDI-TOF mass spectrometry. Radioactivity was determined in the fractions. For the wild-type enzyme fractions containing the peptide with oligosaccharide at Asn-158 or oligosaccharide at Asn-350 showed 109 cpm and 35 cpm, respectively (Table I). This is in accordance with previous studies that found a similar relative distribution of radioactivity in wild-type ASA (16). Because it has been shown that the yield of peptide P18 after reversed-phase HPLC of the tryptic digest is only one-third of the yield of peptide P9, the cpm values in fact reflect a phosphorylation of similar extent (16).
In the K457R-substituted enzyme only the fraction containing the peptide P9 with oligosaccharide Asn-158 displayed 155 cpm, and no radioactivity was associated with oligosaccharide Asn-350. It should be emphasized that set-up of the experiment only makes possible conclusions about differences in the phosphorylation of the two individual oligosaccharides rather than about absolute phosphorylation rates.
Thus, the results show that substitution of Lys-457 affects phosphorylation of oligosaccharide Asn-350 to a greater extent than that of oligosaccharide Asn-158. Because in wild-type ASA both oligosaccharides are phosphorylated to a similar extent (16) one can calculate that loss of phosphorylation of oligosaccharide Asn-350 should account for a 50% reduction of overall phosphorylation. Because the remaining oligosaccharide Asn-158 accounts for only 33% (see Fig. 3, bottom), it can be concluded that phosphorylation of oligosaccharide Asn-158 must also be reduced in the K457R ASA.
Secretion and Phosphorylation of Lysine-substituted Arylsulfatase B-Arylsulfatase B is another lysosomal sulfatase with 29% amino acid identity to arylsulfatase A. Arylsulfatase B has 22 mostly surface exposed lysines. Here we have substituted lysines 470, 497, and 507. Lysines 497 and 507 are based on sequence and structure comparisons closest to residue 457 of ASA. Fig. 4 shows the expression of various alanine-substituted ASB polypeptides. Except for ASB in which the three lysines were substituted simultaneously, all substituted enzymes displayed sufficient amounts of enzyme activity upon transient expression in BHK cells. In three independent experiments ASB activity was quantified in cells and media of transiently transfected cells. None of the substituted ASBs, however, showed increased activity in the media (see Fig. 4). As for ASA, 32 P incorporation into ASB was determined by labeling the cells with radioactive orthophosphate. This experiment yielded comparable phosphorylation rates for wild-type and K507A-and K507A/497A-substituted ASB, respectively (Fig.  3). Thus, substitution of these lysines has no effect on the sorting or phosphorylation of ASB. DISCUSSION We have generated several enzymatically active lysine-substituted ASAs and ASBs to investigate the importance of lysine residues in the lysosomal sorting of these sulfatases. Only substitution of lysine 457 caused an increased secretion and reduced phosphorylation of ASA. Because lysine 302 has previously been identified as an element of the active center (19,26) this residue could be expected not to contribute to the phosphotransferase recognition domain. Of the remaining lysines only residues 393, 433, and 457 are freely accessible to solvent. Lysine 395 is not fully surface-exposed, and lysine 463 is located within a groove, which is consistent with our result that they are not essential for recognition. Of the fully solvent-exposed residues lysine 433 of human ASA is not conserved in the murine ASA, which is in accordance with our result, that it is not essential for recognition of ASA by the phosphotransferase.
In all lysosomal enzymes investigated so far more than one lysine is involved in phosphotransferase recognition (4,5,8,11). Simultaneous substitutions of these residues may (11) or may not (8) have additive effects on the extent of phosphorylation. In ASA substitution of a single lysine reduces phosphorylation by about 70%. This is consistent with the result that this ASA is secreted to an extent that is comparable to an ASA having only the second, poorly phosphorylated N-glycosylation site at Asn-184 (16,23). Thus, a single lysine accounts for the major portion of phosphorylation in ASA.
The residual phosphorylation of 33% may be due to the conservation of the positive charge at position 457. However, experiments with K457S and K457G ASA revealed a comparable reduction of phosphorylation, which allows the conclusion that the phosphotransferase recognizes the particular structure of the lysine residue rather than its positive charge only. The data also suggest that the K457S ASA is slightly better phosphorylated than the K457R-and K457G-substituted enzymes (Fig. 3, bottom). However, to prove that this difference is significant further experiments are needed, because we examined the K457S ASA only in one experiment to support the K457R results.
Although K457S-and K457G-substituted ASAs have only little residual enzyme activity, it is unlikely that reduction of phosphorylation is due to ER retention, because these mutants TABLE I Phosphorylation of individual N-linked oligosaccharide side chains BHK cells were transiently transfected with expression plasmids harbouring the ASA wild-type or the K457R-substituted ASA cDNA. Cells were labeled for 3 h with [ 32 P]orthophosphate. After lysis 40 g of unlabeled ASA was added and ASA was immunopurified from the lysates. Purified ASA was carboxymethylated, desalted, and digested with trypsin. Tryptic peptides were separated by reversed-phase HPLC. The glycopeptides P9 and P18 were localized in the chromatographic fractions by MALDI-TOF mass spectrometry, detecting the hexose microheterogeneity characteristic for glycopeptides, and identified by Nterminal sequencing (FLGIPY and EPALAF for P9 and P18, respectively). The amounts of radioactivity (cpm) coeluting in fractions with P9 or P18 (oligosaccharides Asn-158 or Asn-350, respectively) are given and are corrected for background radioactivity (2-3 cpm per fraction, determined by liquid scintillation counting in the 156 -1700 keV channel). are largely secreted and thus, not intracellularly retained. It should also be mentioned that our results for lysine 463 are not fully informative. Because most of the amino acid substitutions we introduced at Lys-463 did not yield enzymatically active enzyme and the K463R or K463Q ASA showed low secretion rates, we assume that this mutant is at least partially retained in the ER. Thus neither secretion rates nor extent of phosphorylation allow conclusion(s) about the importance of this residue in phosphotransferase recognition. However, in contrast to lysine 457 this residue is not fully surface exposed and is therefore unlikely to play a role in phosphotransferase recognition.
For ASA it has been shown that the oligosaccharide side chains at Asn-158 and Asn-350 are phosphorylated to a similar extent, whereas the oligosaccharide at Asn-184 bears no M6P (16). In the K457R ASA only oligosaccharide at Asn-158 is phosphorylated, whereas no radioactivity was detected in fractions containing oligosaccharide at Asn-350 (Table I). Thus, substitution of lysine 457 abolishes phosphorylation of oligosaccharide Asn-350 completely but still allows for residual phosphorylation of mannose residues residing on oligosaccharide at Asn-158 only.
Lysine 457 may represents only one residue of the recognition domain and its loss may lead to an overall reduction of affinity for the phosphotransferase particularly affecting oligosaccharide Asn-350. Alternatively, ASA may have two independent recognition domains. The one responsible for phosphorylation of oligosaccharide Asn-158 could be lysine independent and located within another surface area of ASA. The latter possibility is supported by the fact that in the three-dimensional structure of ASA, oligosaccharide Asn-350 is closest to lysine 457, whereas oligosaccharide Asn-158 is located on the opposite site of the molecule (see Fig. 5). Introduction of novel N-glycosylation sites at various positions into cathepsin D has demonstrated that an increased distance to the phosphotransferase recognition domain correlates with a decreased phosphorylation of oligosaccharides (27). If lysine 457 is part of an independent recognition domain, which mainly directs the phosphorylation of oligosaccharide Asn-350, its substitution may affect phosphorylation of oligosaccharide Asn-350 more severely than that of oligosaccharide Asn-158.
Given the structural diversity of lysosomal enzymes, it seems unlikely that all of them have developed a similar distinct phosphotransferase recognition domain. This has led to the hypothesis that the phosphotransferase recognizes lysosomal enzymes through the contact to various surface located key residues preferably lysines (8,11). These residues must have a particular position relative to each other that would allow for considerable structural variation in the different lysosomal enzymes. Thus, in cathepsin D and cathepsin L critical lysine residues in both non-homologous enzymes are 34 Å apart (8). Similar values of 34.31 and 35 Å were described for desoxyribonuclease I (14) and aspartylglucosaminidase (11), respectively. These similar distances suggest that the spacing of lysine residues on the surface is important for the recognition by the phosphotransferase.
However, at least in ASA a single lysine is critical. Thus, the suggested interlysine spacing of about 34 Å is also not a conserved feature of a topogenic structure common to all lysosomal enzymes. Even if one assumes that lysines other than 457 may also weakly contribute to phosphorylation, the distance between Lys-457 and the other lysines in no case reaches 34 Å. The largest distance of 25 Å is found between Lys-457 and Lys-367.
With the exception of lysine 54 of cathepsin L, all lysines that have been identified to be essential for phosphotransferase recognition of lysosomal enzymes are located in non-structured loop regions. Based on three-dimensional comparisons it has been suggested that at least in cathepsins A, B, and D and TABLE II Conservation of lysines and potential N-glycosylation sites among sulfatases Part A: six lysosomal sulfatases (arylsulfatase A/ASA, accession number AAB03341; N-acetylgalactosamine-6-sulfatase/GA6S, P34059; arylsulfatase B/ASB, AAA51779; iduronate-2-sulfatase/IDS, P22304; glucosamine-6-sulfatase/GL6S, P15586; sulfoglucosamine sulfaminidase/SPHM, NP 000190), and four non-lysosomal sulfatases (steroidsulfatase/STS, P08842; arylsulfatase D/ASD, I37186; arylsulfatase E/ASE, P51690; and sea urchin arylsulfatase/ASU, SO1793) were individually aligned with the ASA amino acid sequence. The top line shows lysine residues of ASA, and the other lines show the residues, which are found in equivalent positions in the other sulfatases. When a lysine is conserved, this is indicated by bold letters. When a lysine is found within three position towards the N or C terminis of the equivalent residue, this is indicated by italic letters. The first letter/number combination gives the residue found in the equivalent position and the second combination the next adjacent lysine. Part B summarizes the conservation of potential N-glycosylation sites corresponding to Asn-158 and Asn-350 in ASA. If potentially glycosylated asparagines were found within 15 amino acids toward the N or C terminis of the amino acid equivalent to Asn-158 or Asn-350 of ASA, they were considered as conserved. Numbers in parentheses give the number of amino acid residues between the residue aligned to Asn-158 or Asn-350 and the actual asparagine of the nearest potential N-glycosylation site in amino acid residues. ␤-glucuronidase a ␤-hairpin loop may be the common structural determinant of the phosphotransferase recognition domain (12,13). The existence of a critical lysine residue in a helix in ASA demonstrates that this cannot be the case for all lysosomal enzymes. Lysines need not be located in unstructured loop regions to be important for mannose phosphorylation.
The diversity of lysosomal enzymes makes the existence of a discrete structurally conserved topogenic domain unlikely. However, the lysosomal sulfatases are homologous (15), and thus one may expect that during evolution a conserved phosphotransferase recognition domain has developed at least among the members of such an enzyme family. Sulfatases show a high degree of conservation in the N-terminal part, which decreases toward the C terminus (28) Thus, it is surprising to find a lysine (Lys-457) relevant for lysosomal sorting in a region showing the least conservation among sulfatases. Lysine 457 of ASA is strictly conserved in iduronate-2-sulfatase (IDS) only (see Table II for summary). ASB and glucosamine-6-sulfatase have lysine residues in adjacent positions. Lysines in this region, however, are missing in N-acetylgalactosamine-6-sulfatase and sulfoglucosamine sulfaminidase. Instead, lysine 457 of ASA is conserved in the non-lysosomal steroidsulfatase and is found in a closely adjacent position in the secretory sea urchin arylsulfatase. Thus, this residue is neither strongly conserved in nor specific for the lysosomal sulfatases. In contrast, the strong conservation of lysines 123 and 302 among most of the sulfatases is not surprising, because they have been shown to be part of the active center (19,26).
Our data concerning ASB support the conclusion that even structurally similar lysosomal sulfatases do not share a single closely related sorting signal. ASB has also been crystallized, and its three-dimensional structure is very similar to that of ASA (see Fig. 5). Also for ASB it has been shown that lysines are essential for phosphotransferase recognition (9). In the three-dimensional models of ASA and ASB lysine 457 of ASA is in a similar position as lysines 497 or 507 in ASB. Both are found in solvent-accessible positions in the most C-terminal helix of both enzymes (see Fig. 5). In addition, potential Nglycosylation sites of oligosaccharides at Asn-158 and Asn-350 of ASA are located at similar positions in ASB (see Fig. 5 and Table II). However, substitution of lysine 497 or adjacent lysines 470 and 507 in ASB has no effect on the sorting or phosphorylation of this enzyme.
Either the phosphotransferase recognition domains in ASA and ASB are located at different positions, or ASB has more than one domain, the remaining one compensates for the loss of Lys-497/507. The presence of independent recognition domains in the N-terminal and C-terminal lobe has also been demonstrated for cathepsin D (29). Because ASB has many solventaccessible lysines equally distributed on the surface and has a total of five N-linked oligosaccharides, one or more lysines may represent the core of an independent recognition domain, which may lead to the phosphorylation of oligosaccharides not present at homologous positions in ASA.
The process by which the phosphotransferase recognizes lysosomal enzymes is still enigmatic. The only common feature of the recognition domain that has been identified so far is the necessity of lysine residues, but the position of these lysines important within the three-dimensional structure of lysosomal enzymes does not show striking similarities present in all enzymes investigated so far. Finally, our results show that even evolutionary related enzymes such as the sulfatases did not develop a single common recognition domain.