A novel enzyme that cleaves the N-acyl linkage of ceramides in various glycosphingolipids as well as sphingomyelin to produce their lyso forms.

We describe a novel enzyme that hydrolyzes the N-acyl linkage between fatty acids and sphingosine bases in ceramides of various sphingolipids. The enzyme was purified about 300-fold with 5% recovery from the culture filtrate of a newly isolated bacterium (Pseudomonas sp. TK4) by ammonium sulfate precipitation followed by several steps of high performance liquid chromatography. The purified enzyme preparation was completely free of exoglycosidases, sphingomyelinase, and proteases, and showed a single protein band corresponding to a molecular mass of 52 kDa on SDS-polyacrylamide slab gel electrophoresis after staining with Coomassie Brilliant Blue. The enzyme shows quite wide specificity, i.e. it hydrolyzes both neutral and acidic glycosphingolipids, and simple glycosphingolipid cerebrosides to polysialogangliosides such as GQ1b. Furthermore the enzyme also hydrolyzes sphingomyelin to produce the respective lyso form. However, the enzyme shows hardly any activity on ceramides, indicating that it is completely different from the ceramidase (EC 3.5.1.23) reported previously. This enzyme, which is tentatively named sphingolipid ceramide N-deacylase, should greatly facilitate the further study of sphingolipids as well as lysosphingolipids.

In vertebrates, glycosphingolipids (GSLs) 1 are located on the outer leaflet of the plasma membranes and may function as mediators of cell-cell interaction, attachment, proliferation, and differentiation (1). Lyso-GSLs, which are GSLs N-deacylated in the ceramide moiety, have been detected in normal tissues at very low levels but are accumulated in inherited sphingolipid storage diseases (2). Recently, several lines of evidence have suggested the biological significance of lyso-GSLs in cell activities. Tyrosine-specific autophosphorylation of the epidermal growth factor receptor of A431 cells is inhibited by lyso-GM3 as well as by GM3, and both of these are detected in the cells (3). Lyso-GSLs inhibit protein kinase C, and this may be responsible for the pathogenesis of sphingolipidoses (4). Furthermore, sphingosine, a deglycosylated form of lyso-GSLs, has been found to modulate protein kinase C-dependent cell functions (5) as well as a number of other systems (6).
Although one possible mechanism by which intracellular lyso-GSLs may be removed by direct N-acylation has been proposed (7), the molecular mechanism of lyso-GSL generation in situ remains unclear. Recently, Hirabayashi et al. reported the presence of lyso-GSL-generating hydrolase activity in actinomycetes (8). The enzyme was, however, difficult to solubilize from the cells, and thus the enzyme protein has not yet been characterized. In this paper we report that the novel enzyme, purified as an apparently homogeneous protein from a newly isolated bacterium, cleaves the N-acyl linkage of ceramides in various GSLs as well as sphingomyelin to produce their lyso forms. This is the first report describing the generation of lysosphingomyelin from sphingomyelin by a specific hydrolase. Lysosphingomyelin has been shown to exert potent mitogenic activity (9) and to modulate cytosolic protein phosphorylation (10). The enzyme, tentatively designated sphingolipid ceramide N-deacylase, should facilitate the further study of sphingolipid as well as lysosphingolipid functions.

EXPERIMENTAL PROCEDURES
Materials-A mixture of gangliosides was prepared from bovine brain using a method described previously (11). GM1 and asialo GM1 were prepared from a mixture of gangliosides by digestion with neuraminidases isolated from Clostridium perfringens (Sigma) and Arthrobacter ureafaciens (Nakarai Chemical Co., Japan), respectively, followed by purification with DEAE-Sepharose and Iatrobead column chromatography. Other GSLs were purchased from Iatron Laboratories, Inc. (Japan). Sphingomyelin and ceramide from bovine brain were purchased from Matreya, and Triton X-100 was from Sigma. Precoated Silica Gel 60 TLC plates were obtained from Merck (Germany). Endoglycoceramidase was prepared as described previously (12,13) or purchased from Takara Shuzo Co. (Japan).
Enzyme Assay-The activity of sphingolipid ceramide N-deacylase was measured using asialo GM1 as the substrate as described below. The reaction mixture contained 10 nmol of asialo GM1 and an appropriate amount of the enzyme in 20 l of 20 mM sodium acetate buffer, pH 5.0, containing 0.8% Triton X-100. Following incubation at 37°C for 30 min, the reaction was stopped by heating in a boiling water bath for 3 min. The reaction products were freeze-dried by a Speed Vac concentrator (Savant Instruments, Inc.), redissolved in 5 l of chloroform/ methanol (1:2, v/v), and analyzed by TLC using chloroform, methanol, 10% acetic acid (5:4:1, v/v/v) as the developing solvent. GSLs and lyso-GSLs were visualized by spraying the TLC plates with orcinol-H 2 SO 4 reagent and scanning them with a Shimadzu CS-9300 chromatoscanner with the reflectance mode set at 540 nm. The extent of hydrolysis was calculated as follows: hydrolysis (%) ϭ (peak area for lysoasialo GM1 generated) ϫ 100/(peak area for remaining asialo GM1 ϩ peak area for lysoasialo GM1 generated). One enzyme unit was defined as the amount capable of catalyzing the release of 1 mol of lysoasialo GM1/min from the asialo GM1 under the conditions indicated * This work was supported in part by a Grants-in Aid for Scientific Research on Priority Areas 05274107 and 06454657 from the Ministry of Education, Science and Culture of Japan and grants from the Mizutani Foundation for Glycoscience. 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.
FAB-MS Analysis-Lyso-GSLs were analyzed by negative FAB-MS using a JEOL JMS HX-100 mass spectrometer (JEOL Ltd., Japan) with triethanolamine as the matrix. For lysosphingomyelin, analysis was conducted in the positive mode using diethanolamine as the matrix (18).
SDS-Polyacrylamide Slab Gel Electrophoresis and Protein Assay-SDS-Polyacrylamide gel electrophoresis was conducted on a slab gel with 10% acrylamide according to Laemmli (19). The sample was heated at 100°C for 3 min before electrophoresis except for detection of the activity. For this purpose, the sample was left at room temperature for 10 min. The duplicate gel was cut into 4-mm slices without staining. Each gel slice was crushed with a glass bar in an Eppendorf tube containing 1 ml of 20 mM sodium acetate buffer, pH 5.0, containing 0.3% Triton X-100 and shaken at 4°C for 2 h. After centrifugation at 10,000 rpm for 10 min, the supernatant was dialyzed against 2 mM sodium acetate buffer, pH 6.0, in order to remove SDS, and this was found to be effective for restoration of enzyme activity. The enzyme activity was determined by the method described under "Experimental Procedures" using asialo GM1 as the substrate. The incubation time for this experiment was 16 h. The protein was stained with Coomassie Brilliant Blue, and the protein content at each step of purification was determined by the bicinchoninic acid protein assay (Pierce) with bovine serum albumin as the standard.
Sugar Composition Analysis-Lysoasialo GM1 (100 nmol) was hydrolyzed with 2.5 N trifluoroacetic acid at 100°C for 6 h and analyzed using a Dionex HPLC system with a Carbo Pac PA column (Dionex) and pulsed amperometric detection (20).
Isolation and Cultivation of Pseudomonas sp. TK-4 -A strain (TK-4) capable of producing sphingolipid ceramide N-deacylase was isolated from pond water using a synthetic medium containing gangliosides as the sole source of carbon. The bacterium was assigned to the genus Pseudomonas on the basis of morphological and biochemical characteristics, which will be reported in detail elsewhere. In order for this strain to retain its ability to produce the enzyme, it must be maintained in a medium containing gangliosides (0.5% polypeptone, 0.1% yeast extract, 0.2% NaCl, 0.1% bovine brain gangliosides, and 1.6% agar, pH 7.0), as is presently being done at our laboratory. For preparation of sphingolipid ceramide N-deacylase, inocula from an agar slant of the strain TK-4 were introduced into a cotton-plugged 50-ml flask containing 20 ml of sterilized liquid medium (0.5% polypeptone, 0.1% yeast extract, 0.2% NaCl, and 0.1% bovine brain gangliosides) and incubated at 25°C for 1 day with vigorous shaking. The culture was then transferred to a cotton-plugged 5000-ml flask containing 1000 ml of the same medium and incubated at 25°C for 3 days with vigorous shaking.
Purification of Sphingolipid Ceramide N-Deacylase from the Culture Supernatant-The supernatant obtained (1,800 ml) was adjusted to 75% saturation with solid ammonium sulfate and allowed to stand overnight. The precipitate was collected by centrifugation and dissolved in 72 ml of 20 mM sodium acetate buffer, pH 6.0, containing 0.1% (w/v) Lubrol PX (buffer A). A 15-ml aliquot of the enzyme solution from the ammonium sulfate precipitation step was applied to a DEAM column (2.2 ϫ 15 cm; Yamazen Co., Japan), previously equilibrated with buffer A, using a BPLC-600FC HPLC system (Yamazen Co., Japan). The column was washed with 5 bed volumes of buffer A, and sphingolipid ceramide N-deacylase was then eluted from the column with a linear salt gradient generated by buffer A and 1 M NaCl in the same buffer. The active fractions were pooled, concentrated by an Amicon concentrator using a YM10 membrane, and dialyzed against buffer A. The enzyme solution was applied to a gel filtration column of HW-55F (4.4 ϫ 30 cm; Yamazen Co., Japan) using a BPLC-600FC HPLC system. The column was equilibrated and eluted with buffer A containing 0.2 M NaCl. The flow rate was 5 ml/min, and fractions of 5 ml were collected. The active fractions were pooled, concentrated, and applied to a Phenyl-5PW column (8 ϫ 75 mm; Tosoh, Japan) using a GT1 gradient HPLC system (Pharmacia Biotech Inc.). The column was equilibrated with buffer A, and the enzyme was eluted from the column at a flow rate of 0.5 ml/min with a linear gradient generated using 20 mM sodium acetate buffer, pH 6.0, and the same buffer containing 2% Lubrol PX.
The enzyme was finally purified by a DEAE-5PW column (8 ϫ 75 mm; Tosoh, Japan) using a GT1 gradient HPLC system (Pharmacia). The column was equilibrated with buffer A, and the enzyme was eluted from the column at a flow rate of 0.5 ml/min with buffer A. Contaminating proteins were adsorbed on a column and eluted with buffer A containing 1 M NaCl.
Purification of Lysosphingolipids-Products from asialo GM1 after sphingolipid ceramide N-deacylase treatment were purified by reverse phase HPLC using an ODS column (2 ϫ 300 mm; Tosoh, Japan) as described in Ref. 21. Monitoring of lyso-GSLs was conducted by TLC as described above. For products originating from GalCer and sphingomyelin, a silica gel 60 column was used instead of an ODS column. Lysosphingolipids were eluted from the column using a solvent system composed of chloroform/methanol/water (5:4:1, v/v/v).

Purification of Sphingolipid
Ceramide N-Deacylase-In a typical experiment, the enzyme was purified about 300-fold from a culture filtrate of the newly isolated Pseudomonas sp. TK4 strain with 5% recovery. The purified enzyme preparation was completely free from the following enzyme activities: ␣and ␤-galactosidases, ␤-N-acetylhexosaminidase, ␣-N-acetylgalactosaminidase, ␣-N-acetylglucosaminidase, ␣-L-fucosidase, ␣and ␤-mannosidases, ␣and ␤-glucosidases, sialidase, endoglycoceramidase, sphingomyelinase, and proteases. The enzyme preparation showed a single protein band corresponding to a molecular mass of 52 kDa on SDS-polyacrylamide slab gel electrophoresis after staining with Coomassie Brilliant Blue (Fig. 1). The duplicate gel was cut into 4-mm slices, the protein was eluted and dialyzed, and the enzyme activity was measured as described under "Experimental Procedures." The activity was detected only at the position corresponding to the 52-kDa band.
General Properties-The general properties of the enzyme are as follows: optimal activity at pH 5.0 -6.0 and stable between pH 4.0 and 9.0; potently inhibited by Hg 2ϩ , Cu 2ϩ , and Zn 2ϩ (2 mM) but not by Ca 2ϩ , Mn 2ϩ , Mg 2ϩ , and EDTA, all at the same concentration. The enzyme retained 80% of its activity when kept at 60°C for 30 min and can be kept at Ϫ85°C for 2 months without any loss of activity. Addition of Triton X-100 at a concentration of 0.4 -0.8% (w/v) increased the enzyme activity about 10-fold in comparison with that in the absence of the detergent.
Characterization of Enzymatic Digestion Products-To elucidate the action mode of the enzyme, asialo GM1, GalCer, and sphingomyelin were digested with the enzyme, and the digestion products were separately purified by HPLC followed by TLC analysis. The digestion products migrated on the TLC plate more slowly than native sphingolipids and were stained with either orcinol-H 2 SO 4 (those from asialo GM1 and GalCer;  Fig. 2A, lanes 2 and 4) or Coomassie Brilliant Blue (that from sphingomyelin; Fig. 2B, lane 2). The product from GalCer released by the enzyme was identical to the galactosylsphingosine (psychosine) standard on TLC ( Fig. 2A, lanes 4 and 5). All sphingolipids tested were changed to ninhydrin-positive substances after enzyme treatment ( Fig. 2A, lanes 7 and 9, and Fig. 2B, lane 4), whereas the parental sphingolipids were not stained with ninhydrin ( Fig. 2A, lanes 6 and 8, and Fig. 2B,  lane 3). This demonstrated the generation of free amino groups in sphingolipids by the enzyme treatment. Sugar composition analysis of the product from asialo GM1 revealed that the sugar chain was intact even after enzyme treatment (GalN: Gal:Glc ϭ 1.01:2.0:0.87). This result was confirmed by the fact that the sugar chain released from the product by endoglycoceramidase, which releases sugar chains from both GSLs and lyso-GSLs (12), had the same mobility on TLC as that from the parental asialo GM1 released by endoglycoceramidase (Fig. 2C,  lanes 2 and 4). The product from asialo GM1 released by the enzyme was stained with ninhydrin (Fig. 2C, lane 5), but the oligosaccharide released from the product by endoglycoceramidase was not (Fig. 2C, lane 6), suggesting that the acetyl group at the C-2 position in GalNAc could not be removed by the enzyme, i.e. the enzyme is specific to the N-amide linkage in ceramide but not to that in the carbohydrate moiety. Sphingosine, which was stained by ninhydrin, was generated from the product by endoglycoceramidase treatment (Fig. 2C, lane 6). Finally, the products released from asialo GM1, GalCer, and sphingomyelin were identified using a FAB-MS. As shown in Fig. 3, A and B, the characteristic pseudomolecular ions (M-H) Ϫ were found at m/z 989 for the product released from asialo GM1 (M r 1256; C18:0, d 18: 1), and m/z 461 for the product released from GalCer (M r 728; C18:0, d 18:1) using the negative ion mode. On the spectra of the product released from asialo GM1, fragment ions m/z 827 (corresponding to lysoasialo GM2) and m/z 624 (corresponding to lysoasialo GM3) were also observed (Fig. 3A). For the product released from sphingomyelin (M r 732; C:18:0, d 18:1), (M ϩ H) ϩ was found at m/z 467 using

FIG. 3. FAB-MS analysis of the products released from sphingolipids by sphingolipid ceramide N-deacylase.
A, product released from asialo GM1 by the enzyme; B, product released from GalCer by the enzyme; C, product released from sphingomyelin by the enzyme. Analysis was conducted using a negative mode for A and B and a positive mode for C. Details are given under "Experimental Procedures." the positive ion mode (Fig. 3C). Using the negative ion mode, however, the (M-H) Ϫ ion was not detectable, while the characteristic ion triplet for choline-containing lipids was observed at m/z 451, 406, and 380 (18) (data not shown). These triplet ions were characterized as (M-CH 3 ) Ϫ , (M-HN(CH 3 ) 3 ) Ϫ , and (M-CH 2 CHN(CH 3 ) 3 ) Ϫ , respectively. FAB-MS analysis indicated that these products were lyso forms of the respective parental sphingolipids having a d 18:1 sphingosine base as the major molecular species. In summary, it was concluded that the enzyme cleaves the N-acyl linkage between the fatty acid and sphingosine base in ceramides of sphingolipids. We tentatively designate this enzyme sphingolipid ceramide N-deacylase, whose systematic name should be sphingolipid N-acylsphingosine amidohydrolase, based on its unique specificity. It should be emphasized that the enzyme cleaves the N-acyl linkage of ceramides in sphingomyelin to produce lysosphingomyelin, since the sphingomyelinase (EC 3.1.4.12) reported so far cleaves the linkage between ceramide and phosphorylcholine but never that between sphingosine and fatty acid in ceramide of sphingomyelin (22). Therefore, this paper is the first to report a lysosphingomyelin-generating hydrolase.
Specificity of Enzyme- Fig. 4 shows the time course for the degradation by this enzyme of asialo GM1, GM1, globotetraosylceramide, GalCer, sphingomyelin, and ceramide. The time course of the degradation rates of asialo GM1 by the enzyme was similar to that of GM1, suggesting that this enzyme acts on both neutral and acidic GSLs at the almost same reaction velocity. In contrast to the enzyme from actinomycetes (8), this enzyme can hydrolyze GalCer as well as glucosylceramide (Fig.  4, Table I). However, the enzyme shows hardly any activity on ceramides, indicating that it is completely different from ceramidase (EC 3.5.1.23) (23). The extent of hydrolysis of various sphingolipids after exhaustive digestion with the enzyme is summarized in Table I. This enzyme shows quite wide specificity, i.e. it hydrolyzes both neutral and acidic GSLs, including sulfatide, and also a range from simple GSLs (cerebrosides) to complex polysialogangliosides (GQ1b). Furthermore, the enzyme hydrolyzes not only GSLs but also sphingomyelin. It was notable, however, that the enzyme did not hydrolyze completely all sphingolipid substrates tested even after prolonged incubation. The reason for this is unknown at present but may be due to feedback inhibition of the enzyme by the fatty acids generated, since addition of stearic acid to the reaction mixture inhibited the enzyme activity (data not shown). DISCUSSION Lysosphingolipids are present at low levels in normal tissues but are abnormally accumulated in cells in various lysosomal storage diseases (2). For example, in Gaucher's disease, which is caused by a deficiency of glucosylceramidase, abnormal accumulation of glucosylceramide as well as its lyso form, glucosylsphingosine, is observed (24). Intracellular generation of psychosine is seen in cases of Krabbe's disease (25), which is a progressive and fatal neurogenic disorder. Both psychosine and glucosylsphingosine are considered to be synthesized by the glycosylation of sphingosines in situ, although Yamaguchi et al. (26) reported very recently that glucosylsphingosine is formed not only through the glycosylation of sphingosine but also through the deacylation of glucosylceramide in cultured fibroblasts. They suggested the participation of acidic ceramidase in glucosylsphingosine formation. The enzyme presented here, however, seems to be completely different from the ceramidase reported so far, since the enzyme hydrolyzes various sphingolipids efficiently but hardly acts on ceramide (Fig. 4), which is the most favored substrate for ceramidase (23). Therefore, we tentatively designate the novel enzyme sphingolipid ceramide N-deacylase or sphingolipid N-acylsphingosine amidohydrolase to distinguish it from the known ceramidase. The mode of action of sphingolipid ceramide N-deacylase on asialo GM1 along with that of endoglycoceramidase is presented in Fig. 5. It should be noted that sphingolipid ceramide N-deacylase hydrolyzes various sphingolipids including cerebrosides and sphingomyelin, both of which are completely resistant to hydrolysis by endoglycoceramidase (12).
inhibit protein kinase C and have suggested that the accumulation of lyso-GSLs would eventually lead to cell death due to dysfunction of the signal transduction system (4). Besides inherited lysosomal disease, lyso-GM3 has also been found in a human epidermoid carcinoma cell line, A431, and was shown to inhibit EGF-dependent EGF receptor phosphorylation (3). However, the mechanism of lysoganglioside formation in situ has not yet been elucidated. Whether sphingolipid ceramide N-deacylase (or a similar enzyme) responsible for lysoganglioside generation is present in mammalian tissues should be clarified carefully.
The discovery of sphingolipid ceramide N-deacylase should provide advantages for the study of sphingolipids as well as lysosphingolipids. The preparation of lysosphingolipids will become much easier by using this enzyme. To date, the preparation of lyso GSLs has been done using purely chemical procedures (28), which are somewhat troublesome, time-consuming, and give a low yield, especially in the case of polysialogangliosides. Using the sphingolipid ceramide N-deacylase we were able to obtain easily the lyso forms of all species of GSLs tested without any alternation of their carbohydrate and sphingoid moieties, allowing preparation of new GSL derivatives containing appropriately labeled fatty acids. Furthermore, utilizing the amino groups newly generated in lyso-GSLs, they can be coupled with either appropriate proteins or gel matrix for the affinity column. In addition to the GSLs, the fact that the enzyme can be applied to sphingomyelin should be noted.