A Neutral Magnesium-dependent Sphingomyelinase Isoform Associated with Intracellular Membranes and Reversibly Inhibited by Reactive Oxygen Species*

Activation of neutral sphingomyelinase(s) and subsequent generation of ceramide has been implicated in a wide variety of cellular responses. Although this enzyme(s) has not been purified and cloned from higher organisms, one mammalian cDNA has been previously isolated based on its similarity to the bacterial enzyme. To further elucidate the function of this neutral sphingomyelinase, we studied its relationship with enzymes present in mammalian cells and tissues, its subcellular localization, and properties that could be important for the regulation of its activity. Using specific antibodies, it is suggested that the enzyme could represent one of several forms of neutral sphingomyelinases present in the extract from brain particulate fraction. In PC12 cells, the enzyme is localized in the endoplasmic reticulum and is not present in the plasma membrane. The same result has been obtained in several cell lines transfected or microinjected with plasmids encoding this enzyme. The molecular and enzymatic properties of the cloned neutral magnesium-dependent sphingomyelinase, produced using baculovirus or bacterial expression systems, have been analyzed, demonstrating the expected ion dependence and substrate specificity. The enzyme activity also has a strong requirement for reducing agents and is reversibly inhibited by reactive oxygen species and oxidized glutathione. The studies demonstrate that the cellular localization and some properties of this enzyme are distinct from properties previously associated with neutral magnesium-dependent sphingomyelinases in crude or partially purified preparations.

The accumulation of putative second messenger ceramide has been implicated in a variety of cellular responses and, in particular, in response to a number of apoptotic stimuli (1)(2)(3)(4). Although ceramide can be generated by de novo synthesis, for the majority of cellular responses it is produced by sphingomyelin hydrolysis catalyzed by sphingomyelinases. Several types of sphingomyelinase activities have been characterized in mammalian cells according to optimal pH, ion dependence, and localization: an acidic form, a Zn 2ϩ -dependent secreted form, an alkaline form, and both cytosolic Mg 2ϩ -independent and transmembrane Mg 2ϩ -dependent neutral forms (1,4). The most extensively studied of these enzymes are the acidic and neutral Mg 2ϩ -dependent sphingomyelinases, both of which have been considered as signaling enzymes.
Acidic sphingomyelinase (ASM) 1 is a lysosomal enzyme, its main function being the metabolic degradation of sphingomyelin. The lysosomal storage disorder, Niemann-Pick disease, is caused by mutations in ASM (5). The enzyme has been cloned (6), and while there are at least three transcripts, only one mRNA for ASM has been identified (7). The same gene encodes Zn 2ϩ -dependent secreted sphingomyelinase (8). ASM knockout mice develop normally for approximately 10 weeks, after which they develop symptoms consistent with Niemann-Pick disease (9,10). There is also evidence that this enzyme could participate in signaling responses. The production of ceramide via ASM activation has been observed in response to extracellular signals such as TNF␣ (11) and Fas ligand (12,13), which suggested a potential role for this enzyme in apoptosis. However, studies of cells obtained from ASM knockout mice and Niemann-Pick disease patients resulted in observations showing either an absolute requirement for ASM or normal signaling responses in the absence of this gene (1). This discrepancy could be, in part, due to different cell types and extracellular signals used in a specific study.
Neutral Mg 2ϩ -dependent sphingomyelinase (NSM) activation has been implicated in apoptosis (14 -17), cell senescence (18), cell cycle arrest (19), and cell differentiation (20). In the case of TNF␣, which activates both NSM and ASM, it has been shown that different domains on the TNF receptor and distinct adapter proteins are required to stimulate the two enzymes (21). An adapter protein FAN (factor associated with NSM activation), which interacts with the NSM activation domain on the TNF␣ receptor, is likely to provide a link to the stimulation of NSM. This is supported by the observations that overexpression of FAN enhances NSM activity (22) and that FAN-deficient mice are impaired in TNF-induced NSM activation (23). A number of other molecules have been shown to regulate NSM activity, at least in vitro (1,4). For example, recent studies have suggested that glutathione, within a physiological range, could regulate NSM activity by reversibly inhibiting the enzyme (24).
Although an increase in NSM activity has been described in many cellular systems, the purification of NSM and its molec-ular entity have remained elusive. Recently, NSM from mouse and human (referred to as mammalian NSM (mNSM)) were identified by their similarity with the bacterial neutral sphingomyelinases (25). The enzymes were cloned, subsequently expressed, and exhibited similar characteristics to those described for the partially purified brain enzyme. In addition, it has been shown that TNF␣ stimulation of transfected cells resulted only in a moderate increase in ceramide levels, while the TNF␣-triggered apoptosis was not enhanced compared with the control cells (25). This suggested that the cloned molecule may not be the enzyme participating in TNF␣ signaling. More generally, several other critical issues remained, including the relationship of the cloned molecule with the NSM activity in cells, its subcellular localization, and regulatory mechanisms that control the enzyme activity.
In this study, which is focused on the cloned mammalian NSM, we present evidence suggesting that this enzyme represents one of several isoforms of NSM, that it associates with an intracellular membrane rather than the plasma membrane, and that reversible inactivation by reactive oxygen species could provide one level of control of its activity. The vector pBac4x-1 was from Novagen, pEGFP-C1 was from CLON-TECH, and pGEX-2T was from Amersham Pharmacia Biotech. Bacul-oGold linearized baculovirus DNA was from Pharmingen. Anti-calnexin antibodies were from Bioquote Ltd., anti-His antibody was purchased from Qiagen, and conjugated wheat germ agglutinins (WGAs) were from Sigma. Actigel-ALD beads were purchased from Sterogene, and Probond Nickel resin was from Invitrogen. Complete protease inhibitor tablets were from Roche Molecular Biochemicals.

Materials-[methyl-
Constructs for Expression of Mouse NSM-Recombinant mouse NSM was amplified by polymerase chain reaction from a mouse liver cDNA library using primers 5Ј-CGGGATCCATGAAGCTCAACTTTTCTCTA-C-3Ј and 5Ј-CGGAATTCTTAAGCTCTGTCCCCCTCC-3Ј (with incorporated BamHI and EcoRI sites underlined). The polymerase chain reaction product was cloned using the BamHI/EcoRI sites of the mammalian expression vector PLINK to generate Myc-tagged recombinant protein (Myc.NSM). Inserts without any mutations were then subcloned into the pEGFP-C1 vector (to generate Gfp-tagged protein, Gfp.NSM), the baculovirus expression vector pBac4x-1 (to generate His-tagged protein, His.NSM), and the bacterial vector pGEX-2T (to generate a GST fusion protein, GST.NSM).
Generation and Purification of Polyclonal Antibodies-Rabbit anti-NSM antibodies were raised against the synthetic peptide NH 2 -LLEEVWSEQDFQYLR-COOH, which represents residues 46 -60 of both the murine and human NSM coupled to purified protein derivative. To affinity-purify the antibodies, 1 mg of the synthetic peptide was coupled to 5 ml of Actigel-ALD beads. Immune serum (5 ml) was diluted with an equal volume of PBS and loaded onto a column containing 2 ml of resin. Following washing with PBS, antibodies were eluted with 2 ml of 100 mM glycine, pH 2.5, and immediately buffered with 1 M Tris-HCl, pH 8. The purified IgG was analyzed by SDS-PAGE, and the antibodies were stored at Ϫ20°C in the presence of protease inhibitors.
Microinjection and Transfection-For microinjection, PC12 and 3T3 cells were plated at 10 6 on 6-cm dishes for 24 h before use. MDCK cells were grown to confluence. Gfp.NSM DNA (25 g/ml) was microinjected into the nucleus of the cells. Following a 2-h incubation, the cells were analyzed using a Bio-Rad MRC 1024 confocal imaging system configured for FITC fluorescence.
For transfection of COS cells, the cells were plated at 8 ϫ 10 5 on 10-cm dishes and left overnight. The following day, 5.1 g of DNA (diluted to 0.025 g/ml) was added to 34 l of lipofectAMINE and left to form complexes for 15 min. The cells were washed in serum-free media, and the DNA-lipofectAMINE complexes were subsequently added to the cells in serum-free media. Following a 6-h incubation at 37°C, the cells were washed, incubated for 24 h in medium supplemented with serum, and Gfp fluorescence or immunofluorescence was analyzed by confocal microscopy. For analysis of NSM activity, the cells were washed in PBS, scraped from the dishes, and harvested by centrifugation (1000 ϫ g, 5 min).
3T3 cells were plated at 2 ϫ 10 5 and MDCK cells were plated at 2.5 ϫ 10 5 on 6-cm dishes and grown overnight. The following day, DNA was mixed with PLUS reagent in a total volume of 50 l for 15 min as follows: for 3T3, 1.2 g DNA and 8 l of PLUS reagent; for MDCK, 1 g DNA and 8 l of PLUS regent. LipofectAMINE (8 l for 3T3 cells and 10 l for MDCK cells) was added to the DNA/PLUS reagent mixture in a total volume of 50 l (total transfection volume of 100 l). Following washing with serum-free media, the DNA complexes were added to the cells and incubated for 3 h. After 24 h in medium supplemented with serum, cells were analyzed as described for COS cells.
Immunofluorescence Studies-Cells were washed twice with PBS and then fixed for 10 min with 0.4% formaldehyde. The cells were permeabilized for 10 min with 100 mM glycine and 0.1% Triton X-100 (pH 7.0). Nonspecific binding was blocked by a 5-min incubation in PBS containing 0.1% BSA. Primary antibody was added in PBS with 0.1% BSA (anti-calnexin (1:200), anti-NSM (1:500)) and incubated for 1 h at room temperature. Following washing, the Texas Red-or FITC-conjugated antibody was added for 30 min at room temperature. The cells were washed, mounted, and viewed by confocal microscopy.
Transfection and Infection of Sf9 Cells for Production of Recombinant Baculovirus-Sf9 cells were plated at 10 6 cells/well in a six-well plate and allowed to attach for at least 1 h. The cells were then washed twice in serum-free media. 1 g of pBac4x-1/NSM DNA was mixed with 0.1 g of linearized baculovirus DNA (BaculoGold) in a total volume of 12 l. Lipofectin (8 l) was diluted with 4 l of PBS and then mixed with the DNA mixture. Following a 15-min incubation, the DNA-Lipofectin complexes were added to the cells and incubated for 4 h. The cells were then washed twice and incubated in media with serum for 3 days before being transferred to a 75-cm 2 flask. Primary virus was harvested 7-10 days later and used to infect 10 7 cells to generate secondary virus. Tertiary (high titer) virus was similarly produced and used for infection. Large scale virus was produced from suspension cultures.
For infection, cells growing in suspension were diluted to 10 6 cells/ml, and 10 ml of virus stock was added. Cells were harvested 72 h after infection and washed in PBS, and the pellet was stored at Ϫ20°C until required.
Chromatography of PC12 Cell Extract-Confluent PC12 cells from four 15-cm dishes were washed twice with PBS, scraped from the dish, and harvested by centrifugation. Following resuspension in 4 ml of 25 mM Tris, pH 7.5, 0.1% Triton X-100, 1 mM EDTA, 1 mM DTT, and protease inhibitors, the cells were sonicated for 10 s on medium power and then subjected to centrifugation at 1000 ϫ g for 5 min to obtain the postnuclear supernatant (PNS). The PNS was loaded onto a Mono Q 5/5 column and fractionated with a 15-ml NaCl gradient (0 -1 M). Fractions of 0.5 ml were collected.
Partial Purification of Recombinant His.NSM from Insect Cells-Cell pellets were resuspended in (40 ml/liter of original culture) 25 mM Tris, pH 7.5, 0.1% Triton X-100, 1 mM EDTA, 1 mM DTT, and protease inhibitors (buffer A), sonicated, and centrifuged to generate the PNS. The PNS was loaded onto a Mono Q 16/10 column. Activity was eluted in 5-ml fractions with a 300-ml NaCl gradient (0 -0.5 M). The active fractions were pooled and buffer-exchanged into 20 mM sodium phosphate, 500 mM NaCl, 0.2% Triton X-100, pH 7.8. The sample was bound to 6 ml (bed volume) of nickel-resin for 45 min at 4°C. Batch washing/ eluting was carried out for 5 min each time followed by centrifugation. The resin was extensively washed with 20 mM sodium phosphate, 500 mM NaCl, 0.2% Triton X-100, pH 6.0. Activity was eluted with 5 ϫ 1.5 bed volumes of 250 mM imidazole, 20 mM sodium phosphate, 500 mM NaCl, 0.5% Triton X-100, pH 6.0. The active sample was exchanged into buffer A without protease inhibitors and applied to a 5-ml Hi-trap heparin-Sepharose column. The activity was eluted in 1-ml fractions with a 30-ml salt gradient. Pooled active fractions were then concentrated to 300 l. 200 l of sample was loaded onto a Superose-12 10/30 gel filtration column equilibrated in 20 mM Tris, pH 7.5, 150 mM NaCl, 0.1% Triton X-100, 1 mM EDTA, 1 mM DTT. Following a void volume of 7 ml, 250-l fractions were collected. The active fractions were pooled and adjusted to 10% glycerol (which was found to help stabilize the protein) and stored at Ϫ20°C.
Preparation of Recombinant GST.NSM Fusion Protein-GST.NSM production was induced with 0.2 mM isopropyl-1-thio-␤-D-galactopyranoside and grown for 18 h at 25°C. Cells were pelleted by centrifugation and resuspended in (8 ml/liter of original culture) PBS, 2 mM DTT, 1 mM EDTA, 1% Triton X-100, and protease inhibitors. Following sonication and centrifugation (10,000 ϫ g for 10 min), the supernatant was added to glutathione-beads and incubated rotating at 4°C for 90 min. Fusion protein was eluted with 10 mM glutathione for 3 ϫ 30 min at room temperature. Eluted and dialyzed GST.NSM was stored at Ϫ20°C in 10% glycerol.
Preparation of NSM from Mouse Brain-Mouse brains (5 g) were washed four times with PBS. The brains were then homogenized using a pestle glass homogenizer with 50 ml of homogenizing buffer (25 mM Tris-HCl, pH 7.4, 2 mM EDTA, 5 mM DTT, and protease inhibitors). The crude homogenate was subjected to centrifugation for 1 h at 100,000 ϫ g. The resulting pellet was rehomogenized in 50 ml of homogenizing buffer containing 1 M NaCl and stirred for 30 min at 4°C. The pellet obtained after centrifugation was then resuspended in homogenization buffer containing 1% Triton X-100, stirred for 30 min at 4°C, and again subjected to centrifugation. The supernatant was designated the mouse brain membrane protein extract. The membrane protein extract (10 ml) containing 40 mg of protein was loaded onto a Mono Q 5/5 column equilibrated in 25 mM Tris-HCl, pH 7.5, 0.1% Triton X-100, 1 mM DTT, 1 mM EDTA. Protein was eluted with a 20-ml NaCl gradient (0 -0.5 M). Fractions of 0.5 ml were collected.
Gel Filtration Chromatography-This method was used to further purify peak 1, resulting from chromatography on a Mono Q column described above, as well as to determine the molecular size of this preparation and of partially purified His.NSM. Concentrated peak 1 NSM from mouse brain (400 g) or partially purified His.NSM (100 g) were loaded onto a SMART Superose-12 column in a total volume of 50 l. Chromatography was carried out in 25 mM Tris-HCl, pH 7.5, 0.1% Triton X-100, 1 mM DTT, 1 mM EDTA, 250 mM NaCl. Fractions of 80 l were collected. Molecular mass markers, aldolase (158 kDa), BSA (67 kDa), ovalbumin (43 kDa), chymotrypsinogen (25 kDa), and ribonuclease A (13.7 kDa), were subjected to chromatography under identical conditions, i.e. in the presence of 0.1% Triton X-100.
Measurement of Sphingomyelinase Activity-The activities of both NSM and ASM were measured using radiolabeled substrate. 5 Ci of [ 14 C]sphingomyelin was dried under N 2 and resuspended in 1 ml of 2.5 mg/ml cold sphingomyelin in 10 mM Tris, pH 6.8. For NSM activity, the reaction mixture contained 5 l of substrate, 50 mM Tris, pH 7.5, 0.05% Triton X-100, 1 mg/ml BSA, 10 mM MgCl 2 , 10 mM DTT, and enzyme preparation unless otherwise described in a total volume of 50 l. ASM activity was monitored in a similar way, except 50 mM sodium acetate, pH 5.1, was used, and no MgCl 2 or DTT was added. Following a 1-h incubation at 37°C, the reaction was quenched with 250 l of CHCl 3 /MeOH/HCl (100:100:0.6). Phase separation was achieved with the addition of 75 l of 1 M HCl and subsequent centrifugation. A sample (200 l) of the aqueous phase was removed, and 3 ml of scintillant was added for radioactivity counting. One unit of activity is defined as mol of SM hydrolyzed/h. For measurements of hydrolyzing activities toward PC, PI, PIP 2 , PE, and PS, the assay was performed as above. The substrates were made to the same concentration as the SM substrate with the same ratios of radioactive to nonradioactive lipid. When phosphoinositide-specific phospholipase C (isoform ␦1) was used in the assay, the preparation of the enzyme and the assay conditions were as described before (26).
Protein Determination, SDS-PAGE, and Western Blotting-Protein concentrations were calculated from a standard curve generated from BSA standards using a Bradford assay (Bio-Rad). Samples for gel electrophoresis were combined with 4ϫ SDS sample buffer and boiled for 5 min, and the protein was separated on SDS-PAGE gels. Gels were stained as indicated in the figure legends. For Western blotting, following separation by SDS-PAGE, protein was transferred to polyvinylidene difluoride membrane. The membrane was blocked with Tris-buffered saline/Tween containing 5% dried milk powder for 1 h. Primary antibody (anti-His (1:1000), anti-NSM (1:500), anti-GST (1:15,000)) was added for 1 h in Tris-buffered saline/Tween followed by washing. Secondary antibody (anti-rabbit or anti-mouse), which is conjugated to horseradish peroxidase, was added (1:2000) for 30 -60 min. Following washing, ECL was used for detection.

Expression of Mouse NSM and Generation of Specific
Antibody-Both the mammalian acidic and neutral sphingomyelinases hydrolyze sphingomyelin, but they share no sequence similarity (Fig. 1A). Although dissimilar at the level of primary structure, both proteins contain a transmembrane domain(s) near the C terminus. While ASM is significantly glycosylated, which is essential for its function (27,28), the NSM sequence contains a single putative glycosylation site (Fig. 1A). Human and mouse NSM clones of mammalian enzymes have been identified (25), which share 81% identity. For this study, DNA encoding the mouse enzyme was amplified from a mouse liver cDNA library and is designated as mNSM to refer specifically to this clone. The encoded protein was expressed as both a Myc-tagged and Gfp-tagged enzyme in COS cells. As shown in Fig. 1B, the cloned enzyme exhibited magnesium-dependent activity at neutral pH with negligible activity at acidic pH, in contrast to ASM, which showed characteristic high activity at pH 5. These results are consistent with the previous observations (25) and, in addition, demonstrate that the presence of either the Myc or Gfp tag is compatible with the activity.
A polyclonal anti-peptide antibody was raised against the sequence (residues 46 -60 in mouse NSM) present in both the human and mouse enzymes and subsequently affinity-purified. The antibody did not cross-react with any proteins around 40 kDa present in the PNS of mock-transfected COS cells (Fig. 1C,  lane 1), whereas a strong signal is detected in Myc.NSM-trans- fected cells (Fig. 1C, lane 2). The specific band present in Myc.NSM-transfected COS cells was also detected by anti-Myc antibodies (data not shown), further supporting its identity as mNSM.
Localization Studies of mNSM-NSM activity has been described in plasma membranes (29,30), crude microsomes (31), and nuclear membranes (32,33). To study the cellular localization of mNSM, the Gfp-and Myc-tagged constructs were introduced into different cell lines by microinjection and transfection. Gfp.NSM plasmid was microinjected into the nuclei of PC12, Swiss 3T3, and MDCK cells. Following a 2-h incubation, the observed fluorescence indicated the presence of Gfp.NSM in the endoplasmic reticulum (ER) and nuclear membrane in all three cell lines ( Fig. 2A). Identical localization was demonstrated after microinjection of Myc.NSM plasmid (data not shown). To further confirm this observation, NIH 3T3, COS, and MDCK cells were transfected with Gfp.NSM for 24 h and subsequently stained for the ER marker calnexin (Fig. 2B). When Gfp.NSM localization was superimposed on that of calnexin, colocalization was clearly demonstrated in all three cell lines examined (Fig. 2B, Merge).
The localization to ER could involve an ER retention signal (KDEL or the more recently described HDEF signal (34)) or other types of interactions (35). The mNSM lacks recognized ER retention signals, and while this does not exclude such a localization, it was important to establish whether the endogenous enzyme exhibited the same localization as seen in the microinjection and transfection studies. Of the four cell lines utilized for microinjection and transfection studies, PC12 cells contain significant NSM activity compared with the other cell lines (data not shown). Furthermore, when the postnuclear supernatant from PC12 cells was fractionated on a Mono Q column, the presence of a 40-kDa protein recognized by the anti-NSM antibody was detected in fractions containing magnesium-dependent NSM activity (Fig. 3A). Although the majority of NSM activity eluted as a single peak, the presence of other NSM isoforms that are not recognized by the antibody cannot be excluded. The localization of mNSM in PC12 cells was studied by immunocytochemistry using anti-NSM antibody as well as anti-calnexin antibody and WGA as markers. Similar patterns of staining were observed for both endogenous NSM and calnexin in PC12 cells (Fig. 3B). Since anti-NSM and anti-calnexin antibodies are rabbit polyclonals, this prohibited simultaneous staining of the cells. To further examine cell localization, the cells were double stained with either Texas Red-or FITC-conjugated WGA, which strongly stains the Golgi

. Endogenous NSM from PC12 cells is recognized by anti-NSM antibodies and is localized to the ER.
A, PNS from PC12 cells was fractionated on a Mono Q column, and the resulting fractions were assayed for NSM activity in the presence of Mg 2ϩ . The fractions from the Mono Q column were subjected to SDS-PAGE and Western blotting using anti-NSM (inset). B, PC12 cells were incubated with anti-NSM and visualized with FITC-conjugated secondary antibody (green). In these cells, the Golgi and plasma membrane were visualized with Texas Red-conjugated WGA (red). Conversely, cells that were incubated with anti-calnexin antibodies and visualized with Texas Redconjugated secondary antibody (red) were also incubated with FITCconjugated WGA (green). Phase-contrast images were also obtained. C,

COS cells and MDCK cells were treated as in B.
and to a lesser extent the plasma membrane. As shown in Fig.  3B, the staining of NSM and calnexin is similar but clearly distinct from WGA staining. PC12 cells are small cells, which makes localization studies difficult, and the staining could be argued to resemble cytosolic staining. However, the similar staining observed between NSM and calnexin, which is a specific marker for the ER (36), indicates that the endogenous enzyme is localized to the ER, in agreement with the transfection and microinjection studies. Consistent with the analysis of NSM in different cell lines, the anti-NSM antibody was unable to specifically cross-react with proteins from COS and MDCK cells under the conditions used to analyze PC12 cells (Fig. 3C).
Relationships between mNSM and NSM Activity in Mouse Brain Membranes-The specific antibodies to the cloned enzyme were also used to study the relationship between mNSM and NSM from sources such as brain that are rich in this activity and have been used in purification studies. NSM was extracted from mouse brain membranes and fractionated on a Mono Q column. Using a shallow salt gradient, the activity eluted in four distinct peaks (Fig. 4A), which were all stimulated by Mg 2ϩ and hydrolyzed sphingomyelin specifically. A similar profile was also observed when rat brains were used as the source material (data not shown). Western blotting of the individual peaks with anti-NSM revealed that a 40-kDa protein was only recognized in the peak 1 material, and no crossreacting proteins could be detected in the other peaks (Fig. 4B). Competition experiments with the peptide utilized to raise the antibody resulted in the inability of the antibody to detect any protein in peak 1 (Fig. 4C), illustrating that the antibody selectively reacts with the peak 1 enzyme. To exclude the possibility that the lack of correlation between detection of the specific band by the antibody and NSM activity was not due to the presence of inhibitors or activators in different peaks, the fractions were mixed and analyzed for the activity. Additive results obtained in these experiments suggest that regulatory factors are not present. The possibility that the lack of recognition of the activity in peaks 2-4 could be due to degradation of the NSM, resulting in a loss of the epitope recognized by the antibody, was also considered. Chromatography of Myc.NSM under identical conditions, resulted in a single peak of activity corresponding to peak 1 (data not shown). This suggests that degradation is unlikely to explain the chromatographic differences observed between the four peaks obtained when mouse brain extracts were fractionated. Of various tissues tested from mice, the brain contains the most activity despite having lower amounts of mRNA for the cloned NSM than other tissues with less activity (25). The presence of four distinct activities from mouse brain extracts of which only one is recognized by anti-NSM antibodies indicates that there are at least three additional isoforms of the protein. This observation could account for the huge activity present in mouse brain homogenates that does not correlate with the mRNA levels (25).
The NSM activity from peak 1 was purified a further ϳ10fold by gel filtration chromatography (resulting in specific activity of ϳ5 units/mg), with the single peak of activity correlating with the recognition by anti-NSM antibodies.
Purification of mNSM for Further Characterization-In order to further characterize the mNSM, it was subcloned into a baculovirus expression vector containing a six-histidine motif to facilitate the purification of the enzyme. Infection of Sf9 cells resulted in ϳ70-fold increase in NSM activity; however, problems were encountered with the purification of His.NSM. Incomplete solubilization of the enzyme resulted in the loss of significant amounts of activity, and the His tag did not contribute to the purification as expected, resulting in the purification being achieved mainly by conventional chromatography (see "Experimental Procedures"). While the specific activity increased ϳ100-fold throughout the purification (specific activity was ϳ10 units/mg) the protein remained partially purified (Fig. 5A, left  panel), although the enzyme could be easily detected by Western blotting with an anti-His tag antibody (Fig. 5A, right panel) throughout the purification. The band that the anti-His tag antibody recognized represented only 5-10% of the total protein.
The enzyme was also expressed as a GST fusion protein in FIG. 4. Extracted NSM from mouse brain contains several peaks of NSM activity, one of which could correspond to the cloned enzyme. A, NSM extracted from mouse brain membranes was subjected to chromatography on a Mono Q column. Eluted fractions were assayed for neutral activity in the presence of Mg 2ϩ . B, pooled fractions from each of the active peaks were analyzed by SDS-PAGE and Western blotted with anti-NSM antibodies. Myc.NSM, present in the COS cell PNS, was used as the control. C, anti-NSM antibody was incubated with the peptide to which it was raised for 2 h at 4°C prior to using it to probe a Western blot containing samples from all four peaks of activity. bacteria (the specific activity was ϳ0.07 units/mg compared with nondetectable amounts in control cells), and a single step purification was carried out. The fusion protein could be easily detected by standard staining (Fig. 5B), and this band was also recognized by anti-GST and anti-NSM antibody (data not shown). While the preparation appeared to be highly pure, the specific activity of the protein (0.24 units/mg) was poor by comparison with the NSM produced in insect cells. This would imply that either a proportion of the protein was incorrectly folded in the bacteria or post-translational processing could be required for optimal activity. Nevertheless, the GST.NSM exhibited properties that are consistent with the partially purified His.NSM (see below). Both preparations were analyzed further, although the His.NSM preparation was preferentially utilized due to the high specific activity of this material, which was comparable with the peak 1 enzyme from mouse brain.
Properties of mNSM-A preparation of His.NSM from Sf9 cells was subjected to gel filtration (Superose 12) to determine molecular size under nondenaturing conditions (Fig. 6). The profile of activity correlated well with the recognition by anti-NSM antibodies (Fig. 6A), and the enzymes eluted at about 90 -100 kDa (Fig. 6B). These observations indicate that the native enzyme exists as a dimer rather than a monomer. It is possible, however, that the detergent present in the samples may have resulted in aberrant properties, or the enzyme may interact with an additional protein, which would account for the difference in elution profile between our observations and the predicted profile for the monomer. The activity of the cloned enzyme was analyzed in the presence of different metal ions and toward other potential substrates. The activity of His.NSM (Fig. 7A) was dependent on Mg 2ϩ as described previously (25,37). Manganese also stimulated activity although to a lesser extent than Mg 2ϩ . Minimal stimulation by Ca 2ϩ was observed, while both Ni 2ϩ and Zn 2ϩ were inhibitory. Effects of Mg 2ϩ , Mn 2ϩ , and Ca 2ϩ were examined further within a concentration range of 0.1-100 mM. Mg 2ϩ stimulated the enzyme within this range with ϳ5-fold increase in NSM activity at 10 mM MgCl 2 with little further increase in activity in the range of 10 -100 mM. The maximal stimulation by Mn 2ϩ (ϳ2.2-fold) was reached at 3 mM, while higher concentrations were inhibitory. Stimulation by Ca 2ϩ was only up to 10% reached at 10 mM CaCl 2 , and a further increase inhibited NSM, with complete inhibition observed at a concentration of 30 mM. Regarding substrate dependence, the enzyme hydrolyzed sphingomyelin (Fig. 7B) with a K m of 5.9 M without hydrolyzing detectable amounts of phosphatidylcholine (Fig.  7C). This observation is consistent with that for the partially purified rat brain enzyme (37). The data in Fig. 7E show that preparations of recombinant cloned NSM did not have any hydrolyzing activity toward PI, PIP 2 , PE, or PS. The preparations were also free from exo-and endodeoxyribonuclease activity, ribonuclease activity, and phosphatase activity toward p-nitrophenylphosphate. However, although unlikely, the possi- bility that the isolated cDNA could encode some other phosphodiesterase or phophatase activity cannot be completely excluded.
The effect of various reducing agents on NSM activity was also investigated. It has been reported that DTT helps to stabilize the rat brain enzyme (37), while glutathione inhibits both the brain and cloned enzyme (24,25,38). The data presented in Fig. 8A show that in our experiments the reducing agents DTT, ␤-mercaptoethanol, and reduced glutathione all stimulated His.NSM in a dose-dependent manner. Reduced glutathione produced inhibition only when the pH of the reaction mixture was not controlled and was shifted toward acidic values. This is particularly prominent with high glutathione concentrations (e.g. 20 mM) as used by Tomiuk et al. (25), which could account for the discrepancy between our studies of the same enzyme activity.
To analyze the effects of oxidizing compounds, a His.NSM preparation was incubated with either oxidized glutathione (Fig. 8B) or H 2 O 2 (Fig. 8C) for varying lengths of time prior to the addition to the sphingomyelinase assay. Both compounds inhibited the activity of the cloned enzyme. This inhibition could be reversed by subsequent incubation with either DTT or reduced glutathione. Restoration of activity by DTT was more efficient in both cases than reduced glutathione. These results indicate that NSM is potentially regulated by a redox mechanism.
The observed properties using partially purified prepara-tions of His.NSM were confirmed using highly purified GST-.NSM (Table I) despite its low specific activity. The GST.NSM activity was stimulated in the presence of Mg 2ϩ , Mn 2ϩ , DTT, ␤-mercaptoethanol, and reduced glutathione and inhibited by oxidized glutathione. The same properties have been observed with preparations of partially purified peak 1 from mouse brain (data not shown).

DISCUSSION
Activation of NSM has been reported in different cell types after stimulation with a variety of extracellular signals. Furthermore, its activation has been correlated with different cellular responses (1)(2)(3)(4). However, the molecular entities and the role of NSM in these responses have remained poorly understood. In this paper, we examine the properties of mammalian NSM with similarity to bacterial enzymes, including the relationship of this molecule with the NSM activity in cells and tissues, its subcellular localization, and regulatory mechanisms that could control the enzyme activity.
The only eukaryotic cDNA for the NSM has been cloned on the basis of its similarity to known sequences of bacterial enzymes (25). Similar sequences are present in yeast, nematode, mouse, and human genome. It has been previously shown that expression of this mammalian cDNA results in a dramatic increase of NSM activity, consistent with its predicted function (25). Our data, using highly purified preparations, further support this conclusion ( Fig. 7 and Table I). Furthermore, we demonstrate that the antibody, specific for the expressed protein, recognizes one of the peaks of NSM activity resolved by chromatography of mouse brain membrane extracts (Fig. 4). However, although these data and previous reports (24,37,39) suggest the existence of multiple forms of NSM, the full understanding of a relationship between different peaks of activity would require their purification and determination of primary structures, which would allow comparison at a molecular level.
Studies of the subcellular localization of sphingomyelin, sphingomyelin synthase, and NSM suggest that the presence, as well as the synthesis and degradation, of sphingomyelin is not confined to a single cellular compartment. It is widely accepted that the majority of sphingomyelin resides in the outer leaflet of the plasma membrane. Sphingomyelin is also a component of intracellular membranes, including ER, and this intracellular pool could account for 30 -40% of this lipid in cells (40,41). The exact proportion of sphingomyelin in ER, however, is not certain due to problems of subcellular fractionation and the selection of specific markers for this type of study. The same criticism applies to studies of the distribution of sphingomyelin synthase and NSM also found in different membrane fractions including the plasma membrane, nuclear envelope, and Golgi/ER preparations (30,31,33,42,43). Nevertheless, the use of different experimental approaches, metabolic labeling and the treatment of cells with bacterial sphingomyelinase, supports the existence of at least two pools that are metabolically and spatially separated (41,44). Here we show that, at least in nonstimulated cells, protein corresponding to the cloned enzyme is not associated with the plasma membrane and is found associated with ER membranes (Figs. 2 and 3). This cellular localization is not expected for NSM participating in signaling, based on the studies of subcellular localization of agonist-triggered sphingomyelin hydrolysis (41,45,46). For example, it has been shown that there is a reduction in sphingomyelin levels in HL-60 cells after TNF␣ or D3 stimulation in fractions containing plasma membrane and Golgi markers and that this pool is insensitive to the extracellular addition of bacterial sphingomyelinase (41). This led to the suggestion that in these cells the agonists stimulate sphingomyelin hydrolysis in the inner leaflet of the plasma membrane. In contrast, the level of sphingomyelin present in fractions containing the ER marker (representing about 30% of total sphingomyelin) was not affected by these agonists (41). However, these (41,45,46) and other studies (47), suggesting that cellular localization of ceramide generation is critical for signaling responses, are not consistent with the observed effects of bacterial sphingomyelinase and short-chain ceramides, which may not be directed to a particular intracellular compartment (48,49). Assuming that their action is not related to triggering of stress responses through different signaling pathways, these findings indicate that ceramide that is not confined to the inner leaflet of the plasma membrane could also participate in signaling. At present, the question of the biological role of ceramide that could be generated by the cloned NSM in the ER membranes remains open, and its role could be related to interactions with proposed ceramide targets. Alternatively, after transport from the ER to Golgi, ceramide could serve as a precursor for synthesis of other lipids, including a ganglioside recently implicated in ceramideinduced apoptosis (50).
The properties of the cloned enzyme are consistent with the   (25). Also, solubilization of the NSM activity expressed in insect cells or present in brain preparations described here requires prolonged detergent extraction. Similar requirements have been reported in many previous studies of NSM from different biological sources (4). The specificity of the cloned enzyme for sphingomyelin as a substrate is another property previously shown for partially purified NSM preparations (37). However, unlike the partially purified NSM from rat brain (24,37), the cloned enzyme was not inhibited by reduced glutathione (Fig. 8). The inhibition of the rat brain enzyme by either reduced or oxidized glutathione has been emphasized as an important property that could be related to its activation in vivo (37,38). Since a decrease of total glutathione concentrations and efflux of reduced glutathione often accompany cell death and injury, it has been suggested that this drop in concentrations could release the inhibitory effect on NSM exerted in nonstimulated cells. In contrast to the observations with the partially purified enzyme from rat brain, the cloned enzyme and the corresponding activity in mouse brain preparations can be inhibited only by oxidized glutathione or reactive oxygen species such as H 2 O 2 (Fig. 8). Furthermore, this inhibition is reversed by the reduced glutathione and other reducing agents like DTT and ␤-mercaptoethanol, which stimulate this NSM activity (Fig. 8). Thus, these effects of glutathione are not related to the structure of the glutathione molecule itself, as in the case of the rat brain enzyme (24), but due to the ability of glutathione to act as a regulator of the redox state. Similar reversible inactivation has been described for many enzymes that require cysteine residues for their activity. In some instances, it has been suggested that the modification of cysteine residues could underlie regulatory mechanisms mediated by H 2 O 2 and NO in cells. For example, studies of caspase-3 and other cysteine proteases involved in apoptosis have demonstrated that NO, H 2 O 2 , and related molecules (either after addition to the extracellular medium or endogenous production) can regulate enzyme activity by directly modifying the active site cysteine (51)(52)(53)(54)(55)(56). Similar reversible mechanism(s) could be involved in the regulation of the cloned NSM activity, where maintaining the reduced state is required for its activity and the modification of critical residues by reactive oxygen species could lead to inactivation. The data obtained for the NSM corresponding to the cloned enzyme in this and previous studies (25) provided important insights into the molecular and enzymatic properties of this sphingomyelinase. The elucidation of its biological role, however, clearly requires more extensive studies. This is further reinforced by transfection experiments using the cDNA for this enzyme, suggesting the possibility that it could be critical for some cellular systems (57) but without an important role in others (25).