The Cellular Trafficking and Zinc Dependence of Secretory and Lysosomal Sphingomyelinase, Two Products of the Acid Sphingomyelinase Gene*

The acid sphingomyelinase (ASM) gene, which has been implicated in ceramide-mediated cell signaling and atherogenesis, gives rise to both lysosomal SMase (L-SMase), which is reportedly cation-independent, and secretory SMase (S-SMase), which is fully or partially dependent on Zn2+ for enzymatic activity. Herein we present evidence for a model to explain how a single mRNA gives rise to two forms of SMase with different cellular trafficking and apparent differences in Zn2+ dependence. First, we show that both S-SMase and L-SMase, which contain several highly conserved zinc-binding motifs, are directly activated by zinc. In addition, SMase assayed from a lysosome-rich fraction of Chinese hamster ovary cells was found to be partially zinc-dependent, suggesting that intact lysosomes from these cells contain subsaturating levels of Zn2+. Analysis of Asn-linked oligosaccharides and of N-terminal amino acid sequence indicated that S-SMase arises by trafficking through the Golgi secretory pathway, not by cellular release of L-SMase during trafficking to lysosomes or after delivery to lysosomes. Most importantly, when Zn2+-dependent S-SMase was incubated with SMase-negative cells, the enzyme was internalized, trafficked to lysosomes, and became zinc-independent. We conclude that L-SMase is exposed to cellular Zn2+ during trafficking to lysosomes, in lysosomes, and/or during cell homogenization. In contrast, the pathway targeting S-SMase to secretion appears to be relatively sequestered from cellular pools of Zn2+; thus S-SMase requires exogeneous Zn2+ for full activity. This model provides important information for understanding the enzymology and regulation of L- and S-SMase and for exploring possible roles of ASM gene products in cell signaling and atherogenesis.

SMases 1 (SM phosphodiesterase, EC 3.1.4.12) have been implicated in a wide variety of physiologic and pathophysiologic processes, including lysosomal hydrolysis of endocytosed SM (1,2), ceramide-mediated cell signaling (3,4), membrane vesiculation (5,6), alterations in intracellular cholesterol trafficking (5,(7)(8)(9), and atherogenesis (10 -13). One type of mammalian SMase is a magnesium-dependent, membrane-bound neutral SMase, and Tomiuk et al. (14) have recently reported the cloning of an enzyme that has several properties in common with this SMase. Two other types of mammalian SMases are lysosomal SMase (L-SMase) and secretory SMase (S-SMase), both of which arise from the "acid SMase" or "ASM" gene (15,16). Both enzymes are soluble hydrolases that function optimally at acid pH in a standard in vitro micellar assay (16,17), although we have shown that S-SMase can hydrolyze physiologic SM-containing substrates at neutral pH (Ref. 18 and see below). Both L-and S-SMase are absent from the cells of patients with types A and B Niemann-Pick disease, which is due to mutations in the ASM gene, and from the cells of ASM knock-out mice (16).
S-SMase may have significant physiologic roles, since extracellular SM hydrolysis may be involved in some or all of the non-lysosomal processes listed above. For example, several lines of evidence have implicated extracellular SM hydrolysis in atherogenesis. First, treatment of LDL with SMase in vitro leads to LDL aggregation (10,11), which is a prominent event during atherogenesis (19 -21) and one that leads to massive macrophage foam cell formation (10,11,(22)(23)(24). Second, aggregated LDL from human and animal atherosclerotic lesions shows evidence of hydrolysis by extracellular SMase, and LDL retained in rabbit aortic strips ex vivo is hydrolyzed by an extracellular, cation-dependent SMase (12). Third, S-SMase, a leading candidate for this arterial wall enzyme, is secreted by macrophages (16) and endothelial cells (25), cell types found in atherosclerotic lesions. Fourth, S-SMase is able to hydrolyze the SM in atherogenic lipoproteins at neutral pH (18). Other possible roles for S-SMase may be in ceramide-mediated cell signaling (26 -30), perhaps after re-uptake of the secreted enzyme into endosomal vesicles; in extracellular sphingomyelin catabolism after nerve injury and during demyelination (16,(31)(32)(33); and in defense against viruses, many of which are enriched in SM (34,35) and can be inactivated by treatment with SMase in vitro. 2 L-and S-SMase are very similar proteins. Previous work from our laboratories has shown that cells transfected with an ASM cDNA overexpress both L-SMase and S-SMase (16), indicating that S-SMase does not arise by alternative processing of the ASM gene. In addition, antibodies made against L-SMase recognize S-SMase, demonstrating that the common mRNA is translated in the same reading frame, and the molecular weights of the enzymes on Western blot are similar (see Ref. 16 and below). Nevertheless, S-SMase requires exogenously added Zn 2ϩ for activation in in vitro assays, whereas L-SMase isolated from cell or tissue homogenates does not (16). In fact, the lack of stimulation of L-SMase by any cations and its lack of inhibition by EDTA has led to a long-standing body of literature labeling L-SMase as a "cation-independent" enzyme (1).
Despite the widespread interest in mammalian SMases in general and in products of the ASM gene in particular, little is known about cellular itineraries of L-SMase and S-SMase or about the mechanism for their apparent difference in zinc dependence. For example, does S-SMase arise by release or exocytosis of L-SMase from lysosomes or by a separate trafficking pathway, and how could two enzymes that are so similar differ in their requirement for zinc? In this report, we show that S-SMase is secreted through a non-lysosomal secretory pathway, and we present evidence that both forms of the enzyme are zinc-activated. According to our model, L-SMase is exposed to cellular Zn 2ϩ during trafficking to lysosomes, in lysosomes, and/or during cell homogenization. Most likely, the Zn 2ϩ dependence of L-SMase has been overlooked because it is already saturated with Zn 2ϩ upon isolation from cell homogenates and thus does not respond to exogenous Zn 2ϩ at the time of assay. Furthermore, as is the case with known zinc metalloenzymes (cf. Ref. 36), the Zn 2ϩ cannot be stripped from L-SMase by simple exposure to EDTA. In contrast, the pathway targeting S-SMase to secretion appears to be relatively sequestered from cellular pools of Zn 2ϩ . Thus, this enzyme requires Zn 2ϩ during subsequent in vitro assay. The information in this report should prove useful for future studies that explore the enzymology, regulation, and functions of these important SMases.
Cells-Monolayer cultures of J774.A1 cells (from the American Type Culture Collection, see Ref. 40) were grown and maintained in spinner culture with DMEM/HI-FBS/PSG as described previously (9,40). Human skin fibroblasts obtained from a patient with type A Niemann-Pick disease (R496L mutation (41)) were grown in DMEM/HI-FBS/PSG. CHO-K1 cells were grown in Hams' F-12 containing 10% HI-FBS and PSG. CHO cells stably transfected with ASM cDNA 3 were maintained in DMEM/HI-FBS/PSG (16). Cells were plated in 35-mm (6-well) or 100-mm dishes in media containing HI-FBS for 48 h. The cells were then washed 3 times with PBS and incubated for 24 h in fresh serumfree media (1 and 6 ml per 35-mm and 100-mm dishes, respectively) containing 0.2% BSA. This 24-h conditioned medium was collected for SMase assays.
Harvesting of Cells and Conditioned Media-Following the incubations described above and in the figure legends, cells were placed on ice, and the serum-free conditioned media were removed. The cells were washed two times with ice-cold 0.25 M sucrose and scraped into 0.3 and 3.0 ml of this sucrose solution per 35-and 100-mm dishes, respectively. Unless indicated otherwise, the scraped cells were disrupted by sonication on ice using three 5-s bursts (Branson 450 Sonifier), and the cellular homogenates were assayed for total protein by the method of Lowry et al. (42) and for SMase activity as described below. The conditioned media were spun at 800 ϫ g for 5 min to pellet any contaminating cells and concentrated 6-fold using a Centriprep 30 (Amicon; Beverly, MA) concentrator (molecular weight cut-off ϭ 30,000). For the experiment in Fig. 5, CHO-K1 cells were incubated in 100-mm dishes in serum-free media and washed as described above. Cells were then scraped in 5 ml of 0.25 M sucrose and broken open under 500 p.s.i. of nitrogen pressure for 1.5 min using a nitrogen cell disruption bomb (Parr Instruments, Moline, IL). Following disruption, a portion of the cells was subjected to brief sonication as described above; this portion of cells is referred to as the cell homogenate. The remainder of disrupted cells was spun at 1300 ϫ g for 5 min to pellet any remaining intact cells and nuclei. This post-nuclear supernatant was collected, and the volume was increased to 15 ml with 0.25 M sucrose and then spun at 16,000 ϫ g for 30 min. The pellet from this centrifugation was resuspended in 1 ml of 0.25 M sucrose and sonicated as above, and this material, as well as the cell homogenate, was assayed for SMase activity.
SMase Assay-As described previously (16), the standard 200-l assay mixture consisted of up to 40 l of sample (conditioned media or homogenized cells; see above) and a volume of assay buffer (0.1 M sodium acetate, pH 5.0) to bring the volume to 160 l. The reaction was initiated by the addition of 40 l of substrate (50 pmol of [ 3 H]sphingomyelin) in 0.25 M sucrose containing 3% Triton X-100 (final concentration of Triton X-100 in the 200-l assay mix ϭ 0.6%). When added, the final concentrations of EDTA and Zn 2ϩ were 5 and 0.1 mM, respectively, unless indicated otherwise. The assay mixtures were incubated at 37°C for no longer than 3 h and then extracted by the method of Bligh and Dyer (43); the lower, organic phase was harvested, evaporated under N 2 , and fractionated by TLC using chloroform/methanol (95:5). The ceramide spots were scraped and directly counted to quantify [ 3 H]ceramide. Typically, our assay reactions contained approximately 20 g of cellular homogenate protein and a volume of conditioned media derived from a quantity of cells equivalent to approximately 50 g of cellular protein.
SDS-Polyacrylamide Gel Electrophoresis and Immunoblotting-Protein samples were boiled in buffer containing 1% SDS and 10 mM dithiothreitol for 10 min, loaded onto 4 -20% gradient polyacrylamide gels, and electrophoresed for 50 min at 35 mA in buffer containing 0.1% SDS (SDS-PAGE). Following electrophoresis, some gels were fixed in methanol/glacial acetic acid/water (5:2:3, v/v) and then silver-stained using reagents from Bio-Rad. Other gels were electrotransferred (100 V for 1.5 h) to nitrocellulose for immunoblotting. For immunoblotting, the nitrocellulose membranes were incubated with 5% Carnation nonfat dry milk in buffer A (24 mM Tris, pH 7.4, containing 0.5 M NaCl) for 3 h at room temperature. The membranes were then incubated with rabbit anti-FLAG-tagged S-SMase polyclonal antiserum (1:2000) in buffer B (buffer A containing 0.1% Tween 20, 3% nonfat dry milk, and 0.1% bovine serum albumin) for 1 h at room temperature. After washing four times with buffer A containing 0.1% Tween 20, the blots were incubated with horseradish peroxidase-conjugated goat anti-rabbit IgG (1:2000) for 1 h in buffer B at room temperature. The membranes were subsequently washed twice with 0.3% Tween 20 in buffer A and twice with 0.1% Tween 20 in buffer A. Finally, the blots were soaked in the enhanced chemiluminescence reagent (NEN Life Science Products) for 2 min and exposed to x-ray film for 1 min.
Glycosidase Treatments-We followed the procedure described by Hurwitz et al. (44). CHO-K1 cells were incubated overnight with serumfree medium (CHO-S-SFM II from Life Technologies, Inc.). Fifty g of 30-fold-concentrated conditioned medium and 50 g of cell homogenate were diluted 1:1 (v/v) with 50 mM sodium acetate buffer, pH 5.0, containing 2% SDS and 20 mM ␤-mercaptoethanol (44). One set of aliquots of the diluted conditioned medium and cell homogenate was treated for 16 h at 37°C with 4 milliunits of endo H. Another set of aliquots was diluted another 15-fold with 50 mM sodium phosphate buffer, pH 7.2, containing 1% Nonidet P-40 and treated for 16 h at 37°C with 100 milliunits of peptide-N-glycanase F. The endo H digest and a trichloroacetic acid pellet of the peptide-N-glycanase F digest (44) were boiled in SDS/dithiothreitol buffer and then electrophoresed and immunoblotted as described above.
Zinc-Chelate Chromatography-We used a modification of the method of Hortin and Gibson (45). Packed 1-ml columns of chelating Sepharose 6B (iminodiacetic acid coupled to agarose gel via a hydrophilic spacer; from Amersham Pharmacia Biotech) were washed with 10 mM sodium acetate buffer, pH 6.0, containing 10 mM EDTA, to leave the column uncharged, or containing 60 mM ZnCl 2 , to charge the column with Zn 2ϩ . The columns were then equilibrated with 50 mM Hepes buffer, pH 7.4, containing 50 mM NaCl. One-ml samples of a 1:1 (v/v) mixture of this equilibration buffer and unconcentrated conditioned medium from ASM-transfected CHO cells (above) were loaded onto the columns and incubated for 15 min at room temperature. The columns were then washed with 7.5 ml of 50 mM Hepes, pH 7.4, containing either 100 mM NaCl or 1 M NaCl, which was collected as 10 0.75-ml fractions. The columns were eluted with 3.75 ml of 50 mM Hepes, pH 7.4, containing 50 mM EDTA plus 1 mM 1,10-phenanthroline, which was collected as 5 0.75-ml fractions. Aliquots of each of the fractions were spotted on nitrocellulose using a slot-blot apparatus and then immunoblotted using goat anti-human L-SMase polyclonal antiserum as described above.
Statistics-Unless otherwise indicated, results are given as means Ϯ S.D. (n ϭ 3); absent error bars in the figures signify S.D. values smaller than the graphic symbols.

RESULTS
Zn 2ϩ Requirement for S-SMase Does Not Involve a Zn 2ϩ -dependent Cofactor-We sought to address how L-and S-SMase acquire their apparent differences in zinc dependence. One explanation would be that the secreted form requires a Zn 2ϩdependent cofactor. Because many lysosomal enzymes undergo proteolytic activation (46), an obvious candidate for a Zn 2ϩ -dependent cofactor would be a zinc metalloproteinase. Five sets of results, however, ruled out this possibility. First, Zn 2ϩ -activated S-SMase can be subsequently inactivated by Zn 2ϩ chelation (see below); reversibility of Zn 2ϩ -induced activation is not consistent with proteolytic activation. Second, inhibitors of zinc metalloproteinases, such as tissue inhibitor of metalloproteinase-1 (TIMP-1) (47) and two different thiol-based peptide inhibitors, HS-CH 2 -R-CH(CH 2 -CH(CH 3 ) 2 )-C)-Phe-Ala-NH 2 and HO-NH-CO-CH 2 -CH(CH 2 CH(CH 3 ) 2 )-C)-naphthyl-Ala-NH-CH 2 -CH 2 -NH 2 (48), did not affect the ability of Zn 2ϩ to activate S-SMase (data not shown). Third, mammalian zinc metalloproteinases require Ca 2ϩ as well as Zn 2ϩ for activity (49), whereas Ca 2ϩ is not a requirement for the activation of S-SMase (16). Fourth, comparison of Zn 2ϩ -activated S-SMase from CHO cells with that of the intracellular (lysosomal) enzyme by immunoblot analysis showed that the activated secreted form had a somewhat higher, not lower, apparent molecular weight (see control data in Fig. 3, below); in addition, S-SMase not activated with Zn 2ϩ had the same apparent molecular weight as Zn 2ϩ -activated S-SMase (data not shown). Fifth, we found that highly purified S-SMase, obtained by either anti-FLAG immunoaffinity purification of a FLAGtagged S-SMase or by concanavalin A chromatography followed by anti-SMase immunoaffinity purification of S-SMase from ASM-transfected CHO cells (16), was ϳ95% Zn 2ϩ -dependent. Thus, neither a zinc metalloproteinase nor any other Zn 2ϩ -dependent cofactor appears to be involved in the activation of S-SMase, suggesting direct activation of S-SMase by Zn 2ϩ .
To support this conclusion further, we sought to demonstrate that S-SMase directly binds Zn 2ϩ by subjecting conditioned media from ASM-transfected CHO cells (16) to zinc-chelate chromatography (cf. Ref. 45 and "Experimental Procedures"). None of the S-SMase from the conditioned medium bound to an uncharged column, whereas Ͼ95% of the S-SMase bound to a Zn 2ϩ -charged column, even when washed with buffer containing 1 M NaCl; all of the bound material was eluted by EDTA plus 1,10-phenanthroline. These data and the previous data are consistent with the conclusion that S-SMase binds and is directly activated by Zn 2ϩ .
Evidence for Direct Activation of L-SMase by Zn 2ϩ -Despite the long-standing tenet that L-SMase is a cation-independent enzyme (2), several lines of evidence initially suggested to us that L-SMase was a zinc-activated enzyme. First, L-SMase and S-SMase come from the same gene and same mRNA in the same reading frame (16), and S-SMase binds and is directly activated by Zn 2ϩ (above). Second, there are seven aminoacyl sequences in the enzyme that are homologous to Zn 2ϩ -binding sequences in known zinc metalloenzymes (50), including one sequence that is very similar to that in another phosphodiesterase enzyme (Table I). Histidine residues in two of these sequences (His-425 and His-575) are highly conserved and are sites of mutations in Niemann-Pick disease. 3 Moreover, His-421 is conserved in mouse ASM and in a homologue of ASM in Caenorhabditis elegans that is zinc-dependent but not in an ASM homologue in C. elegans that is zinc-independent (51). Third, L-SMase shares two other properties of known zinc metalloenzymes, namely inhibition by phosphate ions (1, 2), which are thought to block the Zn 2ϩ -binding pocket(s) in zinc metalloenzymes (52) and inhibition by high concentrations (e.g. 6 mM) of ZnCl 2 (49,53).
To test directly whether L-SMase is a zinc-activated enzyme, we attempted to chelate Zn 2ϩ away from each enzyme in vitro to determine the effect on catalytic activity. In fact, the conclusion by others (54 -59) that L-SMase is a cation-independent enzyme is based partly on the observation that EDTA does not inhibit activity. Many zinc metalloenzymes, however, bind the metal very tightly and thus require more potent chelation, such as by long term incubation with the Zn 2ϩ chelator 1,10-phenanthroline (36). To begin, we conducted our studies with S-SMase, which we know binds and is directly activated by Zn 2ϩ . In Fig. 1A, conditioned medium from J774 macrophages was incubated with either EDTA or Zn 2ϩ (1st two bars) and then assayed for SMase activity. As we have reportedly previously (16), the S-SMase is markedly activated by Zn 2ϩ . We then took an aliquot of Zn 2ϩ -activated S-SMase and incubated it for 18 h with EDTA plus 1,10-phenanthroline in an attempt to chelate the enzyme-bound Zn 2ϩ . As shown in the 3rd bar in Fig. 1A, this treatment resulted in an approximately 50% loss of activity; treatment with EDTA alone did not affect enzyme activity (data not shown). Finally, the partially inactivated S-SMase was dialyzed against Zn 2ϩ -containing buffer (4th bar), which restored activity to the original level observed when Zn 2ϩ was added initially to the conditioned medium (compare 4th and 2nd bars in Fig. 1A).
Next, we examined cellular (i.e. lysosomal) SMase 4 (Fig. 1B). As reported previously by others (1, 2) and by us (16), and in contrast to the situation with the secreted enzyme, L-SMase in cell homogenates shows maximal activity without added Zn 2ϩ and is not inhibited by EDTA (1st three bars of Fig. 1B). We then incubated the original, active cellular enzyme (i.e. not exposed to exogenous Zn 2ϩ in vitro) with EDTA plus 1,10phenanthroline for 8 h (4th bar). This treatment resulted in almost total inactivation of the enzyme, and substantial activity was restored by dialyzing against Zn 2ϩ (5th bar) or by directly adding back excess Zn 2ϩ (not shown). 5 The actual degree of inhibition by chelation and re-activation by Zn 2ϩ dialysis differed somewhat between the secreted and cellular SMases, perhaps due to a lesser stability of the cellular enzyme under the incubation conditions employed. Nonetheless, the overall patterns of inhibition and reactivation shown in Fig. 1, together with the lines of evidence mentioned earlier, provide strong evidence that L-SMase, like S-SMase, is a zinc-activated enzyme.
The Difference in Requirement for Zn 2ϩ in the in Vitro Assays of L-and S-SMase: Differential Zn 2ϩ Affinity Versus Differential Exposure to Cellular Zn 2ϩ Prior to the Assay-One possible explanation for the difference in Zn 2ϩ requirement in the in vitro assays of L-and S-SMase is that the two enzymes would both be exposed to the same, although limiting, concentration of intracellular Zn 2ϩ but that the lysosomal enzyme would have a higher affinity for the cation, perhaps owing to a difference in post-translation modification. Thus, L-SMase would already have bound Zn 2ϩ at the time of the assay. The secreted enzyme would have lower affinity for Zn 2ϩ , and thus excess exogenous Zn 2ϩ would have to be added for activation in vitro.
We sought to estimate the relative Zn 2ϩ affinities of these two enzymes by assaying their inactivation as a function of increasing exposure to metal chelators (36). Therefore, we incubated a cellular homogenate of J774 macrophages and the conditioned medium from these cells with EDTA plus the 1,10phenanthroline for increasing times at 4°C and then assayed these two fractions for SMase activity at each time point. As expected (above), both enzymes lost activity with increasing duration of chelation (Fig. 2), whereas incubation in the absence of the chelators for 8 h at 4°C resulted in no loss of either secreted or cellular SMase activity (not shown). The data show that cellular SMase activity decreased at a greater rate and to a greater extent than secreted SMase activity, which is not consistent with the hypothesis that L-SMase has a higher affinity for Zn 2ϩ than S-SMase.
The other possibility is that both enzymes bind Zn 2ϩ with similar affinities, but only the lysosomal enzyme would be exposed to pools of intracellular Zn 2ϩ prior to the assay; this exposure to Zn 2ϩ could occur during transit to or residence in lysosomes and/or during preparation of the cell homogenate. Indeed, studies in many different cell types have shown that Zn 2ϩ is distributed in various intracellular organelles, including lysosomes (60) and cytoplasmic vesicles (61). This model makes several assumptions and predictions that we tested experimentally. First, the idea that exposure of L-SMase to Zn 2ϩ could occur during transit to or residence in lysosomes assumes that S-SMase does not simply arise by exocytosis of lysosomal vesicles (cf. Ref. 46). To test this important point directly, we obtained data on the carbohydrates of L-and S-SMase. The lysosomal targeting of L-SMase is typical for most lysosomal enzymes as follows: acquisition of Asn-linked high mannose oligosaccharides (44,62) followed by phosphorylation of some of the mannose residues and shuttling from the trans-Golgi network to early endosomes/late endosomes/ prelysosomes via mannose-phosphate receptor-containing vesicles (46,(63)(64)(65). In the typical (i.e. non-lysosomal) secretory pathway, however, the original high mannose oligosaccharides on the SMase would be expected to undergo processing to complex oligosaccharides during transit through the Golgi (44,46,62,63). Therefore, we incubated aliquots of conditioned medium and homogenates from untransfected CHO cells with endo H, which is specific for high mannose-type Asn-linked oligosaccharides (66); other aliquots were incubated with peptide-N-glycanase F, which cleaves both high mannose and complex Asn-linked oligosaccharides (66). These incubations were then analyzed by anti-SMase immunoblots. As shown in Fig. 3, S-SMase was completely resistant to endo H but susceptible to peptide-N-glycanase F, indicating the presence of complex-type Asn-linked oligosaccharides. In contrast, L-SMase was susceptible to both glycosidases, which confirms that this form of the enzyme has high mannose-type oligosaccharides. In addition, comparison of the N-terminal amino acid sequences of purified S-SMase and L-SMase from CHO cells transfected with FLAGtagged ASM revealed that L-SMase, but not S-SMase, underwent N-terminal proteolytic processing typical of lysosomal enzymes (46) (see following section). These data indicate that S-SMase does not arise via exocytosis of lysosomes or vesicles in transit to lysosomes but rather through the typical secretory 4 Consistent with prior literature (cf. Ref. 2), SMase activity in whole cell homogenates using the standard acidic micellar assay, particularly when EDTA is added, has been equated with "lysosomal" SMase activity. Other types of cellular SMase are not active at acidic pH in this assay, and one of these other SMases also requires Mg 2ϩ for activity (2). 5 In pilot experiments, we found that 1,10-phenanthroline alone was not as effective as EDTA plus phenanthroline in inhibiting the activity of S-and L-SMase. One possible explanation is that the enzymes bind another divalent cation in addition to Zn 2ϩ , and removal of this cation by EDTA facilitates the removal of Zn 2ϩ by 1,10-phenanthroline (cf. Refs. 36 and 53). Whatever the mechanism, the fact that 1,10-phenanthroline alone does not inhibit S-or L-SMase argues against an unlikely alternative interpretation of the data in Fig. 1, namely that 1,10phenanthroline is a direct SMase inhibitor that becomes inactive as an inhibitor when the compound binds Zn 2ϩ .  b X refers to any amino acid. c This histidine residue is also conserved in mouse ASM and in the C. elegans ASM that is zinc-dependent but not in the C. elegans ASM that is zinc-independent (51). d These histidine residues are sites of mutations in Niemann-Pick disease.
pathway. These distinctly divergent pathways provide the opportunity for one of the SMase to be exposed to different levels of cellular Zn 2ϩ than the other form of the enzyme. Second, the model implies that it is the sequestration of S-SMase away from Zn 2ϩ in the lysosomal pathway, not the oligosaccharide processing of S-SMase per se, that is responsible for the dependence of S-SMase on exogenous Zn 2ϩ . To test this idea, we looked for a system in which cells secreted S-SMase that was mannose-phosphorylated but not exposed to the lysosomal pathway. We took advantage of the fact that in transfected cells that massively overexpress a lysosomal enzyme, a substantial portion of the mannose-phosphorylated form of this enzyme saturates the mannose 6-phosphate receptor shuttling mechanism (67). Therefore, these cells secrete lysosomal enzymes that are mannose-phosphorylated but that have not been exposed to the lysosomal targeting pathway or to lysosomes (67). Indeed, we found that ϳ80% of S-SMase from our ASM-transfected CHO cells (above) bound to a mannose 6-phosphate receptor affinity column and could be eluted with mannose 6-phosphate (cf. Ref. 68). This secreted SMase was 98% Zn 2ϩ -dependent, whereas the lysosomal SMase in these cells required no added Zn 2ϩ when assayed in cellular homogenates. Thus, even though the S-SMase from these overexpressing cells underwent the typical carbohydrate modifications of a lysosomal, not a secretory, enzyme, it had the same degree of Zn 2ϩ dependence seen with S-SMase from non-transfected cells. This finding is consistent with our model, since the S-SMase from these cells bypassed lysosomal targeting and thus would not be exposed to cellular pools of Zn 2ϩ . Third, the model predicts that secreted, Zn 2ϩ -dependent S-SMase, when endocytosed by cells, delivered to lysosomes, and assayed in cell homogenates would be exposed to cellular Zn 2ϩ and thus no longer require exogenously added Zn 2ϩ . To test this prediction, we used highly purified secreted FLAG-tagged SMase, which is 99.6% Zn 2ϩ -dependent (see above), and fibroblasts from a patient with type A Niemann-Pick disease, which completely lack both L-and S-SMase activities (16). To introduce the FLAG-S-SMase into intact Niemann-Pick fibroblasts, we added the enzyme to media on living cells and then incubated for 16 h at 37°C. In pilot experiments, we found that these cells can endocytose S-SMase and target it to lysosomes in a catalytically active form, as evidenced by a substantial reduction in lysosomal SM mass, which otherwise accumulates in these mutant fibroblasts (cf. Refs. 44 and 69). 2 After the incubation, media and cells were harvested, and cells were homogenized, and the media and sonicated cell homogenates were assayed for SMase activity (Fig. 4). As expected, the SMase activity that remained in the media (i.e. the portion of the enzyme that was not internalized) was almost entirely Zn 2ϩ -dependent. In contrast, the SMase activity in the sonicated cell homogenates, which originated entirely via internalization of the exogenously added secretory enzyme, was maximally activated in the absence of Zn 2ϩ (Fig. 4). Addition of Zn 2ϩ not only failed to increase the cellular enzymatic activity but for unclear reasons produced a somewhat lower activity. To demonstrate that the Zn 2ϩ -independent SMase activity in the cell homogenates was due to exposure of the enzyme to cellular pools of Zn 2ϩ , homogenates were subsequently incubated for 24 h with EDTA plus 1,10-phenanthroline, to chelate Zn 2ϩ (see Figs. 1 and 2), or with buffer alone as a control. As shown in Fig. 4, the cellular SMase activity was specifically inhib-FIG. 1. Sequential chelation and addition of Zn 2؉ to secreted and intracellular SMase. Serum-free conditioned medium and a cell homogenate from J774 macrophages were prepared as described under "Experimental Procedures." A, conditioned medium (CM) was assayed for SMase activity using 250 nM [ 3 H]sphingomyelin in Triton X-100 micelles in the presence of either 5 mM EDTA (1st bar) or 0.1 mM ZnCl 2 (2nd bar) for 1 h at 37°C at pH 5.0. An aliquot of the zinc-activated conditioned medium was then incubated for 18 h in the presence of 10 mM EDTA, 10 mM 1,10-phenanthroline, and 0.6% Triton X-100 at 4°C (EDTA PHNANTH) and then assayed for SMase activity (3rd bar). After chelation, a portion of this partially inactivated conditioned medium was then dialyzed against zinc-containing buffer C (150 mM NaCl, 10 mM Tris-HCl, 1 mM ZnCl 2 , 0.6% Triton X-100, pH 7.4) for 18 h and then assayed for SMase activity (4th bar). B, an aliquot of cell homogenate (Cell Homgn) was assayed for SMase activity in the presence of assay buffer alone (1st bar), 5 mM EDTA (2nd bar), or 0.1 mM ZnCl 2 (3rd bar). Another aliquot of cell homogenate was incubated for 8 h in the presence of 10 mM EDTA, 10 mM 1,10-phenanthroline, and 0.6% Triton X-100 (EDTA PHNANTH) at 4°C and then assayed for SMase activity (4th bar). An aliquot of this chelator-treated cell homogenate was then dialyzed against zinc-containing buffer C for 18 h and then assayed for SMase activity (5th bar).
ited by Zn 2ϩ chelation. 6 Similar results were obtained when the Niemann-Pick cells were incubated with native (i.e. non-FLAG-tagged) S-SMase (data not shown). These data demonstrate that Zn 2ϩ -dependent secreted SMase can become maximally activated by exposure to intracellular pools of Zn 2ϩ during internalization, intracellular sorting, and/or cellular sonication.
L-SMase Activity from a Lysosome-rich 16,000 ϫ g Pellet of CHO Cells Demonstrates Some Zinc Dependence-The Zn 2ϩ dependence of L-SMase and previous work demonstrating discrete intracellular Zn 2ϩ pools that can change under certain metabolic conditions (cf. Refs. 70 and 71) led us to consider the hypothesis that Zn 2ϩ availability to lysosomes and to L-SMase might be involved in the regulation of this enzyme. A prediction of our hypothesis is that L-SMase may not always be maximally stimulated by intracellular Zn 2ϩ . In the standard L-SMase assay, cells or tissues are completely homogenized, and the cell homogenate is assayed. As shown in Fig. 5A for CHO-K1 cells disrupted by sonication, the intracellular enzyme is maximally activated, and exogenous Zn 2ϩ has no effect (Fig.  5A). To obtain a less damaged lysosomal preparation, a separate aliquot of these CHO cells was disrupted under 500 p.s.i. of nitrogen pressure for 1.5 min, and a 16,000 ϫ g pellet was isolated, which consists of intact lysosomes, as well as mitochondria and peroxisomes (cf. Ref. 59). 2 This 16,000 ϫ g pellet was then sonicated and assayed for SMase activity. Remarkably, under these conditions, the enzyme was only ϳ50% activated and was substantially stimulated by exogenous Zn 2ϩ (Fig. 5A).
To probe this finding further, we isolated a 16,000 ϫ g pellet from CHO cells transfected with FLAG-tagged ASM (see above) and then purified the enzyme by anti-FLAG affinity chromatography, followed by gel filtration and a second round of anti-FLAG affinity chromatography. The purified enzyme migrated as a single band on silver-stained SDS-PAGE slightly below where ASM-S-SMase migrates (data not shown; see Fig. 3). N-terminal amino acid sequence analysis revealed that enzyme began with the sequence GHPARLH, whereas S-SMase, which was purified from the conditioned medium of these cells, began with the sequence HPLSPQGHPARLH. Thus, the enzyme purified from the 16,000 ϫ g pellet meets several criteria for L-SMase as follows: isolation from a lysosome-rich cellular fraction; N-terminal proteolytic processing (46); and more rapid migration on SDS-PAGE than S-SMase, which is due to both proteolytic processing and to differences in oligosaccharide structure (see Fig. 3). We then tested this purified L-SMase for zinc dependence and found that its enzymatic activity was increased 4.7-fold in the presence of Zn 2ϩ (Fig. 5B). When zinc-activated L-SMase was dialyzed extensively against 10 mM EDTA, which does not remove activating Zn 2ϩ from the enzyme (see Fig. 1), and then dialyzed against buffer free of both zinc and EDTA, the enzymatic activity was not changed (data not shown). Thus, the activating effect of zinc cannot be explained by free zinc (i.e. Zn 2ϩ not bound to the SMase) affecting the substrate or some other component of the reaction mixture. These data, together with those in Fig. 5A, have two major implications. First, they suggest that L-SMase in these cells encounters subsaturating levels of Zn 2ϩ during transit to lysosomes and/or after subsequent storage there, raising the possibility of regulation of L-SMase by intracellular Zn 2ϩ availability (see "Discussion"). When cells are disrupted by sonication, we propose that sequestered pools of cellular Zn 2ϩ are released, which leads to saturation of the enzyme with Zn 2ϩ . 7 Second, these data provide further evidence that L-SMase is a zinc metalloenzyme. 6 Note that treatment with chelators results in almost total inhibition of the internalized, activated S-SMase. This near total effect of chelators is similar to that observed with L-SMase but not with zinc-activated S-SMase from conditioned medium, which is only partially inhibited by chelators (Figs. 1 and 2). Therefore, when S-SMase is delivered to lysosomes, it appears to be converted into a form that allows more complete chelation of its Zn 2ϩ . It is possible that this conversion is related to the N-terminal proteolytic processing of SMase in lysosomes (cf. Ref. 46 and data in next section).

FIG. 2. Time course of inactivation of secreted and intracellular SMase by Zn 2؉ chelation.
Serum-free conditioned medium, preactivated by incubation with 0.1 mM ZnCl 2 at 37°C for 10 min, and cell homogenate from J774 macrophages were incubated in the presence of 10 mM EDTA, 10 mM 1,10-phenanthroline, and 0.6% Triton X-100 at 4°C for the indicated times. Each sample was then assayed for SMase activity in the presence of 5 mM EDTA for 1 h at 37°C at pH 5.0. The maximum values (100% on the y axis) for the cell homogenate and conditioned medium were, respectively, 72.0 Ϯ 0.3 and 17.1 Ϯ 0.3 pmol [ 3 H]ceramide/mg protein/h.

FIG. 3. Susceptibility of S-and L-SMase to endo H and peptide-
N-glycanase F. Aliquots of concentrated conditioned media (as a source of S-SMase) and cellular homogenate (as a source of L-SMase) from CHO cells were incubated in the absence or presence of endo H or peptide-N-glycanase F (PNGase F) for 16 h at 37°C and then subjected to SDS-PAGE. The electrophoresed proteins were transferred to a nitrocellulose membrane and then immunoblotted using rabbit anti-FLAG-tagged S-SMase antiserum.

DISCUSSION
Our model to explain the cellular trafficking and apparent difference in Zn 2ϩ dependence of L-SMase versus S-SMase is shown in Fig. 6. Based upon our previous work, the ASM gene gives rise to a common precursor protein (16), which is then modified by typical high mannose oligosaccharide residues (44,62,63). We propose that this mannosylated precursor then traffics into either the lysosomal or the secretory pathway. In the lysosomal pathway, the SMase undergoes modification and trafficking that is typical for lysosomal enzymes: acquisition of mannose-phosphate residues by the sequential action of N-acetylglucosamine-1-phosphotransferase and N-acetylglucosamine phosphodiesterase on the mannose residues of the precursor (Fig. 3 and Refs. 44 and 63). Vesicles containing mannose-phosphate receptors then shuttle this modified SMase to early endosomes or late endosomes/prelysosomes (63,72,73), and we propose that at some point along this pathway the enzyme encounters cellular Zn 2ϩ and thus becomes at least partially activated. As mentioned under "Results," L-SMase, at least in CHO cells, appears to be exposed to subsaturating concentrations of Zn 2ϩ in lysosomes and thus is potentially subject to regulation by changes in Zn 2ϩ availability.
L-SMase has been studied for many years, particularly in the context of its absence in a human disease, namely types A and B Niemann-Pick disease (1,2). Throughout this period of study, the enzyme has been reported to be cation-independent (1, 2). Although plasma emission spectrometry and x-ray crystallography of large amounts of homogeneously purified L-SMase

FIG. 4. Zn 2؉ dependence of S-SMase after internalization by type A Niemann-Pick cells.
Skin fibroblasts from a patient with type A Niemann-Pick disease (NP-A Cells) were grown to 90% confluency in medium supplemented with 10% HI-FBS. The cells were then washed three times with PBS and incubated in serum-free media containing 0.2% BSA and 850 ng of immunoaffinity isolated FLAG-tagged S-SMase (Purified S-SMase)/ml at 37°C for 16 h. The cell homogenate and incubation medium were then prepared as described under "Experimental Procedures," and aliquots were assayed for SMase activity in the presence of either 5 mM EDTA (hatched bars) or 0.1 mM ZnCl 2 (solid bars) for 2 h at 37°C at pH 5.0. Other aliquots of the cell homogenate were incubated in the presence of 0.6% Triton X-100 alone (No Chelators) or in the presence of 10 mM EDTA, 10 mM 1,10-phenanthroline, and 0.6% Triton X-100 (EDTA ϩ Phnanth) for 18 h at 4°C and then assayed for SMase activity in the presence of 5 mM EDTA. All values are total SMase activity in each sample.
FIG. 5. Zn 2؉ dependence of SMase in a sonicated cell homogenate and 16,000 ؋ g pellet and of L-SMase purified from a 16,000 ؋ g pellet. A, CHO-K1 cells were grown to 90% confluency in Hams' F-12 media containing 10% HI-FBS and PSG and then washed 3 times with PBS before incubation in serum-free medium containing 0.2% BSA for 24 h. The cells were then washed 3 times in 0.25 M sucrose, and the total cell homogenate and the 16,000 ϫ g cellular subfraction were isolated as described under "Experimental Procedures." Twelve micrograms of the cell homogenate (Homogenate) and 2.5 g of the 16,000 ϫ g pellet (16K-ϫg Pellet) were assayed for SMase activity in the presence of either 5 mM EDTA (hatched bars) or 0.1 mM ZnCl 2 (solid bars) for 1 h at 37°C at pH 5.0. The total SMase activities (pmol/h) in the cell homogenate were 3.38 Ϯ 0.04 (EDTA) and 3.90 Ϯ 0.06 (zinc), and the total SMase activities in the 16,000 ϫ g pellet were 2.10 Ϯ 0.02 (EDTA) and 3.95 Ϯ 0.13 (zinc). B, L-SMase was purified from the 16,000 ϫ g pellet of CHO cells transfected with FLAG-tagged ASM and assayed for activity in the presence of EDTA or zinc as above.
will be needed to define the stoichiometry and location of zinc interaction with L-SMase, the data in this report strongly support the conclusion that this enzyme is, indeed, a zincactivated enzyme. The most compelling data are those in Fig.  1B, Fig. 5, and Table I, especially footnotes d and e. In fact, the information in Table I raises the possibility that some cases of Niemann-Pick disease may be due to mutations in the zincbinding domain, possibly resulting in defective binding of Zn 2ϩ and thus loss of enzymatic activity. Along these lines, He et al. 3 have shown that chelation of Zn 2ϩ from SMase by 1,10-phenanthroline results in defective SM binding to the enzyme. We believe the reason why this fundamental property of this widely studied enzyme has been overlooked is because the enzyme at the time of isolation from whole cell homogenates, which has been the source of L-SMase for the previous studies (54 -59), is already tightly bound to Zn 2ϩ . In view of the data in Fig. 5, some of this Zn 2ϩ may come from pools of zinc that are released during the homogenization of cells or tissues. Thus, exogenous Zn 2ϩ is not needed for the in vitro assay, and typical short-term EDTA chelation incubations will not strip the enzyme of its metal, similar to findings with other known zinc metalloenzymes (36).
To explain the origin of S-SMase, we propose that a portion of the common precursor, via a potentially regulated process (see below), bypasses N-acetylglucosamine-1-phosphotransferase and thus is directed into the secretory pathway, not the lysosomal targeting pathway (63) (Fig. 6). The difference in susceptibility of S-and L-SMase to endo H (Fig. 3) and the differences in N-terminal proteolytic processing (see "Results") provide direct support for this component of the model. Importantly, our data suggest that SMase in the secretory pathway is not exposed to pools of cellular Zn 2ϩ , thus explaining the requirement for exogenously added Zn 2ϩ when the secreted enzyme is assayed in vitro. As mentioned under "Results," however, the subcellular location of Zn 2ϩ may be subject to cell-type variation or regulation (60,70,71). For example, recent data from Vallee and colleagues (74) suggest that the redox state of the cell may be an important factor in the transfer of zinc from metallothionein, a cellular zinc reservoir, to intracellular zincdependent enzymes. Therefore, it is possible that S-SMase may, under certain circumstances or in certain cell types, be fully or partially Zn 2ϩ -independent. In fact, we have observed that SMase secreted by endothelial cells, unlike that secreted by macrophages (16), is active in the absence of Zn 2ϩ and stimulated only 2-fold by exposure to exogenous Zn 2ϩ (25). According to this model and work by other researchers (46,63), the key step that would determine the fate of SMase is catalysis of the common mannosylated precursor by N-acetylglucosamine-1-phosphotransferase. Extensive work by Kornfeld and colleagues (75)(76)(77) has shown that N-acetylglucosamine-1-phosphotransferase recognizes a particular threedimensional structure of lysosomal enzyme precursors, and induced modifications that alter this structure can have profound effects on lysosomal enzyme modification and targeting. Moreover, these workers have found that at least one enzyme, bovine DNase I, is a suboptimal substrate for the phosphotransferase, thus presumably giving rise to both intralysosomal and secretory forms (78) 8 If the enzymes that undergo secretion by this mechanism can function at neutral pH (see below) or if the cells are in an acidic environment, this process may enable cells to acquire two groups of functions from a single enzyme, namely functions in lysosomes and functions in the extracellular milieu. In the case of S-SMase, there is an additional requirement for extracellular Zn 2ϩ , which is known to exist in sufficient extracellular concentrations in vivo to activate the enzyme (cf. Refs. 16 and 53). Interestingly, we found that certain cytokines increase the secretion of SMase 8 S. Kornfeld, personal communication.
FIG. 6. Model of cellular pathways giving rise to lysosomal and secreted SMase. The ASM gene gives rise to a common mannosylated precursor protein (SMase Precursor-mannose) that gets shuttled into either the lysosomal trafficking pathway (SMase-M6P) or the secretory pathway (SMase) (see data in Fig. 3). According to the model, the enzyme is exposed to cellular pools of Zn 2ϩ in the lysosomal pathway but not in the secretory pathway. As is evident from the data in Fig. 5, however, L-SMase may not always be fully saturated with Zn 2ϩ . Potential sites of regulation are marked by the asterisks; in addition, based upon data with S-SMase from endothelial cells, 3 it is possible that a portion of S-SMase may acquire some Zn 2ϩ prior to secretion (see text for details). In this diagram, we show Zn 2ϩ enrichment of early endosomes/late endosomes/ prelysosomes, which are the acidic vesicles where lysosomal enzymes are initially delivered (63); in the case of late endosomes and prelysosomes, hydrolysis by these enzymes can occur (82). It is possible, however, that mature lysosomes, in addition to or instead of endosomes/prelysosomes, are enriched with Zn 2ϩ . GlcNAc-P-Trf'ase, N-acetylglucosamine-1-phosphotransferase; GlcNAc-PD'ase, N-acetylglucosamine phosphodiesterase; M6P, mannose-6-phosphate; MPR, mannose 6-phosphate receptor. from endothelial cells without affecting L-SMase activity, suggesting that the phosphotransferase reaction or perhaps another critical step responsible for determining the fate of SMase may be subject to specific regulation (25). Finally, C. elegans has two separate genes that encode SMases that are highly homologous to mammalian S-and L-SMase; one of these SMases is almost entirely secreted and the other is mostly intracellular (51). Thus, organisms evidently need both intracellular and extracellular SMases; C. elegans has two genes to meet these needs, and it appears as if higher species (i.e. mammals) have evolved the mechanism described above to meet these needs using one gene.
Our current data and previous work by others (79) indicate difficulties with the prior nomenclature of these SMases. First, we now know that both forms of the enzymes are zinc-activated enzymes, and so our previous designation of the secreted form as "Zn-SMase" (16) is obsolete. Second, the "acid SMase" nomenclature reflects the acid pH optima of the lysosomal and secreted forms of the enzyme in standard in vitro detergentbased micellar assays and the ability of the lysosomal form to function in the acid environment of lysosomes (1). Kinetic studies, however, have shown that acid pH is needed only for proper interaction of the enzyme with the SM in these micelles (i.e. K m ) and that V max for hydrolysis is relatively pH-independent (79). Furthermore, we have recently demonstrated that S-SMase can hydrolyze the SM of certain lipoproteins quite well at neutral pH (18). Thus, the SM in certain physiological substrates may be in an orientation that allows ready interaction with the enzyme at neutral pH, which, based upon the above-mentioned kinetic data, would then result in neutral SM hydrolysis. For these reasons, and since lysosomes and conditioned media of cells contain no other known SMase activity (1,2,16,80,81), we suggest that the nomenclature in this paper (L-SMase and S-SMase) is preferred. To maintain consistency with prior literature, however, we still refer to their common gene of origin as the ASM gene.
The original impetus for the current mechanistic study was evidence gathered by our laboratories supporting a role for an extracellular arterial wall SMase in atherogenesis (10 -13). As outlined in the Introduction, S-SMase is a leading candidate for this arterial wall activity. Furthermore, there is evidence that one or more products of the ASM gene triggers ceramide-mediated cell-signaling processes (14, 26 -30), and S-SMase is a candidate since it would have access to the extracellular leaflet of the plasma membrane, which is where most cellular sphingomyelin is located (42). It is also possible that S-SMase plays roles in extracellular SM catabolism in the central nervous system and in anti-viral host defense mechanisms (see Introduction). In this context, the information reported herein on the cellular trafficking and zinc dependence of the S-SMase and L-SMase should prove useful in further regulatory studies on these enzymes and in designing strategies to test their possible roles in atherogenesis, ceramide-mediated cell signaling, and possibly other physiologic and pathophysiologic processes.