Structural Basis of the Sphingomyelin Phosphodiesterase Activity in Neutral Sphingomyelinase from Bacillus cereus*

Sphingomyelinase (SMase) from Bacillus cereus (Bc-SMase) hydrolyzes sphingomyelin to phosphocholine and ceramide in a divalent metal ion-dependent manner. Bc-SMase is a homologue of mammalian neutral SMase (nSMase) and mimics the actions of the endogenous mammalian nSMase in causing differentiation, development, aging, and apoptosis. Thus Bc-SMase may be a good model for the poorly characterized mammalian nSMase. The metal ion activation of sphingomyelinase activity of Bc-SMase was in the order Co2+ ≥ Mn2+ ≥ Mg2+ » Ca2+ ≥ Sr2+. The first crystal structures of Bc-SMase bound to Co2+, Mg2+, or Ca2+ were determined. The water-bridged double divalent metal ions at the center of the cleft in both the Co2+- and Mg2+-bound forms were concluded to be the catalytic architecture required for sphingomyelinase activity. In contrast, the architecture of Ca2+ binding at the site showed only one binding site. A further single metal-binding site exists at one side edge of the cleft. Based on the highly conserved nature of the residues of the binding sites, the crystal structure of Bc-SMase with bound Mg2+ or Co2+ may provide a common structural framework applicable to phosphohydrolases belonging to the DNase I-like folding superfamily. In addition, the structural features and site-directed mutagenesis suggest that the specific β-hairpin with the aromatic amino acid residues participates in binding to the membrane-bound sphingomyelin substrate.

The catalytic mechanism of the sphingomyelin hydrolytic activity remains to be elucidated in atomic detail, as there are no crystal structures of SMase in complex with the essential divalent metal ions. The sphingomyelin hydrolytic activity of Bc-SMase is believed to proceed in the manner of acid base catalysis, in which His-296 is proposed to generate an activated water and the essential Mg 2ϩ ion at Glu-53 is suggested to stabilize a negatively charged transition state. The proposed catalytic mechanism of Bc-SMase is similar to that of bovine DNase I. In fact, Bc-SMase and bovine DNase I are homologous proteins and share a common architecture of conserved putative catalytic amino acid residues (7). However, the proposed catalytic mechanism does not fully explain the role of the essential divalent metal ion, i.e. the divalent metal ion type dependence for hydrolytic catalysis, because of the lack of the bound essential metal ions in all the currently available structures.
Bc-SMase and neutral sphingomyelinase (nSMase) in mammalian cells share similar metal ion dependence and considerable amino acid sequence identity (20%), including conserved residues involved in divalent metal ion binding, and are thus believed to have a similar hydrolytic mechanism. SMases in mammalian cells are classified into the following three groups according to the optimum pH of the SM hydrolytic activity: neutral SMase, acid SMase, and alkaline SMase (8). The detailed mechanism of the sphingomyelin hydrolysis activity of Bc-SMase may provide insight into sphingolipid metabolism in mammalian cells (9 -11).
The only available structure of SMase is from the bacterium, Listeria ivanovii, reported recently (7). The bacterial SMase was confirmed to be a member of the DNase I-like folding superfamily (12)(13)(14), and the putative active site amino acid residues of the bacterial SMase were found to be geometrically identical to the corresponding amino acid residues of enzymes in the DNase I-like folding superfamily. The bacterial SMase differs, however, in having a unique hydrophobic ␤-hairpin structure.
The L. ivanovii SMase structure allowed elucidation of a number of key features of the catalytic mechanism of bacterial SMases. However, structural information on the role of the essential divalent metal ions in catalysis is still lacking, and the hydrolytic molecular mechanism remains elusive.
Bacterial SMases have been reported to bind to the cell surface when catalyzing hemolysis. The crystal structure of SMase from L. ivanovii indicated that interaction with the membrane would occur at the unique hydrophobic ␤-hairpin region (7). This was inferred based on two indirect observations. First, there are two solvent-exposed aromatic amino acid residues at the top of the ␤-hairpin. Second, the corresponding region of structurally similar enzymes is a short loop. The role of the ␤-hairpin remains to be confirmed experimentally.
In this study we propose that the water-bridged double divalent metal ions are essential in the hydrolytic activity of Bc-SMase based on the high resolution crystal structures of Bc-SMase with the metal ions Co 2ϩ , Mg 2ϩ , and Ca 2ϩ , respectively. It is the divalent metal ion bound to His-296 that activates the water molecule as the nucleophile in sphingomyelin hydrolysis. Simplified assays using very low concentrations of substrate indicated a divalent metal ion-type dependence for the hydrolytic activity of Bc-SMase compatible with the divalent metal ion architecture. In addition, we confirmed experimentally that the ␤-hairpin, the unique feature of hemolytic SMases, participates in not only the binding of Bc-SMase to the cell membrane and SM liposomes but the hydrolytic activity of Bc-SMase.

EXPERIMENTAL PROCEDURES
Expression and Purification-Bc-SMase was overexpressed in Bacillus subtilis ISW1214 transformed with the plasmid vector, pHY300PLK, carrying cDNA of Bc-SMase cloned from B. cereus IAM 1029. The Bc-SMase was secreted into the culture medium. The 80% (w/v) ammonium sulfate fraction of the harvested culture medium was sequentially purified through a Cu 2ϩ column and a DEAE-Sepharose column. The purity of samples was verified using SDS-PAGE stained with Coomassie Brilliant Blue. The Bc-SMase was observed as a single band at 35 kDa.
Preparation of Liposomes-Multilamellar liposomes composed of phospholipid and cholesterol were prepared according to previous reports (15,16). A mixture of phospholipid (0.5 mol) and chloroform was evaporated under reduced pressure to form a lipid film on the wall of a conical bottomed flask. After drying under reduced pressure for 1 h, the lipid film was hydrated by vortexing at 45-50°C or above the phase transition temperature of the SM used, in 100 l of carboxyl fluorescein and 0.9% (w/v) NaCl. The liposome suspensions were centrifuged at 22,000 ϫ g for 15 min at 4°C to remove the non-encapsulated marker and washed three times by centrifugation in 20 mM Tris-HCl buffer (pH 7.5) containing 0.9% (w/v) NaCl (Tris-buffered saline). The resulting liposomes were suspended in 200 l of Tris-buffered saline.
SMase Activity Assay-The hemolytic activity of the enzyme was determined by the amount of hemoglobin released from sheep erythrocytes using a method described previously (17). The reaction mixture containing the enzyme at various concentrations, 3% (w/v) sheep erythrocytes, 20 mM Tris-HCl buffer (pH 7.5), 3 mM MgCl 2 , and 0.9% (w/v) NaCl, was incubated for 30 min at 37°C and then centrifuged at 500 ϫ g for 3 min in order to prepare the test aliquot. Lysis was determined spectrophotometrically at A 550 nm .
The SM liposome disruption activity was determined by the amount of carboxyl fluorescein in the test aliquot. The SM liposome solution containing 20 mM Tris-HCl (pH 7.5), 1 mM MgCl 2 , and 0.9% (w/v) NaCl was incubated with Bc-SMase for 30 min at 37°C. The wavelengths for excitation and measurement were 490 and 530 nm, respectively.
All hydrolytic activities of the enzymes were measured using the 14 Clabeled SM, [N-methyl-14 C]SM (Amersham Biosciences), as substrate (18). The concentration of the labeled SM was well below the critical micelle concentration of SM. The enzyme was dialyzed against a solution containing 20 mM Tris-HCl (pH 7.5), 0.1% (v/v) Triton X-100, and 1 mM EDTA to remove the contaminant metal ions. The reaction mixture containing the dialyzed enzyme (50 ng/ml) and an additional divalent metal ion (5 mM) was preincubated at 37°C for 30 min, and then the reaction mixture was incubated with 1.25 M [N-methyl- 14 C] SM at 37°C for various periods. The reaction was terminated by the addition of the stop solution (CHCl 3 /MeOH, 2:1), and the radioactivity of the labeled product in the chloroform phase was measured by the liquid scintillation counter LSC-6100 (Aloka).
Membrane Binding Assay of Bc-SMase-The membrane binding assay was performed by surface plasmon resonance (SPR) analysis using a Biacore3000 system and the associated analysis software package (Biacore). The enzyme solutions at concentrations of 10, 5, and 0 g of enzyme/ml were applied to the L1 sensor chips coated with or without SM liposome at a flow rate of 10 l/min in a running buffer (20 mM Tris-HCl (pH 7.5), 3 mM CaCl 2 , and 0.9% (w/v) NaCl) at 25°C. Dissociation was monitored for 100 s at least in a constant flow of running buffer without enzyme.
The binding assay using the 125 I-labeled enzyme was performed according to Bolton and Hunter (19). SM liposomes in a buffer (20 mM Tris-HCl (pH 7.5), 0.9% (w/v) NaCl, 3 mM CaCl 2 ) were incubated with 10 ng/ml wild-type or mutant enzymes at 37°C for 30 min. After incubation, the SM liposomes were collected by centrifugation at 20,000 ϫ g for 20 min. The treated liposomes were washed in buffer and then treated with 10% (v/v) trichloroacetic acid, followed by centrifugation at 20,000 ϫ g for 10 min. The pellet was then dissolved in buffer supplemented with 2% (w/v) SDS, boiled for 3 min, and subjected to SDS-PAGE followed by autoradiography using a FLA2000 (Fuji Film).
Crystallization-Crystals of Bc-SMase were obtained by the hanging drop vapor diffusion method at 10°C. The crystallization drop for the calcium acetate-bound form was prepared by mixing equal volumes of protein solution (10 mg/ml Bc-SMase in 20 mM Tris-HCl (pH 7.0)) and reservoir solution (18% (w/v) polyethylene glycol 8000, 0.2 M Ca(O-COCH 3 ) 2 , and 0.1 M sodium cacodylate (pH 6.5)). For the calcium and cobalt chloride-bound crystals, calcium acetate was replaced with calcium chloride or cobalt chloride at a concentration of 0.2 M. In both cases the sodium cacodylate buffer was replaced with 0.1 M MES (pH 6.5). The crystals grew to maximum dimensions of 0.2 ϫ 0.2 ϫ 0.1 mm within a few weeks. The crystals of the magnesium-bound form were obtained in 16% (w/v) polyethylene glycol 8000, 0.2 M MgSO 4 , 0.1 mM CaCl 2 , and 0.1 M MES (pH 6.5).
The crystals of the calcium acetate-bound form were soaked for 12 h in the reservoir solution containing 5 mM samarium acetate to prepare the heavy metal derivative crystals.
The initial phases were calculated by the program SOLVE (22) using the data sets of the calcium acetate and the samarium acetate-derivative crystals at 2.8 Å resolution in the apparent space group C2. Phase extension to 2.5 Å with density modification, manual model building, and refinement calculations were performed using the programs RESOLVE (22), O (23), and CNS (24). The partially refined model was used as a search model for molecular replacement using the program MOLREP (20) in the true space group P1. The structures of four Bc-SMase molecules in the unit cell were refined using CNS (24) and O (23) iteratively. The crystal structures of the cobalt-and magnesium-bound forms were also determined by molecular replacement. The crystallographic statistics of the collected data and the refined models are summarized in Table 1. PyMOL (DeLano Scientific LLC) and QUANTA (Accelrys) were used to prepare the structure figures.
The positions of the bound metal ions of Bc-SMase in the Co 2ϩbound form and the Ca 2ϩ -bound form were determined by the anomalous difference Fourier method to distinguish the bound metal ions from the solvent. The Mg 2ϩ -binding sites were confirmed based on the stereochemical properties between the peaks in the 2͉F o ͉ Ϫ ͉F c ͉ residual electron density map because of the poor anomalous diffraction contribution of Mg 2ϩ at ϭ 1.0000 Å.
Modeling the Substrate-binding Region-Due to the structural and catalytic similarities between Bc-SMase and DNase I, the sphingomye-lin-binding model of the cobalt-bound Bc-SMase was constructed using QUANTA (Accelrys) and O (23) based on the crystal structure of DNase I complexed with DNA (Protein Data Bank code 1DNK) (12). The phosphate group of sphingomyelin was placed at the scissile phosphoester of DNA in the crystal structure of the DNase I ϩ DNA complex superimposed on Bc-SMase. The phosphate group of the model was in the vicinity of the phosphate group at the active site of the L. ivanovii SMase superimposed on Bc-SMase (7), with a distance of about 1.8 Å between the two phosphorus atoms. The position of the phosphate moiety from the surface of Bc-SMase was slightly further than that of L. ivanovii. This is likely to be because the phosphate moiety of the model binds to the active site via the essential divalent cations in contrast to the direct binding of the phosphate groups observed in the L. ivanovii crystal structure lacking the essential divalent cations. The conformation of the sphingomyelin molecule from the phosphocholine to the C-3 carbon mimics the conformation of the phosphodiester backbone of the DNA chain containing the scissile phosphoester bond. Both aliphatic chains lie in the branched trench of Bc-SMase.
Coordinates-The coordinates and the structure factors of the calcium-bound, the cobalt-bound, and the magnesium-bound forms have been deposited in the Protein Data Bank with accession codes 2DDR, 2DDS, and 2DDT, respectively.

RESULTS
Overall Structure-In this study, the crystal structures of Bc-SMase complexed to the functional metal ions Mg 2ϩ , Co 2ϩ , or Ca 2ϩ were determined at 1.8, 1.8, and 1.4 Å resolution, respectively. The overall structure of Bc-SMase consisted of a ␤-sandwich architecture made up of ␣/␤ motifs (Fig. 1A) and is basically the same independent of the bound metal ion. There are three distinct metal ion-binding sites in a long horizontal cleft across the Bc-SMase molecule (Fig. 1B). A double Data set for refinement Co 2ϩ site was identified in the central cleft as the active site located next to the C-terminal end of ␤-strand 2 (site A) and the loop from Asn-291 to Val-298 (site B). The third Co 2ϩ -binding site is located in the edge of the cleft at the N-terminal end of ␣-helix 2 (Fig. 1C). The number of the bound metal ions in the central site is dependent on the divalent metal ion. Two Co 2ϩ ions were bound to the active site compared with only one Ca 2ϩ ion. In contrast, in every case only one metal ion was bound to the edge site (Fig. 2).
Central Metal-binding Site as the Catalytic Site-We investigated the relationship between Bc-SMase hydrolytic activity and the essential divalent metal ions, Co 2ϩ , Mn 2ϩ , Mg 2ϩ , Ca 2ϩ , and Sr 2ϩ . The hydrolytic activity of Bc-SMase was measured using [N-methyl- 14 (Fig. 3). These results show Bc-SMase with bound Co 2ϩ , Mn 2ϩ , or Mg 2ϩ has high catalytic activity, but Bc-SMase with bound Ca 2ϩ or Sr 2ϩ exhibits much lower catalytic activity under the experimental conditions.  green spheres), whereas a further binding site is located at the edge of the protein and contains one Co 2ϩ . C, schematic of the secondary structure components of Bc-SMase. The sequence numbers of the start/end amino acid residues for each secondary structure are shown. The bound Co 2ϩ ions are represented by blue circles, and the water molecule being shared by two Co 2ϩ ions is shown by a red circle. The bound Co 2ϩ ions and the amino acid residues acting as ligands for the Co 2ϩ ions are connected by the dashed lines. The ␤-hairpin (green outline), three cobalt ions, and the solvent-exposed loop (blue outline) are linked closely by binding of Co 2ϩ ions. The bonds with a yellow outline participate in the linkage via the bound metal ions.
Comparison of the binding modes of Co 2ϩ , Mg 2ϩ , and Ca 2ϩ at the central cleft in the Bc-SMase crystal structures revealed water-bridged double Co 2ϩ ions separated by 3.5 Å (Fig. 4A). The bond distance between the bridging water and Co 2ϩ in site A of Glu-53 and site B of His-296 is 2.0 Å in each case. The bond angle between the two Co 2ϩ ions via the bridging water is ϳ120°. The two Co 2ϩ -centered octahedral bi-pyramids tilt at the bridge water by about 66°as the angle between vectors from the carboxylate oxygen of Glu-53 and the imidazole nitrogen of His-296 to each Co 2ϩ ion. The cobalt ions in site A and site B are bound to the carboxylate oxygen of Glu-53 and the imidazole nitrogen of His-296, respectively, as well as coordinated waters. The binding affinity of site A for Co 2ϩ was much higher than that of site B, because the peak heights of the Co 2ϩ ions at Glu-53 (site A) and at His-296 (site B) were 28 and 9 in the anomalous difference Fourier map at ϭ 1.5418 Å, respectively (Fig. 4B). The well defined bound waters form hydrogen bonds with the side chains of the amino acid residues Asn-16 and Asp-295 at site A and Asp-195 and Asn-197 at site B (Fig. 4A). The architecture of the central metal-binding site A of the Mg 2ϩ -bound form was identical to that of the Co 2ϩ -bound form with the same coordination to Glu-53 and the surrounding water molecules (Fig. 4B). The average coordinating bond length for the Co 2ϩ and Mg 2ϩ bound to site A was 2.2 Å in both cases. The architecture of the central metal-binding site B of the Mg 2ϩ -bound form was also identical to that of the Co 2ϩbound form; however, there was no defined Mg 2ϩ ion in the electron density map at the site B (Fig. 4B). As seen for the Co 2ϩ ions, the peak heights of the Mg 2ϩ ion at Glu-53 (site A) was much higher than that at His-296 (site B). In contrast to the double-linked octahedral metal coordination in the Co 2ϩ -and Mg 2ϩ -bound forms, there was only one Ca 2ϩ at the carboxylate oxygen of Glu-53 with five defined bound waters in a hepta-coordinate system in the Ca 2ϩ -bound form (Fig. 4C). The average coordination distance of the Ca 2ϩ ion (2.5 Å) was longer by 0.3 Å than those of the Co 2ϩ and Mg 2ϩ . The bound waters coordinated to the Ca 2ϩ form hydrogen bonds with Asn-16 and Asp-295.
The amino acid residues involved in divalent metal ion binding at the active site of Bc-SMase are Asn-16, Glu-53, Asp-195, Asn-197, Asp-295, and His-296 (Fig. 4A). These six amino acid residues are superimposable on the corresponding conserved amino acid residues of enzymes in the DNase I-like folding superfamily (DNA repair exonuclease III from Escherichia coli (30), DNA repair endonuclease HAP1 from human (31), inositol polyphosphate-5-phosphatase of synaptojanin from Schizosaccharomyces pombe (32), the subunit B of cytolethal distending toxin from Haemophilus ducreyi (33), endonuclease of human Line-1 Orf2P (34) and endonuclease from the telomere-specific long interspersed nuclear element (35)). Almost all these residues are conserved in other bacterial and mammalian nSMases. The exception is Asn-16, which is substituted by Val in some mammalian enzymes (Fig. 5A). This suggests that the mode of metal binding found in Bc-SMase is a common feature among not only nSMases but also members of the DNase I-like folding superfamily (Fig. 5B).
Exposed Hydrophobic Amino Acid Residues-A unique feature of the Bc-SMase structure is the ␤-hairpin (Trp-279 to Tyr-290) (Fig. 1, A and C ) not found in the structure of DNase I. The ␤-hairpin structure protrudes into the solvent with Trp-284 and Phe-285 located at the apex of the ␤-hairpin directly exposed to the bulk solvent. The ␤-hairpin structure of Bc-SMase is similar to that of the SMase from L. ivanovii (7).
In order to investigate the role of Trp-284 and Phe-285 in the ␤-hairpin of Bc-SMase, these residues were replaced with Ala by site-directed mutagenesis. The structure of the variant enzymes was almost identical to  the wild-type Bc-SMase as assessed by CD spectroscopic analysis (data not shown). However the liposome disruption activity of W284A and F285A mutants was a thousand-and a hundredfold lower than that of the wildtype enzyme, respectively (Fig. 6A). In addition, the hemolytic activity of both mutants was a thousandfold lower than that of the wild-type enzyme (Fig. 6B). We examined whether the mutated enzymes hydrolyze solubilized [N-methyl- 14 C]SM at a significantly lower concentration than the critical micelle concentration of SM. The catalytic activity of the variant enzymes was below the detection limit of the assay (greater than a 10-thousand-fold reduction compared with wild-type Bc-SMase).
Binding of W284A and F285A to SM Liposome-We determined whether Trp-284 and Phe-285 play an important role in binding to the membrane. SPR analysis showed that the Bc-SMase response rapidly reached a plateau and returned quickly to base line after the end of the injection of the SM liposome (Fig. 6C). The SPR response of W284A and F285A to SM liposomes showed one-tenth-and one-fifth-fold attenuation, respectively, compared with the wild type (Fig. 6, C, D and E). 125 I-Bc-SMase or the 125 I-mutant enzymes were incubated with SM liposomes in the presence of 3 mM Ca 2ϩ . Binding of W284A and F285A mutants to SM liposomes was a thousand-and a hundredfold lower than the wild type, respectively (Fig. 6F). It is therefore apparent that the affinities of W284A and F285A mutants to SM liposomes are much weaker than the wild-type enzyme.

DISCUSSION
We have determined the crystal structures of Bc-SMase in complex with the divalent metal ions Co 2ϩ , Mg 2ϩ , or Ca 2ϩ . The metal complex structures contained two Co 2ϩ , one Mg 2ϩ , or one Ca 2ϩ in the central cleft (Fig. 4) and in all cases one other metal ion at the side edge of the cleft (Fig. 2). The divalent metal ion bound at Glu-53 is concluded to be essential for hydrolytic activity. The central cleft is likely to be the active site, because Glu-53, Asp-195, and His-296 in the central cleft have been confirmed to be involved in the hydrolytic activity of Bc-SMase (5,36). Glu-53 was reported previously as the binding site of the hydrolytically essential Mg 2ϩ (5), whereas mutation of Asp-195 or His-296 abolishes hydrolytic activity (36).
The binding of both Co 2ϩ and Mg 2ϩ to the central cleft catalytic center showed a double hexa-coordinated architecture with double octahedral bi-pyramids but that of Ca 2ϩ was a hepta-coordinated architecture (Fig. 4). It appears that the binding mode differences between Co 2ϩ /Mg 2ϩ and Ca 2ϩ in the central cleft play a crucial role in the activity of Bc-SMase.
The activities of Bc-SMase bound to various metal ions were in the following order, Co 2ϩ Ն Mn 2ϩ Ն Mg 2ϩ Ͼ Ͼ Ca 2ϩ Ն Sr 2ϩ , compatible to the order of the average Lewis acid strengths of these divalent metal ions (37). The average Lewis acid strength of metal ions represents the empirical acid strength, because the average Lewis acid strength is the value of the oxidation number divided by the average coordinate number in many experimentally determined atomic structures (37). The average Lewis acid strengths of Co 2ϩ , Mn 2ϩ , Mg 2ϩ , Ca 2ϩ , and Sr 2ϩ are reported to be 0.351, 0.344, 0.334, 0.274, and 0.233 valence units, respectively (37) (Fig. 3), with Co 2ϩ and Sr 2ϩ the strongest and weakest Lewis acids of those tested.
The coordination structures of the bound Co 2ϩ , Mg 2ϩ , and Ca 2ϩ at the active site of Bc-SMase support the fact that the actual Lewis acid strengths of the ions in the active site of Bc-SMase are similar to the reported average observed Lewis acid strengths (37). In the catalytic site of Bc-SMase, Mg 2ϩ and Co 2ϩ exhibit a hexa-coordinate structure, whereas Ca 2ϩ exhibits a hepta-coordinate structure (Fig. 4) consistent with the average observed coordination numbers of Co 2ϩ (5.70), Mg 2ϩ (5.98), and Ca 2ϩ (7.31) (37).
The bound divalent metal ion as the Lewis acid at the active site of Bc-SMase would be directly involved in substrate binding. The differences in K m of Co 2ϩ , Mn 2ϩ , Mg 2ϩ , Ca 2ϩ , and Sr 2ϩ are small, in contrast to the V max values that differ widely (Fig. 3). The difference in the Lewis acid strength of Co 2ϩ , Mn 2ϩ , Mg 2ϩ , Ca 2ϩ , and Sr 2ϩ is relatively similar to the observed differences in K m values suggesting that the divalent metal ions acting as the Lewis acid interact with the negatively polarized atoms of the substrate directly, with the result that the K m value changes gradually according to the Lewis acid strength of the metal ion (Fig. 3).
The metal ions examined can be divided into two classes with respect to the efficiency of the Bc-SMase hydrolytic activity. Metal ions effective Dashed lines represent the hydrogen bonds between ligand water molecules and amino acid residues. The amino acid ligands of the bound Co 2ϩ ions are the carboxyl oxygen of Glu-53 (site A) and the imidazole nitrogen of His-296 (site B). B, superimposition of the central metal-binding sites of the Co 2ϩ ion-bound form and the Mg 2ϩ -bound form. The amino acids of the Co 2ϩ -bound form and the Mg 2ϩ -bound form are shown by cyan carbons and green carbons, respectively. The bound Co 2ϩ ion and associated ligand waters are represented by the blue and gray spheres, respectively, and the bound Mg 2ϩ ion and associated ligand waters are represented by the orange and red spheres, respectively. The violet and blue cage models are the ͉F o ͉ Ϫ ͉F c ͉ residual electron density map of the Mg 2ϩ -bound form contoured at 5.5 and the anomalous difference Fourier map of the Co 2ϩ -bound form contoured at 7, respectively. C, the central metalbinding site of the Ca 2ϩ -bound form is superimposed on that of Co 2ϩ -bound form. The amino acids with green and cyan carbons are those of the Ca 2ϩ -bound form and the Co 2ϩ -bound form, respectively. The bound Ca 2ϩ and Ca 2ϩ ligand waters and the bound Co 2ϩ and Co 2ϩ ligand waters are colored in violet, red, blue, and gray, respectively. The blue cage model is the anomalous difference Fourier map of the Ca 2ϩ -bound form contoured at 4.
in hydrolysis are Co 2ϩ , Mn 2ϩ , and Mg 2ϩ , whereas metal ions ineffective in hydrolysis are Ca 2ϩ and Sr 2ϩ (Fig. 3). There are also two clear modes of metal ion binding at the catalytic site. The Mg 2ϩ and Co 2ϩ show a water-bridged metal ion-binding mode, whereas the Ca 2ϩ ion exhibits a single metal ion-binding mode. The architecture of the Mg 2ϩ -bound form in site A is the same as that of the Co 2ϩ -bound form. The Co 2ϩ -FIGURE 5. The amino acid sequence alignment. A, the amino acid sequences of the magnesium-dependent neutral SMases (B. cereus (this work), S. aureus (P09978), L. ivanovii (Q9RLV9), human_SMase: neutral SMase from human (O60906), human_SMase2: neutral SMase2 from human (Q9NY59)) were aligned by the program T-Coffee (29). The completely conserved amino acid residues are indicated by the boxes. The amino acid residues participating in the central metal-binding site and in the edge metal-binding site are shown by red circles and blue triangles, respectively. The amino acid residues of the ␤-hairpin are shown by red characters, and those of the solvent-exposed loop (Asn-92 to Pro-98) before the edge metal-binding site are shown by blue characters. B, superimposition of the amino acid residues participating in the central metal-binding site. The amino acid residues of Bc-SMase are shown by the thick stick models. The amino acid residues from bovine DNase I (Protein Data Bank code 1DNK; blue) (12), DNA repair exonuclease III from E. coli (code 1AKO; red) (30), DNA repair endonuclease HAP1 from human (code 1BIX; orange) (31), inositol polyphosphate-5-phosphatase domain of synaptojanin from S. pombe (code 1I9Z; violet) (32), subunit B of cytolethal distending toxin from H. ducreyi (code 1SR4; pink) (33), endonuclease domain of human Line-1 Orf2P (code 1VYB; yellow) (34),and endonuclease domain from the telomere-specific long interspersed nuclear element (code 1WDU; cyan) (35) are shown as thin stick models.
bound form revealed that the Glu-53 and His-296 coordinate Co 2ϩ at site A and site B of the central cleft, respectively (Fig. 4A). In the Mg 2ϩbound form, Glu-53 coordinated Mg 2ϩ at site A as proposed (5), but Mg 2ϩ ion was not defined on His-296 at site B (Fig. 4B). However, the average distances between Co 2ϩ and the water molecules at site A and the effects of Co 2ϩ on SMase activity were similar to those of Mg 2ϩ (Fig.  3), indicating that the mode of Mg 2ϩ binding in the central cleft in the propagation of the hydrolytic reaction could be the same as that of Co 2ϩ . In fact, the architecture of the water molecules and amino acid residues (Asp-195, Asn-197, and His-296) in site B of the Co 2ϩ -bound form are coincident with those at site B of the Mg 2ϩ -bound form. In addition, the peak height of Co 2ϩ bound to Glu-53 in site A (28) was much higher than the Co 2ϩ bound to His-296 in site B (9) in the anomalous difference Fourier map (Fig. 4B), suggesting that site B has a lower affinity for metal ions than site A. In the 2͉F o ͉ Ϫ ͉F c ͉ electron density map of the Mg 2ϩ -bound form, the height of electron density peak of the Mg 2ϩ bound to the Glu-53 at site A was not very high (6.0), although this is the higher affinity site. We conclude His-296 binds Mg 2ϩ at site B, although the electron density of Mg 2ϩ was not defined in site B of Bc-SMase complex structure of Mg 2ϩ at the pH 6.5 of the crystallization condition. The crystallization pH value is comparable with the pK a of histidine, and a metal ion associated with His-296 is expected to bind with higher occupancy at higher pH values because of less competition with proton ions. Thus, it is highly plausible that Bc-SMase binds Mg 2ϩ at site B of the central cleft as observed in the Co 2ϩbound form (Fig. 4B). In contrast, bound Ca 2ϩ exhibits a hepta-coordination pattern at site A in the central cleft (Fig. 4C), a markedly different pattern from the double hexa-coordination, double octahedral bi-pyramids, seen in the Co 2ϩ -and Mg 2ϩ -bound forms. The average distance between Ca 2ϩ and the bound water molecules is significantly longer than that observed in the Co 2ϩ -and Mg 2ϩ -bound form as shown in Fig.  4C. The differences in coordinate distance are consistent with the ionic radii of hexa-coordinated Mg 2ϩ (0.72 Å) and Co 2ϩ (0.745 Å), compared with hepta-coordinated Ca 2ϩ (1.06 Å) (38). It is likely that the binding of a second Ca 2ϩ at site B in the central cleft of Bc-SMase is restricted by the geometry of the Ca 2ϩ -coordinated water molecules at site A. Taken together, these results indicate that the water-bridged double divalent ions in the central cleft is the essential architecture for SMase activity. The catalytic mechanism of Bc-SMase is proposed as follows based on the SM docking model (Fig. 7). Two metal ions bound to Glu-53 and His-296 at sites A and B of the central cleft of Bc-SMase, respectively, serve as a foothold for binding of SM in the proper orientation to the active site. The divalent cation linked to His-296 provides the general base water. The phosphate moiety of SM binds to the central cleft of Bc-SMase at the site of the water-bridged double metal ions in the model (Fig. 7A). The divalent metal ion at Glu-53 binds to the SM by direct interactions with the amido oxygen and the ester oxygen O-4 between C-1 and the phosphorus of SM (Fig. 7B). The water-bridged double divalent cations and the side chain of Asn-197 bind to the oxygen atoms of the phosphate moiety of SM. The bonds delocalize a negative charge on the phosphate group, resulting in a positively polarized phosphorus. The divalent metal ion at His-296 and the side chains of Asp-195 and Asn-197 lower the pK a value of the bound water molecule (Fig.  7B, Wat1), resulting in an activated water molecule. This activated water molecule attacks the positively polarized phosphorus of SM to form a pentavalent phosphorus. The growing negative charge on the oxygens of the pentavalent phosphorus, the reaction transition state, is delocalized by the double divalent metal ions, resulting in formation of ceramide and phosphocholine.
The model shows that the catalytic site containing the water-bridged metal ions accommodates various transition states with pentavalent phosphorus, as seen in other crystal structures of DNase I-like folding superfamily members (Fig. 7, C, D, and E), and stabilizes each of the negative charges on the transition states. The key residues in binding the water-bridged double Co 2ϩ to the active site of SMase are conserved in inositol polyphosphate-5-phosphatase of synaptojanin from S. pombe (32) and bovine deoxyribonuclease I (12), suggesting that this mechanism may be generally applicable to members of the DNase I-like folding superfamily.
Bc-SMase has an aromatic ␤-hairpin (Trp-279 to Tyr-290) conserved in hemolytic SMases from S. aureus (25) and from L. ivanovii (7) (Fig.  5A). This region is not found in other structurally characterized members of the DNase I-like folding superfamily. Replacement of Trp-284 and Phe-285 with alanine in the ␤-hairpin, W284A and F285A mutants, reduced binding to SM liposomes and disruption of SM liposomes and sheep erythrocytes (Fig. 6, A and B). SPR analysis confirmed the dramatic reduction in membrane binding of these mutants compared with the wild-type enzyme (Fig. 6, C, D, and E). These observations show that Trp-284 and Phe-285 are required for relocation of Bc-SMase to the cell membrane (Fig. 6F). In the crystal structure of the Mg 2ϩ -bound form, a molecule of MES was bound to the aromatic cluster, including Tyr-25, Trp-28, Tyr-242, Trp-284, Tyr-288, and Tyr-290 (Fig. 8A). The top of the ␤-hairpin was twisted in order to hold the bound MES (Fig. 8B). The positive head group of the MES molecule is a structural mimic of the phosphocholine moiety of the substrate, SM (Fig. 8C). Thus the phosphocholine moiety of the substrate may bind to the hydrophobic patch in the same manner as the MES molecule seen in the crystal structure. It can be postulated that the exposed aromatic amino acid residues, Trp-284 and Phe-285, initially interact with the positively charged phosphocholine moiety. Thus the exposed aromatic residues of Bc-SMase are involved in binding to the head group of SM in the cell membrane (Fig. 8D). These results are consistent with the rough sketch of Openshaw et al. (7) based on the crystal structure of SMase from L. ivanovii.
In the amino acid sequence alignments of the Bc-SMase and mammalian nSMases, the solvent-exposed loop (Fig. 5A, blue characters) from Asn-92 to Pro-98 is one of the fingerprint regions for bacterial Bc-SMase. This solvent-exposed loop is located next to Glu-99 and Asp-100, both of which form part of the edge metal-binding site of Bc-SMase. Furthermore, it is linked to another loop containing Phe-55 and Asn-57, which form the remaining part of the edge metal-binding site as seen in the Ca 2ϩ -bound crystal structure (Figs. 1C and 2). This suggests that the solvent-exposed loop may be regulated by Ca 2ϩ binding to the side edge site. Bc-SMase adsorbs to the erythrocyte in a Ca 2ϩdependent manner during hemolysis (2). In fact, Obama et al. (5) demonstrated that the Ca 2ϩ binds to Asp-100 of Bc-SMase. This implies that the solvent-exposed loop may be involved in the Ca 2ϩ -dependent membrane binding (Fig. 8D). The solvent-exposed loop is also conserved in bacterial SMases from S. aureus (25) and L. ivanovii (7), both of which have hemolytic activity, and asparagines corresponding to Asp-100 of Bc-SMase are also conserved (Fig. 5).
In conclusion, we have solved the structure of Bc-SMase complexed with metal ions (Co 2ϩ , Mg 2ϩ , and Ca 2ϩ ) revealing the presence of a double central ion-binding site. The suggested mechanism of action of the enzyme may be applicable to several members of the DNase I-like folding superfamily. A hydrophobic ␤-hairpin containing Trp-284 and Phe-285 and the solvent-exposed region from Asn-99 and Pro-98 are involved in binding of SMase to the cell membrane. FIGURE 8. Bound MES in the Mg 2؉ -bound form of Bc-SMase. A, the exposed aromatic amino acid residues form a binding site for an MES molecule. The bound MES is shown with the 2͉F o ͉ Ϫ ͉F c ͉ omit map contoured at 2.8. The phosphocholine (cyan carbon atoms) is superimposed on the bound MES. B, the ␤-hairpin twists in order to hold the bound MES. The bound MES is shown as a space-filling model. The gray and the cyan ribbons represent the ␤-hairpin of the free enzyme in the Co 2ϩ -bound Bc-SMase and the ␤-hairpin of the Mg 2ϩ enzyme with the bound MES molecule, respectively. C, the electrostatic surfaces of MES and phosphocholine. The blue and the red surfaces show the positive and negative electrostatic potential surfaces, respectively. D, the binding model of Bc-SMase to the SM membrane. Bc-SMase is shown as a gray ribbon model with the solventexposed loop in blue and with ␤-hairpin in green. The amino acid residues participating in the edge metal-binding site and the exposed aromatic amino acid residues of Trp-284 and Phe-285 are shown by the stick models with yellow carbons. The bound metal ion at the edge metal-binding site is colored in blue. The bound SMs are shown as stick models with cyan carbons.