Crystal structure of phosphorylcholine esterase domain of the virulence factor choline-binding protein e from streptococcus pneumoniae: new structural features among the metallo-beta-lactamase superfamily.

Streptococcus pneumoniae is the worldwide leading cause of deaths from invasive infections such as pneumoniae, sepsis, and meningitidis in children and the elderly. Nasopharyngeal colonization, which plays a key role in the development of pneumococcal disease, is highly dependent on a family of surface-exposed proteins, the choline-binding proteins (CBPs). Here we report the crystal structure of phosphorylcholine esterase (Pce), the catalytic domain of choline-binding protein E (CBPE), which has been shown to be crucial for host/pathogen interaction processes. The unexpected features of the Pce active site reveal that this enzyme is unique among the large family of hydrolases harboring the metallo-beta-lactamase fold. The orientation and calcium stabilization features of an elongated loop, which lies on top of the active site, suggest that the cleft may be rearranged. Furthermore, the structure of Pce complexed with phosphorylcholine, together with the characterization of the enzymatic role played by two iron ions located in the active site allow us to propose a reaction mechanism reminiscent of that of purple acid phosphatase. This mechanism is supported by site-directed mutagenesis experiments. Finally, the interactions of the choline binding domain and the Pce region of CBPE with chains of teichoic acids have been modeled. The ensemble of our biochemical and structural results provide an initial understanding of the function of CBPE.

Streptococcus pneumoniae is the most common cause of acute otitis media in children and is responsible for the majority of deaths from invasive infections such as pneumonia, sepsis, and meningitis, accounting worldwide for the death of one million children younger than 5 years old, annually. Treatment is impaired by the increased number of clinical strains resistant to single or multiple antibiotherapies. In addition, the current vac-cine strategies using immunogenic properties of capsular polysaccharides are efficient only against the homologous serotype, and this prevention approach is not very effective in the young and the elderly (1). Consequently, new vaccine strategies focus on the use of additional pneumococcal surface-associated proteins from which protection in all age groups, and against a large spectra of serotypes, could be attained because of high protein sequence conservation among pneumococcal serotypes (1).
Such new strategies require a complex understanding of the bacterium interaction with the host. Two pneumococcal factors involved in virulence processes are the capsular and cell wall polysaccharides. Teichoic acids and lipoteichoic acids are major components of the pneumococcal cell wall and contain phosphorylcholine moieties (PCho) 1 to which several functions have been attributed. One of the functions is binding to the plateletactivating factor receptor whose expression at the surface of epithelial and endothelial cells is up-regulated by local generation of inflammatory factors such as interleukin-1 and tumor necrosis factor ␣. Binding to the platelet-activating factor receptor induces internalization of pneumococci and promotes the transcellular migration through the respiratory epithelium and vascular endothelium (2,3).
Bacterial surface-exposed proteins are also involved in the complex network of interactions between S. pneumoniae and host tissue. The identification of virulence attributes and their characterization as new potential antibiotic and/or vaccine targets have been in progress since the release of complete genome sequences of virulent (TIGR4) and non virulent (R6) isolates of S. pneumoniae (4 -6). Despite an extensive list of putative virulence factors generated by genome-based analysis (sequence scanning or genetic screens), only a limited number of pneumococcal proteins have been identified as playing a role in virulence (7)(8)(9)(10)(11). However, biochemical and structural data for proteins involved in pneumococcal pathogenic processes have only recently started to become available (12)(13)(14)(15)(16)(17)(18).
The presence of PCho molecules at the bacterial surface is a feature shared by most respiratory tract pathogens: streptococcal and clostridial species, Neisseria meningitidis, Pseudomonas aeruginosa, and Haemophilus influenza (19,20). A specific characteristic of S. pneumoniae is the presence of proteins noncovalently attached to PCho; the choline-binding proteins (CBPs) by the means of choline binding domains (CBDs) (21,22). Up to 15 CBPs have been identified in pneumococcal strains (CBPA-G, LytA-C), each containing a CBD located at the N or C terminus, characterized by the repetition of a 20residue consensus motif GWX 6 WYYX 4 GXMX 2 (4,5,23). In addition, CBPs harbor a domain of variable length to which enzymatic properties have been associated in some instances: LytA, LytB, and LytC cleave peptidoglycan bonds, a function also proposed for CBPD, whereas CBPG has been proposed to be a serine protease (24 -28). A role in pneumococcal virulence has been attributed for almost all CBPs: PcpA is thought to be involved in protein-protein and protein-lipid interactions, whereas PspA inhibits complement activation, among other effects (Refs. 7 and 29 and references therein). CBPA is a cell surface adhesin and plays a major role in colonization of the nasopharynx by binding to immobilized sialic acid and lacto-N-neotetraose on cytokine-activated human cells (30). CBPA also interferes with the host immune response by interacting with the polymeric Ig receptor, the C3 complement component, and the complement-control protein factor H (Refs. 7, 12, and 31 and references therein). Finally, CBPD, CBPE, and CBPG have been shown to be involved in nasopharynx colonization. Mutant strains deleted in the cbpE and cbpG genes present, in addition, a significantly lower ability to adhere to human epithelial cells (23). Increasing knowledge regarding nasopharynx colonization and tissue invasion processes has highlighted the participation of surface-associated proteins in both stages of pneumococcal disease, and such molecules are becoming of major interest as potential new antibiotics targets and/or vaccine candidates, as already established for the CBPs (32).
In this work, we focused on CBPE, one of the most important members of the CBP family. The CBD of CBPE is composed of 10 repeats, and the N-terminal region is predicted to harbor a domain with a metallo-␤-lactamase fold (33). This domain, referred to phosphorylcholine esterase (Pce), catalyzes the hydrolysis of the choline-phosphoester bond, releasing PCho molecules from cell wall-associated teichoic and lipoteichoic acids (34 -36). A pneumococcal mutant deleted in cbpE is much less potent than the wild-type strain in colonizing the nasopharynx, harbors a decreased ability to adhere to nasopharyngeal cells, but causes an increase in virulence in a model of intraperitoneal inoculation (23,34). Here, we present the three-dimensional structure of Pce to a resolution of 2 Å. The features of the Pce active site show that this enzyme is unique among the large family of hydrolases harboring the metallo-␤-lactamase fold. Indeed, the orientation and calcium stabilization characteristics of an elongated loop, which lies on top of the active site, suggest that the cleft may be rearranged. Furthermore, the complexed structure of Pce with PCho, together with the characterization of catalytic iron ions in the active site, led us to propose a reaction mechanism that is supported by sitedirected mutagenesis. This work provides a structural and mechanistic insight into the function of CBPE.

Cloning of the pce Gene Encoding the Catalytic Domain of CBPE-
Genomic DNA from the R6 strain of S. pneumoniae was used as a template to amplify the pce gene by conventional PCR methodology. Subsequently, the resulting product was cloned into the pHIS8 vector (pHIS8/ pce). This construct encodes the Pce domain of CBPE (Glu 27 -Ser 334 ), deleted from the peptide signal and from the CBD, fused to a His 8 tag at the N terminus. DNA sequencing (Genome Express, Grenoble) confirmed that no mutations had been introduced during PCR.
Mutagenesis of the pce Gene-The pHIS8/pce construct was employed as a template for QuikChange TM site-directed mutagenesis (Stratagene) following the manufacturer's instructions. The D62A-, D114N-, and H253N-encoding constructs were also subsequently sequenced (Genome Express), and no mutations other than the ones expected were introduced during PCR.
Purification of Wild-type and Mutant Pce Proteins-An overnight culture of Escherichia coli BL21(DE3) strain transformed with pHIS8/ pce was used to inoculate (1:50) 1 liter of Luria Bertani medium supplemented with 30 g/ml of kanamycin. On achieving an optical density at 600 nm equal to 1 at 37°C, protein expression was induced with 1 mM isopropyl ␤-D-1-thiogalactopyranoside while incubating for 16 h at 15°C. Cells were centrifuged at 6,000 ϫ g for 15 min, the pellet was resuspended in 80 ml of a solution of 50 mM Tris-HCl, pH 8.0, 200 mM NaCl, 20 mM imidazole (buffer A), containing 2 tablets of Complete protease inhibitor mixture (Roche Applied Science). Cells were sonicated, the cell lysate was centrifuged at 31,000 ϫ g for 20 min, and the resulting supernatant was loaded onto a 5-ml chelating Sepharose column (Amersham Biosciences), previously loaded with 100 mM NiSO 4 and equilibrated in buffer A. Extensive wash steps (10 column volumes) with buffer A containing successively 50 and 100 mM imidazole were performed preceding the elution of the protein with 300 mM imidazole (buffer A); subsequently, the sample was extensively dialyzed in 50 mM Tris-HCl, pH 8.0, 50 mM NaCl. Pce was incubated with 4 units of thrombin per mg of fusion protein for 1 h at room temperature to cleave the N-terminal His tag. The cleaved product was loaded onto an anionexchange chromatography column (Resource Q; Amersham Biosciences) at a flow rate of 2 ml/min, and a NaCl linear gradient (0 -300 mM) was applied to elute Pce. Finally, gel filtration was performed on Superdex 200 in 25 mM Tris-HCl, pH 8.0, 100 mM NaCl. About 20 mg of pure protein were obtained from a 1-liter E. coli culture.
Expression conditions of Pce mutants, D62A, D114N, and H253N were identical to the Pce wild-type protein. The mutants were purified on Ni 2ϩ affinity columns and subsequently dialyzed against 50 mM Tris-HCl, pH 8.0, 150 mM NaCl. The homogeneity and masses for all purified proteins were verified by SDS-PAGE electrophoresis and ESImass spectrometry.
Native Mass Spectrometry Characterization of the Apo and Reconstituted Forms of Pce-Noncovalent mass spectrometry measurements were performed by using Q-TOF Micro mass spectrometer (Micromass, Manchester, UK) equipped with an electrospray ion source. It operated with a needle voltage of 2.7 kV, sample cone and extraction cone voltages, respectively, of 150 V and 10 V; backing Pirani pressure was set at 6.21 mbar. The mass spectra were recorded in the 2000 -4000 mass-to-charge (m/z) range. Samples were continuously infused at a flow rate of 7 l/min, data were acquired in the positive mode, and calibration was performed using a solution of 0.5 mg/ml CsI in water/ isopropyl alcohol (1:1, v/v). Data were processed with MassLynx 4.0 (Micromass). The apo form of Pce deleted from the His tag was prepared as follows: 20 mM EDTA was added to the protein (28 M) and incubated for 2 h at room temperature; the protein solution was subsequently dialyzed against 20 mM ammonium acetate, pH 6.0. Apoprotein was then incubated with 1 mM Fe 2ϩ , 5 mM dithiothreitol (DTT), 1 mM CaCl 2 , and 10 mM PCho for 2 h at room temperature. Protein solution was again extensively dialyzed against 20 mM ammonium acetate, pH 6.0 to eliminate unbound molecules, leading to the reconstituted Pce protein.
The three Pce samples in 20 mM ammonium acetate, pH 6.0; native (before treatment), apo, and reconstituted forms; were analyzed by mass spectrometry under non-denaturing conditions.
pNP-PC Assays-The PCho esterase activity of Pce wild type and mutants were measured with p-nitrophenyl-phosphorylcholine (pNP-PC, Sigma) as the substrate at 37°C in 50 mM potassium phosphate buffer, pH 8.0, in a total volume of 200 l. The activity was measured by following the increase in absorbance at 405 nm with the production of p-nitrophenyl (pNP), whose quantification as a reaction product of Pce activity had been previously calibrated with a standard curve. Apo forms of wild-type and mutant proteins were prepared by incubation for 2 h at room temperature with 20 mM EDTA and subsequent extensive dialysis against 50 mM Tris, pH 8.0, 150 mM NaCl. The apo forms were used in the enzymatic assays with and without addition of iron, DTT, and calcium. Metal ions and DTT were added in concentrations from 0.01 to 1 mM. Reduced (Fe 2ϩ ) and oxidized (Fe 3ϩ ) forms of iron were generated from Fe(NH 4 ) 2 (SO 4 ) 2 salt (uncolored solution in presence of 10 mM DTT) and FeCl 3 (deep orange-colored solution), respectively. The concentration of pNP-PC varied from 0.16 to 32.8 mM, and each protein was used at concentrations around 70 nM. The kinetic parameters k cat and K m were determined by fitting the data to the Michaelis-Menten equation.
A crystal obtained using MPD as precipitant was soaked for 1 day in 0.1 mM PCho, mounted in a loop, and flash-cooled in liquid nitrogen. X-ray data were obtained in-house using monochromatic CuK␣ x-rays supplied by a Nonius FR591 rotating-anode generator, coupled to a Mar Research Image Plate detector. Another crystal, obtained using PEG550 MME as precipitant, was first transferred to a cryoprotectant solution (reservoir solution containing 20% glycerol), mounted on a loop, and flash-cooled in liquid nitrogen. X-ray data at the 1 (remote) and 2 (peak) edges of iron were collected at the ID29 ESRF beamline. Data sets were processed using MOSFLM and SCALA of the CCP4 program suite (37).
Structure Determination and Refinement-The structure of Pce was solved using the single wavelength anomalous dispersion method (38). Inspection of the in-house collected data set showed an anomalous signal, which could be interpreted as the anomalous signal of iron at 1.5418 Å (fЉ ϭ 3.2 e Ϫ ). The positions of four iron atoms in the asymmetric unit were determined using SOLVE (39). Phase refinement resulted in an initial figure of merit of 0.24. Density modification with noncrystallographic symmetry averaging by RESOLVE (40) increased the figure of merit to 0.65 and gave a map of good quality. Automatic building using RESOLVE resulted in a model that included 76% of the sequence. Multiple rounds of model building with O (41) and refinement with REFMAC (42) generated the final model. Ca 2ϩ assignments were supported by environment coordination, temperature factors, and omit maps for different cations. Conformational torsion angle restraints and charge assignments for the PCho molecule were obtained using CCP4i Libcheck (37).
Phases for the Pce structure were also obtained de novo using synchrotron radiation. The crystal obtained using PEG550 MME as precipitant revealed a strong fluorescence signal at the K-iron edge, which was attributed to the presence of a binuclear iron center. No fluorescence signal was seen at the adsorption edge of zinc or manganese using the same crystal. Using the SAD data from the peak wavelength (1.7396 Å; fЉ ϭ 5.6 e Ϫ ), four iron ions were located in the asymmetric unit using SOLVE. Phase refinement resulted in a figure of merit of 0.22 and of 0.62 after density modification with non-crystallographic symmetry averaging by RESOLVE. Using data collected on the remote wavelength, phases were then extended from 2.40 to 1.95 Å, and automatic building by RESOLVE resulted in a model that included 71% of the sequence. Multiple rounds of model building with O and refinement with REFMAC generated the final model. The quality of the map was very similar to that obtained using the in-house data set. Interestingly, the refined structure obtained from this crystal, which was not soaked in PCho solution, showed a PCho molecule bound to the active site. However, the extra electron density in the active site was weaker than that seen using the in-house data set, indicating a lower ligand occupancy. We therefore only report the refined structure using the in-house data collection, i.e. the Pce-PCho complex with full occupancy.
The Pce catalytic region of CBPE model coordinates and structure factors were deposited in the Protein Data Bank (1WRA). The structural model of the CBD of CBPE was obtained using the Geno3D server with the Cpl-1 structure as the starting model (PDB code 1OBA).

RESULTS
Overall Structure-The CBPE topology is defined upon amino acid sequence and structural data obtained in this work. The full-length protein contains a signal peptide (Met 1 -Gln 26 ), the Pce catalytic region (Glu 27 -Arg 298 ) followed by a linker region (Gly 299 -Ser 334 ), and the C-terminal CBD (Gly 338 -Val 541 ) (Fig. 1A). The crystals obtained with the recombinant Pce protein (Glu 27 -Ser 334 ) belonged to space group P4 1 2 1 2, and the structure was refined to a resolution of 2.0 Å. Two very similar polypeptide chains Mol A (Gly 30 -Ser 334 ) and Mol B (Glu 27 -Ser 334 ) were found in the asymmetric unit (root mean square deviation of 0.13 Å for pairs of C␣ atoms), 2 PCho molecules, a total of 359 water, and 7 MPD molecules were also detected ( Table I).
The / angles of the Asn 51 , Asp 58 , Trp 148 , Asp 176 , Glu 200 , and His 254 residues are located in the disallowed areas of the Ramachandran plot. Asn 51 ( ϭ 77°; ϭ Ϫ71°) connects the ␤2and ␤3-strands. Asp 58 ( ϭ 61°; ϭ 173°) interrupts the ␤3-strand and turns the protein backbone by about 90°, creating a hole filled by 3 water molecules inside the core of Pce. Trp 148 ( ϭ 33°; ϭ Ϫ114°), lying in the active site, is a key residue in the interaction with PCho, while Asp 176 ( ϭ 1°; ϭ Ϫ48°) precedes the ␤7-strand that starts the second half of the catalytic region. Glu 200 ( ϭ 23°; ϭ 111°) belongs to an external fragment (Glu 192 -Lys 203 ) of the loop between ␤8 and ␤9. In both Mol A and Mol B, this fragment is characterized by a relatively high temperature factor and is located at the interface between the two molecules. Finally, His 254 ( ϭ 72°; ϭ Ϫ62°) is one of the metal coordinating residues.
The Shape of the Pce Active Site Possibly Modulated by an Elongated Loop Stabilized by Ca 2ϩ Ions-A striking feature of Pce, unique among reported structures of the metallo-␤-lactamase superfamily, is the presence of an elongated loop, which lies on top of the active site ( Fig. 2A). This loop, which joins the ␤3-strand to the ␣1-helix, starts at Gly 60 and ends at Thr 81 . The terminal fragment, Gly 78 -Thr 81 , delimits the substrate binding pocket. Furthermore, Pro 74 forces residues Arg 76 -Ile 79 to approach Trp 148 , which, in a strained conformation, is at the end of helix G3 ( Fig. 2A). A network of hydrogen bonds between Tyr 63 -Asp 89 , Asp 64 -Tyr 152 , and Glu 80 -His 85 stabilizes the elongated loop ( Fig. 2A). Tyr 63 makes hydrophobic contacts with Tyr 83 and Va l86 , whereas the Phe 65 and Pro 66 side chains form hydrophobic interactions with the side chains of Thr 81 and Tyr 152 , respectively. The interface between the elongated loop and the rest of the protein is mainly hydrophilic and contains 8 buried water molecules. Taken together, the orientation of this elongated loop might modulate the shape of the PCho binding site and/or accessibility to the teichoic acid substrate.
Two Ca 2ϩ ions and several electrostatic and hydrophobic interactions further stabilize the orientation of the elongated loop ( Fig. 2A). The side chains of Glu 61 and Asp 62 , together with the backbone oxygens of Asp 62 and Asp 64 and two water molecules, form the coordination sphere of the first Ca 2ϩ ion ( Fig.  2A). The second Ca 2ϩ ion is located 6.3 Å away from the first: its coordination sphere is formed by the side chains of Asp 62 , Asp 120 , Glu 121 , two water molecules, and the backbone oxygen of Gly 117 ( Fig. 2A). Because of the importance of Asp 62 , which bridges the two Ca 2ϩ ions, the mutant D62A was constructed, and the kinetic parameters were measured. As shown in Table  II, no significant difference was observed between Pce and the D62A mutant, suggesting that the removal of one Ca 2ϩ coordination site out of 6 does not induce major modifications. Despite the fact that the D62A mutant displays the same level of activity toward the synthetic substrate pNP-PC as the wildtype, Ca 2ϩ ions are suggested to stabilize the elongated loop, whose position and orientation may play a role in the selectivity for the natural substrate of CBPE.
Iron Ions Are Essential for Catalysis-The electron density map on Pce crystals indicated the presence of two metals in the active site. From in-house data collection, the anomalous diffraction signal observed with CuK␣ x-rays ruled out the presence of zinc but cannot discriminate between the presence of manganese or iron, whereas strong fluorescence spectra recorded at the European Synchrotron Radiation Facilities (Grenoble) revealed unambiguously the presence of only iron in crystals. It was concluded that the Pce domain had a binuclear center containing iron, although the expected metal was zinc, present in the majority of the members of the metallo-␤-lactamase superfamily.
It appeared then important to verify that the iron ions were present in the purified Pce protein in solution. The apo form of Pce was generated, and the protein was reconstituted by incubation with iron and calcium ions and PCho molecules. The masses of the various forms (native, apo, and reconstituted) measured by mass spectrometry in non-denaturing conditions were 35,848, 35,482, and 35,848 Da, respectively. These results indicated that the native form of Pce, in solution, after purification, is composed by one molecule of the apoPce (35,482 Da) complexed to one PCho molecule (183 Da), to two iron ions (110 Da), and to two calcium ions (80 Da), in accordance with the crystallographic data.
The catalytic relevance of iron and their ionic states have been investigated by enzymatic assays in which the apo form of Pce was tested for pNP-PC hydrolysis in the presence or absence of iron, calcium, and PCho (Fig. 3). Only addition of 1 mM Fe 2ϩ increased the efficiency of catalysis (4-fold) compared with the oxidized iron state (Fe 3ϩ ) or other metals, such as zinc and calcium (Fig. 3). Hydrolysis was doubled when reducing agents such as DTT and Tris(2-carboxyethyl)phosphine were added to the reaction mix (data not shown). It is thus probable that the binuclear center present in the Pce active site is involved in the catalytic reaction and that the reduced state is more favorable for enzymatic activity.
The PCho Binding Site-The structure of Pce was solved in complex with the ligand PCho. The major interaction between Pce and PCho occurs between the phosphate group of PCho and the binuclear center (Fig. 2B). Two of the oxygens of the PCho phosphate group interact directly with iron ions, leaving the third oxygen pointing away from the binding site (Fig. 2B). This phosphate group orientation would allow Pce to bind a substrate bearing an ester moiety linked to PCho. In the active site, residues Arg 146 , Asp 206 , and His 253 further stabilize PCho binding (Fig. 2B).
A unique feature of the active site is the small cavity formed by the side chains of two iron ligands His 112 , Ser 113 , Asp 114 , and Trp 148 and the backbone oxygen of Gly 78 that accommodates the positively charged trimethyl-ammonium group of PCho (Fig. 2B). A cation-interaction is observed between the choline and Trp 148 , a feature found in other choline-binding proteins (44). In addition, Trp 205 and Trp 284 face the active site and might interact with other choline molecules. All the interactions and distances in the active site are represented in Fig. 4.
Structural Features of the Active Site: Pce Is Related to Metallo-phosphatase Enzymes-The metal binding site is situated at a topological equivalent position to that observed in other members of the metallo-␤-lactamase superfamily (Fig. 5) (45). The Pce coordination spheres of the metal ions are close to octahedral and involve the His 110 -His 112 -Asn 208 triad at site 1 and the Asp 114 -His 115 -His 254 triad at site 2. The phosphate group of the PCho molecule bridges the two metals symmetrically; Asp 228 O ␦1 , a water molecule or a hydroxo group bridge complete the coordination sphere of the two metals (Fig. 5).
Pce and ROO from Desulfovbrio gigas, a strict anaerobe, are the only enzymes in the metallo-␤-lactamase fold superfamily to contain a di-iron center and to present only two histidine residues in metal site 1; zinc-containing members GOX, RNaseZ, and FEZ-1 display three histidines (48) (Fig. 5). The latter three enzymes harbor metal cofactors with a tetrahedral or trigonal bipyramidal geometry; human glyoxalase II contains two zinc ions with an octahedral coordination whereas Arabidopsis glyoxalase II contains one zinc and one iron ion (47,50). However, the Pce metal center shows significant differences compared with the ROO structure: in site 2, His 115 replaces a coordinating water molecule whereas the O ␦1 atom of Asp 228 bridges both iron ions (Fig. 5). Finally, despite the same fold and metal cofactors, the structural differences in the Pce and ROO active sites allow these enzymes to have completely different functions: PCho ester hydrolysis or oxygen reduction to water, respectively.
The most striking feature of the Pce structure is the similarity of its binuclear center phosphate group to that of metallophosphatases despite their distinct three-dimensional folds; the active form of purple acid phosphatase (Pap), for example, contains Fe 3ϩ in site 1 and Fe 2ϩ in site 2 (Fig. 5) (51). Coordinate superposition was performed by employing two irons, their two bridging oxygens (hydroxo and O ␦ aspartic acid) and one ligand phosphorus atom. The iron-coordinating triads His-Asp-Tyr and His-His-Asn in Pap correspond in Pce to His 112 -His 110 -Asn 208 and Asp 114 -His 115 -His 254 , respectively (Fig. 5). As a consequence, Pap Tyr 55 , which is responsible for the characteristic purple color of the protein because of the tyrosinate 3 Fe 3ϩ charge transfer transition at 560 nm is replaced by Asn 208 in the Pce site 1 (Fig. 5). Pap His 92 , another residue in the vicinity of the metal center, corresponds to His 253 and occupies an equivalent position, interacting with one phosphate oxygen in both molecules (Fig. 5).
The Catalytic Mechanism of Pce-Over the years, many catalytic mechanisms have been proposed for metallo-phosphatases (52). The most probable mechanism proceeds via nucleophilic attack of the phosphoryl group by a metal-ligated water molecule activated to a hydroxyl; the nature of the metalactivated nucleophile is still under debate. It has been established that Pap hydrolysis of phosphate esters occurs with inversion of stereochemistry of the phosphorus (53). For this enzyme the proposed nucleophile is a hydroxide ion bound to the Fe 3ϩ , the alternative mechanism, using the bridging hydroxo ion as nucleophile, is less probable (54). Based upon the structural similarities between the Pce and Pap active sites, a Pap-like S N 2-type catalytic mechanism can be proposed for Pce (Fig. 6). The PCho ester initially binds to iron site 2 through one of the non-esterified oxygen atoms of the phosphate (step 1), which undergoes nucleophilic attack by a hydroxyl ion bound to iron site 1, inducing an inversion of configuration on the phosphorus atom. These events first generate a pentacoordinate intermediate (step 2) and subsequently the complex (step 3, structure observed) in which the reaction product is bound in the active site (Fig. 6). The enzyme is most probably regenerated by the binding of a water molecule to iron site 1, promoting the release of the product by ligand substitution. Evidence in favor of this hypothesis are the following. (i) The left side of the active site is more accessible to the substrate, supporting an initial interaction of PCho with iron site 2. (ii) A water molecule (Wat11 in MolA and Wat216 in MolB), which makes two hydrogen bonds with Arg 146 and Asp 206 , is in perfect position to interact with one phosphate group oxygen in the pentacoordinate intermediate, and (iii) a conserved role for His 253 in the catalytic mechanism, which may give a proton to the leaving group.
To experimentally verify the proposed reaction mechanism, Pce mutants were constructed and enzymatic assays were performed using pNP-PC as a substrate (Table II). When His 253 was mutated to Asn, the mutant showed much lower enzymatic activity (15% of the wild-type value), which was too low for  Fig. 1B, the enzymatic cavity is viewed 90°C from the right side. Important amino acids and PCho are represented in ball-and-stick, iron and calcium ions as purple and gray spheres, respectively. kinetic parameters to be measured. This result indicates that His 253 is not only involved in the binding of the phosphate group, as evidenced by the structure of the complex, but may play a role in catalysis (Fig. 6). D114N also showed a large decrease in enzymatic efficiency, displaying 21% of the wildtype value (Table II). Asp 114 has two roles in the Pce structure: (i) iron liganding, and (ii) stabilization and activation of the bridging water molecule. An equivalent mutation in Bacteroides fragilis Zn-␤-lactamase generated a mutant enzyme harboring a mixed population with either one or two zinc ions per enzyme molecule, as well as reduced catalytic activity (49-fold decrease using penicillin G) (55).
In this work, the oxidation state of each iron ion was not determined; however, by analogy with Pap and because of the fact that Pce activity increases when a reducing agent is added, it is reasonable to assume that one of the iron ions must be in the reduced state.
Modeling of the CBD of CBPE and Interaction of CBPE with Teichoic Acid Chains-At the C terminus of the Pce structure, residues Gly 299 -Ser 334 form the linker region between the catalytic region and the CBD (Fig. 1A). The well defined structure of this linker region identifies the N terminus of the CBD and suggests the relative positions of Pce and the CBD. All CBDs are formed by the repeat of the consensus motif GWX 6 WYYX 4 GXMX 2 . Two of these consecutive repeats form one choline binding site, whose structural arrangement has been solved in Lyt-A and Cpl-1 proteins (21,22). Sequence identity between the CBDs of CBPE, Cpl-1, and Lyt-A is high (40 -50%). Thus, in comparison with Cpl-1 and Lyt-A structures, a model of the CBD from CBPE has been built (Fig. 7). The only known substrate for Pce are the PCho molecules displayed by the cell wall teichoic and lipoteichoic acids. Modeling of the interaction between CBPE and teichoic acid chains has been performed based on (i) the presence of seven bound MPD molecules on the Pce structure, which may mimic the PCho molecules (56) and (ii) on the four choline-binding pockets of the CBD described above (Fig. 7). It appears that one chain of teichoic acid (chain A) can fit into the enzymatic cavity of Pce; it can then be proposed that the most accessible PCho residue for hydrolysis is the one linked to the N-acetyl-D-galactosaminyl moiety, itself linked to the ribitol phosphate sugar (Fig. 7). This observation suggests that Pce cannot hydrolysis all PCho molecules on the teichoic acid polymers because of structural constraint. The binding of up to two teichoic acid chains, B and C, can be modeled along the CBD, in a way to schematically visualize the attachment of CBPE to the cell wall (Fig. 7). In conclusion, CBPE is a very interesting multifunctional enzyme: the same substrate (PCho on teichoic acids) is recognized by both Pce and CBD regions although the two binding site architectures are not related (even if aromatic residues are always required). Furthermore the functions of these domains are totally different: Pce hydrolyzes the PCho substrate while the CBD remains bound to it. DISCUSSION CBPs are pneumococcal cell wall-associated macromolecules that display activities ranging from adhesion to proteolysis of host proteins. Here, we report the structural and biochemical characterization of Pce, the catalytic domain of CBPE from S. pneumoniae.
The overall structural fold confirms that the Pce region of CBPE belongs to the metallo-␤-lactamase superfamily but major structural and functional features render Pce unique among this group of proteins. First of all, Pce displays a di-iron phosphatase-like reaction mechanism. Iron ions in the Pce active site structure were identified by fluorescence scans on crystals,  and their signal was employed to solve the structure. Furthermore, the presence of iron ions in the structure was verified to ensure that the binding of iron atoms was not a crystal artifact, but de facto physiological. Indeed, solution reconstitution of the Pce native state from the apo form could be achieved by incu-bation with iron, calcium, and PCho molecules. However, because different cations can bind in the metal sites of metallo-␤lactamase like-proteins, we performed enzymatic assays. Under our experimental conditions, all the results reveal that iron ions are essential for the PCho hydrolysis activity of Pce (57). The presence of iron ions and PCho molecules in the active site, as well as the nature and geometry of the coordination ligands, suggest that the catalytic properties of Pce are more closely related to that of metallo-phosphatase enzymes than to those of metallo-␤-lactamase hydrolases. Based on this observation, we proposed a reaction mechanism that was supported by site-directed mutagenesis. The structure of the active site determined in this work corresponds to the oxidized state of the enzyme complexed to the reaction product. Because enzymatic regeneration must proceed by entry of a water molecule in metal site 1, the structure observed may correspond to an inactive enzyme. The reduced state of iron in site 1 must be essential in the catalysis process; this hypothesis is in accordance with the observed increased activity of the native form of Pce in the presence of reducing agents. However, the reduced/oxidized state of each iron ion remains to be elucidated, using, for example, the electron paramagnetic resonance method. Nevertheless, these observations open questions about the optimal physiological conditions for Pce function. For example, is Pce most efficient in an anaerobic environment (inside the host tissue) than in an aerobic environment (on nasopharyngal epithelial cells)?
Kinetic parameters were determined for Pce wild-type and mutant enzymes, using the synthetic substrate pNP-PC. Even for the wild-type protein, the measured values of k cat /K m are in the 10 Ϫ3 s Ϫ1 ⅐M Ϫ1 range, indicating that the protein harbors very low catalytic efficiency (36). This point raises the question of the relevance of the pNP-PC molecule as a substrate for Pce, although the use of this synthetic substrate has been of central importance for the identification of CBPE in pneumococcal cell extracts (34,36).
The Pce active site harbors a structural characteristic not shared by any member of the metallo-␤-lactamase superfamily, the presence of an elongated loop lining the top part of the active site. The orientation, interactions, and calcium stabilization of this loop suggest a role in the accessibility of substrate into the catalytic cavity of Pce. Asp 62 bridges both calcium ions, and although its mutation to Ala had no effect on enzymatic efficiency with pNP-PC as a substrate (neither with exogenous calcium added to the apoPce), this does not exclude the possibility that calcium may play a role in catalysis when native FIG. 7. Proposed model of the recognition of pneumococcal teichoic acids by CBPE. The CBD domain has been modeled. Pce has the same orientation as in Fig. 1B; the surface of the protein is colored in gray. Three chains, A, B, and C of teichoic acids have been modeled, the start and the end of each chain are indicated by s and e, respectively. The polysaccharide chains are colored in red (A), in blue (B), in green (C), and the PCho molecules are all represented in dark blue. The bridge of Bs to an N-acetyl-muramic acid residue of the peptidoglycan is schematically represented, as well as the orientation of teichoic acids and lipoteichoic acids in the cell wall. A repeat unit of teichoic acid is (i) schematically represented in brackets and (ii) docked into the Pce active site (enlarged view of the model). The same color code has been used for clarity of the figure. The phosphate group is colored in pink, the ribitol in wheat, the two N-acetyl-D-galactosaminyl residues in red and orange, the positively charged 2-acetamido-4-amino-2,4,6-trideoxy-D-galactose in green, and the D-glucose in yellow. One PCho molecule is colored in dark blue and the other one in cyan: this latter PCho residue is in a favored interaction with the binuclear center for subsequent hydrolysis. The iron atoms are shown as spheres. The figure was created with Pymol (www.pymol.org/). substrate binds to the active site. In addition, the concentration of free calcium under physiological conditions may also influence the orientation/stabilization of the elongated loop and the active site accessibility.
CBPE has been shown to remove only 20% of the PCho residues from the cell wall teichoic acids (35). The limited activity of the enzyme was suggested as being attributed to the poor spatial accessibility of the PCho molecules; however, this low enzymatic efficiency reflects the catalytic results presented in this work using the synthetic substrate, indicating that in vitro, under aerobic conditions, Pce has low activity. Nevertheless, the virulence effect of CBPE may be the consequence of its enzymatic property, which influences the quantity of PCho (known to bind platelet-activating factor receptor on host cells) and may also regulate the presence of CBPs at the bacterial surface. However, results obtained by Gosink et al. (23) with a pneumococcal strain deleted in cbpE, i.e. decreased nasopharynx colonization and cell adherence, suggest the hypothesis of a direct effect of the CBPE protein in virulence processes. In addition, CBPE may function either bound to the bacterial surface or released in the extracellular medium. By catalyzing the cleavage of PCho molecules, CBPE may proceed to its own release by competition with teichoic acid-bound PCho residues.