Crystal Structure of the 47-kDa Lipoprotein of Treponema pallidum Reveals a Novel Penicillin-binding Protein*

Syphilis is a complex sexually transmitted disease caused by the spirochetal bacterium Treponema pallidum. T. pallidum has remained exquisitely sensitive to penicillin, but the mode of action and lethal targets for β-lactams are still unknown. We previously identified the T. pallidum 47-kDa lipoprotein (Tp47) as a penicillin-binding protein (PBP). Tp47 contains three hypothetical consensus motifs (SVTK, TEN, and KTG) that typically form the active center of other PBPs. Yet, in this study, mutations of key amino acids within these motifs failed to abolish the penicillin binding activity of Tp47. The crystal structure of Tp47 at a resolution of 1.95 Å revealed a fold different from any other known PBP; Tp47 is predominantly β-sheet, in contrast to the α/β-fold common to other PBPs. It comprises four distinct domains: two complex β-sheet-containing N-terminal domains and two C-terminal domains that adopt immunoglobulin-like folds. The three hypothetical PBP signature motifs do not come together to form a typical PBP active site. Furthermore, Tp47 is unusual in that it displays β-lactamase activity (k cat for penicillin = 271 ± 6 s−1), a feature that hindered attempts to identify the active site in Tp47 by co-crystallization and mass spectrometric techniques. Taken together, Tp47 does not fit the classical structural and mechanistic paradigms for PBPs, and thus Tp47 appears to represent a new class of PBP.

Syphilis is a chronic, complex sexually transmitted disease of humans caused by the spirochetal bacterium Treponema pallidum. Humans are the only known reservoir for T. pallidum, and although syphilis is one of the oldest recognized sexually transmitted diseases, a major impediment to research on T. pallidum continues to be the inability to cultivate the organism in vitro. Consequently, despite decades of intensive efforts, many features of T. pallidum ultrastructure, physiology, and membrane biology remain obscure (1).
T. pallidum is exquisitely sensitive to penicillin, which continues to be the drug of choice for syphilotherapy. Penicillin and other ␤-lactams are bactericidal via their ability to inhibit cytoplasmic membrane-bound enzymes (penicillin-binding proteins (PBPs)) 1 involved in peptidoglycan biosynthesis (2). Generally, bacteria contain several PBPs that are classified within two categories (high molecular weight or low molecular weight) (3,4). In Escherichia coli, the high molecular weight PBPs tend to be bifunctional (transglycosylase/transpeptidase activities) and are the lethal targets of ␤-lactams (5). The low molecular weight PBPs can be either monofunctional DD-carboxypeptidases, bifunctional DD-carboxypeptidases/DD-endopeptidases, or monofunctional DD-endopeptidase (6). In T. pallidum, the lethal targets for ␤-lactams are not known. However, two previous studies in which T. pallidum was incubated in vitro with radiolabeled ␤-lactams implicated polypeptides of 94, 80, 63, 58, 47, and 38 kDa (7) or 180, 89, 80, 68, 61, 41, and 38 kDa (8) as PBPs. As a follow-up to an earlier study by us (7), we have shown that the major 47-kDa membrane lipoprotein of T. pallidum (Tp47) is a PBP. More recent genome information (9) has suggested that T. pallidum encodes at least three theoretical PBPs of molecular masses of 71 (TP0500, PBP-1; TP0760, PBP-3) and 98 (TP0705; PBP-2) kDa, but direct biochemical evidence for these proteins as PBPs are lacking. An additional protein putatively has been assigned as a serine-type DDcarboxypeptidase (53-kDa, TP0800), and another as a DDcarboxypeptidase TP0221). No ␤-lactamases have been predicted to be present in T. pallidum (9).
The notion that Tp47 is a PBP has been paradoxical. First, Tp47 has no homologies with any other bacterial or eukaryotic proteins. Second, conventional PBPs contain three conserved motifs, SXXK, S(Y)XN, and KT(S)G, which comprise the active site for the covalent binding of ␤-lactams (10 -12). The serine of the SXXK motif is important for nucleophilic attack on the ␤-lactam ring. Tp47 contains three such appropriately spaced hypothetical motifs (SVTK, TEN, KTG) (13). However, preliminary experiments replacing Ser in the SVTK motif of Tp47 with Gly, Ala, Cys, or Thr all yielded mutant enzymes that still bound ␤-lactam comparable with wild-type Tp47 (14). Finally, lipidation of PBPs also is uncommon (15).
The numerous incongruities surrounding Tp47 as a PBP prompted the current biochemical and biophysical study. Specifically, it was envisioned that precise structural information derived from x-ray crystallography could provide strategic information to guide future biochemical studies on the enzymatic activity of Tp47. In this study, it was found that Tp47 has a crystal structure unique to any other known PBP, and thus it appears to represent an entirely new class of PBP.

Construction of Wild-type and Variant Tp47-Streptavidin Fusion
Proteins-The post-translationally modified N-terminal cysteine of native Tp47 was designated as amino acid 1 (16). To ensure production in E. coli of a nonlipidated version of Tp47, a DNA fragment encoding amino acids 2-415 (residue 415 is the last amino acid before the first TAG termination codon in tp47) (16) was amplified by PCR using T. pallidum genomic DNA (17) as template. The PCR primers were 5Ј-tccCCGCGGCTCGTCTCATCATGAGACGCA-3Ј and 5Ј-catgCCATG-GTTACTACTGGGCCACTACCTCGCA-3Ј. The forward primer contained both a tcc overhang and a SacII site (bold); the reverse primer contained a catg overhang, a NcoI site (bold), and two contiguous stop codons (TTA, CTA). PCR amplification was performed using Vent DNA polymerase (New England Biolabs). Amplified fragments were cleaved with SacII and NcoI and cloned directionally into SacII-and NcoIcleaved pASK-IBA7 vector (Sigma). This construct was designated as wild type, and was verified by DNA sequencing and then transformed into E. coli DH5␣.
Site-directed mutagenesis of tp47 was carried out by a PCR-based method, using two complementary mutation-harboring oligonucleotides for each mutant and the QuikChange site-directed mutagenesis kit (Stratagene). Five different mutant genes were constructed; four encoded proteins with single amino acid substitutions and one contained a double substitution. The mutant proteins expressed were designated as Tp47S100G, Tp47S100C, Tp47K287Q, Tp47C296A, and Tp47H5S/ H9S, based on the amino acid positions involved. Finally, a fusion construct in which C-terminal residues 329 -415 were deleted (corresponding to Domain D; see crystal structure Fig. 2 (Tp47⌬D)) also was constructed by PCR subcloning as described above, except that the reverse primer was 5Ј-catgCCATGGTTACTAATCAGCAACTACGT-CC-3Ј. All resulting mutants were sequenced to verify the specific mutation(s) intended. Mutant tp47 genes were expressed in E. coli, and the cognate proteins were purified as described below for wild-type Tp47. SDS-PAGE analysis revealed that the mutants expressed quantities of proteins comparable with wild-type Tp47, suggesting that none of them was unstable (data not shown).
Expression and Purification of Tp47-E. coli DH5␣ containing the respective cloned tp47 gene fusion was grown at 37°C in LB medium containing 100 g of ampicillin per ml; when the A 600 of the culture reached 0.6, the culture was shifted to 30°C and expression of the Tp47-streptavidin fusion protein was induced (via a tetA promoter) by the addition of 200 g/liter of anhydrotetracycline. After 3 h, cells were harvested by centrifugation and solubilized by B-PER II (Pierce). After centrifugation at 15,000 rpm for 20 min (4°C) to remove cellular debris, the supernatant was loaded onto a StrepTactin-Sepharose column. The fusion protein was then purified according to the Strep-tag II protein expression and purification system manual (Sigma). The yield of purified proteins tended to be about 25 mg/liter of bacterial culture. Purified protein was subjected to buffer exchange with buffer A (20 mM Hepes buffer, pH 7.4, 20 mM NaCl) using a PD-10 column (Amersham Biosciences). The protein was then concentrated to about 15 mg/ml using a Centricon YM-10 device (Amicon). Protein purity was analyzed by SDS-PAGE (18) and by electrospray ionization-mass spectrometry (ESI-MS). The concentration of purified protein was estimated spectrophotometrically using a calculated extinction coefficient of ⑀ 280 ϭ 54,050 M Ϫ1 cm Ϫ1 (19).
␤-Lactam Binding to Tp47-Binding of digoxigenin-labeled ampicillin (Dig-Amp) to Tp47 was determined by a chemiluminescent detection method (13,20); the use of Dig-Amp circumvents problems associated with utilizing radiometric methods for assaying ␤-lactam binding (20). ␤-Lactam binding to Tp47 also was examined by ESI-MS; in these experiments, a typical 100-l reaction mixture contained 100 g of protein, 2 mM ZnCl 2 , and 2 mM ␤-lactam (in buffer A) and was incubated at 37°C for various times. The reaction was terminated by the addition of 30 l of 5% formic acid. Excess ␤-lactam was removed by a Microcon YM-30 device (Amicon), and samples were recovered in 1% formic acid for ESI-MS analysis (21). The peak heights of free and acylated Tp47 were measured from the ESI-MS spectra and the percentage of Tp47 acylation was calculated using the equation: % acylation ϭ [acylated Tp47/(acylated Tp47 ϩ free Tp47)] ϫ 100 (22). In an attempt to identify the Tp47 amino acid involved in penicillin binding, liganded sample was digested with trypsin in 100 mM ammonium bicarbonate (pH 7.8, 37°C); after digestion for various times, samples were subjected to MALDI-TOF MS (23).
Kinetic Analysis of ␤-Lactamase Activity-The hydrolytic activity of Tp47 on various ␤-lactams was assessed at 37°C in buffer A using a Shimadzu UV-1601PC UV-visible spectrophotometer equipped with a thermostated multicell transport system. The molar absorption coefficients used were as follows: penicillin G, ⌬⑀ 235 ϭ Ϫ775 M Ϫ1 cm Ϫ1 ; ampicillin, ⌬⑀ 235 ϭ Ϫ820 M Ϫ1 cm Ϫ1 ; nitrocefin, ⌬⑀ 486 ϭ 16,000 M Ϫ1 cm Ϫ1 . ␤-Lactam solutions were freshly prepared in buffer A. Initial rates were determined from the first 5-10% of the reactions at various substrate concentrations. K m and V max values were determined by fitting all data to the Lineweaver-Burk equation using the program UV Probe (Shimadzu).
Tazobactam inhibition of the hydrolytic activity of Tp47 was performed with penicillin as a competitor substrate in buffer A. Tazobactam at various concentrations was preincubated with Tp47 for 5 min at 37°C before the addition of penicillin. Steady-state rates during the course of penicillin hydrolysis were used to calculate the remaining activity. The inhibition constant (K i ) was deduced from Dixon plots using the UV Probe software.
Protein Crystallization and Data Collection-Wild-type Tp47 described above did not yield crystals in preliminary screening experiments. However, one of the variant versions of Tp47, in which His-5 and His-9 were replaced with Ser (Tp47H5/H9S; Fig. 1), crystallized readily and thus was designated as crystallizable Tp47 (cTp47). Of particular importance, cTp47 retained PBP activity comparable with the wild-type (Fig. 1). cTp47 was crystallized by the hanging-drop vapor diffusion method (24) using 24-well Linbro plates (Hampton Research) at room temperature. Sparse matrix crystallization kits (Hampton Research) were used to screen preliminary crystallization conditions. Crystals of average dimension of 50 m appeared within 3-4 weeks. Further growth of the crystals was hindered because of phase separation/oil formation, and these crystals diffracted poorly to a Bragg spacing (d min ) of 6 Å. Crystallization optimization using dextran sulfate eliminated the phase separation and yielded substantially larger crystals (about 500 m) within 2-4 days that diffracted to better than a d min of 3 Å. Crystals were routinely obtained with drops containing 5 l of protein solution (about 15 mg/ml in buffer A), and 5 l of 32% (w/v) PEG 4000 in 100 mM sodium citrate, pH 5.6, 200 mM ammonium acetate, 3% (w/v) dextran sulfate 8000 (Sigma), and Ϯ100 M ZnCl 2 , equilibrated against 500 l of the latter solution at room temperature. Prior to data collection, crystals were transferred sequentially for 5 min to each of 5, 10, and 15% (v/v) glycerol-enriched reservoir solution for cryogenic conditioning. Diffraction data were collected at 100 K using a Rigaku RU300 rotating copper anode x-ray generator and R-axis IV image plate detector (Molecular Structures Corp., The Woodlands, TX). The diffraction data were indexed, integrated, and scaled in the HKL2000 program package (25).
The cTp47 crystals were found to exhibit the symmetry of space group P3 2 21 with unit cell dimensions of a ϭ b ϭ 129.1 Å, c ϭ 151.5 Å. The crystals contained two molecules per asymmetric unit. The crystal structure of cTp47 was determined by single wavelength anomalous dispersion using a xenon derivative. The xenon derivative of a cTp47 crystal was prepared by exposing a preconditioned native crystal (in glycerol-enriched reservoir solution) in a xenon chamber (kindly provided by Zhenming Wang) at 400 p.s.i. for 15 min at room temperature. The chamber was then depressurized and the crystal flash-cooled in liquid propane within 15 s. Diffraction data to a d min of 2.28 Å were recorded. The data were reduced with the program package HKL2000. Xenon sites were identified and refined to 3.0 Å within the program package CNS (version 1.0) (26), resulting in an overall figure of merit of 0.35. The phases were further improved by density modification in CNS including histogram matching, solvent flipping, and phase extension to a d min of 2.28 Å, resulting in a final figure of merit of 0.95 (Table I).
After the structure was solved, a synchrotron data set on a xenonderivatized cTp47 crystal was collected to a d min of 1.95 Å at the Structural Biology 19-ID beamline at the Advanced Photon Source (Argonne National Laboratory, Argonne, IL). Data collection and single wavelength anomalous dispersion phasing statistics are provided in Table I.
Model Building and Structure Refinement-Model building was performed automatically (arp_warp 5.0) (27) and manually with the program O (28). Structure refinement using the synchrotron data set was carried out within CNS employing cycles of simulated annealing, conjugate gradient minimization, and calculation of individual atomic displacement parameters. An overall anisotropic atomic displacement parameter and bulk solvent correction were used throughout the refinement procedure. Water molecules were added where stereochemically reasonable after the protein part of the model was complete. Inspection of the F obs Ϫ F calc difference density map revealed a large volume of positive difference density extending across the noncrystallographic 2-fold axis, and located in the positively charged cleft between domains B and C of each monomer. This density was modeled as a dextran sulfate polysaccharide with an ␣136 linkage and two sulfate groups (on O-2 and O-3) per glucose. The final model contains residues 7 to 34 and 44 to 414 of molecule A, and residues 7 to 34 and 40 to 413 of molecule B, 14 residues with alternate conformations, five xenon atoms, two complete and two partial sugar moieties of a dextran sulfate polysaccharide, and 407 water molecules. Residues 2 to 6, 35 to 43, and 415 in molecule A, and residues 2 to 6, 35 to 39, and 414 to 415 in molecule B were disordered in the crystal structure and could not be traced in the electron density. The final R free value is 23.5% and the R work value is 21.2% (Table I).
Analytical Ultracentrifugation-Sedimentation equilibrium studies were performed in a Beckman XL-1 Optima analytical ultracentrifuge at 4°C. Tp47 samples corresponding to absorbancies of 0.1, 0.2, and 0.4 at 280 nm in buffer A were used. Samples were centrifuged at 14,000 ϫ g to remove aggregates prior to loading. Experiments were conducted at a rotor speed of 13,000 and 18,000 rpm and the radial scans at 280 nm were recorded until equilibrium was reached. The sedimentation equilibrium data were analyzed using the supplied software.

RESULTS AND DISCUSSION
Expression and Purification of Tp47-Bacterial lipoproteins are membrane proteins by virtue of their three long-chain fatty acids (post-translationally added to an N-terminal cysteine) that serve solely as membrane insertion anchors (29). As such, the long-chain fatty acids do not contribute to the conformation of the protein. The proteins, in the absence of their acyl chains, thus tend to be water soluble (consistent with the polypeptides protruding into the periplasm or extracellular environment). A cloning strategy therefore was implemented in which the leader sequence and N-terminal cysteine codon of tp47 were deleted, ultimately to yield a nonlipidated, water soluble version of Tp47. Finally, soluble Tp47 and its variants were created as fusion proteins with an N-terminal streptavidin tag, which is only 18 amino acids long; the streptavidin tag thus should have minimal, if any, conformational influence on Tp47. This contention was corroborated by the findings that the fusion proteins performed as predicted in PBP assays (see below).
Properties of Mutant Tp47 Enzymes-Ser-100 of a putative SVTK tetrad in Tp47 (13) was altered to cysteine (Tp47S100C); this mutation did not abolish Dig-Amp binding (Fig. 1) or penicillin binding to Tp47 (ESI-MS data not shown). Similarly, conversion of Ser-100 to glycine did not abrogate the binding of penicillin to Tp47, as assessed by ESI-MS (not shown). Thus, as initially proposed (14), it appears that Tp47 does not employ an active-site serine to serve as a nucleophile and subsequent covalent attachment site for ␤-lactams. This is in sharp contrast to what has been observed for other classical PBPs (30, 31). That a mutation of the presumptive active site serine had no influence on the PBP activity of Tp47 provided the first compelling evidence that Tp47 might be dissimilar from other conventional, serine-type PBPs.
The KTG triad also forms a key component of the active site cleft and is highly conserved within PBPs (10 -12). However, if Tp47 is not a serine-type PBP, it was postulated that the KTG motif in Tp47 may be coincidental, or may function in some other unknown manner. For example, the positive charge on the Lys might interact with the carboxylate group of the D-Ala- D-Ala, and hence the carboxylate group of penicillin (30, 32). However, when Lys-287 of the KTG triad in Tp47 was mutated to Gln (Tp47K287Q), the mutant protein retained its penicillin binding activity (Fig. 1). Inasmuch as the mutation of Lys in the KTG motifs of other PBPs typically adversely impacts PBP activity (30), our results further underscore the atypical character of Tp47.
cTp47 Structure-Findings that Tp47 seemed not to rely on either an active site serine (of SVTK) nucleophile or a KTG motif for PBP activity were anomalous. This prompted a structural approach to gain further insights into the structurefunction relationships for Tp47 as perhaps a novel PBP. Initially, crystal growth of cTp47 was hampered by the occurrence of phase separation, and the resulting crystals were small and diffracted poorly to a Bragg spacing (d min ) of 6 Å. Phase separation could be overcome by the addition of dextran sulfate, resulting in larger crystals (up to 500 m in the largest dimension) that diffracted to a d min of 3 Å. These crystals exhibited the symmetry of space group P3 2 21, with two molecules per asymmetric unit. The crystal structure of cTp47 was determined via the single wavelength anomalous dispersion technique using a xenon derivative. Derivatization with xenon not only provided phase information, but also increased the diffraction limit of the cTp47 crystals to a d min of 1.95 Å using synchrotron radiation.
The crystal structure of cTp47 revealed four distinct domains arranged to give the molecule a crab-like appearance (Fig. 2). The first domain (domain A; residues 7 to 34 and 156 to 204) is mainly composed of ␤-strands and is sequentially non-contiguous. The core of this domain is formed by a strand-helixstrand motif (A␤2-A␣3-A␤3) (Fig. 3) in a right-handed superhelical arrangement. Adjacent to A␤2 is a ␤-hairpin (strands A␤4 and A␤5) whose tip interacts with the helix to create a barrel-like structure. The N terminus of cTp47 forms a ␤-strand (A␤1) that inserts between A␤2 and A␤3 to complete a five-stranded, highly twisted, mixed ␤-sheet (order 3, 1, 2, 4, 5). A helix-loop-helix motif (A␣1 and A␣2) next to the ␤-hairpin completes domain A and connects to the adjacent domain B. A structural comparison of this domain using the program DALI (33) did not reveal any similarity with proteins in the Protein Data Bank (highest Z-score of 1.7). The largest recognizable structural motif within this domain is generated by strand A␤1, helix A␣3, and strand A␤3 that forms an anti-parallel two-stranded ␤-sheet with an opposing helix. This motif also has been observed in the Lactobacillus casei Hpr kinase (Protein Data Bank code 1jb1).
Domain B (residues 44 to 155) contains 10 ␤-strands and a single ␣-helix (Figs. 2 and 3). Its main structural feature is a central four-stranded, anti-parallel ␤-sheet (strands B␤1, B␤10, B␤2, and B␤5). This sheet is opposed by an ␣-helix (B␣1) resulting in an arrangement that resembles a right hand, with the strands being the fingers (strand B␤1 is the index finger) and the helix as the thumb. The backside of the sheet forms a flat outer surface. At the N and C termini of strand B␤5 are two large ␤-hairpins (strands B␤3/B␤4 and B␤8/B␤9) that are ori-ented perpendicular to the central sheet. These hairpins, together with large connecting loops and a third ␤-hairpin (strands B␤5/B␤6) in between them, form a second flat outer surface. The central motif consists of strands B␤1, B␤10, and B␤3, and helix B␣1, which is typical of cysteine proteases (34). In fact, the topology of domain B in cTp47, except for the hairpin formed by strands B␤3 and B␤4, is conserved in the cysteine protease staphopain from Staphylococcus aureus (Protein Data Bank code 1cv8). Yet, Tp47 does not appear to be a cysteine protease as a cysteine is not present in a region equivalent to the active site in cysteine proteases. Furthermore, mutation of the sole Cys (Cys-296, which is buried in the hydrophobic core of domain C) to alanine had no effect on PBP activity (Fig. 1), thereby ruling out involvement of this residue in catalysis.
Domain C (residues 205 to 332) is the largest domain (Fig. 2). It is mainly characterized by an immunoglobulin fold with two opposing ␤-sheets that form the typical barrel-like structure. In contrast to the classical immunoglobulin fold, however, domain C of cTp47 has an additional ␤-strand inserted after strand 3. Also, the strands are connected by rather large loops. Helices are inserted between strands 2 and 3 and between strands 4 and 5.
Domain D (residues 333 to 414) also features an immunoglobulin fold. In contrast to domain C, it contains only the characteristic seven-stranded barrel and short loops. As in domain C, a single ␣-helical turn is inserted between strands 2 and 3.
Dimer Formation-In our crystals of cTp47, a dimer was formed between two neighboring molecules (Fig. 4). Domains B and D act as the pincers on a crab that make contact with the pincers of the opposing molecule. The monomer-monomer interface has an area of about 1,830 Å 2 and features a series of polar and hydrophobic interactions as well as six ionic interactions. This finding prompted further assessment of Tp47 dimer formation in free solution by analytical ultracentrifugation. The sedimentation equilibrium data profile produced by analytical ultracentrifugation fit well to a model comprising a single species of molecular mass of 46,178 Da (not shown), consistent with the monomeric mass determined by SDS-PAGE and ESI-MS, supporting the observation that Tp47 displays monomeric characters in free solution. Consequently, Tp47 dimer formation observed within the crystal structure could be a result of crystallization, with the high salt concentration driving a nonspecific association of the hydrophobic surfaces, as has been noted for other proteins undergoing crystal packing (35,36). In fact, when domain D (which is not required for PBP activity) is removed from the buried surface area calculation, only ϳ850 Å 2 surface area is buried at the monomer-monomer interface. This is a value found at the upper limit of buried surface area for nonspecific crystal contacts (37).
Domain Interfaces-The first three domains in cTp47 interact with each other through intimate domain-domain interfaces. Domain A contacts domain B through its N-terminal segment that contains ␤-strand A␤1 and the helix-loop-helix motif, establishing interactions with the loop regions before the first ␤-strand (B␤1) and after the last ␤-strand (B␤10) in domain B. The first linking region between these domains (residues 34 and 44) is disordered in the crystal structure. Domain A also interacts tightly with domain C, involving mainly side chains in helix A␣2 and the loop region between ␤-strands A␤3 and A␤4 in domain A and ␤-strands C␤3␤ and C␤6 as well as the loop region between strands C␤6 and C␤7.
Domain B interacts with domain C via a surface that has a slightly concave, goblet-like shape. The long loops proximal to strand B␤1 and between strands B␤5 and B␤6 form the sides, and helix B␣1 forms the bottom of the goblet. Residues in these FIG. 1. Binding of Dig-Amp to wild-type (Wt) and mutant variants of Tp47. Recombinant versions of Tp47 incubated with Dig-Amp were separated by SDS-PAGE, electrotransferred to nylon membrane, and developed by chemiluminescence (13,20). Dig-Amp binding was assessed in the presence of ZnCl 2 except where noted (ϪZnCl 2 ). ϪDig-Amp, wild-type Tp47 without Dig-Amp treatment.
regions establish a number of polar and hydrophobic interactions with residues at the surface of domain C, which includes strands C␤3, C␤3␤, and C␤4, the loop region between strands C␤5 and C␤6, and helix C␣1. Adjacent to this interaction surface is a deep cleft located between the ␤-hairpin B␤3/B␤4 and the rest of domain B. The tip of this hairpin, as well as the portion of this surface that is not involved in interactions with domain C, are highly positively charged containing five arginines, two histidines, and two lysines.
In contrast to domains A, B, and C, domain D is rather isolated. It interacts only with domain C via an ionic interaction between Arg-330 and Glu-404 in the linker region. Consequently, the relative disposition of domain D is expected to vary. Evidence for a larger degree of domain motion can be found in the higher average displacement factors for the atoms of domain D relative to the first three domains (38.8 versus 57.8 Å 2 for monomer A, 38.0 versus 67.4 Å 2 for monomer B).
Comparison of Tp47 with Other PBP Structures-The three-dimensional structure of a conventional PBP typically is comprised of two structural domains, one of which is predominantly ␣ and another that is ␣/␤ (38) (Fig. 5). The active site is positioned between these two major domains, at the edge of the central ␤-sheet of the ␣/␤ domain. The three signature sequence motifs of classical PBPs that putatively were present in Tp47 do not come together in three-dimensional space to form a typical active site. Given that Tp47 had no similarity to other known PBPs, it was hypothesized that it might represent a new family of PBPs. Consistent with this possibility, DALI did not identify Tp47 as a PBP, but rather had the highest structural homology (Z-score ϭ 6.1) to non-PBPs. Acylation and Deacylation of Tp47-The interaction between PBPs and ␤-lactams generally is described by the equation: E ϩ I u E.I 3 E-I 3 E ϩ P, where E is the PBP enzyme, I is the ␤-lactam, E.I is the Michaelis intermediate, E-I is the covalent acyl-enzyme complex, and P is the reaction product (i.e. cleaved, inactive ␤-lactam) (38). The formation of the enzymatically inactive (covalent) acyl-enzyme complex (E-I) is known as the acylation step. The covalent E-I complex results from the nucleophilic attack of the carbonyl carbon atom of the ␤-lactam ring by the hydroxyl group of the active site serine. The bactericidal efficiency of any ␤-lactam ultimately depends on the stability of the E-I complex. However, hydrolysis of the acylenzyme complex and release of the inactive ␤-lactam (P) occurs by a process known as deacylation; in the case of ␤-lactamases, deacylation is rapid. In former studies, Tp47 bound radiolabeled penicillin (7), and its binding to Dig-Amp subsequently was found to be stimulated by zinc ions (13). In the current study, upon incubation of purified Tp47 for 2 min with penicillin in the presence of zinc, two major peaks of 47,703 Da (free Tp47) and 48,036 Da (penicillin-bound Tp47) were detected by ESI-MS (not shown). The difference of 333 Da between the two molecular masses corresponded with the mass of penicillin (335 Da), indicating the formation of a covalent acyl-Tp47 complex bound predominantly in a 1:1 stoichiometry. Analogous results were obtained using ampicillin, carbenicillin, cefuroxime, and cephalosporin (not shown), indicating that recombinant Tp47 bound a number of ␤-lactams. In the absence of zinc, after 2 min of incubation, 5% of Tp47 became acylated, whereas, in the presence of zinc, 33% of Tp47 was acylated over the same interval (Table II), corroborating previous findings that the PBP activity of Tp47 appears to be stimulated by zinc (13). In the presence of zinc, acylation by penicillin was time-dependent, with maximal binding observed at 6 min (Table II). However, after 6 min, marked deacylation was evident, implying that Tp47 exhibits some intrinsic ␤-lactamase activity.
As shown in a previous study (13) and herein, zinc enhances the binding of ␤-lactams to Tp47. This led to the initial idea that Tp47 was a zinc-dependent PBP (13). Two lines of evidence now challenge this view. First, we now show that rather than promoting acylation, zinc actually inhibits the deacylation of Tp47 (see below). Second, an in vitro carboxypeptidase assay using the synthetic depsipeptide substrate Sle (an analog of D-Ala-D-Ala) initially suggested that Sle was hydrolyzed by Tp47 in the presence of zinc, as indicated by an apparent increase in UV absorption at 254 nm (13). However, subsequent experiments have revealed that this apparent absorption increase is due, at least in part, to scattering caused by Tp47 aggregates that form in the presence of zinc (not shown). Hence, the initial contention that Tp47 might be a zinc-dependent carboxypeptidase (13) remains tenuous at this time.
Mass spectrometry has been employed for the identification of the penicillin-binding site in Staphylococcus aureus PBP 2a (21). Using a similar strategy, liganded Tp47 was digested with trypsin, and peptide fragments were assessed by MALDI-TOF MS. Attempts to identify a particular peptide fragment to which penicillin was bound were unsuccessful, suggesting that the acylated product was unstable during the procedure. One potential explanation for this was the intrinsic ␤-lactamase activity inferred in Table II.  5. Comparison of the cTp47 structure to representative ␤-lactamases and PBPs. Representative structures from the major classes of ␤-lactamases plus a D-Ala-D-Ala-peptidase/PBP are shown with domains A-C of Tp47 (domain D is not required for PBP/␤-lactamase activity). Black arrows highlight the known active sites of the representative structures. The green sphere in the Class B structure represents a Zn 2ϩ ion. The Class A structure is the TEM1 ␤-lactamase from E. coli (Protein Data Bank code 1btl), the Class B structure is the zinc metallolactamase from Bacillus cereus (Protein Data Bank code 1bmc), the Class C structure is the cephalosporinase from Enterobacter cloacae (Protein Data Bank code 2blt), the Class D structure is the Oxa-10 ␤-lactamase from Pseudomonas aeruginosa (Protein Data Bank code 1e3u), and the PBP structure is the D-Ala-D-Ala-peptidase/PBP from P. aeruginosa (Protein Data Bank code 1ceg). Kinetic Parameters for Tp47 ␤-Lactamase Activities-Certain PBPs have intrinsic ␤-lactamase activity (30, 39). Kinetic analysis of ␤-lactam hydrolysis was used to assess whether the deacylation of Tp47 (Table II) was because of a similar intrinsic ability to hydrolyze ␤-lactams. The kinetic parameters of hydrolytic activities of Tp47 were determined for three ␤-lactams and are summarized in Table III. Tp47 exhibited an unexpectedly high level of ␤-lactam hydrolytic activity. Although the turnover rates (k cat ) for ␤-lactam hydrolysis by Tp47 were 10 -20-fold lower than for typical ␤-lactamases (40), they are substantially higher than the ␤-lactamase activity of E. coli PBP5, which has an unusually high ␤-lactamase activity (k cat ϭ 0.07 s Ϫ1 ) (39). On this basis, it could be conjectured that Tp47 is a ␤-lactamase. However, from a biological perspective, this notion is strongly inconsistent with the exquisite sensitivity of T. pallidum to ␤-lactams, particularly when the extraordinary abundance of Tp47 in T. pallidum is taken into account (41). Thus, the biological relevance of the putative Tp47 in vitro ␤-lactamase activity remains suspect, as it may be of little or no consequence to the biology of T. pallidum in vivo (i.e. during human infection). Interestingly, a higher level of penicillin binding to Tp47 was observed in the presence of zinc ( Fig. 1 and Table II). As noted earlier, zinc also induces the aggregation of Tp47 (not shown), which appears as a suppression of in vitro ␤-lactamase activity. Taken together, it is tempting to speculate that the enhanced PBP activity of Tp47 has been observable, at least in part, by virtue of the inhibitory action of zinc on the intrinsic ␤-lactamase activity of Tp47. Finally, the in vitro hydrolytic activity of wild-type Tp47 was inhibited by tazobactam, an inhibitor of class A ␤-lactamases (42,43), suggesting that competitive inhibition is active site directed. The apparent K i value for hydrolysis of penicillin by wild-type Tp47 was 26.95 Ϯ 0.35 nM.
Potential Active Site-Catalytic centers of PBPs have a conserved topology wherein three conserved motifs comprise the catalytic center (10,11). The sequence of Tp47 has three such hypothetical signature motifs (13). However, mutations in the Ser of the putative SVTK motif and Lys of the KTG motif did not abrogate the PBP activity of Tp47 (Fig. 1). Furthermore, all three motifs of classical PBPs initially thought to be present in Tp47 are found in three different domains separated by distances greater than 30 Å (Fig. 2), supporting the contention that the three hypothetical motifs do not comprise the active site for ␤-lactam binding in Tp47. We thus conclude that Tp47 exhibits a unique mechanism for ␤-lactam binding. Further inspection of the structure therefore was undertaken to identify the active site. Emphasis was placed on searching for another reasonable PBP active site cleft, which might contain a Ser nucleophile spatially near another residue suitable for abstraction of a proton from the hydroxyl group of Ser (e.g. a positively charged amino acid such as Lys). Such efforts were not successful.
The predominance of hydrophobic residues and the immunoglobulin fold of domain D suggested that it might be utilized for protein-protein interaction(s) when in its native membrane setting within T. pallidum. In addition, the location, flexibility, and relative disposition of domain D suggests that it might not be involved in PBP and ␤-lactamase activities. In this regard, a domain D deletion mutant of Tp47 (Tp47⌬D) retained wildtype levels of both activities (not shown). Thus, it is reasonable to conclude that domain D has no catalytic role in the PBP activity of Tp47.
An analysis of the charge distribution on the surface of domains A-C of the Tp47 monomer is shown in Fig. 6. A positively charged cleft is found at the intersection of domains B and C, close to the domain B ␤-hairpin formed by strands B␤3 and B␤4. This cleft might function as a binding site for the carboxylate of D-Ala-D-Ala, and hence ␤-lactams. In the crystal structure, this cleft is found near the noncrystallographic 2-fold axis of the dimer. A dextran sulfate polysaccharide with an ␣136 linkage was modeled into the positive difference density found in this cleft. Approximately one-half of the electron density assigned to the polysaccharide is associated with each protein monomer, and the hydrogen-bonding pattern between the protein and each sulfated dextrose monosaccharide is similar. An attempt to model the polysaccharide backbone of naturally occurring peptidoglycan (repeating N-acetylmuramic acid (NAM) ␤134 linked to N-acetylglucosamine) into this density was not successful. An NAM monomer could be modeled into the density, but the ␤134 linkage of NAM-N-acetylglucosamine was inconsistent with the local 2-fold symmetry of the cleft. If Tp47 utilizes this cleft for the interaction of ␤134linked peptidoglycan subunits, it appears that steric constraints dictate that the protein be in the monomeric state, as supported by our sedimentation equilibrium experiments. Whereas small crystals of cTp47 normally can be grown in the absence of dextran sulfate, crystallization with NAM or Nacetylglucosamine monosaccharides in place of dextran sulfate did not yield cTp47 crystals.
Further attempts to identify the active site of Tp47 by cocrystallization and/or soaking of crystals with ␤-lactams were unsuccessful, probably because of the deacylation activity noted earlier. A 3.8-Å data set was obtained from a co-crystallization and soak of cTp47 with the ␤-lactamase inhibitor tazobactam. The electron density map revealed changes in the positively charged cleft that may be because of a partial displacement of the dextran sulfate polysaccharide by the tazobactam, but an unambiguous fit of the inhibitor into this low resolution map was not possible.
Biological Significance and Implications-Tp47 was first noted in early molecular studies of T. pallidum, due largely to its abundance and profound immunogenicity (41). It thus initially was targeted for study as a potential syphilis serodiagnostic reagent (41,44), and many newer generation serological tests for syphilis now include Tp47 as a principal, if not sole, antigenic component (45,46). Tp47 initially also was thought to be an outer membrane protein (41). However, a more extensive  6. The cTp47 monomer has a positively charged cleft. A surface representation of the electrostatic charge distribution for the Tp47 monomer (domains A-C) is shown at the left of the figure and is in the same orientation as in Fig. 2. The central figure was obtained via a rotation of 90°about the horizontal axis of the monomer. For comparison, the charge distribution for the active site cleft in the D-Ala-D-Ala-peptidase/PBP (Protein Data Bank code 1ceg) is shown at the right of the figure. The displayed surface potential varies approximately from Ϫ10 to 10 kT with acidic surfaces in red and basic in blue. The electrostatic surface potential was calculated and rendered in the program GRASP (52). body of work, which has taken into account the previously unrecognized fragility of the unusual T. pallidum outer membrane (1,47), later supported that Tp47 likely is a cytoplasmic membrane lipoprotein that, according to convention, would protrude into the periplasmic space (47,48). This finding was more consistent with earlier studies that implicated it as a PBP (7,13), inasmuch as PBPs reside at the cytoplasmic membrane (2). However, the precise role of Tp47 in the biosynthesis of T. pallidum peptidoglycan remains unclear. Although corroborative data are lacking, it is possible, implicated largely by its molecular mass, that Tp47 is a DD-carboxypeptidase. If so, the marked abundance of Tp47 would imply that it serves to limit the degree of cross-linking in the peptidoglycan of T. pallidum, thereby promoting the rather remarkable, highly flexuous motility pattern of the spirochete (49). Consistent with this view, other preliminary data have suggested that the expression of full-length, lipidated Tp47 in E. coli (13) reduces the degree of cross-linking in E. coli peptidoglycan. 2 Despite both mutagenesis and x-ray crystallography data presented herein, identification of the putative active site of Tp47 for ␤-lactam binding remains unresolved. The three-dimensional structure of Tp47 has revealed a positively charged cleft that may bind monosaccharides and/or possibly tazobactam, and that cleft might function as an interaction site for the relevant carboxylate group of D-Ala-D-Ala (and ␤-lactams), but more conclusive evidence awaits further mutagenesis studies. Regardless, the combined data provide compelling evidence that Tp47 represents a new class of PBP. It also is not known to what extent this novel type of PBP might be found in other bacterial pathogens, but it is anticipated that the burgeoning genomics field will eventually shed additional light on this. Finally, it is noteworthy that although not sharing homology with Tp47, a completely ␣-helical cysteine-rich protein B of Helicobacter pylori recently was described as representing another new class of PBP (50). Although molecular modeling inferred that a site within the ␣-helical cysteine-rich protein B might bind to NAM, the crystal structure also did not definitively reveal the active site. Tp47 and the ␣-helical cysteinerich protein B thus now seem to represent two examples of PBPs that do not satisfy classical PBP paradigms, the ramifications of which remain to be more fully explored.