|Home | About | Journals | Submit | Contact Us | Français|
Bacillithiol (Cys-GlcN-malate, BSH) has recently been identified as a novel low-molecular-weight thiol in Bacillus anthracis, Staphylococcus aureus, and several other Gram-positive bacteria lacking glutathione and mycothiol. We have now characterized the first two enzymes for the BSH biosynthetic pathway in B. anthracis, which combine to produce α-D-glucosaminyl L-malate (GlcN-malate) from UDP-GlcNAc and L-malate. The structure of the GlcNAc-malate intermediate has been determined, as have the kinetic parameters for the BaBshA glycosyltransferase (→GlcNAc-malate) and the BaBshB deacetylase (→GlcN-malate). BSH is one of only two natural products reported to contain a malyl glycoside, and the crystal structure of the BaBshA-UDP-malate ternary complex, determined in this work at 3.3 Å resolution, identifies several active-site interactions important for the specific recognition of L-malate, but not other α-hydroxyacids, as acceptor substrate. In sharp contrast to the structures reported for the GlcNAc—1-D-myo-inositol-3-phosphate synthase (MshA) apo and ternary complex forms, there is no major conformational change observed in the structures of the corresponding BaBshA forms. A mutant strain of B. anthracis deficient in the BshA glycosyltransferase fails to produce BSH, as predicted. This B. anthracis bshA locus (BA1558) has been identified in a transposon site hybridization study as required for growth, sporulation, or germination, suggesting that the biosynthesis of BSH could represent a target for development of novel antimicrobials with broad spectrum activity against Gram-positive pathogens like B. anthracis. The metabolites that function in thiol redox buffering and homeostasis in Bacillus are not well understood, and we present a composite picture based on this and other recent work.
In his recent review on the management of oxidative stress in Bacillus, Zuber (1) concludes that the metabolites that function in redox buffering and thiol homeostasis, and their influence on the oxidative stress response, are not well understood. Earlier work from this laboratory (2) demonstrated that CoASH provided the major low-molecular weight thiol redox buffer in Bacillus anthracis, replacing GSH as had previously been demonstrated for Staphylococcus aureus (3). The likelihood that CoASH plays an important functional role in redox buffering and thiol homeostasis is strengthened by the demonstration that both B. anthracis (4) and S. aureus (3, 5-7) have NAD(P)H-dependent coenzyme A-disulfide reductases (CoADRs).1 We have also shown that the type III pantothenate kinase, which is unusual in its insensitivity to feedback inhibition by CoASH (2), is essential for growth of B. anthracis (8). CoADR, the chimeric CoADR-RHD protein (9), and the type III pantothenate kinase represent three well-characterized adaptations to the CoASH-based redox buffer system for this Gram-positive pathogen.
A new unknown thiol, originally referred to as U12, was also identified in the B. anthracis extract (2); a mass of 398 Da was reported for U12-SH. By combining analytical chemical approaches with mass spectrometry and NMR, the structure of U12 has been determined to be N-cysteinyl-α-D-glucosaminyl L-malate, and U12 has been renamed bacillithiol [BSH (10)]. BSH is present at intracellular concentrations of 0.1-0.35 mM 2 in a number of Bacillus species, including B. anthracis, Bacillus subtilis, and Bacillus megaterium, as well as in S. aureus. While none of the four B. subtilis strains represented in the NCBI database have orthologs of CoADR or any of its isoforms (9), B. anthracis, B. megaterium (11), and S. aureus (3) maintain both CoASH and BSH redox buffer systems, in the absence of GSH. Pöther et al. (12), in a recent study with S. aureus demonstrating that the majority of reversible protein thiol oxidations observed during treatment of cells with diamide are based on S-thiolations with Cys rather than protein disulfide formation, have concluded that Cys also functions as an important thiol redox buffer in S. aureus.
In an initial study of the biosynthesis and functions of BSH (13), we have demonstrated that the B. subtilis YpjH protein (BsuBshA), the enzyme orthologous to B. anthracis ORF BA1558, catalyzes the synthesis of N-acetyl-α-D-glucosaminyl L-malate (GlcNAc-malate) from UDP-GlcNAc and L-malate; the product structure has been confirmed by mass spectrometry (10, 13). We have also demonstrated that the BA1557 gene product (BaBshB) catalyzes the deacetylation of GlcNAc-malate (13), providing the free 2-amino group of the GlcN moiety required for the Cys ligation step that is proposed to complete the synthesis. Deletion of the B. subtilis yllA locus, predicted to be involved in the biosynthetic pathway and initially suggested to encode this putative ligase, does eliminate BSH production; the recombinant protein, however, does not catalyze ATP-dependent BSH formation from Cys and GlcN-malate, as assayed in vitro.
In the absence of functional information, Ruane et al. (14) have reported the structure of the ORF BA1558 (identified in this work as BaBshA) apoenzyme, refined at a resolution of 3.1 Å. Although this enzyme, like the MshA glycosyltransferase that produces GlcNAc-Ins-P in the first step of mycothiol (MSH) biosynthesis (15, 16), is a member of the GT-B and GT4 fold (17) and Carbohydrate-Active enZymes (18) families, respectively, structural homology searches using the MshA apoenzyme structure were unsuccessful in identification of other GT-B family members, including ORF BA1558. This indicates a major conformational difference between the two apoenzymes, despite the similarity in reactions catalyzed. Fadouloglou et al. (19) reported the crystal structure of the Zn2+-binding BcZBP protein from Bacillus cereus, as refined at a resolution of 1.8 Å. BcZBP is the enzyme orthologous to BaBshB, the GlcNAc-malate deacetylase, with 97% sequence identity. A recent kinetic analysis (20) with GlcNAc and several GlcNAc oligomers indicated that, although the biological function and natural substrate for BcZBP were unknown, the enzyme had optimal activity with (GlcNAc)2 (kcat/Km = 3.3 × 107 M-1s-1 at 37°C). The enzyme is a member of the PIG-L superfamily represented by GlcNAc-phosphatidylinositol deacetylase (21), the MshB deacetylase (22) of the MSH biosynthetic pathway, and the mycothiol S-conjugate amidase (23). The active-site Zn2+ of BcZBP is buried at the bottom of a ca. 12 Å deep cavity on the hexamer surface, and Arg140 is prominently positioned at the entry of the cavity. Both R140A and R140E mutants exhibit kcat/Km values 0.2-0.3% that of wild-type enzyme (20). His12, Asp15, and His113 provide the protein ligands to the Zn2+, and Asp14 has been proposed as an acid-base catalyst in the hydrolase reaction. A model of a BcZBP-GlcNAc complex reveals that this substrate is considerably smaller than the volume of the cavity; a hydrophilic patch consisting of the side chains of Asp108, Asn150, and Tyr194 marks an unoccupied region proximal to the Zn2+. The Y194F mutant specificity constant, which is <0.1% that of wild-type (20), reflects a dominant effect on Km. In particular, the anomeric C1-OH of GlcNAc also appears to be accommodated by a relatively large space within the cavity. For the BcZBP apoenzyme structure, three very mobile loops have recently been implicated in determining active-site accessibility and regulating substrate specificity (24).
Given the goal of better defining the thiol redox buffer in B. anthracis and characterizing potential new targets for the development of antimicrobial agents that are selective against B. anthracis, S. aureus, and other Gram-positive pathogens that rely on the BSH-based redox buffer system, we report here detailed kinetic analyses of the BaBshA glycosyltransferase and BaBshB deacetylase. The kinetic studies are interpreted in view of the ORF BA1558 and BcZBP crystal structures, and we also provide the structure of the BaBshA-UDP-malate complex, refined at a resolution of 3.3 Å, which demonstrates the absence of any major conformational change on ligand binding. In addition we show that deletion of the BAS1445 locus encoding BshA in B. anthracis Sterne generates a BSH-deficient strain appropriate for testing in a murine model of inhalational anthrax.
The B. anthracis Sterne 34F2 strain (pXO1+ pXO2-) used in these studies was obtained from Dr. Philip Hanna, University of Michigan Medical School. Cultures were grown and maintained in BHI broth (Difco) and on solid media containing 15 g/L of agar. For preparation of parental and mutant endospores, cultures were grown at 37°C on BHI plates containing antibiotics as appropriate. A single colony was inoculated into 3 mL of BHI broth containing antibiotics as required, and this culture was taken as the inoculum (5%, v/v) into 75 mL of fresh modified G (sporulation) medium (25) without antibiotic, after 8-12 h of growth. After 4 days’ growth at the appropriate temperature endospores were collected by centrifugation, and residual vegetative cells were killed by a 30-min incubation at 65°C. Pellets were washed 3-4 times in deionized water, and endospores were stored (in deionized water) at ambient temperature. The purity of the spore preparations was confirmed by phase-contrast microscopy, and spore titers were determined by serial dilution. Escherichia coli strains DB3.1 (26) and GM272 were used for the cloning and propagation of pBKJ236 and pBKJ223 (27), respectively. DB3.1(pBKJ236) was cultured and maintained in LB broth or on solid media supplemented with 250-400 μg/mL erythromycin. GM272(pBKJ223) was cultured and maintained in LB broth or on solid media supplemented with 15 μg/mL tetracycline.
In-frame deletion mutants of the genes encoding BaBshA (bshA, loci BA1558 and BAS1445 in Ames and Sterne strains, respectively), BaBshB (bshB, loci BA1557 and BAS1444, respectively), BaCoADR [cdr, loci BA1263 (4) and BAS1170, respectively] and BaCoADR-RHD [cdr2, loci BA0774 (9) and BAS0736, respectively] were constructed following the markerless gene replacement protocol (27), using the pBKJ236 and pBKJ223 plasmids. The procedure followed in this work is described in detail for the BASΔ1445 mutant. A modified chromosomal segment containing the BAS1445 deletion was constructed by first amplifying the flanking upstream and downstream regions (>500 bp each) using primer sets BAS1445up (FWD/REV) and BAS1445dwn (FWD/REV), respectively (Table S1). The two amplified products were then ligated using overlap PCR with BAS1445up FWD/BAS1445dwn REV, and the resulting deletion fragment was TA cloned into pCR2.1 (Invitrogen) before subcloning into pBKJ236 digested with NotI. This plasmid, pBKJ236::BAS1445′ was electroporated into methylation-deficient (dam- dcm-) E. coli GM272. E. coli GM272(pBKJ236::BAS1445′) (LB plus 250 μg/mL erythromycin), E. coli helper strain 459 (LB plus 100 μg/mL ampicillin), and B. anthracis Sterne were grown overnight, and aliquots (75 μL) from washed samples of each culture were mixed and spotted onto BHI agar before incubating for 48 h at room temperature. The accumulated growth was resuspended in fresh LB broth, and 150 μL of the suspension was dispensed in a “straight line” on one quadrant of a BHI agar plate supplemented with 5 μg/mL erythromycin and 80 U/mL polymyxin B. After the line was allowed to dry, cells were streaked to isolation and incubated at room temperature for 48 h. Isolated EryR colonies were cultured overnight, again at room temperature, before diluting 1:1000 in fresh BHI broth plus erythromycin and shifting to 37°C, a nonpermissive temperature for plasmid replication. Following this overnight incubation, integrants were isolated using the “straight line” procedure described above, but with incubation at 37°C on BHI agar plus erythromycin. A single EryR colony was grown out, to A600 ca. 0.2, in 25 mL of LB plus glucose (28) at 37°C, with shaking at 100 rpm; cells were harvested by centrifugation at 4°C, resuspended and washed in ice-cold electrotransformation buffer (29), and transferred (400 μL) to a chilled 0.4-cm electrode-gap cuvette. After 10 min incubation on ice, with 5 μL of unmethylated pBKJ223 [isolated from E. coli GM272(pBKJ223)], cells were electroporated at 2.4 kV, 200 Ω and 25 μF. After pulsing, cells were placed back on ice, and 700 μL of ice-cold LB plus glucose was added; the mixture was transferred to a microcentrifuge tube, and cells were allowed to recover for 3 h at 30°C before plating on BHI agar plus 10 μg/mL tetracycline and incubating overnight at 37°C. Several single TetR colonies selected after repeatedly mixing and streaking to isolation were patched onto BHI agar plus erythromycin and BHI agar plus tetracycline and incubated overnight at 37°C. Clones with the desired TetR EryS (loss of pBKJ236) phenotype were validated by colony PCR, repurified by streaking, and patched onto BHI agar with and without tetracycline. Clones cured of TetR (loss of pBKJ223) were again validated by colony PCR to confirm the BAS1445 deletion. ΔbshB, Δcdr, and Δcdr2 mutants were constructed following the same protocol; the ΔcdrΔcdr2 double mutant utilized the Δcdr strain in the initial conjugation step.
An active culture grown to A600 of 0.5 (mid-log) was used to inoculate (1% v/v) 5 mL aliquots of fresh BHI broth containing fosfomycin at incremental concentrations. The wild-type Sterne strain showed robust growth (static conditions in 17 × 100 mm polystyrene tubes) at all fosfomycin concentrations (0-1,000 μg/mL). The ΔbshA mutant grew only in the control (no fosfomycin) culture and at the lowest fosfomycin concentration tested (200 μg/mL), but not at concentrations ≥400 g/mL. The ΔbshB mutant was sensitive to fosfomycin at 800 μg/mL, but not at ≤400 μg/mL.
The codon-optimized BA1558 and BA1557 genes, encoding BaBshA and BaBshB, respectively, were synthesized by GenScript (Piscataway, NJ) and subcloned into pET28a(+) (Novagen). Both genes were expressed in E. coli BL21(DE3) cells using auto-induction media at 37°C (30); BshA was expressed with an N-terminal His-tag (see “Results”), while BshB had a C-terminal His-tag (-LEH6). All steps of both purifications were carried out at 4°C. Harvested cells were disrupted using an Avestin EmulsiFlex-C5 homogenizer, and nucleic acids were removed by addition of 2% streptomycin sulfate and centrifugation (27,000g for 20 min). For BshA, the clarified extract was loaded onto a 25-mL Ni Sepharose High Performance column (GE Healthcare, Piscataway, NJ) in 50 mM sodium phosphate, pH 8.0, containing 0.3 M NaCl and 20 mM imidazole; the protein was eluted with a stepwise increase of imidazole, to 250 mM. Peak fractions were pooled and loaded onto a Q-Sepharose HP column after dialysis against 20 mM HEPES, pH 7.5, containing 50 mM NaCl. The protein was eluted with a linear 375-mL gradient of 0.05 → 1 M NaCl, and the BaBshA pool was dialyzed against the Q-Sepharose loading buffer before concentrating and storing in aliquots at -80°C. For BaBshB, the immobilized metal-ion affinity chromatography column was first converted to the Zn2+ form (31); the clarified extract was loaded onto the 25-mL Zn Sepharose column in 50 mM sodium phosphate, pH 8.0, containing 0.3 M NaCl and 20 mM imidazole. The protein was eluted with a stepwise increase of imidazole, to 250 mM. Peak fractions were pooled and loaded onto a Q-Sepharose HP column after dialysis against 25 mM potassium phosphate, pH 7.0. The protein was eluted with a linear 375-mL gradient of 0 → 1 M NaCl, and the purified BaBshB protein was dialysed against 25 mM Tris-HCl, pH 8.0, containing 50 mM NaCl, concentrated by ultrafiltration, and brought to a final concentration of 20% (v/v) glycerol before being frozen in aliquots at -80°C.
The cysteinyl-tRNA synthetase gene (cysS, BA0089) was amplified from genomic B. anthracis DNA using primers designed with NcoI and XhoI linkers that allowed cloning into pET28a. The recombinant C-terminal His-tagged CysS protein was expressed in E. coli BL21(DE3); cultures grown to mid-log phase in LB plus kanamycin at 37°C were induced with lactose for 3 h. Harvested cells were resuspended in 1.5 volumes of degassed 50 mM Tris-HCl, pH 8.0, containing 0.5 M NaCl, and were lysed with a French press. The extract was clarified by centrifugation and applied to a Ni2+-charged metal-ion affinity column (GE Healthcare) equilibrated with the lysis buffer; the column was washed with this buffer containing 20 mM imidazole before the recombinant CysS protein was eluted with 100 mM imidazole. Pooled fractions were diluted with 5 volumes of lysis buffer and loaded onto a HiTrap Q XL column (GE Healthcare); the column was washed with the lysis buffer before eluting the protein with a linear 50-mL gradient of 0 → 1 M NaCl. The cysteine adenylation activity of B. anthracis CysS was assayed using the ATP-32PPi exchange reaction in the presence of 50 μM Cys (32).
GlcNAc-malate required for kinetic analysis of BaBshB was produced on a preparative scale by incubating 0.5 mg BaBshA at 37°C with 50 μmol each of UDP-GlcNAc and L-malate (substrates purchased from Sigma) in 5 mL of 25 mM HEPES, pH 7.5, containing 100 mM NaCl, 10 mM MgCl2, and 1 mM 2-mercaptoethanol. The reaction was sampled at indicated times; samples were processed as described by Newton et al. (15) for identification of the MshA-catalyzed reaction products (using unlabeled substrates) and were analyzed by HPLC with A260 detection for UDP-GlcNAc and UDP. The conversion of UDP-GlcNAc → UDP was complete in 2 h. The GlcNAc-malate product (50 μmol, ca. 15 mg) was processed for preparative HPLC purification by 1) adding 1 volume of acetonitrile to the reaction mixture and incubating at 60°C for 15 min, 2) clarifying by centrifugation, and 3) adjusting the supernatant pH→3 with trifluoroacetic acid after reducing the volume to 2 mL in a SpeedVac. A Vydac 218TP1022 (1 × 25 cm) reversed-phase preparative column was used with a linear gradient from 0→20% methanol in 0.1% trifluoroacetic acid, over 40 min (flow rate of 5 mL/min), with online detection by mass spectrometry. Fractions containing GlcNAc-malate (ESP-, m/z 336) were combined, and the separation was repeated to obtain the purified product, which was adjusted to pH 6 with NaOH and lyophilized.
Synthetic GlcNAc-malate was maintained at -70°C in H2O at a concentration of 80 mM and contained ca. 1 equiv sodium trifluoroacetate from the HPLC solvent. The sodium trifluoroacetate content was determined in the quantitative pH titration described above. In addition, the disodium GlcNAc-malate content determined by the BaBshB—Accu-Tag method (see below) accounts for ca. 50% of the total mass of the lyophilized material. 1H- and gCOSY NMR spectra were recorded on a Jeol ECA500 at 500 MHz and were referenced to residual solvent (HDO) at 4.8 ppm. The 1H NMR of GlcNAc-malate in D2O (trifluoroacetate is not observable) confirmed the absence of other organic impurities. 13C spectra were collected on a Varian X500 (125 MHz) equipped with an XSens cold probe. All spectra were collected at 23°C. High resolution mass spectrometry analysis was provided by the UCSD Chemistry and Biochemistry Mass Spectrometry Facility using a Thermo Scientific LTQ Orbitrap XL mass spectrometer operating in positive ion mode ESP ionization.
The kinetic properties of BaBshA were first analyzed from initial rates of UDP-GlcNAc → UDP conversion, using the A260-based HPLC assay described above (15) for monitoring GlcNAc-malate synthesis. Initial rates were determined over a range of UDP-GlcNAc concentrations (0→25 mM) at a fixed concentration of 2 mM L-malate (ca. 15 × Km, see “Results”); for each individual assay at 37°C [in 25 mM HEPES, pH 7.5, containing 100 mM NaCl, 10 mM MgCl2, and 1 mM 2-mercaptoethanol], the reaction was sampled four times in order to establish the observed rate of UDP formation. Similarly, initial rates were measured over a series of L-malate concentrations (0→3.2 mM) at a fixed concentration of 3 mM UDP-GlcNAc (ca. 5 × Km). Results representing the mean of triplicate determinations were fit to the Michaelis-Menten equation [equation 1 (33)] where Vmax is the apparent maximal velocity, A is the concentration of the varied substrate, and KA is the Michaelis constant (apparent) for substrate A.
Data described with this single variable equation were analyzed using KaleidaGraph.
In order to analyze the steady-state kinetic mechanism with several fixed concentrations of UDP-GlcNAc and L-malate, respectively, the continuous spectrophotometric assay for glycosyltransferases (34) was adapted, using a 96-well format with a POLARstar OPTIMA (BMG LABTECH) microplate reader. For each assay at 37°C, 30 ng of BaBshA was incubated in 25 mM HEPES, pH 7.5, containing 50 mM KCl, 10 mM MgCl2, 7.5 U pyruvate kinase and 15 U lactate dehydrogenase (both enzymes purchased from Sigma), 0.8 mM phosphoenolpyruvate, and 0.15 mM NADH in a final volume of 200 μL. The concentration of L-malate was varied over the range 0.08-1.2 mM at fixed concentrations of 0.1-1.5 mM UDP-GlcNAc. Reactions were initiated by addition of enzyme and were performed in duplicate. These results (for varying [L-malate]) were fit to the reciprocal form of the velocity equation for a rapid equilibrium random system [equation 2 (33)], using GraFit version 5.
where Don = UDP-GlcNAc and Acc = L-malate. α is a factor representing the effect each substrate has on the binding of the second substrate. Controls showed that the omission of 2-mercaptoethanol and substitution of 50 mM KCl for 100 mM NaCl in the coupled assay had no effect on initial velocity. In addition, any competitive UDP-GlcNAc hydrolysis reaction (measured in the absence of L-malate) was negligible.
The kinetic parameters for BaBshB were determined from initial rates of GlcNAc-malate → GlcN-malate conversion, using an HPLC assay for the AccQ-Tag derivative of the GlcN-malate product, as monitored by fluorescence (13, 31, 35). Initial rates were determined over a series of GlcNAc-malate concentrations (0→2 mM); for each assay at 37°C [in 50 mM HEPES, pH 7.5, containing 50 mM NaCl and 1 equiv of sodium trifluoroacetate per mol GlcNAc-malate], the reaction was sampled at intervals in order to establish the rate of GlcN-malate formation. Results representing the mean of triplicate determinations were fit to equation 1 as described above. Bacillithiol S-conjugate amidase activity was measured using the methylbimane derivative of BSH (BSmB) purified as described (10). The reaction was sampled at intervals; aliquots were processed for analysis, using an adaptation of the fluorescence-based HPLC assay for CySmB (+ GlcN-malate) product formation described previously (23). The initial rate of CySmB formation was determined by sampling the reaction four times in the presence of 320 μM BSmB.
After purification BaBshA was buffer-exchanged into 20 mM HEPES, pH 7.5, containing 150 mM NaCl plus 5 mM dithiothreitol, and concentrated to 16 mg/mL for crystallization. Crystals were obtained in 0.2 M Mg(HCO2-)2 containing 15% PEG3350 with 1 mM each of UDP and L-malate. Crystals were cryoprotected with mother liquor supplemented with 25% glycerol and flash-frozen in liquid nitrogen. BaBshA crystals, which appeared after 4-5 days, diffracted to 3.3 Å. Data sets were collected at SER-CAT beamline 22-BM (Advanced Photon Source, Argonne National Laboratory). Data were indexed and scaled using HKL2000 (36). Molecular replacement was performed using PHASER (37), and model refinement was carried out using the PHENIX software suite (38). The model was adjusted manually using COOT (39) as well as 2Fo-Fc and Fo-Fc electron density maps. The UDP and L-malate model and definition files were generated with PHENIX.
All manipulations of B. anthracis Sterne strains were carried out under Biosafety Level 2 conditions at Wake Forest University School of Medicine and at University of California, San Diego. The detailed methods for culture growth, sample preparation, and analysis, including the determination of thiol redox status, are given in Nicely et al. (2) and Newton et al. (10).
Sequence analyses were performed with the NCBI and PDB databases using BLASTP, and the resources of the Carbohydrate-Active enZymes database (18) were also utilized in this work. Structure superpositions and sequence alignments were performed using DALILITE (40), PyMOL (41) , CLUSTALW (42), and ESPript (43), respectively. Figures of protein molecules and residues were prepared using PyMOL.
Given the structural similarity between bacillithiol and mycothiol (Figure 1) and the well-defined pathway for MSH biosynthesis in Mycobacterium tuberculosis and other Actinobacteria (44, 45), three distinct enzymatic steps have been considered in the pathway for BSH biosynthesis (Figure 2). The first reaction produces GlcNAc-malate from UDP-GlcNAc and L-malate, as indicated:
We have demonstrated (13) that the B. subtilis YpjH gene product catalyzes the synthesis of GlcNAc-malate as given in reaction (1). In B. anthracis the BA1558 locus is the ortholog of B. subtilis ypjH; as shown in the preparative scale assay (Figure S1), ORF BA1558 catalyzes the production of UDP and GlcNAc-malate when incubated with UDP-GlcNAc and L-malate. Isolation of the GlcNAc-malate product from this preparative scale reaction allowed its confirmation by 1H and 13C NMR (Figure S2) and high resolution mass spectrometry. The proton NMR spectra (Figure S2A) indicated that the GlcNAc-malate was pure (see “Experimental Procedures”) and contained the L-malate and D-glucosamine resonances found in bacillithiol (10). The glucosamine N-acetyl group is validated by the acetyl protons (singlet, 2.04 ppm) and the carbon resonances at 174.3 and 21.9 ppm (Figure S2B). The carbon 2 proton (3.9 ppm) of glucosamine (nitrogen bonded carbon) is identified in the gCOSY spectra (Figure S2A) by correlation to the anomeric proton. The anomeric proton (4.89 ppm) coupling constant (3.8 Hz) indicated this was the α-anomer of the malyl glycoside as predicted by the structure of bacillithiol. The high-resolution mass spectra of this compound gave a mass of 360.0902 for the sodium ion of GlcNAc-malate (360.0901, δ 0.3 ppm) corresponding to a formula of C12H19NO10Na, consistent with the product of BaBshA. Confirmation of the novel D-glucosaminyl L-(α)-malate moiety validates the structure of bacillithiol from Deinococcus radiodurans (10) and shows that it is identical to bacillithiol produced in B. anthracis. ORF BA1558 is the BshA glycosyltransferase in B. anthracis and is therefore designated BaBshA (Figure 2).
When the structure of the M. tuberculosis GlcNAc-Ins deacetylase MshB [PDB entry 1Q74 (22)] was used to query the Protein Data Bank with DALILITE, the highest score (Z = 21.6) was observed with the Zn2+-binding BcZBP protein [PDB entry 2IXD; rmsd = 2.3 Å, 24% identity (19)] from B. cereus. This deacetylase is 97% identical in sequence to the B. anthracis ortholog, ORF BA1557. We have recently demonstrated (13) that recombinant ORF BA1557 catalyzes the efficient deacetylation of GlcNAc-malate → GlcN-malate, as given in reaction (2):
The preparative scale reaction, using GlcNAc-malate prepared by enzymatic synthesis with BaBshA and purified by HPLC (see “Experimental Procedures”), allowed isolation of the GlcN-malate product, and deacetylation of the substrate was confirmed by ESP mass spectrometry. ORF BA1557 is the BshB GlcNAc-malate deacetylase in B. anthracis and is therefore designated BaBshB. We have now identified the enzymatic activities, in vitro, for the BaBshA glycosyltransferase and the BaBshB deacetylase. Thus, from UDP-GlcNAc and L-malate precursors, these enzymes combine to produce GlcN-malate, UDP, and acetate; GlcN-malate has also been identified by HPLC analysis of cell-free extracts from B. subtilis (13), confirming this aspect of the BSH biosynthetic pathway as presented in Figure 2 for B. anthracis. While we have not attempted to identify GlcNAc-malate in vivo, these combined results provide very strong evidence that the biosynthetic scheme presented in Figure 2 for BshA and BshB represents the intracellular process in B. anthracis.
These functional assignments, as confirmed experimentally, leave the enzyme(s) responsible for the proposed ATP-dependent Cys:GlcN-malate ligase reaction (Figure 2) unidentified. The mshC gene product in M. tuberculosis is a second cysteinyl-tRNA synthetase isoform that differs from the functional CysS enzyme primarily through the absence of a 76-residue C-terminal extension that is responsible for recognition and specific binding of the tRNACys anticodon (46, 47); the Cys:GlcN-Ins ligase catalyzes the ATP-dependent formation of the amide linkage in the ultimate MSH product (+ AMP + PPi). BLASTP analysis of the B. anthracis genome, however, fails to identify any second CysS isoform, and the ORF BA0089 (annotated as cysteinyl-tRNA synthetase) protein as expressed is active in the standard tRNA synthetase assay. Although deletion of the B. subtilis yllA locus eliminates BSH production with a concomitant increase in intracellular GlcN-malate (the BshB product), the recombinant B. subtilis YllA and S. aureus CysS proteins both fail to catalyze ATP-dependent production of BSH from Cys and GlcN-malate in an in vitro assay (13).
Passalacqua et al. (48) have recently demonstrated that the seven genes corresponding to BA1554-BA1560 (Ames strain) are transcribed as a single mRNA unit in B. anthracis Sterne (BAS1441-BAS1447); this experimental result is consistent with results obtained earlier with the operon prediction algorithm developed by Bergman et al. (49). Given the involvement of both BSH and CoASH in the B. anthracis thiol redox buffer (2, 10), we note that the BA1562-BA1564 loci (panBCD) encoding three key enzymes in the synthesis of the CoASH precursor pantothenate (50) are separated by a single ORF from the bsh operon and have been demonstrated to be cotranscribed together with BA1565 in a second tetracistronic operon (48). In the event that candidate bshC loci might be clustered within the bsh operon, we analyzed the seven protein products for structural homologs of known function, using the PDB and NCBI databases. Table 1 summarizes details of the structural and functional annotations for each of the ORFs in the bsh and pan operons. Although there may be a functional linkage between BSH production and methylglyoxal synthase [ORF BA1556 (51)] activity (see “Discussion”), we find no evidence for candidate bshC loci in either bsh or pan operons in B. anthracis.
In contrast to the soil bacterium B. subtilis, which lacks the CoASH thiol redox buffer (9), the redundancy provided by the CoASH and BSH redox buffers in B. anthracis and S. aureus may provide these two pathogens with some advantage in virulence and/or survival, given the hostile environment posed by host defense mechanisms (52, 53). The results described above predict that individual deletions of the bshA and bshB genes in B. anthracis Sterne (loci BAS1445 and BAS1444, respectively) will disrupt the BSH biosynthetic pathway at the points of GlcNAc-malate and GlcN-malate precursor production, respectively. The two mutants were prepared using the markerless deletion method developed by Janes and Stibitz (27) and were analyzed for low-molecular-weight thiol content in parallel with wild-type B. anthracis Sterne and deletion mutants lacking the cdr and cdr2 genes that encode BaCoADR (4) and the chimeric CoADR-RHD protein (9). The major conclusions from the analytical data presented in Table 2 are as follows: 1) deletion of the bshA locus eliminates BSH production, 2) deletion of the bshB locus does not eliminate BSH production but reduces it by 30% relative to wild-type, 3) BSH redox status (BSH/BSSB) strongly favors the reduced thiol in vegetative B. anthracis cells, and this is not affected by deletions of the cdr and/or cdr2 loci, 4) the value for CoASH is higher (1.9±0.4 μmol/g dry residual weight) than that reported previously [0.87±0.11 μmol/g (2)]; this is attributed to the different wild-type strains used in the two analyses, 5) Cys is a major low-molecular weight thiol, consistent with the recent report of Pöther et al. (12), for S. aureus, and 6) thiosulfate (S=SO32-), a principal source of sulfur for aerobic bacteria (54) and donor substrate for the sulfurtransferase rhodanese (55), is a major inorganic sulfur compound in B. anthracis.
In B. subtilis the primary fosfomycin resistance determinant FosB functions as a thiol-dependent S-transferase that produces a thioether conjugate from the thiol and the antibiotic; until very recently, however, the enzyme had only been demonstrated to work quite poorly in vitro (56), with Cys as thiol substrate. Using a zone of inhibition assay we have recently demonstrated (13) that a B. subtilis mutant deficient in BSH is as sensitive to fosfomycin as a fosB mutant, leading to the conclusion that FosB is a BSH S-transferase. We tested the sensitivity of wild-type and ΔbshA mutant strains of B. anthracis to fosfomycin in liquid culture. While the wild-type strain grew well at 1,000 μg/mL fosfomycin, the bshA mutant did not grow at fosfomycin concentrations ≥400 g/mL. B. anthracis ORF BA4109 is annotated as the FosB protein, and we propose that this enzyme requires BSH as a cofactor for conjugation in the detoxification of fosfomycin (Scheme 1).
Our results with the ΔbshB mutant clearly indicate that other protein(s) capable of recognizing and deacetylating GlcNAc-malate are expressed in B. anthracis. Fadouloglou et al., in their earlier structural analysis of BcZBP (19), had identified ORF Bc3461 as a BcZBP paralog; more recently, this analysis was extended (24) with the identification of the BA630 (BaBshB), BA425 (ORF BA3524), and BA758 (ORF BA3888) orthologs in the B. anthracis genome (Figure 3). ORF Bc3461 has now been expressed and characterized (20); when assayed in parallel with BcZBP, the enzyme shows a preference for (GlcNAc)2 and (GlcNAc)3 as deacetylase substrates. kcat/Km(GlcNAc)2 = 5.6 × 106 M-1s-1, or ca. 20% that of BcZBP. While its biological role is unclear, ORF Bc3461 has potent deacetylase activity. Figure 3 gives a structure-based sequence alignment for BcZBP, BaBshB, ORF BA3888, ORF BA3524, and ORF Bc3461. While BA3888 and BA3524 have limited sequence similarity to BcZBP (26-28% identity), the latter protein is 95% identical in sequence to Bc3461; all three conserve the three Zn2+ protein ligands identified in BcZBP as well as the proposed active-site base and charge-relay dyad (His110/Asp112). These conserved features identify these as Zn2+-dependent hydrolases3; Figure 3 also identifies a series of conserved insertions and deletions that distinguish the true BshB deacetylases BcZBP and BaBshB from the three paralogs. These include a 10-residue insert I1 (relative to BcZBP) after Ser45, two small inserts (1-2 residues) following BcZBP Lys82 and Asp108, and two deletions D1 and D2 of 4-6 residues following Ile129 and Phe205, respectively. The I1 insert falls within the hypermobile BcZBP active-site loop L46, and the D1 deletion falls within the L135 loop identified in the molecular dynamics study (24). Fadouloglou et al. have implicated these loops (plus loop L185) in active-site accessibility and substrate specificity. Further consideration of proposed ORF BA3524 and BA3888 catalytic functions is presented in a following section.
Using a coupled assay, Vetting et al. (16) reported that the dimeric CgMshA catalyzes the formation of GlcNAc-Ins-P via a sequential (ternary complex) kinetic mechanism in which the donor substrate UDP-GlcNAc almost certainly binds first. The steady-state kinetic parameters determined at pH 7.8, 25°C, yield Km(UDP-GlcNAc) = 0.21 mM, Km(Ins-1-P) = 0.24 mM, and kcat = 12.5 s-1. Structural analyses of the apoenzyme and the binary UDP complex demonstrate that nucleotide binding leads to a major rotation of the C-terminal domain relative to the N-terminal domain. kcat/Km(UDP-GlcNAc) is 6 × 104 M-1s-1, and a superposition of the CgMshA apoenzyme and UDP complex structures gives an rmsd of ca. 11 Å, underscoring the magnitude of the conformational change in the UDP complex.
Using an HPLC assay for the UDP product, we demonstrated that the BaBshA reaction was saturable with respect to both donor and acceptor substrates (Figure 4). At pH 7.5 and 37°C, in the presence of 100 mM NaCl, the apparent kinetic parameters (at 2 mM L-malate) for UDP-GlcNAc are Km = 0.37 mM, kcat = 28 s-1, and kcat/Km = 7.6 × 104 M-1s-1. D-Malate, other α-hydroxyacids such as glycolate and D-lactate, and Ins-1-P gave no more than 0.5% of the activity with L-malate when assayed at 0.1 mM (Table S2). The coupled spectrophotometric assay was then employed in an analysis of the steady-state kinetic mechanism for BaBshA. As given in Figure S3, the family of reciprocal plots obtained with varying [L-malate] at different fixed [UDP-GlcNAc] intersects below the x-axis. This suggests that 1) the kinetic mechanism conforms to a rapid equilibrium random bireactant system (33), as contrasted with an ordered system, and 2) the binding of either substrate to free enzyme increases Kapp for the second substrate. From the primary plot and secondary slope and intercept replots, we have determined the kinetic parameters: Km(UDP-GlcNAc) = 0.22 mM, Km(L-malate) = 58 μM, kcat = 15 s-1, and α = 2.6. kcat/Km(UDP-GlcNAc) = 6.8 × 104 M-1s-1, similar to that reported for CgMshA at 25°C.
Figure 5 gives a structure-based sequence alignment for CgMshA and the three known functional BshA enzymes. An initial description of the BaBshA donor substrate binding environment, based on an overlay of the apoenzyme structure with those of the UDP—2-deoxy-2-fluoroglucose complex of the GT4 glucosyltransferase WaaG and the PimA—GDP-mannose complex, identified BaBshA His120, Lys211, Glu282, and Glu290 within the active site (14). Figure 5 demonstrates that His120 is conserved in the functional BshA enzymes; the equivalent of this His (His118) is essential for PimA activity (57). Lys211, interacting with the distal phosphate oxygens of UDP, and the two Glu residues, which recognize the sugar 4-OH and the ribose-2′-OH and 3′-OH, respectively, in the PimA complex (57), are conserved in all four sequences. Guerin et al. demonstrated that other residues within the PimA Glu274-Ile278 signature motif, notably Ser275, Phe276, and Ile278, provide main chain hydrogen bonds to the bound mannose sugar moiety, and this motif (BaBshA Glu282-Leu286, with Leu replacing Ile) is conserved in the BshA and MshA enzymes. Vetting et al. (16) have identified four CgMshA residues (Lys78, Tyr110, Thr134, and Arg154) that coordinate the phosphate of bound Ins-1-P in the ternary complex. As represented in Figure 5, CgMshA Tyr110, Thr134, and Arg154 are conserved in all three known functional BshAs, with the single exception of a conservative Arg→Lys substitution in SaBshA; CgMshA Lys78 is replaced by Val in the sequence alignment. When the BaBshA apoenzyme is used to query the PDB database with DALILITE, the closed UDP (and UDP + Ins-1-P complex) structures for CgMshA give the highest Z-scores [ca. 37; rmsd = 2.4-2.5 Å (respective A-chains)]. Without adjustments, the superposition represented in Figure 6 gives a clear view of the donor substrate binding environment in BaBshA, as described above. In addition, the overlay demonstrates that CgMshA-UDP His133 is the structural equivalent of BaBshA His120 and identifies BaBshA Ile204 (main chain), Asn206, Asp263, and Glu282 as likely recognition elements for the UDP pyrophosphate and for the ribose-2′-OH and uracil oxygens, respectively. There is no evidence for any conformational change on binding of UDP to BaBshA, based on the model deduced from this overlay.
The kinetic parameters for BaBshB were determined using an HPLC assay for the AccQ-Tag derivative of the GlcN-malate product, as monitored by fluorescence (13, 31); Figure 7 gives the resulting initial velocity plot and the fit to the Michaelis-Menten equation. At pH 7.5 and 37°C, in the presence of 50 mM NaCl, the kinetic parameters are Km(GlcNAc-malate) = 0.16 mM, kcat = 42 s-1, and kcat/Km(GlcNAc-malate) = 2.6 × 10 M-1s-1. We also analyzed the deacetylase activity of BaBshB with GlcNAc, at a fixed concentration of 0.15 mM. The observed rate of GlcN production is ca. 10-5 that measured with GlcNAc-malate at the same concentration; the L-malate moiety of GlcNAc-malate is therefore a major determinant of substrate recognition and/or transition state stabilization with BaBshB. When BaBshB was assayed for amidase activity with BSmB (→ CySmB + GlcN-malate), the observed rate at 0.15 mM substrate is only ca. 2 × 10-5 that for GlcNAc-malate deacetylation.
In order to test the model for the enzyme-UDP complex described above, and in an attempt to identify the binding determinants for L-malate, we undertook a crystallographic study with the N-terminal His-tagged BaBshA protein. This His-tag (MGSH6SSGLVPRGSHMASMTGGQQMGRGS-) differs from that for the ORF BA1558 construct (MGSSH6-) analyzed by Ruane et al. (14), which crystallized in space group P21; there were 12 monomers organized into three tetramers within that asymmetric unit. BaBshA crystals belonging to space group P41 were obtained in 0.2 M Mg(HCO2-)2 containing 15% PEG3350 with 1 mM each of UDP and L-malate. The structure determined by molecular replacement using the ORF BA1558 apoenzyme coordinates has eight monomers arranged as a dimer of tetramers within the asymmetric unit (Figure 8); gel filtration analysis of the apoenzyme, in the presence of dithiothreitol, suggests that the biological unit for BaBshA is the dimer. Refinement at 3.3 Å resolution led to a final model with reasonable statistics (Table 3). The final refined structure is missing the residues of the N-terminal His-tag and Met1, as well as residues 11-13, 42-47, and 60-63; the short internal segments correspond closely to three of the four exposed regions in the final ORF BA1558 model (14). Within the BaBshA asymmetric unit, the final model includes two UDP, two malate, and one Mg2+ per tetramer; Vetting et al. (16) reported that Ins-1-P soaks of CgMshA-UDP crystals resulted in a complex with two UDP, two Mg2+, and only one Ins-1-P per dimeric asymmetric unit. The dimeric (substrate-bound) unit of the BaBshA-UDP-malate complex gives an overall Cα rmsd of 1.1 Å for 371 atoms when compared with the corresponding dimeric unit of the ORF BA1558 apoenzyme (Figure 9). In addition, the two subunits per tetramer that are complexed with UDP and malate are covalently linked via an intersubunit Cys241-Cys241′ disulfide (Figure 8); this disulfide is not observed with the complementary unbound subunits within that tetramer, which appear to reflect crystal packing. Focusing on one of the subunits containing bound UDP and malate, Figure 10 gives a final composite omit Fo-Fc map in the vicinity of the substrates, together with the refined model. Considering the 3.3 Å resolution of the BaBshA complex, we have also worked with the possibility of modeling glycerol in two conformations as an alternative to malate. Two conformations of glycerol could be modeled into the density; however, after refinement, additional positive density appears. Upon placing malate in the electron density, no additional Fo-Fc density appears.
Bacillithiol is only the second malyl glycoside reported in the literature to date (10, 58), and the interactions and specificity determinants for L-malate recognition are of significant interest, particularly given the recent report that L-malate is a second preferred carbon source in B. subtilis (59), strongly contributing to repression of the uptake of alternate substrates. An overlay of the CgMshA—UDP-Ins-1-P and BaBshA-UDP-malate complexes (not shown) demonstrates that the positioning and geometry of the Ins-1-P 3-OH and malate 2-OH, relative to the distal phosphates of the respective UDPs, are similar. Neither malate carboxylate is close to the position of the Ins-1-P phosphate, however. Although this interpretation is limited by the 3.3 Å resolution of the BaBshA complex, it is consistent with a chemical mechanism [Scheme 2 (16, 60)] involving nucleophilic attack by the malate 2-OH on the anomeric carbon of UDP-GlcNAc.
The superposition of the two complexes also demonstrates that two residues in the CgMshA phosphate coordination site, Tyr110 and Thr134 (see Figure 5), are structurally conserved in BaBshA. BshA Val64 does occupy the same approximate position as MshA Lys78, but the respective Arg138 and Arg154 residues are not equivalent. Figure 11 gives a stereo representation of the BaBshA-UDP-malate active site, focusing on important polar interactions with both malate and UDP. In particular, this analysis identifies Thr122-OH, Asn206-Nδ2 and -Oδ1, and a distal phosphate oxygen (O1B) of UDP as ligands to the malate C1-carboxylate. The substrate C2-OH interacts with Gly15-N and with both O1B and O2B oxygens of the uridine-pyrophosphate, and Ser16-O stabilizes the C4-carboxylate. All four BaBshA residues are conserved in the other known functional BshA enzymes (Figure 5). It has previously been proposed (61) that an active-site His functions in glycogen phosphorylase and related glycosyltransferases to stabilize an oxocarbenium transition state (e.g., via interaction with His377-O in glycogen phosphorylase) such as that considered for CgMshA. This is thought to be especially important for those GT enzymes operating with negatively-charged (e.g., malate, Ins-1-P) acceptor substrates (61), and the conservation of His120 in the BaBshA catalytic center appears to be an important factor in this regard.
With regard to UDP binding, the structure of the BaBshA ternary complex confirms several aspects of the model deduced from the overlay described in Figure 6. The very favorable comparison of apoenzyme and ternary complex structures (Figure 9) demonstrates that there is no major conformational change on substrate binding to BaBshA.
In this work we have defined the functions of ORF BA1558, previously annotated as a glycosyltransferase, group 1 family protein, and ORF BA1557, previously annotated as a hypothetical protein, as the N-acetyl-α-D-glucosaminyl L-malate synthase (BaBshA) and deacetylase (BaBshB), respectively. The GlcNAc-malate product of BaBshA has been confirmed by 1H and 13C NMR and by high resolution mass spectrometry; deacetylation by BaBshB gives GlcN-malate, as confirmed by ESP mass spectrometry. The kinetic parameters have been determined for both enzymes. BaBshA is specific for L-malate and does not recognize either D-malate or other α-hydroxyacids as acceptor substrates. BaBshA appears to follow a rapid equilibrium random bireactant mechanism in which the binding of either substrate increases Kapp for the second substrate (33). Vetting et al. (16) reported that CgMshA followed a sequential mechanism in which UDP-GlcNAc was proposed to bind first, i.e., as in a rapid equilibrium ordered bireactant scheme. However, Segel has shown that the family of reciprocal plots (1/ν versus 1/[Acc], at several fixed [Don]) for such an ordered sequential mechanism intersects on the y-axis, in the case where substrate A (e.g., UDP-GlcNAc) binds first (33). An examination of the data (16) indicates, in contrast, that reciprocal plots of 1/ν versus 1/[Ins-1-P] for CgMshA intersect in the second quadrant, very close to the x-axis. We suggest that BaBshA and CgMshA both follow rapid equilibrium random bireactant mechanisms; they differ primarily in the fact that α ~1 for CgMshA. While binding of either substrate to BaBshA has the effect of increasing Kapp for the second substrate (α = 2.6), this interpretation does not apply to CgMshA.
The respective kcat/Km(UDP-GlcNAc) values of 6.8 × 104 M-1s-1 (37°C) and 6 × 104 M-1s-1 (25°C) correspond to free energies of activation [ΔGT‡ (62)] of 11.3 kcal/mol and 10.9 kcal/mol. This quantitative similarity is significant, in that Vetting et al. (16) have reported a large conformational change between the CgMshA apoenzyme and UDP complex. Apoenzyme crystals were prepared with an N-terminal His-tagged protein and were of space group P31, while crystals of the complex were prepared with a C-terminal His-tagged protein and were of space group I422. While a crystallographic argument has been presented against the possibility that the apoenzyme conformation does not represent the functional enzyme, this remains an open question. In particular, searches using the CgMshA apoenzyme with the SSM server yielded no significant structural homologs among the GT-B fold family. There is no major conformational change on UDP binding to BaBshA; the possibility of a conformationally distinct BaBshA—UDP-GlcNAc complex also appears unlikely, as the superposition of the BaBshA apoenzyme with the monomeric PimA—GDP-mannose structure reveals similar conformational states (Z = 33.6, rmsd = 3.0 Å). In view of the Km(UDP-GlcNAc) value of 0.22 mM for BaBshA, the intracellular concentration of ca. 2.4 mM reported for B. megaterium (63) predicts that ca. 8% of the enzyme is present in the apo form in the in vivo steady state. With respect to the BaBshA complex, crystal packing stabilizes a closed tetrameric protein form (two UDP and two malate per tetramer; Figure 8) in which bound UDP is solvent inaccessible. Similarly, Vetting et al. showed that crystals of the dimeric CgMshA-UDP complex, when soaked with high concentrations of UDPGlcNAc, do not exchange bound UDP (16).
Using a transposon site hybridization assay protocol, Day et al. (64) concluded (without the benefit of the functional assignments presented in this work) that the bshA and panC genes, as well as BA1559 (pcnB, Table 1), were individually required for optimal growth, sporulation, or germination in B. anthracis strain ΔAmes. As we have demonstrated that deletion of the bshA locus in B. anthracis Sterne eliminates BSH production, this suggests that BSH is important for one or more of these aspects of the life cycle. In this context, we have recently demonstrated that the B. subtilis bshA mutant sporulates with an efficiency ca. 1% that of the wild-type strain (13). In their transcriptional profiling of the B. anthracis life cycle, in which five distinct temporal waves of gene expression were identified from germination through sporulation, Bergman et al. (65) showed that the bshA and bshB genes encoding the first two enzymes in the BSH biosynthetic pathway are upregulated together in waves II and III; the panBCD genes are upregulated in wave V of the germination→sporulation cycle. panB and panC are also upregulated between 1 and 2 h postinfection within host macrophages (52).
We have described the gene products for all seven loci in the bsh operon, which is upregulated during that part of the life cycle corresponding to early outgrowth → rapid growth (65). Aside from the GlcNAc-malate synthase and deacetylase enzymes, we find no evidence for coding sequences that provide for the enzymatic conversion of GlcN-malate → BSH. In connection with the demonstration that BSH is important for fosfomycin resistance in B. anthracis, the coexpression of the mgsA gene encoding methylglyoxal synthase is of interest. Methylglyoxal is cytotoxic in millimolar concentrations and is also mutagenic (51). Still, many bacteria produce methylglyoxal synthase, with a functional implication as a bypass of glycolysis active under conditions of Pi starvation. In most microorganisms, methylglyoxal is detoxified via GSH-dependent conversion to D-lactate; however, we have demonstrated that GSH is absent in B. anthracis (2). Both methylglyoxal (51) and fosfomycin (66) are potent C3-electrophiles known to react with GSH, and our evidence strongly suggests that BSH is the cofactor for FosB in B. anthracis. We have recently demonstrated, using a zone of inhibition assay, that a B. subtilis bshA mutant has a significantly increased sensitivity to methylglyoxal (13). We conclude that BSH may also be the cofactor for an S-transferase reaction converting methylglyoxal to lactoylbacillithiol [CH3CHOHCO-SCH2CH(NH3+)CO-NGlc-malate]. A possible fate of these BSH conjugates is discussed below.
As indicated previously, BshB is both a structural and functional homolog of M. tuberculosis MshB; when assayed in parallel with synthetic GlcNAc-Ins, however, MshB gives kcat of only 0.49 s-1 (31), with kcat/Km(GlcNAc-Ins) = 1.44 × 103 M-1s-1. Both values are ca. 1% of the respective parameters with BaBshB (acting on GlcNAc-malate as substrate). The difference in BshB and MshB catalytic efficiencies is approximately that required by rapidly dividing cells to maintain stable levels of BSH or MSH, respectively, given the 50-fold difference in generation times for B. anthracis and M. tuberculosis (0.5 and 24 h). We have shown that the B. anthracis ΔbshB mutant produces BSH at ca. 70% of the wild-type level, and this has been attributed to an overlapping substrate specificity for two BaBshB paralogs, ORFs BA3524 and BA3888. In the closely related B. cereus 14579, the ORF Bc3461 ortholog of BA3524 has been demonstrated (20) to catalyze deacetylation of both (GlcNAc)2 and (GlcNAc)3, and we propose that this enzyme and BcZBP similarly have overlapping specificities for GlcNAc-malate deacetylation in B. cereus. A BcZBP-deficient mutant would still be expected to produce significant BSH, through the action of ORF Bc3461.
The substrate specificity of the MshB deacetylase overlaps that of the M. tuberculosis mycothiol S-conjugate amidase (Mca), which functions in the MSH-dependent detoxification of thiol-reactive drugs and metabolites by cleaving the Cys-GlcN amide bond of the S-conjugate to give AcCySR plus GlcN-Ins. The optimal amidase substrate for Mca is MSmB, yielding AcCySmB + GlcN-Ins; kcat/Km = 9 × 104 M-1s-1 at 37°C (23), and MshB catalyzes amide cleavage with this substrate, though with kcat/Km = 460 M-1s-1 (31), less than 1% of the value observed with Mca. The M. tuberculosis Rv1170 mutant deficient in MshB continues to produce MSH at ca. 20% of the wild-type level (67), and this has been attributed to the weak deacetylase activity observed with purified Mca in vitro for GlcNAc-Ins [kcat/Km = 0.3 M-1s-1 (23)], ca. 10-4 that of MshB (see above). In addition to providing for GlcNAc-malate deacetylation in the absence of BshB (see above), ORFs Bc3461, BA3524, and BA3888 may function primarily as paralogs of Mca that catalyze bacillithiol S-conjugate amidase reactions related to xenobiotic detoxification. The S-conjugates of fosfomycin and methylglyoxal described above represent two potential substrates (Scheme 3). The charged GlcN-malate product of the amidase reaction can be recycled for BSH synthesis, and a number of AcCySR products from the analogous amidase reactions with mycothiol S-conjugates are known to be lost from the cell (44, 45). Rukmana et al. (68) have recently reported that the Spx regulon is induced in B. subtilis cells challenged by the antibiotics enduracin and bacitracin, suggesting that disulfide stress is occurring under these conditions, but the Spx regulatory system appears to be independent of BSH (13).
Bacterial growth and proliferation, especially during infection of the host, often involves a hostile environment (52, 53) that challenges the defensive mechanisms of the pathogen (e.g., B. anthracis or S. aureus). Among the physiological mechanisms for maintaining thiol-disulfide redox homeostasis are the important functions of thiols such as GSH (69, 70) and MSH (44, 45). B. anthracis (2), S. aureus (3), and B. megaterium (11) are among a number of Gram-positive bacteria that lack both of these thiols. It is clear from several studies, perhaps most closely associated with the Beckwith laboratory (69, 70), that redundancy in these systems is the rule, particularly in E. coli; mutants lacking either TrxR or glutathione reductase grow normally (70, 71).
Figure 12 gives the present working scheme for cytoplasmic disulfide reducing and oxidizing pathways in our model Gram-positive pathogen, B. anthracis; the same or very similar system applies in S. aureus and in the soil bacterium B. megaterium. The three branches include CoASH/CoADR, Trx/TrxR, and bacillithiol (BSH). The recent work of Pöther et al. (12) indicates that cysteine (CySH) represents a fourth branch in S. aureus, accounting for most protein (mixed) disulfide formation during diamide stress. While a specific bacillithiol disulfide reductase has not been identified, the BSH/BSSB redox status is very reduced (redox ratio of 84 versus 2.8 for CySH/CySSR) in vegetative B. anthracis cells (Table 2); this ratio is unchanged in the Δcdr and Δcdr2 mutants, and neither CoADR nor CoADR-RHD is a reductant of oxidized bacillithiol (BSSB). There is, however, no direct evidence as to whether the B. anthracis Trx/TrxR system might reduce BSSB. In both Arabidopsis (72) and Drosophila (73), for example, the Trx/TrxR system either constitutes a functional backup or substitutes for glutathione reductase. In E. coli, the Grx-SSG reductase activity of AhpC* maintains the reduced GSH pool at essentially wild-type levels in ΔgorΔtrxB strains (70).
We recently demonstrated (8) that the B. anthracis cysK-1 locus encoding cysteine synthase A is part of the tricistronic coaX operon that also encodes the type III pantothenate kinase. The B. anthracis pathway from pantothenate → CoASH (Figure 13), as well as the components of the de novo pantothenate biosynthetic pathway (including the panBCD genes of the pan operon; Table 1), have now been established (2, 74). In this work we describe the first two steps in BSH biosynthesis for B. anthracis and the effects of deleting the respective genes on BSH production. While the identification of the proposed Cys:GlcN-malate ligase, the biological functions of the B. subtilis YllA and ORF BA3524 proteins, and the description of the reductase(s) that maintain the reduced status of BSH in vegetative B. anthracis cells remain as priority goals, the materials necessary for genetic analysis of the relative roles of the CoASH/CoADR and BSH systems in both intracellular thiol-disulfide redox homeostasis and in virulence are presently being exploited.
We gratefully acknowledge Dr. Robert Fahey, who first identified bacillithiol and led the structure determination, for his generous support and for many very helpful discussions. We thank Dr. Philip Hanna for hosting one of us (C.P.) in his laboratory and contributing to the construction of mutant strains, Dr. James La Clair for help with NMR work, and Ms. Kelly Lo and Mr. Austin Dosch for technical assistance.
†This work was supported by National Institutes of Health (NIH) Grants GM-35394 (A.C.), AI-78924 (M.R.R.), and AI-72133 (G.L.N.; Robert C. Fahey, Principal Investigator).
‡Coordinates have been deposited with the Protein Data Bank under the file name 3MBO.
1Abbreviations: CoADR, coenzyme A-disulfide reductase; CoADR-RHD, coenzyme A-disulfide reductase isoform with a C-terminal rhodanese homology domain; BSH, bacillithiol; BshA, N-acetyl-α-D-glucosaminyl L-malate synthase; GlcNAc-malate, N-acetyl-α-D-glucosaminyl L-malate; BshB, N-acetyl-α-D-glucosaminyl L-malate deacetylase; GlcN-malate, α-D-glucosaminyl L-malate; MshA, N-acetyl-α-D-glucosaminyl 1-D-myo-inositol-3-phosphate synthase; Ins, myo-inositol; MSH, mycothiol; GT, glycosyltransferase; BcZBP, Bacillus cereus zinc-binding protein; MshB, N-acetyl-α-D-glucosaminyl 1-D-myo-inositol-3-phosphate deacetylase; BHI, Brain Heart Infusion; LB, Luria-Bertani; EryR, erythromycin-resistant; TetR, tetracycline-resistant; HEPES, N-(2-hydroxyethyl)piperazine-N′-2-ethanesulfonic acid; HPLC, high-performance liquid chromatography; ESP, electrospray; gCOSY, gradient correlation spectroscopy; RSmB, methylbimane derivative of thiol, RSH, from reaction with monobromobimane; Mca, mycothiol S-conjugate amidase; TrxR, thioredoxin reductase; Trx, thioredoxin; Grx, glutaredoxin.
2Intracellular concentrations are calculated from BSH contents (μmol/g dry weight), using a value of 2 μL/mg dry weight for the cell volume (Ref 2).
3Purification of the recombinant BcZBP protein employed Ni-NTA chromatography and resulted in Zn2+-containing BcZBP crystals (Ref 19), as analyzed by both X-ray fluorescence and X-ray diffraction. Similar purification of the MshB protein (Ref 31), however, yielded a preparation with no Zn2+ and 0.82 mol Ni2+ per subunit.
SUPPORTING INFORMATION AVAILABLE
Figures S1 and S2, depicting preparative-scale production and NMR characterization of GlcNA-cmalate, Figure S3, providing reciprocal plots of BaBshA activity, and Tables S1 and S2, providing the primer sequences used in the study and describing the acceptor substrate specificity for BaBshA. This material is available free of charge via the Internet at http://pubs.acs.org.