SbnG, a Citrate Synthase in Staphylococcus aureus

Background: Staphylococcus aureus contains a second, iron-regulated citrate synthase. Results: SbnG is a citrate synthase within the phosphoenolpyruvate/pyruvate superfamily. Conclusion: The structural similarity of the SbnG active site to TCA cycle citrate synthase active sites suggests convergent evolution. Significance: SbnG is defined as a new structural class of citrate synthase. In response to iron deprivation, Staphylococcus aureus produces staphyloferrin B, a citrate-containing siderophore that delivers iron back to the cell. This bacterium also possesses a second citrate synthase, SbnG, that is necessary for supplying citrate to the staphyloferrin B biosynthetic pathway. We present the structure of SbnG bound to the inhibitor calcium and an active site variant in complex with oxaloacetate. The overall fold of SbnG is structurally distinct from TCA cycle citrate synthases yet similar to metal-dependent class II aldolases. Phylogenetic analyses revealed that SbnG forms a separate clade with homologs from other siderophore biosynthetic gene clusters and is representative of a metal-independent subgroup in the phosphoenolpyruvate/pyruvate domain superfamily. A structural superposition of the SbnG active site to TCA cycle citrate synthases and site-directed mutagenesis suggests a case for convergent evolution toward a conserved catalytic mechanism for citrate production.

In response to iron deprivation, Staphylococcus aureus produces staphyloferrin B, a citrate-containing siderophore that delivers iron back to the cell. This bacterium also possesses a second citrate synthase, SbnG, that is necessary for supplying citrate to the staphyloferrin B biosynthetic pathway. We present the structure of SbnG bound to the inhibitor calcium and an active site variant in complex with oxaloacetate. The overall fold of SbnG is structurally distinct from TCA cycle citrate synthases yet similar to metal-dependent class II aldolases. Phylogenetic analyses revealed that SbnG forms a separate clade with homologs from other siderophore biosynthetic gene clusters and is representative of a metal-independent subgroup in the phosphoenolpyruvate/pyruvate domain superfamily. A structural superposition of the SbnG active site to TCA cycle citrate synthases and site-directed mutagenesis suggests a case for convergent evolution toward a conserved catalytic mechanism for citrate production.
Staphylococcus aureus is a Gram-positive bacterium that colonizes a variety of mammalian hosts, specifically skin and mucosal membranes of the nasal passage and respiratory tract. In hospital settings, S. aureus has a substantial negative impact because of nosocomial cases of pneumonia and surgical wound infections (1). S. aureus can infiltrate and colonize many human tissue types, leading to a broad range of diseases from boils and sinusitis to more serious conditions like endocarditis and meningitis. Survival of S. aureus depends on the ability to extract iron from the mammalian host. However, iron is actively sequestered as a form of innate immunity against bacterial infection by host proteins such as transferrin, lactoferrin, haptoglobin, and hemopexin (2). In response, S. aureus has evolved multiple iron transport systems to maintain a supply of iron from various host sources (3). One of the most commonly employed strategies by bacterial pathogens is the use of siderophore iron acquisition systems. Siderophores are low molecular weight, ferric iron chelators that are synthesized and secreted to scavenge iron from the extracellular environment (for a review, see Ref. 4).
S. aureus produces two ␣-hydroxycarboxylate siderophores termed staphyloferrin A (SA) 2 and staphyloferrin B (SB) (5,6). Staphyloferrin A is composed of two citrate molecules linked to D-ornithine, assembled by synthetases encoded in the sfa gene cluster (7,8). Staphyloferrin B is synthesized from L-2,3diaminopropionic acid, citrate, and ␣-ketoglutarate, assembled by synthetases encoded in the sbn gene cluster (9). In contrast to SA, SB is not widespread among staphylococcal spp. and is largely limited to the more pathogenic coagulase-positive staphylococci (10). The biosynthetic gene cluster for SB together with its transport operon (sirABC) has only been identified in S. aureus, the Staphylococcus intermedius group (11), and a select few coagulase-negative staphylococci (e.g. Staphylococcus arlettae and Staphylococcus equorum). Furthermore, the biosynthetic genes for SB are among the most highly upregulated under iron restriction and in human blood and serum (12,13), suggesting an important role for SB in staphylococcal pathogenesis.
Growth of S. aureus under iron-restricted conditions results in a global shift in gene expression that is collectively known as the "iron-sparing response" (12). Iron restriction results in the up-regulation of iron uptake systems, including the staphylo-ferrin siderophore pathways. Conversely, many metabolic pathways with iron-containing enzymes are down-regulated in an effort to conserve the iron pool necessary for survival. A notable down-regulation is observed of TCA enzymes including the citrate synthase, CitZ (12,14). Citrate and ␣-ketoglutarate production are essential for the assembly of SB, leading to an apparent, but recently explained, paradox of how SB production is maintained under iron limitation.
Although TCA cycle activity is essential for the synthesis of SA, SB is freed from this dependence by enzymes encoded in the sbn gene cluster, expressed under iron restriction and during glycolytic growth (15,16). Recently, SbnA and SbnB were shown to form a metabolic pathway that produces L-2,3-diaminopropionic acid and ␣-ketoglutarate from Ophospho-L-serine and L-glutamate (15). Both O-phospho-L-serine and L-glutamate can be derived from glycolysis and L-glutamine degradation, pathways independent of the iron sparing response (15). To alleviate the loss of citrate production, a second citrate synthase is encoded by sbnG (17). Both CitZ and SbnG can contribute citrate to the formation of SB; however, only SbnG is up-regulated upon iron starvation (12)(13)(14)16). The precursors to citrate production, oxaloacetate and acetyl-CoA, are also predicted to be available under iron restriction, through anaplerotic reactions and decarboxylation of pyruvate by the pyruvate dehydrogenase complex (pdhABC), respectively (12,16). SbnG, therefore, provides an alternate route for citrate to enter the biosynthetic pathway for SB, independent of TCA cycle activity.
Homologs of SbnG have also been identified in the biosynthetic gene clusters that produce the siderophores achromobactin and vibrioferrin, suggesting that a unique group of citrate synthases may have evolved for siderophore biosynthesis out of the requirement for citrate to be generated during iron restricted growth. Indeed, SbnG represents a unique citrate synthase, originally annotated as a metal-dependent class II aldolase by sequence analysis; instead, SbnG was reclassified as a metal-independent citrate synthase that utilizes oxaloacetate and acetyl-CoA (17).
To gain insight into the catalytic mechanism, we report the x-ray crystal structure of SbnG and an active site variant bound to oxaloacetate. Although SbnG is homologous to metal-dependent class II aldolases, we demonstrate through phylogenetic analysis that SbnG is the archetype of a separate metalindependent clade with homologs from other siderophore biosynthetic gene clusters. Furthermore, structural and functional evidence is presented to support a model for convergent evolution in the active site architecture of SbnG and TCA cycle citrate synthases.

EXPERIMENTAL PROCEDURES
SbnG Overexpression and Purification-The SbnG coding region was cloned into pET28a as previously described (17). SbnG variants (E46Q, H47A, R72A, H96A, E151Q, and D177A) were produced using a modified whole plasmid PCR technique as previously described (18). SbnG variants were introduced into Escherichia coli BL21(DE3), and the mutations were confirmed by DNA sequencing. His 6 -SbnG was expressed from E. coli BL21(DE3) cells inoculated from 5-ml overnight cultures into 2ϫ YT medium supplemented with 25 g/ml kanamycin and grown at 30°C to an A 600 of ϳ0.8. Cultures were then induced with 0.3 mM isopropyl ␤-D-thiogalactopyranoside and incubated for ϳ16 h at 25°C with shaking at 200 rpm. Cell pellets were collected by centrifugation at 4400 ϫ g for 10 min, resuspended in 50 mM Tris (pH 8.0), 100 mM NaCl, and lysed at 10,000 psi with an EmulsiFlex-C5 homogenizer (Avestin). The supernatant was isolated after centrifugation at 39,000 ϫ g for 45 min, and His 6 -SbnG was purified using a 5-ml HisTrap HP column (GE Healthcare) with a linear imidazole gradient (0 -500 mM). Protein fractions were dialyzed into 50 mM Tris (pH 8.0), 100 mM NaCl at 4°C. The His 6 tag was removed by thrombin digestion at a 1:500 mass ratio (thrombin to protein) and incubated over 16 h at 4°C, followed by dialysis into 50 mM Tris (pH 8.0) for 2 h at 4°C. SbnG was further purified by anion exchange chromatography using a Source 15Q column (GE Healthcare) equilibrated with 50 mM Tris (pH 8.0) and eluted with a NaCl gradient (0 -500 mM). SbnG was dialyzed into 20 mM Tris (pH 8.0), concentrated to 20 mg/ml, and stored at Ϫ80°C. Selenomethionine-incorporated SbnG was produced by methods previously described (19) and purified in a similar manner to native SbnG. All SbnG variants (E46Q, H47A, R72A, H96A, E151Q, and D177A) were also purified in a similar manner to native SbnG.
SbnG Crystallization and Structure Determination-Selenomethionine-labeled SbnG crystals were grown by hanging drop vapor diffusion at room temperature in 1:1 mixtures of 20 mg/ml SbnG in 20 mM Tris (pH 8.0) with reservoir solution of 0.2 M calcium acetate, 0.1 M imidazole (pH 9.0), and 4 -6% (w/v) PEG 8000. Crystals were flash frozen in liquid nitrogen following a brief soak in the same buffer supplemented with 20% (v/v) PEG 400 for cryoprotection. Multiwavelength anomalous diffraction data for selenomethionine protein crystals were collected at the Canadian Light Source on Beamline 08B1. The data were processed and scaled using XDS (20,21). Crystals grew in the space group P312 with one molecule in the asymmetric unit. Phase determination and model building were done using AutoSol (initial figure of merit 0.73) and Autobuild (225 of 259 residues built) programs in Phenix (22). Native SbnG was crystallized under similar conditions with the addition of 5 mM citrate to the reservoir, and diffraction data were collected at the Stanford Synchrotron Radiation Laboratory on Beamline 7-1.
SbnG E151Q crystals were grown by sitting drop at room temperature in a 1:1 mixture of 20 mg/ml SbnG E151Q in 5 mM oxaloacetate, 5 mM coenzyme A, and 20 mM Tris (pH 8.0) with reservoir solution of 5% (v/v) tascimate (pH 7.0), 0.1 M HEPES (pH 7.0), 10% (w/v) PEG monomethyl ether 5000 and ϳ40 mM guanidine hydrochloride. SbnG E151Q crystals were soaked for 10 min in 5 mM oxaloacetate/CoA mixture in the same buffer and supplemented with 20% (v/v) PEG 400 for cryoprotection before data collection with a Rigaku MicroMax 007-HF generator, VariMax HR optics, and Saturn CCD 944ϩ detector. The data sets were processed using Mosflm (23) and scaled using SCALA (24). The native structure was solved by molecular replacement using the selenomethionine-labeled coordinates as the search model in Molrep (25) from the CCP4 program suite (26). Manual building was done using Coot (27), and refinement was performed with Refmac5 (28) using translation libration screw (29) parameters with eight translation liberation screw groups. SbnG E151Q was solved by molecular replacement using native SbnG coordinates as a search model in Phaser-MR (30) from Phenix (22). Manual building was done using Coot (27), and refinement was performed with phenix.refine using translation liberation screw refinement. Data collection and refinement statistics are shown in Table 1.
SEC-MALS-SbnG and SbnG E151Q oligomerization were determined by SEC-MALS. Proteins were individually concentrated to 5 mg/ml in 20 mM Tris (pH 8.0) and injected (100 l) into a HPLC 1260 Infinity LC (Agilent Technologies) attached to a Superdex 200 10/300 column (GE Healthcare) with a flow rate of 0.2 ml/min and column temperature set to room temperature. Data were collected with the miniDAWN TREOS multiangle static light scattering device and Optilab T-rEX refractive index detector (Wyatt Technologies). Analysis was performed using the ASTRA6 program (Wyatt Technologies).
SbnG Structure Superposition and Phylogenetic Analysis-SbnG homologs were identified by sequence alignment analysis using BLASTp (31,32) and by tertiary structure superposition using DaliLite (33). A subset of homologs with characterized functions was selected for sequence alignment and phylogenetic analysis. Homologs of known structure were overlaid with SbnG using SSM Superposition (34) in the program Coot (27). Sequence alignments were produced on the EMBL-EBI website using programs ClustalW2 (35) and Clustal Omega (36). The phylogenetic tree was constructed using PHYML (37) in Seaview (38) using the LG substitution model and 100 bootstrap replicates.
The active sites of SbnG and SbnG E151Q were overlaid with the crystal structures of citrate synthase type I and type II using Superpose in the CCP4 suite (34). A total of three active site residues (15 atoms) from SbnG E151Q (His 47 , His 96 , and Asp 177 ) were selected for alignment with the citrate synthase types I and II.
In Vitro Assessment of Citrate Synthase Activity by SbnG Variants-Citrate synthase activity of wild type SbnG and variants was measured using the reagent DTNB to quantify the generation of CoA. The concentration of resulting TNB dianion was monitored at 412 nm on a Varian Cary 50 UV-visible spectrophotometer using an extinction coefficient of 14150 M Ϫ1 cm Ϫ1 . Each reaction contained 50 mM HEPES (pH 8.0), 300 mM NaCl, 0.5 mM oxaloacetate, 0.2 mM acetyl-CoA, and 10 M protein. Each reaction was incubated for 1 h at 37°C followed by the addition of 0.1 mM DTNB reagent. Individual blanks for each reaction contained all required components minus protein. Oxaloacetate-independent activity was recorded by performing the reactions in the absence of oxaloacetate. All reactions were performed in triplicate.
Bacterial Strain and Growth Conditions-All bacterial strains and plasmids employed in this study can be found in Table 2. Unless otherwise indicated, in vivo experiments were performed using S. aureus Newman. For genetic manipulations, E. coli and S. aureus were routinely cultured in Difco Luria-Bertani broth (BD), and Difco tryptic soy broth (BD), respectively. For selection and maintenance of plasmids, medium was supplemented with 100 g ml Ϫ1 of ampicillin for E. coli or 10 g ml Ϫ1 of chloramphenicol for S. aureus. For bacterial growth under iron restriction, RPMI 1640 medium was prepared as directed by the manufacturer (Invitrogen) in sterile polypropylene vessels using water purified with a Milli-Q water filtration system. The nonmetabolizable iron chelator ethylenediamine-di(O-hydroxyphenylacetic acid) (LGC Standards GmbH) was added to the medium, where indicated.

Construction of SbnG Variant Complementation Vectors-
To assess the role of residues implicated in SbnG enzymatic function in the production of SB within the cell, whole plasmid mutagenesis was employed to generate complementation vectors expressing sbnG mutated for these residues. Using the previously constructed psbnG vector (a pALC2073 derivative expressing wild type sbnG) as a template, complementary mutagenesis primers were used to amplify plasmids bearing SbnG variants E46Q, H47A, R72A, H96A, E151Q, and D177A. Following amplification, methylated template DNA was removed through digestion with DpnI, the enzyme was heatinactivated at 80°C for 5 min, and the resulting reaction mixes were used in the transformation of E. coli DC10B. Plasmids bearing the correct mutation were confirmed through sequencing and introduced into S. aureus Newman ⌬citZ sbnG (H2708) through electroporation. The sbnG variants were constitutively expressed, without induction by the leaky P xyl/tetO promoter of pALC2073 (39).
Assessment of SB Production through Agar Plate Bioassays-Concentrated spent supernatants of the citrate-synthase deficient mutant (⌬citZ sbnG) complemented in trans by plasmids bearing the sbnG variants, as described above, were prepared from triplicate 10-ml cultures of these strains grown in RPMI 1640, a medium conducive only to the production of SB (16). Following a 36-h incubation with shaking at 37°C, growth was assessed, and the culture densities normalized to ϳ2, by A 600 . Bacterial cells were removed by centrifugation, and the resulting supernatants were lyophilized overnight. Lyophils were resuspended in 1-ml of sterile double distilled H 2 O, passed through a 0.2-micron syringe filter, and assessed for their ability to promote growth of wild type S. aureus RN6390, as previously described (7,16). In brief, 10 l of the reconstituted supernatants were applied to sterile paper disks, which were then placed on Tris minimal succinate agar plates (40) seeded with ϳ1 ϫ 10 4 colony-forming units ml Ϫ1 of RN690 as a reporter strain and containing 10 M ethylenediamine-di(O-hydroxy-phenylacetic acid). The plates were incubated at 37°C, and growth radius about the disks was measured after 48 h.
In Vivo Assessment of Citrate Synthase-dependent SB Production by SbnG Variants-To assess the role of residues implicated in SbnG enzymatic function in the citrate-dependent production of SB within the cell, sbnG complementation vectors bearing the mutated residues (pE46Q, pH47A, pR72A, pH96A, pE151Q, and pD177A) were assessed for their ability to promote SB production in a previously constructed citrate-synthase deficient strain (H2708; citZ sbnG) of S. aureus Newman (16). Spent culture supernatants from the complemented citZ sbnG mutant strains grown in RPMI 1640 for 36 h and 10ϫ concentrated were assessed for the presence of SB using agar plate bioassays, as previously described (7,16). In brief, S. aureus wild type RN6390 was seeded into Tris minimal succinate agar plates (40), and the above supernatants were applied to sterile paper disks placed onto these plates. SB-dependent growth promotion was assessed by measuring the growth radius about the disks after 48 h of incubation at 37°C.

RESULTS
Overall Structure of SbnG-The crystal structure of SbnG was solved to a resolution of 1.85 Å with a single subunit in the asymmetric unit. The structure consists of residues 4 -258, excluding residues 121-132 that lack clearly defined electron density and form part of an apparent disordered loop. The overall structure of the SbnG protomer revealed an (␣/␤) 8 barrel fold formed by seven ␣-helices that pack against an interior ␤-barrel with an eighth ␣-helix on the C-terminal end projecting outward (Fig. 1A). Dimers related by a crystallographic 2-fold axis resulted in the domain swapping of the C-terminal ␣-helixes (␣8) of each protomer to complete the (␣/␤) 8 fold. The biological hexamer could be reconstructed through 3-fold crystallographic symmetry (Fig. 1B). Additionally, SbnG (a ϳ28-kDa subunit) exists primarily as a hexamer in solution with a calculated mass of 171 kDa as determined by multiangle light scattering ( Fig. 2A). The dimer interface includes one -stacking interaction between equivalent His 244 residues from each protomer, as well as extensive hydrophobic interactions between residues Ile 24 , Leu 28 , Ile 32 , Leu 234 , Ile 241 , Leu 245 , and Leu 249 . The trimeric interface was comprised of a mixture of electrostatic and hydrophobic contacts, including a H-bond between residues His 47 and Asp 86 and several hydrophobic interactions involving residues Val 76 , Leu 120 , Leu 123 , Leu 124 , Pro 185 , and Trp 186 . SbnG Active Site-Based on homology to class II aldolases, the SbnG active site was hypothesized to be situated within a groove at the opening of the (␣/␤) 8 barrel near the 3-fold interface shared with an adjacent SbnG protomer (Fig. 1C). The putative active site was highly polar and is composed in part of two histidine (His 47 and His 96 ), two glutamate (Glu 46 and Glu 151 ), aspartate (Asp 177 ), and arginine (Arg 72 ) residues. In the center of this site, density for a metal ion was identified residing within an acidic patch on the protein surface (Fig. 1D). Because of the high calcium acetate concentration (0.2 M) from which crystals were formed, the density was modeled as a Ca 2ϩ , a known inhibitor of SbnG (17). The Ca 2ϩ is bound by Glu 151 -O⑀1 (2.4 Å) and Asp 177 -O␦2 (2.2 Å), as well as four ordered water molecules (2.3-2.5 Å) that complete the octahedral coordination sphere (Fig. 1C). The coordinated water molecules participate in a H-bond network involving the side chains of residues Glu 46 , His 47 , Arg 72 , and His 96 based on H-bond distances under 3.0 Å. Because of the hexameric oligomerization of SbnG, an adjacent protomer sits atop the active site groove at the opening of the (␣/␤) 8 barrel. A single serine residue (Ser 119 ) from a neighboring protomer adopts two different conformations and also participates in the Ca 2ϩ hydration sphere H-bond network (Fig. 1C). Ser 119 was situated within a large loop that extends above of the active site, part of which was poorly defined in the electron density map (Fig. 1D). Beside the Ca 2ϩ -binding site and within the same protomer lies an extended channel lined with intermittent patches of positive (Arg 72 , Arg 189 , Arg 216 , Arg 238 , and Arg 243 ) and near neutral electrostatic potential (Fig. 1D).
Crystal Structure of SbnG E151Q Variant Bound to Oxaloacetate-Previously, citrate synthase activity of SbnG was shown to be inhibited by low millimolar concentrations of Mg 2ϩ and Ca 2ϩ and not inhibited by the addition of EDTA (17). These observations were in stark contrast to all characterized class II aldolases, which require Mg 2ϩ for activity. Attempts to FIGURE 1. Structure of SbnG. A, SbnG protomer is shown as a cartoon with helices, loops, and ␤-strands colored brown, blue, and yellow, respectively. The C-terminal ␣-helix (␣8) protrudes outwards from the core (␣/␤) 8 barrel. B, SbnG homohexamer is reconstructed through crystallographic symmetry. The hexamer is constructed from a trimer of dimers mediated by the protruding ␣8 helix that completes the (␣/␤) 8 barrel fold. C, SbnG active site residues are shown as sticks with carbon, oxygen, and nitrogen in light blue, red, and blue, respectively. A single Ca 2ϩ (yellow sphere) is bound by Glu 151 and Asp 177 . Water molecules are shown as small blue spheres, and metal-ligand bonds are indicated by dashed lines. D, a space-filling representation of the SbnG active site. The electrostatic potential is shown on the molecular surface of one protomer with a neighboring protomer shown in cartoon form after the hexamer was reconstructed. The Ca 2ϩ is visible as a yellow sphere in an acidic patch within the proposed active site, which includes the directly coordinated water molecules shown as blue spheres. DECEMBER 5, 2014 • VOLUME 289 • NUMBER 49 cocrystallize SbnG in the presence of either acetyl-CoA or oxaloacetate using crystallization conditions lacking calcium acetate proved unsuccessful. Furthermore, soaking experiments of wild type SbnG crystals failed to produce substrate-containing structures, leading to the hypothesis that calcium sterically inhibits substrate binding. An SbnG variant (E151Q) was created to prevent divalent metal ion inhibition and thus favor substrate binding under buffer conditions suitable for crystallization.

SbnG, a New Fold for Citrate Synthase
SbnG E151Q crystals soaked in the presence of oxaloacetate, and CoA diffracted to a resolution of 2.6 Å. Oxaloacetate was modeled in at ϳ80% occupancy into a patch of difference electron density discovered within the proposed active site (Fig.  3A). Oxaloacetate was oriented such that the two carboxylate groups formed H-bonds to residues Glu 46 , His 47 , Arg 72 , His 96 , and Gln 151 , while exposing the carbonyl group to the solvent. Ser 119 , a residue from an adjacent protomer, formed a H-bond to the carbonyl group of oxaloacetate. The E151Q variant also existed as a hexamer in solution with a mass of 169 kDa as determined by SEC-MALS (Fig. 2B).
The wild type SbnG and the substrate bound E151Q variant overlay with a root mean square deviation (r.m.s.d.) of 0.86 Å over all atoms. The largest main chain differences were displacement of two ␣-helices that contained Glu/Gln 151 and Asp 177 . This displacement likely resulted from the loss of Ca 2ϩ coordination and binding of oxaloacetate. Within the active site, Ca 2ϩ present in the wild type SbnG structure occupied an equivalent position to the oxaloacetate observed in SbnG E151Q (Fig. 3B). The amino acid side chain conformations in the active site were similar with the exception of Asp 177 , in which the side chain is rotated ϳ45°about 2 (Fig. 3B).
Structural Comparison of SbnG to the Phosphoenolpyruvate/ Pyruvate Domain Superfamily-A search for similar structures of SbnG in the Protein Data Bank using the DaliLite server (33) identified several members of the metal-dependent class II aldolase (top seven unique entries with Ͻ2.4 Å r.m.s.d. over Ͼ240 C␣ atoms). Examples are the well characterized 2-dehydro-3-deoxy-galactarate aldolase (DDGA) (41) and macrophomate synthase (MPS) (42). These class II aldolases also form hexamers that are assembled by domain swapped dimers. More distantly related but significant matches in the DaliLite search were many structures of the phosphoenolpyruvate/pyruvate domain superfamily as defined in the SCOP database (43). Of these, the most similar were pyruvate kinase and the ␤-subunit of citrate lyase. A list of selected structural superpositions with supporting statistics is presented in Table 3.
To compare active site architectures among SbnG structural homologs, a total of four superpositions were assembled with DDGA (41), MPS (42), the citrate lyase ␤-subunit (CitE) (44), and malate synthase G (GlcB) (45). The four superpositions with SbnG revealed common active site features. Notably, all these structurally characterized homologs contained an Mg 2ϩ in an equivalent position to the Ca 2ϩ present in wild type SbnG (Fig. 4, A and B). To the best of our knowledge, all other mem-  . Oxaloacetate bound to SbnG E151Q. A, stereo view of the active site of SbnG E151Q variant bound to oxaloacetate. The active site SbnG E151Q is represented as sticks with the active site protomer colored gray and the adjacent protomer colored black. Oxaloacetate (OAA) is colored to match its corresponding protomer active site. Omit difference electron density is shown as gray mesh contoured at 2.0 . B, an overlay of wild type SbnG (blue) and E151Q variant (gray) structures. The Ca 2ϩ from wild type SbnG is shown as a yellow sphere. Oxaloacetate from SbnG E151Q is shown as sticks and is colored to match its respective active site residues.
bers of the superfamily employ either Mg 2ϩ or Mn 2ϩ as a cofactor. Also, residues equivalent to Glu 151 and Asp 177 directly coordinate to a divalent metal ion, and these two residues are conserved in the phosphoenolpyruvate/pyruvate domain superfamily. Glu 46 and Arg 72 of SbnG are also conserved in the phosphoenolpyruvate/pyruvate domain superfamily and have been suggested to play a role in ordering water molecules in the active site or in substrate binding (41, 46 -48). Interestingly, amino acid residues equivalent to SbnG His 47 and His 96 show variation across the homologs.  Active site residues are shown as sticks with oxygen and nitrogen colored as red and blue, respectively. Mg 2ϩ and Ca 2ϩ are depicted as spheres and colored to match their respective active site residues. C, unrooted phylogenetic tree of SbnG and homologs. Bootstrap values are presented as percentages at key branch points. Each protein is labeled by the genus, species, and protein name. The dashed box represents a novel clade with the phosphoenolpyruvate/pyruvate domain superfamily for SbnG-like citrate synthases identified in other siderophore biosynthetic gene clusters. DECEMBER 5, 2014 • VOLUME 289 • NUMBER 49

SbnG, a New Fold for Citrate Synthase
SbnG Represents a New Family of Metal-independent Enzymes within the Class II Aldolase Superfamily-To further classify SbnG within the class II aldolase family, a multiple sequence alignment was constructed and used to derive a phylogenetic tree (Fig. 4C). Included in the analysis were functionally uncharacterized homologs identified in achromobactin and vibrioferrin biosynthesis gene clusters (17), in addition to the structurally established homologs described above. Inspection of the tree revealed that SbnG and homologs from other siderophore biosynthetic pathways group together to form a separate clade (91% bootstrap value) from other class II aldolases. Homologs outside the SbnG clade are all known to require divalent metal for activity.
SbnG Active Site Variants Show Reduced SB Production in a Citrate Synthase-deficient S. aureus Strain-Based on the interaction of Glu 46 , His 47 , Arg 72 , His 96 , and Asp 177 with oxaloacetate in the E151Q crystal structure, variants of these putative active site residues were generated, in addition to E151Q, to evaluate their role in citrate synthase activity. SbnG variants E46Q, H47A, R72A, H96A, and D177A were all recombinantly expressed in E. coli BL21 and successfully purified. Production of CoA from acetyl-CoA by wild type and variant forms of SbnG was monitored using DTNB in the presence and absence of oxaloacetate. SbnG activity is oxaloacetate dependent; however, the activity by this assay was modest (ϳ5 nmol CoA released per mg of SbnG) as observed previously (17). The activity of the variants did not differ substantially from wild type enzyme, likely because of the low wild type activity. We hypothesized that physiologically relevant activity of SbnG requires unidentified factors present in S. aureus but not present in the biochemical assay. Thus, an in vivo assay was developed to assess activity of SbnG and variants.
To circumvent the above issue of poor enzyme performance in vitro, SbnG activity was assayed by measuring SB production from a strain of S. aureus in which both citZ and sbnG are inactivated (16). Expression of wild type SbnG from a plasmid in this deletion strain restores SB production as quantified using a disk diffusion assay (Fig. 5). SB production from S. aureus strains expressing each of the six variants was diminished in comparison to wild type SbnG. Notably, the assay indicates that the D177A variant is unable to produce sufficient citrate for SB production to be detected.

DISCUSSION
SbnG is a citrate synthase that functions to provide citrate for SB biosynthesis (17). Previously, SbnG was hypothesized to be an aldolase based on sequence alignments to characterized homologs DDGA and YfaU. These alignments revealed a conserved pair of amino acids (Glu 151 and Asp 177 in SbnG) that in the homologs coordinate the active site Mg 2ϩ required for catalysis (41,47). Thus, SbnG was suggested to have metaldependent aldolase activity and to possibly participate in SB degradation for iron release (17). Furthermore, a homolog found in the biosynthetic operon for the siderophore achromobactin, AcsB, was speculated to play a role in converting an intermediate into pyruvate and an aldehyde, which would eventually feed into siderophore biosynthesis (49). However, recent work by Cheung et al. (17) has revealed SbnG to exhibit citrate synthase activity in the absence of divalent metals like Ca 2ϩ and Mg 2ϩ .
The crystal structure of SbnG contained electron density for a Ca 2ϩ directly coordinated to Glu 151 and Asp 177 in the active site. Ca 2ϩ was modeled instead of Mg 2ϩ because the observed bond lengths were not consistent with Mg 2ϩ coordination, which are typically ϳ2.1 Å (50). Furthermore, 0.2 M calcium acetate was used in the crystallization condition, a sufficiently high enough concentration for full occupancy of bound Ca 2ϩ . In the case of DDGA, the closest structurally characterized homolog to SbnG, Mg 2ϩ , Co 2ϩ , or Mn 2ϩ can be used as cofactors for catalysis (41,51). To the best of our knowledge, no example exists in the literature where Ca 2ϩ serves as a functional cofactor in class II aldolases. Thus, the SbnG-Ca 2ϩ complex is likely an inhibited form of the enzyme, and the binding of Ca 2ϩ may stabilize crystal packing.
Binding sites for oxaloacetate and acetyl-CoA to SbnG were inferred by analysis of the E151Q variant and by comparison to the homolog malate synthase G, respectively. Malate synthase G catalyzes a similar Claisen condensation reaction as citrate synthases using substrates glyoxylate and acetyl-CoA to produce malate (52). Crystals of the SbnG E151Q variant produced in the absence of Ca 2ϩ were soaked in a substrate solution revealing density for oxaloacetate. The possibility that the binding mode of oxaloacetate is altered by the amino acid substitution of Glu 151 cannot be excluded. Nonetheless, the overlay of the oxaloacetate and Ca 2ϩ -binding sites suggests that Ca 2ϩ may be a competitive inhibitor with this substrate. Density for CoA was not identified in the E151Q variant electron density map. Inspection of the structure does reveal a large positively charged channel that lies next to the oxaloacetate binding site that could serve as a binding site for acetyl-CoA, analogous to that observed in the distantly related homolog malate synthase G (45), although this channel is on the opposite side of the active site. Additionally, a disordered loop (residues 121-132) from an adjacent protomer lies directly above the active site and FIGURE 5. SbnG variants are impaired for citrate-dependent production of SB. Agar plate bioassays were performed using supernatant extracts prepared from a citrate synthase-deficient strain of S. aureus Newman (⌬citZ ⌬sbnG) complemented with vectors, as indicated, expressing wild type SbnG, SbnG variants E46Q, H47A, R72A, H96A, E151Q, and D177A or a blank pALC2073 vector control (vehicle). Strains were grown for 36 h in RPMI 1640, a medium conducive to the production of SB (and not SA), and supernatants were assessed for SB through their ability to promote growth of S. aureus RN6390 seeded in iron-restricted Tris minimal succinate agar plates. Growth radius about paper disks, to which the supernatants were applied, was measured after 48 h and reflects the average of three biological replicates for each strain.
could play a role in catalysis by acting as the lid for the reaction. Unlike in SbnG, this loop is well ordered in all structurally characterized homologs and faces away from the adjacent active site.
SbnG shares a similar overall fold with the phosphoenolpyruvate/pyruvate domain superfamily, and features of the active site architecture across this family of proteins are remarkably conserved despite the differences in metal binding and catalytic function. In addition to the polypeptide backbone, the amino acids at the equivalent positions to Glu 46 , Arg 72 , Glu 151 , and Asp 177 are all conserved across this family. In contrast, His 47 and His 96 of SbnG show variation among homologs. Interestingly, in MPS these two residues are also histidines, and both enzymes have specificity for oxaloacetate as a substrate (42,53). A multiple sequence alignment of SbnG and select class II aldolase homologs (25-70% sequence identity; Tables 3 and 4) revealed two residues in close proximity to the metal binding site that are differently conserved between homologs within the SbnG clade and the class II aldolases (Fig. 6). The first residue is Met 149 , which is a glutamine in all representative class II aldolases (Fig. 6). In the MPS crystal structure, this glutamine residue is within H-bonding distance to the pyruvate product and the active site arginine residue (Fig. 7A). The second residue is Ala 175 , which is a proline residue in all representative class II aldolases (Fig. 7B). These substitutions may account for differences in metal binding under physiological conditions. SbnG was shown to catalyze an identical reaction to TCA cycle citrate synthases (17). However, the fold of both proteins are different as SbnG forms an (␣/␤) 8 barrel, whereas citrate synthase types I and II are ␣-helical. Two histidines and an aspartic acid participate in catalysis by TCA cycle citrate synthase by acting as general acids and general bases during catalysis. The active site of SbnG contains the same configuration of these three residues: His 47 , His 96 , and Asp 177 . The side chains of His 47 , His 96 , and Asp 177 (15 atoms total) of the oxaloacetate bound SbnG E151Q structure were superimposed on the equivalent residues of porcine heart citrate synthase (type I) and of Acetobacter aceti citrate synthase (type II) (54,55). The overlays clearly demonstrate a conserved spatial distribution of the three catalytic residues with r.m.s.d. values of 2.2 and 2.3 Å with FIGURE 6. Multiple sequence alignment of select SbnG homologs from the phosphoenolpyruvate/pyruvate domain superfamily. Sequences are identified by the genus, species, and protein name. Listed above the sequence alignment is the secondary structure for SbnG with the arrows indicating ␤-sheets and zigzags indicating ␣-helices. Grayscale highlighting represents the degree of conservation between sequences. Conserved active site residues are highlighted in red. Additional conserved regions found only in SbnG and homologs from other siderophore biosynthetic gene clusters are highlighted in blue boxes and by a blue arrow.  citrate synthases type I and II, respectively (Fig. 8). In SbnG E151Q and the structures of either citrate synthase types I and II, the catalytic Asp residues are not observed to H-bond to oxaloacetate. Additionally, Arg 72 of SbnG E151Q roughly overlays with arginine residues of the two citrate synthases that also interact with bound substrates. However, the orientation of Arg 72 relative to the TCA cycle citrate synthases are ϳ90°apart.
The conformation of this arginine may determine the orientation of oxaloacetate binding in the active site, which is also rotated by ϳ90° (Fig. 8). The acetyl CoA analog (carboxymethyldethia-CoA) bound to citrate synthase from A. aceti does not overlay with the proposed binding site in SbnG. Rather, a similar ϳ90°rotation is required to align the acetyl CoA with the positively charged groove with respect to the oxaloacetate molecule. The structural similarity between the active sites of SbnG and TCA citrate synthases suggests convergent evolution of a clade of the phosphoenolpyruvate/pyruvate domain superfamily. Expression of SbnG active site variants, including the residues His 47 , His 96 , and Asp 177 , had significant reductions in SB production in a citrate synthase-deficient strain of S. aureus, relative to wild type SbnG. A decrease in citrate synthase activity was observed in citrate synthase type I from porcine heart when the equivalent catalytic residues His 274 , His 320 , and Asp 375 were mutated (56). Reduced production of SB by variants with substitutions at residues Glu 46 , Arg 72 , and Glu 151 suggests that these residues are also required for full catalytic activity of SbnG and is supported by their conservation among other SbnG homologs in the new clade defined for the phosphoenolpyruvate/pyruvate domain superfamily (Figs. 4 and 6).
Based on the structural similarity of the active site with those of TCA cycle citrate synthases (Fig. 8) and the mutagenesis data (Fig. 5), the catalytic mechanism for SbnG is likely to be similar to that of TCA citrate synthases. Analysis of oxaloacetate binding in the SbnG E151Q structure did not readily identify the specific catalytic role of His 47 and His 96 . The position of oxaloacetate in the active site may differ in the wild type structure or in presence of acetyl-CoA. Nonetheless, Asp 177 is proposed to act as a general acid and either His 47 or His 96 as a general base to afford the enol of acetyl-CoA. A condensation reaction between the enol and oxaloacetate could be assisted by either His 47 or His 96 acting as a general acid to yield a citryl-CoA intermediate. Hydrolysis gives the products citrate and CoA.
In summary, we have defined SbnG as a new structural class of citrate synthase from S. aureus. Therefore, we suggest reclassifying SbnG as a type III citrate synthase. Though SbnG is formally part of the metal-dependent class II aldolase family, we have demonstrated that it has lost the requirement for a metal cofactor and, together with homologs identified in other siderophore biosynthetic gene clusters, forms a new metal-independent category of class II aldolases.