SbnI is a free serine kinase that generates O-phospho-l-serine for staphyloferrin B biosynthesis in Staphylococcus aureus

Staphyloferrin B (SB) is an iron-chelating siderophore produced by Staphylococcus aureus in invasive infections. Proteins for SB biosynthesis and export are encoded by the sbnABCDEFGHI gene cluster, in which SbnI, a member of the ParB/Srx superfamily, acts as a heme-dependent transcriptional regulator of the sbn locus. However, no structural or functional information about SbnI is available. Here, a crystal structure of SbnI revealed striking structural similarity to an ADP-dependent free serine kinase, SerK, from the archaea Thermococcus kodakarensis. We found that features of the active sites are conserved, and biochemical assays and 31P NMR and HPLC analyses indicated that SbnI is also a free serine kinase but uses ATP rather than ADP as phosphate donor to generate the SB precursor O-phospho-l-serine (OPS). SbnI consists of two domains, and elevated B-factors in domain II were consistent with the open–close reaction mechanism previously reported for SerK. Mutagenesis of Glu20 and Asp58 in SbnI disclosed that they are required for kinase activity. The only known OPS source in bacteria is through the phosphoserine aminotransferase activity of SerC within the serine biosynthesis pathway, and we demonstrate that an S. aureus serC mutant is a serine auxotroph, consistent with a function in l-serine biosynthesis. However, the serC mutant strain could produce SB when provided l-serine, suggesting that SbnI produces OPS for SB biosynthesis in vivo. These findings indicate that besides transcriptionally regulating the sbn locus, SbnI also has an enzymatic role in the SB biosynthetic pathway.

Staphylococcus aureus is a prominent human pathogen that also asymptomatically colonizes a proportion of the human population (1,2). Though colonization is typically not harmful to the host, S. aureus is frequently associated with minor skin and soft-tissue infections. In more serious cases, it is capable of breaching host innate immune responses to gain access to deep tissues, causing more severe and invasive infections, including endocarditis, osteomyelitis, and necrotizing pneumonia (2,3).
Successful iron uptake from the human host is integral to infection and pathogenesis of S. aureus and most other microbial pathogens (4,5). Iron is reactive and tightly regulated in the human body to restrict toxicity and to limit bioavailability to invading microbial pathogens as a type of innate, nutritional immunity (6). As such, iron in the mammalian host is found either within heme and bound to hemeproteins, intracellularly bound to proteins as free ions or iron-sulfur clusters, stored in ferritin, or extracellularly complexed with glycoproteins like transferrin (7,8). S. aureus has evolved several mechanisms to counter iron restriction and exploit a variety of host iron sources. Iron uptake strategies employed include heme uptake via the iron-responsive surface determinant system and through synthesis and secretion of two iron-chelating siderophores, staphyloferrin A (SA) 3 and staphyloferrin B (SB) (9 -11). Siderophores are small molecules with high iron affinity capable of outcompeting extracellular host iron-binding proteins for ferric iron. Iron-bound siderophores are selectively imported by dedicated surface receptors and cognate ABC transporters, HtsABC and SirABC, for SA and SB, respectively (12).
SA and SB have both been implicated in pathogenesis of S. aureus, but SB has been recognized for its importance in severe disease phenotypes and promotion of staphylococcal virulence in abscess and endocarditis models of infection (13,14). Additionally, the SB biosynthetic locus and sirABC are among the most strongly up-regulated genes in bacteria isolated from the iron-restricted host (15)(16)(17). Importantly, SB biosynthesis can occur independent of TCA cycle activity, which is down-regulated in S. aureus as part of the iron-sparing response (18,19). In contrast, SA biosynthesis is dependent on the TCA cycle for citrate precursors (18).
Staphyloferrin biosynthesis is achieved through non-ribosomal peptide synthetase-independent siderophore (NIS) bio-synthesis (20,21). NIS synthesis systems involve intermediates and enzymes that are freely dissociable from each other during assembly. Primary metabolites generally serve as the source of building blocks to generate siderophores and can be first chemically modified before incorporation into the siderophore. In addition to the synthetases required for assembly of the siderophore, NIS biosynthetic gene clusters generally encode enzymes that synthesize the necessary precursors to be assembled by the NIS synthetases (22).
The SB biosynthetic operon consists of nine genes, sbnABC-DEFGHI. SbnC, SbnE, and SbnF are the three synthetases and SbnH is a decarboxylase that together are required for SB assembly from one molecule each of ␣-ketoglutarate (␣-KG) and citrate and two molecules of L-2,3-diaminopropionate (L-Dap) (21). These essential precursor molecules are synthesized by SbnA, SbnB, and SbnG. SbnG is a dedicated citrate synthase that produces citrate from oxaloacetate and acetyl-CoA (18,23,24). ␣-KG and L-Dap are provided by the sequential enzymatic activities of SbnA and SbnB, which use substrates O-phospho-L-serine (OPS) and L-glutamate (25,26). OPS is postulated to be shunted from the serine biosynthesis pathway stemming from 3-phosphoglycerate produced in glycolysis ( Fig. 1) (18,26).
SbnI, the ninth gene product, is a heme-responsive transcriptional regulator for SB production (27). SbnI is required for full expression of sbnD-H and thus controls SB-mediated iron acquisition. SbnI binds a 271-bp DNA fragment in the sbnC coding region. However, SbnI can bind heme, which abrogates interaction with DNA, thereby limiting SB production. A model is proposed by which SbnI is required for transcription of the full SB biosynthetic operon and senses heme to reduce SB synthesis in favor of heme acquisition (27). This model also provides a mechanism for the findings that S. aureus demonstrates a heme-iron preference in vitro (28).
Based on primary sequence analysis, homology was not detected between SbnI and any characterized transcription factors or heme-binding proteins (27), but it is annotated as con-taining an N-terminal ParB-like domain. ParB is an essential component of the chromosome segregation system in bacteria (29). However, the conserved N-terminal ParB-like domain in SbnI is not responsible for interaction with DNA in ParB (30), leaving the role of the ParB-like domain in SbnI unknown.
The aim of this study was to gain insight into how SbnI functions using X-ray crystallography. The SbnI structure revealed striking structural homology to a recently characterized free serine kinase, SerK, from the archaea Thermococcus kodakarensis (31). SerK can phosphorylate free L-serine using ADP to generate OPS for cysteine biosynthesis (32). Herein, we demonstrate that SbnI is also a free serine kinase that uses ATP to phosphorylate L-serine to yield OPS and ADP (Fig. 1). The structure of SbnI, supported by site-directed mutagenesis, suggests that it follows a similar open-close reaction mechanism as proposed for SerK. Additionally, SbnI-generated OPS can be used by SbnA in vitro and serves as the in vivo source of OPS for SB production. To our knowledge, this is the first example of a bacterial free serine kinase and the first described free serine kinase that is ATP-dependent. This enzymatic function earns SbnI an enzymatic role in the SB biosynthetic pathway in addition to its heme-dependent transcriptional regulatory function.

Structure determination of SbnI
Full-length (254-amino acid) SbnI was not amendable for X-ray crystallography due to its propensity to precipitate. To improve solubility and stability in solution, several expression constructs were made containing varying N-and C-terminal truncations based on predicted secondary structure and disorder identified using PSIPRED and DISOPRED, respectively (33,34). One construct containing a 14-amino acid C-terminal truncation, SbnI(1-240), had improved stability in solution and produced well-diffracting crystals. The structure of selenomethionine-labeled SbnI(1-240) was determined to 2.5 Å resolution using single wavelength anomalous dispersion in space

SbnI serine kinase activity and role in SB biosynthesis
group P3 1 with one molecule in the asymmetric unit (Fig. 2a). Data collection and refinement statistics are summarized in Table 1. All 240 residues were modeled with 98% of residues in the most favored regions of the Ramachandran plot.
SbnI(1-240) is composed of two domains, domain I and domain II. Domain I consists of residues Met 1 -Gln 83 and Ile 205 -Ala 240 , and domain II is composed of residues Tyr 84 -Asn 204 . Domain I includes a conserved core ParB/Srx fold, corresponding to the annotated ParB-like domain based on primary sequence. This fold has been described in a functionally diverse ParB/Srx superfamily of proteins. Members are found in varied biological contexts and thus far are described to bind a nucleotide for kinase, ATPase, or DNase activity (32,36). To our knowledge, no ParB/Srx family member has been found to be directly involved in siderophore biosynthesis. The ParB/Srx core domain is composed of a four-strand mixed ␤-sheet and two ␣-helices, ␣2 and ␣3 (Fig. 2a), and contains an absolutely conserved GXXR motif, 59 GVHR 62 in SbnI.
Domain II is composed of a mixed ␣/␤ fold with a central four-stranded antiparallel ␤-sheet surrounded by five ␣-helices. A pair of antiparallel ␤-strands abut the main sheet and serve as a linker to domain I. B-Factor analysis reveals that domain II has a relatively high average B-factor of 109 Å 2 , compared with 75 Å 2 in domain I, suggesting that domain II has more disorder in the crystal and the domains are connected by a flexible linker.
SerK is a free serine kinase from T. kodakarensis that uses ADP to phosphorylate L-serine to generate OPS for cysteine biosynthesis. Of the proteins annotated in the ParB/Srx family, SerK is the only identified kinase, although Osa and Srx both possess ATPase activity (36,40). Overall, the structures of SerK . c, superimposition with the "closed" conformation of SerK (blue, PDB code 5X0E) in schematic putty representation, where the size of the tube depends on the B-factor in each structure. d, structural superimposition of SaSbnI(1-240) (raspberry) and SerK (yellow) (PDB code 5X0E). Selected active-site residues, AMP, and OPS are drawn as sticks. Mg 2ϩ is drawn as a green sphere, and oxygen, nitrogen, and phosphorus atoms are colored red, blue, and orange, respectively. AMP and OPS carbons are colored cyan.

SbnI serine kinase activity and role in SB biosynthesis
and SbnI are very similar. The SerK domain II has a high average B-factor and superimposition of SbnI(1-240) with the SerK structure in a "closed" conformation (PDB code 5X0E) suggests how SbnI may possess similar domain flexibility in solution (Fig. 2c). Additionally, structural superimposition and multiple sequence alignments revealed that several residues important for substrate and product binding identified in the SerK crystal structure (PDB code 5X0E) are conserved in SbnI. Moreover, the activesite architecture is highly conserved between the two proteins ( Fig.  2d and Fig. S1). Of the active-site residues, SerK Glu 30 was identified as a catalytically essential residue, and Asp 69 is required for magnesium ion binding; site-directed mutagenesis of either of these residues abolished SerK kinase activity (31). The homologous residues in SbnI are Glu 20 and Asp 58 . The genomic context of sbnI was analyzed to compare with the genomic neighborhoods of homologs. All staphylococcal sbnI homologs are part of the nine-gene SB biosynthetic cluster. More distant homologs co-occur with sbnA and sbnB homologs, either alone or in combination with different putative siderophore biosynthetic enzymes (Fig. 3). More distant SbnI ho-mologs are shorter, and alignments suggest that they have an abbreviated domain II (Fig. S1). Interestingly, a putative sbnI ortholog was identified upstream of the sbnA-H gene locus in Ralstonia solanacearum, which was previously thought to lack a SbnI homolog but still produces SB (41). A multiple-sequence alignment of SerK and SbnI homologs used in the genomic neighborhood analysis reveals that certain key residues important for catalysis, substrate, and product binding identified in SerK are fully conserved (Fig. S1). Notably, these include the SerK catalytic residue Glu 30 ; magnesium ion-binding residue Asp 69 ; residues implicated in interacting with the ␤-phosphate of ADP or phosphate group of OPS, His 72 and Arg 73 ; and residues that interact with the serine moiety of OPS, Trp 102 and Thr 223 (Fig. S1). More variability is seen with the SerK residues interacting with the adenosine group, raising the possibility that SbnI and other homologs may use a different phosphate donor or binding mode. Overall, the genomic neighborhood and sequence analyses suggest that free serine kinases are found in diverse species belonging to Firmicute and Proteobacteria phyla.

SbnI serine kinase activity and role in SB biosynthesis
Consurf analysis of SbnI(1-240) was used to map conserved regions to the molecular surface (Fig. 4). Highly conserved residues, including those that form the kinase active site, delineate a groove between domain I and II (Fig. 4, a and b). Structural alignment with the structure of the SerK ternary product complex (PDB code 5X0E) revealed that among SbnI homologs, the putative active site is highly conserved, whereas the remainder of the protein surface is variable (Fig. 4c).

SbnI is a dimer in solution
The oligomeric state of SbnI was analyzed using dynamic light scattering (DLS). Because SbnI contains seven Cys residues, the analysis was conducted in the presence of GSH as a reductant. The calculated molecular mass based on amino acid sequence of full-length SbnI is 30 kDa, and the molecular mass measured by DLS was 61 Ϯ 6 kDa with an average of 24 Ϯ 7% of polydispersity, implying that it predominantly forms a dimer in solution (Fig. S2). The molecular mass of SbnI(1-240) by DLS was 28 Ϯ 3 kDa with 34 Ϯ 1% polydispersity (Fig. S2). SbnI(1-240) has a calculated mass of 28 kDa, implying that it is primarily a monomer in solution. These data indicate that the C-terminal 14 amino acids excluded from the SbnI(1-240) construct are important for dimerization of the full-length protein.

SbnI is a serine kinase that uses L-serine and ATP to generate OPS
OPS is a substrate for SbnA in SB biosynthesis (25,26), lending support to our hypothesis that SbnI produces OPS for use by SbnA. To thus test if SbnI is capable of producing OPS, the spectral changes that occur when SbnA binds OPS were used to assay SbnI activity. SbnA has a characteristic absorption maximum at 412 nm attributable to an internal Schiff base formed between its pyridoxal 5Ј-phosphate (PLP) cofactor and an active-site lysine. Adding OPS to SbnA causes a rapid change in UV-visible spectra with the appearance of absorption peaks at 324 and 467 nm (Fig. 5a), characteristic of the formation of an external aminoacrylate intermediate (25). This spectral change is specific to OPS and does not occur with O-acetyl-L-serine or L-serine (25). SbnA incubated with SbnI, L-serine, and ADP resulted in no change in the UV-visible spectrum of SbnA. However, SbnA incubated with SbnI, L-serine, and ATP led to a shift in the UV-visible spectra indicative of OPS production and reaction with SbnA-PLP to form the external aminoacrylate ( Fig. 5, b and c). No spectral change was observed when SbnI was omitted, indicating that only the SbnI enzymatic product could react with SbnA-PLP. Therefore, we conclude that the reaction product is most likely OPS, and SbnI activity is ATP-dependent.
Phosphate acceptors, alternative to L-serine, were tested using a pyruvate kinase/lactate dehydrogenase (PK/LDH) assay for detection of ATP conversion to ADP. L-Threonine, ␣-KG, and L-Dap were not phosphate acceptors (data not shown). Additionally, SbnI did not phosphorylate the serine residue in a His-Ser dipeptide (data not shown).
SbnI-mediated conversion of ATP to ADP and generation of OPS were monitored using HPLC and 31 P NMR. Incubation of SbnI with ATP and excess L-serine led to turnover of ATP to ADP, as detected by HPLC (Fig. 6a), indicating that SbnI has ATPase activity. Unlike SerK, no turnover of ADP to AMP was detected with ADP as a phosphate donor by HPLC (Fig. 6a). ADP was not generated when L-serine was excluded, indicating that this activity requires the presence of L-serine. 31 P NMR was also used to monitor SbnI-mediated ATPase activity and generation of OPS from L-serine. The observed disappearance of the ATP ␥-phosphate 31 P signal with the concomitant appearance of a 31 P signal of OPS demonstrates transfer of the ATP ␥-phosphate to L-serine to yield ADP and OPS (Fig. 7). The chemical shift of SbnI-generated OPS was consistent with the 31 P NMR spectrum measured for an OPS standard (Fig. S3). Attempts to obtain a co-crystal structure of SbnI with identified substrates (or ATP analog, AMP-PNP) or products, both in the presence and absence of heme, have not been met with success.

SbnI active-site variants
The role of Glu 20 and Asp 58 in SbnI kinase function were tested by site-directed mutagenesis. Two mutants, each containing a single alanine substitution, SbnI E20A and SbnI D58A, were generated. Using the PK/LDH assay, we determined that the mutants were incapable of turning over ATP to ADP in the presence of L-serine. Additionally, no ADP could be detected by HPLC in reactions containing the SbnI mutants incubated with ATP and L-serine (Fig. 6b); together, these data allow us to conclude that these mutants are catalytically inactive. These data also correlate with SbnA-PLP UV-visible absorption spectra that demonstrated that these SbnI variants also do not produce OPS (Fig. 5, d and e). The importance of these residues in catalysis is consistent with the reaction mechanism presented for SerK in which substrate binding promotes conformational closure positioning the catalytic Glu 30 (Glu 20 in SbnI) close to the hydroxyl group of bound L-serine. Glu 30 is a catalytic base that deprotonates the hydroxyl group of L-serine. The deprotonated hydroxyl can then attack the phosphorus atom of the ADP ␤-phosphate to yield OPS and AMP (31). Our results suggest SbnI uses a similar two-ligand binding sequential mechanism but instead uses L-serine and ATP to yield ADP and OPS.

Kinetic analysis of SbnI kinase activity
To obtain kinetic parameters for the SbnI kinase activity, SbnI enzymatic turnover was monitored using an established coupled assay for ADP using PK/LDH. The steady-state kinetic parameters of SbnI reaction with ATP and L-serine were determined and are summarized in Table 2. The saturating concentration of L-serine was beyond conditions permissive to the assays, and thus K m could not be accurately determined. A second assay to measure SbnI enzymatic turnover employed SbnA-dependent turnover of OPS coupled to a phosphate release detection assay. SbnA is the most likely in vivo acceptor of OPS generated by SbnI as SbnA requires OPS as a substrate for synthesis of SB precursors, L-Dap and ␣-KG, in concert with SbnB. Using excess concentrations of SbnA, the coupled assay demonstrated that SbnA could use SbnI-generated OPS and supplied L-glutamate to generate its products, N-(1-amino-1carboxyl-2-ethyl)-glutamic acid and inorganic phosphate. The kinetic parameters of SbnI reaction with ATP and L-serine using the SbnA coupled assay were determined and are sum-marized in Table 2. The kinetic parameters measured using both methods agree with rates and K m values in the same order of magnitude. SbnI mutants E20A and D58A were catalytically insufficient to accurately measure enzyme rates. Compared with SerK, SbnI has a K m for ATP 1 order of magnitude lower than SerK has for ADP and a higher K m for L-serine by 2 orders of magnitude.
The enzymatic activity of SbnI(1-240) was also measured because the 14-amino acid C-terminal truncation does not exclude any regions in the SerK structures identified as required for substrate and product binding or catalysis. SbnI(1-240) displayed decreased kinase activity compared with fulllength SbnI. These data correlate with the HPLC and spectral data with SbnA for SbnI(1-240) having intermediate activity relative to full-length SbnI (Figs. 5f and 6b). Plots of initial velocities used for determination of kinetic constants are included in Fig. S4. Given that SbnI(1-240) is a monomer in solution (see above), we conclude that dimerization/oligomerization of SbnI is not essential for its kinase activity.

Physiological function of ATP-dependent serine kinase activity in S. aureus
The serC gene encodes the enzyme responsible for OPS synthesis from 3-phosphohydroxypyruvate in the S. aureus serine biosynthesis pathway. To our knowledge, before this work, SerC activity was the only identified metabolic source of OPS in S. aureus and was the assumed source of OPS for SB biosynthesis (18,26). To test whether other metabolic sources of OPS exist in S. aureus, wildtype USA300 and a serC transposon insertion mutant USA300 strain (serC) were grown in chemically defined media containing glucose with and without L-serine in a TECAN plate reader. In L-serine-replete medium, both wildtype USA300 and the serC mutant grew, although the serC mutant entered stationary phase at a lower cell density and failed to reach biomass equivalent to that of the WT culture (Fig. 8a). However, in stark contrast, in a growth medium lacking L-serine, wildtype USA300 grew, whereas the serC mutant failed to grow altogether (Fig. 8a). Thus, we conclude that the serC transposon mutant is a bona fide serine auxotroph; a serine auxotrophy phenotype was observed previously for an Escherichia coli serC mutant strain (42). We also conclude that SerC activity is the sole source of OPS for serine biosynthesis; because the medium is iron-restricted, SbnI would be expressed, and, if it contributed to OPS biosynthesis for use in the L-serine biosynthetic pathway, we would not have observed L-serine auxotrophy for the serC mutant (see Fig. 1).
To test whether SbnI could serve as a source of OPS for SB biosynthesis, the serC transposon insertion mutant was tested for its ability to make SB. A disc diffusion assay was used to detect SB in spent culture supernatant of serC and sbnI transposon insertion mutants, compared with the wildtype USA300 S. aureus strain. Briefly, spent culture supernatants were applied to sterile filter discs on iron-restricted agar seeded with either sirA or htsABC mutant S. aureus strains. The sirA mutant strain is defective for SB uptake, so growth promotion around the disc in iron-restricted media is dependent on the presence of SA in the supplied culture supernatant. The htsABC mutant strain is unable to take up SA, and thus growth around the disc is reliant on the presence of SB in supplied culture supernatant. Similar to wildtype S. aureus, the serC transposon insertion mutant was capable of producing SB (Fig.  8b), indicating the presence of another pathway for OPS production independent of SerC, almost assuredly via SbnI (Fig. 1).
The sbnI mutant was impaired for SB production (Fig. 8b), consistent with previously published results using the disc diffusion assay (27). Decreased SB production by the sbnI mutant is due to the necessity of SbnI for full expression of sbnDEFGH (27). A serC sbnI double mutant is expected to be SB-deficient. However, as the sbnI mutation is pleiotropic on SB biosynthesis due to its requirement for transcription of sbn biosynthetic genes, assaying for SB production does not yield information about a role in OPS production. Nonetheless, our results support the hypothesis that SbnI serves as a second metabolic route for OPS biosynthesis from L-serine in S. aureus. Production of OPS from L-serine and ATP is a previously unrecognized route for OPS synthesis in bacteria.

Discussion
SbnI is a free L-serine kinase that makes OPS, which serves as a substrate for SbnA and is a precursor for SB biosynthesis. Insight into the kinase function of SbnI was gained based on homology identified by structural similarity with SerK, an enzyme in a biosynthetic pathway for cysteine in the thermophilic archaea T. kodakarensis (31). ATP is used as the phosphate donor in the reaction catalyzed by SbnI, whereas SerK uses ADP. Interestingly, hyperthermophiles generally use ADP in place of ATP in key glycolytic enzymes as an adaptation to life at high temperatures, presumably because ADP is more stable than ATP (43). Whereas SbnI is capable of phosphorylation of free L-serine, it was not capable of mediating phosphotransfer to serine residues within a His-Ser dipeptide. Bacterial Ser/Thr kinases such as S. aureus Stk1 (also named PknB) (44) and cognate phosphatases function as molecular switches that have key roles in bacterial cell signaling as in eukaryotic systems (45). However, SbnI free L-serine kinase activity is functionally and enzymatically distinct from that of Ser/Thr kinases.

SbnI serine kinase activity and role in SB biosynthesis
the C terminus and is monomeric, either the C terminus or dimerization is required for full kinase function. The residues shown to be required for phosphotransfer in SerK (Glu 30 and Asp 69 ) are conserved in SbnI and present in the truncated protein. Moreover, three of four residues interacting with the serine substrate in SerK are conserved (Glu 30 , Trp 102 , and Thr 223 ), and the fourth position is a conservative substitution of His 225 with Phe 203 in SbnI (Fig. 2d). In contrast, interactions of SerK with ADP/AMP are poorly conserved. Of eight residues making key contacts, only three residues (Ser 43 , His 72 , and Arg 73 ) are conserved in SbnI, and two of these interact with the phosphate groups. Attempts to obtain crystals with substrates or products bound to SbnI or SbnI(1-240) have not been met with success, and the binding mode of ATP to SbnI remains elusive. Structural analysis of SerK suggests that the conformational closure upon binding both substrates positions the catalytic glutamate (Glu 30 ) to deprotonate L-serine to attack the terminal phosphate of ADP. With no published structure of substrate or product-free SerK, the apo-SbnI(1-240) structure supports this proposed mechanism in that the unbound form is in an open conformation to expose the binding pocket. Measurement of SbnI enzyme kinetic parameters revealed that, compared with SerK, it has a relatively low selectivity for L-serine. The comparatively low k cat /K m could relate to the physiological role of SbnI in S. aureus. SerK supplies cysteine synthase with OPS to produce cysteine and may represent an ancient heterotrophic mechanism of amino acid metabolism (32). Interestingly, this cysteine synthase is a distant SbnA homolog and a true OPS sulfhydrylase. Also, SerK is postulated to provide an advantage by enabling carbon from serine to be directed to glycolysis and gluconeogenesis by conversion to OPS (32). In vitro evidence suggests that S. aureus uses amino acids to support gluconeogenesis (46). However, serine is used to generate ATP and acetate rather than to facilitate gluconeogenesis as in T. kodakarensis (32,46). In contrast, SbnI kinase activity fulfills a distinct physiological role, providing substrate necessary for SB production. The comparatively low k cat /K m may allow SbnI to respond to a greater range of substrate concentrations, such that at high L-serine concentrations, SbnI increases the rate of OPS production for SB production and possibly other metabolic processes. Although SbnI has lower catalytic efficiency than SerK (Table 2), the related ParB/Srx family member, Srx, is similar to SbnI as it is not a highly efficient enzyme. The Srx ATP-dependent reduction of peroxiredoxin sulfinic acid activities described in mammals and plants so far is slow, with turnover rates of 0.2-0.5 min Ϫ1 for ATP, which are an order of magnitude lower than SbnI (47-49). However, the catalytic efficiencies of Srx are 0.8 -8.4 mM Ϫ1 min Ϫ1 and in the range measured for SbnI (47,49).
SbnI is a sufficient biological source of OPS for SB biosynthesis and contributes to the functional modularity of the sbn locus. A recognized characteristic of bacterial networks is a high degree of modularity and sparse connectivity between individual functional modules, where a functional module refers to a group of biological components that are spatially isolated or chemically specific and work together for a discrete biological function (50). The functional redundancy of enzymes encoded in the sbn locus for generation of precursor substrates, SbnG and SbnI, decreases dependence on central metabolism and contribute to the modularity of the sbn locus. SbnG is functionally redundant with the TCA cycle citrate synthase, CitZ (23). Moreover, SbnG activity allows for SB biosynthesis to occur independent of TCA cycle activity (18). This functional independence is important because S. aureus elicits an ironsparing response during infection resulting in down-regulation of the TCA cycle. We propose that the functional redundancy of SbnI with SerC for OPS production allows SB biosynthesis to occur independent of glycolysis as SerC substrate is funneled from 3-phosphoglycerate. Additionally, the serine biosynthetic pathway is regulated by negative feedback where SerA, metabolically upstream of SerC, is allosterically inhibited by L-serine and could limit the amount of SerC-derived OPS available to support SB synthesis when serine is abundant. Thus, an alternative, SB-dedicated, OPS synthetic route via SbnI is advantageous. Together with SbnG, SbnI allows SB biosynthesis to occur autonomously from glycolysis and TCA cycle activity by generating precursor substrates dedicated to SB biosynthesis.
This observed metabolic redundancy may improve robustness and help buffer environmental perturbations S. aureus encounters during infection. Glucose is the preferred carbon source by S. aureus and available at concentrations to support growth in human blood (51). However, within staphylococcal Figure 8. a, growth kinetics of S. aureus USA300 wildtype (black lines) and serC transposon insertion mutant (red lines) strains in CDMG (solid lines) and CDMG without L-serine (CDMG Ϫ L-Ser) (dashed lines). b, agar plate disc diffusion bioassays were performed using culture supernatants prepared from S. aureus USA300 strains (wildtype and sbnI and serC transposon insertion mutants), as indicated on the x axis, that were grown for 16 h in Chelex-treated Tris minimal succinate (c-TMS) medium. The black dots (labeled SA (⌬sirA)) are a measure for the presence of SA in culture supernatants, and the gray dots (labeled SB (⌬htsABC)) are a measure of SB in culture supernatants based on the growth radius around the disc. The disc radius (3 mm) is subtracted from the reported growth radius. Lines represent the S.D. ***, p Ͻ 0.0002; ****, p Ͻ 0.0001.

SbnI serine kinase activity and role in SB biosynthesis
abscesses, glucose is limiting, and S. aureus probably survives through catabolism of secondary carbon sources, specifically lactate, peptides, and free amino acids (46,52,53). Resultant changes in the central metabolism of S. aureus may not severely hinder SB biosynthesis due to the modularity of the sbn system. Despite variability in growth conditions, one constant in S. aureus survival strategy is iron acquisition.
Serine is among the most abundant amino acids found in fluid obtained from S. aureus-infected prosthetic joints. SB biosynthetic genes, including sbnI, but not SA biosynthetic genes, are up-regulated in this environment (17). Also, L-serine is one of the amino acids most rapidly consumed by S. aureus grown on amino acids in vitro (46). Together, the combination of high L-serine in the extracellular milieu and the rapid consumption of L-serine may limit the activity of the endogenous serine biosynthetic pathway. Under these conditions, OPS from serine biosynthesis is likely to be limiting to feed SB biosynthesis, justifying the necessity of SbnI-generated OPS. Interestingly, SB production is restricted to more invasive coagulase-positive staphylococcal strains, and the sbn locus may represent a lineage-specific innovation to support adaptation to a more invasive lifestyle. In contrast, SA biosynthetic genes are found across both coagulase-positive staphylococci and more commonly commensal coagulasenegative staphylococci.
This newly determined enzymatic role of SbnI generates questions regarding how it carries out heme-dependent transcriptional regulation of the sbn locus. It remains unclear whether SbnI kinase activity impacts its heme-dependent transcriptional regulation of the sbn locus. The structure of SbnI(1-240) did not reveal a prototypical DNA-binding motif or hemebinding site. The functional sites required for kinase activity could be structurally distinct from those involved in the transcriptional regulation and heme binding. Full-length SbnI is a dimer, yet SbnI(1-240) is a monomer in the crystal structure and in solution. The C-terminal 14 amino acids excluded from the SbnI(1-240) construct are probably required for multimerization, which could be important for transcriptional regulatory function or heme binding. Additionally, SbnI has a C-terminal extension of 34 amino acids compared with SerK, and this C-terminal region is truncated in distant SbnI homologs not associated with an SB biosynthetic locus. This raises the possibility that kinase activity may be the core function of the non-SB-associated SbnI homologs. The C-terminal extension in SbnI could be a structural adaptation to facilitate multiple biological functions in SB precursor biosynthesis and sbn locus gene regulation. Heme-dependent SbnI-regulatory activity may further enhance network robustness as a negative feedback loop from an iron acquisition perspective. Characterization of SbnI heme binding and its effect on kinase activity remains the subject of future work.
OPS production by the kinase activity of SbnI represents a new biosynthetic path for production of this metabolite in bacteria. SbnI may serve as a metabolic adaptation to facilitate SB production when growing on non-preferred carbon sources. Furthermore, it demonstrates the metabolic flexibility S. aureus possesses to allow employment of iron uptake strategies in changing host environments.

Cloning and site-directed mutagenesis
Constructs with an N-terminal His 6 tag and thrombin cleavage site were generated in pET28a vectors for recombinant expression of S. aureus full-length SbnI, residues 1-254, with the first codon mutated from the native TTG to a common start codon, ATG, and the S. aureus C-terminal truncated construct SbnI(1-240) (residues 1-240). The S. aureus SbnI nucleotide sequence can be accessed in the GenBank TM database under accession code NC_009641.1 (positions 90178 -90942) (gene locus NWMN_RS00380), and the amino acid sequence can be accessed through NCBI Protein Database under NCBI accession WP_001015549.1. Briefly, a megaprimer-based wholeplasmid synthesis PCR cloning protocol was used to clone constructs amplified from chromosomal DNA from S. aureus strain Newman (54). S. aureus SbnI variants E20A and D58A were produced using a single-primer mutagenesis method (55). Primers used in this study are summarized in Table S1. All clones were introduced into E. coli BL21 (DE3) and confirmed by DNA sequencing. Bacterial strains and plasmids used in this study are summarized in Table 3.

Protein expression and purification
Recombinant full-length SbnI, SbnI(1-240), SbnI E20A, and SbnI D58A constructs were overexpressed in E. coli BL21 (DE3) cells. Cultures were grown in 2ϫ YT medium supplemented with 25 g/ml kanamycin at 30°C to an A 600 of 0.7-0.9. Cultures were then induced with 0.5 mM isopropyl ␤-D-thiogalactopyranoside and grown for an additional 18 h at 20°C. Cells were pelleted by centrifugation at 4,400 ϫ g for 7 min at 4°C and resuspended in buffer containing 50 mM HEPES (pH 7.4), 300 mM NaCl, 5% (v/v) glycerol, 2 mM tris(2-carboxyethyl) phosphine (TCEP), and 10 mM imidazole on ice. Approximately 5 mg of DNase was added to cell suspension before lysis at 4°C using an EmulsiFlex-C5 homogenizer (Avestin). Insoluble material was removed by centrifugation at 39,000 ϫ g for 1 h, and recombinant protein was purified from soluble lysate using a HisTrap nickel affinity column (GE Healthcare) by elution with an imidazole gradient. Protein was dialyzed against 50 mM HEPES (pH 7.4), 100 mM NaCl, 5% (v/v) glycerol, and 2 mM TCEP and then cleaved with thrombin at a 1:500 ratio by weight of His 6 protein to remove the His 6 tag over 18 h at 4°C. Subsequently, recombinant protein was dialyzed into 50 mM HEPES (pH 7.4), 5% (v/v) glycerol, and 2 mM TCEP and further purified by anion-exchange chromatography using a Source 15Q column (GE Healthcare). Purified protein was obtained by elution with a NaCl gradient and further dialyzed into 50 mM HEPES (pH 7.4), 300 mM NaCl, 5% (v/v) glycerol, 2 mM GSH. The sample was concentrated to ϳ20 mg/ml, flash-frozen, and stored at Ϫ80°C. Selenomethionine-incorporated SbnI(1-240) was produced by methods described previously (56) and purified as described above for native SbnI(1-240).
His 6 -tagged SbnA was expressed in E. coli BL21 (DE3) cells from the plasmid pET28a, purified by His tag affinity chromatography, and digested with the thrombin to remove the His 6 tag. The protein was further purified by anion-exchange chromatography using the previously published method for

Crystallization, data collection, and structure determination of SbnI(1-240)
Selenomethionine-labeled SbnI(1-240) crystals were grown by sitting-drop vapor diffusion at 4°C in 2-l drops with a 1:1 mixture of ϳ20 mg/ml SbnI(1-240) in 50 mM HEPES (pH 7.4), 100 mM NaCl, 5% (v/v) glycerol, and 2 mM TCEP with reservoir solution containing 0.18 M HEPES (pH 7.5) and 20% (w/v) PEG 8000. Crystals were briefly soaked in reservoir buffer supplemented with 30% (v/v) glycerol for cryoprotection and flashfrozen in liquid nitrogen. A single-wavelength anomalous diffraction data set was collected at the Canadian Light Source on Beamline 08B1-1 (57). The data were processed and scaled using XDS (58,59). Crystals were of space group P3 1 with one molecule in the asymmetric unit. Five selenomethionine sites were identified for phasing to build a preliminary model using AutoSol (initial figure of merit of 0.35) and Autobuild (187 of 240 residues built) programs in Phenix (60). Manual building was done using Coot (61), and refinement was performed with phenix.refine using translation liberation screw parameters with three groups (62). The refined structure contains Met 1 -Ala 240 , one glycerol, and 12 water molecules. Data collection and refinement statistics are summarized in Table 1. Structure figures were generated in PyMOL (PyMOL Molecular Graphics System, version 1.8, Schrödinger, LLC, New York). Domain analysis was done using the Dali server for comparison of the protein structure against structures in the PDB (63).

Genomic neighborhood and conservation analysis
Protein homology between genomic regions carrying SbnI orthologs was plotted using a custom Biopython script, bio.links.py, based on output from BLASTP 2.2.2.28 (e value Յ 1.00eϪ40) (https://github.com/minevskiy/bioinformatics). 4 Species used for comparison were found by BLAST search of SbnA or SbnI and STRING analysis of SbnI. Orthologous genes are indicated in the same color.
Sequence conservation was mapped onto the SbnI(1-240) structure using ConSurf (64). The multiple sequence alignment used for the analysis was generated using default ConSurf parameters and the SbnI amino acid sequence as the search sequence.

Dynamic light scattering
Samples of SbnI and SbnI(1-240) were analyzed by DLS using a DynaPro Plate Reader (Wyatt Technologies). Protein was diluted to 0.5 mg/ml with 50 mM HEPES (pH 7.4), 300 mM NaCl, 5% (v/v) glycerol, and 2 mM GSH, and results were generated based on averaging five 5-s acquisitions. Data were collected at room temperature. Values reported are an average of data collected.

UV-visible spectrophotometry analysis of SbnI OPS production using SbnA
UV-visible spectra were collected using a Varian Cary 50 UV-visible spectrophotometer. The SbnA-PLP spectrum was recorded at a concentration of 15 M in 50 mM HEPES, pH 7.4, 100 mM NaCl, 5% (v/v) glycerol. The spectrum of SbnA aminoacrylate aldimide complex was recorded immediately after the addition of 30 M OPS (25). The spectral shift observed when SbnA binds OPS was used to evaluate OPS production by SbnI. The spectrum of 15 M SbnA in 50 mM HEPES, pH 7.4, 100 mM NaCl, 5% (v/v) glycerol, 20 mM MgCl 2 , 25 mM L-serine, and 5 mM ATP was recorded before and after the addition of 15 M SbnI, SbnI E20A, SbnI D58A, or SbnI(1-240). Phosphate donor specificity was also examined using 5 mM ADP in place of ATP.

HPLC
Kinase activity of SbnI was detected using HPLC to examine production of ADP from ATP. The reaction mixture was composed of 50 mM HEPES (pH 7.4), 100 mM NaCl, 2.5% glycerol, 50 mM L-serine, 0.25 mM ATP or ADP, 10 mM MgCl 2 , and 5 M SbnI, SbnI E20A or SbnI D58A. The reaction was carried out for 1 h at room temperature (22°C). The protein was removed by centrifugation using a 3K Nanosep column. The filtrate was Table 3 Bacterial strains and plasmids used in this study Em R , Km R , and Tc R designate resistance to erythromycin, kanamycin, and tetracycline, respectively. IPTG, isopropyl 1-thio-␤-D-galactopyranoside.

P NMR spectra of SbnI kinase reaction
The reaction mixture contained 50 mM HEPES (pH 7.4), 100 mM NaCl, 2.5% (v/v) glycerol, 10 mM MgCl 2 , 48 mM L-serine, 5 mM ATP, and 5% D 2 O. NMR spectra were collected at 25°C using a broadband frequency probe with Z-magnetic field gradient in a Bruker Avance III 500-MHz spectrometer. One-dimensional 31 P spectra were recorded at different time points before and after the addition of 4.8 M SbnI until reaction completion. The spectra were referenced to 2,2,6,6-tetramethylpiperidine, which was set to 0 ppm. The spectra were processed and using TopSpin TM (Bruker). The chemical shifts of ATP (65) and OPS were assigned using reference spectra (Fig. S3).

Measurement of serine kinase activity
ATP-dependent serine kinase activity of SbnI was measured using a PK/LDH coupled assay. The assay is based on a reaction in which the regeneration of hydrolyzed ATP is coupled to the oxidation of NADH (66). The rate of NADH absorbance decrease at 340 nm (A 340 nm ϭ 6,220 M Ϫ1 cm Ϫ1 ) is proportional to the rate of ATP conversion to ADP by SbnI kinase activity. Coupled reactions contained 50 mM HEPES, pH 7.4, 100 mM NaCl, 2.5% (v/v) glycerol, 10 mM MgCl 2 , 2 mM phosphoenolpyruvate, 1 ⁄ 50 of the final reaction mixture volume of PK/LDH enzyme (from rabbit muscle, Sigma-Aldrich, catalog no. P-0294), 5 mM ATP, and 100 mM L-serine. The mixture was incubated for 5 min to remove any contaminating ADP. Continuous measurement at 340 nm was recorded for 2 min before the addition of 0.5 M SbnI, SbnI(1-240), SbnI E20A, or SbnI D58A enzyme to start the reaction. The assay was run for 10 min. To determine kinetic parameters, the initial velocities of SbnI and SbnI(1-240) kinase reactions in the presence of varying concentrations of ATP with 100 mM L-serine and in the presence of varying concentrations of L-serine with 5 mM ATP were recorded. Phosphate donor specificity was investigated with 10 mM ADP, and phosphate acceptor specificity was examined with 50 mM L-threonine and 50 mM His-Ser dipeptide. In addition, alternative substrates, 50 mM L-Dap and 50 mM ␣-KG, were tested as phosphate acceptors. All data were collected on a Varian Cary 50 UV-visible spectrophotometer at room temperature (22°C), and a total of three replicates were collected for each reaction condition.
SbnI kinase activity was also measured using an assay with SbnA. In a reaction mixture, SbnI-dependent OPS production was measured using SbnA and L-glutamate to release inorganic phosphate and N-(1-amino-1-carboxyl-2-ethyl)-glutamic acid (26). A coupled enzymatic assay was used to detect inorganic phosphate release from OPS as described previously (67). The reaction mixture contained 50 mM HEPES, pH 7.4, 100 mM NaCl, 2.5% (v/v) glycerol, 10 mM MgCl 2 , 10 mM ATP, 100 mM L-serine, 0.2 unit of purine nucleoside phosphorylase, 400 M 2-amino-6-mercapto-7-methylpurine riboside, 5 M SbnA, 2 mM L-glutamate. The mixture was incubated for 10 min to remove any contaminating inorganic phosphate and establish a baseline. Continuous measurement at 360 nm was recorded for 2 min before the addition of 0.5 M SbnI or SbnI(1-240) enzyme to start the reaction. The assay was run for 15 min. To determine kinetic parameters, the initial velocities of SbnI kinase reactions in the presence of varying concentrations of ATP with 100 mM L-serine and in the presence of varying concentrations of L-serine with 10 mM ATP were recorded. All data were collected on a Varian Cary 50 UV-visible spectrophotometer at room temperature (22°C) and a total of three replicates were collected for each reaction condition. The concentration of inorganic phosphate release from OPS was determined using the extinction coefficient A 360 nm ϭ 11,000 M Ϫ1 cm Ϫ1 (67). Data were fit by nonlinear regression using a Michaelis-Menten model in GraphPad Prism version 6.

S. aureus bacterial strains and growth conditions
Experiments were performed with a derivative of S. aureus USA300 LAC cured of the 27-kb plasmid encoding macrolide resistance (68). The plasmid-cured USA300 LAC is referred to as USA300 throughout. Transposon insertion mutants JE2 serC::⌽N⌺; Em R (SAUSA300_1669) and JE2 sbnI::⌽N⌺; Em R (SAUSA300_0126) were obtained from the Nebraska Transposon Mutant Library containing the resistance cassette ermB, which confers resistance to erythromycin (69). Transposons were transduced to USA300 background strain using phage 80␣ and confirmed by PCR using published methods (70). USA300 serC::⌽N⌺ and USA300 sbnI::⌽N⌺ transposon insertion mutants are referred to as serC and sbnI throughout (Table  3). Bacterial growth curves to test serine auxotrophy were performed in chemically defined medium with 0.4% (w/v) glucose (CDMG) as described previously (18), with and without L-serine. Briefly, colonies of wildtype USA300 or serC transposon insertion mutant were inoculated from tryptic soy agar into 2 ml of CDMG overnight at 37°C. Cells were normalized to an A 600 of 0.1 and washed twice with CDMG lacking L-serine (CDMG-Ser), and 5 l of the resuspension was used to inoculate 200-l aliquots of CDMG or CDMG-Ser in 96-well plates. Cultures were grown in a TECAN plate reader for 24 h at 37°C with 10 s of shaking every 10 min, and the A 600 was assessed every 30 min. Data are representative of three independent experiments, and error bars signify S.E.

Disc diffusion assays to assess siderophore production
Concentrated spent culture supernatants were prepared from 10-ml cultures of S. aureus USA300, serC transposon insertion mutant, and sbnI transposon insertion mutant grown for 16 h in Chelex-100 -treated Tris minimal succinate (TMS), as described previously (71), in a flask/volume ratio of 10:1 at 37°C with shaking at 200 rpm without antibiotic selection. Growth was assessed via A 600 , and culture densities were normalized. Bacterial cells were pelleted by centrifugation, and culture supernatants were filter-sterilized and lyophilized overnight. Dried material was resuspended in 0.5 ml of sterile double-distilled H 2 O. To assess growth promotion of concentrated culture supernatants, S. aureus strain RN6390 sirA SbnI serine kinase activity and role in SB biosynthesis mutant (growth of this mutant is dependent on SA in supernatant) or htsABC mutant (growth of this mutant is dependent on SB in supernatant) derivatives, as described previously (72), were seeded into TMS agar containing 10 M ethylenediamine-N,NЈ-bis(2-hydroxyphenylacetic acid) to 2 ϫ 10 5 cells/ml. 10 l of concentrated supernatant was applied to sterile paper discs placed on TMS agar containing the seeded reporter strains, and growth radii about the discs were measured after 24 h of incubation at 37°C. The reported growth radius has the disc radius (3 mm) subtracted. Statistical analyses were conducted using two-way analysis of variance.