Native mass spectrometry identifies an alternative DNA-binding pathway for BirA from Staphylococcus aureus

An adequate supply of biotin is vital for the survival and pathogenesis of Staphylococcus aureus. The key protein responsible for maintaining biotin homeostasis in bacteria is the biotin retention protein A (BirA, also known as biotin protein ligase). BirA is a bi-functional protein that serves both as a ligase to catalyse the biotinylation of important metabolic enzymes, as well as a transcriptional repressor that regulates biotin biosynthesis, biotin transport and fatty acid elongation. The mechanism of BirA regulated transcription has been extensively characterized in Escherichia coli, but less so in other bacteria. Biotin-induced homodimerization of E. coli BirA (EcBirA) is a necessary prerequisite for stable DNA binding and transcriptional repression. Here, we employ a combination of native mass spectrometry, in vivo gene expression assays, site-directed mutagenesis and electrophoretic mobility shift assays to elucidate the DNA binding pathway for S. aureus BirA (SaBirA). We identify a mechanism that differs from that of EcBirA, wherein SaBirA is competent to bind DNA as a monomer both in the presence and absence of biotin and/or MgATP, allowing homodimerization on the DNA. Bioinformatic analysis demonstrated the SaBirA sequence used here is highly conserved amongst other S. aureus strains, implying this DNA-binding mechanism is widely employed.

The biotin retention protein A (BirA, also known as biotin protein ligase) is responsible for maintaining biotin homeostasis in many bacteria. BirA is a bi-functional protein capable of both enzymatic protein biotinylation as well as serving as a biotin-controlled transcriptional repressor that regulates the expression of the biotin biosynthesis operon (reviewed 1,2 ). By combining both activities in a single protein, BirA is uniquely placed as the key regulator of biotin metabolism. E. coli BirA (EcBirA) has been thoroughly investigated through genetic, biochemical and structural biology studies [3][4][5][6][7][8][9][10][11] . These have provided powerful insights into the maintenance of biotin homeostasis. EcBirA binds to its ligands, biotin and ATP, in an ordered manner 6,12 . Conformational changes induced by biotin binding create the pocket necessary for ATP to bind. ATP binding initiates the synthesis of biotinyl-5′-AMP, which serves as both a reaction intermediate for biotin ligation as well as a co-repressor. The BirA:biotinyl-5′-AMP complex, known as the holo-enzyme, is then available for two alternative activities: either as a biotin protein ligase when there is a substrate available for protein biotinylation or when demand for biotin is low, as a transcriptional repressor through homodimerization 13 . The transcriptional repressor function of EcBirA involves a co-operative interaction between two EcBirA subunits and an inverted palindromic repeat sequence present in the promoter of the biotin biosynthetic operon (bioO). Homodimerization of the holo-enzyme complex (K D 2-1 = 1-6 × 10 −6 M 14,15 ) is a pre-requisite for DNA binding 4,5,16,17 . The unliganded enzyme (i.e. apo-EcBirA) does not dimerize at physiological concentrations (K D 2-1 = 1 × 10 −3 M 14 ) and is unable to bind DNA 9,17 . Mutagenesis studies have further highlighted the importance of homodimerisation in the DNA binding

Results
Biotin-induced repression of transcription. Apo-SaBirA has been shown to bind DNA in vitro 27,29 . We sought to address if the non-liganded protein possessed repressor activity in an in vivo assay of gene expression. As it is technically challenging to achieve the apo-state in vivo, we employed the SaBirA-F123G mutant that impairs homo-dimerization and, therefore, mimics the monomeric properties of the non-liganded protein 29 . Wild-type EcBirA and SaBirA, and the dimerization disrupted EcBirA-R119W mutant 31 , were used as controls. Biotin-induced repressor activity was measured using a bacterial reporter system we have previously described 26 and summarized in Fig. 1A. Briefly, a birA gene under the control of the pLac-UV5 promoter was site-specifically integrated into the lambda phage attachment site (attB) in the chromosome of a biotin auxotrophic strain of E. coli harboring a mutant BirA that is unable to bind DNA 26 . The host strain also produces lac repressor from the strong LacI q promoter, allowing us to repress expression of the BirA variants down to the low concentrations necessary for a physiologically relevant assay. A second construct containing a BirA-target promoter fused to a lacZ reporter gene was then integrated into the HK022 attB phage attachment site. Here, the promoters for bioO from both E. coli (EcbioO) and S. aureus (SabioO) were analyzed alongside promoters for the bioY and the yhfS-yhfT operon in S. aureus 26 . Biotin-mediated gene control was then assayed using β-galactosidase activity as a readout for gene expression. The effect of exogenous biotin on these genetic circuits, engineered into the genome of a strain unable to synthesis its own biotin, was measured by the addition of biotin to the growth media.
The in vivo reporter assay was developed to reconstitute the biotin-induced transcriptional repressor activity of BirA. Previous studies have suggested that E. coli contains ≈10-20 molecules of BirA per cell 32 . Therefore, the assay was designed to minimize artefactual binding to DNA due to the overexpression of BirA. Our previous work has demonstrated that the level of lac repressor provided from a single copy pLacI q -lacI module is able to very significantly, though not completely, repress a single copy plac-UV5 promoter module 33 . As expected, biotin-regulated repression of β-galactosidase was observed when IPTG was omitted from the growth mediaimplying the low level of leaky expression from the repressed lacUV5 promoter provided sufficient SaBirA for activity. Conversely, addition of 10 μM IPTG induced much higher expression of SaBirA, as detected by Western blot, and also abolished biotin-regulated control of the reporter (Supporting Fig. S2). Similarly, the level of EcBirA expression obtained in the absence of IPTG was shown to progressively inhibit β-galactosidase expression with increasing levels of biotin present in the growth media, as expected (Fig. 1B). Hence, all in vivo experiments were performed in the absence of IPTG (Fig. 1B-E).
The concentrations of biotin required to achieve half-maximum repression at equilibrium (R bio ) were calculated for each protein (Table 1). For E. coli, half maximal repression was achieved with 3.8 nM biotin. Likewise, biotin-induced repression was similar for all three S. aureus target promoters, with 7-10 nM biotin required for the same level of response (Table 1). Consistent with previous literature, the EcBirA-R119W mutant that lacks homo-dimerisation activity 22,31 was also devoid of repressor activity in vivo (Fig. 1B, green symbols) yielding www.nature.com/scientificreports www.nature.com/scientificreports/ similar levels of β-galactosidase expression as the no-repressor control (Fig. 1B, grey symbols). In contrast, SaBirA-F123G repressed activity against all three target promoters in a biotin-dependent manner ( Fig. 1C-E, purple curves). For SabioO and SayhfS-yhfT, the R bio values were 2.6-fold and 2.3-fold higher for SaBirA-F123G than wild-type protein, respectively (Table 1) (p = 0.07 WT vs F123G for SabioO; p = 0.2, WT vs F123G for SayhfS-yhfT). This modest decrease in repressor activity is consistent with a 3-fold higher K M for biotin that has been reported for SaBirA-F123G relative to the wild-type enzyme 29 . For SabioY, an accurate estimate of R bio was not achieved as 500 nM of biotin was insufficient to completely inhibit expression down to the background level necessary to generate a concentration-dependent repression curve. At the lowest biotin concentration tested (1 nM), β-galactosidase expression from the SabioO and SayhfS-yhfT promoters was significantly lower than the www.nature.com/scientificreports www.nature.com/scientificreports/ corresponding no-repressor controls (Fig. 1C,E, wild-type blue and SaBirA-F123G purple vs control grey symbols) implying both the wild-type and SaBirA-F123 mutant proteins partially occupied the DNA under these conditions. Together this data suggests monomeric SaBirA (a mimic of the apo-state) is capable of repressing its target genes, which is not evident with EcBirA.
In vitro analysis of apo and holo-BirA oligomeric states. The molecular basis of the unexpected repressor activity observed with SaBirA-F123G was further investigated using biochemical assays. Wild-type EcBirA and SaBirA were purified in their apo-forms alongside the two dimerization-impaired mutant proteins EcBirA-R119W and SaBirA-F123G. Removal of biotin from the recombinant proteins was achieved by incubating cell lysates (containing over-expressed protein) with streptavidin-Sepharose to remove excess biotin, before incubation with a biotin-accepting substrate protein to facilitate protein biotinylation and the concomitant loss of biotinyl-5′-AMP from the BirA active site prior to IMAC purification. The apo-state of the four proteins was addressed using two alternative techniques that independently confirmed that all preparations were devoid of any co-purified biotinyl-5′-AMP; a streptavidin-blot method we have previously described 29 (Supporting Fig. S3) and native nESI-MS that has not been employed previously for the study of BirA. The Streptavidin-blot assay failed to detect biotinyl-transferase activity for all four proteins, as expected for an apo-enzyme. In support, the masses measured by nESI-MS for the apo-preparations were as expected for monomeric BirA devoid of ligand ( Fig. 2A,C,E and G and Supporting Table S4). We confirmed that the mild conditions employed with the native mass spectrometry studies allowed the direct measurement of BirA in complex with biotinyl-5′-AMP, as well as the oligomeric state of holo-BirA ( Apo and dimerization-impaired SaBirA are competent to bind DNA. Native nESI-MS ( Fig. 3) was employed to probe the DNA binding activities of EcBirA and SaBirA. Oligonucleotides containing the recognition sequences from the EcbioO and S. aureus bioY promoters were employed for the nESI-MS analysis. DNA:protein complexes were formed in 200 mM ammonium acetate pH 6.9 buffer by incubating BirA with the double-stranded oligonucleotides on ice for at least one hour prior to mass spectrometry analysis. A comparison of the theoretical and measured molecular masses for the complexes analysed in this study is shown in Supporting Table S5. The specificity of the BirA:DNA interaction was initially confirmed using an oligonucleotide lacking BirA operators. Neither native nESI-MS nor EMSA detected SaBirA binding to an oligonucleotide with the two BirA half sites mutated to random sequences (Supporting Fig. S4).
For holo-EcBirA, the predominant species had a mass consistent with two subunits of EcBirA bound to DNA. For the apo-EcBirA protein, this species was seen only with weak intensity (Fig. 3A,B). For EcBirA-R119W (at a protein concentration of 10 μM) biotinyl-5′-AMP also induced a complex comprising two protein subunits on DNA, a species which was only observed with a weak intensity for apo-EcBirA-R119W (Fig. 3C,D). At equilibrium, the majority of the apo-wild-type EcBirA and EcBirA-R119W was present as monomeric protein, not in complex with DNA. Consistent with the nESI-MS datum, DNA binding was also observed in an EMSA for wild-type holo EcBirA (Fig. 4A). However, holo-EcBirA-R119W mutant binding to DNA could not be measured by EMSA, even at 400 nM BirA (Fig. 4B).

Wild-type
Dimerization-impaired mutant SayhfS-yhfT 10.0 ± 2.0 22.6 ± 9.7 Table 1. In vivo equilibrium binding constants for biotin-induced transcriptional repression. The concentration of biotin required to achieve half-maximum repression (R bio ) at equilibrium in the β-galactosidase reporter assay was calculated from the data presented in Fig. 1B-E. Data was fitted using the one-site specific binding www.nature.com/scientificreports www.nature.com/scientificreports/ For the SaBirA, evidence of two protein subunits in complex with DNA was observed for wild-type and mutant SaBirA-F123G, in both apo and holo-states ( Fig. 3E-H). In both apo-samples minimal free protein was observed, implying most of the SaBirA was in a complex with DNA. This was in contrast to apo-EcBirA where free protein was readily detected. This datum was consistent with EMSA results where both holo-wild-type and holo-SaBirA-F123G were competent to bind DNA (Fig. 4C-H). The EMSA analysis also revealed the holo-wild-type SaBirA binds all three target promoters at similar affinities ( Fig. 4C-E). However, the interaction between holo-SaBirA-F123G and Sa yhfS-yhFT was clearly weaker compared to the SabioO and SabioY operator sequences (Fig. 4E-G) which is likely due to a base pair mismatch in the BirA binding site 26 . Together these findings highlight a key difference between the E. coli and S. aureus BirAs -monomeric (i.e. apo) SaBirA is capable of stable binding to DNA. www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ Ordered assembly of BirA on DNA. The ordered assembly of SaBirA on DNA was finally addressed by native nESI-MS using a SabioO oligonucleotide that contained a single half site (Fig. 5, Supporting Table S7). The apo-SaBirA sample revealed a species consistent with one monomer bound to DNA with a charge state distribution centered around 13+ and 14+ (Fig. 5A). However, no DNA-bound dimer was observed. In contrast, holo-SaBirA revealed free dimer and a species consistent with two holo-SaBirA subunits in complex with DNA (Fig. 5B). Therefore, under high biotin conditions holo-SaBirA can dimerise prior to binding DNA, such as occurs with EcBirA. Like the wild-type protein, apo-SaBirA-F123G was also competent to bind DNA as a single BirA subunit (Fig. 5C). Holo-SaBirA-F123G was able to bind as a single subunit bound to DNA and well as a complex of two protein subunits bound to the mutated SabioO oligonucleotide (Fig. 5D). This datum suggests that once a single subunit of SaBirA is bound, a second subunit can subsequently bind and stabilize the complex. This DNA-induced protein dimerization can proceed despite the inability of SaBirA-F123G to dimerise in solution. EMSA analysis did not detect holo-SaBirA binding to the mutated oligonucleotide containing only a single BirA binding site (Supporting Fig. S6). It is likely that the experimental conditions of an EMSA, whereby a protein:DNA complex must endure slow (minutes to hours) electrophoretic separation, may not favour the relatively short lived interactions that native nESI-MS can measure on the millisecond timescale. Thus, the native nESI-MS data strongly suggests a pathway where SaBirA dimers can assemble on the DNA one subunit at a time.
Bioinformatic analysis of S. aureus BirA sequences. Several groups have now reported studies characterising BirA from S. aureus 24,25,27,29,34 . However, these studies have used genes encoding birA that have been cloned from different strains bearing different sequences making it difficult to directly compare the findings between studies. It was necessary to identify a consensus SaBirA sequence as detailed studies on the E. coli homologue have demonstrated that substitution of certain amino acids can disrupt networks of long-range bonding interactions that can effect homodimerization, DNA binding and transcriptional repression despite not being localized to the dimer interface nor DNA binding domain (Fig. 6A) 18,19,21,35 . To identify the most conserved BirA homolog amongst S. aureus, BirA protein sequences were obtained using available genomic datum. A multiple sequence alignment was performed upon 26 non-redundant SaBirA sequences (Supporting Fig. S7), including Mu50 (used in this study and in 24,29,34 ), Newman (used in the experiments reported in 25,27 ) and the well-studied NCTC 8325 strain. The alignment indicated that the sequence of Mu50 BirA is highly represented amongst the S. aureus genomes analysed. The Mu50 sequence differs from both the Newman and NCTC 8325 strains, specifically at positions 247 (R247 in Mu50, I247 in Newman and NCTC 8325) and 272 (I272 in Mu50, T272 in Newman and NCTC 8325) (Fig. 6B,C). Out of the 26 unique sequences analysed, 20 possessed arginine at 247 (conserved with Mu50) and 22 species had isoleucine at position 272 (conserved with Mu50). These residues are located within the central catalytic domain of SaBirA, within α-helical structures (R247 in α-helix 8 and I272 in α-helix 9, Supporting Fig. S8). Neither residue is implicated in interactions with the substrates, biotin or ATP, nor are they located close to the dimerization interface. However, it is possible these residues may influence long range bonding interactions that can influence protein dimerization and DNA binding. Noteworthy is a tyrosine at position www.nature.com/scientificreports www.nature.com/scientificreports/ 182 that was conserved in 25 strains including Mu50 and Newman, but which is a phenylalanine in NCTC 8325. Recent studies have reported that substitution of the equivalent tyrosine in EcBirA (Y178) disrupts an electrostatic network of binding interactions that alters protein dimerization 18,19 . As the Mu50 sequence employed here is the most highly representative homologue amongst S. aureus, this sequence should be considered the prototypical SaBirA in future experiments.

Discussion
In certain bacteria BirA is a bifunctional protein that is both a biotin protein ligase and transcriptional repressor of biotin biosynthesis. The biosynthesis of biotin is metabolically costly, requiring the activity of at least four gene products and 20 ATP equivalents per biotin molecule 36 . Therefore, the activity of BirA allows cells to find the necessary balance between cellular demand, supply from the environment and de novo synthesis. Here we study the BirA protein from two clinically important bacteria, namely E. coli and S. aureus. E. coli is considered the prototypical bacteria, and its BirA protein has been the subject of many biochemical, structural and genetic studies. The S. aureus homologue has not been investigated in as much detail. The two bacteria colonize different micro niches where the availability of environmental biotin is distinctive. E. coli is a part of the intestinal microflora which contributes to the synthesis and release of substantial amounts of biotin. Recent mouse studies estimate the concentration of biotin in the ileum to be 450-700 nM 37 . Conversely, S. aureus possess the extraordinary ability to colonise a wide range of niche microenvironments and is responsible for various disorders affecting the skin, respiratory organs, soft tissues, bones and joints. The bioavailability of biotin at these sites is lower than in the intestine. Accordingly, we propose that S. aureus BirA has evolved unique properties that allow the bacteria to adapt to low biotin environments. In this study, we performed detailed in vivo and in vitro analyses to obtain new insights into the transcription regulation mechanisms of SaBirA. We utilized the SaBirA from the Mu50 methicillin and vancomycin resistant strain of S. aureus 38 . Our bioinformatics analysis has demonstrated this particular variant of SaBirA represents the most conserved SaBirA sequence among S. aureus strains. Furthermore, Mu50 SaBirA has been the target for the discovery of new classes of antibiotics designed to treat MRSA infections 23,[39][40][41] . We also performed our study alongside the well-characterized BirA from E. coli, in order to highlight the differences in dimerization and DNA-binding mechanisms between the two homologues. www.nature.com/scientificreports www.nature.com/scientificreports/ Previous studies performed on Mu50 SaBirA revealed weak dimerization of the apo-form of this protein (K D 2-1 = 29 ± 0.2 μM) 29 suggesting that in its apo-state, this protein is likely to be monomeric within the intracellular environment. Hence, we proposed that SaBirA-F123G, as a dimerization-deficient mutant in solution, would mimic the oligomeric state of this apo-SaBirA in vivo. Whilst the analogous E. coli mutant was clearly devoid of DNA-binding activity in vivo, as reported in the literature 31 23 ), or that SaBirA partially occupies the DNA even when biotin is limiting, and therefore, the protein is in the apo-state.
In vitro analysis supported the contention that the intrinsically monomeric SaBirA-F123G or wild-type apo-SaBirA can bind DNA, enabling in vivo repression. The Mu50 strain wild-type SaBirA used here was unable to dimerize in solution at a protein concentration of 10 µM in its apo-form, as indicated by native nESI-MS. This www.nature.com/scientificreports www.nature.com/scientificreports/ finding agrees with our previous work which has demonstrated the apo-protein to be monomeric by small angle X-ray scattering assay 29 and crystallises as a single subunit 24 . It also agrees with recent work conducted on SaBirA from the Newman strain reported by Wang and Beckett 25 where analytical ultracentrifugation (AUC) failed to detect homodimerization. This is in contrast to our previous AUC datum where dimerisation of apo-SaBirA from Mu50 was observed 29 . It should be noted that different protein sequences were employed between these two studies (discussed further below). Additionally, this may have been the result of contaminating co-repressor, biotinyl-5′-AMP, that was not detected by the Streptavidin blot method used to assay the purified protein 27 . Here, a superior technique of native nESI-MS has been utilized to unambiguously detect not only the oligomeric state of the protein, but to simultaneously determine whether these species are truly devoid of bound ligands. This is the first report where native mass spectrometry has been employed to study BirA. No dimeric form of SaBirA-F123G was observed in either the apo or holo forms in solution, supporting its characterization as a monomeric mutant.
Furthermore, our results suggest dimerization of apo-SaBirA and SaBirA-F123G is mediated by DNA binding, as native nESI-MS detected species with molecular weights equivalent to two monomers of SaBirA in complex with the SabioY recognition sequence oligonucleotide. Likewise, nESI-MS revealed wild-type and SaBirA-F123G bound a DNA probe containing only a single half site in both the apo and holo-states but with different stoichiometry equivalents. Whilst the apo-proteins only bind as one subunit, biotinyl-5′-AMP allows the dimerization of wild-type SaBirA prior to DNA binding but also allowed SaBirA-F123G to dimerise on the DNA, despite not detectably dimerising in solution. The binding of SaBirA-F123G to DNA agrees with previous studies on apo-SaBirA where the K D for DNA binding was 6-fold lower compared to holo-SaBirA as measured by EMSA 29 , and 60-fold lower as measured by fluorescence anisotropy 27 .
By combining our in vitro and in vivo data, we propose two possible pathways for SaBirA to assemble on DNA, as illustrated in Fig. 7. In high biotin conditions, protein dimerization occurs in solution and precedes DNA binding, as has been well documented for EcBirA 5,7 . The pre-formed homodimer orients the two helix-turn-helix motifs present in the N-terminus of the protein such that they can simultaneously occupy both half sites on the target operator. Here we present evidence that binding to DNA can also proceed via the sequential assembly of two monomers onto the DNA, a pathway which is much less preferred for EcBirA. This alternative pathway is likely be the basis for the lower biotin threshold required to initiate the transcriptional repression observed with SaBirA compared to EcBirA, as highlighted in a previous study 26 and seen with the partial occupation of SaBirA on DNA causing repression at low biotin concentrations (1 nM) in the in vivo reporter assay 26 . Whilst the DNA mediated dimerization pathway has been observed with the SaBirA from the Mu50 strain, which we have identified here as the prototypical example in S. aureus, it remains to be seen whether BirA from other species behave in a similar manner. The findings from this study may provide new insights into how biotin-regulated gene expression occurs in S. aureus allowing it to adapt to the low biotin environments it colonises.
Our bioinformatics analysis has identified multiple polymorphisms amongst S. aureus strains which may give rise to BirA variants with different homo-dimerization and DNA binding activities. Mutagenesis studies that target specific amino acids in EcBirA, together with structural biology and biophysical assays, are helping to understand mechanisms of allosteric communication between distant sites in the protein and the impact upon their function 18,21,35 . Noteworthy is mutation of tyrosine 178 to cysteine (equivalent to Y182 in S. aureus Mu50, and F182 in NCTC 8325) that results in a super-repressor phenotype in whole cell reporter assays of transcription 18 . This amino acid does not reside directly in the dimer interface but, instead, contributes to complex network of electrostatic interactions that are proposed to influence the structural alignment of the central (ligand-binding) and N-terminal (DNA-binding) domains of BirA that effect homo-dimerization and DNA binding 19 . Examination of available crystal structures of the Mu50 SaBirA 23,24,29 likewise revealed that Y182 contributes to a network of electrostatic interactions involving K176 (that resides in a conserved KWPND motif in the active site) and D322 that resides in the DNA binding domain. Our analysis of genomic information also identified two differences in the BirA proteins from Mu50 and Newman at positions 247 and 272. Mutational studies, such as those performed to characterise the EcBirA mutants 18,19 , are required to probe the structure-function relationships in more detail. We propose that the SaBirA variants identified here present a useful system with which to further probe long distance modulation of protein allostery. www.nature.com/scientificreports www.nature.com/scientificreports/ Methods Materials. E. coli strain JD26186 (bioC::Kan), based on host strain KP7600, is a transposon-disrupted bioC biotin auxotroph, and was obtained from the National BioResource Project (NIG, Japan). Bacteria were cultured at 37 °C with vigorous shaking in Luria Bertani media containing the appropriate antibiotic. Plasmid extractions were performed using the Plasmid QIAprep Spin Miniprep Kit (Qiagen) and genomic extractions were performed using the Wizard ® Genomic DNA purification kit (Promega). All molecular biology enzymes (DNA polymerase, ligases and restriction enzymes) and buffers were supplied by New England Biolabs. Oligonucleotides, purchased from Geneworks Ptd Ltd, are shown in Supporting Table S1.
Nucleic acid manipulations. All plasmids employed in this study are shown in Supporting Table S2.
Generation of bacterial strains for in vivo reporter assays. All strains of E. coli employed in this study are shown in Supporting Table S3. β-galactosidase reporter assays were performed on E. coli reporter strains generated in 26 . Both the β-galactosidase reporters and the BirA expression modules were site-specifically integrated in the chromosome in single copy (Fig. 1A). BirA binding sequence was cloned into pLacZ_SHTrim plasmid upstream of the β-galactosidase reporter gene. BirA expression (EcBirA, SaBirA and their dimerization defective mutants EcBirA-R119W and SaBirA-F123G) was under control of the placUV5 promoter. These were constructed using the methods outlined in 26,44 . The β-galactosidase reporter assays were conducted as described 43,45 . Protein methods. Recombinant expression of all BirA proteins employed in this study, as well as their purification by immobilized metal ion affinity chromatography (IMAC), were performed essentially as previously described 29,34 . However, the following modifications were included to ensure removal of biotinyl-5′-AMP from the protein preparations. Prior to IMAC purification, the cell lysate was incubated at 4 °C for a minimum of 1 hour with 10 μL 50% slurry Streptavidin-sepharose High Performance (GE Healthcare Life Sciences) per millilitre of lysate. This was centrifuged at 4000 rpm, 4 °C for 10 minutes, to remove the resin. The cleared lysate was then incubated for 1 hour at 37 °C with 6 mg biotin-accepting substrate GST-SaPC90 per 10 ml of lysate prior to IMAC. Purified proteins were dialyzed overnight in 4 L of storage buffer (50 mM Tris pH 8.0, 100 mM KCl, 1 mM DTT, 5% (v/v) glycerol) and stored at −80 °C until required. Protein SDS-PAGE was performed using NuPage TM Bis-Tris 4-12% gels (Invitrogen).
Native nano-electrospray ionisation-mass spectrometry (nESI-MS). Purified BirA proteins were buffer exchanged into 200 mM ammonium acetate pH 6.85 (Sigma) using Amicon Ultra-0.5 MWCO 10,000 centrifugal filter units (Merck Millipore). Holo-enzyme samples (at least 100 µM, except for EcBirA-R119W that was 30 µM) were prepared by pre-incubating the apo-proteins with 500 µM biotin, 1 mM ATP and 1 mM MgCl 2 on ice for at least 1 hour prior to buffer exchange. HPLC-purified and annealed, double stranded oligonucleotides containing the operator sequences of interest were purchased from Integrated DNA Technology (USA). These were desalted into 200 mM ammonium acetate using Illustra MicroSpin G-25 columns (GE Healthcare). Oligonucleotide concentrations were quantified using a Nanodrop spectrophotometer (Thermo Fisher Scientific). Proteins were quantified following buffer exchange via Bradford assay, and then diluted to 10 µM in 200 mM ammonium acetate for analysis by nESI-MS. Protein:DNA complexes were analyzed by nESI-MS after diluting the oligonucleotide and protein in 200 mM ammonium acetate pH 6.9 to 5 µM and 10 µM respectively, and incubating at 4 °C for at least 1 hour prior to MS analysis.
Mass spectrometry (MS) measurements were performed on a Synapt HDMS system (Waters, UK) with the sample introduced by nano-electrospray ionization in the positive ion MS mode from platinum-coated borosilicate capillaries prepared in-house. Instrument parameters were optimized to remove adducts while preserving non-covalent interactions, and were as follows; capillary voltage, 1.5 kV; cone voltage, 60 V; trap collision energy, 20 V; transfer collision energy, 15 V; source temperature, 50 °C; backing pressure, 3.95 mbar. Data analysis was performed using manual peak finding in MassLynx software (version 4.1).
Electrophoretic mobility shift assay (EMSA). Binding reactions were performed at room temperature for 30 minutes using EMSA buffer (50 mM Tris pH 8.0, 50 mM NaCl, 0.1 mM biotin, 1 mM ATP, 1 mM MgCl 2 and 10% (v/v) glycerol) together with 10 nM of the double stranded oligonucleotide (HPLC-purified as utilized for nESI-MS) and varying concentrations of BirA. Fractionation of the samples was performed using 4-12% TBE polyacrylamide gels (Life Technologies) run at 100 volts (constant) for 45 minutes and stained in GelRed (Biotium) solution for 5 minutes. After washing five times in distilled water, the gels were imaged using a ChemiDoc imager (Bio-Rad). Each EMSA experiment was performed in triplicate.

S. aureus
BirA multiple sequence alignment. BirA sequences of S. aureus were compared as follows: the protein cluster PCLA_885364 (accessed on 15/5/17) containing 71 entries (which encompasses all non-redundant proteins identified as biotin-acetyl-CoA-carboxylase ligase in the Genus Staphylococcus) was obtained from NCBI 46,47 and further filtered for the clade Staphylococcus aureus 19988 to make a list of 26 unique protein sequences. These sequences were then aligned using Clustal Omega 48 . The alignment was then compared using UGENE 49 and the position of the polymorphisms were mapped onto the SaBirA structure (PDB accession