Protein kinase B controls Mycobacterium tuberculosis growth via phosphorylation of the global transcriptional regulator Lsr2

Mycobacterium tuberculosis is able to persist in the body through months of multi-drug therapy. Mycobacteria possess a wide range of regulatory proteins, including the essential protein kinase B (PknB), that control transitions between growth states. Here, we establish that depletion of PknB in replicating M. tuberculosis results in transcriptional adaptations that implicate the DNA-binding protein Lsr2 in coordinating these changes. We show that Lsr2 is phosphorylated by PknB, and that phosphorylation of Lsr2 at threonine 112 is important for M. tuberculosis growth and survival under hypoxic conditions. Fluorescence anisotropy and electrophoretic mobility shift assays demonstrate that phosphorylation reduces Lsr2 binding to DNA, and ChIP-sequencing confirms increased DNA binding of a phosphoablative (T112A) Lsr2 mutant in M. tuberculosis. Altered expression of target genes in T112A Lsr2 compared to wild type Lsr2 M. tuberculosis offers further evidence that phosphorylation mediates expression of the Lsr2 regulon. Structural studies reveal increased dynamics of the Lsr2 DNA binding domain from a T112D phosphomimetic Lsr2 mutant, providing a molecular basis for decreased DNA binding by phosphorylated Lsr2. Our findings suggest that, the essential protein kinase, PknB controls M. tuberculosis growth and adaptations to the changing host environment by phosphorylating the global transcriptional regulator Lsr2.


INTRODUCTION
Mycobacterium tuberculosis is a slow-growing bacterium that can replicate in humans and cause tuberculosis. The pathogen is able to rapidly shut-down its growth to persist in nonreplicating states in infected individuals, which can be modelled in the laboratory 1 . M. tuberculosis adaptation to stressful conditions is accompanied by dramatic changes in global protein phosphorylation but the importance of these modifications is poorly defined 2 . M. tuberculosis has eleven serine/threonine protein kinases which play significant roles in growth, virulence and metabolism 3 . In particular, protein kinase B (PknB) is indispensable for M. tuberculosis growth due to its critical role in the regulation of peptidoglycan biosynthesis 4,5 . It is also important for M. tuberculosis survival in hypoxic conditions and resuscitation during reaeration 6 . However, the molecular mechanism for PknB-mediated adaptation to hypoxia is unknown.
We have recently shown that PknB-depleted M. tuberculosis can grow in a sucrose magnesium medium (SMM) 5 . Comparative phosphoproteomics of PknB-producing and PknB-depleted mycobacteria revealed substantial changes in profiles of phosphorylated proteins. Specifically, the transcriptional regulators Lsr2 and EthR showed increased phosphorylation in PknB-producing compared with PknB-depleted mycobacteria, indicating that these proteins may be PknB substrates. EthR is a TetR family transcriptional regulator that controls expression of EthA, a monoxygenase responsible for activation of the prodrug ethionamide 7 . Protein kinase F was found to phosphorylate EthR in vitro, which negatively regulated EthR binding to DNA 8 .
Lsr2 is a DNA binding protein that combines properties of a nucleoid associated protein 9 and a global transcriptional regulator 10 . Lsr2 has over 1000 binding sites in M. tuberculosis 11,12 . The precise role of Lsr2 in mycobacterial biology remains unclear, nevertheless parallels may be drawn with similar proteins from other bacteria. Lsr2 represents the first example of an H-NS-like protein identified outside Gram-negative bacteria, moreover lsr2 was able to complement the hns mutant in Escherichia coli 13 . Similar to the H-NS proteins, Lsr2 has been proposed to bind to the minor groove of DNA 14 and to possess DNA bridging properties 15 . Deletion of lsr2 in M. tuberculosis resulted in severe growth impairment on solid media, defects in persistence and adaptation to changing oxygen levels 10 . Additionally, Lsr2 has been shown to help protect DNA from reactive oxygen species, and over-expression of Lsr2 improved survival of mycobacteria treated with hydrogen peroxide 16 . However, deletion of lsr2 did not alter survival of M. tuberculosis exposed to hydrogen peroxide or mitomycin C 10 . Transcriptional profiling of an lsr2 deletion mutant revealed global changes involved in cell wall remodelling, respiration and lipid biosynthesis 10 .
Here we show that recombinant PknB phosphorylated Lsr2 at T112, a site that we identified previously by a proteomic approach, implying that PknB is directly responsible for controlling phosphorylation at this site. We demonstrate that phosphorylation of Lsr2 at threonine 112 is important for M. tuberculosis growth and adaptation to hypoxic conditions.
Our results show that phosphorylation of Lsr2, or phosphomimetic T112D mutation, reduce Lsr2 binding to DNA, which is explained by increased dynamics in the DNA binding domain of phosphorylated Lsr2. Chromatin immunoprecipitation and transcriptomics further confirm that the T112A phosphoablative mutant of Lsr2 has an increased number of Lsr2-DNA binding events and modified expression of genes involved in M. tuberculosis stress responses and virulence. Based on our data, we propose that PknB-mediated phosphorylation controls Lsr2 binding to DNA in M. tuberculosis, providing a functional link between serine/threonine protein kinase signalling in replicating bacilli and regulatory networks that enable M. tuberculosis to survive dynamic environments during infection. tuberculosis (pknB-CM) 17 in sucrose-magnesium medium (SMM) to support growth 5 . In total 99 genes were differentially expressed: 65 were induced and 34 repressed by PknBdepletion (Table S1). Two functional classes were identified in the genes overexpressed by PknB depletion: regulatory proteins and proteins involved in lipid metabolism. The predicted transcriptional regulators were csoR, rv1129c, rv1460, rv2017, rv2250c, rv3334, sigB, whiB3 and whiB6 (Table S1). These transcription factors have been shown to regulate copper homeostasis (CsoR) 18 , iron-sulphur cluster biogenesis (Rv1460) 19 , cholesterol catabolism (Rv1129c/PrpR) 20 , the enduring hypoxic response (Rv3334) 21 , multiple stress responses (SigB) 22 , redox stress and complex lipid biosynthesis (WhiB3) 23 and virulence factor expression (WhiB6) 24 . The transcriptional signature of PknB-depletion resembled features of intracellular growth, which significantly overlapped with M. tuberculosis macrophage-derived RNA profiles from several studies as reflected by hypergeometric probability values: 6.7x10 -23 25 ; 7.34x10 -18 26 ; 3.57x10 -17 27 (Fig. 1). This profile was exemplified by induction of pathways involved in mycobactin synthesis (mbtB/C/D), complex lipid phthiocerol dimycocerosate (PDIM) biosynthesis (fadD26, ppsA/B/C/D), metabolism of alternative lipid carbon sources, the glyoxylate shunt (icl), the methylcitrate cycle (prpD/C, prpR) and triacylglycerol synthase (tgs1). The isoniazid inducible genes (iniB/A/C), responsive to cell wall stress and cell wall targeting drugs were also upregulated 28 . Four of the nine genes coding for alternative ribosomal proteins, rpmB1, rpmB2, rpmG1, rpsN2 29 were induced by PknB depletion.

RESULTS
The 34 genes that were significantly repressed in PknB-depleted bacteria included pknB itself (6 fold down-regulated, as expected); nuoA/B/C, encoding subunits of NADH dehydrogenase I, which is part of the aerobic respiratory chain, and several genes involved in intermediary metabolism (Table S1). Overall these changes in replicating bacteria were comparable, in number of differentially expressed genes, to depletion of other regulators, for example DosR 30 . This is in contrast to the large-scale changes in gene expression after treatment with an inhibitor of PknB and PknA that would likely impact M. tuberculosis viability 31 . In summary, PknB depletion in replicating bacteria resulted in co-ordinated changes to the transcriptome, suggesting that PknB may control the induction of alternative gene regulatory pathways.
Application of the Transcription Factor Over-Expression (TFOE) output tool 32 predicted Rv0081 33 , DosR 30 and Lsr2 10 as potential regulators of the observed gene expression patterns (Fig. 1). Rv0081 is a transcriptional hub which coordinates expression of over 50 transcription regulators controlling metabolic adaptation of M. tuberculosis 33 . However, to date there are no reports of Rv0081 regulation by phosphorylation. DosR has been shown to be phosphorylated by PknH 34 but it was not under-phosphorylated during PknB depletion 5 .
According to our data, PknB changed Lsr2 phosphorylation without impacting Lsr2 protein expression levels 5 , suggesting that the transcriptional changes between PknB-depleted and replete M. tuberculosis were likely caused by changes in PknB-mediated phosphorylation of Lsr2. We therefore investigated whether PknB could phosphorylate Lsr2 in vitro.

PknB phosphorylates Lsr2 in vitro
Purified Lsr2 was mixed with the recombinant kinase domain of PknB and incubated at 37°C with or without ATP; phosphorylation was detected using western blotting with an antiphosphothreonine antibody. Phosphorylated-Lsr2 was detected in the ATP-containing sample, demonstrating that PknB could indeed phosphorylate Lsr2 ( Fig. 2A). Interestingly, phosphorylation resulted in a marked change in Lsr2 protein mobility on SDS-PAGE (  (Table S2) and investigated their phenotypes.

Phosphorylation of Lsr2 on threonine 112 is necessary for M. tuberculosis growth
Lsr2 deletion significantly impaired M. tuberculosis growth on solid media and the reintroduction of lsr2 on the pMV306 integrative plasmid (Δlsr2 WT ) into the lsr2 deletion mutant fully complemented this defect ( Fig. 3A & Fig. S1), similarly to a previous study 10 . An Lsr2 deletion mutant containing the empty pMV306 plasmid (Δlsr2 pMV306 ) had the same growth pattern as the lsr2 deletion mutant (Fig. 3A), as expected. Lsr2 variants lacking the phosphorylation sites T8A, T22A and T31A individually restored bacterial growth and colony size to the same extent as wild type Lsr2 (Fig. S1). The T112A Lsr2 mutant (Δlsr2 T112A ), however, failed to complement the Δlsr2 phenotype (Fig. 3A, Fig. S1). Replacement of threonine 112 with an aspartate to mimic phosphorylation (Δlsr2 T112D ) complemented the Δlsr2 phenotype (Fig. 3A).
We next investigated whether T112 mutations of Lsr2 influenced M. tuberculosis growth in liquid culture. As Fig. 3B demonstrates, Δlsr2 pMV306 and Δlsr2 T112A had significant growth defects compared with Δlsr2 WT and Δlsr2 T112D . Growth rates for Δlsr2 pMV306 and Δlsr2 T112A were 0.0157±0.03 h -1 and 0.0173±0.004 h -1 , compared with 0.0414±0.004 h -1 and 0.0385±0.009 h -1 for Δlsr2 WT and Δlsr2 T112D . Thus, the T112A mutant failed to complement the growth defect of Δlsr2 both in liquid and on solid media, suggesting that phosphorylation of Lsr2 at T112 is important for M. tuberculosis growth. Furthermore, western blot analysis using anti-Lsr2 antibody confirmed that wild type, T112A and T112D Lsr2 proteins were produced at similar levels, excluding the possibility that the Δlsr2 T112A phenotype was caused by altered Lsr2 expression (Fig. 3C). The phosphomimetic T112D Lsr2 had lower mobility in SDS PAGE, similar to phosphorylated Lsr2 ( Fig. 3C and Fig. 2B).

Lsr2 phosphorylation at threonine 112 is crucial for survival in hypoxic conditions but not in prolonged stationary phase
We hypothesised that regulation of Lsr2 by PknB could account for the defects in survival of oxygen depletion seen when pknB and lsr2 were perturbed 6,10 . Thus, we assessed survival of M. tuberculosis carrying Lsr2 variants using the Wayne model of non-replicating persistence in hypoxia 35 . The viable counts of Δlsr2 pMV306 dropped dramatically after 7 weekincubation compared to Δlsr2 WT and Δlsr2 T112D (Fig. 3D). Further incubation for 24 weeks resulted in near-complete loss of recoverable Δlsr2 pMV306 bacteria but did not significantly alter the survival of Δlsr2 WT and Δlsr2 T112D . Interestingly, T112A Lsr2 could only partially complement the Δlsr2 survival defect, suggesting a requirement for T112 phosphorylation in microaerobic conditions. After 24 week-incubation, viable counts of Δlsr2 pMV306 and Δlsr2 T112A were almost undetectable (Fig. 3D).
We also assessed the survival of M. tuberculosis expressing Lsr2 variants in late stationary phase of aerobically grown cultures by MPN and CFU counting. Δlsr2 pMV306 showed impaired survival compared to Δlsr2 WT , resulting in a 1.5 order of magnitude difference in viable counts. In this model, the survival of T112A or T112D variants was not significantly different from wild type Lsr2 (Fig. S1B). Our results suggest that Lsr2 phosphorylation of T112 may be specifically required during adaptation to hypoxic conditions but not for survival in prolonged stationary phase.  Table S3). The putative regulon of Lsr2, including genes with an Lsr2 binding site within or immediately upstream of the coding sequence, was defined as 1178 genes (Fig. 4A, Table S3). These genes, potentially directly regulated by Lsr2, are consistent with previously identified Lsr2 binding patterns 14 tuberculosis, we compared phosphomimetic (Δlsr2 T112D ) and phosphoablative (Δlsr2 T112A ) to wild type Lsr2 (Δlsr2 WT ). There were 226 binding events with significantly greater abundance in Δlsr2 T112A in comparison to Δlsr2 WT M. tuberculosis (Fig. S2, Fig. 4A). This potentially affected the expression of 94 genes (Table S4), including pathways involved in cell wall biosynthesis, lipid metabolism, PE/PPE protein synthesis and intermediary metabolism. No binding events were overrepresented in Δlsr2 T112D compared to Δlsr2 WT . Thus, our results suggest that while phosphomimetic Lsr2 (Δlsr2 T112D ) behaved similarly to wild type, phosphoablative Lsr2 (Δlsr2 T112A ) bound DNA to a greater extent.
To probe whether altered binding of Lsr2 variants influenced gene expression we measured the expression of a panel of genes with predicted Lsr2 binding sites in Δlsr2 pMV306 , Δlsr2 WT and Δlsr2 T112A genetic backgrounds. These genes included leuS, rpfA, rpfC sigA, and sodA. According to previously published data rpfA, rpfC and sodA were significantly downregulated in the lsr2 deletion mutant 10 , while expression of sodA and leuS were predicted to be differentially regulated in Δlsr2 T112A in our ChIP-Seq analysis. Lastly, sigA encoding a housekeeping sigma factor, has a putative Lsr2 binding site in the promoter region 36 , which could influence its expression. Fig. 4B shows that 4 genes rpfA, rpfC, sigA and sodA were indeed up-regulated in Δlsr2 WT compared with Δlsr2 pMV306 . However, only expression levels of leuS and sodA, with divergent Lsr2 binding events, were statistically significantly different in Δlsr2 T112A compared with Δlsr2 WT . Hence, our results confirm that the phosphorylation state of Lsr2 may influence gene expression patterns.

PknB-mediated phosphorylation and T112D mutation reduce Lsr2 binding to DNA in vitro
We further investigated the impact of phosphorylation on Lsr2 binding to DNA in vitro using Electrophoretic Mobility Shift Assays (EMSA) with a range of DNA fragments containing putative Lsr2 binding sites. Lsr2 was able to shift all DNA fragments when added at concentrations above 1.9 μM; there was no significant difference in binding patterns between the DNA fragments (Fig. S3). For example, previously published data 36 indicated a putative Lsr2 binding site in the leuS coding region (5'-AATTCGGCAAAATCGGTAAG-3' marked with an asterisk in Fig. 4A); this fragment was shifted in EMSA by Lsr2 (Fig. 5A). A DNA fragment containing a mutated Lsr2 binding site where the central AT rich region was replaced with a GC-rich sequence (5'-AACTCGGCGAGGTCGGTCAG-3') was also shifted by Lsr2 (Fig. 5B). We tested other DNA fragments and all were shifted by Lsr2 regardless of their sequence, demonstrating that Lsr2 can bind non-specifically to DNA as previously suggested 28 and in agreement with our ChIP-Seq analysis.
We next investigated the effect of Lsr2 phosphorylation on binding to the leuS site. As Previous studies demonstrated that Lsr2 BD mainly recognised AT-rich DNA sequences that formed a hook-like structure 14 . We measured Lsr2 BD and Lsr2 BD T112D binding to an Alexa 488N labelled AT-rich DNA fragment (5'-CGCGCATATATGCG-3'), previously identified by Gordon and colleagues 14 (Fig. 5E). At pH 7.5 the Kd value for Lsr2 BD was 0.21±0.06 µM, while the Kd value for Lsr2 BD T112A was 0.02±0.008 µM and the Lsr2 BD T112D showed no significant DNA binding (Kd could not be determined). Next, we tested binding of Lsr2 BD to the wild type and mutated leuS sequences where the central AT rich region was substituted by a GC-rich sequence (as described in the previous section).
The binding of Lsr2 BD to the leuS sequence was characterized with a Kd of 0.4±0.12 µM (Fig. 5 F). No binding was observed for Lsr2 BD to the GC-mutated sequence (Fig. 5G).
Lsr2 BD T112D did not bind to either leuS sequences (Fig. 5F-G). These results show that T112D mutation, mimicking the Lsr2 phosphorylated state, reduced Lsr2 BD T112D binding to DNA in vitro. Hence, phosphorylation controls both Lsr2 and Lsr2 BD binding to DNA.

Phosphomimetic T112D mutation changes conformation entropy of the Lsr2 DNA binding domain
To elucidate a molecular basis for the reduced binding of phosphomimetic Lsr2 to DNA we compared nuclear magnetic resonance (NMR) structures of Lsr2 BD WT and Lsr2 BD T112D. In accordance with the previously published structure 11 , our data confirmed that Lsr2 BD consists of two perpendicular α -helices (α1, residues 78-89; α 2, residues 102-112) linked by a long loop (residues 90-101), and this organisation was preserved in Lsr2 BD T112D (Table S5). However, T112D mutation resulted in a shorter α 2 helix (Fig. S4), which ended with alanine 111 in Lsr2 BD T112D compared with threonine 112 in Lsr2 BD WT ( Fig. 6A and   6B). Our data suggest that both proteins were in monomeric states and the N-terminal segment (residues 66-75) upstream of the DNA binding domain was disordered. We also found that in Lsr2 BD the methyl group of threonine 112 interacted with the tyrosine 108 and the hydroxyl group interacted with tryptophan 86 (Fig. 6C). These interactions did not form in the Lsr2 BD T112D; however, they are unlikely to be important for DNA binding since T112A was able to bind DNA in our anisotropy experiments (Fig. 5E). 15 N heteronuclear NMR relaxation analysis was performed to assess the dynamic behaviour of the two proteins (Fig. S5). In both isoforms two α helices showed similar amplitudes for internal motion. However, in Lsr2 BD mainly the C-terminal helix was affected, while in Lsr2 BD T112D the internal motion was extended to the N-terminal helix. A previous study on the catabolite activator protein (CAP) demonstrated that different protein mutants with the same structure of interaction interface displayed very different affinity for their target DNA 38 . The authors showed that changes of the binding affinity were linked to fast internal dynamics (conformation entropy). Similarly, the T112D mutation and, presumably, phosphorylation of threonine 112, resulted in a shorter α 2 helix and a more mobile loop which increased the Lsr2 BD dynamics and impaired DNA binding.

DISCUSSION
M. tuberculosis can subvert the immune system to survive in the host for many years. This remarkable ability is determined by mechanisms that allow M. tuberculosis to respond to multiple environments and adjust metabolic activity and cell division. One of these mechanisms is protein phosphorylation involving eleven serine/threonine kinases, including protein kinase B. PknB is expressed during active replication and is essential for growth in standard media 17 (Table S1).
Here we demonstrate that Lsr2 could be phosphorylated by PknB on several threonines but only the T112 was essential for growth and survival of M. tuberculosis. The phosphoablative T112A mutant could not complement the growth defect of the lsr2 deletion mutant and was impaired in the Wayne hypoxia model. To investigate the molecular mechanism responsible for the growth defect in Δlsr2 T112A we conducted ChIP-Seq analysis and compared DNA binding patterns in Δlsr2 T112A, Δlsr2 T112D and Δlsr2 WT . In agreement with previously published data we detected multiple occurrences of DNA binding in all three Lsr2 backgrounds supporting both nucleoid shaping and gene regulatory functions of Lsr2.
However, phosphoablative Lsr2 mutation T112A resulted in increased DNA binding, potentially directly affecting expression of 94 genes. Interestingly, most of these genes were not essential for growth, moreover deletion of some of these genes has been previously shown to be advantageous for growth 43

Generation Lsr2 complementation mutants
The Rv3597 (lsr2) coding sequence with additional 200 bp upstream region was amplified from the M. tuberculosis H37Rv genome using primers lsr2-pMV306F and lsr2-pMV306R and cloned into the integrative plasmid pMV306. Lsr2 M. tuberculosis variants were obtained using the GeneArt™ Site-Directed Mutagenesis System (Thermo Fisher Scientific) according to the manufacturer's instructions. For generation of T112A and T112D mutants lsr2-pMV306RA and lsr2-pMV306RD primers were used. All constructs were sequenced before transformation into the previously described lsr2 deletion mutant 10 . Bacterial strains generated in this study are shown in Table S2.

Wayne model of non-replicating persistence
The model was setup as previously described 35 . CFU and MPN counts were determined at 0, 7 and 12-week time points. Statistical analysis was performed using one-way ANOVA (GraphPad Prism) where Δlsr2 pMV or Δlsr2 T112A data were compared to Δlsr2 WT and Δlsr2 T112D data.

Transcriptomic analyses
M. tuberculosis H37Rv PknB conditional mutant (pknB-CM) was grown to OD 580nm ~0.7 in SMM with or without pristinamycin. M. tuberculosis RNA was extracted from three biological replicates using the GTC/Trizol method 53 . The RNA samples were purified and DNasetreated using RNeasy columns (Qiagen). M. tuberculosis RNA (2μg) was enzymaticallylabelled with Cy3 fluorophore and hybridised to a M. tuberculosis complex microarray (ArrayExpress accession number A-BUGS-41) as previously described 54 . Significantly differentially expressed genes (Table S1) were identified using a moderated t-test (p-value <0.05 with Benjamini and Hochberg multiple testing correction), and fold change >1.8 in GeneSpring 14.5 (Agilent Technologies). Hypergeometric probability and TFOE analysis 32 was used to identify significantly enriched signatures.
For quantitative RT-PCR, total RNA was isolated from triplicate M. tuberculosis cultures of Δlsr2 T112A, Δlsr2 T112D, Δlsr2 WT and Δlsr2 pMV as described above and cDNA generated using Superscript Reverse Transcriptase II and mycobacterial genome-directed primers 55 . Q-PCR was performed in a Corbett Rotor Gene 6000 real time thermocycler using Absolute QPCR SYBR Green mix and gene expression values were normalised to 16S rRNA expression.
Statistical analysis was performed by one-way ANOVA where Δlsr2 WT was compared to Δlsr2 T112A and Δlsr2 T112D .

ChIP-Seq analysis
DNA-Lsr2 interactions in Δlsr2 T112A, Δlsr2 T112D, Δlsr2 WT were assayed using methods adapted from Minch et al 12  were crosslinked with 1% formaldehyde, followed by incubation with 125mM glycine for 5 minutes at 37°C. The cells were mechanically lysed and then sonicated to produce 200-500 bp fragments. Input control samples were taken for each genotype before antibody was added to assess antibody specificity. Samples were immunoprecipitated using a polyclonal anti-rabbit anti-Lsr2 antibody and protein-G agarose beads. The Lsr2 complexes were decrosslinked by heating at 65°C overnight and proteins removed by treatment with proteinase K (10 mg/ml) for 2h at 55°C. The DNA samples were column-purified (Qiagen) and the quality of purified IP-Lsr2 DNA verified using the Qubit DNA HS quantification assay and Nanodrop spectrophotometer. Libraries were prepared and sequenced using Illumina HiSeq SE50, 20 million reads (Novogene, Hong Kong). Raw fastq files were aligned to the M.

Generation and purification of recombinant proteins
The M. tuberculosis H37Rv lsr2 coding sequence was amplified using Platinum Taq-HF polymerase (primers in Table S6) and cloned into pET15-TEV. After confirmation by sequencing, the constructs were transformed into E. coli OverExpress™ C41 (DE3) competent cells (strain details are summarised in Table S2).
For protein expression, E. coli cultures were grown in LB with ampicillin to mid-log phase (OD 600 0.6-0.8) at 37°C with shaking at 200rpm before adding 0.5mM isopropyl β -D-1thiogalactopyranoside followed by incubation at 18°C overnight. The recombinant Lsr2 proteins were purified using immobilized-metal affinity chromatography (Ni-NTA agarose, Qiagen) and size exclusion chromatography. The recombinant catalytic domain of PknB was purified using Glutathione Sepharose 4B GST-tagged protein purification resin (GE Healthcare) as previously described 5 . were identified in trypsin-digested proteins using a LTQ-Orbitrap-Velos mass spectrometer.

Electrophoretic mobility shift assay (EMSA)
EMSAs were carried out with DNA fragments amplified by PCR or annealed nucleotides (Table S6) as described previously 15 . Briefly, DNA (1.2 nM) was mixed with indicated amounts of Lsr2 in a total volume of 20µl reaction buffer containing (10 mM Tris-HCl, pH 7.5, 50 mM KCl, 1 mM DTT, 5 mM MgCl 2 , and 2.5% glycerol). The mixture was incubated for 30 min at room temperature followed by native polyacrylamide gel electrophoresis using 8% gels in 0.5 x Tris-Borate-EDTA buffer, pH 7.5 for 24 min at 120 V. The gels were stained with SYBR Safe DNA stain (Thermo Fisher Scientific) and visualised using a ChemiDoc system (Bio-Rad).

DECLARATION OF INTERESTS
Authors declare no competing interests.  62 ). Individual ChIP-seq replicates are shown; the Lsr2 regulon was defined to have binding events in 2 of 3 biological replicates. Genes were included in the predicted regulon if the Lsr2 binding site was immediately upstream of, or within, the coding sequence of each gene. For Lsr2 intergenic binding between divergent genes, both genes were reported. For Lsr2 intergenic binding between convergent genes, neither gene was reported.