High binding affinity of repressor IolR avoids costs of untimely induction of myo-inositol utilization by Salmonella Typhimurium

Growth of Salmonella enterica serovar Typhimurium strain 14028 with myo-inositol (MI) is characterized by a bistable phenotype that manifests with an extraordinarily long (34 h) and variable lag phase. When cells were pre-grown in minimal medium with MI, however, the lag phase shortened drastically to eight hours, and to six hours in the absence of the regulator IolR. To unravel the molecular mechanism behind this phenomenon, we investigated this repressor in more detail. Flow cytometry analysis of the iolR promoter at a single cell level demonstrated bistability of its transcriptional activation. Electrophoretic mobility shift assays were used to narrow the potential binding region of IolR and identified at least two binding sites in most iol gene promoters. Surface plasmon resonance spectroscopy quantified IolR binding and indicated its putative oligomerization and high binding affinity towards specific iol gene promoters. In competitive assays, the iolR deletion mutant, in which iol gene repression is abolished, showed a severe growth disadvantage of ~15% relative to the parental strain in rich medium. We hypothesize that the strong repression of iol gene transcription is required to maintain a balance between metabolic flexibility and fitness costs, which follow the inopportune induction of an unusual metabolic pathway.

Bacterial pathogens often encounter limited availability of carbon, nitrogen and energy sources during their life cycles due to competition with the host organism and the commensal microbiota for nutrients. Increasing evidence suggests that enteric pathogens have developed specific metabolic strategies to overcome those restrictions and thus increase their fitness in the environment and in vivo. Metabolic adaptations of salmonellae during infection include the degradation of ethanolamine cleaved from phospholipids, fucose derived from mucosal glycoconjugates, 1,2-propanediol, the fermentation product of fucose or rhamnose 1 , and other potential substrates not accessible to commensal bacteria or present in competition-free microenvironments.
The ability of Salmonella enterica serovar Typhimurium (S. Typhimurium) to grow with myo-inositol (MI) as its sole carbon and energy source provides another example of an adaptive metabolic pathway with a potential role in vivo. In S. Typhimurium 14028, the iol genes responsible for MI utilization are located on a 22.6-kb genomic island (GEI4417/4436) 2 . The catabolic pathway requires the IolG-and IolE-mediated convertion of MI to 3D-trihydroxycyclohexane-1,2-dione, which is further hydrolysed by IolD. Next, 5-deoxy-glucuronic acid is isomerized by IolB to 2-deoxy-5-keto-D-gluconic acid, which is in turn phosphorylated by the kinase IolC and degraded to dihydroxyacetone phosphate, acetyl coenzyme A and CO 2 . The activator ReiD has been shown to interact with the promoter of iolE and thus to positively regulate iol gene expression 3 .
A remarkable and unique property of S. Typhimurium 14028 cultivated in minimal medium (MM) with MI is the manifestation of a heterogeneous growth phenotype that is abolished in the absence of the iol gene repressor IolR or the presence of at least 0.55% CO 2 4 . This phenotypic heterogeneity correlates with the bistable expression of iolE and iolG, the products of which catalyze the first steps in MI degradation. At the single cell level, fluorescence microscopy and flow cytometry (FC) analyses revealed that increasing numbers of cells exhibit P iolE activation during growth with MI 4 . More recently, we demonstrated that a shift from MI medium to LB medium is characterized by a "memory" of salmonellae for the former medium, and that this hysteresis is successively lost after 8 hours in a medium without MI 5 .
A key player in MI utilization is the negative autoregulator IolR that represses in rich medium all but one promoter of the iol genes involved in MI degradation, including its own promoter 2 and that of the iolT1 gene encoding the predominant MI transporter in S. Typhimurium 14028 6 . An intermediate of MI degradation, 2-deoxy-5-keto-D-gluconic acid 6-phosphate (DKGP), has been shown to antagonize IolR binding, thus inducing the expression of iol genes 7,8 . Here, we investigated IolR in greater detail, as this repressor plays a pivotal role in the coordinated control of MI degradation and the observed phenotypic heterogeneity. Using quantitative growth data of the wildtype and the iolR deletion mutant, the key parameters of growth with MI were delineated. The transcriptional activity of the iolR promoter was analyzed at a single cell level via FC. Electrophoretic mobility shift assays (EMSAs) and surface plasmon resonance (SPR) spectroscopy qualitatively characterized the binding properties of IolR towards its target promoters. Finally, competitive growth assays suggested that induced cells suffer a growth disadvantage in the absence of MI, a constraint that might contribute to the tight repression of iol gene transcription in rich medium.

Results
IolR contributes to lag-phase length during growth with MI. The lag phase of S. Typhimurium in the presence of MI has not previously been evaluated in detail. Therefore, we measured cultures that were inoculated with cells pregrown either in LB medium or on solid MM/MI. An overnight culture of 14028 in LB medium was adjusted to an OD 600 = 0.8 and then diluted 1:500 into MM/MI. When this dilution was further cultivated under standard conditions, a lag phase of approximately 34 h was observed (Fig. 1A). The length of this lag phase was characterized by a low level reproducibility, which was reflective of the stochastic and heterogeneous growth behaviours under this condition as indicated by high standard deviations 4 . In contrast, when strain 14028 was pregrown on MM/MI agar plates, adjusted to an OD 600 = 0.8 and diluted as described, the resulting culture exhibited a much shorter lag phase of 8 h, which could be attributed to the fact that the cells had already adapted to MI utilization. As a control, a growth curve of strain 14028 pregrown in LB medium that was subsequently diluted into fresh LB medium showed a lag phase of < 2 h (Fig. 1A). The standard deviations of data from the experiment using 14028 pregrown on MM/MI were similar to those observed in the control, indicating that strain 14028 grows homogeneously in MM after preadaptation to this medium.
Next, mutant 14028 ∆ iolR, which lacks the repressor IolR, was investigated under the experimental settings described above. When an LB culture was used to inoculate a MM/MI culture, we observed a reduction in the lag phase to 30 h, as well as a reduced population heterogeneity, as indicated by the low standard deviation, relative to the growth of the parental strain (Fig. 1B). When the mutant was pre-grown on MM/MI prior to the dilution Strain 14028 pre-grown in MM/MI and then diluted 1:500 into MM/MI exhibited a much shorter lag phase in comparison with that of a culture inoculated with a pre-culture in LB medium. The growth curve of a culture taken from LB medium and reinoculated into the same medium is shown as a control. (B) Lag phases were shortened by deletion of the repressor gene iolR. Growth curves of 14028 from (A) are displayed in black, and those of 14028 ∆ iolR are indicated by dashed lines. Growth curves were derived as described in (A). Standard deviations were calculated from three biologically independent measurements. into fresh MM/MI, a lag phase of 6 h was observed compared with 8 h observed for strain 14028 under the same conditions. In the LB control experiment, the lag phase of the ∆ iolR mutant was one hour longer than that of strain 14028, possibly because the targeted induction of the MI degradation pathway in 14028 Δ iolR led to weak growth attenuation (see below).
Taken together, the lag phase of S. Typhimurium growth in MM/MI ranges between 6 h and 34 h under the conditions used herein.

Temporal analysis of P iolR transcription by FC.
We have previously demonstrated at the single cell level that the promoter of iolE is bistable 4 . The promoter of iolR, P iolR , is transcriptionally active in the presence of both glucose and MI 2 . To investigate the transcription of iolR in more detail, we performed a temporal FC analysis of strain MvP101 with the transcriptional fusion P iolR ::gfp during growth in MM with MI. Figure 2A shows that all cells exhibit fluorescence below the threshold level at time point zero (i. e. when LB-cultured cells were inoculated into MM/MI medium). During cultivation until the end of the lag phase, the number of cells exhibiting a higher level of fluorescence steadily increased, and two subpopulations with a low and a high expression level were detected. 14 hours after inoculation, a temporary increase of the fluorescence activity was visible. At this time point, the division rate of the cells was still low, possibly resulting in an accumulation of GFP within the cells. To test whether the translation of iolR is also subject to bistability, we measured the fluorescence of strain MvP101 harboring the translational fusion iolR::gfp in the presence of MI. The data shown in Fig. 2B generally resemble those obtained with a transcriptional reporter fusion protein, indicating that the transcriptional bistability is not abolished during IolR production by strain MvP101. Samples of the control strain MvP101 P rpsM ::gfp with constitutive expression were collected and measured for up to 48 h of growth in MM/MI, during which time we observed averages of approximately 93% cells with and 7% cells without gfp expression (Fig. 2C). As the fusion P rpsM ::gfp is expected to be active in viable cells only, the data demonstrate that the majority of cells without significant fluorescence activity are not dead cells.
As an indicator for heterogeneous gene expression within a population, we calculated the variability in the fluorescence intensity, or the noise-level 9 , over all the growth phases in MM/MI of the two reporter strains described  Table 1). The noise-levels of both MvP101 P iolR ::gfp and MvP101 iolR::gfp were highest (1.67 ± 0.13) at the beginning of the lag phase and decreased to minimum levels (0.08 ± 0.0) during the stationary phase, when the cell populations showed more homogeneous gene expression. In contrast, the control strain MvP101 P rpsM ::gfp exhibited an average noise-level of 0.01 ± 0.01 during all growth stages.
To summarize, these findings demonstrate a bistable transcriptional activity of P iolR similar to that of P iolE .
Interaction of IolR-His 6 with iol gene promoters. The binding of IolR to most iol promotor regions was demonstrated previously 2,3,6 . To delineate a molecular model for the interaction of IolR and its target sites, we here performed EMSAs using different-length promoter fragments that represented possible IolR binding regions. The  regulator was overproduced in E. coli BL21(DE3) and purified as described in the Methods. Putative promoter fragments of iolR, iolT1, iolT2, iolA, reiD, iolC1 and iolD1 were incubated without or with increasing amounts of purified IolR-His 6 protein, and the resulting protein-DNA complexes were separated on a polyacrylamide (PAA) gel 10 . The EMSAs demonstrated the specificity of IolR-His 6 to its target DNA and narrowed the binding region of IolR within the promoters P iolR , P iolT1 , P iolT2 , P iolA , P reiD , P iolC1 and P iolD1 to a length of approximately 100 bp (Fig. 3). The molecular excess of protein over DNA for quantitative binding of most promoters tested was in a range between 5 to 21 and thus very low in comparison with other repressors tested under similar conditions 11,12 , pointing to a strong affinity of IolR to its target sequences. The bandshift experiments with the promoters P iolT1 , P iolD1 , P iolR , P iolC1 and P reiD exhibited at least two retarded protein/DNA complexes, suggesting an oligomerization of IolR with increasing protein concentration. We then investigated two fragments of 74 and 81 bp from P reiD with an overlap of only 7 nucleotides by EMSA and observed a binding of IolR-His 6 to both sequences, indicating the presence of two independent IolR target sites (Fig. 3). Moreover, the molar excesses of approximately 20 and 5, respectively, for a complete shift of these two P reiD fragments demonstrate different affinities of IolR to the two binding sites.

Quantification of DNA-binding by IolR-His 6 .
To analyse the interaction of IolR with its target sites identified above in more detail, we performed a kinetic analysis of IolR-His 6 binding to P iolT1 , P iolR and P reiD via SPR spectroscopy using fragments selected from the EMSAs. IolR was confirmed to specifically interact with P iolT1 , P iolR and P reiD , whereas no interaction was detected with the control promoter P pcfA (Fig. 4A-D). Because the maximal binding capacity (R max ) for each sensorgram is > 1.000 responsive units (RU) and the R max for an 1:1 interaction is expected to be approximately 20 RU, these data suggest that IolR binds as a higher oligomer and/or that multiple binding sites for IolR exist within the iol promoter regions. The finding that none of the sensorgrams followed 1:1 binding kinetics, as indicated by the hyperbolic curve shape, underscores the assumption that the sensorgrams represent the sums of different binding events. To calculate reliable binding constants and kinetic parameters, we performed an IM analysis and determined the individual binding events for the SPR spectroscopy curves of IolR-P iolT1 . Based on the sensorgrams, the IM analysis identified two clearly separated peak values (Fig. 4E) that represented a high-affinity interaction, and an interaction with an approximately 400-fold lower affinity. The first peak (blue) had a peak weight of 40.5%, an ON-rate of 6.95 × 10 5 M −1 s −1 and an OFF-rate of 5.78 × 10 −4 s −1 , resulting in an overall high affinity of 0.81 nM ( Fig. 4E and F, upper panel). The second peak (green) had a peak weight of 57.6% and an ON-rate of 1.05 × 10 5 M −1 s −1 . The OFF-rate for this interaction was determined to be 3.2 × 10 −2 s −1 , which was 50-fold higher than the first peak (depicted in blue), corresponding to a lower K D of 312 nM ( Fig. 4E and F, lower panel). We therefore conclude that two IolR binding sites with different affinities exist in the P iolT1 promoter region, which are occupied by a IolR oligomer that influences target gene expression. Because each binding peak accounts for approximately 50%, it is hypothesized that both DNA-binding sites are bound by equal numbers of IolR molecules. According to the OneToTwo evaluation algorithm, the binding kinetics underlying the sensorgrams of IolR-P iolR and IolR-P reiD interactions were nearly identical to that of the IolR-P iolT1 complex. The ON-rates for the IolR-P iolR and IolR-P reiD interaction were determined to be k a1 = 1.3 × 10 5 M −1 s −1 and k a2 = 2.4 × 10 5 M −1 s −1 , and k a1 = 1.5 × 10 5 M −1 s −1 and k a2 = 2.1 × 10 5 M −1 s −1 , respectively. The corresponding OFF-rates were k d1 = 5.5 × 10 −2 s −1 and k d2 = 1.4 × 10 −3 s −1 and of k d1 = 4.9 × 10 −2 /s and k d2 = 1.4 × 10 −3 /s, respectively. For both promoters, these data yielded to two binding events that differed with respect to the overall affinity (K D 1 = 425 nM; K D2 = 6.0 nM, and K D1 = 321 nM; K D2 = 6.6 nM, respectively).
Taken together, IolR binds at two sites within the respective iol gene promoter region, with one site bound with very high affinity in the low nanomolar range, and one site with lower affinity. The binding kinetics of the higher affinity-site with a high ON-rate and a very slow OFF-rate are comparable to those of antigen-antibody interactions 13,14 , thus demonstrating a remarkable high affinity of IolR to the promoter regions of iolT1, iolR, and reiD.
Fitness of cells with constitutive iol gene expression. The high binding affinity of IolR prompted us to speculate that in media lacking MI, cells with an activated MI utilization pathway would exhibit attenuated growth compared to those with silenced iol genes. As demonstrated previously through a promoter probe and FC analysis 2,4 , deletion of iolR for results in constitutive iol gene transcription regardless of the growth conditions. Therefore, we investigated the relative growth fitness of strain 14028 iolR::Kan R in comparison with the kanamycin-sensitive (Kan S ) strain 14028 in a competitive growth experiment. Both strains were grown separately in LB medium to an OD 600 of 0.5 and subsequently mixed in a 1:10 ratio (14028:14028 iolR::Kan R ). To control the ratio of the two strains, aliquots were plated to determine the numbers of colony forming units (cfu)/ml. The mixture ("inoculum") was appropriately diluted into LB medium. After incubation for 24 h without agitation, the numbers of viable Kan R and Kan S cells were determined by plating dilutions of the culture on solid LB medium with or without kanamycin. To determine the longer-term strain fitness, 0.1 ml of the 24-h culture was inoculated 1:500 into fresh LB medium and incubated for 1 day. Then, the cfu/ml of each strain was determined and an aliquot of this culture was used to inoculate fresh medium. This third culture was also incubated for 24 h and analyzed as described above. As a measure of the relative fitness of each strain, we calculated the cfu/ml of each strain after each of the three passages. The experiments started with a roughly 10-fold excess of strain 14028 iolR::Kan R versus 14028 (89.6%, ± 0.62%). Then, the percentage of the iolR-negative strain in the mixture decreased drastically to 29.2% (± 7.46%) after the first passage, to 11.8% (± 2.36%) after the second passage and to 4.2% (± 0.51%) after the third passage (Fig. 5A), indicating a significant reduction in the fitness of strain 14028 iolR::Kan R relative to strain 14028.
To exclude the possibility that the decrease of the iolR mutant fitness was related to the energetic cost of the Kan R cassette, we generated a mutant (14028 dacB::Kan R ) in which dacB was replaced with a Kan R cassette. DacB encodes penicillin binding protein 5, which exhibits DD-carboxypeptidase activity, and does not affect S. Typhimurium growth 15 . Using the kanamycin-sensitive, non-polar deletion mutant 14028 ∆ iolR as a competitor strain, we repeated the described experiment using a strain ratio as mentioned above. Again, we observed a strong growth retardation of the iolR mutant compared to strain 14028 dacB::Kan R , in which iol gene transcription was  ∆ iolR (B). This mixture was used to inoculate 50 ml LB medium in an appropriate dilution of 1:250 to 1:1000. After a 24-h incubation, the cfu/ml were determined by plating culture samples on LB medium with or without kanamycin, and an aliquot of each culture was diluted 1:500 into fresh LB medium. This step was repeated twice. For each passage, the percentages of strains within the cultivated mixture are shown. (C) Competitive growth experiment with complemented strain 14028 iolR::kan R /pBR-iolR against 14028/pBR322, and with strains 14028 and 14028 iolR::Kan R Δ 4418-4436; 14028 and 14028 iolR::Kan R were used as a control. The inoculum was diluted 1:500. Cell numbers were calculated only after the third passage. Standard deviations were calculated from triplicate measurements of three independent experiments. Significance values were below 1% (p < 0.01*) or below 0.01% (p < 0,0001***).
We then tested 14028 iolR::Kan R /pBR-iolR against 14028/pBR322 in a similar manner. The complementation of iolR strongly reduced the growth advantage of strain 14028 over the mutant 14028 iolR::Kan R (initial ratio 2.4:1 versus 441:1 after three passages) to a ratio of 29:1 (Fig. 5C). This data refute the hypothesis that incidental mutations outside the iolR gene played a role to the growth phenotype observed above.
Recently, IolR was reported to activate the expression of genes encoding components of the reversible lysine acetylation system 16 . To therefore exclude the possibility that any function beside iol gene repression contributes to the growth attenuation of 14028 iolR::Kan R , we performed a competitive growth experiment with strains 14028 and 14028 iolR::Kan R Δ 4418-4436. Here, the initial ratio of 1:3 between the strains shifted to a ratio of 2.3:1 after three passages (Fig. 5C) in contrast to the control experiment with strains 14028 and 14028 iolR::Kan R (ratio 2.4:1 versus 441:1), demonstrating that the manifestation of fitness costs of a iolR mutant requires the presence of all genes of the genomic island. As an additional control, the growth of each strain in LB medium was monitored to exclude non-specific growth deficiencies, but no differences in growth were observed ( Figure S1).
Taken together, both 14028 ∆ iolR and 14028 iolR::Kan R , which feature artificially arrested iol gene expression, exhibited a fitness disadvantage in rich medium in comparison with strains 14028 dacB::Kan R or 14028, in which the iol genes are tightly repressed. We conclude that tight repression of the MI utilization pathway in S. Typhimurium is likely a consequence of the fitness costs of inappropriately relaxed iol genes, resulting in a constraint of the number of cells with activated iol genes within an isogenic population.

Discussion
Recent findings suggested that the MI utilisation pathway of S. Typhimurium is very tightly repressed in the absence of MI, and that this repression is mainly achieved by the repressor IolR 2 . This explanation is biologically reasonable as S. Typhimurium can utilize a number of carbon sources, among which MI is not the most preferable 1 . Strong iol gene repression is phenotypically manifested by the start of logarithmic growth of cells that were pre-grown in LB medium diluted 1:500 into MM with MI after only 35 h in culture. In contrast, a much shorter lag phase similar to that observed in rich medium was obtained when Salmonella cells had been pre-grown in MI-containing MM and were thus adapted to the MI-utilisation metabolic pathway. Under this experimental condition the bistable phenotype was abolished, as indicated by homogeneous growth behavior 4 and low variability in the lag phase duration, which resembled that observed in rich medium. We therefore investigated the repressor IolR here in more detail.
The SPR spectroscopy performed in this study demonstrated high binding affinities of IolR to the promoters P iolT1 and P reiD , which control the genes encoding the main MI transporter and the activator of the MI degradation cascade. Additionally, this study revealed the oligomerization of IolR as indicated by the retarded protein/DNA complexes (Fig. 3). IolR oligomerization was also suggested by SPR spectroscopy. Calculating the max binding capacity of the chip surface at a simple one-to one interaction would result in an RU value of ~20. Since a maximal RU of 1200 RU was measured, we conclude that IolR binds as a higher oligomer even if two binding sites are present in a promoter fragment. Analysis of the SPR sensorgrams then identified one high ON/high OFF and one high ON/slow OFF binding sites of IolR within P iolT1 , P iolR and P reiD with different affinities. The detection of these two binding sites with high-affinity (A-site) and low-affinity (B-site) suggests a complex, cooperative DNA-binding mechanism of IolR similar to the so-called AB-BA mechanism that was recently described for the bacterial response regulator YpdB 17 . In this example, subsequent and cooperative promoter binding by several YpdB copies and a rapid successive promotor clearance results in a pulse-like gene expression.
Using the data obtained in this study, we developed a model for iol gene transcription (Fig. 6). We propose that the heterogeneous phenotype of S. Typhimurium growth with MI is essentially related to the mode how IolR interacts with the promoters of those genes that are required for MI utilization. The iolR promoter exhibits a non-linear response to growth in MI with two subpopulations of low and high expression level as indicator for bistability 18 . The bistable growth phenotype might be caused by varying copy numbers of heterogeneously transcribed IolR in a sense that IolR dimers do not bind or bind to one or to both of the binding sites, depending on the copy numbers of the repressor. FC analysis demonstrated that during growth with MI, iolR expression increases. The IolR molecules, however, are not able to bind iol promoters, possibly due to the presence of the catabolic intermediate DKGP that may induce the release of IolR from its target promoters and fine-regulate the repression by IolR. Then, the high amount of inactived IolR permits rapid re-silencing of the iol genes upon DKGP dissociation if additional MI is not available or has been replaced by a more optimal carbon and energy source, thus allowing a quick response to changing environments.
The average number of S. Typhimurium cells harbouring P iolE in the ON status is remarkably low in the absence of MI 4 . Therefore, in light of the high binding affinity of IolR, we speculated that a strong selective force acts on the cells. To test this hypothesis, we used a deletion mutant of iolR in which the repression of most iol genes is abolished, a situation that at least partially mimics that of cells with an activated iolE promoter. In competitive growth experiments with cross-over labelled strains, strain 14028 clearly outcompeted 14028 Δ iolR (Fig. 3). The outcome of this experiment led us to conclude that cells that do not use MI but express iol genes have a significant fitness disadvantage and are outgrown by cells in which IolR tightly represses expression of these genes. We calculated a high growth rate reduction by ~15% of 14028 iolR::Kan R . This is in line with the finding that a misfolded YFP protein of resulted in growth disadvantage of 3.2% in yeast 19 , because up to 16 enzymes are untimely produced in a iolR-negative strain. In a cross-over labelled experiment involving kanamycin-resistant strain 14028 dacB::Kan R and kanamycin-sensitive strain 14028 ∆ iolR, we excluded the possibility that reduced fitness is attributable to the antibiotic resistance 20 . Interestingly, when macrophage infection assays were performed 21 , the survival rate of a iolR knockout mutant was reduced to only 32% in comparison with the parental strain 14028 (data unpublished). This phenotype may be attributed to the growth disadvantage of the repressor mutant revealed here.
While biological costs of antibiotic resistances or reporter proteins have been studied in detail 22,23 , much less reports are available on costly metabolic pathways in case of untimely regulation. Experiments with the lactose operon led the assumption that those costs are associated with transcription and/or translation, but not with the products themselves 24 . More recently, however, this statement was challenged by the hypothesis that the lac permease activity is the main physiological burden to E. coli 25 . A recent example of biological costs of gene induction in a pathogen was studied in a fraction of cells harbouring an active type three secretion system (T3SS), which is involved in the virulence of S. Typhimurium. The authors reported the retarded growth of cells exhibiting the T3SS-positive phenotype relative to those not expressing the T3SS 26 . A similar observation is the growth restriction of Yersinia strains under T3SS-inducing conditions after two generations 27 .
Regulatory mechanisms have probably been evolved to manage tradeoffs for the fitness costs imposed by gene expression and enzymatic activities 28 . We therefore hypothesize here that the strength of the transcriptional repression of the iol genes is an evolutionary result of the balance between the expression of a metabolic capability that would be useful in niche occupation, and the fitness cost imposed by such a pathway when an alternative to the respective substrate is present.

Methods
Bacterial strains, plasmids and growth conditions. Bacterial strains and plasmids used in this study are listed in Table 2. Strain MvP101 was chosen for safety reasons. Comparison with a shotgun sequence of MvP101 confirmed that its GEI4417/4436 sequence is identical to that of ATCC strain 14028. S. Typhimurium strains were grown at 37 °C in Luria-Bertani (LB) medium (10 g/L tryptone, 5 g/L yeast extract and 5 g/L NaCl) or minimal medium (MM). MM comprised M9 medium supplemented with 2 mM MgSO 4 , 0.1 mM CaCl 2 and 55.5 mM (1%, wt/vol) MI or 25.2 mM (0.5%, wt/vol) glucose. A solution of 1.5% agar (wt/vol) was supplemented to yield solid media. For all FC experiments, bacterial strains were grown overnight at 37 °C in LB medium or MM and subsequently diluted to 1:500 into fresh medium. Bacterial growth curves were obtained from cultures incubated at 37 °C without agitation in 250-ml bulb flasks containing 50 ml of medium. The optical density at 600 nm (OD 600 ) was measured at the indicated time points.
Standard procedures. DNA manipulations and chromosomal or plasmid DNA isolation were performed according to standard protocols 10 and the relevant manufacturers' instructions. Plasmid DNA was transformed via electroporation using a Bio-Rad Gene Pulser II as recommended by the manufacturer (Bio-Rad, Hercules, CA, USA) and as described previously 21 . Polymerase chain reactions (PCRs) were performed using Taq polymerase (Fermentas, St. Leon-Rot, Germany). Chromosomal DNA, plasmid DNA or cells from a single colony were used as a PCR template. S. Typhimurium 14028 gene numbers refer to the LT2 annotation (NC 003197). Statistical analyses were performed with ANOVA or t-test; three levels of significance were considered (p < 0.05, p < 0.01 and p < 0.001).
Cloning of gfp reporter strains. The rpsM and iolR promoters defined as the 500 bp upstream of the start codons, and the last 500 bp of the iolR coding sequence were amplified from the chromosomal DNA of strain 14028 by PCR using the oligonucleotides listed in Table S1. The fragments were cloned via KpnI (Fermentas) upstream of gfp into the multiple cloning site of the suicide vector pUTs-gfp(Cm R ) and amplified in E. coli S17.1. Plasmids containing the correct gfp-fusions were isolated, verified by PCR and transferred into S. Typhimurium 14028 via conjugation. Strains with chromosomal insertions were selected, and validated by DNA sequencing (GATC, Konstanz, Germany).
Generation of an iolR deletion mutant. The one-step method based on phage λ Red recombinase cassette 2,29 was used for the allelic replacement of iolR by a kanamycin-resistance cassette (Kan R ). To exclude unspecific recombination events, 5 ml (OD 600 ) of 14028 donor cells harbouring the Kan R were mixed with 5 μ l of P22 phage suspension and incubated for 6 h without agitation. The culture was stored for 2 h at 4 °C, after which the cells were sedimented via centrifugation at 7000 rpm and 4 °C for 10 min. The supernatant was filter sterilized through a 0.2-μ m pore filter and stored at 4 °C. Recipient MvP101 cells were prepared from an overnight culture, mixed at a 1:20 ratio with P22 lysate and incubated for 1 h at 37 °C without agitation. After incubation, the mixture was plated on selective agar-plates and incubated at 37 °C. To obtain phage-free colonies, bacteria were cultivated on green indicator plates 30 . A nonpolar deletion mutant was then obtained upon transformation with pCP20. Gene deletions were verified by PCR analysis and DNA sequencing.   Overproduction and purification of IolR. IolR-His 6 was overproduced from the plasmid pET28b-IolR in E. coli BL21 (DE3) and S. Typhimurium strain 14028 2 . Overnight culture of the strains were diluted 1:100 into 400 ml of LB medium supplemented with 50 μ g/ml kanamycin and incubated for 3 h at 37 °C with rotation at 180 rpm. At an OD 600 of 0.6, IolR production was induced by adding 0.1 mM isopropyl-β -D -1-thiogalactopyranoside (IPTG). After an overnight incubation at 37 °C and 180 rpm, the cells were harvested by centrifugation at 4 °C and 7500 rpm for 20 min. The pellets were each resuspended in 5 ml of native lysis buffer (50 mM NaH 2 PO 4 , 300 mM NaCl and 10 mM imidazole at pH 8.0) and lysed by three passages through a French press (SLM Aminca Instruments, Rochester, NY, USA) at 900 psi; residual cell debris was removed twice by centrifugation at 4 °C and 9000 rpm for 15 min. Following filtration, IolR was isolated using a HisTrap HP colunm with an ÄKTA purifier 10 system (GE Healthcare, Little Chalfont, UK). The protein concentration was determined using RotiQuant solution (Carl, Roth GmbH, Karlsruhe, Germany) according to the Bradford method 32 . The purity of the eluted fractions was analysed by separation on a 15% sodium dodecyl sulfate (SDS)-PAA gel.
Electrophoretic mobility shift assays. Fragments of promotor regions were amplified (for oligonucleotides, see Table S1) as decribed above, and 100 ng of DNA were mixed with increasing amounts of purified IolR-His 6 in 1× Tris/borate/EDTA (EMSA) buffer to a total volume of 20 μ l. After incubation for 45 min, the samples were mixed with 4 μ l of 6× loading dye (MBI Fermentas) and loaded on a 12% native PAA gel prepared in EMSA buffer; the samples were then separated at 120 V and 4 °C for 2 h in the same buffer. Gel-bound DNA was stained using GelRed (Biotium, Hayward, CA, USA) and visualized by ultraviolet (UV) irradiation.
Biotinylated double-stranded (BIOT-ds) DNA fragments. BIOT-ds DNA fragments were obtained by annealing ss-DNA-oligonucleotides or via PCR. BIOT-P iolT1 , BIOT-P iolR and BIOT-P reiD were amplified by PCR using the respective biotinylated DNA oligonucleotides (Sigma-Aldrich, Deisenhofen, Germany) and chromosomal DNA from strain 14028 as a template. To assemble the control promoter region P pcfA from Photorhabdus luminescens 33 , oligonucleotides P4568-btn_fw and P4568_rew were incubated for 5 min at 100 °C, mixed and cooled prior to annealing. BIOT-ds DNA fragments were then captured on SPR sensor chips.
SPR spectroscopy and calibration-free concentration analysis (CFCA). SPR spectroscopy and CFCA assays were performed using a Biacore T200 device (GE Healthcare) and streptavidin-precoated Xantec SAD500-L carboxymethyl dextran sensor chips (XanTec Bioanalytics GmbH, Düsseldorf, Germany). All experiments were conducted at 25 °C with HBS-EP+ buffer [10 mM HEPES pH 7.4, 150 mM NaCl, 3 mM EDTA and 0.05% (v/v) detergent P20] as previously described in detail 17 . Assuming a globular shape, the diffusion coefficient of IolR was calculated using the Biacore diffusion constant calculator and converter webtool (https://www.biacore.com). The diffusion coefficient of IolR was determined to be D = 9.946 × 10 −11 m 2 /s. The initial rates of the dilutions that differed by a factor of at least 1.5 were considered when calculating the concentration of IolR that interacted with the ligand. This "active" protein concentration, which was determined as 3.3 × 10 −8 M (66% of the total protein concentration), was then used to calculate the binding kinetic constants and steady-state affinity.
Interaction Map (IM) analysis. IM analyses were performed as previously described 17,34 . SPR spectroscopy curves were determined as the sums of individual binding curves, each of which represented a monovalent interaction 35 with a unique combination of the association rate k a (ON-rate) and dissociation rate k d (OFF-rate) and the consequently unique real equilibrium dissociation constant K D = k d /k a . This algorithm was used to split the experimental SPR spectroscopy data set into several theoretical monovalent binding curves and the binding curves that, when summed, best fit the experimental data. The association rate k a and dissociation rate k d were plotted within a two-dimensional distribution yield a heterogeneous binding data display in the form of a map wherein each peak corresponded to a single component that contributed to the cumulative binding curve 36 . To further quantify protein/DNA interactions, the OneToTwo evaluation algorithm, which is available in the Trace Drawer Software package (Ridgeview Diagnostics AB, Uppsala, Sweden), was used.