The phosphorelay BarA/SirA activates the non-cognate regulator RcsB in Salmonella enterica.

To survive an environmental stress, organisms must detect the stress and mount an appropriate response. One way that bacteria do so is by phosphorelay systems that respond to a stress by activating a regulator that modifies gene expression. To ensure an appropriate response, a given regulator is typically activated solely by its cognate phosphorelay protein(s). However, we now report that the regulator RcsB is activated by both cognate and non-cognate phosphorelay proteins, depending on the condition experienced by the bacterium Salmonella enterica serovar Typhimurium. The RcsC and RcsD proteins form a phosphorelay that activates their cognate regulator RcsB in response to outer membrane stress and cell wall perturbations, conditions Salmonella experiences during infection. Surprisingly, the non-cognate phosphorelay protein BarA activates RcsB during logarithmic growth in Luria-Bertani medium in three ways. That is, BarA’s cognate regulator SirA promotes transcription of the rcsDB operon; the SirA-dependent regulatory RNAs CsrB and CsrC further increase RcsB-activated gene transcription; and BarA activates RcsB independently of the RcsC, RcsD, and SirA proteins. Activation of a regulator by multiple sensors broadens the spectrum of environments in which a set of genes is expressed without evolving binding sites for different regulators at each of these genes.

Introduction Survival in a hostile environment requires gene products that protect an organism from the particular stresses present in that environment. Phosphorelay systems allow bacteria to respond to environmental insults by activating a regulatory protein that alters the expression of a subset of genes [1,2]. To ensure a specific response to a given stress condition, phosphorelay systems usually restrict activation to their respective cognate regulatory proteins [3][4][5]. However, we now report that a regulatory protein is activated both by cognate and non-cognate phosphorelay proteins depending on the stress experienced by the bacterium Salmonella enterica serovar Typhimurium.
Bacteria and cell wall-containing eukaryotes utilize two-component systems to alter behavior in response to changes in environmental or cellular conditions [6,7]. The prototypical two-component system consists of a sensor that responds to a specific signal(s) by modifying the activity of a cognate regulator through phosphorylation. Phosphorelays are versions of two-component systems in which there is a three-step phosphotransfer, whereby the phosphoryl group is shuttled from the sensor kinase to the response regulator via an intermediary phosphotransferase protein or domain, as opposed to the single phosphotransfer event from a sensor to a regulator in classical two-component systems [8,9].
One of the best-studied phosphorelays in the family Enterobacteriacae is Rcs, which consists of the RcsC, RcsD and RcsB proteins [10,11] (Fig 1). RcsC and RcsD are inner membrane proteins, whereas RcsB is a cytoplasmic DNA binding regulator that alters expression of its target genes when bacteria experience outer membrane stress or cell wall perturbations. These stresses are believed to promote RcsC autophosphorylation at a histidine residue followed by phosphotransfer to an aspartate residue within RcsC, followed by phosphotransfer to a histidine residue in RcsD, and ending in phosphotransfer to an aspartate residue on RcsB. Phosphorylated RcsB activates or represses a subset of genes involved in numerous cellular processes, including motility, biofilm formation, and the general stress response [3,[12][13][14]. Surprisingly, full expression of RcsB-activated genes has been observed in mutants lacking either the rcsC or rcsD gene [15,16], raising the possibility of RcsB being activated by signaling proteins other than RcsC and RcsD.
We now report that the sensor phosphorelay protein BarA activates the non-cognate regulator RcsB in an RcsC-and RcsD-independent manner. We establish that BarA utilizes several mechanisms to increase the amounts of active RcsB protein. And we identify conditions in which RcsB activation requires RcsC and RcsD but not BarA. Our findings indicate that a given regulatory protein can be activated by non-cognate phosphorelay proteins. This activation expands the spectrum of environments in which a given set of genes is expressed without an organism needing to evolve binding sites for different regulatory proteins at the regulatory region of each target gene. Because the proteins that constitute a given phosphorelay are often encoded by independently transcribed genes, they may be more prone to establish physiological interactions with "non-cognate" partners.
To test this hypothesis, we examined the fluorescence of isogenic Salmonella strains with mutations in various rcs genes and harboring a medium copy number plasmid with a transcriptional fusion between the RcsB-activated rprA promoter and a promoterless gfp gene. Although the RcsB protein can form heterodimers with several proteins [10,17,18], rprA transcription is carried out by RcsB homodimers [15], which are favored upon RcsB phosphorylation [10]. Bacteria were grown in regular (i.e., NaCl-containing) Luria-Bertani (LB) media and in LB without NaCl as a means to decrease osmolarity.
Wild-type Salmonella fluoresced when incubated on LB agar plates lacking NaCl at 37˚C for 24 h (Fig 2). Fluorescence was 4-fold higher on regular (i.e., NaCl-containing) LB plates than in LB plates lacking NaCl. These results are in agreement with the notion that increased osmolarity activates the Rcs system [19,20] (Fig 2). The rcsB mutant did not fluoresce on either plate (Fig 2), demonstrating that rprA transcription is fully dependent on RcsB under the two conditions [14,15]. By contrast, the rcsC and rcsD single mutants exhibited wild-type fluorescence on NaCl-lacking LB plates (Fig 2). While displaying half of the wild-type strain fluorescence on LB plates, the fluorescence of the rcsC and rcsD single mutants was higher than that of the rcsB mutant (Fig 2). These results indicate that the RcsB-dependent rprA gene is still expressed in the absence of either RcsC or RcsD [15,16].
The fluorescence of the rcsC rcsD double mutant was 2-fold higher on regular LB than in NaCl-lacking LB plates (Fig 2). Although this mutant had dramatically lower fluorescence than the wild-type strain or the rcsC and rcsD single mutants (Fig 2), its fluorescence was still higher than that of the rcsB mutant (Fig 2), reinforcing the notion that RcsB can be activated independently of RcsC and RcsD. These data argue that increased osmolarity activates RcsB by a non-canonical pathway.
The lipoprotein RcsF responds to envelope stress by activating the Rcs phosphorelay [11,21,22]. However, a rcsF null mutant retained wild-type fluorescence in both regular LB and NaCl-lacking LB plates (Fig 2). Taken together, these results demonstrate that RcsB can be activated by a pathway not involving the established players RcsC, RcsD, and RcsF.

RcsB activation by the phosphorelay sensor BarA
To identify the gene(s) responsible for the RcsB-dependent RcsC-and RcsD-independent activation of the rprA promoter, we screened~9,000 transposon Tn10dTc-generated mutants of the rcsC rcsD strain carrying the rprA-gfp fusion for decreased fluorescence on LB agar plates. Mutants exhibiting less fluorescence than the parental strain were further characterized by transducing the transposon insertion into the original strain, and upon recapitulation of the phenotype, the site of insertion of the Tn10dTc transposon was determined as described in Material and Methods.
One mutant had a total loss of fluorescence and, not surprisingly, harbored a Tn10dTc insertion in the rcsB gene (S1 Fig). Another mutant, exhibiting a slight decrease in fluorescence, had a transposon insertion in barA (S1 Fig), encoding the tripartite sensor kinase BarA that forms a phosphorelay with the regulator SirA (known as UvrY in E. coli) [23,24]. The BarA/SirA system has been implicated in motility, stress resistance, metabolism, and virulence [23,[25][26][27], some of which also require a functional RcsB protein [10,28].
When the barA gene was deleted from the rcsC rcsD double mutant, the resulting triple mutant exhibited less fluorescence than the rcsC rcsD double mutant on LB agar plates ( Fig  3A), recapitulating the behavior of the rcsC rcsD barA::Tn10dTc triple mutant (S1 Fig). These results indicate that the phenotype of the rcsC rcsD barA::Tn10dTc triple mutant is due to loss of barA function (rather than to the barA::Tn10dTc strain specifying a truncated BarA protein with altered activity).
The barA deletion mutant in an otherwise wild-type background had lower fluorescence than wild-type Salmonella on NaCl-lacking LB plates ( Fig 3A) but similar fluorescence on regular LB plates ( Fig 3A). The phenotype of the barA mutant is due to the absence of the barA gene because a plasmid with a copy of the wild-type barA coding region under the control of the cat promoter (pBarA) complemented the mutant, whereas the vector control (pVector) did not (S2 Fig). Fluorescence was also higher in wild-type Salmonella carrying pBarA than in the pVector control strain (S2 Fig). Given that the absence of barA decreases RcsB-dependent gene transcription in the presence of functional RcsC and RcsD proteins, BarA is a physiological activator of RcsB.
The barA rcsC and barA rcsD double mutants displayed lower fluorescence than the isogenic rcsC and rcsD single mutants, respectively, on both LB and NaCl-lacking LB plates ( Fig  3A, S8 Fig and S12 Fig). By contrast, the barA rcsC rcsD triple mutant exhibited less fluorescence than the rcsC rcsD double mutant strain in LB but not in NaCl-lacking LB plates ( Fig  3A).
As a member of the response regulator family, RcsB harbors a conserved aspartate residue that is phosphorylated when a bacterium experiences inducing conditions for its cognate phosphorelay proteins [10]. We have now determined that the BarA-mediated activation of RcsB requires the conserved aspartate in RcsB that is the site of phosphorylation. This is because fluorescence was much higher in wild-type Salmonella carrying the rprA-gfp fusion plasmid The response regulator RcsB can be activated independently of the RcsC, RcsD, and RcsF proteins. Fluorescence from wild-type (14028s), rcsB (EG12925), rcsC (HS1350), rcsD (HS1382), rcsC rcsD (HS1383) and rcsF (HS1326) Salmonella harboring plasmid pRprA-GFP (rprA-gfp) following 24 h of growth on LB solid medium without (-NaCl) or with (+NaCl) NaCl. Data are representative of three independent experiments, which gave similar results. Quantification of the fluorescence is provided on the right panel of the figure. Values derived from three independent experiments (mean ± standard deviation) were statistically analyzed by Prism 8 using two-tailed unpaired t test. Statistical significance is indicated by � P<0.05, �� P<0.01, ��� P<0.001; ns, not significant. Error bars indicate standard deviation. https://doi.org/10.1371/journal.pgen.1008722.g002

PLOS GENETICS
The phosphorelay BarA/SirA activates the non-cognate regulator RcsB in Salmonella enterica and pBarA than with the pVector control (S3 Fig), but similarly low in a rcsBD56Q genetic background (S3 Fig). The rcsBD56Q mutant encodes an RcsB protein that cannot be phosphorylated at the conserved aspartate located at position 56 [14] from the normal chromosomal location and promoter. Thus, RcsB's phosphorylation site is necessary for activation by BarA.
We conclude that RcsB activation by BarA is condition specific, that LB solid media is an activating condition for BarA, that this activation is impacted by the presence of the phosphorelay proteins RcsC and RcsD, and that it requires the site of phosphorylation in RcsB.

BarA activates RcsB in a time-dependent manner
To determine the kinetics of BarA activation of RcsB, we investigated the fluorescence of wildtype Salmonella carrying the rprA-gfp plasmid at different times after bacteria were inoculated into LB broth media with and without NaCl. Fluorescence was higher during growth in LB as BarA is required for full RcsB activation on LB agar plates. Fluorescence from wild-type (14028s), rcsB (EG12925), rcsC (HS1350), rcsD (HS1382), rcsC rcsD (HS1383), barA (HS1520), barA rcsC (HS1521), barA rcsD (HS1522) and barA rcsC rcsD (HS1523) Salmonella harboring plasmid pRprA-GFP (rprA-gfp) following 24 h of growth on LB solid medium without (-NaCl) or with (+NaCl) NaCl. Data are representative of two independent experiments, which gave similar results. (B) BarA activates RcsB in a time-dependent manner. Fluorescence from wild-type (14028s), rcsB (EG12925), rcsC (HS1350), rcsD (HS1382), rcsC rcsD (HS1383), barA (HS1520), barA rcsC (HS1521), barA rcsD (HS1522) and barA rcsC rcsD (HS1523) Salmonella harboring plasmid pRprA-GFP (rprA-gfp) or pVector (empty pFPV25 vector) following 9 h of growth in LB liquid medium without (-NaCl) or with (+NaCl) NaCl. For each strain, relative fluorescence was obtained by subtracting the fluorescence of the strain harboring pVector from the fluorescence of the strain harboring pRprA-GFP. The obtained value was then divided by OD 600 . Error bars represent standard deviation from three independent experiments. Values derived from three independent experiments (mean ± standard deviation) were statistically analyzed by Prism 8 using two-tailed unpaired t test. Significance values (P) are reported in the text. Relative fluorescence values (left axis) are represented by solid lines and OD 600 values (right axis) by dotted lines. (C) Western blot analysis of extracts prepared from wild-type (14028s), barA (HS1520) and rcsB (EG12925) grown in LB liquid medium without (-NaCl) or with (+NaCl) NaCl in late exponential phase (OD 600 of~1.0) following separation on Phos-tag SDS-PAGE to detect RcsB and RcsB-P. Samples were analyzed with antibodies directed to RcsB or AtpB proteins. Crude extracts from rcsBD56Q (HS1483) and rcsC11 (EG14873) strains overnight cultures were used as negative and positive controls for RcsB phosphorylation, respectively. The rcsBD56Q mutant encodes an RcsB protein that cannot be phosphorylated at the conserved aspartate located at position 56 [14] and the rcsC11 strain harbors the rcsC11 point mutation promoting constitutive Rcs system activation [67]. https://doi.org/10.1371/journal.pgen.1008722.g003

PLOS GENETICS
The phosphorelay BarA/SirA activates the non-cognate regulator RcsB in Salmonella enterica compared to LB without NaCl (Fig 3B), in agreement with the results obtained on solid media (Fig 2). The rcsB mutant displayed no fluorescence over the 9 h course of the experiment ( Fig  3B), as expected for the RcsB dependence of the rprA promoter (Fig 2) [14,15].
In NaCl-lacking LB, the rcsC and rcsD single mutants showed 2-to 3-fold more fluorescence than the wild-type strain from 3 to 7 h (P<0.01 for rcsC versus wild-type at 3 h; P<0.001 for rcsD versus wild-type at 3 h; P<0.05 for rcsC versus wild-type at 7 h; P<0.01 for rcsD versus wild-type at 7 h) ( Fig 3B). This result is consistent with the proposed reverse phosphate flow, whereby RcsD removes the phosphate from RcsB, and RcsC from RcsD [10,15], which, intriguingly, is not observed on solid media (Fig 2). After 9 h, however, the rcsC and rcsD single mutants exhibited similar fluorescence as wild-type Salmonella, indicating that RcsB can be activated in the absence of either RcsC or RcsD (Fig 3B).
In contrast to the behavior manifested in NaCl-lacking LB, the rcsC and rcsD single mutants displayed wild-type fluorescence from 3 to 6 h in LB (Fig 3B). The fluorescence of these mutants was two-thirds to one-half that of the wild-type strain from 7 to 9 h (P<0.0001 for rcsC versus wild-type at 7 h; P<0.01 for rcsD versus wild-type at 7 h; P<0.001 for rcsC versus wild-type at 9 h; P<0.001 for rcsD versus wild-type at 9 h) (Fig 3B), arguing that both RcsC and RcsD are necessary for full RcsB activation in LB broth. These data also show that either RcsC or RcsD is sufficient to carry out RcsB activation during the first 6 h.
The rcsC rcsD double mutant exhibited the same fluorescence as the wild-type strain from 3 to 6 h in NaCl-lacking LB, demonstrating that RcsB can be activated independently of both RcsC and RcsD ( Fig 3B). From 7 to 9 h, the fluorescence of the rcsC rcsD double mutant was two-thirds to one-third that of the rcsC and rcsD single mutants or of wild-type Salmonella (P<0.001 for rcsC rcsD versus rcsC at 7 h; P<0.0001 for rcsC rcsD versus rcsD at 7 h; P<0.01 for rcsC rcsD versus wild-type at 7 h; P<0.0001 for rcsC rcsD versus rcsC at 9 h; P<0.0001 for rcsC rcsD versus rcsD at 9 h; P<0.001 for rcsC rcsD versus wild-type at 9 h) (Fig 3B), implying that either RcsC or RcsD is sufficient to achieve full RcsB activation under these conditions. Although wild-type Salmonella and the rcsC rcsD double mutant exhibited a similar fluorescence at the beginning of growth in LB, fluorescence was 3-to 4-fold higher in wild-type Salmonella than in the double mutant from 6 to 9 h (P<0.01 for rcsC rcsD versus wild-type at 6 h; P<0.0001 for rcsC rcsD versus wild-type at 9 h) ( Fig 3B).
The barA mutant strain exhibited lower fluorescence than wild-type Salmonella both in LB (barA/wild-type ratio of 0.5 at 9 h; P<0.001) and NaCl-lacking LB (barA/wild-type ratio of 0.25 at 9 h; P<0.001) (Fig 3B). Consistent with this observation, the ratio of phosphorylated RcsB (RcsB-P) to unphosphorylated RcsB was lower in the barA mutant than in wild-type Salmonella when bacteria were grown in either LB or NaCl-lacking LB to late exponential phase (OD 600 of~1.0) (Fig 3C). Control experiments showed no RcsB phosphorylation in a strain expressing the RcsB D56Q variant, and hyper RcsB phosphorylation in a strain expressing the constitutive rcsC11 allele (Fig 3C).
Fluorescence was lower in the rcsC barA and rcsD barA double mutants than in the rcsC and rcsD single mutants, respectively, both in LB and NaCl-lacking LB media (Fig 3B), in agreement with the behavior displayed on solid media (Fig 2). Fluorescence was slightly lower in the barA rcsC rcsD triple mutant than in the rcsC rcsD double mutant both in LB and NaCllacking LB media (P<0.001 for NaCl-lacking LB at 9 h; P<0.04 for LB at 9 h) (Fig 3B), thus recapitulating the results obtained on solid media (Fig 2).
Cumulatively, the results presented in this section establish that BarA is required for full RcsB activation in LB broth and that this activation is time dependent. As discussed below, BarA uses multiple mechanisms to increase the amount of active RcsB protein but does not appear to target other regulators of the same protein family.

BarA activates RcsB via the SirA-activated regulatory RNAs CsrB and CsrC
To date, BarA's effect on gene expression has been largely ascribed to the regulatory RNAs CsrB and CsrC, which are encoded by genes directly activated by BarA's cognate regulator SirA [23]. CsrB and CsrC act by sequestering CsrA, an RNA-binding protein that binds to the 5' untranslated region (UTR) of hundreds of transcripts, typically decreasing their translation [29][30][31][32]. That BarA activation of RcsB may be mediated by CsrB and CsrC in Salmonella is also supported by the phenotype of a csrB mutant Yersinia pseudotuberculosis grown on solid media, which produced 8-fold less RcsB protein than the wild-type strain [33].
A sirA single mutant and a csrB csrC double mutant displayed half of the fluorescence of wild-type Salmonella when harboring the plasmid with the rprA-gfp fusion following growth on NaCl-lacking LB plates (Fig 4, Fig 5A and Fig 5B), behaving like the barA mutant ( Fig 5B). By contrast, the three mutants behaved similarly to the wild-type strain in regular LB plates (Fig 3, Fig 4, Fig 5A and Fig 5B).
CsrB (and likely CsrC) is largely responsible for the BarA-and SirA-dependent activation of RcsB because a plasmid expressing CsrB from a heterologous promoter complemented not only the csrB csrC double mutant but also the sirA and barA single mutants ( RcsC amounts were higher in the csrB csrC double mutant strain than in wild-type Salmonella at OD 600 of 1.0 (a 1.3-fold difference) and 2.0 (a 1.9-fold difference) in LB broth without NaCl ( Fig 5C). There were similar RcsC amounts at OD 600 of 0.5, indicating that the negative regulation that CsrB and CsrC exert on RcsC is dependent on the growth phase. As a rcsC single mutant exhibited increased RcsB activity compared to wild-type Salmonella under the same experimental conditions (Fig 3B, left panel), one may expect CsrB and CsrC to activate RcsB under low osmolarity conditions by reducing RcsC expression. This is consistent with the observation that the rcsC single mutant did not exhibit reduced activity as compared to rcsC csrB csrC mutant on LB plates without NaCl (Fig 5A and Fig 5B). Moreover, a CsrBexpressing plasmid did not increase the fluorescence of a rcsC csrB csrC mutant (S6 Fig). This is in contrast to the higher fluorescence displayed by the csrB csrC double mutant when carrying pCsrB versus the pVector control (S6 Fig).
By contrast, RcsB amounts were the same for both wild-type Salmonella and the csrB csrC double mutant in LB broth without NaCl (S7A Fig), which is in contrast to previous findings reported with Yersinia pseudotuberculosis [33], and this was also the case for RcsD amounts (S7B Fig). Cumulatively, the results in this section strongly suggest that when wild-type Salmonella experiences low osmolarity, BarA activates RcsB-dependent transcription by reducing RcsC abundance through the action of the SirA-dependent CsrB and CsrC regulatory RNAs.

SirA promotes rcsDB transcription from a rcsD promoter
When grown in LB without NaCl, the fluorescence of wild-type Salmonella harboring the plasmid with the rprA-gfp fusion was higher in cells harboring the plasmid expressing CsrB from a heterologous promoter than with the vector control (S4 Fig). This raised the possibility of the restoration of fluorescence conferred by the CsrB-expressing plasmid upon the sirA and barA mutants hiding other potential mechanisms by which the SirA and BarA proteins activate the RcsB protein. To explore this possibility, we examined fluorescence from the rprA-gfp fusion in isogenic strains with mutations in the csrB and csrC genes.
The csrB csrC rcsC and csrB csrC rcsD triple mutants had higher fluorescence than the rcsC and rcsD single mutants, respectively, on NaCl-lacking LB plates, which is in contrast to the lower fluorescence of the csrB csrC double mutant relative to wild-type Salmonella ( Because the rcsC sirA and rcsC rcsD double mutants exhibited similar fluorescence from rprA-gfp (Fig 4), and because SirA is reported to act primarily as a transcriptional regulator [23,25], we reasoned that SirA controls RcsD abundance. In support of this notion, the fluorescence of the sirA and sirA rcsC mutants strains was three-quarters to two-thirds that of wild-type Salmonella and of the rcsC mutant, respectively, (P<0.01 for sirA versus wild-type; P<0.001 for sirA rcsC versus rcsC at 7 h in NaCl-lacking LB; P<0.001 for sirA versus wildtype; P<0.001 for sirA rcsC versus rcsC at 7 h in LB) when carrying plasmid with a transcriptional fusion between the rcsD promoter and a promoterless gfp gene ( Fig 6A).
When tested on LB plates, wild-type Salmonella and the sirA mutant, both carrying a plasmid expressing sirA from a heterologous promoter and the rcsD-gfp fusion plasmid, showed higher fluorescence than the corresponding strains harboring the vector control ( Fig 6B). However, this was not the case for the barA mutant (Fig 6B), arguing that SirA activation is strictly dependent on BarA under the investigated conditions. By contrast, the sirA-expressing plasmid, but not the vector control, increased the fluorescence of the csrB csrC double mutant ( Fig 6B). The SirA-mediated activation of the rcsD promoter occurs independently of RcsB, a negative regulator of the rcsDB operon [34], as pSirA increased the fluorescence from rcsD-gfp in the rcsB mutant (S9 Fig). Collectively, these results indicate that SirA activates the rcsD promoter independently of the CsrB and CsrC regulatory RNAs and of RcsB, and that this activation relies on SirA's cognate phosphorelay sensor BarA.
To explore how SirA increases RcsD abundance, we mapped the rcsD transcription start site(s) in wild-type Salmonella harboring pSirA versus the pVector following growth in LB

PLOS GENETICS
The phosphorelay BarA/SirA activates the non-cognate regulator RcsB in Salmonella enterica broth to an OD 600 of 2.0. Two transcription start sites were identified 157 (rcsDB -157 ) and 36 (rcsDB -36 ) nucleotides upstream of the rcsD start codon (Fig 6C), in agreement with previous reports [34][35][36]. The abundance of the rcsDB -36 transcript was higher in the pSirA-containing strain than in the pVector-containing strain ( Fig 6C). The SirA-dependent effect is growth phase specific because the isogenic strains behaved similarly when bacteria were harvested at an OD 600 of 0.5 ( Fig 6C). In contrast to the behavior of the rcsDB -36 transcript, the abundance of the rcsD -157 transcript was the same in the pSirA-and pVector-carrying strains at both optical densities (Fig 6C). The SirA-dependent increase in mRNA abundance from the rcsDB -36 transcript resulted in larger amounts of the RcsD (1.4-fold increase) and RcsB (3.2-fold increase) proteins ( Fig 6D). Therefore, by promoting transcription from the rcsDB -36 start site, SirA increases the amounts of the rcsDB polycistronic transcript, resulting in higher amounts of the RcsD and RcsB proteins.
To identify rcsDB promoter element(s) required for SirA-mediated transcription, we examined the fluorescence of wild-type Salmonella harboring the sirA-expressing plasmid or the , csrB csrC rcsD (HS1655) and csrB csrC rcsC rcsD (HS1656) Salmonella harboring plasmid pRprA-GFP (rprA-gfp) following 24 h of growth on LB solid medium without (-NaCl) or with (+NaCl) NaCl. Data are representative of two independent experiments, which gave similar results. (B) Fluorescence from wild-type (14028s), barA (HS1520), sirA (HS1565) and csrB csrC (HS1651) Salmonella harboring plasmid pRprA-GFP (rprA-gfp) following 24 h of growth on LB solid medium without (-NaCl) or with (+NaCl) NaCl. Data are representative of three independent experiments, which gave similar results. Quantification of the fluorescence is provided next to the plate images. Values derived from three independent experiments (mean ± standard deviation) were statistically analyzed by Prism 8 using two-tailed unpaired t test. Statistical significance is indicated by � P<0.05, �� P<0.01; ns, not significant. Error bars indicate standard deviation. (C) Western blot analysis of crude extracts from rcsC-3XFLAG (HS539) and rcsC-3XFLAG csrB csrC (HS2263) strains grown in LB NaCl-free broth. Samples were analyzed with antibodies directed to the FLAG epitope or the RpoB protein. Data are representative of three independent experiments, which gave similar results. RcsC levels for csrB csrC mutant strain relative to wild-type Salmonella are marked below (fold increase ± standard deviation). Values were statistically analyzed by Prism 8 using two-tailed unpaired t test. Statistical significance is indicated by � P<0.05, �� P<0.01; ns, not significant. https://doi.org/10.1371/journal.pgen.1008722.g005

PLOS GENETICS
The phosphorelay BarA/SirA activates the non-cognate regulator RcsB in Salmonella enterica vector control, and a set of isogenic plasmids with an rcsD-gfp transcriptional fusion and various lengths of the putative rcsD promoter region (i.e. positions -270, -235, -220 and -110 relative to rcsD start codon) (Fig 6E). No SirA-dependent activation was observed in strains harboring the rcsD -110 and rcsD -220 fusions (Fig 6E and S10 Fig), similarly high activation with the rcsD -270 and rcsD -293 fusions (Fig 6E and S10 Fig), and intermediate activation with the rcsD -235 fusion (Fig 6E and S10 Fig)  Fluorescence from wild-type (14028s), barA (HS1520), sirA (HS1565) and csrB csrC Salmonella harboring pRcsD -293 -GFP (rcsD-gfp) with pSirA or pVector (empty pACYC184 vector) following 24 h of growth on LB solid medium without (-NaCl) or with (+NaCl) NaCl. Data are representative of two independent experiments, which gave similar results. (C) Primer extension analysis of rcsDB -36 and rcsDB -157 levels in wildtype (14028s) Salmonella harboring pSirA or pVector (empty pACYC184 vector) grown in LB broth. Primer extension reaction was carried out on total RNA samples using primer W4171. (D) Western blot analysis of crude extracts from rcsD-HA (HS1309) and rcsB-FLAG (HS717) Salmonella harboring pSirA or pVector (empty pACYC184 vector) grown in LB broth. Samples were analyzed with antibodies directed to the FLAG or HA epitopes or the RpoB protein. Data are representative of three independent experiments, which gave similar results. RcsB and RcsD levels for pSirA strain relative pVector strain are marked below. Values were statistically analyzed by Prism 8 using two-tailed unpaired t test. Statistical significance is indicated by � P<0.05, �� P<0.01; ns, not significant. (E) Identification of the region upstream of the rcsD coding region required for SirA-mediated activation of rcsDB. Fluorescence from wild-type (14028s) Salmonella harboring pRcsD -293 -GFP (rcsD-gfp), pRcsD -270 -GFP, pRcsD -235 -GFP, pRcsD -220 -GFP or pRcsD -110 -GFP with pSirA or pVector (empty pACYC184 vector) following 24 h of growth on LB solid medium without (-NaCl) or with (+NaCl) NaCl was monitored. SirA activation is defined by increased fluorescence in the pSirA strain as compared to the pVector strain (see S10

PLOS GENETICS
The phosphorelay BarA/SirA activates the non-cognate regulator RcsB in Salmonella enterica presence of SirA binding sites in csrB and csrC promoters 215 and 181 nts upstream of the csrB and csrC transcription starts sites, respectively [23].
BarA activates RcsB independently of the SirA, RcsC, and RcsD proteins, and of acetyl phosphate Unexpectedly, the barA rcsD double mutant exhibited much lower fluorescence from rprA-gfp than the sirA rcsD double mutants in both LB and NaCl-lacking LB plates (S12 Fig). This result suggests that BarA can activate RcsB by a pathway not involving SirA. In agreement with this notion, a barA sirA rcsC rcsD quadruple mutant strain displayed slightly lower fluorescence from rprA-gfp than the sirA rcsC rcsD triple mutant in both LB and NaCl-lacking LB plates ( Fig 7A). The BarA-expressing plasmid complemented the barA sirA rcsC rcsD quadruple mutant, whereas the vector control (pVector) did not ( Fig 7B). Cumulatively, these results argue that BarA is capable of activating RcsB in a SirA-independent fashion.
BarA appears to activate RcsB directly because immunoprecipitation experiments using BarA-HA and in vitro synthesized proteins showed that RcsB-FLAG was pulled down by antibodies recognizing the HA epitope, indicating that BarA binds to RcsB (Fig 7C). The antibodies recognizing the HA epitope also pulled down SirA-FLAG, which was used as a positive control, but not the response regulator PhoP-FLAG, which was used as a negative control ( Fig  7C). These in vitro results demonstrate that BarA binds to RcsB as well as to SirA.
Next, we examined the ability of BarA to autophosphorylate from adenosine triphosphate (ATP) and to serve as phosphodonor to RcsB in vitro. For these experiments, we used the BarA 198-918 variant because the equivalent E. coli variant does autophosphorylate and serve as phosphodonor to its cognate regulator SirA, and also because, lacking amino acids 1-197 corresponding to the transmembrane segments, BarA 198-918 is more suitable for purification than the full-length BarA protein [5]. The Salmonella BarA 198-918 autophosphorylated and served as phosphodonor to the Salmonella SirA protein (S13 Fig), recapitulating the behavior reported for the E. coli sequelogs [5]. By contrast, RcsB was not phosphorylated from BarA 198-918 , behaving like the negative control PhoP (S13 Fig). Consistent with this result, a BarA 198-918expressing plasmid (pBarA 198-918 ) showed a reduced ability to increase fluorescence from rprA-gfp in wild-type Salmonella, and to rescue the rcsC rcsD sirA barA quadruple mutant as compared to the strains expressing the full-length BarA (S15 Fig). The BarA 198-918 variant does not retain the functionality of the full-length BarA because BarA 198-918 -HA bound SirA-FLAG but not RcsB-FLAG in vitro (Fig 7C). This result may account for the inability of BarA 198-918 to serve as phosphodonor do RcsB in vitro. However, the pBarA 198-918 plasmid promoted low level of RcsB activation in the quadruple barA sirA rcsC rcsD mutant strain (S15 Fig), leaving open the possibility of BarA promoting RcsB phosphorylation by a mechanism that does not require these two proteins to interact.
Given that RcsB can autophosphorylate from acetyl-phosphate (AcP), resulting in transcription of RcsB-activated genes [37,38], and that BarA 198-918 failed to promote RcsB phosphorylation in vitro (S13 Fig), we explored the possibility of BarA promoting RcsB phosphorylation from acetyl-phosphate. However, an ackA-pta mutant, which lacks the ability to synthesize acetyl-phosphate, exhibited the same fluorescence from rprA-gfp as wild-type Salmonella on LB plates with and without NaCl (S14 Fig). In addition, the ackA pta barA triple mutant was less fluorescent than the ackA pta double mutant on NaCl-lacking LB (S14 Fig), recapitulating the difference in fluorescence that exists between the barA mutant and wildtype Salmonella (S14 Fig). These data indicate that acetyl-phosphate does not participate in the BarA activation of RcsB under the investigated conditions.

Agents that damage the outer membrane or cell wall activate RcsB in a BarA-independent manner
Polymyxin B activates RcsB by damaging the outer membrane, and mecillinam by damaging the cell wall [11,22,39]. These chemicals activate RcsB in a BarA-independent manner because a similar increase in fluorescence was observed in wild-type and barA Salmonella carrying the rprA-gfp fusion plasmid treated with polymyxin B (2.4-fold increase for wild-type; p<0.01 and 2.7-fold increase for barA; p<0.01 after 2 h of treatment) or mecillinam (2.6-fold increase for wild-type; p<0.0001 and 3.2-fold increase for barA; p<0.0001 after 2 h of treatment) during exponential growth in LB broth (Fig 8A and 8B). The rcsB, rcsC, and rcsD single mutants, and the rcsC rcsD double mutants showed no significant increase in fluorescence upon addition of either agent (Fig 8A and 8B), in agreement with the notion that RcsB activation by polymyxin B and mecillinam is dependent on both the RcsC and RcsD proteins [22,39]

PLOS GENETICS
The phosphorelay BarA/SirA activates the non-cognate regulator RcsB in Salmonella enterica

BarA is dispensable for activation of the regulators ArcA, OmpR, and PhoP under conditions in which it activates RcsB
BarA appears to specifically control the activity of the regulators SirA and RcsB in LB agar with or without NaCl. This is because similar fluorescence was displayed by wild-type and barA mutant Salmonella harboring transcriptional fusions between a promoterless gfp gene and the promoter of the ArcA-repressed lldP gene (S16 Fig). This was also true for strains harboring gfp transcriptional fusions to the PhoP-activated rstA gene (S17 Fig) and the OmpRactivated ompC gene (S18 Fig). The ArcA, PhoP, and OmpR proteins are response regulators of the same protein family as RcsB and SirA.

Discussion
We have now established that the response regulator RcsB is activated by the non-cognate phosphorelay protein BarA (Fig 1) during rapid growth of Salmonella in LB media, (Fig 2 and  Fig 3). BarA activates RcsB in multiple ways: (i) via its cognate response regulator SirA (Fig 4),

PLOS GENETICS
The phosphorelay BarA/SirA activates the non-cognate regulator RcsB in Salmonella enterica which promotes transcription of the rcsDB operon, thus increasing the abundance of the RcsD and RcsB proteins (Fig 6); (ii) via the SirA-activated regulatory RNAs CsrB and CsrC, which decrease the amounts of the phosphorelay protein RcsC (Fig 5); and (iii) by binding RcsB (Fig  7C), which activates RcsB independently of the RcsC, RcsD, and SirA proteins (Fig 7A and  7B). By contrast, BarA is not required for RcsB activation taking place in bacteria experiencing outer membrane stress or cell wall perturbations (Fig 8), which requires RcsB's cognate phosphorelay proteins RcsC and RcsD (Fig 8) [10]. Our findings indicate that different phosphorelay proteins can act on a given regulator, depending on the environment experienced by a bacterium (Fig 1).
BarA activation of RcsB appears to be largely mediated by the SirA-activated CsrB and CsrC regulatory RNAs. This is because a csrB csrC mutant strain behaved the same way as barA and sirA single mutant strains regarding activation of the RcsB-dependent rprA-gfp fusion (Fig 5A and Fig 5B). There are two non-mutually exclusive possibilities about how these two regulatory RNAs may increase RcsB activity. On the one hand, the csrB csrC double mutant has more RcsC protein that wild-type Salmonella when bacteria are grown in LB broth without NaCl (Fig 5C), suggesting CsrB and CsrC activate RcsB by decreasing RcsC amounts. The higher RcsB activity exhibited by a rcsC mutant strain as compared to wild-type Salmonella in NaCl-lacking LB broth (Fig 3B, left panel) may reflect that RcsC can act as a phosphatase that removes a phosphate from RcsB in the absence of RcsC-activating signals [10,15,21,40]. Therefore, by decreasing RcsC amounts, CsrB and CsrC would reduce RcsB dephosphorylation, resulting in RcsB-dependent gene transcription.
The regulatory RNAs CsrB and CsrC operate by binding to the RNA binding protein CsrA, thereby decreasing the CsrA amounts available to control expression of target RNAs [41,42]. Analysis of the rcsC 5' leader region revealed a putative CsrA binding site upstream of the rcsC start codon (S19 Fig) that partly matches the CsrA consensus recognition sequence [43]. While this analysis suggests that CsrA may directly regulate rcsC translation, the rcsC mRNA was not identified as a CsrA target in a recent genome-wide mapping of CsrA binding sites in Salmonella [29]. That this study was conducted under different growth conditions (i.e., stationary phase in regular LB) from those in which the csrB and csrC mutations increase RcsC amounts (i.e., exponential and stationary phases in LB without NaCl) (Fig 5C) may explain why rcsC was not recovered as a CsrA target. Of course, it is also possible that CsrA promotes RcsC expression via a yet to be identified target.
On the other hand, many mRNAs encoding proteins involved in cell envelope integrity have been identified as CsrA targets in both Salmonella and E. coli [29,44,45]. By titrating CsrA away from these mRNAs, the regulatory RNAs CsrB and CsrC could then affect their translation, potentially increasing RcsB activity.
The full-length BarA protein bound to RcsB as well as to SirA in vitro ( Fig 7C) and activated RcsB in the absence of SirA in vivo (Fig 7A and 7B). That BarA activation of RcsB requires RcsB's site of phosphorylation (S3 Fig) suggests that BarA promotes the phosphorylated state of RcsB. However, a truncated BarA protein failed to bind (Fig 7C), and hence, to serve as phosphodonor to RcsB although it did so to SirA (S13 Fig). This raises the possibility that the full-length BarA protein is capable of promoting in vitro RcsB phosphorylation, as suggested by the increased ability of the pBarA plasmid to activate RcsB as compared to the pBarA 198-918 plasmid (S15 Fig). One may also consider the activation of RcsB by BarA independently of RcsC, RcsD, and SirA proteins (Fig 7A and 7B) to be indirect (Fig 1), requiring a yet to be identified factor essential for BarA to promote RcsB phosphorylation.
That BarA activates RcsB when cells are rapidly growing in LB media suggests that some genes from the RcsB regulon fulfill physiological functions when Salmonella is not experiencing the cell envelope perturbations that activate RcsB via the RcsC and RcsD proteins. In agreement with this notion, BarA impacts RcsB activity during logarithmic growth in LB medium (Fig 3A, 3B and 3C) but not in response to outer membrane stress and cell perturbations (Fig 8).
Activation of a regulator by non-cognate phosphorelay proteins is not exclusive to RcsB. This is because the regulator OmpR is activated both by its cognate sensor EnvZ upon high osmolarity [46] and by the non-cognate sensor ArcB when E. coli experiences anaerobic conditions [47]. ArcB forms an anaerobiosis-responsive phosphorelay with its cognate regulator ArcA [48][49][50]. Thus, OmpR-regulated genes respond to both osmolarity and oxygen availability. Yamamoto and colleagues have also reported several cases of in vitro phosphorylation of regulators by non-cognate sensor kinases [51]. However, the incubation times used for the transphosphorylation reactions in this study do not allow us to determine if the observed phosphotransfers between non-cognate proteins are physiologically relevant, as demonstrated by Laub and colleagues [52]. Two-component system proteins are generally encoded in the same operon [6,53,54], allowing their co-expression and making interactions with non-cognate partners less likely [6,53]. Intriguingly, the rcsB, rcsC, and rcsD genes are transcribed in distinct fashions [3,20]. That is, the rcsC gene is convergently transcribed towards the rcsB gene, which is located downstream of rcsD. Although rcsD and rcsB are part of the same transcription unit, a separate promoter located within the rcsD coding region drives rcsB transcription under conditions distinct from those operating on the rcsD promoter [36]. Likewise, the BarA and SirA proteins are encoded by separate genes in all species reported to carry both barA and sirA [5,23]. Given that BarA activates both SirA and RcsB (Fig 1), and that proteins constituting phosphorelays are often encoded by independently transcribed genes, they may be more prone to establish physiological interactions with additional partners. In line with this hypothesis, Acinetobacter baumannii and Allochromatium vinosum encode BarA but neither SirA nor RcsB [23], suggesting that BarA interacts with a yet to be identified regulator.

Bacterial strains, plasmids, primers and growth conditions
Bacterial strains and plasmids used in this study are listed in the table in S1 Table; oligonucleotide sequences are presented in the table in S2 Table. Single gene knockouts and deletions were carried out as described [55]. Mutations generated by this approach were subsequently moved into clean genetic backgrounds via phage P22-mediated transduction as described [56]. Details of strain constructions are presented in SI Materials and Methods. Unless specified, bacteria were grown in LB medium (10 g/L NaCl) at 37˚C. When required, media were supplemented with ampicillin (50 μg/ml), chloramphenicol (20 μg/ml), kanamycin (50 μg/ml), tetracycline (10 μg/ml).
To generate the rcsBD56Q strain, a first PCR product was generated with primers W4387 and W4388 using the plasmid pSLC-242 [57] as a template. The resulting PCR product was then integrated into the chromosome of wild-type S. enterica (14028s) via the one-step inactivation method [55] using plasmid pKD46. Recombinant cells containing the insertion were selected on LB supplemented with 20 μg/ml chloramphenicol at 30˚C. This insertion was subsequently replaced via a second pKD46-mediated recombination of pre-annealed W4391 and W4392 primers into the chromosome. Cells were recovered for 3 h as described [57] and selected on N-minimal medium agar plates [58] containing 50 μM glutamate, 50 μM histidine, 50 μM leucine, 100 μM methionine, 100 μM glutamine, 10 mM MgCl 2 and 30 mM rhamnose as the sole carbon source. The allele replacement was confirmed by PCR with primers W4389-W4390 followed by DNA sequencing.
When required, pCP20 helper plasmid [55] was used to remove antibiotic-resistance markers.

Construction of plasmids
To construct pRprA-GFP, primers W4304-W4305 were used to amplify rprA -160 to +12 region (relative to rprA transcription start site) using Salmonella 14028s genomic DNA as template. The resulting PCR product was digested with EcoRI and BamHI and ligated into pFPV25 plasmid DNA [59] digested with the same restriction enzymes. The ligation reaction was transformed into DH5α cells by electroporation. The identity of rprA insert was verified by DNA sequencing using primer W804.
To construct pLldP-GFP, primers W4587-W4588 were used to amplify lldP -611 to -103 region (relative to lldP ATG translation start site) using Salmonella 14028s genomic DNA as template. The resulting PCR product was digested with EcoRI and BamHI and ligated into pFPV25 plasmid DNA [59] digested with the same restriction enzymes. The ligation reaction was transformed into DH5α cells by electroporation. The identity of lldP insert was verified by DNA sequencing using primer W804.
To construct pOmpC-GFP, primers W4817-W4818 were used to amplify ompC -330 to -71 region (relative to ompC ATG translation start site) using Salmonella 14028s genomic DNA as template. The resulting PCR product was digested with EcoRI and BamHI and ligated into pFPV25 plasmid [59] digested with the same restriction enzymes. The ligation reaction was transformed into DH5α cells by electroporation. The identity of ompC insert was verified by DNA sequencing using primer W804.
To construct pBarA, primers W4640-W4592 were used to amplify barA -18 to + 2763 region (relative to barA ATG start codon) using Salmonella 14028s genomic DNA as template. The resulting PCR product was digested with EcoRI and ligated into pACYC184 plasmid [60] digested with the same restriction enzyme. The ligation reaction was transformed into DH5α cells by electroporation. The identity of barA insert was verified by DNA sequencing using primers W4596-W4597.
To construct pBarA 198-918 , primers W4954-W4592 were used to amplify barA +594 to + 2763 region (relative to barA ATG start codon) using Salmonella 14028s genomic DNA as template. The resulting PCR product was digested with EcoRI and ligated into pACYC184 plasmid [60] digested with the same restriction enzyme. The ligation reaction was transformed into DH5α cells by electroporation. The identity of barA 198-918 insert was verified by DNA sequencing using primers W4596-W4597.
To construct pSirA, primers W4593-W4595 were used to amplify sirA -19 to + 669 region (relative to sirA TTG start codon) using Salmonella 14028s genomic DNA as template. The resulting PCR product was digested with EcoRI and ligated into pACYC184 plasmid [60] digested with the same restriction enzyme. The ligation reaction was transformed into DH5α cells by electroporation. The identity of sirA insert was verified by DNA sequencing using primers W4596-W4597.
To construct pCsrB, primers W4717-W4718 were used to amplify csrB -9 to +407 region (relative to csrB transcription start site) using Salmonella 14028s genomic DNA as template. The resulting PCR product was digested with EcoRI and ligated into pACYC184 plasmid [60] digested with the same restriction enzyme. The ligation reaction was transformed into DH5α cells by electroporation. The identity of csrB insert was verified by DNA sequencing using primers W4596-W4597.
To construct pQE30-BarA 198-918 , primers W4703-W4704 were used to amplify the barA 198-918 (barA coding region; amino acids 198 to 918) using Salmonella 14028s genomic DNA as template. The resulting PCR product was digested with BamHI and ligated into pQE30 Xa (Qiagen) digested with the same restriction enzyme. The ligation reaction was transformed into E. coli M15 cells by electroporation. The identity of barA 198-918 insert was verified by DNA sequencing using primer W4711.
To construct pQE30-RcsB, primers W4705-W4706 were used to amplify the rcsB coding region using Salmonella 14028s genomic DNA as template. The resulting PCR product was digested with BamHI and ligated into pQE30 Xa (Qiagen) digested with the same restriction enzyme. The ligation reaction was transformed into E. coli M15 cells by electroporation. The identity of rcsB insert was verified by DNA sequencing using primer W4711.
To construct pQE30-SirA, primers W4707-W4708 were used to amplify the sirA coding region using Salmonella 14028s genomic DNA as template. The resulting PCR product was digested with BamHI and ligated into pQE30 Xa (Qiagen) digested with the same restriction enzyme. The ligation reaction was transformed into E. coli M15 cells by electroporation. The identity of sirA insert was verified by DNA sequencing using primer W4711.
To construct pQE30-PhoP, primers W4709-W4710 were used to amplify the phoP coding region using Salmonella 14028s genomic DNA as template. The resulting PCR product was digested with BamHI and ligated into pQE30 Xa (Qiagen) digested with the same restriction enzyme. The ligation reaction was transformed into E. coli M15 cells by electroporation. The identity of phoP insert was verified by DNA sequencing using primer W4711.

Mutagenesis with transposon Tn10dTc
Wild-type Salmonella 14028s was transformed with the Tn10 transposase-expressing plasmid pNK972 [61]. A P22 lysate generated in strain TH338 (F::Tn10dTc) was used to transduce the pNK972-carrying strain to randomly mutagenize the wild-type strain 14028s selecting for resistance to tetracycline. About 10,000 transductants were pooled together to generate a Tn10dTc-generated mutant library. The HS1383 strain (ΔrcsC ΔrcsD) carrying the pRprA-GFP plasmid was then transduced with the mutant library and about 9,000 transductants were screened for decreased fluorescence on LB agar plate. Two transductants displayed decreased fluorescence. The two transductants were infected with phage P22 and the the generated lysates were used to transduce HS1383 strain carrying plasmid pRprA-GFP. In both cases, the transductants displayed the phenotype of the original mutants. The identity of the genes responsible for the change in rprA GFP fusion activity, we determined the nucleotide sequence of the Tn10dTc-chromosome joint as described [62] by performing a first PCR reaction with W4448-W4449 primers from genomic DNA extracted from the mutants with DNeasy Blood & Tissue Kits (Qiagen), and then a second PCR reaction using the product of the first reaction with the W4451-W4452 primers. The final PCR product was then sent for sequencing with W4451-W4452 primers.

Colony plate fluorescence imaging and quantification
Salmonella cells expressing plasmid-borne gfp fusions were streaked onto either LB or on LB without NaCl agar plates supplemented with the appropriate antibiotics. Following 24 h of growth, fluorescence was visualized with a dark blue light transilluminator and an amber screen. For fluorescence quantification, isolated colonies were resuspended in 1 ml of phosphate-buffered saline (PBS) and 150 μl were aliquoted in a clear-bottomed 96-well black plates (Corning). Green fluorescence was measured using an Infinite M1000 plate reader (Tecan) with 485-nm excitation and 535-nm emission, and the absorbance was measured at 600 nm.

Fluorescence measurement in liquid cultures
Overnight cultures of Salmonella in LB were diluted 1/1000 in LB or LB without NaCl supplemented with the appropriate antibiotics. Cells were grown in clear-bottomed 96-well black plates (Corning) using a SpectraMax Plus Microplate Reader (Molecular Devices) for agitation. Green fluorescence was measured using a Synergy H1 plate reader (BioTek) with 485-nm excitation and 535-nm emission, and the absorbance in each well was measured at 600 nm.

Pull-down assays with proteins synthesized using an in vitro transcriptiontranslation system
Pull-down assays were performed as previously described [63] with some modifications. Proteins were produced from DNA templates by in vitro synthesis using the PURExpress system (New England Biolabs). To synthesize the DNA templates, primers W4598-W4599 (barA-HA), W4798-W4599 (barA 198-918 -HA), W4608-W4609 (rcsB-FLAG), W4623-W4624 (sirA-FLAG), and W4621-W4622 (phoP-FLAG) were used. Synthesized proteins were mixed in 500 μl of tris-buffered saline (TBS) containing proteoliposomes (0.12 mg/ml) and incubated at room temperature for 2 h. Samples were then pulled-down with anti-HA magnetic beads (Thermo Scientific) at room temperature for 2 h. Samples were then analyzed by Western blot with antibodies directed to the FLAG (Abcam) or HA (Sigma) epitopes.

Polymyxin B and mecillinam assays in liquid cultures
Overnight cultures of Salmonella strains in LB were diluted 1/1000 in LB supplemented with the appropriate antibiotics. Cells were grown into clear-bottomed 96-well black plates until early exponential phase (OD 600 of~0.15) and sublethal concentrations of polymyxin B (0.5 μg/ ml) or mecillinam (10 μg/ml) were added. Water was added to untreated samples. Fluorescence was then monitored for a period of 2 h following the addition of the antimicrobial agents.

Protein expression and purification
E. coli M15 cells co-transformed with pREP4 and the appropriate pQE30 derivatives expressing N-terminally His6-tagged proteins were grown in 250 ml of LB broth supplemented with 100 μg/ml ampicillin and 25 μg/ml kanamycin. Expression of the His6-tagged proteins was induced at mid-exponential phase (A 600nm of 0.5) by the addition of 2 mM isopropyl-β-D-thiogalactopyranoside (IPTG). Cultures were then grown at 30˚C for 4 h and cells were collected by centrifugation (5000 x g at 4˚C for 20 min). Pellets were then washed with 10 ml of cold PBS (5000 x g centrifugation at 4˚C for 20 min) and were stored at -20˚C. His6-tagged proteins were purified from cell pellets using Ni-NTA Fast Start Kit (Qiagen) according to the manufacturer's instructions. The purified proteins were diluted 1/30 in 14 ml of kinase buffer (33 mM HEPES pH 7.5, 50 mM KCl, 5 mM MgCl 2 , 1 mM DTT, 0.1 mM EDTA, 10% glycerol) and were concentrated in Ultra-15 Centrifugal Filter Unit (Amicon). Protein concentrations were determined by direct A280 measurement of the concentrated samples.

Phosphotransfer profiling
Phosphorylation assays were performed as previously described [5] with some modifications. Purified BarA protein (0.96 μM final) was first incubated in kinase buffer (33 mM HEPES pH 7.5, 50 mM KCl, 5 mM MgCl 2 , 1 mM DTT, 0.1 mM EDTA, 10% glycerol) in the presence of 15 Ci/mmol of [γ-32 P]ATP (PerkinElmer) for 30 minutes at room temperature. Purified RcsB, SirA or PhoP proteins were then added to the mixture at final concentrations of 4 and 7 μM (final reaction volume of 10 μl) and the incubation was pursued for 15 min. Reactions were stopped by the addition of 10 μl of 2X LDS sample buffer and samples were run, without heating, on NuPAGE TM 4-12% bis-tris protein gels (ThermoFisher Scientific). The gels were dried at 80˚C for 30 min and exposed to a phosphor screen overnight.

Western blot assay
Cells were grown in LB broth media with or without NaCl. To extract total proteins, cells were precipitated with trichloroacetic acid (5% total volume) and washed with 80% acetone. Samples were resuspended in NuPAGE TM LDS sample buffer (ThermoFisher Scientific) and normalized according to the OD 600 . Protein samples were run on NuPAGE TM 4-12% bis-tris protein gels (ThermoFisher Scientific) and transferred to nitrocellulose membrane using iBlot Gel Transfer Device (ThermoFisher Scientific). Membranes were blocked with 5% milk solution in TBST for 1 h. Membranes were probed with 1:5000 dilution of mouse anti-FLAG (Sigma), rabbit anti-HA (Sigma) or mouse anti-RpoB (BioLegend). Secondary horseradish peroxidase-conjugated anti-rabbit (GE healthcare) or anti-mouse (Promega) was used at 1:5000 dilution. The blots were developed with the Amersham ECL Western blotting detection reagents (GE Healthcare) or SuperSignal West Femto chemiluminescent system (Pierce). Images were acquired with LAS-4000 imager (GE Healthcare). Images were quantified using ImageLabTM software (Biorad).

In vivo detection of phosphorylated RcsB
Cells were grown in LB broth media with or without NaCl. Whole-cell extracts were prepared as previously described [64]. Samples were run on 12.5% polyacrylamide gels containing acrylamide-Phos-tag ligand (Wako Laboratory Chemicals) in standard running buffer [0.4% (w/v) SDS, 25 mM tris, 192 mM glycine] at 150 V at 4˚C for 4 h, transferred to nitrocellulose membranes, and analyzed by immunoblotting using polyclonal rabbit antibodies recognizing RcsB (a kind gift from Anna Vianney, Centre de Recherche en Infectiologie, INSERM, Lyon, France) (1:1000) and polyclonal mouse antibodies recognizing AtpB (Abcam) (1:5000). The blots were developed with the SuperSignal West Femto chemiluminescent reagents (Pierce). Images were acquired with LAS-4000 imager (GE Healthcare).

Primer extension analysis
Cells were grown in LB broth and total RNA was extracted using the hot phenol procedure as previously described [65]. Primer extension reactions were then performed as previously described [66] using 20 μg of total RNA and the primer W4171 annealing with rcsD coding region. Primer extension reactions were run together with a template-specific sequencing ladder generated with primer W4171 and a DNA template corresponding to rcsD -228 to +38 region (relative to rcsD ATG start codon) amplified with primers W3228-W3229 using Salmonella 14028s genomic DNA as template.