The Hybrid Histidine Kinase LadS Forms a Multicomponent Signal Transduction System with the GacS/GacA Two-Component System in Pseudomonas aeruginosa

In response to environmental changes, Pseudomonas aeruginosa is able to switch from a planktonic (free swimming) to a sessile (biofilm) lifestyle. The two-component system (TCS) GacS/GacA activates the production of two small non-coding RNAs, RsmY and RsmZ, but four histidine kinases (HKs), RetS, GacS, LadS and PA1611, are instrumental in this process. RetS hybrid HK blocks GacS unorthodox HK autophosphorylation through the formation of a heterodimer. PA1611 hybrid HK, which is structurally related to GacS, interacts with RetS in P. aeruginosa in a very similar manner to GacS. LadS hybrid HK phenotypically antagonizes the function of RetS by a mechanism that has never been investigated. The four sensors are found in most Pseudomonas species but their characteristics and mode of signaling may differ from one species to another. Here, we demonstrated in P. aeruginosa that LadS controls both rsmY and rsmZ gene expression and that this regulation occurs through the GacS/GacA TCS. We additionally evidenced that in contrast to RetS, LadS signals through GacS/GacA without forming heterodimers, either with GacS or with RetS. Instead, we demonstrated that LadS is involved in a genuine phosphorelay, which requires both transmitter and receiver LadS domains. LadS signaling ultimately requires the alternative histidine-phosphotransfer domain of GacS, which is here used as an Hpt relay by the hybrid kinase. LadS HK thus forms, with the GacS/GacA TCS, a multicomponent signal transduction system with an original phosphorelay cascade, i.e. H1LadS→D1LadS→H2GacS→D2GacA. This highlights an original strategy in which a unique output, i.e. the modulation of sRNA levels, is controlled by a complex multi-sensing network to fine-tune an adapted biofilm and virulence response.


Introduction
The ability of bacteria to survive in specific habitats requires coordination of appropriate gene expression in response to encountered environmental changes. It is interesting to note that the complexity of bacterial regulatory networks and the number of regulatory genes of bacterial genomes proportionally increase with the diversity of environments a bacterial species is able to survive [1]. In order to cope with the various environments they encounter, bacteria have evolved several sensing systems, including two-component systems (TCS) that monitor external and internal stimuli (nutrients, ions, temperature, redox states . . .), and translate these signals into adequate adaptive responses (for a review see, [1]).
A TCS comprises a histidine kinase (HK) protein or sensor mostly inserted into the inner membrane and a cognate partner known as the response regulator (RR). These two proteins function in their simplest version in a two-step phosphorelay mechanism, forming a classical TCS as follows: The detection of the stimulus by the periplasmic or cytoplasmic detection domain of the HK protein triggers autophosphorylation on a conserved histidine (H) residue of the transmitter domain H1. The phosphoryl group is then transferred on a conserved aspartate (D) residue present in the receiver or D domain of the cognate RR [2,3]. In some cases, the phosphorelay mechanism between the HK and the RR requires a four-step phosphorelay. In this case, the HK requires additional domains such as a receiver domain (D1). Although this D1 module could be on a separate protein, it is mostly fused to the HK. Then, an alternative histidine-phosphotransfer domain (Hpt or H2) can either be fused to the HK (H2) or form a third independent component in the cytoplasm called Hpt. The HK carrying both additional D1 and H2 domains are called unorthodox sensors while those carrying only the D1 domain are called hybrid sensors. Autophosphorylation of the first H residue of the H1 domain of hybrid or unorthodox HK initiates a phosphorelay such as H H1 !D D1 !H H2 or Hpt !D RR . In a few cases, the phosphorelay between the HK and the RR can be more complex and involves another TCS to form a multicomponent signal transduction system. The CblSTR signal transduction pathway of Bukholderia cenocepacia, which controls the expression of the cable pili, is such a system [4]. The cblS and cblT genes encode a hybrid and an unorthodox HK respectively, while the cblR gene encodes the cognate RR. While the first two steps of the phosphorelay require the H1 and D1 domains of CblS, the transphosphorylation of the D domain of CblR by CblS requires the H2 domain of CblT, which serves as a bridge component, increasing the complexity of the transduction pathway.
Pseudomonas aeruginosa is a major human pathogen causing severe infections in vulnerable patients such as those with cystic fibrosis or hospitalized with cancer, burns and in intensive care units. It has become a major cause of nosocomial infections. Like other species, P. aeruginosa is able to switch from a planktonic (free swimming) to a sessile (biofilm) lifestyle and several TCSs play a key role in this switch [5][6][7]. Free swimming cells are characterized by an effective production and injection into host cells of effectors of the Type III secretion system (T3SS). In this free swimming lifestyle, they are thought to represent the vast majority of individuals causing acute infections [8] such as in sepsis, ventilator-associated pneumonia, and infections in postoperative wound and burn patients. In contrast, sessile cells are embedded in a biofilm community sealed by a matrix of exopolysaccharides (EPS) and DNA. In this state, the bacteria concomitantly secrete toxins delivered by the H1-Type VI secretion system (H1-T6SS), which are used for killing and competing with other species in this crowded and enclosed community [8][9][10][11]. Cells in biofilms are thought to be in conditions similar to those in chronic infections [12] such as in chronic obstructive pulmonary disease or cystic fibrosis. Several studies have reported an opposing regulation between the expression of molecular determinants involved in acute infection and those involved in chronic infection. Several TCSs have been described as key players controlling this transition, including the central and critical GacS/GacA TCS [13][14][15][16][17] (Fig 1A). GasS is an unorthodox HK with H1/D1/H2 domains. GacA is an RR functioning as a transcriptional regulator, which positively and exclusively controls the expression of two unique target genes encoding two small noncoding RNAs, RsmY and RsmZ [18]. Thus, RsmY and RsmZ have been proposed as key players in controlling the switch between planktonic and biofilm lifestyles [18,19]. These two sRNAs sequester the RNA-binding translational repressor RsmA and thus relieve RsmA binding from its target mRNAs. While bound to target sequences at the site of translational initiation, RsmA exerts a direct translational repression on a limited number of genes grouped in six operons [20], among which are genes encoding the H1-T6SS. Additionally, RsmA has been described as indirectly and positively controlling the expression of a substantial number of genes, including those encoding the T3SS participating in acute infection [20,21]. High expression of rsmY and rsmZ leads to massive biofilm formation due to the production of Pel EPS, and is coupled with H1-T6SS induction and T3SS repression ( Fig 1A) [22]. In a gacS mutant, the absence of expression of these sRNAs results in an impaired biofilm formation and induction of T3SS expression [22]. In P. aeruginosa, expression of these two sRNAs is controlled by a complex and sophisticated regulatory network involving the GacS/GacA TCS but also other TCS pathways. The RetS hybrid HK represses expression of both rsmY and rsmZ genes by interfering with the GacS/GacA TCS activity [23]. The PA1611 hybrid HK induces expression of both rsm genes by counteracting the interfering effect of RetS on GacS [24]. The HptB regulatory pathway, which also intersects with the GacS/GacA TCS, only induces rsmY gene expression [22]. The LadS hybrid HK carrying H1/D1 domains has been shown to activate expression of the rsmZ gene. However, it controls target genes in a reciprocal manner as compared to RetS [25], suggesting that it may also control rsmY gene expression although this has not been demonstrated.
While the GacS/GacA TCS is widely distributed throughout the bacterial kingdom, the molecular switch formed by the hybrid LadS, PA1611 and RetS HKs is unique to the Pseudomonas species, though it can function in very different ways in phylogenetically related Pseudomonas species [26][27][28]. In Pseudomonas fluorescens, it has been proposed that LadS controls rsmX, rsmY and rsmZ expression through GacS, based on the observation that gacS or gacA mutations are epistasic to ladS mutation [29]. In P. syringae, the LadS, GacS and RetS HKs do not control the same targets and this is exemplified for T3SS whose LadS-and RetS-dependent control is GacS-independent in this bacterial species [26,27]. The absence of the D1 domain in P. syringae LadS HK may account for this GacS-independent T3SS regulation [27].
In this network (Fig 1A), the RetS regulatory pathway requires the presence of the GacS/ GacA TCS to control rsm gene expression and this occurs via heterodimer formation between RetS HK and GacS HK [23,24,30], impeding GacS autophosphorylation and thereby preventing phosphorylation of its cognate RR GacA [23]. In P. aeruginosa, the PA1611 hybrid HK structurally related to GacS also interacts with RetS in P. aeruginosa in a very similar manner to GacS and RetS and its action is independent of LadS HK [24]. Furthermore, this interaction does not require the conserved phosphorelay residues of PA1611 [24]. Overall, it is still unclear whether in P. aeruginosa, LadS HK triggers rsmZ and possibly rsmY gene expression through GacS or another unorthodox HK or an Hpt protein connected to another TCS [25].
In the present study, we therefore addressed the questions of whether the P. aeruginosa LadS hybrid HK intersects with the GacS/GacA pathway and at which level by using combined genetic, biochemical and phenotypic approaches. We demonstrated that the LadS HK not only controls the expression of rsmZ but also the expression of rsmY, and thus modulates the production of Pel EPS, H1-T6SS and T3SS targets. We further demonstrated that LadS influences its target genes through the GacS/GacA TCS. Specifically, LadS autophosphorylates on its H1 domain, transfers the phosphoryl group on its D1 domain, and then subsequently to the H2 domain of GacS ( Fig 1B). These results clearly showed that the LadS HK and the GacS/GacA TCS form a multicomponent signal transduction system, functioning in a mechanism clearly distinct from the one proposed for the RetS or the PA1611 hybrid HKs, the other members of the network.

Results
LadS HK signaling converges on RsmY or RsmZ to control its targets In P. aeruginosa, the LadS HK had been shown to trigger rsmZ gene expression [25] but no data were available for its action on rsmY gene expression. We thus first addressed the question of whether LadS is able to control rsmY gene expression.
First, to override variable levels of ladS expression and activation (phosphorylation state) due to LadS signal, overexpression of ladS was first undertaken. This approach mimics a constitutive activation of HKs thereby enabling the study of the corresponding signaling pathway. Using chromosomal rsmY-lacZ and rsmZ-lacZ transcriptional fusions in the PAK genetic background (PAKattB::rsmY-lacZ and PAKattB::rsmZ-lacZ), we overexpressed the full-length ladS HK gene using the pBBRladS plasmid [25]. Overexpression of ladS resulted in a significant increase in activity of both rsm fusions (S1A Fig). Maximal rsm gene expression was reached for both fusions at an OD 600nm of around 3.8 with an 82-fold and a 42-fold increase for rsmY and rsmZ, respectively, upon ladS overexpression. These results were confirmed by RT-qPCR (S1B Fig). Because overexpression of genes encoding HKs can have adverse effects [31], we further assessed by RT-qPCR whether LadS produced at the chromosomal level could have the same impact on rsmY and rsmZ expression in the PAK and its isogenic mutant PAKΔladS strains. Levels of rsmY and rsmZ expression were, respectively, reduced by a 17-fold and an 11-fold The multicomponent signal transduction system made of the LadS hybrid HK and the GacS/GacA TCS. In the presented model sustained by results obtained in the present study, this multicomponent signal transduction system made of the LadS hybrid HK and the GacS/GacA TCS forms a multipleinput system probably reflecting the variability of environmental conditions P. aeruginosa is faced with and may result in a range of gradations of chronic infection. IM (Inner Membrane), P (Periplasm), C (Cytoplasm).
doi:10.1371/journal.pgen.1006032.g001 factor in the ladS mutant as compared to the wild-type strain (S1B Fig). This suggests that the ladS gene is indeed expressed and that LadS signal, although unknown, is present in our testing conditions. This further proved that LadS controls rsmY and rsmZ gene expression at the basal level. The basal level of ladS expression in the wild-type strain PAK was further determined by RT-qPCR. An absolute number of 4,600 copies of ladS gene mRNA copies per μg of total RNA retrotranscribed was monitored in the wild-type PAK strain bearing or not the mild copy empty vector pBBRMCS4 and no copy was detectable in its counterpart ladS mutant. This number increased to 130,000 copies when ladS gene was expressed from the pBBRMCS4 vector (S1C Fig). From these results, overexpression of ladS with a mild copy vector appeared to be a suitable way to investigate the LadS signaling pathway: it can reproduce LadS signaling without adverse effect since the increase in rsm gene expression observed in ladS overexpression conditions was consistent with the one obtained under ladS basal expression (S1A and S1B Fig).
LadS was shown to antagonistically control several of the RetS targets such as Pel, T6SS and T3SS [25]. We next investigated whether it exclusively occurs through the two sRNAs RsmY and RsmZ. The pBBRladS and pBBRMCS4 plasmids were separately conjugated in the PAK, PAKΔrsmY, PAKΔrsmZ and PAKΔrsmYΔrsmZ strains. Overexpression of the ladS HK gene resulted in a 2.65-, 4.3-and 2.8-fold increase in biofilm formation in PAK and in the single rsmY and rsmZ mutants, respectively, while the double mutation abrogated biofilm formation (Fig 2A). Since in the PAK background the biofilm built-in response to the activation of the rsm genes is mostly dependent on pel gene expression [10,22], we further investigated whether the ladS-dependent biofilm formation could rely on the transcriptional activity of the pel locus. The activity of the chromosomal pelA-lacZ transcriptional fusion was therefore assessed in the PAK, PAKΔrsmY, PAKΔrsmZ and PAKΔrsmYΔrsmZ strains transformed with either the pBBRladS or the pBBRMCS4 plasmids. Upon ladS overexpression, the β-galactosidase activity of the pelA transcriptional fusion measured after 4 hours of growth (OD 600nm %3.5) was significantly induced in the wild-type strain, intermediately induced in the rsmY or rsmZ single mutants and abolished in the double rsmYrsmZ mutant (Fig 2B). Another known LadS target is the H1-T6SS, whose production was further checked by immunodetection of the VgrG1 proteins. While ladS HK overexpression induced production of VgrG1 proteins in the PAK strain, these T6SS proteins were undetectable in the PAKΔrsmY, PAKΔrsmZ and PAKΔrsmYΔrsmZ strains ( Fig 2C). Finally, LadS control on T3SS was checked using a chromosomal exoS-lacZ transcriptional fusion in the PAK, PAKΔrsmY, PAKΔrsmZ and PAKΔrsmYΔrsmZ strains, which had received the pBBRladS plasmid or the corresponding empty vector pBBRMCS4. Cells were grown in the presence of EDTA, a Ca 2+ chelator, a condition known to activate T3SS expression. Upon ladS overexpression, the β-galactosidase activity of the exoS fusion was reduced 2-fold in the wild-type strain, 1.65-fold in the rsmY mutant and 1.9-fold in the rsmZ mutant but had no effect in the double rsmYrsmZ mutant (Fig 2D). These ladS-overexpression effects on pelA, vgrG1b and exoS genes were confirmed by RT-qPCR (S1B Fig). The LadS signaling pathway was found to exert the same control on these targets (Pel, T6SS and ExoS) for ladS expression at the basal chromosomal level (S1B Fig). Taken together, these results demonstrated that like RetS, LadS HK controls the expression of both rsmY and rsmZ genes and impacts biofilm formation and Pel production, H1-T6SS and T3SS. Since these sRNAs are the exclusive targets of the GacA RR, our results strongly suggest that such control could occur through the GacS/GacA pathway.

LadS signaling pathway requires GacS and GacA
We next investigated whether in P. aeruginosa, LadS signaling converges on the GacS/GacA TCS, as previously demonstrated for RetS [16,23] and HptB [22] signaling pathways. We first Overexpression of ladS was no longer able to induce rsmY or rsmZ expression in both mutants (Fig 3A), indicating that LadS control of rsm gene expression occurs through the GacS/GacA TCS. This was confirmed by examining phenotypes highly dependent on GacS/ GacA and on the sRNAs RsmY and RsmZ. As illustrated in Fig 3B, the ability of LadS to promote biofilm in the wild-type strain was fully abolished in both gacS and gacA mutants. Similarly, LadS was unable to control the expression of pel, vgrG1b and exoS genes in gac mutants ( Fig 3C). These results confirmed that in P. aeruginosa, LadS-dependent induction of rsmY and rsmZ gene expression and of their targets is strictly dependent on the GacS/GacA TCS.

H1 domain of the LadS HK does not heterodimerize with H1 domains of GacS or RetS HK
It was shown earlier that RetS forms a heterodimer via its H1 domain, with the H1 domain of GacS, and that such heterodimerization prevents GacS autophosphorylation independently of any phosphorelay residue of RetS HK [23]. Because of the antagonism between RetS and LadS HK, we next examined whether LadS HK could form an active heterodimer with GacS HK or an inactive heterodimer with RetS HK counteracting the inhibitory effect of RetS HK on GacS HK by using pull-down and two-hybrid experiments.
The  [23,24,32]. The absence of heterodimer formation involving the H1 domain of LadS HK was further confirmed by using two-hybrid experiments. As shown in Fig 4B, the H1 domain of LadS HK was unable to interact with either the H1 domain of GacS HK or the H1 domain of RetS HK. The H1 domain of each HK, LadS, RetS and GacS was able to homodimerize and interaction was also observed between H1 domains of RetS HK and GacS HK reflecting RetS/GacS heterodimerization as previously reported [23] (Fig 4B shows strains having received both vectors on X-gal-containing plates as well as corresponding levels of measured ß-galactosidase activities). Altogether, these results demonstrate that the H1 domain of LadS HK does not form heterodimers with H1 domains of RetS and of GacS, while H1 domains of RetS and GacS HKs form homo-and heterodimers.
PAKΔrsmYΔrsmZ strains. (A) Biofilm production in glass tubes was illustrated (upper panel) and quantified after Crystal Violet-staining (lower panel). Corresponding levels of biofilm production represent mean values and standard deviations obtained from three independent experiments. (B) Activity of the pelA-lacZ transcriptional chromosomal fusion was monitored in the same strains with the pBBRladS plasmid containing the ladS HK gene (dark violet bars) and the pBBRMCS4 corresponding empty cloning vector (light violet bars) after 4 hours of growth (OD 600nm %3.5). Corresponding β-galactosidase activities are expressed in Miller units and correspond to mean values (with error bars) obtained from three independent experiments. (C) Production of the H1-T6SS VgrG1s proteins was detected in whole cell extracts using western blot with an anti-VgrG1 polyclonal antibody. Numbers on the left side correspond to molecular weight standards (kDa). (D) Activity of the exoS-lacZ transcriptional chromosomal fusion was monitored in the same strains with the pBBRladS plasmid containing the ladS HK gene (dark royal blue bars) and the pBBRMCS4 corresponding empty cloning vector (light royal blue bars) after 6 hours of growth (OD 600nm %4). Corresponding βgalactosidase activities are expressed in Miller units and correspond to mean values (with error bars) obtained from three independent experiments. Wilcoxon-Mann-Whitney tests were performed and *, **, *** and ns referred to p<0.05, p<0.01 and p<0.001 and nonsignificant difference, respectively. LadS signaling pathway requires functional H1 and D1 domains We next tested whether functional H1 and D1 domains of the LadS hybrid HK are required for the LadS-dependent signaling pathway. For that purpose, we engineered a truncated Histagged version of the whole LadS cytoplasmic part of the HK including the H1 and D1 domains   Fig 3A) while the LadSH1 H!Q D1 and LadSH1D1 D!A versions were unable to do so ( Fig 5A, middle panel). Moreover, the wild-type LadSH1D1 version was able to induce the production of VgrG1 proteins while the LadSH1 H!Q D1 and LadSH1D1 D!A versions were not ( Fig 5A, lower panel). To confirm these results, we directly engineered the same point mutations in the ladS gene of the PAK strain, leading to the PAKladSH1 H!Q D1 and PAKladSH1D1 D!A strains. Equivalent results were obtained with point chromosomal ladS mutants of the full-length ladS gene and ladS mutant for biofilm (Fig 5B), expression of rsmY, rsmZ, T6SS-H1 and T3SS genes ( Fig 5C) and EPS production ( Fig 5D). Taken together, these results demonstrated that the LadS signaling pathway requires both the histidine residue of its H1 domain and the aspartate residue of its D1 domain to activate rsm gene expression and further target genes.

LadS requires only the H2 domain of GacS to control its target genes
Since we evidenced that LadS requires its phosphorelay residues for signaling, we next investigated whether such phosphorelay requires the H2 domain of GacS. To test this, we engineered a truncated version of the GacS HK formed from its H2 domain (GacSH2) and its counterpart version GacSH2 H!Q in which the histidine residue in position 859 was mutated into a glutamine residue (S3 Fig), the corresponding H863 residue being crucial for GacS H2 domain activity in P. fluorescens CHA0 [33]. The corresponding gacSH2 and gacSH2 H!Q gene versions as well as the gacS full-length gene (gacS FL ) were introduced into the chromosome of a gacS mutant at the miniTn7 site in which we further introduced the pBBRladS or the corresponding empty vector. GacSH2 was able to trigger a % 600-and a 800-fold induction of rsmY and rsmZ transcript levels in cells overexpressing ladS as compared to cells carrying the corresponding empty vector, respectively. In contrast, GacSH2 H!Q could not transduce LadS signaling ( Fig  6A). Upon ladS overexpression, the production of GacSH2 was sufficient to induce biofilm formation (Fig 6B), pel expression assessed by RT-qPCR (Fig 6C), T3SS gene repression ( Fig 6C) and production of T6SS proteins (Fig 6D) while the GacSH2 H!Q had no ability to affect the rsm-dependent phenotypes tested here. Interestingly, when the gacS FL gene was introduced in the gacS mutant, the complementation level of each rsm-dependent phenotype under ladS overexpression was similar to that obtained with the gacSH2 gene version. As these results were obtained upon LadS overproduction, we further addressed the question of whether the LadS phosphotransfer to GacS occurs at the natural levels of expression of LadS and GacS HKs. We thus engineered the mutations gacSH1 H!Q H2 and gacSH1 H ! Q H2 H ! Q in the wild-type and ladS mutant strains to disable autophosphorylation of the GacSH1 domain and the functionality of the GacsH2 domain, respectively. These strains were evaluated for their capacity to control the expression of rsmY, rsmZ, T6SS, pelA and T3SS genes by RT-qPCR ( Fig 6E). GacSH1 domain autophosphorylation disabled in the strain PAKgacSH1 H!Q led to a reduction of rsmY, rsmZ, pelA and T6SS gene expression and an induction of T3SS gene expression as compared to the PAK strain. In this genetic context, further alteration of GacSH2 domain

In vitro transphosphorylation of GacS H2 domain by LadS HK
The results presented above strongly suggest that LadS HK could use the H2 domain of GacS HK to activate rsm gene transcription, probably via a transphosphorylation mechanism. To further confirm this, in vitro phosphorylation experiments were performed. The C-terminal His-tag forms of LadSH1D1, LadSH1D1 D!A , GacSD1, LadSD1, GacSH1D1, GacSH2 , GacSH2 H!Q and HptA proteins were produced in E. coli and purified close to homogeneity from soluble fractions with nickel affinity columns and subjected to autophosphorylation assays by using [γ- 32 (Fig 7B, left panel). When GacSH2 was further added, we observed that the GacSH2 domain can only receive phosphate for the LadSH1D1 Q ! A / LadSD1 domain and for the GacSH1/GacSD1 domain combinations (Fig 7B, right panel), thereby finally proving that LadS signaling is H1 LadS !D1 LadS !H2 GacS !D2 GacA. From this last experiment, it clearly appeared that GacSH2 phosphorylation was less effective through LadS HK than through GacS HK. As the sampling was done at the same time in both experiments, this strongly suggests that the GacS signaling is faster than the LadS signaling. To confirm this observation, we performed kinetic experiments and clearly observed that phosphotransfer to the GacSH2 domain presented (upper panel) and quantified after crystal violet-staining and extraction (lower panel). Corresponding levels of biofilm production represented by mean values and standard deviations were obtained from three independent experiments. Wilcoxon-Mann-Whitney tests were performed and *, **, *** and ns referred to p<0.05, p<0.01 and p<0.001 and nonsignificant difference, respectively. (C) Transcript levels of RsmY (blue bars), RsmZ (brick-red-colored bars), VgrG1b (T6SS) (green bars) and ExoS (T3SS) (royal blue bars) were monitored in PAK, PAKΔladS and in point chromosomal mutants PAKladSH1 H!Q D1 and PAKladSH1D1 D!A strains using RT-qPCR. Fold induction was presented for the three mutant strains as compared to the PAK strain. Moderated t-tests were performed; *, ** and *** referred respectively to p<0.05, p<0.01 and p<0.001. (D) Activity of the pelA-lacZ (violet bars) transcriptional chromosomal fusion was monitored after 6 hours of growth (OD 600nm %5) and corresponding β-galactosidase activities are expressed in Miller units and correspond to mean values (with error bars) obtained from three independent experiments. Statistical tests were performed and ** referred to p<0.01.
doi:10.1371/journal.pgen.1006032.g005 occurred 0.5-1 min earlier with the GacSH1D1 domain compared to the LadSH1D1 domain as a phosphodonor. This result suggests that the LadS hybrid HK forms a multicomponent signal transduction pathway with the GacS/GacA TCA and adds a supplementary level of regulation triggering chronic infection (Fig 8).

Discussion
In the present study, we demonstrated that in P. aeruginosa the LadS hybrid HK forms a multicomponent signal transduction system with the GacS/GacA TCS. This multicomponent signal transduction system involves an H1 LadS !D1 LadS !H2 GacS !D2 GacA signaling pathway, first through a phosphorelay between the H1 and D1 domains of LadS HK, and thus a transphosphorylation of the H2 domain of GacS HK. This LadS signaling pathway triggers both rsmY and rsmZ gene expression, the two sole direct targets controlled by the RR GacA [18] and further targets such as biofilm formation, H1-T6SS and T3SS expression.
As suggested or observed in other very closely phylogenetically related species to P. aeruginosa such as P. fluorescens [29] but never demonstrated in P. aeruginosa [7,25], our results showed that LadS controls both rsmY and rsmZ gene expression. These results also highlight the fact that LadS signaling occurs through the GacS/GacA TCS, reinforcing the notion that LadS and RetS HK reciprocally regulate the virulence factors under the rsm gene dependency [23,25]. This reciprocal regulation of LadS and RetS converging on GacS seems to be specific to some pseudomonas species such as P. aeruginosa or P. fluorescens but not generalizable to all, since in P. syringae the LadS, GacS and RetS HK do not control the same targets. This is exemplified for T3SS whose LadS-and RetS-dependent control is GacS-independent in this bacterial species [26,27]. In P. syringae, the absence of D1 domain in LadS HK may account for this GacS-independent T3SS regulation [27], while the P. aeruginosa LadS version without its D1 domain is nonfunctional (S5 Fig). Thus, in P. aeruginosa, LadS HK may control T3SS through its D1 domain and GacS HK. Another example of the crucial involvement of the H1D1 subdomain of P. aeruginosa LadS HK in LadS signaling is provided by the observed specific and natural ladS mutation in the PA14 strain, a 49 nucleotide duplication that leads to possible frameshift and results in a truncated and nonfunctional LadS protein lacking H1 and D1 domains [34]. In PA14, this nonfunctional mutation of the ladS gene can be reversed by transcomplementation with the PAK ladS gene.
The RetS hybrid HK [16,35,36] control of GacS involves a heterodimer formation between their H1 domains, which leads to suppression of GacS autophosphorylation in a phosphorelayindependent manner [23]. RetS conserved phosphorelay residues have been found dispensable [23] or not fully required [30] to control GacS, probably depending on the genetic background  ) and RsmZ (brick-red-colored bars) transcript levels were monitored using RT-qPCR and fold induction was presented in the strains PAKΔgacS::miniTn7gacSH2 (gacSH2) and PAKΔgacS::miniTn7gacSH2 H!Q (gacSH2 H!Q ) as compared to the PAKΔgacS::miniTn7 strain (miniTn7). (B) Biofilm production in glass tubes was illustrated (upper panel) and quantified after crystal violet-staining (lower panel). Corresponding levels of biofilm production represent mean values and standard deviations obtained from three independent experiments. Wilcoxon-Mann-Whitney tests were performed; ** and ns referred to p<0.01 and nonsignificant difference. C. PelA (violet bars) and ExoS (royal blue bars) transcript levels were monitored using RT-qPCR and fold induction was presented in the strains PAKΔgacS::miniTn7gacSH2 (gacSH2) and PAKΔgacS::miniTn7gacSH2H!Q (gacSH2H!Q) as compared to the PAKΔgacS::miniTn7 strain (miniTn7). (D) Production of the H1-T6SS Hcp1 proteins was detected in whole cell extracts using western blot with an anti-Hcp1 polyclonal antibody. Numbers on the left side are molecular weight standards (kDa). Moderated t-tests were performed and *, **, *** and ns referred to p<0.05, p<0.01 and p<0.001 and nonsignificant difference, respectively. (E) Transcript levels of RsmY (blue bars), RsmZ (brick-red-colored bars), VgrG1 (green bars), PelA (violet bars) and ExoS (royal blue bars) were monitored using RT-qPCR. Fold induction was presented in the strains PAK, PAKΔladS, PAKgacSH1 H!Q , PAKgacSH1 H!Q ΔladS, PAKgacSH1 H ! Q H2 H ! Q and PAKgacSH1 H ! Q H2 H ! Q ΔladS in order to disable autophosphorylation of the GacSH1 domain and the functionality of the GacsH2 domain, respectively. Moderated t-tests were performed and *, **, *** and ns referred to p<0.05, p<0.01 and p<0.001 and nonsignificant difference, respectively.  This confirms that the LadS signaling pathway exclusively use the Hpt module (the H2 domain) of the GacS unorthodox HK to control its targets.
Once the GacS/GacA signaling pathway is activated upon a signal that remains to be identified, P. aeruginosa is engaged in a chronic infection lifestyle characterized by biofilm and H1-T6SS production and the shutting down of the T3SS. The multicomponent signal transduction system made of the LadS hybrid HK and the GacS/GacA TCS (Fig 1B) is therefore able to integrate at least two different signals, one from GacS and the other from LadS, which also remains to be identified. LadS contains a putative 7-transmembrane (7TMR) region anchoring the HK into the inner membrane and a periplasmic sensor domain (diverse intracellular signaling module extracellular 2, DISMED2), whose predicted fold exhibits a putative binding site, highly conserved in carbohydrate-binding modules (CBMs) [37]. The activity of the multicomponent transduction system described here and the subsequent output response were also certainly modulated by the ratio of kinase to phosphatase activity [38,39]. Phosphatase-like activity leading to catalyzed dephosphorylation of phospho-response regulators described for unorthodox and hybrid sensors by reverse phosphotransfer [40][41][42] can be embodied in the cognate sensor kinase itself [42], carried out by the response regulator itself [40] or by another partner protein [43]. Thus, whether phosphatase-like activity leading to GacA dephosphorylation is assumed by GacS, LadS, any other HK of the network, by itself or by a partner protein requires extensive additional studies.
Thus, P. aeruginosa probably shares with P. fluorescens but not with P. syringae a unique molecular switch controlling Rsm sRNA-dependent virulence made of a central unorthodox HK and three hybrid HKs among which LadS HK makes a unique multicomponent system with the TCS GacS-GacA. This multiple input system probably reflects the variability of the environmental conditions P. aeruginosa faces and may result in a range of gradations of chronic infection through the integration of variable environmental signals.

Truncated versions of LadS and GacS HK
DNA fragments corresponding to the cytoplasmic LadS1D1 part of the LadS hybrid HK (H1 and D1 subdomains), the LadSD1 domain, the cytoplasmic GacSH1D1 part of the GacS unorthodox HK (H1 and D1 subdomains), the GacSH1, the GacSD1 and the GacSH2 domains all fused to a His tag were amplified by PCR using appropriate oligonucleotide pairs (S1 Table). The DNA fragments corresponding to LadSH1D1 and GacSH2 were cloned into pCR2.1 vector yielding respectively pCR2.1ladSH1D1 and pCR2.1gacSH2 plasmids while LadSD1, GacSH1D1, GacSH1 and GacSD1 were cloned into pLic03 vector yielding pLic03_ladSD1, pLic03_gacSH1D1, pLic03_gacSH1 and pLic03_gacSD1, respectively. After DNA sequencing, digestion was performed using EcoRI/BamHI for LadSH1D1 and BamHI/HindIII for GacSH2 for subcloning into pBBRMCS4 and pUC18-miniTn7 vectors, respectively, yielding pBBRLadSH1D1, referred to as LadSH1D1, and pUC18-miniTn7-gacSH2, referred to as GacSH2. Site-directed mutations in the DNA sequence of LadSH1D1 and GacSH2 were introduced respectively into pCR2.1ladSH1D1 and pCR2.1gacSH2 plasmids by Quick exchange site-directed mutagenesis method. Briefly, the conserved histidine residue at position 428 and the conserved aspartate residue at position 718 of LadS HK were changed into glutamine and alanine residues, respectively generating LadSH1 H!Q D1 and LadSH1D1 D!A variants. The conserved histidine residue at position 859 of GacS HK was changed into glutamine, leading to the GacSH2 H!Q variant. This was done by using pCR2.1ladSH1D1 and pCR2.1gacSH2 vectors as matrices and by PCR using pfu turbo DNA polymerase (Stratagene) and 39-mer primers that incorporated appropriate mismatches to introduce the expected mutations (S1 Table). The resulting PCR products were digested with DpnI for 1 hour. After DNA sequencing, each DNA sequence corresponding to LadSH1 H!Q D1 and LadSH1D1 D!A or GacSH2 H!Q variants cloned into pCR2.1 were released by EcoRI/BamHI or BamHI/HindIII digestion, respectively, and inserted into pBBRMCS4 or pUC18-miniTn7, respectively, to generate pBBRladSH1 H!Q D1 and pBBRladSH1D1 D!A and pUC18-miniTn7-gacSH2 H!Q .

Construction of deletion mutants
PCR was used to generate a 500 bp DNA fragment upstream (Up) and a 500 bp DNA fragment downstream (Dn) of the PA4112 and PA4982 and of the LadSD1 domain of the ladS gene  [25] miniCTX-lacZ Tc r lacZ + ; self-proficient integration vector with tet, V-FRT-attPMCS, ori, int, and oriT [51] miniCTX-rsmY-lacZ Promoter region of rsmY gene inserted into miniCTX-lacZ, Tc R This study miniCTX-rsmZ-lacZ Promoter region of rsmZ gene inserted into miniCTX-lacZ, Tc R This study LadS Signalling Pathway through GacS/GacA using the appropriate pairs of primers (S1 Table). Each PCR upstream and downstream product was linked together by overlapping PCR and products were cloned into pCR2.1. The linked DNA fragment was digested with XbaI and SpeI and cloned in the suicide vector pKNG101, yielding pKNG101ΔPA4112 and pKNG101ΔPA4982, respectively. The suicide plasmids were introduced into P. aeruginosa via a three-partner procedure and the deletion mutants were obtained by double selection on LB agar supplemented with Irgasan (25 μg/mL) and streptomycin (1000 μg/mL) at 37°C and NaCl-free LB agar containing 6% sucrose at 30°C. The PAKΔhpAtΔhptBΔhptC triple mutant was constructed as follows: the hptA mutator cloned into the suicide pKNG101 vector [22] was introduced by mating into the PAKΔhptB leading to PAKΔhptAΔhptB. The hptC mutator cloned into the suicide pKNG101 vector was further introduced by mating into the PAKΔhptAΔhptB strain leading to the PAKΔhptAΔhptBΔhptC strain.

Construction of chromosomal punctual mutants
For engineering strains: 1) PAKladSH1 H!Q D1 and PAKladSH1D1 D!A harboring in their chromosomal copy of ladS gene a point mutation of the conserved histidine residue at position 428 or of the conserved aspartate residue at position 718, respectively, and 2) PAKgacSH1 H!Q and PAKgacSH1 H!Q H2 H!Q harboring in their chromosomal copy of gacS gene a point mutation of the conserved histidine at position 293 or of both conserved histidine residues at positions 293 and 859, respectively, the upstream and downstream sequences (approximately 500 bp) were amplified from PAK genomic DNA using the appropriate pairs of primers (S1 Table). Each PCR upstream and downstream product was linked together by overlapping PCR. For pKT25-gacSH1 Two-hybrid plasmid containing cyaAT25-gacS H1 domain fusion [23] pKT25-gacSH2 Two-hybrid plasmid containing cyaAT25-gacS H2 domain fusion [49] pKT25-rocS1H2 Two-hybrid plasmid containing cyaAT25-rocS1 H2 domain fusion [49] pKT25-rocS2H2 Two-hybrid plasmid containing cyaAT25-rocS2 H2 domain fusion [49] pKT25-hptA

Biofilm assay
The P. aeruginosa adherence assay was performed in individual glass tubes containing 1 mL of medium as described previously [22]. After 5 hours of incubation at 30°C, the cultures were incubated with 1% Crystal Violet for 10 min and washed twice. Staining was extracted by treatment with 400 μL 95% ethanol. Subsequently, 600 μL of water was added and OD 570nm was measured. All quantification assays were performed at least in triplicate.

Pull-down experiments
A DNA fragment corresponding to H1 domains of GacS HK, RetS HK and LadS HK was amplified by PCR with an N-terminal FLAG or Strep tag (see S1 Table) and cloned in pCR2.

Overexpression and purification of proteins
Recombinant His-tagged LadSH1D1, LadSH1D1 D!A , LadSD1, GacSH1D1, GacSH1, GacSD1, GacSH2, GacSH2 H!Q and HptA proteins were purified from soluble extracts of the TG1 strain containing either pJF_ladSH1D1, pJF_ladSH1D1 D!A , pJF_gacSH2, pLic03_ gacSH1D1, pLic03_ gacSH1, pLic03_gacSD1, pLic03_ladSD1, pJF_gacSH2 H!Q or pJF_hptA. Cultures were grown aerobically at 37°C until OD 600nm 0.6 and induced for 3 hours with 1 mM IPTG for recombinant proteins produced by the pJF119EH vector or 250 μM IPTG for recombinant proteins produced by the pLic03 vector. A one-step purification via affinity chromatography was facilitated by the presence of a His_Tag at the C-terminal extremity of LadSH1D1, LadSH1D1 D!A , GacSH2, GacSH2 H!Q and HptA and at the N-terminal extremity of LadSD1, GacSH1D1, GacSH1 and GacSD1, using nickel columns (HiTrap HP chelating column) as described by the manufacturer (GE healthcare). Proteins were eluted in an imidazole gradient buffer (20 mM to 500 mM) and analyzed by SDS-PAGE.

In vitro phosphorylation assay
Evidence of a phosphotransfer between LadS variants (LadSH1D1 and LadSH1D1 D!A ) or GacS variants (GacSH1D1 and GacSH1) and GacSH2 variants (GacSH2 and GacSH2 H!Q ), GacSD1 domain, LadSD1 domain or HptA protein was tested by in vitro phosphorylation assays. These assays were carried out in 10 μL of reaction buffer ( (v/v) β-mercaptoethanol). All samples were analyzed by SDS-PAGE and radioactivity was revealed 12 hours after exposition by using a PhosphorImager screen (Molecular Dynamics).

Phosphotransfer kinetic experiments
The phosphotransfer kinetic between the LadSH1D1 or GacSH1D1 and GacS H2 domains was further followed in vitro. These assays were conducted as follows: 2 mM of purified LadSH1D1 or GacSH1D1 proteins was incubated at room temperature for 20 min in 10 μL of reaction buffer (50 mM Tris-HCl [pH 7.6], 50 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol containing 0.1 mM [γ-32 P] ATP). Then 2 mM of purified GacSH2 protein was added and the transphosphorylation reaction was stopped after 0, 0.5, 1, 2, 5, 10 or 30 min by adding 5 μl of loading buffer as described above. All samples were analyzed by SDS-PAGE and radioactivity was revealed 10 hours after exposition by using a PhosphorImager screen (Molecular Dynamics).

Western blot
Bacterial cell pellets were resuspended in loading buffer (

Measurements of β-galactosidase activity
Strains carrying the lacZ transcriptional fusions were grown in LB under agitation at 37°C. The bacterial cells were collected by centrifugation at different growth times. The β-galactosidase activity was measured using the method of Miller. Experiments with strains carrying the exoS-lacZ fusion were performed similarly except that EGTA (5 mM) and MgCl 2 (20 mM) were added in the growth medium.

RT-qPCR
The PAKΔgacS::miniTn7gacSH2 and PAKΔgacS::miniTn7gacSH2 H!Q strains that had received the pBBRladS plasmid or the corresponding pBBRMCS4 empty vector and the PAK, PAK-ΔladS, PAKladSH1 H!Q D1 and PAKladSH1D1 D!A strains were grown in the presence of EDTA at 37°C under agitation until OD 600nm reached 4. Total cellular RNA from 10 mL of cultures was isolated, using the PureYield RNA Midiprep System (Promega), cleaned up and concentrated using the RNeasy kit (Qiagen). The yield, purity and integrity of RNA were further evaluated on Nanodrop and Experion devices. Reverse transcription was performed on 2 μg of RNA by using the SuperScript III first-strand synthesis system (Invitrogen). Real-time PCR runs were carried out on a CFX96 Real-Time System (Bio-Rad). Cycling parameters of the real-time PCR were 98°C for 2 min, followed by 45 cycles of 98°C for 5 s and 60°C for 10 s, ending with a melting curve from 65°C to 95°C to determine the specificity of the amplification.
To determine the amplification kinetics of each product, the fluorescence derived from the incorporation of EvaGreen into the double-stranded PCR products was measured at the end of each cycle using a SsoFast EvaGreen Supermix 2X Kit (Bio-Rad). The results were analyzed using Bio-Rad CFX Manager Software 3.0 (Bio-Rad). The uvrD gene was used as a reference for normalization, in particular because transcription of uvrD is fairly stable in bacteria exposed to antibiotics even at relatively high concentrations [47].
Supporting Information S1 Fig. (A) Activities of the rsmY-lacZ (blue circles) and rsmZ-lacZ (brick-red-colored triangles) transcriptional chromosomal fusions were monitored at different growth stages in the PAK strain, which had received the pBBRMCS4 (empty symbols) or the pBBRladS (filled symbols) vectors. The hptA, hptB, hptC, rocS1H2, rocS2H2, PA4112H2, PA4982H2 and gacSH2 DNA regions were cloned into the two-hybrid pKT25 and the ladS-D1 DNA region was cloned into pUT18C. All of the pKT25 construction as well as the empty vector were cotransformed in BTH101 cells with pUT18C vector containing or not ladS-D1 DNA regions and β-galactosidase activities were measured after 16 hours of growth. All experiments were carried out in at least triplicate, and error bars represent standard deviation. (B) The pBBRladS plasmid containing the ladS HK gene (dark bars) and the pBBRMCS4 corresponding empty cloning vector (light bars) were conjugated in the PAK, PAKΔhptA, PAKΔhptB, PAKΔhptC, PAKΔhptAΔhptBΔhptC, PAKΔrocS1, PAKΔrocS2, PAKΔPA4112, PAKΔPA4982 and PAKΔ-gacS strains. Activity of the rsmZ-lacZ (brick-red-colored) transcriptional chromosomal fusion was monitored after 6 hours of growth (OD 600nm %4) and corresponding β-galactosidase activities are expressed in Miller units and correspond to mean values (with error bars) obtained from three independent experiments. (TIF) S1 Table. Oligonucleotides used in this study. (DOCX)