Introduction

Pseudomonas fluorescens CHA0, a model biocontrol bacterium originally isolated from a tobacco field in Switzerland, is an effective antagonist of plant-pathogenic fungi and nematodes that cause root diseases. Like other biocontrol bacteria, strain CHA0 colonizes the rhizosphere of important crop plants and produces several antibiotic compounds and lytic exoenzymes, which are important biocontrol factors accounting for suppression of root diseases (Haas and Défago, 2005; Mark et al., 2006; Loper et al., 2007; Mercado-Blanco and Bakker, 2007). The expression of biocontrol factors depends on the Gac/Rsm signal transduction pathway (Laville et al., 1992; Zuber et al., 2003). In this regulatory cascade, the GacS/GacA two-component system activates the transcription of three small RNAs (sRNAs) termed RsmX, RsmY and RsmZ when cells reach high population densities (Heeb et al., 2002; Valverde et al., 2003; Kay et al., 2005). These sRNAs avidly bind two sRNA-binding proteins belonging to the RsmA/CsrA family, named RsmA and RsmE (Reimmann et al., 2005). In this way, translational repression exerted by these proteins can be relieved and target mRNAs become accessible to ribosomes for translation (Valverde et al., 2004; Lapouge et al., 2007, 2008). Typical target genes are, on the one hand, genes involved in biocontrol such as hcnA (for hydrogen cyanide (HCN) synthesis) and phlA (for synthesis of the antifungal metabolite 2,4-diacetylphloroglucinol (DAPG)) and, on the other hand, the rpoS gene encoding the stress and stationary phase sigma factor σ38, which is involved in the response of P. fluorescens to oxidative stress (Blumer et al., 1999; Heeb et al., 2005; Kay et al., 2005). Thus, mutants affected in the GacS/GacA two-component system produce dramatically reduced amounts of secondary metabolites and are more sensitive to hydrogen peroxide, compared to the wild type (Laville et al., 1992; Zuber et al., 2003; Heeb et al., 2005).

In pseudomonads, the activity of the Gac/Rsm pathway is regulated by autoinducing signal molecules whose chemical structures are unknown. These signals are thought to activate phosphorylation of the unorthodox GacS sensor (Heeb et al., 2002; Zuber et al., 2003; Dubuis et al., 2007). Phosphorylated GacS then activates the response regulator GacA via a phosphorelay mechanism, for which experimental evidence has been obtained in Pseudomonas aeruginosa (Goodman et al., 2009). In this organism, two additional sensors provide input into the Gac/Rsm pathway. These hybrid sensors termed RetS (for regulator of exopolysaccharide and type III secretion) and LadS (for lost adherence) were discovered in screens for mutants that form increased or decreased amounts of biofilm polysaccharides. It has been shown that RetS inhibits and LadS activates the activity of the Gac/Rsm pathway (Goodman et al., 2004; Laskowski and Kazmierczak, 2006; Ventre et al., 2006). There is evidence that both RetS and LadS physically interact with GacS (Goodman et al., 2009; Workentine et al., 2009). However, the mechanisms by which these sensors communicate with one another and thereby determine the output of the system are not known. Moreover, the function of the Gac/Rsm pathway can be influenced by environmental cues. For instance, in Escherichia coli low pH values inhibit the activity of the BarA/UvrY proteins, which are GacS/GacA homologues (Mondragón et al., 2006).

In various biocontrol strains of P. fluorescens it has been observed that incubation temperatures around 35 °C have a negative effect on biocontrol efficacy in vivo and on the expression of biocontrol factors such as DAPG and phenazine-1-carboxylic acid in vitro (Shanahan et al., 1992; Slininger and Shea-Wilbur, 1995; Schmidt et al., 2004). We found that in strain CHA0, too, the production of antibiotic compounds and HCN was reduced at 35 °C, by comparison with the production at 30 °C. This suggested to us that some component of the Gac/Rsm pathway might be sensitive to elevated temperature. The aim of this study, therefore, was to find this component. This led us to examine the roles of the RetS and LadS sensors in strain CHA0; RetS was found to be involved in temperature control. Using a genetic approach, we show that mutations in gacS or gacA override the effects of mutations in retS or ladS. These findings are consistent with GacS interacting directly with RetS and LadS.

Materials and methods

Bacterial strains, plasmids and growth conditions

The bacterial strains and plasmids are listed in Table 1. Strains of E. coli and P. fluorescens were grown in nutrient yeast broth (NYB) with shaking or on nutrient agar plates (Stanisich and Holloway, 1972). When required, antibiotics were added at the following concentrations: ampicillin (Ap), 100 μg ml−1 (only for E. coli); gentamicin (Gm), 10 μg ml−1; kanamycin (Km), 25 μg ml−1; tetracycline (Tc), 25 μg ml−1 (for E. coli) or 125 μg ml−1 (for P. fluorescens). In the mobilization of suicide plasmids (pME3087 derivatives) from E. coli to P. fluorescens, chloramphenicol (Cm) at 10 μg ml−1 and Tc were used to select for the recipient having integrated the suicide plasmid. Enrichment for Tc-sensitive strains, from which the suicide plasmid had been excised, was performed by exposing cells growing in NYB (at approximately 108 cells per ml) to Tc (20 μg ml−1) for 1 h, followed by the addition of cycloserine (1.6 mg ml−1) and further incubation for 5 h. For detection of lacZ constructs, 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal) was added to plates at a final concentration of 0.02%. Routine incubation temperatures were 30 °C for P. fluorescens and 37 °C for E. coli. Alternatively, P. fluorescens was grown at 35 °C to test the temperature sensitivity of the Gac/Rsm signal transduction pathway and to improve its capacity to accept heterologous DNA originating from E. coli.

Table 1 Bacterial strains, plasmids and oligonucleotides used in this study
Full size table

DNA manipulation

Small-scale plasmid extractions were done with the QIAprep Spin Miniprep Kit (Qiagen, Basel, Switzerland), whereas large-scale preparations were performed with the Jetstar kit (Genomed GmbH, Basel, Switzerland). Chromosomal DNA from P. fluorescens was prepared as previously described (Gamper et al., 1992). DNA manipulations were carried out by standard techniques (Sambrook and Russell, 2001). DNA fragments were purified from agarose gels with the MinElute or QIAquick Gel extraction kits (Qiagen), depending on the fragment size. Electroporation of bacterial cells with plamid DNA was done as described (Farinha and Kropinsky, 1990). Conditions for amplifying PCR fragments were as follows: 200 ng of genomic DNA was dissolved in a final volume of 20 μl containing 200 μM of each of the four dNTPs, 20 pmol of each of two primers, 2 U of GoTaq polymerase (Promega, Catalys, Wallisellen, Switzerland) and 1 × GoTaq buffer (Promega). The PCR cycle was 2 min at 95 °C, 30 × (45 s at 95 °C, 45 s at 50–60 °C (depending on the G+C content and length of the primers), 0.5-2 min at 72 °C (depending on the length of the amplicon)) and a final elongation step of 5 min at 72 °C. The reaction products were purified on an agarose gel and the purified fragments were sequenced with an automatic sequencer.

Plasmid and strain constructions

These were facilitated by the fact that strain CHA0 is very similar to the completely sequenced strain Pf-5 of P. fluorescens (Paulsen et al., 2005; Loper et al., 2007). To generate the retS mutant CHA1202, a 2.8-kb fragment was deleted in-frame in the chromosomal retS gene. For this purpose, a 620-bp BamHI-EcoRI fragment upstream of retS was amplified by PCR from strain CHA0, using primers RetF1 and RetF2. A 640-bp EcoRI-HindIII fragment including the last 9 bp of retS and the adjacent downstream region was amplified by PCR with primers RetR1 and RetR2. The resulting upstream and downstream fragments were cloned by a triple ligation into pBluescript II KS digested with BamHI and HindIII, giving plasmid pME7704 (Table 1). The 1.2-kb BamHI-HindIII insert was excised and cloned into the suicide plasmid pME3087 digested with BamHI and HindIII, producing pME7705 (Table 1). This plasmid was integrated into the chromosome of strain CHA0 by triparental mating, using E. coli HB101/pME497 as the mobilizing strain. Clones, in which excision of the vector by a second crossing-over event had occurred, were isolated after enrichment for tetracycline-sensitive cells. The ΔretS mutation in the recombinant strain was verified by PCR using primers RetN1 and RetN2.

An analogous gene replacement strategy was followed to create a 2.5-kb ladS deletion in strain CHA1204. Using CHA0 DNA as a template, fragments flanking the ladS gene were amplified by PCR with primer pairs LadF1-LadF2 and LadR1-LadR2, respectively. The 575-bp upstream and 625-bp downstream fragments obtained were digested with BamHI-EcoRI and with EcoRI-HindIII, respectively, and cloned into BamHI-HindIII-digested pBluescript II KS, resulting in plasmid pME7708 (Table 1). The 1.2-kb BamHI-HindIII insert was excised from pME7708 and cloned into pME3087 digested with the same restriction enzymes, giving plasmid pME7709 (Table 1), which served to delete the ladS gene in strain CHA1204. The ΔladS mutation in the recombinant strain was verified by PCR using primers LadN1 and LadN2.

The ΔretS ΔladS double mutant CHA1305 was obtained by using the suicide plasmid pME7705 to delete the retS gene in the ladS mutant CHA1204. The ΔretS ΔgacS double mutant CHA1301 (Table 1) was obtained similarly in the gacS mutant CHA19. The ΔladS ΔgacS double mutant CHA1302 (Table 1) was obtained by using the suicide plasmid pME7709 to delete the ladS gene in the gacS mutant CHA19. Likewise, pME7705 and pME7709 were used to delete retS and ladS in the gacA::Kmr mutant CHA89, generating the gacA::Kmr ΔretS and the gacA::Kmr ΔladS double mutants CHA1303 and CHA1304, respectively (Table 1). The retS and ladS mutations were verified by PCR.

The mini-Tn7-Gm carrier plasmid pME6182 is a derivative of the previously described Tn7-delivery vector pME3280a (Zuber et al., 2003), which carries transcription stop signals at both ends of its multiple cloning site. It was generated in several steps. First, a 0.36-kb BfrI-SalI fragment carrying the transcription stop signal located between the P. aeruginosa pchDCBA operon and the downstream ssb gene (http://www.pseudomonas.com) was cloned between the PstI and SalI sites of pUC18; the BfrI and PstI ends were made compatible by T4 DNA polymerase treatment. The transcription stop signal was then excised as 0.36-kb SphI-XbaI fragment and inserted between the SphI and SpeI sites of the pME3280a polylinker. To generate the chromosomal insertion of rsmX-lacZ, a 3.5-kb rsmX-lacZ fusion fragment was first recovered by digesting pME7317 with EcoRI and XhoI, blunted (with 5 U T4-DNA polymerase (Promega), 100 μM dNTPs, 20 min at room temperature), and cloned into plasmid pME6182 digested with SmaI. The resulting plasmid pME7698 and the Tn7 transposition helper plasmid pUX-BF13 were then coelectroporated into different recipient strains, with selection for the mini-Tn7. Likewise, the chromosomal insertion of rsmY-lacZ was constructed by digesting pME6916 with EcoRI and XhoI to recover a 3.5-kb rsmY-lacZ fusion fragment, which was blunted and cloned into the plasmid pME6182 digested with SmaI. The construct obtained, pME7699, and pUX-BF13 were coelectroporated into different recipient strains.

Antibiotic and HCN assays

Antibiotic production by strain CHA0 was assessed with Bacillus subtilis 168M as the indicator as described by Dubuis and Haas (2007) on plates containing, per liter, 10 g proteose peptone, 4.6 g glycerol, 0.75 g K2HPO4, 0.75 g MgSO4·7H2O, 27 mg FeCl3·6H2O and 20 g agar (pH 7.0). HCN concentrations in culture supernatants were determined after oxygen-limited growth in NYB according to a protocol previously described (Kay et al., 2006).

β-Galactosidase assays

P. fluorescens strains containing lacZ constructs were grown in 20 ml NYB (amended with 0.05% (vol/vol) Triton X-100) in 100-ml Erlenmeyer flasks with shaking. β-Galactosidase activities were quantified by the method of Miller (1972), using cells permeabilized with 5% (vol/vol) toluene. All experiments were performed in triplicate.

Results

Temperature sensitivity of the Gac/Rsm pathway

In rich liquid medium (NYB) and during exponential growth, strain CHA0 had a similar doubling time (33±1 min) at 30 °C as well as at 35 °C. No growth occurred above 37 °C. At an incubation temperature of 35 °C, P. fluorescens CHA0 produced only a small amount of antibiotic compounds on rich solid medium, as revealed by a small inhibition zone using a B. subtilis overlay. By contrast, at the standard incubation temperature of 30 °C, antibiotic production was markedly stronger. We presume that the antibiotics that are produced under these conditions include DAPG as a major component. We also noted that strain CHA0 produced less HCN at 35 °C than at 30 °C. Antibiotic and HCN data will be shown in the last section of Results. The expression of the hcnA and phlA genes involved in the biosynthesis of HCN and DAPG, respectively, revealed the extent of temperature sensitivity. This was seen with translational hcnA′-′lacZ and phlA′-′lacZ fusions (Figures 1a and b). An rpoS′-′lacZ fusion was also tested and found to be less active at 35 °C than at 30 °C (Figure 1c). Together, these results suggested that it might be the Gac/Rsm pathway that responds to temperature as an environmental cue. This hypothesis was confirmed by the observation that the expression of transcriptional lacZ fusions to the rsmZ, rsmY and rsmX sRNA genes was lower at 35 °C than that found at standard 30 °C in NYB medium (Figures 2a–c). Note that strain CHA0 did not tolerate fully induced expression of the rsmY-lacZ and rsmX-lacZ fusions when these were carried by plasmids having about six copies. In our experience, it is difficult to detect specific activities exceeding 40 000 Miller units, which were the activities found at the end of growth in the wild type carrying pME6916 (rsmY-lacZ) or pME7317 (rsmX-lacZ; data not shown). To overcome this problem, these fusions were introduced into the chromosomal Tn7 attachment site, whereas rsmZ-lacZ was assessed on a plasmid construct.

Figure 1
figure 1

Expression of target genes of the Gac/Rsm cascade in the wild-type CHA0 as a function of growth temperature. Strains were grown in nutrient yeast broth (NYB) at 30 or 35 °C. β-Galactosidase activities were determined for (a) a ptac-hcnA′-′lacZ fusion (on pME6530), (b) a phlA′-′lacZ fusion (on pME6259) and (c) an rpoS′-′lacZ fusion (on pME6355). □, wild type at 30 °C; ▪, wild type at 35 °C. Experiments were done three times; each value is the average±standard deviation.

Full size image
Figure 2
figure 2

Influence of the growth temperature on transcriptional expression of the small regulatory RNAs RsmZ, RsmY and RsmX in the wild-type CHA0. Strains were grown in nutrient yeast broth (NYB) at 30 or 35 °C. β-Galactosidase activities were determined for (a) a plasmid-borne chromosomal rsmZ-lacZ fusion (on pME6091), (b) a chromosomal rsmY-lacZ fusion and (c) a chromosomal rsmX-lacZ fusion. □, wild type at 30 °C; ▪, wild type at 35 °C. Experiments were done three times; each value is the average±standard deviation.

Full size image

RetS as a temperature-sensitive element

To test which component of the Gac/Rsm pathway was responsible for the temperature effect, we first tested the rsmZ-lacZ fusion in the gacSΔ76 mutant CHA19.8, which expresses GacA-dependent genes at constitutive high levels (Zuber et al., 2003). However, the rsmZ-lacZ construct displayed no temperature sensitivity (data not shown), indicating that the activities of the GacS sensor and the GacA response regulator were not compromised at high temperature. This led us to inspect the potential influence of the two accessory sensors LadS and RetS, which had previously been described in P. aeruginosa (Goodman et al., 2004, 2009; Laskowski and Kazmierczak, 2006; Ventre et al., 2006). To this end, we constructed mutants deleted for ladS (PFL_5426 in the closely related strain Pf-5) or retS (PFL_0664 in strain P. fluorescens Pf-5) in the wild-type P. fluorescens CHA0. In a ladS mutant, the lacZ fusions to rsmZ, rsmY and rsmX were all expressed at levels (Figures 3a–c) that were roughly half of those found in the wild type (Figures 2a–c). However, in the ladS background temperature sensitivity essentially persisted (Figure 3). By contrast, in a retS mutant, temperature sensitivity was lost and all three fusion constructs were expressed at high levels, well above those observed in the wild type (Figures 4a–c). In a retS ladS double mutant, the same result was obtained as in a retS mutant (data not shown). A control experiment using a transcriptional lacZ fusion to the constitutive tac promoter confirmed that transcription per se was not compromised at 35 °C (data not shown). Taken together, these results indicate that LadS and RetS have positive and negative effects, respectively, on the Gac/Rsm cascade of P. fluorescens CHA0 and that temperature sensitivity depends essentially on RetS.

Figure 3
figure 3

Influence of the growth temperature on transcriptional expression of the small regulatory RNAs RsmZ, RsmY and RsmX in a ladS mutant. Strains were grown in nutrient yeast broth (NYB) at 30 or 35 °C. β-Galactosidase activities were determined for (a) a plasmid-borne rsmZ-lacZ fusion (on pME6091), (b) a chromosomal rsmY-lacZ fusion and (c) a chromosomal rsmX-lacZ fusion. ○, ladS mutant at 30 °C; •, ladS mutant at 35 °C. Experiments were done three times; each value is the average±standard deviation.

Full size image
Figure 4
figure 4

Influence of the growth temperature on transcriptional expression of the small regulatory RNAs RsmZ, RsmY and RsmX in a retS mutant. Strains were grown in nutrient yeast broth (NYB) at 30 or 35 °C. β-Galactosidase activities were determined for (a) a plasmid-borne rsmZ-lacZ fusion (on pME6091), (b) a chromosomal rsmY-lacZ fusion and (c) a chromosomal rsmX-lacZ fusion. ▵, retS mutant at 30 °C; ▴, retS mutant at 35 °C. Experiments were done three times; each value is the average±standard deviation.

Full size image

Mutations in gacS or gacA are epistatic to mutations in ladS and retS

Recent studies indicate that LadS and RetS physically interact with GacS (Goodman et al., 2009; Workentine et al., 2009). The following experiments lend support to these findings. The expression of rsmZ, rsmY and rsmX was very low in both gacS and gacA mutants of strain CHA0, compared with the wild type, in agreement with previous results (Kay et al., 2005). In ladS gacS, ladS gacA, retS gacS and retS gacA double mutants, the expression of the three sRNA genes was equally low (Supplementary Figure S1). By contrast, as shown above (Figures 2, 3 and 4), in a simple ladS background all reporter fusions were expressed at levels that were intermediate between the basal (gacS or gacA) level and the wild-type level, whereas very high levels occurred in a simple retS mutant. Thus, LadS and RetS functioned as modulators of the Gac/Rsm pathway in P. fluorescens CHA0. The fact that gacS and gacA mutations were epistatic to retS and ladS mutations are in agreement with a model in which RetS and LadS make contacts with GacS.

A retS mutant shows diminished temperature regulation of biocontrol factor expression

If RetS–GacS interaction is regulated by temperature, we would expect that in a retS mutant high temperature should have little or no effect on the expression of biocontrol factors. This was the case. The expression of translational lacZ fusions to the hcnA and phlA target genes was consistent with the expression of the rsmZ, rsmY and rsmX sRNA genes: a ladS mutation lowered the expression, whereas a retS mutation increased the expression of these lacZ fusions (Supplementary Figure S2), although the hcnA′-′lacZ reporter on plasmid pME6530 could not be measured in the retS mutant, because β-galactosidase levels were above the tolerated upper limit. The translational expression of rpoS also followed the same pattern (Supplementary Figure S2). Antibiotic production was enhanced in the retS mutant CHA1202, relative to that found in the wild type. This effect was most pronounced at 35 °C where the wild type had low antibiotic activity, as revealed by a small zone of inhibition of B. subtilis (Figure 5). The effects of the ladS mutation on antibiotic production (Figure 5) were also consistent with sRNA expression data (Figure 3). Finally, HCN production was enhanced and advanced in the retS mutant both at 30 °C and at 35 °C, whereas in the wild-type HCN formation was delayed and strongly reduced at 35 °C (Figure 6). The model shown in Figure 7 summarizes the findings and will be presented in Discussion.

Figure 5
figure 5

Influence of the growth temperature on the production of antibiotics in the Pseudomonas fluorescens wild-type strain CHA0, the retS mutant CHA1202, the ladS mutant CHA1204 and the gacS mutant CHA19. Antibiotic production was assessed by spotting 2-μl samples of an overnight culture (adjusted to OD600=2.0) onto solid plates (see Materials and methods). After incubation at 30 °C (a) or 35 °C (b) for 12–14 h, P. fluorescens was killed by UV irradiation and Bacillus subtilis strain 168 M was overlaid in 0.5% nutrient yeast broth (NYB) agar. Plates were incubated at 30 °C until inhibition halos appeared.

Full size image
Figure 6
figure 6

Influence of the growth temperature on hydrogen cyanide (HCN) production in the Pseudomonas fluorescens wild-type strain CHA0 and in the retS mutant CHA1202. HCN production was measured in nutrient yeast broth (NYB) medium in which P. fluorescens had been grown with oxygen limitation. □, wild type at 30 °C; ▪, wild type at 35 °C; ▵, retS mutant at 30 °C; ▴, retS mutant at 35 °C. Experiments were done five times; each value is the average±standard deviation.

Full size image
Figure 7
figure 7

Integrative model of the Gac/Rsm signal transduction pathway in Pseudomonas fluorescens CHA0. ↓, positive control; ⊥, negative control; dotted lines, indirect control. Evidence for this pathway comes from this work and from previous studies (Heeb et al., 2002, 2005; Valverde et al., 2003; Kay et al., 2005; Reimmann et al., 2005).

Full size image

Discussion

The principal aim of this study was to shed light on the mechanism involved in thermoregulation of biocontrol factor expression in P. fluorescens CHA0. In precedent work on various biocontrol strains of P. fluorescens phenotypic evidence had been obtained for this type of regulation (Shanahan et al., 1992; Slininger and Shea-Wilbur, 1995; Schmidt et al., 2004), but the mode of action remained unknown. Our observation that key regulatory elements, the sRNA genes rsmX, rsmY and rsmZ, all showed diminished transcription at 35 °C, relative to that seen at the standard growth temperature of 30 °C (Figure 2), led to the hypothesis that some component of the Gac/Rsm signal transduction pathway might be temperature-sensitive. We found that a retS mutant had lost temperature sensitivity and expressed the sRNA genes at constitutive high levels (Figure 4). This finding suggests that RetS activity is influenced by temperature and that the GacS/GacA two-component system itself remains fully functional at elevated temperature. The ecological significance of temperature-sensitive expression of biocontrol factors is speculative at the moment. If we admit that the production of biocontrol traits normally confers a selective advantage on the producer strains (Haas and Keel, 2003), then we can assume that this advantage may not be relevant at temperatures around 35 °C where the growth of many competing soil microorganisms including some pathogenic fungi is inhibited. Under these conditions, P. fluorescens can conceivably afford to dedicate its metabolic energy essentially to primary metabolism, without having to pay the full cost of secondary metabolism, which is necessary for producing biocontrol factors.

We also tested pH as a potential cue, as a previous study (Mondragón et al., 2006) had shown pH-sensitivity of the BarA/UvrY two-component system in E. coli. However, at pH 6.2, the lowest pH value allowing good growth of strain CHA0, the expression of lacZ fusions to rsmZ, rsmY and rsmX was similar to that measured at standard neutral pH (data not shown).

The specificity of the GacS–GacA interaction has previously been substantiated by genetic analyses conducted in several Pseudomonas species (Rich et al., 1994; Heeb and Haas, 2001) and by biochemical evidence for phosphotransfer between a soluble form of GacS and GacA in P. aeruginosa (Goodman et al., 2009). In P. aeruginosa, the LadS and RetS sensors have been reported to modify the activity of the GacS/GacA system in vivo (Goodman et al., 2004; Laskowski and Kazmierczak, 2006; Ventre et al., 2006). LadS contains eight putative transmembrane segments, an autophosphorylation (kinase) domain and a response regulator domain. Mutation of the ladS gene strongly diminished rsmZ expression (rsmY expression was not tested) and resulted in reduced production of adhesive extracellular polysaccharides and enhanced expression of the type III secretion apparatus in P. aeruginosa (Ventre et al., 2006). Thus, LadS appears to activate the GacS/GacA system although the mechanism by which this effect is brought about has not been elucidated. In P. fluorescens CHA0, there is evidence that LadS and GacS may interact physically (Workentine et al., 2009). As we have shown here, LadS positively controls rsmX, rsmY and rsmZ expression, albeit less strongly than GacS (Figures 2, 3; Supplementary Figure S2). As both gacS and gacA mutations were epistatic to a ladS mutation (Supplementary Figure S1), it is likely that LadS acts upstream of GacS. This might be achieved if LadS physically interacted with GacS and thereby facilitated autophosphorylation of GacS (see model in Figure 7). A direct interaction between LadS and GacA appears less plausible as in this case the expression of the three sRNAs should be more strongly affected by the double ladS gacS mutation than by single ladS or gacS mutations; however, this was not observed (Supplementary Figure S1).

The RetS sensor of P. aeruginosa also contains eight putative transmembrane segments, a kinase domain and an adjacent response regulator domain. However, RetS has an additional response regulator domain located at the C terminus, and this domain is most important for biological activity. A retS mutant of P. aeruginosa overexpressed rsmZ and biofilm exopolysaccharides and had a downregulated type III secretion apparatus; rsmY expression was not tested (Goodman et al., 2004; Laskowski and Kazmierczak, 2006). Mutation in gacS was found to be epistatic to mutation in retS (Goodman et al., 2004). RetS acts as an antagonist of GacS in vitro (Goodman et al., 2009). In P. fluorescens CHA0, we found that the RetS homologue had a negative effect on the expression of all three GacA-controlled sRNA genes, rsmX, rsmY and rsmZ (Figures 2 and 4). Consequently, a retS mutation resulted in strongly enhanced promoter activities of the three sRNA genes and hence the expression of target genes and of biocontrol factors was strongly elevated as well (Figures 5, 6; Supplementary Figure S2). These data are consistent with a model of direct RetS–GacS interaction (Figure 7), for which there is in vivo evidence in strain CHA0 (Workentine et al., 2009). However, the biochemistry of the antagonistic interaction between RetS and GacS is still uncertain. It has been postulated that RetS could have phosphatase activity on GacS (Laskowski and Kazmierczak, 2006), but this has not been confirmed in vitro (Goodman et al., 2009). There are precedents of interacting membrane sensor proteins: various dimeric chemoreceptor proteins of enteric bacteria are known to assemble as trimers in the cytoplasmic membrane, allowing the bacteria to integrate several signals in the chemotactic response (Hazelbauer et al., 2008). Note that the model shown in Figure 7 differs from another recently published model in which GacS, LadS and RetS are all assumed to interact with GacA in P. aeruginosa (Gooderham and Hancock, 2009).

Temperature is an important environmental cue. Some bacteria can sense it via transmembrane sensor kinases other than RetS. For example, a temperature-responsive sensor regulates the production of the chlorosis-inducing toxin coronatine in the soybean pathogen Pseudomonas syringae pv. glycinea PG4180. Toxin production occurs at 18 °C but not at 28 °C, the optimal growth temperature (Palmer and Bender, 1993). This thermoregulation is mediated at the transcriptional level by a regulatory system consisting of a histidine protein kinase, CorS, and two transcriptional activators, CorR and CorP. The C-terminal cytosolic region of CorS appears to act as a temperature sensor; it is believed to respond to intracellular temperature changes via autophosphorylation and to transduce the signal to the response regulator CorR via phosphorylation (Braun et al., 2008). Another example is provided by the thermal control of fatty acid synthesis. The fraction of unsaturated phospholipid acyl chains in phospholipids increases when the growth temperature decreases. This adaptation improves membrane fluidity. In B. subtilis, a two-component system composed of a membrane-associated kinase, DesK, and a soluble transcriptional activator, DesR, regulates the transcription of the des gene coding for a Δ5-fatty acid desaturase. DesK is a sensor having both kinase and phosphatase activities. At 37 °C, when membrane lipids are in a disordered fluid state, the phosphatase mode of DesK is dominant. After a temperature downshift to 25 °C, the proportion of ordered membrane lipids (that is, a nonfluid state) predominates, leading to an increase of the kinase mode of DesK. This results in autophosphorylation and transfer of the phosphoryl group to DesR. Phosphorylated DesR activates transcription of des, and the Des enzyme introduces a double bond into the acyl chains of membrane phospholipids (Aguilar et al., 2001). Experimental evidence points to membrane fluidity being a stimulus of the N-terminal transmembrane domain of DesK (Hunger et al., 2004).

In this study, we have shown that RetS is a stronger antagonist of GacS at 35 °C than at 30 °C. This mechanism of temperature sensing might involve a change in membrane fluidity, enabling a stronger contact between the two sensors at 35 °C. Previous observations in Pseudomonas species have indeed suggested that changes of membrane fluidity can modify quorum sensing regulation (Baysse and O’Gara, 2007).