Modulation of antibiotic sensitivity and biofilm formation in Pseudomonas aeruginosa by interspecies signal analogues

Pseudomonas aeruginosa, a significant opportunistic pathogen, can participate in inter-species communication through signaling by cis-2-unsaturated fatty acids of the diffusible signal factor (DSF) family. Sensing these signals leads to altered biofilm formation and increased tolerance to various antibiotics, and requires the histidine kinase PA1396. Here, we show that the membrane-associated sensory input domain of PA1396 has five transmembrane helices, two of which are required for DSF sensing. DSF binding is associated with enhanced auto-phosphorylation of PA1396 incorporated into liposomes. Further, we examined the ability of synthetic DSF analogues to modulate or inhibit PA1396 activity. Several of these analogues block the ability of DSF to trigger auto-phosphorylation and gene expression, whereas others act as inverse agonists reducing biofilm formation and antibiotic tolerance, both in vitro and in murine infection models. These analogues may thus represent lead compounds to develop novel adjuvants improving the efficacy of existing antibiotics.

A ntibiotic resistance, coupled with limited development of new antibiotic agents, poses a significant global threat to public health, underscoring the need to find alternative strategies to fight infection [1][2][3][4] . These strategies include targeting the signaling pathways that regulate the synthesis of microbial virulence factors and approaches to improve the efficacy of existing antibiotics. Bacterial cell-cell communication (quorum sensing), biofilm formation, and cyclic di-GMP signaling were proposed as potential targets for interference by small molecules [1][2][3][4] . Anti-virulence factors are attractive since they do not influence bacterial growth, which may reduce the selective pressure for developing resistance.
This study addresses the modulation of Pseudomonas aeruginosa by inter-species signaling. P. aeruginosa, a widespread opportunistic human pathogen, uses two N-acyl homoserine lactones (3-oxo-dodecanoyl-and butanoyl-HSL) and the quinolone signal PQS (2-heptyl-3-hydroxy-4(1H)-quinolone) as quorum-sensing molecules that regulate the synthesis of virulence factors [5][6][7][8] . There are various reports of small molecule inhibition of these pathways with effects on virulence factor synthesis [9][10][11] . In addition to intra-species signaling, P. aeruginosa can participate in inter-species signaling mediated by molecules of the diffusible signal factor (DSF) family, which are cis-2-unsaturated fatty acids. In particular, P. aeruginosa senses cis-11-methyl-2-dodecenoic acid (DSF) and cis-2-dodecenoic acid (BDSF), which are produced by other bacteria such as Burkholderia species and Stenotrophomonas maltophilia, but not by P. aeruginosa 12,13 . Sensing occurs through the PA1396 histidine kinase and results in altered biofilm formation and increased tolerance to several antibiotics including polymyxin B 13,14 . Deletion of PA1396 in the model strain PAO1 and several clinical isolates also leads to increased tolerance to polymyxins B and E, suggesting DSF negatively modulates PA1396 activity 13,14 . Detection of DSF family molecules in polybacterial infections, such as those associated with cystic fibrosis (CF) where P. aeruginosa is present together with other bacterial species, suggests interspecies signaling occurs in vivo and may therefore lead to reduced efficacy of antibiotic therapy 13,14 .
Here, we examine in more detail the role of PA1396 in sensing DSF family signals and the potential of structural analogs of these signals to modulate PA1396 action. In particular, we focus on the effects of DSF analogs on the PA1396-regulated functions of biofilm formation and antibiotic tolerance, both in vitro and in murine infection models. We demonstrate that two of the five transmembrane helices of the input domain of PA1396 are required for DSF sensing, and DSF binding is associated with enhanced PA1396 auto-phosphorylation. Furthermore, we found that certain analogs of DSF block its ability to trigger autophosphorylation, which coincide with specific alterations in the expression of a subset of P. aeruginosa genes, reduction in biofilm formation and antibiotic tolerance using both in vitro and in vivo infection models. Altogether, our findings reveal that DSFmediated interspecies interactions may be influenced pharmacologically to improve the treatment of chronic P. aeruginosa infections.

Results
Examining the sensor input domain of PA1396 for DSF recognition. Although DSF perception by PA1396 requires its membrane-associated sensor input domain 13,14 , the underlying molecular mechanism is unknown. As a first step to better understand DSF sensing by PA1396, topological models of the protein were generated with various prediction programs. Analyses with TMHMM, MEMSAT, SOSUI, and HMMTOP suggested a topological model for PA1396 comprising five transmembrane helices (TMHs), and with the N-and the Cterminus located in the periplasm and cytoplasm, respectively (Fig. 1a). In contrast, TOPPRED predicted a relatively large cytosolic loop and a different location in the protein for some of the TMHs. To experimentally assess the PA1396 topology, we constructed translational fusions expressing hybrid proteins of the PA1396 input domain with either alkaline phosphatase (PhoA) or β-galactosidase (LacZ) reporters as topological probes for periplasmic and cytoplasmic locations, respectively. The hybrid proteins were designed such that at least one fusion was placed in each predicted loop facing either the periplasm or the cytoplasm (Fig. 1b).
To test whether the predicted TMHs are important for DSF sensing, we constructed PA1396 derivatives with progressive truncations at the N-terminus. Each truncated PA1396 gene was cloned into pBBR1MCS and the constructs introduced into a PA1396 deletion mutant (herein ΔPA1396). Bacteria expressing each of these transconjugants were examined for their ability to respond to exogenous DSF by measuring pmrAB gene expression using a pmr-gfp fusion, and for polymyxin B resistance (see Methods). Removal of the first three TMHs did not affect the ability of PA1396 to respond to DSF when compared to wild type (Fig. 1c). In contrast, any further truncation led to a loss of DSF responsiveness (Fig. 1c). Western blot analysis of the truncated proteins with antisera against the His 6 tag showed they were all expressed to the same level as the wild type ( Supplementary  Fig. 1). These results suggested TMHs IV and V have a role in DSF sensing by PA1396.
To identify specific residues necessary for DSF binding and signal transduction, we first focused on those conserved in PA1396 homologs from different bacterial species. Conserved or semi-conserved residues in TMHs IV and V ( Supplementary  Fig. 2) were examined for functionality by constructing alanine replacements (or serine in the case of alanine residues). Residues  Y116, L117, T121, L123, L128, T135, P136, W138, A140, P142,  M144, L148, M149, V154, I155, P156, and Y158 were replaced and the resulting variants assayed for their ability to confer response to DSF. Bacteria expressing many of the variants did not exhibit significant changes in DSF responsiveness. We identified several variants exhibiting decreased sensitivity to DSF, such that higher concentrations were required to activate the pmr-gfp reporter, although significant activation occurred at concentrations in excess of 50 µM. These variants had alanine replacements at T121, L123, L128, P142, L148, M149, V154, and I155. Except for P142, all the variants with reduced DSF sensitivity were altered in positions within the last two TMHs of PA1396. Variants with combinations of alanine substitutions (such as T121A/L123A/L128A) had an even further reduced sensitivity ( Fig. 1d). The observed differences in the responses to DSF by the reporter strain could not be attributed to differential protein expression since all replacements were expressed to the same level as the parental protein ( Supplementary Fig. 1). Altogether, we conclude residues in TMH-IV and TMH-V of PA1396 play a critical role in the recognition of the DSF signal.
Physiochemical features of DSF important for recognition by PA1396. The DSF family comprises cis-2-unsaturated fatty acids of differing chain length, methyl substitution and saturation 12 . Individual bacteria can produce multiple DSF family signals, although often a single chemical species predominates. Previous work showed that the configurational isomer trans-11-methyl-2dodecenoic acid was 500-fold less effective than DSF in activation of pmrAB, whereas the corresponding unsaturated fatty acids have little or no activity 13 . To examine the structural features of the DSF signals required for PA1396 sensing in more detail, we synthesized a panel of analogs differing from the DSF 'parent' molecule in chain length, position and presence of methyl branching or with a trans configuration of the double bond. Additional molecules were created by esterification, conversion to hydroxamic acids or amides, and by incorporation of stable aromatic substituents to replace the carboxylic acid moieties (see Fig. 2a). This library of structural derivatives provided a platform to define key structural features for PA1396 sensing and signal transduction activity.
The activity of the molecules in this panel was assayed by determining the relative fluorescence of the reporter normalized by bacterial cell density (Fig. 2b). As the minimum concentration of DSF (cis-11-methyl-2-dodecenoic acid) for significant induction of pmr-gfp expression was 0.01 µM, the 27 DSF derivatives were examined at this concentration. The results revealed that all fatty acids with a trans configuration of the double bond at the 2position render the molecule inactive (Fig. 2b), consistent with previous reports indicating this configuration is critical for DSF activity [13][14][15] . However, the presence or the position of the methyl group at C-11 in DSF was dispensable for activity. Indeed, cis-2tridecenoic acid (C9) and cis-2-dodecenoic acid (C2, BDSF) without the methyl substitution, as well as cis-7-methyl-2dodecenoic acid (C6) and cis-8-methyl-2-dodecenoic acid (C5) where the methyl group was transposed to the C-7 and C-8 positions, were equally effective in the bioassay. In contrast, the fatty acid chain length played a major role in determining bioactivity. cis-2-unsaturated fatty acids with chain lengths of 12-14 carbons were active whereas those with longer chains were not (Fig. 2b). The introduction of a second double bond in the chain also led to reduced activity, while derivatives with the carboxylic acid esterified, converted to a hydroxamic acid or amide derivative or replaced with a 2-amino pyridine moiety were all inactive. For each of these derivatives, no activity was seen in the concentration range from 0.01 to 1000 µM. Interestingly, cis-2-tetradecenoic acid (C11), which is a naturally occurring analog  I  II  III  IV  V   I  II  III  IV  V   I  II  III  IV  V   I  II  III IV  V   I  II  III  IV  V   III IV   first detected in Xylella fastidiosa 15 , was most effective in activating the bioassay (Fig. 2c). Strains carrying the T121A/ L123A/L128A replacement mutants of PA1396 showed similar reduced responsiveness to both C11 and DSF (Fig. 2c).
DSF analogs interfere with DSF signaling in P. aeruginosa. The structural analogs unable to trigger DSF-dependent responses were assessed for inhibitory activity against DSF signaling. For these experiments, the molecules were screened at 10μM using P. aeruginosa grown with 50 μM DSF. Several compounds, especially C12, C23, and C24 effectively repressed pmr-gfp expression (Fig. 3a), possibly acting as PA1396 receptor antagonists. None of the compounds affected planktonic bacterial growth (Supplementary Fig. 3). Dose-response assays for each of the three compounds revealed that compound 23 ((Z)-11-methyl-N-(methylsulfonyl)dodec-2-enamide; C23) was the most potent DSF antagonist showing substantial inhibition at 0.1 μM (Fig. 3b).
To examine if these compounds interfere directly with signal transduction through PA1396, we assessed the effects of DSF and the putative antagonists on PA1396 auto-phosphorylation. For these experiments, MycHis-tagged PA1396 was expressed, purified under native conditions, and reconstituted in liposomes where the protein adopts an inside-out orientation in which the cytoplasmic histidine acceptor, histidine kinase and receiver domains are surface exposed. The incorporation of PA1396 into liposomes was confirmed by western blot using anti-Myc antibody, whereas the inside-out orientation can be surmised from the accessibility of ATP to the kinase site, which allows auto-phosphorylation without disruption of the liposomes. Basal auto-phosphorylation of PA1396 in the liposomes was seen in the absence of any signal molecule (Fig. 3c and Supplementary  Table 2). However, the addition of DSF to PA1396-loaded liposomes led to a robust increase in the level of autophosphorylation, an effect not seen in the presence of C23 ( Fig. 3c and Supplementary Table 2). These findings suggest C23 blocks the recognition of DSF by PA1396, which normally mediates increased kinase activity. The observation that the T121A/L123A/L128A mutant protein had a similar level of basal auto-phosphorylation as the parental PA1396, which could not increase upon addition of DSF, also supports this conclusion ( Fig. 3c and Supplementary Table 2). Further, the latter result agrees with the observation that T121A/L123A/L128A has a substantially reduced response to DSF in vivo as measured by the pmr-gfp reporter bioassay (Fig. 1d).
We previously found that DSF alters the expression levels of P. aeruginosa genes implicated in virulence, biofilm formation and stress tolerance 13 . In this study, we assessed the effect of C23 on DSF-regulated functions by quantitative reverse transcription PCR (qRT-PCR) using a subset of 11 DSF-regulated genes. These included genes involved in iron uptake (PA4358, PA4359) and antibiotic resistance (PA4599, PA4774-PA4777). For these experiments, P. aeruginosa was grown in artificial CF sputum medium with and without DSF supplementation, and with both DSF and C23. Cultures were assayed at early log phase (OD 600 of  The biological activity of DSF and its derivatives in activation of the pmr-gfp fusion. Samples were tested at a concentration of 0.01 µM (the minimum concentration of cis-11-methyl-2-dodecenoic acid required for activation of the bioassay). Values are the average of three technical repeats (mean ± standard deviation). Asterisks indicate values that are significantly different from the appropriate wild type (p < 0.05, Student's t-test). c Dose-response of DSF (cis-11-methyl-2-dodecenoic acid) and C11 (cis-2-tetradecenoic acid) on activation of pmr-gfp transcriptional fusion in wild-type PA1396 and a strain expressing the variant T121A/L123A/L128A with alteration in conserved amino acids of TMH-IV. The data of GFP fluorescence shown is from one representative experiment (n = 3). Differences seen between wild type and DSF, and mutant and DSF was statistically significant (p-value < 0.05, ANOVA). Differences seen between wild type and C11, and mutant and C11 was statistically significant (p-value < 0.05, ANOVA) 0.6). For several of the genes tested, the addition of C23 in the growth medium reduced the level of expression seen with DSF alone and in some cases (e.g. PA2966, PA5505) the effects on expression were reversed (Fig. 3d). These experiments were repeated with the ΔPA1396 mutant strain, demonstrating that the addition of DSF or C23 had no effect on the level of gene expression ( Supplementary Fig. 4). Altogether, the results from in vitro auto-phosphorylation experiments and transcriptional profiling of a subset of PA1396-regulated genes indicate C23 interferes with the PA1396 ability to sense DSF.
Interference with DSF sensing decreases biofilm formation and alters virulence of P. aeruginosa in vivo. DSF stimulates the development of Pseudomonas aeruginosa biofilms in vitro and promotes persistent infection in vivo [12][13][14] . We examined if C23 interferes with biofilm formation using a co-culture system in which P. aeruginosa biofilms were allowed to form on human lung epithelial cells upon addition of 1 µM DSF 14 . C23 at concentrations as low as 0.5 µM led to reduced biofilm formation in the presence of DSF (Fig. 4a). Intriguingly, addition of C23 alone also caused a reduction of biofilm development (Fig. 4a). In contrast, ΔPA1396 formed more biofilm than the parental strain, but biofilm formation was not significantly altered by either DSF or C23 (Supplementary Fig. 5a). In a second set of assays, the effects of DSF and C23 on biofilm formation on glass surfaces were examined. For this, bacteria were grown either in glass tubes or in µ-chambers 16 and biofilm development quantified by crystal violet staining. Biofilm biomass in glass tubes was quantified by dye elution whereas biofilms in µ-chambers were visualized by confocal microscopy and biomass quantified by COMSTAT software (see Methods). DSF and C23 had comparable effects on biofilm biomass in glass tubes as seen in the co-culture experiments with epithelial cells: DSF addition led to an increased biofilm biomass, whereas C23 reversed this effect (Fig. 4b). Furthermore, C23 alone was able to inhibit biofilm formation, but not to the same extent as in the presence of DSF. These effects were not seen in experiments using ΔPA1396 mutant strain ( Supplementary Fig. 5b). In µ-chambers, the wild-type strain formed a flat mat biofilm ( Supplementary Fig. 6). Addition of DSF led to the development of a denser flat biofilm whereas Compound Addition of C23 alone also caused a reduction of biofilm and the formation of microcolonies ( Supplementary Fig. 6).
To determine whether C23 modulates bacterial behavior during infections, C57BL/6 mice were inoculated intranasally with suspensions of P. aeruginosa supplemented with C23, DSF, or C23 with DSF. As a control, bacterial inocula were suspended in PBS alone. At 24 h post infection, the P. aeruginosa PAO1 bacterial load was quite considerable in the control animals. However, in the presence of C23, the number of bacteria was considerably reduced. With DSF alone, higher numbers of bacteria were seen after 24 h than in the control (Fig. 4c), in agreement with previous trends seen in other in vivo models 14 . With C23 in combination with DSF, the bacterial load was reduced when compared to DSF alone (Fig. 4c). Additionally, C23 alone appeared to slightly reduce the bacterial load (Supplementary Fig. 7). Thus, C23 reduced the bacterial load in this mouse model in the presence or absence of DSF.
To verify that C23 functions by inhibiting DSF signaling in vivo, we used microarray to compare the effects of adding either DSF alone or DSF with C23 on P. aeruginosa gene expression during mouse lung infection relative to the gene expression of P. aeruginosa in the mouse lung with no compound. RNA was derived from organisms isolated directly from the lung homogenates of infected mice. The results showed that the expression of 68 genes was significantly altered (>1.25-fold, p adj < 1 × 10 −5 ) by the addition of DSF (Fig. 5a and Supplementary Table 3). While many of these differentially expressed genes were originally annotated as hypothetical proteins, several encode factors required by P. aeruginosa for growth and/or host colonization of the host; these genes included norB (PA0524), popD (PA1709), sphA (PA2659) and narK (PA3876). Additional regulated genes encoded factors that contribute to P. aeruginosa virulence and biofilm formation such as the mexC (PA4599), pmr operon (PA4774, PA4776, PA4777) and exoS (PA3841).
Addition of DSF and C23 significantly altered expression of 51 genes (Fig. 5a). Comparison with the effects of DSF alone showed expression changes in 21 genes occurred in both treatments (DSF vs. DSF plus C23) with all 21 of these genes altered in the same direction. The expression level (fold change in expression) of these genes was reduced in DSF plus C23 treatment when compared with the DSF treatment alone (Supplementary Table 3). Further, some genes significantly regulated by DSF showed no significant alteration in the presence of DSF and C23. Examples of this group of genes include popD (PA1709) and pmr operon genes (PA4774, PA4776, PA4777). Also, various genes showed a significant alteration in expression in response to DSF plus C23, but not to DSF alone (Fig. 5a, b). Selected genes were examined using qRT-PCR to confirm the alteration in expression as revealed by the microarray (Fig. 5b). Altogether, the combined results of the experiments described above show that C23 not only can inhibit DSF-activated responses in vitro and in vivo, but also has a broader effect on gene expression independent of DSFregulated responses.
The effect of C23 alone on gene expression was examined in a separate experiment on bacteria grown in artificial sputum medium in the presence or absence of the compound at 50 μM. RNA was extracted from these cultures and the level of expression of selected genes examined by qRT-PCR. Of the genes examined, gcdH (PA0447) and murA (PA4450) appeared to have elevated expression as a response of C23 addition to P. aeruginosa PAO1, but showed little or no response to DSF (Supplementary Fig. 8).
The DSF analog C23 improves antibiotic efficacy. Perception of DSF by P. aeruginosa mediated by PA1396 leads to increased expression of the pmrAB operon and the concomitant increased level of resistance to the cationic antimicrobial peptides polymyxin B and E 12,13 . Further, DSF increases tolerance to polymyxin B in P. aeruginosa biofilms-airway epithelial cells cocultures 14 . These observations prompted us to examine the influence of C23 on the sensitivity of P. aeruginosa to antibiotics. As expected, addition of DSF led to pmrAB up-regulation and an increased resistance to polymyxin B (Fig. 6a, b). By contrast, addition of C23 plus DSF led to a marked reduction in pmrAB expression and reduced resistance to polymyxin B. Although the pmrAB expression level in response to C23 alone was close to Biofilm biomass ** ** **  wild type, C23 did appear to slightly enhance the activity of polymyxin B against PAO1 (Fig. 6a). We also examined the influence of DSF and C23 singly and in combination in resistance to other classes of antibiotics used in treatment of P. aeruginosa. Addition of DSF to the cultures led to increased resistance to tobramycin and nalidixic acid but had no effect on carbapenem resistance. The effect of DSF on increased tobramycin and nalidixic acid resistance was reversed by C23 ( Supplementary Fig. 9).
To validate these results in other strains, we examined a panel of P. aeruginosa clinical isolates and the data revealed that addition of DSF led to a significant increase in the expression of pmrAB in each of these strains (Fig. 6b). However, addition of C23 reversed this DSF-induced effect in all strains, whereas C23 alone gave only modest or no increase in expression. In all of these cases, changes in pmrAB expression were associated with a concomitant alteration in resistance to polymyxin B, as seen with PAO1 ( Supplementary Fig. 9). Further, addition of C23 also reversed DSF-induced tobramycin tolerance in the clinical isolates, which decreased in all strains ( Supplementary Fig. 10). The ΔPA1396 mutant showed increased resistance to polymyxin B, an effect that could not be reversed by addition of C23, as expected. To determine whether PA1396 was also required for the response of clinical isolates of P. aeruginosa to C23, we created PA1396 gene disruption mutants in five strains, which were selected because of their intrinsically higher level of pmrAB expression and polymyxin resistance. None of the mutant strains responded to C23, confirming the C23 antagonist effect on DSF is mediated by PA1396 (Supplementary Fig. 11).
We also assessed the effect of C23 on the efficacy of tobramycin treatment in the C57BL/6 mouse lung infection model. Mice were infected with 1 × 10 7 CFU of PAO1 and treated by inhaling PBS with or without C23. Tobramycin was administered at a concentration of 30 mg/kg 1 h after infection, as previously reported 17,18 . The lung bacterial load was determined 24 h post infection. Control mice infected with P. aeruginosa and inhaled PBS exhibited considerable colonization (Fig. 6c), which was reduced by tobramycin treatment. However, adding C23 with tobramycin resulted in a larger decrease of the bacterial load than that seen with tobramycin alone (Fig. 6c). The kinetics of clearance of ΔPA1396 from mice not treated with antibiotics was similar to that of mice infected with PAO1 ( Supplementary   Fig. 12). Altogether, these results demonstrate that C23 treatment leads to reduced resistance of P. aeruginosa laboratory and clinical strains to distinct antibiotics both in vitro and in vivo during infection, an effect manifested through a functional PA1396 sensor kinase.

Discussion
This study provides a detailed characterization of the membrane topology of PA1396, including the identification of its input domain, which has considerable sequence similarity with the input domain of RpfC, the sensor histidine kinase for DSFmediated cell-to-cell signalling in Xanthomonas campestris pv. campestris 12,19 . However, significant differences between PA1396 and RpfC also exist. RpfC contains a short N-terminal periplasmic region implicated in DSF binding 20 , suggesting DSF is captured in the periplasmic space. Residue E19, which is conserved in all homologs with the RpfC-related input domain, is among various key residues for DSF binding in RpfC 12,20 . In P. aeruginosa PA1396, the corresponding residue is E8, also in the N-terminal periplasmic region. However, deletion of a significant part of the N-terminal region of PA1396 has no effect on DSF sensing by P. aeruginosa. Unlike RpfC, our results indicate the DSF binding capability of PA1396 requires residues in TMHs IV and V. This suggests that DSF likely interacts with residues in these TMHs at the membrane level.
The PA1396 recognition of DSF shares features with the CqsS system of Vibrio cholerae, which detects the quorum-sensing signal CAI-1 (S-3-hydroxytridecanone). Both CqsS and PA1396 have complex membrane spanning domains with multiple TMHs and periplasmic and cytoplasmic loops of limited size and both recognize ligands with alkyl chains of medium length and an amphipathic nature. Nevertheless, substantial differences between the systems occur. In CqsS, conserved residues in the first three (of six) TMHs are obligatory for CAI-1 ligand binding and subsequent signal transduction. Further, residues in the fourth transmembrane helix enable CqsS to discriminate between CAI-1 and amino-CAI-1 and residues in the fifth TMH have roles in restricting ligand head group size and tail length 3,21 . In contrast, PA1396 truncation experiments suggest that the first three TMHs of the protein are dispensable for DSF perception, but whether Previous work on interspecies signaling in P. aeruginosa demonstrated that the cis-2-unsaturated fatty acids DSF and BDSF trigger pmr gene expression, but weak or no activity was observed with the trans-derivatives and related saturated fatty acids. Our analysis provides detailed structural requirements for interspecies signaling. All trans-unsaturated fatty derivatives, irrespective of chain length, had little or no signaling activity on pmr gene expression. Similarly, any derivative lacking a free carboxylic acid group (through esterification, conversion to a hydroxamic acid or substitution with a 2-amino pyridine moiety) was inactive. The signaling activity was not significantly affected by the position of the methyl group (at position 11 in DSF) but was partially reduced by introducing a second double bond. Notably, a derivative of DSF with a second double bond (cis, cis-11-methyldodeca-2,5-dienoic acid) occurs naturally in various Xanthomonas species, but also has reduced bioactivity compared to DSF to trigger DSF-regulated genes 22 . Further, interspecies signaling in P. aeruginosa depends on the chain length of the cisunsaturated fatty acid; molecules with chain lengths longer than 14 (this work) and less than 12 12,13,23 are inactive. In particular, PA1396 does not respond to cis-2-decenoic acid, a molecule produced by P. aeruginosa and which is implicated in biofilm dispersal 12,13,20 .
We demonstrate that DSF and cis-unsaturated fatty acid analogs enhance the tolerance of P. aeruginosa to a range of antibiotics, including polymyxins, tobramycin, and nalidixic acid; this action requires the sensor kinase PA1396. In contrast, Deng et al. 24 , reported that both DSF and its analogs promote sensitivity of several Gram-positive pathogens including Bacillus cereus and Staphylococcus aureus to various antibiotics, including kanamycin and gentamicin. The molecular mechanisms underlying these effects in B. cereus are unclear, but it is intriguing that cis-and trans-2-unsaturated fatty acids are equally effective in inducing a response.
We reasoned that if DSF signaling activates expression of genes involved in antibiotic tolerance, structural analogs of the molecule could act as antagonists. Accordingly, we show that various effects of DSF on P. aeruginosa, including induction of pmr gene expression and persistence in the mouse model is antagonized by the C23 analog. Antagonism of DSF action may have an application to control of P. aeruginosa in multispecies infections with pathogens that produce DSF family signals. An example is infection associated with cystic fibrosis (CF), where P. aeruginosa can occur together with S. maltophilia and Burkholderia species, both of which produce DSF and BDSF. The detection of DSF and BDSF signals at physiologically relevant levels in the sputum of CF patients supports the contention that interspecies signaling may occur in these infections, where it may affect the antibiotic sensitivity of P. aeruginosa as suggested by in vitro experiments.
Intriguingly, C23 has additional modulating effects on P. aeruginosa that cannot be attributed to its role as a DSF antagonist. For example, C23 affects gcdH and murA expression, two genes that do not respond to DSF. C23 also affects biofilm formation, both in co-culture with lung epithelial cells and in 'complex' Fig. 6 Effects of DSF and C23 on antibiotic resistance of P. aeruginosa strains. a Effect of addition of DSF, C23, and DSF together with C23 on tolerance of P. aeruginosa to polymyxins. Time-courses of killing of P. aeruginosa by 4 μg/ml polymyxin B were established for bacteria suspended in sodium phosphate buffer. Bacteria for these experiments were grown in BM2 medium with glucose supplemented with 2 mM Mg 2+ . DSF and C23 were added to these cultures to a final concentration of 50 μM. Values are the average of three technical repeats (mean± standard deviation). The observed values of addition of DSF and C23 were significantly different from the addition of DSF alone (p < 0.05, ANOVA). b Effect of addition of DSF, C23, and DSF with C23 on antibiotic resistance gene expression in clinical isolates of P. aeruginosa as compared to the laboratory strain PAO1. Transcript levels of PA4775 in PAO1, the isogenic ΔPA1396, and a selection of clinical isolates of P. aeruginosa (CF3, CF9, CF14, CF21, and CF 22) were examined using qRT-PCR as described in Methods. The qRT-PCR data were normalized to proC and are presented as the fold change relative to the wild type for each gene in each strain with no addition. Data (means ± standard deviation) are from one representative experiment (n = 3). A pvalue of <0.05 was considered statistically significant and is designated in the figures with an asterisk. Double asterisks indicate p-values of <0.01. c Effect of administration of C23 on P. aeruginosa clearance by tobramycin treatment in the mouse airway. C57BL/6 mice were infected intranasally with 1 × 10 7 CFU P. aeruginosa PAO1 and treated by inhaling PBS with or without 50 μM C23. Tobramycin was administered at a concentration of 30 mg /kg 1 h after infection. After 24 h of infection, the mice were harvested, and bacterial loads were determined in lung homogenates. Each data point shows the results from an individual mouse. A p-value of <0.05 was considered statistically significant and is designated in the figures with an asterisk. Double asterisks indicate p-values of <0.01 medium, in a DSF-independent manner and contributes to a more rapid clearance of P. aeruginosa from the lungs of mice. More detailed molecular knowledge of the activities of PA1396 may afford insight into these different actions of C23. For example, can PA1396 sense other environmental signals for biofilm formation (in addition to DSF) perhaps through TMHs I-III? Is transduction of these signals also blocked by C23? Can C23 alone cause reduced auto-phosphorylation of PA1396 in vivo? Work has indicated the two-component regulator PA1397 is involved in signal transduction beyond PA1396 14 . However, the findings are not consistent with a simple model in which DSF-induced phosphorylation of PA1397 directly activates pmr gene expression, indicating that other regulatory elements are involved. Future work will be directed at identifying these components.
In conclusion, pharmacological inhibition of DSF-mediated interspecies signaling in P. aeruginosa using a DSF analog potential has identified a potential lead compound to develop molecules that could be used as antibiotic adjuvants improving the efficacy of existing antibiotics and help control diseases caused by this human pathogen.
DNA manipulation. Molecular biological methods such as isolation of plasmid and chromosomal DNA, PCR, plasmid transformation, as well as restriction digestion were carried out using standard protocols. PCR products were cleaned using the Qiaquick PCR purification kit (Qiagen) and DNA fragments were recovered from agarose gels using Qiaquick minielute gel purification kit (Qiagen). Oligonucleotide primers were purchased from Sigma-Genosys. Primer sequences are provided in Supplementary Table 5.
Cloning the PA1396 gene. The DNA fragments encoding the full-length PA1396 protein or truncation of interest were synthesized by Gene Oracle (Santa Clara, USA) in pGOv4 and sub-cloned into pET47b, pME6032 or pBAD/Myc-His before transformation into E. coli BL21 (DE3). Genomic regions are described in Supplementary Table 5. BL21 (DE3) cells were grown in LB media and induced with 0.25 mM IPTG; protein overexpression was carried out at 37°C for 1 h. Purification was achieved by Ni 2+ affinity chromatography using the N-terminal His6 tag.
Suitable fusion sites in the periplasmic and cytoplasmic loops were identified using a consensus based on the TMS prediction data obtained ( Supplementary  Fig. 1). The sites selected correspond to amino acids E4, N6, V37, E38, G66, A70, G107, G109, A137, Q139, Q141, and L205. The DNA fragments encoding the proteins of interest were synthesized by Gene Oracle (Santa Clara, USA) in pGOv4 and sub-cloned into topology reporter plasmids phoA (pRMCD28) and lacZ (pRMCD70). Genomic regions are described in Supplementary Table 5.
The fusion junction of each construct was confirmed by sequencing. Alkaline phosphatase assays were carried out according to Daniels et al. 31 . β-Gal assays were carried out according to Baker et al. 32 , except that overnight cultures were subcultured and grown to OD < 0.6 before activity assays. Samples were assayed in triplicate over at least three independent experiments.
Synthesis of DSF analogs. The starting point for the synthesis of cis-2-dodecenoic acid C2 (BDSF) and 11-methyldodec-2-enoic acid C1 (DSF) involved a Swern oxidation of the starting alcohol with dimethyl sulfoxide (DMSO) and oxalyl chloride at -78°C to corresponding aldehydes decanal and 11-methyldecanal 33 . A subsequent Wittig reaction of the purified aldehyde with the modified Horner-Wadsworth-Emmons phosphonate salt ethyl [bis(2,2,2-trifluoroethoxy)-phosphinyl]acetate in the presence of sodium hydride (NaH) in THF afforded both the cisand trans-α,β-unsaturated esters, ethyl dodec-2-enoate and ethyl 11-methyldodec-2-enoate 34 . Hydrolysis of the cis-α,β-unsaturated esters with lithium hydroxide (LiOH) in THF:MeOH:H2O (2:1:1 (v/v/v)) gave required products C2 and C1 35,36 . Treatment of C1 with propylphosphonic anhydride and cyclopropylamine gave the required amide analog 37 . DSF antagonist C23 was prepared by way of an EDCImediated coupling of C1 with methanesulfonamide in the presence of dimethylaminopyridine and subsequent isolation of the desired cis-isomer by reverse phase HPLC. All the chemicals and reagents were purchased from Sigma-Aldrich unless otherwise stated. Additional synthetic procedures and analytical data used in this study are delineated in Supplementary Note.
DSF analog structure analysis. Nuclear magnetic resonance spectra 1 H, 13 C, 1 H-1 H COSY, and DEPT were recorded on a Bruker Avance 400 NMR Spectrometer (400 MHz for 1 H) and a Bruker Avance 500 NMR Spectrometer (500 MHz for 1H and 125 MHz for 13 C) with trimethylsilylchloride as an internal standard in CDCl 3. 1 H-13 C correlated HMBC and HMQC spectra were performed by Bruker Avance 500 spectrometer. Mass spectra ESI-MS and high-resolution mass spectra ESI-MS were performed on a Waters/Micromass: LCT Premier Time of Flight and a Quattro Micro triple quadrupole instruments, respectively. Infrared spectra were measured using NaCl plates on a Perkin Elmer paragon 1000 Fourier-transform infrared spectroscopy (FT-IR) spectrometer.
DSF bioassays. The original DSF bioassay is based on its ability to restore endoglucanase production to rpfF mutant of Xcc as described in ref. 38 . DSF activity was expressed as the fold increases in endoglucanase activity over the control. We also used the bioassay previously described 13 that relies on DSF-dependent induction of fluorescence (pmr-gfp) in P. aeruginosa PAO1.
Reconstitution and phosphorylation of PA1396-His in liposomes. Using a variation in the method previously described 39,40 , E. coli strains containing pBAD/ Myc-His (PA1396-pBAD/Myc-His) were induced with 0.2% arabinose and purified through nickel columns according to the manufacturer's instructions (Qiagen).
Liposomes were reconstituted as previously described [40][41][42] . Briefly, 50 mg of E. coli phospholipids (44 μl of 25 mg/ml; Avanti Polar Lipids) were evaporated and then dissolved into 5 ml of potassium phosphate buffer containing 80 mg of Noctyl-β-D-glucopyranoside. The solution was dialyzed overnight against potassium phosphate buffer. The resulting liposome suspension was subjected to freeze-thaw in liquid nitrogen. Liposome size was analyzed by dynamic light scattering. Liposomes were stored at 4°C. Liposomes were then destabilized by the addition of 26 mg of dodecylmaltoside, and 5 mg of PA1396-His (dual Myc-and His-tagged protein) was added, followed by stirring at room temperature for 10 min.
Two-hundred-sixty milligrams of Biobeads (BioRad) were then added to remove the detergent, and the resulting solution was allowed to incubate at 4°C overnight. The supernatant was then incubated with fresh Biobeads (BioRad) for 1 h in the morning. The resulting liposomes containing reconstituted PA1396-His were frozen in liquid N2 and stored at −80°C until used. The orientation of HKs in the liposome system was established by other groups and can be confirmed from the accessibility of ATP to the kinase site and anti-Myc antisera (Invitrogen R950-25, 1:1000 dilution) to the C-terminal PA1396-MycTag without disruption of the liposomes.
Twenty microliters of the liposomes containing PA1396-His were adjusted to 10 mM MgCl2 and 1 mM DTT, and various concentrations of agonist or antagonist, frozen and thawed rapidly in liquid nitrogen, and kept at room temperature for 1 h (this allows for the signals to be loaded within the liposomes).
[γ32P]dATP (0.625 μl) (110 TBq/mmol) was added to each reaction. To some reactions, 10 μg of PA1396-His was added. At each time point (0, 10, 30, 60, or 120 min), 20 μl of SDS loading buffer was added. For all experiments involving PA1396 alone, a time point of 10 min was used. The samples were run on SDS/PAGE without boiling and visualized using a Molecular Dynamics PhosphorImager. These were quantified by ImageJ software.
Biofilm assays. Biofilm was assessed by attachment to glass and was determined by crystal violet staining. Log-phase-grown bacteria were diluted to OD 600 nm = 0.02 in LB broth and 5 ml was incubated at 37°C for 24 h in 14-ml glass tubes. After gently pouring off the media, bacterial pellicles were washed twice with water and were then stained with 0.1% crystal violet. Tubes were washed and rinsed with water until all unbound dye was removed. Bound crystal violet was eluted in ethanol and measured at OD 595 . Three independent assays were carried out for each strain. For P. aeruginosa biofilms developed on cystic fibrosis bronchial cells (CFBE) cells using a co-culture model system as previously described 14 . In short, epithelial cells were seeded at a concentration of 10 6 cells/well in 6-well tissue culture plates or 2 × 10 5 cells/well in 24-well tissue culture plates and maintained in minimal essential medium (MEM) with 10% fetal bovine serum, 2 mM l-glutamine, 50 U/ml penicillin, and 50 μg/ml streptomycin. The cells were grown at 37°C and 5% CO 2 until~90% confluence before inoculation with bacteria. P. aeruginosa was inoculated at a concentration of~2 × 10 7 CFU/ml into 1.5 ml MEM/well (without fetal bovine serum, penicillin, or streptomycin. After 1 h, the supernatant was replaced with fresh MEM supplemented with 0.4% arginine, and the system was incubated at 37°C and 5% CO2 for 4 h. To determine the numbers of biofilm CFU, we washed the cells two or three times with phosphate-buffered saline (PBS) to remove planktonic bacteria and treated the cells with 0.1% Triton X-100 for 10 to 15 min to lyse the epithelial cells. For P. aeruginosa biofilms developed in μ-well chambers, compounds (0.5 µM) were added to in ABTGC media and biofilms were developed for 16 h in chambers (Ibidi) as previous described 16 . Confocal laser scanning microscopy was used to image biofilms and biofilm biomass was quantified using COMSTAT 14 .
Polymyxin killing curves. Killing curves were carried out at 37°C temperature as previously described 13,19 . In short, mid-log-phase cultures (OD 600 0.4-0.6) of PAO1 wild-type strain, PAO1 supplemented with 50 µM DSF or the ΔPA1396 mutant grown in BM2 medium supplemented with 0.5 mM and 2 mM Mg 2+ were diluted 1:100 into 30 mM sodium phosphate buffer, pH 7.0, containing 4 μg/ml of polymyxin B sulfate and 5 μg/ml polymyxin E sulfate (Sigma). Samples were shaken gently, and aliquots removed at specified time intervals were assayed for survivors by plating appropriate dilutions onto LB agar.
Infection and treatment of animals. The mouse infection experiments carried out were assessed by the Queen's University Belfast animal welfare and ethical review body (AWERB) committee and conducted under a license issued by the UK Home Office under the Animals (Scientific Procedures) Act 1986, amended in 2012.
Mouse infection was performed using a variation on the mouse model described in refs. 17,19 . Briefly, P. aeruginosa strains were grown in Luria broth at 37°C overnight with shaking, after which bacteria were collected by centrifugation and resuspended in PBS. The exact number of bacteria was determined by plating serial dilutions of each inoculum on pseudomonas isolation agar plates. Female C57BL/6 mice (~8-weeks-old) were anesthetized and intranasally inoculated with 20 μl of the bacterial suspension (1 × 10 7 CFU/mouse) in PBS with PAO1, PAO1/DSF, PAO1/ or selected DSF analog or ΔPA1396. Mice were killed at 24 h post infection by intraperitoneal injection of 0.3 ml of 30% pentobarbital. Lungs were harvested aseptically and homogenized in sterile PBS. A 10-fold serial dilution of lung homogenates was plated on Pseudomonas isolation agar. The results, mean ± standard deviation are expressed as CFU/lung.
For animal experiments examining the effectiveness of tobramycin treatment against P. aeruginosa during respiratory infection, the mice were intranasally inoculated with 20 μl of the bacterial suspension in PBS with PAO1, PAO1/DSF, PAO1/or selected DSF analog (1 × 10 7 CFU/mouse). Each mouse received was administered tobramycin 1 h post infection. Twenty-four hours after infection, the lungs were removed, homogenized with PBS and plated on agar to determine the number of viable bacteria. The results (mean ± standard deviation) are expressed as CFU/lung. Gene expression-profiling experiments. For the in vivo RNA samples, infected animals as described above were euthanized~24 h post infection 43,44 . RNA was isolated from airway homogenate pellets using Trizol according to the manufacturer's instructions. Samples were treated with DNAse, then rRNA was removed from 1 to 5 μg of total RNA using the RiboZero kit for Gram-negative bacteria (Epicenter). rRNA-depleted samples were concentrated by EtOH precipitation. cDNA synthesis and hybridization to Affymetrix GeneChip P. aeruginosa genome arrays were carried out by a commercial Affymetrix Genechip service supplier (Conway Institute of Biomolecular & Biomedical Research, UCD, Ireland). Microarray data analysis was performed using Biomedical Genomics Workbench software (Qiagen).
Quantitative real-time PCR assays. P. aeruginosa strains (including mutants) were inoculated at an OD 600 of 0.1 in LB (with 50 µM DSF-like signal molecules if used). Cultures were harvested during mid-log-phase (OD 600 = 0.6) and RNA was extracted using the High Pure RNA Isolation Kit (Roche). The expression of genes was monitored by qRT-PCR, as described previously 17,19 Similar fold changes with respect to the wild type for each gene was seen regardless of normalized to 16s rRNA and proC. Only proC normalization is reported.
Statistical analysis. Significant differences between the means ± the standard deviation of two different groups were examined using a Student's t-test; for more than two groups one-way ANOVA was used. To determine the differences between infected mice group Mann-Whitney U-test was used. A p-value of <0.05 was considered statistically significant and is designated in the figures with an asterisk. Double asterisks indicate p-values of <0.01.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The microarray data have been submitted to the Gene Expression Omnibus (GEO) database with accession code GSE110126 . Source data underlying Figs. 1c, d, 2b, c, 3a, b, 3d, 4a-c, 5b, 6a-c, Supplementary Figs. 1, 3-12, and Supplementary Tables 1, 2 are provided as a Source Data file. All the data produced for this study are available from the corresponding authors upon request.