We previously reported the molecule chlorolactone (CL) (Fig. 1A) as an inhibitor of the LasR/RhlR homolog CviR from Chromobacterium violaceum (21). Based on the CL structure, we designed and synthesized a focused library of ~30 molecules (21) (SI Appendix, Fig. S3A). Here, we present our analyses for CL and three additional molecules [chloro-thiolactone (CTL), meta-chloro-thiolactone (mCTL), and meta-bromo-thiolactone (mBTL); Fig. 1A]. First, we tested the molecules for inhibition of quorum sensing in WT P. aeruginosa PA14 using the quorum-sensing–regulated production of the virulence factor pyocyanin as the readout (Fig. 1C). Following activation by LasR:3OC12–HSL, the RhlR:C4–HSL complex activates expression of the pyocyanin operon (15, 22–24) (Fig. 1B). Therefore, P. aeruginosa PA14 produces pyocyanin at high cell density, and limited pyocyanin production occurs in a lasR or a rhlR null mutant and in the lasR, rhlR double mutant (Fig. 1C).

Neither CL nor CTL inhibited pyocyanin production in vivo (Fig. 1C). By contrast, mCTL and mBTL showed potent inhibition of pyocyanin (Fig. 1C) while not affecting P. aeruginosa PA14 growth. We calculated 50% inhibitory concentration (IC₅₀) values of 8 μM (±2) for mBTL and 9 μM (±2) for mCTL (Fig. 1D). Inhibition was not due to interaction of these molecules with pyocyanin, as incubation of pyocyanin-containing cell-free culture fluids from untreated P. aeruginosa PA14 with 100 μM mBTL or mCTL did not alter pyocyanin levels (SI Appendix, Fig. S1). We synthesized four previously reported LasR inhibitors for comparison in our pyocyanin assay (SI Appendix, Fig. S2A) (18–20). Two compounds, itc-13 and V-06–018, inhibited pyocyanin production in P. aeruginosa PA14 with higher IC₅₀ values [56 μM (±10) and 18 μM (±2), respectively] than mBTL and mCTL (SI Appendix, Fig. S2 B and C). The other molecules (PD-12 and B7) showed limited or no inhibition in vivo. (SI Appendix, Fig. S2B). Thus, mBTL is more potent than mCTL and these previously identified inhibitors.

To determine the enantiomer of mBTL responsible for inhibition, we performed a chiral separation. The S enantiomer is active (IC₅₀ = 4 μM), whereas the R enantiomer displays residual activity (IC₅₀ = 100 μM) (SI Appendix, Fig. S3 A and B). Because the racemic mixture has an IC₅₀ value only twofold higher than the isolated S enantiomer, we conclude that the R enantiomer does not influence inhibition. For ease of synthesis, we use racemic mixtures for the remainder of the work.

mBTL contains a four-carbon linker (Fig. 1A). We synthesized and tested mBTL derivatives with two- to six-carbon linkers and lactone versions of mBTL harboring four- or six-carbon linkers, containing or lacking a 3-oxo-moiety. None of these compounds exhibited increased potency (SI Appendix, Fig. S3 A, C, and D). Thus, mBTL remains the most potent in vivo inhibitor in our collection.

We investigated whether mBTL interacts with LasR, RhlR, or both receptors using recombinant Escherichia coli strains producing the receptor proteins and containing target gfp reporter fusions (rsaL–gfp for LasR and rhlA–gfp for RhlR). In the absence of ligand, neither receptor activates expression of the target–gfp fusion (Fig. 2 A and B). When autoinducer is supplied at 95% maximal effective concentration (EC₉₅), LasR and RhlR activate gene expression. mBTL partially antagonizes this effect (22 ± 4% and 43 ± 10% inhibition, respectively; Fig. 2 A and B). In the absence of autoinducer, mBTL, when provided at the EC₉₅ of the native ligand for each receptor, acts as a partial agonist of both LasR and RhlR (72 ± 3% and 80 ± 25%, respectively). As we have previously reported (21), molecules such as mBTL that resemble autoinducers induce conformational changes in LuxR-type receptors that impair their ability to interact with RNA polymerase which lowers their transcriptional activation potential. In the presence of autoinducer, this reduction in transcription potential manifests as antagonism. In the absence of autoinducer, some transcription activation can still occur. Thus, without autoinducer present, molecules such as mBTL appear as partial agonists.

Open in a separate window

Fig. 2.

mBTL binds and inhibits LasR and RhlR, and the primary in vivo target of mBTL is RhlR. (A) LasR activation of expression of rsaL–gfp in E. coli. (B) RhlR activation of expression of rhlA–gfp in E. coli. In each panel, gfp expression in the presence of the cognate autoinducer (100 nM 3OC12–HSL or 20 μM C4–HSL) is set to 100%. mBTL was tested for inhibition at 1 mM. Agonism was examined at 100 nM for LasR and at 20 μM for RhlR. Error bars represent SD of three replicates. (C) SDS/PAGE analysis of whole cell (WC) and soluble (S) extracts from E. coli cultures expressing LasR in the presence of DMSO, 100 μM 3OC12–HSL, or 100 μM mBTL. An uninduced control (UN) is shown for comparison. (D) Same as C with RhlR in the presence of DMSO, 100 μM C4–HSL, or 100 μM mBTL. (E) Hierarchical clustering, heat maps, and the RMS of the fold change (log₁₀) of mBTL-treated (+) or DMSO-treated (−) WT P. aeruginosa, lasR, rhlR, and rhlI mutants. Dendrogram to the left of the map indicates average Euclidean linkage distances between the gene expression profiles. Blue and yellow indicate decreased and increased expression, respectively. Data are the average of three independent biological experiments, one in which the Cy3 and Cy5 dyes were swapped.

LuxR-type proteins require cognate autoinducers to fold (3, 4). Consistent with this, LasR and RhlR are insoluble in the absence of autoinducer and are present in the whole cell (WC) fraction but not the soluble (S) fraction following SDS/PAGE (Fig. 2 C and D). Addition of 100 μM 3OC12–HSL or mBTL solubilizes LasR (Fig. 2C). mBTL also solubilizes RhlR; surprisingly, C4–HSL, its native autoinducer, does not (Fig. 2D). We suspect that autoinducer-bound RhlR is not particularly soluble when overproduced in E. coli, and the protein likely aggregates. mBTL appears to protect the protein from aggregation. We have previously demonstrated that similar protection occurs with CviR bound to CL (21).

To verify that mBTL functions by inhibiting quorum sensing in vivo, we used microarrays to examine the consequences of administering mBTL to WT P. aeruginosa and to the lasR and rhlR mutants. Treatment of WT P. aeruginosa PA14 caused alterations in expression of many of the known quorum-sensing targets (SI Appendix, Table S1). For example, LasR–regulon genes including rpoS and nor were down-regulated (SI Appendix, Tables S1 and S2). RhlR-controlled virulence genes, for example, those encoding rhamnolipids (rhlA and rhlB), and phenazines (phzA2, phzB1, and phzB2) were also repressed (SI Appendix, Table S1 and S3). Indeed, the profile of WT cells treated with mBTL matches well with the combined profiles of the lasR and rhlR mutants (SI Appendix, Table S1). However, the fold activation and fold repression is not as dramatic in the mBTL-treated WT as in the mutants, confirming that mBTL does not fully inhibit either regulator (SI Appendix, Table S1–S3). Thirteen genes are activated twofold or more in WT P. aeruginosa PA14 when treated with mBTL (SI Appendix, Table S4). By contrast, 213 genes are down-regulated twofold or more when WT is treated with mBTL (SI Appendix, Table S1). These data indicate that the major role of mBTL in WT P. aeruginosa PA14 is as an antagonist that exerts control over virulence through partial inhibition of LasR and RhlR, not via up-regulation of other genes.

The most important comparisons for defining the target of mBTL are the mBTL-treated and untreated lasR and rhlR mutants (SI Appendix, Tables S5 and S6 and Fig. 2E). In the absence of LasR, mBTL treatment still alters expression of some genes. By contrast, there is essentially no difference between the untreated and mBTL-treated rhlR mutant profiles. Thus, although some of the mBTL effects occur through LasR, all of the mBTL effects depend on RhlR (SI Appendix, Tables S5 and S6 and Fig. 2E). Thus, mBTL functions in vivo by inhibiting quorum sensing via RhlR.

Given that mBTL acts as a partial agonist of RhlR in recombinant E. coli when the cognate autoinducer C4–HSL is not present (Fig. 2B), we explored further the extent of mBTL agonism of RhlR in vivo. To do this, we treated a P. aeruginosa rhlI null strain with mBTL and performed microarrays. mBTL activated twofold or higher expression of 44 genes (SI Appendix, Table S7). For comparison, RhlR bound to C4–HSL activated nearly 300 genes (SI Appendix, Table S3). In every case in which a gene could be activated by C4–HSL or by mBTL, C4–HSL was a much more potent activator. Importantly, only one gene that was activated by mBTL in the rhlI mutant was activated by mBTL in WT P. aeruginosa (SI Appendix, Table S7 and S4). These data indicate that mBTL can act as a weak RhlR agonist in vivo, but only when the native autoinducer is absent, a situation that is not likely to occur in WT bacteria.

To determine if mBTL can limit virulence, we used a Caenorhabditis elegans fast-kill infection assay (8, 25). WT P. aeruginosa PA14 and the lasR mutant rapidly kill C. elegans: 77 and 90% of worms die after 24 h, respectively (Fig. 3A), indicating that LasR is not required for pathogenesis in this assay. Much less killing occurs when the nematodes are exposed to the rhlR or the lasR, rhlR double null strain (31 and 17%, respectively, died in 24 h; Fig. 3A), showing that RhlR is required for virulence in this assay. Treatment of WT and the lasR mutant P. aeruginosa PA14 strains with 50 μM mBTL protects C. elegans from killing (23 and 50% death, respectively; Fig. 3B and SI Appendix, Fig. S4). Together, these results confirm that the relevant in vivo target of mBTL is RhlR and, importantly, that inhibiting RhlR could form the basis of an antibacterial therapeutic strategy.

Open in a separate window

Fig. 3.

mBTL inhibits P. aeruginosa PA14 virulence toward C. elegans and human A549 lung cells. (A) C. elegans were applied to lawns of E. coli HB101 (circles), WT P. aeruginosa PA14 (squares), lasR mutant (triangles), rhlR mutant (inverted triangles), and lasR, rhlR double mutant (diamonds) strains. The percent live worms was calculated every hour for the first 5 h and again at 24 h. Error bars represent SEM of three replicates. (B) Same as A. E. coli HB101 (circles), WT P. aeruginosa PA14 (squares), and WT P. aeruginosa in the presence of 50 μM mBTL (asterisks). (C) The percent cell death was calculated using propidium iodide uptake into A549 lung cells after 8 h and normalized to cells lysed with detergent. Error bars represent SEM of three replicates.

We tested whether mBTL could improve the outcome for mammalian cells during infection using the human lung carcinoma cell line A549. mBTL at 100 μM is not toxic to A549 cells (SI Appendix, Fig. S5). WT, lasR, and rhlR single mutants and the lasR, rhlR double mutant P. aeruginosa are all capable of killing A549 cells (Fig. 3C). Treatment with mBTL reduces killing by the WT and the lasR mutant strain but not by the rhlR or the lasR, rhlR double mutant strains (Fig. 3C). These results validate the conclusion reached on the basis of the microarray data presented above. Specifically, the relevant target of mBTL is present in the WT and the lasR mutant but is not in the rhlR single and lasR, rhlR double mutant strains. Thus, RhlR appears to be the major in vivo target.

Beyond being a clinically relevant pathogen, P. aeruginosa is an industrial and medical nuisance because it causes blockages in filtration devices and stents. P. aeruginosa PA14 also clogs microfluidic chambers that model such devices (26). Clogging is due to biofilms that produce exopolysaccharide-containing streamers that act as sieves to catch passing cells (26). Compared with WT, lasR and rhlR single and double null mutants exhibit dramatically delayed clogging, demonstrating that quorum sensing is required to form blockages (Fig. 4A). In the presence of 100 μm mBTL, WT P. aeruginosa PA14 exhibits a phenotype indistinguishable from the mutants (Fig. 4A). We also examined the ability of mBTL to influence static biofilm formation in P. aeruginosa PA14. WT P. aeruginosa PA14 forms biofilms with an average height of 27.5 ± 11.5 μm (Fig. 4B). Treatment of WT with 100 μM mBTL results in a decrease in the average height to 10 ± 4 μm (Fig. 4B). Treatment of the lasR null strain with mBTL also decreases the average height of the biofilm (15.7 ± 4.9 μm to 6.6 ± 2.3 μm). However, treatment of the rhlR null strain (25 ± 10.6 μm untreated vs. 22.1 ± 7.5 μm treated) or the lasR, rhlR double null strain (11.9 ± 5.5 μm untreated vs. 16.2 ± 5.9 μm treated) did not influence the height of the biofilm, indicating that the action of mBTL relies on the presence of RhlR (Fig. 4B). Taken together these findings show that mBTL inhibits quorum-sensing–controlled biofilm formation in P. aeruginosa PA14.

Open in a separate window

Fig. 4.

mBTL inhibits quorum-sensing–regulated clogging of microfluidic chambers and biofilm formation in static cultures. (A) Time to clogging was measured for the designated P. aeruginosa PA14 strains and for the WT in the presence of 100 μM mBTL. Error bars represent SD of six replicates. (B) Biofilms were grown in static cultures at the base of a glass-bottom microtiter plate in the presence or absence of 100 μM mBTL. Biofilm thickness was measured using confocal microscopy. Error bars indicate SD of five to eight biological replicates.

This work exploring interference with quorum sensing demonstrates that mBTL, an analog of the native P. aeruginosa autoinducers, represses expression of the genes encoding the virulence factor pyocyanin, prevents biofilm formation, and protects C. elegans and human lung epithelial cells from killing by P. aeruginosa. Both the LasR and RhlR quorum-sensing receptors are partially inhibited by mBTL; however, in the contexts that we have examined, RhlR, not LasR, is the relevant in vivo target. Much emphasis has been placed on the discovery of LasR inhibitors because LasR is situated at the top of the P. aeruginosa quorum-sensing cascade (27). Our results suggest that the critical in vivo target for modulation is RhlR. Experiments in Drosophila melanogaster also demonstrate that the P. aeruginosa PA14 lasR mutant strain is virulent, whereas the rhlR mutant strain is attenuated (28) suggesting, as do our results, that RhlR is the pertinent in vivo quorum-sensing regulator.

It is curious that mBTL, our most potent in vivo inhibitor, is a partial agonist/partial antagonist of RhlR and LasR in the recombinant E. coli assay (Fig. 2 A and B) and according to the P. aeruginosa microarrays (SI Appendix, Table S1). This feature of mBTL may be critical for its ability to function in vivo. Because LasR and RhlR act reciprocally to control key virulence genes (Fig. 1B), molecules that function as pure antagonists of both receptors might not be superior antivirulence therapies. Consistent with this notion, CL is a more potent inhibitor of both LasR- and RhlR-directed transcription in E. coli than is mBTL, and CL does not agonize either receptor (SI Appendix, Fig. S6). Nevertheless, CL exhibits no in vivo inhibition of pyocyanin production (Fig. 1 B and C), and it does not attenuate killing of A549 lung cells by P. aeruginosa. We suspect that a “sweet spot” exists in terms of antagonist potency for inhibitors of LasR and/or RhlR and that mBTL has the appropriate level of potency to be effective in vivo. Consider a case in which LasR activates and RhlR represses a particular virulence factor (see Fig. 1B). A highly potent, dual LasR–RhlR antagonist could result in derepression and thus production of the virulence factor. Many examples exist in which LasR and RhlR act reciprocally in their regulation of particular genes, including virulence genes (9, 11, 17).

It is noteworthy that, in the lung cell assay, the lasR, rhlR double mutant causes more cell death than does the WT (Fig. 3C). This result may stem from misregulation of virulence factors that are controlled in opposing directions by RhlR and LasR (Fig. 1B). In the preceding paragraph, we argue that potent inhibition of both LasR and RhlR could result in increased pathogenicity. The phenotype of the double mutant in the lung cell assay supports this idea. Complete inhibition of LasR and RhlR would yield the same phenotype as deletion of both receptors. Apparently, that phenotype is high virulence. Consistent with our results and interpretation, the quadruple ΔlasI, ΔlasR, ΔrhlI, ΔrhlR mutant remains virulent in a mouse lung infection model (29). Lastly, we note that the double lasR, rhlR null mutant is not hypervirulent in the nematode assay. The quorum-sensing–controlled virulence factors required for pathogenicity in mammalian cells are not precisely identical to those that are essential for virulence in nematodes which presumably accounts for this discrepancy (5, 8). What is critical is that, despite these differences, mBTL attenuates virulence of WT P. aeruginosa PA14 in both assays and in a RhlR-dependent manner.

In the nematode and lung cell experiments shown here, we pregrew the bacteria with mBTL and supplied a dose of inhibitor at the start of infection. Virulence was not reduced when P. aeruginosa PA14 was pregrown in the absence of inhibitor (SI Appendix, Fig. S7), indicating that mBTL functions as a prophylactic. The use of prophylactics, specifically for P. aeruginosa, is an established procedure, so anti–quorum-sensing prophylactic molecules could find use in the clinic. Further optimization of mBTL could tune efficacy and delivery and possibly broaden use beyond prophylaxis. An immediate focus is the head group given that thiolactones are susceptible to hydrolytic ring opening.

Our results showing that mBTL prevents biofilm formation and clogging in microfluidic devices (Fig. 4) suggest that, in addition to the implications of mBTL as an antiinfective, deploying anti–quorum-sensing molecules could prevent the failure of devices that are prone to fouling by biofilms. Thus, such molecules could be useful in industrial settings and in contexts requiring medical implants. Here we administered the inhibitor in the liquid phase; a next goal is to embed mBTL-like molecules in materials used for production of such devices, resulting in products that are innately resistant to biofilms. Taken together, our data concerning mBTL make an excellent case for the effectiveness of quorum-sensing modulators for the attenuation of quorum-sensing–controlled phenotypes in pathogenic bacteria.