The contribution of MvfR to Pseudomonas aeruginosa pathogenesis and quorum sensing circuitry regulation: multiple quorum sensing-regulated genes are modulated without affecting lasRI, rhlRI or the production of N-acyl- l-homoserine lactones

Production of AHLs is not affected in the mvfR mutant

Although the MvfR transcriptional regulator influences the expression of a number of QS-controlled genes and phenotypes in P. aeruginosa, the transcription of lasR and rhlR is not affected in a mvfR mutant (Cao et al., 2001). To better understand how MvfR modulates the expression of QS-controlled genes and possibly determine its mode of action, we first assessed whether the regulatory activity of MvfR might occur at a post-transcriptional level. As presented in Fig. 1A, the expression of lasI′–lacZ and rhlI′–lacZ translational fusions is the same in PA14 wild-type and mvfR mutant backgrounds, suggesting that AHL synthesis is normal in the mvfR mutant. To further confirm that mvfR does not modulate QS-regulated genes by affecting lasRI and rhlRI gene expression and resulting activity, we performed LC/MS analysis for oxo-C₁₂-HSL and C₄-HSL in culture supernatants. As shown in Fig. 1B, the total normalized concentrations of these AHLs are not significantly different between cultures of PA14 and the mvfR mutant. Similar results are obtained using the Chromobacterium violaceum CV26 bioassay (data not shown).

Presence of HAQs, including PQS, does not overcome the absence of mvfR or pqsE

We have recently shown that MvfR controls the synthesis of HAQs, a large family of extracellular compounds that includes HHQ, the precursor of PQS (Déziel et al., 2004). To assess whether any of these compounds, which comprise signalling molecules, actually mediate the regulatory activity of MvfR, we investigated the effect of the three principal members of the various congener series of HAQs on the transcription of the phz1 operon, a target of the MvfR regulatory system involved in the synthesis of pyocyanin (Déziel et al., 2004; and Table 1). HHQ, its hydroxylated product PQS or its N-oxide derivative HQNO all fail to activate transcription from the phzA1 promoter in mvfR and pqsE mutant backgrounds (Fig. 2). These results indicate that the function of MvfR is not limited to the control of HAQ synthesis and, furthermore, that pqsE is essential for the activity of the MvfR regulatory system.

Identification of genes expressed differently in an mvfR mutant versus wild-type PA14

Transcriptome studies were therefore carried out to identify genes controlled by the transcriptional regulator MvfR, in order to gain insight into the function of this regulator and its role in the QS circuitry and P. aeruginosa virulence. Using the P. aeruginosa GeneChip^®, whole-genome expression profiles were obtained for the wild-type PA14 strain and the isogenic mvfR mutant, at three time points corresponding to different bacterial growth stages from two independent experiments. Using the criteria outlined in Experimental procedures, we identified 141 genes as being differentially regulated, with 121 showing higher expression in wild-type PA14 compared with mvfR mutant, and 22 showing higher expression in the mvfR mutant relative to PA14 at least in one time point (Table 1). The signal values for these genes and the whole data set can be viewed using a GuiGraph template (see Experimental procedures) available as supplementary material (TableS1).

The K-means and GeneTree methods were used to cluster the differentially transcribed genes. In general, both clustering methods give results that agree well with each other, and most known co-regulated genes and transcriptional units cluster together. These analyses reveal the sequential activation of genes involved in the production of HAQs, phenazines, and other toxic and virulence-related factors, most of which are also known to be regulated by QS (Fig. 3). The salient features of these clustering analyses are summarized below.

Interestingly, rsmA, which encodes a QS system regulator, and lasR, which encodes a QS regulator function, are differentially transcribed at one of the time points. Also, mvfR itself exhibits lower transcription in the mvfR mutant background (Table 1).

Within this analysis, no further generalization can be made regarding the remaining genes that are not present in previously reported QS transcriptome lists (results described below). Most of them display a differential expression at only one of the sampling times and/or encode hypothetical/unknown functions. Some of these genes may be related to QS or represent secondary effect derived from the differential expression of primary responsive genes.

Relationship between genes affected by MvfR and P. aeruginosa genes identified in previous genome-scale studies

We compared our list of differentially expressed genes with several previously reported large-scale studies of P. aeruginosa: a genome-wide transposon mutagenesis to identify genes involved in QS by reporter gene construct activation (Whiteley et al., 1999); a microarray analysis of genes differentially expressed in biofilm (Whiteley et al., 2001); GeneChip^® analyses of genes differentially expressed in mucoid strains (Firoved and Deretic, 2003), under iron starvation (Ochsner et al., 2002), in las/rhl QS regulatory mutants (Hentzer et al., 2003; Schuster et al., 2003; Wagner et al., 2003), in a vqsR mutant (Juhas et al., 2004), under conditions that induce type III secretion systems (Wolfgang et al., 2003); and proteome analyses of magnesium starvation (Guina et al., 2003) and QS (Arevalo-Ferro et al., 2003).

As seen in Table 1 and Fig.S1, a large overlap exists between the genes differentially regulated by MvfR and previously reported QS transcriptomes (Hentzer et al., 2003; Schuster et al., 2003; Wagner et al., 2003). While this overlap is incomplete, the data sets in these three studies also show very significant areas of disagreement between each other (Fig.S1) (Hentzer et al., 2003; Vasil, 2003). Comparison of transcription profiling with the recently reported virulence and QS regulator VqsR (Juhas et al., 2004) shows a limited number of genes shared with the MvfR-modulated genes that are also QS regulated (22 in common out of 66 for mvfR and 57 for vqsR). As the inactivation of vqsR inhibits the production of AHLs (Juhas et al., 2004), it appears to modulate QS-regulated genes through a different route from the MvfR system. Beyond the differences in experimental designs and data analysis methodologies, use of a different strain from PAO1 in the latter study and our own might also have contributed to the observed differences. Nevertheless, we find a strong correlation between QS-controlled genes displaying reduced expression in the mvfR mutant and rhl-dependent genes requiring both autoinducers for full induction. These genes include the hcnABC and phzA1-G1 operons, and phzS, which were originally described as class III/IV QS-controlled transcriptional units (Whiteley et al., 1999). Other previously reported rhl-dependent QS-controlled genes/operons include lecA and lecB, rhlAB, chiC, PA2066-PA2069 and phzM (Winzer et al., 2000; Hentzer et al., 2003; Medina et al., 2003a), all of which belong to the mvfR transcriptome. Many of these QS-controlled genes encode virulence factors whose production had previously been shown to be reduced in mvfR mutants, such as pyocyanin, hydrogen cyanide (HCN) and LecA lectin (Rahme et al., 1997; Gallagher et al., 2002; Diggle et al., 2003). As MvfR controls the synthesis of PQS (Cao et al., 2001; Gallagher et al., 2002), this corroborates the recent observation that PQS is involved for the expression of primarily rhl-regulated genes (Diggle et al., 2003).

Although we had previously reported that lasR and rhlR mutations do not affect the activity of a mvfR–lacZ reporter (Cao et al., 2001), recent transcriptome studies have shown that LasR actually induces the transcription of mvfR in very early growth stages (Hentzer et al., 2003; Schuster et al., 2003). We recently resolved this inconsistency by the absence in our original lacZ construct of a critical las box found unusually high upstream in the mvfR promoter region (G. Xiao and L. G. Rahme, unpubl. data). Accordingly, the production of HAQs/PQS is reduced in lasR, but not rhlR, mutants (Pesci et al., 1999; Gallagher et al., 2002; Déziel et al., 2004). These observations collectively indicate that the MvfR system upregulates a subset of QS genes requiring both LasR and RhlR, while the las system is required for the expression of mvfR.

Finally, no significant overlaps with genes induced by mucoidy, genes preferentially upregulated in cells growing as biofilms, or genes differentially expressed during conditions inducing the type III secretion system, are observed. As discussed by the authors, there are major differences when comparing QS proteome with microarray data (Arevalo-Ferro et al., 2003). Therefore, it is not surprising that we see very limited correlation with our list, as only five QS-regulated proteins correspond to MvfR-modulated genes.

Gene fusion reporter validation of GeneChip^® data

To confirm the transcriptomic results for genes not previously reported to be under MvfR (or PQS) regulation, we performed reporter expression assays. Using lacZ fusions, we show that transcription from the promoter region of mexGHI-opmD is almost abolished in the mvfR mutant (Fig. 5), in agreement with the microarray results (Fig. 4A). Furthermore, there is no activity in rhlR and pqsE mutant backgrounds, and delayed expression in a lasR mutant. Among the primary extracellular products whose synthesis depends on the rhl QS system are rhamnolipids (Ochsner et al., 1994; Pearson et al., 1997). A rhlA′–lacZ fusion reveals a delayed expression in the mvfR mutant (Fig. 6A), also manifested by a slightly postponed halo formation on blue rhamnolipid indicator plates (data not shown), in agreement with the presence of rhlB in Table 1 and previous observations (Diggle et al., 2003). We also investigated the expression of two regulators found in Table 1. Expression of the rsmA′–lacZ translational reporter is strongly reduced in both mvfR and pqsE mutant backgrounds (Fig. 6B), while mvfR seems to slightly autoregulate its own transcription (Fig. 6C). That the activity of the rsmA′–lacZ translational reporter is nearly abolished suggests an additional post-transcriptional level of regulation of this gene, as the transcription of rsmA is only slightly attenuated (Table 1). RsmA, an RNA-binding protein, acts as a post-transcriptional negative regulator of extracellular products by effecting message decay (Pessi et al., 2001). Whereas rsmA inactivation had only minor effects on elastolytic and staphylolytic activities as well as HCN, lectin and pyocyanin production, overexpression of this gene resulted in severe downregulation of these QS-regulated factors (Pessi et al., 2001). Interestingly, it has not been reported to be QS-regulated in any of the previous whole-genome profiling studies. More work is required to elucidate how the MvfR system might regulate RsmA activity and to which extent the mvfR regulon is affected by this regulator.

Reduced pathogenicity of pqs operon mutants, but not of a rhlR mutant

lasR, mvfR and pqsB mutants display reduced pathogenicity in plants, nematodes, insects and mice (Rahme et al., 1997; Mahajan-Miklos et al., 1999; Tan et al., 1999; Jander et al., 2000; Cao et al., 2001). Accordingly, we find that pqsA and pqsE mutants also display attenuated virulence in mice with mortality rates similar to the mvfR mutant (Fig. 7), thus confirming the importance of these mvfR-dependant genes in mammalian pathogenesis. Importantly and unexpectedly, an isogenic rhlR mutant does not display a reduced virulence phenotype under the same experimental conditions (Fig. 7). These data indicate that the impaired virulence of the mvfR mutant in this model is mediated by factors regulated through the MvfR system and not overlapping with the rhl QS regulon.

MvfR is part of the QS system but is not involved in AHL-mediated gene regulation

The most striking feature of the mvfR transcriptome is the over-representation of QS-controlled genes (Table 1). Collectively, about 55% of the MvfR-dependent genes are regulated by QS in PAO1, whereas QS-controlled genes represent no more than 6–7% of the PAO1 genome (Schuster et al., 2003; Wagner et al., 2003), supporting our assertion that the MvfR system serves as a modulator of QS-regulated genes in P. aeruginosa (Cao et al., 2001). The moderate differences on transcription levels we see for a number of these QS-influenced genes possibly reflect the fact that these genes are under the control of more than one system, i.e. mvfR system and QS.

While the maximal expression of mvfR requires LasR, and also itself (Fig. 6C), several arguments indicate that the mechanism used by the MvfR system to regulate many QS-controlled genes does not operate via lasRI and rhlRI. First, while MvfR regulates the expression of genes requiring both LasR and RhlR, it does not alter the expression of these regulators (Cao et al., 2001) or, as we show in this study using a very precise LC/MS-based method of analysis, the extracellular levels of the C₄-HSL and oxo-C₁₂-HSL autoinducers. Our results agree with recent observations that mvfR inactivation and loss of PQS production actually do not alter rhlI transcription (Gallagher et al., 2002) and C₄-HSL production (Diggle et al., 2003). Second, the mvfR mutation nearly abolishes the expression of many RhlR-dependent genes (e.g. phz1), suggesting that they require both rhlR and mvfR, while other RhlR-dependant genes seem less or not dependant on MvfR activity. For instance, we see a limited effect of MvfR on rhlAB expression (Table 1 and Fig. 6A) and rhamnolipid production, while the expression of this operon and rhamnolipid production are nearly abolished in a rhl mutant background (Ochsner et al., 1994; Pearson et al., 1997).

Results from this study and earlier reports (Pesci et al., 1999; McKnight et al., 2000) indicate that LasR controls HAQ and PQS synthesis on at least two separate levels. First, mvfR is only maximally transcribed in the presence of LasR and second, LasR controls pqsH (Whiteley et al., 1999), which is required for PQS synthesis (Gallagher et al., 2002; Déziel et al., 2004). Nevertheless, that the MvfR system does not operate through modulation of the las and rhl AHL-based QS systems implies an additional, independent level of regulation for many QS-controlled genes.

Reduced expression of mexGHI-opmD and other efflux pump-related genes

Besides the co-regulated operons involved in HAQ synthesis and response (pqsABCDE, phnAB) (Déziel et al., 2004), the other group of early differentially expressed genes in the mvfR mutant consist of the PA4205-PA4208 transcriptional unit, which corresponds to the mexGHI-opmD operon (Aendekerk et al., 2002) of unknown function. The expression of this operon is upregulated by both oxo-C₁₂-HSL and C₄-HSL in lasI/rhlI mutants (Whiteley et al., 1999; Schuster et al., 2003; Wagner et al., 2003). Inactivation of this putative efflux pump in PAO1 was reported to result in the decreased production of multiple virulence factors regulated by QS (Aendekerk et al., 2002). We show here that this operon is completely dependent on pqsE and rhlR, while its expression is only delayed in a lasR mutant (Fig. 5). Such a postponed response of lasR mutants has been observed previously for the production of pyocyanin (Diggle et al., 2002) and PQS (Diggle et al., 2003). As suggested by the microarray and lacZ reporter results, some delayed and very low transcription seems to occur in the mvfR mutant (Fig. 5). This might explain why we fail to observe any difference in extracellular AHL concentrations between PA14 and the mvfR mutant, whereas mexGHI-opmD was previously postulated to be involved in AHL homeostasis (Aendekerk et al., 2002). Further studies should be designed to address the role of this operon in the MvfR system.

Other interesting observations that enhance the findings on the mexGHI-opmD operon is that four putative transcriptional regulators of the TetR (tetracycline repressor) family, PA2885, PA3678, PA3721 and PA3973, display lower expression levels in the mvfR mutant. TetR family members are typically autoregulated repressors that bind specific ligand molecules. PA3678 encodes MexL,  a  transcriptional  repressor  of  the  adjacent mexJK-encoded efflux pump (Chuanchuen et al., 2002), and PA3721 is disrupted in nalC-type mutants of P. aeruginosa that hyperexpress the MexAB-OprM efflux pump (Cao et al., 2004). At least seven efflux pumps have been identified in P. aeruginosa, MexAB-OprM, MexCD-OprJ, MexEF-OprN, MexGHI-OpmD, MexJK-OprM, MexVW-OprM and MexXY-OprM (Poole and Srikumar, 2001; Li et al., 2003; Schweizer, 2003). The intended natural substrates for these pumps are unlikely to be only anti-microbial/toxic compounds as they are expressed by both antibiotic-susceptible and -resistant bacteria (Poole and Srikumar, 2001; Schweizer, 2003). The other two putative TetR regulators exhibiting lower expression in the mvfR mutant have no known function and are not located close to efflux pump genes. Although we might expect that a number of efflux pumps would be derepressed in the mvfR mutant, this is not reflected in the transcriptome data, highlighting the complexity of the global regulatory circuits of efflux pumps in P. aeruginosa.

The MvfR system controls the virulence of P. aeruginosa independently of the rhl QS system

Besides the predicted involvement of QS in pathogenesis via the regulation of virulence factors (Van Delden and Iglewski, 1998; de Kievit and Iglewski, 2000), experimental evidences obtained in various animal models demonstrate the implication of the QS system in the pathogenesis of P. aeruginosa (Tang et al., 1996; Rumbaugh et al., 1999; Tan et al., 1999; Pearson et al., 2000; Wu et al., 2001; Smith et al., 2002; Lesprit et al. 2003). Remarkably, however, our studies show that a rhlR mutant is not attenuated in the burn mouse model whereas mvfR (Cao et al., 2001), pqsA and pqsE mutants all display a similarly reduced virulence phenotype (Fig. 7). Earlier studies reporting decreased virulence were conducted with lasR, lasI, rhlI or lasIrhlI mutants, and there are no reports using a rhlR mutant. Although one might expect rhlR and rhlI mutants to display similar behaviours, a recent study has proposed that RhlR might actually act as a repressor in the absence of its cognate autoinducer C₄-HSL (Medina et al., 2003b). Hence, it is likely the expression of genes co-regulated by the rhl system and the MvfR system would be inhibited in a rhlI mutant but would simply fail to be upregulated in a rhlR mutant. Alternative explanations include compensatory secondary mutations that could arise in rhlR mutant in vivo as has been reported for rhlI and lasI mutants in vitro (Beatson et al., 2002).

Our results support a model where a subset of genes regulated by the MvfR system but not overlapping with the rhl regulon is important in P. aeruginosa pathogenesis.

pqsE is essential for the regulatory activity of the MvfR system and the full virulence of P. aeruginosa

MvfR controls the transcription of two co-regulated operons, pqsABCDE and phnAB, the products of which, with the exception of pqsE, mediate the synthesis of a large array of HAQs, including HHQ, the direct precursor of PQS (Déziel et al., 2004). Intriguingly, a pqsE mutant is not defective in the production of HAQs or PQS (Gallagher et al., 2002; Déziel et al., 2004), but still is deficient in production of pyocyanin, lectin and HCN, and expression of the corresponding synthetic operons phz1, lecA and hcnABC (Gallagher et al., 2002; Diggle et al., 2003; and data not shown). We are now adding mexGHI-opmD to the list of pqsE-dependent genes (Fig. 5). Although the above-mentioned factors are also controlled by the rhl system, production of C₄-HSL and expression of rhlI and rhlR are not affected in a pqsE mutant, as it is for a mvfR mutant (Gallagher et al., 2002; Diggle et al., 2003; and data not shown). Basically, apart from the production of HAQs, a pqsE mutant seems to display the same phenotypes as a mvfR mutant.

Pseudomonas quinolone signal or HQNO addition to cultures of any pqs or mvfR mutants, or induction of PQS synthesis by HHQ addition, all fail to complement the absence of pqsE. Although the production of HAQs (including PQS) is unaffected in the pqsE mutant, the pathogenic fitness of this mutant is as reduced as the mvfR and pqsA/pqsB mutants (Fig. 7). Genomic context and blast analysis suggest that pqsE probably encodes an hydrolase. Further investigations to elucidate how pqsE mediates the function of the MvfR system are warranted.

While PQS and other HAQs appear to act as intercellular signals (Déziel et al., 2004), and might be involved in the activity of PqsE (Diggle et al., 2003), their ultimate function is still unresolved. Nevertheless, here we demonstrate that the loss of PqsE function, and not only PQS production, is responsible for the phenotypes of lasR or mvfR mutants previously solely attributed to PQS production deficiency.

Strains, plasmids and growth conditions

Table 2 lists strains and plasmids. Bacteria were routinely grown in Luria–Bertani (LB) broth at 37°C in a roller drum or on a gyratory shaker at 250 r.p.m. LB plates contained 1.5% agar.

Isogenic knockout mutants were generated by allelic exchange using sucrose counterselection (Schweizer, 1992; Alm and Mattick, 1996), using described constructs (Beatson et al., 2002).

Standard methods were used to manipulate DNA. The parent plasmid pQF50 was used to construct the mvfR–lacZ and mexG–lacZ transcriptional fusion reporters pGX1 and pED1 respectively. To ensure that the entire regulatory region is included, 746 bp upstream to the 160 bp downstream of the mvfR translational initiation site and 586 bp upstream to the 119 bp downstream of the mexG translational initiation site were amplified. The promoter fragments were generated from genomic DNA using polymerase chain reaction with primers 5′-GGATCGGTACCAGTCGCTACACCTGAAGGC-3′ and 5′-CTAGGAAGCTTCCCGACGGACCAGCTCCAC-3′ for mvfR and with 5′-GAAGATCTTCGCACTACGGAGCCAGA GC-3′ and 5′-GGGGTACCCCTGGCCTGATAGTCGAACA-3′ for mexG. The resulting fragments were digested with KpnI and HindIII for mvfR and with BglII and KpnI for mexG, then inserted into pQF50 digested with the corresponding enzymes. The intended subcloning event was confirmed by DNA sequencing.

Plasmids were introduced by electroporation (Smith and Iglewski, 1989). For P. aeruginosa, carbenicillin (300 µg ml⁻¹), tetracycline (100 µg ml⁻¹), gentamicin (100 µg ml⁻¹) and X-Gal (40 g ml⁻¹) were included when required.

β-Galactosidase activity assays

Pseudomonas aeruginosa strains harbouring lacZ fusions were typically grown overnight with the appropriate antibiotic, then subcultured without antibiotic in LB to a starting OD₆₀₀ of 0.05. Culture samples were obtained at regular intervals for determination of growth (OD₆₀₀) and β-galactosidase-specific activity (Miller, 1972). Experiments were performed in triplicate.

For the HAQ addition experiment, PQS and HHQ were synthesized as described (Lépine et al., 2003) and HQNO was obtained from Sigma-Aldrich (Oakville, Canada). Stocks were prepared in methanol.

RNA isolation and generation of transcriptomic data

Pseudomonas aeruginosa PA14 and the isogenic mvfR mutant were grown in 1 l Erlenmeyer flasks containing 100 ml of LB incubated at 37°C and 200 r.p.m. Culture samples were harvested at three growth stages (OD₆₀₀ = 1.5, 2.5 and 3.5) and the RNA was immediately stabilized with RNAprotect Bacteria Reagent (Qiagen, Valencia, CA). Samples were stored at −80°C. Total RNA was isolated with the RNeasy spin column (including an on-column DNase digestion step) according to the manufacturer (Qiagen), treated with RQ1 DNase I (Promega, Madison, WI) for 1 h at 37°C, and repurified through an RNeasy column. Samples were labelled according to the manufacturer (Affymetrix, CA) and hybridized to the Affymetrix GeneChip^®P. aeruginosa Genome array for 24 h at 50°C using the GeneChip^® hybridization oven at 60 r.p.m. Chip washing, staining and scanning was performed according to instructions from Affymetrix. Experiments were performed in duplicate for independently performed experiments.

Analysis of GeneChip^® expression data

The data files of scans of the PAO1 GeneChip^® arrays hybridized with different probes were converted to cell intensity files (.CEL files) with the Microarray suite 5.0 (MAS, Affymetrix, CA). The data were scaled to target intensity of 500, and for each time point all possible pair-wise array comparisons of the replicates were performed (i.e. four combinations) with mutant array as baseline, using a change call algorithm of MAS 5.0. Genes that had (i) change calls in the same direction in each experiments and (ii) change calls of same direction in three of four comparisons, with additional constraints that in each comparison a signal value difference greater than 100, non-zero signal log ratio and one of the two samples being compared was not called ‘Absent’, were scored as differentially modulated between the mutant and wild-type bacteria. The ratio of differential expression is the average of all four ratios. For genes showing higher expression in mvfR mutant (i.e. decreased in the wild type), the average of the inverse of the ratios was calculated, and scored as the negative change value.

GeneTree analysis (Eisen et al., 1998) of signal values (averages of duplicate experiments) normalized to median values per gene across all time points in log ratio mode using Pearson correlation was performed using Gene Spring (version 5.1, Silicon Genetics, CA). Ten clusters K-means clustering (Hartigan and Wong, 1979) was also performed on the log ratio of signal values normalized to median for each gene across all time points using Pearson correlation, and they were used to colour tree branches.

Transcriptional units in Table 1 were based on predicted open reading frames (ORFs) from http://www.cifn.unam.mx/moreno/pub/TUpredictions/Predictions/ (Salgado et al., 2000; Moreno-Hagelsieb and Collado-Vides, 2002) as of 20 May 2003. One operon, PA3478-PA3479, was added to the list and minor errors were corrected. The list of transcriptional units used and the template for selecting the genes satisfying the above-mentioned rules are available at http://genetics.mgh.harvard.edu/RahmeWeb/home.htm

GuiGraph: a tool for visualization of gene expression data

In many instances gene expression data can be better perceived by looking at both the signal values and the fold change. In order to facilitate visualization for large data sets, we have developed GuiGraph, a tool to aid interpretation of transcriptome data. The GuiGraph version adapted for the Affymetrix P. aeruginosa GeneChip is available at http://www.genetics.mgh.harvard.edu/RahmeWeb/

AHL and HAQ quantification

An estimate of AHL production was obtained using the C. violaceum CV26 bioassay (McClean et al., 1997). Quantification was performed by LC/MS analysis as follows. Aliquots of culture samples were taken at regular intervals, and diluted with an equivalent volume of methanol containing 2% of acetic acid and 1000 mg l⁻¹ tetradeutero-PQS, as internal standard. After centrifugation, 20 µl of the supernatants were injected for LC separation onto an Agilent HP1100 (Agilent Canada, Montreal, Canada) equipped with a 150 × 3 mm C8 Luna (Phenomenex, Torrance, CA) reverse phase column, using a water (A)/acetonitrile (B) gradient both containing 1% acetic acid. A 30% to 80% B gradient over 20 min was followed by 100% B for two additional minute. The HPLC flow rate was 400 µl, split to 10% with a Valco Tee. The mass spectrometer was a Micromass Quattro II (Waters Canada, Montreal, Canada) triple quadrupole operating in positive electrospray ionization mode. Data acquisition was performed in full scan mode with a scanning range of 130–350 Da. Precise quantification of C₄-HSL (Sigma-Aldrich Canada, Oakville, Canada) and oxo-C₁₂-HSL (Quorum Science, Coralville, IA) was performed in MS/MS mode by monitoring the intensity of daughter ions produced by collisional activation of their pseudomolecular ions at m/z 172 and 298 respectively. Argon was the collision gas at 2 × 10⁻³ mTorr, and the collision energy was 13 eV. HAQs, including PQS, were also measured using LC/MS (Lépine et al., 2003).

Mice mortality studies

Animal pathogenicity was assessed using the thermal injury mouse model (Stevens et al., 1994) as described (Rahme et al., 1995). Animal protocols were reviewed and approved by the Institutional Animal Care and Use Committee of the Massachusetts General Hospital.