A Quadruple Knockout of lasIR and rhlIR of Pseudomonas aeruginosa PAO1 That Retains Wild-Type Twitching Motility Has Equivalent Infectivity and Persistence to PAO1 in a Mouse Model of Lung Infection

It has been widely reported that quorum-sensing incapable strains of Pseudomonas aeruginosa are less virulent than wild type strains. However, quorum sensing mutants of P. aeruginosa have been shown to develop other spontaneous mutations under prolonged culture conditions, and one of the phenotypes of P. aeruginosa that is frequently affected by this phenomenon is type IV pili-dependent motility, referred to as twitching motility. As twitching motility has been reported to be important for adhesion and colonisation, we aimed to generate a quorum-sensing knockout for which the heritage was recorded and the virulence factor production in areas unrelated to quorum sensing was known to be intact. We created a lasIRrhlIR quadruple knockout in PAO1 using a published technique that allows for the deletion of antibiotic resistance cartridges following mutagenesis, to create an unmarked QS knockout of PAO1, thereby avoiding the need for use of antibiotics in culturing, which can have subtle effects on bacterial phenotype. We phenotyped this mutant demonstrating that it produced reduced levels of protease and elastase, barely detectable levels of pyoverdin and undetectable levels of the quorum sensing signal molecules N-3-oxododecanoly-L-homoserine lactone and N-butyryl homoserine lactone, but retained full twitching motility. We then used a mouse model of acute lung infection with P. aeruginosa to demonstrate that the lasIRrhlIR knockout strain showed equal persistence to wild type parental PAO1, induced equal or greater neutrophil infiltration to the lungs, and induced similar levels of expression of inflammatory cytokines in the lungs and similar antibody responses, both in terms of magnitude and isotype. Our results suggest, in contrast to previous reports, that lack of quorum sensing alone does not significantly affect the immunogenicity, infectiveness and persistence of P. aeruginosa in a mouse model of acute lung infection.


Introduction
The role of quorum sensing (QS) and biofilm formation in Pseudomonas aeruginosa biology is clear, and previous reports have suggested that knocking out QS systems in P. aeruginosa results in decreased virulence in animal models of infection [1,2,3,4,5] and has impacts on bacterial growth rate in culture. The deletion of both the synthase genes and the responder genes for both 3-oxo-N-dodecanoyl-L-homoserine lactone (3OC12HSL) and the other principal homoserine lactone QS signal produced by P. aeruginosa, N-butyryl-L-homoserine lactone (C4HSL) should result in a quadruple knockout that will not respond to or produce 3OC12HSL or C4HSL. Previously reported lasI/rhlI knockouts including PAO1-JP2 have been reported to be less virulent than wild type P. aeruginosa PAO1 in a range of animal models [3,4,5,6,7,8,9,10,11]. However, QS mutants of P. aeruginosa have been shown to develop spontaneous mutations under prolonged culture conditions [12,13], and one of the phenotypes of P. aeruginosa that is frequently affected by this phenomenon is type IV pili-dependent motility, referred to as twitching motility [12]. In fact, twitching motility was originally reported to be controlled by QS [14] but was later confirmed to be independently regulated [15]. Interestingly, JP-2 has been reported to be incapable of twitching motility [14] which suggests that this strain is likely to contain secondary mutations in a gene (or genes) required for twitching motility. It has been shown that, unlike the PAO1 parental strain, a PAO1 mutant with defective pili is unable to bind to respiratory epithelial cells and cannot induce production of the proinflammatory and neutrophil-recruiting chemokine interleukin (IL)-8 [16]. It has also been reported in many studies that type-IV pili are integral for adherence to and colonization of mucosal surfaces (reviewed in Hahn et al. [17]). It is therefore important that twitching motility remains intact in any QS knockout P. aeruginosa used for in vivo infection studies assessing the role of QS in bacterial infectivity and persistence.
Because of the inherent and spontaneous mutations that occur during standard laboratory subculturing of P. aeruginosa, we created a lasIRrhlIR knockout in a strain for which the heritage was recorded and the virulence factor production in areas unrelated to QS was known to be intact. We used a technique described by Hoang et al. [18], which allows for the deletion of antibiotic resistance cartridges following mutagenesis, to create an unmarked QS knockout of PAO1. Such a knockout strain has two benefits: firstly an unmarked knockout can be further genetically modified without the need for multiple resistance cassettes and secondly, it has been found that subinhibitory concentrations of antibiotics can profoundly influence the phenotype of P. aeruginosa [19,20,21]. A knockout strain that does not require antibiotic selection for maintenance of mutation would allow the comparison of QS knockout strains to isogenic wild-type bacteria without any possibility of phenotypic changes being generated in the bacteria because of the presence of antibiotics in the culture medium.
After creation of the quadruple lasIRrhlIR knockout, we tested it in a mouse model of P. aeruginosa lung infection to compare the course of infection with that of the parental PAO1. We found that the knockout showed equal persistence, induced comparable or greater inflammation, and resulted in identical antibody responses in terms of magnitude and immunoglobulin isotype to parental PAO1.

Bacterial Strains, Plasmids and Media
The wild type P. aeruginosa used as the basis for this study was PAO1 strain ATCC 15692. E. coli strain DH5a was used in all genetic manipulations and in the preparation of DNA sequencing templates, and E. coli S17-1 was used as the donor strain in bacterial conjugation for allelic exchange mutagenesis. All bacterial strains used in this study are listed in Table 1.
P. aeruginosa and E. coli were cultured in LB-Lennox broth (LB) or cation-adjusted Mueller Hinton broth (CAMHB) or on LB solidified with 1.5% agar (LBA). Antibiotic concentrations used for selection of E. coli were 100 mg/mL ampicillin, 12.5 mg/mL tetracycline, and 50 mg/mL kanamycin and for P. aeruginosa were 250 mg/mL carbenicillin, and 200 mg/mL tetracycline.

Construction of PAO1DlasIR, PAO1DrhlIR, and PAO1DlasIRrhlIR
Unmarked deletion mutants of PAO1 were constructed using the Flp-FRT recombination system for site specific excision of chromosomal sequences described [18]. Briefly, 1 kb sections that flanked the lasIR and rhlIR region(s) to be deleted from PAO1 chromosomal DNA were PCR amplified and cloned into pGEMTeasy and sequenced to ensure that no mutations in these flanking regions had been introduced by PCR. The upstream and downstream flanking regions of the las or rhl regions were ligated and then cloned into pOK12 (creating pJL005 and pJL006 respectively) and the FRT/Tet R cassette from pCBW108 cloned into the internal SacI site resulting in pOK12 constructs pJL007 and pJL008 containing the flanking regions of lasIR or rhlIR regions separated by the FRT::Tet R cassette. The resultant clones were then digested with SpeI and cloned into the suicide vector pRIC380. This vector carries the genes sacBR, which promote sensitivity to sucrose, and oriT which enables conjugal transfer.
The resultant clones were transformed into the E. coli donor strain S17-1 in preparation for mating with P. aeruginosa PAO1. Following conjugation, the transconjugants were plated onto LBA without sodium chloride and supplemented with 5% sucrose and containing tetracycline to select for colonies in which the plasmid had excised while leaving the homologously recombined lasIR::Tet R or rhlIR::Tet R alleles in the chromosome. The Tet R gene was then excised using the pFLP2 plasmid that expresses the Flp recombinase as described previously [18] creating P. aeruginosa strains with the lasIR or rhlIR regions deleted and replaced with an FRT sequence. Allelic exchange deletion mutants were confirmed by both PCR and Southern hybridization of isolated chromosomal DNA. A list of all plasmids used in this study is given in Table 1 and all primers used are listed in Table 2.

Assays of Bacterial Phenotypes
Growth curves. P. aeruginosa strains were cultured overnight in CAMHB at 37uC shaking at 250 r.p.m. Quadruplicate 200 mL aliquots of 1:100 dilutions of overnight cultures were added to wells of a sterile 96 well plate (BD Biosciences, Mountainview, CA), which was then incubated for 10 h at 37uC shaking in a BIO-TEKH Synergy TM HT microplate luminometer (BIO-TEK, Winooski, VT, USA). Optical density readings (605 nm) were taken every 20 min and subtracted from the blank (CAMHB only).
Extraction of acylated homoserine lactones (AHLs) from culture supernatant and AHL bioluminescence assay. The supernatant of 8 h cultures of P. aeruginosa strains cultured in LB at 37uC shaking at 250 r.p.m. was collected by centrifuging the culture at 30006 g for 10 min. The supernatant was sterilised though a 0.22 mm filter and AHLs extracted using a method adapted from that described by Rice et al. [22]. Equal volumes of culture supernatant and 0.01% glacial acetic acid in ethyl acetate were mixed, centrifuged at 100006 g for 2 min and the organic layer containing the AHL retained. This was repeated at least three times and the collected organic phase dried overnight at room temperature. The sediment was resuspended in 50 mL of ethyl acetate.
3OC12HSL and C4HSL were quantified using bioluminescence reporter strains as described previously [23,24]. Briefly, 20 mL of culture supernatant or AHL standards were added to the wells of a white opaque microtiter plate (PerkinElmer, Wellesley, MA, USA), and 180 mL of mid-log phase cultures of E. coli containing pSB1073 or pSB406 was added to the microtiter plate, gently agitated and incubated at room temperature for 1 h, after which the luminescence was measured using a Veritas Microplate Luminometer (Veritas, Sunnyvale, CA) or a BIO-TEKH Synergy TM HT microplate luminometer (BIO-TEK). The concentration of AHL in culture supernatants was determined using dilutions of synthetic AHL as a standard curve. The results were standardized to the expression from a vehicle control (the limit of detection for the assay), and because the interassay variation was large, the results are presented as a percentage of the wild-type.
Twitching motility assay. Twitching motility was assayed using the subsurface stab assay described previously [25]. Briefly, the P. aeruginosa strain to be tested was stab inoculated through a plate of LB solidified with 1% agar to the underlying Petri dish and incubated at 37uC with saturated humidity. After 20 h at 37uC the diameter of the interstitial biofilm was measured.
Rhamnolipid synthesis assay. Triplicate 5 mL aliquots of overnight culture were applied to a dry rhamnolipid assay agar (prepared as described by Kohler et al. [26]) and allowed to dry. The plate was incubated at 37uC for 18 h then chilled overnight at 4uC. The plates were then examined for the presence or absence of a zone of clearance. Skim milk protease assay. This method was adapted from that described by Sokol et al. [27]. In brief, three 10 mL droplets of overnight broth culture were applied to a dry skim milk agar plate and grown for 18 h at 37uC. The zone of clearance was measured with the diameter of the colony subtracted.
Elastase assay. The determination of elastolytic activity of 16 h P. aeruginosa cultures was performed using a protocol based on those described by Ohman et al. [28] and Rust et al. [29]. In brief, P. aeruginosa strains were cultured in CAMHB and then centrifuged at 30006 g for 10 min. The supernatant was then filter sterilized using a 0.22 mm syringe filter and either used immediately or stored at 4uC for up to 24 h. In triplicate, 10 mg of elastin Congo red (ECR), 500 mL of ECR buffer (0.1 M TrisCl and 10 mM CaCl 2 ) and either 500 mL of the sterile spent supernatants or 500 mL of a CAMHB negative control were added to 5 mL polystyrene tubes (BD FalconH, BD Biosciences) and incubated at 37uC for 6 h with 250 r.p.m. shaking. The incubation was stopped by the addition of 100 mL of 0.12 M Na 2 EDTA and the remaining solids were pelleted by centrifugation at 30006 g for 10 min. The soluble Congo red in the aqueous supernatant was quantitated using duplicates of each supernatant in a 96-well microtiter plate (BD Biosciences) with the CAMHB negative control as a blank and measuring the absorbance at 495 nm in a Biomek spectrophotometer (Beckman Coulter, La Brea, CA, USA).
Pyocyanin extraction and assay. The pyocyanin production from 16 h cultures of P. aeruginosa was determined by first extracting the pyocyanin from the culture supernatant as described by Hassett et al. [30]. To assay pyocyanin, duplicate 100 mL aliquots of the HCl extraction were added to microtiter wells and the absorbance at 520 nm was measured in a Biomek spectrophotometer (Beckman Coulter).

Mouse Lung Infection Models
Ethics statement. All mouse studies were conducted in accord with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes, and were approved by the Animal Experimentation and Ethics committees of CQUniversity (approval No. A09/02-242) and University of New South Wales (approval Nos. 04/105A and 03/10). Animals were anesthetized for infection as described below, and were euthanized for tissue collection. All experiments were designed to use the minimum number of animals consistent with statistical validity, and all animals were monitored daily and euthanized if signs of distress were detected.
Acute infection model. For studies of acute infection and inflammation, groups of five male BALB/c mice 8-12 weeks old (ARC, Perth, WA, Australia) were inoculated intratracheally at day 0 with 1610 6 colony forming units CFU) of PAO1 or PAO1DlasIRrhlIR in 20 mL PBS containing 0.03% ethanol. A group of mice were also sham treated (anesthetized and cannula inserted, but no treatment given). Mice were sedated with a tail vein injection of Alfaxan at 12 mg/kg body weight (90 mg alphaxalone, 30 mg alphadolone acetate; Schering-Plough Animal Health, Baulkham Hills, NSW, Australia), and treatment administered by means of a 20 G paediatric cannula (Terumo Medical, Macquarie Park, NSW, Australia) inserted in the trachea. Mice from all treatment groups were then sacrificed at 4 h, 24 h or 10 days after infection, and lung tissue and BAL collected for analysis. For analysis of antibody responses, groups of five mice were given 1610 3 CFU of PAO1 or PAO1DlasIRrhlIR in a 0.5% agar slurry administered intratracheally by the method described above. These mice were bled by cardiac puncture under terminal anesthesia at day 21. Blood was collected in a 1.1 mL Z-gel tube (Sarstedt, Mawson Lakes, SA, Australia) that was then left at room temperature for at least 30 min before centrifuging at 20006 g for 10 min. The serum was collected and kept frozen at 220uC until use for evaluation of anti-Pseudomonas serum antibody concentrations.
Bronchoalveolar lavage (BAL). To obtain BAL, 0.5 mL of sterile PBS was slowly flushed into the lungs of euthanized mice, and as much as possible of the PBS was recovered. Differential cell counts were performed on cytospin preparations of the BAL fixed with DPX mountant (Scharlau, Barcelona, Spain) and stained using the Diff-Quik Staining Kit 64851 (a modified version of the Wright's stain; Veterinary Medical Surgical Supply, Maryville, NSW, Australia) according to the manufacturer's instructions.
Bacterial enumeration in the lung tissue and BAL. A section of the mouse lung was homogenized in 2 mL sterile PBS using a tissue homogenizer. The bacterial CFU in the BAL fluid and homogenized lung were determined by making serial dilutions and plating onto LBA.
Preparation of cDNA from mouse lung tissue. Total RNA was extracted from the lung tissue using TRIzolH Reagent (Invitrogen, Mulgrave, VIC, Australia) with a maximum of 100 mg of tissue per mL of TRIzolH Reagent. Tissue was removed from RNAlaterH storage solution and homogenized by bead dissociation in TRIzolH using 2.4 mm zirconia/silica beads (Daintree Scientific, St Helens, TAS, Australia). Beads and cellular debris were pelleted by centrifugation at 10 0006 g for 10 min at 4uC. Approximately 1 mL of supernatant was removed to a fresh tube and total RNA extracted as per the TRIzolH protocol. Isolated RNA was resuspended in 50 mL of DEPC-treated water (Ambion) and DNAse treated using a TURBO DNA-free kit TM (Ambion) according to the manufacturer's instructions. The concentration and purity of recovered RNA was quantified using a NanoDrop TM 1000 spectrophotometer (Thermo Scientific, Scoresby, VIC, Australia). cDNA was synthesized from 500 ng of total RNA using the Superscript TM III Reverse Transcriptase kit according to the manufacturer's instructions.
Quantitative reverse transcription real time PCR (RT-qPCR) for inflammatory markers in mouse lung. Primer sequences for reference gene b-actin [31], TNFa, IL1b (both [32]) and IL6 ( [33]) have been previously published. Primer sequences for mouse keratinocyte cytokine (KC) (analogue of human IL-8) (forward 59-GCTGGGATTCACCTCAAGAA-39, reverse 59-AGGTGCCATCAGAGCAGTCT-39) were designed from Gen-Bank sequences using Primer 3-web version 0.3.0 [34]. Primer pairs were checked for reaction efficiency using triplicate serial dilutions of template cDNA with efficiency calculated as described in Pfaffl, 2001 [35]. Quantitative PCR (qPCR) was performed using the PlatinumHSYBRHGreen qPCR Supermix-UDG (Invitrogen). Each qPCR reaction contained 5 mL of cDNA diluted 1:5-1:10 with DEPC-treated H 2 O, and final primer concentrations of 300 nM. The reactions were performed on a RotorGene 3000H (Corbett Life Science) with an initial incubation of 50uC for 10 min then 95uC for 10 min followed by 40 reaction cycles of 95uC for 15 sec, 60uC for 30 sec, and acquiring fluorescence at 72uC for 20 sec. All products underwent melt curve analysis.
Replicates showing errors were omitted from further analysis. The average of the technical replicates was used to determine the ratio of the target to the reference genes (b-actin and GAPDH) and corrected for the primer efficiency as described by Pfaffl [35]. Interassay variation was calculated by comparing the Ct values of a standard incorporated into all qPCR reactions. A coefficient of variation was calculated using PRISMH v.5 software (Graph Pad Inc. La Jolla, CA, USA) and values below 10% deemed to be acceptable.

Detection of anti-P. aeruginosa Antibodies by Enzymelinked Immunosorbent Assay (ELISA)
To prepare PAO1 sonicate to use as the capture antigen for the ELISA, PAO1 bacterial suspension was autoclaved for 15 min at 121uC and 15 kPa. The suspension was sonicated on ice three times with a Branson SonifierH S250D digital sonicator using a fine point probe (Branson Ultrasonics Corp., Danbury, CT, USA) before the cellular debris was pelleted at 4uC and 30006 g for 10 min. The supernatant was then aliquotted and stored at -80uC. A bicinchoninic acid assay was used to determine the protein concentration. The concentration of the unknown was calculated against a BSA standard curve.
Preparation of P. aeruginosa IgG and IgM positive and negative sera. Heat-killed bacteria were diluted to the equivalent of 2610 9 CFU/mL and 50 mL (1610 8 CFU equivalents) were injected intraperitoneally into male BALB/c mice. Anti-P. aeruginosa IgM-positive sera was obtained after 8 days, and anti-P. aeruginosa IgG-positive was obtained after 24 days with repeat injections on day 8 and day 16. Negative control serum was collected from unimmunised mice. The positive sera were arbitrarily defined as containing 1610 6 U/mL of anti-P. aeruginosa antibody.
Anti-P. aeruginosa IgG and IgM ELISA. Ninety-six well Polysorp microtiter plates (Nunc, Rochester, NY, USA) were coated with 100 mL/well of P. aeruginosa sonicate diluted to 10 mg/ mL in coating buffer (1.59 g/L Na 2 CO 3 and 2.94 g/L NaHCO 3 in diH 2 O pH 9.6). The plates were sealed and incubated overnight at 4uC. The plates were then washed three times in Dulbecco's modified phosphate buffered saline (DPBS) supplemented with 0.05% v/v Tween-20 then blocked with reagent diluent (DPBS with 5% skim milk powder, prepared on the day). All samples were prepared in reagent diluents and a standard curve was generated using the positive control serum.
After blocking, the plate was washed three times with washing buffer and then 100 mL/well of the samples added in duplicate to the plate, which was sealed and incubated at room temperature for 2 h. Following this incubation, the plate was washed with DPBS-Tween and 100 mL/well of the diluted detection antibodies added. The detection antibodies were diluted in reagent diluent to the following concentrations: biotinylated anti-mouse IgG1 heavy chain (1 mg/mL); biotinylated anti-mouse IgG2a heavy chain (1 mg/mL); and biotinylated anti-mouse IgM heavy chain (0.25 mg/mL) (all AbD Serotec, Oxford, UK) and allowed to incubate at room temperature for at least 10 min. The plate was sealed and incubated at room temperature for 2 h.
The plate was washed with DPBS-Tween and 100 mL/well of 0.2 mg/mL Streptavidin-HRP (Amersham Biosciences) diluted in 1% w/v BSA in DPBS was added, the plate sealed and incubated at room temperature for 1 h. After the incubation, the plate was washed with DPBS-Tween and 100 mL/well of 3,3,5,59-tetramethylbenzidine diluted to 110 mg/mL in citrate acetate (pH 9.6) supplemented with 0.2% v/v H 2 O 2 , and the colour left to develop for 5 min. The development was stopped by the addition of 100 mL/well of 0.16 M H 2 SO 4 . The absorbance was then read at 450 nm with a reference filter of 520 nm on a BioRad microplate reader and the software package Microplate Manager v 5.2.1 was used to calculate the standard curve. The least dilute sample with an absorbance reading that fit on the linear portion of the curve was used to calculate the antibody concentration of the sample.
Statistical analysis. All statistical analysis was performed using GraphPad Prism v 5. Distribution of all data was tested for normality using the Kolmogorov-Smirnov test. Data sets that were normally distributed are expressed as mean 6 standard deviation, and differences between groups were evaluated using ttest or ANOVA. Data sets (antibody titers) that were not normally distributed are presented as median with individual points. Differences between groups were evaluated using the Mann-Whitney U test.

Phenotypic Characterisation of PAO1DlasIR, PAO1DrhlIR and PAO1DlasIRrhlIR Deletion Mutants
To re-examine the role of quorum sensing in infection, we generated an unmarked deletion mutant of PAO1 that lacks both the las and rhl quorum sensing systems. We also further minimized any chance of other mutations arising by limiting its subculturing prior to characterization in various phenotypic assays and mouse models of infection. Because of the hierarchical nature of the las and rhl quorum sensing systems the PAO1DrhlIR mutant was created first as a lasIR knockout would also be deficient in many of the phenotypes that we used to confirm the rhlIR knockout.
An unmarked PAO1DrhlIR deletion mutant was generated and a number of phenotypic assays were performed to confirm that this strain displayed phenotypes consistent with previously published observations for mutants of rhlI and rhlR and to confirm that it had not acquired secondary mutations that result in defective twitching motility. As expected, our PAO1DrhlIR mutant produced reduced levels of elastase (Fig. 1a) and protease (Fig. 1b) relative to PAO1, did not produce rhamnolipid (Fig. 1e), produced barely detectable levels of pyocyanin (Fig. 1c), and retained wildtype twitching motility (Fig. 1d).
After verification of the PAO1DrhlIR mutant phenotypes, PAO1DlasIR and PAO1DlasIRrhlIR deletion mutants were created. These mutants were assayed to verify that they had phenotypes consistent with published observations and that they had not acquired secondary mutations that result in defective twitching motility. PAO1DlasIR showed reduced levels of elastase production whereas the PAO1DlasIRrhlIR quadruple knockouts produced barely detectable levels of elastase (Fig. 1a). Protease production by PAO1DlasIR was significantly reduced compared with the wildtype control, and was even lower in the quadruple knockout PAO1DlasIRrhlIR (Fig. 1b). PAO1DlasIR did not show a significant defect in pyocyanin production whereas PAO1DlasIRrhlIR produced barely detectable levels of pyocyanin (Fig. 1c). Importantly, both PAO1DlasIR and PAO1DlasIRrhlIR retained wild type levels of twitching motility (Fig. 1d).
All mutants were also tested to determine the levels of production of C4HSL or 3OC12HSL in the culture supernatant using bioluminescence reporters specific for C4HSL [23] and 3OC12HSL [24]. The concentration of AHLs produced was calculated using a standard curve of synthetic AHL, but because the interassay variation was large, the results are presented as a percentage of the wild-type (Fig. 2a, b). The rhlIR mutant produced no detectable C4HSL and higher levels of 3OC12HSL than wild type PAO1; the lasIR mutant produced similar levels of C4HSL to wild type and no detectable 3OC12HSL, and the lasIRrhlIR mutant produced no detectable C4HSL or 3OC12HSL.

Characteristics of Lung Infection in BALB/c Mice
To evaluate the effects of knocking out both lasIR and rhlIR on the course of an infection, groups of BALB/c mice were infected with either PAO1 or PAO1DlasIRrhlIR, and followed for 10 days. Mice were euthanized at 4 h, 24 h and 10 days after infection, and bacterial numbers, lung levels of inflammatory leukocytes, and  lung expression of mRNA for inflammatory cytokines TNFa, IL-1b and mKC (IL-8 analogue) were evaluated. The bacterial numbers recovered from lung tissue and BAL at each time point are shown in Fig. 3, and indicate that at each time point there were no significant differences between wild type PAO1 and the PAO1DlasIRrhlIR mutant, although there was a trend to recovery of higher numbers of PAO1DlasIRrhlIR than PAO1 at 4 h and 24 h. Low levels of infection with both strains were maintained out to 10 days.
We also investigated leukocyte infiltration into the lung by examining leukocytes in BAL at all time points (Fig. 4). At 4 h and 24 h, the infiltrate was dominated by neutrophils (80-90%), whereas at day 10, neutrophils constituted only about 10% of infiltrating cells, with monocytes and other cells (mostly lymphocytes) both representing approximately 45% of cells. The only significant differences between mice inoculated with PAO1 and PAO1DlasIRrhlIR were in total white cells and neutrophils at 4 h and 24 h post infection, where neutrophil numbers (and hence total white cell counts) were significantly higher in mice infected with PAO1DlasIRrhlIR (p,0.01, unpaired t test).
We used RT-qPCR to investigate the level of expression of three cytokines characteristic of acute inflammation, TNFa, IL1b and mKC, a mouse analogue of human IL-8, at 24 h post infection. The results, shown in Fig. 5, demonstrate that there was no significant difference in the level of expression of any of these cytokines between mice infected with PAO1 and those infected with PAO1DlasIRrhlIR.
Lastly, in a different set of experiments, mice were infected with PAO1 or PAO1DlasIRrhlIR embedded in agar slurry, euthanized at 22 days after initial infection and after a second dose of bacteria at day 21, and serum levels of P. aeruginosa-specific IgM, IgG1 and IgG2a measured by ELISA. Figure 6 shows that there was no significant difference in the titres of any of the isotypes between mice infected with PAO1 and those infected with PAO1DlasIRrh-lIR.

Discussion
The literature contains numerous examples of animal models of P. aeruginosa infection. In many of these, a comparison of a wild type (usually PAO1) and a variety of QS-deficient strains suggest that QS-controlled virulence factors are important in the pathogenesis of infection, and that most QS mutants are less virulent than wild type P. aeruginosa. It is unquestionable that QS controlled biofilm formation and virulence factor production plays an important role in bacterial virulence, particularly in clinical settings such as cystic fibrosis (CF) (e.g. [7]). However, because many QS mutants are known to acquire other mutations, particularly in twitching motility, which could affect infection and dissemination, we decided to create a rigorously designed quadruple knockout, and further minimize any chance of other mutations arising by limiting its culturing. Such a knockout would allow the separation of effects of bacterial adherence and infectivity from the effects of QS-controlled virulence.
Creation of the PAO1DlasIRrhlIR was a multi-step process. A PAO1DrhlIR mutant was created first because of the hierarchical nature of the las and rhl quorum sensing systems: a lasIR knockout would be deficient in many of the phenotypes used to confirm a rhlIR knockout, making the phenotypic confirmation of a rhlIR knockout on a PAO1DlasIR background more difficult. However, creating the rhlIR knockout first enabled us to confirm that its phenotype was as expected, with loss of pyocyanin production being the key phenotype change. The deletion of the lasI and lasR genes also resulted in the deletion of rsaL, but we predicted that an rsaL deletion would not affect the mutant significantly because its product is a lasI repressor, and this was supported by the phenotypic studies on the PAO1DlasIRrhlIR and PAO1DlasIR mutants generated. The characteristics of the PAO1DlasIRrhlIR knockouts clearly demonstrate that they are unable to produce a number of key QS-regulated virulence factors, and cannot respond to exogenous AHLs, but that they retain full twitching capability.
The results of the studies in the mouse model of lung infection suggest that there is no significant difference between the persistence of PAO1 and PAO1DlasIRrhlIR in mouse lungs, or in their ability to induce an inflammatory response in the host, although there is a suggestion from the neutrophil counts that PAO1DlasIRrhlIR actually induces a stronger inflammatory infiltrate than the parental PAO1, and the trend to higher bacterial numbers recovered in PAO1DlasIRrhlIR infected mice is consistent with the slightly faster growth rate of this strain (results not shown). Wu et al. also observed a stronger inflammatory response to a PAO1 lasIrhlI knockout strain (PAO1-JP2) in a rat model, but unlike in our study, this was accompanied by a more rapid clearance of bacteria [11]. However, as PAO1-JP2 is known to be deficient in twitching motility [14], this is likely to affect its ability to colonise lungs and hence its ability to persist. However, other studies in rats and mice have also suggested that quorum sensing-defective mutants are cleared more effectively than wild type bacteria [11,36]. For example, a study using a neonatal BALB/c model of infection reported that a PAO1DlasR mutant was virtually avirulent, causing no mortality, no evidence of lung pathology and only 15% of mice showed evidence of bacterial replication in vivo [37], although the authors found that PAO1-DlasR were still recovered in the lungs 24 h post infection in numbers similar to PAO1, indicating that although the bacteria could persist they did not cause pathology [37]. In our study, the bacterial numbers in PAO1DlasIRrhlIR-infected and wild type PAO1-infected mice were similar at each time point, with low numbers persisting out to at least 10 days after infection. Analysis of the cytokine mRNA in the lungs indicates that there is little difference in the ability of the two strains to stimulate effective proinflammatory cytokine responses, although analysis of the PMN recruitment to the lung indicates that PAO1DlasIRrhlIR is more effective at stimulating innate immunity.
An explanation for some of the differences between our results and those of others may be that, although PAO1 is the laboratory standard for P. aeruginosa, there is a great deal of genetic diversity between laboratories in the PAO1 strains used [13]. For example, twitching motility is one of the phenotypes commonly lost during ongoing laboratory culture, because of the development of point mutations in other regulators such as vfr [12]. For example, the PAO1 lasIrhlI knockout strain (PAO1-JP2) used in many studies has been reported to lack twitching motility [14] and this can reduce the ability of the bacteria to colonize mammalian epithelia. Thus, some of the reported reduced virulence of QS mutant P. aeruginosa may be the consequence of impaired colonization by twitching-deficient bacteria. Care was taken in the creation of the PAO1DlasIRrh-lIR mutant used in this study to ensure that such secondary mutations had not occurred, increasing confidence that the results from this study were in fact a reflection of the differences in quorum sensing ability of PAO1DlasIRrhIR and wild type PAO1, and not of other mutations in PAO1.
High antibody titres against P. aeruginosa have been associated with decreases in lung function and poorer outcomes of infection [38,39]. There were detectable titres of anti-P. aeruginosa IgM antibodies at 10 days post lung infection in the acute infection model, but no detectable IgG1 or IgG2a, and no difference between PAO1-infected mice and those infected with PAO1DlasIRrhIR (results not shown). To evaluate later immune responses, we generated a more chronic infection by giving mice P. aeruginosa in an agar slurry, which delays bacterial clearance, and collecting serum at day 22 after an acute bacterial challenge on day 21. P. aeruginosa-specific IgM, IgG1 and IgG2a were all easily detectable at this time point, but there was no difference in the titers of any of the isotypes between PAO1DlasIRrhlIR-infected and PAO1-infected mice. This is in contrast to a previous report using a rat model of infection suggesting that the IgG1 response was higher in rats infected with PAO1 compared with rats infected with a QS signal-deficient mutant (PAO1DlasIrhlI) [40]. As IgG1 is associated with prohumoral Th2 type responses in mice, this implied that QS-deficient mutants generated a more Th2-biased response, but we saw no such difference. The difference between our results and those reported by Wu et al. could be due to species variation, because of the differences in persistence between PAO1 and PAO1DlasIrhlI observed in Wu et al.'s study, where the mutant was cleared more effectively than the wild type, or could be related to differences in the bacterial strains other than those defined.
The mouse model used in this study was an acute lung infection, and our results do not exclude the likelihood that QS-mediated induction of biofilm formation and virulence factor production play important roles in chronic infections. It would be of interest to repeat these studies in models of chronic infection, particularly immune deficiency or cystic fibrosis, which in humans, results in increased susceptibility to P. aeruginosa infection. It remains entirely possible that QS mutants may be less virulent in such situations than wild type P. aeruginosa, for a combination of reasons related to both defects in biofilm formation and virulence factor production and to host factors and the characteristics of the inflammatory and immune responses to infection, which are known to be aberrant in people with cystic fibrosis (CF) [41]. In this context, although it is clear that P. aeruginosa exists in biofilms in CF lung [42,43] and the bacterial populations present in CF lung are heterogeneous and always include some QS-competent bacteria [42], it is relevant to note that it has been reported that QS-deficient mutants arise spontaneously after some years in people with CF with chronic P. aeruginosa lung infection [7]. The most frequently observed mutation in the genes of the QS systems is in lasŖ which could suggest that while QS systems are important for initiating and maintaining chronic infection, ''social cheating'' by subpopulations of QS-deficient P. aeruginosa could confer a survival advantage in the environment of the chronically-infected and damaged CF lung.
In conclusion, our results suggest that careful control of culture and propagation of QS deficient PAO1 derivatives to prevent additional mutations and the loss of such important characteristics as twitching motility may eliminate a significant proportion of the previously observed differences in infectivity and persistence of QS-deficient bacteria and wild-type P. aeruginosa in animal models, and allow a more precise evaluation of the importance of QS and QS-regulated gene expression in bacterial virulence and pathogenesis of infection. (QSneg, open bars) was processed, total RNA extracted and cDNA prepared. Expression of cytokine genes was assessed in RT-qPCR using b-actin as the reference gene. All results are expressed as gene expression relative to b-actin, and are presented as the mean 6 SD of five mice per group. A control group of lungs from mice that underwent anaesthesia and intratracheal infusion of PBS without bacteria was included (grey bars). Expression of all cytokines was significantly higher in mice infected with bacteria than in sham-infected mice (P,0.01, all groups, ANOVA) but there were no significant differences between mice infected with PAO1 or PAO1DlasIRrhlIR (QSneg). doi:10.1371/journal.pone.0060973.g005 Figure 6. Anti-Pseudomonas antibody production by mice infected with PAO1 or PAO1DlasIRrhlIR. Mice were infected intratracheally with 1610 3 CFU of PAO1 or PAO1DlasIRrhlIR in an 0.5% agar slurry, and bled on day 22. Sera were tested in ELISA against antigen prepared from whole killed PAO1, using isotype-specific detection antibodies for mouse IgM, IgG1, and IgG2a. Results are expressed as individual titers with a line representing the median values. There were no significant differences (Mann-Whitney U test) between mice infected with PAO1 or PAO1DlasIRrhlIR (QSneg). doi:10.1371/journal.pone.0060973.g006