RhlR-Regulated Acyl-Homoserine Lactone Quorum Sensing in a Cystic Fibrosis Isolate of Pseudomonas aeruginosa.

The opportunistic pathogen Pseudomonas aeruginosa is a leading cause of airway infection in cystic fibrosis (CF) patients. P. aeruginosa employs several hierarchically arranged and interconnected quorum sensing (QS) regulatory circuits to produce a battery of virulence factors such as elastase, phenazines, and rhamnolipids. The QS transcription factor LasR sits atop this hierarchy and activates the transcription of dozens of genes, including that encoding the QS regulator RhlR. Paradoxically, inactivating lasR mutations are frequently observed in isolates from CF patients with chronic P. aeruginosa infections. In contrast, mutations in rhlR are rare. We have recently shown that in CF isolates, the QS circuitry is often rewired such that RhlR acts in a LasR-independent manner. To begin understanding how QS activity differs in this rewired background, we characterized QS activation and RhlR-regulated gene expression in P. aeruginosa E90, a LasR-null, RhlR-active chronic infection isolate. In this isolate, RhlR activates the expression of 53 genes in response to increasing cell density. The genes regulated by RhlR include several that encode virulence factors. Some, but not all, of these genes are present in the QS regulon described in the well-studied laboratory strain PAO1. We also demonstrate that E90 produces virulence factors at similar concentrations as PAO1, and in E90, RhlR plays a significant role in mediating cytotoxicity in a three-dimensional lung epithelium cell model. These data illuminate a rewired LasR-independent RhlR regulon in chronic infection isolates and suggest further investigation of RhlR as a possible target for therapeutic development in chronic infections.IMPORTANCE Pseudomonas aeruginosa is a prominent cystic fibrosis (CF) pathogen that uses quorum sensing (QS) to regulate virulence. In laboratory strains, the key QS regulator is LasR. Many isolates from patients with chronic CF infections appear to use an alternate QS circuitry in which another transcriptional regulator, RhlR, mediates QS. We show that a LasR-null CF clinical isolate engages in QS through RhlR and remains capable of inducing cell death in an in vivo-like lung epithelium cell model. Our findings support the notion that LasR-null clinical isolates can engage in RhlR QS and highlight the centrality of RhlR in chronic P. aeruginosa infections.

In laboratory strains of P. aeruginosa, deletion or deleterious mutation of lasR results in attenuated virulence in various animal models of infection (10,11). Despite the importance of LasR in regulating virulence, several studies have shown that lasR mutations are commonly observed in isolates collected from the lungs of chronically infected CF patients (12)(13)(14). In some patients, the frequency of isolates with a mutant lasR has been reported to be greater than 50% (13,15). Although a few singlenucleotide substitutions in lasR still yield functional protein, most mutations in either the signal-or DNA-binding domains yield a nonfunctional polypeptide (16). These findings led to the notion that QS is not essential during chronic stages of infection, dampening enthusiasm for QS inhibitors as potential therapeutics. Contrary to this idea, we and others have shown that many LasR-null P. aeruginosa chronic infection isolates remain capable of engaging in QS activity through the RhlI/RhlR circuit (16)(17)(18). Given the plasticity of the QS hierarchy, there are likely several mechanisms through which LasR-null clinical isolates can maintain RhlR activity. Recently, Kostylev et al. proposed that RhlR-active clinical isolates may emerge from a LasR-null background via mutation of the gene encoding the transcription factor MexT (19). Although it is apparent that CF strains "rewire" their QS circuitry so that RhlR is the key transcription factor, the RhlR regulon in a rewired background had not yet been described.
We are interested in the regulatory remodeling of QS that occurs in isolates of P. aeruginosa from chronic infections, including those in CF. To begin to understand how RhlR-mediated QS in clinical isolates might be different from that in laboratory strains (15,16,18,20), we studied a CF isolate called E90 (21), which contains a single-base-pair deletion in lasR at base 170, and uses RhlR to mediate QS. E90 produces QS-regulated virulence factors at levels comparable to those of PAO1. We used transcriptome sequencing (RNA-seq) to analyze the RhlR regulon of this isolate by comparing its transcriptome with that of an isogenic RhlR deletion mutant. We determined that the E90 RhlR regulon consists of more than 83 genes, including those that encode virulence factors. Using a three-dimensional tissue culture model, we also observed that E90 induces cell death in an RhlR-dependent manner. Together our data provide a much broader picture of the rewiring of QS that can take place in CF-adapted P. aeruginosa while also providing a basis for elucidating RhlR-specific gene regulation without the confounding effects of the QS hierarchy.
synthase and is strongly activated by LasR (23). Green fluorescent protein (GFP) fluorescence in E90 transformed with this reporter plasmid was unmeasurable and mirrored that of a PAO1 ΔlasR mutant (Fig. 1A). As a complementary approach, we measured the concentration of 3OC12-HSL produced by E90 using a bioassay. We found that E90 after overnight growth produced 40 nM 3OC12-HSL, a very small amount compared to that by PAO1 (1.5 M) and in contrast to that by PAO1 ΔlasR, for which no 3OC12-HSL was detected (Fig. 1B). Together, these data suggested that E90 produces a small amount of 3OC12-HSL in a LasR-independent manner but that the 3OC12-HSL was not important for activation of QS-regulated genes. In contrast, E90 produced approximately 8.3 M C4-HSL after overnight growth, comparable to what we measured for PAO1 (9.8 M). Altogether, these data confirmed that E90 encodes a nonfunctional LasR, and suggested that if QS was active in this isolate, it was regulated by either RhlR, PqsR, or both transcription factors.
To determine if E90 retained AHL-dependent QS, we examined the expression of several well-studied quorum-regulated genes in the presence or absence of AiiA lactonase, an enzyme that degrades AHL signals (24). Using reverse transcriptionquantitative PCR (qRT-PCR), we observed that expression of lasB and rhlA were increased in the presence of AHLs (Fig. 2). These genes, which encode the exoprotease elastase and a rhamnosyltransferase involved in rhamnolipid production, were identified as QS regulated in PAO1 (3,25). The gene rhlI, which encodes the C4-HSL synthase, was also AHL regulated (Fig. 2). Next, we asked if the small amount of 3OC12-HSL produced by E90 (Fig. 1B) has a role in QS. Using a transcriptional reporter assay, we observed that expression from the lasB and rhlA promoters in the E90 ΔlasI background was similar to that of the wild type, demonstrating that 3OC12-HSL does not contribute to the QS activity observed in E90 (see Fig. S1 in the supplemental material).
Previous work has shown that RhlR activity can be uncoupled from LasR regulation in LasR-null backgrounds (19,26,27). The rhlR gene in E90 encodes a polypeptide that is identical to that of PAO1, although it contains three synonymous mutations, and there is no difference in the rhlR promoter at the Vfr binding sites or the putative Las box. Given that C4-HSL production is robust in E90 (Fig. 1B), we queried if this strain similarly engaged in RhlR-dependent QS activity. To address this question, we engineered an rhlR deletion in the E90 background to observe its effect on quorumregulated phenotypes. We found that the E90 ΔrhlR deletion mutant, like an rhlI mutant, exhibited undetectable rhlA promoter activity and produced little to no exoprotease and pyocyanin, consistent with the idea that RhlR regulates QS activity in E90 (Fig. 3). The phenotype was complemented in the rhlI mutant by addition of C4-HSL. We also verified that differences in exoprotease and pyocyanin production FIG 2 In isolate E90, expression of several canonical QS-regulated genes is AHL dependent. The following target genes were measured in the presence or absence of AiiA lactonase using qRT-PCR: lasI, 3OC12-HSL signal synthase; lasB, elastase; rhlI, C4-HSL signal synthase; rhlR, RhlR; pqsA, coenzyme A ligase involved in Pseudomonas quinolone signal synthesis; chiC, chitinase; aprA, alkaline metalloprotease. The differences in threshold cycle (ΔC T ) are measured relative to the housekeeping gene rplU. Fold changes in gene expression (table on right) are reported relative to cultures incubated with AiiA lactonase. *, P Ͻ 0.05 by t test. Error bars represent the standard deviations for results from three independent experiments.

FIG 3
RhlR regulates QS in E90. Production of pyocyanin (A) or protease (B) in either PAO1, E90, PAO1 ΔlasR, E90 ΔrhlR, E90 ΔrhlI, or E90 ΔrhlI with 10 M C4-HSL. The dashed line indicates the detection limit for the pyocyanin assay, which is 0.2 g/ml. The detection limit for the protease assay is approximately 0.004 g/ml. (C) rhlA-gfp reporter activity over time. Data from the first 5 h were excluded, because cell density measurements were below the limit of detection of the plate reader. Error bars represent the standard deviations for results from three independent experiments. In some cases, error bars are too small to be seen. Both the PAO1 ΔlasR mutant and E90 produced concentrations of pyocyanin or exoprotease that were significantly different from those of PAO1. ****, P Ͻ 0.0001 by t test.
observed between PAO1, E90, and PAO1-or E90-derived strains were not due to differences in growth under the conditions used for this assay (see Fig. S2). As a whole, our results showed that the LasR-null isolate E90 retains QS activity in an RhlR-and C4-HSL-dependent manner, and suggested that regulation by RhlR in this strain parallels that of LasR in PAO1. Because RhlR-dependent QS regulation appears to be common in CF isolates (16)(17)(18), we reasoned that a study of the genes regulated by RhlR in this background would give insight into which QS-regulated gene products might be important in chronic CF infections. Furthermore, because rhlR is not regulated by LasR in E90 (and other clinical isolates), a study of the E90 QS transcriptome has the potential to disentangle genes that are regulated solely by RhlR from those that require both LasR and RhlR.
Identification of the RhlR regulon of E90. To determine which genes are regulated by RhlR, we performed an RNA-seq-based differential gene expression (DE) analysis comparing RNA collected from cultures of the parent strain E90 to that from the isogenic RhlR deletion mutant. First, we sought to generate a de novo-assembled genome for E90 to use as an RNA-seq mapping reference which would account for the potential genomic differences between E90 and reference strains of P. aeruginosa. Using a hybrid approach combining both short-and long-read high-throughput sequencing, we were able to assemble the genome of E90 into a single circular contig of approximately 6.8 Mb that harbors 6,650 annotated features ( Fig. 4A) (6,650 features total, 6,503 protein-coding sequences). In addition to being roughly 550 kb larger than the published sequence of laboratory strain PAO1 (28), the genome of E90 includes 862 features with no homology to PAO1. Also present is a 4.4-Mb inversion relative to PAO1, which includes an internal reorder of roughly 250 kb. The inversion appears to be the result of a recombination event between two roughly 5-kb repeat regions that do not have homology to PAO1 but flank the rrnA-rrnB region previously implicated in restructuring of the P. aeruginosa genome (28). A brief search of the E90 genome for P. aeruginosa genes previously reported to be under purifying selection in CF isolates revealed a nonsynonymous mutation in the gene coding for the probable oxidoreductase MexS (locus PAE90_2949/PA2491; nonsynonymous single nucleotide polymorphism [SNP]), as well as the resistance-nodulation-division multidrug efflux membrane fusion protein precursor MexA (locus PAE90_0464/PA0425; 33-bp deletion) (13).
Next, to facilitate a comparison to previously published studies of QS regulons that also implemented an RNA-seq-based approach (29), we grew strains in LBmorpholinepropanesulfonic acid (MOPS) to an optical density at 600 nm (OD 600 ) of 2.0. The growth of E90 and E90 ΔrhlR is indistinguishable in this medium under the conditions used for this experiment (see Fig. S3). Our DE analysis identified 53 genes that were upregulated in the E90 versus E90 ΔrhlR comparison (Tables 1 and S1). Forty-four (83%) of these genes were identified as QS regulated in a previous microarray study of PAO1 (3) (Fig. 4B), and 21 belong to the core quorum-controlled genome characterized in reference 29.
We also identified several well-known virulence genes, including those that encode biosynthetic machinery required for rhamnolipid (rhlAB), hydrogen cyanide (HCN; hcnABC), elastase (lasB), and pyocyanin (phzABC1) synthesis. Elastase is an exoprotease known to degrade various components of the innate and adaptive immune system, including surfactant proteins A and D (30,31). Rhamnolipid and pyocyanin have also been previously appreciated for their roles in airway epithelium infiltration and damage (32,33). In addition, our RNA-seq analysis revealed hsiA2, the first gene in the cluster encoding the second type VI secretion system, which facilitates the uptake of P. aeruginosa by lung epithelial cells (34).
While QS control of the phenazine biosynthesis pathway was reported previously, only one of the two "redundant" operons ("phz1;" phzA1-G1) was identified in our study (3). Interestingly, our transcriptome analysis found that RhlR also regulates the first two genes of the second phenazine operon ("phz2;" phzA2-G2) in E90, albeit at a slightly lower level than phz1. Both operons encode nearly identical sets of proteins, each with the capacity to synthesize the precursor (phenazine-1-carboxylic acid) of many downstream phenazine derivatives, including the virulence factor pyocyanin (35). Despite their seemingly redundant function, phz1 and phz2 do not appear to be regulated in concert. In strain PA14, although phz1 is more highly expressed than phz2 in liquid culture, similar to what we observed in the E90 RhlR regulon, phz2 actually contributes more to overall phenazine production in liquid culture (36). Furthermore, phz2 is the only active phz operon in colony biofilms and was the only phz operon implicated in lung colonization in a murine model of infection (36).
Moreover, we observed that the RhlR regulon included genes that likely confer a growth advantage in the CF lung. For example, cbpD encodes a chitin-binding protein shown to contribute to the thickness of biofilms, the development of which is impor- This circular representation of the E90 genome includes rings indicating the following features described from the outer-most to inner-most rings: annotated features of CDS (blue) or rRNA genes (gray) on the forward (outer) or reverse (inner) strands; all-by-all blastn hits (red) in a comparison against PAO1_107 (nucleotide identity Ͼ40%); GC content deviation (black); GC skew (ϩ, green; Ϫ, purple). Additional outer partial rings indicate the 4.4-Mb inversion (bright red) and the 250-kb reorder (light blue). (B) Venn diagram highlighting genes shared between the E90 RhlR-induced regulon and the PAO1 QS-induced regulon as determined via microarray in reference 3. Lobes are scaled to approximate the relative size of each regulon. tant for nutrient acquisition and stress resistance (37). The gene encoding the monodechloroaminopyrrolnitrin 3-halogenase PrnC was also present in the E90 RhlR regulon, which has not been reported in previous P. aeruginosa transcriptomes and is not present in the PAO1 reference genome. Halogenase PrnC was only previously described in Pseudomonas protegens (formerly Pseudomonas fluorescens), where it is involved in the synthesis of pyrrolnitrin, an antifungal antibiotic (38).
Among the most highly regulated genes (3,29) were those belonging to a conserved nonribosomal peptide synthetase (NRPS) pathway (PAE90_1770 to -1779; PA3327 to -3336). The products of this NRPS pathway have been identified as azetidinecontaining alkaloids referred to as azetidomonamides (39). The biological significance of this widely conserved NRPS pathway in Pseudomonas species or what roles azetidomonamides may play in virulence or interspecies interaction is not well understood, but regulation by QS appears to be a common feature.
Our interrogation of the E90 RhlR regulon also revealed 30 genes that were RhlR repressed; none of these genes were reported in previous reports of QS-repressed genes (3) and 19 are not present in the PAO1 genome. We found two genes of the alpBCDE lysis cassette, alpB and alpC, were repressed by RhlR in E90 under the conditions of our experiments. While induction of alpBCDE, via derepression of the alpA gene, has been shown to be lethal to individual cells, it may benefit infecting cells at the population level (40). We also observed downregulation of the gene encoding the posttranscriptional regulatory protein RsmA by RhlR in E90. RsmA is nested in a host of regulatory machinery important in infection, and mutation of RsmA has been observed to favor chronic persistence and increased inflammation in a murine model of lung infection (41). Lastly, we identified RhlR regulation of phage loci not found in the PAO1 genome. The RhlR-repressed phage loci correspond to E90 genes PAE90_2433 through PAE90_2442.
RhlR is the primary driver of cytotoxicity in a lung epithelium model. LasR-null laboratory strains are less virulent than the wild type (WT) in acute infection settings  (10,11,42). However, as the RhlR-dependent QS regulon of E90 includes several factors implicated in virulence (Table 1), we queried if E90 might be capable of inducing host cell death. To address this question, we incubated an in vivo-like three-dimensional (3D) lung epithelial cell culture model (A549 cell line) (43) with either PAO1, E90, or engineered QS transcription factor mutants. The 3D lung cell model possesses several advantages over the standard A549 monolayer as an infection model, including increased production of mucins, formation of tight junctions and polarity, decreased expression of carcinoma markers, and physiologically relevant cytokine expression and association of P. aeruginosa with the epithelial cells (43,44). Following an incubation period of 24 h, we measured cell death of the 3D cell cultures via cytosolic lactate dehydrogenase (LDH) release. Consistent with prior studies, WT PAO1 cytotoxicity was abrogated in a LasR deletion mutant; however, cytotoxicity of a PAO1 RhlR-null mutant was similar to that of the wild type, because in this assay, the secreted products responsible for cytotoxicity are LasR regulated in PAO1, with little or no contribution from RhlR. Strikingly, the opposite was true for E90: deletion of RhlR significantly reduced cytotoxicity. Addition of exogenous C4-HSL to an E90 ΔrhlI deletion mutant restored cytotoxicity to WT levels (Fig. 5). This RhlR-dependent cytotoxicity might be related to the different timing of RhlR activation in E90, the specific set of genes regulated by RhlR in this strain, or both. Together, these results highlight the restructuring of QS gene regulation in this clinical isolate and underscore implications for virulence during chronic infection.

DISCUSSION
A substantial body of literature now suggests that the QS hierarchy of P. aeruginosa is adaptable and that LasR mutants can be rewired to be AHL QS proficient (16,26,27,45). These rewired LasR-null clinical isolates retain the QS regulation of several exoproducts through the RhlI/RhlR circuit (16,17). Prior studies examining an RhlRdependent variant of PAO1 (19), and another LasR-null and RhlR-active clinical isolate (17), showed that the parent strain outcompetes RhlR-null derivatives when grown in coculture (17,19). These findings support the notion that there is something inherently disadvantageous about mutation of RhlR and point to RhlR as a key QS transcription factor in chronic infections such as CF.
We do not know the mechanism or genetic modifications that resulted in Lasindependent RhlR activity in isolate E90. In strain PAO1, in which the hierarchy of QS was initially described, LasR mutants can readily evolve an independent RhlR QS system through inactivating mutations of mexT, which encodes a non-QS transcriptional regulator (19,27). However, this is not the case in isolate E90, which possesses a functional mexT allele. We did observe that rhlI expression is upregulated by RhlR in E90 unlike in PAO1, where rhlI expression is predominately LasR regulated (46). These data suggest that in E90 RhlR and RhlI may constitute a positive autoregulatory loop that may facilitate Las-independent RhlR activity. We are interested in investigating alternate mechanisms, other than inactivation of mexT, through which RhlR escapes LasR regulation in these rewired backgrounds.
In the present study, we aimed to identify which genes comprise the RhlR regulon in a clinical isolate, which may shed light on factors important for establishment or continuation of a chronic infection. Our RNA-seq analysis revealed that the E90 RhlR regulon bears a substantial amount of overlap with the suite of AHL-regulated genes previously identified in PAO1 (3) and consists of virulence factors that are likely advantageous in the context of the CF lung.
A portion of the genes found to be RhlR regulated in our transcriptomic analysis of E90 were not previously reported to be QS regulated. These genes include vqsR (PAE90_2723; PA2591) and two genes of the phz2 operon (PAE90_3614 and -3615; PA1899 and -1900). VqsR is itself a LuxR homolog that serves to augment QS gene regulation, possibly through activation of the orphan QS receptor QscR, although the precise mechanism and biological outcomes of this interaction are still mysterious (47,48). Our finding that the phz2 operon, in addition to phz1, is activated by RhlR may reflect ongoing QS adaptation in our selected CF isolate. While E90 appears to produce slightly less bulk pyocyanin in broth culture than PAO1, pyocyanin production by E90 may be comparatively greater in the biofilm lifestyle of the CF lung. The phz2 locus, while showing roughly 98% nucleotide identity with phz1, has been shown to be responsible for nearly all the pyocyanin produced in biofilms by PAO1 and is the dominant contributor to murine lung colonization between the two loci (36). It is possible that some of these previously unreported QS-regulated genes were excluded from earlier transcriptome analyses (3,29) due to different analysis approaches or methodology. Of particular note, we compared an RhlR deletion mutant to the parent strain to derive our transcriptome, while some of these previous studies used signal synthase mutants with and without signal, which has been demonstrated, in the case of RhlR QS, to yield a different phenotype (49).
We also discovered RhlR-QS regulation of many genes that are not present in the PAO1 genome. This list includes several hypothetical proteins activated as much as 15-fold in E90 compared to that in the RhlR mutant. The list also includes the gene encoding the halogenase PrnC, a protein involved in production of the antifungal antibiotic pyrrolnitrin, which may be important in interspecific interactions in the CF lung (50). Our finding that RhlR-QS in E90 also appears to repress genes in the programmed cell death cassette alpBCDE points to additional potential for QS regulation of population-level interactions in CF-adapted strain E90.
Although we do not yet fully understand the biological significance of the RhlRmediated suppression of the phage identified in this study, we are interested in exploring its role, if any, in fitness and inter-and intraspecies competition in the near future. We note that had we used the PAO1 genome, as opposed to the E90 de novo genome, for read alignment, we would have failed to identify the phage loci and a few other genes. These findings therefore argue in favor of using de novo genomes to improve comprehensive transcriptome analyses of clinical and environmental isolates moving forward.
Strikingly, we found that in E90, RhlR but not LasR is the critical determinant of cytotoxicity in a three-dimensional lung epithelium cell aggregate model. Though our study did not reveal exactly which virulence factors are important for cell death in this model, our results nevertheless challenge the idea that LasR-null isolates are avirulent. Instead, our data argue that some virulence activity is conserved in rewired isolates, but that RhlR and not LasR is the primary regulator of several such functions.
The scope of our analysis is limited by our examination of a single clinical isolate and the use of laboratory growth conditions for RNA-seq analysis; but, our data provide a basis for understanding regulatory remodeling of QS activity and provide avenues for future investigation. Several important questions remain about QS in clinical isolates, including whether or not there is a "core" regulon that is common to isolates that use either LasR or RhlR as the primary QS transcription factor. Our work also serves as a starting point to test hypotheses regarding the role of RhlR-regulated genes during chronic infection, the possible fitness advantage associated with LasR-independent RhlR activity, and mediators of sustainable chronic infections. In summary, our work sheds light on RhlR-specific gene regulation and reveals the potential breadth of QS activity and virulence functions retained in LasR-null CF-adapted isolates.

MATERIALS AND METHODS
Bacterial strains and growth conditions. Bacterial strains and plasmids used in this study are described in Table S3 in the supplemental material. E90 is part of a collection of clinical isolates obtained via the Early Pseudomonas Infection Control observational (EPIC Obs) study (21). The isolates are from oropharyngeal and sputum samples from 5-to 12-year-old patients. Further details regarding the EPIC Obs study design and results were described previously (21,51).
For the transcriptional reporter assays as well as pyocyanin and AHL measurements, overnight cultures were started from single colonies grown in 3 ml of Luria-Bertani (LB) broth buffered with 50 mM morpholinepropanesulfonic acid (MOPS) in an 18-mm culture tube. For the cytotoxicity experiments, overnight cultures were started from single colonies in 5 ml of unbuffered LB broth. When appropriate, antibiotics were added at the following concentrations: 10 g/ml gentamicin or 100 g/ml ampicillin for Escherichia coli, and 100 g/ml gentamicin for P. aeruginosa. Cells were grown at 37°C with shaking at 250 rpm unless stated otherwise. For complementation experiments, synthetic C4-HSL (Cayman Chemical) was added to E90 ΔrhlI deletion mutant cultures to a final concentration of 10 M.
LasR and RhlR activity. LasR-and RhlR-specific promoter fusions constructed in pPROBE-GT were described previously (16) and are listed in Table S3. Electrocompetent P. aeruginosa cells were prepared through repeated washing and resuspension of cell pellets in 300 mM sucrose (52). Transformants were obtained by plating on LB agar supplemented with gentamicin and verified by PCR.
Experimental cultures were prepared as follows: first, overnight cultures were grown with the addition of 100 g/ml gentamicin and 100 g/ml AiiA lactonase, the latter inhibiting AHL-mediated QS (24). The addition of AiiA lactonase eliminates residual GFP fluorescence that would otherwise arise from previously induced reporter gene expression during overnight growth. Overnight cultures were then diluted to an optical density at 600 nm (OD 600 ; 1-cm pathlength) of 0.001 (approximately 1 ϫ 10 6 to 5 ϫ 10 6 CFU/ml) in 3 ml MOPS-buffered LB supplemented with AiiA lactonase in 18-mm culture tubes. After these cultures grew to an approximate OD 600 of 0.2, they were diluted to an OD 600 of 0.001 in 400 l of MOPS-buffered LB alone in a 48-well plate with a clear bottom (Greiner Bio-One). To prevent evaporation, strains were only grown in wells that did not line the edges of the plate and all empty wells were filled with 400 l water. We monitored GFP fluorescence and OD 600 at 30-min intervals for 15 h using a BioTek Synergy HI microplate reader (excitation, 489 nm; emission, 520 nm; gain, 80). All strains were grown at 37°C with shaking for the duration of the assay. To account for differences in growth, results were normalized to OD 600 values. As a negative control, each strain was electroporated with an empty vector, which was used to establish a baseline level of background fluorescence. The fluorescence intensity was calculated by subtracting the background fluorescence from the total fluorescence measured at every time point. All experiments were performed in biological triplicates.
Construction of the E90 ⌬rhlR mutant. A homologous recombination approach was used to generate an in-frame deletion mutant (53,54). Fragments flanking rhlR were PCR amplified from E90 genomic DNA and cloned into pEXG2 to yield pEXG2.E90 ΔrhlR, which was then transformed into E. coli S17-1 in order to facilitate conjugal transfer of pEXG2.E90 ΔrhlR into E90. Transconjugants were selected by plating on Pseudomonas isolation agar supplemented with gentamicin, and deletion mutants were counterselected by plating onto LB agar with 10% (wt/vol) sucrose. Deletion of rhlR was confirmed by PCR and targeted sequencing.
AHL signal extraction and measurement. Experimental cultures were prepared from overnight cultures diluted to an OD 600 of 0.001 in 3 ml of MOPS-buffered LB in an 18-mm culture tube. Experimental cultures were grown with shaking until they reached an OD 600 of 2.0. Then, AHL signals were extracted from experimental cultures using acidified ethyl acetate as described elsewhere (55). We used an E. coli DH5␣ strain containing either pJN105L and pSC11 in conjunction with the Tropix Galacto-Light chemiluminescent assay (Applied Biosystems) to measure 3OC12-HSL or containing pECP65.1 to measure C4-HSL (25,56,57). The bioassay strains and plasmids are listed in Table S3.
Protease and pyocyanin measurements. Experimental cultures were prepared from overnight cultures by diluting to an OD 600 of 0.001 in 3 ml MOPS-buffered LB in 18-mm culture tubes. For secreted protease measurements, experimental cultures were grown with shaking for 18 h. Then, cells were pelleted, and 100 l of filtered supernatant was collected to measure protease production using fluorescein isothiocyanate (FITC)-casein for the Pierce fluorescent protease assay kit (Thermo Fisher Scientific). For pyocyanin measurements, experimental cultures were grown with shaking for 18 h, and pyocyanin was extracted from cultures as described previously (16). We grew strains in MOPS-buffered LB to remain consistent with the growth conditions used for RNA-seq analysis.
Cytotoxicity of three-dimensional A549 cell cultures. A three-dimensional lung epithelial cell culture model was generated by culturing A549 cells (ATCC CCL-185) on porous microcarrier beads in a rotating well vessel (RWV) bioreactor system, as described previously (43). A549 cells were grown in GTSF-2 medium (GE Healthcare) supplemented with 2.5 mg/liter insulin transferrin selenite (ITS) (Sigma-Aldrich), 1.5 g/liter sodium bicarbonate, and 10% heat-inactivated fetal bovine serum (FBS) (Invitrogen) and incubated at 37°C under 5% CO 2 , Ͼ80% humidity conditions. Infection studies were performed on cultures grown for 11 to 14 days in the RWV. Thereafter, 3D cell cultures were equally distributed in a 48-well plate at a concentration of 2.5 ϫ 10 5 cells/well (250-l volume), and infected with the different strains at a targeted multiplicity of infection of 30:1 as described previously (44). All infection studies were performed in the above-described cell culture medium, with the exception that no FBS was added given the interference of serum compounds with QS signaling (58). After 24 h of infection, the release of cytosolic lactate dehydrogenase (LDH) from 3D lung epithelial cell cultures was determined using an LDH activity assay kit (Sigma-Aldrich) according to the manufacturer's instructions. A standard curve using NADH was included. The positive control (theoretical 100% LDH release) was obtained by lysing 2.5 ϫ 10 5 cells with 1% Triton X-100. All LDH release values were expressed as a percentage of the positive control.
RNA isolation and qRT-PCR. Overnight cultures were started from single colonies grown in 3 ml of MOPS-buffered LB in 18-mm culture tubes. Experimental cultures were prepared by diluting overnight cultures to an OD 600 of 0.01 in 25 ml of MOPS-buffered LB in 125-ml baffled flasks. Experimental cultures were grown at 37°C with shaking at 250 rpm. Approximately 1 ϫ 10 9 cells were pelleted at an OD 600 of 2.0 and mixed with RNA Protect Bacteria reagent (Qiagen) before being stored at Ϫ80°C. Thawed cell pellets were resuspended in QIAzol reagent and mechanically lysed by bead beating. To extract RNA, we used the RNeasy kit (Qiagen) according to manufacturer's instructions. Isolated RNA was then treated with Turbo DNase (Ambion) and purified using the MinElute cleanup kit (Qiagen). Three biological replicates were processed for each strain (E90 and E90 ΔrhlR). Next, cDNA was prepared using the iScript cDNA Synthesis kit (Bio-Rad). Then, expression of target genes was analyzed by following the protocol for the iQ SYBR green SuperMix (Bio-Rad) on a CFX96 real-time PCR cycler for a total of 40 cycles. We analyzed expression of the following genes: lasI, lasB, rhlI, rhlR, rhlA, pqsA, chiC, aprA, vqsR (PAE90_2723), PAE90_0133, PAE90_0837, and PAE90_2705. We used rplU as a reference gene. Primers used for qRT-PCR are listed in Table S3.
For RNA-seq experiments, cultures were prepared and RNA was extracted and purified as described above for qRT-PCR with 2 biological replicates per treatment. Genewiz, LLC, performed rRNA depletion, library generation, and sequencing for all samples. RNA reads were obtained using the Illumina HiSeq platform with an average of 15.3 million 150-bp paired-end raw reads per sample, which were then groomed using Trim Galore (v0.4.3; https://github.com/FelixKrueger/TrimGalore). Reads were then aligned against the E90 genome and counted using the Subread/featureCounts suite of command line tools to produce a final count matrix of 4 by 6,478, which was then loaded into the R statistical environment (66,67). Differential expression (DE) analysis was performed using DESeq2 using a fold change cutoff of 2 and an adjusted P value of 0.05 (68).
Data availability. The E90 genome as well as raw reads and count matrix associated with this transcriptome analysis have been deposited in the NCBI Sequence Read Archive under BioProject accession PRJNA559863.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.