Instantaneous Within-Patient Diversity of Pseudomonas aeruginosa Quorum-Sensing Populations from Cystic Fibrosis Lung Infections

Abstract

In the opportunistic pathogen Pseudomonas aeruginosa, acyl-homoserine lactone (acyl-HSL) quorum sensing (QS) regulates biofilm formation and expression of many extracellular virulence factors. Curiously, QS-deficient variants, often carrying mutations in the central QS regulator LasR, are frequently isolated from infections, particularly from cystic fibrosis (CF) lung infections. Very little is known about the proportion and diversity of these QS variants in individual infections. Such information is desirable to better understand the selective forces that drive the evolution of QS phenotypes, including social cheating and innate (nonsocial) benefits. To obtain insight into the instantaneous within-patient diversity of QS, we assayed a panel of 135 concurrent P. aeruginosa isolates from eight different adult CF patients (9 to 20 isolates per patient) for various QS-controlled phenotypes. Most patients contained complex mixtures of QS-proficient and -deficient isolates. Among all patients, deficiency in individual phenotypes ranged from 0 to about 90%. Acyl-HSL, sequencing, and complementation analyses of variants with global loss-of-function phenotypes revealed dependency upon the central QS circuitry genes lasR, lasI, and rhlI. Deficient and proficient isolates were clonally related, implying evolution from a common ancestor in vivo. Our results show that the diversity of QS types is high within and among patients, suggesting diverse selection pressures in the CF lung. A single selective mechanism, be it of a social or nonsocial nature, is unlikely to account for such heterogeneity. The observed diversity also shows that conclusions about the properties of P. aeruginosa QS populations in individual CF infections cannot be drawn from the characterization of one or a few selected isolates.

Pseudomonas aeruginosa is a ubiquitous opportunistic pathogen that causes a range of acute and chronic infections in immunocompromised individuals, including those suffering from the genetic disorder cystic fibrosis (CF). In these patients, P. aeruginosa persistently colonizes the lungs, forming antibiotic-resistant biofilms (40, 58, 65). Many of the factors implicated in P. aeruginosa virulence are controlled by quorum sensing (QS) (3, 15, 70). This includes biofilm formation, although the impact of QS is environmentally conditional (8, 26, 70). Thus, QS has been considered a promising alternative drug target (2, 6). In QS, bacteria are able to coordinate diverse group behaviors through the production of chemical signals in a density-dependent manner (66). In P. aeruginosa and other gram-negative bacteria, the majority of these signals are acyl-homoserine lactones (acyl-HSL) (14, 68).

The P. aeruginosa acyl-HSL QS circuitry is particularly well understood. It is comprised of two acyl-HSL synthases, LasI and RhlI, that generate N-3-oxododecanoyl-HSL (3OC₁₂-HSL) and N-butyryl-HSL (C₄-HSL), respectively. As the bacterial population increases, the signals accumulate intra- and extracellularly until they bind to their cognate transcriptional regulators, LasR and RhlR, respectively, to activate the expression of target genes (23, 56). These two QS systems are arranged hierarchically, with the LasR-LasI circuit in control of the RhlR-RhlI circuit (30, 49), although under certain conditions, expression of RhlR-RhlI can occur independently (9, 35). Another orphan regulator, QscR, exists, which also responds to 3OC₁₂-HSL (31).

Transcriptome analysis of P. aeruginosa quorum-controlled genes revealed that both systems together control the expression of hundreds of genes (17, 57, 64), many of which encode virulence factors. In animal models of acute and chronic P. aeruginosa infections, mutations in central regulatory gene lasR result in attenuated virulence compared to that of the wild-type strain (32, 46, 52, 72). However, although QS is required for full virulence in many model systems, clinical and environmental P. aeruginosa isolates are often QS deficient (7, 13, 18, 20, 60, 62). Loss-of-function mutations in the central regulator lasR gene are most frequent. This apparent paradox may be explained by a physiological advantage of the lasR mutant under certain growth conditions (7, 19). Alternatively, lasR mutants may emerge by social exploitation, as initially shown in vitro. Under conditions that require QS, lasR mutants evolve from a wild-type ancestor during long-term culturing and have a selective advantage when cocultured with the wild-type parent (10, 54). These mutants also increase in frequency during acute P. aeruginosa infection of mechanically ventilated patients (27) and during experimental coinfection in a mouse burn wound model (51). This is because lasR mutants benefit from the extracellular products made by the QS-proficient wild type without paying the metabolic cost. Thus, QS-deficient variants can arise because of, rather than despite, the importance of QS. Social conflict should be of particular significance in a localized, long-term infection such as CF, which according to evolutionary theory, would result in increased competition and selection for cheaters (10, 67). If QS-deficient variants in CF lung infection indeed arise by social exploitation, then populations of P. aeruginosa cells isolated from individual patients should consist of mixtures of wild-type and mutant cells. Moreover, if there is a single benefit to QS mutants, be it social or nonsocial, then we would expect to find one predominant mutant phenotype and little diversity.

Although several studies have analyzed QS in CF isolates, they have generally focused on either between-patient diversity or longitudinal within-patient diversity, rarely involving more than one isolate per patient at any one time (13, 53, 55, 60, 62). Very little attention has been given to the issue of instantaneous within-patient heterogeneity, that is, diversity among isolates collected from a single patient at any one time. One recent study screened an average of three P. aeruginosa isolates from 58 young CF patients, many concurrently isolated, for the presence of lasR mutations (20). Overall, 31% of the isolates collected were mutated in lasR. In several cases, individual patients harbored both wild-type and lasR mutant isolates at the same time. The distinguishing feature of the present study is that a large number of concurrent isolates were comprehensively characterized, phenotypically and genotypically, from individual adult CF patients, allowing for a more complete description of instantaneous diversity and types of QS-related features.

MATERIALS AND METHODS

CF patients, bacterial strains, and plasmids.

Laboratory strains and plasmids used in this study are shown in Table 1. Clinical P. aeruginosa isolates were obtained concurrently from the sputa of eight adult CF patients at Oregon Health and Science University (OHSU). Sputum samples were collected from CF patients by expectoration during a routine hospital visit. Patients were between 20 and 48 years old. Infection with P. aeruginosa had been documented since their treatment at OHSU (between 1 and 10 years), although actual infection is likely to be longer. Patients' lung functions ranged from mildly to severely obstructed. Lung function was measured by spirometry as forced expiratory volume during the first second (FEV1) (38). Greater than 90% FEV1 is considered normal, 70 to 89% FEV1 is considered mild obstruction, 49 to 69% FEV1 is considered moderate obstruction, and less than 40% FEV1 is considered severe obstruction (38).

TABLE 1.

Bacterial strains and plasmids

Lab strain or plasmid	Relevant property(ies)	Reference
Strains
PAO1	Pseudomonas aeruginosa wild type	22
PDO300	mucA22 derivative of PAO1	34
PAO lasR	PAO1 derivative; lasR::Tetr	50
PAO rhlR	PAO1 derivative; rhlR::Genr	50
PAO lasR rhlR	PAO1 derivative; lasR::TetrrhlR::Genr	50
PAO lasI rhlI	PAO1 derivative; lasI::TetrrhlI::Tn501	69
Plasmids
pJN105	araC-pBAD cloned in pBBR1 MCS-5; Gmr	41
pJN105.lasR	lasR in pJN105	31
pJN105.lasRD65G	lasRD65G from patient B isolate in pJN105	This study

Bacterial isolation and species identification.

Sputum samples were instantly stored on ice, express shipped from OHSU to Oregon State University, and processed immediately. Sputa were liquefied by the addition of an equal volume of Sputolysin (Calbiochem), diluted, and plated onto Pseudomonas isolation agar (BD Biosciences), which permits growth of P. aeruginosa and other Pseudomonas species. Twenty isolates were randomly picked from a single sputum sample per patient and saved as frozen stock cultures. PCR amplification with P. aeruginosa-specific 16S ribosomal DNA primers was used to confirm species identity (61). Only isolates identified as P. aeruginosa were used for this study. Mucoidy was evaluated based on distinct colony morphology (34). Relative abundance of P. aeruginosa compared with that of other microbial flora was estimated based on routine four-quadrant plating of CF sputum samples by the Kaiser Permanente Clinical Microbiology Laboratory (Portland, OR).

Production of extracellular proteases, rhamnolipids, and nucleoside hydrolase.

Assays for the production of extracellular proteases, rhamnolipids, and nucleoside hydrolase were performed in high-throughput format. Skim milk proteolysis (54), rhamnolipid production (28), and growth on adenosine, indicative of nucleoside hydrolase production (19, 54), were determined semiquantitatively through the use of agar plate assays. Isolates were initially grown on Lennox LB agar plates at 37°C for 24 to 48 h. Colonies were subsequently replica plated onto the respective agar test plates and evaluated after 24 h of growth. The scoring scheme was as follows. For adenosine plates, growth was scored as “+,” while the absence of growth was scored as “−.” For skim milk proteolysis and rhamnolipid production, the formation of a halo similar to that of the PAO1 wild type was scored as “+,” the formation of a halo significantly smaller than that of the PAO1 wild type was scored as “○,” and the absence of a halo was scored as “−.” Severe growth deficiencies were indicated as “no growth.”

LasA and LasB protease activities were measured from Lennox LB liquid cultures grown in deep-well titer blocks for 24 h. All cultures reached stationary phase by the time of measurement, as determined by the optical density at 600 nm (OD₆₀₀). Assays were adapted for high-throughput analysis in 96-well plate format using a microplate reader (Infinite M200; Tecan). LasA protease activity was determined by measuring the rate at which filter-sterilized P. aeruginosa culture supernatants could lyse boiled Staphylococcus aureus cells (11, 24). Twenty microliters of bacterial supernatant was added to each well of a 96-well plate, followed by the addition of 180 μl of boiled S. aureus cells to each well. The OD₆₀₀ was measured every 5 min for 75 min. LasB (elastolytic) activity of bacterial supernatants was determined using an elastin Congo red assay as previously described (11). Twenty-two microliters of bacterial supernatant was added to 200 μl of elastin congo red buffer in 96-well plates. Absorbance was read at 495 nm. For both assays, ≥80% activity compared with that of the PAO1 wild-type control was scored as “+,” 10 to 79% activity was scored as “○,” and <10% activity was scored as “−” (see Table S1 in the supplemental material). Additional controls included in all assays were defined as lasR, rhlR, and lasR rhlR mutants (Table 1).

For our initial screening, we chose the aforementioned approach of assaying QS-controlled phenotypes of all clinical isolates at one particular time, although these isolates exhibited a range of different growth rates. Few grew as fast as the wild type, and most grew significantly slower. However, when dealing with 135 samples in our initial screening, it would not have been feasible to specifically adjust the time of harvesting for every single isolate. Instead, we chose a time (24 h) at which even the slowest isolate had well reached stationary phase. Because the quorum-controlled products assayed are typically induced at the transition from logarithmic to stationary phase, every isolate that did not produce QS factors was thereby correctly classified as QS negative. Individual isolates that produced QS factors essentially served as internal “controls” for other isolates from the same patient that were scored as QS negative but exhibited very similar growth characteristics (see Table S1 in the supplemental material). For the next step—the in-depth characterization of isolates from two selected patients—we chose an approach that takes growth rate differences into account (see below). These results nevertheless were in good agreement with the initial screening results.

Acyl-HSL detection.

Production of 3OC₁₂-HSL and C₄-HSL was quantified from ethyl acetate extracts of culture supernatants using Escherichia coli bioassays (45, 47, 48). MOPS (morpholinepropanesulfonic acid)-buffered Lennox LB cultures were inoculated to a starting OD₆₀₀ of 0.01 and grown until stationary phase. To correct for differences in growth, growth curves were determined in a preliminary experiment for all clinical isolates from patients A and B. Doubling times for most patient A and B isolates were 60 min compared with 30 min for those with the PAO1 control strain. Thus, for acyl-HSL analysis, most clinical isolates were grown for up to 25 h compared with 15 h for PAO1 and mutant derivatives. At the indicated times, all strains spent the same amount of time in stationary phase and reached similar culture densities. Diluted acyl-HSL extracts were added to E. coli biosensor strain MG4/pKDT17 to detect 3OC12-HSL and to strain DH5α/pECP61.5 to detect C₄-HSL. β-Galactosidase production by these E. coli strains was measured with a microplate reader using a luminescence assay (Galacto-Light Plus; Tropix) (54, 69).

RAPD analysis.

Random amplified polymorphic DNA (RAPD) analysis was performed as previously described using primer 208 and 40 ng of genomic DNA per PCR (33, 36). Chromosomal DNA was isolated from P. aeruginosa liquid cultures using the Puregene DNA purification kit (Gentra Systems). The PCR-amplified DNA products were resolved by agarose gel electrophoresis and stained with ethidium bromide. Gel images were taken using the BioDoc-It imaging system (UVP). Banding patterns were analyzed using GelCompar II software (Applied Math). This program aligns sample lanes to a reference ladder using pattern recognition. It then creates a dendrogram based on the similarity of banding patterns between samples. This entire process was performed twice with independently amplified DNA. In both cases, very similar results were obtained.

DNA sequencing.

PCR fragments amplified from isolated P. aeruginosa DNA were directly sequenced at the Center for Genome Research and Biocomputing at Oregon State University. Each allele was sequenced twice using independently amplified DNA. The 1.28-kb region of vfr was amplified using forward primer 5′-AAG GCT TCG CAG CTC TCC ACC G-3′ and reverse primer 5′-CTG GTT GGC GAG GCG TTC ATG CGG-3′. The 1.24-kb region of lasR was amplified using primers las1 and las2 (19). The 1.1-kb region of lasI was amplified using forward primer 5′-GGA AGT TCG GTG TGA CCT CCC GC-3′ and reverse primer 5′-CGG ATT CGG CAT CGA CGC CAG G-3′. The 1.35-kb region of rhlR was amplified using forward primer 5′-GGA GCC TTG CTG CCA TCG TGC G-3′ and reverse primer 5′-CCA AGT CCC CGT GTC GTG CCG-3′. The 1.11-kb region of rhlI was amplified using forward primer 5′-TGC TGA CCC AGA AGC TGA CCG ACC-3′ and reverse primer 5′-GCA GGG GAA CGC CCG ATG ACG C-3′. Appropriate sequencing primers were included with each sequencing reaction.

Complementation analysis.

Experimental cultures were grown as described for acyl-HSL analysis. Chemical complementation of patient A isolates was performed by addition of 3OC₁₂-HSL, C₄-HSL, both signals, or neither signal to cultures upon inoculation. The concentrations of 3OC₁₂-HSL and C₄-HSL were 2 and 10 μM, respectively, as previously described (57). 3OC₁₂-HSL was custom synthesized by RTI International (Research Triangle Park, NC), and C₄-HSL was custom synthesized by Vertex Pharmaceuticals (Coralville, IA). For lasR complementation of patient B isolates, we used the low-copy-number plasmids pJN105 and pJN105.lasR. In the latter, expression of lasR is under the control of the arabinose-inducible araBAD promoter (31, 41). Plasmids were introduced into clinical strains by transformation of chemically competent cells. For strains containing plasmids, 50 μg/ml gentamicin and 5 mM l-arabinose were added to the medium. The concentration of arabinose chosen results in LasR expression levels similar to those of the native chromosomal gene in the PAO1 wild type (data not shown). LasB elastase activity in culture supernatants was measured using the elastin Congo red assay, as described above.

Cloning and characterization of lasR D65G.

The lasR gene from an isolate from patient B harboring the D65G mutation was amplified with flanking primers (31) containing EcoRI and XbaI restriction sites and cloned into appropriately digested pJN105 (31). Proper construction was confirmed by restriction analysis and DNA sequencing. The resulting plasmid, pJN105.lasRD65G, was transformed into the defined lasR mutant, PAO lasR::Tc^r (50). This transformant and appropriate controls were analyzed for proteolysis on 4% skim milk plates supplemented with arabinose and gentamicin as described previously (54).

RESULTS

Assembly of a library of P. aeruginosa isolates from CF patients.

The main objective of this study was to determine the proportion of QS-deficient and QS-proficient clinical isolates within an individual CF lung infection. Our goal was to characterize numerous (up to 20) concurrent isolates per patient to obtain sufficient information about instantaneous QS diversity. To our knowledge, such a broad collection of concurrent isolates from single patients does not exist, necessitating new isolation of P. aeruginosa from CF patients. Sputum samples were collected from adult patients. The propensity for the emergence of QS-deficient variants is expected to be highest in these patients with long-term lung infection. In the sputum samples, P. aeruginosa was present at typical bacterial loads (10⁵ to 10⁸ CFU/ml) (21) and was in most cases the predominant, although not exclusive, organism. P. aeruginosa was isolated by plating on Pseudomonas isolation agar. Twenty candidate isolates were randomly picked, irrespective of morphotype, for further analysis. Species identity was verified by 16S rRNA analysis (61). In most cases, less than 20 of the original sputum isolates were confirmed as P. aeruginosa. Some P. aeruginosa populations were exclusively mucoid or nonmucoid, while others contained mixtures of both (Table 2; see also Table S1 in the supplemental material).

TABLE 2.

Characteristics of P. aeruginosa populations from CF patients

Patient	Severity of lung obstruction	Bacterial load in sputa (CFU/ml)	No. of isolates	% Mucoidy	% of negative isolates in the QS assay indicated
					Adenosine	Skim milk	Rhamnolipid	Staphylolysis	Elastin congo red
A	Mild	108	18	0.0	89	83	83a	83	83
B	Severe	107	18	100	78	78	78a	83	78
C	Mild	105	19	79	95	58	95a	5.3	5.3
D	Moderate	107	19	0.0	0.0	0.0	0.0	0.0	0.0
E	Moderate	106	20	100	15	70	70	30	20
F	Severe	108	9	0.0	0.0	78	44	44	44
G	Moderate	106	14	79	71	14	64a	64	71
H	Moderate	105	18	83	83	83	89	61	67

^aIncludes “no growth” isolates (see Table S1 in the supplemental material).

CF patients are commonly infected with a large subpopulation of QS-deficient variants.

Our collection of P. aeruginosa CF isolates was screened for various QS-dependent phenotypes. All isolates were tested for growth on adenosine, skim milk proteolysis, rhamnolipid production, elastolytic (LasB) activity, and staphylolytic (LasA) activity. In the PAO1 reference strain, utilization of adenosine as the sole carbon source and skim milk proteolysis are las-dependent phenotypes (54); LasB and LasA activities are dependent on both las and rhl systems, whereas rhamnolipid production is primarily rhl dependent (57). We identified mixed populations of isolates with either partially or fully attenuated QS phenotypes in seven out of eight patients (Table 2; see also Table S1 in the supplemental material). Many isolates were fully deficient in some but not all of the QS phenotypes tested. All P. aeruginosa isolates from patient D displayed QS phenotypes similar to those of the PAO1 control strain. However, there was no apparent correlation between QS proficiency and total bacterial load. The QS-proficient P. aeruginosa population from patient D was not more abundant than other, largely QS-deficient populations.

Some isolates did not grow on adenosine medium and also did not grow on rhamnolipid detection medium (see Table S1 in the supplemental material). As these assays utilize minimal media, it is possible that some mutants did not grow because they are amino acid auxotrophs. Thus, lack of growth on adenosine does not necessarily indicate a QS defect. For the purpose of this study, we felt that it was unnecessary to distinguish between the two possibilities, as our phenotypic analysis included several other, independent assays that allowed us to evaluate QS proficiency.

P. aeruginosa populations from two selected patients are largely clonal, are deficient in acyl-HSL production, and harbor mutations in lasR, lasI, and rhlI.

Most P. aeruginosa isolates from two patients (patients A and B) were negative for all of the QS phenotypes tested (see Table S1 in the supplemental material), indicating mutations in central QS regulatory genes. These isolates were therefore selected for further in-depth analysis. Clonal relatedness among isolates was assessed by RAPD analysis (Fig. 1). All isolates from patient A were highly related (>80%). Most isolates from patient B were also highly related, whereas two isolates, B6 and B18, showed lower relatedness to other patient B isolates (40% and 70%, respectively). There were no strain variants common to both patients. Importantly, there was no apparent correlation between RAPD profiles and QS phenotypes. There were QS-deficient and -proficient (partially) isolates with the same RAPD type, suggesting that both types evolved from a common ancestor during infection.

FIG. 1.

RAPD analysis of P. aeruginosa isolates from patients A and B. Isolates A1 to A18 and B1 to B18 are isolates 1 to 18 from patient A and patient B, respectively. The upper scale next to the dendrogram indicates relatedness of isolates in percentages, whereas the lower scale indicates the sizes of DNA fragments in base pairs.

Next, we quantified production of 3OC₁₂-HSL and C₄-HSL using an E. coli bioassay (Fig. 2). Patient A isolates fell into two groups. Some produced very little to no 3OC₁₂-HSL and C₄-HSL, while others produced some 3OC₁₂-HSL but very little to no C₄-HSL (Fig. 2A). Complete lack of acyl-HSL production is consistent with mutations in either lasI and rhlI or lasR and rhlR. Most isolates from patient B showed substantially reduced 3OC₁₂-HSL levels but only slightly reduced C₄-HSL levels (Fig. 2B). This acyl-HSL profile is similar to that of a defined lasR mutant. Curiously, isolates scored as partially QS proficient (see Table S1 in the supplemental material) also produced little to no acyl-HSL (patient A isolates 3, 17, and 18 and patient B isolates 16, 17, and 18).

FIG. 2.

Acyl-HSL assay of P. aeruginosa isolates. Patient A (A); patient B (B). Black and gray bars designate 3OC₁₂-HSL and C₄-HSL, respectively. Acyl-HSL levels were normalized to the OD₆₀₀ of the respective culture at the time of measurement. The genotypes of the respective P. aeruginosa mutant control strains are indicated. Values represent the average results of three independent replicates. Error bars indicate standard deviations of the means.

To obtain further insights into the mutations responsible for the observed phenotypes, we sequenced lasR, rhlR, lasI, and rhlI from selected patient A isolates, and we sequenced lasR from selected patient B isolates (Table 3). These isolates were selected based on their QS phenotypes and acyl-HSL profiles (Fig. 2; see also Table S1 in the supplemental material). All sequenced isolates from patient A have missense mutations in rhlI, resulting in a Leu substitution of conserved Phe 28 and a Glu substitution of nonconserved Asp 83. Substitution F28L in rhlI was predicted to not be tolerated by the SIFT algorithm (42). QS-negative isolates 5 and 13 also harbor a frameshift mutation in lasI, resulting in a truncated protein. QS-negative isolate 8, which produced some 3OC₁₂-HSL, and partially QS-proficient isolate 18, which produced no acyl-HSL, showed no mutation in lasI. Of note, whereas defined rhlR mutants are completely deficient in the standard rhamnolipid plate assay employed, rhlI mutants are only partially deficient (29) (data not shown). This could explain why isolate 18 was scored as partially rhamnolipid proficient (see Table S1 in the supplemental material), despite carrying a mutation in rhlI.

TABLE 3.

Sequence analysis of selected P. aeruginosa CF isolates

Isolatea	Gene	Nonsynonymous mutation(s) (position)b	Change(s)c
Patient A
5	lasR	None	None
5	lasI	A insertion (+43)	Truncation
5	rhlR	None	None
5	rhlI	T→C (+82), C→A (+249)	F28L*†, D83E
8	lasR	None	None
8	lasI	None	None
8	rhlR	None	None
8	rhlI	T→C (+82), C→A (+249)	F28L*†, D83E
13	lasR	None	None
13	lasI	A insertion (+43)	Truncation
13	rhlR	None	None
13	rhlI	T→C (+82), C→A (+249)	F28L*†, D83E
18	lasR	None	None
18	lasI	None	None
18	rhlR	None	None
18	rhlI	T→C (+82), C→A (+249)	F28L*†, D83E
Patient B
4	lasR	A→G (+194)	D65G†
4	vfr	None	None
6	lasR	A→G (+194)	D65G†
6	vfr	None	None
7	lasR	A→G (+194)	D65G†
7	vfr	None	None
17	lasR	A→G (+194)	D65G†

^aIndividual clinical isolates.

^bNucleotide substitution or insertion at the indicated position relative to translational start site.

^cAmino acid changes relative to the PAO1 protein sequence. *, the respective amino acid residue is conserved in LuxR and LuxI-type proteins from a range of different bacterial species (68); †, the respective amino acid residue is conserved in P. aeruginosa PAO1, PA7, PA14, and LESB58, according to www.pseudomonas.com.

All sequenced isolates from patient B carry a missense mutation in lasR, resulting in an Asp-to-Gly substitution at position 65, which is adjacent to highly conserved Tyr 64. The fact that the same mutation was also present in partially QS-proficient isolate 17 suggests that it may not be fully responsible for the observed loss-of-function phenotypes in QS-deficient isolates, although this mutation was predicted to not be tolerated by the SIFT algorithm (42).

Complementation analysis confirms the significance of lasR and acyl-HSL-dependent phenotypes, but not of the lasR mutation itself.

Complementation analysis was performed in order to determine whether the identified mutations are responsible for the observed phenotypes. LasB elastase production was used as a representative QS assay because it is controlled by both the las and rhl QS systems. Selected patient A isolates mutated in lasI or rhlI (Table 3) were chemically complemented using synthetic acyl-HSL signals (Fig. 3A). Isolates 5 and 13, which carry mutations in both lasI and rhlI, showed the highest elastase production by the addition of both 3OC₁₂-HSL and C₄-HSL. Isolates 8 and 18, which carry a mutation in rhlI only, could be complemented by the addition of C₄-HSL alone. Curiously, the low-level elastase production by isolate 8 was completely inhibited by the addition of 3OC₁₂-HSL.

FIG. 3.

Complementation analysis of selected P. aeruginosa isolates. Elastolytic (LasB) activity of culture supernatants was measured using the elastin Congo red assay and was normalized to culture density (OD₆₀₀). (A) Complementation of patient A isolates with exogenous acyl-HSL. Black bars indicate the absence of acyl-HSL, dark gray bars indicate the presence of 3OC₁₂-HSL, light gray bars indicate the presence of C₄-HSL, and white bars indicate the presence of both 3OC₁₂-HSL and C₄-HSL. (B) Complementation of patient B isolates with lasR. Black bars indicate the presence of pJN105, and gray bars indicate the presence of pJN105.lasR. Values represent the average results of three independent replicates. Error bars indicate standard deviations of the means. Statistical significance of the data was determined using a two-tailed unpaired t test, with “***” indicating P values of <0.017 (treatment-level Bonferroni correction to adjust for multiple comparisons), “**” indicating P values of ≥0.017 but <0.05, and “*” indicating P values of ≥0.05 but <0.1. Within each isolate, expression in the presence of either one or both acyl-HSL signals was compared to that in the presence of no acyl-HSL (A), and expression with pJN105.lasR was compared to that with the vector control (B).

Isolates from patient B were complemented with a plasmid (pJN105) carrying lasR under the control of an arabinose-inducible promoter (31). Although the relevance of the lasR D65G mutation itself was questionable, we nevertheless reasoned that the observed loss-of-function phenotypes were lasR dependent because acyl-HSL profiles in many isolates were similar to that of a defined lasR mutant. Indeed, in QS-deficient isolates 4 and 6 (Tables 2 and 3; see also Table S1 in the supplemental material), elastase production increased significantly by introducing lasR in trans (Fig. 3B). Moreover, the lasR D65G allele, when cloned in pJN105, was able to fully complement skim milk proteolysis of a defined lasR mutant (Fig. 4). This result indicates that the D65G mutation itself is not responsible for the observed QS phenotype. Instead, our results indicate that lasR expression is abolished due to an upstream regulatory defect, because (i) lasR expression from a heterologous promoter was able to restore QS and (ii) there were no mutations in the lasR promoter region. We found no mutations in vfr, which encodes a major regulator of lasR transcription that is often mutated in CF isolates (1, 60), suggesting a defect in another regulatory gene (Table 3).

FIG. 4.

Complementation of a defined lasR mutant with the lasR D65G allele. Proteolysis was measured on skim milk plates.

DISCUSSION

In the present study, we characterized P. aeruginosa populations concurrently isolated from chronically colonized adult CF patients to obtain insight into the within-patient diversity of QS. In seven out of eight patients, we found mixtures of isolates with distinct QS-proficient and QS-deficient phenotypes. Deficiency in individual QS-dependent phenotypes was not the result of attenuated growth of some of the clinical isolates. In our initial phenotypic screenings, isolates were grown long enough to reach the stationary phase of growth; the PAO1 control strain shows high-level expression of QS products as soon as cells transition from logarithmic to stationary phase. In subsequent acyl-HSL and complementation assays performed with patient A and B isolates, culturing times were carefully adjusted according to growth rate, and data were normalized to culture density.

We found that the diversity of QS phenotypes is high among patients but also within a given patient. This result suggests that the selective forces that shape the evolution of QS during CF lung infection are also diverse, which is not surprising if one considers the complexity of this infection. Previously, between-patient diversity has been inferred mainly from the characterization of single isolates from different infections (13, 53, 55, 62). It is thought to result from individual differences in the genetic makeup and virulence potential of the infecting strain(s) (63), in the coinfection with other microbial species (16), in the severity and context of the genetic defect (25), and in host immune function (39). Within-patient QS diversity, on the other hand, has been much less well recognized but is in fact a prerequisite to accurately assessing between-patient diversity. Within-patient diversity may result from the heterogeneity of the lung environment. Given the compartmentalized nature of the lung, and the high viscosity of CF mucus, P. aeruginosa cells even within a single infection are exposed to distinct microenvironments. These include steep nutrient and oxygen gradients in the mucus lining the CF airways that impact P. aeruginosa metabolism (71, 73), differences in the exposure to host inflammatory responses, and microbial communities of different species compositions in different lobes of the lung (59). Taken together, all these factors may contribute to the diversification of P. aeruginosa QS populations in the CF lung. The observed diversity also confirms the notion that conclusions about P. aeruginosa populations cannot be determined through the characterization of only a select few isolates. Instantaneous within-patient heterogeneity of P. aeruginosa from CF lung infections has been recognized for some time (63, 74) but has primarily been restricted to distinct morphotypes, most notably mucoid and nonmucoid colony variants.

Two specific mechanisms, social and nonsocial, have been proposed to explain the frequent emergence of QS variants during infection. Previous studies demonstrated that social cheating can select for QS mutants in vitro, in an animal infection model, and presumably also in acute human infection (10, 27, 51, 54). Here, QS-deficient bacteria benefit from the production of QS-controlled extracellular products by the QS-proficient majority. QS deficiency may also confer an innate (nonsocial) physiological advantage under certain growth conditions. For example, lasR mutants have a higher growth yield on certain carbon sources, including amino acids, and show increased resistance to alkaline lysis in stationary phase compared to that of the wild-type parent (7, 18). These properties may be relevant in vivo, as amino acids are abundant in CF sputum (44), and amino acid catabolism may lead to localized alkalinization.

Consistent with social cheating, many CF sputum samples contained mixed populations of QS-proficient and QS-deficient isolates; no sample contained 100% fully deficient QS mutants. However, in several cases, the proportion of QS-deficient phenotypes was very high (≥83% for patient A, ≥78% for patient B, and ≥61% for patient H) (Table 1). If intact QS were indeed required for chronic persistence, then it appears questionable whether the small fraction of QS-proficient cells would be able to sustain the entire P. aeruginosa population. Based on recent in vitro and in vivo studies, high proportions of QS-deficient variants are predicted to significantly decrease fitness under conditions that require QS (10, 51, 54). We also found that predominantly QS-deficient P. aeruginosa CF populations persisted and colonized the CF lung to levels similar to those of QS-proficient populations and that patients harboring these QS-deficient populations had significantly impaired lung function (much of which is generally attributed to P. aeruginosa-triggered inflammation). Although more data will be needed to substantiate this finding, one possible interpretation is that QS is not important for chronic persistence in some individuals and, consequently, that social conflict is not a major selective force. This notion is consistent with a recent longitudinal screening of P. aeruginosa CF isolates for mutations in lasR, suggesting that lasR mutants are associated with lung disease progression (20). In addition, acyl-HSL signals are often not detectable in sputum samples from CF patients colonized with P. aeruginosa. 3OC₁₂-HSL could not be detected in 22%, 38%, and 46% of all samples in three previous studies (5, 12, 37), respectively, while C₄-HSL could not be detected in 74% of samples (12). These data are in accordance with the view that invasive functions (which include many secreted quorum-controlled products and other “traditional” virulence factors) may become dispensable in progressive infections, which often occur in damaged tissues where nutrients are more accessible and immune responses are diminished (43). In this case, the efficacy of QS as an antivirulence drug target appears questionable. It is conceivable that QS, and hence, social cheating, is more important early in chronic infection, for example, for initial colonization of the CF lung. Data from chronic infection models indeed show that QS contributes to lung colonization (72). Thus, in addition to multiple selective pressures acting on QS evolution in the CF lung at any one time, there may be multiple forces acting at different times during the course of infection.

Many of our isolates were fully defective for some but not all of the phenotypes tested, suggesting either that the responsible mutations are not in central regulatory genes or that the mutations are in central genes but secondary mutations partially restore expression of QS phenotypes. Global QS deficiencies in patient A and B isolates were due to mutations in lasI and rhlI as well as impaired lasR expression. Complementation analysis of several patient B isolates harboring a presumed loss-of-function lasR allele indicated that an upstream regulatory defect, rather than the lasR mutation itself, is the underlying cause, emphasizing the importance of experimental verification. While we ruled out a defect in one important regulator of lasR expression, Vfr (1), which is commonly mutated in CF isolates (60), there are several other candidates (GacAS/RsmAZ, RelA, and RpoS) (56) and presumably more that remain to be discovered. Chemical and genetic complementations with acyl-HSL signals and lasR, respectively, significantly increased elastase production, although absolute levels were, with one exception, lower than that of the PAO1 reference strain. This could mean that the genetic capacity for maximal elastase expression is simply lower in the clinical isolates or that additional, specific mutations contribute to low-level expression. The latter is plausible since these clinical strains grew significantly slower in rich medium, indicating pleiotropic defects beyond QS. If social exploitation was the primary selective force in the evolution of QS variants, one might not expect to see lasI and rhlI mutants as well as other lasR-independent QS deficiencies. Mutations in the las system could be favored because it is atop the QS hierarchy, although as recently shown, this regulatory hierarchy is dependent on growth conditions (9, 35). In particular, lasR mutants should have a selective advantage over lasI mutants. lasR mutants deficient in the expression of hundreds of QS-dependent genes would benefit from their production by others, whereas lasI mutants would benefit only from the 3OC₁₂-HSL produced by others.

Although our prediction that global QS deficiencies are due to mutations in central QS genes turned out to be true, the correlation among individual QS phenotypes, acyl-HSL production, and underlying mutation in clinical isolates was not always as straightforward as one might expect from studies with the PAO1 reference strain. For example, patient A isolates 3, 17, and 18 were partially proficient in expressing QS-controlled factors (Table 3 and Fig. 3A; see also Table S1 in the supplemental material), although they did not produce detectable levels of acyl-HSL (Fig. 2A). This suggests that in some cases, QS-target genes are expressed independently of a functional QS circuitry, perhaps due to second-site mutations in QS-deficient backgrounds. If this is true, then the restoration of individual QS-regulated functions would in some cases be beneficial for chronic persistence. Similar observations have been made in other studies. In a study by Tingpej et al., a small proportion of CF isolates produced the QS-controlled factors rhamnolipid and chitinase, although they did not produce detectable acyl-HSL (62). In a study by Schaber et al., one isolate from a urinary tract infection produced high levels of pyocyanin, although it produced less acyl-HSL than the PAO1 reference strain (55). Cabrol et al. reported a lack of correlation between the growth-phase-dependent transcription patterns of lasR and selected target genes, but there was a reasonable correlation between the presence of a mutation in lasR predicted to impair function and QS target gene expression (4).

Deviation in QS regulatory control from the PAO1 paradigm, along with the instantaneous within- and between-patient diversities reported here, adds to the complexity of the role of QS in CF lung infection. Our results suggest that any one selective mechanism alone cannot account for the evolution of such diversity. An integrative cross-sectional and longitudinal approach will be needed to fully elucidate the forces driving QS evolution during chronic persistence of P. aeruginosa.