Expression of Pseudomonas aeruginosa Antibiotic Resistance Genes Varies Greatly during Infections in Cystic Fibrosis Patients

The lungs of individuals with cystic fibrosis (CF) become chronically infected with Pseudomonas aeruginosa that is difficult to eradicate by antibiotic treatment. Two key P. aeruginosa antibiotic resistance mechanisms are the AmpC β-lactamase that degrades β-lactam antibiotics and MexXYOprM, a three-protein efflux pump that expels aminoglycoside antibiotics from the bacterial cells.

ABSTRACT

The lungs of individuals with cystic fibrosis (CF) become chronically infected with Pseudomonas aeruginosa that is difficult to eradicate by antibiotic treatment. Two key P. aeruginosa antibiotic resistance mechanisms are the AmpC β-lactamase that degrades β-lactam antibiotics and MexXYOprM, a three-protein efflux pump that expels aminoglycoside antibiotics from the bacterial cells. Levels of antibiotic resistance gene expression are likely to be a key factor in antibiotic resistance but have not been determined during infection. The aims of this research were to investigate the expression of the ampC and mexX genes during infection in patients with CF and in bacteria isolated from the same patients and grown under laboratory conditions. P. aeruginosa isolates from 36 CF patients were grown in laboratory culture and gene expression measured by reverse transcription-quantitative PCR (RT-qPCR). The expression of ampC varied over 20,000-fold and that of mexX over 2,000-fold between isolates. The median expression levels of both genes were increased by the presence of subinhibitory concentrations of antibiotics. To measure P. aeruginosa gene expression during infection, we carried out RT-qPCR using RNA extracted from fresh sputum samples obtained from 31 patients. The expression of ampC varied over 4,000-fold, while mexX expression varied over 100-fold, between patients. Despite these wide variations, median levels of expression of ampC in bacteria in sputum were similar to those in laboratory-grown bacteria. The expression of mexX was higher in sputum than in laboratory-grown bacteria. Overall, our data demonstrate that genes that contribute to antibiotic resistance can be highly expressed in patients, but there is extensive isolate-to-isolate and patient-to-patient variation.

INTRODUCTION

Pseudomonas aeruginosa infects the lungs of individuals with cystic fibrosis (CF), with infections lasting for years or decades (1–3). Chronic P. aeruginosa infections in patients with CF are associated with progressive clinical decline and increased morbidity, leading to respiratory failure and premature death (4–6). Aggressive antibiotic treatment can eliminate P. aeruginosa from the lower airways during the initial stage of infection (1, 3). However, once chronic infection becomes established, existing antibiotic treatment regimens usually fail to eradicate infection, although they reduce the bacterial burden (7, 8). Treatment differs between patients and between CF care centers but typically involves the use of one or more aminoglycosides, such as tobramycin; a carbapenem, including meropenem; a fluoroquinolone, such as ciprofloxacin; and in some cases, colistin, a polymyxin (9, 10). P. aeruginosa undergoes genetic changes during the course of chronic infection that help it adapt to the CF airway microenvironment, including mutations associated with increased antibiotic resistance (11, 12). CF is characterized by periods of clinical stability interspersed with pulmonary exacerbations that are associated with reduced pulmonary lung function and require hospitalization, though the relationship between bacterial infection and exacerbation is unclear (1).

P. aeruginosa has multiple antibiotic resistance mechanisms (13–15). The expression of a number of these is induced by substrate antibiotics leading to increased resistance in a process known as adaptive resistance (13, 16–18). Two key antibiotic-inducible systems are AmpC β-lactamase, which confers resistance to members of the β-lactam class of antibiotics (14), and an efflux pump, MexXYOprM, that is associated with resistance to aminoglycosides and fluoroquinolones, among other antibiotics (19, 20). β-Lactam antibiotics increase expression of the ampC gene, and aminoglycoside substrates increase expression of the mexXY genes, which are expressed as an operon (21). P. aeruginosa isolates cultured from individuals with CF commonly possess genetic mutations that give rise to increased ampC and mexXY expression and are associated with increased β-lactam and aminoglycoside antibiotic resistance (22–31).

The physiology of P. aeruginosa growing in pure culture in laboratory medium is likely to be quite different from that in the CF airway where they exist in biofilms within the context of the infected host, with multiple other microbial species present (32, 33). These physiological differences may have profound effects on bacterial gene expression and antibiotic resistance (34, 35). However, there is limited information on the expression of antibiotic resistance genes as it occurs during infection. In a pioneering study, RNA was extracted from P. aeruginosa in samples of sputum from a single patient without any ex vivo incubation or laboratory culture of the bacteria in order to analyze bacterial gene expression in vivo in the patient (36). The expression of multiple genes, including ampC and the mexXY efflux genes, occurred at a much higher level during infection in the patient than in bacteria isolated from the patient and grown under laboratory conditions.

The overall aims of this study were to measure the expression of ampC and mexX antibiotic resistance genes during infection in CF, and to determine how gene expression during infection relates to expression and antibiotic resistance of bacteria in laboratory culture.

RESULTS

Expression of antibiotic resistance genes in P. aeruginosa in sputum from patients with CF.

Expectorated sputum was obtained from 36 patients with CF and P. aeruginosa infection of different ages (range, 11 to 58 years; median age, 24 years) attending four different CF centers (Dunedin, Hobart, and two independent hospitals in Brisbane). Patient demographics are listed in Table 1. Two of the patients were transiently infected with P. aeruginosa, with the remainder of the patients being chronically infected.

TABLE 1.

Isolates of Pseudomonas aeruginosa used in this study

Isolatea	Source and/or description	Patient sexb	Patient age (yr)	Reference or source
Patients with cystic fibrosis
001C	Dunedin	M	51	This study
002C	Dunedin	M	25	This study
003B	Dunedin	F	36	This study
004	Dunedin	M	20	This study
005A	Dunedin	M	28	This study
009B	Dunedin	M	30	This study
012-4	Dunedin	F	24	This study
012-2	Dunedin	F	24	This study
014-2A	Dunedin	M	43	This study
015A	Dunedin	F	28	This study
021-3A	Dunedin	M	23	This study
030-1	Dunedin	M	58	This study
036-1	Dunedin	F	31	This study
039-1	Dunedin	M	22	This study
284	Tasmania	F	20	This study
451	Tasmania	F	23	This study
310A	Tasmania	F	26	This study
454A	Tasmania	M	24	This study
399	Tasmania	F	28	This study
254	Tasmania	M	50	This study
409	Tasmania	F	22	This study
001-A1	Brisbane	F	16	This study
002-A2	Brisbane	F	11	This study
003-A1	Brisbane	F	11	This study
006-A2	Brisbane	F	15	This study
007-A1	Brisbane	M	16	This study
008-A1	Brisbane	F	12	This study
403-102	Brisbane	F	22	This study
403-110	Brisbane	F	22	This study
403-103	Brisbane	M	38	This study
403-104	Brisbane	M	28	This study
403-105	Brisbane	F	19	This study
403-106	Brisbane	F	36	This study
403-107	Brisbane	F	26	This study
403-108	Brisbane	F	23	This study
403-109	Brisbane	M	21	This study
CF07	Melbourne	M	5	54
CF13	Auckland	M	5	54
CF21	Sydney	M	2	54
CF23	Sydney	F	2	54
Non-cystic fibrosis strains
PAO1	Melbourne; wound isolate			55
PW8616	PAO1 with transposon insertion phoAwp06q2E07 in ampD gene			56
PW4501	PAO1 with transposon insertion lacZbp01q1C12 in mexZ gene			56
ATCC 25873	USA; blood culture			57
Pa4	Belgium; urinary tract infection			58
Pa5	Belgium; intensive care unit			58
Pa6	Belgium; urinary tract infection			58

^aMatching sputum samples were available for all CF isolates except CF07-CF23. All CF isolates were from chronically infected patients, except for 002-A2, 003-A1, and CF07-CF23.

^bM, male; F, female.

The expression of ampC and mexX antibiotic resistance genes was measured by extracting RNA from expectorated sputum and then carrying out RT-qPCR. Gene expression was quantified for 31 of the 36 samples analyzed, with the remaining samples either giving no detectable expression of the P. aeruginosa genes tested (2 samples) or showing expression at a level that was too low for quantification (3 samples; see Table S1 in the supplemental material). Among the quantifiable samples, the expression of ampC varied over 4,000-fold (range, 0.004 to 24.6; median, 1.74; 25th to 75th percentiles, 0.34 to 3.71, relative to the reference genes), while mexX expression varied by approximately 100-fold (range, 0066 to 6.65; median, 1.77; 25th to 75th percentiles 0.426 to 3.06) (Fig. 1A and Table S1). There was no evidence of correlation between the expression of ampC and the expression of mexX (P = 0.559) (Fig. 1B). Taken together, these data indicate that both ampC and mexX are expressed during CF airway infection, but there is very marked variation in the expression of these genes across sputum samples from different patients.

FIG 1.

Expression of antibiotic resistance genes in P. aeruginosa in sputum. Expression of ampC and mexX was measured in 31 sputum samples from patients with cystic fibrosis. (A) The individual expression levels of ampC and mexX (circles), as well as boxplots showing medians, 25th and 75th percentiles, and minimum and maximum values, are shown. (B) Comparison of expression of ampC and mexX. Expression of ampC and mexX for each sample is shown.

Correlations between resistances to different antibiotics.

At the same time that samples were collected for analysis of gene expression, P. aeruginosa was isolated from sputum samples. We first investigated whether MICs for different antibiotics were correlated. MICs were determined for 6 antibiotics for isolates from 28 patients. The antibiotics selected represented three commonly used classes of anti-Pseudomonas antibiotics used in the treatment of CF, comprising β-lactams (aztreonam, ceftazidime, meropenem, and ticarcillin-clavulanic acid [Timentin]), an aminoglycoside (tobramycin), and a fluoroquinolone (ciprofloxacin). MICs are shown in Table S1. Using the EUCAST breakpoints (37), 21% of the strains were resistant to aztreonam, 32% were resistant to ceftazidime, 63% were resistant to meropenem, 25% were resistant to ticarcillin-clavulanic acid, 40% were resistant to tobramycin, and 25% were resistant to ciprofloxacin. Four isolates showed resistance to all six agents, and a further three strains were sensitive to only one antibiotic.

Correlation analysis was carried out to test the hypothesis that there is a correlation between levels of resistance to different antibiotics (Table 2). There was a strong correlation between MICs for the four β-lactam antibiotics aztreonam, ceftazidime, meropenem, and ticarcillin-clavulanic acid (Spearman's correlation coefficients = 0.74 to 0.93, P < 0.001 in all cases). There was also a significant, though lower, correlation between MICs for antibiotics from different classes. For example, resistance to meropenem correlated with resistance to tobramycin (Spearman's correlation coefficient = 0.63, P < 0.001), as well as with resistance to ciprofloxacin (Spearman's correlation coefficient = 0.60, P < 0.001).

TABLE 2.

Correlations between antibiotic MICs^a

Antibiotic	Aztreonam	Ceftazidime	Meropenem	Ticarcillin-clavulanic acid	Ciprofloxacin	Tobramycin
Aztreonam	1.00
Ceftazidime	0.84 (<0.001)	1.00
Meropenem	0.80 (<0.001)	0.91 (<0.001)	1.00
Ticarcillin-clavulanic acid	0.88 (<0.001)	0.82 (<0.001)	0.74 (<0.001)	1.00
Ciprofloxacin	0.35 (0.0646)	0.55 (0.0026)	0.60 (0.0008)	0.36 (0.0806)	1.00
Tobramycin	0.40 (0.0334)	0.44 (0.0202)	0.63 (<0.001)	0.25 (0.2309)	0.58 (0.0030)	1.0000

^aMICs were measured for 28 isolates of P. aeruginosa from CF patients (Table S1), and Spearman's correlation coefficients for pairs of antibiotics are shown. P values for correlations between values are shown in parentheses. Correlation analysis between MICs of meropenem and tobramycin included data for an additional 11 isolates (Table S1).

The same antibiotics were examined for their ability to induce the expression of ampC and mexX in P. aeruginosa reference strain PAO1. Meropenem was the strongest inducer of ampC expression (8.6-fold), and tobramycin was the strongest inducer of mexX expression (6.5-fold) (Fig. S1A).

Expression of antibiotic resistance genes in laboratory-grown bacteria.

We also measured the in vitro expression of ampC and mexX among P. aeruginosa isolates cultured from sputum and grown on laboratory agar in both the presence and absence of two antimicrobial agents. The results are listed in Table S1 and summarized in Fig. 2. In the absence of antibiotics, the expression of ampC varied over 20,000-fold between isolates (range, 0.002 to 46.5; median, 1.28; 25th to 75th percentiles, 0.178 to 10.9). There was less variation in the expression of mexX (range, 0.004 to 7.18; median, 0.189; 25th to 75th percentiles, 0.019 to 1.09), but nonetheless, there was a 2,000-fold difference in the expression of mexX between the lowest- and highest-expressing strains. The median expression values for the five non-CF strains were lower (0.121 for ampC and 0.011 for mexX) than for the CF strains. Mutations in ampD or mexZ, which are known regulators of ampC and mexX, respectively (14, 38), resulted in increased ampC and mexX gene expression in the reference strain PAO1, as expected (Fig. S1B). There was a correlation between the expression of ampC and mexX for the isolates of P. aeruginosa from CF patients (Spearman's correlation coefficient = 0.51, P = 0.002) (Fig. 2C). This is consistent with the strong correlation between resistance to tobramycin and to meropenem (Table 2).

FIG 2.

Expression of antibiotic resistance genes in laboratory-grown bacteria. Expression of ampC and mexX was determined for 36 isolates of P. aeruginosa from different CF patients cultured in the absence of antibiotic and for 14 isolates cultured in the presence of sub-MICs of antibiotic. The individual expression of ampC and mexX (circles), as well as boxplots showing medians, 25th and 75th percentiles, and minimum and maximum values, are shown. (A) ampC. Median expression of 0.82 (absence of meropenem) and 4.06 (presence of meropenem). (B) mexX. Median expression of 0.79 (absence of tobramycin) and 1.01 (presence of tobramycin). (C) Comparison of expression of ampC and mexX in the absence of antibiotics.

β-Lactam antibiotics increase the expression of ampC, and aminoglycoside substrates increase expression of the mexXY genes (21). To determine whether this was also the case for isolates in our study, gene expression was also measured for a subset of 14 isolates from CF patients grown with sub-MICs of meropenem or tobramycin (Fig. 2A and B). The presence of meropenem increased ampC expression by, on average, 5-fold across the 14 isolates (Wilcoxon signed-rank test, P = 0.001). There was also a strong intrastrain correlation between gene expression with and without meropenem (Spearman's correlation coefficient = 0.78, P < 0.001) (Fig. S2A). Likewise, the presence of tobramycin increased the expression of mexX by, on average, 1.3-fold (P = 0.048). The intrastrain correlation between gene expression with and without added tobramycin was very strong (Spearman's correlation coefficient = 0.90, P < 0.001) (Fig. S2B).

The expression of ampC and mexX was tested for correlation with the MIC of each strain for meropenem and tobramycin, respectively. The expression of ampC was correlated with the MIC for meropenem (Spearman's correlation coefficient = 0.49, P = 0.003 for bacteria grown in the absence of meropenem) (Fig. S3A), with the correlation being stronger for bacteria grown in the presence of meropenem (Spearman's correlation coefficient = 0.61, P = 0.020) (Fig. S3B). These findings are consistent with those of previous studies indicating that the expression of ampC is correlated with β-lactam resistance (22, 30, 39–41), although meropenem is usually not a good substrate for AmpC β-lactamase (42). It remains to be determined whether ampC gene expression in our samples correlates more strongly with resistance to antibiotics that are better substrates for the enzyme. The expression of mexX did not correlate with the MIC for tobramycin when the bacteria were grown in the absence of antibiotic (Spearman's correlation coefficient = 0.19, P = 0.287) (Fig. S3C) but had a non-statistically significant tendency when the bacteria were grown in the presence of tobramycin (Spearman's correlation coefficient = 0.51, P = 0.065) (Fig. S3D). A number of studies have demonstrated that the MexXYOprM efflux pump contributes to aminoglycoside resistance (reviewed in reference 38); for example, the tobramycin MIC of mutants of P. aeruginosa strain PAO1 lacking mexXY is 4-fold (43) or at least 8-fold (44) lower than that of isogenic mexXY-containing bacteria. However, resistance to tobramycin is multifactorial (14, 21), and a recent study indicates that some isolates of P. aeruginosa with high expression of the mexXYoprM efflux pump are nonetheless susceptible to tobramycin (45). It is likely that the effects of differences in the expression of mexX on tobramycin resistance are masked by other resistance mechanisms for the isolates of P. aeruginosa examined in this study.

Relationship between resistance gene expression in sputum and laboratory-grown bacteria.

The expression of ampC was lower in bacteria in the sputum samples than for the corresponding bacteria grown on laboratory agar without antibiotic or in the presence of meropenem, but the difference was not statistically significant (Wilcoxon signed-rank test, P = 0.052 and 0.062, respectively). However, there was some correlation between the expression of ampC in sputum samples and the expression of ampC by the corresponding P. aeruginosa isolates grown without antibiotics (Spearman's correlation coefficient = 0.37, P = 0.046) (Fig. 3A). These findings indicate a degree of relationship between gene expression in sputum and expression in the corresponding bacteria grown in culture, although the different physiology of P. aeruginosa during infection may involve lower expression of ampC. The expression of ampC in sputum samples was not significantly correlated with the expression of ampC in bacteria grown on meropenem-containing agar (Rho = 0.464, P = 0.151).

FIG 3.

Comparison of antibiotic resistance gene expression in bacteria in sputum and in cultured bacteria. (A) Expression of ampC in cultured bacteria grown in the absence of meropenem was correlated with expression in the corresponding sputum sample (n = 31, Spearman's correlation coefficient = 0.367, P = 0.046). (B) Expression of ampC in cultured bacteria grown in the presence of meropenem was not correlated with expression in the corresponding sputum sample (n = 11, Spearman's correlation coefficient = 0.464, P = 0.151). (C) Expression of mexX in cultured bacteria grown in the absence of tobramycin was not correlated with expression in the corresponding sputum sample (n = 31, Spearman's correlation coefficient = 0.007, P = 0.973). (D) Expression of mexX in cultured bacteria grown in the presence of tobramycin was not correlated with expression in the corresponding sputum sample (n = 11, Spearman's correlation coefficient = 0.264, P = 0.433).

Conversely, the expression of mexX was higher in sputum samples than when the corresponding bacteria were grown in the laboratory without antibiotics (Wilcoxon signed-rank test, P = 0.002). It was also higher than for bacteria grown on agar containing tobramycin, although the difference was not statistically significant (P = 0.328). The expression of mexX in bacteria in sputum did not correlate with expression of mexX by the corresponding P. aeruginosa grown on agar (Fig. 3C and D, Spearman's correlation coefficient, P = 0.811 and 0.433).

Gene expression of bacteria in sputum was also compared with the MICs of the corresponding P. aeruginosa isolates. There was no significant correlation between the expression of ampC in sputum and the MIC of isolated bacteria for meropenem (Spearman's correlation coefficient, P = 0.252), nor between the expression of mexX in sputum and the MICs of isolated bacteria for tobramycin (Spearman's correlation coefficient, P = 0.116).

DISCUSSION

In this research, we investigated the expression of two key P. aeruginosa antibiotic resistance genes, ampC and mexX, during infection in CF. Samples from patients at four different hospitals (three different cities) and with a wide range of ages were analyzed to represent a diversity of CF patients. Both ampC and mexX were expressed during infection, but there was a remarkably wide range of expression of both genes. The wide range of gene expression extended to bacteria isolated from sputum and grown in laboratory culture. In pairwise comparisons, the expression of ampC in bacteria in sputum was not significantly different from that in laboratory-grown bacteria; the expression of mexX was slightly higher in sputum than for bacteria grown on laboratory agar.

There was high variation in the expression of both ampC (over 20,000-fold) and mexX (2,000-fold) among isolates of P. aeruginosa grown on laboratory agar. Other researchers have also reported extensive variation in the expression of these genes among isolates of P. aeruginosa from a number of clinical settings (25, 27, 30, 39–41), reflecting the frequent occurrence of mutations that result in upregulation of ampC and mexX in clinical isolates of P. aeruginosa (22–31).

Meropenem increased the expression of ampC, and tobramycin increased expression of mexX in the isolates in our study. These findings are consistent with antibiotic-inducible expression of ampC and mexX in the type strain P. aeruginosa PAO1 (14, 38), although to the best of our knowledge, the effects of sub-MIC amounts of antibiotics on the expression of ampC and mexX in clinical isolates of P. aeruginosa have not been examined previously. There was also a correlation between the expression of ampC and that of mexX for the laboratory-grown bacteria. ampC and mexX are not known to share a regulatory mechanism. It is plausible that prolonged exposure to antibiotics during chronic CF infection selects for independently arising mutations that cause higher expression of the ampC and mexX genes.

This possibility is consistent with the strong correlation between resistances to different antibiotics (β-lactams, an aminoglycoside, and a fluoroquinolone) for the isolates tested, both within and between antibiotic classes (Table 2), something that has been observed previously (46, 47). Correlation, especially between different classes of antibiotics, could reflect patient treatment with prolonged administration of multiple antibiotics selecting for mutants with increased resistance to different antibiotic classes due to multiple independent resistance mechanisms. However, a correlation between resistance to different antibiotics may also result from shared resistance mechanisms, such as reduced access of multiple β-lactams to P. aeruginosa (14, 30), or enhanced activity of efflux pumps that export multiple antibiotics, including different antibiotic classes (14).

The expression of antibiotic resistance genes in sputum samples varied over 4,000-fold for ampC and approximately 100-fold for mexX. The occurrence of sputum samples with low levels of expression of ampC and/or mexX indicates that at least under some circumstances, even low levels of transcription are sufficient to allow survival in CF. The wide variation in gene expression in P. aeruginosa in laboratory culture indicates that strain-to-strain variation is at least part of the cause for the range of gene expression in sputum samples. The expression of ampC was similar in bacteria in sputum to that in bacteria grown in laboratory medium, and there was a modest correlation between ampC expression levels in the two environments. Conversely, the expression of mexX was higher in bacteria in sputum, and expression in laboratory culture did not show evidence of correlation with expression in sputum. These differences most likely reflect at least in part the difference in physiological states of P. aeruginosa bacteria in the two different environments. For example, gene expression may be affected by different growth rates of bacteria in the CF lung and in laboratory culture, by growth in complex biofilms during infection, by the presence of other bacteria, and by oxidative stress due to the action of the immune system. Gene expression in sputum could also be affected by changes in the lung environment during infection, such as the transition between stable infection and exacerbation, and, as antibiotics influence gene expression, differences in the antibiotic treatment received by each patient prior to sample collection. Furthermore, data from each sputum sample represent the average of the infecting P. aeruginosa population, which can have considerable genetic and phenotypic heterogeneity (11). The individual isolates tested in laboratory culture may therefore not be representative of the infecting population. Further research analyzing multiple samples and isolates from individual patients and determining whether there is a correlation between patient treatment or clinical status and resistance gene expression will be required to distinguish these possibilities.

Overall, our findings demonstrate a remarkably wide range of P. aeruginosa antibiotic resistance gene expression during infection in CF patients and between isolates of the bacteria. The reasons for this high patient-to-patient variability, which may influence the effectiveness of treatment, remain to be determined. The presence of antibiotics increased the expression of resistance genes of clinical isolates of P. aeruginosa in vitro and will be important to determine the extent to which antibiotic treatment affects resistance gene expression in infections. The differences in expression of ampC and mexX during infection and in laboratory-grown bacteria emphasize the different physiology of P. aeruginosa in the two environments and the importance of investigating the physiology of P. aeruginosa within the human host.

MATERIALS AND METHODS

Subjects and sampling.

Sputum samples were collected under the approval of the Southern Tasmanian Health and Medical Research Ethics Committee (H9813), the Queensland Children's Health Services (RCH) Human Research Ethics Committee (HREC/11/QRCH/163), the Metro North Hospital and Health Service Human Research Ethics Committee of The Prince Charles Hospital (HREC/08/QPCH/87), and the New Zealand Health and Disability Ethics Committees (NTY/10/12/106). Written informed consent was obtained for all study participants. Individuals with CF were recruited from outpatient clinics and inpatients being treated for pulmonary exacerbations. P. aeruginosa was isolated from samples of sputum using cetrimide agar (Oxoid) or by growth on nonselective agar followed by confirmatory PCR testing, as described previously (48). For isolation of RNA, sputum was expectorated directly into RNAlater (Qiagen) and stored at 4°C. A total of 36 sputum samples for RNA analysis were collected from CF patients in Dunedin (New Zealand), Brisbane (Queensland, Australia), and Hobart (Tasmania, Australia) (Table 1). Patients were 83% adults, with an age range of 11 to 58 years and median age of 24 years (interquartile range, 21 to 28 years). All of these patients except for three (those with isolates 002-A2, 003-A1, and 030-1) were chronically infected with P. aeruginosa. P. aeruginosa was analyzed from a further four samples from CF patients for which sputum for RNA analysis was unavailable. Five reference strains of P. aeruginosa from non-CF infections (Table 1) were included for comparison in some analyses. Expectorated sputum was obtained from 36 patients attending four different CF centers (Dunedin, Hobart, and two independent hospitals in Brisbane).

Antibiotic sensitivity testing.

Antibiotic sensitivity testing was carried out using the protocol of Weigand et al. (49). Overnight cultures of P. aeruginosa grown in King's B broth (50) were diluted to approximately 10⁶ CFU/ml, and 5-μl portions were inoculated as spots onto plates of Mueller-Hinton agar (Becton Dickinson) containing doubling dilutions of antibiotic. After the spots had dried, the plates were incubated at 37°C for 24 h. The MIC was defined through visual examination as the lowest concentration of antibiotic that inhibited the growth of the bacteria.

RT-qPCR.

RNA was extracted from sputum samples that had been collected into RNAlater, as described previously (51). RNA was also extracted from bacteria that had been spread onto Mueller-Hinton agar, supplemented as required with meropenem or tobramycin (one-third of the MIC), and incubated overnight at 37°C. Bacteria were collected into 1 ml of a 2:1 mixture of RNAprotect (Qiagen) and King's B broth. Following centrifugation, the bacterial cell pellet was resuspended in 200 μl of Tris-EDTA (TE) buffer containing lysozyme (1 mg/ml) and incubated for 5 min at room temperature to lyse the bacteria. RNA was then extracted as described above.

Aliquots of RNA were reverse transcribed to make cDNA, and transcripts were quantified by RT-qPCR, as described previously (51, 52). Gene-specific primers for P. aeruginosa ampC (5′-TCG CCT GAA GGC ACT GGT-3′ and 5′-GGT GGC GGT GAA GGT CTT G-3′), mexX (5′-GGC CCT GGT CGC CCT ATT C-3′ and 5′-TCC TCG TAC AGG CGA CGG-3′), and the reference genes clpX and oprL, established previously (51, 52), were used. The use of reference genes corrected for differences in bacterial load across different samples. There was a very strong correlation (Pearson correlation coefficient [R²] = 0.91) between the crossing point (C_p) values obtained with clpX and oprL with the sputum samples used in this study. Primer amplification efficiencies were determined using serial dilutions of genomic DNA templates from the reference strain PAO1 and the clinical isolates DUN002C, DUN005A, 002-A2, and 403-110; these were between 1.8 and 2.0 for all primer pairs with all templates. Quantitative PCR was carried out using SYBR green I master mix in conjunction with the LightCycler 480 platform (Roche). All reactions were carried out in duplicate. The presence of correct products was confirmed by melt curve analysis and by agarose gel electrophoresis, and the amplification efficiencies of each reaction were confirmed using LinRegPCR (53). The expression of antibiotic resistance genes was calculated relative to the geometric mean of clpX and oprL using the LightCycler software (51).

Statistical analyses.

Pairs of outcomes were examined for evidence of monotonic associations using Spearman's correlation coefficients (i.e., for values of one outcome showing a positive or negative, but not necessarily linear, association against the other outcome). Levels of expression were compared between pairs of outcomes within study participants using Wilcoxon matched-pairs signed-rank tests to determine if within-participant differences tended to be positive or negative. For these analyses, only participants with both outcomes measured were able to be included. Analyses were performed using Stata 15.1, and a two-sided P value of <0.05 was considered statistically significant in all cases. Some outcomes were unavailable for some participants, and all available data were used for each particular analysis.