Characterization of Hypermutator Pseudomonas aeruginosa Isolates from Patients with Cystic Fibrosis in Australia

Hypermutable Pseudomonas aeruginosa isolates (hypermutators) have been identified in patients with cystic fibrosis (CF) and are associated with reduced lung function. Hypermutators display a greatly increased mutation rate and an enhanced ability to become resistant to antibiotics during treatment.

ABSTRACT

Hypermutable Pseudomonas aeruginosa isolates (hypermutators) have been identified in patients with cystic fibrosis (CF) and are associated with reduced lung function. Hypermutators display a greatly increased mutation rate and an enhanced ability to become resistant to antibiotics during treatment. Their prevalence has been established among patients with CF, but it has not been determined for patients with CF in Australia. This study aimed to determine the prevalence of hypermutable P. aeruginosa isolates from adult patients with CF from a health care institution in Australia and to characterize the genetic diversity and antibiotic susceptibility of these isolates. A total of 59 P. aeruginosa clinical isolates from patients with CF were characterized. For all isolates, rifampin (RIF) mutation frequencies and susceptibility to a range of antibiotics were determined. Of the 59 isolates, 13 (22%) were hypermutable. Whole-genome sequences were determined for all hypermutable isolates. Core genome polymorphisms were used to assess genetic relatedness of the isolates, both to each other and to a sample of previously characterized P. aeruginosa strains. Phylogenetic analyses showed that the hypermutators were from divergent lineages and that hypermutator phenotype was mostly the result of mutations in mutL or, less commonly, in mutS. Hypermutable isolates also contained a range of mutations that are likely associated with adaptation of P. aeruginosa to the CF lung environment. Multidrug resistance was more prevalent in hypermutable than nonhypermutable isolates (38% versus 22%). This study revealed that hypermutable P. aeruginosa strains are common among isolates from patients with CF in Australia and are implicated in the emergence of antibiotic resistance.

INTRODUCTION

Pseudomonas aeruginosa frequently causes chronic pulmonary infections that are associated with increased morbidity and mortality among patients with suppurative lung diseases, especially those with cystic fibrosis (CF) (1, 2). Patients with CF have frequently been found to harbor hypermutable P. aeruginosa strains (3–6), a finding generally associated with poorer patient outcomes (7, 8). These hypermutators show an increased mutation rate of up to 1,000-fold compared to that of wild-type strains (6) and this, together with the short bacterial generation time, allows them to rapidly adapt to a variety of stressful environments (9). Hence, hypermutable P. aeruginosa strains can quickly adapt to antibiotic exposure (4), and these strains have been strongly linked with antibiotic resistance within patients with CF (3, 4, 6, 7).

The hypermutator phenotype results from the mutation of genes involved in DNA repair, especially genes involved in the mismatch repair (MMR) system (mutS, mutL, and uvrD) (10, 11). Mutations in the 7,8-dihydro-8-oxo-deoxyguanosine (8-oxodG or GO) system genes, mutM, mutY, and mutT, have also been found to produce the hypermutator phenotype in P. aeruginosa (12). P. aeruginosa cells show a variety of diverse phenotypic changes associated with adaptation to the CF lung, including conversion to a mucoid phenotype (13, 14), inactivation of quorum-sensing functions (15–17), motility loss (18, 19), and various auxotrophies (20). Common mutation-mediated antibiotic resistance mechanisms include the increased expression of Mex-efflux pumps following mutations in regulator genes (21–23), antibiotic-binding-site modifications (24), increased production of antibiotic inactivating β-lactamase enzymes (25), and inactivation of outer membrane porins that lead to cell membrane permeability changes and reduced intracellular drug accumulation (23). These phenotypic changes may occur more readily in hypermutator strains due to the increased mutation rates.

The prevalence of P. aeruginosa hypermutators in isolates from patients with CF has been examined across several clinics in Europe (15% to 54%; average, 27%) (4, 7, 8, 10, 26–29) and the Americas (17% to 42%; average, 29%) (9, 30, 31). However, the prevalence and genetic diversity of hypermutator isolates obtained from patients with CF in Australia have received much less attention. Previously, four Australian isolates were included in a European multicenter study; however, only mutation frequencies (MF) were determined for these isolates, and there was no description of the individual results (32). Another multinational study, which assessed eight P. aeruginosa isolates belonging to a single widespread clone (clonal complex 274 [CC274]), including three hypermutators obtained in 2007 to 2008 from different geographical locations in Australia, determined the phylogeny, interpatient dissemination, and hypermutator and resistance genotypes (33).

Antibiotic resistance in P. aeruginosa is a major problem in Australian CF centers (34). Given the importance of hypermutator P. aeruginosa strains for antibiotic resistance and poorer patient outcomes, we aimed to determine the prevalence of hypermutators among isolates obtained in 2013 from chronically infected adult patients with CF at a major health care center in Australia and to characterize and evaluate the cause of hypermutation. Each of the isolates was tested for susceptibility to a panel of antibiotics, and whole-genome sequencing (WGS) with comparative genomic and phylogenetic analyses was used to further characterize the hypermutators. This report describes for the first time the prevalence of hypermutable P. aeruginosa isolates from patients with CF in an Australian health care center and the comprehensive genotypic and phenotypic characterization of these isolates.

RESULTS AND DISCUSSION

Prevalence and resistance profiles of P. aeruginosa hypermutators in patients with CF from Australia.

Of a total of 59 consecutive P. aeruginosa isolates (designated CW1 to CW59) from 37 patients with CF (22 patients had both a mucoid and nonmucoid colony variant included from their sputum culture) who attended the Alfred Hospital in 2013, 13 isolates (22%) from 10 patients (27%) were found to be hypermutable (Table 1). Of the mucoid and nonmucoid colony variant pairs from 22 patients, in three patients both were hypermutators, in six patients only one of them was a hypermutator, and in 13 of the patients both were nonhypermutators. All of the isolates defined as hypermutable showed at least a 48-fold increase in MF compared to that of PAO1, while the highest MF in the nonhypermutator isolates was 8-fold that of PAO1. Furthermore, analysis of the total number of rifampin (RIF)-resistant mutants per 10⁸ bacterial cells (Fig. 1) showed that the definition of hypermutation (rifampin MF ≥ 20-fold higher than that for PAO1 [4]) allows a very clear distinction between hypermutable and nonhypermutable isolates. Of the 13 hypermutable isolates, 9 showed more than a 100-fold MF increase, with CW35 showing the highest MF of 470-fold that of PAO1. Our results agreed very well with those for patients with CF in Spain, where on average 21% (range 15% to 29%) of isolates were hypermutable and 30% (24% to 37%) of patients harbored hypermutable strains (4, 26, 27).

TABLE 1.

The bacterial characteristics of 59 clinical P. aeruginosa isolates from patients with cystic fibrosis

Isolatea	Patient no.	Mucoid phenotypeb	RIF MF (fold change)c	MIC for:
				ATMd,e	CAZd,f	CIPd,e	MEMd,e	TOBd,e
CW1	1	N	0.8	4	>256	0.75	8	12
CW2	2	Mi	0.4	0.016	0.032	0.047	0.016	2
CW3	3	M	6.2	0.125	0.125	0.25	0.25	0.75
CW4	4	N	1.5	8	2	1.5	0.25	6
CW5	5	N	285.2	6	>256	1.5	>32	3
CW6	5	M	0.5	24	>256	4	12	3
CW7	6	Mi	1.2	1.5	2	2	2	0.032
CW8	6	N	141.0	1.5	0.75	4	1.5	0.75
CW9	7	N	1.0	16	4	0.25	0.19	0.75
CW10	7	Mi	1.0	6	>256	4	0.125	0.75
CW11	8	M	2.3	8	12	0.094	0.75	3
CW12	9	M	2.2	0.19	0.5	0.75	0.125	1
CW13	10	N	180.5	0.5	24	4	8	1.5
CW14	10	M	0.6	0.047	0.25	1.5	0.125	1
CW15	11	M	4.4	4	1.5	2	0.094	1.5
CW16	12	M	4.8	0.125	0.75	0.75	0.125	3
CW17	13	N	7.5	0.5	4	4	4	3
CW18	13	M	3.2	0.19	1	1.5	0.064	1
CW19	14	M	76.9	1	16	0.5	0.75	1
CW20	14	N	0.7	8	>256	0.25	16	8
CW21	15	M	2.7	0.125	0.5	2	0.5	1.5
CW22	15	N	0.5	12	16	1	>32	32
CW23	16	M	2.6	0.064	0.25	0.5	0.032	0.38
CW24	16	N	1.0	0.5	1	0.5	0.047	2
CW25	17	M	1.9	0.032	0.125	0.75	1.5	1
CW26	18	N	1.4	0.75	2	1.5	>32	24
CW27	18	M	1.9	0.19	0.38	2	2	3
CW28	19	N	125.9	12	>256	3	>32	3
CW29	19	M	1.0	0.75	2	0.19	4	2
CW30	20	N	84.9	0.032	0.064	1	0.016	0.75
CW31	20	Mi	143.9	0.064	0.094	0.094	0.004	1
CW32	21	N	2.8	16	12	1.5	>32	2
CW33	21	M	6.2	0.19	0.75	0.75	0.125	3
CW34	22	N	240.0	0.38	1	4	0.5	8
CW35	22	Mi	472.5	0.19	1.5	3	0.5	8
CW36	23	N	6.5	0.75	4	1.5	0.75	>256
CW37	23	M	2.3	0.5	0.38	0.5	0.125	>256
CW38	24	M	1.7	0.75	0.75	0.5	0.25	3
CW39	24	N	1.1	4	1	2	0.5	16
CW40	25	N	2.1	1	0.75	>32	0.64	24
CW41	26	Mi	48.4	0.5	0.75	4	0.75	16
CW42	26	N	54.5	2	>256	6	>32	48
CW43	27	M	0.9	0.19	0.5	0.25	1	3
CW44	28	Mi	122.6	1.5	2	0.19	0.5	1
CW45	29	N	232.5	16	>256	1.5	>32	8
CW46	29	M	0.8	3	8	1.5	2	2.5
CW47	30	Mi	5.4	0.094	0.125	0.25	0.023	4
CW48	30	M	4.4	0.023	0.125	0.5	0.23	>256
CW49	31	N	3.1	0.19	0.75	0.094	0.32	4
CW50	31	M	3.0	0.19	0.75	0.125	0.125	3
CW51	32	Mi	4.2	0.19	0.5	0.5	0.032	2
CW52	33	Mi	2.7	0.25	0.5	0.75	0.032	1.5
CW53	34	M	7.8	0.38	2	4	0.19	1.5
CW54	34	N	0.4	0.5	2	8	3	1.5
CW55	35	N	0.7	32	>256	6	0.125	1.5
CW56	36	M	5.1	0.38	0.5	4	0.125	3
CW57	36	N	1.7	12	0.75	4	0.38	1
CW58	37	Mi	2.0	0.125	0.75	1	1.5	2
CW59	37	N	1.8	0.125	0.75	1	1.5	2

^aHypermutator strains are indicated in bold.

^bMucoid phenotype; M, mucoid; N, nonmucoid; Mⁱ, nonmucoid revertants, i.e., mucoid isolates that reverted to nonmucoid. There were significantly more nonmucoid revertants from mucoid phenotype for hypermutators than for nonhypermutators (one-tailed Mann-Whitney U test, P = 0.004).

^cMutation frequencies (MF) were determined as the fraction of the resistant population quantified on rifampin (300 mg/liter)-containing CAMHA plates compared to the total population quantified on drug-free CAMHA plates. Hypermutation was defined by the rifampin MF being ≥20-fold higher than that of PAO1.

^dMICs were interpreted based on EUCAST susceptibility breakpoints (56). MICs indicative of resistance are underlined and MICs indicative of intermediate resistance are shown in italics. Antibiotics tested: ATM, aztreonam; CAZ, ceftazidime; CIP, ciprofloxacin; MEM, meropenem; TOB, tobramycin.

^eHypermutators were not significantly more resistant (Mann-Whitney U test, P > 0.05).

^fHypermutators were significantly more ceftazidime resistant (one-tailed Mann-Whitney U test, P = 0.021).

FIG 1.

The number of isolates for each range of rifampin-resistant mutants per 10⁸ bacterial cells, determined for each of the 59 clinical isolates. PAO1 showed 2.5 mutants/10⁸ cells and PAOΔmutS showed 2,631.2 mutants/10⁸ cells.

The Etest MIC values (Table 1) were used to determine multidrug resistance (MDR; nonsusceptibility to ≥1 agent from ≥3 different antimicrobial categories [35]). The percentages of nonsusceptible isolates in hypermutable compared to nonhypermutable strains, respectively, were 77% versus 65% for ciprofloxacin (CIP), 46% versus 17% for ceftazidime (CAZ), 46% versus 33% for aztreonam (ATM), 38% versus 20% for meropenem (MEM), and 38% versus 22% for tobramycin (TOB). These results are in agreement with previous studies that found greater resistance to antimicrobial agents for hypermutators compared to nonhypermutators (4, 31). However, in the present study, only reduced susceptibility to ceftazidime was found to be significantly greater for hypermutators compared to nonhypermutators (one-tailed Mann-Whitney U test, P = 0.021). Multidrug-resistant strains were more abundant among hypermutators (38%) than nonhypermutators (22%), in agreement with a previous study (7), although this difference did not reach statistical significance (one-tailed Mann-Whitney U test, P = 0.096).

Phylogenomic analysis.

The evolutionary relatedness of the 13 hypermutable clinical isolates, as determined using draft WGS, compared with 63 publicly available P. aeruginosa genomes (including those of PAO1 and ATCC 27853 wild-type strains) is displayed in Fig. S1. A maximum likelihood phylogeny was determined by comparison of all core genome single-nucleotide polymorphisms (cgSNPs), with PAO1 used as the reference genome. The 13 hypermutable isolates were spread widely across the phylogenetic tree, indicating that they arose from diverse P. aeruginosa strain backgrounds. Of the three paired hypermutator isolates (CW30 and CW31, CW34 and CW35, and CW41 and CW42), each of the individual isolates of a pair clustered closely together, indicating that they are likely to be true isogenic derivatives. Indeed, the average number of cgSNPs between all of the nonpaired hypermutator isolates was 24,641, while the average number of cgSNPs between the paired isolates was more than 30-fold less (774). However, the three hypermutator pairs were not closely related, indicating that they each arose from different backgrounds. CW8 and CW13 were found to be clustered closely with an Australian epidemic strain, AES-1R_NZ (36). CW41 and CW42 clustered closely with the Denmark epidemic strain DK2 (37, 38). Similarities with P. aeruginosa genomes from different environments around the world demonstrated diversity between these isolates (Fig. S1).

Mutations responsible for the hypermutator phenotype.

Genes likely to be involved in the hypermutator phenotypes are presented in Table S1. Complementation assays confirmed that for 9 of the 13 hypermutable isolates, the cause of hypermutation was linked to defects in either mutL (6 isolates) or mutS (3 isolates) (Table 2). Six of the isolates had inactivating mutations within mutL (CW8, CW19, and CW45 had frameshift mutations, CW30 and CW31 had an identical nonsense mutation, and CW44 had a large deletion), one isolate (CW5) had a frameshift mutation within mutS, and isolates CW34 and CW35 harbored two previously observed mutations that lead to missense substitutions in MutS (C224R and T287P) (33). Conversely, mutations within mutL in isolates CW41 and CW42 were confirmed to not be involved in the hypermutator phenotype, although the mutation at nucleotide (nt) 247 results in a K83E substitution in the ATPase domain (ATPlid) of MutL (39). Finally, several attempts to electroporate the plasmids for the complementation assays into isolate CW13 failed, and thus the effect of the A470V substitution within MutL could not be assessed.

TABLE 2.

Genetic analysis of mutS (PA3620) and mutL (PA4946) in the 13 hypermutator isolates

Isolate	DNA sequence changea or predicted amino acid changes resulting from mutations inc:		Complementation withb:
	mutS	mutL	mutS	mutL
CW5	nt1312Δ13	A213V	+	−
CW8		nt700Δ4	−	+
CW13		A470V	ND	ND
CW19		nt1444ins1	−	+
CW28			−	−
CW30		Q341X	−	+
CW31		Q341X	−	+
CW34	C224R, T287P		+	−
CW35	C224R, T287P		+	−
CW41	P249L	K83E	−	−
CW42		K83E	−	−
CW44		nt691Δ378	−	+
CW45		nt565Δ1	−	+

^aSequence variations from the reference strain PAO1.

^bND, not determined.

^cInactivating mutations are shown in bold. nt, nucleotide; X, early STOP codon; Δ, nucleotide deletion; ins, nucleotide insertion.

The observation that mutations in mutL were the most common cause of hypermutability is in agreement with another study, where 60% of hypermutable isolates were attributed to mutations in mutL and the remaining 40% to mutS (31). However, other studies have found that hypermutation was more commonly due to a defective mutS gene (6, 11). Interestingly, our mutS-associated hypermutator isolates (CW5, CW34, and CW35) had the highest rifampin MF.

Apart from mutations within the mutS and mutL genes, several unique mutations leading to missense amino acid changes were documented in other mutator-associated genes (shown in Table S1, online supplement) that may also contribute to the hypermutator phenotype. The mechanism of hypermutation in paired weak hypermutators CW41 and CW42 may be linked to mutations within these other mutator genes, while in CW28 it may be due to mutations in unknown genes that are related to a strong hypermutator phenotype. Further studies are required to determine the definite cause for hypermutation in these strains.

Mutations in genes associated with the mucoid phenotype and quorum sensing.

The overview of mutations in genes involved in the mucoid phenotype and quorum sensing is shown in Table S1. Mucoidy is a common trait of P. aeruginosa isolates obtained from respiratory infections in patients with CF and occurs when there is increased alginate biosynthesis (28, 31, 40). This can result from inactivating mutations in the mucA gene, encoding the anti-sigma factor (anti-σ²²), increased expression of the sigma factor (σ²²) algU (or algT), or various mutations within the multicomponent signal transduction genes (algR or algP), which give rise to increased alginate production through the activation of the algD promoter (14).

Within the whole CF collection, 34 isolates (58%) exhibited a mucoid phenotype, and 10 of these displayed nonmucoid reversion (Table 1). Nonmucoid revertants are mucoid isolates that revert to the nonmucoid phenotype after acquiring additional mutations, likely in genes associated with alginate regulation and synthesis, along with the selective pressure for nonmucoid cells in aerobic, stationary cultures, like those experienced after isolation (41). Five of the hypermutators (38%) were found to show a mucoid phenotype, with four of these being nonmucoid revertants. In contrast, 29 nonhypermutable isolates (63%) exhibited a mucoid phenotype, with only six reverting to the nonmucoid phenotype (Table 1). The nonmucoid revertants were significantly more common for hypermutators than for nonhypermutators (one-tailed Mann-Whitney U test, P = 0.004). This relationship between hypermutation and nonmucoid reversion has been observed previously, where only hypermutators exhibited this reversion (42). It is likely that this reversion occurs more easily in hypermutators due to the increased chance of a spontaneous mutation arising in the mucoid-associated genes. The mucoid, hypermutable CW19 strain had a nonsense mutation in mucA (encoding the anti-sigma factor) and a large deletion in algP, which encodes the alginate transcriptional regulator (Table S1). Nonmucoid and nonmucoid revertant hypermutator isolates also contained inactivating mutations in a range of mucoid phenotype-related genes. Indeed, over half (54%) of the hypermutable isolates had frameshift or nonsense mutations in mucA. Of note, the nonmucoid revertant CW35 strain had no mutation in mucA, suggesting that mutations in other alginate biosynthesis or regulatory genes may have resulted in the original mucoid conversion and subsequent reversion. CW5, CW30, and CW45 showed a defective mucA gene with no observable mucoid conversion displayed, suggesting that these isolates underwent further adaptive mutations in the lungs of patients, making them nonmucoid revertants prior to their collection.

Four hypermutable isolates (CW30, CW31, CW41, and CW42) had no identifiable lasR gene (Table S1), which encodes the major quorum-sensing regulator (16). Three hypermutable isolates had inactivating mutations in lasR; CW28 had a frameshift mutation, while CW5 and CW45 had an identical nonsense mutation. CW19, CW34, CW35, and CW44 all had missense mutations in the section of lasR that encodes the DNA-binding domain, with the paired isolates CW34 and CW35 displaying an identical missense mutation. All of these mutations likely resulted in nonfunctional or highly reduced quorum-sensing activity (43). There has been discordance in the literature over whether the mucoid phenotype and quorum sensing are associated with a hypermutator phenotype (8, 28, 31). One of these studies found that mutations in mucA and lasR occurred prior to the acquisition of the hypermutator phenotype of P. aeruginosa isolates from patients with CF (28). A limitation of the present study is that we cannot determine the order of appearance of the mutations in our isolates to comment on whether the acquisition of the hypermutator phenotype or the mutations came first.

Mutations involved in antibiotic resistance.

We analyzed the whole-genome sequences of the hypermutable strains for mutations associated with specific antibiotic resistance mechanisms (Fig. 2). The genes encoding structural components and regulators of the four main efflux systems of P. aeruginosa were examined at a genetic level (Table S1). Mutations were found in the efflux regulator genes mexR, mexT, mexZ, and nfxB in most of the hypermutator strains, which may result in increased expression of MexAB-OprM, MexEF-OprN, MexXY-OprM, and MexCD-OprJ efflux pumps, respectively (44). Our study showed that all isolates with increased tobramycin MIC values had inactivating mutations within the MexXY-OprM repressor, mexZ. This was similar to results of a previous study, which found that 82% of multidrug-resistant P. aeruginosa isolates overproduced MexXY-OprM (45). Additionally, CW19 had a nonsense mutation early in mexC that may have contributed to the susceptible phenotype to ciprofloxacin despite having an inactivating mutation in nfxB (negative regulator of the efflux pump MexCD-OprJ). CW44 had a one-nucleotide deletion in mexX, causing a frameshift that likely contributed to the measured low MIC to tobramycin. CW34 and CW35 had an identical deletion of one nucleotide in mexB, which resulted in a frameshift that likely contributed to the susceptibility to β-lactam antibiotics. However, expression levels of efflux pump components cannot be ascertained from this genomic data. Ultimately, transcriptomic analysis is required to elucidate the expression levels and full effect of these mutations.

FIG 2.

The mutations in the resistome (resistance-associated genes) examined via whole-genome sequencing and bioinformatics analysis for the 13 hypermutable P. aeruginosa clinical isolates from patients with cystic fibrosis and the well-characterized PAO1 wild-type strain. The strongest hypermutators for high rifampin mutation frequency (RIF MF) are shown in the darkest purple.

Resistance to β-lactams often results from the overproduction of the chromosomally encoded cephalosporinase AmpC, which results from mutations in the peptidoglycan-recycling genes, ampD, dacB, and ampR (46, 47). In two isolates, frameshift mutations were identified in ampR, encoding a regulator, and one isolate had a nonsense mutation in ampD (Table S1). Another gene that is involved in the recycling of cell wall components and has been found to increase expression of AmpC when mutated is mpl, which encodes UDP-N-acetylmuramate:L-alanyl-γ-d-glutamyl-meso-diaminopimelate ligase (48). Five isolates contained frameshift mutations in mpl, but reduced β-lactam susceptibilities did not always correspond with the presence of these mutations; validation through β-lactamase assays would be able to provide additional insight.

A link was apparent between reduced aztreonam susceptibility and the presence of missense mutations in ftsl, which encodes penicillin-binding protein 3 (PBP3); two ftsl mutations leading to the amino acid substitutions R504C and F533L in PBP3 have been previously described (49). We also examined the gene oprD, whose altered expression is responsible for reduced susceptibility to carbapenems (50). Eight of the isolates were found to have inactivating mutations in oprD, with four being found in strains with meropenem resistance (Table 1 and Table S1). One meropenem-intermediate resistant isolate also had a missense mutation in OprD that may have contributed to reduced meropenem susceptibility. The frameshift mutations in oprD observed for a number of hypermutators did not always lead to greatly reduced susceptibility; however, it has been suggested that reduced susceptibility requires multiple resistance mechanisms (45).

Fluoroquinolone resistance is usually associated with missense mutations in gyrA, gyrB, parC, and parE, encoding fluoroquinolone targets (51, 52). Ten hypermutators were found to be ciprofloxacin resistant (Table 1); this was likely due to previously described mutations leading to missense substitutions in GyrA at residues 83 and 87 for most of the isolates (53, 54), while three hypermutator strains had mutations in gyrB, leading to previously described substitutions (Q467R and E468D) that are likely responsible for resistance (33, 54, 55). Aminoglycoside resistance can occur with missense mutations in elongation factor G (fusA1); previously described mutations in fusA1 (resulting in V93A, K504E, Y552C, and T671I substitutions) were found in all tobramycin-resistant hypermutators (Table S1) (33).

In conclusion, this study has for the first time determined the prevalence of P. aeruginosa hypermutators in adult patients with CF from a health care center in Australia and established that the majority of the hypermutator isolates come from divergent lineages, with a small number of very closely related pairs among them. Furthermore, mutations in mutL and mutS were determined to be the cause of the hypermutator phenotype in the majority of these isolates. Hypermutable isolates had a higher proportion with MDR and were more often resistant to each of the tested antibiotics in comparison to nonhypermutable isolates. These findings indicate that hypermutable P. aeruginosa isolates are playing an important role in the antibiotic resistance problem in patients with CF in Australia.

MATERIALS AND METHODS

Bacterial strains and media.

We characterized 59 P. aeruginosa clinical isolates collected from the sputa of 37 consecutive patients with CF (in 2013) (Table 1). For 22 patients, a mucoid and nonmucoid colony variant was present from the same culture and one of each was selected for analysis. The P. aeruginosa PAO1 WT strain and the hypermutable PAOΔmutS strain (mutS knock out of PAO1) (9) were used as controls with these clinical isolates. All susceptibility studies and viable counting were performed on cation-adjusted Mueller-Hinton agar (CAMHA; containing 25 mg/liter Ca²⁺ and 12.5 mg/liter Mg²⁺; Medium Preparation Unit, The University of Melbourne, Parkville, Australia). Drug-containing agar plates were prepared using CAMHA (BD, Sparks, MD) supplemented with the appropriate amount of antibiotic. The rifampin (Sigma-Aldrich, Sydney, Australia) stock solution was prepared in dimethyl sulfoxide (DMSO; Sigma-Aldrich, Sydney, Australia) and subsequently filter sterilized using a 0.22-μm polyvinylidene difluoride (PVDF) syringe filter (Merck Millipore, Cork, Ireland). The mucoid or nonmucoid phenotype of growth exhibited after 24 h of incubation at 37°C on antibiotic-free CAMHA of the 59 clinical isolates was recorded.

Mutation frequencies.

Rifampin (300 mg/liter) MF were determined as previously described (4). This determination was carried out in at least 4 experiments, and the average was used as the MF.

Antibiotic susceptibility.

The MICs of ceftazidime (test range, 0.016 to 256 mg/liter), ciprofloxacin (0.002 to 32 mg/liter), meropenem (0.002 to 32 mg/liter), aztreonam (0.016 to 256 mg/liter), and tobramycin (0.016 to 256 mg/liter) for each clinical isolate were determined using Etest strips (bioMérieux, North Ryde, Australia). The manufacturer’s protocol was followed with alterations as described previously (27), and readings were taken at 20, 24, 36, and 48 h for comparison. The MICs at 24 h were evaluated based on EUCAST susceptibility breakpoints (56) to determine which isolates had MDR (resistance to ≥1 agent from ≥3 different antimicrobial categories [35]).

Genetic basis for hypermutation.

Whole-genome sequencing was performed on all 13 hypermutators. Genomic DNA (gDNA) was purified from each hypermutable clinical isolate using the GenElute bacterial genomic DNA kit (NA2110-IKT; Sigma-Aldrich, Castle Hill, NSW, Australia) according to the manufacturer’s protocol, except that 40 μl of sterile distilled water was used instead of the elution solution. The gDNA was sequenced utilizing the paired-end 150-bp protocol on an Illumina NextSeq instrument at Micromon Next-Gen Sequencing Facility (Monash University, Clayton, VIC, Australia).

De novo assembly of the raw read data (average read depth of 39 to 51×) was performed using CLC Genomics Workbench v7.03 software. Automatic annotations of the de novo assemblies were produced using Rapid Annotations using Subsystem Technology (RAST) version2, accessed November 2016 (57).

To further explore the genetic basis of hypermutation, complementation studies were performed. Briefly, plasmids pUCPMS and pUCPML, harboring PAO1 wild-type mutS and mutL genes, and plasmid PUCP24, a control cloning vector, were used to transform the mutator isolates. Complementation was demonstrated by reversion of the increased MF for rifampin resistance in two independent transformant colonies for each hypermutator isolate.

Phylogenetic analysis.

The comparative phylogenetic analysis was performed on the de novo assembled genomes together with 63 publicly available P. aeruginosa genomes obtained from Pathosystems Resource Integration Centre (PATRIC) (58). The phylogenetic tree to determine evolutionary relatedness was created using Parsnp, a fast core genome multialigner software (Harvest) (59). Furthermore, an SNP analysis was performed via Nullarbor v1.01 (60).

Comparative genomic analysis.

The paired reads were reference assembled against PAO1 using CLC Genomics Workbench. The nucleic acid sequences were extracted for the MMR and other mutator-associated genes mutS, mutL, mutT, mutY, mutM, uvrD (mutU), radA, pfpI, sodM, oxyR, polA, ung, dnaQ (mutD), tRNAGly (PA2583, PA2819.1, and PA2819.2), mfd, and sodB; the mucoid phenotype genes mucA, algP, algU, and algD; the quorum sensing gene, lasR; and resistance-associated genes gyrA, gyrB, parC, parE, fusA1, galU, ampR, ampC, ampD, ftsI (PBP3), dacB, dacC, mpl, oprD, rpoB, phoP, phoQ, pmrA, pmrB, pagL, glpT, mexR, armR, mexA, mexB, nalC, nalD, mexZ, armZ, mexX, mexY, oprM, parR, parS, amgR, amgS, nfxB, mexC, mexD, oprJ, mexT, mexS, mexE, mexF, oprN, and oprF. These nucleic acid sequences were converted to amino acid sequences using the ExPasy online translational tool on the Bioinformatics Research Portal (Swiss Institute of Bioinformatics); subsequently, the amino acid sequences from each isolate were compared to that of PAO1 using the Clustal Omega multiple sequence alignment program (61–63). DNA sequence changes or predicted amino acid changes resulting from mutations were evaluated as evidence for loss of function.

Statistics.

A Mann-Whitney U test was used for the comparison of resistance rates, distributions of MICs, and mucoid reversion between hypermutators and nonhypermutators. A P value of <0.05 was considered statistically significant. Where the sample size was less than 30, a Fisher exact test was used to refine the P value.

Data availability.

The Fastq files for each strain have been deposited in GenBank under BioProject accession number PRJNA510693.