sign in sign up
<<
Home , Evolution of bacteria

Evolutionary genomics of epidemic and nonepidemic strains of Pseudomonas aeruginosa

Jeremy R. Dettmana,b,1, Nicolas Rodriguec, Shawn D. Aarond, and Rees Kassena,b

aDepartment of Biology, bCentre for Advanced Research in Environmental Genomics, and dOttawa Hospital Research Institute, University of Ottawa, Ottawa, ON, Canada K1N 6N5; and cDepartment of Mathematics and Statistics, University of Calgary, Calgary, AB, Canada T2N 1N4

Edited by Ralph R. Isberg, Howard Hughes Medical Institute and Tufts University School of Medicine, Boston, MA, and approved November 8, 2013 (received for review April 26, 2013) Pseudomonas aeruginosa is an opportunistic of virulence factor expression (6), presumably an to and is a major cause of morbidity and mortality in patients with escape detection by the host , and increased ac- cystic fibrosis (CF). Prolonged infection of the respiratory tract can tivity of efflux pumps (7) associated with resistance to many lead to adaptation of the pathogen to the CF lung environment. antibiotics used in treatment. To examine the general patterns of adaptation associated with The genetic changes responsible for these have chronic infection, we obtained genome sequences from a collection recently begun to be investigated. Evidence from longitudinal of P. aeruginosa isolated from airways of patients with CF. Our studies suggests that a large number of candidate genes can ac- analyses support a nonclonal epidemic population structure, with crue substitutions during adaptation of P. aeruginosa (6–9). Any a background of unique, recombining genotypes, and the rare given infection involves only a small subset of those substitutions, occurrence of successful epidemic clones. We present unique ge- and most virulence-related phenotypes are not observed uni- nome sequence evidence for the intercontinental spread of an formly (10), making it difficult to say with confidence which ge- epidemic strain shared between CF clinics in the United netic changes are most commonly associated with adaptation to and North America. Analyses of core and accessory genomes identified candidate genes and important functional pathways the CF lung. A more comprehensive view therefore requires the associated with adaptive . Many genes of interest were comparison of a larger sample of diverse clinical isolates. To this involved in biological functions with obvious roles in this pathos- end, we obtained whole-genome sequence data from a collection ystem, such as biofilm formation, antibiotic , pathogene- of P. aeruginosa isolated from the airways of patients with CF sis, transport, reduction/oxidation, and secretion. Key factors driving to investigate general patterns of adaptation associated with the adaptive evolution of this pathogen within the host appear to chronic infection. We summarize the broad-scale patterns of be the presence of oxidative stressors and antibiotics. Regions of genomic variation and identify the most likely genetic targets of the accessory genome unique to the epidemic strain were enriched adaptation causing contemporary infection in our isolates. Our for genes in transporter families that efflux heavy metals and anti- collection includes nonepidemic strains and a transmissible, ep- biotics. The epidemic strain was significantly more resistant than idemic strain recently reported within North America (1). In- nonepidemic strains to three different antibiotics. Multiple lines of fection with an epidemic strain confers a poorer prognosis than evidence suggest that selection imposed by the CF lung environ- infection with a nonepidemic strain, such as reduced lung func- fl ment has a major in uence on genomic evolution and the genetic tion (11) and increased risk of or lung transplant (1). Our characteristics of P. aeruginosa isolates causing contemporary comparative analyses of genomic variation identified some unique infection. Significance virulence | multidrug resistance | redox balance

The bacterium Pseudomonas aeruginosa is an opportunistic Pseudomonas aer- he ubiquitous, Gram-negative bacterium pathogen of humans and is the leading cause of death in uginosa T can be found free-living in a wide range of environ- patients with cystic fibrosis (CF). We sequenced the genomes of ments or in association with various animal and species. P. aeruginosa isolated from respiratory tracts of patients with P. aeruginosa is also an opportunistic pathogen of humans and is CF to investigate general patterns of adaptation associated the most common species isolated from the respiratory tracts of with chronic infection. Selection imposed by the CF lung envi- fi adult patients with cystic brosis (CF). Chronic endobronchial ronment has had a major influence on genomic evolution and infections caused by P. aeruginosa are found in up to 80% of the genetic characteristics of isolates causing contemporary adult patients with CF (1, 2) and are associated with increased infection. Many of the genes and pathways implicated in morbidity and mortality (3). Once established in the lung, adaptive evolution within the host had obvious roles in the P. aeruginosa infections are notoriously difficult to eradicate and pathogenic lifestyle of this . Genome sequence data are highly resilient to repeated rounds of intensive, prolonged indicated that an epidemic strain, with increased virulence and antibiotic treatment. multidrug resistance, has spread between clinics in the United There is growing evidence that the often decades-long in- Kingdom and North America. teraction between the host and P. aeruginosa during chronic in- fection results in adaptation of the pathogen to the CF lung Author contributions: J.R.D., N.R., S.D.A., and R.K. designed research; J.R.D. and N.R. environment (4). The CF lung is an inhospitable place charac- performed research; J.R.D., N.R., and R.K. analyzed data; and J.R.D., N.R., and R.K. wrote the paper. terized by osmotic and oxidative stress, limited , high The authors declare no conflict of interest. concentrations of antibiotics, and constant assault by the host immune system. Although the hurdles posed by these factors are This article is a PNAS Direct Submission. fl Data deposition: The draft genome sequences reported in this paper have been deposited dynamic and uctuate through time, they are likely the main in the DNA DataBank of Japan (DDBJ)/European Molecular Biology Laboratory/GenBank drivers of adaptive evolution of bacteria in CF airways. Studies database, www.ncbi.nlm.nih.gov (accession nos. AWYJ00000000–AWZG00000000). have documented some repeatable patterns of adaptation to the 1To whom correspondence should be addressed. E-mail: [email protected]. EVOLUTION CF lung environment such as loss of associated with This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. growth as an unattached biofilm or microcolony (5), reduced 1073/pnas.1307862110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1307862110 PNAS | December 24, 2013 | vol. 110 | no. 52 | 21065–21070 Downloaded by guest on September 24, 2021 genomic features that distinguish the epidemic strain from Phylogeny of Isolates. Our sampling scheme was designed to allow nonepidemic strains. detailed analyses of our focal population (Ontario) and to place these isolates within the broader phylogenetic framework of the Results and Discussion global species. The established P. aeruginosa multilocus sequence Isolate Sources, Genome Sequencing, and Alignment. We take a typing (MLST) (16) system was applied to the 24 Ontario iso- population genomics approach to study the evolution of the op- lates; a total of 14 sequence types were identified, half of which portunistic pathogen, P. aeruginosa. Our study is unique in using were previously undescribed (SI Appendix, Table S3). The MLST whole-genome sequence data to examine the cross-sectional ge- system covers only 0.18% of the genome and the loci have low nome diversity of P. aeruginosa at the population and intrastrain levels of nucleotide variation, providing limited power for inferring level. Isolates were sampled from the sputum of 24 different phylogenetic relationships. Instead, we constructed a maximum patients attending CF clinics in six cities in Ontario, Canada (ref. likelihood phylogeny from the core genome alignment (Fig. 2 and 1, SI Appendix,TableS1). Clinics were located within 520 km of SI Appendix Fig. S1). One notable feature of this phylogeny is the each other, so this sample represents a geographically restricted two well-supported (A and B) that correspond to trans- population. Reference-guided mapping may skew the detection missible, epidemic strains of P. aeruginosa (see next section), of variation in genomic content, so draft genomes were assembled a designation supported by the tight phylogenetic clustering of de novo (average of 69-fold sequence coverage) from whole- isolates within clades. The core-genome nucleotide diversity genome sequence data for the 24 Ontario isolates. To place the within clades A and B is only 2.5% and 3.0%, respectively, of Ontario population in the evolutionary context of other members that in the entire sample. of the species, eight previously sequenced genomes were also Another feature of the phylogeny is long terminal branches included in our analyses (four from non-CF sources; SI Appendix, with a backbone of short internodes (Fig. 2 and SI Appendix, Fig. Table S2). A 32-genome global alignment was produced and lo- S1). Despite using genome-wide data to construct cally collinear blocks of conserved sequence were identified. the phylogeny, many internodes remain weakly supported. To fl Genomic regions possessed by all 32 isolates were classified as investigate if this poor resolution was due to con icting phylo- core genome, and those occurring in only some isolates were genetic signals caused by recombination, recombinant regions of fi the core genome were identified (17) and removed in an isolate- classi ed as accessory genome. The size of the core genome was fi 5.35 Mb, or 83.5% of the average full P. aeruginosa genome speci c manner. The maximum likelihood phylogeny built from (6.41 Mb). When restricted to genes only, 74.6% were consid- this nonrecombinant core genome still had short, weakly sup- ered core. Summing all nonredundant, accessory regions gives ported internodes (SI Appendix, Fig. S1), indicating that recom- a total of 6.89 Mb, but each isolate contains an average of only bination is not solely responsible for the poor resolution of the 1.06 Mb (15.4%) of this accessory genome pool. basal relationships. An alternate explanation is rapid radiation and diversification Nucleotide Polymorphism. The core genome alignment was 5.06 among clades during adaptation to the CF lung environment. All Mb in length and contained 137,941 (2.7%) variable nucleotide but one of the ingroup isolates (PA01) originated from a patient positions. Core nucleotide diversity of the Ontario population with CF, in comparison to the well-supported outgroup − (π = 0.0041, SE = 1.5 × 10 5) was 79% of the total for all isolates from non-CF sources. A previous study of longitudinally sampled − (π = 0.0052, SE = 1.6 × 10 5), confirming that much of the P. aeruginosa isolates (9) proposed that rates of adaptation are worldwide can be present within a limited geographical initially high and decline through time, a pattern commonly region (12). Nucleotide diversity was fairly even along the core observed in laboratory evolution experiments (18). Evidence with only a few hotspots of polymorphism (Fig. 1). The most suggests that the initial burst of adaptation is driven by positive selection during colonization of the CF airways, whereas the divergent region contained genes for the synthesis and uptake of subsequent prolonged period of persistence is characterized by iron-scavenging siderophores (pyoverdine), which are important purifying selection (6, 9). The structure of the phylogeny is for virulence and are known to be under diversifying selection consistent with this dynamic selection regime; however, in- (13, 14). Other divergent regions included genes for production creased genome sampling of isolates from environmental and of flagella or (O-antigen), both of which non-CF sources is needed to confirm that these patterns are elicit host immune response and are likely to be under di- CF specific. versifying selection (15). These results suggest that these clinical No evidence for phylogeographic structure was found. Closely P. aeruginosa isolates have been evolving in close association related isolates were not sampled from the same geographic with hosts for long enough to leave imprints of di- location, and isolates from Ontario were not exclusively mono- versifying selection on genomic polymorphism. phyletic. Population differentiation was not detected between Ontario versus non-Ontario populations (coefficient of differ- − entiation = 0.12, SE = 5.6 × 10 4), or between subpopulations (cities) within Ontario (0.005, SE = 0.001).

Intercontinental Spread of a Transmissible, Epidemic Strain. Clade A was composed of the Liverpool epidemic strain (LES) (19) and eight isolates of a transmissible, epidemic strain from Ontario named “strain A” (1), representing the only case where an Ontario isolate was nearly identical to a non-Ontario isolate. These two strains have the same pulsed-field gel electrophoresis (PFGE) and MLST patterns and share characteristics of trans- missibility and greater risks of negative clinical outcomes (1). Based on the high levels of genome-wide sequence similarity found here, we conclude that LES and epidemic strain A are equivalent and hereafter refer to them collectively as “Liverpool/ epidemic strain A” (LESA). Our genomic data therefore confirm Fig. 1. Nucleotide diversity (π) along the core genome alignment. Plotted val- cross-infection of an epidemic strain within Ontario and, im- ues are means from sliding windows of 4 kb in length with a step size of 2 kb. portantly, between the United Kingdom and North America.

21066 | www.pnas.org/cgi/doi/10.1073/pnas.1307862110 Dettman et al. Downloaded by guest on September 24, 2021 Fig. 2. (Left) Maximum likelihood phylogeny constructed from the core genome alignment (5.06 Mb; 137,941 SNPs). All internal branches have bootstrap percentages >70 except those shown in gray. The rooting branch is four times longer than shown. Isolates with names starting with JD were sampled from Ontario (Ham, Hamilton; Kit, Kitchener; Lon, London; Ott, Ottawa; Sud, Sudbury; Tor, Toronto). (Right) Example of a BLAST atlas with a LESA isolate as the reference genome. Coding genes in the LESB58 genome are indicated along the Upper row as blue bars. For every other genome, the presence/absence of the corresponding LESB58 gene is indicated by presence/absence of a bar. Arrows along the bottom indicate regions of genome that were identified as LESA specific; named regions correspond roughly with regions identified in ref. 19.

It may be that LESA is not in fact an “epidemic” strain but is and direction of selection among core genes (ω range = 0–1.42; SI just highly prevalent in the surrounding environments. Two lines Appendix,Fig.S2). We investigated the characteristics of the top of evidence argue against this interpretation. First, it is extremely 5% of genes with the highest ω-values using gene ontology (GO) unlikely that the same high-frequency strain is found in both annotation and enrichment analyses. Three ontology classes were Ontario and the United Kingdom, especially given the large strongly represented in this top-ω sample: oxidoreductase activity, amount of found within Ontario, let alone on secretion, and heterocycle metabolism (SI Appendix, Table S4). a global geographic scale (12). Second, in a collection of 65 GO information is available for only a subset of genes in the P. aeruginosa isolates sampled from medical equipment and sinks genome, so more comprehensive manual annotation of predicted from CF clinics and patients’ homes in Ontario (1), all were function or biological role was performed. The list of the top 50 found to be nonepidemic. ω-ranked annotated genes contained many genes with reduction/ Clade B consists of five isolates of an epidemic strain that have oxidation (redox), regulatory, secretion, or transport functions not been sampled outside of Ontario and were classified as Ontario (Table 1). The most common class, redox, comprised 36% of the “strain B” on the basis of their distinct PFGE profiles (1). We refer top 50 genes, a proportion significantly greater than that for the to this strain as “epidemic strain B” (ESB). Three additional iso- full reference genome (Fisher’s exact, P < 0.0001). lates that were originally identified as strain B did not fall within the Redox genes of interest included those with roles in the den- main ESB clade, demonstrating the superior utility of whole- trification pathway (nosY and nirQ) and antioxidation (glutathione genome analyses to distinguish phylogenetic relationships. S-transferases, epoxide hydrolase). Other genes were involved in type-III secretion of (exoY and pscQ), a key bacterial Adaptation to the CF Lung. To identify genes putatively involved in virulence mechanism that has been demonstrated in P. aeruginosa adaptation to the CF lung, we examined the entire 32-genome (20). The annotated gene with the second highest ω was plcR, sample to (i) quantify the average selective regime on core a regulator of phospholipase C secretion, which is a determinant genes and (ii) establish the distribution and identity of accessory of virulence and pathogenesis in numerous bacteria (21). The genome content. annotated gene with the highest ω was erbR, which encodes Selection on core genes. Confidently aligned core genes were ana- a response regulator of biofilm formation, ethanol oxidation, lyzed individually to quantify the strength of selection using the and drug resistance (22). Signatures of elevated nonsynonymous relative rate of nonsynonymous to synonymous substitution (dN/ rates in genes with these particular molecular functions and dS = ω). Here, ω is a global parameter for each gene, and the biological processes suggest they are involved in adaptation to average ω for 2,827 analyzed genes was 0.099 (SE = 0.002). the CF lung environment. EVOLUTION Although this indicates a general background of purifying Characteristics of the accessory genome. Gain and loss of accessory (negative) selection, there was substantial variation in the strength genome regions via horizontal transfer determines the evolution

Dettman et al. PNAS | December 24, 2013 | vol. 110 | no. 52 | 21067 Downloaded by guest on September 24, 2021 Table 1. Functional categorization of the top 50 core genes with the highest relative rates of nonsynonymous substitution (with automated or manual annotation) Class No. of genes Examples of genes and functional roles

Redox 18 Nitrous oxide reductase (nosY) Regulation of denitrification (nirQ) Ethanol oxidation (erbR, eraS) Glutathione S-transferases Epoxide hydrolase Oxidases Dehydrogenases Oxidoreductase Dioxygenase Transport 7 MFS transporters ABC transporters Multidrug efflux protein (norA) Secretion 5 Type III secretion system (pscQ) Phospholipases Fig. 3. Diagram of the proposed functional relationships between sources (exoY) of oxidative stress, regulation of NO, and candidate pathways identified in Regulatory 7 Two-component response regulators this study. Transcriptional regulators RNA polymerase sigma-70 factor Hydrolysis 8 Beta-lactamases as the terminal electron acceptor, which produces NO as a tox- Hydrolases ic intermediate. Miscellaneous 5 Kinases Antibiotic exposure. Several classes of antibiotics appear to cause Transferases bacterial cell death by oxidative damage. The proposed central mechanism involves elevated levels of intracellular ROS (27) but the details of the mechanism are still under debate (28). Also, bacterial NO and oxidative stress interact to affect sensitivities to a wide of genome content. Horizontal acquisition of novel genetic ele- spectrum of antibiotics (29). The interplay between these two phe- ments is often associated with adaptation to new niches or - nomena requires further study, but they demonstrate a role for oxi- styles (23), so the nonessential, accessory genome of P. aeruginosa dative stress and redox balance in antibiotic resistance. may contain adaptive factors that contribute to successful coloni- Taken together, these three lines of evidence suggest there can zation and persistence within CF lungs. We visualized the location, be strong selective pressure for adequate mitigation of oxidative distribution, and relative sizes of accessory genome regions by insult and maintenance of proper redox balance within the bac- constructing BLAST atlases (example in Fig. 2). Accessory ge- terial cell (Fig. 3). In particular, the regulation of the free radical nome content was commonly, but not always, shared among iso- NO may be a central mediator of oxidative stress management. In lates that were closely related in the core phylogeny. P. aeruginosa, the accumulation of NO is prevented by reduction Gene identification predicted 3,299 ORFs within the 4.4 Mb of to nitrous oxide (N2O), which is itself an RNI and is further re- analyzed accessory genome sequence. Compared with the entire duced to nontoxic dinitrogen gas. This pathway is regulated by fi genome, the accessory genome was signi cantly enriched for nu- nirQ, and the last step is performed by nosY (nitrous oxide re- merous functional roles (SI Appendix,TableS5). The hypothesis ductase), both of which were on the list of genes with highest that the accessory genome helps P. aeruginosa adapt to a diversity relative nonsynonymous rates. of environments, particularly the CF lung, was supported by many We hypothesized that isolates from CF sources may be able to of the significantly overrepresented biological processes, such as withstand greater concentrations of external NO than isolates pathogenesis, antibiotic metabolism, and transmembrane trans- from non-CF sources. Consistent with this expectation, minimum port. The top molecular function and were both inhibitory concentrations (MICs) of NO (reduction product of oxidation/reduction, further suggesting that genes with redox func- NaNO2) varied among CF isolates from Ontario, with over half tionality have played an important role in the evolution of clinical (15/24) (SI Appendix, Fig. S3) being equally or more resistant P. aeruginosa strains. than the non-CF outgroup isolate, PA14. However, median MICs of CF and non-CF isolates (PA01 and PA14) did not differ Oxidative Stress and Redox Balance. How might selection on significantly (Z = 0.55; P > 0.59). These results should be treated P. aeruginosa genes with redox functions be associated with an as preliminary, as laboratory assay conditions may not accurately adaptive benefit in the CF lung? The CF airway is characterized simulate some aspects of the CF lung that affect NO resistance by multiple disparate sources of oxidative stress (Fig. 3), three of and the non-CF sample consisted of just two laboratory isolates. which are described below. A stronger test requires more extensive sampling of environ- Oxidative burst by host. The host’s immune response includes the mental isolates from Ontario. production of several reactive oxygen species (ROS) and reactive nitrogen intermediates (RNIs). The oxidative burst by activated LESA. Laboratory and clinical studies have revealed differences macrophages releases nitric oxide (NO) and superoxide, creating between epidemic and nonepidemic strains in drug resistance oxidative stress that functions to protect the host’s lung from profiles, virulence in infection models, propensity for superin- infection (24). Lungs of patients with CF also have reduced fection (colonization by a new strain of a patient already infected levels of the major pulmonary antioxidant, glutathione (25). with another strain), and clinical outcomes for patients with CF Anaerobic respiration. The thick, viscous mucus of the CF lung has (1, 10, 30–33). We investigated to what extent genes and genomic reduced oxygen permeability, and growth as a biofilm further regions distinguish LESA from all other strains by analyzing (i) reduces the oxygen availability to bacterial cells. Under such LESA divergence in core gene phylogenies and (ii) the func- conditions, P. aeruginosa respires anaerobically (26) using nitrate tional characteristics of the LESA-specific accessory genome.

21068 | www.pnas.org/cgi/doi/10.1073/pnas.1307862110 Dettman et al. Downloaded by guest on September 24, 2021 LESA divergence in core gene phylogenies. Many nucleotide poly- The abundance of LESA-specific transporters also suggested morphisms within the core genome phylogenetically differentiate that LESA may have a greater ability to survive in harsh envi- LESA from other isolates (Fig. 2). To determine which genes ronments with high concentrations of cations or heavy metals. To contributed most to LESA divergence, we ranked core genes test this hypothesis, we measured the MIC of NaCl, CuSO4, based on the relative length of the LESA branch and summa- ZnCl2, and CoCl2 for the 24 sequenced Ontario isolates. Of the rized the functional roles of the top 5% (SI Appendix, Table S6). four tested cations, LESA isolates were significantly more re- The most represented molecular function and biological process sistant to only one, cobalt (P < 0.02, SI Appendix, Table S10), was kinase activity and arginine metabolism, respectively. Add- potentially due to the two LESA-specific genes in the cobalt– ing manual annotation reinforced these results (SI Appendix, zinc–cadmium (CzcA) resistance family (SI Appendix, Table S9). Table S7), but also showed that the 50 top-ranked, annotated Our evolutionary genomic analyses of an epidemic P. aeruginosa fi genes were signi cantly overrepresented by redox functions [12 strain focused on the most prevalent strain in our Ontario pop- < of 50 (24%), P 0.001]. ulation, LESA. Other transmissible, epidemic strains are common LESA-divergent genes of interest included several involved in in different regions of the world, such as Denmark (9) and Australia arginine metabolism (aguB, spuF, and aotP) and two with roles (40). When population-level genomic data become available for fi in bio lm matrix formation (pelF and pelG). A gene required these epidemic strains, and similar analyses are performed, it will be for the biosynthesis of pyrroloquinoline quinone (pqqE), a redox very interesting to compare the results with those described here. cofactor and free-radical scavenger, was also found. Two genes The addition of other epidemic strains will help to determine if whose products are secreted by type-III systems were also dis- the characteristics of LESA observed here are shared among all covered: exotoxin T (exoT) and a pore-forming hemolysin, which epidemic strains or are simply strain specific. causes the lysis of host red blood cells, both of which contribute to virulence (34). Conclusions The importance of arginine metabolism can be explained by its We have demonstrated that P. aeruginosa has a nonclonal epi- multiple roles in several pathways relevant to CF lung adapta- demic population structure, with a background of many unique, tion. First, both bacteria and macrophages synthesize NO from recombining genotypes and the rare occurrence of successful arginine precursors. Second, P. aeruginosa preferentially catab- epidemic clones that spread to higher prevalence under certain olizes arginine (into ornithine) to provide energy for anaerobic conditions. We present unique genome sequence evidence that growth in conditions that mimic the CF lung, a pathway pre- viously implicated in adaptation during chronic infection (35, the most common epidemic strain in the United Kingdom is the 36). Third, the synthesis of the polyamine putrescine from argi- same as that in North America. This highlights the importance of nine precursors produces agmatine as an intermediate, which comprehensive surveillance of dominant, CF-associated strains inhibits NO synthases of macrophages (37) and diminishes the that occur worldwide. Furthermore, it emphasizes the need for macrophage-derived oxidative burst. An imbalance of arginine identifying the genetic factors that characterize epidemic strains and polyamine metabolism in the CF lung has been implicated in of P. aeruginosa and similar . the pathophysiology of CF lung disease (38). The many LESA- Resident P. aeruginosa can reproduce for hundreds of thousands ’ divergent genes with roles in utilization and transport of argi- of generations during a patient s lifetime of chronic infection, nine, ornithine, agmatine, and putrescine (SI Appendix, Tables adapting for survival and persistence in the CF lung. As opposed to S6 and S7) suggest LESA may be particularly well adapted to this longitudinal sampling of a single infection, we examined a broader aspect of the CF lung environment. sample of strains from a larger number of individual CF patients, Characteristics of the LESA-specific accessory genome. We identified allowing us to identify the trends that emerge from multiple in- five broad regions of LESA-specific accessory genome (regions dependent bouts of adaptation to CF lungs. By analyzing the possessed by all LESA isolates but by no other isolates; Fig. 2), genomes of a diverse set of isolates that have successfully infected summing to 85 kb. We studied the functions of the 72 included and persisted within CF lungs, we could infer the genes and ORFs and found that the two most overrepresented biological functional pathways that were important for the evolution of this processes were protein secretion and cation transport (SI Ap- human pathogen. Consistent with a long history of host–patho- pendix, Table S8). The largest LESA-specific accessory region gen interaction, the selection pressures imposed by the CF lung was 30 kb in length and corresponded to a portion of a previously environment have had a major influence on the genomic evo- described genomic island in the LES (LESGI-3) (19). This re- lution and genetic characteristics of clinically important isolates. gion contained nine transport-related genes and three genes with Many of the highlighted candidate genes were involved in bi- to transcriptional regulators of heavy metal resistance ological functions with obvious roles in this pathosystem, such as operons (SI Appendix, Table S9). Many of the LESA-specific biofilm formation, transmembrane transport, pathogenesis, he- transporters were efflux pumps belonging to the major facilitator molysis, and secretion. However, the presence of oxidative superfamily (MFS) and resistance–nodulation–division (RND) families, both of which have a wide substrate range, with affini- ties for binding and exporting sugars, metal ions, and multiple Table 2. Comparison of MICs of three antibiotics with different classes of antibiotics (39). These transporters actively pump the modes of action toxic compound out of the bacterial cell to avoid contact with Wilcoxon rank target sites. sum test Enrichment of these types of efflux pumps may confer in- creased resistance to antibiotics. To test this prediction directly, Antibiotic Strain group Median MIC (μg/mL) Z Prob>jZj we determined the MICs of three antibiotics with different fl mechanisms of action—the quinolone ciprofloxacin, the β-lactam Cipro oxacin LESA 4 1.97 0.049 ceftazidime, and the aminoglycoside gentamicin—for 48 clinical (Quinolone) Non-LESA 2 isolates from Ontario and two reference isolates. Resistance to Ceftazidime LESA 32 2.14 0.033 each of the three antibiotics was significantly greater for LESA (Beta-lactam) Non-LESA 12 than non-LESA isolates (Table 2). These results lend experi- Gentamicin LESA 128 3.28 0.001 mental support to the idea that higher levels of efflux-mediated (Aminoglycoside) Non-LESA 8 EVOLUTION multidrug resistance may be a major factor contributing to the LESA isolates (n = 10) were significantly more resistant than non-LESA iso- success of this epidemic strain. lates (n = 40) to all three antibiotics. MICS, minimum inhibitory concentrations.

Dettman et al. PNAS | December 24, 2013 | vol. 110 | no. 52 | 21069 Downloaded by guest on September 24, 2021 stressors and antibiotics appear to be key factors that have driven regarding the functions and processes that are important for in- the adaptive evolution of this pathogen within the host. fection and survival. Of course, detailed genetic manipulations and The evolution of generalized resistance to oxidative stress, whether tests in infection models are required to verify genes of particular the source is macrophages, anaerobic respiration, or antibiotics, importance. By identifying the key genetic pathways involved in would confer a strong adaptive benefit for survival in the CF lung. We bacterial adaptation to the CF lung, research on new therapeutic hypothesize that bacteria can adapt simultaneously to these three approaches can be targeted toward blocking the common adaptive similar selection pressures, in part, because of the interconnections routes taken by the pathogen. between the underlying genetic pathways that are affected. For ex- ample, resistance to antibiotics may differ between anaerobic and Materials and Methods aerobic conditions (41), and a regulatory connection exists be- Sources of isolates and genome sequencing methods are presented in SI tween oxidation sensing and antibiotic resistance (42). This may Appendix. Methods for de novo genome assembly, multiple-genome explain why eliminating P. aeruginosa infections from the CF lung alignment, core and accessory genome calculation, and phylogenetic re- is so difficult. Survival in untreated CF lungs may require the construction are described in SI Appendix. Assessment of polymorphism, ability to neutralize oxidative stressors, a characteristic that pre- population differentiation, core recombination, strength of selection, and adapts the bacteria to survive extended periods of antibiotic ex- gene ontology information were performed as described in SI Appendix. posure during clinical treatment. Methods for measuring minimum inhibitory concentrations are also de- scribed in SI Appendix. The molecular bases for transmissibility, colonization, and per- sistence of in the CF lung remain poorly understood. As ACKNOWLEDGMENTS. This work was supported by the Natural Sciences and studies of other CF airway pathogens continue to reveal additional Engineering Research Council of Canada and the Canadian Institutes for candidate genes (e.g., Burkholderia) (43), a consensus will form Health Research.

1. Aaron SD, et al. (2010) Infection with transmissible strains of Pseudomonas aerugi- 23. Jackson RW, Vinatzer B, Arnold DL, Dorus S, Murillo J (2011) The influence of the nosa and clinical outcomes in adults with cystic fibrosis. JAMA 304(19):2145–2153. accessory genome on bacterial pathogen evolution. Mobile Genet Elements 1(1): 2. LiPuma JJ (2010) The changing microbial epidemiology in cystic fibrosis. Clin Microbiol 55–65. Rev 23(2):299–323. 24. MacMicking J, Xie QW, Nathan C (1997) Nitric oxide and macrophage function. Annu 3. Hauser AR, Jain M, Bar-Meir M, McColley SA (2011) Clinical significance of microbial Rev Immunol 15:323–350. fi infection and adaptation in cystic fibrosis. Clin Microbiol Rev 24(1):29–70. 25. Roum JH, Buhl R, McElvaney NG, Borok Z, Crystal RG (1993) Systemic de ciency of fi – 4. Folkesson A, et al. (2012) Adaptation of Pseudomonas aeruginosa to the cystic fibrosis glutathione in cystic brosis. J Appl Physiol (1985) 75(6):2419 2424. airway: An evolutionary perspective. Nat Rev Microbiol 10(12):841–851. 26. Schobert M, Jahn D (2010) Anaerobic physiology of Pseudomonas aeruginosa in the fi – 5. Sriramulu DD, Lünsdorf H, Lam JS, Römling U (2005) Microcolony formation: A novel cystic brosis lung. Int J Med Microbiol 300(8):549 556. biofilm model of Pseudomonas aeruginosa for the cystic fibrosis lung. J Med Micro- 27. Kohanski MA, Dwyer DJ, Hayete B, Lawrence CA, Collins JJ (2007) A common mech- anism of cellular death induced by bactericidal antibiotics. Cell 130(5):797–810. biol 54(Pt 7):667–676. 28. Liu Y, Imlay JA (2013) Cell death from antibiotics without the involvement of reactive 6. Smith EE, et al. (2006) Genetic adaptation by Pseudomonas aeruginosa to the airways oxygen species. Science 339(6124):1210–1213. of cystic fibrosis patients. Proc Natl Acad Sci USA 103(22):8487–8492. 29. Gusarov I, Shatalin K, Starodubtseva M, Nudler E (2009) Endogenous nitric oxide 7. Poole K (2005) Efflux-mediated . J Antimicrob Chemother protects bacteria against a wide spectrum of antibiotics. Science 325(5946): 56(1):20–51. 1380–1384. 8. Cramer N, et al. (2011) of the major common Pseudomonas aerugi- 30. McCallum SJ, et al. (2001) Superinfection with a transmissible strain of Pseudomonas fi – nosa clones C and PA14 in cystic brosis lungs. Environ Microbiol 13(7):1690 1704. aeruginosa in adults with cystic fibrosis chronically colonised by P aeruginosa. Lancet 9. Yang L, et al. (2011) Evolutionary dynamics of bacteria in a human host environment. 358(9281):558–560. – Proc Natl Acad Sci USA 108(18):7481 7486. 31. Salunkhe P, et al. (2005) A cystic fibrosis epidemic strain of Pseudomonas aeruginosa 10. Mowat E, et al. (2011) Pseudomonas aeruginosa population diversity and turnover displays enhanced virulence and antimicrobial resistance. J Bacteriol 187(14): in cystic fibrosis chronic infections. Am J Respir Crit Care Med 183(12):1674–1679. 4908–4920. 11. Al-Aloul M, et al. (2004) Increased morbidity associated with chronic infection by an 32. Beaudoin T, Aaron SD, Giesbrecht-Lewis T, Vandemheen K, Mah TF (2010) Charac- epidemic Pseudomonas aeruginosa strain in CF patients. Thorax 59(4):334–336. terization of clonal strains of Pseudomonas aeruginosa isolated from cystic fibrosis 12. Pirnay JP, et al. (2009) Pseudomonas aeruginosa population structure revisited. PLoS patients in Ontario, Canada. Can J Microbiol 56(7):548–557. ONE 4(11):e7740. 33. Carter MEK, et al. (2010) A subtype of a Pseudomonas aeruginosa cystic fibrosis ep- 13. Takase H, Nitanai H, Hoshino K, Otani T (2000) Impact of siderophore production idemic strain exhibits enhanced virulence in a murine model of acute respiratory on Pseudomonas aeruginosa infections in immunosuppressed mice. Infect Immun infection. J Infect Dis 202(6):935–942. 68(4):1834–1839. 34. Shafikhani SH, Engel J (2006) Pseudomonas aeruginosa type III-secreted toxin ExoT 14. Smith EE, Sims EH, Spencer DH, Kaul R, Olson MV (2005) Evidence for diversifying inhibits host-cell division by targeting cytokinesis at multiple steps. Proc Natl Acad Sci selection at the pyoverdine locus of Pseudomonas aeruginosa. J Bacteriol 187(6): USA 103(42):15605–15610. 2138–2147. 35. Hoboth C, et al. (2009) Dynamics of adaptive microevolution of hypermutable 15. Raymond CK, et al. (2002) at the O-antigen biosynthetic locus in Pseudomonas aeruginosa during chronic pulmonary infection in patients with fi – Pseudomonas aeruginosa. J Bacteriol 184(13):3614–3622. cystic brosis. JInfectDis200(1):118 130. 16. Jolley KA, Maiden MCJ (2010) BIGSdb: Scalable analysis of bacterial genome variation 36. Palmer KL, Aye LM, Whiteley M (2007) Nutritional cues control Pseudomonas aeru- ginosa multicellular behavior in cystic fibrosis sputum. J Bacteriol 189(22):8079–8087. at the population level. BMC Bioinformatics 11:595. 37. Regunathan S, Piletz JE (2003) Regulation of inducible nitric oxide synthase and 17. Marttinen P, et al. (2012) Detection of recombination events in bacterial genomes agmatine synthesis in macrophages and astrocytes. Ann N Y Acad Sci 1009:20–29. from large population samples. Nucleic Acids Res 40(1):e6. 38. Grasemann H, et al. (2012) L-ornithine derived polyamines in cystic fibrosis airways. 18. Dettman JR, et al. (2012) Evolutionary insight from whole-genome sequencing of PLoS ONE 7(10):e46618. experimentally evolved microbes. Mol Ecol 21(9):2058–2077. 39. Saier MH, Jr., et al. (1998) Evolutionary origins of multidrug and drug-specificefflux 19. Winstanley C, et al. (2009) Newly introduced genomic prophage islands are critical pumps in bacteria. FASEB J 12(3):265–274. determinants of in vivo competitiveness in the Liverpool Epidemic Strain of Pseudo- 40. Armstrong DS, et al. (2002) Detection of a widespread clone of Pseudomonas aeru- – monas aeruginosa. Genome Res 19(1):12 23. ginosa in a pediatric cystic fibrosis clinic. Am J Respir Crit Care Med 166(7):983–987. 20. Hauser AR (2009) The type III secretion system of Pseudomonas aeruginosa: Infection 41. Hill D, et al. (2005) Antibiotic susceptabilities of Pseudomonas aeruginosa isolates – by injection. Nat Rev Microbiol 7(9):654 665. derived from patients with cystic fibrosis under aerobic, anaerobic, and biofilm 21. Schmiel DH, Miller VL (1999) Bacterial phospholipases and pathogenesis. Microbes conditions. J Clin Microbiol 43(10):5085–5090. Infect 1(13):1103–1112. 42. Chen H, et al. (2008) The Pseudomonas aeruginosa multidrug efflux regulator MexR 22. Beaudoin T, Zhang L, Hinz AJ, Parr CJ, Mah TF (2012) The biofilm-specific antibiotic uses an oxidation-sensing mechanism. Proc Natl Acad Sci USA 105(36):13586–13591. resistance gene ndvB is important for expression of ethanol oxidation genes in 43. Lieberman TD, et al. (2011) Parallel bacterial evolution within multiple patients Pseudomonas aeruginosa biofilms. J Bacteriol 194(12):3128–3136. identifies candidate pathogenicity genes. Nat Genet 43(12):1275–1280.

21070 | www.pnas.org/cgi/doi/10.1073/pnas.1307862110 Dettman et al. Downloaded by guest on September 24, 2021