Enhanced in vivo fitness of carbapenem-resistant oprD mutants of Pseudomonas aeruginosa revealed through high-throughput sequencing

Significance

It is thought antibiotic resistance carries a fitness cost and reduces microbial virulence. Using high-throughput sequencing analysis of a transposon insertion bank in Pseudomonas aeruginosa, we found enhanced fitness for in vivo mucosal colonization and systemic spread of strains with transposon insertions in the oprD gene. This conferred resistance to carbapenem antibiotics as well as enhanced resistance to killing at acidic pH and by normal human serum along with increased cytotoxicity against murine macrophages. RNA-sequencing analysis revealed that oprD deficiency led to transcriptional changes in numerous genes that may contribute to the enhanced in vivo fitness observed. Thus, if carbapenem resistance develops during antibiotic therapy of P. aeruginosa infections, it may lead to enhanced fitness and virulence in infected hosts.

Abstract

An important question regarding the biologic implications of antibiotic-resistant microbes is how resistance impacts the organism’s overall fitness and virulence. Currently it is generally thought that antibiotic resistance carries a fitness cost and reduces virulence. For the human pathogen Pseudomonas aeruginosa, treatment with carbapenem antibiotics is a mainstay of therapy that can lead to the emergence of resistance, often through the loss of the carbapenem entry channel OprD. Transposon insertion-site sequencing was used to analyze the fitness of 300,000 mutants of P. aeruginosa strain PA14 in a mouse model for gut colonization and systemic dissemination after induction of neutropenia. Transposon insertions in the oprD gene led not only to carbapenem resistance but also to a dramatic increase in mucosal colonization and dissemination to the spleen. These findings were confirmed in vivo with different oprD mutants of PA14 as well as with related pairs of carbapenem-susceptible and -resistant clinical isolates. Compared with OprD⁺ strains, those lacking OprD were more resistant to killing by acidic pH or normal human serum and had increased cytotoxicity against murine macrophages. RNA-sequencing analysis revealed that an oprD mutant showed dramatic changes in the transcription of genes that may contribute to the various phenotypic changes observed. The association between carbapenem resistance and enhanced survival of P. aeruginosa in infected murine hosts suggests that either drug resistance or host colonization can cause the emergence of more pathogenic, drug-resistant P. aeruginosa clones in a single genetic event.

Serious political, medical, and public health measures are being implemented to address the significant problems associated with drug-resistant pathogens. National and international campaigns have been instituted to reduce and restrict antibiotic use to achieve decreases in infections by antibiotic-resistant organisms (1), which have been met with some success (2). Although one reason for implementing antibiotic stewardship programs is to reduce the selective pressure for emergence of drug-resistant microbes, an additional expected consequence is a reduction in extant-resistant microbes from community and hospital environments (3). Extensive studies indicate drug-resistant microbes have decreased fitness and virulence, reflected by an impairment in growth in infected hosts (4), lower transmission rates (5), and reduced virulence manifested by diminished invasiveness and higher clearance rates (6). A general decrease in the occurrence of drug-resistant organisms can thus potentially be achieved as the more fit antibiotic-susceptible organisms displace resistant strains over time (6). However, if antibiotic-resistant mutations can lead to enhanced fitness and virulence, this would challenge the prevailing paradigm that there is a negative fitness cost to drug resistance.

Pseudomonas aeruginosa is a classical example of a bacterial pathogen that is often found associated with extremely difficult-to-treat infections that resist antibiotic therapies. This organism frequently emerges as a threat to neutropenic, immunosuppressed patients undergoing treatment for cancer wherein one usually observes the spread of antibiotic-resistant organisms from gastrointestinal (GI) sites into the blood stream. It has also been observed in the setting of cystic fibrosis (CF) that reversion or displacement of resident drug-resistant P. aeruginosa strains does not occur even when antibiotic treatment is intermittent (7). This observation suggests that some mechanisms leading to antibiotic resistance could also enhance the fitness of P. aeruginosa in vivo and thus contribute to persistent infections.

Transposon (Tn) insertion-site sequencing (Tn-seq) is a powerful analytical method that in various formats has been called “INSeq” (insertion-site sequencing) (8, 9), “Tn-seq” (10), or “high-throughput insertion tracking by deep sequencing” (HITS) (11). These methods allow one to measure the fitness of collections of insertion mutants under a given growth condition by using deep sequencing to efficiently quantitate the levels of junction sequences that mark unique Tn-insertion sites on the bacterial chromosome. To test the hypothesis that changes in the antibiotic resistance profile of P. aeruginosa could be associated with enhanced host colonization particularly in the context of immunosuppression (12), we used INSeq to identify P. aeruginosa genes whose inactivation promotes increased fitness in neutropenic mice. We constructed a saturated TnSAMDGm (8) insertion library in the well-characterized P. aeruginosa strain PA14 (13) and subjected the bank to in vivo selection in a mouse model for mucosal colonization and dissemination. Tracking of various mutants indicated that a strong positive selection for the loss of different gene functions occurred in vivo. Remarkably, loss of OprD function resulted both in enhanced host colonization and dissemination, as well as resistance to the carbapenem antibiotics.

Results and Discussion

We have recently reported that antibiotic treated mice can be used in principle to analyze the GI fitness of P. aeruginosa Tn mutants (9). In brief, to establish P. aeruginosa GI colonization, mice received streptomycin and penicillin in sterile drinking water for 5 d to clear the indigenous commensal GI microbial flora, after which the PA14 TnSAMDGm insertion bank, grown overnight in lysogeny broth (LB, designated “input LB”) containing 15 mg gentamicin per L was added to sterile water containing penicillin and gentamicin. The drinking water containing bacteria was renewed after 72 h and administered to mice for a total of 6 d after which sterile drinking water was given for 24 h before mice were killed and their ceca removed. During this time, the strains with Tn insertions that could colonize and survive in the cecum of the GI tract were selected (designated “output cecum”). Following induction of neutropenia after 6 d of mucosal colonization, the strains with Tn insertions that were able to disseminate systemically over 24 h were recovered from the spleen (designated “output spleen”) (14). Splenic dissemination after induction of neutropenia was considered to indicate fitness for systemic dissemination.

Tn-seq analysis of the population of organisms derived from growing the Tn-insertion bank overnight in LB showed there were no Tn insertions mutants that yielded more than 8,000 sequencing reads out of 10⁶ normalized reads (0.8%) indicating that the Tn insertions were relatively evenly distributed in the genome and no Tn insertions were significantly overrepresented in the bank (Fig. 1A). However, strains with Tn insertions in 13 different genes were strongly overrepresented in the output cecum samples (Fig. 1B). Twelve of these were in genes resulting in the loss of type IVa pili (15) comprising 380,000 total reads (38%). Strikingly, strains bearing Tn insertions in the oprD gene, whose loss leads to acquisition of the resistance to the carbapenem antibiotics, represented 42% of the output cecum population. The number of sequencing reads of the various oprD mutants increased from 515 (0.05% of the total reads) in LB to 420,621 in the cecum (Fig. 1B). More dramatically, 94% of the strains that disseminated to the spleen of neutropenic mice (Fig. 1C) had Tn insertions in the oprD gene (947,397 sequencing reads in the spleen output group). Whereas a large number of Tn-insertion mutants showed a significant decrease in their recovery from each in vivo environment, mutants carrying Tn insertions in the oprD gene were dramatically increased in in vivo fitness [Z = 972.29, P < 10^-16, Kal’s statistical test (16)]. To avoid any bias due to an overrepresentation of the oprD mutants in water over time, we sequenced the bank of mutants after bacteria resided in water containing penicillin and gentamicin for 48 h at room temperature. Tn insertions in oprD represented less than 0.5% of the viable mutants present in such water samples. As the drinking water containing bacteria was changed after 72 h, the lack of overrepresentation of the oprD mutant strains at any time point in the water was confirmed by plating the bank of mutants that survived for 72 h in water on plates with or without carbapenems (3 mg/L). We confirmed by PCR that only strains with a Tn inserted in the oprD gene were able to grow on the carbapenem-containing plates and they represented only 0.5% of the total strains. In parallel, we observed no difference in the growth of an oprD mutant strain in vitro (Fig. S1).

Fig. 1.

Analysis of in vivo fitness of Tn insertions in the P. aeruginosa PA14 genome. (A) Circos plot of Tn insertions into strains grown overnight in LB with their ordered representation (inner blue circle). Gray circular lines represent 2,000, 4,000, 6,000, or 8,000 sequencing reads recovered from the input LB sample_. Outermost circle represents the full PA14 genome. (B and C) Analysis of Tn insertions into genes within the PA14 chromosome revealed strains with increased in vivo fitness. (B) Ordered representation of the in vivo fitness for cecal colonization of all of the strains with Tn insertions able to grow overnight in LB. (C) Ordered representation of the in vivo fitness for splenic dissemination of all of the strains with Tn insertions able to colonize the output cecum. Total number of reads recovered from the input LB and output cecum and spleen samples were normalized to 1,000,000.

Increased fitness was not specific for Tn insertions into genes encoding for all outer membrane proteins (OMPs) although, as among the Tn insertions in the 262 genes predicted to encode for OMPs (17, 18), only 68 were able to colonize the cecum (Fig. 2A). Among them, only the Tn insertions in the oprD gene displayed an enhanced fitness for colonization (Fig. 2A).

Fig. 2.

Characterization of oprD mutants. (A) Fitness for dissemination of all of the Tn insertions in genes encoding for predicted OMPs that are able to colonize the cecum. Positive fitness (more than twofold increase) was detected only for Tn insertions into the oprD gene (PA14_51880) (green bar). All of the insertions in the genes for the remaining OMPs had a reduced fitness (red bars) ranging from a 10-fold to a >10,000-fold decrease in the ratio of reads in the cecum versus spleen. Black circular lines within the gray circle represents baseline, and thin gray circular lines represent 10-fold changes (i.e., a log₁₀ scale). Outermost circle represents the full PA14 genome with a 10× magnification of the regions of interest. (B) Analysis of the OMP expression. oprD mutant strains Tn_oprD (spleen, 6-4, 8-1), clinical isolates (48-2 and 51-2), and PA14 ΔoprD did not express the OprD protein. OprD (arrow) is seen in the OMP extract from WT PA14, clinical isolates 48-1 and 51-1, and in the complemented (PoprD) strains. M, molecular mass standard. (C) Swarming motility of the oprD mutant strains. WT PA14, Tn-oprD::Spleen, and Tn-oprD::Spleen (PoprD) strains showed normal motility, whereas the PA14_Tn-pilE strain lacking pili has defective motility and the PA14_Tn-fliC strain has no motility.

To further analyze the comparative colonization and dissemination abilities of wild-type (WT) P. aeruginosa PA14 and strains with Tn insertions in the oprD gene, we characterized several oprD-deficient variants of strain PA14. Two of these strains, PA14_Tn-oprD:6–4 and PA14_Tn-oprD:8–1, were retrieved from an ordered Tn library (19) and had distinct Tn insertions in the oprD gene. An additional PA14 strain, PA14_Tn-oprD::Spleen, was recovered from a neutropenic mouse spleen. All of these oprD Tn-insertion strains showed an enhanced resistance to the carbapenem antibiotic imipenem, with a minimum inhibitory concentration (MIC) of ≥6 mg/L compared with 1 mg/L for the WT PA14 strain.

To acquire correlative data from infected human patients, we obtained two additional strains of P. aeruginosa resistant to carbapenems from a collection of clinical isolates. Pulsed-field gel electrophoresis (PFGE) (20) was used to identify related pairs of strains from two patients (Fig. S2A) in which earlier clinical isolates (strains 48-1 and 51-1) were carbapenem susceptible (imipenem MIC < 1 mg/L), and later isolates (48-2 and 51-2) carbapenem resistant (imipenem MIC ≥ 32 mg/L). Sequencing of the oprD genes from these strains confirmed that the carbapenem-susceptible isolates had an intact oprD gene, whereas the carbapenem-resistant strain 48-2 had acquired the insertion sequence element IS Pa1328 at nucleotide 50 in the oprD gene and the carbapenem-resistant strain 51-2 had a 12-bp deletion (base pairs 579–590 included) and a single nucleotide change (G→A at position 1148 leading to a stop codon) in the oprD gene (Fig. S2B). We further constructed complemented strains for the PA14_Tn-oprD::Spleen strain and the 48-2 and 51-2 strains by conjugating into them a plasmid bearing a full-length oprD gene (PoprD). The complemented strains had MICs to imipenem of ≤1 mg/L. Finally, to further study the phenotypes associated with oprD deficiency, 22 additional clinical strains were selected (CS1–CS22): 10 strains (CS1–CS10) with a normal level of transcription of the oprD as determined by quantitative (q)RT-PCR, and 12 strains (CS11–CS22) with a reduced level of expression (Dataset S1). The three controls used for the qRT-PCRs were the PA14 WT strain, a strain with a clean deletion of the oprD gene, and a strain overexpressing oprD.

Analysis of OprD expression using SDS/PAGE showed that the 48.4 kDa OprD protein was readily seen in extracts from WT PA14, the clinical isolates 48-1 and 52-1 and the corresponding transcomplemented strains (Fig. 2B), whereas no evidence of an intact OprD protein was found in PA14 ΔoprD strain (used as a control), the Tn insertions designated 6-4, 8-1, Tn-oprD::Spleen, and the clinical strains 48-2 and 51-2 (Fig. 2B). This result was confirmed using a sensitive silver-staining reagent (Fig. S3) that also showed the lack of difference of expression in the other OMPs under these conditions, and suggested that OprD-truncated protein was not present in OprD mutant strains (Fig. S3).

As Tn insertions unable to produce type IVa pili were also positively selected for enhanced GI colonization, the swarming and the twitching motilities of P. aeruginosa PA14 Tn::oprD were assessed (15). No change in motility was associated with the loss of production of OprD (Fig. 2C and Figs. S4 and S5). Therefore, the basis for enhanced fitness for GI colonization of oprD mutants in mice is not the result of a defect in the production of type IVa pili.

Attempts to mark WT P. aeruginosa PA14 with either streptomycin or tetracycline resistance for in vivo tracking resulted in a diminution in their ability to colonize the murine GI tract in comparison with unmarked WT P. aeruginosa PA14 (Fig. S6), emblematic of the fitness cost usually attributed to the acquisition of antibiotic resistance (6). Thus, to compare the oprD-deficient strains with WT PA14, the unmarked WT strain and the oprD-deficient strains (gentamicin resistant) or the WT and PoprD-complemented strains (also gentamicin resistant) were mixed together in drinking water at a ratio of ∼1:1 and the relative levels of WT and mutant strains in the output samples were determined by plating cultures on LB agar and LB agar with 15 mg gentamicin per L. We subtracted the colony forming units (cfu) determined from the latter plates from those on the nonselective plates to obtain the level of colonization with WT P. aeruginosa PA14. This ratio was further confirmed by subculturing 100 separate colonies that grew on the LB-agar plates with and without gentamicin. Analysis of the oprD gene by PCR for each gentamicin-resistant colony confirmed that they were oprD mutants and not spontaneous gentamicin-resistant strains. A similar approach was used to differentiate between oprD-deficient and PoprD-complemented strains, however, carbapenem plates were used instead of gentamicin plates as both strains were gentamicin resistant.

Comparing selective fitness during cecal colonization of the PA14_Tn-oprD::Spleen isolate with the WT PA14 strain and the complemented PA14_Tn-oprD::Spleen (PoprD) strains revealed that the PA14_Tn-oprD::Spleen strain out-competed both the WT and complemented strains, constituting >90% of the isolates recovered from the cecum (Fig. 3A). Comparing the WT PA14 strain and complemented PA14_Tn-oprD::Spleen (PoprD) strains in the cecal colonization setting showed that both colonized the mice at comparable levels (Fig. 3A), supporting the conclusion that the loss of expression of OprD conferred an in vivo GI colonization advantage to the PA14_Tn-oprD::Spleen strain over the WT PA14 (Fig. 3A). No effect of the empty vector used for complementation was found in the GI colonization model (Fig. S7).

Fig. 3.

Analysis of the fitness for cecal colonization and systemic dissemination of the oprD mutant strains. (A) Ratio of cecal colonization between indicated oprD mutant strain and a strain with intact oprD. (B and C) In vivo competitive index (CI) for GI tract colonization (B) and systemic dissemination (C) of the PA14 oprD mutants 6-4 and 8-1 versus WT and of the oprD mutant clinical strains 48-2 and 51-2 versus a related strain with intact oprD (48-1 and 51-1, respectively). Each point represents the CI for a single mouse. The medians are shown as a solid line. A CI > 1 indicates increased fitness. P. aeruginosa strains with an intact oprD gene were rarely recovered (0–5%) from the cecum and, consequentially, rarely recovered from the spleen when in competition with oprD mutants.

Further confirmation that loss of OprD enhanced GI colonization was obtained by evaluating the competitive colonization efficacies of WT PA14 against the two carbapenem-resistant strains from the ordered Tn library, PA14_Tn-oprD:6-4 and PA14_Tn-oprD:8-1. The PA14_Tn-oprD:6-4 and PA14_Tn-oprD:8-1 strains were stronger cecal colonizers than the PA14 WT strain. Following induction of neutropenia, more than 90% of the isolates recovered from the spleen were oprD mutants (Fig. 3 B and C). Comparative analysis of the competitive colonization capacities of the P. aeruginosa clinical isolates with an intact oprD gene, 48-1 and 51-1 with their corresponding related oprD mutants, 48-2 and 52-2, showed the oprD mutants constituted >95% of the recovered strains (Fig. 3 B and C). Thus, in all circumstances, loss of OprD and the consequent acquisition of carbapenem resistance resulted in enhanced GI colonization and systemic dissemination during neutropenia. To ascertain whether the overrepresentation of the oprD mutants was not merely due to enhanced survival as opposed to enhanced fitness of these strains, we colonized antibiotic-treated mice with monocultures of either WT PA14 or streptomycin- or tetracycline-resistant variants, or with the PA14_Tn-oprD::Spleen strain, and found no difference in the levels of any of these variants in the ceca after 6 d of colonization (Fig. S8). Thus, in the absence of competition from the OprD-deficient strains, WT strains were able to achieve levels of GI colonization comparable to that of the oprD mutants.

We next evaluated several phenotypes that could be related to the oprD deficiency and account for the increased in vivo fitness of the oprD mutants by analyzing in vitro survival of WT PA14 or Tn-insertion mutants in the presence of major host innate immune factors in the serum (21), as well as survival in the acidic conditions found in the GI tract (pH∼5 in mice) (22). After 180 min in 50% serum, all of the strains lacking OprD survived better than their isogenic or related parental or trans-complemented partners containing an intact oprD gene (Fig. 4A). Similarly, except for one strain (CS23 strain), all of the clinical strains with a low level of oprD transcription were significantly more resistant to serum-mediated killing compared with the clinical strains with a high level of oprD transcription (Fig. 4A). All of the oprD mutant strains tested were found to survive better at pH 5 than isogenic or related strains with an intact oprD gene (Fig. 4B). Furthermore, the PA14_Tn-oprD::Spleen strain displayed higher levels of cytotoxicity against murine macrophages than either this mutant complemented with PoprD or to the PA14 WT strain (Fig. 4C).

Fig. 4.

Phenotypes associated with increased in vivo fitness of oprD mutant strains. (A) Resistance to the antibacterial action of serum of P. aeruginosa strains with or without oprD deficiencies. OprD deficiency was confirmed by assessing the level of OprD expression (Fig. 2A) or the level of oprD mRNA transcription (strains CS1–S23; Dataset S1). The P values were determined by t test. (B) Survival for 60 min at pH 5 of either WT P. aeruginosa strains or those with mutations in the oprD gene. Bars indicate mean percent survival compared with isogenic or related strains with intact oprD. Error bars indicate a single SD of the data. *P value of less than 0.05 (t test) between the oprD intact and mutant strains; **P value of less than 0.05 between the oprD mutant and isogenic or related oprD complemented strain. (C) Role of the oprD gene in the cytotoxicity of PA14 against murine macrophages after 1 h of incubation. Two multiplicities of infection were tested: 20:1 (Upper) and 100:1 (Lower). The PA14 ΔexoU strain was used as a negative control. Error bars indicate the SD of the data.

To define a mechanism that could explain all these phenotypes associated with oprD deficiency, we used RNA-seq (23) to determine the transcriptional profiles of WT PA14 and PA14_Tn-oprD::Spleen strains after in vitro growth. As shown Fig. 5, the PA14 in vitro transcriptome was quite uniform in that it showed little variability throughout the whole genome. In contrast, the PA14_Tn-oprD::Spleen transcriptome showed considerable variability with elevated and reduced transcriptional levels detected in many different chromosomal locations compared with WT PA14 (Fig. 5). In total, 97 genes that were clearly transcribed in WT PA14 were transcriptionally silent in PA14_Tn-oprD::Spleen when grown in vitro (Dataset S2). In contrast, 60 genes had transcription levels increased more than 10-fold in the PA14_Tn-oprD::Spleen strain compared with WT PA14 (Dataset S3).

Fig. 5.

Global analysis of transcript levels in PA14 WT and Pa14_Tn-oprD::Spleen by RNA-seq. Sequencing reads for each of the 5,977 genes of PA14 WT (blue dots) or PA14_Tn-oprD::Spleen (green dots) corresponding to the transcript expression of each gene of these two strains grown in LB. A total of 97 genes expressed in PA14 WT were not expressed in PA14_Tn-oprD::Spleen strain.

Interestingly, increases in transcript levels were not found for the genes encoding most known virulence factors in the P. aeruginosa genomes (www.mgc.ac.cn/VFs/main.htm). To validate RNA-seq results, we performed individual qRT-PCR determinations that confirmed the lack of significant increases in the transcription level between the WT PA14 and the PA14_Tn-oprD::Spleen strains for genes representative of the following virulence factors: type 1, 2, 3, and 6 secretion systems, type IVa and IVb pili, exopolysaccharides alginate, PEL, and LPS, as well as rhl, quinolone quorum-sensing systems and adhesins, flagellins, rhamnolipids, pyochelin, pyoverdin, and pyocyanin (Dataset S4). Overall, transcriptional changes in genes needed for production of well-established P. aeruginosa virulence factors did not account for the enhanced fitness of OprD-deficient strains in GI colonization and dissemination. Thus, either the other transcriptional differences seen in the oprD mutant contributed to enhanced in vivo fitness or, alternatively, the loss of OprD itself causes posttranscriptional changes in phenotypic properties that enhance colonization and dissemination in the host.

We also examined published studies to ascertain if infections with oprD-mutant, carbapenem-resistant P. aeruginosa strains were associated with worse clinical outcomes. We hypothesized that the observed increased fitness of the oprD mutant strains in laboratory settings could be associated with more severe clinical outcomes in human infections. Peña et al. (24) reported significantly greater mortality after 7 and 30 d of infection in patients with carbapenem-resistant P. aeruginosa bloodstream infections, of which ∼95% were due to oprD mutations (25). In a related analysis, Cabot et al. (26) found that oprD mutations underlay the acquisition of high-risk infections due to extensively drug-resistant P. aeruginosa infections. In addition, a recent study analyzing different mechanisms of antibiotic resistance that occur in clinical isolates of P. aeruginosa causing severe bloodstream infections concluded that acquisition of resistance did not lead to decreased fitness (27). Consistent with the observations reported here that mutations in the oprD gene in P. aeruginosa can increase fitness for infection, other studies have reported that oprD-inactivating mutations can arise in the absence of carbapenem treatment (28), suggestive of a survival benefit conferred by OprD loss. Notably, some oprD mutations occur in isolates with MICs to imipenem or meropenem of 0.06–4 μg/mL, considered to be within the susceptible range (25), suggesting that increased fitness and full carbapenem resistance may be separable properties of the OprD protein. This might be explained by the findings of Eren et al. (29) who reported that multiple outer-membrane carboxylate channels (Occ) like OprD with different substrate specificities are found among various Gram-negative bacteria, raising the possibility that some strains of P. aeruginosa might possess additional Occ channels involved in carbapenem uptake.

Although almost all prior reports indicated no obvious fitness cost for causing severe infections by OprD deficient/carbapenem-resistant P. aeruginosa, our studies suggest there might be enhanced in vivo fitness of P. aeruginosa as a result of acquisition of oprD-inactivating mutations. Mechanistically, this is likely due to be the collective properties of OprD-deficient P. aeruginosa including enhanced serum resistance, a better ability to survive in hostile environments such as the acidic pH of the stomach, increased cytotoxicity against phagocytes, and other yet-to-be-discovered properties contributing to enhanced fitness. Notably, OprD-deficiency does not appear to confer any enhanced fitness in noninfectious settings, as a previous study (30) on 328 unrelated P. aeruginosa isolates from 69 localities in 30 countries on five continents collected from diverse clinical (human and animal) and environmental habitats over the last 125 y found no OprD-deficient strains among environmental isolates. However, the prevalence of OprD-deficient strains in CF and non-CF patients from the same study was 16% and 10%, respectively.

In a broader context, our findings imply that carbapenem treatment and the consequent selection of P. aeruginosa oprD mutant strains can lead to enhanced in vivo fitness and potentially to an increase in virulence. If additional studies validate an increased fitness and/or virulence of OprD-deficient P. aeruginosa, the findings might impact decisions related to infection control and treatment of P. aeruginosa infections. Thus, a careful evaluation might need to be made regarding the empiric use of carbapenems in hospitals, particularly in intensive care units and hematology departments, if there is a potential that treatment of patients with suspected but unconfirmed P. aeruginosa infection might, in the long run, be more harmful than beneficial.

Materials and Methods

The murine model of GI tract colonization and systemic dissemination by P. aeruginosa and the DNA preparation for Illumina sequencing were as described (8, 14). The PFGE was interpreted according to previously used criteria (20). The bacterial OMPs were detected by SDS/PAGE. Twitching motility was assessed using 1.5% agar LB plates and swarming assays performed using supplemented fresh plates of M9 minimal medium. The serum killing experiments were performed with pooled human serum. The cytotoxicity experiments used RAW264.7 cells. The library for the RNA-seq was prepared using Encore Complete Prokaryotic RNA-Seq DR Multiplex Systems kit (NuGEN), and cDNA fragmentation was done using Qsonica Sonicator Q800R. A full description of methods is available in SI Materials and Methods. Bacteria, plasmids, and primers used in this study are presented in Tables S1 and S2 and Dataset S5.