(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Hypermutator strains of Pseudomonas aeruginosa reveal novel pathways of resistance to combinations of cephalosporin antibiotics and beta-lactamase inhibitors [1]

['Augusto Dulanto Chiang', 'Bacterial Pathogenesis', 'Antimicrobial Resistance Unit', 'Lcim', 'Niaid', 'Nih', 'Bethesda', 'Maryland', 'United States Of America', 'Prashant P. Patil']

Date: 2022-11 

Hypermutation due to DNA mismatch repair (MMR) deficiencies can accelerate the development of antibiotic resistance in Pseudomonas aeruginosa. Whether hypermutators generate resistance through predominantly similar molecular mechanisms to wild-type (WT) strains is not fully understood. Here, we show that MMR-deficient P. aeruginosa can evolve resistance to important broad-spectrum cephalosporin/beta-lactamase inhibitor combination antibiotics through novel mechanisms not commonly observed in WT lineages. Using whole-genome sequencing (WGS) and transcriptional profiling of isolates that underwent in vitro adaptation to ceftazidime/avibactam (CZA), we characterized the detailed sequence of mutational and transcriptional changes underlying the development of resistance. Surprisingly, MMR-deficient lineages rapidly developed high-level resistance (>256 μg/mL) largely without corresponding fixed mutations or transcriptional changes in well-established resistance genes. Further investigation revealed that these isolates had paradoxically generated an early inactivating mutation in the mexB gene of the MexAB-OprM efflux pump, a primary mediator of CZA resistance in P. aeruginosa, potentially driving an evolutionary search for alternative resistance mechanisms. In addition to alterations in a number of genes not known to be associated with resistance, 2 mutations were observed in the operon encoding the RND efflux pump MexVW. These mutations resulted in a 4- to 6-fold increase in resistance to ceftazidime, CZA, cefepime, and ceftolozane-tazobactam when engineered into a WT strain, demonstrating a potentially important and previously unappreciated mechanism of resistance to these antibiotics in P. aeruginosa. Our results suggest that MMR-deficient isolates may rapidly evolve novel resistance mechanisms, sometimes with complex dynamics that reflect gene inactivation that occurs with hypermutation. The apparent ease with which hypermutators may switch to alternative resistance mechanisms for which antibiotics have not been developed may carry important clinical implications.

Funding: ADC, PPP, LB, AL, PPK, and JPD are supported by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases (NIAID). RAB reports extramural funding from NIAID under Award Numbers R01AI100560, R01AI063517, and R01AI072219. RAB is also supported in part by funds and/or facilities provided by the Cleveland Department of Veterans Affairs, Award Number 1I01BX001974 from the Biomedical Laboratory Research & Development Service of the VA Office of Research and Development, and the Geriatric Research Education and Clinical Center VISN 10. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Ceftazidime/avibactam (CZA) is an expanded spectrum antimicrobial consisting of an antipseudomonal cephalosporin (ceftazidime) and a novel beta-lactamase inhibitor (avibactam) [ 23 – 25 ]. The development of CZA added a valuable option against many MDR P. aeruginosa, given that avibactam is a potent inhibitor of the P. aeruginosa PDC (AmpC) cephalosporinase, which is often overexpressed in beta-lactam-resistant P. aeruginosa isolates [ 2 ]. However, CZA-resistant P. aeruginosa isolates were described within a year following its introduction, mostly associated with the development of point mutations in the chromosomal PDC cephalosporinase and overexpression of the RND-class MexAB-OprM efflux systems [ 26 – 35 ]. Recently, we demonstrated that MMR-deficient P. aeruginosa isolates with an inactivated mutS gene can rapidly develop high-level CZA resistance in an in vitro adaptive evolution model [ 19 ]. In that study, lineages derived from a laboratory strain P. aeruginosa MPAO1 (MPAO1-WT) and an MPAO1 strain containing a transposon insertion in the mutS gene (MPAO1-mutS Tn ) were passaged through a gradient of increasing CZA concentrations. While all lineages evolved CZA resistance, the MPAO1-mutS Tn isolates developed clinical levels of CZA resistance more rapidly than the MPAO1-WT isolates (median passages to MIC = 16 μg/mL for MPAO1-mutS Tn = 2.5, range 2 to 4; versus MPAO1-WT = 15, range 7 to 16) [ 19 ]. In the present work, we employ large-scale genomic and transcriptional analyses in combination with genetic engineering to study the mechanistic basis of resistance to cephalosporin/beta-lactamase inhibitor combination antibiotics in these MMR-deficient hypermutators.

Pseudomonas aeruginosa is a leading cause of serious infections in humans. A distinguishing feature of this pathogen is its remarkable ability to develop resistance to most classes of antibiotics through chromosomal mutations, without the need for horizontal gene transfer [ 1 – 3 ]. Consequently, multidrug resistant (MDR) P. aeruginosa can emerge rapidly with treatment in a number of important clinical contexts, and an understanding of the mechanisms by which this resistance evolves is critical to developing the next generation of antipseudomonal agents. A frequently observed phenomenon during chronic infections with P. aeruginosa is the development of hypermutation, usually caused by inactivation of genes involved in DNA repair [ 4 – 6 ]. The resulting DNA repair deficiencies can elevate spontaneous mutations rates by 100- to 1,000-fold, with distinct mutational spectra depending on the specific DNA repair pathway affected [ 7 , 8 ]. Hypermutation due to deficiencies in the mismatch repair (MMR) system in particular has been shown to accelerate the emergence of antimicrobial resistance (AMR) both in vitro and in vivo and is thus of considerable clinical concern [ 9 – 11 ]. Hypermutator P. aeruginosa isolates have been found in up to 50% of cystic fibrosis respiratory specimens, and hypermutation has been linked to development of MDR phenotypes over periods of years to decades in this context [ 4 , 12 – 18 ]. Recent work has also demonstrated that hypermutation can lead to the evolution of antibiotic resistance over the time course of days in the context of acute systemic infection [ 19 ]. Though the link between hypermutation and the development of AMR is well documented, the detailed mutational steps leading to resistance to many different classes of antibiotics in hypermutators have not been well characterized [ 20 – 22 ].

A caveat to these calculations is that they are based only on NCBI genome sequences without functional verification, and we cannot rule out sequencing errors and assembly artifacts, though the prevalence would have to be extraordinarily high to explain these findings. Another caveat is that isolates from cystic fibrosis respiratory specimens, where hypermutators are common, may be overrepresented in the genomic databases [ 4 , 12 – 18 ]. However, even if hypermutators and MexAB-OprM efflux deficient isolates are somewhat less common than our estimates of ≥2.3% and ≥3.9%, this would still suggest that the co-occurrence of hypermutation and preexisting MexAB-OprM functional deficiencies may be substantially more common than appreciated. We thus asked how many assemblies in the database would be predicted to be both putative hypermutators and have a nonfunctional MexAB-OprM. Surprisingly, 91 out of 260 putative hypermutators (35%) were also found to have at least 1 sequence variant suggestive of an inactive MexAB-OprM, raising the possibility that hypermutation and MexAB-OprM inactivation may in fact occur frequently together. Once again, in order to adjust for clonality in this subset of assemblies, we conservatively assumed each individual disruptive variant (n = 47) to represent a single clonal lineage, obtaining a lower bound estimate of 47/260 (18%). This analysis suggests that concurrent hypermutator phenotypes and MexAB-OprM efflux pump disruption coexist in a small, but easily detectable, proportion of the clinically relevant P. aeruginosa population. These isolates may be expected to evolve alternative resistance mechanisms under selection as did the isolates in the experiments in this work. However, detailed examination of resistance mechanisms in these isolates is limited due to lack of susceptibility testing data. Out of 7,492 isolates in the NCBI Pathogen Detection database, only a small minority (149 genomes, 2%) had any antimicrobial susceptibility data, with 27 (0.3%) having CZA susceptibility data available. Only 7 out of the 260 putative hypermutator isolates as defined above had AST phenotypic data available, and only 1/7 had CZA data. These data do not allow specific conclusions about resistance phenotypes in these isolates to be drawn.

To evaluate how commonly hypermutation phenotypes and MexAB-OprM efflux pump deficiency occur in P. aeruginosa, we used the NCBI Pathogen Genome Database to search for disruptive mutations (e.g., frameshifts and stop codons, mutations that result in complete loss or severe truncation of the affected protein) in the corresponding genes (see Methods ). Inactivation of a number of different genes can result in hypermutator phenotypes in P. aeruginosa, including MMR genes mutS, mutL, and uvrD, base excision repair (BER) genes mutT, mutM and mutY, and others. We identified severely disruptive variants in at least one of these genes in 260 out of 6,805 assemblies (3.8%) ( S9 Table ). Given that some isolates can be clonal, and this might inflate this estimate, a conservative lower bound estimate of 155/6,805 assemblies (2.3%) was calculated assuming that all repeated variants are clonal. We similarly estimated the frequency of disruptive variants in mexB, mexA, or oprM that would be expected to result in truncated or mistranslated proteins with functional inactivation of the MexAB-OprM efflux pump. We found 86 unique disruptive variants in the mexA gene, 148 in the mexB gene, and 33 in the oprM gene ( S10 Table ). In total, 629 (9.2%) isolates had a variant predicted to be severely disruptive in at least 1 of the 3 genes resulting in a nonfunctional MexAB-OprM complex. Again, a conservative lower bound estimate of 267/6,820 (3.9%) assemblies was calculated assuming that all isolates carrying a given variant represented a single clonal lineage.

The introduction of the MexB W753R substitution significantly reduced the aztreonam MIC in all 3 strains, consistent with disruption of MexAB-OprM function. In the MexAB-OprM-overexpressing 1C12-1 isolate, the introduction of MexAB W753R also decreased the CZA and ceftazidime MIC to pre-evolved WT baseline levels ( S11 Fig ), suggesting that this isolate’s CZA and CAZ resistance were entirely due to MexAB-OprM overexpression, and that the MexB W753R mutation was fully inactivating. These findings support the hypothesis that MexB W753R was not compatible with the evolutionary mechanisms the WT isolates used to generate resistance to CZA under selection, and suggests that this early inactivating mutation in the MexAB-OprM efflux pump, a primary mediator of CZA resistance, potentially drove an evolutionary search for alternative resistance mechanisms. Interestingly, the CZA and CAZ resistance were not affected by this mexB mutation when introduced in a mexV -82:T>C + MexW E36K background ( S11 Fig ), providing further support that MexVW can confer cephalosporine resistance through a MexAB-OprM-independent mechanism.

As noted above, the MPAO1-mutS Tn strain was found to differ from the MPAO1-WT strain by 54 variants (50 SNVs and 4 indels), with 33 of these variants present at a majority frequency in the starting stock and an additional 21 that occurred during the first passage, but before CZA resistance appeared ( S1 Data ). All 50 of the SNVs were transitions, consistent with the expected outcome of MutS-deficient hypermutation. Though these mutations in aggregate do not account for the CZA resistance that subsequently developed, it was possible that they may have played unappreciated roles that influenced the likelihood of evolving resistance through major efflux pathways. Importantly, these mutations included a transition variant in the mexB gene in the MPAO1-mutS Tn isolate (resulting in the W753R substitution), and we hypothesized that this substitution may have inactivated the MexAB-OprM efflux pump, blocking further evolution that would have used up-regulation of this transporter to confer resistance. To test whether this mutation would have been compatible with the MexAB-OprM-based CZA resistance that ultimately evolved in the WT strains, we introduced it into 3 different strains: (1) an MPAO1-WT strain; (2) an evolved CZA-resistant WT isolate with increased MexAB-OprM expression (1C12-1); and (3) an engineered MPAO1 mexV -82:T>C + MexW E36K background. The aforementioned 1C12-1 isolate was obtained during serial CZA passage and contains an intergenic SNV between mexR and mexA (471932 C>T), upstream from the MexAB-OprM operon. This intergenic variant appears to be responsible for the increase in MexAB-OprM expression seen in this lineage by RNA-seq ( Fig 3B ).

Given that mutations in the genes encoding the MexVW efflux pump have not commonly been associated with resistance to cephalosporin/beta-lactamase inhibitor combinations, we sought to understand the frequency with which these variants occurred in publicly available P. aeruginosa genomes. BLAST was used to search for sequence homologues of the mexW CDS and intergenic sequence upstream from mexV in a set of 7,492 P. aeruginosa genomes from the NCBI Pathogen Detection database [ 51 ] (see S8 Table and Methods ). Neither of the 2 specific variants identified in our study was found in this search, and the neighboring amino acid and nucleotide sequences were relatively conserved ( S8 Table ). The apparent rarity of the specific mutations uncovered in this adaptive evolution experiment suggests that sequence space may harbor a large universe of potential resistance mechanisms available to hypermutators.

( A ) Gradient diffusion (E-test) MICs for MPAO1-WT, mexV -82:T>C, MexW E36K, and double mutant (n = 10 per antibiotic/genotype combination). The boxes display the median with the lower and upper hinges corresponding to the first and third quartiles. Brackets with asterisks above the plot indicate statistical significance with respect to MPAO1-WT (Wilcoxon 2-sided p-value < 0.05). All pairwise comparisons between mutants were statistically significant as well ( S7 Table ). ( B ) Growth curves of MexVW engineered strains at 37°C in LB broth with a range of ceftazidime concentrations with fixed 4 μg/mL avibactam concentration. Lines represent the median OD 600 of 7 biological replicates per time point, with shaded envelope representing the standard error of the mean. ( C ) Scatterplot showing the relative contribution of the mexV -82:T>C and MexW E36K mutations to resistance to each antibiotic. Each colored point represents the mean ratio of gradient diffusion MIC in the corresponding mutant strain to that in the parental MPAO1 strain, with error bars represent the corresponding standard error of the mean. A dashed line of equality is included. The underlying data to generate this figure can be found in S2 Data . AZN, aztreonam; CAZ, ceftazidime; CTX, cefotaxime; CZA, ceftazidime/avibactam; FEP, cefepime; MEM, meropenem; PIP, piperacillin; PIP/TAZ, piperacillin/tazobactam; C/T, ceftolozane/tazobactam; WT, wild type.

Based on the foregoing combination of genomic and transcriptomic analysis, we focused on the MexVW efflux pump as potentially contributing to CZA resistance in the hypermutators. To test the effects of the 2 identified mutations, we introduced them individually and in combination into an MPAO1-WT strain to generate 3 derivative strains: MPAO1 mexV -82:T>C, MPAO1 MexW E36K, and MPAO1 mexV -82:T>C + MexW E36K. We performed susceptibility testing to a variety of classes of antimicrobials and categories of beta lactams on all 3 strains (Figs 4A and S9 and S5 Table ). This demonstrated that each mutation individually conferred a 1.5- to 2-fold increase in CZA MIC, and the 2 mutations in combination conferred a statistically significant 4- to 6-fold increase in MICs to ceftazidime, CZA, C/T, and cefepime. Additional testing demonstrated no changes in susceptibility to aminoglycosides, meropenem or fluoroquinolones with respect to WT MPAO1 ( S5 Table and S9 Fig ). Standard tests of AMR were supplemented with an assessment of growth kinetics in the presence of CZA. These results corroborated the ability of engineered mutants to grow under higher CZA concentrations, with the double mutant being able to grow at CZA concentrations 4 times that of the wild type ( Fig 4B ). The relative contributions of the mexV -82:T>C and MexW E36K to resistance to each of the tested antibiotics are displayed in Fig 4C . Extended antimicrobial susceptibility testing was also performed on the evolved hypermutator isolates for comparison, with the caveat that interpretation is complicated by the growth of satellite colonies within the zones of inhibition (see S10 Fig and S6 Table ).

Three lineages of MPAO1-WT and MPAO1-mutS Tn were selected for transcriptomic analysis and 3 biological replicates from 1 isolate per passage underwent RNA sequencing. (A) A clustered heatmap of the top 750 most variable genes across CZA-resistant isolates is shown. Rlog normalized levels of expression of each gene (rows) were scaled within each row. Isolates (columns) are grouped by lineage and arranged by genotype with WT on the left and hypermutators on the right. Within each lineage, isolates are arranged from left to right in order of increasing number of passages under CZA selection, with the corresponding MIC indicated at the bottom. 1X and 2X represent the MPAO1-WT and MPAO1-mutS Tn parental strains of the respective lineages. (B) LFC with respect to their corresponding ancestor of the terminal MPAO1-WT isolates (first 3 columns) and CZA-adapted MPAO1-mutS Tn isolates (last 9 columns). The rows represent 6 major beta-lactam resistance genes in P. aeruginosa from 4 major resistance pathways (encoding the MexAB-OprM efflux pump, OprD outer membrane porin, PDC beta-lactamase, and penicillin-binding-protein 3 FtsI). Data for 2 differentially expressed RND efflux pumps in the MPAO1-mutS Tn lineages (MexVW and MexGHI-OpmD) are additionally shown. (C) Expression levels of RND efflux pumps and outer membrane porins in MPAO1-mutS Tn lineages during the course of the CZA adaptation experiment. The vertical axis represents the LFC of expression levels compared to the early isolates, and the horizontal axis shows the passage number. CZA MIC is indicated with color tiles under the axis. Lines are colored and labeled if the respective gene reached a LFC > 1.0 in the terminal isolates of the lineage. The underlying data to generate this figure can be found in S2 Data . CZA, ceftazidime/avibactam; MMR, mismatch repair; WT, wild type.

Comparison of transcriptomes demonstrated substantial differences between the evolved CZA-resistant MPAO1-WT and MPAO1-mutS Tn isolates. While the overall numbers of differentially expressed genes (DEGs) were not significantly different between the terminal WT and MMR-deficient groups ( S5 Fig ), principal component analysis ( S6 Fig ) demonstrated that their content differed, and transcriptomes from the 2 groups of isolates were well separated in the first 2 principal components. These global differences in DEG content between WT and MMR-deficient lineages is also evident from a heatmap representation ( Fig 3A ). Analysis of expression of known resistance genes revealed up-regulation of the MexAB-OprM efflux pump components in the MPAO-WT transcriptomes (mexA >1.5 and mexB >1.3 log 2 fold change (LFC) in all 3 WT lineages versus WT ancestor, p < 0.001) ( Fig 3B ). These expression changes provide an explanation for CZA resistance in the WT isolates that is consistent with both initial genomic analysis and previously described mechanisms. In contrast, the MPAO1-mutS Tn isolates did not show changes in expression of mexA, mexB, oprM or in other previously described major resistance genes, including oprD, ftsI, or the PDC beta-lactamase gene ( Fig 3B ). Additionally, there were no baseline differences in expression of the 6 major resistance genes listed above between the MPAO1-WT and MPAO1-mutS Tn , using thresholds of p adjust < 0.01 and/or LFC >1 ( S7 Fig ). To narrow down possible mechanisms conferring CZA resistance, we identified 91 genes that were consistently up-regulated and 142 consistently down-regulated in MPAO1-mutS Tn lineages ( S8 Fig ). This analysis revealed an important detail: The MexVW operon demonstrated increased expression with early evolution of CZA resistance in all 3 MPAO1-mutS Tn lineages, and this coincided with acquisition of fixed mutations in this operon, including the E36K mutation in MexW with moderate to high conservation and impact scores ( Fig 2 ). For comparison, we examined the expression of other RND transporter systems in these isolates and found that MexVW was the only transporter that was overexpressed consistently across lineages ( Fig 3C ). This increased expression of mexV and mexW was statistically significant when comparing the MPAO1-mutS Tn terminal isolates with the MPAO1-WT terminal isolates (LFC > 3; p adjust < 4 × 10 −15 ). We also noted that mexG, mexH, and mexI were down-regulated in all MPAO1-mutS Tn lineages.

Given the surprising finding that the CZA-resistant MPAO1-mutS Tn lineages largely lacked fixed mutations in genes known to be involved in CZA and third generation cephalosporin resistance, we next sought to examine whether the observed mutations had resulted in overexpression of known resistance genes such as the MexAB-OprM efflux pump and PDC beta lactamase through unexpected transcriptional control mechanisms. We thus performed RNA-seq on 3 evolved CZA-resistant MPAO1-WT isolates representing 3 different WT lineages, as well as on a series of MPAO1-mutS Tn isolates with different MICs over the clinically relevant range from the 3 hypermutator lineages and their respective ancestors. At total of 61 mid-log phase RNA-seq datasets were obtained from 22 isolates with an average of 6.1 million reads per transcriptome ( S3 Data ).

To explore whether the CZA-adapted MMR-deficient isolates demonstrated gross growth defects due to mutational accumulation, we measured growth curves in LB broth for CZA-evolved isolates at different points along antibiotic adaptation from all 3 MPAO1-mutS Tn lineages and 1 MPAO1-WT lineage for comparison (Methods, S4 Fig ). While there is a reduction in the growth rate in lineage 2D at passage 12 and a progressive trend to slower growth along passages in lineage 2A, the growth curves of most isolates were not substantially different from those of MPAO1-WT 1A.

To examine whether some of the observed variants may have been selected by conditions in the serial passaging experiment other than antibiotic pressure, we compared variants that occurred under CZA selection to the variants that occurred under the same conditions in the absence of CZA for 11 sequential passages. We found 5 shared genes with at least 1 variant among isolates passaged with and without CZA ( S4 Table ). Out of 12 variants in these 5 shared genes (pdtA, amaB, PA2462, pilY, and rfaE), only 1 variant was identical between isolates passaged with and without CZA (G insertion at position 1114264, in the amaB gene). The pdtA and PA2462 genes have unusually long open reading frames (>10 kb), which increases the likelihood of observing repeated mutations in these genes by chance and might explain why mutations were observed in these genes under both conditions. The 5 variants seen in pdtA, amaB, PA2462, and pilY consisted of short indels in GC-rich contexts, which undergo higher rates of mutation in the mismatch-repair deficient P. aeruginosa isolates [ 49 , 50 ]. None of these variants were shared between multiple evolved CZA-resistant lineages.

Given the high mutation rates in the hypermutators, a large number of background mutations that do not improve fitness mutations are expected to be co-selected along with mutations that improve fitness. We thus expect that some proportion of the observed fixed variants (possibly the majority) did not contribute directly to antibiotic resistance. As a first step towards understanding which subset of fixed variants in the MPAO1-mutS Tn isolates were those contributing to antibiotic resistance, we extended the comparative genomic analysis by adding measures of relative conservation and predicted impact at each position [ 40 , 48 ] ( Fig 2 , Methods). This analysis predicted a moderate to high impact for a number of fixed variants in the MPAO1-mutS Tn lineages, indicating a number of potential functional targets. An additional indicator of the importance of a specific variant or gene target for CZA resistance is independent selection and fixation in different lineages. Overall, 8 genes were mutated in more than 1 MPAO1-mutS Tn lineage: PA1545 (hypothetical protein belonging to Pseudomonas ortholog group 3815) [ 40 ]; glucose-6-phosphate isomerase, pgi; pdtA (phosphate depletion regulated TPS partner A); the β-subunit of RNA polymerase, rpoB; mexW (subunit of the MexVW RND-type efflux pump); the phoQ subunit of the 2-component sensor PhoP/Q; ftsI (PBP3); and the ATP-binding subunit of the Clp protease, clpA ( S3 Fig ). An additional noncoding variant was shared across the 3 lineages, located 82 nucleotides upstream from the MexVW operon (position 4903384 in the PAO1 reference). Five of the variants in these 8 common targets (RpoB D521G, Pgi G164D, a frameshift in PA1545, MexW E36K, and a SNV 82nt upstream from MexV) were identical between at least 2 lineages, suggesting that these variants may have been present at low proportions in the heterogeneous starting population from which each lineage was derived in the microevolution experiment, as opposed to independent de novo occurrence during adaptation. Another possibility is early-stage cross-contamination between lineages in the experiment. We were unable to find supportive genomic evidence identifying a cross-contamination event to explain the sharing of these mutations, though this case may be hard to distinguish from a variant that is shared by the starter populations at low frequency. Nevertheless, the fact that these variants had initially low population frequency but subsequently became dominant in multiple lineages under increasing CZA selection suggests that they may have provided fitness advantages.

We next looked at the fixed variants in the MPAO1-mutS Tn lineages as CZA resistance evolved. Strikingly—and in contrast to the behavior of the WT isolates—almost all fixed variants acquired as CZA resistance evolved through the clinical breakpoint of 16/4 μg/mL were located in genes that have not been previously described to be involved in CZA or third generation cephalosporin resistance ( Fig 2 ). Though mutations in these lineages did not occur in the “classical” resistance genes, we identified 9 mutations across 6 genes for which some level of prior experimental evidence exists supporting a role in resistance to third generation cephalosporins: PA1436 (mexN of MexMN efflux pump), acsB, ftsL, clpA, argJ, PA0478-PA0479 [ 30 , 42 – 47 ]. All of these mutations first appeared in passages with CZA MIC equal to or greater than the clinical resistance breakpoint of 16 μg/mL, and thus contributed to resistance only at these higher MICs. Notably, one of the 54 “early” variants present in the parental strain of the MPAO1-mutS Tn lineages was located within MexB (W753R). This variant, discussed more fully below, was present by the end of the first day of passage in all sequenced isolates.

Fixed variants are plotted versus passage (horizontal axis tick marks) and CZA MIC (μg/mL), with a filled tile indicating presence of the variant at the corresponding passage. The shade of blue represents the proportion of isolates (allele frequency) in a given passage carrying the corresponding variant. Up to 3 isolates per lineage per passage underwent WGS. The sidebars to the right of each plot represent functional characteristics of the variant and its protein targets as follows: coding/noncoding (NC) variant; variant impact (IM) as predicted by SnpEff; conservation score (CN) as relative conservation of a mutated AA position within each protein defined as a within-CDS percentile of the Jensen–Shannon divergence scores calculated over a set of 100 Pseudomonas genomes; breadth (BR) as the percentage of 100 Pseudomonas genomes in which the CDS is present; average AAI of the CDS with respect to its orthologs. Genes marked with a Ψ symbol were also mutated in the no antibiotic control experiments. Genes marked with a diamond ◆ are those for which experimental evidence supports a role in third generation cephalosporin resistance. An asterisk (*) represents a passage with no sequenced isolates. FS = Frameshift variant. ΔOprN = 7-kb deletion including the oprN gene (partial deletion of mexF, oprN, PA2496, PA2497, PA2498 (yahD), PA2499 (ykoA), PA2500 (cynX), PA2501 and part of PA2502; PAO1 Reference coordinates 2,812,525–2,819,849). CZA MIC was determined by E-test. The underlying data to generate this figure can be found in S2 Data . AAI, amino acid identity; CZA, ceftazidime/avibactam; MMR, mismatch repair; WGS, whole-genome sequencing.

The majority of fixed variants in lineages from both WT strains (16/21 in MPAO1-WT and 9/14 in PT) were located within genes previously described in association with beta-lactam resistance, including efflux pumps and their transcriptional regulators (mexR, MexA, MexB, nalD, mexE, mexF, oprN) or in the Ω-loop of the PDC chromosomal cephalosporinase ( Fig 2 and S3 Table ). Of note, a 7 kilobase deletion involving 8 genes, including part of mexF, oprN, PA2496, PA2497, PA2498 (yahD), PA2499 (ykoA), PA2500 (cynX), PA2501, and part of PA2502 (PAO1 Reference coordinates 2,812,525 to 2,819,849), emerged in passage 6 in the MPAO1-WT 1D lineage. Lineage 3B in the PT isolate acquired a 21-nucleotide deletion in the gene encoding the PDC beta-lactamase gene, which has previously been demonstrated to confer CZA and ceftolozane-tazobactam (C/T) resistance [ 27 ]. Mutations in the ATP-binding subunit of the Clp protease clpA, as well as in the PhoP/PhoQ phosphorelay system were also observed in independent lineages (clpA in all WT, PT, and mutS lineages except WT lineage 1B; phoQ in lineages 1A, 1D, 2A, 2B, and 2D, see Fig 2 and S3 Table ).

( A ) Proportions and types of fixed variants acquired during the course of CZA adaptation for each genotype. The total number of fixed variants acquired across lineages per genotype is indicated in parenthesis. ( B ) Evolutionary conservation of genes with fixed variants acquired in MPAO1-mutS Tn- (yellow) and MPAO1-WT (red) CZA adaptation experiments in the Pseudomonas genus. Each P. aeruginosa PAO1 CDS is plotted as a dot representing average amino acid identity and the proportion of genomes containing the given CDS (breadth of coverage) among a set of 100 complete Pseudomonas genomes (see S6 Data and Methods ). Genes without fixed mutation are represented in blue. The marginal density plots and histograms show the distribution of a corresponding conservation measure in all CDSs (gray) and genes containing fixed variants (blue), respectively. The underlying data to generate this figure can be found in S2 Data . CZA, ceftazidime/avibactam; WT, wild type.

We identified a total of 14, 21, and 139 fixed variants that emerged under CZA selection in the PT, MPAO1-WT, and MPAO1-mutS Tn lineages, respectively, representing a mean of 3.5, 5.3, and 46.3 fixed variants per lineage (Figs 1A and S1 ). In agreement with previous work [ 36 – 39 ], fixed mutations in the MutS-deficient lineages demonstrated a strong transition bias (119/120 SNV), compared to the WT lineages (5/11 SNV; Fisher’s exact test p-value < 0.0001). To evaluate the global properties of the observed fixed mutations, we applied a comparative genomics approach using the Pseudomonas ortholog database and compared average amino acid identity versus gene conservation for the coding sequences with non-synonymous substitutions [ 40 , 41 ] (see Methods ). This analysis demonstrated a broad distribution of mutated genes in the hypermutator lineages and confirmed that mutations were not confined to poorly conserved accessory genes but were distributed across genes at all conservation levels ( Fig 1B ). Further Gene Ontology analysis of the variants in the hypermutator lineages demonstrated enrichment in a number of functional classes including purine nucleotide metabolism and cell wall biosynthesis ( S2 Fig ).

To characterize mutations that emerged under CZA selection in Khil and colleagues [ 19 ], we performed Illumina whole-genome sequencing (WGS) of evolved isolates. Three lineages (MPAO1-WT, MPAO1-mutS Tn , and a comparator clinical P. aeruginosa isolate, PT) were grown in CZA concentrations ranging from 0.5 μg/mL to 256 μg/mL (see Methods and Khil and colleagues [ 19 ] for details). Liquid cultures that displayed detectable growth were plated, and up to 3 colonies per passage were selected for sequencing with a focus on isolates straddling the clinically relevant CZA concentrations. Following quality control, WGS data from 303 isolates were included in the analysis, with a median of 3.7 million reads and median coverage of 51× per sample ( S1 and S2 Tables). To prioritize variants that were likely targets of selection and to account for potential mutations introduced during lab manipulations prior to sequencing in hypermutator strains, we focused on SNVs that were retained in terminal isolates as CZA resistance evolved. These “fixed” variants were defined as those that were found in at least 2 isolates of a given lineage including at least 1 terminal isolate, but that were not present in the parental (starting) strain of the lineage. For this purpose, the set of variants present in the parental MPAO1-mutS Tn was assessed by sequencing of the bulk starting bacterial stock solution and 6 isolates spanning 2 of the lineages at the end of the first passage day. This approach was chosen to take into account the high spontaneous mutation rate in these lineages and revealed 54 “early” variants that were dominant before the end of the first passage day ( S1 Data , Methods). Notably 50/50 (100%) of the early SNVs that occurred were transitions, consistent with a MutS deficiency-driven hypermutation spectrum [ 36 – 39 ].

Discussion

In this work, we have employed in vitro adaptive evolution in combination with genomic sequencing, transcriptional profiling, and genetic engineering to study how resistance to cephalosporin/beta-lactamase inhibitor combinations evolves in MMR-deficient P. aeruginosa. Sequencing of in vitro evolved CZA-resistant MMR-deficient isolates revealed that—in contrast to WT isolates—almost all fixed variants were located within genes not previously associated with cephalosporin resistance. Additionally, while WT isolates demonstrated significant up-regulation of the MexAB-OprM efflux pump, providing an explanation for CZA resistance [52,53], the MMR-deficient isolates did not show altered expression of MexAB-OprM or of 6 other major beta-lactam resistance genes. Further investigation revealed that these isolates had paradoxically generated an early inactivating mutation in the MexAB-OprM efflux pump, likely driving an evolutionary search for alternative resistance mechanisms. The mutation was a transition variant within the mutational spectrum expected for mutS deficiency, similar to the 49 other early SNVs, and therefore likely directly acquired as a consequence of the MutS-deficient hypermutation. The affected residue in MexB is located in the second large periplasmic loop of the protein, in a region postulated to be important for trimerization [54], potentially explaining its effect on function.

In addition to a number of other changes, expression of the less well-studied MexVW efflux pump machinery was determined to be up-regulated, and mutations were present upstream of the MexVW operon and within the coding sequence of MexW. Engineering of these 2 mutations in a WT genetic background increased the MICs to ceftazidime, CZA, C/T, and cefepime by 4- to 6-fold, demonstrating a previously unappreciated mechanism of resistance to cephalosporin/beta-lactamase inhibitor combination antibiotics in P. aeruginosa. These observations suggest that the MexVW efflux pump operon can be mutationally modified and used as a compensatory resistance mechanism to overcome MexAB-OprM loss or inhibition. Recent studies have also shown that P. aeruginosa strains lacking 6 major RND efflux pumps ΔmexAB-oprM, ΔmexCD-oprJ, ΔmexX, ΔmexEF-OprN, ΔmexJKL ΔtriABC can utilize the minor RND efflux pumps including MuxABC-OpmB and MexGHI-OpmD to gain resistance to certain antibiotics [55], and other recent work has demonstrated the previously unappreciated role of several less studied efflux pump complexes in conferring resistance to novel beta-lactam/beta-lactamase inhibitor combinations [56]. In aggregate, these findings may have implications for the development of efflux pump inhibitors for clinical use.

The MexVW complex is a member of the resistance-nodulation-division (RND) family of efflux pumps and was originally described in association with resistance to fluoroquinolones, tetracyclines, chloramphenicol, cefpirome, and erythromycin by Li and colleagues [57], but has not been previously described as involved in conferring resistance to cephalosporin/beta-lactamase inhibitor combinations such as CZA or C/T. The mexV and mexW genes appear to form a single operon without an adjacent outer membrane porin gene, though the work of Li and colleagues suggested an association with OprM to form a tripartite MexVW-OprM complex [57]. Our analysis of transcription uncovered a progressive increase in expression of the components of the MexVW efflux pump with increasing MIC that was not observed in the WT counterparts. The exact mechanism by which the MexVW mutations examined in this study confer resistance to cephalosporins remains to be defined, but direct efflux by this pump appears likely. In addition to the up-regulation of MexVW, we observed down-regulation of the MexGHI-OpmD RND efflux pump across the MMR-deficient lineages. This pump is under the control of the redox-responsive SoxR regulon [52], but a potential role in CZA resistance would require further evaluation.

Comparison of the other mutations observed in our study with previous publications provides additional insights. Work on in vitro evolution of ceftazidime (CAZ) and CZA resistance in P. aeruginosa has been performed in a PA14 genetic background by Sanz-Garcia and colleagues [30]. While strains in their work developed many variants in previously described resistance genes (including nalD, mexB, ftsI, and large chromosomal deletions containing the MexXY-OprM operon [1,2,33,58–61]), the authors describe genes that were targeted only in lineages exposed to CZA but not CAZ. These included the RND efflux pump components mexM (PA14_45910) and mexN (PA14_45890). We observed a relatively conservative A44V substitution within the mexN (PA1436) gene in one of the hypermutator lineages. Determining whether the MexMN efflux pump plays a role in CZA resistance would require further study.

Other work in WT and MMR-deficient backgrounds also reported an enhanced rate of evolution of resistance. A study by Cabot and colleagues [8] analyzed variants that developed following in vitro exposure to CAZ (without avibactam), meropenem, and ciprofloxacin. Ceftazidime-resistant isolates acquired mutations in targets previously described in association with beta-lactam resistance, including dacB, ampD, PDC, ampR, and galU [8]. These ceftazidime-resistant isolates largely did not acquire cross-resistance to carbapenems, consistent with an increase in the PDC beta-lactamase expression as a mechanism of resistance. Notably, this evolutionary pathway to resistance may be less available in the presence of a beta-lactamase inhibitor such as avibactam. There were no exact variants shared between the CAZ-resistant isolates observed in our study and those of Cabot and colleagues, though we did observe mutations in shared genes: dnaX, fdnG, pgi, and pvdL. As these genes were mutated in the background of a very large number of total mutations, their role in CAZ resistance would need to be further studied before any conclusions are drawn regarding their potential contribution. Similarly, in a study by Gomis and colleagues [62], isolates passaged under selection with imipenem and imipenem-relebactam evolved mutations in previously described genes associated with beta-lactam resistance, in particular nonsense mutations in oprD. The addition of relebactam decreased the rate of acquisition of resistance and was associated with an increase in mexB expression as opposed to an increase in PDC expression as observed in the imipenem-resistant isolates. This finding would again be consistent with the effect of the beta-lactamase inhibitor on the PDC beta-lactamase, impeding opportunities for the development of resistance through this pathway, similar to conditions seen in our experiments. Indeed, in the presence of PDC inhibition, we observed evidence that MexAB-OprM overexpression contributed substantially to resistance in the WT isolates, a pathway to resistance not available to the hypermutators with the MexB W753R substitution.

In this study, we performed susceptibility testing with CZA on 1 sequenced isolate per passage to establish MICs by a standard method (E-test). Discrepancies were present between the MIC calculated from the highest CZA concentration in which growth of the population met the OD 600 threshold in the serial passage experiment and the MICs of individual isolates cultured from this passage (S4 Data). In these cases, the isolate MIC (measured by E-test) was usually higher than the MIC calculated from the greatest CZA concentration in which there was growth passing the threshold, most often by a 2-fold dilution, though in some cases more substantial discordances were observed. These discrepancies are in part expected, as the serial passage experiment used different conditions than those used for the E-test (growth in LB broth with a quantitatively different starter inoculum, measured with an arbitrary OD 600 threshold cutoff). Additionally, E-tests were performed on single isolates, whereas heterogeneous populations were measured by well growth. It should also be noted as a general point that our experiments were not designed to identify the exact association between new variants and resistance phenotypes at all intermediate passage steps. Given the large number of total mutations, we focused only on those variants that were retained and fixed as resistance to CZA developed. It is possible that transient variants that did not fix contributed incrementally to resistance at intermediate stages, and our analysis would not have captured these variants. Similarly, we cannot rule out the possibility of epigenetic or other nongenetic mechanisms of resistance playing a role in the observed increased in MIC in our selection experiment [63,64].

We believe the findings in this work have important potential clinical implications. Firstly, clinicians should be aware of the possibility that hypermutator P. aeruginosa strains will be encountered in the hospital, and that these isolates may rapidly and predictably develop resistance under CZA selection. Secondly, in the course of such antibiotic selection, hypermutators may dynamically modify the set of available evolutionary pathways through mutational inactivation of classical resistance genes, leading to the evolution of novel resistance mechanisms for which antibiotics have not been developed. Alternatively, mutations leading to hypermutation may occur within the background of preexisting MexAB-OprM deficiency, which we found was relatively common among P. aeruginosa genomes in the NCBI Pathogen Database. These mechanisms may also confer cross resistance to other related antibiotics, including cefepime and C/T, as seen in this study. At present the exact clinical significance of these findings is unclear, but merits further investigation. Similar analyses performed with other agents that are used to treat P. aeruginosa in MMR-deficient hypermutator strains may reveal additional important insights. Given these findings, identification of hypermutator phenotypes in the clinical microbiology lab, either through direct phenotypic testing or through targeted gene amplification and sequencing, may in the future inform optimized treatment decisions based on these differences in behavior, but additional clarifying work on the role of hypermutation in the evolution of clinical resistance is needed.

There are a few important caveats that flow from the design of this study. The first is that the experiments were performed under in vitro conditions, where the predominant selection force was CZA concentration. In the natural context of in vivo infection, a number of other selection forces, including host defense mechanisms factor into the evolutionary equation, and it is possible that bacterial strains may face more stringent global purifying selection under these conditions that might limit the kind of mutational fixation we observed here. Secondly, the hypermutator experiments were performed in a single lab-strain genetic background, and it is possible that the resistance mechanisms uncovered will turn out to be dependent upon unique features of the lab strain genetic background and will not translate into general mechanisms conferring resistance in genetically dissimilar P. aeruginosa strains. Thirdly, although both elevated mutation rate and the ease with which non-beneficial mutations may fix (such as MexB W753R) may contribute to the accelerated evolution of resistance through alternative mechanisms in hypermutators, we are unable to distinguish the relative contributions of these 2 factors. As noted in the foregoing discussion, we believe that MutS-deficient hypermutation likely contributed the initial inactivating mexB mutation and then accelerated the search for alternative resistance mechanisms that was required by this mutation. These 2 events are thus inextricably linked in the actual process that occurred in our adaptive evolution experiments, and we anticipate that similar behavior may be observed more generally in hypermutators. Separately, we observed a relatively high frequency of mutations that would be expected to inactivate MMR, BER (≥2.3%), and MexAB-OprM (≥3.9%) in publicly available P. aeruginosa genome data. These findings suggest that hypermutation may co-occur with a preexisting MexAB-OprM mutation leading to similar activation of alternative AMR resistance pathways as we observed in our work. In this context, we think it is important to note that in many studies describing MexAB-OprM mutations associated with antibiotic resistance, there is no functional corroboration of the mutant allele phenotypes through the construction of isogenic mutants or other methods. It is conceivable that some previously described MexAB-OprM variants in antibiotic-adapted strains may actually be inactivating, and the isolates may have developed resistance through less appreciated pathways due to the unavailability of MexAB-OprM.

While the MexVW mutations characterized above account for only a small proportion of the high level of resistance that evolved in the MMR-deficient isolates, they demonstrate conclusively that hypermutators can acquire antibiotic resistance through mechanisms not commonly used by WT isolates. Furthermore, the rarity of the observed mutations in public databases implies that the universe of mechanisms that can confer beta lactam resistance in P. aeruginosa is likely much larger than previously appreciated. The apparent ease with which hypermutators may switch to alternative resistance mechanisms for which antibiotics have not been developed may carry important therapeutic implications. Further study of the gene targets altered in these isolates may reveal additional unappreciated mechanisms, including potentially novel antibiotic targets.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001878

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/