Variation in the response to antibiotics and life-history across the major Pseudomonas aeruginosa clone type (mPact) panel

ABSTRACT Pseudomonas aeruginosa is a ubiquitous, opportunistic human pathogen. Since it often expresses multidrug resistance, new treatment options are urgently required. Such new treatments are usually assessed with one of the canonical laboratory strains, PAO1 or PA14. However, these two strains are unlikely representative of the strains infecting patients, because they have adapted to laboratory conditions and do not capture the enormous genomic diversity of the species. Here, we characterized the major P. aeruginosa clone type (mPact) panel. This panel consists of 20 strains, which reflect the species’ genomic diversity, cover all major clone types, and have both patient and environmental origins. We found significant strain variation in distinct responses toward antibiotics and general growth characteristics. Only few of the measured traits are related, suggesting independent trait optimization across strains. High resistance levels were only identified for clinical mPact isolates and could be linked to known antimicrobial resistance (AMR) genes. One strain, H01, produced highly unstable AMR combined with reduced growth under drug-free conditions, indicating an evolutionary cost to resistance. The expression of microcolonies was common among strains, especially for strain H15, which also showed reduced growth, possibly indicating another type of evolutionary trade-off. By linking isolation source, growth, and virulence to life history traits, we further identified specific adaptive strategies for individual mPact strains toward either host processes or degradation pathways. Overall, the mPact panel provides a reasonably sized set of distinct strains, enabling in-depth analysis of new treatment designs or evolutionary dynamics in consideration of the species’ genomic diversity. IMPORTANCE New treatment strategies are urgently needed for high-risk pathogens such as the opportunistic and often multidrug-resistant pathogen Pseudomonas aeruginosa. Here, we characterize the major P. aeruginosa clone type (mPact) panel. It consists of 20 strains with different origins that cover the major clone types of the species as well as its genomic diversity. This mPact panel shows significant variation in (i) resistance against distinct antibiotics, including several last resort antibiotics; (ii) related traits associated with the response to antibiotics; and (iii) general growth characteristics. We further developed a novel approach that integrates information on resistance, growth, virulence, and life-history characteristics, allowing us to demonstrate the presence of distinct adaptive strategies of the strains that focus either on host interaction or resource processing. In conclusion, the mPact panel provides a manageable number of representative strains for this important pathogen for further in-depth analyses of treatment options and evolutionary dynamics.

E volving human pathogens are a major global challenge for human well-being.The recent COVID-19 pandemic has been a dramatic reminder of this challenge.Bacterial pathogens pose a similar threat because of their ongoing adaptation to the human host and particularly the emergence and spread of antimicrobial resistance (AMR).A recent worldwide analysis revealed that approximately 5 million deaths per year are associated with infections caused by AMR-expressing pathogens, and of these, a total of 1.27 million deaths per year can be directly attributed to AMR (1).AMR has, thus, been declared as one of the main areas requiring urgent action by the World Health Organization (2).The Gram-negative bacterium Pseudomonas aeruginosa is an example of a top priority pathogen, for which new treatment options are urgently needed (3) and which is part of the highly problematic group of the ESKAPE pathogens (4,5).This bacterium is a ubiquitous opportunistic human pathogen that is found in diverse environments and can cause various infections in humans, often acquired in a hospital context (6).It is a frequent cause of both wound and lung infections, the latter of which often developing in patients with chronic conditions, such as cystic fibrosis (CF), bronchiectasis, or chronic obstructive pulmonary disease (COPD) (6).This pathogen commonly expresses AMR, often against multiple antibiotics, making it difficult and, in some cases, impossible to treat.
Due to its medical importance, P. aeruginosa has become a model for studying the mechanistic basis as well as the evolution of AMR.A particular focus is on understanding the genetics of AMR.This pathogen is notorious for having a large as well as flexible genome with diverse AMR genes, often found on mobile genetic elements, especially integrative and conjugative/mobilizable elements (ICEs/IMEs) and also plasmids (7)(8)(9).Pseudomonas aeruginosa also readily evolves AMR through spontaneous mutation of chromosomal genes (e.g., drug target genes, efflux pump genes/regulators, or betalactamases) (10)(11)(12)(13)(14), a phenomenon which may be exacerbated by the bacterium's propensity to form hypermutator lineages in situ (15)(16)(17).Genes responsible for AMR against diverse antibiotics have been identified, including widely used antibiotics like those from the group of beta-lactams (18,19) as well as reserve antibiotics, such as colistin (20), or newly developed drugs, such as ceftolozane (21).This bacterium is further able to express phenotypic resistance, which may be based on forming biofilms, producing bacterial persisters, or through the expression of heteroresistance (22).These phenotypic resistance mechanisms often complicate diagnostics and make it difficult to predict AMR from the genome sequence alone (23)(24)(25).Moreover, the majority of the functional analyses use two main laboratory strains, PAO1 and PA14, which were both isolated more than 25 years ago, have both adapted to laboratory conditions, and are unlikely representative of the P. aeruginosa strains currently infecting patients (26)(27)(28)(29)(30). Thus, we often lack information on whether the identified mechanisms underlying AMR in these two laboratory strains apply to the clinically relevant pathogen strains and can assist the improvement of diagnostics as well as treatment designs.
The aims of the current study are to characterize a panel of P. aeruginosa strains, which are representative of the genomic diversity encountered in this species and which are still of a manageable size for in-depth functional as well as evolutionary analyses.It builds on the strains which have been isolated previously by the "Pseudomonas Genomics" group at Hannover Medical School and includes the major P. aeruginosa clone types (mPact) (31)(32)(33).These mPact strains have been isolated from different environments as well as different types of human infections and vary in virulence toward distinct hosts (33,34).In the current study, we describe these strains and their genomic distribution.We further characterize variation in AMR in detail using distinct, commonly applied approaches, including diagnostic automated susceptibility testing (AST), Etests, and disc diffusion assays, in consideration of a range of clinically relevant antibiotics.We also examine the presence of microcolonies and variation in the shape of time-kill curves as indicators of additional defense responses to antibiotics and relate these characteris tics to variation in general growth parameters (i.e., lag time, exponential growth rate, yield).In addition, we reconstructed metabolic models and predicted life history traits that could predict growth and clinical parameters.These characterizations yield new insights into the variation in AMR and related responses in a representative set of strains for P. aeruginosa and thereby provide a reference for future in-depth molecular as well as evolutionary analysis of AMR or virulence in the mPact panel.The mPact panel is available to the scientific community as a resource for future research.

RESULTS
The major P. aeruginosa clone type (mPact) strain panel covers a large part of the genomic diversity of the species We focused our work on the strain panel originally selected to represent the interclonal diversity of the major clones in the present-day P. aeruginosa population (31,33).Using our recent whole-genome sequence analysis, we could show that the mPact panel was distributed across a large part of the species' phylogeny (Fig. 1; Table 1).It inclu ded members of all three main phylogroups.Clinical and environmental mPact strains were available for both main phylogroups A and B. Some strains were more closely related to the two main reference strains PA14 (mPact strain H02) and PAO1 (mPact strain H11), while the remaining strains were placed in distinct phylogenetic groups.Genome size varied among mPact strains, ranging from 6.26 Mb to 7.27 Mb (Table 1).The larger genomes generally harbored more mobile genetic elements.In detail, three strains contained one plasmid, while one other strain had three plasmids.Up to six integrative conjugative elements (ICEs) were found in the mPact genomes.Overall, the wide phylogenetic distribution of the included strains and their origin from both clinical and environmental samples makes the mPact panel a small, representative sample for in-depth analysis.

Variation in antibiotic resistance across mPact strains is related to origin but not phylogroup
Antimicrobial resistance varied substantially across the panel strains (Fig. 2A).Using EUCAST resistance breakpoints, 6 of 22 strains were classified as clinically resistant against at least one substance.Resistance against ciprofloxacin was most frequent (6 cases), while resistance to carbapenems was rare.All strains exhibiting resistance were clinical in origin, while the environmental strains showed only slight variation in overall high drug susceptibility.No pattern of resistance matching phylogroup distribution was apparent.All strains showed low MICs against the last resort drugs ceftazidime/avibac tam, ceftolozane/tazobactam, cefiderocol, and also colistin.Resistance measurements using gradient strip and disc diffusion methods on M9 minimal agar yielded simi lar results but did not confirm resistance to the beta-lactam drugs piperacillin and ceftazidime (Fig. S1).The quantitative results of disc diffusion and gradient strip MICs generally covaried well (Fig. 1B), with some restrictions in this strain set for colistin and ceftolozane/tazobactam.
Microcolony formation inside zones of inhibition of the agar-based methods was assessed to find indications of monoclonal heteroresistance, a phenomenon in which a genetically clonal bacterial population exhibits a range of MICs against an antibiotic in the form of individual colonies growing inside the inhibition zones.Microcolonies were observed in numerous strain/drug combinations (Fig. 2A, rightmost panel), but mainly resulted in a less than twofold MIC increase, with the notable exception of H15 and ceftolozane/tazobactam, where colonies resulted in a 32-fold MIC increase.Some drugs caused little to no microcolony formation (piperacillin/tazobactam, tobramycin), while it was frequent in others, especially ceftazidime.Similarly, some strains exhibited almost no microcolony formation (H13, H20), while others reacted to many drugs in this manner (H15).Overall, possible tendencies toward heteroresistance could be observed.Strong MIC increases due to microcolonies could be observed in two cases.
Resistance phenotypes were tested for their stability against spontaneous loss by repeatedly culturing individual strains in drug-free medium.While the majority of resistances remained stable and fluctuated only mildly around the ancestral MICs, strain H01 showed a dramatic loss of all tested resistances (Fig. 3).This loss resulted in an MIC change to one-quarter of the ancestral value for piperacillin/tazobactam, meropenem and ciprofloxacin, and an even more pronounced loss of aminoglycoside resistance.This strong effect hints at a strain-specific configuration of resistance mechanisms with unstable expression or high fitness cost in the absence of selective pressure.The increased MIC of strain H12 against meropenem also fell below the resistance breakpoint (Fig. 3).

Substantial variation in time-kill dynamics among mPact strains toward high concentrations of distinct antibiotics
The kinetics of bacteria-antibiotic interactions over time can give indications of a bacterial strategy for the survival of lethal drug doses, which is distinct from phenotypic resistance.A slow decrease in cell density after the addition of a drug despite above-MIC concentrations can hint at the presence of a tolerant phenotype, in which survival is based on a decrease in cellular activity in order to avoid or repair the cellular effects of the drug, rather than the expression of a resistance phenotype.We created time-kill curves of the panel strains against four distinct antibiotics at drug concentrations at least 32-fold the strain MICs and quantified the populations' responses by calculating the area under the time-kill curve (AUC) (Fig. 4A).Strong differences between the antibiotics were observed, with ciprofloxacin and gentamicin resulting in rapid killing and, thus, small AUCs, while the beta-lactam antibiotics ceftazidime and meropenem killed populations more slowly and sometimes not completely (Figure 4).
None of the examined strains exhibited a clear increase in AUC against all drugs, but some cases of variable response were observed.For example, strain H06 showed above average AUC against gentamicin, but below average AUC against the beta-lactams, indicating drug-specific responses.No correlations between strain MIC and AUC were observed, with the exception of ceftazidime (Fig. S2), in which three strains showed very rapid killing.These strains had low, but not the lowest ceftazidime MICs.Because this response to antibiotics could also be linked to overall bacterial growth characteristics, we next examined growth in the absence of drugs.

mPact strains vary in general growth characteristics
Several panel strains grew rapidly in M9 minimal medium (Fig. 5; Table S1), with short lag times and high growth rates (H05, 07, 10,11,12,13).In contrast, some strains were markedly slow (H01, H15), with opposing parameters.This indicates a link between these main growth characteristics, which in combination results in matching differences in area under the growth curve.However, this trend was not universal.H14, for example, showed a relatively low growth rate, but a very short lag phase, resulting in an overall large AUC.Maximum optical density (OD) also varied significantly, which could be taken as an indicator of variation in carrying capacity.
We tested the observed variation of growth parameters against a number of strain characteristics and phenotypes.For a cumulative measurement of strain resistance to use in these analyses, we tested multiple aggregate scores.On the one hand, the sum of disc diffusion diameters resulted in a symmetrical distribution with a mean value of 301 mm (Fig. S3, purple curve).Sorting the strains by this score confirmed that the most resistant strains were of clinical origin even though some environmental strains also showed a certain level of resistance (Fig. S4).In addition, calculating the sum of gradient test MICs yielded a strongly skewed distribution with two major outliers (Fig. S3, cyan curve), which persisted after transforming the data as relatives to the drug-specific breakpoints and log2-transforming the MICs to compensate for the doubling dilution series of MIC measurements (Fig. S3, orange curve).Neither of these cumulative resistance scores covaried significantly with the growth parameters (Fig. S5 and Fig. S6).No significance test was performed for the MIC-based resistance score due to H01 being a strong outlier.This was due to the aminoglycoside MICs in strain H01 being extremely high, while disc diffusion diameters were closer to those of the other strains, indicating a poor correlation of these methods for very high aminoglycoside concentrations.

Resistance phenotypes can be related to resistance genes in the sequenced genomes
The genome sequences of the mPact strains have been recently published (8), now allowing us to assess whether specific genome characteristics (e.g., inferred phylogroup, origin, presence, and number of mobile genetic elements) or the presence of known AMR genes are associated with the phenotypes described above.We found no significant correlations between phylogroup, origin, number of plasmids and number of ICEs, on the one hand, and cumulative resistance and growth characteristics, on the other hand, with one exception.A higher number of ICEs correlated with a longer lag phase (Fig. S7), possibly indicating a growth cost of these mobile elements.All tested correlations are summarized in Table S3.
quinolone-resistance determining regions (QRDR) gyrA and parC were found in strains with ciprofloxacin resistance.AMRfinder also found the sulfonamide-resistant variant of dihydropteroate synthase (sul1) and the antiseptic resistance conferring variant of qacE (qacEdelta1).Neither of these phenotypes were tested in our study, but were found in highly resistant clinical isolates, indicating further adaptations to a hospital environ ment.No clear beta-lactam resistance genes were identified.We, therefore, performed a manual search of the known beta-lactam resistance genes blaPDC, ampD, ampDh3, ampR, dacB, ftsI, and oprD.Genetic distance in their amino acid sequences, in some cases, partially matched the pangenome phylogroups (ampDh3, oprD) but was not associated with beta-lactam resistance except in one case (Fig. S8).We identified a frameshift mutation in dacB in strain H06.Loss of function of this gene, which likely results from this mutation, is a known beta-lactam resistance mechanism in P. aeruginosa, matching this strain's elevated MICs against piperacillin and ceftazidime.No confirmed resistance mechanism was found in H03, which showed beta-lactam resistance in the automated susceptibility testing only (Fig. 1).It should be noted that the only strain with distinct reduction in carbapenem susceptibility was not included in this analysis due to an incomplete genome.Overall, the identified resistance genes explain the majority of observed resistance.2A).Resistance gene hits from AMRfinder are presented in the grid on the right.A frameshift in dacB was additionally identified in one strain through amino acid sequence alignment.Antibiotics abbreviations see legend of Fig. 2. Details of identified genes see text.Please note that a sixth strain with resistance (strain H12) is not included here because of incomplete genome sequence information, thus precluding inference of AMR genes in its genome.The color of strain names indicates their clinical (purple) or environmental (green) origins.

Genome sequence analysis predicts life history characteristics, growth, and virulence phenotypes
We further explored the genomic capacity of the mPact panel by reconstructing metabolic models using a recently introduced pipeline called Bactabolize that espe cially accounts for differences between strains (supplemental information).While the predictions by flux balance analysis did not identify variation in the strain's growth rates, we found differences in the number and types of metabolic reactions contained in the reconstructed metabolic models for the mPact strains (total reactions: 2,000, shared reactions: 1,960; Fig. S9).Therefore, we performed an enrichment analysis to assess whether the mPact strains varied in the presence/absence of specific metabolic pathways.We found an enrichment of reactions involved in carbohydrate, amino acid, and secondary metabolite metabolism for environmentally obtained strains (Fig. 7E).Next, we inferred microbial life history traits from genomes using GO terms, pathways, subsystems, secondary metabolites, carbohydrate-active enzymes, and virulence genes (see Materials and Methods; supplemental information).We identified traits that could predict microbial phenotypes in a multivariate analysis.In detail, the source of isola tion (environment or clinic) could be predicted with an accuracy of 0.95 (determined by leave-one-out cross-validation): the capacity for posttranslational modification and osmotic stress response was indicative for clinical isolates, whereas lipid metabolism, carbohydrate degradation, and CRISPR genes were characteristic of environmental isolates (Fig. 7A).Interestingly, we found contrasting directions in amine and polyamine metabolism with degradation in environmental and biosynthesis in clinical isolates (Fig. 7A).Similarly, the virulence of strains in a mouse model (high, moderate, low) was estimated with an accuracy of 0.74 by the number of effector delivery systems (high virulence) and carbohydrate degradation pathways (low virulence; Fig. 7B).Moreover, high growth rates were associated with host interaction processes and heat response, whereas slow growth was linked to cell cycle processes and the number of pseudogenes (Fig. 7C).Compared to the experimental data, the predicted traits captured the growth variance accurately (Spearman correlation R = 0.86, P-value < 0.01; Fig. 7D).In sum mary, life history analysis suggested partitioning of mPact strains into environmentallyderived, slower-growing carbohydrate degraders and clinically derived, host-associated, stress-tolerating, fast-growing pathogens.

DISCUSSION
Our study identified substantial variation in AMR and related traits in the mPact strain panel, representative of the genomic diversity encountered in the opportunistic, high priority human pathogen P. aeruginosa.In detail, we identified consistent variation in resistance against a range of clinically relevant antibiotics.Importantly, we also found variation in traits indicating additional responses to antibiotics, including the presence of microcolonies and time-kill dynamics under high drug concentrations.The variation in resistance could be linked to the presence of known AMR genes in the included strains, which are usually found on mobile genetic elements, especially ICE/IMEs or plasmids.Surprisingly, we could not infer strong correlations between the distinct resistance phenotypes and also not between these and the described general growth characteris tics.The only exception is the observed relationship between ceftazidime resistance and ceftazidime time-kill dynamics.Moreover, we developed a new approach that integrates data on resistance, growth, virulence, and life-history characteristics and that demon strates the presence of distinct adaptive strategies among the mPact strains.Overall, we here provide a reference data set for a strain panel that is representative of the genomic diversity of the species P. aeruginosa, covering different environmental as well as clinical origins.The mPact strain panel is of a manageable size and clearly smaller than alterna tive P. aeruginosa strain panels (36)(37)(38), thus facilitating in-depth molecular as well as evolutionary analyses in the future that go beyond the two main laboratory strains PAO1 and PA14.
The results of our different antimicrobial resistance measurements correlated strongly, indicating that phenotypic resistance is robustly expressed in liquid and agar culture as well as minimal medium.Minor limitations to this agreement were observed for some beta-lactam antibiotics and colistin.Automated susceptibility testing evaluated four strains to be resistant to piperacillin or ceftazidime, which was not confirmed by the other measurements.However, reduced susceptibility was observed in all cases, which suggests that the issue lies with the resistance breakpoints rather than an inconsistency in resistance expression.The lack of strongly resistant strains against the drugs in the panel exacerbates this problem.Reliable colistin resistance measurements, on the other hand, are notoriously difficult (39).We identified genetic resistance determinants for the observed fluoroquinolone and aminoglycoside resistance.Specifically, point mutations in DNA gyrase and topoisomerase are frequently encountered in quinolone-resistant strains of many species, including P. aeruginosa (40).Different aminoglycoside-modifying enzymes (AMEs) were found in the two highly aminoglycoside-resistant strains H03 and H06.In contrast, no established beta-lactam resistance genes were found although our manual analysis found a frameshift mutation in dacB (penicillin-binding protein 4), which is involved in cell wall recycling and strongly induces expression of the chromosomal beta-lactamase AmpC (blaPDC) when inactivated by mutation (41).In general, betalactam resistance in P. aeruginosa can result from a multitude of mechanisms including porin loss, efflux pumps, beta-lactamases, and target modification, making reliable prediction difficult (42).
The resistance phenotypes were found to be stably expressed in our drug-free evolution experiments, with one notable exception.Strain H01 exhibited strong loss of beta-lactam and especially aminoglycoside resistance.While we found no distinct beta-lactam resistance genes, a unique AME was identified in ant(2″)-Ia (synonym aadB).This aminoglycoside adenylyltransferase was previously reported on mobile genetic elements in P. aeruginosa and other Gram-negative pathogens (43).This loss of resistance suggests high costs to the cell, which was, indeed, confirmed for H01 by the observed growth characteristics in the absence of antibiotics (long lag phase and low values for growth rate, final OD, and AUC; Fig. 5), and which should then result in rapid loss of the phenotype once the selective pressure is lifted during evolution.In many cases, this scenario is associated with location of the resistance genes on a costly plasmid, which may get lost after removal of the selective constraint.This scenario cannot explain our finding since H01 does not seem to bear a plasmid and, thus, the exact underlying reasons still remain to be determined in future work.Irrespective of the exact cause, such rapid resistance losses may have implications for clinical diagnostics and experimentation where the studied strains are repeatedly precultured without drugs, favoring lineages with low copy number or even loss of the entire mechanism (44)(45)(46).While we also observed five instances of changed breakpoints moving strains into higher resistance categories, these were confined to combinations in which the ancestral MICs were already very close to the resistance breakpoints.These cases may, therefore, be artifacts caused by minor fluctuations in generally stable resistance.Overall, the observed resistances did not coincide with phylogenetic relatedness.Instead, the most resistant strains had a clinical background, suggesting that individual strain histories influence their resistances (although a few environmental strains also produced moderate levels of resistance).This general pattern is supported by our reconstruction of genetic determinants for surface disinfectant resistance in the resistant clinical strains.
We additionally examined the strains for the presence of microcolonies in the zones of inhibition and their effect on MICs.This phenomenon indicates that a bacterial population, despite being isogenicly founded, has members with increased resistance.When resulting in extensive increases in the overall resistance of the population, this is called heteroresistance (47) and can have strong implications for diagnostics.For one, microcolonies may not always be visible for inspection after the regular 18-24 h incubation, leading to underestimation of MICs and potentially misclassification as susceptible.Furthermore, heteroresistance is a possible stepping stone toward full resistance (47,48), meaning the population may be in the process of losing its suscept ibility.The exact clinical implications of this phenomenon are not fully understood.Unexpectedly, we found comprehensive variation among strains and drugs, suggesting a variety of underlying mechanisms rather than a drug-specific effect or a strain-spe cific mechanism, such as increased mutation rates (49).Microcolonies were commonly found in the presence of three antibiotics (i.e., ceftazidime, ciprofloxacin, amikacin; Fig. 2A), all from different classes.Moreover, a few strains expressed microcolonies toward several antibiotics, including strains PA14, H02, H03, and especially H15.The latter strain H15 produced microcolonies in the presence of six of the tested antibiotics, including beta-lactam drugs, a fluoroquinolone, and aminoglycosides (Fig. 2A).Interestingly, the growth characteristics for H15 showed a comparatively high lag phase and very low values for growth rate, final OD, and AUC, all suggesting a possible fitness deficit for this strain.It is tempting to speculate that such a fitness deficit may be a consequence of expressing microcolonies toward diverse antibiotics.Further investigation of these phenotypes and their consequences for adaptation to antimicrobials is clearly warranted.
Variation in strain responses to antibiotics was also found in time-kill kinetics.After exposing the panel strains to above-MIC drug concentrations, the decrease in cell density occurred at different speeds.Generally, slow decrease in cell number could be an indication that strains handle antimicrobial stress through tolerance rather than phenotypic resistance (50).We observed a pattern of drug-specific differences in time-kill kinetics, with the beta-lactams ceftazidime and meropenem generally reducing cell density more slowly than ciprofloxacin or gentamicin.This is likely, at least in part, the result of their mechanism of action.Beta-lactam antibiotics inhibit peptidoglycan synthesis, affecting both cell wall maintenance and cell division.Therefore, the speed of bactericidal action may depend on the growth speed and conditions.Pharmacodynamic parameters may also play a role in these kinetics, because beta-lactams primarily bind their targets covalently, meaning that their effective concentration decreases over time.This effect is also the basis for the so-called inoculum effect (51).We further observed a correlation of ceftazidime MIC and area under the time-kill curve, suggesting that more resistant strains are killed more slowly, despite the fact that all strains were treated with at least 32× their MIC.In contrast, no correlation was found between growth characteris tics and time-kill dynamics, indicating that tolerance by generally slow growth does not play a role in these observations.We generally found only few significant correlations among the distinct phenotypes.This is surprising because high resistance often comes at a growth cost under drug-free conditions (44), as at least observed for the strain H01, and because antibiotic tolerance may be achieved by reduced growth rates (50).The general lack of an association could be explained by emergence and spread of compensatory mechanisms that counter the resistance-induced fitness defects.Alternatively, it may be explained by different evolutionary trajectories to resistance or life history trait optimization that can lead to different types of trade-offs in different genomic backgrounds.
We further explored the genomic potential of the mPact panel by employing metabolic and ecological modeling.We used the available whole-genome sequences to infer life history characteristics and predict phenotypes of the strains.On the one hand, we found stress-tolerance traits consistently associated with clinical isolation source, higher virulence, and faster growth rate.Interestingly, effector delivery systems such as secretion systems, potentially involved in antibiotic resistance, were found to be predictive for virulent strains.On the other hand, carbohydrate degradation was prominent for environmentally isolated strains and lower virulence.In recent years, life history trait analysis has been applied to analyze environmental and antibiotic-resistant microorganisms (52,53).In particular, stress-response such as resistance to antibiotics are known to impose a burden on microbial physiology with consequences for growth and resource acquisition (54,55).Although we did not see an impact of stress-response on growth rate, we found less capacity for carbohydrate acquisition for virulent and clinical strains.Higher growth rates were associated with the enrichment of genes involved in heat response and host interactions, which might link to host adaptation and a strategy specific for pathogens (56,57).In addition, investment in versatile carbohydrate degradation pathways can negatively impact growth (58) and shapes the life history strategy for many environmental microbes (59).Therefore, we hypothesize that mPact strains follow distinct adaptive strategies characterized by either increased degradation or stress and host response.Moreover, we found an opposing pattern in polyamine metabolism.Environmental strains showed an increased capacity to degrade polyamine, and clinical strains showed a higher capacity toward biosynthesis.Polyamines, such as putrescine and spermidine, are ubiquitously present in all organisms, play a prominent role in many processes like gene regulation, cell proliferation, stress response, and can affect host lifespan (60)(61)(62).Several host-associated microbes can produce polyamines (63) and especially pathogens are known to interfere with host polyamine metabolism (64,65).Therefore, the opposing pattern of polyamine degradation and biosynthesis further supports our proposed classification into degradation and host-related strategies for the mPact strains.To our knowledge, this is the first application of life history theory to microorganisms on the strain level.
Altogether, we here present a panel of 20 strains of P. aeruginosa plus two laboratory strains, representing the species' genomic diversity and a variety of resistance pheno types and growth characteristics.The mPact panel is of suitable size for the in-depth analysis of any phenotypes of interest, without constraining the results to the usual lab strains and, therefore, greatly expanding the predictability of general application.Most strains are highly susceptible to a range of clinically relevant antibiotics, making them suitable for the examination of the trajectories of resistance evolution of variable genetic backgrounds using experimental evolution.All but one genome have been fully sequenced and are publicly available.Life histories analysis suggests differences in strain adaptation dependent on isolation source or virulence.The mPact panel is, therefore, a useful resource for fundamental and translational research of this important human pathogen.

MATERIALS AND METHODS
Experiments were conducted with a strain panel of Pseudomonas aeruginosa which represents genomic diversity observed in the species.Details on strain origin, microassay classification, and whole-genome sequencing have previously been published (8,31,33).The reference strains PA14 (27,29) and PAO1 (26) were also included.Bacteria were grown in M9 minimal medium supplemented with glucose (2 g/L), citrate (0.58 g/L), and casamino acids (1 g/L) or on M9 minimal agar (1.5%), unless otherwise indicated.Antibiotics were added as specified in the respective sections.Cultures and plates were incubated at 37°C.
Statistical analysis was performed in R, Version 4.3.1.All R scripts are available via the Github account of the Schulenburg lab under the following link: https://github.com/evoecogen/mPact-panel.

Phylogenomics and identification of mobile genetic elements
Phylogenetic trees were constructed as described in Botelho et al. (8).In brief, a total of 5,468 P. aeruginosa genomes were downloaded from RefSeq's NCBI database using PanACoTA v1.2.0 (66).After removal of low-quality assemblies, duplicates, and misclassi fied assemblies, 1,991 genomes were retained.The genomes of 19 mPact strains were added to this, resulting in a total of 2,010 genomes.A maximum likelihood tree was inferred with the general time reversible model of nucleotide substitution in IQ-TREE v2.1.2(67) and visualized with iTOL v6 (68).To search for ICEs/IMEs in the mPact strains genomes, we used the annotated files generated by prokka (69) as input in the standalone-version of ICEfinder (70).

Antimicrobial susceptibility
The minimum inhibitory concentration of antibiotics was measured using (i) a gradient test strip (71) and (ii) the disc diffusion test (72).Bacterial cultures were incubated at 37°C for 18 h, with subsequent inoculation of M9 plates with a cotton swab of overnight culture (OD 600 at 0.08).(i) MIC test strips (Liofilchem) were placed onto the plate, and the plate was incubated for 24 h, after which the MIC was read at the intersection of the zone of inhibition and the test strip.(ii) Three antimicrobial discs (Mast Diagnostica) per plate were applied with a disc dispenser (MDD65, Mast Diagnostica) and incubated for 24 h, after which the inhibition zones were analyzed by standardized image capture using a flatbed scanner (Epson Perfection V600 pro).Image analysis was performed with the program Antibiogram J (73) (always in triplicate).We also performed automated susceptibility testing using the VITEK2 system (bioMérieux) and analyzed using the EUCAST clinical breakpoint table Version 13.1 (74).Breakpoints for gentamicin were taken from Version 9, the last version to contain values for this drug in P. aeruginosa (75).

Presence of microcolonies
MIC test strip plates were visually inspected for the presence of microcolonies inside the zones of inhibition.If present, MICs with and without these colonies were recorded.Plates were incubated for another 24 h to improve colony detection.No differences were observed between the two time points (data not shown).

Assessment of resistance stability
Resistance stability was assessed after evolution in the absence of drug.Populations of three independent parallel cultures per strain were seeded in antibiotic free medium and grown for 18 h.They were then diluted to 0.08 and 100 µL was plated onto plates without antibiotics.Every 24 h, 20% of the bacteria were transferred to a fresh plate without antibiotic.The experiment lasted a total of 15 days.Evolved populations were frozen in dimethyl sulfoxide on the last day.Due to logistic reasons, only 18 out 22 strains of the panel were used and had to be split in two separate evolution experiments.As controls, two strains were included in both.Results for them were found to be consistent across the experiments (Kolmogorov-Smirnov test).Resistance stability was calculated as log2-transformed means of replicates minus log2-transformed ancestral MICs.

Time-kill curves
Measures for antibiotic tolerance were based on time-kill curves (50).In brief, each strain was cultured in M9 medium overnight from single colonies, and the resulting culture was adjusted to OD 600 0.08-0.12.After that, antibiotics (at least 10 times higher than MIC, meropenem: 15 µg/mL, gentamicin: 30 µg/mL, ciprofloxacin: 15 µg/mL, ceftazidime: 200 µg/mL) were added to the diluted bacterial culture and the tubes incubated at 37°C with shaking.At different time points, the number of cells was monitored.Serial dilution and plating were used to determine the number of colony forming units.The survival frequency was calculated by the number of colonies at different time points divided by that of starting point.To exclude phenotypic resistance effects on the time-kill curve analysis, strains where the effective concentrations were less than 32× the MIC, were excluded from analysis.For all other combinations, the area under the curve was calculated using the AUC function of the DescTools package in R with the trapezoid approach.

Growth curves
Growth curves were assessed in 96-well plates.In short, we inoculated one colony overnight for 20 h in LB-Medium at 37°C with shaking.These overnight cultures resulted in equivalent cell numbers, which did not vary significantly across the mPact strains (Table S2).0.5% of the overnight liquid culture were subsequently inoculated in the 96-well plates in a fully randomized way containing M9 medium.The growth kinetics were assessed with continuous shaking (shake mode: double orbital; orbital frequency: 807 cpm [1 mm]; orbital speed: fast), and optical density measurements at 600 nm in a plate reader (Epoch 2, Agilent) for 20 h, with 15 min time points.For each strain, a minimum of 8 biological replicates were tested.Growth parameters were modeled using the method by Hall et al. (35), implemented in the growthrates package in R (easylinear method), yielding maximum growth rates and length of lag phase.Replicates with model fit scores <0.9 were excluded.Maximum OD was recorded and taken as a possible indicator of carrying capacity.AUC was calculated as mentioned above, using a spline approach.

Antimicrobial resistance genes
Genomic resistance genes were retrieved by parsing the assembled and closed genomes into AMRFinderPlus (76).Hits encountered in all strains were filtered manually.For additional resistance genes, the amino acid sequences of the genes in question were extracted from the genomes' RefSeq annotations by downloading the gbff files from Genbank, splitting the contigs into individual files, and loading them into a common data set using the genbankr package in R with bioconductor 3.17 (77).Coding sequences were extracted with the GenomicFeatures package and aligned using the msa command with default settings.Neighbor-joining trees were generated with the nj command of the "ape" package and plotted using tidytree.

FIG 1
FIG 1 Distribution of the mPact strains across the genome-sequence-inferred phylogeny of P. aeruginosa.Maximum likelihood of the softcore-genome alignment of 2009 Pseudomonas aeruginosa isolates.The scale bar indicates genetic distance.Outside colored ring represents phylogroup placement within the species pangenome.The mPact panel strains are indicated and placed according to their phylogenetic position; their origin is denoted by color.Figure modified from Botelho et al. (8) and used with permission of Elsevier/eBioMedicine.

FIG 2
FIG 2 Variation in resistance across the mPact panel.(A) Resistance measurements in (from left to right) automated susceptibility testing (AST, VITEK2), reserve antibiotics on Mueller-Hinton agar (gradient strip, Etest), and Etest on M9 minimal medium agar, displayed as their fold ratio of minimum inhibitory concentration (MIC) to EUCAST resistance breakpoint; disc diffusion on M9 minimal medium agar, displayed as zone of inhibition diameter minus the EUCAST resistance breakpoints in mm; and effect of microcolonies on M9 Etest resistance data.Antibiotic abbreviations: PIP = piperacillin, PIT = piperacillin + tazobactam, CTZ = ceftazidime, CEP = cefepime, IMI = imipenem, MER = meropenem, AZT = aztreonam, CIP = ciprofloxacin, AMI = amikacin, GEN = gentamicin, TOB = tobramycin, COL = colistin, CTV = ceftazidime + avibactam, CTT = ceftolozane + tazobactam, CID = cefiderocol.Box labels in the heatmaps indicate clinical resistance classification according to EUCAST, where appropriate: S = susceptible, I = susceptible at increased exposure, R = resistant, iR = inferred resistance from related substances despite a minimal inhibitory concentration below the breakpoint.IE: insufficient evidence for clinical use.(B) Regression of M9 gradient strip MIC and M9 disc diffusion diameter.R 2 is the coefficient of determination of the linear regression model.Symbols in the upper right corners indicate the P-values of the t-test of R 2 = 0, with *P < 0.05, **P < 0.01, ***P < 0.001.

FIG 4
FIG 4 Variation in time-kill response among mPact strains toward distinct antibiotics.(A) Illustration of test principle.Strains were exposed to drug concentra tions at least 32× above the respective MICs and cell numbers counted after 2.5 and 5 h.The area under the time-kill curve (AUC) is used to compare strains.(B) AUCs per strain, split by test antibiotic.Missing bars indicate combinations that were not tested due to high strain MICs.The dotted lines indicate the respective average values for the laboratory strain PA14 and dashed lines those for the laboratory strain PAO1.Error bars represent standard deviation.The color of strain names indicates their origin.

FIG 5
FIG 5 Variation in growth characteristics across mPact strains.Bacterial growth was characterized in the absence of antibiotics in M9 medium.The characteristics of bacterial growth were assessed as (i) the growth rate (top left panel) and (ii) lag time (top right panel), inferred from a log-linear growth model (35) using the growthrates package in R, (iii) maximum optical density (OD), providing a possible indicator of carrying capacity (bottom left panel), and (iv) area under the curve (AUC) of OD measurements across the entire growth period.The dashed and dotted lines indicate the average values for the laboratory strains PAO1 and PA14, respectively.The color of strain names highlights their origins, as indicated.Error bars represent standard deviation of at least eight replicates.Further details are provided in TableS1.

FIG 6
FIG6 Genomic resistance determinants are associated with resistance phenotypes.Strains are arranged according to their core genome phylogeny and shaded by phylogroup (left part).The heatmap indicates strain resistance to the given drugs plotted as the distance of disc diffusion diameter to the drug resistance breakpoint, resulting in resistant values given as negative values (heatmap as in Fig.2A).Resistance gene hits from AMRfinder are presented in the grid on the

FIG 7
FIG7  Genome sequence analysis and metabolic models predict life history characteristics of mPact strains.(A) Prediction of strain isolation source (environment or clinic) by life history traits using a lasso classifier.Accuracy was determined by leave-one-out-cross-validation (LOOV).(B) Inference of virulence (high, moderate, low) of mPact strains by life history traits using random forest classifier.Accuracy was determined by LOOV.(C) Prediction of strain maximal growth rates by life history traits using random forest regression.Root-mean-square error (RMSE) and R 2 were determined by LOOV.(D) Correlation of experimental maximal growth rates and the growth rates predicted by the random forest regression model from panel C. (E) Enrichment of metabolic pathways found in metabolic models that were isolated from the environment compared to clinical isolates.

TABLE 1
(33)ral information on the origin and phylogenomic characteristics of the mPact panel a Missing due to incomplete sequencing; for strain H12, the BioSample ID of the sequencing done in Hilker et al. is given instead(33).
bNo exo gene identified.c