Adaptive Responses of Pseudomonas aeruginosa to Treatment with Antibiotics

ABSTRACT

Pseudomonas aeruginosa is among the highest priority pathogens for drug development because of its resistance to antibiotics, extraordinary adaptability, and persistence. Antipseudomonal research is strongly encouraged to address the acute scarcity of innovative antimicrobial lead structures. In an effort to understand the physiological response of P. aeruginosa to clinically relevant antibiotics, we investigated the proteome after exposure to ciprofloxacin, levofloxacin, rifampicin, gentamicin, tobramycin, azithromycin, tigecycline, polymyxin B, colistin, ceftazidime, meropenem, and piperacillin-tazobactam. We further investigated the response to CHIR-090, which represents a promising class of lipopolysaccharide biosynthesis inhibitors currently under evaluation. Radioactive pulse-labeling of newly synthesized proteins followed by two-dimensional polyacrylamide gel electrophoresis was used to monitor the acute response of P. aeruginosa to antibiotic treatment. The proteomic profiles provide insights into the cellular defense strategies for each antibiotic. A mathematical comparison of these response profiles based on upregulated marker proteins revealed similarities of responses to antibiotics acting on the same target area. This study provides insights into the effects of commonly used antibiotics on P. aeruginosa and lays the foundation for the comparative analysis of the impact of novel compounds with precedented and unprecedented modes of action.

INTRODUCTION

Pseudomonas aeruginosa is a common Gram-negative pathogen that frequently establishes severe infections in critically ill patients. Particularly in neutropenic patients, P. aeruginosa often causes bacteremia associated with high morbidity and mortality rates (1, 2). Owing to the natural resistance of Pseudomonas strains to several antibiotic classes and their outstanding adaptability, medicine is facing an alarming scarcity of effective antipseudomonal drugs today (3, 4). On that basis, P. aeruginosa was recently listed as one of the pathogens for which new drugs most urgently need to be developed (5). Despite extensive efforts in the past, no new antibiotic classes with efficacy against Gram-negative pathogens were approved in more than 50 years. Derivatives of the current pool of structural classes and antibiotic adjuvants dominate late stages of clinical development (6). Drugs that were discontinued because of severe and frequent side effects like polymyxins and colistin have been reintroduced into medical practice since treatment options became severely limited (7). With panresistant P. aeruginosa strains emerging (8), drug discovery efforts are directed at replenishing the pipeline with novel lead structures (9, 10).

Gram-negative bacteria excel in preventing antibiotics such as vancomycin or daptomycin from accessing their targets. Their envelope features two membranes, a cytoplasmic membrane and an outer membrane with a lipopolysaccharide (LPS) layer in the outer leaflet, which have different and in part opposing permeability properties (11). Targeting this specific defense by combining frontline antibiotics with polymyxins, which enhance permeability of the outer membrane (12), has proven a viable treatment option that may achieve bacterial clearance. Several promising envelope-related targets are under investigation. The inhibition of LPS synthesis at committed steps results in the accumulation of toxic lipid A precursors that perturb the cell envelope (13). Among numerous LPS synthesis inhibitors studied, CHIR-090 was the first lead that demonstrated potent antipseudomonal activity alone and in combination in a mouse implant biofilm infection model (14–16). Another target that was pursued in Gram-positive bacteria, the type I signal peptidase (SPase), can now be targeted in Gram-negative bacteria with arylomycin G0775 (17). SPase is essential for the secretion of membrane proteins, including proteins involved in cell wall synthesis and envelope maintenance (18).

Developing novel antibiotic classes and getting them approved is a lengthy process involving a high financial risk, since the requirements for efficacy and safety are high. A repositioning of nonantibiotic drugs could significantly accelerate development, since detailed knowledge on safety and pharmacology is already available (19). Existing drugs, as well as lead structures under investigation for other disease areas, have been recognized as sources for novel antimicrobial lead structures (20, 21). For instance, ebselen, a synthetic organoselenium compound tested for treatment of bipolar disorder, has been reported to target the thioredoxin system in Gram-negative bacteria (22, 23), and the antipsychotic phenothiazine trifluoperazine exhibits bactericidal activity and increases host survivability (24).

To gain insights into the bacterial response to antibiotics and to support antibiotic research by providing a system-based resource for mechanism of action assessment, we recently published a comparative proteomic study on the Gram-positive model organism Bacillus subtilis (25). We monitored the initial changes in the cytosolic proteome in response to antibiotic treatment with the exquisite and unparalleled sensitivity of radioactive pulse-labeling of newly synthesized proteins using two-dimensional polyacrylamide gel electrophoresis (2D-PAGE). The acute proteomic responses provide insights into impaired pathways, processes, and cellular structures as they are tailored to overcoming the physiological challenge inflicted by the antibiotic (26). A mathematical Comparison of Proteomic Responses (CoPR) allowed researchers to rapidly assess the similarity of more than 90 proteomic response profiles based on the upregulated marker proteins (25). These marker proteins deeply enhance our understanding of how bacteria cope with antibiotic exposure. From these marker proteins, protein signatures were delineated that are indicative of the general target area. These signatures together with the cosine similarity facilitate a rapid sorting of novel antimicrobial agents into those with precedented and unprecedented modes of action.

Here, we lay the foundation for a complementary reference compendium for Gram-negative bacteria using P. aeruginosa to facilitate research into experimental drugs that target Gram-negative-specific molecules and processes and to provide system-based information on the response of P. aeruginosa to antibiotic treatment. We report proteomic responses to 12 agents currently used clinically, including (i) the membrane-disrupting colistin and polymyxin (27, 28), (ii) the cell wall biosynthesis inhibitors ceftazidime (29), meropenem (30), and piperacillin-tazobactam (31), (iii) the topoisomerase inhibitors ciprofloxacin and levofloxacin (32, 33), (iv) the transcription inhibitor rifampicin (34), (v) gentamicin and tobramycin, which interfere with ribosomal proofreading (35, 36), and (vi) the translation inhibitors tigecycline (37) and azithromycin (38), the latter of which also interferes with quorum sensing (39). We further investigated the proteomic response to the experimental LPS biosynthesis inhibitor CHIR-090. We compared proteomic responses based on marker proteins using CoPR similarity scoring and pathway representation to extract signature proteins that represent integral effectors of the pseudomonal stress response for each target area. The presented P. aeruginosa proteomic responses are intended to provide unique insights into the initial antibiotic stress response to clinically relevant drugs and to serve as starting point for an antibiotic response library that will aid in antipseudomonal drug discovery.

RESULTS AND DISCUSSION

Proteomic response library for antibiotic treatment.

To build a proteomic response library, we selected representatives of the most important antibiotic classes currently used to combat pseudomonal infections: polymyxins (colistin/polymyxin B), fluoroquinolones (ciprofloxacin/levofloxacin), cell wall biosynthesis inhibitors (ceftazidime, meropenem, and piperacillin-tazobactam), and aminoglycosides (gentamicin/tobramycin). We complemented the selection with macrolides (azithromycin) and glycylcyclines (tigecycline) and the LPS biosynthesis inhibitor CHIR-090 to extend the mode of action coverage.

To generate the proteomic response profiles, we followed the approach established for B. subtilis (25). To strike a balance between exerting antibiotic stress and allowing the cells to respond to the physiological challenge, physiologically effective concentrations were identified that cause a 20% to 50% inhibition of growth during exponential phase (see Fig. S1 in the supplemental material). Physiologically effective concentrations could be determined for all tested compounds except for the piperacillin-tazobactam treatment, which did not inhibit growth of exponentially growing cultures even at 256 μg/mL piperacillin and 51.2 μg/mL tazobactam (see Table S1). Piperacillin-tazobactam treatment did, however, effectively inhibited P. aeruginosa growth in an MIC assay performed in the same medium (MIC = 0.64/0.128 μg/mL). For proteome analysis, cells were grown in chemically defined medium to early exponential phase and then treated with physiologically effective concentrations of antibiotics. Starting 10 min after antibiotic addition, newly synthesized proteins were labeled with l-[³⁵S]methionine during a 5-min pulse. Cells were harvested and lysed, l-[³⁵S]methionine incorporation rates were determined, and proteins were separated by 2D-PAGE. Autoradiographs of the 2D gels were subjected to image analysis to determine relative synthesis rates of proteins. Proteins consistently upregulated by a factor of 2 or greater in each of three independent replicate experiments were designated marker proteins, provided that they accumulated sufficiently to allow mass spectrometry-based identification.

The impact of acute antibiotic treatment on protein biosynthesis rates reflects changes in translation capacity (40). In B. subtilis a reduction in protein biosynthesis rate is typically associated with ribosome inhibition, disturbance of replication, or transcription. For P. aeruginosa we found that translation inhibitors azithromycin and tigecycline reduced protein biosynthesis rates by 50 to 60%. Other antibiotics only had minor effects on protein synthesis rates, including the gyrase inhibitors levofloxacin and ciprofloxacin, causing 7 to 26% reduction, and colistin reaching 30% (Fig. 1A). Most strikingly, rifampicin, which reduced protein biosynthesis rates very efficiently in B. subtilis (41), had very little effect on protein biosynthesis rates in P. aeruginosa during the pulsing window from min 10 to 15 after treatment.

FIG 1.

(A and B) Effect of antibiotics on protein biosynthesis rates (A) and induction of marker proteins (B). P. aeruginosa cultures were treated in early log phase with physiologically effective concentrations of antibiotics. l-[³⁵S]methionine incorporation during a 5-min pulse starting 10 min after antibiotic addition was set in relation to untreated controls. Cell wall biosynthesis inhibitors and rifampicin did not elicit proteomic responses.

Based on the 2D-PAGE analysis of all antibiotics, we identified a total of 102 marker proteins, 4 of which are functionally uncharacterized. The proteomic data for all compounds is provided in the supplemental material (see Fig. S2 to S14 and Tables S2 to S19 in the supplemental material). The number of marker proteins per treatment condition ranged from 0 to 30 marker proteins (Fig. 1B). As previously observed for the Gram-positive model organism B. subtilis (25), inhibition of late steps of cell wall biosynthesis inhibition by beta-lactams and cephalosporins yielded no marker proteins. The antibiotics ceftazidime, meropenem, and piperacillin-tazobactam showed good efficacy in an MIC assay (see Table S1 in the supplemental material). However, only meropenem also had a low physiologically effective concentration, leading to lysis after ∼100 min of treatment (see Fig. S9). To cause a partial inhibition of growth of exponentially growing cell cultures treated at an optical density at 578 nm (OD₅₇₈) of 0.4, high concentrations of ceftazidime (512 μg/mL) were needed, and piperacillin-tazobactam did not inhibit growth of these cultures at 256 μg/mL piperacillin/51.2 μg/mL tazobactam. Compared to the MIC assay, at the time of antibiotic addition, the cultures have a 3 to 4 orders of magnitude higher cell density. A nitrocefin-based assay showed that beta-lactamase activity was detectable in cell extracts of control cultures, as well as cultures treated for 15 min with 512 μg/mL ceftazidime or with piperacillin-tazobactam (256 μg/mL piperacillin and 51.2 μg/mL tazobactam) (see Fig. S15 in the supplemental material). The enzyme activity was equally high in untreated and ceftazidime-treated cells but only approximately 23% after piperacillin-tazobactam treatment, likely due to tazobactam action. Notably, no difference in beta-lactamase protein levels was detected between untreated and treated cultures in the proteome analysis. Taken together, these data indicate that at the comparably high cell densities, cells are sufficiently protected from ceftazidime and piperacillin action through the time of pulse-labeling by beta-lactamase levels present already under control conditions, i.e., in the absence of antibiotic. The question remains as to why there are no marker proteins for meropenem treatment. We speculated earlier that the cells either do not sense or are unable to respond rapidly to the structural loss of cell wall integrity (26), and for Gram-positive cells it was shown that they die as they continue to initiate cell division (42). Again, strikingly, transcription inhibition by rifampicin did not result in upregulation of marker proteins in P. aeruginosa, whereas 25 marker proteins were described for rifampicin-treated B. subtilis (26). Growth was affected differently as well in both species. While growth of B. subtilis stopped abruptly but temporarily, in P. aeruginosa the growth rate was reduced gradually over time (see Fig. S12). It will be interesting to further investigate the reasons underlying this difference. It may be related to the comparatively slow growth of P. aeruginosa in chemically defined M9 medium (doubling times of approximately 180 versus 50 min for B. subtilis), to the long mRNA half-life in this medium (43), or perhaps to a more active posttranscriptional regulation in B. subtilis.

To assess the similarity of the responses to the different antibiotics, pairwise comparisons based on cosine similarity were performed using the regulation factors of marker proteins as described recently (25). The resulting similarity scores (CoPR scores) are displayed as a similarity matrix (see Fig. S16 in the supplemental material). High CoPR scores are reflective of shared target areas, and any score greater 0 is worth investigating since it indicates that there is some similarity between two responses that may or may not be related to a shared target area. The highest similarity was observed for the membrane-disrupting antibiotics polymyxin B and colistin (CoPR score of 0.69) and agents interfering with ribosomal proofreading gentamicin and tobramycin (CoPR score of 0.69). The responses to both aminoglycosides also gave high CoPR scores of 0.62 and 0.67, respectively, compared to the previously published response to allicin (44), a natural product from garlic that is known to cause protein thiol stress (45). High CoPR scores were obtained for the gyrase and topoisomerase inhibitors ciprofloxacin and levofloxacin (CoPR score of 0.34) and for the translation inhibitors azithromycin and tigecycline (CoPR score of 0.22). The comparably lower CoPR scores for the ciprofloxacin/levofloxacin and azithromycin/tigecycline pairs indicate that there are compound-specific aspects to the responses. Possible reasons include differences in the ratio of gyrase versus topoisomerase IV inhibition, and levofloxacin has been described to have a higher lytic activity than ciprofloxacin (46). Differences between azithromycin and tigecycline are discussed in more detail below. The proteomic response to the LPS biosynthesis inhibitor CHIR-090 was most similar to polymyxin B and colistin responses, but reflective of its unique mechanism the CoPR scores were low (0.13 and 0.11, respectively).

To gain insights into the physiological effects of antibiotic treatment and the strategies of P. aeruginosa to cope with the stress, the proteomic responses to antibiotics were evaluated with regard to cellular pathways represented by the upregulated marker proteins (Fig. 2). The association of marker proteins with biological processes (GO categories) (see Table S20) was supplemented with assignments to cellular functions based on literature and data bank entries (see Table S21), and pathway representation was calculated using the scaled distribution of marker proteins among cellular pathways (Fig. 2A). This crude analysis revealed that the aminoglycosides’ interference with ribosomal proofreading elicited the most focused response, namely, the upregulation of chaperones, indicating that misfolded proteins represent the biggest physiological challenge. In response to inhibition of the elongation phase of translation by azithromycin and tigecycline, translation-related proteins were upregulated, likely as a countermeasure to the decreased translation capacity. The inhibition of replication by fluoroquinolones was countered by the upregulation of proteins involved in nucleotide and precursor biosynthesis and the SOS response. Polymyxin B and colistin both evoked the upregulation of proteins involved in fatty acid metabolism, while proteins involved in protein folding, secretion, and respiration were upregulated only in response to polymyxin B.

FIG 2.

Comparison of proteomic responses. (A) Pathway representation in proteomic responses. Dendrograms were generated based on similarity scores of proteomic responses using the Ward method (ward.D2). Pathway representation was calculated using the scaled distribution of marker proteins among cellular pathways and summarized in a heatmap. Color coding represents the weighted proportion of marker proteins in relation to all other cellular processes. Marker proteins were assigned to their cellular function based on literature and data bank entries. For a full list of protein pathway affiliations, see Table S27 in the supplemental material. Most strongly represented pathways are indicated by darker colors. (B) Marker proteins upregulated in response to both antibiotics representing a target area are listed. The target area specific marker proteins are highlighted with black lines. AIR, 5′-phosphoribosyl 5-aminoimidazole; MG, methylglyoxal; SU, subunit; Fao, fatty acid oxidation.

Extraction of proteomic signatures.

A proteomic signature consists of a protein or a group of proteins that is upregulated specifically in response to a defined physiological condition (47). The diagnostic power of proteomic signatures can be used to identify antibiotic target areas of structurally novel compounds with unknown mechanisms of action. When using chemical compounds to identify proteins indicative of a certain physiological state, in addition to proteomic responses that are related directly or indirectly to target inhibition, there may also be compound-specific effects. To delineate proteomic signatures for target areas, we analyzed the response to two structurally different inhibitors and disregarded marker proteins that were only upregulated in response to one (Fig. 2B). Proteomic signatures could be further refined by analyzing additional compounds or conditional mutants (48). Polymyxins represented by polymyxin B and colistin are basic polypeptides that act as cationic detergents causing alterations of membrane structure and fluxes across the membranes, eventually leading to cell death (27). Two enzymes associated with fatty acid biosynthesis and degradation, FabB and FaoA, respectively, together with the pyruvate decarboxylase subunit B (OadA), which is involved in NADH supply, e.g., for fatty acid biosynthesis, serve as a signature for impairment of the cytoplasmic and outer membranes.

The protein signature for replication inhibition, based on the responses to ciprofloxacin and levofloxacin, encompasses three proteins involved in the synthesis of nucleotides, namely, the catalytic subunit of the phosphoribosylaminoimidazole carboxylase (PurE) and two proteins of the pentose phosphate shunt (PgI and Rpe).

Translation is targeted at various stages by different antibiotic classes. The aminoglycosides, which are most relevant clinically for treatment of P. aeruginosa infections, are represented by gentamicin and tobramycin, which cause conformational changes at the A-site of the ribosomal 30S subunit that lead to the incorporation of mismatched amino acids into the polypeptide chain (35). The upregulated chaperone systems GroESL and DnaK/GrpE, which were already shown to be upregulated in response to tobramycin in a previous study (49), are integral components of the RpoH-controlled heat shock response (50) and serve as proteomic signature for compounds that affect protein homeostasis (51).

Tigecycline and azithromycin inhibit the elongation step of translation—tigecycline by blocking the A-site and thus aminoacyl-tRNA entry (52) and the macrolide by blocking the exit tunnel of the nascent polypeptide chain, leading to ribosome stalling (38). The elongation factor EfP and ribosomal protein L25 (RplY) serve as signature proteins for elongation inhibitors. RplY binds to the E-loop of the 5S rRNA as a part of the central protuberance. It further forges direct contact with helix 38 of the 23S rRNA, which catalyzes the peptidyl transferase reaction, and is thought to stabilize this functionally important structure to ensure ribosome function (53). There are also significant differences in the responses to tigecycline and azithromycin. In response to tigecycline, but not azithromycin, the ribosomal protein S1 (RpsA) and elongation factor Tu (TufA) are upregulated. RpsA acts as RNA chaperone unwinding most structured mRNAs (except leaderless mRNAs) at the ribosome to increase binding efficiency (54). Like TufA, it is required for tmRNA binding to the 30S subunit and thus proposed to play a role in ribosome rescue by trans-translation and fine-tuning of translation (55). In response to azithromycin, but not to tigecycline, peptide chain release factor 2 (PrfB) and ribosomal protein S6 (RpsF) were upregulated. RpsF localizes in the decoding region of the central platform of the 30S subunit and interacts with the 16S rRNA (56). PrfB is recruited by the alternative ribosome rescue factor ArfA that can rescue stalled ribosomes by terminating translation by UGA and UAA stop codon recognition (57, 58). ArfA synthesis is tightly regulated by tmRNA (59) and is only induced when trans-translation is insufficient to restore ribosome function (60). It is therefore considered a major fail-safe mechanism for ribosome rescue (61). While it has been shown that the upregulation of peptide chain release factors is indicative of an exit tunnel block (62), it remains to be investigated whether upregulation of RpsA is indicative of an A-site block. This would be interesting, since it would connect trans-translation with an A-site block and alternative ribosome rescue mechanisms with an exit tunnel block.

Notably, tigecycline treatment also elicits the upregulation of fatty acid biosynthesis. Previous studies in B. subtilis demonstrated that the atypical tetracyclines chelocardin and anhydrotetracycline (25, 63) have dual mechanisms of action targeting translation and the cell membrane. While the proteomic data presented here could indicate that tigecycline also has a dual mechanism, using a fluorescence microscopy-based assay (Live/Dead BacLight; Invitrogen), we found no evidence of membrane damage (see Fig. S17). Exponentially growing cultures were treated with the physiologically effective concentration of tigecycline for 15 min or 2 h prior to staining with SYTO 9, which stains all cells, and with propidium iodide, which stains cells suffering an impaired membrane integrity. In contrast to the positive controls (cells treated with 70% ethanol) or cells treated with the physiologically effective concentration of colistin, there was no increase in tigecycline-treated cells that were stained with propidium iodide.

The stress response to CHIR-090 is distinct from that to frontline antibiotics.

Gram-negative specific cellular processes such as LPS biosynthesis are emerging as potent target areas for drug development, which was the main motivation for investigating the proteomic responses of a Gram-negative pathogen to antibiotics. Barb et al. demonstrated that the inhibition of LpxC, an essential metal-dependent UDP-3-O-(R-3-hydroxymyristoyl)-N-acetylglucosamine deacetylase, by CHIR-090 is a highly effective strategy for combatting Gram-negative pathogens including P. aeruginosa (14). LpxC catalyzes the committed step of lipid A biosynthesis, and its inhibition leads to membrane perturbation and impairment of the assembly of outer membrane proteins (14, 64). We analyzed the proteomic response to gain insights into the effects of CHIR-090 on bacterial physiology and to aid mode-of-action studies of structurally novel compounds that might impact on the same target area. The MIC in chemically defined minimal medium of CHIR-090 was 0.3 μg/mL. In growth experiments, the same concentration reduced the final OD of cultures treated in early log phase by approximately 30% (Fig. 3A), whereas higher concentrations caused cell lysis. CHIR-090 treatment affected protein synthesis rates only marginally (Fig. 1A). As briefly stated above, the proteomic response to CHIR-090 (Fig. 3B; see also Tables S2 to S19 in the supplemental material) was distinct from the response profiles of the other antibiotics with low similarities to polymyxin B and colistin responses. As for polymyxin, among the highly represented pathways in CHIR-090-treated P. aeruginosa were respiration and pyruvate metabolism, and the respective marker proteins AtpD and OadA are shared with polymyxin B. In addition, several marker proteins associated with iron homeostasis, virulence, and lipid and LPS biosynthesis were identified, which is consistent with an intricate protein-protein interaction network of LPS, phospholipid, and fatty acid biosynthesis enzymes (65) (Fig. 2B).

FIG 3.

Proteomic response to LPS biosynthesis inhibitor CHIR-090. (A) Impact of 0.3 μg/mL CHIR-090 on the growth. Antibiotic addition is marked by an arrow. (B) The protein synthesis rate as determined by l-[³⁵S] methionine incorporation is shown in relation to the untreated control. (C) The proteomic response profile is an overlay of autoradiographs representing l-[³⁵S]methionine pulse-labeled cytosolic proteins of an untreated control culture (green) and a CHIR-090-treated culture (red) that were separated by 2D-PAGE. A “#” followed by a number indicates the marker protein was identified in multiple spots. (D) Selected marker proteins discussed in the text with regulation factors (see Table S2 in the supplemental material for full list). RNAP, RNA polymerase.

Damage of the membrane and cell wall, as well as impairment of outer membrane protein assembly and transport processes, are discussed as primary consequences of the enrichment of detergent-like intermediates of the LPS synthesis pathway upon LpxC inhibition (67). Indeed, several marker proteins for CHIR-090 can be linked to cell envelope-associated processes (Fig. 3c). GlmU stands out among those, as it converts glucosamine 1-phosphate (GlcN-1-P) to UDP N-acetylglucosamine (UDP-GlcNAc), which serves as precursor both for LPS and peptidoglycan biosynthesis (68). LpxC inhibition thus leads to upregulation of precursor supply on the protein level as a direct countermeasure to the antibiotic action. Further marker proteins that are more indirectly involved in precursor supply include (i) the glycogen debranching enzyme GlgX and ADP sugar pyrophosphatase (AspP), which prevents glycogen synthesis (50), (ii) triosephosphate isomerase (TpiA), which connects glucose metabolism with phospholipid metabolisms by converting glyceraldehyde 3-phosphate to dihydroxyacetone phosphate (69), and (iii) pyruvate decarboxylase subunit B (OadA), which we discussed in the context of polymyxins as one of the signature proteins for impairment of the cytoplasmic and outer membranes. The negative effects of lipid A synthesis inhibition on cell envelope integrity and assembly of essential outer membrane proteins (67) is congruent with the upregulation of several proteins that are part of membrane-associated machineries or depend on export processes, including AtpD, a subunit of ATP synthase localized in the cytoplasmic membrane, β-lactamase PIB-1, and type VI secretion system protein PA2365 (TssB3). Other upregulated proteins are involved in replication (ParE), protection from ROS (Dps), translation (RpsF, AspS, and DksA), protein degradation (ClpA), and iron acquisition (PvdA and PvdO), indicating that multiple cellular processes are perturbed in response to CHIR-090 treatment.

Cell envelope-associated processes such as cellular respiration and recycling of cell envelope components (degradation of glycogen and glycolysis) were enriched pathways not only after exposure to CHIR-090 but also polymyxin treatment. Indeed, membrane and lipid A regeneration and modification are known to be relevant for polymyxin survival (70). As for polymyxins, we hypothesize that cell envelope damage is the primary challenge after CHIR-090 treatment.

Conclusion.

P. aeruginosa has been identified as one of the most difficult to treat bacterial species. Here, we investigated the proteomic response of the strain PAO1 to antibiotics commonly used in antipseudomonal therapy, as well as the experimental anti-Gram-negative agent CHIR-090. The mathematical comparison of proteomic responses allowed the grouping of aminoglycosides, of fluoroquinolones, of translation elongation inhibitors (tigecycline and azithromycin), and of polymyxins. Ceftazidime and piperacillin-tazobactam were only weakly active against exponentially growing high-density liquid cultures due to beta-lactamase activity, and meropenem, while active, did not result in the upregulation of proteins. Most surprisingly, no marker proteins were detected for rifampicin treatment despite a decent growth inhibition.

MATERIALS AND METHODS

Experimental details and citations for all methods and data collected from the literature are provided in the supplemental material. Described here are methods that were adapted for mode-of-action analysis in P. aeruginosa PAO1. Unless stated otherwise, experiments were performed in triplicates.

MIC testing.

Antimicrobial susceptibility was assessed in M9 minimal medium using the test tube assay, as described previously (26).

Radioactive pulse-labeling.

P. aeruginosa subcultures were grown in parallel in M9 minimal medium at 37°C under steady agitation at 200 rpm. In the mid-logarithmic-growth phase, antimicrobial compounds were added in sublethal physiologically effective concentrations that reduce growth to 50 to 80% of control conditions (see the supplemental material). We added 1.8 MBq l-[³⁵S]methionine 10 min after stress induction, and pulse-labeling was stopped after 5 min. Cells were harvested and washed with Tris-EDTA prior to cell disruption by ultrasonication. Incorporation rates were monitored by scintillation counting.

Proteomics.

Published protocols were followed for 2D gel-based proteomics (71). In brief, protein separation was conducted by isoelectric focusing on immobilized pH gradient strips (pH 4 to 7) in the first dimension and by SDS-PAGE in the second dimension. Relative protein synthesis rates were determined using autoradiography and spot densitometry using Decodon Delta 2D 4.1 image analysis software (26). Marker proteins that were at least 2-fold upregulated in each of three biological replicates were cut from nonradioactive gels and processed via tryptic in-gel digestion for mass spectrometry. Protein identification was conducted as described by Wüllner et al. (44). Protein identification data were uploaded to the Pride repository (PXD022879).

Proteomic response similarity clustering.

Proteomic responses were compared based on marker protein induction factors using the CoPR (Comparison of Proteomic Responses) similarity scoring as described previously (25). Results were displayed in force-directed networks using Cytoscape 3.5.2. Calculations for pathway representations were performed using R 4.0.0 using the ComplexHeatmap package in version 3.10 (72).

Beta-lactamase assay.

Beta-lactamase activity was determined in cell extracts from cultures treated with antibiotics for 15 min using a nitrocefin-based assay (73, 74). To this end, cells were grown and treated as described for 2D-PAGE analysis. After 15 min of treatment, cells from 1 ml of culture volume were harvested by centrifugation, washed twice with phosphate buffer (pH 7) and lysed using the VialTweeter as described for 2D-PAGE analysis. The cell debris was separated by centrifugation and discarded. The assay was performed by mixing 150 μL of protein extract with 150 μL of a 0.02-μg/mL nitrocefin (Cayman Chemical Company, Ann Arbor, MI) solution that was prepared from a 0.5-μg/mL stock solution in DMSO diluted in phosphate buffer (pH 7). The absorbance was recorded in an EnSpire Mutimode plate reader (Perkin-Elmer) to investigate.

Cell membrane integrity.

Membrane integrity after tigecycline, azithromycin, and colistin treatment was monitored using a Live/Dead BacLight bacterial viability kit (Invitrogen) according to the manufacturer’s instructions and as described previously (75). The assay uses the fluorescent dyes SYTO 9 to stain all cells and propidium iodide to stain cells with impaired membrane integrity. Cells were treated with physiologically effective concentrations of antibiotics, with 70% ethanol, or left untreated as controls prior to fixation with low-melting-point agarose prior to fluorescence microscopy.

Data availability.

Data for protein identification by nUPLC-ESI-MS were uploaded to the Pride repository (project “Adaptive Responses of Pseudomonas aeruginosa to Treatment with Antibiotics”; project accession no. PXD029948).