Lipid A in outer membrane vesicles shields bacteria from polymyxins

Abstract

The continuous emergence of multidrug-resistant bacterial pathogens poses a major global healthcare challenge, with Klebsiella pneumoniae being a prominent threat. We conducted a comprehensive study on K. pneumoniae's antibiotic resistance mechanisms, focusing on outer membrane vesicles (OMVs) and polymyxin, a last-resort antibiotic. Our research demonstrates that OMVs protect bacteria from polymyxins. OMVs derived from Polymyxin B (PB)-stressed K. pneumoniae exhibited heightened protective efficacy due to increased vesiculation, compared to OMVs from unstressed Klebsiella. OMVs also shield bacteria from different bacterial families. This was validated ex vivo and in vivo using precision cut lung slices (PCLS) and Galleria mellonella. In all models, OMVs protected K. pneumoniae from PB and reduced the associated stress response on protein level. We observed significant changes in the lipid composition of OMVs upon PB treatment, affecting their binding capacity to PB. The altered binding capacity of single OMVs from PB stressed K. pneumoniae could be linked to a reduction in the lipid A amount of their released vesicles. Although the amount of lipid A per vesicle is reduced, the overall increase in the number of vesicles results in an increased protection because the sum of lipid A and therefore PB binding sites have increased. This unravels the mechanism of the altered PB protective efficacy of OMVs from PB stressed K. pneumoniae compared to control OMVs. The lipid A-dependent protective effect against PB was confirmed in vitro using artificial vesicles. Moreover, artificial vesicles successfully protected Klebsiella from PB ex vivo and in vivo. The findings indicate that OMVs act as protective shields for bacteria by binding to polymyxins, effectively serving as decoys and preventing antibiotic interaction with the cell surface. Our findings provide valuable insights into the mechanisms underlying antibiotic cross-protection and offer potential avenues for the development of novel therapeutic interventions to address the escalating threat of multidrug-resistant bacterial infections.

Keywords: antimicrobial peptides (AMP); bacterial extracellular vesicles; bacterial resistance mechanisms; last‐resort antibiotic; lipid A; multi‐drug resistance (MDR); polymyxins.

FIGURE 1

Stress conditions alter the size and quantity of Klebsiella pneumoniae (Kp) outer membrane vesicles (OMVs). (a‐d) K. pneumoniae was subjected to the indicated stress conditions or left untreated as a control for 90 min. Bacterial quantification was performed by plating K. pneumoniae on agar plates, while the size and amount of released OMVs were measured using nFCM in the sterile‐filtered supernatant. (a) OMV release was calculated relative to the amount of K. pneumoniae and normalized to the untreated control (ctr). (b) Average size of OMVs released by stressed and control K. pneumoniae. (c, d) K. pneumoniae was incubated with increasing doses of Polymyxin B (PB). OMV release was calculated relative to the quantified K. pneumoniae after 90 min of PB treatment (c), and cytotoxicity was assessed by propidium iodide staining (d). The percentage of dead K. pneumoniae was calculated relative to heat‐killed (HK) bacteria. Bars show mean values +SEM of three to five independent experiments. Statistics: one‐way ANOVA (Tukey's multiple comparisons test); **p < 0.01, ***p < 0.001, ****p < 0.0001 (a, b, d) compared to control (ctr) or (c) to 0.25 μg/mL PB; PB: Polymyxin B, ns: not significant, ctr: control, n ≥ 3.

FIGURE 2

OMVs mediate protection against Polymyxin B (PB) and Colistin. K. pneumoniae growth kinetics were determined under continuous shaking at 37°C, and the optical density at 600 nm wave length (OD₆₀₀) was measured every 30 min using a plate reader. All biological replicates were performed in technical triplicates. (a) K. pneumoniae was incubated with 4 μg/mL PB with or without the addition of OMVs from control K. pneumoniae (Kp‐ctr OMVs) or K. pneumoniae stressed with 1 μg/mL PB (Kp‐PB OMVs) or left untreated as a control. (b) The OMV dose‐dependent protection effect of Kp‐PB OMVs against PB is shown. K. pneumoniae was incubated with 4 μg/mL PB and decreasing concentrations of Kp‐PB OMVs (ranging from 10 μg/mL to 4 μg/mL Kp‐PB OMVs) or left untreated for control. The x‐fold increase in time [h] to reach the mid‐log phase (OD₆₀₀ = 0.4) is presented, calculated by comparing each Kp‐PB OMV dose with untreated control K. pneumoniae. (c) Protection of K. pneumoniae against Colistin mediated by Kp‐ctr OMVs or Kp‐PB OMVs is shown as an x‐fold change normalized to the untreated control K. pneumoniae. (d, e) K. pneumoniae was incubated with 35 μg/mL Gentamicin (d) or 50 ng/mL Meropenem (e) and Kp‐ctr OMVs or Kp‐PB OMVs. Untreated Klebsiella served as control. Graphs (a‐e) show mean values ± SEM (a, b, d, e) of three to four independent experiments. Statistics: (a, b, d, e) one‐way ANOVA (Tukey´s multiple comparisons test), compared to control without antibiotic and as indicated by brackets; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns: not significant; n ≥ 3.

FIGURE 3

Protein profile of antibiotic‐stressed Klebsiella. K. pneumoniae was incubated for 90 min with 1 μg/mL Polymyxin B (PB) or 20 μg/mL Gentamicin (Genta) with or without Kp‐ctr OMVs or left untreated as a control. The bacterial protein profile was analyzed by shotgun proteomics and liquid chromatography‐mass spectrometry. (a) Principal component analysis of label‐free quantitative proteomics data, including all significantly regulated proteins compared to untreated control. (b) Heatmap of all differentially expressed proteins compared to untreated Klebsiella is shown for all replicates. Hierarchical clustering was performed. (c) Volcano plots of proteins identified in the differently treated bacteria compared to the untreated control K. pneumoniae. The log2 fold change (cut‐off: 1) versus ‐log10 p‐value (cut‐off: 2) is shown. (d) Gene ontology analysis of the highest enriched pathways in K. pneumoniae treated with PB in comparison to K. pneumoniae treated with PB in the presence of Kp‐ctr OMVs is shown. Significantly regulated proteins of both conditions were compared. The top three regulated biological processes (upper) and cellular components (lower) for PB‐treated K. pneumoniae are shown. Graphs show the results of three biological independent experiments. ctr: control, Genta: Gentamicin, PB: Polymyxin B, PC: principle component.

FIGURE 4

OMVs confer inter‐family protection and exhibit ex vivo as well as in vivo protection against PB. (a, b) The protection against PB mediated by OMVs from control or PB‐stressed bacteria was determined by performing bacterial growth kinetics. Growth kinetics were determined under continuous shaking at 37°C, and the OD₆₀₀ was measured every 30 min using a plate reader. All biological replicates were performed in technical triplicates. The highest concentration of PB that did not significantly alter bacterial growth compared to untreated control bacteria was used to calculate the fold change of Polymyxin B protection, indicating how much more PB the bacteria can tolerate in the presence of the respective OMVs. (a) Heatmap of the protection of different Enterobacteriaceae acceptor bacteria against PB in the presence of different Enterobacteriaceae OMVs isolated from control (ctr) or PB‐stressed donor bacteria. (b) Protection of Pseudomonas aeruginosa or Legionella pneumophila against PB mediated by Kp‐PB OMVs or Kp‐ctr OMVs. (c) Precision‐cut lung slices (PCLS) were infected with K. pneumoniae for 4 or 8 h, respectively, and additionally treated with 6 μg/mL PB with or without Kp‐PB OMVs, each in technical duplicates. As a control, PCLS were infected with K. pneumoniae without any treatment. Bacterial load at the respective time points is displayed in CFU/mL. (d) Galleria mellonella were infected with K. pneumoniae by intra‐hemocoel injection through the last leg pair and additionally treated with either PB or PB in combination with OMVs. As controls, G. mellonella was infected with K. pneumoniae or OMVs together with PB. Survival was monitored for 7 days. Per group and replicate, 10–15 larvae were used. Graphs show mean values of 3–5 independent replicates ± SEM (c, d). Statistics: (c) two‐way ANOVA (Sidak`s multiple comparisons test); compared to respective control or (d) unpaired t‐test; compared as indicated by brackets; *p < 0.05, **p < 0.01; ns: not significant, PB: Polymyxin B, ctr: control; Kp: commercially available Klebsiella pneumoniae strain (MGH 78578); Kp‐i: multi‐drug resistant Klebsiella pneumoniae clinical isolate; Ec: Escherichia coli; Sal: Salmonella enterica serovar Typhimurium; n ≥ 3.

FIGURE 5

OMVs can bind and become saturated with PB. (a) Schematic illustration (created with biorender.com) of experimental setup used in b and c. (b, c) OMVs were pre‐incubated with or without PB and subjected to ultracentrifugation. The vesicle‐free supernatant (b) or the OMV‐containing pellet (c) was added to K. pneumoniae growth experiments with or without the addition of fresh PB while bacterial growth. Bacterial replication was determined by colony forming units (CFU). (d) Kp‐ctr OMVs or Kp‐PB OMVs were incubated with increasing amounts of PB (5‐500 μg/well) and 1‐N‐phenylnapthylamine. Fluorescence intensity was measured at 405 nm using a plate reader. Results are shown relative to Saponin, which was used as a positive control for vesicular lysis and is depicted as 100%. (e) Kp‐ctr OMVs or Kp‐PB OMVs were incubated with increasing amounts of PB (0‐10 μg/mL). After ultracentrifugation, the bound PB in the vesicle fraction was quantified via mass spectrometry. (b, e) Bars and curves are mean values of three to four independent experiments ± SEM. Statistics: (b, c) one‐way ANOVA (Tukey's multiple comparisons test) and (d) two‐way ANOVA (Sidak`s multiple comparisons test), compared to the sample only containing Kp or as indicated by brackets (b, c) or Saponin (d); * p < 0.05, ***p < 0.001, ****p < 0.0001; n ≥ 3.

FIGURE 6

Lipid composition of OMVs and their donor bacteria. (a‐g) K. pneumoniae (Kp) and PB‐resistant K. pneumoniae (KpR) were cultured in the presence or absence of PB. OMVs and bacteria were isolated, and their lipids extracted via chloroform/methanol‐extraction (a‐d) or chloroform/ methanol‐extraction combined with a mild hydrolysis (e‐g) and analyzed via mass spectrometry. (a) Heatmap illustrating the hierarchical clustering of lipid classes across all samples and detected lipids. (b) Principal component analysis of lipidomics data. (c) Volcano plot of up‐ and down‐regulated lipids between Kp‐ctr OMVs and Kp‐PB OMVs. The log2 fold change versus ‐log10 p‐value is shown. (d) Comparison of lipid class abundancies between Kp‐ctr OMVs and Kp‐PB OMVs. (e, f) The normalized peak area for lipid A in OMVs (e) or bacterial (f) samples. Lipid A was normalized to the amount of OMVs (e) or the biomass (f) used for lipid extraction. Graphs show the results of four biological independent experiments. (g) Heatmap of the quantification of all detected nominal masses of the m/z ratio of lipid A molecules found in OMV samples. Statistics: (e) one‐way ANOVA (Tukey`s multiple comparisons test) of log10 transformed data, compared to Kp‐ctr OMVs; **p < 0.01, n ≥ 3; ns: not significant, TG: Triacylglycerol, SM: Sphingomyelin, PI: Phosphatidylinositol, PG: Phosphatidylglycerol, PEtOH: Phosphatidylethanol, PE: Phosphatidylethanolamine, PC: Phosphatidylcholine, LPE: Lysophosphatidylethanolamine, LPC: Lysophophatidylcholine, HexCer: Hexosylceramide, Hex2Cer: Dihexosylceramide, HBMP: Hemibismonoacylglycerophosphate, DGDG: Digalactosyldiacylglycerol, DG: Diacylglycerol, CoQ: Coenzyme Q, Cer: Ceramide, BMP: Bismonoacylglycerophosphate.

FIGURE 7

Artificial vesicles containing Kdo2‐lipid A protect K. pneumoniae from PB. (a‐f) Artificial vesicles with increasing amounts of Kdo2‐lipid A (KLA, 0%–40%) were generated via membrane extrusion. Size distribution profiles from nFCM (a) and representative TEM images (b) of artificial vesicles are shown. (c, d) The in vitro protection potency of artificial vesicles against PB was determined by performing growth kinetics with K. pneumoniae. Growth kinetics were determined under continuous shaking at 37°C, and OD₆₀₀ was measured every 30 min using a plate reader. All biological replicates were performed in technical triplicates. (d) A heatmap of the time [h] K. pneumoniae need to reach OD₆₀₀ = 0.4 (mid of the logarithmic phase) in the presence or absence of increasing amounts of PB (0‐2 μg/mL) and with or without artificial vesicles with increasing amounts of KLA (10%‐40%) is shown. Graph shows mean values of three independent experiments. (e) Precision‐cut lung slices (PCLS) were infected with K. pneumoniae for 4 h, and additionally treated with 6 μg/mL PB with or without artificial vesicles containing 40% KLA, each in technical duplicates. As a control, PCLS were infected with K. pneumoniae without any treatment. Bacterial load at 4 h post infection is displayed in CFU/mL. Graph shows four biological independent experiments. (f) Galleria mellonella were infected with K. pneumoniae by intra‐hemocoel injection through the last leg pair and additionally treated with either PB or PB in combination with artificial vesicles containing 40% KLA. As controls, G. mellonella were injected with PBS (solvent control), K. pneumoniae or artificial vesicles together with PB. Survival was monitored for 7 days post‐infection (p.i.). Per group and replicate, 10–15 larvae were used. (c‐f) Graphs show mean values of three to five independent experiments ± SEM. Statistics: (c, e, f) one‐way ANOVA (Tukey`s multiple comparisons test), compared as indicated by brackets; *p < 0.05, **p < 0.01, ****p < 0.0001; ns: not significant.