Antibiotic resistance alters the ability of Pseudomonas aeruginosa to invade bacteria from the respiratory microbiome

Generation of antibiotic-resistant P. aeruginosa strains

A green fluorescent protein (GFP)-tagged strain of P. aeruginosa PAO1 (PAO1-GFP; Vogwill et al., 2016) was used so that reads in the GFP channel could measure P. aeruginosa invasion ability (excitation: 485/20, emission: 516/20). This strain was created by insertion of GFP and a gentamicin resistance cassette into the Tn7 transposon insertion site of the reference strain of P. aeruginosa PAO1 (Vogwill et al., 2016). Three independent spontaneous resistant mutants were selected on each of the three antibiotics (ciprofloxacin, ceftazidime, and meropenem) at the clinical breakpoints according to the EUCAST Clinical Breakpoint Tables v.12.0 (ciprofloxacin 0.5 μg/ml, ceftazidime 8 μg/ml, and meropenem 8 μg/ml) (EUCAST, 2024). These antibiotics were chosen to represent different classes of clinically important antipseudomonal drugs for treating P. aeruginosa infections (fluoroquinolones, cephalosporins, and carbapenems).

To generate spontaneous resistant mutants, PAO1-GFP was grown from glycerol stock on LB (Lysogeny Broth) Miller with agar (Sigma-Aldrich) plates supplemented with 15 μg/ml gentamicin overnight at 37 °C. Single independent colonies were grown in LB Miller broth overnight at 37 °C in biological triplicate with shaking at 225 rpm, then 150 ul of overnight culture was plated on LB Miller agar plates supplemented with either 0.5 μg/ml ciprofloxacin, 8 μg/ml ceftazidime, or 8 μg/ml meropenem. Spontaneous resistant mutants of PAO1-GFP successfully grew on both 0.5 μg/ml ciprofloxacin and 8 μg/ml ceftazidime. A single colony from each plate was inoculated into LB Miller broth supplemented with the corresponding antibiotic at 50% of the selective concentration (i.e., 0.25 μg/ml ciprofloxacin and 4 μg/ml ceftazidime) for overnight growth. Three independent resistant mutants for each antibiotic were stored as glycerol stocks at −80 °C (Supplementary Table S1). No resistant mutants of PAO1-GFP were able to grow on meropenem 8 μg/ml, and so PAO1-GFP was serially passaged through progressively higher meropenem concentrations over 3 days to generate meropenem-resistant mutants, ramping from 1 μg/ml to 2 μg/ml to 4 μg/ml with transfer every 24 hr before plating on 8 μg/ml meropenem. The same procedure for stocking was then followed as above. A total of nine resistant mutants of PAO1-GFP were generated across the three antibiotics which are hereby referred to as: ciprofloxacin-resistant mutants 1–3 (cipR1-cipR3), ceftazidime-resistant mutants 1–3 (cefR1-cefR3), and meropenem-resistant mutants 1–3 (merR1-merR3).

Antibiotic resistance phenotyping via minimum inhibitory concentration assays

Isolates were grown from glycerol stocks on LB Miller agar plates overnight at 37 °C. Single colonies were then inoculated into tryptic soy broth (Sigma-Aldrich) for overnight growth at 37 °C with shaking at 225 rpm, after which overnight suspensions were serial diluted to ~5 × 10⁵ CFU/ml. Antibiotic resistance phenotyping was carried out as minimum inhibitory concentration (MIC) assays via broth microdilution as defined by EUCAST guidelines (EUCAST, 2022), with the alteration of tryptic soy broth for the growth media to replicate the media used for the invasion assays. Minimum inhibitory concentration was calculated using twofold dilution series for meropenem (from 0.5 to 64 μg/ml), ciprofloxacin (from 0.03125 to 4 μg/ml), and ceftazidime (from 0.5 to 64 μg/ml) designed to capture ranges either side of the clinical breakpoint (EUCAST, 2024). Growth inhibition was defined as OD₅₉₅ < 0.200 and we calculated the MIC of each isolate as the median MIC score from three biological independent assays of each isolate. For some strains, the MIC was the lowest concentration in this range. However, as we were using this assay to determine increased resistance around the clinical breakpoint that sat in the midpoint of these ranges, this was not an issue.

Growth rate assays

Growth rate assays were used to characterize growth of the antibiotic-resistant mutants compared to the ancestral strain (PAO1-GFP) in tryptic soy broth in the absence of any respiratory microbes. Pseudomonas aeruginosa strains were grown from glycerol stock on LB Miller agar plates overnight at 37 °C, then inoculated into tryptic soy broth for 18–20 hr overnight growth at 37 °C with shaking at 225 rpm. Overnight suspensions were diluted to an OD₅₉₅ of ~0.05 and placed within the inner 60 wells of a 96-well plate equipped with a lid. To assess growth, isolates were grown in tryptic soy broth at 37 °C in a BioTek Synergy 2 microplate reader set to moderate continuous shaking for 24 hr and reads in the GFP channel (excitation: 485/20, emission: 516/20) were taken at 10-min intervals. Relative fluorescence units (RFU) of P. aeruginosa growth over 24 hr were plotted using the Growthcurver package in R (Sprouffske & Wagner, 2016).

Invasion assays of GFP-tagged P. aeruginosa strains into the respiratory microbiome strains

An invasion assay in which endpoint measurements of fluorescence in the GFP channel (excitation: 485/20, emission: 516/20) at 24 hr was used to measure the ability of PAO1-GFP and the nine spontaneous antibiotic-resistant mutants to invade dense overnight cultures of the six respiratory microbiome strains. To conduct this assay, the 10 P. aeruginosa strains (PAO1-GFP, cipR1, cipR2, cipR3, cefR1, cefR2, cefR3, merR1, merR2, and merR3) and six respiratory microbiome strains (S. epidermidis, R. mucilaginosa, S. gordonii, S. lugdunensis, S. oralis, and S. salivarius) were grown from glycerol stock on LB Miller agar plates overnight at 37 °C. Single colonies of each P. aeruginosa strain were inoculated into the inner 60 wells of a 96-well plate with wells filled with 200 ul tryptic soy broth for overnight growth at 37 °C and shaking at 225 rpm, for six biological replicates per strain. Single colonies of each respiratory microbiome strain were inoculated into 14 ml falcon tubes filled with 7 ml tryptic soy broth for overnight growth at 37 °C with shaking at 225 rpm, for six biological replicates per strain.

The next day, a black-sided 96-well plate was labeled for each respiratory microbiome strain, and the following plate layout was used to arrange respiratory microbe invasions (Supplementary Figure S1). In brief, each plate contained six biological replicates of a respiratory microbiome strain that had been allowed to grow for 24 hr. Six biological replicates of the 10 P. aeruginosa strains were then inoculated into the wells containing the pre-established respiratory microbe cultures, with the invasion set up as a 1/20 dilution of P. aeruginosa. Three 96-well plates were used for tryptic soy broth controls, in which the same six replicates of the 10 P. aeruginosa strains were inoculated (1/20 dilution) into sterile tryptic soy broth for three technical replicates of P. aeruginosa growth. This paired growth of P. aeruginosa invasion inoculations was done in order to calculate P. aeruginosa invasion into respiratory microbe culture as a percentage of the growth achieved in rich media in the absence of that respiratory microbe. The plates were incubated for 24 hr at 37 °C with shaking at 225 rpm. After 24 hr, OD₅₉₅ and GFP (excitation: 485/20, emission: 516/20) of the respiratory microbiome invasions and the tryptic soy broth paired growth assays were read in a BioTek Synergy 2 microplate reader.

Plates in which PAO1-GFP did not grow in 24 hr were then incubated for a further 12 hr, and OD₅₉₅ and GFP were measured again at 36 hr. Invasion ability is given as percentage RFU (arbitrary unit) of the invasion assay standardized against RFU in the absence of the respiratory microbe. Fluorescence is proportional to cell density across the range used in our experiments. Prior to these invasion assays in which endpoint measurements were taken, invasion was first continually monitored in a BioTek Synergy 2 microplate reader taking reads at 10-min intervals for 24 hr. This preliminary data were used to inform the endpoint assays and produce invasion growth curves for illustrative purposes (Supplementary Figure S2), which was done using the Growthcurver package in R (Sprouffske & Wagner, 2016).

Genome sequencing and variant detection of spontaneous antibiotic-resistant mutants of P. aeruginosa

The ten P. aeruginosa strains (PAO1-GFP, cipR1, cipR2, cipR3, cefR1, cefR2, cefR3, merR1, merR2, and merR3) were sequenced using Illumina short-read sequencing. DNA extraction, library preparation, and sequencing were performed by MicrobesNG (Birmingham, UK) according to their protocols. Briefly, ~4–6 × 10⁹ cells were concentrated and suspended in 500 μl of DNA/RNA Shield (Zymo Research) in 2 ml screw cap tubes, and stored at room temperature prior to sample submission. Between 5 and 40 μl of cell suspension was lysed with 120 μl TE buffer containing lysozyme (final concentration 0.1 mg/ml) and RNase A at 37 °C for 25 min. Protein digestion was performed using Proteinase K and SDS with incubation at 65 °C for 5 min. Genomic DNA was purified using an equal volume of paramagnetic beads (SPRI beads, Beckman Coulter) and resuspended in EB buffer (10mM Tris–Hcl, pH 8.0). DNA was quantified with QuantiT dsDNA HS Kit (ThermoFisher Scientific) using an Eppendorf AF2200 plate reader and diluted to an appropriate concentration for library preparation.

Library preparation was performed using the Nextera XT Library Prep Kit according to the manufacturer’s protocol with modifications (i.e., twofold increase in input DNA, PCR elongation increased to 45 s). DNA quantification and library preparation were performed using a Hamilton Microlab STAR liquid handling system. Libraries were sequenced using the Illumina NovaSeq6000 platform using a 250 bp paired-end protocol. Read adapter trimming was performed using Trimmomatic version 0.30 (Bolger et al., 2014) with a sliding window quality cutoff of Q15. Trimmed reads were aligned to the PAO1-UW reference genome (NCBI RefSeq accession NC_002516.2) (Klockgether et al., 2010). Variants in PAO1-GFP and the resistant strains were called using the breseq 0.36.1 pipeline (Deatherage & Barrick, 2014). Variants reported in the main text are those found in the resistant strains compared to the PAO1-GFP background and called at 100% (Supplementary Table S2). Gene descriptions in Supplementary Table S2 are added from the Pseudomonas genome database (Winsor et al., 2016).

Spent media assays with S. lugdunensis

To better characterize the interaction between P. aeruginosa and S. lugdunensis, we repeated the invasion assays using clinical isolates of P. aeruginosa from two patients where meropenem resistance evolved during infection and the de novo resistance mutations are known (Supplementary Table S1; Wheatley et al., 2021, 2022). A spent media assay was used for this invasion assay as the clinical isolates were not GFP-tagged (Dos Santos et al., 2022). The invasion assay method used previously was conducted with the following alterations. After overnight growth of S. lugdunensis cultures, spent media was prepared via centrifugation at 16,000 g/min for 1 min to pellet the S. lugdunensis cells followed by filter sterilization of the supernatant through a 0.22 μM filter. Pseudomonas aeruginosa strains were inoculated into this cell-free spent media (1/20 dilution) in a 96-well plate as described above alongside paired replicate inoculations into sterile tryptic soy broth. For each clinical P. aeruginosa genotype (ST782-WT, oprD mexR ST782, ST17-WT, and oprD ST17), three biological replicates of three genetically identical isolates from the patient were measured (Supplementary Table S1), and PAO1-GFP was included in these assays as a control.

To test for the presence of potential growth inhibition molecules in the supernatant of S. lugdunensis, we prepared fresh cell-free spent media as described above. The media was subsequently used to either (a) spot 10 µl directly onto, or (b) inoculate a sterile filter disc placed onto, plates where PAO1-GFP had been spread using sterile beads. These plates were inoculated with dilutions of PAO1-GFP and we selected the dilution that resulted in confluent CFU growth. This was done in triplicate, using ddH₂O as negative control. Plates were incubated overnight and inspected for the presence of zones of inhibition.

Invasion of P. aeruginosa-resistant mutants into pre-established respiratory microbe cultures

We measured the ability of the P. aeruginosa-resistant mutants to invade the pre-established cultures of the respiratory microbiome strains by measuring GFP at 24 hr after P. aeruginosa inoculation (Figure 3A). We calculated invasion ability as the growth of P. aeruginosa in the presence of the respiratory microbe compared to the paired growth of P. aeruginosa in tryptic soy broth (in the absence of a respiratory microbe). Invasion assay results are shown as percentage growth in respiratory microbe invasion compared to the media control (Figure 3A). Prior to these assays in which endpoint measurements were taken, invasion was first continually monitored in a BioTek Synergy 2 microplate reader (Supplementary Figure S2).

Open in a separate window

Figure 3.

Schematic of invasion assay method (A) and invasion of P. aeruginosa into respiratory microbiome strains in 24 hr (B, C, D, F, G, and H) and 36 hr (E). Each bar chart shows invasion ability measured as percentage RFU compared to the media control at that same timepoint (y-axis), for invasion into pre-established cultures of: (B) R. mucilaginosa, (C) S. epidermidis, (D) S. lugdunensis (24 hr), (E) S. lugdunensis (36 hr), (F) S. gordonii, (G) S. salivarius, and (H) S. oralis. Values obtained in S. gordonii, S. salivarius, and S. oralis were considered negligible and below the limits of the assay (below 2%). Differences in invasion ability between the P. aeruginosa antibiotic-resistant strains and PAO1-GFP were assessed using a one-way ANOVA followed by a post hoc Dunnett’s test, and significance (Dunnett’s test: p < 0.05) is indicated by an asterisk (*). Bar charts show the mean of six biological replicates ± standard error of the mean.

All respiratory microbe cultures provided some degree of resistance to P. aeruginosa growth, and this effect was particularly strong for the Streptococcus species and S. lugdunensis (Figure 3). Pseudomonas aeruginosa was best able to invade R. mucilaginosa, with PAO1-GFP reaching ~86% (±4%) of the RFU reached in the absence of R. mucilaginosa (Figure 3B). Second to this was S. epidermidis, with PAO1-GFP reaching ~70% (±2%) of the RFU reached in the absence of S. epidermidis (Figure 3C). The ceftazidime- and meropenem-resistant mutants showed no change in invasion ability into R. mucilaginosa and S. epidermidis (Figure 3B, ​,C).C). However, the ciprofloxacin-resistant mutants had an impaired invasion ability in both R. mucilaginosa (2/3: cipR1, cipR3) and S. epidermidis (3/3: cipR1, cipR2, and cipR3) (Figure 3B, ​,C),C), suggesting a clear cost to resistance in this context.

The ceftazidime-resistant mutants showed no change in invasion ability into cultures of S. lugdunensis, but ciprofloxacin- and meropenem-resistant mutants demonstrated an increased invasion ability (Figure 3D). While the ancestral PAO1-GFP strain was unable to invade S. lugdunensis in 24 hr, five of the resistant mutants (cipR1, cipR3, merR1, merR2, and merR3) were able to grow in this same timeframe (defined as at least 2% of RFU in media in absence of S. lugdunensis) (Figure 3D). The invasion of the resistant mutants into S. lugdunensis was near the lower limit of detection (Figure 3D). We extended this assay to 36 hr. In 36 hr, PAO1-GFP was now able to invade the pre-established S. lugdunensis culture (Figure 3E). The five resistant mutants (cipR1, cipR3, merR1, merR2, and merR3) demonstrated a trend toward increased invasion ability (Figure 3E), suggesting a benefit to resistance in this context. It should be noted that statistical significance as assessed by a one-way ANOVA followed by a post hoc Dunnett’s test (p < 0.05) was only observed for cipR1 at 24 hr and merR1/merR2 at 36 hr (Figure 3E).

Finally, the Streptococcus spp. (S. gordonii, S. salivarius, and S. oralis) showed strong resistance to P. aeruginosa growth, which was maintained regardless of the antibiotic resistance genotype (Figure 3F, ​,G,G, ​,H).H). Pseudomonas aeruginosa was unable to invade in 24 hr, and this effect was preserved when we extended the assay to 36 hr (Supplementary Figure S3). Across these assays (Figure 3), changes to invasion ability were observed for the ciprofloxacin- and meropenem-resistant mutants, but the three ceftazidime-resistant mutants (cefR1-cefR3) associated with mutations in dacB were not found to alter invasion into any of the six respiratory microbes.

Figures within this publication were created with Biorender.com (Figures 1, ​,3A,3A, and ​and4A).4A). We would like to thank the Calleva Research Centre for Evolution and Human Sciences at Magdalen College (Oxford) for their support of this work. We would like to thank MicrobesNG (Birmingham, UK) for the generation and initial processing of sequencing data. We would like to thank Jason L Brown and Gordon Ramage (Oral Sciences Research Group, University of Glasgow), Angela Nobbs (University of Bristol), and Josh Thomas and Ashleigh Griffin (University of Oxford) for strains provided and used in this study. We would like to thank Alexandre Figueiredo and Erik Bakkeren for their comments on the manuscript.