Experimental models and subject detailsBacterial strains and growth conditions

Bacterial strain information is provided in Table 1. Where listed, growth media were prepared according manufacturer recommendations: LB broth and LB broth supplemented with 0.3 mM thymine (BD Biosciences 244610, Alfa Aesar A15879), 0.3 mM methionine (Sigma-Aldrich M9625), 0.3 mM inosine (EMD Millipore 4060), cation-adjusted Mueller–Hinton II broth (CAMHB) (BD 212322) or 0.3 mM thymidine (Sigma-Aldrich T1895), Gutnick minimal media (1.0 g l−1 K2SO4, 13.5 g l−1 K2HPO4, 4.7 g l−1 KH2PO4, 0.1 g l−1 MgSO4·7H2O, 10 mM NH4Cl as a nitrogen source and 0.4% w/v glucose as a carbon source)65.

Animal models

For pharmacokinetics determination, care and handling of male CD-1 mice approximately 6–8 weeks old conformed to institutional animal care and use policies as carried out at Pharmaron.

For P. aeruginosa thigh infection model and maximum tolerated dose (MTD) studies, care and handling of female 5–6-week-old CD-1 conformed to Institutional Animal Care and Use Committee (IACUC) policies as carried out at the University of North Texas Health Science Center (Fort Worth, Texas) under UNTHSC IACUC-approved protocol nos. IACUC-2021-0003 and IACUC-2020-0039. Rodents were fed either base chow (Envigo) or base diet supplemented with 1.8 g kg−1 thymidine (Sigma T9250) starting at 2 days before infection. All procedures were conducted in accordance with the UNTHSC IACUC-approved protocol. Animals were housed in rooms undergoing 10–15 air changes per hour. Air provided to the animal rooms was controlled for temperature and humidity and is fully monitored 24 h a day. Air pressure was balanced according to room use. Lighting was controlled in individual rooms by automatic timers with a standard 12 h on, 12 h off cycle.

MIC results

MIC was defined as the lowest concentration of antibiotic at which no visible growth was detected after 16 h at 37 °C. Overnight cultures were diluted 1:150 in LB broth and added to a 96-well plate. Antibiotics were serially diluted 1:2 and added to columns of the 96-well plate and grown at 37 °C with continuous shaking. Cell growth was measured using optical density at 600 nm (OD600). MIC assays were performed in either BioTek Synergy HT or Tecan InfiniteM200 Pro microplate readers.

For MIC calculations performed by WuXi, MIC was calculated as the lowest concentration that inhibits visible growth after 18 h. Bacterial colonies (4–8) of strains of interest were vortexed in saline and adjusted to an OD600 of 0.2. Strains were diluted 1:200 into CAMHB media in 96-well plates. Antibiotics were serially diluted 1:3 in DMSO, and 1 μl of each dilution was added to bacteria. The plates were incubated for 18–20 h at 37 °C before observation.

Colony-forming units counting

P. aeruginosa PA14 overnight cultures were diluted 1:100 and grown to exponential phase (OD600 = 0.4–0.6). Cultures were diluted 1:10 and treated with antibiotic. At each time point, 150 μl of culture was removed, serial diluted 1:10 six times and plated onto LB agar plates. Plates were grown overnight at 37 °C, after which visible colonies were counted. C.f.u.s ml−1 are reported from dilutions in which ~10–100 single colonies were visible.

Membrane potential and permeability assay

Overnight P. aeruginosa PA14 or E. coli lptD4213 cultures were diluted 1:100 and grown to mid exponential phase at 37 °C. Cultures were diluted 1:10 into PBS and treated with antibiotics for 15 min. P. aeruginosa PA14 was stained with TO-PRO-3 (640 nm excitation, 670/30 nm emission) to measure cell membrane integrity. E. coli lptD4213 was stained with both TO-PRO-3 and DiOC2(3) (ThermoFisher B34950) to measure cell membrane integrity and membrane potential. DiOC2(3) was evaluated as a ratio of green (488 nm excitation, 525/50 nm emission) to red (488 nm excitation, 610/20 nm emission)66. The LSRII flow cytometer (BD Biosciences) at the Flow Cytometry Resource Facility, Princeton University, was used to measure the fluorescent intensities of both dyes in response to antibiotic treatment. 100,000 events were recorded for each data file. Gates for permeabilization were determined using polymixin B (Sigma-Aldrich P1004) and untreated controls. Gates for depolarization were determined using carbonyl cyanide m-chlorophenyl hydrazone (CCCP) as a positive control. Data were analysed using FlowJo v.10 software.

MexAB-OprM transposon mutants

P. aeruginosa PA14 transposon mutants were generated by the Ausubel Lab (https://pa14.mgh.harvard.edu/cgi-bin/pa14/home.cgi)67. The MICs of fluorofolin and TMP against strains with disrupted MexA, MexB and OprM were determined as above and compared to the parental strain. As transposon mutants in MexB were represented twice in this collection, the MIC was confirmed across both mutant strains.

Haemolysis

Defibrinated sheep red blood cells (Lampire 50414518) were diluted to 6 × 106 cells per ml, pelleted and washed 3× with PBS. Samples were treated at 37 °C with shaking for 1 h and then centrifuged. Supernatants were collected and absorbances were measured at 405 nm in a Tecan InfiniteM200 Pro microplate reader. Percentage haemolysis was calculated compared to 100% lysis by Triton X-100 (1% v/v) (Sigma-Aldrich X100RS).

Mammalian cell cytotoxicity

HK-2 (500 cells per well) (ATCC CRL-2190), HLF (500 cells per well) (Cell Applications 506K-05a), WI-38 (500 cells per well) (ATCC CCL-75) or PBMC (5,000 cells per well) (TPCS PB010C) cells were seeded in white opaque 384-well plates. After 24 h, DMSO or compounds were added in 3-fold dilutions and cells incubated for 72 h. For PBMC, CellTiter-Glo reagent was added in equal volume and incubated for 30 min, after which luminescence was read. For other cell types, CyQUANT detection reagent was added in equal volume and cells incubated for 1 h, after which fluorescence was read with standard green filter set (508/527 nm excitation). Cell toxicity was evaluated by Pharmaron.

Metabolomics

Overnight P. aeruginosa PA14 cultures were diluted 1:150 in Gutnick minimal media and grown to early–mid exponential phase (OD600 = 0.4–0.6). Cultures were treated with either 6.3 μg ml−1 fluorofolin (2× MIC) or 250 μg ml−1 trimethoprim (2× MIC) (Chem-Impex 01634) for 15 min. Metabolites were extracted by vacuum filtering 15 ml of treated cells using 0.45 μm HNWP Millipore nylon membranes and placing the filters into an ice-cold quenching solution of 40:40:20 methanol:acetonitrile:H2O. Extracts were kept on dry ice for 1 h and centrifuged at 16,000 g for 1 h at 4 °C. The supernatant was kept at −80 °C until mass spectrometry analysis.

Liquid chromatography–mass spectrometry (LC–MS) analysis of metabolites was performed on an Orbitrap Exploris 240 mass spectrometer coupled with hydrophilic interaction liquid chromatography (HILIC)68. HILIC was on an XBridge BEH Amide column (2.1 mm × 150 mm, 2.5 μM particle size; Waters 196006724), with a gradient of solvent A (95 vol. % H2O, 5 vol. % acetonitrile, with 20 mM ammonium acetate and 20 mM ammonium hydroxide, pH 9.4) and solvent B (acetonitrile). Flow rate was 0.15 ml min−1 and column temperature was set at 25 °C. The LC gradient was: 0–2 min, 90% B; 3–7 min, 75% B; 8–9 min, 70% B; 10–12 min, 50% B; 12–14 min, 25% B; 16–20.5 min, 0.5% B; 21–25 min, 90% B. The orbitrap resolution was 180,000 at an m/z of 200. The maximum injection time was 200 ms and the automatic gain control target was 1,000%. Raw mass spectrometry data were converted to mzXML format by MSConvert (ProteoWizard). Pick-peaking was done on El Maven (v.0.8.0, Elucidata).

In vitro DHFR E. coli

As described previously26, purified E. coli dihydrofolate reductase (FolA) was purified by Genscript. Enzyme activity was measured on a QuantaMaster 40 spectrophotometer (Photon Technology) using the DHFR reductase assay kit with slight modifications. E. coli FolA was diluted 1:1,000 into 1× assay buffer. Of this mixture, 100 μl with or without compound was added to a BRAND UV cuvette (Sigma-Aldrich BR759200) and sample transmitted light intensity at 340 nm was measured for 100 s at 1 kHz sampling. Readings were averaged for every 1 Hz and the activity of each sample was calculated from the slope (β) of a linear regression of the log-transformed intensity measurements on MATLAB R2022B. To account for enzyme stability, measurements were normalized to a standard condition (60 μM NADPH and 100 μM DHF) measured immediately before the sample of interest. The relative activity was calculated as (βsample − βnoEnzyme)/(βstandard − βnoEnzyme).

Human DHFR in vitro assay

Human purified DHFR was purchased from R&D Systems (8456-DR). DHFR activity was assayed by monitoring the decrease in absorbance by NADPH at 340 nM. DHFR enzyme (0.5 µg ml−1), dihydrofolic acid (100 µM) and different concentrations of methotrexate, fluorofolin or DMSO control were dissolved in 200 µl of Tris buffer (pH 7.5, Tris salt concentration 25 mM). Reaction was initiated by adding NADPH in a 10× stock (1 mM for a final concentration of 100 µM), and absorbance at 340 nM was monitored over time using a Cytation 5 reader (Agilent). Activity was normalized to the DMSO control.

Molecular docking

As the structure of P. aeruginosa DHFR has not been solved, we used AlphaFold30,31 to derive the enzyme’s three-dimensional structure from its sequence (UniProt ID: 6XG5)32. Following the acquisition of the protein structure, we introduced the coenzyme NADPH to the structure, aligning it on the experimentally characterized human DHFR in a complex with IRS-17, providing a structural reference for subsequent steps.

The structures of fluorofolin and trimethoprim were translated from SMILES representation using the RDKit chemoinformatic package. After preparing the enzyme, coenzyme and ligand structures, we defined a cubic grid box of dimensions 20 × 20 × 20 Å centred around the active site of the reference human DHFR–IRS-17 complex. This box serves as the search space for potential binding sites in our docking simulations. We executed the docking simulation using the AutoDock Vina33,34 forcefield, with an exhaustiveness parameter set to 64 to ensure comprehensive sampling of the search space.

Checkerboard assay

Cells were seeded in a similar manner as described above for MIC calculations. Sulfamethoxazole (Chem-Impex 00821) was diluted 1:2 down the rows of the plate, while fluorofolin was diluted 1:2 down the columns of the plate. Fractional inhibitory concentrations (FICs) were calculated as [fluorofolin]/MICFluorofolin + [SMX]/MICSMX, where [fluorofolin] and [SMX] are the concentrations of compounds in a given well, which were divided by the concentration of drug at the MIC for each compound. FIC values less than or equal to 0.5 were considered synergistic.

Growth competition assay

Overnight cultures of P. aeruginosa PA14 or E. coli MG1655 were diluted 1:150 into LB broth in the presence of DMSO or 50 µg ml−1 fluorofolin with or without TMI supplementation and grown for 16 h at 37 °C. Cultures of each species were grown separately as well as being mixed 1:1. Cultures were plated onto LB agar or Pseudomonas Selection Agar (Sigma-Aldrich 17208) and c.f.u. counting was caried out as described above. To control for appropriate Pseudomonas selection, E. coli MG1655 was plated onto Pseudomonas Selection Agar and an absence of colonies was observed. The number of colonies on Pseudomonas Selection Agar plates is reported as the c.f.u. ml−1 of P. aeruginosa. To calculate c.f.u.s ml−1 of E. coli MG1655, c.f.u.s ml−1 were determined from LB agar plates and the c.f.u.s ml−1 of P. aeruginosa were subtracted from these values.

Drug accumulation assay

Overnight PA14 cultures were back-diluted to early–mid exponential phase (OD600 = 0.4–0.6). The assay was initiated with treatment of the culture with either 5.0 µM fluorofolin or 5.0 µM trimethoprim. A DMSO-treated culture was utilized as a control. At time points of 30, 60 and 90 min, a 10 ml aliquot was collected out of the 120 ml parent culture (in triplicate) and pelleted by centrifugation at 2,000 g at 4 °C. The supernatant was then removed and the pellet was washed with ice-cold 0.85% NaCl solution. Following suspension of the cell pellet in 1 ml of 2:2:1 CH3CN:MeOH:H2O, samples were subjected to four cycles of freeze-thaw cell lysis using dry ice in 95% ethanol/ice water. Before each freeze phase, samples were vortexed for 10 s to ensure adequate mixing. Samples were subsequently pelleted at 16,000 g for 5 min, with the supernatant being subjected to filtration using a 0.22 µm SpinX centrifuge tube filter. The resulting cell lysate samples were analysed, utilizing verapamil as an internal standard. For LC–MS analysis, sample components were separated using a Chromolith SpeedRod column, using a gradient of 10–100% CH3CN/H2O acidified with 0.1% (v/v) formic acid, with an Agilent 1260 liquid chromatograph coupled to an Agilent 6120 quadruple mass spectrometer.

Compound accumulation was realized using the selective ion monitoring mode to quantify peak integration for a compound and the internal standard using their respective m/z values. Compound peaks were confirmed using a scanning mode that detected the compound peak using an m/z range of 100–1,000. Peak area integration values were determined and a ratio of the peak area for the compound to the peak area for the verapamil internal standard was calculated, and compound concentration was then determined from the compound calibration curve. The calculated concentration of the compound in each sample was then normalized using the bacterial culture OD600 value. Compound accumulation versus time plots were generated using GraphPad Prism v.9.4.1. Compound accumulation area under the curve (AUC, calculated in Microsoft Excel v.16.65) was determined for each bacterial strain–compound combination and these were compared via statistical analysis (unpaired t-test) in GraphPad Prism v.9.4.1.

Fluorofolin resistance screens

For resistance passaging, P. aeruginosa PA14 was grown overnight at 37 °C in a 96-well plate similarly to MIC assays in duplicate. The wells corresponding to 0.5× MIC was selected and struck out on LB agar plates in the absence of antibiotic to select for stable resistance. Single colonies were picked and inoculated into fresh LB broth. This process was repeated for a total of ten passages. At each passage, the MIC was recalculated and compared to a culture that had not been previously exposed to fluorofolin, which was also grown as a control to confirm antibiotic potency. Cells from each passage were stored as a frozen stock.

To isolate resistant mutants on a plate, 108 c.f.u.s of P. aeruginosa PA14 were plated on LB agar plates containing 4× MIC fluorofolin, ciprofloxacin, meropenem or gentamycin. We plated 109 c.f.u.s on 10 plates containing 4× MIC fluorofolin with 4× the minimal concentration of SMX at which synergy was observed. Plates were grown at 37 °C for 48 h, after which individual colonies were picked and restruck onto fresh plates and grown at 4× MIC of the relevant antibiotic LB broth to confirm resistance. Resistant mutants were maintained as a frozen stock.

To confirm the identity of resistance mutations, whole-genome sequencing was performed and compared to the parental strain of PA14. Briefly, genomic DNA was isolated from a strain of interest using the DNeasy blood and tissue kit (Qiagen 69504). Once DNA was extracted and its quality was confirmed, the DNA was sequenced using an Illumina NextSeq 2000 system. Sequencing and variant calling were performed at SeqCenter.

RNA sequencing

RNA was extracted from overnight cultures of wildtype PA14 or mutant strains. Cultures were pelleted at 4 °C, resuspended in Trizol (Ambion 10296010) and incubated at 65 °C before addition of chloroform. The aqueous layer was collected and RNA was isolated using a mirVana RNA extraction kit (ThermoFisher AM1560). DNAse treatment was performed on RNA using recombinant DNase I (Sigma-Aldrich 04716728001). RNA samples were sent to SeqCenter for sequencing using the Illumina Stranded RNA library preparation with RiboZero Plus rRNA depletion kit.

Pharmacokinetic analysis

Pharmacokinetic properties were determined after a single dose of 200 mg kg−1 fluorofolin given orally (PO). Plasma samples were taken from three mice at times 0.083, 0.25, 0.5, 1, 2, 4, 8 and 24 h, and quantitative analysis was performed using LC–MS/MS. Half-life was determined from plasma concentration after fluorofolin levels reached pseudo-equilibrium (4 h for mouse 1 and 2, and 2 h for mouse 3). Pharmacokinetic values were estimated using a non-compartmental model generated from WinNonlin 6.1. Pharmacokinetic analysis was carried out by Pharmaron.

Serum binding

Fluorofolin (1 μM) was added to a mouse plasma solution or to a buffer-only control. An initial t = 0 sample was collected. Fluorofolin was incubated with plasma for 6 h, after which dialysis was performed. After dialysis, supernatant was collected and the amount of fluorofolin was determined by LC–MS/MS to determine the percentage of unbound fluorofolin. Serum binding parameters were determined as follows:

1.

% Bound = 100% − % Unbound

2.

logK = log(% Bound/100 − % Bound)

3.

% Remaining = Area-ratio6h/Area-ratio0h × 100%

4.

% Recovery was determined as (Area-ratiobuffer-chamber + Area-ratioplasma-chamber)/(Area-ratiototal sample) × 100

Serum binding analysis was performed by Pharmaron.

C. elegans maintenance and toxicity assay

C. elegans N2 worms were maintained on E coli OP50-coated nematode growth medium (NGM) plates before experiments. For P. aeruginosa-coated plates, overnight cultures were diluted to OD600 = 1, spread onto NGM plates, incubated overnight at 37 °C and equilibrated to 25 °C. To synchronize worms for virulence assays, young adult hermaphrodites were bleached to obtain eggs and synchronized L4 worms were collected at 2 days post bleaching. For virulence assays, synchronized L4 worms were transferred to P. aeruginosa plates. Worms were counted at time t = 0, 30, 40, 50 and 60 h to assess viability. Worms were declared dead if they lacked movement after being gently poked with forceps. P values were calculated using a Mantel–Cox test in Prism 9 to compare mutant virulence to wildtype PA14 virulence.

Clinical isolate panels

Clinical isolates46 were inoculated into LB broth in a 96-well flat-bottom plate and grown overnight to stationary phase at 37 °C. The following day, strains were diluted 1:150 into fresh LB broth with fluorofolin or ciprofloxacin at 50 µg ml−1 or vehicle control wells and incubated at 37 °C overnight. Percent growth was calculated by dividing the OD600 of fluorofolin antibiotic-treated wells by that of vehicle-only wells.

Pyocyanin production

Overnight cultures of PA14, PA14 ΔpqsA69 and each mutant were grown in biological triplicate and the OD600 of the cultures were measured. Cell-free supernatants were calculated by centrifugation, and 100 μl were added to a 96-well plate in duplicate. The integrated absorbance spectrum from 306 to 326 nm was taken in a Tecan InfiniteM200 Pro plate reader to determine pyocyanin levels in each sample. A ΔpqsA sample was used to subtract out any background values, and pyocyanin levels were normalized by the ratio of OD600 between wildtype PA14 and the mutants to account for any slight differences in growth.

Maximum tolerated dose

MTD was determined by administration of compounds at increasing dosage until the maximum dose before adverse reactions were observed. Doses were increased in a stepwise manner from 1 mg kg−1 to 5, 10, 25 and 50 mg kg−1. Mice were observed for adverse effects including respiration, piloerection, startle response, skin colour, injection site reactions, hunched posture, ataxia, salivation, lacrimation, diarrhoea, convulsion and death. The maximum tolerated dose of fluorofolin was determined to be 25 mg kg−1 administered subcutaneously. MTD was evaluated by the UNTHSC.

In vivo PA14 infection

Female 5–6-week-old CD-1 mice were rendered neutropenic by intraperitoneal cyclophosphamide treatment (Cytoxan) before this study. On day 0, mice were infected intramuscularly with 5.33 log c.f.u. per thigh PA14. Mice were treated subcutaneously with fluorofolin at 1 and 12 h post infection. For co-treatment, mice were treated with SMX intraperitoneally at 1 and 12 h post infection. Mice were euthanized using CO2 at 24 h post infection, after which thighs were removed and placed into sterile PBS, homogenized and serially diluted onto brain heart infusion (BHI) and charcoal plates for c.f.u. counting. In vivo efficacy was evaluated by the UNTHSC.

Statistical information

For all assays showing error bars, we used the arithmetic mean and standard deviation across multiple biological replicates as our measures of centre and spread. The number of replicates for each experiment type and the type of statistical test used to determine significance are included in respective figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.