Ethics statement

All E. coli isolates used in this study were collected at the RLUH (Liverpool, UK) as part of routine clinical diagnostics procedures. The bacterial isolates were identified in the hospital biobank database as having interesting resistance profiles by the Consultant Microbiologist. Isolates were retrieved from the Microbiology Laboratory by Thomas Edwards (Liverpool School of Tropical Medicine), who has an NHS research passport enabling work in the hospital laboratories. Antimicrobial susceptibility data and the treatment data were anonymised, unlinked to patient identifiers and data produced in this study was not used for the treatment or management of patients, therefore requirement for ethical approval and informed patient consent was not required. This was confirmed using the online NHS Research Ethics Committee review tool http://www.hra-decisiontools.org.uk/ethics/.

Bacterial isolates, media and antibiotics

Clinical isolates of E. coli isolated from blood cultures between 2010 and 2017 at the RLUH (Liverpool, UK) which were found to be carbapenem and cephalosporin susceptible but TZP resistant using the disk diffusion method of antimicrobial susceptibility testing (AST) were initially identified from isolate records. Isolate records were then searched for a corresponding carbapenem, cephalosporin and TZP susceptible isolates, isolated in the same or a previous infection episode from the same patient. Using these criteria, we identified five paired clinical isolates of E. coli. All isolates had been stored at the time of blood culture isolation in glycerol broth at −80 °C.

All isolates were grown on LB (Lennox) agar at 37 °C for 18 h followed by growth in LB, (Sigma, UK), ISO (Oxoid, UK) or M9 (50% (v/v) M9 minimal salts (2x) (Gibco, ThermoFisher Scientific, USA), 0.4% D-glucose, 4 mM magnesium sulphate (both Sigma, UK) and 0.05 mM calcium chloride (Millipore, USA)) at 37 °C for 18 h at 200 rpm.

Piperacillin, tazobactam (both Cayman Chemical, USA), gentamicin (GEN), and amoxicillin trihydrate:potassium clavulanate (4:1, AMC) was solubilised in molecular grade water (all Sigma, UK), while CHL and tetracycline (TET) (both Sigma, UK) were solubilised in ethanol (VWR, USA) and ciprofloxacin (CIP) was solubilised in 0.1 N hydrochloric acid solution (both Sigma, UK). All stock solutions of antibiotics were filter sterilised through a 0.22 µm polyethersulfone filter unit (Millipore, USA). In all assays, unless stated, tazobactam was used at a consistent concentration of 4 µg/ml and the piperacillin concentration was altered.

Restriction enzyme digestion

RFLP analysis of 1 µg of 16S rRNA PCR amplicon from the 10 putative clonal isolates were digested with AlwNI, PpuHI and MslI (all New England Biolabs (NEB), USA) and 1 µg of long fragment genomic DNA extracts of 153964, 152025, 190693 and 169757 were digested with SpeI and MslI (both NEB, USA) for 1 h at 37 °C. Both RFLP digest reactions were incubated for 5 min at 80 °C and immediately run on a 1% agarose gel. Enzyme digest of 500 ng of plasmid DNA extracted from 190693 and 169757 was performed with PpuMI and XhoI (both NEB, USA) and immediately run on a 1% agarose gel following incubation for 1 h at 37 °C. RFLP was performed on a single DNA extract from each bacterial isolate.

Antimicrobial susceptibility testing

Initial AST for cefpodoxime (CPD), cefoxitin (FOX), TZP, meropenem (MEM), CIP, cefotetan (CTT), amikacin (AMK), ertapenem (ETP), AMC, CHL and ampicillin (AMP) was performed in RLUH clinical laboratory using the disk diffusion method, according to either British Society for Antimicrobial Chemotherapy (BSAC) or EUCAST guidelines for Antimicrobial Susceptibility Testing^41,42,43.

MIC for TZP, GEN, CIP, CHL, AMC and TET were performed using the broth microdilution method, in cation adjusted Mueller Hinton Broth (CA-MHB, Sigma, UK), following EUCAST Guidelines⁴⁴. Efflux pump inhibition was performed using PAβN as a supplement in CA-MHB at a final concentration of 50 µM, inhibition of TEM-1B was determined with piperacillin with increasing concentrations of tazobactam (4, 8 and 16 µg/ml) and inhibition of OXA-1 was determined with TZP plus 100 mM sodium chloride (Sigma, UK). Each MIC was performed in triplicate using independent bacterial cultures, each with three technical replicates.

Nitrocefin assay

β-lactam hydrolysis was evaluated using a colorimetric nitrocefin assay. Cell lysates were obtained from triplicate cultures of 190693 and 169757 in LB, adjusted to an optical density at 600 nm (OD₆₀₀) of 0.1. Cultures (10 ml) were centrifuged at 14,000 × g for 5 min, the supernatant discarded, and the pellet resuspended in 5 ml phosphate-buffered saline (PBS). The cultures were sonicated for three intervals of ten seconds, on ice, using a Soniprep 150 plus (MSE centrifuges, UK). The lysed cultures were centrifuged at 14,000 × g for 5 min, and the supernatant taken as the culture lysate.

A total of 90 µl of this lysate was then added to 10 µl of 0.5 mg/ml nitrocefin solution (Sigma, UK) in a 96-well microplate, in triplicate. The absorbance of the plate was read at an optical density of 450 nm (OD₄₅₀) every 20 s for 25 min, using a SPECTROstar OMEGA spectrophotometer (BMG lab systems, Germany). This assay was performed in triplicate using independent bacterial cultures.

Whole-genome sequencing and bioinformatics

Whole-genome sequencing was performed from a single DNA extract for each isolate. Illumina MiSeq 2 × 250 bp short-read sequencing of long fragment DNA extractions from isolates 190693 and 169757, as well as adapter trimming of the sequencing reads, were provided by MicrobesNG (MicrobesNG, UK).

The same long fragment DNA extracts were processed using the SQK-LSK109 ligation and SQK-RBK103 barcoding kit and sequenced on an R9.4.1 flow cell with an ONT MinION. Sequencing reads were basecalled during the sequencing run using MinKNOW (v19.05.0), de-multiplexing and adapter trimming of the basecalled reads were performed using Porechop (v0.2.4) and finally sequencing reads were filtered for a quality score of 10 via Filtlong (v0.2.0).

Both Illumina short-read and ONT long-read sequences were assembled using Unicycler (v0.4.7⁴⁵), with the quality of the assembly assessed using QUAST (v5.0.2⁴⁶), annotated using Prokka (v1.14.0⁴⁷) and visualised using Bandage (v0.8.1⁴⁸).

Sequence type and serotype of both 190693 and 169757 were determined using Multi-Locus Sequence Typing (MLST, v2.0.4⁴⁹) and SerotypeFinder (v2.0.1⁵⁰), respectively. The relatedness of the two genomes were compared using MUMmer (v3.23⁵¹) and the ANI was calculated using OrthoANI (v0.93.1⁵²). Presence of acquired antimicrobial resistance genes within the two genomes were assessed using ResFinder with minimum threshold of 90% and a minimum length of 60% (v3.2⁵³) and segments of the two genomes were characterised using SnapGene® software (v3.3.4, from GSL Biotech; available at snapgene.com). Finally, plasmid replicons were identified using PlasmidFinder (v2.0.1)⁵⁴.

Long-read sequencing reads of 190693 and 169757 were mapped to a predicted structure of three TUs in tandem using BWA MEM⁵⁵ (https://github.com/lh3/bwa, v0.7.17-r1188). Aligned sequencing reads were converted to FASTQ using samtools (ref. ⁵⁶ v1.10) and then used to a build a consensus sequence with Medaka (https://github.com/nanoporetech/medaka, v0.11.5) and compared to the predicted tandem TU structures in EasyFig (ref. ⁵⁷ v2.2.2).

Competent cell preparation

The TZP-susceptible isolate was made competent according to Chung et al.⁵⁸.

Quantitative PCR

Changes in gene copy number of bla_TEM-1B, bla_OXA-1, aac(3)-lla, aac(6’)-lb-cr, tet(D) were calculated via qPCR, using the ΔΔCT method for relative quantitation of these genes against the single copy uidA housekeeping gene.

Each qPCR reaction contained 6.25 µl QuantiTect® SYBR Green PCR buffer (Qiagen, UK), 0.4 µM forward and reverse primers (Supplementary Table 4), 1 ng of extracted DNA, and molecular grade water to a final volume of 12.5 µl. Reactions were processed using a Rotor-Gene Q (Qiagen, Germany), using the following protocol; an initial denaturation step of 95 °C for 5 min, followed by 40 cycles of; DNA denaturation at 95 °C for 10 s, primer annealing at 58 °C for 30 s, and primer extension at 72 °C for 10 s with fluorescence monitored in the FAM channel. High resolution melt analysis was carried out over a temperature range of 75 °C to 90 °C, increased in 0.1 °C increments, in order to confirm specific amplification. Fluorescence thresholds were set manually for calling Ct values, at 5% of the difference between baseline and maximum fluorescence.

The mean qPCR Ct value for the uidA gene from each strain was taken using at least three replicates qPCR reactions, and the ΔΔCT method was utilised to determine to fold change using at least triplicate qPCR reactions for each AMR gene from a single DNA extract of each isolate.

In vitro evolution of susceptible isolate

The clinical isolate 190693 (TZP-susceptible isolate) and 190693 transformed with pHSG396:IS26 were subcultured into 10 ml LB and 10 ml LB plus 35 µg/ml CHL, respectively, and incubated at 37 °C for 18 h at 200 rpm. Following incubation 10 µl of 190693 was subcultured into 10 ml LB and 10 ml LB plus 8/4 µg/ml TZP and 10 µl of 190693 with pHSG396:IS26 was subcultured into 10 ml LB plus 35 µg/ml CHL and 10 ml LB plus 35 µg/ml CHL and 8/4 µg/ml TZP and incubated at 37 °C for 24 h at 200 rpm. Each evolution experiment was performed in triplicate using independent cultures. Genomic DNA from each of the four cultures from each replicate were extracted for qPCR following the protocol described in Supplementary Methods 1, except a single replicate of each of the three biological replicates were performed.

TU capture

The TZP-susceptible isolate 190693 transformed with pHSG396:IS26 was grown in the presence of TZP and CHL as previously stated. Following selection, the culture was serially diluted 1/10 in PBS down to 10⁻⁷ dilution and 50 µl of each dilution was plated out on to LB agar supplemented with 35 µg/ml CHL and 16/4 µg/ml TZP. Five single colonies were selected and subcultured into 10 ml LB plus 35 µg/ml CHL and 16/4 µg/ml TZP for 18 h at 37 °C and 200 rpm and the plasmid extracted following the protocol in Supplementary Methods 1. The purified plasmids were transformed into NEB® 5-alpha competent E. coli (NEB, US) following the protocol in Supplementary Methods 1 and plated out on to LB agar supplemented with 35 µg/ml CHL and 16/4 µg/ml TZP and incubated at 37 °C for 18 h. A single colony from each transformation was subcultured into 10 ml LB supplemented with 35 µg/ml CHL and 16/4 µg/ml TZP and incubated at 37 °C, 200 rpm for 18 h and the plasmid extracted following the protocol in Supplementary Methods 1. The initial pHSG396:IS26 plasmid extract and pHSG396: IS26 plasmid selected in TZP and extracted from NEB® 5-alpha E. coli were digested with XhoI (NEB, US) and EcoRI for 1 h at 37 °C, followed by a 20 min incubation at 65 °C and run on a 1% agarose gel. A single DNA extract of pHSG396:IS26 selected in TZP was sequenced on a MinION, adapters removed, and sequencing reads filtered as stated above, except using a quality score of 50. Long-read sequencing reads of the captured TU in pHSG396:IS26 were mapped to three predicted plasmid structures, pHSG396:IS26, pHSG396:IS26 plus TU and tandem pHSG396:IS26 plus TU and used to build a consensus sequence using the method described above.

Comparative fitness

The relative fitness of 169757 (TZP-resistant) and 190693 (TZP-susceptible) grown in the presence of 8/4 µg/ml TZP, compared to 190693 and 190693 grown in the absence of TZP, respectively, were assessed comparatively in LB, ISO and M9. Each culture was diluted to an OD₆₀₀ of 0.1 in the respective media, then further diluted 1/1000 in the same media and 150 µl of each diluted culture added to a flat bottom, 96-well microtitre plate in duplicate as well as 150 µl of the media as a negative control. The 96-well plate was incubated at 37 °C and the OD₆₀₀ of each well was measured with 100 flashes every 10 min over 24 h, with orbital shaking at 200 rpm between readings, using a Clariostar Plus microplate reader (BMG Labtech, Germany). The relative fitness compared to either 190693 or 190693 grown in the absence of TZP between absorbance values 0.02 and 0.08 and a minimum R value of 0.9905 was estimated using BAT version 2.1⁵⁹. Comparative fitness was performed in triplicate with independent bacterial cultures, with two technical replicates for each replicate.

Statistical analysis

Statistical analysis of comparison for the qPCR assay of the antibiotic resistance genes was performed using the two-way ANOVA with Uncorrected Fisher LSD test. Statistical analysis of the nitrocefin assay was performed using the two-way ANOVA test. Statistical analysis of relative fitness of 169757 and 190693 grown in the presence of TZP was performed using ordinary one-way ANOVA with Uncorrected Fisher LSD test. All statistical tests were performed using GraphPad Prism version 8.2.1. Means and standard error of the mean of all data presented in this study are available in Supplementary Table 3.

Reporting summary

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