Ethics statement

This study was approved by the Sapporo Medical University Hospital Institutional Review Board (IRB no. 272-70) and Sapporo Medical University Animal Care and Use Committee (nos. 17-137, 18-083, and 20-006).

Clinical epidemiology

Clinical epidemiology analysis was performed using 695 Kp infections reported from 2017 to 2022 at Sapporo Medical University Hospital, including 393 Colonisation and 302 Infection cases. The Infection cases classified according to the presence or absence of immunosuppression, site of infection, and presence or absence of bacteraemia. The contingency tables were analysed by Fisher’s exact test. A p-value < 0.05 was considered to indicate statistical significance.

Bacterial isolation, antimicrobial susceptibility testing, and string test

A total of 277 Kp strains were isolated from clinical specimens derived from hospitalised patients at Sapporo Medical University Hospital between 2017 and 2021. These clinical specimens included 100 urine samples, 113 respiratory samples, 12 blood samples, and 52 other samples (drainage, tongue coating, skin, vaginal lubricant, pus, and bile). Identification of Kp (K. pneumoniae subsp.) was performed using MALDI Biotyper (Bruker Corporation, Billerica, MA, USA). BIDMC_1, a carbapenem-resistant Kp strain isolated at the Beth Israel Deaconess Medical Centre (BIDMC), was provided by BEI Resources (NIAID, NIH, USA).

The antimicrobial susceptibility of the Kp strains was tested via the broth microdilution method, and the results were interpreted according to the Clinical and Laboratory Standards Institute (CLSI) recommendations⁵⁰. In this study, the following antimicrobial agents were used: ciprofloxacin (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan), ciprofloxacin hydrochloride monohydrate (Tokyo Chemical Industry, Tokyo, Japan), amikacin (FUJIFILM Wako Pure Chemical Corporation), kanamycin (FUJIFILM Wako Pure Chemical Corporation), and meropenem (FUJIFILM Wako Pure Chemical Corporation).

HMV strains were defined as giving a positive string test result, as previously described⁵¹. A single colony grown overnight on Mueller-Hinton II (MHII) agar was collected, and the formation of a string > 5 mm in length was defined as a positive result. For the detection of hypervirulence factors (serotypes K1 and K2, rmpA, rmpA2, iutA, iroN, and the IncHIB plasmid), multiplex PCR was performed as previously described¹⁸.

Serum susceptibility

In this study, we used commercially available human serum from individual healthy donors (Cedarlane Laboratories Ltd, Burlington, Canada). The serum MIC was defined as the minimum serum concentration (%) that prevented the visible microorganism growth. We set the resistance breakpoint at 32%, and isolates whose serum MIC was higher than 48% of serum MIC were defined as serum-resistant isolates because this concentration is the serum composition of human whole blood (around 40%). For the measurement of serum MICs, the concentration was increased twofold from 0.5 to 32 mg/L. From 32 to 64 mg/L, the increase was set at increments of 16 mg/L to better understand serum resistance levels and to avoid misjudging strains with serum MICs ranging from 40 to 64 mg/L as being judged as having a serum MIC of 32 mg/L, which would incorrectly classify them as serum-susceptible. For the measurement of serum MICs in the serial passaging experiment, the concentration was increased in more detail, in increments of 4%, from 4% to 72%, due to the monitoring of bacterial evolution.

Kp strains were grown in 0.5 ml of tryptic soy broth (TSB) from an overnight culture. The strains were diluted 10⁻⁴-fold (10⁵ CFU/ml) and incubated in plates with different concentrations of serum in each well. After 20 h, the bacterial growth was visually confirmed in the wells with high serum concentrations due to the high optical density (OD600 nm).

Determination of mutation frequency

The mutation frequency was determined via a rifampicin assay⁵². The Kp isolates were cultured overnight in TSB. The solution was concentrated 10-fold and plated onto MHII agar plates supplemented or not with 100 mg/L rifampicin, and the plates were subsequently cultured at 37 °C for 24 h. After cultivation, the colony-forming units (CFUs) that grew on the agar plates were counted. Gene mutation frequency was calculated as [CFUs on the rifampicin-supplemented MHII agar plate]/[CFUs on the unsupplemented MHII agar plate]. We defined the mutator types as hyper (> 10⁻⁷), high (from 10⁻⁸ to 10⁻⁷), moderate (from 5 × 10⁻⁸ to 10⁻⁸), and low (< 5 × 10⁻⁸). Student’s t test was used for statistical analysis. A p-value < 0.05 was considered to indicate statistical significance.

Serial passaging experiments

Serial passaging experiments were performed by incubating Kp isolates (SMKP838, SMKP590, and BIDMC) in 96-well plates with MHBII containing various concentrations of human serum (the concentration was increased in increments of 4%, from 4% to 72%) or antimicrobial agents (ciprofloxacin, amikacin, and meropenem), as previously described⁵³. For the experiments using the other Kp clinical isolates, we selected 19 serum-susceptible Kp isolates (with serum MICs ranging from 8 to 16%) from among the hyper- (n = 1), high- (n = 9; containing one K. quasipneumoniae), and low-mutators strains (n = 9) from the serial passaging experiment in the presence of human serum. We selected the well with the highest concentration (sub-MIC) of human serum or antimicrobial agent in which the bacteria grew and diluted the bacterial culture 100-fold with 0.85% NaCl. Then, 1 µL of the diluted solution was inoculated in 96-well plates containing 100 µL of MHBII with various concentrations of human serum or antimicrobial agents for cultivation at 37 °C for 24 h. Serial passaging was repeated for 20 days in triplicate.

Time-killing assay

Single colonies of SMKP838 and the mutS mutant strains were grown overnight in TSB media. The culture mixtures were adjusted to a final concentration of 1 × 10⁵ CFU/ml and incubated for 0–24 h with each serum-containing solution (1/4 × MIC, 1 × MIC, or 2 × MIC) or in solution without serum at 37 °C without shaking. The assay results were determined at 0 min, 30 min, 1, 3, 6, and 24 h.

WGS

Genomic DNA was isolated with a DNeasy Blood & Tissue Kit (Qiagen, Hulsterweg, The Netherlands). The DNA library was prepared with a Nextera XT DNA Library Preparation Kit (Illumina, San Diego, CA, USA) to sequence 300 bp paired-end reads according to the manufacturer’s protocol. An Illumina MiSeq was used for WGS. The CTX-M genes were identified by Resfinder (https://www.genomicepidemiology.org) using the assembled genome data. MLST was performed using the Institute Pasteur MLST database and software (https://bigsdb.pasteur.fr/klebsiella/). Fast average nucleotide identification (FastANI) against the type strain genome was utilised for species identification. Core genome single nucleotide polymorphism (SNP)-based phylogenetic analysis was conducted, using the Kp ATCC 35657 genome (accession number: CP015134.1) as a reference for mapping. Mapping and core genome extraction were performed using BWA version 0.7.17 with the bwasw option, SAMtools version 1.6 with the mpileup option, and VERSCAN version 2.3.9 with the mpileupcns option. Estimated homologous recombination regions were excluded using ClonalFrameML version v1.11-2. Snp-dists was used to determine the pairwise SNP distance. A phylogenetic tree was generated using FastTree version 2.1.11 and FigTree version 1.4.4 (http://tree.bio.ed.ac.uk/software/figtree/). The number of gene mutations accumulated during serial passaging experiments was analysed by mapping the genome reads to the reference genome (wild-type strain on day 0) obtained via WGS, followed by basic variant detection using CLC Genomics Workbench 21 (QIAGEN).

Determination of bacterial growth

Bacterial growth was monitored by measuring the turbidity (OD600) using an Infinite M200 PRO multimode microplate reader (Tecan, Kawasaki, Japan). Strains were grown in 0.5 ml of TSB (Becton Dickinson, Franklin Lakes, NJ, USA) overnight at 37 °C, and 1 × 10⁵ CFU/ml bacteria were cultured in 0.1 ml of MHBII broth (Becton Dickinson) in a 96-well plate at 37 °C with shaking at 140 rpm for 16 h. Bacterial growth curves were generated from the measurements taken every 10 min for 16 h. Bacterial growth of the mutants derived after 20 days of the serial passaging experiment in the presence of human serum was also evaluated by counting the viable bacterial number as CFU at 37 °C with shaking at 140 rpm for 0, 1, 3, 6, and 16 h. Each 1 ml of the 10-fold serial dilution with 0.85% NaCl was plated on Easy Plate AC (Kikkoman Corp., Chiba, Japan), and CFUs were counted after 48 h of cultivation at 37 °C.

TraDIS

The SMKP838 transposon library was constructed using the EZ-Tn5™ Tnp Transposome™ Kit (Epicentre, Madison, WI, USA). Bacteria with transposase introduced by electroporation (2.5 kV/cm, 200 Ω, and 25 μF) were selected on the basis of the formation of colonies on MHII agar containing 50 mg/L kanamycin. Over 100,000 colonies were collected, pooled, and frozen at − 80 °C in TSB with 10% glycerol as stock solutions until use. The transposon mutant library (10⁶ CFUs/ml) was inoculated into 1 ml of plain MHBII, MHBII containing 4% or 8% serum, or MHBII containing 40 mg/L SPA and cultured at 37 °C for 20 h. Total DNA was isolated using the Wizard Genomic DNA Purification Kit (Promega, Madison, WI, USA). Total DNA (500 ng) was used to prepare the DNA library for TraDIS using the NEBNext Ultra II FS DNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA). After fragmentation, end repair, 5’ phosphorylation, dA-tailing, adaptor ligation, and size selection (275-475 bp) according to the manufacturer’s protocol, the transposon-inserted genes were amplified via PCR using NEBNext Ultra II Q5 Master Mix (New England Biolabs), 20 nM primers [NEBTnF2fas (5’-TCGACCTGCAGGCATGCAAGCTTCAGGGTTGAGATGTG-3’) and NEBTn5-700 (5’-GTGACTGGAGTTCAGACGTGTGCTCTTCCGATC-3’)], and 20 ng of fragmented DNA as the template under the following conditions: initial denaturation at 98 °C for 30 sec, 22 cycles of 98 °C for 10 sec and 72 °C for 1 min and 15 sec, and a final extension at 72 °C for 2 min. After PCR product was purified using AMPure XP beads (Beckman Coulter, Brea, CA, USA), enrichment PCR was performed using the KAPA HiFi HotStart Library Amplification Kit (Roche, Basel, Switzerland), 20 nM NEBNext i700 primers [including NEBNext Multiplex Oligos for Illumina (New England Biolabs) and NEBTn5-501-3 (5’- AATGATACGGCGACCACCGAGATCTACACTATAGCCTACACTCTTTCCCTACACGACGCTCTTCCGATCTTCGACCTGCAGGCATGCAAGCTTC-3’)], and 20 ng of the purified DNA as the template under the following conditions: initial denaturation at 98 °C for 45 sec, 10 cycles of 98 °C for 15 sec, 60 °C for 30 sec, and 72 °C for 10 sec, and a final extension at 72 °C for 30 sec. The PCR products were purified and selected on the basis of size (average: 650 bp) using AMPure XP beads. These products were pooled, and a NovaSeq600 was used for TraDIS. TraDIS analysis was performed according to a previous study³⁴, and a false discovery rate-adjusted p-value (FDRp) < 0.05 (vs. unsupplemented medium) was considered significant.

Genes with significantly lower detected levels in serum or SPA samples (> 2-fold vs. unsupplemented medium) were considered putative serum or SPA resistance genes.

Construction of the gene deletion mutants

The mutS-deletion SMKP838 mutant and each putative serum resistance-associated gene were generated via the λ-Red recombinase system, as previously described, using pKD46-hyg^54,55. Each gene was replaced with Mini genes containing kanamycin resistance cassettes (Gene Bridges, Heidelberg, Germany) and 50 bases corresponding to the upstream and downstream regions of the target genes. Gene deletion was confirmed via PCR using the specific primers listed in Supplemental Data Dataset 7.

Construction of the isogenic non-HM-Kp mutants

The isogenic SMKP838 strains that developed gene mutation(s) during the serial passaging experiment in serum were constructed by pORTMAGE⁵⁶. Hygromycin-integrated pORTMAGE (pOSTMAGE-hyg) was generated as previously described²¹. The pooled 90 bp oligonucleotides of putative serum- and antimicrobial-resistance gene mutations (listed in Supplemental Data Dataset 7) were electroporated into the Kp SMKP838 strain harbouring pORTMAGE-hyg, and approximately 90 clones were selected after 4-6 pORTMAGE cycles. The serum- and antimicrobial-susceptibilities of the clones were evaluated by determining the MICs as described in the Serum susceptibility and antimicrobial susceptibility testing sections.

Prevalence of serum resistance gene mutations in public databases

To investigate the frequency of gene depletion or mutation in clinical isolates of Kp, we downloaded 3447 genome sequences of Kp isolates from the BV-BRC database (https://www.bv-brc.org/). The genome sequences retrieved for use in this study were derived from human isolates between 2019 and 2023 and registered in the database by the end of 2023. We also excluded strains with abnormal genome sizes, high contamination values or low completeness using checkM. The genome sequences were translated into amino acid sequences using getorf included in the EMBOSS package for all open reading frames (ORFs) longer than 50 amino acids. We subsequently used BLASTp in a local environment to search for the amino acid sequences of the nine proteins of the clinical strain SMKP838. Strains possessing an ORFs with sequence lengths matching each query protein were determined to be conserved. The protein sequences that matched those of the SMKP838 strain were classified as identical, whereas the proteins with at least one amino acid mutation were labelled as mutated. In the initial BLASTp searches, strains that did not possess an ORF with an appropriate length were further analysed by examining the ORF with the highest bitscore to identify sequence alterations due to nonsense or indel mutations. These were classified as not conserved. Strains for which no matching results were found were labelled not-found.

Mouse models of lung and bloodstream infection

Ten- to 12-week-old female BALB/c mice were anaesthetised and infected transbronchially with 50 μl of a 1 × 10⁸ CFU/ml solution using a microsprayer (TORAY PRECISION, Tokyo, Japan). The mice were immunosuppressed by administrating cyclophosphamide monohydrate (lot no. SKE6784; FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) via intraperitoneal injection as follows: 250 mg/kg five days prior to infection and 125 mg/kg one day prior to infection as a previous study²⁸. In the treatment group, the mice were injected subcutaneously with 100 mg/kg ciprofloxacin monohydrochloride (Tokyo Chemical Industry, Tokyo, Japan) 1 and 24 h after infection. After 48 h (or 32 h when the prelethal critical endpoint had been reached) of infection, the mice were euthanized by cervical dislocation. The lungs were washed with sterile phosphate-buffered saline (PBS) and homogenised with a gentleMACS dissociator (Miltenyi Biotec, Bergisch Gladbach, Germany). The homogenates were plated to determine the number of CFUs per lung. Blood was collected by puncturing the jugular vein. One drop of blood was added to 1 ml of PBS, vortexed, and then cultured on an appropriate plate. In the experiments, wild-type-derived strains were selected by seeding on MHII agar supplemented with 100 mg/L ampicillin sodium, and the ΔmutS-derived strains were selected on MHII agar supplemented with 50 mg/L kanamycin to eliminate the effects of other indigenous and environmental bacteria. Fifty colonies from each sample were randomly selected, and their serum MICs were determined.

Evaluation of the extent of Klebsiella infiltration

H&E-stained slides were examined under a microscope to evaluate the extent of Klebsiella infiltration in the following tissues: alveoli, capillaries, and interstitium of the lungs, glomeruli, capillaries, and interstitium of the kidneys; and the Gleason sheath, sinusoids, and central veins of the liver. The degree of infiltration was classified as follows: none, mild (small amounts in 1-2 locations), moderate (less confluent than severe), and severe (large amounts in multiple locations). A pathologist (AT) and a trained researcher blinded to the conditions evaluated the whole slides. Conflicting results were discussed, and a consensus was reached.

Construction of mutS-mutated BIDMC_1

Since the λ-Red recombination system was unable to generate mutS-deficient strains of BIDMC1, we used pORTMAGE and constructed mutS-mutated BIDMC1 strain (BIDMC1 MutS_Tyr37STOP)²¹. The transformants were produced via electroporation. Oligonucleotides (90 bp) for mutS containing the C111T mutation were designed using the MEGA Oligo Design Tool: MegamutS,CGCAACATCCTGACATTCTGCTGTTTTACCGGATGGGGGATTTTTAaGAGCTATTTTATGACGATGCGAAACGCGCCTCGCAGCTGCTCG; where the bold letter a indicates an introduced base. Gene replacement was confirmed by direct DNA sequencing.

Statistical analysis

All the data were analysed by using GraphPad Prism 9 (GraphPad Software, Inc.; San Diego, CA, USA), R version 4.4.0 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria, 2024, https://www.R-project.org) and JMP Pro 17.0.0 (SAS Institute Inc., Cary, NC, USA). One-way ANOVA, An unpaired, two-tailed Student’s t test or a two-tailed Mann‒Whitney U test was used to compare two groups, and Dunn’s comparison test, followed by the Kruskal–Wallis test, was used to compare three or more groups. In addition, the log-rank test was used for survival data analysis. Statistical methods and p values are given in each figure legend. A p-value < 0.05 was considered to indicate statistical significance. In addition, Fisher’s exact test and Student’s t test were used to compare the two groups in the clinical analysis. Multivariate logistic regression analysis was performed to assess factors associated with bacteraemia, immunosuppression and death within 60 days, and age, sex, body mass index (BMI), haemoglobin, albumin, smoking status, and overall length of hospital stay were included as covariates. Individual models of clinical outcomes were used to account for collinearity.

Reporting summary

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