Epidemiological characterization of chromosome-mediated colistin resistance in hypervirulent carbapenem-resistant Klebsiella pneumoniae Authors: Yanghua Xiao, Jingwen Zhang, Feng Nie, Keyi Li, Xingyu Tao, Tianxin Xiang, Ping Li Published: 2025 ABSTRACT BackgroundHypervirulent and carbapenem-resistant Klebsiella pneumoniae (hv-CRKP) poses a major clinical threat. Colistin is a last-resort agent against hv-CRKP, yet resistance driven by chromosomal mutations remains poorly characterized in its prevalence and functional impact.MethodsFrom 2020 to 2023, 239 non-duplicate clinical hv-CRKP isolates were collected from a tertiary hospital in China. Isolates were analyzed by multilocus sequence typing, antimicrobial susceptibility testing, conventional PCR screening, and targeted sequencing of mcr, pmrAB, phoPQ, crrAB, and mgrB. RT-qPCR assessed lipid A modification gene expression. Whole-genome sequencing was performed on colistin-resistant isolates. Functional roles of novel pmrB mutations were validated using CRISPR-Cas9 genome editing. Growth kinetics, competition assays, Galleria mellonella infection models, and transcriptomic profiling were conducted to assess fitness, virulence, and gene expression changes.ResultsMost isolates (210/239, 87.87%) belonged to ST11 lineage, and no mcr genes were detected. Mutations in pmrB were highly prevalent (232/239, 97.07%), with R256G representing a lineage-associated polymorphism, while specific variants (P95L, D150H, T157P, S203P) associated with elevated colistin MICs. mgrB alterations were identified in 6.28% (15/239) of isolates and were exclusively observed in colistin-resistant strains, whereas isolates carrying crrB variants (17/239, 7.11%) remained susceptible to colistin. Mutations in the HAMP and HisKA domains of pmrB were associated with increased colistin MICs and upregulation of lipid A modification genes. Functional validation using CRISPR-Cas9 confirmed that two novel pmrB variants (D150H and S203P) conferred resistance in ST11-KL25 isolates, as reversion restored colistin susceptibility and impaired bacterial growth under colistin pressure. Transcriptomic analysis demonstrated that reversion suppressed lipid A modification pathways (arnA-D, eptA). Despite fitness advantages under colistin pressure, virulence assays in the G. mellonella infection model showed that pmrB mutants retained full pathogenic potential.ConclusionsChromosomal pmrB mutations and disruptive mgrB alterations are the predominant mechanisms of colistin resistance in clinical hv-CRKP. The emergence of colistin-resistant ST11-KL64/KL25 hv-CRKP highlights the urgent need for ongoing genomic surveillance and the development of effective therapeutic strategies to mitigate their growing public health threat.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12866-025-04615-y. Background Klebsiella pneumoniae is a Gram-negative opportunistic pathogen and a leading cause of nosocomial infections globally [1]. In recent years, the emergence of hypervirulent and carbapenem-resistant K. pneumoniae (hv-CRKP) has become a growing public health concern [2]. These strains commonly carry virulence determinants such as rmpA, rmpA2, and the iucABCD-iutA operon, which are implicated in the regulation of capsule production and siderophore-mediated iron uptake [3, 4]. In China, the sequence type (ST) 11 lineage has become the predominant clone among hv-CRKP isolates, associated with high mortality, substantial clinical burden, and limited treatment options due to extensive antimicrobial resistance [5, 6]. With the continued erosion of effective antibiotics, polymyxins, notably colistin (polymyxin E), have been reintroduced as last-resort therapeutics for infections caused by extensively drug-resistant Gram-negative pathogens [7, 8]. Colistin exerts its bactericidal activity primarily through electrostatic interaction with the lipid A moiety of lipopolysaccharide within the bacterial outer membrane, disrupting membrane integrity and leading to cell death [9]. However, the increasing prevalence of colistin-resistant isolates of Enterobacteriaceae critically undermines the efficacy of this essential antimicrobial agent [10, 11]. The mechanisms of colistin resistance involve structural modifications to lipid A that diminish its net negative charge, thereby reducing colistin binding affinity [12]. These modifications are conferred either by mobile colistin resistance (mcr) genes or by chromosomally encoded mutations [12]. The predominant chromosomal mechanisms of colistin resistance in K. pneumoniae involve mutations affecting regulatory pathways that control lipid A modification [13]. Mutations in pmrB, which encodes the sensor kinase of the PmrA/PmrB two-component system, can result in constitutive activation of this pathway, leading to upregulation of the arn operon and eptA gene, and subsequent lipid A modification with 4-amino-4-deoxy-L-arabinose (L-Ara4N) and phosphoethanolamine (pEtN), thereby reducing colistin binding affinity [14, 15]. Insertion sequence (IS)-mediated disruption of mgrB, a key negative regulator of the PhoQ/PhoP two-component system, is also common [16, 17]. Loss of mgrB function constitutively activates PhoQ/PhoP, promoting similar lipid A modifications (primarily pEtN addition) and resulting in increased colistin resistance [18]. Traditionally, the acquisition of antimicrobial resistance has been thought to incur a fitness cost, reducing bacterial adaptability [19, 20]. However, the widespread dissemination of ST11 K. pneumoniae, which has successfully integrated carbapenem resistance and hypervirulence while retaining high transmissibility, indicates that resistance and virulence can in some cases coexist without obvious fitness compromise [21]. In this study, we investigated the molecular epidemiology and functional impact of chromosomal mutations conferring colistin resistance in clinical hv-CRKP isolates, with a particular focus on the epidemic ST11 lineage. We characterized the prevalence and distribution of mutations in key regulatory systems (PmrAB, PhoPQ, CrrAB) and the negative regulator MgrB, and analyzed their associations with colistin susceptibility. Furthermore, we functionally validated novel pmrB variants, assessed their effects on bacterial fitness, competitive dynamics, virulence, and transcriptomic profiles. Materials and methods Isolate collection and identification Between 2020 and 2023, a total of 1755 non-duplicate K. pneumoniae isolates were collected from clinical specimens at the Department of Clinical Laboratory, First Affiliated Hospital of Nanchang University (Jiangxi, China). Antimicrobial susceptibility testing for imipenem and meropenem was initially performed using the Kirby-Bauer disk diffusion method. Isolates were classified as CRKP if the inhibition zone diameter for either antibiotic was ≤ 19 mm, according to CLSI guidelines (M100-34) [22]. Based on this criterion, 472 isolates were identified as CRKP. To determine hypervirulent phenotypes, CRKP isolates were screened by PCR for established virulence-associated genes (rmpA, rmpA2, iucA, iroB, and peg-344) using published primers [23, 24]. The hypermucoviscous phenotype was assessed using the string test, with a positive result defined as the formation of a viscous filament > 5 mm when a bacterial colony was stretched with a sterile inoculating loop. Isolates that carried at least one virulence-associated gene and tested positive in the string test were classified as hypervirulent CRKP (hv-CRKP). By these criteria, 239 hv-CRKP isolates were identified and included for subsequent analyses. Multilocus Sequence Typing (MLST) and mutation detection Genomic DNA was extracted from cultures of 239 hv-CRKP isolates grown in LB at 37 °C with shaking at 220 rpm for 12 h using the Bacterial Genomic DNA Extraction Kit (Tiangen Biotech, Beijing, China) according to the manufacturer’s instructions. Multilocus sequence typing (MLST) was performed by PCR amplification of seven housekeeping genes (gapA, infB, mdh, pgi, phoE, rpoB, and tonB) using primers as previously described [25]. Amplified products were sequenced, and sequence types (STs) were determined using the MLST database (https://pubmlst.org/organisms/klebsiella-pneumoniae). To investigate colistin resistance determinants, PCR screening was conducted for mcr-1 to mcr-9 genes using established protocols on both genomic and plasmid DNA, with mcr-positive Enterobacteriaceae as controls [26]. Chromosomal genes mgrB, pmrA, pmrB, phoP, phoQ, ccrA, and ccrB were amplified and sequenced, with primer sequences listed in Supplementary Table S1. Nucleotide and translated amino acid sequences were aligned against the reference strain K. pneumoniae subsp. pneumoniae MGH 78,578 (GenBank accession no. CP000647.1) to identify mutations. Antimicrobial susceptibility testing Colistin MICs were assessed by broth microdilution in cation‑adjusted Mueller-Hinton broth (CAMHB). Cultures were grown for 12 h at 37 °C with shaking at 220 rpm, adjusted to 0.5 McFarland using a densitometer (bioMérieux DensiCHEK Plus), diluted 1:100 to ~ 1 × 106 CFU/mL, and inoculated (100 µL) into 96‑well plates containing two‑fold serial dilutions of colistin sulfate (0.25–128 µg/mL). Plates were incubated aerobically at 35 °C for 16–20 h, with MICs defined as the lowest concentration preventing visible growth. Interpretive criteria followed EUCAST breakpoints for K. pneumoniae: susceptible (S) ≤ 2 mg/L, resistant (R) > 2 mg/L [27]. Quality control was ensured using Escherichia coli ATCC 25,922 and K. pneumoniae NTUH-K2044 within each test, and a clinically collected mcr-1-positive colistin-resistant K. pneumoniae isolate as the positive control. Real-time fluorescence quantitative PCR K. pneumoniae isolates were cultured in CAMHB at 37 °C with shaking at 220 rpm until reaching the exponential growth phase (approximately 4 h). Total RNA was extracted using the Total RNA Purification Kit (Sangon Biotech, Shanghai, China) according to the manufacturer’s instructions. RNA samples were treated with a gDNA eraser included in the PrimeScript RT reagent kit (Takara Bio, Japan), followed by reverse transcription to synthesize complementary DNA (cDNA) using the same kit. Real-time quantitative PCR (RT-qPCR) was carried out with TB Green Premix Ex Taq II (Takara, Japan) on the QuantStudio™ 5 system, with data collected and analyzed using QuantStudio™ Real-Time PCR Software. Gene expression levels were normalized to the housekeeping gene rpoB. Relative quantification of target gene expression was calculated using the 2−ΔΔCt method [27]. Each reaction was performed in triplicate to ensure reproducibility. Primer sequences used for RT-qPCR are listed in Supplementary Table S1. Whole-genome sequencing and bioinformatic analysis Genomic DNA was extracted from K. pneumoniae isolates harboring chromosomal mutations using the Bacterial Genomic DNA Extraction Kit (Tiangen Biotech, Beijing, China). Whole-genome sequencing was performed on the Illumina NovaSeq 6000 platform (Illumina, USA). Adapter sequences and low-quality bases were removed using Trimmomatic v0.39 [28]. Assembly was performed with SPAdes v4.0.0 under default parameters [29]. Assembly quality was assessed with QUAST v5.2.0 [30], and contaminated assemblies were excluded. Capsular serotypes, antimicrobial resistance determinants, and virulence factors were identified using Kleborate v3 [31]. For phylogenetic analysis, single-nucleotide polymorphism (SNP) calling was performed using Snippy v4.6.0 (https://github.com/tseemann/snippy). A maximum-likelihood tree was constructed using IQ-TREE and visualized in iTOL [32, 33]. Mutations in colistin resistance-associated genes (pmrAB, phoPQ, crrAB, mgrB) were validated by aligning assembled contigs against the colistin-susceptible K. pneumoniae MGH 78,578 reference genome (GenBank: CP000647.1) using Snippy v4.6.0. Growth curve analysis Overnight cultures of K. pneumoniae were adjusted to 0.5 McFarland (~ 1.5 × 108 CFU/mL) and diluted 1:100 in CAMHB with or without colistin. Controls included vehicle-treated and sterile medium blanks. Aliquots (200 µL) were dispensed into 100‑well honeycomb microplates and incubated at 37 °C with continuous shaking (220 rpm) in a Bioscreen C Pro system (Growth Curves USA). Optical density at 600 nm was recorded hourly for 24 h. Each condition was tested in triplicate, and experiments were repeated independently three times. Construction of pmrB site-directed mutant Site-directed mutagenesis of the pmrB gene was performed in hv-CRKP strains KP7145 and KP11396, both of which exhibited a colistin MIC of 32 mg/L. A CRISPR-Cas9 genome editing system, utilizing the pCasKP-apr and pSGKP-spe plasmids, was employed as previously described to introduce specific point mutations (H150D and P203S) into the pmrB gene via a two-step homologous recombination strategy [34]. In the first step, linear donor DNA and a spacer-containing pSGKP-spe plasmid were co-transformed into pCasKP-apr-containing cells that had been induced with L-arabinose. This process facilitated the deletion of approximately 300 bp of the pmrB locus surrounding the target mutation site. Transformants with partial gene deletions were selected on LB agar plates containing 50 mg/L apramycin and 5% sucrose, followed by incubation at 30 °C to promote plasmid curing. In the second step, the wild-type pmrB sequence was amplified and reintroduced into the chromosomal locus using pSGKP with a spacer targeting the flanking regions of the deleted segment. After the sequential curing of both pCasKP and pSGKP plasmids, individual colonies were screened by colony PCR. The presence of the desired point mutations was confirmed by Sanger sequencing. Colistin MICs for the resulting mutant strains were determined by broth microdilution. All primers used in this study are listed in Supplementary Table S1. Time-kill assay Time-kill assays were conducted to assess the bactericidal activity of colistin against wild-type and pmrB reversion strains of K. pneumoniae [35]. Log-phase cultures grown in CAMHB at 37 °C with shaking at 220 rpm for approximately 4 h were adjusted to ~ 1 × 106 CFU/mL in 10 mL CAMHB containing 4 mg/L colistin. Cultures were incubated at 37 °C with agitation, and 100 µL samples were collected at 0, 4, 8, 12, and 24 h, serially diluted in saline, and plated onto antibiotic-free MHA. Colonies were counted after overnight incubation at 37 °C to determine viable CFU over time. Each condition was tested in triplicate. Competition assay Competition experiments were conducted to assess the relative fitness of the wild-type and pmrB reversion strains. Log-phase cultures of both strains were obtained by incubating in CAMHB at 37 °C with shaking at 220 rpm for approximately 4 h, adjusted to the same density (1 × 106 CFU/mL) in CAMHB, mixed in equal volumes, and incubated at 37 °C for 24 h. At 0 and 24 h, 100 µL samples of the mixed culture were serially diluted and plated onto two types of MH agar plates: one without antibiotic and the other supplemented with 16 mg/L colistin. Prior validation confirmed that the reversion strain is unable to grow on 16 mg/L colistin-containing plates, whereas the wild-type strain is not inhibited. Plates were incubated overnight at 37 °C, and colonies were counted. Relative fitness (RF) was calculated using the following formula [36]: RF = log10(N24/N0) / log10 (M24/M0), where N and M represent the CFUs of the reversion and wild-type strains, respectively, at 0 h and 24 h. All experiments were performed in biological triplicates. Galleria mellonella survival assay Virulence was assessed using a G. mellonella infection model. Strains were grown in LB broth at 37 °C with shaking at 220 rpm for 12 h, harvested, washed, and resuspended in sterile saline. Suspensions were adjusted to 1 × 106 CFU/mL (confirmed turbidimetrically), and 10 µL (1 × 104 CFU) was injected into the last left proleg of each larva using a Hamilton syringe. Larvae were incubated at 37 °C, and survival was recorded every 12 h. The hypervirulent reference strain NTUH‑K2044 and sterile saline solution served as positive and negative controls, respectively. Death was defined as the lack of response to gentle touch. Each group contained ≥ 10 larvae, and experiments were repeated in triplicate with independent biological replicates. RNA-Seq and data analysis Wild-type and pmrB reversion strains of K. pneumoniae were cultured in CAMHB at 37 °C with shaking at 220 rpm to exponential phase (OD600 ≈ 0.6; approximately 4 h). Total RNA was extracted using the Total RNA Purification Kit (Sangon Biotech, Shanghai, China). Ribosomal RNA was depleted using a probe-based depletion method, and the remaining mRNA was purified for cDNA library construction. The quality of cDNA libraries was assessed using an Agilent Bioanalyzer 2100 system (Agilent Technologies, USA). High-throughput sequencing was performed on the Illumina NovaSeq platform, generating 150 bp paired-end reads. Clean reads were mapped to the K. pneumoniae HS11286 reference genome (NCBI accession: GCF_000240185.1) using Bowtie2 v2.2.3 [37]. Differentially expressed genes (DEGs) were identified with edgeR (adjusted p < 0.05, |log2FC| ≥ 1). Functional enrichment analysis of DEGs was performed using the KEGG PATHWAY database via custom R scripts, and Gene Ontology (GO) enrichment analysis was conducted with Goatools software using Fisher’s exact test. Statistical analysis Statistical analyses were conducted using GraphPad Prism 8 (GraphPad Software Inc., San Diego, CA, USA). Comparisons of colistin MICs and gene expression levels between two groups were performed using the Wilcoxon rank-sum test or unpaired t-test, respectively. For comparisons among three groups, one-way analysis of variance (ANOVA) was applied. A p-value of < 0.05 was considered statistically significant. Results Prevalence of mutations in PmrAB, PhoPQ, CrrAB two-component systems and the regulator MgrB Multilocus sequence typing revealed that the majority of 239 non-duplicate clinical hv-CRKP isolates belonged to the epidemic ST11 lineage (210/239, 87.87%). Other detected sequence types included ST23 (6/239, 2.51%), ST1 (3/239, 1.26%), ST147 (3/239, 1.26%), ST307 (3/239, 1.26%), ST86 (3/239, 1.26%) (Fig. 1A). No mcr genes (mcr-1 to mcr-9) were detected in any hv-CRKP isolate by PCR screening. The distribution and frequency of mutations in pmrA, pmrB, phoP, phoQ, crrA, crrB, and mgrB are summarized in Fig. 1B and C. Among these genes, pmrB exhibited the highest mutation rate, with nonsynonymous substitutions detected in 232 out of 239 isolates (232/239, 97.07%). Of these, 231 harbored missense mutations (231/239, 96.65%), while one isolate carried a frameshift mutation. Mutations in pmrA were identified in 17 isolates (17/239, 7.11%), exclusively as missense variants. No mutations were detected in phoP, and only two isolates carried missense mutations in phoQ (2/239, 0.84%). Interestingly, the CrrAB two-component system was not universally present. PCR analysis showed that all ST11 isolates carried incomplete or truncated crrA and crrB genes. Instead, crrAB was detected in only 17 isolates, belonging to diverse sequence types including ST1 (n = 3), ST147 (n = 3), ST307 (n = 3), ST37 (n = 2), ST1107 (n = 1), ST16 (n = 1), ST218 (n = 1), ST29 (n = 1), ST3161 (n = 1), and ST86 (n = 1). Among these crrAB-positive isolates, no mutations were identified in crrA, whereas all 17 carried missense mutations in crrB (17/239, 7.11%). The negative regulator mgrB was mutated in 15 isolates (15/239, 6.28%), comprising IS element insertions in 11 isolates (11/239, 4.60%), premature stop codons in 3 isolates (3/239, 1.26%), and a single missense variant (1/239, 0.42%). Fig. 1Distribution and prevalence of mutations in 239 hv-CRKP Clinical Isolates. A Sequence type distribution of 239 hv-CRKP isolates in this study. B Frequency of mutations identified in seven key chromosomal loci (pmrA, pmrB, phoP, phoQ, crrA, crrB and mgrB) associated with colistin resistance. C Categorization of mutation types. Mutation types were categorized as insertion/deletion/frameshift (ins, del, fs), premature stop codons (stop), and other single nucleotide polymorphisms (other snp) Distribution and prevalence of mutations in 239 hv-CRKP Clinical Isolates. A Sequence type distribution of 239 hv-CRKP isolates in this study. B Frequency of mutations identified in seven key chromosomal loci (pmrA, pmrB, phoP, phoQ, crrA, crrB and mgrB) associated with colistin resistance. C Categorization of mutation types. Mutation types were categorized as insertion/deletion/frameshift (ins, del, fs), premature stop codons (stop), and other single nucleotide polymorphisms (other snp) Distribution of chromosomal mutations and their association with colistin mics in hv-CRKP To evaluate the impact of specific mutations on colistin susceptibility, we determined the colistin MICs of all 239 hv-CRKP isolates and analyzed their association with mutations in pmrA, pmrB, phoQ, crrA, crrB, and mgrB. Thirty different mutations were identified across these genes (Fig. 2A). Two phoQ variants were identified, V450D and R16C, with each found in only one colistin-resistant isolate. pmrB exhibited the highest mutational diversity, with R256G being the most prevalent substitution (215/239, 89.96%). This variant (R256G) was detected in all ST11 strains (210/215, 97.67%) as well as in ST147 strains (3/215, 1.40%) and ST37 strains (2/215, 0.93%). Notably, specific pmrB mutations were exclusively detected in colistin-resistant isolates, including T157P (14/239, 5.86%), P95L (4/239, 1.67%), S205P (3/239, 1.26%), D150H (2/239, 0.84%), D150V (1/239, 0.42%), S203P (1/239, 0.42%), and L339fs (1/239, 0.42%). Conversely, pmrB mutations G345E (3/239, 1.26%), L213M (3/239, 1.26%), P344L (2/239, 0.84%), T240M (2/239, 0.84%), and V257A (1/239, 0.42%) were exclusively identified in colistin-susceptible isolates. pmrA mutations showed variable associations with colistin resistance. Substitutions at glycine 53 (G53S/A/V (7/239, 2.93%) were consistently observed in colistin resistant isolates. In contrast, Y155H, A41T, and D86E occurred exclusively in susceptible strains. Within the CrrAB system, five substitutions were observed in crrB, including C68S (17/239, 7.11%), V284I (3/239, 1.26%), I66V (2/239, 0.84%), I27V (2/239, 0.84%), and T276A (2/239, 0.84%). All of these variants were exclusively detected in colistin-susceptible isolates (Fig. 2A). Disruptive mgrB mutations were only detected in colistin-resistant isolates, including ISkpn26 insertions (10/239, 4.18%), Q30* mutations (3/239, 1.26%), and IS903B insertion (1/239, 0.42%). An additional variant, C16R (1/239, 0.42%), was also identified in a resistant strain. Colistin MIC distributions further elucidated genotype-phenotype relationships (Fig. 2B). Fig. 2Chromosomal mutations and colistin susceptibility profiles of hv‑CRKP isolates. A Distribution of 30 distinct mutations in phoQ, pmrA, pmrB, crrA, crrB and mgrB among 239 isolates. Bars represent the number of isolates carrying each mutation, with colors indicating resistant and susceptible phenotypes. B Colistin MIC distributions stratified by mutation site. Each dot denotes an individual isolate. The dashed line represents the EUCAST resistance breakpoint (> 2 mg/L) Chromosomal mutations and colistin susceptibility profiles of hv‑CRKP isolates. A Distribution of 30 distinct mutations in phoQ, pmrA, pmrB, crrA, crrB and mgrB among 239 isolates. Bars represent the number of isolates carrying each mutation, with colors indicating resistant and susceptible phenotypes. B Colistin MIC distributions stratified by mutation site. Each dot denotes an individual isolate. The dashed line represents the EUCAST resistance breakpoint (> 2 mg/L) Distribution of pmrB mutations and their association with colistin MICs in hv-CRKP Domain annotation identified that PmrB contains two transmembrane helices, a HAMP domain (residues 90–142), a HisKA phosphoacceptor domain (143–203), and a C-terminal HATPase_c domain (250–358) (Fig. 3A). Mapping these mutations onto a predicted PmrB tertiary structure showed their spatial distribution across key functional domains (Fig. 3B). Pro95 is located in the HAMP domain, while Asp150, Thr157, and Ser203 are positioned within the HisKA domain essential for autophosphorylation. Ser205 lies at the boundary between the HisKA and HATPase_c domains. Fig. 3Structural localization and effects of pmrB mutations in colistin-resistant hv‑CRKP isolates. A Domain structure of PmrB, illustrating the transmembrane (TM, residues 13–35 and 67–89), HAMP (residues 90–142), HisKA (residues 143–203), and HATPase_c (residues 250–358) domains. Each domain is color-coded as follows: TM (blue), HAMP (pink), HisKA (green), and HATPase_c (orange). B Predicted tertiary structure of PmrB, with the resistance-associated mutations mapped onto their respective positions. C Colistin MIC distributions of pmrB mutants grouped by the domain location of mutations. D Relative expression levels of arnB, arnC, arnT, and eptA Structural localization and effects of pmrB mutations in colistin-resistant hv‑CRKP isolates. A Domain structure of PmrB, illustrating the transmembrane (TM, residues 13–35 and 67–89), HAMP (residues 90–142), HisKA (residues 143–203), and HATPase_c (residues 250–358) domains. Each domain is color-coded as follows: TM (blue), HAMP (pink), HisKA (green), and HATPase_c (orange). B Predicted tertiary structure of PmrB, with the resistance-associated mutations mapped onto their respective positions. C Colistin MIC distributions of pmrB mutants grouped by the domain location of mutations. D Relative expression levels of arnB, arnC, arnT, and eptA To assess whether the location of pmrB mutations correlates with colistin MIC levels and the expression of pmrB-regulated genes (arnB, arnC, arnT, and eptA), we categorized the pmrB missense mutants in our study into three groups based on mutation location: the HAMP group (P95L), the HisKA group (D150H, D150V, T157P, S203P), and the Other group (S205P, L213M, T240M, V257A, P344L, G345E). As shown in Fig. 3C, strains with mutations in the HAMP or HisKA domains exhibited significantly elevated MICs (median 32 mg/L and 24 mg/L, respectively) compared to strains with mutations in other domains (median 1 mg/L; P < 0.01 and P < 0.0001, respectively). RT-qPCR analysis revealed significantly higher expression of arnB, arnC, arnT, and eptA in strains with HAMP/HisKA mutations compared to those with mutations in other domains (P < 0.0001, Fig. 3D). Epidemiological and genomic characteristics of colistin-resistant hv-CRKP with chromosomal mutations To explore the molecular characteristics of colistin-resistant hv-CRKP, we performed whole-genome sequencing on all hv-CRKP isolates harboring at least one resistance-associated mutation in pmrA, pmrB, phoQ, or mgrB (47 isolates in total). A core genome-based maximum likelihood phylogenetic tree was constructed and annotated with isolate metadata (Fig. 4). All 47 isolates belonged to the epidemic ST11 clone. Capsular typing revealed that the majority of isolates were either KL64 (25/47, 53.19%) or KL25 (21/47, 44.68%). Carbapenemase profiling revealed universal carriage of blaKPC, with KPC-2 predominating (46/47, 97.87%); a single isolate carried KPC-71. β-lactamase gene analysis revealed that 27 isolates (27/47, 57.45%) harbored blaSHV−12, 9 isolates (9/47, 19.15%) harbored blaSHV−11, and 2 isolates (2/47, 4.26%) harbored blaSHV−129. The majority of isolates (44/47, 93.62%) only carried blaCTX−M−65, while 2 isolates (2/47, 4.26%) harbored both blaCTX−M−3 and blaCTX−M−65. None of the isolates carried NDM-type carbapenemases. Virulence gene analysis demonstrated that all isolates harbored both rmpA2 and iucABCD, two hallmark virulence loci associated with hypermucoviscosity and aerobactin-mediated iron acquisition, respectively. Most isolates (45/47, 95.74%) harbored a single colistin resistance-associated mutation, while two isolates carried combinatorial mutations. Specifically, one isolate harbored both phoQ V450D and an ISkpn26 insertion in mgrB, whereas another isolate carried phoQ R16C and pmrB mutations (P95L and L339fs). These two dual-mutant isolates exhibited high colistin MICs (64 mg/L) (Fig. 4). Fig. 4Phylogenetic analysis and molecular characteristics of 47 colistin-resistant hv-CRKP isolates harboring chromosomal mutations. A core genome maximum-likelihood phylogenetic tree is shown with integrated metadata, including multilocus sequence type (MLST), capsular (KL) type, virulence genes (rmpA2 and iucABCD), and KPC, SHV, and CTX-M variants. Resistance-associated mutations are marked by colored circles. Colistin MICs are displayed in the right panel Phylogenetic analysis and molecular characteristics of 47 colistin-resistant hv-CRKP isolates harboring chromosomal mutations. A core genome maximum-likelihood phylogenetic tree is shown with integrated metadata, including multilocus sequence type (MLST), capsular (KL) type, virulence genes (rmpA2 and iucABCD), and KPC, SHV, and CTX-M variants. Resistance-associated mutations are marked by colored circles. Colistin MICs are displayed in the right panel Functional validation of novel pmrB point mutations in colistin resistance To assess the contribution of pmrB mutations to colistin resistance, we focused on the nonsynonymous substitutions D150H and S203P. These mutations, located within the HisKA signaling domain of PmrB (Fig. 3), were exclusively identified in colistin-resistant ST11-KL25 hv-CRKP isolates KP7145 (D150H) and KP11396 (S203P), and their functional roles had not previously been validated. Using a CRISPR-Cas9 genome-editing system, isogenic revertant strains were generated by restoring the mutated residues to their wild-type codons. The revertants, KP7145-Rev and KP11396-Rev, were confirmed through Sanger sequencing (Fig. 5A), which verified successful codon reversion and restoration of wild-type sequences. Broth microdilution assays revealed that KP7145 and KP11396 exhibited high-level colistin resistance with MICs of 32 mg/L (Fig. 5B). However, the isogenic revertants displayed a significant reduction in colistin MICs, with values ≤ 4 mg/L. To further validate whether pmrB D150H and S203P contribute to colistin resistance, we individually introduced these point mutations into the colistin-susceptible strain NTUH-K2044 (MIC = 1 mg/L). The engineered mutants carrying either D150H or S203P substitutions exhibited elevated colistin MICs of 4 mg/L, confirming the resistance-conferring effect of both mutations. Fig. 5Functional validation of colistin resistance-associated pmrB mutations. A Sanger sequencing chromatograms confirming the successful reversion of D150H and S203P mutations in the pmrB gene to their wild-type codons. B Colistin MIC values determined by broth microdilution for wild-type strains (KP7145 and KP11396) and their corresponding revertants. C Growth kinetics of wild-type and revertant strains under increasing concentrations of colistin (1, 2, and 4 mg/L) Functional validation of colistin resistance-associated pmrB mutations. A Sanger sequencing chromatograms confirming the successful reversion of D150H and S203P mutations in the pmrB gene to their wild-type codons. B Colistin MIC values determined by broth microdilution for wild-type strains (KP7145 and KP11396) and their corresponding revertants. C Growth kinetics of wild-type and revertant strains under increasing concentrations of colistin (1, 2, and 4 mg/L) Adaptive fitness and pathogenic potential of pmrB variants Growth kinetics experiments revealed distinct fitness characteristics associated with the two mutations (Fig. 5C). In the absence of colistin, KP7145 and KP7145-Rev exhibited similar growth rates, while KP11396-Rev showed slightly faster growth than KP11396. Under sub-inhibitory concentrations of colistin (1–4 mg/L), both original strains sustained robust growth while their revertants showed significantly impaired growth. Time-kill analysis demonstrated sustained bacterial survival in the wild-type strains, KP7145 and KP11396, during exposure to 4 mg/L colistin (Fig. 6A). KP7145-Rev exhibited rapid bactericidal killing with CFUs dropping below detection limits within four hours, while KP11396-Rev showed initial reduction followed by regrowth. Competition assays revealed no fitness advantage for revertants in antibiotic-free conditions, but showed a significant decline under sub-MIC colistin exposure (1 mg/L) (Fig. 6B). KP7145-Rev decreased from 51.96% to 18.32%, while KP11396-Rev declined from 52.47% to 16.35%. In the G. mellonella infection model, both original and revertant strains caused comparable mortality rates (Fig. 6C), suggesting that pmrB mutations do not significantly affect virulence. Fig. 6Fitness and competitive dynamics of pmrB mutations in hv‑CRKP. A Time-kill curves under 4 mg/L colistin exposure over 24 h. B Competitive experiment of revertant versus wild-type strains in co-culture without/with sub-MIC colistin (1 mg/L). C Survival curves of G. mellonella larvae infected with wild-type and revertant strains Fitness and competitive dynamics of pmrB mutations in hv‑CRKP. A Time-kill curves under 4 mg/L colistin exposure over 24 h. B Competitive experiment of revertant versus wild-type strains in co-culture without/with sub-MIC colistin (1 mg/L). C Survival curves of G. mellonella larvae infected with wild-type and revertant strains Transcriptomic analysis reveals reversion of pmrB mutations suppresses lipid a modification pathways RNA-seq analysis revealed that reversion to wild-type pmrB resulted in significant transcriptional changes. In KP7145-Rev versus KP7145, 37 genes were upregulated and 153 were downregulated (FDR < 0.05, |log2FC| > 1) (Fig. 7A). The most pronounced downregulation occurred in lipid A modification genes, including arnB (log2FC = -5.67), arnC (-4.31), arnD (-4.23), arnA (-4.08), eptA (-3.46), pmrA (-3.12), and pmrB (-2.68). Similarly, in KP11396-Rev versus KP11396, 19 genes were upregulated and 129 were downregulated (Fig. 7B), with key resistance genes arnB (-3.50), eptA (-1.95), arnC (-2.05), arnA (-1.46), and arnD (-1.53) significantly suppressed. Pathway enrichment analysis showed that cationic antimicrobial peptide resistance was the most enriched pathway in both comparisons (Fig. 7C and D). The most significantly enriched biological processes were cobalamin biosynthesis, tetrapyrrole metabolism, and water-soluble vitamin biosynthesis (Fig. 7E). Cellular component analysis revealed enrichment in bacterial microcompartment organization (Fig. 7E). Intersection analysis identified 17 consistently downregulated genes across both revertant-mutant pairs, including the core lipid A modification components arnA, arnB, arnC, arnD, and eptA (Fig. 7F). Additional downregulated genes included bacA (peptidoglycan biosynthesis), cybB (NADPH oxidase), fadE (fatty acid degradation), glnK (nitrogen regulation), and mgtE (magnesium transport). Only one gene, KPHS_37020 (an uncharacterized gene), was consistently upregulated in both comparisons. Fig. 7Transcriptional profiling of pmrB D150H and S203P reversion mutations. A, B Volcano plots of differentially expressed genes (DEGs) in KP7145 vs. KP7145‑Rev and KP11396 vs. KP11396‑Rev. Genes with |log2FC| >1 and FDR < 0.05 are highlighted (blue, upregulated; red, downregulated). C, D KEGG pathway enrichment of DEGs from the two strain pairs. E Biological process and cellular component analysis of DEGs from both revertant-mutant pairs. F Venn diagram identifying shared differentially expressed genes between both comparisons. Overlapping sections indicate common up/downregulated genes Transcriptional profiling of pmrB D150H and S203P reversion mutations. A, B Volcano plots of differentially expressed genes (DEGs) in KP7145 vs. KP7145‑Rev and KP11396 vs. KP11396‑Rev. Genes with |log2FC| >1 and FDR < 0.05 are highlighted (blue, upregulated; red, downregulated). C, D KEGG pathway enrichment of DEGs from the two strain pairs. E Biological process and cellular component analysis of DEGs from both revertant-mutant pairs. F Venn diagram identifying shared differentially expressed genes between both comparisons. Overlapping sections indicate common up/downregulated genes Discussion Colistin resistance in K. pneumoniae poses a critical threat to global health, as it undermines one of the few remaining therapeutic options for carbapenem-resistant infections [12]. Mutations in two-component regulatory systems, particularly PmrAB, PhoPQ, and CrrAB, as well as disruptions of the negative regulator MgrB, have been identified as key contributors to colistin resistance [16, 38, 39]. However, distinguishing lineage-specific background polymorphisms from true resistance determinants remains a major challenge, limiting the predictive power of genomic surveillance and complicating clinical diagnostics. Here, we investigated the sequence types and chromosomal mutations associated with colistin resistance in 239 hv-CRKP isolates. In line with global epidemiology trends, most isolates belonged to the epidemic ST11 lineage. Although previous studies have frequently reported pmrB R256G as a resistance-associated mutation [38, 40], our data indicate that R256G represents a lineage-specific polymorphism (in ST11, ST147, and ST37) rather than causal determinant of resistance. This finding highlights the importance of overestimating resistance determinants when relying on genomic data alone. Similarly, several pmrA mutations (Y155H, A41T, and D86E) and pmrB mutations (G345E, L213M, P344L, T240M, and V257A) were detected exclusively in colistin-susceptible strains, suggesting that these amino acid substitutions are likely neutral polymorphisms rather than functionally significant alterations. This re-interpretation emphasizes the importance of integrating whole-genome data from colistin-susceptible isolates into comparative genomic analyses to avoid misclassification of lineage-specific polymorphisms as resistance markers. Misclassification of such naturally occurring variants could lead to erroneous assumptions about resistance mechanisms and hinder accurate surveillance of colistin resistance. In our study, colistin resistance was primarily mediated by mutations within key functional domains of pmrB, particularly the HAMP and HisKA regions. Potentially disruptive alterations in mgrB and rare substitutions in phoQ were also observed. Recent studies simulating K. pneumoniae infections have shown that colistin resistance can emerge rapidly, with mutations in regulatory genes pmrB and crrB occurring within 24 h under therapeutic drug concentrations [41]. These findings are consistent with our observation that pmrB mutations represent a common resistance mechanism. However, although crrB variants were detected across diverse sequence types in this study, they were confined to colistin-susceptible isolates. Together, these results emphasize the necessity of integrating functional evidence with genomic surveillance to avoid misclassification of background polymorphisms as causal mutations. Capsular polysaccharides are recognized as major virulence determinants in K. pneumoniae, modulating immune evasion, serum resistance, and mucosal colonization [42]. Variations in capsular structure and composition significantly affect pathogenicity, with certain KL types showing strong associations with hypervirulent phenotypes [43]. The increasing prevalence of ST11 CRKP carrying virulence plasmids has established KL47 and KL64 as the predominant capsular types in China [44, 45]. In this study, whole-genome sequencing shown that all colistin-resistant isolates belonged to ST11, with KL64 and KL25 as the major capsule types. The identification of ST11-KL25 as an emerging lineage carrying both carbapenem resistance and hypervirulence is particularly significant, as it suggests a potential shift in the dominant epidemic clone [46]. This finding extends previous reports focused on KL64 and raises the possibility that KL25 provides unique adaptive advantages in the exposure of antibiotic and host pressures. Future investigations should test whether capsule type influences the balance between resistance acquisition and fitness costs. Functional validation revealed that the novel mutations pmrB D150H and S203P confer colistin resistance by constitutively activating lipid A modification gene expression. These pmrB mutations did not impose significant fitness costs under standard in vitro conditions. Instead, they conferred a clear survival advantage under colistin pressure without attenuating virulence. This result suggests that resistance acquisition may not be inevitably linked to reduced fitness or virulence [47, 48]. It suggests that colistin-resistant hv-CRKP harboring chromosomal pmrB mutations may circumvent the usual trade-off between resistance and pathogenicity. Inactivation of mgrB can confer colistin resistance while simultaneously enhancing virulence by reducing susceptibility to host antimicrobial peptides and attenuating early immune activation [49]. However, other studies have reported that mgrB loss may compromise gastrointestinal colonization fitness [50]. Insertion sequence (IS) element insertions represent a well-documented mechanism of mgrB inactivation leading to colistin resistance [16]. IS elements are highly mobile, and their activity may generate genetic heterogeneity within clonal populations, thereby accelerating the emergence of resistance under antibiotic pressure [51]. A prolonged outbreak in Brazil demonstrated that 84.7% of ColR-CRKP isolates belonged to a single cluster with mgrB alterations predominantly mediated by IS903B, with cross-transmission occurring even among patients who had not received polymyxin therapy [52]. In our study, IS-mediated disruption emerged as the predominant form of mgrB alteration, consistent with previous observations [16, 53]. Interestingly, while most disruptive mgrB mutations conferred high-level colistin resistance (MICs ≥ 16 mg/L), the isolate with an IS903B insertion exhibited a comparatively lower resistance level (MIC = 8 mg/L, Fig. 4). This suggests that different IS elements may vary in their impact on gene disruption and resistance expression, potentially reflecting differences in insertion site, orientation, or secondary regulatory effects. Further functional analyses will be needed to clarify the mechanistic basis of these discrepancies. Taken together, these findings reveal that different chromosomal mechanisms of colistin resistance have divergent effects on bacterial resistance, fitness and pathogenesis, which may explain the heterogeneous distribution of resistance pathways across clinical isolates. This study has some limitations. First, while the G. mellonella model offered useful insights into virulence, mammalian models are needed to better assess the pathogenic impact of resistance in more complex physiological contexts. Second, we did not directly compare the fitness effects of different resistance pathways. Variants such as pmrB mutations and disruptive changes in mgrB may carry distinct biological costs. Despite these limitations, this study characterizes the molecular epidemiology of colistin resistance in hv-CRKP and underscoring their importance for surveillance and control. Conclusion Our study demonstrates that colistin resistance in hv-CRKP is primarily driven by chromosomal alterations in pmrB and mgrB, with novel pmrB variants (D150H, S203P) functionally validated as resistance determinants. The emergence of colistin-resistant ST11 hv-CRKP highlights the urgent need for ongoing genomic surveillance and the development of effective therapeutic strategies to mitigate their growing public health threat. Supplementary Information Supplementary Material 1. Supplementary Material 1.