Clinical Characteristics and Whole-Genome Sequencing Analysis of Carbapenem- and Colistin-Resistant Hypervirulent Klebsiella pneumoniae

Introduction

Klebsiella pneumoniae is an important opportunistic pathogen within the Enterobacteriaceae family and a major cause of nosocomial infections.¹ It is generally categorized into classical K. pneumoniae (cKP) and hypervirulent K. pneumoniae (hvKP) based on distinct genetic and phenotypic characteristics. cKP primarily affects immunocompromised individuals and readily acquires antimicrobial resistance determinants, contributing to the widespread emergence of carbapenem-resistant K. pneumoniae (CRKP).² In contrast, hvKP is defined by the presence of specific virulence-associated genes, including peg-344, rmpA, rmpA2, iroB, and iucA, which confer enhanced invasiveness, enabling severe community- and hospital-acquired infections in otherwise healthy hosts, often involving multiple organs.^3,4 The convergence of hypervirulence and carbapenem resistance has led to the rise of hypervirulent carbapenem-resistant K. pneumoniae (hv-CRKP), a formidable “dual threat” that combines high virulence with limited therapeutic options.⁵ These strains significantly complicate clinical treatment and are associated with increased morbidity and mortality.^6,7

Polymyxins, including polymyxin B and polymyxin E (colistin), are nonribosomal peptide antibiotics that have historically been reserved as one of the last-line treatment options for carbapenem-resistant Gram-negative bacterial infections because of concerns regarding neurotoxicity and nephrotoxicity.⁸ However, the global emergence of colistin-resistant hv-CRKP (ColR-hv-CRKP) further exacerbates the clinical burden.^9,10 It is well established that resistance to polymyxins in Gram-negative bacteria is primarily mediated by modifications to lipopolysaccharide (LPS), which increase the net positive charge on the bacterial surface.¹¹ This mechanism is known to reduce electrostatic interaction between LPS and the cationic colistin, thereby diminishing antimicrobial efficacy.^12,13 Previous molecular studies have demonstrated several resistance mechanisms, including mutations or constitutive activation of two-component regulatory systems (TCSs) such as PhoP/Q, PmrA/B, and CrrAB; inactivation or mutation of the negative regulator mgrB; and the horizontal transfer of mcr genes.^14,15 Despite these insights, comprehensive data on the clinical and genomic features of ColR‑hv‑CRKP remain limited, and the specific chromosomal mutations contributing to colistin resistance in the hv-CRKP isolates have not been systematically characterized. Therefore, the primary objective of this study was to identify clinical risk factors associated with ColR-hv-CRKP infection. The secondary objectives were to characterize the genomic diversity of ColR-hv-CRKP isolates and to elucidate the chromosomal mechanisms underlying colistin resistance.

In this study, we investigated the prevalence, clinical characteristics, and risk factors associated with ColR‑hv‑CRKP infections. We also performed genomic analyses of 92 ColR‑hv‑CRKP isolates to characterize their molecular features, phylogenetic relationships, resistance gene profiles, and chromosomal mutations related to colistin resistance.

Materials and Methods

Isolate Collection and Identification

A total of 326 non-duplicate clinical isolates of hypervirulent carbapenem-resistant K. pneumoniae (hv-CRKP) and corresponding clinical data were collected from the Clinical Microbiology Laboratory of the First Affiliated Hospital of Nanchang University between January 2021 and December 2024. The isolates were obtained from various clinical specimens, including sputum, blood, bronchoalveolar lavage fluid (BALF), ascites, pus, secretion, urine and drainage. Only the first isolate per patient during the study period was included to avoid duplication. Bacterial identification was performed using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). Carbapenem and colistin resistance was determined according to the Clinical and Laboratory Standards Institute (CLSI M100-34) guidelines,¹⁶ and isolates resistant to at least one carbapenem agent (imipenem (MIC ≥ 4 mg/L), meropenem (MIC ≥ 4 mg/L), or ertapenem (MIC ≥ 2 mg/L)) were defined as carbapenem-resistant. Among these isolates, 92 strains were further identified as colistin‑resistant hv‑CRKP and selected for subsequent antimicrobial susceptibility testing and whole genome sequencing. Hypervirulence was assessed using a combination of phenotypic and genotypic criteria. The hypervirulent phenotype was initially screened using the string test, with a positive result defined as the formation of a viscous string ≥ 5 mm in length when a colony was stretched by an inoculation loop. Isolates that were positive in the string test and harbored known virulence-associated genes (rmpA, rmpA2, iucA, and iroB) by PCR were defined as hypervirulent.³ All isolates were stored in Luria-Bertani (LB) broth supplemented with 20% glycerol at –80 °C for further analysis.

Antimicrobial Susceptibility Testing

Initial antimicrobial susceptibility testing for all isolates was performed using the VITEK 2 automated system (bioMérieux, Marcy-l’Étoile, France). Isolates exhibiting resistance to any carbapenem antibiotic (imipenem, meropenem, or ertapenem) and to colistin, based on VITEK 2 interpretation, were designated as putative ColR-hv-CRKP isolates. Subsequently, the antimicrobial susceptibility of the 92 ColR-hv-CRKP isolates was determined using the broth microdilution method for the following antibiotics: imipenem (1–64 μg/mL), meropenem (1–64 μg/mL), cefotaxime (1–64 μg/mL), ceftriaxone (1–64 μg/mL), ceftazidime (1–64 μg/mL), cefazolin (1–64 μg/mL), levofloxacin (0.5–32 μg/mL), ciprofloxacin (0.5–32 μg/mL), aztreonam (1–64 μg/mL), ticarcillin/clavulanate (16/2-128/2 μg/mL), piperacillin/tazobactam (2/4-128/4 μg/mL), ceftazidime/avibactam (2/4-128/4 μg/mL), tobramycin (1–64 μg/mL), amikacin (2–128 μg/mL), doxycycline (1–64 μg/mL), minocycline (1–64 μg/mL), tigecycline (2–16 μg/mL), trimethoprim/sulfamethoxazole (2/38-8/152 μg/mL), and colistin (1–64 μg/mL). Antibiotic powders were obtained from commercial sources (MedChemExpress, USA), and serial twofold dilutions were prepared according to the manufacturer’s instructions. All testing procedures and result interpretations were conducted according to the CLSI M100‑Ed34.¹⁷ According to CLSI breakpoints for K. pneumoniae, isolates with colistin MICs > 2 mg/L were defined as resistant. K. similipneumoniae ATCC 700603, K. pneumoniae ATCC 13883 and Escherichia coli ATCC 25922 were used as the quality‑control strain for all assays.

Clinical Data Collection and Definitions

Clinical and demographic data were retrospectively collected from electronic medical records. Patients were eligible for inclusion if they were hospitalized at the First Affiliated Hospital of Nanchang University between January 2021 and December 2024 and had a confirmed hv-CRKP isolate isolated from a clinical specimen. Patients of all age groups were included. Only cases considered to represent infection were enrolled. Infection was defined based on compatible clinical manifestations together with microbiological evidence, as determined by treating physicians and confirmed by medical record review. Patients were excluded if they had multiple hospital admissions during the study period, in which case only the first admission was included, if medical information or key clinical data were incomplete, or if the isolate was considered a contaminant. Surgery was defined as any operative procedure conducted under general or regional anesthesia within 90 days before isolate collection. Invasive procedures encompassed mechanical ventilation, central venous catheterization, urinary catheterization, tracheostomy, and drainage interventions. Prolonged hospitalization was defined as a continuous hospital stay lasting more than 30 days prior to the recovery of the isolate.

DNA Extraction and Whole Genome Sequencing

Genomic DNA was extracted from overnight cultures of colistin‑resistant isolates using the TIANamp Bacteria DNA Kit (Tiangen Biotech, China) following the manufacturer’s protocol. DNA quality and concentration were evaluated using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA) and 1% agarose‑gel electrophoresis. Whole‑genome sequencing (WGS) was performed on the Illumina NovaSeq 6000 platform (Illumina, USA) generating 150 bp paired‑end reads. Raw sequencing reads were trimmed using Trimmomatic v0.39 to remove low-quality bases and adapter sequences.¹⁷ High-quality clean reads were then assembled de novo using SPAdes v4.0.0.¹⁸ Assembly quality was evaluated with QUAST v5.2.0.¹⁹ Genome annotation was performed on the high-quality assemblies using Bakta v1.6.0.²⁰ In silico multilocus sequence typing (MLST) was carried out using mlst v2.11.0, with allele profiles retrieved from the PubMLST database (https://pubmlst.org/). Capsular polysaccharide (K) and lipopolysaccharide (O) locus types were identified using Kaptive v2.0.7.²¹ Antimicrobial resistance genes, including mobilized colistin resistance genes (mcr-1 to mcr-10) using ResFinder v4.1, and virulence genes were identified using the Virulence Factor Database (VFDB),^22,23 respectively. The virulence scores of the strains were assigned using the Kleborate tool (https://github.com/katholt/Kleborate).²⁴

Phylogenetic Analysis of Colistin‑Resistant Hv-CRKP Isolates

Core-genome single-nucleotide polymorphisms (SNPs) were identified using Snippy v4.6.0 (https://github.com/tseemann/snippy) with default parameters, employing the ST11 K. pneumoniae reference strain HS11286 (GCF_000240185.1) as the reference genome. The resulting core genome alignment was filtered to remove recombinant regions using Gubbins v3.2.1 to ensure accurate phylogenetic inference.²⁵ SNP‑sites v2.5.1 was then used to extract biallelic SNP positions from the recombination‑free alignment.²⁶ Pairwise SNP distances between isolates were computed with snp‑dists v0.8.2 (https://github.com/tseemann/snp-dists), and isolates differing by ≤ 16 core SNPs were defined as potential clonal clusters.²⁷ A maximum‑likelihood phylogenetic tree was reconstructed from the filtered core‑SNP alignment using IQ‑TREE2 v2.2.6, with automatic model selection and 1,000 ultrafast bootstrap replicates.²⁸ The final tree was visualized and annotated with Interactive Tree of Life (iTOL) (https://itol.embl.de).²⁹

Identification of Colistin Resistance‑Associated Mutations

To investigate the molecular mechanisms of colistin resistance, each genome was aligned against the colistin‑susceptible K. pneumoniae reference strain MGH 78578 (RefSeq accession GCF_000016305.1) and further compared with the colistin‑susceptible reference HS11286 (RefSeq accession GCF_000240185.1) to exclude polymorphisms unrelated to colistin resistance. Snippy v4.6.0 (https://github.com/tseemann/snippy) was used with default parameters to identify single‑nucleotide polymorphisms (SNPs) and insertion‑deletion (indel) mutations within the PmrA/B, PhoP/Q, and mgrB loci. Variants were subsequently annotated with SnpEff v5.1 to assess their predicted functional impact.³⁰ Insertion sequences (IS) were identified using the online tool ISfinder (https://www-is.biotoul.fr/).

Statistical Analysis

Statistical analyses were performed using SPSS (version 26.0, IBM Corp). Categorical variables were compared using Chi-square or Fisher’s exact test, while continuous variables were analyzed using Student’s t-test or Mann–Whitney U-test as appropriate. Risk factors for ColR-hv-CRKP infection were identified through the above differential analysis, with variables showing P <0.05 subsequently included in multivariate logistic analysis. A P value < 0.05 was considered statistically significant. GraphPad Prism 8 was used for data visualization.

Results

Clinical Characteristics of Patients Infected with ColR-Hv-CRKP

A total of 326 non‑duplicate hv‑CRKP isolates were collected from the First Affiliated Hospital of Nanchang University between 2021 and 2024. Among these, 92 isolates (92/326, 28.22%) demonstrated resistance to colistin and were classified as ColR‑hv‑CRKP, whereas 234 isolates (71.78%) remained susceptible to colistin and were classified as colistin-susceptible hv-CRKP (ColS‑hv‑CRKP). The demographic and clinical characteristics of patients infected with ColR‑hv‑CRKP and ColS‑hv‑CRKP are summarized in Table 1. Patients with ColR‑hv‑CRKP infection were more frequently admitted to the ICU (53.26% vs 33.76%, P = 0.013) and had a higher prevalence of prior ICU stays (89.13% vs 76.90%, P = 0.013). Sputum specimens accounted for a higher proportion in the ColR-hv-CRKP group than in the ColS-hv-CRKP group (78.26% vs 57.69%, P = 0.001). Prolonged hospitalization (> 30 days; 64.13% vs 49.60%, P = 0.019) and invasive procedures (94.57% vs 86.70%, P = 0.049) were significantly more common among ColR‑hv‑CRKP cases. However, the proportions of patients with hospitalization within the previous month or long‑term steroid/immunosuppressant exposure did not differ between groups. Prior antibiotic exposure differed significantly between the two groups (χ²-test, P < 0.001). Carbapenem-based regimens were more frequently used in the ColS-hv-CRKP group than in the ColR-hv-CRKP group (54.27% vs. 39.13%). In contrast, prior exposure to colistin-based regimens was more common in the ColR-hv-CRKP group compared with the ColS-hv-CRKP group (36.96% vs. 8.97%). Multivariable analysis identified prior colistin administration (OR = 4.58, 95% CI 2.41–8.69) and sputum specimen source (OR = 2.44, 95% CI 1.35–4.42) as independent risk factors for ColR-hv-CRKP infection (Table 2). Other variables, including prolonged hospitalization, ICU admission, or prior invasive procedures, were not independently associated after multivariate logistics regression analysis.

Table 1 Demographic and Clinical Characteristics of Patients Infected with ColR-Hv-CRKP and ColS-Hv-CRKP

Table 2 Multivariate Analysis of Risk Factors for ColR-Hv-CRKP Infection

Distribution and Molecular Characteristics of ColR-Hv-CRKP Isolates

A total of 92 ColR-hv-CRKP isolates were identified between 2021 and 2024, including 14 (15.22%) isolates in 2021, 17 (18.48%) in 2022, 42 (45.65%) in 2023, and 19 (20.65%) in 2024 (Figure 1A). Most ColR-hv-CRKP isolates were recovered from the Intensive Care Unit (49/92, 53.26%), followed by Neurosurgery (16/92, 17.39%), Orthopedics (7/92, 7.61%) (Figure 1B). Regarding specimen types, the majority of isolates originated from sputum (72/92, 78.26%), while a minority were isolated from blood (5/92, 5.43%) (Figure 1C). To investigate the molecular characteristics of ColR-hv-CRKP, whole-genome sequencing was performed on all 92 isolates. MLST analysis revealed that the vast majority (90/92, 97.83%) belonged to sequence type ST11, the predominant high-risk clone associated with carbapenem resistance in China, while the remaining two isolates were classified as ST15. Capsular typing identified two major capsule loci: KL64 (52/92, 56.52%) and KL25 (36/92, 39.13%), with two isolates each belonging to KL107 (2/92, 2.17%) and KL2 (2/92, 2.17%). Virulence gene profiling showed that all isolates carried the hypermucoviscosity-associated regulator rmpA2 and the aerobactin biosynthetic cluster (iucABCD) along with its receptor gene iutA. Kleborate analysis indicated that all isolates belonged to the iuc1 lineage, and the majority exhibited virulence scores of ≥3 (90/92, 97.83%), with both ST15 isolates displaying the highest score of 5. Notably, 90 out of 92 isolates (97.83%) did not carry the salmochelin operon (iroBCDN), whereas both ST15 isolates harbored this operon. Plasmid replicon typing revealed that IncFII (pHN7A8) was the dominant plasmid backbone, detected in most isolates (90/92, 97.83%).

Figure 1 Distribution of 92 colistin‑resistant hypervirulent carbapenem‑resistant K. pneumoniae (ColR‑hv‑CRKP) isolates. (A) Annual number of ColR‑hv‑CRKP isolates collected from 2021 to 2024. (B) Departmental distribution of the ColR‑hv‑CRKP isolates. (C) Specimen‑type distribution of the ColR‑hv‑CRKP isolates. Abbreviation: BALF, bronchoalveolar lavage fluid.

Phylogenetic Analysis of ColR-Hv-CRKP Isolates

To explore the genetic relationships among the ColR-hv-CRKP isolates, a core‑genome phylogenetic tree was generated based on SNP alignments (Figure 2). Notably, ST11-KL25 clade comprised all 36 isolates of this capsule type and formed a highly homogeneous monophyletic cluster with minimal intra-clade diversity and strong bootstrap support. These KL25 isolates were predominantly recovered from the ICU (22/36, 61.11%) and neurosurgery (6/36, 16.67%) wards and were mainly collected during 2023. In contrast, the KL64 isolates exhibited greater genetic heterogeneity and were distributed across multiple subclades. The two ST15 isolates carrying the KL2 capsule locus were phylogenetically distinct from the ST11-KL64‑dominated lineages and formed a small, independent branch on the periphery of the tree.

Figure 2 Maximum-likelihood phylogenetic tree based on core-genome SNPs of 92 ColR-hv-CRKP isolates, with metadata annotations indicating capsular type (KL type), specimen source, hospital department and year of isolation. Bootstrap support values are represented by the size of circles at branch nodes. Red branches represent KL25 isolates.

To investigate genetic relatedness of ColR-hv-CRKP within the hospital environment, we performed core-genome SNP analysis and visualized pairwise SNP distances in heatmaps stratified by capsular type (Figure 3). Using a conservative threshold of ≤16 SNPs to define putative recent transmission events, several clusters of highly related isolates were identified in both groups. In the KL64/KL107 group (Figure 3A), multiple closely related isolates exhibited minimal SNP divergence. Notably, the KL25 group showed tighter genetic clustering (Figure 3B). A particularly homogeneous cluster of 11 isolates differed by no more than 11 SNPs (Figure 3B), suggesting recent hospital-associated spread from a common source. Two additional clusters containing 11 and 7 isolates, respectively, also exhibited SNP distances within the ≤13 SNP range. Overall, KL25 isolates exhibited significantly smaller pairwise SNP distances than KL64/KL107 isolates (median 15 [IQR 11–19] vs 24 [IQR 10–60]; Mann–Whitney U-test, P < 0.0001) (Supplementary Figure S1).

Figure 3 SNP analysis of ColR-hv-CRKP isolates. Heatmaps showing pairwise SNP distances between (A) ST11-KL64/KL107 isolates and (B) ST11-KL25 isolates. The color scale represents the number of SNP differences.

Resistance Phenotypes and Resistance Gene Profiles Among ColR‑hv‑CRKP Isolates

Antimicrobial susceptibility testing of the 92 ColR‑hv‑CRKP isolates demonstrated universal resistance to a wide spectrum of β‑lactam antibiotics, including cefotaxime (CTX), ceftriaxone (CRO), ceftazidime (CAZ), and cefazolin (CZO) (Figure 4). In addition, all isolates exhibited resistance to β‑lactam/β‑lactamase inhibitor combinations ticarcillin/clavulanate (TCC) and piperacillin/tazobactam (TZP), and to the monobactam aztreonam (ATM). Nearly all isolates (91/92, 98.91%) were resistant to the fluoroquinolones levofloxacin (LEV) and ciprofloxacin (CIP). Aminoglycoside resistance was also prevalent, with 84 out of 92 isolates (91.30%) resistant to tobramycin (TOB) and amikacin (AMK). Most isolates were resistant to tetracyclines (doxycycline and minocycline); however, susceptibility to tigecycline (TGC) was retained in 76 isolates (82.61%). Resistance to ceftazidime/avibactam (CZA) was observed in 23 isolates (25.00%). The MIC values, along with MIC₅₀ and MIC₉₀, for all tested antimicrobial agents are provided in Supplementary Table S2.

Figure 4 Antibiotic susceptibility profiles and resistance gene distribution among 92 ColR-hv-CRKP isolates. The left panel shows antimicrobial susceptibility testing results for various antibiotics. The right panel displays the presence or absence of resistance genes detected by whole-genome sequencing. Abbreviations: Antibiotic abbreviations: IPM, imipenem; MEM, meropenem; COL, colistin; CTX, cefotaxime; CRO, ceftriaxone; CAZ, ceftazidime; CZO, cefazolin; LEV, levofloxacin; CIP, ciprofloxacin; ATM, aztreonam; TCC, ticarcillin/clavulanate; TZP, piperacillin/tazobactam; CZA, ceftazidime/avibactam; TOB, tobramycin; AMK, amikacin; DOX, doxycycline; MNO, minocycline; TGC, tigecycline; SXT, trimethoprim/sulfamethoxazole.

Whole‑genome sequencing revealed that carbapenem resistance was primarily mediated by bla_KPC-2, detected in 98.91% (91/92) of isolates; one isolate carried bla_KPC-71, and another isolate co-harbored bla_NDM-5 (Figure 4). A variety of other β-lactamase genes were prevalent, including bla_CTX-M-65 (94.57%), bla_SHV-12 (78.26%), bla_SHV-182 (19.57%), bla_TEM-1B (92.39%), and bla_CTX-M-3 (3.26%). Genes conferring resistance to other antimicrobial classes were also widespread. The aminoglycoside resistance gene rmtB was present in 92.39% of isolates, while aadA2 and aadA16 were detected in 81.52% and 3.26%, respectively. The quinolone resistance gene qnrS1 was found in 95.65% of isolates, and aac(6′)-Ib-cr in 3.26%. The tetracycline resistance gene was was present in tet(A) (94.57%). The chloramphenicol resistance gene catA2 was present in 53.26% of isolates, and the floR gene in 3.26%. Trimethoprim resistance genes dfrA14 and dfrA27 were present in 91.30% and 3.26% of isolates, respectively. The sulfonamide resistance gene sul2 was found in 92.39%, and sul1 in 3.26% of isolates. The macrolide resistance gene mph(A) was detected in 3.26% of isolates. Notably, none of the 92 isolates carried colistin resistance genes (mcr‑1 through mcr‑9), suggesting that colistin resistance in these strains likely arose from chromosomal mutations rather than mcr‑type gene acquisition. Overall, the resistance genotypes identified showed high concordant with the phenotypic antimicrobial susceptibility profiles. The high prevalence of bla_KPC-2 and extended-spectrum β-lactamase genes aligned with the widespread resistance to β-lactam agents, while the frequent detection of rmtB, qnrS1, and tet(A) was corresponded to the observed resistance to aminoglycosides, fluoroquinolones, and tetracyclines, respectively.

MgrB and Two‑component Regulatory System Mutations in ColR‑hv‑CRKP Isolates

To elucidate the genetic mechanisms underlying colistin resistance in the 92 ColR‑hv‑CRKP isolates, mutations were analyzed in the two‑component regulatory systems PmrA/PmrB and PhoP/PhoQ, as well as in their negative regulator MgrB (Supplementary Table S1). As shown in Table 3, most ColR-hv-CRKP isolates (81/92, 88.04%) harbored colistin resistance‑associated mutations. No mutations in mgrB or the two-component regulatory systems were identified in the remaining 11 isolates (11.96%). mgrB alterations were identified in 33 isolates (35.87%), predominantly attributable to insertional inactivation by various insertion sequences (ISs), including ISKpn26 (14/92, 15.22%), IS903B (5/92, 5.43%), ISKpn14 (5/92, 5.43%), and ISkpn18 (3/92, 3.26%). In addition, nonsense mutations introducing premature stop codons were detected, comprising Gln30* (3/92, 3.26%) and Trp6* (2/92, 2.17%). A single missense substitution, Cys16Arg, was also identified in one isolate (1.09%). Mutations within the PmrA/PmrB system were comparatively frequent. PmrA substitutions were observed in 10 isolates (10.87%), including Gly53Ser (7/92, 7.61%), Gly53Val (2/92, 2.17%), and Gly53Ala (1/92, 1.09%). Multiple amino acid changes were detected in pmrB, with Thr157Pro being the most prevalent (19.57%), followed by Ser205Pro (8/92, 8.70%), Pro95Leu (4/92, 4.35%), Ser203Pro (4/92, 4.35%), Ala246Thr (2/92, 2.17%), Asp150His (2/92, 2.17%), Asp150Val (1/92, 1.09%), Asp150Tyr (1/92, 1.09%), and Ser270Arg (1/92, 1.09%). One frameshift mutation, Leu339fs, was also detected (1/92, 1.09%). In the phoQ gene, three missense variants were identified: Arg16Cys, Asn378Tyr, and Val450Asp, each detected in one isolate (1.09%). However, no mutations were detected in the phoP gene in any of the isolates.

Table 3 Mutations Identified in the Two-Component Regulatory Systems and Their Regulator (MgrB) Among 92 ColR‑hv‑CRKP Isolates

Discussion

The prevalence of colistin resistance among hv‑CRKP isolates in this study reached 28.22%, which is relatively higher than the rates reported in other regions, where it generally remains below 20%.^9,31 This disparity likely reflects regional variations in antimicrobial use and the selective pressure associated with increased colistin administration in our healthcare setting. Whole‑genome sequencing revealed that the vast majority of ColR‑hv‑CRKP isolates belonged to ST11, the predominant high‑risk clone in China, representing more than 70% of clinical CRKP isolates nationwide.^32,33

Capsular typing identified KL64 and KL25 as the dominant capsule loci among the ColR‑hv‑CRKP isolates. Although KL64 replaced KL47 as the prevailing capsular type in China after 2016, the relatively high proportion of KL25 isolates (39.13%) observed in our study may indicate an emerging subclonal transition.³⁴ Distinct capsular types are known to confer varying abilities in immune evasion, environmental persistence, and transmissibility.³⁵ SNP analysis revealed several conserved clusters among KL25 isolates, suggesting possible recent patient‑to‑patient transmission. Moreover, the narrower SNP distances within the KL25 lineage compared with KL64 may reflect a recent clonal expansion and potentially higher transmissibility in the hospital setting. Consistent with our genomic observations, Wang et al demonstrated that ST11-KL25 strains display enhanced resistance to phagocytosis, superior biofilm formation, and improved survival under antibiotic pressure, features that may explain their epidemiological success.³⁵ However, the absence of environmental and healthcare-worker sampling in this study limits our ability to definitively determine the transmission routes, and alternative pathways cannot be excluded. The clonal expansion of ST11‑KL64 and ST11‑KL25 ColR‑hv‑CRKP strains highlights the urgent need for comprehensive infection‑control strategies, including molecular surveillance for early detection, strict adherence to contact‑precaution protocols, prudent use and reduced duration of invasive procedures, and implementation of reinforced protective measures for high‑risk patient populations.³⁶

Genomic analysis revealed that colistin resistance in our ColR-hv-CRKP isolates was primarily mediated by chromosomal mutations rather than mcr genes. Mutations in the two-component regulatory systems PmrA/B and PhoP/Q, along with disruptions in the regulator mgrB, were identified in 88.04% of isolates. The predominance of the PmrB Thr157Pro substitution (19.57%) is particularly notable and consistent with previous studies identifying this mutation as a recurrent mutation associated with colistin resistance.^37–39 The PmrB alteration modifies lipid A structure by upregulating the pmrHFIJKLM operon, thereby reducing the electrostatic interaction between polymyxins and the bacterial outer membrane.^37,40 Another common mechanism identified was ISKpn26‑mediated mgrB gene inactivation (15.22% of isolates). ISKpn26 was the most prevalent insertion sequence in our collection and has been experimentally demonstrated to disrupt mgrB and thereby confer colistin resistance.^41,42 Moreover, ISKpn26‑carrying plasmids often harbor carbapenemase genes, contributing to the development of extensively drug‑resistant K. pneumoniae phenotypes.⁴³ In addition, ISKpn26-mediated mgrB disruption, mgrB Gln30*pmrA Gly53Ser, pmrA Gly53Val, pmrA Gly53Ala, pmrB Thr157Pro, pmrB Ser203Pro, pmrB Asp150His, and pmrB Asp150Tyr have been experimentally validated to contribute to colistin resistance.^{37,42,44–49} Several mutations identified in this study, including mgrB Trp6*mgrB Cys16Arg, phoQ Arg16Cys, phoQ Val450Asp, phoQ Asn378Tyr, pmrB Pro95Leu, pmrB Asp150Val, pmrB Ser270Arg, pmrB Ala246Thr, and pmrB Leu339fs require further experimental validation to clarify their specific contributions to colistin resistance. No mcr genes were detected in any of the isolates, a finding that diverges from several international reports.^50,51 This pattern likely reflects the epidemiological context in China, where mcr-mediated resistance may be less prevalent due to the ban on colistin as a growth promoter in animal husbandry since 2017.⁵² The low prevalence of mcr gene in hv-CRKP may be attributed to the associated fitness cost and reduced stability of the plasmid after carrying mcr.⁵³ The diversity of chromosomal mutations observed indicates that multiple evolutionary routes to colistin resistance may exist within the ST11 lineage under similar selective pressures.

Multivariate analysis identified prior colistin exposure as the strongest independent predictor of ColR-hv-CRKP infection (OR = 4.576, P < 0.001). This finding is consistent with previous case–control studies demonstrating that colistin exposure independently increases the likelihood of the in vivo emergence of colistin resistance in CRKP and other Enterobacteriaceae.^54,55 Although colistin use imposes substantial selective pressure and should not be underestimated as a driver of resistance, horizontal transmission of ColR‑CRKP may also contribute to the increasing detection of these isolates.⁵⁶ These observations suggest that colistin resistance may develops through two complementary pathways: selection during polymyxin therapy and transmission between patients, underscoring the importance of both antimicrobial stewardship and rigorous infection-control measures. Isolation from sputum specimens was also independently associated with ColR-hv-CRKP infection (OR = 2.444, P = 0.003), which may reflect the antimicrobial selective pressure associated with pulmonary infections rather than a direct causal relationship between respiratory infection itself and colistin resistance. This association may be partly explained by the suboptimal alveolar concentrations achieved by intravenously administered polymyxins, which can facilitate the emergence of colistin-resistant subpopulations during prolonged therapy. Taken together, these observations support the important role of antimicrobial selective pressure in the development of colistin resistance. Accordingly, these results underscore the importance of targeted antimicrobial stewardship aimed at prudent colistin use, as well as strengthened surveillance of high-risk patients to enable early detection and timely infection-control interventions.⁵⁷

The universal presence of key virulence factors, including the aerobactin biosynthetic cluster (iucABCD/iutA) and the hypermucoidy regulator rmpA2, confirms the hypervirulent nature of ColR‑hv‑CRKP isolates. The aerobactin‑mediated iron acquisition system, almost universally detected in this collection, is a critical virulence determinant that promotes bacterial growth under iron‑limited conditions and facilitates invasive infection. The absence of the salmochelin operon (iroBCDN) in ST11-KL64/25 suggests that these isolates may represent a distinct evolutionary lineage from the classical hypervirulent K1 and K2 serotypes, which typically harbor both siderophore systems.⁵⁸ The convergence of extensive drug resistance and hypervirulence therefore represents a particularly alarming evolutionary trend in K. pneumoniae, potentially enabling these strains to evade antimicrobial therapy while causing more severe and tissue‑invasive infections.⁵⁹

This study has several limitations. First, the relatively small number of colistin-resistant isolates and the lack of hospital environmental samples limited our ability to fully elucidate the transmission dynamics. Second, although whole-genome sequencing provided valuable insights into the genetic basis of colistin resistance, future studies should incorporate experimental validation to confirm the functional effects of the identified mutations. Third, colistin resistance in 11 isolates could not be explained by currently recognized mechanisms, highlighting an important knowledge gap and suggesting the existence of additional or unconventional resistance pathways. Fourth, clinical outcomes were not systematically assessed in this study; therefore, the potential association between colistin resistance and patient prognosis could not be determined.

Conclusions

This study demonstrates that ColR‑hv‑CRKP is primarily driven by chromosomal mutations, particularly in mgrB and the two‑component regulatory systems, rather than through acquisition of mcr genes. The predominance of the ST11‑KL64 and ST11‑KL25 lineages, along with evidence of recent clonal expansion and hospital transmission, highlights their epidemic potential. Prior colistin exposure was the strongest independent predictor of ColR‑hv‑CRKP infection, emphasizing the selective pressure exerted by colistin use. These findings emphasize the critical importance of strengthened antimicrobial stewardship and infection-control measures, as well as continued genomic surveillance and functional validation of resistance mechanisms, to mitigate the further dissemination of these high-risk clones.