Reversible Evolution of Ceftazidime-Avibactam Resistance Driven by bla_KPC-71 and Aerobactin Loss in ST11-KL64 Hypervirulent Klebsiella Pneumoniae

Introduction

Klebsiella pneumoniae is an important opportunistic pathogen that colonizes the human gastrointestinal and respiratory tracts, as well as hospital environments.^1,2 It is a leading cause of severe infections such as pneumonia, urinary tract infections, and bloodstream infections.³ The widespread and often inappropriate use of antibiotics has contributed to the emergence and rapid dissemination of drug-resistant K. pneumoniae strains, particularly carbapenem-resistant K. pneumoniae (CRKP) carrying K. pneumoniae carbapenemase (KPC), which has become a major global public health threat.⁴

Among the various mechanisms of carbapenem resistance, KPC-type β-lactamases are the most widely disseminated globally, with KPC-2 and KPC-3 being the predominant variants.⁵ Ceftazidime-avibactam (CZA), a novel combination of a third-generation cephalosporin and a non-β-lactam β-lactamase inhibitor, has been introduced into clinical practice and has demonstrated potent activity against most KPC-producing organisms.^6,7 With increasing clinical use of CZA, resistance has frequently arisen through point mutations within the bla_KPC gene, particularly in regions affecting the active site or regulatory domains of the enzyme, which alter its affinity for substrates and inhibitors.^8,9 Several CZA-resistant KPC variants have been identified, such as KPC-8, KPC-31, and KPC-33, which evade avibactam inhibition by modifying the conformation of the Ω-loop or catalytic pocket.^10–12 Nevertheless, the in vivo evolution of KPC variants during clinical infection remains poorly characterized, particularly with respect to associated modifications in virulence, representing an important gap in current understanding.

Importantly, increasing virulence among CRKP strains has raised significant concern in recent years.¹³ The ST11-KL64 clone, in particular, is frequently associated with hypervirulence genes such as iucABCD, iutA, and rmpA2, resulting in so-called “superbugs” that combine high levels of resistance and virulence, posing serious challenges to clinical management.¹⁴ However, it remains unclear whether CRKP can modulate virulence under strong antibiotic selection pressure, for example by attenuating pathogenic potential to preserve resistance. The interplay between resistance evolution, virulence dynamics, and plasmid adaptation therefore warrants further investigation.

In this study, we conducted a longitudinal genomic analysis of seven CRKP isolates collected from a single hospitalized patient over a 147-day clinical course. Within‑host longitudinal investigations offer a valuable opportunity to directly observe the adaptive evolution of pathogens in real time, capturing the dynamic interplay between antibiotic exposure, resistance development, and virulence modulation that is often overlooked in cross‑sectional surveillance. To our knowledge, this is the first clinical observation of the in vivo evolution of a KPC-2 variant, KPC-71. Our findings revealed that the emergence of KPC-71 was closely associated with CZA exposure and was accompanied by distinct changes in antimicrobial susceptibility and virulence. Through whole-genome sequencing, phylogenetic analysis, plasmid profiling, and functional validation, we elucidated the adaptive evolutionary mechanisms of CRKP during antibiotic treatment.

Materials and Methods

Isolate Identification and Culture

A total of seven K. pneumoniae isolates were sequentially collected from a single hospitalized patient over a 147-day period at the Department of Clinical Laboratory, First Affiliated Hospital of Nanchang University (Jiangxi, China). Clinical specimens were obtained from sputum and bronchoalveolar lavage fluid (BALF) during routine diagnostic procedures. All samples were processed in accordance with standard microbiological protocols. Initial bacterial isolation was performed using blood agar. Colonies with typical Klebsiella morphology were subcultured for further testing. Species-level identification was conducted using matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) (Bruker Daltonics, Germany).

Antimicrobial Susceptibility Testing

Antimicrobial susceptibility testing of the seven CRKP isolates was performed using the broth microdilution method with commercially prepared, pre-loaded microdilution panels (Autobio, China), following the manufacturer’s instructions. Minimum inhibitory concentrations (MICs) were determined for a wide range of antimicrobial agents, including meropenem, imipenem, fosfomycin, ceftazidime, cefuroxime, cefepime, ceftazidime-avibactam, cefoperazone-sulbactam, piperacillin-tazobactam, amikacin, gentamicin, levofloxacin, moxifloxacin, tigecycline, trimethoprim-sulfamethoxazole, and polymyxin B. Interpretation of MIC results was based on the Clinical and Laboratory Standards Institute (CLSI) guidelines (2024).¹⁵ For tigecycline, interpretive criteria were applied according to the US Food and Drug Administration (FDA) breakpoints, and for polymyxin B, European Committee on Antimicrobial Susceptibility Testing (EUCAST) breakpoints were used. Escherichia coli ATCC 25922 and K. pneumoniae ATCC 13883 was used as a quality control strain throughout the testing process. Each susceptibility test was performed in triplicate to ensure the reproducibility and reliability of the results.

Whole-Genome Sequencing and Bioinformatics Analysis

Genomic DNA was extracted from each isolate using a commercial DNA extraction kit (Qiagen, Germany) following to the manufacturer’s instructions. Whole-genome sequencing (WGS) was performed for all seven CRKP isolates using the Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA) with paired-end 150 bp reads. In addition, three representative isolates (CRKP36, CRKP55, and CRKP103) underwent long-read sequencing on the PacBio Sequel platform (Pacific Biosciences, Menlo Park, CA, USA). Genome assemblies were generated using Unicycler v0.5.0,¹⁶ with Illumina short‑read assemblies for all isolates and hybrid assemblies for the three representative strains (CRKP36, CRKP55, and CRKP103). Antimicrobial resistance genes, virulence determinants, plasmid replicon types, serotype predictions, and multilocus sequence types (MLST) were identified using the Kleborate v3.1.3, VRprofile2, and PlasmidFinder v2.1 databases,^17–20 applying identity ≥ 90% and coverage ≥ 90% as thresholds. Circular plasmid maps were generated using BRIG (BLAST Ring Image Generator) to visualize genome architecture and plasmid content.²¹ To investigate the structural differences between KPC variants, amino acid sequence alignment and secondary structure prediction of KPC-2 and KPC-71 were performed using ESPript 3.0.²² For phylogenetic analysis, a core genome alignment was constructed using CRKP isolates with K. pneumoniae HS11286 (RefSeq: GCF_000240185.1) as the reference genome, using snippy v.4.6.0.²³ Recombination regions were identified and removed with Gubbins v3.2.1.²⁴ A maximum-likelihood phylogenetic tree was then inferred using FastTree v2.1.11 under the GTR + CAT substitution model,²⁵ and visualized using the Interactive Tree of Life (iTOL) platform for annotation and presentation.²⁶

Capsule Quantification

Uronic acid content was quantified following a previously described protocol.²⁷ Overnight bacterial cultures were grown in Luria-Bertani (LB) broth at 37 °C with shaking and subsequently diluted 1:100 into fresh LB medium. The cultures were incubated at 37 °C for 6 hours to reach exponential phase. A 500 μL aliquot of each culture was mixed with 100 μL of 1% Zwittergent 3–12 detergent (in 100 mM citric acid buffer, pH 2.0) and incubated at 50 °C for 20 minutes to lyse bacterial cells and release capsular material. The mixture was then centrifuged at 13,000 × g for 5 minutes, and 300 μL of the resulting supernatant was transferred to a new tube and mixed with 1.2 mL of absolute ethanol. After incubation at room temperature for 5 minutes, samples were centrifuged again at 13,000 × g for 5 minutes to pellet the capsular material. The resulting pellet was air-dried and resuspended in 200 μL of sterile distilled water. To quantify uronic acid content, 1.2 mL of sodium tetraborate solution (12.5 mM in concentrated sulfuric acid) was added to each sample, followed by incubation at 100 °C for 5 minutes. Samples were then immediately transferred to ice and cooled for at least 10 minutes. After cooling, 20 μL of m-hydroxydiphenyl reagent was added, and the mixture was incubated at room temperature for 5 minutes. Absorbance was measured at 520 nm using a microplate reader, and uronic acid concentrations were calculated from a standard curve prepared with D-glucuronic acid as the reference compound.

Quantitative Siderophore Production Assay

Quantitative chrome azurol S (CAS) agar assay was performed as previously described.²⁸ This assay measures the ability of bacteria to chelate iron from a dye–iron complex, resulting in a visible color change. Briefly, bacterial strains were cultured in iron-limited M9 minimal medium until mid-logarithmic phase (OD₆₀₀ ≈ 0.6). A 1 μL aliquot of the culture was spotted onto freshly prepared CAS agar plates, which contain chrome azurol S dye, FeCl₃, and hexadecyltrimethylammonium bromide (HDTMA) to form a blue iron–dye complex. Plates were incubated at 37 °C for 24 hours under aerobic conditions. Following incubation, the formation of a distinct orange halo around the bacterial colony was recorded. The diameter of the orange zone (in millimeters) was measured using a digital caliper and used as a semi-quantitative indicator of siderophore activity. Each assay was performed in triplicate, and the results were reported as mean ± standard deviation.

Serum Resistance Assay

Bacterial cultures were grown in LB broth to mid-logarithmic phase (OD₆₀₀ ≈ 0.6) and adjusted to a final concentration of 1.5 × 10⁶ CFU/mL in phosphate-buffered saline (PBS). An aliquot of 100 μL bacterial suspension was mixed with 300 μL of pooled normal human serum (NHS) obtained from healthy adult volunteers who had provided informed consent. The mixture was incubated at 37 °C for 2 hours under gentle shaking to allow interaction between bacteria and complement proteins. Following incubation, samples were serially diluted in PBS and plated onto LB agar to determine the number of viable bacteria by colony counting. Plates were incubated overnight at 37 °C, and CFU counts were recorded. Each assay was performed in biological triplicates.

Galleria Mellonella Infection Model

G. mellonella larvae were obtained from a commercial supplier (Tianjin Huiyude Biotech, Tianjin, China) and stored in the dark at 4 °C prior to use. Only healthy larvae weighing approximately 250–300 mg, with uniform creamy coloration and no signs of melanization, discoloration, or physical damage, were included in the experiments. Bacterial suspensions were grown to mid-log phase, washed twice in PBS, and adjusted to a final concentration of 1×10⁶ CFU/mL. Each larva was injected with 10 μL of bacterial suspension through the last left proleg using a fine Hamilton syringe. Control groups were injected with 10 μL of sterile PBS. A minimum of 30 larvae per group were used, divided into three Petri dishes (10 larvae per dish) to account for potential environmental variation. Infected larvae were incubated at 37 °C in the dark, and survival was monitored every 12 hours over a 72-hour period. Larvae were considered dead when they failed to respond to gentle tactile stimulation. Survival data were analyzed using Kaplan-Meier survival curves, and statistical differences between groups were evaluated with the log‑rank (Mantel-Cox) test. Two well-characterized reference strains were included: NTUH-K2044, a hypervirulent strain used as a positive control, and ATCC 700603, a classical strain used as a negative control. All experiments were repeated at least three times.

Construction of pDK6-KPC Plasmids

Plasmid constructs carrying different bla_KPC variants were generated to evaluate their functional impact. The complete bla_KPC-2 and bla_KPC-71 genes, including their native promoter regions, were amplified using PrimeSTAR DNA Polymerase (Takara, Japan; Cat. No. R045A).The complete bla_KPC-2 and bla_KPC-71 genes, including their native promoter regions, were amplified with PrimeSTAR DNA Polymerase (Takara, Japan; Cat. No. R045A) using the primers 5′-ACAGGAAACAGAATTCGAGCTCTTGTTCAGGAAACGCGTGTA-3′ and 5′-AGCCAAGCTTGCATGCCTGCAGGGTGGGCCAATAGATGATTTTC-3′. The PCR products were purified and ligated into the pDK6 expression vector using the NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs, USA; Cat. No. E2621), following the manufacturer’s instructions. The recombinant plasmids (pDK6-KPC-2 and pDK6-KPC-71) were introduced into chemically competent E. coli DH5α cells via heat-shock transformation. Transformants were selected on LB agar plates supplemented with appropriate antibiotics, and positive clones were verified by colony PCR targeting the bla_KPC gene followed by Sanger sequencing. Confirmed plasmids were then extracted from E. coli DH5α and subsequently introduced into K. pneumoniae ATCC 13883 via electroporation using a Gene Pulser Xcell system (Bio-Rad, USA). Transformants were selected and further validated by plasmid profiling and PCR to ensure successful introduction of the KPC variants.

Statistical Analysis

All statistical analyses were performed using GraphPad Prism version 9.0 (GraphPad Software Inc., San Diego, CA, USA). Data were expressed as mean ± standard deviation (SD) from at least three independent biological replicates. Comparisons between multiple groups were conducted using one-way analysis of variance (ANOVA). A p-value < 0.05 was considered statistically significant.

Result

Emergence and Evolution of a KPC-2 Variant (KPC-71) in a Hospitalized Patient

We collected seven CRKP isolates longitudinally from a single 65‑year‑old male patient who was admitted to the intensive care unit with severe pneumonia and sepsis secondary to chronic obstructive pulmonary disease. The isolates were recovered from respiratory specimens, including sputum and bronchoalveolar lavage fluid, during a 147-day hospitalization period. The patient initially received empirical piperacillin‑tazobactam therapy during the first 12 days of hospitalization for ventilator‑associated pneumonia. After the isolation of CRKP, ceftazidime‑avibactam (CZA) was introduced on day 13 and continued for 17 days. Meropenem, ceftazidime, cefoperazone‑sulbactam, and tigecycline were administered intermittently during the later course of hospitalization in response to recurrent infection episodes (Figure 1A). Each isolate underwent whole-genome sequencing, resistance gene profiling, and antimicrobial susceptibility testing to characterize the molecular trajectory of resistance. The earliest isolate, CRKP17, recovered on day 13 of hospitalization, harbored the bla_KPC-2 gene and exhibited high-level resistance to carbapenems (imipenem/meropenem) while remaining susceptible to CZA (Table 1).

Table 1 Antimicrobial Susceptibility Profiles of 7 CRKP Strains in This Study

Figure 1 Longitudinal evolution and genomic characterization of CRKP isolates from a hospitalized patient. (A) Timeline of seven CRKP isolates collected over a 147-day hospitalization. (B) Phylogenetic analysis and genomic features of the seven CRKP isolates. The maximum-likelihood phylogenetic tree was constructed from core genome SNPs, with K. pneumoniae HS11286 as the reference.

Following the initiation of CZA therapy, a genotypic and phenotypic shift was observed. On hospital day 30, CRKP22 was isolated and found to harbor a KPC variant, KPC-71, which conferred resistance to CZA (MIC >16/4 μg/mL) and reduced carbapenem MICs (imipenem and meropenem, MIC = 2 μg/mL). Interestingly, subsequent isolates CRKP24 (day 42) and CRKP36 (day 61) reverted back to KPC-2, accompanied by restoration of CZA susceptibility (MIC = 4/4 μg/mL), but with a return to high-level carbapenem resistance (MICs >16 μg/mL). Later in the treatment course, with the reintroduction of CZA, three additional KPC-71-producing strains, CRKP55 (day 93), CRKP86 (day 117), and CRKP103 (day 140), were recovered. These isolates again exhibited CZA resistance (MIC >16/4 μg/mL) and intermediate susceptibility to imipenem and meropenem (MIC = 2 μg/mL).

All seven isolates displayed multidrug-resistant phenotypes, with resistance to all tested β-lactam/β-lactamase inhibitor combinations (including piperacillin-tazobactam and cefoperazone-sulbactam), extended-spectrum cephalosporins (ceftazidime, cefepime, cefuroxime), aminoglycosides (amikacin and gentamicin), fluoroquinolones (levofloxacin and moxifloxacin), trimethoprim-sulfamethoxazole, polymyxin B, and fosfomycin. Tigecycline remained the sole antimicrobial agent to which all isolates were consistently susceptible (Table 1).

Genomic Characteristics and Phylogenetic Analysis of CRKP Strains

To further explore the genetic relatedness and evolutionary dynamics of the seven CRKP isolates, a phylogenetic analysis based on core genome SNPs was performed (Figure 1B). Genome assembly statistics for these isolates are summarized in Supplementary Table S1. All strains belonged to sequence type ST11 and capsular type KL64. The three KPC-2-producing strains (CRKP17, CRKP24, CRKP36) were genetically closer to the reference strain HS11286. The four KPC-71-producing isolates (CRKP22, CRKP55, CRKP86, CRKP103) formed a separate cluster, suggesting clonal expansion of a KPC-71-producing sublineage. These strains exhibited extremely low SNP differences, ranging from 2 to 4.

Plasmid replicon typing revealed that all isolates harbored multiple plasmid types, including ColRNAI, IncFII (pCRY), IncFII (pHN7A8), and IncHI1B (pNDM-MAR). Resistance gene analysis showed that all isolates carried a broad spectrum of antibiotic resistance genes, including bla_CTX-M-65, bla_TEM-1B, catA2, dfrA14, fosA6, qnrS1, rmtB, and tet(A). All strains harbored either bla_KPC-2 or the variant bla_KPC-71, while the presence of bla_SHV-12 and bla_SHV-11 varied among isolates. The amino acid alignment and predicted secondary structure modeling (Figure 2) revealed that KPC-71 differs from the wild-type KPC-2 carbapenemase by an insertion of a serine residue between positions 181 and 182. This structural alteration may influence the binding affinity to β-lactam–β-lactamase inhibitor combinations, potentially contributing to ceftazidime-avibactam resistance. Importantly, most isolates retained key virulence-associated genes commonly linked to hypervirulent K. pneumoniae, including the aerobactin siderophore system (iucABCD and iutA), mucoid phenotype regulator (rmpA2), and yersiniabactin biosynthetic cluster (ybt genes). In contrast, two isolates, CRKP86 and CRKP103, did not carry iucABCD or iutA.

Figure 2 Comparative amino acid sequence alignment and secondary structure prediction of KPC-2 and KPC-71.

Chromosomal and Plasmid-Mediated Antimicrobial Resistance in CRKP36, CRKP55, and CRKP103 Strains

Whole-genome and plasmid profiling of CRKP36, CRKP55, and CRKP103 revealed a conserved chromosomal backbone (~5.46 Mb, GC content ~57.4–57.8%) and five plasmids per strain, collectively encoding a broad spectrum of resistance and virulence genes (Table 2). All three isolates harbored key chromosomal resistance determinants, including bla_SHV-11, sul1, aadA2, and fosA6. Notably, CRKP103 also carried several additional chromosomal resistance genes, including sul2, catA2, tet(A), bla_LAP-2, and qnrS1. Plasmid comparisons revealed the presence of three major replicon types with highly conserved backbones across the strains, as visualized in the circular maps (Figure 3). A large IncHI1B-type plasmid (Figure 3A) was detected in all isolates and carried rmpA2 in all three strains, while encoding the aerobactin siderophore system (iucABCD/iutA) in CRKP36 and CRKP55 but not in CRKP103. This deletion is clearly visible in the outermost rings of the plasmid map, where the iucABCD gene cluster is absent in CRKP103-1 but present in CRKP36-1 and CRKP55-1 (Figure 3A).

Table 2 Genomic Characteristics of CRKP36, CRKP55, and CRKP103, Including Chromosomal and Plasmid-Borne Resistance and Virulence Determinants

Figure 3 Comparative plasmid maps showing structural conservation and gene content in CRKP36, CRKP55, and CRKP103. (A) Alignment of IncHI1B-type plasmids (~201 kb) with reference plasmid pKP1064WHY-1 (CP084706.1). (B) Alignment of IncFII(pHN7A8)/IncR plasmids (~135 kb) with reference plasmid pKP2722-2 (CP116905.1). (C) Alignment of IncFII(pCRY)-type plasmids (~87 kb in CRKP36/CRKP55, ~44 kb in CRKP103) with reference plasmid pKP37_4 (CP082756.1). Black circles mark the positions corresponding to the iucABCD/iutA locus.

The IncFII (pHN7A8)/IncR plasmid (Figure 3B) was nearly identical in all strains and carried bla_KPC-2 in CRKP36 or bla_KPC-71 in CRKP55 and CRKP103. This plasmid also encoded bla_SHV-12, bla_TEM-1B, bla_CTX-M-65, and rmtB, conferring resistance to carbapenems, extended-spectrum cephalosporins, and aminoglycosides. The structural conservation of this plasmid is evident in Figure 3B, where all three CRKP strains show high sequence identity with the reference plasmid pKP2722-2 (GenBank accession: CP116905.1). A third multidrug resistance plasmid (~87 kb in CRKP36 and CRKP55, ~44 kb in CRKP103; Figure 3C), identified as IncFII (pCRY)-type, contained dfrA14, catA2, qnrS1, bla_LAP-2, tet(A), and sul2 in CRKP36 and CRKP55, while CRKP103 harbored only dfrA14. The missing resistance genes (sul2, catA2, tet(A), bla_LAP-2, and qnrS1) were instead found integrated into the chromosome of CRKP103 (Table 2), explaining the reduced coverage observed in the outermost ring of the plasmid map. Additionally, all strains carried two small plasmids (~12 kb and ~5.5 kb) that did not encode any identifiable resistance or virulence determinants. Detailed comparison of the local genetic context surrounding the bla_KPC genes revealed a highly conserved IS26-tnpR-ISKpn27-bla_KPC-ISKpn6 cassette across all strains (Figure 4).

Figure 4 Comparison of the local genetic environment surrounding blaKPC-2 (CRKP36) and blaKPC-71 (CRKP55 and CRKP103) with the reference plasmid K. pneumoniae pHSKP43-2 (GenBank accession: CP100100.1).

CRKP103 Exhibited Lower Virulence Compare CRKP36

Virulence-associated phenotypes were assessed using CPS quantification, siderophore production, serum resistance, cytotoxicity, and in vivo pathogenicity in G. mellonella larvae. In the CPS quantification assay (Figure 5A), both CRKP36 and CRKP55 produced significantly higher uronic acid levels (374.33 ± 11.34 mg/L and 370.17 ± 20.61 mg/L, respectively) than the low-virulence control ATCC700603 (130.83 ± 13.24 mg/L), while remaining slightly below the hypervirulent reference NTUH‑K2044 (431.17 ± 14.59 mg/L). In contrast, CRKP103 exhibited markedly reduced capsule production (336.87 ± 16.04 mg/L). In the siderophore production assay (Figure 5B), CRKP103 displayed a smaller halo diameter (14.78 ± 0.79 mm) relative to CRKP36 (20.97 ± 0.94 mm) and CRKP55 (20.78 ± 0.86 mm), both of which were comparable to NTUH-K2044. In the serum resistance assay (Figure 5C), CRKP103 demonstrated a reduced survival rate in 50% human serum, with bacterial counts declining over time. In contrast, CRKP36 and CRKP55 maintained stable bacterial loads comparable to NTUH-K2044, indicating robust serum resistance. The LDH release assay (Figure 5D) revealed that CRKP103 induced significantly lower cytotoxicity toward A549 epithelial cells (38.28 ± 2.81%) compared with CRKP36 (73.14 ± 1.56%) and CRKP55 (73.48 ± 2.43%), which again did not differ significantly from NTUH‑K2044 (83.95 ± 3.08%). In the G. mellonella infection model (Figure 5E), larvae infected with CRKP103 exhibited significantly higher survival rates (50% at 72 h) than those infected with CRKP36 or CRKP55 (each 10%), further confirming the attenuated in vivo virulence of CRKP103. No significant difference in larval mortality was observed between CRKP36 and CRKP55. Collectively, these results demonstrate that CRKP103 exhibited consistent attenuation of virulence across all assays, in line with the loss of key virulence loci.

Figure 5 Phenotypic comparison of three CRKP isolates with different resistance and virulence profiles. (A) Capsule production measured by uronic acid quantification. (B) Siderophore production assessed by chrome azurol S assay. (C) Serum resistance. (D) Cytotoxicity evaluated by LDH release from A549 epithelial cells after 6 hours of infection. (E) In vivo virulence evaluated using the G. mellonella infection model. NTUH-K2044 and ATCC 700603 were used as hypervirulent and low-virulence controls, respectively. Data are presented as mean ± standard deviation. P < 0.01 (**), P < 0.0001 (**), and “ns” indicates no significant difference.

K. Pneumoniae Carrying bla_KPC-71 Exhibited High-Level Resistance to CZA

Compared to the wild-type KPC-2 carbapenemase, KPC-71 contains a disruptive in-frame insertion of three nucleotides (c.540_542dupATC), resulting in the duplication of a serine residue at position 181 (p. Ser181dup). This insertion lies immediately adjacent to the canonical Ω-loop (residues 164–179), a structural motif essential for substrate recognition and inhibitor binding. To assess the phenotypic impact of KPC-71, we cloned bla_KPC-71 into the pDK6 vector and transformed it into K. pneumoniae ATCC 13883. Isogenic strains carrying bla_KPC-2 or the empty vector were constructed in parallel (Table 3). Antimicrobial susceptibility testing revealed that KPC-71-producing strains exhibited high-level CZA resistance (MIC >16/4 μg/mL), in contrast to the KPC-2-producing counterpart (MIC = 4/4 μg/mL). Similarly, ceftazidime MICs increased from 4 μg/mL (KPC-2) to 64 μg/mL (KPC-71). Interestingly, KPC-71 expression led to a significant reduction in carbapenem MICs, with both imipenem and meropenem MICs decreasing to 2 μg/mL, compared to ≥16 μg/mL in the KPC-2 background. Furthermore, KPC-71 restored susceptibility to several β-lactam/β-lactamase inhibitor combinations previously ineffective in the KPC-2 background, including piperacillin-tazobactam and cefoperazone-sulbactam.

Table 3 Antimicrobial Susceptibility Profiles of K. Pneumoniae ATCC 13883 Strains Expressing KPC-2, KPC-71, or Empty Vector

Discussion

The evolutionary dynamics of CRKP remain a major challenge in antimicrobial resistance research.^29,30 Previous studies have reported that point mutations in bla_KPC can mediate CZA resistance, including the emergence of KPC-71.³¹ Our study expands this knowledge by providing the first longitudinal clinical evidence that the insertional variant KPC-71 can arise through a stepwise and reversible evolutionary process in vivo. By analyzing seven ST11-KL64 CRKP isolates collected over 147 days, we traced the adaptive trajectory from KPC-2 to KPC-71 and back, revealing how selective antibiotic exposure drives alternating phenotypic and genotypic states.

The emergence of KPC-71 coincided with CZA therapy and was associated with high-level resistance to CZA but reduced carbapenem MICs (Table 1). Previous work demonstrated that the Ser181 duplication in KPC‑71 diminishes hydrolytic activity while increasing substrate affinity for ceftazidime, thereby compromising the inhibitory effect of avibactam and potentially explaining the resistance phenotype observed in our isolates.³¹ Conversely, following CZA withdrawal, several isolates reverted to KPC-2, restoring carbapenem resistance and resensitization to CZA. This evolutionary pattern suggests repeated selection and re-emergence of the bla_KPC-71 variant in response to CZA exposure, and its reversion to bla_KPC-2 in the absence of pressure. This cyclical pattern underscores the capacity of CRKP to exploit a resistance trade-off, dynamically balancing susceptibility to carbapenems and β-lactam/β-lactamase inhibitor combinations.³² Similar trade-offs have been described for Ω-loop variants such as KPC-31 and KPC-50, which confer resistance to CZA but reduce carbapenem hydrolysis.^11,33 However, unlike these irreversible point mutations, our longitudinal analysis highlights that insertional variants such as KPC-71 can both emerge and regress in vivo, reflecting a more flexible evolutionary strategy.^34,35 These findings suggest that selective pressure from CZA monotherapy may inadvertently promote recurrent selection of resistant variants, highlighting the potential clinical risk of treatment failure and resistance cycling.

Both KPC-2 and KPC-71 were carried on highly conserved IncFII(pHN7A8)/IncR plasmids, embedded within a stable IS26-tnpR-ISKpn27-bla_KPC-ISKpn6 transposon, suggesting that the emergence of KPC-71 is driven by local mutational events rather than horizontal acquisition. However, adaptive plasmid remodeling was particularly evident in the final isolate CRKP103. This strain exhibited loss of the iucABCD/iutA aerobactin cluster from its IncHI1B plasmid, accompanied by reduced siderophore production, impaired serum resistance, diminished capsular polysaccharide synthesis, and attenuated virulence in the G. mellonella model, suggesting a potential reduction in hypervirulence potential despite maintaining extensive antimicrobial resistance. While the loss of the iucABCD/iutA cluster likely contributes to the observed virulence attenuation, other factors such as plasmid burden and associated metabolic costs may also play a role in shaping these phenotypic changes. These changes suggest that resistance evolution may be coupled with virulence attenuation, allowing CRKP to optimize survival while reducing fitness costs in the host environment.

Furthermore, truncation of the IncFII(pCRY) plasmid in CRKP103 (from ~87 kb to ~44 kb) was associated with the loss of accessory resistance genes (catA2, qnrS1, bla_LAP-2, tet(A), sul2). Interestingly, several of these genes were subsequently integrated into the chromosome (Table 2), reflecting a genomic reorganization strategy that may stabilize essential resistance determinants while reducing plasmid burden. The absence of iucABCD/iutA and partial loss of resistance elements in CRKP103 suggest progressive plasmid remodeling and potential virulence attenuation during host adaptation or antimicrobial pressure. The role of plasmid remodeling observed in CRKP103 also aligns with reports that resistance gene loss or chromosomal integration can stabilize bacterial fitness in the absence of strong selective pressure.^36,37 Zongo et al recently demonstrated that chromosomal integration of plasmid-borne resistance genes can be actively promoted by an antiplasmid system (ApsAB), which destabilizes costly plasmids while preserving ARGs in carbapenemase-producing E. coli.³⁶ Similarly, Abe et al showed that chromosomally integrated carbapenemase genes can act as long-term reservoirs and may be released back onto plasmids under carbapenem pressure, amplifying resistance and dissemination.^38,39 Our data are consistent with these observations, as multiple resistance genes originally plasmid-borne in CRKP36/55 were integrated into the chromosome of CRKP103. Together, this suggests that CRKP employs both plasmid remodeling and chromosomal assimilation to balance resistance stability, fitness costs, and transmissibility.

From a clinical perspective, these findings carry important implications. The “evolutionary pendulum” we observed underscores the risk that CZA monotherapy may foster cycles of resistance gain and loss, ultimately complicating treatment outcomes and promoting persistence of resistant subclones.⁴⁰ This echoes recent reports where CZA-resistant but carbapenem-susceptible isolates emerged during therapy, only to revert under carbapenem pressure.⁴¹ The decoupling of resistance and virulence traits, as demonstrated by CRKP55 and CRKP103, indicates that low-virulence but highly resistant clones may persist as hidden reservoirs in immunocompromised patients, facilitating silent dissemination of resistance determinants within healthcare settings.

Despite the detailed genomic and phenotypic analyses, several limitations should be acknowledged. First, this work was based on isolates from a single patient, and the limited sample size restricts the ability to generalize our findings to broader epidemiological settings. Second, although plasmid remodeling coincided with virulence attenuation, causality cannot be firmly established without validation in larger cohorts and vertebrate animal infection models, since the G. mellonella model has inherent limitations in recapitulating the pathogenicity of K. pneumoniae in humans. Finally, the impact of different therapeutic regimens on the observed evolutionary trajectory was not assessed and warrants further investigation.

In conclusion, this study provides the first clinical evidence for the reversible in vivo evolution of the insertional KPC-71 variant in CRKP under antibiotic pressure. We demonstrate how bla_KPC mutations and plasmid remodeling jointly shape a dynamic balance between resistance and virulence. The integration of genomic surveillance into clinical practice could inform antimicrobial stewardship and support precision-guided therapeutic decision-making to curb the emergence of resistant subclones. These findings extend current knowledge of adaptive evolution in multidrug-resistant pathogens and underscore the need for longitudinal resistance monitoring and precision-based therapeutic strategies, particularly during CZA treatment.