Introduction

Antibacterial resistance is one of the most pressing public health challenges worldwide, further exacerbated by the rapid spread of multidrug-resistant (MDR) and extensively drug-resistant (XDR) bacteria1. Among Gram-negative pathogens, carbapenem-resistant Enterobacterales (CRE) are of critical importance and have been classified as a priority group in the WHO’s list of critical pathogens2. The Enterobacterales order includes a diverse range of bacteria, such as Escherichia coli, Klebsiella spp., and Enterobacter spp. From an antimicrobial resistance perspective, these pathogens are particularly significant as they are major etiological agents of both community-acquired and healthcare-associated infection3. Carbapenems are considered last-resort antibiotics and have traditionally been used to treat infections caused by extended-spectrum β-lactamase (ESBL)-producing E. coli and Klebsiella pneumoniae. However, resistance to these agents has risen more rapidly than to other antibiotic classes, leading to increased mortality and the urgent need for alternative treatments4,5. Few antibiotics remain effective against CRE due to carbapenemase production, which hydrolyze most β-lactams, and the frequent coexistence of resistance mechanisms to other antibiotic classes, including fluoroquinolones and aminoglycosides. In response, new β-lactam/β-lactamase inhibitor (BL/BLI) combinations, such as ceftazidime/avibactam (CAZ/AVI), meropenem/vaborbactam, imipenem/relebactam and aztreonam/avibactam, have been introduced into clinical practice in recent years6,7,8. In this study, we characterized the whole genome of a KPC-producing K. pneumoniae and a concurrent E. coli isolate from a hospitalized patient who had received multiple antibiotic treatments, including CAZ/AVI. The two isolates shared three common resistance genes, including blaKPC-3. Furthermore, we sequenced the PBP3 of the E. coli strain to identify mutations and their potential correlation with resistance to BL/BLI therapies.

Results

Clinical case

A 77-year-old patient with chronic obstructive pulmonary disease (COPD), idiopathic dilated cardiomyopathy, heart failure with preserved ejection fraction, and paroxysmal atrial fibrillation on direct oral anticoagulants with an implantable cardioverter-defibrillator was admitted to Ischia Hospital due to an acute exacerbation of COPD. He was managed with inhaled glucocorticoids, bronchodilators, and a 12-day course of piperacillin/tazobactam (PIP/TAZO). Four months later, the patient was readmitted to the Surgery Department with sepsis secondary to acute calculous cholecystitis. Blood cultures performed during hospitalization yielded negative results, while a rectal swab tested positive for the blaKPC gene, and the swab culture yielded K. pneumoniae. The patient underwent cholecystectomy and was again treated with PIP/TAZO, leading to clinical stabilization. Five days after discharge, the patient was admitted to the Medicine Department with sepsis. Blood cultures from this admission tested positive for carbapenemase-producing E. coli. Antimicrobial susceptibility test results for E. coli are reported in Table 1. Given the patient’s clinical deterioration, empirical therapy was escalated to meropenem and linezolid. Following susceptibility testing, treatment was adjusted to CAZ/AVI (2.5 g 8-hourly) for seven days, leading to clinical improvement and normalization of laboratory findings. A subsequent rectal swab for CRE remained positive for the blaKPC gene, and swab culture again identified K. pneumoniae.

Table 1 Antimicrobial susceptibility testing result of Escherichiacoli strain isolated from the patient. MIC: minimum inhibitory concentration.
Full size table

Strains characterization

The multi-locus sequence typing (MLST) analysis showed the presence of Sequence Type (ST) 252, K locus KL51 and O locus O1ab for K. pneumoniae and ST131, phylogroup B2, serotype O25:H4 and fumC40/fimH30 for E. coli. Resistance genes analysis showed that K. pneumoniae and E. coli shared two resistance genes: blaKPC-3 and blaOXA-9. K. pneumoniae harbors one copy of the blaKPC-3 gene on Tn4401a, while E. coli carries two copies on two Tn4401a transposons.

PBP3 analysis

Alignment of the E. coli reads to the reference genome (RefSeq NC_002695.2) identified two mutations in the PBP3 gene (also known as ftsI), specifically A233T and I332V.

Analysis of antimicrobial resistance-associated gene mutations

Whole-genome variant calling of Illumina sequencing revealed several mutations in antimicrobial resistance-associated genes in both isolates. In E. coli, we detected two missense mutations in acrB (p.His596Asn and p.Ser1043Asn), one in acrA (p.Thr104Ala), two in marR (p.Gly103Ser and p.Tyr137His ) and one in marA (p.Ser127Asn). In K. pneumoniae, two missense mutations were identified in acrA (p.Val313Ile and p.Ala188Thr) (Table 2).

Table 2 Summary of missense mutations identified in efflux-related antimicrobial resistance (AMR) genes in E. coli and K. pneumoniae. For each variant, the gene, amino acid change, and a brief annotation are reported.
Full size table

Study of mobile genetic elements

PlasmidFinder identified several plasmids, including an IncFII(K) plasmid carrying the blaKPC gene. Assembly and contig analysis using MOB_suite detected a transposon harboring the blaKPC gene within a plasmid homologous to pECAZ147_KPC from E. coli strain Ecol_AZ147 (GenBank accession number CP018992) and pKPC from K. pneumoniae strain D1 (GenBank accession number CP043971). Comparative analysis revealed a high degree of homology (99.85%) between these plasmids. Due to the limitations of short-read sequencing with Illumina technology, we were unable to fully reconstruct the plasmid. Oxford Nanopore long-read sequencing allowed for full plasmid characterization in both strains. In K. pneumoniae, three plasmids were identified (ranging from 50 to 150 kb), while E. coli carried two plasmids of approximately 60 kb and 120 kb. In both cases, blaKPC-3 was located on ~ 120 kb plasmids, which also included replicase genes, confirming their co-localization on the same contig. MOB-suite analysis assigned these plasmids to the IncFII incompatibility group, frequently linked to antimicrobial resistance, and detected MOBP-type relaxases, supporting their mobilization potential. Furthermore, both plasmids were classified as conjugative, and Mash distance analysis showed high similarity (distance < 0.01) between the blaKPC-3-carrying plasmids of E. coli and K. pneumoniae, strongly suggesting horizontal plasmid transfer between the two species. (Supplementary Tables 1 and 2).

Discussion

The susceptibility of bacteria to β-lactam antibiotics depends on a complex interplay of factors, including target binding affinity, resistance to β-lactamase degradation, and, in Gram-negative bacteria, outer membrane permeability12. Bacterial antibiotic resistance can arise through intrinsic or acquired mechanisms13. Intrinsic resistance involves genes naturally present on the host’s chromosome, such as AmpC β-lactamase in Gram-negative bacteria and efflux pump systems13. Acquired resistance, on the other hand, may result from mutations in antibiotic target genes or the horizontal transfer of resistance elements via plasmids, bacteriophages, transposons, and other mobile genetic elements13. Conjugation is the primary mechanism by which bacteria acquire external genetic material, with plasmids and transposons playing a crucial role in this process14. These resistance mechanisms often coexist within the same pathogen, contributing to multidrug resistance or increasing resistance levels to specific antibiotic classes14. Another key resistance mechanism involves mutations in penicillin-binding proteins (PBPs), which are essential for peptidoglycan synthesis and serve as primary targets for β-lactam antibiotics15. Structural modifications in PBPs, resulting in reduced affinity for β-lactams, are well-documented in Gram-positive bacteria but are increasingly reported in Gram-negative species16. Specifically, penicillin-binding protein 3 (PBP3), a high-molecular-weight transpeptidase, is a major component of E. coli and plays a key role in catalyzing peptide cross-bridges between the glycan chains of the peptidoglycan17,18. Mutations in PBP3, often driven by antibiotic pressure, can decrease its affinity for β-lactams, including carbapenems, complicating treatment strategies19. Ceftazidime and aztreonam preferentially inhibit PBP3, meaning that mutations in this protein can significantly impact bacterial susceptibility to these agents, potentially leading to increased minimum inhibitory concentration (MIC) values20. Additionally, the production of β-lactamase enzymes, particularly extended-spectrum β-lactamases (ESBLs) and carbapenemases, further contributes to multidrug resistance, severely limiting therapeutic options for Enterobacterales infections21. Klebsiella pneumoniae carbapenemases (KPCs) belong to class A serine β-lactamases and are the most widespread carbapenemases globally, predominantly found in K. pneumoniae but also in other Gram-negative species22,23. These enzymes hydrolyze penicillins, cephalosporins, aztreonam, and carbapenems and are frequently located on mobile plasmids within the Tn4401 transposon24. Consequently, in addition to clonal expansion, the blaKPC gene can spread through plasmid- and transposon-mediated horizontal gene transfer25. With the introduction of new beta-lactamase inhibitors, the emergence of KPC variants resistant to these therapies has been observed. Among them, the most common determinants are KPC-2 and KPC-3, which exhibit point mutations, insertions, or deletions that alter their resistance levels to those molecules26. The presented clinical case may represent an example of plasmid-mediated conjugation. Although Illumina sequencing did not allow for the complete reconstruction of the plasmid, the high homology (99.85%) between the plasmid harbored in K. pneumoniae and that in E. coli strongly suggests that the transposon carrying the blaKPC-3 gene was transferred between the two strains via the same plasmid. The blaKPC gene is typically located within the mobile transposon Tn440127. In our study, the K. pneumoniae strain harbored a single copy of the blaKPC-3 gene within Tn4401a. This transposon variant, designated as ‘isoform a’, is characterized by a 100 bp deletion between blaKPC and istB. While this isoform is commonly associated with blaKPC-2, it has also been identified in certain blaKPC-3-producing isolates from Israel and Italy28. In contrast, the E. coli strain carried the same blaKPC-3 variant in two copies, each located on separate instances of Tn4401a. A similar occurrence was previously reported in an E. coli strain of the same sequence type (ST) in the United States29. Additionally, a study demonstrated that E. coli DH5α transformants harboring an IncX3 plasmid with two copies of blaKPC-3:Tn4401a exhibited significantly MICs of carbapenems, cephalosporins, and BL/BLIs30. The Tn4401a transposon has been linked to increased carbapenem resistance due to modifications in the promoter region that enhance blaKPC-3 expression31. Compared to other transposon variants, Tn4401a exhibits higher blaKPC expression, likely due to a 99-bp deletion in the intervening sequence between the P1 and P2 promoters. This deletion may contribute to a more stable RNA structure, thereby enhancing gene expression27. The increased promoter activity of Tn4401a correlates with elevated carbapenem resistance, as demonstrated by higher meropenem and cefepime MICs in E. coli transformants compared to those harboring other transposons27. Consistently, our E. coli isolate exhibited high MIC values of carbapenems, cephalosporins and traditional BL/BLIs further supporting the role of Tn4401a in enhanced resistance to these agents. The two amino acid substitutions in the PBP3 of our E. coli, namely A233T and I332V, have been previously reported19. The presence of these PBP3 mutations, which reduce the affinity of ceftazidime for PBP3, together with other genetic alterations such as mutations in acrR and marR, and overexpression of sdiA, contribute to a significant increase in ceftazidime MIC in E. coli mutants19. Beta-lactam antibiotics mimic a transpeptidase substrate and bind to the catalytic domain of PBP, where they function as suicide inhibitors by forming a covalent adduct with the catalytic serine37. It can be hypothesized that the A233T mutation, which involves the substitution of alanine (a non-polar residue) with threonine (a polar residue), may induce conformational changes in the active site of PBP3, leading to reduced β-lactam binding due to altered electrostatic interactions. Similarly, the I332V mutation, where isoleucine is replaced by valine (a smaller residue), could impact PBP3 interactions with other cell division components, potentially influencing its function and antibiotic susceptibility19. In addition to these point mutations, recent literature highlights the significant role of PBP3 insertions in conferring reduced susceptibility to BL/BLI combinations. Specifically, the tetrapeptide duplication of amino acids 334–337 (YRIN) and its single-amino acid variant YRIK have been associated with decreased susceptibility to both aztreonam-avibactam and cefiderocol; furthermore, duplications such as PYRI (amino acids 333–336) and TIPY (amino acids 331–334) have similarly been linked to reduced susceptibility to aztreonam-avibactam38. We did not identify any missense mutation in acrR, while we detected two missense mutations in acrB (p.His596Asn and p.Ser1043Asn), one in acrA (p.Thr104Ala), two in marR (p.Gly103Ser and p.Tyr137His ) and one in marA (p.Ser127Asn). AcrR is a repressor of the acrAB operon32, and acrA and acrB are key components of an efflux system in E. coli, which actively expels a wide range of toxic compounds, including antibiotics, thereby contributing to intrinsic antimicrobial resistance33. MarR is a repressor of marRAB operon and therefore also a repressor of marA expression34. The marA locus is needed for production of a multiple-antibiotic-resistant (Mar) phenotype in E. coli, which include resistance to cephalosporins35,36. Among the variants we observed, acrB p.His596Asn is located in the predicted transmembrane domain and could theoretically affect substrate transport; however, this mutation has not been previously reported and its functional impact remains speculative. The acrB p.Ser1043Asn and acrA p.Thr104Ala substitutions are located in periplasmic or helical regions, with no known role in antibiotic resistance. Similarly, marA p.Ser127Asn lies in the C-terminal region, potentially involved in DNA binding, but no evidence supports an effect on acrAB expression or regulatory activity. The two marR (p.Gly103Ser and p.Tyr137His ) missense mutations are associated with resistance to quinolones and carbapenems, respectively19,37,38. Despite the potential overproduction of KPC-3, the presence of PBP3 mutations, and the missense mutations in marR, our E. coli isolate exhibited a low MIC for CAZ/AVI (Table 1). Given that the β-lactamase inhibitor avibactam is virtually devoid of intrinsic antibacterial activity39 and its ability to restore ceftazidime activity against KPC-producing Enterobacterales depends on its capacity to inhibit KPC β-lactamase activity40, the concurrent presence of PBP3 mutations (considering that PBP3 is the primary target of ceftazidime), and mutations in marR (which could potentially confer cephalosporin resistance) would theoretically be expected to result in resistance to CAZ/AVI. However, this was not observed in our E. coli isolate. There are two possible explanations for the low MIC of CAZ/AVI in this strain. The first is the absence of potential concurrent intrinsic resistance mechanisms13. The second, more speculative yet intriguing hypothesis is that avibactam, by binding to E. coli PBPs, particularly PBP241,42, acts just like a β-lactam and could compensate for the reduced binding of ceftazidime to the mutated PBP3. This interaction may help achieve a critical level of PBP binding, which is determined by both the number of PBPs bound and the affinity of the β-lactam for its targets43. Such an effect could potentially restore the full antimicrobial activity of the CAZ/AVI combination and is conceptually similar to the principle of enhanced bacterial killing through complementary target redundancy in Gram-negative bacteria44. Our research group proposed this mechanism in a commentary on the ATTACK trial45,46, suggesting that simultaneous inhibition of multiple PBPs could enhance bacterial killing by overcoming resistance mechanisms. This hypothesis was later confirmed by in vitro time-kill studies conducted by Veeraraghavan et al.47, which demonstrated a synergistic effect between sulbactam-durlobactam and imipenem against carbapenem-resistant Acinetobacter baumannii calcoaceticus complex strains47. These findings reinforce the clinical relevance of exploiting complementary PBP targeting to restore or enhance the efficacy of β-lactam-based therapies, particularly against multidrug-resistant Gram-negative pathogens but further studies are needed to confirm this hypothesis.

In conclusion, this study underscores the multifactorial nature of antibiotic resistance in the analyzed E. coli strain, involving both plasmid-mediated dissemination of the blaKPC-3 gene and potential contributions from PBP3 and marR mutations. While the exact impact of these mutations on MICs remains unclear due to the lack of comparative data, the findings highlight the need for further studies to clarify their role and better understand the interplay between plasmid dynamics, transposon activity, and target protein modifications in resistance mechanisms.

Materials and methods

Patient and samples

In this study, we extracted demographic and clinical data from the medical records of a patient admitted to Ischia Hospital in June 2024. The study was approved by the local Institutional Review Board and all research was performed in accordance with relevant guidelines and regulations. Informed consent was obtained from the patient for the publication of this information. A KPC-positive E. coli isolate from a blood culture and a KPC-positive K. pneumoniae isolate from a surveillance swab, both obtained from this patient, were characterized. The strains were provided as cryopreserved stocks, which were streaked onto standard blood agar plates and incubated at 37 °C to obtain single colonies. These colonies were then inoculated into tryptic soy broth and incubated overnight at 37 °C to prepare bacterial cultures for subsequent molecular analysis.

Antimicrobial susceptibility testing

Antimicrobial susceptibility testing was performed using the Vitek®2 system (bioMérieux, Marcy-l’Étoile, France). Disk diffusion on agar was used exclusively for cefiderocol.

DNA isolation

Genomic DNA was isolated using the Quick-DNA Fungal/Bacterial Kit (Zymo Research), supplementing the lysis buffer with lysozyme. The DNA concentration and quality were determined using a NanoDrop 2000 spectrophotometer and a Qubit Flex Fluorometer (Thermo Fisher Scientific).

Library preparation and illumina sequencing

Library preparation was conducted starting from 300ng of genomic DNA using the FX DNA Library Preparation Kit (Qiagen) and the NexteraTM XT Index kit (Illumina), following manufacturer’s instruction. Sequencing was performed on an Illumina MiSeq System using a 2 × 500 bp paired-end run.

Illumina data analysis

Data analysis was carried out using a series of specialized tools. Quality control (QC) of raw sequencing reads was performed using FastQC and MultiQC to evaluate the quality of reads both before and after filtering with fastp. Taxonomic classification was conducted using Kraken2. De novo assembly of reads was performed with Shovill and contig annotation was achieved using Prokka. The MLST tool was used to determine sequence types from the PubMLST database, while ABRicate identified resistance genes (CARD database) and virulence factors (VFDB database), applying coverage and identity thresholds of 90%. MOB_suite enabled plasmid reconstruction from contigs, while PlasmidFinder provided plasmid identification and typing. Additionally, after aligning reads to the reference sequence with bwa mem algorithm, EMBOSS transeq was used to convert DNA sequences to amino acid sequences in order to identify amino acid mutations.

Long-read whole genome sequencing with nanopore technology and data analysis

Library preparation was performed using the 4bases Microbiome WGS kit, a shotgun-based solution optimized for long-read sequencing. The protocol enables multiplexed library construction compatible with Oxford Nanopore platforms. Sequencing was performed on a GridION system using FLO-MIN114 flow cells, following the manufacturer’s instructions. Raw reads quality was assessed using FastQC. The presence of plasmids and the genes necessary for conjugation or mobilization as well as the presence of antimicrobial resistance genes were identified using the MOB-suite pipeline48,49,50. Variations were characterized with respect to a reference assembled genome (E. coli = GCA_000005845 and K. pneumoniae = GCA_000016305, NCBI) using snippy (https://github.com/tseemann/snippy). Identified SNPs and INDELs were then annotated with respect to gene, effect, and amino acid substitutions using snpEff51.