Tracing the evolutionary dynamics of carbapenem-resistant Escherichia coli in recurrent and multi-site infections
- Open Access
- 01.12.2025
- Research
Erschienen in:
Abstract
Background
Escherichia coli is a common pathogen causing various infections, including urinary tract infections (UTIs), bacteremia, pneumonia, meningitis, and intra-abdominal infections. Carbapenem-resistant Escherichia coli (CREC) poses a significant challenge due to its resistance to carbapenems, often the last line of defense against multidrug-resistant (MDR) bacteria, and constitutes a serious global public health threat [1]. In Taiwan, the incidence of CREC increased from 0.4 to 0.8% between 2011 and 2020 [2]. By 2012, E. coli had become the second most common carbapenem-resistant Enterobacteriaceae (CRE) isolated in hospitals across Asia [3].
The spread of carbapenem resistance occurs through horizontal gene transfer (HGT) of plasmids, promoting genetic diversification, and potentially contributing to bacterial adaptation, evolution, and persistence [4, 5]. Among CREC clinical isolates, metallo-β-lactamases (MBLs), such as blaNDM, are the most prevalent, followed by oxacillinases, including blaOXA-48-like [4]. Phylogroups A and D are the most prevalent among CREC, with three sequence types (STs), ST167, ST410, and ST131, commonly carrying carbapenemase genes [4]. Notably, transferable plasmids carrying resistance genes can recombine with virulence plasmids by integration or fusion, resulting in hybrid plasmids that not only exhibit antibiotic resistance but also enhance bacterial virulence [6‐8].
Anzeige
The host’s harsh and complex environment and the prolonged use of antibiotics can induce gene mutations in pathogens within the host. Additionally, genetic exchanges, such as transferable plasmids, enable pathogens to adapt to the constantly changing environment within the host, ultimately leading to persistent infections [5, 9, 10]. Through whole-genome sequencing (WGS) analysis, Yang et al.. identified mutations and insertions/deletions (INDELs) occurred during the recurrent infection process of Staphylococcus aureus. These mutations were associated with changes in toxin expression, carbon metabolism, and stress responses [11].
Among the 99 recurrent urinary tract infection (rUTI) patients, 41 experienced rUTI episodes involving repeated infections with a single highly genetically related clone [12]. Our previous study suggested that antibiotic use can trigger phenotypic shifts in E. coli strains responsible for rUTIs and drive plasmid evolution within the host [13]. These findings demonstrate how genetic evolution within the host, driven by selection pressure from antibiotics and/or immune responses, can shape strain phenotypes, potentially resulting in prolonged recurrent infections and multi-site infections by the same pathogen. In this study, we focus on examining the evolution of CREC in 18 patients with recurrent or multi-site infections, aiming to explore the relationship between adaptive genotypic and phenotypic evolution and infection outcomes.
Methods
Identification of carbapenem-resistant Escherichia coli isolates
The En Chu Kong Hospital Institutional Review Board has reviewed and approved this study, with the approved accession numbers ECKIRB1120801 and ECKIRB1131205. A total of 151 CREC were isolated from patients at En Chu Kong Hospital between August 2011 and July 2022. These isolates were identified in the clinical laboratory following the manufacturer’s recommendations, using colony morphology, Gram staining, biochemical tests, and the Vitek 2 system (bioMérieux, Marcy l’Etoile, France). The susceptibility of E. coli isolates to third-generation cephalosporins (ceftazidime or ceftriaxone, 30 µg/disc, BD BBL™ Sensi-Disc™, Sparks, MD, USA) was determined using the disk diffusion method on Mueller-Hinton (MH) agar plates (Bio-Rad, Marne la Coquette, France) according to CLSI guidelines (2021, M100-S31) [14]. Third-generation cephalosporin-resistant isolates were further tested for susceptibility to carbapenems, including imipenem, ertapenem, meropenem, and doripenem (10 µg/disc, BD BBL™). Among the 151 CREC isolates, 71 from 24 patients were chosen for further study based on the following criteria: (1) isolates obtained from different body sites of a single patient, or (2) isolates collected from two or more infection episodes of a single patient. These isolates were stored at − 80 °C in lysogenic broth (LB) containing 20% glycerol (v/v) until further use.
Pulsed field gel electrophoresis (PFGE)
PFGE was performed following previously established methods [12]. XbaI-digested genomic DNA was subjected to PFGE using a CHEF Mapper XA apparatus (Bio-Rad Laboratories, Inc., Hercules, CA, USA) and a 1% agarose gel (Seakem Gold agarose; FMC BioProducts, Rockland, ME, USA) in 0.5× Tris-Borate-EDTA (TBE) buffer. Electrophoresis was conducted for 19 h at 14 °C, with pulse times ranging from 6.76 to 35.38 s at 6 volts/cm. Salmonella enterica serovar Braendrap H9812 was used as the quality control strain, and its DNA pattern was utilized as the DNA marker for PFGE gels. Gels were stained with ethidium bromide (EtBr), and images were captured using UV transillumination. PFGE profiles were analyzed with GelCompar II software, version 2.0 (Unimed Healthcare, Inc., Houston, TX, USA), with a tolerance of 0.9% and an optimization parameter of 0.9%. Isolates were considered genetically highly related if their restriction fragment patterns showed > 85% similarity [15].
Anzeige
Phylogenetic grouping
CREC isolates were classified into phylogroups A, B1, B2, C, D, E, F, and clade I using a previously described method [16]. Classification was based on identifying the genetic biomarkers arpA, chuA, yjaA, and TspE4.C2 through quadruplex phylo-typing, combined with allele-specific PCR to identify strains belonging to phylogroups C and E [16]. The primers used in phylogroup analysis are listed in Table S1.
Multilocus sequence typing (MLST)
CREC isolates were genotyped using the seven-gene Achtman MLST scheme as described on the website https://pubmlst.org/ [17]. Allelic numbers and corresponding genotype assignments were provided by the MLST website.
Antimicrobial susceptibility tests
The antimicrobial susceptibility of CREC isolates was evaluated using the disk diffusion method, with results interpreted according to CLSI guidelines (2021, M100-S31) [14]. The tested discs included ampicillin (AM, 10 µg), ampicillin/sulbactam (SAM, 10/10 µg), amoxicillin/clavulanic acid (AMC, 25/10 µg), amikacin (AN, 30 µg), aztreonam (ATM, 30 µg), cefazolin (CZ, 30 µg), cefuroxime (CXM, 30 µg), cefmetazole (CMZ, 30 µg), ceftriaxone (CRO, 30 µg), ceftazidime (CAZ, 30 µg), cefepime (FEP, 30 µg), cefoxitin (FOX, 30 µg), chloramphenicol (C, 30 µg), ciprofloxacin (CIP, 5 µg), colistin (CL, 10 µg), doripenem (DOR, 10 µg), ertapenem (ETP, 10 µg), fosfomycin (FOS, 200 µg), gentamicin (GM, 30 µg), imipenem (IPM, 10 µg), levofloxacin (LVX, 5 µg), meropenem (MEM, 10 µg), piperacillin/tazobactam (TZP, 100/10 µg), sulfamethoxazole/trimethoprim (SXT, 23.75 µg/1.25 µg), tetracycline (TE, 15 µg), and tigecycline (TGC, 15 µg) (BD BBL™). Tigecycline susceptibility was interpreted according to EUCAST guidelines (< 18 mm indicating resistance), with colistin breakpoints referenced from the study by Galani et al. [18]. E. coli ATCC 25,922 was used as the quality control strain for the disk diffusion tests. Isolates were classified as MDR, extensively drug-resistant (XDR), and pandrug-resistant (PDR) based on criteria from a previous study [19].
Phenotypic detection of carbapenemase-producing E. coli by mCIM/eCIM
The mCIM and eCIM tests were conducted on CREC isolates following CLSI guidelines and previous studies to detect carbapenemase production [14, 20]. The mCIM result is interpreted as negative if the zone size is ≥ 19 mm, positive if the zone size is 6 to 15 mm, or intermediate (considered positive) if pinpoint colonies are present within a 16 to 18 mm zone. An isolate is positive for metallo-carbapenemase production if the eCIM zone size increases by ≥ 5 mm compared to the mCIM zone size; it is considered negative if the increase is < 4 mm [21, 22]. Internal controls for the mCIM and eCIM tests included Klebsiella pneumoniae ATCC BAA-1706 (carbapenemase negative), K. pneumoniae ATCC BAA-1705 (blaKPC positive), and K. pneumoniae ATCC BAA-2146 (blaNDM positive). The mCIM and eCIM tests were performed in duplicate to ensure reproducibility.
Carbapenemase gene detection
Bacterial genomic DNA was extracted from overnight cultures grown at 37 °C in 3 ml LB using the boiling method [2]. The DNA-containing supernatant was transferred to a new Eppendorf tube, and samples were stored at 4 °C until testing. Five major carbapenemase genes (blaKPC, blaNDM, blaIMP, blaVIM, and blaOXA–48) were detected by PCR amplification using a GeneExplorer Thermal Cycler (BIOER, China) with Fast-Run™ 2× Taq Master Mix (Protech, Taipei, Taiwan). The primers and PCR procedures used for detecting carbapenemase genes are listed in Table S1. PCR products were analyzed by electrophoresis on 1.2% agarose gels in 0.5× TBE buffer. After electrophoresis, the gels were stained with EtBr, and the PCR products were visualized using a UV transilluminator. Clinical isolates of lab stock carrying blaKPC, blaNDM, blaIMP, blaVIM, and blaOXA− 48 served as positive controls for PCR [2]. The PCR products were further subjected to Sanger sequencing to determine the carbapenemase NDM and KPC variants.
Virulence genes detection
We performed PCR to identify 31 virulence-associated genes, including those related to adhesion (fimH, afa, iha, foc, papGI, papGII, papGIII, papA, papC, sfa), toxin activity (hlyA, sat, usp, cnf1), iron acquisition (iroN, iutA, ireA, feoB, sitA, irp2, fyuA), capsule production (neuA, kpsMTII, kpsMTIII), serum resistance (iss, traT), auto-aggregation (agn43), curli production (csgA), maltose uptake (malX-PAI), as well as the protease gene ompT, and invasion of brain endothelium protein A gene ibeA. The annealing temperatures for the primers and the anticipated sizes of PCR products are detailed in Table S1. The PCR products were separated on a 1.2% agarose gel and visualized under UV light following EtBr staining. The virulence score was calculated by summing the number of virulence-related genes present in each strain.
Larvae infection assay
The virulence of E. coli isolates was assessed using the Galleria mellonella larvae infection model, following the procedure described by Wang et al. [23]. Larvae weighing between 160 and 200 mg were selected for the experiment. Before injection, the larvae were deprived of food and kept in darkness at 37 °C for one day. Overnight bacterial cultures were resuspended in 1x PBS, adjusted to an optical density 600 nm (OD600) of 1, and diluted to a concentration of 1 × 107 colony-forming unit (CFU)/ml. Subsequently, 10 µl of the bacterial suspension (1 × 105 CFU) was injected into the second-to-last left proleg of larvae using a Model 701 cemented needle syringe (Hamilton Company, Reno, Nevada, USA). The number of bacterial cells injected was confirmed by enumerating CFUs on LB agar plates. Additionally, ten larvae injected with 10 µl of sterile PBS served as the negative control group. Fifteen larvae were infected with each E. coli isolate, and the viability of the larvae was monitored for seven days post-infection at 24-hour intervals. Larvae that exhibited no response to physical stimuli were considered dead.
Biofilm formation assay
The biofilm formation of E. coli was evaluated using the crystal violet staining method as described previously [23]. Two different growth media, LB and M9 minimal medium, were used to assess the biofilm formation ability. Briefly, overnight bacterial cultures were diluted in both media to an OD600 of 0.2 and then inoculated into a 96-well plate. Biofilm formation was assessed after 1, 2, and 3 days of incubation at 37 °C. Following incubation, biofilms were fixed with 95% methanol and stained with 0.1% crystal violet for 30 min. After drying, 75% ethanol was used to solubilize the bound dye, and the absorbance of the solution was measured at 590 nm using a Synergy HTX Multi-Mode Reader (BioTek, USA). The experiment was performed using three biological replicates, each measured in technical triplicates.
Anzeige
Siderophore detection using chrome azurol S
The chrome azurol S (CAS) agar was prepared according to methods from a previous study [24], with slight modifications, to detect E. coli siderophore production. Bacteria were cultured in LB at 37 °C overnight and subsequently refreshed for 4 to 5 h the following day. The bacterial suspensions were adjusted to an OD600 of 0.6, and 20 µl of each suspension was inoculated onto CAS agar plates. E. coli strains CYEC029 and CYEC034 were used as positive and negative controls [23], respectively. The plates were incubated at 37 °C and observed for the development of an orange halo after 12, 24, and 36 h. Siderophore production ability was quantified by measuring the diameter of the orange halo. The experiment was conducted in biological triplicates.
Swimming motility assay
Single colonies of CREC isolates were inoculated onto soft agar plates containing LB with 0.3% agar using a 200-µl pipette tip [23]. The diameter of the resulting swimming zone was measured after incubation periods of 8, 16, and 24 h at 37 °C. The experiment was conducted in technical and biological triplicates to ensure the reproducibility of the results.
Human serum killing assay
A 50% serum was prepared by diluting the pooled human serum (Pel-Freez Biologicals, Rogers, Arkansas, USA) with PBS buffer. Then, 50 µl of the diluted serum was mixed with 50 µl of the bacterial suspension (OD600 of 1). The mixture was then incubated in Eppendorf tubes at 37 °C with agitation at 200 rpm. CFUs were counted at 0, 45, 90, and 180 min post-incubation, with the total bacterial count at 0 min used as the 100% control. Bacterial survival at each time point was compared against the total bacterial count at 0 min. The serum was heat-inactivated by incubating it in a 56 °C water bath for 30 min, serving as a negative control for serum complement-killing assay.
Cell culture and adhesion assay
The bladder epithelial cell line 5637 cells were cultured in 12-well plates with RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin until 80% confluency was reached. Before infection, the culture medium was replaced with fresh medium without penicillin-streptomycin, and the cells were infected with E. coli at a multiplicity of infection (MOI) of 10. The plates were centrifuged at 600 xg for 5 min to enhance bacteria-host cell contact and then incubated at 37 °C for 2 h. After incubation, the cells were washed three times with PBS to remove non-adherent bacteria. The cells were then lysed by incubation with saponin at 37 °C for 15 min, and the lysates were plated on LB agar plates. The number of colonies was counted to quantify adherent bacteria [25].
Anzeige
Determination of IL-1β, IL-6, and TNF-α mRNA by reverse-transcription quantitative PCR (RT-qPCR)
The assay was performed with modifications based on a previous study [26]. 5637 cells (1 × 106/well) were grown for three day in 60 × 15 mm culture dishes to approximately 80% confluence. E. coli was added to the wells at a MOI of 100 without centrifugation and incubated for 4 h for the cellular IL-1β, IL-6, and TNF-α gene expression assay. Each dish with 5637-E. coli coculture was washed three times with PBS buffer. The total mRNA expressed in E. coli-infected 5637 cells was extracted using TRIzol (Sigma-Aldrich, St. Louis, MO, USA) according to the manufacturer’s instructions. For cDNA synthesis, 1 µg of RNA was reverse-transcribed using SuperScript II RNase H-reverse transcriptase (Bionovas Biotechnology Co., Ltd, Canada). The expression levels of IL-1β, IL-6, and TNF-α mRNA were quantified by RT-qPCR, conducted on a StepOnePlus™ system (Thermo Fisher Scientific Inc., USA) with SYBR Green Master Mix (Ampliqon, Denmark) in 96-well qPCR plates (MB-Q96-LP and MB-QSM; Gunster Biotech, New Taipei City, Taiwan). The primers used for RT-qPCR are listed in Table S1. IL-1β, IL-6, and TNF-α expression levels were normalized to β-actin mRNA. RT-qPCR was performed in triplicate with a total reaction volume of 20 µl, including 10 µl of 2x SYBR Green PCR Master Mix, 0.4 µl of each primer, 8.7 µl of distilled H2O, and 0.5 µl of cDNA per sample. The reaction involved an initial heating step at 95 °C for 15 min, followed by 40 cycles of amplification. Quantification was carried out using the 2-ΔΔCt method, where the Ct value represents the PCR threshold cycle. Lipopolysaccharide (LPS) treatment at a final 5 µg/ml concentration was used as the positive control.
Phage isolation, identification, and host range test
E. coli-targeting phages were isolated from water samples collected from hospitals and the environment. Single plaques formed on LB agar were selected and enriched through co-cultivation with the host strain. Pure phage lysate was obtained by filtration through a 0.45 μm filter membrane, and the phage titer was determined using a plaque assay as described previously [27]. The phages were preserved in LB containing 20% glycerol at − 80 °C.
Phage identification was conducted using transmission electron microscope (TEM). Phage particles were applied onto a formvar film on a carbon-coated copper grid (300 mesh) for 2 min. After washing the grids twice with MQ water, the phages were negatively stained with 0.1% uranyl acetate for 30 s [28]. The phages were then observed using a JEM-1400 Plus microscope (JEOL, Tokyo, Japan), and images were captured using a camera. The host range of the phages, indicating their ability to infect different CREC strains, was determined through plaque formation assays [29].
Whole-genome sequencing and analysis
The genomic DNA (gDNA) of CREC was extracted using the Presto™ gDNA Bacteria Advanced Kit (Geneaid Biotech, Ltd., Taiwan), following the gram-negative bacteria protocol with a 5 ml LB culture. The complete genome sequence of CREC was determined using the Nanopore WGS platform [30]. A sequencing library was prepared using 1 µg of gDNA and the Ligation Sequencing Kit (SQK-NBD114.24, Oxford Nanopore Technologies, Oxford, UK). The CREC genomes were sequenced on the Oxford Nanopore MinION MK1C, with 300 ng of gDNA loaded onto an R10.4.1 flow cell. Quality assessment of the generated reads was performed using FastQC v0.11.5. Raw signals were processed into DNA sequences using the ONT Gussy basecalling program (version 4.2.3). The genomes were assembled using the Flye de novo assembler (version 2.9), with options enabled for backward compatibility (plasmids) and uneven coverage mode (meta) in the analysis [31].
Anzeige
The assembled CREC genomes were submitted and deposited in the NCBI GenBank database. Genome annotation was conducted using the NCBI Prokaryotic Genome Annotation Pipeline (PGAP, version 6.1) [32]. To identify antimicrobial resistance genes (ARGs), ResFinder [33] was employed. PlasmidFinder databases were utilized to identify plasmid types, and pMLST was employed to determine the plasmid replicon [34]. VirulenceFinder and Virulence Factor Database were used to identify virulence-associated genes [35, 36].
Comparison and phylogenetic analysis of closely related plasmids
The assembled plasmid sequences of CREC with identical replicons and similar sequence sizes were chosen for further analysis. Phylogenetic trees were constructed using ClustalW alignment and the Maximum Likelihood method with 500 bootstrap, implemented through the Molecular Evolutionary Genetics Analysis software (MEGA, version 11) [37].
Statistical analyses
Statistical analyses were performed using data from three independent biological assays. Standard deviations were calculated and presented. Statistical analyses for biofilm formation, CAS assay, motility, and serum resistance were performed using two-way ANOVA with Tukey’s multiple comparisons test. Cell adhesion data were analyzed using ordinary one-way ANOVA followed by Tukey’s multiple comparisons test. Kaplan-Meier survival curves were used to assess larvae survival, and statistical significance was evaluated using the Mantel-Cox log-rank test. Cytokine (TNF-α, IL-6, IL-1β) expression levels were analyzed using ordinary one-way ANOVA, followed by Dunnett’s multiple comparisons test to compare each experimental group with the first isolate. P-values less than 0.05 were considered statistically significant. Graphs and statistical analyses were generated using GraphPad Prism software (version 10.4.0, GraphPad Software Inc., USA).
Results
Clonality, phylogroups, and sequence types among CREC isolates from the same patient
To explore the evolution of CREC within hosts, we continuously collected 151 CREC isolates from a single regional hospital between 2011 and 2022. We further focused on investigating multiple CREC isolates from individual patients, resulting in a total of 71 CREC strains isolated from 24 patients between 2016 and 2022. Specimens included urine, sputum, blood, wound pus, and bronchial washings (Fig. 1).
Anzeige
Fig. 1
Genetic relatedness of 65 CREC strains isolated from 21 patients based on PFGE analysis. Strains with PFGE patterns showing > 85% similarity (indicated by red dashed lines) are considered highly related. The sample specimens, collection dates, phylogroups, and sequence types for the CREC isolates are shown in the figure. The abbreviations for the sample sources of the strains are as follows: U, urine; S, sputum; B, blood; WP, wound pus; BW, bronchial washing
Using rUTIs as an example, rUTIs are defined as ≥ two episodes of uncomplicated UTI in the previous six months or ≥ three episodes within a year. Although our study is not limited to UTIs, we use rUTIs to illustrate the concept of recurrent infections. Our study aims to include both strains responsible for recurrent infections and those isolated from multiple infection sites within a single host. This approach was designed to investigate the evolution of bacterial traits responsible for recurrent and site-specific infections. Therefore, all CREC strains isolated from a single patient were enrolled in the subsequent analysis.
Initially, we used PFGE to evaluate the genotypic relatedness of CREC strains isolated from 24 individual patients, defining strains with more than 85% genetic similarity as highly related (Fig. 1). This cutoff was based on a previous study [15]. Six strains from three patients were excluded from further analysis due to the unavailability of clear band patterns by PFGE. Among the remaining 65 CREC strains from 21 patients, highly related strains (a total of 51 CREC strains) were repeatedly isolated from 18 patients. Strains from patients 6 (433 and 464), 15 (R25 and R28), and 23 (618 and 766) exhibited genetic dissimilarity. Additionally, strains R15 and R43 from patient 17, 774 and 755 from patient 18, and R18, 455, 639, and 661 from patient 5, were not highly related to other CREC strains from the same patients. Notably, patients 6 and 18 experienced persistent infections with two distinct CREC strains in succession (Fig. 1). Therefore, a total of 20 strain groups (highly genetically related strains from the same patient) were categorized and included in the subsequent tests.
According to phylogenetic group typing findings, phylogroups B2 and D are predominantly associated with extraintestinal infections, such as urinary tract infections, neonatal meningitis, and sepsis [38, 39]. We further analyzed the phylogroup distribution of the 20 CREC groups, finding that 7 (35%) belonged to phylogroup B1, 6 (30%) to phylogroup A, 4 (20%) to phylogroup B2, and 3 (15%) to phylogroup D. (Fig. 1). MLST analysis revealed that ST617 dominated among the 20 groups (7 groups, 35%), followed by ST405 (6 groups, 30%), ST5229 and ST131, each present in 3 CREC groups (15%), and one group belonged to ST1193 (5%).
Anzeige
We performed a preliminary analysis of plasmid content in 51 CREC strains using the Kado-Liu plasmid extraction method to evaluate the possibility of in vivo plasmid acquisition or loss. Among the 20 CREC groups, strains from patients 20 and 24 showed a reduction in plasmid numbers within the host. In contrast, isolates 687 and 700 from patient 18 acquired an additional plasmid, while isolate 636 lost one plasmid compared to the initially isolated strain (data not shown).
Antimicrobial susceptibility changes within the host and carbapenem resistance mechanisms
Antibiotic use can drive mutations and promote antibiotic resistance in pathogens within the host. Therefore, we analyzed the susceptibility of 51 strains against 26 antibiotics. Among 20 groups, there were variations in susceptibility to amikacin, gentamicin, cefmetazole, cefepime, tetracycline, aztreonam, colistin, imipenem, ertapenem, and doripenem in 7, 1, 4, 1, 3, 2, 1, 2, 1, and 1 strain groups, respectively (Table S2). In addition, among these 20 strain groups, we found that 9 groups were MDR, 6 groups consisted of XDR strains, and interestingly, 4 groups (patients 10, 16 (group 16 − 2), 18 (group 18 − 2), 20) had strains that transitioned from XDR to MDR. Furthermore, for group 18 − 1, 2 strains were MDR, and 5 were XDR (Table S2).
The mean age of the 18 patients was 79.3 years, with only seven males (38.9%). Only patients 18, 20, 21, and 22 had no history of chronic diseases (Table S3). Medication records during the strain collection period indicate that patient 7 had used cefmetazole, and strain 447 demonstrated greater resistance to cefmetazole compared to the initial strain 439 (Table S2 & S3). Similarly, strains 446 and 536 from patient 8 exhibited higher resistance to cefepime compared to strains 463 and 532, which may also correlate with their medication history (Table S2 & S3).
To explore the mechanisms driving carbapenem resistance and the dissemination of carbapenemase-carrying plasmids in CREC, we initially conducted phenotypic mCIM/eCIM analyses to determine the types of carbapenemases present. As shown in Table S4, among the 20 groups, 8 carried class A or D carbapenemases, while 10 harbored class B carbapenemases. Strains 305 and 306 tested negative in the mCIM analysis. mCIM/eCIM testing showed that strain 488 from patient 11 carried class A or D carbapenemases, while strain 492 carried class B carbapenemases.
Additionally, PCR detection targeting five common carbapenemase genes generally correlated with the mCIM/eCIM results. Interestingly, both strains 488 and 492 from patient 11 carry blaOXA-48 and blaNDM-5. However, their mCIM-eCIM results differ, possibly due to differences in gene expression levels. The interpretation of mCIM/eCIM outcomes was complicated by the presence of multiple carbapenemase genes from different classes. Among these 20 groups, 9 carried blaNDM-5, 4 carried blaKPC-2, 2 carried blaOXA-48, 2 carried both blaKPC-2 and blaOXA-48, 1 carried blaNDM-1, and 1 carried both blaNDM-5 and blaOXA-48 (Table S4). Importantly, no acquisition or loss of carbapenemase genes was observed within the host during the study.
Distribution of virulence-associated genes in highly related strains
To assess virulence scores and analyze changes in the distribution of virulence-associated genes in CREC strains during persistent or multi-site infections, we analyzed the presence of 31 virulence-associated genes across 51 strains using PCR (Table S5). The virulence score was determined by the cumulative number of detected virulence genes. Our results showed that, in all 20 groups, the distribution of virulence genes remained consistent with their respective initial isolates. Among these groups, only three from patients 11 and 18 carried fewer than 10 virulence-associated genes (all belonging to phylogroup B1), while those from patients 7, 8, and 24 exhibited the highest scores, with up to 20 detected genes. The strains isolated from patients 7 and 8 belong to phylogroup B2, while the strain from patient 24 belongs to phylogroup B1. All 20 clones consistently carried feoB, agn43, and csgA, whereas papG1 (found in 2 groups) and neuA (found in 3 groups) were the least prevalent (Table S5). None of the 51 strains harbored sfa, foc, cnfI, ireA, or ibeA. Notably, no acquisition or loss of virulence-associated genes of CREC was observed within the host.
Variation in virulence among closely related CREC strains in the Galleria mellonella larvae model
The G. mellonella larvae model is widely used to assess bacterial virulence, leveraging its innate immune components, such as hemocytes and antimicrobial peptides, which are functionally analogous to certain aspects of the human innate immune system [40, 41]. Given the diverse clinical origins of our CREC strains, this model provided a viable alternative to a mouse infection model. Our results showed significant differences in larval survival after seven days of infection with CREC strains from six groups (patients 5, 7, 8, 11, 16, and 18) relative to their initial isolates (Fig. 2 & Table S6). Strains from patients 7 and 8 demonstrated a gradual increase in virulence toward larvae over time. In contrast, virulence decreased in strain 492 (patient 11) and strain 613 (patient 16, group 16 − 2) relative to their initial isolates, strains 488 and 571, respectively. Similarly, strains 572 and 580 from patient 5 exhibited lower virulence than strains 505 and 605, while strains 587, 611, and R7 from patient 18 (group 18 − 1) displayed higher virulence compared to strains 624, 636, 687, and 700 (Fig. 2 & Table S6).
Fig. 2
Analysis of the virulence of six highly related CREC groups in larvae from (A) patient 5, (B) patient 7, (C) patient 8, (D) patient 11, (E) patient 16, and (F) patient 18. Larval survival rates were monitored for up to seven days post-infection, with 15 larvae infected per strain. The first isolate from each group is colored in red. The Kaplan-Meier method was used to evaluate differences in overall larval survival. Statistical analyses were performed by comparing each isolate to the first isolate from the same patient. Exact p-values were calculated, and statistical significance is indicated as follows: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001
To investigate the mechanisms underlying the differences in virulence among strains from these six groups, we analyzed the expression levels of IL-1β, IL-6, and TNF-α cytokine genes in the strains from all six groups through cell infection experiments (Table S7). Infection of 5637 cells with these CREC isolates led to a 52.7–241.9-fold increase in IL-1β expression, a 17.9–183.6-fold increase in IL-6, and a 38.8–311.1-fold increase in TNF-α. However, no significant differences were observed among other isolates from the same patient.
Variations in virulence-associated phenotypes among strains from the same patients
Variations in virulence to larvae were observed among strains within the same group, suggesting that genomic evolution may influence bacterial phenotypes and pathogenicity. Therefore, we also analyzed the virulence-associated phenotypes of 51 CREC strains. In our prior research, we noted that highly related strains from a single patient exhibited morphological changes on blood agar plates (BAP) [13]. In this study, we found that among the 20 CREC groups, strains from 10 patients consistently displayed a smooth phenotype on BAP, 3 exhibited a rough phenotype, and 1 showed a mucoid phenotype. Notably, colonies from patients 5, 11, 16 (group 16 − 1, strains 540 and 609), and 18 (group 18 − 2, strains 739 and 753) transitioned from rough to smooth. Strain 558 from patient 14 had a more pronounced mucoid phenotype compared to other isolates, while the strain from patient 17 transitioned from smooth to mucoid (Fig. S1 & Table S6).
Given the role of biofilms in bacterial pathogenesis, we evaluated the biofilm-forming ability of CREC strains in both LB and M9 minimal media using crystal violet staining. While no differences were observed among strains isolated from the same patient in M9 medium, significant variation was found in LB across 10 groups. Strains from patients 2, 16 (group 16 − 1), 18 (group 18 − 1), 22, and 24 showed an increasing trend in biofilm formation according to the order in which the strains were isolated. Conversely, strains from patients 14 and 18 (group 18 − 2) exhibited a gradual decrease in biofilm formation ability. Unique trends were observed in patient 5, where only strain 572 formed more biofilm than the first isolate. Strains 463 and 532 from patient 8 and strain 621 from patient 9 formed less biofilm compared to their first isolates (Table S6).
Iron is essential for bacterial growth and gene regulation, and pathogens have developed mechanisms, such as siderophore, to compete for iron in host environments. Using a CAS assay to measure siderophore production, we observed significant changes in the CREC of 8 out of 20 groups compared to their initial isolates. Strains from patients 2, 10, and 12 showed a progressive decline in siderophore production. Conversely, strains from patients 11, 22, and 24 demonstrated a gradual increase in siderophore production. Strain 558 from patient 14 and strains 572 and 605 from patient 5 exhibited reduced iron acquisition compared to their initial isolates (Table S6).
Motility and adhesion are crucial during the early stages of infection. Among the 20 groups tested, strains from 10 groups showed no motility under the test conditions. The motility of strains from patients 3 and 16 gradually increased, with group 16 − 2 notably transitioning from non-motile to motile, whereas strains from patient 11 exhibited the opposite trend, transitioning to less motile. In 5637 cell adhesion assays, only isolate 609 from patient 16 demonstrated enhanced adhesion ability compared to isolate 540 (Table S6).
Serum resistance, for bacterial survival during bacteremia and the progression of invasive infections, was evaluated using serum-killing assays. Strains from patients 12, 16 (both groups 16 − 1 and 16 − 2), 17, and 22 gradually lost their serum resistance, while only the strains from patients 7 and 8 demonstrated a gradual increase in virulence toward larvae over time, potentially indicating adaptive mechanisms or genetic modifications enhancing pathogenicity. After 3 h of serum treatment, strain 439 from patient 7, isolated from bronchial washing, had only 0.13% bacterial survival, whereas strain 447, isolated from blood, exhibited a 4.1-fold increase in bacterial growth. Surprisingly, after 3 h of serum treatment, the first strain from patient 16, strain 571, isolated from urine, showed a 25.9-fold increase in bacterial growth, while strain 613 isolated from blood, had only 37% bacterial survival. Additionally, strains 572 and 580 from patient 5 showed reduced serum resistance compared to strains 505 and 605. Similarly, strain 687 from patient 18 (group 18 − 1) almost completely lost its serum resistance (Table S6).
Phage typing of CREC strains
Phage typing and molecular techniques, such as whole-genome sequencing, offer a comprehensive strain identification and differentiation method within a species [42, 43]. Furthermore, due to their ability to lyse pathogens, lytic phages have the potential to combat CREC [44]. In this study, eight lytic phages were isolated from environmental water and hospital wastewater using the E. coli reference strain EC25922 (ATCC 25922) and clinical isolate EC35 as hosts. Transmission electron microscope identified all eight phages as belonging to the Myoviridae family, regardless of their source (Fig. S2 & Table S8). We further assessed the lytic activity of these phages against CREC strains within the same group (Fig. 3). Notably, CREC strain 740, isolated from patient 24, exhibited reduced resistance to the phages compared to its earlier isolates. In contrast, CREC strains from patients 7, 16, 17, and 22 developed increased resistance following evolutionary adaptation within the host. For patient 5, only strain 572 was resistant to the phages, while strains 624, 636, R7, and 700 from patient 18 showed heightened susceptibility compared to their initial isolate, strain 587 (Fig. 3). These findings reveal the variability in phage susceptibility among isolates from different patients, as well as between sequential isolates from the same patient, reflecting the complex interplay of bacterial adaptation and phage efficacy.
Fig. 3
Lytic activity of eight isolated phages against 20 groups of closely related CREC strains. The color gradient, ranging from white to dark blue, indicates bacterial susceptibility levels to the phage, with white representing high susceptibility and dark blue representing low susceptibility
The horizontal gene transfer drives genomic evolution of CREC within the host
We further conducted whole-genome analysis on six groups of CREC strains that exhibited altered virulence to larvae to elucidate the potential mechanisms underlying their virulence transition. Although strains R18, 455, 639, and 661 from patient 5, as well as strains 755 and 774 from patient 18, did not show genetic relatedness to other strains isolated from the same patient (Fig. 1), these strains were included in the WGS analysis to explore the in vivo transfer of plasmids between different strains.
We isolated a total of eight CREC and one Pseudomonas aeruginosa (strain R44) from patient 5, with R18 being the only strain obtained from a blood sample, while the others were isolated from urine samples. Strains 505, 572, 580, and 605 exhibited > 85% similarity by PFGE and were collected between November 2019 and October 2020. In contrast, R18, 455, 639, and 661 were isolated in December 2020, May 2019, February 2021, and April 2021, respectively (Fig. 4A). Interestingly, the first isolated strain, 455, carried an IncFIA plasmid, with highly similar sequences identified in strains 505, 572, 580, and 605. The IncC plasmid initially found in strain 455 was later detected in strains 572, 580, and 639. Similarly, the IncI1-I(Alpha) plasmid from strain 455 was observed in strains 505, 572, and 580. Notably, the IncN plasmid was consistently found across all four highly related strains (Table S9). However, the genome of the P. aeruginosa R44 strain did not reveal any plasmids similar to those found in other patient 5 isolates (data not shown).
Fig. 4
The timeline of bacterial isolates for patients 5, 16, and 18, and the phylogenetic tree analysis of related plasmids among these strains. (A) Bacteria icons of the same color indicate high genetic relatedness as demonstrated by PFGE analysis. Only the R44 strain from patient 5 is Pseudomonas aeruginosa, while the other strains are E. coli. Patients 5, 16, and 18 were sequentially infected with 5 (A1–A5), 2 (B1–B2), and 4 (C1–C4) distinct CREC clones, respectively. This figure was created using BioRender software. Phylogenetic trees of plasmid replicons IncC (B), IncFIA (C), IncI1-I(Alpha) (D), IncR (E), and IncFII(pRSB107) (F) were constructed using Molecular Evolutionary Genetics Analysis (MEGA, version 11.0) with Maximum Likelihood method. Scale bar equals percent sequence divergence. The numbers on the nodes indicate the frequency with which each node was recovered out of 100 bootstrap replicates, from a total of 500 replications
Strain 439 from patient 7 was isolated from wound pus, while strain 447 was obtained from blood. All four strains from patient 8 were isolated from sputum. Although the strains within these two groups exhibited low genetic relatedness (Fig. 1), they all carried similar IncFIA and p0111 plasmids, with the IncFIA plasmids harboring the blaKPC-2 gene. Moreover, all six strains belonged to ST131 and phylogroup B2. Interestingly, the p0111 plasmid was also identified in ST405/phylogroup D strains 540 and 609 from patient 16. These two strains showed high PFGE similarity and carried eight and seven potential plasmid circular contigs, respectively, in their genomes, including IncC, IncFIA, IncC/O/K/Z, p0111, and IncX4 (Table S9). However, no antibiotic resistance genes or virulence genes were identified in the p0111 plasmid.
Interestingly, the chromosome of strain 447 from patient 7 harbored the sitA gene, which was absent in strain 439. This gene is associated with manganese and iron uptake as well as resistance to oxidative stress, factors that influence bacterial virulence (Table S9). Similarly, strain 609 from patient 16 possessed the espY4 gene, a type III secretion system effector protein, which was absent in the highly genetically related strain 540.
In another CREC group from patient 16, strains 571 and 613 contained five potential plasmid circular contigs, including replicons IncC, IncFIA, IncI1-I(Alpha), and IncI2(Delta). In contrast to the broader variation observed in strains from other patients, strains 488 and 492 from patient 11 (isolated from wound pus and urine, respectively) had relatively conserved genomes, likely due to the proximity of their isolation times. Both strains carried IncFIA and ColKP3 plasmids (Table S9).
From patient 18, a total of 11 strains were isolated from urine. Among them, strains 587, 611, 624, 636, R7, 687, and 700, collected between June 2020 and December 2021 (Fig. 4A), showed high genomic similarity and were grouped as 18 − 1. Strains 739 and 753, isolated between January and March 2022, also exhibited high genomic similarity, forming group 18 − 2. Strain 755 was isolated in March 2022, and strain 774 in June 2022, with no significant overlap in their isolation periods (Fig. 4A). In general, strains in the 18 − 1 group carried IncC, IncI1-I(Alpha), IncFIA, and IncI2(Delta) plasmids. However, strain 636 lacked the IncFIA plasmid, and strain 687 lacked the IncI1-I(Alpha) plasmid but carried additional IncFII(pRSB107) and IncB/O/K/Z plasmids. Strain 700 also carried IncFII(pRSB107) and IncR plasmids. Notably, the IncR plasmid was also found in subsequent strains outside the 18 − 1 group, including strain 739 (group 18 − 2) and strain 755.
Among the six patients, we identified three (patients 5, 16, and 18) carrying IncC plasmids (Table S9). Overall, the phylogenetic tree results show that these plasmids can be divided into two main clusters. For patient 18, the R7 and 587 strains, which were the earliest isolates from this patient, have IncC plasmids with high similarity. Interestingly, the IncC plasmids from these two strains also closely resemble those from patient 5’s 580, 572, and 455 strains. Additionally, the IncC plasmids from patient 18’s strains 700, 611, and 624 showed high similarity to the plasmid from patient 5’s strain 639. For patient 16, the strains were divided into two clones based on PFGE > 85% similarity: the IncC plasmids from strains 609 and 540, as well as those from strains 571 and 613, exhibited high similarity within their respective groups (Fig. 4B). All 19 IncC plasmids were classified as ST3 or putative ST3 plasmids. The IncC plasmids of strains 540, 609, and 755 carried the cheD (methyl-accepting chemotaxis protein) and tepI genes (negative regulator of the major pilin), although strain 755 lacked the ipaH (invasion plasmid antigen) gene. Additionally, strains 687, 753, and 774 harbored fepC (iron-enterobactin transporter ATP-binding protein) and bsaO (type III secretion system protein), with the IncC plasmid of strain 687 also containing pilV (shufflon system plasmid conjugative transfer pilus tip adhesin), and that of strain 753 carrying cyaB (cyclolysin secretion ATP-binding protein) and rtxB (RTX toxin transporter, ATPase protein). Regarding resistance genes, most IncC plasmids carried aadA2, aac(6’)-Ib3, floR, sul1, sul2, and dfrA12. However, differences in resistance genes were observed even among plasmids with high sequence similarity on the phylogenetic tree. For instance, the IncC plasmid of strain 540 contained aph(6)-Id, which was absent in the plasmid of strain 609. These findings suggest the evolution of plasmids within the host environment.
After sequencing 31 CREC strains from six patients, we found that 29 strains carried IncFIA plasmids, which can be divided into two major clusters according to the phylogenetic tree (Fig. 4C). Interestingly, strains 455 and 639 from patient 5, although belonging to different strains, exhibited high similarity in their IncFIA plasmids. Additionally, strains 755, 687, 753, 739, and 611 from patient 18, along with strains 572 and 580 from patient 5, showed high similarity in their IncFIA plasmids despite having < 85% similarity in PFGE (Fig. 1). However, IncFIA plasmids of strains 611, 687, and 753 lacked the tet(B) gene compared to other plasmids. Similarly, the IncFIA plasmids from strains 774, 700, 624 of patient 18, strains 605 and 505 of patient 5, and strain 571 of patient 16 also exhibited high similarity. Notably, strains 700 and 774 lacked the traJ gene in their IncFIA plasmids compared to other strains, while strain 605 carried an additional sul2 gene (Table S9). Strain 447 from patient 7 and strains 446, 536, and 532 from patient 8 showed plasmid similarity, yet strain 532 lacked the traJ gene in its IncFIA plasmid compared to the others. Similarly, although strains 455 and 639 from patient 5 had similar IncFIA plasmids, strain 639’s plasmid lacked the aadA2, blaTEM−176, sul1, and dfrA12 resistance genes. This suggests the transmission of these plasmids between different CREC strains within patients and may imply the importance of IncFIA plasmids in persistent infections and their evolution within the host.
The IncI1-I(Alpha) plasmid was detected in CREC strains from patients 5, 6, and 18. The similarity of the IncI1 plasmids in patient 5’s strains corresponded to their PFGE pattern similarity results (Figs. 1 and 4D). Strains 455 and 580 were closely related, while the plasmids of strains 572 and 505 were similar. For patient 18, the IncI1 plasmids were primarily divided into two clusters: strains 636, R7, 700, and 624 had evolutionarily similar IncI1 plasmids, whereas the plasmids of strains 753, 587, 739, and 774 were closely related. However, these strains belong to three different PFGE clones (Fig. 4A). The plasmids isolated from patient 5 were relatively smaller compared to those from other patients (except for strain 739 from patient 18) (Table S9). These plasmids carried blaCMY−2 and the shufflon system plasmid conjugative transfer pilus tip adhesin pilV gene, except for the IncI1-I(Alpha) plasmid in strain 753, which was inserted into the chromosome and lacked both blaCMY−2 and pilV. Additionally, strain 774 from patient 18 lacked the tssM gene (type VI secretion system membrane subunit), compared to others isolated from the same patient.
We identified an IncR plasmid in strains 700, 739, and 755 from patient 18, as well as in strain R18 from patient 5 (Table S9). The phylogenetic tree revealed significant differences between the IncR plasmid from strain R18 and the other three plasmids (Fig. 4E). The IncR plasmid in R18 was 135,517 bp in size and carried antimicrobial resistance genes (aadA2, aadA1, blaTEM−1B, cmlA1, sul3, drfA12) and the conjugation-related gene traJ (a positive regulator of the conjugal transfer operon). However, no transfer of this plasmid was observed among other strains from patient 5. In contrast, the IncR plasmids in strains 755 and 739 were 51,502 bp, while the plasmid in strain 700 was 51,503 bp. Although these three IncR plasmids lacked the traJ gene and other related genes, they were found across three different clones in patient 18.
Whole-genome sequencing data also revealed that strains 687, 700, 753, 755, and 774 from patient 18 carried an IncFII(pRSB107) plasmid with highly similar sequences (size 56,950 − 69,238 bp) (Table S9). Strain 700 carried a 56,950 bp plasmid, which exhibited evolutionary differences compared to the other four plasmids (Fig. 4F). Interestingly, this plasmid was absent in strains isolated before strain 687 (strains 587, R7, 611, 624) but was transmitted among four different CREC clones (Fig. 4A & Table S9). Although this plasmid does not carry antimicrobial resistance genes, VirulenceFinder analysis results indicated that all five plasmids contained mcbA (bacteriocin microcin B17) and traT (outer membrane protein complement resistance) (Table S9). However, only plasmids from strains 687 and 755 harbored the traJ gene, suggesting their potential transferability.
We further performed conjugation experiments using E. coli C600 as the recipient, employing a single protocol reported in a previous study [25], to assess the transferability of the following plasmids: the IncC plasmid from strain 540, the IncC plasmid from strain 571, the IncFIA plasmid from strain 455, and the IncI1 plasmids from strains R7 and 587 (data not shown). Due to the absence of antimicrobial resistance determinants, the IncR plasmid from strain 700 and the IncFII plasmid from strain 687 (shown in Fig. 4E and F) could not be tested in this assay. Our results showed that only the IncFIA plasmid from strain 455 was successfully transferred to C600. Under our experimental conditions, the IncC plasmid from strain 571 was not transferable; however, its IncFIA plasmid carrying blaNDM−5 could be transferred. Similarly, the IncI1 plasmids from strains 587 and R7 were not transferable, but their respective IncFIA plasmids carrying blaNDM−5 were successfully transferred. Notably, these IncFIA plasmids all harbored tra genes. Interestingly, the IncC plasmid from strain 540 was also non-transferable under our conditions. In contrast, its IncB/O/K/Z plasmid, which carries blaCMY−2, was able to transfer either independently or co-transfer with the IncFIA plasmid. To evaluate the impact of these plasmids on bacterial virulence, we tested the transconjugants in the G. mellonella larvae model (data not shown). However, none of the plasmids enhanced the virulence of C600; all transconjugants exhibited 100% survival in larvae at 7 days post-infection, similar to the original C600 recipient strain. Therefore, under our experimental conditions, these plasmids did not directly influence virulence of E. coli C600.
Discussion
In this study, we collected CREC strains from the same patient during different episodes and within intervals of less than three months to investigate pathogen trait evolution across various clinical specimens. The sample size in this study is relatively small due to several constraints: (1) Our sample collection specifically targeted CREC rather than more antibiotic-sensitive E. coli. (2) Patients needed to return to the same hospital for follow-up sample collection, which limited the number of multiple samples obtained. Bacterial adaptation to specific niches or persistent colonization within the host likely drives phenotypic variability [45], including alterations in antimicrobial resistance profiles and virulence factors [46]. These adaptations enable bacteria to survive in diverse host environments, which vary in nutrient availability, immune responses, and physical conditions [47]. Song et al.. reported that carbapenem-resistant hypervirulent K. pneumoniae (CRhvKP) can transition from hypermucoviscosity to hypomucoviscosity, enhancing its persistence in the urinary tract in both patients and antibiotic-free mouse models [48]. This mucoid switch, driven by mutations in the rmpA and wcaJ genes, mediated by ISKpn26 insertions or deletions, reduced sepsis virulence while increasing adherence to epithelial cells and biofilm formation, aiding colonization. Such transitions are key to CRhvKP’s survival in the urinary tract and its potential spread. In contrast, when moving from the urinary tract to the bloodstream, selective pressures like complement-mediated killing can drive further mutations and alter virulence gene expression. Therefore, traits like biofilm formation, iron acquisition, and immune evasion are critical for colonization and persistence at specific sites [49]. The potential for evolutionary changes across different body sites underscores the challenges of treating infections caused by CREC, as the pathogen may exhibit varying phenotypes depending on its location within the host. This may partially explain the discrepancies between in vitro phenotypes and the phenotypes exhibited by pathogen when invading different sites within the host.
Our findings highlight the significant proportion of CREC strains causing persistent infections across multiple body sites, which contrasts with our previous observations and the typical behavior of rUTI-causing E. coli strains [12]. Approximately 41% of patients experienced rUTI with a highly related E. coli strain, suggesting that the host immune system or antibiotic treatment often cannot eradicate these infections. Such interventions may drive bacterial evolution within the host. In contrast, rUTIs in another subset of patients are often caused by different E. coli strains, especially in immunocompromised or elderly individuals. However, CREC appears to cause a higher incidence of recurrent infections with the same strain, indicating that it is more challenging to eradicate. This increased persistence may stem from several factors, including the bacterium’s ability to acquire and retain resistance genes, adapt to diverse host environments, and evade immune responses. The high rate of recurrent infections caused by CREC might also reflect the characteristics of the patient population in this study, which likely includes individuals with compromised immune systems or chronic conditions, making them more vulnerable to persistent infections. As we found, the mean age of the 18 patients was 79.3 years, and only patients 18, 20, 21, and 22 had no history of chronic diseases. The capacity of CREC to establish long-term infections poses significant challenges for treatment strategies, as it suggests that standard antibiotic regimens may be insufficient, leading to prolonged and recurrent disease. For recurrent infections caused by the same strain, it is crucial to combine the action of the immune system with antibiotic therapy and identify the niche of the strain within the host to develop effective treatment approaches.
The distribution of phylogroups is shaped by the source and antibiotic susceptibility of the bacterial isolates, as well as the geographic regions where they were obtained [4, 50]. Phylogroups B1 and A are typically considered less virulent compared to phylogroups B2 and D, which often carry more virulence factors and are frequently associated with extraintestinal infections such as UTIs and sepsis [23, 51, 52]. Overall, phylogroup A dominates in CREC [4]. However, in Asian countries, phylogroup F and A account for the majority of CREC in China and Thailand, respectively [4]. Among carbapenemase-producing E. coli isolates in Iran, urinary isolates primarily belonged to phylogroups B2 (41.7%) and D (25%), while other clinical isolates were classified into B1 (25%) and A (8.3%) [53]. In this study, focusing on CREC repeatedly isolated from the same host, we found phylogroup B1 and A to be the most common. However, the differences in characteristics between these strains and other B1 and A strains warrant further investigation to understand how they enhance their ability to cause persistent infections and resist treatment.
The clinical significance of E. coli ST131 and ST1193 lies in their association with antimicrobial resistance and virulence [15, 54]. ST131 is a globally prevalent, high-risk clone frequently linked to extended-spectrum β-lactamases (ESBLs), such as CTX-M enzymes, and fluoroquinolone resistance. Comparative genomic analyses have revealed that ST131 strains possess unique virulence traits and exhibit multidrug resistance, contributing to their successful dissemination and the complexity of treating infections they cause [55]. Similarly, ST1193 has emerged as a significant fluoroquinolone-resistant clone within phylogenetic group B2 [54, 56]. Studies indicate that ST1193 shares characteristics with ST131, including multidrug resistance and potential for widespread transmission [54]. These two sequence types were also found in our patients, suggesting that their virulence may be associated with recurrent or multi-site infections. Additionally, recent studies have identified ST617 and ST410 as prevalent in CREC, with carbapenemase genes such as blaNDM being the most common [4], consistent with our findings. The global spread of these sequence types also requires close monitoring.
A limitation of this study lies in the method of strain collection. Similar to other studies on bacterial collection, we only gathered a single colony from each clinical specimen. By selecting a single colony for analysis, we may have overlooked the presence of hetero-populations within the bacterial community, which could include subpopulations with varying resistance profiles and virulence characteristics. This phenomenon is evident in our phenotypic assays, where the strains collected from a single patient showed fluctuating susceptibility to antimicrobials and phages. These hetero-populations may undergo different evolutionary paths and enhance the overall survival rate of the bacteria under adverse conditions, such as antibiotic pressure. However, failing to account for this diversity could result in an incomplete understanding of the evolutionary dynamics of CREC within the host. The presence of multiple subpopulations might also explain some of the discrepancies observed in resistance patterns and virulence traits, as different subpopulations may respond differently to selective pressures. The ability of CREC strains to evolve in response to host conditions underscores the need for personalized treatment approaches that take into account the dynamic nature of bacterial populations within the host. Moreover, recent research has indicated that the transition of nosocomial carbapenem-resistant K. pneumoniae from hypervirulent strains to less harmful commensal bacteria, yet with a higher ability for cell attachment, is closely linked to mutations in the rmpA promoter region [57]. This finding underscores the dynamic interplay between bacterial genome mutations and the horizontal transfer of plasmids within the host.
Our phenotypic assays also emphasize the influence of antibiotic treatment history on the evolution of bacterial resistance. The use of specific antibiotics can drive the selection of resistant strains, leading to shifts in the resistance profile of bacterial populations over time [58]. These results emphasize the importance of careful antibiotic stewardship, as inappropriate or excessive use of antibiotics can accelerate the evolution of resistant strains, making infections more challenging to treat. Antibiotic-induced selection pressure not only affects immediate resistance profiles but can also contribute to changes in virulence and further dissemination of resistant strains within the patient and the broader community [59]. Monitoring antibiotic usage and its impact on resistance patterns is critical for devising effective treatment strategies and controlling the spread of resistance.
In the context of persistent bacterial infections, we found certain isolates exhibited reduced antimicrobial resistance or diminished serum resistance. This phenomenon can be linked to bacterial fitness and adaptive strategies within the host environment [60, 61]. Antimicrobial resistance mechanisms often impose fitness costs on bacteria, such as reduced growth rates or competitiveness in the absence of antibiotics [61, 62]. Consequently, the loss of resistance can enhance bacterial fitness under antibiotic-free conditions, facilitating persistence within the host [63]. For example, studies have demonstrated that E. coli strains with rifampicin resistance mutations exhibit reduced fitness; however, compensatory mutations can restore fitness without reverting resistance [64]. Serum resistance mechanisms, such as modifications to the bacterial outer membrane, enable evasion of host immune responses [65]. However, these modifications may carry fitness costs. Loss of serum resistance could result in structural changes that reduce immune recognition, thereby enhancing bacterial survival. For instance, alterations in LPS structure have been associated with decreased serum resistance but increased survival in certain host environments. Variations in LPS, particularly in the O-antigen, can help pathogens evade both humoral and cell-mediated host immune defenses. In summary, reductions in antimicrobial or serum resistance can be adaptive strategies that enhance bacterial fitness and persistence within the host. These adaptations may involve complex trade-offs between resistance mechanisms and overall virulence, warranting further investigation to elucidate their roles in chronic infections.
Since our samples were not all derived from urine, using only the 5637 cell line for cell adhesion and cytokine production assays may not fully represent the characteristics of other strains. In this study, the virulence score was limited to assessing whether the number of virulence factors varied within the same patient. However, due to the large number of strains, we did not examine genome or transcriptome differences among strains from the same patient. The expression levels of virulence genes or their regulators and post-translational modifications of proteins can significantly impact protein function and bacterial virulence [66, 67]. This was reflected in our findings, where patients 7, 8, and 24 carried the highest number of virulence factors, yet larvae survival rates remained high. This discrepancy may result from these genes not being highly expressed in vivo. Although this study utilized WGS to analyze the transfer of plasmids within the host and the genomic and phenotypic changes of bacteria, the impact of these genomic alterations on bacterial transcriptomes remains an important area for future research. Understanding how these genetic changes influence transcription could provide valuable insights into the bacterial adaptation process within the host and its gene expression dynamics. In particular, comparing bacterial transcriptomes in both in vitro and in vivo conditions could shed light on how transcriptional profiles shift during infection, helping to uncover key regulatory mechanisms that drive bacterial survival and pathogenicity within the host. Additionally, because the strains were derived from diverse sources, we did not perform animal experiments to assess toxicity in mice.
Strain 563 from patient 17 was more susceptible to phages than strain 588. We also observed a morphological shift on BAP, with the bacteria transitioning from a smooth to a mucoid appearance. However, phage resistance caused by mutations requires further investigation. Fang et al.. reported that phage-resistant mutants exhibited reduced capsule production and decreased virulence [68]. Mutations in genes such as mshA and wcaJ are involved in capsule polysaccharide synthesis, and mutations in epsJ, encoding exopolysaccharide synthesis, cause phage resistance. Due to the long-term colonization and potential accumulation of mutations, we focused on plasmid transfer rather than conducting SNP analysis through hybrid-WGS to study genome evolution within the host. Additionally, in this study, phage susceptibility is considered as an observation of a phenotype. We aim to explore how strain evolution within the host may influence changes in phage susceptibility. Therefore, the phage receptors, resistance mechanisms, and treatment outcomes after mouse infection were not further explored in this study.
The evolution of bacterial virulence within the host is shaped by complex adaptive mechanisms and genetic modifications driven by host-pathogen interactions, selective pressures, and genomic plasticity. Bacteria must continuously adapt to the host immune system, environmental stresses, and antimicrobial exposure, leading to dynamic alterations in virulence phenotypes. These adaptations may involve regulatory shifts in gene expression, loss or gain of virulence-associated genetic elements, and structural genome modifications. A key factor influencing virulence trends is the balance between pathogen fitness and host immune evasion. Attenuation of virulence has been observed in some strains, likely as a result of prolonged host colonization, where selective pressure favors reduced immune activation to promote persistent infection. For instance, studies on Salmonella enterica have demonstrated that chronic infections can drive the selection of variants with reduced virulence to evade host defenses and maintain long-term persistence [69]. Conversely, virulence can also be enhanced through horizontal gene transfer, particularly via the acquisition of virulence plasmids or genomic islands that encode factors promoting immune evasion, tissue invasion, or increased bacterial fitness. For example, hypervirulent K. pneumoniae strains have been shown to acquire large virulence plasmids, resulting in increased capsule production and enhanced pathogenic potential, which may contribute to more severe infections. The absence of a consistent phenotypic trajectory during prolonged colonization further underscores the highly context-dependent nature of bacterial virulence evolution. Variability in host immune responses, bacterial genetic background, and environmental pressures likely contribute to diverse infection outcomes, highlighting the need for further longitudinal studies to elucidate the molecular mechanisms underlying these adaptive processes.
In our study, we observed dynamic changes in plasmid carriage within bacterial populations in the host, where certain strains either lost or acquired plasmids. This phenomenon may be driven by selective pressures that favor the retention of plasmids conferring fitness advantages, such as enhanced virulence or improved stress tolerance, while dispensable plasmids are eliminated to reduce metabolic burden. The evolutionary trajectory of plasmid maintenance is shaped by the balance between fitness costs and benefits. A previous study discusses how plasmid-mediated gene acquisition facilitates bacterial adaptation by promoting the selection of advantageous traits, including antibiotic resistance and virulence [70]. Additionally, Harrison and Brockhurst highlight that plasmid stability is governed by host-plasmid compatibility and the dynamic interplay between horizontal transfer and selection pressures [71].
Lurie-Weinberger et al.. reported three cases of patients concurrently colonized by NDM-producing E. coli and K. pneumoniae, demonstrating interspecies transfer of the blaNDM-1 gene via an 87,450 bp IncM2 multidrug resistance plasmid. This highly conjugative plasmid highlights the risk of horizontal plasmid dissemination, which may drive the emergence of high-risk clones and under-recognized multi-species outbreaks [72]. In this study, we observed the potential for plasmid transfer between different strains within the same patient, a phenomenon of bacterial plasmid transfer and the subsequent spread of resistance within the host that has been previously reported [46, 73, 74]. However, whether plasmids are commonly shared among CREC strains remains uncertain. Interestingly, the IncI1-I(Alpha) plasmid in strain 573 from patient 18 was integrated into the chromosome. Our findings suggest that plasmid transfer between different CREC strains within the same host niche may occur in vivo, facilitating the acquisition of new traits and driving genome evolution. This process could enhance bacterial survival and pathogenicity, potentially explaining the observed variations in virulence and resistance profiles among different strains from the same patient. Bacterial fitness may also shape the population dynamics [75], potentially leading to the loss of non-essential plasmids [76]. However, we cannot definitively confirm plasmid transfer within the host, as these findings could also be attributed to the widespread prevalence of these plasmids.
It is worth mentioning that both CREC strains from patients 7 and 8 are ST131, and both carry the IncFIA and p0111 plasmids, along with other plasmids of similar sizes. The isolation periods of the strains from these two patients are also relatively close. Whether this strain is circulating within the hospital remains to be determined. Additionally, this strain can persistently colonize and be isolated from different patients, making the characteristics of these strains and the role of these two plasmids in bacterial virulence worthy of further investigation.
Although under our testing conditions, only the IncFIA plasmid from strain 455 was successfully transferred to E. coli C600 among the targeted 5 plasmids predicted to be transferable between strains, we also observed that the IncC plasmid from strain 540 could not be transferred. However, its IncB/O/K/Z plasmid, which carries blaCMY-2 but lacks tra genes, was still capable of transferring independently or co-transferring alongside the IncFIA plasmid. This observation is consistent with previous reports indicating that even if a single plasmid lacks tra genes, co-residing tra-carrying plasmids within the same bacterial host can facilitate the mobilization of tra-negative plasmids [77], potentially promoting the spread of antimicrobial resistance and virulence plasmids. Additionally, under our experimental conditions, we did not observe a direct impact of the transferred plasmids on bacterial virulence in the G. mellonella model. Nevertheless, the influence of these plasmids on other bacterial phenotypes and gene expression requires further investigation in future studies.
Recent professional research on CREC underscores its growing threat to public health due to the rapid spread of resistance genes, such as blaKPC, blaNDM, and blaOXA-48, often found on mobile genetic elements [78]. Advanced genomic techniques and rapid detection methods, including next-generation sequencing and PCR, enhance our ability to identify and monitor these pathogens [79‐82]. Moreover, innovative approaches, such as machine learning for surveillance and alternative treatments like phage therapy, are being explored to combat CRE and inform effective antimicrobial stewardship programs. Integrating our findings within the broader context of CREC epidemiology has significant implications for infection control, antibiotic stewardship, and public health policy. The evolutionary dynamics of CREC strains highlight the necessity of adaptive infection control strategies tailored to mitigate the persistence and dissemination of these pathogens. Antibiotic stewardship programs must account for the evolutionary potential of CREC to develop resistance, ensuring that treatment regimens do not inadvertently drive further resistance selection. Effective coordination between infection control and antibiotic stewardship programs is essential to counteract the hospital environment’s role in facilitating the emergence and spread of MDR gram-negative bacteria. Furthermore, our study underscores the need for a deeper understanding of bacterial characteristics and niche adaptation within the host, particularly in patients with recurrent infections. A more comprehensive analysis of pathogen persistence mechanisms and host-microbe interactions is crucial for developing targeted therapeutic strategies aimed at complete pathogen eradication.
Conclusion
This study reveals the significant role of genetic changes in shaping bacterial phenotypes, allowing pathogens to adapt and persist within the dynamic host environment. The presence of different bacterial strains within the host may promote genetic exchange, driving genomic evolution and enhancing bacterial adaptability. This adaptability enables bacteria to survive under specific selective pressures from the host, facilitating the establishment of hetero-population dynamics and long-term infections.
Acknowledgements
The authors thank the En Chu Kong Hospital staff for their support in collecting CRE isolates. This work was supported by the Higher Education Sprout Project of the National Yang Ming Chiao Tung University and Ministry of Education (MOE), Taiwan.
Declarations
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Ethical approval and consent to participate
This study was approved by the En Chu Kong Hospital Institutional Review Board under the approval number ECKIRB1120801 and ECKIRB1131205.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
