Comprehensive genomic analysis of carbapenem-resistant Citrobacter freundii and characterization of a co-carrying bla_KPC−2 and bla_NDM−1 isolate in Zhejiang Province， China

Abstract

Background

Despite the clinical importance of carbapenem-resistant Citrobacter freundii (CRCF), its resistance genomics is poorly understood. This study systematically characterizes the genomic features and evolution of clinical CRCF isolates, with detailed analysis of a bla_KPC−2 and bla_NDM−1 co-harboring strain.

Methods

Clinical CRCF isolates were collected from multiple tertiary hospitals in China, and their basic clinical data were obtained. Phenotypic characterization was performed through antimicrobial susceptibility testing and plasmid conjugation transfer assays. Bioinformatic analysis was conducted based on whole-genome sequencing data.

Results

Clinical data from 37 CRCF isolates in Zhejiang Province, China, revealed that infections primarily occurred among elderly patients (≥ 70 years, 65%) and in high-risk departments such as the ICU. The majority of isolates were recovered from urine samples (67%). Antimicrobial susceptibility testing indicated high resistance to carbapenems and most β-lactams, whereas susceptibility to amikacin remained moderate (21.6%) and no resistance against tigecycline and colistin was detected. Resistance mechanism analysis revealed that 97.3% of isolates carried bla_NDM, and bla_CMY was present in all strains. IncX3 (59.5%), IncFIB (54.1%), and IncFII (51.4%) were the predominant plasmid types. MLST identified 17 sequence types, with ST22 and ST116 being most prevalent. Both STs carried bla_NDM−1, clustered phylogenetically, and were persistently isolated from 2019 to 2024, indicating long-term clonal circulation. Complete genome analysis of a rare bla_KPC−2 and bla_NDM−1 co-harboring strain (T1001) confirmed that both genes were located on conjugative plasmids, with IS26 and ISKpn19 facilitating their integration.

Conclusion

The dissemination of CRCF in Zhejiang Province, China, is driven by both clonal expansion of successful lineages (ST22, ST116) and horizontal transfer of resistance plasmids. These clones, persisting from 2019 to 2024, carry bla_NDM−1 and show high carriage rates of IncX3 and IncFIB plasmids, suggesting a possible association between these replicon types and the dominant clones. Although amikacin, tigecycline, and colistin remain effective in vitro, the complex recombination mechanisms of resistance genes—particularly IS26/ISKpn19-mediated modular dissemination—pose a serious nosocomial threat. Enhanced infection control measures are urgently needed to curb the spread of these multidrug-resistant bacteria and their resistance plasmids.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12866-026-05080-x.

Introduction

The persistent increase in carbapenem-resistant Enterobacteriaceae (CRE) infections poses a severe challenge to global public health security. These bacteria exhibit broad resistance not only to carbapenems but also to a wide range of other antibiotics. Their rapid global spread has significantly complicated clinical treatment and heightened patient mortality risk. While clinically common CRE species include Klebsiella pneumoniae and Escherichia coli, antimicrobial resistance in other members of the Enterobacteriaceae family, such as Citrobacter spp., is often overlooked [1, 2].

As an important member of the Citrobacter genus, Citrobacter freundii (C. freundii) is often considered a colonizing bacterium. However, it can act as an opportunistic pathogen causing nosocomial infections in specific populations, such as neonates, the elderly, and immunocompromised patients, potentially leading to life-threatening conditions [3]. In recent years, the detection rate of carbapenem-resistant Citrobacter freundii (CRCF) has been increasing, with related cases successively reported in hospitals across certain regions of China [4, 5]. Furthermore, multiple nosocomial outbreaks caused by this microorganism have been documented worldwide [6–8], highlighting its growing epidemiological threat.

The primary resistance mechanism of CRCF is the production of carbapenemases. To date, a variety of carbapenemase types have been reported in CRCF, including KPC [9], NDM [10], IMP [11], VIM [12], and OXA-48-like enzymes [13]. The prevalence of these enzymes exhibits significant geographical heterogeneity, with distinct types dominating in different regions. Notably, the emergence of strains co-harboring multiple carbapenemase genes has been increasingly documented. For instance, a strain co-producing NDM-1 and VIM-48 was reported in Tunisia [12]. In China, strains co-producing KPC-2 and NDM-1 have been identified across multiple regions [14–17]. Additionally, a strain carrying three carbapenemase genes (bla_KPC−2, bla_NDM−1, and bla_IMP−4) was identified in Henan Province [18]. This coexistence of multiple carbapenemase genes further expands the resistance profile of the strains, posing even greater challenges for clinical management.

To systematically investigate the epidemiological characteristics and resistance mechanisms of CRCF in Zhejiang Province, China, this study collected clinical isolates from multiple tertiary hospitals in the area. A comprehensive analysis was conducted on their clinical distribution, resistance phenotypes, resistance genes, plasmid types, molecular profiles, and genomic evolutionary patterns. Furthermore, an in-depth characterization was performed on one strain co-harboring the bla_NDM−1 and bla_KPC−2 genes, aiming to elucidate the genetic structure and evolutionary features of its mobile resistance elements. This work provides scientific evidence to support precise clinical control and rational antibiotic use against such resistant pathogens.

Materials and methods

Sample collection, species identification and phenotypic assays

During the period from 2016 to 2024, CRCF isolates were collected from inpatients in seven tertiary hospitals located in Zhejiang province, China (Table S1). Duplicate isolates from the same patient and the same specimen type were excluded. Corresponding patient demographic and clinical data, including gender, age, specimen type, and isolation department, were also collected. Initial identification of all isolates was performed using Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry (MALDI-TOF MS, bioMérieux, France). To further confirm species classification, genome sequences were analyzed by calculating Average Nucleotide Identity (ANI, FastANI v1.34 ) [19] against the C. freundii reference strain MSB1_1H-sc-2,280,393 (GenBank accession no.: LR890181). Isolates with an ANI value greater than 95% were classified as C. freundii. Antimicrobial susceptibility testing was conducted using the VITEK 2 Compact automated system (bioMérieux, France), and the results were interpreted according to the Clinical and Laboratory Standards Institute (CLSI) 2024 guidelines [20].

Sequencing and sequence assembly

Genomic DNA from all strains was extracted using the Gentra Puregene Yeast/Bacteria Kit (Qiagen, Valencia, CA, USA). Sequencing libraries were prepared with the TruePrepTM DNA Library Prep Kit V2 for Illumina and the SQK-LSK109 Ligation Sequencing Kit for Nanopore. Libraries were then sequenced on an Illumina HiSeq X Ten platform (Illumina, San Diego, CA, USA) and a GridION X5 device (Oxford Nanopore Technologies, Oxford, UK). In order to improve the reliability of data processing, NanoFilt was used to filter the raw data of the GridION X5 platform to obtain high-quality clean reads. Efficient data was obtained by using fastp (https://github.com/OpenGene/fastp) to remove adapters from NovaSeq sequencing data and filter the low quality. Paired-end short Illumina reads and the long Nanopore reads were assembled de novo utilizing Unicycler (v0.5.1) (https://github.com/rrwick/Unicycler).

Screening for antibiotic resistance genes (ARGs), multi-locus sequence typing (MLST) profiles, and plasmid types

Acquired antibiotic resistance genes were identified using abricate v2.0.1 (run on Linux platform) with the ResFinder database (database version 2023-01-18, https://cge.cbs.dtu.dk/services/ResFinder/), applying thresholds of 90% coverage and 90% identity. A heatmap of detected antibiotic resistance genes was generated using the pheatmap package (version 1.0.12) in R (version 3.27.1; https://www.r-project.org). MLST was performed by comparing sequences of seven housekeeping genes (arcA, aspC, clpX, dnaG, fadD, lysP, mdh) against the Citrobacter PubMLST database (http://pubmlst.org) to assign sequence types (STs). Plasmid replicon typing was performed using a customized database, which was constructed by supplementing the PlasmidFinder database (https://cge.food.dtu.dk/services/PlasmidFinder/) with additional plasmid sequences derived from our in-house sequencing data.

Phylogenetic analysis

The complete chromosomal sequence of the reference strain MSB1_1H‑sc‑2,280,393 (GenBank accession LR890181) was aligned with all strain genomes using MUMmer v3.2 (https://mummer.sourceforge.net/) to identify core single‑nucleotide polymorphisms (SNPs). Repetitive DNA regions were detected with RepeatMasker (http://www.repeatmasker.org/), and SNPs within these regions were excluded from subsequent analysis. The remaining recombination‑filtered core SNPs were used to construct a maximum‑likelihood phylogenetic tree under the GTR model with 1000 bootstrap replicates in RAxML (v8.2.13). Citrobacter koseri ATCC BAA‑895 (accession CP000822) was included as an outgroup. The resulting tree was visualized using the Interactive Tree of Life (iTOL) online tool (https://itol.embl.de).

Genomic annotation and comparative analysis

Genome annotation was initially performed using the RAST 2.0 online platform server to obtain a baseline prediction. To ensure high-resolution annotation of mobile genetic elements and resistance gene contexts, we subsequently performed manual curation using the following databases: CARD [21], ResFinder [22], ISFinder [23], and DANMEL [24]. The comparative structure of plasmids, mobile elements and features of resistance regions were visualized using Inkscape (https://inkscape.org).

Conjugal transfer

A conjugation assay was conducted using sodium azide-resistant Escherichia coli J53 as the recipient strain and C. freundii T1001 as the donor strain. Single colonies of donor strain T1001 and recipient strain J53 were separately inoculated into 5 mL of LB broth and incubated at 37 °C with shaking at 200 rpm for 12–18 h. Subsequently, 20 µL of donor culture and 20 µL of recipient culture were mixed and transferred into 4 mL of fresh LB broth, followed by static incubation at 35 °C for 18 h. A volume of 100 µL of the overnight mating mixture was evenly spread onto selective Mueller‑Hinton agar plates supplemented with 200 µg/mL sodium azide (to select against the J53 recipient) and 0.5 µg/mL imipenem (to select for transconjugants carrying bla_NDM‑1 and bla_KPC‑2 genes). After incubation for 12–18 h, suspected transconjugant colonies were identified by MALDI-TOF MS; transfer of resistance genes and plasmid replication initiation gene repA was assessed by PCR and antimicrobial susceptibility testing.

Nucleotide sequence accession number

The complete genome sequence of strain T1001, comprising its chromosome and eight plasmids, has been deposited in the GenBank database under accession numbers CP186042-CP186050. The accession numbers for the remaining 36 strains are provided in Table S1.

Results

Species identification and clinical characterization

A total of 37 CRCF isolates were collected from multiple tertiary hospitals in Zhejiang Province, China. All 37 isolates were sequenced using Illumina technology. Among them, only strain T1001 was found to carry both bla_NDM−1 and bla_KPC−2. Subsequently, this strain was subjected to further sequencing on the Oxford Nanopore platform to facilitate detailed genomic characterization. The sequencing depth and coverage for these genomes were > 100× and > 98%, respectively. Average Nucleotide Identity (ANI) values between the isolates and the C. freundii reference genome (accession no. LR890181) ranged from 98.39% to 99.04% (mean 98.74%), confirming their identification as C. freundii.

Demographic data indicated a predominance of male patients (n = 26, 70%) over female patients (n = 11, 30%) (Fig. 1A). Patient ages ranged from 45 to 85 years, with a majority being elderly: 65% (n = 24) were aged ≥ 70 years. The 70–79‑year‑old group represented the largest proportion (n = 14, 38%) (Fig. 1B). Urine specimens accounted for most isolates (n = 25, 67%), followed by sputum (n = 5, 13%). The remaining isolates were obtained from blood (n = 2), wound secretion (n = 1), vaginal secretion (n = 1), stool (n = 1), throat swab (n = 1), and ascites (n = 1) (Fig. 1C). In terms of ward distribution, the highest number of isolates came from the Intensive Care Unit (ICU, n = 6), followed by the Department of Urology (n = 5), Gastroenterology (n = 4), and Neurology (n = 4) (Fig. 1D).

Fig. 1.

Distribution of clinical characteristics of 37 C. freundii strains (n = 37). A Pie chart showing patient gender distribution: males (70%) and 11 females (30%). B Bar chart showing patient age distribution. C Pie chart showing the distribution of sample sources. D Pie chart showing the distribution of the C. freundii isolates recovered from hospital departments

Antimicrobial susceptibility profiling

Antimicrobial susceptibility testing was performed on all 37 CRCF isolates (Table S2). All isolates were resistant to meropenem, and 97.3% were resistant to imipenem, indicating high‑level resistance to carbapenems. Among other β‑lactam agents, resistance rates were as follows: ceftazidime 97.3%, piperacillin‑tazobactam 100%, ticarcillin‑clavulanic acid 100%, and cefoperazone‑sulbactam 94.6%. In contrast, cefepime and aztreonam exhibited lower resistance rates of 59.5% and 43.2%, respectively.

Regarding fluoroquinolones, 62.2% of isolates were resistant to ciprofloxacin and 43.2% to levofloxacin. Sulfamethoxazole resistance was observed in 54.1% of isolates. Among aminoglycosides, resistance to amikacin was 21.6% and to tobramycin 27.0%, representing relatively lower resistance levels. Notably, all isolates remained susceptible to tigecycline and colistin.

Plasmid replicon profiles and distribution of antibiotic resistance genes in 37 CRCF Isolates

Screening for plasmid replicons in the genomic data identified IncX3 (22/37, 59.5%), IncFIB (20/37, 54.1%), and IncFII (19/37, 51.4%) as the most prevalent types, followed by IncC (12/37, 32.4%). Other replicon types including IncHI2, IncN1, IncN3, IncQ1, IncR, and IncU were also detected but at lower frequencies (Fig. 2).

Fig. 2.

Heatmap of plasmid replicon types distribution in 37 C. freundii strains. Presence of a plasmid replicon is indicated by yellow boxes, while absence is represented by grey boxes. Strain names are listed on the right (e.g. S3026, or N14), and plasmid replicon types are labeled at the bottom

Analysis of ARGs across the 37 isolates revealed a diverse resistome. A total of 64 distinct ARGs were identified, associated with 10 antibiotic classes: aminoglycosides, β-lactams, chloramphenicols, fosfomycins, macrolides, quinolones, rifamycins, sulfonamides, tetracyclines, and trimethoprim, as well as quaternary ammonium compounds. The highest diversity of resistance genes was observed for aminoglycosides (n = 17) and β-lactams (n = 16).

Carbapenem resistance was primarily associated with the presence of bla_NDM-1 (29/37, 78.4%), bla_NDM‑5 (7/37, 18.9%), and bla_KPC‑2 (2/37, 5.4%). Notably, isolate T1001 co‑carried both bla_KPC‑2 and bla_NDM‑1. The AmpC‑type β‑lactamase gene bla_CMY was detected in all 37 isolates. Other frequently identified genes included sul1 (26/37, 70.3%) and mph(A) (20/37, 54.1%) (Fig. 3).

Fig. 3.

Heatmap of antimicrobial resistance gene distribution across 37 C. freundii strains, outgroup strain and the reference strain. Each row represents a bacterial strain, and each column represents a specific antimicrobial resistance gene. The presence of a gene is indicated by red square, and its absence by a blue square

Multilocus sequence typing and phylogenetic analysis of 37 CRCF isolates

MLST identified a total of 17 distinct sequence types (STs) among the 37 isolates (Table S1). The most prevalent types were ST22 (9/37, 24.3%) and ST116 (6/37, 16.2%), followed by ST1098 (4/37, 10.8%) and ST91 (2/37, 5.4%) (Fig. 4, Table S1). Phylogenetic analysis revealed that the isolates formed two major clusters (Cluster 1 and Cluster 2). The SNP differences within Cluster 1 range from 86 to 16,007, within Cluster 2 range from 1 to 17,003, and between Cluster 1 and Cluster 2 range from 18,959 to 22,193 (Fig. S1). Isolate T1001, which co‑carried bla_KPC‑2 and bla_NDM‑1, was located within Cluster 2. An association was observed between specific STs and phylogenetic clustering. For instance, ST22 and ST116, both of which carried bla_NDM−1, were predominantly found within the same phylogenetic cluster and were persistently isolated from 2019 to 2024. This indicates the long-term circulation of these successful clones in the region.

Fig. 4.

Phylogenetic analysis of 37 C. freundii clinical isolates. The maximum-likelihood phylogenetic tree was constructed based on core genome alignment. Citrobacter koseri ATCC BAA-895 (accession no. CP000822) was used as an outgroup, and the C. freundii reference strain MSB1_1H-sc-2,280,393 (accession no. LR890181) is included for comparison. The tree is annotated with the following characteristics (from inner to outer rings): (1) Colored ranges indicating the two major clusters; (2) Patient gender; (3) Carbapenem resistance genotypes; (4) Sequence types (STs); and (5) Isolation time

Genomic overview of C. freundii T1001

Long‑read whole‑genome sequencing of the dual carbapenemase‑producing isolate T1001, carrying both bla_KPC‑2 and bla_NDM‑1, generated a complete genome assembly. The genome comprises one chromosome of 5,248,772 bp (GenBank accession CP186042) with a GC content of 51.7%, encoding 4,973 predicted open reading frames (ORFs). Six resistance genes (aadA1, aacA4, bla_CMY‑48, qnrS1, arr‑3, and dfrA1) were identified on the chromosome. The isolate was assigned to sequence type ST1098 (Table 1).

Table 1.

Whole genome information of T1001

Strain	Sequence	Mean G + C content (%)	Length (bp)	Total number of ORFs	MLST	Inc type	Accession number	Resistance genes
T1001	cT1001	51.7	5,248,772	4,973	ST1098	-	CP186042	aadA1, aacA4, blaCMY−48, qnrS1, arr-3, dfrA1
	pT1101-KPC	49.8	45,707	57	-	Untypable	CP186043	bla KPC−2
	pT1101-NDM	52.4	57,078	74	-	N1	CP186044	blaNDM−1, dfrA14
	pT1131-armA	51.1	154,852	189	-	C	CP186050	aphA1, armA, catB8, sul1
	pT1131-NR1	53.5	94,374	126	-	FII: FIB	CP186045	-
	pT1131-NR2	58.6	14,762	20	-	Untypable	CP186046	-
	pT1131-NR3	52.1	5,409	6	-	Untypable	CP186047	-
	pT1131-NR4	54.9	4,921	5	-	Untypable	CP186048	-
	pT1131-NR5	51.5	4,665	4	-	Untypable	CP186049	-

-, not applicable

In addition, eight circular plasmids were identified in T1001, three of which are resistant plasmids: pT1001‑KPC (accession CP186043), pT1001‑NDM (accession CP186044), and pT1001‑armA (accession CP186050).

pT1001‑KPC is 45,707 bp in length with a GC content of 49.8% and contains 57 ORFs. This plasmid carries only the bla_KPC‑2 gene and could not be assigned to a known replicon type (Untypable). Following a BLAST search of the plasmid replicon sequence against the GenBank database, a total of 24 related plasmids were identified. Among these, 23 were found to carry the bla_KPC−2 gene. All of these plasmids were identified in various genera of Enterobacterales isolated from different provinces in China, with 14 originating from Zhejiang Province. A subsequent BLAST comparison of the complete pT1001-KPC plasmid sequence revealed that 19 of these 23 plasmids exhibited high similarity (identity ≥ 99%, coverage ≥ 90%) (Table S4). Sequence analysis revealed that the plasmid contains a conserved backbone and an inserted segment carrying bla_KPC−2. A 13.9-kb conjugation transfer region was identified within this segment, which was found to be structurally complete, containing 14 open reading frames (ORFs) that encode a full set of transfer-associated proteins.

pT1001‑NDM is 57,078 bp with a GC content of 52.4% and 74 ORFs. It carries bla_NDM‑1 and dfrA14 and belongs to the IncN1 incompatibility group. Besides the backbone, the plasmid contains two inserted resistance regions: the bla_NDM‑1 region and the dfrA14 region. These insertions split the fipA gene associated with conjugation, thereby fragmenting the transfer region.

pT1001‑armA is 154,852 bp in length (GC 51.1%, 189 ORFs) and carries aphA1, armA, catB8, and sul1. It belongs to the IncC2 group and harbors four inserted regions: IS5D, a multidrug‑resistance (MDR) region, an ars locus‑IS1F segment, and an insertion sequence belonging to the IS3 family (Table 2; Fig. 5).

Table 2.

Whole genome information of T1001

Strain	Sequence	Mean G+C content (%)	Length (bp)	Total number of ORFs	MLST	Accession number	Resistance genes
T1001	cT1001	51.7	5,248,772	4,973	ST1098	CP186042	aadA1, aacA4, blaCMY−48, qnrS1, arr-3, dfrA1
	pT1101-KPC	49.8	45,707	57	-	CP186043	bla KPC−2
	pT1101-NDM	52.4	57,078	74	-	CP186044	blaNDM−1, dfrA14
	pT1131-armA	51.1	154,852	189	-	CP186050	aphA1, armA, catB8, sul1
	pT1131-NR1	53.5	94,374	126	-	CP186045	-
	pT1131-NR2	58.6	14,762	20	-	CP186046	-
	pT1131-NR3	52.1	5,409	6	-	CP186047	-
	pT1131-NR4	54.9	4,921	5	-	CP186048	-
	pT1131-NR5	51.5	4,665	4	-	CP186049	-

-, not available

Fig. 5.

A Circular maps of pT1001-KPC; B Circular maps of pT1001-NDM; C Circular maps of pT1001-armA. Genes are denoted by arrows and colored based on gene function classification. The innermost circle presents GC-skew [(G-C)/(G + C)], with a window size of 500 bp and a step size of 20 bp. Next-to-innermost circle presents GC content. The third inner circles black denotes plasmid backbone regions, and gray denotes exogenous insertions

Analysis of mobile genetic elements harboring resistance genes in C. freundii T1001

The genetic environment of key resistance determinants in T1001 was analyzed in detail. The bla_NDM‑1 region in plasmid pT1001‑NDM spans 13.55 kb and is flanked by IS26 sequences. Internally, it consists of a ISKpn19‑mediated resistance unit that carries a truncated transposon Tn125 (Fig. 6a). The truncated Tn125 was captured by ISKpn19 and subsequently mobilized by IS26. The bla_KPC‑2 region in plasmid pT1001‑KPC is 16.66 kb in length. Its core structure corresponds to the transposon Tn6296, which carries bla_KPC‑2. The right‑side truncated mcp gene is absent and replaced by an ISKpn19 insertion (Fig. 6b).

Fig. 6.

Comparative genetic analysis of mobile genetic elements harboring resistance genes in C. freundii T1001. Genes are denoted by arrows. Genes, mobile elements and other features are colored based on their functional classification. Shading denotes regions of homology (nucleotide identity ≥ 95%)

The multidrug‑resistance (MDR) region of plasmid pT1001‑armA extends over 18.66 kb and contains four resistance genes: aphA1, armA, catB8, and sul1. The aphA1 gene is carried within the composite transposon Tn4352, while the other resistance genes are primarily associated with the insertion sequence ISCR1 (Fig. 6c).

The chromosome cT1001 harbors two principal resistance modules: one is an ISKpn19‑mediated unit carrying qnrS1, aacA4, and arr‑3; the other is the unit transposon Tn7, which carries aadA1 and dfrA1 (Fig. 6d).

Conjugative transfer of plasmids

A conjugation assay was performed using C. freundii T1001, which carries both bla_NDM-1 and bla_KPC-2, to evaluate the transferability of its resistance plasmids. Three distinct types of transconjugants were successfully obtained, as confirmed by colony PCR and antimicrobial susceptibility testing: (i) transconjugants harboring only the bla_NDM-1-positive plasmid (transfer frequency 7.61 × 10^− 3), (ii) those harboring only the bla_KPC-2-positive plasmid (transfer frequency 1.72 × 10^− 4), and (iii) those harboring both plasmids (pT1001_NDM and pT1001_KPC, transfer frequencies 8.34 × 10^− 6). This finding indicates that the two resistance plasmids are capable of both independent transfer and co-transfer into the same recipient E. coli J53 strain. Plasmid stability was assessed by serial passage of the three transconjugant types for 15 consecutive days without antibiotic selection. The plasmids were retained in the tested colonies after 15 passages, indicating that these plasmids are stable in E. coli J53 in the absence of selective pressure. All three transconjugants exhibited resistance to imipenem and meropenem (Table S3).

Discussion

In this study, a systematic phenotypic and genomic analysis was performed on 37 CRCF isolates collected from multiple tertiary hospitals in Zhejiang Province, China. Clinical data analysis revealed that infections primarily occurred in elderly patients (≥ 70 years old, accounting for 65%), with urine specimens being the predominant source (67%). The isolates were mainly distributed in high-risk departments such as the intensive care unit (ICU) and urology.

A detailed analysis of the antimicrobial susceptibility profiles and resistance gene repertoires revealed that, despite showing high-level resistance to carbapenems and most β-lactams, the isolates displayed moderate resistance to amikacin and showed no resistance against tigecycline and colistin. However, despite the favorable in vitro activity of amikacin and colistin, these agents are not generally recommended as first-line therapy for urinary tract infections or as monotherapy against carbapenem-resistant strains, due to concerns over efficacy, toxicity, and the risk of treatment failure [25].

From the perspective of resistance mechanisms, carbapenem resistance in this study was almost entirely mediated by the bla_NDM genes. This contrasts with a previous epidemiological study of CRCF in China, which reported KPC enzymes as the predominant carbapenemase type (9/17, three ST396, three ST116, and one each of ST111, ST62, and ST18) [26]. Notably, the earlier investigation primarily included isolates from northern China, whereas the current collection originated from the southeast region. This suggests that carbapenemase determinants in clinical Citrobacter freundii isolates from China may exhibit geographical variation and the circulation of distinct clonal strains. Furthermore, the ubiquitous presence of the bla_CMY gene (encoding an AmpC β-lactamase) across all isolates may contribute to the intrinsic resistance of C. freundii to first- and second-generation cephalosporins, such as cefazolin and cefuroxime. The high prevalence of sul1 (70.3%), which typically comprises the 3’-conserved segment of class 1 integrons, corresponded with the detection of integrase genes in 21 out of 37 isolates (56.8%), confirming the widespread presence of integron structures in this collection. These integrons are frequently linked to epidemic plasmids or transposons, significantly contributing the accumulation and dissemination of resistance genes [27].

Plasmid replicon analysis identified IncX3, IncFIB, IncFII and IncC as the most prevalent plasmid types in this study. Due to their high conjugation efficiency and broad host range, IncX3-type plasmids are recognized as major vectors for bla_NDM and other carbapenemase genes, significantly contributing to the global spread of NDM-type carbapenemases [28]. In this study, all strains harbored the bla_NDM gene. However, due to the inherent limitations of short-read sequencing, complete plasmid assembly was not achievable, precluding the definitive assignment of the bla_NDM gene to a specific replicon type. Nevertheless, based on the high frequency of the IncX3 replicon among these bla_NDM-positive isolates and its established role as a primary vehicle for this gene, it is plausible to infer that IncX3 represents the most common plasmid lineage harboring bla_NDM in this dataset (Fig. 2). In contrast, IncF-type plasmids are known for their ability to carry large resistance gene islands and frequently recombine with chromosomal DNA, often being associated with multidrug-resistant phenotypes [29]. The co-existence of these prevalent plasmid types facilitates the horizontal transfer of various resistance genes, including carbapenemase genes, among strains and even across different bacterial species.

MLST revealed that ST22 and ST116 were the dominant sequence types. Both carried bla_NDM-1, clustered together phylogenetically, and were persistently recovered from 2019 to 2024. This finding indicates that, in addition to horizontal transfer via plasmids and mobile genetic elements, clonal dissemination has played a substantial role in the regional spread of CRCF. Notably, both ST22 and ST116 primarily carried the IncX3 plasmid. In addition, ST116 was also found to predominantly harbor the IncFIB plasmid (Fig. 2).

In this study, an in‑depth genomic analysis of the clinical C. freundii isolate T1001, which co-carries bla_KPC-2 and bla_NDM-1, revealed a complex genetic architecture. The bla_NDM-1-bearing plasmid pT1001‑NDM belongs to the IncN1 type, a well‑documented and efficient vector for the global dissemination of carbapenemase genes [30, 31]. Structural examination of this plasmid indicated that the insertion of the bla_NDM‑1 and dfrA14 regions resulted in truncation of the fipA gene. Disruption of the fipA gene has been reported in several IncN1-type plasmids. Its encoded protein functions in fertility inhibition of IncP plasmids, and previous studies have demonstrated that the fipA gene is not essential for conjugative transfer. Consistent with this, our conjugation assays confirmed that pT1001-NDM remains transmissible [32, 33]. The plasmid pT1001-KPC (45,707 bp) exhibits a compact structure and carries only the bla_KPC‑2 resistance gene. Its conjugation region is intact, and transfer experiments confirmed its potential for horizontal dissemination.

The mechanisms underlying the capture and dissemination of resistance genes are notably exemplified in the genome of T1001. Analysis revealed that insertion sequences IS26 and ISKpn19 play a central role in assembling multi‑drug resistance gene modules. IS26 is known to continuously integrate resistance modules from diverse sources, forming dense and overlapping resistance gene arrays on bacterial plasmids or chromosomes [34]. Similarly, ISKpn19 has been reported to contribute to the formation and spread of structurally complex resistance units [35, 36]. In plasmid pT1001‑NDM, bla_NDM‑1 is located within a canonical module bounded by IS26, while its internal structure consists of an ISKpn19‑mediated truncated Tn125 transposon. Likewise, in pT1001‑KPC, bla_KPC‑2 is embedded in a Tn6296 transposon, with structural variations also involving ISKpn19. These observations suggest that ISKpn19 serves as a key driver in capturing and rearranging carbapenemase‑gene‑bearing modules. On the chromosome, ISKpn19 integrates a resistance unit containing qnrS1, aacA4, and arr‑3. In contrast, the MDR region of pT1001-armA contains two ISCR1 elements. ISCR1 is characterized by a rolling-circle replication mechanism that allows it to capture and co-mobilize downstream resistance genes [37, 38]. Therefore, the presence of these ISCR1 elements is likely associated with the carriage of adjacent resistance determinants, including armA, catB8, and sul1, on plasmid pT1001-armA.

Conclusion

Our results indicate that carbapenem resistance among C. freundii clinical isolates in Zhejiang Province, China, is primarily mediated by the production of NDM‑type carbapenemases, mobilized by a diverse repertoire of mobile genetic elements and integron structures. Predominant plasmid types (IncX3, IncF, and IncC) are critical vectors for bla_NDM dissemination. Furthermore, in‑depth analysis of a rare clinical isolate co‑carrying bla_KPC‑2 and bla_NDM‑1 revealed that insertion sequences, notably IS26 and ISKpn19, could play a central role in organizing and disseminating modular resistance units containing carbapenemase genes. Together, these findings provide a molecular and epidemiological foundation for guiding clinical and public health strategies aimed at controlling the spread of CRCF.

This study has several main limitations. First, our collection of isolates was limited to a specific area/region, which may not capture the full picture of the regional genomic epidemiology of CRCF. Second, as most isolates were sequenced solely by second-generation sequencing technologies, this prevented the differentiation of carbapenem resistance plasmid types.

Abbreviations

ANI

Average Nucleotide Identity

CLSI

Clinical and Laboratory Standards Institute

SNPs

Single-nucleotide polymorphisms

ARGs

Antibiotic Resistance Genes

MLST

Multi-Locus Sequence Types

ORFs

Open reading frames

C. freundii

Citrobacter freundii

CRCF

Carbapenem-resistant Citrobacter freundii

Authors' contributions

Conceptualization, XH.L., DR.D., J.Z and YL.Z.; methodology, XH.L, JB.Y., YL.Z. , J.L., HM.C., and YT.J.; data analysis, XH.L., YL.Z., DR.D and J.L.; resources, J.Z., JB.Y. and DR.D.; writing-original draft, XH.L and YL.Z.; writing-review and editing, DR.D and J.Z. All authors read and approved the final manuscript.

Funding

This work was supported by the Medical and Health Science and Technology Project of Zhejiang Province (2025KY461 and 2025HY1462), the Natural Science Foundation of Zhejiang Province (LTGY23H190003) and the Foundation of Taizhou Science and Technology Bureau (23ywa22) .

Data availability

The datasets generated and analyzed during the current study are available in the public NCBI GenBank database, under the following accession numbers: GCA_052077515.1, GCA_052071035.1, GCA_052159415.1, GCA_052070995.1, GCA_054254375.1, GCA_053902465.1, GCA_053902485.1, GCA_053902505.1, GCA_053902525.1, GCA_052076245.1, GCA_052070955.1, GCA_052070975.1, GCA_052072345.1, GCA_052071015.1, GCA_052071075.1, GCA_052070915.1, GCA_052077575.1, GCA_052159455.1, GCA_052159435.1, GCA_052159375.1, GCA_052159475.1, GCA_052159515.1, GCA_052159495.1, GCA_052174295.1, GCA_052174715.1, GCA_052174275.1, GCA_052174315.1, GCA_052174355.1, GCA_052174335.1, GCA_052174795.1, GCA_052174765.1, GCA_052174755.1, GCA_052174835.1, CP186042, GCA_052174635.1, GCA_052174655.1, GCA_052174815.1.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki. The use of human specimens and all related experimental protocols was reviewed and approved by the Ethics Committee of Taizhou Municipal Hospital, Zhejiang, China, in accordance with the medical research regulations of the Ministry of Health, China. Research and all related procedures involving biohazardous materials were approved by the Biosafety Committee of Taizhou Municipal Hospital. The Ethics Committee of Taizhou Municipal Hospital, Zhejiang, China granted a waiver of the requirement for informed consent for this study. This research was conducted in China.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.