Klebsiella spp. used in this work were isolated from wastewater collected from a wastewater treatment facility between 2019 and 2020 from Qilu Hospital in Qingdao, China, as described in our previous publication (Li et al., 2022). Klebsiella strains were purified by growth overnight at 37°C on MacConkey agar plates without antibiotics.

Antibiotic MICs were determined using the broth microdilution method in 96-well plates, following the Clinical and Laboratory Standards Institute (CLSI) and the European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines. The antibiotics tested included Imipenem, Tigecycline, Polymyxin E, Ampicillin, Cefotaxime, Kanamycin, Streptomycin, Trimethoprim, Tetracycline, Ciprofloxacin, Gatifloxacin, Meropenem, and Sulfisoxazole. Escherichia coli ATCC 25922 was used as a quality control strain for antimicrobial susceptibility testing.

The genomic DNA of Klebsiella strains was extracted using the TIANamp DNA Kit (Tiangen Biotech (Beijing) Co., Ltd., Beijing, China) according to the manufacturer’s instructions. The purity and quantity of DNA were determined using a Qubit™ 4.0 fluorometer (Thermo Fisher Scientific, MA, US). The sequencing library was constructed from 150 ng of genomic DNA using the Oxford Nanopore rapid barcoding kit SQK-RBK114.96. It was sequenced on the Nanopore p2solo platform (Oxford Nanopore Technologies, Oxford, UK) with an R10.4.1 flow cell. To obtain raw data, Basecalling was performed using Dorado version 0.5.3 (https://github.com/nanoporetech/dorado/).

The analysis of whole-genome sequencing (WGS) data for all isolates was conducted utilizing various bioinformatics tools. Flye version 2.8.1-b1676 was employed to assemble the genomes from long reads (Kolmogorov et al., 2019) and to determine sequence circularity. Additionally, Quast version 5.0.2 and CheckM2 version 1.0.2 were utilized to assess the assembly’s quality and completeness and check for contamination (Mikheenko et al., 2018; Chklovski et al., 2023). GTDB-Tk version 2.1.1 was used to determine the taxonomic classification of genomes. The genomes were annotated with the Prokaryotic Genome Annotation Pipleline (Tatusova et al., 2016). AMRFinder version 3.11.26 and Plasmidfinder version 2.1.1 were utilized to identify antimicrobial resistance genes (ARGs) and plasmid replicon types (Carattoli et al., 2014; Feldgarden et al., 2019). Multilocus sequence typing (MLST) and serotypes were identified using kleborate.

Raw reads were subjected to quality control using fastp (v0.23.4). Adapter sequences and low-quality bases (Q < 20) were trimmed, and reads shorter than 500 bp after trimming were removed. The quality of clean reads was checked with FastQC, ensuring that >90% of bases reached Q30. Genome assemblies were subsequently evaluated with QUAST (v5.2.0) to obtain metrics including N50, GC content, and genome size. Assembly completeness and contamination were assessed using CheckM (v1.0.2).

Heavy metal resistance assays were performed by agar dilution using Mueller-Hinton Agar plates. Overnight cultures of strains were diluted and adjusted to an OD₆₀₀ value of 0.08–0.1 using LB medium. A 5 µL aliquot of the diluted cultures was inoculated onto MH Agar plates containing various concentrations of metals. Metal salts, including K₂TeO₃, CuSO₄, CoCl₂, AgNO₃, K₂Cr₂O₇, NiSO₄, and HgCl₂, were added to the media, and the plates were incubated at 37 °C for 24 hours. Experiments were performed with three (n=3) independent technical replicates. K. pneumoniae ATCC13883 was used as the control strain.

Thirty-three Klebsiella isolates were obtained from the wastewater of Qilu Hospital Qingdao as part of a large-scale surveillance of wastewater bacterial communities previously reported by our group (Li et al., 2022). Four of the isolates, 1-76, 2-28, 3-82, and 3-88, are Klebsiella quasipneumoniae strains, while the remaining strains are K. pneumoniae strains. Third generation Nanopore sequencing was performed to obtain the whole genome sequences of these isolates. The long reads of this technology led to the assembly of high quality genomes, with the generation of near-perfect plasmid maps. All 33 isolates have their chromosomes well assembled to their circular forms, with the sizes of 5.34 ± 0.11 Mb, in agreement with the common sizes of Klebsiella chromosomes. Only four of the isolates were free of plasmids, all of which are K. pneumoniae. Analysis of plasmid-containing Klebsiella isolates suggested that they belong to 24 genomically distinct strains, as determined by whole-genome sequence similarity and assigned LIN codes using the Pathogenwatch cgMLST-based classification system (Supplementary Table 1). These strains were subject to further studies. Features of these strains are documented in Table 1. These strains belong to various and diverse sequence types and serotypes, showing high levels of heterogeneity. On average, they carry 3.4 plasmids per strain.

Among the 81 plasmids identified from the 24 Klebsiella strains, 28 (34.6%) were considered novel (Supplementary Table 1). Plasmid novelty was assessed by BLASTn comparison against the NCBI NT database, using <80% backbone sequence identity and <70% coverage as thresholds. The comparison was made with previously reported plasmids in public databases. These findings, consistent with our earlier report (Xu et al., 2024), suggest ongoing plasmid diversification and structural rearrangement within Klebsiella populations.

Fifteen of the 24 isolated Klebsiella strains carry antimicrobial resistance determinants on their plasmids (Supplementary Table 1), consistent with our previous finding that K. pneumoniae acquire antibiotic resistant plasmids for its antibiotic resistance (Li et al., 2018). A total of 22 such antibiotic resistant plasmids were identified. Five Klebsiella strains carried more than one antibiotic resistant plasmid, agreeing with the hypothesis that acquisition of multiple antibiotic resistant plasmids can lead to multidrug resistance in Klebsiella. To further evaluate their phenotypic resistance, minimum inhibitory concentrations (MICs) of representative antibiotics were determined for all strains. The MIC data, summarized in Table 2, reveal substantial variation in resistance levels among the isolates, with several strains exhibiting high-level resistance to multiple antibiotics.

A high level prevalence of integrons were found to be associated to ARGs. Of the 22 antibiotic resistant plasmids, 12 carry integrons, four of which are complex Class I integrons (Table 3). Two of the integrons are unconventional. The integron carried by pKP265–1 has the exact structure of a common Class I integron, but does not carry an integrase-coding gene. Instead, it carries a relaxase-coding gene (Figure 1A). Whether this gene can encode an enzyme that functions similarly to an integrase is unknown. K. pneumoniae 2–67 carries an IncFII(K)-type antibiotic resistant plasmid p267–1 that also carries an unusual but not unprecedented Class I integron with the gene cassette array ending with qacL and lacking sul1 (Figure 1B) (Alves et al., 2025).

Four antibiotic resistant plasmids carry complex Class I integrons that are diverse in their structures (Figure 1C). Of particular interest is pKQ228-1, an unreported potentially conjugative plasmid that carries a very large complex Class I integron with a novel gene cassette array harboring 12 ARGs, including bla_NDM-1, which is located within the ISCR1-associated integron region. Comparative analysis with regional bla_NDM-1 plasmids carrying complex class 1 integrons revealed distinct integron structures in pKQ228-1 (Supplementary Figure S1). Complex Class I integrons carrying bla_NDM-1 were previously reported in Proteus mirabilis (Li et al., 2023), Enterobacter hormaechei (Doualla-Bell et al., 2021), Enterobacter cloacae (Zhu et al., 2020), Pseudomonas aeruginosa (Kostyanev et al., 2020), and Raoultella ornithinolytica (Yu et al., 2020). To our knowledge, bla_NDM-1 has not previously been reported within complex Class I integrons in Klebsiella species. The identification of such a structure in clinical Klebsiella strain may suggest a new approach of dissemination of carbapenem resistance in Klebsiella.

K. pneumoniae 1–74 hosts a conjugative pKP174–2 plasmid that carries mcr-8.1 and tporJ1-tmexCD1. This plasmid may serve as a vehicle for the dissemination of resistance to polymyxin and tigecycline, both considered last-line antibiotics. Indeed, K. pneumoniae 1–74 is resistant to both polymyxin (MIC = 8 μg/ml) and tigecycline (MIC = 8 μg/ml). To verify transferability, a conjugation assay using K. pneumoniae 1-74 (donor) and E. coli BW25113+pRedCas9 (recipient) was performed. Transconjugants selected on MacConkey agar with streptomycin (200 µg/mL) and polymyxin E (2 µg/mL) were PCR-positive for mcr-8.1, confirming plasmid transfer. The transconjugant exhibited elevated MICs (Polymyxin E: 64 μg/mL; Tigecycline: 8 μg/mL), indicating that pKP174–2 confers transferable resistance. This plasmid is closely related to pHNAH8I-1 from which tigecycline-resistant tporJ1-tmexCD1 was first identified (Lv et al., 2020). The two plasmids share conserved regions carrying mcr-8.1, tmexCD1-toprJ1, and conjugation-associated genes, highlighting their structural similarity and potential for horizontal dissemination (Supplementary Figure S2). However, K. pneumoniae strain AH8I that carried pHNAH8I-1, along with the other four reported similar strains were from chicken fecal samples (Lv et al., 2020). The identification of K. pneumoniae 1–74 that is from hospital samples suggested that this plasmid has now entered clinical settings and poses a direct threat to patients.

The extent of transferability of ARGs found in Klebsiella strains in this work, either in the form of plasmids or integrons, showed that Klebsiella strains quickly exchange and acquire antibiotic resistance with mobile genetic elements. This is important for the clinical setting, particularly in hospitals, because the chances of antibiotic resistant strains meet and exchange antibiotic resistance are much higher than in communities. Antibiotic resistant genetic elements can accumulate to a high level in these settings, as can be confirmed by the finding of plasmid- and integron-rich multidrug resistant Klebsiella strains in this work.

One alarming finding is the extent of heavy metal resistance genes in isolated Klebsiella strains. Eighteen out of 24 strains carry heavy metal resistance genes, more than that carry antibiotic resistance genes (Table 4). Resistance genes for eight metals and metalloids were found: chromate, tellurium, silver, copper, nickel, arsenate, cobalt, and mercury. These resistance genes were identified on 24 distinct plasmids, indicating their potential for horizontal transfer and contribution to metal resistance dissemination. This number is also larger than the 22 antibiotic resistant plasmids found. This reflects that heavy metal resistance is even more prevalent than antibiotic resistance in clinical Klebsiella strains isolated in this work.

The enriched transferrable heavy metal resistance genes in Klebsiella showed that metal resistance is even worse than, or at least comparable with, antibiotic resistance in the hospital setting studied in this work. This makes sense, because in hospitals heavy metals are commonly used for medication. In the case of traditional Chinese medicine, herbs are also often contaminated by heavy metals (Yang et al., 2018). This may explain the level of heavy metal resistance observed in this study. Previous research has shown that hospital effluents often contain heavy metals and other pollutants. For example, analyses of hospital wastewater from hospitals in Indonesia during the COVID-19 pandemic, detected heavy metals and pharmaceuticals that may exert co-selection pressure on bacteria (Sakina et al., 2023). Although specific data for Chinese hospitals remain limited, these findings suggest that hospital-originated Klebsiella strains may be exposed to both antibiotic and heavy metal selective pressures, warranting further investigation.

The level of transferability for metal resistance genes is high. Not only are they hosted by plasmids, 58.3% (14) of these plasmids also carry transconjugation gene cassettes (Table 4). This prompted us to wonder whether heavy metal resistance, similarly to antibiotic resistance, is also enriched in clinical Klebsiella strains under constant selection pressure, and induced by high transferability of the heavy metal resistance determinants.

Twelve plasmids were found to host both antibiotic resistance genes and heavy metal resistance genes. This accounts for 54.5% of antibiotic resistant plasmids, and 50% of heavy metal resistant plasmids. This suggests co-selection of antibiotic and metal resistance on plasmids (Figure 2). Previous studies have shown that exposure to antibiotics or heavy metals can promote co-selection of resistance determinants across different stressors (Wales and Davies, 2015; Mustafa et al., 2021), indicating that environmental contaminants can drive the aggregation and dissemination of multiple resistance traits. We suspect that antibiotic or heavy metal stress can induce the evolution and aggregation of not only the resistance against the exposed stress, but also resistance against the other stress.

Heavy metal resistance genes were organized in a limited number of gene clusters in the metal resistant plasmids found (Figure 3). This is another piece of evidence suggesting the transferability of heavy metal resistance among Klesiella strains. Identical metal resistant clusters could be found in different plasmid types, under different genetic contexts, in different strains, and even in different species (K. pneumoniae and K. quasipneumoniae).

The genetic organization of heavy metal resistance gene clusters were inspected. All chromate resistance was generally associated with the presence of chrA that encodes a membrane-bound efflux pump (Caballero-Flores et al., 2012). Three genetic structures were found associated with chrA, with a IS6100-eal-padR-chrA structure present in both K. pneumoniae and K. quasipneumoniae (Figure 3A). Three conserved genetic structures were found to be related to tellurium resistance, with the gene cluster in pKP388–1 being a variant of that in pKP228-1 (Figure 3B). All silver-resistant gene clusters are similar. Those present in pKP32-1, pKP33-1, and pKP371–1 are IS903 insertion variants of those found in pKQ228-2. In hospitals, silver is widely used as disinfectants for its antibacterial properties (Li and Xin, 2025). Therefore, the prevalence of its resistance in hospital wastewaters is not a surprise. Similarly, a copper resistant gene cluster (like in pKP228-2) and its ΔpcoE2 variant led to copper resistance. Both nickel and cobalt resistance were encoded in the same gene cluster in pKQ228-2 (Figure 3E and 3G), whereas in three plasmids nonspecific transporter corA may be responsible for cobalt resistance (Gibson et al., 1991). Arsenic resistance was prevalent and three forms of gene clusters were found to encode arsenic resistance (Figure 3F). Arsenic trioxide is a well-known medicine for acute promyelocytic leukemia, explaining the prevalence of its resistance mechanisms (Paul et al., 2023). Resistance gene clusters for mercury, an environmental pollution and widely used dental filling (Iqbal and Asmat, 2012), was also found in four plasmids in three formats. While some resistant strains lacked the corresponding genes, the observed associations are supported by previous reports and suggest potential genotype–phenotype correlations.

Phenotypic analysis of heavy metal resistance was performed for Klebsiella strains that host heavy metal resistance genes, except for arsenic resistance because arsenic-containing compounds are heavily regulated and cannot be purchased. A high level of concordance was found on the carriage of heavy metal resistance and heavy metal resistance phenotypes (Supplementary Figure S3). The detailed MIC results for different heavy metals are summarized in Table 5, further supporting the observed genotype–phenotype consistency. Several exceptions exist: strains 2-28, 2-65, 3-2, and 4–33 carry chromate resistance genes but did not show chromate resistance; strain 3–82 carries mercury resistance genes but didn’t show mercury resistance. The case for 3-82 (pKQ382-2) should be due to the incompleteness of the mercury resistance cluster in strain 3-82 (Figure 3).

The analysis of antibiotic resistance and heavy metal resistance can provide answers to the questions we aimed to answer in this work. Plasmids play a major role in heavy metal resistance in Klebsiella species, as heavy metal resistance plasmids are similarly prevalent in the clinical Klebsiella strains studied. Antibiotic resistance and heavy metal resistance showed a high level of correlation, and co-selection of the two resistance types could be taking place. Similarly to antibiotic resistance plasmids, a large portion (45.8%) of heavy metal resistance plasmids are new. This work showed how little we know about the structure and types of antibiotic and heavy metal resistance plasmids in Klebsiella, and provides rationale for truly large scale surveillance. We believe this work serves as a starting point for such investigations.