Introduction

Klebsiella pneumoniae, a member of the Enterobacteriaceae family, is a major cause of healthcare-associated infections, including pneumonia, bloodstream infections (BSIs), and urinary tract infections1,2,3,4. BSIs are associated with high morbidity and mortality, especially in immunocompromised patients2,3. The global dissemination of multidrug-resistant (MDR) K. pneumoniae, particularly carbapenem-resistant K. pneumoniae (CRKP), is a serious clinical and public health challenge5. CRKP is designated as a critical priority pathogen by the WHO and a priority concern by the US Centers for Disease Control and Prevention, requiring urgent public health intervention6,7. The rise in CRKP has led to wider use of colistin in human and veterinary medicine. However, this is contributing to the development and spread of colistin-resistant CRKP (Col-CRKP), an alarming trend given that colistin is a last-line therapeutic agent. This rising resistance to both carbapenem and colistin increases the risk of treatment failure, mortality, and rapid nosocomial transmission, which can potentially lead to the emergence of pan-drug-resistant strains5,8,9.

In Thailand, the Department of Medical Sciences (DMSc) established the National Antimicrobial Resistance Surveillance Center (NARST) in 1997 to monitor antimicrobial resistance (AMR) in humans. NARST collects laboratory data and pathogenic bacterial isolates through a nationwide network, which currently includes 141 hospitals, for confirmatory AMR testing. NARST data indicate that the prevalence of CRKP in Thai hospitals has been steadily increasing, with meropenem resistance rising from 2.1% in 2015 to 15.5% in 202310. Moreover, in 2022, CRKP was linked to extremely high rates of in-hospital mortality (52.9% of patients with hospital-origin CRKP BSIs and 33.1% of patients with community-origin CRKP BSIs), making it one of the deadliest antimicrobial-resistant pathogens in the country11. Due to its growing clinical burden, CRKP has been designated as a priority pathogen for infection reduction in Thailand’s AMR National Action Plan (NSP on AMR 2023–2027).

Thailand is part of the Southeast Asian region where carbapenem-resistant Enterobacteriaceae (CRE) are considered endemic, and carbapenemase-producing K. pneumoniae clones carrying blaNDM and blaOXA−48-like genes have been reported to circulate widely in the country12,13. However, the genomic characteristics of these clones, including their sequence types (STs), AMR genes, and plasmids, as well as their transmission dynamics, are still largely unknown. This information is crucial for understanding the epidemiology of this high-risk pathogen at the regional, national, and international levels4,12. Expanding the genomic surveillance of CRKP will generate key information for the development of antimicrobial stewardship programs, infection prevention and control, and policies designed to address the regional and national threats posed by CRKP.

The existing knowledge gap can be reduced considerably by performing whole-genome sequencing (WGS), which provides deeper insights than phenotypic testing alone. Hence, we aimed to analyze the whole genomes of K. pneumoniae BSI isolates collected by the NARST between 2020 and 2024, including CRKP, colistin-resistant (ColR-KP), and colistin- and carbapenem-resistant (Col-CRKP) isolates, to identify the clonality, AMR genes, plasmid content, virulence potential, and genetic relatedness of isolates from different regions. Although real-time WGS was initially limited, integrating genomic and phenotypic surveillance revealed critical insights into resistance and transmission dynamics, thereby underscoring its value for broader implementation14,15. Our findings have implications for public health responses, infection control, and the clinical management of AMR in Thailand in accordance with the NSP on AMR (2023–2027). In ongoing efforts, we are investigating the implementation of near-real-time sequencing to detect emerging high-risk clones, with the objective of developing targeted onsite diagnostics for clinical surveillance and tracking of outbreak-related activity in the future.

Results

Antibiotic-resistant bacteria recorded in the NARST database between 2020 and 2024

Between 2020 and 2024, 5,783 presumptive antibiotic-resistant bacterial isolates were collected from NARST network hospitals for confirmatory AMR testing. Based on the regional classification of the National Statistical Office of Thailand, the majority originated in Central Thailand (49.8%), followed by Northern (22.5%), Northeastern (16.2%), and Southern (11.4%) Thailand (Table 1). This distribution reflects not only the number of provinces in each region but also regional differences in healthcare access, population density, and sample submission practices. K. pneumoniae was the most prevalent species (2,318 isolates), accounting for 40.1% of the isolates, followed by Escherichia coli (907 isolates, 15.8%) (Table 1). Other Enterobacteriaceae species, such as Enterobacter cloacae, Citrobacter freundii, and Klebsiella aerogenes, were also identified in several regions, albeit less frequently.

Table 1 Distribution of the antibiotic-resistant bacteria most frequently isolated from patients in the NARST hospital network between 2020 and 2024.
Full size table

K. pneumoniae was the predominant species in all regions. Most of the K. pneumoniae isolates were obtained from sputum (n = 926) and urine (including catheterized; n = 817) samples, followed by blood (n = 248) and pus (n = 139) samples. Interestingly, a high prevalence (93.15%) of carbapenem resistance was observed among the K. pneumoniae BSI isolates, and it exceeded the prevalence recorded in the isolates from urine (91.31%) and sputum (91.79%) samples. Of the 248 K. pneumoniae BSI isolates, 227 culturable isolates were subjected to AST to confirm their resistance profiles, followed by WGS for genomic characterization. The distribution of the study isolates across Thailand is shown in Fig. 1A.

Fig. 1
figure 1

Geographical distribution and antimicrobial resistance characteristics of the K. pneumoniae bloodstream infection isolates included in this study. (A) Geographical distribution of the isolates. (B) The carbapenem and colistin resistance profiles of the isolates. (C) The proportions of colistin and aminoglycoside resistance among carbapenem-resistant isolates. The map was generated by the authors using Microsoft Excel (version 2507, build 16.0.19029.20136) with embedded geographic data derived from OpenStreetMap (© OpenStreetMap contributors, https://www.openstreetmap.org). OpenStreetMap data are made available under the Open Database License (ODbL) by the OpenStreetMap Foundation (OSMF).

Full size image

Antimicrobial susceptibility of the K. pneumoniae BSI isolates

The antimicrobial susceptibility results of the study isolates are summarized in Supplementary Table S1. Among the 227 isolates, 199 were CRKP, seven were ColR-KP (resistant to colistin but remained susceptible to carbapenem), 21 were not resistant to either carbapenems or colistin which were included for comparison. Among the CRKP isolates, 38 were resistant to ertapenem only, while 158 showed pan-carbapenem resistance (i.e., resistance or intermediate resistance to all tested carbapenems). The remaining three isolates exhibited other carbapenem resistance patterns (Fig. 1B).

Among the CRKP isolates, 65 (32.7%) were also resistant to colistin (referred to as ColR-CRKP). The pan-carbapenem-resistant isolates exhibited a higher rate of colistin resistance (35.4%) than those resistant to ertapenem only (21.1%; Fig. 1C). Regarding aminoglycoside resistance, 33.2% of the CRKP isolates were resistant to at least one of the tested aminoglycosides. Interestingly, the pan-carbapenem-resistant isolates had a slightly lower aminoglycoside resistance rate (31.6%) than those resistant to ertapenem only (42.1%) (Fig. 1C).

When we examined the resistance of the isolates to the other tested antibiotics, we found that most of the CRKP isolates (93.0%) exhibited resistance or intermediate resistance to all tested β-lactam/β-lactamase inhibitors, cephalosporins, and fluoroquinolones. In addition, 85.9% of the CRKP isolates were resistant to trimethoprim-sulfamethoxazole. In contrast, most of the non-CRKP isolates were susceptible to these antibiotics (Supplementary Table S1).

Distribution of STs among CRKP BSI isolates

The 227 K. pneumoniae BSI isolates were subjected to WGS, and the genome assembly quality metrics and taxonomic identification results are summarized in Supplementary Table S2. Using MLST, based on the Pasteur scheme, we identified 40 distinct STs (Supplementary Table S1). Notably, isolates P0097 (non-CRKP) and P0121 (CRKP) contained novel STs. Subsequent submission of the data to the BIGSdb-Pasteur database (https://bigsdb.pasteur.fr/) resulted in P0097 being classified as an ST8922 isolate, and P0121 as an ST8923 isolate.

Overall, 31 STs were identified in the CRKP isolates, and their regional distribution is shown in Fig. 2. ST16, ST147, and ST231 were widespread across all regions. In the central region, ST16, ST147, and ST231 predominated, with ST11 also highly represented. Several minor STs, including ST15, ST340, ST395, and ST394, occurred at lower frequencies, while ST15 was uniquely detected in the central region. In the northeastern region, ST16, ST147, and ST231 were also prevalent, alongside ST340, ST20, and ST395. The STs detected in the northern region exhibited less diversity, with ST16, ST147, and ST231 as the predominant lineages. ST231 was less common in the northern region than in the central and northeastern regions. Interestingly, ST37 was uniquely detected in the northern region and was found in 16% of the isolates. In contrast, the STs detected in the southern region displayed the greatest heterogeneity, with no single dominant ST. The detected STs included ST11, ST16, ST48, ST147, ST231, ST273, ST268, and ST337. Notably, a substantial proportion of the isolates (29.2%) from the southern region contained rare STs, with each rare ST detected in a single isolate. This finding indicates considerable genomic diversity among CRKP in the southern region. Overall, ST16, ST147, and ST231 were the dominant lineages nationwide. They accounted for the highest proportions of the STs detected in the isolates from the central, northeastern, and northern regions, whereas the STs detected in the isolates from the southern region showed more diversity and no dominant lineage.

Fig. 2
figure 2

Regional distribution of the sequence types (STs) identified in the carbapenem-resistant K. pneumoniae bloodstream infection isolates. The STs that were found in single isolates are grouped together in the “Other” category.

Full size image

Whole-genome SNP-based phylogenetic analysis, AMR gene and plasmid profiling, and determination of virulence potential

The whole-genome SNP-based phylogenetic tree of the 227 K. pneumoniae BSI isolates and the reference genome is shown in Fig. 3. Clustering consistent with the ST classification can be observed, with the major clusters dominated by high-risk clones (e.g., ST16, ST147, and ST231 isolates) and low SNP divergence within each cluster, suggesting clonal expansion. These clusters were geographically dispersed across multiple regions, implying inter-regional transmission.

The ST16 isolates had MDR profiles, with most showing resistance to all tested carbapenems. Additionally, 59% (23/39) of ST16 isolates were colistin-resistant, which is a markedly higher rate compared to other STs. Resistance to amikacin and gentamicin was relatively low (5.1% and 12.8%, respectively), whereas resistance to netilmicin was notably higher at 76.9% (Supplementary Table S1). Two carbapenemase genes, blaNDM−1 and blaOXA−232, were detected in nearly all isolates. Importantly, 66.7% (26/39) of the ST16 isolates harbored both blaNDM−1 and blaOXA−232. In addition, mutations in the ompK36 gene were identified in nearly all ST16 isolates. Despite the high rate of colistin resistance, only three ST16 isolates harbored mutations in crrB, mgrB, pmrA, or pmrB, and no known colistin resistance genes were found to be associated with the resistant phenotype. Regarding aminoglycoside resistance, all netilmicin-resistant isolates carried the aac(6’)-Ib gene, while all susceptible isolates lacked this gene, suggesting that it is involved in mediating netilmicin resistance. The ST16 isolates also exhibited conserved plasmid replicon profiles, frequently harboring Col440I, ColKP3, and variants of IncFIA, IncFIB, and IncFII (Fig. 3).

The ST231 isolates exhibited variations in their carbapenem resistance patterns and carbapenemase gene profiles, with blaOXA−232 and blaOXA−48 being predominant and associated with pan-carbapenem-resistant phenotypes. Additionally, mutations in the ompK36 gene were identified in most isolates. The colistin MIC varied among the ST231 isolates; however, no known colistin resistance genes were detected. Most of the ST231 isolates (87.5%) were resistant to all tested aminoglycosides, with most harboring both aac(6’)-Ib and rmtF1 genes. All isolates harboring rmtF1 showed resistance to all tested aminoglycosides, indicating a strong association between this gene and the aminoglycoside-resistant phenotype. ST231 isolates also harbored a conserved plasmid replicon set, which included ColKP3, and variants of IncFIA, IncFIB, and IncFII (Fig. 3).

The ST147 isolates exhibited considerable variation in their antibiotic resistance patterns, ranging from susceptible to all tested carbapenems, colistin, and aminoglycosides; to pan-resistance against all three antibiotic classes. The resistance genes and plasmids present in these isolates were also highly diverse. Most of the pan-carbapenem-resistant isolates (75%; 30/40) carried blaNDM−1, while blaNDM−5 was also found in certain isolates. In addition, blaOXA−48, blaOXA−181, and blaOXA−232 were detected at low frequencies. Interestingly, a subgroup of ST147 isolates carrying both blaNDM−1 and blaOXA−48 was observed. Most isolates within this subgroup exhibited pan-carbapenem resistance. Furthermore, a unique mutation in pmrB was universally present among ST147 isolates; however, this mutation was not associated with colistin resistance. For aminoglycosides, 29.4%, 35.3% and 56.9% of ST147 were resistant to amikacin, gentamicin, and netilmicin, respectively. aph(3’)-VI was the predominant aminoglycoside resistance gene identified in this ST. Interestingly, most of the isolates carrying armA exhibited resistance to all tested aminoglycosides. The plasmid analysis revealed high diversity, suggesting that plasmid-mediated dissemination of AMR genes may occur among ST147 strains (Fig. 3).

Other STs were detected at low frequencies and were often limited to specific geographic regions. The isolates with these STs displayed substantial diversity in their AMR gene profiles and plasmid content. In addition to blaNDM−1, blaNDM−4 and blaNDM−5 were also identified in rare ST isolates. Despite their low prevalence, five isolates with these uncommon STs harbored mcr genes, and this is of particular concern due to their strong association with phenotypic colistin resistance. Isolate P0273 (ST127) carried mcr-1 and displayed colistin resistance, along with an unusual carbapenem resistance profile, being resistant to meropenem only. Isolates P0188 (ST881) and P0264 (ST3167) harbored mcr-3 and were resistant to colistin; however, both were susceptible to all tested carbapenems. Two ST273 isolates (P0130 and P0192) carried mcr-8. P0130 was susceptible to all tested carbapenems and showed a moderate colistin MIC (2 µg/mL), whereas P0192 was resistant to ertapenem, doripenem, and meropenem, showed intermediate resistance to imipenem, and was resistant to colistin (Supplementary Table S1, Fig. 3). Four of the five isolates harboring mcr genes originated from the central region, and one was isolated from a northern province adjacent to the central region, suggesting potential regional dissemination.

Two novel STs, ST8922 (isolate P0097) and ST8923 (isolate P0121), were also identified. ST8923 clustered within a clade containing the reference lineage ST38, although it remained clearly separated from the major high-risk clones. This isolate exhibited pan-carbapenem resistance and intermediate resistance to colistin, while remaining susceptible to all tested aminoglycosides, and was characterized by the presence of blaNDM−4 alongside multiple plasmid replicons. In contrast, ST8922 grouped with ST882 within the broader MDR-associated clade that includes ST231; however, this isolate was susceptible to all tested antibiotics (Fig. 3). A sub-tree showing the phylogenetic placement of these novel STs, along with the allelic profiles used for ST designation, is provided in Supplementary Figure S1.

The virulence potential of all K. pneumoniae isolates based on the presence of specific virulence genes was evaluated using Kleborate virulence score (from 0 to 5). The ST23 and ST268 isolates were classified as hypervirulent (virulence score: 5). Both isolates carried the full repertoire of hvKp-associated loci, including yersiniabactin (ybt), aerobactin (iuc), salmochelin (iro), and colibactin (clb), as well as rmpA2. Capsule locus typing revealed that the ST23 isolate belonged to the K1 type, whereas the ST268 isolate corresponded to the K20 type. The ST231 and ST3167 isolates were considered virulent (virulence score: 4). All remaining isolates were classified as nonvirulent (virulence score ≤ 1; Supplementary Table S3).

Fig. 3
figure 3

The single-nucleotide polymorphism-based phylogenetic tree, antimicrobial susceptibility, antimicrobial resistance genes, and plasmid profiles of the K. pneumoniea bloodstream infection isolates collected in Thailand between 2020 and 2024. Antibiotic abbreviations: ETP, ertapenem; DOR, doripenem; MEM, meropenem; IMP, imipenem; AMK, amikacin; GEN, gentamicin; NET, netilmicin; COL, colistin. AST = antimicrobial susceptibility testing. Susceptibility categories: S, susceptible; I, intermediate; R, resistant. ST = Sequence type,.

Full size image

Association between AMR genes and phenotypes

We performed a binary logistic regression analysis to evaluate the associations between the presence of carbapenemase genes and phenotypic resistance in the K. pneumoniae isolates. Multicollinearity among predictor variables was evaluated using variance inflation factors (VIF). All included gene predictors demonstrated VIF values < 3, indicating no significant multicollinearity (Supplementary Table S4). All logistic regression models included 227 isolates with complete genotypic and phenotypic data. The numbers of resistant versus susceptible isolates for each antibiotic were: ertapenem (202/25), doripenem (160/67), meropenem (168/59), imipenem (159/68), and pan-carbapenem resistance (158/69). Significant correlations were observed between the presence of specific carbapenemase genes and carbapenem-resistant phenotypes (Table 2). blaNDM−1 was strongly associated with resistance to imipenem (aOR = 283.44, q < 0.0001), meropenem (aOR = 103.34, q = 0.0002), doripenem (aOR = 240.82, q < 0.0001), ertapenem (aOR = 14.19, q = 0.0298), and the pan-carbapenem-resistant phenotype (aOR = 304.24, q < 0.0001). Additionally, perfect associations were identified between blaNDM−4 and blaNDM−5 and resistance to all tested carbapenems (i.e., present in all resistant isolates and absent in all susceptible isolates). When all blaNDM variants were pooled, these associations remained consistent. blaOXA−232 was significantly associated with resistance to all tested carbapenems, most notably to ertapenem (aOR = 14.96, q = 0.0298). blaOXA−48 and blaOXA−181 were also significantly associated with resistance to all carbapenems except ertapenem. When blaOXA variants were pooled, consistent associations were observed across all individual carbapenems and the pan-carbapenem-resistant phenotype.

When we analyzed the associations between AMR genes and colistin resistance (72 resistant and 155 susceptible isolates), only one mutation in pmrB (pmrB_T157P) showed a significant association (aOR = 5.20, q = 0.0388). Among the seven isolates carrying pmrB_T157P, six were colistin resistant. All isolates that harbored mcr-1 or mcr-3 were resistant to colistin. Of the two isolates that carried mcr-8, one was resistant and one was susceptible to colistin. In most cases (62/72 isolates), resistance to colistin could not be attributed to known resistance genes or mutations.

Table 2 Binary logistic regression analysis of carbapenemase genes and resistant phenotypes.
Full size table

Discussion

In this nationwide genomic study, we analyzed the epidemiology and molecular mechanisms of the resistance of K. pneumoniae BSI isolates collected in Thailand between 2020 and 2024. The high proportion of CRKP (93.15%) found in K. pneumoniae BSI specimens compared with other specimen types in the NARST dataset underscores the threat that this pathogen poses in Thai healthcare settings and highlight the growing prevalence of MDR bacteria in Southeast Asia12. The prevalence of co-resistance to colistin in the CRKP isolates (32.7%) is particularly alarming, as colistin is a last-line antimicrobial agent used to treat infections with MDR bacteria16.

Notably, the CRKP isolates exhibited moderate rates of resistance to amikacin and gentamicin (33.2% and 45.2%, respectively), which were lower than the rates of resistance to the other antibiotic classes. These findings are consistent with national antibiogram data10. Furthermore, there are also previous reports of CRKP exhibiting moderate resistance to amikacin and gentamicin, with rates below 50% reported in several countries17,18,19. This suggests that aminoglycosides are still partially effective against CRKP and are a therapeutic option when treatment choices are limited20,21.

Our results demonstrate that high-risk ST16, ST147, and ST231 clones, which appear to be endemic in Thailand, predominate and are widely distributed across the country, particularly in the central, northern, and northeastern regions. These observations are consistent with the classification of such clones as global epidemic lineages associated with MDR and hospital outbreaks13,22,23,24. In contrast, the southern region exhibited highly diverse lineages, suggesting that a distinct epidemiological dynamic exists in this region. The absence of dominant clones in the south may reflect the occurrence of localized outbreaks or multiple introductions rather than sustained clonal expansion25. Several factors may contribute to this elevated lineage diversity in southern Thailand. The region shares an active land border and substantial population movement with Malaysia, and cross-border healthcare utilization, travel, and trade may introduce additional K. pneumoniae lineages26,27,28. Ecological factors may also play a role, as southern Thailand has extensive coastal and aquaculture activities, and K. pneumoniae can persist in brackish water, marine sediments, and aquaculture systems. Human interaction with these settings may provide opportunities for transmission of environmentally adapted or rare STs into clinical settings29,30,31,32. In addition, lower sequencing density relative to other regions indicates that additional sampling may be needed to fully characterize the population structure in the south. Collectively, such diversity and the unique influencing factors that differ from other regions highlight the importance of continuous genomic surveillance in the south, as it may serve as a reservoir for emerging resistance determinants or facilitate horizontal gene transfer events33.

After exploring regional differences in lineage distribution, we next examined the genomic and antimicrobial resistance profiles of all sequence types, with particular focus on the dominant high-risk clones. Most of the ST16 isolates exhibited pan-carbapenem resistance, which was mediated primarily by the co-occurrence of blaNDM−1 and blaOXA−23234,35. The co-occurrence of these genes in ST16 isolates has been reported in multiple countries, including Thailand36,37,38. The co-existence of these genes appears to be driven by the co-introduction of different plasmids, each carrying a distinct carbapenemase gene37,38. Most of the ST16 isolates also harbored mutations in ompK36, which contributed to porin loss39. Furthermore, ST16 emerged as the predominant Col-CRKP clone, with 59% of the ST16 isolates exhibiting colistin resistance despite the absence of acquired colistin resistance genes such as mcr genes40, or chromosomal mutations in pmrAB, crrB, or mgrB41. These findings underscore the complexity of colistin resistance and the limitations of current genotypic predictors, suggesting the involvement of uncharacterized genetic factors and, hence, the need for further investigation. Moreover, the ST16 isolates exhibited extensive resistance profiles, as well as conserved AMR genes and plasmid replicons (e.g., Col440I, ColKP3, IncFIA, IncFIB, and IncFII). These findings, which are consistent with previous reports42,43,44, indicate that this lineage has undergone clonal expansion and achieved broad geographic dissemination.

ST231 was the most prevalent ST in this study and was detected in multiple regions. In ST231 clones, carbapenem resistance is primarily mediated by blaOXA−48 or blaOXA−23235. The rate of aminoglycoside resistance was notably high among the ST231 isolates, with 87.5% showing resistance to all tested aminoglycosides. The copresence of aac(6’)-Ib and rmtF1 genes in the majority of isolates suggests that the resistance was mediated by plasmids via both the acetylation and methylation of ribosomal targets45. The perfect correlation between rmtF1 and aminoglycoside resistance underscores the clinical importance of this gene46. Furthermore, the ST231 isolates contained conserved plasmid replicons (e.g., IncFIA, IncFIB, and ColKP3), indicating the presence of stable mobile elements47. The high AMR rate, substantial clonal expansion, and widespread geographic distribution of the ST231 clones analyzed in this study indicate that ST231 is an established epidemic lineage in Thailand, and this is consistent with findings from previous surveillance studies13.

ST147 was the second most common ST detected in this study. The ST147 isolates were characterized by remarkable genomic diversity and a broader range of resistance profiles compared to the ST16 and ST231 isolates. These isolates displayed variations in their carbapenem, colistin, and aminoglycoside resistance, with many exhibiting a pan-resistant phenotype. The majority of the carbapenem-resistant isolates harbored blaNDM−1, and blaNDM−5, whereas blaOXA−48, blaOXA−181 and blaOXA−232 were occasionally detected. The copresence of blaNDM−1 and blaOXA−48 was observed, particularly in pan-carbapenem-resistant isolates. A pmrB mutation (R256G) was detected in all ST147 isolates; however, its role in colistin resistance remains uncertain, as only 31.37% of isolates of this ST were resistant. Previous studies have reported an association between pmrB mutations and polymyxin B resistance48; however, other investigations have shown that the presence of pmrB mutations does not always correspond to colistin resistance49,50. Aminoglycoside resistance in the ST147 isolates was driven by a range of genetic factors. The most commonly detected gene was aph(3’)-VI, while a subset of isolates harbored armA, which showed a strong correlation with high-level resistance to all tested aminoglycosides45. The heterogeneity of the AMR genes and plasmid replicons within the ST147 isolates reflects the genomic plasticity of this lineage and suggests that it may serve as a reservoir for AMR genes51.

In addition to the dominant STs, several rare STs were detected, with some isolates carrying mcr genes that confer plasmid-mediated colistin resistance40. Specifically, one ST127 isolate carried mcr-1, one ST881 isolate and one ST3167 isolate each carried mcr-3, and two ST273 isolates carried mcr-8. Although these STs were detected at low frequencies, and the isolates were mostly susceptible to carbapenems, the presence of these mcr variants in BSI isolates underscores the potential for horizontal gene transfer of colistin resistance within and across bacterial species52. Notably, STs associated with mcr genes have also been found in animals. For example, ST127 has been detected in horses53, and ST881 was previously reported in pigs in Europe54. Interestingly, ST3167 has been found in pigs from Thailand55. This overlap indicates that mcr-harboring clones can be spread across different ecological niches and may facilitate the spread of colistin resistance between bacteria that infect animals and those that infect humans56. Such findings reinforce the need for integrated genomic surveillance across human, animal, and environmental settings to monitor the evolution and transmission of mcr genes56,57. In contrast, ST273 appears to be a hospital-associated lineage58,59. Notably, a study in China also identified an ST273 isolate carrying the mcr-8 gene on a plasmid from a patient with bacteremia60. In this study, we identified an ST273 isolate harboring mcr-8 that exhibited resistance to colistin, ertapenem, doripenem, and meropenem, as well as intermediate resistance to imipenem. These findings raise concern that ST273 may serve as a reservoir for plasmid-mediated colistin and carbapenem resistance and could emerge as a high-risk clone in clinical settings.

The ST23 and ST268 isolates analyzed in this study were classified as hvKp based on their Kleborate virulence score of 5. Both isolates carried the full repertoire of hvKp-associated loci, including rmpA2, which is associated with the hypermucoid phenotype61. Capsule locus typing identified the ST23 isolate as K1 and the ST268 isolate as K20. ST23 has been recognized as a globally disseminated hvKp lineage, and it has been detected in Southeast Asia24,62. While carbapenem-resistant and MDR ST23 isolates have been reported in several countries63,64,65, the ST23 isolate identified in this study was susceptible to all tested carbapenems, and no known carbapenemase genes were detected. ST268 has also been reported as a global hvKp lineage, with multiple studies describing its association with extended-spectrum β-lactamase and carbapenemase genes66,67,68. In this study, the ST268 isolates exhibited pan-carbapenem resistance, harboring either blaNDM−4 or blaNDM−5. Moreover, an ST268 isolate obtained from the southern region was also nonsusceptible to all tested aminoglycosides, which underscores clinical concerns about this lineage. The ST231 and ST3167 isolates were classified as virulent (virulence score: 4). ST231 is recognized as a globally distributed resistant lineage and was the most prevalent ST detected in our CRKP isolates. Although ST231 clones harbor multiple virulence genes, several studies have suggested that they generally exhibit low to moderate virulence69,70. In this study, the ST3167 isolate was susceptible to carbapenems but carried mcr-3 and displayed colistin resistance. Little is currently known about the virulence potential of this lineage, and further investigation is required to clarify its pathogenic and epidemiological significance. The isolates with other STs were classified as nonvirulent, which is consistent with the findings of previous studies suggesting that MDR K. pneumoniae strains generally exhibit low virulence71. We determined the virulence potential of the isolates by assessing the presence of virulence-associated genes in the genome; however, virulence phenotypes and clinical outcome data were not assessed. Future genomic surveillance, combined with virulence phenotypes and clinical outcome analysis, will provide a more comprehensive understanding of the virulence and AMR of K. pneumoniae in clinical settings.

We identified two novel STs in this study. The ST8922 isolate exhibited susceptibility to all tested antibiotics, and no plasmid replicons were detected in this isolate. In contrast, the ST8923 isolate exhibited resistance to all tested carbapenems and carried blaNDM−4 along with multiple plasmids. Notably, the ST8923 isolate clustered separately from the other isolates, indicating an evolutionary distinction. It is possible that the ST8923 lineage has a nonclinical origin, such as an animal or environmental reservoir72,73. The detection of novel STs highlights the importance of performing continuous genomic surveillance, which enables the early detection and tracking of previously unrecognized high-risk lineages and provides insights into the evolutionary dynamics of AMR74.

Our binary logistic regression analysis showed strong associations between specific carbapenemase genes and phenotypic resistance. blaNDM−1 showed high aORs for resistance to all tested carbapenems. This finding is consistent with global reports highlighting blaNDM−1 as a key driver of high-level carbapenem resistance75. Moreover, blaNDM−4 and blaNDM−5 showed perfect associations with carbapenem resistance. A pooled analysis of blaNDM variants yielded similarly robust associations across all carbapenems. Among the OXA-48-like genes, blaOXA−232 was significantly associated with resistance to all tested carbapenems, particularly ertapenem, whereas blaOXA−48 and blaOXA−181 were significantly associated with resistance to doripenem, imipenem, and meropenem. Pooled OXA-48-like variants also demonstrated consistent associations with resistance to each carbapenem and with the pan-carbapenem resistant phenotype. Although OXA-48-like genes showed significant associations with carbapenem resistance in our regression models, this association likely reflects co-existing mechanisms in the genetic background of the isolates particularly the frequent co-harboring of blaNDM−1 and ompK36 mutations rather than the effect of OXA-48-like enzymes alone. Previous studies have demonstrated that OXA-48–like enzymes exhibit limited carbapenemase activity and typically confer low-level carbapenem resistance76,77. Thus, the high-level resistance observed in these isolates is expected to result from the combined effects of carbapenemase co-carriage and porin loss78,79.

For colistin resistance, only one mutation in pmrB (pmrB_T157P) showed a significant association, and this variant was present in only a small subset of isolates. Notably, the majority of colistin-resistant isolates did not carry any known resistance determinants, such as mcr genes or mutations in pmrAB, mgrB, or crrB80,81. This observation suggests that additional or uncharacterized mechanisms may contribute to colistin resistance in our collection. Several factors may account for the absence of known resistance markers. Colistin resistance can emerge through alternative Lipid A modification pathways, potentially involving genes or mutations that remain unannotated81. The phenomenon of heteroresistance, in which only a subset of cells displays increased tolerance, may also contribute and is not readily identified through genomic analysis alone82. Furthermore, incomplete annotation of regulatory networks, including two-component systems involved in lipid A modification, may limit detection of relevant mutations81,83,84. Additional factors, such as efflux pumps or other poorly defined stress response pathways, may also underlie colistin resistance in these isolates84. Together, these observations highlight the complexity of colistin resistance mechanisms and emphasize the need for more comprehensive genomic and multiomics analyses to fully elucidate the factors contributing to this phenotype.

This study offers a detailed genomic overview of CRKP and ColR-KP circulating in Thailand; however, several limitations should be acknowledged. This study was based exclusively on short-read sequencing, which limited the ability to resolve complete plasmid structures and distinguish plasmid-borne from chromosomal elements. The absence of patient-level clinical information also restricted interpretation of how specific genomic features relate to disease severity or treatment outcomes. In addition, the cross-sectional sampling approach provides a snapshot of circulating lineages but does not capture temporal or directional transmission events. Future studies that incorporate long-read sequencing, clinical metadata, and longitudinal sampling will enable a more comprehensive view of CRKP and ColR-KP dissemination and its clinical consequences.

Despite these limitations, our study demonstrates the utility of integrating WGS into routine AMR surveillance. The capacity to analyze genomic data not only increased our understanding of resistance mechanisms and transmission dynamics but also enabled the detection of high-risk clones with epidemic potential. While retrospective in nature, our findings provide a strong rationale for implementing prospective, near-real-time genomic surveillance to inform infection control and antimicrobial stewardship strategies in Thailand.

Conclusion

This nationwide genomic surveillance study revealed that high-risk CRKP is widespread in Thailand. The predominance of ST16, ST147, and ST231 clones, coupled with their extensive AMR, underscores the critical threat posed by these lineages. The detection of mcr genes in isolates with rare STs and the emergence of a hypervirulent, carbapenem-resistant clone further emphasizes the evolving complexity of AMR in this pathogen. Integrating WGS into national surveillance provides vital insights into resistance mechanisms, clonal dynamics, and evolutionary trends. Furthermore, expanding the use of genomic surveillance and molecular diagnostics is necessary to control further dissemination and to reach the goals outlined in Thailand’s AMR National Strategic Plan (2023–2027).

Methods

K. pneumoniae isolates and laboratory data collection

All K. pneumoniae isolates included in this study were recovered from glycerol stocks stored at − 80 °C as part of routine AMR surveillance performed by NARST. Between 2020 and 2024, the nationwide surveillance program collected 248 antibiotic-resistant K. pneumoniae isolates from blood cultures of hospitalized patients. A total of 227 isolates were successfully recovered and included in this study. Non-viable isolates were excluded, and no additional selection was applied. All isolates were re-identified by standard biochemical methods following the Manual of Clinical Microbiology85. The location at which each isolate was obtained, and the year of isolation are summarized in Supplementary Table S1.

Antimicrobial susceptibility testing

Antimicrobial susceptibility testing (AST) was performed via broth microdilution using customized THAN2F Sensititre plates (Thermo Fisher Scientific, Waltham, MA, USA). The panel included 21 antibiotics: amikacin, amoxicillin-clavulanate, ampicillin, ampicillin-sulbactam, cefepime, cefotaxime, cefoxitin, ceftazidime, ceftriaxone, cefuroxime, ciprofloxacin, colistin, doripenem, ertapenem, gentamicin, imipenem, levofloxacin, meropenem, netilmicin, piperacillin-tazobactam, and trimethoprim-sulfamethoxazole. Briefly, isolated colonies cultured on sheep blood agar plates were suspended in 5 mL of cation-adjusted Mueller–Hinton broth (Thermo Fisher Scientific, Waltham, MA, USA). The bacterial suspension was then adjusted to a 0.5 McFarland standard using a nephelometer. A 30 µL aliquot of this suspension was added to 11 mL of Mueller–Hinton broth to prepare the inoculum. Using the AIM Sensititre system (Thermo Fisher Scientific, Waltham, MA, USA), 50 µL of the inoculum was dispensed into each well of the THAN2F plates. The plates were incubated at 35 °C ± 2 °C for 18–24 h and read using the Sensititre ARIS 2X system (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s instructions. Minimum inhibitory concentrations (MICs) were calculated and interpreted using the Clinical and Laboratory Standards Institute (CLSI) M100 guidelines for AST86.

DNA extraction and whole genome sequencing

All K. pneumoniae isolates were freshly cultured on sheep blood agar plates and incubated overnight at 37 °C. Genomic DNA was extracted using the Presto™ Mini gDNA Bacteria Kit (Geneaid Biotech, New Taipei City, Taiwan) according to the manufacturer’s protocol. DNA purity and concentration were initially assessed using a NanoDrop™ 1000 spectrophotometer (Thermo Scientific, Waltham, MA, USA) and further quantified using a Qubit™ 3.0 Fluorometer (Invitrogen, Carlsbad, CA, USA). DNA integrity was evaluated by gel electrophoresis. Sequencing libraries were prepared from 100 ng of genomic DNA using Illumina DNA Prep kit (M) Tagmentation (Illumina, San Diego, CA, USA), following the manufacturer’s instructions. Libraries were sequenced on the Illumina NextSeq platform with the P1 XLEAP-SBS High Output Reagent Kit (Illumina, San Diego, CA, USA), generating 150-bp paired-end reads.

Genomic assembly and annotation

The raw sequencing reads were initially assessed for their quality and trimmed using Trim Galore v0.6.5 and then assembled using the SPAdes genome assembler v4.0.0 via the BV-BRC web resources using default parameters87. The genome assemblies were evaluated using QUAST v5.0.288, and only those with a minimum coverage of 30× were included in further analyses. Taxonomic classification based on the average nucleotide identity and alignment fraction was performed using GTDB-Tk v2.1.189. Multilocus sequence typing (MLST) of seven housekeeping genes (gapA, infB, mdh, pgi, phoE, rpoB, and tonB) was conducted using MLST v2.23.090. AMR genes and plasmid replicons were identified using AMRFinderPlus v3.12.891 and PlasmidFinder v 2.1.692, respectively. Virulence potential was assessed using Kleborate v2.3.2, which assigns a virulence score (0–5) based on the presence of specific virulence genes61.

Phylogenetic analysis and visualization

A phylogenetic tree was constructed using the 227 assembled genomes and the reference genome K. pneumoniae MGH 78,578 (GCF_000016305.1). Core single-nucleotide polymorphisms (SNPs), including substitutions, deletions, and insertions, were identified using Snippy v4.6.0 (https://github.com/tseemann/snippy). Recombinant regions were detected and masked using Gubbins93. Maximum likelihood phylogenetic analysis was then performed in IQ-TREE using the GTR + G nucleotide substitution model with 1,000 bootstrap replicates94. The resulting phylogenetic tree was visualized and annotated using the iTOL v6 web tool95.

Logistic regression analysis of AMR genes and phenotypic characteristics

Multicollinearity among predictor variables was evaluated using variance inflation factors. Associations between antimicrobial susceptibility and the presence of AMR genes were evaluated using multivariable binary logistic regression. For each antibiotic, phenotypic resistance (resistant vs. susceptible) was modeled as the dependent variable, with AMR gene presence as independent predictors. All 227 isolates with complete genotype and phenotype data were included in each model, with non-resistant isolates serving as controls. Adjusted odds ratios with 95% confidence intervals were calculated, and multiple testing correction was applied using the Benjamini–Hochberg false discovery rate (FDR), with significance defined as q < 0.05. All statistical analyses were performed in R (v4.3.1) (https://www.R-project.org/).