In-depth characterization of multidrug-resistant NDM-1 and KPC-3 co-producing Klebsiella pneumoniae bloodstream isolates from Italian hospital patients

ABSTRACT Bloodstream infection (BSI) caused by carbapenem-resistant Klebsiella pneumoniae (KP) poses significant challenges, particularly when the infecting isolate carries multiple antimicrobial resistance (AMR) genes/determinants. This study, employing short- and long-read whole-genome sequencing, characterizes six New Delhi metallo-β-lactamase (NDM) 1 and KP carbapenemase (KPC) 3 co-producing KP isolates, the largest cohort investigated in Europe to date. Five [sequence type (ST) 512] and one (ST11) isolates were recovered from patients who developed BSI from February to August 2022 or February 2023 at two different hospitals in Rome, Italy. Phylogenetic analysis revealed two distinct clusters among ST512 isolates and a separate cluster for the ST11 isolate. Beyond blaNDM-1 and blaKPC-3, various AMR genes, indicative of a multidrug resistance phenotype, including colistin resistance, were found. Each cluster-representative ST512 isolate harbored a blaNDM-1 plasmid (IncC) and a blaKPC-3 plasmid [IncFIB(pQil)/IncFII(K)], while the ST11 isolate harbored a blaNDM-1 plasmid [IncFII(pKPX1)] and a blaKPC-3 plasmid [IncFIB(K)/IncFII(K)]. The blaNDM-1 plasmids carried genes conferring resistance to clinically relevant antimicrobial agents, and the aminoglycoside resistance gene aac(6′)-Ib was found on different plasmids. Colistin resistance-associated mgrB/pmrB gene mutations were present in all isolates, and the yersiniabactin-encoding ybt gene was unique to the ST11 isolate. In conclusion, our findings provide insights into the genomic context of blaNDM-1/blaKPC-3 carbapenemase-producing KP isolates. IMPORTANCE This study underscores the critical role of genomic surveillance as a proactive measure to restrict the spread of carbapenemase-producing KP isolates, especially when key antimicrobial resistance genes, such as blaNDM-1/blaKPC-3, are plasmid borne. In-depth characterization of these isolates may help identify plasmid similarities contributing to their intra-hospital/inter-hospital adaptation and transmission. Despite the lack of data on patient movements, it is possible that carbapenem-resistant isolates were selected to co-produce KP carbapenemase and New Delhi metallo-β-lactamase via plasmid acquisition. Studies employing long-read whole-genome sequencing should be encouraged to address the emergence of KP clones with converging phenotypes of virulence and resistance to last-resort antimicrobial agents.

In a 5-year genomic surveillance study involving public hospitals in Singapore, plasmid-mediated transmission accounted for ~50% of dissemination of carbapene mase-producing KP (and other Enterobacterales), whereas another ~40% met clonal (i.e., clone mediated) transmission criteria (20).Regarding KP, whole-genome sequencing (WGS) is the most cost-efficient approach to deeply characterize clinical isolates (16), and bioinformatic tools facilitate the analysis and interpretation of genomic data (21).Here, we reported on the WGS-based characterization of KP isolates co-producing NDM-1 and KPC-3 from hospitalized patients in Italy.

Resistome profiling of study isolates
Figure 2 shows the SR-based resistome analysis, which revealed the presence of the putative chromosomal fosA (fosfomycin resistance) and oqxA/oqxB (fluoroquinolone resistance) genes as well as mutated/truncated genes, such as ompK35/ompK36 (carbapenem resistance), mgrB/pmrB (colistin resistance), or gyrA/parC (fluoroquinolone resistance), in all six isolates.Regarding the putative acquired resistome, along with bla NDM-1 and bla KPC-3 genes, other β-lactam (bla OXA-9 ), trimethoprim (dfrA12/dfrA14), or phenicol (catA1/catB3) resistance genes, and at least one gene that confers aminogly coside resistance [e.g., aac(6′)-Ib] were detected in all six isolates.The sulphonamide (sul1) or erythromycin (mphA) resistance genes were detected in all but BSI_389-23 isolates, whereas the β-lactam resistance gene bla CTX-M-15 was only detected in the BSI_389-23 isolate.The AMR determinants (mutations and/or acquired genes) identi fied were consistent with the above-described MDR phenotype for all six isolates.No CAZavi resistance-associated mutations (19) were found in the isolates' bla KPC-3 genes, suggesting that co-production of a (less sensitive) NDM-type MβL could have caused the CAZavi resistance phenotype of isolates.No plasmid-borne colistin resistance gene mcr-1 (23) was detected in any of the isolates, supporting the complexity of some AMR mechanisms in KP organisms.

Plasmid characterization in cluster-representative isolates
The isolates CRKP2202 and CRKP2205 were selected along with BSI_389-23 for a long-read (LR) sequencing approach.Subsequent hybrid genome assemblies allowed us to assess the gene content, plasmid content, and AMR or virulence determinants for the three isolates, respectively.The number of annotated coding DNA sequences per genome was 5,484 for CRKP2202, 5,744 for CRKP2205, and 5,401 for BSI_389-23.Of 25 AMR genes detected in total (see also Fig. 2), five were in the chromosome, and 20 were in plasmids.

Molecular determinants of colistin resistance or virulence
A detailed description of colistin resistance determinants in the three isolates (CRKP2202, CRKP2205, and BSI_389-23) is provided in Fig. 5.A Kleborate-based AMR analysis allowed us to search for missense mutations (alterations to the amino acid sequence), nonsense mutations (premature stop codon), or frameshift mutations (by insertions/deletions) in the mgrB and pmrB genes (23).As a result, we found that MgrB was absent in CRKP2202 (complete deletion) but was present in CRKP2205 (6% truncation) and BSI_389-23 (60% truncation; in this case, an insertion was followed by a stop codon).Regarding PmrB, all three isolates had the R256G mutation, whereas CRKP2205 also had the T157P mutation, and BSI_389-23 also had the T140P mutation.

DISCUSSION
This study represents the first comprehensive WGS-based characterization of KP-BSI isolates from Italian hospital patients co-producing NDM-1 and KPC-3 carbapenemases.Beyond bla NDM-1 and bla KPC-3 , the six isolates, five from clone ST512 (CRKP2201 to CRKP2205) and one from clone ST11 (BSI_389-23), exhibited an array of AMR genes and chromosomal mutations, indicative of an MDR phenotype observed in vitro.While isolates (five and one) formed clusters corresponding to the respective hospitals where BSIs were diagnosed, genomic similarities within ST512 or ST11 clusters and even between clusters highlighted shared AMR plasmids.This underscores that, although certain AMR genes and plasmids are linked to specific clones (ST512 or ST11) in hospital-acquired BSI isolates, some plasmid similarities extend beyond these clones and healthcare facilities, indicating broader dissemination.

High-risk KP clones and clinical implications
The KP clones ST258 and its derivative ST512 are a cause of concern due to their spatiotemporal expansion in European hospitals, linked to nosocomial transmission and genomic detection of carbapenem resistance genes (9).Shropshire et al. (24) used hybrid SR and LR genome assemblies to dissect the accessory genome of carbapenem-resistant KP isolates (CG258 and co-circulating CG307) in Houston, TX, hospitals.CG307 isolates lacked siderophore genes, suggesting that convergence of MDR and hypervirulence molecular determinants may be a hallmark of CG258.The Shropshire et al. 's study (24) reported higher rates of BSI and 30-day mortality in CG258-infected patients compared to CG307-infected patients.In our study, all patients (5/5) patients with ST512 KP-BSI succumbed, highlighting the challenges in managing carbapenem-resistant (and MDR) Additionally, T140P (denoted as a green dot) or T157P (denoted as a blue dot) mutation was identified in BSI_389-23 and CRKP2205, respectively.KP isolates.Notably, the patient with ST11 KP-BSI, treated with cefiderocol (to which the isolate was susceptible in vitro), survived.
The KP clone ST11, derived from ST258 by recombination, differs geographically (17).The Shropshire et al. 's study (24) reported that 2 of 37 CG258 isolates were thirdgeneration cephalosporin resistant due to a bla CTX-M-15 -harboring plasmid and were carbapenem resistant due to a bla KPC-3 -harboring plasmid.Only one isolate in our study (ST11) had a plasmid-borne bla CTX-M-15 .These findings underscore the diversity of AMR genes within the same clone across locations, emphasizing the role of plasmid content in phenotypic resistance.

Uncommon coexistence of multiple carbapenemases in KP-BSI isolates
Since the first identification of KPC in the USA in 1996, NDM-1 and other non-KPC-type carbapenemases, such as MβLs, have emerged but to a lesser extent (43 KPC variants versus 28 NDM variants) (10,25).Alarmingly, NDM-1, especially in combination with KPC-2/KPC-3, is on the rise.In 2021, the Pan American Health Organization/World Health Organization warned of NDM/KPC combinations in Enterobacterales, including KP, in Latin America and the Caribbean (26).This may be linked to increased broad-spectrum antibiotic use in COVID-19 patients, impacting microbiological diagnosis and antimicro bial agent use.Gao et al. ( 14) studied seven carbapenem-resistant KP isolates in China from 2012 to 2017.They suggested that extensive CAZavi use [amplified by COVID-19; see reference (27)] may have counter-selected isolates with only bla KPC-2/3 , while highly transferable bla NDM-1 plasmids contributed to intra-hospital/inter-hospital adaptation and transmission of isolates carrying both bla KPC-2/3 and bla NDM-1 .In our study, similar clinical contexts could have selected for KPC-2/KPC-3 carbapenem-resistant KP isolates and induction of co-produced KPC and NDM via plasmid acquisition.

Cluster-associated plasmids and AMR profiles
Our analysis of hybrid genome assemblies in three cluster-representative KP isolates (CRK2202, CRKP2205, and BSI_389-23) revealed that bla NDM-1 and bla KPC-3 were located on distinct but cluster-associated plasmids.The bla NDM-1 plasmids harbored genes conferring resistance to various antimicrobial agents, including aminoglycoside [aac(6′)-Ib)], trimethoprim (dfrA14), phenicol (catB3), bleomycin (ble), and sulphonamide (sul1).In contrast to BSI_389-23, CRK2202 and CRKP2205 carried an additional AMR plasmid that did not harbor bla NDM-1 or bla KPC-3 .Further analysis showed that while the plasmids in CRK2202 and CRKP2205 were identical, they differed from those in BSI_389-23, not only in the number and type of AMR genes but also in the genetic structure surrounding these genes.In our study, ST512 isolates (CRK2202 and CRKP2205) carried IncC plasmids with bla NDM-1 and IncFIB(pQil)/IncFII(K) plasmids with bla KPC-3 .On the other hand, the ST11 isolate (BSI_389-23) carried an IncFII(pKPX1) plasmid with bla NDM-1 and an IncFIB(K)/IncFII(K) plasmid with bla KPC-3 .These findings align with previous studies (17,24,28,29), suggesting that bla NDM-1 may be associated with a higher diversity of plasmids (IncC or IncF in our study) than the bla KPC-3 gene and that bla NDM-1 may be co-located with bla CTX-M-15 on IncF plasmids.However, homologous regions between the plasmids of all three clones suggest possible links or movements of patients between the two hospitals.
Our analysis found that genes in KP isolates, whether located on plasmids or chromosomes, impacted several antimicrobial agents crucial for BSI.According to the Elias at al. 's study (23), the mgrB/pmrB genes conferring resistance to colistin, a last-resort treatment for MDR Gram-negative bacterial infections (4), displayed mutations across all three isolates (CRK2202, CRKP2205, and BSI_389-23).Despite suggestions of poten tial cross-resistance between colistin and cefiderocol (30), the colistin-resistant isolate BSI_389-23 remained phenotypically susceptible to cefiderocol.This aligns with evidence indicating that co-expression of multiple β-lactamases, often in combination with porin (OmpK35/OmpK36) mutations, may not suffice to induce cefiderocol resistance in KP isolates (30).Notably, except for BSI_389-23, which harbored the ybt gene, no simultane ous presence of AMR and virulence determinants was found in all isolates.

Strengths and limitations
Our study, based on WGS analysis of a limited number of KP isolates, aligns with similar research on KPC/NDM co-producing carbapenem-resistant KP isolates from hospital ized patients (14,31).Despite its retrospective nature and limitations in drawing firm conclusions on isolate clustering due to the lack of detailed patient movement and environmental culture data, our findings are relevant for strengthening strategies against the dissemination of MDR organisms in Italian hospitals.To address the global concern of plasmid-mediated carbapenem resistance gene spread, we employed LR WGS on 50% of isolates, chosen based on core-genome clustering.Similar strategies have been used in related studies (32).Recognizing the limitations of Kleborate analysis (21,24), we supplemented it with AMRFinderPlus for a more comprehensive characterization of AMR and virulence determinants from genome data.

Conclusions
Our findings enhance understanding of NDM/KPC co-producing KP within CG258 (ST512 and ST11), which may display a BSI-associated MDR phenotype.WGS analysis revealed plasmid localization of most AMR genes, including bla KPC-3 and bla NDM-1 , with strong evidence of clonal transmission among patients.This emphasizes the critical role of genomic surveillance in mitigating the emergence and spread of AMR genes or plasmids associated with these isolates in hospital settings.

Study isolates
Of KP isolates included in the study, five (CRKP2201 to CRKP2205) were from BSIs that occurred in COVID-19 patients at the GVM-ICC COVID-3 hospital in Rome, Italy, during a 6-month period (February to August) in 2022.A sixth isolate (BSI_389-23) was from BSI that occurred in a non-COVID-19 patient at the Fondazione Policlinico Universitario A. Gemelli IRCCS hospital in Rome, Italy, in February 2023.Isolates were identified at the species level using the Bruker Daltonics (Bremen, Germany) Biotyper matrix-assisted laser desorption ionization-time of flight mass spectrometry system.
The susceptibility of the isolates to all clinically relevant antimicrobial agents was tested using broth microdilution or disk diffusion methods (Table S1), and the minimum inhibitory concentration results were interpreted accord ing to the European Committee on Antimicrobial Susceptibility Testing 13.0 version breakpoints (https://www.eucast.org/fileadmin/src/media/PDFs/EUCAST_files/Breakpoint_tables/v_13.0_Breakpoint_Tables.pdf).Among β-lactams, imipenem and meropenem had a minimum inhibitory concentration of >8 µg/mL.All isolates were found to co-produce NDM-and KPC-type carbapenemases using the NG-Test Carba 5 immunochromatographic assay (NG Biotech, Guipry, France) and to exhibit resistance to CAZavi.
Five patients died within 3-6 days after BSI onset (i.e., when diagnostic blood cultures were taken and patients were empirically treated with broad-spectrum antibiotics), while the sixth patient (i.e., infected with the BSI_389-23 isolate) survived and received cefiderocol as antimicrobial therapy.

DNA extraction and whole-genome sequencing
Before sequencing, KP isolates (n = 6) were grown overnight at 37°C on 5% sheep blood agar and subjected to DNA extraction using the DNeasy Blood and Tissue kit (QIAGEN, Hilden, Germany) protocol.The concentration and purity of extracted DNA were determined using a NanoDrop2000 spectrophotometer (Thermo Fisher, Waltham, MA).

Phylogenetic analysis
SR assemblies allowed us to determine the ST of each isolate using a seven-loci-based MLST scheme (35) and to create an ad hoc cgMLST scheme using chewBBACA (v.2.6.0)(36).This resulted in a scheme of 864 loci, based only on genes present in 100% of the genomes that we analyzed.A minimum spanning tree was generated from the cgMLST scheme profiles and visualized using GrapeTree (v.1.5.0) (37).

FIG 1
FIG 1 Phylogenetic trees showing differences between the six KP isolates from patients with BSI at two hospitals in Italy.The minimum spanning trees were created based on cgMLST (panel A) or core genome SNPs (panel B).In each tree, nodes are denoted as circles, with different colors indicating the three clusters formed by isolates.Except for one (larger) node (panel B), which includes two isolates, each of (remaining) nodes represents a single isolate.Numbers indicate allele (panel A) or SNP (panel B) differences between nodes.In both trees, isolates in cluster A (CRKP2202 and CRKP2203) or cluster B (CRKP2201, CRKP2204, and CRKP2205) differed markedly from the isolate in cluster C (BSI_389-23).

FIG 2
FIG 2 Resistome profiling for the six KP isolates in the study.A matrix of the distribution of AMR determinants (x axis) among the isolates within the indicated cgMLST clusters (A, light blue; B, dark blue; and C, green) is shown.According to different colors, blocks indicate the presence of an acquired gene (red, aminoglycoside resistance; blue, β-lactam resistance; turquoise, trimethoprim resistance; yellow, sulphonamide resistance; orange, fluoroquinolone resistance; green, phenicol resistance; pink, fosfomycin resistance; magenta, macrolide resistance) identified using AMRFinderPlus, the presence of a mutated/truncated gene (dark gray) identified using Kleborate, or the absence of any AMR determinant (light gray).Mutations included truncation of OmpK35 and GD insertion in the OmpK36 β-strand loop (both known to contribute to the carbapenem resistance phenotype), truncations in MgrB/PmrB (known to confer colistin resistance), and mutations in the quinolone resistance-determining region of GyrA (S83L) and ParC (S80I) (both known to confer fluoroquinolone resistance).

FIG 3
FIG 3 Bandage plots of the plasmids identified in the three KP isolates (CRKP2202, CRKP2205, and BSI_389-23) for which both long-read and short-read sequencing data were available.(A) In the CRKP2202 isolate, three of five plasmids (size range, 1,666-149,749 bp) were circular, and two were non-circular.(B) In the CRKP2205 isolate, seven of seven plasmids (size range, 13,636-207,079 bp) were circular.(C) In the BSI_389-235 isolate, four of eight plasmids (size range, 727-173,173 bp) were circular, and four were non-circular.

FIG 5
FIG 5 Schematic representation of colistin resistance molecular determinants in the three KP isolates (CRKP2202, CRKP2205, and BSI_389-23) for which both long-read and short-read sequencing data were available.Multiple amino acid sequence alignments of MgrB (A) and PmrB (B) from the three hybrid assemblies against the reference (Kleborate retrieved) wild type were performed.(A) MgrB from BSI_389-23 is truncated at amino acid position 34, MgrB from CRKP2205 is truncated at amino acid position 3, and MgrB is absent in CRKP2202.(B) PmrB from all three isolates showed the R256G mutation (denoted as a red x).

TABLE 1
Summary of the characteristics of the plasmids identified in three K. pneumoniae isolates included in the study a a "-" means the absence of data about the indicated characteristic.