Multiplicity of blaKPC Genes and pKpQIL Plasmid Plasticity in the Development of Ceftazidime-Avibactam and Meropenem Coresistance in Klebsiella pneumoniae Sequence Type 307

ABSTRACT In 2021, Klebsiella pneumoniae sequence type 307 (ST307) strains causing pulmonary and bloodstream infections identified in a hospital in Rome, Italy, reached high levels of resistance to ceftazidime-avibactam (CZA). One of these strains reached high levels of resistance to both CZA and carbapenems and carried two copies of blaKPC-3 and one copy of blaKPC-31 located on plasmid pKpQIL. The genomes and plasmids of CZA-resistant ST307 strains were analyzed to identify the molecular mechanisms leading to the evolution of resistance and compared with ST307 genomes at local and global levels. A complex pattern of multiple plasmids in rearranged configurations, coresident within the CZA-carbapenem–resistant K. pneumoniae strain, was observed. Characterization of these plasmids revealed recombination and segregation events explaining why K. pneumoniae isolates from the same patient had different antibiotic resistance profiles. This study illustrates the intense genetic plasticity occurring in ST307, one of the most worldwide-diffused K. pneumoniae high-risk clones.

KPC plasmids but related pKPN-like plasmids carrying bla CTX-M-15 (10). On pKPN, up to five putative virulence clusters were identified: the lac operon, the Fec-like iron (III) dicitrate and the glutathione ABC-transport systems, the urea transport system, and the cluster for glycogen synthesis. ST307 also carried the yersiniabactin siderophore, mobilized by an integrating conjugative element (ICE) (8,10,11).
In 2016, a novel combination of a third-generation cephalosporin (ceftazidime), with the non-beta-lactam beta-lactamase inhibitor (avibactam), was introduced into clinical practice to treat infections by carbapenem-resistant Enterobacterales isolates producing class A and D carbapenemases (12).
In 2019, we described the emergence of the first ST307 strains showing ceftazidimeavibactam (CZA) resistance at the University Hospital Policlinico Umberto I (PUI) in Rome, Italy (13). The bla KPC-31 gene located on the Tn4401 transposon, was the most frequent (11 strains of 32) CZA resistance mechanism identified in that study. Among KPC-31 producers, ST307 (5/11) and ST101 (4/11) were the most represented STs, while in 17 CZA-resistant ST512 strains, 6 different KPC variants were observed, and only one strain produced KPC-31. KPC-31 is characterized by a single amino-acid substitution (a D-to-Y change at position 179 [D179Y]) that confers CZA resistance at the expense of carbapenemase activity (14). Therefore, the CZA-resistant ST307 strains identified in 2019 were susceptible to carbapenems.
In 2021, despite epidemiological association linking the cases, a KPC-31-producing ST307 isolate identified at the PUI showed interesting differences in resistance levels to CZA and carbapenems. In this study, we describe the CZA resistance mechanisms and plasmid evolution that occurred in contemporary ST307 strains, compared with the genomes of historical strains isolated from the same hospital and with a selection of ST307 genomes from the public NCBI data set.
A midpoint-rooted phylogenetic tree, based on 4,362 core genes, showed at least three different ST307 lineages (separated by .60 single nucleotide polymorphisms [SNPs]) circulating in the hospital in the 2019-2021 period (Fig. 1). One lineage included the carbapenem-susceptible strains 1802 and 1203 and the KPC-3-producing strain 3, differing from each other by only 6 SNPs. A second lineage included the CZA-resistant strains identified at the hospital. Strain 0213-2021 was related to 21 and 27B (15 SNPs). Strains 1001-2021 and 0323-2021, isolated from the same patient, were highly related to each other (1 SNP difference in the core genome) but differed by 45 to 77 SNPs from the other CZA strains (Fig. 1). Both lineages carried the yersiniabactin (ybt) siderophore sequence type 384 (YbST384) (11), mobilized by the integrating conjugative element ICEKp4. On the same branch with these two lineages, there were 2 genomes of OXA-48-producing isolates (ASM1337825 and ASM1337827) from pets in Switzerland (15) and one from South Korea (ASM1296340). These isolates also carried the YbST384 locus. The genome of the susceptible strain 0603 represented the third lineage at the hospital, with a mean distance of 81 SNPs. It carried a different YbST157 ybt siderophore, mobilized by an ICEKp3 integrating conjugative element. The isolate showing the closest phylogenetical relationship with 0603 was sampled in Rome at a different hospital and did not carry any virulence determinant (ASM1340838).
Resistome and mobilome of ST307 strains. No carbapenemase genes were present in strains 1802, 1203, or 0603, showing reduced susceptibility to both CZA and MEM. These isolates carried several antimicrobial resistance genes (Data set S2), including bla CTX-M- 15 . In 1802 and 1203, this gene is located on indistinguishable pKPN plasmids (GenBank accession number OM489434), but pKPN in 1203 is integrated into the chromosome (OM489428) ( Table 2; Fig. S1).

FIG 1
Phylogenetic tree based on a core genome alignment of Klebsiella pneumoniae ST307 isolates. The figure shows an unrooted maximum likelihood phylogenetic tree based on a concatenated core gene alignment (4,362 genes) of 39 K. pneumoniae genomes belonging to ST307. Of the 39 genomes, 30 (indicated by empty circles) were downloaded from GenBank, 3 genomes (3, 21, and 27B) were already described (13), and 6 genomes (0213-2021, 0323-2021, 1001-2021, 1202, 1803, and 0603) were obtained in this study. Carbapenem-susceptible strains 1802, 1203, and 0603 are indicated by green dots, the KPC-3-producing strain 3 is indicated by a yellow dot, CZA-resistant KPC-31-producing strains are identified by blue dots, and the CZA-MEM-resistant strain 0323-2021 is indicated by a red dot. The colored squares indicate the presence and empty squares the absence of the feature in the respective metadata column. Darker colors indicate the presence of multiple copies of the feature in the respective genome. Colored squares in the data set report the major resistance genes and replicons: gray, replicons; pink, betalactamase; magenta, aminoglycoside resistance; blue, tetracycline resistance; cyan, quinolone resistance; and green, sulfonamide resistance.
CZA-resistant ST307 strains carry bla KPC-31 , located on transposon Tn4401a in pKpQIL-like plasmids. Strains 21 and 27B carried the gene on a large plasmid composed of a fusion of pKPN and pKpQIL, as previously described (13). Comparison of the original pKpQIL (GenBank accession number GU595196) (6) and pKpQIL in strain 3 (99.82% nucleotide identity; 100% coverage), showed several nucleotide mutations: 63 SNPs in the traC gene (positions 77821 to 80440; ON002623.2), causing 9 amino acid substitutions in the deduced TraC protein sequence; 14 SNPs in the traD gene (positions 61389 to 61586; ON002623.2), causing 3 amino acid substitutions in the deduced TraD protein sequence; and other SNPs identified in 5 genes encoding hypothetical proteins. The traC SNPs were observed only in pKpQIL of strain 3, while the traD gene allele was conserved in all ST307 pKpQIL sequences analyzed in this study.
Strain 0323-2021 produced ambiguous hybrid Illumina and ONT read assemblies, including a version of pKpQIL carrying three copies of the bla KPC gene, two copies of bla KPC-3 , and one copy of bla KPC-31 . The complete plasmid assembly required further evaluation, as described below.
In 0603, wild-type OmpK35 and OmpK36 were detected. In 1203 and 1802, frameshift mutations were present in the deduced protein sequences of OmpK35 and OmpK36, predicting premature termination of translation. All CZA-resistant ST307 strains showed wildtype OmpK35 and OmpK36 porins, except strain 1001-2021, where a frameshift mutation in the deduced OmpK36 protein sequence predicted the premature termination of translation ( Table 2).
Analysis of plasmids in the ST307 0323-2021 CZA-MEM-resistant strain. Because of the complexity of the repeated elements in strain 0323-2021, the hybrid Illumina and ONT read assembly needed experimental confirmation. ONT was applied to the original K. pneumoniae strain 0323-2021 using purified plasmid DNA as the template. Running ONT on a good plasmid extract, the read length graph showed highly represented reads covering the entire plasmid length (Fig. S2). At least three reads visible as higher bars in the graph were interpreted as plasmids of 83 kb, 136 kb, and 140 kb, coexisting in this K. pneumoniae strain.
To discern these plasmids, purified plasmid DNA from the K. pneumoniae strain was used to transform chemically competent Escherichia coli DH5-a cells. The following three types of transformants were obtained from strain 0323-2021 ( Fig. 2; Table S2). 1. Type 1 transformants (prototype 0323-37-TR) showed MEM and CZA resistance. ONT sequencing showed a complete, circular plasmid of 136 kb, carrying 2 copies of bla KPC-3 and 1 copy of bla KPC-31 (pKpQIL-0323). One of the two bla KPC-3 copies was in the Tn4401a identified in the same position as in pKpQIL-3 (copy 1 in Fig. 2). The second Tn4401a::bla KPC-31 (copy 2) was inserted in the opposite orientation as copy 1, downstream of the traU gene. The third Tn4401a::bla KPC-3 copy (copy 3) was integrated in the same orientation as copy 1 in the traS gene. We hypothesize that recombination between copies 1 and 2 in opposite orientations caused the inversion of the pKpQIL region comprising the traT, traD, traI, traX, and finO genes up to the FII(K) replicon. Moreover, recombination between copy 2 and copy 3 could have led to the inversion of the transfer locus genes downstream of the traU integration site (genes traS to traN).
The 5-bp target-site duplications (TSDs) that originated by transposition of the Tn4401a copies were identified (Fig. S3). To complete the plasmid analysis, the location of the Tn4401a::bla KPC-3 and Tn4401a::bla KPC-31 copies in pKpQIL-0323 was also confirmed by PCR and Sanger sequencing of the amplicons, using primers designed in the traS, traN, and bla KPC genes (Table S1). In pKpQIL-0323, the klebicin B-encoding genes were inserted between the finO gene and the FII(K) replicon. These genes were not present in other genomes of our collection (Fig. 2). 2. Type 2 transformants (prototype 0323-1-TR) showed MEM resistance and CZA susceptibility. In these transformants, a plasmid of 83 kb was detected, carrying a single copy of the Tn4401a::bla KPC-3 . This plasmid was probably generated by the recombination of pKpQIL-0323, which occurred between the directly oriented Tn4401a::bla KPC-3 and Tn4401a::bla KPC-31 copies 1 and 3, resulting in the loss of the Tn4401a::bla KPC-31 copy and the deletion of the pKpQIL backbone at the traU gene. 3. Type 3 transformants (prototype 0323-11-TR) showed MEM susceptibility and CZA resistance. In these transformants, the same 83-kb plasmid identified in the type 2 transformants was observed, but the recombination between the two Tn4401a::bla KPC-3 and Tn4401a::bla KPC-31 copies resulted in the maintenance of a single Tn4401a::bla KPC-31 copy and the deletion of the backbone at the traU gene. This plasmid was identical to the 83-kb plasmid identified in K. pneumoniae strain 1001-2021 (Fig. 2).
These results demonstrate that the 83-kb and 136-kb versions of pKpQIL were coresident within K. pneumoniae strain 0323-2021. The higher 140-kb read observed in the ONT profile (Fig. S2) was interpreted as pKPN, also coresident with the two pKpQIL variants in this K. pneumoniae strain.

DISCUSSION
This study highlights the consequences of the strong CZA and MEM selective pressure on K. pneumoniae ST307. This high-risk clone shows sharp plasticity of plasmids circulating in multiple lineages, flowing within the hospital.
Reduced susceptibility to MEM in ST307 ancestor strains 1802 and 1203, negative for carbapenemase genes, was conferred by the production of CTX-M-15 in association with frameshift mutations predicted in the OmpK35 and OmpK36 porin sequences, as previously described (16). The same mechanisms could also contribute to the slight increase of the CZA MICs observed in these strains, since no other known resistance mechanisms were detected in their genomes. The bla CTX-M-15 gene was located on pKPN in strain 1203, integrated into the chromosome. These strains belonged to the same subclade of strain 3 that developed MEM resistance by the acquisition of bla KPC-3 on pKpQIL. The gain of this plasmid requires a fully functional OmpK36 for conjugation (17). Yet, given the predicted depletion of this protein in isolates 1802 and 1203, the ancestor of pKpQIL-positive strains was presumably an isolate akin to 1802 or 1203, with a wild-type OmpK36 porin.
ST307 isolates gained CZA resistance by the acquisition of bla KPC-31 located on pKpQIL plasmids. In the historical ST307 strains of our collection, pKpQIL was observed in fusion with pKPN. In the recent isolates, bla KPC-31 was still located on pKpQIL, but the plasmid had evolved by transposition, recombination, and deletion events, associated with the loss of self-transmissibility.
Among the CZA-resistant isolates, strain 0323-2021 reached high resistance levels to both CZA and carbapenems by acquisition of three copies of Tn4401a on the same pKpQIL, carrying bla KPC-31 and bla KPC-3 genes. However, this configuration was unstable. In the same strain, smaller 83-kb pKpQIL versions were observed, probably due to homologous recombination between directly oriented Tn4401a copies, leading to recombined plasmids carrying Tn4401a::bla KPC-31 or Tn4401a:bla KPC-3 copies, respectively.
The 83-kb pKpQIL, carrying the Tn4401a::bla KPC-31 copy, was identified in strain 1001-2021 isolated from a blood culture from the same patient from whom strain 0323-2021 was recovered. This is an important finding because it demonstrates that the recombination and segregation of the 83-kb pKpQIL is not an artifact that arose in transformants in vitro; instead, it was an event that occurred in the patient in vivo. This plasmid confers CZA resistance, but the loss of the Tn4401a::bla KPC-3 copy is expected to restore susceptibility to MEM. However, strain 1001-2021 still showed an MEM MIC of 8 mg/L, probably due to the lack of the OmpK36 porin, that was not observed in the other KPC-31-producing strains, which showed lower MICs for carbapenems.
It should be noted that plasmids from both strains 1001-2021 and 0323-2021 all lacked a complete transfer locus because of the observed rearrangements. Consequently, the identified plasmids are expected not to be self-conjugative and cannot be transferred to other bacteria but can be propagated vertically within the ST307 clone.
The genetic complexity of the bacteria described in this study needed a lot of experimental work to be correctly described. It was not discerned by standard whole-genome sequencing (WGS) but was also at the limit of resolution of the hybrid ONT-Illumina combined approach.
Accurate molecular methods and monitoring of phenotypical changes can help us trace the elevated capacity of high-risk clones to evolve toward resistance under positive selection. Still, the careful use of novel antibiotics is a crucial action to prevent the development of resistance to life-saving antimicrobial agents in these highly flexible clones.

MATERIALS AND METHODS
Strain isolation and susceptibility testing. Bacteria were isolated from samples processed for routine diagnostics; the species was identified by the matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) system (Bruker Daltonik GmbH, Bremen, Germany). Antimicrobial susceptibility was tested using the MicroScan WalkAway system (Beckman Coulter Inc., Brea, CA, USA), and the CZA and meropenem-vaborbactam (MVA) MICs were confirmed using Etest strips (Liofilchem, Roseto degli Abruzzi, Italy). The cefiderocol MICs were tested using the ComASP compact antimicrobial susceptibility panel for 0.008 to 128 mg/L cefiderocol (Liofilchem). All CZA-resistant strains were tested for bla KPC genes (GeneXpert, Cepheid, CA, USA).
ST307 analyzed in this study. Strains 0213-2021, 0323-2021, and 1001-2021 were isolated from two patients hospitalized at the neurological surgery unit of PUI hospital (Table 1). A historical ST307 collection from 2019 was added to the study for comparison and included the following: strains 0603, 1203, and 1802, not producing carbapenemases; the KPC-3-producing strain 3; and the KPC-31-producing strains 27B and 21. Strains 27B and 21 were representative of the 5 CZA-resistant ST307 strains identified in 2019 at PUI (13). All strains resistant to CZA were susceptible to MVA (MIC, ,0.2 mg/L) and cefiderocol (MIC, #2 mg/L).
Whole-genome sequencing. Whole-genome sequencing (WGS) was carried out on purified genomic DNA (Macherey-Nagel DNA extraction kit; Düren, Germany) using Illumina MiSeq (San Diego, CA, USA) sequencing. The Illumina reads were assembled through the public European Galaxy server (https:// usegalaxy.eu/) using the SPAdes v3.15.3 pipeline (18). Oxford Nanopore Technologies (ONT) sequencing was performed as previously described (13). One round of polishing of the ONT reads was carried out using the Flye assembly (19), while short-read assembly was automatically performed using the SPAdes assembler in the Galaxy wrapper of the Unicycler v0.4.8.0 tool. Hybrid assembly was performed using normal bridging mode (20), and standard polishing parameters (lowest k-mer size, expressed as a fraction of the read length, of 0.2; filtering out contigs lower than 0.25 of the chromosomal depth) were adopted. The Staramr tool (21) was used on the resulting assemblies to detect antimicrobial resistance genes. The BacAnt tool (22) was used for annotation of transposable elements.
Plasmid DNA from 100 mL LB liquid broth cultures of E. coli DH5-a transformants and K. pneumoniae strains was purified using a plasmid midiprep purification kit (PureYield; Promega Italia Srl, Milan, Italy) and concentrated using Microcon 100 centrifugal filters (Merck KGaA, Darmstadt, Germany). Plasmid libraries were prepared with 400 ng of purified plasmid DNA and sequenced using ONT. The ONT reads were assembled using the Flye tool (19).
Phylogenetic analysis. A total of 2,040 completely assembled K. pneumoniae genomes were downloaded from the public NCBI data set (https://www.ncbi.nlm.nih.gov/datasets/docs/v1/how-tos/genomes/ download-genome/). The Kleborate tool (23) was used to assess the STs of the downloaded sequences, and 30 genomes belonging to ST307 were retrieved and annotated using Prokka v1.14.6 (24). The resulting general feature formats (GFFs) were analyzed using Roary v3.11.3 (25). The core genome was defined as the genes found in all 39 isolates (25). Removal of recombining regions from the pangenome produced by Roary was carried out using Gubbins (26), generating a maximum likelihood phylogenetic tree using RAxML with default parameters (27). The tree and metadata were visualized with Microreact (28) and adjusted using the open-source software Inkscape.
Plasmid transformation. Purified plasmid DNA was obtained from overnight 50-mL LB liquid broth cultures of K. pneumoniae isolates. Plasmid extraction was performed using the PureYield plasmid midiprep system (Promega Italia Srl). Plasmid DNA was transformed into chemically competent Escherichia coli DH5-a cells (Life Technologies, Thermo Fisher Scientific, Waltham, MA, USA), selecting transformants on LB agar plates containing ceftazidime (CAZ, 6 mg/L). After 24 h, colonies were screened for bla KPC using the KPC FW and KPC RV primer pair, as previously described (13).
The CAZ and MEM MICs of the KPC-positive E. coli DH5a transformants were tested using microdilution (MicroScan system; Beckman Coulter Inc.). The CZA MICs were determined using Etest strips (Liofilchem).
The plasmid content of transformants showing different CZA and MEM resistance was sequenced using ONT. To confirm the position of bla KPC-31 and bla KPC-3 at the Tn4401 copies flanking the traS and traN genes, respectively, PCR was performed with the KPC INT /TraN RV and KPC INT /TraS FW primer pairs, followed by nested PCR performed with the KPC INT and KPC RV primer pair. The bla KPC amplicons were sequenced using the Sanger method with both the KPC INT and KPC RV primers (see Table S1 in the supplemental material).
Ethics. Procedures performed in the study were in accordance with the ethical standards of the Institutional and National Research Committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.6 MB.