Ceftazidime–avibactam resistance in Klebsiella pneumoniae sequence type 37: a decade of persistence and concealed evolution

The first reports of carbapenem-resistant Enterobacterales in our hospital date back to 2006. In that period, few ertapenem-resistant but meropenem-susceptible Klebsiella pneumoniae isolates belonging to sequence type (ST) 37 were retrieved from clinical samples. These strains produced the CTX-M-15 extended spectrum β-lactamase, OmpK35 was depleted due to a nonsense mutation, and a novel OmpK36 variant was identified. Yet, starting from 2010, Klebsiella pneumoniae carbapenemase (KPC)-producing ST512 isolates started prevailing and ST37 vanished from sight. Since 2018 the clinical use of the combination of ceftazidime–avibactam (CZA) has been introduced in clinical practice for the treatment of bacteria producing serine-β-lactamases, but KPC-producing, CZA-resistant K. pneumoniae are emerging. In 2021, four CZA-resistant ST37 isolates producing KPC variants were isolated from the same number of patients. blaKPC gene cloning in Escherichia coli was used to define the role of those KPC variants on CZA resistance, and whole genome sequencing was performed on these isolates and on three ST37 historical isolates from 2011. CZA resistance was due to mutations in the blaKPC genes carried on related pKpQIL-type plasmids, and three variants of the KPC enzyme have been identified in the four ST37 strains. The four ST37 isolates were closely related to each other and to the historical isolates, suggesting that ST37 survived without notice in our hospital for 10 years, waiting to re-emerge as a CZA-resistant K. pneumoniae clone. The ancestor of these contemporary isolates derives from ST37 wild-type porin strains, with no other mutations in chromosomal genes involved in conferring antibiotic resistance (parC, gyrA, ramR, mgrB, pmrB).


INTRODUCTION
Carbapenem-resistant Enterobacterales are one of the greatest threats to public health systems worldwide [1]. In Italy, the European Centre for Disease Prevention and Control surveillance system reports dramatic increases of carbapenem-resistant Klebsiella pneumoniae from 1 % in 2009 to 34 % in 2016 (currently, 2020, 29.5 %; https://atlas.ecdc.europa.eu/). In Italy, the main representative of this threat is K. pneumoniae carbapenemase (KPC)-producing isolates, especially the sequence types (STs) 512, 307 and 101 [2]. The resistance to carbapenems is the outcome of decades of evolution, mainly based on horizontal gene transfer of plasmids carrying multiple resistance genes [3,4].
In the Policlinico Umberto I (PUI) hospital, carbapenem resistance in Gram-negative bacteria was not reported until 2006. However, a few ertapenem-resistant, meropenem-susceptible K. pneumoniae strains were identified at that time [5], which were assigned to ST37 based on multi-locus sequence typing (MLST [6]). These strains carried the genes coding for the CTX-M-15 extended spectrum β-lactamase (ESBL) and showed the depletion of OmpK35 due to a nonsense mutation, associated with a novel OmpK36 variant [5]. Resistance to meropenem was observed in the ESBL-producing ST37 strains that did not produce either the OmpK35 or the OmpK36 porin [5].
Carbapenem-resistant ST37 strains were not reported after the introduction in the PUI hospital of isolates belonging to clonal group (CG) 258 carrying bla KPC . ST37 K. pneumoniae seemingly disappeared, substituted by KPC-producing clones [7].
A decade later, the spread of carbapenem resistance in K. pneumoniae had not ceased. Since 2018 the clinical use of the combination of ceftazidime-avibactam (CZA) has been introduced in the hospital for treatment of bacteria producing serine-β-lactamases. Yet, KPC-producing CZA-resistant K. pneumoniae strains are emerging [8,9].
Here, starting from the report of four KPC-producing CZA-resistant ST37 K. pneumoniae isolated in 2021 causing respiratory tract and bloodstream infections, we endeavour to reconstruct the evolution of this clone at PUI and compare contemporary and historical isolates.

Bacterial strain isolation and susceptibility testing
Isolated colonies were identified by a MALDI-TOF MS system (Bruker Daltonik). Antimicrobial susceptibility was determined by the MicroScan WalkAway system (Beckman Coulter). Isolates showing a carbapenem-resistant phenotype were tested using the real-time PCR assay Xpert Carba-R kit for the GeneXpert system (Cepheid). CZA minimal inhibitory concentrations (MICs) were determined using the CZA gradient test (Liofilchem). CZA-resistant isolates carrying bla KPC were subjected to complete DNA sequencing of the bla KPC gene obtained by amplification, as previously described [10].

Whole genome sequencing and assembly
Whole genome sequencing (WGS) was obtained by Illumina MiSeq (Illumina) after genomic DNA purification following the Isolate II genomic DNA Extraction Kit procedure (Bioline) for four contemporary and three historical ST37 isolates. The Nextera XT DNA sample preparation kit generated paired-end libraries with the 2×300 paired-end protocol (Illumina). Illumina reads were assembled at the public Europe Galaxy Server (https://usegalaxy.eu/) by version 3.15.3 of the SPAdes [11] pipeline.
Representative isolates of contemporary and historical isolates were also subjected to Oxford Nanopore Technologies (ONT) sequencing, as previously described [10].

Impact Statement
Owing to its capacity to collect multiple antimicrobial-resistance determinants, Klebsiella pneumoniae is one of the greatest threats to public health. Typically, the dissemination of multidrug-resistant K. pneumoniae is driven by a few high-risk clones, such as those belonging to clonal groups (CGs) 101, 147, 258 and 307. Even though K. pneumoniae isolates belonging to sequence type (ST) 37 have been listed as a high-risk multidrug-resistant clone, typically linked to the dissemination of the 16S rRNA methylases RmtB and ArmA, this ST is poorly represented in both genomic databases and the scientific literature. In this study, we dissected an outbreak of ceftazidime-avibactam-resistant isolates belonging to ST37, comparing their genomes with others from different times and places belonging to the same ST. ST37 was the endemic clone in our hospital 10 years ago, when it was overshadowed by K. pneumoniae carbapenemase (KPC)-producing CG258 isolates. Today, ST37 is striking back, with four KPC-producing isolates resistant to ceftazidime-avibactam.

Phylogenesis and synteny
Thirty-nine genome sequences (seven obtained in this study and 32 downloaded from the GenBank database) were annotated using Prokka version 1.14.6 [13] and the resulting general feature formats (GFFs) were analysed using Roary v3.11.3 [14] to build a core genome alignment and a gene presence/absence tabular file.
A consensus phylogenetic tree based on 100 000 bootstraps was generated with IQ-TREE using the TIM+F+I model of substitution [15]. The tree and metadata were visualized with Microreact [16] and adjusted using the open-source InkScape software.
The temporal signal resulting from these phylogenetic data was evaluated by TempEst v1.5.3 using a root-to-tip regression analysis as a function of the sampling time [17]. This preliminary analysis yielded a low correlation coefficient (0.1871 best root), and thus no Bayesian-based analysis was performed to infer a time-scaled phylogeny for these isolates. Synteny maps were created by blastn and visualized using the Circos tool [18], adjusting the resulting plot with the open-source InkScape software.

Analysis of the core and of the accessory genome
Single Nucleotide Polymorphisms (SNPs) were analysed by the Snippy tool (https://github.com/tseemann/snippy) to determine the number and the position of the SNPs between isolates 9362 and 1020. Given the lack of core-genome distance between contemporary isolates, isolate 1020 was arbitrarily chosen as the reference. All chromosomal SNPs were plotted over the 1020 genome. Furthermore, we performed a Clustal Ω alignment [19] on proteins encoded by specific chromosomal genes associated with antimicrobial resistance (gyrA and parC for fluoroquinolone resistance, ramR for tigecycline resistance, ompK35 and ompK36 for β-lactam resistance, mgrB and pmrB for colistin resistance). Capsular polysaccharide (CPS) and the lipopolysaccharide (LPS) loci were analysed using the Kaptive tool [20].
To evaluate which genes were present in the contemporary isolates and which were present in the historical ones, we deployed the Scoary tool [21]. In addition, we performed a function-and a sequence-based analysis by the Rapid Annotation of microbial genomes using Subsystems Technology (RAST) and the associated SEED tools [22]. Major genomic differences, defined by the presence or absence of at least five consecutive coding sequences (CDSs), were manually curated.

pKpQIL plasmid transformation
Plasmid DNA was extracted from overnight growth of 10 ml liquid LB broth of K. pneumoniae isolate 1020 by the PureYield plasmid midiprep system (Promega Italia). Purified plasmid DNA was used to transform chemically competent Escherichia coli DH5-α cells (Life Technologies, Thermo Fisher Scientific), selecting transformants on LB agar plates containing ceftazidime (6 mg l −1 ). After 24 h, colonies were screened for bla KPC and replicon content (PBRT 2.0; Diatheva). E. coli DH5-α transformants positive for both bla KPC and the FIB replicon specific for the pKpQIL plasmid were tested in triplicate for several antibiotics by microdilution (MicroScan system). CZA MICs were tested in triplicate by the CZA gradient test (Liofilchem).

Cloning of bla KPC mutants in pCR-Blunt II TOPO vector
Each isolate carrying a different bla KPC mutant was used as template for an AccuPrime Pfx DNA polymerase-based PCR as previously described [10] and ligated into the pCR-Blunt II TOPO vector (ThermoFisher). Ligation mixtures were introduced by transformation in chemically competent TOP10 cells (Life Technologies, Thermo Fisher Scientific) and transformants were screened on LB agar plates containing kanamycin (ImMedia Kan agar; Invitrogen, Thermo Fisher Scientific). The accuracy of the cloning was assessed by PCR and Sanger sequencing of the amplicons as previously described [10]. E. coli TOP10 cells producing KPC variants were tested in triplicate for several antibiotics by microdilution (MicroScan system). CZA MICs were tested in triplicate by the CZA gradient test (Liofilchem).

Klebsiella pneumoniae ST37 collection
In this study, seven K. pneumoniae isolates were studied and WGS was performed. Strains from the historical collection (SC26, 9362 and 4011) were isolated in 2011 during a KPC-positive and KPC-negative case-control study performed in different hospitals of Rome, Italy [26]. Among the KPC-negative isolates, ST37 9362 (ertapenem-susceptible) and 4011 (ertapenem-resistant) strains were identified at the PUI, while isolate SC26 (ertapenem-resistant) was from the San Camillo Hospital.
Bacterial isolates from the contemporary collection (1015, 1016, 1020, 1021) were isolated from samples collected during routine microbiological processes in PUI hospital in 2021. The four bacterial strains were from four different infected patients hospitalized in the same ward, causing respiratory tract infections or bacteraemia ( Table 1).

Retrieval of ST37 isolates from the GenBank database
As of June 2022, 2934 completely assembled K. pneumoniae genomes were downloaded from the public NCBI dataset (https://www. ncbi.nlm.nih.gov/datasets/docs/v1/how-tos/genomes/download-genome/) and screened by the Kleborate [23] and the Kaptive [20] tools. Thirty-two univocal isolates belonging to ST37 were retrieved and included in the study for genomic comparison.

Analysis of the core genome in contemporary and historical ST37
Thirty-nine genome sequences (seven obtained in this study and 32 downloaded from the GenBank database) were used to build a core genome-based alignment.
The ST37 phylogenetic tree shows two deep branches (Fig. 1). One less populated branch (eight genomes) appears to be rather heterogeneous, while the most populated one (31 genomes) is made up of three major branches. The historical and the contemporary ST37 isolates from Rome displayed a rather conserved core genome and were in one of these three branches (coloured in pink in Fig. 1), sharing a tight phylogenetic bond (coloured in blue in Fig. 1). On this branch, three genomes isolated from patients in the USA (New York, ASM1417001; Chicago, ASM1202989; Washington, ASM1690345) demonstrated close phylogenetic relationships with historical and contemporary ST37 isolates from Italy.
Chromosomal differences were calculated between isolates 9362 and 1020 as representatives of the historical and contemporary collections, respectively. This analysis identified 470 total SNPs accumulated in the decade separating the two compared strains, corresponding to an average of 47 mutations/genome per year (Dataset SD1 available in the online version of this paper). As a term of comparison, we used the previously described K. pneumoniae evolutionary rate calculated as being 10.1 substitutions/ genome/year [27]. Therefore, the accumulation of mutations in ST37 is observed to be higher, despite the fact that estimation of the mutations gained in short timeframes may vary according to several factors (i.e. mutation rate, generation time and natural selection), greatly magnifying, or reducing, the evaluation of this evolutionary measure [28].

Comparison of capsule and other chromosomal genes
In K. pneumoniae, the diversity of LPS antigens is currently restricted to 13 different serotypes. Regarding CPS the situation is more complex, since there are 77 serologically defined K-types [20,29]. In silico analysis [20] revealed that contemporary and historical isolates from Rome shared the same outer LPS (O3b) and CPS (KL38) antigens. Analysing the 32 retrieved genomes, O3b LPS antigen was identified in nine other ST37 genomes. By contrast, KL38 was detected only in the three above-mentioned genomes from the USA (Fig. 1). There is no co-linear evolution between historical and contemporary isolates considering chromosomal genes linked to antimicrobial resistance. While the three historical isolates presented a premature stop codon at position 36 in the outer membrane protein OmpK35, the contemporary isolates expressed the wild-type porin (9 % coverage because of the frameshift, 100 % identity in the 35 translated amino acids). Furthermore, OmpK36 in ertapenem-resistant SC26 and 4011 strains presented a DT insertion at position 135-136 compared to the wild-type porin, identified in the ertapenem-susceptible 9362 genome (100 % coverage and 99.42 % identity in the amino acid sequence). These OmpK36 variants were previously described as associated with ertapenem resistance [5] and were not found in contemporary isolates. This observation suggests that the ancestor of the contemporary KPC-producing isolates identified at PUI came from an ertapenem-susceptible, wild-type porin strain. Our hypothesis is that the ST37 clone which has been able to survive up to the acquisition of bla KPC was the one with fewer mutations, and not the most resistant one. In fact, despite the outer membrane protein alterations conferring higher MICs for carbapenems, they have a cost to bacterial fitness [30].
This line of reasoning can also be applied to the novel variant (K84E) of ParC (100 % coverage and 99.87 % identity in the amino acid sequence) and GyrA (S83Y and D87A) (100 % coverage and 99.77 % identity in the amino acid sequence) identified in the ST37 historical isolates that were not mutated in contemporary isolates.
No mutations were identified in both contemporary and historical isolates in colistin and tigecycline resistance associated with MgrB, PmrB and RamR proteins (100 % coverage and 100 % identity in the amino acid sequence).

Plasmids, integrative conjugative elements and phages in ST37
The three historical isolates harboured an IncF [F2:A-:B-] plasmid, carrying the bla CTX-M-15 , bla TEM-1 and bla OXA-1 β-lactamase genes. In isolate SC26, IncF was simultaneously present within the same cell with an IncR plasmid, while in 9362 and 4011 strains small plasmids were detected (Table S1). Historical isolates carried a resistome coding for ESBLs, resistance to aminoglycosides, folate pathway inhibitors, chloramphenicol, macrolides and tetracyclines (Table 1). Remarkably the armA gene, already described in other ST37 isolates from 2014 [31], was detected in the 9362 genome located in the pKPN plasmid, indicating deletion of the entire transfer locus (Table 1, Fig. S1). Any of these plasmids and the relative resistome were detected in the contemporary isolates, but which acquired the pKpQIL plasmid carrying the bla KPC gene. Beside pKpQIL, these strains carried the pKPN plasmid not harbouring any resistance gene (Table S1). The pKpQIL plasmid was not identified in the other 32 ST37 genomes retrieved from GenBank (Fig. 1).
In the chromosome, major genomic differences consisted of the presence or absence of prophages, restriction/modification systems (yeeA, yeeB and yeeC genes) and in the acquisition of the siderophore yersiniabactin (ybt) locus, which was located within the mobile genetic element ICEKp [32] in contemporary isolates (Table S2). In these isolates the structural variant ICEKp3, carrying the ybt lineage 9, integrated inside a tRNA-Asn located between the genes coding for a Na + /H + antiporter and a nitrogen assimilation transcriptional regulator (Fig. 2a). The ybt ST414 was determined based on the combination of the alleles of the genes belonging to the ybt locus. To determine its diffusion in K. pneumoniae we performed a scan with Kleborate [23] on the 2934 completely assembled genomes downloaded from GenBank. Among these genomes, only the phylogenetically related ST37 strain from New York (ASM1417001) carried a similar ybt locus combination (single nucleotide variant, SNV, of ybt ST414).
Two prophages (φ1 and φ2 in Fig. 2a) were shared by all recent and old isolates from Rome, while prophages φ3 and φ5 were specific for the historical isolates, and φ4 for the contemporary isolates (Fig. 2a, Table S2). Neither the contemporary nor the historical isolates carried any CRISPR-Cas system.
These discontinuities between 9362 and 1020 genomes were confirmed by the Scoary tool [21] and by the analysis of the RAST [22] results (Table S2).

Acquisition of bla KPC in contemporary isolates
The contemporary ST37 isolates acquired bla KPC on identical pKpQIL plasmids (100 % coverage and 99.99 or 100 % identity in the nucleotide sequence), with differences only in the bla KPC genes, coding for the KPC variants KPC-31 (isolates 1016 and 1021), KPC-70 (isolate 1015) and KPC-110 (isolate 1020) (100 % coverage and 99.66 % identity in the amino acid sequence between KPC-31 and the KPC-70 and KPC-110 variants). Synteny analysis was performed among pKpQIL_1020 (from strain 1020, NCBI accession no. OL744329), pKpQIL_17B, a pKpQIL plasmid of an ST512 isolate from PUI (MT809697) and the FIA plasmid belonging to the phylogenetically closest ST37 isolate producing KPC-3 from Chicago (plasmid pCRE-236-2, CP061382) (Fig. 2b). The related strain from New York was positive for the bla KPC-2 gene located on a plasmid that had no similarity with those in Italian ST37.
The homology between pKpQIL_1020 and pCRE-2 was low, the two plasmids being significantly different in replicon content but both carrying the Tn4401 transposon with bla KPC-3 -like genes. The pKpQIL_1020 and pKpQIL_17B plasmids, by contrast, were almost identical except for a 4.5 kb region, located between the IncFII repA replicase and the fertility inhibition finO genes. This 4.5 kb region, present in the ST37 plasmids but not in other pKpQIL plasmids circulating in PUI, coded for: klebicin B, a colicin immunity protein, a phospholipase D family protein and three hypothetical proteins. The presence of this specific region has been certified for the first time in KPC-producing E. coli carrying pKpQIL [33], but a GenBank search of the region revealed its presence in several K. pneumoniae isolates, belonging to multiple STs [34].
Variants of the KPC enzyme displaying only the D179Y substitution have been well described and documented, and the role of mutations inside the Ω-loop (amino acids 164-179 [35]) in the development of CZA resistance (paired with a restored susceptibility toward carbapenems) has been extensively reported [8,9].

CONCLUSIONS
The ST37 clone was widely represented in the Italian epidemiology before the spread of the bla KPC -harbouring isolates belonging to CG258. Several descriptions of ST37 in hospitals of Rome were made in 2010 [5,7,26].
To the best of our knowledge, between 2011 and 2021, there were very few ST37 isolates described in Italian hospitals. In that decade, the Italian epidemiology was characterized by carbapenem-resistant K. pneumoniae isolates belonging to ST307, ST258 and ST512 [37][38][39][40]. The presence of bla KPC-3 in ST37 was reported only in one isolate from the centre of Italy in the nationwide collection of carbapenem-resistant invasive isolates, obtained in 2016 in the framework of the National Antibiotic-Resistance Surveillance (AR-ISS) [2]. Additionally, the number of complete K. pneumoniae ST37 genomes available in the GenBank database is very limited, suggesting narrow diffusion of this clone worldwide.
Given this context, we hypothesize that the ST37 clone was present in PUI for over a decade, probably within the wider group of carbapenem-susceptible K. pneumoniae strains. It was subsequently reported by the surveillance system when it shifted in the CZA-resistant pool by acquisition of the bla KPC gene variants. We consider that the historical ertapenem-resistant strains from 2010 [5] were not the direct ancestors of the KPC-producing CZA resistant strains from 2021.
It is fascinating to observe how the genomic analysis highlighted the evolutionary pathways of these isolates, including the acquisition of ICEKp harbouring the ybt loci, the differences in phage content and the gain of bla KPC harbouring the pKpQIL plasmid.
It has recently been demonstrated that the wild-type OmpK36 porin is relevant in stabilizing the mating pair during conjugation of the pKpQIL plasmid. Mating pair stabilization occurs by interaction between OmpK36 and the TraN factor of pKpQIL [41]. In ertapenem-resistant isolates 4011 and SC26, the insertion at the glycine-aspartic acid site, observed in the OmpK36 L3 loop, could have limited the acquisition by conjugation of pKpQIL in these strains. This could explain why ST37 from the 2010s, which was a highly diffused clone at PUI, did not participate significantly in the acquisition of bla KPC on pKpQIL. Instead, the introduction and spread of CG258 occurred intensively in the hospital [5,7]. The acquisition of pKpQIL-bla KPC happened in more recent times, starting from a susceptible ST37 strain, carrying a wild-type OmpK36.
Interestingly, ST37 could have acquired the pKpQIL encoding KPC-31, and during the outbreak it led to two further variants in a month. The KPC-70 and KPC-110 variants, compared to KPC-31, showed an increase in CZA MIC. This implies that ST37 has high mutability capacity, as also suggested by the high numbers of SNPs identified between historical and contemporary strains. Because of the increasing use of CZA for treating carbapenem-resistant K. pneumoniae infections, it can be expected that ST37 will spread in the hospital in the near future.